# DEVELOPMENT OF THE HYPOTHALAMUS

EDITED BY: Gonzalo Alvarez-Bolado, Valery Grinevich and Luis Puelles PUBLISHED IN: Frontiers in Neuroanatomy

#### *Frontiers Copyright Statement*

*© Copyright 2007-2015 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88919-634-0 DOI 10.3389/978-2-88919-634-0

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **DEVELOPMENT OF THE HYPOTHALAMUS**

### Topic Editors:

**Gonzalo Alvarez-Bolado,** University of Heidelberg, Germany **Valery Grinevich,** German Cancer Research Center (DKFZ), Germany **Luis Puelles,** University of Murcia, Spain

Updated prosomeric model, from Fig. 10B of Puelles and Rubenstein (2015) "A new scenario..." (this volume).

COVER IMAGE: Background: Detection of oxytocin-neurophysin I by the PS38 antibody (red) and oxytocin-intermediate forms by the VA10 antibody (green) in the mouse hypothalamic paraventricular nucleus. At E16.5 (upper rows) both forms are detected. At P0 (lower rows) virtually all oxytocin neurons of this nucleus co-express oxytocin-neurophysin I and immature oxytocin forms. Courtesy of Francoise Muscatelli, unpublished. Foreground: Hmx3 (Nkx5-1) expression in the adult mouse hypothalamus, from "The Gene Expression Nervous System Atlas (GENSAT) Project, NINDS Contracts N01NS02331 & HHSN271200723701C to The Rockefeller University (New York, NY)" (public domain).

The hypothalamus is the region of the brain in charge of the maintenance of the internal milieu of the organism. It is also essential to orchestrate reproductive, parental, aggressivedefensive, and other social behaviors, and for the expression of emotions. Due to the structural complexity of the hypothalamus, however, many basic aspects of its ontogenesis are still mysterious.

Nowadays we assist to a renewal of interest spurred in part by the growing realization that prenatal and early postnatal influences on the hypothalamus could entail pathological conditions later in life. Intriguing questions for the future include: do early specification phenomena reflect on adult hypothalamic function and possibly on some kinds of behavior? Can early events like specification, migration or formation of nuclei influence adult hypothalamic function?

A change in morphological paradigm, from earlier columnar interpretations to neuromeric ones, is taking place. Concepts long taken for granted start to be challenged in view of advances in developmental and comparative neurobiology, and notably also in the molecular characterization of hypothalamic structures. How should we understand the position of the hypothalamus in relation to other brain regions? Should we bundle it together with the thalamus, a functionally, genetically and developmentally very different structure? Does the classic concept of "diencephalon" make sense, or should the hypothalamus be separated? Does the preoptic area belong to the hypothalamus or the telencephalon? The answer to these questions in the context of recent causal molecular analysis will help to understand hypothalamic evolution and morphogenesis as well as its adult function and connectivity.

In this Research Topic we have reviewed the fundamentals of hypothalamic ontogenesis and evolution, summarizing present-day knowledge, taking stock of the latest advances, and anticipating future challenges.

**Citation:** Alvarez-Bolado, G., Grinevich, V., Puelles, L., eds. (2015). Development of the Hypothalamus. Lausanne: Frontiers Media. doi: 10.3389/978-2-88919-634-0

# Table of Contents



Irina G. Makarenko


Dominic Landgraf, Christiane E. Koch and Henrik Oster


Nora-Emöke Szabó, Roberta Haddad-Tóvolli, Xunlei Zhou and Gonzalo Alvarez-Bolado

# Editorial: Development of the hypothalamus

#### Gonzalo Alvarez-Bolado<sup>1</sup> \*, Valery Grinevich<sup>2</sup> \* and Luis Puelles <sup>3</sup> \*

*<sup>1</sup> Department of Neuroanatomy, University of Heidelberg, Heidelberg, Germany, <sup>2</sup> Schaller Research Group on Neuropeptides, German Cancer Research Center and University of Heidelberg, Heidelberg, Germany, <sup>3</sup> Department of Human Anatomy and IMIB, University of Murcia, Murcia, Spain*

#### Keywords: circadian, MCH, Notch, oxytocin, prosomeric, Shh, thyroid, zebrafish

The hypothalamus is the region of the brain in charge of homeostasis as well as homeostatic behaviors like eating and drinking. The anatomical, connectional and physiological complexity of this region matches the importance and intricacy of its functions. Perhaps because of this, research on the developing hypothalamus has lagged behind that on the cortex or hippocampus. The realization that current pathological conditions like, e.g., some forms of obesity, hypertension and hormonal dysfunctions have its origin in developmental alterations of the hypothalamus has turned the focus again on this region. In the present Research Topic we try to give an idea of a variety of approaches, morphological, comparative and genetic, to key developmental questions related to the hypothalamus. In this collection, a focus on essential questions on the nature of this brain region and its specification is obvious.

We start logically with papers explaining how the hypothalamus can be subdivided. Differentially with all earlier monographs on the hypothalamus, we reflect here the increasing perception in the field that the classic columnar model is outdated and incompatible with accruing molecular data on the hypothalamus. Many workers are turning to the updated prosomeric model as an instrument for morphologic and causal interpretation. This model has been presented in several articles, and several progressive changes and improvements made to it (Puelles et al., 1987, 2012; Puelles and Rubenstein, 1993, 2003; Rubenstein et al., 1994; Puelles, 1995; Shimamura et al., 1995). The model proposes that the hypothalamus is the rostralmost part of the neural tube, since the developmental forebrain length axis ends in the terminal wall of the hypothalamus (i.e., the telencephalon is a dorsal outgrowth of the alar hypothalamus). The resulting peduncular and terminal hypothalamic subdivisions of the rostral part of the neural tube are admittedly divergent with those traditionally taught under the columnar viewpoint. The hypothalamus is perhaps the brain region undergoing the most puzzling changes in the model, and this curiously turns out to the advantage of causal explanations. For this reason, the place of the hypothalamus in the prosomeric model and corresponding molecular subdivisions were recently the subject of a long and scholarly book chapter (Puelles et al., 2012). This includes the so-called updated prosomeric model.

Here, Puelles and Rubenstein offer a more succint and very clear explanation of the changes introduced in their updated model, tracing their fundament to establish a solid basis of explicit reasonable assumptions, pointing out also some novel morphologic highlights, like the interpretation of the course of the fornix tract, or the newly-defined acroterminal domain (Puelles and Rubenstein, 2015). If this prosomeric paradigm is correct, gene expression patterns should support it. Therefore, the authors of the model mined the most extensive developmental gene expression database, the Allen Developing Mouse Brain Atlas (Allen-Institute-for-Brain-Science, 2009), in order to find additional support for the proposed Peduncular Hypothalamus, Terminal Hypothalamus, and Acroterminal Domain (Ferran et al., 2015). They find a number of early expression domains not only supporting the proposed subdivision, but also contributing to explain why certain structures are formed within certain domains.

Edited and reviewed by: *Javier DeFelipe, Cajal Institute, Spain*

#### \*Correspondence:

*Gonzalo Alvarez-Bolado, alvarez@ana.uni-heidelberg.de; Valery Grinevich, v.grinevich@dkfz-heidelberg.de; Luis Puelles, puelles@um.es*

> Received: *19 May 2015* Accepted: *09 June 2015* Published: *23 June 2015*

#### Citation:

*Alvarez-Bolado G, Grinevich V and Puelles L (2015) Editorial: Development of the hypothalamus. Front. Neuroanat. 9:83. doi: 10.3389/fnana.2015.00083*

The hypothalamus is a brain region having intimate implication in the life story of animal species: heat production, eating and drinking behavior, reproduction, all are regulated by it. For this reason the hypothalamus might be thought to be very species-specific. Are the gene expression patterns and distinct morphological fates on which the model is based specific of mammals? Or are they on the contrary ancestral to vertebrates and thus widely shared across different species, irrespective of evolutionary variation? To answer this question it was important to test the current prosomeric model of the hypothalamus, built mostly on observations on mammalian brains, in other vertebrate classes. Here we show two contributions in this direction. Domínguez et al. reviewed characteristic gene expression patterns (genoarchitectonic approach) in amphibians and in reptiles, and found shared traits as well as differences that underline the anamniote/amniote transition (Domínguez et al., 2015). Other workers focused similarly on the hypothalamus of an elasmobranch, the cat-shark, noting that a prosomeric organization is also found in cartilaginous fishes, so that it can be concluded that the model subdivisions are ancestral to jawed vertebrates (Santos-Durán et al., 2015).

Then we move on to the early steps in hypothalamic regionalization. One key gene in the prosomeric subdivision is Sonic hedgehog (Shh), and it could not be absent here. A concise review highlights the main characteristics of expression, progenitor domains and roles of Shh in hypothalamic development (Blaess et al., 2014). Shh signaling exerts its effects through the Gli family of transcription factors. These can act either as activators or repressors of downstream genes, and the balance between these opposite functions of the Gli proteins (the "Shh-Gli code") is the basis of the effects of Shh on differentiation. Earlier analyses have unraveled the Shh-Gli code in the mouse spinal cord (see for instance Ericson et al., 1997; Ruiz i Altaba et al., 2003; Bai et al., 2004; Stamataki et al., 2005). Here, Haddad-Tóvolli et al. (2015) have approached the question of the particular roles of Gli2 and Gli3 as activators or repressors and their possible contribution to hypothalamic regional differentiation (the hypothalamic Shh-Gli code). As part of their analysis, the authors mapped their results onto the hypothalamus model, using it for the first time as a tool of phenotype analysis. Among other novel results, this study shows that hypothalamic regional diversity depends on differentially stringent requirements for Gli2 and Gli3. Ware et al. (2014) then review the available evidence on early hypothalamic neurogenesis. They conclude that a genetic network including members of the Notch pathway as well as proneural transcription factors is essential for the regulation of the initial neurogenesis occurring within the SHH-controlled hypothalamic basal plate.

The zebrafish is a key animal model to understand early hypothalamus development. The advantages of the zebrafish as developmental model system are well-known: embryos are very abundant, are transparent and develop rapidly and outside the mother. This makes it possible to physically manipulate them as well as live-imaging complex morphogenetic processes. Neural development studies, and in particular approaches to the early specification of regions like the hypothalamus, have benefited from the ease and speed of the genetic analysis in this model, for instance through gene knockdown or overexpression as well as the generation of transgenic fish (carrying inheritable genetic alterations). This has made the zebrafish model particularly useful to uncover gene function. As a result of the sequencing of the zebrafish genome, we know that around 70% of human genes have zebrafish orthologues. Many of these orthologues are involved in basic processes of brain specification and patterning. Hundreds of zebrafish mutant phenotypes have been identified through large-scale genetic screens, including many resembling human pathological conditions. Finally, having another vertebrate model system besides mammals and birds (Gallus gallus) to study the developing brain adds to the richness and variety of our knowledge in the field of comparative neuroanatomy.

In this collection, three different contributions approach important questions by analyzing early patterning events in zebrafish mutants. Manoli and Driever (2014) investigated the role of the transcription factor genes Nkx2.1, Nkx2.4a, and Nkx2.b in hypothalamic specification. They show that these regulators have differential roles in the development of the preoptic, intermediate and caudal hypothalamic regions.

Analyzed the developmental expression of peptide genes in the. Herget et al. (2014) recently defined the "neurosecretory preoptic area" (NPO) as the zebrafish larval paraventricular nucleus. Here, Herget and Ryu (2015) generate a complex map of peptide gene expression of this region as the basis for approaches to the neuroendocrine hypothalamus in the developing zebrafish.

Biran et al. (2015) review evidence on embryonic and postnatal expression of regulatory genes in the hypothalamus of zebrafish and mouse. They find that a number of transcription factors important for correct development are later redeployed in order to regulate key adult functions. Diaz et al. (2014) use peptide expression to characterize progenitor domains in the early hypothalamic ventricular zone of the mouse, then carefully map the migration of specific peptide-expressing neuronal subpopulations to their final settling place. They find that intricate tangential and radial migration routes as well as a considerable degree of mixing between the original subpopulations underlie the complex adult anatomy of the hypothalamus.

The previously unsuspected prevalence of tangential migrations illuminates the complexity of cell typology in many hypothalamic areas, and raises the possibility of functional alterations due to abnormal migrations. These peptides indeed have important functions, and oxytocin has come to be seen as the model for all of them, given its importance in behavior. Valery Grinevich and his coworkers have reviewed the development of the central oxytocin system and its alteration in patients afflicted with neurodevelopmental diseases (such as Prader-Willi syndrome and autism spectrum disorders) and respective animal models (Grinevich et al., 2014). At the end of the article, the authors discuss the therapeutic use of oxytocin for treatment of deficits in psychosocial and affiliative behaviors.

Croizier et al. (2014) review the development of the major axonal tracts coursing through the hypothalamus. They find that one important source of orientation for the main projection pathways is the earliest generated hypothalamic mantle. Later, these pathways will guide the formation of other afferent and efferent tracts of the prosencephalon. Makarenko (2014) has collated the available data on the time schedule and navigation pathways of the main hypothalamic tracts in the rat brain. She elaborates a precise calendar of projection development, and finds that each of the tracts is characterized by very specific spatio-temporal parameters.

Alkemade (2015) reviews contributions showing that, although the fetus is dependent on thyroid hormone from the mother, it can locally regulate its concentration not only in serum but also in the hypothalamus. The interplay between hormone sources and modulatory mechanisms achieves the precisely timed maturation of the hypothalamus (not too early, not too late).

We cannot focus on the development of all hypothalamic nuclei, but approach here two particularly emblematic ones. Landgraf et al. (2014) succinctly review the development

### References


of the circadian clocks. They conclude that, although we have good knowledge of the anatomical formation of the suprachiasmatic nucleus, there is still considerable debate about the prenatal emergence of rhythmicity. Sanchez-Arrones et al. (2015) use experimental fate-mapping together with prosomeric and genoarchitectonic principles in order to elucidate the developmental origin of the adenohypophysis in the chicken.

Finally, the hypothalamus and its heterogeneity can be used as a model to investigate key brain development problems. Szabó et al. (2015) approached a neglected developmental problem, the specific aggregation of neurons to form neuronal nuclei, using the mammillary body as a model. The authors show that two information systems, based on birthdates and adhesion mechanisms, work together in order to guarantee appropriate neuronal aggregation as well as specific axonal fasciculation. Both systems are implemented on the basis of sequential and hierarchical functions of classical cadherins.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Alvarez-Bolado, Grinevich and Puelles. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A new scenario of hypothalamic organization: rationale of new hypotheses introduced in the updated prosomeric model

#### Luis Puelles <sup>1</sup> \* and John L. R. Rubenstein<sup>2</sup>

<sup>1</sup> Department of Human Anatomy, School of Medicine, University Murcia and Instituto Murciano de Investigación Biosanitaria, Murcia, Spain, <sup>2</sup> Nina Ireland Laboratory of Developmental Neurobiology, Department of Psychiatry, University of California, San Francisco, San Francisco, CA, USA

In this essay, we aim to explore in depth the new concept of the hypothalamus that was presented in the updated prosomeric model (Puelles et al., 2012b; Allen Developing Mouse Brain Atlas). Initial sections deal with the antecedents of prosomeric ideas represented by the extensive literature centered on the alternative columnar model of Herrick (1910), Kuhlenbeck (1973) and Swanson (1992, 2003); a detailed critique explores why the columnar model is not helpful in the search for causal developmental explanations. In contrast, the emerging prosomeric scenario visibly includes many possibilities to propose causal explanations of hypothalamic structure relative to both anteroposterior and dorsoventral patterning mechanisms, and insures the possibility to compare hypothalamic histogenesis with that of more caudal parts of the brain. Next the four major changes introduced in the organization of the hypothalamus on occasion of the updated model are presented, and our rationale for these changes is explored in detail. It is hoped that this example of morphological theoretical analysis may be useful for readers interested in brain models, or in understanding why models may need to change in the quest for higher consistency.

Keywords: peduncular hypothalamus, terminal hypothalamus, acroterminal domain, genoarchitecture, anteroposterior pattern, dorsoventral pattern, length axis, tracts

### Introduction

The hypothalamus is a brain region whose name is familiar to all neurobiologists, though not many claim to understand perfectly its position, limits and inner structure in the context of surrounding forebrain territories. Indeed, there is controversy even among experts about the morphological model that best accounts for its complexity. How the hypothalamus is regionalized during development is still largely a matter of conjecture, despite various lines of insight, such as its ancestral origin in chordates, an ample number of neurogenetic and genoarchitectonic studies, and identification of various candidate patterning mechanisms. Our anatomic knowledge of the complex nuclear composition of the hypothalamus is still redolent of the frustrating "potatoes- in-a-potatosack" approach, though modern genoarchitectonic analysis has introduced a measure of order and promises rational classification. As a consequence of the remarkable structural heterogeneity of the hypothalamus, the logic of its intrinsic circuitry at the service of various functional systems

#### *Edited by:*

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### *Reviewed by:*

Charles R. Watson, Curtin University, Australia Salvador Martinez, University Miguel Hernandez, Spain

#### *\*Correspondence:*

Luis Puelles, Department of Human Anatomy, School of Medicine, University of Murcia, Campus Espinardo s/n, 30071, Murcia, Spain puelles@um.es

#### *Received:* 23 January 2015 *Paper pending published:*

09 February 2015 *Accepted:* 23 February 2015 *Published:* 19 March 2015

#### *Citation:*

Puelles L and Rubenstein JLR (2015) A new scenario of hypothalamic organization: rationale of new hypotheses introduced in the updated prosomeric model. Front. Neuroanat. 9:27. doi: 10.3389/fnana.2015.00027 operating throughout the brain and beyond (e.g., neurohumoral functions) remains obscure. However, we do know that the hypothalamus is an important central station involved in networked neural control of organismic humoral homeostasis, circadian neural activity patterns, self-placing computation, motor control and central drives. We clearly need deeper understanding of the genetic causal mechanisms that organize primarily hypothalamic structure and function, prior to the intervention of postnatal epigenetic plasticity. This requires an appropriate morphological model, pregnant with suggestions about the spatial dimensions and limits of potential causal signaling effects, which can be tested experimentally. There is a recently updated version of the prosomeric model (Allen Developing Mouse Brain Atlas reference atlases and ontology; Martínez et al., 2012; Puelles et al., 2012b, 2013, 2014; Puelles, 2013) that includes novel anatomical hypotheses about hypothalamic organization (**Figure 1**). These hypotheses possibly need an explanatory commentary, and this is the aim of the present essay.

### Antecedents of the Updated Prosomeric Model

Hypothalamic studies during the last 100 years were largely interpreted using the columnar morphological model, which holds that the hypothalamus is the ventralmost longitudinal column of the diencephalon, and is intercalated between the telencephalon rostrally and the midbrain caudally. This concept was introduced by Herrick (1910) in amphibians (**Figure 2**), and was elaborated by Kuhlenbeck (1927, 1973) and others (e.g., Swanson, 1992, 2003; Alvarez-Bolado and Swanson, 1996) for vertebrates in general (**Figure 3**). We hold that this model is incorrect as applied to the forebrain, since its fundamental underpinning holds that the length axis of the neural tube ends beyond the diencephalon in the telencephalon (a position that we regard as arbitrary, and devoid of developmental correlation with axial mesodermal structures). In the original Herrick model the hypothetized columnar sectors of the forebrain neural wall were delimited by ventricular sulci (**Figure 2**); in general, such landmarks do not coincide with the boundaries of gene expression discovered in recent times (see Figure 3 in Puelles and Rubenstein, 1993). Molecular boundaries are thought to be much stronger and comparatively conserved limits, since they reflect primary causal regionalization features; the differential molecular identities of the limited territories control all subsequent histogenesis such as proliferation, neurogenesis and mantle development. On the other hand, ventricular sulci form as tertiary epiphenomena of mantle development; they emerge later, between the variously bulging parts of the differentiating mantle layer. It also was held in columnar theory that the resulting forebrain subdivisions—epithalamus, dorsal thalamus, ventral thalamus, hypothalamus—are associated with sensorimotor viscero-somatic functions analogous to those of brainstem columns; this tenet has aged considerably in the meantime. Importantly, the columnar model offers no account about the developmental mechanisms that might generate the postulated organization, nor explains causally finer regionalization within the columns (e.g., nuclear subregions). The recent loss of favor of this model has been accelerated by

FIGURE 1 | Updated prosomeric model as applied to the adult mouse brain. Hindbrain rhombomeres and cryptorhombomeres (r0–r11) are in blue, midbrain mesomeres (m1–m2) in green, diencephalic prosomeres (p1–p3) in yellow, and hypothalamo-telencephalic prosomeres (hp1–hp2) in red and orange, respectively. The roof, alar, basal and floor parts are not differentiated, for simplicity, but exist in every case (note the anterior commissure represents the rostralmost roof domain; the rostralmost floor corresponds to

the mamillary area-M). Abbreviations: ac, anterior commissure; cc, corpus callosum; VPall, LPall, DPall, MPall, ventral, lateral, dorsal and medial pallial sectors; PallSe, pallial septum; SPallSe, subpallial septum; OB, olfactory bulb; POA, preoptic area; THy, terminal hypothalamus; PHy, peduncular hypothalamus; PTh, prethalamus; Th, thalamus; PT, pretectum; M, mamillary body; APit, anterior pituitary; PPit, posterior pituitary; pc, posterior commissure; tc, tectal commissure; icc, intercollicular commissure.

its apparent inability to integrate meaningfully the accruing variety of gene expression patterns observed in the developing diencephalon and hypothalamus (Puelles et al., 2004, 2012b; Shimogori et al., 2010; Diez-Roux et al., 2011).

A modified version of the columnar model is favored by Swanson (1987, 1992, 1993, 2003, 2012), and Alvarez-Bolado and Swanson (1996), who adapted to it the alar/basal concepts of His (1893a,b, 1895, 1904), extrapolating them from the diencephalon into the telencephalon, while maintaining the standard columnar axis of Herrick/Kuhlenbeck. In this model the telencephalic subpallium is held to be "basal," and the pallium "alar," while the hypothalamus represents the "basal" part of the diencephalon (**Figure 3**). However, there is no clearcut molecular evidence or causal underpinning that supports these changed notions of His' alar and basal plates (His, 1889, 1892, 1893a,b, 1894, 1895, 1904), which apparently simply answer to the preconceived idea of the axis ending in the telencephalon. Swanson and colleagues have performed extensive connectivity studies addressing the rat hypothalamus in the context of major forebrain functional circuits, significantly advancing functional analysis of the observed systems (Swanson, 1987, 2000a,b, 2003, 2007, 2012; Petrovich and Swanson, 2001; Sawchenko et al., 2000; Thompson and Swanson, 2003, 2010). We hold that all conclusions on these topics can be reinterpreted without significant loss using our non-columnar morphological model—the prosomeric model (**Figure 1**). Other recent authors who developed their own variant version of a columnar model with an axis ending in the telencephalon are Altman and Bayer (1986, 1988, 1995). These authors contributed extensive reports on developmental neurogenetic patterns in the rat hypothalamus. Part of their data and conclusions also can be translated into prosomeric terms, though some of their interpretations on diencephalic regionalization (e.g., the extent of the thalamic progenitor domain—in which both the prethalamus and the pretectum, and some aspects of hypothalamic partition, are misrepresented) seem inconsistent with recent gene expression studies and the phenotypes of some mouse mutants.

The abundance of molecular, genetic and developmental data now provides opportunity to investigate more fully the developmental organization of the hypothalamus. The title of this essay— "A new scenario of hypothalamic regionalization"—refers to the prosomeric model approach, which emphasizes a return to the length axis originally defined by His (1889, 1892, 1893a, 1894, 1895, 1904), and a detailed morphologic and molecular analysis of the hypothalamus along redefined dorsoventral and anteroposterior dimensions (**Figure 1**). This novel approach appears capable of improving experimental developmental analysis and, therefore, causal understanding.

## Comparison of the Explanatory Capacity of Columnar and Prosomeric (Neuromeric) Models

It usually is not recognized that the columnar model, perhaps because of its selective functional orientation, was not a helpful morphological framework toward understanding the mechanisms underlying forebrain development. Notably, this paradigm admitted over the years numerous inconsistencies and points of

impasse (see examples below). We believe its assumptions have represented an occult obstacle to progress, mainly due to the pragmatic introduction by Herrick (1910) of an arbitrary, noncausally underpinned axis concept, which can be tracked in the literature of the next 100 years as a subconsciously implemented dogma that never was criticized or corrected by his followers. Indeed, when Herrick (1910, 1933, 1948) and Kuhlenbeck (1973) highlighted the idea of diencephalic columns, they unwittingly discarded the causal explanatory advantages of the earlier axial structural concept of His (1893a,b, 1895, 1904). This author had underlined the epichordal longitudinal position of the histogenetically precocious forebrain basal plate, whose growth mechanically causes the emergence of the alar/basal sulcus limitans; this longitudinal basal zone was held to end rostrally at the tuberal suboptic hypothalamus (**Figure 4**). Wilhelm His, one of the pioneers of neuroembryology, also defined the telencephalon as a dorsal outgrowth of the rostral alar plate, while his nascent term "hypothalamus" was restricted to the underlying forebrain basal plate (His, loc. cit.; review in Puelles et al., 2012b). These early ideas implicitly supported the transversal nature and general comparable dorsoventral patterning of the mesencephalic, pretectal, thalamic, prethalamic, and hypothalamic parts of the alar plate, as was also understood by Kappers (1929; 1947; **Figure 5**). This pioneering forebrain axial concept was based on comparative neuroembryological analysis in various vertebrates (His, 1889, 1892), and significantly included a topographic correlation of the floor and basal plates with the underlying notochord (ulteriorly shown experimentally to be causal). In contrast, Herrick (1910) did not adopt his new axis ending in the telencephalon because of any developmental discoveries. He rather assumed this ending against available developmental knowledge, because it was a "convenient" measure (Herrick, 1948), in order to be able to call "longitudinal" diverse diencephalic and telencephalic regions separated by ventricular sulci, which otherwise would be transverse in the model of His (**Figure 4**; compare Kappers, 1929, 1947; **Figure 5**). In this case the interpretation clearly preceded the concept. Herrick was interested in exploring the potential of forebrain regions delimited by sulci as longitudinal histogenetic and functional entities (so that they could be described as columns, and be assigned columnar functions extrapolated from the hindbrain), and he accordingly postulated the axis that was convenient to that purpose. This was done implicitly, without any argumentation, simply by saying the diencephalic sulci were longitudinal. He never defined his forebrain axis expressly, or discussed in detail the reasons for his modification of the axis. The functions alluded by Herrick were represented by the then novel insights about sensorimotor viscerosomatic functional specialization of the hindbrain nuclear columns associated to cranial nerves. Herrick thought that these notions might illuminate as well the functions of the forebrain, if it was similarly composed

of four longitudinal columns, like the hindbrain. The potential functional light thus brought into the forebrain was the apparent reason why the axial change was accepted by the scientific community, though probably with little understanding of the morphologic price paid (with exceptions, such as Kappers, loc. cit.). Neuroanatomists understandably were then not much preoccupied with a developmental understanding of how a given structure was organized, or evolved. The emphasis was on gaining functional understanding, and this was also the apparent reason why the pre-existent segmental/neuromeric morphological models of brain development and structure (reviewed in Von Kupffer, 1906; Ziehen, 1906) practically fell into disuse after 1910: no function was attached to the observed neuromeres, whereas the columns apparently were functional entities. Curiously, nobody ever articulated thoughts about why longitudinal ventricular sulci should correlate with functional regionalization. Nowadays, the rare authors still attached to sulci have not reacted yet to the fact that developmental genetic patterning does not correlate topographically with ventricular sulci, a point already emphasized by Puelles and Rubenstein (1993). Indeed, sulci do not seem to be genomically coded—why should they? In other words, why should Nature select for brain ventricular sulci?

As causal developmental neurobiology finally advanced in more recent times, a harmful effect of the now dogmatically accepted columnar axis was to change the expected source of anteroposterior patterning effects from the front of the hypothalamus (as implied by the neuromeric models, following His, 1893a,b) into the front of the telencephalon (as implied by columnar models) (**Figures 6A,B**). Secondly, the columnar idea of the hypothalamus as a ventral part of the diencephalon significantly handicapped the interpretation of dorsoventral patterning effects in this area as we understand them now, introducing much confusion. The columnar dorsoventral dimension of the diencephalon is not orthogonal to the essential axial landmarks, the forebrain floorplate and the notochord (**Figure 6B**).

The arbitrary columnar axis became an undoubted dogma after nearly a century of columnar thinking and publication. Most neuroscientists regard it as an established fact, rather than as a conjecture. Consequently, visualization of alternative interpretive causal possibilities was handicapped, and even the fact that this was happening was unnoticed among authors, reviewers and journal editors. This hidden effect that promotes wrong morphologic and causal assumptions may be easily traced in the relevant literature dealing with forebrain patterning, even up to contemporaneous reports; there is much inconsistent or nonsubstantiated axis-referred reasoning that distorts or misdirects causal analysis.

An example of such unnoticed explanatory inconsistency is the following: the columnar hypothalamus was postulated as the ventralmost diencephalic column, continuous with the telencephalic subpallium rostrally and the midbrain tegmentum caudally, though there is no notochord "under" the subpallium and the hypothalamus, as there is under the midbrain and hindbrain tegmentum (presently, the notochord represents the known causal agent of floor plate and basal plate induction, and resulting ventralization of the ventral part of the neural wall; Echelard et al., 1993; Roelink et al., 1994; Marti et al., 1995; Müller et al., 2000; Rastegar et al., 2002). The postulated rostral "basal" tissue is thus implied to arise on the whole out of dissimilar causal conditions (question marks in **Figure 6B**). Nobody apparently feels the need to explain this singularity. Another well-known example of inconsistency is that the columnar model does not explain the holoprosencephalic syndrome, in which abnormal rostral forebrain patterning causes the telencephalon and eyes to lose their bilateral division. Eventually, they can be completely lost, accompanied by the hypothalamus, leaving a stunted diencephalon remnant. Such a major patterning defect would be predicted to alter dorsoventral patterning of the whole diencephalon, since dorsalized structures normally equilibrate with ventralized counterparts, but the patterning of the prethalamus, thalamus and epithalamus appears normal in this syndrome. A final example of inconsistency: the basal plate is widely held to be "motor" in function, but the sensory eyes develop out of the hypothalamic region, as revealed by the position of the eye stalks and the arrival of the optic nerves at the optic chiasm, consistently with results from fate mapping the neural plate in several species (Rubenstein et al., 1998); therefore, the columnar viewpoint that

FIGURE 5 | Brain subdivisions in a generalized vertebrate model as conceived by Kappers (1947). Note conservation of the sulcus limitans of His ("sillon limitant"), and the clear concept of the sensory and motor longitudinal brainstem zones plus the floor plate (P.AL; P.BAS; PL.V). These bend uniformly around the cephalic flexure and end at the rostrally placed hypothalamus. The middle and ventral thalamic sulci (S.TH.M; S.TH.VEN) are represented as strictly transversal relative to the longitudinal dimension, which obviously does not end in the telencephalon, but in the hypothalamus. Note the notochordal tip contacting the mamillary pouch.

models (thick black line).

cases. The postulated hypothalamic and diencephalic neuromeres are

unconscious evasive action (e.g., Swanson, 1992, 1993, 2003, 2012; see also Alvarez-Bolado and Swanson, 1996; their Figure 17). The cause of the problem is the use of Herrick's hundredyear-old incorrect longitudinal axis. Dozens of such inconsistent stumbling blocks can be pointed out, of which the scientific community does not seem to be aware, since attention to them is clearly not demanded by the peer review system.

Weighty molecular and experimental patterning evidence now shows that Herrick's diencephalic "columns" are not organized developmentally as dorsoventrally arranged structures, but as alar parts of transverse neuromeric units, or brain segments, which are themselves arranged rostrocaudally along the

the hypothalamus is the basal part of the diencephalon (Swanson, 1992; **Figure 3**) implicitly holds that the sensory eye is basal in origin, and we are forced to accept that the central optic pathway first enters and connects with the hypothalamus, the pretended diencephalic basal plate, before it reaches dorsally placed centers of analysis in the alar plate (note no other brain sensory input does this). No one in the ample columnar literature has ever discussed this point.

Columnar tradition has thus by action or omission caused scientific thought to stop at these and many other impasse points, since the columnar model and its cryptic axis only allows the thought "how odd," but no further line of reasoning, causing histogenetic axis defined by His (1893a,b) (**Figure 6A**). Each neuromere possesses its own sector of roof, alar, basal and floor longitudinal zones, a shared basic structural feature which makes all neuromeres fundamentally comparable developmental units, that is, metameres, irrespective of their mutual differences. We also know why this is so: all neuromeres share crucial dorsoventral patterning mechanisms that start at neural plate stages, which relate to comparable antagonistic floor plate and roof plate patterning signals throughout the axial dimension of the neural tube (**Figure 6A**; Puelles, 1995, 2013; Shimamura et al., 1995; Puelles and Rubenstein, 2003; Martínez et al., 2012). The floor plate is induced by the notochord (Echelard et al., 1993; Roelink et al., 1994; Marti et al., 1995; Müller et al., 2000; Rastegar et al., 2002; Sanchez-Arrones et al., 2009). This explanation does not apply in the columnar model for the diencephalon and telencephalon, since these pretended AP units do not share DV patterning mechanisms, either among themselves, or with the midbrain and hindbrain (**Figure 6B**).

It was repeatedly underlined (Puelles and Rubenstein, 1993; Shimamura et al., 1995; Puelles, 1995; Puelles et al., 2004, 2012b, 2014) that a thin longitudinal band expressing the transcription factor Nkx2.2 courses through midbrain, diencephalon and hypothalamus along the apparent alar-basal boundary, or next to it; the topography of this band is comparable in all vertebrates investigated so far. This pattern emerges at neural plate stages, before the neural tube axis starts bending (Shimamura et al., 1995), and remains topologically invariant as the cephalic flexure forms (see Hauptmann et al., 2002). Secondarily, the band deforms around the transverse Shh-positive core of the zona limitans intrathalamica (ZLI; **Figure 7**). This very robust result (the neuronal derivatives of the band can be traced into the adult brain) reflects the common position of the alar-basal border throughout the forebrain and midbrain, and is clearly inconsistent with the columnar axis and the attached concept of diencephalic columnar subdivision. The Nkx2.2 expression pattern obviously is longitudinal except at the ZLI, but does not enter the telencephalon (as predicted by Swanson, 2003, 2012). Moreover, it consistently divides serially the midbrain, diencephalon and hypothalamus into alar and basal moieties (**Figure 7**). An alar hypothalamus is impossible in the columnar model (thus the problem with the eyes). Nowadays we know that Nkx2.2 expression is induced at high concentrations of the diffusing SHH morphogen at the border of basal plate (midbrain and forebrain) and floor plate (hindbrain and spinal cord) expression of Shh, which in its turn depends causally on notochordal signals (Echelard et al., 1993; Roelink et al., 1994; Marti et al., 1995; Rastegar et al., 2002). This result falsates the columnar postulate of the hypothalamus and telencephalic subpallium as an entirely basal region (Swanson, 1992, 2003, 2012).

Finally, the modern hypothalamus is not a homogeneous territory. Puelles et al. (2012b) mapped molecularly 33 discrete hypothalamic progenitor areas, and suggested that these areas produce a minimum of 150 derived nuclei or distinct cell populations. More recent data suggest that many hypothalamic areas are capable of sequentially producing several cell types over time; this extends significantly the list of different derivatives (Díaz et al., 2014). In contrast, columns were theoretically expected to be homogeneous (their components supposedly being unified by their dedication to a shared function; Kuhlenbeck, 1973). Obviously, this simplistic early columnar idea has evolved into the present-day concept of functional columnar subsystems of the hypothalamus, formed by strings of nuclei (Swanson, 2003, 2012), but remains devoid of any developmental explanation of how the hypothalamic column becomes subdivided into the 150 nuclei.

As mentioned, the nuclear structure of the hypothalamus is quite varied in terms of molecular profiles, neuronal aggregates and characteristic cell types (e.g., Swanson, 1987, 2003, 2012; Shimogori et al., 2010; Puelles et al., 2012b; Puelles, 2013). Its genoarchitectural profile, when interpreted within the updated

prosomeric model, highlights a series of molecularly distinct parallel progenitor bands arranged along the dorsoventral dimension, that is, stacked one upon another between the dorsal hypothalamo-telencephalic boundary and the ventral hypothalamic floor plate (Morales-Delgado et al., 2011; Puelles et al., 2012b; Díaz et al., 2014; Domínguez et al., 2015; Santos-Durán et al., 2015). These progenitor bands clearly must result from the interplay of early antagonistic dorsalizing and ventralizing patterning mechanisms (**Figures 8A,B**). The corresponding developmental units underlie causally the well-differentiated adult paraventricular, subparaventricular, tuberal, perimamillary, and mamillary histogenetic areas, and the respective derived nuclear regions. Note the first two bands are alar, whereas the remaining three are basal, clearly emphasizing antecedent DV patterning, i.e., dorsalization vs. ventralization (**Figure 8**). These dorsoventral anatomic regions were always characterized instead as anteroposterior ones in columnar studies on the hypothalamus, in the absence of any postulates on corresponding AP patterning mechanisms that would cause such an organization (e.g., Swanson, 1987, 2003, 2012). We believe that the columnar viewpoint does not help causal explanation of this structural arrangement that typically subdivides the hypothalamus, because, surprisingly, no columnar assumptions whatsoever exist that predict and causally explain that columns eventually may show anteroposterior subdivisions ("how odd," again). This problem is resolved by the neuromeres postulated in the prosomeric models, but these units are not admitted in columnar thinking. Examination of **Figure 8B** reveals that the dorsoventral alar and basal domains contemplated in the prosomeric model would require outrageously unparsimonious treatment in order to be interpreted as segmental units relative to the columnar axis. Again in this case the wrong axis impedes appropriate causal explanations to be conceived, and the issue is left unresolved.

Apart of the cited DV pattern, the updated prosomeric model contemplates also a general anteroposterior (AP) partition of the hypothalamus into terminal and peduncular transverse territories across the cited 5 DV bands (THy; PHy; **Figure 8B**). This partition implies the existence of an intrahypothalamic interneuromeric limit that separates the hypothalamo-telencephalic prosomeres 1 and 2 (hp1, hp2 in **Figure 1**; Pombal et al., 2009; Martínez et al., 2012; Puelles et al., 2012b, 2014; Puelles, 2013). These novel concepts take into consideration two classical telencephalic regions—the non-evaginated (impar) and evaginated (hemispheric) parts—that respectively complement the two hypothalamic AP regions at the dorsalmost part of the alar plate and the corresponding roof plate. This updated prosomeric pattern is rooted in the pioneering view of His (1893a,b), who already connected hypothalamus and telencephalon, seeing them jointly as a rostral forebrain unit; this feature changes significantly relative to earlier versions of the prosomeric model (**Figure 10B**: Puelles et al., 1987a, 2004, 2007; Puelles and Rubenstein, 1993, 2003; Rubenstein et al., 1994, 1998; Puelles, 1995, 2001, 2009). The hypothalamus + telencephalon forebrain unit (also known as secondary prosencephalon) is held to be the rostralmost transversal part of the forebrain, lying in front of the trineuromeric diencephalon proper (diencephalic prosomeres 1–3; p1–p3); the latter recovers in this model the original tegmental portions conceived by His (1893a,b, 1895), which had been arbitrarily ascribed to either hypothalamus or midbrain in the columnar tradition (e.g., compare Dong, 2008 with Puelles et al., 2012b; see also **Figures 3** and **10B**).

The axial rostral neural tube sequence postulated in the prosomeric model accordingly runs: secondary prosencephalondiencephalon-midbrain, each unit representing complete rings of the neural tube and of the forebrain (the hypothalamus/telencephalon composite is a modified ring, since it is closed

FIGURE 8 | (A) Summary of antagonistic dorsoventral patterning effects spreading from the roof plate, including its rostralmost portion at the anterior commissure, and the floor plate, including its rostral hypothalamic sector. These effects presumably establish the alar-basal boundary (red line), as well as the telencephalo-hypothalamic boundary. The blue boxed area is examined in detail in (B). (B) Map of the known dorsoventral molecular regionalization of the alar and basal hypothalamus, held to result from graded finer interactive effects within the primary dorsoventral pattern. The alar-basal boundary is marked by the thick red line. The alar longitudinal domains are represented by the paraventricular area

(subdivided into dorsal, central, and ventral microzones) and the subparaventricular area (this relates to the optic chiasm and the initial course of the optic tract). The basal hypothalamus consists of similarly dorsoventrally related tuberal and mamillary regions (sensu lato). The updated terminology proposes distinguishing tuberal (Tu) from retrotuberal (RTu) areas, as well as perimamillary and mamillary sensu stricto (PM, M) from periretromamillary and retromamillary sensu stricto areas (PRM, RM), respectively belonging to THy and PHy. Note the Tu/RTu complex can also be subdivided dorsoventrally into dorsal, intermediate and ventral microzones (TuD, TuI, TuV; RTuD, RTuI, RTuV).

rostrally by the terminal wall (Swanson, 1992). This singular morphologic feature can be visualized topologically via fate mapping at neural plate stages (**Figure 9A**). Its causal explanation relates to the fashion in which the floor plate and the roof plate end rostrally at neural plate stages (Shimamura et al., 1995; Puelles, 1995, 2013; Cobos et al., 2001; Sanchez-Arrones et al., 2009; Puelles et al., 2012b). Note that the only forebrain locus that fuses together during neurulation is the prospective roof plate (Cobos et al., 2001), and the telencephalon field present at neural plate stages is demonstrably parallel to this longitudinal zone (loc. cit.; **Figure 9A**). The telencephalon accordingly no longer can be explained as representing by itself the rostral end of the neural tube (compare **Figure 9B**). It is not more rostral than the eye and the whole hypothalamus; analogously to the eye, it should be considered a giant derivative that buds bilaterally out of the dorsal alar hypothalamus. Its patterning into subpallial and pallial moieties likewise starts at neural plate stages, this being a topologically AP pattern (the subpallium field lies rostral to the pallium field; **Figure 9A**); note columnar tradition has wrongly defined pallium/subpallium as a DV pattern, again causing unrecognized problems in causal interpretation (**Figure 9B**). The telencephalon accordingly is best understood as a hypothalamic derivative (also from the evolutionary point of view). Topologically, the hypothalamus is no hypothalamus, but a hypotelencephalon.

### New Aspects of the Updated Prosomeric Model

The expression "new scenario" is used in the title because significant changes were introduced with regard to the preceding model version of Puelles and Rubenstein (2003) by Puelles et al. (2012b); these novelties also appeared in the Allen Developing Mouse Brain Atlas reference atlases and related ontology (www.developingmouse.brain-maps.com, online since 2009). Among the recent model changes are included various aspects that are not relevant for the hypothalamus, such as a better systematic treatment given to the telencephalic subpallium (see Puelles et al., 2013), a redefinition of pallial sectors and the concept of the claustrum (Puelles, 2014), and the introduction of the m2 mesomere and the cryptorhombomeres r7–r11 (**Figure 1**; Alonso et al., 2012; Puelles et al., 2012a; Tomás-Roca et al., 2015). We address here only the novelties that affect the hypothalamus, generally offering solutions for nagging conundrums that had resisted previous analysis. Our concern with

subpallial) relates ventrally with the hypothalamus and dorsally with the septal

difference is observed in the rostral limit of the midbrain (green area).

some of these unresolved issues was expressed explicitly in the Puelles and Rubenstein (2003) review.

The old difficulties we now believe to have solved with the update are three: (1) the early topographic relationship of the hypothalamus with the notochord; our new analysis led us to molecular and causal redefinition of the hypothalamic floor plate, and we discovered its epichordal character throughout (important corollaries: there is no prechordal part of the neural tube, and the well-known median displacement of prechordal plate cells occurs ventrodorsally in front of the terminal hypothalamic wall); (2) we resolved satisfactorily the dorsalward course of the transverse intrahypothalamic boundary across the telencephalic field, in order to connect it with the roof plate (impasse on this in Puelles and Rubenstein, 2003); its ending at the floor plate was also modified; consequently, this limit acquires the topologic properties of a complete neuromeric border (see Puelles and Rubenstein, 2003) and the hypothalamus + telencephalon complex (the secondary prosencephalon) results divided in prosomeres hp1 and hp2; (3) the topologic position of the mamillary/retromamillary and tuberal regions in the basal hypothalamus was reconsidered, reaching the novel conclusion that both regions are longitudinal, rather than transversal (as we thought before); this led to the proposal of a novel partition, the retrotuberal area, as well as to the distinction of a similarly longitudinal intercalated domain between tuberal/retrotuberal and mamillary/retromamillary regions, the perimamillary/periretromamillary area [note we write "mamillary" with a single "m," since we believe, following Rose (1939); Bleier (1961); Berman (1968), and (Jones, 1985), that the descriptor derives from the Latin term "mamilla" (nipple); otherwise we consequently should use "mammary" instead, if we held the descriptor derives from "mamma" (breast), but nobody does this].

A further significant change was applied to the updated concept of hypothalamus (Puelles et al., 2012b), attending to a difficulty that had not been noticed before, namely, (4) the need to explain the unique rostromedian hypothalamic specializations, a task achieved via the definition of the acroterminal hypothalamic domain.

### Rationales on These Points

### Relationship of the Hypothalamus With the Notochord (Hypothalamic Floor Plate)

In earlier versions of the prosomeric model, including Puelles and Rubenstein (2003), we held that the diencephalon and midbrain were epichordal (i.e., their floor plate was causally influenced by the underlying notochord), while the secondary prosencephalon, represented ventrally by the hypothalamus, was a prechordalrostral part of the neural tube (i.e., its floor plate lacked notochordal influences, and related causally instead to the prechordal plate mesoderm; **Figure 10A**). The implied prechordal floor region included retromamillary, mamillary and tuberal (median eminence, infundibulum and neurohypophysis) neighborhoods (**Figure 10A**). The histologic and functional variety shown by these regions was bewildering and difficult to explain

FIGURE 10 | Schematic comparison of the earlier prosomeric model version of Puelles and Rubenstein (2003) in (A) with the updated version of Puelles et al. (2012b) in (B). The (A) schema was slightly modified, repositioning more conveniently the anterior commissure, and eliminating for simplicity all unnecessary details in the present context. The (B) schema illustrates changes in the intrahypothalamic boundary, which now extends from the roof plate into the floor plate, distinctly separating the hp1 and hp2 prosomeres and the PHy and THy parts of the hypothalamus. The telencephalic subpallium is identified as a blue field; note its POA, Dg, Pal, and St parallel subdivisions. The alar hypothalamus remains essentially unchanged, apart the introduction of the paraventricular and subparaventricular areal names. The basal hypothalamus is deeply changed, due to our recognizing the mamillary area as occupying an extreme rostral and ventral longitudinal position, consistently with the new floor concept, and the tip of the notochord. This pushes the whole tuberal area, including the median eminence, infundibulum and neurohypophysis (NH), out of the hypothalamic floor (compare A) and into the rostral end of the basal plate. It represents now a fully longitudinal domain. The novel retrotuberal area (RTu) lies caudally to the tuberal area sensu stricto (Tu), and extends back to the prethalamic (p3) tegmentum, dorsally to the periretromamillary area (PRM). Rostral to PRM lies the perimamillary band (PM).

causally, since there was no known property of the postulated prechordal plate induction that would account for these different structural fates. This was definitely a "how odd" situation needing attention within the prosomeric model. A wider concern lay in considering potentially unsatisfactory a forebrain axis that was defined by two different axial causes, the notochord up to the diencephalon and the prechordal plate more rostrally, insofar as these mesodermal derivatives are themselves molecularly distinct cell populations, though sharing secretion of the SHH morphogen. In the background of this concern was the apparently hard result suggesting that the entire forebrain vesicle of Amphioxus is epichordal (Hatscheck, 1882; Von Kupffer, 1893; Lacalli, 1996; Nieuwenhuys, 1998).

Our understanding of this difficulty was unexpectedly illuminated by the experiments of García-Calero et al. (2008) on temporally-stepped extirpation of the prechordal plate in early chick embryos. It was found that complete deletion of the prechordal tissue immediately after its formation caused a loss of the differentiation of the basal plate throughout the expanded forebrain (secondary prosencephalon, diencephalon, and midbrain), in addition to holoprosencephaly and massive molecular dorsalization of remnant tissue. Selective deletion of the notochordal tip only caused loss of the floor plate. Prechordal plate deletions performed after increasing time intervals—thus allowing prechordal cells to act upon the neural primordium during the interval— "saved" progressively the basal plate fate in caudorostral order (e.g., first midbrain, then diencephalon, finally hypothalamus; we had a selective early marker of the mamillary anlage, which was the last basal locus to appear). Even later deletions saved the holoprosencephaly syndrome, and finally also the loss of Shh expression in the subpallium (these clearly are prechordal effects on the alar plate). The conclusion was reached that the prechordal plate, a migrating cell population derived from the node (Izpisúa-Belmonte et al., 1993), sequentially exerts diverse inductive effects as it relates topographically to a sequence of basal and alar neural domains. Initially it is needed for the specification of the basal plate rostral to the isthmus, probably acting in parallel to notochordal and floor plate signaling upon this domain, but it does not itself induce floorplate-like structures, a feature particularly noted in the terminal wall, where prechordal signals work without accompanying chordal effects. The ventrodorsal migration of prechordal cells along the median terminal wall allows them to have additional specific effects, first on the hypothalamic rostromedian basal plate (where the tuberal infundibulum and the retrochiasmatic anterobasal area emerge), and then on the rostromedian alar plate, leading to separation of the eyes (chiasmatic area) and of the telencephalic vesicles (terminal lamina), ending, finally, with the specification of the preoptic patch of Shh expression, important for subpallial regionalization. The association of prechordal inducing effects to the basal zone, first, and to the alar zone, afterwards, was consistent with the novel idea that, topologically, the movement of prechordal plate cells is not rostralward, as it appears to naïve inspection, but dorsalward relative to the terminal wall (progressing from the tip of the floor plate to the tip of the roof plate). On the other hand, the notochordal founder cells also emerge from the node, but represent non-motile cells which incorporate sequentially to the caudally elongating chordal primordium along the body midline (the axis) as the node recedes caudalwards; vertical chordal signaling is known to induce specifically the differentiation of the floor plate in the neural ectoderm, a phenomenon starting already at neural plate stages (Echelard et al., 1993; Roelink et al., 1994; Marti et al., 1995; Rastegar et al., 2002; Sanchez-Arrones et al., 2009). These results led us to the conviction that causal underpinning of both the forebrain length axis and the floor plate should be only ascribed to the notochord, and we classified any prechordal plate patterning effects as separate terminal non-axial DV patterning mechanisms produced by a motile signal source.

Our attention next turned to where lies precisely the rostral tip of the notochord relative to the hypothalamic primordium. We explored this issue in the literature, as well as via genoarchitectural analysis. We found that the literature is often vague and inconclusive about this point. Evidently, the notochord (or head process) only contacts the median floor of the neural primordium at very early stages (neural plate, early neural tube; see **Figure 11A**), since the morphogenetic appearance of the cephalic flexure soon causes the separation of these two tissues. Nevertheless, several credible images reported on such later stages show that the tip of the notochord usually contacts or approaches the mamillary pouch (e.g., Romanoff, 1960; Figures 85, 105, 335; diverse images in Kuhlenbeck, 1973; e.g., his Figure 48C). Consistently with this result, most workers describing the earliest topography of the notochord relative to the forebrain underlined a rostral end at or under the prospective mamillary pouch (e.g., His, 1893a,b; Von Kupffer, 1894, 1906; Jurand, 1974; Morris-Kay and Tuckett, 1987; Saucedo and Schoenwolf, 1994; Sulik et al., 1994; Puelles, 1995; Alvarez-Bolado and Swanson, 1996; Barteczko and Jacob, 2002; Bardet, 2007; Sanchez-Arrones et al., 2009). This agrees with observations of Johnston (1923) on the existence of a modified floor plate rostral to the isthmus, all the way to the mamillary area, a result which we reproduced with whole-mount histochemical labeling of an AChE-positive epichordal floor-plate strip ending at the mamillary area (Puelles et al., 1987a).

If we return to the provisional conclusion reached above that only the notochord induces a floor plate fate in the neural primordium (this can be correlated with incipient molecular differentiation of the floor plate already at open neural plate stages; Sanchez-Arrones et al., 2009), the literature data on the notochordal tip topography jointly point out that the forebrain floor plate must end beyond the prosomeric diencephalon, within the hypothalamus, and specifically at the mamillary pouch.

Moreover, we searched the Allen Developing Mouse Brain Atlas for floor-plate-specific gene markers, and found that not only Shh (which is directly induced in the floor by the notochord), but also Foxa1, Lmx1b, Ntn1, and Nr4a2, appeared expressed at the forebrain floor, with an identical rostral end. At E11.5, labeling ended rostrally at a small outpouching of the midline, which subsequently transformed into the mamillary pouch at E13.5 (**Figures 11B–F**). This genoarchitectural finding was revolutionary for both the columnar and earlier prosomeric models. In columnar models, the mamillary hypothalamic area is held to be a caudal diencephalic region (**Figure 3**), whereas the new results strongly support a position at the rostral end of the forebrain

(encompassing the rostralmost floor). On the other hand, earlier versions of the prosomeric model (**Figure 10A**) had assumed that the forebrain floor reached the tuberal infundibular area, whereas the new results negated this possibility, suggesting that the tuberal region must be a rostromedian component of the basal plate (**Figure 10B**; see below).

Retrospectively, it may be noticed that the position of the mamillary area in the prosomeric model always was a difficulty. We had it initially in p4, caudal to p5 and p6 —Bulfone et al. (1993), Puelles and Rubenstein (1993)—, possibly due to the influence of His (1893a,b, 1895). Subsequently, we progressively felt the need to push it to a more rostral topologic position, as done in Puelles and Rubenstein (2003) (**Figure 10A**). Finally, we surprisingly found that it falls nicely at the absolute rostral end of the hypothalamic floor plate, coherently with various other novel morphologic features (**Figure 10B**; Allen Developing Mouse Brain Atlas; Puelles et al., 2012b). The resulting updated hypothalamic floor plate is thus shorter than previously imagined, but is entirely epichordal, thus parsimoniously unifying the causal underpinning of the forebrain axis throughout. Unexpectedly, our new interpretation also becomes consistent with the epichordal position of the entire brain in Amphioxus. Note we can now tentatively start to explain why other rostromedian territories in the hypothalamus (and beyond) differentiate distinctly than the floor plate, since they develop alternatively within the median basal hypothalamus (tuberal area), the median alar hypothalamus (chiasmatic area) or the median preoptic telencephalon (terminal lamina) (**Figure 10B**). All these nonfloor median forebrain areas are sequentially influenced by prechordal signals in the absence of notochordal signals, and their distinctive structural and molecular profiles can be attributed confidently to DV patterning (not possible in the columnar model; **Figures 6A,B**; **8A**). On the other hand, the true epichordal hypothalamic floor still shows two different regions—the mamillary and retromamillary floor domains—, an aspect which turns out to be consistent with our postulate of two hypothalamic prosomeres (**Figure 10B**; see below).

### Rostral End of the Roof Plate and Full Course of the Intrahypothalamic Boundary (=Neuromeric Border between Hypothalamic Prosomeres hp1 and hp2)

As reviewed in Shimamura et al. (1995) and Puelles (1995), the lateral border of the neural plate with the primitive non-neural ectoderm represents the prospective roof plate of the neural tube. The process by which the plate halves hinge upwards, and the bilateral borders then fuse together at the midline, forming the roof plate, is known as neurulation. The anterior and posterior neuropores are transiently open sites where the neurulation process has not yet finished. Puelles et al. (1987b) previously discussed the discrepant views in the literature about the closure of the anterior neuropore, bearing on the identification of the rostralmost roof plate point. They also performed a crucial experiment aimed to test the main hypotheses, by marking the rostromedian end of the anterior neuropore with a black plastic thread at successive stages in chick embryos. The results revealed that there is a single caudorostral sequence of closure of the anterior neuropore (other authors, as e.g., Swanson, 1992, still propose a double closure mechanism that so far lacks experimental support). It was suggested that the rostralmost roof plate roughly coincides with the prospective locus of the anterior commissure, that is, it would correspond to the telencephalon (earlier views had speculatively suggested several other possibilities apart this one, notably the optic chiasma; e.g., His, 1893a,b; Alvarez-Bolado and Swanson, 1996, their Figures 4, 16). Ulterior fatemapping experiments on the median end of the roof plate were performed by Cobos et al. (2001) using quail-chick homotopic grafts; the results fully corroborated the earlier result of Puelles et al. (1987b), and distinctly identified the bed of the anterior commissure as the rostralmost locus of the forebrain roof plate. The crossing of the anterior commissure appears in all vertebrates at the upper end of the terminal lamina. It is simultaneously understood to represent the bottom end of the septal commissural plate, though it actually represents its rostral end, as indicated by these experimental data. These fate-mapping data about the morphologic signification of the median bed of the anterior commissure inescapably imply that the entire septal midline belongs to the forebrain roof plate (**Figure 10B**), contradicting the popular assumption that the septum is a "ventral component" of the subpallium (the paramedian septum containing bilaterally the major septal nuclei belongs instead to the telencephalic alar plate). The same fate-mapping data indicate that the preoptic terminal lamina is neither roof- nor floor-plate-derived, but a rostromedian terminal alar differentiation of the secondary prosencephalon, corresponding, jointly with the chiasmatic area, to the place where the right and left alar telencephalic fields are primarily continuous in the neural plate (i.e., there is no fusion here, since the continuity already exists in the open neural plate; **Figures 9A**, **13**).

Insofar as the prosomeric model postulates that the whole telencephalon is an alar derivative of the secondary prosencephalon that is topologically superposed dorsally to the alar hypothalamus, it poses no problem to realize that the roof plate corresponding to the hypothalamus is the telencephalic roof (**Figures 10A,B**). The same results lead to inconsistent and unparsimonious interpretations within the columnar model, wherein the basal plate is held to reach the septum (**Figure 9B**; Swanson, 1992).

Now, coming to our problem, if the hypothalamus is subdivided anteroposteriorly in two domains, as considerable morphologic evidence suggests (Puelles and Rubenstein, 2003; Puelles et al., 2012b), then the separating intrahypothalamic boundary might represent an interprosomeric limit. This is only possible, theoretically, in the case that this boundary was complete, that is, was traceable all the way from the floor plate into the roof plate (according to the criterion formulated by Puelles and Rubenstein, 2003). Therefore, it is not enough to show that the intrahypothalamic boundary divides the hypothalamus transversely; it needs to be shown that it also divides the overlying telencephalic field, and reaches the local roof plate. This is the point at which we stumbled with earlier versions of the prosomeric model, since we did not find a convincing solution for how this boundary might satisfy this criterion (several alternative options were considered in Bulfone et al., 1995; Shimamura et al., 1997; and Puelles, 2001; finally Puelles and Rubenstein, 2003 acknowledged an impasse; see **Figure 10A**). The reason of these failures turned out to be an error in our assumption of where was the bed of the anterior commissure in terms of telencephalic subpallial domains. Up to 2007 we had assumed that this median locus corresponded to the anterior entopeduncular area or AEP (now renamed diagonal area or Dg; see Allen Developing Mouse Brain Atlas, and Puelles et al., 2013; **Figures 10A,B**). This implied that the septal roof plate ended within the AEP/Dg, while the preoptic area was thought not to participate at all in the septal roof plate, lying wholly in the alar plate.

However, more precise genoarchitectural mappings (notably of the Shh expression pattern) performed in the mouse (Flames et al., 2007; Allen Developmental Mouse Brain Atlas) and the chick (Bardet, 2007; Puelles et al., 2007; García-López et al., 2008; Bardet et al., 2010; Medina and Abellán, 2012) eventually disclosed that the preoptic area shows dorsally a median spikelike region that reaches the rostralmost septal roof—the bed of the anterior commissure—in between the right and left diagonal domains (**Figure 13**). This allowed to relocate the anterior commissure, and, accordingly, the rostral end of the roof plate, to this dorsomedian preoptic region, which can be conveniently named septo-commissural preoptic area (SCPO; Allen Developmental Mouse Brain Atlas; note Medina and Abellán, 2012 identify this domain as "commissural preoptic area," or POC, a term that in our opinion loses the semantic reference to a simultaneous ascription to the septum). The median preoptic nucleus that develops in the SCPO mantle zone is widely mapped in rodent atlases as surrounding frontally the anterior commissure in the median plane, consistently with this new interpretation (MnPO; **Figure 13**; see also Puelles et al., 2013).

As a consequence, it soon became obvious that this conceptual change at the preopto-septal intersection allowed to extend the intrahypothalamic boundary into the roof plate according to a new possibility which had not been considered before, namely, following the boundary between the preoptic area and the diagonal area (the preopto-diagonal border; dash-line in **Figure 10B**). This solution of the old conundrum seemed satisfactory for various reasons. First, the boundary separates the non-evaginated preoptic area (the classic telencephalon impar) from the evaginated telencephalic vesicle; theoretically, this allows a tentative causal explanation of this morphogenetic difference as related to differential neuromeric molecular identities. Second, the preoptic area within hp2 is corroborated as a distinct telencephalic territory that relates intimately to the optic area (the evaginated eye vesicle and the optic chiasma), representing its immediate dorsal neighbor within the anterior part of the alar secondary prosencephalon, whereas the evaginated telencephalon within hp1, placed altogether caudally to the preoptic area, limits separately with the paraventricular hypothalamic alar area; this represents the frontier that is traversed selectively by the cerebral peduncle (**Figure 12**). Third, the well-known course of the fornix tract in front of the interventricular foramen, as it passes bilaterally behind the anterior commissure to enter the hypothalamus,

suddenly acquired morphologic meaning, that is, the possibility of a causal explanation (there must be reasons for the course of any brain tract). Indeed, it can be hypothesized that, during their growth beyond the end of the hippocampal fimbria, the fornix tract fibers first elongate longitudinally along the paramedian septal commissural plate, that is, parallel to the roof plate; however, once they reach the preopto-diagonal boundary, most of them seem unable to cross it, and turn topologically 90◦ ventralward (forming the postcommissural fornix), to grow thereafter dorsoventrally along the caudal aspect of the intrahypothalamic boundary all the way to the retromamillary floor plate, where a number of the fornix fibers deccusate (**Figure 12**; see also Stanfield et al., 1987). Fourth, the new concept also apparently explains why the septal commissural plate consists of two different sectors, a caudal one containing the hippocampal and callosal commissures, and a rostral one containing the anterior commissure (**Figures 10B**, **12**). Within the updated prosomeric model, the reason is that we deal here with the roof plate domains of two different neuromeres, hp1 and hp2, where distinct axonal navigational guidance mechanisms are expected. No previous explanation background existed before for the remarkable course of the fornix. Curiously, this background wholly disappears as soon as this solution for the completeness of the intrahypothalamic boundary is abandoned (returning to earlier prosomeric model versions, or to columnar models).

The hypothalamo/telencephalic roof plate (evolutionarily it was hypothalamic before it was telencephalic) is accordingly divided into preoptic and hemispheric sectors by the extended intrahypothalamic border, and, as mentioned above, mamillary and retromamillary sectors are distinguished at the hypothalamic floor plate. This boundary at the floor plate is likewise underlined by the behavior of the fornix tract, which seems to be guided dorsoventrally through the whole hypothalamus by the intrahypothalamic boundary (Bardet, 2007; Puelles et al., 2012b). The invariant dorsoventral course of the fornix tract ends with a crossing of the hypothalamic floor plate just caudally to the mamillary body, that is, in the retromamillary area (Edinger and Wallenberg, 1902; Ramón y Cajal, 1911; Valenstein and Nauta, 1959; Stanfield et al., 1987; Köhler, 1990). Note the latter is identified in columnar descriptions as the "supramamillary area," though this traditional prefix is semantically inconsistent within the modified columnar schema used by Swanson (1987, 1992, 2003, 2012), in which the area is as retromamillary as in the prosomeric model, since it forms part of the same longitudinal zone, the postulated basal plate, topologically caudal to the mamillary body (**Figures 3**, **9**). Interestingly, neurons of the retromamillary area project reciprocally to the dentate gyrus, which implies an inverse ventrodorsal route via the fornix (Pasquier and Reinoso-Suarez, 1978; Haglund et al., 1984; Nitsch and Leranth, 1994). Since the intrahypothalamic border lies just rostral to the fornix tract, it neatly separates the molecularly distinct mamillary and retromamillary areas, allowing these likewise to be explained as differential neuromeric AP phenomena (**Figure 10B**). An explanation of why these two neighboring regions are structurally and molecularly distinct had never been offered before. Of course, this interpretation of the fornix implies that the fornix fibers that target the mamillary nuclei (and several other cell populations, such as the ventromedial shell formation) must cross the intrahypothalamic border rostralwards to reach them (**Figure 12**; see in this respect Stanfield et al., 1987).

As a consequence of being able to define this transverse boundary all the way from the roof plate into the floor plate, using the fornix as a crucial anatomic landmark (apart other anatomic features summarized graphically by Díaz et al., 2014; their Figure 1), we postulated that the secondary prosencephalon (or hypothalamo-telencephalic complex) is divided into two prosomeres, identified as "hypothalamic prosomeres 1 and 2" (hp1, hp2). The numbering proceeds in caudo-rostral order, continuing the caudo-rostral sequence of the diencephalic prosomeres 1-3. We abstained purposefully from continuing the cardinal list of prosomeres—e.g., naming them p4 and p5—, since this surely would lead to confusion with our earlier (now obsolete) p1–p6 model (Bulfone et al., 1993; Puelles and Rubenstein, 1993), in which the hypothalamus was subdivided in three quite non-comparable prosomeres p4–p6, including a misconceived floor region.

We came up with the idea to call the hp1-hypothalamus "peduncular hypothalamus" (PHy), referring to its clearcut and constant relationship in all vertebrates with the dorsoventral hypothalamic course of the cerebral peduncle (**Figure 12**; note the observable basal bending of the peduncle caudalwards is not understood within the columnar conception, which holds the whole tract is longitudinal). The caudal boundary of the peduncle while it courses through the hypothalamus thus roughly marks the limit between the PHy and the diencephalic prethalamus (check the topology in **Figures 9A**, **10B**, **12**). The advantage of the non-topographic "peduncular" term is that it intentionally evades referring to the controversial axis, while alluding to a well-known landmark present in all vertebrates. Accordingly, it can be used by any neuroscientist, irrespective whether he/she believes the hypothalamic course of the peduncle is transverse (prosomeric model) or longitudinal (columnar model). For the hp2-hypothalamus we considered for a time the use of "prepeduncular" as descriptor, but discarded it because it would be more precise to say "prefornical," since the fornix tract is the immediate peduncular landmark behind the intrahypothalamic frontier. Eventually, we chose to name this hypothalamic region "terminal hypothalamus" (THy; Allen Developing Mouse Brain Atlas; Puelles et al., 2012b, 2013; Puelles, 2013), in order to emphasize the relative position of this transverse unit at the topologic rostral end of the forebrain, leading to its implication in the "terminal wall." The latter term was apparently introduced by Swanson (1992), aptly referring to the rostromedian region that closes rostrally the neural tube (**Figure 11A**; see below more details about this median locus).

THy is continuous dorsally with "its" telencephalic sector, the preoptic area (**Figures 1**, **10B**); well-known terminal hypothalamic derivatives include in dorsoventral order the supraoptic, lateral anterior, suprachiasmatic, anterior, anterobasal, ventromedial, arcuate, and mamillary nuclei; there is also a terminal part of the dorsomedial nucleus, placed immediately caudal to the arcuate nucleus. Paradoxically, the terminal dorsomedial nucleus lies ventral to the ventromedial nucleus (this semantically confusing situation represents collateral damage of the columnar axis, to which all these classic terms refer; the new scenario demands complete revision and adjustment to the prosomeric "natural" axis of all positional descriptors in hypothalamic nomenclature).

On the other hand, PHy is continuous dorsally with the whole evaginated telencephalon (**Figure 1**), and includes as significant derivatives (again in dorsoventral order) the major part of the paraventricular nucleus, the peduncular part of the dorsomedial nucleus and the retromamillary area. Recently we have been searching the Allen Developing Mouse Brain Atlas for early gene expression patterns that are restricted to either the THy or the PHy, thus collectively defining molecularly the intrahypothalamic boundary. Part of these data are presented in this Issue by Ferran et al. (2015).

Interestingly, genoarchitectural data (Puelles et al., 2004, 2012b, 2014; Shimogori et al., 2010; Diez-Roux et al., 2011) show that PHy and THy are patterned dorsoventrally into a shared series of longitudinal zones across the respective alar and basal territories (**Figure 8B**). The alar-basal boundary is continuous with the diencephalic one (as was already recognized in the earliest versions of the prosomeric model; **Figures 10A,B**), and is marked by the dorsal boundary of the basal expression of Shh in the ventricular zone, which is partially overlapped by the abovementioned longitudinal band expressing Nkx2.2 (**Figure 7**). This molecular border, which roughly coincides with the sulcus limitans concept of His (1893a,b), reaches on both sides the terminal wall under the optic chiasm.

Leaving aside the alar telencephalic fields of hp1 and hp2, the subjacent alar hypothalamus shows a common longitudinal zonal division into a paraventricular area (Pa; we previously called it "supraopto-paraventricular area," but later discovered that the supraoptic nucleus only appears within THy) and a subparaventricular area (SPa) (**Figures 8B**, **10B**). The former is differentially labeled by Otp and Sim1, and lacks expression of Dlx or Arx genes, which are characteristic both of the overlying telencephalic subpallium and the underlying subparaventricular area. The peduncular paraventricular sector (PPa) is much broader than its companion terminal sector (TPa), and typically shows a tripartite triangular shape (DPa+CPa+VPa in **Figure 8B**). Its expands dorsoventrally caudalwards, toward the hypothalamo-diencephalic border, where it ends (it contacts there the prethalamic reticular nucleus and the overlying prethalamic eminence). PPa produces the largest part of the paraventricular nucleus complex, plus a radially migrated dorsal entopeduncular population. In contrast, the rather thin terminal paraventricular portion (TPa) relates to smaller parts of the paraventricular complex, namely the subpial supraoptic nucleus, the lateral anterior nucleus and the anterior periventricular area. Note the so-called "tuberal supraoptic nucleus," which we prefer to call "tuberal suboptic nucleus," according to its true position relative to the optic tract, lies in the underlying basal plate, though its neurons apparently migrate tangentially into this position from TPa origins (Morales-Delgado et al., 2011).

The underlying subparaventricular area differentially produces GABAergic neurons and also shows differently sized terminal and peduncular sectors (TSPa, PSPa; **Figure 8B**, **10B**). In this case, TSPa produces more voluminous derivatives, including the suprachiasmatic nucleus and the main (classic) anterior hypothalamic nucleus. The PSPa component forms a posterior tail of the anterior hypothalamic nucleus, an area that can be also described topographically as a "preincertal area" (corresponding to the "subincertal area" of some rodent brain atlases), since it is continuous with the prethalamic zona incerta formation, with which the SPa shares various gene markers (Puelles et al., 2004, 2012b, 2014; Shimogori et al., 2010; Puelles, 2013).

The basal territories of hp1 and hp2 are very extensive dorsoventrally, compared with those of the rest of the forebrain, and, interestingly, basal THy is much larger than basal PHy (**Figures 8B**, **10B**). This aspect may be due to early patterning influences of the prechordal plate, in concert with the predominant terminal expression of the early neural gene Six3 (Lagutin et al., 2003). This basal domain was classically divided into tuberal and mamillary regions, traditionally interpreted as anteroposterior items within the columnar model. In all the prosomeric model versions advanced up to the Puelles and Rubenstein (2003) review (**Figure 10A**), we tentatively accepted an anteroposterior arrangement of these two regions within the hypothalamic basal plate, consistently with our misguided concept of the floor plate extent (see above). Nevertheless, there was dim awareness of unresolved problems there. Eventually a satisfactory solution was found for this aspect, which accordingly was changed in the Allen Developing Mouse Brain Atlas, as well as in Martínez et al. (2012), Puelles et al. (2012b, 2013, 2014), and Puelles (2013), as is explained in the next section.

### The Topologic Position of the Mamillary/Retromamillary and Tuberal Regions in the Basal Hypothalamus

The background for the search of a better solution for the hypothalamic basal pattern was represented in the first place by our noticing of the fact that some longitudinal lines extending rostralwards from the cephalic flexure seem to end by sweeping neatly around the mamillary region to meet the terminal wall (then supposed to be the floor plate). This implied an inconsistency ("how odd" situation), since a longitudinal line in the lateral wall should not meet the floor plate, being topologically parallel to it. For instance, Kuhlenbeck (1973) always traced the sulcus limitans of His into such a perimamillary ending; this feature of his thinking led him to define the tuberal hypothalamus as an alar plate derivative (a point recently taken again by Diez-Roux et al. (2011) on the basis of genoarchitectural considerations). A similar curve is also traced by the longitudinal course of the mamillotegmental tract. Vertebrate species showing a clearcut hypothalamic ventricular organ (a linear circumventricular specialization which is unremarkable in mammals; see review in Puelles et al., 2012b) likewise provide evidence suggesting that this organ curves longitudinally around the mamillary region. The dorsal premamillary nucleus bends similarly around the mamillary body, and so does the tuberomamillary population of histaminergic neurons. The conundrum to resolve obviously was that the mamillary region cannot be a longitudinal domain and simultaneously display a transversal border with the tuberal region (**Figure 10A**). One of these aspects must be illusory, and both required attention.

Our previous conclusion that the hypothalamic floor plate ends precisely at the mamillary area (see above) was significant in this regard, since the floor plate is a primary longitudinal reference. This result by itself weighs importantly in favor of considering the mamillary/retromamillary region a longitudinal zone, consistently with the course parallel to the floor of the mamillotegmental tract and the band of perimamillary grisea. Dlx and Isl1 gene expression within the tuberal region distinctly limits the negative mamillary region along a curve that parallels the local floor plate (see Puelles et al., 2012b, their Figures 8–10). The same longitudinal boundary is underlined from the other side by genes selectively expressed within the mamillary and/or retromamillary areas, such as Otp and Foxb1 (ibid). Otp expression highlights a curved tissue band within the mamillary region sensu lato that limits with the Dlx/Isl1-positive tuberal region. This is the band that produces the dorsal perimamillary nucleus within its terminal portion, and it was identified as the "perimamillary/periretromamillary area" (PM/PRM; **Figures 8B**, **10B**; Simeone et al., 1994; Puelles et al., 2012b; Puelles, 2013; Allen Developing Mouse Brain Atlas; note the implied two parts correspond to THy and PHy, respectively). Close examination of these relationships suggested that the tuberal region sensu lato, which is quite massive rostrally (THy), extends longitudinally all the way to the hypothalamo-diencephalic boundary (PHy) via a gradually diminishing caudal portion placed over the PM/PRM; this "caudal tuberal" region in principle separates the mamillary region from the overlying alar-basal boundary (**Figures 8B**, **10B**). This observation made it possible to regard the tuberomamillary boundary as purely longitudinal.

The same as the mamillary region sensu lato decomposes dorsoventrally into the dorsal PM/PRM and the ventral mamillary/retromamillary (M/RM) areas sensu stricto, the tuberal region sensu lato also can be subdivided dorsoventrally into three longitudinal subdomains, identified by Puelles et al. (2012b) as dorsal, intermediate and ventral, across both PHy and THy (**Figure 8B**). The dorsal subdomain encompasses the precociously differentiating cells of the classic hypothalamic cell cord, aggregated into the anterobasal and posterobasal areas (ABas, PBas; **Figure 10B**). The intermediate subdomain includes as its own derivatives the dorsomedial nucleus (which has peduncular and terminal parts) and the arcuate nucleus (also terminal), and receives as a migrated entity the ventromedial nucleus, which is produced at the dorsal subdomain (see Puelles et al., 2012b on this previously unknown feature). Finally, the ventral (or tuberomamillary) subdomain is rather thin and corresponds to the hypothalamic ventricular organ, being likewise the restricted source of histaminergic neurons (which partly invade neighboring mamillary areas (see Puelles et al., 2012b for data supporting this new point). It limits ventrally with the PM/PRM areas.

This analysis implies that the hypothalamic basal plate is patterned dorsoventrally into 5 longitudinal zones, all of which expand rostralwards in a fan-shaped configuration into their respective ends at the terminal wall (**Figure 8B**). The large intermediate tuberal subdomain significantly encompasses rostrally the median eminence, infundibulum and neurohypophysis. This solution of the hypothalamic basal plate problem is clearly satisfactory in that it allows to understand the whole alar and basal (plus telencephalic) patterning of the rostral forebrain as a special case of standard dorsoventral patterning, implying antagonistic dorsalizing and ventralizing signals diffusing from the roof and floor plates (**Figures 6A**; **8A**), as occurs elsewhere in the neural tube (notably in the diencephalon and midbrain, where various relevant DV gene patterns are shared). The columnar model forbids such an explanation, due to its unhelpful axis reaching the telencephalon (**Figures 6A,B**), and does not provide a parsimonious alternative explanation.

We also reflected that the name tuberal area (Tu) strictly was meant originally only for the terminal (THy) sectors of these tuberal subdomains, since this term refers to the external bulge of the median eminence and infundibulum. The caudal, molecularly-defined "tuberal" extension into the peduncular (PHy) territory hardly relates to these rostromedian specializations, as it relates instead to the overlying peduncle. We therefore distinguished the caudal part of this basal complex with a novel term, the retrotuberal area (RTu), in analogy to the retromamillary neighbor (**Figures 8B**, **10B**). Thus, within basal PHy we have the RTu and RM, with their respective five dorsoventral subdivisions (RTuD, RTuI, RTuV, PRM, RM), and within basal THy there appear the Tu and M regions, with their own five dorsoventral subdivisions (TuD, TuI, TuV, PM, M). See Shimogori et al. (2010), Puelles et al. (2012b) and Ferran et al. (2015) for details of differential gene expression patterns throughout these diverse areas. An unexpected singularity that also emerged from the Puelles et al. (2012b) analysis is that the ventral premamillary nucleus, which in the adult appears within the TuI subdomain, intercalated between the dorsal premamillary nucleus and the ventromedial nucleus, originates from the RM area within PHy, from where its cell population migrates tangentially en masse into the definitive THy locus.

At first glance it may seem that the complex molecular and fate regionalization of the hypothalamic basal plate is out of the ordinary, but recent detailed genoarchitectural studies of dorsoventral patterning in the basal spinal cord have similarly disclosed a diversity of molecularly distinct dorsoventral progenitor domains (actually also 5 in number), where characteristic cell types are produced (Ulloa and Briscoe, 2007; Dessaud et al., 2010; Grossmann et al., 2010). Similar studies of the hindbrain and midbrain basal plate likewise detect diverse dorsoventrally disposed progenitor domains or microzones (e.g., Sieber et al., 2007; Gray, 2008; Storm et al., 2009; Puelles et al., 2012a; Puelles, 2013). Such results probably also can be extrapolated to the diencephalic tegmentum [e.g., the dopaminergic cell populations are continuously produced along a mesodiencephalic tegmental continuum, which also produces various other cell populations, such as neurons associated to the fasciculus longitudinalis medialis, Nkx6.1/6.2-positive elements of the pre-Edinger–Westphal nucleus (Puelles et al., 2012a), cells associated to the red nucleus and to the medial terminal nucleus of the accessory optic tract (Puelles, 2013)]. Such heterogeneity hardly would result from a homogeneous basal progenitor population. Therefore, the complexity we see at the hypothalamic basal plate may be just a differentially developed (expanded) version of the general case along the whole neural tube. There surely is a differential role of the patterning effects exerted here by the prechordal plate (and the adenohypophysis afterwards) in explaining any properties that selectively apply to the basal hypothalamus. Further study is needed to investigate whether the five longitudinal basal subzones presently postulated in the hypothalamus can be extrapolated individually backwards into thinner corresponding domains in the other brain areas, examining as well how far the respective genoarchitecture is shared throughout (vs. regional differences). Since the hypothalamic basal plate is the largest dorsoventrally, it may well occur that ventralizing signals diffusing from the floor plate dorsalward, and secondary antagonistic interactions between transient early gene patterns (such as those observed in the spinal cord), can be read out by the responding basal matrix cells into more distinct levels of genomically significant signal concentrations. This would imply that smaller basal plates might have not only thinner, but perhaps also less longitudinal microzones. This issue will no doubt be cleared in the near future.

A final issue that should be commented in this section is the proposal of Kuhlenbeck (1973) that the tuberal/retrotuberal region is alar in nature, being separated from the mamillary/retromamillary region sensu lato by an alternative alar-basal boundary. This conclusion was also reached recently by Diez-Roux et al. (2011), due to the expression within the tuberal region of a number of genes otherwise characteristic of the alar plate, such as Dlx and Arx genes. This hypothesis certainly simplifies the concept of the hypothalamic basal plate, reducing it to the M/RM and PM/PRM longitudinal domains, but complicates instead the schema of the alar plate, which would then have five longitudinal zones (Pa, SPa and the three Tu/RTu subdomains). This hypothesis implies a lack of linearity (a step) in the alar-basal boundary at the preincertal/incertal hypothalamoprethalamic border, an issue that will need additional analysis. The shared "alar" gene patterns in the tuberal region appear associated topographically to the sites where GABAergic neurons are produced (Puelles et al., 2012b). While no definitive explanation seems presently available for the fact that genes otherwise characteristic of the alar plate, such as Arx and Dlx, are expressed as well (with some differential characteristics) in the Shh-positive tuberal/retrotuberal territory, it is by no means extraordinary that, due to differential enhancer effects, the same gene can be activated independently in morphologically unrelated domains (for instance, Shh itself, held to be a ventral marker—even a floor plate marker by some authors—, is expressed also separately in the alar preoptic area). Such ectopic peculiarities should not confuse morphologic analysis. We are forced to take into account a variety of arguments in order to reach the most meaningful interpretations. Any single gene signal does not have a straightforward morphologic meaning. In this case, we hold as significant that there is a precocious molecular alar-basal division detectable in the forebrain already at neural plate stages, according to the early floor and basal expression of Shh plus a limiting band of Nkx2.2 expression (Shimamura et al., 1995). This creates a primary pattern that was recently corroborated by progeny analysis of Shhand Foxb1-derived populations (Zhao et al., 2008; Szabo et al., 2009), as well as chicken fate-mapping data (Sanchez-Arrones et al., 2009). These results pinpoint the dorsal limit of hypothalamic Shh expression, with the overlapping band of Nkx2.2 expression (**Figures 7A**, **10B**), as the primary alar-basal boundary. The added basal expression of Arx, Dlx and other gene markers listed by Diez-Roux et al. (2011) appears relatively later in development (after E10.5). We submit that this phenomenon may relate to differentiative decisions leading to the GABAergic phenotype adopted by many basal cells, mainly along the TuI/RTuI domain (the histaminergic neurons produced at the TuV/RTuV also share analogous markers).

### The *Acroterminal Hypothalamic Domain* as a Necessary Causal Background for the Unique Rostromedian Hypothalamic Specializations

As mentioned above, the rostromedian hypothalamic midline stretching between the mamillary region (end of floor plate) and the anterior commissure (end of roof plate)—see **Figure 9A** is singular in being patterned dorsoventrally (as opposed to anteroposteriorly, as is dictated by the columnar model— **Figure 9B**). Though neuroanatomic literature traditionally interprets this territory as extended along the length axis, due to the assumptions of the columnar model, its molecular patterning, which is already visible at neural plate stages (Puelles, 1995; Shimamura et al., 1995; Sanchez-Arrones et al., 2009) indicates instead that it should be understood as a singular rostromedian continuity of the lateral walls of the neural tube, representing the unpaired median place where the lateral walls—alar+basal primarily meet each other rostrally, on top of the rostralmost floor plate and under the rostralmost roof plate (**Figure 9A**). This peculiar rostromedian domain belonging to the THy shows in the adult various structural specializations (**Figure 13**). In its alar subregion there is dorsally the terminal lamina and the median preoptic nucleus (TL; MnPO), as well as the optic chiasm (OCH), ventrally; the terminal lamina is fixed dorsally to the anterior commissure (roof plate) and ventrally to the optic chiasm. At the latter transitional neighborhood, the terminal lamina shows an intensely vascularized median circumventricular organ (the organum vasculosum laminae terminalis; OVLT). The ventral aspect of the optic chiasm relates intimately to the postoptic decussations (these are topologically rather "suboptic," though they used to be named "supraoptic" in reference to the columnar axis); they apparently lie just above the alar-basal boundary (this is merely a tentative interpretation at this point, pending more detailed genoarchitectural analysis).

In its turn, the terminal median basal plate also shows a sequence of specializations: there is dorsally (close to the postoptic decussations) a median portion of the anterobasal area (ABasM; this is the primitive rostral end of the precociously differentiated hypothalamic cell cord, which used to be known as the "retrochiasmatic area," e.g., in Puelles et al., 1987a; note ABas is a prosomeric-consistent term, though it was introduced by Altman and Bayer, 1986, whereas RCH is columnar). The ABas is horseshoe-shaped and displays bilateral wings within the TuD

preoptic derivatives, ending with the organum vasculosum laminae terminalis

area (TM), finishing with the median mamillary area (MnM).

area of THy (**Figure 10B**). More ventrally, coinciding with the median part of TuI, there appears the median eminence and the associated arcuate nucleus (ME, Arc), as well as the infundibulum and the neurohypophysis (NH), whereas the underlying median TuV is represented by the tuberomamillary recess area (TM; **Figure 13**). Part of the median mamillary region (MnM) possibly participates of this extensive rostromedian (transverse) territory, immediately dorsal to the mamillary floor plate. Additional median or paramedian structures close to those described above also may be ascribed to the structurally singular rostromedian territory: for instance, the optic vesicles and their stalks (optic nerves), and the suprachiasmatic nucleus (SCH; **Figure 13**). Our criterion for adding the SCH to this singular region is that it is limited to a rostromedial sector of the THy, and does not reach the intrahypothalamic border. Accordingly, it needs a special causal underpinning, which most probably relates to the differential molecular profile of the rostromedian terminal area (see Ferran et al., 2015).

Indeed, these specializations in principle belong all to the THy, but they occupy a radially distinct territory at its rostralmost end, and none of them extend caudalwards across the whole THy, reaching the intrahypothalamic border. Their development must obey specific causes restricted to the rostromedian alar and basal midline and its immediate paramedian neighborhood. The differential histogenetic patterns observable at the standard THy entities that do reach the intrahypothalamic boundary (see list above) vs. the corresponding rostromedian specializations at each dorsoventral level are corroborated by the existence of developmental gene expression patterns distinguishing these two THy subregions (see Ferran et al., 2015). This might be construed eventually as evidence that the rostromedian hypothalamic terminal subdomain represents an extra, atypic hypothalamic neuromere (hp3). Though granting this possibility, we caution that finding support for this hypothesis would require to re-examine again the roof and floor plates, in order to verify that the requirement for a "complete" interneuromeric border can be satisfied (Puelles and Rubenstein, 2003). We have not obtained such evidence yet, so we keep this territory within hp2 and THy.

Meanwhile, it was thought convenient to have a specific name for this territory within the ampler concept of the terminal hypothalamic wall. Puelles et al. (2012b) proposed the novel term "acroterminal hypothalamic domain" (ATD), referring to its topologic location at the tip (Greek, acron) of the terminal wall. Accordingly, the descriptor "acroterminal" can be applied unambiguously to any of the mentioned specialized structures of this territory, as well as to the whole subregion, eschewing the continuous use of circumlocutions. Note the ATD is shared by the hypothalamus and the preoptic telencephalon (**Figures 9A**, **13**). It is well possible that the ATD is a direct consequence of the signaling activity of the prechordal plate along the median part of the terminal wall.

Interestingly, both alar and basal parts of the ATD seem to develop signaling properties, due to the localized expression of several members of the fibroblast growth factor family (Fgf8, Fgf10, Fgf18; see Ferran et al., 2015; **Figure 14**). Diffusion of these morphogens caudalwards from the ATD into the hypothalamus may be relevant for its segmentation into hp1 and hp2, and/or for detailed anteroposterior patterning of the alar and basal hypothalamic territories. For instance, the difference between the median eminence/arcuate nucleus complex and the terminal dorsomedial domain might obey to FGF signaling from the local basal ATD area. The alar part of the ATD shows bilateral spots of Fgf8 expression at the base of the optic stalks and a median line of Fgf18 expression along the terminal lamina (see Ferran et al., 2015). Signaling spreading from these alar loci might be relevant for the differential specification of the median OVLT and the bilateral SCH nuclei.

### Coda

Looking into the rationale of the novel aspects in the prosomeric model possibly has brought us to consider quite unexpected morphological and developmental results, which seem relevant one way or other for underpinning solidly our assumptions about forebrain structure, including that of the hypothalamus, in a realistic causal background. Progress apparently lies in increasing our awareness of such relevant developmental phenomena and their spatial and molecular characteristics, incorporating them coherently after due analysis into the model's assumptions and predictions. This surely improves its overall consistency and sturdiness, to the advantage of potential morphologic interpretations and causal explanations. Our take-home message is that a morphologic model helps us to think all the better, the deeper its roots extend into causal foundations.

A good model points out the apparent best options for our dealings with complex reality (either the planning of our research, or the analysis of results), but certainly does not represent a definitive Truth that stops us from considering heterodox novel ideas and possible changes to the model. Models must adapt to progress in knowledge, or will be superseded. In the past, neuroanatomic models first aimed to encompass gross aspects of adult brain structure as they appeared in dissections, and accordingly were very much man-made and wanting in precision. Then they incipiently started to consider aspects of dorsoventral and anteroposterior developmental pattern (columnar versus neuromeric models), but were hampered by the low resolutive power of the research methods available, and possibly also by misguided (premature) attention to functions. Finally, the progress of molecular biology, genomics and mechanistic developmental biology has brought in masses of new relevant data, leading us to the consequent need of models capable of encompassing causal mechanisms of structure in three dimensions. We can no longer accept that the brain longitudinal axis, or any other fundamental structural component, be defined arbitrarily (e.g., merely implied by the use of given descriptors), without express reference to known molecular aspects of developmental causation, irrespective whether we only have tentative solutions, or seemingly solid ones. This is the modern, promising way in which we look at the hypothalamus now, in the new molecular scenario.

Since we have not yet collected or analyzed all possible data, we must be ready to change our assumptions as the model evolves in response to new techniques, additional experimental results and more detailed thought. Importantly, the morphological model

of the hypothalamus should not be conditioned by functional preconceptions, as happened with the columnar model. Our justified interest in brain functions should find its proper place in the experimental analysis of the biology of living brain structure. Morphological models are important primarily as instruments to understand developing (evolving) brain structure. They allow us to produce increasingly detailed maps where causal mechanisms, differentiation patterns, connective pathways, synaptic fields and even neuro-pharmacological properties can be correlatively inscribed, first bi-dimensionally, and later in 3 dimensions. This complex and as yet incompletely fulfilled endeavor eventually should allow us to conceive multi-dimensional representations, which might be relevant for functional analysis, even though brain functions per se, representing dynamic capabilities

### References


of distributed interactive neural networks relative to the body and the world, hardly can find a fixed place in a morphological brain model.

**Quaerendo invenitis** (by asking, you will find) [J. S. Bach]

### Acknowledgments

This work was funded by the Spanish Ministry of Economy and Competitiveness grant BFU2008-04156 and the SENECA Foundation contract 04548/GERM/06 (no. 10891) to LP. Infrastructure support provided by the University of Murcia and the IMIB is also acknowledged.

Altman, J., and Bayer, S. A. (1986). The development of the rat hypothalamus. Adv. Anat. Embryol. Cell Biol. 100, 1–178. doi: 10.1007/978-3-642- 71301-9

Altman, J., and Bayer, S. A. (1988). Development of the rat thalamus: I. Mosaic organization of the thalamic neuroepithelium. J. Comp. Neurol. 275, 346–377. doi: 10.1002/cne.902750304

Altman, J., and Bayer, S. A. (1995). Atlas of Prenatal Rat Brain Development. Boca Raton, FL: CRC Press.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Puelles and Rubenstein. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Molecular codes defining rostrocaudal domains in the embryonic mouse hypothalamus**

*José L. Ferran1 \*, Luis Puelles <sup>1</sup> and John L. R. Rubenstein2*

<sup>1</sup> Department of Human Anatomy and Psychobiology, School Medicine, University of Murcia and IMIB (Instituto Murciano de Investigación Biosanitaria), Murcia, Spain, <sup>2</sup> Department of Psychiatry, Rock Hall, University of California, San Francisco, CA, USA

The prosomeric model proposes that the hypothalamus is a rostral forebrain entity, placed ventral to the telencephalon and rostral to the diencephalon. Gene expression markers differentially label molecularly distinct dorsoventral progenitor domains, which represent continuous longitudinal bands across the hypothalamic alar and basal regions. There is also circumstantial support for a rostrocaudal subdivision of the hypothalamus into transverse peduncular (caudal) and terminal (rostral) territories (PHy, THy). In addition, there is evidence for a specialized acroterminal domain at the rostral midline of the terminal hypothalamus (ATD). The PHy and THy transverse structural units are presently held to form part of two hypothalamo-telencephalic prosomeres (hp1 and hp2, respectively), which end dorsally at the telencephalic septocommissural roof. PHy and THy have distinct adult nuclei, at all dorsoventral levels. Here we report the results of data mining from the Allen Developing Mouse Brain Atlas database, looking for genes expressed differentially in the PHy, Thy, and ATD regions of the hypothalamus at several developmental stages. This search allowed us to identify additional molecular evidence supporting the postulated fundamental rostrocaudal bipartition of the mouse hypothalamus into the PHy and THy, and also corroborated molecularly the singularity of the ATD. A number of markers were expressed in Thy (Fgf15, Gsc, Nkx6.2, Otx1, Zic1/5), but were absent in PHy, while other genes showed the converse pattern (Erbb4, Irx1/3/5, Lmo4, Mfap4, Plagl1, Pmch). We also identified markers that selectively label the ATD (Fgf8/10/18, Otx2, Pomc, Rax, Six6). On the whole, these data help to explain why, irrespective of the observed continuity of all dorsoventral molecular hypothalamic subdivisions across PHy and THy, different nuclear structures originate within each of these two domains, and also why singular structures arise at the ATD, e.g., the suprachiasmatic nuclei, the arcuate nucleus, the median eminence and the neurohypophysis.

#### *Edited by:*

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### *Reviewed by:*

Kenji Shimamura, Kumamoto University, Japan Isabel Rodríguez-Moldes, University of Santiago de Compostela, Spain

#### *\*Correspondence:*

José L. Ferran, Department of Human Anatomy and Psychobiology, School of Medicine, University of Murcia, Murcia 30071, Spain jlferran@um.es

> *Received:* 30 January 2015 *Accepted:* 24 March 2015 *Published:* 17 April 2015

#### *Citation:*

Ferran JL, Puelles L and Rubenstein JLR (2015) Molecular codes defining rostrocaudal domains in the embryonic mouse hypothalamus. Front. Neuroanat. 9:46. doi: 10.3389/fnana.2015.00046 **Keywords: peduncular hypothalamus, terminal hypothalamus, acroterminal domain, genoarchitecture**

### **Introduction**

The developing mouse hypothalamus has been extensively investigated as regards its genoarchitecture in recent years (Shimogori et al., 2010; Diez-Roux et al., 2011; Puelles et al., 2012). Many results corroborated a fundamental organization in terms of a dorsoventral array of longitudinal zones showing characteristic molecular profiles. Thus, the alar hypothalamus appears divided dorsoventrally into the paraventricular and subparaventricular areas, and the basal hypothalamus likewise displays dorsoventrally arranged tuberal and mamillary areas (Morales-Delgado et al., 2011, 2014; Puelles et al., 2012; Díaz et al., 2015, this Issue). Each of these primary longitudinal domains can be subdivided into thinner longitudinal subdomains, or microzones (Puelles, 2013). The pattern bespeaks of an important role of dorsoventral patterning in the anatomical structure of the hypothalamus. It allows us to explain similarities in differentiation patterns (and eventually functions) shared along these histogenetically characteristic regions [e.g., precocious neurogenesis along a dorsal tuberal microzone (the classic hypothalamic cell cord), distribution of hypophysiotropic cell populations along the paraventricular area, the continuum formed by the dorsomedial and arcuate nuclei, origin of histaminergic neurons along the ventral tuberal microzone, and the distribution of *Otp*-expressing neurons within the mamillary region; the adult arrangement of glutamatergic and GABAergic cell populations also relates to the same dorsoventral pattern of longitudinal zones (Puelles et al., 2012)].

However, conventional analysis of nuclear structure in the hypothalamus already underlines the existence of grisea that apparently only develop within the rostral or caudal moieties of the cited longitudinal bands. Clearcut examples of this are the anterolateral, suprachiasmatic, anterior, ventromedial, ventral/dorsal premamillary and arcuate nuclei, only present rostrally, and the main paraventricular nucleus, jointly with the entopeduncular nuclei and the subthalamic nucleus, only present caudally. Moreover, distinct mamillary (rostral) and retromamillary (caudal) formations have been distinguished within the mamillary region. Such differences already suggested an anteroposterior bipartition of the hypothalamus (Puelles and Rubenstein, 2003; Puelles et al., 2012) finally postulated that such bipartition responded to the existence of two distinct neuromeric fields stretching through the hypothalamus and telencephalon continuum (hypothalamic prosomeres hp1 and hp2, named in caudorostral order) (**Figure 1A**). The corresponding hypothalamic parts were identified as *peduncular* and *terminal* hypothalamus (PHy, THy; note there is a rough correspondence of these *transverse* parts with historical use by Herrick, 1910 and others of the notion of "dorsal" and "ventral" hypothalamus, of course referring to a completely different length axis; this use has been abandoned by columnar authors in recent times, e.g., see Swanson, 1992, 2003; other authors identified THy as "hypothalamus" and PHy as "subthalamus," e.g., Reinoso-Suárez, 1960; Richter, 1965; see discussion in Puelles et al., 2012).

In any case, these antecedents suggest that, apart of longitudinal genoarchitectonic patterns, there is also room for anteroposterior molecular differences across PHy and THy (that is, hp1 and hp2), without forgetting the existence of a singular rostromedian transverse territory within THy, recently identified as the acroterminal domain (ATD; Puelles et al., 2012), where specialized formations are characteristic (**Figures 1A,B**). One expects some differential molecular pattern across adjacent neuromeres that share other determinants, particularly so when the respective adult derivatives are anatomically diverse. In that case, the crucial expected observation is that various partial zonal or microzonal patterns restricted to each of the units show a common transverse boundary that agrees with the postulated interneuromeric border. This corresponds mainly to the intrahypothalamic boundary in our field of interest, but also, secondarily, to the acroterminal vs. terminal boundary.

We already had partial knowledge of some such molecular anteroposterior differences (Shimogori et al., 2010; Puelles et al., 2012), but we resolved to perform a wider datamining study within the Allen Developing Mouse Brain Atlas database (developingmouse.brain-map.org), whose results we report here. We searched for gene markers expressed at early stages (E11.5, 13.5) whose expression domain either at the ventricular or mantle zones was significantly restricted to the terminal or peduncular moiety of one or several longitudinal zones. Interesting candidates were also examined at older stages, to assess the correlation of specific nuclei to the molecular pattern.

### **Materials and Methods**

We report here on 37 gene expression patterns (**Table 1**), analyzed from *in situ* hybridization images downloaded from the *Allen Developing Mouse Brain Atlas*. <sup>1</sup> These are mostly sagittal sections; while this section plane is appropriate for the analysis of spatial dorsoventral and anteroposterior topographic

**Abbreviations:** aATD, Alar acroterminal domain; ABas, Anterobasal nucleus; ABasM, Median acroterminal ABas; ABb, Alar basal boundary; AH, Anterior hypothalamic nucleus; AHy, Adenohyphophysis; AP, Alar plate; Arc, Arcuate nucleus; ATHy, Acroterminal hypothalamus; BP, Basal plate; CPa, Central portion of the peduncular paraventricular area; Di, Diencephalon; DM, Dorsomedial hypothalamic nucleus; DM-P, Dorsomedial hypothalamic nucleus—Peduncular part; DMcP, Core portion of the peduncular DM; DMcT, Core portion of the terminal DM; DMsP, Shell portion of the peduncular DM; DMsT, Shell portion of the terminal DM; DPa, Dorsal portion of the peduncular paraventricular area; EPV, Ventral entopeduncular nucleus; FP, Floor plate; Hb, Habenula; hp1, Hypothalamo-telencephalic prosomere 1; hp2, Hypothalamo-telencephalic prosomere 2; M, Mamillary region; M/RM-FP, Mamillary/retromamillary floor plate; Mb, Midbrain; MCLH, Magnocellular lateral hypothalamic nucleus; ME, Median eminence; mtg, Mamillotegmental tract; NHy, Neurohypophysis; NHyS, Neurohypophyseal salk; OCH, Optic chiasm; os, Optic stalk; Pa, Parventricular hypothalamic complex; Pall, Pallium; PBas, Posterobasal area and nucleus; PHy, Peduncular hypothalamus; PM, Perimamillary region; PM/ATD, Acroterminal part of perimamillary region; POA, Preoptic area; PPa, Peduncular part of paraventricular area; PRM, Periretromamillary region; PSPa, Peduncular part of subparaventricular area; PSTh, Para-subthalamic nucleus; p1, Prosomere 1; p2, Prosomere 2; p3, Prosomere 3; PT, Pretectum; PTh, Prethalamus; rf, Retroflex tract; RM, Retromamillary region; RP, Roof plate; RPa, Rostral paraventricular nucleus; RTu, Retrotuberal area; SCH, Suprachiasmatic nucleus; SPa, Subparaventricular area; SPall, Subpallium; STh, Subthalamic nucleus; T/ATD, Acroterminal domain of the tuberal region; Tel, Telencephalon; Th, Thalamus; THy, Terminal hypothalamus; TPa, Terminal part of paraventricular area; TSPa, Terminal part of subparaventricular area; TSO, Terminal supraoptic nucleus; Tu, Tuberal region; TuD, Dorsal tuberal domain; TuI, Intermediate tuberal domain ; TuSbO, Tuberal suboptic nucleus; VM, Ventromedial hypothalamic nucleus; VMc, Core portion of the VM; VMs, Shell portion of the VM; VPa, Ventral portion of the peduncular paraventricular area; VPM, Ventral premamillary nucleus.

<sup>1</sup>http://developingmouse.brain-map.org/

**genoarchitectonic subdivisions of the hypothalamus are detailed. (A)** Schema showing the two rostral diencephalic prosomeres (p2 and p3) and the hypothalamo-telencephalic prosomeres 1 and 2 (hp1 and hp2). Note hp1 contains the peduncular hypothalamic region (PHy), whereas hp2 includes the terminal hypothalamic region (THy) and the rostralmost, median acroterminal domain (ATHy). **(B)** Schema of the main hypothalamic progenitor areas distributed across the dorsoventral and anteroposterior dimensions. The longitudinal alar/basal boundary (ABb) is indicated as a thick dark line. The hypothalamic area is subdivided rostrocaudally into neuromeric THy and PHy parts (pink and green, respectively). Alar territories (AP) are shown on the left, and basal ones on the right. The alar hypothalamus is

central and ventral subdivisions (DPa, CPa, VPa). The basal hypothalamus is also subdivided dorsoventrally into the large tuberal/retrotuberal (Tu/RTu) area and the primary mamillary/retromamillary (M/RM) area, plus the corresponding acroterminal regions. The THy/PHy parts of the hypothalamic floor lie underneath (FP). Moreover, the Tu/RTu region is subdivided into three dorsoventral parts: TuD/RTuD, TuI/RTuI, and TuV/RTuV, and the primary M/RM area is subdivided into perimamillary/periretromamillary area (PM/PRM) and the secondary M/RM area. Some well-known nuclear elements of the hypothalamus are represented within their respective topography relative to the molecular domains; note some of these positions are postmigratory (see the list for abbreviations).

relationships needed in our analysis, the visualization of some anatomic landmarks may be compromised. We recurred to careful analysis of all sagittal (eventually also coronal) section planes shown at the *Allen Atlas*, as well as to our extensive experience with multiple planes of sections through the mouse hypothalamus. We correlated the positions of labeled cells at the time


**TABLE 1 | 37 genes analyzed, classified according to their mouse genome informatics ID, gene family classification (HUGO) and functional gene ontology (GO).**

points available at the *Allen Atlas* (embryonic days E11.5, E13.5, E15.5, and E18.5, and postnatal day P4) with the genoarchitectonically distinct areas and conventional nuclei, following the model of Puelles et al. (2012).

### **Results**

In this section we will describe selected examples that best represent expression restricted to either PHy or THy, or to the ATD.

#### **Acroterminal Patterns**

We deal first with observations at the ATD, that is, with genes restricted in expression to the ATD, or combining ATD and THy patterns.

The gene *Rax*, for instance, appears selectively expressed at E11.5 and E13.5 in ventricular cells along the tuberal sector of the ATD; the neurohypophysis primordium lies centered within this domain, and is also positive (**Figures 2A,B**, **6B**). It is unclear whether this tuberal domain includes the complete TuD subdomain, or rather only its ventralmost part (see **Figure 2B**;

**FIGURE 2 | Sagittal, parasagittal (A,B–E,H–Z) and horizontal (F,G) sections through the mouse secondary prosencephalon and diencephalon at E11.5, E13.5, and E15.5, showing relevant examples of selective hypothalamic gene expression at the acroterminal domain (ATD) and terminal territory:** *Rax* **(A–C),**

*Prdm12* **(D–E'),** *Nr5a1* **(F,G,L),** *Pomc* **(H–J,M),** *Otx2* **(K),** *Six3* **(N–P),** *Fgf8* **(Q,R),** *Six6* **(S–U),** *Fgf18* **(V,Y,Z), and** *Fgf10* **(W).** All images were downloaded from the Allen Developing Mouse Brain Atlas (http://developingmouse.brain-map.org/). For abbreviations, see the list.

compare other patterns shown in **Figures 2E,I**). In contrast, the perimamillary and mamillary parts of the ATD remain free of label, as does the whole alar ATD. At E13.5, some *Rax* signal also appears spread out into the adjacent terminal tuberal territory (TuI; **Figure 2C**), but was not observed to reach any part of the PHy.

The gene *Prdm12* is expressed already at E11.5 at the dorsal tuberal part of THy (TuD), restricted to the mantle zone, and clearly encompassing the corresponding median ATD territory (**Figures 2D,D**- ). Two days later, this pattern is amplified throughout the anterobasal derivatives of TuD (**Figures 2E,E**- ), and additional signal appears more laterally, distributed along the alar subparaventricular band (both THy and PHy; not shown).

At E11.5 and E12.5, there is expression of *Nr5a1* that is clearly restricted to the TuD sector of the ATD (TuD/ATD, or ABasM area; **Figures 2F,G,L**). At later stages this marker also appears within the adjoining TuD zone of THy (TuD/THy, or ABas area; **Figures 4H,J**, **6F**); Puelles et al. (2012) commented on the relationship of these TuD-originated primordia in the development of the migrated ventromedial nucleus.

*Pomc* was first described in an early median population by Shimogori et al. (2010); these cells apparently migrate subsequently into the arcuate nucleus area (TuI/ATD), as was corroborated by Puelles et al. (2012). The latter authors ascribed the initial *Pomc*-positive locus to the ABasM (the TuD/ATD sector). We examined this pattern again at E11.5, confirming positive mantle cells within ABasM, but accompanied by more dorsal cells within the overlying alar sectors of the ATD; the signal seemed restricted to the ventricular zone at these alar loci (**Figures 2H,M**); it was no longer present at E13.5, when *Pomc* cells densely aggregate at the ABasM mantle (**Figure 2I**). This result opens the possibility that the real origin of *Pomc* cells lies in the alar ATD. Long dorsoventral tangential migrations of peptidergic hypothalamic neurons are by no means rare (see Díaz et al., 2015; this Issue). At E15.5 the ABasM population appears spread out dorsoventrally, consistently with its postulated final migration into the arcuate region (**Figures 2J**, **6B**).

*Otx2* appears expressed selectively along the mamillary and tuberal sectors of the ATD up to and including the neurohypophysis primordium at E11.5; interestingly, the selected midsagittal section also shows widespread labeling of the whole roof plate, all the way to its end at the anterior commissure locus, as well as of the whole floor plate, all the way to the mamillary region; this causes here continuity of floor- and ATD-related signal (**Figures 2K**, **6D**). At E13.5 this ATD expression of *Otx2* is much reduced, but some signal still remains at the neurohypophysis and at the perimamillary sector of the ATD, next to novel signal appeared within the mamillary area of THy (**Figure 5N**).

*Six3*, a gene already expressed in the rostral neural plate, is well-known to become restricted subsequently as a marker of the rostromedian telencephalon and the alar and basal hypothalamus, but excluding the mamillary region (Puelles et al., 2012). This ATD-related distribution is shown here at E13.5 (**Figures 2N**, **6E**); more lateral sections reveal that *Six3* signal also spreads into adjacent alar and tuberal basal parts of THy, but not into PHy (**Figures 2O,P**, **6E**). Curiously, irrespective of this restriction of its neural tube expression domain along both DV and AP dimensions, the whole hypothalamus disappears in the null mutation (Lagutin et al., 2003). As nuclear structure develops, *Six3* signal appears mainly at the suprachiasmatic nucleus (one of the ATD derivatives; not shown).

On the other hand, *Six6* appears more selectively expressed at the alar ATD, as well as at the TuD sector of the basal ATD, at E11.5 (weak; not shown) and E13.5 (**Figures 2S,T**, **6A**). The pattern is unchanged at E15.5, though the signal seems to predominate in the ATD regions of the alar paraventricular and subparaventricular areas (**Figure 2U**).

We examined all *Fgf* genes, and found some relevant aspects. At E11.5 *Fgf8* is expressed strongly at the perimamillary ATD sector, and more weakly along the tuberal ATD sector, just including the neurohypophysis anlage (**Figures 2Q**, **6C**). However, this pattern appears much reduced at E13.5, with signal remaining only (weakly) at the PM/ATD (**Figure 2R**). Note this small sector also appears labeled at E13.5 with *Otx1* (**Figure 5M**), *Otx2* (**Figure 5N**) and *Cadm1* (**Figure 5R**). *Fgf10* roughly reproduces the initial tuberal ATD expression domain of *Fgf8*, with stronger signal, though the PM/ATD sector is in this case inconspicuous (**Figures 2W**, **6C**); at later stages the *Fgf10* signal becomes restricted to the neurohypophysis (not shown). Finally, *Fgf18* transiently appeared expressed at E11.5 within the alar ATD region (**Figures 2V,Y**, **6A**); but at E13.5 a weak signal is observed at the PM/ATD (**Figure 2Z**).

The gene *Nkx6.2* was found expressed only bilaterally at the ABas area (TuD/THy) at E11.5 (**Figures 4F**, **7A**), but no signal was detected in two intervening sections depicting the midsagittal plane (not shown). We conclude from this that the ATD is specifically excluded from this expression pattern.

A further gene pattern highlighting the ATD is *Sall3*. This pattern covers at E13.5 essentially the whole basal ATD, starting with rather weak signal at its TuD sector. The expression is weakest at the neurohypophysis anlage and then becomes strong ventral to it, along the TuI, TuV (TM) and mamillary sectors (**Figure 5P**). This gene is also expressed within ventricular and mantle zone of the mamillary area of THy, but distinctly not at the mantle zone of the retromamillary area of PHy (data not shown).

### **Terminal Hypothalamus Patterns**

*Zic5*, belonging to the *Zic* gene family related to the specification of alar neural regions, shows an interesting dynamic expression pattern. At E13.5 its signal appears restricted to the ventricular zone of the alar paraventricular area across both THy and PHy (extending into the eye stalk, which is an ATD component). However, this expression is distinctly weaker at PHy than at THy (**Figures 3A,B**). At E15,5, the paraventricular mantle zone shows strong expression (some cells also disperse into the subjacent subparaventricular area), but only at the THy moiety (**Figures 3C**, **6F**). This situation clearly remains unchanged at P4 (**Figure 3D**; note the blank main paraventricular nucleus and the negative suprachiasmatic nucleus. For comparison we inserted the pattern of *Otp*, which shows an equally dense labeled mantle in THy and PHy (**Figure 3E**), and the pattern of *Tbr1* at P4,

which emphasizes the PHy derivative of the paraventricular area (**Figure 3H**, to compare with **Figure 3D**).

**and peduncular (PHy) territories:** *Zic5* **(A–D),** *Otp* **(E),** *Zic1* **(F),** *Ascl1*

*Zic1* also seems predominantly expressed in the terminal paraventricular area (**Figure 3F**).

*Ascl1* is selectively expressed at E13.5 within the terminal subparaventricular area mantle (**Figure 3G**).

dotted line: PHy/p3 boundary.

*Rgs4* seems first restricted to the peduncular paraventricular area at E13.5 (**Figures 3I,J**), though additional signal

**FIGURE 4 | Sagittal, parasagittal (A–H,J–L,O,P,T) and transversal (I,M,N,Q–S) sections through the mouse secondary prosencephalon and diencephalon at E11.5, E13.5, E15.5, and E18.5, showing relevant examples of the hypothalamic genes expressed selectively at the terminal (THy), peduncular (PHy) or acroterminal (ATD)**

*Nr5a1* **(H–J),** *Satb2* **(K–O),** *Tcf7l2* **(P–S), and** *Rprm* **(T).** All images were downloaded from the Allen Developing Mouse Brain Atlas (http://developingmouse.brain-map.org/). For abbreviations, see the list. Red dotted line: THy/PHy boundary. Blue dotted line: PHy/p3 boundary.

appears within the terminal basal territory. Various positive basal populations of THy are distinguished at E15.5 (**Figures 3K**, **7C**). Practically the whole terminal tuberal and mamillary regions show *Rgs4* signal at P4, contrasting with only few and dispersed positive cells within the basal PHy (**Figure 3L**; note in particular that the ventral premamillary nucleus, a migrated derivative of

**FIGURE 5 | Sagittal and parasagittal (A–T) sections through the mouse secondary prosencephalon and diencephalon at E11.5, E13.5, E15.5, E18.5, and P4, showing relevant examples of hypothalamic genes expressed selectively at the terminal (THy), peduncular (PHy) or acroterminal (ATD) territories:** *Plagl1* **(A–D),** *Foxp1* **(E),** *Irx1* **(F,G),**

*Irx5* **(H),** *Irx3* **(I,J),** *Erbb4* **(K),** *FoxB2* **(L),** *Otx1* **(M,O),** *Otx2* **(N),** *Sall3* **(P),** *Cadm1* **(Q,R), and** *Calb1* **(S,T).** All images were downloaded from the Allen Developing Mouse Brain Atlas (http://developingmouse.brain-map.org/). For abbreviations, see the list. Red dotted line: THy/PHy boundary. Blue dotted line: PHy/p3 boundary.

**patterns in the hypothalamus, illustrating studied patterns selective for the acroterminal (ATHy), terminal (THy) or peduncular (PHy) hypothalamic domains.** Compare subdivisions with **Figure 1**. **(A)** The Fgf18 and Six6 domains overlap within alar ATHy domains; but only Six6 (red tag) is detected at the acroterminal TuD domain. **(B)** Rax and Pomc were detected along the acroterminal tuberal domains; but only Rax was observed at the NHy (red tag). **(C)** Fgf8 and Fgf10 were detected along the intermediate and ventral basal tuberal acroterminal domains, including NHy, (Continued)

#### **FIGURE 6 | Continued**

extending also into the acroterminal perimamillary area. **(D)** Otx2 was observed at the acroterminal TuI, TuV, PM, and M domains, as well as along the THy and PHy floor plate. **(E)** Six3 was throughout the alar domains of ATHy and THy; but at the basal plate its expression was restricted to the Tu region of ATHy and THy, and ATHy of the PM domain. **(F)** Zic5 expression was detected at the alar TPa (and corresponding ATHy area) and TSPa

the retromamillary area according to Puelles et al., 2012, appears completely negative within the tuberal zone).

We found expression of *Fgf15* restricted at E11.5 to the terminal alar hypothalamus, though leaving the acroterminal optic stalk negative (**Figure 3M**). *Gsc* expression was observed selectively in the terminal paraventricular zone at E13.5 and E15.5 (**Figure 3Q**). The terminal pattern of *Nkx6.2* was described above (**Figure 4F**).

It is unclear whether *Lef1* expression in the tuberal hypothalamus shows a difference between THy labeling (which seems extensive) and PHy labeling (which seems minoritary, restricted to a small tail-like area; **Figure 4G**). Note no signal corresponds to the ventromedial nucleus, possibly due to the ectopic migrated nature of this tuberal mass, originated at the overlying TuD domain (Puelles et al., 2012).

The VM primordium can be selectively labeled with the *Nr5a1* marker, as advanced above (**Figures 4H–J**, **6F**); note the crosssection shows the migratory course, leading TuD–originated cells into TuI. A similar labeling pattern restricted to THy is provided by *Satb2*, whose initial expression is associated to TuD (ABas/THy and ABasM/ATD) (**Figures 4K–N**) and later the cells translocate analogously to *Nr5a1*–expressing cells into the definitive VM nucleus, particularly into its rounded dorsomedial subnucleus (**Figure 4O**). Another gene marker associated to the VM is *Tcf7l2*, a massive marker of thalamus and pretectum, which in addition already shows at E15.5 label at the TuD area of THy and ATD, from where labeled cells translocate subsequently into the migrated VM (**Figures 4P,Q**).

In addition, *Tcf7l2* also labels selectively the hypophyseal stalk within ATD, as well as the perimamillary band within THy (**Figures 4R,S**). The perimamillary THy band was also selectively labeled by other markers, such as *Zic1* (**Figure 3F**), *Foxp1* (**Figure 5E**) and *Otx1* (**Figures 5O**, **7B**).

A surprising finding was the expression of *Rprm* at the parasubthalamic nucleus, while the subthalamic nucleus itself remains wholly devoid of this signal (**Figure 4T**). It had been previously supposed that the PSTh originated jointly with the STh at the peduncular retromamillary area (Puelles et al., 2012). The present result suggests that perhaps the PSTh is terminal in origin, or at least comes from a different origin than the STh nucleus.

Several genes show expression restricted to the terminal mamillary body complex, leaving the retromamillary area negative. However, these patterns may show restriction within the mamillary area itself, sometimes to quite small cell groups difficult to interprete. One mystifying cell group appears labeled selectively by *FoxB2* at E13.5 (**Figures 5L**, **7F**). It parallels the origin of the mamillotegmental tract and apparently represents domains of THy (but respecting the local acroterminal suprachiasmatic nucleus). Additionally, Zic5 signal also appeared restricted to the PM region of THy. **(G)** Fgf15 and Gsc were detected jointly at the TPa area; but only Gsc was detected in the PSPa domain (red tag). **(H)** Nr5a1 expression was detected in the TuD domain across ATHy and THY, and the migrated derivatives of these areas entering the ventromedial nucleus also expressed this gene within TuI.

only a minor part of the medial mamillary nucleus. *Otx1* labels faintly the mamillary ventricular zone (**Figure 5M**), while *Otx2* also labels the mantle zone (**Figure 5N**). *Cadm1* only labels a compact superficial population visible at relatively lateral section levels, possibly related to the lateral mamillary nucleus (**Figure 5Q**). Finally, *Calb1* is also expressed in a subpopulation of the mamillary body which lies deep to the brain surface at intermediate section levels (**Figures 5S,T**); note there is also *Calb1* expression restricted to the VM nucleus.

#### **Peduncular Hypothalamus Patterns**

As already mentioned, at early stages *Rgs4* shows restricted labeling of the peduncular component of the paraventricular area. At E13.5 this seems to imply the whole dorsoventral extent of the main paraventricular nucleus (**Figures 3I,J**), though at E15.5 the expression seems restricted to its ventral component (**Figures 3K**, **7C**).

*Lmo4* and *Mfap4* likewise appear expressed at the peduncular paraventricular area, predominantly at its central and ventral subdivisions (**Figures 3N–P**, **7E**).

*Gsc* appears expressed selectively at E15.5 within the peduncular portion of the subparaventricular area, apart separate expression in the terminal paraventricular area is observed (**Figures 3Q**, **6G**). Another marker distinguishing this PHy subparaventricular domain is *Meis2*. At E13.5 its expression seems to extend partly into the THy (**Figure 4A**), but at E18.5 the signal is limited strictly to PHy (**Figures 4B**, **7F**). Moreover, *Vax1* shows at E15.5 differential subparaventricular labeling, with weaker and more disperse signal within the THy (suprachiasmatic and anterior hypothalamic nuclei), including some ventrally migrated cells in the underlying tuberal area, and more compact signal within the PHy component (**Figure 4C**).

*Plagl1* expression is restricted at E15.5 to the basal PHy, involving the retromamillary area (RM), the periretromamillary area (PRM) and the peduncular dorsomedial nucleus (DM-P, belonging to the RTu area), but not the posterobasal area (**Figures 3R**, **5A–D**, **7D**).

The precociously differentiating superficial basal cell group known as the magnocellular lateral hypothalamic nucleus (MCLH), a derivative of the posterobasal area (the PHy component of the RTuD area) appears selectively marked with *Pmch*, as shown here at E15.5 and E18.5 (**Figures 4D,E**, **7G**).

Several genes show restricted expression at the retromamillary area. For instance, *Irx1* (**Figures 5F,G**, **7H**), *Irx5* (**Figures 5H**, **7H**), *Irx3* (**Figures 5I,J**, **7H**), *Erbb4* (**Figures 5K**, **7H**), and *Sall3* (**Figure 5P**; compare with **Figure 5J**).

**FIGURE 7 | Schematic maps of characteristic genoarchitectonic patterns in the hypothalamus, illustrating studied patterns selective for the terminal (THy) or peduncular (PHy) hypothalamic domains.** Compare subdivisions with **Figure 1**. **(A)** Nkx6.2 was detected in the TuD domain of THy. **(B)** Otx1 expression was observed in the PM, M, and FP areas across ATHy and THy. **(C)** Rgs4 was detected at the basal TuI, TuV, and M areas from THy, as well as throughout the Pa area from PHy. **(D)**

Plagl1 was detected at the basal RTuI, RTuV, PRM, and RM from PHy. **(E)** Lmo4 and Mfap4 were expressed selectively at the CPa and VPa paraventricular subdivisions within alar PHy. **(F)** Meis2 signal (red tag) was detected at the small PSPa domain; FoxB2 (red tag) was observed in a previously unknown dorsal subdivision of M domain (THy). **(G)** Pmch signal was detected only at the RTuD domain from basal PHy. **(H)** Erbb4, Irx1, Irx3, and Irx5 were detected at the RM and PHy FP.

### **Discussion**

Our results clearly illustrate that, as expected, the acroterminal, terminal, and peduncular subregions of all the shared dorsoventral domains are further differentially specified by a number of genes. These molecular subdivisions probably underpin causally the development of particular nuclei within these areas. Most of the genes studied code DNA-binding proteins, which probably are involved in fate and identity specification. These observations in general agree with additional data reported by Shimogori et al. (2010), Diez-Roux et al. (2011) and Puelles et al. (2012).

While in most cases the observed restricted expression domains were limited to a single longitudinal zone, in the cases of *Rgs4* and *Plagl1* a large part of the basal plate area, including the TuI/RTuI, PM/PRM, and M/RM domains was neatly divided into THy and PHy moieties. In all cases the labeled domains stopped at the postulated intrahypothalamic boundary, corroborating its role as a general interneuromeric boundary.

Remarkably, the acroterminal domain appeared also well delimited molecularly from the remaining part of THy, so that we can speak now of a distinct acro-terminal border. This domain also showed a diversity of molecular profiles along its dorsoventral dimension, illustrating that its different sectors producing characteristic derived structures are also differentially specified by the antagonism of dorsalizing and ventralizing patterning mechanisms.

Most of the gene expression patterns analyzed are consistent with neuroepithelial compartments previously delimited in the prosomeric model (Puelles et al., 2012); in contrast, these expression patterns cannot be explained or classified on the basis of columnar ideas, since this alternative model was not developed to the point of postulating such fine subdivisions. From the viewpoint of the updated prosomeric model these results demonstrate that the postulate of two hypothalamo-telencephalic prosomeres (hp1, hp2), which cut transversely across the set of five main longitudinal dorsoventral zones previously distinguished in the hypothalamus (**Figure 1**), dividing each of them in two molecularly distinct anteroposterior domains, satisfies the assumption that differential gene expression patterns should characterize each of these prosomeres, irrespective that they evidently share the fundamental dorsoventral molecular regionalization of the hypothalamus.

The supposed evolutionary advantage of neuromeres lies precisely in their capacity to maintain a common organization of shared properties along the dorsoventral dimension of a series of developmental units (metamerism), while varying subtly the molecular profile (molecular identity) of each neurogenetic unit along the anteroposterior axis, that is, varying the sorts of neurons produced within similar environments. This allows generic properties, such as the capacity to guide the longitudinal growth of the optic tract, to be combined with the differential development of specific target areas for this tract within given prosomeres; this multiplies with a minimum of genetic instructions the sorts of operational algorithms that the brain can perform on any given afferent signal.

We also found some unexpected data, representing molecularly distinct mantle subdivisions that are smaller than the partitions contemplated so far in the updated prosomeric model (e.g., *FoxB2* expression). These results suggest that there may be grounds to further analyze morphologically and developmentally the regions of the model where this occurs, an endeavor that may lead eventually to a more advanced and comprehensive version of the model.

The specialized median acroterminal domain postulated by Puelles et al. (2012) was a logical necessity, since the differential structures observable in this domain, both as regards the dorsoventral dimension (roof, alar, basal, and floor sectors) and with respect to the caudally adjacent structural elements of the THy, only could be conceived if the underlying causal conditions, including molecular ones, are likewise differential. Though some evidence regarding our prediction of differential gene markers within the ATD was adduced already by Puelles et al. (2012), recollected from earlier reports cited therein, it was conceptually important to fill all the remaining theoretical pigeonholes with some corresponding singular markers. This aim was satisfactorily fulfilled with our present analysis of available data in the Allen Developmental Mouse Brain Atlas database, though many other selective markers probably will appear in the future. Traditionally it had been habitual to look away from this area in gene mapping studies, probably because it seemed to behave differently from the basic structural schema of the hypothalamus found in textbooks. The acroterminal concept and the presently available evidence of its distinct molecular signature, simultaneously immersed in the general molecular pattern of the hypothalamus, hopefully will attract attention to it and to its particular mode of morphogenesis.

### **Acknowledgments**

This work was funded by the Spanish Ministry of Economy and Competitiveness grant BFU2008-04156 and the SENECA Foundation 04548/GERM/06 (no. 10891) to LP, and Nina Ireland, NIMH R01 MH081880, and NIMH R37 MH049428 to JR. Infrastructure support provided by the University of Murcia is also acknowledged.

### **References**


the mouse embryo. *PLoS Biol.* 9:e1000582. doi: 10.1371/journal.pbio.10 00582


forebrain development. *Genes Dev*. 17, 368–379. doi: 10.1101/gad.10 59403


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Ferran, Puelles and Rubenstein. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Patterns of hypothalamic regionalization in amphibians and reptiles: common traits revealed by a genoarchitectonic approach

### **Laura Domínguez, Agustín González and Nerea Moreno\***

Faculty of Biology, Department of Cell Biology, University Complutense of Madrid, Madrid, Spain

#### **Edited by:**

Luis Puelles, Universidad de Murcia, Spain

#### **Reviewed by:**

Nobuaki Tamamaki, Kumamoto University, Japan Luis Puelles, Universidad de Murcia, Spain Alino Martinez-Marcos, Universidad de Castilla, Spain

#### **\*Correspondence:**

Nerea Moreno, Faculty of Biology, Department of Cell Biology, University Complutense of Madrid, Av. José Antonio Novais 2, Madrid E-28040, Spain e-mail: nerea@bio.ucm.es

Most studies in mammals and birds have demonstrated common patterns of hypothalamic development highlighted by the combination of developmental regulatory genes (genoarchitecture), supporting the notion of the hypothalamus as a component of the secondary prosencephalon, topologically rostral to the diencephalon. In our comparative analysis we have summarized the data on the expression patterns of different transcription factors and neuroactive substances, used as anatomical markers, in the developing hypothalamus of the amphibian Xenopus laevis and the juvenile turtle Pseudemys scripta. This analysis served to highlight the organization of the hypothalamus in the anamniote/amniotic transition. We have identified supraoptoparaventricular and the suprachiasmatic regions (SCs) in the alar part of the hypothalamus, and tuberal and mammillary regions in the basal hypothalamus. Shared features in the two species are: (1) The supraoptoparaventricular region (SPV) is defined by the expression of Otp and the lack of Nkx2.1/Isl1. It is subdivided into rostral, rich in Otp and Nkx2.2, and caudal, only Otp-positive, portions. (2) The suprachiasmatic area contains catecholaminergic cell groups and lacks Otp, and can be further divided into rostral (rich in Nkx2.1 and Nkx2.2) and a caudal (rich in Isl1 and devoid of Nkx2.1) portions. (3) Expression of Nkx2.1 and Isl1 define the tuberal hypothalamus and only the rostral portion expresses Otp. (4) Its caudal boundary is evident by the lack of Isl1 in the adjacent mammillary region, which expresses Nkx2.1 and Otp. Differences in the anamnio-amniote transition were noted since in the turtle, like in other amniotes, the boundary between the alar hypothalamus and the telencephalic preoptic area shows distinct Nkx2.2 and Otp expressions but not in the amphibian (anamniote), and the alar SPV is defined by the expression of Otp/Pax6, whereas in Xenopus only Otp is expressed.

**Keywords: hypothalamus, prosencephalon, forebrain patterning, development, evolution**

#### **THE HYPOTHALAMUS WITHIN THE CURRENT PROSOMERIC MODEL**

The hypothalamus is considered the forebrain territory par excellence dedicated to control homeostatic processes, and its neuroanatomical regionalization has been a much debated topic

**Abbreviations:** BH, basal hypothalamus; CeA, central amygdala; CPa, central portion of the paraventricular area; CT, caudal tuberal region; Dg, diagonal domain; DPa, dorsal portion of the paraventricular area; EPTh, prethalamic eminence; hp1, hypothalamic prosomeric domain 1; hp2, hypothalamic prosomeric domain 2; Hyp, hypophysis; M, mammillary region; Ma, mammillary area proper; MeA, medial amygdala; oc, optic chiasm; P1, prosomere 1; P2, prosomere 2; P3, prosomere 3; P3b, basal plate of P3; Pa, paraventricular nucleus; Pal, pallidum; PM, perimammillary area; PO, preoptic region; POC, preoptocommissural area; POH, preoptohypothalamic boundary; PPa, peduncular domain of Pa; PSPa, peduncular domain of SPa; PT, pretectum; PTh, prethalamus; PR, perimammillary area; PRM, periretromammillary area; RM, retromammillary area; RMa, retromammillary region; RT, rostral tuberal region; RTu, retrotuberal region (peduncular); RtuD, retrotuberal dorsal domain; in recent years. The term "hypothalamus" was coined during the last century with the beginning of neuroanatomical studies (His, 1893a,b), following a columnar conception of the brain (Herrick, 1910; Kuhlenbeck, 1973). This was based on the idea that the forebrain is organized in longitudinal functional units homologous to the ones in the brainstem and it was considered that the ventricular sulci marked the straight longitudinal axis of the forebrain, ending somewhere in the telencephalon (Herrick, 1948; Kuhlenbeck, 1973). Following this concept and the analysis of classical "transverse" sections, the hypothalamus was defined

RTul, retrotuberal lateral domain; RTuV, retrotuberal ventral domain; SC, suprachiasmatic region; SCc, caudal suprachiasmatic region; SCr, rostral suprachiasmatic region; Spa, subparaventricular area; SP, subpallium; SPV, supraoptoparaventricular region; SPVc, caudal supraoptoparaventricular region; SPVr, rostral supraoptoparaventricular region; Str, striatum; VPa, ventral portion of the paraventricular area; Th, thalamus; TPa, terminal domain of Pa; TPaC, central portion of TPa; TPaD, dorsal portion of TPa; TPaV, ventral portion of TPa; TSPa, terminal domain of SPa; Tu, tuberal as a diencephalic region beneath the thalamus (from the old Greek ÿpó: under). However, the hypothalamus is formed, as the rest of the forebrain, from the anterior neural plate through complex processes of morphogenesis. As a result, this brain region in the mature brain is highly distorted, mainly by the sharp flexure of the longitudinal brain axis and by differential degree of development of its components. These phenomena make it difficult to identify the basic units or subdivisions in the mature hypothalamus and understand the topological relationships between them. Moreover, the variable degree of elaboration and differentiation of structures in the hypothalamus of the different vertebrates obscures the interpretation of anatomical data and the comparison across species, and greatly complicates studies of forebrain evolution (Butler and Hodos, 2005; Bruce, 2008; Hodos, 2008; Nieuwenhuys et al., 2008).

Twenty years ago, the first proposal of the prosomeric model pointed out several evidences to discard the columnar paradigm of the forebrain organization, revealing the discrepancy between the traditional anatomical landmarks and the morphogenetic organization of the brain, what eventually led to refute the boundary role of the ventricular sulci (Puelles and Rubenstein, 1993; Rubenstein et al., 1994; Puelles, 1995). In this model, the forebrain is organized into transverse segments (prosomeres) and longitudinal zones defined by differential expression patterns of several developmental regulatory genes that establish the internal boundaries. According to the original prosomeric model and its subsequent revisions (Puelles and Rubenstein, 1993, 2003; Puelles, 1995, 2001; Puelles et al., 2012a) the hypothalamus is excluded from the diencephalon, which is composed of three neuromeres (prosomeres P1–P3). The rostralmost forebrain is designated the secondary prosencephalon that contains the hypothalamus (rostral to the diencephalic P3), the telencephalon impar, and the telencephalic hemispheres (Puelles and Rubenstein, 2003). The interpretation of the parts of the secondary prosencephalon is fraught with difficulties, mainly derived from the early optic and hemispheric evaginations and the different degree of development shown across vertebrates that disturb the primary pattern of this region (Nieuwenhuys et al., 2008). However, morphological, molecular, and hodological results have progressively contributed to highlight the organization of the main parts of the secondary prosencephalon and its subdivisions, particularly in mice, where different organization models have been proposed (Figdor and Stern, 1993; Puelles and Rubenstein, 2003; Shimogori et al., 2010; Diez-Roux et al., 2011; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012a). Recently, in our group we applied similar developmental gene expression criteria to the identification of hypothalamic components in amphibians and reptiles (Moreno et al., 2012; Domínguez et al., 2013, 2014). We selected representative species of these vertebrate classes for their importance in evo-devo studies with a phylogenetic perspective. Amphibians constitute the only group of tetrapod anamniotes and represent a key model in anamniote/amniote

transition, as they share features with other tetrapods (amniotes) and also with other anamniotes. In turn, reptiles occupy a crucial position, especially turtles, which were reported to be the most closely related to the extinct therapsids from which mammals arose (Northcutt, 1970), although, alternatively, they have been considered the sister group to crocodiles and birds (Zardoya and Meyer, 2001a,b). Therefore, the study of these vertebrate groups appears particularly relevant since the colonization of land by tetrapod ancestors is presumably one of the evolutionary events that could involve more neural changes.

Within the current anatomical context, we now define the hypothalamic boundaries with its neighboring forebrain areas on the basis of distinct molecular profiles during development. Thus, gene expression data have highlighted that the preoptic area does not belong to the hypothalamus but it is part of the subpallial telencephalic territory (Flames et al., 2007; Medina, 2008; Garcia-Lopez et al., 2009; Sánchez-Arrones et al., 2009; Zhao et al., 2009; Roth et al., 2010) and is topologically adjacent to the dorsal hypothalamic territory. The caudal hypothalamo-diencephalic boundary is highlighted by the distinct Six3, Lhx9, Arx and Dlx expression in the prethalamic territory (P3), as well as the Otx2 expression in the diencephalon, but not in the hypothalamus (Puelles et al., 2012a,b).

The longitudinal domains of the alar and basal plates, which extend along the neuraxis, also extend to the hypothalamus and the alar–basal boundary is considered to end rostrally just behind the optic chiasm in all vertebrates (Puelles, 1995). The expression of the gene Nkx2.2 along the alar–basal boundary in the caudal prosencephalon continues rostrally in the hypothalamus, which allows distinguishing between alar and basal territories (Shimamura et al., 1995). The recently updated prosomeric model in mammals (see **Figure 1**) holds that the hypothalamus is subdivided dorsoventrally into alar, basal, and floor longitudinal domains and separates rostrocaudally, by the intrahypothalamic boundary (IHB), into two transverse regions called terminal hypothalamus (THy; rostral; hp2: hypothalamic prosomeric domain 2) and peduncular hypothalamus (PHy; caudal, hp1:hypothalamic prosomeric domain 1). The main forebrain bundles course dorsoventrally along PHy, which is also characterized by the generation of highly characteristic structures such as the main paraventricular nucleus, the retromammillary area and the migrated subthalamic nucleus (Puelles et al., 2012a). The THy contains the main tuberal and mammillary regions, as well as the supraoptic, suprachiasmatic, and retrochiasmatic areas. The THy includes a rostromedian subdomain recently named acroterminal area, with specializations such as the lamina terminalis (and related vascular organ), suprachiasmatic, and chiasmatic alar areas, and the anterobasal, arcuate, median eminence, and infundibular/neurohypophysial basal areas (Puelles et al., 2012a). During development, Six6 and Foxb1 gene expression apparently delineates the entire acroterminal territory. Although the structures included in the acroterminal part are obviously present in reptiles and amphibians (see ten Donkelaar, 1998a,b), developmental studies did not reveal specific markers for the

region (terminal); TuD, dorsal tuberal domain; Tul, lateral tuberal domain; TuV, ventral tuberal domain; Tub, tuberal region.

origin of this hypothalamic part (Moreno et al., 2012; Domínguez et al., 2013, 2014).

In the alar hypothalamus, the dorsal domain is adjacent to the telencephalic preoptic area and expresses the transcrition factors Tbr1, Sim1 and Otp (Medina, 2008; Puelles et al., 2012a; Morales-Delgado et al., 2014). This subdomain is subdivided into terminal (and acroterminal) and peduncular areas, and among others, produces the paraventricular, supraoptic and periventricular nuclei. The second alar subdomain is located ventrally and is mainly characterized by the expression of Dlx genes. It is also subvivided into terminal (and acroterminal), and peduncular areas, which give rise to the suprachiasmatic nucleus, the anterior hypothalamic nucleus and the subparaventricular zone (Puelles et al., 2012a). The basal hypothalamus classically includes the tuberal (Tub) and mammillary regions (M), both mainly characterized by the expression of the morphogen Shh and the transcription factor Nkx2.1, with minor exceptions (Puelles et al., 2012a). The tuberal region (Tub) is divided into a proper tuberal area (terminal and acroterminal) and a retrotuberal area (peduncular), which produce the anterobasal and posterobasal nuclei in the dorsal portion, the ventromedial, dorsomedial and the arcuate nuclei in the intermediate portion and ventral portions. In the mammillary region, in addition to the proper mammillary and retromamillary regions (in the terminal and peduncular segmentes, respectively) a topologically dorsal band has been defined (called with the prefix peri-) that also included terminal and peduncular portions (see **Figure 1**).

### **HYPOTHALAMIC ORGANIZATION IN THE ANAMNIO-AMNIOTIC TRANSITION: EVOLUTIONARY TRAITS ON HYPOTHALAMIC REGIONALIZATION**

The achievement of new tools in developmental neuroanatomy for the analysis of the genoarchitecture of particular brain regions has led to the precise interpretation of the hypothalamic regionalization, and the definition of different hypothalamic progenitor domains, which were traditionally linked to anatomical landmarks that not always coincided with the molecular boundaries. Thus, the analysis of the patterns of distribution of main regulatory transcription factors and proteins involved in neural patterning, and that are also expressed after development, have allowed to determine the extent of the different hypothalamic histogenetic divisions. In addition, the molecular boundaries with the adjacent non-hypothalamic territories could be assessed. Comparative studies using the same sets of markers in different vertebrates, particularly amphibians and reptiles, have highlighted that the main molecular features of the subdivisions topologically identified in the hypothalamus have been highly conserved (Medina, 2008; Domínguez et al., 2010, 2011, 2013, 2014; Moreno and González, 2011; Morales-Delgado et al., 2011, 2014; Moreno et al., 2012). In the following sections the molecular characteristics of each of the main hypothalamic regions, as well as the boundaries with the neighboring areas are detailed for *Xenopus* and *Pseudemys* and will be compared with the situation found in other vertebrates (see **Table 1**).

#### **PREOPTOHYPOTHALAMIC BOUNDARY (POH)**

The preoptic region (PO) was traditionally included within the hypothalamus until genoarchitectonic studies revealed that this region is derived from the FoxG1-positive telencephalic neuroephitelium (Tao and Lai, 1992; Murphy et al., 1994; Bourguignon et al., 1998; Zhao et al., 2009; Roth et al., 2010), revealing its subpallial nature (Flames et al., 2007; Medina, 2008; Garcia-Lopez et al., 2009; Moreno and González, 2011). In *Xenopus*, the PO is a Dll4, Isl1, Shh and Nkx2.1 positive territory that limits ventrally with the Otp expressing supraoptoparaventricular region (SPV, the dorsal part of the alar hypothalamus; Domínguez et al., 2013). This gene expression profile of the subpallial PO seems to be largely shared by reptiles (*Pseudemys*: Moreno et al., 2010) and the rest of amniotes (Puelles et al., 2000). However, some differences can be noted regarding its ventral boundary with the hypothalamus. In *Xenopus*, the Nkx2.1/Shh positive PO is in contact with the Otp/Nkx2.2 positive SPV (**Figures 2A–D**; Domínguez et al., 2013). In contrast, a narrow Nkx2.2 positive territory has been observed in the turtle (and not in *Xenopus*) between the Isl1/Nkx2.1 positive PO and the SPV-Otp expressing region (**Figures 2E–H**; Moreno et al., 2012). In this context, recent studies have described a Dlx/Nkx2.2 expressing band in mammals and birds that represents the boundary between the PO and the hypothalamus (Bardet et al., 2006, 2010; Flames et al., 2007), like in the turtle but in contrast to the situation found in amphibians and fishes (Domínguez et al., 2011, 2013; Moreno et al., 2012). Thus, the presence of this Nkx2.2 positive territory (preopto-hypothalamic boundary) supposes a relevant acquisition during the anamnio-amniotic transition.


(Continued)


Note that we only have indicated negative expressions to remark a difference between groups of vertebrates that is not due to the lack of data in the literature (empty squares).

**Table**







**1 | Continued**

#### **SUPRAOPTOPARAVENTRICULAR REGION (SPV)**

The SPV is the most dorsal region in the alar hypothalamus and it is defined by the expression of Otp/Sim1 and the lack of Dlx/Shh/Nkx2.1 expression in all vertebrates analyzed (reviewed in Markakis, 2002; Medina, 2008; Moreno and González, 2011; Puelles et al., 2012a), from anamniotes (Brox et al., 2003; Del Giacco et al., 2006; Blechman et al., 2007; Bardet et al., 2008; Domínguez et al., 2010, 2013; Machluf et al., 2011; Martínez-dela-Torre et al., 2011; Herget et al., 2014) to amniotes (Acampora et al., 1999; Flames et al., 2007; Bardet et al., 2008; Morales-Delgado et al., 2011, 2014).

Both in the amphibian *Xenopus laevis* and in the turtle *Pseudemys scripta* the extent of the SPV is particularly well defined by the expression of the transcription factor Otp, and its dorsal limit with the PO is defined by the lack of Isl1 expression (**Figures 3A,G,I,N**; Bardet et al., 2008; Moreno et al., 2012; Domínguez et al., 2013). The boundaries of the Otp positive SPV with the adjacent prethalamic and SC territories are also discernible by the lack of Isl1 in the SPV (**Figures 3G,H,H',N,O,O'**; Moreno et al., 2012; Domínguez et al., 2013). The caudal boundary with the Tbr1-expressing prethalamic eminence (EPTh in P3) is also extremely conserved in the anamnio-amniotic transition (**Figures 3G,N**; Moreno et al., 2012; Domínguez et al., 2013). Thus, the lack of Dlx, Shh, Isl1, and Nkx2.1 in the SPV of all the vertebrates analyzed is a constant feature in tetrapods that allows the distinction of the SPV territory from the adjacent diencephalic prethalamus (PTh) and SC area (Flames et al., 2007; Medina, 2008; van den Akker et al., 2008; Domínguez et al., 2010, 2013; Moreno and González, 2011; Moreno et al., 2012; Puelles et al., 2012a; Morales-Delgado et al., 2014).

Interestingly, in some groups of fishes such as teleosts and lungfishes (the closest living relatives of tetrapods) the expression of Otp in the SPV territory has been reported (Del Giacco et al., 2006; Blechman et al., 2007; Machluf et al., 2011; Moreno and González, 2011). Moreover, a very recent study of the molecular and neurochemical features of this hypothalamic region in zebrafish identified a territory homologous to the mammalian paraventricular nucleus called neurosecretory preoptic-hypothalamic area (NPO), which is characterized by the expression of Otp and the lack of Isl1, Dlx5 and Arx, which define its anatomical boundaries (Herget et al., 2014).

Furthermore, both in *Xenopus* and *Pseudemys* the expression of the transcription factor Nkx2.2, allowed the rostro-caudal subdivision of the SPV into two different progenitor domains (Domínguez et al., 2011, 2013; Moreno et al., 2012). The rostral domain (SPVr) is characterized by the ventricular and subventricular expression of Otp and Nkx2.2, whereas the caudal domain (SPVc) only expresses Otp (**Figures 3B,C,G,L,N**; Moreno et al., 2012; Domínguez et al., 2013). In the case of *Xenopus*, Lhx5 has also been distinctly observed in the rostral domain of the SPV (**Figure 3D**; Domínguez et al., 2013), in agreement with descriptions of the SPV of mouse and chick (Bulfone et al., 1993; Abellán et al., 2010). Actually, attending to its internal molecular organization, this region in mammals and birds was divided rostrocaudally into terminal and peduncular portions

(Bardet et al., 2008; Puelles et al., 2012a). In addition, regarding to the localization of the Nkx2.2, positive cells, they were reported partially overlapping the Sim1 expression domain in the anterior hypothalamus of the mouse, where the Nkx2.2 cell population seems to occupy a rostral position (see Figures 5K, 7D,E in Caqueret et al., 2006). And a population of Nkx2.2, expressing cells located in the ventral part of the paraventricular area in mouse has been described (Puelles et al., 2012a). In mammals, it was described that Nkx2.2 cells from the dorsal portion of the peduncular tuberal hypothalamus migrated very early colonizing this paraventricular nucleus, overlapping the Otp expressing cells (Puelles et al., 2012a).

The SPV of mammals and birds is also characterized by the expression of Pax6 and Tbr1 (Michaud et al., 1998, 2000; Puelles and Rubenstein, 2003; Flames et al., 2007; Medina, 2008). In the turtle, Pax6 expression was demonstrated in the ventricular zone of the SPV region (**Figures 3J,K,N**; Moreno et al., 2012), whereas it was not observed in *Xenopus* (Moreno et al., 2008a, 2012; Domínguez et al., 2013).The lack of Pax6 expression in the SPV has also been reported in other anamniotes (Murakami et al., 2001; Moreno and González, 2011). This situation might reflect differences in the specification of this area between amniotes and anamniotes. Therefore, the expression of Pax6 seems to have appeared for the first time in amniotes, and it could be related to the specific size and functionality of this area, being Pax6 involved in the dorsoventral brain organization, as it has been demonstrated in the mammalian forebrain (Toresson et al., 2000).

Of interest, some studies have recently described in amniotes a contingent of Otp positive cells generated in the SPV that migrate into the medial amygdala (Bardet et al., 2008; Abellán et al., 2010; García-Moreno et al., 2010; Bupesh et al., 2011a,b; Medina et al., 2011). This most likely represents a conserved feature that arose early in phylogeny since in lungfishes, anurans, and reptiles Otp expressing cells have been consistently observed in the region identified as the medial amygdala (González and Northcutt, 2009; Moreno et al., 2010; Domínguez et al., 2013), and a similar migratory pathway from the SPV was suggested (Moreno and González, 2011).

Regarding its neurochemical profile, the SPV of amphibians and reptiles contains different groups of cells secreting several neuropeptides such as vasotocine, mesotocine CRH, and TRH (Smeets et al., 1990; Propper et al., 1992; D'Aniello et al., 1999; Domínguez et al., 2008; López et al., 2008) that constitute part of the neuroendocrine hypothalamic system, which seems to be very conserved during the anamnioamniotic transition. The TRH positive population in *Xenopus* and turtle, is specifically located within the Otp expressing territory (Domínguez et al., 2008; López et al., 2008), suggesting that this transcription factor might be involved in the specification of the TRH phenotype during the anamnio-amniotic transition, as describing in other amniotes (Goshu et al., 2004; Del Giacco et al., 2008; Morales-Delgado et al., 2014). Moreover, in *Xenopus* a correlation between the emergence of somatostatin and mesotocine positive neurons and the presence of Otp was observed in the SPV **(Figures 3E,F**; Domínguez et al., 2013), highlighting the role of this transcription factor in the specification of these postmitotic cell populations and, consequently, in the differentiation of independent nuclei within this territory. In mice it was shown that Otp is involved in the specification of the somatostatin expressing cell populations (Morales-Delgado et al., 2011) and, therefore, the specification of the somatostatin phenotype in the SPV seems to be conserved during the vertebrtae evolution. Of note, in fishes Otp might also be involved in the specification of multiple neurosecretory hypothalamic cell populations such as those containing somatostatin, vasoticin-neurophysin and isotocinneurophysin (the latter two are homologous of mammalian vasopressin and oxytocin, respectively), as suggested by the overlapping of these neurosecretory populations in the Otp expressing domain (Blechman et al., 2007; Eaton and Glasgow, 2007; Tessmar-Raible et al., 2007; Eaton et al., 2008; Herget et al., 2014). Finally, a population of dopaminergic cells was described in the paraventricular nucleus of the SPV in reptiles (**Figure 3M**; Smeets et al., 1987), whereas catecholaminergic cells (TH positive) were not found in the SPV of *Xenopus* (Domínguez et al., 2013).

#### **SUPRACHIASMATIC REGION (SC)**

The SC constitutes the ventral part of the alar hypothalamus and contains important neuroendocrine cell groups. This region is characterized by the expression of Dlx/Arx genes in all vertebrates analyzed (Bachy et al., 2002; Brox et al., 2003; Puelles and Rubenstein, 2003; Flames et al., 2007; Bardet et al., 2008, 2010; Medina, 2008; Domínguez et al., 2011, 2013; Martínez-de-la-Torre et al., 2011). The SC abuts ventrally the basal hypothalamus, characterized by the expression of Shh/Nkx2.1 genes, defining the alar–basal boundary, according to the prosomeric model (Puelles et al., 2012a). Both in amphibians and reptiles, this ventral boundary is highlighted by the Otp expression in the most rostral part of the Tub (**Figures 4G,H,L,M**; Moreno et al., 2012; Domínguez et al., 2013). Caudally, in *Xenopus*, the SC region is adjacent to the Dll4 positive PTh in the diencephalon, whereas in the juvenile turtle the SPV Otp-expressing cells extends reaching the alar–basal boundary (**Figures 4G,L**; Moreno et al., 2012; Domínguez et al., 2013).

This region in mammals, identified as subparaventrcular area (Puelles et al., 2012a), is defined as a Dlx+/Nkx2.1-territory (Puelles and Rubenstein, 2003; Medina, 2008; Puelles et al., 2012a). Furthermore, a recent study in mice has defined a territory called intrahypothalamic diagonal band, based primarily on the differential expression of Lhx1, Lhx6, Lhx7 and Lhx8, constituting the territory from which the suprachiasmatic populations of interneurons would arise (Shimogori et al., 2010). The domain recently defined in mammals as the liminal subparaventricular subdomain distinctively contains alar mantle cells expressing Nkx2.1, not present at the supraliminal subdomain (Shimogori et al., 2010; Puelles et al., 2012a). It could be comparable to those results showed in our models, but in our case the Nkx2.1 expression is also detected in the ventricular cells.

The acroterminal domain in front of the terminal region was proposed to be the source of the proper suprachiacmaitc nucleus (Puelles et al., 2012a). Both in *Xenopus* and *Pseudemys*, no distinct labeling paterns could help in the deliniation of the acroterminal part of the alar hypothalamus but the specialized derivatives of this region have been classically described, such as the optic chiasm, postchiasmatic commissures and suprachiasmatic nuclei (for review, see ten Donkelaar, 1998a,b). Therefore, it is likely that the origin in the terminal region would be conserved, but further studies are needed in order to identify genoarchitectonically this territory.

In birds, Nkx2.1 is also expressed in the subparaventricular nucleus, that belongs to the suprachiasmatic domain, which also expresses Nkx2.2, Lhx6/7 and Lhx8 (Abellán and Medina, 2009; Bardet et al., 2010). This region of *Xenopus* and turtle is Isl1-positive (**Figures 4A,G,I,L**; Moreno et al., 2008b, 2012; Domínguez et al., 2013) and, in the case of *Xenopus* in which the expression of several Dlx genes has been analyzed, the Isl1 and Dlx expression domains overlap in almost all prosencephalic regions including the entire SC territory (Brox et al., 2003; Domínguez et al., 2010, 2013). The transcription factors Nkx2.1 and Nkx2.2, are also expressed in the SC (**Figures 4B,G,J,L**; Domínguez et al., 2010, 2011) and the combination of both markers allowed the identification of rostro-caudal subdivisions in *Xenopus* and *Pseudemys*. Thus, only in the rostral part (SCr) Nkx2.1 and Nkx2.2 are found in the ventricular zone, in contrast to the caudal portion (SCc) that is devoid of expression (**Figures 4B,J**;

Moreno et al., 2012; Domínguez et al., 2013). Of note, in *Xenopus*, the expression pattern of the morphogen Shh in SC runs parallel to the Nkx2.1 expressing domain (**Figures 4C,G**), suggesting a regulatory role of Shh through Nkx2.1 actions also in amphibians (Domínguez et al., 2013). Also in *Xenopus*, the expression of Lhx1 and Lhx7 is restricted to the rostral SC domain (**Figures 4D,E,G**; Moreno et al., 2004; Domínguez et al., 2013), which would be comparable to the subparaventricular nucleus described in chicken (Abellán and Medina, 2009; Bardet et al., 2010). In the case of fishes, expression of Dlx and Lhx7 has been detected in comparable regions to the SC territory in *Medaka* (Alunni et al., 2004). In addition, expression of Dlx genes in the SC primordium has been observed in the lamprey (Martínez-de-la-Torre et al., 2011), and Isl1 and Nkx2.1 expressions have been detected in the SC region of lungfishes (Moreno and González, 2011). However, the region expressing Shh/Nkx2.1 in fishes extends to the entire SC territory and subdivisions were not observed (Rohr et al., 2001).

Regarding the Nkx2.1 expression in the alar hypothalamus, it appears that is gradually restricted especially in the SC, from amphibians through amniotes (**Figure 7**). In mammals, the Lhx6 + intrahypothalamic diagonal band, proposed by Shimogori et al. (2010) corresponds to the Nkx2.1 expressing band of Puelles in the subparaventricular region of the alar domain (see Figure 8.9D in Puelles et al., 2012a). Also in mammals, like in Xenopus and turtle, this small band matches the Nkx2.2 expression (see Figure 8.9F in Puelles et al., 2012a). It could be related to the liminal subparaventricular area proposed by Puelles et al. (2012a). Thus, Nkx2.1 and Shh are expressed in almost the SC territory in non-tetrapod anamniotes like the zebrafish (Rohr et al., 2001), whereas in non-mammalian tetrapods like the anamniote *Xenopus* and the amniote reptiles and birds the Nkx2.1 expression is restricted to just a SC subdomain (Medina, 2008; van den Akker et al., 2008; Abellán and Medina, 2009; Moreno et al., 2012; Domínguez et al., 2013). Consequently, the evolutionary tendency of the disappearance of both developmental regulators in the SC begins at the origin of the tetrapod phylogeny. Furthermore, this progressive disappearance of Shh/Nkx2.1 expression in the alar hypothalamus/SC region has been related to the pallial expansion that takes place in amniotes (Bruce and Neary, 1995; Striedter, 1997), and the reduction of the alar hypothalamus in amniotes in contrast to anamniotes at the expense of the thalamic expansion (van den Akker et al., 2008; for review, see Medina, 2008).

In functional terms, the SC region is known to belong to the neuroendocrine system and therefore consists of multiple neuropeptide-secreting cell populations. In mammals and birds, the SC is characterized by the presence of TRH positive cells, among others, that have also been reported in anamniotes such as anurans and fishes (Domínguez et al., 2008), but that were not detected in reptiles (López et al., 2008). However, the SC region in amphibians and reptiles is characterized by the presence of catecholaminergic cell groups (González and Smeets, 1991; González et al., 1993) restricted to the rostral domain (**Figures 4F,G,K,L**; Morona and González, 2008; Moreno et al., 2012; Domínguez et al., 2013). This feature appears to be conserved throughout vertebrate evolution, being observed in amniotes and anamniotes (Hökfelt et al., 1984; González et al., 1993; Smeets and González, 2000; Moreno et al., 2012; Domínguez et al., 2013). In addition, the GABAergic expression has been analyzed in the SC of *Xenopus* showing that it is widely distributed along the entire Dlx expressing zone (Domínguez et al., 2013), suggesting that Dlx could be involved in the GABAergic specification, like in mammals (Price et al., 1991; Bulfone et al., 1993; Marín and Rubenstein, 2001).

#### **TUBERAL REGION (TuB)**

This region is currently considered to extend in the dorsal part of the basal hypothalamus, and like the rest of the hypothalamus has been postulated that posses acroterminal, terminal, and peduncular portions (**Figure 1**; Puelles et al., 2012a). The Tub is primarily characterized by the expression of Shh, which is directly involved in the organization of the basal hypothalamus through Nkx2.1 action (Kimura et al., 1996; Puelles et al., 2004), and thus Shh/Nkx2.1 expression has been observed in all the vertebrates analyzed (reviewed in Medina, 2008; Moreno and González, 2011). In amphibians and reptiles this territory is defined by the expression of Nkx2.1 and Isl1 (**Figures 5A,B,F,H–J,M**; Moreno et al., 2008b, 2012; Domínguez et al., 2014). Moreover, in *Xenopus* the Tub is also characterized by the ventricular expression of Shh, along with Nkx2.1 (**Figures 5B,F**; Domínguez et al., 2010, 2014), suggesting that also in *Xenopus* Shh could be implicated in the hypothalamic organization by the action of Nkx2.1, as in amniotes (Kimura et al., 1996; Puelles et al., 2004; reviewed in Medina, 2008). Consistently, both in *Xenopus* and *Pseudemys* the transcription factor Otp is exclusively located in the rostral tuberal portion (RT), within the Isl1-positive basal territory (**Figures 5C,F,K–M**).

However, in mammals the dorsal portion of the rostral terminal Tub is the only Otp expressing zone (Morales-Delgado et al., 2014), and the transcription factor Isl1 is expressed in the ventromedial and arcuate nuclei, where it is involved in the hypothalamic development and the regulation of the reproductive behavior (Davis et al., 2004). Moreover, the intermediate terminal and peduncular Tubs of mammals are also characterized by the expression of Dlx genes, involved in the specification of the GABAergic cell fate (Yee et al., 2009). Importantly, the transcription factor Otp is expressed in the dorsal Tub of mouse and in clusters of cells that from the acroterminal domain give rise to part of the nucleus arcuatus (Puelles et al., 2012a). Thus, the particular expression of Otp in a subdomain of the Tub appears as a conserved feature of the basal hypothalamus in amniotes and anamniotes equivalent to the region that gives rise to part of the arcuate nucleus in amniotes (Bardet et al., 2008; Puelles et al., 2012a).

Distinctly, the caudal tuberal part (CT) of *Xenopus* and *Pseudemys* is characterized by the lack of Otp expression (**Figures 5C,F,K–M**) and, expression of Nkx2.2 (**Figures 5C–F**), a transcription factor typically located in basal territories and necessary to maintain the ventral phenotype (Briscoe and Ericson, 1999; Sander et al., 2000; Garcia-Lopez et al., 2004; Puelles et al., 2004). Moreover, in mammals, Nkx2.2 in this Tub has been detected in the ventromedial nucleus and the core portion of the dorsomedial peduncular hypothalamic nucleus, in the terminal and peduncular domains respectively (Puelles et al., 2012a), where it is involved in the specification of the basal phenotype and the specification of the ventromedial fate (Kurrasch et al., 2007). Thus, it is possible that the Otp/Nkx2.1/Nkx2.2 expression in Xenopus likely resembles the mouse situation in which Otp is expressed in the dorsal domain of the terminal tuberal hypothalamus (likely including the acroterminal domain; Puelles et al., 2012a; Morales-Delgado et al., 2014), whereas Nkx2.1 is found in the terminal and peduncular intermediate portion (TuI), which gives rise to the ventromedial and dorsomedial nuclei (Puelles et al., 2012a), and likely corresponds to the region with Nkx2.2, Lhx1 and Dll4 expression found in our models, thus resembling the subdivision proposed (**Figures 5D,F**; Domínguez et al., 2014).

The boundary between the tuberal and mammillary territories in the basal hypothalamus of amphibians and reptiles is mainly defined by the lack of Isl1 expression in the mammillary region within the continuous Nkx2.1 positive tuberomammillar region (**Figures 5F,G,M,N**; Moreno et al., 2012; Domínguez et al., 2014). In addition, both regions can be distinguished by the differential expression of the transcription factor Otp, which is expressed in the mammillary region and not in the caudalmost tuberal domain (**Figures 5C,F,L,M**; Moreno et al., 2012; Domínguez et al., 2014). In *Pseudemys* a thin Nkx2.1 positive band (Moreno et al., 2012)

can be detected between the Isl1 positive caudal tuberal zone and the Otp expressing mammillary region, defining specifically the tuberomammillary boundary (Moreno et al., 2012). In mammals, a thin ventral band expressing Dlx5/Nkx2.1/Arx+ but without Otp expression has been recently described in the ventral tuberal domain (Puelles et al., 2012a; Morales-Delgado et al., 2014). This is associated with a longitudinal circumventricular organ and it is the source of histamine in the hypothalamic cells (Puelles et al., 2012a).

Compared to tetrapods, there are only a few data about the hypothalamic organization in fishes, mainly attending to expression patterns and development. Recent studies in lunghfishes have revealed that Nkx2.1 and Isl1 are expressed in the entire Tub, whereas Otp expression is restricted to the most rostral and dorsal part, sustaining similar subdivisions in the tuberal territory to the ones described in tetrapods, using the same markers (Moreno and González, 2011). Data obtained in agnathans (lampreys) demonstrated that a substantial number of Dlx expressing cells occur in the tuberal hypothalamic nucleus and the tuberomammillary region, similar to anurans (Martínezde-la-Torre et al., 2011).

The chemoarchitecture and neuronal specification processes in the Tub seem to be largely conserved throughout vertebrate evolution. In *Xenopus*, in the Otp-positive rostral Tub a population of somatostatin expressing cells has been observed (**Figures 5E,F**; Domínguez et al., 2014) suggesting the implication of Otp in the specification of this neurons in anurans, as has been previously reported in mammals (Acampora et al., 1999; Wang and Lufkin, 2000). In the mouse, the anterobasal nucleus, in the acroterminal domain defined by Puelles et al. (2012a), has been described to be the source of somatostatin cells to the ventromedial and arcuatus nucleus, where Otp would be specifically involved in the specification of these neurons (Morales-Delgado et al., 2011). The neuropeptide TRH, has been traditionally located in the dorsomedial nucleus (Hökfelt et al., 1975; Lechan et al., 1986; Tsuruo et al., 1987; Merchenthaler et al., 1988) and in the lateral hypothalamic area of mammals, where a recent study has proved its role in the arousal generation (Horjales-Araujo et al., 2014). However, a recent study has revealed that these tuberal TRH positive populations are likely generated in the SPV alar region (Morales-Delgado et al., 2014). The neuromodulator TRH has also been found in the periventricular hypothalamic nuclei of reptiles and in the Tub of anurans (Domínguez et al., 2008; López et al., 2008).

Finally, the anatomical position of the Dll4 expressing cell group located in the most caudal tuberal part of *Xenopus* is closely related to the GABAergic positive population (unpublished data), suggesting an implication of Dll4 in the specification of the GABAergic phenotype in the Tub, as occurs in the majority of the histogenetic domains where both markers colocalize (Price et al., 1991; Bulfone et al., 1993; Marín and Rubenstein, 2001). A distinct feature of the neurochemical profile of the Tub in birds is the presence of catecholaminergic populations originated under the control of Shh. Actually, in the chicken a dopaminergic positive population located in the Tub exists that is specified by Shh in a Six3-dependent manner (Ohyama et al., 2005). However, in anurans and reptiles there are not catecholaminergic cells in the Tub (Moreno et al., 2012; Domínguez et al., 2014).

#### **MAMMILLARY REGION (M)**

In the current prosomeric model, the mammillary region is interpreted as formed by mammillary-terminal and retromammillary-peduncular regions (see **Figure 1**; Puelles et al., 2012a). In addition, immediately dorsal to them, corresponding peri-mammillary and peri-retromammillary regions (RMas) were considered. The latter form in mouse a rostral (peri-) band, which has been described based on the Otp/Nkx2.1 expression and the lack of Dlx genes (Puelles et al., 2012a).

In recent years, the regionalization of the this area has been under analysis and the terminology used for its various subdivisions and their actual extent in the basal hypothalamic region have progressively varied with the appearance of the molecular approach (Shimogori et al., 2010; Puelles et al., 2012a). By means of the combinatorial expression of Shh and Nkx2.1, in *Xenopus* two different regions were identified, the mammillary area proper (Ma) that is Otp/Nkx2.1+/Shh-, and the RMa, where the expression patterns of Shh and Nkx2.1 are inverted, being Nkx2.1-/Shh+ (**Figures 6A,G**; Domínguez et al., 2014). In contrast to *Xenopus*, in the turtle the Nkx2.1 expression is continuous throughout the mammillary band, abutting directly the Pax7 + p3b (**Figures 6J,P**; Moreno et al., 2012). However, in the turtle, but not in *Xenopus* (**Figures 6B,C**), within the Nkx2.1 expressing region (**Figure 6O**) a portion that is Isl1-/Otp- (**Figure 6K**, asterisk in K') can be distinguished between the Isl1 + Tub and the Otp + and Nkx2.2 + Ma region (**Figures 6K,L**). It has been discussed (see above) that this region could correspond to the ventral tuberal portion proposed in mammals (Puelles et al., 2012a). In addition, the Ma of both reptiles and amphibians shows scattered Pax7 + cells in the subventricular zone late in development that likely originate in the Pax7 expressing population of the adjacent basal plate of P3 (**Figures 6D,G–I,O,Q,Q'**; Moreno et al., 2012; Domínguez et al., 2014), coincident with the Otp expression observed in the mammillary region (defined by the lack of Isl1 and the expression of Nkx2.1 and Otp; Moreno et al., 2012; Domínguez et al., 2014). This suggests that the longitudinally organization proposed in mammals, and specifically the rostral mammillary band, could also be present in Xenopus and turtle. In this line, in our models Nkx2.2 cells have been observed in the Ma Otp + zone (**Figure 5L**) and similarly Dlx and GABA expressing cells (Domínguez et al., 2014), likely from the adjacent p3b. In this context, it could also be possible that in mammals some of the Nkx2.2 expressing cells along the alar–basal boundary could reach the periretromammilar region, where Otp is observed in contrast to the perimammillar portion where only Otp is found (see Figure 8.26D in Puelles et al., 2012a).

The mammillary area is also characterized by the expression of genes of the LIM-HD family, whose combinatorial expression pattern led to propose a new regionalization of this territory in mammals (Shimogori et al., 2010). Thus, distinct nuclei were proposed including a supramammillary nucleus Irx5+, a premammillary nucleus expressing Lhx9/Lef1, a mammillary nucleus expressing Lhx1, and a tuberomammillary terminal zone positive for Lhx6, which is continuous with the diagonal band that separates the alar and basal hypothalamic regions (Shimogori et al., 2010). Comparatively, in *Xenopus* Lhx7 is expressed in the alar hypothalamus and continues ventrally into the mammillary territory (Moreno et al., 2004; Domínguez et al., 2013), and this led to propose the existence of a comparable tuberomammillary terminal zone in anurans (Domínguez et al., 2014). The expression of Lhx1 in the mammillary portion of the Xenopus hypothalamus was defined based on the lack of Isl1 in this portion and the Otp expression (**Figure 5D**; Domínguez et al., 2014). In addition, Lhx1 is also expressed in the mammillary portion of mouse hypothalamus (Bachy et al., 2002; Shimogori et al., 2010). Of note, in their hypothalamic analysis the group of Shimogori et al. (2010) used the Lhx1 expression to define the MM region (mammillary). However, in the current prosomeric interpretation (Puelles et al., 2012a) the expression of Lhx6 is interpreted in the ventral portion of the tuberal hypothalamus (see Figure 8.9 in Puelles et al., 2012a). Independently of its exact anatomical localization, its position in our models suggests, along with the Nkx2.2 and Pax7 expressing cells described before (**Figures 5H,L,Q**), a comparable longitudinal hypothalamic band that could be comparable to the rostral perimammillar and periretromammilar band (Puelles et al., 2012a).

Several studies in fishes have also reported the differential expression of LIM genes, such as Lhx6 and Lhx1/5, in the

basal hypothalamic territory (Osorio et al., 2005; Menuet et al., 2007). Moreover, a recent study has analyzed the mammillary organization proposed by Shimogori et al. (2010) in zebrafish, finding different domains based on the differential expression of Lef1, Lhx6, Irx5 and Foxb1 (Wolf and Ryu, 2013). Thus, the presence of Otp in the Nkx2.1 positive region of the mammillary band was reported to be playing a crucial role in the establishment of different mammillary domains and is involved in the specification of posterior hypothalamic neurons regulating the expression of Fezf2 and Foxb1.2 in the putative mammillary region (Wolf and Ryu, 2013).

The lack of Shh expression in RMa has also been reported in the chicken, where the Shh becomes downregulated in the tuberomammillary primordium, but not in the RMa, at a specific point during development (Martí et al., 1995; Shimamura et al., 1995; Crossley et al., 2001; Patten et al., 2003; Manning et al., 2006), what seems to confer the hypothalamic fate to these cells (Manning et al., 2006). In the mammillary territory of mammals and birds, two rostro-caudal portions have been described on the basis of the differential expression of Nkx2.1 and Shh. Thus, there is a Nkx2.1+/Shh-region that is also positive for Otp (Bardet et al., 2008; Morales-Delgado et al., 2011), and a Nkx2.1-/Shh+ RMa region (García-Calero et al., 2008;

#### **FIGURE 7 | Continued**

non mammalian amniotes and in anamniotes; Otp is expressed in the mammillary region of all vertebrates analyzed (no data in the lamprey are available). However, most differences in the scheme are due to the absence of data in the literature. The numbers 1–4 in the scheme represent the main evolutionary events regarding to the hypothalamic organization, as follows: (1) Nkx2.1 expression restriction in SC. (2) POH Nkx2.2 expression. (3) Pax6 expression in SPV for the first time. (4) Pallial and thalamic expansion at the expense of the alar hypothalamic reduction. Note that the developmental stages used in the scheme are not equivalent for all species.

Morales-Delgado et al., 2011). Attending to the Shh/Nkx2.1 expression pattern in the mammillary territory, it represents an exception within the prosencephalon, being the only forebrain area where Shh and Nkx2.1 are not expressed in parallel because Shh expression becomes secondarily downregulated at some point in the development (Shimamura et al., 1995; Crossley et al., 2001; Manning et al., 2006).

Thus, regarding the situation in fishes, some studies have described the presence of Shh in the basal hypothalamus, although so far there are no data about the specific location of Shh expression within this basal hypothalamic territory. It has been reported the presence of two hedgehog genes, expressed in a Sonic Hh-like pattern, in the basal hypothalamus of lamprey (Osorio et al., 2005; Kano et al., 2010). In addition, expression of Shh has been reported in the basal hypothalamus of cavefish and zebrafish (Menuet et al., 2007; Wolf and Ryu, 2013), where its expression seems to be limited in the tuberal territory, although no specific distinction of tuberal and mammillary regions was described.

In terms of chemical specification, the amphibian and reptilian mammillary region is characterized by the presence of a rich catecholaminergic cell population (Smeets et al., 1987; Smeets and González, 2000; Moreno et al., 2012; Domínguez et al., 2014) that co-expresses Nkx2.1 (**Figures 6E,G,M,O**) and Otp (**Figures 6F,G,N,O**), in line with previous studies in other vertebrates and suggesting a conserved role of both transcription factors in the specification of the dopaminergic phenotype during the anamnio-amniotic transition (Kawano et al., 2003; Del Giacco et al., 2006; Blechman et al., 2007; Ryu et al., 2007; Löhr et al., 2009).

Regarding the nuclear specification, controversy exists regarding the origin of the different neuronal groups and several data support the contribution of diencephalic areas to the mammillary territory. In mammals, the retromammillary area was considered a caudoventral hypothalamic specification located between the diencephalic tegmentum (in P3; see for review Puelles et al., 2012a), giving rise to the subthalamic nucleus. In birds, recent fate map studies have described that the basal plate of P3 generates the retromammillary tegmentum and the subthalamic nucleus (Garcia-Lopez et al., 2009). In reptiles and anurans, Pax7 expressing cells likely originated in P3 colonize the mammillary region (Moreno et al., 2012; Bandín et al., 2013; Domínguez et al., 2014), suggesting a diencephalic contribution to the formation of the hypothalamic mammillary territory. Moreover, the subthalamic nucleus in mammals was identified by the expression of Pax7 (Stoykova and Gruss, 1994; see mouse developmental Allen Brain Atlas), thus the Pax7 positive cells found dispersed in the mammilary territory in *Xenopus* could suggest the existence of a forerunner of the subthalamic nucleus in anurans (Domínguez et al., 2014).

### **CONCLUDING REMARKS**

The organization of the brain undergoes evolutionary/adaptative changes during the anamnio-amniotic transition. The evolutionary leap from amphibians to reptiles involves relevant adaptation changes to conquer a new environment that have clear consequences on brain organization. However, it seems that during the transition from aquatic to terrestrial life the hypothalamus has maintained a major general pattern of organization, but with subtle differences that could be related to the new requirements for adaptation to the new environment. These variations in hypothalamic organization/regionalization highlighted in the present comparative genoarchitectonic analysis appear to have occurred gradually during the anamnio-amniotic transition starting with amphibians, which are the first tetrapods that arose, being anamniotes (**Table 1**; **Figure 7**). Considering the data gathered on the organization of the hypothalamus, it seems that there is a mostly common general pattern shared by all vertebrates that includes the following main features: (1) it belongs to the secondary prosencephalon and is topologically rostral to the diencephalon; and (2) it is formed by alar and basal regions that show genoarchitectonic patterns during development that are generally conserved across vertebrates, especially in the basal territories.

In the evolutionary context (**Table 1**; **Figure 7**), our results in amphibians and reptiles add information to the known features of the hypothalamic organization in birds and mammals and point out to some main features shared by all tetrapods: (1) each alar (SPV, SC) and basal (Tub, M) territory is also subdivided rostrocaudally into two different domains based on molecular criteria; (2) the expression of Nkx2.1 that characterizes the entire SC region in fishes starts to be restricted in amphibians and is gradually reduced through mammals where the SC virtually lacks expression of this transcription factor, what could be related to the gradual pallial and thalamic expansion that take place in the amniotes. In addition, there are some features in the organization of the hypothalamus that seem to have emerged with the amniotes (see **Figure 7**): (1) the existence of the preoptohypothalamic boundary observed in amniotes starts in reptiles; (2) also in reptiles, as in birds and mammals, Pax6 is expressed in the SPV, whereas such expression is not observed in anamniotes. These facts highlight the relevance of the studies involving species of amphibians and reptiles for elaborating a complete evolutionary story of the hypothalamus.

Comparative studies of the hypothalamus across vertebrates encompass many difficulties because the different degree of topographical modification of its parts, due to diverse forces during development that lead to the final different anatomy in each group (**Figures 7**, **8**). The forces involved in the hypothalamic final conformation might be of different nature. If we consider the situation in mammals, in a "non-disturbed" neural tube at the level of the prosencephalon (**Figure 8A**) the

force (1) to act is the flexure of the neural tube **(A)**. In mammals, the longitunal axis bends almost 90◦ forming a sharp flexure and the rostral tube is moved to a "ventral" position **(B)**, whereas in non-mammals this angle seems to be less pronounced **(C)**. Then, a second morphological force acts over this longitudinal axis that is already partially bent, which

evagination (3). In mammals this third strength is contributing to the elongation of the hypothalamic territory and, in the case of non-mammals this force is also contributing to pronounced hypothalamic modification.

alar hypothalamus is in the most rostral portion along with the telencephalic prospective territories, which will give rise to the telencephalic vesicles and the telencephalon impar during development. The neural tube suffers a second morphological strength given the flexure of the neural tube, which is maximum at the level of the diencephalic basal plate, thus at the boundary with the hypothalamic basal region. In addition, those rostral regions of the brain are under the direct morphological strength that produces the evagination of the telencephalic vesicles. Specially in mammals, the pallium is enormously expanded dramatically increasing in size and literally pushing the adjacent regions, like the alar hypothalamus. Therefore, in mammals due to the drastic expansion of the pallium, together with the strong flexure of the brain that bends the longitudinal axis almost 90◦ , the hypothalamus acquires a "ventral" position (**Figure 8B**). In the case of non-mammalian vertebrates, and specially in anamniotes, these developmental changes due to morphological pressures are, in general, less significant (**Figure 8C**). Telencephalic development is less massive and the cephalic flexure less pronounced, varying in the different vertebrates. However, in spite of the different topography of the hypothalamus, studies such as ours reveal that comparable subdivisions are contained in the hypothalamus of each group. Therefore, the main final conclusion of the comparative analysis of the region of the hypothalamus in vertebrates is probably the high degree of conservation of this region in evolution, as expected given its functional importance in the animal survival. Developmental forces during the ontogeny of each vertebrate group would be responsible for the different topographical arrangement of the hypothalamic regions, which otherwise are similarly specified by gene expression patterns throughout vertebrates.

### **AUTHOR CONTRIBUTIONS**

All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. This review is based on previous studies in which the three authors were involved (Moreno et al., 2012; Domínguez et al., 2013, 2014).

#### **ACKNOWLEDGMENTS**

This work was supported by the Spanish Ministry of Economy and Competitivity (MINECO, grant BFU2012-31687) and the UCM-B. Santander (grant GR3/14). **Figures 2**–**6** contain modified images already used, and included here thanks to the Journal of Comparative Neurology policy that allows free use to authors in their own publications.

#### **REFERENCES**


combinatorial expression of transcriptional codes. *J. Neurosci.* 27, 9682–9695. doi: 10.1523/jneurosci.2750-07.2007


functional subdivisions. *J. Comp. Neurol.* 472, 52–72. doi: 10.1002/cne. 20046


homeobox gene Nkx2.1/Titf-1 in forebrain evolution. *J. Comp. Neurol.* 506, 211– 223. doi: 10.1002/cne.21542


Zhao, X. F., Suh, C. S., Prat, C. R., Ellingsen, S., and Fjose, A. (2009). Distinct expression of two foxg1 paralogues in zebrafish. *Gene Expr. Patterns* 9, 266–272. doi: 10.1016/j.gep.2009.04.001

**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 01 October 2014; accepted: 09 January 2015; published online: 03 February 2015*.

*Citation: Domínguez L, González A and Moreno N (2015) Patterns of hypothalamic regionalization in amphibians and reptiles: common traits revealed by a genoarchitectonic approach. Front. Neuroanat. 9:3. doi: 10.3389/fnana.2015.00003 This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Domínguez, González and Moreno. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Prosomeric organization of the hypothalamus in an elasmobranch, the catshark *Scyliorhinus canicula*

*Gabriel N. Santos-Durán1, Arnaud Menuet2, Ronan Lagadec3, Hélène Mayeur3, Susana Ferreiro-Galve3, Sylvie Mazan3, Isabel Rodríguez-Moldes1 and Eva Candal1\**

*<sup>1</sup> Centro de Investigaciones Biológicas, Department of Cell Biology and Ecology, University of Santiago de Compostela, Santiago de Compostela, Spain, <sup>2</sup> Centre National de la Recherche Scientifique, Experimental and Molecular Immunology and Neurogenetics, University of Orleans, UMR7355, Orleans, France, <sup>3</sup> Centre National de la Recherche Scientifique, FR2424, Development and Evolution of Vertebrates Group, Sorbonne Universités – Université Pierre et Marie Curie, Roscoff, France*

#### *Edited by:*

*Luis Puelles, Universidad de Murcia, Spain*

#### *Reviewed by:*

*Manuel A. Pombal, University of Vigo, Spain Alino Martinez-Marcos, Universidad de Castilla, Spain*

#### *\*Correspondence:*

*Eva Candal, Centro de Investigaciones Biológicas, Department of Cell Biology and Ecology, University of Santiago de Compostela, Campus Vida, Avenida Lope Gómez de Marzoa, s/n, Santiago de Compostela E-15782, Spain eva.candal@usc.es*

> *Received: 01 December 2014 Paper pending published: 12 January 2015 Accepted: 09 March 2015 Published: 08 April 2015*

#### *Citation:*

*Santos-Durán GN, Menuet A, Lagadec R, Mayeur H, Ferreiro-Galve S, Mazan S, Rodríguez-Moldes I and Candal E (2015) Prosomeric organization of the hypothalamus in an elasmobranch, the catshark Scyliorhinus canicula. Front. Neuroanat. 9:37. doi: 10.3389/fnana.2015.00037* The hypothalamus has been a central topic in neuroanatomy because of its important physiological functions, but its mature organization remains elusive. Deciphering its embryonic and adult organization is crucial in an evolutionary approach of the organization of the vertebrate forebrain. Here we studied the molecular organization of the hypothalamus and neighboring telencephalic domains in a cartilaginous fish, the catshark, *Scyliorhinus canicula*, focusing on *ScFoxg1a*, *ScShh, ScNkx2.1*, *ScDlx2/5*, *ScOtp,* and *ScTbr1* expression profiles and on the identification αacetylated-tubulin-immunoreactive (ir), TH-ir, 5-HT-ir, and GFAP-ir structures by means of immunohistochemistry. Analysis of the results within the updated prosomeric model framework support the existence of alar and basal histogenetic compartments in the hypothalamus similar to those described in the mouse, suggesting the ancestrality of these subdivisions in jawed vertebrates. These data provide new insights into hypothalamic organization in cartilaginous fishes and highlight the generality of key features of the prosomeric model in jawed vertebrates.

Keywords: chondrichthyan, forebrain patterning, evolution, development, prosomeric model, *Shh*, *Nkx2.1, Otp*

### Introduction

Biological diversity emerges, at least in part, through changes in development. Organisms are different because their developmental process differ and, what is more, because their developmental process also evolve (Kutschera and Niklas, 2004; Müller, 2007; Medina et al., 2011). Thus, the

**Abbreviations:** ABB, alar-basal boundary; ac, anterior commissure; AHy, alar hypothalamus; At, acroterminal region; AP, alar plate; BHy, basal hypothalamus; BP, basal plate; CAHy, caudal part of the alar hypothalamus; CBHy, caudal part of the basal hypothalamus; D, diencephalon; F, forebrain; FP, floor plate; HDB, hypothalamo-diencephalic boundary; hp1, prosomere hp1 or peduncular hypothalamus; hp2, prosomere hp2 or terminal hypothalamus; IHB, intrahypothalamic boundary; MM, mammillary area; MTT, mammillo-tegmental tract; mz, marginal zone; os, optic stalk; p1, prosomere 1; p2, prosomere 2; p2Tg, tegmental part of prosomere 2; p3, prosomere 3; p3Tg, tegmental part of prosomere 3; P, pallium; PM, perimammillary area; POA, preoptic area; PPa, peduncular paraventricular area; PRM, periretromammillary area; PSPa, peduncular subparaventricular area; PThE, prethalamic eminence; RAHy, rostral part of the alar hypothalamus; RBHy, rostral part of the basal hypothalamus; Rh, rhombencephalon; RM, retromammillary area; rmc, retromammillary commissure; RP, roof plate; RTu, retrotuberal domain; sot, supraoptic tract; Sp, subpallium; T, telencephalon; TPa, terminal paraventricular area; TPOC, tract of the postoptic commissure; TSPa, terminal subparaventricular area; Tu, tuberal domain; vz, ventricular zone.

understanding of the development of the vertebrate brain becomes fundamental to comprehend its structure and evolution. In this context, the hypothalamus has been both a central and elusive topic. The hypothalamus is a conserved integrative center that coordinates autonomic, endocrine, and limbic responses (Sarnat and Netsky, 1981; Kandel and Schwartz, 2001; Butler and Hodos, 2005). Its development, at the base of the vertebrate forebrain (prosencephalon), involves complex patterning processes dependent on different signaling events that converge at this point. It also undergoes a complex morphological deformation during development, which misleads its topological (*vs*. topographic) location (Shimamura et al., 1995; Puelles and Rubenstein, 2003; Puelles, 2009; Puelles et al., 2012). As a result, the hypothalamic organization remains a matter of debate (Figdor and Stern, 1993; Puelles and Rubenstein, 2003; Shimogori et al., 2010; Diez-Roux et al., 2011; Puelles et al., 2012). Cross-species comparisons can be important to resolve this issue, and an important effort to understand the underlying unity of hypothalamic embryonic and adult organization across vertebrates has been made recently (Shimogori et al., 2010; Domínguez, 2011; Morales-Delgado et al., 2011, 2014; Moreno et al., 2012; Domínguez et al., 2013, 2014; Herget et al., 2014).

The prosomeric model (Puelles and Rubenstein, 2003; Puelles et al., 2004, 2012; Medina, 2008; Puelles, 2009) has become a key reference in such comparative studies, since it offers a mechanistic paradigm of the vertebrate brain structure and organization. Initially based on analyses of amniotes, this model defines for the first time anatomical structures as developmental hierarchical units based on specification mechanisms that determine longitudinal and transverse axis orientation, segmental structure, transcription factor expression profiles and the emergence of differential histogenetic domains (Puelles and Rubenstein, 2003; Puelles, 2009; Martínez et al., 2012).

A major interest and novelty of this model is that it puts emphasis on developmental criteria (including topological relationships among certain morphological landmarks, regulatory gene expression patterns and signaling molecules). Testing their conservation across vertebrates is a powerful approach for the correct establishment of homologies between embryonic territories beyond amniotes (Puelles and Medina, 2002). The underlying notion is that formation of the vertebrate brain involves a conserved core of highly constrained, invariant mechanisms and genetic networks, which are the basis for homology establishment. This in no way excludes the emergence of diversifications through evolution, which are the source of the neuroanatomic diversity observed among vertebrates.

Latest updates of the model provide novel views on the organization of the rostral-most (secondary) prosencephalon, and its telencephalic and hypothalamic moieties (Puelles and Rubenstein, 2003; Pombal et al., 2009; Puelles et al., 2012). Detailed studies in different vertebrate groups are necessary to validate the model assumptions. Cartilaginous fishes or chondrichthyans are crucial in this task because they are among the most basal extant groups of gnathostomes (jawed vertebrates). Because of its phylogenetic position as the closest outgroup to osteichthyans (the other major phylum of gnathostomes, which includes bony fish and tetrapods), chondrichthyans are essential to reconstruct gnathostome ancestral characteristics through comparisons with other vertebrate models. Here we studied the molecular histogenetic organization of the hypothalamus and directly adjoining territories of an elasmobranch representative of one of the most basal extant gnathostome lineage, the catshark *Scyliorhinus canicula*, and analyzed them under the updated prosomeric framework. We have integrated data from neuroepithelial specification codes (based on the expression of catshark orthologues of *Foxg1a*, *Shh*, *Nkx2.1*, *Dlx2/5*, *Otp,* and *Tbr1*), and from the distribution of α-acetylated-tubulinimmunoreactive (-ir) and TH-ir cell groups, neuron-fiber tracts (5-HT-ir) and glial-processes (GFAP-ir). In the search of conserved traits among jawed vertebrates, we compared our data in *S. canicula* with that obtained in murine models. Our analysis reveals a strikingly high degree in the conservation of hypothalamic histogenetic compartments between chondrichthyan and murine models. Furthermore, we identified some of the boundaries and confirmed some of the assumptions predicted by the prosomeric model. However, some differences and discrepancies also exist mainly concerning the neuroepithelial specification genetic codes of the basal hypothalamus (BHy). Similar studies are required in other basal species to figure out if these differences should prompt the model update or they are the consequence of shark specialization.

### Materials and Methods

### Phylogenetic Reconstructions

Sequence alignments of the sequences listed in Table S1 were constructed using the alignment editor Seaview 3.0 and the MUSCLE algorithm. *S. canicula* sequences were retrieved by tblastn searches in transcriptomic databases obtained by Sanger and Illumina sequencing of embryonic and adult cDNA libraries (stages 8–25 and mixed adult tissues). Maximum-likelihood trees were inferred using the PhyML program version 3, the LG-F+-12+I substitution model and the SPR algorithm. Posterior probabilities (PPs) supporting groupings were calculated using the aLRT algorithm implemented in PhyML and are displayed as percentages at the corresponding nodes. Only PP >80% are indicated. The trees were viewed and edited using Mega6.

### Experimental Animals

Some embryos of the catshark (lesser spotted dogfish; *S. canicula*) were supplied by the Marine Biological Model Supply Service of the CNRS UPMC Roscoff Biological Station (France) and the Estación de Bioloxía Mariña da Graña (Galicia, Spain). Additional embryos were kindly provided by the Aquaria of Gijón (Asturias, Spain), O Grove (Pontevedra, Spain) and the Aquarium Finisterrae (A Coruña, Spain). Embryos were staged by their external features according to Ballard et al. (1993). For more information about the relationship of the embryonic stages with body size, gestation and birth, see Table 1 in Ferreiro-Galve et al. (2010). Thirty-seven embryos from stages 12 to 31 were used in this study. Eggs from different broods were raised in seawater tanks in standard conditions of temperature (15–16◦C), pH (7.5–8.5) and salinity (35 g/L). Adequate measures were taken to minimize animal pain or discomfort. All procedures conformed to the guidelines established by the European Communities Council Directive of 22 September 2010 (2010/63/UE) and by the Spanish Royal Decree 53/2013 for animal experimentation and were approved by the Ethics Committee of the University of Santiago de Compostela.

#### Tissue Processing

Embryos were deeply anesthetized with 0.5% tricaine methanesulfonate (MS-222; Sigma, St. Louis, MO, USA) in seawater and separated from the yolk before fixation in 4% paraformaldehyde (PFA) in elasmobranch's phosphate buffer [EPB: 0.1 M phosphate buffer (PB) containing 1,75% urea, pH 7.4] for 48–72 h depending on the stage of development. Subsequently, they were rinsed in phosphate buffer saline (PBS), cryoprotected with 30% sucrose in PB, embedded in OCT compound (Tissue Tek, Torrance, CA, USA), and frozen with liquid nitrogen-cooled isopentane. Parallel series of sections (12–20 μm thick) were obtained in transverse and sagittal planes on a cryostat and mounted on Superfrost Plus (Menzel-Glasser, Madison, WI, USA) slides.

### Single and Double Immunohistochemistry on Sections and Whole Mounts

For heat-induced epitope retrieval, sections were pre-treated with 0.01 M citrate buffer (pH 6.0) for 30 min at 95◦C and allowed to cool for 20–30 min at room temperature (RT). Sections were then rinsed twice in 0.05 M Tris-buffered saline (TBS; pH 7.4) for 5 min each and incubated overnight with the primary antibody (rabbit anti-serotonin [anti-5-HT] polyclonal antiserum, DiaSorin, Immunostar, Hudson, WI, USA, diluted 1:5000; polyclonal rabbit anti-Sonic Hedgehog [anti-Shh], Sta. Cruz Biotechnology, Santa Cruz, CA, USA, diluted 1:300; polyclonal rabbit anti-glial fibrillary acidic protein [anti-GFAP], Dako, Glostrup, Denmark, diluted 1:500; and monoclonal mouse antityrosine hydroxilase [anti-TH], Millipore, Billerica, MA, USA, diluted 1:500). Appropriate secondary antibodies [horseradish peroxidase (HRP)-conjugated goat anti-rabbit and anti-mouse, BIORAD, diluted 1:200] were incubated for 2 h at RT. For double immunohistochemistry (IHC) experiments, cocktails of primary antibodies were mixed at optimal dilutions and subsequently detected by using mixtures of appropriate secondary antibodies. Sections were rinsed in distilled water (twice for 30 min), allowed to dry for 2 h at 37◦C and mounted in MOWIOL 4- 88 Reagent (Calbiochem, MerkKGaA, Darmstadt, Germany). All dilutions were made with TBS containing 15% donkey normal serum (DNS; Millipore, Billerica, MA, USA), 0.2% Triton X-100 (Sigma) and 2% bovine serum albumin (BSA, Sigma). Double IHC with primary antibodies raised in the same species was performed as described in Tornehave et al. (2000).

For whole mounts embryos were prepared as previously described in Kuratani and Horigome (2000) with minor modifications. After fixation with 4% PFA in 0.01 M PBS at 4◦C for 2 days, embryos were washed in 0.9% NaCl in distilled water, dehydrated in graded series of methanol solutions (50, 80, 100%) and stored at −20◦C. Samples to be stained were placed on ice in 2 mL of dimethyl sulfoxide (DMSO)/methanol (1/1) until they sank. Then, 0.5 mL of 10% Triton X-100/distilled water was added, and the embryos were incubated for 30 min at RT. After washing in 0.05 M TBS with 0.1% Triton X-100 (TST, pH 7.4) the samples were sequentially blocked using spin-clarified aqueous 1% periodic acid and 5% non-fat dried milk in TST (TSTM). Primary antibody (monoclonal mouse anti-α-acetylated-tubulin, Sigma, 1:1000) was diluted in TSTM containing 0.1% sodium azide for 2–4 days at RT with gently agitation on a shaking platform. The secondary antibody HRP-conjugated goat anti-rabbit, BIORAD, dilution 1:200 in TSTM) was incubated overnight. After a final washing in TST, the embryos were pre-incubated with 0.25 mg/mL diaminobenzidine tetrahydrochloride (DAB, Sigma) in TST with 2.5 mg/mL nickel ammonium sulfate for 1 h, and then allowed to react with DAB in TST containing 2.5 mg/mL nickel ammonium sulfate and 0.00075% H2O2 for 20–40 min at RT. The reaction was stopped using Tris-HCL buffered saline and specimens were post-fixed with 4% PFA overnight at 4◦C. Epidermis and mesodermic derivatives were carefully removed and specimens were rinsed in graded series of glycerol (25, 50, 75, and 100%) in order to directly observe the neural tube under the stereomicroscope.

### Controls and Specificity of the Antibodies

No immunostaining was detected when primary or secondary antibodies were omitted during incubations. Controls and specificity of anti-TH and anti-5-HT were performed as described in Pose-Méndez et al. (2014). The primary anti-α-acetylated-tubulin antibody has been shown to label early differentiated neurons and their processes in the embryonic nervous system (Piperno and Fuller, 1985; Chitnis and Kuwada, 1990). The polyclonal anti-Shh antibody (Santa Cruz Biotechnology Inc, CA, USA) was raised in rabbit against the amino acids 41–200 of Shh human protein. The *in situ* hybridization (ISH) results were similar to those obtained by IHC, and therefore validate the specificity of the anti-Shh antibody used here.

### *In Situ* Hybridization on Whole Mount Embryos and on Sections

We applied ISH for *ScFoxg1a*, *ScShh* (Compagnucci et al., 2013; Quintana-Urzainqui, 2013), *ScNkx2.1* (Quintana-Urzainqui et al., 2012; Quintana-Urzainqui, 2013), *ScDlx5* (Compagnucci et al., 2013; Debiais-Thibaud et al., 2013), *ScOtp* (Quintana-Urzainqui, 2013), *ScTbr1* (Quintana-Urzainqui, 2013), and *ScDlx2* (Quintana-Urzainqui et al., 2012; Compagnucci et al., 2013; Debiais-Thibaud et al., 2013; Quintana-Urzainqui, 2013) genes. These probes were selected from a collection of *S. canicula* embryonic cDNA library (mixed stages 9–22), constructed in pSPORT1, and submitted to high throughput EST sequencing. cDNA fragments were cloned in pSPORT vectors. Sense and antisense digoxigenin-UTP-labeled and fluorescein-UTP-labeled probes were synthesized directly by *in vitro* transcription using as templates linearized recombinant plasmid DNA or cDNA fragments prepared by PCR amplification of the recombinant plasmids. ISH in whole mount and on cryostat sections was carried out following standard protocols (Coolen et al., 2009). Briefly, sections were permeabilized with proteinase K, hybridized with sense or antisense probes overnight at 65◦C (in sections) or 70◦C (whole mount) and incubated with the alkaline phosphatase-coupled anti-digoxigenin and anti-fluorescein antibody (1:2000, Roche Applied Science, Manheim, Germany) overnight at 4◦C. The color reaction was performed in the presence of BM-Purple (Roche). Control sense probes did not produce any detectable signal.

### Inhibition of the *Shh* Pathway

Inhibition of the Shh pathway was performed by *in ovo* injection of the pharmacological inhibitor cyclopamine in order to test whether, as in osteichthyans, the initiation of *ScNkx2.1* expression in the forebrain is dependent on Shh. First, 200 μL of a solution containing 1x PBS, 500 μM cyclopamine and 5% DMSO were injected through the shell of stage 15–16 *S. canicula* eggs. This solution was replaced by the same volume of 5% DMSO in 1x PBS for control embryos. The eggs were maintained for 3 days in oxygenated sea water at 17◦C, with viabilities higher than 90%. Embryos reached stage 18 in these conditions. They were dissected, fixed in PFA 4%, dehydrated and stored in methanol 100% prior to ISH.

### Image Acquisition and Analysis

Light field images were obtained with an Olympus BX51 microscope equipped with an Olympus DP71 color digital camera. *In toto* embryos were analyzed in the Olympus SZX12 stereomicroscope fitted to an Olympus DP12 color digital camera. Photographs were adjusted for brightness and contrast and plates were prepared using Adobe Photoshop CS4 (Adobe, San Jose, CA, USA).

### Results

### Identification of Catshark Orthologues of the Genes Studied

Exhaustive phylogenetic characterizations of the catshark *Dlx* gene repertoire have been previously published (Debiais-Thibaud et al., 2013), confirming the identity of *ScDlx2* and *ScDlx5*. In order to unambiguously identify the catshark orthologues of *Foxg1*, *Shh*, *Nkx2.1*, *Otp,* and *Tbr1*, we conducted systematic phylogenetic analyses of the corresponding vertebrate gene families, including all the vertebrate classes derived from duplications of a single ancestral chordate orthologue (**Figure 1**). In each case, phylogenies were constructed from alignments containing deduced amino acid sequences of all paralogous sequences retrieved from catshark transcriptomic databases and from a representative sampling of actinopterygians and sarcopterygians. The trees were rooted using a *Branchiostoma floridae* sequence, except in the case of Otp which could not be found in the amphioxus Ensembl database. In the case of Foxg1, three strongly supported classes (posterior probability or PP > 90%), each containing a catshark and several osteichthyan sequences, were retrieved, highlighting for the first time the presence of three gnathostome Foxg1 classes (**Figure 1A**). These classes were termed Foxg1a, Foxg1b, and Foxg1c, respectively. One coelacanth and several actinopterygian sequences, but no amphibian or amniote sequence, were found in the Foxg1b and Foxg1c classes, suggesting a loss of their representatives in tetrapods. We focused the expression analysis on *ScFoxg1a*, the catshark

analyzed in this study. Phylogenetic trees for the Foxg1, Hedgehog, Nkx2.1/Nkx2.4, Otp and Tbr1/Tbx21/Eomes families are shown in (A–E) respectively. The number of substitutions per site is indicated at the bottom of each tree, on the left. *S. canicula* genes are displayed in red. Abbreviations used: Hs, *Homo sapiens* (human); Gg, *Gallus gallus*

(African clawed frog); Lc, *Latimeria chalumnae* (coelacanth); Lo, *Lepisosteus oculatus* (spotted gar); Ol, *Oryzias latipes* (medaka); Dr, *Danio rerio* (zebrafish); Cm, *Callorhinchus milii* (elephant shark); Sc, *Scyliorhinus canicula* (catshark or lesser spotted dogfish); Bf, *Branchiostoma floridae* (amphioxus).

orthologue of the only *Foxg1* gene retained in all major gnathostome lineages including tetrapods. The tree topology obtained for the Hedgehog family confirmed the presence of the three gnathostome classes, corresponding to the Indian Hedgehog, Desert Hedgehog and Sonic Hedgehog classes already reported in osteichthyans, and confirmed *ScShh* as the representative of the latter (**Figure 1B**). Concerning the Nkx2.1/Nkx2.4 family, a single catshark gene could be identified and it was unambiguously assigned to the Nkx2.1 class based on the strongly supported grouping of its deduced amino acid sequence with teleost, chick, and human Nkx2.1 sequences (PP <sup>=</sup> 97%; **Figure 1C**). This gene is therefore referred to as *ScNkx2.1* hereafter. A single catshark Otp related sequence, termed ScOtp, could be found and as expected, it clustered with the elephant shark sequence annotated as Otp in the reconstruction shown in **Figure 1D**. Finally, the Tbr1, Tbx21, and Eomes classes were retrieved with high statistical support (PP = 83, 100, and 99%, respectively) within the Tbr1/Tbx21/Eomes family. Each class contained a single catshark sequence at the expected position, allowing an unambiguous identification of the *ScTbr1* gene analyzed in this study (**Figure 1E**).

### Preliminar Considerations Concerning Vertebrate Segmental Prosencephalic Organization

The organization of the shark hypothalamus has been analyzed in the framework of the updated prosomeric model (Puelles et al., 2012). **Figure 2** summarizes the general architecture of the hypothalamus in mouse according to the updated prosomeric model (Puelles et al., 2012). This model is mainly inspired in murine data though it is usually assumed that it can be extrapolated to all vertebrates because it also integrates information from other vertebrates (Puelles and Rubenstein, 2003; Pombal et al., 2009; Puelles, 2009). Indeed, this model represents a useful developmental and comparative framework since it makes use of concepts, nomenclature and topological references that can be used across different vertebrate species.

The prosomeric model establishes that hypothalamus and telencephalon are part of the secondary prosencephalon, which is understood as a segmental unit at the rostral-most point of the neural tube, the hypothalamus being located ventral to the telencephalon and rostral to the diencephalon (see **Figure 2A**).

The model also postulates that the rostral-most point of the brain, referred as the acroterminal region (At), lies at the rostral border of the secondary prosencephalon. This region is restricted to the frontal border of the neural tube where left and right alar and basal plates meet. This border expands dorso-ventrally from the rostral-most roof plate (which is telencephalic) to the rostralmost floor plate (which is hypothalamic). Thus, every structure classically considered being dorsal or ventral to these points (see arrowheads in **Figure 2A**), should be considered as caudal in this framework. Of note, the anterior commissure, located in the rostral-most roof plate, is a clear landmark of both the dorso-ventral and rostro-caudal axis (Puelles et al., 2012; see also **Figures 2A,B**).

illustrated according to Figure 8.5B in Puelles et al. (2012). (A) Longitudinal

prosencephalon according to the prosomeric model. For abbreviations, see list.

The secondary prosencephalon presents two true segments rostro-caudally arranged (**Figure 2B**): hp2 (rostral or terminal) and hp1 (caudal or peduncular). Each segment harbors telencephalic and hypothalamic derivatives (**Figures 2B,C**). However, the telencephalon harbors only roof and alar plates while the hypothalamus harbors alar, basal, and floor plate derivatives. The existence of these segments is supported by several genes differentially expressed in the rostro-caudal axis, the location of commissures in the roof and floor plates (anterior and retromammillary commissures, respectively), and the course of important tracts [medial forebrain bundle (mfb); lateral forebrain bundle (lfb); and fornix (fx)] running by a common path at the rostral border of hp1, through alar and basal plates. These data, in turn, support the existence of an intersegmental boundary that separates terminal and peduncular subdivisions of both telencephalon and hypothalamus, which is referred as the intrahypothalamic boundary (IHB; **Figures 2B,C**). Caudally, the secondary prosencephalon limits with the diencephalon at the hypothalamic diencephalic border (HDB), another intersegmental limit among hp1 and p3, though it should be noticed that part of the caudal limit of the secondary prosencephalon does correspond to the telencephalon (Puelles et al., 2012; see also **Figure 2C**).

The model considers the adult hypothalamic organization arranged in different histogenetic territories defined by neuroepithelial specification codes and radial units (Puelles and Medina, 2002; Puelles et al., 2012). These codes reveal that telencephalon and hypothalamus belong to different histogenetic territories being the preoptic area (POA) the unique terminal territory of the telencephalon (**Figure 2C**). Of note, the POA also harbors the anterior commissure (Puelles et al., 2012; see also **Figures 2B,C**).

#### *ScFoxg1a* Expression

In mice, *Foxg1* is one of the earliest transcription factors expressed specifically in the part of the neural plate that gives rise to the telencephalon and it remains expressed throughout the telencephalon during embryonic development (see Manuel et al., 2011). In an attempt to discriminate telencephalic and underlying hypothalamic domains throughout *S. canicula* development, we have analyzed the expression of *ScFoxg1a* in the developing nervous system of this species. At stage 18, *ScFoxg1a* expression was found in the dorsal-most portion of the secondary prosencephalon including the optic cup, extending from the level of the optic stalk (which is located rostrally, within the At) up to a caudal point in the roof plate, which has been tentatively identified as the dorsal border between the telencephalon and the diencephalon (**Figure 3A**). At stage 22, *ScFoxg1a* was observed in the telencephalon and in the nasal part of the optic cup (**Figure 3B**). The expression in the telencephalon was maintained until late stages of development (**Figure 3C**), which allowed identifying the border between the telencephalon and the hypothalamus.

#### *ScShh* Expression

*ScShh* expression was detected during gastrulation (stage 12) in the caudal midline of the embryo (data not shown). At stage 14, during early neurulation, it has been detected in the axial mesoderm of the notochord and the prechordal plate and in the ectoderm of the caudal midline (data not shown). After the closure of the neural tube (stage 17), the signal was detected as a ventral longitudinal continuous band that extends from the caudal end of the spinal cord to the At of the forebrain, roughly at the level of the optic stalk (**Figure 3D**). As in other vertebrates (Shimamura et al., 1995), the expression of *ScShh* can be used to define the alar-basal boundary (ABB; **Figures 3E–H**). At stage 19, *ScShh* expression became downregulated in the forebrain to progressively give rise to a caudal and a rostral domain (arrow in **Figures 3E–H**). The narrow transverse and dorsally directed stripe of *ScShh*-expressing cells within the caudal domain was identified as the developing zona limitans intrathalamica (zli; arrowhead in **Figure 3G**). The rostral border of the *ScShh* caudal domain, in turn, was somewhat extended rostral to the HDB (Puelles et al., 2012), which at this stage was identified as the point where the neural tube expands to acquire the distinctive shape of the ventral hypothalamus. Therefore, the BHy appeared to be divided in three domains: two positive for *ScShh* (one rostral and other caudal) and one (intermediate) negative for *ScShh* (arrow in **Figures 3G,H**; see also Figure 5H in Compagnucci et al., 2013). Of note, the dorsal border of the rostral domain (presumably corresponding to the ABB) seems to codistribute with α-acetylatedtubulin-immunoreactive (-ir) longitudinal tracts (arrowheads in **Figure 3I**). At stage 24 (**Figure 3H**), a new domain emerged within the telencephalon. This short domain (arrowhead in **Figure 3H**) extended from a region located dorsally to the optic stalk without reaching the prospective territory of the anterior commissure (that can be identified at early development by means of <sup>α</sup>-tubulin-immunoreactivity; asterisk in **Figures 3H,I**). A clear gap of expression was observed between this telencephalic domain and the rostral hypothalamic one (**Figure 3H**). The telencephalic domain was located medially while the hypothalamic one also expanded laterally (not shown). From stage 27 onward the zli expanded dorsally toward the roof plate (arrowhead in **Figure 4A**). At stage 29 the medio-lateral histologic organization of the developing walls of the forebrain become more evident. As in previous developmental stages, Shh immunoreactivity was clearly identified in the basal plate of the diencephalon entering the caudo-ventral part of the BHy (arrow in **Figures 4A,B,B-** ) and in the rostro-dorsal part of the BHy (**Figures 4A,B**), so that a clear negative gap of Shh-immunoreactivity occupied most of the caudal BHy (CBHy; **Figures 4A,B**) and part of the rostral BHy (RBHy). In the telencephalon, Shh-immunoreactivity expanded caudally beyond the prospective territory of the anterior commissure (arrowhead in **Figure 4B**; compare with **Figure 3H**). Of note, from late stage 30 onward, Shh-immunoreactivity is downregulated in the CBHy and basal diencephalon, except in the zli (data not shown).

#### *ScNkx2.1* Expression

The expression of *ScNkx2.1* was first detected at stage 18 in the rostro-ventral portion of the forebrain, in a longitudinal band which extended ventral to the optic stalk (arrowhead in **Figure 3J**). At stage 23 *ScNkx2.1* was expressed in most of the BHy (**Figure 3K**). Differently from *ScShh*, *ScNkx2.1* delimited the ABB even in the CBHy (**Figure 3K**;

#### FIGURE 3 |Continued

Regionalization of the hypothalamus and neighbor territories in embryos of *S. canicula* from stages 18–29 based on the expression pattern of *ScFoxg1a* (A–C), *ScShh* (D–H), *ScNkx2.1* (J–L), *ScDlx5* (M–O), *ScOtp* (P,Q), *ScTbr1* (R) genes and **α**-acetylated-tubulin immunoreactivity (I). In all panels, dotted lines define the hypothalamo-telencephalic boundary (HTB), dashed lines indicate the caudal border of the secondary prosencephalon and red lines indicate the ABB. (A–C) *ScFoxg1a* expression in the secondary prosencephalon at indicated stages. The arrowheads in (A) mark the caudo-dorsal and rostro-ventral limit of *ScFoxg1a* expression. (D–H) *ScShh* expression at the indicated stages. The arrowhead in (D) marks the rostral-most point of *ScShh* expression in the forebrain. The arrows in (E–H) indicate the downregulation of *ScShh* expression in the hypothalamus. The arrowhead in (G) points to the developing zli. The arrowhead in (H) points to a novel domain in the telencephalon. The asterisk in (H) marks the prospective territory of the anterior commissure. (I) Anti-α-acetylated-tubulin IHC to show three sets of tracts at stage 25. These tracts are classically referred as sot, TPOC and MTT. The asterisk indicates the territory of the developing anterior commissure. The arrowheads point to the longitudinal TPOC. The arrow points to the rostral-most extension of the MTT. (J–L) *ScNkx2.1* expression at the indicated stages. The arrowhead in (J) points to the rostral-most point of *ScNkx2.1* expression at stage 18, which was restricted to a short longitudinal domain ventrally to the optic stalk. The arrow in (K,L) points to a small *ScNkx2.1*-negative domain at the most caudo-ventral BHy. The asterisk in (K,L) marks the prospective territory of the anterior commissure. The arrowhead in (K,L) points to a domain in the telencephalon that spread rostro-caudally. (M–O) *ScDlx5* expression at the indicated stages. The arrowheads in (M,N) indicate *ScDlx5* expression in the olfactory placode and the anterior part of the telencephalon. The asterisk in (N) indicates the prospective territory of the anterior commissure. The arrowheads in (O) point to the ventral and caudal expansion of *ScDlx5* expression in the telencephalon. This domain was fairly continuous with a longitudinal band of *ScDlx5* over the ABB. The arrows in (O) point to *ScDlx5*-expressing domains that spread into the BHy. (P,Q) *ScOtp* expression at the indicated stages. The arrowhead in (P) indicates a restricted domain of *ScOtp* expression ventrally located with respect to the optic stalk. The expression of *ScOtp* in the hypothalamus was faint compared to that of the Rh. The white arrowhead in (Q) points to *ScOtp* expression in the AHy. Two additional *ScOtp*-expressing domains were observed in the BHy. (R) *ScTbr1* expression at stage 25 was found in part of the telencephalon and at the dorsal-most part of the rostral diencephalon (white arrowhead). The asterisk indicates the prospective territory of the anterior commissure. For abbreviations, see list.

compare with **Figure 3G**), though a small gap of expression was observed within the caudo-ventral part of the BHy (arrow in **Figure 3K**). A second domain emerged in the telencephalon at this stage (arrowhead in **Figure 3K**). This domain was restricted to the rostral-most portion of the telencephalon and extended from a region located dorsally to the optic stalk to the prospective territory of the anterior commissure (asterisk in **Figure 3K**). A clear gap of expression was observed between the telencephalic and the hypothalamic domains. At stage 25 (**Figure 3L**), as in previous developmental stages, *ScNkx2.1* expression was lacking in a small domain located within the caudo-ventral BHy (arrow in **Figure 3L**). This region seems to fit with the rostral border of a basal α-acetylated-tubulin-ir tract [see mammillo-tegmental tract (MTT) in **Figure 3I**]. In the telencephalon, *ScNkx2.1* expression became caudally expanded beyond the prospective territory of the anterior commissure (asterisk in **Figure 3L**). At stage 29, as in previous developmental stages, *ScNkx2.1* was observed throughout most of the BHy, except in a small wedge-shaped domain within the caudo-ventral BHy (arrowhead in **Figures 4C,D**). Groups of *ScNkx2.1*-expressing cells in the caudo-ventral portion of the hypothalamus were observed along the marginal zone (arrows in **Figures 4C,C- ,D**). In the telencephalon, *ScNkx2.1* expression expanded beyond the territory it occupied at previous developmental stages (asterisk in **Figure 4D**). Of note, Shh immunoreactivity was overlapping with *ScNkx2.1* expression beyond the anterior commissure (**Figure 4D**).

In order to test whether, as in osteichthyans, the initiation of *ScNkx2.1* expression in the forebrain is dependent on Shh, we used *in ovo* injections of the Shh inhibitor cyclopamine. All control embryos (*n* = 4) exhibited the expected *ScNkx2.*1 signal in the rostral-most and ventral-most portion of the forebrain (**Figure 5**). This signal was lost in all embryos dissected following cyclopamine treatment (*n* = 3), supporting the conclusion that Shh signaling is required for the initiation of *ScNkx2.1* expression in *S. canicula* as in osteichthyans.

#### *ScDlx2/ScDlx5* Expression

We analyzed the expression of *ScDlx5* from stage 18 onward and the expression of *ScDlx2* from stage 29 onward. Fairly identical results were observed with both markers in the brain of *S. canicula* from stage 29 onward, so we use *ScDlx2/5* at these stages to refer indistinctly to both.

General features of *ScDlx5* expression and detailed profiles in the developing branchial arches have been previously described from stage 15 to stage 27 in Compagnucci et al. (2013) and from stage 15 to stage 25 in Debiais-Thibaud et al. (2013). We revisited these data focusing on the developing forebrain. At stage 18, *ScDlx5* expression was found in the most anterior part of the neural tube (**Figure 3M**; compare with **Figure 3A**; see also Figure 5C1 in Debiais-Thibaud et al., 2013). From stage 21 to 25, *ScDlx5* becomes mostly restricted to the anteriormost part of the telencephalon and to the olfactory placodes (**Figure 3N**; see also Figure 4G in Compagnucci et al., 2013 and Figure 5C'1 in Debiais-Thibaud et al., 2013). However, at later stages (**Figure 3O**), *ScDlx5* expression spread caudally and ventrally (arrowheads in **Figure 3O**) and reached the rostral-most portion of the optic stalk (see also Figure 9C in Debiais-Thibaud et al., 2013). This domain was fairly continuous with a longitudinal band of *ScDlx5* that crossed through the hypothalamus over the ABB (compare with **Figures 3H,L**) and entered p3 (**Figure 3O**). Therefore, this domain delineates the ABB along the hypothalamus. Of note, the longitudinal domain appeared to codistribute with α-acetylated-tubulin tracts (arrowheads in **Figure 3I**). Although both *ScDlx5* domains were continuous, a wedge-shaped area of reduced signal intensity was observed between the dorsal (telencephalic) and the ventral (hypothalamic-diencephalic) domains (**Figure 3O**). Two bands of cells were additionally observed, which were ventrally located with respect to the longitudinal domain (arrows in **Figure 3O**). One was located at its caudal end and spread ventral-ward at the rostral end of p3. The other spread perpendicularly to the ABB from the caudo-dorsal part of the BHy up to the rostral hypothalamus (**Figure 3O**). Of note, this *ScDlx5*-expressing

FIGURE 4 | Continued

#### FIGURE 4 | Continued

Regionalization of the shark hypothalamus from stages 29–31 based on immunoreactivity to Shh (A,B,B**-** ,D,I–N), TH (B**-** ), 5-HT (I–L), GFAP (O, P) and expression of *ScNkx2.*1 (C,C**-** ,D,G**--**), *ScDlx2/5* (E,E**-** ,F,M) and *ScOtp* (G,G**-** , G**--**, H, H**-** , N) genes by means of single immunohistochemistry (IHC; A,B,O,P), double IHC (B**-** ,I–L), single ISH (C,E,E**-** ,F,G, G**-** ,H,H**-** ) and/or combined with IHC (D,M,N) on sagittal (A,B,B**-** ,C–G,G**--**,H–J,M,N) or transverse sections (C**-** ,E**-** ,G**-** ,H**-** ,K,L,O,P). Image in (G**--**) results from the overlapping of two parallel sections respectively hybridized with *ScOtp* and *ScNkx2.1* probes*.* Color for *ScNkx2.1* was digitally converted to brown to ease comparison. Dotted lines define the hypothalamo-telencephalic boundary (HTB), dashed lines indicate the caudal border of the secondary prosencephalon, red lines indicate the ABB and continuous black lines represent the path followed by 5-HT-ir fibers. (A) Shh-immunoreactivity was observed in the most caudo-ventral part of the CBHy (arrow) and in the RBHy. The arrowhead points to the zli. (B) Shh-immunoreactivity was observed in the telencephalon (arrowhead) beyond the territory of the anterior commissure (asterisk). The arrow points to Shh-immunoreactivity in the caudal-most CBHy. (B**-** ) Detail of the squared area in (B) to show Shh- and TH-immunoreactivity in its most caudo-ventral part (arrow). (C,D) *ScNkx2.1* expression was observed in the hypothalamus and telencephalon. A detail of a transverse section at the level indicated in (C) is shown in (C**-** ). The arrows in (C,C**-** ,D) point to *ScNkx2.1*-expressing cells in the mantle zone of the most caudo-ventral CBHy. The arrowheads in (C,D) indicate a wedge-shaped domain lacking *ScNkx2.1* expression. The asterisk in (D) indicates the territory of the anterior commissure. (E,F) *ScDlx2/5* expression was observed in p3 and in the secondary prosencephalon. A detail of a transverse section at the level indicated in (D) is shown in (D**-** ). The arrow in (E,E**-** ) points to *ScDlx2/5*-expressing cells in the mantle of the p3Tg. The star in (E,E**-** ) indicates the prospective territory of the PThE. The black asterisks in (E,F) indicate a gap of *ScDlx2/5* expression in the telencephalon. The arrowheads in (E,F) point to *ScDlx2/5* expression in the BHy. (G,H) *ScOtp* expression in the hypothalamus. Details in (G**-** ,H**-** ) correspond to transverse sections at the levels indicated in (G,H). Detail in (G**--**) correspond to the squared area in (G). The arrows in (G,G**--**) point to the ventricular domain expressing *ScOtp* in the caudal CBHy. The black arrowheads in (G,G**--**, H**-** ) point to *ScOtp*-expressing cells in the mantle of the most caudo-ventral part of CBHy and p3Tg. The arrow in (G**--**) indicates a domain expressing *ScNkx2.1* alone. The white asterisks in (G**-** ,H) indicate *ScOtp*-expressing cells in the mantle zone. The black arrowhead in (H) points to *ScOtp*-expressing cells in the AHy and the white arrowhead in (H) points to *ScOtp*-expressing cells between the alar and basal domains. (I–L) Double Shh- and 5-HT-immunoreactivity. (K,L) correspond to trasverse sections at the level indicated in (J). 5-HT-ir fibers in (I) are observed in the basal plate of the secondary prosencephalon. In the rostral hypothalamus such fibers coursed among RAHy and RBHy, and were located dorsally to Shh-immunoreactivity (I,J). Note the presence of 5-HT-ir fibers in the Sp (I,J,L). The white arrowhead in (J,K) points to 5-HT-ir fibers that course in the rostral CAHy. The black arrowhead in (J,L) points to 5-HT-ir fibers in the telencephalon. The arrow in (J,K) points to 5-HT-ir fibers decussating in the CBHy. The arrow in (L) points to the faint Shh immunoreactivity in the zli. (M,N) Shh IHC combined with *ScDlx2/5* expression (M) or *ScOtp* expression (N). A gap of expression is observed between the rostral and caudal domains of *ScDlx2/5* and *ScOtp* expression, which appeared to coincide with the path followed by 5-HT-ir fibers. (O) GFAP-ir processes at the level shown in (J). The arrowhead points to the GFAP-ir processes among RAHy and Sp. The arrow points to the GFAP-ir processes in the CBHy. (P) GFAP-ir processes at the level shown in (J). The arrow points to GFAP-ir processes in the Sp. For abbreviations, see list.

domain appeared to delineate *ScShh* expression in the rostral hypothalamus (compare with **Figure 3H**). This pattern became more patent at stage 29 (**Figures 4E,F**). At this stage, *ScDlx2/5* expression was observed in the telencephalon (subpallium). The telencephalic domain is almost continuous at medial levels (black asterisk in **Figure 4F**) while a clear gap of expression was observed at the level of the anterior commissure in parasagittal sections (black asterisk in **Figure 4E**). *ScDlx2/5* expression was also observed in a longitudinal domain outlining the ABB, which crossed through the alar hypothalamus (AHy) and entered p3 (**Figure 4E**). A wedge-shaped negative domain separated the telencephalic and hypothalamic-diencephalic domains (**Figure 4E**). A transverse band of non-ventricular *ScDlx2/5* expressing cells was observed extending ventral-ward from the alar plate (arrow in **Figures 4E,E-** ), along the rostral-most p3Tg. This domain appeared to overlap with the *ScShh*-expressing domain from which the zli emerged (compare with **Figure 4A**). As observed previously, an additional domain (arrowhead in **Figures 4E,F**) cut across the BHy perpendicularly to the ABB. This pattern was maintained until late stages of development (**Figure 4M**).

#### *ScOtp* Expression

At stage 19, *ScOtp* signal was detected in the rhombencephalon and in the rostral-most and ventral-most portion of the optic stalk (arrowhead in **Figure 3P**) though a faint *ScOtp* expression was also observed in part of the BHy and in the AHy. At stage 25, three domains of *ScOtp* expression were observed in the forebrain. The rostral one was restricted to the rostralmost region of the forebrain, expanding ventrally from the optic stalk along the RBHy without reaching the prospective

neurohypophysis. The second domain abutted the HDB at the intersection with the ABB and spread from this region up to the rostro-ventral hypothalamus. The third domain overlaid the ABB from the optic stalk up to the alar p3 and was poorly stained compared to the other two domains (arrowhead in **Figure 3Q**). From stage 28 onward these three domains were respectively identified in the RBHy (in **Figure 4G**), in an arched domain that spread from the CBHy (arrow in **Figures 4G,G-** ) and in the AHy (black arrowheads in **Figure 4H**). In the BHy, a small domain containing *ScNkx2.1* alone was identified caudal to the *ScOtp*-expressing domain (arrow in **Figure 4G**).

A novel domain of non-ventricular scattered *ScOtp-*expressing cells was also detected entering p3Tg from the most caudoventral part of the BHy (arrowheads in **Figures 4G,G--,H-** ). Scattered *ScOtp*-expressing cells were observed between the alar and basal domains of the caudal hypothalamus (white arrowhead in **Figure 4H**). In the AHy, *ScOtp*-expressing cells were only observed at parasagittal levels (**Figure 4H**; compare with **Figure 4G**). Of note, non-ventricular *ScOtp*-expressing cells appeared to codistribute with the longitudinal band of *ScDlx2/5* expression over the ABB (**Figure 4H**; compare with **Figure 4F**). *ScOtp*-expressing cells were also observed in the telencephalon (data not shown). This pattern is maintained until late stages of development. At stage 30 (**Figure 4N**), a gap of *ScOtp* expression was observed at parasagittal levels that divide the AHy in rostral and caudal domains.

### *ScTbr1* Expression

*ScTbr1* signal was detected at stage 25 in dispersed cells that spread the telencephalic vesicle (**Figure 3R**), except for a rostral domain that extended from the optic stalk to the prospective territory of the anterior commissure (asterisk in **Figure 3R**). Of note, α-acetylated-tubulin-ir tracts reaching the telencephalon seem to define the boundary between telencephalic *ScTbr1* positive and negative domains (see sot in **Figure 3I**; compare with **Figure 3R**). *ScTbr1* expression was additionally observed in the dorsal most part of the alar p3, abutting the *ScDlx5* domain that entered p3 (**Figure 3R**; compare with **Figure 3O**). At stage 29, the extension of the *ScTbr1*-expressing domain became decreased in the telencephalic vesicle (not shown).

## 5-HT **+** Shh Immunoreactivity

Anti-5-HT immunoreactivity was coanalyzed with anti-Shh immunoreactivity at stage 30 to better understand the segmental organization of different fiber bundles, which in turn contributes to the understanding of the organization of the hypothalamus. Immunoreactivity for both markers was examined at stage 30 when the first 5-HT-ir fibers reach the telencephalon (Carrera et al., 2008). Positive fibers were observed coursing parallel to the ABB (**Figure 4I**). Besides, some fibers were detected ascending throughout the AHy (white arrowhead in **Figures 4J,K**) toward the telencephalon (black arrowhead in **Figures 4J,L**) from the caudal part of the hypothalamus. This pathway seemed to concur with negative domains for *ScDlx2/5* and *ScOtp* genes at the same developmental stage (**Figures 4M,N**). A group of decussating fibers was detected close to the most caudo-ventral region of the hypothalamus (arrow in **Figures 4J,K**) that coincides with the caudal-most and ventral-most domain of *ScNkx2.1* in the BHy (**Figure 4J**; compare with **Figures 4C,D**) and also with the rostral end of the ventral-most α-acetylated-tubulin-ir tract at stage 25, that could represent a pioneering tract, the mammillo-tegmental tract (MMT in **Figure 3I**). Of note, from stage 30 onward, Shh immunoreactivity was only observed in the rostral portion of the zli but not in the caudo-ventral part of the BHy nor the diencephalic basal plate (**Figure 4J** and arrow in **Figure 4L**).

### GFAP Immunoreactivity

Glial fibrillary acidic protein immunoreactivity was also analyzed in the forebrain at stage 31. Radial and longitudinal GFAP-ir processes were detected through the whole forebrain (**Figures 4O,P**). Ascending fibers to the telencephalon were detected in a similar pathway to that described above for 5-HT (black arrowhead in **Figure 4O**; compare with white arrowhead in **Figure 4K**). Some GFAP-ir processes were also observed in the same point where 5-HT-ir fibers decussate in the caudo-ventral part of the BHy (arrow in **Figure 4O**; compare with **Figure 4K**). GFAPir processes were also observed in the subpallium (arrow in **Figure 4P**).

#### Santos-Durán et al. Prosomeric organization of the catshark hypothalamus

### Discussion

#### Alar Hypothalamus

According with the updated prosomeric model (Puelles et al., 2012; see also **Figure 6A**) the AHy, together with the telencephalon, are the rostral-most regions of the alar plate. The AHy is located ventral to the *Foxg1*-expressing telencephalon, dorsal to the BHy (which is characterized by the expression of *Nkx2.1* in its whole extension except in its caudal-most portion), and rostral to the alar p3 (characterized by the complementary expression of *Dlx* and *Tbr1* genes). Within the territory delimitated by the above-mentioned genes, two longitudinal (dorso-ventrally arranged) histogenetic domains are defined based on the complementary expression of *Dlx* and *Otp* genes. The dorsal-most domain is termed paraventricular domain (Pa), and expresses *Otp* but not *Dlx* genes. The ventral-most is the subparaventricular domain (SPa), which expresses *Dlx* genes but not *Otp*. *Dlx* is expressed beyond the HDB in the alar p3. The alar HDB is defined by the clear-cut expression among genes restricted to the AHy (*Sim1*, *Otp)* or to the alar p3 (*Lhx9*, *Arx*, *Olig2* among others). The alar p3, besides, includes the prethalamic eminence (PThE) and express genes in a complementary manner (PThE: *Tbr1*, *Lhx9*, *Gdf10*; alar p3: *Dlx* genes, *Arx*, *Gsh2*). Moreover, according with the prosomeric model both, Pa and SPa domains, present terminal (hp2) and peduncular (hp1) domains (TPa, PPa; TSPa, PSPa respectively). However these subdomains cannot be genetically identified without additional markers. Thus, the AHy presents at least four different histogenetic domains although some work on the development of hypothalamic peptidergic cells point to much more subdivision (**Figure 6A**; see also Morales-Delgado et al., 2011, 2014; Puelles et al., 2012).

In the shark, we have studied the expression of *ScFoxg1a*, *ScShh*, *ScNkx2.1*, *ScDlx2/5*, *ScOtp,* and *ScTbr1*, which allow us to identify an AHy harboring Pa-like and SPa-like histogenetic domains and their boundaries. The early expression territories of *ScFoxg1a, ScNkx2.1* and *ScOtp/ScTbr1* define the dorsal, ventral and caudal limits of the AHy, respectively (**Figures 3B,K,Q,R**; see also **Figure 6B**). Thus, *ScFoxg1a* was expressed from early stages of development in the presumptive telencephalon (**Figure 3C**), leading to the identification of the telencephalon-AHy border (see also **Figure 6B**). *ScNkx2.1* was expressed in most of the BHy (**Figure 3L**) and delineates the ABB (see also **Figure 6B**). The caudal border of the AHy could be delimited at stage 25 by the caudal-most expression of *ScOtp* at parasagittal levels (**Figure 3Q**; see also **Figure 6B**). Both *ScOtp* and *ScDlx5* were expressed from stage 25 onward within the AHy, though the intensity of the expression increases in late stages of development. At stage 29, both *ScOtp* and *ScDlx2/5* were observed in the AHy. Of note, two mutually exclusive histogenetic domains dorsoventrally arranged could be readily observed at ventricular levels (**Figures 4E,H**; see also **Figure 6B**), though non-ventricular *ScOtp*-expressing cells were also observed within the *ScDlx2/5* domain. These observations allowed us to identify the TPa/PPalike and the TSPa/PSPa-like domains (**Figure 6B**). The former, *ScOtp*-expressing one, abutted a *ScTbr1* domain that occupied the dorsal-most part of p3 (the PThE; **Figure 6B**), which is also negative for *ScDlx2/5* (compare with **Figures 4E,E-** ). The latter, *ScDlx2/5*-expressing one, as in mouse, was continuous with the transverse *ScDlx2/5*-expressing domain in p3 (**Figures 4E** and **6B**).

In the mouse, and using markers homologous to those studied here, termino-peduncular compartments can be differentiated only by the late expression of *Tbr1* in the mantle of PPa (Puelles et al., 2012). While *ScTbr1* marker did not reveal rostro-caudal differences in the shark, we discerned these compartments by means of the identification of the mfb, which coursed caudally to the IHB (see below).

### Basal Hypothalamus

The BHy is the rostral-most territory of the basal and floor plates. It is located ventrally to the AHy and rostral to the basal p3 (p3Tg; **Figure 6A**). The BHy is characterized by the expression of *Nkx2.1* in its whole extension with the exception of its caudal-most region. It harbors three longitudinal domains dorso-ventrally arranged: tuberal/retrotuberal region (Tu/RTu); perimammillary/periretromammillary region (PM/PRM) and mammillary/retromammillary region (MM/RM). These domains harbor terminal and peduncular parts separated by the IHB. The peduncular domains are distinctly referred in this basal region with the prefix "retro" while the terminal lack prefix. In the mouse (**Figure 6A**), the Tu/RTu compartment is characterized by the expression of *Dlx5* although it also presents subdomains expressing *Shh* (in part of the Tu and the whole RTu); the PM/PRM is characterized by the expression of *Nkx2.1* and *Otp* although *Shh* is also expressed in the PRM; the MM only expresses *Nkx2.1* (being the caudal and ventral-most portion of the *Nkx2.1* domain) while the RM express *Shh* and other genes in a complementary manner to *Nkx2.1* (see Morales-Delgado et al., 2011, 2014; Puelles et al., 2012). The basal HDB is defined by the clear-cut expression among genes restricted to the caudal part of the hypothalamus (*Pitx2*, *Lhx5* among others) or to the p3Tg (*Nr5a1*, *Nkx6.1*; Puelles et al., 2004, 2012; Medina, 2008).

In the shark, the analysis of the expression of *ScShh*, *ScNkx2.1*, *ScDlx2/5,* and *ScOtp* highlighted the presence of Tu/RTu-like, PM/PRM-like and MM/RM-like histogenetic domains within the BHy. However, the relative organization of some of these expression territories differed between the catshark and mammals. While *ScNkx2.1* expression territories were well established in the BHy from early stages of development, the other markers analyzed here showed a more dynamic pattern through development. We chose to focus on stage 29 for the interpretation of these data, all studied markers exhibiting a strong expression at this stage, with sharp boundaries.

In mouse, according with the prosomeric model, the *Nkx2.1* territory in the basal plate located dorsal to the *Otp*, is considered to represent the Tu/RTu domain. *Dlx* genes are widely expressed in this domain (**Figure 6A** and Puelles and Rubenstein, 2003), though some tuberal areas lack *Dlx* gene signal (Morales-Delgado et al., 2011). Accordingly, in *S. canicula*, the BHy located dorsal to the *ScOtp*-expressing domain was interpreted as the Tu/RTu-like domain (**Figure 6B**). Within this territory, *ScOtp* was expressed in a restricted stripe at the most rostro-dorsal part of the basal plate spreading ventral-ward from the ABB without reaching the neurohypophysis (see **Figures 3Q** and **6B**). *ScShh* overlapped with this domain, extended caudally and abutted a *ScDlx2/5* expressing domain that extended from the ABB up to the neurohypophysis. Therefore, in sharks, the Tu/RTu-like domain would be composed by four subdomains: a rostro-dorsal subdomain coexpressing *ScShh*, *ScOtp* and *ScNkx2.*1; a second subdomain expressing *ScShh* and *ScNkx2.1*; a third subdomain expressing *ScDlx2/5* and *ScNkx2.1;* and a ventral subdomain expressing *ScNkx2.1* alone (compare **Figures 4C,D**; see also **Figure 6B**). As in mouse, and with the markers studied here, it was not possible to establish a clear distinction between the Tu-like and the RTu-like compartment. It is noteworthy that the Tu/RTulike compartment has a different histogenetic identity to that described in the mouse. On one hand, *ScShh* was down-regulated in a large portion of the BHy. On the other, *ScDlx2/5* expression appeared much more restricted than in mouse, and different subdomains could be delineated, including a domain expressing *ScNkx2.1* alone in midsagittal sections (**Figures 4C,D**; see also **Figure 6B**). In mouse, the single domain in the BHy that contains *Nkx2.1* expression alone was identified as the MM region. The *ScNkx2.1* expressing region ventral to *ScDlx2/5* is unlikely to correspond to the MM for two reasons. First, it is not the only compartment where we detect the expression of *ScNkx2.1* alone. Indeed, in mid-sagittal sections, a small domain can be observed ventral to *ScOtp*, which expressed *ScNkx2.1* alone (**Figures 4G--** and **6B**). Second, the presence of a MM-like domain at this location would imply the interruption of longitudinal compartments (MM-like located dorsally to PM-like) and the redefinition of terminal domains within the BHy. Since it has been previously reported that, in mouse, some tuberal areas lack *Dlx* gene signal (Morales-Delgado et al., 2011), the most parsimonious interpretation implies that this *ScDlx2/5*-lacking region belongs to the Tu/RTu-like area and that *ScDlx2/5* cannot be used to identify the ventral border of the Tu/RTu-like domain, at least in mid-sagittal sections.

According to the prosomeric model, in mouse, adjacent to the Tu/RTu area, a distinct PM/PRM histogenetic domain exists in which *Otp* expression is selectively found. In *S. canicula*, *ScOtp* expression was observed in an arched domain that spread from the CBHy at the ABB/HDB junction and entered the rostral hypothalamus (**Figures 4G-** and **6B**). At its caudal-most portion, this domain seems to express *ScOtp* only in the ventricular zone (asterisk in **Figures 4G- ,H**; see also **Figure 6B**), while in its rostral-most portion *ScOtp* is mainly expressed on mantle cells. This *ScOtp*-expressing domain (including either ventricular or mantle cells) was therefore interpreted as a PM/PRM-like domain, where the PRM-like domain is likely to correspond to caudal *ScOtp* expression in cells at the ventricular zone, and the PM-like would mainly correspond to rostral *ScOtp* expression in mantle cells.

As in mouse, a small domain containing *ScNkx2.1* alone was identified ventral to the *ScOtp*-expressing domain (**Figures 6A,B**). This domain abutted a *ScShh*-expressing domain which expanded caudally through the diencephalic basal plate (**Figure 6B**). Based on these expression similarities, the *ScNkx2.1* and *ScShh*-expressing territories were respectively identified as the MM-like and the RM-like domains. The caudal border of the RM-like domain abutted a transverse band of non-ventricular *ScDlx2/5*-expressing cells at the p3Tg (**Figures 4A,E**; see also **Figure 6B**).These *ScDlx2/5*-expressing cells are just ventral to the *ScDlx2/5*-expressing domain in the alar p3, supporting their assignment to p3Tg. Furthermore, these cells are in the same position that Pax6-ir cells in the caudal posterior tuberculum of the shark, described by Ferreiro Galve (2010) at equivalent developmental stages. Similar diencephalic basal plate *Dlx*-expressing cells have not been described in the mouse but *Dlx2-* and *Pax6*-expressing cells have been described in the basal plate of zebrafish as belonging to the preglomerular complex (p3Tg), suggesting that their presence may be an ancestral characteristic of jawed vertebrates lost in mammals (Ishikawa et al., 2007; Vernier and Wullimann, 2008). Accordingly to this, the HDB in the basal plate lie rostral to non-ventricular *ScDlx2/5*-expressing cells in p3Tg.

### Posterior Tuberculum

The shark hypothalamus and the posterior tuberculum have been analyzed before, mainly under neurochemical and topographical approaches. The posterior tuberculum in chondrichthyans extends caudally from the posterior recess (or mammillary recess). This recess lies just at the caudal and ventral border of *ScNkx2.1* expression in the MM-like domain and the rostral and ventral border of *ScShh* expression in the RM-like domain (**Figures 4A,C,D**; see also **Figure 6B**). A posterior tuberculum harboring TH-ir cells has been classically related to the hypothalamus (Smeets, 1998) although modern studies addressed it as belonging to p3Tg (Carrera, 2008; Carrera et al., 2008, 2012; Ferreiro-Galve et al., 2008; Quintana-Urzainqui et al., 2012). Our present genoarchitectonic analysis suggests that the bigger part and rostral-most located portion of TH-ir cells and fibers belongs to the RM-like (*ScShh*-expressing) domain while the caudal-most located portion of these TH-ir cells and fibers belongs to p3Tg (**Figure 4B-** ). Thus, the rostral-most portion of TH-ir cells should be understood as hypothalamic. This configuration seems to fit parsimoniously with the prosomeric model under the light of the following facts. TH-ir ascending fibers from the rostral posterior tuberculum to the telencephalon seem to arise from these THir hypothalamic groups (compare with Figures 4A,B in Carrera et al., 2012). As discussed below, these ascending TH-ir tracts seem to respect and follow the intersegmental boundary between hp2 and hp1 as proposed by the updated prosomeric model (Puelles et al., 2012), since they course in the most rostral part of hp1. Furthermore, equivalent cells in the posterior tuberculum and their ascending fibers to the telencephalon seem to exist across different vertebrate species (Vernier and Wullimann, 2008) and so, the situation observed in the shark is likely to occur in other vertebrates. Besides having a hypothalamic location, this TH-ir cell population seems to have a hypothalamic origin among different vertebrates. Except in reptiles and mammals, these cells emerge concurrently with those of the rostral hypothalamus in all vertebrates studied so far, supporting a conserved hypothalamic origin (Carrera et al., 2012). On the other hand, in *S. canicula*, *ScOtp* is expressed in the PM/PRM-like area from stage 25 onward (**Figure 3Q**) just before TH-ir cells emerge in the rostral posterior tuberculum (Carrera et al., 2012). Interestingly, in zebrafish, mutants lacking *Otp* expression in the hypothalamus also lack hypothalamic and posterior tubercular TH-ir groups (Ryu et al., 2007). Finally, at late stages in *S. canicula*, as in zebrafish and mouse (Ryu et al., 2007; Puelles et al., 2012), scattered *ScOtp*-expressing cells were observed outside the ventricular zone of the PM/PRM-like region entering the marginal zone of the RM-like and p3Tg (**Figures 4H** and **6B**), which support a PM/PRM-like origin for *ScOtp*-expressing cells in p3Tg.

Thus, it appears that, among different vertebrates, at least some populations of the RM-like and p3Tg emerge from hypothalamic domains expressing *Otp*, and that TH-ir cells of the RM-like domain send ascending projections to the telencephalon caudally to the IHB.

### Intrahypothalamic Boundary Identification

The updated prosomeric model proposes a secondary prosencephalon divided in two segments, hp2 and hp1, separated by the IHB. Both segments include a hypothalamic and a telencephalic counterpart (**Figure 6A**). This boundary is supported by (i) the existence of the anterior and retromammillary commissures in the roof and floor plates, respectively, (ii) the restricted expression of several genes to one or other segment, and (iii) the course of main tracts [medial forebrain bundle (mfb), lateral forebrain bundle (lfb) and fornix (fx)] separating both segments. Thus, there is a correlation among histogenetic data and fiber tract data. In fact, it was argued that the course of tracts is a powerful test for brain models since they are also guided by mechanisms related to those involved in histogenetic patterning (Puelles et al., 2012).

In the shark (**Figure 6B**), the IHB is also supported by (i) the existence of commissures in the alar and floor plates, (ii) the restricted expression of *ScShh* and *ScNkx2.1* in particular subdomains within either the rostral or the caudal segment, and (iii) the presence of different neurochemical populations of fibers, topologically homologue to those described in the model, which divide the hypothalamus in a caudal (peduncular, hp1) and a rostral (terminal, hp2) domain. Thus, we have tentatively defined an IHB-like based on genetic or histogenetic data and partial neurochemical or fiber tract data. We also discuss the congruence between the two kinds of data, when possible.

### The Presence of Commissures

In the roof plate, according to the prosomeric model, the IHB ends caudal to the anterior commissure. This commissure has been identified in *S. canicula* by means of α–acetylated-tubulin immunoreactivity in early embryos (**Figure 3I**) and GFAPimmunoreactivity at late development (**Figure 4P**). In the floor plate, the IHB coincides with the border between MM and RM-like subdomains. In mouse, the IHB at this point can be associated with the retromammillary commissure (also known as Forel's ventral tegmental commissure) where fx tracts cross at its rostral-most portion. This commissure seems to be continuous and indistinguishable from the prethalamic (p3Tg) commissure, a bit more caudally arranged, where fibers (partly p3Tg and partly from the raphe nuclei) cross (see Puelles et al., 2012). Based on 5-HT-ir tracts, we have identified in *S. canicula* a conspicuous commissure rostrally located with respect to the HDB, as predicted in the frame of the prosomeric model. A hypothalamic commissure has been also identified at this point in adult chondrichthyan specimens, which has been referred as postinfundibular commissure and extends through the ventral and rostro-caudal extension of the posterior tuberculum. This postinfundibular commissure presents differential rostro-caudal connectivity and no other commissures have been described at this point, closely resembling the scenario found in the mouse. While the rostral portion (or pars superior) connects hypothalamic cell masses, fibers of the tract of the saccus vasculossus decussate in the caudal part (or pars inferior; Smeets, 1998). Previous work on the shark reveals 5-HT-ir and GAD-ir fibers crossing in the rostral and caudal extension of this commissure, respectively (Sueiro et al., 2007; Carrera, 2008). It has been proposed that these 5-HT-ir fibers belong to 5-HT-ir cells of the posterior tuberculum (Sueiro et al., 2007; Carrera, 2008) although this fact has not been confirmed. However, in mammals, on which 5-HT-ir projecting cells are only located in the brainstem, a similar commissure has been described and referred as supramammillar commissure (Botchkina and Morin, 1993) which we assumed as equivalent to the prosomeric retromammillary commissure (Puelles et al., 2004). While it remains unclear whether the postinfundibular commissure of sharks is homologous to the retromammillary commissure of mouse, it appears that its location fit with the floor plate limit of the IHB (**Figure 6B**).

### Correspondence to Histogenetic Domains

Either the expression of different genes (that would help identifying different histogenetic domains) or, accordingly, the course of the IHB through the telencephalon, has not been analyzed here. In the AHy, the prosomeric model proposes that the IHB can be delineated just caudal to the optic stalk, at the caudal border of expression of genes like *Six3*, *Neurog3*, *Six6*, *Nkx2.6* or de rostral border of *Tbr1*, *Uncx4.1*, *Sim1*, *Olig2*, *Foxb1*, *Nr5a1*, which is the same point whereby the mfb, lfb, and fx course (Shimogori et al., 2010; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012; see below). In the BHy, the model proposes that the IHB can be delineated just caudal to the caudal border of genes as *Nkx2.1*, *Olig2*, *Foxb1* or *Nr5a1* or de rostral border of genes as *Lmx1b*, *Lhx5*, *Ptix2*, *Lhx1*, *Lhx6*, *Lhx9*, *Arx* or *Irx5*, which is the same point whereby the fx run (Shimogori et al., 2010; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012). Therefore, the PPa, PSPa, RTu, PRM, and RM domains belong to hp1, while TPa, TSPa, Tu, PM, and MM domains belong to hp2 (**Figure 6A**).

In the AHy of *S. canicula*, as discussed above, no evidence for subdivisions along the rostro-caudal axis could be found on the basis of *ScOtp* or *ScDlx2/5* expression. In the BHy, two distinct territories could be inferred from *ScNkx2.1* and *ScShh* expressions, the MM-like (rostral, *ScNkx2.1*-expressing) and the RM-like (caudal, *ScShh* expressing) domains (**Figure 6B**), which is consistent with the predictions of the prosomeric model. According to the prosomeric model, the IHB can be delineated between both domains. The other compartments (Tu/RTu and PM/PRM) are presumably divided by the IHB, but any other gene among those used here serves as a caudal (hp1) or rostral (hp2) marker, except for *ScShh* in the Tu/RTu-like domain that, in *S. canicula* (but not in mouse) seems to be restricted to the rostral (hp2) domain.

### Main Tracts Coursing the Chondrichthyan Alar and Basal IHB

In the mouse, the fx is the only tract coursing from the alar to the basal plate that additionally decussates in the hypothalamic floor plate by the retromammillary commissure. The mfb is also a transverse peduncular tract whose rostral border follows the IHB (Puelles et al., 2012). In chondrichthyans, a fx counterpart has not been successfully confirmed to date (Smeets, 1998). Only a counterpart of the mfb has been described, which is referred as the basal forebrain bundle or *fasciculus basalis telencephali* in chondrichthyans literature [the sot in zebrafish literature (Smeets, 1998; Carrera, 2008; Carrera et al., 2008, 2012; Puelles et al., 2012)]. Ascending and descending projections between the telencephalon and the superior and caudal part of the inferior lobes (which probably correspond to the lateral and caudal part of the Tu/RTu-like domain defined here) have been experimentally confirmed coursing through the mfb of different adult chondrichthyans (Smeets, 1998; Hofmann and Northcutt, 2008). These facts support the existence of tracts coursing caudally to a hypothetic IHB. Interestingly, Carrera et al. (2012) have identified TH-ir fibers ascending from the posterior tuberculum to the telencephalon through the mfb, as in other vertebrates (see above; Vernier and Wullimann, 2008; Carrera et al., 2012). We argued above that the TH-ir cells of the rostral posterior tuberculum previously described as belonging to p3Tg in sharks, probably belong to what we identified here as the RM-like compartment, and that the situation observed in the shark is likely to occur in other vertebrates. Thus, TH-ir fibers likely ascending from the RM-like to the telencephalon seem to course caudally to the IHB (see above), which additionally support the identification of the rostral border of hp1. Therefore, the presence of tracts in the rostral border of the RM-like domain parsimoniously fits with the predictions of the prosomeric model respect the course of fiber tracts in the rostral border of the hp1prosomere and caudally to a hypothetic IHB (**Figure 6B**). Of note, 5-HT-ir and GAD-ir cells have been also identified in the posterior tuberculum of the shark and other vertebrates. Whether ascending projections from this source coursed to the telencephalon has not been determined to date (Barale et al., 1996; Mueller et al., 2006; Carrera, 2008; Carrera et al., 2008; Lillesaar, 2011). However, these tracts are likely to join those mfb tracts that ascend to the telencephalon just caudal to the optic stalk, indirectly drawing the boundary among the Tu-like and RTu-like.

In the alar plate, 5-HT-ir, GAD-ir, TH-ir and GFPA-ir fibers of the mfb, arising from different points of the brain, have been observed ascending to the telencephalon by a common path just caudal to the optic stalk, which could correspond to the alar IHB (Carrera, 2008; Carrera et al., 2012, **Figures 4J,K,O**). The most conspicuous of these tracts were 5-HT-ir tracts recognizable at stage 30 (Carrera, 2008; Carrera et al., 2008, and **Figures 4I–L** in present work). This path is also followed by GFAP-ir processes which, besides, cross at the anterior commissure (**Figures 4O,P**). Since, as reported in mouse, they coursed just caudal to the optic stalk, we propose that these tracts in *S. canicula* course caudally to the IHB, separating the TPa and TSPa (hp2) from the PPa and PSPa (hp1; **Figure 6B**). While, with the markers used here, we cannot ascertain if this boundary in the AHy is also supported by histogenetic domains, these tracts appear to course through gaps of *ScDlx2/5* and *ScOtp* expression in the telencephalon (**Figures 4M,N**, respectively).

To summarize, our data support the conclusion that, in *S. canicula*, an IHB topologically homologous to that proposed by the updated prosomeric model, courses from the anterior commissure (in the telencephalon) to the postinfundibular commissure (in the hypothalamus) through the mfb. Note that, in contrast to mouse, the IHB in *S. canicula* can be traced by partial genetic and fiber tract evidences, since tracts coursing through the telencephalon to the floor plate have not been demonstrated. In the alar plate the mfb courses topologically caudal to the optic stalk and is known to be composed by different neurochemical systems arising from different points of the brain. In the basal plate the mfb is known to be composed, at least, of TH-ir fibers which arise upward from the rostral posterior tuberculum.

### Conclusion

We have revisited and reinterpreted the organization of the developing hypothalamus in a chondrichthyan model, *S. canicula*,within a prosomeric and histogenetic framework. These data reveal striking similarities with the organization described in the mouse by means of the updated prosomeric model (Puelles et al., 2012). In the AHy we have tentatively identified TPa/PPalike, TSPa/PSPa-like histogenetic domains and their boundaries. In the BHy we have identified similar histogenetic domains to those observed in the mouse (Tu/RTu, PM/PRM, RM/MM-like). The fact that *ScShh* was downregulated in a large portion of the BHy and *ScDlx2/5* expression was much more restricted than in mouse have allowed us to identify different subdomains within the Tu/RTu-like area. Furthermore, we have identified an IHB separating terminal and peduncular portions of telencephalon and hypothalamus, as the model predicts, based partially on genetic and fiber tract data. Altogether, these data show that the prosomeric model in its latest version provides an adequate reference to describe the molecular organization of the catshark developing hypothalamus, thus highlighting the underlying unity of this complex anatomical structure across jawed vertebrates.

### Acknowledgments

This work was supported by grants from the Spanish Dirección General de Investigación-FEDER (BFU2010- 15816), the Xunta de Galicia (10PXIB200051PR, CN 2012/237), European Community-Research Infrastructure Action under the FP7 "Capacities" Specific Programme (ASSEMBLE 227799), the Région Centre, Région Bretagne (EVOVERT grant number 049755; PEPTISAN project), National Research Agency (grant ANR-09-BLAN-026201), CNRS, Université d'Orléans and Université Pierre et Marie Curie. GNSD would like to thank Spanish SEPE for its funding support.

### References


## Supplementary Material

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fnana.2015.00037/ abstract

*Markers*. Ph.D. thesis, Universidad de Santiago de Compostela, Santiago de Compostela.


insights on the basal ganglia. *Brain Behav. Evol.* 80, 127–141. doi: 10.1159/000 339871


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Santos-Durán, Menuet, Lagadec, Mayeur, Ferreiro-Galve, Mazan, Rodríguez-Moldes and Candal. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# NEUROANATOMY

**Sandra Blaess <sup>1</sup>\*, Nora Szabó<sup>2</sup> , Roberta Haddad-Tóvolli <sup>3</sup> , Xunlei Zhou<sup>3</sup> and Gonzalo Álvarez-Bolado<sup>3</sup>\***

<sup>1</sup> Neurodevelopmental Genetics, Institute of Reconstructive Neurobiology, University of Bonn, Bonn, Germany

<sup>2</sup> Department of Neurobiology and Development, Institut de Recherches Cliniques de Montréal, Montréal, QC, Canada

<sup>3</sup> Department of Medical Cell Biology, Institute of Anatomy and Cell Biology, University of Heidelberg, Heidelberg, Germany

#### **Edited by:**

Valery Grinevich, German Cancer Research Center DKFZ and University of Heidelberg, Germany

#### **Reviewed by:**

Sara Ferrando, University of Genoa, Italy Clemens Martin Kiecker, King's

### College London, UK

#### **\*Correspondence:**

Sandra Blaess, Neurodevelopmental Genetics, Institute of Reconstructive Neurobiology, University of Bonn, Sigmund-Freud-Straße 25, D-53127 Bonn, Germany e-mail: sandra.blaess@uni-bonn.de; Gonzalo Álvarez-Bolado, Department of Medical Cell Biology, Institute of Anatomy and Cell Biology, University of Heidelberg, Im Neuenheimer Feld 307, D-69120 Heidelberg, Germany e-mail: alvarez@ ana.uni-heidelberg.de

The expression pattern of Sonic Hedgehog (Shh) in the developing hypothalamus changes over time. Shh is initially expressed in the prechordal mesoderm and later in the hypothalamic neuroepithelium—first medially, and then in two off-medial domains. This dynamic expression suggests that Shh might regulate several aspects of hypothalamic development. To gain insight into them, lineage tracing, (conditional) gene inactivation in mouse, in ovo loss- and gain-of-function approaches in chick and analysis of Shh expression regulation have been employed. We will focus on mouse studies and refer to chick and fish when appropriate to clarify. These studies show that Shh-expressing neuroepithelial cells serve as a signaling center for neighboring precursors, and give rise to most of the basal hypothalamus (tuberal and mammillary regions). Shh signaling is initially essential for hypothalamic induction. Later, Shh signaling from the neuroepithelium controls specification of the lateral hypothalamic area and growth-patterning coordination in the basal hypothalamus. To further elucidate the role of Shh in hypothalamic development, it will be essential to understand how Shh regulates the downstream Gli transcription factors.

#### **Keywords: chick, development, Gli, hypothalamus, lineage, mouse, phenotype, Shh**

### **EXPRESSION PATTERN OF Shh IN THE DEVELOPING HYPOTHALAMUS**

Already the first studies characterizing mouse embryonic *Sonic Hedgehog* (*Shh*) expression showed that it is dynamic in the developing forebrain, a fact that immediately led to speculation as to the function of these domains (Echelard et al., 1993; Chang et al., 1994; Shimamura et al., 1995; Goodrich et al., 1996; Platt et al., 1997; **Figures 1A–D**). Based on these early studies and our own detailed hypothalamus expression analysis (Szabó et al., 2009; Alvarez-Bolado et al., 2012), we divide the expression of *Shh* in the mouse neural tube into distinct patterns (**Figures 1A–D**). Similar patterns are found in chick (Dale et al., 1997; Ohyama et al., 2004, 2008; Placzek and Briscoe, 2005; Manning et al., 2006) and zebrafish (Barth and Wilson, 1995; Mathieu et al., 2002) embryos. Note that the dates in embryonic days (E) are approximate, and, for some events, differences of up to 1 day can be found between studies. The influence of Shh on the hypothalamus starts with the onset of Shh expression in the underlying head process (**Figures 1A,E**, early pattern) (Aoto et al., 2009) (chick HH stage [st] 4). This is accompanied by the almost immediate onset of *Gli1* expression in the overlying neural ectoderm, starting at E7.5 (**Figure 1E**; Hui et al., 1994). Since *Gli1* expression is diagnostic of Shh pathway activation (Goodrich et al., 1996; Marigo and Tabin, 1996; Marigo et al., 1996; Lee et al., 1997), *Gli1* expression indicates that Shh signaling is involved in specifiying this neuroectodermal region. At E8.5, the neuroectodermal cells in the ventral midline start to express Shh (i.e., neuroepithelial *Shh* makes its appearance) (chick st 7–10; zebrafish 5 somites) while *Gli1* expression is downregulated medially (Christ et al., 2012; **Figure 1F**). Later, *Shh* expression is downregulated in the ventral midline of the basal hypothalamus and *Shh* is expressed in two domains bilaterally to the midline (**Figure 1G**, late pattern) (chick st 15 and later; zebrafish 22–28 somites). We still lack detailed studies of *Gli1* expression in the alar and basal hypothalamus at E9.5–E10.5. The scarce expression data found in the literature (Furimsky and Wallace, 2006; Aoto et al., 2009) are imprecise in terms of hypothalamic region as well as plane of section. In chick, the off-medial *Shh*-expressing cells migrate anteriorly from the diencephalic/mesencephalic junction and start to express Shh once they reach their final position in the hypothalamic primordium (Manning et al., 2006). It is important to note the expression of Shh is restricted to the

**FIGURE 1 | Shh expression and lineage during forebrain development**. **(A)** Shh expression in the prechordal plate seen on a transverse section of the head folds of an E8.0 embryo. pm, prechordal mesoderm. **(B–D)** In situ hybridization detection of Shh on wild type mouse brains. In **(C)** and **(D)**, the heads have been saggitally halved to show the neuroepithelium-ventricular zone on the inner side of the brain; rostral to the left. **(B)** Continuous Shh expression domain in the ventral forebrain at E9.0. **(C)** Shh expression is downregulated in the basal hypothalamus at E9.5–E10.0. **(D)** Shh hypothalamic expression at E12.5. **(E–G)** Diagrams of Shh (black) and Gli1 (pink) expression domains in the presumptive hypothalamus at three embryonic ages (only two for Gli1), together with their progeny represented in schemas of transverse sections through the hypothalamus (middle, blue in **(F)**; late, yellow in **(G)**). (Right side panels in **(F)**, **(G)**, are from Alvarez-Bolado et al., 2012). Abbreviations: ARH, arcuate nucleus; DMH, dorsomedial nucleus; LH, lateral hypothalamic area; ME, median eminence; mth, mammillothalamic tract; ne, neuroectoderm; nf, neural folds; nt, neural tube; pm, prechordal mesoderm; VMH, ventromedial nucleus.

ventricular zone, i.e., differentiated neurons or glia cells cease to express Shh. From here on, we will refer to the spatialtemporal classification of Shh expression described above to explain the results of lineage tracing and gene inactivation studies.

#### **LINEAGE OF Shh-EXPRESSING AND -RESPONDING PROGENITORS IN THE HYPOTHALAMUS**

*Shh*-expressing cells in the ventral midline (floor plate) of the spinal cord and hindbrain appear to function solely as organizers, i.e., they induce and pattern neighboring precursors but do not contribute progeny to the developing neural tissue (Joksimovic et al., 2009). In contrast, *Shh*-expressing cells in the mesencephalic floor plate give rise to neurons (Joksimovic et al., 2009; Blaess et al., 2011). To investigate whether *Shh*expressing cells contribute to the hypothalamus and whether precursors in the medial vs. lateral *Shh*-expressing domain give rise to different hypothalamic regions, we used the *Shh-CreER* mouse line in combination with Cre-inducible reporters for an inducible genetic fate mapping approach (Alvarez-Bolado et al., 2012; **Figures 1F,G**). We mapped the distribution of fatemapped cells in late embryonic and adult brains after tamoxifen administration (TM) (to activate Cre and induce reporter gene expression) at selected stages between E7.5 and E12.5. Based on these data, we were able to characterize the changing lineages that are derived from *Shh*-expressing cells. The timing of these changes is largely consistent with the early, middle and late patterns of *Shh* expression. We found that initially (TM at E7.5, corresponding to early pattern), very few cells are derived from *Shh*-expressing precursors, all of them localized to the posterior hypothalamus. With TM at E8.5 (corresponding to middle pattern), the progeny of *Shh*-expressing precursors contribute to the mammillary and tuberal regions. At later stages (TM at E9.5, E10.5, E11.5 or E12.5; corresponding to late pattern), *Shh*-expressing precursors give rise to cells in the preoptic area and in the posterior hypothalamic anlage, but to only few cells in the mammillary nucleus and ventral midline of the tuberal region. Precursors labeled with TM at E12.5 generate relatively few labeled cells and most of these have a morphology indicative of astrocytes. Interestingly, *Shh*-expressing precursors do not contribute any progeny to the anterior hypothalamic region at any of the labeling stages.

The telencephalic domain of *Shh* expression corresponds to the preoptic area and the medial ganglionic emminence. *Shh*expressing precursors in the medial ganglionic eminence give rise to the globus pallidus (Flandin et al., 2010; Nóbrega-Pereira et al., 2010) while the *Shh*-expressing domain in the preoptic area generates septal neurons (Wei et al., 2012) and preoptic astrocytes (Alvarez-Bolado et al., 2012). Lineage tracing of the lateral *Shh*-expressing precursor domain in the chick show that one of the neuronal populations generated from this domain are hypothalamic dopaminergic neurons (Ohyama et al., 2005).

A detailed characterization of the lineage of Shh-responding (*Gli1*-expressing) cells in the hypothalamus is lacking. Aoto et al. (2009) used genetic inducible fate mapping with a *Gli1CreER* line to investigate the lineage of *Gli1*-expression in craniofacial tissues, but the authors also provide some results on the tuberal hypothalamus. *Gli1*-expressing precursors labeled with TM at E8.5 give rise to progeny in the mediolateral tuberal hypothalamus (except for the midline). At later stages (TM at E9.5 and E11.5), *Gli1*-expressing precursors do not contribute to the tuberal hypothalamus. More anterior or posterior areas of the hypothalamus or the fate of *Gli1*-expressing cells before E8.5 or after E11.5 have not been assessed in this study.

### **PHENOTYPES OF MOUSE MUTANTS LACKING Shh EXPRESSION IN THE FOREBRAIN/FUNCTION OF Shh SIGNALING IN THE DEVELOPMENT OF THE BASAL HYPOTHALAMUS**

The phenotype of the *Shh* knock-out mouse mutant was published in 1996 (Chiang et al., 1996) and famously showed the loss of the ventral portion of the neural tube. The hypothesis that Shh was a key ventralizer, initially based on the *Shh* expression domain, was confirmed. The loss of ventral midline structures leads to a particularly severe forebrain phenotype: *Shh* knockout mice have a single fused telencephalic vesicle and a single fused optic cup (holoprosencephaly) (Chiang et al., 1996). Studies in chicken uncovered a key role of Shh (together with other determinants like bone morphogenetic proteins (BMPs)) secreted by the prechordal plate in hypothalamic induction and specifically in the specification of the hypothalamic ventral midline (Dale et al., 1997; Pera and Kessel, 1997).The dissection between the effects of Shh secreted by the axial mesoderm in the mouse however (notochord and prechordal plate; nonneural Shh) or by the neural tube itself (neural Shh) started more than a decade later. Analysis of *Shh* knock-out embryos and embryos chimeric for *Shh* knock-out cells showed that Shh signaling is non-cell autonomously required to maintain the prechordal mesoderm. In the absence of Shh, but also after lesions of the prechordal mesoderm, the forebrain midline does not develop (Aoto et al., 2009), indicating that Shh signaling from the prechordal mesoderm is essential to induce forebrain midline structures. To investigate the role of Shh secreted from the hypothalamic primordium, several conditional knockout mouse models were established. Using a *Shh* floxed allele (Dassule et al., 2000; Lewis et al., 2004) and a *Foxb1-Cre* mouse line (Zhao et al., 2007), which drives Cre expression in the forebrain neuroepithelium starting around E8.0, Szabó et al. (2009) generated a neuroepithelial specific inactivation of Shh in the hypothalamic primordium. Analysis of these conditional *Shh* mutants showed that neural Shh is required for the specification of the lateral hypothalamus and in particular of the hypocretin/orexin neurons. The medial portion of the basal hypothalamus (tuberal and mammillary regions) was severely reduced in size and showed specification defects (Szabó et al., 2009).

*Nkx2-1* is a specific marker of the hypothalamic primordium starting to be expressed at E8.0 (Shimamura et al., 1995). Using a *Nkx2-1-Cre* mouse line (Xu et al., 2008) to inactivate *Shh*, Shimogori et al. (2010) generated a second conditional mutant mouse model lacking *Shh* expression in the hypothalamic primordium. The phenotype of these mutants confirmed the requirement of neural Shh for normal hypothalamic growth and for the specification of the basal portion; alterations in lateral hypothalamic markers were also found. In addition, the developmental transcriptome of the mouse hypothalamus was analyzed with microarrays and was used to generate a genomic atlas resource for the study of hypothalamus development (Shimogori et al., 2010).

Zhao et al. (2012) inactivated *Shh* in the hypothalamic neuroepithelium using the SBE2-Cre mouse line. SBE2 is a hypothalamus-specific upstream regulatory element of *Shh* (Jeong et al., 2006). The results of the mutant analysis confirmed previous results (Szabó et al., 2009; Shimogori et al., 2010). Additionally, the study demonstrated that deficiency in neural Shh causes septo-optic dysplasia, a congenital brain anomaly that leads to defects in the pituitary, the optic nerve, and the midline of the forebrain.

### **THE Shh-Gli CODE IN THE DEVELOPMENT OF THE HYPOTHALAMUS**

Shh signaling is transduced by the two transmembrane receptors Patched (Ptch) and Smoothened and the Gli zinc finger transcription factors (Gli1-3). In the presence of Shh, Gli2 acts as a strong activator in the pathway. Gli3 acts primarily as a repressor of Shh target genes; in presence of Shh the Gli3 repressor function is attenuated and Gli3 can even function as a weak activator. Gli1 (**Figures 1E–G**, right panels, pink) contributes to the activation of the pathway, but it is only expressed in cells in which Gli2 (or Gli3) activator is already present and, as described above, can be used as a readout for the pathway (Fuccillo et al., 2006). In the spinal cord, midbrain and telencephalon, the dissection of the relative contribution of Gli activator and repressor function downstream of Shh have given considerable insight into the role of Shh signaling in the specification of these regions. The analysis of this Shh-Gli code in the hypothalamus is still rudimentary. Although *Shh* is essential for maintenance of prechordal plate and induction of ventral midline (Aoto et al., 2009), analysis of *Gli2* null mutants indicates that *Shh* expression in the hypothalamus is independent of Shh signaling from underlying prechordal mesoderm (Matise et al., 1998). In *Gli2* knock-out mutants, the bilateral Shh expression domain normally observed after E9.5 (**Figure 1G**) is fused to a single midline domain in the basal hypothalamus at E11.5, indicating that Gli2-mediated Shh signaling is required for the induction of the ventral midline, but not for the induction of *Shh* expression in the tuberal hypothalamic neuroepithelium *per se* (Park et al., 2000). *Nkx2-1* expression is reduced and shifted medially in these mutants. Additional inactivation of *Gli1* on the *Gli2* knock-out background results in a more severe phenotype; the *Nkx2-1* and *Shh* expression domains are completely missing and the size of the tuberal hypothalamic primordium is severely reduced. *Gli1* appears to be primarily important for the partial rescue of *Gli2* loss-of-function: mice in which *Gli1* is inactivated and which are heterozygous for *Gli2* do not have an obvious phenotype in the tuberal hypothalamus (Park et al., 2000).

Evidence that *Gli3* plays a role in hypothalamic development comes from a human malformation syndrome. Pallister-Hall syndrome is associated with several malformations including polydactyly, imperforated anus and hypothalamic hamartomas (a non-cancerous tumor in the tuberal hypothalamus), which develop typically during early gestation (33–41 days) (Clarren et al., 1980; Hall et al., 1980). The underlying mutation in the *Gli3* gene results in the production of a truncated form of Gli3 protein acting as a constitutive repressor (Meyer and Roelink, 2003) indicating that the precise regulation of Gli3 repressor levels is essential for normal hypothalamic development in humans. However, mice generated to have this mutation in *Gli3* do not develop hypothalamic hamartomas, even though they have most of the other malformations characteristic of the Pallister-Hall syndrome (Böse et al., 2002). Evidence that Shh-signaling mediated regulation of Gli3 repressor levels is essential for the induction of ventral forebrain structures comes from the analysis of *Shh* knock-out and *Shh/Gli3* double mutant mice: in addition to the loss of ventral midline structures in *Shh* knock-out mice (*Nkx2-1* positive) ventrolateral domains (*Dlx2* and *Gsh2* positive), are severely reduced. If *Gli3* is inactivated in addition to *Shh*, all three markers are expressed in their normal ventral domains (Rallu et al., 2002b). Whether defects in hypothalamic development in *Shh* knock-out mice can be rescued by removal of *Gli3* has not been investigated. The effects of *Gli3* loss-of-function on hypothalamic development in mouse have not been assessed directly, but histological sections through the diencephalon point to the absence of a severe phenotype in *Gli3* null mutants (Theil et al., 1999). In chick, the antagonistic functions of Gli3 activator and repressor have been analyzed by means of their effect on *Pax7* expression in the developing hypothalamus. *Pax7* is expressed in the lateral hypothalamus, but only at stage 30. Gli3 repressor induces *Pax7* expression, while Gli activator inhibits *Pax7* expression. *Gli3* expression is upregulated by BMP7 and in addition to Gli3 repressor, activation of BMP signaling is necessary for the induction of the *Pax7* positive cell fate (Ohyama et al., 2008).

### **NEURAL Shh IN THE PREOPTIC AREA AND alar HYPOTHALAMUS**

The preoptic area is classically assigned to the hypothalamus, according to its adult functionality, but embryologically it develops from the telencephalon (Puelles et al., 2012). Both this region and the classical "anterior" (or postoptic) region belong to the alar portion of the forebrain. *Shh* is expressed in this region of the neural tube as an isolated domain corresponding to the medial ganglionic eminence and preoptic area. These regions generate neurons that migrate to other areas of the telencephalon (Flandin et al., 2010; Nóbrega-Pereira et al., 2010; Wei et al., 2012) as well as local preoptic astrocytes (Alvarez-Bolado et al., 2012). The neural-*Shh*-specific mutants mentioned above do not show strong phenotypes in the anterior portions of the hypothalamus beyond reduced growth and mild patterning defects (Szabó et al., 2009; Shimogori et al., 2010) and the phenotype of these anterior regions has not been investigated in *Gli* mutant mice (Theil et al., 1999; Park et al., 2000). Evidence for an important role of non-neural Shh in the development of the preoptic area was provided by a recent study that identified Lrp2, a member of the LDL receptor gene family, as a component of the Shh signaling pathway important for the induction of the preoptic ventral midline (Christ et al., 2012). *Lrp2* is expressed broadly in the forebrain starting at E7.5, but expression gets restricted to the ventral midline by E9.5. Lrp2 sequesters secreted Shh protein and forms a complex with Ptch that regulates internalization and intracellular trafficking of Shh. In absence of *Lrp2*, the ventral tissue in the preoptic region fails to respond to Shh secreted from the prechordal mesoderm and medial *Shh* expression is not induced. *Shh* expression is shifted to two off-medial domains and *Bmp4* and *Gli3* are expressed in the ventral midline of the preoptic area in E10.5 *Lrp2* knock-out mice. While Shh signaling is obviously required for the development of the preoptic area and alar hypothalamus (see also Pabst et al., 2000; Rallu et al., 2002a,b), we consider that the precise requirements for neural Shh in these regions have been insufficiently analyzed.

#### **REGULATION OF Shh EXPRESSION IN THE HYPOTHALAMUS**

The dynamic expression pattern of Shh during hypothalamic development raises the question on how this pattern is regulated. The transcription factor Six3 regulates *Shh* expression directly, and haploinsufficiency in *Six3* results in the loss of *Shh* expression in the ventral midline, but only in the alar portion of the hypothalamus (Geng et al., 2008; Jeong et al., 2008; **Figures 2A,B**). Similarly, Sox2 has been shown to bind directly to the SBE2 enhancer and Sox2 and Sox3 have dose dependent effects on *Shh* expression in the midline of the alar hypothalamus (Zhao et al., 2012). Whether there is a similar transcription factor code required to induce *Shh* expression in the ventral midline of the basal hypothalamus is unknown, but the mechanisms for the later downregulation of *Shh* expression in the midline have been investigated. In chick, BMPs act upstream of the transcription factor Tbx2 to inhibit *Shh* expression in the ventral midline of the basal hypothalamus, allowing this region to acquire its fate as a precursor area (Manning et al., 2006). In zebrafish, Wnt inhibition induces hypothalamic cell fate (Kapsimali et al., 2004) and *in vitro* experiments indicate that, in chick, BMPs act probably by antagonizing Wnt signaling in this region (Manning et al., 2006). A transcriptional mechanism that integrates Shh and Wnt signaling to regulate ventral specification has been suggested before in the spinal cord (Lei et al., 2006). In mouse, Tbx3 appears to be required to downregulate *Shh* in the ventral midline and both Tbx2 and Tbx3 have the ability to suppress *Shh* expression by sequestering the transcription factor Sox2 away from the SBE2 forebrain enhancer (Trowe et al., 2013).

#### **CONCLUSION**

Ever since the first *Shh* knockout mouse (Chiang et al., 1996), our knowledge of the precise roles of Shh, prechordal or neuroepithelial, on hypothalamic specification and growth has steadily increased. Experimental approaches in chick and the phenotypical analysis of a variety of zebrafish and mouse mutants have demonstrated that prechordal Shh is necessary for midline specification, while neuroepithelial Shh plays a role in hypothalamic growth and in the specification of the lateral hypothalamic area. In addition to the functional insights, lineage tracing of neuroepithelial Shh in the hypothalamus has provided further important clues on hypothalamic development. Areas still in need of work are among others the specific Shh-Gli code for the different hypothalamic regions as well as elucidating the role of Shh on the alar hypothalamus (anterior hypothalamic region) and the preoptic area (a telencephalic area functionally included in the adult hypothalamus).

#### **REFERENCES**


globus pallidus but few neocortical interneurons. *J. Neurosci.* 30, 2812–2823. doi: 10.1523/JNEUROSCI.4228-09.2010


**Conflict of Interest Statement**: The Guest Associate Editor Valery Grinevich declares that, despite being affiliated to the same institution as authors Roberta Haddad-Tóvolli, Xunlei Zhou and Gonzalo Álvarez-Bolado, the review process was handled objectively and no conflict of interest exists. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 31 October 2014; paper pending published: 19 November 2014; accepted: 02 December 2014; published online: 06 January 2015*.

*Citation: Blaess S, Szabó N, Haddad-Tóvolli R, Zhou X and Álvarez-Bolado G (2015) Sonic hedgehog signaling in the development of the mouse hypothalamus. Front. Neuroanat. 8:156. doi: 10.3389/fnana.2014.00156*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Blaess, Szabó, Haddad-Tóvolli, Zhou and Álvarez-Bolado. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Differential requirements for *Gli2* and *Gli3* in the regional specification of the mouse hypothalamus

#### *Roberta Haddad-Tóvolli1 , Fabian A. Paul2, Yuanfeng Zhang1, Xunlei Zhou1, Thomas Theil3, Luis Puelles4,5\*, Sandra Blaess2\* and Gonzalo Alvarez-Bolado1\**

*<sup>1</sup> Department of Medical Cell Biology and Neuroanatomy, University of Heidelberg, Heidelberg, Germany, <sup>2</sup> Laboratory of Neurodevelopmental Genetics, Institute of Reconstructive Neurobiology, Life and Brain Center, University of Bonn, Bonn, Germany, <sup>3</sup> Centre for Integrative Physiology, University of Edinburgh, Edinburgh, UK, <sup>4</sup> Department of Morphology, Instituto Murciano de Investigación Biosanitaria, School of Medicine, University of Murcia, Murcia, Spain, <sup>5</sup> Facultad de Medicina, University of Murcia, Murcia, Spain*

Secreted protein Sonic hedgehog (Shh) ventralizes the neural tube by modulating the crucial balance between activating and repressing functions (GliA, GliR) of transcription factors *Gli2* and *Gli3*. This balance—the Shh-Gli code—is species- and contextdependent and has been elucidated for the mouse spinal cord. The hypothalamus, a forebrain region regulating vital functions like homeostasis and hormone secretion, shows dynamic and intricate *Shh* expression as well as complex regional differentiation. Here we asked if particular combinations of *Gli2* and *Gli3* and of GliA and GliR functions contribute to the variety of hypothalamic regions, i.e., we wanted to approach the question of a possible hypothalamic version of the Shh-Gli code. Based on mouse mutant analysis, we show that: (1) hypothalamic regional heterogeneity is based in part on differentially stringent requirements for *Gli2* or *Gli3*; (2) another source of diversity are differential requirements for Shh of neural vs. non-neural origin; (3) the medial progenitor domain known to depend on *Gli2* for its development generates several essential hypothalamic nuclei plus the pituitary and median eminence; (4) the suppression of Gli3R by neural and non-neural Shh is essential for hypothalamic specification. Finally, we have mapped our results on a recent model which considers the hypothalamus as a transverse region with alar and basal portions. Our data confirm the model and are explained by it.

#### Keywords: embryo, *Gli1, Gli2, Gli3*, hypothalamus, mouse, mutant, Shh

### Introduction

Sonic hedgehog (Shh) is a morphogen required for ventral neural tube specification (Echelard et al., 1993; Ericson et al., 1995, 1997; Chiang et al., 1996). Shh acts through the Gli transcriptional activators (GliAs) and repressors (GliRs); the balance between GliA and GliR specifies ventral differentiation and proliferation (Lee et al., 1997; Ruiz i Altaba, 1997; Brewster et al., 1998). This "Shh-Gli code" is known for the mouse spinal cord [reviewed in Ruiz i Altaba et al. (2003), Dessaud et al. (2008), Ingham et al. (2011)] and brainstem (Wang et al., 1995; Blaess et al., 2006, 2011; Feijoo et al., 2011).

The Shh expression domain in the forebrain is more extensive and elaborate than in the spinal cord, has become more intricate and dynamic during phylogenesis and is considered a motor of

#### *Edited by:*

*Agustín González, Universidad Complutense de Madrid, Spain*

#### *Reviewed by:*

*Kenji Shimamura, Kumamoto University, Japan Pierre-Yves Risold, Université de Franche-Comté, France*

#### *\*Correspondence:*

*Gonzalo Alvarez-Bolado, Department of Medical Cell Biology and Neuroanatomy, University of Heidelberg, Im Neuenheimer Feld 307, Abt. Neuroanatomie (3.OG), 69120 Heidelberg, Germany alvarez@ana.uni-heidelberg.de;*

#### *Sandra Blaess,*

*Laboratory of Neurodevelopmental Genetics, Institute of Reconstructive Neurobiology, Life and Brain Center, University of Bonn, Sigmund Freud Street 25, 53127 Bonn, Germany sandra.blaess@uni-bonn.de;*

*Luis Puelles, Department of Morphology, Facultad de Medicina, Instituto Murciano de Investigación Biosanitaria, School of Medicine, University of Murcia, Campus de Espinardo, 30100 Murcia, Spain puelles@um.es*

#### *Received: 16 January 2015 Paper pending published: 02 February 2015*

*Accepted: 09 March 2015 Published: 25 March 2015*

#### *Citation:*

*Haddad-Tóvolli R, Paul FA, Zhang Y, Zhou X, Theil T, Puelles L, Blaess S and Alvarez-Bolado G (2015) Differential requirements for Gli2 and Gli3 in the regional specification of the mouse hypothalamus. Front. Neuroanat. 9:34. doi: 10.3389/fnana.2015.00034* brain evolution (Osorio et al., 2005). Work on thalamic development supports the notion that regional variation of the Shh-Gli code underlies forebrain complexity (Haddad-Tovolli et al., 2012). In the same way, the canonical Shh-Gli code shows interspecies variation (Ruiz i Altaba, 1997; Aberger and Ruiz, 2014). On the basis of mutant phenotype analysis at different rostrocaudal levels of the mouse spinal cord and the hindbrain it has been proposed that *Gli2* and *Gli3* have partially overlapping roles and that their relative contributions to ventral specification shows regional variation (Motoyama et al., 2003; Lebel et al., 2007).

The hypothalamus regulates homeostasis, endocrine secretion, and reproductive behavior (Saper, 2006; Puelles et al., 2012; Sternson, 2013) and its alterations can cause conditions like obesity and high blood pressure (Caqueret et al., 2005; McMillen et al., 2008). Complex gene expression pattern combinations underlie hypothalamic regional specification (Shimogori et al., 2010; Puelles et al., 2012). On the basis of classical neuroanatomy studies, the adult hypothalamus has been traditionally described as subdivided into four regions (preoptic, anterior, tuberal, and mamillary) arranged rostro-caudally and ventrally in the brain (**Figure 1A**) and flanked by the lateral hypothalamic area (LHA), a large region essential to regulate behavioral state and arousal (Swanson, 1987). The modern view considers the adult hypothalamus as part of a behavioral control column (Swanson, 2000).

preoptic; ANT, anterior; TUB, tuberal; MAM, mamillary. (B) Model of the hypothalamus considering Shh expression (pink) as basal (ventral) marker. The POA is part of the telencephalon; the alar hypothalamus (yellow) corresponds to the anterior region; the tuberal and mamillary regions are not "caudal" but basal (ventral). ac, anterior commissure; hp, hypophysis; PTh, prethalamus; ZLI, zona limitans.

Here we analyze the hypothalamic phenotypes of mouse mutants to ascertain which combinations of GliA and GliR specify the mouse hypothalamic regions and which Gli genes perform these functions. We examined embryos primarily after midgestation so that we could assess which hypothalamic nuclei are affected when the GliA/R code is affected. We have mapped our results on a model of the developing hypothalamus (Puelles et al., 2012; **Figure 1B**) built around the observation that, since *Shh* is indispensable to ventralize the neural tube, and it is expressed in a long domain stretching the entire length of it, it follows that during development the *Shh* expression boundary separates dorsal (alar) from ventral (basal). The rostral end of the developing neural tube is closed by a transverse structure called acroterminal region, which does not share the typical characteristics of the floor plate. Beyond mamillary level, the acroterminal region extends all the way through the tuberal region, alar hypothalamus and preoptic region and up to the anterior commissure, it is transversally oriented (has alar and basal portions) and strongly patterned (probably by the underlying prechordal plate) and would generate the median eminence, infundibulum, neurophypophysis, preoptic terminal lamina, eyes, optic chiasma, and suprachiasmatic area. Two progenitor domains, medial and lateral, give rise to the basal part of the hypothalamus: the medial domain generates median eminence and neurohypophysis, medial portions of the ventromedial and arcuate nuclei, and the mamillary body; the lateral originates most of the ventromedial nucleus, the dorsomedial nucleus and the LHA (Alvarez-Bolado et al., 2012).

We show that, in the basal hypothalamus, the medial progenitor domain requires non-neural Shh acting through Gli2A. The lateral progenitor domain is patterned by neural Shh acting through Gli3R and Gli2A or Gli3A. In the presence of Shh signaling, the Gli3R function is not required for hypothalamus specification. Neither *Gli2* nor *Gli3* are required for overall patterning of the alar hypothalamus and preoptic area. These data confirm the main tenets of the model (Puelles et al., 2012), since they strongly support a subdivision of the developing hypothalamus into alar and basal domains.

## Materials and Methods

### Mice and Mouse Lines

Animals were housed and handled in ways that minimize pain and discomfort, in accordance with German animal welfare regulations (TierSchG) and in agreement with the European Communities Council Directive (2010/63/EU). The authorization for the experiments was granted by the Regierungspräsidium Karlsruhe (state authorities) and the experiments were performed under surveillance of the Animal Welfare Officer responsible for the Institute of Anatomy and Cell Biology. To obtain embryos, timed-pregnant females were sacrificed by cervical dislocation; the embryos were decapitated.

### *Gli2*zfd/+ (*Gli2* Zinc Finger-Deleted) Mutant Mice

This *Gli2* null mutant mouse line was generated (Mo et al., 1997) by replacing the exons encoding for zinc fingers 3–5. The deletion leads to an out-of-frame mutation causing disrupted transcription from the deletion site to the 3 end of the *Gli2* gene. This results in translation of a truncated protein unable to bind to DNA, since the zinc fingers 4 and 5 are essential for DNA binding (Pavletich and Pabo, 1993). The *Gli2zfd*/*zdf* are null mutants for *Gli2*; the *Gli2zfd*/<sup>+</sup> have normal phenotypes and are used as controls.

#### *Gli3Xt-J/+* (Extra-Toes) Mutant Mice

This line carries a 50 kb deletion that removes the exons encoding zinc fingers 3–5 and the complete 3 part of the *Gli3* gene (Hui and Joyner, 1993; Maynard et al., 2002; Genestine et al., 2007). The *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* are null mutants for *Gli3*.

We have not been able to obtain double *Gli2-Gli3* mutant embryos (*n* = 4 litters, 1 at E9.5, 2 at E10.5, and 1 at E12.5).

### *Gli3-nlacZ* Mutant Mice

The *Gli3-nlacZ* knock-in mouse line was generated by partially replacing the first coding exon of *Gli3* with the nlacZ cDNA. Thus, expression of lacZ is controlled by the endogenous *Gli3* promoter/enhancer elements and can be used to monitor the expression pattern of *Gli3* (Garcia et al., 2010).

#### *Foxb1-Cre* Mutant Mice

Express Cre in the thalamic and hypothalamic neuroepithelium (Zhao et al., 2007, 2008). We used only heterozygous *Foxb1-Cre* mice, which show a normal phenotype (Zhao et al., 2007, 2008), *Foxb1 Cre/Cre* homozygotes were not used in this study.

#### *Foxb1-Cre;Shhf/+* Mutant Mice

To obtain mice specifically deficient in Shh expressed in the neural tube (conditional knock-out for neural *Shh)*, we crossed our *Foxb1-Cre* mice (Zhao et al., 2007, 2008) with *Shhf* /<sup>+</sup> conditional mutants in which exon 2 of the *Shh* gene was flanked by loxP sites (Dassule et al., 2000; Lewis et al., 2001). The *Shhf* /<sup>+</sup> conditional mutants were generated in the laboratory of Dr. Andrew McMahon (University of Harvard) and were obtained through Jackson Labs (www.jax.org). The *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants lack all Shh expression in the forebrain neuroepithelium (Szabo et al., 2009a,b).

### *Foxb1-Cre;Shhf/+;Gli3Xt-J/+* Mutant Mice

The double homozygous mutants for *Shh* expressed in the neural tube (neural *Shh* or n*Shh*) and *Gli3* were generated by crossings between *Foxb1Cre;Shhf* /<sup>+</sup> mice; and *Gli3Xt*−*J*/<sup>+</sup> mice. The double mutants (*Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/ *Xt*−*<sup>J</sup>* ) do not survive beyond birth.

#### *In Situ* Hybridization

Embryos or embryonic brains were dissected, fixed in 4% paraformaldehyde, and embedded in paraffin. Non-radioactive ISH was performed on paraffin sections (7 μm for E10.5, 10 μm for E12.5, and 14 μm for E18.5 embryos) that were fixed in 4% paraformaldehyde and acetylated after sectioning. RNA *in situ* hybridization was performed as described (Blaess et al., 2011).

#### BrdU Labeling

Pregnant mice (E12.5) from appropriate crossings were injected intraperitoneally with 5- -bromo-2- -deoxyuridine (BrdU; Sigma; 50 μg/g body weight) at 12:00 h. Three hours after the injection, embryos were collected and fixed overnight in 4% PFA in PBS at 4◦C. Cell proliferation was detected by means of antibody (rat anti-BrdU; AbCam; 1:100) after epitope retrival in Tris-EDTA buffer pH = 9.0 for 20 min in pressure cooker. The nuclear marker 4- 6-diamidino-2-phenylindole dihydrochloride (DAPI; Invitrogen) was used as a counterstain. For cell counting, 10 μm paraffin sections were analyzed under a confocal microscope (LSM700 -Zeiss) and DAPI and BrdU-positive cells were counted in 100-μm-wide bins encompassing the thickness of the neuroepithelium (apical to basal side) at four hypothalamic sites (preoptic area, alar hypothalamus and tuberal and mamillary regions) on two histological sections per level in three animals per age and genotype (WT, *Gli2zfd*/*zfd*, *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* , and *Foxb1-Cre;Shhf* /+*;Gli3Xt*−*J*/<sup>+</sup> double mutants). The BrdUlabeling index (BrdU-labeled cells as percentage of total cells) was then calculated (Takahashi et al., 1993; Warren et al., 1999; Ishibashi and McMahon, 2002).

### Cloning of Constructs

In an expression vector driven by pCAGGS (Niwa et al., 1991) we inserted either EmGFP (kind gift of Dr. Boris Fehse, University of Hamburg; Weber et al., 2010) or tdTomato (kind gift of Dr. Roger Y. Tsien, UCSD) as reporters. On vectors carrying the tdTomato reporter we then inserted (upstream an internal ribosomal entry site and the reporter) a mutated form of human PTCH1 in which we deleted part (between MfeI and NsiI) of the second large extracellular loop (PTCH1---loop2), as was done in Briscoe et al. (2001).

#### *In Utero* Electroporation

This procedure was carried out as described (Saito and Nakatsuji, 2001; Saito, 2006; Haddad-Tovolli et al., 2012) with added caveats for hypothalamus targeting (Haddad-Tovolli et al., 2013). Pregnant mice at E12.5 were anesthetized with a mixture of Halothane (Isoflurane, Baxter) and oxygen (0.5 l/min) administered with a Komesaroff Anaesthetic Machine. The uterus was exposed and the DNA solution (1 μg/μl) was injected with a glass micropipette in the third ventricle of the embryo brain through the uterine wall. Electric pulses were administered with a CUY21 electroporator (Nepagene, Japan; 5 square-wave pulses, 50 V, 50 ms on/950 ms off) and a stainless steel needle electrode (CUY550-10) used as positive pole and a round flat electrode (CUY700P4L) as negative pole. After the surgery, the embryos were allowed to develop *in utero* for 6 days and collected at E18.5 for analysis. The embryonic brains were dissected, fixed overnight in 4% PFA in PBS at 4◦C and then protected with sucrose (20; 30%) and embedded in OCT mounting medium (Tissue Tek). Blocks were sectioned into 20 μm thick sections in a cryostat (Leica CM3050S) and observed and photographed with a Zeiss LSM 700 confocal microscope. We used laser line 488 nm for the green reporter EmGFP (excitation maximum 487 nm, emission maximum 509 nm) and laser line 555 nm for the red reporter tdTomato (excitation maximum 554, emission maximum 581). Since our readouts are based in the comparison between numbers of cells counted on confocal images (see next paragraph), it was imperative to relie on strictly comparable data. To guarantee comparability, the images of experimental and control brains were obtained under the exact same conditions and with the exact same confocal settings.

### Experimental Design of *In Utero* Electoporation Experiments

Because each *in utero* electroporation experiment results in a different number of neuroepithelial cells being transfected, the experiments are not directly comparable with each other. For this reason we do "two-reporter-experiments" (Haddad-Tovolli et al., 2012). The two reporters answer two problems. The green reporter construct (GFP) is an internal control. It will label every one of the transfected neuroepithelial cells and their progeny. In this way, for each single electroporated mouse embryo we know how many cells have been transfected. The second question is the actual scientific question: "does *Ptch1-delta-loop2* reduce proliferation?" For this, we have a second construct expressing a dominant loss-of-function version of the Ptch1 receptor (Ptch1 delta-loop2, see above Cloning of constructs) and, in the same construct, a red reporter (tandem dimer tomato, tdTomato). We use a ratio of 2 (GFP):1 (Ptch1-delta-loop2+red) in order to introduce some bias in the results, so that the readout of the experiment is the ratio between green cells and red cells. In principle there must be, after electroporation, a very few cells which are only green: they happen to express only GFP (not Ptch1-delta-loop2+red), proliferate normally and generate numerous green neurons, otherwise presumably normal. If the cells coexpressing the green plus the red (= experimental) constructs proliferate less, we will see less green + red neurons.

In parallel, we performed control experiments transfecting a 2:1 mixture of GFP construct and tdTomato construct (without loss-of-function Ptch1 protein) in order to evaluate how many only green and how many green-plus-red neurons we obtain in normal circumstances (i.e., without introducing any dominant loss-of-function). Those are the gray bars in **Figure 10J**. Additionally, these control experiments remove a possible concern related to the relative brightness of the green and the red reporters. In principle, a green cell could have been transfected also with some red (experimental) constructs in a number to small to be detected (since EmGFP is brighter than tdTomato). This possible source of imprecision can be disregarded since our readout is not absolute but relative (comparison between gray bars and black bars; **Figure 10J**).

#### Statistics

Statistical assessment of the BrdU and electroporation data was performed with Prism 6 software (Graph Pad Software, San Diego, CA, USA).

#### Morphological Interpretive Model

The results of mutant analysis were interpreted and mapped using the updated prosomeric model (Puelles et al., 2012) and the Allen Brain Atlas (Allen-Institute-for-Brain-Science, 2009).

## Results

### Developing Hypothalamic Expression of *Shh* and *Gli* can be Broadly Subdivided into at least Three Stages

Our purpose was to determine for each of the mouse hypothalamic regions which member of the Gli family performs the GliA and which one the GliR function, and which combinations of GliA and GliR specify these regions—in short, the hypothalamic Shh-Gli code. The expression of *Gli1, Gli3,* and Shh has been assessed at several stages in the developing chick hypothalamus, but in mouse the data are less comprehensive (Aoto et al., 2002; Ohyama et al., 2008). Thus, the first requisite for our study was to ascertain a detailed spatial-temporal expression map for the three mammalian *Gli* genes and *Shh* in the developing hypothalamus of the mouse (**Figures 2 and 3**). Although inactivation of *Gli1* does not result in an abnormal phenotype (Park et al., 2000; Bai et al., 2002), *Gli1* expression is a readout for Shh signaling [see references in Lewis et al. (2001)] and for this reason it was important to analyze its expression domain too. It has been described that, in the mouse neural plate, expression of *Gli* genes is first detected at E7.5 (neural fold); in this early stage of *Gli* expression, *Gli1* is expressed only in the midline of the neural fold, while *Gli2* and *Gli3* expression is widespread in the entire ectoderm (Hui et al., 1994) and *Shh* is expressed in the underlying mesoderm (non-neural Shh; Echelard et al., 1993).

We started our investigation of *Gli* expression patterns after neurulation, when they become more complex and at the same time more relevant to our study. At E8.5 (middle stage; **Figure 2**), *Gli1* and *Gli2* were expressed in overlapping patterns in the lateral domain (Alvarez-Bolado et al., 2012; **Figures 2A,B**), while *Gli3* was expressed in a more peripheral, non-hypothalamic domain (**Figure 2C**) and *Shh* was expressed in the medial domain (neural *Shh,* medial expression; **Figure 2D**), in coincidence with the medial progenitor domain (Alvarez-Bolado et al., 2012). The presumptive hypothalamus was defined by expression of specific marker *Nkx2-1* (**Figures 2E,J**).

At E10.5 (late stage; **Figure 3**) the Gli expression pattern had changed again. While *Gli2* expression was absent from the hypothalamic primordium (**Figures 3G–I**), *Gli3* and Shhactivation diagnostic marker *Gli1* showed overlapping expression domains in the medial domain (**Figures 3A–C,M–O**), suggesting a potential activator function of *Gli3* (Gli3A) in the midline at this age. *Shh* was expressed in a lateral domain corresponding to the lateral progenitor domain (Alvarez-Bolado et al., 2012; neural *Shh*, lateral expression). We concluded that the hypothalamic expression of *Shh* and the *Gli* genes can be broadly subdivided into at least three stages (summarized in **Figure 11A**).

### Deficiency in *Gli2* or *Gli3* does not Alter the Overall Specification of the Alar Hypothalamus

Sonic hedgehog is required to specify hypothalamic structures and the preoptic area (Chiang et al., 1996; Pabst et al., 2000; Rallu et al., 2002). In mouse mutants lacking *Shh* expression in the neural tube (*Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants), however, the preoptic and

alar hypothalamus have only a moderate phenotype, mostly evident in their reduced size (Szabo et al., 2009a; Zhao et al., 2012), indicating that they are specified by Shh of non-neural origin (e.g., from the prechordal plate or the notochord). Here we asked what is the role of Gli factors in those two hypothalamic regions by analyzing mutants in which *Gli2, Gli3* or both neural *Shh* and *Gli3* were inactivated. Expression of transcription factor *Nkx2-1,* an early preoptic marker (Shimamura et al., 1995; Xu et al., 2008), was preserved in the *Gli2zfd*/*zfd* and *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutants (black arrowheads in **Figures 4A–C**).Incidentally, some non-preoptic telencephalic expression domains were missing in the mutants

(white arrowheads in **Figures 4A–D**). *Arginin-Vasopressin (Avp)* is specifically expressed by the supraoptic and paraventricular nuclei (Swanson and Sawchenko, 1983) and shows robust expression in both mutants (**Figures 4E–G**). The transcription factor gene *Lhx1* is a marker of the suprachiasmatic nucleus (Szabo et al., 2009a), and this pattern remains essentially unchanged in the mutants (**Figures 4I–K**). Finally, analysis of double *Foxb1- Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutants (lacking both neural *Shh* and *Gli3*) showed robust marker expression (**Figures 4D,H,L**).

These results could indicate that, downstream of Shh of nonneural origin at the early stage, *Gli2* and *Gli3* can fully substitute for each other's activator function in the alar portions of the hypothalamus or, alternatively, that the specification of the alar hypothalamus depends on suppression of Gli3R by non-neural *Shh* (Rallu et al., 2002).

### *Gli2* is Required for the Development of Medial Tuberal and Mamillary Regions

In order to analyze the *Gli2zfd*/*zfd* phenotype in the basal hypothalamus (tuberal and mamillary regions), we examined expression of *Shh* and *Gli* genes as well as regional markers at E8.5 and E10.5. At E8.5, expression of *Gli1*, *Gli3* and the regional marker *Nkx2- <sup>1</sup>* was not changed in the *Gli2zfd*/*zfd* mutant (**Figures 2F,H–J**), except of course for the disappearance of the *Gli2* domain (**Figure 2G**). Expression of Shh, however, seemed expanded (**Figure 2I**). At E10.5, expression of *Gli1* and *Gli3* was strongly downregulated in the midline around the infundibular area (red arrowheads in **Figures 3D,E,P,Q**) in the *Gli2zfd*/*zfd* mutants. At mamillary levels, however, the two lateral expression domains seem to have fused in a thickened midline [**Figures 3F,R**; this is also true of the expression of the truncated (inactive) form of *Gli2* in the mutant (**Figure 3L**)]. At this age, *Shh* expression is normally downregulated in the medial domain of the tuberal region (Manning et al., 2006; arrow in **Figure 5A**). In the *Gli2zfd*/*zfd* mutant this *Shh*-negative domain was absent (arrow in **Figure 5B**). *Nkx2-1,* a transcription factor gene defining regional specification of the basal hypothalamus (Kimura et al., 1996; Puelles et al., 2004, 2012), was expressed in an appropriate but smaller domain, with stronger expression shifted into the medial domain (**Figures 5C,D**). *Six3* is a transcription factor required for initiation of hypothalamic specification (Kobayashi et al., 2002). It is normally expressed strongly along the entire medial domain and flanking hypothalamus, except the mamillary part. *Six3* expression was severely reduced at both the infundibular (**Figures 5E,F**) and median eminence levels (**Figures 5G,H**). Together with the alterations in gene expression, we observed again a thickening of the medial domain of the tuberal region (arrowheads in **Figures 5F,H**). Analysis of *Six3* expression on sagittal sections at E12.5 (**Figures 5I,J**) confirmed *Six3* downregulation and a thickened medial domain of the *Gli2zfd*/*zfd* mutant (arrowheads in **Figures 5I,J**). Since expression of Six3 (**Figures 5E,F**) indicated alterations of the infundibulum, which is essential for pituitary development, we then examined the expression of appropriate gene markers for this region (**Figure 6**). Infundibular expression of *Tbx2* (Manning et al., 2006) and *Fgf8* (Ericson et al., 1998; **Figures 6A–D**), as well as expression of pituitary markers *Lhx3* (**Figures 6E,F**), and *Pitx2* **(Figures 6 G,H)**

hypothalamus of E10.5 *Gli2zfd*/<sup>+</sup> and *Gli2zfd*/*zfd* mouse embryos as indicated. For each gene, three levels are shown, from pituitary/infundibulum (top row)

(D,E,P,Q) show downregulation of *Gli1* and *Gli3* in the *Gli2zfd*/*zfd* midline. Scale bars, 200 μm.

was completely lost in the *Gli2zfd*/*zfd* presumptive hypothalamus at E10.5 (see also Park et al., 2000). These results indicated that *Gli2* is required for appropriate development of the medial domain in the basal hypothalamus and for the development of the neurohypophysis.

### Arcuate, Ventromedial, and Mamillary Nuclei are Severely Reduced in Size in the *Gli2zfd/zfd* Mutant

We next analyzed the differentiation of the tuberal and mamillary regions in *Gli2zfd*/*zfd* brains at E18.5 (at this stage, characteristic neuronal nuclei are recognizable in the wildtype). *Npy*-expressing and *Pomc*-expressing neurons are specifically present in the arcuate nucleus (tuberal region; Elias et al., 1998; **Figures 7A,C**). In the *Gli2zfd*/*zfd* brain, the arcuate nuclei were not preserved as two distinct left and right domains. Instead, one single specifically labeled area was observed, unpaired and medial, sitting in the midline at the level of the tuberal area (dashed circle in **Figures 7B,D**). The third ventricle was abnormally absent at the site of this unpaired structure. Expression of *SF-1* (nuclear receptor *Nr5a1*) specifically labels the ventromedial nucleus of the hypothalamus (Ikeda et al., 1995; **Figure 7E**). In the *Gli2zfd*/*zfd* brain, *SF-1* was expressed in a median, unpaired group of cells (dashed circle in **Figure 7F**). The transcription factor *Nkx2-1* is specifically expressed in the lateral part of the wildtype ventromedial nucleus (Nakamura et al., 2001; **Figure 7G**), but formed one single medial domain in the *Gli2zfd*/*zfd* brain (**Figure 7H**). The transcription factor genes *Lhx1*, *Otp,* and *Sim1* are specifically expressed in the mamillary body (mamillary region) in the wildtype (Szabo et al., 2009a) but this expression was completely lost in the *Gli2zfd*/*zfd* mutant (**Figures 7I–N**). Together with the observations shown in **Figures 5** and **6,** these results indicate that *Gli2* is essential for the specification of the medial progenitor domain (Alvarez-Bolado et al., 2012) of the basal hypothalamus. The *Gli2zfd*/*zfd* mutant mice showed an altered latero-medial organization of the molecular pattern of the basal hypothalamus consistent with a loss of the medial markers (notably reduced *Six3* and loss of *Tbx2*, *Otp*, *Sim1,* and *Lhx1*) and derivatives (median eminence and neurohypophysis, mamillary body). The latter were substituted at the mutant midline by markers and derivatives typical of the lateral domain at this age, like *Nkx2.1, Npy, Pomc,* and *SF-1*. That the neurohypophysis is a derivative from this region has been described before (Pearson et al., 2011; Pearson and Placzek, 2013).

### *Gli2* in the Medial Progenitor Domain

In the early presumptive hypothalamus (E7.5 to E8.5), an unpaired medial progenitor domain ("med" in **Figures 2A** and **11A**) is specified, which gives rise to medially located nuclei like most of the arcuate nucleus, the medial portion of the ventromedial nucleus, the median eminence and the mamillary body (Alvarez-Bolado et al., 2012; derivatives of the acroterminal

region, Puelles et al., 2012). This early domain and its lineage are strongly affected in the *Gli2zfd*/*zfd* mutant (**Figures 5–7**). *Gli2* expression overlaps with *Gli1* in the medial domain at E7.5 (Hui and Joyner, 1993). Since *Gli1* expression is diagnostic of Shh pathway activation, this indicates a Gli2A function. The strong *Gli2zfd*/*zfd* midline phenotype must be due to a requirement for *Gli2* expression in the medial domain at E7.5, since this domain does not show *Gli2* expression at later stages (**Figures 2** and **3**). Moreover, expression of *Gli1* and *Gli3,* normally absent from the midline at E8.5 (**Figures 2A,C**), is not ectopically upregulated in the *Gli2zfd*/*zfd* mutant (**Figures 2F,H**; i.e., no rescue). At E10.5, *Gli3,* and *Gli1* expression overlap in the midline (black arrowheads in **Figures 3A–C,M–O**) suggesting an activator role of Gli3 (Gli3A). However, both genes are strongly downregulated in the *Gli2zfd*/*zfd* midline at E10.5 (red arrowheads in **Figures 3D,E,P,Q**), again making a rescue of the *Gli2zfd*/*zfd* phenotype by a Gli3A function impossible.

### No Abnormal Phenotype in the *Gli3Xt-J/Xt-J* Basal Hypothalamus

We went on to analyze the developing *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* basal hypothalamus. At E18.5, expression of specific marker genes *Npy*, *Pomc* and *SF-1* in the tuberal region (**Figures 8A–F**) and of *Lhx1* in the mamillary region (**Figures 8M,N**), showed that a Gli3A function in presence of Gli2A is dispensable for the specification of the basal hypothalamus.

We next addressed the question of a possible Gli3R function in the developing hypothalamus. Gli3R function often results in negative regulation of tissue growth (by inducing cell death and reducing proliferation) and counteracting the ventralizing influence of Shh (Persson et al., 2002; Ruiz i Altaba et al., 2003). Therefore, loss of Gli3R could result in ventralization and/or an increased size of hypothalamic nuclei. Since Shh signaling counteracts the processing of Gli3 protein into its repressor form, experimental abolition of Shh signaling might result in overabundance of Gli3R. We investigated this possibility by analyzing mouse mutants lacking *Shh* expression in the neural tube (*Foxb1- Cre;Shhf* /*<sup>f</sup>* mutant, Szabo et al., 2009a). This mutant showed strong downregulation of three tuberal marker genes (*Npy*, *Pomc*, *SF-1*; **Figures 8G–I**) and had a more severe phenotype than the *Gli2zfd*/*zfd* mutants (**Figures 7B,D,F,H**; see also Szabo et al., 2009a; Shimogori et al., 2010). Given that Gli2A acts primarily at early stage of hypothalamic induction (see above) and following the logic of the Shh-Gli code (Bai et al., 2004), a possible explanation of this difference could be that, in the absence of neural *Shh*, formation of Gli3R is not prevented. A prediction of this hypothesis is that, in the absence of both neural *Shh* and *Gli3* the

tuberal region would have a less marked phenotype. We tested this prediction by analyzing mutants deficient not only in neural *Shh* but also in *Gli3* (*Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutants). These showed essentially correct marker expression in the arcuate and ventromedial nuclei (although the expression domains appeared somewhat reduced and distorted; **Figures 8J–L**) suggesting that the phenotype in *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants is at least partially caused by upregulated Gli3R activity. The same reasoning applies to the mamillary body (mamillary region), which is extremely reduced in the *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutant (Szabo et al., 2009a) but appears normal in the *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutant and in *Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* double mutants (**Figures 8M–O**). We conclude that *Gli3* is dispensable for overall hypothalamic specification. Moreover, it is likely that the upregulation of Gli3R is the main contributor to the defects in the tuberal and mamillary hypothalamus when neural *Shh* is inactivated, which would be consistent with the classical Shh-Gli code in the spinal cord (Bai et al., 2004).

### A Possible *Gli3* Activator Function in the Lateral Hypothalamic Area of the *Gli2zfd/zfd* Mutant

The LHA is a large and morphologically complex region and with key functions in the regulation of behavioral state and arousal mechanisms [reviewed in Swanson (2000)]. Analysis of *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants has shown that expression of *Shh* by the forebrain is essential for its specification (Szabo et al., 2009a). *Hypocretin/orexin (Hcrt;* Hungs and Mignot, 2001; **Figures 9A–D**) and *pro-melanin-concentrating hormone (Pmch;* Croizier et al., 2013; **Figures 9E–H**)*,* essential modulators of metabolism and behavior, are among the very few specific marker genes of restricted groups of LHA neurons. The *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>*

brain did not show changes in *Hcrt* (**Figure 9C**) or *Pmch* expression (**Figure 9G**), indicating that Gli3A is normally not involved in the specification of the LHA. In the *Gli2zfd*/*zfd* mutant mice, only a few scattered *Hcrt*-expressing cells were present, and they were displaced toward the midline from their normal lateral position (arrowheads in **Figure 9B**). The number of *Pmch*expressing neurons in the *Gli2zfd*/*zfd* mutant seemed not altered, but the cells tended to gather in the midline, similar to *Hcrt* cells (**Figure 9F**). This indicates that *Gli2* is dispensable for the generation of *Pmch*-expressing cells, their altered position being rather a phenotypic consequence of the missing medial domain in this mutant (**Figures 5** and **6**). The phenotype of *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants in this area (*Hcrt* cells are absent, and *Pmch* cells severely reduced Szabo et al., 2009a) is stronger that that of *Gli2zfd*/*zfd* mutants. We went on to address the possibility that a compensatory Gli3A function could explain the relatively mild LHA phenotype of *Gli2zfd*/*zfd* mutant mice. To test this hypothesis, we examined double mutants deficient in neural *Shh* and *Gli3 (Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* ) and found a phenotype similar to that of the *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants (*Hcrt* cells absent, *Pmch* cells severely reduced; **Figures 9D,H**), but more pronounced than that of *Gli2zfd*/*zfd* mutants (**Figures 9B,F**). This indicates that, in the LHA of the *Gli2zfd*/*zfd* brain, Gli3A might compensate for the loss of Gli2A. This would be consistent with *Gli1* still being expressed in the lateral domain of *Gli2zfd*/*zfd* mutants at E8.5 (**Figure 2E**).

We concluded that, for the specification of the LHA progenitors within the lateral progenitor domain, a Gli2A function is needed which can be partially substituted for by Gli3A.

### *Gli3* in Mamillary Neurogenesis

Differences in size can be due to quantitative changes in precursor generation (symmetric cell divisions) at an early stage or to later changes in neuron generation (asymmetric cell divisions). Shh is essential for the expansion of neural precursors in the early development of this region (Rowitch et al., 1999; Ishibashi and McMahon, 2002).

Here, we wanted to address the contribution of Shh-Gli to the neurogenesis of hypothalamic nuclei.

*In situ* analysis of *Gli* family expression at E12.5 (**Figures 10A–C**) showed *Gli1* and *Gli3* (arrowheads in **Figures 10A,C**) expression in the mamillary region. *Gli2* expression on the contrary was very low or absent (arrowhead in **Figure 10B**). In agreement, a *Gli3* reporter mouse line (*Gli3 nLacZ* knock-in) showed strong beta-galactosidase labeling in the mamillary region (arrowhead in **Figure 10D**).

Therefore, we labeled proliferating cells in the neural tube by injecting BrdU in pregnant mice at E12.5 [i.e., at the peak of neurogenesis in the mouse hypothalamus (Ishii and Bouret, 2012)] and collecting the embryonic brains 3 h later. Our results (**Figure 10E**) show that proliferation during the neurogenic period in the mamillary region was reduced in the *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutant*. Gli1* expression is diagnostic of *Shh* pathway activation, and can be found in the mamillary region at E10.5 (**Figure 10F**, upper panel), but it is lost in the *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutant (**Figure 10F**, lower panel), indicating failure of this domain to activate the *Shh* pathway in the mutant, in agreement with the reduced proliferation. BrdU labeling in this region was also reduced in the *Gli2zfd*/*zfd* mutant (**Figure 10E**), which we

nucleus; VMH, ventromedial nucleus. Scale bars, 500 μm. (O) The proposed hypothalamic model (see Figure 1B) showing in black the area affected by the *Gli2* null mutation. Black arrows indicate the arcuate nucleus (A–D) or the ventromedial nucleus (E,F,H). Dashed circles in (B,D,E–G,I,K,M) outline hypothalamic nuclei (as indicated).

interpret as a consequence of the defect in midline development in these mutants (**Figures 5** and **6**). Intriguingly, deficiency in both neural *Shh* and *Gli3* increased neurogenesis, particularly in the mamillary region (**Figure 10E**; see Discussion).

We then approached this issue experimentally by specifically blocking the Shh pathway in the hypothalamus of mouse embryos developing *in utero*. We electroporated wildtype embryos at E12.5 (**Figures 10G–I**) with *EGFP*-expressing reporter constructs mixed with constructs expressing the loss-of-function Shh receptor *Ptch1-*Δ*-loop2* plus the red fluorescent reporter *tdTomato* [similar to the one used by Briscoe et al. (2001); see Materials and Methods]. The results show that cells expressing high levels of *Ptch1-delta-loop2* plus *tdTomato* in the mamillary region are less proliferative, while the same experiment did not alter neurogenesis in the alar hypothalamus (**Figure 10J**). We concluded that proliferation during the neurogenic period in the alar portion of the hypothalamus is not directly affected by Shh-Gli, while *Gli3* has a role as an activator inducing neurogenesis at later stages in the basal regions.

### Discussion

We asked which combinations of GliA and GliR specify different hypothalamic regions, and which members of the Gli family perform GliA or GliR functions. Therefore we interpreted mutant phenotypes with the help of Shh and Gli expression patterns (**Figure 11A**), hypothalamic progenitor domains (Alvarez-Bolado et al., 2012; **Figure 11B**), the Shh-Gli code established for the spinal cord (Bai et al., 2004) and a hypothalamic model (**Figure 1B**; Puelles et al., 2012). We uncovered strong differences in *Gli* gene requirements between alar and basal hypothalamus as predicted by the model. Null mutations of *Gli2* or *Gli3* do not alter the overall specification of the alar portions, including preoptic (actually telencephalic) and alar hypothalamus (or "anterior region"). In the basal regions (tuberal and mamillary), however, *Gli2* is indispensable for the development of the medial progenitor domain and its derivatives but it is partly dispensable for the lateral progenitor domain (**Figures 11B,D**). *Gli3* is dispensable for overall specification of the wildtype hypothalamus, but Gli3A has a late influence on mamillary proliferation. Finally, medial progenitor domain specification is dependent on Shh of non-neural source (prechordal plate, notochord), while the lateral progenitor domain is strongly dependent on neural Shh, in whose absence ectopic upregulation of Gli3R causes a severe phenotype. In this way, the notorious anatomical complexity of the hypothalamus depends on combinations of specification timing, progenitor domain, Shh source, *Gli* gene dependence, and alar vs. basal position (**Figure 11D**).

### Gli2A, Gli3A, and repression of Gli3R in hypothalamus specification

*Gli2zfd*/*zfd* mutants lack a floor plate and its flanking cells from spinal cord to midbrain (Matise et al., 1998; Park et al., 2000). We extend this result to the rostral end of the floor plate (the mamillary region Puelles et al., 2012), and beyond this point, through the medial progenitor domain of the entire basal hypothalamus. Therefore, in the *Gli2zfd*/*zfd* mutant, the median eminence, pituitary and mamillary body are missing as well as part of the arcuate and ventromedial nuclei (**Figure 11B**).

In spinal cord, *Gli3* has important repressor (Gli3R; Litingtung and Chiang, 2000; Persson et al., 2002) and, in the absence of Gli2A, activator (Gli3A) functions (Bai et al., 2004). The lack of a pronounced phenotype in the *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutant hypothalamus indicates that under normal conditions either Gli3A is not important or that Gli2A can substitute for Gli3A. However, in the *Gli2zfd*/*zfd* mutant, the lateral progenitor domain is still able to produce part of the ventromedial and the arcuate nuclei and also the LHA (with only minor alteration; **Figures 9B,F**) – this can either be explained through rescue by a Gli3A function or it indicates that GliA is not essential for the induction of the lateral progenitor domain. A partial compensation through Gli3A (asterisk in **Figure 11D**) is supported by *Gli1* expression in the lateral progenitor domain of the *Gli2zfd*/*zfd* mutants (**Figure 2F**), where

indicate the arcuate nucleus. Black arrows in L indicate the ventromedial

it partially overlaps with *Gli3* (**Figure 2H**). The suppression of Gli3R function by non-neural and neural Shh appears to be essential for the development of the hypothalamus (**Figure 11D**), since the phenotype of the ventral forebrain in *Shh* null mutants (Chiang et al., 1996) is more severe than in *Gli2zfd*/*zfd* mutants and the phenotype of the lateral progenitor domain in *Foxb1- Cre;Shhf* /*<sup>f</sup>* mutants (Szabo et al., 2009a) is more severe than in *Gli2zfd*/*zfd* mutants. In addition, the medial and lateral progenitor domains of *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants are largely rescued in *Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mice (**Figures 8G–I,J–L**). Finally, *Hcrt*<sup>+</sup> neurons (part of the LHA) are reduced in *Gli2zfd*/*zfd* mutants and completely lost in *Foxb1-Cre;Shhf* /*<sup>f</sup>* mutants and even in *Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* brains, suggesting that their progenitors are uniquely specified by GliA or need GliA

ventromedial nucleus. Scale bars 500 μm.

FIGURE 9 | *Gli* mutant phenotypes in the LHA. RNA *in situ* detection of lateral hypothalamic markers *Hcrt/Orexin* and *Pmch* on E18.5 mouse brain sections, genotypes as indicated. Arrowheads in (B,H), point at smaller groups of cells. Scale bars, 500 μm.

later or for a longer time than other progenitors in the lateral domain (**Figure 11D**). This is consistent with the very restricted place and time of neurogenesis of *Hcrt*+ neurons (Amiot et al., 2005). As mentioned in Section "Materials and Methods," we have not been able to obtain double mutant *Gli2- Gli3* embryos.

As for the alar hypothalamus and preoptic area, they are strongly dependent on Shh for their development, while *Gli2zfd*/*zfd* or *Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* show no – or only subtle – phenotypes in these areas. This indicates that either Gli2A and Gli3A can fully substitute for each other in these regions, or they depend on suppression of Gli3R rather than induction of GliA by *Shh* for their specification (Chiang et al., 1996; Rallu et al., 2002).

### The Hypothalamic Version of the Shh-Gli Code

In the spinal cord (**Figure 11C**), signaling by notochordal Shh is sufficient to generate the proper pattern of ventral progenitor gene expression (Matise et al., 1998; Jeong and McMahon, 2005), whereas ongoing Shh signaling from the floor plate (neural *Shh*) is necessary to maintain progenitor domain formation during neurogenesis (Dessaud et al., 2010) and for oligodendrocyte specification (Yu et al., 2013). We show that Shh of non-neural origin specifies the medial progenitor domain through Gli2A at an early stage, while neural Shh specifies the lateral progenitor domain at a later stage, probably by counteracting ectopic GliR function and, in the case of LHA progenitors, by inducing Gli2A (see above).

In both spinal cord (Bai et al., 2004; **Figure 11C**) and hypothalamus (**Figures 11B,D**), *Gli2* performs the main GliA function. *Gli2,* however, is required for the induction of *Shh* expression in the floor plate (Matise et al., 1998), but not in the hypothalamus (**Figure 2I**).

Opposite gradients of GliA and GliR underlie the precise dorsoventral polarity of the spinal cord (Litingtung and Chiang, 2000; Persson et al., 2002) and hypothalamic specification requires counteracting Gli3R by Shh. Additionally, in the chicken hypothalamus, Gli3R activity is involved in *Pax7* de-repression in some progenitors (Ohyama et al., 2008).

Finally, a Gli3A function is required for mamillary proliferation during the neurogenic phase (**Figure 10E**). The mamillary region overgrowth in *Foxb1-Cre;Shhf* /*<sup>f</sup> ;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutants (**Figure 10E**) parallels the abnormally increased size of the spinal cord in *Gli2zfd*/*zfd;Gli3Xt*−*J*/*Xt*−*<sup>J</sup>* mutants. It remains unclear why inactivation of all Shh signaling results in a proliferation increase (Bai et al., 2004).

#### Acroterminal Region vs. Floor Plate

We have mapped the *Gli2zfd*/*zfd* phenotype (**Figure 7O**) on a genetic-molecular model of the developing hypothalamus (Puelles et al., 2012) in which the ventral and dorsal midlines do not meet at a hypothetical "tip" of the neural tube (**Figure 1B**). Rather, the model proposes that the rostral end of the tube is closed by a "lid" in the form of a transverse structure called acroterminal region, which does not share the typical characteristics of the floor plate—e.g., it does not express *Foxa2* (Ruiz i Altaba et al., 1995; Dale et al., 1999), undergoes complex, specific regulation (Ohyama et al., 2005; Manning et al., 2006; Ohyama et al., 2008; Trowe et al., 2013) and, as we show here, it is strongly neurogenic, not a property of the floor plate (except in the midbrain Kittappa et al., 2007; Ono et al., 2007; Bonilla et al., 2008). The histologically recognizable floor plate expresses *Shh*, *Ntn1*, *Lmxb1*, *Foxa1*, and *Nr4a2* (Allen-Institute-for-Brain-Science, 2009; Puelles et al., 2012), is

an E12.5 *Gli3-nlacZ* knock-in mouse brain section; arrowhead indicates the mamillary region. (E) BrdU-labeled cells per bin at E12.5, genotypes as indicated. Unpaired *t*-test, two-tailed, mean ± SD; <sup>∗</sup>*p* ≤ 0.05, ∗∗*p* ≤ 0.01, ∗∗∗*p* ≤ 0.001. (F) *Gli1* expression in the medial domain of the hypothalamus (arrows) on E10.5 horizontal sections of WT (upper

GFP and Ptch-Δ-loop-tomato DNA constructs at E12.5. White arrowheads show double-labeled cells. (J) Percent of GFP-expressing cells co-expressing red reporter "tomato" after *in utero* electroporation of control (white bars) or experimental (black bars) constructs, in two different regions, as indicated. Unpaired *t*-test, two-tailed, mean ± SD; ∗∗*p* ≤ 0.01; n.s., non-significant.

induced by the underlying notochord, and it ends rostrally at mammillary level (Puelles et al., 2012). Beyond mammillary level, the acroterminal region extends all the way through the tuberal region, alar hypothalamus and preoptic region and up to the anterior commissure, it is transversally oriented (has alar and basal portions) and strongly patterned (probably by the underlying prechordal plate) and generates, among other,

the median eminence, infundibulum, neurophypophysis, and eyes.

The dorso-ventral and rostro-caudal axes of the embryonic neural tube, considered in this way, are at a 90◦ angle with those of the adult brain as they are usually considered; i.e., the adult rostro-caudal axis would be the dorso-ventral axis in our model. If this discrepancy will eventually be corrected remains

of the mouse. (A) Diagram showing expression domains of the Gli factors and *Shh* in the presumptive hypothalamus at the early, middle, and late phases. "lat" and "med," lateral and medial domains, respectively (early phase according to Hui et al., 1994). (B) Summary diagrams of progenitor domains (neuroepithelium) of the basal hypothalamus in WT and mutants as deduced from phenotype analysis in the present study. (C) Diagrams comparable to those in (B) representing the progenitor

literature (as indicated). Question marks indicate that the V0/V1 domains have not been investigated. (D) Specific contribution of Gli proteins to the specification of the medial and lateral progenitor domains in three successive stages of development. Dotted square, possible influence of Gli2A on lateral progenitors before E8.5. The asterisk (∗) means that loss of GliA2 could be compensated by Gli3A. MBO, mamillary body. See Discussion for details.

open. The connectivity and function of the classical regions of the hypothalamus and the behavioral control column (Swanson, 2000) are not otherwise challenged by the proposed nomenclature (Puelles et al., 2012).

### The *Gli2zfd/zfd* Phenotype and the Hypothalamic Model

The *Gli2zfd*/*zfd* hypothalamic phenotype can be cleanly mapped (**Figure 7O**) on the model of the embryonic hypothalamus (Puelles et al., 2012), which in turn receives experimental confirmation from our work. The basal regions depend specifically on *Gli2*. The alar hypothalamus and preoptic region, on the contrary, are not strictly dependent on Gli2A or Gli3A and are therefore genetically different. In this way, the basal/alar boundary, one main insight of the model, is confirmed. The basal part has unique genetic requirements, as much in the floor plate as in the acroterminal region, which are difficult to reconcile with a conventional rostral-caudal hypothalamic orientation (**Figure 1A**). Moreover, the medial and lateral progenitor domains of the basal hypothalamus (Alvarez-Bolado et al., 2012) can be mapped on the

### References


model too, corresponding to acroterminal and terminal hypothalamus (Puelles et al., 2012). Mapping other mutant phenotypes will refine the model and reveal fundamental aspects of brain development and organization. The patterning of the acroterminal region by the prechordal plate, for instance, is an open question.

### Acknowledgments

This work was supported by the Deutsches Forschungsgemeinschaft Grant AL603/2-1 (to GAB); the North-Rhine-Westphalia Repatriation Program, Ministry for Innovation, Science and Research of North Rhine Westphalia (to SB); BFU2008-04156 (to LP); and Grant MR/K013750/1 from the Medical Research Council, MRC (to TT). RHT was the recipient of a DAAD Fellowship. We gratefully acknowledge the financial support of the Deutsche Forschungsgemeinschaft and Ruprecht-Karls-Universität Heidelberg within the funding programme Open Access Publishing.


the emergence of pattern in the embryonic anterior pituitary. *Development* 125, 1005–1015.


in *The Rat Nervous System*, ed. G. Paxinos (San Diego, CA: Academic Press), 3–25.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Haddad-Tóvolli, Paul, Zhang, Zhou, Theil, Puelles, Blaess and Alvarez-Bolado. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Notch signaling and proneural genes work together to control the neural building blocks for the initial scaffold in the hypothalamus

### **Michelle Ware, Houda Hamdi-Rozé and Valérie Dupé \***

Institut de Génétique et Développement de Rennes, Faculté de Médecine, CNRS UMR6290, Université de Rennes 1, Rennes, France

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Andrea Wizenmann, University of Tuebingen, Germany Pierre-Yves Risold, Université de Franche-Comté, France

#### **\*Correspondence:**

Valérie Dupé, Institut de Génétique et Développement de Rennes, Faculté de Médecine, CNRS UMR6290, Université de Rennes 1, IFR140 GFAS, 2 Avenue du Pr. Léon Bernard, 35043 Rennes Cedex, France e-mail: valerie.dupe@ univ-rennes1.fr

The vertebrate embryonic prosencephalon gives rise to the hypothalamus, which plays essential roles in sensory information processing as well as control of physiological homeostasis and behavior. While patterning of the hypothalamus has received much attention, initial neurogenesis in the developing hypothalamus has mostly been neglected. The first differentiating progenitor cells of the hypothalamus will give rise to neurons that form the nucleus of the tract of the postoptic commissure (nTPOC) and the nucleus of the mammillotegmental tract (nMTT). The formation of these neuronal populations has to be highly controlled both spatially and temporally as these tracts will form part of the ventral longitudinal tract (VLT) and act as a scaffold for later, follower axons. This review will cumulate and summarize the existing data available describing initial neurogenesis in the vertebrate hypothalamus. It is well-known that the Notch signaling pathway through the inhibition of proneural genes is a key regulator of neurogenesis in the vertebrate central nervous system. It has only recently been proposed that loss of Notch signaling in the developing chick embryo causes an increase in the number of neurons in the hypothalamus, highlighting an early function of the Notch pathway during hypothalamus formation. Further analysis in the chick and mouse hypothalamus confirms the expression of Notch components and Ascl1 before the appearance of the first differentiated neurons. Many newly identified proneural target genes were also found to be expressed during neuronal differentiation in the hypothalamus. Given the critical role that hypothalamic neural circuitry plays in maintaining homeostasis, it is particularly important to establish the targets downstream of this Notch/proneural network.

**Keywords: early axon scaffold, forebrain, differentiation, tract of the postoptic commissure, mammillotegmental tract, hypothalamus patterning, ASCL1**

#### **INTRODUCTION**

The hypothalamus is an evolutionary ancient structure in the rostral brain that plays a central role in the regulation of physiological processes such as hunger, thermoregulation, reproduction and behavior in adult vertebrates. The adult hypothalamus is subdivided into regions, each containing well documented clusters of neurons with defined functions (Simerly, 2004). Countless work involving physiological and genetic studies has focused on signaling molecules and transcription factors that control hypothalamus morphogenesis and the emergence of different neuronal subtypes (Shimogori et al., 2010). However, relatively little attention has been paid to the process through which the initial neurons are induced and specified in the primordium of the vertebrate hypothalamus, despite their key roles in pioneering the major axon pathways in the forebrain (Wilson et al., 1990; Mastick and Easter, 1996; Ware and Schubert, 2011). The first differentiating cells of the hypothalamus will give rise to neurons that form the nucleus of the tract of the postoptic commissure (nTPOC) and the nucleus of the mammillotegmental tract (nMTT). Recent advances in the chick model has established that a Notch/proneural regulatory loop is implicated very early during the differentiation of these neurons (Ratié et al., 2013). The aim of this review is to highlight a role for Notch signaling during nTPOC and nMTT differentiation; including key findings from zebrafish, chick and mouse models, which has contributed to our understanding of this field. A potential cascade involving *Ascl1* and target genes will be discussed to determine the possible regulation of these initial hypothalamic neurons.

#### **PATTERNING OF THE VERTEBRATE HYPOTHALAMIC PRIMORDIUM**

During early embryogenesis the hypothalamus develops within the secondary prosencephalon (Puelles and Rubenstein, 2003;

**Abbreviations:** AH, anterior hypothalamus; bHLH, basic helix-loop-helix; MH, mammillary hypothalamus; MTT, mammillotegmental tract; TH, tuberal hypothalamus; NPC, neural progenitor cell; nTPOC, nucleus of the tract of the postoptic commissure; nMTT, nucleus of mamillo-tegmental tract; TPOC, tract of the postoptic commissure.

Martinez-Ferre and Martinez, 2012; Puelles et al., 2012). Developmental studies performed in zebrafish, chick and mouse indicate Sonic Hedgehog (SHH), secreted by the underlying prechordal plate mesendoderm, induces the formation of the hypothalamus (Dale et al., 1997; Mathieu et al., 2002; Aoto et al., 2009). Loss of *Shh* leads to missing ventral structures including the hypothalamus in zebrafish (Varga et al., 2001) and mouse (Chiang et al., 1996). In humans, mutations in the *Shh* gene results in holoprosencephaly, the most frequent human brain malformation that includes hypothalamic defects (Mercier et al., 2011). However, SHH alone is not sufficient to induce specific hypothalamus identity. The prechordal plate expresses numerous other secreted proteins that are involved in the development of the overlying hypothalamus primordium including Wnt antagonists, NODAL and Bone Morophogenic Proteins (BMP; Pera and Kessel, 1997; Kiecker and Niehrs, 2001; Mathieu et al., 2002; Manning et al., 2006; Cavodeassi and Houart, 2012).

Specific patterning of the hypothalamus begins when the hypothalamic primordium expresses the transcription factor *Nkx2.1* from Hamburger and Hamilton stage (HH)8 in chick and embryonic day (E)8 in mouse (Shimamura et al., 1995; Pera and Kessel, 1998; Sussel et al., 1999; Crossley et al., 2001). This expression of *Nkx2.1*, along with *Nkx2.2* is dependent on the presence of *Shh* in the prechordal plate (Barth and Wilson, 1995; Pera and Kessel, 1997; Rohr et al., 2001; Mathieu et al., 2002). SHH is then required to coordinate tissue growth and acquisition of anteroposterior (AP), dorsoventral (DV) and mediolateral patterning of the hypothalamus (Manning et al., 2006; Szabó et al., 2009).

At HH10, *Shh*, *Nkx2.1* and *Nkx2.2* expression expands in the basal plate of the chick prosencephalon, with the same rostral expression at the level of the presumptive anterior hypothalamus (AH) that corresponds to the prospective chiasmatic area (also called suboptical domain) (Crossley et al., 2001). A new *Nkx2.1* expression domain develops at HH12, just rostral to the hypothalamus in the basal telencephalon called the postoptic area (POA). In zebrafish and mouse, the same dynamic expression patterns of *Shh, Nkx2.1* and *Nkx2.2* is present within the hypothalamus (**Figure 1**). By HH13 in chick and E9.5 in the mouse, *Shh* and *Nkx2.1* expression has expanded further and the hypothalamic primordium is morphologically evident. Studies in chick show that once the hypothalamic primordium is established, SHH down-regulation mediated by local production of BMPs is necessary for establishing region-specific transcriptional profiles (Patten and Placzek, 2002; Manning et al., 2006; Ohyama et al., 2008). This leads to the subdivisions of the primordial hypothalamus into three regions, the AH, the tuberal hypothalamus (TH) and the mammillary hypothalamus (MH), with each region expressing specific markers (**Figure 1**; Alvarez-Bolado et al., 2012; Wolf and Ryu, 2013).

### **INITIAL NEUROGENESIS IN THE VERTEBRATE HYPOTHALAMUS**

The first neurons that differentiate in the vertebrate brain give rise to the highly conserved early axon scaffold (Chitnis and Kuwada, 1990; Wilson et al., 1990; Easter et al., 1993; Mastick and Easter, 1996; Barreiro-Iglesias et al., 2008; Ware and Schubert, 2011; Ware et al., 2014). This is an important structure for the guidance of later, follower axons allowing more complex connections to form. Predating the mature hypothalamic neuronal clusters, two small GABAergic positive populations differentiate within the hypothalamus (**Figure 1**; Patel et al., 1994). The first neurons in the hypothalamic primordium differentiate to give rise to the nTPOC (also termed the ventro-rostral cluster (vrc) in anamniotes) at 16 hpf in zebrafish (**Figure 1A**; Chitnis and Kuwada, 1990; Ross et al., 1992) and HH13 in chick (**Figure 1B**; Ware and Schubert, 2011). An early birth-dating study has shown that hypothalamic neurogenesis in the mouse begins at E10 (Shimada and Nakamura, 1973). However, it is well-known that the initial nTPOC neurons arise at E9.5, suggesting neurogenesis begins earlier than previously thought (**Figure 1C**; Easter et al., 1993; Mastick and Easter, 1996; Ricaño-Cornejo et al., 2011). From the nTPOC neurons, axons extend and project caudally within the basal plate. The tract of the postoptic commissure (TPOC) axons project into the mesencephalon where these axons form part of the ventral longitudinal tract (VLT) along with the medial longitudinal fascicle (MLF) and later the mammillotegmental tract (MTT; Ware and Schubert, 2011). The MTT forms from a second set of neurons (nMTT) that differentiate later in the caudal hypothalamus of amniotes from HH14 in chick and E10 in mouse (**Figures 1E,F**; Puelles et al., 1987; Easter et al., 1993; Mastick and Easter, 1996). While the presence and location of the nTPOC is conserved in all vertebrates studied, the nMTT is not present in zebrafish, at least during early development (Barreiro-Iglesias et al., 2008). Neurons do however form later in the zebrafish MH, but it is not possible to comment on the homology with the nMTT (Wolf and Ryu, 2013). The postoptic commissure (POC) forms by 24 hpf, projecting axons from the nTPOC rostrally to form a commissure across the rostral midline connecting the left and right sides of the neural tube (**Figure 1D**; Ross et al., 1992; Bak and Fraser, 2003). The POC is likely to form in chick and mouse at later stages but has not been studied exhaustively (Croizier et al., 2011; Ware and Schubert, 2011).

The prosomeric model and hypothalamic markers such as *Shh*, *Nkx2.1* and *Nkx2.2* confirms the nTPOC and nMTT neurons form within the hypothalamus (**Figure 1**; Hjorth and Key, 2001; Puelles and Rubenstein, 2003). The nTPOC arises just below the optic stalk at the midline of the AH area and the nMTT in the lateral edge of the caudal hypothalamus in the MH (**Figure 1**; Easter et al., 1993). The nTPOC neurons differentiate along the boundaries of many gene expression areas in zebrafish (Macdonald et al., 1994), however the mechanism by which these neurons differentiate has been overlooked.

Some zebrafish and mouse mutants are available where the formation of these hypothalamic axon tracts is affected. Some genes are implicated in the differentiation of the neurons such as *Six3* (Ando et al., 2005), but many studies focus on the effect of gene inactivation on axon guidance, including *Fgf8* (Shanmugalingam et al., 2000), *Pax6* (Mastick et al., 1997; Nural and Mastick, 2004), Slits and Robos (Ricaño-Cornejo et al., 2011)

**FIGURE 1 | Organization of the hypothalamic primordium in the rostral vertebrate brain. (A–C)** The first nTPOC neurons arise in the hypothalamus in zebrafish **(A)** at 16 hpf, chick **(B)** at HH13 and mouse **(C)** at E9.5. **(D)** Axons project caudally from the nTPOC forming the TPOC and rostrally from the nTPOC to form the POC in zebrafish at 24 hpf. **(E, F)** The nTPOC neurons begin projecting axons and the first nMTT neurons arise in chick at HH14 **(E)** and in mouse **(F)** at E10. The hypothalamus is specifically marked by three genes, Nkx2.1 (light blue), Nkx2.2 (green) and Shh (red stripes). **(A, D)** Gene expression in zebrafish is based on the following studies: Shh, Nkx2.2 (Barth and Wilson, 1995; Hjorth and Key, 2001) and Nkx2.1a (Rohr and Concha, 2000; Rohr et al., 2001). **(B, E)** Gene expression in chick is based on the following studies: Shh (Bardet et al., 2010), Nkx2.1 (Ratié et al., 2013) and Nkx2.2 (Gimeno and Martinez, 2007). **(C, F)** Gene expression in mouse is based on the following studies: Shh (Shimamura et al., 1995; Alvarez-Bolado et al., 2012), Nkx2.1 and Nkx2.2 (Shimamura

et al., 1995). **(E, F)** The three subdivisions of the hypothalamus (AH, TH and MH) in chick and mouse is based on Shh expression (Alvarez-Bolado et al., 2012; Ratié et al., 2013). **(E)** Asterisk, Shh negative region, overlapping where the ventral MLF neurons differentiate. For all schematics, subdivisions of the brain is based on the prosomeric model (Mastick and Easter, 1996; Hauptmann et al., 2002; Puelles and Rubenstein, 2003; Ware and Schubert, 2011). **(D–F)** Zona limitans intrathalamica (ZLI) marks the p2/p3 boundary. Although other neuronal populations are present in the brain at these stages, they are not added to focus on the hypothalamic neurons. AH, anterior hypothalamus; mes, mesencephalon; MH, mammillary hypothalamus; nMTT, nucleus of the tract of the mammillotegmental tract; nTPOC, nucleus of the tract of the postoptic commissure; os, optic stalk; POA, postoptic area; POC, postoptic commissure; p1–p3, prosomeres 1–3; TPOC, tract of the postoptic commissure; TH, tuberal hypothalamus; tel, telencephalon.

and *Sim1/Sim2* (Marion et al., 2005). Functionally, the TPOC is important for the guidance of other axon tracts. Ablation of the TPOC axons in the zebrafish embryo affects the patterning of the early axon scaffold (Chitnis and Kuwada, 1991). In zebrafish *Cyclops* mutants, the TPOC does not form, leading to the misguidance of the tract of the posterior commissure

(TPC) axons (Patel et al., 1994). More recently, a study in the embryonic mouse has shown later hypothalamic axons from the melanin-concentrating hormone (MCH) neurons use the TPOC for guidance (Croizier et al., 2011). While the potential function of the TPOC neurons is not known, lypophilic tracing shows that the TPOC axons project into the hindbrain, although the target of these axons remains a mystery (Ware and Schubert, 2011). It is also unclear whether these neurons are still present postnatally, it could be that their sole purpose is to provide axons for guidance and then simply die after connections are made in the adult brain (Easter et al., 1993). The MTT may also function in the guidance of other tracts but this has not been studied exhaustively. In mouse, the MTT is likely to guide the mammillothalamic tract (MTH) that forms later in mouse contributing to the principle mammillary tract (Marion et al., 2005). The MTH is not known to form in zebrafish or chick. The MTT axons project to the tegmentum and are described as having a role in visceral function and processing special information in the adult human brain (Alpeeva and Makarenko, 2007; Kwon et al., 2011).

No attention has been brought to the mechanism by which the nTPOC and nMTT neurons differentiate, until 2013, when Notch components were first described as being present very early in the hypothalamus of the developing chick embryo (Ratié et al., 2013). A basic PubMed search of the key words Notch and hypothalamus generated very few publications and many of which are based in adult models or describe differentiation of late forming embryonic neurons (Chapouton et al., 2011; Aujla et al., 2013). This indicates a surprising lack of investigations surrounding neurogenesis of the initial hypothalamus neurons, when considering these early neurons have been described through-out the 1990s in different vertebrate species. As these neurons contribute to the early axon scaffold and are essential for the set-up of more complex connections, it is surely essential to understand how they differentiate and how they are specified. Finally, considering Notch along with the proneural network is a well-known signaling pathway, little is known about the implication of Notch signaling or neurogenic factors involved in the formation of the nTPOC and nMTT neurons.

#### **NEUROGENESIS AND THE NOTCH/PRONEURAL NETWORK**

Notch signaling is an evolutionary conserved signaling pathway involved in cell-cell communication regulating multiple processes throughout development. The Notch signaling pathway has previously been reviewed in detail, here a brief outline is described (Pierfelice et al., 2011). First identified in *Drosophila*, the Notch pathway has been confirmed to have similar roles in vertebrates (Coffman et al., 1990; Artavanis-Tsakonas and Simpson, 1991; Artavanis-Tsakonas et al., 1999). The core pathway consists of the interaction between a transmembrane Notch receptor anchored in one cell, with a transmembrane Notch ligand (Delta or Serrate/Jagged) in a neighboring cell. Upon receptor-ligand binding a series of proteolytic cleavages are triggered that releases the intracellular domain of Notch (NICD), which forms a nuclear complex with recombination signal binding protein for immunoglobulin kappa J region (RBPJ). This complex activates the transcription of target genes (Tamura et al., 1995; Fortini, 2009). The best characterized direct targets of the NICD/RBPJ complex are the Hes (Hairy-Enhancer of Split) and Hey (Hes related type) genes (Jarriault et al., 1995; Maier and Gessler, 2000). They are class-C basic helix-loop-helix (bHLH) proteins that function as transcriptional repressors and can function together as homodimers or heterodimers (Iso et al., 2003).

One function of Notch relies on lateral induction, which is defined as the process by which a ligand-expressing cell stimulates those cells nearby to upregulate ligand expression, promoting ligand propagation and coordinated cell behavior (Eddison et al., 2000). The other function of Notch is lateral inhibition, whereby a ligand-expressing cell inhibits the expression of the ligand in the neighboring cells, therefore preventing those cells from adopting the same fate and generating a patched cellular pattern (Bray, 2006). It is associated with salt-and-pepper like patterns of gene expression (Fior and Henrique, 2009). For example, these two modes of Notch pathway operation coexist during inner ear development. Each mode relies on an associated gene regulatory network (Kiernan, 2013; Neves et al., 2013). Expression and functional studies suggest that lateral induction and lateral inhibition are associated with different Notch ligands that initiate signaling (Brooker et al., 2006; Saravanamuthu et al., 2009; Petrovic et al., 2014). The association of DLL1 with lateral inhibition is a general theme during neural development (Henrique et al., 1995; Adam et al., 1998; Kageyama et al., 2010).

Notch signaling has a very well-known role in neurogenesis, controlling the balance between proliferation of neural progenitor cells (NPCs) and differentiation of NPCs into neuronal and glial cells (Campos-Ortega, 1993; Chitnis et al., 1995; de La Pompa et al., 1997; reviewed by Paridaen and Huttner, 2014). In the neuroepithelium, neuron production is mostly controlled by lateral inhibition, where a regulatory loop is formed, with proneural genes controlling the expression of Notch ligands (Bertrand et al., 2002). The ligand, DLL1, can bind and active NOTCH in neighboring cells. When the Notch signaling pathway is activated, transcriptional repressors (such as Hes or Hey genes) are expressed that prevent expression of proneural genes, inhibiting differentiation and therefore cells remain as progenitors. Cells expressing the ligand and therefore lacking Notch signaling can no longer express transcriptional repressors, leading to the upregulation of bHLH proneural transcription factors such as *Ascl1* or *Neurog1/2*. Under this Notch/proneural network the cell can exit the cell cycle and undergoes neural differentiation (Bertrand et al., 2002). This differentiation step is controlled by several classes of transcription factors that determine the identity of the neuron produced. Among them, a number of bHLH differentiation genes are switched on, such as *Nhlh1* or *NeuroD4*, followed by specific neuronal genes.

### **NOTCH SIGNALING IN THE VERTEBRATE HYPOTHALAMUS PRIMORDIUM**

There are numerous studies investigating the expression and function of Notch components, proneural genes and downstream targets. However, as mentioned previously there is very little data describing the role of Notch signaling during the differentiation of the nTPOC and nMTT neurons. When Notch signaling is inhibited in the developing chick embryo, the number of nTPOC neurons increases, along with ectopic expression of many genes within the hypothalamus, confirming Notch has a role during hypothalamic neurogenesis at this early stage (Ratié et al., 2013). This study describes a typical neurogenic phenotype expected for the loss of Notch function, working by lateral inhibition.

For the first time, the Notch components *Dll1*, *Hes5* and *Hey1* are shown to be expressed just before HH11 in the presumptive AH of the chick embryonic brain where the first nTPOC neurons will differentiate at HH13 (Ware and Schubert, 2011; Ratié et al., 2013). Expression of Notch components during initial neurogenesis in the zebrafish and mouse has been extensively studied, however for much of the data, it is difficult to interpret the expression in the hypothalamic primordium as no special attention was given to this area at early stages. The expression of Notch receptors in zebrafish are first described at 16 hpf in the prosencephalon and appear to overlap in the area where the nTPOC forms (Bierkamp and Campos-Ortega, 1993; Dyer et al., 2014). In mouse, while *Notch3* is ubiquitously expressed in the neuroectoderm from E8.0, *Notch2* and *Hes1* are expressed in the ventral prosencephalon from E8.5 and *Notch1* is expressed from E9.5 (Reaume et al., 1992; Williams et al., 1995; Koop et al., 1996). Remarkably, little information is present in the literature about when these genes are first expressed in the developing hypothalamus (Bettenhausen et al., 1995; de La Pompa et al., 1997; Leimeister et al., 1999; Barsi et al., 2005). Therefore, in this review, expression of Notch components are analyzed using *in situ* hybridization data to deal with this deficiency (**Figure 2**). *Dll1, Hes5* and *Hey1* mRNA probes are used to show the presence of Notch activity, focusing more specifically in the hypothalamus. At E8.0, *Dll1*, *Hes5* and *Hey1* are not expressed in the mouse presumptive hypothalamus (**Figures 2A–C**), it is only from E8.5, before the initial neurons differentiate that *Dll1* and *Hes5* expression is first observed (**Figures 2E,E',F,F'**, arrowheads). Flat-mounted preparations of the ventral midline reveal a salt-and-pepper like pattern for these genes in the rostral hypothalamus (**Figures 2E',F'**). Expression continues in the AH at E9 for *Dll1* and *Hes5* (**Figures 2I,J**, arrowhead), while *Hey1* expression first starts to be expressed in the same region (**Figure 2K**, arrowhead). At E9.5, *Dll1*, *Hes5* and *Hey1* are first expressed in the MH where the nMTT neurons will differentiate at E10 (**Figures 2M–O**, unfilled arrowhead). It is important to note that the genes analyzed here are not specific for either the nTPOC or nMTT, but are also expressed by other early developing neurons such as those in the nucleus of the mesencephalic tract of the trigeminal nerve (nmesV; **Figure 2**). This is not surprising as Notch is a very general pathway involved in neuron progenitor expansion (Kageyama et al., 2009). Flatmounted preparations performed at E9.5 confirm the localized expression of *Hes5* (**Figure 2N**) and *Hey1* (**Figure 2O**) in the AH.

Bringing together the data from the literature and *in situ* hybridization of mouse embryos presented here, this highlights that Notch signaling is active very early in the AH and MH where the nTPOC and nMTT neurons will develop respectively (Mastick and Easter, 1996; Ratié et al., 2013). The data also suggests that redundancy could be strong between the direct Notch target genes as multiple transcriptional repressors such as *Hes1*, *Hes5* and *Hey1* are expressed in the developing hypothalamus.

Like with the expression studies described in this section, no functional data about neurogenesis in the early hypothalamus is available in zebrafish and mouse. There are several models lacking Notch signaling, which exhibit an increase in neurons throughout the embryo (de La Pompa et al., 1997; Itoh et al., 2003). For example, in the zebrafish and mouse mindbomb/Mib1 mutants, *Dll1* ubiquitination is affected and aberrant neurogenesis due to lower expression of *Hes1* and *Hes5* is observed throughout the embryo (Itoh et al., 2003; Barsi et al., 2005; Koo et al., 2005). A similar phenotype is also present in RBPj mutant mice, where Notch activity is absent (Oka et al., 1995; de La Pompa et al., 1997). As all these mutant mice display early lethality, no description is available to indicate whether neurogenesis is disturbed in the hypothalamus. Conditional lossof-function mice lacking RBPJ, using *Nkx2.1*-Cre to specifically knock-out Notch signaling in the hypothalamus, shows that Notch signaling is essential for the differentiation of late arcuate hypothalamic neurons in the mouse from E13.5 (Aujla et al., 2013). This study did not identify a role in the initial neurons, but we would assume there would be an increase in the number of nTPOC and nMTT neurons in these mutant mice.

Many other knock-out or ectopic expression studies of Notch components describe an effect on neurogenesis throughout the embryo. The *Dll1* mutant mouse has not been well studied for a neurogenesis phenotype (Hrabe de Angelis et al., 1997; Przemeck et al., 2003). However, *Dll1* does regulate primary neurogenesis in the *Xenopus* embryo (Chitnis et al., 1995).

There appears to be much redundancy between genes of the Notch pathway, which could explain why a function for Notch during nTPOC neuronal differentiation has not been described before in the mouse. For example, *Hes5* does not show any phenotype in single mutants (Cau et al., 2000). Double or triple knock-out mice produce more obvious phenotypes and prove redundancy occurs between these genes (Hatakeyama et al., 2004; Kageyama et al., 2008a). The absence of both *Hes1* and *Hes5* leads to aberrant neuronal localization. Interestingly, expression of *Dll1* and *Ascl1* is highly upregulated in the ventral diencephalon of E9.5 *Hes1/Hes5* double mutants as are the number of βIII-tubulin (Tuj1) positive cells (Hatakeyama et al., 2004). The capacity of these bHLH proteins to do the same job, may also explain why there is discrepancy between their expressions in chick compared with mouse. For example, flat-mounted preparations of chick embryos at HH15 (**Figure 3A**, arrowhead) and HH14 (Ratié et al., 2013) confirm specific expression of *Hey1* in the AH, whereas expression is throughout the developing hypothalamus in the mouse (**Figures 2O,O'**).

### **PRONEURAL GENE EXPRESSION IN THE VERTEBRATE HYPOTHALAMUS PRIMORDIUM**

Induction of the Notch/proneural loop is essential in the developing hypothalamus as this will eventually lead to the correct number of cells differentiating into nTPOC and nMTT neurons as well as maintaining the progenitor population.

*Ascl1* is a well-studied proneural bHLH transcription factor, its expression and function during early embryogenesis has been well described in many vertebrates (Johnson et al., 1990; Ferreiro et al., 1993; Guillemot and Joyner, 1993; Jasoni et al., 1994; Mcnay et al., 2006). In zebrafish, *Ascl1* expression appears early

at 12 hpf in cells prior to the appearance of markers indicative of overt differentiation, by 16 hpf, *Ascl1* expression overlaps with the nTPOC (Allende and Weinberg, 1994; Ando et al., 2005). In the Notch inhibited chick model, embryos display an upregulation of *Ascl1* expression in the AH, overlapping the nTPOC (Ratié et al., 2013). Flat-mounted preparations of HH15 chick hypothalamus show that *Ascl1* expression overlaps with *Shh* expression in the AH (**Figure 3B**, arrowhead). The expression of *Ascl1* in the hypothalamus is examined further by *in situ* hybridization between E8 and E9.5 in the mouse embryo, like with the Notch components, expression of *Ascl1* has been badly interpreted in this region (**Figures 2D,H,L,P**). *Ascl1* starts to be expressed in the developing hypothalamus at E9.0 (**Figure 2L**, arrowhead). At E9.5, flat-mounted preparations indicate that *Ascl1* is specifically expressed in a salt-and-pepper like pattern in the AH (**Figures 2P,P'**, arrowhead) and in the MH (**Figures 2P,P'**, unfilled arrowhead). *Ascl1* is important for the differentiation of late hypothalamic neurons because in *Ascl1* knock-out mice differentiation of neuroendocrine neurons is disturbed (Mcnay et al., 2006). Although the authors did not specifically look at the

**FIGURE 3 | Hey1**, **Ascl1 and Tagln3 expression overlaps with Shh in the chick hypothalamus primordium**. Double labeling of Hey1, Ascl1 and Tagln3 (Purple, Digoxigenin labeled probes) at HH15 with the dynamic hypothalamic marker, Shh (Red, Fluorescein labeled mRNA probe) in chick confirms expression of these genes in the hypothalamic domains. **(A)** Hey1 expression is specifically expressed in the anterior hypothalamus (AH) (arrowhead), overlapping the area where the nTPOC neurons will differentiate. Dashed lines indicate the boundaries of the hypothalamic domains, while the solid line marks the diencephalic-mesencephalic boundary (DMB). **(B)** Ascl1 expression is located in the AH (arrowhead). Asterisk labels the nTPOC located in p1 (Ware and Schubert, 2011). **(C)** Tagln3 expression is located in the AH (arrowhead) and in the mammillary hypothalamus (MH), overlapping where the nMTT neurons differentiate (unfilled arrowhead). Asterisk labels the ventral medial longitudinal fascicle (nMLF) located in p2 (Ware and Schubert, 2011). mes, mesencephalon; p1-p3, prosomeres 1–3; TH; tuberal hypothalamus.

nTPOC or nMTT neurons it can be assumed these neurons will be affected.

During initiation of neuronal differentiation various proneural genes are recruited, but the specific proneural genes involved could be different between species and neuronal populations. Here, the expression of other proneural genes has been researched in the developing hypothalamus. Remarkably, as *Neurog1/2* are not expressed in the hypothalamus (Ratié et al., 2013), *Ascl1* appears to be the only proneural gene expressed in the ventral chick AH, at least during early development. Lateral inhibition is the process controlling differentiation of these neurons but the precise mechanisms is different between chick and mouse. There are several lines of evidence to suggest this including restriction of *Ascl1* expression to the AH in chick (**Figure 2B**), where in mouse *Ascl1* is expressed in both the AH and MH (**Figure 2P**). To date, no other proneural gene has been described in the developing chick MH. *Neurog1* and *Neurog2* are not found in the ventral hypothalamus of zebrafish and mouse (Ando et al., 2005; Mcnay et al., 2006; Osório et al., 2010), but a third member of the neurogenin family, *Neurog3* has been identified, specifically expressed in the AH (Wang et al., 2001; Villasenor et al., 2008; Pelling et al., 2011). *Neurog3* expression is regulated by *Ascl1* (Mcnay et al., 2006), but in *Neurog3* mutant mice there is no effect on early neurogenesis in the hypothalamus (Pelling et al., 2011; Anthwal et al., 2013). This lack of phenotype could be due to redundancy between the two proneural genes. It would be interesting to analyses *Ascl1/Neurog3* mutant mice to determine whether there is an additional defect in the formation of the nTPOC.

Additionally, in zebrafish and mouse, *Ascl1* and *Neurog3* may act together to control the processes of lateral inhibition leading to the differentiation of the nTPOC, whereas in chick differentiation is specifically regulated by *Ascl1*.

As the capacity to regulate differentiation steps during neurogenesis is shared by all the proneural genes (Guillemot, 2007), it may explain why neuronal differentiation in the vertebrate hypothalamus is not conserved.

### **DESCRIPTION OF PRONEURAL TARGET GENES WITHIN THE HYPOTHALAMIC PRIMORDIUM**

In the absence of Notch activity during nTPOC differentiation in the chick hypothalamus, *Ascl1* is upregulated and induces expression of a wide spectrum of neuron specific genes (Castro et al., 2011; Ratié et al., 2013). While upregulation of some neuronal genes like *Nhlh1* or *Stmn2* is expected in tissue lacking Notch signaling, other genes identified are not associated with a role in hypothalamic development, such as Transgelin 3 (*Tagln3*) and Chromogranin A (*Chga*). *Nhlh1* and *Chga* mutant mice are available, but there is no phenotype or effect on neurogenesis, suggesting redundancy with other genes (Krüger and Braun, 2002; Hendy et al., 2006; Schmid et al., 2007). *Tagln3* appears to be a good marker because it is strongly expressed in the areas where both the nTPOC and nMTT form. In a flat-mounted preparation of HH15 chick hypothalamus, double labeling with *Shh* and *Tagln3* reveals expression of *Tagln3* in the AH and MH where the nTPOC and nMTT neurons are respectively located (**Figure 3C**, arrowhead and unfilled arrowhead). *Tagln3* is also expressed in the ventral MLF population, which are the first neurons to develop in the brain (**Figure 3C**, asterisk). While these target genes are all expressed in post-mitotic neurons (Theodorakis et al., 2002; Pape et al., 2008; Xie et al., 2008; Burzynski et al., 2009; Ratié et al., under review) no specific function can be attributed to these genes during nTPOC and nMTT development.

Another set of genes are specifically upregulated in the chick AH when Notch signaling is inhibited, *Slit1* and *Robo2*, which are well-known components involved in axon guidance (Chisholm and Tessier-Lavigne, 1999). They guide the TPOC axons through the hypothalamus (Devine and Key, 2008; Ricaño-Cornejo et al., 2011) and the regulation of these genes is Notch dependent (Ratié et al., 2013).

Analysis of the promoter regions in *Slit1*, *Robo2*, *Tagln3* and *Chga* reveal binding sites of *Hes5*, *Hey1*, *Ascl1* and *Nhlh1*

**FIGURE 4 | Network of Notch/proneural genes and initial expression of downstream targets in the developing chick hypothalamus**. Whole-mount in situ hybridization or immunohistochemistry of markers in chick. Immunohistochemistry protocol has been described elsewhere (Lumsden and Keynes, 1989). Anti-HuC/D mouse (1:500; molecular probes;

A21271) primary antibody was detected with a peroxidase-conjugated rabbit-anti-mouse secondary antibody (1:2000; Jackson ImmunoResearch; 315-035-045). Probes were obtained from cDNA and subcloned into pCRII-TOPO (Invitrogen) to make RNA probes or plasmids were obtained from other sources: Dll1 and Notch1 (kind gifts from Dr Frank Schubert). **(A, B)** Frontal view of the AH in the developing hypothalamus where the nTPOC neurons will differentiate. This network of genes is based on in silico results and data from Ratié et al., 2013. Expression of all markers, except

providing further evidence these target genes are part of the Notch/proneural regulatory network involved in neuronal differentiation in the hypothalamus (Ratié et al., 2013).

### **MOLECULAR CASCADE OF NEUROGENESIS ONSET IN THE CHICK HYPOTHALAMUS PRIMORDIUM**

In order to corroborate this network of genes, the expression of Notch components and target genes is analyzed by *in situ* hybridization in the chick hypothalamus to provide further evidence for the existence of a molecular cascade that is Notch/proneural dependent.

The molecular cascade begins with the expression of Notch components and proneural genes followed by other bHLH transcription factors, target genes and well-known neuronal markers (**Figure 4**). *Notch1*, *Hes5*, *Dll1*, *Ascl1*, *Nhlh1*, *NeuroD4*, *Stmn2*, HuC/D and *Chga* are examples chosen to evaluate the stage of their first expression in the AH (**Figure 4**). At HH10, the first components to be expressed in the developing hypothalamus are *Notch1*, *Dll1* and *Ascl1*, followed by *Hes5* that form a regulatory loop (**Figure 4A**; Ratié et al., 2013). This mechanism has been well described in the literature for the induction of neurogenesis by lateral inhibition (Bertrand et al., 2002; Kageyama et al., 2008b).

*Notch1* is present in the AH, ubiquitously expressed (**Figure 4A**) compared with *Dll1*, *Ascl1* and *Hes5* that are expressed in a salt-and-pepper like pattern with a horseshoe shape (**Figure 4**). This is in agreement with a lateral inhibition Notch1 have a horseshoe shape. **(A)** Notch network loop in NPCs. Notch1 expression is ubiquitous throughout the hypothalamus at HH10. Dll1 and Ascl1 expression in the hypothalamus at HH10. Hes5 expression in the hypothalamus at HH10+. **(B)** Genes are upregulated in post-mitotic differentiating neurons. Expression of Nhlh1 at HH11<sup>+</sup> and NeuroD4 at HH12. Stmn2 expression at HH12+, HuC/D expression at HH13 and Chga expression, first appears at HH13++ in very few cells in the developing hypothalamus. Genes are expressed in a salt-and-pepper like pattern (arrowhead). Expression confirms Notch components and Ascl1 are expressed first, followed by the expression of downstream targets. Arrows represent activation of downstream targets. Barred lines represent repression of downstream targets. A single line represents direct binding between ligand and receptor.

model taking place in the AH. In this model, when *Ascl1* is active in a NPC, this can upregulate other bHLH genes such as *Nhlh1* (Ratié et al., 2013). *Nhlh1* and *NeuroD4* are analyzed as they are known markers of differentiation and expressed in the hypothalamus (Murdoch et al., 1999; Abu-Elmagd et al., 2001; Ratié et al., 2013). These genes are expressed from HH12, with *Nhlh1* expression appearing slightly earlier at HH11++ (**Figure 4B**). Other genes are upregulated from around HH13 such as, the well-known neuronal markers *Stmn2* and HuC/D but also new markers such as *Chga* (**Figure 4B**; Ratié et al., 2013).

BrdU labeling suggests *Dll1* expressing cells have exited the cell cycle (Henrique et al., 1995; Myat et al., 1996) therefore NPCs destined to become nTPOC neurons exit the cell cycle around HH10 as seen with *Dll1* expression (**Figure 4A**). It suggests that as early as HH10, the *Dll1* positive cells of the hypothalamus are destined to become neurons several stages before they become mature neurons expressing markers such as *Stmn2* or HuC/D at HH13. These results provide further evidence that the Notch/proneural loop is active in the hypothalamus from a very early stage before the first neurons appear.

#### **CONCLUDING REMARKS**

In this review, data has been discussed implicating the Notch/proneural network with a role during the differentiation of the first two groups of neurons that develop in the hypothalamus, the nTPOC and nMTT. There is still specific functional data lacking in the hypothalamus to conclude the specific mechanisms in which these neurons differentiate, but a general picture using expression data and interpretation of other functional models has been achieved. The same Notch/proneural network is likely to regulate differentiation of nTPOC neurons in zebrafish, chick and mouse. Considering the conservation of the TPOC axon tract and the Notch signaling pathway, this is not surprising. Data regarding proneural gene expression in the chick MH is still too scarce to conclude, but some of the components of the Notch/proneural network are expressed in the mouse MH before the nMTT neurons differentiate. This expression suggests the same mechanisms occur between the nTPOC and nMTT, only the players for nMTT differentiation are yet to be found in chick.

It is still not known what triggers this Notch/proneural loop in these hypothalamic NPCs. Neuronal specification during spinal cord development is initially generated by activities of two competing signaling pathways: SHH and BMP/Wnt (Ericson et al., 1997; Jessell, 2000; Liem et al., 2000). Evidence is emerging to suggest that SHH and BMP may play a similar role in the differentiation of the early hypothalamic neurons (Manning et al., 2006; Ahsan et al., 2007; Szabó et al., 2009; Alvarez-Bolado et al., 2012). However, how these signaling pathways integrate the Notch/proneural network has to be investigated in the developing hypothalamus. Future work will require a study to identify transcription factors that are necessary for the patterning of the AH and MH very early during vertebrate development.

One thing is clear, this review highlights lots of open questions regarding initial neuronal differentiation in the hypothalamus as well as general patterning of the hypothalamic regions. We hope that this review will encourage the scientific communities to investigate the phenotype of their mutants during earlier stages when the nTPOC and nMTT neurons develop.

A final thought, distinct late hypothalamic cell types dysfunction can lead to metabolic or homeostatic disorders and there is evidence that this is the case in congenital obesity (Gibson et al., 2004; Bingham et al., 2008). Therefore, could a defect in the induction and specification of the initial neurons lead to such disorders as these neurons are essential to the axon tract formation of the late hypothalamic neurons (such as the MCH neurons) (Croizier et al., 2011).

#### **AUTHOR CONTRIBUTIONS**

Michelle Ware and Valérie Dupé set up and designed the experiments. Michelle Ware and Houda Hamdi-Rozé performed the experiments. Michelle Ware and Valérie Dupé wrote the manuscript. All authors read, discussed and edited the manuscript.

#### **ACKNOWLEDGMENTS**

We would like to thank the members of the David laboratory for suggestions and comments. This work was supported by the Agence Nationale de la Recherche (grant no. ANR-12-BSV1- 0007-01, Valérie Dupé). We also thank the animal house platform ARCHE (SFR Biosit, Rennes, France). We are grateful to receive the following plasmids: mouse *Ascl1* (Dr Francois Guillemot, London, UK), chick *Dll1* and *Notch1* (Dr Frank Schubert, Portsmouth, UK).

### **REFERENCES**


implicated in neurogenesis. *Dev. Genet.* 24, 165–177. doi: 10.1002/(SICI)1520- 6408(1999)24:1/2<165::AID-DVG15>3.0.CO;2-V


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 01 October 2014; paper pending published: 21 October 2014; accepted: 10 November 2014; published online: 02 December 2014*.

*Citation: Ware M, Hamdi-Rozé H and Dupé V (2014) Notch signaling and proneural genes work together to control the neural building blocks for the initial scaffold in the hypothalamus. Front. Neuroanat. 8:140. doi: 10.3389/fnana.2014.00140 This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2014 Ware, Hamdi-Rozé and Dupé. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## *nkx2.1* and *nkx2.4* genes function partially redundant during development of the zebrafish hypothalamus, preoptic region, and pallidum

### *Martha Manoli <sup>1</sup> and Wolfgang Driever 1,2\**

*<sup>1</sup> Developmental Biology, Faculty of Biology, Institute Biology I, University of Freiburg, Freiburg, Germany*

*<sup>2</sup> Centre for Biological Signaling Studies (BIOSS), University of Freiburg, Freiburg, Germany*

#### *Edited by:*

*Gonzalo Alvarez-Bolado, University of Heidelberg, Germany*

#### *Reviewed by:*

*Thomas Mueller, Kansas State University, USA Steffen Scholpp, Karlsruhe Institute of Technology, Germany*

#### *\*Correspondence:*

*Wolfgang Driever, Institute of Biology 1, Albert-Ludwigs-University Freiburg, Hauptstrasse 1, D-79104 Freiburg, Germany e-mail: driever@biologie. uni-freiburg.de*

During ventral forebrain development, orthologs of the homeodomain transcription factor Nkx2.1 control patterning of hypothalamus, preoptic region, and ventral telencephalon. However, the relative contributions of Nkx2.1 and Nkx2.4 to prosencephalon development are poorly understood. Therefore, we analyzed functions of the previously uncharacterized *nkx2.4-*like *zgc:171531* as well as of the presumed *nkx2.1* orthologs *nkx2.1a* and *nkx2.1b* in zebrafish forebrain development. Our results show that *zgc:171531* and *nkx2.1a* display overlapping expression patterns and a high sequence similarity. Together with a high degree of synteny conservation, these findings indicate that both these genes indeed are paralogs of *nkx2.4*. As a result, we name *zgc:171531* now *nkx2.4a,* and changed the name of *nkx2.1a* to *nkx2.4b*, and of *nkx2.1b* to *nkx2.1.* In *nkx2.1, nkx2.4a,* and *nkx2.4b* triple morpholino knockdown (nkx2TKD) embryos we observed a loss of the rostral part of prosomere 3 and its derivative posterior tubercular and hypothalamic structures. Furthermore, there was a loss of rostral and intermediate hypothalamus, while a residual preoptic region still develops. The reduction of the ventral diencephalon was accompanied by a ventral expansion of the dorsally expressed *pax6,* revealing a dorsalization of the basal hypothalamus. Within the telencephalon we observed a loss of pallidal markers, while striatum and pallium are forming. At the neuronal level, nkx2TKD morphants lacked several neurosecretory neuron types, including avp, *crh*, and *pomc* expressing cells in the hypothalamus, but still form *oxt* neurons in the preoptic region. Our data reveals that, while *nkx2.1, nkx2.4a*, and *nkx2.4b* genes act partially redundant in hypothalamic development, *nkx2.1* is specifically involved in the development of rostral ventral forebrain including the pallidum and preoptic regions, whereas *nkx2.4a* and *nkx2.4b* control the intermediate and caudal hypothalamus.

**Keywords: hypothalamus, preoptic region, pallidum, diencephalon, telencephalon, neural patterning, neuroendocrine neurons, zebrafish**

### **INTRODUCTION**

Regional specification of the central nervous system (CNS) commences during gastrulation and at neural plate stage, when patterning along the anteroposterior as well as the dorsoventral axes is initiated. The anteroposterior patterning assigns rostrocaudal regional identities (Lumsden and Krumlauf, 1996), and together with the dorsoventral patterning influences cell fate specification and consequently the generation of different neuronal subtypes (Tanabe and Jessell, 1996). Several *Nkx* genes act during patterning of the ventral CNS, and contribute to a molecular code for neuronal differentiation (Shimamura et al., 1995; Ericson et al., 1997; Pabst et al., 1998). Nkx2.1 is a member of the vertebrate Nkx homeobox transcription factor family (Pera and Kessel, 1998; Small et al., 2000; van den Akker et al., 2008). It is also known as thyroid transcription factor 1 (TTF-1) or Thyroid-specific enhancer-binding protein (T/ebp) because of its involvement in thyroid development (Guazzi et al., 1990; Mizuno et al., 1991; Elsalini et al., 2003). Two *nkx2.1* genes in the zebrafish genome have been previously described as paralogs*, nkx2.1a* and *nkx2.1b* (Rohr et al., 2001). Expression of *nkx2.1a* and *b* in the zebrafish CNS was reported to initiate toward the end of gastrulation in a rostrocaudal stripe in the medial anterior neural plate, giving rise to hypothalamus, preoptic region and ventral telencephalon (Rohr et al., 2001). In the medial neural plate and ventral neural tube, expression of mammalian *nkx2.1* has distinct anteroposterior and dorsoventral boundaries (Shimamura et al., 1995; Briscoe et al., 2000; Puelles and Rubenstein, 2003). *nkx2.1a* and *b* expression domains display a common posterior border, which resides in the basal part of prosomere 3 located ventrally to the prethalamus. Both are expressed in the posterior tuberculum and basal hypothalamus. However, the preoptic region, the basal telencephalon around the anterior commissure, and the alar preoptic region exclusively express *nkx2.1b* (Rohr et al., 2001; Lauter et al., 2013). The medial hypothalamus is characterized by *nkx2.1a* expression and low or absent *nkx2.1b* expression.

In mice, development of ventral hypothalamus and telencephalic medial ganglionic eminence depend on NKX2.1. In homozygous *Nkx2.1* knockout mice, ventral forebrain developmental abnormalities start anteriorly in the septal area and extend to the mammillary body of the hypothalamus (Kimura et al., 1996). Loss of NKX2.1 also causes a transformation of medial ganglionic eminence into lateral ganglionic eminence structures (Sussel et al., 1999). In addition, both in mouse and Xenopus embryogenesis removal of *nkx2.1* causes a dorsalization of the basal plate forebrain (Sussel et al., 1999; van den Akker et al., 2008). *Nkx* genes are also involved in neuronal differentiation. In the spinal cord the combinatorial expression of *Nkx* transcription factors (*Nkx6.1, Nkx2.2*, and *Nkx2.9*) specifies ventral identity of neurons (Briscoe et al., 2000; Sander et al., 2000). Less is known about potential involvements of Nkx2 factors in forebrain neuronal differentiation, where Nkx2.1 has been shown to contribute to cortical interneuron subtype specification (Butt et al., 2008).

Here, we report the identification, expression and functional analysis of the zebrafish *nkx2.4a* gene. During embryogenesis *nkx2.4a* is expressed in the hypothalamus in a manner similar to the gene previously reported as *nkx2.1a.* Phylogenetic analysis suggests that *nkx2.4a* resulted from an *nkx2.1* gene duplication event that is not restricted to teleosts (Price, 1993; Small et al., 2000; Wang et al., 2000). In zebrafish, like in mouse (Marín et al., 2002), *nkx2.4a* is expressed in a restricted area of the hypothalamus. In contrast to *nkx2.1*, *nkx2.4a* is not expressed in the telencephalon. Synteny and phylogenetic analysis reveals that the previously reported *nkx2.1a* is in fact a paralog of *nkx2.4*, and not of *nkx2.1,* and thus will be named *nkx2.4b* here in accordance with zebrafish nomenclature. Combined inactivation of *nkx2.1, nkx2.4a*, and *nkx2.4b* function by triple Morpholino knockdown (here called nkx2TKD) revealed their contribution to neural patterning and neuronal differentiation. In contrast to single knockdown, the ventral diencephalon of nkx2TKD morphants is dorsalized, revealing that *nkx2.1*, *nkx2.4a*, and *nkx2.4b* act redundantly in hypothalamic patterning. All three genes are required for development of specific subsets of neurosecretory neurons in the hypothalamus.

#### **MATERIALS AND METHODS**

#### **ZEBRAFISH HUSBANDRY**

Zebrafish breeding and maintenance were carried out under standard conditions at 28.5◦C (Westerfield, 2000). We used AB-TL wildtype zebrafish, the enhancer trap *ETvmat2:GFP* (Wen et al., 2008), and *smob*<sup>641</sup> (Barresi et al., 2000), which were identified morphologically. To inhibit pigmentation, embryos were incubated in egg water containing 0.2 mM 1-phenyl-2-thiourea. Embryos were staged according to Kimmel et al. (1995).

#### **INJECTION OF MORPHOLINOS AND RNA**

The following morpholinos were used (Gene Tools LLC): *nkx2.4a* TBMO (ATG): 5 -GCTCAGCGACATGGTTCAGCCC GCA-3 , *nkx2.4a* SBMO (e1i1): 5 -TGCATTAGAAGAACTTAC TTGTTGA-3 . The standard control SCMO, p53 ATG (Robu et al., 2007), *nkx2.1* (previously published as *nkx2.1b*) and *nkx2.4b* (published as *nk2.1a-1*) ATG morpholinos have been described (Elsalini et al., 2003). Morpholinos were diluted in H2O containing 0.05% phenol red or 0.05% rhodamine dextran. *nkx* morpholinos were injected at the one cell stage in different combinations at a total amount of 7.5 ng per embryo, plus an additional 2.5 ng p53 MO. As controls, 7.5 ng standard control morpholino (SCMO) and 2.5 ng p53 MO per embryo were injected.

The efficiency of the *nkx2.4a-e1i1*-SBMO was verified by RT-PCR using cDNA synthesized from 1, 2, and 3 dpf embryos injected with SCMO or *nkx2.4a-e1i1*-SBMO+*nkx2.4b*-MO+*nkx2.1-*MO and *p53*-MO at 2.5 ng each (same conditions as used for all nkx2TKD; **Figure 3B**). The PCR products where sequenced using primers spanning the whole *nkx2.4a* coding region: *nkx2.4a* F: 5 -ATGTCGCTGAGCCCAAAG-3 , *nkx2.4a* R: 5 -CTACCACGTTCTGCCATAAAGC-3 .

To analyse the specificity and efficacy of the translation blocking MOs (TBMOs) directed against the translation start sites of *nkx2.1*, *nkx2.4a*, or *nkx2.4b*, pCS2+-gfp-reporter plasmids were created which each harbor the respective morpholino target sequence, fused in frame to the GFP ORF. *gfp-*reporter mRNAs were generated from plasmids pCS2+-5 UTR-nkx2.4bgfp, pCS2+-5 UTR-nkx2.1-gfp and pCS2+-5 UTR-nkx2.4a-gfp linearized with NotI, and transcribed using the SP6 mMessage mMachine kit (Ambion). *gfp*-reporter mRNAs were co-injected into one-cell stage embryos in combination with SCMO or the respective specific targeting morpholino. At epiboly stages, embryos were assayed for GFP fluorescence (**Figure 4**). We also attempted to validate the knockdown phenotypes by rescue experiments injecting mRNAs encoding the Nkx2.1 and Nkx2.4, however broad overexpression of Nkx2.1 or Nkx2.4 from injected mRNAs caused severely abnormal development likely because of gain-of-function effects due to ectopic expression (data not shown).

To verify the results obtained with the *nkx2.4a* splice blocking MO, we repeated the nkx2TKD substituting the *nkx2.4a* SBMO with the translation blocking *nkx2.4a* TBMO in the nkx2TKD mix, and analyzed *tyrosine hydroxylase* (*th*), *lhx5*, and *lhx6* expression (**Figure 6**). Both *nkx2.4a* SBMO or TBMO at 2.5 ng per embryo in combination with *nkx2.4b* and *nkx2.1* TBMO generated a similar phenotype.

Synthetic *lefty1* mRNA (mMessage mMachine kit, Ambion) was injected as described (Barth and Wilson, 1995; Thisse and Thisse, 1999).

#### *IN SITU* **HYBRIDIZATION AND TUNEL STAINING**

Whole-mount *in situ* hybridization (WISH) (Lauter et al., 2011) and fluorescent WISH with immunohistochemistry (Filippi et al., 2010) were performed as described. The following digoxigeninlabeled riboprobes were synthesized: *dbx1a* (Fjose et al., 1994), *dlx5a* (Akimenko et al., 1994), *foxa2* (Strähle et al., 1993), *lhx5* and *lhx6* (Toyama et al., 1995), *shha* (Ekker et al., 1995), *otpa* (Del Giacco et al., 2006; Ryu et al., 2007), *pax6* (Krauss et al., 1991b), *pomca* (Herzog et al., 2003), *nkx2.2* (Barth and Wilson, 1995). For generation of *nkx2.4b* (RefSeq NM\_131589.1), *nkx2.1(b)* (NM\_131776.1), and *nkx2.4a* (NM\_001111166.1) probes, the coding sequences were PCR amplified and cloned into the TOPO PCRII vector using the primers *nkx2.4b* F: 5 -ATGTCCTTGA GCCCCAAAC-3 ; *nkx2.4b* R: 5 -TCACCATGTTCTGCCGTACA-3 ; *nkx2.1* F: 5 -ATGTCGATGAGCCCTAAGCA-3 ; *nkx2.1* R: 5 - TCACCACGTCCTGCCATA-3 (for *nkx2.4a* see above).

TUNEL assay was performed using the Apoptag *in situ* apoptosis detection kit (Chemicon) (Ryu et al., 2005). Classification of signal: "absent"—no blue stained cell detected in specific brain region; "few cells"—small number of blue stained cell (1–10) dispersed in specific brain region (example telencephalic region in **Figures 3I,J**); "many cells" accumulation of *>*10 blue stained cells, often clustered, in specific brain region.

#### **MICROSCOPY AND IMAGE ANALYSIS**

Transmitted light images were acquired using a Zeiss Axioskop compound microscope. Fluorescently labeled embryos were documented by confocal image stacks using a Zeiss LSM510. Images shown are z-projections of defined sets of consecutive focal planes assembled with the Zeiss Zen 2012 software.

#### **SEQUENCE ALIGNMENTS AND SYNTENY**

NKX protein sequences were aligned and analyzed with CLC Genomics Workbench 5 (www.clcbio.com) using the distance based method to generate a phylogenetic tree. The NCBI accession numbers are: NK2.4b [Danio rerio] NP\_571664.1, NK2.1(b) [Danio rerio] NP\_571851.1, Nkx-2.1 [Mus musculus] NP\_001139670.1, Nkx-2.1 isoform 2 [Homo sapiens] NP\_003308.1, Nkx-2.1 isoform 1 [Xenopus tropicalis] XP\_002935383.1, Nkx-2.4a [Danio rerio] NP\_001104636.1, Nkx-2.4 [Mus musculus] NP\_075993.1, Nkx-2.4 [Homo sapiens] NP\_149416.1, Nkx-2.4 [Xenopus tropicalis] XP\_002939478.1, Nkx-2.2 [Homo sapiens] NP\_002500.1, Nkx2-2 protein [Mus musculus] AAI38160.1, Nkx-2.2 isoform X1 [Xenopus tropicalis] XP\_002939477.1, Nkx-2.2a [Danio rerio] NP\_571497.1, Nkx-2.3 [Homo sapiens] NP\_660328.2, Nkx2-3 [Mus musculus] CAA72002.1, Nkx-2.3 [Xenopus tropicalis] XP\_002937234.1, Nkx2.3 [Danio rerio] AAC05228.1, Nkx-2.5 [Mus musculus] NP\_032726.1, Nkx-2.5 isoform 1 [Homo sapiens] NP\_004378.1, Nkx2.5 [Danio rerio] AAC05229.1, Nkx2-5 [Xenopus tropicalis] AAI60531.1

Synteny between the mouse and zebrafish *nkx2* genes was visualized using Cinteny (http://cinteny*.*cchmc*.*org).

#### **RESULTS**

#### *Nkx2.1* **AND** *Nkx2.4* **GENES IN ZEBRAFISH**

We systematically searched for zebrafish genes that may interact and be co-expressed with *nkx2.1* transcription factors during ventral diencephalon development, and noticed that the *zgc:171531* gene (ENSDARG00000075107) encodes an NKX2 family member. *zgc:171531* has recently been reported to be expressed in the hypothalamus during late somitogenesis stages (Armant et al., 2013), but its function has not been studied. We analyzed the phylogenetic relationship of Zgc:171531 to other members of the Nkx protein family, and could confirm the suggestion by Armant et al. (2013) that Zgc:171531 is an ortholog of mouse NKX2.4 (**Figure 1**). Phylogenetic tree analysis showed that zebrafish Zgc:171531 is the closest relative of mammalian NKX2.4, and that Zgc:171531and Nkx2.1 are more closely related

to each other than to any other Nkx2 transcription factor family member (**Figure 1A**). Therefore, we refer to Zgc:171531 as Nkx2.4a.

To explore the evolutionary history of the *nkx2.1* and *nkx2.4* genes, we searched for synteny conservation. We found that the gene previously reported as zebrafish *nkx2.1b* is located in a microsyntenic chromosomal region related to mouse *Nkx2.1* (**Figure 1B**), thus *nkx2.1b* is indeed the ortholog of mouse *Nkx2.1*. Zebrafish *zgc:171531* resides in a microsyntenic chromosomal region related to mouse *Nkx2.4.* Importantly, also zebrafish *nkx2.4b* (previously named *nkx2.1a*) shares synteny with mouse *Nkx2.4* (**Figure 1C**). Thus, we conclude that *nkx2.4a* and *nkx2.4b* are indeed paralogous NKX2.4 genes derived from the teleost specific genome duplication. In contrast, the previous *nkx2.1a* and *nkx2.1b* are not paralogs, and *zgc:171531* is now named *nkx2.4a*, while *nkx2.1a* is named *nkx2.4b,* and *nkx2.1b* is named *nkx2.1.* This change in nomenclature was approved by the zebrafish gene nomenclature committee (see www.zfin.org).

We analyzed expression of *nkx2.4a* in comparison to *nkx2.1* (**Figure 2**) (Rohr et al., 2001; Tessmar-Raible et al., 2007). At 1 dpf *nkx2.4a* is expressed within the forebrain in basal prosomere 3 including the hypothalamus and the posterior tuberculum, and with decreasing expression levels into the preoptic hypothalamic areas of the basal secondary prosencephalon (**Figure 2E**). In contrast to *nkx2.1* (**Figures 2C,D**), expression of *nkx2.4a* was not detected in the alar hypothalamic region (anterior preoptic region) or the pallidum of the secondary prosencephalon. At 2 dpf *nkx2.4a* expression is downregulated in the basal hypothalamic preoptic area, but continues in the posterior tuberculum and caudal hypothalamus (**Figure 2F**). At 3 dpf *nkx2.4a* expression persists in the posterior tuberculum and hypothalamus (**Figure 2G**), similar to *nkx2.4b* (**Figures 2A,B**). This expression pattern is maintained until 4 dpf (data not shown). Thus, all three *nkx2* genes are expressed in the posterior tuberculum as well as in the basal hypothalamus. However, *nkx2.1* is also expressed in the basal telencephalon around the anterior commissure as well as in the alar hypothalamus preoptic region. *nkx2.4a,* in contrast to *nkx2.4b* (Elsalini et al., 2003), is not expressed in the thyroid gland (**Figure 2H**). Double fluorescent WISH of *nkx2.4a* expression in combination with *nkx2.4b* or *nkx2.1* revealed broad coexpression of *nkx2.4a* and *nkx2.4b* (**Figure 2J**), whereas *nkx2.4a* and *nkx2.1* coexpression is restricted mainly to the intermediate hypothalamus (**Figures 2I,K**).

#### **NODAL AND Shh SIGNALING CONTROL** *Nkx2.4a* **EXPRESSION**

Nodal signaling has been revealed essential for initiation, and Shh for maintenance of *nkx2.1* and *nkx2.4b* expression (Rohr et al., 2001). Smoothened is a key transmembrane effector essential for transmitting Hedgehog signals into the cell (Rohatgi and Scott, 2007). *smoothened homolog (smo*−*/*−*)* zygotic mutant embryos have severely reduced Shh signaling, although no complete loss occurs due to residual maternal Smo activity (Barresi et al., 2000; Rohr et al., 2001). We detected *nkx2.4a* expression at reduced levels in *smo*−*/*<sup>−</sup> embryos (**Figures 2L,M**), indicating that Hedgehog signaling activity is required to maintain *nkx2.4a* expression. However, Shh is not required for *nkx2.4a* initial induction, similar to what has been shown for *nkx2.4b*.

We then overexpressed *lefty1* mRNA, encoding an inhibitor of Nodal activity, which resulted in complete loss of *nk2.4* expression in *>*50% of experimental embryos (**Figures 2N,O**). Thus, Nodal is strictly required to induce *nkx2.4a* expression, while Shh contributes to its maintenance.

#### *Nkx2.1* **AND** *Nkx2.4a/4b* **COMBINED KNOCKDOWN SEVERELY AFFECTS THE HYPOTHALAMUS**

To examine the role of Nkx2.4a we performed gene knockdown using antisense MOs. The *nkx2.4a* SBMO is complementary to the exon1-intron1 boundary (**Figure 3A**). We have amplified *nkx2.4a* mRNA by RT-PCR from *nkx2.4a* SBMO injected embryos, and found intron1 to be completely retained in the mature mRNA (**Figure 3B**). Sequencing of the splice-blocked mRNA revealed a premature stop codon truncating the Nkx2.4a protein before the homeodomain. To determine the efficiency of the *nkx2.4a* TBMO a *nkx2.4a-gfp* reporter mRNA with the MO binding site at the GFP ATG position was co-injected with the *nkx2.4a* TBMO and shown to completely suppress GFP expression (**Figure 4**). Thus, *nkx2.4a* SBMO and TBMO efficiently inhibit Nkx2.4a expression. Knocking down *nkx2.4a* alone did not result in any detectable morphological phenotype (data not shown).

The similarity in amino acid sequences and the overlapping expression patterns suggest that *nkx2.1, nkx2.4a*, and *nkx2.4b* may act redundantly. Therefore, we aimed at combined knockdown of all three *nkx2* genes. We decided to use Morpholino antisense knockdown to study the combined activity of these three genes, because mutations in none of these genes are available, and even if mutations may become available at some point, the genetic analysis of triple mutations providing the desired phenotype only in one out of 64 embryos is cumbersome. ATG-morpholinos specific to both *nkx2.1* genes have been published (Elsalini et al., 2003). We confirmed their efficacy by knockdown of GFP expression from injected reporter mRNAs (**Figure 4**). *nkx2.4b* TBMO, *nkx2.1* TBMO, and *nkx2.4a* SBMO

**FIGURE 2 | Expression and regulation of** *nkx2.1, nkx2.4a***, and** *nkx2.4b***. (A,B)** *nkx2.4b*, **(C,D)** *nkx2.1* and **(E–H)** *nkx2.4a* mRNA expression analyzed by whole mount *in situ* hybridization at indicated stages. **(I–K)** Expression of *nkx2.4a* and *nkx2.4b* and *nkx2.1* were detected by double fluorescent whole mount *in situ* hybridization of 2 dpf embryos, magenta and green channels as indicated in panel headings. Shown are 140–190µm Z-projections of confocal image stacks. **(L,M)**

Expression of *nkx2.4a* in WT and *smo* homozygous mutant embryo. Arrow: residual *nkx2.4a* expressions (**M**, *n* = 10). *nkx2.4a* expression in controls **(N)** and embryos with Nodal signaling inactivated by *lefty1* mRNA injection (**O**: 65% of embryos showed a complete loss, 25% a reduction, 10% normal *nkx2.4a* expression; *n* = 25). **(A1–G1,L-O)** lateral views; **(A2–G2,I–K)** dorsal views. Anterior to left. Scale bar in **A1** = 100µm (for **A–H**), in **I** = 20µm (for **I–K**), in **L** = 100µm (for **L–O**).

**morphants. (A)** Schematic representation of the *nkx2.4a* gene with the morpholino binding site and the stop codon (12 bases into intron) terminating the protein when intron excision is blocked by the SBMO. **(B)** RT-PCR for *nkx2.4a mRNA* prepared from 3 dpf embryos injected with 7.5 ng SCMO or with 2.5 ng each of *nkx2.4a-e1i1*-MO, *nkx2.4b* and *nkx2.1* ATG MOs. The *nkx2.4a-*e1i1 MO effectively blocks the splice donor site at the exon1-intron1 boundary, giving rise to a longer cDNA. The mature *nkx2.4a* mRNA is nearly completely eliminated under these conditions. **(C–F)** Morphological *in vivo* phenotype of nkxTKD knockdown combined with p53MO knockdown

**(E2–F2)** Close-ups with the ventral border of the diencephalon outlined by dotted lines. **(G,H)** nkxTKD larvae survived at least until 6 dpf (swim bladder is not inflated in fish in **H** indicated by an arrow). **(I–M)** TUNEL staining for apoptotic cells in 30 hpf and 2 dpf SCMO plus p53MO-injected control and nkxTKD plus p53MO morphant embryos (arrows in **I** and **J**: hypothalamus) (**K–L,K2,L2** show close-ups). **(M)** Quantification of number of TUNEL stained apoptotic cells in nkxTKD+p53 (*n* = 12) morphant larvae compared to SCMO+p53MO injected larvae (*n* = 19). Lateral views, anterior at left. Abbreviations see **Table 1**. Scale bars = 100µm except **G**: scale bar = 400µm.

were co-injected at 2.5 ng each per embryo in combination with p53 morpholino (for exact amounts used in each figure panel see Supplemental Table 1). Given that some morpholinos cause cell death irrespective of knockdown of specific gene function (Robu et al., 2007), we included p53 MO in our nkx2TKD knockdown mix in all experiments. Live nkx2TKD embryos and larvae were analyzed for morphological abnormalities using transmitted light microscopy (**Figures 3C–H**). At 2 dpf nkx2TKD embryos developed a severe loss of hypothalamic structures, as well as abnormalities in the preoptic region (**Figures 3C,D**). At 3 dpf the massive loss of ventral forebrain tissue became more prominent (**Figures 3E,F**). With the exception of the ventral forebrain, nkx2TKD embryos and larvae developed anatomical structures morphologically similar to control WT embryos, and overall progress of development was not affected. nkx2TKD larvae established blood circulation and survived at least until 6 dpf (**Figures 3G,H**), but frequently did not fully inflate the swim bladder.

To test whether nkx2TKD embryos may develop apoptosis, we performed TUNEL assays. Apoptosis was clearly increased in

#### **Table 1 | Anatomical abbreviations.**


the hypothalamus of morphants compared to SCMO injected embryos at 30 hpf (**Figures 3I,J,M**). However, at 2 dpf apoptosis levels in nkx2TKD morphants were similar to SCMO injected control larvae (**Figures 3K,L**). Thus, nkx2TKD causes increased apoptosis in those brain regions of *nkx2.1/2.4* expression specifically when the hypothalamus forms, but not at later stages.

#### **KNOCKDOWN OF** *Nkx2.1/4a/4b* **SEVERELY AFFECTS VENTRAL FOREBRAIN PATTERN FORMATION**

Shh is a key determinant of ventral pattern formation in the neural plate. Studies have shown that its expression in the basal telencephalon depends on *nkx2.1* function in mice (Sussel et al., 1999). In wild-type zebrafish embryos *shha* (**Figure 5A**) is expressed in floor plate and underlying axial mesoderm, as well as in the zona limitans intrathalamica (ZLI; Scholpp et al., 2006), which is a narrow transverse region between the prethalamus and thalamus (Shimamura et al., 1995; Kiecker and Lumsden, 2004). In *nkx2.1* morphants *shha* expression was mostly normal, with only a minor reduction in the hypothalamus (**Figure 5B**). In *nkx2.4a/4b* double morphants (**Figure 5C**) we detected a more severe reduction of the hypothalamic *shha* expression. In nkx2TKD embryos *shha* expression in the hypothalamus was almost completely eliminated (**Figure 5D**), while in the ZLI and caudal to it *shha* expression appeared normal. Thus, all three *nkx2* genes redundantly regulate *shha* expression in the hypothalamus. Caudal to prosomer 3 *shha* expression was not affected in nkx2TKD morphants. This is consistent with the finding that the expression pattern of *foxa2*, a marker for medial and lateral floor plate caudal to prosomere 2 (Odenthal and Nusslein-Volhard, 1998), appeared unaffected in nkx2TKD morphants with some enlargement of basal prosomeres 1 and 2 (**Figures 5O,P**). With respect to basal prosomere 3, the *nkx2.1* expression extends only into the rostral half of prosomere 3, while *dbx1a* is expressed in the caudal part of basal prosomere 3 (Lauter et al., 2013). *dbx1a* expression was not affected in nkx2TKD morphants, indicating normal development of the caudal portion of prosomer 3 (**Figures 5M,N**). This finding demonstrates that brain regions caudal to the limit of *nkx2.1/4a/4b* expression are not affected upon nkx2TKD.

A target of mouse NKX2.1 in the telencephalon is *Lhx6* (Sussel et al., 1999; Du et al., 2008). In our study, in *nkx2.1* morphants pallidal *lhx6* expression was strongly reduced, but not the diencephalic *lhx6* expression (**Figures 5E,F**). In *nkx2.4a* and *nkx2.4b* double morphants however pallidal *lhx6* expression was maintained, whereas in the diencephalon the caudal posterior tubercular expression was lost (**Figure 5G**). In nkx2TKD morphants, both the pallidal and posterior tubercular *lhx6* expression were severely reduced, while *lhx6* expression in the preoptic supraoptoparaventricular (SPV) region was retained (**Figure 5H**). These results were confirmed by analysing the more broadly expressed *lhx5* (**Figure 5I**). The *lhx5* domains in posterior tuberculum and caudal and rostral hypothalamus were depleted in nkx2TKD morphants, while expression in the alar preoptic region was still detectable (**Figure 5L**). In the telencephalon, the dorsal *lhx5* expression domain expanded ventrally into the pallidal area, from which *lhx5* was absent in wildtype. This ventral expansion of the telencephalic *lhx5* expression was also visible in *nkx2.1* morphants (**Figure 5J**), which otherwise had an unaltered *lhx5* expression pattern. Double *nkx2.4a* and *nkx2.4b* morphants did not develop the telencephalic *lhx5* expansion, but exhibited a reduction of the hypothalamic expression pattern (**Figure 5K**). The nkx2TKD expression changes of *lhx5* and *lhx6* were also validated in triple knockdown embryos using the *nkx2.4a* TBMO (**Figure 6**). The expression of *otpa* in the posterior tuberculum was strongly reduced in nkx2TKD morphants, while expression was still detected in the anterior preoptic area (**Figures 5Q,R**). Analysis of *dlx5a* reveals that the dorsal subpallium (striatum) still forms in nkx2TKD morphants (**Figures 5S,T**). In contrast, the caudal hypothalamic *dlx5a* domain was absent in triple morphants. Similar to what was observed for *lhx6* expression in nkx2TKD morphants, the *dlx5a* expression domain in the alar preoptic region appears expanded ventro-caudally (**Figures 5S,T**). *nkx2.2a* has a rostro-caudal domain extending approximately along the alar-basal boundary of the fore- and midbrain (Puelles and Rubenstein, 1993; Hauptmann and Gerster, 2000). This boundary remained unaffected in the triple morphants. Also, the prethalamic and thalamic as well as preoptic *nkx2.2a* expression domains appeared largely normal, while the caudal hypothalamic expression was reduced (**Figures 5U,V**). Analysing *emx1* expression, we did not observe any effect of nkx2TKD on the pallium (**Figures 5W,X**). In summary, these results show that

**FIGURE 4 | Experimental validation of** *in vivo* **knockdown efficiency of translation blocking morpholinos used in this study.** MO knockdown efficiency was demonstrated by co-injection of GFP reporter mRNAs with the morpholino binding sites engineered at the start ATG of GFP. The following

GFP reporter were injected at 100 pg mRNA per embryo: **(A,B)** *5 UTR-nkx2.1-gfp* mRNA, **(C,D)** *5 UTR-nkx2.4a-gfp* mRNA, and **(E,F)** *5 UTR-nkx2.4b-gfp* mRNA. Morpholinos used are indicated in boxes at *(Continued)*

#### **FIGURE 4 | Continued**

left of each panel and were injected at 5 ng each, controls containing the same amount of SCMO. *nkx2.1*ATGMO and *nkx2.4b*ATGMOmorpholinos have been described previously (Elsalini et al., 2003). All embryos were also co-injected with rhodamine-dextran, which was used to sort embryos after 4–5 h. of development for homogenous amounts and distribution of injected material. Knockdown efficiency was then documented between 4 and 5 hpf by pictures

taking with the epifluorescence dissecting microscope in the red channel to verify that embryo were injected, in the green channel to control efficiency of GFP knockdown, and in the transmitted light channel to control morphology and viability of embryos. *nkx2.4b* and *nkx2.1* and *nkx2.4a*TBMO morphants were rhodamine-dextran fluorescence levels similar to control-injected embryos but showed no expression of the respective GFP reporters, indicating that these TBMOs effectively bind and block translation of target mRNAs *in vivo*.

loss of *nkx2.1/4a/4b* activity mainly affects three regions of the forebrain. In the caudal diencephalon, there is a loss of the rostral part of basal prosomer 3 and derivative posterior tubercular and hypothalamic structures, while the alar basal boundary appears non-shifted. Further rostrally, there is a loss of rostral and intermediate hypothalamic markers, while a residual preoptic region including the supraoptoparaventricular region develops in nkx2TKD morphants. Within the telencephalon, there is a loss of pallidal markers, while striatum and pallium are forming.

#### **KNOCKDOWN OF** *nkx2.1/4a/4b* **LEADS TO A DORSALIZATION OF THE VENTRAL DIENCEPHALON**

Studies in mice suggested that *Nkx2.1* knockout causes a ventral expansion of dorsal diencephalic marker genes (Kimura et al., 1996). Therefore, we tested whether similar fate shifts occur in nkx2TKD embryos by analysing *pax6a* expression, which is predominantly expressed in dorsal parts of the diencephalon (Krauss et al., 1991a), as well as in cells at the pallial-subpallial boundary (Wullimann and Rink, 2001). In nkx2TKD embryos, *pax6* expression was expanded ventrally in its prethalamic domain, suggesting a loss of basal prosomere 3 (**Figures 5Y–AA**). Together with the data obtained for *dbx1a* expression (**Figures 5M,N**; caudal part of prethalamus and caudal part of basal prosomer 3; Lauter et al., 2013), it appeared that the rostral part of basal prosomere 3 was completely dorsalized, while the caudal *dbx1a* expressing part of prosomere 3 was not affected. Also, it appears that alar *pax6a* expression did not invade into the basal prosomere 1 and 2 regions, which in morphants appeared more pronounced caudal to prosomere 3 (see also 2 dpf *foxa2* expression **Figures 5O,P**). In contrast, when only *nkx2.1* and *nkx2.4b* were knocked down, no obvious ventral expansion of the *pax6a* domain into the hypothalamus was detectable. Thus, *nkx2.4a* can compensate loss of *nkx2.1* and *nkx2.4b* in hypothalamic patterning. In summary these results suggest that the loss of *nkx2.1/4a/4b* activity leads to a dorsalization of the basal prosomer 3 as well as of the pallidum.

#### **Nkx2.1/4a/4b CONTROL EXPRESSION OF** *nkx2.1* **AND** *nkx2.4* **GENES**

We next analyzed whether Nkx2.1/4a/4b activity is required for continued expression of these genes during embryonic development. Compared to controls, in nkx2TKD morphants the expression domains of all three genes were restricted to more medial forebrain regions, most pronounced for *nkx2.4a* and *nkx2.4b* (**Figure 7**). At 2 dpf, expression of all three genes was maintained in the posterior tubercular portion of the prosomer 3 derived basal hypothalamus in nkx2TKD embryos. In contrast, expression of all three *nkx2* genes was depleted in the intermediate and caudal hypothalamus. *nkx2.1/4a/4b* expression was reduced in the preoptic region. *nkx2.1* expression domains in the alar plate preoptic telencephalon and pallidum were maintained, albeit the size of the domain and the expression level appeared slightly reduced (**Figures 7A,B**). These results suggest that *nkx2.1* and *nkx2.4a/b* activity may be required to maintain their expression in specific domains of the developing forebrain. An alternate explanation may be that some hypothalamic tissue may get lost due to apoptosis (see **Figures 3I,J**).

#### **nkx2TKD AFFECTS HYPOTHALAMIC NEUROSECRETORY POPULATIONS**

We next investigated which neuroendocrine systems were affected by loss of *nkx2.1* or *nkx2.4* activity. Expression of *pomca* (Hansen et al., 2003; Lohr and Hammerschmidt, 2011) is lost in the arcuate nucleus as well as the pituitary of nkx2TKD morphants (**Figures 8A,B**). To further characterize differential effects of nkx2TKD on ventral neuroendocrine vs. preoptic hypothalamus, we analyzed expression of the neuroendocrine hormone genes *corticotropin releasing hormone* (*crh), oxytocin (oxt),* and *arginine vasopressin* (*avp)* at 2 and 3 dpf. *crh* expression (Chandrasekar et al., 2007) in the posterior tuberculum and hypothalamus of nkx2TKD morphants was completely absent (**Figures 8C,D**), while *crh* expressing neurons in all other regions developed normally. *oxt* is exclusively expressed in the preoptic region (**Figures 8E,F**), and similar to the preoptic domain of *avp* (**Figures 8E,F**) was not affected in triple morphants. In contrast, *avp* expression in the neuroendocrine ventral hypothalamus was strongly reduced or eliminated in nkx2TKD morphants (**Figures 8G,H**). We also occasionally observed *oxt*-expressing cells at ectopic locations posterior to the domain in the anterior hypothalamus in several nkx2TKD morphants (**Figure 8F**, arrow). In summary, nkx2TKD affects neuroendocrine development selectively in the hypothalamus, but not in the preoptic region.

### **DISCUSSION**

The hypothalamus harbors many highly conserved neuroendocrine and neuromodulatory systems vital for the control of fundamental behavioral patterns and physiology. Hypothalamus organization and major developmental control centers have been well described (Puelles and Rubenstein, 2003; Moreno and Gonzalez, 2011). However, the complex organization and dynamic morphogenesis of the hypothalamus have hindered a more detailed understanding of molecular mechanisms controlling patterning and neuronal differentiation. *Nkx2.1* genes have a crucial role in development of the ventral forebrain, but analysis of mutant mice (Kimura et al., 1996; Sussel et al., 1999; Marín et al., 2002) and knockdown in *Xenopus* (van den Akker et al., 2008) have resulted in phenotypes that differentially affect the alar preoptic region and the basal hypothalamus. Here, we characterize a second *nkx2.1*-related gene in zebrafish, *nkx2.4a.*

**FIGURE 5 |** *nkx2.1***,** *nkx2.4a***, and** *nkx2.4b* **knockdown affects forebrain patterning.** Zygotes were injected with single or combinations of Nkx2 gene morpholinos as indicated in boxes above the panels, and incubated until 30 hpf **(A–D,M,N)** or 2 dpf **(E–L,O–AA)**. Expression of patterning genes were analyzed by WISH. **(A–D)** Hypothalamic *shha* expression in **(B)** *nkx2.1* morphants is largely normal, with a minor reduction (*n* = 10*.*10); **(C)**

*nkx2.4a/4b* double morphants (*n* = 10*.*10) is more severely reduced; **(D)** nkx2TKD embryos is severely reduced in the rostral forebrain (*n* = 29*.*29). **(E–H)** *lhx6*, and **(I–L)** *lhx5* expression in morphants at 2 dpf (**E**, *n* = 42*.*42; **F**, *n* = 11*.*11; **G**, *n* = 4*.*4; **H**, *n* = 67*.*67; **I**, *n* = 30*.*30; **J**, *n* = 7*.*7; **K**, *n* = 9*.*9; **L**, *n* = 23*.*23) (red arrowhead in **K** indicates affected area/reduced expression). *(Continued)*

#### **FIGURE 5 | Continued**

**(M,N)** *dbx1a* expression was not perturbed by nkx2TKD at 30 hpf. **(O,P)** nkxTKD leads to a slight anterior expansion of *foxa2* expression in basal prosomeres 1 and 2 (**O**, *n* = 22*.*22; **P**, *n* = 11*.*11). **(Q,R)** *otpa* and **(S,T)** *dlx5a* expression were affected in the posterior tuberculum and hypothalamus of nkx2TKD morphants at 2 dpf (**Q**, *n* = 34*.*34; **R**, *n* = 60*.*60; **S**, *n* = 16*.*16; **T**, *n* = 12*.*12). **(U,V)** nkx2TKD affects the expression pattern of *nkx2.2* only in the caudal hypothalamus (*n* = 14*.*14). **(W,X)** The pallial telencephalic *emx1* expression is not affected in nkx2TKD. **(Y,AA)** In nkx2TKD morphants the prethalamic *pax6* expression domain expands into the ventral hypothalamus (inset in **AA**, *n* = 13*.*13), while double *nkx2.1* and *nkx2.4b* morphant have a less severe ventral *pax6* expansion (inset in **Z**, *n* = 25*.*25). Lateral views, anterior is to the left. Abbreviations see **Table 1**. lateral views. Black arrows point at positions of anatomical structures labeled by abbreviation at start of arrow. Scale bar = 100µm in **(A)** for **(A–X)**, in **(Y)** for **(Y–AA)**.

Partial redundancies between *nkx2.1* and *nkx2.4a*, which has not been previously knocked out in mice, may explain difficulties in understanding the role of NKX2.1 factors in ventral forebrain development.

*zgc:171531 / nkx2.4a* encodes a novel zebrafish Nkx2 homeobox transcription factor of high sequence similarity to Nkx2.1 and Nkx2.4. In light of synteny and phylogenetic tree analysis, the previously reported zebrafish *nkx2.1a* is more closely related to *nkx2.4a* than to *nkx2.1*. Recent phylogenetic analysis of rainbow trout Nkx2-4 also suggested zebrafish the previously *nkx2.1a* named gene to be a *nkx2.4* paralog (Uemae et al., 2014). Analysis of *nkx2.4a* expression in relation to the *nkx2.1* genes also reveals a strong similarity to the expression pattern in the brain of the gene previously reported as *nkx2.1a. nkx2.4a* is expressed in the posterior tuberculum and hypothalamus. *nkx2.4a* is not expressed in the preoptic region or pallidum, as *nkx2.1* is. For zebrafish, our data indicate that the gene previously named *nkx2.1a* is indeed a paralog of *nkx2.4*, and the two paralogous genes will be named *nkx2.4a* and *nkx2.4b* as orthologs of mammalian *Nkx2.4*, while *nkx2.1* is the only *Nkx2.1* ortholog in zebrafish.

Expression patterns similar to zebrafish have been reported for *nkx2.4* in *Xenopus* (Small et al., 2000; Ermakova et al., 2007). Mouse *Nkx2.4* expression (Price, 1993) is also restricted to the hypothalamus and absent from the preoptic region (Marín et al., 2002). Thus, it appears that the *nkx2.4* expression pattern is conserved throughout evolution, and based on the high sequence

similarity in the homeodomain, may contribute to hypothalamic development in a redundant manner with *nkx2.1*. The two "parallel" *nkx2.1* and *nkx2.4* genes may also explain other riddles, for example *LhjTTF-1/LjNkx2.1* reported for *lampreta* to lack telencephalic expression (Ogasawara et al., 2001) may indeed be a *nkx2.4* ortholog.

Analysis of nkx2TKD embryos revealed that most *nkx2.1/4a/4b* expression in the basal hypothalamus depends on *nkx2.1/4a/4b* activity, except for the posterior tubercular domain. In contrast, it appears that the preoptic and pallidal domain do not strictly depend on *nkx2.1/4a/4b* activity. All three genes are reduced in the medio-lateral extent of their expression in nkx2TKD embryos, which would be in line with a reduced midline-derived activity like Shh required to establish and maintain these domains. Our findings are consistent with reports that *nkx2.1* expression is maintained in a smaller more medial-ventral domain in mice (Sussel et al., 1999), and in *Xenopus* (van den Akker et al., 2008) *nkx2.1* loss-of-function embryos. However, in mouse, expression of *nkx2.4* appears to strictly depend on *nkx2.1* activity (Marín et al., 2002), which would suggest changes in regulatory interactions in evolution.

While morpholino antisense oligonucleotide based knockdown of individual *nkx2.1/4* genes had only subtle effects on hypothalamus development, the combined knockdown of *nkx2.1, nkx2.4a* and *nkx2.4b* expression revealed that these genes together control specification and patterning of the basal hypothalamus. **Figure 9** gives an overview of anatomical changes as revealed by marker gene analysis in nkx2TKD morphants. Triple morphant embryos were devoid of the basal hypothalamus including the rostral half of prosomere 3 and extending to the optic commissure. The defects in the nkx2TKD basal hypothalamus are thus limited to the region of *nkx2.1/4a/4b* expression, which extends caudally into the rostral half of basal prosomer 3 (Lauter et al., 2013). In nkx2TKD morphants, the caudal half of prosomer 3 expresses *dbx1a* normally, revealing that the *nkx2.1/4* activity has no effects caudal to its expression limits.

The loss of basal hypothalamic structures upon nkx2TKD is clearly revealed by loss or reduction of the expression domains of *lhx6*, *lhx5*, and *dlx5a* in this region. This basal plate tissue is dorsalized, as the prethalamic expression domain of *pax6a* expands to the ventral edge of the forebrain. However, the ventral expansion of dorsal fates does not reflect a global expansion of alar territories, as the dorsoventral positioning of the *nkx2.2* expression domain, which correlates with the alar-basal boundary, is not shifted. Thus, our data are consistent with the previously published dorsalization of ventral prethalamus in *nkx2.1* mutant mice (Sussel et al., 1999; Marín et al., 2002). They potentially explain discrepancies reported between mouse and *Xenopus* patterning mechanisms. The *xnkx2.1* knockdown has been reported to develop a much weaker basal hypothalamus phenotype (van den Akker et al., 2008) as compared to mice. This could be due to the fact that the *Xenopus nkx2.4* gene (Ermakova et al., 2007) could compensate loss of *nkx2.1*, while in mice *nkx2.4* expression

depends on *nkx2.1* and thus the *nkx2.1* phenotype is much stronger.

(*n* = 7*.*7). **(E,F)** Formation of *oxt*-expressing cells in the PO (arrowheads) is

In nkx2TKD morphants the preoptic region is less severely affected, and the SPV appears to develop normally. Both the basal hypothalamus and the preoptic region express *nkx2.1*, except for a small alar plate region of the SPV area characterized by *Orthopedia (otp)* expression (Puelles and Rubenstein, 2003). *otp* is also expressed in the posterior tuberculum and in the arcuate nucleus region. Interestingly, in nkx2TKD morphants the *otpa* expressing domain in the preoptic SPV area was not affected, while the posterior tubercular *otpa* domain was absent. This is consistent with *otpa* in the preoptic SPV not being co-expressed with *nkx2.1/4a/4b*, while in the posterior tuberculum *nkx2.1* is coexpressed (Ryu et al., 2007). The differentiation of neuroendocrine cells in the preoptic region has been studied in detail (Kurrasch et al., 2009; Machluf et al., 2011; Herget et al., 2014) and aids in dissecting the neuroanatomical structures effected in *nkx2.1* mutants. The less severe preoptic phenotype is in line with our finding that neuroendocrine cells develop in the preoptic region of nkx2TKD morphants. In contrast, development of essentially all tested basal hypothalamic neuroendocrine cells depends on *nkx2.1/4a/4b* activity.

**Table 1**. Scale bars = 100µm in **(A1)** for **(A1–J1)** and in **(A)** for **(A2–J2)**.

The major subdivisions of the larval zebrafish telencephalon, pallium, striatum (dorsal part of area ventralis telencephali, Sdd) and pallidum (ventral part of area ventralis telencephali, Sdv), have been defined by marker gene expression including *lhx6* and *dlx2a* (Mueller et al., 2008; Ganz et al., 2011). Based on loss of *lhx6* expression, triple morphants have a severely reduced pallidum. Mammalian *lhx6* contains a highly conserved Nkx2.1 binding site in its promoter (Du et al., 2008), suggesting a direct and

evolutionary conserved regulation. The ventral expansion of the striatal *lhx5* expression domain suggests that in the absence of *nkx2.1* activity, striatal fate expands into the pallidum, resulting in a dorsalization of the ventral telencephalon similar to that observed in *nkx2.1* mutant mice (Sussel et al., 1999; Marín et al., 2002). When we analyzed *emx* expression, we did not observed an expansion of the pallium, again similar to *nkx2.1* mutant mice (Sussel et al., 1999; Marín et al., 2002), but distinct from reports for *nkx2.1* knockdown in Xenopus (van den Akker et al., 2008).

In summary, our data reveals that *nkx2.1*, *nkx2.4a*, and *nkx2.4b* genes act partially redundant in zebrafish hypothalamic development. *nkx2.1* is specifically involved in the development of rostral ventral forebrain including the pallidum and preoptic regions, but its function in the basal hypothalamus appears redundant with both *nkx2.4* genes. In contrast, *nkx2.4a* and *nkx2.4b* control aspects of basal hypothalamus development including the intermediate and caudal hypothalamus, where loss of *nkx2.4* activity is not fully compensated by *nkx2.1*.

### **AUTHOR CONTRIBUTIONS**

Wolfgang Driever and Martha Manoli designed the study. Martha Manoli performed all experiments and documentation. Martha Manoli and Wolfgang Driever carried out data analysis and wrote and edited the manuscript. Wolfgang Driever obtained funding and supervised the project.

#### **ACKNOWLEDGMENTS**

We are grateful to the zebrafish community for plasmids and probes. We thank Thomas Müller, Jörn Schweitzer, Meta Rath, and Alida Filippi for discussion and comments on the manuscript. S. Götter and R. Schlenvogt provided expert care of the fish. This work was supported by the DFG (EXC294- BIOSS), GRK1104 (Martha Manoli), and by the European Commission (FP7, mesDANEURODEV 222999, DOPAMINET 223744, ZFHEALTH 242048) (Wolfgang Driever).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fnana. 2014.00145/abstract

### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 05 October 2014; paper pending published: 21 October 2014; accepted: 14 November 2014; published online: 02 December 2014.*

*Citation: Manoli M and Driever W (2014) nkx2.1 and nkx2.4 genes function partially redundant during development of the zebrafish hypothalamus, preoptic region, and pallidum. Front. Neuroanat. 8:145. doi: 10.3389/fnana.2014.00145*

*This article was submitted to the journal Frontiers in Neuroanatomy. Copyright © 2014 Manoli and Driever. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Coexpression analysis of nine neuropeptides in the neurosecretory preoptic area of larval zebrafish

### **Ulrich Herget 1,2 and Soojin Ryu<sup>1</sup>\***

<sup>1</sup> Developmental Genetics of the Nervous System, Max Planck Institute for Medical Research, Heidelberg, Germany

<sup>2</sup> The Hartmut Hoffmann-Berling International Graduate School of Molecular and Cellular Biology, University of Heidelberg, Heidelberg, Germany

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Matthias Carl, University of Heidelberg, Germany Uwe Strähle, KIT ITG, Germany

#### **\*Correspondence:**

Soojin Ryu, Developmental Genetics of the Nervous System, Max Planck Institute for Medical Research, Jahnstr. 29, 69120 Heidelberg, Germany e-mail: soojin.ryu@ mpimf-heidelberg.mpg.de

The paraventricular nucleus (PVN) of the hypothalamus in mammals coordinates neuroendocrine, autonomic and behavioral responses pivotal for homeostasis and the stress response. A large amount of studies in rodents has documented that the PVN contains diverse neuronal cell types which can be identified by the expression of distinct secretory neuropeptides. Interestingly, PVN cell types often coexpress multiple neuropeptides whose relative coexpression levels are subject to environment-induced plasticity. Due to their small size and transparency, zebrafish larvae offer the possibility to comprehensively study the development and plasticity of the PVN in large groups of intact animals, yet important anatomical information about the larval zebrafish PVNhomologous region has been missing. Therefore we recently defined the location and borders of the larval neurosecretory preoptic area (NPO) as the PVN-homologous region in larval zebrafish based on transcription factor expression and cell type clustering. To identify distinct cell types present in the larval NPO, we also generated a comprehensive 3D map of 9 zebrafish homologs of typical neuropeptides found in the mammalian PVN (arginine vasopressin (AVP), corticotropin-releasing hormone (CRH), proenkephalin a (penka)/b (penkb), neurotensin (NTS), oxytocin (OXT), vasoactive intestinal peptide (VIP), cholecystokinin (CCK), and somatostatin (SST)). Here we extend this chemoarchitectural map to include the degrees of coexpression of two neuropeptides in the same cell by performing systematic pairwise comparisons. Our results allowed the subclassification of NPO cell types, and differences in variability of coexpression profiles suggest potential targets of biochemical plasticity. Thus, this work provides an important basis for the analysis of the development, function, and plasticity of the primary neuroendocrine brain region in larval zebrafish.

**Keywords: neuroendocrine system, hypothalamus, preoptic region, paraventricular nucleus, zebrafish, coexpression**

#### **INTRODUCTION**

The vertebrate neuroendocrine system is controlled by specific hypothalamic nuclei. The most extensively studied ones are the paraventricular nucleus (PVN) and supraoptic nucleus (SON) in rodents, which contain neuroendocrine cells expressing secretory neuropeptides. These peptides are transported to the pituitary gland, where they are directly released into the general circulation via the neurohypophysis, or released into the portal system of the adenohypophysis, where they trigger the release of hormones into the general circulation. The two peptides released in the neurohypophsis are arginine vasopressin (AVP), and oxytocin (OXT), and they are produced by distinct magnocellular neurons. Releasing and inhibiting factors acting on adenohypophyseal secretion are instead produced in parvocellular neurons, and include corticotropin-releasing hormone (CRH), thyrotropinreleasing hormone (TRH), methionine or leucine enkephalin (mENK/lENK), neurotensin (NTS), vasoactive intestinal peptide (VIP), cholecystokinin (CCK), and somatostatin (SST) (Swanson and Sawchenko, 1983; Mezey et al., 1985; Simmons and Swanson, 2009).

Coexpression of several of these peptides in the same cell is a common phenomenon in the mammalian PVN. For example, AVP colocalizes with mENK in neurohypophysial terminals in the rat (Martin and Voigt, 1981). Also in rats, CRH can be coexpressed with AVP, OXT, NTS, ENK, VIP, and/or CCK (Burlet et al., 1983; Hökfelt et al., 1983; Roth et al., 1983; Sawchenko et al., 1984a; Mezey et al., 1985; Piekut and Joseph, 1986; Swanson et al., 1986; Sawchenko, 1987; Whitnall and Gainer, 1988; Ceccatelli et al., 1989; Swanson and Simmons, 1989; Arima et al., 2001; Dabrowska et al., 2011). Cells producing OXT were found to also synthesize CRH, CCK, and ENK (Martin and Voigt, 1981; Vanderhaeghen et al., 1981; Rossier et al., 1983; Levin and Sawchenko, 1993). Not all peptide combinations occur, however. The populations of cells producing AVP, OXT, or SST are completely separate in the rat (Swanson and Sawchenko, 1983; Swanson et al., 1986).

The molecular development of the mammalian hypothalamus is largely conserved in zebrafish (Machluf et al., 2011). We have recently defined the location and borders of the larval neurosecretory preoptic area (NPO) as the PVN-homologous region based on transcription factor expression and cell type clustering in larval zebrafish. To identify distinct cell types present in the larval NPO, we also generated a comprehensive 3D map of 9 zebrafish homologs of typical neuropeptides found in the mammalian PVN (*avp*, *crh*, *oxt*, *proenkephalin a* (*penka*), *proenkephalin b* (*penkb*), *nts*, *vip*, *cck*, and *somatostatin* (*sst1.1*); Herget et al., 2014). The larval NPO featured 9 out of 10 peptides also found in the mammalian PVN. One exception was *trh*, which was found to be outside the boundaries of the larval NPO as we defined them. We found a small group of cells producing *cck* at the rostral border of the NPO and cells producing *avp*, *oxt*, *crh*, *penka*, *nts*, or *sst1.1* as dense and intermingled clusters. In contrast, cells producing *penkb* or *vip* appeared to reside in separate subregions of the NPO.

Several of these neuropeptides are coexpressed in the same cells in the mammalian PVN, and extensive coexpression also in the larval NPO seemed likely based on the spatial proximity of cells after 3D registration. We reasoned that definitive classification of distinct cell types cannot be assigned in the larval NPO based on the expression of one neuropeptide alone. Therefore, we analyzed in this study the degree of coexpression of two neuropeptides in the same cell by performing systematic pairwise comparisons of coexpression of *avp*, *crh*, *oxt*, *penka*, *penkb*, *nts*, *cck*, *vip*, and *sst1.1* in the larval zebrafish NPO. Our results show that many of the peptides produced by densely intermingled cells of the larval zebrafish NPO are not coexpressed, while some neuropeptide combinations show occasional, low or moderate levels of coexpression. Interestingly we observed high degrees of coexpression for certain neuropeptide combinations such as *avp* + *crh* and *cck* + *penkb*. These results illustrate that information about coexpressed peptides is essential to identify subclasses of cell types and the classification of cell types should not be based on the expression of one peptide alone within the NPO.

Plastic changes in the coexpression profile of a cell allow it to acquire additional regulatory functions. Indeed, in mammals the coexpression properties of PVN cells are subject to stressinduced plasticity, with different types of stress influencing the expression levels of different neuropeptides (Swanson et al., 1986; Harbuz and Lightman, 1989; Swanson, 1991). The larval zebrafish offers an attractive system to dissect the mechanistic basis of such environment-induced plasticity in the hypothalamus because of the possibility to study the architecture, development, and function of distinct neural circuits in intact and transparent animals. The basal characterization of coexpression profiles is a prerequisite to classify and identify distinct NPO cell types and their plasticity in response to environmental challenges.

## **MATERIALS AND METHODS**

#### **ZEBRAFISH PREPARATION**

Zebrafish maintenance and breeding were carried out under standard conditions at 28.5◦C (Westerfield, 2000). To avoid pigmentation, embryos were incubated in 0.2 mM 1-phenyl-2-thiourea (Sigma-Aldrich). AB/TL zebrafish larvae were fixed at 5 days post fertilization (dpf) in 4% paraformaldehyde (PFA, Merck; in phosphate buffered saline (PBS), pH 7.2–7.3) overnight. All animals were raised under constant conditions and fixed quickly to avoid chronic environmental or acute handling stress. On the following day, larvae were washed briefly with PBST (phosphate-buffered saline with 0.1% Triton X-100, Merck and Roth), then dehydrated with increasing methanol (Merck) concentrations (25%, 50%, 75%, 100%, in PBST, 5 min steps), and stored in 100% methanol at −20◦C. All procedures were performed according to the guidelines of the German animal welfare law and approved by the local government.

#### **WHOLE-MOUNT FLUORESCENT IN SITU HYBRIDIZATION**

*In situ* hybridization (ISH) probes for *avp* (Eaton et al., 2008), *oxt* (Unger and Glasgow, 2003), *sst1.1* (Devos et al., 2002), *trh*, *crh* (Löhr et al., 2009), *vip* (Wolf and Ryu, 2013), *cck*, *penka*, *penkb*, and *nts* (Herget et al., 2014) were previously described. Riboprobes were synthesized from linearized plasmids following the instructions provided with the digoxygenin labeling mix (Roche). Fluorescent ISH was performed based on a previously published protocol (Lauter et al., 2011).

#### **MICROSCOPY AND IMAGE PROCESSING**

For imaging, larval heads were cleared in 80% glycerol (Gerbu) in PBS for 1 h. Dorsal confocal stacks of larval heads were recorded using a Leica SP5 confocal microscope with a Nikon 20x glycerol objective. Each channel was recorded sequentially, using alternating excitation wavelengths specific for each tyramide, to reduce interfering signals from overlapping emission spectra. Acquisition settings were adjusted for each stack to obtain the optimal image quality of the desired volume. Stacks were evaluated using Amira 5.4 (Visualization Sciences Group) to create maximum intensity projections that were restricted to the volume of interest, excluding signals from planes above or below. Staining signal was analyzed plane by plane within the NPO. Brightness and contrast were adjusted for each channel. Any accumulation of signal with the proper shape and size of a typical cell was included in the analysis and compared to the signal in co-stained channels at the same location. Thus, coexpression was determined by the spatial overlap of cells stained for different peptide markers. Images of single planes and maximum projections were exported from Amira and arranged into figures using Adobe Illustrator. All images show dorsal views of substacks or single planes, with the rostral direction on the left side, unless indicated otherwise.

#### **RESULTS**

To comprehensively analyze the degree of coexpression of two peptides in the same cell, we performed cell by cell comparisons of pairwise combinatorial ISH staining of nine peptide markers that we had previously identified to be expressed in the 5 dpf larval NPO. The NPO is defined by the dense clustering of cells expressing these peptides within the transcription factor *orthopedia a (otpa)*-positive preoptic area (**Figure 1**). Altogether we examined 36 pairwise combinations of neuropeptides, analyzing a minimum of 5 animals per pair, but the sample size varied and was larger for some peptide combinations. All

**FIGURE 1 | Schematic lateral view of a 5 dpf larval zebrafish brain showing the location of the NPO (dashed line) within the otpa-positive part (dark gray line) of the preoptic area, and the spatial distribution of nine cell types expressing the indicated neuropeptides**. Cells clustering within the NPO are opaque. For more details, the reader is referred to our previously reported chemoarchitectural map (Herget et al., 2014). Abbreviations: ac, anterior commissure; d, dorsal; H, hypothalamus; Ha, habenula; NPO, neurosecretory preoptic area; oc, optic chiasm; PO, preoptic area; poc, postoptic commissure; PT, posterior tuberculum; PTh, prethalamus; r, rostral; Tel, telencephalon; TeO, optic tectum; Th, thalamus.

data was obtained from 5 dpf larvae, since the original NPO cell type map was generated for that stage. We found broadly three different categories of coexpression extent: (1) Absence of coexpression in all animals; (2) Occurance of coexpression in a single, few or several cells only in some animals analyzed ("variable coexpression"); (3) Coexpression in several cells in all animals analyzed ("consistent coexpression").

#### **NO COEXPRESSION IN ALL ANIMALS**

Many of the neuropeptide combinations (16/36) showed no coexpression in the same cell, and often these neuropeptides were expressed in spatially separate clusters. The rostralmost clusters formed by cells expressing *avp*, *crh*, *cck*, or *sst1.1* did not show any overlap with the caudalmost cluster formed by cells expressing *vip* (**Figures 2A–D"**, 5–8 animals analyzed). Among the rostral group, *crh* and *cck* were not coexpressed in the same cells, and *cck* expression was also separate from the large *nts*-positive cluster (**Figures 2E–F"**, 6–7 animals analyzed). The *penka*-expressing cluster extended as far caudally as the small group of *vip*-positive cells, which however occupied a more lateral region (**Figures 2G–G"**, 8 animals analyzed). *oxt*-expressing cells formed a large central cluster, and they did not overlap with the rostrally located *cck*-expressing cells, nor with the caudally located *vip*-expressing group (**Figures 2H–I"**, 6–9 animals analyzed). The *vip*-positive cluster was also separate from cells expressing *penkb* (**Figures 2J–J"**, 7 animals analyzed). The rostral cluster of *cck*-positive cells was close to, but separate from the *avp*-positive and *penka*-positive clusters (**Figures 2K–L"**, 7 animals analyzed). *penkb*-positive cells surrounded the more central *nts*-positive cluster (**Figures 2M–M"**, 5 animals analyzed). Within the center of the NPO, cells expressing *avp* or *oxt* were found intermingled in the same region, but did not overlap (**Figures 2N–O"**, 8 animals analyzed). The central and intermingled clusters of cells expressing *oxt* or *nts* did not show coexpression of these two peptides (**Figures 2P–P"**, 10 animals analyzed). Similar spatial proximity without coexpression was also found for cells expressing *oxt* or *sst1.1* (**Figures 2Q–R"**, 8 animals analyzed).

#### **VARIABLE COEXPRESSION**

Occasionally, a single cell was found coexpressing two neuropeptides in a larva, although most of the animals analyzed did not show coexpression. We observed such rarely coexpressing single cells in 10/36 neuropeptide combinations. A single *avp*positive cell was occasionally found within *nts*-positive cells (**Figure 3A**, 3/11 animals). Similarly, in rare cases, single cells showed coexpression of *crh* and *oxt* (**Figure 3B**, 3/17 animals). The clusters of cells producing *penka* or *penkb* were spatially separate, but in one animal, coexpression could be found in a single cell (**Figure 3C**, 1/16 animals). The intermingled clusters of cells expressing *penka* or *sst1.1* also in rare cases showed one cell with coexpression (**Figure 3D**, 2/14 animals). The *crh*-positive cluster occupied the rostral half of the region covered by both the *penka*-positive and *nts*-positive clusters, and in rarely occuring cells, *crh* was coexpressed with *penka* (**Figure 3E**, 4/14 animals) or *nts* (**Figure 3F**, 1/6 animals). While the *cck*-positive cluster was rostral, occasional coexpression of *sst1.1* was observed (**Figure 3G**, 3/7 animals). The *vip*-producing cluster was caudal and lateral, but a cell that was more rostromedial did in one case coexpress *nts* (**Figure 3H**, 1/9 animals). The central cluster of *penka*-positive cells was intermingled with the clusters of cells expressing *avp* or *oxt*, and occasionally single *penka*-positive cells coexpressed *avp* (**Figure 3I**, 4/21 animals) or *oxt* (**Figure 3J**, 4/33 animals).

The extent of coexpression for other peptide combinations was somewhat higher where more than one cell per animal showed coexpression. In the combination of *crh* and *penkb* staining, many animals showed no coexpression, but in one animal, coexpression was observed in few cells (**Figures 4A–A"'**, 1/12 animals). Similarly few *avp* and *penkb* coexpressing cells were found in some animals analyzed (**Figures 4B–B"'**, 5/15 animals). In those cases in which *crh* and *sst1.1* were coexpressed, we found such coexpression in several cells (**Figures 4C–C"'**, 2/9 animals). Few cells coexpressing *penkb* and *sst1.1* were also found in one animal (**Figures 4D–D"'**, 1/5 animals). Few *avp* and *sst1.1* coexpressing cells were also found in some animals (**Figures 4E–E"'**, 6/25 animals). Some animals showed few *oxt* and *penkb* coexpressing cells (**Figures 4F–F"'**, 4/7 animals). Although in some animals, *penka* and *nts* were not coexpressed, in most animals we found coexpression of these peptides in few cells (**Figures 5A–A"'**, 8/15 animals).

#### **CONSISTENT COEXPRESSION IN ALL ANIMALS**

Consistent coexpression in all animals analyzed was found for three peptide combinations. The combination of *nts* and

**FIGURE 2 | Cell type staining combinations showing no coexpression**. **(A)** Cells expressing avp **(A')** or vip **(A")** form neighboring but separate clusters. **(B)** Cells expressing crh **(B')** or vip **(B")** are similarly separated. **(C)** Cells expressing cck cluster rostrally **(C')**, and are therefore distant from vip-positive cells **(C")**. **(D)** Cells expressing sst1.1 **(D')** also cluster in a separate region from vip-positive cells **(D")**. **(E)** cck-positive cells **(E')** are

rostrally neighboring crh-positive cells **(E")**. **(F)** Cells expressing cck **(F')** or nts **(F")** are separate. **(G)** penka-positive cells **(G')** are scattered, but do not overlap with vip expression **(G")**. **(H)** oxt-positive **(H')** and cck-positive cells **(H")** are also separated. **(I)** The clusters formed by cells expressing oxt **(I')** or vip **(I")** do not show coexpression. **(J)** vip **(J')** and penkb **(J")** are also not (Continued)

#### **FIGURE 2 | Continued**

coexpressed. **(K)** The rostral location of cck-positive cells **(K')** is separate from the neighboring avp-positive cluster **(K")**. **(L)** penka-positive cells **(L')** also are close to the cck-positive population **(L")**, but still separate. **(M)** Cells positive for penkb **(M')** surround the central nts-positive cluster **(M")**. **(N)** Cells expressing avp **(N')** or oxt **(N")** are intermingled, but the peptides are not coexpressed (single planes: **O–O"**). **(P)** Cells expressing nts **(P')** or oxt **(P")** have similar locations, but these peptides are not coexpressed. **(Q)** Cells expressing oxt **(Q')** or sst1.1 **(Q")** are also intermingled and do not coexpress these peptides (single planes: **R–R"**). All images show maximum intensity projections, unless indicated otherwise. Scale bar: 50 µm.

*sst1.1* showed consistent coexpression in few cells (**Figures 5B– B"'**, 5/5). High degree of coexpression was consistently found for the rostralmost clusters formed by cells producing *penkb* or *cck* (**Figures 5C–C"'**, 13/13 animals). Coexpression was also consistently high for the combination of *avp* and *crh* staining (**Figures 5D–D"'**, 19/19 animals). These results are summarized in **Figure 6**, in which we also indicate the range of coexpression observed (minimum and maximum degrees).

### **DISCUSSION**

Our results show that many of the peptides produced by densely intermingled cells of the larval zebrafish NPO are not coexpressed. Occasional coexpression in one cell can be observed for other peptide staining combinations, and some combinations with low or moderate coexpression appear more frequently. Only three of the 36 combinations show consistent coexpression in all animals analyzed, reaching a high degree only for *avp* + *crh* and *cck* + *penkb*. It should be noted that the presence of RNA does not necessarily mean that the peptide will also be synthesized. For example, there is a distinct cluster of *avp*-expressing cells outside the NPO in the ventral hypothalamus, which can clearly be labeled by ISH, but not by immunohistochemistry (IHC; Eaton et al., 2008; Herget et al., 2014). Still, the information presented here about coexpressed peptide transcripts suggests the existence of subclasses of cell types which can have very different functions.

The large amount of information available on the degree of coexpression of different neuropeptides in the mammalian PVN allows comparisons of our results with those obtained in mammals. Interestingly, many peptide combinations show similar degrees of coexpression both in larval zebrafish and in mammals. The absence of coexpression of *avp* with *oxt* we observed is in line with rat data (Swanson and Sawchenko, 1983). The coexpression of *avp* with *crh* we found to vary between moderate and high levels was also shown to be low or high in the rat (Sawchenko et al., 1984b; Whitnall et al., 1985, 1987; Aubry et al., 1999; Arima et al., 2001; Simmons and Swanson, 2009),

**FIGURE 3 | In some cell type staining combinations, occasional and low coexpression can be observed**. **(A)** Cells expressing avp or nts only overlap in rare cases. **(B)** Cells expressing oxt or crh are usually separate, but occasially these peptides are coexpressed in a single cell. **(C)** Cells expressing penkb cells usually surround the penka-positive cluster, but in few animals, one cell shows coexpression. **(D)** penka-positive cells and sst1.1-positive cells are intermingled, and occasionally show low coexpression. **(E)** crh-positive cells are intermingled with penka-positive cells, but some rare occurences of coexpression were found. **(F)** crh can

also be coexpressed with nts in rare cases. **(G)** cck-positive cells appear to faintly coexpress sst1.1 in few of the animals. **(H)** One isolated cell was sometimes found coexpressing vip and nts. **(I)** Cells expressing avp or penka are intermingled and usually these peptides are not coexpressed, but single coexpressing cells do occur. **(J)** Stainings for oxt and penka often show no coexpressing cells, but sometimes these peptides are coexpressed in few cells. Images show maximum intensity projections, insets show split channels of single confocal planes with coexpressing cells (arrowheads). Scale bar: 50 µm.

**(A')** single plane) of crh **(A")** and penkb **(A''')** rarely occurs, but can be found in more than one cell. **(B)** In some animals, coexpression (**(B)** maximum intensity projection; **(B')** single plane) of penkb **(B")** and avp **(B''')** can be found. **(C)** Moderate coexpression (**(C)** maximum intensity projection; **(C')** single plane) of sst1.1 **(C")** and crh **(C''')** can be

penkb **(D''')**. **(E)** Coexpression is found in few cells (**(E)** maximum intensity projection; **(E')** single plane) labeled for sst1.1 **(E")** and avp **(E''')**. **(F)** In some animals, moderate coexpression (**(F)** maximum intensity projection; **(F')** single plane) of penkb **(F")** and oxt **(F''')** can be found. Arrowheads mark coexpressing cells. Scale bar: 50 µm.

and was low in the mouse or sheep (Rivalland et al., 2005; Biag et al., 2012). Recently, it was found that only magnocellular CRH-positive cells coexpress AVP in the rat (Dabrowska et al., 2013). Coexpression of CRH and AVP was also found in other teleosts (Yulis and Lederis, 1987; Olivereau et al., 1988; Fryer, 1989). We found rare coexpression of *oxt* with *penka*, and low coexpression of *oxt* with *penkb*, and coexpression of OXT with ENK was also seen in the rat (Martin and Voigt, 1981; Rossier et al., 1983). OXT and SST do not overlap in the rat (Swanson and Sawchenko, 1983), and we saw the same absence of coexpression in the fish. We observed rare coexpression of *crh* with *penka*, and low coexpression of *crh* with *penkb*, while ENK coexpression with CRH seems to be higher in rats (Hökfelt et al., 1983; Ceccatelli et al., 1989; Pretel and Piekut, 1990), or sheep (Rivalland et al., 2005). However, one more recent study also found only low coexpression of ENK in CRH-producing cells in the rat (Dabrowska et al., 2013), which is more in line with our results. NTS and ENK are coexpressed in low degrees in the rat (Ceccatelli et al., 1989), and we also saw consistent coexpression of *nts* with *penka*, but not with *penkb*. It was reported that very few NTS cells coexpress VIP in the rat (Ceccatelli et al., 1989), and the coexpression of *nts* and *vip* was also rare here.

Other peptide coexpression profiles we observed in larval zebrafish deviate from previously established mammalian data. AVP and ENK are thought to be coexpressed in the rat (Martin and Voigt, 1981), and moderately coexpressed in the sheep (Rivalland et al., 2005), but *penka* was only rarely coexpressed with *avp*, and *penkb* showed only low coexpression with *avp* in our case. We did not find coexpression of *avp* with *cck*,

but there are many such cells in the rat (Mezey et al., 1986). Our observation showed low coexpression of *avp* with *sst1.1*, but those are non-overlapping cell types in the rat (Swanson

(**(B)** maximum intensity projection; **(B')** single plane) of sst1.1 **(B")** and nts **(B''')**. **(C)** High coexpression can be found

> and Sawchenko, 1983). *oxt* coexpression with *crh* was rare here, but variably low or high in the rat (Sawchenko et al., 1984a; Dabrowska et al., 2013), and low in the mouse (Biag et al., 2012).

crh-positive clusters **(D''')**. Arrowheads mark coexpressing cells.

Scale bar: 50 µm.


**FIGURE 6 | Overview of coexpression profiles of cells expressing avp**, **oxt**, **crh**, **penka**, **penkb**, **nts**, **cck**, **vip, or sst1.1**. Coexpression degrees indicate complete absence (−), rare occurrence (+), low coexpression (++), or high coexpression (+++). In many combinations, coexpression was variable, and the minimum and maximum degrees are indicated (min/max). Differences between the coexpression of one peptide in one cell type and vice versa are caused by differences in cell cluster sizes.

In the zebrafish, one study reported colocalization of *oxt* and *crh* in the preoptic area (Chandrasekar et al., 2007), but with our increased resolution, we can clarify that those cells only reside in the same region, and are in fact intermingled. OXT and CCK are coexpressed in the rat (Vanderhaeghen et al., 1981; Bondy et al., 1989; Levin and Sawchenko, 1993), but we saw no such coexpression in the zebrafish. CRH coexpression with NTS was reported to be low (Sawchenko et al., 1984a) or moderate in the rat (Ceccatelli et al., 1989), but according to our results coexpression only occurs rarely. CRH and CCK are coexpressed in some cells in the rat (Mezey et al., 1986; Ceccatelli et al., 1989), but not at all in our data. CRH and VIP were reported to be coexpressed in low levels in the rat (Ceccatelli et al., 1989), but they are not coexpressed here. ENK and VIP are coexpressed in the rat (Hökfelt et al., 1987), but neither *penka* nor *penkb* overlap with *vip* in our results. NTS and CCK show very low coexpression in the rat (Ceccatelli et al., 1989), but no coexpression in our results. All observed differences can originate from differences in both species and age, since we compare larval zebrafish data with adult mammalian results. Changes in coexpression levels during continued development from the larval stage detailed here into adulthood can also be expected.

To our knowledge, several peptide combinations were never addressed as far as coexpression is concerned. *avp* coexpression with *nts* was not discussed before, and rarely occurs here. Also, *crh* coexpression with *sst1.1* was not addressed before, and is rare here. We found that *penka* is not coexpressed with *cck*, which was not addressed before. Coexpression of *penkb* with *cck* was high, and is reported here for the first time. Similarly, coexpression of *penka* or *penkb* with *sst1.1* was not reported before, and we found rare or low coexpression. Also the coexpression of *sst1.1* and *nts* is shown here for the first time. Coexpression of *cck* and *vip* was never addressed before, and we found that they do not coexpress. We also report rare coexpression of *cck* with *sst1.1* for the first time. Lastly, we also show that the expression of *vip* does not overlap with that of *oxt* or *sst1.1*.

The rarely observed coexpression in only few cells or few animals is a deviation from the otherwise complete absence of coexpression seen in most cells or animals for these combinations of stainings. Potential causes of such rare coexpression could be transitional stages in the development of these cells, mismapping due to lack of resolution, or stochastic variation. The confocal microscopy used allowed sufficient depth resolution to spatially separate cells in all three dimensions and therefore mismapping is unlikely.

The high degree of coexpression of *avp* and *crh* observed has immediate functional implications. Similar to CRH, AVP stimulates adrenocorticotropic hormone (ACTH) secretion (Gillies and Lowry, 1979; Rivier and Vale, 1983). A wide range of coexpression levels was reported in different fish species. Coexpression of *avp* and *crh* was very high, reported as 100%, in *Catostomus*, but absent in *Anguilla* (Yulis and Lederis, 1987; Olivereau and Olivereau, 1990). In humans, AVP/CRH coexpression increases with age, and a connection with stress has been implied (Raadsheer et al., 1993). In addition to CRH and AVP, CCK can also stimulate ACTH release, and CCK acting in concert with AVP has a similar ACTH-releasing potency as CRH alone (Mezey et al., 1986). We saw no coexpression of *cck* with *avp* or *crh*. In contrast to CRH, AVP, and CCK, OXT generally inhibits stress responses, while local, endogenous OXT potentiates hypothalamopituitary-adrenal (HPA) axis activity, suggesting a dual mechanism of OXT released within the PVN (Neumann, 2002). In our results, coexpression of *oxt* and *crh* was very rare, and only present in single cells when it occurred. The spatial proximity of the densely intermingled cells positive for *oxt* or *crh* suggests that local release of Oxt could have immediate effects on *crh*-producing cells. ENK, CRH, and AVP were found to be colocalized within the same secretory vesicles (Hisano et al., 1987), suggesting that coexpressed peptides are packaged into common vesicles and coreleased at the synapse. We found that either *penka* or *penkb* can be coexpressed in both *crh*-positive and *avp*-positive cells. The concerted action of coreleased CRH and ENK is thought to fine-tune stress regulation (Pretel and Piekut, 1990).

In the rat, CRH-positive innervation of the neurohypophysis was suggested by observations of CRH in neurohypophyseal terminals (Bloom et al., 1982). The coexpression of CRH in magnocellular cells innervating the neurohypophysis could allow direct CRH release to the general circulation. The stress axis is also affected by VIP (Westendorf et al., 1983; Tilders et al., 1984), and NTS (Gudelsky et al., 1989). In the parvocellular PVN, coexpression of NTS and CRH was found, and the fractions of CRH-positive cells that coexpress AVP or NTS form different discrete subsets (Sawchenko et al., 1984a). The reported segregation of cells expressing AVP or CRH in rats cannot be found in larval zebrafish, and *vip* is not coexpressed with *crh* here, but we did find coexpression of *nts* and *crh*.

The CRH-producing cells coexpressing AVP were reported to be stress-responsive, while the AVP-negative CRH-producing cells do not respond to stress (Whitnall, 1993). Under osmotic stress, ENK and CRH levels are increased, but restraint and swimming stress only elevated CRH, not ENK (Harbuz and Lightman, 1989). Such findings demonstrate the functional implications of coexpression differences. For a cell, switching on the coexpression of another peptide is an elegant strategy to change its function without the need for major structural rearrangement of input or target projections, which could impose much stronger metabolic demands. Neurochemical switching in response to alterations in environmental conditions has been suggested as a relevant biological phenomenon in PVN neurons (Kiss, 1988; Swanson, 1991). Dynamic adaptation of neuroendocrine transcription to altered supply or demand for neuropeptides has also been suggested in zebrafish larvae (Kurrasch et al., 2009). With the coexpression profiles established here for the larval zebrafish under basal conditions, alterations triggered by environmental challenges can be studied in a model organism that has important advantages for the analysis of neural structure and function in intact and genetically tractable animals.

#### **ACKNOWLEDGMENTS**

We are grateful for helpful comments and critical discussion of the data by Rodrigo De Marco, Jose Arturo Gutierrez-Triana, and Colette vom Berg. We appreciate the suggestions and critical reading of the manuscript by Mario Wullimann. We also thank Regina Singer for technical assistance, Gabi Shoeman, Christiane Brandel and Angelika Schoell for fish care, and Mona Friedrich for additional staining and imaging.

#### **REFERENCES**


*Spring Harb. Symp. Quant. Biol.* 48(Pt. 1), 393–404. doi: 10.1101/sqb.1983.048. 01.043


median eminence in normal rats. *Neuroendocrinology* 45, 420–424. doi: 10. 1159/000124768


**Conflict of Interest Statement**: The Guest Associate Editor Gonzalo Alvarez-Bolado and Review Editor Matthias Carl declare that, despite being affiliated to the same institution as author Ulrich Herget, the review process was handled objectively and no conflict of interest exists. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 09 November 2014; accepted: 07 January 2015; published online: 12 February 2015*.

*Citation: Herget U and Ryu S (2015) Coexpression analysis of nine neuropeptides in the neurosecretory preoptic area of larval zebrafish. Front. Neuroanat. 9:2. doi: 10.3389/fnana.2015.00002*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Herget and Ryu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

# Role of developmental factors in hypothalamic function

#### Jakob Biran , Maayan Tahor, Einav Wircer and Gil Levkowitz \*

Departments of Molecular Cell Biology, Weizmann Institute of Science, Rehovot, Israel

The hypothalamus is a brain region which regulates homeostasis by mediating endocrine, autonomic and behavioral functions. It is comprised of several nuclei containing distinct neuronal populations producing neuropeptides and neurotransmitters that regulate fundamental body functions including temperature and metabolic rate, thirst and hunger, sexual behavior and reproduction, circadian rhythm, and emotional responses. The identity, number and connectivity of these neuronal populations are established during the organism's development and are of crucial importance for normal hypothalamic function. Studies have suggested that developmental abnormalities in specific hypothalamic circuits can lead to obesity, sleep disorders, anxiety, depression and autism. At the molecular level, the development of the hypothalamus is regulated by transcription factors (TF), secreted growth factors, neuropeptides and their receptors. Recent studies in zebrafish and mouse have demonstrated that some of these molecules maintain their expression in the adult brain and subsequently play a role in the physiological functions that are regulated by hypothalamic neurons. Here, we summarize the involvement of some of the key developmental factors in hypothalamic development and function by focusing on the mouse and zebrafish genetic model organisms.

#### Edited by:

Valery Grinevich, German Cancer Research Center (DKFZ), Germany

#### Reviewed by:

Soojin Ryu, Max Planck Institute for Medical Research, Germany Françoise Muscatelli, Institut National de la Santé et de la Recherche Médicale, France

#### \*Correspondence:

Gil Levkowitz, Department of Molecular Cell Biology, Weizmann Institute of Science, PO Box 26, Rehovot 76100, Israel gil.levkowitz@weizmann.ac.il

> Received: 09 February 2015 Accepted: 27 March 2015 Published: 21 April 2015

#### Citation:

Biran J, Tahor M, Wircer E and Levkowitz G (2015) Role of developmental factors in hypothalamic function. Front. Neuroanat. 9:47. doi: 10.3389/fnana.2015.00047 Keywords: homeostasis, neuroendocrine, Otp, SIM1, PAC1, SF-1, zebrafish model system, neuropeptides

### Introduction

The hypothalamus is a key regulator of homeostasis in animals. It does so by integrating internal and external sensory signals, processing them, then exerting regulatory autonomic signals and neuroendocrine releasing peptides to maintain homeostasis (Pearson and Placzek, 2013). Hypothalamus-related neuropeptides were identified in ganglions of lower evolutionary animals such as corals and clams (Twan et al., 2006; Takayanagi and Onaka, 2010) and ontogenetic conservation of hypothalamus-related structures in the brains of Annelids and zebrafish has been demonstrated (Tessmar-Raible et al., 2007). In vertebrates, the hypothalamus resides ventrally to the thalamus, dorsally to the anterior pituitary and is structurally composed of several nuclei of interconnected cell populations. Each nucleus contains several neuronal types, and these work in an orchestrated manner within and between nuclei to regulate physiological functions including metabolism, water balance, satiety, reproductive physiology, circadian rhythm, and emotional responses (Machluf et al., 2011; Pearson and Placzek, 2013). Developmental abnormalities of the hypothalamus are associated with alterations in body growth and sexual development leading to adverse physiological and neurological conditions such as infertility, depression, chronic stress, autism and obesity (Michaud, 2001; Swaab, 2004; Silveira et al., 2010). Moreover, interaction of hypothalamic neurons with neighboring neuronal, astroglial and endothelial cells is highly important for sensing blood-borne hormones and metabolites. Failure to generate these interactions might lead to impairment in negative feedback signals, release of trophic neuropeptides and abnormalities in the structure of the neurohypophysial portal system (de Seranno et al., 2010; Gutnick et al., 2011).

Major efforts in recent years have been focused on the roles of hypothalamic transcription factors (TF) and signaling molecules, as well as identification and characterization of the molecular and biochemical mechanisms they regulate during the organization of distinct hypothalamic regions, their differentiation, and circuit connectivity (reviewed by Machluf et al., 2011; Pearson and Placzek, 2013). Interestingly, some of these essential developmental factors are also expressed in the mature hypothalamus (Bardet et al., 2008; Tolson et al., 2010). A few studies have directly addressed the non-developmental role of these factors in the proper functioning of mature hypothalamic nuclei, suggesting that proper regulation by these factors is essential for normal activation of the hypothalamus. These findings support the hypothesis that developmental and post-developmental impairment of these components may lead to hypothalamus-related disorders, such as infertility, obesity, depression and chronic stress.

Here, we summarize some key developmental factors involved in hypothalamic development and function. The periventricular zone of the hypothalamus contains several nuclei including the paraventricular nucleus (PVN), arcuate nucleus (Arc), supra-chiasmatic nucleus, and the anterior periventricular nucleus (aPV; Szarek et al., 2010). These hypothalamic areas are well characterized for their physiological roles, allowing the investigation of both developmental and functional regulation. Hence, we have focused our review mainly on TFs expressed in this brain region. To gain better insights on the molecular mechanisms conveyed by these TFs, we mainly focused on the mouse and zebrafish genetic models, in which specific genetic perturbations have unveiled the functions of these factors. Lastly, we discuss the possible link between factors that regulate hypothalamic development to neurodevelopmental disorders that disrupt both physiological and psychological homeostasis.

### Comparative Neuroanatomy of the Hypothalamus

In recent years, several publications emphasized the importance and relevance of non-mammalian model organisms for the study of hypothalamic development with zebrafish as the prominently utilized model (reviewed in Machluf et al., 2011; Pearson and Placzek, 2013; Wircer et al., in press). The general organization of the vertebrate brain is evolutionarily conserved, however several events during tetrapod and mammalian evolution led to neuroanatomical changes (Suárez et al., 2014). Importantly, it has been shown that key genetic factors driving the patterning and specification of major hypothalamic nuclei are evolutionarily conserved (Tessmar-Raible et al., 2007). Hence, understanding the neuroanatomical homology of the various hypothalamic nuclei is important in order to integrate the information, which has been obtained from various model organisms.

One approach for comparative identification of the hypothalamic nuclei is based on expression patterns of mRNAs and proteins of evolutionarily conserved TFs and neuropeptides. In this regards, we will discuss findings from two prevalent genetic models namely the mouse and zebrafish, focusing on the Arc, PVN and supraoptic nuclei (SON) of the mouse hypothalamus and the ventral zone of the periventricular hypothalamus (Hv), neurosecretory preoptic area (NPO) and ventral posterior tuberculum (vPT) of the zebrafish. Schematic localization of these nuclei in the adult brains of these animals is illustrated in **Figure 1**.

Several studies have demonstrated that the Hv of the zebrafish (also known as nucleus lateralis tuberis; NLT) and the mammalian Arc are homologous as both nuclei express Neurokinin B (Ramaswamy et al., 2010; Biran et al., 2012; Ogawa et al., 2012), Kisspeptin (Ramaswamy et al., 2010; Servili et al., 2011; Ogawa et al., 2012), GHRH (Farhy and Veldhuis, 2004; Castro et al., 2009), αMSH and AgRP (Forlano and Cone, 2007; Guzmán-Ruiz et al., 2014). The NPO of the fish and its homologous mammalian PVN were shown to express oxytocin (OXT; Wang and Lufkin, 2000; Goodson et al., 2003; Unger and Glasgow, 2003; Löhr et al., 2009; Gutnick et al., 2011; Fernandes et al., 2013; Herget et al., 2014), Arginine vasopressin (AVP; Wang and Lufkin, 2000; Eaton et al., 2008; Löhr et al., 2009; Fernandes et al., 2013; Herget et al., 2014), Corticotropin-releasing hormone (CRH; Wang and Lufkin, 2000; Löhr et al., 2009; Amir-Zilberstein et al., 2012; Fernandes et al., 2013; Herget et al., 2014) and Somatostatin

FIGURE 1 | Hypothalamic nuclei in vertebrates. Schematic lateral view of the zebrafish (A) and mouse (B) brains representing the projected 2D anatomy of multiple sagittal planes. Color matched areas represents the presumed homology between specific hypothalamic areas of zebrafish and mouse (see text). Arc, arcuate nucleus; CC, crista cerebellaris; CCe, corpus cerebelli; Hv: ventral zone of periventricular hypothalamus; Hc, caudal zone of periventricular hypothalamus; NPO, neurosecretory preoptic area; OB, olfactory bulb; PT, posterior tuberculum; PVN, paraventricular nucleus; SON, supraoptic nucleus; TeO, tectum opticum; VMN, ventromedial nucleus.

(SST; Wang and Lufkin, 2000; Blechman et al., 2007; Russek-Blum et al., 2008; Löhr et al., 2009; Fernandes et al., 2013; Herget et al., 2014).

A recent study illustrates the high homology between the zebrafish NPO and the mammalian PVN by both peptidergic and specific NPO TFs (Herget et al., 2014). However, no piscine hypothalamic nucleus is recognized as homologous to the mammalian SON, a key hypothalamic nucleus expressing the neurohypophyseal hormones OXT and AVP (Wircer et al., in press). Moreover, the piscine NPO was recently suggested as the common evolutionary ancestor of the vertebrate magnocellular neuronal cluster, which later anatomically partitions to generate the PVN and SON in mammals (Gutnick et al., 2011; Herget et al., 2014; Knobloch and Grinevich, 2014). However, since mammalian PVN and SON neurons differ in their origin from the preoptic area neurons (Altman and Bayer, 1978a,b; Markakis, 2002), this hypothesis should be carefully considered. Interestingly, the PT in fish is another brain region that is not considered a classical hypothalamic region (Wullimann and Rink, 2001, 2002), but a careful look at the literature might suggest otherwise. Firstly, the periventricular zone of PT (vPT) was shown to express neuropeptides which are characteristic of the mammalian periventricular hypothalamus such as AVP (Wang and Lufkin, 2000; Hatae et al., 2001; Goodson et al., 2003; Eaton et al., 2008; Fernandes et al., 2013), CRH (Wang and Lufkin, 2000; Amir-Zilberstein et al., 2012; Fernandes et al., 2013) and Neurokinin B (Hatae et al., 2001; Biran et al., 2012). Secondly, catecholaminergic (tyrosine hydroxylase expressing) cells of the vPT were suggested to be homologous to hypothalamic group A11 of dopaminergic cells (Ryu et al., 2007; Löhr et al., 2009; Filippi et al., 2014). Notably, Puelles and Rubenstein define the boundary between the caudal hypothalamus and diencephalic prosomere 3 in rodents by the expression of several genes, including the TFs Single minded (Sim1) and Orthopedia (Otp), which are expressed exclusively in the hypothalamus (Puelles and Rubenstein, 1993, 2003). As both Sim1 and Otp are expessed in the zebrafish vPT (Borodovsky et al., 2009; Löhr et al., 2009; Fernandes et al., 2013) we propose a revised prosomere subdivision of the zebrafish forebrain (**Figure 2**), in which the boundary between the caudal hypothalamus and prosomere 3 is shifted to the caudal limit of Sim1 and Otp domains, so that the vPT is regarded as part of the teleostian hypothalamus. Future comparative gene expression analyses, fate-mapping experiments and other comparative anatomy experiments are required to further establish this refined model.

## Transcription Factors Regulating the Development of the Vertebrate Hypothalamus

The hypothalamus contains anatomical partitioning and its various neuronal cell populations form elaborate connectivity with virtually all parts of the nervous system. This raises many questions regarding the mechanisms that underlie the development of hypothalamic brain nuclei and the

specification of the neuronal populations that inhabit the hypothalamus.

### Otp−

The homedomain-containing TF Otp is well conserved across species. The deduced amino acid sequence of the homeodomain of the human protein is 99% homologous to the mouse Otp, and demonstrates high degree of conservation when compared to sea urchin, drosophila (Lin et al., 1999) and planaria (Umesono et al., 1997). Additionally, the existence of several evolutionarily conserved non-coding sequences (ECR) was recently demonstrated in the Otp promoter by Gutierrez-Triana et al. (2014). Using zebrafish as their model, the authors have further demonstrated that OtpaECR6 specifically regulates the expression of Otp in the NPO of the zebrafish (Gutierrez-Triana et al., 2014). Taken together, this conservation suggests an evolutionarily conserved functional role for Otp in vertebrates. Otp is expressed in conserved hypothalamic domains (Simeone et al., 1994; Bardet et al., 2008; Del Giacco et al., 2008), where it plays an important role in the differentiation of several neurohormone—secreting nuclei including the aPV, PVN, SON, Arc and ventromedial nucleus (VMN; Acampora et al., 1999; Blechman et al., 2007; Eaton et al., 2008). In the zebrafish, the expression of Otp in the NPO and PT is regulated by the zinc-finger-containing TF Fezf2 (Blechman et al., 2007; Machluf et al., 2011; Yang et al., 2012; Wolf and Ryu, 2013). Non-hypothalamic expression of Otp is also detected in the medial amygdaloid nucleus (MeA), hindbrain and spinal cord (Simeone et al., 1994; Acampora et al., 1999). Importantly, Otp positive neurons inhabiting the murine MeA are of diencephalic origin. These neurons are generated in the hypothalamus and migrate during brain development through the diencephalic-mesencephalic junction into their final position in the MeA (García-Moreno et al., 2010).

It has been demonstrated that Otp is crucial for proper development of diencephalic dopaminergic neurons in zebrafish and mouse. Zebrafish embryos lacking the Otpa protein are devoid of dopaminergic neurons in the hypothalamus and the PT while overexpression of Otp can induce ectopic expression of dopaminergic markers, such as TH and dopamine transporter indicating that Otp can instruct dopaminergic identity (Ryu et al., 2007; Fernandes et al., 2013). Otp cooperates with another TF, Sim1 (see below) to regulate the expression of TH, CRH, TRH, SST, OXT and AVP in the NPO and PT of zebrafish (Eaton et al., 2008; Borodovsky et al., 2009; Löhr et al., 2009; Fernandes et al., 2013).

As in the zebrafish, Otp−/<sup>−</sup> mouse embryos lack diencephalic dopaminergic neurons of the diencephalospinal dopaminergic system (Acampora et al., 1999; Wang and Lufkin, 2000). Homozygous Otp−/<sup>−</sup> mutant mice die soon after birth and display progressive impairment of crucial neuroendocrine developmental events such as reduced cell proliferation, abnormal cell migration and failure in terminal differentiation of neurons of the PVN, SON, and Arc (Acampora et al., 1999). Further analysis of Otp mutants revealed that Otp contributes to the patterning of the hypothalamus and preoptic region, and is required for differentiation of specific OXT, AVP, CRH and SST expressing cells (Acampora et al., 1999; Wang and Lufkin, 2000). Analysis of Otp−/<sup>−</sup> mouse embryos demonstrated that Otp-expressing cells fail to properly migrate from the hypothalamus to the amygdaloidal complex, leading to structural impairments in several amygdaloidal nuclei (García-Moreno et al., 2010). This data suggest that in addition to its role as a developmental regulator of several neuroendocrine lineages, Otp is also an important regulator of migratory processes of other neuronal populations.

### Sim1 and Arnt2−

Sim1 and Arnt2 are two PAS (PER-Arnt-Sim) containing TFs belonging to the large basic loop-helix-loop (bHLH) family of TFs (Ema et al., 1996; Fan et al., 1996). Sim proteins present considerable sequence divergence from the Drosophila melanogaster protein. Sim form a heterodimeric protein complex with the aryl hydrocarbon receptor nuclear translocator (Arnt) to activate or repress their target genes containing the so called central midline enhancer (CME) or hypoxic response element (HRE) repeats (Moffett and Pelletier, 2000; Woods et al., 2008). Data from mice and zebrafish suggest that the heterodimeric Sim1-Arnt2 complex regulates hypothalamic differentiation in vivo (Michaud et al., 2000; Löhr et al., 2009).

Sim1 null and Arnt2 null mice die shortly after birth and analysis of newborn brains indicates the lack of the hypothalamic aPV, PVN, and SON nuclei, phenocopying the Otp null phenotype (Michaud et al., 1998; Keith et al., 2001). However, Otp null mice show a dramatic decrease of almost 30% in brain size (Wang and Lufkin, 2000) while Sim1, Sim2 and Arnt2 null mice display developmental impairments that are correlated with deficits in neuronal migration and differentiation (Michaud et al., 1998, 2000; Goshu et al., 2004).

Sim1, Arnt2 and Otp function along parallel pathways, as they are all required for Sim2 expression in the PVN for the differentiation of the neurons secreting TRH, and in the aPV for the differentiation of the neurons that secrete SST. In the PVN and SON nuclei, these TFs are required for the maintenance of Brn2 expression, a POU domain TF necessary for the development of OXT, AVP, and CRH producing neurons (Schonemann et al., 1995; Michaud et al., 1998, 2000; Acampora et al., 1999; Wang and Lufkin, 2000; Keith et al., 2001; Goshu et al., 2004).

### Steroidogenic Factor 1 (SF-1)

The VMN of the hypothalamus is involved in the regulation of many homeostatic functions, such as the maintenance of energy balance, sexual behavior, anxiety and circadian rhythms (McClellan et al., 2006; Cheung et al., 2013). However, the development of the VMN is less characterized in comparison to the highly studied periventricular zone of the hypothalamus. The finding that SF-1 null mice lack a recognizable hypothalamic VMN, now allows the investigation of neuroanatomical and functional development of this non-peptidergic hypothalamic nucleus. SF-1 is an orphan nuclear receptor (also known as adrenal 4-binding protein; Ad4BP or Nr5a1). It was originally identified as a transcriptional regulator of cytochrome P450 steroid hydroxylases enzymes that are involved in the biosynthesis of steroid hormones (Omura and Morohashi, 1995; Parker and Schimmer, 1995). It is highly conserved both in structure and function throughout evolution (Luo et al., 1994; Achermann et al., 1999; Takase et al., 2000; Allen and Spradling, 2008). SF-1 is expressed in steroidogenic cells in the gonads, adrenal cortex and spleen as well as in the VMN and the anterior pituitary. At E11.0 SF-1 is expressed by diencephalic cells which will form the VMN (Ikeda et al., 2001; Tran et al., 2003). It is expressed as early as E9.0 during embryonic development of the mouse by cells which at later stages, will form the gonads in both sexes. At around E12.5, it shows sexual di-morphism with a higher expression in testes than ovaries (Luo et al., 1994; Ikeda et al., 2001; Sekido and Lovell-Badge, 2008). Notably, SF-1 is important for endocrine cell-fate specification (Lee et al., 2011). Mice lacking SF-1 expression die shortly after birth of adrenocortical insufficiency (Luo et al., 1994; Parker et al., 2002). These knock-out mice do not develop an adrenal gland or gonads. Similar defects in adrenal and gonadal development are also apparent in humans who carry mutations in the sf-1 gene (Luo et al., 1994; Achermann et al., 1999). Disruption of SF-1 also results in structural and neuronal connectivity alterations in the VMN (Shinoda et al., 1995; Tran et al., 2003; Zhao et al., 2008; Cheung et al., 2013). Yet it seems that in the absence of functional SF-1 the initial migration and proliferation of the neuronal precursors remains unaffected, whereas it is important for terminal differentiation of these hypothalamic neurons (Tran et al., 2003). Interestingly, in vitro analysis of various promoters reveals potential SF-1 binding sites in the fezf1, A2bp1, Nkx2-2, Slitrk1 and Slitrk5 genes that are involved in neuronal differentiation and patterning (Kurrasch et al., 2007).

## Receptors and Ligands Regulating Hypothalamic Development

The migration of a specific neuronal type into the hypothalamus, settlement of specific neuronal populations in distinct hypothalamic nuclei as well as their proper connectivity with a variety of target sites further requires extrinsic signals such as growth factors, neuropeptides and their receptors. These act concomitantly with the aforementioned intrinsic TFs to regulate the above processes. Moreover, several recent findings indicate that some neuropeptides, are involved in the development of neural circuits in which they function in the mature hypothalamus. This suggests that at least some neuropeptides act as ''developmental autoregulators''. This section demonstrates the importance of these extrinsic factors for the patterning and assembly of neurocircuits in the developing hypothalamus.

### Extrinsic Developmental Factors

Sonic hedgehog (SHH) which is probably the most characterized morphogen was shown to be crucial for the growth and axial patterning of the hypothalamus (Mathieu et al., 2002; Szabó et al., 2009; Alvarez-Bolado et al., 2012). Moreover, SHH was shown to directly regulate the expression of its cognate receptor Patched 1 (PTCH1) through which it probably signals to promote anterior-dorsal hypothalamic fate and to antagonize Nodal activity in the development of the posteriorventral hypothalamus (Concordet et al., 1996; Koudijs et al., 2008; Szabó et al., 2009). SHH cooperates with Nodal in the maintenance of the anterior-dorsal hypothalamus (Mathieu et al., 2002) and with bone morphogenetic proteins (BMPs) to drive hypothalamic dopaminergic neuronal specification (Ohyama et al., 2005). Genetic disruption of BMP-receptor1a from olig1 cell lineage led to a decreased number of dopaminergic and proopiomelanocortin (POMC) neurons and increased neuropeptide Y (NPY) neurons in the Arc leading to hypophagic phenotype (Peng et al., 2012). Although this genetic manipulation led to increased expression of the orexigenic AgRP, there was a profound impairment in their fiber numbers (Peng et al., 2012). As both Nodal and BMP are members of the transforming growth factorβ (TGF-β) superfamily, this data suggests interaction between SHH and TGF-β signaling pathways during hypothalamic development.

Wnt and its receptor Frizzled are also important regulators of hypothalamic differentiation and Wnt signaling components such as wnt8b, Frizzled8a and Lef1 were shown to regulate the patterning, neurogenesis and differentiation of posterior hypothalamic cells of zebrafish (Kim et al., 2002; Lee et al., 2006; Russek-Blum et al., 2008). In addition, other members of the Wnt cascade were identified in the Arc of the mouse (Benzler et al., 2013). As the expression of SHH, Nodal, BMP and Wnt is maintained in the mature brain it would be interesting to see whether these molecules are released in a synaptic manner, or maintain their activity via a diffusion mechanism in the adult hypothalamus.

### Peptides and Their Receptors

PAC1 (A.K.A ADCYAP1R1) which is the most specific (i.e., high-affinity) receptor for the pleiotropic neuropeptide pituitary adenylate cyclase-activating polypeptide (PACAP) was shown to regulate the development of zebrafish dopaminergic and OXT neurons by controlling the rate of Otp protein synthesis (Blechman et al., 2007). The anorexigenic peptide Leptin is known to regulate metabolic related homeostatic functions in the hypothalamus through its cognate Leptin receptor (LepR; reviewed by Münzberg and Morrison, 2015). Interestingly, leptin was shown to directly regulate developmental neurite formation of Arc neurons (Bouret et al., 2004). It regulates the neural projections of both orexigenic NPY and anorexigenic POMC neurons (Bouret et al., 2012). This seemingly contradictory effect in which leptin regulates the plasticity of two neuronal populations with opposing effects on energy balance, could be partly explained by the developmental switch in leptin regulation of NPY neurons during the postnatal weaning maturational period (Baquero et al., 2014).

The roles of PACAP, Leptin and their receptors in regulating adult brain and hypothalamic functions were thoroughly investigated in the last decades (reviewed by Blechman and Levkowitz, 2013; Matsuda et al., 2013; Münzberg and Morrison, 2015). The above examples show that these peptidergic systems, which were previously considered as regulators of mature functions, also play a role in the developing hypothalamus. Thus, at least some neuropeptides regulate not only the function but also the development of the systems they control.

## Adult Functions of Developmental Factors

Several key developmental factors have been found to be expressed in the mature hypothalamus, however, their postdevelopmental roles in homeostatic regulation remain elusive. This section presents the current data regarding the adult function of the developmental-related TFs described in previous sections of this manuscript.

### Otp−

It has been suggested that a common ancestor of all rayfinned fish experienced a whole genome duplication event early in evolution, about 350 million years ago (Amores et al., 1998; Vandepoele et al., 2004; Dehal and Boore, 2005; Brunet et al., 2006). As a result, the zebrafish genome contains two Otp genes, otpa and otpb, which share high sequence and expression-pattern homology and present partial redundancy in function (Ryu et al., 2007; Fernandes et al., 2013). Otp expression is maintained in the mature hypothalamus of mouse and zebrafish (Amir-Zilberstein et al., 2012; Herget et al., 2014; **Figure 3**). This suggests that alongside its crucial role in embryonic development, Otp is also involved in adult hypothalamic function. While Otp-null mice die shortly after birth and conditional allele for the gene currently does not exist, the duplication and partitioning of the gene in zebrafish allows both otpa−/<sup>−</sup> or otpb−/<sup>−</sup> fish to survive into adulthood (Ryu et al., 2007; Fernandes et al., 2013). This enables the investigation of Otp's role in post-developmental and adult hypothalamic function. Data acquired from adult Otp-null zebrafish, point to the homeostatic activities Otp regulates in adult brains. Adult Otpa-null zebrafish demonstrated impaired anxiety-like behavior compared with their wild-type siblings in response to ''novel-tank'' diving test. Mutant fish spent more time in the top zone of the tank during the first 2 min of the assay suggesting that Otp is involved in the regulation of novelty related stressors (Amir-Zilberstein et al., 2012; Blechman and Levkowitz, 2013). Otp mutant zebrafish also display deficits in the molecular response of the hypothalamopituitary-adrenal (HPA) axis to stressors. These include the regulation of CRH transcription and cortisol response. Thus, Otp directly regulates CRH gene expression in zebrafish and mouse (Amir-Zilberstein et al., 2012). Moreover, Otp indirectly regulates the alternative splicing of the aforementioned PAC1 receptor during stress adaptation phase that follows various homeostatic challenges (Amir-Zilberstein et al., 2012). Thus in addition to its role in hypothalamic development, Otp may act as a cellular sensor, which mediate between a given homeostatic challenge (''input'') and the following hypothalamic hormonal response (''output''). This assumption is reinforced by the finding that Otp-positive neurons in the zebrafish NPO modulate the visual motor response through the regulation of the melanopsin 4a (opn4a) receptor. This apparent Otp-mediated ''non-visual'' deep brain light-sensing system indicates that Otp neurons serve as an extra-ocular photoreception center in dark photokinesis behavior (Fernandes et al., 2012).

Taken together with its known role in regulating several types of neuropeptidergic neurons in the NPO/PVN, these findings suggest that Otp orchestrates the physiological response to environmental challenges. Given the importance of Otp in hypothalamic function, comprehensive research regarding its expression patterns and stress-induced molecular targets in response to physiological and psychological challenges, is further required.

### Sim1−

Sim1 displays haploinsufficiency unveiling its function in metabolic regulation. Sim1+/<sup>−</sup> mice possess a hypocellular PVN and are hyperphagic and obese with increased linear growth, hyperinsulinemia and hyperleptinemia, mainly under high fat diet conditions (Michaud et al., 2001; Holder et al., 2004; Kublaoui et al., 2006). Sim1 heterozygotes display normal energy expenditure, and treatment with the melanocortin receptor

Immunofluorescence staining of Otp (red) and oxytocin (OXT) EGFP reporter (Blechman et al., 2011) (green) in a two year-old zebrafish brain. The image shows tiled maximum intensity projection of a mid-sagital

and OXT in the NPO. CC, crista cerebellaris; CCe, corpus cerebelli; NPO, neurosecretory preoptic area; OB, olfactory bulb; Tel, telencephalon; TeO, tectum opticum. Scale bar, 200 µm.

agonist, MTII, increases energy expenditure in both WT and Sim1 heterozygous mice (Kublaoui et al., 2006). This phenotype is further supported by data from humans with balanced chromosomal translocations or genomic mutations, which interrupts the Sim1 gene. In spite of their normal basal metabolic rate, these subjects usually suffer from early onset obesity, increased food intake, and display evidence of neurobehavioral abnormalities (Holder et al., 2000; Ramachandrappa et al., 2013).

Sim1+/<sup>−</sup> mice display reduced expression of several hypothalamic neuropeptides such as TRH, CRH, AVP, and SST (Kublaoui et al., 2008). In line with their low OXT levels, Sim1 heterozygotes also show higher sensitivity to the orexigenic effect of the OXT receptor antagonist, OVT (Kublaoui et al., 2008). Furthermore, intracerebroventricular administration of OXT to Sim1+/<sup>−</sup> rescues the hyperphagic phenotype and reduces the characteristic weight increase of heterozygotes. Postnatal chemical ablation of Sim1 expressing neurons leads to hyperphagic obesity and reduced expression of OXT and TRH. However, while Sim1 heterozygotes or post-developmental knockouts display normal energy expenditure, ablation of Sim1-positive neurons leads to decreased energy expenditure (Xi et al., 2012). Postnatal PVN-specific ablation of Sim1 combined with chow diet leads to a hyperphagic obesity phenotype. However, when fed with high-fat diet, the trend is reversed and these mice display reduced food intake and weight loss. Since PVN specific ablation of Sim1 neurons also leads to increased Sim1 expression in the Amygdala, the key regulating region of this phenotype remains to be determined (Xi et al., 2013).

In view of the above data, the authors suggested a model arguing that Sim1 heterozygous phenotype of obesity and hyperphagia occur due to the Sim1 regulatory effect on OXT which is also severely depleted in Sim1 heterozygotes and melanocortin recptor-4 (Mc4r), both of which are known to function in appetite regulation (Kublaoui et al., 2006, 2008; Tolson et al., 2010). The accumulating data further support the fact that at least some of the phenotypes correlated with lack of Sim1 are caused by the perturbation of its mature brain functioning rather than from developmental impairments.

### SF-1−

SF-1 positive neurons are directly involved in the regulation of body weight in the mature brain. This function is mediated, among others, by their responsiveness to the hormone leptin. Hence, specific deletion of leptin receptor in SF-1 neurons results in defects in the ability to maintain normal energy homeostasis and these mice display increased body weight (Dhillon et al., 2006). In addition, there are some indications that SF-1 may directly regulate the expression of the brain-derived neurotrophic factor (BDNF; Tran et al., 2003, 2006), a growth factor which is involved in energy balance (Xu et al., 2003). Yet, it is not clear whether SF-1 indeed regulates BDNF in vivo, and if so—what the physiological significance of this regulation is (Dhillon et al., 2006).

Alongside its role in the maintenance of energy balance SF-1 may be involved in the modulation of the HPA axis in response to stress. SF-1 heterozygous mice develop normal VMN and pituitary, but have defects in their adrenal and in their stress response. These mice display abnormal circulating levels of ACTH and corticosterone during the day and following stressful challenges (Bland et al., 2000). In addition, mice with CNS-specific SF-1 knock-out display anxiety-like behaviors in response to various environmental challenges (Zhao et al., 2008). Beyond the developmental defects that disrupt the normal stress response, in vitro experiments indicate the involvement of SF-1 in the direct regulation of the CRH receptor-2 gene, and in SF-1 knock-out mice there is a marked reduction in CRH receptor expression in the VMN (Zhao et al., 2008).

## Concluding Remarks

The hypothalamus regulates brain and body functions by controlling the activity of a variety of neuropeptidecontaining cell types. By doing so, it allows vertebrates to orchestrate multiple homeostatic processes in order to adapt to the ever-changing environment, thereby maintaining the organism's survival and reproduction. Therefore, mechanistic understanding of the patterning and differentiation of the hypothalamus may further shed light on the function of the mature hypothalamus.

The increased use of zebrafish as a model organism allows the combination of powerful genetic tools with highresolution imaging techniques to advance our knowledge regarding the molecular pathways governing hypothalamic development. One obvious advantage in utilizing both mammalian and non-mammalian models is the ability to gain wider knowledge on evolutionarily conserved regulatory pathways. However, a prerequisite for practical use of the accumulating scientific knowledge is to understand the neuroanatomical organization of the hypothalamus in the key models.

Although the knowledge regarding nucleus-specific markers and TFs continue to expand, many hypothalamic nuclei remain ''uncharted''. Thus, the identification of new regulators of hypothalamic development and function is of fundamental importance. In this regards, the introduction of innovative genome editing methodologies, such as the transcription activator-like effector nuclease (TALEN) or the clustered regularly interspaced short palindromic repeats (CRISPR) genome editing methods now allow efficient genetic manipulation and analysis of new candidate genes (Doudna and Charpentier, 2014; Wright et al., 2014).

Recent findings have demonstrated that some genetic pathways that are involved in hypothalamic development, also play a role in mature hypothalamic functions (summarized in **Table 1**). Thus, key developmental factors maintain their expression in the mature hypothalamus. As developmental perturbation of these genes might lead to lethality, the use of conditional and inducible knockout models is necessary. For example, Sim1 post-developmental knockouts demonstrate the valuable scientific knowledge that can be achieved when using such models, unveiling the highly important role of


#### TABLE 1 | Key functions demonstrated for transcription factors expressed in both developing and mature hypothalamus.

Sim1 in feeding and metabolic regulation in the mature hypothalamus.

### Acknowledgments

We thank Dr. Natalia Borodovsky for her comments on the prosomeric scheme (**Figure 2**). The research in the Levkowitz lab is supported by the Simons Foundation Autism Research

### References


Initiative (SFARI #240085); Israel Science Foundation (ISF); Legacy Heritage Fund for Biomedical Science Partnership; Israel Ministry of Agriculture; Sparr Foundation (in the frame of the Weizmann Research Council); The Henry Chanoch Krenter Institute for Biomedical Imaging and Genomics; The Maurice and Vivienne Wohl Biology Endowment. G.L. is an incumbent of the Elias Sourasky Professorial Chair.


dopaminergic and neuroendocrine cell specification in larval zebrafish. PLoS One 8:e75002. doi: 10.1371/journal.pone.0075002


the hedgehog signaling pathway. BMC Dev. Biol. 8:15. doi: 10.1186/1471- 213X-8-15


Federation (INF) Masterclass Series: Molecular Neuroendocrinology: ''From Genome to Physiology'', eds D. Murphy and H. Gainer (Chichester: Wiley & Sons, Ltd).


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Biran, Tahor, Wircer and Levkowitz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Ontogenesis of peptidergic neurons within the genoarchitectonic map of the mouse hypothalamus

### **Carmen Díaz <sup>1</sup>\*, Nicanor Morales-Delgado<sup>2</sup> and Luis Puelles <sup>2</sup>**

<sup>1</sup> Department of Medical Sciences, School of Medicine and Institute for Research in Neurological Disabilities, University of Castilla-La Mancha, Albacete, Spain <sup>2</sup> Department of Human Anatomy and Psychobiology, University of Murcia, School of Medicine and IMIB (Instituto Murciano de Investigación Biosanitaria), Murcia, Spain

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Stuart Tobet, Colorado State University, USA Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **\*Correspondence:**

Carmen Díaz, Department of Medical Sciences, School of Medicine and Institute for Research in Neurological Disabilities, University of Castilla-La Mancha Almansa Street, 14 Albacete 02006, Spain e-mail: Carmen.Diaz@uclm.es

During early development, the hypothalamic primordium undergoes anteroposterior and dorsoventral regionalization into diverse progenitor domains, each characterized by a differential gene expression code. The types of neurons produced selectively in each of these distinct progenitor domains are still poorly understood. Recent analysis of the ontogeny of peptidergic neuronal populations expressing Sst, Ghrh, Crh and Trh mRNAs in the mouse hypothalamus showed that these cell types originate from particular dorsoventral domains, characterized by specific combinations of gene markers. Such analysis implies that the differentiation of diverse peptidergic cell populations depends on the molecular environment where they are born. Moreover, a number of these peptidergic neurons were observed to migrate radially and/or tangentially, invading different adult locations, often intermingled with other cell types. This suggests that a developmental approach is absolutely necessary for the understanding of their adult distribution. In this essay, we examine comparatively the ontogenetic hypothalamic topography of twelve additional peptidergic populations documented in the Allen Developmental Mouse Brain Atlas, and discuss shared vs. variant aspects in their apparent origins, migrations and final distribution, in the context of the respective genoarchitectonic backgrounds. This analysis should aid ulterior attempts to explain causally the development of neuronal diversity in the hypothalamus, and contribute to our understanding of its topographic complexity in the adult.

**Keywords: hypothalamus, neuropeptides, genoarchitecture, progenitor areas, migrations**

#### **INTRODUCTION**

The hypothalamus is a complex forebrain structure that regulates vital processes and various visceral and somatic behavior. Numerous hypothalamic peptides are involved in modulating such functions. Advances on the molecular mechanisms associated to the ontogeny of peptidergic neurons are helping us

**Abbreviations:** 3v, Third ventricle; A/B, Alar-basal boundary; ac, Anterior commissure; ABas, Anterobasal nucleus; ABasM, Median acroterminal ABas; ABasW, Wing (terminal) portion of ABas; Agrp, Agouti-related peptide; AH, hypothalamic nucleus; AHy, Adenohypophysis; Arc, Arcuate nucleus; Arcc, Core portion of the Arc; ArcM, Median acroterminal arcuate nucleus; Arcs, Shell portion of the Arc; ArcW, Wing (terminal) portion of Arc; Avp, Arginin-vasopressin; ATB, Acroterminal border; BFB, Basal-floor boundary; Cartpt, Cocaine- and amphetamine-regulated transcript prepropeptide; ch, Optic chiasm; CPa, Central portion of the peduncular paraventricular area; Crh, Corticotrophin-releasing hormone; DBLH, Dorsobasal lateral hypothalamic area; DM, Dorsomedial hypothalamic nucleus; DM-P, Peduncular DM; DMcP, Core portion of the peduncular DM; DMcT, Core portion of the terminal DM; DMsP, Shell of the peduncular DM; DMsT, Shell portion of the terminal DM; DM-T, Terminal DM; DPa, Dorsal portion of the peduncular paraventricular area; DPM, Dorsal premamillary nucleus; EPD, Dorsal entopeduncular nucleus; EPV, Ventral entopeduncular nucleus; fx, Fornix tract; Gal, Galanin; Ghrh, Growth hormone-releasing hormone; HDB, to understand developmental defects that cause metabolic and neuroendocrine disorders (Michaud, 2001; Caqueret et al., 2005). Murine experimental studies in which some transcription factors were inactivated, showed that gene products such as GSH1, MASH1, SIM1, SIM2, ARNT2, BRN-2, and OTP are crucial for the differentiation of the parvicellular GHRH, SST, TRH, CRH neuroendocrine cell types, as well as the magnocellular OT and VP secreting neurons (Nakai et al., 1995; Schonemann et al., 1995; Li et al., 1996; Michaud et al., 1998; Acampora et al., 1999; Wang and Lufkin, 2000; Hosoya et al., 2001; Goshu et al., 2004; Caqueret et al., 2005; McNay et al., 2006; Szarek et al., 2010).

Hypothalamo-diencephalic boundary; Hcrt, Hypocretin/orexin; IHB, Intrahypothalamic boundary; LH, Lateral hypothalamus; LM, Lateral mamillary nucleus; LPa, Lateral paraventricular nucleus; M, Mamillary region; MCLH, Magnocellular lateral hypothalamic nucleus; ME, Median eminence; mtg, Mamillotegmental tract; NHy, Neurohypophysis; Npy, Neuropeptide Y; opt, Optic tract; os, Optic stalk; Oxt, Oxytocin; Pa, Paraventricular hypothalamic complex; PBas, Posterobasal area and nucleus; Pdyn, Prodynorphin; ped, Peduncle (lateral and medial forebrain bundles); Penk, Preproenkephalin; PHy, Peduncular hypothalamus; PM, Perimamillary area; Pmch, Pro-melanin-concentrating hormone; PMS, Superficial perimamillary nucleus; POA, Preoptic area; Pomc, Pro-opiomelanocortin α; PPa, Peduncular part of Pa; PRM, Periretromamillary area; PRML, Lateral

Furthermore, developmental gene expression patterns allowed a molecular characterization of specific areal subdivisions within the alar and basal regions of the hypothalamus. Such regional analysis produces *genoarchitectonic maps* (Ferran et al., 2009; Puelles and Ferran, 2012), which illuminate the variety of molecular mechanisms controlling the specification of differential neuronal fates at each particular hypothalamic subregion. Each neuroepithelial area of the hypothalamus becomes distinct by its unique gene expression profile. As a progenitor domain, it is capable of controlling differentially over time its proliferative and neurogenetic activity, producing specific neuronal populations, and simultaneously generating signals for axonal and neuronal navigation. Different neuron types may originate sequentially at the same domain, due to temporal changes in the local molecular profile. Some neurons migrate tangentially within the mantle layer, invading other hypothalamic domains before they finish developing their phenotype and acquire functional roles within particular circuits.

Tangential migrations from nearby forebrain areas into the hypothalamus, as well as internal migrations, are not exceptional during hypothalamic development (e.g., Keyser, 1972; Alvarez-Bolado et al., 2000; Skidmore et al., 2008; Zhao et al., 2008). Puelles et al. (2012) postulated several other tangential migrations, including distinct cell streams forming the ventromedial nucleus and the ventral premamillary nucleus. Our recent studies likewise suggested widespread tangential migrations of some peptidergic neurons (Morales-Delgado et al., 2011, 2014). As a result, a frequent feature of conventional hypothalamic nuclei is their content of mixed neuron types, using a variety of neuropeptides and neurotransmitters; this intermixing may be functionally relevant (Rhodes et al., 1981; Sofroniew and Glasmann, 1981; Markakis and Swanson, 1997; Broberger et al., 1998; Elias et al., 1998; Sawchenko, 1998; Broberger, 1999; Simmons and Swanson, 2009; Shimogori et al., 2010; Puelles et al., 2012; Tobet and McClellan, 2013).

Puelles et al. (LP, online reference atlases and ontology of the *Allen Developing Mouse Brain Atlas*, 2009; Puelles et al., 2012, 2013) recently updated the prosomeric model, particularly as

part of PRM; PRMM, Medial part of PRM; PRtLH, Prereticular lateral hypothalamus; PSPa, Peduncular part of SPa; PSPa lim, Liminal subarea of PSPa; PTh, Prethalamus; PThE, Prethalamic eminence; RM, Retromamillary region; RPa, Rostral paraventricular nucleus; Rt, Reticular nucleus of prethalamus; RTu, Retrotuberal area; RTuD, Dorsal retrotuberal domain; RTuD sub, Subliminal subarea of RTuD; RTuI, Intermediate retrotuberal domain; RTuV, Ventral retrotuberal domain; SCH, Suprachiasmatic nucleus; SCHs, Shell of SCH; sm, Stria medullaris; SPa, Subparaventricular area; SPall, Subpallium; Sst, Somatostatin; STh, Subthalamic nucleus; Th, Tyrosine hydroxylase; Thal, Thalamus; THy, Terminal hypothalamus; TPa, Terminal part of Pa; TPaC, Central portion of TPa; TPaD, Dorsal portion of TPa; TPaV, Ventral portion of TPa; Trh, Thyrotrophin-releasing hormone; TSO, Terminal supraoptic nucleus; TSPa, Terminal part of SPa; Tu, Tuberal region; TuD, Dorsal tuberal domain; TuD sub, Subliminal subarea of TuD; TuI, Intermediate tuberal domain; TuSbO, Tuberal suboptic nucleus; TuV, Ventral tuberal domain; VBLH, Ventrobasal lateral hypothalamic area; Vip, Vasoactive intestinal polypeptide; VM, Ventromedial hypothalamic nucleus; VMc, Core portion of the VM; VMs, Shell portion of the VM; VPa, Ventral portion of the peduncular paraventricular area; VPM, Ventral premamillary nucleus; ZI, Prethalamic zona incerta.

regards the hypothalamus. It was defined as a bi-neuromeric rostral forebrain territory, lying in front of the diencephalon proper and ventral to the telencephalon (which can be seen as a part of it; **Figure 1A**). The *Mash1/Dlx/Arx/Isl1*-expressing preoptic area was ascribed to the subpallial telencephalon. The neighboring *dorsal* part of the alar hypothalamus—the paraventricular areasingularly expresses the genes *Otp* and *Sim1*. There is thus a clearcut molecular hypothalamo-telencephalic boundary, which is longitudinal (Shimogori et al., 2010; Puelles et al., 2012). At its *ventral* end, the hypothalamus is represented by the mamillary and retromamillary areas, which include at their median plane the *rostral end* of the forebrain floor plate. The hypothalamus has no rostral neighbor, since it represents the rostralmost part of the neural tube. As a consequence, the right and left lateral walls of the neural tube are here continuous one with another *primarily* (from neural plate stages onwards), creating an unique *median* alar + basal subregion recently named "acroterminal area" (Puelles et al., 2012). Terminal lamina, optic chiasma and tuberomamillary midline specializations (e.g., median eminence and neurohypophysis) develop there. The roof plate is telencephalic (partly choroidal and partly septocommissural), and ends at the anterior commissure.

The **Figures 1A–C** illustrate the hypothalamic progenitor domains defined within the prosomeric model, with correlative major derived nuclei, and basic genoarchitectonic patterns. Additional such data will be provided below, as each domain is considered. There is a rostrocaudal partition of the hypothalamus and attached telencephalon into two transverse (neuromeric) parts (hypothalamic prosomeres 1 and 2, or hp1/hp2; Pombal et al., 2009; Nieuwenhuys, 2009; Puelles et al., 2012). To avoid confusion with older terminologies, the resulting two transverse parts of the hypothalamus were renamed as *terminal* and *peduncular* hypothalamus (THy, PHy), referring to the terminal rostral position of the former and to the association of the latter with the course of peduncular telencephalic fibers (Puelles et al., 2012; for genoarchitectonic analysis of these rostrocaudal hypothalamic divisions, see Ferran et al., under review).

THy and PHy are both divided dorsoventrally into longitudinal alar, basal and floor territories, which display differential molecular profiles and various microzonal subdivisions. These longitudinal zones are continuous *caudally* with diencephalic counterparts, being a result of shared dorsoventral patterning. The alar hypothalamus is divided into the *paraventricular area* (continuous with the prethalamic eminence and the prethalamic reticular nucleus), and the *subparaventricular area* (continuous with the prethalamic zona incerta) (TPa, PPa; TSPa, PSPa; **Figures 1A–C**; Puelles et al., 2012). The narrow TPa contains the eye vesicle, the optic stalk and the chiasma. The SPa contains subpially the optic tract.

The basal hypothalamus largely corresponds to the classical *tuberal* and *mamillary* hypothalamic regions. However, these two regions belong exclusively to the voluminous basal THy. The new terms *retrotuberal* and *retromamillary* regions are needed for the corresponding PHy basal territories; accordingly, the tuberal and retrotuberal domains compose a hypothalamic longitudinal column, and the mamillary and retromamillary domains another (Tu/RTu; M/RM; **Figures 1A–C**; note "retromamillary"

#### **FIGURE 1 | Schematic partial prosomeric model of the forebrain representing the general position, morphologic organization and principal nuclear and genoarchitectonic subdivisions of the**

**hypothalamus**. The rostrocaudal (R-C) and dorsoventral (D-V) spatial directions are indicated. **(A)** Schema of the main hypothalamic progenitor areas distributed across the dorsoventral and rostrocaudal dimensions. The longitudinal alar/basal boundary (A/B), and the intrahypothalamic (IHB) and acroterminal (ATB) boundaries are indicated respectively as thick pink, orange and blue lines. The hypothalamic area is subdivided rostrocaudally into neuromeric peduncular and terminal parts (PHy, THy). Alar territories are shown on the left (yellow) and basal territories on the right (blue). The alar hypothalamus is subdivided into the paraventricular (TPa/PPa) and subparaventricular (TSPa/PSPa) areas (each pair of areas refers to THy and PHy components of a longitudinal zone), plus corresponding acroterminal subregions. The paraventricular area shows a general tripartition into dorsal, central and ventral parts (TPaD, TPaC, TPaV, DPa, CPa, VPa). The subparaventricular area appears subdivided into supraliminal and liminal parts (referring to the alar-basal limit; the liminal domain expresses

Nkx2.2). The basal hypothalamus is also divided dorsoventrally into the large tuberal/retrotuberal (Tu/RTu) area and the primary mamillary/retromamillary (M/RM) area, plus the corresponding acroterminal subregions. The THy/PHy parts of the hypothalamic floor lie underneath (white). Moreover, the Tu/RTu region is subdivided into three dorsoventral parts: TuD/RTuD, TuI/RTuI and TuV/RTuV. The TuD/RTuD contains a dorsal subliminal part (which also expresses Nkx2.2). The primary M/RM area is subdivided into a perimamillary/periretromamillary band (PM/PRM), and the underlying secondary M/RM complex proper. **(B)** Map of the main alar and basal hypothalamic nuclei represented upon the diagram shown in **(C)**. **(C)** Schematic color-coded map of the gene expression patterns distributed across the main dorsoventral hypothalamic subdivisions of E13.5 mouse embryos. Varied combinations of gene markers create a characteristic molecular profile for each territory, as represented in tabular form on the right. Genes expressed in more than one dorsoventral domain are grouped in the white area and genes expressed in an unique domain are shown in the blue area. For abbreviations, see the list.

substitutes for the older "supramamillary" term). Tu/RTu further subdivides dorsoventrally into dorsal, intermediate and ventral longitudinal microzonal domains (TuD/RTuD, TuI/RTuI, TuV/RTuV; **Figures 1A–C**). The TuD/RTuD domain corresponds to the classic, precociously differentiated "hypothalamic cell cord' (Gilbert, 1935; Keyser, 1972). The TuI/RTuI contains the ventromedial, dorsomedial and arcuate nuclei, whereas the narrow TuV/RTuV produces the hypothalamic histaminergic neurons (Puelles et al., 2012). The underlying mamillary/retromamillary basal territory also shows a molecularly distinct dorsal microzonal subdivision, the perimamillary/periretromamillary domain (PM, PRM; **Figures 1A–C**; Bardet et al., 2008; Puelles et al., 2012; Morales-Delgado et al., 2014; Allen Developing Mouse Brain Atlas).

This anteroposterior and dorsoventral map of the hypothalamus provides a scenario where the genoarchitectonically characterized progenitor domains of specific peptidergic neuron groups can be precisely identified (**Figures 1C**, **2**). Moreover, jointly with results from knockout and other transgenic mice, this modern hypothalamic scenario potentiates the analysis of patterning mechanisms implicated in neuronal type specification.

### **MATERIAL**

We examined the development of twelve peptidergic cell populations (*Agrp, Avp, Cartpt, Gal, Hcrt, Npy, Oxt, Penk, Pdyn, Pmch, Pomc, Vip*; **Table 1**), analyzed from *in situ* hybridization images downloaded from the *Allen Developing Mouse Brain Atlas*. <sup>1</sup> These are mostly sagittal sections; while this section plane is appropriate for the analysis of potential dorsoventral and anteroposterior tangential migrations, the visualization of some anatomic landmarks may be compromised. We recurred to careful analysis of all sagittal (eventually also coronal) section planes shown at the *Allen Atlas*, as well as to our extensive experience with multiple planes of sections through the mouse hypothalamus. We correlated the positions of peptidergic cells at the time points available at the *Allen Atlas* (embryonic days E11.5, E13.5, E15.5 and E18.5, and postnatal day P4) with the genoarchitectonically distinct areas (**Figure 1C**) and conventional nuclei (**Figure 1B**), following the model of Puelles et al. (2012). We also recorded aspects of areal heterochrony (**Table 2**), and apparent tangential migrations (**Table 3**; **Figures 3**–**5**). Note that

<sup>1</sup>http://developingmouse.brain-map.org/

#### **Table 1 | Peptide mRNAs mapped in this work**.

Agouti-related peptide (Agrp) Arginin-vasopressin (Avp) Cocaine- and amphetamine-regulated transcript prepropeptide (Cartpt) Galanin (Gal) Hypocretin/orexin (Hcrt) Neuropeptide Y (Npy) Oxytocin (Oxt) Preproenkephalin (Penk) Prodynorphin (Pdyn) Pro-melanin-concentrating hormone (Pmch) Pro-opiomelanocortin α (Pomc) Vasoactive intestinal polypeptide (Vip)

All transcript patterns correspond to the Allen Developing Brain Mouse Atlas database.

any interpretations of "migration" extracted from our descriptive material are necessarily hypothetic, open to experimental testing, though we hold that these conclusions momentarily represent the most parsimonious interpretations of the data. For overview, we added the similar data on *Sst*, *Ghrh*, *Trh* and *Crh* neurons of Morales-Delgado et al. (2011, 2014) in our present **Figure 2**; **Tables 2**, **3**. Our discussion accordingly contemplates sixteen peptidergic cell types, allowing some general conclusions.

### **RESULTS AND DISCUSSION**

**Figure 2** maps the locations where the different peptide markers listed in **Table 1** first appear expressed, adding the results of Morales-Delgado et al. (2011, 2014) and Morales-Delgado (2012). It reveals that given dorsoventrally disposed longitudinal zones dominate as major sources of neuropeptidic populations. These are the Pa area in the alar hypothalamus and the TuD/RTuD and PM/PRM areas in the basal hypothalamus. It is also apparent that multiple peptidergic cell types emerge within these domains, and, remarkably, some cell types appear independently in two or all three of them. A shared characteristic of the molecular profiles of these three areas is the expression of the transcription factor *Otp* (**Figure 1C**; Morales-Delgado et al., 2011, 2014; Morales-Delgado, 2012; Puelles et al., 2012). This was already shown to be required for the differentiation of a number of peptidergic cell types (Acampora et al., 1999; Wang and Lufkin, 2000). A few other progenitor domains, such as SCH within the acroterminal part of alar TSPa, and DM-P, DM-T, and Arc within the basal hypothalamus, give rise to other peptidergic derivatives, in a more restricted mode. On the other hand, there are hypothalamic progenitor domains that so far do not represent sources of peptidergic neurons. These results are also represented in tabular form within **Figure 2** and in **Table 2**.

#### **ALAR HYPOTHALAMIC SOURCES: THE PARAVENTRICULAR AREA**

This area consists of a large, triangular peduncular paraventricular subarea (PPa) defined by diverse deep populations of the *main paraventricular nucleus* (Pa), and a thin terminal paraventricular subarea (TPa), which develops the deep *rostral paraventricular nucleus* (RPa; the conventional "anterior paraventricular area", or aPV), and the subpial *supraoptic nucleus* (PPa; TPa; RPa; TSO; **Figure 1A**). The main Pa is subdivided into dorsal, central, and ventral parts (DPa, CPa, VPa; **Figure 1B**). Puelles et al. (2012) showed that the PPa/TPa essentially contains glutamatergic cells.

The paraventricular area originates peptidergic cell types expressing *Gal*, *Penk*, *Cartpt*, *Avp* and *Oxt* (present data) as well as *Sst*, *Trh*, *Ghrh*, or *Crh* cells (**Figure 2**; Jing et al., 1998; Morales-Delgado et al., 2011, 2014), with timing differences between the peduncular and terminal sectors. Some peptides may coexist in the same neurons (Swanson, 1987). OXT and GAL, e.g., co-localize in neurons of the adult paraventricular and supraoptic nuclei (Landry et al., 1991; see Rossmanith et al., 1996; Foradori et al., 2006; Furutani et al., 2013, and Bartzen-Sprauer et al., 2014 for other examples). As regards its molecular profile, this domain co-expresses *Otp* and interactive *Sim1*/*Arnt2* genes, within a *Pax6*-positive and *Dlx/Arx/Shh/Nkx2.1*-negative background (**Figure 1C**; Puelles et al., 2012). Mice lacking *Otp* or *Sim1*/*Arnt2* functions lost the differentiation of OXT, AVP, SST, CRH, and TRH neurons at the paraventricular and supraoptic nuclei, as well as at the RPa (Michaud et al., 1998; Acampora et al., 1999; Wang and Lufkin, 2000; Michaud, 2001; Caqueret et al., 2005). Unfortunately, *Gal*, *Penk*, *Cartpt* and *Ghrh* cell types were not studied in these reports. *Ghrh* cells possibly were disregarded because in the adult this phenotype largely seems restricted to the arcuate nucleus. However, *Ghrh* cells separately originate at the basal Arc area, and the alar peduncular Pa, whose derivatives secondarily invade massively the basal plate (**Figure 2**; Morales-Delgado et al., 2014). In *Otp* null mice, *Ghrh* cells were apparently absent in the basal areas that receive tangentially migrated alar *Ghrh* elements, while the intrinsic *Ghrh* cells at the tuberal Arc were not affected.

The topographic and rostrocaudal distribution of *Trh-*, *Ghrh-*, *Cartpt-, Sst-*, *Avp-, Penk, Crh*-, *Oxt-* and *Gal*-expressing cells within the Pa highlights dorsoventral microzonal differences, which are most evident at the main peduncular Pa (DPa, CPa, VPa; TPaD, TPaC, TPaV; **Figure 2**). The earliest differentiated paraventricular *Trh, Ghrh, Cartpt, Sst*, *Avp* and *Penk* cells are circumscribed to the VPa/TPaV subdomains, whereas the earliest *Crh*, *Oxt* and *Gal* cells appear restricted to the CPa/TPaC subdomains (**Figures 3A,B,F,G,I,K,L,P,V; Table 2**; see also Morales-Delgado et al., 2011, 2014). Curiously, none of the analyzed peptidergic populations arises primarily in the DPa/TPaD subdomains, though these are secondarily invaded by some of them. A rostrocaudal peculiarity of the distribution of these paraventricular cells is that the peduncular populations (CPa, VPa sources) largely remain close to the ventricle, forming the *Pa* complex (**Figures 3I,J,L,N,O,U–W,Y**), though there are relatively more superficial elements in the LPa nucleus (radially superficial to VPa) and in the *dorsal entopeduncular nucleus* (EPD; interstitial within the cerebral peduncle) (**Figures 3F,G,S**). In contrast, the terminal paraventricular derivatives (TPaC, TPaV sources) largely eschew the deep RPa (aPV), excepting the *Sst* cells, which are abundant there, and mostly aggregate instead within the subpial *terminal supraoptic nucleus*(TSO; **Figures 3G,H,J',K,M,X**). Others incorporate into a conspicuous cell stream that migrates early on into the tuberal area, forming the subpial *tuberal suboptic nucleus* (TuSbO; this new term corrects the misleading one




Light gray areas are devoid of peptidergic cells, whereas dark gray areas produce two or more types simultaneously. Orange background: Alar domains. Reddish background: Basal domains. Bold letters: peptidergic cells studied in Morales-Delgado et al., 2011, 2014; Morales-Delgado, 2012. All other peptidergic cell types were analyzed from Allen Developing Mouse Brain Atlas datasets.

#### **Table 3 | Alar/basal topography of the different peptidergic cell populations, according to their primary progenitor sources and their perinatal distribution, after tangential migration**.


Bold tags: populations that move from alar into basal domains. Asterisks: populations that move within their primary domain.

"tuberal supraoptic nucleus"; Puelles et al., 2012), and the related deeper *tuber cinereum* (TCi) cell population (**Figures 3H,J,L,M,X**;

see TuSbO and TCi in Puelles et al., 2012; their Figures 8.27, 8.32, 8.33).

*Trh* cells first appear in the VPa at E10.5, and later are found also in the TPaV (E12.5), as well as in the CPa and DPa subnuclei. In the adult rat *Trh* cells were recently described likewise in the tuberal lateral hypothalamus (Horjales-Araujo et al., 2014). *Ghrh* cells also have early E10.5 origins in the VPa and appear in the TPaV at E12.5 (**Table 2;** Morales-Delgado et al., 2014). Some of these *Ghrh* and *Trh* elements then move tangentially ventralwards into basal territories, such as the PBas/ABas, the TuSbO nucleus, and the shells of the VM and DM-P nuclei (Morales-Delgado et al., 2014; their Figure 13). *Cartpt* cells were first identified at VPa at E11.5 (where they persist at P4) and appeared at the TPaV at E13.5 (**Figures 3A,B**; **Table 2**). Some peduncular elements disperse into neighboring domains underneath, such as the alar PSPa and the basal PBas (compare **Figures 3A,C**), while terminal elements seem to migrate into the alar TSPa (**Figures 3B**, **5D**). In newborn mice, *Cartpt*-expressing cells were described as associated to *Lhx6* expression in a population at the ventral part of the PSPa (Shimogori et al., 2010); these might correspond to our migrated *Cartpt* elements.

*Sst* cells emerge within VPa at E12.5 and within TPaV at E13.5, where they largely remain visible, though the deep elements of the terminal RPa possibly disperse partly into TPaC

<sup>2</sup>http://mouse.brain-map.org

and TPaD (Morales-Delgado et al., 2011; *Allen Developing Mouse Brain Atlas*). Singularly, *Avp* cells were first observed at the superficial LPa/EPD nuclei (peduncular Pa domain) at E13.5; deeper cells appear at the VPa and TPaV at E15.5 (**Figures 3F–I**; **Table 2**). Later, *Avp* cells move from VPa into CPa, where they subsequently predominate (**Figures 3J**, **5C**; *Allen Developing Mouse Brain Atlas*), and from the TPaV (E15.5) into the shell of the suprachiasmatic nucleus (SCH) and the TuSbO; in this case, *Avp* cells nearly disappear from their ventral paraventricular sources postnatally (**Figures 3H,J,J"**, **5C**; **Table 2**). *Penk* cells likewise appear in VPa at E13.5 and TPaV at E18.5 (**Figures 3P,R**); the VPa ones remain at this location thereafter, whereas the TPaV counterparts seem to invade the SCH shell (**Figures 3S,T**, **5E**).

<sup>3</sup>http://mouse.brain-map.org

migrations are indicated with a question mark. Filled red ovals identify progenitor areas where the relevant peptidergic cells persist and

abbreviations, see the list.

population at its source, as well as at individual recipient areas. For

On the other hand, among the peptidergic populations originated at the central portion of the peduncular Pa, *Crh* cells were first detected there at E13.5 (Morales-Delgado et al., 2014), whereas *Oxt* and *Gal* cells made their earliest appearance at E15.5 (**Figures 3L,V**; **Table 2**). At the usually retarded TPaC area, similar *Oxt* cells emerge at E15.5 and *Gal* cells were seen at E18.5. From E18.5 onwards some peduncular *Oxt* cells appear also in the VPa (and the LPa cell group, which appears radially displaced relative to the VPa), and other cells seem to enter the DPa (compare **Figures 3L,N,O**, **5A**); terminal *Oxt* cells become ventrally displaced into the SCH shell and the TuSbO nucleus (**Figures 3L,M,O**, **5A**), jointly with some terminal *Gal* cells (**Figures 3X**, **5B**). *Oxt* cells also appeared at the TSO at E15.5, while a less dense *Gal* cell population was found there from E18.5 onwards (**Figures 3K,M,X**, **5A,B**; **Table 2**; Jing et al., 1998). The body of terminal paraventricular cells migrating into the TuSbO is observable as a characteristic cellular arch passing dorsoventrally deep to the optic tract, extending from the alar TSO into the basal TuSbO between E11.5 and E13.5. We only illustrate it for *Avp* cells (**Figures 3H**, **5C**). In the adult hypothalamus, *Gal* and *Crh* transcripts in mice (and CRH peptide in rats) are still concentrated mainly at the CPa, with a minority of cells dispersed into VPa and DPa and the cited basal areas; this indicates that migratory dispersion of these elements is limited (*Allen Developing Mouse Brain Atlas*; compare Figure 10 of Swanson, 1987, and Figure 4 of Simmons and Swanson, 2009; note these authors use the old columnar axis ending in the telencephalon; they accordingly describe as "rostral" and "caudal" our prosomeric "dorsal Pa" and "ventral Pa" positions; Puelles et al., 2012).

We mentioned above that the transcriptional functions of *Otp* and *Sim1/Arnt2* are required for the development of at least some of these paraventricular peptidergic neuron types (Michaud et al., 1998; Acampora et al., 1999; Wang and Lufkin, 2000; Michaud, 2001; Goshu et al., 2004). However, the partially overlapping domains of expression of these genes across the peduncular and terminal Pa area do not seem sufficient to explain the above described dorsoventral dissociation of central and ventral paraventricular progenitor subdomains in the paraventricular complex. Some of these differential effects can perhaps be attributed to the *Brain-2* (*Brn2*), and *Sim2* genes, which seem likewise involved in the differential specification of peptidergic cell lineages in this region. See below for data on *Sim2*. The *Brn2* and *Sim1* expression domains selectively overlap within the CPa domain at E12.5 and E13.5, whereas VPa only expresses *Brn2*, and DPa only displays *Sim1* signal (Figure 7 of Michaud et al., 1998; note their columnar rostrocaudal axis again corresponds to our prosomeric dorsoventral axis).

The reported observation that *Cartpt* cells are diminished in number in *Sim1*−/<sup>−</sup> mice at E12.5 (Caqueret et al., 2006) is difficult to understand, since these cells apparently originate at the VPa, where *Sim1* is not expressed. Moreover, in newborn mice, BRN2 protein co-localizes with *Crh*-, *Avp*- and *Oxt*-expressing cells, which occupy CPa, but not significantly with *Trh* cells, which are produced at the selectively *Brn2*-expressing VPa area (Schonemann et al., 1995; Morales-Delgado et al., 2014). These two apparently inconsistent results may be explained by our earlier (Morales-Delgado et al., 2014) and present data indicating that the *Trh* and *Cartpt* cells born at the VPa normally migrate into the retrotuberal basal plate. The reported absence of *Cartpt* cells in the Pa complex of *Sim1*−/<sup>−</sup> mice (Caqueret et al., 2006) may be spurious, if these cells were produced normally (no *Sim1* normally at the VPa), and simply migrated away. The same interpretive error perhaps occurred with *Trh* cells with co-localized BRN2 protein, which probably were not searched for where they lie *after migration*. Moreover, in *Brn2-*null mice, no *Crh*, *Avp* and *Oxt* cells were detected in the Pa and TSO nuclei, while the *Sst* and *Trh* mRNA expression in VPa/TPaV was unaffected (Schonemann et al., 1995).

Conversely, *Trh* and SST cells coincide with the alar expression domain of *Sim2*, which is largely restricted to VPa/TPaV, while there is no topographic correlation between the *Sim2* expressing domain and *Crh-*, *Avp-* and *Oxt-*positive cells in neonatal mice (Goshu et al., 2004). As expected, the number of *Trh* and SST cells was reduced in *Sim2* null mice, whereas *Crh*-, *Avp*- and *Oxt*-expressing cells were not affected (Goshu et al., 2004).

In summary, the development of *Crh*, *Oxt* and *Avp* cells within CPa seems to occur under control of *Otp* and *Sim1/Arnt2*, probably associated to maintained *Brn2* expression. In contrast, *Trh* and *Sst* cells, and probably also *Cartpt* cells, emerge in the VPa/TPaV, controlled by overlapping signals of *Otp*, *Brn2* and *Sim2* (in absence of *Sim1*). Unfortunately, no experimental data are available with respect to the possible defects in the distribution of alar *Ghrh*, *Cartpt, Penk* or *Gal* cells in the *Otp*, *Sim1*, *Sim2* and *Brn2* mutants.

Additional spatially restricted signals possibly are also implicated in the fate determination of these and other peptidergic cell lineages in the paraventricular area. For instance, CPa strongly expresses *Dickkopf 3* (*Dkk3*) from E11.5 onwards (see *Allen Developing Mouse Brain Atlas*), making this transcription factor another candidate bearing on the differentiation of CRH, galanin and/or OXT cells.

#### **ALAR HYPOTHALAMIC SOURCES: THE SUBPARAVENTRICULAR AREA**

The subparaventricular domain (TSPa/PSPa; Puelles et al., 2012) is primarily negative for *Otp*/*Sim1*, and positive for *Dlx*, *Arx, Isl1, Lhx6* and *Vax1* (**Figure 1C**; Shimogori et al., 2010; Puelles et al., 2012); the early expression of *Arx* is subsequently downregulated, whereas *Dlx* family signals remain expressed into perinatal stages, associated to the differentiation of gabaergic neurons in the entire domain (Hallonet et al., 1998; Shimogori et al., 2010; Puelles et al., 2012; *Allen Developing Mouse Brain Atlas*). This domain produces the classic anterior hypothalamus (massively developed at the TSPa, compared to the thin PSPa); the derivatives include the periventricular subparaventricular nucleus, the anterior hypothalamic nucleus, and the acroterminal suprachiasmatic nucleus (**Figure 1B**; Puelles et al., 2012). The latter selectively shows *Vip*-, *Avp-* and *Oxt-*expressing cells (**Figure 2**), detected from E18.5 onwards (SCH; **Figures 3D,E,J,J',O**; **Table 2**; *Allen Developing Mouse Brain Atlas*). Jing et al. (1998) and VanDunk et al. (2011) reported *Avp* mRNA at the mouse SCH already at E16.5 and E17.5, respectively. At P4, *Avp* and *Oxt* cells were more abundant than *Vip* cells, the latter being restricted to a superficial SCH locus (SCHs; **Figure 3T**). In the adult, VIP cells largely aggregate at the gabaergic SCH core (Moore and Speh, 1993; Castel and Morris, 2000), while the VP/OXT cells lie within the glutamatergic SCH shell (Silverman and Pickard, 1983; Sofroniew, 1985; VanDunk et al., 2011; Puelles et al., 2012; *Allen Developing Mouse Brain Atlas*).

Due to its acroterminal topography, the SCH primordium is distinct from its nearby subparaventricular neighbor, the anterior hypothalamic nucleus, in that it selectively expresses *Six3*, *Six6, Ror*α and *Lhx1*, a profile that persists in the adult mouse SCH, with *Ror*α in the shell and *Lhx1* in the core; transient selective expression of *Nkx6.2*, *Fzd5* and *Rx* occurs as well in this area (Conte et al., 2005; Shimogori et al., 2010; VanDunk et al., 2011; Puelles et al., 2012). Differential programming of neurogenesis at this rostral SPa locus is therefore to be expected. *Ror*α signal progressively becomes restricted to the SCH shell postnatally, with a distribution overlapping that of *Avp/Oxt* cells. However, neither *Avp* or *Vip* expression was affected in *Ror*α null mice (VanDunk et al., 2011). With respect to *Six3*, this gene first appears widely expressed in the rostral neural plate, down to the prospective isthmo/mesencephalic border (Oliver et al., 1995; Bovolenta et al., 1998; Kobayashi et al., 2001; Sánchez-Arrones et al., 2009; Dutra de Oliveira Melo, 2011; VanDunk et al., 2011). Afterwards, *Six3* is progressively downregulated at its caudal end; from E11.5 onwards, *Six3* appears restricted to the terminal hypothalamic prosomere (hp2; Puelles et al., 2012) (*Allen Developing Mouse Brain Atlas*; Dutra de Oliveira Melo, 2011; VanDunk et al., 2011). Complete *Six3* inactivation stunts the whole rostral prosencephalon (telencephalon and hypothalamus; Carl et al., 2002; Lagutin et al., 2003; Lavado et al., 2008). A recent study using a *Nestin*-Cre transgenic line to limit floxed *Six3* loss to neural progenitors produced unexplained variable results at E15.5–E19.5. Some specimens showed absence of *Ror*α and *Avp* expression specifically at the SCH, while *Avp* expression continued to be present in the neighboring RPa and the supraoptic nucleus (TPa derivatives) (VanDunk et al., 2011).

The recent developmental study of VanDunk et al. (2011) clearly represented a significant advance in our understanding of this specialized area. We argue nevertheless that their columnar descriptions would benefit from a translation into prosomeric terms, in so far as prosomeric theory allows fine dorsoventral and anteroposterior regionalization and description of the hypothalamus, whereas columnar theory does not; that is, there is no precise columnar answer to the question ¿which part of the hypothalamic primordium is occupied by the suprachiasmatic nucleus? In contrast, its position within the prosomeric genoarchitectonic map and the framework of possible signaling mechanisms is clearcut in **Figure 1B**. In spite of the VanDunk et al. (2011) analysis, the cascade of regulators involved in the local differentiation of *Avp* and *Oxt* cell lineages remains unclear (in comparison with the origin of such cells in an *Otp*-expressing molecular background), and so does the mechanism that segregates *Avp/Oxt* cells into the SCH shell (whose glutamatergic profile surprises within the SPa); *Vip* cells unproblematically settle within the SCH gabaergic core. Though VanDunk et al. (2011) concluded there is a single neuroepithelial source for all SCH neurons, experience accrued so far elsewhere in the brain suggests that the same progenitors normally do not produce both glutamatergic and gabaergic cells, leaving apart exceptional observations of neurons displaying both neurotransmitters, whose progenitor mechanisms remain unknown (e.g., Jarvie and Hentges, 2012). Parallel sources for these two components, at the TPa or ABasM, and the TSPa, respectively, should perhaps be considered.

Puelles et al. (2012) conjectured that the vasopressinergic and oxytocinergic cell types (and other glutamatergic cells) of the SCH shell might originate in the suprajacent acroterminal TPa progenitor area, which also expresses *Six3*, and is characterized by *Otp*/*Sim1/Brn2* expression (plausibly a necessary genetic background for these peptidergic phenotypes; note the underlying ABasM area also expresses *Otp* in conjunction with basal markers). According to this hypothesis, prospective *Ror*α, *Avp* and *Oxt* shell components might migrate tangentially at early stages from the acroterminal TPa into the subjacent SCH primordium within TSPa (**Figures 5A,C**). *Ror*α expression is most dense next to the TPa/TSPa boundary in sagittal sections, and the early *Ror*α cells form a marginal stratum covering early *Lhx1* cells, supporting such a nearby source and migration mechanism (Figures 2B, 4 of VanDunk et al., 2011). The variant hypothesis may be contemplated that prospective SCH *Avp/Oxt* cells are produced within the TPa, and a combination of *Six3* with specific SPa markers (*Dlx genes, Arx, Isl1, Vax1*) and/or selective SCH regional markers (*Six6, Ror*α*, Lhx1, Nkx6.2*, *Fzd5*, *Rx)* defines a SCH domain that selectively *attracts* these *Avp/Oxt* cells from the TPa into the incipient SCH shell subregion (question marks in **Figures 5A,C**).

#### **BASAL HYPOTHALAMIC PEPTIDERGIC CELL SOURCES**

As mentioned above, the basal hypothalamus is primarily divided dorsoventrally into the tuberal/retrotuberal (Tu/RTu) and the primary mamillary/retromamillary (M/RM) regions (**Figure 1**). The Tu/RTu territory is subdivided into dorsal, intermediate and ventral subdomains (TuD/RTuD; TuI/RTuI; TuV/RTuV; Puelles et al., 2012). The TuD/RTuD produces glutamatergic cells for the periventricular anterobasal and posterobasal nuclei (ABas, PBas), plus some intermediate populations such as the nucleus of the TCi, the magnocellular lateral hypothalamic nucleus (MCLH), and several ventralwards migrating populations, including the massive ventromedial nucleus (VM; Puelles et al., 2012); ABas can be further subdivided into acroterminal/median—AbasMand wing—AbasW- portions. Leaving apart the migrated ventromedial nucleus, the TuI/RTuI contains intrinsic gabaergic populations belonging to the terminal and peduncular parts of the dorsomedial nucleus *shell domains* (DMsT, DMsP), as well as to the acroterminal arcuate nucleus *shell* (Arcs); these formations all have glutamatergic *core* cell aggregates whose presumptive migratory origins are under study (Puelles et al., 2012). In addition, there exists the glutamatergic ventral premamillary nucleus, which migrates from the retromamillary area into DMsT (VPM; **Figure 1B**; Puelles et al., 2012). The remaining thin TuV/RTuV domain is related to production of the hypothalamic histaminergic neurons (Puelles et al., 2012). On the other hand, the primary M/RM area subdivides into the *Otp/Sim1*-positive perimamillary/periretromamillary band, and the secondary M/RM complex proper, both of which produce exclusively glutamatergic neurons.

#### **BASAL HYPOTHALAMIC SOURCES: THE DORSAL TUBERAL AND RETROTUBERAL AREAS**

The acroterminal part of the anterobasal area (ABasM; TuD), where *Six3* is selectively expressed, develops an early *Otp*-positive cell population, some of whose elements differentiate into diverse peptidergic cell types (e.g., Morales-Delgado et al., 2011). Jointly with the peduncular RTuD, the TuD subdomain is characterized molecularly by the expression of general basal markers such as *Shh* and *Nkx2.1*, and more restricted expression of other markers such as *Vax1*, *Lmo3*, *Enc1*, *Vat1l*, and *Cnr1* (**Figure 1C**; *Allen Developing Mouse Brain Atlas*). A subliminal dorsal section of this domain also expresses *Nkx2.2* (Puelles et al., 2012; their Figure 8.14B).

The first peptidergic cell types detected at the ABasM are *Sst* and *Pomc* cells, first detected at E10.5 and E11.5, respectively (ABasM; **Figure 4A**; **Table 2**; present data; McNay et al., 2006; Morales-Delgado et al., 2011). *Cartpt-*, *Penk*- and *Npy*expressing cells next appear throughout ABas at E13.5 (ABasM and ABasW; TuD; **Figures 3B,P**, **4D**; **Table 2**), whereas *Pmch* cells first appear at E11.5 at the PBas (RTuD; not shown; see *Allen Developing Mouse Brain Atlas*; Croizier et al., 2011), and *Sst*, *Penk, Npy* and *Hcrt* cells next emerge there between E13.5 and E15.5 (**Figures 4D,H**; **Table 2**). The peptidergic cell types mentioned at the ABas differentiate as subtypes within the preexistent abundant population of *Otp/Nkx2.1*-positive postmitotic neurons, whereas, in contrast, the peduncular PBas area contains few *Otp/Nkx2.1* cells (**Figures 1C**, **2**; Morales-Delgado et al., 2014). At subsequent stages, peptidergic cells like those of the ABas appear to invade the subjacent tuberal (acroterminal) arcuate nucleus, as well as the shell of the migrated VM nucleus, which also originates from the TuD area (**Figures 5D–G**; Morales-Delgado et al., 2011; Puelles et al., 2012). As is further discussed below, we conclude that ABas is the source of these topographically tuberal *Sst*, *Cartpt, Penk*, *Pomc*, and *Npy* cells (which also are accompanied by migrated *Otp*-positive cells). In *Otp*-null mice, cells expressing *Sst* mRNA or producing SST peptide were absent in the Arc and in an "adjacent area", probably the VM shell (Acampora et al., 1999; Wang and Lufkin, 2000). It would be interesting to analyze the other peptidergic cell populations originated at ABas/PBas in these mice, in order to check their dependency on *Otp* expression and function.

The ABas domain expresses additionally *Nr5a1* (*SF1*) at E9.5–E11.5 (Figure 7 of Ikeda et al., 2001), coinciding with the appearance of the earliest *Sst* and *Pomc* cells (**Table 2**; Shimogori et al., 2010; Morales-Delgado et al., 2011). SF1- and POMCimmunoreactive neurons overlap at the acroterminal ABasM at E10.5 (Figure 1C of McNay et al., 2006; these authors identified it as "retrochiasmatic nucleus", which is a classic synonym for ABas). Subsequently, at 13.5, *Nr5a1* (*SF1*) labels also the AbasW, as well as a VM subpopulation that moves massively ventralwards into TuI (see *Allen Developing Mouse Brain Atlas*; Puelles et al., 2012). Separately, *Pomc* labels since E12.5 a parallel dorsoventral migration from the acroterminal ABasM into the acroterminal Arc area (**Figures 4B,C**, **5G**; Shimogori et al., 2010; Puelles et al., 2012; *Allen Developing Mouse Brain Atlas*). These migrations thus start selectively at the dorsal tuberal ABasW and ABasM subareas, respectively. A number of *Cartpt*, *Sst*/*Otp* and *Npy* cells accompany both migrations, and thus reach as a common terminus the VM and Arc shell regions (TuI; **Figures 3C**, **4D',E–G**, **5D,F**; Morales-Delgado et al., 2011). In murine *Nr5a1* knockout mutants, the distribution of cells expressing NPY, GAL and estrogen receptor α was altered, but the differentiation of these phenotypes was unaffected (Dellovade et al., 2000). The VM nucleus itself appeared subtly altered in this mutant, with partial fading of its normally sharp boundary, allowing some intermixing of peripheral GAD67 cells, as well as disgregated GFP-labeled SF1-/- neurons and redistribution of other components (e.g., ISL1, BDNF, NKX2.1, NPY immunoreactive cells; Dellovade et al., 2000; Tran et al., 2003; Davis et al., 2004). On the whole, these studies suggest that the final topography of the diverse peptidergic or non-peptidergic neuronal types within and surrounding the VM is altered as a consequence of the dismorphogenesis caused by *Nr5a1* mutation. It may be speculated that other peptidergic cell types originated at the ABas may be also partially relocated in the *Nr5a1*-null phenotype.

In the adult rat, NPY cells are generally ascribed to the Arc, plus transient *Npy* expression at other sites. Our data in the mouse indicate there is in addition early appearance of *Npy* cells within the ABas and PBas areas at E13.5 (**Figures 4D,D'**). Subsequently some of these *Npy* cells migrate into the intermediate tuberal and retrotuberal areas, targeting the Arc and DM nuclei, and the shell of VM (**Figures 4E–G**, **5F**). These migrated elements may correspond with the reported rat dorsomedial, perifornical and lateral hypothalamic *Npy* cells (Singer et al., 2000; Grove et al., 2001).

*Cartpt* cells first appeared at ABas at E13.5, with an apparent later expansion of the source into the peduncular PBas, and subsequent dispersion into terminal and peduncular Tu/RTu regions such as the VM shell and the DM shell (**Figures 3B,C**, **5D**), consistently with Koylu et al. (1997), Broberger (1999), Vrang et al. (1999), and Elias et al. (2001). *Cartpt* cells were first observed at the Arc at P14. We also observed *Cartpt*-positive cells in the mammillary area from P4 to P56, a localization which was not previously reported (M; **Figures 3C**, **5D**; *Allen Developing Mouse Brain Atlas*). This late *Cartpt* cell population may be interpreted either as due to cell dispersion from TuI, with a more dorsal origin (TuD), or as late timing of the expression of the marker.

*Mash1* (*Ascl1*) codes a transcription factor, whose expression generally overlaps and parallels in regional intensity that of *Dlx* genes (SPa ventricular and mantle zones in the alar hypothalamus, and the ventricular zone of the whole basal Tu/RTu region). In the absence of *Mash1* function, a general reduction of cell numbers was observed at the VM and Arc nuclei, affecting the POMC, NPY, GHRH and dopaminergic lineages (McNay et al., 2006). At E10.5 there is already a dramatic reduction in the number of SF1/MASH1 and POMC cells at the ABas, whereas at later stages migrated SF1 (*Nr5a1*) cells in the VM, as well as migrated NPY cells and intrinsic dopaminergic neurons in the Arc are significantly reduced (McNay et al., 2006). As regards the differentiation of GHRH cells at the Arc, see comments below on requirement of *Msh1* and *Gsh1* functions.

The early restricted expression of *Pomc* at the ABasM (acroterminal TuD) without any additional expression at the ABasW and PBas may be related to the local restricted signal of *Six3* (or other genes mentioned above) within the acroterminal area. *Six3* is reportedly necessary for *Shh* activation in the terminal hypothalamus (Geng et al., 2008). General basal *Shh* and *Nkx2.1* expression is necessary for the development of the tuberal hypothalamic regions, though, curiously, the ABas itself is not affected in mice lacking *Nkx2.1* function, whereas the Arc and VM nuclei are absent or abnormal (Kimura et al., 1996). Moreover, the *Neurog3* (*Ngn3*) gene is expressed selectively at ABas as early as E9.5 (described as "Arc/VM region of the hypothalamus" by Pelling et al., 2011). *Mash1* apparently acts upstream of *Ngn3* to regulate neurogenesis in the basal hypothalamus (McNay et al., 2006). Pelling et al. (2011) showed in *Ngn3*-Cre mice that a set of POMC, NPY, TH and SF1 cells originate from *Ngn3* progenitors, presumably at the ABas area. This result suggests that the arcuate TH-positive population of dopaminergic cells may also migrate tangentially down from the ABas domain (an idea never proposed before, but presently corroborated by our finding of ABas *Th*-expressing cells at E11.5 (*Allen Developing Mouse Brain Atlas*). Moreover, loss of *Ngn3* function leads to a significant reduction of POMC cells, combined with an increase of TH and NPY cells (Pelling et al., 2011), implying that this factor promotes the POMC phenotype but represses the NPY and TH fates.

Basal *Pmch* cells were already present superficially at E11.5, restricted to the subliminal dorsal retrotuberal PBas area (not shown; *Allen Developing Mouse Brain Atlas*); we identified this population according to Puelles et al. (2012) as the conventional *magnocellular lateral hypothalamic nucleus* (MCLH; **Figure 4L**); similar cells were included in the lateral hypothalamus by Croizier et al. (2011). Additional PBas *Pmch* cells later disperse between E15.5 and P4 within the *dorsobasal and ventrobasal sectors of the lateral hypothalamus*, deep to the lateral forebrain bundle (see these sectors in Puelles et al., 2012; their Figure 8.32), deep to the migrated subthalamic nucleus. Some of these elements also invade the DM-P shell and the dorsobasal/ventrobasal perifornical nuclei (MCLH, LH, DBLH, VBLH; **Figures 4N–P**, **5H**; see P56: **Figure 4Q**).

The *Pmch* cells found in the adult rat lateral hypothalamus, including the MCLH, were labelled by acute BrdU at E10-E11 (Brischoux et al., 2001; Croizier et al. (2011). The latter authors mapped the earliest MCH neurons in rat embryos, compared their positions with relevant regional molecular markers (emphasizing their origin within a *Nkx2.2/Nkx2.1/Shh*-positive band), and studied the course of their efferent axons. We are essentially in agreement with their primary data, but qualify the interpretation given by these authors about the molecular environment of the early *Pmch* cells. This is partly because we use the more elaborated model of molecularly distinct progenitor domains reported by Puelles et al. (2012), which was not available to these authors. Croizier et al. (2011) convincingly showed in E14 rat embryos that the early *Pmch* cells occupy a restricted posterior sector of the classical longitudinal hypothalamic cell cord, which is formed by precociously differentiating basal neurons (Gilbert, 1935; Keyser, 1972; His, 1892; see also Puelles et al., 1987, 2014). Consistently with this interpretation, this band expresses *Shh* and *Nkx2.1* (typical overall basal plate markers), as well as *Nkx2.2* (an early longitudinal marker expressed *across* the alar-basal boundary, thus labeling a subliminal part of the hypothalamic cell cord, as well as a liminal part of the overlying subparaventricular area in the alar plate; **Figure 1A**; Puelles and Rubenstein, 2003; Puelles et al., 2004, 2012; Figure 8.14B of the 2012 reference). Within our model, the basal subliminal band falls specifically within the TuD/RTuD progenitor area. The restricted *peduncular* position of the early *Pmch* cells, corroborated by our present mouse results, jointly with the molecular environment provided by Croizier et al. (2011), is consistent with a subliminal PBas (RTuD) origin, a site with selective expression of *Lhx9* (**Figures 1**, **5H**; Shimogori et al., 2010).

However, Croizier et al. (2011) assumed that the early basal cell cord is identical with the "intrahypothalamic diagonal" of Shimogori et al. (2010), thinking it correlates with the longitudinal band that expresses *Nkx2.2*. This interpretation seems wrong in two ways. First, the *Nkx2.2* band has alar and basal parts (as contemplated in Puelles et al., 2004, 2012), and the differentiating MCH cells are restricted to the basal component, as was clearly indicated by the overlapping expression of basal markers such as *Shh* and *Nkx2.1* (Croizier et al., 2011; their Figures 3K–P). Second, these authors apparently misinterpreted the description of the molecular profile of the intrahypothalamic diagonal given by Shimogori et al. (2010), insofar as these authors did not identify their diagonal area as expressing *Shh* and *Nkx2.1*. The intrahypothalamic diagonal is the distinct *alar* longitudinal SPa territory that overlies the basal hypothalamic cell cord (TuD/RTuD); it expresses *Dlx*/*Arx*/*Vax1/Isl1* transcripts and some *Lhx* genes, but lacks significant *Shh* or *Nkx2.1* signal (Shimogori et al., 2010; Puelles et al., 2012; **Figure 1C**). A ventral part of this area—the *liminal* SPa subarea- shows *Nkx2.2* signal on top of the cited molecular profile (Puelles et al., 2012; see their Figure 8.14B). The *Pmch* cells clearly lie under it (Croizier et al., 2011; their Figures 3K–P), within the underlying *Nkx2.1/Nkx2.2*-positive subliminal TuD/RTuD basal domain, which corresponds to the true hypothalamic cell cord.

Hypothalamic *Hcrt/orexin* cells also differentiate selectively within the peduncular dorsal retrotuberal region, though later, at E15.5 (**Figures 2**, **4H**). We think that this occurs specifically within the ventral PBas subarea that does not express *Nkx2.2* (present data). Progressively, most *Hcrt/orexin* neurons disperse radially and tangentially into the (retrotuberal) *ventrobasal sector of the lateral hypothalamus*, where *Nkx2.2* is not expressed (VBLH; **Figures 4H,I**, **5I**; compare Figure 8.29 of Puelles et al., 2012), though some remain more dorsally, within their source area PBas. In the adult rat brain, *Hcrt* cells were found just rostral to the zona incerta, within the "lateral and dorsal hypothalamus" and the "perifornical nucleus" (Peyron et al., 1998; their Figures 3, 8 and 14; compare **Figure 1**). This location lies just dorsally to our PRM area, pinpointing the RTuI, or ventrobasal, part of the LH. The singularity of PBas as a cell source, as opposed to the terminal ABas domain, might be due to its distinctive genoarchitectonic properties, such as the lack of *Nr5a1* expression (Puelles et al., 2012; present data) and the coincidence with selective *Lhx9* expression (Shimogori et al., 2010). The latter authors reported that *Lhx9* is co-expressed with *Hcrt* and partially also with *Gal* cells in newborn mice. In conditional *Shh* mutant mice in which *Shh* was selectively abolished in the basal hypothalamus, the entire Tu/RTu region was very reduced, and *Hcrt* cells were not detected (Szabó et al., 2009); on the whole, this supports our conclusion of a restricted *basal Hcrt* source at the ventral part of PBas. In contrast, Zhao et al. (2008) suggested an alar prethalamic origin of at least some *Hcrt* cells, based on the presence of a few LH hypocretin /orexin-expressing cells that coincide with *Foxb1-*derived progeny in newborn mice. This possibility needs additional analysis, since it postulates both a caudorostral (prethalamo-hypothalamic) and dorsoventral (alarbasal) translocation of such cells, which would need to be distinguished from alternative *Foxb1-*derived progeny potentially produced at the basal mamillary domain, some of which move into periretromamillary positions, just under the VBLH (Zhao et al., 2008).

#### **BASAL HYPOTHALAMIC SOURCES: THE INTERMEDIATE TUBERAL AND RETROTUBERAL AREAS**

This section centers on the arcuate and dorsomedial domains. The tuberal arcuate nucleus, an acroterminal basal locus characterized by overlapping *Nkx2.1*/*Dlx/Six3* markers, combined with early downregulation of its initial basal *Shh* expression (Manning et al., 2006), is an important source of some peptidergic neurons, such as *Ghrh* (Morales-Delgado et al., 2014) and *Agrp* cells (present data). We postulated above that the adult local dopaminergic (*Th*), and *Pomc* cell types originate more dorsally at the *Ngn3* expressing ABas domain (ArcM, ArcW; **Figure 5G**; Pelling et al., 2011). It is less clear whether *Npy* cells are included in the same migratory pathway. Such cells do appear first at the ABas area, and later at the Arc. It was established that most adult POMC cells are not gabaergic cells (consistently with the postulated ABas origin), whereas a majority of NPY neurons are gabaergic, which suggests instead a local Arc origin, though they may be complemented with some migrated cells (Horvath et al., 1997; Ovesjö et al., 2001; Hentges et al., 2004; Puelles et al., 2012). Significantly, about one quarter of the mature NPY neurons in the Arc share a common progenitor with POMC cells (Padilla et al., 2010). Accordingly, two subgroups of Arc NPY cells probably exist: 25% is glutamatergic, originated at the *Ngn3*/*Nr5a1/Six3* positive ABas locus, and a majority (75%) gabaergic population is generated within the Arc, in clear correspondence with local *Dlx* and *Gad67* expression (Yee et al., 2009; Puelles et al., 2012; **Figure 5F**).

Dispersed *Ghrh* cells first appear at E13.5 at the wing (nonacroterminal) part of the arcuate nucleus, extending later into the median acroterminal Arc (ArcM) where they form an important *Ghrh* cell population (ArcW, ArcM; **Table 2**; Morales-Delgado et al., 2014; *Allen Developing Mouse Brain Atlas*). On the other hand, *Agrp* cells first develop at the median part of the arcuate nucleus (ArcM) somewhat later, from E15.5 onwards. *Agrp* cells thereafter remain fully restricted to the Arc during postnatal development (**Figures 4J,K**; Nilsson et al., 2005). Some of the genes expressed at the Arc locus, such as *Six3*, *Nkx2.1*, *Mash1*, *Hmx2*, *Hmx3*, *Gsh1*, *Aldh1a2*, and *Isl1* were implicated in the differentiation of GHRH cells. *Nkx2.1* is expressed throughout the mantle of the hypothalamic basal plate at prenatal stages, and conditional *Nkx2.1* mutants are defective in the Arc derivatives (Mastronardi et al., 2006). The NKX2.1 transcription factor acts upstream of *Mash1*, *Hmx2*/*Hmx3* and *Gsh1* (Caqueret et al., 2005). *Mash1* and *Hmx2*/*Hmx3* control of *Gsh1* expression, which is required for the specification of GHRH neurons (Wang et al., 2004; McNay et al., 2006). Unfortunately, these studies did not explore the role of those genes in the differentiation of *Agrp* cells. Interestingly, *Agrp* cells were unaffected in null mutants of *Ngn3* (Arai et al., 2010), a transcript restricted to ABas, consistently with their apparent Arc origin.

Caudal to the Arc, the intermediate Tu/RTu subregion shows the mutually similar DM-T and DM-P areas, which share *Shh*, *Dlx*, *Nkx2.1, Isl1, Cnr1, Peg10* and *Lef1* expression, in absence of *Lhx9/Lhx6* signals (DM-T, DM-P; **Figures 1B,C**). Both DM portions (as well as the Arc) are subdivided into core (glutamatergic) and shell (gabaergic) portions; the core cells were suspected of migrating into these areas from a neighboring origin, probably the TuD/RTuD, whereas the gabaergic shell elements are intrinsic, consistently with local *Dlx* family genes and *Gad67* (Puelles et al., 2012).

Interestingly, the DM-P domain is a major source of *Gal* cells; the latter were first identified there at E15.5 (**Figure 3V**; *Allen Atlas* data). Similar early *Gal* cells were reported to appear in a vaguely defined "precursor hypothalamic region" at E16 in the rat, which we estimate to be peduncular and retrotuberal (RTuI) in topography (Figure 1A of Gundlach et al., 2001). Starting at E18.5, a less important *Gal* population appears likewise at the DM-T area, jointly with a possibly migrated perimamillary subpopulation (not shown; see *Allen Atlas* data). Postnatally these cells concentrate at the DM *shell* domains (Puelles et al., 2012), a pattern that suggests a local origin within TuI/RTuI and a gabaergic nature (DMsP, DMsT, **Figures 3V,W,Y,Z**, **5B**; **Table 2**; Puelles et al., 2012). Additional *Gal* cells appear at the Arc and the ventrobasal lateral hypothalamus postnatally (VBLH; **Figures 3Z**, **5B**; compare Melander et al., 1986). This suggests partial migratory dispersion of DM-P/DM-T *Gal* cells into the PM, Arc and VBLH areas (**Figure 5B**).

#### **BASAL HYPOTHALAMIC SOURCES: THE PERIMAMILLARY AND PERIRETROMAMILLARY AREAS**

Another basal progenitor domain producing peptidergic cell types is the arc-shaped perimamillary/periretromamillary area (PM/PRM), which lies ventral to the linear *Nkx2.1*/*Dlx*/*Arx*/*Lhx6/Wnt8b*-expressing TuV/RTuV domain (tuberomamillary terminal of Shimogori et al., 2010), where histaminergic neurons selectively originate (Puelles et al., 2012; **Figure 1C**). The PRM/PM area represents a domain with precocious neurogenesis, comparable in this regard to the hypothalamic cell cord (TuD/RTuD; see Puelles et al., 2014). It is molecularly distinct from the retarded retromamillary and mamillary areas by its selective expression of *Otp* and *Sim1*, in curious parallelism with the alar paraventricular area (**Figure 1C**; Puelles et al., 2012). *Sst, Penk, Npy* and *Pmch* phenotypes seem to originate from both the peduncular (PRM) and terminal (PM) subdivisions of this domain, in each case with a particular temporal sequence (**Figure 2**; **Table 2**; Morales-Delgado et al., 2011). It is unclear whether the PM/PRM *Gal* cells interpreted above as migrated from the DM-P area might originate instead locally (this point can be decided by examining their gabaergic vs. glutamatergic nature, since PM/PRM produces only the latter type; Puelles et al., 2012). *Sst* and *Npy* cells appeared selectively at the PRM at E13.5 and were detected only at E18.5 and P4, respectively, within the PM; this delay raises the possibility of their tangential translocation from PRM into PM (**Figures 4D–G**, **5F**; **Table 2**; Morales-Delgado et al., 2011; Morales-Delgado, 2012). *Penk* cells were not found at E15.5, but were already well represented at both PRM and PM at E18.5 (**Figure 3R**), which suggests an earlier appearance. At the later stage, as well as postnatally, the *medial PRM nucleus* (PRMM; **Figure 3R**) and the *dorsal premamillary nucleus* (DPM; **Figure 3R**; this is the main PM derivative) show massive *Penk* cell populations, and additional *Penk* cells were found at a previously undescribed rounded perimamillary cell aggregate superficial to the DPM, named here *superficial perimamillary nucleus* (PMS; **Figure 3U**; see also the Allen Developing Mouse Brain Atlas, transversal P56 series [64881286], section 61, for DPM and PMS). The neighboring *ventral premamillary nucleus* (VPM) also contains *Penk-*positive cells; however, it lies in the TuI area, intercalated between the ventromedial nucleus and the PM band (not shown; see the P56 Allen image cited above). Though these tuberal cells might have a different, more dorsal basal origin (see above results on ABas/PBas *Penk* cells), we mention them here because of the possibility that these VPM cells may have migrated either from the underlying PM area or from the more distant PRM area, accompanying the migration of RM cells into VPM (**Figure 5E**; Puelles et al., 2012).

Medial hypothalamic *Pmch* cells were also clearly identified along the characteristic PRM/PM band from E13.5 onwards (**Figures 1**, **2**, **4M–P**; **Table 2**). A neighboring *Pmch* cell group forms a shell around the migrated ventral premamillary nucleus (TuI area) from E15.5 onwards (VPM; **Figures 4N–P**). Remarkably, at postnatal stages few *Pmch* cells remain visible at the PRM/PM locus, suggesting an earlier translocation into the suprajacent TuI/RTuI areas, mainly the DM-P and DM-T nuclei, as well as the mentioned VPM shell neighborhood, where they increasingly become visible (**Figures 4N–P**, **5H**). The PRM/PM *Pmch* cells seem unrelated to the more dorsal and precocious ones described above, which contribute to the MCLH and LH populations (see **Figures 4Q**, **5H**). While Croizier et al. (2011) did not identify this separate set of PM/PRM MCH cells, (Brischoux et al. (2001); their Figure 5E; compare with our **Figure 1**) showed a drawing of a sagittal section, in which rat *Pmch* cells at E18.5 clearly occupy a band that does not agree in shape and position with any part of the basal hypothalamic cell cord (TuD/RTuD), but agrees with our PRM/PM area.

*Otp* and *Sim1*/*2* are potential key genes in the differentiation of the mentioned PRM/PM peptidergic lineages, since they are selectively co-expressed very early at this locus (supplementary figures of Shimogori et al., 2010; Morales-Delgado et al., 2011, 2014; Puelles et al., 2012; *Allen Developing Mouse Brain Atlas*); however, this hypothesis was not tested so far. Other genes expressed at the PRM/PM domain, such as *Shh, Nkx2.1, Pou3f3 (Brn1), Ebf3 and* *Peg10* (**Figure 1C**), might be also involved in the neurogenesis and/or differentiation of the peptidergic cell types originated in this territory.

#### **PRODYNORPHIN PROGENITOR AREAS**

We also studied *Pdyn* cells, which may be a singular case of delayed appearance of the peptidic marker expression in neurons that previously migrated tangentially. *Pdyn* cells first appear at E13.5 within the alar VPa area (peduncular hypothalamus); later they invade the overlying CPa and DPa areas at E15.5 and E18.5, respectively (**Figures 4R–U**, **5J**; **Table 2**). Basal *Pdyn*expressing cells were first observed at the DM-P and migrated (terminal) VM from E13.5 onwards, and at the DM-T at P4 (**Figures 4R',S–U**, **5J**). At perinatal stages, the basal elements became progressively concentrated within the core domains of these nuclei, known to be glutamatergic (e.g., VMc; DMcT in **Figures 4T,U**). Puelles et al., 2012 argued that there might be a relationship of *Pdyn* cells with *Nkx2.2*-expressing immature migrating elements, which all arise from the longitudinal band mentioned above, which overlaps the alar-basal boundary; this band divides into a *liminal* alar sub-band and a *subliminal* basal sub-band, whose derivatives co-express differential alar vs. basal gene markers (**Figures 1A,C**, **2**). Puelles et al. (2012) showed that, from E10.5 onwards, peduncular liminal *Nkx2.2*-positive neurons selectively migrate dorsalwards into the VPa nucleus, from where some of them later move into the LPa, CPa and DPa nuclei. This sequential pattern recalls our present data on early peduncular *Pdyn* cells, though these are first detected at the VPa at E13.5, and not at the *Nkx2.2* band (**Table 2**). If they are identical with the earlier migrating *Nkx2.2*-positive cells, then they apparently start to express *Pdyn* in a delayed manner, after their initial migration into VPa ends. In that case, their proper origin would not be the VPa area, but the underlying liminal PSPa subarea (**Figure 5J**). Co-expression and lineage studies are needed to resolve this issue; *Nkx2.2* continues being expressed in the paraventricular complex even in the adult mouse (Puelles, unpublished observations). It is also relevant to check the potential gabaergic vs. glutamatergic phenotype of these *Pdyn* cells, since local VPa cells are expected to be glutamatergic, whereas PSPa cells are held to be gabaergic (Puelles et al., 2012, note their Figure 8.18 shows gabaergic cells at the VPa and DPa nuclei).

On the other hand, terminal subliminal (basal) *Nkx2.2* positive cells (from the upper TuD area) migrate between E13.5 and E15.5 ventralwards into the VM nucleus, constituting one of its diverse migrated subpopulations (Puelles et al., 2012; their Figure 8.26 and text; **Figure 5J**). In this case, we see *Pdyn* cells emerging at the VM locus at E13.5 (**Figure 4R'**), possibly representing migrated subliminal *Nkx2.2* cells that again express the peptidic marker in a delayed manner, after migration. This interpretation is supported by the fact that all other known populations of the VM migrate from the overlying TuD area. Accordingly, the latter may be suspected to be the true origin of these VM *Pdyn* cells.

The *Pdyn* cells emerging at the DM-P may correlate likewise with peduncular subliminal *Nkx2.2*-positive cells (from the upper RTuD area), since some of them also migrate ventralwards a short distance into the DM-P area at E14.5 (Puelles et al., 2012; their Figure 8.26F). This is where we saw *Pdyn* cells at E15.5, possibly again due to delayed differentiation (**Figure 4S**; **Table 2**). Unfortunately, the Allen Atlas lacks clearcut data on *Nkx2.2* expression at this locus at later embryonic stages (few labelled cells appeared at the DM-P locus at E18.5), or postnatally (no significant signal). Contrarily, *Pdyn* cells become increasingly visible postnatally at the TuI/RTuI area, largely aggregated within the developing core portions of the DM-P and DM-T nuclei, as shown here at E18.5 and P4 (**Figures 4T,U**, **5J**; note that due to its obliquity at 45 degrees relative to the ventricle, the DMcP appears in more lateral sections than the DMcT, which is parallel to the ventricle**)**. We likewise lack evidence that any *Nkx2.2-*expressing cells invade the TuI area where the DMcT *Pdyn* cells emerge postnatally. In defense of the migratory hypothesis, it may be conjectured that migration of such cells into the DM areas might imply downregulation of *Nkx2.2* before the *Pdyn* signal appears. Lineage studies are thus needed to test our conjecture that *Pdyn* cells derive systematically from either alar or basal *Nkx2.2*-espressing progenitors. Note the potential origin of the DM *Pdyn* cells at the overlying TuD/RTuD area would explain their glutamatergic phenotype within the gabaergic DM shell environment (Puelles et al., 2012).

#### **CONCLUSIVE COMMENTS**

**Table 2** summarizes our findings about the earliest topography of the different hypothalamic peptides, separated into terminal and peduncular progenitor areas. We subsumed under the terminal progenitor domains the rostral specialized acroterminal areas (**Figure 1B**; Puelles et al., 2012), which nevertheless show some independent behavior (e.g., SCH, Arc, as indicated in parentheses in **Table 2**).


early *Cartpt* cells appear aligned with the prospective anterior hypothalamic nucleus, whereas the later *Vip, Avp* and *Oxt* cells emerge at the neighboring acroterminal suprachiasmatic nucleus; moreover, the *Avp/Oxt* types have a question mark in **Table 2**, because we suspect these elements may come from the overlying TPa area (see text above). A sequential pattern nevertheless is truly present in other areas; this implies a general fate regulatory mechanism that is heterochronic and position-sensitive (i.e., at a given locus, different cell types are generated over time); moreover, the same cell type may differentiate sooner or later in different areas.


### **ACKNOWLEDGMENTS**

This work was funded by the Local Government of Castilla-La Mancha grant PII1I09-0065-8194 to CD, the Spanish Ministry of Economy and Competitiveness grant BFU2008-04156 and the SENECA Foundation 04548/GERM/06 (no. 10891) to LP. Infrastructure support provided by the University of Castilla-La Mancha and the University of Murcia is also acknowledged.

### **REFERENCES**

Acampora, D., Postiglione, M. P., Avantaggiato, V., Di Bonito, M., Vaccarino, F. M., Michaud, J., et al. (1999). Progressive impairment of developing neuroendocrine cell lineages in the hypothalamus of mice lacking the Orthopedia gene. *Genes Dev.* 13, 2787–2800. doi: 10.1101/gad.13.21.2787


and the lateral hypothalamic area. *J. Comp. Neurol.* 402, 442–459. doi: 10. 1002/(sici)1096-9861(19981228)402:4<442::aid-cne2>3.3.co;2-i


postnatal development in the mouse. *Brain Res. Dev. Brain Res.* 155, 147–154. doi: 10.1016/j.devbrainres.2005.01.009


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 30 October 2014; accepted: 12 December 2014; published online: 12 January 2015*.

*Citation: Díaz C, Morales-Delgado N and Puelles L (2015) Ontogenesis of peptidergic neurons within the genoarchitectonic map of the mouse hypothalamus. Front. Neuroanat. 8:162. doi: 10.3389/fnana.2014.00162*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Díaz, Morales-Delgado and Puelles. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Ontogenesis of oxytocin pathways in the mammalian brain: late maturation and psychosocial disorders

#### *Valery Grinevich1 \*†, Michel G. Desarménien2†, Bice Chini 3†, Maithé Tauber 4,5† and Françoise Muscatelli 6,7\*†*


#### *Edited by:*

*Gonzalo Alvarez-Bolado, University of Heidelberg, Germany*

#### *Reviewed by:*

*René Hurlemann, University of Bonn, Germany Harold Gainer, National Institutes of Health, USA*

#### *\*Correspondence:*

*Valery Grinevich, Schaller Research Group on Neuropeptides (V078), German Cancer Research Center, CellNetwork Cluster of Excellence, University of Heidelberg, Im Neuenheimer Feld 581 (TP4), Office 3.301, D-69120 Heidelberg, Germany e-mail: v.grinevich@ dkfz-heidelberg.de; Valery.Grinevich@ mpimf-heidelberg.mpg.de; Françoise Muscatelli, Institut de Neurobiologie de la Méditerranée, INMED UMR U901, INSERM, Parc Scientifique de Luminy, Aix-Marseille Université, INMED UMR 901, 13273 Marseille, France e-mail: francoise.muscatelli@ inserm.fr*

Oxytocin (OT), the main neuropeptide of sociality, is expressed in neurons exclusively localized in the hypothalamus. During the last decade, a plethora of neuroendocrine, metabolic, autonomic and behavioral effects of OT has been reported. In the urgency to find treatments to syndromes as invalidating as autism, many clinical trials have been launched in which OT is administered to patients, including adolescents and children. However, the impact of OT on the developing brain and in particular on the embryonic and early postnatal maturation of OT neurons, has been only poorly investigated. In the present review we summarize available (although limited) literature on general features of ontogenetic transformation of the OT system, including determination, migration and differentiation of OT neurons. Next, we discuss trajectories of OT receptors (OTR) in the perinatal period. Furthermore, we provide evidence that early alterations, from birth, in the central OT system lead to severe neurodevelopmental diseases such as feeding deficit in infancy and severe defects in social behavior in adulthood, as described in Prader-Willi syndrome (PWS). Our review intends to propose a hypothesis about developmental dynamics of central OT pathways, which are essential for survival right after birth and for the acquisition of social skills later on. A better understanding of the embryonic and early postnatal maturation of the OT system may lead to better OT-based treatments in PWS or autism.

**Keywords: oxytocin, oxytocin receptor, ontogenesis, somatodendritic release, axonal release, autisn, Prader-Willi syndrome**

*†These authors have contributed equally to this work.*

#### **INTRODUCTION**

Oxytocin (OT) is a neuropeptide which is synthesized in defined nuclei of the hypothalamus, paraventricular (PVN), supraoptic (SON), and accessory (AN) nuclei and secreted both in the blood circulation as a hormone and in the brain as a neuromodulator. During the last decade, hundreds of original reports and reviews have been published, clearly showing that the OT system is a key regulator of all the aspects of social behavior, including those involved in reproduction and care of the offspring (Lee et al., 2009). Notably, in humans, OT facilitates the processing of social information and improves cognitive emphatic abilities, representing a potential new approach for the treatment of autism spectrum disorders, as extensively reviewed elsewhere (Meyer-Lindenberg et al., 2011). Even if OT is currently regarded as a "prosocial" drug, this does not implicate *per se* a "positive" connotation. OT has been found to increase trust and generosity (Kosfeld et al., 2005), but also gloating and envy (Shamay-Tsoory et al., 2009), which are also manifestations of complex "social" behaviors. Similarly, OT increases in-group but not out-group cooperation, an ethnocentric behavior that can promote prejudice, xenophobia, and intergroup violence (De Dreu et al., 2011). Furthermore, OT has been found to induce adverse behavior in borderline-personality disorders (Bartz et al., 2011). These results suggest that the selection of patients will play a crucial role in determining the outcome of the symptomatic treatment with this neuropeptide, the use of which should be restricted to those individuals who could benefit from it. An understanding of the OT-mediated cellular and molecular mechanisms on neuronal networks underlying social and feeding behavior is needed for a successful employment of this peptide in neuropsychiatric and genetic disorders. However, despite the large body of knowledge about physiology and central pathways of OT (see reviews from Landgraf and Neumann, 2004; Ross and Young, 2009; Lee et al., 2009; Knobloch and Grinevich, 2014), the developmental aspect of the OT system remains the *"locus minoris"* of neuroscience and neuroendocrinology. Here, we collected virtually all of the available information, including our personal ongoing research, and provide some general conclusions regarding the dynamics of central OT pathways in prenatal and early postnatal ontogenesis. Furthermore, we clearly demonstrate, by investigating a mouse model for the Prader-Willi syndrome (PWS), that changes in the developmental maturation of the OT system cause metabolic and social alterations. These can be significantly improved or even fully compensated by OT treatment in the first days after birth, opening the door for a powerful pharmacological therapy for PWS in early infancy.

### **THE BIRTH OF OT-ERGIC NUCLEI**

In many species (rodent, human, zebrafish, chicken), all oxytocin (OT) neurons are generated from the proliferative ("convoluted") neuroepithelium of the diamond shaped third ventricle (**Table 1** and **Figure 1**). Birth-dating studies revealed that these hypothalamic neurons are generated in the second half of the gestational period in rodents, within the first quarter of the gestational period (E30-43; length of pregnancy ∼165 days) in macaques (Markakis, 2002) and at the middle of pregnancy in humans (Swaab, 1995).

In rodents, the SON and PVN appear very early. At embryonic day (E) 12.5 dpc (days post-coîtum), two groups of cells are identified in the mouse: one near the third ventricle and the other moved lateral to the surface pial to give rise to the SON (Dongen and Nieuwenhuys, 1989). At E14.5 dpc, the PVN and the SON are settled (Nakai et al., 1995), while AN are recognized later (probably due to their small size and relatively small number of cells; Altman and Bayer, 1978a,b,c). At this stage, an antibody recognizing the Neurophysin-I (the carrier protein for OT) reveals a positive immunosignal (**Figure 1B**), consistent with the expression of the OT-prohormone, not yet detected at E12.5 dpc. Furthermore, an antibody recognizing the intermediate forms of OT (VA10) reveals a positive immunosignal of the OT-intermediate forms at E16.5 dpc (**Figure 1C**), but not at E14.5 dpc. Importantly, the mature, amidated form of OT is detected at birth (see **Table 1**) and co-exists with immature forms of OT during the entire postnatal life (**Figure 1D**). The role of an earlier production of the non-mature forms of OT has not been studied, although a functional role of these forms during the embryonic development has been suggested (Tribollet et al., 1988).

In humans, the SON and PVN are completely formed at 25 weeks of gestation (Dörner and Staudt, 1972) and OTimmunoreactivity is first detected at the age of 26 weeks (Wierda et al., 1991). At that age, the number of stained OT neurons is relatively similar in the fetal and adult hypothalamus (Van der Woude et al., 1995), while the morphological analysis of individual magnocellular neurons suggests that these cells are still immature, as can be seen by the gradual increase of their nuclear volume (Rinne et al., 1962; was also observed in the rat by Crespo et al., 1988).

#### **GENETIC HIERARCHY OF OT NEURON FORMATION**

The signaling molecules and transcription factors that are involved in the determination and differentiation of OT neurons are not well known, only a few mouse studies reported factors


involved in the early stages of development of the hypothalamicneurohypophyseal system (Carrel and Allen, 2000; Caqueret et al., 2006; Szarek et al., 2010).

The early patterning of hypothalamus has been shown to depend on a cascade of transcription factors (**Figure 1A**): firstly

the sonic hedgehog (SHH) is necessary to establish the hypothalamus primordium (E9.5 dpc in mouse) (Mathieu et al., 2002; see also Blaess et al. (2014) in this Research Topic of Frontiers in Neuroanatomy); the local production of the bone morphogenetic proteins (BMPs) is then required to down-regulate SHH and establish region specific transcription profiles (Ohyama et al., 2008). The bHLH-PAS (basic helix-loop-helix PER-ART-SIM) transcription factor SIM1 is expressed in the incipient PVN, SON and AN from E10.5 dpc (Caqueret et al., 2006) where it dimerises with ARNT2 (Michaud et al., 2000; Hosoya et al., 2001). A key downstream target of SIM1/ARNT2 is Brn2, a POU domain transcription factor required for OT (as well as for two other types of neuroendocrine neurons, expressing arginine-vasopressin (AVP) and corticotropin releasing hormone neurons of the PVN/SON/AN and expressed from E11.5 dpc, Nakai et al., 1995; Schoneman et al., 1995). In a parallel or convergent SIM1/ARNT2 pathway, the homeobox orthopedia (OTP) factor is also necessary for Brn2 expression (Caqueret et al., 2006), which is still expressed at E15.5 dpc, with Nkx2.2. All of these factors are required to define at E12.5 dpc the prospective PVN domain, however, from this stage on, the factors that will specify the parvocellular1 and magnocellular OT neurons have not been identified. The ablation of Brn2 results in a loss of all neurons of the PVN, SON and presumably of the AN (Nakai et al., 1995; Schoneman et al., 1995). Importantly the lack of axonal projections of magnocellular OT and AVP neurons in the Brn2 knock-out mice to the pituitary (Schoneman et al., 1995), as also observed in the Arnt2-knockout mice (Hosoya et al., 2001), leads to progressive loss of pituicytes (pituitary astrocyte-like glial cells). All together, these results suggest a role of OT and/or AVP in the formation of the neurohypophysis. Indeed, in the absence of OT, the neurovascular interface in the neurohypophysis does not form in zebra fish (Gutnick et al., 2011).

One approach to identify such factors involved in the determination and differentiation of OT cells is the analysis of the OT promoter and its upstream and downstream sequences. Interestingly, genomic DNA constructs, including more than 500 base pairs (bp) upstream and 3.6 kb downstream of the OT gene, enabled an expression of OT or EGFP-reporter in OT-neurons only (Young et al., 1990; Belenky et al., 1992) in adult transgenic rodents. Later, a genomic analysis of the OT-promoter (500 bp) defined a minimal promoter sufficient for expression in OT-neurons (magnocellular neurons) of adult rodents (Fields et al., 2012; Gainer, 2012). Several binding sites for transcription factors (ERE, COUP-TF, SF1) have been identified in this sequence, but not the ones for the transcription factors described above (**Figure 1A**), suggesting that the DNA sequence required for the development of OT cells is not present in these 500 bp.

In conclusion, from the E12.5 dpc stage, the transcriptional factors that will specify the parvocellular and magnocellular OT neurons have not yet been characterized.

Several other genes such as *fibroblast growth factor 8* (Brooks et al., 2010), the *Mage-D1*, *Necdin* and *Magel2* genes, *CD38*, *Peg3* (**Figure 1** and **Table 3**), are known to be expressed in mouse hypothalamus during development. Their knock-out (KO) alters the number and/or function of OT-neurons, but their role in OT-neurons is not clarified yet.

#### **THE BIOSYNTHESIS OF OT IN ONTOGENESIS**

In the rat and mouse, the OT can be detected by several techniques (such as radioimmunoassay, enzyme immunoassay, immunohistochemistry) from the beginning of the second gestational week (**Table 1**). However, all of these techniques are based on the use of OT antibodies. The specificity of these antibodies, in particular against the different forms of OT, is not always established. The most reliable and relevant studies have been performed with the use of antibodies characterized by Harold Gainer's laboratory or using Mass Spectrometry Ànalysis, which, unfortunately, has not been used in developmental studies. Based on the tools used for OT detection, the OT prohormone is found in embryos just after the appearance of the OT mRNA, while the mature OT peptide is only released from birth on (**Table 1** and **Figure 1**). The OT gene encodes for the Pre-Pro-OT-Neurophysin I (pre-pro-hormone), which is cleaved by different enzymes to give rise to different OT intermediate forms and to the Neurophysin I, and finally to the mature amidated form that is released (**Figure 2**). It has been shown that the steady state of the mature OT form can be controlled by an oxytocinase (P-LAP) that is produced in periphery and centrally by the OTmagnocellular neurons. Noticeably, P-LAP is also expressed in parvocellular OT neurons and in other brain structures (Tobin et al., 2014).

Before birth, there is a delay in OT prohormone processing with an accumulation of the OT intermediate forms in OT neurons until birth. This delay was not observed for AVP processing as the amidated AVP form is detected as early as E16.5 dpc (Whitnall et al., 1985; Altstein and Gainer, 1988). Furthermore, morphological and electrophysiological properties of the magnocellular OT neurons are not mature at birth; their morphological and electrophysiological properties develop progressively during the first two postnatal weeks (see below, as well as Swaab, 1995, 1997a,b). Thus, although functional at birth, the OT-ergic system undergoes a progressive maturation during early postnatal life and is not mature before weaning in rodents and, most likely, in human infants as well. However, a comparison of immature and mature forms of OT in human development has never been studied.

#### **ELECTRICAL PROPERTIES OF OT NEURONS AFTER BIRTH**

To our knowledge, no electrophysiological studies have been performed on embryonic OT neurons. However, the groups of Francoise Moos and Michel G. Desarmenien systematically analyzed electrical activity of magnocellular neurons (although without distinction between OT and AVP) in the SON of rats during early postnatal life. Three main features arise from these studies: the importance of the second postnatal week (PW2), the role of autocontrol by OT, presumably released from soma and dendrites, and the determinant role of the development of electrophysiological features on the morphological maturation of the neurons. At birth and during PW1, the membrane potential is unstable and cells fire small and large erratic action potentials. During PW2, the membrane potential stabilizes to a more

<sup>1</sup>Parvocellular OT neurons belong to a relatively small population of PVN cells, which project to the brainstem and spinal cord (Sawchenko and Swanson, 1982) and control appetite, nociception and autonomic functions. The development of these particular neurons has not been studied.

hyperpolarized value, the spontaneous activity becomes organized and the action potential progressively increases in size and decreases in duration, leading to a decrease in action potentialevoked calcium entry (Widmer et al., 1997; Chevaleyre et al., 2000). At the end of PW2 and during PW3, a switch in regulation of intracellular free calcium from extrusion to sequestration into the reticulum also occurs (Lee et al., 2007a). These two modifications of calcium entry and regulation may have important consequences on peptide secretion. The PW2 is also the period during which the chloride equilibrium potential becomes hyperpolarized, GABA becomes inhibitory and glutamatergic activity appears (Chevaleyre et al., 2001), together with an increase in NMDA receptor expression (Hussy et al., 1997). Most interestingly, this period of action potential and synaptic activity maturation is concomitant with the appearance of a major feature of SON neurons: autocontrol (**Figure 3**).

OTR has been shown to play animportant rolein autoregulation of magnocellular neuronal activity (Richard et al., 1997) and it is well established that OTRs are expressed in adult PVN (van Leeuwen et al., 1985; Freund-Mercier et al., 1987; Tribollet et al., 1988;Yoshimuraetal.,1993;Adanetal.,1995).Duringontogenesis, cells expressing OTR mRNA have been observed in SON and PVN starting from P1 (Yoshimura et al., 1996), a finding consistent with the faint and diffuse expression of OTR present throughout the hypothalamus at this stage (Tribollet et al., 1989). An increase in OTR mRNA expression is detected from P7 and is then maintained fairly stably to adulthood (Yoshimura et al., 1996).

Consistently with changes in OTR expression during PW2, OT and its related analog are most efficient in increasing electrical activity on one-third of SON neurons (supposed to be OT-ergic neurons since they were insensitive to AVP) and somatodendritic release of native OT (Chevaleyre et al., 2000).

The somatodendritic release of OT not only activates action potential firing, it is also determinant for the maturation of glutamatergic synaptic activity and of the neuronal morphology. Indeed at birth, supraoptic neurons display an oblong soma from which 2-3 dendrites with few proximal branches arise (Chevaleyre et al., 2001). However, during PW2, the interplay between incoming glutamatergic inputs and autocontrol induces an intense sprouting of dendritic branches (Chevaleyre et al., 2002), as if neurons were exploring the environment to establish new connections. This sprouting is transient and the neurons acquire their mature morphology (Randle et al., 1986) by the end of PW2. Although partial and concerning only SON OT neurons, these data point to a determinant role of autocontrol of OT neurons during PW2 in rats. This information should be taken into account in our understanding of how OT treatments during infancy can have lifelong consequences on OT-related social diseases (see the chapter below).

### **ESTABLISHMENT OF OT CIRCUITRY**

Since neuroendocrinology was established as a new discipline, it has been clearly demonstrated that magnocellular (both OT and AVP) neurons of adult vertebrate species, including mammals, are primarily projecting to the posterior pituitary lobe to release these hormones into the systemic blood stream (Knobloch and Grinevich, 2014 and references therein). However, embryogenesis of pituitary OT projections remains unexplored. In fact, only two studies demonstrate such projection in rats without discrimination between OT and AVP components. The first study by the Ann-Judith Silverman group (Silverman et al., 1980) showed the existence of neurophysin-positive (i.e. without

In the rat SON during the second PN week, locally released OT promotes calcium mobilization and OT release, and favors the maturation of

activity and mobilization of calcium from intracellular stores and promotes growth of new dendritic branches.

discrimination between OT and AVP) fibers in the posterior pituitary. Another study by the Michael Ugrumov and André Calas groups (Makarenko et al., 2000, 2002), employing DiIbased retrograde tracing in fixed brains, showed certain dynamics of these projections: first connections between the main part of the SON and pituitary are established earlier (detected at E15—earliest time point of the experiment), while the PVN and retrochiasmatic parts of the SON project to the pituitary later at E17 (Makarenko et al., 2000). Intriguingly, the AN, composed mostly of OT neurons, projects to the pituitary only after birth (Makarenko et al., 2002).

Remarkable work on zebra fish larvae showed that OT is essential for the formation of an effective neurovascular interface (i.e., contacts of axonal OT terminals with fenestrated capillaries in the posterior pituitary; Gutnick et al., 2011). Unfortunately, such observation has not been extended to mice lacking OT or OTR, in which the analysis of the pituitary structure has not been reported.

Although central projections of OT neurons in adult rodents (mice, rats, and voles) have been significantly explored during last years (Ross et al., 2009; Knobloch et al., 2012; Dölen et al., 2013), the literature lacks reports on embryogenesis and early postnatal development of OT projections. This concerns both ascending projections of magnocellular OT neurons to the forebrain and descending projections of parvocellular OT neurons terminating in the brain stem and spinal cord.

The major inputs carrying visceral sensory information to the SON and PVN are relayed by catecholaminergic and noncatecholaminergic (glucagon-like peptide 1) neurons whose soma are located in the nucleus of the solitary tract and ventrolateral medulla. Both direct and indirect arguments indicate that these afferents are not functional at birth and develop during the first postnatal week (see review by Rinaman, 2007). Similarly, although no anatomical studies on the dynamics of innervation of OT neurons by glutamate and GABA have been performed, electrophysiological data summarized above suggest long-term postnatal maturation of synaptic inputs.

#### **THE DEVELOPMENTAL TRAJECTORIES OF OT RECEPTOR EXPRESSION**

The developmental trajectories of OT receptor (OTR) expression in the nervous system have been investigated in mice, rats, voles and humans by different experimental approaches such as autoradiography, *in situ* hybridization and transcriptomic analysis, which provide complementary information. While autoradiography maps high affinity OTR at subcellular sites, which grossly correspond to receptor site(s) of activity on neuronal cell bodies and processes, *in situ* hybridization maps the neuronal bodies in which OTR mRNA accumulates. Transcriptomic analysis provides clues on brain regions actively synthesizing OTR mRNA without giving details on the cell types and subpopulations involved (see **Table 2** for the available literature).

Among mammalian species, the rat has been by far the most extensively investigated. In this animal, the three aforementioned techniques resulted in highly comparable and consistent results, which made it possible to trace a developmental trajectory of OTR ontogenesis. This has been schematically summarized in **Table 3** and **Figure 4** (Shapiro and Insel, 1989; Snijdewint et al., 1989;


#### **Table 2 | Studies of OTR trajectories in development.**


**Table 3 | Brain structures expressing OTR in different periods of prenatal and early postnatal life in rats.**

◦*In the adult, binding in the accumbens has been reported to almost completely disappear (Tribollet et al., 1989) or to be greatly reduced as compared to its pick at PN20 (Shapiro and Insel, 1989).*

*\*The appearance of OTR in the hypothalamic ventromedial nucleus has been reported to appear at PN1 (Tribollet et al., 1989) or to emerge only in the adult brain (Shapiro and Insel, 1989).*

Tribollet et al., 1989; Yoshimura et al., 1996; Lukas et al., 2010; Workman et al., 2013). Throughout the embryonic development and the first post-natal days, OTR progressively appears in several brain regions, reaching a well-defined "infant" pattern of distribution around PN10. After PN13, an abrupt decline of OTR density is observed in several areas, accompanied by expression in novel brain regions; this phase has been referred to as the first transition to the adult pattern and is basically completed at PN18. Around and after weaning, a second transition occurs, characterized by a novel reshaping of OTR expression, which slowly disappears from some areas and increases in others. Finally, the adult pattern of OTR expression is achieved at P60-90. Brain structures expressing OTR in different periods of prenatal and early postnatal life in rats (according to Shapiro and Insel, 1989; Tribollet et al., 1989; Lukas et al., 2010) are reported in **Table 3**.

OTR appears as early as E14 pcd is the posterior portion of the neuronal tube that will become the vagal motor nucleus. Even though the mature OT is not detected until much later in development (see above), immature C-terminal extended forms of OT have been visualized much earlier (E16 pcd, see above). It is thus tempting to speculate that some immature OT forms may have an unrecognized role during early development in the vagal subregion. Mechanistically, the delivery of immature OT to the vagal nucleus can be achieved either via the neuropeptide diffusion through the brain tissue after somatodendritic release or via the transventricular pathway (Knobloch and Grinevich, 2014).

Strongly labeled areas of the "infant pattern" around PN10, schematically reported in **Figure 4B**, include the anterior olfactory nucleus, caudate putamen, accumbens, cingulate cortex, bed nucleus of the stria terminalis, some septal, thalamic and amygdaloid nuclei and the dorsal peduncular cortex. A faint but specific signal is observed at this stage in the hypothalamic ventral medial nucleus, a region that will become strongly labeled in the adult brain, while binding in the lateral septum and bed nucleus of the stria terminalis will undergo an intra-regional reshaping during post-natal development. Of relevance, between PN10 and PN25, is the disappearance of binding in the dorsal subiculum, accompanied by the concomitant appearance of increasingly strong binding in the ventral subiculum. During the same period, binding drops in the thalamus, cingulate cortex and CA1 but increases sharply in the accumbens, where it peaks at PN20 to subsequently decline during the second phase of transition to the adult brain. Changes in selected brain regions between week 5 and week 8 have been reported, such as an increase in OTR binding in the ventromedial hypothalamus and a decrease in the lateral septum (Lukas et al., 2010). However, a detailed, comprehensive study of OTR expression trajectories around weaning in males and females is, at present, missing.

Unfortunately, the studies available in literature for mice, voles and humans do not allow for similar comprehensive compilation of OTR expression trajectories during development. Nevertheless, a recent autoradiography study in mice revealed that regionspecific trajectories of OTR expression are present in this species as well (Hammock and Levitt, 2013). A qualitative inspection of OTR binding from P0 to PN60 outlined a progressive strong increase of OTR from P0 to PN14, followed by region-specific up and down regulation of receptor expression. Different temporal expression profiles were reported in the three areas of the brain in which the OTR binding profile was quantified (hippocampus, lateral septum, and neocortex) with no significant differences between sexes. A very high degree of individual variability in OTR expression during ontogenesis was also reported in prairie voles, hampering the detailed delineation of developmental trajectories in this species; however, as observed in mice and rats, a higher density of OTR was observed in some forebrain regions (the septo-hippocampal nucleus and the hippocampus) between PN6 and PN21 as compared to PN1 and PN60 (Prounis and Ophir, 2013). Similarly, in the lateral septum, OTR density increased during the post-natal age, reaching the adult level at weaning; in particular, the binding increased more rapidly in mountain than in prairie voles, resulting in species differences at weaning and adulthood (Wang and Young, 1997). Finally, it is worth noting that in the different species, the overall pattern of the OTR expression during development does not match completely. The

strong transient expression of OTR in the neocortex reported in mice was suggested to account for a species-specific role played by OTR during brain development (Hammock and Levitt, 2013). However, it should be mentioned that a transient OTR expression in the parietal cortex has also been reported in the rat brain (Snijdewint et al., 1989).

As in humans, the transcriptomic analysis of OTR reported a progressive increase in OTR mRNA during embryonic life in five out of six brain areas analyzed (Kang et al., 2011 and http:// hbatlas*.*org/). Remarkably, the receptor level appeared to reach a maximum already before birth and to remain quite stable thereafter, at least in the first 5 years of life, although with some wide individual variations.

An important consideration when comparing developmental trajectories in translational medicine approaches is that the neurodevelopmental stage of the human brain at birth corresponds to rat and mouse brain at PN10 (Workman et al., 2013 and related web tool at http://www*.*translatingtime*.*net/home). Consistently, the human infant pattern of OTR expression is achieved before parturition while, in the rat, the infant pattern of OTR expression is achieved around PN10, and, in mice, between PN7 and PN14. To extrapolate the effects of pharmacological and environmental manipulation of the OT/OTR system in the human newborn brain from experiments performed in mice and rats, treatments should thus be performed in rodents around PN10, when a comparable maturation of the brain and of the OTR system has been reached in these species.

Several environmental factors have been reported to affect OTR expression in embryogenesis and early postnatal life, among which a predominant role is played by social and sensorial experiences, as more extensively discussed in the next paragraph. Furthermore, exposure to drugs and toxic agents can also modulate the OT/OTR system, as outlined by the interesting finding that nicotine and ethanol administration to pregnant rats increases OTR binding in the nucleus accumbens and in the CA3 region of the hippocampus of male offspring (Williams et al., 2009). Recently, an up-regulation of OTR in the nucleus accumbens, medial anterior olfactory nucleus and central and medial nuclei of amygdala has been reported in µ-opioid receptor knock out mice (Gigliucii et al., 2014), which suggests a close link between OTR expression and the opioid reward system.

#### **OT AND SOCIAL EXPERIENCE IN DEVELOPMENT**

Effects of OT during embryogenesis and early postnatal ontogenesis on social life are extensively summarized in recent reviews (Carter, 2014; Hammock, 2014), but the reverse—i.e., the effects of social stimuli on maturation of OT system—are less explored. It has been reported that early social experience tremendously affects physiology of the OT system (Bales and Perkeybile, 2012; Hammock, 2014). Recently it has been demonstrated that early sensory experience (in the newborn) regulates development of sensory cortices via OT-signaling (Zheng et al., 2014).

This oxytocin early regulation may have long-term consequences in adults. It has been known for many years that exogenous OT in neonates can revert the long-term behavioral effects of prenatal stress (Lee et al., 2007b) and has consequences on other endocrine systems (i.e., the estrogen receptor, Pournajafi-Nazarloo et al., 2007) as well as on blood pressure (Holst et al., 2002) in adults. The effects of OT, or of the environmental and familial circumstances resulting in increased production of mature neonatal OT, on adult social behavior mainly remain to be investigated. With respect to OTR expression, increased levels have been observed in offspring after communal rearing (Curley et al., 2012), increased maternal licking/grooming (Champagne et al., 2001) and social enrichment (Champagne and Meaney, 2007). On the contrary, late weaning has been reported to reduce OTR density in selected, socially relevant, brain regions (Curley et al., 2009). Furthermore, maternal separation has been found to induce a complex modulation of OTR expression with regionspecific up and down regulation of OTR (Lukas et al., 2010). It is important to note, at the end of this section, that exploration of early life experience on OT/OTR system in animals provides a scientific basis for child care and new therapeutical approaches to ameliorating social alterations occurring in adult patients afflicted with Autism Spectrum Disorders or the PWS (see below).

#### **OT AND DEVELOPMENTAL NEUROLOGICAL DISORDERS**

Many reviews are exhaustive on the involvement of OT in shaping and regulating the social brain (Meyer-Lindenberg et al., 2011; Striepens et al., 2011; McCall and Singer, 2012) or in learning and memory (Chini et al., 2014). The strongest line of demonstration of the OT system in social behavior, feeding behavior and maternal care is via the study of knockout mice in which either the *OT* or the *OTR* genes are inactivated (**Table 4**). Mice constitutively lacking OT (*Oxt*−*/*−) are unable to release milk and have impaired social memory (Ferguson et al., 2000). Mice

#### **Table 4 | OT and neurodevelopmental disorders.**

with constitutive ablation of the OTR gene (*Oxtr*−*/*−*)* display a behavior very similar to the *Oxt*−*/*<sup>−</sup> mouse, mainly marked by social deficits (Takayanagi et al., 2005). In addition, although learning is normal in *Oxtr-/-* mice, reversal learning is strongly decreased, indicating impaired cognitive flexibility reminiscent of the ASD syndrome (Sala et al., 2011). Importantly, even a 50% loss of the OTRs also leads to an impairment of social behavior, suggesting that a fine tuning of the OT system is necessary to control behavior (Sala et al., 2013). Other mouse models, in which the knock-out of a specific gene induced a disruption in the OT system that has been linked to a pathological phenotype, reinforce the role of OT in neurodevelopmental disorders (**Table 4**). Recently, it has been shown that disruption of the neonatal surge of OT coming from the mother during delivery results in autistic-like features in the adults (Tyzio et al., 2014). All these data suggest that an early postnatal injury or dysfunction of the OT system has consequences in infant and adult behaviors. Unfortunately, the contribution of early life experience to OT signaling, in a pathological context or following a trauma, has not been extensively studied in the brain of highly social mammalian species such as voles or monkeys. A limited number of reports (Bales et al., 2007, 2011; Ahern and Young, 2009) demonstrated that social deprivation and enrichment paradigms induce changes in OT synthesis and OTR binding in voles (Bales et al., 2007, 2011; Ahern and Young, 2009). While it is not yet clear how these changes occur, the most likely scenario is the environmental influence on the gene expression through epigenetic mechanisms.

In humans, autistic spectrum disorder (ASD) is a broadlydefined disorder that mainly affects behavior and cognition. Social interaction impairments are the most characteristic deficits in ASD. Interestingly, several lines of evidence suggest a role of OT in the etiology of ASD (Harony and Wagner, 2010; LoParo and


Waldman, 2014) and OT's therapeutic effects have been observed in social communication (Aoki et al., 2014), however, they are still debated (Guastella et al., 2014). Furthermore, it is also acknowledged that an alteration in the OT-system might be involved in neurodevelopmental disorders marked by social cue deficits such as Fragile X Syndrome, Williams Syndrome and PWS (see **Table 3** and Francis et al., 2014).

PWS is one of the best reported examples of a neurodevelopmental disease characterized mainly as an eating disorder with behavioral and social disturbances (**Figure 5**; Butler et al., 2011; Dykens et al., 2011; McAllister et al., 2011; Cassidy et al., 2012; Jauregi et al., 2013). PWS is a rare genetic disease with an estimated prevalence worldwide of 1 in 10,000–30,000 individuals (Cassidy et al., 2012). Patients with PWS exhibit a complex and progressive phenotype. Their eating behavior is mainly characterized by two opposite stages (Butler et al., 2010; Miller et al., 2011). During phase I of the syndrome, apparent at birth, the suckling activity is weak or absent and babies show little interest in feeding during the first few months of their lives. After 2 years, it is characterized by a true hyperphagia with obesity. In fact, PWS children initially display anorexia as neonates and then switch to hyperphagia with obesity (Tauber et al., 2014).

In parallel with the eating problem, PWS patients have mild to moderate intellectual disability and behavioral alterations (Ho and Dimitropoulos, 2010; Chevalere et al., 2013), including emotional outbreaks (temper tantrums) and compulsive traits (Dimitropoulos et al., 2006). Repetitive and ritualistic behaviors and difficulty with routine changes (Holland et al., 2003; Greaves et al., 2006; Dykens et al., 2011), similar to those found in autistic spectrum conditions, have also been described. Indeed all PW patients share some features of ASD (Koenig et al., 2004). Some patients, who are diagnosed as autists, share some features observed in PWS and are defined as PWS-like patients (Schaaf et al., 2013).

Nevertheless, only a few patients with PWS are diagnosed as autists. Patients with PWS present greater overall behavior disturbance than age-matched mentally retarded patients, but score comparably to patients with psychiatric disorders. Indeed, they frequently display anxiety traits and anxious mood comparable to those of patients presenting anxiety disorder or schizophrenia. They also show pronounced emotional liability and a striking inability to control their emotions, which results in frequent emotional outbreaks (temper tantrums), which can occur due to an impaired capacity to understand the motivations of others in the social environment, possibly indicating deficits in "theory of mind" (the ability to attribute mental states to others) and empathy (the ability to infer emotional experiences) (Lo et al., 2013). Importantly, Tauber et al. (2011) reported the first clinical trial on OT that showed that a single intranasal administration of OT rescue some behavioral features, such as increased trust and decreased signs of depression as well as emotional outbreaks, in patients with PWS (**Figure 5**). However, a double-blind randomized cross-over trial of OT nasal spray performed in 22 PWS patients (12–29 years old) did not find statistically significant effects using 18–40 IU of intranasal OT twice daily for 8 weeks (Einfeld et al., 2014). The authors reported an increased number of temper tantrums in the patients receiving the high dose of OT and no effect was observed in patients treated with low doses. The discrepancies between these two studies may be explained by the dose used, the short wash-out period (15 days), the duration of treatment and the sex ratio (most patients were male in the report of Einfeld and colleagues compared to the report of Tauber's team).

Much of the phenotype of PWS is consistent with a hypothalamic defect (Swaab et al., 1995; Swaab, 1997a,b) characterized by a reduction of OT expressing neurons in the PVN, mostly represented by parvocellular OT neurons (Swaab et al., 1995; Swaab, 1997a,b). These cells project to the brainstem nuclei, including the nucleus of the solitary tract, where OT acts as a powerful anorexic peptide (Atasoy et al., 2012). Normal OT plasma levels have been detected in 17 PWS adult patients (Höybye et al., 2003), but elevated cerebrospinal fluid (CSF) OT levels have been reported in five PWS patients (Martin et al., 1998). This high level of OT in the CSF most likely relies on OT released from magnocellular neurons, contacting both the vasculature of the posterior pituitary and the lumen of the third ventricle (Knobloch and Grinevich, 2014). However, such discrepancy between OT released into the blood and CSF needs further clarification, especially with respect to methodological limitations for measuring OT in biological samples (McCullough et al., 2013).

Genetically, PWS results from the lack of expression of several contiguous imprinted genes located in the 15q11–q13 region. These genes are regulated by genomic imprinting: a mechanism leading to the transcriptional expression of the paternal alleles of these genes only, the maternal alleles being silent (i.e., not expressed). Recently, pathogenic mutations of *MAGEL2* alone have been reported in four patients (Schaaf et al., 2013), causing a classical PWS in one patient and PWS-like phenotypes in the other three. All of the cases described were also diagnosed with early feeding problems and ASD, without severe obesity. These results underline *MAGEL2's* contributing role in cognitive and behavioral alterations in PWS and early feeding in general.

Francoise Muscatelli's group created a mouse model deficient for *Magel2*. These mice showed an altered onset of suckling activity and subsequent impaired feeding leading to 50% of neonatal lethality, affecting both males and females (Schaller et al., 2010). Impressively, there is an obvious alteration of production of mature OT in the PVN of *Magel2*-deficient pups, while a single administration of OT, in a restricted time window after birth, allows resetting of the feeding behavior and consequently rescues the life of all pups (Schaller et al., 2010, p. 24). Furthermore, inactivation of *Magel2* induces in adult males (but not in females) a deficit in social recognition and social interaction as well as a reduced learning ability with an alteration of social and spatial memory. A daily administration of oxytocin (OT) in the first week of life is sufficient to restore suckling activity at birth and to restore a normal social behavior and learning abilities in adult mutant males (Meziane et al., in press).

Altogether, these results suggest that an alteration of the OT system around birth has early and long term consequences on feeding and social behaviors and on cognition. Importantly, an OT treatment of Magel2-deficient pups in the first post-natal week partially restores a normal anatomy of the OT system and prevents deficits in social behavior and learning in adults. This concept opens the door to a powerful pharmacological therapy in early infancy for the PWS and might be considered for other pathologies such as autism spectrum disorders. Importantly, a clinical trial on OT treatment of PWS infants has been initiated by Maithé Tauber and her team.

#### **OT PATHWAYS IN PERINATAL ONTOGENESIS: PHYSIOLOGY, PATHOLOGY, AND TREATMENT**

While the picture of OT pathways development is far from complete, at least three conclusions can be made (see **Figure 6**):


The delayed maturation of the central OT system in rodents (and humans) may be caused or may correlate with the feature of these species, which are born relatively immature (termed "altricial" species). From a comparative point of view, it would be interesting to compare the dynamics of OT system maturation with the animals which are relatively mature and mobile from the moment of birth (termed "precocial" species, such as guinea pigs).

Since the distant/extrahypothalamic OT axons during neonatal period in studied (altricial) species are absent (at least not reported), but OT neurons were responsive to externally applied OT, it is most likely that neonate rodents operate by endogenous OT released from somato-dendritic compartments of OT cells. While PVN OT cells are closely located to the ependymal layer of the third ventricle, OT may diffuse also to the cerebrospinal fluid, resembling the evolutionarily old "transventricular" pathway of OT action (Knobloch and Grinevich, 2014). Moreover, OT can

reach the brain directly via brain capillaries, as the blood brain barrier is not formed at that age (Ugrumov, 2010 and references therein), allowing peripherally administered OT to efficiently reach the brain (Meziane et al., in press).

In line with the "diffusion-like" mode of OT action, the report on newborn and young mice (P0–P14; Zheng et al., 2014) showed no axonal projections in the somatosensory cortex, while the effects of OT on electrical activity of cortical neurons as well as on sensory processing were very prominent (Zheng et al., 2014). Keeping this contradiction in mind, the authors speculated about "diffusible" OT reaching the cortex from the hypothalamus. Despite the lack of literature, it is tempting to speculate that magnocellular OT axons will grow to forebrain regions and parvocellular OT neuronal axons to brain stem/spinal cord only after weaning, to execute addressed OT release aimed to orchestrate autonomic and behavioral responses respectively (see above).

Animal models of social deficiency (such as OT and OTR knockout mice) and human cases of OTR gene duplication (Bittel et al., 2006) support the critical role of neonatal OT signaling in the development of an adequate social skill. The model of PWS—Magel2 knockout mice has pronounced OT deficiency with delayed expression of mature OT on the day of birth.

The discovery that peripheral OT administration to neonate Magel2 knockout mice rescues both the feeding problem in neonates and social impairments in adults, gives the reason to dissect alteration in sequences of OT signaling (i.e., in steps of OT expression in OT cells, OT transport via axons and their releasing capacity, expression of OTR etc.), which can be considered as target(s) for exogenous OT. Keeping in mind the potential OT sensitivity of OT neurons, and many other brain cells, as well as long-lasting OT effects, someone may hypothesize that exogenous OT stimulates transcriptional, electric and secretory activity of OT neurons, which start to release larger amounts of OT to fill in the brain. Additionally, upregulation of central OTR expression can facilitate OT signaling. However, the testing of these scenarios requires further comprehensive research.

#### **ACKNOWLEDGMENTS**

The preparation of this review was supported by the Chica and Heinz Schaller Research Foundation, German Research Foundation (DFG) grant GR 3619/4-1, SFB 1134, Royal Society Edinburgh Award, and German Academic Exchange service (DAAD) program for partnership between German and Japanese Universities, PHC PROCOP program (DAAD and Campus France) (to Valery Grinevich), Telethon Foundation grant GGP12207 (to Bice Chini), ANR and MESR (to Michel G. Desarménien), INSERM, the European community (grant #512136 PWS), Prader-Willi France (to Maithé Tauber and Françoise Muscatelli) and Fondation Jerôme LeJeune (to Françoise Muscatelli). The authors thank Thomas Splettstoesser (SciStyle; www*.*scistyle*.*com), Vivien Chevaleyre, and Muriel Asary for their help with the preparation of Figures and Anne Seller for proof reading the manuscript.

#### **REFERENCES**


of the rat during labour. *J. Endocrinol*. 86, 221–229. doi: 10.1677/joe.0. 0860221


with autism. *J. Intellect. Dosabil. Res*. 50, 92–100. doi: 10.1111/j.1365- 2788.2005.00726.x


late-onset obesity. *Neuroreport* 19, 951–955. doi: 10.1097/WNR.0b013e328 3021ca9


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 14 November 2014; paper pending published: 01 December 2014; accepted: 17 December 2014; published online: 20 January 2015.*

*Citation: Grinevich V, Desarménien MG, Chini B, Tauber M and Muscatelli F (2015) Ontogenesis of oxytocin pathways in the mammalian brain: late maturation and psychosocial disorders. Front. Neuroanat. 8:164. doi: 10.3389/fnana.2014.00164 This article was submitted to the journal Frontiers in Neuroanatomy.*

*Copyright © 2015 Grinevich, Desarménien, Chini, Tauber and Muscatelli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Characterization of a mammalian prosencephalic functional plan

#### **Sophie Croizier † , Sandrine Chometton, Dominique Fellmann and Pierre-Yves Risold\***

EA 3922, SFR FED 4234, UFR Sciences Médicales et Pharmaceutiques, Université de Franche-Comté, Besançon, France

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Luis Puelles, Universidad de Murcia, Spain Gonzalo Alvarez-Bolado, University of Heidelberg, Germany José L. E. Ferran, University of Murcia, Spain

#### **\*Correspondence:**

Pierre-Yves Risold, EA 3922, SFR FED 4234, UFR Sciences Médicales et Pharmaceutiques, Université de Franche-Comté, 19 rue Ambroise Paré, 25030 Besançon cedex, France e-mail: pierre-yves.risold@ univ-fcomte.fr

#### **†Present address:**

Sophie Croizier, The Saban Research Institute, Neuroscience Program, Children's Hospital Los Angeles, University of Southern California, Los Angeles, California 90027, USA

#### **INTRODUCTION**

As a whole, the hypothalamus is involved in an extremely large range of functions, including neuroendocrine and visceral responses, thermogenesis, circadian or seasonal cycles, sleep or general arousal, the expression of specific instinctive behaviors, the control of rhythmic cortical (hippocampal) neuron firing, emotion and reward. Therefore, the hypothalamus is a complex structure composed of dozens of cell groups or nuclei that are often involved in several of these responses. Classically, the hypothalamus has been divided into four anteroposterior regions (preoptic, anterior, tuberal and posterior regions) and three longitudinal zones (periventricular, medial and lateral zones) (Swanson, 1987). This organizational scheme has been widely accepted by anatomists and physiologists during the past decades but is not satisfactory, as most of the borders are not clear and are often arbitrarily drawn. In light of anatomical findings acquired during the late 1980s and early 1990s, the organization of the hypothalamus has been revised around the concept of a behavioral control column that is composed of the medial zone nuclei (**Figure 1A**; Swanson, 2000, 2005). Following this new concept, each medial zone nucleus is involved in pathways that include the tectum, thalamus and telencephalon, referencing the classical circuit described by Papez in 1937 (Papez, 1995). This new view

Hypothalamic organizational concepts have greatly evolved as the primary hypothalamic pathways have been systematically investigated. In the present review, we describe how the hypothalamus arises from a molecularly heterogeneous region of the embryonic neural tube but is first differentiated as a primary neuronal cell cord (earliest mantle layer). This structure defines two axes that align onto two fundamental components: a longitudinal tractus postopticus(tpoc)/retinian component and a transverse supraoptic tract(sot)/olfactory component. We then discuss how these two axonal tracts guide the formation of all major tracts that connect the telencephalon with the hypothalamus/ventral midbrain, highlighting the existence of an early basic plan in the functional organization of the prosencephalic connectome.

**Keywords: melanin concentrating hormone, cell cord, pioneer tracts, lateral hypothalamic area, medial forebrain bundle**

> of the hypothalamic organization is interesting as it suggests that this region is fully integrated within the complex prosencephalic networks that control behavioral expression. Therefore, the hypothalamus is capable of influencing telencephalic centers, including the cerebral cortex, as well as being influenced by descending projections (Risold and Swanson, 1996, 1997; Risold et al., 1997). Conspicuous convergences have appeared between the organization of these connections and those of the classical striato-nigral and mesotelencephalic circuits, and a revision of the telencephalic organization has been proposed (**Figure 1B**; Risold et al., 1997; Swanson, 2000, 2003; Risold, 2004).

> Although interesting, this concept primarily involves the medial zone nuclei at hypothalamic levels but ignores large sections of this structure, especially the entire hypothalamic lateral zone (lateral hypothalamic area, LHA) (**Figure 1A**). The LHA is a poorly differentiated region that has always been viewed as a rostral extension of the brainstem reticular formation or a bed nucleus of the medial forebrain bundle (mfb). The mfb is the major fiber tract of the basal prosencephalon that passes through the LHA and bidirectionally connects more than fifty cell groups in the brainstem and telencephalon (Nieuwenhuys et al., 1982; Swanson, 1987). The mfb is a specific attribute of the LHA

throughout its entire alar-basal extent. However, the LHA is not homogeneous. The basal portion of the LHA contains abundant cell populations that are characterized by the expression of specific peptides, such as melanin-concentrating hormone (MCH) and hypocretins/orexins, and widespread projections from the cerebral cortex to the spinal cord.

Tremendous progress in expanding the general knowledge of the forebrain embryonic development has been made over the last twenty years. Recent studies, alongside anatomical data, have led to a better understanding of the organization of the vertebrate forebrain, which has allowed for a better comprehension of its evolution (Puelles, 2001; Aboitiz, 2011). To understand the organization of the basal portion of the LHA, our group analyzed the comparative anatomy and development of hypothalamic neurons that produce MCH (Croizier et al., 2013). In the present analysis, we revised some of these observations from past and recent developmental studies regarding the forebrain to better understand the relative role of the LHA within the context of a putative general prosencephalic framework. We observed that the whole ventral prosencephalon is organized around a precocious structure, previously named the cell cord and from which the LHA differentiates, in a timely manner. This primary structure defines two axes that align onto two fundamental components: a tractus postopticus (tpoc)/retinian component and a supraoptic tract (sot)/olfactory component. These two axes determine the path of the mfb and provide what can be described as a basic structural framework for a prosencephalic "functional plan".

### **DEVELOPMENTAL GENE PATTERNS AND HYPOTHALAMIC SUBDIVISIONS**

Very complex molecular interactions occur at the origin of the hypothalamic regions, which are very heterogeneous, even at the earliest stages. These patterns have been extensively analyzed in many more detailed works to which the reader may refer (Shimamura et al., 1995; Nieuwenhuys et al., 1998; Puelles et al., 2012). Longitudinal and transverse axes in the embryos have been revised on the basis of these patterns of gene expression (Puelles et al., 2012). Although the terminology proposed by Puelles and Rubenstein is widely used in the developmental field (see for instance "prethalamus"), in the field of adult neuroanatomy the traditional axes and nomenclature are maintained. Since we strive to be clear for all interested readers, "adult" or "developmental", we often hesitate to use one or the other name for a structure or spatial relation. For this practical reason, and in particular to respect the main information flows in the adult brain, we have used sometimes similar rostrocaudal and dorsoventral axes in the embryonic brain as are used in the adult brain (for example see in **Figure 3**). We also use the terminology prethalamus-ventral thalamus, in this way adding "prethalamus", as is often done in developmental studies, for the presumptive regions of the zona incerta and the ventral lateral geniculate nuclei. The preoptic region (POA) (actually a part of the telencephalon, see below), the anterior region (or alar hypothalamus) that is supraoptic and regions that are posterior (basal) or postoptic represent no problem (**Figure 2**).

From the general literature in this field one important observation retained our attention: very early in development, these gene expression patterns bear resemblances to and are contiguous with those in structures adjacent to the hypothalamus, which suggests that some hypothalamic boundaries are not sharply delineated (**Figure 2A**). Consider the following examples.

**FIGURE 2 | (A)** The distribution patterns of transcription factors agree well with the alar-basal divisions of the hypothalamus into preoptic, anterior and posterior/postoptic regions. However, these patterns extend outside the borders of the hypothalamus, involving the ventral telencephalon, prethalamus-ventral thalamus and ventral midbrain. MCH neurons and neurons of the VMH are generated from Nkx2.1/Nkx2.2 expressing neuroepithelial zones in the postoptic region. **(B)** Based on immunohistochemical analysis of a horizontal section of an E15 rat embryonic hypothalamus, the VMH clearly express both Nkx2.1 and Nkx2.2. **(C)** Figures from Keyser (1972) illustrating the differentiation of the early neurogenic zone in E12 and E13 Chinese hamster embryos: the first appearance of a longitudinal zone in the ventral mesencephalon and dorsal hypothalamus correspond to the cell cord **(A)**. At E13, neurogenesis involved larger regions in the preoptic/ventral telencephalon; in other figure from Keyser that is not shown here, this author observed these regions forming one single continuum (as in **D**). This continuum takes the shape of (Continued)

#### **FIGURE 2 | Continued**

an inverted Y. **(D,E)** Figure adapted from Croizier et al., 2011 illustrating the distribution of neurons generated at E11 on an E13 rat embryo. BrdU was injected into the pregnant dam at E11, and embryos were taken 2 days later at E13. BrdU was detected by immunohistochemistry on horizontal sections. The distribution pattern of these nuclei is schematized on a sagittal section in **(D)**. BrdU-labeled nuclei follow an inverted Y pattern. In **(E)** pictures are arranged from dorsal (1) to ventral (4). **(F)** Gradients of neurogenesis in the ventral diencephalon. Left side: Schematic representation of the neurogenic gradients in the hypothalamus and prethalamus-ventral thalamus, as described by Altman and Bayer (1986). The LHA is generated between E11and E13, the medial hypothalamus from E13 to E15 and the periventricular zone from E14 to E17. Note the medial to lateral gradient in the prethalamus-ventral thalamus (generated from E13 to E15). Right side: Drawing summarizing the gradients in the ventral diencephalon: lateral to medial gradients (red arrows) in the hypothalamus suggest the apparent or passive migrations of neurons in lateral territories (LHA or VMHvl), but the prethalamus-ventral thalamus requires the effective migration of cells away from the ventricular surface (black arrow). Abbreviations: ANT: anterior hypothalamic area; cpd: cerebral peduncle; fx: fornix; LHA: lateral hypothalamic area; MCH: melano-concentrating hormone expressing neurons; MGE: medial ganglionic eminence; MM: mammillary body; mtt: mammillothalamic tract; opt: optic tract; PAL: pallidum; POA: preoptic area; PRO: presumptive preoptic area; RCH: retrochiasmatic region; THv: prethalamus or ventral thalamus; TUB: presumptive tuberal hypothalamic region; VMH: ventromedial hypothalamic nucleus; VMHvl: ventrolateral part of the VMH; ZI: zona incerta; zli: zona limitans intrathalamica; V3: third ventricle.

#### **THE HYPOTHALAMIC/TELENCEPHALIC BORDER**

The POA shares many characteristics and developmental expression patterns with the pallidum, and has been considered a telencephalic structure (Moreno and González, 2011; Puelles et al., 2012). The preoptic anlage expresses the telencephalic marker Foxg1, distinguishing it from hypothalamic structures. Shh and Nkx2.1 are co-distributed in the POA, but their expression patterns extend into the pallidal anlage. Although the POA expresses Shh, it does not express Nkx2.2, whereas the hypothalamic regions do express Nkx2.2. Defining the telencephalo-hypothalamic limit has always been problematic, despite functional and connectional studies in adult animals that have been interpreted in favor of including the POA as part of the hypothalamus (Swanson, 1987; Risold et al., 1997). Earlier authors often viewed this region to be a medial and unevaginated section of the telencephalon or telencephalon impar (His, 1893; Herrick, 1910; Kuhlenbeck, 1929). These early claims have been strengthened by recent studies of gene expression patterns (Moreno and González, 2011). This definition is further reinforced by recent findings that, similar to the lateral and medial ganglionic eminences (LGE and MGE, respectively), the POA produces GABAergic (gamma aminobutyric acid) interneurons that tangentially migrate into the cerebral cortex (Brown et al., 2011; Gelman et al., 2011; Vitalis and Rossier, 2011). In addition, migrating cells from the olfactory epithelium colonize the medial septal and preoptic areas. Many of these cells express GnRH (gonadotropin-releasing hormone) and are neuroendocrine neurons, including in the septal region. Neuroendocrine functions are a hallmark of the hypothalamus, but a putative neuroendocrine zone extends beyond the actual rostral border of the hypothalamus. *Therefore, a* *putative hypothalamo/telencephalic limit at the septal/preoptic border is not delineated by the distribution pattern of neuroendocrine GnRH neurons, the origin of GABAergic interneurons migrating into the pallium, or developmental gene expression patterns*.

#### **ANTERIOR HYPOTHALAMUS/PRETHALAMUS-VENTRAL THALAMUS**

At early embryonic stages, Pax6 is expressed in a continuum that includes the presumptive prethalamus-ventral thalamus and extends into the optic vesicle (Stoykova et al., 1996; Puelles et al., 2013). In the hypothalamus, this pattern involves a strip of tissue between the optic stalk and the prethalamus-ventral thalamus (**Figure 2A**). The *Pax6* expression pattern (encompassing the alar hypothalamus and prethalamic eminence) clearly suggests an elongation of the primary embryonic brain (see the interesting paper of (Suzuki et al., 2014) about the rise of the eyes in chordates). As previously noted, Pax family members are involved in the formation of the retina and are also involved in the guidance of retinal projections and the differentiation of retinorecipient structures into the suprachiasmatic nucleus (SCN). In vertebrates, such as lampreys or batrachians, the SCN is adjacent to the prethalamus. In the anuran, Dominguez confirmed this close and continuous positioning of the anterior hypothalamus and prethalamus (Dominguez et al., 2013). In some species, projections from the prethalamus and from the tectum reach the retina through the optic tract. In mammals, the ventral geniculate body has retained a strong retinal afferent, and the intergeniculate leaflet, a small structure of prethalamicventral thalamic origin, has strong bidirectional connections with the SCN, reminiscent of the adjacent positions of the SCN and the prethalamus in non-amniote vertebrates (see the recent work of Suzuki et al., 2014 as well). *Therefore*, *early Pax6 expression patterns prefigure optic related pathways that in the alar hypothalamus and in the prethalamus-ventral thalamus*.

#### **THE HYPOTHALAMUS AS A ROSTRAL STRUCTURE**

Also of particular interest for the development of the hypothalamus are the longitudinal expression patterns of Shh and Nkx2.2, which label a band of hypothalamic neuroepithelial tissue that rostrally extends from a similar band in the ventral mesencephalon (Shimamura et al., 1995; Alvarez-Bolado et al., 2012). When the anterior neuropore closure occurs, the initial expression patterns of Shh and Nkx2.2 involve the differentiating zona limitans intrathalamica (zli), at the junction between the prethalamus and thalamus. At roughly the same stage, corresponding to the beginning of neurogenesis, Shh expression (but not Nkx2.2) appears in a telencephalic region. The domain of Nkx2.2 overlaps partially with both Pax6 and Nkx2.1 expression domains (Croizier et al., 2011). The Pax6/Nkx2.2-rich region gives rise to anterior hypothalamic structures whose composition in the adult are not yet completely clear. However, the Nkx2.1/Nkx2.2 regions give rise to vast portions of the basal hypothalamus. We have clearly shown (Croizier et al., 2011) that neurons that produce MCH are generated and differentiate under the control of Shh (Szabó et al., 2009; Alvarez-Bolado et al., 2012) in this sector of the embryonic wall. In the model of Puelles, this is the RTu-I portion of the basal hypothalamus (Puelles et al., 2012); in this way, the MCH cells would represent a precocious peduncular superficial derivative of the dorsal retrotuberal basal domain.

The ventromedial hypothalamic nucleus is produced by a more rostral portion of the Nkx2.1/Nkx2.2 region (**Figure 2B**, and see (Altman and Bayer, 1986) for the origin of this nucleus). Shimogori identified this region as the "intrahypothalamic diagonal", on the basis of multiple gene expression patterns ("diagonal", that is, neither columnar nor prosomeric, but somewhat in the middle of both, rather confusingly) (Shimogori et al., 2010). The Nkx2.2 and Shh expression patterns extend into the brainstem and are known to be involved in the genesis of other very early defined neurons, including serotonergic neurons, which have diffuse projection patterns similar to MCH neurons (Ye et al., 1998). Somewhat later, Shh is also involved in the differentiation of dopaminergic ventral midbrain neurons (Riddle and Pollock, 2003; Perez-Balaguer et al., 2009). *Therefore, the co-expression of two primary markers of the basal neural tube extends into the postoptic (i.e., basal) hypothalamus*, and we observe the early production of specific neuron populations with diffuse projection patterns as MCH and serotonergic neurons in corresponding hypothalamic regions and hindbrain.

The mammillary nuclei and regions of the very ventromedial hypothalamus (VMH, arcuate nucleus) are generated by an Nkx2.1 expressing neuroepithelial zone (Puelles and Rubenstein, 2003). This appears to be the only pattern that does not show any sign of extension outside of the hypothalamic borders (although, see Puelles et al., 2013).

From all these observations we can conclude that the hypothalamus has diverse origins. Patterns of gene expressions are very complex, and precise combinations of gene expression are associated with specific cell groups or nuclei. However, at the very early stages, the patterns of Shh, Pax6, Nkx2.2 and Nkx2.1 expression indicate that the POA is a part of the telencephalon (Puelles et al., 2012) and the anterior (alar hypothalamus) and postoptic regions (basal hypothalamus) share some gene expression patterns with the prethalamus and the ventral (basal) brainstem.

### **EARLY NEUROGENESIS IN THE HYPOTHALAMUS—EVIDENCE OF A PRIMARY STRUCTURE**

In the embryonic neural tube, neurogenesis (neuron production and therefore the formation of a postmitotic "mantle layer") begins in the ventral hindbrain, behind the cephalic flexure. This neurogenic zone extends both caudally and rostrally. In more rostral regions, Keyser very precisely depicted the patterns of morphologic modifications that occur in the hypothalamic periventricular and mantle layers of the Chinese hamster (Keyser, 1972), describing the development of a "matrix" that can be translated into a very dynamic view of the pattern of neurogenesis in the diencephalon (**Figure 2C**, right and left diagrams). These observations by Keyser can easily be correlated with neurogenesis studies using tritiated nucleotides or BrdU. Keyser showed that the early pattern of neuron production is not uniform throughout the hypothalamus. Neurogenesis begins in a column of cells that was named the "cell cord" by Gilbert in 1935 in the human embryo (cited in Keyser, 1972), and it was more recently observed again in the mouse embryo by Croizier (Croizier et al., 2011; **Figures 2D,E**). The position of the cell cord on the model of Puelles is probably basal, immediately under the alar-basal boundary (Puelles et al., 2012). This column of early neurogenesis gives rise to the first generated hypothalamic neurons that ultimately form the postchiasmatic lateral hypothalamus, and extends into the ventral midbrain. MCH expressing neurons are among the first generated cells in this region, and we showed their early differentiation within this cell cord (Croizier et al., 2011). From this original sector, neurogenesis involves more rostral territories (presumptive entopeduncular nucleus (Altman and Bayer, 1986)). Therefore, the early mantle layer forms the shape of an inverted Y in the hypothalamic primordium (**Figures 2C,D**). The vertical (postoptic) limb and stem of the Y are longitudinal and correspond with the Shh/Nkx2.2/Pax6 expression patterns. The supraoptic arm, however, does not respect these longitudinal patterns and is transversally oriented (longitudinal and transverse here in the sense of the model by Puelles et al., 2012). A chronological correlation could be made between the development of this supraoptic arm and the differentiation of the zli (another transverse feature), the differentiation of the telencephalic vesicle or the telencephalic expression of Shh. However, causal links have not yet been demonstrated.

Keyser also observed that the development of the mantle layer extends from these hypothalamic initial regions, in both the ventral (hypothalamic) and dorsal (ventral thalamic/prethalamic) directions (Keyser, 1972). Altman and Bayer described a lateral to medial gradient of neurogenesis in the hypothalamus (lateral to periventricular) that is clearly in agreement with the observations of Keyser (**Figure 2F**; Altman and Bayer, 1986). A similar gradient was also clearly demonstrated for MCH expressing neurons (Brischoux et al., 2001; Croizier et al., 2010). On the contrary, both Altman and Bayer (1986) in the rat, and Keyser (1972) in the Chinese hamster, described a medial to lateral gradient of neurogenesis in the prethalamus that generates the zona incerta (adult ventral thalamus). This gradient is opposite to the hypothalamic gradient, although both structures are generated during the same period (between E11 and E16 in the rat). Therefore, these two opposing gradients involving two adjacent structures, lateral to medial for the hypothalamus and medial to lateral for the zona incerta, clearly designate a unique sector of origin, and the region giving rise to the cell cord is a good candidate. The opposite (lateral-medial and medial-lateral) gradients of genesis in the hypothalamus and prethalamus-ventral thalamus suggest distinct strategies of cell migration; the lateral to medial hypothalamic gradient suggests a dominant passive migration, as was shown for MCH expressing neurons in the dorsal hypothalamus and is also evident for the VMH. However, the medial to lateral gradient of the zona incerta indicates that lateral neurons must actively migrate far from the ventricular surface (**Figure 2F**; Keyser, 1972; Altman and Bayer, 1986).

Therefore, the initial neurogenesis in the hypothalamus produces a primary inverted Y-shaped structure named here the cell cord (although this structure is slightly different but encompasses the original cell cord). The prethalamus-ventral thalamus and medial hypothalamus are then produced through inverted gradients that are dorsal and ventromedial to this initial cell cord, respectively.

### **THE STRUCTURAL WIRING OF THE HYPOTHALAMUS AND THE CHRONOTOPIC DIFFERENTIATION OF THE WHOLE PROSENCEPHALON**

Gene expression patterns indicate that the development of the hypothalamus is a multifactorial process, but gradients of neurogenesis show that time is a key parameter. The study by Altman and Bayer (1986) emphasized this point when these authors described three waves of genesis to form the three longitudinal zones of the hypothalamus, even if "three waves" might not to be literally considered (Alvarez-Bolado et al., 2012). Time is also a critical parameter according the description of the "matrix" (mantle layer) and cell cord by Keyser (1972). Relying on the development of MCH expressing neurons, we have more recently illustrated the importance of chronology in the organization of this conspicuous neuron population in the posterior hypothalamus in both rat and mouse (Croizier et al., 2010, 2011). Therefore, an analysis of the hypothalamic wiring is inseparable from the timely and sequential events that lead to the morphofunctional organization of the whole prosencephalon.

Tract formation in the hypothalamus immediately follows neurogenesis and accompanies the differentiation of hypothalamic regions. The main axonal bundles of the prosencephalon, transverse and longitudinal, have been illustrated on the prosomeric model by Puelles et al. (see their Figure 8.34) (Puelles et al., 2012). Pioneer tracts have been well described in a series of papers (Herrick, 1910; Easter et al., 1993; Mastick and Easter, 1996). Pioneer tract organization is very well conserved in the young embryo of all vertebrates, from fishes to mammals. The tpoc is the first prosencephalic tract. It is composed of commissural axons (postoptic commissure) and axons running toward the ventral midbrain, parallel to the Nkx2.2 expression domain. This tract joins the medial longitudinal fasciculus (mlf) in the midbrain. The sot and the stria medullaris are formed in the preoptic/entopeduncular primordium. The sot joins the tpoc by passing over the optic stalk (Anderson and Key, 1996), while the stria medullaris runs toward the dorsal diencephalon. The long projections of the hypothalamus subsequently organize along these pioneer tracts, and several stages can be recognized. *Each stage is correlated with a different degree of organization in the embryonic brain and is reflected in the structure of the adult hypothalamus*.

– The first stage is concerned with the initial formation of the pioneer tracts. Guidance cues, such as Slit/ROBO family members, and transcription factors, such as Pax6, play important roles in constraining the paths of pioneer tracts (Mastick et al., 1997; Nural and Mastick, 2004; Ricaño-Cornejo et al., 2011). Both tpoc and sot clearly recall the early neurogenic pattern and travel along the inverted Y-shaped cell cord.

Initially, these tracts are composed of descending axons. The first MCH expressing neurons and the first neurons in the ventrolateral VMH are generated during this preliminary stage (**Figure 3A**). MCH and SF-1 are expressed in neurons within the dorsal and ventral cell cords, respectively, and their axons have been traced in the tpoc running toward the mesencephalon (Croizier et al., 2011; Cheung et al., 2013). In the adult hypothalamus, this first stage is represented by spinally projecting MCH neurons (Brischoux et al., 2001; Croizier et al., 2010, 2011) that are located very laterally in the rat LHA (**Figure 3E**). Neurons in the lateral region of the adult VMH, where the first SF1-labeled cells settle, send abundant projections through the supraoptic commissures (Canteras et al., 1994), and this pattern is also clearly reminiscent of descending SF-1 projections in the tpoc.

– The second stage is characterized by the growth of ascending projections along the tpoc and sot. This growth is particularly well illustrated by the differentiation of ascending projecting MCH expressing neurons (**Figure 3A**). During this stage, large bundles of ascending axons containing neurotransmitters, such as serotonin and dopamine, develop from the brainstem. The projections from hindbrain serotonergic or ventral midbrain dopaminergic neurons are initially longitudinal as they follow the tpoc, changing course in the basal hypothalamus and becoming transversally oriented in order to migrate towards the telencephalon (**Figures 3B–D**). MCH expressing neurons settle in the hypothalamic region, where these axons change direction. Moreover, in the rat, the phenotype of MCH expressing neurons change drastically as the mesotelencephalic dopaminergic pathway develops. As mentioned above, the first MCH expressing neurons send descending axons to the spinal cord. However, MCH expressing neurons produced during the second stage, as the dopaminergic mesotelencephalic axons progress in the mfb, project axons toward the telencephalon but not the spinal cord in the adult rat (Brischoux et al., 2002; Croizier et al., 2010, 2011; **Figure 3E**).

The mechanisms responsible for the change in the axial organization of the MCH population appear to be related to the differentiation of the telencephalic vesicles (Croizier et al., 2011). The growth and differentiation of telencephalic vesicles involves complex interactions between morphogenic molecular actors, such as Fgf8 and Wnts (Rubenstein et al., 1998; Aboitiz, 2011). These proteins are produced by organizing centers and may diffuse and act far from their production sites. Croizier also detected a sharp increase in the production of the chemoattractant Netrin1 in the telencephalon following the onset of neurogenesis (Croizier et al., 2011). Therefore, the telencephalon exerts a strong influence on the developing rostral brainstem as it differentiates. This influence likely increases as the rostral brainstem becomes involved in very active neurogenesis (**Figure 4A**). Although not fully understood, the telencephalic organizing centers, such as the ventral and cortical hems, and the diencephalic organizing centers, such as the zli, interact through the production of morphogenic protein gradients (Marín et al., 2002; Pottin et al., 2011; Rash and Grove, 2011). These processes are important for the coordinated growth of the cortex and thalamus and for the establishment

the telencephalon, along DA and 5HT axons (see Croizier et al., 2011). **(B–D)** Distribution of serotonin, MCH (MCH-GFP, revealed with an anti-GFP antibody; see Croizier et al., 2011) and tyrosine hydroxylase (dopamine) in three adjacent sections cut in the parasagittal plane and passing through the mfb of an E14 mouse embryonic brain. Serotonergic and dopaminergic axons from respectively the hindbrain and midbrain travel along the tpoc and arch rostrally at the level of the posterior hypothalamus, where MCH expressing

of corticothalamic and thalamocortical connections. However, the second stage also corresponds to an outburst of neurogenesis throughout the hypothalamus and the ventral mesencephalon. The dopaminergic neurons of the substantia nigra and ventral tegmental area are representative of this second stage. Their soma migrate through pre-existing premotor or motor midbrain structures, such as the Edinger-Westfall or oculomotor nuclei, to settle in the ventral midbrain. Their ascending projections through the lateral mfb to the striatum define them anatomically. The outburst of neurogenesis in

projections in the adult rat central nervous system. Abbreviations: DA: dopaminergic neurons; Hyp: hypothalamus; MCH: melano-concentrating hormone containing neurons; mfb: medial forebrain bundle; och: optic chiasm (or presumptive position of the optic tract in **A**); SN: substantia nigra; sot: supraoptic tract; Str: presumptive striatal region; THv: ventral thalamic region; tpoc: tractus postopticus; VTA: ventral tegmental area (presumptive); 5HT: serotonergic neurons.

the hypothalamus includes cortically projecting MCH expressing neurons and most of the hypothalamic medial zone that is generated between E13 and E15 in the rat (Altman and Bayer, 1986). During the same stage, Cheung reported ascending SF-1 expressing axons from the VMH in the mfb (Cheung et al., 2013).

Therefore, the second stage in the development of the hypothalamus is characterized by the differentiation of most of the hypothalamic lateral and medial cell groups and is concomitant with the differentiation of the dopaminergic ventral midbrain.


### **HYPOTHALAMIC TRACT TOPOGRAPHIC ORGANIZATION**

The anatomical dispositions of all the major tracts crossing or bounding the hypothalamus of the adult animal are summarized in **Figure 4B** and supplementary information. From this, it appears that all of the major tracts that originate in the dorsal or ventral telencephalon or in the retina converge in the anterior hypothalamic region (alar hypothalamus). However, upon reaching the caudal hypothalamus, they all follow adjacent pathways. This topographical organization closely follows the initial scaffolding provided by the tpoc in the caudal hypothalamus, as well as that provided by the sot for descending tracts from the telencephalon. The more ventral of these tracts (fornix, stria terminalis, ventral lateral hypothalamic tract, neuroendocrine tracts) end in the neurohypophysis and caudal hypothalamus, while the others (mfb, cerebral peduncle) take divergent routes at the mesencephalic limit. The relative path of these tracts can be illustrated on a schematic sagittal view of the embryonic brain (**Figure 4C**). We therefore observed that the descending tracts from the telencephalon traveled along a transverse path related to the sot, but as they joined the optic tract at the level of the posterior hypothalamus, their course became longitudinal, coincident with that of the tpoc and optic tract. Only the stria medullaris escaped from this general scheme.

### **STRUCTURAL ORGANIZATION OF THE HYPOTHALAMUS AND THE PROSENCEPHALIC FUNCTIONAL PLAN**

The development of the hypothalamus and the adjacent "prethalamus-ventral thalamus" involves several stages. The first stage includes the differentiation of the cell cord. The inverted Y-shaped arrangement is the first differentiated structure of the prosencephalon and guides the first pioneer tracts, including the tpoc and sot. Following the differentiation of the cell cord, neurogenesis becomes generalized in the prosencephalon, preceding the formation of all of the major fiber tracts that connect the telencephalon with the hypothalamus and mesencephalon. Therefore, as new neurons settle medially or dorsally to the cell cord, fiber tracts topographically organize dorsally or medially to the early mfb. The tpoc also guides the optic tract and the supraoptic commissures. Therefore, while the supraoptic arm of the initial inverted Y-shaped pattern appears to guide the olfactory projections in the hypothalamus, the postoptic guides tracts parallel to retinal projections (**Figure 5A**).

These first and second stages in the forebrain differentiation leave traces that are found in the adult anatomical organization of the hypothalamus. In the adult brain, neurons that were initially derived from the cell cord form a large part of the reticularly organized hypothalamus, including the LHA. This structure can be considered, at a functional level, a rostral extension of the primary medial mantle layer that originates in the hindbrain and from which serotonergic neurons are produced. The concept of a deep structure in the brain with a reticular like appearance that is involved in general arousal in all vertebrates and that forms a reticular core is quite ancient. For example, this reticular core was termed the isodentritic core by Ramon-Moliner and Nauta and was also compared to the deep ancient brain described by McLean (Ramón-Moliner and Nauta, 1966; Nieuwenhuys et al., 1998; Swanson, 2003). Pfaff recently argued that primitive mechanisms involving the reticular formation of all vertebrates are important for initiating the activation of behaviors (Pfaff et al., 2012). The hypothalamic cell cord along with a more caudal cell cord are reminiscent of such primitive structures. In the adult brain, structures along the cell cord serve synchronizing and patterning functions; MCH and the cognate hypocretin expressing neurons play roles in the sleep/wake cycle. MCH knockout mice have modified locomotor activities, and in humans, the absence of hypocretin in the dorsal hypothalamus is associated with narcolepsy (Peyron et al., 2000; Verret et al., 2003). This cell cord could also have important functions during all stages of brain development. Serotonergic neurons, which are among the very first generated cells in the hindbrain, act as pacemakers to synchronize the electrical activity of local motoneurons (Moruzzi et al., 2009). Later, the sot and tpoc are the precocious frames for the mfb. The early mfb contains dopamine and serotonin projections, and both neurotransmitters play key roles in the development of telencephalic structures. Dopamine modulates the cell cycle and proliferation in the ganglionic eminences and influences the maturation of the local circuitry (Diaz et al., 1997; Goffin et al., 2010). Serotonin has well recognized developmental effects. Alterations of early serotonergic or dopaminergic pathways lead to pathological conditions, such as autism or schizophrenia (Herlenius and Lagercrantz, 2001; Kinast et al., 2013). MCH is also suspected to have trophic actions (Cotta-Grand et al., 2009).

However, the second stage of forebrain development is more specifically associated with the differentiation of the hypothalamic medial regions and structures belonging to the

**FIGURE 4 | (A)** Schematic comparison of neurogenesis in the hypothalamus and telencephalon in the rat (see text for details). Neurogenesis in the hypothalamus is described as involving three stages: an early stage that produces only the lateral zone; a second that is concomitant to neurogenesis in the telencephalon and produces neurons in all hypothalamic longitudinal zones, but mostly the medial; a late third stage that concerns mainly periventricular zone neurons. Note that MCH neurons are produced during all three stages. **(B)** Illustration of the primary fiber tracts in the hypothalamus; these tracts originate in the telencephalon or retina and converge at the diencephalon-telencephalon limit at preoptic level (top drawing). In the postoptic hypothalamus (bottom drawing), these tracts are all aligned according to an axis determined by the dashed line. See text and supplementary information for details. **(C)** Schematic

representation of the primary descending tracts in the ventral prosencephalon on a sagittal view of the embryonic brain. The tracts are topographically organized. Descending pathways from the telencephalon end in more posterior regions as they are distributed more dorsally in the hypothalamus. The dashed line recalls the ventro-medial/dorso-lateral axis, as in **(B)**. Abbreviations: cpd: cerebral peduncle; fx: fornix; GnRH: gonadotropin-releasing hormone; MCH: melano-concentrating hormone; mes: mesencephalon; mfb: medial forebrain bundle; MM: mammillary body; NG: nucleus Gemini; NH: neurohypophysis; periV: periventricular; PMv: ventral premmamillary nucleus; SN: substantia nigra; st: stria terminalis; tel: telencephalon; tpoc: tractus postopticus; vlt: ventrolateral hypothalamic tract; VTA: ventral tegmental area; zli: zona limitans intrathalamica.

classical striato-nigral pathways. These structures share genetic, topographic and chronotopic characteristics during development. This developmental stage coincides with the concept of a convergence in the anatomical organization of circuits involving these regions in the adult. It therefore becomes very attractive to describe the circuits connecting the pallium, striatum, pallidum and rostral brainstem as a series of parallel interacting loops (**Figures 5B,C**). Alexander described several putative circuits

involving dorsal and ventral striatal components (Alexander et al., 1990). The Swanson group described several other circuits, involving medial and posterior striatal/pallidal structures and hypothalamic medial zone nuclei, suggesting the existence of a basic organizational plan (see Section Introduction; Swanson, 2003; Thompson and Swanson, 2010). We believe that we can now hypothesize that these sets of circuits are developmentally linked and can identify this as to be basic mammalian forebrain functional plan. Each of these circuits shows specific cytoarchitectonic characteristics and are either reticularly or nuclearly organized, likely under the control of the specific expression and localization of adherence molecules (CAM, cadherins) along the corresponding pathways (for example, for the nuclearly organized amygdala and medial hypothalamic nuclei connected through the stria terminalis). Some may even be characterized by the expression of specific transcription factors (again, as illustrated for the amygdala and hypothalamus, concerning reproductive and defensive pathways—Choi et al., 2005).

"Classic" authors have already suggested that well differentiated structures of highly organized brains must have emerged during evolution from primordial reticularly organized forms. Ramon-Moliner and Nauta used of the term "phylogenetic segregation" to characterize these evolutionary processes (Ramón-Moliner and Nauta, 1966). Pre- and postoptic hypothalamic structures have been observed in amphibians, as has a dopamine rich posterior tuberculum; however, laterally organized structures cannot be found in these species. Obviously, lateral and medial hypothalamic structures are phylogenetically recent but made of neurons that derivate from phylogenetically ancient populations (see in Croizier et al., 2013 for MCH and the dorsal hypothalamus), as are the substantia nigra/ventral tegmental area in the ventral midbrain. It is functionally relevant that these structures, including the mammillary nuclei (which contain head direction cells and are part of the Papez circuit), evolved in parallel. A larger behavioral repertory in mammals, especially related to reproductive and agonistic behaviors, is related to the differentiation of the hypothalamic medial and lateral zones, but is also likely associated with increased voluntary motor controls allowed by the extrapyramidal pathways.

To conclude, it has become obvious that the classical hypothalamus with its four regions does not constitute one single neurological entity, at least from the developmental point of view. The divergence in the origins of the collection of nuclei and areas that are usually gathered between its arbitrary borders can be traced to the earliest patterns of gene expression. However, here we contend that early neurogenesis gives rise to a first mantle layer, with longitudinal and transverse components (i.e., Y-shaped) that serves as a foundation for the formation of the whole forebrain connectivity, guiding most ascending and descending tracts appearing later. These observations demonstrate that the structures of the hypothalamic region are intimately implicated within complex networks along the extrapyramidal pathway and act together for the expression of behaviors. They also suggest that these circuits that involve the telencephalon and hypothalamus share a basic organizational plan.

#### **ACKNOWLEDGMENTS**

The authors are particularly grateful to Dr Gonzalo Alvarez-Bolado (Anatomisches Institut, Universität Heidelberg, Germany) and Philippe Ciofi (Institut Magendie, Bordeaux, France) for helpful discussions and careful reading of the manuscript.

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fnana. 2014.00161/abstract

#### **REFERENCES**


regulation of paradoxical sleep. *BMC Neurosci.* 4:19. doi: 10.1186/1471-2202- 4-19


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 30 September 2014; accepted: 09 December 2014; published online: 06 January 2015*.

*Citation: Croizier S, Chometton S, Fellmann D and Risold P-Y (2015) Characterization of a mammalian prosencephalic functional plan. Front. Neuroanat. 8:161. doi: 10.3389/fnana.2014.00161*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Croizier, Chometton, Fellmann and Risold. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## DiI tracing of the hypothalamic projection systems during perinatal development

### **Irina G. Makarenko\***

Laboratory of Cellular and Molecular Basis of Histogenesis, Koltzov Institute of Developmental Biology, Russian Academy of Sciences, Moscow, Russia

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Loreta Medina, Universidad de Lleida, Spain Nora Szabo, Institut de Recherches Cliniques de Montréal, Canada

#### **\*Correspondence:**

Irina G. Makarenko, Laboratory of Cellular and Molecular Basis of Histogenesis, Koltzov Institute of Developmental Biology, Russian Academy of Sciences, Vavilov Street 26, Moscow 119334, Russia e-mail: imakarenk@mail.ru; irina-makarenko@mail.ru

The hypothalamus is the higher neuroendocrine center of the brain and therefore possesses numerous intrinsic axonal connections and is connected by afferent and efferent fiber systems with other brain structures. These projection systems have been described in detail in the adult but data on their early development is sparse. Here I review studies of the time schedule and features of the development of the major hypothalamic axonal systems. In general, anterograde tracing experiments have been used to analyze short distance projections from the arcuate and anteroventral periventricular nuclei (Pe), while hypothalamic projections to the posterior and intermediate pituitary lobes (IL) and median eminence, mammillary body tracts and reciprocal septohypothalamic connections have been described with retrograde tracing. The available data demonstrate that hypothalamic connections develop with a high degree of spatial and temporal specificity, innervating each target with a unique developmental schedule which in many cases can be correlated with the functional maturity of the projection system.

**Keywords: DiI, hypothalamus, intrahypothalamic, septal, mammillary, mammillothalamic, mammillotegmental, rat**

#### **INTRODUCTION**

The hypothalamus is an important structure of the brain positioned as a higher part of the vegetative nervous system and a part of the limbic system. It is involved in realization of numerous neuroendocrine, endocrine, somatomotor and behavioral functions which help an organism to survive and adopt to the environment (Swanson, 2000; Simerly, 2004). The background of such complex functioning lies in multiple intrinsic and external axonal connections. The adult hypothalamic projection systems have been examined in modern times by means of axonal tracing techniques based on stereotaxic injections (Vercelli et al., 2000; Makarenko, 2008) which cannot be applied to the study of embryos or early postnatal animals. For this reason, the early formation of hypothalamic connections is unsufficiently known. However, the beginning of axogenesis and tract formation are important events, as they begin the complex task of forming

**Abbreviations:** AL, anterior pituitary lobe; ARC, arcuate hypothalamic nucleus; AT, anterior thalamic nuclei; AVPV, anteroventral periventricular hypothalamic nucleus; BSTp, principal nucleus of the bed nuclei of the stria terminalis; C, circular accessory hypothalamic nucleus; DL, dorsolateral accessory hypothalamic nucleus; DM, dorsomedial hypothalamic nucleus; IL, intermediate pituitary lobe; LHA, lateral hypothalamic area; LS, lateral septal nucleus; ME, median eminence; MFB, medial forebrain bundle accessory hypothalamic nucleus; MBO, mammillary body (bodies); MPN, medial preoptic nucleus; MTeg, mammillotegmental tract; MTh, mammillothalamic tract; Ne, neuroepithelium; OC, optical commissure; OVLT, organum vasculosum laminae terminalis; PA, preoptic area; Pe, periventricular hypothalamic nucleus; PL, posterior pituitary lobe; PV, paraventricular hypothalamic nucleus; RC, retrochiasmatic hypothalamic nucleus; S, septum; SC, suprahiasmatic hypothalamic nucleus; SO, supraoptic hypothalamic nucleus; Teg, tegmental midbrain nuclei.

the network on which neural function depends (Easter et al., 1993). The formation of axonal connections and the time schedule of these processes in different neural systems have been considered to be the least well-known events in brain development (Keshavan and Murray, 1997) but such knowledge may be important to investigate potential therapeutic interventions, for instance in pathological obesity (Grayson et al., 2006). As the hypothalamus is an integrator of homeostatic processes required for survival, disruption of its development may cause pathological conditions in the adult (Caqueret et al., 2005). Here I will review the current knowledge concerning the time schedule and specific features of formation of several hypothalamic axonal systems using the most commonly used available method, DiI tracing.

#### **METHODOLOGICAL CONSIDERATIONS**

Carbocianine dye tracing was introduced for studies of the brain connections since first works of Honig and Hume (1986, 1989) and Godement et al. (1987). Although there are several carbocyanine dyes (whose advantages and disadvantages I have compared in a previous review (Makarenko, 2008)) one of them has been used predominantly for tracing developing hypothalamic connections: 1,1<sup>0</sup> -dioctadecyl-3,3,3<sup>0</sup> 3,3<sup>0</sup> -tetramethyl-indocarbocyanine perchlorate (DiI) (Molecular Probes, Eugene, OR). It diffuses along the lipid layer of the axonal membrane both in the anterograde and retrograde directions. The general steps of DiI tracing have been described before (see for instance (Molnar et al., 2006; Makarenko, 2008)). Anterograde DiI tracing can only be used in hypothalamic regions which do not have reciprocal connections as is the case for the arcuate nucleus projections to the paraventricular hypothalamic nucleus (Bouret et al., 2004). The existence of reciprocal connections makes the interpretation of the results difficult as it is not possible to find out if fibers were labeled anterogradely or retrogradely. Unwanted retrograde labeling of axons ending on our nucleus of interest can be assumed when neuronal cell bodies are found labeled in a different nucleus. Sometimes, authors deal with this fact by indicating that retrogradely labeled neurons have been omitted for clarity (Hutton et al., 1998) or are not numerous (Polston and Simerly, 2006). To make things worse, neuronal cell bodies are only poorly labeled at 37◦C (the temperature at which DiIlabeled brains are stored according to many current protocols), making retrogradely labeled axons difficult to recognize. In my experience however, storage at room temperature usually labels numerous cell bodies with full dendritic trees clearly recognizable (Makarenko, 2008). For these reasons, the only sure method to evaluate anterogradely labeled fibers ending on a specific nucleus is to visualize labeled terminal arborizations and synaptic boutons.

The small dimensions of the hypothalamic nuclei at early developmental stages require very precise application of the marker, sometimes even single crystals of appropriate size were used (Hutton et al., 1998; Bouret et al., 2004). The need for long storage of the brain with DiI application at room temperature is one of the restrictions of this kind of tracing axonal connections but it provides excellent results. We have used storage times of at least 3–4 months at room temperature for the fetal rat brains and up to 6 months for rat postnatal material. The retrograde DiI labeling was very well preserved even after 1–2 years storage in 4% paraformaldehyde (Makarenko, 2007). Storage of the brain with DiI insertion at +37–40◦C accelerates labeling (Lukas et al., 1998) and can give excellent results (Bouret et al., 2004; Polston and Simerly, 2006). In our hands however, these temperatures have failed to give satisfactory results. Sometimes it takes 12 or 26 weeks even with the heating (Bader et al., 2012).

Sectioning and coverslipping are also important. Thick vibratome sections (80–100 µm) sections are the most appropriate for the analysis of DiI tracing results using conventional or confocal fluorescence microscopy. Frozen sections can not be recommended for the DiI tracing because of the fast lateral diffusion of the marker from the axon in any type of the mounting medium. This can be overcome only by mounting and drying cryocut sections without coverslipping (Makarenko et al., 2001). As for the medium for mounting the sections, it is common to use buffered glycerol (Magoul et al., 1994; Hutton et al., 1998; Bouret et al., 2004), but others consider Mowiol (Calbiochem, Germany) more convenient (Makarenko, 2007; Alpeeva and Makarenko, 2009; Bader et al., 2012). Sections coverslipped using Mowiol can be stored for a long time without bleaching of the label and drying of the medium (Makarenko, 2008).

Additional control DiI insertions in the regions adjacent to the studied source of projections are necessary and useful because they usually resulted in labeling patterns that are distinct from those in the main experiments. They help in the analysis of specificity and interpretation of the results (Bouret et al., 2004; Alpeeva and Makarenko, 2007). The importance of methodological details and careful analysis in DiI tracing studies can be demonstrated on one example. Kouki and Yamanouchi (2007) set out to analyze the postnatal development of the lateral septal projections to the midbrain, and found septo-hypothalamic connections developing on 8 week-old rats. However, several poor methodological decisions make it difficult to evaluate their results. By using a DiI solution in dimethylformamide on the lateral septum, they labeled a wide region including not only the lateral septal nucleus but also adjacent regions; they incubated the tissue under excessive temperature (+40◦C), and their analysis was performed on sections that were too thick (200 µm) and photographed at too low magnification with conventional fluorescence microscopy. As a result, they described only few fibers from the lateral septum in the preoptic region at birth and did not find labeled neurons there. Conversely, our own data obtained after similar experiments demonstrates that the first neurons of the lateral septal nucleus send axons to the preoptic area of the hypothalamus as early as E14–E15 and on E18–E20 these neurons have prominent ramified dendrites (Makarenko, 2007). According our preliminary results reversed hypothalamic projections to lateral septum also developed prenatally at least from E18 on (Makarenko, 2008). Perhaps the poor quality of the labeling in the mentioned work was associated with age of rats. Based on our work on the innervation of the anterior thalamic nuclei (Alpeeva and Makarenko, 2009), we support the view that DiI tracing works better on the prenatal and early postnatal material than in adults (Balthazart et al., 1994; Vercelli et al., 2000). The quality of the preterminal labeling in the thalamus decreased significantly after postnatal day (P)20 and on P60 distinct terminal arborizations could not be distinguished at all.

### **ANTEROGRADE TRACING OF DEVELOPING HYPOTHALAMIC CONNECTIONS**

The first DiI tracing studies of the hypothalamic connections were performed on adult rats. It was shown that projections from the arcuate nucleus to the bed nucleus of the stria terminalis follow through the stria terminalis and this route coincides with the orientation of ACTH-immunoreactive axons (Magoul et al., 1994). DiI was used also in studies analyzing the integration in the host brain of fetal septal-preoptic grafts (Silverman et al., 1992). Retrograde DiI labeling has also been used in combination with immunocytochemical visualization of gonadotropin-releasing hormone (GnRH) in the hypothalamic neurons projecting to the mediobasal hypothalamus in postnatal Djungarian hamsters (Buchanan and Yellon, 1993). The authors noted the difficulties of such combination because the saponin treatment necessary for their immunocytochemical staining diminished DiI labeling. They described a decreasing number of GnRH neurons projecting to the median eminence in the process of sexual maturation of Djungarian hamster. Such a result may be due to the fact that visualization of DiI-labeled neurons in the adult brain is worse than in the early stages of development (Balthazart et al., 1994; Vercelli et al., 2000; Makarenko, 2008). A special modification of the method has been proposed for adult human material (Sparks et al., 2000).

#### **SUPRACHIASMATIC NUCLEUS**

Anterograde DiI tracing of the suprachiasmatic nucleus connections in the hamster has revealed very rapid and simultaneous development of the efferents to the principal targets known to be innervated by the adult suprachiasmatic nucleus (medial preoptic nucleus, the paraventricular nucleus (PV) of the thalamus, the parvicellular division of the hypothalamic PV, the subparaventricular zone, the dorsomedial nucleus of the hypothalamus, the ventral lateral septum, the intergeniculate leaflet, the bed nucleus of the stria terminalis) on the first postnatal day (P1) with significant increase on P2–P3 (Müller and Torrealba, 1998). These projections begin to form before the ontogenetic neuron death on P3. Retinal afferents reach the suprachiasmatic nucleus later only on P6–P7.

#### **BED NUCLEUS OF THE STRIA TERMINALIS**

Projections from the bed nucleus of the stria terminalis to the anteroventral periventricular nucleus of the preoptic region are established between postnatal day 9 (P9) and P10 in male rats but not in females although projections from the bed nucleus of stria terminalis (BSTp) to the preoptic region were revealed earlier on P4–P7 both in males and females (Hutton et al., 1998).

#### **ARCUATE NUCLEUS OF THE HYPOTHALAMUS**

Special attention was paid to the hypothalamic circuitry that modulates feeding and energy expenditure as the disturbances in their normal formation during the critical periods of the perinatal development may have long-term consequences on feeding behavior and body weight management (Grove et al., 2005). As the arcuate nucleus is one of the most important hypothalamic nuclei involved in regulation of body weight and responding to leptin signals in adults but not during two postnatal weeks, specific attention was paid for its connections with other hypothalamic regions. The study of their development has been performed on postnatal mice using anterograde DiI tracing (Bouret et al., 2004) because early postnatal administration of leptin did not cause inhibition of food intake (Bouret, 2010). The authors found the anatomical basis for this fact: arcuate nucleus projections to the dorsomedial and paraventricular hypothalamic nuclei and lateral hypothalamic area are immature at birth and develop from P6, P8 and P10 on, reaching an adult-like pattern at P16–P18. Bouret et al. (2004) attempted to quantify the innervation of these hypothalamic nuclei using confocal images. Such experiments are difficult, since it is not easy to standardize marker application on the brains at different developmental stages. An additional problem is that DiI labels all terminal arborizations of the axons but not only nerve terminals. Regardless of this their results provide a very good demonstration of increasing fiber density in the target nuclei. Additionally they found that arcuate nucleus axons reach the lateral hypothalamic area on P10, the medial preoptic nucleus on P12 and later they reach more rostral extrahypothalamic regions (the BSTp, paraventricular nucleus of the thalamus, or ventral part of the lateral septum). Thus all arcuate nucleus projections are formed primarily during the second postnatal week and appeared to be fully developed by P18. These results explain why there were no labeled neurons in the arcuate nucleus following DiI insertion in the lateral septal nucleus of the rat fetuses (Makarenko, 2007).

#### **ANTEROVENTRAL PERIVENTRICULAR NUCLEUS OF THE HYPOTHALAMUS**

Later a new methodological approach for quantifying of the anterograde DiI tracing was used to study the projections from the anteroventral Pe to the GnRH neurons close to the organum vasculosum laminae terminalis (OVLT; Polston and Simerly, 2006). They used CM-DiI (Molecular Probes) to shorten the time of the brain storage in the fixative and for combination of the DiI tracing with GnRH immunocytochemistry. The authors carefully analyzed synaptic oppositions labeled with CM-DiI on the dendrites of GnRH neurons and developed a detailed method for their quantification on the volume reconstructions of the confocal images. This hard work did not reveal representative differences between E19, P0 and P60 brains. Rostral projections of anteroventral Pe to the BSTp and lateral septal nucleus develop later during second and third postnatal weeks. Caudal projections of anteroventral Pe to the parvicellular part of the paraventriculer hypothalamic nucleus were revealed also postnatally. Analysis of these data led authors to the idea that structures containing neurons that express neuroendocrine releasing factors appeared to be innervated first.

### **RETROGRADE TRACING OF HYPOTHALAMIC PROJECTIONS TO THE PITUITARY AND MEDIAN EMINENCE**

The main purpose of our own serial studies was to perform a systematic evaluation of the first steps of the development of main hypothalamic fiber tracts during perinatal ontogenesis in rats using carbocyanine dye tracing. The first studied system was the hypothalamo-pituitary tract that consists of several projection systems well described in the adult. The magnocellular hypothalamic nuclei produce hypothalamic peptidergic hormones (vasopressin and oxytocin) transported by their axons to the posterior pituitary (PL; Hoffman et al., 1986; Hatton, 1990; Falke, 1991). The neurons of the parvicellular hypothalamic nuclei produce adenohypophysiotropic neurohormones, peptides and dopamine and send axons to the median eminence (ME) providing the pathway for the neurohormone transfer via hypophysial portal circulation to the adenohypophysial target cells (anterior lobe, AL; Zimmerman et al., 1984; Swanson, 1986). Some of the parvicellular neurons innervate also the intermediate pituitary lobe (IL).

Because DiI was used in most of experiments as a retrograde tracer, it was inserted in the terminal regions of the projections of interest (PL, ME or IL) (Makarenko, 2008). DiI was usually distributed strictly in the posterior lobe (**Figure 1A**) in the cases were the boundary between PL and IL was not damaged (Makarenko et al., 2000). Larger incisions in the PL that penetrated this boundary let the marker spread into both PL and IL (**Figure 1B**). Separate IL labeling was achieved by pasting DiI crystals on the IL after dissection of the anterior pituitary lobe in postnatal rats (Makarenko et al., 2005; Makarenko, 2008).

#### **PROJECTIONS TO THE POSTERIOR PITUITARY LOBE (PL)**

DiI insertion into the PL (**Figure 1A**) revealed labeled neurons specifically in the regions of presumptive and later differentiated

**median eminence**. DiI distribution in the pituitary after insertion into the posterior pituitary lobe (PL) without damage of the boundary between the PL and intermediate (IL) pituitary lobes **(A)**. DiI insertion with labeling in the PL and part (arrow) of IL **(B)**. DiI insertion in the median eminence (ME) with few labeled neurons in the arcuate nucleus **(C)**. AL – anterior pituitary lobe. Scale bar = 100 µm.

magnocellular peptidergic nuclei on all developmental stages analyzed (Makarenko et al., 2000): supraoptic nucleus (SO) and PV. The most intriguing fact was that on E15–E16 neurons sending axons to the PL were revealed in the anterior hypothalamus from the wall of diamond like swelling of the third ventricle and occupied all space ventrolaterally down to the location of the future SO which looks very large (**Figure 2A**). This location of the DiI-labeled cells resembles the distribution of magnocellular vasopressin- and oxytocinergic neurons originating from the neuroepithelium of the diamond-like swelling of the third ventricle, in rats at E12–E14 and migrating to the places of their final destination (Altman and Bayer, 1986; Buijs, 1992; Bayer and Altman, 1995). Later on E17–E18, the major DiIlabeled neurons were gathered along the lateral edges of the optic chiasm and tracts, i.e., in the SO (**Figure 2B**), whereas only a few fluorescent neurons still were seen between the PV. Thus it can be assumed that the labeled cells described earlier between the third ventricle and SO represent migrating magnocellular neurons which axons already reached the posterior lobe of pituitary. Previous immunocytochemical studies failed to visualize vasopressin- or oxytocin- immunopositive axons in the ME and the PL until E17 and E18, respectively (Whitnall et al., 1985; Buijs, 1992).

In contrast to the SO, only occasional labeled neurons were seen in the primordium of the PV at E16 and E17 suggesting that the axons of differentiating neurons of the proper PV reach the posterior lobe significantly later than axons from the SO. These data appear to be consistent with earlier observations of the slower settling and differentiation of magnocellular neurons of the PV when compared to those of the SO (Laurent et al., 1989). On E19 and onwards, the PV contains distinct groups of labeled neurons located in the magnocellular part of the nucleus (**Figure 2C**). A moderate number of the labeled neurons projecting to PL from the magnocellular part of PV in young rats is in agreement with earlier observations of a rather limited number of axonal projections from the PV to the PL even in adults (Arai et al., 1990).

Magnocellular neurons projecting axons to the PL have been described in adult rats additionally in the accessory peptidergic nuclei (Sofroniew et al., 1980; Ju et al., 1986) and other mammals (Grinevich and Polenov, 1994) were revealed using DiI tracing on the different stages of perinatal development (Makarenko et al., 2002). The retrochiasmatic nucleus that is functionally close to the SO and contains significant number of vasopressinergic neurons sends axons to the PL first on E16–17 and continues development postnatally (Makarenko et al., 2000, 2002). Labeled neurons in the other accessory nuclei such as dorsolateral nucleus (**Figure 2D**), nucleus of the medial forebrain bundle, circular and commissural nuclei were visualized only postnatally (P2–P10) (Makarenko et al., 2002). Their role is rather questionable and the number of the neurons revealed during perinatal period was significantly lower than in adult rats (Grinevich and Polenov, 1994).

#### **HYPOTHALAMIC PROJECTIONS TO THE INTERMEDIATE PITUITARY (IL)**

The IL receives dopaminergic innervation from the periventricular (Pe) and arcuate hypothalamic nuclei in adult mammals (Kawano and Daikoku, 1987; Goudreau et al., 1992; Vanhatalo et al., 1995). As it was very difficult to perform separate DiI labeling of the IL we analized cases with insertions of the marker in the deep incisions through the PL that reached the IL (**Figure 1B**). As a result labeling in the hypothalamus resembled that described following the DiI insertions in the PL (Makarenko et al., 2000) but with specific visualization of the parvicellular neurons (Makarenko et al., 2005). The first axons arrive to the IL from the Pe in the late prenatal period in coincidence with the first appearance of TH-immunoreactive neurons in the Pe (Jaeger, 1986; Ugrumov et al., 1989). The number of labeled cell bodies in the Pe increased progressively in postnatal rats, in coincidence with the first appearance of the tyrosine hydroxylaseimmunoreactive fibers in the IL (Gary and Chronwall, 1992).

Single DiI labeling of the IL in P2–P3 rats revealed neurons in the parvicellular but not magnocellular part of the Pe (Makarenko et al., 2005). Besides, the periventricular and parvicellular PV the labeled neurons were regularly seen in the arcuate nucleus of the neonatal rats following the DiI staining of both IL and PL or only IL. These data shows that IL receives afferents from the

following DiI insertions into the posterior **(A,B,C,D)** and intermediate pituitary lobes on different prenatal stages. **(A)** E15, neurons in the supraoptic nucleus in the periventricular hypothalamic nucleus after DiI insertion in the intermediate pituitary lobe. Scale bars = 100 µm.

parvicellular hypothalamic neurons in rats shortly after the birth. The number of the hypothalamic DiI-labeled neurons and fibers increased progressively from P10 (**Figure 2E**) to P20. Additional DiI labeled neurons projecting to the IL were revealed in young rats along the dorsolateral border of the ventromedial nucleus. Thus, the hypothalamic parvicellular hypothalamic neurons form innervation of the IL mainly during first three weeks of postnatal life.

#### **HYPOTHALAMIC PROJECTIONS TO THE MEDIAN EMINENCE (ME)**

Most hypothalamic nuclei projecting to the ME are parvicellular neurons, producing adenohypophysotropic neurohormones, peptides and dopamine (see for reference Ugrumov, 1991). They project axons to the ME providing the pathway for the neurohormone transfer via hypophysial portal circulation to the adenohypophysial target cells (Ugrumov, 1992). The axons containing adenohypophysiotropic neurohormones or the enzymes of their synthesis were first detected by immunocytochemistry not earlier than on E16 (Daikoku et al., 1986; Ishikawa et al., 1986; Okamura et al., 1991). However, immunostaining defined the accumulation of neurohormones in the axons or terminals that could be visualized immunocytochemically but not the first arrival of the axons in the ME. DiI inserted into the median eminence rostrally to pituitary stalk (**Figure 1C**) labels both fibers and neurons in the hypothalamus from the first rare cells on E14 with subsequent increase during the next few days (Makarenko et al., 2001). They were widely distributed through the hypothalamus and in the ventromedial region of the more rostral forebrain resemble such distribution after PL insertion on E15 because of the DiI labeling of the axons of the hypothalamopituitary tract. On E20 most labeled neurons were concentrated mainly in distinct hypothalamic nuclei: the PV (dorsal and medial parvicellular parts), the arcuate nucleus and to a lesser extent in the medial preoptic nucleus, the SO, the diagonal band,

**(A)** E15, neuronal bodies on the left side of the septum after ipsilateral DiI insertion. **(B)** E18, asymmetrical distribution of DiI in the preoptic

magnification. **(F)** E.18, DiA-labeled multipolar neurons in the lateral septum. Scale bars =100 µm.

and the retrochiasmatic nucleus. In neonates, DiI-labeled neurons appeared additionally in the accessory dorsolateral nucleus, medial preoptic area lateral to the diagonal band, anterior hypothalamic area, and in the anterior Pe. Thus, the axons of differentiating neurons of parvicellular hypothalamic nuclei invade the ME during prenatal development and continue in neonates.

### **RETROGRADE TRACING OF THE LATERAL SEPTAL (LS) PROJECTIONS TO THE HYPOTHALAMUS DURING PRENATAL DEVELOPMENT**

The septum is an integrative relay station of the limbic system, connecting the limbic forebrain with the hypothalamus and brainstem (Risold and Swanson, 1997; Herman et al., 2002). Its extensive reciprocal connections with the hypothalamus have been visualized and described in adult rats using different methods. Projections from the lateral septal nucleus to the hypothalamus have been described in detail in adult rats (Risold and Swanson, 1997). These connections are topographically organized and belong to specific hypothalamic behavioral systems. The time of origin of the septo-hypothalamic projections has been studied in rats using DiI and DiA as retrograde tracers (Makarenko, 2007). The tracers were applied to different hypothalamic regions: preoptic area, mediobasal hypothalamus and mammillary bodies or adjacent regions of the posterior hypothalamus in rat fetuses on E14.5–E21. The

number of retrogradely-labeled neurons revealed in the septum correlates with the region and size of the DiI application (**Figures 3B–D**). The first septal neurons sending axons to the preoptic area were visualized in the septum on E14.5– E15 but there were no clear differentiation of the septal nuclei and the neuroepithelial layer was very thick (**Figure 3A**). On E18 the number of the neurons in the lateral septal nucleus projecting to the preoptic area increased significantly and they have surprisingly large dendritic tree when visualized by DiI (**Figures 3C–E**) that is characteristic for adult rat (Risold and Swanson, 1997). In contrast another carbocyanine dye 4- (4-dihexadecylaminostyryl)-N-methylpyridinium iodide (DiA) visualized using fluoresceine filter revealed the comparable number of neurons in the lateral septum but the marker was accumulated mainly in neuronal bodies with very light staining of the dendrites of septal neurons and afferent hypothalamic axons (**Figure 3F**). The lateral septum projections to the mediobasal hypothalamus are not so numerous at all studied ages and the number of neurons visualized in the lateral septum was always lower than those projecting to the preoptic area. This was confirmed by a double labeling study, when DiA was inserted into the preoptic region and DiI into the mediobasal hypothalamus on E18–E21 (Makarenko, 2007). These insertions resulted in visualization of few DiI-labeled cells and numerous DiA-labeled neurons in the lateral septal nucleus. Anterogradely labeled fibers were also revealed in the septum on E18 and E20 with concentration in the lateral septal nucleus following DiI insertion in the preoptic area. They can be regarded as afferent hypothalamic fibers innervating the septum. No septal projections to the posterior hypothalamus specifically the mammillary bodies were revealed prenatally, although they were described in adult rats (Risold and Swanson, 1997). According to the data that septal projections to the hypothalamus participate in regulation of several neuroendocrine functions associated with water and salt intake, food intake, thermoregulation, aggressiveness, and sexually related behaviors (for review, see Herman et al., 2002; Sheehan et al., 2004) these results are of interest as providing knowledge that a significant part of septal projections to the preoptic area develop prenatally but other connections are not numerous or even absent before birth.

Additionally our preliminary data on the development of the innervation of the lateral septum by the hypothalamic neurons shows that these axons originate at E18 if not before (Makarenko, 2008) and they are already well differentiated at birth.

### **DEVELOPMENT OF THE TRACTS OF THE MAMMILLARY BODY (MBO)**

In humans, the mammillary bodies are paired brain structures located in the ventral part of the hypothalamus and surrounded by a "capsule" of nervous fibers. They are part of the limbic system and in particular of the classical "Papez circuit" that plays a major role in emotional and motivational activity as well as in memory formation (Mark et al., 1995; Morgane et al., 2005). In rodents, the MBO forms a single, medially located mass of cells sending very specific bilateral efferent projection systems: the short principal mammillary tracts (left and right) divide into two massive efferent fiber bundles, the mamillotegmental tract, directed to the midbrain (Cruce, 1977; Allen and Hopkins, 1990; Hayakawa and Zyo, 1991), and the mamillothalamic tract, innervating the anterior thalamic nuclei (Watanabe and Kawana, 1980; Seki and Zyo, 1984). Afferent projections to the MBO from the ventral and dorsal tegmental nuclei are provided by small fiber system called mammillary peduncle, but not through the mammillotegmental tract (Allen and Hopkins, 1990). In rodents, the MBO is visible on the ventral surface of the hypothalamus and therefore easy to target with DiI crystals. The surrounding capsule prevents diffusion of the marker outside the MBO from E19 on (in the rat). Cases with accidental DiI application in adjacent regions can be used as a control and provide a quite different distribution of labeled structures.

#### **MAMMILLOTEGMENTAL TRACT (MTeg)**

The MTeg is one of the earliest developing brain tracts (Easter et al., 1993; Mastick and Easter, 1996). It consists of MBO efferent axons innervating tegmental nuclei and does not contain reciprocal fibers from the midbrain (Cruce, 1977; Allen and Hopkins, 1990; Hayakawa and Zyo, 1990). We have used anterograde and retrograde DiI tracing to study the MBO-midbrain connections (Alpeeva and Makarenko, 2007). In the rat, at E14 numerous labeled axons of MBO neurons were visible in the principal mammillary tract and in the Mteg, curving along the mesencephalic flexure (**Figure 4A**). In some cases, a few retrogradely labeled neurons could be seen on coronal brain sections as bilateral groups in the caudal midbrain area (**Figure 5A**). The labeled growth cones indicate that these are growing efferents. At E19–E21, in cases when the place of DiI crystal insertion was very precisely in the MBO, a distinct group of retrogradely labeled neurons was visualized on sagittal sections in the midbrain, very likely the primordium of the tegmental nuclei (**Figure 4B**). These neurons have long dendrites and are surrounded by a network of labeled fibers (**Figure 5B**). DiI application on the same tegmental region visualized retrogradely labeled axons in the MTeg and numerous neuronal bodies in the MBO on E16–19 (**Figure 4C**) and later.

#### **MAMMILLARY PEDUNCLE (MP)**

Afferent axons from the tegmental neurons reach the MBO through the MP (Allen and Hopkins, 1989; Hayakawa and Zyo, 1991). DiI insertions in the MBO labeled both efferents in the MTeg at early prenatal stages and afferents from tegmental nuclei that were visualized from E19 on (Alpeeva and Makarenko, 2007;

**FIGURE 5 | Tegmental neurons innervating MBO**. DiI labeled neurons in the tegmentum following DiI insertion into the MBO on different prenatal stages: **(A)** E14, bilateral groups of neurons on a coronal section; **(B)** E19, confocal image of the same neurons, in sagittal section; **(C)** P1, the ventral tegmental nucleus on sagittal section; **(D)** P10, ventral tegmental nuclei on coronal section. Scale bars: **(A,B)** = 80 µm, **(C,D)** = 100 µm.

mammillothalamic tract and innervation of the anterior thalamic nuclei following DiI insertion in the mammillary body. **(A)** E19.5, axonal growth

the MTeg through the formation of the mammillary peduncle around E19 to the final differentiation of the tegmental nuclei

**(D)** P10, terminal network of the mammillary body axons in the anteromedial and anteroventral thalamic nuclei. Scale bars: **(A,C,D)** = 20 µm, **(B)** = 80 µm.

**MAMMILLOTHALAMIC TRACT (MTh)**

postnatally.

The mammillothalamic tract is a well-organized projection system described in different mammals and innervating three anterior thalamic nuclei (anteromedial, anteroventral and anterodorsal). It is formed by collaterals of the MTeg as has been shown in adult rats with double tracer injections (Hayakawa and Zyo, 1989), and we have confirmed it with DiI in the developing rat (Alpeeva and Makarenko, 2007). The mechanisms and regulation of the development of the MTh has also been investigated by genetic methods in the mouse (Alvarez-Bolado et al., 2000; Szabó et al., 2011).

The development of the MTh has been studied in the rat using anterograde DiI tracing (Alpeeva and Makarenko, 2009; Makarenko, 2011). From E15 to P5, DiI insertions in the MBO resulted in its distribution in all mammillary body

**Figure 4D**), although they were not always visible on the same sagittal section. Since retrogradely labeled neurons were observed already at E14–15 among the fibers of Mteg, it is possible that the tegmental afferents develop earlier. Their axons cross the MTeg and formed a bundle (mammillary peduncle) reaching the MBO ventral and lateral to the MTeg. The neurons of the midbrain tegmental area are formed between E13 and E15 in the rat (Altman and Bayer, 1980). At E19, the first fibers from the MTeg turn dorsally and start to form terminal arborizations. Full morphological differentiation of the tegmental nuclei takes place postnatally. From P2 to P10, a single tegmental group divides into two (the dorsal and ventral tegmental nuclei) (**Figures 5C,D**).

It has been proposed that reciprocal connections between MBO and midbrain tegmental nuclei function as a feedback loop (Gonzalo-Ruiz et al., 1999; Kocsis et al., 2001), which would be an important element of the brain limbic system. DiI anterograde and retrograde tracing has shown that these connections undergo a protracted development from the early emergence of


#### **Table 1 | The time schedule of the development of hypothalamic projection systems studied using DiI tracing**.

nuclei and projections of the separate MBO nuclei can not be analyzed. The first collaterals of the MTeg became visible at E17 and a short MTh was observed at E18. Numerous fibers grow simultaneously as a tightly packed bundle through the hypothalamus and ventral regions of the thalamus. Confocal scanning of the rostral end of the MTh revealed typical growth cones on the ends of MBO axons (**Figure 6A**). At E20–21 the MTh grows into the ventromedial region of the anterior thalamic nuclei (**Figures 4D**, **6B**). This fact is in agreement with the evidence that the different nuclear components of the anterior thalamus can be distinguished in the rat at E21 (Coggeshall, 1964). Thus the MBO innervation of the anterior thalamic nuclei develops from late prenatal stages (innervation of the ipsilateral anteromedial nucleus at E20–21). On P1–P2, MBO axons start to innervate the ipsilateral anteroventral thalamic nucleus. At the same time, a separate bundle courses forward in the direction of the anterodorsal thalamic nuclei ipsi- and contralaterally forming the thalamic decussation. At P3–P4, labeled fibers of the MTh filled the whole volume of the anteroventral and began to form the terminal network in the anterodorsal thalamic nuclei. The terminal network consisted of fine beaded nervous processes tightly surrounding all neurons inside the anterior thalamic nuclei (**Figure 6C**). Immunocytochemical visualization of synapsin, a marker of functioning brain synapses (Melloni and DeGennaro, 1994; Castejón et al., 2004), revealed immunolabeling in the anterior thalamic nuclei as small dots in the neuropil around neurons and density of their distribution correlated with density of the DiI labeled terminal network in each nucleus (Alpeeva and Makarenko, 2009). By P5–P6 and P10 the density of fluorescent Mth terminal arborizations in all three nuclei of the anterior thalamus has grown significantly (**Figure 6D**).

The differential contribution of MBO subnuclei to the MTh can be distinguished already on P6–P8 (Alpeeva and Makarenko, 2009). Separate labeling of the medial mammillary nucleus or lateral mammillary nucleus with little spread outside revealed that the axons from the lateral MBO innervate both anterodorsal thalamic nuclei (left and right) and the axons from the medial MBO innervated anteroventral and anteromedial thalamic nuclei on the side of DiI insertion. This is in agreement with the data obtained in adult rat (Watanabe and Kawana, 1980; Seki and Zyo, 1984; Hayakawa and Zyo, 1989; Guison et al., 1995; Gonzalo-Ruiz et al., 1998). The adult innervation pattern of the anterior thalamic complex by the MBO is reached around P5–P10.

### **CONCLUSIONS**

The available data, summarized in (**Table 1**), demonstrate that hypothalamic connections develop with a high degree of spatial and temporal specificity, innervating each target with a unique developmental schedule which in many cases can be correlated with the functional maturity of the projection system.

It can be useful to distinguish three groups of hypothalamic projection systems according to the period of their formation

	- Mammillary body projections to the tegmentum (mammillotegmental tract)
	- Projections of parvicellular hypothalamic neurons (Pe, paraventricular and arcuate nuclei) to the median eminence
	- Projections of magnocellular hypothalamic neurons (SO and PV) to the PL and median eminence
	- Lateral septal projections to the preoptic area

#### 2. *Projections starting late prenatally and differentiating postnatally*


#### 3. *Projections developing postnatally*


### **REFERENCES**


on tracing distance and combination with immunocytochemistry. *J. Histochem. Cytochem.* 46, 901–910. doi: 10.1177/002215549804600805


**Conflict of Interest Statement**: The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 27 September 2014; accepted: 13 November 2014; published online: 04 December 2014*.

*Citation: Makarenko IG (2014) DiI tracing of the hypothalamic projection systems during perinatal development. Front. Neuroanat. 8:144. doi: 10.3389/fnana.2014. 00144*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2014 Makarenko. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Thyroid hormone and the developing hypothalamus

### **Anneke Alkemade \***

Amsterdam Brain and Cognition Center, University of Amsterdam, Amsterdam, Netherlands

#### **Edited by:**

Valery Grinevich, German Cancer Research Center DKFZ and University of Heidelberg, Germany

#### **Reviewed by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany Josef Köhrle, Charité-Universitätsmedizin Berlin, Germany

#### **\*Correspondence:**

Anneke Alkemade, Amsterdam Brain and Cognition Center, University of Amsterdam, Nieuwe Achtergracht 129, 1018WS Amsterdam, Netherlands e-mail: jmalkemade@gmail.com

Thyroid hormone (TH) plays an essential role in normal brain development and function. Both TH excess and insufficiency during development lead to structural brain abnormalities. Proper TH signaling is dependent on active transport of the prohormone thyroxine (T4) across the blood-brain-barrier and into brain cells. In the brain T4 undergoes local deiodination into the more active 3,3<sup>0</sup> ,5-triiodothyronine (T3), which binds to nuclear TH receptors (TRs). TRs are already expressed during the first trimester of pregnancy, even before the fetal thyroid becomes functional. Throughout pregnancy, the fetus is largely dependent on the maternal TH supply. Recent studies in mice have shown that normal hypothalamic development requires intact TH signaling. In addition, the development of the human lateral hypothalamic zone coincides with a strong increase in T3 and TR mRNA concentrations in the brain. During this time the fetal hypothalamus already shows evidence for TH signaling. Expression of components crucial for central TH signaling show a specific developmental timing in the human hypothalamus. A coordinated expression of deiodinases in combination with TH transporters suggests that TH concentrations are regulated to prevent untimely maturation of brain cells. Even though the fetus depends on the maternal TH supply, there is evidence suggesting a role for the fetal hypothalamus in the regulation of TH serum concentrations. A decrease in expression of proteins involved in TH signaling towards the end of pregnancy may indicate a lower fetal TH demand. This may be relevant for the thyrotropin (TSH) surge that is usually observed after birth, and supports a role for the hypothalamus in the regulation of TH concentrations during the fetal period anticipating birth.

**Keywords: hypothalamus, brain development, thyroid hormone, paraventricular nucleus, deiodinase**

#### **INTRODUCTION**

Thyroid hormone (TH) is crucial for normal fetal growth and maturation including the brain (Eayrs and Taylor, 1951; Morreale de Escobar et al., 1983; Legrand, 1984). Developmental TH deficiency impairs growth, and compromises adaptation to life outside the womb (Forhead et al., 1998; Hillman et al., 2012; Sferruzzi-Perri et al., 2013; Forhead and Fowden, 2014). The importance of TH during development is widely recognized and reflected in government interventions, such as iodine supplementation programs, as recommended by the World Health Organization (e.g., universal salt iodisation), and population wide screening for TH deficiency via the heel prick (Ford and LaFranchi, 2014). TH deficiency during development causes irreversible damage, which can largely be prevented by TH replacement therapy (Dubuis et al., 1996; Grüters and Krude, 2011).

During uterine development the fetus is largely dependent on the maternal TH supply, and TH receptors (TRs) are already present before the thyroid gland becomes functional (Morreale de Escobar et al., 2004, 2007). TH is implicated in many developmental processes in the brain including cell cycling, synaptogenesis, migration, plasticity, and myelination (Bernal, 2007). In both humans and rodents, brain development is not completed at, and continues after birth. From the moment of birth, the offspring no longer benefits from maternal TH supply. Insufficient TH availability, also during the postnatal developmental period, leads to severe neurological damage and mental retardation. TH insufficiency can be the result of iodine deficiency, or of defective formation of the thyroid gland (Deladoëy et al., 2007; Nilsson and Fagman, 2013; Szinnai, 2014). Interestingly, many studies have focused on the importance of TH in brain development, but much less is known about the development of the hypothalamic feedback loop of TH. When studying intrauterine development, it is not only the fetal hypothalamus that is in play, it is also that of the mother. In view of the cross-talk between systems it is often not possible to distinguish between the different (neuro- )endocrine systems in play. Rodent studies have been crucial for understanding TH signaling, but the research is complicated by interspecies differences that are observed between rodents, humans and other mammals. In the present review I will focus mainly on TH signaling in the developing human hypothalamus.

#### **TH SIGNALING IN THE HYPOTHALAMUS**

Under normal conditions, TH signaling is controlled via a classic negative feedback pathway at the level of the anterior

pituitary and hypothalamus (**Figure 1A**). Hypothalamic TH signaling is also involved in circadian rhythmicity, feeding and adaptation to environmental challenges (Costa-e-Sousa and Hollenberg, 2012). Interestingly, studies in anencephalic humans have shown that when the hypothalamus is not formed at all, pituitary-thyroid function still develops (Beck-Peccoz et al., 1992). Following normal brain development, hypophysiotropic neurons of the paraventricular nucleus (PVN) of the hypothalamus expressing thyrotropin releasing hormone (TRH) project to the portal system, via which it reaches the thyrotropin (TSH) producing cells of the anterior pituitary. TSH, which is subsequently released, binds to its receptor in the thyroid where TH is produced and released. TH is released predominantly as thyroxine (T4), a prohormone, and to a lesser extent as the active 3,3<sup>0</sup> ,5-triiodothyronine (T3). T4 is converted locally into T3 providing a negative feedback at the level of the pituitary as well as the hypothalamus (**Figure 1A**). T3 mainly acts by regulation of gene expression via binding of nuclear TRs, although nongenomic effects of T3 have been described as well (Cheng et al., 2010; Davis et al., 2013).

A number of specific proteins are required for normal TH signaling in the brain including the hypothalamus. These proteins have been studied mainly in humans and rodents. TH cannot enter or exit the cell via passive diffusion and requires facilitated transport across the cell membrane, as well as the membrane of the cell nucleus (Visser et al., 2008; Heuer and Visser, 2009; Wirth et al., 2014). Proteins capable of transporting TH belong to a number of families including the sodium dependent organic anion tranporters (NTCP), organic anion transporting polypeptides (OATP), heterodimeric amino acid transporters (HAT) including light chains LAT1 and 2, and the monocarboxylate transporters MCT8 and 10 (Friesema et al., 1999, 2001; Hennemann et al., 2001; Roberts et al., 2008). The distribution patterns of these transporters throughout the body show significant overlap, and most of them are able to transport other molecules than iodothyronines as well. MCT8 appears to be specific for transporting iodothyronines (Friesema et al., 2003). OATP1C1 is expressed in endothelial cells of the blood brain barrier, as well as the choroid plexus (Roberts et al., 2008). MCT8 is also expressed in the human as well as the rodent choroid plexus (Heuer et al., 2005; Friesema et al., 2012). LAT1 and 2 do not appear to play any major role in the human brain (Wirth et al., 2014).

T4 is converted to T3 via intracellular outer ring deiodination. In the central nervous system conversion of T4 into T3 is mainly dependent on type 2 deiodinase (D2), and inactivation of T4 and T3 is dependent on inner ring deiodination by type 3 deiodinase (D3). Type 1 deiodinase (D1) does not appear to play a major role in rodent or human brain (Bianco et al., 2002). Interestingly, D2 is mainly expressed in glial cells of the hypothalamus. These include astrocytes and specialized glial cells called tanycytes, which are located at the ependymal layer of the third ventricle (Tu et al., 1997; Alkemade et al., 2005a). *Dio2 mRNA* expression in tanycytes shows a clear increase in response to hypothyroidism and iodine deficiency (Tu et al., 1997; Peeters et al., 2001).

In the cell nucleus, T3 binds the nuclear receptors encoded by the THRA and THRB genes, which give rise to TRα1, α2, β1 and β2 (Sap et al., 1986; Weinberger et al., 1986; Benbrook and Pfahl, 1987; König and Moura Neto, 2002). TRα2 does not bind T3, but exerts dominant negative effects on TRH transcription (Guissouma et al., 2014). Studies in knockout mice indicate that TRβ2 is crucial for the negative feedback loop controlling TH concentrations in the body (Abel et al., 2001). In addition, TRβ1 is also involved in regulating serum TSH (Guissouma et al., 2006). It is not entirely clear what the role for TRα in the negative feedback loop is. In the human and rat hypothalamus TRs are expressed in a number of nuclei, including the PVN where the hypophysiotropic TRH neurons are located (Lechan et al., 1994; Alkemade et al., 2005b). We, and others have proposed models for TH signaling in the hypothalamus, which involves both glial cells and neurons (Guadaño-Ferraz et al., 1997; Tu et al., 1997; Diano et al., 2003; Lechan and Fekete, 2004; Alkemade et al., 2005a; **Figure 1B**).

Hypothalamic TH signaling extends beyond the classical feedback loop affecting the TRH neurons of the PVN. In addition, functioning of other nuclear receptors such as the liver X receptor is dependent on thyroid status (Ghaddab-Zroud et al., 2014). We found in our previous studies on the human hypothalamus that the individual TR isoforms are expressed in a number of hypothalamic nuclei and studies in rats have shown comparable results. In addition, the supraoptic (SON), infundibular (IFN), tuberomamillary (TMN), and the lateral tuberal nucleus (NTL) express TRs. In rodents, TR isoforms have been described in the arcuate nucleus (ARC), the rodent equivalent of the IFN, as well as SON and PVN (Bradley et al., 1989; Cook et al., 1992; Lechan et al., 1994). The SON showed higher TRα than TRβ expression (Bradley et al., 1989). Studies in a transgenic mouse line with a GFP-labeled TRα1 (Wallis et al., 2010) have shown that the distribution of the TRα1 is even more widespread than we concluded from our studies on the human hypothalamus. In mice TRα1 was expressed in the majority of cerebral neurons, including many hypothalamic nuclei, as well as in tanycytes. Several explanations are possible for the discrepancy between our findings and findings in this transgenic mouse strain. It is possible that the antibodies we used for studying TR expression in the human hypothalamus lacked sensitivity to detect all TRs in the human hypothalamus, or, alternatively, interspecies differences may exist. Wallis et al. (2010) further showed TRα1 expression in nonneuronal cells including hypothalamic oligodendrocytes. In our earlier studies in humans we did not report on oligodendritic expression of TRs (Alkemade et al., 2005b). Interestingly, the absence of a functional MCT8 results a persistent hypomyelination in humans (López-Espíndola et al., 2014). The findings by Wallis et al. fit with the ubiquitous expression of the TH transporter OATP1C1 throughout the human hypothalamus, in both neurons as well as glial cells (Alkemade et al., 2011). In rodents Oatp1C1 does not show neuronal expression in the hypothalamus (Roberts et al., 2008). It is possible that other TH transporters are involved in the transport of TH in and out of TRα1 expressing neurons and glia in rodents. In addition, we have shown that MCT8 is expressed in both neurons and glial cells of the human hypothalamus. Interestingly, in humans MCT10 expression was confined to neurons in the majority of hypothalamic nuclei. The distinct distribution patterns support different roles for the individual TH transporters.

### **TH SIGNALING IN THE DEVELOPING HUMAN HYPOTHALAMUS**

Very few studies have investigated the distribution of protein expression in the developing human hypothalamus, therefore, little is known about its chemoarchitecture. Human brain material obtained is not readily available, and tissue from embryo's, fetuses and children is even scarcer. Koutcherov et al. (2002), studied 33 human hypothalamic specimens ranging in age from 9 weeks of gestation to 3 weeks after term birth. This study describes the development of human hypothalamic zones and individual nuclei and provides a clear schematic representation of the development of the human hypothalamus. Hypothalamic differentiation starts already during the first trimester of pregnancy. The lateral zone is the first zone to differentiate, and gives rise to the lateral hypothalamus (LH), Posterior hypothalamic area (PH), lateral tuberal hypothalamic nucleus (LTu) and perifornical hypothalamic nucleus (PeF). The core zone consists of a heterogeneous collection of nuclei positioned between the LH and midline structure, including medial preoptic nucleus (MPO), ventromedial nucleus (VMH), supramamillary nucleus (SUM) and mammillary bodies (Mb). The midline hypothalamus consists of structures differentiating in close proximity of the ventricular wall, including the suprachiasmatic nucleus (SCN), IFN, PVN and SON. These structures become evident during late gestation.

We have described the developmental expression of proteins involved in TH signaling in 15 fetal and infant human hypothalami obtained at various stages of development (Friesema et al., 2012). To my knowledge comparable studies investigating fetal functional neuroanatomy underlying TH signaling are not available for rodents. In our study on the developing human hypothalamus the first sampling point was at 17 weeks of gestation. Although we did not study TR expression in our human brain specimens, at this time both TRs and the ligand T3 have been reported in the human brain and TR mRNA increases strongly from 10–18 weeks gestation (Iskaros et al., 2000). Interestingly, during development TRα1 is not expressed in the ventricular zone in mice where neurons are born and proliferate (Wallis et al., 2010). These studies showed that TRα1 appears to be expressed in immature neurons, preceding the expression of NeuN. The TRα1 expression increases when cells reach their destination and differentiate. This means that TRα1 acts after cell cycle exit. The specific oligodendrocytic expression of TRα1 fits with data showing that oligodendrocyte precursors differentiate after T3 activation of a transiently expressed TR (Billon et al., 2002). This indicates that in mice TRα1 expression is required only during a specific time window in development (Wallis et al., 2010). Whether a similar pattern is also present in the (human) hypothalamus is unknown.

At 17 weeks of gestation the mamillary bodies become prominent and the LH, SON and IFN can already be distinguished, as well as the fornix and the anlage of the PVN (Koutcherov et al., 2002). In our human studies, we observed some MCT8 expression in blood vessels at 17 weeks, indicating possible presence of TH signaling (Friesema et al., 2012). We only observed few D3 positive neurons in the IFN, whereas D2 did not show any staining. These findings indicate T3 degradation, but not production in the human fetal hypothalamus at this moment in development. This is in line with the prevention of TH exposure during early stages of brain development, which may cause untimely maturation of brain cells as observed in rodents (Obregon et al., 2007). Interestingly, excess TH as a result of deletion of the *Dio3* gene in mice results in premature cerebellar differentiation, as well as a central hypothyroidism associated with defective TRH regulation, which persists throughout life (Hernandez et al., 2006, 2007; Peeters et al., 2013).

At 18 weeks gestation all T3 in the brain is produced via local deiodination (Ferreiro et al., 1988), which fits with D2 activity and T3 content in the cerebral cortex during the second trimester (Kester et al., 2004). At this time TH transporters are also expressed in the cerebral cortex (Chan et al., 2014). It does not appear that at this time point the human hypothalamus is capable of converting T4 into T3. This finding could reflect timing differences between distinct brain areas. At 18–23 weeks of gestation the PVN develops, and the PeF area as well as the LH become discernable. At the end of this period the SCN becomes visible (Koutcherov et al., 2002). At present no data are available on TH signaling at this time-point during development.

At 24–33 weeks the fetal human hypothalamus takes and adult like appearance. At 25 5/7 weeks, we found MCT8 expression in the PVN and IFN, which should now also express Neuropeptide Y (Koutcherov et al., 2002). In addition, MCT8 neurons showed a scattered pattern throughout the hypothalamus, as well as in tanycytes and blood vessels (Friesema et al., 2012). MCT10 showed expression in SON, PVN and IFN. D3 and OATP1C1 were also expressed in the IFN. At this time D2 was still very low and only found surrounding blood vessels. At 27 weeks MCT8 expression in PVN and LH persisted, and MCT10 showed more widespread expression in PVN, LH and SON. OATP1C1 was also present in LH, PVN and IFN. Hardly any D2 was observed, whereas the PVN showed prominent D3 expression.

At 27 2/7 weeks of gestation TH signaling appeared to increase. MCT8 was present in neurons of the PVN, IFN and LH, as well as in tanycytes. MCT10 was strongly expressed in the LH, as well as in neurons of the IFN, PVN and SON. OATP1C1 was present in blood vessels of the organum vasculosum laminae terminalis (OVLT) and in LH, IFN, PVN and SON neurons. D2 at this stage showed expression in tanycytes. In addition D2 staining surrounded blood vessels of the OVLT. Now D3 expression was not observed (Friesema et al., 2012). It is possible that the absence of D3 expression indicates an increased need for TH at this stage of development persisting to late gestation, after which TH signaling appears to decrease again. At 28 3/7 weeks gestation MCT8 staining was present in the IFN, PVN and LH. MCT10 and OATP1C1 also showed expression in the LH. In addition, OATP1C1 was also present in the IFN and in tanycytes, which also expressed D2. Again D3 showed very little expression.

From 34 weeks gestation term gestation, the NTL and TMN can develop further (Koutcherov et al., 2002). At 34 5/7 weeks the expression of all tested TH transporters was low. What is interesting to note, is that during late gestation (35 weeks), D2, MCT8 and OATP1C1 expression was very weak, and MCT10 expression was not observed. This was in contrast to the strong D3 expression observed in the IFN, TMN and LH. These observations led us to speculate that these observations may be related to the TSH surge observed in humans associated with birth, and that the hypothalamus may play a role in giving rise to this TSH surge. Although highly speculative, decreased T3 availability due to decreased outer ring deiodination and TH transport, together with increased inner ring deiodination by D3 could lead to upregulation of TRH, which is negatively regulated by T3. A very similar staining pattern was observed in hypothalami of children that were born at term. Inverse hypothalamic expression levels of D2 and 3 have also been observed in birds and mammals in the context of photoperiod setpoints (Watanabe et al., 2007).

At *2 months of age* staining for MCT8 in SON, PVN, and in the lining of the third ventricle was found in humans. MCT10 immunoreactivity was present in SON and PVN. Sporadic expression was found in the LH. OATP1C1 showed a more scattered pattern and increased staining was observed in SON, PVN, TMN, IFN and the ependymal layer. D2 expression was present in blood vessels. D3 showed expression in PVN as well as other hypothalamic nuclei. In children *aged 20–29 months of age*, we found moderate MCT8 and moderate to strong MCT10 staining in PVN, SON, IFN and weak staining in the LH. OATP1C1 was present in PVN, SON, IFN and in some cells in the ependymal layer of the third ventricle. In addition, scattered OATP1C1 positive cells were present, as well as staining in NTL, TMN, SCN and LH. D2 was expressed in blood vessels especially in the IFN and in tanycytes. At this time point stainings very much resembled our observations in the adult hypothalamus (Alkemade et al., 2005a, 2011; **Figure 2**).

Hypothalamic expression patterns throughout the developmental period studied are summarized in **Figure 3**. This figure indicates that regulation of proteins involved in TH signaling may play an important role in timing of TH availability at different stages of development.

### **DEFECTIVE HYPOTHALAMIC TH SIGNALING**

Defects in proteins involved in hypothalamic TH signaling are potentially reflected in alterations in serum TH and TSH concentrations. A number of mutations in, TRs and TH transporters has been described. These syndromes are described excellently in a recent review (Dumitrescu and Refetoff, 2013), and here only findings relevant for HPT-axis regulation and the brain are recapitulated.

#### **TRs**

Complete deletion of all TRs in mice causes serum TSH levels that are 500-fold higher than those of the WT mice, and T4 concentrations 12-fold above the average normal mean (Gauthier et al., 1999). In humans both mutations in the THRA and THRB genes have been reported.

#### **RESISTANCE TO TH**α

The first subject described suffering from Resistance to THα (RTHα), was diagnosed only recently (Bochukova et al., 2012). RTHα is caused by a mutation in the *THRA* gene, which encodes TRα1 and α2. The mutation caused a truncation of the protein, which lacked the C-terminal α-helix. This 6-year old girl showed low-normal or subnormal levels of total T4 and free T4, high-normal or elevated levels of total T3 and free T3, in addition to normal levels of TSH (Bochukova et al., 2012). Since the description of this young girl, additional patients with RTHα have been described in literature. RTHα patients show cognitive impairment and macrocephaly (Bochukova et al., 2012; van Mullem et al., 2013). In addition, the phenotype observed in a patient with mutations both in TRα1 and α2 was very similar (Moran et al., 2014). Serum TH concentrations do not suggest any major impairment of the classic hypothalamic feedback system. This is corroborated by findings in mice. Mouse models with a mutant TRα1, show a range of phenotypes, dependent on the specific mutation or deletion. Brain changes have been described as well (Wallis et al., 2008; Mittag et al., 2010), and changes resulting from defects in the autonomic nervous system including impaired cardiovascular response to stress, activity and environmental temperature changes. The hypothalamus has only been studied in detail in one of these mouse models, showing the absence of a parvalbumin positive neuronal population in the anterior hypothalamic area with consequences for the regulation of the cardiovascular system

indicate sites of TR isoform expression. Upper panels: rostral level; lower panels: caudal level. III, Third ventricle; AC, anterior commissure; BST, bed nucleus of the stria terminalis; DBB, diagonal band of Broca; FO, fornix; LV, tract; SCN, suprachiasmatic nucleus; SDN, sexually dimorphic nucleus; SON, supraoptic nucleus; TMN, tuberomammilary nucleus (Taken from Alkemade et al., 2005b).

(Mittag et al., 2013). The HPA-axis was not affected in these mutant mice as assessed by hypothalamic corticotropin releasing hormone (CRH) and pro-opiomelanocortin (POMC) expression (Mittag et al., 2010).

#### **RESISTANCE TO TH**β

Patients harboring mutations in the TRβ1 show more pronounced changes in serum TH concentrations. For the THRB gene number of mutation hotspots have been defined (Dumitrescu and Refetoff, 2013). Characteristic are elevated freeT4 levels, and to a lesser extent T3. In addition, TSH is normal or slightly increased, and responsive to TRH. Patients generally do not display the usual metabolic symptoms associated with hyperthyroidism, although they often present with goiter (Refetoff et al., 1993). TRβ2 knockout mice show a severe disruption of the hypothalamic TH feedback system (Abel et al., 2001).

#### **TH TRANSPORTERS**

In MCT8 deficient subjects serum TSH is modestly increased, which fits with the decreased serum T<sup>4</sup> concentration but not with the elevated serum T<sup>3</sup> level. Since MCT8 is expressed in

the hypothalamus and pituitary, it is likely that inactivation of the gene interferes with the negative feedback at the level of the hypothalamus (Fliers et al., 2006). In addition, defects in brain development have been described in patients lacking MCT8 (López-Espíndola et al., 2014). Rodent models for MCT8 mutations do not reproduce the severe psychomotor phenotype observed in humans, but do faithfully reproduce the biochemical phenotype (Trajkovic et al., 2007). In *Mct8*KO mice, hypothalamic TRH expression is markedly increased and high T<sup>3</sup> doses are needed to suppress it. In humans there is clear expression of MCT10 and OATP1C1 already in the second trimester of pregnancy. The presence of these transporters is does compensate for the absence of MCT8 as evidenced by the neurological defects observed in patients lacking MCT8. Wirth et al. suggest that Lat2 might compensate for the Mct8 deletion in mice. This is unlikely in humans, since developing neurons in the human brain only show very low LAT2 expression (Wirth et al., 2009). It is possible that in rodents other transporter variants such as Oatp1a4 and Oatp1a5 may further compensate for MCT8 defects. These transporters appear not to have orthologs in the human brain (Suzuki and Abe, 2008).

#### **TH CONVERSION DEFECTS**

Selenocysteine insertion sequence (SECIS) binding protein 2 (SBP2) plays an important role in insertion of selenocysteine into selenoproteins such as deidonases. Defects in the *SBP2* gene therefore interfere with the production of these conversion enzymes, resulting in high T4 and rT3, low T3, normal or slightly elevated TSH concentrations (Dumitrescu et al., 2005). Genetic variants in *DIO2* in humans do not show any clear phenotypic changes (Zevenbergen et al., 2014).

#### **CONCLUSION**

TH signaling involves facilitated transport, local conversion and receptor binding. TH signaling is therefore dependent on a number of proteins, each of which can become defective, thereby affecting a plethora of processes modulated by TH. The effects of mutations in TH signaling on the functional neuroanatomy, which underlies the classic negative feedback of TH on the hypothalamus is dependent on the affected gene, the type of mutation, as well as the compensatory mechanisms, which appear to differ between species. These effects are largely assessed by evaluation of biochemical parameters. Detailed studies on the functional neuroanatomy underlying hypothalamic TH signaling are scarce, and are complicated by interspecies differences, as well as the limited availability of human postmortem brain material for research purposes. Future studies on the expression of TR isoforms in the developing (human) hypothalamus would strongly improve our understanding of central TH signaling during development.

#### **REFERENCES**


isoforms in regulating TRH transcription. *Neurosci. Lett.* 406, 240–243. doi: 10. 1016/j.neulet.2006.07.041


**Conflict of Interest Statement**: The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 30 September 2014; accepted: 02 February 2015; published online: 20 February 2015*.

*Citation: Alkemade A (2015) Thyroid hormone and the developing hypothalamus. Front. Neuroanat. 9:15. doi: 10.3389/fnana.2015.00015*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Alkemade. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Embryonic development of circadian clocks in the mammalian suprachiasmatic nuclei

#### **Dominic Landgraf <sup>1</sup> , Christiane E. Koch<sup>2</sup> and Henrik Oster <sup>2</sup>\***

<sup>1</sup> Center of Circadian Biology and Department of Psychiatry, University of California, San Diego, and Veterans Affairs San Diego Healthcare System, San Diego, CA, USA

<sup>2</sup> Chronophysiology Group, Medical Department I, University of Lübeck, Lübeck, Germany

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Manuel A. Pombal, University of Vigo, Spain Paul A. Gray, Washington University, USA

**\*Correspondence:** Henrik Oster, Chronophysiology Group, Medical Department I, University of Lübeck, Ratzeburger Allee 160, 23538 Lübeck, Germany

e-mail: henrik.oster@uksh.de

In most species, self-sustained molecular clocks regulate 24-h rhythms of behavior and physiology. In mammals, a circadian pacemaker residing in the hypothalamic suprachiasmatic nucleus (SCN) receives photic signals from the retina and synchronizes subordinate clocks in non-SCN tissues. The emergence of circadian rhythmicity during development has been extensively studied for many years. In mice, neuronal development in the presumptive SCN region of the embryonic hypothalamus occurs on days 12–15 of gestation. Intra-SCN circuits differentiate during the following days and retinal projections reach the SCN, and thus mediate photic entrainment, only after birth. In contrast the genetic components of the clock gene machinery are expressed much earlier and during midgestation SCN explants and isolated neurons are capable of generating molecular oscillations in culture. In vivo metabolic rhythms in the SCN, however, are observed not earlier than the 19th day of rat gestation, and rhythmic expression of clock genes is hardly detectable until after birth. Together these data indicate that cellular coupling and, thus, tissue-wide synchronization of single-cell rhythms, may only develop very late during embryogenesis. In this mini-review we describe the developmental origin of the SCN structure and summarize our current knowledge about the functional initiation and entrainment of the circadian pacemaker during embryonic development.

**Keywords: suprachiasmatic nucleus, circadian clocks, embryonic and fetal development, entrainment, clock genes**

#### **INTRODUCTION**

Endogenous circadian clocks facilitate the adaptation of behavior and physiology to the 24-h rhythm of day and night. In mammals, the circadian timing is organized by pacemaker cells in the hypothalamic suprachiasmatic nuclei (SCN). This SCN master clock is reset by photic time cues, or *Zeitgebers*, perceived through the retina and transmitted via the retino-hypothalamic tract (RHT). At the molecular level, the cellular clocks in the SCN and other tissues are built from self-sustained interlocked transcriptional-translational feedback loops of clock genes/proteins characterized by rhythmic transcription patterns. While clock function and rhythm generation have been extensively studied in adults, there is still no agreement on how circadian rhythms emerge during embryonic development.

#### **ANATOMICAL DEVELOPMENT OF THE SCN**

The ontogeny of the SCN has been extensively described in rodents (**Figure 1**), while only few data on primate SCN development are available (Weinert, 2005). The rat SCN is derived from the neuroepithelium of the preoptic recess of the third ventricle and becomes discernable as a discrete structure at embryonic day E17 (Altman and Bayer, 1978b). Neurogenesis in the rat mainly occurs between E12 and E18 with a maximum at E16 (Ifft, 1972; Altman and Bayer, 1978a). In mice neurogenesis begins earlier and is restricted to days E10–15, peaking at E12 (Shimada and Nakamura, 1973; Kabrita and Davis, 2008). In the hamster SCN, neurons are born even earlier at E9.5–13 (Davis et al., 1990; Antle et al., 2005). In squirrel monkeys, SCN neurons have been described at late gestational stages and, in humans, the SCN is discernable as a discrete structure around the 18th–30th week of pregnancy (Reppert and Schwartz, 1984; Reppert et al., 1988; Swaab et al., 1990).

The ontogeny of non-neuronal cell types within the SCN also follows temporal programs of formation (Botchkina and Morin, 1995; Antle et al., 2005). In the hamster SCN, radial glial cells start to develop at E8 and appear at high density in the SCN at E13. At post-natal day 0 (P0), the density of these cells in the hamster SCN is drastically reduced and by P5 most of them are replaced by neurons. Hamster SCN astrocytes start to form at E15 and development continues at least until P21 (Botchkina and Morin, 1995). In the rat, astrocytes first appear at E20 and rapidly increase after birth, and radial glia cells are present at E17, which was the earliest stage investigated in this study (Munekawa et al., 2000), and begin to disappear at P0.

#### **ACTIVATION OF SCN MARKER GENE EXPRESSION**

The neuropeptides vasopressin (AVP) and vasoactive intestinal polypeptide (VIP) are strongly expressed in the SCN and contribute to its role as circadian pacemaker (Vosko et al., 2007;

Kalsbeek et al., 2010). In the hamster SCN, AVP expressing cells first appear on the day of birth (Romero and Silver, 1990). In the mouse, AVP neurons first appear at E12 at the peak of neurogenesis (Okamura et al., 1983). However, AVP mRNA appears later at E17.5 of mouse development and AVP protein is only detectable at postnatal stages (VanDunk et al., 2011). From the 27th week of pregnancy, AVP is detectable in the human embryonic hypothalamus (Swaab et al., 1990). From then on, AVP levels constantly increase until 1–1.5 years after birth.

shown in purple. Furthermore, the neurogenesis of each species occurring

In the developing hamster SCN, VIP expressing neurons are detectable after completion of neurogenesis at E13–14 and VIP expression increases substantially until P10 (Romero and Silver, 1990; Botchkina and Morin, 1995). VIP mRNA first appears only after neurogenesis at E18.5 in the developing mouse SCN and VIP protein is detectable after birth (VanDunk et al., 2011). In contrast, in the rat SCN, *VIP* mRNA first appears directly after completion of neurogenesis at E18 and increases after birth until stage P20 (Ban et al., 1997). VIP protein in the rat SCN is first expressed no later than E20 (Laemle, 1988). However, earlier embryonic stages have not yet been investigated. The first VIP neurons in the human SCN are observed at 31*th* week of pregnancy with increasing numbers until the age of 3 years. Spatial re-organization of VIP neurons, however, continues throughout puberty (Swaab et al., 1994).

highlighted during the postnatal development of mice. For details see text.

In mice, a combination of transcription factors shows distinct spatial and temporal patterns during prenatal and postnatal SCN development (VanDunk et al., 2011). The SCN anlage emerges from a specific region of the neuroepithelium expressing *Six homeobox 3* and *6* (*Six3*, *Six6*), *frizzled-5* (*Fzd5*), and transient *retinal homeobox gene* (*Rx*). Throughout embryonic development *Six3*, *Lim homeodomain transcription factor 1* (*Lhx1*) and *RARrelated orphan receptor alpha* (*Rora*) expression remains restricted to the SCN region. Conditional deletion of *Six3* at early stages prevents SCN development completely (VanDunk et al., 2011). In addition, *Six6* is required for the normal development of the optic nerves and the SCN. *Six6*-null mice display abnormal entrainment behavior, due to the lack of functional optic nerves, but also show abnormal circadian rhythms under freerunning conditions, indicating the absence of a functional SCN. Indeed, immunohistochemical stainings for VIP and AVP confirmed the absence of a defined SCN structure in *Six6*-null mice (Clark et al., 2013). A selective deletion of SCN-enriched *Lhx1* in the developing anterior hypothalamus, including the developing SCN, leads to reduced SCN-enriched neuropeptides involved in circadian function and causes loss and death of neurons in the developing mouse SCN (Bedont et al., 2014). Furthermore, the loss of *Lhx1* during SCN development leads to reduced cellular synchrony in the SCN and results in disrupted circadian rhythms in adult mice (Bedont et al., 2014; Hatori et al., 2014).

### **DEVELOPMENT OF CIRCADIAN PACEMAKER FUNCTION IN THE SCN**

At which embryonic stage endogenous SCN rhythms emerge, is still a matter of debate. Shortly before birth, glucose utilization in the embryonic SCN of squirrel monkeys shows diurnal rhythms (Reppert and Schwartz, 1983, 1984). Similarly, expression of the neuronal activity marker *c-fos* in the embryonic SCN of sheep at gestational day 135 shows diurnal fluctuations (Constandil et al., 1995). However, the mother animals used for these studies had functional circadian clocks and were housed in rhythmic light-dark (LD) conditions during the entire or at least during large parts of gestation. Thus, maternal or external timing cues driving diurnal rhythms observed in fetuses cannot be excluded and the existence of an endogenous fetal circadian clock cannot be concluded from these studies. Contrary, in rats it was shown that embryonic SCN rhythms emerge even in the absence of maternal clocks and under constant housing conditions. Glucose utilization and also neuronal firing of cultured embryonic rat SCNs at E22 exhibit circadian rhythms *in vitro*, suggesting that these oscillations are evoked by an endogenous clock (Shibata and Moore, 1987, 1988). In addition, the offspring of SCN-lesioned rat mothers displays circadian behavioral rhythms indicating that a functional circadian system develops in the absence of rhythmic maternal signals (Reppert and Schwartz, 1986). In humans, preterm infants of 29–35 weeks of age show circadian temperature rhythms under constant lighting and feeding conditions, which suggests the existence of a functional pacemaker during human gestation (Mirmiran and Kok, 1991).

Additional indirect evidence of functional embryonic SCN clocks comes from transplant studies. Implanting embryonic SCN cells or grafts into the brain of SCN-lesioned adult hamsters restores free-running rhythms soon after transplantation (Silver et al., 1990; Kaufman and Menaker, 1993). Importantly, the outcome was independent of the age of the donor embryo suggesting that the SCN is already fully functional during midgestation.

In contrast, examination of clock gene or protein expression in fetal SCNs yields ambiguous results. In constant darkness, mRNA expression of *Per1*, but not *Per2*, is rhythmic at E17 in mice (Shimomura et al., 2001). At E18, PER1 and PER2 protein levels show circadian rhythms in the mouse embryonic SCN when mothers are kept in constant darkness, albeit with low baseline levels and moderate amplitude (Ansari et al., 2009). In rats kept under LD conditions, *Per2* and *Bmal1* mRNA expression is not rhythmic at E19, but on E21 (Houdek and Sumová, 2014). A different study has shown that *Per1* and *Per2* mRNA is already rhythmic in rat embryos at stage E20 (Ohta et al., 2002, 2003). Interestingly, although *Per2* and *Bmal1* expression is arrhythmic at E19 in the rat SCN, *nuclear receptor subfamily 1, group D, member 1* (*Nr1d1*), *c-fos*, *Avp*, and *Vip* mRNAs already show significant circadian rhythms that are likely driven by maternal signals (Houdek and Sumová, 2014). Rhythmic *Bmal1* and *Per2* expression is seen in fetal SCNs of capuchin monkeys at late stages of gestation (Torres-Farfan et al., 2006). However, since the mother animals used in all these studies had a functional circadian clock and/or were kept in an LD cycle, an influence of rhythmic maternal signals on rhythmic clock gene expression in the embryonic SCN cannot be ruled out. Accordingly, other studies have shown that when pregnant rats are kept in constant darkness, *Per1*, *Per2*, and other clock gene mRNAs and proteins are not rhythmic in the embryonic SCN at E19/20 and rhythms only emerge around P1 (Sládek et al., 2004; Kováciková et al., 2006). However, cultured explants of PERIOD2::LUCIFERASE (PER2::LUC) mice, which carry the firefly luciferase gene in the wild-type circadian clock gene *Per2* as a reporter for circadian rhythms (Yoo et al., 2004), revealed the capability of early embryonic SCNs of stage E15 to express self-sustained circadian rhythms *in vitro*. With advancing age of the embryos, the rhythms became gradually stronger (Wreschnig et al., 2014). Whether the observed rhythms reflect circadian oscillations that were already present *in vivo* or whether rhythms were induced by the tissue preparation procedure could, however, not be clarified.

#### **ENTRAINMENT OF THE FETAL SCN**

A hallmark of circadian clocks is their ability to entrain to external *Zeitgebers*. Circadian rhythms in the embryonic SCN of different species, including humans, strongly react to different signals including light, melatonin, food, and dopamine.

Depending on the lighting conditions, rat embryos show different phasing in glucose utilization rhythms (Reppert and Schwartz, 1983). When pregnant rats are housed under normal or inverted LD cycles, the embryonic SCN entrains to the external conditions. SCNs of squirrel monkey embryos from mothers housed in different LD cycles stay in phase with the maternal rhythm (Reppert and Schwartz, 1984). However, since synaptic connections from the retina to the SCN are only formed after birth (see below) these studies suggest an indirect entrainment of embryonic SCN clocks by light, probably via signals derived from the mother and passed on through the placental barrier. One candidate for a maternal signal, which entrains the embryonic clock, is melatonin released by the pineal gland (Wurtman et al., 1964) as daily melatonin injections can reset embryonic rhythms in pregnant SCN-lesioned hamsters (Davis and Mannion, 1988). A single injection of melatonin one day before birth is sufficient to shift the phase of the offspring (Viswanathan and Davis, 1997). In pregnant pinealectomized rats arrhythmicized by constant light melatonin injections influence the phase of *Avp* and *c-fos* expression in the offspring (Houdek et al., 2014). Melatonin effects on fetal SCNs are mediated through locally expressed melatonin receptors. In capuchin monkey fetuses, melatonin 1 receptor (MT1) is strongly expressed in the SCN and suppression of maternal melatonin changes oscillating expression patterns of *Bmal1*, *Per2* and *MT1* (Torres-Farfan et al., 2006). At around midgestation, the human embryonic SCN also expresses melatonin binding sites (Reppert et al., 1988).

Periodic feeding of pregnant SCN-lesioned rats synchronizes drinking behavior in the offspring, suggesting that feeding can entrain the SCN clock of fetuses during embryogenesis (Weaver and Reppert, 1989). Even in pregnant rats housed in LD with an intact circadian clock, time-restricted feeding shifts *Per1* expression in the fetal SCN by several hours when food is restricted to a 4-h period in the inactive phase of the animals (Ohta et al., 2008). In contrast, *c-fos* and *Avp* expression patterns are not affected. However, when the circadian clocks of the rat mothers are disrupted by constant light exposure, restricted feeding resets *c-fos* and *Avp* expression of the newborns suggesting that light may be a stronger *Zeitgeber* for the fetal SCN than food-related signals (Nováková et al., 2010).

Dopamine is a potent regulator of the molecular circadian clock machinery (Yujnovsky et al., 2006). Timed treatment with SKF 38393, a dopamine receptor agonist, shifts the activity phase of the offspring from SCN-lesioned hamster mothers (Viswanathan et al., 1994; Viswanathan and Davis, 1997). Interestingly, melatonin injections have similar effects, but lead to opposite shifts in offspring activity. Dopamine and lightinduced glutamatergic signaling converge on shared intracellular kinase pathways (Schurov et al., 2002; Govindarajan et al., 2006; Colwell, 2011). However, light and the melatonin phase response curves are about 12 h out of phase (Lewy et al., 1998), which may explain the opposite directional effects on offspring activity.

### **EMBRYONIC AND POSTNATAL DEVELOPMENT OF SCN INPUT PATHWAYS**

After birth, light becomes the most important *Zeitgeber* to entrain circadian rhythms and also influences the maturation of the SCN itself. Irradiance information is received by melanopsin-expressing intrinsically photosensitive retinal ganglion cells (ipRGCs). ipRGCs directly sense light, but also integrate input from rod and cone photoreceptors and signal via the RHT to the SCN. Melanopsin, the photopigment of the ipRGCs, is detectable in the retina of prenatal rodents—mice E11.5 (Tarttelin et al., 2003), rats E18 (Fahrenkrug et al., 2004) and ipRGCs become light responsive directly after gestation (Sekaran et al., 2005; Tu et al., 2005). However, at this time ipRGCs are randomly distributed throughout the ganglion cell layer and the inner nuclear layer of the mouse retina (Tu et al., 2005). Between birth and adulthood, re-organization as well as reduction of about 65% of ipRGCs occurs. During maturation ipRGCs separate into distinct regions of the inner plexiform layer. At P4, most of them are located in this region, but further separation occurs during the following days. This re-organization is accompanied by a profound loss of melanopsin-positive cells (Sekaran et al., 2005; Tu et al., 2005) that is associated with a 10 fold increase in photic sensitivity at P4-6. This increase is partially, but not solely, induced by a gradual increase in retinal melanopsin expression (Tu et al., 2005). Almost simultaneously, the first circadian rhythm of melanopsin gene (*OPN4*) expression is observed around P5 (González-Menéndez et al., 2009) indicating a further maturation of the ipRGCs.

Paralleling ipRGC development, the RHT that is principally functional directly after birth at P0 in mice (Lupi et al., 2006) and P1 in rats (Leard et al., 1994), maturates during the first 1–2 weeks after birth to reach full functionality in rats (Takahashi and Deguchi, 1983; Duncan et al., 1986). Whereas the development of the RHT of hamsters starts at P4 and reaches the adult pattern by P15, its development in rats initiates prenatally at E21–22. First connections to the ventral part of the rat SCN appear at P1, reaching maximal density at P4 and prune back to the adult pattern by P10. At that stage, the first gating of light induced *c-Fos* production is detectable in the SCN of rat pups (Bendová et al., 2004). While the maturation of the light signaling cascade is mostly finished by P10 in rats (Speh and Moore, 1993), it was shown in mice that the adult-like light response is detectable not until around P14 when the eye opening occurs (Muñoz Llamosas et al., 2000), indicating that opening the eyes and removing the light dampening cover over it enables the rodent to fully respond to the environmental light.

Whether changes in pre- or postnatal lighting conditions may interfere with the functional development of ipRGCs, the RHT or the SCN itself is not known. However, light input during postnatal phase may affect neuropeptide expression in the SCN and the circadian system. CBA/J mice, which show increased photosensitivity compared with C57Bl/6 mice, display elevated VIP and AVP levels in the SCN (Ruggiero et al., 2010) that are paradoxically associated with attenuated phase shifting behaviors (Yoshimura et al., 1994; Ruggiero et al., 2009). Additionally, constant postnatal light or darkness conditions affect the stability of circadian behavioral rhythms in mice (Canal-Corretger et al., 2000, 2001), possibly due to changes in astrocyte development in the SCN (Canal et al., 2009). Darkness exposure leads to more and larger astrocytes whereas constant light reduces SCN astrocyte numbers associated with an increased stability of circadian rhythms and overall increased running-wheel activity (Canal et al., 2009).

### **CONCLUSION**

While the structural development of the SCN is relatively well understood, the question whether endogenous circadian rhythmicity in the SCN develops before birth is still matter of debate and studies based on SCN output or clock gene expression in the SCN provide different results. Species-specific traits make it difficult to draw a generally valid conclusion and the majority of studies investigating the existence of embryonic clocks are carried out in rhythmic environments and/or with pregnant females, which have a functional circadian clock. Consequently, environmental or maternal rhythmic signals driving diurnal rhythms in the embryonic SCN cannot be excluded, and the presence of an endogenous, self-sustained fetal SCN clock cannot be demonstrated. Studying pup development in arrhythmic mothers continually housed in constant environments during pregnancy may be a feasible approach to this problem. It was shown that behavioral rhythms of pups born under such conditions are not synchronized to each other. This complicates the investigation of rhythmic gene expression in embryonic SCNs because most techniques applied to measure gene activity only allow investigating one time point per animal. An alternative approach using animals carrying circadian-controlled reporters for continuous recording of transcriptional/translational activity from one specimen was conducted. However, to clarify whether clock gene oscillations of *in vitro* SCN explants reflect rhythms that were already present *in vivo* or that were rather initiated by the tissue culture procedure, SCN tissues may be collected and cultured at different daytimes. If explant phasing was determined by *in vivo* rhythms, phasing for all groups should be roughly the same and should be independent of the preparation time. In combination with improved *in vivo* imaging techniques this may finally facilitate the determination of the emergence of SCN and peripheral tissue clock function during ontogeny and allow a clear distinction between maternal and embryo-derived rhythms.

#### **REFERENCES**


Yoshimura, T., Nishio, M., Goto, M., and Ebihara, S. (1994). Differences in circadian photosensitivity between retinally degenerate CBA/J mice (rd/rd) and normal CBA/N mice (+/+). *J. Biol. Rhythms* 9, 51–60. doi: 10. 1177/074873049400900105

Yujnovsky, I., Hirayama, J., Doi, M., Borrelli, E., and Sassone-Corsi, P. (2006). Signaling mediated by the dopamine D2 receptor potentiates circadian regulation by CLOCK:BMAL1. *Proc. Natl. Acad. Sci. U S A* 103, 6386–6391. doi: 10. 1073/pnas.0510691103

**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 23 September 2014; accepted: 13 November 2014; published online: 01 December 2014*.

*Citation: Landgraf D, Koch CE and Oster H (2014) Embryonic development of circadian clocks in the mammalian suprachiasmatic nuclei. Front. Neuroanat. 8:143. doi: 10.3389/fnana.2014.00143*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2014 Landgraf, Koch and Oster. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

## Origin and early development of the chicken adenohypophysis

#### **Luisa Sánchez-Arrones <sup>1</sup>† , José L. Ferrán<sup>1</sup> , Matías Hidalgo-Sanchez <sup>2</sup> and Luis Puelles <sup>1</sup>\***

<sup>1</sup> Faculty of Medicine, Department of Human Anatomy, School of Medicine and IMIB (Instiuto Murciano de Investigación Biosanitaria), University of Murcia, Murcia, Spain

<sup>2</sup> Department of Cell Biology, Faculty of Science, University of Extremadura, Badajoz, Spain

#### **Edited by:**

Gonzalo Alvarez-Bolado, University of Heidelberg, Germany

#### **Reviewed by:**

Kenji Shimamura, Kumamoto University, Japan Diego Echevarria, University of Miguel Hernandez (UMH-CSIC), Spain

#### **\*Correspondence:**

Luis Puelles, Faculty of Medicine, Department of Human Anatomy, School of Medicine and IMIB (Instiuto Murciano de Investigación Biosanitaria), University of Murcia, Campus Espinardo s/n, Murcia, 30071 MU, Spain e-mail: puelles@um.es

#### **†Present address:**

Luisa Sánchez-Arrones, Centro de Biología Molecular Severo Ochoa, Consejo Superior de Investigaciones Científicas - Universidad Autónoma de Madrid, Madrid, Spain

The adenohypophysis (ADH) is an important endocrine organ involved in the regulation of many physiological processes. The late morphogenesis of this organ at neural tube stages is well known: the epithelial ADH primordium is recognized as an invagination of the stomodeal roof (Rathke's pouch), whose walls later thicken and differentiate as the primordium becomes pediculated, and then fully separated from the stomodeum. The primordium attaches to the pial surface of the basal hypothalamus, next to the neurohypophyseal field (NH; future posterior pituitary), from which it was previously separated by migrating prechordal plate (pp) cells. Once the NH evaginates, the ADH surrounds it and jointly forms with it the pituitary gland. In contrast, little is known about the precise origin of the ADH precursors at neural plate stages and how the primordium reaches the stomodeum. For that reason, we produced in the chicken a specific ADH fate map at early neural plate stages, which was amplified with gene markers. By means of experiments labeling the mapped presumptive ADH, we were able to follow the initial anlage into its transformation into Rathke's pouch. The ADH origin was corroborated to be strictly extraneural, i.e., to lie at stage HH4/5 outside of the anterior neural plate (anp) within the pre-placodal field. The ADH primordium is fully segregated from the anterior neural border cells and the neighboring olfactory placodes both in terms of precursor cells and molecular profile from head fold stages onwards. The placode becomes visible as a molecularly characteristic ectodermal thickening from stage HH10 onwards. The onset of ADH genoarchitectonic regionalization into intermediate and anterior lobes occurs at closed neural tube stages.

**Keywords: adenohypophysis, anterior pituitary, fate map, pre-placodal ectoderm, placodes, gene markers, Rathke's pouch**

#### **INTRODUCTION**

The pituitary gland of vertebrates is a key regulator of the endocrine system, required for the maintenance of reproduction, growth, homeostasis and metabolism. It develops by apposition of two embryonic primordia: the posterior pituitary gland, or neurohypophysis (NH), which is a dock for release of the hypothalamic neurohormones vasopressin and oxytocin are into the bloodstream, and the anterior pituitary gland, or adenohypophysis (ADH), which represents a major endocrine control organ (Herzog et al., 2004; Guner et al., 2008). The NH develops as a digitiform median evagination from the

**Abbreviations:** abb, alar basal boundary; anp, anterior neural plate; ANR, anterior neural ridge; bp, basal plate; Di, diencephalon; dt, digestive tube; ec, ectoderm; HN, Hensen's node; Hypoth, hypothalamus; Inf, infundibulum; mb, midbrain; N, notochord; nf, neural fold; nh, neurohypophysis; och, optic chiasm; oe, oral ectoderm; olf, olfactory placode; os, optic stalk; ov, optic vesicle; Pall, pallium; pp, prechordal plate; ppr, preplacodal region; PS, primitive streak; Pt/Th, Pretectum/Thalamus; Rh, Rhombencephalon; rm, retromamilar; rp, Rathke's pouch; SP, secondary prosencephalon; SPall, subpallium; Tel, telencephalon.

hypothalamic tuberal neuroepithelium, at the area known as "median eminence" (Cobos et al., 2001; Puelles and Rubenstein, 2003; Herzog et al., 2004).

The ADH origin is still under discussion, since different developmental scenarios have been conjectured in the light of fate-mapping experiments in diverse vertebrates. The controversy centers on whether the ADH origin is neural or non-neural. Various fate-mapping studies performed at neurula stages in zebrafish, frog, and chick embryos reported that the ADH arises at the rostral border of the neural plate (known as the "anterior neural ridge", ANR; Takor and Pearse, 1975; Couly and Le Douarin, 1985; Eagleson et al., 1986; Couly and Le Douarin, 1987; elAmraoui and Dubois, 1993; Dubois and Elamraoui, 1995; Whitlock and Westerfield, 2000; Whitlock et al., 2003; Herzog et al., 2004). Specifically, Couly and Le Douarin (1988) thought that this early neural ADH primordium was continuous with the presumptive hypothalamic NH primordium (possibly suggesting, but not reasoning out, an implicit answer to the question about why ADH and NH later come to be joined in the pituitary gland).

In contrast, other experimental work done in mammals and birds suggested that the primary ADH domain is not neural and arises from a median placodal ectodermal domain placed in front of (outside) the ANR. For instance, Cobos et al. (2001) concluded that the avian ANR domain proper contains exclusively prospective telencephalic cells (never giving rise to NH or ADH cells when grafted selectively). The ADH primordium was labeled only when the ectoderm forming the outer non-neural slope of the ANR was included in the grafts. This distinction was not made by earlier authors reporting such experiments. These results implied that the non-neural ADH primordium is separated from the prospective tuberal NH by several interposed neural domains, such as the telencephalic preoptic area (later represented at the midline by the lamina terminalis) and the hypothalamic chiasmatic, retrochiasmatic and tuberal areas. Their subsequent meeting must be due to parallel morphogenetic changes in position of the ADH and NH primordia. These results about the strictly telencephalic character of the ANR neural plate domain were corroborated later by more recent fate maps of the chicken neural plate (Fernández-Garre et al., 2002; Sanchez-Arrones et al., 2009, 2012; Cajal et al., 2012, 2014); see also zebrafish and frog fate maps (Houart et al., 1998; Eagleson and Dempewolf, 2002).

Everybody in the field agrees that Rathke's pouch, a dorsal evagination of the stomodeal roof, represents the immature ADH later in development (Cobos et al., 2001; Rizzoti and Lovell-Badge, 2005). It is also widely accepted that the stomodeal roof is ectodermal in character, rather than endodermal, thus excluding the possibility of an endodermal origin of the ADH. Nevertheless, it was found that the foregut endoderm underlying the early ANR is required during the specification of both the anterior neural border and the ADH primordium (Withington et al., 2001; Sanchez-Arrones et al., 2012).

There is little information, though, about how the early ADH primordium, be it neural or non-neural, comes to occupy the position of Rathke's pouch under the hypothalamus. This suggests the need of a detailed fate map and morphogenetic follow-up at several early developmental stages, in order to understand more fully the early development of the anterior pituitary gland (**Figure 7**). In this essay we re-examined in detail the origin and subsequent changing position of the ADH epithelial plate with regard to neighboring tissues at various stages. To address this issue, we first examined the early ADH field by fate mapping at neural plate stages (HH4/5) and survival up to closed neural tube stages (HH12), using the chick embryo as a model system. These data showed that the ADH precursors are located within an area distant 260– 290 µm from the node in the median non-neural ectoderm rostral (one might rather say "dorsal", or "peripheral") to the ANR. This ADH primordium is a close neighbor of the rostral midline telencephalic domain (prospective subpallial cells; see Cobos et al., 2001; Sanchez-Arrones et al., 2009; Cajal et al., 2014); the prospective ADH ectoderm was thus again found to lie just outside the neural plate and quite distant from the prospective NH. Moreover, the ADH primordium directly overlies the boundary between the extraembryonic hypoblast and the rostralmost foregut endoderm precursors (Kimura et al., 2006).


#### **MATERIALS AND METHODS**

#### **FATE MAP OF THE ROSTRAL MIDLINE ECTODERM**

Fertilized chicken eggs were incubated at 38◦C to reach the desired stage. Fate-mapping experiments (**Tables 1**, **2**) were performed at stage HH4/5 (Hamburger and Hamilton, 1951), with focal injections (Selleck and Stern, 1991) of carbocyanine dyes DiI (1, 1<sup>0</sup> -dioctadecyl-3,3,3<sup>0</sup> ,30 -tetramethylindocarbocyanine perchlorate; Invitrogen, D282) and/or DiO (3,3<sup>0</sup> dioctadecyloxacarbocyanine perchlorate; Invitrogen, D275) into the rostromedian neural and non-neural ectoderm in Newcultured chick embryos (New, 1955; protocol modified by Stern and Bachvarova, 1997). The injection was performed by applying buccal pressure to tubing connected with glass micropipettes filled with the dye solution. In some cases, homotopic grafts labeled with green fluorescent 5-(and-6)-carboxyfluorescein diacetate, succinimidyl ester (CFSE) (Molecular Probes) were performed as detailed by Fernández-Garre et al. (2002). The operated embryos were recorded photographically under fluorescent illumination using an Axiocam camera (Carl Zeiss Vision; München-Hallbergmoos), immediately after the graft or injection. The graft was detected by the green/red (DiI/DiO) fluorescent signal of the carbocyanines, respectively (Hatada and Stern, 1994). The embryos were cultured further until they reached stages HH10–14 and were then fixed overnight in cold 4% paraformaldehyde (phosphate buffered pH 7.4, 0.1M, 4◦C).

#### **PHOTO-OXIDATION OF CARBOCYANINE-DYE-LABELED CELLS**

To visualize permanently the dye-labeled cells, the fluorescence was photo-converted to an insoluble diaminobenzidine (DAB) by photo-oxidation (described by Selleck and Stern, 1991). The embryos were removed from the PFA and incubated with DAB solution (DAB in 0.1M Tris pH 7.4) in the dark, at room temperature, for 1 h. Each specimen was placed in a fairly deep glass cavity slide, covered with a coverslip. Under microscope epifluorescence the regions containing labeled cells were illuminated until all fluorescence disappears. After that, most operated specimens were systematically processed for *in situ* hybridization (ISH) with one of the mRNA probes described below.

#### **IN SITU HYBRIDIZATION (ISH)**

The embryos were hybridized following the protocol described by Streit and Stern (2001), Ferran et al. (in press). For this study we normally used digoxigenin-uridin triphosphate (UTP) labeled antisense chicken riboprobes. The standard visualization procedure with nitro blue tetrazolium (NBT)/5-bromo-4 chloro-3<sup>0</sup> -indolyphosphate p-toluidine salt (BCIP) solution as chromogenic alkaline phosphatase substrate gave a dark blue reaction product for digoxigenin-UTP. The gene markers analyzed were *Pitx2*, *Shh*, *Fgf8*, *Raldh3*, *Tcf4*, *Cytoqueratin-8*, *Hesx1, Dlx3, Pax6* and *Six3*.

**Table 2 | Correlation of medio-lateral and dorso-ventral extent of derived graft domains with boundaries of gene expressions in the extra and neural areas.**


The relative amount of labeled cells filling the respective AP (medial-lateral) or DV (dorsoventral) territories is estimated roughly as "few cells" (+), "up to half the domain is labeled" (++) and "most or all the domain is labeled" (+++).

labeled partially the anterior neural border and/or the ADH field. **(B)** The immediate bright field whole-mount view of case RPA-09 shows that the DiI injection was located within the neural plate, ranging between the prospective preoptic telencephalon and the more ventral basal hypothalamus. **(C)** At stage HH10, the embryo was whole-mount-

#### **IMMUNOHISTOCHEMITRY (IHC) AND TISSUE PROCESSING**

Cells derived from the CFSE-labeled grafts were visualized after the ISH reaction. The embryos were immunoreacted with anti-fluorescein Fab fragments conjugated to horseradish peroxidase (anti-fluorescein-peroxidase, horseradish (POD); 1:500; Boehringer; Mannheim) following standard protocols (Fernández-Garre et al., 2002; Ferran et al., in press). Afterwards the embryos were cryoprotected overnight in 10% sucrose in PBS, and embedded in 10% gelatin and 10% sucrose in PBS. The blocks were cryostat-sectioned 10–12 µm-thick in the sagittal plane and mounted with Mowiol.

#### **IMAGING**

Images were captured with an Axiocam digital camera (Carl Zeiss Vision; München-Hallbergmoos). Digital images were processed with Photoshop® CS4 11.0.2 and ImageJ (Fiji) software. Representative images were used to draw vectorial schemata with Adobe Illustrator CS4.

As previously described (Streit, 2002), embryos subjected to ISH shrink significantly when heated in the presence of formamide and detergent. Two approaches were used to

neural tube. In mid-sagittal sections, the injection-derived cells partially overlap the Shh positive cells of the hypothalamic basal plate. Laterally, positive cells appear in alar hypothalamic and telencephalic domains, where Shh expression is absent. Scale bars: 250 µm in **(B,C)**; 125 µm in **(D–F)**.

overcome this problem when comparing the position of DiIlabeled cells. In some embryos the images obtained from *in situ* hybridized embryos were increased by 10% and aligned to landmarks on the fluorescence image to produce a montage. In other embryos, the DiI fluorescence was photoconverted with DAB (Izpisúa-Belmonte et al., 1993) before ISH. This approach allows DiI labeled cells to be seen directly in the whole-mount *in situ* processed embryo and confirms the assumptions made by adjusting the magnification of separately obtained images, as described above (Sanchez-Arrones et al., 2012).

#### **RESULTS AND DISCUSSION**

#### **EXPERIMENTAL FATE MAPPING AND EARLY MORPHOGENETIC TRACING OF THE PROSPECTIVE ADH**

During chick development, the ADH placode (presumptive Rathke's pouch) becomes morphologically identifiable as a thickened patch of epithelium at stages HH12–14 (Romanoff, 1960). At this moment the ADH is molecularly segregated from the adjacent olfactory placode, both being ectodermal placodal specializations that lie outside the neural tube. In order to know precisely the position of the prospective

ADH primordium relative to the neural telencephalic and neurohypophyseal primordia at earlier neural plate stages, when the ADH is not histologically distinct, we performed fatemapping experiments combined with mappings of neural and non-neural maker genes. To this end, the anterior midline ectoderm was labeled systematically using small fluorescent grafts or small single injections of the fluorescent dyes DiI or DiO at stage HH4/5 at different dorsoventral levels (the node represents the ventralmost median position and the median non-neural ectoderm lies topologically dorsal to the ANR (Fernández-Garre et al., 2002; Sanchez-Arrones et al., 2009). The derived domains of these grafts and injections were examined in comparison to a selected gene marker between stages HH10 and HH12, when the placodal cells start to be morphologically distinguishable (thickened and invaginated) from the adjacent ectoderm. The experimental cases selected for this analysis are illustrated in the **Figures 1**–**4**. These show schemata which respectively locate the set of grafts and injections grouped at the midline which produced derivatives at either the anterior *median neural domain* (preoptic telencephalon and hypothalamus; **Figure 1**), the anterior *neural border domain* (ANR; **Figure 2**), anterior median *pre-placodal domain* (ADH; **Figure 3**) and anterior nonplacodal *ectoderm* and *endoderm* (**Figure 4**). The size, distance from the node, and angular radial position of the grafts and injections relative to the nodal median radius appears listed in **Table 1**. We will describe below six representative cases out of a total of 23, ordered according to their progressively more dorsal fates; neural fate (RPA-09), ANR (RPA-08 and RPA-30), anterior median pre-placodal domain (RPA-27, RPA-02

and RPA-80) and anterior non-placodal ectoderm/endoderm (RPA-06).

#### **Anterior median neural cells**

In the case RPA-09, a DiI injection was placed across the prospective alar/basal boundary of the anterior neural plate (anp), roughly 170 µm distant from the node (**Figures 1A,B**, compare (Sanchez-Arrones et al., 2009; **Table 1**). At stage HH10, labeled cells derived from the injection appeared precisely across the molecular limit separating the alar and basal plates (bp) of the secondary forebrain (**Figure 1C**). This specimen was processed to detect *Shh* signal, which represents a gene expressed selectively at this stage throughout the floor and bp of the secondary prosencephalon (SP; Bardet et al., 2010). Some *Shh*-positive cells overlapped with brown-labeled cells derived from the injection (implying the latter fell partially upon prospective basal hypothalamus), whereas other injection derivatives overlapped with more dorsal *Shh*-negative cells located in the median alar SP (prospective alar hypothalamus and/or preoptic subpallial telencephalon (**Figures 1D–F**; see Cobos et al., 2001; Bardet et al., 2010)).

#### **Anterior neural border cells**

In the case RPA-08, a DiI injection targeting the ANR was placed more dorsally across the border between neural and nonneural (preplacodal) domains at a distance of 286 µm from the node (**Figures 2A,B**; see Sanchez-Arrones et al., 2012). At stage HH11, the DiI-labeled cells were located at the midline of the prospective septal roof plate of the subpallial telencephalon and

**FIGURE 3 | Adenohypophyseal fate: analysis of the experimental cases RPA-27, RPA-02 and RPA-80. Initial median cell movements. (A)** Map of all studied grafts or injections corresponding with partial adenohypophyseal fate. **(B,D,F,H)** Combined fluorescent and bright field images of representative HH4 chick embryos in which a graft **(B)** or either a DiI **(D)** or DiO injection **(F,H**) was placed at the median non-neural ectodermal region. **(C,C')** At stage HH11, the graft-derived cells were located in the rostromedian non-neural ectoderm, largely coinciding with the Fgf8-negative Rathke's pouch rudiment. **(E,E')** The prospective Rathke's pouch tissue, flipped over into external

contact with the neural terminal wall, already has elongated to the level of the Shh-positive prospective hypothalamic basal plate. **(F,G)** In case RPA-80, the labeled non-neural cells of the ADH primordium clearly appeared at the medial head ectoderm, within the space that separates the bilateral Raldh3-positive olfactory placodes, whose cells remained unlabeled. **(I,J)** Simple fluorescent and combined fluorescent and bright field images of the case illustrated in **(H)**, counterstained with Dlx5 whole-mount ISH, showing the elongated median labeled trace of ectodermal tissue extending from the (Continued)

#### **FIGURE 3 | Continued**

preplacodal field (ppf) into the ventrally displaced ADH primordium, at stage HH6. Arrowheads in **(I)** mark the site of the original injection (see also Sanchez-Arrones et al., 2012). Scale bar: 250 µm in **(B,C)**; 125 µm in **(C')**.

also formed a thin stripe extending along the midline of the ADH placode (Rathke's pouch) and neighboring head ectoderm (**Figure 2C**). In other experimental embryos of this group, such as case RPA-30, in which a fluorescent graft was inserted at the rostral neural border approximately 273 µm distant from the node (**Figures 2A,D**), slightly laterally to the RPA-08 injection, the grafted cells appeared at stage HH12 (closed neural tube) exclusively at the prospective median telencephalon, represented by the *Fgf8*-expressing ANR area of the presumptive septum and the preoptic area, without labeling the ADH primordium (**Figure 2E**; see Cobos et al., 2001). Some labeled (grafted) cells were observed as well at the optic stalk (alar hypothalamus), which is also *Fgf8* positive (**Figure 2E**; Crossley et al., 2001). The positional similarity of the experiments producing both neural and placodal derivatives vs. those producing only neural derivatives suggests close vicinity between both domains across the neural/non-neural border.

#### **Anterior preplacodal domain; prospective ADH precursors**

The ADH precursors were labeled specifically in the next three cases: RPA-27 RPA-02 and RPA-80. In case RPA-27, a graft was placed more dorsally along the midline ectoderm, at a 260 µm distance from the node (**Figure 3B**), whereas in RPA-02 a DiI injection was introduced at 280 µm from the node (**Figure 3D**). In the whole-mounted embryos photographed at stage HH11, the graft-derived cells were found at the rostral midline ectoderm site which contains the thickened Rathke's pouch (see Cajal et al., 2012); the latter and the surrounding rostral ectoderm do not express *Fgf8* (**Figures 3C,C'**). Rathke's pouch, already occupying the stomodeal roof, lies below the prechordal plate (pp), which in its turn underlies the prospective hypothalamic bp (*Shh*-positive domain; **Figures 3E,E'**; García-Calero et al., 2008). The grafted/injected ADH areas showed intercalation of labeled cells with unlabeled ones, suggesting that a longitudinal intercalation mechanism is involved in the early morphogenesis of the ADH placode (see also Sanchez-Arrones et al., 2009). Laterally, further cells derived from the graft extended into the rostral head ectoderm (shown in the RPA-02; **Figures 3E,E'**), but not into the olfactory placode (Raldh3-positive cells; shown in the RPA-80 case; **Figure 3G**).

Consistently with this idea, we recently reported that a longitudinal median intercalation mechanism also obtains during cell movements of the anterior proneural zone at neural plate stages. In fact, the median cells are compactly stretched across the entire ectodermal midline, including the neural and non-neural territories (shown in **Figures 3H–J**; see also Sanchez-Arrones et al., 2012); this suggests that midline cells display a differential intercalative behavior compared with nearby more lateral (caudal) cells.

#### **Anterior ectoderm and endoderm**

In case RPA-06, a single DiO injection (green fluorescence) was placed at HH5 at the rostromedian non-neural ectoderm and underlying rostral endoderm (**Figures 4A,B**; see Stalsberg and DeHaan, 1968; Kimura et al., 2006). The embryo survived until HH12, and the resulting fluorescent-labeled cells were photooxidated, while the specimen was hybridized for *Shh* (**Figure 4C**). Most labeled cells appear located in or superficial to the rostral digestive tube (dt; **Figure 4D**. The *Shh* expression is visible at the prospective basal and floor plates of the SP (hypothalamus). Axial mesodermal tissues including the pp and the notochord are also positive for *Shh* (Withington et al., 2001; García-Calero et al., 2008). Some DiO-labeled cells were detected in the head ectoderm located in front of the pp (**Figure 4D**).

These results indicate that the fates of ectodermal cells occupying the rostral midline are well segregated at early neural plate stages. The ADH precursors occupy strictly extraneural positions (**Figure 7**). These cells separate the more dorsal midline ectoderm (prospective oral and head ectoderm) from the anterior neural border cells (ANR; prospective roof and median alar subpallial domains of the telencephalon). The latter territories are relatively distant from the prospective NH, located much more ventrally at neural plate stages, within the prospective basal hypothalamus. Once the neural tube is closed and the telencephalic vesicles start to emerge bilaterally, the ADH precursors restricted to the extra-neural midline (as well as the underlying rostral endoderm) are bent by the cephalic fold as by a hinge into the stomodeal roof. Both the preopto-hypothalamic acroterminal midline and the median cephalic ectoderm are stretched considerably by the growth of the head and associated cell intercalation phenomena (Sanchez-Arrones et al., 2009, 2012). The ADH primordium thus results separated from the median telencephalon and develops a new close relationship with the basal hypothalamus (prospective NH), thanks to the vacation of the space previously occupied by the prechordal mesoderm.

#### **MOLECULAR TRACING OF ADENOHYPOPHYSIAL DEVELOPMENT**

The development of the pituitary gland is a multistep morphogenetic process controlled by a genetic program (Sheng et al., 1997; Treier et al., 1998). Though we know various molecular traits involved in specification and differentiation of the different cell types of the ADH (Sheng and Westphal, 1999), little is known about the molecular patterns activated during the earliest stages of ADH development. Assuming a non-neural placodal nature of the ADH primordium (see results above, and references cited in the Introduction), our next aim was to examine the molecular signals that first specify differentially the rostral pre-placodal ectodermal region vs. the anterior neural ectoderm at neural plate stages; secondly, we wanted to study the initial regionalization of the presumptive preplacodal territory into ADH vs. olfactory placodes at early neural tube stages, during the formation of Rathke's pouch; finally, our attention centered on the onset of ADH molecular regionalization into the anterior and intermediate pituitary gland lobes.

#### **Early ADH development: paraneural pre-placodal primordium from HH4 to HH12**

In order to visualize the potential ADH primordium we mapped the *Pitx2* transcription factor. This gene is involved in head mesoderm patterning (Bothe et al., 2011), and in the establishment of embryonic left-right asymmetry and the fate of precardiac cells during cardiogenesis (García-Castro et al., 2000; Lopez-Sanchez et al., 2009). It also plays a key role during

**FIGURE 5 | Early gene expression at the ADH primordium**. Neural plate and neural tube whole mounts hybridized with Pitx2, Dlx3, Hesx1 and Raldh3. The images are shown in a ventral view. The node is marked by an asterisk in some of the images. Scale bar: 250µm.

ADH development at neural tube stages, correlative with the emerging Rathke's pouch (chick, Sjödal and Gunhaga, 2008; Parkinson et al., 2010; Xenopus, Schweickert et al., 2001; mouse, Drouin et al., 1998; Suh et al., 2002). Other *Pitx* family members as *Pitx1* and *Pitx3* were well characterized during early ADH development, at pre-placodal stages (mouse, Lanctôt et al., 1997; zebrafish, Dutta et al., 2005); however, the early *Pitx2* expression was not yet addressed. We thus examined *Pitx2* expression by ISH in whole-mount chick embryos and cross-sections, comparing with other genes known to be involved in pre-placodal/placodal specification, such as *Dlx3* (expressed in non-neural ectoderm; Dutta et al., 2005; Khudyakov and Bronner-Fraser, 2009), *Hesx1* (an ADH placode marker; Hermesz et al., 1996), and *Raldh3* (an olfactory placode marker; Sabado et al., 2012; **Figure 5**). Initially, at HH4, *Pitx2* signal was only present in the head mesenchyme (**Figure 5A**). Shortly afterwards, at HH6+/7-HH8, when the anterior neural border becomes sharp (Sanchez-Arrones et al., 2012), an ectodermal *Pitx2*-positive domain was detected in the rostral head fold, which coincides with the *Dlx3*-positive paraneural pre-placodal area (**Figures 5B,C'**); a similar result was reported during early placodal differentiation in mouse embryos and zebrafish larvae. At early closed neural tube stages, HH10-HH11, this paraneural *Pitx2* signal overlaps the expression domain of the placodal marker *Hesx1*, precisely in the thicker median placodal cells, which are held to represent the adenohypophyseal placode (**Figures 5D,E**; see also Sjödal and Gunhaga, 2008). *Raldh3*-positive cells were detected instead bilaterally in the prospective olfactory placode cells and the surrounding ectoderm, whereas *Raldh3* expression was completely absent at the ADH placode (**Figure 5F**). Accordingly, at stages HH10–11 the *Dlx3/Pitx2*-expressing primary paraneural pre-placodal ectodermal domain becomes regionalized in cellular and molecular terms, forming rostrally the median *Pitx2/Hesx1* positive ADH placode and bilaterally the *Raldh3*-positive olfactory placodes (Bailey et al., 2006).

#### **Later development of the ADH: Rathke pouch rudiment from HH14 to HH18**

The developing pre-placodal domain subsequently undergoes changes in cell shape that drive the histologic differentiation of the placodes. The first change observed is a thickening of the placodal cells, which is generally accompanied by incipient invagination. The ADH placode starts to transform into a recognizable Rathke's pouch during stage HH12 (Romanoff, 1960), when the neural tube is bending at the cephalic flexure. Parallel growth of the cephalic fold, with rostroventral protrusion of the embryonic head and intercalative elongation of the median preopto-opto-hypothalamic part of the neural plate, causes the oro-pharyngeal plate (originally formed rostral to the pre-placodal ectodermal band) to be internalized into the depth of the growing stomodeum. The median part of the placodal band sharply hinges under the closing rostral neuropore into the stomodeal roof, thus inverting its apico-basal orientation as it also undergoes intercalative elongation (**Figure 7**). The median stomodeal cells derived from the primary ADH placodal domain (prospective Rathke's pouch) thus result placed underneath the tuberal hypothalamus (prospective neurohypophyseal primordium), though the two primordia are still mutually separated by the respective basal membranes and interposed prechordal mesoderm. Closure of the rostral neuropore and differential morphogenesis of the non-median tissues (mainly the evaginating telencephalic and eye vesicles, plus the olfactory placodes) increasingly separates the stomodeal ADH primordium from its earlier telencephalic neighbors (by stretching and growth of the interposed non-neural ectoderm), and the fused ANR transforms into the neural commissural septal plate (Puelles et al., 1987; Cobos et al., 2001). By stage HH14 the invaginated ADH primordium progressively forms the roof of Rathke's pouch under the median tubero-infundibular forebrain area. From this stage onwards the ADH cells lie progressively closer to the presumptive NH (though the evaginated NH organ only emerges later, at stage HH26 in chick embryos), and molecular regionalization of the ADH primordium starts, as the stalk of Rathke's pouch degenerates and disappears.

We illustrate in **Figure 6** the correlative changes of some molecules involved in ADH and NH development, namely

**FIGURE 7 | Schematic summary of proposed morphogenetic events leading to the development of the adenohypophysis across neural plate and neural tube stages**. These schemata each represent in general the changing sagittal midline of chick embryos at different developmental stages. The dashed lines delimit the main rostrocaudal subdivisions of the neural tube. **(HH4)** The initial schema illustrates a median section through the early neural plate, marking tissue corresponding to hypoblast (black), rostral endoderm (pale gray) and neural plate or tube (gray). The prospective ADH, representing median non-neural ectoderm, is coded in deep orange. The deduced flat fate map of the median anterior head domain is illustrated in the insert at the side, showing in deep red the labeled sites identifying by fate the ADH field, in contrast to rostral neuroectoderm sites (gray field) and non-ADH head ectoderm sites (light red). **(HH6)** This schema is similar to the HH4 one, though we distinguish now the endodermal primordium of the anterior intestinal portal (yellow) and the notochordal tissue (deep blue). **(HH8)** The head fold develops,

causing the ADH field to flip over in a hinge-like motion under the anterior intestinal portal, though it still remains transiently attached by head ectoderm to the anterior neural ridge (rostral neuropore still open). **(HH11)** Neurulation is nearly finished at this stage; we identify the head ectoderm (weak orange) and the prechordal plate mesoderm (weak blue) at the tip of the notochord (the latter lies under the prospective mammillary pouch); note the prechordal tissue transiently separates the ADH rudiment from the basal hypothalamus. **(HH14,HH16)** The prechordal tissue gradually migrates dorsalward in front of the terminal hypothalamic wall, allowing the stomodeum to form and the ADH to approach the hypothalamus. **(HH18)** The prechordal cells have migrated away, being now in contact with the alar median hypothalamus (chiasmatic region) and the prospective preoptic telencephalon (terminal lamina); they leave the basal tuberal hypothalamus free for close contact with the ADH primordium; this rostroventral tuberal territory includes the prospective neurohypophysis (NH), whose evagination is first observable several stages later, at HH26.

the markers with *Tcf4, Pitx2, Shh*, *Cytoqueratin-8, Pax6, Raldh3* and *Six3*. At HH14, the oral stomodeal ectoderm containing Rathke's pouch is distinguished molecularly by selective expression of *Tcf4*, *Pitx2* and *Shh* (**Figures 6A–C**; Sjödal and Gunhaga, 2008). At HH16-18, the ADH pouch underlies the NH primordium and the prospective tuberal region (see sagittal sections, **Figures 6E–I**; see also Bardet et al., 2010), where *Pax6* and *Cytokeratin-8* are expressed (**Figures 6D,G,J**). Furthermore, we noted the onset of ADH regionalization into two domains, the future anterior and intermediate lobes (Sheng and Westphal, 1999; Rizzoti and Lovell-Badge, 2005; Reyes et al., 2008). Whereas *Tcf4* and *Pitx2* signals delineated the presumptive anterior part of the ADH (**Figures 6E–G,J**), *Six3* and *Raldh3* expression appears restricted to the future intermediate ADH lobe, which underlies the presumptive NH or posterior pituitary lobe (*Six3* and *Shh*-positive domain; **Figures 6H–J**).

These data jointly suggest that *Pitx2* is an early preplacodal marker that intervenes during molecular regionalization of the rostral preplacodal domain into olfactory and ADH placodes. The ADH rudiment is later partitioned molecularly into two subdomains, the prospective intermediate and anterior lobes, which lie underneath the prospective NH or posterior lobe, contained for a while within the tubero-infundibular hypothalamic wall.

#### **CONCLUSION**

Our analysis corroborates the median extraneural locus of the ectodermal ADH primordium at HH4, coinciding with the molecularly distinct median preplacodal domain. Data were obtained indicating that this midline locus is involved in particularly strong dorsoventral intercalative movements, oriented orthogonally to the neural/nonneural boundary. Such movements apparently represent part of the forces that quickly separate the ADH primordium from the ANR (as observed already at HH6; see **Figures 3I,J**), complementing other forces derived from surrounding tissues involved in head fold development, forebrain neurulation and brain and eye growth. At the end of median intercalation (e.g., HH8), the ectodermal ADH anlage appears hinged downwards under the head fold, and it already occupies a position close to its subsequent apposition to the basal hypothalamus. However, this contact only occurs somewhat later (HH12), after the initially intervening prechordal plate cells move past the median basal hypothalamus into more dorsal neighborhoods (**Figure 7**).

#### **ACKNOWLEDGMENTS**

We are grateful to P. Bovolenta, S. Martinez, H. Nakamura, J. Rubenstein, A. Simeone, and M. Studer for kindly provided cDNA clones. This work was supported by the Spanish Ministry of Economy and Competitiveness grant BFU2008-04156 and the SENECA Foundation 04548/GERM/06 (no. 10891) to Luis Puelles.

#### **REFERENCES**


ectoderm cells in early neural development. *Development* 141, 4127–4138. doi: 10.1242/dev.107425


and pituitary development. *Mech. Dev.* 107, 191–194. doi: 10.1016/s0925- 4773(01)00461-0


**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 27 November 2014; paper pending published: 15 December 2014; accepted: 12 January 2015; published online: 17 February 2015*.

*Citation: Sánchez-Arrones L, Ferrán JL, Hidalgo-Sanchez M and Puelles L (2015) Origin and early development of the chicken adenohypophysis. Front. Neuroanat. 9:7. doi: 10.3389/fnana.2015.00007*

*This article was submitted to the journal Frontiers in Neuroanatomy*.

*Copyright © 2015 Sánchez-Arrones, Ferrán, Hidalgo-Sanchez and Puelles. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms*.

# Cadherins mediate sequential roles through a hierarchy of mechanisms in the developing mammillary body

Nora-Emöke Szabó<sup>1</sup> , Roberta Haddad-Tóvolli <sup>2</sup> , Xunlei Zhou<sup>2</sup> \* and Gonzalo Alvarez-Bolado<sup>2</sup> \*

<sup>1</sup> Department Neurobiology and Development, Neural Circuit Development Unit, IRCM, Montréal, QC, Canada, <sup>2</sup> Department of Neuroanatomy, University of Heidelberg, Heidelberg, Germany

Expression of intricate combinations of cadherins (a family of adhesive membrane proteins) is common in the developing central nervous system. On this basis, a combinatorial cadherin code has long been proposed to underlie neuronal sorting and to be ultimately responsible for the layers, columns and nuclei of the brain. However, experimental proof of this particular function of cadherins has proven difficult to obtain and the question is still not clear. Alternatively, non-specific, non-combinatorial, purely quantitative adhesive differentials have been proposed to explain neuronal sorting in the brain. Do cadherin combinations underlie brain cytoarchitecture? We approached this question using as model a well-defined forebrain nucleus, the mammillary body (MBO), which shows strong, homogeneous expression of one single cadherin (Cdh11) and patterned, combinatorial expression of Cdh6, −8 and −10. We found that, besides the known combinatorial Cdh pattern, MBO cells are organized into a second, nonoverlapping pattern grouping neurons with the same date of neurogenesis. We report that, in the Foxb1 mouse mutant, Cdh11 expression fails to be maintained during MBO development. This disrupted the combination-based as well as the birthdate-based sorting in the mutant MBO. In utero RNA interference (RNAi) experiments knocking down Cdh11 in MBO-fated migrating neurons at one specific age showed that Cdh11 expression is required for chronological entrance in the MBO. Our results suggest that neuronal sorting in the developing MBO is caused by adhesion-based, noncombinatorial mechanisms that keep neurons sorted according to birthdate information (possibly matching them to target neurons chronologically sorted in the same manner). Non-specific adhesion mechanisms would also prevent cadherin combinations from altering the birthdate-based sorting. Cadherin combinations would presumably act later to support specific synaptogenesis through specific axonal fasciculation and final target recognition.

#### Keywords: neuronal birthdates, cell sorting, combinatorial, differential adhesion, mamillary body

### Introduction

The mammalian brain is formed by a large variety of neuronal aggregates organized as layers, nuclei and subnuclei. The diversity of forms found in animal tissues is considered to be largely the result of conserved morphogenetic processes and mechanisms (Lecuit, 2008). If and how

#### Edited by:

Agustín González, Universida Complutense de Madrid, Spain

#### Reviewed by:

Herbert Hildebrandt, Hannover Medical School, Germany Irina Georgievna Makarenko, Russian Academy of Sciences, Russia

#### \*Correspondence:

Xunlei Zhou and Gonzalo Alvarez-Bolado, Department of Neuroanatomy, University of Heidelberg, Im Neuenheimer Feld 307 69120, Heidelberg, Germany xzhou@ana.uni-heidelberg.de; alvarez@ana.uni-heidelberg.de

Received: 26 January 2015 Paper pending published: 13 February 2015 Accepted: 25 February 2015 Published: 19 March 2015

#### Citation:

Szabó N-E, Haddad-Tóvolli R, Zhou X and Alvarez-Bolado G (2015) Cadherins mediate sequential roles through a hierarchy of mechanisms in the developing mammillary body. Front. Neuroanat. 9:29. doi: 10.3389/fnana.2015.00029 these underlie brain histogenesis is not well understood. Differential cell-cell adhesive interactions are essential drivers of morphogenesis (Edelman, 1988). Classical cadherins are transmembrane proteins mediating cell-cell adhesion with roles in cell sorting and in axonal connectivity (Takeichi, 2007). The intriguing combinatorial cadherin expression patterns in brain regions (see for instance (Hertel et al., 2008, 2012; Krishna-K et al., 2011)) have been proposed (Redies and Takeichi, 1996) to underlie the sorting of specific neuronal subpopulations. As an additional function, a combinatorial mechanism underlying appropriate connectivity/synaptogenesis has been suggested (Suzuki et al., 1997; Bekirov et al., 2002; Treubert-Zimmermann et al., 2002) since, in some systems, projecting neurons express the same cadherin combinations as their targets.

If combinations of cadherins confer adhesion specificity (or synaptic specificity), homophilic adhesion (e.g., Cdh11 would bind only, specifically, to Cdh11) would be indispensable. Only in that way could combinations specifically recognize each other. Data from a variety of experimental systems has proven the importance of homophilic binding of one cadherin (not a combination) in morphogenesis (Gumbiner, 2005; Suzuki and Takeichi, 2008), axonal fasciculation (Treubert-Zimmermann et al., 2002), synapse formation (Manabe et al., 2000; Elia et al., 2006; Paradis et al., 2007; Suzuki et al., 2007) and guidance of migrating neurons (Luo et al., 2004). The role of cadherin combinations in neuronal sorting has been experimentally proven in chicken hindbrain motoneurons (Astick et al., 2014). Still, that cadherin combinations form a specific code underlying brain histogenesis is far from clear.

To complicate things, the study of cell sorting phenomena in tissue aggregates in vitro suggests an additional, nonmolecularly-specific source of histogenetic order. This consists of physical forces like the surface tension of cell aggregates, resulting from the ratio between adhesion and cortical tension (Steinberg, 1962a,b,c; Krieg et al., 2008; Manning et al., 2010). Indeed, nonspecific adhesion differentials can mediate cadherin-dependent cell sorting in culture (Steinberg and Takeichi, 1994; Duguay et al., 2003) and determine the antero-posterior body axis of the Drosophila embryo (Godt and Tepass, 1998; González-Reyes and St Johnston, 1998). This paradigm presupposes heterophilic binding and, consistently, cadherins exhibit actually little binding specificity (Shimoyama et al., 1999, 2000; Niessen and Gumbiner, 2002; Foty and Steinberg, 2005; Prakasam et al., 2006; Krieg et al., 2008; Shi et al., 2008). However, if this paradigm can be applied at all to migrating neurons in the developing brain is open to question, and the possible role of non-specific adhesion forces in brain histogenesis has to our knowledge never been approached.

In summary, the questions of the actual role of the intricate cadherin combinations in brain cell sorting, and the relative importance of specific (homophilic) vs. non-specific (heterophilic) mechanisms are still mysterious.

Here we have tested in the developing mouse brain in utero the role of cadherins on neuronal aggregation. Our model is the developing mammillary body (MBO), a large, compact and well-delimited paired neuronal structure with

FIGURE 1 | Age of neurogenesis and Cadherin expression do not match. (A) The MBO in a transverse section of E18.5 mouse brain labeled with in situ hybridization for Cdh11. Inset: position of the MBO on a sagittal diagram of the mouse brain, rostral to the left. (B--E) In situ hybridization (ISH) for cadherins (as indicated) on transverse sections of E18.5 MBO. (F) Summary diagram of cadherin expression in the MBO after the data in (B--E). (G--K) Diagrams of transverse sections through E18.5 MBO labeled with anti-BrdU antibody after BrdU injection at E9.5 (G), E10.5 (H), E11.5 (I), E12.5 (J) and E13.5 (K). (L) Summary diagrams of BrdU-labeled cells corresponding to the data in (G--K).

defined functions (Vann and Aggleton, 2004) located in the hypothalamus (**Figure 1A**) and showing ubiquitous expression of Cdh11 and patterned expression of Cdh6, 8, and 10 (Kimura et al., 1996; Suzuki et al., 1997). Each MBO is medio-laterally subdivided into medial and lateral mammillary nuclei (Allen and Hopkins, 1988). We first explored the relation between neuronal birthdate and specific cadherin expression in MBO neurons. Then we analyzed cell sorting upon loss of Cdh11 expression over the entire MBO during development. Finally, we used in utero electroporation and RNAi to reduce Cdh11 expression in all MBO neurons born at a certain specific age and analyzed their position several days later.

Our results suggest that neuronal sorting inside brain nuclei is caused by adhesion-based, non-combinatorial mechanisms that keep neurons sorted according to birthdate information matching them to target neurons chronologically sorted in the same manner. Non-specific adhesion mechanisms would also prevent cadherin combinations from altering the birthdatebased sorting. Cadherin combinations would presumably act later to support specific synaptogenesis through specific axonal fasciculation and final target recognition.

### Materials and Methods

#### Mice

Animals were housed and handled in ways that minimize pain and discomfort, in accordance with German animal welfare regulations (TierSchG) and in agreement with the European Communities Council Directive (2010/63/EU). The authorization for the experiments, including in utero electroporation, was granted by the Regierungspräsidium Karlsruhe (state authorities) and the experiments were performed under surveillance of the Animal Welfare Officer responsible for the Institute of Anatomy and Cell Biology. To obtain embryos, timed-pregnant females were sacrificed by cervical dislocation; the embryos were decapitated.

Wild type observations and electroporation experiments were carried out on C57BL/6 mice. Additionally, two mouse lines carrying null mutations of Foxb1 were used, the Foxb1 tauLacZ (Alvarez-Bolado et al., 2000a), with beta-galactosidase as reporter, and the Foxb1-Cre-GFP (Zhao et al., 2007), with green fluorescent protein (GFP) as reporter. By crossing heterozygotes of both lines, Foxb1 homozygous mice were generated carrying one b-galactosidase-expressing Foxb1 null allele and one GFP-expressing Foxb1 null allele. In this way, the homozygotes as well as half of the heterozygotes carried only one b-galactosidase-expressing allele and so the intensity of beta-galactosidase expression could be compared between them in order to evaluate the size and shape of the MBO (see below).

#### Size and Shape Measurements of the MBO

The brains of E18.5 homozygotes and beta-galactosidaseexpressing heterozygote embryos (see above) were collected (three brains per age and genotype), embedded in agarose and cut sagittally with a vibration microtome into 100 µm thick sections. The sections were stained with the X-gal reaction (Zhao et al., 2007), then fixed and photographed. The sections were assigned to one of four medio-lateral regions of the MBO, and the section area (in arbitrary units) labeled by the X-gal reaction in the mammillary region was measured with Cell-F software (Olympus Soft Imaging Solutions, Münster, Germany). The combined section areas for every mediolateral region were used as a proxy for the size of the region.

#### Cell Density Measurement in the MBO

Twenty five µm thick sections of E18.5 Foxb1-Cre-GFP homozygous and heterozygous brains were labeled with the nuclear marker 4',6-diamidino-2-phenylindole (DAPI) as well as an anti-GFP antibody to specifically stain the MBO. Two square regions (100 µm side) were defined in the medial and in lateral part of the MBO and the number of cells in each of them was counted by the optical dissector method (Coggeshall and Lekan, 1996).

#### In Utero Electroporation

We have described the procedure in detail elsewhere (Haddad-Tóvolli et al., 2013). Timed-pregnant (E12.5) mice were anesthetized and the uterus surgically exposed. Plasmid encoding small hairpin RNAs (shRNA) (1,5 µg/µl) (see below) was mixed with pCAGGS-GFP reporter vector (0,8 µg/µl), and approximately 1 µl of this DNA mixture was injected with a pulled micropipette into the third ventricle of each embryonic brain. Five pulses of square-wave current were applied (50 V, 50 ms on, 950 ms off) to each injected embryonic brain using a CUY21EDIT electroporator (Nepagene), and the pregnant mice were allowed to recover. The embryo brains were collected at E18.5 and those showing strong fluorescence in the mammillary region were prepared for further analysis. Some brains were fixed in 4% paraformaldehyde for 1--2 h at RT, embedded in gelatine-albumin and cut into 100--200 µm thick sections. The sections were then analyzed under a fluorescent microscope. Some brains were cryostat-sectioned at 20 µm for immunohistochemistry.

#### Immunohistochemistry

We followed a published protocol (Szabó et al., 2011) on paraffin sections (15 µm). We used the following antibodies: anti-Cadherin11 (1:80) (monoclonal, Zytomed), anti-GFP (1:1000) (rabbit polyclonal, Invitrogen) and (1:500) (rabbit polyclonal, Abcam), anti-beta Galactosidase (1:500) (polyclonal, Abcam), anti-nestin (1:200) (monoclonal, Chemicon), anti-2H3 (1:5) (Developmental studies Hybridoma bank, monoclonal). Then we photographed the results with a Leica TCS SP5 confocal microscope.

#### RNA Interference Plasmids

DNA plasmids encoding shRNA designed to interfere with Cdh11 mRNA were purchased from Sigma (NM\_009866). The following three were tried in culture:


Successful interference (see below) was obtained with shRNA-3.

#### Quantitative PCR Control of RNAi in Culture

HEK293T cells were plated (200,000 cells per 3.5 cm well). After 24 h in culture they reached 50% confluence and were transfected with one of the shRNA plasmids (either shRNA-2, -3 or -4, see above) plus a ''target and control'' plasmid carrying CAG promoter---mCdh11 cDNA--IRES--EGFP--poly A--SV40 promoter--neomycine phosphotransferase II (neo)--poly A. A total of 2 µg of DNA per well were transfected (1.8 µg of shRNA plasmid plus 0.2 µg of ''target and control'' plasmid). Forty eight hours after transfection RNA was extracted, treated with DNAse I and reverse transcribed with the Superscript kit (Invitrogen) (2 µg RNA per reaction). The RNA was quantitated by PCR (StepOne Plus, Applied Biosystems) using the neo transcript to normalize. The transfections were done in triplicate and the quantitative RT-PCR was repeated three times per transfection.

### RNAi Complementation ("Rescue") Experiments

A complementation construct was cloned carrying a human CDH11 cDNA and the GFP reporter under the control of the CAG promoter (Niwa et al., 1991; **Figures 11A,B**). We performed this deletion on human CDH11 cDNA, whose nucleotide sequence is not 100% identical with the mouse Cdh11, to maximize the probability of the complementation construct not to be recognized by the shRNA.

To make this CDH11 immune to RNAi by the shRNA-3, the ''seed sequence'' (Lai, 2002; Lewis et al., 2003), required for target recognition by the shRNA-3 and subsequent degradation was deleted (**Figures 11A,B**). The deleted seed sequence encodes three amino acids in the extracellular domain EC5 in principle not involved in the adhesive or signaling function (Leckband and Prakasam, 2006; Ciatto et al., 2010; Harrison et al., 2010) of cadherins. The complementation construct was mixed with shRNA-3 construct (1:1) and then transfected into the developing MBO by in utero electroporation. The results were analyzed as before.

### Birthdate Analysis in the Wild Type and Foxb1 −/− MBO

Pregnant mice were intraperitoneally injected with bromodeoxyuridine (BrdU) (RPN201; GE Healthcare) (50 µg/g body weight) at the appropriate gestational age (from E9.5 to E13.5). The injections took place at 12:00 P.M., 3:00 P.M., and 6:00 P.M. (Takahashi et al., 1993) and the fetuses were collected at E18.5. We detected cell proliferation on cryosections (20 µm) by means of anti-BrdU antibody M0744 (1:100) (Dako), after epitope retrieval in 2 M HCl for 30 min at 37◦C.

### In Situ Hybridization (ISH) on Sections

Nonradioactive ISH was performed on cryosections (20 µm thick) that were fixed in 4% paraformaldehyde and acetylated after sectioning. Prehybridization, hybridization, and washing steps were performed with the help of an automatic liquid-handling unit (Genesis RSP 200; Tecan), and the digoxigenin-labeled probe was detected by a dual-amplification procedure.

#### Quantitative Real-Time PCR

The posterior ventral part of the hypothalamus of Foxb1 −/− and wild type animals was dissected, the tissue was homogenized and mRNA extracted with the Dynabeads mRNA DIRECT kit (Invitrogen). Reverse Transcription was performed with the Transcriptor First Strand cDNA Synthesis kit (Roche) using anchored-oligo(dT) and random hexamer primers. The cDNA was amplified in a Bio-Rad iCycler using SYBR Green Supermix (Bio-Rad) and the following gene-specific primers:


The PCR was performed in triplicates for each sample with three samples per genotype and normalized to house-keeping gene EF1 alpha as control.

#### Apoptosis Detection

We sectioned (20 µm thickness) with a cryostat E14.5, E15.5 and E16.5 brains electroporated at E12.5. We selected the sections containing the MBO, pretreated them with proteinase K (1.5 µg/ml, 5 min) at room temperature and labeled the apoptotic cells with the ApopTag TUNEL (terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling) kit (Millipore Bioscience Research Reagents) according to the instructions of the manufacturer. We used DAPI as counterstain and counted the absolute number of apoptotic cells in the posterior ventral part of the hypothalamus under 20x magnification in three histological sections per animal and in three individuals per treatment.

#### Proliferation After In Utero Electroporation of shRNA-3

Mouse embryos were transfected by in utero electroporation with GFP-control plasmid alone or together with shRNA-3 at E12.5, received BrdU at E13.5 (through intraperitoneal injection of the pregnant dam) (see above, Neuronal birthdate analysis) and their brains were collected at E14.5. Three control and three experimental embryonic brains were analyzed. For each of them, five horizontal sections (20 µm thick) through the MBO were treated with anti-BrdU and anti-GFP antibodies and examined under the confocal microscope. We counted BrdUlabeled cells in 100 µm × 200 µm bins covering the width of the neuroepithelium in the GFP-positive area of the neuroepithelium of the mammillary recess next to the MBO.

### Proliferation in the Foxb1 Mutant

We injected pregnant dams intraperitoneally with BrdU at E12.5 and collected the embryos for analysis either 3 h or 6 h later. Three embryos of each genotype (homozygotes vs. and heterozygotes) were analyzed. For each of them, seven to twelve horizontal sections (12 µm thick) through the MBO were reacted with anti-GFP antibody to identify the mammillary neuroepithelium (in this mutant, expression of reporter gene GFP is a proxy for Foxb1 transcriptional activation) as well as with anti-BrdU antibody and nuclear marker DAPI. We counted all cells on the apical border of the GFP-expressing mammillary neuroepithelium and scored them as BrdU-labeled or unlabeled.

#### Statistical Analysis

We used Prism 6 software (GraphPad Software Inc., La Jolla, California) to calculate the one-way ANOVA. The results are represented as mean ± Standard Deviation (SD).

### Results

#### Neuronal Birthdate Pattern does not Match Cdh Expression Pattern in the MBO

We hypothesized a simple mechanism to build the MBO. Since neurons fated for a certain specific MBO subnucleus are born during the same wave of neurogenesis (neurogenetic paradigm, Altman and Bayer, 1988; Bayer and Altman, 1995a), these neurons would then express the same cadherin combination and so they would aggregate together. This hypothesis predicts that the patterns of birthdating and cadherin expression combinations should match each other. That is, an MBO subnucleus would be born at a specific time and express a specific cadherin combination. This would be a direct and immediate way to prove that cadherin combinations underlie brain architecture.

We chose E18.5 as the age of analysis, since at this age all MBO neurons have been born, have completed migration and have settled in their final position; in addition, cadherin expression in the MBO gradually decreases and becomes less patterned after birth and through the adult stage (data not shown).

We first used ISH to label expression of the four classical cadherins of Type II present in the developing MBO (Kimura et al., 1996; Suzuki et al., 1997) at E18.5 (**Figures 1A--F**). We then labeled embryonic brains with proliferation marker BrdU at E9.5 through E13.5 and mapped the labeled cells at E18.5 (**Figures 1G--L**). We found that, on transverse sections at this age, MBO neurons are arranged in bands or strata (**Figure 1L**) according to an ''outside-in'' model. The neurons born first (E9.5) settled most laterally (''outside'') and younger neurons would settle gradually more medial, with the last born (E13.5) in the medialmost position by the third ventricle. Additionally, analysis on sagittal sections (not shown) indicated an anterior-lateral-dorsal (early born) to posterior-medial-ventral (late born) gradient, consistent with classical descriptions (Altman and Bayer, 1986). The outside-in chronological arrangement matches as well the latero-medial partition of the MBO into histological subnuclei (Allen and Hopkins, 1988).

the Foxb1 mutant MBO. (A) Position of the MBO on a sagittal diagram of the mouse brain, rostral to the left. (B--G) Sagittal sections (rostral to the left) of Foxb1 +/− (B--D) and Foxb1 −/− (E--G) E18.5 MBO. Antibody against reporter tree (arrows) in both Foxb1 +/− (B) and Foxb1 −/−(E). Antibody against Cdh11 labels MBO (arrowhead) and axonal bundle (arrow) in the Foxb1 +/− (C,D) but not in the Foxb1 −/− (F,G).

Comparison of the two data sets revealed that the combinatorial domains of cadherin expression do not match the birthdating bands revealed by BrdU (**Figures 1F,L**). Instead, each MBO neuron seems to belong at the same time to two different, intersecting groups, one of them determined by birthdate and the other by cadherin combination. Therefore, the chronological arrangement of MBO subnuclei cannot be due to cadherin expression combinations. Intriguingly, however, one characteristic was common to the entire MBO, and this was the intense expression of Cdh11 (**Figure 1B**; Allen and Hopkins, 1988). We hypothesized that this one cadherin could somehow be the ''universal glue'' keeping together the two intersecting systems of the MBO.

### Cdh11 Expression in the MBO is Maintained by Transcription Factor Foxb1

Next we looked for ways to study MBO architecture in conditions of reduced Cdh11 expression. Foxb1 is a transcription factor gene specifically expressed in the developing MBO (Kaestner et al., 1996; Alvarez-Bolado et al., 2000b) and essential for the development of the mammillary axons (Alvarez-Bolado et al., 2000a; Kloetzli et al., 2001; Szabó et al., 2011). Since Cdh11 has been implicated in axonal development and circuit formation (Marthiens et al., 2005; Paradis et al., 2007; Ross et al., 2012), we asked if Foxb1 could be involved in the regulation of Cdh11 expression in the MBO. Cdh11 protein was absent from the Foxb1 −/− MBO at E18.5 (**Figure 2**). Since we can detect Cdh11 mRNA in the mutant MBO at E12.5, E14.5 and E16.5 by ISH (**Figures 3A--F**) as well as quantitative RT-PCR (**Figures 3G--I**), but we cannot detect it anymore at E18.5 (**Figure 2**), we assume that Foxb1 is necessary not for activating Cdh11 expression in the MBO but only for its maintenance. This is a previously unreported role of transcription factor Foxb1 in the development of this part of the hypothalamus. The residual expression of Cdh11 in the MBO at E18.5 by quantitative RT-PCR (**Figure 3I**) is probably due to a periventricular layer (outside the MBO) which does not change in the mutant (arrowheads in **Figures 3A--F**). Additionally, Cdh6 and Cdh8 showed a slight reduction in expression after E16.5 in the Foxb1 mutant (**Figure 3I**).

We concluded that Foxb1 is required for maintenance of Cdh11 expression in the developing MBO, adding to the list of forkhead-regulated cadherin genes like E-cadherin (Cdh1) (Cha et al., 2007), Cdh3 (Habashy et al., 2008), Cdh5 (Kalinichenko et al., 2002) and Cdh7 (Dottori et al., 2001).

### Decrease in Cdh11 Expression in the Entire MBO Alters Cell Sorting

Next we wanted to use the Foxb1 mutant in order to test the hypothesis that intense expression of Cdh11 could be acting as a general glue, overriding any in principle possible effect of the cadherin combinations. Therefore we analyzed cadherin expression and birthdate of the different cell populations in the Foxb1 mutant MBO at E18.5 (**Figure 4**). The strong decrease in Cdh11 expression in the MBO at E18.5 (**Figure 2F**) was confirmed by ISH (**Figure 4A**; compare with **Figure 1B** for a control). The domains of expression of Cdh10, 8 and 6 in the mutant MBO were rearranged (**Figures 4B--D**, summarized in E; compare to **Figures 1A--F**). Comparison of the distribution of neuronal birthdates in the mutant (**Figures 4F--K**) and in the wild type MBO (**Figures 1G--L**) was very informative. In the mutant, neurons sharing a birthdate were spread over a large area, did not form separate bands, and were mixed with neurons of different birthdate (**Figure 4K**). We concluded that the late, gradual loss of Cdh11 expression in the entire developing MBO leads to disruption of the chronological arrangement of the MBO neurons.

### Decrease in Cdh11 Expression Alters Morphology but not Size of the MBO

Since cell sorting is an important mechanism underlying the development of a typical, characteristic shape of the different organs (reviewed in (Lecuit and Lenne, 2007)), we expected the overall morphology of the Foxb1 mutant MBO to change. To confirm this prediction, we took advantage of the two existing null mutant alleles of Foxb1, which carry different reporters. The Foxb1-tauLacZ (Alvarez-Bolado et al., 2000a) produces beta-galactosidase as reporter, while expression of the Foxb1-Cre allele is reported by EGFP (Zhao et al., 2007). By crossing these mutants, we generated Foxb1 heterozygous mutants carrying one allele expressing beta-galactosidase (and a wild type one of course), and homozygous mutants carrying also only one allele expressing beta-galactosidase (and another expressing EGFP). In this way, the amount of beta-galactosidase expressed per cell is the same in heterozygotes and homozygotes, and as a consequence we can use beta-galactosidase as a marker for comparison (we discarded the heterozygotes expressing GFP). We know that Foxb1 is expressed in the entire MBO (Alvarez-Bolado et al., 2000b), and therefore, expression of betagalactosidase is a good reporter of MBO morphology and size. Observation of transverse sections of the MBO of both genotypes labeled with antibody against beta-galactosidase showed a change in MBO morphology in the mutant (diagram in **Figure 5A**). Measuring the size of every one of four mediolateral regions of the MBO (see Materials and Methods section) revealed significant reduction in the most medial region and significant enlargement in the most lateral region of the homozygous MBO (**Figures 5B,C**). Remarkably, the overall size of the mutant MBO was not different from that of the heterozygous (**Figure 5D**). To further support this claim, we ascertained that there is no difference in cell density (**Figure 5E**) or in proliferation (**Figure 5F**) in the mutant MBO. As expected after an alteration of cell sorting, the MBO morphology was affected while its overall size remained unaffected.

#### Knocking Down Cdh11 by RNA Interference

Based on the hypothesis that Cdh11 expression keeps the chronological arrangement of MBO neurons, we then predicted that MBO-fated migrating neurons lacking Cdh11 would fail to enter a wild-type, Cdh11-expressing MBO. To test the prediction, we decided to use RNA interference (RNAi; Paddison et al., 2002)

in utero in order to reduce Cdh11 expression in MBO neurons born at a specific time point, then analyze their position several days later. We tested different commercially obtained plasmidencoded small hairpin (sh)RNA against Cdh11 in culture (see Methods section for details) and found that transfection of shRNA-3 resulted in powerful knockdown of Cdh11 in culture (**Figure 6A**). We then used in utero electroporation to transfect shRNA-3 into the neuroepithelium lining the mammillary recess of the third ventricle, where MBO neurons are born (**Figure 6B**). We chose E12.5 as time point for the experiment, since at this age the MBO is accessible to DNA transfection through in utero electroporation (Haddad-Tóvolli et al., 2013). Transfection of GFP-expressing reporter plasmid at E12.5 into the mammillary recess resulted in an abundance of labeled neuroepithelial cells as can be seen in horizontal sections (**Figure 6C**). Cdh11 could be detected with antibodies in the same cells (**Figures 6D,E**). A very different picture could be seen when Cdh11 mRNA was knocked down in the neuroepithelium. Although numerous neuroepithelial cells were labeled with GFP (**Figure 6F**), Cdh11 protein could not be detected in them (**Figures 6G,H**).

Furthermore, Cdh11 protein could not be detected in MBO neurons of Cdh11-knockdown brains either (**Figures 7A--F**). Finally, by screening for GFP expression on

transverse Vibratome sections of transfected MBO we detected a clear and consistent pattern alteration after transfection with shRNA-3 (**Figures 7G--I**).

In conclusion, at this point we were able to specifically knockdown Cdh11 expression in culture and in the MBO developing in utero.

### Control-Transfected MBO-Fated Neurons Form a Defined Group Inside the MBO

We then used antibody detection of GFP on horizontal sections in order to analyze the position of control-transfected neurons at different time points (**Figure 8**). Transfection was performed at E12.5 (**Figures 8A,B**). Two days after transfection, a number of GFP-labeled neurons was present in the MBO forming a well-defined stream extending from rostral to caudal through the MBO (**Figures 8C--E**). These neurons were placed at the most medial side of the MBO, as expected

FIGURE 5 | Shape, but not size, altered in the Foxb1 −/− MBO. (A) Superimposing the outline of the labeled area in transverse section of the wild type and Foxb1 −/− MBO suggests a change in morphology in the mutant. (B) Thick sagittal sections of Foxb1 +/− and−/− labeled with X-gal reaction and compared according to 4 arbitrary latero-medial regions of equal size (numbered 1--4) confirm this impression (in heterozygotes as well as homozygotes, only one allele expressed β-galactosidase, see Methods section for details). (C) The combined stained area of the sections in (B), for each of the 4 latero-medial regions shows significantly increased lateral size and significantly reduced medial size for the Foxb1 −/− MBO. (D) No significant differences in total volume of the MBO between Foxb1 +/− and Foxb1 −/− (combined stained area of all sections). (E) No significant differences in cell density difference between Foxb1 +/− and Foxb1 −/−. (F) No significant difference in proliferation in the mammillary neuroepithelium between Foxb1 +/− and Foxb1 −/−. The cells were counted either 3 h (left) or 6 h (right) after BrdU injection at E12.5. (D--F) One-way ANOVA; mean ± SD; n.s., not significant.

following the general ''outside-in'' settling pattern typical of the hypothalamus. The latest arrived neurons appose themselves to earlier populations from the medial side, so that the oldest neurons (born at E9.5) will end up forming the most lateral (''outside'') part of the nucleus and the youngest (born at E13.5) the most medial (**Figures 1G--L**). As expected, neurons born before transfection age (E12.5) had arrived earlier to the MBO, occupied more lateral positions and were unlabeled (**Figures 8C--E**).

Labeled cells arrived to the MBO through E16.5 (**Figures 8F--H**) and E18.5 (**Figures 8I--K**) and they remained recognizable as a stable, compact group on the lateral side of the nucleus.

These results show that we can use in utero transfection to label neurons born at a certain age and that these neurons form

an identifiable group consistently entering the MBO as a stream and consistently settling in a position corresponding to their birthdate.

### Cdh11-Knockdown Neurons Accumulate Outside the MBO

Cdh11-knockdown transfected neurons behaved in a quite different way (**Figure 9)**. Already at E14.5 the stream of labeled cells did not form a straight, rostro-caudally oriented group but seemed deformed in the medial direction, towards the midline (arrow in **Figures 9C--E**). At the same time, labeled cells started to abnormally accumulate on the rostral side of the MBO (double arrowhead in **Figures 9D,E**). Two days later (E16.5), only few labeled cells were still to be found in the MBO (arrow in **Figure 9G**). The labeled cells abnormally gathering rostral to the MBO formed an elongated, medio-laterally oriented aggregate (double arrowhead in **Figures 9G,H**). At E18.5 there were virtually no labeled cells in the MBO. The rostral border of this nucleus however was easy to recognize because of a large accumulation of labeled cells (double arrowhead in **Figures 9I--K**). We checked for apoptosis and proliferation effects in order to discard these phenomena as causes of the decrease in labeled cells in the MBO (**Figure 10**).

Arrowheads in D-F show GFP-expressing, non-Cdh11-expressing cells. (G) Transverse section of E18.5 brain showing Cdh11 detection by ISH. The frame includes the MBO (identified by strong Cdh11 expression) and indicates the approximate area of the image in (H,I). (H,I) Transverse vibratome sections of wild type E18.5 MBO after electroporation of plasmids containing GFP reporter alone (H) or together with shRNA-3 plasmid (I), with an added outline of the MBO for reference. Arrowheads indicate equivalent positions in the control (H) and knockdown MBO (I). The arrow in (I) indicates an abnormal band of labeled cells outside the knockdown MBO. Scale bar (in A) 100 µm.

The definitive control for an RNAi experiment is the rescue by expression of a form of the target gene resistant to siRNA (Anonymous-Editorial, 2003). Accordingly, we performed rescue experiments based on co-transfection of DNA constructs expressing non-interferable Cdh11 (see Material and Methods for details). These experiments consist of introducing in the cells a form of Cdh11 that has been mutated in such a way that it preserves its adhesive domains while losing the domain that is recognized by the shRNA (**Figures 11A,B**). We would expect that cells transfected in this way would show a lesser effect of the shRNA interference, since shRNA will be able to degrade endogenous Cdh11, but not the ''non-interferable'' Cdh11

that we are cotransfecting. The results of these experiments (**Figures 11C--E**) show many more GFP-labeled cells inside the MBO in ''rescued'' animals than in animals treated only with shRNA-3. In this way, we confirmed the specificity of our previous RNAi experiments. We concluded that experimental reduction of Cdh11 expression in one specific MBO neuronal subpopulation during development causes that subpopulation to accumulate outside the MBO.

### Discussion

Three insights have been considered as key to understand cell sorting in brain development--the importance of information encoded in neuronal birthdates (Bayer and Altman, 1987) and in cadherin combinations (Suzuki et al., 1997; Price et al., 2002), and in (D,E,G,H,J,K), 50 µm.

the importance of non-specific adhesion phenomena (Foty et al., 1996). In this work we combine for the first time these insights by showing: (1) that one-cadherin adhesion has the power to organize the neurons of a brain nucleus according to dates of neurogenesis; and (2) that cadherin combinations and onecadherin-adhesion have different roles and different mechanisms

working sequentially to fulfill different roles through different mechanisms.

In the developing MBO, two sources of information, i.e., neuronal birthdates and cadherin combinations, work to secure appropriate connections between MBO neurons and their anterior thalamic targets. Neuronal birthdates ensure appropriate medio-lateral correspondence between MBO subdivisions and the anterior thalamic nuclei that are their specific targets. Cadherin combinations presumably take care of the last step in navigation, identifying the individual target neurons inside the thalamus.

These two sources of information are maintained through a hierarchy of adhesions. First, Cdh11 allows entrance of the successively arriving neurons into the target nucleus,

well.

then, again Cdh11 keeps them organized chronologically. Cdh11 prevents also weaker, combination-based adhesion forces from intermixing the birthdate-based organization. Finally, the cadherin combinations would underlie appropriate fasciculation

amino acids (Asn-Arg-His) in the resulting human Cadherin11 protein. (B) The missing three aminoacids are part of the juxtamembrane domain (JMD) of the

(blue), GFP plus shRNA#3 (red) and GFP plus shRNA#3 plasmid plus rescue plasmid (green). of axons projecting to same area within a target region (Wöhrn et al., 1999; Treubert-Zimmermann et al., 2002), and could be

responsible for the final identification of the target neurons as

Frontiers in Neuroanatomy | www.frontiersin.org March 2015 | Volume 9 | Article 29 |

### Cdh11, One Cadherin to Rule Them All

Our first finding, that birthdates and cadherin combinations do not coincide in the MBO (**Figure 1**) (reminiscent of similar results in the avian and mouse striatum (Redies et al., 2002; Heyers et al., 2003)) is surprising. How could neurons born on a certain date aggregate together (Bayer and Altman, 1995b) if not by expressing specific combinations of adhesive molecules (Redies and Puelles, 2001). A possible answer can be found in **Figures 4**, **5**, which show that abolition of Cdh11 expression (Foxb1 mutant) alters sorting in the MBO, causing a mixing of the combinatorial groups as well as the birthdate groups. These results suggests that intense, generalized Cdh11-based adhesion would make all MBO neurons homogeneously highly adhesive overriding the effect of the subordinate interactions based on the combinations of Cdh6, 8 and 10 otherwise present in MBO neurons. In this way, Cdh11 would ensure that the newly arrived neurons appose themselves from the medial side to the previously arrived (''outside-in'' arrangement) rather than mixing with each other based on weaker variegated interactions based on the combinations. The results of the knockdown experiments (**Figures 6**--**9**) reinforce this insight by showing that, without Cdh11 expression, newly arrived neurons are excluded from the MBO. This is reminiscent of the DAH prediction that less adhesive cells will remain on the periphery (Steinberg, 1963). MBO neurons keep a medio-lateral correspondence with their targets in the anterior thalamus- the most medial MBO neurons project to the most medial anterior thalamic neurons, and the most lateral to the most lateral (Seki and Zyo, 1984). We suggest that the function of Chd11-based adhesion is to keep the MBO subdivisions approximately in register with their targets in the ATC, making sure that their axons enter the target region in the appropriate neighborhood. Accordingly, when the developing mammillothalamic axonal tract reaches the anterior thalamus, its axons separate into three bundles which innervate their targets sequentially from medial to lateral (Alpeeva and Makarenko, 2009).

### Role of the Cadherin Combinations

What would then be the role of the cadherin combinations? We propose that, in the MBO, combinatorial adhesion adds one further layer of specificity to the connections between MBO and anterior thalamus. After appropriate, medio-laterally organized entrance of MBO axons in their target region, the anterior thalamus, and since anterior thalamic neurons express combinations of Cdh6, 8 and 10 (Suzuki et al., 1997; Bekirov et al., 2002) matching those expressed by the incoming MBO axons, these can rely on the combinatorial code for the final target identification. In this way, the two intersecting, cadherinbased sorting systems of the MBO guarantee appropriate neighborhood targeting (Cdh11) and fine-grained ''address'' targeting (combinations). Cdh6, 8, 10 and 11 have all been shown indispensable for appropriate synaptic connectivity in a variety of systems (Suzuki et al., 1997; Paradis et al., 2007; Osterhout et al., 2011; Williams et al., 2011; Ross et al., 2012). Cadherindependent specific fasciculation, experimentally demonstrated in other systems (Treubert-Zimmermann et al., 2002) could play a role also here. Appropriate connectivity based on birthdates has been suggested as a general principle in brain development (Bayer and Altman, 1995b).

#### A Hierarchy of Homotypic Interactions

Incidentally, the proposed role of Cdh11 as ''central clasp'' can be understood as non-specific, that is, based on stronger adhesion, not on combinations (despite being homophilic, i.e., Cdh11- Cdh11). This means that Cdh11 is not part of the combinations but overrides them all. On this basis we can predict that, when expression of Cdh11 decreases (i.e., in the Foxb1 mutant), MBO neurons recover their underlying differential adhesivities, which are due to differential expression intensity of various adhesion molecules other than Cdh11, and reorganize accordingly. A key assumption for this interpretation is that the combinations provide weaker adhesion than the homogeneous expression of Cdh11. This conjecture is borne out by the phenotype. In addition, the appearance of the Cdh11-knockdown neurons gathered at the boundary of the MBO (**Figures 8**, **9**) brings to mind the DAH prediction that the least adhesive cells will remain on the surface of a more adhesive ''bulk'' (Steinberg, 1963). Perhaps the dicotomies ''homophilic vs. heterophilic'' and ''specific vs. non-specific'' should be substituted by more flexible concepts.

#### Caveats

Cdh11-knockdown neurons could simply be migrationimpaired, since cadherins have a role in migration (Geisbrecht and Montell, 2002; Cavallaro and Dejana, 2011) and Cdh11 is specifically required for migration in some models (Kiener et al., 2006, 2009; Kashef et al., 2009; Huang et al., 2010; Kaur et al., 2012). However, our Cdh11-knockdown cells are able to reach the boundary of the MBO, indicating that Cdh11 is not essential for their migration. The lack of an abnormal phenotype in the MBO of the Cdh11 mutant mouse (Manabe et al., 2000), can be attributed to early compensatory effects through other adhesive proteins (Nadeau, 2003; Barbaric et al., 2007). The phenotype can be due to other adhesion molecules being downregulated in the Foxb1 mutant. However, that does not change, rather would reinforce, the main finding---that there are two intersecting systems, and cadherin combinations underlie one of them. For specific synapse formation, other molecules, like the nectins, are also important (Takeichi, 2007).

### Conclusions

We propose that neuronal sorting inside brain nuclei, based on cell body-to-cell body interactions and responsible for brain cytoarchitecture, is caused by adhesion-based, noncombinatorial mechanisms, one important function of which would be to keep neurons sorted according to birthdate information. Additionally, non-specific adhesion mechanisms would prevent cadherin combinations from altering the birthdate-based sorting through weaker, combination-based mechanisms. The most likely role for cadherin combinations in the developing brain is to support specific synaptogenesis through specific axonal fasciculation and final target recognition.

### Acknowledgments

This work was supported by the Deutsches Forschungsgemeinschaft (AL603/2-1). NES and RHT were

### References


the recipients of DAAD Fellowships. We acknowledge the financial support of the Deutsche Forschungsgemeinschaft and Ruprecht-Karls-Universität Heidelberg within the funding programme Open Access Publishing.


Geisbrecht, E. R., and Montell, D. J. (2002). Myosin VI is required for E-cadherinmediated border cell migration. Nat. Cell Biol. 4, 616--620. doi: 10.1038/ ncb830

Godt, D., and Tepass, U. (1998). Drosophila oocyte localization is mediated by differential cadherin-based adhesion. Nature 395, 387--391. doi: 10.1038/ 26493

González-Reyes, A., and St Johnston, D. (1998). The Drosophila AP axis is polarised by the cadherin-mediated positioning of the oocyte. Development 125, 3635--3644.


in the tectofugal pathways of the chicken embryo. Neuroscience 90, 985--1000. doi: 10.1016/s0306-4522(98)00526-0

Zhao, T., Zhou, X., Szabó, N., Leitges, M., and Alvarez-Bolado, G. (2007). Foxb1 driven Cre expression in somites and the neuroepithelium of diencephalon, brainstem and spinal cord. Genesis 45, 781--787. doi: 10.1002/dvg.20356

**Conflict of Interest Statement**: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Szabó, Haddad-Tóvolli, Zhou and Alvarez-Bolado. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## ADVANTAGES OF PUBLISHING IN FRONTIERS

FAST PUBLICATION Average 90 days from submission to publication

COLLABORATIVE PEER-REVIEW

Designed to be rigorous – yet also collaborative, fair and constructive

RESEARCH NETWORK Our network increases readership for your article

#### OPEN ACCESS

Articles are free to read, for greatest visibility

#### TRANSPARENT

Editors and reviewers acknowledged by name on published articles

GLOBAL SPREAD Six million monthly page views worldwide

COPYRIGHT TO AUTHORS

No limit to article distribution and re-use

IMPACT METRICS Advanced metrics track your article's impact

SUPPORT By our Swiss-based editorial team

EPFL Innovation Park · Building I · 1015 Lausanne · Switzerland T +41 21 510 17 00 · info@frontiersin.org · frontiersin.org