# BONE MARROW ADIPOSE TISSUE: FORMATION, FUNCTION, AND IMPACT ON HEALTH AND DISEASE

EDITED BY: William P. Cawthorn and Erica L. Scheller PUBLISHED IN: Frontiers in Endocrinology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2017 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 978-2-88945-245-3 DOI 10.3389/978-2-88945-245-3

## 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

# **BONE MARROW ADIPOSE TISSUE: FORMATION, FUNCTION, AND IMPACT ON HEALTH AND DISEASE**

## Topic Editors:

**William P. Cawthorn,** University of Edinburgh, Edinburgh, United Kingdom **Erica L. Scheller,** Washington University, United States

Micrograph of a rat femoral head (male FBN rat, 30-months old). The section was stained with haematoxylin and eosin. The image was then stylised in Adobe Photoshop by processing through the 'Oil Paint' Filter, followed by HDR Toning. Adipocytes (white circles) are clustered within pockets of bone marrow (purple), sorrounded by trabecular bone (pink). Tissue sectioning and staining was done by Dr. Karla Suchacki and Mr. Alexandre Lafond (University of Edinburgh), with image processing done by Dr. William Cawthorn.

Adipocytes are a major component of the bone marrow, accounting for up to 70% of total bone marrow volume in healthy humans. Indeed, this bone marrow adipose tissue (often referred to as 'MAT' or 'BMAT') accounts for at least 5% of total adipose tissue mass in lean, healthy humans, suggesting a role in normal physiology and development. Bone marrow adiposity further increases with ageing and in diverse clinical conditions, including major public health challenges such as osteoporosis. Yet despite this abundance and compelling clinical potential, bone marrow adipocytes have received surprisingly little attention from the biomedical research community.

Thankfully, this is now beginning to change. Research over the past decade has begun to increase our knowledge of BMAT, including the conditions associated with altered bone marrow adiposity and the potential physiological and pathological functions of bone marrow adipocytes. The articles within this e-Book highlight many of these recent developments, underscoring our increasing knowledge of BMAT formation and function; showcasing emerging techniques for basic and clinical BMAT analysis; and highlighting key questions and future directions for this burgeoning and increasingly diverse field.

The editors would like to express their thanks to the authors for contributing the articles within this e-Book; to the senior editors at Frontiers in Endocrinology for their guidance; and to the staff at Frontiers for their helpful input throughout.

**Citation:** Cawthorn, W. P., Scheller, E. L., eds. (2017). Bone Marrow Adipose Tissue: Formation, Function, and Impact on Health and Disease. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-245-3

# Table of Contents

*06 Editorial: Bone Marrow Adipose Tissue: Formation, Function, and Impact on Health and Disease*

William P. Cawthorn and Erica L. Scheller

## **Chapter 1 – Key Aspects of BMAT Biology in Health and Disease**


Christian Cordes, Thomas Baum, Michael Dieckmeyer, Stefan Ruschke, Maximilian N. Diefenbach, Hans Hauner, Jan S. Kirschke and Dimitrios C. Karampinos

## **Chapter 2 – BMAT Origins and Development**

*33 The Bone Marrow-Derived Stromal Cells: Commitment and Regulation of Adipogenesis*

Michaela Tencerova and Moustapha Kassem

*45 Fatty Infiltration of Skeletal Muscle: Mechanisms and Comparisons with Bone Marrow Adiposity*

Mark W. Hamrick, Meghan E. McGee-Lawrence and Danielle M. Frechette

*52 Vanadate Impedes Adipogenesis in Mesenchymal Stem Cells Derived from Different Depots within Bone*

Frans Alexander Jacobs, Hanél Sadie-Van Gijsen, Mari van de Vyver and William Frank Ferris

## **Chapter 3 – Metabolic and Endocrine Aspects of BMAT**

*64 Changes in Skeletal Integrity and Marrow Adiposity during High-Fat Diet and after Weight Loss*

Erica L. Scheller, Basma Khoury, Kayla L. Moller, Natalie K. Y. Wee, Shaima Khandaker, Kenneth M. Kozloff, Simin H. Abrishami, Brian F. Zamarron and Kanakadurga Singer

*77 Hypothalamic Leptin Gene Therapy Reduces Bone Marrow Adiposity in ob/ob Mice Fed Regular and High-Fat Diets*

Laurence B. Lindenmaier, Kenneth A. Philbrick, Adam J. Branscum, Satya P. Kalra, Russell T. Turner and Urszula T. Iwaniec

## *86 Marrow Adipose Tissue Expansion Coincides with Insulin Resistance in MAGP1-Deficient Mice*

Tezin A. Walji, Sarah E. Turecamo, Alejandro Coca Sanchez, Bryan A. Anthony, Grazia Abou-Ezzi, Erica L. Scheller, Daniel C. Link, Robert P. Mecham and Clarissa S. Craft


Olfa Ghali, Nathalie Al Rassy, Pierre Hardouin and Christophe Chauveau

*112 Increased Circulating Adiponectin in Response to Thiazolidinediones: Investigating the Role of Bone Marrow Adipose Tissue*

Richard J. Sulston, Brian S. Learman, Bofeng Zhang, Erica L. Scheller, Sebastian D. Parlee, Becky R. Simon, Hiroyuki Mori, Adam J. Bree, Robert J. Wallace, Venkatesh Krishnan, Ormond A. MacDougald and William P. Cawthorn

## **Chapter 4 – BMAT and Cancer**


Carolyne Falank, Heather Fairfield and Michaela R. Reagan

*151 New 3D-Culture Approaches to Study Interactions of Bone Marrow Adipocytes with Metastatic Prostate Cancer Cells*

Mackenzie Katheryn Herroon, Jonathan Driscoll Diedrich and Izabela Podgorski

# Editorial: Bone Marrow Adipose Tissue: Formation, Function, and Impact on Health and Disease

*William P. Cawthorn1 \* and Erica L. Scheller <sup>2</sup>*

*1University/British Heart Foundation Centre for Cardiovascular Science, The Queen's Medical Research Institute, University of Edinburgh, Edinburgh, United Kingdom, 2Division of Bone and Mineral Diseases, Department of Medicine, Washington University, Saint Louis, MO, United States*

Keywords: bone marrow adipocytes, bone marrow adipose tissue, bone remodeling, osteoporosis, bone marrow stromal cell, endocrinology and metabolism, bone metastases, multiple myeloma

**The Editorial on the Research Topic**

**Bone Marrow Adipose Tissue: Formation, Function, and Impact on Health and Disease**

Enthusiasm surrounding bone marrow (BM) adiposity has accelerated in recent years, motivated by numerous factors: adipocytes are abundant within the BM, BM adiposity increases in diverse pathophysiological states, and BM adipocytes can exert diverse effects both within and beyond the skeleton. This diversity has attracted a broad cross-section of scientists to the field and we are delighted to showcase their work in this special edition of *Frontiers in Endocrinology*.

## WHAT'S IN A NAME?

*Edited and Reviewed by: Jonathan H. Tobias, University of Bristol, United Kingdom*

> *\*Correspondence: William P. Cawthorn w.cawthorn@ed.ac.uk*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 02 May 2017 Accepted: 09 May 2017 Published: 29 May 2017*

#### *Citation:*

*Cawthorn WP and Scheller EL (2017) Editorial: Bone Marrow Adipose Tissue: Formation, Function, and Impact on Health and Disease. Front. Endocrinol. 8:112. doi: 10.3389/fendo.2017.00112*

Bone marrow adipocytes (BMAs) are large, roughly spherical cells containing a unilocular lipid droplet. They have gone by many names. Early work aptly recognized that BM has different colors, red or yellow, depending on its location—thus the terms *red marrow* and *yellow marrow* came into use (1). Red marrow consists of blood-forming cells with scattered adipocytes, whereas yellow marrow is filled almost entirely with adipocytes. It was not until 1950s–1960s that BM adipocytes became recognized as a fat depot with the potential for adipose tissue-like characteristics (2, 3). This caused the term *yellow marrow* to be replaced by "*marrow fat"* or "*fatty marrow"*. Even today, radiological studies still refer to the "bone marrow fat fraction" (BMFF) as a measure of marrow lipid content (Cordes et al.); however, as marrow adipocytes gain more recognition as a distinct, functional cell type, such designations are expanding beyond references to simple lipid. Herein, several options and abbreviations are introduced, including "BMAs" (Hardouin et al.; Ghali et al., ), "marrow adipose tissue" (MAT) (Scheller et al.; Walji et al.; Pino et al.; Sulston et al.), "bone marrow adipose tissue" (BMAT) (Hardouin et al.), and the "constitutive" and "regulated" subtypes of these (i.e., cMAT/rMAT or cBMA/rBMA). Clearly, reaching a consensus on this nomenclature will be important in providing a consistent framework on which this burgeoning field can progress. Based on the publications to date and the overlap of the acronym "MAT" with "muscle adipose tissue", we propose adopting "BMAT" to refer to the tissue and "BMA" to refer to the adipocytes therein.

## ORIGINS AND EXPANSION OF BMAT

Bone marrow adipose tissue develops postnatally and accounts for 50–70% of BM volume in healthy adult humans; hence, BMAT development is a normal physiological process. BMAT further accumulates with aging and in diverse clinical conditions, suggesting pathological implications of aberrant BMAT formation and function. In their comprehensive review, Hardouin et al. further discuss these issues, which underscore the need to better understand BMAT in physiology and disease. Several other articles herein provide additional focus on BMAT developmental origins and the mechanisms underlying BMAT expansion. Tencerova and Kassem review the differentiation of BMAs from BM stromal cells (BMSCs), discussing lineage tracing studies that are identifying BMSCs committed to the adipogenic lineage; signaling pathways that regulate this commitment; and the heterogeneity of committed BMSC subpopulations. The latter is underscored by Jacobs et al., whose study reveals site-specific differences in BMSC adipogenic potential, while Lindenmaier et al. provide further evidence for the important role of leptin in regulating adipogenic and osteogenic differentiation of BMSCs. Elsewhere, Hamrick et al. discuss BMAT expansion by raising a compelling question: are there common mechanisms underlying adipocyte accumulation in muscle and bone? This unique perspective highlights the many conditions with similar effects on BMAT expansion and intramuscular adipocyte accumulation, and how these common mechanisms might be targeted for therapeutic benefit.

Several other papers in this Research Topic address BMAT expansion in adverse metabolic states. As reviewed by Ghali et al., increased BMAT during caloric restriction is now well established, although the causes and consequences of this remain incompletely understood. At the other end of the metabolic spectrum, studies by Scheller et al. and by Lindenmaier et al. provide additional evidence that, in mice, BMAT accumulates during obesity, and further reveal that this is prevented by leptin treatment or weight loss. This is reminiscent of exercise's ability to prevent BMAT expansion during obesity or other conditions, as discussed by Pagnotti and Styner. Finally, Walji et al. reveal that, in *Mfap2−/−* mice, BMAT expansion coincides with the development of insulin resistance, but not of obesity or hyperglycemia.

Despite such progress, more research is needed to identify committed BMA progenitor(s) in humans and to clarify the mechanisms underlying gain or loss of BMAT in normal development and disease.

## BMAT—GOOD, BAD, OR SOMEWHERE IN BETWEEN?

These questions regarding BMAT formation are related to a broader question: what is the function of BMAT? Given that it is a feature of normal anatomy and development, it would be surprising if BMAT did not fulfill at least some physiological functions. Yet, it also seems likely that, in adverse contexts, BMAT can contribute to disease pathogenesis. Notably, increased BMAT often coincides with decreased bone mass, suggesting that bone formation and marrow adiposity are linked. One possibility, touted throughout this Research Topic, is that a common BM progenitor undergoes adipogenesis at the expense of osteogenesis (Hardouin et al.; Pino et al.; Tencerova and Kassem; Jacobs et al.; Pagnotti and Styner). However, other contributors show that alterations in BMAT quantity do not always coincide with opposite changes in bone mass, whether in response to obesity (Scheller et al.), insulin resistance (Walji et al.), leptin treatment (Lindenmaier et al.), or caloric restriction/anorexia (Ghali et al.). Another possibility is that BMAs secrete factors to directly regulate bone remodeling, even without gross changes in BMAT quantity. Other potential secretory functions of BMAT are further explored by Sulston et al., who investigate BMAT as a source of the hormone adiponectin. Such endocrine properties might allow BMAT to act outside the skeleton, exerting systemic effects on metabolism and beyond.

A final concept gaining increasing attention is that BMAs modulate tumor growth and metastasis. Morris and Edwards provide a comprehensive review, while Falank et al. offer an in-depth focus on the interplay between BMAT and multiple myeloma cells. Complementing these articles, Herroon et al. describe novel methods for studying interactions between BMAs and prostate metastases. Together, these articles emphasize that some BMAderived factors can stimulate tumor progression, while other secretory products can exert inhibitory effects. Targeting these mechanisms may thereby represent a novel therapeutic avenue in the battle against myeloid cancers and metastatic bone disease.

## TECHNOLOGICAL ADVANCEMENTS

Recent advances in quantitative imaging of BMAs have accelerated our understanding of this relatively inaccessible fat depot in both rodents and humans. In the article by Hardouin et al., the evolution of these techniques is described in detail, including magnetic resonance imaging, proton magnetic resonance spectroscopy, and high-resolution computed tomography. In humans, synthesis, adaptation, and modification of existing techniques have allowed researchers to non-invasively monitor BMAT development and expansion in parallel with changes to the healthy and diseased skeleton (reviewed in Cordes et al.). In rodents, imaging of skeletal BMAT with osmium tetroxide and computed tomography has emerged as the current gold standard, facilitating both volumetric quantification of BMAT and spatial analysis of BMAT patterning within bone. Longitudinal application of these techniques in rodents is leveraged in this edition to demonstrate that bone loss precedes BMAT expansion in two models of obesity and diabetes (Scheller et al.; Walji et al.), leading to the conclusion that, in this context, accumulation of BMAT may be more closely linked to peripheral adipose tissue dysfunction than bone turnover.

Unlike widespread advances in imaging, techniques for robust molecular analysis and genetic manipulation of BMAs are still lacking. However, the work by Herroon et al., presented in this edition, provides a glimpse of the future possibilities. The authors detail two novel *in vitro* approaches to study the interactions of primary BM-derived adipocytes and tumor cells in a threedimensional coculture. Given the challenges inherent in studying BMAs within their native skeletal microenvironment, *ex vivo* and *in vitro* systems that recapitulate key aspects of BMAT biology will be crucial to elucidation of BMA function.

## QUALITY VS QUANTITY

Emerging evidence supports the concept that not all BMAs are created equal and that they have the potential for maladaptation with age and disease. Indeed, it may not be the total amount of BMAT, but rather its context-specific phenotype that dictates the relationships between BMAT, bone and whole-body metabolism. In this edition, our authors highlight the ability of BMAT to undergo pathological change in diseases such as osteoporosis (Ghali et al.). Morris and Edwards also discuss the relevance of this concept for tumor metastasis, which may explain context-specific tumor-promoting and tumor-suppressive effects of BMAT. A key aspect of this phenotypic switch or maladaptation appears to be shifts in the lipid composition of BMAs (discussed in Pino et al.). While the source of these differences remains unknown, current hypotheses suggest consideration of site-specific progenitor populations (Jacobs et al.) and microenvironmental programming of both developing and mature BMAs. Future work will undoubtedly be needed to explore the differences in BMAT between skeletal sites and to determine its implications for bone loss and metabolic health.

## PROSPECTUS

This edition provides a comprehensive overview of the state of the field of BM adiposity, details novel hypotheses, and provides opportunities for development and growth. Based on the work presented, it is clear that future studies are warranted to define the biological functions of BM adipocytes—having implications for

## REFERENCES


bone turnover, hematopoiesis, systemic metabolic homeostasis, tumor metastasis, energy storage, and beyond.

## AUTHOR CONTRIBUTIONS

WC and ES jointly wrote this editorial. ES initially wrote the section entitled "What's in a Name?", with WC then contributing further edits. WC then focused on the sections entitled "Origins and Expansion of BMAT" and "BMAT—Good, Bad, or Somewhere in between?", while ES focused on "Technological Advancements", "Quality vs Quantity," and the "Prospectus".

## ACKNOWLEDGMENTS

WC and ES are grateful to all of the authors for their contributions to this Research Topic.

## FUNDING

WC is supported by a Career Development Award (MR/ M021394/1) from the Medical Research Council (UK) and by a Chancellor's Fellowship from the University of Edinburgh. ES is supported by grant R00-DE024178 from the National Institutes of Health.

**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 © 2017 Cawthorn and Scheller. 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.*

# Bone Marrow Adipose Tissue: To Be or Not To Be a Typical Adipose Tissue?

*Pierre Hardouin, Tareck Rharass and Stéphanie Lucas\**

*Laboratory of Pathophysiology of Inflammatory Bone Diseases PMOI, University of Littoral-Opale Coast ULCO, Boulogne sur Mer, France*

Bone marrow adipose tissue (BMAT) emerges as a distinct fat depot whose importance has been proved in the bone–fat interaction. Indeed, it is well recognized that adipokines and free fatty acids released by adipocytes can directly or indirectly interfere with cells of bone remodeling or hematopoiesis. In pathological states, such as osteoporosis, each of adipose tissues – subcutaneous white adipose tissue (WAT), visceral WAT, brown adipose tissue (BAT), and BMAT – is differently associated with bone mineral density (BMD) variations. However, compared with the other fat depots, BMAT displays striking features that makes it a substantial actor in bone alterations. BMAT quantity is well associated with BMD loss in aging, menopause, and other metabolic conditions, such as anorexia nervosa. Consequently, BMAT is sensed as a relevant marker of a compromised bone integrity. However, analyses of BMAT development in metabolic diseases (obesity and diabetes) are scarce and should be, thus, more systematically addressed to better apprehend the bone modifications in that pathophysiological contexts. Moreover, bone marrow (BM) adipogenesis occurs throughout the whole life at different rates. Following an ordered spatiotemporal expansion, BMAT has turned to be a heterogeneous fat depot whose adipocytes diverge in their phenotype and their response to stimuli according to their location in bone and BM. *In vitro*, *in vivo*, and clinical studies point to a detrimental role of BM adipocytes (BMAs) throughout the release of paracrine factors that modulate osteoblast and/or osteoclast formation and function. However, the anatomical dissemination and the difficulties to access BMAs still hamper our understanding of the relative contribution of BMAT secretions compared with those of peripheral adipose tissues. A further characterization of the phenotype and the functional regulation of BMAs are ever more required. Based on currently available data and comparison with other fat tissues, this review addresses the originality of the BMAT with regard to its development, anatomy, metabolic properties, and response to physiological cues.

Keywords: fat–bone association, marrow fat, adipokine, bone marrow adiposity, skeletal adipocyte, osteoporosis, bone fragility

Besides being the main energy storage sites, adipose tissues have revealed through these last decades their diversity regarding their cellular composition, anatomical location, and pathophysiological properties. Indeed, adipocytes are typically classified into three categories: white, brown, and beige types (1). White adipocytes store excessive energy supply in a unilocular triglyceride droplet to release fatty acids in periods of energy depletion. White adipocytes also exert an endocrine function

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Eleni Douni, Agricultural University of Athens, Greece Ann Schwartz, University of California San Francisco, USA*

*\*Correspondence: Stéphanie Lucas stephanie.lucas@univ-littoral.fr*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 03 May 2016 Accepted: 21 June 2016 Published: 30 June 2016*

#### *Citation:*

*Hardouin P, Rharass T and Lucas S (2016) Bone Marrow Adipose Tissue: To Be or Not To Be a Typical Adipose Tissue? Front. Endocrinol. 7:85. doi: 10.3389/fendo.2016.00085*

9

through the secretions of various adipokines that mainly regulate metabolism and inflammation (2). Conversely, brown adipocytes are multilocular, rich in mitochondria, and dissipate the energy into heat through the uncoupling protein-1 (UCP-1). Their high glucose uptake and oxidative capacities make them key players in the energy balance. At last, beige (also called "brite") adipocytes are brown-like adipocytes with UCP-1 expression that arise within white fat depots in response to cold or catecholaminergic stimulation (3).

Most adipose tissues in adult humans consist of white adipose tissue (WAT) that encompasses major subcutaneous depots (85% of total adipose tissue) in the lower or abdominal body parts and visceral fat depots (~10%) with omental, mesenteric, or retroperitoneal distributions. According to that anatomical location and the pathophysiological contexts, WATs exhibit differences in their development pattern, lipogenic and lipolytic activities, "browning" ability, or endocrine functions. This is best exemplified by the development of visceral adipose tissue that is considered as a strong predictive factor for the emergence of obesity co-morbidities, notably through the secretion of proinflammatory cytokines and an enhanced lipolysis (4). Brown adipose tissue (BAT) exhibits a more diffuse distribution with discrete small depots. Owing to the recent reassessement of BAT in adults, stimulating BAT and beige adipocyte recruitment have become promising strategies for the management of metabolic diseases (3, 5).

In addition to WAT and BAT, bone marrow adipose tissue (BMAT) emerges as a "new" fat depot that could represent up to 5% of total fat mass in adults. The presence of BMAT – also referred to as "yellow" bone marrow (BM) – was of a long-standing knowledge but has been surprisingly disregarded for many decades. Meanwhile, bone has been revealed as a target and a regulator of energy metabolism: the two main adipokines, adiponectin and leptin, modulate bone mass through indirect and direct mechanisms (6, 7) and undercarboxylated osteocalcin, the bone-derived hormone, positively impacts on whole-body glucose metabolism (8, 9). Moreover, numerous clinical studies indicate strong relationships between BMAT amount and bone loss emphasizing its potential pathophysiological role in osteoporosis. Other lines of evidence support the involvement of BMAT in hematopoiesis regulation (10) and in the pathophysiology of myeloma (11) and bone metastases (12). Such bone–fat connections have contributed to renew interest in BMAT. Yet, the anatomical dissemination of BM fat and the difficulties to study adipocytes inside bones have considerably hampered our understanding of BM adipocyte (BMA) function and its relative contribution to pathophysiological processes compared with extramedullary fat depots. In this review, we aim at highlighting the current knowledge of BMAT development and phenotype to pinpoint its original features as an adipose tissue.

## BMAT DEVELOPMENT IS BOTH PHYSIOLOGICAL AND PATHOLOGICAL

## Exploration of BMAT Development

The first descriptions of "hematopoietic red BM" replacement by the "fat yellow BM" were brought by histomorphometric studies of iliac crest biopsies in humans (13, 14) or other bone sites in animals. BMAT quantification has been, thus, performed using specific but static parameters (adipocyte number and diameter, percent of adipocyte volume per tissue volume), which precludes a reliable and dynamic assessment of adipocyte evolution according to the pathophysiological conditions. Magnetic resonance imaging (MRI) has been being of considerable interest to map non-invasively the distribution of hematopoietic BM and fatty BM in clinical studies (15). Moreover, proton magnetic resonance spectroscopy (1H MRS) allows the relative assessment of the saturated and unsaturated fatty acid composition of the fat fraction to monitor the lipid content changes. Combined with bone mineral density (BMD) and bone structure measurement by dual energy X-ray absorptiometry (DEXA) and high-resolution peripheral quantitative computed tomography, respectively, MR techniques have been being instrumental to follow BMAT development and to evaluate relationship between bone quantity and BMAT amount in diverse cohorts. Finally, using the lipid affinity of the opaque agent osmium tetroxide, a three-dimensional quantification of BMAT (whole amount, individual volume, spatial distribution) can be achieved by micro- or nanocomputerized tomography (μCT, nanoCT) in animal decalcified bones (16).

## Physiological BM Cellular Conversion to BMAT

Bone marrow adiposity development has been shown to be age, bone site, and gender dependent. At birth, bone cavities mainly contain active hematopoietic red marrow. BMAT accretion then occurs in an orderly and centripetal way: the process begins in the terminal phalanges around birth, continues in the appendicular skeleton (from the diaphysis to the distal and proximal extremities of the long bones) and finally arises in the axial skeleton (15, 17, 18). By the age of 25 years, BMAT is considered to occupy 50 (17) to 70% (16) of the BM volume, while hematopoietic BM is mainly restricted to the axial skeleton, ribs, sternum, and proximal metaphyses of humerus and femur. Afterward, the BM conversion into BMAT slowly progresses throughout the adulthood. Interestingly, women exhibit less BMAT amount compared with age-matched men (19, 20) prior the menopause age, while the following period is associated with a sharp increase of BM adiposity (21).

The development pattern in rodents is considered to be similar though it is far to be as well characterized as in humans. Histological studies of femur or tibia show the presence of BMA at adult age (22, 23), which further increases with aging (24, 25). Of note, the percentage of BM adiposity (or BMA density) appears low in rodents when compared with humans and varies according to the mouse strain. Moreover the presence of different BMA subsets could be suspected from early (26) to more recent studies (27, 28). However, it is with the new introduction of osmium tetroxide staining combined with μCT visualization that analysis of BMAT development has led to the first characterization of two BMA subpopulations in rodents (29). Scheller and collaborators propose to distinguish constitutive BMA (cBMA) – which arise first and early in life Hardouin et al. The Originality of BMAT

in distal tibia and caudal vertebrae to constitute a rather dense fat depot – from regulated BMA (rBMA) whose formation is late, increases with age, and occurs in a more scattered way in the proximal tibia, distal femur, and lumbar vertebrae (29). Importantly, this distinction based on a spatiotemporal distribution also corresponds to a different metabolic pattern and to different bone remodeling regions (see ''Specific properties of BMA versus other adipocytes''). How such classification can be extrapolated to humans is still difficult although one can suggest that cBMA reside in the feet and hands, whereas the rBMA develops in the proximal femur and lumbar vertebrae (16, 29).

## BMAT Development Differs from that of Extramedullary Adipose Tissues

Formed during gestation, BAT quantity and activity are maximal at birth to provide an efficient thermogenesis during the first weeks. Beyond puberty that is characterized by an important BAT activity, cold-activated BAT incidence remains high in young adults but rapidly declines by the age of 30 to almost disappear in the elderly (30). BAT function and WAT

Table 1 | Comparison of the main characteristics of brown adipose tissue (BAT), white adipose tissues (WAT), and bone marrow adipose tissue (BMAT) in rodents.


*a Values from Sprague Dawley female rats aged between 19 and 24 weeks (29, 43). refers to "increased," refers to "decreased," refers to "unchanged." "Whitening" refers to BAT with a changed morphology as occurrence of unilocular adipocytes with increased triglyceride storage. "Beiging" refers to WAT with a changed morphology as the occurrence of multilocular adipocytes with UCP-1 expression.*

browning also decrease with age in rodents (31) as shown in **Table 1**. Thus, the development pattern of BMAT differs from the recruitment of thermogenic adipose tissues during the lifespan and other physiological conditions [exercice, cold exposure (3, 5)], supporting that BMAT development responds to different cues.

Early WAT formation is achieved through two early periods of intense precursor proliferation with adipocyte differentiation and lipogenesis in the meanwhile. By the end of adolescence, an adipocyte number set point specific to each individual is considered to be reached and to remain constant throughout life (44), with a rather low adipocyte renewal rate estimated at 8% per year (45). Indeed, the tremendous expandability of WAT primarily relies on adipocyte hypertrophy (cell size increase). Hyperplasia (cell number increase) occurs secondly when adipocyte storage capacity is exceeded and preferentially in subcutaneous AT. WAT growth is, thus, observed throughout the adult lifespan with a maximal mass achieved at middle–early old age. In advanced old age, white fat depots are redistributed with a subcutaneous fat loss in favor of visceral fat accumulation and ectopic fat deposition in other tissues (46). BMAT development with its long-lasting hyperplasia and concomitant hypertrophy (14) throughout life exhibits some striking differences compared with WAT. Moreover, BMAT amount is generally not correlated to usual anthropometric parameters of adiposity, such as waist-to-hip ratio, amount of visceral or subcutaneous fat, or even body mass index (47–49). Altogether, BMAT formation also appears independently regulated from extramedullary WAT both in humans and rodents (**Table 1**).

## Adipose Tissues and Relationship with Skeletal Fragility: The Importance of BMAT Development

Beyond being a simple witness of age, diffuse or local BMAT accumulation has been described in several types of osteoporosis notably that associated with aging, menopause, anorexia nervosa, or glucocorticoid treatment. In humans, BMAT amount is even found inversely correlated with bone quantity in aged (47, 50), post-menopausal (51), and anorexia nervosa (52) subjects at various bone sites. Enhanced BMAT formation is also depicted in animal models of aging (53), ovariectomy (27), calorie restriction (22), or following glucocorticoid administration (54). Bone loss results from an altered bone remodeling with either decreased number and/or mineralizing function of osteoblasts as in senile osteoporosis, or with increased bone resorption by osteoclasts that overwhelms bone formation as in post-menopausal osteoporosis (55). BMAT is suspected to contribute to this unbalanced bone remodeling through the adipogenesis process *per se* or its paracrine activity (see Section ''BM Adipogenesis'' and ''Specific Properties of BMA''). However, it has to be emphasized that blocking BMAT formation failed to generate any bone modification in some mouse models (56, 57).

The impact of other adipose tissues on bone has of course been the subject of intensive clinical research that led to a complex picture. Whereas inconsistent results are drawn with subcutaneous adipose tissue measurements, visceral adipose Hardouin et al. The Originality of BMAT

tissue level, when directly quantitated, is often found negatively associated with BMD and bone quality (58). Indeed, excess visceral fat as in obesity is paralleled by an altered adipokine secretion with increased pro-inflammatory cytokines that have been suggested to interfere with bone remodeling (59). A positive association between BAT volume and BMD has been reported in a few studies in humans (60–62). Yet the prevalence and activity of BAT in humans remain difficult to measure. The beneficial impact of active BAT on bone has also been reported in mouse models (63, 64) and could rely on direct [derived-BAT adipokines (64)] or indirect mechanisms (61, 63). Moreover, bone fragility is a comorbidity of several metabolic diseases and BMAT evolution is obviously of interest in that context.

## BMAT Development in Metabolic Diseases

Regarding with type 1 diabetes, genetically or streptozotocininduced insulin deficiency results in increased BMAT in the long bones of mice (65). However, preventing BMAT formation does not impact on the bone loss (66) inherent to these models and the disease. Moreover, as reported in one study, BMAT content measured at different bone sites was unchanged in diabetic patients compared with control subjects (67). Thus, the involvement of BMAT in type 1 diabetes remains unclear and deserves further explorations.

The skeletal health in obesity has been a controversial subject (68, 69). However, most clinical and epidemiologic studies have reported an alteration of bone quality in obesity leading to an increased fracture incidence at specific bone sites (59, 70–72). Even though visceral adipose tissue quantity can be negatively associated with bone microarchitecture and strength (73), BMAT content has been poorly examined in obesity. So far, only one clinical study performed in obese premenopausal women has reported a positive correlation between visceral adiposity and vertebral BMAT (74). The Ob/Ob mouse model with extreme obesity due to spontaneous leptin disruption shows increased BM adipogenesis in the long bones (75). Nevertheless, this model remains complex to analyze because of the pleiotropic effects of leptin on bone and metabolism. The BMA amount in the long bones also increases in models of high-fat diet-induced obesity (39, 41, 76) (**Table 1**). However, divergent results make it difficult to conclude about the bone phenotype in these models and no associations have been drawn between metabolic parameters and the BMAT rise yet. Of note, animal age and diet duration appear to influence BMAT increment (41, 76), so that older mice would develop more quickly and at a higher level BMA.

Type 2 diabetic patients exhibit an increased fracture risk that is predominantly linked to a compromised bone quality (77, 78). In post-menopausal women, BMAT content in vertebrae is found unchanged by the diabetic state (79, 80), even though it is higher in patients who experienced prolonged hyperglycemia reflected by a HbA1c level above 7% (79).

Weight loss is well known to trigger bone loss, as exemplified in anorexia nervosa but also in bariatric surgery in obesity care (59). In anorexia nervosa (52), BMAT unexpectedly develops while subcutaneous and essentially visceral adipose tissues are extremely depleted. Bariatric surgery (81) or a 4-week-calorie restriction (82) in obese women leads to a marked or a more modest reduction of extramedullary fat depots, respectively. Yet, these decreases are not accompanied by significant changes in BMAT volume. These later studies rely on a limited number of subjects but suggest that the initial metabolic status [diabetes (81) or abdominal adipose tissue distribution (82)] can impact on the BMAT amount changes.

In summary, compared with other fat depots, BMAT could, thus, be considered as a strong and reliable indicator of bone integrity that could open new clinical perspectives in the management of osteoporosis (83). Its measurement rather easily performed using MRI should be more systematically analyzed particularly in metabolic diseases considering that most available data have been generated in animal models. In response to metabolic variations, BMAT development differs from other fat depots in both humans and rodents. Yet, it remains conceivable that BMAT expansion may not be completely independent from the WAT distribution. In this connection, BMAT accretion could be triggered when WAT redistribution occurs as in aging. However, the paradoxical situation of anorexia nervosa supports that BMAT formation is dissociated from that of other fat tissues. A mechanistic pathway shared by all physiopathological conditions to explain BMAT development is, thus, still missing.

## BM Adipogenesis

Bone marrow adipogenesis of resident mesenchymal stem cell (MSC) is classically controlled through a transcriptional cascade involving the key transcriptional factors PPARγ and c/EBPα. Within the BM, MSCs reside in the perivascular compartment, at the endosteal surface, and in the marrow space. As in other tissues, BM MSCs represent a heterogeneous cell population with regard to their cell surface marker expression and differentiation potential toward the osteoblast, adipocyte, or chondrocyte lineages (84, 85). It has to be emphasized that a great diversity of adipocyte precursors (distinct MSC populations, pericyte progenitors) for WAT and BAT has been revealed and debated this last decade (3, 44, 86). Interestingly, whereas most adipocytes also develop from adipose tissue-resident progenitors, engrafted BM-derived cells have been shown to generate new adipocytes within WAT in both humans (87) and rodents (88), which argues for another interaction between BM and WAT.

Importantly, adipogenesis is well considered as a competitive process for osteoblastogenesis within the BM. Haplo-insufficiency or overexpression of PPARγ in BM progenitor cells results in increased osteoblastogenesis or adipogenesis, respectively (89). Both *in vitro* and *in vivo* studies have shown that various conditions – such as elevated glucocorticoid levels, estrogen withdrawal, oxidative stress, and immobilization – that promote adipogenesis limit osteoblastogenesis. Conversely, pro-osteogenic factors [growth hormone, insulin-like growth factor 1 (IGF1), Wnt proteins, estrogens, mechanic inputs] decrease adipogenesis (85, 89, 90). Thus, the BM MSC shift toward either lineage results from a complex interplay of systemic and local mediators.

However, the classical view of an unbalanced adipogenesis at the expense of osteoblastogenesis is challenged by other findings. Hardouin et al. The Originality of BMAT

The coexistence within the mouse BM of independent precursor cells committed to each lineage has been shown (91). Both adipocytes and osteoblasts can fully develop from mouse BM MSC in a co-differentiation medium without any alteration of either lineage (92). As described in Section "Physiological BM Cellular Conversion to BMAT," a part of BMAT development occurs when the bone mass peak is achieved at human puberty and cBMA fully arise without bone loss in rodents. Bone mass can evolve independently of BM adipogenesis in some models (56, 57) or mouse strain (e.g., the C3H/HeJ strain) (93).

In this context, BMAT development cannot represent the only detrimental mechanism. BMA phenotype – which underlies its function – should be considered as a pivotal aspect in bone alterations as emphasized below.

## SPECIFIC PROPERTIES OF BMA VERSUS OTHER ADIPOCYTES

## Basic Anatomy of the "BMAT"

Compared with subcutaneous or visceral fat lobules composed of ~80% well-packed adipocytes, BMAs arise loosely connected and scattered among hematopoietic cells. In that respect, such adipocyte distribution may not be referred to as a true "adipose tissue". Whereas cBMA constitute a rather dense adipocyte area, rBMA are primarily located at the trabecular site where bone remodeling is active, which suggests their involvement in this process. The sympathetic cue is crucial in bone homeostasis (94) but the nerve distribution toward BMAs has not been depicted yet. Compared with hematopoietic areas, BMA-enriched regions are considered to be less vascularized with a lower perfusion as observed in the hip of normal-aged subjects (95). Moreover, vascularization could differ between cBMA and rBMA areas with, surprisingly, a higher capillary density in the cBMA regions in mouse (96).

Bone marrow adipocytes appear filled by a large unilocular lipid vacuole, a typical morphology of white adipocytes. Human BMA diameter can vary from 40 to 65 μm according to the fat infiltration (14), the cellular composition of BM (97), or the analyzed age (14, 98). BMAs are, thus, smaller than subcutaneous and visceral adipocytes in both humans and rodents, with cBMA larger than rBMA (**Table 1**).

## A White or Brown/Beige Phenotype?

Determining the white or brown/brite phenotype of BMA is crucial with regard to a better understanding of their regulation and function. The development factors (**Table 1**) and the unilocular morphology of BMAs are more reminiscent of a white adipocyte. However, the preferential distribution of BM adiposity in the appendicular skeleton following a temperature gradient led some authors to propose a thermogenic role for the adipocytes (99, 100). Several gene markers specific for the "brown/beige" lineage were found expressed in the whole tibia BM in young mice. Based on the expression of both white and brown markers upon rosiglitazone treatment, although UCP-1 expression was barely detected, the authors proposed that BMAT has the two phenotypes (101). A high glucose uptake detected within the BM of cold-exposed patients combined with the immuno-detection of UCP-1 protein in the BMA of very young mice convincingly support the potential presence of functional brown/beige BMA in vertebrae (102). However, these results are contradicted by other studies. Several genes of the fatty acid oxidation pathway were found more weakly expressed in isolated BMA from long bones in comparison with isolated visceral adipocytes in mouse (23). Similarly, human BMA isolated from the iliac crest had lower mRNA levels of typical markers of the brown phenotype compared with subcutaneous adipocytes (103). Importantly, a long-term cold exposure of mice led to a marked decline of BMA density in tibiae but a concomitant browning of other WAT depots (29). Altogether, one may suggest that the BMA phenotype varies according to their localization: BMAs that arise in the long bones most likely belong to the white lineage, whereas some BMAs in the vertebrae can exhibit brown-like features. The presence of such thermogenic adipocytes could explain the disappearance of the "yellow BM" in the caudal vertebrae during a temperature rise (104). Indeed, hematopoiesis remains quite efficient in the vertebra BM and brown adipocytes have been proposed to support myelopoiesis through the secretion of some specific adipokines (102).

Most studies characterizing BMA have been performed from long bones or iliac crest samples and have raised some puzzling specificities regarding their paracrine/endocrine and metabolic properties.

## Secretory Profile of BMA

The two main adipokines adiponectin and leptin are detected in isolated mature BMA, could locally promote osteoblastogenesis, and impact on osteoclastogenesis or hematopoiesis (89). However, their mRNA expression levels were found markedly low compared with extramedullary adipocytes in adult healthy donors (103) and normal-aged mice (23).This finding may relate to a lower state of differentiation of BMA as proposed in aging mice (23). Yet BMAT explants from rabbit (distal tibia) or human patients (tibia) were shown to secrete more adiponectin than other WAT depots. Moreover, through the characterization of a mouse model of defective BM adipogenesis, BMAT was revealed as a significant source of serum adiponectin during caloric restriction (34). Even though this contribution to the calorie restriction-induced adiponectinemia rise was not confirmed in rabbits (35), this study underlies a true endocrine status of BMA which can modulate metabolism at the systemic level.

Given the importance of inflammatory factors in the bone loss of post-menopause or aging, the cytokine production by BMA has begun to be assessed. A first antibody-based array study indicates that the cytokine content of BMA diverges from that of subcutaneous adipocytes with a more pro-adipogenic and proapoptotic profile in aging mice (105). In another independent transcriptomic study in mice, BMAs were reported to express inflammatory genes, such as TNFα and IL6, at higher levels than visceral adipocytes (23). Surprisingly, in both studies, their expression levels were downregulated during aging, suggesting that this pro-inflammatory profile of BMA is rather early and transient. Of note, another transcriptional analysis failed to detect any further upregulation of these genes in the BMA from normalaged mice fed with a high-fat diet (24). Interestingly, primary human femoral BMAs express the pro-osteoclastogenic factor RANKL and mediate through a direct cell contact the differentiation of osteoclast precursors (106, 107). RANKL expression was also shown to be associated with BMA differentiation and with Pref1-expressing preadipocytes in the BM of aging mice (108).

Several *in vitro* studies using notably co-culture models (109, 110) of BM MSC-derived adipocytes and osteoblasts strongly support a deleterious paracrine role of BMA. BM MSCderived adipocytes have also been reported to secrete factors that alter osteoblastogenesis and favor adipogenesis, such as Wnt signaling inhibitors (111) and chemerin (112) (**Figure 1**).

## Metabolic Profile of BMA

Adipocytes synthesize and hydrolyze triglycerides in response to insulin and catecholamines, the respective prototypical lipogenic and lipolytic factors. Functional assays to establish the relative capacity of BMA for lipogenesis and lipolysis compared with other fat depots remain scarce due to technical difficulties. Only one early study in rabbits has reported that isolated femoral adipocytes incorporate palmitate into triglycerides at level equivalent to perirenal adipocytes (113). Catecholamine-induced lipolysis appeared lower in femur compared with omental adipose tissue in dogs (114). More indirectly, but in line with their small size, mouse BMAs were shown to express weaker expression of key genes involved in lipogenesis and lipolysis compared with visceral adipocytes (23). Moreover, the metabolic activity of BMA has been indirectly assessed by the adipocyte size and density determination and seems to vary according to their bone localization and the different subtypes in rodents. Indeed, only BMA in the lumbar spine were affected by a chronic administration of a β3-adrenergic agonist (27). In the tibia, cBMA were reported resistant to a long-term cold-induced catecholamine challenge compared with the regulated subpopulation (29). Whether these observations result from the development timing of each BMA subset, a different innervation/vascularization of the BM regions or intrinsic divergences among BMA subtypes has to be further explored. Nevertheless, the relative amount of BMAT compared with other adipose tissues supports that the metabolic influence of BMA is locally confined.

Importantly, the fatty acid composition of BMA is emerging as a key signature of their phenotype. Compared with subcutaneous adipose tissue, lipid extracts of human BM (mainly from proximal tibia or femur) display a higher proportion of saturated fatty acids and a lower amount of monounsaturated ones (115). Based on 1H MRS exploration, the unsaturation index of the

Figure 1 | The bone marrow adipocyte as an active cell in bone remodeling. The bone marrow adipocyte (BMA) can secrete leptin and adiponectin, which directly stimulate the proliferation and the differentiation of mesenchymal stem cell (MSC) into osteoblast. Other factors, such as the Wnt signaling inhibitor sFRP1 or chemerin, can be released by the BMA: these factors were shown to promote adipogenesis while negatively regulating osteoblastogenesis. As a source of RANKL and TNF-α, the BMA can also support osteoclastogenesis. The BMA appears enriched in saturated fatty acids that can trigger the dysfunction and apoptosis of osteoblasts. Saturated fatty acids could impact on osteoclast differentiation and survival but results are conflicting. The BMA could release these different factors in response to catecholamines, glucocorticoids, and cytokines.

vertebral fat fraction is decreased compared with controls and inversely associated with bone loss (51) or fragility fractures (80) in post-menopausal women. Moreover, in healthy adult women, a gradual increase of the unsaturation index is observed from the proximal femur to the most distal part of the tibia, suggesting a differential fatty acid distribution in the BMA subpopulations. Accordingly, this was further examined in several bone sites in rats to reveal a higher unsaturation index in cBMA compared with regulated ones (29). These studies strongly support that BMA associated with bone alterations preferentially store saturated fatty acids. This has a strong pathophysiological relevance since *in vitro* chronic exposure of osteoblasts to saturated fatty acids triggers their dysfunction and apoptosis (116–118). Results are conflicting for the resorbing cells since saturated fatty acids were reported either to reduce osteoclastogenesis (119) or to exert beneficial effects on mature osteoclasts by preventing their apoptosis (120). Though the amount of fatty acids released by BMA appears rather low *in vitro* (121), BMA could contribute to bone alterations through a detrimental fatty acid-mediated process referred as lipotoxicity. Moreover, a specific lipid pattern of BMA may be considered as a discriminative trait, which is rather puzzling considering that dietary fat intake usually impacts on the fatty acid composition of bone (122), adipose tissues, and blood (123). The variation in the BMA fatty acid content could reflect an adaptation to systemic metabolic alterations (79, 80) and/or an intrinsic characteristic through the expression of desaturases (29).

## Regulation of BMA Function

Directly assessing the factors regulating the BMA function as done through *ex vivo* experiments with other fat depots is hampered by the dissemination and the scarcity of BMAT in animal models. Alternatively, indirect measurements of BMA activity through the size and density determination have often been performed using long-term experimentations (27, 29, 40). However, in such studies (27, 40), experiments are often carried out while BM adipogenesis is also initiated, which may interfere with the interpretation of the mature BMA response. One may propose to perform tests when BM adiposity (and mostly rBMA) is fully developed: indeed, a few interventional studies have already uncovered for example a sensitivity to estradiol for BMAT in post-menopausal women (124, 125) or to growth hormone for tibial rBMA in dwarf rats (126). Moreover, using radiolabeled glucose or fatty acid analogs with Positron Emission Tomography imaging could be helpful to decipher the metabolic function and regulation of BMA during acute challenges. So far, information regarding the regulation of BMA function mainly comes from *in vitro* studies (**Figure 1**).

As mentioned before, like for other WAT, catecholamines or agonists of the β-adrenergic receptors stimulate fatty acid release from BMA both *in vitro* (121, 127, 128) and *in vivo* (114). Surprisingly, insulin addition does not seem necessary for human BM MSC adipogenesis and subsequent lipogenesis *in vitro* (109, 127, 129). However, insulin acutely stimulates glucose uptake (130) and modifies the adipokine pattern (127, 131) *in vitro* and in organotypic cultures supporting that BMAs are sensitive to insulin.

Dexamethasone triggers lipolysis (118) and stimulates the secretion of leptin from human BM MSC-derived adipocytes (118) or human primary cultivated BMAs (131) like reported for extramedullary white adipocytes. Regarding to adiponectin secretion, discrepancies are obtained according to the studied model (118, 131). Dexamethasone also stimulates the expression of RANKL in human primary BMAs (106).

Similarly to white adipocytes, cytokines – including notably TNFα – alter leptin secretion in two cultured models (131, 132), supporting that BMAs are also sensitive to a pro-inflammatory environment.

## CONCLUSION

Bone marrow adipose tissue is metabolically distinct from other fat depots as indicated by several evidence related to its development and properties. Numerous data from clinical studies, *in vivo*, and vitro models point to an involvement of BMAT in bone remodeling resulting in bone alterations. Besides the participation of the adipogenesis process, BMAs are active cells whose phenotype is strikingly heterogeneous according to the bone site or BM area. Compared with cBMA, rBMA appear highly responsive to diverse stimuli. The first characterizations of BMA as well as their sensitivity to catecholamines and glucocorticoids support that BMA behaves more as a white-like adipocyte whose fatty acid or adipokine release could impact on BM cells (**Figure 1**). Moreover, its phenotype could also evolve when bone remodeling is altered. Owing to its specific microenvironment, it is expected that osteoblasts, osteoclasts, hematopoietic, and other bone cells reciprocally regulate BMA function. In that respect, a description of BMAT vascularization, innervation, and interaction with BM cells is needed to better comprehend its regulation. Moreover, the relative contribution of BMAT secretions compared with those of extramedullary adipose tissues remains poorly examined, which emphasizes the necessity to deepen the characterization of BMA. So far considering the different physiopathological situations of BMAT development together has not led to a shared hypothesis defining the role of BMA. This actually suggests that, as a true responsive cell, BMA function and regulation could vary according to the pathophysiological context and should be analyzed accordingly.

## AUTHOR CONTRIBUTIONS

SL drafted the review. TR and SL made the figure. PH, TR, and SL contributed to the design of the review, critically revised it, approved the final version to be published, and agreed to be accountable for all aspects of the work.

## ACKNOWLEDGMENT

Writing of this review was enabled by ULCO University.

## FUNDING

The study was supported by ULCO University and The French Society of Rhumatology (SFR).

## REFERENCES


fat after a four-week calorie restriction in obese women. *J Magn Reson Imaging* (2015) 42:1272–80. doi:10.1002/jmri.24908


**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 © 2016 Hardouin, Rharass and Lucas. 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.*

# Qualitative Aspects of Bone Marrow Adiposity in Osteoporosis

*Ana María Pino1 , Melissa Miranda1 , Carolina Figueroa2 , Juan Pablo Rodríguez1 and Clifford J. Rosen2 \**

*<sup>1</sup> Laboratory of Cell Biology, INTA, University of Chile, Santiago, Chile, 2Maine Medical Center, Portland, ME, USA*

The function of marrow adipocytes and their origin has not been defined although considerable research has centered on their presence in certain conditions, such as osteoporosis. Less work has focused on the qualitative aspects of marrow fat. Bone marrow serum is composed of multiple nutrients that almost certainly relate to functional aspects of the niche. Previous studies using non-invasive techniques have shown that osteoporotic individuals have more marrow fat and that the ratio of saturated: unsaturated fatty acid is high. We recently reported that bone marrow sera from osteoporotic patients with fracture showed a switch toward decreased content of total saturated versus unsaturated fatty acids, compared to patients without fracture highlighting a dynamic relationship between the composition of fatty acids in the bone microenvironment and the metabolic requirements of cells. The relative distribution of fatty acids differed considerably from that in the serum providing further evidence that energy utilization is high and that marrow adipocytes may contribute to this pool. Whether these lipids can affect osteoblast function in a positive or negative manner is still not certain but will require further investigation.

#### *Edited by:*

*Xinhua Qu, Shanghai Ninth People's Hospital, China*

#### *Reviewed by:*

*Jan Tuckermann, University of Ulm, Germany Jawed Akhtar Siddiqui, University of Nebraska Medical Center, USA*

> *\*Correspondence: Clifford J. Rosen cjrofen@gmail.com*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 04 August 2016 Accepted: 10 October 2016 Published: 25 October 2016*

#### *Citation:*

*Pino AM, Miranda M, Figueroa C, Rodríguez JP and Rosen CJ (2016) Qualitative Aspects of Bone Marrow Adiposity in Osteoporosis. Front. Endocrinol. 7:139. doi: 10.3389/fendo.2016.00139*

Keywords: lipids, fatty acids, unsaturated, bone marrow cells, bone marrow examination, adipocytes

## THE BONE MARROW MICROENVIROMENT

In adults, the rigid bone structure of the cortex encloses a dynamic and flexible cell organization sustaining continuous bone and bone marrow stroma remodeling, as well as blood formation. These processes rest on preserving specific adult stem cells (SCs) that are characterized by their capacity for self-renewal and multilineage differentiation. These unique properties of SCs are not only cellautonomous *in vivo*, but also controlled by their surrounding microenvironment, which is currently called the SC niche. The original Schofield's proposition for hematopoietic stem cells (HSCs) in the bone marrow emphasized the input from other marrow cells types to maintain SCs behavior and prevent maturation (1). Such cell- depending interactions provide a specialized microenvironment ("niche") that allows cell lodging while maintaining self-renewal of SCs; loss of such an association would lead to cell differentiation. At present, the niche concept has been assumed to explain the behavior of SCs in several tissue types while evolving to take account of specific cell types, anatomical sites, soluble molecules, signaling cascades and gradients, as well as physical factors, such as weight bearing, shear stress, oxygen tension, and temperature (2–5). Among the most studied somatic SCs in mammals are the HSC and mesenchymal stromal/stem cells (MSCs) that integrate a unique niche in the bone marrow made of heterotypic SC pairs (6). Therefore, the bone marrow niche concept comprises a complex microenvironment, providing spatial and temporal coordinated signals to support SC function; balanced inputs from the niche can sustain homeostatic SC self-renewal and differentiation, but under pathological conditions they could constrain SCs functioning.

## MESENCHYMAL STEM CELLS OR MARROW STROMAL CELLS

In the bone marrow, MSCs mainly localize lining blood vessels, in particular on the sinusoids, the characteristic vessels of bone marrow microcirculation (7, 8), often in close association with HSCs (6, 9, 10). Then, the sinusoidal wall brings together both HSCs and MSCs, framing and sustaining a mutual niche (6, 11). Such perivascular localization has also been demonstrated for MSCs in several tissues through the body, which support the proposition that MSCs are fundamental to the healing of many tissues. MSCs in the bone marrow are ontogenically distinct populations taking over various biological functions, such as niches for HSCs (6, 12, 13), progenitor cells for bone formation during bone remodeling or repair (14), for cartilage and adipocyte formation, and vascular support (15–17).

*Ex vivo*, hMSCs are relatively easy to obtain from a small bone marrow aspirate and they can simply be isolated from other marrow nucleated cells by their adherence to plastic dishes. Subsequently, these cells can be expanded under *in vitro* conditions in which they display varying proliferation and differentiation potential; however, as cells are expanded consecutively under standard conditions, they lose some of these capacities. Clonal analysis demonstrated that MSCs obtained from the bone marrow are a heterogeneous mixture of cells that differ in their stage of lineage commitment and extent of differentiation (18–21). Therefore, MSCs populations in the bone marrow or those that are isolated and maintained in culture are not homogeneous, but rather assemble progenitor cells with diverse biological properties, such that not all MSCs are alike.

In theory, the lineage fate of MSCs appears to be determined during cell "commitment," at very early stages of cell differentiation. During this almost unknown period, both intrinsic (genetic) and environmental (local and/or systemic) conditions interplay to outline the cell's fate toward one of the possible lineages. Factors, such as age (22), culture condition (23), microenvironment (24), mechanical strain (25), and some pathologies (26, 27), appear to affect the intrinsic activity of MSCs.

Multilineage differentiation capacity has been related to wide range gene expression at intermediary levels, which along cell commitment and differentiation shifts to a selective mode of restricted genes expressed at high level (21, 28–30). Several fundamental signaling pathways participate in regulating the lineage commitment of MSCs, including transforming growth factor-beta (TGFβ)/bone morphogenic protein (BMP) signaling, wingless type MMTV integration site (Wnt) signaling, Hedgehogs (Hh), Notch, and fibroblast growth factors (FGFs) (31, 32).

## RELATIONSHIP BETWEEN THE OSTEO/ ADIPOGENIC PROCESSES THE OBESITY THEORY OF OSTEOPOROSIS

The formation, maintenance, and repair of bone tissue depend on fine-tuned interlinks in the activities of cells derived from the two SC types housed in the bone marrow interstice, above described. HSCs along the myeloid differentiation lineage generate osteoclasts (33), whereas osteoblasts derive from MSCs, which are also progenitor cells for adipocytes (34). Bone diseases, such as osteopetrosis, osteopenia, and osteoporosis, show imbalance between bone formation and resorption (33).

Since MSCs can generate several cell types, the control of lineage commitment of MSCs appears critical. Alteration in such control has been observed in some bone diseases resulting in abnormal bone remodeling, which show divergent commitment of MSCs to adipocytes and osteoblasts. For instance, increased marrow fat content has been demonstrated in osteoporosis patients, the most common bone remodeling disorder (35, 36). Also, this alteration of the osteo/adipogenic processes is observed in other bone loss conditions, such as aging, immobilization, microgravity, ovariectomy, diabetes, and glucocorticoid or thiazolidinedione treatments, highlighting the harmful consequence of marrow adipogenesis in osteogenic disorders (37–39).

Osteoblasts and adipocytes originate from a common precursor, MSCs; preserving bone tissue requires adequate osteoblastic differentiation while minimizing adipogenesis. Commitment and differentiation of MSCs into a specific phenotype *in vivo* is postulated to be controlled by hormonal and local factors (paracrine/autocrine) regulating the expression and/or activity of master differentiation genes (32, 40). Research on MSCs differentiation *in vitro* showed that activation of PPARγ2 and C/ EBPs match to the master transcription factors for adipogenic differentiation (41–43), while Runx2 and osterix are required for osteogenic differentiation (44). PPARγ2 positively regulates adipocyte differentiation while acting as a dominant negative regulator of osteogenic differentiation (45, 46). By contrast, an increase in bone mass density was observed in PPARγ2-deficient mice model (47). On the other hand, Runx2 expression by MSCs inhibits their differentiation into adipocytes, as shown by experiments in Runx2/calvarial cells, which spontaneously differentiate into adipocytes (48). However, due to the heterogeneity of the earliest MSC, cell expression of only adipogenic or osteogenic markers is unlikely. Indeed, osterix positive marrow adipocytes have been shown in lineage tracing studies, and marrow adipocytes can be traced to peroxiredoxin 1 (Rosen, personal communication). Moreover, some Runx2-positive cells isolated in high marrow adiposity states also have large lipid droplets and express perilipin (49).

Thus, in most cases, there is a reciprocal relationship in the regulation of these differentiation processes whose alteration would facilitate adipose accretion in the bone marrow, at the expense of osteoblast formation, decreasing bone mass (50–52). Such altered conditions would prevail in the bone marrow of osteoporotic patients and other bone loss conditions, disrupting the activities of MSCs and their microenvironment (37, 40, 50). This proposition has been termed the obesity theory of osteoporosis.

According to this proposition, observations in iliac crest biopsies from elderly women show considerable accumulation of adipocytes; i.e., 70–80% of the total marrow volume in osteoporotic patients, compared to that of healthy elderly women. More recently, observations done by magnetic resonance imaging (MRI) allowed similar conclusion in patients with low bone density (53–55). However, the origin of marrow adipocytes and their true nature has been more difficult to discern.

## HUMAN BONE MARROW FAT

In newborn mammals, there is scarce marrow fat; however, adipocyte number increases with age such that in humans older than 30 years, most of the femoral cavity is occupied by adipose tissue (56). The function of human marrow fat was largely overlooked; at first, it was considered "filler" for the void left by trabecular bone during aging or after radiation treatments. Early observations from the 1970s highlighted the different characteristics between adipocytes within the red and yellow marrow, suggesting that marrow adipocytes may have a region-specific lipid composition (57, 58). In humans, adipocytes accumulate within the yellow marrow at or slightly before birth, regardless of prematurity, and accelerate between 4 and 8 weeks of age (59, 60). Early adipose tissue is established in distal skeletal regions, including the hands, feet, and distal tibia. This is often called constitutive marrow fat. After attaining some bulk, distal marrow adipose tissue histologically resembles peripheral white adipose tissue, and is relatively devoid of active hematopoiesis. During adulthood, marrow fat accretion continues in areas of red, hematopoietic marrow (61), which show histologically single adipocytes interspersed with sites of active hematopoiesis. These regions appear to be spatially distinct in rats and mice; but in human, both types of adipose tissue may allocate in the same skeletal region (62).

In a recent study in mice, Scheller et al. (62) related regionspecific fat differences in development to functional differences, such as regulation, adipocyte size, lipid composition, gene expression, and genetic determinants of marrow adipocytes, proposing that distal fat areas embody constitutive marrow adipose tissue (cMAT), while proximal adipose tissue contains regulated marrow adipose tissue (rMAT). Although cMAT showed histologic similarities with white extra-medullar adipose tissue (WAT), the lipid composition, cold regulation, and gene expression of WAT is more related to that of rMAT, concluding that lipid metabolism in WAT and rMAT adipocytes may be similar. Of note, cMAT showed increased unsaturation index compared to both, rMAT and WAT. Scheller's study and those of others highlight that not all marrow adipocytes are equivalent, sustaining connotation for the marrow niche and its relationship to skeletal and whole-body metabolism.

Bone marrow fat has a role in systemic and local energy metabolism (63–66). When dysfunctional, it may disrupt the complex relationships in the bone marrow microenvironment, constraining functioning of both HSCs and MSCs, and thereby affecting their progeny (67, 68).

Bone marrow adipocytes have endocrine, autocrine, and paracrine effects; the activity of these cells on neighboring marrow result from the production of adipokines, steroids, cytokines, and free fatty acids (51, 64, 69–71), which could sustain or suppress the hematopoietic and osteogenic processes (10, 50, 51, 63, 64, 72).

The modification of bone marrow adipose tissue, either in size or function, may have similar consequences to that of extra medullary fat, which by unbalanced production of signaling products underlay in several human diseases, including obesity, lipodystrophy, atherogenesis, diabetes, and inflammation.

## MARROW FAT COMPOSITION AND OSTEOPOROSIS

Current information on bone marrow fat points to diverse adipocytes that have development-dependent skeletal distribution, showing distinct properties, such as lipid composition, gene expression, and probably subjected to different peripheral and local regulation. Underlying the relationship between such bone marrow fat and bone function is a balance in the number and quality of adipose cells; hence, the interest in characterizing bone marrow fat and searching for the quality of human marrow fat, for instance, in a bone remodeling disorder like osteoporosis.

Skeletal and marrow cells utilize fatty acids as substrates for their energy needs (73), although the contribution of marrow adipocytes to such requirements is most unknown; as it is the role of fatty acids stored in the marrow adipocytes in bone remodeling and hematopoiesis. Moreover, the lipid composition in the interstitial compartment surrounding bone marrow cells, its physiological significance, its variation under a bone stress situation, or its relationship with systemic lipids are undefined.

Studies on the composition of bone marrow fat suggested that it could have clinical relevance, in addition to marrow fat size and volume. Such conclusion is derived from *in situ* non-invasive MRI-based analysis of bone marrow, in patients with chronic diseases, such as osteoporosis (54), type 2 diabetes mellitus (74), or diabetic and non-diabetic women with prevalent fragility fractures (75). These studies demonstrated significantly lower unsaturation of bone marrow lipids in patients than corresponding controls. Moreover, Patsch and colleagues showed that low unsaturation and high saturation levels of lipids was accentuated in diabetic patients with prevalent fractures (75).

Physiological levels of molecules in the human bone marrow milieu are practically unknown; measurement has been particularly difficult not only because of tissue seclusion, but because of the complicated anatomy and blood perfusion of bone requiring invasive bone marrow sampling. Thus, knowledge on the availability to bone marrow cells of metabolites or regulatory compounds is scarce, limited to some pathologic condition or estimated from measurements in plasma (69, 76–78). A soluble fraction, named "bone marrow serum (BMS)," obtained after spinning human bone marrow aspirates allows measurement of physiologically significant concentrations in the bone cell microenvironment, which are dissimilar from circulating plasma or whole blood (70, 79, 80).

In such fraction, we tested the hypothesis that BMS reflects the uniqueness of the marrow compartment in respect to the sequestration and utilization of fatty acids, by studying the composition of total soluble fatty acids in BMS and plasma samples from control and osteoporotic women with and without hip fracture (81). Results revealed a specific pattern of fatty acid composition in the bone marrow milieu that was characterized by higher saturated and decreased unsaturated fatty acids, compared to that of the circulation. The proportion of fatty acids in the BMS varied in the range of lipids in plasma, but its relative levels of unsaturated and saturated fatty acids compares well to those observed in the human marrow fat tissue (62, 75, 82), implying a distinct lipid content in BMS that replicates relevant fatty acids derived from the activity of adipocytes and other marrow cells, rather than those provided by blood plasma. The lipid content in BMS of all women studied suggests an active exchange of lipids between marrow cells; thus, the surplus of saturated fatty acids in BMS could result from palmitic and stearic acids supplied mainly by adipocytes, while marrow cells apparently preserve their pool of unsaturated fatty acids, except for oleic acid that is higher in BMS than in plasma (81).

By contrast, when the content of fatty acids in BMS and blood plasma was evaluated according to women's bone mass density, no consistent differences in fatty acid composition were apparent in BMS, nor in plasma, which is analogous to former observations on the composition of bone fat tissue (53, 54, 74). Notwithstanding, there was a trend toward higher saturation and lower unsaturation in BMS fatty acids as compared to plasma. The analysis of the relative proportion of individual fatty acids suggests an active distribution of unsaturated fatty acids.

The former conclusion is supported by further analysis of data on the composition of fatty acids in both BMS and plasma in the osteoporosis group, in relation to the presence or not of a hip fracture (**Figure 1**). A switch toward decreased content of total saturated versus unsaturated fatty acids was observed in BMS of women with fractures, emphasizing a dynamic relationship between the composition of fatty acid in the bone microenvironment and the metabolic requirements of cells. In women with fractures, stearic acid content significantly decreased concomitant with increased oleic acid content, implying a substrate to product relationship to fulfill specific requirements for unsaturated fatty acids. Moreover, increased activity of cyclooxygenase (COX), COX-2 could be proposed from the total removal of polyunsaturated eicosatrienoic and arachidonic fatty acids (81).

Figure 1 | Relationship between hip fracture and fatty acid composition in the bone marrow serum of osteoporotic women.

## REFERENCES


It is difficult to define the origin (adipocyte or other marrow cell production) and type of lipids that change the fatty acid content of the BMS fraction after fracture, in part because the measurement was limited to the total fatty acid pool in the soluble fraction of the bone marrow. After a hip fracture bone cell metabolism may change in order to provide extra energy for damage repair by osteoblasts, or to suppress excessive inflammation. Interestingly, the lipid fraction in BMS of fractured women appears adjust to such conditions, thus diminished level of saturated fatty acid could result from increased supply of energetic substrate to marrow cells (bone, fat, and hematopoietic), while an increased exchange of oleic and polyunsaturated fatty acids could be related to their regulatory functions.

In conclusion, in this report, we summarize evidence on the role that adipocytes may play in the bone marrow in health and disease. Bone marrow adipose tissue appears as distinct from other extra marrow fat tissues, heterogeneous in its development, and allocated in subpopulations that seem to have different physiological properties. The regulation of the number and quality adipocytes is a necessity to preserve a functional bone marrow microenvironment to sustain SCs. In a bone disease like osteoporosis, there is increased differentiation of the precursor MSCs toward adipocytes, altering the local marrow microenvironment but leading to bone loss. In addition to fat mass, the quality of lipids could also participate in disrupting local regulation; hence, the interest in knowing the fatty acid composition of marrow adipose tissue. The measurement of the fat composition in BMS portrays a dynamic exchange of fatty acids in this fraction implying the connotation of locally produced lipids. The fatty acid composition in the BMS is enriched in saturated fatty acids and decreased in unsaturated fatty acids, as compared to blood plasma. Therefore, it is clear that qualitative aspects of marrow adiposity provide significant insights into the metabolism of the bone marrow niche, and may offer clues as to the pathophysiology of bone diseases, such as osteoporosis.

## ETHICS STATEMENT

This study was approved by the INTA, University of Chile IRB.

## AUTHOR CONTRIBUTIONS

JR did the research and helped prepare the manuscript. AP did the background work. CF also performed some of the experiments.

## FUNDING

This work was supported by a grant from the Fondo Nacional de Ciencia y Tecnología (FONDECYT·#1160214) and NIDDK R24 092759.


bone-marrow under conditions of hypothermia is not associated with stimulation of erythropoiesis. *Mutat Res* (2008) 656:8–13. doi:10.1016/ j.mrgentox.2008.06.016


transcription of osteocalcin in osteoblasts. *J Biol Chem* (2003) 278:23270–7. doi:10.1074/jbc.M211610200


**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 © 2016 Pino, Miranda, Figueroa, Rodríguez and Rosen. 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.*

# MR-Based Assessment of Bone Marrow Fat in Osteoporosis, Diabetes, and Obesity

*Christian Cordes1 \*, Thomas Baum1 , Michael Dieckmeyer1 , Stefan Ruschke1 , Maximilian N. Diefenbach1 , Hans Hauner2 , Jan S. Kirschke3 and Dimitrios C. Karampinos1*

*1Department of Diagnostic and Interventional Radiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany, 2Else Kröner Fresenius Center for Nutritional Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany, 3Section of Diagnostic and Interventional Neuroradiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany*

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Graham J. Galloway, Translational Research Institute, Australia Miriam A. Bredella, Massachusetts General Hospital, USA*

> *\*Correspondence: Christian Cordes c.cordes@tum.de*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 19 April 2016 Accepted: 14 June 2016 Published: 27 June 2016*

#### *Citation:*

*Cordes C, Baum T, Dieckmeyer M, Ruschke S, Diefenbach MN, Hauner H, Kirschke JS and Karampinos DC (2016) MR-Based Assessment of Bone Marrow Fat in Osteoporosis, Diabetes, and Obesity. Front. Endocrinol. 7:74. doi: 10.3389/fendo.2016.00074*

Bone consists of the mineralized component (i.e., cortex and trabeculae) and the non-mineralized component (i.e., bone marrow). Most of the routine clinical bone imaging uses X-ray-based techniques and focuses on the mineralized component. However, bone marrow adiposity has been also shown to have a strong linkage with bone health. Specifically, multiple previous studies have demonstrated a negative association between bone marrow fat fraction (BMFF) and bone mineral density. Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) are ideal imaging techniques for non-invasively investigating the properties of bone marrow fat. In the present work, we first review the most important MRI and MRS methods for assessing properties of bone marrow fat, including methodologies for measuring BMFF and bone marrow fatty acid composition parameters. Previous MRI and MRS studies measuring BMFF and fat unsaturation in the context of osteoporosis are then reviewed. Finally, previous studies investigating the relationship between bone marrow fat, other fat depots, and bone health in patients with obesity and type 2 diabetes are presented. In summary, MRI and MRS are powerful non-invasive techniques for measuring properties of bone marrow fat in osteoporosis, obesity, and type 2 diabetes and can assist in future studies investigating the pathophysiology of bone changes in the above clinical scenarios.

Keywords: bone marrow, magnetic resonance imaging, magnetic resonance spectroscopy, osteoporosis, diabetes, obesity

## INTRODUCTION

Bone consists of the mineralized (i.e., cortex and trabeculae) and the non-mineralized component (i.e., bone marrow). The interaction of the mineralized and non-mineralized components plays an important role in bone loss pathophysiology. Quantitative measurements of the mineralized component have been traditionally performed by using dual-energy-X-ray-absorptiometry (DXA) or quantitative computed tomography (QCT) assessing bone mineral density (BMD) (1). BMD is used in clinical routine to determine osteoporosis-associated fracture risk. Osteoporosis is defined as a skeletal disorder characterized by compromised bone strength predisposing an individual to an increased risk for fractures (2). It is classified as a public health problem, since osteoporosis-related fractures are associated with a reduction in quality of life and an increased morbidity and mortality (3, 4). However, BMD accounts for only 60–70% of the variation in bone strength (5) and the BMD values of subjects with and without osteoporotic fracture overlap (6). Measurements of trabecular bone microstructure, based on high-resolution imaging techniques, in addition to BMD have shown to improve the prediction of the variation in bone strength (7–9).

Despite the focus on the mineralized bone component for osteoporosis diagnostics, recent studies have highlighted the potential role of the non-mineralized bone component in bone health (10–14). Bone marrow fills the cavities of trabecular bone and it primarily consists of adipocytes (yellow marrow regions) or adipocytes and hematopoietic red blood cells (red marrow regions). It is well known that osteoporosis is associated with an increased bone marrow fat mass due to a shift of differentiation of mesenchymal stem cells to adipocytes rather than to osteoblasts (11, 13, 15). Multiple studies have shown that higher bone marrow fat fraction (BMFF) values are associated with lower BMD values (12, 14, 16–26).

Bone marrow fat has a distinctly different function compared to other white and brown adipose tissue depots or ectopic fat depots in the human body (27) and might play a role in the pathophysiology of metabolic disorders. Metabolic diseases, including obesity and diabetes, are known to have a complex and still poorly understood relationship to bone health. There seems to be a higher fracture-related morbidity in obese than in nonobese women (28), and more fractures in diabetic than in healthy subjects (29–32). In addition, there is a growing interest to study the relationship between bone marrow fat mass and fat in other depots regarding obesity and type 2 diabetes mellitus (T2DM).

In this context, imaging biomarkers are emerging to non-invasively study the properties of the non-mineralized component of bone (33). Dual-energy CT (DECT) allows to simultaneously determine bone marrow fat and BMD (34). In contrast to DECT, magnetic resonance (MR) allows the quantitative assessment of bone marrow without radiation exposure. MR enables reliable measurements of bone marrow water-fat and fatty acid composition with different methods including magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) (35, 36).

The purpose of the present work is to review the currently available literature on MR-based assessment of bone marrow fat in the context of osteoporosis, T2DM, and obesity.

## MR-BASED ASSESSMENT OF BONE MARROW FAT

## Literature Research

Electronic searches in PubMed (http://www.ncbi.nlm.nih.gov/ pubmed) were performed up to March 2015 to identify relevant studies for this review. No starting date was entered for the electronic search to obtain the entire literature available in PubMed. Search terms used included "Bone Marrow," "Bone Marrow Fat," "Bone Marrow Adipose Tissue," "Magnetic Resonance Imaging," "Magnetic Resonance Spectroscopy," "Osteoporosis," "Obesity," and "Diabetes." The search was restricted to studies in humans. The reference lists of relevant articles were also screened.

## MR Methods

T1-weighted imaging (T1WI), MRS, and chemical shift encoding-based water–fat imaging have been previously used to assess BMFF.

## T1-Weighted Imaging

T1-weighted imaging is not technically demanding and has been mostly applied on the pelvis, hip, and spine. A recent study has even introduced a score analog to the DEXA *t*-score based on the mean signal-to-noise ratio of the L1–L4 vertebral bodies in T1WI (37). Measurements of bone marrow fat volume have been also proposed by applying thresholds on T1WI to extract bone marrow fat voxels (25). The applied threshold was usually set at the same gray-scale level as subcutaneous adipose tissue. The intra- and interobserver reproducibility for the assessment of bone marrow fat volume in T1WI images expressed as coefficient of variation (CV) amounted 0.9% (intraobserver) and 2.2% (interobserver) (for the post-processing) (38). The main error source for the calculation of bone marrow fat volume based on T1WI results from partial volume effects and threshold selection, especially in regions with red marrow.

### Magnetic Resonance Spectroscopy

Proton-MRS (1H-MRS) is considered as the MR gold standard for bone marrow fat quantification (**Figure 1**). Point-resolved spectroscopy (PRESS) and stimulated echo acquisition mode (STEAM) single-voxel 1H-MRS sequences have been most commonly used for the characterization of the fat spectrum in the bone marrow at the pelvis, spine, and hip. Based on MR spectra, fat fraction and fatty acid composition parameters can be determined (36). The average CV of vertebral BMFF was reported to be 1.7% (on the same day with repositioning). Griffith et al. recruited 36 subjects who underwent MRS in the femoral neck and head, as well as in the subtrochanteric region of the femoral shaft to measure bone marrow fat content (39). They found the best reproducibility (between two measurements within 1 week) in the femoral head (interclass correlation 0.85), followed by the femoral shaft (interclass correlation 0.83), and the femoral neck (interclass correlation 0.78). Another recent study that examined reproducibility of 1H-MRS vertebral BMFF found a CV of 9.9% (SD 0.08) for a 6-week reproducibility and a CV of 12.3% (SD 0.10) for a 6-month reproducibility (40).

However, most of the above studies did not account for the difference in T2 relaxation times between the water- and fat components and measured a signal-weighted fat fraction at a single echo time (TE), dependent on the employed TE value and field strength. A multi-TE MRS measurement combined with T2 estimation routines instead removes the confounding effects of water–fat T2 differences and results in the proton density fat fraction (PDFF), which is independent of the employed experimental settings (41).

In addition, bone marrow spectra are characterized by strongly overlapping water and fat peaks due to susceptibility effects from the trabecular bone (41, 42): water and olefinic fat peaks overlap significantly, especially in the spine. The *a priori* knowledge of the chemical structure of triglycerides can be employed to improve the robustness of water peak extraction

by constraining the area of the olefinic fat peak to the main fat peaks (41–43). However, the quantification of fat unsaturation relies on the extraction of the olefinic fat peak on an individual spectrum basis and remains a major technical challenge in the presence of a strong water peak (i.e., in the spine). Previous studies measuring vertebral bone marrow fat unsaturation based on short-TE spectra have reported moderate reproducibility values (same day, with repositioning) (36). Increasing the TE (44) or adding diffusion-weighting (45, 46) have been recently proposed to reduce the water peak height and more robustly extract the olefinic peak in vertebral bone marrow spectra.

## Water–Fat Imaging

In contrast to single-voxel MRS, chemical shift encoding-based water–fat imaging allows the spatially resolved assessment of the BMFF (**Figure 1**). High-resolution BMFF mapping is highly advantageous due to the heterogeneous distribution of bone marrow in most regions (e.g., proximal femur, spine). However, several confounding factors have to be considered when measuring bone marrow PDFF by using water–fat imaging, including the presence of multiple peaks in the fat spectrum (47, 48), T1-bias (49, 50) and T2\*-decay effects (47, 51). The presence of trabecular bone most importantly shortens the T2\* of water and fat components, inducing a rapid decay of the measured gradient echo signal with echo time (42, 52). After correcting for T2\* decay effects, a good agreement was reported *in vivo* between MRSbased and imaging-based PDFF in both the proximal femur (42) and spine (52).

An *ex vivo* study was performed in trabecular bone specimens filled with water–fat emulsions of different fat fractions, showing an excellent agreement between imaging-based PDFF and the known PDFF of the emulsions (53). To validate the results of bone marrow water–fat imaging with a non-MRI reference, Arentsen et al. (54) and MacEwan et al. (55) performed studies comparing water–fat imaging with histological examinations. Water–fat imaging in both studies resulted in vertebral BMFF and water fraction values, which showed good correlation (*r* = 0.76–0.77) with the histological results.

As water–fat imaging methodologies become more broadly available, studies have been recently performed to evaluate the significance of a spatially resolved fat fraction map (56–60). In a recent study, Baum et al. investigated whole body spine water–fat imaging on 28 healthy subjects (56). They found an increased BMFF from the cervical to the lumbar level. Absolute precision error amounted to 1.7% on the average.

Despite a considerable amount of literature using MR methods to assess bone marrow properties in the pelvis and the long bones, vertebral bone marrow remains the region most extensively investigated in bone marrow studies of aging, osteoporosis, diabetes, and obesity (**Table 1**).

## Bone Marrow Fat in Aging

The BMFF is well known to increase with age (**Figure 1**). The exact age dependence has been shown to differ between male and female subjects (61–63). A recent study has also emphasized the importance of T2-correction when using MRS to investigate the age dependence of BMFF (41). However, less is known about the age dependence of bone marrow fat unsaturation. Huovinen et al. recently investigated bone marrow fat unsaturation in young adults (15–27 years) (64). Thirty-five young adults from 27 to 35 years were included in their study, and they showed an increase of bone marrow fat unsaturation with age. The results


#### TABLE 1 | Summary of MR-based studies investigating vertebral bone marrow fat properties in osteoporosis, diabetes, and obesity.

suggest an association of bone marrow fat unsaturation with age in early adulthood and may represent the normal maturation of bone marrow.

## Bone Marrow Fat in Osteoporosis

#### Bone Marrow Fat Fraction

The MR-based vertebral bone marrow fat parameter studied most frequently in the context of osteoporosis is the fat fraction. Representative data from the authors own work, using a methodology similar to Baum et al. (65) and highlighting the effect of age and osteoporosis on bone-marrow fat fraction, are shown in **Figure 1**.

Most previous studies evaluating the relationship between BMD and BMFF have been based on single-voxel MRS. Griffith et al. performed a study in 103 female subjects to investigate the association between bone marrow fat and osteoporosis *in vivo* (20). They acquired DXA to stratify the group into healthy (*n* = 18), osteopenic (*n* = 30), and osteoporotic subjects (*n* = 55). MRS was performed at L3. The vertebral marrow fat fraction was significantly increased in osteoporotic subjects (67.8 ± 8.5%) compared with healthy subjects (59.2 ± 10.0%). Similar results were found in men. In another study, 90 men (42 subjects with normal BMD, 23 osteopenic, and 17 osteoporotic subjects) were recruited and obtained a measurement of the bone marrow fat content in L3 using MRS (19). The vertebral marrow fat fraction was significantly increased in subjects with osteoporosis (58.2 ± 7.8%) and osteopenia (55.7 ± 10.2%) compared to subjects with normal BMD (50.5 ± 8.7%).

Karampinos et al. recently examined ten vertebrae from human cadavers using MRS to assess BMFF, multi-detector computed tomography (MDCT) to determine BMD and trabecular bone microstructure parameters, and biomechanical testing to assess vertebral bone strength (21). They reported significant correlations between the MRS-based fat fraction and MDCTbased parameters (up to *r* = −0.72), and MRS-based fat fraction and vertebral failure load (*r* = −0.77). Thus, this study demonstrated that bone marrow fat volume is negatively associated with both bone microstructure and bone strength. However, further studies with larger number of specimens are needed in order to investigate whether the BMFF has an effect on bone strength after correcting for the contribution of BMD.

Kühn et al. performed one of the first bone marrow high-resolution fat fraction mapping studies in patients with osteoporosis. Their study included 51 patients who underwent DXA as well as water-fat imaging of the lumbar spine (57). According to the DXA results, the participants were divided into three subgroups: 92 healthy, 47 osteopenic, and 34 osteoporotic. The obtained PDFF was greater in osteoporotic than healthy vertebrae (62.4 ± 11.0% versus 56.3 ± 14.8%). Thus, water–fat imaging could be an alternative to the most commonly used MRS to assess bone marrow fat in the context of osteoporosis.

#### Bone Marrow Fat Unsaturation

The second MR-based bone marrow fat parameter, which has been linked to bone matrix loss, is bone marrow fat unsaturation. Yeung et al. analyzed bone marrow fat composition in the context of osteoporosis (26). They included 53 women over 60 years of age and 12 young controls in their study. The subjects underwent DXA and MRS of the lumbar spine. Interestingly, the fat unsaturation level was significantly decreased in osteoporotic (0.091 ± 0.013) and osteopenic (0.097 ± 0.014) subjects compared to healthy subjects (0.114 ± 0.016) and young controls (0.127 ± 0.031). Patsch et al. investigated the vertebral bone marrow fat composition in diabetic and non-diabetic postmenopausal women with and without fragility fractures (66). In consistency with Yeung et al., the prevalence of fragility fractures was associated with −1.7% lower unsaturation levels and +2.9% higher saturation levels.

## Bone Marrow Fat in T2DM

Baum et al. performed a study on 26 subjects, including 13 non-diabetic and 13 treated diabetic women (16). The subjects underwent blood parameter analysis including plasma glucose and HbA1c. Images from mid T12 to mid L4 were acquired using MDCT to obtain SAT (subcutaneous adipose tissue), VAT (visceral adipose tissue), and TAT (total adipose tissue). MRS was obtained at L1, L2, and L3 to assess BMFF and unsaturation. The mean vertebral BMFF was similar in the diabetic women and the healthy controls (69.3 ± 7.5% versus 67.5 ± 6.1%; *P* > 0.05). However, the mean unsaturation level was significantly lower in the diabetic group (6.7 ± 1.0% versus 7.9 ± 1.6%; *P* < 0.05). Adjusted SAT and TAT correlated significantly with mean vertebral bone marrow fat content in the whole study population (*r* = 0.538 and *r*= 0.466; *P*< 0.05). Interestingly, significant correlations of mean vertebral bone marrow fat content with adjusted VAT and HbA1c were observed only in the diabetic group (*r* = 0.642 and *r* = 0.825; *P* < 0.05).

Similar results to the study by Baum et al. (16) were reported by Patsch et al. (66): diabetes was associated with a −1.3% lower unsaturation and +3.3% higher saturation levels. Diabetics with fractures showed the lowest marrow unsaturation and highest saturation. Therefore, the authors suggested that vertebral bone marrow fat unsaturation beyond fat fraction might have a potential for assessing the BMD-independent fracture risk.

## Bone Marrow Fat in Obesity

Bone marrow fat fraction has been shown to be a negative predictor of bone microarchitecture and mechanical properties in obese men (67). BMFF has been also shown to be positively associated with ectopic and serum lipid levels in obese men and women (68) and to increase after a 6-month growth hormone administration in obese women (69).

To get further insight into the relationship between bone marrow fat and obesity, Bredella et al. performed a study in 47 pre-menopausal women (70). The authors measured IGF-1 and growth hormone blood levels, and performed MRS at L4 and MDCT scans at the level of L4 for VAT determination. The vertebral BMFF was positively associated with VAT and inversely associated with IGF-1. The authors concluded that visceral fat might have detrimental effects on bone health, which may be mediated in part by IGF-1 as an important regulator of the fat and bone lineage. Wongdee et al. concluded in a recent review that obesity and insulin resistance because of hyperglycemia in T2DM may induce osteoblast and osteoclast dysfunction with a resulting lower bone turnover (71). However, the underlying cellular and molecular mechanisms of insulin resistance in osteoblasts, osteoclasts, and osteocytes remain unknown and require further investigation.

Cordes et al. examined a group of twenty obese women who underwent a 4 week calorie restriction of 800 kcal/day (72). They performed MRS of the bone marrow and liver. Furthermore, they determined blood fat values and acquired two-point Dixon images of the abdomen to measure SAT and VAT. Despite a significant reduction of the body mass index (BMI), SAT, VAT, triglycerides, and LDL (low density lipoproteins)-cholesterol, they observed no significant reduction of the vertebral BMFF. However, absolute BMFF changes were positively associated with SAT volume (*r* = 0.489) and negatively associated with non-adipose tissue volume (*r* = −0.493) before dietary intervention (67).

Schafer et al. examined the changes of the BMFF and bone mass after gastric bypass surgery (22). They included eleven women (six diabetic, five non-diabetic) who underwent bariatric surgery (Roux-y-gastric-bypass) and underwent L3/L4 MRS, anthropometric measurements, whole body fat, and BMD measurements. VAT was determined on a single slice at the level of L4 using MDCT. In consistency with previous studies, they reported a positive correlation between age and bone marrow fat content. Interestingly, mean bone marrow fat decreased (−7.5%, *p* = 0.05) in the diabetic subjects, while the non-diabetic women showed only a small change (+0.9%, *p* = 0.84). Despite the small sample size, this study highlighted that bone marrow fat behaves differently compared to other fat depots in patients without diabetes after gastric bypass surgery.

## CONCLUSION AND PERSPECTIVES

In summary, MR-based assessment of bone marrow fat provides interesting insights into the pathophysiology of osteoporosis, T2DM, and obesity as well as the association of bone and metabolic disturbances. Currently available MR methods including MRS and water–fat imaging enable the non-invasive extraction of the BMFF and unsaturation. Very little is known about the underlying mechanisms. Furthermore, new research questions are evolving on the role of bone marrow adipocytes in bone remodeling, hematopoietic stem cell differentiation, and whole body homeostasis (73, 74), on novel non-invasive MR biomarkers specific to the distribution, composition, microstructure, and function of bone marrow adipocytes requiring further investigations.

## AUTHOR CONTRIBUTIONS

Collection of literature: CC, TB, and DK. Development of technical parts: MD, SR, MND, and DK. Development of clinical parts: CC, TB, HH, JK, and DK. Manuscript preparation/revision: CC, TB, MD, SR, MND, HH, JK, and DK.

## FUNDING

This work has received funding from Philips Healthcare (to DK), the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme [grant agreement No 677661, ProFatMRI (to DK)] and the Deutsche Forschungsgemeinschaft [DFG BA 4085/2-1 (to JK) and BA 4906/1-1 (to TB)]. This work was supported by the German Research Foundation (DFG) and the Technical University of Munich (TUM) in the framework of the Open Access Publishing Program.

## REFERENCES


with varying bone density: MR evaluation. *Radiology* (2006) 241(3):831–8. doi:10.1148/radiol.2413051858


**Conflict of Interest Statement:** DK receives Grant Support from Phillips Healthcare. The remaining 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 © 2016 Cordes, Baum, Dieckmeyer, Ruschke, Diefenbach, Hauner, Kirschke and Karampinos. 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.*

# The Bone Marrow-Derived Stromal Cells: Commitment and Regulation of Adipogenesis

*Michaela Tencerova1,2\* and Moustapha Kassem1,2,3*

*1Department of Molecular Endocrinology, Odense University Hospital, University of Southern Denmark, Odense, Denmark, 2Danish Diabetes Academy, Novo Nordisk Foundation, Odense, Denmark, 3Stem Cell Unit, Department of Anatomy, Faculty of Medicine, King Saud University, Riyadh, Saudi Arabia*

Bone marrow (BM) microenvironment represents an important compartment of bone that regulates bone homeostasis and the balance between bone formation and bone resorption depending on the physiological needs of the organism. Abnormalities of BM microenvironmental dynamics can lead to metabolic bone diseases. BM stromal cells (also known as skeletal or mesenchymal stem cells) [bone marrow stromal stem cell (BMSC)] are multipotent stem cells located within BM stroma and give rise to osteoblasts and adipocytes. However, cellular and molecular mechanisms of BMSC lineage commitment to adipocytic lineage and regulation of BM adipocyte formation are not fully understood. In this review, we will discuss recent findings pertaining to identification and characterization of adipocyte progenitor cells in BM and the regulation of differentiation into mature adipocytes. We have also emphasized the clinical relevance of these findings.

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Jan Tuckermann, University of Ulm, Germany Luca Vanella, University of Catania, Italy*

#### *\*Correspondence:*

*Michaela Tencerova mtencerova@health.sdu.dk*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 04 May 2016 Accepted: 05 September 2016 Published: 21 September 2016*

#### *Citation:*

*Tencerova M and Kassem M (2016) The Bone Marrow-Derived Stromal Cells: Commitment and Regulation of Adipogenesis. Front. Endocrinol. 7:127. doi: 10.3389/fendo.2016.00127*

Keywords: bone marrow stem cells, adipogenesis, secreted factors, bone marrow microenvironment, bone marrow stem cell subpopulations

## INTRODUCTION

Bone marrow (BM) is an important compartment of bone, which regulates bone homeostasis. BM can also be perceived as an immune organ as it contains many different cell types secreting a large number of cytokines and immune modulatory factors. Finally, BM is a metabolic organ and has recently been demonstrated to regulate a whole body energy metabolism (1, 2).

The cellular composition of BM is complex as it contains hematopoietic stem cells giving rise to myeloid lineage including osteoclasts and lymphoid lineage giving rise to immune cells. It also contains a stroma compartment containing bone marrow stromal stem cells (BMSC) (also known as skeletal or mesenchymal stem cells) and their differentiated progeny of adipocytes and osteoblasts as well as endothelial cells, pericytes and neuronal cells. The cellular composition of BM changes with age, gender, and metabolic status (3, 4).

Bone turnover/remodeling is very dynamic and energetically demanding process that consists of two main phases: bone formation mediated by osteoblasts recruited from BMSC and bone resorption mediated by osteoclasts recruited from hematopoietic progenitors. Bone resorption and bone formation are *coupled* in time and space and there is a balance between the amount of bone resorbed by osteoclasts and the amount of bone formed by osteoblasts. These processes of *coupling and balance* are tightly regulated via several factors present in BM microenvironment and also via sympathetic central nervous system (2) (**Figure 1**). Imbalance between bone resorption and bone formation leads to metabolic bone diseases, including age-related bone loss and osteoporosis.

During the recent years, there has been an increasing interest in understanding the biology of BM adipocyte for a number of reasons. First, it is an abundant cell type in adult BM (5). Second, an increased BM adipose tissue mass has been reported in the conditions of low bone mass, suggesting an abnormal differentiation of BMSC as a possible pathogenetic mechanism to be investigated. Finally, the biological role of BM adipocytes and their differences and similarities with extramedullary adipocytes are not known and may be relevant to bone tissue homeostasis.

In this review, we will present an overview of the BM adipocyte differentiation and its regulation by a number of factors. We will also outline a number of specific signaling pathways that regulate BMSC lineage commitment to adipocytes versus osteoblasts and that can be targeted to enhance bone formation and increase bone mass.

## FROM BONE MARROW STEM CELLS TO COMMITTED ADIPOCYTIC CELLS IN THE BONE MARROW

*In vitro*, BMSC are plastic-adherent cells, present within BM stroma and capable of clonal expansion and differentiation to osteoblasts and adipocytes. Currently, human BM adipocytes are thought to differentiate from BMSC as evidenced by a large body of both *in vitro* and *in vivo* studies (5). In mice, recent lineage tracing studies employing genetically modified mice, provided evidence for the common stem cell hypothesis for the presence of a common stem cells for osteoblastic and adipocytic cells (6, 7). **Table 1** summaries the main characteristics of recently reported BMSC and progenitor cells identified and characterized based on lineage tracing studies employing expression of a number of markers.

servier.com/Powerpoint-image-bank

Table 1 | List of different skeletal progenitor cells in the bone marrow identified by specific cell surface markers and mediators.


Single cell clonal analysis of cultured BM stromal cells revealed, in addition to BMSC, the presence of independent committed precursors for osteoblasts and adipocytes. Our study of Post et al. demonstrated the existence of these precursors within the murine BM stroma (16). Based on clonal selection, we were able to isolate and characterize the phenotype of two different precursor cell lines isolated from BM of 2- to 3-monthold mice: one cell line is committed to adipocyte differentiation and termed adipoMSC and another one committed to osteoblast and chondrocyte differentiation and termed boneMSC. The two cell lines exhibit distinct morphology and molecular signature. Based on the differentiation marker expression between these two cell lines, we have recently reported that CD34 is differentially expressed in boneMSC and not in adipoMSC and that it can be used to prospective isolation of osteoblastic committed BMSC (15). Examples of studies using lineage tracing to identify committed precursor cells include the study of Yue et al. (11) that employed leptin receptor (LepR) to identify precursor cells with adipocyte differentiation potential. Mice with conditionally deleted LepR in limb bones enhanced osteogenesis and improved fracture healing (11). Another example is the study of Holt et al. (10) that demonstrated the presence of RANKL + preadipocytes in aged mice BM that support osteoclastogenesis. Similar findings have been confirmed in human BM (10, 17). Interestingly, some lineage tracing studies show heterogeneity of BMSC populations with respect to their differentiation potential. Gremlin 1 (Grem1) is a secreted BMP inhibitor and involved in the regulation of adipogenesis in peripheral adipose acting via PPARγ (18). Recent study of Worthley et al. identified Gremlin 1 positive precursor cells that can differentiate into osteoblasts, chondrocytes, and reticular cells, but not to adipocytes (9). All these studies demonstrate the presence of complex cellular heterogeneity within the BM microenvironment and the presence of different committed BMSC subpopulations with a different differentiation potential, whose functions in the BM can be modulated independently during aging and metabolic diseases. It is also possible that these precursor cells number and functions are modulated by physiological conditions and energy demand. Since most of the above-mentioned studies were performed in Table 2 | Cellular and molecular characteristics of bone marrow and extramedullary adipocytes.


mice, it is important to define the relevance of these findings to normal human physiology.

## BONE MARROW ADIPOCYTES VERSUS EXTRAMEDULLARY ADIPOCYTES

Biological differences between BM medullary adipocytes and extramedullary adipocytes are not completely delineated. Several investigators have reported similarities between stromal cells isolated from the BM and subcutaneous stromal cells with respect to molecular phenotype and adipocyte differentiation capacity (19–22). However, significant differences have been observed in their responsiveness to adipocyte differentiation signals, as subcutaneous adipose tissue-derived stem cells (ASC) are better to differentiate to adipocytes compared to BMSC. BM and extramedullary adipocytes are different in cell size, expression of stem cell markers (e.g., Sox2, Nanog, Klf4), presence of extracellular matrix, free fatty acid content, proportion of immune cells in contact with adipocytes, levels of cytokine, and adipokine expression (23–28). Also BMSC have higher expression of inflammatory genes compared to ASC (29). Interestingly, the molecular signature analysis reveals that stromal cells from different tissue compartment are imprinted by their tissue of origin and ASC are enriched in adipogenesis-associated genes compared to MSC from other tissue compartments (19, 30–35). Other differences on cellular and molecular level have been reported (see listed in **Table 2**).

A recent paper of Liaw et al. has reported significant differences in lipid composition among different adipocytic cell lines derived from a variety of sources, including white and brown adipose tissue in mice and in adipocytes differentiated from 3T3-L1 cell line and ear mesenchymal cells (36). Unfortunately, the authors did not include BM adipocytes in their studies. This data highlight the differences in lipid metabolism in different adipocyte compartments, which may be mediated by exposure to different bioactive molecules in their microenvironment.

## BONE MARROW STROMAL STEM CELL COMMITMENT TO ADIPOCYTIC LINEAGE AND REGULATORY FACTORS

Bone marrow stromal stem cell commitment to adipocytic lineage is a complex process, which is tightly controlled via several positive and negative regulatory factors activated possibly in a


Table 3 | List of selected regulatory factors for adipocyte differentiation in bone marrow and adipose-derived stem cells.

sequential cascade. These factors include steroid hormones, secreted cytokines/adipokines, and transcription factors. Most of these signaling molecules are known for their regulation of adipogenesis in bona fide ASC. **Table 3** summarizes the effects of selected regulatory factors on adipocyte differentiation in BMSC and ASC that are discussed in the following paragraphs.

The most characterized transcription factor and a key regulator of adipogenesis is peroxisome proliferated-activated receptor gamma (PPARγ) (37, 66). PPARγ belongs to a nuclear receptor superfamily, which is activated by lipophilic ligands. The activation of PPARγ is necessary and sufficient for adipocyte differentiation and also required for maintenance of differentiated state in BMSC and ASC (67–69). Inhibition of PPARγ *in vitro* impairs adipogenesis, while enhancing osteoblast differentiation in BMSC (67). In mice PPARγ deficiency leads to impaired development of adipose tissue when fed a high-fat diet (HFD) (70). PPARγ is also a target for insulin sensitizing drugs, such as thiazolidinediones in diabetes. However, their use for diabetic patients is associated with a decreased bone mass and increases a risk for fracture. The role of PPARγ activation in age-related increase of BM adipogenesis and decreased osteoblastogenesis has been discussed previously [for more information, see the reviews: Ref. (3, 38, 68, 71)].

Additional transcription factors involved in the regulation of adipogenesis are members of CAAT enhancer binding proteins (C/EBP) family: C/EBPα, C/EBPβ, C/EBPγ and C/EBPδ. Based on the studies performed in 3T3 cell line, C/EBP activation during adipocyte differentiation is synchronized in a temporal manner where early activation of C/EBPβ and C/EBPδ leads to induction of C/EBPα. In BMSC, the function and activation of individual transcription factors exhibited a different pattern (72). Moreover, it has been shown that an isoform of C/EBPβ, liver-enriched inhibitory protein (LIP), which lacks transcriptional binding domain, induces activation of Runx2 and promotes osteogenesis in BMSC (39). C/EBPs crosstalk with PPARγ and regulate each other via a feedback loop (38, 68). C/EBP deficient mice exhibited impaired adipogenesis and insulin sensitivity (73–75). Moreover, C/EBPβ-deficient mice displayed reduced bone mineral density with decreased trabecular number (76, 77). These findings confirm an important role of C/EBPs in the early stage of MSC differentiation and their commitment (78).

The PPARγ-regulated adipokines: leptin and adiponectin are primarily secreted by adipocytes and can regulate adipogenesis (79, 80). *In vitro* leptin inhibits adipogenesis and enhances osteoblastogenesis in human stromal marrow cells (43). On the other hand, leptin-deficient mice *ob/ob* and LepR-deficient *db/ db* mice exhibit an increased BM adiposity and low bone mass (79). Leptin regulates bone mass negatively indirectly via sympathetic nervous system (44). Interestingly, selective inhibition of LepR in osteoblastic cells has no effects on bone mass, whereas hypothalamic deletion of LepR leads to a phenotype similar to that of *ob/ob* mice (81), suggesting that the main effects of leptin on bone are centrally mediated. In extramedullary adipocytes leptin impairs adipocyte function (e.g., insulin responsiveness and lipid metabolism) and inhibits lipogenesis (45). In addition, Aprath-Husmann et al. reported no effect of leptin on adipocyte differentiation in ASC of lean and obese subjects (82). Also, leptin levels increase with obesity and diabetes, diseases associated with bone fragility. Thus, leptin seems to exert multiple functions with direct and indirect effects on BM adipocytes and extramedullary adipocytes.

Adiponectin is an adipocyte-secreted factor with insulin sensitizing and anti-inflammatory effects. Adiponectin blocks adipocyte differentiation of BMSC and ASC suggesting an autocrine or paracrine negative feedback loop (40). *In vitro* adiponectin enhances osteoblast differentiation, increases osteoblast proliferation and maturation via cyclooxygenase 2 (Cox2)-dependent mechanism, and inhibits osteoclastogenesis (41, 83). However, *in vivo* effects on bone mass are more complex. Adiponectin regulates bone formation via opposite central and peripheral mechanisms through FoxO1 transcriptional factor. Adiponectin-deficient mice exhibit increased bone mass in young age but low bone mass during aging. This effect is explained by local inhibition of osteoblast proliferation and enhanced osteoblast apoptosis. During aging, this effect is antagonized by adiponectin-mediated effects on hypothalamic neurons that lead to decreased sympathetic tone and, consequently, increased bone mass and decreased energy expenditure (80). Overexpression of adiponectin in adipose tissue causes impairment of adipogenesis and increased preadipocyte factor 1 (Pref-1) expression, which inhibits adipogenesis in mice (42).

In obesity and type 2 diabetes, circulating levels of leptin and adiponectin are differentially regulated (up- and downregulated, respectively). However, their role in the regulation of BM adipogenesis has not been determined. Recently, Yue et al. demonstrated LepR signaling in BMSC promotes adipogenesis and inhibits osteoblastogenesis in response to diet (11). By contrast, activation of adiponectin receptor R1 (AdipR1) in osteoblasts results in enhanced bone formation via GSK-3β/β-Catenin signaling (84). AdipoR1 deficient mice exhibit impaired osteoblast differentiation. The study of Yu et al. highlighted the importance of adiponectin signaling in BMSC mobilization and recruitment during bone fracture repair via increased secretion of stromal cell-derived factor 1 (SDF-1) in mice (85). Cawthorn et al. recently reported increased secretion of adiponectin from BM adipocytes in caloric restriction state that can contribute to its circulating levels (86), suggesting an important endocrine role of BM adipocytes in the regulation of whole body energy metabolism. However, this observation needs further investigation.

Novel factors, which have been identified in our laboratory based on proteomic analysis of secreted factors by committed BM adipocytic cells (adipoMSC) and committed BM osteoblastic cells (boneMSC), are secreted frizzled-related protein 1 (sFRP1) and Delta-like 1, also known as preadipocyte factor 1 (Dlk1/Pref-1) (16).

Secreted frizzled-related protein 1 is an inhibitor of Wnt signaling that sequester Wnts from their receptors. *In vitro* it inhibits osteoblastogenesis and promotes adipogensis of BMSC by blocking the Wnt signaling (46) *In vivo* sFRP1 inhibited bone formation. Similar effects of sFRP1 have been reported in preadipocytes and primary adipose tissue-derived cells (87, 88).

 Delta-like 1/preadipocyte factor 1 is a transmembrane protein, which belongs to a family of epidermal-growth-factor-like repeats containing proteins. Its extracellular domain is proteolytically cleaved by ADAM17/TACE and released as soluble factor circulating in body fluids include amniotic fluid and, hence, its name fetal antigen A (FA1) (89–91). Pref-1 is highly expressed in preadipocytes and its expression decreases during differentiation. Pref-1 overexpression in 3T3-L1 cells blocks adipogenesis (92). Pref-1 regulates adipocyte differentiation via FOXA2 (93), KLF6 (94), and KLF2 (95). Our group has reported that overexpression of Dlk1 in human BMSC inhibits adipocyte and osteoblastic differentiation (48). Interestingly, Dlk1/Pref-1 inhibited differentiation of MSC downstream of C/EBPβ during adipocytic differentiation and Cbfa1/Runx2 during osteoblastic differentiation, suggesting that Dlk1/Pref-1 maintains MSC in a progenitor state. Importantly, we showed a negative effect of soluble FA1 on bone formation in *ex vivo* neonatal calvaria organ cultures. Also transgenic mice with Dlk1 overexpression had reduced fat and bone mass (96). We have also recently reported that FA1 acts as a link between bone and whole body energy metabolism and it interacts with osteocalcin (97).

Additional negative regulators of adipocyte differentiation of BMSC include molecules in Wnt signaling pathway that consists of several ligands, receptors, co-receptors and transcriptional mediators, e.g., β-catenin, which blocks PPARγ and its downstream-regulated genes (98, 99).

Low-density lipoprotein receptor-related protein 5 (LRP5) is a Wnt co-receptor and is involved in activation of canonical Wnt signaling (49, 50, 100). It has been shown to inhibit adipogenesis and promote osteoblastogenesis in BMSC. A gain of function mutation in Lrp5, a clinical condition known as a high-bone-mass phenotype, leads to inhibition of adipogenesis and enhances osteoblastogenesis, which is associated with increased bone mass. On the other hand, Lrp5 loss of function mutation causes severe osteoporosis in mice and humans (49, 50, 100, 101). Recent paper of Loh et al. (49) demonstrated enhanced adipogenesis in lower body fat, e.g., gluteal adipose tissue in high-bone-mass phenotype patients. This finding confirms the results of Palsgaard et al., who reported impaired adipogenesis in LRP5-deficient preadipocytes due to an interaction between LRP5 and insulin receptor (51).

Bone morphogenetic proteins (BMPs) are members of the transforming growth factor β (TGFβ) superfamily that were originally identified based on their ability to induce ectopic bone formation. BMPs have pleiotropic developmental actions, important for stem cell self-renewal and lineage commitment and differentiation (102). BMP4 promotes adipogenesis in peripheral adipose tissue progenitors by increasing transcription activity of PPARγ (18, 103). However, dependent on BMP concentration and receptor activation, they exert different lineage differentiation effects (104). Low concentrations of BMP-2 and BMP-7 induce adipocytic differentiation, whereas high concentrations promote differentiation toward chondrocytes and osteoblasts (105, 106). BMP signaling through type IB BMP receptor (BMPR-IB) plays a crucial role in mediating osteoblast differentiation of BMSC by a Dlx5/Runx2-mediated pathway, while activation of the type IA BMP receptor (BMPR-IA) in BMSC induces PPARs expression and promotes adipocyte differentiation (107, 108).

Another signaling pathway involved in the regulation of adipogenesis is insulin/insulin-like growth factor (IGF-1) pathway (55, 56, 109). IGF-1 has pleiotropic functions in several tissues with regulatory effects on cell proliferation and cell differentiation. It is the most abundant growth factor in the bone matrix. Deletion of IGF-1 in osteocytes caused impaired developmental bone growth in mice (110). IGF-1-osteoblast deficient animals exhibit impaired bone formation and reduction in bone mass (111). Mice treated with recombinant IGF-1 exhibited enhanced bone formation and osteogenesis via activation of mTOR signaling (112). In adipose tissue, IGF-1 plays an important role in adipocyte differentiation, especially in lineage commitment stage where insulin acts predominantly through IGF-1 receptors, which are highly expressed in preadipocytes compared to insulin receptors (113). Blocking downstream molecules in insulin signaling pathway inhibits adipogenesis in preadipocytes (56). Importantly, IGF-1 is also involved in the regulation of energy metabolism and glucose uptake in insulin responsive cells. Therefore, IGF-1 exerts multiple functions dependent on cellular energy needs. However, its role in the regulation of energy metabolism in BMSC and BM adipocytes is not known.

Muscle-secreted proteins, known as myokines, represent a group of regulatory molecules with expanding role in crosstalk between muscle and different organs in the body and with regulatory functions on bone and energy metabolism.

Irisin is a newly identified myokine released from skeletal muscle during exercise through peroxisome proliferator-activated receptor gamma coactivator 1 (PGC1) activation that mediates expression of membrane protein Fibronectin type III domaincontaining protein 5 (FNDC5). This is a precursor molecule for irisin, which is subsequently cleaved as the myokine (114). Irisin has a browning effect on white adipose tissue via upregulating uncoupling protein 1 (UCP1) through p38 MAPK and ERK (57, 114). Moreover, recent studies demonstrated that irisin positively affects skeletal system, i.e., it enhances osteoblastogenesis *in vitro* and *in vivo* (58, 59).

Fibroblast growth factor 21 (FGF21) has been identified as a circulating hepatokine with effects on glucose and lipid metabolism. FGF21 is also secreted from adipose tissue and skeletal muscle (115). It exerts a positive effect on adipogenesis via activation of PPARγ in BMSC and adipose tissue progenitors. FGF21-deficient mice exhibit high bone mass and decreased fat formation. Reciprocally, mice overexpressing FGF21 exhibit reduced bone mass (60, 61). In older men, high serum levels of FGF21 are associated with low bone mass (116).

Inflammatory cytokines are mostly produced by immune cells and play an important role in the regulation of bone remodeling and adipocyte formation.

Transforming growth factor beta is a cytokine of the TGF superfamily that promotes preadipocyte proliferation and inhibits adipocyte differentiation. Overexpression of TGFβ in mice leads to impaired adipose tissue development. On the other hand, TGFβ-deficient mice display impaired bone growth and mineralization (117). TGFβ mediates its inhibitory function on adipogenesis via SMAD3, which acts on C/EBPα (56, 118).

The effects of the pro-inflammatory cytokines: interleukin 1 (IL1), interleukin 6 (IL6), and tumor necrosis factor alpha (TNFα) on preadipocytes and BMSC are similar (62). They inhibit adipogenesis by reducing PPARγ and C/EBPα expression and by blocking insulin action via decreasing Glut4 expression in preadipocytes (119). TNFα and IL1 suppress adipocyte differentiation by activation of the TAK1/TAB1/NIK cascade, which in turn inhibits PPARγ activity (120). IL1 and TNFα inhibit adipocyte cell proliferation by activation of several distinct intracellular signaling pathways (e.g., JNK, p38 MAPK) (119, 121, 122). Moreover, IL6 maintains BMSC in undifferentiated state through ERK1/2-mediated mechanism during bone fracture healing (123). On the other hand IL6 enhances osteoblast differentiation of BMSC by decreasing Sox2 expression (124). In estrogen-deficient mouse model, IL6-deficient mice are protected from ovariectomy-induced bone loss (125), suggesting a role in mediating estrogen-deficiency-related bone loss (125).

Heme-oxigenase 1 (HO-1) is a rate-limiting enzyme with antiinflammatory properties, activated by oxidative stress, which was reported to regulate commitment of human BMSC differentiation to osteoblastic cells. HO-1 acts as an inhibitor of adipogenesis by enhancing Wnt signaling (64). Similar effects were observed in ASC of obese mice, which were treated with HO-1 inducer that led to decreased adiposity in peripheral adipose and BM along with a positive effect on insulin sensitivity (65).

Taken together, the above-mentioned regulatory factors share similar signaling pathways for the regulation of adipocyte differentiation in BMSC and ASC. However, some of these factors, e.g., BMP, IGF-1, and LRP5 display different effects depending on the origin of MSCs, suggesting an important role of local microenvironment. These findings are relevant to the design of potential drugs for targeting BMSC in the context of regulation of bone and energy metabolism. **Figure 1** summarizes factors regulating BMSC differentiation and their associated signaling pathways.

## MICRO RNA AND REGULATING GENETIC NETWORKS IN ADIPOCYTIC DIFFERENTIATION

Micro RNAs (miRNA) are evolutionary conserved short noncoding RNA molecules (containing about 22 nucleotides) that function in RNA silencing and post-transcriptional regulation of gene expression. Accumulating evidence suggests that miRNA regulate fate decisions of stem cells, including self-renewal and differentiation (126, 127).

Several groups have employed global miRNA gene expression profiling during differentiation of human BMSC to identify several miRNAs that regulate BMSC fate and that act as a molecular switch to control adipocyte and osteoblast differentiation fate. Most of miRNAs regulate gene expression of key molecules involved in BMSC differentiation, such as PPARγ, C/EBP, Runx2, Wnt/β-catenin, Lrp5/6, and so on. For example, recent study of Hamam et al. identified miR320 family, whose upregulation in human BMSC enhanced adipocyte differentiation. The biologically relevant gene targets for miR-320c are RUNX2, MIB1 (mindbomb E3 ubiquitin protein ligase 1), PAX6 (paired box 6), YWHAH, and ZWILCH (128). Other miRNAs that have been reported as regulators of adipogenesis include miR-143, -24, -31, -30c, and -642a-3p. More detailed description of their function in the regulation of BMSC differentiation is summarized in these recently published reviews (127, 129, 130). Thus, targeting different miRNAs represents a potential tool for a molecular therapy to regulate BMSC differentiation fate.

## BONE MARROW ADIPOGENESIS IN AGING, OSTEOPOROSIS, AND METABOLIC DISORDERS

*In vivo*, it has been shown an inverse relationship between bone and fat formation in the BM cavity (1, 131, 132). For examples, observed abnormalities of bone remodeling during aging, osteoporosis, estrogen deficiency, chronic glucocorticoid (GC) treatment, immobilization, anorexia nervosa, and Cushing disease are associated with increased adipose tissue accumulation in the BM and decreased bone mass (3, 133–135). One of the cellular explanations that has been put forward to explain this inverse relationship between bone and fat tissue mass in the BM is differentiation reprograming of BMSC toward adipocyte instead of osteoblastic fate (136–138). Moerman et al. reported that aging activates adipogenic and suppresses osteogenic differentiation programs in BMSC in mice (139). Molecular mechanism behind this reprograming machinery is not completely delineated. It has been shown that several extracellular signaling proteins have overlapping roles in BMSC adipogenesis and osteoblastogenesis by modulating the expression and/or activity of adipocyte-specific (e.g., PPARs) or osteoblast-specific (e.g., Runx2 and osterix) transcription factors. Some of these factors play opposing roles in lineage determination, while others function in complementary fashion. Schilling et al. using whole genome analyses identified several genes that could play a role in osteoblast versus adipocyte differentiation of BMSC (138).

Obesity and diabetes are highly prevalent diseases, in which bone mass is also affected (140). Several studies have demonstrated that metabolic complications of diabetes are associated with increased risk for bone fractures. However, it is not clear whether these effects are mediated by changes in BM adipose tissue. Bredella et al. found positive correlation between visceral adipose tissue and BM adiposity as measured in vertebrae of obese premenopausal women. Interestingly, this finding correlates with decreased BMD, even after correcting for the degree of obesity (141). An *in vitro* study reported that incubating BMSC with sera obtained from overweight persons promotes *in vitro* adipocyte differentiation and diminishes osteoblast differentiation (142), suggesting that secreted factors/nutrients present in circulation can affect the differentiation process of BMSC (142). In HFD-induced obesity in mice, an increased osteoclastic bone resorption associated with a lower trabecular bone mass is observed (143). Indeed, saturated fatty acids impair osteoblastogenesis, enhance adipogenesis, and affect cell survival and proliferation of human BMSC (144, 145). Other animal study reported impairment of mitochondrial function and apoptosis of BMSC in obese mice (146).

An increased BM adiposity has been reported in type 1 and type 2 diabetes (T1D and T2D). However, in T1D, there is a decrease in bone mass, whereas T2D is characterized by no change or higher bone mass and paradoxically increased risk for osteoporotic fractures (147–150). Studies are underway to examine the role of impaired glucose metabolism and its associated hyperglycemia and hyperinsulinemia on the biological functions of BMSC. This area of research has also been strengthened by the discovery of osteocalcin as a bone secreted hormone that regulates insulin secretion, proliferation of β-cells, and overall energy metabolism in mice (151, 152) and FA1 as a negative regulator of osteocalcin-induced hypoglycemia (97).

## TARGETING BONE MARROW ADIPOCYTES TO INCREASE BONE MASS

Several studies have examined the possibility of reverting the adipocyte differentiation fate of BMSC to bone forming osteoblasts as an approach to increased bone mass during aging and in osteoporosis. A number of molecular studies have investigated therapeutic potential of several factors as regulators of BMSC differentiation fate, which included hormone replacement therapy/

small molecules with antagonistic/agonistic effect or neutralizing antibodies.

An example is sclerostin (SOST), which is a glycoprotein produced by osteocytes and acts as an inhibitor of Wnt signaling. SOST inhibits bone formation and increases bone adiposity through possibly targeting differentiation of BMSC (153, 154). Treatment with humanized antibodies against SOST is currently in phase III trials for osteoporosis management (155). Another molecule with a similar function as SOST is Dickkopf-1 (DKK1), which negatively regulates Wnt signaling (156). Growth factors, such as BMPs or activin A, represent other anabolic agents for a potential treatment (157). Recent study of Florio et al. reported promising results on the use of bispecific antibody targeting SOST and DKK1 with an enhanced effect on bone formation in rodents and non-human primates (158). However, further clinical studies are needed to investigate the effectiveness of combined treatment, especially in patients with severe osteoporosis.

Some anti-diabetic drugs designed to improve insulin sensitivity and adipogenesis in the peripheral tissues have unfortunately the side effects on bone mass with increased fracture risk, e.g., thiazolidinediones due to partly enhanced BM adiposity (71, 159, 160). Thus, one of the research goals is to design antidiabetic drugs with minimal negative effects on bone. Recently used anti-diabetic drugs in the clinic include the incretin-based therapies (GLP-1 receptor agonists, DPP-4 inhibitors) and drugs targeting sodium-glucose co-transporter 2 (SGLT2)-inhibitors. However, their effects on bone mass and fracture risk need to be determined (161).

## CONCLUSION/FUTURE PERSPECTIVES

MSC commitment to differentiate into osteoblasts or adipocytes and, consequently, the balance between bone mass and BM adipose tissue mass is a complex and dynamic process, which is regulated and fine tuned by a large number of bioactive molecules. The sequential cascade of these processes and how they modulate BMSC differentiation is currently under intensive investigation. Future studies need to identify the biological functions of BM adipocytes not only in relation to bone remodeling but also as part of the overall regulation of energy metabolism. The findings of novel secreted molecules involved in the regulation of cellular fate of BMSC provide new possible anabolic therapies for treating clinical conditions of low bone mass and possibly disturbances in energy metabolism.

## AUTHOR CONTRIBUTIONS

MT and MK researched data, wrote manuscript and reviewed the final manuscript.

## ACKNOWLEDGMENTS

This work was supported by The Novo Nordisk Foundation NNF15OC0016284, Interregional EU grant Bonebank 16-1.0-15, Inovationsfonden, Denmark and OUH Research grant 15-A845. MT is supported by a Postdoctoral fellowship from the Novo Nordisk Foundation through the Danish Diabetes Academy.

## REFERENCES


of bone marrow-derived and adipose tissue-derived stromal cells. *Arthritis Rheum* (2003) 48:418–29. doi:10.1002/art.10767


via adiponectin and prostaglandins. *J Clin Invest* (2002) 109:1303–10. doi:10.1172/JCI0214506


proliferator-activated receptor gamma. *Proc Natl Acad Sci U S A* (2012) 109:3143–8. doi:10.1073/pnas.1200797109


BMP signaling pathways. *Aging Cell* (2004) 3:379–89. doi:10.1111/j.1474- 9728.2004.00127.x


**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 © 2016 Tencerova and Kassem. 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.*

# Fatty Infiltration of Skeletal Muscle: Mechanisms and Comparisons with Bone Marrow Adiposity

#### *Mark W. Hamrick1 \*, Meghan E. McGee-Lawrence1 and Danielle M. Frechette2*

*1Department of Cellular Biology and Anatomy, Medical College of Georgia, Augusta, GA, USA, 2Department of Biomedical Engineering, Stony Brook University, Stony Brook, NY, USA*

Skeletal muscle and bone share common embryological origins from mesodermal cell populations and also display common growth trajectories early in life. Moreover, muscle and bone are both mechanoresponsive tissues, and the mass and strength of both tissues decline with age. The decline in muscle and bone strength that occurs with aging is accompanied in both cases by an accumulation of adipose tissue. In bone, adipocyte (AC) accumulation occurs in the marrow cavities of long bones and is known to increase with estrogen deficiency, mechanical unloading, and exposure to glucocorticoids. The factors leading to accumulation of intra- and intermuscular fat (myosteatosis) are less well understood, but recent evidence indicates that increases in intramuscular fat are associated with disuse, altered leptin signaling, sex steroid deficiency, and glucocorticoid treatment, factors that are also implicated in bone marrow adipogenesis. Importantly, accumulation of ACs in skeletal muscle and accumulation of intramyocellular lipid are linked to loss of muscle strength, reduced insulin sensitivity, and increased mortality among the elderly. Resistance exercise and whole body vibration can prevent fatty infiltration in skeletal muscle and also improve muscle strength. Therapeutic strategies to prevent myosteatosis may improve muscle function and reduce fall risk in the elderly, potentially impacting the incidence of bone fracture.

Keywords: bone marrow adipogenesis, myosteatosis, intramyocellular lipid, exercise

## INTRODUCTION

Osteoporosis affects ~10 million people in the U.S. and results in over 1.5 million bone fractures per year. Hip fractures are a major cause of morbidity and mortality among the elderly: ~40% of those suffering a hip fracture will end up in a nursing home and 20% will never walk again. In addition, the 1-year mortality of hip fractures at age 70 is ~30%. Muscle weakness and postural instability are major contributors to the incidence of falls among the elderly, and falling is the primary etiological factor in more than 75% of hip fractures (1). Loss of muscle and bone mass with age is therefore a significant public health problem, as the morbidity that accompanies fractures in the elderly is costly both in terms of financial burden and quality of life. The mechanisms underlying loss of muscle and bone strength with age are complex and multifactorial in nature, but evidence suggests that common factors regulate the integrated growth, development, and degeneration of these two tissues. For example, skeletal muscle and bone share common embryological origins from mesodermal cell populations and also display common growth trajectories early in life. Moreover, muscle and bone

#### *Edited by:*

*Erica L. Scheller, Washington University School of Medicine, USA*

#### *Reviewed by:*

*Gretchen Meyer, Washington University, USA Amanda L. Lorbergs, Harvard Medical School, USA*

#### *\*Correspondence:*

*Mark W. Hamrick mhamrick@augusta.edu*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 02 May 2016 Accepted: 07 June 2016 Published: 20 June 2016*

#### *Citation:*

*Hamrick MW, McGee-Lawrence ME and Frechette DM (2016) Fatty Infiltration of Skeletal Muscle: Mechanisms and Comparisons with Bone Marrow Adiposity. Front. Endocrinol. 7:69. doi: 10.3389/fendo.2016.00069*

are both mechanoresponsive tissues, and the mass and strength of both tissues decline with age. Importantly, the decline in muscle and bone strength that occurs with aging is accompanied in both cases by an accumulation of adipose tissue. This accumulation of fat in non-adipose depots, such as bone, liver, and muscle, is now recognized as a common feature of aging (2). The processes driving the accumulation of bone marrow adipocytes (ACs) are becoming more well understood (3, 4); however, the factors leading to the accumulation of fat in skeletal muscle (myosteatosis) with age are not yet as well defined. Evidence, to date, does suggest that many of the factors that have been observed to stimulate bone marrow adipogenesis, such as estrogen deficiency, glucocorticoid treatment, and disuse atrophy, also induce myosteatosis. In this study, we review these findings to highlight potential therapeutic strategies for the prevention of age-related myosteatosis as an approach for reducing fall risk and hence the likelihood of bone fracture.

## FACTORS CONTRIBUTING TO BONE MARROW ADIPOGENESIS

Bone cell populations are heterogeneous and include cells of both hematopoietic (e.g., megakaryocytes and osteoclasts) and mesenchymal (e.g., osteoblasts and AC) origin. Aging is accompanied by an accumulation of AC as well as increase in ACs size within the bone marrow cavity (5). Adipose tissue represents ~20% of bone marrow tissue before the third decade in life but increases to nearly 50% by the ninth decade (6). This accumulation of bone marrow fat shows a strong association with bone loss, reduced bone formation, and fracture risk (6–9). Mesenchymal progenitors (MSCs) within bone marrow can be directed toward the AC or osteoblast lineage, and conditions that favor adipogenesis such as estrogen depletion (10), disuse (11), anorexia/calorie restriction (12, 13), and exposure to microgravity (14) are also associated with reduced osteoblast differentiation.

In addition, there are a number of pharmaceutical treatments that can mediate bone marrow adipogenesis. For example, glucocorticoids and PPAR gamma agonists will stimulate adipogenesis in mesenchymal progenitors (15, 16), whereas lipid-lowering statins can inhibit adipogenic differentiation (17). Importantly, the microenvironment of the MSCs plays a key role in modulating this reciprocal switch between adipogenic or osteogenic differentiation, particularly with aging, as young MSCs transplanted into old animals or young MSCs exposed to serum of old donors will tend to differentiate down the adipogenic pathway rather than become osteogenic (18, 19). Finally, epigenetic programing also appears to play an important role in modulating bone marrow adipogenesis. For example, conditional deletion of Hdac3 in preosteoblasts increases marrow AC number and lipid storage in preosteoblasts (20). It is worth noting that marrow ACs are themselves not homogenous in their gene expression and secretory profile. For example, some marrow ACs are similar to "white" fat in being rich in saturated fatty acids, whereas other marrow ACs are more "beige-like" fat in having greater thermogenic potential (4).

## FATTY INFILTRATION IN SKELETAL MUSCLE: CELLULAR AND MOLECULAR MECHANISMS

Aging in humans is accompanied by a loss of subcutaneous fat but an accumulation of AC and lipids in non-adipose depots, such as bone marrow, liver, and skeletal muscle (2). Fatty infiltration of skeletal muscle (myosteatosis) has, in particular, been recognized as an important component of aging and frailty (21–26). Lipid accumulation in muscles of the lower limb is also associated with increased fracture risk in the elderly (27). The cellular origins of fatty accumulation in muscle arise through several different pathways (**Figure 1**). One direct route is *via* the accumulation of lipid within myofibers themselves, known as intramuscular fat or intramyocellular (IMC) lipid (28–30). Accumulation of IMC lipid is now known to be associated with insulin insensitivity, inflammation, and functional deficits in skeletal muscle. Accumulation of the sphingolipid ceramide appears to have a particularly

detrimental effect on skeletal muscle function (30). Recent data also suggest that the lipid metabolites diacylglycerols (DAG) are responsible for mediating insulin resistance in skeletal muscle through disrupting the insulin signaling pathway (31).

Another pathway for myosteatosis is an accumulation of AC within skeletal muscle, known as intermuscular fat. There are several stem cell populations in skeletal muscle, the most well defined being muscle satellite cells (SCs), which lie below the basil lamina of muscle fibers and contribute to myogenesis during the process of muscle regeneration. A second, more recently described, population of cells is termed fibro/adipogenic progenitors (FAPs) or mesenchymal interstitial cells [**Figure 1**; Ref. (32–35)]. These cells are distinct from SCs and lack Pax7 expression but are Sca-1 and PDGFRα positive. SCs are generally resistant to adipogenic differentiation, whereas FAPs readily differentiate into ACs under various conditions such as muscle injury or glucocorticoid treatment (34, 36). Endogenous glucocorticoid levels increase with age (37), which may contribute not only to accumulation of bone marrow ACs but also to the deposition of intermuscular fat with age. Multipotent mesenchymal stem cells and other progenitors may also contribute toward skeletal muscle adipogenesis. For example, PW1<sup>+</sup> interstitial cells (PICs) have shown adipogenic potential *in vitro* (38); however, the extent to which this population overlaps with FAPs is unclear. Additionally, type-1 pericytes expressing PDGFRα have been shown to commit to the adipogenic lineage *in vivo* in the presence of glycerol (39).

Just as glucocorticoids can stimulate adipogenesis in both bone and muscle, other signaling pathways appear to be shared that regulate adipogenesis in muscle and bone (**Figure 2**). Wnt10b is well recognized to inhibit adipogenesis and stimulate bone formation in bone tissue (40). Wnt10b also suppresses the accumulation of IMC lipid in myofibers, increases insulin sensitivity, and inhibits adipogenic differentiation of aged, muscle-derived stem cells (41, 42). Similarly, inhibition of histone deacetylases (HDAC) can inhibit the adipogenic differentiation of MSCs *in vitro* and enhance their differentiation to osteoblasts (43), and HDAC inhibitors also inhibit the adipogenic differentiation of FAPs during the process of muscle regeneration (44). Altered leptin signaling, either due to absence of leptin or leptin receptors, is associated with increased bone marrow fat (45) as well as increased intra- and intermuscular fat (**Figure 3**). The leptin receptor is a key marker of bone marrow mesenchymal stem cells that mediate marrow adipogenesis (46), and the leptin receptor is also expressed in skeletal muscle (47). Whether or not the accumulation of inter- and intramuscular fat is directly mediated by the leptin receptor is, however, not well understood. Leptin deficiency associated with calorie restriction results in increased marrow adiposity (12), as does anorexia nervosa (48), but calorie restriction decreases lipid stores and lipid droplet size in skeletal muscle (49).

Unloading through either prolonged bedrest or spaceflight increases bone marrow adipogenesis (11, 14), and prolonged bedrest also decreases muscle strength and increases IMC lipid in skeletal muscle (50), which can ultimately lead to postural instability (51). Finally, estrogen deficiency is implicated in bone loss and marrow AC accumulation in women. Estrogen deficiency increases lipid content in skeletal muscle, the expression of adipogenic genes, and decreases relative satellite cell proportions in ovariectomized rodents (52, 53). Androgen deprivation therapy also increases fatty infiltration of skeletal muscle in men with prostate cancer, although CT imaging does not enable a distinction between IMC or intermuscular lipid accumulation and so

reverse fatty deposition in muscle and bone.

the actual site of lipid deposition is not clear in this case (54). Together, these findings indicate that many of the conditions that induce marrow adipogenesis and bone loss in men and women such as disuse, sex steroid deficiency, altered leptin signaling, and glucocorticoid treatment also stimulate the accumulation of ACs and IMC lipid in skeletal muscle (**Figure 2**).

## FUNCTIONAL CONSEQUENCES OF FATTY INFILTRATION IN MUSCLE

Protein synthesis enhances muscle hypertrophy and the maintenance of muscle strength, whereas impaired protein synthesis contributes to muscle atrophy. Insulin is an anabolic factor for skeletal muscle, and accumulation of muscle ACs and IMC lipid decreases insulin sensitivity, impairing the capacity for normal protein synthesis in skeletal muscle (30). Thus, decreased insulin sensitivity with fatty infiltration in skeletal muscle is one pathway by which fatty infiltration can directly affect muscle mass and muscle strength. The accumulation of IMC lipid with aging or with disuse is not homogenous across different muscles or different fiber types. This may be analogous to the unequal distribution of ACs throughout bone marrow in the appendicular skeleton, where fatty infiltration begins at more distal skeletal locations (55). For example, within the posterior compartment of the leg, the gastrocnemius accumulates more lipid with age than other calf muscles (21). Type I fibers, also referred to as "slow-twitch oxidative fibers," tend to accumulate more IMC lipid with age in human subjects than fast-twitch oxidative fibers (23, 25), and fast-twitch fibers typically show greater atrophy with age than type I fibers (23, 56). It is possible that lipid accumulation alone may even support a transition of type II fibers to more of a type I phenotype (57). These changes ultimately lead to muscles with impaired contractile capacity of both type I and type II fibers, which together lead the dramatic decrease in muscle power (product of force and speed) observed with age (58).

Aging and disuse can induce the accumulation of IMC lipid, but muscle injury is associated with a marked accumulation of intermuscular fat (ACs), likely derived from the FAPs referenced above. This phenomenon has been most well described in patients with Duchene muscular dystrophy (DMD), where the prolonged cycle of muscle injury and regeneration that accompanies dystrophin deficiency ultimately results in an accumulation of ACs and fibrous tissue in areas where muscle fibers are lost (59, 60). The loss of muscle fibers and replacement with fatty and fibrous tissues leads to muscle weakness. The extent to which muscle injury with aging, which might occur with frequent eccentric muscle contractions, contributes to accumulation of intermuscular fat is not well documented. Fatty infiltration of skeletal muscle is also common following rotator cuff muscle injury and is a major factor that limits functional recovery (61). Attenuation of fatty infiltration following rotator cuff injury with statin treatment can have a protective effect on muscle atrophy in rats (62); however, a number of studies in human subjects indicate that fatty infiltration and muscle atrophy after rotator cuff repair is very difficult to reverse (63, 64). Hyperlipidemia and type 2 diabetes are independent risk factors for rotator cuff injury (65). It is certainly possible that these risk factors may not only increase the risk of rotator cuff injury but also may contribute to an attenuated repair response following treatment by exacerbating fatty infiltration of the injured rotator cuff muscles.

## DISCUSSION: TARGETING ADIPOGENESIS AND LIPID ACCUMULATION IN MUSCLE TO PREVENT FRACTURE

One of the most effective countermeasures against fatty infiltration of muscle with aging is physical activity and regular exercise. Previous work indicates that 6 months of regular aerobic exercise combined with weight loss reduced low-density muscle (lipid measurement) and improved glucose tolerance in men aged 60+ years compared with those who just exercised alone (66). Resistance training 3 days/week in adults' age 55+ years decreased thigh intramuscular adipose tissue (67), and 1 year of brisk walking prevented fatty infiltration of muscle in older subjects (68). Importantly, resumption of physical activity following periods of sedentary activity could reverse the fatty infiltration that occurred in older adults following cessation of resistance training (69). Fracture risk in women declines with higher levels of weekly physical activity (70), and hip fracture in men is more common in those individuals with low physical activity compared with men with higher levels of physical activity (71). Resistance exercise increases leg strength and power in both older (aged 70 years) men and women (72), and this increase is associated with increased muscle fiber size (73). While the effects of exercise on bone and perhaps bone marrow ACs are more modest (74), resistance training may have a positive effect on reducing fracture risk by reducing intramuscular fat and increasing muscle strength and power.

Alternative forms of mechanical signals that are safe and can help prevent accumulation of muscular or bone marrow fat may

## REFERENCES


be desirable, particularly, for the elderly or injured who are unable to exercise or have increased risk of fracture. Low magnitude (<1 *g*; *g* = earth's gravitational field), whole body vibration has been observed to reduce adipose tissue as well as the expression of adipogenic genes in muscle (53, 75) while also acting as an anabolic signal and increasing muscle fiber area (76). Similarly, vibration has reduced bone marrow adiposity in a model of postmenopausal osteoporosis (77) and reduced bone marrowderived mesenchymal stem cell commitment to the adipogenic lineage (78). Reduced indices of adipogenesis with the application of these mechanical signals as seen in both muscle and bone may occur through a similar mechanism – bias of mesenchymal stem cell or fate away from the fat differentiation pathway. These findings suggest that mechanical stimulation in a relatively low magnitude, high-frequency domain may have the potential to preserve muscle function with age by reducing the accumulation of lipids and ACs in skeletal muscle.

## AUTHOR CONTRIBUTIONS

MH wrote the initial draft and prepared the manuscript illustrations. MM-L contributed additional narrative material on bone marrow adipogenesis and edited the manuscript. DF contributed narrative material on exercise and whole body vibration and on myosteatosis. DF also edited the manuscript.

## ACKNOWLEDGMENTS

The authors are grateful to E. Scheller and W. Cawthorn for the opportunity to prepare this contribution. Funding to MH is provided by the National Institute on Aging (NIA AG036675) and funding to MM-L is provided by the American Diabetes Association (1-16-JDF-062). The authors thank Donna Kumiski in the Electron Microscopy and Histology Core Facility for her assistance with the staining shown in **Figure 2**.


loss impacts glucose tolerance in older men. *J Clin Endocrinol Metab* (2007) 92:880–6. doi:10.1210/jc.2006-2113


**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 © 2016 Hamrick, McGee-Lawrence and Frechette. 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.*

# Vanadate Impedes Adipogenesis in Mesenchymal Stem Cells Derived from Different Depots within Bone

*Frans Alexander Jacobs, Hanél Sadie-Van Gijsen, Mari van de Vyver and William Frank Ferris\**

*Department of Medicine, Division of Endocrinology, Faculty of Medicine and Health Sciences, Stellenbosch University, Cape Town, Western Cape, South Africa*

Glucocorticoid-induced osteoporosis (GIO) is associated with an increase in bone marrow adiposity, which skews the differentiation of mesenchymal stem cell (MSC) progenitors away from osteoblastogenesis and toward adipogenesis. We have previously found that vanadate, a non-specific protein tyrosine phosphatase inhibitor, prevents GIO in rats, but it was unclear whether vanadate directly influenced adipogenesis in bone-derived MSCs. For the present study, we investigated the effect of vanadate on adipogenesis in primary rat MSCs derived from bone marrow (bmMSCs) and from the proximal end of the femur (pfMSCs). By passage 3 after isolation, both cell populations expressed the MSC cell surface markers CD90 and CD106, but not the hematopoietic marker CD45. However, although variable, expression of the fibroblast marker CD26 was higher in pfMSCs than in bmMSCs. Differentiation studies using osteogenic and adipogenic induction media (OM and AM, respectively) demonstrated that pfMSCs rapidly accumulated lipid droplets within 1 week of exposure to AM, while bmMSCs isolated from the same femur only formed lipid droplets after 3 weeks of AM treatment. Conversely, pfMSCs exposed to OM produced mineralized extracellular matrix (ECM) after 3 weeks, compared to 1 week for OM-treated bmMSCs. Vanadate (10 μM) added to AM resulted in a significant reduction in AM-induced intracellular lipid accumulation and expression of adipogenic gene markers (PPARγ2, aP2, adipsin) in both pfMSCs and bmMSCs. Pharmacological concentrations of glucocorticoids (1 μM) alone did not induce lipid accumulation in either bmMSCs or pfMSCs, but resulted in significant cell death in pfMSCs. Our findings demonstrate the existence of at least two fundamentally different MSC depots within the femur and highlights the presence of MSCs capable of rapid adipogenesis within the proximal femur, an area prone to osteoporotic fractures. In addition, our results suggest that the increased bone marrow adiposity observed in GIO may not be solely due to direct effect of glucocorticoids on bone-derived MSCs, and that an increase in femur lipid content may also arise from increased adipogenesis in MSCs residing outside of the bone marrow niche.

Keywords: mesenchymal stem cells, bone, vanadate, adipogenesis, glucocorticoids

## *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Jan Tuckermann, University of Ulm, Germany Antonia Sophocleous, University of Edinburgh, UK*

> *\*Correspondence: William Frank Ferris wferris@sun.ac.za*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 03 May 2016 Accepted: 22 July 2016 Published: 03 August 2016*

#### *Citation:*

*Jacobs FA, Sadie-Van Gijsen H, van de Vyver M and Ferris WF (2016) Vanadate Impedes Adipogenesis in Mesenchymal Stem Cells Derived from Different Depots within Bone. Front. Endocrinol. 7:108. doi: 10.3389/fendo.2016.00108*

## INTRODUCTION

Glucocorticoids (GCs) are widely used to treat a variety of inflammatory disorders, but chronic GC use can reduce bone mineral density (BMD) and lead to GC-induced osteoporosis (GIO) through an increase in bone resorption and a decrease in bone formation (1, 2). The GC-induced decrease in formation may arise from a reduction in the pool of mesenchymal stem cells (MSCs) available for osteoblastic differentiation, resulting in smaller numbers of mature bone-forming osteoblasts (1, 2).

Osteoblasts arise from multipotent MSCs, which can also differentiate into other cell-types, including adipocytes and chondrocytes (3, 4), with several lines of evidence indicating an inverse reciprocal relationship between osteoblast and adipocyte differentiation (5). Several mechanisms have been identified whereby GCs may reduce the number of osteoprogenitor cells in bone. *In vitro* studies have suggested that GCs may reduce the number of MSCs destined for osteoblastogenesis by shifting the differentiation potential of MSCs preferentially toward adipogenesis (6, 7), and this is in agreement with *in vivo* observations of GC-induced increases in marrow adiposity (8). In addition, we have previously documented the anti-proliferative effects of GCs in Naïve adipose-derived MSCs and in partially differentiated osteoblasts derived from these cells (9). Evidence from *in vivo* studies in mice and *in vitro* studies on osteoblast cell-lines indicate that GCs may also directly cause apoptosis in osteoprogenitors, osteoblasts, and osteocytes (10, 11). Taken together, these findings indicate that GCs have many deleterious effects on MSCs and on cells committed to the osteoblastic lineage, and that the prevention of these effects may restore both the number of MSCs available for osteoblastic differentiation and the number of functional osteoblasts in bone.

Earlier work in animal models found that GIO could be prevented by the non-selective protein tyrosine phosphatase inhibitor sodium orthovanadate (12), although the mechanism was not described in detail. It was subsequently shown that vanadate inhibited apoptosis of cultured osteoblasts *in vitro* and of osteocytes *in vivo* (10). Vanadate has also been shown to counteract the anti-proliferative effects of GCs on Naïve MSCs and pre-osteoblasts (9), and to inhibit adipocytic differentiation in the 3T3-L1 pre-adipocyte cell line (13). Vanadate is therefore an attractive candidate for the prevention of GIO by reversing or counteracting the effects of GCs on osteoblast precursors, possibly through a variety of mechanisms.

The aim of this study was to examine the effects of supraphysiological doses of GCs and vanadate on primary bonederived MSCs, with a view of further understanding the etiology of GIO and how vanadate ameliorates GC-induced bone disease. Bone-derived MSCs can be isolated and cultured from various parts of long bones, including bone marrow (14), compact bone (15), the periosteal layer (16), or the epiphysis of long bones (17). However, many of these studies reported subtle differences between bone marrow-derived MSCs (bmMSCs) and MSCs derived from other areas of bone, suggesting that not all MSCs populations within bone are identical and that these different populations may respond differently during disease progression and pharmacological treatments (15–17). The typical bmMSC isolation procedure utilizes the diaphysis (shaft) of the femur, thereby excluding the proximal region of the femur, which is particularly susceptible to fracture during GIO compared to other bone sites (18, 19). We have therefore isolated MSCs from the proximal region of the femur (pfMSCs) and compared these cells with bmMSCs in order to investigate whether these two osteoprogenitor populations may differ in their responses to GCs and vanadate.

## MATERIALS AND METHODS

## Experimental Animals

Experiments involving animals were approved by Stellenbosch University Ethics Committee and performed in accordance with the South African Medical Research Council Guidelines on Ethics for Medical Research in compliance with the South African Animal Protection Act (Act No. 71 of 1962). Adult male Wistar rats [12 weeks old, ~250 g mass, housed at the Stellenbosch University Animal Facility and fed *ad libitum* on standard laboratory chow (Rat and Mouse Breeder Feed, Animal Specialities, Pty. Ltd., Klapmuts, South Africa)], were used for all experiments. Animals were sacrificed *via* intraperitoneal injection with 12 mg kg<sup>−</sup><sup>1</sup> sodium pentobarbitone (Eutha-naze, Bayer, South Africa) and the femora subsequently excised.

## Cell Culture Isolation of MSCs

Bone marrow-derived MSCs were isolated using a protocol that was adapted from Nadri et al. (14) and Zhu et al. (15): The tendon and muscle tissue on the femoral surface were manually removed with sterile gauze. The ends of the bones were cleaved off at the greater trochanter and saved for the isolation of pfMSCs. The bone marrow cavity was then flushed with cell isolation media [Dulbecco's Modified Eagle Medium (DMEM) (Lonza, Verviers, Belgium)] containing 1% Penicillin/Streptomycin (P/S) (Lonza) and 20% fetal bovine serum (FBS) (Biochrom, Berlin, Germany) into a 100-mm cell culture dish using a syringe. Bone marrow from both femora was pooled and subsequently incubated overnight at 37°C in 95% humidified air containing 5% CO2. The following day, the culture was washed once with PBS to remove any non-adherent cells and debris, and subsequently cultured in growth media (DMEM with P/S supplemented with 10% FBS). For the isolation of pfMSCs, the protocol was adapted from Zhu et al. (15). The proximal region of the femur was macerated into 1 mm3 fragments and digested at 37°C for 60 min in 10 ml Hanks' Balanced Salt Solution (Lonza) containing 0.075% (w/v) collagenase I (#CLS1, Worthington, Lakewood, NJ, USA) and 1.5% bovine serum albumin (BSA). The digested proximal femur fragments were then washed five times in DMEM and seeded in a cell culture dish with cell isolation media. After 24 h, non-adherent material was washed off with PBS and the media replaced with standard growth media.

## Subculturing and Maintenance

All cell cultures were maintained at 37°C with 5% CO2 and 95% humidified air. Once the MSC cultures reached approximately 80% confluence, the cells were dissociated using 0.5% trypsin solution (Lonza) and sub-cultured at a dilution of 1:4. MSC cultures were expanded to passage 3 before being used for further experiments.

## MSC Differentiation and Cytological Staining

All constituents of the differentiation media were purchased from Sigma-Aldrich (Schnelldorf, Germany). Differentiation experiments were performed in six-well cell culture-treated dishes (Porvair, UK). For osteoblastic differentiation, the protocol was adapted from Jaiswal et al. (20). In brief, bmMSCs were cultured to confluence before being treated with osteogenic media (OM), which consisted of growth media supplemented with 50 μM ascorbic acid, 10 mM β-glycerophosphate, and 10 nM dexamethasone, with a final ethanol concentration of 0.1%. For negative vehicle controls, cells were treated with growth media containing 0.1% ethanol. OM was replaced twice a week for the duration of osteoblastic differentiation. Once osteoblastic differentiation was complete (after day 7 for bmMSCs and day 21 for pfMSCs), mineralized extracellular deposits were stained with Alizarin Red S (Amresco, USA), using a protocol adapted from Gregory et al. (21). Briefly, cells were fixed in 70% ethanol for 10 min and stained with 40 mM Alizarin Red S dissolved in water (pH adjusted to 4.1 with 5% NH4OH). Excess stain was removed after 1 h, and the samples were washed twice with water. PBS (1 ml) was added to each stained well before image capturing using an Olympus CKX41 microscope (CKX41, CachN 10×/0.25 PhP objective) and a Canon EOS 600D camera.

For adipocytic differentiation, the protocol was adapted from Ogawa et al. (22). MSCs at passage 3 were cultured to 2 days post-confluence before being treated with adipogenic media (AM), which consisted of growth media supplemented with 1 μM dexamethasone, 50 μM ascorbic acid, 56 μM indomethacin, 10 μM insulin, and 0.5 mM 3-isobutyl-1-methylxanthine. The AM was replaced every 2–3 days for the duration of differentiation. For concomitant vanadate treatments, a stock solution of 10 mM sodium orthovanadate was dissolved in water and boiled as per manufacturer's instructions (Sigma-Aldrich, Schnelldorf, Germany), and aliquots were stored at −20°C. Intracellular lipid droplets were stained with filtered 0.7% (w/v) Oil Red O (Sigma-Aldrich) in 70% isopropanol for 30 min and subsequently washed three times with water before images were captured as described above (23). For quantification of staining, isopropanol was used to extract the bound Oil Red O, and the absorbance of the extracted dye was measured at 510 nm. To correct for differences in cell density, the cells were re-stained with 0.1% (w/v) Crystal Violet (CV) nuclear stain, which was extracted with 75% ethanol, and the absorbance of the extracted dye was measured at 570 nm. Oil Red O absorption values were normalized against corresponding crystal violet absorption values (ORO/CV) and expressed as the relative triglyceride content.

## Characterization of MSC Populations using Flow Cytometry

Bone marrow-derived MSCs and pfMSCs (80% confluent; passage 3) were harvested by trypsinisation and re-suspended in PBS containing 1% BSA (Sigma-Aldrich, Berlin, Germany). Cell suspensions at a concentration of 1 × 106 cells per 100 μl were co-labeled with mouse anti-rat Alexa Fluor 647-conjugated CD106 (AbDSerotec #MCA4633A647T), FITC-conjugated CD90 (BD Pharmingen, # 554897), V450-conjugated CD45 (BD Horizon, # 561587), and PE-conjugated CD26 (BD Pharmingen, # 559641). Flow cytometry was performed on a BD FACS Canto II instrument using FACSDiva software. A total of 15 000 events were recorded prior to data analysis. Since a multicolor cytometric analysis was carried out, fluorescent compensation settings were established through a compensation experiment, and regions of positive and negative staining were determined through a fluorochrome minus one (FMO) experiment. An unstained control sample was used as a negative control for gating purposes, and to measure forward and side scatter. Data analysis was performed using Flow Jo Vx (Treestar, Oregon, USA) software.

## Cell Viability

Cells at passage 3 were grown to 2 days post-confluence before being treated with 1 μM dexamethasone, in the absence or presence of 10 μM vanadate, for 7 days. For crystal violet (CV) staining, samples were fixed in 70% ethanol, stained with 0.1% CV for 5 min, and washed three times with PBS before images were captured as described above (see MSC Differentiation and Cytological Staining). Crystal violet was then extracted with 75% ethanol and the absorbance measured at 570 nm. The MTT assay protocol was adapted from the MTT-based *in vitro* toxicology assay kit from Sigma-Aldrich. Cells at passage 3 were seeded into 96-well plates and treatments commenced at 2 days post-confluence. Cells were treated for 7 days, after which 10 μl of 5 mg/ml MTT stock solution was added to each well and incubated for 2 h. The color reaction was stopped by the addition of 100 μl of 10% Triton X-100 plus 0.1N HCL in isopropanol, and samples were incubated on a plate shaker until the color product was completely dissolved. The color development was quantified spectrophotometrically at 570 nm, and background absorbance at 690 nm was subtracted from each well.

## Apoptosis Detection

For the detection and quantification of apoptosis, pfMSCs were expanded to passage 3 and grown to 2 days post-confluence before being treated with 1 μM dexamethasone and/or 10 μM vanadate for 7 days. Cells were harvested by trypsinization, subjected to centrifugation at 400 × *g* for 5 min, and cell pellets were re-suspended in PBS. Apoptosis was measured according to the PE Annexin V Apoptosis Detection Kit I (BD Pharmingen, #559763). Approximately 1 × 106 cells were co-stained with PE-conjugated annexin V (a marker of early apoptosis) and 7-amino-actinomycin (7-AAD: a marker for the loss of membrane integrity and cell death) and incubated in the dark at room temperature for 15 min. Flow cytometry was performed on the FACSCanto II flow cytometer with FACSDiva software, analyzing a total of 10,000 events per sample. Further data analysis was performed on Flow Jo Vx (Treestar) software where early apoptotic cells were characterized as annexin V<sup>+</sup>/7-AAD<sup>−</sup>. Gating strategy: Unstained cells were used as a negative control, whereas cells treated with 50 μg/ml cycloheximide (4, 24, 48 h) to induce apoptosis were used as a positive control for Annexin V staining (data not shown). The positive control for the nuclear stain, 7-AAD, was obtained through prolonged exposure of cells to trypsin in order to permeabilize the cell membrane and allow 7-AAD to penetrate the cells (data not shown).

## Gene Expression Analysis (qRT-PCR)

For all gene expression analyses, cells were plated into six-well plates at passage 3. Cells were prepared for RNA extraction by lysis in Qiagen (Berlin, Germany) buffer RLT and then stored at −80°C until required. Total RNA was purified using the Qiagen RNeasy Mini kit (#74106), and 1 μg total RNA was treated with 1 U of Promega (Madison, Wisconsin, USA) RQ1 RNase-free DNase, as per manufacturer's instructions. The DNase-treated RNA samples were used as a template for cDNA synthesis, using 20-(dT) as a primer with Promega ImProm-II reverse transcriptase. Real-time semi-quantitative RT-PCR (qRT-PCR) was performed on a Rotor-Gene 6000 (Corbett Life Science) using the Quantace Sensimix No-ROX kit (Bioline, London, England). Sequences for gene-specific primers can be found in **Table 1**. Relative gene expression levels were calculated according to the ΔCt method (24) and normalized to the expression of the housekeeping gene ARBP (25).


*a Forward primer designed by authors.*

*bPrimers designed by authors.*

*c Primers derived from Qiagen RT2 Adipogenesis PCR array.*

## Statistical Analysis

All statistical analyses were performed with GraphPad Prism version 5.01. All data were expressed as average ± SD, and were analyzed using one-way ANOVA and Bonferroni's multiple comparison test.

## RESULTS

## Characterization of Isolated MSC Populations

The cell populations isolated from bone marrow and the proximal femur were analyzed at passage 3 using flow cytometry in order to assess the homogeneity of the populations as well as the expression of cell surface markers associated with the MSC phenotype. Both bone marrow-derived and proximal femur-derived cell cultures constituted single homogeneous populations (**Figure 1A**), and both cell-types expressed the MSC cell surface markers CD90 (bmMSCs 90 ± 2.5%; pfMSCs 61 ± 9%) and CD106 (bmMSCs 72 ± 4%; pfMSCs 80 ± 5%), but not the hematopoietic marker CD45 (bmMSCs 8 ± 3%; pfMSCs 12 ± 2%) (**Figure 1B**). Despite biological variability, the overall expression of the fibroblast marker, CD26, was slightly higher in pfMSCs (53 ± 14%) (*n* = 8) than in bmMSCs (37 ± 11%) (*n* = 9) (**Figure 1B**).

## Comparison of Differentiation Potential between pfMSCs and bmMSCs

Bone marrow-derived MSCs and pfMSCs were treated with OM and AM to assess the osteoblastic and adipocytic differentiation potential of the cells. BmMSCs rapidly responded to OM treatment and had completely and uniformly mineralized the extracellular matrix within 7 days, while pfMSCs formed individual mineralized nodules after 21 days of OM treatment (**Figure 2**). In contrast, pfMSCs exhibited a strong adipogenic response, with large numbers of lipid-filled cells apparent by day 7, while bmMSCs required 21 days to form intracellular lipid droplets in a few individual cells (**Figure 3**).

## The Effect of Vanadate on Adipogenesis in Bone-Derived MSCs Cytological Staining

In order to determine whether vanadate had an effect on adipogenesis in bone-derived MSCs, pfMSCs and bmMSCs were treated with AM in the absence or presence of 10 μM vanadate (treatment group AMV). Oil Red O staining was significantly decreased by approximately 50% in AMV-treated pfMSCs and bmMSCs, compared to AM-treated cells (**Figure 4**).

## Gene Expression Analysis

We hypothesized that the observed partial inhibition of AM-stimulated lipid accumulation may be either due to an effect on one or more molecules in lipid sequestering or synthesizing pathways, or by globally affecting the process of adipogenesis. To ascertain whether vanadate inhibits adipogenesis, and to gain further insight into the possible mechanism(s) involved, the effect

of vanadate on the expression of several adipogenesis-related genes was examined using qRT-PCR.

The expression of the pivotal activator of adipogenesis, PPARγ2, could not be consistently observed in untreated, Naïve pfMSCs, and was completely undetectable in Naïve bmMSCs. However, PPARγ2 expression was induced in pfMSCs after 3 days of AM treatment and was maintained until day 7 (the last time-point examined for pfMSCs) (**Figure 5A**). Vanadate did not affect the initial upregulation of PPARγ2 expression by day 3, but reduced AM-induced PPARγ2 expression by approximately 70% by day 7 (**Figure 5A**). In contrast to the rapid upregulation of PPARγ2 in pfMSCs, PPARγ2 expression was only induced in bmMSCs between 14 and 21 days of AM treatment, and this induction was reduced by 30–60% by vanadate (**Figure 5A**). C/ EBPα expression was readily detectable in untreated pfMSCs and bmMSCs (Ct values between 25 and 29), and was upregulated in response to AM at all time-points tested (**Figure 5B**), although the magnitude of the response to AM varied between cell isolates. In pfMSCs, C/EBPα expression exhibited a stronger induction at day 7 of AM treatment than at day 3, and a significant downregulation with vanadate at day 7 but not at day 3 (**Figure 5B**), similar to that observed for PPARγ2 (**Figure 5A**). In bmMSCs, vanadate also had very little effect on AM-induced C/EBPα expression at earlier time-points (days 7 and 14), but caused a 60–70% reduction in AM-induced C/EBPα expression by day 21 (**Figure 5B**). The late adipogenic markers, aP2/FABP4 (adipocyte protein-2, also known as fatty acid binding protein-4), fatty acid synthase (Fasn) and adipsin, were strongly upregulated in response to AM at day 7 in pfMSCs and day 21 in bmMSCs, although the magnitude of the responses varied dramatically. As was found for the early adipogenic markers examined, vanadate caused a significant decrease in the AM-induced expression of late adipogenic markers in both cell-types (**Figure 6**).

## Inherent Differences in Gene Expression Underlying the Variation in Differentiation Potential between pfMSCs and bmMSCs

Given the similarities in the AM-mediated upregulation of adipocyte gene expression in pfMSCs and bmMSCs, and our observations that vanadate had similar inhibitory effects on lipid accumulation and adipogenic gene expression in both cell types, we concluded that it was likely that the mechanisms governing adipogenesis and lipid accumulation were largely the same in both cell types, even though adipogenesis proceeded more rapidly in pfMSCs than in bmMSCs. However, we questioned whether differential expression of specific genes before the initiation of adipogenesis could account for the increased rate of lipid

accumulation in pfMSCs, compared to bmMSCs. First, we compared the expression of adipocyte-specific genes between Naïve pfMSCs and bmMSCs using qRT-PCR. Although the expression of C/EBPα and fatty acid synthase was readily detectable in both cell types, neither was found to be differentially regulated (**Figure 7A**). Furthermore, the expression levels of aP2/FABP4 in Naïve cells varied considerably, but did not correlate with adipogenic potential (**Figure 7A**), and GLUT4 expression was found to be very low in both cell types (data not shown). It therefore appeared unlikely that the increased lipid accumulation rate of pfMSCs could be attributed to elevated adipocyte-specific gene expression in the Naïve state, and consequently the expression of anti-adipogenic genes was examined. The anti-adipogenic/ pro-osteogenic factors Wnt1, Wnt3a, and sonic hedgehog (Shh) (30) were undetectable by qRT-PCR in either cell type (data not shown). Expression levels of the pro-osteogenic transcription factors Runx2 and Msx2 also did not differ between Naïve pfMSCs and bmMSCs (**Figure 7B**), although the antiadipogenic/pro-osteogenic factor Wnt10b (31) was expressed in Naïve bmMSCs, but not pfMSCs (**Figure 7B**). Subsequently it was found that Wnt10b expression in AM-treated bmMSCs was downregulated after 14 days (**Figure 7C**), therefore preceding the induction of PPARγ2 expression. Wnt10b expression

remained downregulated in AM-treated bmMSCs by day 21, but was unexpectedly also found to be downregulated by vanadate treatment (**Figure 7C**).

taken at 20× magnification, with the size bar = 1 mm.

## The Effects of GCs on Lipid Accumulation and Viability of pfMSCs and bmMSCs

Given that GC-induced osteoporosis is often associated with increased bone marrow adiposity (8), it was questioned whether GCs could directly induce adipogenesis in bone-derived MSCs. PfMSCs and bmMSCs were treated with 1 μM dexamethasone (Dex) for 21 days, but no lipid accumulation was observed (**Figure 8**). It was also hypothesized that GCs may reduce the viability of bone-derived MSCs, as it had been previously shown that glucocorticoids induced apoptosis in bone cells *in vivo* and *in vitro* (10). In-well staining of bone-derived MSCs with crystal violet or MTT demonstrated a significant decrease in cell density of pfMSC cultures that had been treated with 1 μM Dex for 7 days (**Figures 9A,B**). In addition, results from the annexin V apoptosis assay indicated an increase in the percentage of apoptotic cells in Dex-treated pfMSC cultures (**Figure 9C**). As previous work had found that vanadate could be protective against GC-induced apoptosis in an immortalized bone cell line (10), we examined

Figure 5 | The effect of vanadate on the expression of early markers of adipogenesis. (A,B) Cells were treated with adipogenic media (AM) or AM plus 10 μM vanadate (AMV) for the number of days as indicated, and the expression of PPARγ2 (A) and C/EBPα (B) were measured by qRT-PCR. All target gene measurements were normalized to the housekeeping gene acidic ribosomal phosphoprotein (ARBP). (A) As PPARγ2 expression was undetectable in Naïve cells, PPARγ2 expression in day 3 AM-treated pfMSCs or day 21 AM-treated bmMSCs was set as 1. The graph presents the combined data of *n* = 3 for bmMSCs and *n* = 3 for pfMSCs. (B) C/EBPα expression in control cells was set as 1. The graph presents representative data from one experimental repeat out of three repeats for bmMSCs and one out of four repeats for pfMSCs.

whether vanadate could counteract the apoptotic effects of Dex on pfMSCs. However, both cytological staining and apoptosis measurements showed that vanadate could not rescue pfMSCs from the negative effects of Dex treatment (**Figures 9A–C**). Vanadate treatment also had no effect on the viability of bmMSCs (**Figures 9A,B**).

## DISCUSSION

In the present study, we report for the first time the isolation and characterization of an MSC population, residing in the proximal end of rat femurs (pfMSCs) that is phenotypically similar to, but functionally distinct from bmMSCs. Compared to bmMSCs, pfMSCs exhibited a more rapid adipogenic response but a delayed and impaired osteogenic response. However, despite these apparent differences in differentiation responses, gene expression analysis indicated that the mechanisms involved in adipogenesis and lipid accumulation are likely to be the same in both cell types. It was also found that GCs strongly reduced the viability of pfMSCs by stimulating apoptosis in these cells, while bmMSCs were resistant to the cytotoxic effects of GCs. Vanadate partially inhibited adipogenesis and lipid accumulation in both cell types, but could not reverse the GC-induced cytotoxicity in pfMSCs.

Cell surface marker analysis indicated that both pfMSCs and bmMSCs were strongly positive for the mesenchymal markers CD90 (32) and CD106 (33), and negative for the hematopoietic marker CD45 (34). However, it was found that CD26 was expressed at higher levels in pfMSCs than in bmMSCs. CD26

is considered to be a fibroblast marker, and earlier work has described that CD26 and CD106 may be used to distinguish between bmMSCs and fibroblasts (33, 35). However, fibroblasts are terminally differentiated cells, whereas the maintenance of an undifferentiated phenotype, proliferative- and bipotential differentiation capacity shown in the bmMSCS and pfMSCs in the present study indicate that these are Naïve progenitor cells, suggesting that CD26 expression is not absolutely indicative of the differentiated fibroblastic phenotype.

The most dramatic difference observed between pfMSCs and bmMSCs was in the adipocytic and osteoblastic differentiation potential, with pfMSCs being highly adipogenic and bmMSCs being highly osteogenic. Unexpectedly, despite these differences in differentiation potential, neither cell types exhibited spontaneous differentiation in culture, and we found that the expression of pro-adipogenic (PPARγ2, C/EBPα) (36) and pro-osteogenic transcription factors (Runx2, Msx2) (37, 38), as well as factors related to adipocyte function (aP2/FABP4, fatty acid synthase, GLUT4) (39, 40), did not differ between Naïve pfMSCs and bmMSCs. However, we found that Wnt10b was expressed in bmMSCs, but was absent in pfMSCs.

It is well established that an inverse relationship exists between osteoblastogenesis and adipogenesis in MSCs, with signals stimulating the one process often reciprocally inhibiting the other (5, 41). Wnt10b was first identified as a crucial "offswitch" for adipogenesis *in vitro* in 3T3-L1 pre-adipocytes (42) and subsequently *in vivo* in transgenic mice (43). Conversely, it was found that Wnt10b promoted osteoblastogenesis *in vitro* in bipotential ST2 cells (31, 44) and *in vivo* in transgenic mice (44, 45). Furthermore, proximal femur bone samples from osteoporotic women with hip fractures exhibit decreased Wnt10b expression (46), providing evidence for a relationship between Wnt10b expression, the maintenance of osteoblastic differentiation and bone strength. In addition, calvarial cultures from transgenic mice with reduced Wnt10b expression displayed dominant adipogenesis and a reduction in osteoblastogenesis (47), similar to the association between Wnt10b expression and osteoblastic differentiation seen in pfMSCs and bmMSCs in the present study. It may therefore be likely that the difference in differentiation potential between pfMSCs and bmMSCs is driven, at least in part, by a dissimilarity in Wnt10b expression. However, apart from Wnt10b expression and the concomitant delay in the onset of adipogenesis in bmMSCs, compared to pfMSCs, the profile of adipogenesis-related genes expressed in response to AM did not differ between these two cell types, indicating that the mechanism of adipocyte development from these two distinct progenitor populations was highly similar.

Glucocorticoid-induced osteoporosis is often associated with a decrease in osteoblast number and an increase in bone marrow adiposity (8). Previously, it has been found that vanadate prevented GC-induced osteoporosis *in vivo* in rats by restoring osteoblast numbers and preventing osteocyte apoptosis (10, 12). However, to our knowledge, there is no information available on the effects of vanadate on adipogenesis and lipid accumulation in bone-derived MSCs. The results presented here indicate that vanadate was able to inhibit lipid accumulation and the expression of early and late adipogenesis markers in pfMSCs and bmMSCs, demonstrating that vanadate inhibited the adipogenic program, and not just the later phases of lipid accumulation, similar to that previously found in immortalized 3T3-L1 pre-adipocytes (13). However, it should be noted that AM-induced expression of the early adipogenesis markers PPARγ2 and C/EBPα was only inhibited by vanadate at later time-points, after the expression of these genes had been upregulated by AM, suggesting the possibility that vanadate-mediated inhibition of adipogenesis involves an upstream signal that is induced during the initial phases of the

adipogenic response, with consequential effects only occurring later in the differentiation process. Furthermore, we observed that the vanadate-mediated inhibition of adipogenesis in bmMSCs was not associated with upregulated expression of the antiadipogenic Wnt10b (42, 43), but that Wnt10b was unexpectedly downregulated by AM as well as by vanadate. The inhibition of adipogenesis in MSCs by vanadate is therefore not underpinned by increased Wnt10b expression, suggesting that other signals are activated by vanadate that can inhibit adipogenesis without a requirement for Wnt10b.

Our finding that bmMSCs showed no differences in cell viability when cultured in a high concentration of Dex (1 μM) suggests that cultured primary bmMSCs are resistant to the detrimental effects of high doses of GCs, even though other studies have described decreased viability, proliferation and osteogenic potential in bmMSCs isolated from GC-treated animals (7, 48). It is therefore possible that these effects of GCs are not the result of direct GC actions at a cellular level, but that they may be mediated systemically. In contrast, pfMSCs were found to be exquisitely sensitive to the cytotoxic and apoptotic effects of GCs, but these effects could not be counteracted by vanadate co-administration. This is in contrast with previous findings, which demonstrated that *in vivo* vanadate treatment could rescue osteoblasts and osteocytes from the negative effects of GCs (10, 12), again suggesting that the effects of GCs and vanadate on bone may be systemic rather than direct.

It has been noted that GC-induced osteoporosis may result in a greater increase in fractures of the vertebrae and proximal femur, compared to fractures at other sites (18, 19), and that GC-induced fat conversion takes place in the proximal femur (8). In addition, studies on patients with GC-induced osteonecrosis of the proximal femur found a decreased proliferative capacity in isolated MSCs (49) and a decrease in progenitor cell numbers in regions adjacent to the necrotic areas (50). Moreover, it was demonstrated in rabbit models that GC treatment induced both marrow fat cell hyperplasia and hypertrophy in the proximal femur (51, 52), and that GC-induced osteonecrosis was also associated with marrow fat cell hypertrophy in this region of bone (53). While our results indicated that GCs alone could not induce lipid accumulation in either pfMSCs or bmMSCs, GC treatment reduced the viability and increased apoptosis in pfMSCs, but not in bmMSCs. These findings, taken together with the highly adipogenic nature of pfMSCs, compared to bmMSCs, suggest that distinct MSC populations may exist in the proximal femur that are more susceptible to adipogenic and apoptotic signals than other bone-derived MSC populations, such as bmMSCs. In addition, it may also be possible that the fat content of the proximal femur, which is usually labeled as "marrow adiposity," may actually arise from distinct pfMSCs, rather than from bmMSCs from the diaphyseal region.

## AUTHOR CONTRIBUTIONS

FJ, HS, and MV were responsible for the conceptual design of the study, experimental procedures, sample collection and analysis, interpretation of data, and preparation of the manuscript. WF contributed to the conceptual design and editing of the manuscript.

## FUNDING

This work was supported by research funding from the South African National Research Foundation (NRF), the South African

## REFERENCES


Medical Research Council (MRC) and the Harry Crossley Foundation. FJ was supported by a PhD bursary from the NRF and MV was supported by a postdoctoral fellowship from the Faculty of Medicine and Health Sciences, Stellenbosch University.


**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 © 2016 Jacobs, Sadie-Van Gijsen, van de Vyver and Ferris. 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.*

# Changes in Skeletal Integrity and Marrow Adiposity during High-Fat Diet and after Weight Loss

*Erica L. Scheller1,2\*, Basma Khoury3 , Kayla L. Moller1 , Natalie K. Y. Wee4 , Shaima Khandaker2 , Kenneth M. Kozloff3 , Simin H. Abrishami5 , Brian F. Zamarron6 and Kanakadurga Singer5*

*1Division of Bone and Mineral Diseases, Department of Medicine, Washington University, St. Louis, MO, USA, 2Department of Molecular and Integrative Physiology, University of Michigan, Ann Arbor, MI, USA, 3Department of Orthopaedic Surgery, University of Michigan, Ann Arbor, MI, USA, 4Osteoporosis and Bone Biology Division, Garvan Institute of Medical Research, Darlinghurst, Sydney, NSW, Australia, 5Division of Pediatric Endocrinology, Department of Pediatrics and Communicable Diseases, University of Michigan Medical School, Ann Arbor, MI, USA, 6Graduate Program in Immunology, University of Michigan, Ann Arbor, MI, USA*

## *Edited by:*

*Ann Schwartz, University of California San Francisco, USA*

#### *Reviewed by:*

*Jan Josef Stepan, Charles University in Prague, Czech Republic Roberto Jose Fajardo, University of Texas Health Science Center at San Antonio, USA*

*\*Correspondence:*

*Erica L. Scheller scheller@wustl.edu*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 16 May 2016 Accepted: 08 July 2016 Published: 27 July 2016*

#### *Citation:*

*Scheller EL, Khoury B, Moller KL, Wee NKY, Khandaker S, Kozloff KM, Abrishami SH, Zamarron BF and Singer K (2016) Changes in Skeletal Integrity and Marrow Adiposity during High-Fat Diet and after Weight Loss. Front. Endocrinol. 7:102. doi: 10.3389/fendo.2016.00102*

The prevalence of obesity has continued to rise over the past three decades leading to significant increases in obesity-related medical care costs from metabolic and nonmetabolic sequelae. It is now clear that expansion of body fat leads to an increase in inflammation with systemic effects on metabolism. In mouse models of diet-induced obesity, there is also an expansion of bone marrow adipocytes. However, the persistence of these changes after weight loss has not been well described. The objective of this study was to investigate the impact of high-fat diet (HFD) and subsequent weight loss on skeletal parameters in C57Bl6/J mice. Male mice were given a normal chow diet (ND) or 60% HFD at 6 weeks of age for 12, 16, or 20 weeks. A third group of mice was put on HFD for 12 weeks and then on ND for 8 weeks to mimic weight loss. After these dietary challenges, the tibia and femur were removed and analyzed by micro computed-tomography for bone morphology. Decalcification followed by osmium staining was used to assess bone marrow adiposity, and mechanical testing was performed to assess bone strength. After 12, 16, or 20 weeks of HFD, mice had significant weight gain relative to controls. Body mass returned to normal after weight loss. Marrow adipose tissue (MAT) volume in the tibia increased after 16 weeks of HFD and persisted in the 20-week HFD group. Weight loss prevented HFD-induced MAT expansion. Trabecular bone volume fraction, mineral content, and number were decreased after 12, 16, or 20 weeks of HFD, relative to ND controls, with only partial recovery after weight loss. Mechanical testing demonstrated decreased fracture resistance after 20 weeks of HFD. Loss of mechanical integrity did not recover after weight loss. Our study demonstrates that HFD causes long-term, persistent changes in bone quality, despite prevention of marrow adipose tissue accumulation, as demonstrated through changes in bone morphology and mechanical strength in a mouse model of diet-induced obesity and weight loss.

Keywords: obesity, bone, marrow adipose tissue, marrow fat, weight loss, leptin, high-fat diet, fracture

## INTRODUCTION

Over the past two decades, the prevalence of obesity has increased in Western countries (1, 2). In the United States, currently ~68.6% of adults and approximately one-third (~31.8%) of children are overweight or obese (3). Obesity is associated with comorbidities including cardiovascular and metabolic disease, autoimmune disorders and some cancers (1, 4–6). Recent work has suggested that obesity is also detrimental to bone health (7–11), with skeletal changes that can persist even after weight loss (10, 12).

Previously, it was assumed that obesity had a purely positive effect on bone mass (13–15); increased body weight provides mechanical stimulation, resulting in skeletal loading and bone accrual. However, juxtaposed to this, there is a newly recognized metabolic component, as the adipose tissue itself can exert a negative influence on bone (14). Indeed, increases in body mass index (BMI) have been associated with decreased bone mineral density (BMD) and increased fracture risk in obese adolescents and adults (9, 16), and in obese children (17). The effect of obesity on fracture risk is site specific. The presence of soft-tissue padding from fat may contribute to decreased fracture risk in some areas (e.g., hip) while unprotected sites, such as the extremities (e.g., humerus and ankle), have increased risk (18–20).

The cross-sectional nature of previous clinical studies can only identify associations between obesity and bone, thus, rodent models are widely utilized to explore the mechanisms underlying the relationship between obesity and the skeleton. It is well established that high-fat feeding of mice leads to a reduction in cancellous bone mass (7, 12, 21, 22). This may be mediated by leptin-induced sympathetic tone, which has been implicated as strong mediator of cancellous bone loss (23–25). By comparison, the cortical phenotype in response to high-fat diet (HFD) in rodents remains unclear, with some studies indicating an increase (11), no change (12, 21, 22, 26) or a reduction in cortical bone mass (10, 27). Located within the skeleton are the bone marrow adipocytes; recent studies suggest that marrow adipose tissue (MAT) expansion occurs during high-fat feeding (28, 29). Whether MAT expansion and bone loss are somehow linked during obesity is still unclear; some studies suggest that these lineages are correlated (29–31) while Doucette et al. recently reported MAT expansion during diet-induced obesity that occurred independently of a bone phenotype (28).

In addition to the effects of obesity on bone, weight loss interventions have also been shown to have detrimental effects on bone metabolism, as reviewed by Brzozowska et al. (32). There are a range of interventions including calorie-restricted diets, exercise regimens, medications, and bariatric surgery (32, 33). Each of these interventions aim to reduce body fat and improve metabolic disease; the full extent to which these processes may alter MAT and bone mass in the context of obesity are largely unknown. Surgical interventions of bariatric surgery (Roux-en Y gastric bypass, laparoscopic adjustable gastric banding, and sleeve gastrectomy) have all been associated with a decline in bone mass despite improvements in metabolic health (32). In contrast to surgical weight loss, exercise has been shown to be quite beneficial on bone density due to increased muscle loading (34–36). The most common initial intervention clinically is calorie restriction or "dieting." Few studies have looked at weight loss in rodent models through interventions of "switching" diet. One study performed showed that switching back to a chow diet following high-fat feeding could rescue bone loss (12); however, the response of MAT and the interaction of MAT with bone loss in these models was not examined.

The objective of this study was to investigate the interaction between MAT and bone in the context of high-fat feeding and to examine the response of these tissues to dietary weight loss. We demonstrate that high-fat feeding leads to excess peripheral adiposity, MAT expansion, a reduction in bone mass and impaired bone strength. Weight loss led to a significant reduction in whole body adiposity and blocked MAT expansion; however, it failed to completely rescue defects in skeletal morphology and biomechanics. This work begins to address the potential of adipose tissue within the skeleton to have an impact on bone – working, unlike peripheral fat, from the inside out.

## MATERIALS AND METHODS

## Animals

Male C57Bl6/J mice (Jackson Laboratories) were given a normal chow diet (ND) (13.5% calories from fat; LabDiet 5LOD) or 60% high fat diet (HFD) (Research Diets D12492) at 6 weeks of age for a duration of 12, 16, or 20 weeks. A third group of mice was put on HFD for 12 weeks and then on ND for 8 weeks [weight loss (WL) group]. Animals were housed in a specific pathogen-free facility with a 12-h light/12-h dark cycle at ~22°C and given free access to food and water. All animal use was in compliance with the Institute of Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and approved by the University Committee on Use and Care of Animals at the University of Michigan. The tibia was selected for our longitudinal analyses since it can be used to simultaneously monitor changes in rMAT (proximal tibia) and cMAT (distal tibia) within one sample (37). To compare the changes in bone within the tibia to those in the femur, as reported previously (12), we also analyzed the femurs in the 20-week groups.

## Micro Computed-Tomography

Tibiae were fixed in formalin for 48-h and then placed in phosphate buffered saline (PBS). Specimens were embedded in 1% agarose and placed in a 19-mm diameter tube, and the length of the bone was scanned using a Micro Computed-Tomography (microCT) system (μCT100 Scanco Medical, Bassersdorf, Switzerland). Scan settings were: voxel size 12 μm, medium resolution, 70 kVp, 114 μA, 0.5 mm AL filter, and integration time 500 ms. Density measurements were calibrated to the manufacturer's hydroxyapatite phantom. Analysis was performed using the manufacturer's evaluation software.

Femurs were removed and frozen after wrapping in PBS-soaked gauze and then analyzed by microCT. Femora were scanned in water using cone beam computed tomography (explore Locus SP, GE Healthcare Pre-Clinical Imaging, London, ON, Canada). Scan parameters included a 0.5° increment angle, four frames averaged, an 80 kVp and 80 μA X-ray source with a 0.508 mm AI filter to reduce beam hardening artifacts, and a beam flattener around the specimen holder. All images were reconstructed and calibrated at an 18 μm isotropic voxel size to manufacturer-supplied phantom of air, water, and hydroxyapatite (38).

## Biomechanical Assessment

Following microCT scanning, femurs were loaded to failure in four-point bending using a servohydraulic testing machine (MTS 858 MiniBionix, Eden Prairie, MN, USA). All specimens were kept hydrated in lactated ringers solution-soaked gauze until mechanical testing. In the same mid-diaphyseal region analyzed by μCT, the femur was loaded in four-point bending with the posterior surface oriented under tension. The distance between the wide, upper supports was 6.26 mm, and the span between the narrow, lower supports was 2.085 mm. The vertical displacement rate of the four-point bending apparatus in the anterior–posterior direction was 0.5 mm/s. Force was recorded by a 50 lb load cell (Sensotec) and vertical displacement by an external linear variable differential transducer (LVDT, Lucas Schavitts, Hampton, VA, USA), both at 2000 Hz. A custom MATLAB script was used to analyze the raw force-displacement data and calculate all four-point bending parameters. Combining anterior–posterior bending moment of inertia data from μCT with mechanical stiffness from four point bending, the estimated elastic modulus was calculated using standard beam theory as previously described (38). The modulus of elasticity was derived based on previous methods with "L" set at 3.57 and "a" at 0.99 (39).

## Quantification of Trabecular and Cortical Parameters with microCT

*Tibia.* Regions of interest (ROI) was located for both cortical and trabecular parameters. Analyses were performed with MicroCT software provided by Scanco Medical (Bassersdorf, Switzerland). A mid-diaphyseal cortical ROI was defined as ending at 70% of the distance between the growth plate and the tibia/fibula junction. A ROI spanning 360 μm (30-slices) proximal to this region was analyzed with standard plugins using a threshold of 280. The trabecular ROI was defined as starting 60 μm (5-slices) distal to the growth plate and ending after 600 μm total (50-slices). Trabecular analyses were performed with standard Scanco plugins with a threshold of 180.

*Femur.* ROI was located for both cortical and trabecular parameters. A diaphyseal cortical ROI spanning 18% of total femur length was located midway between the distal growth plate and third trochanter. Cortical bone was isolated with a fixed threshold of 2000 Hounsfield Units for all experimental groups. Parameters including cortical thickness, endosteal and periosteal perimeter, cross sectional area, marrow area, total area, anterior–posterior bending moment of inertia, and tissue mineral density (TMD) were quantified with commercially available software (MicroView v2.2 Advanced Bone Analysis Application, GE Healthcare Pre-Clinical Imaging, London, ON, Canada). A trabecular ROI 10% of total femur length was located immediately proximal to the distal femoral growth plate and defined along the inner cortical surface with a splining algorithm. Trabecular metaphyseal bone was isolated with a fixed threshold of 1200 Hounsfield Units.

## Quantification of Marrow Adipose Tissue

Marrow adipose tissue volume within the tibia was assessed as described previously (37, 40). After the initial microCT scan, bones were decalcified in 14% EDTA solution, pH 7.4 for 14 days at 4°C. Decalcified bones were stained with 1% osmium tetroxide solution in Sorensen's phosphate buffer pH 7.4 at room temperature for 48 h. Osmium-stained bones were re-scanned using the Scanco microCT settings described above. For analysis of MAT within the tibia, four regions were defined as follows: (1) the proximal epiphysis between the proximal end of the tibia and the growth plate, (2) the proximal metaphysis, beginning 60 μm (5-slices) distal to the growth plate and ending after 600 μm total (50-slices), (3) the growth plate to the tibia/fibula junction (GP to T/F J), and the distal tibia between the tibia/fibula junction and the distal end of the bone. MAT volume analyses were performed with standard Scanco plugins with a threshold of 500.

## Statistics

Statistical comparisons were performed in GraphPad Prism (GraphPad Software, Inc., La Jolla, CA, USA). The following planned comparisons were performed on the graphs in **Figures 1**, **2**, **3** and **5**: 12-week ND vs. HFD (two-tailed *t*-test); 16-week ND vs. HFD (two-tailed *t*-test); 20-week ND vs. HFD vs. WL (1-way ANOVA); 12-, 16-, 20-week ND (1-way ANOVA); 12-, 16-, 20-week HFD (1-way ANOVA); 12-week HFD vs. 20-week WL (two-tailed *t*-test). These results were corrected for multiple comparisons using the Benjamini-Hochberg procedure as described previously (41). For comparisons in **Figures 4**, **6** and **7**, a one-way ANOVA with Tukey's correction was applied. In **Figure 8**, linear regression was applied to test the significance of the correlations. Raw data for the skeletal morphology, marrow fat quantification, and biomechanical testing is available in Data Sets 1–3 in Supplementary Material.

## RESULTS

## Increases in Body Mass with High-Fat Diet Are Rescued by Weight Loss

Mice were fed normal chow diet (ND) or 60% high-fat diet (HFD), starting at 6 weeks of age, for 12, 16, or 20 weeks. A separate group of mice received 12 weeks of HFD, followed by 8 weeks of ND to mimic weight loss (**Figure 1A**). Comparison of the 12-week HFD group to the WL group was used to determine if weight loss reversed changes that were already present at 12 weeks, or, rather, prevented further deterioration induced by continued HFD.

Increases in body mass relative to ND control were apparent after 12-weeks of HFD (**Figure 1B**). Increases in body mass relative to ND control persisted at the 16- and 20-week time points (**Figure 1B**). Over time, body mass continued to increase from 12 to 16 weeks of HFD but stabilized between 16 and 20 weeks (**Figure 1B**). Body mass returned to normal after weight loss (**Figure 1B**). Weight gain was due, at least in part, to increases in liver, inguinal and gonadal white adipose tissue (WAT) mass after 20-weeks of HFD (**Figure 1C**). With weight loss, liver mass returned to normal; however, WAT mass was only partially rescued (**Figure 1C**).

same diet. "c" – significant vs. 12-week HFD. \**p* < 0.050 for the indicated comparison.

## Marrow Adipose Tissue Expansion after High-Fat Diet Feeding Is Inhibited with Weight Loss

Within the tibia, MAT expansion became significant, relative to ND control, after 16 weeks of HFD (**Figures 2A,B**). With HFD, changes in the proximal tibial epiphysis mimicked what was observed between the growth plate and tibia/fibula junction, with a 5.5- and 4.3-fold increase at 16 weeks, relative to 12 weeks, respectively (**Figure 2B**). The distal tibia was similar, though there was only a 2.1-fold increase between 12 and 16 weeks of HFD, likely owing to the higher baseline MAT in this region (**Figure 2B**). No additional MAT accrual in any region of the tibia was observed between 16 and 20 weeks of HFD (**Figure 2B**).

In the weight loss group, the MAT in the regions of the proximal epiphysis and GP to T/F J was indistinguishable from that of the 20-week ND group (**Figure 2B**). However, it was also similar in magnitude to the 12-week HFD group, suggesting that switching to ND was sufficient to block HFD-induced MAT expansion – rather than reversing MAT accrual that had already occurred.

In the distal tibia, age-associated increases in MAT were noted from 12- to 16-weeks of age in the ND group. Prior to correction for multiple comparisons, MAT within the distal tibia in the WL group was higher than chow (*p* = 0.019) but less than HFD (*p* = 0.050). This suggests that weight loss blunted, but did not entirely prevent, HFD-induced MAT expansion in the distal tibia at 20-weeks (**Figure 2B**). Raw data for these comparisons is available as Data Set 1 in Supplementary Material.

## In the Tibia, Trabecular Bone Quality Decreases with High-Fat Diet and Partially Improves with Weight Loss

Consistent with previous reports (42, 43), we observed an age-related decrease in trabecular bone volume fraction (BVF) and trabecular number, with a corresponding increase in spacing, in the proximal tibial metaphysis of ND control mice (**Figures 3A,C,E,G**).

Relative to ND controls, mice fed a HFD had a significant decrease in trabecular BVF, bone mineral content (BMC), and number after 12-weeks of diet (**Figures 3C,D,E**). Structure model index and trabecular spacing were reciprocally increased (**Figures 3F,G**). Trabecular thickness remained unchanged (Data Set 1 in Supplementary Material). Loss of trabecular BVF and BMC with HFD, relative to control ND, persisted at the 16- and 20-week time points (**Figures 3C–G**). Weight loss partially rescued decreases in trabecular BVF, BMC, and number (**Figures 3C–E**) and increases in spacing (**Figure 3G**).

Unlike loss of trabecular bone at 12-weeks, MAT volume was not significantly increased relative to ND control until 16- and 20-weeks of HFD (**Figure 3B**). Over time, MAT increased by 3.5 fold from 12- to 16-weeks of age in the HFD group (**Figure 3B**).

tibia bone (light gray). (B) Region-specific quantification of MAT in the proximal tibial epiphysis, between the growth plate to tibia/fibula junction (GP to T/F J), and the distal tibia. All graphs are mean ± SEM. *N* = 3-6 per group for the proximal epiphysis; low *N* is due to accidental removal/fracture of the proximal epiphysis during processing. *N* = 6–8 for all other groups. ND, normal chow diet; HFD, high-fat diet; WL, weight loss. "a" – significant vs. 12-week on same diet. "b" – significant vs. 16-week on same diet. "c" – significant vs. 12-week HFD. \**p* < 0.050 for the indicated comparison.

No further increases were present from 16- to 20-weeks of age. Weight loss completely prevented HFD-induced MAT accumulation from 12- to 20-weeks (**Figure 3B**).

## In the Femur, Trabecular Bone Quality Decreases with High-Fat Diet and after Weight Loss

Relative to ND control, HFD caused MAT expansion within the femur (**Figure 4A**). MAT expansion was absent in the WL group (**Figure 4A**). The absolute amount of MAT in the distal femoral metaphysis was ~90% less than the proximal tibia (*p* < 0.0001, *t*-test) (**Figures 3B** and **4A**). The magnitude of the loss of BVF with HFD at 20-weeks was comparable between femur and tibia (50 vs. 45%) (**Figure 4B**; Data Sets 1 and 2 in Supplementary Material). There was also a significant, HFD-induced decrease in trabecular number and increase in trabecular spacing (**Figures 4D,F**). Trabecular thickness and tissue mineral density remained unchanged (**Figures 4C,E**). With weight loss, there were no statistically significant differences in trabecular morphology,

relative to HFD, in the femur (**Figures 4B–F**). There was a nonsignificant trend toward a decrease in trabecular spacing in the WL group relative to HFD (**Figure 4F**).

## Changes in Cortical Bone after High-Fat Diet and Weight Loss

Within the tibia, there were no statistically significant changes in mid-diaphyseal cortical morphology after 12, 16, or 20 weeks of HFD or after weight loss (**Figures 5A–F**). However, slight differences may have been missed after statistical correction for multiple comparisons. For example, with standard one-way ANOVA at the 20-week time point only, there was a slight decrease in cortical thickness in the 20-week HFD group relative to ND control (*p* = 0.045) (**Figure 5D**). Raw data are available in Data Set 1 in Supplementary Material.

In the femur, 20-week HFD caused a significant decrease in cortical thickness relative to ND control (**Figure 6D**). Cortical tissue mineral density was also decreased with HFD (**Figure 6F**). With weight loss, cortical thickness and total mineral density improved relative to ND (**Figures 6D,F**). No differences in total area, marrow area, cortical area, or Iyy were noted (**Figures 6A–F**). Raw data are available in Data Set 2 in Supplementary Material.

## Marrow Adipose Tissue Expansion Correlates with Bone Loss in the Tibia

In the control ND group, pooled over all ages (12, 16, and 20 weeks of diet), there was a significant inverse correlation between MAT volume in the proximal metaphysis and measures of trabecular morphology including BVF, trabecular thickness, and BMC – but not with trabecular number (**Figures 7A–D**). By contrast, in the HFD group, MAT volume was negatively correlated with trabecular BVF, thickness, and BMC in addition to trabecular number (**Figures 7A–D**). Cortical thickness was significantly negatively correlated with GP to T/F J MAT volume in the HFD group only (**Figure 7E**). Cortical TMD did not correlate with MAT volume in either group (**Figure 7F**).

## High-Fat Diet Causes Persistent Decreases in Biomechanical Properties of the Femur

Four-point bending was performed to assess the biomechanical integrity of HFD and WL femurs. The femurs from the 20-week HFD and WL groups broke under a reduced maximum load relative to ND, trending toward less total work to induce fracture (**Figures 8A,B**). This indicates that despite recovery of cortical

thickness and mineral density with WL (**Figure 6**), the bone quality remains impaired, leading to decreased fracture resistance and poor post-yield behavior. The yield load and post-yield work trended toward a decrease relative to ND in the HFD and WL groups, respectively (*p* < 0.1) (**Figures 8D,E**). The stiffness, post-yield displacement, and modulus of elasticity (39) were not significantly different between groups (**Figures 8C,F**) (Data Set 3 in Supplementary Material).

## DISCUSSION

To our knowledge, this is the first study that has measured, within the same bone, HFD-induced MAT expansion and changes in skeletal morphology. By incorporating a weight loss group, we were also able to inhibit MAT expansion, and thus examine the impact of HFD on bone quality in the absence of MAT accumulation. In this study, HFD caused increases in body mass at 12-weeks, indicating accumulation of peripheral adiposity. This occurred prior to increases in MAT, supporting the hypothesis that dysfunction of peripheral tissues (e.g., insulin resistance) occurs prior to HFD-induced MAT expansion.

As adipocytes and osteoblasts arise from the same mesenchymal progenitor cell, the notion of "fate-switching" whereby one lineage is favored over the other has been suggested (29–31). The inhibition of osteoblast differentiation may subsequently lead to increased adipocyte production, thus presenting with a situation of reduced bone mass and increased MAT (44). It is of note that this concept fails to capture the complexity of skeletal progenitors – some of which have the capacity to differentiate into osteoblasts but not adipocytes (45). In our study we observed the well-documented inverse correlation between MAT volume and bone mass/density in the tibia (**Figure 7**). However, despite this correlation, our data do not support the hypothesis that MAT expansion is the sole mediator of bone loss with HFD. Specifically, deterioration of trabecular architecture occurred as early as 12-weeks after HFD in the tibia, while changes in MAT did not become statistically significant until 16 weeks (**Figures 2** and **3**). Furthermore,

though switching from HFD to chow at 12-weeks completely prevented HFD-induced MAT accumulation in the WL group (**Figure 3B**), loss of trabecular number and corresponding increases in trabecular spacing beyond that of controls still occurred (**Figures 3E,G**). Thus, in this context, inhibition of MAT expansion by weight loss was not sufficient to block HFD-induced decreases in trabecular bone within the tibial metaphysis.

There are many MAT-independent effects with the potential to regulate bone during high-fat feeding, the presence of which may contribute to the cancellous and cortical bone loss observed in this model. Increased fat mass is associated with increased systemic markers of oxidative stress in both humans and mice (4). Increased peroxide (H2O2) and reduced endothelial nitric oxide synthase in a genetic model of obesity was associated with cancellous bone loss (46). Reactive oxygen species have been found to promote the association of the transcription factors FoxO with β-catenin, subsequently leading to a reduction in Wnt signaling and osteoblastic differentiation (47). Although the direct effects of leptin may also promote osteoblast proliferation and differentiation (25, 48), the central effects of leptin have been shown to mediate the opposite effects, promoting cancellous bone loss *via* the sympathetic nervous system (23, 24). Another central pathway involving increased neuropeptide Y (NPY) arising from leptin resistance during obesity is implicated in bone metabolism as mice with increased central NPY have concurrent obesity with bone loss (49) and NPY deficiency in *ob/ob* mice leads to improved cortical bone mass (50).

Lastly, there is an increase in systemic inflammation with obesity that might directly affect bone marrow osteoclasts. A major source of obesity-induced inflammation stems from an increase in bone marrow macrophages and their progenitors (51). These bone marrow-derived macrophages during obesity mediate an inflammatory environment that has been shown to stimulate osteoclastogenesis and reduce osteoblast development (52, 53), possibly due to the expansion of the common monocyte-osteoclast progenitor (54). Recently, Yue et al. have also demonstrated that leptin produced from obese adipose tissue can directly bind to leptin receptors on mesenchymal stem cells promoting differentiation of adipocytes and inhibiting osteoblast formation (29). Altogether there are a number of MAT-independent variables involved in coordinating the relationship between diet-induced obesity and bone.

Though it is not the sole mediator of bone loss with HFD, our study does not rule out the possibility that MAT, particularly when present in large excess, may exert detrimental effects on bone. Indeed, the magnitude of cancellous bone loss in the

for the indicated diet.

tibia (**Figure 3**) and cortical bone loss in the femur (**Figure 6**) was significantly greater in the 20-week HFD group with MAT expansion than the WL group in which MAT expansion failed to occur. Comparisons between the femur and tibia provide further clues as to this relationship. Consistent with a previous report, 12-weeks of HFD followed by 8-weeks of normal chow diet (WL group) did not prevent cancellous bone loss in the distal femoral metaphysis (12). By contrast, in the same animals, weight loss partially prevented HFD-induced deterioration of trabecular BVF and BMC in the tibia (**Figure 3**). It is possible

that this discrepancy may be explained by differences in MAT. After 20-weeks of HFD, the volume of MAT in the metaphysis of the tibia was 9.8-fold greater than in the femur. This is similar to previous work by Halade et al., despite substantial differences in their model system (10% corn oil diet for 24-weeks in 12-month-old female mice) (44). Thus, it is possible that this increase in MAT contributed to additional bone loss in the tibia, beyond that observed in the femur, subsequently leading to a difference between the HFD and WL groups. However, given previous work, the nuances of this observation remain unclear (44).

Direct interactions between MAT and bone may influence bone loss during high fat feeding. Recently, MAT was found to be a significant contributor of circulating adiponectin during calorie restriction (55), this emphasizes the potential of MAT to influence not only bone but also whole body homeostasis. Direct adipose-bone pathways have been demonstrated to influence bone mass; the main two adipokines implicated are leptin (29, 48, 56–58) and adiponectin (59–61). More locally within the bone microenvironment, *in vitro* experiments have demonstrated that the release of free fatty acids from adipocytes inhibited osteoblast differentiation and promoted apoptosis through ROS production (62). Interestingly, co-cultures of osteoblasts and osteoclasts with adipocytes suggest that in addition to reducing osteoblastogenesis, osteoclastogenesis may be increased with increased adiposity, resulting in reduced bone mass (31).

Biomechanically, weight loss after 12-weeks of HFD was insufficient to rescue impaired fracture resistance. Indeed, the maximum load endured by the HFD and WL femurs was nearly identical – despite almost complete recovery of body mass and prevention of MAT expansion in the WL group. Comparable stiffness and modulus of elasticity in the ND, WL, and HFD groups indicates that the elastic properties of the bone were not affected. However, the failure properties were similarly reduced in both the HFD and WL groups, despite differential rescue of tissue mineral density and cortical thickness, implying that femur architecture fails to explain the impaired biomechanics. This may point to dysfunction within the organic properties of the bone, such as impaired crosslinking of collagen (26), as a potential mediator of persistent HFD-induced fracture risk.

Our study demonstrates that HFD causes long-term, persistent changes in bone quality. We started HFD at an age in which skeletal development is still highly active, likely contributing to impaired bone accrual during growth. Indeed, diet-induced obesity causes greater damage in growing bones (63). This is an important finding given the rise of obesity in pediatric populations (2, 64). Furthermore, these data demonstrate that MAT is not necessary for HFD-induced bone loss; however, MAT expansion, when present, may contribute to additional skeletal deterioration. It is likely that changes within the bone microenvironment including the adipocytes themselves are being altered but this was not examined in the current study (31, 65) and will need to be evaluated with future mechanistic investigations.

Given the rise in obesity across the age spectrum, this is a critical area of research and future studies are needed to determine the effects of weight loss (dietary or surgical) on bone density and to understand the mechanisms that drive changes in bone health. Even with the limitations a clear finding in this study is that there are some reversible and some permanent changes with HFD, followed by WL. Different regimens may be required to maintain bone health after WL, possibly with a focus on activity and diet (36).

## REFERENCES


## AUTHOR CONTRIBUTIONS

ES and KS were involved in designing studies, completion of studies, data interpretation and analysis, and manuscript preparation. BK, KM, SK, KK, SA, and BZ were involved in completion of studies, data analysis, and reviewed the final manuscript.

## FUNDING

This work was supported by grants from the National Institute of Health, K99-DE024178 and R00-DE024178 (ES) and K08-DK101755 (KS).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fendo.2016.00102


**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 © 2016 Scheller, Khoury, Moller, Wee, Khandaker, Kozloff, Abrishami, Zamarron and Singer. 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.*

# H

*Laurence B. Lindenmaier1 , Kenneth A. Philbrick1 , Adam J. Branscum2 , Satya P. Kalra3 , Russell T. Turner1,4 and Urszula T. Iwaniec1,4\**

*1Skeletal Biology Laboratory, School of Biological and Population Health Sciences, Oregon State University, Corvallis, OR, USA, 2Biostatistics Program, School of Biological and Population Health Sciences, Oregon State University, Corvallis, OR, USA, 3Department of Neuroscience, McKnight Brain Institute, University of Florida, Gainesville, FL, USA, 4Center for Healthy Aging Research, Oregon State University, Corvallis, OR, USA*

#### *Edited by:*

*Erica L. Scheller, Washington University in St. Louis, USA*

#### *Reviewed by:*

*Robin Mark Howard Rumney, University of Sheffield, UK Mark Hamrick, Georgia Health Sciences University, USA*

*\*Correspondence: Urszula T. Iwaniec urszula.iwaniec@oregonstate.edu*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 19 May 2016 Accepted: 02 August 2016 Published: 16 August 2016*

#### *Citation:*

*Lindenmaier LB, Philbrick KA, Branscum AJ, Kalra SP, Turner RT and Iwaniec UT (2016) Hypothalamic Leptin Gene Therapy Reduces Bone Marrow Adiposity in ob/ob Mice Fed Regular and High-Fat Diets. Front. Endocrinol. 7:110. doi: 10.3389/fendo.2016.00110*

Low bone mass is often associated with elevated bone marrow adiposity. Since osteoblasts and adipocytes are derived from the same mesenchymal stem cell (MSC) progenitor, adipocyte formation may increase at the expense of osteoblast formation. Leptin is an adipocyte-derived hormone known to regulate energy and bone metabolism. Leptin deficiency and high-fat diet-induced obesity are associated with increased marrow adipose tissue (MAT) and reduced bone formation. Short-duration studies suggest that leptin treatment reduces MAT and increases bone formation in leptin-deficient *ob/ob* mice fed a regular diet. Here, we determined the long-duration impact of increased hypothalamic leptin on marrow adipocytes and osteoblasts in *ob/ob* mice following recombinant adeno-associated virus (rAAV) gene therapy. Eight- to 10-week-old male *ob/ob* mice were randomized into four groups: (1) untreated, (2) rAAV-Lep, (3) rAAV-green fluorescent protein (rAAV-GFP), or (4) pair-fed to rAAV-Lep. For vector administration, mice were injected intracerebroventricularly with either rAAV-leptin gene therapy (rAAV-Lep) or rAAV-GFP (9 × 107 particles) and maintained for 30 weeks. In a second study, the impact of increased hypothalamic leptin levels on MAT was determined in mice fed highfat diets; *ob/ob* mice were randomized into two groups and treated with either rAAV-Lep or rAAV-GFP. At 7 weeks post-vector administration, half the mice in each group were switched to a high-fat diet for 8 weeks. Wild-type (WT) controls included age-matched mice fed regular or high-fat diet. High-fat diet resulted in a threefold increase in MAT in WT mice, whereas MAT was increased by leptin deficiency up to 50-fold. Hypothalamic leptin gene therapy increased osteoblast perimeter and osteoclast perimeter with minor change in cancellous bone architecture. The gene therapy decreased MAT levels in *ob/ ob* mice fed regular or high-fat diet to values similar to WT mice fed regular diet. These findings suggest that leptin plays an important role in regulating the differentiation of MSCs to adipocytes and osteoblasts, a process that may be dysregulated by high-fat diet. However, the results also illustrate that reducing MAT by increasing leptin levels does not necessarily result in increased bone mass.

Keywords: adipocyte, osteoblast, osteoclast, stem cell, osteoporosis, histomorphometry, rAAV

## INTRODUCTION

Adipose tissue in bone marrow may contribute to metabolic health through its effects on local energy balance and/or its actions as an endocrine organ (1). Low bone mass is often associated with elevated bone marrow adiposity (2, 3). Bone marrow contains mesenchymal stem cells (MSCs) capable of differentiating into cells of the osteoblastic and adipocytic lineages (4). A reciprocal relationship between the number of mature osteoblasts and bone marrow adipocytes may occur as a consequence of differentiation of MSCs toward one lineage at the expense of the other lineage (5, 6). Mechanistically, peroxisome proliferator-activated receptor gamma (PPARγ), a transcription factor that plays a key role in regulation of adipogenesis and lipid uptake, has been implicated as a factor controlling MSC differentiation. Inhibition of PPARγ increases osteoblast differentiation while decreasing adipocyte differentiation. Osteoblast-targeted overexpression of PPARγ inhibits bone mass gain in male mice and increases ovariectomy-induced osteopenia in female mice (5). Furthermore, cultured adipocytes release factors capable of inducing osteoblast linage cells to differentiate into an adipocyte-like phenotype (7). Thus, excessive marrow adipose tissue (MAT) has the potential to directly and/or indirectly reduce bone mass by inhibiting osteoblast differentiation. As such, an increase in MAT may contribute to osteoporosis in conditions, such as menopause, skeletal disuse, alcohol abuse, and eating disorders (3, 8, 9).

Interestingly, anorexia and obesity both result in increased MAT (10, 11). Anorexia is commonly associated with reduced bone mineral density (BMD) and increased fracture risk. By contrast, being overweight is generally associated with increased BMD and reduced fracture risk. However, recent studies suggest that morbid obesity has negative effects on bone quality (12, 13). Thus, it is possible that the increase in MAT plays a role in the detrimental skeletal changes associated with both extremes in body weight.

Leptin, an adipokine produced in proportion to fat mass, plays a critical role in central nervous system-mediated regulation of energy homeostasis (14). Additionally, leptin plays an important positive role in skeletal growth and maturation (15, 16). Indeed, leptin signaling deficiency results in abnormal growth plate development (17) and mild osteopetrosis (18).

Leptin suppresses PPARγ expression and increases lipolysis in adipose tissue in rodents (19). In parallel, leptin increases longitudinal bone growth, osteoblastogenesis and bone formation (15, 16). Thus, the low leptin levels resulting from insufficient adipose tissue observed in anorexia may contribute to increased MAT and decreased bone formation. In contrast to anorexia, obesity typically results in increased leptin levels which, in turn, should suppress MAT. Many authors have concluded that leptin resistance plays an important role in hyperphagia and weight gain associated with obesity (20). Thus, leptin resistance may counteract the expected response to elevated leptin levels and thereby contribute to an increase in MAT during obesity. However, some new studies question whether resistance to endogenous leptin contributes to development of diet-induced obesity in mice (21, 22). If leptin resistance is a major contributor to the etiology of obesity, it may be overcome in normal rodents, at least in part, by increasing hypothalamic leptin levels. Whatever its precise role in diet-induced obesity, it is clear that leptin resistance resulting from loss of function of the leptin receptor (*db/db* mice), in addition to inducing morbid obesity, results in profound negative effects on the skeleton. Furthermore, some of these negative skeletal effects (e.g., reduced bone formation) are recapitulated, without impacting energy metabolism, following adoptive transfer of bone marrow from *db/db* mice into wild-type (WT) mice (16).

Leptin-deficient *ob/ob* and leptin receptor-deficient *db/db* mice exhibit excessive MAT in long bones (16). Short-term delivery of leptin into the hypothalamus was shown to reduce peripheral fat depots as well as MAT (23, 24). However, the longterm effects of increased leptin levels on MAT have not been well characterized. Hypothalamic leptin gene therapy has been shown to result in life-long reductions in body weight in *ob/ob* mice (25). The goal of the present study was to determine the long-duration effects of increased hypothalamic leptin, using recombinant adeno-associated virus leptin gene therapy (rAAV-Lep), on bone marrow adiposity in morbidly obese *ob/ob* mice. Given that the energy density of a diet impacts weight gain and MAT levels (26), the effects of a regular and high energy density (high fat) diet were also evaluated.

## MATERIALS AND METHODS

## Experimental Animals

Eight- to 10-week-old male WT C57BL/6J (B6) and leptindeficient *ob/ob* mice on the same genetic background were obtained from Jackson Laboratory (Bar Harbor, ME, USA). This age corresponds to peak cancellous bone volume fraction in the femur metaphysis (27). The mice were maintained in accordance with the NIH Guide for the Care and Use of Laboratory Animals and the experimental protocols were approved by the Institutional Animal Care and Use Committee at the University of Florida. The mice were housed individually in a temperature (21–23°C) and light-controlled room (lights on 6:00 a.m. to 6:00 p.m.) under specific pathogen-free conditions.

## Experiment 1: Effects of 30 Weeks of Hypothalamic Leptin Gene Therapy on Marrow Adiposity and Cancellous Bone Histomorphometry in *ob/ob* Mice

Following arrival, *ob/ob* mice were randomized by weight into four treatment groups: (1) untreated (*n* = 6), (2) control vector encoding green fluorescent protein (rAAV-GFP, *n* = 7), (3) rAAV-Lep (*n* = 8), or (4) pair-fed to rAAV-Lep (*n* = 6). The mice were maintained on standard mouse chow (LM-485, Teklad, Madison, WI, USA) and sacrificed 30 weeks following vector administration at 38–40 weeks of age. This age corresponds to a period immediately prior to a drastic increase in mortality in *ob/ob* mice – median lifespan in *ob/ob* mice is 55 weeks compared to 131 weeks in WT mice (25). The effects of treatment on hypothalamic leptin gene expression, body weight, food intake, hormone levels, organ weights, and cancellous and cortical bone architecture evaluated by microcomputed tomography in this study are detailed elsewhere (25, 28).

## Experiment 2: Effects of 15 Weeks of Hypothalamic Leptin Gene Therapy and 8 Weeks of High-Fat Diet on Marrow Adiposity and Cancellous Bone Histomorphometry in *ob/ob* Mice

Experiment 2 was conducted using WT and *ob/ob* mice. Following arrival, WT mice (*n*= 12) were maintained on regular chow (LM-485, Teklad, Madison, WI, USA; caloric density 3.4 kcal/g, 11% of kcal from fat) until 15–17 weeks of age and then randomized by weight into two groups: (1) control (*n* = 3) or (2) high-fat diet (*n* = 9). Mice in the control group continued to consume the regular diet *ad libitum* while mice in the high-fat group were placed on a high-fat diet (caloric density 4.7 kcal/g; 45% of kcal from fat, primarily from lard; Research Diets, New Brunswick, NJ, USA) fed *ad libitum*. The mice were sacrificed 8 weeks later at 23–25 weeks of age – an age corresponding to cessation of linear growth in B6 mice (27).

In conjunction, *ob/ob* mice were randomized by weight into two treatment groups: rAAV-Lep (*n*= 16) or control vector rAAV-GFP (*n* = 14). At 7 weeks post-vector administration, rAAV-GFP and rAAV-Lep mice were each divided into two groups: one group continued to consume regular diet and the other was switched to a high-fat diet as described above for WT mice. The mice were sacrificed 8 weeks later at 23–25 weeks of age (15 weeks following vector administration). The effect of the rAAV-Lep pretreatment and high-fat diet on hypothalamic leptin gene expression, body weight, food intake, organ weights, hormone levels, and cancellous and cortical bone architecture determined by microcomputed tomography are detailed elsewhere (28–30).

## Construction and Packaging of rAAV Vectors

rAAV-leptin gene therapy and rAAV-GFP vectors were constructed and packaged as previously described (31). In brief, the vector pTR-CBA-Ob *Eco*RI fragment of pCR-rOb containing rat leptin cDNA was subcloned into rAAV vector plasmid pAAVβGEnh after deleting the *Eco*RI fragment carrying the β-glucoronidase cDNA sequence. The control vector, rAAV-GFP, was similarly constructed to encode the GFP gene.

## Vector Administration

For vector administration, the mice were anesthetized with sodium pentobarbital (60 mg/kg, i.p.), placed on a Kopf stereotaxic apparatus with mouse adapter for intracerebroventricular injection, and injected intracerebroventricularly with either rAAV-Lep (9 × 107 particles in 1.5 μl) or rAAV-GFP (9 × 107 particles in 1.5 μl). The coordinates employed for microinjector placement in the third cerebroventricle were 0.3 mm posterior to bregma, 0.0 lateral to midline, and 4.2 mm below the dura (29).

## Tissue Collection and Analyses

At the end of each experiment mice were anesthetized with sodium pentobarbital (60 mg/kg; i.p.) and euthanized by exsanguination. Femora were excised, cleaned of soft tissue, and stored in 70% ethanol. Femora were prepared for histomorphometric evaluation as described (32). In brief, distal femora were dehydrated in graded increases of ethanol and xylene and embedded undecalcified in methyl methacrylate. Frontal sections (4 μm thick) were cut with a vertical bed microtome (Leica 2165) and affixed to slides precoated with a 1% gelatin solution. One section/ animal was stained for tartrate-resistant acid phosphatase and counterstained with toluidine blue (Sigma, St Louis, MO, USA) for assessment of bone and cell-based measurements.

Histomorphometric data were collected using the OsteoMeasure System (OsteoMetrics, Inc., Atlanta, GA, USA). The sampling site for the distal femoral metaphysis was located 0.25–1.25 mm proximal to the growth plate and 0.1 mm from cortical bone. Cancellous bone measurements included bone area fraction (bone area/tissue area, %) and the derived architectural indices of trabecular number (mm<sup>−</sup><sup>1</sup> ), trabecular thickness (micrometer), and trabecular separation (micrometer). Measurements of MAT included overall marrow adiposity (adipose area/tissue area, %), adipocyte density (mm<sup>−</sup><sup>2</sup> ), and adipocyte size (micrometers2 ). Adipocytes were identified as large circular or oval-shaped cells bordered by a prominent cell membrane and lacking cytoplasmic staining due to alcohol extraction of intracellular lipids during processing. This method has been validated by fat extraction and analysis (33). Osteoblast and osteoclast perimeters were also measured and expressed as % of total bone perimeter. Osteoblasts were identified as plump cuboidal cells immediately adjacent to a thin layer of osteoid in direct contact with the bone perimeter. Osteoclasts were identified as multinucleated (two or more nuclei) cells with acid phosphatase positive (red-stained) cytoplasm in contact with the bone perimeter. Data are reported using standard two-dimensional nomenclature (34).

## Statistical Analysis

Mean responses for Experiment 1 were compared among the untreated, rAAV-GFP, rAAV-Lep, and pair-fed groups using one-way analysis of variance. For Experiment 2, the effects of treatment and diet were assessed using two-way analysis of variance. Pairwise comparisons were made using *t*-tests or the Wilcoxon–Mann–Whitney test. The required conditions for valid use of linear models were assessed using Levene's test for homogeneity of variance, plots of residuals versus fitted values, normal quantile plots, and the Anderson–Darling test of normality. The Benjamini and Hochberg method (35) for maintaining the false discovery rate at 5% was used to adjust for multiple comparisons. Data analysis was performed using R version 3.3.2.

## RESULTS

## Experiment 1: Effects of 30 Weeks of Hypothalamic Leptin Gene Therapy on Marrow Adiposity and Cancellous Bone Histomorphometry in *ob/ob* Mice

The effects of treatment on body weight and on cancellous bone in the distal femur metaphysis are shown in **Table 1**. rAAV-Lep treatment resulted in lower body weight compared to untreated,


TABLE 1 | Effects of hypothalamic leptin gene therapy (rAAV-Lep) on body weight and cancellous bone architecture in distal femur metaphysis in *ob/ob* male mice at 30 weeks post-vector administration.

*Data are mean* ± *SE.*

*\*Previously reported (28).*

*a Different from untreated, P* < *0.05.*

*bDifferent from rAAV-GFP, P* < *0.05.*

*c Different from pair-fed, P* < *0.05.*

rAAV-GFP-treated, and pair-fed mice. Significant differences in cancellous bone area fraction, trabecular number, or trabecular spacing were not detected with treatment. However, trabecular thickness was lower in rAAV-Lep-treated mice compared to untreated and rAAV-GFP-treated mice, but did not differ from pair-fed *ob/ob* mice.

The effects of treatment on MAT in the distal femur metaphysis are shown in **Figure 1**. Marrow adiposity (adipose area/tissue area) (**Figure 1A**), adipocyte density (**Figure 1B**), and adipocyte size (**Figure 1C**) were lower in rAAV-Lep-treated mice compared to untreated, rAAV-GFP-treated, and pair-fed mice. Significant differences among untreated, rAAV-GFP-treated, and pair-fed mice were not detected for any of the MAT measurements. The effects of rAAV-Lep treatment on marrow adiposity can be readily appreciated in **Figures 1D–G**.

The effects of treatment on osteoblast perimeter and osteoclast perimeter in the distal femur metaphysis are shown in **Figure 2**. Osteoblast perimeter (**Figure 2A**) and osteoclast perimeter (**Figure 2B**) were higher in rAAV-Lep-treated mice compared to untreated, rAAV-GFP-treated, and pair-fed mice. Significant differences among untreated, rAAV-GFP-treated, and pair-fed mice were not detected for either of the cellular endpoints evaluated. The effects of rAAV-Lep treatment on osteoblast and osteoclast perimeter can be appreciated in **Figures 2C–F**.

## Experiment 2: Effects of 15 Weeks of Hypothalamic Leptin Gene Therapy and 8 Weeks of High-Fat Diet on Marrow Adiposity and Cancellous Bone Histomorphometry in *ob/ob* Mice

The effects of rAAV-Lep pretreatment and high-fat diet on body weight and on cancellous bone in distal femur metaphysis are shown in **Table 2**. Body weight was higher in WT mice fed highfat diet compared to WT mice fed regular diet. Body weight was also higher in *ob/ob* mice fed high fat compared to *ob/ob* mice fed regular diet and rAAV-Lep treatment resulted in lower body weight. Cancellous bone area fraction, trabecular thickness, and trabecular number were higher and trabecular spacing was lower in WT mice fed high-fat diet compared to WT mice fed regular diet. rAAV-Lep treatment in *ob/ob* mice resulted in lower cancellous bone area fraction and trabecular thickness. Significant differences in trabecular number or trabecular spacing were not detected with treatment in the *ob/ob* mice. With the exception of trabecular thickness, which was lower, significant differences between WT mice and rAAV-Lep-treated *ob/ob* mice fed regular diets were not detected for any of the remaining cancellous endpoints evaluated.

The effects of treatment on marrow adiposity in distal femur metaphysis are shown in **Figure 3**. Marrow adiposity and adipocyte size were higher in WT mice fed high-fat diet compared to mice fed regular diet. Significant differences in adipocyte density were not detected with diet in WT mice. rAAV-Lep treatment in *ob/ob* mice resulted in lower marrow adiposity due to lower adipocyte density as well as lower adipocyte size. Significant differences in marrow adiposity, adipocyte density, or adipocyte size were not detected between WT and rAAV-Lep-treated *ob/ob* mice fed regular diet.

## DISCUSSION

Leptin-deficient *ob/ob* mice were heavier and had increased MAT in distal femur metaphysis compared to WT mice. Consumption of high-fat diet resulted in increased body weight in both WT mice and *ob/ob* mice but increased MAT and cancellous bone in WT mice only. rAAV-Lep treatment decreased MAT in *ob/ ob* mice. The reduction in MAT in rAAV-Lep-treated *ob/ob* mice was accompanied by increases in osteoblast-lined and osteoclastlined bone perimeter but not by an increase in cancellous bone.

Continuous and once daily intracerebroventricular administration of leptin were similarly effective in reducing MAT in long bones of *ob/ob* mice (23, 24, 36, 37). Based on lower adipocyte number and size and increased concentration of apoptosis marker caspase-3 in bone marrow adipocytes, the reduction in MAT was likely due to a combination of reduced adipocyte differentiation, increased fat oxidation, and increased adipocyte apoptosis. A similar reduction in MAT was observed following subcutaneous leptin administration (36).

In normal female rats, hypothalamic delivery of rAAV-Lep was shown to maintain lower body weight, WAT weight, and serum leptin levels (2.7 ± 0.3 versus 1.0 ± 0.1 ng/ml) for at least 18 weeks following vector administration. By contrast, rAAV-Lep transiently reduced MAT; MAT levels were reduced at 5 weeks but returned to normal levels by 10 weeks following vector

administration (38). Also, hypothalamic rAAV-Lep gene therapy was ineffective in lowering MAT in ovariectomized rats (39). In the present study, MAT levels in rAAV-Lep-treated *ob/ob* mice, evaluated 15 and 30 weeks following vector administration, were much lower than age-matched *ob/ob* controls and comparable to WT mice in Experiment 2. These findings suggest that (1) while very important, the physiological actions of leptin on MAT are primarily manifested at low hormone levels and (2) hyperleptinemia has little further effect on MAT. If correct, this could help explain why some studies fail to detect a relationship between blood leptin levels and MAT (1). The findings regarding the actions of leptin on MAT are remarkably similar to the actions of the hormone on bone growth, maturation, and turnover. Whereas hypothalamic rAAV-Lep gene therapy corrected the skeletal abnormalities in *ob/ob* mice, it had minimal long-term impact on bone in rodents capable of producing leptin (16, 18, 28, 39).

It was initially hypothesized that the complex skeletal phenotype of *ob/ob* mice was due to opposing actions of peripheral and central leptin on bone formation (40, 41). However, subcutaneous and intracerebroventricular delivery of leptin were found to similarly increase bone formation in *ob/ob* mice (36). Additionally, longduration hypothalamic leptin gene therapy was shown to normalize bone microarchitecture in *ob/ob* mice; specifically, increasing hypothalamic leptin levels resulted in increased femur length and total femur bone volume but decreased cancellous bone volume fraction in lumbar vertebra (28). These latter findings imply that, in addition to increasing longitudinal bone growth, delivery of leptin into the hypothalamus results in increased bone formation as well as increased bone resorption. Thus, an imbalance between bone formation and resorption related to local environment (e.g., precursor cell populations, mechanical loads, paracrine factors, etc.) potentially explains the contrasting phenotypes that have been identified in bones of the limb and spine in *ob/ob* mice (41).

Bone- and bone compartment-specific changes in microarchitecture in response to hormonal regulators of bone metabolism and mechanical loading environment are not unique to leptin. For example, by regulating longitudinal and radial bone growth and bone turnover balance, estrogens, and androgens contribute to sexual dimorphism of the skeleton. In this regard, administration of estrogen to growing ovariectomized rats results in shorter bones with lower total bone mass but higher site-specific cancellous bone volume (42). Ovariectomized rats also experience increased MAT expansion (39). *ob/ob* mice of both genders are hypogonadal due to reduced GnRH secretion (43), a defect that is reversed following leptin treatment (25, 29). Thus, it is possible that hypogonadism contributes to MAT expansion in *ob/ob* mice. Expansion of MAT during caloric restriction in WT mice was associated with increased circulation of glucocorticoids, while caloric restriction resulted in a further increase in the already high levels of MAT in leptin-deficient *ob/ob* mice (1, 18). These findings provide evidence that multiple factors, including leptin, regulate MAT levels.

As previously mentioned, short-duration delivery of leptin into the hypothalamus increased bone formation (36). Similarly, hypothalamic leptin gene therapy increased serum osteocalcin levels and osteoblast perimeter in lumbar vertebra of *ob/ob* mice (16, 44). In the present study, hypothalamic leptin gene therapy increased osteoblast perimeter in distal femur metaphysis in *ob/ ob* mice 30 weeks following vector administration. These findings indicate that leptin promotes higher levels of bone formation prior to and following restoration of normal body weight and bone mass in *ob/ob* mice (28).

The increased cancellous bone volume fraction observed at selected skeletal sites (lumbar vertebrae) in *ob/ob* mice was initially attributed to increased bone formation, suggesting that leptin was antiosteogenic (45). However, subsequent studies consistently reported decreased bone formation in *ob/ob* mice and leptin receptor-deficient *db/db* mice, and increased bone formation following intracerebroventricular delivery of leptin, leptin gene therapy, or subcutaneous administration of leptin in *ob/ob* mice (16, 37, 44). Leptin signaling-deficient (*ob/ob* and *db/db*) mice have normal or increased osteoclast number but exhibit evidence for impaired osteoclast function (16, 18). As a consequence, these mice exhibit impaired skeletal maturation due to defective resorption of calcified cartilage. Specifically, the high cancellous bone volume fraction represents mild osteopetrosis. In the present study in *ob/ob* mice, rAAV-Lep resulted in increased osteoclast-lined bone perimeter. Thus, the failure to detect an increase in cancellous bone volume fraction in the femur metaphysis in response to higher leptin levels is likely due to parallel increases in bone formation and bone resorption.

High MAT levels in *ob/ob* mice are associated with low cancellous bone turnover (16). In the present study, rAAV-Lep resulted in increases in osteoclast-lined perimeter as well as osteoblastlined bone perimeter and greatly reduced MAT with minimal change in cancellous bone area fraction. High MAT levels are not unique to leptin deficiency. Growth hormone deficiency in rats and mice induced by hypophysectomy or deletion of the gene for growth hormone, respectively, is also associated with high MAT and low bone turnover. In the case of growth hormone deficiency, parathyroid hormone was found to increase bone formation in hypophysectomized rats without impacting MAT levels, demonstrating that bone formation induced by bone anabolic agents is not suppressed by high levels of MAT (33). Similarly, although bone formation was increased, the absence of MAT in *kitW/W-v* mice did not protect against ovariectomy-induced bone loss (46). Taken together, these findings suggest that bone resorption as well as bone formation can be impacted during changes in MAT levels and interventions that target MAT may not necessarily change bone turnover balance.

TABLE 2 | Effects of high-fat diet and rAAV-Lep pretreatment and high fat diet on terminal body weight and cancellous bone architecture in distal femur metaphysis in male WT and *ob/ob* mice, respectively.


*Data are mean* ± *SE.*

*\*Previously reported (30).*

*a Different from WT mice fed regular diet.*

A positive association between body weight and bone mass was observed in *ob/ob* as well as WT mice (30, 47). However, leptin appears to sensitize the skeleton to bone mechanical loading. This may explain why leptin-deficient mice have a low total bone mass even though they are morbidly obese and why the massive weight loss in *ob/ob* mice following leptin treatment is actually associated with a net increase in bone mass (28).

Conditional knockout of the leptin receptor in bone marrow stromal cells has been reported to result in local increases in osteogenesis and decreased adipogenesis (48). It is difficult to reconcile these findings with the skeletal phenotype of leptin receptor-deficient mice (16, 49) or with the present results demonstrating that increasing hypothalamic leptin levels profoundly reduces MAT levels. Evidence that factors secondary to leptin deficiency are responsible for the discrepancy, as suggested by Yue et al., are presently lacking and correction of many of the metabolic abnormalities associated with leptin deficiency (e.g., hyperglycemia), rather than improving, actually worsens the skeletal phenotype of *ob/ob* mice (18). Also inexplicable by Yue et al. is the finding that adoptive transfer of leptin receptor-deficient *db/db* bone marrow cells into WT mice recapitulates the low bone formation skeletal phenotype of *db/db* mice without impacting food intake or weight gain (16). It is possible that the role of leptin receptors in regulating bone metabolism depends upon stage of stromal cell differentiation (50), but this requires additional research.

A limitation of the present study is that MAT measurements were performed at a single skeletal site and MAT subtypes were not evaluated. The lipid composition and physiological function of MAT can vary with location and/or regulatory factors, such as growth hormone status (33, 51). A further limitation of most studies, including the present study, is that they have been performed housing mice at room temperature. Thermoneutral (temperature range where basal rate of energy production is at equilibrium with heat loss) in mice ranges from 26 to 34°C (52). Mild cold stress induced by room temperature housing results in dramatic cancellous bone loss at the femur metaphysis. Interestingly, mice housed at 32°C consumed ~40% less food (fed *ad libitum*) but did not differ from room temperature-housed mice in weight (53). In addition to higher bone mass, mice housed at 32°C had greatly reduced UCP-1 gene expression in brown fat, higher serum leptin, higher MAT levels due to increased adipocyte number, higher bone formation rate to due higher osteoblast perimeter, and lower osteoclast perimeter. These findings suggest that non-shivering thermogenesis substantially influences the association between leptin, MAT, and bone cells.

In summary, hypothalamic leptin gene therapy maintained low MAT levels in *ob/ob* mice fed regular or high-fat diet. The reduction in MAT was accompanied by an increase in osteoblast-lined bone perimeter but not an increase in cancellous bone volume fraction. The increase in osteoclast-lined bone perimeter suggests that the increase in bone formation was matched by an increase in bone resorption. These findings provide further evidence that a deterministic model where reducing MAT will invariably lead to increased bone volume is not tenable. As a consequence, interventions targeted at reducing MAT may not be an effective strategy for increasing bone mass.

## REFERENCES


## AUTHOR CONTRIBUTIONS

Study design: SK. Data acquisition: LL, KP, and RT. Data analysis: AB. Data interpretation: LL, KP, AB, SK, RT, and UI. Drafting of manuscript: LL, RT, and UI. Revising manuscript content: LL, KP, AB, SK, RT, and UI. Approving final version of manuscript: LL, KP, AB, SK, RT, and UI. All authors take responsibility for the integrity of this work.

## ACKNOWLEDGMENTS

We thank Dr. S. Boghossian for executing the studies.

## FUNDING

This study was supported by NIH AR060913, NASA NNX12AL24, and USDA 38420-17804.


**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 © 2016 Lindenmaier, Philbrick, Branscum, Kalra, Turner and Iwaniec. 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.*

# Marrow adipose Tissue expansion coincides with insulin resistance in MagP1-Deficient Mice

*Tezin A. Walji1†, Sarah E. Turecamo1†, Alejandro Coca Sanchez2 , Bryan A. Anthony3 , Grazia Abou-Ezzi3 , Erica L. Scheller4 , Daniel C. Link3 , Robert P. Mecham1 and Clarissa S. Craft1,4\**

*1Department of Cell Biology and Physiology, Washington University School of Medicine, St. Louis, MO, USA, 2Department of Medicine and Medical Specialties, Faculty of Medicine and Health Sciences, University of Alcala de Henares, Madrid, Spain, 3Department of Medicine, Oncology Division, Washington University School of Medicine, St. Louis, MO, USA, 4Department of Medicine, Bone and Mineral Diseases Division, Washington University School of Medicine, St. Louis, MO, USA*

#### *Edited by:*

*Ann Schwartz, University of California San Francisco, USA*

#### *Reviewed by:*

*Katherine A. Staines, The Royal Veterinary College, UK Beata Lecka-Czernik, University of Toledo, USA*

#### *\*Correspondence:*

*Clarissa S. Craft clarissa.craft@wustl.edu*

*† Tezin A. Walji and Sarah E. Turecamo contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 05 May 2016 Accepted: 22 June 2016 Published: 30 June 2016*

#### *Citation:*

*Walji TA, Turecamo SE, Sanchez AC, Anthony BA, Abou-Ezzi G, Scheller EL, Link DC, Mecham RP and Craft CS (2016) Marrow Adipose Tissue Expansion Coincides with Insulin Resistance in MAGP1- Deficient Mice. Front. Endocrinol. 7:87. doi: 10.3389/fendo.2016.00087*

Marrow adipose tissue (MAT) is an endocrine organ with the potential to influence skeletal remodeling and hematopoiesis. Pathologic MAT expansion has been studied in the context of severe metabolic challenge, including caloric restriction, high fat diet feeding, and leptin deficiency. However, the rapid change in peripheral fat and glucose metabolism associated with these models impedes our ability to examine which metabolic parameters precede or coincide with MAT expansion. Microfibril-associated glycoprotein-1 (MAGP1) is a matricellular protein that influences cellular processes by tethering signaling molecules to extracellular matrix structures. MAGP1-deficient (*Mfap2*−/−) mice display a progressive excess adiposity phenotype, which precedes insulin resistance and occurs without changes in caloric intake or ambulation. *Mfap2*−/− mice were, therefore, used as a model to associate parameters of metabolic disease, bone remodeling, and hematopoiesis with MAT expansion. Marrow adiposity was normal in *Mfap2*−/− mice until 6 months of age; however, by 10 months, marrow fat volume had increased fivefold relative to wild-type control at the same age. Increased gonadal fat pad mass and hyperglycemia were detectable in *Mfap2*−/− mice by 2 months, but peaked by 6 months. The development of insulin resistance coincided with MAT expansion. Longitudinal characterization of bone mass demonstrated a disconnection in MAT volume and bone volume. Specifically, *Mfap2*−/− mice had reduced trabecular bone volume by 2 months, but this phenotype did not progress with age or MAT expansion. Interestingly, MAT expansion in the 10-month-old *Mfap2*−/− mice was associated with modest alterations in basal hematopoiesis, including a shift from granulopoiesis to B lymphopoiesis. Together, these findings indicate MAT expansion is coincident with insulin resistance, but not excess peripheral adiposity or hyperglycemia in *Mfap2*−/− mice; and substantial MAT accumulation does not necessitate a proportional decrease in either bone mass or bone marrow cellularity.

Keywords: marrow adipose tissue, obesity, insulin resistance, bone remodeling, hematopoiesis, microfibrilassociated glycoprotein-1

**Abbreviations:** ECM, extracellular matrix; MAT, marrow adipose tissue; MAGP1, microfibril-associated glycoprotein-1.

## INTRODUCTION

Adipose tissue exists in multiple variations, the most extensively studied being: brown, white, and beige. Brown adipocytes specialize in utilizing energy to produce heat (thermogenesis), white adipocytes are energy-storing reservoirs, and beige adipocytes principally store lipids but can be stimulated to transdifferentiate into a "brown-like" state (1). Less studied is a fourth lipid storing depot, located within the skeleton, marrow adipose tissue (MAT). These skeleton-associated adipocytes arise from a unique progenitor (2) and occupy approximately 70% of the marrow space within adult human bones (3). There is an evidence that MAT adipocytes exist in two forms: regulated marrow adipocytes (rMATs) that respond in size and number to physiological challenge, and the constitutive marrow adipocytes (cMATs) that persist despite challenge (4). Marrow adipocytes are not inert space-filling cells. Rather, the ability to produce leptin and adiponectin classifies them as an endocrine organ (5, 6). These cells also release fatty acids *via* lipolysis, suggesting that they may fuel local cellular processes (7).

Pathologic MAT expansion, as a consequence of metabolic dysfunction, has been studied in several contexts, including caloric restriction, high fat diet feeding, and leptin deficiency (6, 8–10). Unfortunately, the rapid change in peripheral fat and glucose metabolism associated with these models impedes our ability to determine which metabolic parameters precede or coincide with MAT expansion. The objective of this study was to correlate parameters of metabolic disease, bone remodeling, and hematopoiesis with MAT volume. Thus, we chose a model of progressive metabolic dysfunction – mice lacking the extracellular matrix (ECM) protein microfibril-associated glycoprotein-1 (MAGP1).

The ECM is a milieu of structural proteins, proteoglycans, and adhesive glycoproteins (11). MAGPs are a non-structural component of the ECM. The biology of this family has been recently reviewed (12). Briefly, MAGP1 (*Mfap2*) is broadly expressed, including both adipose tissue and bone (13, 14), and typically associates with fibrillin-rich microfibrils. MAGP1 may also interact with collagen-VI fibers (15). Characterization of MAGP1-deficient mice (*Mfap2*−/−) suggests that MAGP1 does not contribute to the mechanical integrity of its associated ECM fibers (16). Rather, MAGP1 functions as a gatekeeper of signal transduction. A significant body of evidence has demonstrated that MAGP1 interacts with ligands of the transforming growth factor beta (TGFβ) superfamily; sequestering them in the ECM (13, 14, 16, 17). In mice, MAGP1 deficiency (*Mfap2*<sup>−</sup>/<sup>−</sup>) results in adipocyte hypertrophy in peripheral white adipose tissue (WAT), diabetes, and reduced bone mass (13, 14). These phenotypes are progressive, developing with age.

Using the *Mfap2*−/− model, we demonstrate that MAT expansion occurs concurrent with insulin resistance, not excessive peripheral adiposity or hyperglycemia. Bone loss in the MAGP1 deficient mice occurs without corresponding increases in MAT. Furthermore, a fivefold increase in MAT in the proximal tibia does not negatively affect bone marrow (BM) cellularity.

## MATERIALS AND METHODS

## Animals and Diet

All animals were C57BL/6 background males, housed in a pathogen-free animal facility and fed standard chow *ad libitum*. MAGP1-deficient mice were generated using C57BL/6-derived ES cells [*Mfap2*tm1a(KOMP)Wtsi] purchased from KOMP Repository (Davis, CA, USA). ES cells were injected into blastocyst from C57BL/6 donors and transferred into pseudopregnant C57BL/6 females. Offspring were maintained on the Jackson Laboratories C57BL/6J strain. These mice are a newly derived line and are therefore different from the *Mfap2*−/− mice used in previous studies (13, 14, 16, 17). All animals were treated in accordance with animal protocols approved by the Animal Studies Committee at Washington University.

## Tissue Collection

Tissues were harvested at 2, 6, and 10 months. Mouse fur was sprayed with 70% EtOH, then gonadal white adipose tissue (gWAT), and tibias were collected. Tissues were cleaned thoroughly of all contaminants (hair, connective tissue, etc.), then frozen or stored in 10% neutral buffered formalin.

## Insulin Tolerance Test

For insulin tolerance tests (ITTs), mice were fasted for 6 h and then given an injection of 0.75 U/kg Humulin-R insulin (Lilly, Indianapolis, IN, USA). Insulin was delivered by intraperitoneal injection, and tail blood glucose concentration was measured using Contour strips and meters (Bayer, Whippany, NJ, USA) at the indicated intervals.

## Osmium Staining and Micro-Computed Tomography

Micro-computed tomography (μCT) was completed on tibias from 2-, 6-, and 10-month-old mice. Bones were fixed in 10% neutral buffered formalin and then embedded in 2% agarose gel. Tibias were scanned at 20 μm voxel resolution using a Scanco μCT 40 (Scanco Medical AG, Zurich, Switzerland) calibrated using a hydroxyapatite phantom. Measurements of both trabecular and cortical bones were made based on reported guidelines (18). For trabecular bone, 50 slices below the growth plate were contoured to exclude the cortical bone, allowing trabecular bone volume/ tissue volume (BV/TV) and bone mineral density (BMD) to be determined. For cortical bone, 20 slices – 2 mm proximal to the tibia–fibula junction were analyzed to determine cortical tissue mineral density (TMD). A threshold of 175 for trabecular bone and 260 for cortical bone (on a 0–1000 scale) was maintained. Note: two 10-month-old *Mfap2*<sup>−</sup>/<sup>−</sup> bones were excluded from study due to suspected fracture.

Tibias were then decalcified in 14% EDTA for 3 weeks. Demineralized bones were incubated in a solution containing 1% osmium and 2.5% potassium dichromate for 48 h at room temperature. After thorough washing (water), tibias were embedded in 1% agarose gel. Osmium stained bones were then scanned as above, using a Scanco μCT 40, but at 10 μm voxel resolution. Analysis was performed in the same region as the trabecular bone (1,000 μm distal to the growth plate, 100 slices), using a threshold of 350 (on a 0–1000 scale). Osmium volume (OV) is quantification of the total OV within the region of interest. OV/ TV is division of the OV by the volume of the marrow cavity within the same region.

## Flow Cytometry

Blood, spleen, and BM (one each of pelvis bone, femur, and tibia) were harvested from Jackson Laboratories c57BL/6J WT and *Mfap2*<sup>−</sup>/<sup>−</sup> mice using standard techniques. Single cell suspensions were made, and red blood cells were lysed. Blood cells were counted using a Hemavet. BM and spleen cells were counted using a Cellometer. Ten million cells were stained and washed in FACS buffer (PBS, 1% BSA, 0.1% sodium aziade, and 2 μm EDTA). For monocyte, macrophage, and neutrophil analyses, cells from each tissue were stained with Brilliant Violet 421-conjugated F4/80 (BM8), peridinin chlorophyll protein complex (PerCP)-Cy5.5-conjugated MHC Class II (M5/114.15.2), phycoerythrin (PE)-Cy7-conjugated CD11c (N418), allophycocyanin (APC)-conjugated CD19 (1D3), APC-conjugated CD45R/B220 (RA3-6B2), APC-eFluor780 conjugated Ly-6G (Gr1, RB6-8C5), and PE-conjugated CD115 (AFS98). Two million live events were collected in the live gate and analyzed *via* FlowJo. Gating strategies can be found in Figure S1 in Supplementary Material.

## Statistical Analysis

Statistical comparisons were performed using Graphpad Prism® (GraphPad Software, Inc., La Jolla, CA, USA). In **Figures 1**–**3**, with the exception of panel 2D, comparisons between wild-type (WT) and *Mfap2*<sup>−</sup>/<sup>−</sup> mice at each time point and over time within groups (WT or *Mfap2*<sup>−</sup>/<sup>−</sup>) were performed using a regular two-way ANOVA with Holm–Sidak correction for multiple comparisons. For the ITTs in panel 2D, repeated measures two-way ANOVA with Sidak's correction was used. In **Figures 4** and **5**, results were compared with a two-tailed *t*-test. Comparisons with a *P*-value ≤0.05 were considered statistically significant.

Figure 2 | Characterization of metabolic disorder in *Mfap2***−**/**−**mice. (A) Gonadal white adipose tissue (WAT) was harvested at 2 months (*n* = 10, 8), 6 months (*n* = 6, 8), and 10 months (*n* = 8, 8). (B) Total body weight at 2 months (*n* = 10, 8), 6 months (*n* = 7, 9), and 10 months (*n* = 8, 8). (C) Fasted (6 h) blood glucose concentration at 2 months (*n* = 10, 8), 6 months (*n* = 6, 8), and 10 months (*n* = 8, 8). (D) Insulin sensitivity results at 2 months (*n* = 10, 8), 6 months (*n* = 5, 8), and 10 months (*n* = 8, 8). Blood glucose data were normalized to time point 0 (before insulin injection). Mean ± SEM; statistical significance in (A–C) at each time point and within groups (WT or *Mfap2*−/−) over time was determined by two-way ANOVA with Holm–Sidak's correction for multiple comparisons. Statistical significance in (D) was determined by repeated-measured two-way ANOVA with Sidak's correction (\**p* ≤ 0.05, *a* = significant relative to 2 months of same genotype, and *b* = significant relative to 6 months of same genotype). Sample sizes shown above are for WT and *Mfap2*−/− mice, respectively.

## RESULTS

## Pathologic MAT Expansion in *Mfap2***−**/**−** Mice

To compare parameters of metabolic disease with MAT expansion, a longitudinal study was performed in MAGP1-deficient (*Mfap2*<sup>−</sup>/<sup>−</sup>) mice that are susceptible to metabolic syndrome. Marrow adiposity was determined by μCT imaging of osmium stained lipid within the proximal tibia (**Figure 1A**). Quantification of total marrow lipid OV and OV normalized to total marrow cavity volume (OV/TV) is shown in **Figures 1B,C**. Marrow adiposity was found to be indistinguishable in either young (2 months) or adult (6 months) *Mfap2*<sup>−</sup>/<sup>−</sup> mice, relative to WT mice. However, aged 10-month-old *Mfap2*<sup>−</sup>/<sup>−</sup> mice displayed a fivefold increase in marrow lipid (OV and OV/TV) relative to WT animals. WT C57BL/6 mice have only a modest, statistically insignificant, increase (54%) in MAT between 6 and 10 months.

## Characterization of Metabolic Dysfunction in *Mfap2***−**/**−** Mice

Previously, we demonstrated that MAGP1 deficiency results in excess adiposity and reduced bone mass (13, 14, 16). Consistent with previous results, 2-month-old *Mfap2*<sup>−</sup>/<sup>−</sup> mice had only a modest 15% increase in white fat (as measured by gonadal fat pad mass, **Figure 2A**). However, by 6 months, *Mfap2*<sup>−</sup>/<sup>−</sup> mice had a 2.7-fold increase in fat mass, relative to WT mice. Extending our study to 10 months of age, we found that the differential in fat mass stabilizes, indicated by a 1.8-fold increase *Mfap2*<sup>−</sup>/<sup>−</sup> gonadal fat pad mass relative to WT mice. Total body weight reinforced this finding (**Figure 2B**). *Mfap2*<sup>−</sup>/<sup>−</sup> body weight was insignificantly increased at 2 months, but a significant (17%) increase in *Mfap2*<sup>−</sup>/<sup>−</sup> weight was found at 6 months. Similar to the fat mass measurement, the weight differential between WT and *Mfap2*<sup>−</sup>/<sup>−</sup> mice stabilized, remaining increased by 17% relative to baseline, at 10 months. Six-hour fasting blood glucose was significantly elevated in *Mfap2*<sup>−</sup>/<sup>−</sup> mice by 2 months (12%, **Figure 2C**). By 6 months, fasted blood glucose was 63% higher in *Mfap2*<sup>−</sup>/<sup>−</sup> mice relative to WT mice. However, by 10 months, there was only a 35% increase in *Mfap2*<sup>−</sup>/<sup>−</sup> blood glucose relative to WT. To characterize metabolic function, ITTs were performed on WT and *Mfap2*<sup>−</sup>/<sup>−</sup> mice at all ages (**Figure 2D**). In contrast to fat mass, body weight, and fasting blood glucose, insulin resistance did not reach statistical significance until the *Mfap2*<sup>−</sup>/<sup>−</sup> mice were 10-month old. Thus, *Mfap2*<sup>−</sup>/<sup>−</sup> mice

groups (WT or *Mfap2*−/−) over time was determined by two-way ANOVA with Holm–Sidak's correction for multiple comparisons (\**p* ≤ 0.05, *a* = significant relative to 2 months of same genotype, and *b* = significant relative to 6 months of same genotype). Scale bar is 100 μm. Note: two 10-month-old *Mfap2*−/− bones were excluded from study due to suspected fracture.

are severely hyperglycemic by 6 months in age, but not insulin resistant until 10 months.

## Relationship between MAT Expansion and Bone Mass

Tibial bone morphology was assessed by μCT prior to osmium staining for MAT quantification (**Figure 3A**). Trabecular bone analysis was performed using the same region of interest analyzed for MAT (**Figures 3B,C**). Reduced trabecular bone volume (BV/ TV) and BMD were detectable in *Mfap2*<sup>−</sup>/<sup>−</sup> mice by 2 months. Aging of WT mice from 6 to 10 months was associated with statistically significant reductions in trabecular bone volume (−32%, **Figure 3B**) and mineral density (−19%, **Figure 3C**), but not a statistically significant increase in MAT (**Figure 1C**). During this time, *Mfap2*<sup>−</sup>/<sup>−</sup> mice had a 22% reduction in BV/

TV (**Figure 3B**), but a 5.2-fold increase in MAT volume (OV/ TV, **Figure 1C**). Cortical bone density was maintained during aging and unchanged by MAT accumulation (**Figure 3D**). In fact, cortical BMD was highest at 10 months in both WT and *Mfap2*<sup>−</sup>/<sup>−</sup> mice. Additional μCT indices of bone structure can be found in Tables S1 and S2 in Supplementary Material.

## MAT Expansion and Hematopoiesis

To address the impact of pathologic MAT expansion on hematopoietic function, we measured the cellularity of the BM and spleen as well as the number of circulating white blood cells (WBCs) at 10 months of age. BM cellularity was not significantly reduced in *Mfap2*<sup>−</sup>/<sup>−</sup> mice, despite the fivefold increase in MAT (**Figure 4A**). However, spleen cellularity was slightly elevated and WBC count was decreased in *Mfap2*<sup>−</sup>/<sup>−</sup> mice (**Figures 4B,C**). To further investigate these findings, flow cytometry was used to assess hematopoietic lineage distribution. Within the BM of *Mfap2*<sup>−</sup>/<sup>−</sup> mice, there was a modest increase in macrophages, myeloid-derived dendritic cells (MDCs), and B-cells (+61, +78, and +18%, respectively) and a corresponding decrease in neutrophils relative to WT mice (−13%) (**Figure 5A**). In the spleen, there was a significant increase in MDCs (+250%) and decrease in neutrophils (−72%), with a trend to increased B-cells (**Figure 5B**). Finally, in the blood, we observed an increase in monocytes (+216%) and a modest decrease in B-cells (−15%) in *Mfap2*<sup>−</sup>/<sup>−</sup> mice (**Figure 5C**).

## DISCUSSION

Once viewed as inert, space-filling structures, BM adipocytes are now recognized as metabolically active cells with the potential to influence bone remodeling and hematopoiesis. The relationship between MAT volume and peripheral fat mass has been addressed previously (9, 10). In rodents, weight gain resulting from HFD feeding is associated with significant MAT expansion. However, extreme peripheral fat loss due to caloric restriction also correlates with increased marrow adiposity (6). The rapid change in adiposity impedes our understanding of whether changes in peripheral adipose tissue lipid storage shift lipids to the marrow, or if metabolic changes such as insulin resistance contribute to MAT expansion. To address this question, we performed a longitudinal study of mice predisposed to excess, age-associated fat accumulation without dietary intervention.

Microfibril-associated glycoprotein-1 is a matricellular protein that associates with ECM structures but does not contribute to their biophysical properties. Instead, MAGP1 affords fibrillar structures the capacity to regulate signal transduction by tethering growth factors to the ECM [reviewed in Ref. (12)]. MAGP1 is expressed in adipose tissue. MAGP1 deficiency in mice (*Mfap2*<sup>−</sup>/<sup>−</sup>) causes adipocyte hypertrophy in peripheral WAT, which contributes to obesity and then insulin resistance (13). MAGP1 is also highly expressed by osteoblasts in the bone. Loss of MAGP1 *in vivo* leads to increased osteoclast number and reduced bone mass (14). In this study, MAT expansion did not occur coincident with bone loss or excess peripheral adiposity in *Mfap2*<sup>−</sup>/<sup>−</sup> mice; osteopenia was detectable by 2 months and excess adiposity by 6 months in *Mfap2*<sup>−</sup>/<sup>−</sup> mice whereas pathologic MAT expansion was not detectable until 10 months of age. This suggests that MAGP1 deficiency in bone is sufficient to cause abnormal bone remodeling but is unlikely to be mediating the changes in MAT. Instead, MAT expansion coincided with insulin resistance in *Mfap2*<sup>−</sup>/<sup>−</sup> mice. Our findings are of importance because they address three confounding factors of MAT biology: the mechanisms that drive MAT expansion, the relationship between MAT expansion and bone loss, and whether MAT is detrimental to hematopoiesis.

The coincidence of MAT expansion with insulin resistance suggests that insulin may limit MAT adipocyte size or number. Insulin signaling in marrow adipocytes has not been reported, thus, we cannot comment on insulin's direct action on lipid storage or adipogenesis in these specialized cells. However, the pathophysiology of insulin resistance could induce a secondary complication (e.g., hyperlipidemia) that drives subsequent

MAT expansion. Insulin is a potent inhibitor of lipolysis, the breakdown of triglyceride stores, in traditional white adipocytes. Insulin resistance is therefore associated with elevated circulating free fatty acids (FFA) and excess energy storage in atypical sites (e.g., muscle and liver) (19). It is possible that marrow adipocytes respond to elevated FFA associated with insulin resistance by storing them as triglycerides. Rodent models of type 1 diabetes (T1DM) also display significant MAT accumulation (20–22), providing a second example of perturbed insulin signaling associating with hyperlipidemia and MAT expansion.

The relationship between marrow adiposity and bone mass is highly debated. Published reports have shown that MAT expansion can coincide with increased, reduced, or unchanged bone volume (9, 10, 23–27).

Substantial *in vitro* evidence demonstrates that adipocytes and osteoblasts arise from the same mesenchymal progenitor. If true, maturation of mesenchymal stem cells through the adipocyte lineage should be at the expense of the osteoblast lineage. However, *in vivo* evidence is conflicting. Specifically, lineage tracing of marrow adipocytes has demonstrated that the osteoblasts are derived from a Gremlin-1 positive mesenchymal progenitor cell, but these Gremlin-1 positive cells do not differentiate into adipocytes (28). This would allow osteoblast cell numbers to exist independent of marrow adipocyte number. The marrow cavity is also a relatively defined space, and it would, therefore, seem that expansion of MAT must occur at the expense of something else, such as bone. Our study supports a disconnection between MAT volume and bone volume. When WT and *Mfap2*<sup>−</sup>/<sup>−</sup> mice age from 6 to 10 months, there is an increase in MAT and decrease in BV/TV. However, loss of trabecular BV/TV from 6 to 10 months is actually greater in WT mice, despite *Mfap2*<sup>−</sup>/<sup>−</sup> mice having a greater accrual of MAT. Specifically, in WT mice, BV/TV loss is 32% and MAT volume increases 54%; however, MAT volume in *Mfap2*<sup>−</sup>/<sup>−</sup> mice is increased 530% but BV/TV is decreased only 22%.

Marrow adipocytes are considered as members of the hematopoietic niche. Marrow adiposity and hematopoiesis have been studied in the context of marrow ablation by irradiation and chemotherapy. From these studies, it was concluded that the filling of the marrow with fat impedes marrow reconstitution and hematopoiesis (29–31). The current study investigated the consequence of pathologic MAT expansion on existing hematopoietic cells (not reconstitution). We found a significant reduction in circulating WBCs and a small increase in spleen cellularity, indicators of BM failure and extramedullary hematopoiesis. Further study is needed to determine whether MAT expansion in this model leads to a loss of functional hematopoietic stem cells. Unexpectedly, however, total BM cellularity was unchanged. This may be due to shifts in marrow composition to favor loss of fluid space and vascular tone with MAT expansion to allow subsequent preservation of hematopoiesis. We also examined whether fat expansion within the marrow would cause a recruitment (or retention) of macrophages. Macrophages are recruited to peripheral adipose tissue to allow remodeling of adipose tissue during expansion, and because of this, obesity is linked to fat inflammation (32). Indeed, we found a slight increase in macrophage frequency

## REFERENCES


within *Mfap2*<sup>−</sup>/<sup>−</sup> BM. However, the implications of this finding are unclear. Future studies are needed to examine the remodeling of the MAT adipocyte niche in response to obesity.

The goal of this project was to determine the association between pathologic MAT expansion and peripheral indicators of metabolic dysfunction without the use of modified diet. We also addressed the relationship between MAT expansion and bone mass, and the consequence of pathologic MAT expansion on hematopoiesis. We found that MAT volume was not linked directly to peripheral fat accumulation and, instead, coincided with insulin resistance. Furthermore, we found that bone loss occurred, regardless of MAT expansion, in *Mfap2*<sup>−</sup>/<sup>−</sup> relative to WT at all ages. Lastly, we show that MAT expansion is associated with modest alterations in basal hematopoiesis, including a shift from granulopoiesis to B lymphopoiesis.

## AUTHOR CONTRIBUTIONS

TW and ST contributed equally to this work. TW, ST, AS, and BA designed and performed the experiments as well as analyzed the data. TW, ST, and CC wrote the manuscript. GA-E and DL analyzed and interpreted FACS data. CC, ES, RM, and DL provided research materials and contributed to editing of the manuscript.

## ACKNOWLEDGMENTS

We thank Dan Leib for his help in developing the osmium μCT imaging technique, and Kayla Moller for her generation of the 3-dimensional μCT images. This work was supported by National Institutes of Health grants HL-053325 and HL-105314 to RM, and K99-DE024178 to ES. An American Diabetes Association grant 7-13-JF-16 to CC. Washington University Nutrition and Obesity Research Center award P30-DK-056341 to CC, and Washington University's Musculoskeletal Research Center award JIT2014\_Craft\_1 to CC. Technical support was provided by the Mouse Phenotyping Core, the Adipocyte Biology and Molecular Nutrition Core of the Nutrition and Obesity Research Center (grant P30-DK-056341), the Morphology Core of the Digestive Disease Research Core Center (grant P30-DK-52574), and the Washington University Musculoskeletal Research Center structure and strength core (grant P30-AR-057235).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fendo.2016.00087


associated with increased circulating glucocorticoids and not with hypoleptinemia. *Endocrinology* (2016) 157:508–21. doi:10.1210/en.2015-1477


**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 © 2016 Walji, Turecamo, Sanchez, Anthony, Abou-Ezzi, Scheller, Link, Mecham and Craft. 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.*

# Exercise Regulation of Marrow Adipose Tissue

#### *Gabriel M. Pagnotti1 and Maya Styner2 \**

*1Department of Biomedical Engineering, Stony Brook University, Stony Brook, NY, USA, 2Department of Medicine, University of North Carolina, Chapel Hill, NC, USA*

Despite association with low bone density and skeletal fractures, marrow adipose tissue (MAT) remains poorly understood. The marrow adipocyte originates from the mesenchymal stem cell (MSC) pool that also gives rise to osteoblasts, chondrocytes, and myocytes, among other cell types. To date, the presence of MAT has been attributed to preferential biasing of MSC into the adipocyte rather than osteoblast lineage, thus negatively impacting bone formation. Here, we focus on understanding the physiology of MAT in the setting of exercise, dietary interventions, and pharmacologic agents that alter fat metabolism. The beneficial effect of exercise on musculoskeletal strength is known: exercise induces bone formation, encourages growth of skeletally supportive tissues, inhibits bone resorption, and alters skeletal architecture through direct and indirect effects on a multiplicity of cells involved in skeletal adaptation. MAT is less well studied due to the lack of reproducible quantification techniques. In recent work, osmium-based 3D quantification shows a robust response of MAT to both dietary and exercise intervention in that MAT is elevated in response to high-fat diet and can be suppressed following daily exercise. Exercise-induced bone formation correlates with suppression of MAT, such that exercise effects might be due to either calorie expenditure from this depot or from mechanical biasing of MSC lineage away from fat and toward bone, or a combination thereof. Following treatment with the anti-diabetes drug rosiglitazone – a PPARγ-agonist known to increase MAT and fracture risk – mice demonstrate a fivefold higher femur MAT volume compared to the controls. In addition to preventing MAT accumulation in control mice, exercise intervention significantly lowers MAT accumulation in rosiglitazone-treated mice. Importantly, exercise induction of trabecular bone volume is unhindered by rosiglitazone. Thus, despite rosiglitazone augmentation of MAT, exercise significantly suppresses MAT volume and induces bone formation. That exercise can both suppress MAT volume and increase bone quantity, notwithstanding the skeletal harm induced by rosiglitazone, underscores exercise as a powerful regulator of bone remodeling, encouraging marrow stem cells toward the osteogenic lineage to fulfill an adaptive need for bone formation. Thus, exercise represents an effective strategy to mitigate the deleterious effects of overeating and iatrogenic etiologies on bone and fat.

Keywords: exercise, marrow adipose tissue, quantitative image analysis, bone microarchitecture, lipid, PPAR**γ**, rosiglitazone, exercise

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Petra Simic, Massachusetts Institute of Technology, USA Jonathan Gooi, The University of Melbourne, Australia*

#### *\*Correspondence:*

*Maya Styner mstyner@med.unc.edu*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 21 May 2016 Accepted: 04 July 2016 Published: 14 July 2016*

#### *Citation:*

*Pagnotti GM and Styner M (2016) Exercise Regulation of Marrow Adipose Tissue. Front. Endocrinol. 7:94. doi: 10.3389/fendo.2016.00094*

## MARROW ADIPOSE TISSUE

Increased marrow adipose tissue (MAT) is associated with states of impaired bone formation (1, 2) and dysfunctional hematopoiesis (3–5), although its physiological role remains unclear. In humans, pathologists have noted that MAT increases in healthy subjects with age, beginning in the distal long bones and accruing proximally such that by age 25, approximately 70% of the marrow space is filled with MAT (4). In addition to physiologic MAT, which accrues with aging, this fat depot – housed within bone – is abundant in states of low bone density: osteoporosis (6), anorexia nervosa (7), skeletal unloading (8, 9), and anti-diabetes therapies (10), conditions that are also associated with skeletal fractures. Adipocytes within the marrow originate from the mesenchymal stem cell (MSC) pool that also gives rise to osteoblasts, chondrocytes, and myocytes, among other cell types (11, 12). Recent work suggests that increased marrow fat can also be demonstrated in the setting of preserved or increased bone density (high-fat feeding or obesity) (13–15) and, thus, challenges the premise that the relationship between MAT and bone volume is reciprocal. The MAT/ bone relationship is further complicated by the identification of a new population of Grem1<sup>+</sup> MSC (16), a phenotype capable of differentiating into osteoblasts and chondrocytes, but not adipocytes: the Grem1<sup>+</sup> population differs from the LepR<sup>+</sup> MSC, which do generate marrow adipocytes. Whether senile marrow invasion with adipocytes represents a later predominance of a LepR<sup>+</sup> MSC population is unknown but complicates considerations as to the physiologic and/or pathologic role of MAT.

In the case of diet-induced obesity, marrow fat also increases compared to normal weight controls, but whether this contributes to bone fragility is unclear (17). Nevertheless, if the burden of fat across the marrow space is inevitable, then perhaps what's more worthy of an investigation is the quality of the MAT being generated, possibly representing a direct reflection of the health of the surrounding bone. Importantly, the unsaturation index of MAT increases with aging, and thus, this feature of MAT may shed light on its physiology; nonetheless, unsaturation index of MAT is unaffected by physical activity (18). While subcutaneous white fat depots store excess energy and provide a clear evolutionary advantage during times of scarcity (19), MAT's purpose remains indeterminate, harboring characteristics of both white and brown fat (20). WAT serves as a source of adipokines and inflammatory markers that have both positive (e.g., adiponectin) (21) and negative (22) effects on metabolic and cardiovascular endpoints. Visceral abdominal fat is a distinct depot of WAT that is proportionally associated with negative metabolic and cardiovascular morbidity (23), regenerates cortisol (24), and has been linked to reduced bone formation (25, 26). WAT substantially differs from brown adipose tissue (BAT), as defined by a panel of proteins that support BAT's thermogenic role (27). MAT, by virtue of its specific marrow location and its adipocyte origin from at least LepR<sup>+</sup> marrow MSC, is clearly demarcated from non-bone fat depots by higher expression of bone transcription factors (28) and likely represents a unique fat phenotype (29). Recently, MAT was noted to produce a greater proportion of adiponectin – an adipokine associated with improved metabolism – than WAT (30), suggesting an endocrine function for MAT as distinct depot, akin, but different from that of WAT. Moreover, deficiency of histone deacetylase 3 (Hdac3), known to play a major role in skeletal development and lipid metabolism, increases MAT volume, implicating this important transcriptional regulator in MAT development (31). Potentially, MAT might serve multiple functions, reflecting those of both white and brown fat, storing lipid in preexisting adipocytes, secreting adipokines, and generating heat. Exercise universally affects the metabolism of both WAT (32) and more recently BAT (33, 34); thus, exercise intervention can be harnessed as a powerful tool to query the poorly understood physiology of MAT.

## MEASUREMENT AND QUANTIFICATION OF MARROW ADIPOSE TISSUE

In order to quantify and characterize the effects of exercise on MAT, various analytic methods were considered. Until recently, qualitative measurements of MAT have relied on bone histology (35, 36), which is subject to site selection bias and cannot adequately quantify the volume of fat in the marrow. Nevertheless, histological techniques and fixation make possible *in situ* visualization of MAT, quantification of adipocyte size, and MAT's association with the surrounding endosteum, milieu of cells, and secreted factors (37–39).

Recent advances in cell surface and intracellular marker identification and single-cell microfluidic analyses have led to greater resolution and high-throughput *ex vivo* quantification. Flow cytometric quantification can be used to purify adipocytes from the stromal vascular fraction of most fat depots (40). Early research with such machinery cited adipocytes as too large (50–200 mm) and fragile for cytometer-based purification, as their cytoskeleton lacks rigidity, rendering them susceptible to lysis; however, recent advances have been made to mitigate this (41). One may distinguish discrete adipocyte subpopulations from other cells by utilizing internal lipid content and surface biomarker identification. Filtration of the marrow (pore size = 150 μm) permits adequate flowthrough for adipocytes and smaller cellular contents. Subsequent centrifugation of the suspension aids in isolating adipocytes. Maintaining laminar flow and optimal temperatures when sorting has led to greater viability and precision. However, accumulation of lipid and protein content can adhere to the sheath tubing, thereby clogging the instrumentation (42). High-binding affinity of protein to antibodies bound by fluorescent probes used in FACS has made the identification of MAT and the cells that cohabitate the marrow increasingly specific, though these measurements provide little information on adipocyte location within the marrow microenvironment.

To improve our understanding of MAT, novel imaging techniques have recently been developed as a means to visualize and quantify MAT, *in situ*. Although proton magnetic resonance spectroscopy (1H-MRS) has been used with success to quantify vertebral MAT in humans (43), it is more difficult to employ in laboratory animals (44). Magnetic resonance imaging (MRI) provides MAT assessment in the vertebral skeleton (45) in conjunction with μCT-based marrow density measurements (46). A volumetric method to identify, quantify, and localize MAT in rodent bone has been recently developed, requiring osmium staining of bones and μCT imaging (47), followed by image analysis of osmium-bound lipid volume (in cubic millimeter) relative to bone volume (see **Figure 1**) (13, 48). Briefly, femurs stripped of connective tissue are decalcified, immersed in osmium tetroxide, and placed in potassium dichromate (49). Bones are scanned using μCT imaging (resolution 10 μm × 10 μm × 10 μm). Image processing consists of rigid image coalignment (**Figures 1B,D**, Slicer) (50), regional masking, allowing consistent regional measurements and superimposed visualizations. Additional bone masks are established in a semiautomatic contouring of the femur. As osmium is significantly more dense than bone [Hounsfield units (HU) ~ 700–2000], HU thresholds are set to capture low osmium from 2000 to 3000 HU, mid osmium from 3000 to 4000 HU (**Figures 1E,F**, green), and high osmium from 4000 to 5000 HU (**Figures 1E,F**, blue), and quantified accordingly (cubic millimeter). The lowest threshold is set above dense cortical bone (2000 HU) (51), and, thus, the contribution of potentially mislabeled cortical bone to the osmium volume is negligible (51, 52). Following quantification, the femur is subdivided into anatomical regions wherein regional osmium volume is normalized to bone volume. Aligned bone images are then averaged across all images as a reference for visualization. Average images are also computed for each group to obtain color-coded visualizations of the osmium densities to allow additional visual comparison of MAT between groups (**Figure 2**). This technique provides reproducible quantification and visualization of MAT, enabling the ability to quantify changes in MAT with diet, exercise, and agents that constrain precursor lineage allocation.

## EXERCISE REGULATION OF MARROW ADIPOSE TISSUE IN THE SETTING OF HIGH-FAT FEEDING

Marrow adipose tissue volume has recently been demonstrated to increase during short-term, high-fat feeding in rodents relative to the total bone volume (13), rising more rapidly than calorie induction of visceral fat depot size (53). A positive association between obesity and MAT was noted in rodents (15) and humans (43). In fact, obese individuals generally present with higher bone density (54–62) and are unlikely to experience classical osteoporotic fragility fractures (63). Because MAT has been associated with states of low bone density and increased fracture risk (6, 8, 64), the impact of fat in the marrow likely varies depending on its etiology. As obesity associates with increased bone density and lower fracture, MAT in obesity could represent a distinct fat depot that supports skeletal anabolism.

The *Nurse's Health Study* showed that exercise affords a morbidity and mortality benefit that is independent of weight loss or calorie expenditure (65). One well-known effect of exercise is to improve bone strength, limiting the impact of postmenopausal osteoporosis (66). Exploring the salutary effects of exercise on bone and MAT in the setting of HFD, small animal models have been employed to address the degree of responsivity under said conditions. In one such experiment, 6 weeks of daily voluntary running exercise suppresses MAT accumulation in mice fed both a regular and HFD (13), in contrast to their non-exercised counterparts. While HFD does not perturb trabecular bone parameters as compared to control, significant gains in trabecular bone

volume and trabecular thickness are noted in tibiae of exercised animals. Cortical bone parameters are unaltered by HFD and exercise in this short-term study. Thus, the trabecular compartment might be more receptive to both metabolic disarray and to mechanical signals from exercise than cortical bone, especially when challenged with elevated MAT *via* HFD.

## EXERCISE REGULATION OF MARROW ADIPOSE TISSUE IN THE SETTING OF ROSIGLITAZONE

It is accepted that elevated MAT due to PPARγ-agonist treatment in mice (48, 49, 67) and humans (68, 69) is due to PPARγ-induction of adipogenesis from marrow MSC. *In vitro*, rosiglitazone (PPARγagonist)-treated MSCs resist mechanically stimulated adipogenesis when treated with a dynamic load (70). *In vivo*, it was previously hypothesized that exercise might not overcome pathologically induced MAT in the setting of rosiglitazone; however, interestingly, exercise suppresses MAT even in the face of powerful adipogenic biasing (*via* MSC) in rosiglitazone-treated mice (48). While rosiglitazone significantly elevates cortical porosity in proximal tibiae, bone quantity is unaffected in these young mice. Exercise associates with increased trabecular bone volume fraction and trabecular thickness in control and rosiglitazone-treated mice (48). This further highlights the positive impact of exercise and mechanical signals on healthy bone as well as bone challenged by pharmacologic agents shown to facilitate marrow adiposity as well as cortical porosities.

## EXERCISE AND MECHANICAL REGULATION OF BONE QUALITY AND QUANTITY

## Exercise and Musculoskeletal Mass

Of the many health benefits attributed to exercise – improved neurological endpoints (71, 72), cardiovascular health (73), reduced inflammation (74–77), and decreased risk for chronic disease development, perhaps, most immediately and visually apparent is its ability to augment musculoskeletal mass (78–81). Exercise is known to encourage anabolic responses in musculoskeletal tissue (i.e., bone, muscle, ligament, and tendon) (82) as a consequence of successive bouts of musculoskeletal-loading. Indeed, skeletal tissue is known to adapt to meet loading demands (83), altering its bone remodeling strategy to sustain maximum loads. Prominent and varied forces are exerted across the appendicular and axial skeleton (84) through exercise: the musculoskeletal construct, thereby, acts as a conduit for transducing muscle contractive perturbations at the bone surface at both low frequencies (e.g., bicep flexion) and high frequencies (type II fast-twitch muscle). At the cellular level, these responses are mediated by a wide spectrum of mechanical stimuli of both high- and low-magnitude stresses, information that is internalized by cells through cytoskeletal and transmembrane-bound integrins linking the extracellular environment with the genetic machinery encased within the nucleus (85–90). Thus, these mechanical factors transcribe osteo-, chondro-, or myogenic (91) growth factors (RunX2) while deterring pathways conducive to adipogenesis (PPARγ) (70, 92–95). We now know that these signals propagate across the Wnt/β–Catenin transduction pathways (96), upregulating expression of genes that drive osteogenic (RunX2) and chondrogenic (SOX9) growth, while also being positioned across other complex signaling pathways involving other signals of interest, such as MAPK, pRB, FGFs, and TGF-β. The ability of mechanical stimuli to regulate musculoskeletal mass is likely multifactorial, occurring through repression of fat generation as well as bone resorptive pathways (97), while at the same time stimulating musculoskeletal anabolism. Additionally, exercise has been shown to improve resistance to fracture, specifically through the separation of exercise into shorter regimens (98).

## Skeletal Unloading

The importance of mechanical information on bone adaptation can also be exemplified by observing the response of musculoskeletal tissue and the underlying cellular dynamics in the Pagnotti and Styner Exercise and Marrow Fat

absence of mechanical cues (99–101). Whereas exercise delivers large quanta of mechanical information, as a function of disuse (chronic bed rest, microgravity, or reduced physical activity) (102, 103), regulation of musculoskeletal tissue homeostasis is compromised, instead elevating conditions wherein muscle, tendon, and ligament (and fat) undergo catalysis and rapid resorption of bone. Together, these tissue-level responses heighten the occurrence of osteoporotic bone and degree of fracture risk: these outcomes are in direct response to lapses in mechanical input (104–107). Whether chronic skeletal unloading encourages a specific MAT phenotype remains unclear, yet, studies have definitively shown that extended bed rest drives an increased marrow adipogenesis (108). In the absence of mechanical cues, PPARγ and receptor activator of nuclear factor-Kappa-B ligand (RANKL), which promotes osteoclast-mediated bone resorption, are both elevated, indicating an effect that could be stemmed upon reintroduction of mechanical stimuli.

## Mechanical Effect on Bone and Fat Precursors

Both *in vitro* and *in vivo* studies have demonstrated that MSCs and early, non-committed progenitors exhibit unquestionable responsivity to mechanical loading (93, 109). Osteocytes (88, 110), osteoblasts (111, 112), and pre-osteoblast MC3T3 cells (113) in the marrow are other known mechanosensitive cells and contribute to the complex transduction of mechanical information driving osteogenic gene expression. While mechanical signals can inhibit osteoclastogenesis and subsequent bone resorption through direct effects on osteocyte and MSC expression of RANKL, it is known that mechanical effects on bone remodeling also involve regulation of MSC differentiation toward osteogenesis (93, 114, 115) facilitated by Wnt/β–Catenin signaling and uncommitted precursors. This is, in part, due to the plasticity of stem cells to differentiate specifically toward one mesenchymal lineage over another as dictated by environmental cues, such as exogenous mechanical stimuli (93), local substrate rigidity (113, 116, 117), and regional cytokine signaling gradients (11, 118). These factors drive MSC and other resident marrow cells toward fulfilling their role in musculoskeletal homeostasis by promoting formation of bone and other critical tissues that support skeletal health in lieu of engaging pathways conducive to adipogenesis. Exercise not only encourages MSC proliferation but downstream lineages are also influenced as well: lipid droplets and adipocyte cell diameters are reduced while driving osteogenic potential through upregulated alkaline phosphatase activity (119). *In vivo* studies show decreased adipocytes and increased pre-osteoblasts in the marrow of running rats (38) and climbing mice (120).

Rodent studies highlight increased bone formation rates in response to exercise and mechanical signals *via* dynamic histomorphometry (104), including running exercise (121–124). These responses persist through incorporation of non-exercise mechanical loading interventions [low-magnitude mechanical signals (LMMS)], which have been demonstrated to increase bone formation rates in loaded tibiae of mice (125, 126). In consideration of the phenotypic differences in lineage subtypes across niches, the bone marrow-derived MSC has recently been suggested to have unique, focal-specific properties (127). Therefore, it is important to weigh the contribution of precursor cells and other progenitors in the presence of mature adipocytes or the marrow, when considering the effect mechanical stimuli may have on their interaction with the surrounding milieu.

## Low-Magnitude Mechanical Signals Effect on Bone and Marrow Adiposity

While physical activity presents an ideal strategy to introduce exogenous low-frequency, high-magnitude mechanical cues to musculoskeletal tissue, this approach is impractical for those patients with compromised bone microarchitecture (i.e., osteoporosis, osteopenia) (105, 109, 128–130) or muscle instability (131–133), populations that could benefit the most from their effects. Alternatively, mechanical signals delivered to the skeleton in the form of high-frequency, LMMS (fast-twitch muscles controlling balance and posture) can be introduced outside of the context of physical activity (134). For instance, low strain (<100 μs) displacements contribute more toward maintaining musculoskeletal health than higher magnitude strains and can be delivered whole-body using platforms to oscillate in the high-frequency domain (20–100 Hz), while maintaining a lowmagnitude (i.e., sub-gravitational) acceleration. In doing so, these platforms partially reintroduce the spectral content of muscle contraction (105, 135), thereby exerting a beneficial quotient of exercise without risking fracture to bone resulting from extreme loads prevalent in exercise. Moreover, when separated into multiple administrations, the effects of both exercise and LMMS are amplified (95), encouraging enhancement in the responsiveness of MSC. Importantly, in humans, the non-pharmacologic therapy LMMS prevent bone loss due to Crohn's disease (136), in children recovering from various cancers (137), and in other disabling conditions where bone losses are apparent (138, 139).

## EXERCISE AND THE BROWNING OF MARROW ADIPOSE TISSUE

Upon initiation of exercise, there is an increase in uptake and oxidation of lipids in skeletal muscle (140). When exercise intensity increases, fuel selection shifts toward an increase in carbohydrate and decrease in fat utilization. In contrast, endurance training is associated with a shift toward an enhanced lipid utilization (140). BAT, initially observed in hibernating mammals and human infants, dissipates energy in the form of heat through non-shivering thermogenesis (141). Inducible brown fat depots – beige fat – have been discovered within the white adipose tissue of adult humans (142). On exposure to cold or β-adrenergic stimulation, these beige/brite fat cells express high levels of mitochondrial uncoupling protein UCP1 and fat globules become multilocular (143), characteristics of the brown fat phenotype. Irisin, a muscle-derived hormone induced by exercise, also activates UCP1 expression and browning of white adipose tissue (33): coactivator PPAR-γ coactivator-1 α (PGC1-α) stimulates irisin, and transgenic mice overexpressing PGC1-α exhibit increased energy expenditure despite no changes in food intake or activity (33). Overall, there is evidence that fat depots can alter phenotype to serve functional demands. Since exercise browns white adipose depots (33), it is conceivable that exercise might result in analogous browning of MAT. Indeed, running exercise increased UCP1 in bone mRNA (48). UCP1 is localized in the mitochondrial inner membrane of mammalian BAT and is, therefore, a specific marker for BAT (144). This increase in UCP1 with exercise may indicate a brown phenotype within MAT adipocytes; however, this requires further confirmation. Interestingly, exercise-induced increases in UCP1 expression may correlate with increases in irisin (33), although irisin's role in exercise physiology remains unclear (145). Finally, recent work suggests that irisin may have direct effects on bone in addition to its known effect on adipocytes, and thus irisin's role in skeletal health remains an area of active investigation (146–148).

## CONCLUSION

Marrow adipose tissue, housed within bone and interspersed with hematopoietic elements, remains a poorly understood fat depot, likely due to its anatomic location, rendering it inaccessible and thus challenging to quantify. Clinicians are particularly interested in the physiology of this fat depot due to its association with low bone density states and pharmacologic agents that increase fracture risk. As more robust volumetric imaging and quantification tools emerge [e.g., osmium-μCT Ref. (13, 48)], precise determinations can be made regarding MAT physiology and relationship to bone health. Interestingly, these methodologies have pointed to an increase in MAT in the setting of high-fat feeding (13), without an impact on bone quantity. Additionally, we are now able to visualize the effect of pharmacologic PPARγ activation with rosiglitazone on bone and to appreciate the significant encroachment of marrow fat (**Figure 2**) as well as to quantify these dramatic findings. Interestingly, both regimented exercise and LMMS serve to counter the effects of obesity and

## REFERENCES


potent pharmacologic agents on bone remodeling. Further, in evaluating the response of MAT to mechanical stimuli, we highlight a positive effect toward normalizing bone parameters and, in doing so, constraining expansion of MAT across the marrow. It is possible that exercise serves to "brown" MAT as indicated in Ref. (48); however, further work is needed to establish the metabolic purpose of MAT in the setting of exercise. It remains unclear whether exercise-induced bone formation biases MSCs away from adipogenesis in order to recruit osteoblasts or whether alternative mechanisms are involved. *In vitro*, MSCs are highly responsive to mechanical signals during differentiation; indeed, mechanical loading slows adipogenesis (38, 92, 93, 149–151). Conversely, MSCs under microgravity conditions decrease osteogenic differentiation in favor of fat formation, an event accompanied by elevated nuclear expression of PPARγ (152). Thus, in the setting of mechanical input or exercise, bone formation is increased and marrow fat is suppressed, highlighting a likely mechanistic relationship between MAT and bone in the setting of mechanical stimulation. Other pathways are likely involved in the exercise regulation of MAT: lipogenesis, lipid uptake, skeletal anabolism, regulation of hematopoiesis in the bone marrow, and regulation of adipokines and cytokines. The elucidation of these pathways and their role in MAT/bone regulation in the setting of exercise remains an area of active investigation.

## AUTHOR CONTRIBUTIONS

MS and GP wrote this review together as a team. Each author contributed approximately 50% to the writing of this review.

## FUNDING

Funding for this work was provided by NIAMS: AR062097 and NIDDK: P30DK056350.


in the water maze and modulates adult hippocampal neurogenesis in mice. *Neurobiol Learn Mem* (2016) 131:26–35. doi:10.1016/j.nlm.2016.03.002


differentiation and inhibition of adipogenesis of human mesenchymal stem cells. *J Bone Miner Res* (2007) 22(11):1720–31. doi:10.1359/jbmr.070721


**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 © 2016 Pagnotti and Styner. 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.*

# Increased Bone Marrow Adiposity in a Context of Energy Deficit: The Tip of the Iceberg?

*Olfa Ghali1,2, Nathalie Al Rassy1,2, Pierre Hardouin1,2 and Christophe Chauveau1,2\**

*<sup>1</sup> Laboratoire de Physiopathologie des Maladies Osseuses Inflammatoires, Université de Lille, Boulogne-sur-Mer, France, 2 Laboratoire de Physiopathologie des Maladies Osseuses Inflammatoires, Université du Littoral Côte d'Opale, Boulogne-sur-Mer, France*

Elevated bone marrow adiposity (BMA) is defined as an increase in the proportion of the bone marrow (BM) cavity volume occupied by adipocytes. This can be caused by an increase in the size and/or number of adipocytes. BMA increases with age in a bone-site-specific manner. This increase may be linked to certain pathophysiological situations. Osteoporosis or compromised bone quality is frequently associated with high BMA. The involvement of BM adipocytes in bone loss may be due to commitment of mesenchymal stem cells to the adipogenic pathway rather than the osteogenic pathway. However, adipocytes may also act on their microenvironment by secreting factors with harmful effects for the bone health. Here, we review evidence that in a context of energy deficit (such as anorexia nervosa (AN) and restriction rodent models) bone alterations can occur in the absence of an increase in BMA. In severe cases, bone alterations are even associated with gelatinous BM transformation. The relationship between BMA and energy deficit and the potential regulators of this adiposity in this context are also discussed. On the basis of clinical studies and preliminary results on animal model, we propose that competition between differentiation into osteoblasts and differentiation into adipocytes might trigger bone loss at least in moderate-to-severe AN and in some calorie restriction models. Finally, some of the main questions resulting from this hypothesis

are discussed.

Keywords: bone marrow adiposity regulation, osteoporosis, gelatinous bone marrow transformation, mesenchymal stem cell differentiation, anorexia nervosa

## INTRODUCTION

The bone marrow (BM) is predominantly composed of two fractions, namely the hematopoietic and stromal fractions. Mesenchymal progenitor cells reside in the stromal fraction of the BM and can differentiate primarily into osteoblasts and adipocytes (1).

Bone marrow adiposity (BMA) is defined as the proportion of the BM cavity volume occupied by adipocytes. A rise in the BMA can be caused by an increase in the size and/or number of adipocytes (2). BMA levels differ for males vs. females; they are usually higher in women than in men, and higher in female animal models than in males (3). It is also well known that BMA increases with age in a bone-site-specific manner. At birth, the BM is largely hematopoietic (red marrow) but is gradually converted to fatty (yellow) marrow over the lifespan (4–7). The age-related progression of BMA in the axial skeleton differs from that observed in the appendicular skeleton.

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Elaine Dennison, University of Southampton, UK Jan Josef Stepan, Charles University in Prague, Czech Republic Maureen Devlin, University of Michigan, USA*

*\*Correspondence:*

*Christophe Chauveau chauveau@univ-littoral.fr*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 04 May 2016 Accepted: 30 August 2016 Published: 16 September 2016*

#### *Citation:*

*Ghali O, Al Rassy N, Hardouin P and Chauveau C (2016) Increased Bone Marrow Adiposity in a Context of Energy Deficit: The Tip of the Iceberg? Front. Endocrinol. 7:125. doi: 10.3389/fendo.2016.00125*

The fat fraction in iliac crest biopsies is relatively low during childhood but increases gradually with age (8). In contrast, BMA in long bones of the limbs starts to rise during the first few years of life, with progressive colonization of the medullar cavity from the diaphysis. This process is complete at the age of 25 years or so (9).

These observations suggest that (i) the presence of adipocytes in the BM is a physiological, age-related phenomenon, and (ii) these adipocytes may have positive effects. In fact, adipocytes can supply energy for the high level of bone remodeling required during puberty (10, 11). Furthermore, BM adipocytes may be involved in thermogenesis and heat dissipation because they display some of the characteristic features of brown adipocytes (12). However, an increase in BMA may be linked to certain pathophysiological situations. In fact, osteoporosis or compromised bone quality related to mechanical unloading (13), aging (14), menopause (15), diabetes (16, 17), obesity (18, 19), or anorexia nervosa (AN) (20) is frequently associated with high BMA [for a review, see the article by Hardouin et al. (21) in this special issue].

Two hypotheses have been proposed to explain the bone alterations induced by BMA increase: (i) the elevated BMA observed in osteoporotic patients is caused by a shift in mesenchymal stem cell commitment from the osteogenic pathway to the adipogenic pathway at the expense of bone formation (22–24) and (ii) BM adipocytes can regulate the BM microenvironment and act negatively on the balance between bone formation and bone resorption (11, 25–27) and hematopoiesis (28–31).

A better understanding of the quantitative and/or qualitative changes in BM adipocytes during osteoporosis might enable the development of novel therapeutic strategies. Of the various pathologic contexts associated with osteoporosis, AN is of particular interest. Although osteoporosis in AN has been extensively described, the present article discusses changes in BMA and the regulation of this process in a context of energy deficit and, then, proposes some perspectives for research in this specific field.

## BMA AND ENERGY DEFICIT

Anorexia nervosa is a major public health concern; it is the psychiatric disease with the highest mortality rate. Given that AN often coincides with puberty, this condition markedly interferes with bone mass gain and has a long-term impact on bone quality (32). In particular, studying AN may reveal novel, hitherto unsuspected links between BMA and bone mass. First, secondary osteoporosis in AN is often associated with an increase in BMA that contrasts with a decrease in the adiposity of other fatty tissues (**Table 1**). Second, some studies (but not others) have found that levels of inflammation markers are not elevated in patients with AN (despite amenorrhea) (33, 34). Given that most patients with AN have a low bone mass, inflammation-independent mechanisms may be involved in this bone loss.

However, the fact that a link between energy deficit and increased BMA is not always observed raises questions as to the relevance of BMA in AN. Indeed, several studies of osteoporotic patients with AN did not observe elevated BMA (**Table 1**). In fact, the studies having found an increase in BMA tended to feature patients with a higher body mass index (BMI) than those in studies that did not observe this increase. Moreover, a number of studies that categorized patients into BMI classes highlighted a relationship between the severity of body weight loss and changes in the BMA (40, 43, 44). Interestingly, Abella et al.'s histological study of patient biopsies described a hypoplastic BM in which an increase in both the fat fraction and adipocyte diameter was associated with moderate body weight loss (40). The researchers also observed areas of unaffected or hypoplastic BM and focal gelatinous degeneration in patients with intermediate levels of body weight losses (40). Last, the most severe cases featured a high microadipocyte count in a hyaluronic acid matrix. It is noteworthy that, in one of the patients in Abella et al.'s study, the BM's appearance normalized after the resumption of food intake and the return to an appropriate bodyweight (40). Vande Berg et al. used magnetic resonance imaging to highlight the presence of serous-like BM at different skeletal sites in anorectic patients (44). They distinguished between two groups as a function of the presence (group A) or absence (group B) of a water-like pattern in the marrow spaces. The individual BMI values in the group A were all under 13.5 kg/m2 , whereas six of the eight patients in group B had a BMI above 15 kg/m2 . Abella et al. did not find an intergroup difference in disease duration, which suggests that body weight loss is a key factor in BM alterations (39–41, 46).

This short overview of the relationship between BMA alterations and the severity of the body weight loss highlights the complexity of the processes leading to these alterations and probably, thereafter, to bone loss. Studies of animal models should provide some clues to these processes. In rodent models of food restriction or calorie restriction, only four published studies have focused on BMA (**Table 2**). First, Hamrick et al. studied the effects of a 40% food restriction in 14-week-old male mice (60). The 10-week protocol led to a body weight loss of 30% (relative to control mice) and the total disappearance of BM adipocytes in the distal femur and lumbar vertebra (60) (two adipocyte-poor medullar sites). Despite this lack of adipocytes, the cortical thickness fell in the femur and lumbar vertebra, and the trabecular thickness fell in the femur (60). However, Devlin et al.'s 2010 study of the effects of a 9-week, 30% food restriction in 3-week-old male mice highlighted decreases of 11 and 27% in the bone volume/ total volume (BV/TV) ratio and in cortical thickness, respectively (61). Although this protocol induced a relative body weight loss of 40% (compared with control mice), the food-restricted animals were nevertheless 80% heavier at the end of the protocol than at the start. The main impact on bone microarchitecture was observed in the femur, where the BM adipocyte density was 700% higher than in control animals. Third, Baek and Bloomfield fed 6-month-old Sprague-Dawley female rats a calorie-restricted diet for 12 weeks (62). At the end of this period, the calorie-restricted rats displayed body weight losses of 20% (relative to day 0) and 25% (relative to control rats). The volumetric bone mineral density of the proximal tibia was 14% below that of control mice, while the BMA in the proximal femur had doubled. More recently, Cawthorn et al. showed that 6 weeks of a 30% calorierestricted diet in 9-week-old female mice induced a final increase in tibia BMA of 700% vs. mice able to feed *ad libitum* (27). In our separation-based model of AN (SBA), 8-week-old female mice were housed singly and submitted to time-restricted feeding (in


#### Table 1 | Variations in BMA and the presence of gelatinous BM transformation in patients with AN, as reported in case–control studies and case reports.

*BMI, body mass index; BMA, bone marrow adiposity; GBMT, gelatinous bone marrow transformation; H-MRS, proton magnetic resonance spectroscopy; ND, not determined; MRI, magnetic resonance imaging; CT, computed tomography; ND, not determined; AN, anorexia nervosa.*

*Arrows: the variations in patients with AN differed significantly from those observed in control subjects.*

*a Data from 15 case reports on GBMT in patients with AN (46–59).*

*bExcept for two cases where MRI was applied.*

#### Table 2 | Variations in BM adipose tissue (BMAT) content in calorie-restricted rodent models.


order to avoid a compensatory increase in food intake and, thus, attain a daily food intake close to that of *ad libitum* mice) (63). The calorie-restricted animals rapidly lost around 25% of their initial body weight, which corresponded to a body weight loss of 40% (relative to control mice) after 10 weeks of the protocol. The bone mass gain observed in control mice was curtailed after 2 weeks of the protocol (63). After 10 weeks, the tibia trabecular BV/TV ratio, trabecular thickness, and cortical thickness were, respectively, 32, 31, and 15% lower in SBA mice than in control mice. However, the BMA level in the proximal tibia was similar to that observed in control mice (unpublished data). These results showed (as did Hamrick et al.'s study) that energy deficit can induce bone alterations in the absence of an obvious increase in BMA (60). Interestingly, Cawthorn et al. showed very recently that, in rabbit, CR leads to bone loss, even without increases in BMA (64).

To learn more from these animal datasets about BMA regulation, one can attempt to compare them. The five above-mentioned studies assessed rodents of different genders and at different ages. The well-known differences between male and female physiology and the differences in BMA variation with age make it difficult to compare males with females (65). For the studies of male animals, the difference in observed BMA changes between Hamrick et al.'s study and Devlin et al.'s study might be linked to the age difference. One might conclude that calorie restriction induces a large increase in BMA in young male rodents only. However, it is also noteworthy that there is a late-onset BMA increase in male control mice; this might reduce the possible BMA increase in older animals (as observed at the end of Hamrick et al.'s study). When considering the three studies of female rodents, Baek's work was performed on rats aged up to 9 months. It is difficult to compare old female rats with the 8- and 9-week-old C57Bl/6 female mice used by Cawthorn et al. and Zgheib et al., respectively. The main differences consisted in the body weight changes (relative to the start of the study) and the BMA changes in the tibia. Cawthorn et al. observed a stable body weight and a 700% increase in BMA, whereas Zgheib et al. observed a 25% decrease in body weight but no significant change in BMA. Given that these two models were similar enough to enable a valid comparison, one can hypothesize that the magnitude of body weight loss influences the change in BMA most strongly in young adult female mice.

## POTENTIAL REGULATORS OF BMA IN A CONTEXT OF ENERGY DEFICIT

As described above, it appears that BMA may be modulated by age, gender, and the severity of the energy deficit. However, it is important to bear in mind that BMA can also be affected by many different biological factors, including hormones, growth factors, and pharmacologic agents. Although the molecular mechanisms involved in BMA regulation have not been fully defined, they appear to converge on the modulation of peroxisome proliferator activated receptor gamma 2 (PPARγ2) expression and/or activity.

When considering potential regulators of BMA in subjects with a severely reduced calorie intake, it has been shown that dysregulation of the growth hormone insulin-like growth factor 1 axis and low leptin levels can stimulate BM adipogenesis (19, 61, 66). Moreover, starvation prompts the mobilization of fat stores, which might activate PPARγ2 and stimulate adipocyte differentiation in the BM (27). Furthermore, calorie restriction in a mouse model resulted in low expression and activity of runtrelated transcription factor 2 (Runx2) and tafazzin (TAZ), both of which are pro-osteogenic transcriptional factors in BM stromal cells (67, 68).

Several other investigators have established links between BMA and pre-adipocyte factor 1 (Pref1). Indeed, circulating Pref1 concentrations are higher in women with AN than in healthy controls (66). In women who have recovered from AN, Pref1 levels are significantly lower than in normal-weight controls (66). Hence, it is possible that elevated Pref1 levels in AN may be involved in the increase in BMA.

It is also known that the production of reactive oxygen species (ROS) is greater in AN patients than in control subjects (69). ROS greatly influence the generation and survival of bone cells (70). *In vitro* studies of bone metabolism have shown that oxidative stress inhibits osteoblastic differentiation and induces apoptosis (71, 72).

Interestingly, women with AN have higher cortisol levels than healthy controls (73, 74). Cortisol levels are also predictive of a low bone mineral density in AN (73). In animal models, Cawthorn et al. showed that both BMA and glucocorticoids increase during CR in mice, but neither of these increases during CR in rabbits, suggesting a narrow link between these two phenomena (64). Experiments *in vitro* suggest that corticosteroids lead to adipogenesis by activating PPARγ2 and blocking the Wnt signaling pathway (75, 76). These findings also suggest that cortisol, Pref1, and leptin are all potential BMA regulators in a context of energy deficit.

Furthermore, women with AN have low estrogen levels (77, 78). Postmenopausal estrogen deficit has been linked to high BMA (79). *In vitro* studies have demonstrated that estradiol induces a pro-osteogenic lineage shift in BM stromal cells, whereas estrogen deficit in aging mice was found to induce a decrease in the expression and activity of the pro-osteogenic, anti-adipogenic factor sirtuin type 1 (80, 81). One can, thus, conclude that estrogen is likely to be involved in the regulation of BMA.

Moreover, it is important to note that BM adipose tissue (BMAT) expansion contributes significantly to the elevated serum adiponectin levels and skeletal muscle adaptation observed during CR (27). Indeed, the study by Cawthorn et al. suggested that BMAT is a major source of circulating adiponectin in states of leanness, and that (through endocrine functions) BMAT can have extraskeletal, systemic effects (27). Thus, adiponectin may also regulate BMA in this context.

Last, it is probable that ongoing and/or future research will prompt us to take account of new factors that regulate adiposity in a context of energy deficit.

## PERSPECTIVES

This short overview of the relationship between BMA alterations and the severity of body weight loss highlights the complexity of the underlying processes and the mechanism of subsequent bone loss.

With a view to designing BMA-focused therapeutic strategies against osteoporosis in a context of energy deficit, one major question is "*do BM adipocytes trigger bone loss in a context of energy deficit?*" Several studies of patients with AN have shown that average-to-severe body weight loss tends to lead to normal or decreased BMA. This suggests that (i) the BMA increase observed in patients with mild body weight loss may be a late marker of changes that occur in BM and that result in low bone mass, and (ii) the absence of an increase in BMA does not necessarily mean that the skeleton is healthy. Some studies of CR animal models have shown that bone quality can decrease when the BMA is stable or even when it falls. This finding suggests that (at least in these models) hypotheses based on the effect of factors released by mature adipocytes on the BM microenvironment are not the most realistic. If the latter phenomenon is nevertheless involved, it might have a secondary role in the downregulation of bone physiology. Thus, BM adipocytes might not trigger bone loss in moderate-to-severe AN and in some CR models. Next, hypotheses involving competition between differentiation into osteoblasts and differentiation into adipocytes should be considered. Interestingly, our analysis of the differentiation capabilities of BM stromal cells from SBA and control mice found that the SBA cells displayed a huge increase in adipogenesis at the expense of osteoblastogenesis (unpublished data). This was observed in SBA mice that displayed normal or low BMA, which led us to suppose that the BM of these mice contains many pre-adipocytes that are not able to synthesize and store lipids *in vivo*. These preliminary results strengthen the second hypothesis, although further experiments will be required to confirm these observations and to determine which factors in the culture medium trigger this pre-adipocyte maturation and adipocyte activity. The glucose concentration appears to be a relevant candidate factor. However, favoring hypotheses based on the importance of pre-adipocytes raises another question: "*Do pre-adipocytes influence their microenvironment?*" The characterization of stromal cells at the earliest stage in culture might help to answer this question.

## REFERENCES


If one acknowledges that a shift in the differentiation of BM stromal cells toward the adipocyte lineage is the main BM event during bone mass loss in a context of energy deficit, the next question would be "*how is this commitment regulated?*" One way of addressing this question would be to develop CR models with different levels of body weight loss. One could then compare BMA, bone quality/turnover, concentrations of circulating hormones known to have effects on adipogenesis and osteoblastogenesis, and the BM stromal cells' differentiation capabilities.

Tools are already available for studying how the differentiation of BM stromal cells is reprogrammed in animal models of CR and how differentiation is linked to the associated low bone mass. However, the development of novel tools for the *in vivo* assessment of specific cell types in the BM (such as pre-adipocytes) remains a true challenge.

## AUTHOR CONTRIBUTIONS

OG, NAR, PH, and CC discussed the concept, compiled the literature, and wrote the paper.

## FUNDING

NAR's PhD is founded by ULCO and Lebanon CNRS.


cell differentiation. *J Bone Miner Metab* (2008) 26(3):203–12. doi:10.1007/ s00774-007-0824-2


dietary energy restriction. *J Appl Physiol (1985)* (2012) 113(11):1792–801. doi:10.1152/japplphysiol.00187.2012


hypothalamic amenorrhea and anorexia nervosa. *J Clin Endocrinol Metab* (2009) 94(12):4710–6. doi:10.1210/jc.2009-1046


**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 © 2016 Ghali, Al Rassy, Hardouin and Chauveau. 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.*

# Increased Circulating Adiponectin in Response to Thiazolidinediones: Investigating the Role of Bone Marrow Adipose Tissue

*Richard J. Sulston1 , Brian S. Learman2 , Bofeng Zhang2 , Erica L. Scheller2 , Sebastian D. Parlee2 , Becky R. Simon3 , Hiroyuki Mori2 , Adam J. Bree2 , Robert J. Wallace4 , Venkatesh Krishnan5 , Ormond A. MacDougald2,3,6 and William P. Cawthorn1,2,5\**

*1University/British Heart Foundation Centre for Cardiovascular Science, The Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK, 2Department of Molecular & Integrative Physiology, University of Michigan Medical School, Ann Arbor, MI, USA, 3Program in Cellular and Molecular Biology, University of Michigan Medical School, Ann Arbor, MI, USA, 4Department of Orthopaedics, University of Edinburgh, Edinburgh, UK, 5Musculoskeletal Research, Lilly Research Laboratories, Indianapolis, IN, USA, 6Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA*

#### *Edited by:*

*Basem M. Abdallah, University of Southern Denmark, Denmark*

#### *Reviewed by:*

*Melissa Orlandin Premaor, Universidade Federal de Santa Maria, Brazil Jan Tuckermann, University of Ulm, Germany*

> *\*Correspondence: William P. Cawthorn w.cawthorn@ed.ac.uk*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 14 June 2016 Accepted: 05 September 2016 Published: 21 September 2016*

#### *Citation:*

*Sulston RJ, Learman BS, Zhang B, Scheller EL, Parlee SD, Simon BR, Mori H, Bree AJ, Wallace RJ, Krishnan V, MacDougald OA and Cawthorn WP (2016) Increased Circulating Adiponectin in Response to Thiazolidinediones: Investigating the Role of Bone Marrow Adipose Tissue. Front. Endocrinol. 7:128. doi: 10.3389/fendo.2016.00128*

Background: Bone marrow adipose tissue (MAT) contributes to increased circulating adiponectin, an insulin-sensitizing hormone, during caloric restriction (CR), but whether this occurs in other contexts remains unknown. The antidiabetic thiazolidinediones (TZDs) also promote MAT expansion and hyperadiponectinemia, even without increasing adiponectin expression in white adipose tissue (WAT).

Objectives: To test the hypothesis that MAT expansion contributes to TZD-associated hyperadiponectinemia, we investigated the effects of rosiglitazone, a prototypical TZD, in wild-type (WT) or *Ocn*-Wnt10b mice. The latter resist MAT expansion during CR, leading us to postulate that they would also resist this effect of rosiglitazone.

Design: Male and female WT or *Ocn*-Wnt10b mice (C57BL/6J) were treated with or without rosiglitazone for 2, 4, or 8 weeks, up to 30 weeks of age. MAT content was assessed by osmium tetroxide staining and adipocyte marker expression. Circulating adiponectin was determined by ELISA.

Results: In WT mice, rosiglitazone caused hyperadiponectinemia and MAT expansion. Compared to WT mice, *Ocn*-Wnt10b mice had significantly less MAT in distal tibiae and sometimes in proximal tibiae; however, interpretation was complicated by the leakage of osmium tetroxide from ruptures in some tibiae, highlighting an important technical consideration for osmium-based MAT analysis. Despite decreased MAT in *Ocn-*Wnt10b mice, circulating adiponectin was generally similar between WT and *Ocn*-Wnt10b mice; however, in females receiving rosiglitazone for 4 weeks, hyperadiponectinemia was significantly blunted in *Ocn-*Wnt10b compared to WT mice. Notably, this was also the only group in which tibial adiponectin expression was lower than in WT mice, suggesting a close association between MAT adiponectin production and circulating adiponectin. However, rosiglitazone significantly increased adiponectin protein expression in WAT, suggesting that WAT contributes to hyperadiponectinemia in this context. Finally, rosiglitazone upregulated uncoupling protein 1 in brown adipose tissue (BAT), but this protein was undetectable in tibiae, suggesting that MAT is unlikely to share thermogenic properties of BAT.

Conclusion: TZD-induced hyperadiponectinemia is closely associated with increased adiponectin production in MAT but is not prevented by the partial loss of MAT that occurs in *Ocn-*Wnt10b mice. Thus, more robust loss-of-MAT models are required for future studies to better establish MAT's elusive functions, both on an endocrine level and beyond.

Keywords: bone marrow adipose tissue, white adipose tissue, brown adipose tissue, thiazolidinedione, rosiglitazone, adiponectin, beige adipocyte, UCP1

## INTRODUCTION

Adipose tissue is typically classified into two broad subtypes, white adipose tissue (WAT) and brown adipose tissue (BAT). WAT is commonly known for its role in energy storage and release and is now established as a major endocrine organ. BAT also mediates some endocrine functions but is more widely known for its ability to mediate adaptive thermogenesis (1). Through these functions, both WAT and BAT impact metabolic homeostasis; hence, the global burden of obesity and metabolic disease has motivated extensive study of these tissues (2).

In addition to WAT and BAT, adipocytes also exist in the bone marrow, and such marrow adipose tissue (MAT) has been estimated to account for over 10% of total adipose tissue mass in lean, healthy humans (3). We recently revealed that MAT characteristics are region-specific, such that MAT can be classified into two broad subtypes: regulated MAT (rMAT), which exists in more proximal skeletal sites and consists of adipocytes interspersed with hematopoietic BM; and constitutive MAT (cMAT), which exists in more distal regions (e.g., distal tibia, caudal vertebrae) and appears histologically similar to WAT, with few visible hematopoietic cells (4). These MAT subtypes also differ in their lipid composition and response to external stimuli (4). Both rMAT and cMAT appear to be developmentally and functionally distinct to WAT and BAT, and therefore MAT may represent a third general class of adipose tissue (5, 6). Despite these advances, knowledge of MAT formation and function remains relatively limited (2). However, research from others and us suggests that, like WAT, MAT is an endocrine organ that can exert local and systemic effects (3, 7).

The appreciation of WAT as an endocrine organ derives largely from the discovery in the mid-1990s of two adipocyte-derived hormones, leptin and adiponectin, each of which impacts metabolic homeostasis (8–10). These two hormones have since been mentioned in over 40,000 publications, reflecting the extensive depth of WAT research. Circulating leptin concentrations correlate directly with body fat percentage and are therefore increased in obesity. In contrast, adiponectin concentrations are decreased in obesity and insulin-resistance; hence, hypoadiponectinemia is now an established biomarker for increased risk of cardiometabolic disease. Conversely, circulating adiponectin is elevated in conditions of leanness and insulin sensitivity, such as during caloric restriction (CR) (2). This counterintuitive observation has been dubbed the "adiponectin paradox": why should circulating adiponectin increase when the amount of WAT, the presumed source of adiponectin, is decreased? Moreover, CR can cause hyperadiponectinemia without increasing expression or secretion of adiponectin from WAT (2), suggesting that other tissues contribute to this effect. Our recent research has highlighted a potential explanation for this paradox, revealing, unexpectedly, that MAT is a source of circulating adiponectin during CR (3).

Our studies into MAT and CR were motivated by the striking observation that, in contrast to WAT and BAT, MAT accumulates during CR (3, 11, 12). The biochemical phenotype of this CR-responsive MAT remains to be elucidated; however, we have since shown that these increases occur predominantly in regions of rMAT rather than cMAT (2). This phenomenon led us to investigate if MAT contributes to increased circulating adiponectin during CR. After confirming that MAT expresses and secretes adiponectin (3), we then tested if MAT accumulation is required for CR-associated hyperadiponectinemia. To do so, we used *Ocn*-Wnt10b mice, a transgenic model in which the secreted ligand, Wnt10b, is expressed in osteoblasts from the *Ocn* promoter (13). Because Wnt10b simulates osteoblastogenesis and inhibits adipogenesis, these mice have increased bone formation and decreased bone marrow volume (3, 13). We found that *Ocn-*Wnt10b mice also resist CR-associated MAT expansion, both in rMAT and cMAT (2, 3). Notably, hyperadiponectinemia during CR is also blunted in these mice, despite no differences in adiponectin expression in WAT (3). Finally, impaired MAT expansion also coincides with blunted hyperadiponectinemia in a separate mouse model of CR (14) and during CR in rabbits (12). Together, these observations suggest that MAT expansion is required for CR-induced hyperadiponectinemia, supporting the conclusion that MAT is a source of circulating adiponectin in this context.

To further investigate this endocrine function, we sought to determine if MAT also influences circulating adiponectin beyond CR. For example, increases in both MAT and circulating adiponectin occur in many other conditions, including aging, estrogen deficiency, type 1 diabetes, cancer treatment, and in response to fibroblast growth factor-21 or glucocorticoid therapy (2). Notably, these increases also coincide during treatment with thiazolidinediones (TZDs), a class of insulin-sensitizing, antidiabetic drugs that act as agonists for the nuclear hormone receptor, peroxisome proliferator-activated receptor gamma (PPARγ) (15, 16). Binding and subsequent activation of PPARγ causes it to activate the expression of its transcriptional targets, including adiponectin and other lipid-metabolism-associated genes such as fatty acid-binding protein 4 (*Fabp4*) (17). Thus, TZDs, such as rosiglitazone, significantly increase circulating adiponectin in rodent models and human patients (18, 19). Importantly, preclinical studies suggest that such hyperadiponectinemia is required for the full insulin-sensitizing effects of TZDs (20). Despite the beneficial effects of TZDs in improving glucose tolerance, their clinical use has been restricted owing to increased risk of myocardial infarction and bone fractures (21, 22); the latter may relate to the ability of TZDs to drive MAT expansion. If so, MAT expansion may be detrimental to the clinical utility of these drugs. However, it is notable that TZDs can increase circulating adiponectin without increasing adiponectin expression in WAT (23, 24), and that some studies suggest that TZD action is independent of WAT (25). Thus, given the contribution of MAT to increased circulating adiponectin during CR, we hypothesized that MAT also contributes to TZD-induced hyperadiponectinemia. If so, MAT expansion may play a role in the beneficial insulin-sensitizing effects of TZDs.

Herein, we investigated this hypothesis by feeding wild-type (WT) and *Ocn*-Wnt10b mice a Western diet to induce obesity and glucose intolerance, followed by treatment with rosiglitazone, a prototypical TZD. We postulated that, as in CR, these mice would resist TZD-induced MAT expansion. As secondary analyses, we also investigated the contribution of WAT to TZD-mediated hyperadiponectinemia and the possibility, suggested previously (26), that MAT has BAT-like properties. Our results shed light on MAT's characteristics and highlight technical considerations that will be important for future research into MAT formation and function.

## MATERIALS AND METHODS

## Animals and Animal Care

*Ocn*-Wnt10b mice (Wnt10b) or non-transgenic controls (WT) were on a C57BL/6J background and were bred in-house, as described previously (13). The University of Michigan Committee on the Use and Care of Animals approved all animal experiments, with daily care of mice and rabbits overseen by the Unit for Laboratory Animal Medicine (ULAM).

## Diets and Rosiglitazone (TZD) Treatment

Male WT (*n* = 18), male Wnt10b (*n* = 22), female WT (*n* = 20), and female Wnt10b (*n* = 15) mice were fed a standard laboratory chow diet (Research Diets, D12450B) from weaning until 10 weeks of age. From 10 to 22 weeks of age, all mice were fed a high-fat, high-sucrose Western diet (Research Diets D12079B), with the goal of promoting diet-induced obesity and glucose intolerance. The rationale for this was that, if *Ocn*-Wnt10b mice resisted hyperadiponectinemia, they might also be less responsive to the metabolic effects of TZD treatment. Thus, mice were fed a Western diet prior to TZD treatment, with the aim of allowing detection of metabolic improvements upon TZD administration. At 18 weeks of age, mice were fasted overnight; body mass and fasting glucose were then recorded and body fat, lean mass, and free fluid were measured in conscious mice using an NMR analyzer (Minispec LF90II; Bruker Optics, Billerica, MA, USA). These measurements were then used as a basis to assign mice to four evenly matched groups (with similar body mass, fat mass, and fasting glucose), with each group corresponding to a different duration of rosiglitazone (TZD) treatment. Mice were then treated with or without TZD from 22 to 30 weeks of age. TZD was administered in the diet (Research Diets D12112601) by supplementing diet D12450B with rosiglitazone at 0.175 mg/g diet; based on daily consumption of D12450B, this concentration was estimated to give a final dose per mouse of 15 mg/kg body mass per day. This approach is similar to that used in previous studies investigating rosiglitazone-induced MAT accumulation (26–28). The four experimental groups were as follows: control mice (0 weeks' TZD), which continued to receive Western diet D12079B from 22 to 30 weeks of age; 2 weeks' TZD mice, which received D12079B from 22 to 28 weeks and D12112601 from 28 to 30 weeks; 4 weeks' TZD mice, which received D12079B from 22 to 26 weeks and D12112601 from 26 to 30 weeks; and 8 weeks' TZD mice, which received D12112601 from 22 to 30 weeks. Numbers of mice per group and diet durations are described further in **Table 1**. At 29.5 weeks of age, blood glucose concentrations were recorded after an overnight fast to assess the effects of TZD treatment. At 30 weeks of age, serum was sampled, mice were humanely euthanized, and tissues were isolated for further analysis.

## Blood Collection and Serum Adiponectin Analysis

Blood was sampled from the lateral tail vein of mice using Microvette CB 300 capillary collection tubes (Sarstedt, Newton,


NC, USA). Blood glucose was measured using an Accu-Chek Aviva glucometer. To obtain serum, blood samples were allowed to clot on ice for 2 h before centrifuging at 3,800 RCF for 5 min at 4°C. Serum adiponectin was determined using an ELISA kit (catalog no. MRP300) from R&D Systems (Bio-Techne Ltd., Abingdon, UK) according to the manufacturer's instructions.

## Osmium Tetroxide Staining and **μ**CT Analysis

Tibiae were isolated and, after removal of external soft tissue, fixed in formalin at 4°C. Fixed tibiae were decalcified in 14% EDTA for 14 days and then washed in Sorensen's Phosphate buffer (81 mM KH2PO4, 19 mM Na2HPO4 ⋅ 7H2O, pH 7.4). Decalcified tibiae were stored in Sorensen's Phosphate buffer at 4°C until ready to be stained with osmium tetroxide. To do so, osmium tetroxide solution (2% w/v; Agar Scientific, UK) was diluted 1:1 in Sorensen's Phosphate buffer. Tibiae were then stained in this 1% osmium tetroxide solution for 48 h at room temperature, then washed, and stored in Sorensen's Phosphate buffer at 4°C prior to micro computed tomography (μCT) analysis.

## Micro Computed Tomography Analysis

Layers of four to five stained tibiae were arranged in parallel in 1% agarose in a 30-mL universal tube and mounted in a Skyscan 1172 desktop micro CT (Bruker, Kontich, Belgium). The samples were then scanned through 360° using a step of 0.40° between exposures. A voxel resolution of 12.05 μm was obtained in the scans using the following control settings: 54 kV source voltage, 185 μA source current with an exposure time of 885 ms. A 0.5-mm aluminum filter and two-frame averaging were used to optimize the scan. After scanning, the data were reconstructed using Skyscan software NRecon v1.6.9.4 (Bruker, Kontich, Belgium). The reconstruction thresholding window was optimized to encapsulate the target image. Volumetric analysis was performed using CT Analyser v1.13.5.1 (Bruker, Kontich, Belgium).

## Real-time Quantitative PCR

RNA was extracted from tissue using RNA STAT60 reagent (Tel-Test, Inc.) according to the manufacturer's instructions. Synthesis of cDNA was done using TaqMan reverse transcription reagents (Thermo Fisher Scientific) using 1 μg of RNA template per reaction, as per manufacturer's instructions. Transcript expression was then analyzed by quantitative PCR (qPCR) in 10 μL duplicate reactions using qPCRBIO SyGreen Mix (part number PB20.11; PCR Biosystems, UK) and 1–4 μL of cDNA template. Reactions were loaded into 384-well qPCR plates (part number 72.1985.202; Sarstedt, UK) and run on a Light Cycler 480 (Roche). Transcript expression was calculated based on a cDNA titration loaded on each plate and was presented relative to expression of the housekeeping gene *Ppia.* Primers for *Adipoq* and *Ppia* were described and validated previously (3).

## Immunoblot Analysis

Frozen tissue was processed as described previously (12). The resulting protein lysates were separated by size using gradient (4–12%) polyacrylamide gels (BioRad). Protein was then transferred to Immobilon-FL membrane (Millipore) for 150 min at 350 mA, 4°C, using a Criterion wet-transfer system (BioRad). Post-transfer, the membranes were blocked in 5% milk for 1 h at room temperature, then immunoblotted with primary antibody in 5% bovine serum albumin overnight at 4°C. Membranes were then incubated in 1:15,000-diluted fluorescently labeled secondary antibody (LiCor) for 1 h at room temperature. Signal was detected using the LiCor Odyssey system and band intensities quantified using LiCor Image Studio Lite software. The following primary antibodies were used: rabbit anti-adiponectin antibody (Sigma, A6354-200UL) diluted 1:1,000 in 5% BSA; rabbit antiuncoupling protein 1 (UCP1) antibody (Sigma, U6382) diluted 1:10,000 in 2.5% milk; and rabbit anti-β-actin antibody (Abcam, ab8227) diluted 1:1,000 in 5% BSA.

## Data Presentation and Statistical Analysis

Data are presented as box and whisker plots overlaid with individual data points for each animal. Boxes indicate the 25th and 75th percentiles; whiskers display the range; and horizontal lines in each box represent the median. Group sizes are described in **Table 1**. Statistical analysis was done using GraphPad Prism 6 software, with significant differences assessed by two-way ANOVA using a Tukey or Sidak *post hoc* test for multiple comparisons, as appropriate. A *P*-value <0.05 was considered statistically significant. A significant influence of TZD treatment, across all treatment groups, is indicated by †. For multiple comparisons, asterisks (\*) indicate significant differences between genotypes within each TZD group, while hash signs (#) indicate significance between TZD-treated and non-TZD-treated controls within each genotype.

## RESULTS

## Effects of TZD on WAT Mass and Fasting Glucose Do Not Differ between WT and *Ocn-*Wnt10b Mice

Circulating adiponectin is decreased in states of obesity and glucose intolerance (29), while TZDs modulate both fat mass and glucose homeostasis. Therefore, after treating WT and *Ocn-*Wnt10b mice with or without rosiglitazone for 2, 4, or 8 weeks, we first assessed body mass, fat mass, and fasting glucose, primarily as readouts of TZD action but also as parameters that might influence circulating adiponectin. While body masses of female or male mice did not differ with TZD treatment (**Figures 1A,C**), across both genotypes, TZD was associated with significant loss of gonadal WAT (gWAT), a visceral adipose depot, in both females (*P* = 0.027) and males (*P* = 0.006) (**Figures 1B,D**). Conversely, the mass of inguinal WAT (iWAT), a subcutaneous depot, significantly increased with TZD treatment in males (*P* = 0.002) (**Figure 1D**). Analysis of fasting glucose revealed no effects of TZD in female mice (**Figure 1E**), whereas TZD significantly decreased fasting glucose across WT and *Ocn-*Wnt10b males (**Figure 1F**). These findings are consistent with previous studies demonstrating that rosiglitazone decreases gWAT mass (26, 27), increases subcutaneous WAT (30), and ameliorates hyperglycemia (16);

percentage of total body mass. (E,F) Fasting blood glucose, as recorded at 29.5 weeks of age. For (A–F), a significant influence of TZD treatment, across all treatment groups, is indicated by † (*P* < 0.05), †† (*P* < 0.01) or ††† (*P* < 0.001). Within each genotype, significant differences between untreated controls (0-week TZD) and individual durations of TZD treatment are indicated by ### (*P* < 0.001). Within each TZD treatment group (0-, 2-, 4-, or 8-week TZD), there were no statistically significant differences between WT and *Ocn*-Wnt10b mice.

however, it is unclear why rosiglitazone influenced iWAT mass and fasting glucose in males but not in females. In contrast to these effects of rosiglitazone, body mass, WAT mass, or fasting glucose did not differ between WT and *Ocn*-Wnt10b mice within each TZD treatment group (**Figures 1A–F**). Thus, differences in peripheral adiposity or glucose homeostasis, which can influence circulating adiponectin, did not occur between WT and *Ocn*-Wnt10b mice.

## TZD Increases rMAT and cMAT Volume in WT Mice, and These Effects Are Partially Blunted in *Ocn-*Wnt10b Mice

Having characterized these peripheral metabolic parameters, we next investigated the impact of TZD on MAT expansion and whether this differs between WT and *Ocn-*Wnt10b mice. To do so, we first used an approach based on staining tibiae with osmium tetroxide, which covalently binds to unsaturated lipids within bone marrow and thereby acts as a strong contrast agent for MAT detection by μCT (31). As shown in **Figure 2A**, osmium staining in the proximal diaphysis, corresponding to rMAT, was markedly increased in females of both genotypes following 4- or 8-week TZD treatment. Statistical analyses confirmed that, across all groups of females, TZD significantly influenced the volume of rMAT, as well as that of cMAT and total MAT (*P* < 0.0001 for each) (**Figures 2B,C**). Multiple comparisons were then made to assess the effect of each duration of TZD treatment in each genotype of mice, relative to untreated controls. This revealed that, in female WT mice, TZD treatment for 4 or 8 weeks significantly increased rMAT volume in proximal tibiae and cMAT volume in distal tibiae (**Figures 2A,B**). Thus, total tibial MAT volume was also significantly increased by 4- or 8-week TZD in WT females (**Figure 2C**). Similar effects of TZD were also observed in *Ocn*-Wnt10b females (**Figures 2A–C**), and rMAT volume did not significantly differ between genotypes (**Figure 2B**). However, cMAT volume was markedly lower in *Ocn-*Wnt10b females within each TZD treatment group (**Figure 2B**). Total tibial MAT volume was also significantly decreased in *Ocn*-Wnt10b compared to WT females following 8 weeks' TZD (**Figure 2C**).

Compared to the female mice, osmium staining was less pronounced in the rMAT of TZD-treated males (**Figure 2D**). Nevertheless, statistical analysis across all groups of males revealed a significant influence of TZD on the volumes of rMAT (*P* = 0.0015), cMAT (*P* = 0.0006), and total MAT (*P* = 0.0003) (**Figures 2E,F**). Multiple comparisons further revealed that, in WT males, treatment with TZD for 8 weeks, but not shorter durations, significantly increased volumes of rMAT, cMAT, and total MAT (**Figures 2E,F**). This also occurred for cMAT and total MAT, but not rMAT, in *Ocn-*Wnt10b males (**Figures 2E,F**). As found for female mice, male *Ocn-*Wnt10b mice had significantly decreased cMAT volume compared to their WT counterparts (**Figure 2E**), and this also occurred for total MAT volume in the 0-, 4-, and 8-week TZD groups (**Figure 2F**).

## Cortical Bone Ruptures Significantly Decrease Detectable rMAT Volume

The above observations suggest that *Ocn*-Wnt10b males and females partially resist TZD-induced MAT expansion, predominantly as a result of a lower cMAT volume than WT mice. Indeed, rMAT volume did not differ between genotypes in any of the TZD treatment groups. However, when analyzing the μCT scans, we noticed that the decalcified cortical bone had ruptured in tibiae from all male mice and in a subset of females. The ruptures always occurred in proximal tibial diaphysis, as indicated by representative scans of non-ruptured (intact) and ruptured tibiae (**Figure 3A**). Cross sections show that ruptures allowed escape of the bone marrow, resulting in decreased osmium staining compared to intact tibiae (**Figure 3A**). This loss of signal suggests that the ruptures might have confounded measurement of MAT volume. To address this, we quantified the volumes of cMAT, rMAT, and total MAT in female tibiae, comparing bones from non-TZD-treated mice (all of which were intact) with intact or ruptured bones from TZD-treated mice. The volume of cMAT did not differ between intact and ruptured bones (**Figure 3B**), indicating that these ruptures did not affect detection of cMAT in the distal tibia. However, in TZD-treated WT mice, the detectable volume of rMAT and total MAT was significantly lower in ruptured compared to intact bones (**Figure 3B**). Indeed, in both WT and *Ocn-*Wnt10b mice, TZD-induced expansion of rMAT and total MAT was strongly significant for intact bones, but not when ruptures were present (**Figure 3B**). This impaired rMAT detection likely explains why TZD-induced rMAT expansion was less marked in male than in female mice (**Figure 2B** vs. **Figure 2E**), because all male bones were ruptured. Importantly, for intact tibiae, TZD-induced expansion of rMAT and total MAT was significantly blunted in *Ocn-*Wnt10b compared to WT mice (**Figure 3B**). This suggests that, in addition to decreased cMAT (**Figures 2B,E**), *Ocn-*Wnt10b mice are at least mildly resistant to rMAT expansion during TZD treatment.

## Adipocyte Transcript Expression in Tibiae Confirms TZD-Induced MAT Expansion

Given the confounding influence of these bone ruptures, we next pursued other approaches to further assess tibial MAT content. MAT expansion during CR or rosiglitazone treatment has previously been monitored by qPCR analysis of adipocyte marker expression in intact bones, including transcripts for adiponectin (*Adipoq*) and fatty acid-binding protein 4 (*Fabp4*) (3, 26). Thus, we next used qPCR to assess *Adipoq* and *Fabp4* expression in whole tibiae. As shown in **Figures 4A–D**, across all groups of female or males, expression of *Adipoq* and *Fabp4* was significantly affected by TZD treatment (*P* < 0.0001 for each). Multiple comparisons confirmed that 4- or 8-week TZD significantly increased *Adipoq* in WT males and females (**Figures 4A,C**) and *Fabp4* in males and females of each genotype (**Figures 4B,D**). TZD treatment for 2 weeks was also associated with increased *Adipoq* in *Ocn-*Wnt10b males (**Figure 4C**); increased *Fabp4* in WT females (**Figure 4B**); and increased *Fabp4* in males of each genotype (**Figure 4D**). Transcript expression within each TZD treatment group generally did not differ between WT and *Ocn-*Wnt10b mice, with the exception of decreased *Adipoq* in 4-week TZD females (**Figure 4A**) and decreased *Fabp4* in 8-week TZD males (**Figure 4D**). However, two-way ANOVA revealed that genotype significantly influenced *Fabp4* in females (*P* = 0.0184) and males (*P* = 0.0339), while there was a significant genotype–TZD interaction that influenced *Adipoq* expression in females (*P* = 0.0193) and males (*P* = 0.0297).

Collectively, these results suggest that TZD increases tibial adipocyte content and that there is some level of resistance to this effect in *Ocn-*Wnt10b mice. Thus, together with the osmium

FIGURE 3 | Cortical bone ruptures significantly decrease detectable rMAT volume. All male and some female tibiae had ruptured cortical bone near the proximal end, causing a large reduction in signal. (A) Representative 3D models of intact (non-ruptured) and ruptured tibiae from 8-week-TZD-treated females, with cross-sectional images showing loss of MAT signal (dark regions within bone marrow cavity). Gray squares across the 3D models indicate the slices from which cross-sectional images are taken. Scale bars = 500 μm. (B) Tibiae from female WT or *Ocn*-Wnt10b mice were categorized into bones from mice untreated with TZD, all of which were intact; intact bones from TZD-treated mice; and ruptured bones from TZD-treated mice. The volume of cMAT, rMAT, and total MAT in these bones was determined from μCT scans. Statistically significant differences between untreated (intact), TZD-treated (intact), and TZD-treated (ruptured) samples are indicated by ## (*P* < 0.01) or ### (*P* < 0.001). Within each group, statistically significant differences between WT and *Ocn-*Wnt10b mice are indicated by \* (*P* < 0.05), \*\* (*P* < 0.01), or \*\*\* (*P* < 0.001).

tetroxide analyses of MAT volume (**Figures 2** and **3**), it seems likely that *Ocn-*Wnt10b mice partially resist TZD-induced MAT expansion, but this is not prevented entirely.

## TZD Induces Hyperadiponectinemia to a Similar Extent in WT and *Ocn*-Wnt10b Mice

Given this mild resistance of *Ocn-*Wnt10b mice to MAT expansion, we next investigated if TZD-induced hyperadiponectinemia was altered in these mice. To do so, we used an ELISA to measure serum adiponectin concentrations. Across both genotypes and all TZD groups, treatment with TZD was associated with significantly increased circulating adiponectin (*P* < 0.0001 for males or females) (**Figure 5**). Genotypic differences were not detected in males; however, in females, there was a significant influence of genotype (*P* = 0.0466) and a significant interaction between genotype and TZD treatment (*P* = 0.0248), indicating that effects of TZD differ between WT and *Ocn-*Wnt10b females. Consistent with this, multiple comparisons showed significantly

presented normalized *Ppia* mRNA expression. Statistical significance is presented, as described in Figure 2*,* for the influence of TZD across all groups, and for differences between TZD-treated and untreated mice (within each genotype) or WT and *Ocn*-Wnt10b mice (within each TZD group)*.*

decreased circulating adiponectin in *Ocn-*Wnt10b compared to WT females following 4 weeks' TZD (**Figure 5**). Thus, while TZD-induced hyperadiponectinemia was similar between WT and *Ocn-*Wnt10b males, this effect of TZD was mildly blunted in female *Ocn-*Wnt10b mice.

## TZD Increases Adiponectin Protein Expression in WAT

The above observations demonstrate that, despite having significantly decreased cMAT and being partially resistant to TZD-induced rMAT expansion, *Ocn-*Wnt10b mice undergo

hyperadiponectinemia to a similar extent to their WT counterparts. Thus, factors beyond MAT might have a greater influence over TZD-induced hyperadiponectinemia. To address this possibility, we next investigated the effects of TZD and genotype on adiponectin expression in WAT. In WT and *Ocn-*Wnt10b females, TZD did not influence *Adipoq* expression in iWAT or gWAT, either when comparing individual treatment durations or when effects were assessed across all groups (**Figure 6A**). However, in female iWAT, *Adipoq* transcripts varied significantly by genotype (*P*= 0.0095), with expression diverging as TZD duration increased (**Figure 6A**). Given that Wnt10b transgene expression is restricted to osteoblasts, this genotypic difference in WAT was unexpected. Across all groups of males, TZD treatment had a minor influence on *Adipoq* transcript levels in both iWAT and gWAT (*P* = 0.0269 and *P* = 0.0461, respectively), although multiple comparisons found no significant effects of TZD between individual treatment groups (**Figure 6B**). As for females, male iWAT also exhibited genotypic variation (*P* = 0.0096), although in this case *Adipoq* expression tended to be greater in *Ocn-*Wnt10b than in WT mice (**Figure 6B**).

These findings show that WAT *Adipoq* expression is influenced only modestly by TZD, in contrast to the marked TZD-mediated increases in circulating adiponectin (**Figure 5**). Indeed, our above observations in female mice are consistent with previous reports demonstrating that TZDs can increase circulating adiponectin without increasing *Adipoq* expression in WAT (23, 24). However, other studies have reported a disparity between adiponectin expression at the transcript and protein level (32, 33). Therefore, we next used fluorescence-based immunoblotting to detect and quantify adiponectin protein expression. As shown in **Figure 7A**, across all groups of female mice, there was a significant influence of TZD on adiponectin protein expression in iWAT and gWAT (*P* < 0.0001 for both), with adiponectin typically increased in WAT of TZD-treated mice compared to untreated controls. Multiple comparisons further confirmed that, for females of each genotype, TZD treatment for 2 or 8 weeks significantly increased adiponectin protein in gWAT or iWAT, respectively (**Figure 7A**). As for females, across all groups of male mice, TZD had a significant influence on adiponectin protein in iWAT (*P* = 0.004) and gWAT (*P* = 0.0004) (**Figure 7B**). Based on multiple comparisons, adiponectin was significantly elevated in iWAT and gWAT of WT males after 2 weeks of TZD, and in gWAT of *Ocn*-Wnt10b males after 2 or 4 weeks of TZD (**Figure 7B**). These results demonstrate that rosiglitazone significantly increases expression

of adiponectin protein in WAT of female and male mice. Thus, it seems likely that, with this regimen of rosiglitazone treatment, WAT makes at least a partial contribution to TZD-induced hyperadiponectinemia.

## TZD Induces UCP1 Protein Expression in BAT but Not in WAT or Tibiae

While these studies focused on the relationship between MAT and adiponectin, they also provided the opportunity to assess other properties of MAT; indeed, it remains unclear to what extent MAT's characteristics overlap with those of WAT and BAT. The key function of BAT is to mediate adaptive thermogenesis *via* uncoupled respiration, which is dependent on expression of uncoupling protein 1 (UCP1). TZDs dose-dependently increase UCP1 protein and *Ucp1* transcripts in BAT and can upregulate *Ucp1* in WAT and whole tibiae (26, 34). Based on the latter, it has been suggested that MAT may have BAT-like characteristics (26). However, others have argued that elevated *Ucp1* expression alone, without assessment of UCP1 protein, is insufficient evidence for a tissue's thermogenic capacity (35). Indeed, the relative protein expression of UCP1 between MAT and BAT remains to be firmly established. Thus, to further study the BAT-like properties of MAT, we next analyzed UCP1 expression in BAT, WAT, and tibiae, both at the transcript and protein level.

As shown in **Figures 8A,B**, across all groups of female or male mice, TZD strongly influenced *Ucp1* expression in BAT (females, *P* = 0.0002; males, *P* < 0.0001). Multiple comparisons further confirmed that, relative to untreated controls, each duration of TZD significantly increased BAT *Ucp1* in males of each genotype (**Figure 8B**), while this also occurred for females with 2 weeks' TZD treatment (**Figure 8A**). Across-group effects of TZD were also detected in female gWAT (*P* = 0.014), female tibiae (*P* < 0.0001), male gWAT (*P* = 0.0051), and male tibiae (*P* = 0.0017), wherein TZD-treated groups typically had greater *Ucp1* expression than untreated controls (**Figures 8A,B**). Analysis of genotypic differences *via* multiple comparisons confirmed that, relative to untreated mice, *Ucp1* was significantly elevated in *Ocn-*Wnt10b female tibiae or gWAT after 4 or 8 weeks' TZD, respectively, and in WT male gWAT, after 8 weeks' TZD (**Figures 8A,B**). Each of these groups also had significantly greater *Ucp1* expression than their untreated WT or *Ocn*-Wnt10b counterparts, respectively; however, such genotypic differences were uncommon, and genotype generally did not influence *Ucp1* expression in any of the tissues analyzed.

These data demonstrate that rosiglitazone increases *Ucp1* expression in BAT, gWAT, and tibiae, consistent with results in previous studies (26, 34). However, relative to expression in BAT, *Ucp1* expression in untreated mice was over 500-fold lower in gWAT, 1,500-fold lower in iWAT, and 10,000-fold lower in tibiae (**Figures 8A,B**). Thus, even with TZD treatment, *Ucp1* expression in these tissues is negligible, raising questions over its functional relevance. To further address this, we next analyzed UCP1 expression at the protein level. As shown in **Figure 8C**, in females and males of either genotype, TZD robustly increased UCP1 expression in BAT; however, in gWAT, iWAT, or whole tibiae, we could not detect UCP1 protein expression (data not shown). Together, these observations suggest that MAT is unlikely to possess the thermogenic properties of BAT.

## DISCUSSION

Herein, we pursued studies in *Ocn*-Wnt10b mice to investigate the hypothesis that MAT expansion contributes to TZD-induced hyperadiponectinemia. We found that, regardless of TZD treatment, *Ocn-*Wnt10b mice have significantly less cMAT than WT mice and likely resist TZD-induced rMAT accumulation. Despite this loss of cMAT and mild resistance to rMAT expansion, circulating adiponectin is generally similar between WT and *Ocn-*Wnt10b mice. Moreover, we found that TZD significantly increases adiponectin protein expression in WAT. Together, these findings suggest that WAT, rather than MAT, is the key mediator of TZDinduced hyperadiponectinemia, at least under these conditions of TZD treatment. However, as discussed in Section "Contribution of MAT and WAT to TZD-Induced Hyperadiponectinemia," further consideration of these findings suggests that MAT makes at least some contribution to TZD-induced hyperadiponectinemia. As a secondary aim, we also addressed the hypothesis that MAT has BAT-like properties. We found that, in contrast to BAT, UCP1 protein is undetectable in tibiae, suggesting that MAT is unlikely to have the thermogenic capacity of BAT.

As further discussed below, these findings shed new light on the function of MAT and have implications worth considering for future MAT research.

## Contribution of MAT and WAT to TZD-Induced Hyperadiponectinemia

We previously identified MAT as a source of increased circulating adiponectin during CR, a conclusion based, in part, on the finding that *Ocn-*Wnt10b mice resist both CR-associated MAT expansion and hyperadiponectinemia (2, 3).

Herein, we find that, regardless of TZD treatment, cMAT volume is significantly lower in *Ocn-*Wnt10b than in WT mice. Unfortunately, tibial ruptures confounded measurement of rMAT volume; however, analysis of intact tibiae, without ruptured samples, suggests strongly that *Ocn*-Wnt10b mice also partially resist rMAT accumulation in response to TZD treatment. Despite this loss of cMAT and mild resistance to rMAT expansion, circulating adiponectin is generally similar between WT and *Ocn*-Wnt10b mice. Superficially, this suggests that MAT expansion *per se* does not contribute to hyperadiponectinemia under these TZD treatment conditions. However, this conclusion is less certain when several other key factors are taken into account.

One such factor is the likely contribution of WAT. Unlike during CR, adiponectin protein expression in WAT is increased under these conditions of TZD treatment, which may override any effects of suppressed MAT expansion. These increases are more pronounced for adiponectin protein than for *Adipoq* transcripts, with TZD affecting the *Adipoq* transcripts only in male WAT (**Figure 6**). These findings echo previous reports of sex-specific differences in WAT adiponectin expression, and of a disconnect between adiponectin expression at the transcript and protein level (32, 33). Notably, these effects on WAT are inconsistent with the finding that TZDs can increase circulating adiponectin without increasing adiponectin expression in WAT (23, 24). This disparity is likely a result of the higher dose of rosiglitazone used in our study. Indeed, WAT adiponectin expression is unaltered by lower TZD doses (24, 36) but significantly increased at higher rosiglitazone concentrations (18). Thus, as discussed further below ("Implications of TZD Dose on the Capacity of *Ocn*-Wnt10b Mice to Resist MAT Expansion and Hyperadiponectinemia"), a lower dose of rosiglitazone would likely be required if we were to further investigate the contribution of MAT to TZD-associated hyperadiponectinemia.

Considering these findings, it seems likely that WAT makes at least some contribution to hyperadiponectinemia under these conditions of rosiglitazone treatment. However, the effect of TZD on WAT adiponectin expression (**Figures 6** and **7**) is far less consistent and pronounced than that which occurs for circulating adiponectin concentrations (**Figure 5**). In contrast, these changes in circulating adiponectin are closely reflected by the increases in tibial adiponectin expression (**Figure 4**). For example, when comparing *Ocn-*Wnt10b and WT mice, circulating adiponectin is lower only in the 4-week TZD females (**Figure 5**); hence, it is striking that this is also the only group in which tibial adiponectin expression is lower in *Ocn-*Wnt10b mice (**Figure 4A**). Together, these observations show that circulating adiponectin is closely associated with tibial adiponectin expression and less tightly associated with gross changes in rMAT and cMAT volume. This underscores the importance of analyzing MAT through multiple approaches, including transcriptional markers, rather than focusing solely on MAT volume *via* osmium tetroxide staining. Perhaps, more importantly, this close association suggests that adiponectin production from MAT contributes, at least in part, to TZD-induced hyperadiponectinemia.

## Implications of TZD Dose on the Capacity of *Ocn*-Wnt10b Mice to Resist MAT Expansion and Hyperadiponectinemia

Previous studies have used a wide range of doses of rosiglitazone to investigate its impact on adiponectin or bone marrow adiposity. For example, Nawrocki et al. treated mice with 10 mg/ kg/day to examine the effects on circulating adiponectin and glucose homeostasis (20), while 20 mg/kg/day was used in two more recent studies investigating effects of rosiglitazone on bone marrow adiposity (26, 27). A much lower dose of 3 mg/kg/day was also found to cause MAT expansion in mice, but whether this promotes hyperadiponectinemia was not reported (28, 37). Thus, in the present study, we used 15 mg/kg/day to ensure robust effects on MAT and circulating adiponectin. As in the above studies (26–28), we did so by administering rosiglitazone in the diet, based on measurements of body mass and estimated daily food intake. However, because of variation in body masses and daily food consumption, some mice may have exceeded the target dose of 15 mg/kg/day. As discussed above, this relatively high dose may explain the increased adiponectin expression in WAT, thereby complicating interpretation of any contribution from MAT. Such high doses may also have overwhelmed the ability of *Ocn*-Wnt10b mice to more robustly resist MAT expansion. Thus, a lower rosiglitazone dose may limit these confounding effects and thereby be more suitable for addressing our hypothesis in *Ocn-*Wnt10b mice.

## Limitations of *Ocn-*Wnt10b Mice as a Model Resistant to MAT Expansion

One reason that we have not pursued such follow-up studies is that *Ocn*-Wnt10b mice may be too limited a model in which to robustly address MAT's endocrine functions. For example, while CR- or TZD-induced MAT expansion is blunted in these mice, such expansion is still marked in comparison to untreated *Ocn*-Wnt10b controls (**Figures 2** and **3**) (3). Moreover, distal tibiae of *Ocn-*Wnt10b mice remain laden with MAT (**Figures 2A,D**) and, while cMAT volume is lower in *Ocn*-Wnt10b mice, this likely reflects restrictions imposed by decreased bone marrow volume (3), rather than resulting from a direct inhibition of adipogenesis. Thus, as we have recently argued elsewhere (7), future research of MAT would benefit enormously from development of new mouse models that more robustly resist MAT formation, developmentally and/or in response to CR, TZDs, or other stimuli that promote MAT expansion.

## Confounding Effects of Bone Ruptures on Osmium Tetroxide-Based rMAT Analysis

The present study reveals that cortical bone ruptures can confound osmium-based MAT detection. This issue, which has not been reported previously, provides further support for taking a multifaceted approach to MAT analysis. While TZDs can cause bone loss and fractures (38), it is unlikely that the ruptures occurred *in vivo* over the course of rosiglitazone treatment: such an injury would have dramatically impacted mobility and behavior of the mice, which was not apparent during daily inspections; and the ruptures also occurred in tibiae of males untreated with rosiglitazone, demonstrating that they are not a consequence of TZD treatment. Instead, it is likely that ruptures occurred *ex vivo* as a result of methodological issues. In particular, after fixing tibiae in formalin post-necropsy there was a very long time span, ~30 months, before most of these bones were decalcified and analyzed by osmium tetroxide staining. This holdup largely resulted from delays associated with one of our lead authors moving to a new institution, during which time the fixed tibiae were stored in PBS at 4°C and shipped by air from the USA to the UK. Although these storage conditions would not be expected to promote cortical ruptures, it is notable that no ruptures occurred in a subset of female tibiae that were stored in PBS for a shorter duration (~12 months), without air transport, before decalcification and osmium analysis. Moreover, at both the University of Michigan and the University of Edinburgh, we have analyzed hundreds of osmium-stained mouse bones from other studies, none of which has undergone such long-term storage or air transport, and among which no ruptures have ever been detected. Thus, one possibility is that exposure to low air pressure during air transport can cause a pressure differential that promotes bone ruptures. It remains possible that other factors also contributed to this phenomenon; however, our findings strongly suggest that fixed bones should not undergo long-term storage or exposure to low atmospheric pressure if MAT volume is to be determined accurately using osmium tetroxide staining.

## BAT-Like Thermogenic Function Is Unlikely in MAT

There is extensive interest in "beige" or "brite" (brown-in-white) adipocytes, which develop in WAT in response to diverse external stimuli, including cold exposure or TZD treatment, and which share some properties of brown adipocytes (39–41). This socalled "browning" of WAT can increase energy consumption and may thereby contribute to TZD-associated improvement in metabolic health. Therefore, whether MAT can also undergo browning or has BAT-like characteristics has been a subject of some interest. Krings et al. found that, in mice treated for 4 weeks with rosiglitazone (20 mg/kg/day), transcript expression of brown adipocyte markers is increased in tibiae, supporting the possibility that MAT has BAT-like characteristics (26). This is consistent with our finding that rosiglitazone influences tibial *Ucp1* mRNA expression (**Figure 8**). However, Krings et al. further highlighted that, compared to BAT, *Ucp1* expression is 16,000-fold lower in tibiae, which agrees closely with our data showing a 10,000- to 25,000-fold decrease in tibiae compared to BAT. In addition, even with rosiglitazone treatment, we find that *Ucp1* transcripts remain 400- to 7,000-fold lower in tibiae than in BAT (**Figures 8A,B**). Crucially, at this level of *Ucp1* expression, we could not detect UCP1 protein, whereas rosiglitazone-induced increases in UCP1 protein in BAT are readily detectable (**Figure 8C**).

UCP1 is the hallmark of brown adipocytes and the key mediator of adaptive thermogenesis, and therefore the above findings strongly suggest that bone marrow adipocytes are unlikely to possess BAT-like thermogenic properties. One limitation of our study, and that of Krings et al.*,* is that we analyzed whole bones rather than isolated marrow adipocytes. However, both brown and beige adipocytes are further defined by their high mitochondrial content and the multilocular morphology of their lipid droplets (41), neither of which is observed in bone marrow adipocytes (3, 4, 12, 42). This raises further doubts about the notion that MAT has BAT-like characteristics, an issue that we recently discussed in greater depth elsewhere (5). Nevertheless, it remains possible that, under a different TZD treatment regimen or in response to other stimuli, MAT is capable of undergoing browning. Indeed, while we were unable to detect UCP1 protein in WAT or whole tibiae, previous studies have detected UCP1 in WAT following TZD treatment (41). Moreover, effects of TZDs on *Ucp1* expression in BAT are dose-dependent (34). This supports the possibility that TZD-mediated browning of WAT and MAT requires an optimal regimen of TZD treatment, in which case the conditions in our study may simply have been unsuitable for inducing a browning response. Thus, there is ample scope for future studies to further investigate whether MAT has BAT-like properties, both at the molecular and functional levels.

## Other Strengths and Limitations of This Study

This study is the first to address the novel hypothesis that MAT contributes to hyperadiponectinemia in conditions beyond CR. The experimental design is strengthened by including both male and female mice, as well as by analysis of several durations of TZD treatment. The discovery of the confounding effects of cortical ruptures is partly a strength, in that it highlights an important technical consideration for future MAT research; however, presently this was a limitation because it prevented thorough measurement of rMAT volume. As discussed above, another limitation is that the dose of rosiglitazone may have been too high, and may have varied slightly between mice because of differences in body mass and daily food intake. Administration of lower doses by daily injection or oral gavage could be one approach to overcome this issue. Finally, a key limitation of this study is that our findings are based on only a single cohort of mice. Ideally, these experiments would be repeated in a second cohort, with particular care taken to avoid cortical ruptures; this would allow more robust determination of rMAT volume in *Ocn-*Wnt10b mice. However, as noted above ("Limitations of *Ocn*-Wnt10b Mice as a Model Resistant to MAT Expansion")*, Ocn-*Wnt10b mice may not be sufficiently robust as a model of impaired MAT formation. For example, adipocyte marker expression in tibiae is generally similar between WT and *Ocn*-Wnt10b mice (**Figure 4**). This limitation of the *Ocn-*Wnt10b model undermines the rationale for repeating the present studies in a second cohort. Instead, we feel strongly that it would be more productive and scientifically beneficial to focus efforts on developing new mouse models that more robustly resist MAT expansion. Such models could then be used to better address the present hypothesis, as well as other aspects of MAT function.

Beyond these particular issues, one broader limitation is that our experiments were based only in mice; it remains unclear if the relationship between TZD-induced MAT expansion and hyperadiponectinemia also exists in humans. It is well established that TZDs increase circulating adiponectin in humans (16), but their effects on MAT are less clear. Indeed, one study finds that rosiglitazone *decreases* bone marrow adiposity in the lumbar vertebrae (43), whereas a more recent report finds that pioglitazone, another TZD, increases femoral and lumbar vertebral bone marrow adiposity (44). Unfortunately, neither of these studies assessed circulating adiponectin, and there remain very few clinical studies assessing the effect of TZDs on MAT. Thus, there is a strong rationale for future research to establish how TZDs affect MAT accumulation in humans and whether this is associated with changes in circulating adiponectin.

## CONCLUSION

There are four main conclusions from our study. First, under these TZD treatment conditions, it is likely that both WAT and MAT make some contribution to TZD-induced hyperadiponectinemia. Second, UCP1 expression in MAT is negligible, and therefore MAT is unlikely to have the thermogenic capacity of BAT. Third, cortical bone ruptures confound measurement of MAT volume by osmium tetroxide staining, and therefore care must be taken when processing bones for such analysis. Moreover, osmiumbased assessment of MAT volume can differ from measurement of MAT content based on expression of bone marrow adipocyte transcripts; hence, multiple approaches should be used to assess MAT content, including analysis of molecular markers, rather than relying solely on osmium tetroxide staining. Finally, while *Ocn-*Wnt10b mice have been useful as a model resistant to MAT formation, development of more robust loss-of-MAT models would be of great benefit to future MAT research.

## AUTHOR CONTRIBUTIONS

RS and WC designed all figures and wrote the manuscript, with ES, HM, and OM providing additional critical revisions. WC, BL, ES, VK, and OM contributed to the conception and design of the experiments. BL oversaw mouse breeding and colony management. RS, BL, BZ, ES, SP, BS, HM, AB, RW, OM, and WC contributed to data acquisition, analysis, and/or interpretation. All authors gave final approval for publication of the manuscript and agree to be accountable for all aspects of the work presented herein.

## FUNDING

This work was supported by grants from the National Institutes of Health (R24 DK092759 to OM; K99-DE024178 to ES; R25 DK088752 to BZ; S10-RR026475-01 to the University of Michigan School of Dentistry microCT Core; and P30 DK089503 to the Michigan Nutrition Obesity Research Center, which oversaw NMR analysis of mouse body composition and provided a Pilot/Feasibility grant to HM). RS is supported by a British Heart Foundation 4-year PhD Studentship. WC is supported by a Career Development Award (MR/M021394/1) from the Medical Research Council (UK) and by a Chancellor's Fellowship from the University of Edinburgh, and previously by a Lilly Innovation Fellowship Award and a Postdoctoral Research Fellowship from the Royal Commission for the Exhibition of 1851 (UK). SP and BS were supported by a Training Grant from the University of Michigan Training Program in Organogenesis (T32-HD007505). Analysis of microCT imaging data was supported by a Bioinformatics Award provided through a British Heart Foundation Centre of Research Excellence grant.

## REFERENCES


reduced responsiveness to peroxisome proliferator-activated receptor gamma agonists. *J Biol Chem* (2006) 281(5):2654–60. doi:10.1074/jbc.M505311200


reveals a population of thermogenically competent, UCP1-containing adipocytes molecularly distinct from classic brown adipocytes. *J Biol Chem* (2010) 285(10):7153–64. doi:10.1074/jbc.M109.053942


controlled trial. *Eur J Endocrinol* (2012) 166(6):1087–91. doi:10.1530/ EJE-11-1075

**Conflict of Interest Statement:** RS, BL, BZ, ES, SP, BS, HM, AB, and RW have nothing to disclose. WC held a postdoctoral fellowship funded by Eli Lilly and Company. VK is employed by Eli Lilly and Company. OM has received research funding from Eli Lilly and Company.

*Copyright © 2016 Sulston, Learman, Zhang, Scheller, Parlee, Simon, Mori, Bree, Wallace, Krishnan, MacDougald and Cawthorn. 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.*

# Bone Marrow Adipose Tissue: A new Player in Cancer Metastasis to Bone

*Emma V. Morris1 and Claire M. Edwards1,2\**

*1Nuffield Department of Surgical Sciences, University of Oxford, Oxford, UK, 2Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, UK*

The bone marrow is a favored site for a number of cancers, including the hematological malignancy multiple myeloma, and metastasis of breast and prostate cancer. This specialized microenvironment is highly supportive, not only for tumor growth and survival but also for the development of an associated destructive cancer-induced bone disease. The interactions between tumor cells, osteoclasts and osteoblasts are well documented. By contrast, despite occupying a significant proportion of the bone marrow, the importance of bone marrow adipose tissue is only just emerging. The ability of bone marrow adipocytes to regulate skeletal biology and hematopoiesis, combined with their metabolic activity, endocrine functions, and proximity to tumor cells means that they are ideally placed to impact both tumor growth and bone disease. This review discusses the recent advances in our understanding of how marrow adipose tissue contributes to bone metastasis and cancer-induced bone disease.

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Melissa Orlandin Premaor, Federal University of Santa Maria, Brazil Michaela Ruth Reagan, Maine Medical Center Research Institute, USA*

#### *\*Correspondence:*

*Claire M. Edwards claire.edwards@ndorms.ox.ac.uk*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 26 May 2016 Accepted: 29 June 2016 Published: 14 July 2016*

#### *Citation:*

*Morris EV and Edwards CM (2016) Bone Marrow Adipose Tissue: A New Player in Cancer Metastasis to Bone. Front. Endocrinol. 7:90. doi: 10.3389/fendo.2016.00090*

Keywords: cancer-induced bone disease, marrow adipose tissue, multiple myeloma, prostate cancer, bone metastasis

## INTRODUCTION

For the majority of cancers, surgical removal of an isolated primary tumor can be curative. However, once tumor cells establish residence in other organs patient mortality is markedly increased. Once detached from the primary site, a single or cluster of tumor cells can circulate the body and establish secondary lesions in distant sites. Metastasis is a highly inefficient process with <0.1% of disseminating cells surviving to form secondary lesions (1). This implies that healthy tissues display a level of hostility toward invading tumor cells, and so in order to survive and repopulate new sites these cells need to have adapted to seek out more permissive environments to occupy. Interestingly, it has long been recognized that each type of tumor has a distinct pattern of dissemination. In 1889, Stephen Paget published a paper proposing that disseminating cancer cells or "seeds" would only colonize secondary sites or "soils" that were compatible with their growth (2). This theory was challenged in the 1920s by James Ewing who proposed that the colonization of secondary sites could be explained purely by circulatory patterns between the primary and secondary sites (3). In fact, these theories are not mutually exclusive, as there is evidence to support both (1). Systemic breast cancer commonly metastasizes to the bones, lungs, liver, and brain most of which do not have a direct circulatory link to breast tissue. Advanced prostate cancer also selectively metastasizes to bone. By contrast, patients suffering from colon cancer often develop initial metastases in the liver (4). Therefore, mechanical factors that govern the amount of cancer cells delivered to an organ, and compatibility factors such as whether the organ preferentially supports the growth of a specific tumor cell type, contribute to cancer spread (5). Once a cancer cell has successfully seeded itself, it is then dependent on the molecular interactions between it and the environment of the new organ. The more permissive the environment, the more likely the tumor cell is able to take advantage and thrive.

## THE BONE MICROENVIRONMENT

Bone or bone marrow is a major target organ for metastasis, providing a fertile "soil" for circulating tumor cells to settle and repopulate. It is the preferential site of a number of solid tumor metastases including breast and prostate, of which 70–80% of advanced cases have bone involvement (6, 7). It is also the site of the fatal hematological malignancy, multiple myeloma. The bone consists of a number of different cell types, including endothelial cells and nerve cells, cells of hematopoietic origin, such as hematopoietic stem cells, osteoclasts macrophages, and lymphocytes, as well as cells from mesenchymal origin, such as chondrocytes, osteoblasts, and adipocytes. Together these cells provide a supportive niche via cellular cross-talk that maintains healthy bone. It is thought that cancer cells are able to hijack this support network to transform what was initially a beneficial niche into a deadly one. Considerable research efforts have been made, attempting to understand the complexity of the bone microenvironment and defining the exact contribution of each cell type in tumor growth and metastasis. Mundy and colleagues greatly broadened our understanding of this by introducing the concept of the "vicious-cycle" involving bi-directional interactions of tumor cells and bone cells, resulting in osteolysis and in turn, tumor growth. Breast cancer cells produce parathyroid hormonerelated peptide (PTHrP) that activates osteoblasts to produce receptor activator of nuclear factor kappa-B ligand (RANKL) and downregulate osteoprotegerin (OPG). This in turn activates osteoclast precursors, leading to osteolysis. The breakdown of the bone matrix releases bone-derived growth factors, such as transforming growth factor-β and insulin-like growth factor 1, and raises extracellular calcium concentrations. The growth factors bind to receptors on the cell surface of the tumor cells and activate SMAD and MAPK signaling, extracellular calcium binds and activates calcium pumps leading to tumor cell proliferation and the production of PTHrP, thereby causing a vicious cycle of events that result in osteolytic lesions and the progression of cancer metastasis (8–10). This theory explained how bone cells and cancer cells could have a reciprocal relationship. However, it does not take into consideration the contribution of the other resident cells of the bone marrow, such as the cells of the immune system, which have also been shown to play an important part in cancer progression (11, 12). In recent years, the contribution of bone marrow adipocytes (BMAs) has also come into question. Cancer primarily occurs in older people whose bone marrow is highly populated by adipocytes. In early childhood, the bone marrow exists in a predominantly red/hematopoietic osteogenic state (13) and there are few if any detectable adipocytes, however, by the age of 25, ~70% of bone marrow volume is filled with adipose tissue (14). This constitutes around 2 kg of marrow adipose tissue in a healthy adult that represents over 10% of total adipose mass (15). Longitudinal analysis of patients from birth to 90 years have demonstrated that even after this initial exponential accumulation, there continues to be a gradual increase in adipocyte number throughout adult life (16). These cells were previously thought to be inert space filling bystanders (17). However, recently that way of thinking has changed and it is becoming clear that adipocytes have a definite and significant contribution to the bone microenvironment and, thus, the establishment of metastatic disease.

## BONE MARROW ADIPOCYTES

Over the last century, our knowledge and understanding of the role of BMAs has been slowly growing. At the beginning of the nineteenth century, it was determined that the skeleton is filled with areas consisting of yellow or red marrow that are distributed in a defined pattern (18). A number of decades later, it was shown using elegant ultrastructural studies that BMAs develop from a unique progenitor when compared to white adipocytes (19), and that there are two distinct populations of BMAs that respond differently to hematopoietic demands (20). This phenomenon was also shown by Scheller and colleagues that marrow adipose tissue may exist in two isoforms, "constitutive" that forms early on in development in sites such as the distal tibia, and "regulated" that tends to develop with aging and is located interspersed with hematopoietic cells at sites such as the proximal femur and lumbar vertebrae (21). The fact that marrow adipose tissue develops in a conserved, spatial, and temporal manner implies that it is the product of a defined developmental event; suggesting that it has a distinct physiological function. Its phenotype resembles both brown and white adipose tissue; and unlike white adipose tissue, it is not strictly associated with BMI or body fat. It is known to be increased in patients suffering from anorexia nervosa, despite their lean appearance (22). It is also elevated in young mice in response to calorie restriction (23). The main functions of bone marrow adipose tissue are to serve as an energy reservoir and secrete fatty acids, cytokines, and adipokines. BMAs store energy in the form of lipid and release triglycerides and fatty acids in response to energy demands. They are smaller than their visceral counterparts; however, due to enhanced triacylglycerol synthesis the net effect of fatty acid uptake is similar (24). BMAs are both an endocrine target and have endocrine-like functions: responding to growth hormones insulin and thyroid hormone, as well as releasing cytokines such as IL-6, IL-1β, and TNF-α (25). They also secrete adipokines, among them leptin and adiponectin, which regulate calorie uptake and insulin sensitivity, respectively. They can potentially influence neighboring cells via autocrine, paracrine, and endocrine signaling (26) making them an influential component of the bone microenvironment. BMAs arise from multipotent mesenchymal progenitors; these cells have the capacity to differentiate into several cell types, including myocytes, chondrocytes, osteoblasts, and adipocytes (27, 28). It is thought that downstream of these multipotent progenitors, bipotent osteoblast–adipocytes progenitors form an intermediate in the process of cell commitment to these two cell lineages (29). These bipotent cells are stimulated to commit to either cell lineage by the presence of adipogenic vs. osteogenic factors within the bone microenvironment that activate their respective transcriptional programs (30, 31). This creates an inverse reciprocal relationship between osteoblastogenesis and adipogenesis, with factors that promote one of these processes usually inhibiting the other. Many Morris and Edwards Marrow Adipocytes and Bone Metastasis

have likened it to a see-saw, when one process goes up the other often comes down and vice versa (32). It has also been suggested that mature osteoblasts and adipocytes are able to dedifferentiate and transdifferentiate from one phenotype to the other (33). Work by Martin and colleagues also demonstrated that BMAs are able to influence their environment by secreting extracellular vesicles containing adipogenic mRNA transcripts. These could be taken up and expressed by neighboring osteoblasts, thereby weakening their lineage commitment (34, 35). The net result of this close relationship is that if the balance tips in flavor of adipogenesis, osteoblast number may be reduced, resulting in a decrease in bone, so compromising bone strength. BMAs are not only negative regulators of bone formation but also serve to inhibit hematopoiesis. The number of BMAs has been shown to be inversely correlated with the hematopoietic activity of the marrow (36). The evidence to support the notion that BMAs play a metabolic role in the bone is quite clear; however, the influence they have in the context of cancer development and progression is still poorly understood.

## WHITE ADIPOSE TISSUE; A GROWING CONTRIBUTOR TO CANCER RISK AND PROGRESSION

White adipose tissue (WAT) was the first adipose tissue type to be heavily studied in relation to cancer development and progression. In 2002, the International Agency for Research on Cancer (IARC) conducted an evaluation on whether there was a link between weight and cancer. They concluded that the risk of developing some cancers was increased with weight gain (37). Since then, there have been numerous studies investigating the association between increased adiposity and cancer, suggesting that obesity is associated with an increased risk of developing a number of different tumor types, such as colon, breast, and endometrial cancer (38–40). As adipose tissue expands, adipocytes enlarge to store excess energy intake. This causes an increased production of a number of different adipokines and inflammatory cytokines coupled with a decrease in adiponectin, as well as a diminished ability of adipocytes to store surplus-free fatty acids. These changes are associated with dysfunctional WAT that often leads to insulin resistance. In turn, the suppression of lipolysis by insulin is inhibited in insulin resistance, resulting in an increased release of free fatty acids, thereby setting up a vicious cycle of events (40). Lipids are crucial for malignant tumors as they are necessary for the synthesis of membrane constituents; they are also an effective bioenergetic source when metabolic demands are high. In 2011, Nieman et al. demonstrated that adipocytes–ovarian cancer cell co-culture led to the direct transfer of lipids from adipocytes to ovarian cancer cells and promoted *in vitro* and *in vivo* tumor growth. Furthermore, co-culture induced lipolysis in adipocytes and β-oxidation in cancer cells, suggesting that the adipocytes act as an energy source for the cancer cells (41). Adipocytes were also shown to promote the direct migration of prostate cancer cells (42), as well as promote colon cancer proliferation (43). Similar findings have been reported in breast cancer studies, which demonstrate that adipocytes located close to invasive cancer cells are essential for breast tumor development and progression (44). Moreover, adipocytes promote drug resistance in HER2 positive breast cancer cells, suggesting that they bestow a level of protection upon tumor cells (45). Dysfunctional WAT appears to have a tumor-promoting role; it may be that cancer cells that have been in close proximity to adipocytes are primed to settle in an adipocyte-rich secondary site, making the bone a hospitable and permissive environment (**Figure 1**).

## BONE MARROW ADIPOCYTES; FUELING CANCER PROGRESSION

Once WAT had been identified as a driver of cancer progression, the contribution of adipocytes from different anatomical sites came into question. There are now a number of publications highlighting the importance of BMAs as a lipid source that can be utilized by cancer cells to promote proliferation, migration, and invasion (46, 47). A key factor implicated in Nieman's 2011 study was fatty acid binding protein 4 (FABP4); a lipid chaperone that mediates lipid trafficking and transfer of free fatty acids, predominantly expressed in adipocytes, macrophages, and endothelial cells (41). Herroon et al. took these findings further using a mouse model of diet-induced marrow adiposity and demonstrated that FABP4 along with interleukin 1β and its target gene, oxidative stress protein, heme oxygenase 1 (HMOX-1) was also upregulated in prostate cancer cells that were in direct contact with BMAs (46). Taken together, these studies support the fact that cancer cells are able to utilize marrow adipocyte-supplied lipids to thrive in skeletal sites. They also open up a number of questions as to whether BMAs are utilized in the same manner as WAT or whether they offer different environmental advantages to some cancer types over others.

## ADIPOKINES/CYTOKINES

Similarly to WAT, BMAs are not only an effective source of energy they also secrete a plethora of bioactive substances, such as IL-1β, IL-6, leptin, adiponectin, VCAM-1, TNF-α, and VEGF (25). As well as being important in the context of maintaining healthy bone, these secreted cytokines, adipokines, and growth factors can also influence cancer cell behavior and survival. Increased IL-1β secretion coupled with increased leptin expression was shown to recruit breast cancer cells to colonize in bone marrow adipose tissue (47). IL-6, TNF-α, CXCL12, and leptin were shown to promote cell proliferation and migration, as well as inhibit apoptosis and activate autophagy to promote chemotherapy resistance in multiple myeloma (25, 48–51). In prostate cancer, the chemokines, CXCL1 and CXCL2, have been implicated in the progression of associated bone disease by activating osteoclastogenesis and thus promoting tumor cell survival (52). Similarly, in melanoma increased IL-6 triggered an increase in osteoclastogenesis resulting in tumor cell proliferation (53). BMAs clearly influence cancer cell establishment and progression in bone. However, they also play a known tumor-suppressive role. BMAs are the largest source of circulating adiponectin in the body, far more than white adipose tissue (54). Adiponectin has been shown

to exert anti-tumor effects (55, 56), and its levels are inversely correlated with a number of different cancers, such as myeloma, breast, prostate, colon, and endometrial cancer (57, 58), and are associated with poor prognosis (59). Interestingly, unlike other adipokines, such as leptin, circulating adiponectin levels are also decreased in obesity, which appears to be somewhat paradoxical as the numbers of adipocytes are increased. Reduced circulating adiponectin in obesity is likely derived from dysfunctional adipocytes that have decreased adiponectin expression and secretion, which may be the result of mitochondrial dysfunction or increased inflammation, hypoxia or endoplasmic reticulum stress due to the challenge of a high fat diet (60). These findings trigger the question as to what causes the downregulation of adiponectin expression in cancer? Is it a consequence of cancer progression or a cause? Do cancer cells secrete factors that directly regulate adiponectin levels? Are adiponectin levels compromised before cancer development by conditions, such as obesity, that allow for a more permissive environment? More research needs to be done to address these questions and to validate whether adiponectin signaling may be a future therapeutic target.

## TARGETING BONE MARROW ADIPOCYTES

Given the mounting evidence to support the negative role of BMAs in cancer progression, as is further discussed in the context of myeloma by Falyank et al. (61), targeting these cells either alone or in combination with conventional therapeutics may be a promising approach. Due to the ability of cancer cells to induce lipidation of BMAs to support their energy expenditure, drugs are now being developed that target essential molecules of fatty acid synthesis and uptake. Chemical or RNAi mediated inhibition of fatty acid synthase, acetyl-CoA-carboxylase or ATP-citrate lyase was shown to attenuate cancer cell proliferation and cell survival in both *in vitro* and *in vivo* models (62–64). Another approach is to regulate the balance between osteogenesis and adipogenesis by targeting PPARγ or glucocorticoid receptor signaling, thereby preventing an increase in marrow adiposity (65). The modulation of adipokines although still in its infancy has also shown encouraging effects. Leptin receptor antagonists were shown to inhibit tumor cell growth in a model of triple-negative breast cancer (66). Enhancement of circulating adiponectin levels by the apolipoprotein mimetic L-4F caused cancer cell death in a mouse model of multiple myeloma (55). Furthermore, a peptide-based adiponectin receptor agonist ADP 355 was shown to suppress the growth of orthotopic human breast cancer xenografts by ~31% (67). As we begin to gain more knowledge of the exact contribution of BMAs to disease progression, targeting strategies will inevitably become clearer. First, though we need to understand whether BMAs are simply power stations that fuel an inevitable process? If they were removed from the equation, would the cancer cells still home to the bone and thrive? Or are BMAs an integral part of the metastatic process, fertilizing the soil so circulating cancer cells can settle and progress. Interestingly, marrow adipose tissue unlike white adipose tissue increases in times of starvation. Calorie restriction has been show to not only increase marrow adiposity but also extend lifespan in a number of diverse species, such as rodents and primates (68, 69), and is associated with a decrease in cancer risk (70). These findings suggest that BMAs have a protective function that is contradictory to the evidence to suggest that they promote cancer progression. However, one question that remains unresolved is whether the BMAs generated

## REFERENCES


by weight gain and sedentary behavior are the same as those generated by calorie restriction? It may be that BMAs generated by calorie restriction are more metabolically aware and, therefore, less responsive to cancer cell manipulation. However more research needs to be done to explore these possibilities. Targeting BMAs as part of a combination therapy may prove to be a valuable tool; however, first there needs to be a greater understanding of the balance between the tumor-promoting and tumor-suppressive effects of these cells.

## CONCLUSION

As the field of marrow adiposity advances and we come to fully appreciate and understand the influence BMAs have on disease progression, drugs that target these cells may start to come into their own and open up new therapeutic avenues. These advances will hopefully move us one step closer to successfully treating lethal metastatic cancer.

## AUTHOR CONTRIBUTIONS

All authors listed have made substantial, direct, and intellectual contribution to the work and approved it for publication.

## FUNDING

This work was supported by Bloodwise, the International Myeloma Foundation and a Marie Curie Career Integration Grant from the European Union Seventh Framework Programme.


receptor gamma (PPARgamma) by rosiglitazone suppresses components of the insulin-like growth factor regulatory system in vitro and in vivo. *Endocrinology* (2007) 148(2):903–11. doi:10.1210/en.2006-1121


70. Hursting SD, Smith SM, Lashinger LM, Harvey AE, Perkins SN. Calories and carcinogenesis: lessons learned from 30 years of calorie restriction research. *Carcinogenesis* (2010) 31(1):83–9. doi:10.1093/carcin/bgp280

**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 © 2016 Morris and Edwards. 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.*

# Signaling Interplay between Bone Marrow Adipose Tissue and Multiple Myeloma cells

#### *Carolyne Falank1 , Heather Fairfield1 and Michaela R. Reagan1,2,3 \**

*1Reagan Laboratory, Maine Medical Center Research Institute, Scarborough, ME, USA, 2School of Biomedical Sciences and Engineering, University of Maine, Orono, ME, USA, 3School of Medicine, Tufts University, Boston, MA, USA*

In the year 2000, Hanahan and Weinberg (1) defined the six Hallmarks of Cancer as: self-sufficiency in growth signals, evasion of apoptosis, insensitivity to antigrowth mechanisms, tissue invasion and metastasis, limitless replicative potential, and sustained angiogenesis. Eleven years later, two new Hallmarks were added to the list (avoiding immune destruction and reprograming energy metabolism) and two new tumor characteristics (tumor-promoting inflammation and genome instability and mutation) (2). In multiple myeloma (MM), a destructive cancer of the plasma cell that grows predominantly in the bone marrow (BM), it is clear that all these hallmarks and characteristics are in play, contributing to tumor initiation, drug resistance, disease progression, and relapse. Bone marrow adipose tissue (BMAT) is a newly recognized contributor to MM oncogenesis and disease progression, potentially affecting MM cell metabolism, immune action, inflammation, and influences on angiogenesis. In this review, we discuss the confirmed and hypothetical contributions of BMAT to MM development and disease progression. BMAT has been understudied due to technical challenges and a previous lack of appreciation for the endocrine function of this tissue. In this review, we define the dynamic, responsive, metabolically active BM adipocyte. We then describe how BMAT influences MM in terms of: lipids/metabolism, hypoxia/angiogenesis, paracrine or endocrine signaling, and bone disease. We then discuss the connection between BMAT and systemic inflammation and potential treatments to inhibit the feedback loops between BM adipocytes and MM cells that support MM progression. We aim for researchers to use this review to guide and help prioritize their experiments to develop better treatments or a cure for cancers, such as MM, that associate with and may depend on BMAT.

Keywords: bone marrow adipose, BMAT, MAT, adipocyte, microenvironment, multiple myeloma, fatty acids, bone metastasis

## INTRODUCTION

Within the last few years, researchers have begun to explore the mechanistic relationship between bone marrow (BM) adipose and adjacent tumors such as multiple myeloma (MM), which is a cancer characterized by clonal proliferation of transformed plasma cells (3). The clinical potential of such a research avenue is yet unknown, but preclinical data suggest that targeting BM adipose tissue (BMAT) could be an effective cancer treatment. BMAT also interacts with bone cells and other immune cells, highlighting indirect ways in which BMAT may affect MM disease progression

#### *Edited by:*

*Erica Lynn Scheller, Washington University School of Medicine, USA*

#### *Reviewed by:*

*Graziana Colaianni, University of Bari, Italy Izabela Podgorski, Wayne State University School of Medicine, USA*

> *\*Correspondence: Michaela R. Reagan mreagan@mmc.org*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 03 May 2016 Accepted: 03 June 2016 Published: 17 June 2016*

#### *Citation:*

*Falank C, Fairfield H and Reagan MR (2016) Signaling Interplay between Bone Marrow Adipose Tissue and Multiple Myeloma cells. Front. Endocrinol. 7:67. doi: 10.3389/fendo.2016.00067*

(**Figures 1** and **2**). Clearly, there needs to be more research in this area. MM cells accumulate within the BM and are highly dependent on this unique biochemical and cellular niche, as we have recently reported (4). Only recently, the idea that adipocytes may accelerate or support MM has come to researchers' attention. The BM adipocyte may play a role in MM bone homing, tumor progression, drug resistance, recurrence, or osteolysis, due to local paracrine, endocrine, or metabolic signals. Just as understanding the relationship between osteoclasts and tumor cells led to the development of highly effective antiresorptive agents (bisphosphonates), and understanding the relationship between osteoblasts and MM cells has led to bone anabolic agent research, we propose that a clearer perception of the BMAT–MM cell relationship would identify novel ways to more effectively treat or prevent MM or MM-associated bone disease.

As adipose tissue is one of the main components within the BM niche, especially in old age, obesity, and upon radiation, there is clearly a need to characterize BMAT–MM relations. In this review, we discuss the current evidence regarding the signaling pathways driving effects of BMAT on myelomagenesis and progression. This review should guide future research strategies toward developing novel therapies to target MM or MM-induced bone disease through focusing on BMAT and its derivatives. For an overview of the contributions of the other components of the BM, we refer the reader to a few other recent reviews (4–6).

## DEFINING MULTIPLE MYELOMA AND MYELOMA-ASSOCIATED BONE DISEASE

Multiple myeloma is a cancer resulting from the accumulation of genetic mutations within an immune cell, called a plasma cell. Along the uncontrolled myeloma cell growth, MM also causes disruption of the BM and cancer-induced bone disease (4). Myeloma accounts for ~1–2% of cancers and ~13–15% of all blood cancers (7) and is characterized by clonal proliferation of tumor cells in the BM, monoclonal protein spikes in the blood or urine, and organ shutdown (3). In August 2015, a revised staging system was released for myeloma from the International Myeloma working group that categorized MM as stage I, II, or III, based on disease risk levels, such as chromosomal abnormalities and serum lactate dehydrogenase (LDH) levels (8). At a median follow-up of 46 months, the society found a 5-year overall survival rate of 82% in stage I, 62% in stage II, and 40% in stage III. The 5-year progression-free survival rates were 55, 36, and 24%, respectively, for these groups. Although treatments for MM have significantly improved since the disease was first

multiple myeloma (MM) growth in the marrow through indirect mechanisms, such as influences on other cells in the marrow, or direct mechanisms. BMAT has some evidence of inhibiting osteoblasts and the anticancer effects of immune cells and supporting osteoclasts and MM cell. White adipocytes, the basis of white adipose tissue (WAT), may also contribute to tumor growth in the bone marrow through systemic signaling pathways. MM cells also induce apoptosis in osteocytes, which may support MM cells. Bone lining cells and mesenchymal stromal cells (MSCs), as well as osteoclasts, support MM while osteoblasts may induce dormancy in MM cells.

named in 1873 by J. von Rustizky (9), MM remains considered an incurable cancer. The disease is more common in males than females, African–Americans than Caucasians, older rather than younger people (the median age at diagnosis is 70), and in individuals with a family history of lymphatohematopoietic cancers (3). Obesity also has been found to be risk factor for MM in numerous studies and a pooled analysis of 20 prospective studies (10).

Myeloma arises from an asymptomatic precursor disease termed monoclonal gammopathy of undefined significance (MGUS) that progresses to smoldering myeloma and, eventually, overt, symptomatic myeloma (3). While early chromosomal abnormalities, such as immunoglobulin heavy chain translocations or trisomies, are present in both MGUS and MM, secondary translocations or mutations involving oncogenes (e.g., *MMSET, MYC, MAFB, IRF4*, *FGFR3*, *RAS* family members, among many others) (11) or tumor suppressors (e.g., *CDKN2A, CDKN2C*, or *TP53*) are unique to MM and absent in MGUS (12). Interestingly, deep sequencing of 203 tumor–normal paired samples revealed intratumor genetic heterogeneity with recurrent mutation occurring early or late during tumor evolution to be common in MM (12). Other pathways, such as the phosphatidylinositol 3-kinase (PI3K) pathway (important for cell division, growth, survival, and motility), can also be hyperactivated in MM (due to external signaling from the bone milieu) and serve as a good target, despite a lack of mutations in the pathway (13). Cells from the immune system also appear to be abnormal in MM and contribute to MM progression through expression of proteins such as TNFSF14 (6, 14) or by inducing T-cell immunosenescence (15). In sum, the genetic heterogeneity in MM may limit effectiveness of tumor-targeted therapy, indicating that better results may be obtained by targeting the bone microenvironment to impede MM and MM-induced bone disease.

Multiple myeloma-induced bone disease is the general term for the destruction of bone (associated with severe pain, pathologic fractures, and spinal cord compression) that occurs during myeloma colonization of the BM. Upon engrafting within the BM niche, MM cells accelerate osteoclastogenesis through expression of molecules, such as RANKL, MMP-13 (16), and Decoy receptor 3 (DcR3), a member of the tumor necrosis factor (TNF) receptor superfamily (17). MM cells also inhibit osteoblastogenesis, disrupting the normal equilibrium between these two processes (18), through expression of Dickkopf-1 (DKK-1) and inducing upregulation of SOST in local osteocytes. Chemokines and cytokines associated with osteolysis in MM include CCL3, CCL20, and Activin-A (19). Increased osteoclastic activity leads to hypercalcemia (elevated calcium in the blood) and bone lesions. Therefore, the mnemonic for the signs and symptoms of MM is CRAB: C, elevated Calcium in the blood stream; R, renal failure due to elevated circulating protein (immunoglobulin); A, anemia, or lack of red blood cells due to tumor crowding into the BM; and B, bone lesions (4). Much research has been directed toward inhibiting the "vicious cycle" of osteoclast activation using bisphosphonates, OPG, or RANKL antibodies (denosumab) (6, 20–22). Using bone anabolic agents to regrow bone by stimulating osteoblasts (23) is another therapy for healing bone lesions and potentially inducing quiescence in MM cells (24). Lately, research has also focused on targeting MM cell homing to the BM, either through targeting the unique BM vasculature (25, 26), the molecules (e.g., sugars) and proteins on this vasculature (27, 28), or the chemokines (e.g., SDF1) within the BM (29–31). Other marrow cellular components, such as mesenchymal stromal cells (MSCs) (5, 32–34), osteocytes (35), and adipocytes, as described in this review, are also potential new avenues to regrow bone, inhibit bone loss, or inhibit MM survival or proliferation.

## DEFINING THE BM ADIPOCYTE

The anatomy and physiology of adipose tissue, as reviewed by Colaianni et al. (36), can direct energy storage (in white adipose), energy use (in brown adipose, for heat generation), or a combination of these and other functions yet to be discovered, as seen in BMAT. BMAT is a distinct adipose depot distinguishable from other adipose depots based on differences in phenotype, stress and diet response, physiological roles, gene expression, and origin. It has been found to affect the disease course of cancer, osteoporosis, and other pathologies of the bone (37). Composed of BM adipocytes and infiltrating inflammatory cells, BMAT has a gene expression pattern that overlaps with both white adipose tissue (WAT) and brown adipose tissue (BAT) (38). Like WAT, BMAT stores energy in the form of unilocular intracellular lipid droplets, opposed to multilocular droplets, as seen in BAT (39). Yet, WAT and BMAT are different in some other regards: BMAT expression of certain proteins [e.g., Dio2, peroxisome proliferator-activated receptor (PPAR) gamma coactivator 1-alpha (PGC-1α), and FOXC2] (40) is much higher than WAT expression, and while WAT volume decreases during starvation, BMAT volume increases perhaps highlighting its evolutionary role as the last energy store during starvation (41, 42). Gene expression level is also different for WAT and BMAT, as seen in the following genes: uncoupling protein 1 (UCP1), type II iodothyronine deiodinase (Dio2), PGC-1α, PR domain containing 16 (PRDM16), Forkhead box protein C2 (FOXC2), and leptin (43). Yet, these adipose depots are similar in other regards. For example, in response to obesity in mice and humans, both WAT and BMAT volumes increase due to increased adipocyte size and quantity, suggesting that both may act as reservoirs for excess energy storage (44, 45). Overall, due to the hard-to-access location of BMAT, its interspersion with many other BM cells, and its absence from hematoxylin and eosin stain histology slides due to processing challenges, BMAT has been inadvertently ignored in the BM niche for years and is thus poorly understood relative to other adipose depots.

Adipose depot properties also diverge within the BM and are both cell- and microenvironment-dependent. Adipose in the distal long bone BM is termed "constitutive marrow adipose tissue" (cMAT) and proximal adipose is termed "regulated marrow adipose tissue" (rMAT), as it is commonly "regulated," or modified, rather than constitutively present (37). This suggests that BM adipocytes may be either location dependent or composed of two subpopulations of adipocytes; this remains under investigation. In rabbits, humans, and mice, MAT develops differently based on its location in the skeleton (46). *cMAT,* often termed "yellow adipose" due to its yellow appearance in the marrow, is found in the distal tibia and tail (caudal vertebra) of rodents and forms at birth, whereas *rMAT* accumulates with aging in proximal femora and more proximal vertebrae. *cMAT* volume can be measured by MRI in humans or by osmium microcomputed tomography in rodents and is constitutively present (47, 48). *cMAT* is proportional to bone mass in many cases; for example, the distal tibia, which is loaded with *cMAT* relative to the proximal tibia, and the caudal vertebrae, again loaded with cMAT relative to the lumbar vertebrae, also have more trabecular bone mass (46, 49). Interestingly, these sites with high cMAT/yellow MAT (distal tibia metaphysis, first lumbar vertebra), compared to regions with more red marrow (proximal tibia metaphysis or fifth caudal vertebra), also appear protected from bone loss induced by ovariectomy in rats (50).

Constitutive marrow adipose tissue may negatively impact hematopoiesis and maintain hematopoetic stem cells (HSCs) in a quiescent state (51). *rMAT* is often, but not always, correlated with low bone mass and is regulated by factors including diet, drugs, age, and other endocrine and paracrine influences (42, 52–56). Interestingly, both cell-autonomous factors and the BM microenvironment appear to govern BMAT formation. In one study, although differentiation potential was found to be generally decreased in BM-MSCs, donor age was found to affect osteogenic differentiation of BM MSCs more than it affects adipogenic differentiation (57, 58). In another study, human adiposederived stem cells showed a shift in favor of adipogenesis with increased age (59). Yet, as demonstrated in a transplant study of BM cells into old and young mice, researchers found older hosts induced greater adipogenic lineage allocation than younger hosts did for the same transplanted MSCs, demonstrating the context and source influences on adipogenesis (60).

Lineage tracing experiments demonstrate that BMAT arises from an osterix-positive BM mesenchymal progenitor cell, common to osteoblasts, chondrocytes, and other BM stromal cells (61) (**Figure 2**). Interestingly, BM adipocytes cells are more closely related to osteoblasts and chondrocytes than are peripheral WAT adipocytes (62). One study found that a quiescent, leptin receptor-positive (LepR<sup>+</sup>) progenitor cell [stem cell factor (SCF) and CXCL12 expressing, and Nestin low] is the progenitor cell for most BM adipocytes, osteoblasts, and chondrocytes. This cell is also the progenitor to new cells formed after irradiation or fracture in the bone (61). These progenitors also express Prx1, PDGFRα, and CD51 markers expressed by BM-MSCs, emphasizing the need for more thorough bone progenitor classification (61). The plasticity or elasticity between different progenitors and their progeny may complicate the unequivocal identification of phylogenic lines, and differences between mouse and human cells and proteins may also further complicate these studies. A better understanding of the lineage pathways of BM cells would provide insight into a wide array of pathophysiologies.

## BONE MARROW ADIPOCYTE INFLUENCES ON MM

High body mass index (BMI) is correlated with an increased risk of developing MM and is associated with higher levels of BM adiposity, perhaps creating an optimal microenvironment, or "soil," in which MM can engraft and grow (63–65). BM adipocytes isolated from MM patient femoral biopsies have been shown to support myeloma growth *in vitro* and may protect MM cells from chemotherapy-induced apoptosis (66, 67). These results suggest that elevated adipocyte numbers support MM advancement. By excreting free fatty acids (FFAs) and producing a plethora of signaling molecules [e.g., adipokines (leptin, adiponectin, adipsin, etc.) and growth factors (e.g., IL-6, TNFα, MCP-1, insulin-like growth factor 1 (IGF-1), and insulin)], BM adipocytes are both an energy source and an endocrine signaling factory (**Figures 3** and **4**). Many of these BMAT-derived signaling molecules may promote myelomagenesis and enhance tumor growth (42, 68) (**Figure 3**). In this section, we explore the potential contributions of BMAT to MM progression.

## Lipids and Cellular Metabolism

When metastatic ovarian cells colonize the omentum (the fatty membrane surrounding the stomach and abdominal organs), they induce adipocytes to release lipids, which are subsequently utilized as energy for tumor cell proliferation. This process transforms the soft, flexible omentum fat pad into a hardened, tumor-infiltrated membrane with few remaining adipocytes in a process termed "omental caking" (69). This same phenomenon may occur in adipose-rich BM cavities, and fuel-switching in MM cells and the use of fatty acids could prove advantageous to MM cells owing to the high energy content of lipids and lipidinduced cell signaling changes that lead to drug resistance. Yet,

Figure 3 | Tumor-supportive effects of BMAT. Many factors from BMAT may induce MM tumor growth and disease progression. Lipids may serve as a fuel source for tumor cells, antigens to stimulate precursor disease initiation [i.e., monoclonal gammopathy of undefined significance (MGUS)], or inhibitors of the immune system. IGF-1 and insulin can accelerate tumor proliferation. IL-1 and IL-15 can have effects on immune cells and inflammatory molecules to support MM growth and immune evasion. Complex interactions between TNFα, IL-6, leptin, PAI-1, and MCP-1 can lead to osteoclast activation, thrombosis, and JAK/Stat/MAPk signaling to cause osteolysis, thrombosis and tumor cell migration, drug resistance, and proliferation. Glycolytic and pentose phosphate pathway enzyme upregulation, potentially found in high energy states, can also lead to melphalan resistance in MM cells.

this story is not clear cut. For example, despite the fact that obesity correlates with increased risk of MM, one study found that obese and severely obese patients had superior overall survival and progression-free survival after high-dose melphalan and autologous hematopoietic stem cell transplantation compared with normal and overweight patients (70). Yet, other research found that melphalan-resistant MM cells upregulate glycolytic and pentose phosphate pathway (PPP) enzymes and downregulate tricarboxylic acid (TCA) cycle proteins (71). Together, these reports suggest that high BMI patients fuel MM cells via fatty acids, while hyperglycolytic diabetic patients could support MM cells via glycolysis, and that either metabolic pathways could support drug resistance. Additionally, some lipids, such as palmitic acid, have shown direct anti-myeloma effects (72).

Recent new data suggest that certain drugs, such as arsenic trioxide (As2O3), may induce anti-MM effects by affecting the sphingolipid pathways in MM cells. U266 MM cells treated with As2O3 displayed decreased lipid metabolites in this pathway including dihexosylceramide (Hex2Cer), sphingosine-1-phosphate (S1P), and sphinganine-1-phosphate (dhS1P) (73). As sphingolipids are a major group of membrane bioactive lipids, these changes could not only affect FFA metabolism but also membrane fluidity and cell–cell signaling. Further, complexity arises from the fact that sphingolipids and their metabolites also act as signal transduction messengers, regulating diverse cellular events such as cell cycle arrest or apoptosis, proliferation, cancer development, and multidrug resistance, as recently reviewed in Ref. (74). Increased fatty acid levels (saturated and n-6 polyunsaturated fatty acids) have also been observed in MM patient versus healthy donor blood serum (75). Lipid profiles differ between MM cells and plasma cells, such as the levels of glycerophospholipids [specifically phosphatidylcholine (16:0/20:4)] (76), which suggest potential therapeutic avenues based on lipid biochemistry.

Autophagy, the process by which intracellular proteins and organelles are degraded in lysosomes, is a protective process through which MM cells protect themselves from unfolded or misfolded proteins (77). Certain lipids can induce autophagy in hematological malignancies, but other lipids can induce tumor cell survival, proliferation, or cell death, so it is important to understand how different sphingolipids and their metabolizing enzymes cooperatively exert their functions (74). Modulating cholesterol metabolism in myeloma cells, in particular the sterols zymosternol and desmosterol, has also been shown to mediate autophagy signaling (78). Overall, it is clear that lipids may affect autophagy of MM cells.

New data also suggest that lipids may be drivers of monoclonal gammopathies, such as MM and MGUS, by acting as antigens for plasma-cell-derived antibodies (**Figure 3**). Evidence of this comes from data showing that clonal immunoglobulin in 33% of sporadic human monoclonal gammopathies is specific for the lysolipids lysoglucosylceramide (LGL1) and lysophosphatidylcholine (LPC) (79). Nair et al. reported that substrate reduction ameliorated Gaucher's disease-associated gammopathy in mice and suggest that long-term immune activation by lysolipids may underlie both sporadic monoclonal gammopathies and Gaucher's disease-associated gammopathies (79). This work was built on genetic analyses over the past two decades of immunoglobulin mutations in MM cells that found myelomagenesis to be an antigen-driven process (80). Implications of these findings are that decreasing key lipids responsible for myeloma initiation potentially represents a novel preventative measure for at-risk populations. Moreover, the recent evidence finds that adipocyte-derived lipids, rather than adipokines, mediate obesity-related changes in macrophage phenotypes, highlighting the influential effects of adipocyte-derived lipids of the microenvironment (81).

Lipids also function as PPARγ agonists, and the PPARγ pathway has evident tumor-promoting properties in multiple cancers, as recently reviewed in Ref. (82) (**Figures 3** and **4**). Although the receptor-independent effects of PPARγ ligands compound our understanding of PPARγ in MM, the PPARγ agonist function of certain lipids likely creates a positive feedback loop both accelerating BM adipogenesis and directly supporting MM. Recent data have also found that the PPARγ agonist pioglitazone (PIO) enhances the cytotoxic effect of the histone deacetylase inhibitor (HDACi) and valproic acid (VPA) on MM cells, *in vitro* and *in vivo*, suggesting that agonizing PPARγ while inhibiting HDACs could decrease MM growth (83). Similarly, the PPARγ agonist rosiglitazone (RGZ) suppressed the expression of angiogenic factors in MM cells (HIF-1α and IGF-1) and inhibited proliferation and reduced viability of RPMI-8226 cells in a concentration- and time-dependent manner (84). RGZ also inhibited the expression of pAKT and downregulated the expression levels of phosphorylated extracellular signal-regulated kinase (pERK) in MM cells (84). However, PPARγ has a strong osteoclastogenic effect that would likely worsen osteolysis for MM patients, highlighting a downside of using RGZ in MM.

In contrast to the above, the PGC-1α is upregulated in myeloma cells grown in a high glucose media (modeling myeloma growth in hyperglycemic patients). It also contributes to chemotherapy (dexamethasone or bortezomib) resistance. These two properties suggest that inhibiting, rather than activating, the PPARγ pathway in MM cells (and controlling hyperglycemia) may improve the efficacy of chemotherapy in MM patients with diabetes. PGC-1α also increases vascular endothelial growth factor gene (VEGF) and GLUT-4 expression in MM cells suggesting that inhibition of PGC-1α in MM cells could decrease angiogenesis and glucose uptake, potentially slowing MM cell proliferation (85). Despite the growing knowledge in this area, it is still unclear how best to modulate the PPARγ pathway to inhibit MM disease progression in patients.

## Adipocyte Cell Signaling Pathways

In addition to lipid molecules, there are a vast number of proteins derived from adipocytes that may influence MM tumor growth, as described here.

## Adipokine and Growth Factors Affecting MM Cells

Adipocyte-derived cytokines (adipokines) within the local microenvironment may also uniquely stimulate the growth of MM cells or contribute to other aspects of the disease (86). Some of the major humoral factors and adipokines that WAT and BMAT secrete are TNFα*,* monocyte chemoattractant protein-1 (MCP-1), plasminogen activator inhibitor-1 (PAI-1), resistin, leptin, and adiponectin (87, 88) (**Figure 3**). TNFα is a known MM-supportive, osteoclast-activating, and osteoblast-inhibitory factor (89). TNFα treatments induce MEK and AKT phosphorylation in MM cells and stimulate the production of IL-6. This causes a forward feedback loop that drives MM cell growth and survival (90). An autocrine TNFα-MCP-1 loop has also been identified in MM cells, which was found to stimulate MM cell migration (91) (**Figure 3**).

Plasminogen activator inhibitor-1 causes increased risk of thrombosis, as it inhibits fibrinolysis, the physiological process that degrades blood clots (**Figure 3**). PAI-1 has been shown to be elevated in MM patients and appears to contribute to the greater risk of pulmonary embolism and blood clots in these patients (92). Some results suggest that patients with MM have decreased fibrinolytic activity mainly due to increased PAI-1 activity (92). In sum, these data suggest a link between adipocyte-specific cytokines, autocrine signaling, and obesity-linked cancer.

## Adipocyte-Derived Hormones

Body weight is controlled by energy intake and expenditure, which are tightly regulated by communication between the brain and adipose depots through molecules such as adipocytederived hormones. Some hormones signal satiety (leptin) and represent high energy stores; others indicate hunger resulting from low blood glocose, inducing caloric intake as the hypothalamus receives these signals and regulates behavioral responses (93). Key adipokines such as adiponectin, leptin, and resistin are often present in skewed levels in various disease states (94–98). Abnormal adipokine levels and leptin-induced changes in gene expression profiles have been observed in MM, suggesting that these may be drivers or useful biomarkers of the disease (99–103).

## *Adiponectin*

Adiponectin is an anti-inflammatory cytokine primarily produced by adipocytes but found to be secreted by additional cell types, including osteoblasts and BM MSCs (104). It is decreased in obesity (105–107) and has been shown to inhibit MM disease progression (100, 108) (**Figure 4**). In fact, low levels of adiponectin are associated with obesity, cardiovascular disease, and diabetes and are a risk factor for breast cancer (109). Circulating adiponectin was also decreased in patients with MGUS who then progressed to overt, symptomatic MM when compared to those with MGUS that did not develop MM (110). This study also showed that C57Bl6/KaLwRijHsd mice, which are permissive to 5T murine myeloma cells, have significantly lower adiponectin gene expression and adiponectin protein in their BM and lower total serum adiponectin compared to the non-permissive, but closely related C57BL6/J mice (110). Moreover, pharmacological stimulation of adiponectin in tumor-bearing mice led to a decrease in tumor burden and increased survival (110). Importantly, in humans, low circulating adiponectin and resistin, but not leptin, are associated with MM (99, 100, 108). Adiponectin has been shown to inhibit proliferation of MM through an increase in cell death *via* activation of the protein kinase A/AMP-activated pathways (111) (**Figure 4**). In sum, these are important findings that demonstrate the potential relevance of increasing adiponectin for MM and associated bone disease therapy.

Bone marrow adipose tissue, WAT, and BAT-derived adipocytes express relatively similar amounts of the anti-myeloma protein adiponectin on the mRNA level (40), but on the protein level, and *in vivo*, adiponectin secretion is greater from MAT than from WAT (42). Moreover, BMAT specifically increases its production of adiponectin in times of starvation and in patients with cancer therapy (42). Expression of *adipoq,* the gene encoding adiponectin, in tibiae and femurs has been found to mirror changes in serum adiponectin, which suggests that circulating adiponectin levels are directly related to adiponectin production from BMAT (42). Therefore, adiponectin appears to be one of the major BMAT-derived molecules responsible for signaling from BMAT to MM cells.

## *Leptin*

Leptin, a peptide hormone produced and secreted by adipocytes, has primarily been characterized for its role in the regulation of hunger response and metabolic activity (112). The main signaling capability of leptin is through the long form of its receptor, which is expressed in peripheral and brain tissues, although its primary function has been identified as signaling through the hypothalamus (112). Signaling through its receptor, leptin stimulates JAK/ STAT cascades, mainly JAK2/STAT3 and JAK2/STAT5, to signal satiety (**Figure 3**). Congenital leptin deficiency in both mice and humans results in early obesity due to severe hyperphagia, but can be corrected with leptin replacement therapies (113, 114). In patients with obesity, circulating leptin levels are significantly higher than in normal age- and sex-matched patients, suggesting that a level of leptin resistance exists in these obese patients (115). Plasma leptin levels were found to be increased in both newly diagnosed male and female MM patients compared to healthy controls (100), and leptin levels are decreased in response to disease treatment (102). Similar to the effects of lipids mentioned above, autophagy, can also be induced by adipocyte-derived hormones (116) (**Figure 3**). Adipocytes have been found to upregulate the expression of autophagic proteins in MM cells *via* leptin and adipsin, leading to chemoresistance, suppression of caspase cleavage, and suppression of apoptosis in melphalan-treated MM cells *in vitro* and *in vivo* (67).

*Resistin, Insulin, Insulin-Like Growth Factor 1, and Androgens* Data on resistin do not translate as well from mice to human as leptin appears to, and the relationship between resistin and adiposity is not consistent between humans and mice (117).

Still, in both species, resistin is elevated in obesity, regulates insulin sensitivity, and is positively associated with insulin resistance and glucose tolerance (118). In clinical studies, low circulating resistin levels are associated with MM risk (108). Yet, another study found no significant differences in circulating serum resistin levels between newly diagnosed MM patients and healthy controls (100). Insulin and IGF-1 are, however, both adipose-derived growth factors that stimulate proliferation for MM cells (68, 119) (**Figure 3**). Lastly, adipose tissue is one of the major sources of aromatase, an enzyme also expressed in the gonads, which synthesizes estrogens from androgen precursors. Adipose-derived aromatase and the subsequent synthesis of estrogen could contribute to MM growth, as certain MM cells have been found to express estrogen receptors and proliferate in response to estrogen (78). However, the bone anabolic effects of estrogen suggest that this enzyme could combat myelomainduced bone disease. In sum, the net effects that adipocytederived hormones potentiate on MM and MM-induced bone disease are currently an open area of research.

## BMAT and Hypoxia: Tumor Growth and Drug Resistance

The relationship between BMAT and hypoxia is likely an important, dynamic, and bidirectional relationship that contributes to MM development and drug resistance. As oxygen tension ranges from 21 (in normoxia) to 12% in peripheral blood and ~1.3 to 3% (hypoxia) in the BM, based on the proximity to the vasculature and endosteum (120), it is probable that BMAT-MM *in vitro* experiments, and perhaps all BM cultures, will give more translational data if they are performed in hypoxic rather normoxic conditions (121). This is because hypoxia can drive proliferation of stem cells *via* HIF1 signaling (122), induce drug resistance in MM cells, and affect MM cell homing and egress from the BM (123–126). Some data demonstrate that hypoxia decreases adipogenic differentiation (127), and severe hypoxia (1% O2) inhibits adipogenic, chondrogenic, and osteogenic differentiation of human BM-MSCs (128). Pachón-Peña et al. found that hypoxia increased adipose-derived stem cell (hASC) proliferation and migration from lean, but not obese, patients (129), so patient type is likely important in how cells respond to hypoxia. hASC donor BMI has also been found to dictate adipogenic potential, immunophenotypic profile, and response to oxygen tension *in vitro* (129). Other studies have confirmed that obesity, and FFAs specifically, decrease stem cell multipotency (130). Overall, there appears to be an interaction coefficient between donor BMI/lipids and response to hypoxia for stem cells, suggesting that multiparameter experiments should be designed to capture these complex, non-linear interactions.

Hypoxia itself is an important factor in tumor drug resistance and is associated with poor prognosis. However, due to the challenges associated with measuring oxygen tension within the BM, it is not yet clear how, or if, the oxygen gradients in the BM specifically dictate the locations of osteolysis (131). Hypoxia activates the VEGF (132), a major stimulator of angiogenesis and neovascularization, as well as a direct inducer of MM cell growth, survival, and migration (133). Neovascularization is common in the bones of myeloma patient and in mice in areas infiltrated with myeloma cells and provides more exit routes for tumor cell intravasation and increased nutrient delivery to sustain tumor growth (134). Targeting vasculogenesis and VEGF signaling has been found to be successful to decrease tumor burden in *in vivo* models (25). VEGF concentration in the BM significantly correlates with BM microvascular density, percentage of tumor cells in bone biopsy, and hypercalcemia (135). VEGF is also significantly increased in patients after treatment who progress versus those with a partial or complete remission (135). Since adipose tissue has been shown to express high levels of VEGF, it is likely that BMAT is an important source for VEGF family members in the BM, supporting aberrant microvessel growth and neovascularization and directly fueling MM cell proliferation (136, 137). Paracrine signaling of VEGFA from BMAT to MM cells may also be fueled through autocrine signaling, as MM cells also demonstrate high VEGFA expression and production levels (124).

As MM cells are often resistant to hypoxia-induced cell death, antiangiogenic factors do not seem to be highly effective for this type of tumor cell, despite the correlations between BM vessels and disease progression. Hypoxia protects tumor cells from apoptosis through an increase in local VEGF concentrations and subsequent increases in tumor cell MAPK/ERK signaling (138). In MM cells, hypoxia increases HIF1α and activates the PI3K/ Akt/mammalian target protein of rapamycin (mTOR) pathway (139). MM cells in the BM also show high glucose uptake, similar to most tumors, as demonstrated by 18F-FDG PET imaging and increased glucose transport protein 3 (GLUT3) expression (140). As the metabolic shift from oxidative metabolism to glycolysis occurs based on both energy and oxygen sources, it is clear that the fuel type (lipid versus glucose), expression of glucose transporter, and glycolytic enzymes, as well as oxygen tension, direct tumor cell metabolism and fuel switching. Therefore, lipids and adipose tissue affect MM cell metabolism depending on oxygen availability. Specifically, decreasing local lipid concentrations may simply switch tumor cell metabolism from fatty-acid oxidation to glycolysis and not necessarily decrease tumor proliferation, or, fuel switching coupled with oxygen tension control may prove a viable therapeutic avenue through which to tackle MM. As a final consideration here, intermittent hypoxia also affects adipose tissue macrophage polarization and tumor infiltration, suggesting that immune changes should also be considered when investigating metabolic and hypoxic-based interventions in MM (141).

## Bone Marrow Adipocytes and Skeletal Remodeling

The growing evidence associating elevated BMAT with low bone density suggests that BM adipocytes may contribute to bone loss in MM or that bone loss may contribute to increased BM adiposity. Either dynamic could support MM growth and increased risk of fracture (**Figure 2**). In humans (142–145) and rodents (146–149), there is often an inverse correlation between BMAT and bone quantity. Decreased bone volume or mass coinciding with higher BMAT is consistently observed across sexes, ages, models, and underlying disease etiologies (54). Moreover, many pharmacologic strategies cause opposing effects on bone and adipose tissue [glucocorticoids, hormone replacement therapies, radiation, and thiazolidinediones (TZD)] (150). Higher BMAT has been found to correlate with lower trabecular bone mineral density (BMD) in older women, but not men, and higher marrow fat is associated with prevalent vertebral fracture in men, even after adjustment for BMD (145). Lumbar spine BMD has been found to negatively correlate with BMAT (151). High BMAT also leads to disrupted hematopoiesis and reduced BMD in other studies and may increase the risk of bone metastasis, potentially resulting from an increase in receptor activator of NFκB-ligand (RANKL) and downregulation of osteoprotegerin, as observed in aging-related marrow adipogenesis (44, 152, 153).

In moving beyond correlation into causation, recent evidence demonstrates that adipocytes actively inhibit osteogenesis, based on lower mineralization, alkaline phosphatase activity, and expression of osteogenic (Runx2, osteocalcin) mRNA markers, using conditioned media experiments with hMSCs (154). Adipocytes can also induce osteoblast apoptosis (154). One pathway found to govern the effects of adipokines on osteoblasts is the PI3-kinase-FoxO1 pathway (155). Both decreased osteoblast function and induced apoptosis were enhanced by dexamethasone treatment of adipocytes, and both processes appear to be driven by the lipotoxic effect of two FFAs, stearate and palmitate, which may act as PPARγ-ligands (inhibiting osteogenesis), and can induce ROS in human cells (154). These findings demonstrate that increased BMAT may decrease osteogenesis, thus contributing to bone disease in MM patients, although this has not yet been explored in myeloma patient MSCs. Overall, the effects of BMAT specifically on MM-induced bone disease and osteolysis may be substantial and promising as a new therapeutic target.

## Bone Marrow Adipocytes and Hematopoiesis

As BM adipocytes are interspersed throughout the vascular and endosteal niches responsible for guiding the lineage commitment of HSCs, they may also affect hematopoiesis both *via* local and systemic effects. Research on human iliac crest-derived marrow adipocytes found that these cells have the ability to support CD34<sup>+</sup> hematopoietic progenitor cells *in vitro* (39). BMAT is also intimately associated with the blood-forming marrow. Primary human BMAT adipocytes, purified from the iliac crest, have the ability to support differentiation of CD34<sup>+</sup> hematopoietic progenitor cells in long-term culture *in vitro* (39). Yet, other data suggest that BMAT may be inhibitory toward hematopoiesis; this has been observed in mouse experiments where BMAT induced hematopoietic cell quiescence and decreased the number of progenitor marrow cells (51). Adipocyte-derived factors are also known to inhibit B lymphopoiesis (156).

The number of adult BM adipocytes was found to correlate inversely with the hematopoietic activity of the marrow and decrease marrow transplant cell engraftment after irradiation (51). Yet, in another study, mice treated with a TZD called "Troglitazone," which causes massive BMAT expansion, hematopoietic progenitor frequency was not altered, and, in fact, preadipocytes were found to support hematopoetic cells *in vitro* (77). Thus, it is unclear if MAT always has a negative influence on the hematopoietic niche, or if this is time, location, or disease dependent.

## INFLUENCES OF MYELOMA ON BMAT

Bone marrow MSCs can give rise to BM adipocytes and osteoblasts, as dictated through expression of proteins in major transcriptional regulatory pathways such as PPARγ and Wnt, respectively. It is not well understood how MM cells alter BMAT or MSC cell fate, but a study from 2007 revealed that MM-MSCs retain their capacity to differentiate down adipogenic and osteogenic lineages, although quantification of this differentiation (e.g., with oil red O or alizarin red staining) was not performed (157). Studies since then have observed a decreased ability for MM-MSCs to proliferate and undergo osteogenic differentiation (5, 32), suggesting that their adipogenic capacity may be altered. It is also possible that MM cells utilize the lipids stored in BMAT to fuel their proliferation and migration, as other tumor cells (ovarian cells) have been found to do in other adipose depots (the omentum) (158). This utilization would decrease the amount of lipid stored in these cells, though this is an observation that has yet to be examined. Research into the bidirectional communication between MM cells and BMAT is needed to determine how MM cells affect BMAT as well as the ramifications of these interactions on tumor growth and osteolysis.

## LINKING BMAT AND SYSTEMIC INFLAMMATION

Bone marrow adipose tissue is linked to systemic inflammation through mechanisms that include the production of proinflammatory cytokines and lipids able to undergo oxidation. Obesity and aging both correlate with increased systemic inflammation, increased risk of MM, and increased BMAT. This leads to a few potential hypotheses: (1) that BMAT drives MM through local and/or systemic effects (e.g., on inflammation), or (2) that elevated BMAT and MM correlate because both are driven by a common or linked underlying mechanism, e.g., obesity, aging, or decreased immune function. Currently, either hypothesis could prove true. While WAT imparts systemic/endocrine influences, BMAT may produce systemic as well as local, paracrine, and cell–cell contact-based effects on tumors. The close proximity of BMAT and MM cells suggests potential contact-mediated bidirectional signaling between these cells, which is absent from WAT–MM cell interactions. However, WAT appears to be comprised of cells that derive from the marrow (up to 35%) (159); the signaling parallels and lineage tracing links between WAT and BMAT confound determining the specific contributions of each toward MM progression or myelomagenesis. Although more research examining the specific contributions of each depot are needed, much evidence suggests that immune system alterations resulting from elevated BMAT or WAT could contribute to MM progression (44, 152, 160).

In breast cancer, obesity-related host factors, such as components of the secretome (e.g., insulin, IGF-1, leptin, adiponectin, steroid hormones, cytokines, vascular regulators, and inflammationrelated molecules), explain the causative link between increased risk of breast cancer in postmenopausal women and poor prognosis in pre- and postmenopausal women (161). Many of these same factors are also systemic signals that could explain the link between obesity and increased MM risk. However, proinflammatory cytokines that are derived from adipose tissue, such as IL-1 (162), can be difficult to identify as anti-myeloma or myelomasupportive, because of the complex roles of the immune system in cancer. In general, immune cells attack and can eliminate myeloma cells. But systemic inflammation can also contribute to tumor growth if regulatory T-cells or myeloid suppressor cells (which are cells that suppress other immune cells) are increased. Other adipocyte-derived factors are proinflammatory and support natural killer cells, such as IL-15 (163). As genetically modified, *ex vivo*-expanded natural killer cells are being used as a treatment for MM and many cancers, IL-15 and adipocyte-induced support of NK cells may in fact have anti-myeloma consequences (164). Yet, IL-15, along with other angiogenic factors (VEGF, IL-6, and HGF), is also significantly increased in MM patient blood serum reflecting a correlation between angiogenesis and MM (164). From this perspective, IL-15 and these other adipocytederived factors appear to support tumor growth through both direct effects and also increased tumor vascularization. MSC adipogenic differentiation has also been found to be modulated by natural killer cells (165), suggesting that a forward feedback loop between inflammation and adipogenesis may be at work. These data suggest that adipocytes not only are affected by, but also affect, the immune system. For a review on systemic and BMAT-induced inflammation and its contributions to tumor growth and survival, dysregulated bone remodeling, and activation of inflammatory pathways in tumor cells (e.g., CCL2/ CCR2- and COX-2-dependent pathways), refer to the review by Hardaway et al. (166).

Lipids are essential components of cell membranes and represent an energy-rich fuel source. However, lipids are frequently targeted by reactive oxygen species (ROS), such as free radicals. This leads to the oxidation of lipids in a chain reaction known as lipid peroxidation, which has been associated with a wide range of diseases, including cancer, diabetes, and neurological disorders (167). Many of the products of free radical chain oxidation are unstable, but stable isoprostanes have become the gold standard measurable biomarker for oxidative stress (167). One well-studied lipid electrophile, 4-NHE, is generated from lipid peroxidation and mediates a variety of biological processes (e.g., DNA damage, mutagenesis, inflammatory response, cell growth, and apoptosis) through a range of pathways (ER stress, stress-responsive MAP kinase signaling, NF-kB signaling, and DNA damage response signaling) (167). Malondialdehyde (MDA) is another product of lipid peroxidation; it is highly mutagenic (168). MDA and 4-NHE are two molecules responsible for lipid-initiated genetic disruption that could support MM development through numerous pathways, such as the oxidative stress-driven activation of the PI3K/AKT pathway and inactivation of the tumor suppressor gene *PTEN* (169). Oxidative stress can also lead to increased PPAR, Cox-2, MAPK, and PKC signaling; any of these pathways could support myelomagenesis or disease progression (170). As antioxidants can abrogate oxidative-stress-induced apoptosis of osteoblasts, they may represent a potential therapeutic avenue in MM (154).

## TREATMENTS TARGETING BMAT

There is immense potential in targeting BMAT or BMATderived factors, to combat myeloma initiation, progression, relapse, chemoresistance, and osteolysis. Based on preclinical data regarding the roles of adiponectin in MM, recombinant or biologically isolated adiponectin treatment for MM patients with low adiponectin levels may hold great potential as a therapeutic treatment. Similarly, decreasing BMAT-derived factors that are MM-supportive using inhibitors or antibodies may be a potential future BMAT-targeted therapy. Another way to target BMAT may be to target those signaling pathways that push MSCs down the adipogenic rather than osteogenic lineage, thus flipping the commitment lineage switch. One such pathway is the Wnt signaling pathway, which supports osteogenic differentiation and inhibits adipogenic differentiation. As we know that sclerostin, a Wnt inhibitor, is elevated in the BM of MM patients, it is possible that antisclerostin antibodies would not only increase bone volumes but also decrease BMAT in MM patient marrow, creating a less hospitable microenvironment for MM cells to colonize (171, 172). Other potential target lineage switches that induce osteogenesis and limit adipogenesis are parathyroid hormone receptor (PTH), TAZ/YAP (173), and numerous zinc finger proteins (174).

It is important to consider the link between BMAT and bone when analyzing adipose-directed therapies, because treatments that affect bone could affect BMAT (and *vice versa*). As there appears to be a reciprocal relationship between BMAT and bone formation in both healthy and diseased conditions (175, 176), increasing bone mass may be one novel way to decrease BMAT and also strengthen bones that are weakened by MM. It is becoming clear that the skeleton has a complex, non-linear, and genotype-dependent relationship with energy utilization and MAT (151, 177). Exercise has been shown to significantly suppress BMAT volume and induce bone formation in certain mouse models, suggesting that a healthy diet and increased exercise or strength training program could create a two-pronged attack to strengthen bones and decrease BMAT in MGUS or MM patients (178). The antidiabetic drug metformin can also decrease BMAT in mice that are fed with a high fat diet (Michaela R. Reagan and CJ Rosen, unpublished data). It can also modestly improve bone volume (179) as well as directly affect metabolism of tumor cells (180). These data suggest that metformin may be another potential multidimensional therapeutic. The topic of metformin effects on cancer has been reviewed recently (181).

Altering lipid levels, ratios, or content systemically or in the BM may also hold great promise as an anti-myeloma treatment. For instance, Abdi et al. demonstrated that omega-3 fatty acids [n-3 polyunsaturated eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)] induced apoptosis and increased sensitivity to bortezomib in MM cells preclinically, without affecting normal human peripheral mononuclear cells viability (182). These lipids modulated multiple signaling pathways including NFκB, Notch, Hedgehog, oxidative stress, and Wnt. They also induced apoptosis through mitochondrial perturbation and caspase-3 activation (182). Combined with the data above on oxidative stress, these data suggest that supplements such as vitamins (antioxidants) and fish oil, and/or diets rich in fish, fruits, and vegetables, should be explored as preventative measures in the development of MM. However, carefully designed trials are necessary to best optimize treatment regimes, as some antioxidants, such as vitamin C and flavonoids in vegetables, fruits, and green tea, can neutralize and should not

## REFERENCES

1. Hanahan D, Weinberg RA. The hallmarks of cancer. *Cell* (2000) 100:57–70. doi:10.1016/S0092-8674(00)81683-9

be used with bortezomib, a commonly prescribed anti-myeloma proteasome inhibitor (183).

## CONCLUSION

As reviewed herein, BMAT appears to affect MM through an array of different mechanisms. We have described what is currently understood about the BM adipocyte and BMAT. We next highlighted the ways in which BMAT may support MM, for example, through bioactive lipids (as a fuel source, signaling molecule, and a substrate for lipid peroxidation), and myelomasupportive adipokines (e.g., IL-6, TNFα, MCP-1, PAI-1, IL-6, resistin, and leptin). We also provided an overview of adiponectin, a protein that is decreased during obesity and has anti-myeloma properties making it an attractive potential therapeutic in MM. The complex relationship between hypoxia, BMAT, angiogenesis, and myeloma in the BM was discussed. Influence of BMAT on bone health and osteogenesis was delineated, and our current understandings of potential ways in which MM cells may affect BMAT were outlined. The review investigates the relationship between BMAT and systemic inflammation in relation to MM. Lastly, we suggested possible therapeutic avenues through which BMAT could be targeted, similarly to how osteoblasts and osteoclasts, and factors derived from these cells, have been successfully targeted in MM. Targeting lipid metabolism of cancer cells and adipocytes in combination with standard antimyeloma therapies will likely reveal novel therapeutic avenues through which to attack hematological malignancies. In sum, we are optimistic about the development of new combination therapies and preventative methods that take into account the roles of the BM adipocyte in MM and other bone-metastatic cancers. The path toward improved therapies will be built on basic scientific research of BMAT roles in cancer.

## AUTHOR CONTRIBUTIONS

CF, HF, and MR contributed to the conception, drafting, writing, and editing of this work.

## ACKNOWLEDGMENTS

The authors thank Dr. Michael Erard, Scientific Editor and Writing consultant at Maine Medical Center Research Institute (MMCRI) for editorial assistance. We apologize to colleagues whose work could not be cited due to space constraints.

## FUNDING

The authors' work is supported by Start-up funds, a pilot project grant and support from NIH/NIGMS (P30 GM106391 and P30GM103392), and the NIH/NIDDK (R24 DK092759-01) at Maine Medical Center Research Institute.


effects of recombinant leptin on OB-R, leptin expression and cytokine secretion by human hematopoietic malignant cells. *Cytokine* (2009) 48:203–11. doi:10.1016/j.cyto.2009.07.006


and human adipose tissue. *FASEB J* (2015) 30:1096–108. doi:10.1096/ fj.15-278994


**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 © 2016 Falank, Fairfield and Reagan. 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.*

# New 3d-Culture Approaches to study Interactions of Bone Marrow Adipocytes with Metastatic Prostate Cancer Cells

*Mackenzie Katheryn Herroon1 , Jonathan Driscoll Diedrich1,2 and Izabela Podgorski1,2\**

*1Department of Pharmacology, Wayne State University School of Medicine, Detroit, MI, USA, 2Karmanos Cancer Institute, Wayne State University School of Medicine, Detroit, MI, USA*

Adipocytes are a major component of the bone marrow that can critically affect metastatic progression in bone. Understanding how the marrow fat cells influence growth, behavior, and survival of tumor cells requires utilization of *in vitro* cell systems that can closely mimic the physiological microenvironment. Herein, we present two new threedimensional (3D) culture approaches to study adipocyte–tumor cell interactions *in vitro*. The first is a transwell-based system composed of the marrow-derived adipocytes in 3D collagen I gels and reconstituted basement membrane-overlayed prostate tumor cell spheroids. Tumor cells cultured under these 3D conditions are continuously exposed to adipocyte-derived factors, and their response can be evaluated by morphological and immunohistochemical analyses. We show via immunofluorescence analysis of metabolism-associated proteins that under 3D conditions tumor cells have significantly different metabolic response to adipocytes than tumor cells grown in 2D culture. We also demonstrate that this model allows for incorporation of other cell types, such as bone marrow macrophages, and utilization of dye-quenched collagen substrates for examination of proteolysis-driven responses to adipocyte- and macrophage-derived factors. Our second 3D culture system is designed to study tumor cell invasion toward the adipocytes and the consequent interaction between the two cell types. In this model, marrow adipocytes are separated from the fluorescently labeled tumor cells by a layer of collagen I. At designated time points, adipocytes are stained with BODIPY and confocal *z*-stacks are taken through the depth of the entire culture to determine the distance traveled between the two cell types over time. We demonstrate that this system can be utilized to study effects of candidate factors on tumor invasion toward the adipocytes. We also show that immunohistochemical analyses can be performed to evaluate the impact of direct interaction of prostate tumor cells with adipocytes. Our models underline the importance of using the appropriate culture conditions to mimic physiological interactions between marrow adipocytes and metastatic tumor cells. These systems have a potential to be utilized for analyses of various factors that may be regulated by the adipocytes in bone. Their application likely extends beyond metastatic prostate cancer to other tumors that colonize the bone marrow microenvironment.

#### *Edited by:*

*William Peter Cawthorn, University of Edinburgh, UK*

#### *Reviewed by:*

*Stuart Rushworth, University of East Anglia, UK Bonnie Loveless King, Stanford University, USA*

*\*Correspondence:*

*Izabela Podgorski ipodgors@med.wayne.edu*

#### *Specialty section:*

*This article was submitted to Bone Research, a section of the journal Frontiers in Endocrinology*

*Received: 14 May 2016 Accepted: 20 June 2016 Published: 06 July 2016*

#### *Citation:*

*Herroon MK, Diedrich JD and Podgorski I (2016) New 3D-Culture Approaches to Study Interactions of Bone Marrow Adipocytes with Metastatic Prostate Cancer Cells. Front. Endocrinol. 7:84. doi: 10.3389/fendo.2016.00084*

Keywords: bone marrow adipocytes, bone metastasis, prostate cancer, 3D culture, 3D invasion, *in vitro* models

## INTRODUCTION

Adipocytes constitute a significant portion of adult bone marrow and their number increases with age, obesity, and metabolic dysfunction (1–3). Growing evidence positively links the abundance of fat cells in the marrow with metastatic progression. Adipocyte-rich bone marrow appears to contribute to skeletal colonization and growth in a number of secondary cancers, including prostate (4, 5), breast (6), multiple myeloma (7, 8), and melanoma tumors (9). It is believed that adipocytes enhance the fertility of the bone metastatic niche by serving as a source of growth factors, chemokines, and lipid mediators (10, 11). Specifically, they have been shown to (1) upregulate lipid transporters and drive lipid uptake by tumor cells (5), (2) promote osteoclast differentiation and maturation (4, 9), and (3) induce authophagy-driven tumor cell survival, all processes that ultimately allow the metastatic cancers to thrive in the bone marrow niche (8). Despite these emerging data clearly pointing to marrow fat cells as one of the critical determinants of tumor cell fate in bone, their functional contribution to the growth and aggressiveness of metastatic tumors in bone is not well understood. Studies investigating the interactions between the tumor cells and adipocytes in the bone marrow have been limited and thorough mechanistic evaluations on how fat cells affect the phenotype, metabolism, and function of the surrounding cells in the metastatic niche are lacking.

The majority of the studies examining adipocyte–tumor cell interactions to date have utilized pre-adipocyte cell lines or adipocytes derived from visceral or breast adipose tissues (12–16) depots, which are known to be distinctively different from bone marrow fat (17). There have only been a handful of studies, including our own, that have examined the interactions of bone marrow mesenchymal cell-derived or primary bone adipocytes with metastatic tumor cells (4, 5, 7–9). Although all of these investigations resulted in important findings linking marrow adipocytes with metastatic progression, the caveat is that they have all been performed using two-dimensional (2D) culture approaches. It is becoming increasingly recognized that 2D layer cultures, although convenient and reasonably inexpensive, do not adequately mimic the limited diffusion-driven access to nutrients, growth factors, and signaling molecules in the tumor microenvironment (18). Under physiological conditions, exposure of solid tumors to microenvironmental factors, such as oxygen, nutrients, stress, and therapeutic treatments, is heterogeneous and regulated by their three-dimensional (3D) spatial conformation (19). The importance of employing 3D models to model tumor architecture has proven critical to understanding the mechanisms behind tumor phenotype, behavior, and response to therapy (19–22). Emphasis has also grown on considering the contribution of host cells in the tumor microenvironment to cancer progression, and various *in vitro* models that focus on stromal–epithelial interactions and immune cell involvement have emerged (21, 23–27).

Three-dimensional, multi-cellular cell culture models have become well-accepted tools for dissecting complex molecular mechanisms of tumor progression that may not be possible to dissect *in vivo*. There have also been many advancements in the development of 3D culture systems that mimic specific tumor niches, including very complex and dynamic microenvironments such as bone (28). Models intended to interrogate the mechanisms of skeletal metastases range from culture of tumor cells and specific bone-derived cells on biologically derived or synthetic matrices, through the use of patient-derived xenografts and direct culture of tumor cells with bone explants (21, 28). However, aside from one recently reported *in vitro* system designed to evaluate bone marrow adipose colonization by breast cancer cells (6), there have been no *in vitro* 3D models that consider involvement of marrow adipocytes.

Here, we describe new *in vitro* approaches designed to study the interaction of prostate cancer cells with bone marrow-derived adipocytes. Our methods employ murine bone marrow mesenchymal cells differentiated into adipocytes in 3D collagen I gel and grown in a Transwell system with 3D-cultures of prostate carcinoma cells. We show that in this system, which allows continuous exchange of factors between the two cell types, adipocytes promote 3D growth of tumor spheroids. We also demonstrate that the cell culture approaches we are employing in this model allow for easy manipulation and are suitable for imunocytochemical analyses. We show examples of immunofluorescence analyses of metabolism-associated factors, such as carbonic anhydrase 9 (CA9) and hexokinase 2 (HK2) that reveal distinctively different expression profiles between 2D and 3D cultures exposed to adipocytes. We also demonstrate the suitability of our model to study proteolysis by live prostate carcinoma cells and potentially other components of bone marrow microenvironment, such as bone marrow macrophages. Finally, we also describe a design of a 3D invasion assay that allows direct monitoring of the attraction of prostate tumor cells to marrow adipocytes and can be utilized to evaluate potential inhibitors that target this interaction. Our models provide new approaches to dissect the functional role of marrow-derived adipocytes in tumor cell growth and aggressiveness.

## MATERIALS AND METHODS

## Materials

Dulbecco's modified Eagle's medium (DMEM), minimum essential medium (MEMα), and other chemicals, unless otherwise stated, were obtained from Sigma (St. Louis, MO, USA). HyClone fetal bovine serum (FBS) was from ThermoFisher (Pittsburg, PA, USA). Trypsin–EDTA, Alexa Fluor 488-conjugated goat anti-rat and anti-rabbit IgG, MitoTracker Deep Red FM, CellTracker Orange (CTO), DQ collagen type IV, BODIPY (493/503), Hoechst Dye, and Gentamicin (G418) were from Invitrogen (Carlsbad, CA, USA). StemXVivo Adipogenic Supplement was from R&D Systems (Minneapolis, MN, USA). Rosiglitazone was from Cayman Chemical Company (Ann Arbor, MI, USA). PureCol collagen type I was from Advanced Biomatrix (San Diego, CA, USA). Cultrex™ (rBM; reduced growth factor) was from Trevigen (Gaithersburg, MD, USA). Rat monoclonal F4/80 was from Abcam (Cambridge, MA, USA). Rabbit monoclonal carbonic anhydrase 9 (CA9) and hexokinase II (HK2) were from Cell Signaling Technology (Danvers, MA, USA). Transwell systems (Costar™ Transwell™ Permeable Supports with 0.4-μm pore size) were from Corning (Corning, NY, USA). FABP4 inhibitor (BMS309403) was from Calbiochem (San Diego, CA, USA). Atglistatin was from Axon Medchem (Groningen, Netherlands).

## Cell Lines

PC3, an androgen independent cell line derived from a bone metastasis of a high-grade prostate adenocarcinoma, was purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). The PC3-DsRed cell line was established by stable transfection with pDsRed2-N1 vector (Clontech Laboratories, Palo Alto, CA, USA), containing the neomycinresistant gene. Transfection was performed using Lipofectamine 2000 and pooled populations of stable cells were selected, expanded, and maintained in medium supplemented with 400 mg/ml of G418 (29). L929 cells (source of M-CSF for macrophages, purchased from ATCC) were cultured until confluent and conditioned medium was collected, centrifuged, and stored at −80°C until ready for use. PC3 and L929 cells were cultured in DMEM supplemented with 10% FBS, 10 mM HEPES, and 100 U/ ml penicillin–streptomycin. All cells were maintained in a 37°C humidified incubator ventilated with 5% CO2.

## Isolation and Preparation of Primary Murine Bone Marrow Adipocytes and Macrophages

Collection of murine cells was performed in accordance with the protocol approved by the institutional Animal Investigational Committee of Wayne State University and NIH guidelines (Protocol # 15-12-025; IP, PI). Primary mouse bone marrow stromal cells (mBMSC) were isolated from femurs and tibiae of 6- to 8-week-old FVB/N mice and induced to become bone marrow adipocytes according to our previously published protocols (5). Specifically, bone marrow from each tibia and femur was flushed with of DMEM containing 20% FBS (4 ml/bone) using a 26-gauge needle. Marrow suspension was mixed and broken apart using a 20-gauge needle and seeded into a 6-well cell culture plate (~3 ml/ well). After 24 h, non-adherent cells were removed by replacing the medium. Cells were cultured to confluency by changing the medium every 2–3 days and then expanded to larger dishes as needed. For adipogenic differentiation, mBMSCs (~600,000/ well) were mixed with bovine collagen I and seeded in six-well plates (500 μl/well) for Transwell coculture or 60-mm dishes (500 μl, ~1,300,000/dish) for 3D invasion assays. Approximately 48–72 h later, upon reaching confluency, cells were treated with adipogenic cocktail (30% StemXVivo Adipogenic Supplement, 1 μM insulin, 2 μM Rosiglitazone; DMEM and 10% FBS) for 8–10 days (5).

For the preparation of bone marrow macrophages (BMMs), bone marrow was flushed from femurs and tibiae of 10- to 12-week-old FVB/N male mice with BMM growth medium (MEMα containing 20% FBS and 30% L929-conditioned media as the source of M-CSF; 4 ml/bone) using a 26-gauge needle (30). The cell suspension was mixed with an additional 20 ml of BMM medium using a 20-gauge needle and plated onto three 100-mm Petri dishes (12 ml/dish). Cells were cultured for 4–5 days in a 37°C humidified incubator ventilated with 5% CO2 to obtain differentiated BMMs.

## 2D Tumor Cell-Adipocyte Coculture Using a Transwell System

Bone marrow adipocytes were differentiated in six-well plates as described above. On the day of the Transwell setup, tumor cells were plated on acid-washed glass coverslips (12 mm in diameter) coated with reconstituted basement membrane (rBM; Cultrex™; Trevigen). Briefly, 100 μl of 1 mg/ml rBM was added per coverslip and allowed to polymerize for 45 min at RT, then the excess was removed via suction, and the coverslips were allowed to dry for 15 min at RT before cells were plated. PC3 cells (30,000–50,000) were plated in 500 μl of normal growth medium and allowed to settle for a minimum of 4 h at 37°C. Adipocytes were prepared for Transwell coculture by washing 3× with PBS and adding 2 ml of DMEM supplemented with 5% FBS. Transwell membranes were placed above the adipocyte culture according to manufacturer's instructions and coverslips with attached tumor cells (2× coverslips per transwell) were placed on each membrane with medium gently added on top (2 ml). The control coverslips were cultured separately in DMEM supplemented with 5% FBS. All cells were allowed to grow for 48 h at 37°C.

## 3D Tumor Cell-Adipocyte Coculture Using a Transwell System

Adipocytes were prepared as described above for the 2D system. The experimental approach to prepare 3D cultures of prostate carcinoma cells has been adapted from the original protocols for breast cancer progression models developed by the Brugge laboratory (31). Non-diluted rBM (15.35 mg/ml) was used for the preparation of all cultures. In brief, acid-washed coverslips were placed in 35 mm dishes, and 45 μl of rBM was carefully added on top of each coverslip, making a continuous surface without going over the edge of the coverslip. The dishes were placed in 37°C for 15 min to polymerize, then 60 μl of PC3 cell suspension was placed on top of each rBM-coated coverslip (30,000–35,000 cells/coverslip). The dishes were placed in a 37°C incubator for 45 min–1 h to allow the cells to settle and attach to the matrix. After cells were attached, 2–3 ml of media (5% FBS, 2% rBM overlay) was added to the 35-mm dish control coverslips. For 3D Transwell cultures, one 3D coverslip (covering only ~24% of the membrane and thus allowing free flow of factors and nutrients) was placed on each Transwell membrane positioned above the differentiated adipocyte culture, and 2 ml of medium was gently added (**Figure 1**). Cells were allowed to grow for 48–120 h depending on experimental design.

## Immunofluorescence Analyses

For the analysis of the expression and localization of carbonic anhydrase 9 (CA9) (**Figure 2**), control and Transwell coverslips from 2D and 3D cultures were washed with PBS, fixed with cold methanol, and incubated with rabbit monoclonal anti-CA9 antibody (1:50) at 4°C overnight. To label the mitochondria, coverslips were washed with PBS, and incubated with 200 nM MitoTracker Deep Red for 30–45 min at 37°C prior to fixation.

For immunodetection of hexokinase 2 (HK2) (**Figure 3**), cells were washed with PBS, fixed with cold methanol, and stained with rabbit monoclonal anti-HK2 antibody (1:100) at 4°C overnight. Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:1000) was used as a secondary antibody for both CA9 and HK2 immunostaining. For 2D cultures, DAPI was used as a nuclear stain, and coverslips were mounted using Vectashield mounting medium (Vector Laboratories) before imaging with Zeiss LSM 510 META NLO confocal microscope using 40× oil immersion lens. Coverslips from 3D cultures were left in PBS in 35-mm dishes, Hoechst Dye (1:1000) was added for labeling of nuclei, and imaging was performed at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

## Imaging the Infiltration of Bone Marrow Macrophage into 3D Tumor Spheroids

Primary BMMs were differentiated from murine bone marrow cells, as described above. To examine their ability to infiltrate the pre-formed tumor spheroids in the absence or presence of adipocytes, the 3D tumor-adipocyte Transwell cocultures were set up and cultured for 3 days, as described above. To distinguish tumor cells from macrophages, PC3 cells were pre-labeled with CTO (1:1000 in serum-free media for 1 h at 37°C) prior to plating on rBM-covered coverslips. After 3-day culture, medium with rBM overlay was removed from the Transwell or control dishes, 60 μl of cell suspension containing 150,000 BMMs was plated on top of each coverslip and allowed to settle for 45min-1 h at 37°C, then new medium (containing 2% rBM overlay) was reapplied. Cells were allowed to grow for additional 2 days (**Figure 4**) before cultures were fixed in 3.7% formaldehyde and immunostained for murine macrophage marker F4/80 using rat anti-mouse F4/80 antibody (1:50). Alexa Fluor 488 conjugated goat anti-rat IgG (1:1000) was used as a secondary antibody. Hoechst Dye (1:1000) was added for labeling of nuclei, and imaging was performed at an extended depth of focus with Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

## Imaging Proteolysis by Live Prostate Tumor Cells in the Absence or Presence of Infiltrating BMMs

Cleavage of DQ-collagen IV substrate by live PC3 cells grown in a control or Transwell coculture with adipocytes

in the presence or absence of BMMs was assayed in real time and quantified based on published protocols (32–34). Briefly, single cell suspensions of tumor cells (30,000–35,000) were plated on top of coverslips coated with rBM containing DQ-collagen IV (1:30) and overlayed with 2% rBM. Cells were grown alone or in Transwell coculture with adipocytes for 3 days prior to addition of pre-labeled BMMs (**Figure 5**). After an additional 2-day culture, Hoechst dye was added as nuclear marker and DQ-IV proteolysis under all culture conditions was imaged live at an extended depth of focus

with Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

## Coculture of Bone Marrow Macrophages with Prostate Tumor Cells in a 3D Transwell System

To examine the effects of BMMs on 3D growth and proteolysis by prostate tumor cells cocultured with marrow adipocytes, we pre-labeled differentiated BMMs by incubating them with a

1:1000 dilution of with CellTracker Orange (CTO) in serum-free medium for 1 h at 37°C. Labeled BMMs were then washed 3× with PBS and maintained in normal BMM medium overnight until use. On the day of 3D Transwell setup, adipocytes and prostate tumor cells were prepared as described above. Coverslips were coated with rBM and 75,000 BMMs mixed with 30,000 PC3 cells were plated together on top of a coated coverslip, as described above. As before, one 3D coverslip was placed on each Transwell membrane positioned above differentiated adipocyte culture, and 2 ml of medium was gently added (**Figure 6**). Cells were allowed to grow for 3 days, and DQ-IV proteolysis was imaged live at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

## 3D Invasion Assays

Bone marrow-derived mBMSC cells were mixed with collagen I matrix and plated in 60-mm dishes (500 μl/dish; ~1,300,000/500 μl). Cells were differentiated into adipocytes over the course of 8–10 days as described above. Postdifferentiation, each plate was washed 3× with PBS, with all PBS carefully removed before 1 ml of collagen I was added on top and allowed to polymerize at 37°C for 45 min–1 h. A total

of 4 ml of medium, containing 400,000 dsRed-expressing PC3 cells, was gently added on top. Cultures were maintained in a 37°C humidified incubator ventilated with 5% CO2 for up to 96 h. At designated time-points, plates were washed with PBS, fixed for 45 min with 3.7% formaldehyde at RT, stained with BODIPY (493/503) (1:1000, 1 h at RT), nuclei were marked with Hoechst dye, and imaged at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 20× dipping lens (**Figure 7**). For inhibitor studies (**Figure 8**), FABP4 inhibitor (BMS309403; 1 μM) and Atglistatin (10 μM) were added to both the collagen I layer between the adipocytes and the tumor cells as well as the medium, and were replenished in the medium daily.

## Image Reconstruction and Fluorescence Quantification

All images were captured using a Zeiss LSM510 META NLO confocal microscope (Carl Zeiss AG, Göttingen, Germany). For 2D cultures, a 40× oil immersion lens was used, and a single slice was captured. 3D cultures were imaged in 35 mm dishes with a 40× water immersion lens, and *z*-stacks of 2–4 μm optical

slices were captured. 3D reconstruction of optical slices was performed using Volocity software (PerkinElmer, Waltham, MA, USA). For quantification of DQ-IV proteolysis, Volocity software was used to determine integrated intensity per image by dividing the total sum of green fluorescence signal (thresholded using intensity) by the number of nuclei in the image. At least three replicates were measured per condition. The totals in each group were averaged and shown as mean ± SEM.

## Measurement and Quantification of Invasion

Invasion cultures in 60 mm dishes were imaged using a 20× water immersion lens, and *z*-stacks of 4 μm increments were captured. 3D reconstruction of optical slices was performed using Volocity software. For quantification of distance traveled by the tumor cells, images were opened in Zeiss LSM Image Browser. After removing the DIC channel, the Ortho tool was used to select and measure the distance from the middle of PC3 (red) layer to the middle of the adipocyte (green) layer. Distances between PC3 and adipocyte layers were recorded from at least three biological triplicates. For inhibitor studies, distances were shown as percent of distance at baseline (4-h control).

## Statistical Analyses

All data analyses were performed using GraphPad Prism Software version 6.05. Data were presented as mean ± SD and statistically analyzed using unpaired Student's *t*-test. For three or more groups, one-way analysis of variance was used.

## RESULTS

## Three-Dimensional Culture of Prostate Tumor Spheroids in the Absence or Presence of Bone Marrow-Derived Adipocytes

Adipocytes are metabolically active cells that have been shown to promote growth and aggressiveness of several cancers, particularly those that grow in adipocyte-rich microenvironments

such as the breast or ovary (10, 35, 36). The fat cells residing specifically in the bone marrow have been increasingly credited with the ability to support and promote metastatic growth in bone (4–9). We have shown previously that exposure of prostate tumor cells to marrow adipocyte-supplied factors increases their proliferation and invasiveness (5). Here, to better mimic physiological conditions and to allow exploration of molecular mechanisms of adipocyte involvement in metastatic tumor growth in bone, we developed a cell culture model that allows 3D growth of prostate cancer cells exposed to adipocytes.

Seeding of the single-cell suspensions of PC3 prostate carcinoma cells on rBM-coated coverslips with 2% rBM overlay results in a formation of tumor spheroids (**Figure 1**). Cells cultured in their normal growth medium (CONTROL) form compacted, rounded spheroids after 3-day culture (**Figure 1B**, top panels). In contrast, tumor cells grown in a Transwell coculture with marrow adipocytes (TRANSWELL) form much larger and more disorganized clusters (**Figure 1C**, top panels). Longer, 5-day culture of PC3 cells results in further growth under either of the conditions, with Transwell cultures exhibiting significantly more robust proliferation and acquiring much more disordered morphology compared with control cultures (**Figures 1B,C**, bottom panels). Quantification of 3D volumes of the spheroids confirms the growth-promoting effects of adipocytes on tumor cells and indicates potential utility of this experimental system for interrogation of adipocyte–tumor cell interactions (**Figure 1D**).

## Modeling Metabolic Responses to Marrow Adipocytes

One way that adipocytes can affect tumor cell behavior is by regulating cancer cell metabolism (37). Adipocyte-supplied lipids have been shown to feed into the glycolytic pathway and (38–40) and induce the Warburg effect in tumor cells (41–45). One of the master regulators of metabolic reprograming is hypoxia inducible factor (HIF), a key driver of hypoxic stress (46). Bone marrow adipocytes have a capability of inducing HIF-1α signaling in prostate tumor cells grown in 2D monolayer culture (Podgorski

et al., unpublished results1 ). It is, however, well-recognized that tissue dimensionality and associated oxygen status have a potential to profoundly affect hypoxic response in tumor cells (19, 47). Therefore, we examined the utility of our 3D Transwell model to measure hypoxia response to adipocytes. Our 3D cultures established on glass coverslips are easily fixable and suitable for immunocytochemical analyses. We used carbonic anhydrase 9 (CA9), a HIF-1α target gene (48), as a measure of hypoxic response in our system. The immunofluorescence analysis of CA9 protein revealed a significant increase and typical membrane localization in 2D tumor cell cultures exposed to adipocytes (**Figures 2A,B**), indicating activation of HIF-1α signaling. Interestingly, cells grown in 3D culture showed robust CA9 expression even under control conditions (**Figure 2D**). However, although tumor cells grown in transwell coculture with adipocytes expectedly formed significantly larger, disorganized structures (**Figures 2E,H,I,L,M**), there was no visible increase in CA9 expression compared with spheroids grown under control conditions (**Figures 2D,F,G,J,K**). This was further confirmed by the quantification of CA9 fluorescence/nuclei in each of the 3D structures (data not shown). 3D growth alone appears to activate HIF-1α signaling, which is consistent with the reports of heterogeneous oxygen distribution in 3D cultures (47). This

speaks to the importance of considering tissue architecture while examining metabolic responses within the tumor microenvironment. Further studies are underway to validate these findings.

One important consequence of HIF-1α activation is the induction of glycolytic phenotype in tumor cells. We performed immunofluorescence analysis of expression and localization of hexokinase-2 (HK2), a critical enzyme in the first step of glycolysis, that elicits its functions by binding to the voltage-dependent anion channels (VDAC) in the outer mitochondrial membrane (49, 50). Low HK2 expression was detected in monolayer cultures of PC3 cells (**Figure 3A**), with robust increases in HK2 fluorescence upon exposure to adipocytes (**Figure 3E**). This is consistent with adipocyte-induced Warburg phenotype in these cells (see footnote text 1). Mitotracker labeling (**Figures 3B–D,F–H**) showed no observable differences in the number of mitochondria between the control and Transwell conditions. The majority of the HK2 appeared to co-localize with mitochondria (**Figures 3C,G**), which is consistent with the literature evidence demonstrating that approximately 80% of total HK2 is bound to the mitochondrial VDAC (51).

In stark contrast to the monolayer cultures, a significant HK2 expression was observed in PC3 3D spheroids even under control conditions (**Figures 3I,K–N**). This indicates a potential enhancement of glycolytic phenotype, consistent with activation of HIF-1α signaling by 3D culture (**Figure 2**). High HK2 fluorescence closely mirroring mitochondrial pattern was also observed in 3D cultures grown under Transwell conditions (**Figures 3J,P–S**).

<sup>1</sup>Diedrich J, Rajagurubandara E, Herroon MK, Mahapatra G, Huttemann M, Podgorski I (unpublished results).

However, HK2 fluorescence intensity was not significantly higher than in control cultures, indicating no significant enhancement of HK2 levels by coculture with adipocytes (data not shown).

## Examining Effects of Marrow Adipocyte Presence on Tumor Response to Bone Marrow Macrophages

Bone represents a complex, dynamic microenvironment and establishment of metastatic lesions in the skeleton requires tumor cell interactions with a number of cell types in the marrow. To examine the suitability of our cell culture system to study multi-cell interactions, we incorporated BMMs into 3D tumor spheroids growing under control and Transwell conditions (**Figure 4A**). Pre-labeling of PC3 tumor cells with CTO allowed visualization of the tumor growing in 3D culture (**Figures 4B,C**). We tracked the incorporation of BMMs into the tumor spheroid by F4/80 immunofluorescence staining. Tumors growing under Transwell conditions with adipocytes attracted more macrophages than those cultured under control conditions (**Figures 4B,C**; green fluorescence). Incorporation of BMMs into the spheroid was also visible in the optical slices taken in the middle of the spheroid (**Figures 4D–G**). These observations suggest that adipocytes might be inducing phenotypic changes in the tumor cells and/or macrophages to promote the BMM affinity for the tumor. Further studies are needed to confirm this hypothesis.

## Imaging Live Tumor Cell-Driven Proteolysis in the Presence of Marrow Adipocytes

To grow and thrive in their surrounding environments, tumors have to interact with the extracellular matrix (ECM) via biochemical and mechanical processes. Proteolysis of the ECM is one of the key aspects of tumor invasion (52, 53), and we have previously shown that exposure to adipocyte-derived factors increases invasive potential of prostate carcinoma cells (5). One way to effectively visualize ECM proteolysis by the tumor cells and other cell types residing in the tumor microenvironment is through a confocal microscopy assay utilizing quenchedfluorescent (DQ) proteins, an approach developed by the Sloane laboratory over a decade ago (53). We have previously utilized this technique to demonstrate that prostate carcinoma cells are capable of degrading quenched fluorescent derivatives of collagen IV (basement membrane) and collagen I (organic matrix of the bone) (33). Therefore, we used this approach to determine whether exposure to adipocyte-derived factors affects the collagen IV degradation in our model. PC3 cells plated on rBM containing DQ-IV were allowed to establish into spheroids by growing for 3 days under control or Transwell conditions before adding CTO-labeled BMMs and allowing cultures to continue for additional 2 days (**Figure 5A**). Degradation of the DQ-collagen IV, observed as green fluorescence, was captured and quantified at the end of 5-day culture. In agreement with our previous findings (33), fluorescent cleavage products of DQ-collagen IV were found both intracellularly and pericellularly (**Figures 5B–E**). Similar levels of fluorescence were detected in PC3 cells cultured alone (without BMMs) regardless if they were under control or Transwell conditions (**Figures 5B,D,F,H,J**), suggesting no measurable contribution of adipocytes to tumor-driven proteolysis.

Since macrophages are important sources of proteases and key contributors to tumor-associated proteolysis (34, 53), we also measured and quantified the DQ-IV fluorescence following the 2-day coculture of pre-established tumor spheroids exposed to macrophages (**Figures 5C,E,G,I,J**). Surprisingly, despite the significant infiltration of BMMs into tumor spheroids exposed to adipocytes in a Transwell coculture and visible BMM-associated proteolysis (**Figures 5E,I**), there was no significant increase in overall tumor-associated DQ-IV degradation compared with cultures grown without macrophages (**Figure 5J**).

Interestingly, when macrophages were incorporated into the culture at the time of seeding the tumor cells (**Figure 6A**), significant changes were observed in the morphology of the resulting spheroids and in the tumor-associated proteolysis. Specifically, under control conditions, addition of BMMs led to formation of less compacted and more disorganized tumor spheroids compared with tumor cells cultured alone (**Figures 6B,F** vs. **Figures 6C,G**). This change in morphology was also associated with a significant increase in DQ-IV proteolysis (**Figure 6J**). Even more pronounced effects of BMMs were observed in cultures exposed to adipocytes in the Transwell system (**Figure 6D,H** vs. **Figures 6E,I**). Formation of highly disorganized tumor cell/ BMM clusters with high levels of DQ-IV fluorescence (**Figure 6J**) suggested potentially important contribution of adipocytes to tumor- and BMM-driven proteolytic activity in this model.

## Quantifying Tumor Cell Invasion Toward Adipocytes in Three-Dimensional System

Our previous work has shown that prostate carcinoma cells exposed to media conditioned by marrow adipocytes take up the fat cell-supplied lipids in a process that results in accelerated tumor growth and invasiveness (5). Prostate tumor cells have been reported to be attracted to adipocyte-rich areas of the bone marrow, where metastases commonly occur (54, 55), and the translocation of adipocyte-stored lipids has been linked to increased tumor cell motility (56). The mechanisms behind the tumor cell attraction to marrow fat cells, however, remain poorly understood. Here, we have designed a simple 3D system in which tumor cell invasion toward adipocytes can be easily visualized, measured, and quantified (**Figure 7A**). We used collagen I, the predominant component of bone, as the matrix separating BODIPY-labeled bone marrow adipocytes (bottom) and DsRedexpressing PC3 cells (top). Tumor cells invaded through collagen I matrix over the course of 96 h (**Figures 7B–E**), with complete merging of the two cell types at the 96-h time point.

We next tested two compounds with potential to affect adipocyte–tumor cell interactions: a selective inhibitor of fatty acidbinding protein 4 (FABP4) (BMS309403) (5), and Atglistatin, a selective inhibitor of Adipocyte Triglyceride LIpase (ATGL) (57). We chose to inhibit FABP4 based on our previous studies demonstrating upregulation of this lipid transporter in prostate tumor cells exposed to marrow adipocytes, and the apparent efficacy of BMS309403 in inhibiting prostate tumor cell invasion in the 2D Boyden chamber assay (5). We also used Atglistatin as the means of inhibiting adipocyte-driven lipolysis, which we have shown to be induced by prostate tumor cells (see footnote text 1). As expected, under control conditions, the distance between PC3 cells and marrow adipocytes was significantly reduced after 70 h, compared with the baseline at 4 h (**Figures 8A,B**). Atglistatin alone did not seem to have any effect on tumor cell invasion toward the adipocytes (**Figure 8C**). Despite potent inhibition of PC3 invasion in 2D (5), only modest inhibition of tumor invasion was observed with FABP4 inhibitor in a 3D assay (**Figure 8D**), a result potentially due to differences in matrix and experimental conditions between the two assays. Notably, however, the use of both inhibitors in combination significantly inhibited tumor invasion and restored the distance between the two cell types to the levels measured at 4 h (**Figures 8E,F**). This suggests a potential link between lipid hydrolysis and transport between adipocytes and tumor cells that needs further investigation.

## DISCUSSION

The functional role of bone marrow adipocytes in the growth and aggressiveness of skeletal tumors is an understudied and not well-understood area of cancer research. There is a great need for *in vitro* 3D models that can address the complexity of the bone tumor microenvironment, yet allow dissection of the mechanisms behind the contribution of specific cell types to the metastatic growth. There has been a growing effort to generate 3D models that would represent a bridge between the 2D monolayer cultures and *in vivo* tumor microenvironment, with the idea of capturing the physiological complexity, multiplicity of cell types, ECM composition, and temporal and spatial distribution of soluble factors in the bone microenvironment (21, 28, 58–61). Many of these models have been key to demonstrating the importance of 3D tumor architecture in mechanistic understanding of basic cancer biology and, especially, in evaluating tumor response to therapy (22, 58, 61). Regretfully, however, none of these models to date have addressed the contribution of fat cells, one of most abundant cell types in the adult bone marrow to growth, phenotype, and behavior of metastatic tumor cells in bone.

The *in vitro* approaches we describe herein provide a simple way of examining interactions between bone marrow adipocytes and metastatic tumor cells in a physiologically relevant manner. Our Transwell system combines 3D culture of bone marrow-derived adipocytes with 3D culture of prostate tumor cells to allow paracrine interactions and sharing of nutrients and secreted factors by the two cell types. We show via immunofluorescence analysis of metabolism-associated proteins that under 3D conditions tumor cells have significantly different metabolic responses to adipocytes than tumor cells grown in a monolayer culture. This underlines the importance of employing 3D culture conditions to mimic physiological interactions between marrow adipocytes and metastatic tumor cells. It is noteworthy that the simple design of our 3D system allows easy manipulation, processing, and analysis of the resulting cultures. Because tumor cell cultures are established on glass coverslips they are immediately available for immunocytochemistry and confocal imaging. These samples can also be recovered from the coverslips and processed for protein and RNA analyses (data not shown), allowing for additional data to support the confocal imaging results. We have also shown the suitability of this system to image proteolysis by living cells. We used DQ Collagen IV as a representative component of the basement membrane, but this approach can be easily modified for use with other DQ-labeled matrices, such as DQ Collagen I [a major protein in the bone (33, 62)], or with activity-based probes (63).

Contribution of other bone marrow-derived cell types is an important consideration in the design of physiologically relevant cell culture models. Our data show that BMMs can be easily incorporated into our coculture system. More importantly, we reveal that macrophages are attracted to tumor spheroids exposed to adipocyte-derived factors, and their coculture with tumor cells results in a more aggressive morphology and an enhanced ECM proteolysis by both cell types. This speaks to the importance of multi-cellular design in modeling the effects of specific cell types on tumor phenotype and behavior. Our system allows the manipulation of the design to vary cell types, numbers, and ECM components. Studies are currently ongoing to examine the effects of direct interaction of marrow macrophages with adipocytes on tumor cell growth and aggressiveness.

Previous studies have indicated that prostate tumor cells are attracted to adipocyte-rich areas in the bone marrow (54, 55). Our 3D invasion assay results confirm this attraction and provide a simple approach to evaluate potential targets involved in this process. We utilized our invasion system to inhibit the lipid transporter FABP4, and the main adipocyte lipase ATGL and revealed that ATGL-driven lipolysis and FABP4-driven lipid transport might be important for tumor cell–adipocyte interaction in the bone marrow microenvironment. This not only opens new avenues for investigation but also speaks to the potential utility of this assay as an initial screen to identify factors behind tumor cell invasiveness toward the marrow adipocytes. Many G proteincoupled receptors (GPRCs), including chemokine receptors and receptors of bioactive lipids, have been implicated in tumor cell growth and invasiveness (64). Given the fact that adipocytes are an abundant source of chemokines and lipids, their role in modulating the invasive capacity of tumor cells via GPCRs could be studied in this system. It is also important to mention that this approach is easily modifiable in terms of matrix type and thickness, time of invasion and cell types used, representing a useful

## REFERENCES


tool to examine the functional effects of marrow adipocytes on tumor cell invasiveness.

We acknowledge that our models have limitations. First, we are utilizing adipocytes differentiated from bone marrow mesenchymal cells, which may not be entirely representative of adipocyte population in bone. Emerging data now show that there might be differences in bone marrow fat depending on localization in the bone (17), and we would not be able to capture these differences with our model. Second, we are utilizing murine bone marrow cells in combination with human tumor cells, which may not be entirely clinically relevant. Species-specific osteotropism is an important factor to be considered in design of physiologically relevant models (65). This would be particularly important when utilizing this model to study the effects of adipocyte-derived cytokines and chemokines on tumor cells. Adipokines, such as hepatocyte growth factor (HGF) or interleukin-6 (IL-6) are known to activate their receptors in a species-specific manner (66, 67). Nevertheless, it is important to mention that murine cells in our system can be easily replaced with human derived mesenchymal bone marrow cells to eliminate this issue.

Our model does not entirely recapitulate the bone tumor microenvironment found in the patient with metastatic disease, but it provides a controllable, testable system for examination of the molecular mechanisms behind adipocyte involvement in tumor progression in bone. We believe that the models described herein provide a good compromise between currently utilized 2D adipocyte-tumor cell cocultures and *in vivo* mouse models, where dissection of the specific contribution of adipocytes is not possible. Further development of these *in vitro* culture systems and combining them with other bone metastasis-focused cell culture models will open new strategies to increase our understanding of the role of marrow adipose tissue in metastatic progression.

## AUTHOR CONTRIBUTIONS

MH carried out most of the data acquisition and analyses and participated in writing of the manuscript. JD carried out some data acquisition and participated in data analyses and writing of the manuscript. IP wrote the manuscript and participated in some of the data acquisition and analyses.

## ACKNOWLEDGMENTS

The authors thank Dr. Kamiar Moin and the Microscopy, Imaging and Cytometry Resources Core (MICR) for assistance with confocal microscopy analyses. Grant support was provided by NIH/ NCI 1 R01 CA181189-01 (IP, PI), NIH 5T32CA009531-29 (JD, trainee), and P30 CA 22453 (MICR).


**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 © 2016 Herroon, Diedrich and Podgorski. 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.*