Edited by: Marnie Blewitt, Walter and Eliza Hall Institute of Medical Research, Australia
Reviewed by: Claudio V. Mello, Oregon Health & Science University, USA; Claudia Vianna Maurer-Morelli, University of Campinas, Brazil
Specialty section: This article was submitted to Neurogenomics, a section of the journal Frontiers in Neurology
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.
Autism spectrum disorders (ASD) are a heterogeneous group of neurodevelopmental disorders characterized by problems with social communication, social interaction, and repetitive or restricted behaviors. ASD are comorbid with other disorders including attention deficit hyperactivity disorder, epilepsy, Rett syndrome, and Fragile X syndrome. Neither the genetic nor the environmental components have been characterized well enough to aid diagnosis or treatment of non-syndromic ASD. However, genome-wide association studies have amassed evidence suggesting involvement of hundreds of genes and a variety of associated genetic pathways. Recently, investigators have turned to epigenetics, a prime mediator of environmental effects on genomes and phenotype, to characterize changes in ASD that constitute a molecular level on top of DNA sequence. Though in their infancy, such studies have the potential to increase our understanding of the etiology of ASD and may assist in the development of biomarkers for its prediction, diagnosis, prognosis, and eventually in its prevention and intervention. This review focuses on the first few epigenome-wide association studies of ASD and discusses future directions.
Autism spectrum disorders (ASD) are defined diagnostically by impaired social communication, restricted interests, and repetitive behaviors, defined hereafter as endophenotypes. Such endophenotypes are thought to result from disordered neurodevelopment, although the precise etiology is unknown (
The reference standard tools for diagnosing ASD using a multidisciplinary team include the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Statistical Classification of Diseases and Related Health Problems (
Autism spectrum disorders, like all other human conditions and diseases, are likely caused by a combination of genes, environment, and the interaction between the two (Figure
Abbreviation | Name | Protein function |
---|---|---|
ADNP | Activity-dependent neuroprotector homeobox | Vasoactive intestinal peptide, neuroprotective factor, transcription factor (E) |
ANK2 | Ankyrin 2, neuronal | Cytoskeletal and cell membrane protein |
ARID1B | AT rich interactive domain 1B (SWI1-like) | ATP-dependent chromatin remodeller (E) |
ASH1L | Ash1 (absent, small, or homeotic)-like (Drosophila) | Transcriptional activator, cell-cell tight junctions (E) |
ASXL3 | Additional sex combs like 3 (Drosophila) | Possible regulator of transcription (E) |
CHD8 | Chromodomain helicase DNA binding protein 8 | Transcriptional repressor involved in early development (E) |
DYRK1A | Dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A | Protein kinase involved in signaling and early development |
GRIN2B | Glutamate receptor, inotropic, |
Glutamate receptor involved in long-term potentiation and synaptic transmission |
POGZ | Pogo transposable element with ZNF domain | Possible transposase and transcription factor (E) |
PTEN | Phosphatase and tensin homolog (mutated in multiple advanced cancers 1) | Tumor suppressor involved in signaling and mitochondrial function |
SCN2A | Sodium channel, voltage-gated, type II, alpha subunit | Sodium channel expressed in the brain |
SETD5 | SET domain containing 5 | Likely chromatin protein (E) |
SHANK3 | SH3 and multiple ankyrin repeat domains 3 | Postsynaptic density synapse scaffold protein |
SUV420H1 | Suppressor of variegation 4-20 homolog 1 (Drosophila) | Likely chromatin protein (E) |
SYNGAP1 | Synaptic Ras GTPase activating protein 1 | Postsynaptic density synapse protein |
TBR1 | T-box, brain, 1 | Likely transcription factor associated with early cortical development (E) |
It has been proposed that, rather than resulting from dysfunction of specific genes, ASD result from the dysfunction of specific genetic pathways (
Going beyond genetic associations, alterations in other cellular processes have begun to be found in ASD. Studying changes to physiology, gene expression, and the epigenetic states that contribute to ASD phenotype along with genetics, has begun to broaden our understanding of ASD and will eventually, in combination, lead to better methods of diagnosis, prognosis and even treatment of ASD. For the rest of this review, we focus on these mechanisms, briefly reviewing physiology and gene expression and focusing mostly on epigenetics, and in particular, the genome-wide studies of ASD published to date.
Autism spectrum disorders are currently defined on the basis of behavioral observations only. However, evidence is growing rapidly that is beginning to define ASD in physiological terms (reviewed in Ref. (
At the center of the ASD-associated physiological definition is the process of inflammation. Evidence for the association of inflammation with ASD comes from a relationship between familiar autoimmune disorders and ASD, discovery of a physiologically defined shift to a pro-inflammatory state (exemplified by cytokines) in the brain and blood in ASD, from animal models of ASD and from the positive response to immune suppressive medications (e.g., corticosteroids) in those affected by ASD (
Two other main physiological characteristics of ASD have been reported and reviewed: elevated levels of oxidative stress and mitochondrial dysfunction in the blood and brain (reviewed in Ref. (
A further pathway implicated in ASD is intestinal dysbiosis as evidenced by a reduced microbiome complexity in ASD (reviewed in Ref. (
As the expression levels of a gene are largely influenced by the epigenetic state of its surrounding regulatory regions, it is first worth briefly reviewing genome-wide studies of gene expression in ASD. For a more extensive review, please refer to a paper written by Voineagu and colleagues (
Reference | Samples (cases/controls) |
Tissue source | Pathways identified |
---|---|---|---|
( |
3/3 |
LCLs | Neurodevelopment |
( |
49/12 | PBLs | Immune and inflammatory response (mediated by NK cells), cytotoxicity |
( |
15/15 | LCLs | Cell communication, immune and inflammatory response |
( |
20/– | LCLs | Steroid hormone metabolism |
( |
86/30 | LCLs | Steroid hormone metabolism, circadian rhythm |
( |
52/27 | PBLs | immune and inflammatory response (mediated by NK cells) |
( |
20/22 | LCLs | Neurodevelopment, synaptic function (long-term potentiation) |
( |
10/23 | CB, PFC, CN | Synaptic function |
( |
6/6 | TC | Immune and inflammatory response |
( |
19/17 | FC, TC, CB | Synaptic function, immune and inflammatory response |
( |
32/40 | PFC, FC | Microglial function, immune response, neuronal activity |
( |
70/60 | PBLs | Neurodevelopment; signaling; skeletal development |
( |
20/18 | PBLs | Ribosome function, spliceosome function, mitochondrial, immune and inflammatory response, calcium signaling |
( |
170/115 | PBLs | Neurotrophic signaling, notch signaling; synaptic function (long-term potentiation) |
( |
60/68 | PBMCs | Immune and inflammatory response; hemoglobin metabolism |
( |
3/3 |
LCLs | Neurodevelopment, skeletal development, gastrointestinal development, steroid hormone metabolism, circadian rhythm |
Epigenetics describes the molecular factors that form complexes at regulatory regions of DNA to influence genetic activity without changing the primary DNA sequence. Such factors are usually inherited through mitosis but their meiotic (transgenerational) inheritance is controversial due to the expansive epigenetic remodeling that happens twice per generation – during gametogenesis and during very early embryonic development (
DNA methylation is the most widely studied and widely understood epigenetic mark and involves the covalent attachment of a methyl (CH3) molecule to the cytosine of a CpG dinucleotide (
Both genetic and environmental variation, and an interaction between the two, can influence epigenetic change (
A statistically significant association of an epigenetic state with a disorder after its onset could mean one of two things: that the epigenetic state mediates the cause of, or results from, the disorder. Although the former is a more attractive conclusion, evidence must be accumulated that test this hypothesis. For example, if the same epigenetic state were present prior to onset of symptoms, ideally at birth, this would support the hypothesis. Furthermore, if the environmental cause was linked with both epigenetic state and the disease phenotype via an approach such as Mendelian randomization (
Epigenetic state can be used as a biomarker of disease risk, diagnosis, prognosis, and response to treatments (reviewed in Ref. (
Epigenetic modification is increasingly thought to play a role in ASD, based on the findings discussed below. As mentioned previously, genes that play a role in epigenetic pathways constitute a sizable proportion of ASD candidate genes identified through genetic screens (
When we study disease-associated epigenetic marks such as DNA methylation we are usually looking for two things: (1) clues to the cause and mechanisms of a disorder and (2) biomarkers for its risk, diagnosis, or prognosis (
Therefore with ASD, we have a number of options when planning epigenetic studies. We can search for (1) signatures of causative genetic or environmental factors; (2) clues to the physiological mechanisms predisposing or resulting from the onset of ASD; (3) biomarkers that will contribute toward a better prediction of risk prior for ASD prior to diagnosis; (4) biomarkers that will aid in diagnosis, including endophenotyping and (5) biomarkers that may help in predicting symptom severity or diversity. However, before we do this, there are a number of considerations we need to make.
ASD is thought to be primarily a disorder of neurodevelopment involving multiple regions of the brain including frontal, temporal, and occipital lobe cortices and the cerebellum (
Finally, genome-wide methylation profile in buccal epithelium shows a higher similarity than blood in comparison with the brain (
Although they are being superseded by genome-wide approaches (Section “
The oxytocin receptor (OXTR) is a G-protein coupled receptor for the peptide hormone and neurotransmitter oxytocin. It plays a role in anxiety, social memory and recognition, sexual and aggressive behaviors, and maternal-offspring bonding.
Glutamate decarboxylase 1 (
Engrailed-2 (
Reelin (RELN) is a secreted extracellular matrix glycoprotein involved in neuronal migration and positioning in the developing brain and modulates synaptic plasticity in adult brain. It has a score of 2 (“strong candidate”) in the SFARI ASD gene database (
The evidence for the association of each of the above five candidate genes with ASD is summarized in Table
Gene | Genetic evidence | Methylation reference | Diagnostic method | Tissue | Samples |
Largest effect size |
Expression | Protein | Other data |
---|---|---|---|---|---|---|---|---|---|
Weak | ( |
DSM-IV, ADI-R | PBLs | 20/20 | +23% | No | No | Endophenotype |
|
10/10 |
+38.9% | ||||||||
Temporal cortex | 10/10 |
+41.6% |
yes | No | |||||
Weak | ( |
Not given | Cerebellum | 10/10 | +3% |
Yes |
No | Animal models, MECP2 binding | |
Minimal | ( |
DSM-IV | Cerebral cortex | 13/13 | +10–20% |
Yes | Yes | ||
Strong | ( |
Not given | Cerebellum | 10/10 | Not quantifiable | Yes | Yes | MECP2 binding | |
Syndromic | ( |
ADI-R, ADOS | Frontal cortex | 14/14 | +12%, +10% |
Yes | Yes | Animal model |
Seven recent studies have gone beyond candidate genes to study levels of DNA methylation on a genome-wide scale, commonly termed as epigenome-wide association studies (EWAS) (
Epigenome-wide association studies methods fall into two major classes: those based on the affinity of a molecule for methylated cytosine and those based on the sequence difference resulting from conversion of only non-methylated cytosine to uracil using sodium bisulfite (
There are a number of “best practices” when it comes to planning, conducting, and analyzing genome-wide methylation analysis that will assist in the interpretation of the studies reviewed below (reviewed in Refs. (
Even when these conditions are satisfied, findings need to be replicated in independent cohorts. While the use of a different method of genome-wide analysis is desirable, comparison between different platforms can be difficult for reasons including differences in genomic coverage and resolution. One further consideration is the functional relevance of disease-associated differences in DNA methylation. The “best case” scenario is when methylation correlates with expression of a nearby gene that has relevance for the specific disorder, but such associations are often not found or not looked for. Having said that, as discussed earlier, functional relevance is not an absolute requirement for an epigenetic biomarker.
At the design stage, accurate phenotyping and study power are important issues to consider. As with other “omics” studies, bigger is usually better, although the proportion of variance in phenotype explained by single epigenetic variants appears to be much larger than genetic variants (e.g., Ref. (
At the analysis stage, potential confounders, such as age, sex, genetic factors such as ethnicity, biological (e.g., heterogeneity of tissues such as blood), and technical variation, need to be queried. Issues of cause vs effect and which tissue to analyze have already been discussed.
Below we review, to our knowledge, all the genome-wide studies of DNA methylation conducted on ASD samples (summarized in Table
Reference | Samples |
Tissue | Participant age (years) | Diagnostic method | Method of analysis | DMR/DMP analysis |
Effect size cut off |
Adjustment for multiple testing | Validation |
Expression data |
---|---|---|---|---|---|---|---|---|---|---|
( |
9/9 |
Occipital cortex | 1–60 | DSM-IV, ADOS, and/or ADI-R | HM27 | DMP | No | Yes | No |
Yes |
Cerebellar hemispheric cortex | ||||||||||
( |
12/21 | Prefrontal cortex | 17–35 | ADI-R and/or ADOS | HM450 | DMR | No | Yes | No | No |
16/21 | Temporal cortex | 21–40 | Yes |
|||||||
13/21 | Cerebellum | 14–17 | No |
|||||||
( |
11/11 | Anterior cingulate gyrus | 16–51 | ADI-R | HM450 | DMP | >5% difference | Yes | Yes | Yes |
12/12 | Prefrontal cortex | |||||||||
( |
3 |
LCL | 2–19 | ADI-R | MIRA | DMR | No | Yes | Yes | Yes |
( |
5/5 | PBLs | 6–12 | DSM-IV, MINI instrument | MeDIP | DMR | >1.5-fold change | No | Yes | Yes |
( |
6 |
PBLs | 15 | CAST | HM27 | DMP | No | No | No | No |
16/22 | Yes | |||||||||
6/10 |
No | |||||||||
50 |
No | |||||||||
( |
47/48 | Buccals | 1–28 | Not stated | HM450 | DMR | No | Unclear |
Yes | No |
Using DNA from the cerebellar cortex and occipital cortex from nine men with ASD (diagnosed using DSM-IV, ADOS, and/or ADI-R) and nine age-matched controls, Ginsberg and colleagues used HM27 arrays to identify ASD-specific probes at an FDR of 0.05 (
Ladd-Acosta and colleagues analyzed DNA methylation, using Infinium HM450 arrays, in the dorsolateral prefrontal cortex (
The authors focused on the three DMRs with the lowest adjusted
A second temporal cortex DMR was located in the 3′ untranslated region (3′UTR) of the proline-rich transmembrane protein 1 (
A third temporal cortex DMR was located in a region ~3.5 kb upstream of the zinc finger gene
A further DMR was found in cerebellum using the above methods within a 1 kb region around the promoter region of the succinate dehydrogenase complex subunit A flavoprotein pseudogene 3 (
Ladd-Acosta and colleagues also investigated the possibility that differences in cellular heterogeneity between ASD and control brains influenced methylation levels in their data (
Nardone and colleagues analyzed DNA methylation, also using Infinium HM450 arrays, in the anterior cingulate gyrus and prefrontal cortex from 11 and 12 matched ASD (diagnosed using ADI-R) and control pairs respectively (
The first study of genome-wide DNA methylation in ASD on peripheral tissue analyzed DNA from B-cell derived lymphoblastoid cell lines (LCLs) from three pairs of twins discordant for severity of ASD (
Wang and colleagues analyzed DNA methylation, using MeDIP and promoter arrays, in PBLs from five children with DSM-IV-diagnosed ASD and five age- and sex-matched controls (
Using PBLs from 50 pairs of monozygotic (MZ) twins, Wong and colleagues used Infinium HM27 arrays to identify specific differentially methylated CpG sites associated with ASD (
In the within-pair analysis of six pairs of ASD-discordant MZ twins, a ranked list of top 50 DMPs were generated from six MZ twin pairs by combining significance (
A case-control analysis (
To identify any epigenetic differences between sporadic (one twin of each pair with ASD,
Wong and colleagues also analyzed probes associated with the autistic trait scores of social, repetitive behaviors and interests, and communication. DNA methylation at multiple CpG sites were found to be correlated with CAST scores, including one in the putative promoter of the neurexin 1 (
In addition, a small number of genes appeared in more than one of the above analysis. For example,
Berko and colleagues analyzed DNA methylation using HM450 arrays and DNA from buccal epithelium from 47 cases of ASD (diagnostic tool not described) and 48 unaffected controls (
First, the low numbers of cases and controls in all studies are striking and way below the ideal of >100. Replication of potential ASD-specific methylation biomarkers must be attempted in much larger sample numbers. Next, an extremely diverse variety of biological samples were used in the different studies. Although the identification of the same gene in different tissues, methods and studies (namely
Despite their weaknesses, the seven recent studies of DNA methylation in ASD provide some useful insights into its etiology. Firstly, a small number of replicated, potential methylation biomarkers for ASD have emerged. For
Two differentially methylated gene pathways were found in multiple studies: neurodevelopment and immune and inflammatory response (
To summarize, we present a possible model of the findings reviewed above. Figure
The five candidate genes whose methylation correlates with a diagnosis of ASD and for which there is additional evidence of involvement with ASD warrant attempts at replication, ideally in larger sample numbers and multiple tissues. Priority should be given to those with the largest effect sizes (
Studies of genome-wide DNA methylation in ASD need to be larger, phenotyping needs to be standardized between studies, analysis platforms need to move toward true genomic coverage, genetic variation needs to be taken into account, and analytical methods need to be standardized to include differentially methylated probes and regions. In addition, with sufficient sample size, endophenotypes should be studied to reduce the genetic and epigenetic complexity. This will lead to systems-based approaches of integration of data from genomic, epigenomic, transcriptomic, proteomic and other platforms in more accurate statistical models (classifiers) of diagnosis, prognosis and risk estimation. This will also lead to a better understanding of gene–gene (
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.
This work was supported by grants from the Australian National Health and Medical Research Council (grant numbers 437015, 607358); the Financial Markets Foundation for Children (grant number 032-2007). JC and YL are also grateful to past support from the MCRI, which is supported in part by the Victorian Government’s Operational Infrastructure Support Program. We thank Katrina Williams for her useful comments on the manuscript and the two anonymous reviewers for their constructive comments.