Schizophrenia Research Forum - A Catalyst for Creative Thinking

Epigenetic Analysis Finds Widespread DNA Methylation Changes in Psychosis

19 March 2008. A genome-wide scan for alterations in DNA methylation has found several differences in brain tissue from people with schizophrenia and bipolar disorder compared to unaffected people. The observed changes, affecting one in 200-odd genes surveyed, are consistent with the epigenetic theory of schizophrenia, which holds that the pathophysiology of schizophrenia could stem from changes in gene expression driven by aberrant chromatin structure, rather than from genetic changes in DNA sequence. The study, from the lab of Arturas Petronis and colleagues at the Center for Addiction and Mental Health in Toronto, Ontario, Canada, did not pinpoint any one or two genes, but produced a host of candidates for further study. The work appears in the March issue of the American Journal of Human Genetics.

Epigenetic influences on gene expression include methylation, acetylation, and other modifications (see SRF Current Hypothesis by Grayson and colleagues, and SRF related news story). Methylation at cytosine-guanine (CG)-rich "islands" surrounding genes is important for silencing, and is necessary to create stable configurations of gene expression in differentiated cells. A recent paper analyzing methylation at 50 genetic loci in human cerebral cortex showed that DNA methylation is dynamic over time, involves differentiated neurons, and affects a substantial portion of genes (Siegmund et al., 2007). In schizophrenia, changes in the methylation status of a few genes have been reported (see SRF related news story), but the new study is the first large-scale look at changes in methylation in the disease.

For the initial analysis, first author Jonathan Mill and colleagues extracted genomic DNA from postmortem brain tissue of 105 subjects (from the Stanley Medical Research Institute collection), of which one-third had schizophrenia, one-third had bipolar disorder, and one-third were demographically matched controls. The unmethylated fraction of DNA was enriched using methylation-sensitive restriction enzymes and hybridized to gene chip arrays covering 12,000 CpG islands, spanning the GC-rich regions around genes where methylation occurs. The relative hybridization signals indicated genes that were hyper- or hypomethylated in the different groups.

Changes in methylation were quite common between affected and unaffected subjects, and the affected genes included many involved in glutamatergic and GABAergic neurotransmission, and neuronal development. There were sex differences in methylation, but males and females with schizophrenia had similar patterns overall. The researchers singled out hypermethylation in two genes in this regard: RPP21, a component of ribonuclease P involved in tRNA maturation, and KEL, a blood group glycoprotein, whose abnormal expression has been tied to symptoms of schizophrenia. Both genes were also hypermethylated in females with bipolar disorder. However, in the bipolar group, there was no significant correlation of methylation between males and females, suggesting stronger sex-specific factors at play in that disease. No association was found between demographics and methylation at any one gene, with the exception that hypomethylation upstream of the MEK1 gene correlated with lifetime antipsychotic use in the male schizophrenia group. A network analysis of the methylation data in cases versus controls suggested that there might be a widespread epigenetic imbalance in major psychosis.

Methylation appeared to reflect lower gene expression in some cases where mRNA analysis had been done on the same Stanley collection samples for other studies. Comparing methylation data with gene expression data, the investigators found that 82 percent of loci that were hypermethylated had lower gene expression in at least one study, with one quarter showing significant downregulation across several studies. For hypomethylation, the concordance with higher gene expression was not as tight.

From the genes implicated in the chip analysis, the researchers chose 10 for a closer look at the sites of methylation. They did this by treating the DNA with sodium bisulfite, which converts normal cytosine residues to thymidine, leaving the methylated cytosines intact, and then quantifying the degree of methylation across selected regions by Pyrosequencing. The results confirmed the chip analysis in a number of loci.

Using the sequencing approach, the investigators also looked at methylation in a set of candidate genes chosen for their implication in schizophrenia. They failed to find psychosis-associated methylation differences in any of these genes, which included two, COMT and RELN, where hypomethylation had been reported before (see SRF related news story and Grayson et al., 2005). The reason for this discrepancy is unclear. Identification of epigenetic changes could be confounded by variations in the tissue samples, where distinct neuronal types have their own methylation profiles (Ruzicka et al., 2007).

These types of epigenetic studies are still in their infancy, and much remains to be done. Going forward, the authors conclude that, “The unbiased microarray approach was far more productive in identification of differentially methylated loci than was the focused candidate-gene approach; this has implications for the design of future epigenetic studies for complex disease.”—Pat McCaffrey.

Mill J, Tang T, Kaminsky Z, Khare T, Yazdanpanah S, Bouchard L, Jia P, Assadzadeh A, Flanagan J, Schumacher A, Wang SC, Petronis A. Epigenomic profiling reveals DNA-methylation changes associated with major psychosis. Am J Hum Genet. 2008 Mar;82(3):696-711. Abstract

Comments on News and Primary Papers
Comment by:  Dennis Grayson
Submitted 26 March 2008
Posted 27 March 2008
  I recommend the Primary Papers

The paper by Mill et al. is one of the first comprehensive attempts to examine changes in methylation across the entire genome in patients with various diagnoses of mental illness. The study is well designed, extensive, and uses fairly new technology to examine changes in methylation profiles across the genome. In the frontal cortex, the authors provide evidence for psychosis-associated differences in DNA methylation in numerous loci, including those involved in glutamatergic and GABAergic transmission, brain development, and other processes linked with disease etiology. Methylation in the frontal cortex of the BDNF gene is correlated with a non-synonymous SNP previously associated with major psychosis. These data provide further support for an epigenetic origin of major psychosis, as evidenced by DNA methylation-induced changes likely important to gene expression.

In many ways, this seems reminiscent of the trend in genetics several years ago when the inclination was to move from single gene loci association and linkage studies to genomewide scans. The only downside of the approach is that what one gains in information, one (at least initially) loses in biology. That is given the wealth of new findings uncovered; we now need to go back and examine these results in light of what we know regarding gene function in neurobiology and cognition. Of course, this is the trend, now that microarrays have increased our capacity to look at all things at the same time. The flipside is that it will take several large-scale studies of this sort to better understand which findings are replicable and which are not. That is, do the results of the Mill paper agree with data obtained and carried out by laboratories using the methyl DIP or MeCP2 ChIP assays coupled with microarrays. While these experiments ask different questions, the implication is that there may be some degree of overlap in comparing these different methodologies. While this may be premature, there is a sense that this information will be available shortly.

Finally, I would like to focus on recent findings regarding the methylation of the reelin promoter. These authors (Mill et al.) and Tochigi and colleagues (Tochigi et al., 2008) have found that the reelin promoter is not hypermethylated in patients with schizophrenia. In fact, Tochigi et al., 2008, found that the reelin promoter is not methylated at all using pyrosequencing. However, several groups (Grayson et al., 2005; Abdolmaleky et al., 2005; Tamura et al., 2007; Sato et al., 2006) have shown that the human reelin promoter is methylated in different circumstances. Interestingly, there is little consensus in the precise bases that are methylated in these latter studies. Our group (Grayson et al., 2005) performed bisulfite treatment of genomic DNA and sequencing of individual clones. Moreover, we analyzed two distinct patient populations. The clones were sequenced at a separate facility. What was intriguing was that the baseline methylation patterns in the two populations was different, and yet several sites stood out as being relevant in both. We mapped methylation to the somewhat rare CpNpG sites proximal to the promoter. Interestingly, these bases were located in a transcription factor-rich portion (Chen et al., 2007) of the promoter and in a region that shows 100 percent identity with the mouse promoter over a 45 bp stretch. We have also been able to show that changing one of these two bases to something other than cytosine reduces activity 50 percent in a transient transfection assay. So the question becomes, How do we reconcile these disparate findings regarding methylation? As suggested by Dr. McCaffrey, the answer may lie in regional differences that arise due to the nature of the material available for each study. We have found a degree of reproducibility by using human neuronal precursor (NT2) cells for many of our studies. At the same time, this cell line is somewhat artificial and cannot be used to reconcile differences found in human tissue. Perhaps it might be prudent to examine material taken by using laser capture microdissection to enrich in more homogenous populations of neurons/glia. In moving ahead, it might be best to now focus on the mechanism for these differences in methylation patterns and try to understand the biology associated with the new findings (Mill et al., 2008) as a starting point.


Abdolmaleky HM, Cheng KH, Russo A, Smith CL, Faraone SV, Wilcox M, Shafa R, Glatt SJ, Nguyen G, Ponte JF, Thiagalingam S, Tsuang MT. Hypermethylation of the reelin (RELN) promoter in the brain of schizophrenic patients: a preliminary report. Am J Med Genet B Neuropsychiatr Genet. 2005 Apr 5;134(1):60-6. Abstract

Chen Y, Kundakovic M, Agis-Balboa RC, Pinna G, Grayson DR. Induction of the reelin promoter by retinoic acid is mediated by Sp1. J Neurochem. 2007 Oct 1;103(2):650-65. Abstract

Grayson DR, Jia X, Chen Y, Sharma RP, Mitchell CP, Guidotti A, Costa E. Reelin promoter hypermethylation in schizophrenia. Proc Natl Acad Sci U S A. 2005 Jun 28;102(26):9341-6. Abstract

Mill J, Tang T, Kaminsky Z, Khare T, Yazdanpanah S, Bouchard L, Jia P, Assadzadeh A, Flanagan J, Schumacher A, Wang SC, Petronis A. Epigenomic profiling reveals DNA-methylation changes associated with major psychosis. Am J Hum Genet. 2008 Mar 1;82(3):696-711. Abstract

Sato N, Fukushima N, Chang R, Matsubayashi H, Goggins M. Differential and epigenetic gene expression profiling identifies frequent disruption of the RELN pathway in pancreatic cancers. Gastroenterology. 2006 Feb 1;130(2):548-65. Abstract

Tamura Y, Kunugi H, Ohashi J, Hohjoh H. Epigenetic aberration of the human REELIN gene in psychiatric disorders. Mol Psychiatry. 2007 Jun 1;12(6):519, 593-600. Abstract

Tochigi M, Iwamoto K, Bundo M, Komori A, Sasaki T, Kato N, Kato T. Methylation status of the reelin promoter region in the brain of schizophrenic patients. Biol Psychiatry. 2008 Mar 1;63(5):530-3. Abstract

View all comments by Dennis Grayson

Comments on Related News

Related News: SfN 2005: Nature or Nurture—Epigenetics in Neuronal Responses

Comment by:  Robert Fisher
Submitted 24 December 2005
Posted 3 January 2006

Dr. Eric Nestler at UT Southwestern Medical Center, Dallas, has long postulated addiction as a 50/50 percent split between genetic predisposition and biological changes as a result of substance abuse and the brain's accommodation to the same. I have found this particular hypothesis to be quite useful in treating addicts and alcoholics in inpatient and outpatient settings. They usually are able to easily grasp the concepts on a reasonably scientific level. This approach allows them to avoid guilt, shame, ect., that are typically strong predictors of success or failure in treatment. It is an even more successful tool in working with co-occurring Axis I disorders.

View all comments by Robert Fisher

Related News: New Genetic Variations Link Schizophrenia and Bipolar Disorder

Comment by:  Mary Reid
Submitted 28 September 2006
Posted 29 September 2006

It's of interest that Vazza and colleagues suggest that 15q26 is a new susceptibility locus for schizophrenia and bipolar disorder. I have suggested that reduced function of the anti-inflammatory SEPS1 (selenoprotein S) at 15q26.3 may reproduce the neuropathology seen in schizophrenia.

View all comments by Mary Reid

Related News: New Genetic Variations Link Schizophrenia and Bipolar Disorder

Comment by:  Patricia Estani
Submitted 5 October 2006
Posted 6 October 2006
  I recommend the Primary Papers

Related News: Epigenetic Forces May Blaze Divergent Heritable Paths From Same DNA

Comment by:  Shiva SinghRichard O'Reilly
Submitted 2 February 2009
Posted 3 February 2009

The methylation difference between twins is clearly demonstrated using newer methods in this publication. However, conceptually it’s an old story now. A quick PubMed search for "monozygotic twins and non-identical" yielded a total of 7,653 publications. There is no doubt that the more we look, the more difference we will find between monozygotic twins. Also, monozygotic twin differences in methylation and gene expression are expected to increase with age. It is also affected by a variety of genetic and environmental factors. We have come a long way in genetic research on twins and the time has come to modify our thinking about monozygotic twins as "non-identical but closest possible" rather than as "identical." They started from a single zygote, but have diverged during development and differentiation including upbringing.

The implication of the published results is that the methylation (epigenetic) differences (in monozygotic twins) will be powerful in any genetic analysis of disease(s). Once again, it is probably more problematic than usually assumed. Also, it is particularly problematic for behavioral/psychiatric disorders including schizophrenia. The reason is multi-fold and includes the effect of (known and unknown) environment including pregnancy, upbringing, drugs, life style, food, etc. All these are known to affect DNA methylation and gene expression. As a result, they add unavoidable confounding factors to the experimental design. It does not mean that epigenetics is not involved in these diseases. Rather, directly establishing a role for methylation in schizophrenia will be challenging. A special limitation is the fact that methylation is known to be cell-type specific, and perfectly matched affected and normal (twin) human brain (region) samples for necessary experiments are problematic and methylation studies on other cell types may or may not be informative.

View all comments by Shiva Singh
View all comments by Richard O'Reilly

Related News: Twins Tell Story of Epigenetic Alterations in Schizophrenia and Bipolar Disorder

Comment by:  Schahram Akbarian
Submitted 7 October 2011
Posted 7 October 2011

The genetic risk architecture is still very difficult to "capture" for a large majority of patients diagnosed with schizophrenia or related diseases. Therefore, studies like that of Dempster et al., who profiled DNA methylation (a type of "epigenetic" modification of cytosine that residues mostly at sites of CpG dinucleotides in the genome) in blood cells of monozygotic twins discordant for schizophrenia, provide an important additional layer of information. The idea is that the disease process in the affected twin leaves behind a molecular signature (in the study by Dempster et al., this would be a change in DNA methylation) that is not found in the healthy twin, with the implication that this signal is related to disease etiology or disease process and treatment, etc.

Dempster and colleagues screened approximately 20 twin pairs. I believe the Illumina bead system they used probes primarily annotated promoters; on a genomewide level, they found, overall, quite subtle changes. One of the more prominent findings is hypomethylation of one specific CpG dinucleotide associated with ST6GALNAC1, a gene regulating protein glycosylation, with additional changes in a dozen or so genes. Hypomethylation of ST6GALNAC1 (which the authors further verified in postmortem brains of subjects with schizophrenia) at sequences proximal to promoters is generally associated with increased expression, but it is not clear if this is the case in the twin blood or the postmortem brain. Interestingly, as the authors point out, the same gene (ST6GALNAC1) may harbor an excess copy in some subjects on the psychosis spectrum. This is a good example of the possibility that some of the genes that are potentially linked to psychosis because of DNA sequence and copy number changes may also show up in epigenetic studies such as that of Dempster and colleagues. Whether a genetic variation or mutation somewhere else in the genome "drives" the epigenetic changes at ST6GALNAC1 and other genes is, of course, hard to prove.

The most important challenge to the epigenetics and psychosis fields is, at least in my opinion, that of reproducibility and independent replication. In other words, will we be able to replicate, in independent studies, epigenetic alterations in blood or postmortem brain tissue, reported to be "highly significant" for cohorts comprising a few dozen or fewer cases? With fewer than five published studies (to the best of my knowledge) that measured DNA methylation on a genomewide scale in mood and psychosis spectrum disorders, it is still hard to predict whether DNA and histone modification mapping will provide valuable clues to the underlying neurobiology of disease. I take an optimistic view, and would like to predict that DNA methylation and histone modification mapping will provide a very important additional layer of information when paired with whole-genome sequencing and transcriptome (RNA expression) profiling.

View all comments by Schahram Akbarian