Not really identical: Epigenetic differences in monozygotic twins and implications for twin studies in psychiatry


  • F. Nipa Haque,

    Search for more papers by this author
    • F. Nipa Haque is completing her MSc in pharmacology at the University of Toronto with Dr. Albert Wong at the Centre for Addiction and Mental Health (CAMH) in Toronto, Ontario. The focus of her research is gene–environment interaction in the development of neuropsychiatric disease.

  • Irving I. Gottesman,

    Search for more papers by this author
    • Irving I. Gottesman holds the Irving and Dorothy Bernstein Professorship in Adult Psychiatry and is a Senior Fellow in the Department of Psychology at the University of Minnesota Medical School. He is also Sherrell J. Aston professor of Psychology Emeritus at the University of Virginia. One of the founders of post-WW II behavioral genetics, he began his first twin study on the genetics of personality in 1957 which became his doctoral dissertation at the University of Minnesota. His strategies with normal and psychiatric twins have ranged from heritability to “epigeneticability” in partnership with Arturas Petronis and Albert Wong, and to unexpressed genotypes in the offspring of schizophrenic twins with Aksel Bertelsen.

  • Albert H.C. Wong

    Corresponding author
    • Centre for Addiction and Mental Health, 250 College Street, Room 711, Toronto, ON M5T 1R8, Canada.
    Search for more papers by this author
    • Albert H.C. Wong is a neuroscientist and psychiatrist at CAMH, and an associate professor of Psychiatry at the University of Toronto. His main interests are the regulation of behavior by genetic and epigenetic factors, transcription, and gene–environment interactions.

  • How to cite this article: Haque FN, Gottesman II, Wong AHC. 2009. Not really identical: Epigenetic differences in monozygotic twins and implications for twin studies in psychiatry. Am J Med Genet Part C Semin Med Genet 151C:136–141.


Classical twin studies in the field of psychiatry generally fall into one of two categories: (1) those designed to identify environmental risk factors causing discordance in monozygotic (MZ) twins and (2) those geared towards identifying genetic risk factors. However, neither environment nor differences in DNA sequence can fully account for phenotypic discordance among MZ twins. The field of epigenetics – DNA modifications that can affect gene expression – offers new models to understand discordance in MZ twins. In the past, MZ twins were regarded as genetically-identical controls for differing environmental conditions. In contrast, the evolving current concept is that epigenetic differences between MZ twins may modulate differences in diverse phenotype, from disease to personality. In this article, we review some twin studies, and discuss the dynamic interactions between stochastic, environmental, and epigenetic variables that influence neurobiological phenotypes. © 2009 Wiley-Liss, Inc.

“Why, here begins his morning story right: These two Antipholuses, these two so like And the two Dromios, one in semblance-Besides her urging of her wreck at sea-These are the parents to these children Which accidentally are met together.”

William Shakespeare, The Comedy of Errors Act 5, Scene 1, 348–353


In 1865, Francis Galton wrote that twins offer a “means of distinguishing between the effects of tendencies received at birth, and of those that were imposed by the circumstances of their after lives; in other words, between the effects of nature and nurture.”

The standard assumption of genetic studies on twins has been that greater disease concordance rates in monozygotic (MZ) versus dizygotic (DZ) twins is evidence for a genetic susceptibility component [Boomsma et al., 2002; Kato et al., 2005]. However, findings from recent clinical and molecular research have widened this narrow view. In this article, we review clinical studies of twins that have addressed nature versus nurture issues, and how these studies do not account for all the observed phenotype discordance among twin pairs. Next, we discuss epigenetic evidence that although MZ twins have the same or very similar DNA sequence, gene expression and DNA modification patterns can differ significantly. This challenges the conventional paradigm that MZ twins are identical genetic controls in which environment is the only differing variable [Cardno and Gottesman, 2000]. These new insights about epigenetic DNA modifications and their effects on gene expression and phenotype may increase our understanding of diverse phenotypes from personality traits to neuropsychiatric disease. The new paradigm is not one of nature versus nurture, but of a complex and dynamic interaction between DNA sequence, epigenetic DNA modifications, environment, gene expression, and environmental factors that all combine to influence phenotype.

These new insights about epigenetic DNA modifications and their effects on gene expression and phenotype may increase our understanding of diverse phenotypes from personality traits to neuropsychiatric disease. The new paradigm is not one of nature versus nurture, but of a complex and dynamic interaction between DNA sequence, epigenetic DNA modifications, environment, gene expression, and environmental factors that all combine to influence phenotype.

Classical Twin Study Designs: Genes and Environment

Over the last century, Galton's nature versus nurture dichotomy influenced classical twin studies by stimulating two categories of inquiry: (1) studies aimed at identifying environmental risk factors causing discordance between MZ twins [Kato et al., 2005], and (2) studies aimed at identifying genetic determinants of disease or other phenotypes. In the first category is the landmark Minnesota Study of Twins Reared Apart, which challenged the notion that environment plays a significant role in determining a wide variety of phenotypes [Bouchard et al., 1990]. Extensive physical and psychological evaluations of MZ and DZ twin pairs, some separated and raised apart since early childhood, were performed to quantify the degree of phenotypic discordance among MZ twin pairs [Bouchard et al., 1990; Markon et al., 2002]. The intraclass correlation of scores on numerous tests (R) within pairs of twins raised apart: MZA (RMZA) or raised together: MZT (RMZT) was expressed as a ratio (RMZA/RMZT). As expected, adult MZ twins were similar on many physiological and psychological traits. The surprising discovery was that for some phenotypes, this similarity was present to the same degree whether the MZ twins were raised together or not, suggesting that environmental factors had limited influence, at least in producing within-MZ-twin differences. For example, correlations within MZT and MZA twin pairs on personality measurements were almost identical (e.g., RMZA = 0.50 and RMZT = 0.49 on the Multidimensional Personality Questionnaire (MPQ)). The RMZA/RMZT ratio for the MPQ was 1.02, compared with 1.01 for fingerprint ridge counts.

Other studies have explored the role of prenatal and postnatal environmental influences on neuropsychiatric disease penetrance. For example, Torrey et al. 1994 studied MZ twin pairs discordant for schizophrenia and found that only 30% of the MZ twin pairs had neurological or behavioral differences by the age of 5. These results suggest that while significant prenatal and neonatal events may have an influence on adult-onset schizophrenia, these events cannot entirely account for the discordance in diagnosis [Torrey et al., 1994]. The alternate hypothesis is that factors other than DNA sequence and major environmental events affect susceptibility to diseases such as schizophrenia.

Twin studies aimed at determining the magnitude of genetic influence on disease susceptibility have often relied on quantitative estimates of heritability: the proportion of total phenotypic variation attributable to additive genetic effects. Heritability (h2) can be calculated as twice the difference between MZ and DZ concordance, based on the assumption that MZ twins have identical genomes, while DZ twins have half the genetic variation present in unrelated individuals [Boomsma et al., 2002; Visscher et al., 2008]. It is also presumed that MZ and DZ co-twins have a similar degree of difference between their pre- and postnatal environments [Guo, 2001]. A greater phenotype concordance rate in MZ versus DZ twins therefore provides evidence for a genetic component in the phenotype of interest [Boomsma et al., 2002; Kato et al., 2005]. Using this approach, heritability estimates for schizophrenia have been reported to be as high as 80% [Cardno and Gottesman, 2000]; with heritability for bipolar at 62–79% [Bertelsen, 2004], and depression 21–45% [Kendler et al., 1992]. However, these relatively high heritability estimates can obscure the substantial discordance among a large proportion of MZ twin pairs; for schizophrenia, the MZ concordance rate is in the range of 41–65% [Cardno and Gottesman, 2000].

Clinical diagnosis in psychiatry is of course somewhat subjective, but objective brain phenotypes such as volume show similar heritability. Rijsdijk et al. 2005 compared the brain volumes (whole brain, hippocampus, third and lateral ventricles) of MZ twins concordant and discordant for schizophrenia, healthy MZ twins, discordant sibling pairs, concordant sibling pairs, and healthy control subjects. These comparisons generated heritability estimates of 88% for whole-brain volume, while for lateral ventricle size, 67% of the variation was attributed to common environmental effects [Rijsdijk et al., 2005]. Brain morphology changes in schizophrenia twins were also examined in a recent Dutch study, which reported that intracranial and whole-brain corrected frontal lobe volumes are smaller in discordant MZ twins compared to healthy MZ twins [Baare et al., 2001]. This study also reported that both MZ and DZ discordant twins have smaller whole-brain, parahippocampal, and hippocampal volumes than healthy twins, and that affected twins have yet smaller whole-brain volumes than their non-schizophrenic co-twins. The hippocampal size findings were replicated in a Finnish twin study [van Erp et al., 2004]. Together, these observations indicate the presence of a genetic influence on brain volumes, but the additional reduction in whole-brain volume observed in the affected compared to the unaffected co-twins suggests that other factors must also modulate brain volume [Baare et al., 2001].

It is now clear that structural genes comprise only a small proportion of the genome (∼1%) [Lander et al., 2001], and that phenotype can be profoundly affected by changes in the amount, timing, and location of gene transcription, which are regulated by both genetic and environmental factors [Sullivan et al., 2003; Marcus, 2004; Gibson, 2008; Ramos and Olden, 2008]. In retrospect, it is naïve to suppose that direct, linear relationships between DNA sequence and phenotype should arise at all. Global phenotypes are the product of complex networks of genes, proteins, and tissues that accept environmental inputs, so phenotypes represent an emergent property of a dynamic biological system rather than the deterministic output of either genetic or environmental inputs [Bhalla and Iyengar, 1999; Koch and Laurent, 1999]. This is especially relevant to diseases of the brain, since many complex brain functions are not anatomically localized and, even when controlled by a defined brain structure, arise from dynamic patterns of activation of large neural networks.

Traditional twin studies are based in part on the assumption that there are deterministic effects of gene sequence and environmental events on phenotype. This perspective does not account for the stochastic nature of many biological and biochemical processes, nor the complex interaction between environmental influences and phenotype. One molecular mechanism that could be both a source of stochastic variation in gene expression and a mediator of environmental effects on phenotype is epigenetics. Epigenetic factors that may cause MZ twin pairs to diverge and account to a large degree for some of their phenotypic discordance include skewed X-inactivation in female MZ (FMZ) twins, imprinting, and other modifications of chromosomal DNA and gene expression, especially through DNA methylation [Gringras and Chen, 2001; Boomsma et al., 2002; Rosa et al., 2008].

Epigenetic factors that may cause MZ twin pairs to diverge and account to a large degree for some of their phenotypic discordance include skewed X-inactivation in FMZ twins, imprinting, and other modifications of chromosomal DNA and gene expression, especially through DNA methylation.


There are differing definitions of the term “epigenetics,” but a common notion is that epigenetic phenomena are characterized by “modifications in gene expressions that are controlled by heritable but potentially reversible changes in DNA methylation and/or chromatin structure” [Henikoff and Matzke, 1997; Petronis et al., 2000]. Epigenetic mechanisms are dynamic processes that are influenced by developmental stage, tissue type, environmental factors, and stochastic events [Petronis et al., 2003]. Both de novo methylation for altering gene expression and maintenance methylation occur. Methylation patterns can be maintained during DNA replication and inherited across generations [Goto and Monk, 1998]. Methylation of genomic DNA may affect a variety of processes related to gene expression, such as X-inactivation, imprinting, or expression of specific genes [Singh et al., 2002]. Since DNA modification is a dynamic process, it is only partially stable and there is a great potential for epigenetic variation within MZ twin pairs [Wong et al., 2005].

There is recent evidence that MZ twins have epigenetic discordance. Kaminsky et al. mapped differences in DNA methylation at a locus displaying a range of co-twin variability in three types of tissues—white blood cells, buccal epithelial cells, gut (rectum) biopsies and—from MZ and DZ twin pairs using microarrays [Schumacher et al., 2006]. They found a large degree of MZ co-twin DNA methylation variation in all the tissue samples investigated, validating their findings using sodium bisulfite modification based mapping of methylated cytosines in CpG islands [Frommer et al., 1992; Kaminsky et al., 2009]. There are a number of specific epigenetic mechanisms that may alter phenotype, including skewed X-inactivation in FMZ twins, imprinting (differential expression of genes inherited from the mother or father), and DNA methylation, that are discussed below [Gringras and Chen, 2001; Boomsma et al., 2002; Rosa et al., 2008].

Differential Epigenetic Profiles in Twins: Skewed X-Inactivation

Skewed X-inactivation via epigenetic modifications may be responsible for some of the phenotypic discordance observed in FMZ twin pairs. X-inactivation refers to the process of inactivation of one of the X chromosomes in females to achieve dosage compensation of X-linked genes with males [Goto and Monk, 1998; Petronis, 2003]. Such inactivation often involves epigenetic modifications such as methylation of DNA at CpG islands in the 5′ region of X-linked genes [Goto and Monk, 1998; Csankovszki et al., 2001] or hypoacetylation of histone H4 [Jeppesen and Turner, 1993; Csankovszki et al., 2001]. In most cases, X-inactivation is a random process where either X chromosome can be inactivated, but some studies have produced evidence for “skewed” X-inactivation where the X chromosome of one parental origin is preferentially inactivated [Goto and Monk, 1998]. However, not all genes on the inactivated X chromosome remain silent. The differential expression of X-linked genes on both the active and inactivated X chromosomes is regulated temporally and spatially via differential DNA methylation, influenced by both environment and stochastic factors [Singh et al., 2002]. Numerous studies have identified skewed X-inactivation in FMZ twin pairs discordant for X-linked medical conditions such as fragile X-syndrome [Kruyer et al., 1994], Duchenne muscular dystrophy [Zneimer et al., 1993], color blindness [Jørgensen et al., 1992], X-linked immunodeficiencies, Lesch–Nyham disease, and hemophilia [Rosa et al., 2008].

Skewed X-inactivation can aid in the investigation of complex diseases, including the broad spectrum of psychiatric illnesses. Comparing concordance rates between FMZ and male MZ (MMZ) twin pairs can potentially identify X-linked loci that influence the manifestation of disease states. If polymorphic X-linked loci are involved, FMZ pairs should be more discordant than MMZ due to skewed X-inactivation [Rosa et al., 2008]. This was observed by Loat et al. 2004 in a study that tested a large sample of same-sex twin pairs on several social, behavioral, and cognitive measures: concordance rates were higher in almost all categories in MMZ twins than in FMZ twins. Traits showing higher concordance rates in males included pro-social behavior, peer problems, and verbal ability, all of which are influenced by genes on the X chromosome [Loat et al., 2004]. In schizophrenia, some evidence points to the possibility that the disease may be X-linked and thus modulated by skewed X-inactivation: An excess of X-chromosome aneuploidies (XXX and XXY) among patients has been found in a subgroup of familial cases [DeLisi et al., 1994; Rosa et al., 2008], and MZ twin concordance for psychosis is slightly higher in males than females [Rosa et al., 2008].

Differential Epigenetic Profiles in Twins: Genomic Imprinting

In addition to skewed X-inactivation, genomic imprinting may play a role in the differential epigenetic profiles of MZ twin pairs. Genomic imprinting occurs when a gene is preferentially expressed based on its parental origin; only one of either the paternal or the maternal allele is expressed. It is generally accepted that this marking occurs through epigenetic modifications of chromosomal DNA during gametogenesis [Goto and Monk, 1998; Monk, 1995]. Discordant MZ twins have been studied in the context of genomic imprinting. One such study of MZ twins discordant for Beckwith–Wiedemann syndrome (BWS) found that only the affected twin had an imprinting defect at KCNQ1OT1 on 11p15 [Weksberg et al., 2002].

So far, studies that have investigated imprinting in psychiatric disease have been disappointing. For example, De Luca et al. 2007 found no evidence of a parent-of-origin effect in the serotonin receptor HTR2A gene T102C polymorphism in association with psychosis in schizophrenia or bipolar disorder, so imprinting at this locus is unlikely to account for the complex inheritance pattern seen in major psychoses. While two studies in mouse suggested gene imprinting to be associated with loci related to endophenotypes observed in psychiatric patients [Luedi et al., 2005; Zhao et al., 2006], a subsequent study with human brain samples did not find strong evidence of imprinting of DISC1 [Hayesmoore et al., 2008]. DISC1 is a strong candidate gene for schizophrenia and mood disorders, but in 148 human brain mRNA samples studied, only one showed unequal expression of paternal and maternal transcripts [Hayesmoore et al., 2008]. Although genomic imprinting has not yet been demonstrated to be important in psychiatric conditions, imprinting could help to explain inconsistent genetic association data in the context of parent-of-origin effects [Hayesmoore et al., 2008].

Epigenetic Regulation of Gene Expression

Although skewed X-inactivation and genomic imprinting are important sources of MZ twin discordance, they occur much less frequently than other forms of epigenetically regulated gene expression [Goto and Monk, 1998]. Recent studies show that DNA methylation may account to a large degree for MZ twin discordance and may influence susceptibility to bipolar disorder and schizophrenia [Petronis, 2001, 2006]. Studies investigating epigenetic differences in MZ twin pairs with bipolar disorder are rare, possibly because MZ twins completely discordant for bipolar disorder are difficult to find [Bertelsen, 2004; Kato et al., 2005]. Some empirical work on the genomes of MZ twins discordant for schizophrenia has been published which shows epigenetic differences reflected in different DNA methylation patterns between MZ co-twins.

A study by Tsujita et al. 1998 illustrates possible differential DNA methylation in MZ twins discordant for schizophrenia. Using restriction landmark genome scanning (RLGS), a method consisting of two-dimensional electrophoresis of genomic DNA fragments that are end-labeled and have been digested with a restriction enzyme [Asakawa, 2008], this study revealed genomic differences in a schizophrenia discordant MZ twin pair. Two areas were identified at which the MZ twin pair had differing intensities indicative of either genetic or epigenetic differences between their genomes. The limitation of this study is that Tsujita et al. 1998 used the methylation-sensitive restriction enzyme NotI, which cuts at sites often located in CpG islands, near the promoters of genes. Thus, although the group's results could be due to discordance in DNA sequence, the confounder is that they could instead reflect differences in methylation status at NotI-flanking sites [Tsujita et al., 1998; Kato et al., 2005].

Additional evidence for epigenetic differences between MZ twins discordant for schizophrenia comes from two studies by Deb-Rinker et al. 1999, 2002. The group identified and sequenced two retroviruses in the affected twins from MZ twin pairs discordant for schizophrenia using representational difference analysis (RDA). RDA locates differences between similar genomes by enriching sequences present in one complex DNA sample, but absent or substantially deleted from another [Lisitsyn and Wigler, 1993; Ushijima et al., 1997; Deb-Rinker et al., 2002]. The retroviruses identified were schizophrenia-related retrovirus-1 (SZRV1) and SZRV2. These are both heavily methylated and placentally expressed endogenous retroviral-related human genome sequences. Although no direct evidence exists for any causative effect of SZRV1 and 2 on schizophrenia, these retroviruses may be abnormally expressed in the cerebrospinal fluid of individuals with recent-onset schizophrenia or schizoaffective disorder [Karlsson et al., 2001]. In one affected patient, but none of the controls, loss of DNA methylation of SZRV2 was observed [Deb-Rinker et al., 2002], suggesting that DNA methylation may restrict placentally expressed retroviral sequences as part of natural nuclear host defences against parasitic sequence elements such as viruses [Deb-Rinker et al., 2002].

Some work has been published on the role of MZ twin methylation discordance in relation to specific schizophrenia candidate genes, such as the dopamine D2 receptor gene (DRD2) or catechol-O-methyltransferase gene (COMT). One study investigated DNA modification in two MZ twin pairs, one concordant and one discordant for schizophrenia, in a fragment of the 5′ regulatory region of DRD2, a gene for which specific polymorphisms have been associated with schizophrenia [Petronis et al., 2003]. Several DNA methylation differences were identified in the analyzed region both within and between the pairs of MZ twins. This study used direct sequencing of sodium bisulfite-treated DNA to map methylated cytosines [Frommer et al., 1992]. Epigenetic distance, a measure of the number of differences at each CpG site [Yatabe et al., 2001; Petronis et al., 2003], was calculated for each subject and used for the comparison of twin DRD2 methylation profiles. They found that the affected twin from the pair discordant for schizophrenia was epigenetically closer to the affected concordant twins than to his unaffected MZ co-twin. This suggests a role for DNA methylation in a candidate gene for schizophrenia and demonstrates association between epigenetics and phenotypic MZ twin discordance.

They found that the affected twin from the pair discordant for schizophrenia was epigenetically closer to the affected concordant twins than to his unaffected MZ co-twin. This suggests a role for DNA methylation in a candidate gene for schizophrenia and demonstrates association between epigenetics and phenotypic MZ twin discordance.

A second study investigated MZ twin methylation at CpG sites within COMT, and found discordance between twin pairs in their methylation profiles ranging from approximately 1% in one pair to 42% in the most discordant pair [Mill et al., 2006].


The evidence reviewed makes it clear that MZ twins have substantial differences in obvious phenotypes like disease, and in epigenetic DNA modification patterns. Earlier twin studies were based on the premise that MZ twins are genetically identical, and that phenotypic differences must arise from non-shared environment. However, knowledge of epigenetic mechanisms such as differential DNA methylation, skewed X-inactivation, and imprinting provides a new model to understand MZ twin discordance, and potentially to discover more general mechanisms affecting psychiatric disease susceptibility.

Advances in technology, including automated high-throughput DNA sequencing, microarray-based CpG island methylation measurement, RLGS, and RDA, make it possible to study MZ twin discordance on a genomic level. Although twin studies represent one of the earliest approaches in genetic epidemiology, new methods continue to rely on this naturally occurring genetic experiment. Galton's conception of nature versus nurture may be recast as nature and nurture in combination, with epigenetic and stochastic factors both mediating and initiating the interactions between genes and the environment.


We thank Silvia Odorcic for critical reading of this manuscript and the Canadian Institutes of Health Research for salary support through a Clinician-Scientist Fellowship to AHCW.