Comprehensive interrogation of CpG island methylation in the gene encoding COMT, a key estrogen and catecholamine regulator

Catechol-O-methyltransferase (COMT) enzyme has broad biological functions, principally the catabolism of biologically active or toxic catechols, including catecholamines and catecholestrogens [1]. As a result of its ubiquity, COMT has been implicated in a wide range of human conditions, including cancer [2], pain sensitivity [3, 4], schizophrenia [5], affective [6], addictive [7, 8], impulse control disorders [9], and Parkinson’s disease [10]. The clinical significance of these conditions has stimulated growing interest in COMT in recent decades, particularly following the discovery that COMT enzyme activity level in human tissues is genetically polymorphic, conferring low, intermediate, and high activities [11, 12]. A great deal of research has focused on a common, functional, single nucleotide polymorphism (SNP) of COMT, Val ¹⁵⁸ Met (rs4680), that is the most studied variant, due to its location within the exon 4 coding region. Specifically, the substitution of a methionine (Met) for a valine (Val) at position 158 results in three- to four-fold reduced activity of the COMT enzyme due to reduced protein stability [13–15]. COMT is the primary regulator of dopamine clearance in extrastriatal brain regions, including the prefrontal cortex [16–18], which has helped to motivate research into associations of Val ¹⁵⁸ Met with neuropsychiatric disorders since the mid-1990’s [19]. Early neuroimaging studies found the COMT 158^Val allele to be associated with impaired prefrontal cognition and physiology, which could contribute to schizophrenia risk [5], and to modulate pain perception and brain responses to pain [4]. However, observed associations between various conditions and the 158^Val allele are modest and often inconsistent. Thus, more recent work has identified COMT haplotypes that are associated with more profound changes in COMT activity, in part by effects on COMT protein expression [20, 21].

Finally, other studies have tried to shed light on additive contributions to disease state by considering COMT polymorphisms in combination with other polymorphic gene loci. For example, genotype × genotype analyses involving COMT Val ¹⁵⁸ Met with the methylenetetrahydrofolate reductase (MTHFR) C⁶⁷⁷T polymorphism have shown strong interactive effects associated with elevated total plasma homocysteine in a case control study of elders with and without dementia [22] as well as on dopamine signaling [23], executive function [24], and cognition [25] in persons with schizophrenia. COMT × MTHFR and other multigene interactions have also been explored in breast cancer [26–28].

In addition to the reported genetic effects on COMT expression, activity, brain function, and associations with behavior or disease risk, several non-genetic factors have been reported to impact COMT function, either in isolation or via interactive effects on genetic associations. These include an age-dependent rise in COMT activity in the prefrontal cortex [29], interacting effects of Val ¹⁵⁸ Met genotype and age on impulsive choice [30], and sexually dimorphic associations with COMT in humans [31, 32].

In addition, numerous studies have reported interactions between Val ¹⁵⁸ Met genotype and environmental stressors, impacting everything from prefrontal function [33], and affect-modulated startle [34], to risk of alcoholism [35], posttraumatic stress disorder [36, 37], and impulsive aggression [38]. How these non-genetic factors modulate COMT effects is largely unknown, but epigenetic regulation at the level of DNA methylation is one potential mechanism. In fact, investigations of the association between alcohol use and DNA methylation is a rapidly expanding area of research, although no studies to date have specifically investigated methylation within the COMT gene, beyond a few CpGs [39].

COMT polymorphisms have been broadly explored, not only within neurobiology, but also for their role in carcinogenesis, particularly in hormonally distinct cancers of the uterus and breast [2, 31, 40]. As in multiple neurobiobehavioral studies, the associations between COMT Val ¹⁵⁸ Met and breast cancer risk have been inconsistent or modest [41–43]. One meta-analysis of COMT Val ¹⁵⁸ Met based on more than 30,000 cases and 38,000 controls, found an increased risk for breast cancer only when the sample was stratified by race [44]. Given the predominance of estrogen receptor negative breast cancer in African American women, it is plausible that the He et. al. (2012) finding may more precisely reflect varied COMT influences that are dependent on the estrogen receptor status of the tumor.

The rationale for studying the influence of COMT on hormonally influenced cancer relates to the role of COMT on catecholamines and in estrogen metabolism. Specifically, the catecholamine neurotransmitters dopamine, norepinephrine, and epinephrine are synthesized primarily in the adrenal glands, and are derived metabolically from the amino acids tyrosine and phenylalanine [45]. The COMT enzyme effects the degradation of both catecholamines and catecholestrogens (important intermediary metabolites in estrogen induced cancers), by the addition of a methyl group [46]. The methyl (CH3) group with which COMT carries out targeted destruction of catechol compounds is provided by S-adenosyl methionine (SAM), a key methyl donor in the folate metabolic pathway with a pivotal role in epigenetic alterations in general, [45] and in epigenetic changes in cancer in particular [47, 48].

Past studies have focused on genetic variations, particularly SNPs as vulnerability factors for cancer [40, 44], schizophrenia [5], pain [49], emotional processing [50], and broad cognitive functioning [51]. While the discovery of COMT genetic polymorphisms has illuminated a host of biologic vulnerabilities ranging from cancer to neurobehavioral pathology, recent evidence suggests that epigenetic changes, particularly DNA methylation of CpG (e.g., “C – phosphate - G” on the same DNA strand) dinucleotide sequences in the COMT gene, may also have an important impact on COMT function [52, 53]. DNA methylation may be inherited (via imprinting for example), and/or may result from somatic changes due to a wide array of environmental influences, such as exposure to stress [54], diet, alcohol and tobacco use [47, 48]. Once effected, somatic DNA methylation changes are perpetuated in successive cell generations during cell replication and division [55]. DNA methylation can affect gene transcription via interactions with DNA packaging chromatin proteins, and/or by interfering with the binding of enhancers, transcription factors, or other proteins involved in transcriptional regulation [15]. In such cases, DNA methylation may exert its effects either synergistically or independent of known genetic variations.

Based on recent findings showing evidence of differential COMT DNA methylation [56], and associations between COMT methylation state and vulnerabilities to a spectrum of conditions ranging from schizophrenia [57, 58] to the cognitive effects of stress [33], we sought to provide the first comprehensive assessment of DNA methylation throughout the COMT gene, simultaneously considering Val ¹⁵⁸ Met genotype in association with methylation changes. We hypothesized that differential methylation (DM) would be associated with demographic factors, reflecting a possible environment × gene interaction, and with variation in COMT genotype, which would support COMT allele-specific methylation. Moreover, we hypothesized that different levels of alcohol use would also be associated with COMT methylation, possibly in a COMT genotype-dependent manner. We tested these hypotheses using saliva samples collected from a non-clinical community sample of adult social drinkers (ages 18–40) with no known history of neurologic or psychiatric illness, and a spectrum self-reported alcohol use. The epigenetic variation in peripheral COMT detected in DNA derived from human saliva has been shown to be similar to those found in COMT in the brain [58]. In addition, we were able to test for associations between COMT DM and observed variation in COMT gene expression in breast cancer cell lines. We interrogated over 200 CpGs in 13 amplicons spanning the 5’ UTR to the last exon of the CpG dinucleotide-rich COMT gene in n = 48 subjects, n = 11 cell lines and 1 endogenous 18S rRNA control. We report DM at multiple COMT loci in association with Val ¹⁵⁸ Met genotype, socioeconomic status (SES), ethnicity, alcohol use, and gene expression.

We studied DNA methylation throughout the large COMT gene, interrogating over 200 CpG dinucleotides spanning more than 27,000 base pairs. To our knowledge, this is the first comprehensive assessment of COMT DNA methylation, and also the most comprehensive test of associations between such methylation and non-genetic factors, the Val ¹⁵⁸ Met genotype, and their interactions. With the exception of the hypomethylated amplicon 001 covering part of the 5’UTR and exon 1 of COMT (Figure 1), the remaining 12 amplicons were relatively hypermethylated, with an observed dynamic range of methylation typically between 50–90 percent.

By employing the use of the EpiTYPER^® MassARRAY platform to precisely quantify methylation for each CpG unit, we were able to pinpoint specific, differentially methylated CpG sequences that were highly informative between sample types and by subject features. Indeed, Figures 3B-C underscore the importance of consecutive interrogation of CGs throughout CpG islands in that we were able to identify transitional CpGs in amplicons 005 and 006 where the balance of methylation shifted from a methylated to a relatively unmethylated state. Notably, we identified significant associations with COMT Met alleles in these transitional regions of both amplicons 005 and 006. Additionally, we identified what, to our knowledge, is a newly described ~40 bp repeat sequence motif that begins in exon 1-v2. This repeat motif spans ~1,500 bp, and given its exonic location, is of potential significance in that it may have an effect on MB-COMT expression. What is certain, however, is that this repeat motif made PCR amplification of the COMT_002 amplicon challenging, and rendered amplification of COMT_004 nearly impossible due to stuttering of Taq polymerase through multiple repeats over large bp distances. Thus the DM within some regions of COMT remains unknown, although future studies may be able to overcome this technical challenge.

Although the size of our sample of salivary DNA from a community sample of self-identified social drinking adults, albeit representing a continuum from light to heavy drinkers, was not large (n = 48), we note that these data replicate the previously reported finding of significant DM in the putative COMT promoter region (designated amplicon COMT_006 here, particularly 006_3) [33, 56]. Like Ursini and colleagues, we found an effect of Val ¹⁵⁸ Met genotype on methylation in this promoter region, however, that study found increasing methylation with Met allele number, whereas we observed reduced methylation in white samples that was present only in Met carriers, but no independent effect of Met alleles. Discrepancies may reflect sample differences: Ursini et al.’s sample was all white and predominantly female, whereas ours was ethnically mixed and half female. It could also reflect analytical strategy differences, as we assayed at the level of individual CpGs, and evaluated partial correlations, which controlled for effects of other factors, while Ursini et al. did not. Our 006_avg results may be the closest measure to theirs, and while we did find a significant effect of Val ¹⁵⁸ Met on methylation of 006_avg, our observed relationship was in the opposite direction (less methylation with more Met alleles). Future work with larger samples allowing for well-powered stratification for sex and ethnicity may help resolve this issue. The robustness of the finding of DM in the COMT_006 region, including here in two small independent samples, coupled with our novel finding of significant associations between COMT_006 methylation with COMT expression highlights the importance of further work in this area. We also report novel associations between methylation in COMT_001 with SES, and COMT_003, 005, and 006 with ethnicity (Table 3), with many of these effects interacting with Val ¹⁵⁸ Met genotype. These latter findings indicate the importance of controlling for these factors in future studies. Viewed in light of reported associations with stress and both race/ethnicity and SES (e.g. [68]), and the numerous recent reports of interacting effects of stress and Val ¹⁵⁸ Met genotype on a large variety of conditions [33–38], differential COMT methylation state as a function of stress in Met carriers is a plausible underlying molecular mechanism for these interacting associations and warrants further study.

The relationship between DNA methylation and alcohol use in human samples is a topic of growing interest, although to date, our understanding is far from complete [39]. For example, less global DNA methylation is observed in postmortem brain samples from alcoholics relative to controls [67]. In contrast, samples from living human subjects have found reduced global methylation among ever-drinkers relative to non-drinkers [69], elevated global methylation among alcoholics compared with controls [70], or no differences in DNA methylation based on alcohol use [71, 72]. Studies investigating sites within candidate genes have shown mixed results, although together, findings suggest associations between DNA methylation and alcohol use or alcohol use disorders in human samples, although the relationships appear complex [39]. To our knowledge, only two such studies have interrogated CpGs within the COMT gene [73, 74], which both interrogated 3 CpGs in the promoter region (along with 381 other CpGs in other candidate genes), finding no association between these three sites and alcohol use. However, as we found here, such associations may interact with COMT genotype, which was not accounted for in those studies. Specifically, we found significant hypomethylation of amplicon 001 with increasingly hazardous alcohol use only among Met carriers. Notably, we also observed significant hypomethylation at a locus within amplicon 001 with lower SES specific to Met carriers. Given the finding that Met carriers show increasing risk of alcoholism with greater childhood adversity (which negatively correlates with SES) [35], it is tempting to speculate that this gene × environment interaction could derive in part from additive hypomethylation in the promoter regions of the COMT gene in Met carriers as a consequence of both childhood adversity and high alcohol intake. Further work investigating the functional consequences of such methylation is needed, although our gene expression data showed a trend toward decreased COMT expression with increasing methylation. Thus, hypomethylation would be predicted to result in increased COMT expression, reducing tonic levels of frontal dopamine in Met carriers.

In summary, we report the first comprehensive analysis of DM within the COMT gene using precisely quantified methylation for each CpG unit. In doing so, we have pinpointed specific, differentially methylated CpG sequences that were highly informative between sample types and by individual differences between samples. Some of these findings solidify existing literature, while other data identifies novel sites at which regulation of COMT expression may be achieved by specific biological or environmental factors. Given the broad function of COMT and its implication in a wide spectrum of clinical conditions, a thorough understanding of how expression of this gene is regulated may be therapeutically transformative in diseases as diverse as alcoholism and breast cancer. The present findings make a substantial contribution in that direction.