DNA methylation in peripheral tissue of schizophrenia and bipolar disorder: a systematic review

Background Increasing evidence suggests the involvement of epigenetic processes in the development of schizophrenia and bipolar disorder, and recent reviews have focused on findings in post-mortem brain tissue. A systematic review was conducted to synthesise and evaluate the quality of available evidence for epigenetic modifications (specifically DNA methylation) in peripheral blood and saliva samples of schizophrenia and bipolar disorder patients in comparison to healthy controls. Methods Original research articles using humans were identified using electronic databases. There were 33 included studies for which data were extracted and graded in duplicate on 22 items of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement, to assess methodological precision and quality of reporting. Results There were 15 genome-wide and 18 exclusive candidate gene loci investigations for DNA methylation studies. A number of common genes were identified as differentially methylated in schizophrenia/bipolar disorder, which were related to reelin, brain-derived neurotrophic factor, dopamine (including the catechol-O-methyltransferase gene), serotonin and glutamate, despite inconsistent findings of hyper-, hypo-, or lack of methylation at these and other loci. The mean STROBE score of 59 % suggested moderate quality of available evidence; however, wide methodological variability contributed to a lack of consistency in the way methylation levels were quantified, such that meta-analysis of the results was not possible. Conclusions Moderate quality of available evidence shows some convergence of differential methylation at some common genetic loci in schizophrenia and bipolar disorder, despite wide variation in methodology and reporting across studies. Improvement in the clarity of reporting clinical and other potential confounds would be useful in future studies of epigenetic processes in the context of exposure to environmental and other risk factors.


Background
Schizophrenia (SZ) and bipolar disorder (BD) share some common genetic vulnerability [1,2] and environmental risk factors [1,3]. Only a small portion (approximately 23 %) of the variance in risk for these disorders can be accounted for by common variation in the genome [4,5]. The role of epigenetic processes (affecting gene expression) may thus account for substantial variation in the development of SZ and BD [6], and is consistent with evidence for non-genetic risk factors (e.g., obstetric complications [7,8] and viral infections [9]) which may confer risk for these disorders via epigenetic processes.
Epigenetic modifications to the genome refer to changes in the physical structure of the chromatin, without a change in the DNA sequence itself [10]. The most widely studied epigenetic modification is DNA methylation, characterised by covalent linking of a methyl (CH3) group to a cytosine residue [11], almost exclusively occurring at cytosines within CpG dinucleotides. These CpGs are clustered in 'CpG islands' that tend to be located in regulatory elements of the gene, such as promoters or enhancers [12]. Methylation at CpG islands usually results in transcriptional silence of the associated gene [12]. In recent studies of psychiatric phenomena, the functional impact of stressrelated hypomethylation of genetic loci known to regulate stress responses (e.g., FK506 binding protein 5 (FKBP5)) suggest that this process may be relevant to many stressrelated disorders [13]. While there are a number of postmortem studies reporting differentially methylated genes in these disorders, these findings have been recently reviewed elsewhere [14,15]. This review instead focused on the growing evidence base for differential DNA methylation in peripheral (i.e. blood and saliva) samples, which may minimise confounding effects related to tissue quality and stability [16], and importantly allow the study of epigenetic processes in living humans. Recent comparison of within-subject methylation patterns across blood and brain suggest the utility of peripheral blood in human epigenetic studies [17].
With the increasing use of peripheral tissue for the study of methylation patterns in psychotic disorders, the aim of this study was to perform a systematic review of evidence from observational case-control studies investigating differential DNA methylation in the peripheral tissues (blood or saliva) of SZ and/or BD patients, in comparison to a healthy control (HC) group. Assessment of the quality, consistency and strength of evidence reported across studies was undertaken for all studies using accepted criteria, using a validated tool for assessing methodological precision and quality of reporting.

Literature search: inclusion/exclusion criteria
Included are peer-reviewed, observational case-control studies investigating DNA methylation in the peripheral tissues (blood, saliva) of SZ (including schizoaffective disorder) and/or BD (type I and II) in comparison to a HC group. Excluded studies reported other types of epigenetic modifications (i.e. hydroxymethylation), mRNA gene products of the methylation pathway, or DNA methylation in germ line cells or post-mortem brain tissue, for which results have recently been reviewed elsewhere [14,15].
Searches were limited to studies published in English, conducted in humans, and published after the year 2000 to minimise the methodological inconsistencies seen in the earliest studies of DNA methylation (e.g. improvements in polymerase chain reaction based DNA methylation methods) [18].

Study selection
A preferred reporting items for systematic reviews and meta-analysis (PRISMA) flowchart of the search and selection processes of the included studies is presented in Fig. 1. All decisions relating to study inclusion were completed independently by two authors (NT and LG) with any disagreements resolved by discussion with MG.

Quality assessment and study characteristics
Information relating to data quality were graded in duplicate (authors NT and MG) using 22 items listed in the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [19], to assess the risk of bias within studies and across studies, methodological consistency and precision, as well as reporting transparency and comprehensibility. The STROBE statement consists of a 22 item checklist which allows a systematic and critical assessment of the strengths and weakness of the study design, conduct and analysis [19]. Thresholds for determining study quality were determined by authors, as per the STROBE guidelines: scores on the STROBE checklist of ≥ 66 % were considered to be reflective of high study quality, ≤ 33 % of low quality reporting, and scores in between this range were of moderate quality. Thresholds for categorising significant change in methylation status for hypermethylation, hypomethylation and differential methylation (i.e. non-specified direction of difference), as well as no significant group difference in methylation were protocol dependent, but largely reflect convention of significance level being set at 0.05, with appropriate corrections applied according to the number of analyses being conducted. Study characteristics are summarised in Tables 1, 2 and 3 and include sample characteristics, methods of quantifying DNA methylation, as well as any relevant data (e.g., medication status) that could contribute to the investigation of subgroups.

Search results: included and excluded studies
The systematic search strategy identified a total of 908 publications, of which 622 were duplicates (i.e., 286 unique studies); an additional 18 publications were found by hand searching reference lists and advance access publications (See Fig. 1). These 304 studies were screened for relevance by title and abstract, resulting in the removal of 178 studies. Full text screening of the remaining 126 studies excluded a further 75 studies which did not meet inclusion criteria (see Fig. 1); of these, 41 were conference abstracts/ reviews, 22 did not include SZ or BD participants, two did not have a comparison group consisting of healthy unrelated subjects, seven reported indirect measures of DNA methylation (i.e. mRNA expression of DNA methylation products), two investigated other types of epigenetic modification and one study did not investigate DNA methylation in the genome. An additional 18 studies were excluded which conducted DNA methylation analyses using only germ line cells or post-mortem tissue. A final total of 33 studies, which fulfilled inclusion criteria, were evaluated in this systematic review.

Study quality assessment
The STROBE ratings suggested that the available evidence for differential methylation in SZ and BD ranged from low (29.5 % minimum) to high quality (77 % maximum) with the mean of all scores at 59 % (SD: 2.36), suggesting moderate quality of available evidence and moderate probability of reporting bias.

Methylation analyses and genes investigated in schizophrenia and bipolar disorder
Genome-wide DNA methylation analyses (including three methylome-wide association study; MWAS) were conducted in 15 out of 33 studies (comprising 13 SZ studies, one BD sample, and one combined SZ/BD sample). Two of these 15 genome-wide DNA methylation studies (one SZ and one BD) reported no difference in DNA  [23,42], while one study found genome-wide hypomethylation in SZ [40]. Of the 15 genome-wide studies, only four reported estimates of 'global' methylation changes across the entire genome (i.e., % differential methylation without reference to specific genes). The remaining 18 studies focused exclusively on candidate gene loci (9 SZ, 6 BD and 3 SZ/BD studies). There was a total 163 different genes investigated, with four genes investigated in more than one study. These included reelin (RELN) (2 SZ studies), brain-derived neurotrophic factor (BDNF) (3 SZ and 3 BD studies), catechol-O-methyltransferase (COMT) (1 SZ/BD, 3 SZ studies) and hydroxytryptamine serotonin 1A receptor (HTR1A) (2 SZ/BD studies). Out of the 33 studies, 3 provided a raw results database for download [32,41,51].

Evidence for DNA methylation in schizophrenia and bipolar disorder
Across all studies of SZ and/or BD, there were 21 sites reported as hypermethylated, seven sites of hypomethylation, and 135 genetic loci reported as differentially methylation.
The most common genes identified as differentially methylated in SZ/BD were different receptors, transporters and neurotransmitters related to RELN, BDNF, dopamine, serotonin and glutamate (see Table 4); this consisted of 14 candidate gene loci studies (one RELN, five BDNF, five dopamine, two serotonin and one glutamate) and 10 genome-wide studies (one RELN, one BDNF, three dopamine, four serotonin and one glutamate). For these genes, there was evidence of both hyper-and hypo-methylation in both SZ and BD, as well as some evidence for lack of differential methylation. There were also several studies reporting DNA methylation of genes previously linked to SZ, including: hypermethylation of gamma-aminobutyric acid receptor beta-2 (GABRB2) [39], discs large homolog 4 (DLG4) and the gene disrupted in schizophrenia 1 (DISC1) [32], as well as differential methylation of major histocompatibility complex class C (HLA-C) and calcium homeostatis modulator 1 (CALHM1) [41]. The results of specific genetic loci reported in more than one study are discussed in further detail below. In addition, 11 studies (7 SZ, 3 BD and 1 SZ/BD study) reported no differences in methylation in a number of genes (see Tables 1, 2 and 3).

Reelin
Differential methylation for RELN was reported for intron 1 in SZ [20], although another study also reported a lack of differential methylation of the RELN promoter in SZ [22].

Brain-derived neurotrophic factor
Methylation investigations for BDNF in SZ and BD were reported only for promoter regions. In BD (both type I and II) there was consistent reporting of hypermethylation of the BDNF exon 1 promoter in two studies [44,45], although one other study of BD (unspecified-type) reported lack of differential methylation at this site [43]. In SZ, the results were mixed with hypermethylation of BDNF promoter I [25], differential methylation of an unspecified BDNF promoter [30] and a lack of differential methylation of BDNF promoter IV [25].

Dopamine
There was mixed evidence for methylation status of genes associated with dopamine transporters in SZ, which included hypermethylation of dopamine active transporter 1 (DAT1) [30], hypomethylation of solute carrier family 6 transporter member 3 (SLC6A3) [37] and differential methylation of vesicular monoamine transporter 2 (SLC18A2) [27]. The other inconsistent results for genes associated with dopamine were for COMT with studies reporting hypomethylation of membrane-bound (MB-) COMT in SZ and BD [52], hypomethylation of COMT (isoform not specified) in SZ (but not BD) [51], hypermethylation of soluble (S-) COMT in SZ [34], differential methylation of COMT domain containing 1 (COMTD1) promoter in SZ [37] and a lack of differential methylation of S-COMT promoter in SZ [36]. One global DNA methylation study also reported differential methylation of glial cell line-derived neurotrophic factor family receptor alpha 2 (GFRA2) in SZ [28], which indirectly affects dopaminergic neurons.

Subgroup analyses
Subgroup analyses and/or the study of covariates were reported in 28 out of 33 studies. However, only 20 studies reported significant effects of age, sex, pharmacological (antipsychotic/antidepressant) treatment, symptom severity, and/or smoking/alcohol abuse. Further analyses of ethnicity effects on DNA methylation was absent in all but one study [33]. Other notable findings, which were only reported in one study, include a significant association of gene expression with DNA methylation in SZ [41], and cellular heterogeneity of white blood cells as a major confounder in DNA methylation analyses also in SZ [26].

Discussion
This review highlights findings of moderate quality, showing mixed evidence of hyper-and hypomethylation of several common genetic loci in 22 studies of schizophrenia and/or bipolar disorder, from a total of 33 reviewed studies. Differential methylation converged on  Differential methylation was also reported in several genes (e.g. Fc fragment of IgA (FCAR), cyclic AMP-responsive element-binding protein 1 (CREB1), lymphocyte transmembrane adaptor 1 (LAX1)) related to immune system function and the inflammatory response in SZ [20,33], consistent with recent evidence for shared genetic risk (for SZ and BD) in common variants of the major histocompatibility complex [53].

Genes implicated in schizophrenia and bipolar disorder
The most commonly reported sites of epigenetic changes were in regions known to regulate the availability of neurotrophins, dopamine and serotonin. For example, BDNF is a neurotrophin involved in neuroplasticity and dopaminergic neuron survival [54], for which peripheral blood levels have been found to be decreased in both SZ and BD patients [55,56]. However, in BD, there were two studies reporting hypermethylation of the BDNF gene (exon 1 promoter) that was associated with pharmaceutical treatment and mood states [44,45], while another study reported a lack of differential methylation at this site in BD patients with a history of psychosis [43]. In SZ, there was evidence of differential methylation at several other BDNF sites, including promoter I [25], and an unspecified BDNF promoter [30], with one study also reporting no difference in methylation of BDNF promoter IV [25]. On the COMT gene, there was consistent evidence for hypomethylation of MB-COMT in SZ and BD [52] and an unspecified COMT isoform in SZ only [51]; other SZ studies reported mixed findings including hypermethylation of S-COMT [34], differential methylation of COMTD1 promoter [37] and a lack of significant differential methylation of S-COMT promoter [36]. The mixed evidence for methylation of dopamine transporter genes in SZincluding hypermethylation of DAT1 [30], hypomethylation of SCL6A3 [37] and differential methylation of SLC18A2 [27], is interesting in the context of previous evidence of genome-wide differential methylation of GFRA2 in SZ [28], a receptor for glial cell-derived neurotrophic factor (GDNF) which manages dopaminergic neuronal maintenance while also being implicated in SZ and BD [57,58]. Non-specific, differential methylation of serotonin transporter sites were evident in BD (SLC6A4) [48] and SZ (HTR1E) [37], while hypermethylation of 5-HTR1A was reported in two SZ and BD studies [49,51]. These results converge with the numerous reports of variation in serotonin transporter gene (5-HTT or SLC6A4) interacting with stressful life events to result in psychiatric (usually mood) disorder [59,60]. However, there was also evidence for lack of differential methylation of 5-HTT in SZ [34], and HTR2A in SZ and BD [50].
Finally, a number of glutamate receptor genes (GRIA2, GMR2, GMR5, GMR8 and GRIA3) were found to be hypermethylated in SZ [21,29] while in BD there was no such evidence. This is intriguing given that recent genome-wide association studies (GWAS) have implicated genes associated with glutamate neurotransmitter dysfunction as relevant to risk for both disorders [61,62]. In SZ, there was also a finding of hypermethylation of DLG4 [33], a gene which has downstream regulatory effects on glutamate receptors implicated in SZ pathophysiology [63]. The few studies of methylation in the promoter region of RELN in SZ should be mentioned as consistent with post-mortem evidence [64], while there were some other notable findings for hypermethylation of DISC1 [32], differential methylation of HLA-C and CALHM1 [41], and hypermethylation of GABRB2 [39] which have each been identified as risk variants for SZ in previous work [65][66][67].

Associations with demographic and clinical variables
Only 20 of 33 studies examined the effects of age, sex, medication, symptom severity, and/or smoking/alcohol abuse on methylation patterns, with mixed findings. However, there were consistent trends emerging for no significant associations between methylation status of various genes and age [23,25,34,38,45,47,49] (particularly in SZ studies [23,25,34,38]), while a handful of other studies suggest that differential methylation increases with age [50][51][52]. There was also a trend for hypermethylation being more prevalent in females (see Tables 1, 2 and 3). Previous studies have reported altered DNA methylation in SZ and BD following treatment with antipsychotics and mood stabilisers such as haloperidol [34], clozapine [68], lithium and valproate [44], but these variables were inconsistently reported in the studies reviewed here.

Limitations
There are a number of limitations to this review. The most obvious was the inability to conduct a metaanalysis owing to the diversity of experimental protocols (there were 16 different methods reported across 33 studies). Methodological variability also precludes interpretation of results for the most commonly reported genes across these studies. In addition, the lack of consistency in reporting the potential effects of clinical symptoms, age, sex, medication, and ethnicity, precluded adequate interpretation of findings across studies. For example, factors such as diet [69], exercise [70], smoking [71], trauma [72], emotional state [73] and ethnicity [74] are known to effect DNA methylation status, but were not adequately reported in many studies. Variability in DNA extraction methods and blood cell composition may have also affected the results of included studies [35], for which details are not included in this review. For example, the cellular heterogeneity of white blood cells has been considered to confound DNA methylation analyses [26], despite associations between gene expression and DNA methylation in whole blood samples suggesting that differences are minimal. However, methods for conducting methylation analyses are known to vary in efficacy and sensitivity, and may have affected the pattern of results revealed here. For example, bisulfite sequencing (conducted in eight of 33 included studies), is prone to PCR amplification bias [75], with at least some incomplete conversions of cytosine to uracil resulting in a higher number of methylated CpGs being recorded [76]. Moreover, methods for determining the appropriate significance threshold to determine differential methylation status was dependent on the experimental protocol employed in each study, such that the strength of the results reported in these studies may be equivocal. Another potential bias lies in the sample size differences between studies of candidate gene and global DNA methylation studies, of which the latter require larger sample sizes with respect to multiple testing issues. In addition, the results of global DNA methylation studies are simply not comparable with approaches such as MBD protein-enriched genome sequencing which is more sensitive than 27 k/450 k arrays [21].
Finally, this study did not directly compare the methylation status of particular genes arising from studies of post-mortem versus peripheral tissue, but included studies using DNA derived from blood or saliva (only two studies used saliva). While methylation patterns in saliva cells may be affected by oral hygiene, we note that similar patterns of methylation were reported in saliva and post-mortem tissue in both these studies [50,52]. There are certain advantages and limitations to using both post-mortem and peripheral tissues for DNA methylation studies. While the brain is the primary organ of pathology in SZ and BD, methylation analyses using post-mortem tissue may be affected by pH, postmortem interval and variability of different neuronal cell types and brain regions [77]. On the other hand, peripheral tissue is easily accessible in a minimally invasive and of low cost procedure, thus allowing for collection of larger sample sizes to overcome cellular heterogeneity of methylation patterns and facilitating longitudinal studies. Consistent methylation results across brain and blood tissues have been reported for particular promoter CpG islands in other studies not included in this review [17]. Notably, two of the included studies showed comparable methylation results for major histocompatibility complex 9 (HCG9) [46] and synapsin 3 (SYNIII) [35] in both post-mortem brain tissue as well as a blood-derived DNA.

Conclusions
Moderate quality evidence shows differential DNA methylation in peripheral tissue of SZ and BD participants, with some common genes affected despite the direction of methylation at common sites not always being consistent. While it remains questionable as to whether the differences in statistical thresholds between genome-wide to candidate gene loci studies are too great to compare results from such studies, we could not systematically address the effects of these methodological difference on general patterns of findings across disorders because of the limited data available for some methods in some groups. We also note that the role of DNA methylation in modifying gene expression has only been explored in the last 20 years [78], and it is likely that the inconsistency of the results reported here reflects the numerous methods available for determining DNA methylation alterations in peripheral tissues of these disorders, which can vary according to cell type. Reliable patterns in methylation alterations specific to SZ or BD are likely to emerge with improved, cost-effective and standardised technology that also account for cellular heterogeneity. Apart from methodological issues, another potential explanation for variability among findings is that none of the studies of psychotic samples to date have addressed the impact of early life experiences (such as childhood trauma) on differential methylation patterns in SZ and BD; more consistent methylation patterns may be revealed in the context of common lifetime environmental exposures (e.g., childhood maltreatment, birth complications, cannabis use), genomic structure, and/or mRNA expression profiles. The consideration of these additional factors will be necessary in future research to clarify the contribution of environmental effects on epigenetic processes in the development of psychosis. DLG4: Discs large homolog 4; DISC1: Disrupted in schizophrenia 1; HLA-C: Major histocompatibility complex class C; CALHM1: Calcium homeostatis modulator 1; DAT1: Dopamine active transporter 1; SLC6A3: Solute carrier family 6 transporter member 3; SLC6A4: Solute carrier family 6 transporter member 4; SLC18A2: Vesicular monoamine transporter 2; GFRA2: Glial cell line-derived neurotrophic factor family receptor alpha 2; GRIA2: Glutamate receptor ionotropic alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid 2; GRIA3: Glutamate receptor ionotrophic alpha-amino-3-hydroxy-5methyl-4-isoxazole propionic acid 3; GMR2: GMR5 and GMR8, glutamate metabotrophic receptors 2, 5 and 8 respectively; FCAR: Fc fragment of IgA; CREB1: Cyclic AMP-responsive element-binding protein 1; LAX1: Lymphocyte transmembrane adaptor 1; HCG9: Major histocompatibility complex 9; SYNIII: Synapsin 3.