A comparison of feature selection methodologies and learning algorithms in the development of a DNA methylation-based telomere length estimator.

BACKGROUND: The field of epigenomics holds great promise in understanding and treating disease with advances in machine learning (ML) and artificial intelligence being vitally important in this pursuit. Increasingly, research now utilises DNA methylation measures at cytosine-guanine dinucleotides (CpG) to detect disease and estimate biological traits such as aging. Given the challenge of high dimensionality of DNA methylation data, feature-selection techniques are commonly employed to reduce dimensionality and identify the most important subset of features. In this study, our aim was to test and compare a range of feature-selection methods and ML algorithms in the development of a novel DNA methylation-based telomere length (TL) estimator. We utilised both nested cross-validation and two independent test sets for the comparisons. RESULTS: We found that principal component analysis in advance of elastic net regression led to the overall best performing estimator when evaluated using a nested cross-validation analysis and two independent test cohorts. This approach achieved a correlation between estimated and actual TL of 0.295 (83.4% CI [0.201, 0.384]) on the EXTEND test data set. Contrastingly, the baseline model of elastic net regression with no prior feature reduction stage performed less well in general-suggesting a prior feature-selection stage may have important utility. A previously developed TL estimator, DNAmTL, achieved a correlation of 0.216 (83.4% CI [0.118, 0.310]) on the EXTEND data. Additionally, we observed that different DNA methylation-based TL estimators, which have few common CpGs, are associated with many of the same biological entities. CONCLUSIONS: The variance in performance across tested approaches shows that estimators are sensitive to data set heterogeneity and the development of an optimal DNA methylation-based estimator should benefit from the robust methodological approach used in this study. Moreover, our methodology which utilises a range of feature-selection approaches and ML algorithms could be applied to other biological markers and disease phenotypes, to examine their relationship with DNA methylation and predictive value.

DNA methylation and inflammation marker profiles associated with a history of depression.

Depression is a common and disabling disorder, representing a major social and economic health issue. Moreover, depression is associated with the progression of diseases with an inflammatory etiology including many inflammatory-related disorders. At the molecular level, the mechanisms by which depression might promote the onset of these diseases and associated immune-dysfunction are not well understood. In this study we assessed genome-wide patterns of DNA methylation in whole blood-derived DNA obtained from individuals with a self-reported history of depression (n = 100) and individuals without a history of depression (n = 100) using the Illumina 450K microarray. Our analysis identified six significant (Sidák corrected P < 0.05) depression-associated differentially methylated regions (DMRs); the top-ranked DMR was located in exon 1 of the LTB4R2 gene (Sidák corrected P = 1.27 × 10-14). Polygenic risk scores (PRS) for depression were generated and known biological markers of inflammation, telomere length (TL) and IL-6, were measured in DNA and serum samples, respectively. Next, we employed a systems-level approach to identify networks of co-methylated loci associated with a history of depression, in addition to depression PRS, TL and IL-6 levels. Our analysis identified one depression-associated co-methylation module (P = 0.04). Interestingly, the depression-associated module was highly enriched for pathways related to immune function and was also associated with TL and IL-6 cytokine levels. In summary, our genome-wide DNA methylation analysis of individuals with and without a self-reported history of depression identified several candidate DMRs of potential relevance to the pathogenesis of depression and its associated immune-dysfunction phenotype.

Gene expression and epigenetic discovery screen reveal methylation of SFRP2 in prostate cancer.

Aberrant activation of Wnts is common in human cancers, including prostate. Hypermethylation associated transcriptional silencing of Wnt antagonist genes SFRPs (Secreted Frizzled-Related Proteins) is a frequent oncogenic event. The significance of this is not known in prostate cancer. The objectives of our study were to (i) profile Wnt signaling related gene expression and (ii) investigate methylation of Wnt antagonist genes in prostate cancer. Using TaqMan Low Density Arrays, we identified 15 Wnt signaling related genes with significantly altered expression in prostate cancer; the majority of which were upregulated in tumors. Notably, histologically benign tissue from men with prostate cancer appeared more similar to tumor (r = 0.76) than to benign prostatic hyperplasia (BPH; r = 0.57, p < 0.001). Overall, the expression profile was highly similar between tumors of high (≥ 7) and low (≤ 6) Gleason scores. Pharmacological demethylation of PC-3 cells with 5-Aza-CdR reactivated 39 genes (≥ 2-fold); 40% of which inhibit Wnt signaling. Methylation frequencies in prostate cancer were 10% (2/20) (SFRP1), 64.86% (48/74) (SFRP2), 0% (0/20) (SFRP4) and 60% (12/20) (SFRP5). SFRP2 methylation was detected at significantly lower frequencies in high-grade prostatic intraepithelial neoplasia (HGPIN; 30%, (6/20), p = 0.0096), tumor adjacent benign areas (8.82%, (7/69), p < 0.0001) and BPH (11.43% (4/35), p < 0.0001). The quantitative level of SFRP2 methylation (normalized index of methylation) was also significantly higher in tumors (116) than in the other samples (HGPIN = 7.45, HB = 0.47, and BPH = 0.12). We show that SFRP2 hypermethylation is a common event in prostate cancer. SFRP2 methylation in combination with other epigenetic markers may be a useful biomarker of prostate cancer.

IGFBP7 promoter methylation and gene expression analysis in prostate cancer.

PURPOSE: IGFBP7 belongs to a family of insulin-like growth factor-1 regulatory binding proteins. IGFBP7 hypermethylation is associated with its down-regulation in various carcinomas. In prostate cancer IGFBP7 down-regulation has been widely reported but to our knowledge the mechanisms behind this event are unknown. We performed a denaturing high performance liquid chromatography screening and validation strategy to profile the methylation status of IGFBP7 in prostate cancer. MATERIALS AND METHODS: We combined denaturing high performance liquid chromatography and bisulfite sequencing to examine IGFBP7 methylation in a panel of prostate cancer cell lines. Quantitative methylation specific polymerase chain reaction was used to determine methylation levels in prostate tissue specimens of primary prostate cancer, histologically benign prostate adjacent to tumor, high grade prostatic intraepithelial neoplasia and benign prostatic hyperplasia. IGFBP7 gene expression was measured by quantitative methylation specific polymerase chain reaction in cell lines and tissue specimens. RESULTS: IGFBP7 was methylated in the 4 prostate cancer cell lines DU145, LNCaP, PC-3 and 22RV1. Quantitative methylation specific polymerase chain reaction analysis revealed that promoter methylation was associated with decreased IGFBP7 expression. Quantitative methylation specific polymerase chain reaction showed that IGFBP7 methylation was more frequently detected in prostate cancer (60% (31/52)) and high grade prostatic intraepithelial neoplasia (40% (6/15)) samples compared to histologically benign prostate adjacent to tumor (10%) and benign prostatic hyperplasia (0%) samples. CONCLUSIONS: To our knowledge this is the first report of aberrant IGFBP7 promoter hypermethylation and concurrent IGFBP7 gene silencing in prostate cancer cell lines. Results demonstrate that CpG methylation of IGFBP7 may represent a novel biomarker of prostate cancer and pre-invasive neoplasms. Thus, future examination of IGFBP7 methylation and expression in a larger patient cohort, including bodily fluids, is justified to further evaluate its role in a diagnostic and prognostic setting.

In silico analysis and DHPLC screening strategy identifies novel apoptotic gene targets of aberrant promoter hypermethylation in prostate cancer.

BACKGROUND: Aberrant DNA methylation has been implicated as a key survival mechanism in cancer, whereby promoter hypermethylation silences genes essential for many cellular processes including apoptosis. Limited data is available on the methylation profile of apoptotic genes in prostate cancer (CaP). The aim of this study was to profile methylation of apoptotic-related genes in CaP using denaturing high performance liquid chromatography (DHPLC). METHODS: Based on an in silico selection process, 13 genes were screened for methylation in CaP cell lines using DHPLC. Quantitative methylation specific PCR was employed to determine methylation levels in prostate tissue specimens (n = 135), representing tumor, histologically benign prostate, high-grade prostatic intraepithelial neoplasia and benign prostatic hyperplasia. Gene expression was measured by QRT-PCR in cell lines and tissue specimens. RESULTS: The promoters of BIK, BNIP3, cFLIP, TMS1, DCR1, DCR2, and CDKN2A appeared fully or partially methylated in a number of malignant cell lines. This is the first report of aberrant methylation of BIK, BNIP3, and cFLIP in CaP. Quantitative methylation analysis in prostate tissues identified 5 genes (BNIP3, CDKN2A, DCR1, DCR2 and TMS1) which were frequently methylated in tumors but were unmethylated in 100% of benign tissues. Furthermore, 69% of tumors were methylated in at least one of the five-gene panel. In the case of all genes, except BNIP3, promoter hypermethylation was associated with concurrent downregulation of gene expression. CONCLUSION: Future examination of this "CaP apoptotic methylation signature" in a larger cohort of patients is justified to further evaluate its value as a diagnostic and prognostic marker.

Altered lipid metabolism in Hfe-knockout mice promotes severe NAFLD and early fibrosis.

The HFE protein plays a crucial role in the control of cellular iron homeostasis. Steatosis is commonly observed in HFE-related iron-overload disorders, and current evidence suggests a causal link between iron and steatosis. Here, we investigated the potential contribution of HFE mutations to hepatic lipid metabolism and its role in the pathogenesis of nonalcoholic fatty liver disease. Wild-type (WT) and Hfe knockout mice (Hfe(-/-)) were fed either standard chow, a monounsaturated low fat, or a high-fat, high-carbohydrate diet (HFD) and assessed for liver injury, body iron status, and markers of lipid metabolism. Despite hepatic iron concentrations and body weights similar to WT controls, Hfe(-/-) mice fed the HFD developed severe hypoxia-related steatohepatitis, Tnf-α activation, and mitochondrial respiratory complex and antioxidant dysfunction with early fibrogenesis. These features were associated with an upregulation in the expression of genes involved in intracellular lipid synthesis and trafficking, while transcripts for mitochondrial fatty acid ß-oxidation and adiponectin signaling-related genes were significantly attenuated. In contrast, HFD-fed WT mice developed bland steatosis only, with no inflammation or fibrosis and no upregulation of lipogenesis-related genes. A HFD led to reduced hepatic iron in Hfe(-/-) mice compared with chow-fed mice, despite higher serum iron, decreased hepcidin expression, and increased duodenal ferroportin mRNA. In conclusion, our results demonstrate that Hfe(-/-) mice show defective hepatic-intestinal iron and lipid signaling, which predispose them toward diet-induced hepatic lipotoxicity, accompanied by an accelerated progression of injury to fibrosis.

Risk and protective genetic variants in suicidal behaviour: association with SLC1A2, SLC1A3, 5-HTR1B &NTRK2 polymorphisms.

BACKGROUND: Suicidal behaviour is known to aggregate in families. Patients with psychiatric disorders are at higher risk for suicide attempts (SA), however protective and risk genetic variants for suicide appear to be independent of underlying psychiatric disorders. Here we investigate genetic variants in genes important for neurobiological pathways linked to suicidal behaviour and/or associated endophenotypes, for association with SA among patients with co-existing psychiatric illness. Selected gene-gene and gene-environment interactions were also tested. METHODS: DNA was obtained from bloods of 159 patients (76 suicide attempters and 83 non-attempters), who were profiled for DSM-IV Axis I psychiatric diagnosis. Twenty-eight single nucleotide polymorphisms (SNPs) from 18 candidate genes (COMT, 5-HT2A, 5-HT1A, 5-HTR1B, TPH1, MAO-A, TPH2, DBH, CNR1, BDNF, ABCG1, GABRA5, GABRG2, GABRB2, SLC1A2, SLC1A3, NTRK2, CRHR1) were genotyped. Genotyping was performed by KBioscience. Tests of association between genetic variants and SA were conducted using Chi squared and Armitage Trend tests. Binary logistical regression analyses were performed to evaluate the contribution of individual genetic variants to the prediction of SA, and to examine SNPs for potential gene-gene and gene-environment interactions. RESULTS: Our analysis identified 4 SNPs (rs4755404, rs2269272, rs6296 and rs1659400), which showed evidence of association with SA compared to a non-attempter control group. We provide evidence of a 3-locus gene-gene interaction, and a putative gene-environment interaction, whereby genetic variation at the NTRK2 locus may moderate the risk associated with history of childhood abuse. CONCLUSION: Preliminary findings suggest that allelic variability in SLC1A2/3, 5-HTR1B and NTRK2 may be relevant to the underlying diathesis for suicidal acts.

Genome-wide DNA methylation meta-analysis in the brains of suicide completers.

Suicide is the second leading cause of death globally among young people representing a significant global health burden. Although the molecular correlates of suicide remains poorly understood, it has been hypothesised that epigenomic processes may play a role. The objective of this study was to identify suicide-associated DNA methylation changes in the human brain by utilising previously published and unpublished methylomic datasets. We analysed prefrontal cortex (PFC, n = 211) and cerebellum (CER, n = 114) DNA methylation profiles from suicide completers and non-psychiatric, sudden-death controls, meta-analysing data from independent cohorts for each brain region separately. We report evidence for altered DNA methylation at several genetic loci in suicide cases compared to controls in both brain regions with suicide-associated differentially methylated positions enriched among functional pathways relevant to psychiatric phenotypes and suicidality, including nervous system development (PFC) and regulation of long-term synaptic depression (CER). In addition, we examined the functional consequences of variable DNA methylation within a PFC suicide-associated differentially methylated region (PSORS1C3 DMR) using a dual luciferase assay and examined expression of nearby genes. DNA methylation within this region was associated with decreased expression of firefly luciferase but was not associated with expression of nearby genes, PSORS1C3 and POU5F1. Our data suggest that suicide is associated with DNA methylation, offering novel insights into the molecular pathology associated with suicidality.

Methylation QTLs in the developing brain and their enrichment in schizophrenia risk loci.

We characterized DNA methylation quantitative trait loci (mQTLs) in a large collection (n = 166) of human fetal brain samples spanning 56-166 d post-conception, identifying >16,000 fetal brain mQTLs. Fetal brain mQTLs were primarily cis-acting, enriched in regulatory chromatin domains and transcription factor binding sites, and showed substantial overlap with genetic variants that were also associated with gene expression in the brain. Using tissue from three distinct regions of the adult brain (prefrontal cortex, striatum and cerebellum), we found that most fetal brain mQTLs were developmentally stable, although a subset was characterized by fetal-specific effects. Fetal brain mQTLs were enriched amongst risk loci identified in a recent large-scale genome-wide association study (GWAS) of schizophrenia, a severe psychiatric disorder with a hypothesized neurodevelopmental component. Finally, we found that mQTLs can be used to refine GWAS loci through the identification of discrete sites of variable fetal brain methylation associated with schizophrenia risk variants.

DNA Methylation Profiling in Inflammatory Bowel Disease Provides New Insights into Disease Pathogenesis.

BACKGROUND AND AIMS: Inflammatory bowel diseases (IBDs) are heterogeneous disorders with complex aetiology. Quantitative genetic studies suggest that only a small proportion of the disease variance observed in IBD is accounted for by genetic variation, indicating a potential role for differential epigenetic regulation in disease aetiology. The aim of this study was to assess genome-wide DNA methylation changes specifically associated with ulcerative colitis (UC), Crohn's disease (CD) and IBD activity. METHODS: DNA methylation was quantified in peripheral blood mononuclear cells (PBMCs) from 149 IBD cases (61 UC, 88 CD) and 39 controls using the Infinium HumanMethylation450 BeadChip. Technical and functional validation was performed using pyrosequencing and the real-time polymerase chain reaction. Cross-tissue replication of the top differentially methylated positions (DMPs) was tested in colonic mucosa tissue samples obtained from paediatric IBD cases and controls. RESULTS: A total of 3196 probes were differentially methylated between CD cases and controls, while 1481 probes were differentially methylated between UC cases and controls. There was considerable (45%) overlap between UC and CD DMPs. The top-ranked IBD-associated PBMC differentially methylated region (promoter region of TRIM39-RPP2) was also significantly hypomethylated in colonic mucosa from paediatric UC patients. In addition, we confirmed TRAF6 hypermethylation using pyrosequencing and found reduced TRAF6 gene expression in PBMCs of IBD patients. CONCLUSIONS: Our data provide new insights into differential epigenetic regulation of genes and molecular pathways, which may contribute to the pathogenesis and activity of IBD.

Anxiety is associated with higher levels of global DNA methylation and altered expression of epigenetic and interleukin-6 genes.

OBJECTIVES: Anxiety is associated with elevated levels of the inflammatory cytokine interleukin-6 (IL-6) and an increased risk for diseases with an inflammatory aetiology. In cancer, higher levels of IL-6 have been associated with increased expression of the epigenetic enzymes DNMT1 and Enhancer of Zeste Homolog 2 (EZH2). However, the relationship between IL-6 and DNA methyltransferases (DNMTs) and EZH2 expression has not previously been examined in anxious individuals. METHODS: Global DNA methylation levels were measured using the Methylflash Methylated DNA Quantification Kit and gene expression levels of the DNMT and EZH2 genes in anxious (n=25) and nonanxious individuals (n=22) were compared using quantitative real-time PCR. Specifically, we investigated whether global DNA methylation or aberrant expression of these genes was correlated with IL-6 mRNA and protein serum levels in anxious individuals. RESULTS: Anxious participants had significantly higher levels of global DNA methylation compared with controls (P=0.001). There were no differences in the mean mRNA expression levels of the DNMT1/3A/3B, EZH2 and IL-6 genes in anxious individuals compared with controls. However, the expression of DNMT1/3A, EZH2 and IL-6 genes increases with increasing Hospital Anxiety and Depression Scale-Anxiety scores in the anxious cohort only. Interestingly, IL-6 gene expression was correlated strongly with DNMT1/3A/3B and EZH2 expression, highlighting a potential relationship between IL-6 and important epigenetic regulatory enzymes. CONCLUSION: This study provides novel insight into the relationship between anxiety, epigenetics and IL-6. Moreover, our findings support the hypothesis that changes in DNA methylation profiles may contribute to the biology of anxiety.

Methylomic markers of persistent childhood asthma: a longitudinal study of asthma-discordant monozygotic twins.

BACKGROUND: Asthma is the most common chronic inflammatory disorder in children. The aetiology of asthma pathology is complex and highly heterogeneous, involving the interplay between genetic and environmental risk factors that is hypothesized to involve epigenetic processes. Our aim was to explore whether methylomic variation in early childhood is associated with discordance for asthma symptoms within monozygotic (MZ) twin pairs recruited from the Environmental Risk (E-Risk) longitudinal twin study. We also aimed to identify differences in DNA methylation that are associated with asthma that develops in childhood and persists into early adulthood as these may represent useful prognostic biomarkers. RESULTS: We examined genome-wide patterns of DNA methylation in buccal cell samples collected from 37 MZ twin pairs discordant for asthma at age 10. DNA methylation at individual CpG sites demonstrated significant variability within discordant MZ twin pairs with the top-ranked nominally significant differentially methylated position (DMP) located in the HGSNAT gene. We stratified our analysis by assessing DNA methylation differences in a sub-group of MZ twin pairs who remained persistently discordant for asthma at age 18. The top-ranked nominally significant DMP associated with persisting asthma is located in the vicinity of the HLX gene, which has been previously implicated in childhood asthma. CONCLUSIONS: We identified DNA methylation differences associated with childhood asthma in peripheral DNA samples from discordant MZ twin pairs. Our data suggest that differences in DNA methylation associated with childhood asthma which persists into early adulthood are distinct from those associated with asthma which remits.

Methylomic analysis of monozygotic twins discordant for childhood psychotic symptoms.

Childhood psychotic symptoms are associated with increased rates of schizophrenia, other psychiatric disorders, and suicide attempts in adulthood; thus, elucidating early risk indicators is crucial to target prevention efforts. There is considerable discordance for psychotic symptoms between monozygotic twins, indicating that child-specific non-genetic factors must be involved. Epigenetic processes may constitute one of these factors and have not yet been investigated in relation to childhood psychotic symptoms. Therefore, this study explored whether differences in DNA methylation at age 10 were associated with monozygotic twin discordance for psychotic symptoms at age 12. The Environmental Risk (E-Risk) Longitudinal Twin Study cohort of 2,232 children (1,116 twin pairs) was assessed for age-12 psychotic symptoms and 24 monozygotic twin pairs discordant for symptoms were identified for methylomic comparison. Children provided buccal samples at ages 5 and 10. DNA was bisulfite modified and DNA methylation was quantified using the Infinium HumanMethylation450 array. Differentially methylated positions (DMPs) associated with psychotic symptoms were subsequently tested in post-mortem prefrontal cortex tissue from adult schizophrenia patients and age-matched controls. Site-specific DNA methylation differences were observed at age 10 between monozygotic twins discordant for age-12 psychotic symptoms. Similar DMPs were not found at age 5. The top-ranked psychosis-associated DMP (cg23933044), located in the promoter of the C5ORF42 gene, was also hypomethylated in post-mortem prefrontal cortex brain tissue from schizophrenia patients compared to unaffected controls. These data tentatively suggest that epigenetic variation in peripheral tissue is associated with childhood psychotic symptoms and may indicate susceptibility to schizophrenia and other mental health problems.