GWAS SNPs Research Articles

Identifying causal models between genetically regulated methylation patterns and gene expression in healthy colon tissue

BackgroundDNA methylation is involved in the regulation of gene expression and phenotypic variation, but the inter-relationship between genetic variation, DNA methylation and gene expression remains poorly understood. Here we combine the analysis of genetic variants related to methylation markers (methylation quantitative trait loci: mQTLs) and gene expression (expression quantitative trait loci: eQTLs) with methylation markers related to gene expression (expression quantitative trait methylation: eQTMs), to provide novel insights into the genetic/epigenetic architecture of colocalizing molecular markers.ResultsNormal mucosa from 100 patients with colon cancer and 50 healthy donors included in the Colonomics project have been analyzed. Linear models have been used to find mQTLs and eQTMs within 1 Mb of the target gene. From 32,446 eQTLs previously detected, we found a total of 6850 SNPs, 114 CpGs and 52 genes interrelated, generating 13,987 significant combinations of co-occurring associations (meQTLs) after Bonferromi correction. Non-redundant meQTLs were 54, enriched in genes involved in metabolism of glucose and xenobiotics and immune system. SNPs in meQTLs were enriched in regulatory elements (enhancers and promoters) compared to random SNPs within 1 Mb of genes. Three colorectal cancer GWAS SNPs were related to methylation changes, and four SNPs were related to chemerin levels. Bayesian networks have been used to identify putative causal relationships among associated SNPs, CpG and gene expression triads. We identified that most of these combinations showed the canonical pathway of methylation markers causes gene expression variation (60.1%) or non-causal relationship between methylation and gene expression (33.9%); however, in up to 6% of these combinations, gene expression was causing variation in methylation markers.ConclusionsIn this study we provided a characterization of the regulation between genetic variants and inter-dependent methylation markers and gene expression in a set of 150 healthy colon tissue samples. This is an important finding for the understanding of molecular susceptibility on colon-related complex diseases.

GESLM algorithm for detecting causal SNPs in GWAS with multiple phenotypes.

With the development of genome-wide association studies, how to gain information from a large scale of data has become an issue of common concern, since traditional methods are not fully developed to solve problems such as identifying loci-to-loci interactions (also known as epistasis). Previous epistatic studies mainly focused on local information with a single outcome (phenotype), while in this paper, we developed a two-stage global search algorithm, Greedy Equivalence Search with Local Modification (GESLM), to implement a global search of directed acyclic graph in order to identify genome-wide epistatic interactions with multiple outcome variables (phenotypes) in a case-control design. GESLM integrates the advantages of score-based methods and constraint-based methods to learn the phenotype-related Bayesian network and is powerful and robust to find the interaction structures that display both genetic associations with phenotypes and gene interactions. We compared GESLM with some common phenotype-related loci detecting methods in simulation studies. The results showed that our method improved the accuracy and efficiency compared with others, especially in an unbalanced case-control study. Besides, its application on the UK Biobank dataset suggested that our algorithm has great performance when handling genome-wide association data with more than one phenotype.

177 Integrative computational analysis identifies eQTL-lincRNA pairs associated with alopecia areata

Alopecia Areata (AA) is an autoimmune disorder characterized by hair loss ranging from patchy disease to total body hair loss. AA is one of the most prevalent autoimmune diseases in the US, with a lifetime risk of 1.7%. Our previous GWAS studies and subsequent meta-analysis in AA patients revealed many SNPs associated with increased disease risk. In complex polygenic diseases like AA, >90% of GWAS SNPs map to non-coding regions, and at least 60% of them reside in cis-regulatory elements (CREs) such as lincRNAs, miRNA binding sites and CTCF sites. Recently, multiple studies reported autoimmune disease associated GWAS SNPs in lincRNAs impacting disease through reshaping three dimensional (3D)-genome architecture. However, the impact of GWAS SNPs in lincRNAs and their impact on 3D-genome structure in AA has not been evaluated. We systematically screened AA associated GWAS SNPs in CREs using publicly available studies in the GWAS catalog and retrieved 44 SNPs. We selected 27 AA GWAS SNPs with p-value < 5E-8 for downstream analysis, suggesting these SNPs may function as eQTLs. These SNPs were further screened for CREs in the UCSC genome browser. A total of 5 AA SNPs (rs7299099, rs2155219, rs694739, rs4348998, rs9479482) were found in the CREs and exhibited chromatin signatures for enhancer RNAs. Further, single-tissue eQTL analysis using GTEx portal identified 3 eQTL-lincRNAs pairs, out of which 2 (rs694739-AP003774.2 and rs9479482-AL355497.2) exhibited the highest p-values, P=5.37E-21 and P=1.12E-12 for skin, respectively. By using publicly available chromatin conformation capture studies in dermal fibroblasts, we visualized these eQTL-lincRNAs pairs in Juicebox and found them located in the same topological associated domain (TAD). It is known that CREs in the same TAD tend to operate together at the 3D-genome level, predicting that these SNPs may influence the 3D genome structure in AA. To validate our computational findings, we are performing hi-C sequencing in AA scalp skin to define the impact of GWAS SNPs in CREs on chromatin looping in AA.

Transcriptome-wide high-throughput mapping of protein\u2013RNA occupancy profiles using POP-seq

Interaction between proteins and RNA is critical for post-transcriptional regulatory processes. Existing high throughput methods based on crosslinking of the protein–RNA complexes and poly-A pull down are reported to contribute to biases and are not readily amenable for identifying interaction sites on non poly-A RNAs. We present Protein Occupancy Profile-Sequencing (POP-seq), a phase separation based method in three versions, one of which does not require crosslinking, thus providing unbiased protein occupancy profiles on whole cell transcriptome without the requirement of poly-A pulldown. Our study demonstrates that ~ 68% of the total POP-seq peaks exhibited an overlap with publicly available protein–RNA interaction profiles of 97 RNA binding proteins (RBPs) in K562 cells. We show that POP-seq variants consistently capture protein–RNA interaction sites across a broad range of genes including on transcripts encoding for transcription factors (TFs), RNA-Binding Proteins (RBPs) and long non-coding RNAs (lncRNAs). POP-seq identified peaks exhibited a significant enrichment (p value < 2.2e−16) for GWAS SNPs, phenotypic, clinically relevant germline as well as somatic variants reported in cancer genomes, suggesting the prevalence of uncharacterized genomic variation in protein occupied sites on RNA. We demonstrate that the abundance of POP-seq peaks increases with an increase in expression of lncRNAs, suggesting that highly expressed lncRNA are likely to act as sponges for RBPs, contributing to the rewiring of protein–RNA interaction network in cancer cells. Overall, our data supports POP-seq as a robust and cost-effective method that could be applied to primary tissues for mapping global protein occupancies.

The Long Life Family Study: Sequencing Exceptional Pedigrees for Rare Protective Variants

The Long Life Family Study (LLFS) has longitudinally measured key aging phenotypes on 4,953 participants (539 pedigrees) in the USA and Denmark selected for exceptional familial longevity. On average, both generations of the LLFS sample are healthier than average for their age/sex, for many phenotypes. However, the pedigrees are heterogeneous, with different families showing familial clustering of protection for different phenotypes. Linkage analyses identified extremely strong genetic linkage peaks for many cross-sectional as well as longitudinal trajectory rates of change phenotypes. These peaks are NOT explained by GWAS SNPs (either measured or imputed). Pedigree specific HLODs and preliminary deep sequencing suggests that these peaks are driven by rare, protective variants running in selected pedigrees. Whole Genome Sequencing, a third longitudinal visit, and extensive OMICs (transcriptomics, epigenomics, metabolomics and proteomics) will help us resolve the mechanisms behind these protective genetically linked variants, and could illuminate new biology and enable new therapeutics.

Whole Genome Linkage Scan Identifies a Novel Locus on 3q28 for TG/HDL-C Ratio Change Over Time

TG/HDL-C ratio (THR) represents a single inherited surrogate predictor of hyperinsulinemia or insulin resistance that is associated with premature aging processes, risk of diabetes and increased mortality. To identify genetic loci for THR change over time (ΔTHR), we conducted a whole genome linkage scan among subjects of European ancestry who had complete data from two exams collected about seven years apart from the Long Life Family Study (LLFS, n=3091), a study with familial clustering of exceptional longevity in the US and Denmark. Subjects with diabetes or using medications for dyslipidemia were excluded from this analysis. ΔTHR was derived using growth curve modeling, and adjusted for age, sex, PCs, familial membership, and then log-transformed to approximate normality. Our linkage scan was built on haplotype-based IBD estimation with 0.5 cM average spacing. Heritability of ΔTHR was moderate (46%), and evidence for significant linkage (LOD>3) was identified on 3q28 (LOD=4.1). This locus harbors ADIPOQ among several other promising candidate genes. Interestingly, several studies previously reported suggestive evidence of linkage at this locus for relevant traits including adiponectin, dementia, AD and SBP. This linkage signal was not explained by significant GWAS SNPs for LPL or those under the peak (LOD attenuated to 3.7). In conclusion, we found a novel genetic locus on 3q28 for ΔTHR in subjects without diabetes selected for exceptional survival and healthy aging. Further query of sequence elements including rare functional and regulatory variants at this locus is underway which may reveal novel insights on insulin resistance mechanisms for aging.

Discovering genetic variants contributing to sleep dysfunction in presymptomatic Alzheimer’s mice

AbstractBackgroundDisordered sleep is frequently observed in Alzheimer’s disease patients. Sleep loss is associated with increased Aβ and tau accumulation in both AD patients and model organisms. Furthermore, cognitively intact individuals with impaired sleep are at higher risk for AD. Although individual differences in sleep are known to be highly heritable, little is known of the genetic factors associated with sleep dysfunction in aging and AD and if these contribute to disease development.MethodSleep was assessed at a presymptomatic age (6‐7 months) in a panel of genetically diverse mice carrying familial AD mutations, which better models complex etiology of human AD (AD‐BXDs; Neuner et al 2019). Sleep‐wake activity was measured for five consecutive days using the PiezoSleep tracking system across 25 AD‐BXD strains and their nontransgenic littermates (Ntg‐BXD).ResultAD‐BXD mice slept significantly less during the dark cycle compared to Ntg‐BXDs. Additionally, reduction of sleep was observed during the transition from dark to the light cycle, suggesting disruption of circadian function. Females slept less than males regardless of AD mutation status. We identified a significant quantitative trait locus (QTL) on chromosome 9 associated with total sleep, this region is syntenic with a regions on the human chromosome 11, harboring six sleep‐associated human GWAS SNPs, demonstrating the relevance of this region to human biology. Additionally, we identified a QTL for hourly sleep at ZT14 that mapped to intron 1 of the gene Kirrel3 of chromosome 9 (∼20 mb downstream of the sleep‐QTL). Differential gene expression implicates Kirrel3 as a novel target‐gene for validation.ConclusionWe provide the first evidence that genetic factors interact with AD mutations to influence sleep behavior, and identified two regions causally associated with variation in sleep in the AD‐BXDs. Mouse‐to‐human comparison suggest that both region are relevant to humans, demonstrating the value of the AD‐BXDs as a translational AD model. Future work will leverage genome engineering to correct sleep‐associated variants to test the delay or prevention of cognitive deficits onset apparent in the AD‐BXDs by 14 months of age. These studies may identify novel therapeutic targets to treat AD dementia.

Cytokines mapping for tissue-specific expression, eQTLs and GWAS traits

Dysregulation in cytokine production has been linked to the pathogenesis of various immune-mediated traits, in which genetic variability contributes to the etiopathogenesis. GWA studies have identified many genetic variants in or near cytokine genes, nonetheless, the translation of these findings into knowledge of functional determinants of complex traits remains a fundamental challenge. In this study we aimed at collection, analysis and interpretation of data on cytokines focused on their tissue-specific expression, eQTLs and GWAS traits. Using GO annotations, we generated a list of 314 cytokines and analyzed them with the GTEx resource. Cytokines were highly tissue-specific, 82.3% of cytokines had Tau expression metrics ≥ 0.8. In total, 3077 associations for 1760 unique SNPs in or near 244 cytokines were mapped in the NHGRI-EBI GWAS Catalog. According to the Experimental Factor Ontology resource, the largest numbers of disease associations were related to ‘Inflammatory disease’, ‘Immune system disease’ and ‘Asthma’. The GTEx-based analysis revealed that among GWAS SNPs, 1142 SNPs had eQTL effects and influenced expression levels of 999 eGenes, among them 178 cytokines. Several types of enrichment analysis showed that it was cytokines expression variability that fundamentally contributed to the molecular origins of considered immune-mediated conditions.

What Have We Learned from GWAS for Atopic Dermatitis?

GWASs have revealed multiple loci associated with atopic dermatitis (AD). Some have confirmed pre-existing knowledge, including the role of skin barrier and type 2 inflammation in AD pathogenesis, whereas others have provided newer insights, including evidence of autoimmunity and previously unrecognized genes controlling epidermal differentiation. The majority of risk loci are in intergenic regions for which functional mechanism(s) remain unknown. These loci require detailed molecular studies carried out in cells and tissues of relevance to AD. Genomic findings to date account for ∼30% of AD heritability, therefore, considerable further work is needed to fully understand individual risk.

THU0225 INTEGRATIVE PLASMA METABOLOME AND TRANSCRIPTOME ANALYSIS REVEALED THE IMPORTANCE OF HISTIDINE HOMEOSTASIS IN SLE PATHOGENESIS WITH POTENTIAL FOR IMPROVED SLE PATIENTS STRATIFICATION

Background:Recently, immunometabolism has gathered attention of many immunologists. It has been widely recognized that metabolic reprogramming in each immune cell brings different effects on different cells and is important for regulating their functions. Along with the progress of statistical genetics, serum metabolites were shown to be under genetic regulations1). Metabolic changes are now considered not only to be mere phenotypes of cells but also to be key factors for controlling immune cell differentiation, proliferation and function through regulating gene expressions eventually. Although genome-wide association studies have brought deep insights into SLE pathogenesis, the precise pathway from genome to metabolome has been largely unknown, and vice versa.Objectives:The aim of this study is to investigate metabolomic regulation in SLE in relation to gene expressions by integrating plasma metabolome data and transcriptome data.Methods:We collected plasma samples from patients with SLE (n=57) who met the 1997 American College of Rheumatology criteria for SLE. Gender- and age-matched healthy controls (HCs) (n=56) were recruited. Metabolic profiles focusing on 39 amino acids were analyzed with liquid chromatography (LC)-mass spectrometry. Transcriptome data of SLE patients were obtained from our RNA-sequencing data of each immune cell subset (total 19 subsets). Whole-genome sequencing was also performed.Results:Our previous experiment showed that about 160 peaks were detected from comprehensive LC-TOFMS and amino acids were useful for distinguishing SLE patients from HCs. Both partial least squares discriminant analysis (PLS-DA) and random forest, a machine learning algorithm, revealed the importance of histidine (His), one of the essential amino acids, to classify SLE patients from HCs, whose plasma level was lower in SLE patients. In addition, inverse correlation between His level and titer of ds-DNA as well as damage index (SDI) was detected. His level was correlated neither with PSL dosage nor with type I interferon (IFN) signature. Receiver operating characteristic (ROC) analysis showed the best predictability for SLE with the combination of specific amino acids including His. Our transcriptome analysis has revealed the significance of oxidative phosphorylation (OXPHOS) in B cells for SLE pathogenesis. Interestingly, OXPHOS signature was inversely correlated with His level in SLE B cells.Conclusion:His may be an important factor for SLE pathogenesis especially in B cells independently from IFN signal. SLC15A4, a transporter of His on lysosome, is one of the SLE GWAS SNPs and has been reported to play an important role in IFN production in B cells through regulation of TLR7/9 activation 2). We also identified that SLE patients with risk allele of SLC15A4 had tendency to show higher plasma His level, indicating His homeostasis could become a novel treatment target for SLE. Moreover, the inverse correlation of His level to SDI as well as OXPHOS signature suggests that His might play a key role for promoting organ damages in SLE.

Determinants of transcription factor regulatory range

Characterization of the genomic distances over which transcription factor (TF) binding influences gene expression is important for inferring target genes from TF chromatin immunoprecipitation followed by sequencing (ChIP-seq) data. Here we systematically examine the relationship between thousands of TF and histone modification ChIP-seq data sets with thousands of gene expression profiles. We develop a model for integrating these data, which reveals two classes of TFs with distinct ranges of regulatory influence, chromatin-binding preferences, and auto-regulatory properties. We find that the regulatory range of the same TF bound within different topologically associating domains (TADs) depend on intrinsic TAD properties such as local gene density and G/C content, but also on the TAD chromatin states. Our results suggest that considering TF type, binding distance to gene locus, as well as chromatin context is important in identifying implicated TFs from GWAS SNPs.

Connecting GWAS SNPs to Target Genes in primary human immune cell types

Abstract Expression quantitative trait loci (eQTLs) studies have provided associations with gene expression for many non-coding genetic variants associated with human disease and evolution. However, they fail to pinpoint the functionally important eQTLs or provide mechanistic insights. Also, most eQTL discovery studies restrict analysis to SNPs within ±1 Mb of the TSS of the gene tested to minimize corrections for multiple association tests. Using proximity ligation-assisted H3K27ac ChIP-seq (HiChIP) assays, here we mapped eQTLs that overlap active cis-regulatory elements and interact with their target gene promoters (promoter-interacting eQTLs, pieQTLs) in five common immune cell types. This allowed us to narrow down functionally important eQTLs and show mechanisms that explain their cell type restriction. A subset of the identified pieQTLs overlapped non-transcribed promoters, which interact with their target gene promoters and function as enhancers as validated by CRISPRi. In addition, we devised a novel approach to eQTL discovery that relied on HiChIP-based promoter interaction maps as a structural framework for deciding what SNPs to test for association with gene expression. We discovered ultralong distance pieQTLs (&amp;gt;1Mb from target gene), including several disease-risk variants and new eGenes such as FHIT for which we validated the role of an ultralong distance pieQTL by CRISPRi. Our study provides insights into cell-specific gene regulatory mechanisms and highlights the potentially functional non-coding variants (pieQTLs) linked to human diseases.

Chromatin interaction maps of active cis-regulatory elements in primary human immune cells reveal extensive overlap with disease-associated genetic variants

Abstract Surveys of purified human immune cell types have revealed that common genetic variants have profound effects of gene expression in a highly cell-specific manner. However, it has been challenging to determine the mechanisms behind cell-specific regulatory effects as well as to pinpoint the potentially functional variants in dense haplotype blocks. Here, we performed genome-wide ChIP-Seq and proximity ligation-assisted ChIP-seq assays (HiChIP) for the H3K27ac modification to capture physical interactions between active cis-regulatory elements and their target gene promoters for 5 common and diverse immune cell types from 6 donors. The generated interaction maps identified long-range loops (over 20kb) for more than 90% of active cis-regulatory regions. We found that majority of the cis-regulatory interactions were highly cell-specific, which in turn correlated with cell-specific transcription factor binding motifs and gene expression patterns. Nearly 10% of expression quantitative trait loci (eQTLs), including disease-associated eQTLs and GWAS SNPs, were located at active cis-regulatory regions that interacted with the target genes of eQTLs (eGenes), thus allowing us to narrow down potentially functional/causative QTLs. Majority of cis-eQTLs that are specific to only a single cell type overlapped cis-regulatory elements that showed interactions with the promoter of their target eGenes only in that specific cell type. Our study thus provides a comprehensive interaction map of active regulatory elements for each of the 5 immune cell types across multiple donors and highlights the cell type-specificity of potentially functional variants and interactions linking them to genes in human disease.

Studies of liver tissue identify functional gene regulatory elements associated to gene expression, type 2 diabetes, and other metabolic diseases

BackgroundGenome-wide association studies (GWAS) of diseases and traits have found associations to gene regions but not the functional SNP or the gene mediating the effect. Difference in gene regulatory signals can be detected using chromatin immunoprecipitation and next-gen sequencing (ChIP-seq) of transcription factors or histone modifications by aligning reads to known polymorphisms in individual genomes. The aim was to identify such regulatory elements in the human liver to understand the genetics behind type 2 diabetes and metabolic diseases.MethodsThe genome of liver tissue was sequenced using 10X Genomics technology to call polymorphic positions. Using ChIP-seq for two histone modifications, H3K4me3 and H3K27ac, and the transcription factor CTCF, and our established bioinformatics pipeline, we detected sites with significant difference in signal between the alleles.ResultsWe detected 2329 allele-specific SNPs (AS-SNPs) including 25 associated to GWAS SNPs linked to liver biology, e.g., 4 AS-SNPs at two type 2 diabetes loci. Two hundred ninety-two AS-SNPs were associated to liver gene expression in GTEx, and 134 AS-SNPs were located on 166 candidate functional motifs and most of them in EGR1-binding sites.ConclusionsThis study provides a valuable collection of candidate liver regulatory elements for further experimental validation.

Genomic and epigenomic mapping of leptin-responsive neuronal populations involved in body weight regulation.

Genome wide association studies (GWAS) in obesity have identified a large number of noncoding loci located near genes expressed in the central nervous system. However, due to the difficulties in isolating and characterizing specific neuronal subpopulations, few obesity-associated SNPs have been functionally characterized. Leptin responsive neurons in the hypothalamus are essential in controlling energy homeostasis and body weight. Here, we combine FACS-sorting of leptin-responsive hypothalamic neuron nuclei with genomic and epigenomic approaches (RNA-seq, ChIP-seq, ATAC-seq) to generate a comprehensive map of leptin-response specific regulatory elements, several of which overlap obesity-associated GWAS variants. We demonstrate the usefulness of our leptin-response neuron regulome, by functionally characterizing a novel enhancer near Socs3, a leptin response-associated transcription factor. We envision our data to serve as a useful resource and a blueprint for functionally characterizing obesity-associated SNPs in the hypothalamus.

Evolution and Comprehensive Analysis of DNaseI Hypersensitive Sites in Regulatory Regions of Primate Brain-Related Genes.

How the human brain differs from those of non-human primates is largely unknown and the complex drivers underlying such differences at the genomic level remain unclear. In this study, we selected 243 brain-related genes, based on Gene Ontology, and identified 184,113 DNaseI hypersensitive sites (DHSs) within their regulatory regions. To performed comprehensive evolutionary analyses, we set strict filtering criteria for alignment quality and filtered 39,132 DHSs for inclusion in the investigation and found that 2,397 (~6%) exhibited evidence of accelerated evolution (aceDHSs), which was a much higher proportion that DHSs genome-wide. Target genes predicted to be regulated by brain-aceDHSs were functionally enriched for brain development and exhibited differential expression between human and chimpanzee. Alignments indicated 61 potential human-specific transcription factor binding sites in brain-aceDHSs, including for CTCF, FOXH1, and FOXQ1. Furthermore, based on GWAS, Hi-C, and eQTL data, 16 GWAS SNPs, and 82 eQTL SNPs were in brain-aceDHSs that regulate genes related to brain development or disease. Among these brain-aceDHSs, we confirmed that one enhanced the expression of GPR133, using CRISPR-Cas9 and western blotting. The GPR133 gene is associated with glioblastoma, indicating that SNPs within DHSs could be related to brain disorders. These findings suggest that brain-related gene regulatory regions are under adaptive evolution and contribute to the differential expression profiles among primates, providing new insights into the genetic basis of brain phenotypes or disorders between humans and other primates.

46 - THE POLYGENIC EPISTASIS RISK SCORE DEMONSTRATES SIGNIFICANT ROLE OF GENE INTERACTION IN BIPOLAR DISORDER

Background Gene-gene interaction occurs when the association of one locus to a phenotype depends on the genotype at a different locus. GWAS has made possible the discovery of dozens of genes for psychiatric illnesses generally using a model in which there is no gene-gene interaction. Even a modest GWAS may include 1 M SNPs, making an exhaustive search for interaction vastly underpowered. Yet, interaction effects are frequently found in defined biological systems, though they may have significant effect sizes, none can survive correction for genome-wide comparisons. Though polygenic models such as the polygenic risk score, achieve high reproducibility, they explain only a small portion of the genetic variance and do not incorporate interaction. A more realistic model is that in addition to a large number of independent marginal effects, there are also a very large number of interaction effects. In order to test this, I propose an extension to the polygenic risk score and demonstrate a significant contribution of gene-gene interaction in bipolar disorder. Methods The 1500 most significant SNPs in the PGCBD1 GWAS were selected for interaction testing. The model dataset was the GAIN bipolar sample of 995 cases and 1,023 controls. The test sample was a similarly collected set of subjects termed the TGEN sample of 1,199 cases and 403 controls. Interactions among these SNPs were detected using a case-case analysis as implemented in the plink –fast epistasis option. A routine was written in R, to estimate the odds ratios of each genotypic combination in each interaction detected. This array of odds ratios was then used to weight the nine genotypic combinations as seen in each individual in the test dataset. The log10 of these odds ratios then were summed over all interactions to produce a cumulative score representing genetic loading due to interactions. This I have termed the Polygenic Epistasis Risk Score (PERS). An R package will be made available. Results 11,605 interactions were detected in the GAIN sample amongst the 1500 SNPs tested that met a criterion of p Discussion The highly significant enrichment of the TGEN sample for interactions found in the GAIN sample argues for a substantial contribution of interaction in bipolar disorder. It also argues for the model of a very large number of interactions of small effect that add or otherwise combine together. This score can be combined with conventional PRS in order to capture a larger portion of the genetic variance and improve the power of such analyses.

Abstract 234: Understanding melanoma susceptibility through GWAS of risk phenotypes

Abstract Melanoma susceptibility SNPs can be categorized by their impact on known susceptibility phenotypes (pigmentation, nevus count, telomere length, or other). In an ongoing meta-analysis of melanoma, at this stage involving 17,800 cases but soon expanding to more than 30,000 cases, 26 SNPs have achieved genome-wide significance. UK Biobank is a population-based cohort of 500,000 UK persons aged at recruitment 40-69 years; participants reported pigmentation characteristics including ease of tanning, natural hair color, skin color (in the absence of tan), and number of childhood sunburns. To conduct analyses, we separately analyzed the red hair phenotype while ordering the other hair colors from black to blond. Among the 26 SNPs from the melanoma GWAS, 15 were associated with at least one pigmentation factor assessed in UK Biobank (each at p &amp;lt;10-5) with 5 being associated with all pigmentation factors (e.g., SLC45A2, ASIP, MC1R) while others were associated with a subset of factors (e.g., RAD23B/TAL2 was strongly associated with hair color alone). 13 of the melanoma SNPs were associated with "ease of tanning." Detailed analysis of "ease of tanning" SNPs found 29 genome-wide significant in a subset of UK Biobank, of which 13 were associated with melanoma risk (p &amp;lt;0.001) but notably 4 of the SNPs showed no evidence of melanoma risk even when being among the strongest effects for "ease of tanning" and the remainder showing weaker evidence of association. For all pigmentation factors (except having red hair), the effect sizes for a SNP on melanoma risk and pigmentation were highly correlated. We examined the 26 melanoma SNPs in recent GWAS of nevus count and telomere length, and similarly observed overlaps with these phenotypes. While the telomere, nevus and pigmentation instruments were largely distinct, we found evidence of some overlap in genetic susceptibility (for instance, the PLA2G6 SNP rs2092180 is strongly associated with nevus count, ease of tanning and skin color [all p &amp;lt;10-12]). Combined analysis of the melanoma GWAS results with the GWAS of each of these phenotypes provides overwhelming evidence of further susceptibility loci, indicating that GWAS of intermediate phenotypes can benefit identification of the underlying genetic profile. Supported by Cancer Research UK (C588/A19167) &amp; NIH (CA083115). This research has been conducted using the UK Biobank Resource. Citation Format: David T. Bishop, Mark M. Iles, Julia A. Newton-Bishop, Jennifer H. Barrett, Adam Trouwer, Matthew Law, Stuart MacGregor, David Duffy, Nicholas G. Martin, Nicholas Hayward, Anne Cust, Maria T. Landi, Alisa Goldstein, Jianxin Shi, Mitch Machiela, Florence Demenais, Peter Kanetsky, GenoMEL &amp; MELANOSTRUM. Understanding melanoma susceptibility through GWAS of risk phenotypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 234.

Abstract 4317: Cell-type specific epigenetic effects of tobacco smoking on the immune cell methylome

Abstract Exposure to tobacco smoke dramatically alters DNA methylation in blood cells and this immune cell modulation may mediate smoking-associated complex diseases. To identify cell-type specific, smoking-associated differentially methylated regions (SM-DMRs) in hematopoietic cells from smokers and non-smokers, we used Illumina 450/850K arrays, reduced representation (RRBS) and whole genome bisulfite sequencing (WGBS) to measure methylation on DNA extracted from whole blood and 7 leukocyte subtypes (CD4+, CD8+, CD14+ monocytes, CD15+ granulocytes, B cells, NK cells and CD34+ progenitor cells). SM-DMRs preferentially occurred at enhancers (H3K27ac), and/or CpG island shores and were enriched for transcription factor (TF) binding suggesting crosstalk between TFs, DNA methylation, and tobacco smoke exposure. SM-DMRs in or near genes AHRR, ALPPL2, and GFI1 were highly significant in CD34+, CD14+, CD15+, B cells and NK cells but were absent in CD4+ and CD8+ T cells. We also identified genomic regions with distinct cell-type and lineage-specific differences in smoking-associated methylation and gene expression. AHRR displayed increased levels of both noncoding RNA and mRNA gene expression only in CD14+ monocytes. The SM-DMR in GPR15 was highly significant in CD4+ and CD8+ T cells, B cells, and NK cells but absent in myeloid cells. To explore possible links between smoking, DNA methylation/gene expression and immune cell functions, we carried out pathway analysis on smoking-affected methylomes and transcriptomes in each cell type. The smokers' transcriptome of CD56+ NK cells showed perturbation of genes either altered in leukemias, linked with leukemia GWAS SNPs, or under various types of innate immune stimulation. In particular, the locus containing HLA-DQA1 and HLA-DQB1 is highly significantly linked with chronic lymphocytic leukemia and these genes were strongly down-regulated in NK cells of smokers. The NK cell methylome of smokers also displayed perturbation of genes altered in leukemia cell lines. To identify and characterize rare blood cells among cell-type subsets that have smoking-altered transcriptomes, we are carrying out single cell RNA-seq. These studies may reveal new biological connections between tobacco smoke exposure and smoking-associated disease etiology. Citation Format: Xuting Wang, Ma Wan, Brian D. Bennett, Gary S. Pittman, Michelle Campbell, Dan Su, Isabel J. Thompson, Suzanne N. Martos, Douglas A. Bell. Cell-type specific epigenetic effects of tobacco smoking on the immune cell methylome [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4317.

Genotyping of the Alzheimer's Disease Genome-Wide Association Study Index Single Nucleotide Polymorphisms in the Brains for Dementia Research Cohort.

The Brains for Dementia Research project is a recently established longitudinal cohort which aims to provide brain tissue for research purposes from neuropathologically defined samples. Here we present the findings from our analysis on the 19 established GWAS index SNPs for Alzheimer's disease, in order to demonstrate if the BDR sample also displays association to these variants. A highly significant association of the APOEɛ4 allele was identified (p = 3.99×10-12). Association tests for the 19 GWAS SNPs found that although no SNPs survive multiple testing, nominal significant findings were detected and concordance with the Lambert et al. GWAS meta-analysis was observed.