INTRODUCTION Translational impact. The majority of IBD genetic loci identified by GWAS resides in the non-coding region and is thought to have regulatory roles. Discovering the functional outcome of these regulatory variations is integral to understanding IBD susceptibility. In this report, we identify shared and unique eQTLs in the two key tissues of IBD the blood and intestine, which were assessed in the same set of CD patients. These eQTLs pinpoint the association between IBD genetic loci and the downstream genes under influence, as well as their magnitude and direction of genetic effect (ie., does the disease risk allele up- or downregulate downstream genes). Alongside the deeper understanding of the context of IBD susceptibility, eQTLs also provide guidance in designing new therapeutic treatments. In this way, eQTLs, either as protective or pathogenic variants, directly or indirectly could be candidates for further consideration as drug targets. EQTLs may also be predictive of drug response, as reported in other disease studies,1, 2, 3 4
Inflammatory bowel disease (IBD) is comprised of two different disease sub-types, Crohn's disease (CD) and ulcerative colitis, which are distinguished by specific clinical and pathological features. Inflammation in Crohn's is patchy, affecting the distal ileum and/or colon with peri anal involvement; whereas in ulcerative colitis, inflammation manifests in a continuous manner from the rectum toward the proximal region of the colon. Crohn's disease is characterized by transmural inflammation, whereas the inflammation in ulcerative colitis is confined to the mucosa. IBD can be initiated by an environmental trigger in a genetically susceptible individual altering immune homeostasis in the intestinal mucosa, often resulting in compromised intestinal function.5
Understanding the genetic and environmental interaction in the intestine is critical in uncovering the basis for IBD onset, progression, and remission.5 6 per se . In fact, the majority of this genetic contribution to disease is thought to come from expression quantitative trait loci (eQTL) or in other words, genomic loci that contribute to variation in the expression levels of mRNAs. An eQTL is the genetic polymorphism whose genotype is significantly associated with an mRNA's expression level, and that genetic polymorphism that in most cases is a SNP is in turn called an expression SNP or eSNP. An important distinction is made as to the location of the eQTL relative to the physical location of the gene whose expression level it is associated to. If the eQTL is close to the gene (which codes the mRNA), it is defined as a cis-eQTL. In contrast, if the eQTL is distant (typically >500 kb away or on another chromosome altogether) from the gene, it is defined as a trans-eQTL. In the current paper, we also applied this definition. A cis-eQTL is usually stronger (of larger genetic effect size) than trans-eQTLs. Also, the number of tests performed in discovering cis-eQTLs is much less than for trans-eQTLs because given the nature of the definition we only search the chromosome region approximate to a gene to discover cis-eQTLs but search the entire genome for trans-eQTLs. The heavier multiple testing burden means studies inherently have lower statistical power in detecting trans-eQTL, and in fact, eQTL studies routinely define significantly more cis-eQTLs than trans-eQTLs.7, 8, 9
The outcome of any GWA study of a case and control design is a statistical association between a region of the genome (of a certain size) and the disease of interest. This means, GWAS do not reveal the tissue, organ or cell type where the associated SNPs are functional or how they may modify disease risk. One way to capture this information is via integration with eQTL analysis. EQTL is a powerful approach that empirically links the GWAS SNPs to pathologically relevant genes expressed in a tissue of interest. In this study the blood and intestine were studied given the importance of these tissues in IBD disease biology. By using gene expression as a phenotype and examining how DNA polymorphisms contribute to both gene expression and disease phenotype, true causal relationships can be dissected.10, 11, 12 13 9, 12, 14
METHODS Dataset. Blood and biopsy data were collected at baseline from anti-TNF resistant Crohn's disease (CD) patients enrolled in the Ustekinumab trial previously described.15 15
Total RNA including micro ribonucleic acids was isolated with PAXgene Blood RNA MDx Kit plus customized reagent BM3 (Qiagen Inc., Valencia, CA, USA). RNA quantity and quality were determined with a LabChip GX (Caliper Life Sciences, Hopkinton, MA, USA). Gene expression were carried out using Affymetrix GeneChip HTHGU133+ (Affymetrix, Santa Clara, CA, USA Cat# 901262) according to the manufacturer's protocol with the exception that dimethyl sulphoxide replaced the Tetramethylammonium Chloride Solution in the hybridization buffer. Arrays were washed and stained on the Affymetrix GeneChip Array Station then scanned on an HTAPS scanner. Genotyping was performed at the Medical Genetics Institute as Cedars-Sinai Medical Center using Illumina OmniExpress chips (Human610-Quadv1 Chips; Illumina, San Diego, CA, USA). Genotypes were determined based on clustering of the raw intensity data for the two dyes using Illumina BeadStudio software. Six samples performed in duplicate yielded >99% concordance. In total, 733,120 SNPs were successfully genotyped.
Data analysis workflow. In Figure 1a and Figure 1b we report a schematic of the data analysis workflow and subsequent eQTL results mining, respectively. In brief, after quality control filtering of the data, we collected blood and intestine biopsies from 152 anti-TNF-resistant IBD patients. Gene expression was quantified by microarray on all samples, whereas blood samples were used for genotyping. Genotype and gene expression data, separately for blood and intestine, were then used for eQTL mapping. One the two eQTL sets were computed, we applied several data mining techniques to facilitate interpretation of the results in the context of disease including: testing for whether eQTLs (blood and intestine) were enriched for disease GWAS loci, overlap with known CD and IBD risk loci, Ingenuity Pathway Analysis and cell type enrichment analysis. Details about data QC, eQTL mapping and downstream mining of results are reported in the Supplementary Methods section.
Figure 1:
Data analysis workflow. (a ) Molecular data collection and mapping of expression Quantitative Trait Loci (eQTLs). (b ) Downstream eQTLs results mining.
RESULTS Discovery of new eQTLs in blood and intestine. Discovery of eQTLs were based on genetic and microarray gene expression profiles collected from a cohort of anti-TNF-resistant CD patients.15 Table 1 ).
Table 1: Summary of 10% false discovery rate expressional quantitative trait loci statistics
Results from the two tissues showed a high degree of overlap, with >60% of the intestine genes also reaching significance in whole blood (Figure 2 ). The lower number of eQTLs discovered in intestine as compared with the blood was mainly due to the smaller sample size of unique individuals in the intestine dataset (Table 1 ). The newly mapped gene expression controlling loci (eSNPs henceforth) are enriched in the regulomeDB database16 Table 1 ), indicating their concentration in regulatory regions and further supporting their putative functional relevance.
Figure 2:
Overlap between SNP-controlled genes in blood and intestine. A large degree of overlap was observed between discovered SNP-controlled genes from blood and intestine tissues.
Within the 149 analyzed blood samples, we quantified the eQTLs effects sizes as quantified the fraction of gene expression variability explained by the eSNP genotype (termed as R 2 ). For the 203 intestine biopsy samples it was not obvious how to quantify R 2 owing to repeated measures from only 54 independent individuals, and we thus omitted them from this specific analysis. Several eQTLs showed a large R 2 . For example, ERAP2 expression (Endoplasmic Reticulum Aminopeptidase 2, which has been previously reported in multiple studies6, 17 R 2 was distributed between 8.5 and 74.5% within blood cis-eQTLs, and between 37.6 and 69.7% within blood trans-eQTLs, indicating the powerful effect of these eSNPs on gene expression.
One of the primary utilities of eQTLs is to assign functional relevance to findings from GWAS. We employed different data mining strategies to embed our eQTL results in the context of human diseases and traits in general, and of CD in particular (see Figure 1b ): enrichment for disease-associated loci; remapping of eQTLs within known CD risk loci; Ingenuity Pathway Analysis for biological pathway insights; and finally cell type enrichment analysis of remapped genes controlled by CD risk SNPs. The results of these analyses are reported in the following sections.
Blood and intestine eQTLs are enriched in disease risk loci. To embed our eQTL results in the context of human disease we first surveyed the SNPs in the National Human Genome Research Institute (NHGRI) GWAS catalog18, 19, 20 P value thresholds based on the strict guidelines of replication. We detected significant enrichment for the eQTLs discovered in this study (P value=6.2e-203; OR=3.25), indicating that GWAS hits (as a whole regardless of the disease/trait) were more likely to be eQTLs than non-risk associated loci: 8.8% of the NHGRI catalog loci were significant eQTLs in either blood or intestine in our dataset, compared with 2.9% for the rest of the tested genome.
Second, we focused on diseases/traits with at least 90 GWAS association loci documented in NHGRI GWAS catalog (Figure 3a and Supplementary Table S4). The rationale of limiting the analysis within diseases/traits with many GWAS hits (i.e., ≥90) is to ensure sufficient statistical power in enrichment test. The tests showed that IBD GWAS hits were highly enriched for both intestine and blood (Figure 3a ). In other words, a significant fraction of IBD risk loci were influencing gene expression (i.e., as eQTLs) in blood and intestine tissue. In total, 57 out of 201 IBD risk loci were supported by eQTLs in at least one tissue. As already observed for the entire set of mapped eQTLs, many of these (14 loci) were shared between blood and intestine. In all, 40 IBD risk loci were supported by blood-specific eQTLs, and three loci (rs11741861, associated with ZNF300P1 gene expression; rs12677663, associated with SBSPON gene expression; and rs1728918, associated with 240330_at probeset expression level) were intestine-only eQTLs (Figure 3 and Supplementary Table S4). Furthermore, there was a high ranking of eQTL enrichment in IBD-associated disease traits indicating that blood and intestine are highly relevant tissues for this disease. Some of these eQTL enriched trait categories included growth and development (e.g., metabolite levels, bone mineral density, and height) and inflammatory diseases (e.g. IBD and type-1-diabetes), indicating the role of these two tissue types in these diseases. In contrast, psychiatric and neurological disease GWAS loci were under-present among blood and intestine eQTLs, indicating that these tissues are less relevant to these conditions. It should be noted that the background of enrichment test for each disease/trait (Figure 3a ) was the NHGRI catalog SNPs as a whole. As GWAS hits are enriched among eQTLs (regardless of disease type), the individual disease/trait enrichment test and point estimations (i.e., OR) tend to be conservative (Figure 3a and Supplementary Table S4). But the ranking of the disease traits, which is the main purpose of the analysis, was not affected.
Figure 3:
eQTLs enrichments for disease-associated loci. (a ) We tested the overlap between known and validated disease/trait-associated loci (from the NHGRI GWAS catalog) and expression Quantitative Trait loci discovered from the blood and intestine of anti-TNF-resistant patients. Disease traits were sorted according to odds ratio of enrichment for disease/trait-associated loci within intestine eQTLs. In parentheses, we present the number of overlaps in the following order: number of disease/trait loci overlapping with intestine-specific eQTLs; number of disease/trait loci overlapping with eQTLs shared by blood and intestine; number of disease/trait loci overlapping with blood-specific eQTLs; total number of loci for the considered disease/trait. The heatmap color indicates the overlap log-odds ratio, with red, white, and blue indicating overlap above, in line with, and below what would be expected by random chance. (b ) Intersection between IBD-associated loci (from a genome-wide analysis of the WTCCC dataset) and blood/intestine eQTLs. Highlighted some shared (green circles), blood-specific (red circle), and intestine-specific (blue circle) eQTL peaks.
Importantly, known IBD GWAS loci were shown as tissue-specific or shared eQTLs (Figure 3b ). For example, SLC22A5 (Solute Carrier Family 22 Member 5) loci act as eQTLs in both blood and intestine, and the genetic control of gene expression follow similar pattern in blood and various sections of intestine (Supplementary Table S4 and Figure 3a ). NOD2 locus was a blood-specific eQTL, where the rs17221417-G allele was associated with lower expression level in blood (Figure 3b ), whereas the rs17221417-NOD2 association was inconsistent among the intestine sections. ZNF300P1 locus was an intestine-specific eQTL, and the eQTL pattern was highly consistent in intestine sections (Figure 3c ), but not in blood.
Supplementary Table S2 presents the genome-wide eQTLs discovery, which is subject to heavy multiple testing penalty and consequently relatively low statistical power. We conducted a secondary eQTL discovery only focus on the known IBD risk loci. By these means, we limited the number of tests and therefore achieved higher statistical power (than genome-wide search). In total, we found that 110 unique genes were influenced by IBD risk loci (Table 2 ), and provided empirical evidence of directionality and magnitude of modulation through which IBD risk loci modify disease susceptibility. The GWA studies typically propose genes affected by association loci based on physical location and proximity on the genome. In many cases, several genes are close to the association peak and it is difficult to identify the gene(s) that mediate genetic effect. For example, the Chr17q12 locus showed strong association with IBD (lead SNP rs12946510, P value=4.00E-38), and several genes (IKZF3, ZPBP2, GSDMB, ORMDL3, GSDMA, etc.) were proposed as genes affected by this variant in this locus.6 Table 2 ). In total, 10.5% of the GSDMB gene expression variance was explained by this single eSNP. Furthermore GSDMB was linked to other inflammatory diseases (e.g. asthma) by GWAS,21 9 22 P value=6E-13) in GWAS,6 P value=2.95E-38 in blood and P value=1.50E-13 in intestine) and LNPEP (eQTL P value=3.2E-16 in blood but not significant in intestine) through cis-eQTLs (Table 2 ). ERAP2 and LNPEP are both aminopeptidases, which are involved in antigen presentation. ERAP2, a highly conserved, interferon γ inducible gene,23
Table 2: Intersection of blood and intestine eQTLs with GWAS hits
Table 2: Continued
Table 2: Continued
Table 2: Continued
Table 2: Continued
Table 2: Continued
Pathway enrichment analysis of genes controlled by CD risk loci in blood and intestine. Whereas the NHGRI GWAS catalog and most publications only report the top signals as in those a passing genome-wide significant threshold of 5e-8, most of the existing GWAS only have moderate statistical power, therefore many true disease-associating SNPs may only show a suggestive P value (e.g., P value<1e-3). Thus, it is generally considered that eQTLs can guide identification of susceptibility genes and active pathways by utilizing a full dataset rather than just the published top hits of GWAS signals. In this paper, we leveraged the full results of large IBD GWAS,6 P value ≤0.001 (Supplementary Information), and termed these as “IBD SNPs”. Importantly, through integrating these “IBD SNPs” with our newly discovered intestine and blood eQTLs (10% false discovery rate) we obtained a list of 96 and 335 genes, respectively, that were controlled by these genetic loci. These blood and intestine genes were subsequently submitted for biological pathway analysis using Ingenuity Pathway Analysis to determine their enrichment in canonical pathways and predicted upstream regulators.
Genes influenced by IBD GWAS SNPs in blood (through eQTLs) were enriched for several canonical pathways: T-cell signaling, antigen presentation, JAK/STAT signaling, interferon response, and pathways related to inositol metabolism (Figure 4a ). Although the pathways enriched in blood and intestine eSNPs were highly overlapping (Figure 4b ), there were also a number of pathways that showed tissue-specific enrichment. Enrichment in common pathways between blood and intestine are driven by shared eSNPs in human leukocyte antigen and JAK/STAT signaling genes, whereas pathways uniquely enriched in blood eSNPs included protein ubiquitination, Systemic Lupus Erythematosis signaling and PPARa/RXRa activation (Supplementary Tables S6 and S7).
Figure 4:
Ingenuity pathway analysis of genes controlled by IBD-associated loci. (a ) Heatmap displaying clustering of canonical pathways enriched genes (enrichment P value <0.01) whose blood expression is controlled by an IBD-associated locus (‘eGenes’ henceforth); pathways on y axis and genes driving enrichment on the x axis. Dark blue squares indicate a gene is present in the corresponding pathway listed to the right. (b ) Venn diagram of the intersection of Ingenuity pathways enriched in blood and intestine (refer to Supplementary Table S6 and Supplementary Table S7).
Compared with blood eSNPs, intestine eSNPs gave rise to fewer enriched pathways with less coverage (Supplementary Table S6 vs. Supplementary Table S7), therefore we have focused on the blood eSNPs for upstream regulator analysis. In this analysis Ingenuity Pathway Analysis software predicts for any given gene list, the potential upstream regulators or transcription factors of these genes, based on their databases containing literature supported downstream targets of genes Because our submitted gene list did not include information on changes in gene expression, the predicted upstream regulators do not have an associated likelihood of being activated or inhibited. For blood eSNPs, the top predicted upstream regulators included IFNγ, lipopolysaccharide, and genes involved in antigen presentation and T-cell signaling (Supplementary Table S11-A). Interestingly, the top predicted upstream master transcription factors included the forkhead (Fox) family members FOXD1 and FOXJ1, both of which have been predicted to have roles in inhibiting inflammation through negative regulation of NFAT (nuclear factor of activated T cells) and NF-kB (nuclear factor kappa-light-chain-enhancer of activated B cells) gene pathways (Supplementary Table S11-B).
Cell type-specific enrichment of genes influenced by eQTLs. Given that blood and intestinal mRNA expression analysis is the summation of expression of mRNA's from potentially multiple cell types within a particular tissue type, we attempted to further refine our eQTL analysis by determining the cell types most likely expressing the genes associated with IBD SNPs. To do this we took advantage of the (GEB) Gene Expression Barcode database, which provides reliable absolute measures of expression for most annotated genes for over 100 human tissue types, including diseased tissue. To determine cell types associated with genes with eSNPs we retrieved the GEB 3.0 dataset, which assigns each probeset targeting a gene, to the tissues and cell types where that probeset is expressed.24 P value<1e-3 to their corresponding cell types and gene labels. The summary of all eSNP probesets according to GEB classification and cell type-specificity is shown in Table 3 . Blood eSNP probesets showed enrichment in 24 cell types (Supplementary Methods), whereas intestine eSNP probesets did not show significant enrichment (e-value<0.05) in any cell type, possibility owing to the smaller number of intestine eSNPs than blood and consequently lower statistical power. Figure 5 shows the mapping of blood eSNP genes to the 24 significantly enriched cell types. Blood eSNP genes showed strongest enrichment in T cells, neutrophils, monocytes, and natural killer cells (Figure 5 ). Of note, the enriched T-cell subsets are specific to naive T cells, regulatory T cells and CD161+CD8 T cells. Genetic perturbation of pathways associated with regulatory T-cell development or cytokines produced by regulatory T cells are strongly implicated in IBD pathogenesis in murine models of colitis, monogenic very early-onset IBD and GWAS loci identified in adult-onset IBD patients25, 26 27, 28
Table 3: Mapping of eSNP probesets to gene expression barcodes cell types
Figure 5:
Cell type enrichment of blood eSNPs. Heatmap displaying clustering of cell types enriched in blood eSNPs (enrichment P value <0.05); cell types on y axis and genes driving enrichment on the x axis. Dark red squares indicate a gene is present in the corresponding pathway listed to the right.
DISCUSSION In this paper, we report two key results.1 2
Both sample availability and tissue relevance have been taken into account when designing this eQTL study. Although a much smaller number of subjects was available for the study of the intestine limiting power to detect association, the intestine-derived eQTLs, location of disease manifestation, proved crucial to orient the disease role of the GWAS hits, which were otherwise unsupported by the whole blood data. Our results indicate that the analysis of tissues more directly involved in a disease (e.g., the intestine of Crohn's patients) can highlight pathways that would not be otherwise detected in different tissues, even if bigger sample sizes were available. For instance, in Supplementary Tables S6 and S7 we show pathways that are enriched in either blood or intestine, but not in both. This can possibly be due to that disease mechanisms are not active in all tissues.
In the current study, we utilized blood and intestine tissue samples15 29, 30, 31, 32 32 Figure 3a ), however we also replicated and discovered a large number of blood- and intestine-specific eQTLs underlying disease GWAS hits, pointing to tissue-specific roles of an IBD genetic variant. We also observe a few eQTLs we discovered in this study overlap with previously reported eQTLs by Kabakchiev et al. 17 et al. 33
More specifically, we compared our newly discovered eQTLs with previously published eQTLs from a number of tissues and from subjects of different disease conditions (Supplementary Table S10). Besides the already highlighted high overlap between intestine and whole-blood eQTLs of the present study, we observed a particularly high overlap with previously published ileum and rectum eQTLs (odds ratios=3.09 and 2.93, respectively) obtained from independent populations composed of UC17 17 33
It is important to highlight that our study cohort differs from a generic cohort of Crohn patients in that they were all resistant to anti-TNF therapy. This might have impacted our findings in at least two ways: (1) the genetic background of anti-TNF resistant CD patients might be slightly different from that of anti-TNF responders; and (2) exposure to anti-TNF therapy before the collection of blood and intestine samples, if unsuccessful in treating CD itself, might well have impacted the microbiome of the patients, thus revealing expression patterns not otherwise found in a more general patients' population.
Besides observing statistically significant overlap of eQTLs discoveries from different tissues from a same set of subjects (i.e., between intestine and blood eQTLs in the present study) and between subjects of different disease conditions (e.g., Supplementary Table S10), we also observed some degree of heterogeneity across different intestine sections within our cohort. Even though our small sample size does not allow us to report statistically significant differences across intestine sections in a genome-wide survey, we can report section-specific association patterns when we zoom on specific loci of interest. One such example is the IBD risk locus NOD2 (rs17221417, Figure 6b ), for which terminal ileum and rectum display the same pattern of expression observed in whole blood, whereas ascending, transverse, descending, and sigmoid colon display a rather unclear, if opposite, direction of this same effect. This inconsistent association pattern of the risk locus with NOD2 gene expression allowed as to report a significant eQTL from the blood only, but not from the pooled intestine data analysis.
Figure 6:
Tissue specificity of IBD-associated eSNPs. Some example of eSNPs that were previously reported to be associated with IBD, with different degree of tissue specificity in their gene expression signal. Genotype on the horizontal axis, mean-centered expression levels on the vertical axis. (a ) Polymorphism rs11741861 shows a significant association with ZNF300P1 mRNA levels in all the intestine sections, but not in the blood; (b ) polymorphism rs17221417 shows different effects on NOD2 in different tissues; (c ) polymorphism rs2188962 controls SLC22A5 expression uniformly and with a similar effect across different tissues.
As in complex diseases, multiple pathways and gene networks are altered in IBD. We identified many genes relevant to IBD in the analysis of intersection between eQTLs and published association signals from GWAS studies focusing on IBD, UC and Crohn's (Table 2 ). The large number of biologically plausible findings indicate the eQTLs as powerful tools to interpret and dissect GWAS signals. In the vast majority of the genes listed in Table 2 , the direction of transcription regulation (i.e., up- or downregulation) caused by GWAS SNPs on the genes, agree with our current understanding of IBD molecular etiology. To help the reader interpret our observations we have organized the genes influenced by IBD GWAS hits (mediated by blood and intestine eQTLs) by functional categories and discussed their potential relevance below. These types of insights are what we hope gives momentum to the task of unraveling the underlying causes of IBD and then how one could potentially intervene therapeutically.
Epithelial barrier. Increased levels of ADCY3 (Adenylate Cyclase 3) in intestine and blood eQTLs34 35 TSPAN14 (tetraspanin 14), whose gene expression is controlled by an IBD GWAS SNP (rs6586030) in both blood and intestine (Table 2 ). While also expressed in immune cells, TSPAN14 has an intestinal epithelial role as a positive regulator of ligand induced ADAM10-mediated Notch1 activation. ADAM10 is a sheddase expressed on the surface of intestinal epithelial cells and is important for intestinal epithelial cell lineage specification and crypt cell columnar cell maintenance. This eQTL, resulting in a decrease in transcript in IBD, could contribute to epithelial barrier and restitution defects.36, 37, 38
Metabolism. In accordance with the analysis of the highest trait enrichment in IBD for NHGRI GWAS, we identified several eQTLs involved in metabolism. We reproduced, from previous reports, (Repnik and Potocnik, 2011), a significant cis-eQTL for the solute carrier family 22 member 5 (SLC22A5) in both the blood and intestine in our Crohn's cohort. The T allele, which is associated with increased risk of CD/IBD was found to associate with decreased expression of SLC22A5. This decrease in expression in the intestine was also seen when each of the regions sampled were calculated separately (see Figure 6c ). Similar observations were seen in a Slovenian cohort of refractory CD patients whereby lower expression of SLC22A5 gene was associated with disease-susceptible genotypes for two IBD disease-associated SNPs. SLC22A5 encodes a polyspecific organic cation transporter and has a high affinity for carnitine.39 40 40, 41, 42, 43
Monocytes/mcrophages. Several blood and intestine eQTLs have been identified as having specific roles in monocyte and macrophage function such as differentiation, polarization, and migration. LST1 is upregulated by Crohns GWAS risk alleles and has been reported as being a pro inflammatory marker expressed in the colon of IBD patients where its expression has a role in intestinal epithelial cells, monocyte differentiation, and nanotube formation.44, 45 Table 2 ), suggesting different behavior of IRF5 splicing isoforms (Supplementary Table S9). IRF5 is a marker for M1 macrophages and leads to production of IL-12 p40, IL-12 p35, and IL-23p19 while repressing IL-10. These macrophages create an environment conducive for a potent T- helper type-1 T(H)1-T(H)17 response,46
Several identified GWAS enriched eQTLs are involved in migration. AIF1 is upregulated by Crohn GWAS risk allele, and has been shown to increase chemokine expression in monocytes in addition to having a putative role in T cells47 Table 2 ), is involved in cytoskeletal re-arrangement and macrophage migration for integrin induced responses. Other eQTLs are less studied, such as TM9SF4, which is upregulated by IBD GWAS risk allele and has a conserved immune function in adhesion and phagocytosis in drosophila.48
Eosinophils. GSDMB and ORMDL3 are controlled by both blood and intestine eQTLs, and are also associated with UC and Crohns in GWAS studies. This gene locus has been linked to alterations in eosinophillic number and hyperinflammation in asthma.49, 50 Table 2 ). ITLN1 is also a serum biomarker for disease activity with lower levels found in IBD patients.51
Interferon response. As seen through the enrichment in the JAK STAT pathway of eQTLs of this TNF refractory population, genetic susceptibility in cytokine signaling can lead to dysregulated immune homeostasis in the intestine. EMSY (c11orf30), downregulated by IBD GWAS risk allele, is a transcriptional repressor in the PI3K/AKT pathway of the interferon response. Other interferon-responsive eQTLs are IRF1, whose upregulation has also been reported in IBD52 53
NF-κB. Both the blood-specific eQTLs pathway enrichment and predicted regulators involve ubiquitination and NF-κB. CUL2 is part of a ubiquitin ligase complex involved in modulating the NF-KB pathway through proteosomal degradation of sub units.54, 55 56, 57 58, 59 60
Autophagy, inflammasome, and endoplasmic reticulum stress. IPMK that is upregulated by a risk allele identified in IBD and Crohn GWAS is involved in induction of early response genes in serum through stabilization of mechanistic target of rapamycin while acting as a co-activator for CREB-binding protein/E1A-binding protein p300.61 62 63 64 65 66, 67, 68 69 70
Lymphocytes. Through our eQTL enrichment analysis in cellular specific gene expression signatures, genes related to T-cell function and T-regulatory cells, in particular, were among the top hits in blood in Crohn's. CD244 (2B4), a non MHC-binding inhibitory receptor on activated NK and CD8 memory cells interacts with CD48 and has a role in regulating tolerance.71 72 73 74,75 76 in vitro experiments, FOSL2 overexpression was shown to decrease the number of IL17A producing T cells, whereas FOSL2 deficiency enabled IFNγ production in Th17 and Th2 cell cultures when Th17 cells were subsequently exposed to Th1-skewing conditions.77, 78 79 80
In summary, we systematically characterized eQTL architecture in two IBD relevant tissues (blood and intestine), and identified genes influenced by IBD GWAS loci known to date. Our approach revealed a large number of genes driven by IBD loci which potentially mediate the disease etiology. For example, rs26528-C was strongly associated with IBD (P =1.00E-21),6 6 Table 2 ), and informed 109 out of 372 IBD GWAS SNPs documented in NHGRI catalog. Many of the genes mapped to these eQTLs have experimentally validated roles in cell types contributing to intestinal inflammation. These findings reflect our most up-to-date and comprehensive understanding of transcriptome regulation of IBD GWAS, and will serve as a valuable resource in further dissecting the molecular basis of this disorder.
Study Highlights Acknowledgments. We thank the patients and clinicians that participated in the in the CERTIFI study. This study was approved by IRB review, protocol number C0743T26.
CONFLICT OF INTEREST Guarantor of the article: Ke Hao, ScD.
Specific author contributions: Conception and design: Antonio F. Di Narzo, Joel Dudley, Judy Cho, Eric E. Schadt, Andrew Kasarskis, Mark Curran, Radu Dobrin, and Ke Hao; data generation, analysis, and interpretation: Antonio F. Di Narzo, Lauren A. Peters, Carmen Argmann, Aleksandar Stojmirovic, Jacqueline Perrigoue, Katherine Li, Shannon Telesco, Brian Kidd, Jennifer Walker, Radu Dobrin, and Ke Hao; drafting the manuscript for important intellectual content: Antonio F. Di Narzo, Lauren A. Peters, Carmen Argmann, Aleksandar Stojmirovic, Jacqueline Perrigoue, Katherine Li, Shannon Telesco, Radu Dobrin, and Ke Hao.
Financial support : This work is partially supported by Janssen R&D, LLC, including study design, data generation, analysis, interpretation of the results, and in the writing of the report.
Potential competing interests : None.
1. Guo C, D'Ippolito AM, Reddy TE. From prescription to transcription: genome sequence as drug target. Cell 2015;
162: 16–17.
2. Soccio RE, Chen ER, Rajapurkar SR
et al . Genetic variation determines PPARgamma function and anti-diabetic drug response
In Vivo . Cell 2015;
162: 33–44.
3. Mathews RJ, Robinson JI, Battellino M
et al . Evidence of NLRP3-inflammasome activation in rheumatoid arthritis (RA); genetic variants within the NLRP3-inflammasome complex in relation to susceptibility to RA and response to anti-TNF treatment. Ann Rheum Dis 2014;
73: 1202–1210.
4. Krausgruber T, Saliba D, Ryzhakov G
et al . IRF5 is required for late-phase TNF secretion by human dendritic cells. Blood 2010;
115: 4421–4430.
5. Neurath MF. Cytokines in inflammatory bowel disease. Nat Rev Immunol 2014;
14: 329–342.
6. Jostins L, Ripke S, Weersma RK
et al . Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease. Nature 2012;
491: 119–124.
7. Schadt EE, Molony C, Chudin E
et al . Mapping the genetic architecture of gene expression in human liver. PLoS Biol 2008;
6: e107.
8. Greenawalt DM, Dobrin R, Chudin E
et al . A survey of the genetics of stomach, liver, and adipose gene expression from a morbidly obese cohort. Genome Res 2011;
21: 1008–1016.
9. Hao K, Bosse Y, Nickle DC
et al . Lung eQTLs to help reveal the molecular underpinnings of asthma. PLoS Genet 2012;
8: e1003029.
10. Emilsson V, Thorleifsson G, Zhang B
et al . Genetics of gene expression and its effect on disease. Nature 2008;
452: 423–428.
11. Lamontagne M, Timens W, Hao K
et al . Genetic regulation of gene expression in the lung identifies CST3 and CD22 as potential causal genes for airflow obstruction. Thorax 2014;
69: 997–1004.
12. Battle A, Montgomery SB. Determining causality and consequence of expression quantitative trait loci. Hum Genet 2014;
133: 727–735.
13. Howie B, Fuchsberger C, Stephens M
et al . Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nat Genet 2012;
44: 955–959.
14. Hou L, Zhao H. A review of post-GWAS prioritization approaches. Front Genet 2013;
4: 280.
15. Sandborn WJ, Gasink C, Gao L-L
et al . Ustekinumab induction and maintenance therapy in refractory Crohn's disease. N Engl J Med 2012;
367: 1519–1528.
16. Xie D, Boyle AP, Wu L
et al . Dynamic trans-acting factor colocalization in human cells. Cell 2013;
155: 713–724.
17. Kabakchiev B, Silverberg MS. Expression quantitative trait loci analysis identifies associations between genotype and gene expression in human intestine. Gastroenterology 2013;
144: 1488–1496, 1496.e1-3.
18. Hindorff LA, Sethupathy P, Junkins HA
et al . Potential etiologic and functional implications of genome-wide association loci for human diseases and traits. Proc Natl Acad Sci USA 2009;
106: 9362–9367.
19. Laurie CC, Doheny KF, Mirel DB
et al . Quality control and quality assurance in genotypic data for genome-wide association studies. Genet Epidemiol 2010;
34: 591–602.
20. Welter D, MacArthur J, Morales J
et al . The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic Acids Res 2014;
42: D1001–D1006.
21. Moffatt MF, Gut IG, Demenais F
et al . A large-scale, consortium-based genomewide association study of asthma. N Engl J Med 2010;
363: 1211–1221.
22. Luo W, Obeidat M, Di Narzo AF
et al . Airway epithelial expression quantitative trait loci reveal genes underlying asthma and other airway diseases. Am J Respir Cell Mol Biol 2016;
54: 177–187.
23. Christodoulou K, Wiskin AE, Gibson J
et al . Next generation exome sequencing of paediatric inflammatory bowel disease patients identifies rare and novel variants in candidate genes. Gut 2013;
62: 977–984.
24. McCall MN, Jaffee HA, Zelisko SJ
et al . The Gene Expression Barcode 3.0: improved data processing and mining tools. Nucleic Acids Res 2014;
42: D938–D943.
25. Mayne CG, Williams CB. Induced and natural regulatory T cells in the development of inflammatory bowel disease. Inflamm Bowel Dis 2013;
19: 1772–1788.
26. Uhlig HH. Monogenic diseases associated with intestinal inflammation: implications for the understanding of inflammatory bowel disease. Gut 2013;
62: 1795–1805.
27. Le Bourhis L, Mburu YK, Lantz O. MAIT cells, surveyors of a new class of antigen: development and functions. Curr Opin Immunol 2013;
25: 174–180.
28. Serriari NE, Eoche M, Lamotte L
et al . Innate mucosal-associated invariant T (MAIT) cells are activated in inflammatory bowel diseases. Clini Expe Immunol 2014;
176: 266–274.
29. Petretto E, Bottolo L, Langley SR
et al . New insights into the genetic control of gene expression using a Bayesian multi-tissue approach. PLoS Comput Biol 2010;
6: e1000737.
30. Di Narzo A, Cheng H, Lu J
et al . Meta-eQTL: a tool set for flexible eQTL meta-analysis. BMC Bioinformatics 2014;
15: 392.
31. Sul JH, Han B, Ye C
et al . Effectively identifying eQTLs from multiple tissues by combining mixed model and meta-analytic approaches. PLoS Genet 2013;
9: e1003491.
32. Flutre T, Wen X, Pritchard J
et al . A statistical framework for joint eQTL analysis in multiple tissues. PLoS Genet 2013;
9: e1003486.
33. Singh T, Levine AP, Smith PJ
et al . Characterization of expression quantitative trait loci in the human colon. Inflamm Bowel Dis 2015;
21: 251–256.
34. Hulur I, Gamazon ER, Skol AD
et al . Enrichment of inflammatory bowel disease and colorectal cancer risk variants in colon expression quantitative trait loci. BMC Genom 2015;
16: 138.
35. Garg P, Vijay-Kumar M, Wang L
et al . Matrix metalloproteinase-9-mediated tissue injury overrides the protective effect of matrix metalloproteinase-2 during colitis. AM J Physiol Gastrointest Liver Physiol 2009;
296: G175–G184.
36. Dornier E, Coumailleau F, Ottavi JF
et al . TspanC8 tetraspanins regulate ADAM10/Kuzbanian trafficking and promote Notch activation in flies and mammals. J Cell Biol 2012;
199: 481–496.
37. Tsai YH, VanDussen KL, Sawey ET
et al . ADAM10 regulates Notch function in intestinal stem cells of mice. Gastroenterology 2014;
147: 822–834.e13.
38. Hong SH, Goh SH, Lee SJ
et al . Upregulation of adenylate cyclase 3 (ADCY3) increases the tumorigenic potential of cells by activating the CREB pathway. Oncotarget 2013;
4: 1791–1803.
39. Longo N, Amat di San Filippo C, Pasquali M. Disorders of carnitine transport and the carnitine cycle. Am J Med Genet Part C: Semin Med Genet 2006;
142C: 77–85.
40. Odievre MH, Chretien D, Munnich A
et al . A novel mutation in the dihydrolipoamide dehydrogenase E3 subunit gene (DLD) resulting in an atypical form of alpha-ketoglutarate dehydrogenase deficiency. Hum Mutat 2005;
25: 323–324.
41. Quintana E, Pineda M, Font A
et al . Dihydrolipoamide dehydrogenase (DLD) deficiency in a Spanish patient with myopathic presentation due to a new mutation in the interface domain. J Inherit Metab Dis 2010;
33 (Suppl 3): S315–S319.
42. Uchida H, Yamamoto H, Kisaki Y
et al . D-lactic acidosis in short-bowel syndrome managed with antibiotics and probiotics. J Pediatr Surg 2004;
39: 634–636.
43. Stamatakos M, Douzinas E, Stefanaki C
et al . Ischemic colitis: surging waves of update. Tohoku J Exp Med 2009;
218: 83–92.
44. Heidemann J, Kebschull M, Tepasse PR
et al . Regulated expression of leukocyte-specific transcript (LST) 1 in human intestinal inflammation. Inflamm Res 2014;
63: 513–517.
45. Schiller C, Diakopoulos KN, Rohwedder I
et al . LST1 promotes the assembly of a molecular machinery responsible for tunneling nanotube formation. J Cell Sci 2013;
126: 767–777.
46. Krausgruber T, Blazek K, Smallie T
et al . IRF5 promotes inflammatory macrophage polarization and TH1-TH17 responses. Nat Immunol 2011;
12: 231–238.
47. Cano-Martinez D. P052 Allograft inflammatory factor-1 (AIF-1) stimulates Th1 differentiation of human T-cells and protects them from apoptosis. J Chron's Colitis 2015;9 (Suppl 1)S103.
48. Bergeret E, Perrin J, Williams M
et al . TM9SF4 is required for Drosophila cellular immunity via cell adhesion and phagocytosis. J Cell Sci 2008;
121: 3325–3334.
49. Kang MJ, Yu HS, Seo JH
et al . GSDMB/ORMDL3 variants contribute to asthma susceptibility and eosinophil-mediated bronchial hyperresponsiveness. Hum Immunol 2012;
73: 954–959.
50. Ha SG, Ge XN, Bahaie NS
et al . ORMDL3 promotes eosinophil trafficking and activation via regulation of integrins and CD48. Nat Commun 2013;
4: 2479.
51. Kerr SC, Carrington SD, Oscarson S
et al . Intelectin-1 is a prominent protein constituent of pathologic mucus associated with eosinophilic airway inflammation in asthma. Am J Respir Criti Care Med 2014;
189: 1005–1007.
52. Tang R, Yang G, Zhang S
et al . Opposite effects of interferon regulatory factor 1 and osteopontin on the apoptosis of epithelial cells induced by TNF-alpha in inflammatory bowel disease. Inflamm Bowel Dis 2014;
20: 1950–1961.
53. Gonsky R, Deem RL, Targan SR. Multiple activating and repressive cis-promoter regions regulate TNFSF15 expression in human primary mononuclear cells. Cytokine 2013;
63: 36–42.
54. Mahrour N, Redwine WB, Florens L
et al . Characterization of Cullin-box sequences that direct recruitment of Cul2-Rbx1 and Cul5-Rbx2 modules to Elongin BC-based ubiquitin ligases. J Biol Chem 2008;
283: 8005–8013.
55. Mao X, Gluck N, Li D
et al . GCN5 is a required cofactor for a ubiquitin ligase that targets NF-kappaB/RelA. Genes Dev 2009;
23: 849–861.
56. Pacaud R, Sery Q, Oliver L
et al . DNMT3L interacts with transcription factors to target DNMT3L/DNMT3B to specific DNA sequences: role of the DNMT3L/DNMT3B/p65-NFkappaB complex in the (de-)methylation of TRAF1. Biochimie 2014;
104: 36–49.
57. Qiao YQ, Shen J, Gu Y
et al . Gene expression of tumor necrosis factor receptor associated-factor (TRAF)-1 and TRAF-2 in inflammatory bowel disease. J Digest Dis 2013;
14: 244–250.
58. Fransen K, Visschedijk MC, van Sommeren S
et al . Analysis of SNPs with an effect on gene expression identifies UBE2L3 and BCL3 as potential new risk genes for Crohn's disease. Hum Mol Genet 2010;
19: 3482–3488.
59. Fu B, Li S, Wang L
et al . The ubiquitin conjugating enzyme UBE2L3 regulates TNFalpha-induced linear ubiquitination. Cell Res 2014;
24: 376–379.
60. Palmieri G, Bergamo P, Luini A
et al . Acylpeptide hydrolase inhibition as targeted strategy to induce proteasomal down-regulation. PloS One 2011;
6: e25888.
61. Kim E, Tyagi R, Lee JY
et al . Inositol polyphosphate multikinase is a coactivator for serum response factor-dependent induction of immediate early genes. Proc Natl Acad Sci USA 2013;
110: 19938–19943.
62. Jang CW, Chen CH, Chen CC
et al . TGF-beta induces apoptosis through Smad-mediated expression of DAP-kinase. Nature Cell Biol 2002;
4: 51–58.
63. Lei Y, Wen H, Ting JP. The NLR protein, NLRX1, and its partner, TUFM, reduce type I interferon, and enhance autophagy. Autophagy 2013;
9: 432–433.
64. Patergnani S, Marchi S, Rimessi A
et al . PRKCB/protein kinase C, beta and the mitochondrial axis as key regulators of autophagy. Autophagy 2013;
9: 1367–1385.
65. Wang LJ, Huang HY, Huang MP
et al . The microtubule-associated protein EB1 links AIM2 inflammasomes with autophagy-dependent secretion. J Biol Chem 2014;
289: 29322–29333.
66. Parkhouse R, Boyle JP, Mayle S
et al . Interaction between NOD2 and CARD9 involves the NOD2 NACHT and the linker region between the NOD2 CARDs and NACHT domain. FEBS Lett 2014;
588: 2830–2836.
67. Sokol H, Conway KL, Zhang M
et al . Card9 mediates intestinal epithelial cell restitution, T-helper 17 responses, and control of bacterial infection in mice. Gastroenterology 2013;
145: 591–601.e3.
68. Lees CW, Barrett JC, Parkes M
et al . New IBD genetics: common pathways with other diseases. Gut 2011;
60: 1739–1753.
69. Cooke J, Zhang H, Greger L
et al . Mucosal genome-wide methylation changes in inflammatory bowel disease. Inflamm Bowel Dis 2012;
18: 2128–2137.
70. Schram AW, Baas R, Jansen PW
et al . A dual role for SAGA-associated factor 29 (SGF29) in ER stress survival by coordination of both histone H3 acetylation and histone H3 lysine-4 trimethylation. PloS One 2013;
8: e70035.
71. McNerney ME, Lee KM, Kumar V. 2B4 (CD244) is a non-MHC binding receptor with multiple functions on natural killer cells and CD8+ T cells. Mol Immunol 2005;
42: 489–494.
72. Roncagalli R, Hauri S, Fiore F
et al . Quantitative proteomics analysis of signalosome dynamics in primary T cells identifies the surface receptor CD6 as a Lat adaptor-independent TCR signaling hub. Nat Immunol 2014;
15: 384–392.
73. Garcia Santana CA, Tung JW, Gulnik S. Human treg cells are characterized by low/negative CD6 expression. Cytometry A 2014;
85: 901–908.
74. Altin JA, Daley SR, Howitt J
et al . Ndfip1 mediates peripheral tolerance to self and exogenous antigen by inducing cell cycle exit in responding CD4+ T cells. Proc Natl Acad Sci USA 2014;
111: 2067–2074.
75. Beal AM, Ramos-Hernandez N, Riling CR
et al . TGF-beta induces the expression of the adaptor Ndfip1 to silence IL-4 production during iTreg cell differentiation. Nat Immunol 2012;
13: 77–85.
76. Xue L, Qiu H, Ma J
et al . ZNF300, a recently identified human transcription factor, activates the human IL-2Rbeta promoter through the overlapping ZNF300/EGR1 binding site. Cell Mol Biol Lett 2010;
15: 530–540.
77. Rani A, Greenlaw R, Runglall M
et al . FRA2 is a STAT5 target gene regulated by IL-2 in human CD4 T cells. PloS One 2014;
9: e90370.
78. Ciofani M, Madar A, Galan C
et al . A validated regulatory network for Th17 cell specification. Cell 2012;
151: 289–303.
79. Kennedy JM, Fodil N, Torre S
et al . CCDC88B is a novel regulator of maturation and effector functions of T cells during pathological inflammation. J Exp Med 2014;
211: 2519–2535.
80. Nasir A, Norton JD, Baou M
et al . ZFP36L1 negatively regulates plasmacytoid differentiation of BCL1 cells by targeting BLIMP1 mRNA. PloS One 2012;
7: e52187.
Supplementary Information accompanies this paper on the Clinical and Translational Gastroenterology website (http://www.nature.com/ctg
