ORIGINAL ARTICLE
Extraintestinal Manifestations and Complications in Inflammatory Bowel Disease: From Shared Genetics to Shared Biological Pathways Suzanne van Sommeren, MD,* ,† Marcel Janse, MD,† Juha Karjalainen, MSC,* Rudolf Fehrmann, MD, PhD,* Lude Franke, PhD,* Jingyuan Fu, PhD,* and Rinse K. Weersma, MD, PhD†
Background: The clinical presentation of the inflammatory bowel diseases (IBD) is extremely heterogenous and is characterized by various extraintestinal manifestations and complications (EIM). Increasing genetic insight for IBD and EIM shows multiple shared susceptibility loci. We hypothesize that, next to these overlapping genetic risk loci, distinct disease pathways are shared between IBD and EIM.
Methods: The overlapping genetic risk loci for IBD and its EIM were searched in literature. We assessed shared disease pathways by performing an extensive pathway analysis by protein–protein interaction and cotranscriptional analysis, using both publicly available and newly developed databases.
Results: Reliable genetic data were available for primary sclerosing cholangitis, ankylosing spondylitis, decreased bone mineral density, colorectal carcinoma, gallstones, kidney stones, and deep venous thrombosis. We found an extensive overlap in genetic risk loci, especially for IBD and primary sclerosing cholangitis and ankylosing spondylitis. We identified 370 protein–protein interactions, of which 108 are statistically specific. We identified 446 statistically specific cotranscribed gene pairs. The interactions are shown to cluster in specific biological pathways.
Conclusions: We show that the pathogenetic overlap between IBD and its EIM extends beyond shared risk genes to distinctive shared biological pathways. We define genetic background as a risk factor for IBD-EIM alongside known mechanisms such as malabsorption and medication. Clustering patients based on distinctive pathways may enable stratification of patients to predict development of EIM. (Inflamm Bowel Dis 2014;20:987–994) Key Words: extraintestinal manifestations, complications, inflammatory bowel disease, genetic risk, biological pathways
T
he inflammatory bowel diseases (IBD) are complex diseases encompassing multiple specific subphenotypes. Included in IBD’s disease spectrum are various extraintestinal manifestations and complications (together named EIM). In addition to having a high incidence in patients with IBD, the EIM can actually cause more morbidity than IBD itself. Frequently occurring EIM are immune-mediated and involve multiple organ systems like the
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.ibdjournal.org). Received for publication January 31, 2014; Accepted February 19, 2014. From the Departments of *Genetics, and †Gastroenterology and Hepatology, University Medical Centre Groningen, University of Groningen, Groningen, the Netherlands. R. K. Weersma is supported by a VIDI grant (016.136.308) from the Netherlands Organization for Scientific Research (NWO) and a MLDS grant (WO 11–72) of the Dutch Digestive Foundation. J. Fu is supported by a VENI grant (863.09.007). L. Franke received a Horizon Breakthrough grant from the Netherlands Genomics Initiative (93519031) and a VENI grant from NWO (ZonMW grant, 916.10.135). The research leading to these results has received funding from the European Community’s Health Seventh Framework Program (FP7/2007–2013) under grant agreement 259867. This study was financed in part by the SIA-raakPRO subsidy for project BioCOMP. Reprints: Rinse K. Weersma, MD, PhD, Department of Gastroenterology and Hepatology, University of Groningen, University Medical Centre Groningen, P.O. Box 30001, Groningen 9700 RB, the Netherlands (e-mail: [email protected]). Copyright © 2014 Crohn’s & Colitis Foundation of America, Inc. DOI 10.1097/MIB.0000000000000032 Published online 15 April 2014.
Inflamm Bowel Dis Volume 20, Number 6, June 2014
skin (erythema nodosum, pyoderma gangrenosum), the eyes (uveïtis, episcleritis), the liver (primary sclerosing cholangitis, PSC), or spine (ankylosing spondylitis, AS).1 The skin and eye diseases occur during active intestinal disease, whereas PSC and AS can also occur while IBD itself is in a quiescent state. Decreased bone mineral density (BMD), gallstones and kidney stones are considered to be mainly driven by the (metabolic) consequences of malabsorption and chronic use of steroids. Longstanding active disease predisposes to the development of colorectal carcinoma (CRC), necessitating frequent surveillance colonoscopies.2 Despite this knowledge, it remains unclear why the IBD phenotype includes so many EIM, in what way disease pathogenesis is overlapping between IBD and EIM and why certain patients develop a particular EIM and others do not. One possible answer to these questions is an overlap in genetic architecture; this would explain why these EIM co-occur so often with IBD and moreover variation in the genetic make-up in individuals could partly explain the variability in disease phenotype. It has been demonstrated before that many immune-mediated diseases share a genetic background, and because IBD and its EIM co-occur so frequently we hypothesize that they have a common genetic background.3 It has become clear that IBD is not caused by an abnormality in a single gene leading to a single uniform disease, but is a consequence of the perturbations of complex pathways www.ibdjournal.org |
987
van Sommeren et al
leading to multiple specific subphenotypes.1 Recently there has been tremendous progress in unraveling the genetic background of ulcerative colitis and Crohn’s disease, to date 163 independent genetic susceptibility loci have been identified. The identified single genetic risk factors have been shown to cooperate in disease relevant pathways.4 Similarly, genetic variants predisposing to many EIM have been established, of which some are shared with IBD. We hypothesize that, next to these overlapping genetic risk loci, disease pathways are shared between IBD and EIM. To understand how these genetic variants predispose to IBD-EIM, we performed a survey of overlapping genetic loci. For this we included the following EIM or complications of IBD: AS, PSC, decreased BMD, colorectal carcinoma, erythema nodosum, pyoderma gangrenosum, uveïtis, episcleritis, kidney stones, gallstones, and venous thrombosis. Next, we analyzed coregulation between all pairs of genes within IBD loci and EIM loci (with available and reliable genetic data), for protein–protein interactions (PPI) and cotranscription using large databases (including .80,000 human Affymetrix mRNA expression data sets). We show that the pathogenetic overlap between IBD and its EIM is partly driven by a shared genetic predisposition and extends beyond purely shared risk genes to distinct shared biological pathways.
MATERIALS AND METHODS Literature Search and Selection of Associated Loci Genetic association data for IBD was extracted from the most recent analyses.4 An extensive literature search was performed in Pubmed and the GWAS catalog (www.genome.gov/ gwastudies) till October 2011 to assess available data on genetic susceptibility of the EIM. For EIM, all performed GWAS and GWAS meta-analyses were included. We also included large candidate studies if limited GWAS data were available. From GWAS, we included single nucleotide polymorphisms (SNPs) with reported genome wide significance (in most studies defined as P , 5E-8). For candidate studies, we included SNPs with P , 1E-4. For PSC, we also included the GPBAR1 locus because this locus showed functional evidence.5 When there were only small candidate gene studies available or with conflicting results, we did not include the identified susceptibility loci for the pathway analyses. All well-established genetic loci in IBD and the EIM were included and genetic overlap was assessed as presented in Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456.
Mapping SNPs to Genes For the cotranscriptional and PPI analyses, all candidate genes in the loci were included. Therefore, we assessed which genes are located at associated loci in IBD and EIM. The diseaseassociated SNPs were linked to proximate candidate genes in linkage disequilibrium (LD) with them, using a previously described approach.6 We downloaded the recombination hotspot
988
| www.ibdjournal.org
Inflamm Bowel Dis Volume 20, Number 6, June 2014
and LD information from www.hapmap.org for CEU population (release 28) and genome build hg18. The information of gene positions was based on genome build hg18 and was downloaded from UCSC Genome Browser. For each of the associated SNPs, we first defined the disease locus as the region containing the SNPs with LD r2 . 0.5 to the associated SNP and then extended it to the nearest recombination hotspot. This region was further extended 100 kb on each side to include the potential regulatory regions for genes. If any transcript isoform of a gene overlaps with the defined disease locus, this gene was included as a candidate gene. Thus an ith IBD locus with n number of candidate genes was defined as IBDi {g1, g2, ., gn}; and a jth EIM locus with z number of candidate genes was defined as EIMj {ga, gb, ., gz}.
PPI Between IBD and EIM Loci We used the PPI database that was described by Rossin et al6 and Lage et al7 and extracted all the direct interactions between the candidate genes in IBD loci and the candidate genes in EIM loci, for example, between the given ith IBD loci IBDi {g1, g2, ., gn} and the given jth EIM loci EIMj {ga, gb, ., gz}. To assess the specificity of the interactions between loci, we further calculated P values empirically by testing interactions between the EIMj loci and the mimicked random IBDi locus with similar number of genes: if n # 10, the mimicked locus must contain the same n number of genes; if n . 10, the mimicked random locus could contain the number of genes within 10% variation (i.e., gene numbers within 0.9*n and 1.1*n). The random IBDi locus was mimicked 1000 times. We then scored, out of 1000 random loci, how many times we could observe at least 1 interaction between the EIMj locus and the random IBDi locus and calculated the empirical P value. The significance threshold was controlled at 0.05.
Cotranscriptional Interactions Between IBD and EIM Loci To assess the similarities in transcription between genes in IBD loci and EIM loci, we used a new gene cotranscriptional network based on gene expression (www.genenetwork.nl). First, to derive a regulatory model of the human transcriptome, we performed principal component analysis on microarray expression data on 3 species (homo sapiens, mus musculus, and rattus norvegicus) publicly available in Gene Expression Omnibus. We used data for samples hybridized to HG-U133 Plus 2.0, HG-U133A, MG-430 2.0, and RG-230 2.0 Affymetrix Genechip platforms and performed principal component analysis on the probe correlation matrix of each of the 4 platforms. We ended up with 777, 377, 677, and 375 robust principal components and eigenvectors, respectively; each component describing an underlying factor that regulates expression. We treated each of these components as being equally interesting: Although the first components from each of the 4 platforms explain most variation in expression, they overshadow other (subtle) relationships that also exist within these data, which have
Inflamm Bowel Dis Volume 20, Number 6, June 2014
Shared Biological Pathways in IBD and Its EIM
been captured by other components. Conventional coexpression analysis, which uses, e.g., Pearson’s correlations between the expression levels of pairs of genes and does not use principal component analysis, is likely to miss such subtle relationships. Here, we use a correlation measure that uses the 2200 components between pairs of genes and is able to identify such relationships. First, we mapped each probe set on each of the 4 platforms to a gene and averaged eigenvector coefficients over probe sets mapping to the same gene. Probes with no or unambiguous mapping (due to e.g., cross-hybridization) were excluded from further analysis. Second, we calculated for each pair of human genes, a Pearson product–moment correlation coefficient over the 2200 eigenvector coefficients of each platform that contained probe sets for the genes or, for the mouse and rat platforms, their most similar orthologs. Both the probe set and gene ortholog mapping information were downloaded from Ensembl Biomart (Ensembl release 65). The correlation coefficients were further converted to Z-scores to account for different numbers of available eigenvectors for pairs of genes because of missing orthologs. The Z-score for a pair of genes can be positive or negative and describes the similarity of their regulation. The genes (19,997 in total) form the nodes and the correlation coefficients forms the edges of the network. We used this cotranscriptional network to find significantly cotranscribed genomic regions implicated by SNPs associated with IBD or EIM. First, we determined regions of interest for each phenotype. For each SNP associated with a phenotype, we defined implicated genes by identifying furthest SNPs in both the 30 and 50 directions in LD with it (r2 . 0.5 based on CEU HapMap 2) and extending the region first to the nearest recombination hotspot and then an additional 100 kb in both directions. Each gene in the coregulation network overlapping with this region was considered implicated by the SNP and part of the locus. Overlapping loci were merged together. We then examined the potential cotranscription between the regions associated with IBD and those associated with each of the EIM by finding pairs of genes that show the strongest cotranscription in either direction in each pair of loci between the phenotypes. To ascertain whether genes in a locus were often cotranscribed with genes in other regions, we ran a permutation test with 1000 permutations randomly picking a region from the genome with a similar centimorgan range and a similar number of genes as the IBD locus and repeating the procedure with the random locus replacing the IBD locus. We then applied a 5% of false discovery rate to eliminate potential false-positive gene pairs.
RESULTS
Biological Interpretation of Interactions To identify pathways in which the interacting genes of IBD and the different EIM play roles, we used DAVID Bioinformatics database.8 DAVID uses several resources like GO terms, KEGG—the Kyoto encyclopedia of genes and genomes,9 and BioCarta (www.biocarta.com) to cluster genes together in pathways.
Genetic Overlap We first assessed genetic overlap between IBD and all EIM: AS, PSC, BMD, CRC, erythema nodosum, pyoderma gangrenosum, uveïtis, episcleritis, kidney stones, gall stones, and venous thrombosis. Most robust genetic data were available for AS, CRC, BMD, and PSC. For gallstones, kidney stones, and deep venous thrombosis, limited genetic data were available and only candidate gene studies were available, whereas for the skin EIM (erythema nodosum, pyoderma gangrenosum) and eye EIM (uveïtis, episcleritis), no genetic data were available (Table 1). Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456, shows all associated loci per EIM. The classical immune-driven diseases AS and PSC have the largest number of risk loci that are shared with IBD. In AS, 13 of 18 known risk loci are shared with IBD; for PSC this is 10 of 14. In CRC, 3 of 15 risk loci are shared with IBD and for BMD, this is 1 of 23. For deep venous thrombosis, 2 loci are identified, the traits gallstones and kidney stones both have 1 identified genetic risk locus. None of these loci are also associated with IBD. Figure 1 shows all the shared loci between any EIM and IBD per chromosome. Because the HLA and CARD9 loci are shared between IBD, AS, and PSC, there are 25 unique shared loci across all EIM. Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456 lists all established EIM loci.
PPI and Cotranscriptional Regulation We then searched for statistical significant PPIs and cotranscriptional interactions between IBD genes and EIM genes. From the PPI database, we identified 370 PPIs between IBD and EIM loci, excluding PPIs between shared IBD-EIM loci. The (extended) HLA locus is involved in 57 of 370 interacting loci. After 1000 permutations of testing PPIs between the tested EIM locus and a mimicked IBD locus 86/370 PPIs turned out to be significantly specific (Fig. 2). Interacting IBD-EIM loci can hold multiple PPIs between different genes in the loci. These 370 interacting loci hold
TABLE 1. Available Genetic Data for EIM and Complications of IBDs EIMs/Complications CC Decreased BMD AS PSC Gall stones Kidney stones Venous thrombosis Erythema nodosum Pyoderma gangrenosum Uveïtis Episcleritis
No. of Genetic Studies 910–18 919–27 428–31 432–35 136 137 138 0 0 0 0
www.ibdjournal.org |
989
van Sommeren et al
Inflamm Bowel Dis Volume 20, Number 6, June 2014
FIGURE 1. Chromosomes with risk loci for IBD and its EIM. Each dot represents the position of an associated locus on the chromosome. For shared risk loci, the candidate gene is depicted. Complete loci information can be found in Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456.
a total of 915 PPIs between IBD and EIM genes (441 of these PPIs involve a gene originating in the HLA locus). From the cotranscriptional network we created, we identified 10,890 significant cotranscribed genes between IBD and its EIM. After 1000 permutations, 446 turned out to be significantly specific (Fig., Supplemental Digital Content 2, http://links.lww.com/IBD/A457).
Pathway Annotation We then annotated the biological pathways in which the interacting genes of IBD and the different EIM play roles, by using the DAVID Bioinformatics database. To ensure the reliability of the observed pathways, an extensive literature search was performed. Because of the small amount of associated loci in gallstones, venous thrombosis and kidney stones, few interactions were found and no defined pathways were established. For AS, PSC,
990
| www.ibdjournal.org
CRC, and BMD, we find several interesting pathways where genes with interactions through PPI or cotranscription play roles. These are highlighted in (Fig. 3), and we discuss them in detail further on.
DISCUSSION We aimed to give an overview of the current genetic data on IBD and its EIM and provide additional downstream pathway analyses using publicly available and new locally developed databases. The data as shown here are first of all important to understand the pathogenetic mechanisms that lead to the co-occurrence of immune-mediated extraintestinal manifestations or the development of long-term complications. Second, in the future, clustering patients based on distinctive involved pathways may enable stratification of patients to predict development of EIM and investigate specific screening protocols.
Inflamm Bowel Dis Volume 20, Number 6, June 2014
Shared Biological Pathways in IBD and Its EIM
FIGURE 2. PPI between IBD loci and EIM loci. Each dot represents the position of a locus associated to IBD or one of the EIM on the chromosomes. Names of the candidate genes in the associated loci are presented on the outer side of the circle. The colored lines represent the PPIs that are significant after permutations between IBD genes and EIM genes; different colors are used for the different EIM. The gray lines represent PPIs that are not significant after permutations.
Colorectal Carcinoma The most prominent shared pathway between IBD and CRC is the regulation of the intrinsic epithelial barrier integrity. Central in this pathway is CDH1; encoding E-cadherin, which anchors cells
together in adherens junctions and is associated to both IBD and CRC.39 Other interacting genes influence epithelial barrier function by regulating the cellular actin cytoskeleton (RHPN2, FMN1, RHOA, ARPC2, CAPN10) or establishing attachment of the cytoskeleton to www.ibdjournal.org |
991
van Sommeren et al
Inflamm Bowel Dis Volume 20, Number 6, June 2014
FIGURE 3. Summary of shared biological pathways in IBD and its EIM. Each square represents an associated gene. An IBD gene with multiple interactions across the EIM is depicted multiple times. A blue line represents the most significant cotranscriptional interaction between 2 genes. PPIs significance after permutations are represented by black lines, nonsignificant PPIs by gray lines. Pathways are encircled, subpathways within a bigger pathway are depicted in gray circles.
the basement membrane (DAG1) or in the basement membrane (LAMB1, LAMA5).40 The TGF-b signaling pathway harbors many genes associated to CRC or IBD, illustrating the crucial regulatory role TGF-b plays in different relevant pathways including epithelial barrier function and the immune system.41 The regulatory functions of TGF-b are shared with Wnt signaling, which is an established pathway in CRC and has recently been associated with UC-related CRC.42
992
| www.ibdjournal.org
Bone Mineral Density Wnt signaling genes are abundantly present in BMDassociated loci (WNT16, CTNNB1, GPR177, LRP5) and have interactions with IBD genes like CDH1 and SMAD3. Wnt signaling has been shown to influence the RANK/RANKL/OPG (TNFSF11, TNFRSF11A, TNFRSF11B) pathway.43 TNFSF11 is associated with BMD and was previously identified as a CD
Inflamm Bowel Dis Volume 20, Number 6, June 2014
locus44; however, in the most recent IBD analyses,4 this locus could not be replicated. Therefore, the association of this locus with IBD remains questionable. TNFSF11 has interactions with its receptor TNFRSF11A and the inhibiting decoy receptor TNFRSF11B, which are both associated to BMD. This pathway is involved in both bone homeostasis as regulating T-cell— dendritic cell communications, dendritic cell survival, and lymph node organogenesis, crucial in IBD pathogenesis.45 We see many coregulation between genes in this pathway and genes that are involved in the adaptive immune system (e.g., PTPN22, CCR6). These findings highlight possibilities for e.g., drug repositioning for Denosumab (Amgen/GlaxoSmithKline), which targets TNFSF11 and is a marketed drug for the treatment of postmenopausal women at high risk of fracture with osteoporosis.46 The gene ITGB7, encoding integrin beta-7, which associates with alpha-4 to form integrin alpha-4/beta-7, resides in a BMD-associated locus and interacts with the IBD gene CDH1. Natalizumab (Biogen, Weston, MA), registered for use in Crohn’s disease, blocks homing of lymphocytes to vascular endothelial cells of the gastrointestinal tract through MADCAM1 on the endothelium and integrin alpha-4/beta-7 on lymphocytes.47 These findings highlight the importance of investigating pathways associated with the broader IBD-EIM phenotype.
Primary Sclerosing Cholangitis and Ankylosing Spondylitis Most striking in the analysis for PSC and AS is the large number of overlapping loci. Even if we focus on PSC and AS genes that are not shared with IBD, we find many interactions in pathways that are shared with IBD. PSC genes and their connected IBD genes play several roles in T-cell signaling, like T-cell apoptosis (UBASH3A, BCL2L11, FOXO1 and IRF8) and the JAK-STAT signaling pathway (SOCS1, JAK2, STAT3 and TYK2).48,49 For AS, the largest shared pathway is the T-cell apoptosis pathway, with many interactions with IBD genes. The AS-associated genes TAPBPL and NPEPPS function in the same process as the AS-IBD shared gene ERAP1, namely the process of protein antigen binding to the MHC-I molecules, which is likely to be crucial to both AS and IBD pathogenesis.50 Given the fast moving field of IBD genetics, we do not provide a definitive analysis. Very recently, the number of shared genes between IBD and other diseases has increased with the publication of several articles using a customized GWAS chip (Immunochip, Illumina, San Diego, CA) focussed on immunemediated diseases.4,51 At the time of this analyses, these data were not available but the number of shared pathways is expected to increase substantially. Large-scale analyses of completely phenotyped IBD cohorts (including reliable data on EIM) are necessary to further investigate whether patients with specific EIM are enriched for risk genes or risk pathways as presented here. In conclusion, we show that the pathogenetic overlap between IBD and its EIM or complications extends substantially beyond purely shared risk genes to shared biological pathways. We identified numerous statistically significant interactions
Shared Biological Pathways in IBD and Its EIM
clustering in several distinct biological pathways between IBD and the different EIMs. Hereby, we further highlight the genetic background as risk factor for IBD-EIM next to known mechanisms as malabsorption, chronic inflammation, or medication.
ACKNOWLEDGMENTS Author contributions: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma have contributed equally. Study design: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma; Data analysis and interpretation: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma; Manuscript preparation: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma.
REFERENCES 1. Podolsky DK. Inflammatory bowel disease. N Engl J Med. 2002;347: 417–429. 2. Rothfuss KS, Stange EF, Herrlinger KR. Extraintestinal manifestations and complications in inflammatory bowel diseases. World J Gastroenterol. 2006;12:4819–4831. 3. Zhernakova A, van Diemen CC, Wijmenga C. Detecting shared pathogenesis from the shared genetics of immune-related diseases. Nat Rev Genet. 2009;10:43–55. 4. 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. 5. Hof JR, Keitel V, Laerdahl JK, et al. Mutational characterization of the bile acid receptor TGR5 in primary sclerosing cholangitis. PLoS One. 2010;5:e12403. 6. Rossin EJ, Lage K, Raychaudhuri S, et al. Proteins encoded in genomic regions associated with immune-mediated disease physically interact and suggest underlying biology. PLoS Genet. 2011;7:e1001273. 7. Lage K, Karlberg EO, Størling ZM, et al. A human phenome-interactome network of protein complexes implicated in genetic disorders. Nat Biotechnol. 2007;25:309–316. 8. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009;4:44–57. 9. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27–30. 10. Houlston RS, Cheadle J, Dobbins SE, et al. Meta-analysis of three genome-wide association studies identifies susceptibility loci for colorectal cancer at 1q41, 3q26.2, 12q13.13 and 20q13.33. Nat Genet. 2010;42: 973–977. 11. Cui R, Okada Y, Jang SG, et al. Common variant in 6q26-q27 is associated with distal colon cancer in an Asian population. Gut. 2011;60: 799–805. 12. Tomlinson IP, Webb E, Carvajal-Carmona L, et al. A genome-wide association study identifies colorectal cancer susceptibility loci on chromosomes 10p14 and 8q23.3. Nat Genet. 2008;40:623–630. 13. Houlston RS, Webb E, Broderick P, et al. Meta-analysis of genome-wide association data identifies four new susceptibility loci for colorectal cancer. Nat Genet. 2008;40:1426–1435. 14. Tomlinson IP, Webb E, Carvajal-Carmona L, et al. A genome-wide association scan of tag SNPs identifies a susceptibility variant for colorectal cancer at 8q24.21. Nat Genet. 2007;39:984–988. 15. Zanke BW, Greenwood CM, Rangrei J, et al. Genome-wide association scan identifies a colorectal cancer susceptibility locus on chromosome 8q24. Nat Genet. 2007;39:989–994. 16. Tenesa A, Farrington SM, Prendergast JG, et al. Genome-wide association scan identifies a colorectal cancer susceptibility locus on 11q23 and replicates risk loci at 8q24 and 18q21. Nat Genet. 2008;40:631–637. 17. Jaeger E, Webb E, Howarth K, et al. Common genetic variants at the CRAC1 (HMPS) locus on chromosome 15q13.3 influence colorectal cancer risk. Nat Genet. 2008;40:26–28. www.ibdjournal.org |
993
van Sommeren et al
18. Broderick P, Carvajal-Carmona L, Pittman AM, et al. A genome-wide association study shows that common alleles of SMAD7 influence colorectal cancer risk. Nat Genet. 2007;39:1315–1317. 19. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, et al. Multiple genetic loci for bone mineral density and fractures. N Engl J Med. 2008;358: 2355–2365. 20. Rivadeneira F, Styrkarsdottir U, Estrada K, et al. Twenty bone-mineraldensity loci identified by large-scale meta-analysis of genome-wide association studies. Nat Genet. 2009;41:1199–1206. 21. Cho YS, Go MJ, Kim YJ, et al. A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing eight quantitative traits. Nat Genet. 2009;41:527–534. 22. Richards JB, Rivadeneira F, Inouye M, et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet. 2008;371:1505–1512. 23. Guo Y, Zhang LS, Yang TL, et al. IL21R and PTH may underlie variation of femoral neck bone mineral density as revealed by a genome-wide association study. J Bone Miner Res. 2010;25:1042–1048. 24. Timpson NJ, Tobias JH, Richards JB, et al. Common variants in the region around Osterix are associated with bone mineral density and growth in childhood. Hum Mol Genet. 2009;18:1510–1517. 25. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, et al. New sequence variants associated with bone mineral density. Nat Genet. 2009;41:15–17. 26. Xiong DH, Liu XG, Guo YF, et al. Genome-wide association and follow-up replication studies identified ADAMTS18 and TGFBR3 as bone mass candidate genes in different ethnic groups. Am J Hum Genet. 2009;84:388–398. 27. Kung AW, Xiao SM, Cherny S, et al. Association of JAG1 with bone mineral density and osteoporotic fractures: a genome-wide association study and follow-up replication studies. Am J Hum Genet. 2010;86:229–239. 28. Evans DM, Spencer CC, Pointon JJ, et al. Interaction between ERAP1 and HLA-B27 in ankylosing spondylitis implicates peptide handling in the mechanism for HLA-B27 in disease susceptibility. Nat Genet. 2011;43:761–767. 29. Australo-Anglo-American Spondyloarthritis Consortium (TASC), Reveille JD, Sims AM, et al. Genome-wide association study of ankylosing spondylitis identifies non-MHC susceptibility loci. Nat Genet. 2010; 42:123–127. 30. Danoy P, Pryce K, Hadler J, et al. Association of variants at 1q32 and STAT3 with ankylosing spondylitis suggests genetic overlap with Crohn’s disease. PLoS Genet. 2010;6:e1001195. 31. Wellcome Trust Case Control Consortium, Burton PR, Clayton DG, et al. Association scan of 14,500 nonsynonymous SNPs in four diseases identifies autoimmunity variants. Nat Genet. 2007;39:1329–1337. 32. Folseraas T, Melum E, Raush P, et al. Extended analysis of a genomewide association study in primary sclerosing cholangitis detects multiple novel risk loci. J Hepatol. 2012;57:366–375. 33. Janse M, Lamberts LE, Franke L, et al. Three ulcerative colitis susceptibility loci are associated with primary sclerosing cholangitis and indicate a role for IL2, REL, and CARD9. Hepatology. 2011;53:1977–1985.
994
| www.ibdjournal.org
Inflamm Bowel Dis Volume 20, Number 6, June 2014
34. Melum E, Franke A, Schramm C, et al. Genome-wide association analysis in primary sclerosing cholangitis identifies two non-HLA susceptibility loci. Nat Genet. 2011;43:17–19. 35. Karlsen TH, Franke A, Melum E, et al. Genome-wide association analysis in primary sclerosing cholangitis. Gastroenterology. 2010;138: 1102–1111. 36. Buch S, Schafmayer C, Völzke H, et al. A genome-wide association scan identifies the hepatic cholesterol transporter ABCG8 as a susceptibility factor for human gallstone disease. Nat Genet. 2007;39:995–999. 37. Thorleifsson G, Holm H, Edvardsson V, et al. Sequence variants in the CLDN14 gene associate with kidney stones and bone mineral density. Nat Genet. 2009;41:926–930. 38. Trégouët DA, Heath S, Saut N, et al. Common susceptibility alleles are unlikely to contribute as strongly as the FV and ABO loci to VTE risk: results from a GWAS approach. Blood. 2009;113:5298–5303. 39. Van Roy F, Berx G. The cell-cell adhesion molecule E-cadherin. Cell Mol Life Sci. 2008;65:3756–3788. 40. Turner JR. Intestinal mucosal barrier function in health and disease. Nat Rev Immunol. 2009;9:799–809. 41. Massagué J, Blain SW, Lo RS. TGFbeta signaling in growth control, cancer, and heritable disorders. Cell. 2000;103:295–309. 42. Van Dekken H, Wink JC, Vissers KJ, et al. Wnt pathway-related gene expression during malignant progression in ulcerative colitis. Acta Histochem. 2007;109:266–272. 43. Häusler KD, Horwood NJ, Chuman Y, et al. Secreted frizzled-related protein-1 inhibits RANKL-dependent osteoclast formation. J Bone Miner Res. 2004;19:1873–1881. 44. Franke A, McGovern DP, Barrett JC, et al. Genome-wide meta-analysis increases to 71 the number of confirmed Crohn’s disease susceptibility loci. Nat Genet. 2010;42:1118–1125. 45. Tilg H, Moschen AR, Kaser A, et al. Gut, inflammation and osteoporosis: basic and clinical concepts. Gut. 2008;57:684–694. 46. Sanseau P, Agarwal P, Barnes MR, et al. Use of genome-wide association studies for drug repositioning. Nat Biotechnol. 2012;30:317–320. 47. Ghosh S, Goldin E, Gordon FH, et al. Natalizumab for active Crohn’s disease. N Engl J Med. 2003;348:24–32. 48. O’Shea JJ, Gadina M, Schreiber RD. Cytokine signaling in 2002: new surprises in the Jak/Stat pathway. Cell. 2002;109:S121–S131. 49. Hughes PD, Belz GT, Fortner KA, et al. Apoptosis regulators Fas and Bim cooperate in shutdown of chronic immune responses and prevention of autoimmunity. Immunity. 2008;28:197–205. 50. Chang SC, Momburg F, Bhutani N, et al. The ER aminopeptidase ERAP1, trims precursors to length of MHC class I peptides by a “molecular ruler” mechanism. Proc Natl Acad Sci U S A. 2005;102:17107– 17112. 51. Liu JZ, Hov JR, Folseraas T, et al. Dense genotyping of immune-related disease regions identifies nine new risk loci for primary sclerosing cholangitis. Nat Genet. 2013;45:670–675.
Extraintestinal Manifestations and Complications in Inflammatory Bowel Disease: From Shared Genetics to Shared Biological Pathways Suzanne van Sommeren, MD,* ,† Marcel Janse, MD,† Juha Karjalainen, MSC,* Rudolf Fehrmann, MD, PhD,* Lude Franke, PhD,* Jingyuan Fu, PhD,* and Rinse K. Weersma, MD, PhD†
Background: The clinical presentation of the inflammatory bowel diseases (IBD) is extremely heterogenous and is characterized by various extraintestinal manifestations and complications (EIM). Increasing genetic insight for IBD and EIM shows multiple shared susceptibility loci. We hypothesize that, next to these overlapping genetic risk loci, distinct disease pathways are shared between IBD and EIM.
Methods: The overlapping genetic risk loci for IBD and its EIM were searched in literature. We assessed shared disease pathways by performing an extensive pathway analysis by protein–protein interaction and cotranscriptional analysis, using both publicly available and newly developed databases.
Results: Reliable genetic data were available for primary sclerosing cholangitis, ankylosing spondylitis, decreased bone mineral density, colorectal carcinoma, gallstones, kidney stones, and deep venous thrombosis. We found an extensive overlap in genetic risk loci, especially for IBD and primary sclerosing cholangitis and ankylosing spondylitis. We identified 370 protein–protein interactions, of which 108 are statistically specific. We identified 446 statistically specific cotranscribed gene pairs. The interactions are shown to cluster in specific biological pathways.
Conclusions: We show that the pathogenetic overlap between IBD and its EIM extends beyond shared risk genes to distinctive shared biological pathways. We define genetic background as a risk factor for IBD-EIM alongside known mechanisms such as malabsorption and medication. Clustering patients based on distinctive pathways may enable stratification of patients to predict development of EIM. (Inflamm Bowel Dis 2014;20:987–994) Key Words: extraintestinal manifestations, complications, inflammatory bowel disease, genetic risk, biological pathways
T
he inflammatory bowel diseases (IBD) are complex diseases encompassing multiple specific subphenotypes. Included in IBD’s disease spectrum are various extraintestinal manifestations and complications (together named EIM). In addition to having a high incidence in patients with IBD, the EIM can actually cause more morbidity than IBD itself. Frequently occurring EIM are immune-mediated and involve multiple organ systems like the
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.ibdjournal.org). Received for publication January 31, 2014; Accepted February 19, 2014. From the Departments of *Genetics, and †Gastroenterology and Hepatology, University Medical Centre Groningen, University of Groningen, Groningen, the Netherlands. R. K. Weersma is supported by a VIDI grant (016.136.308) from the Netherlands Organization for Scientific Research (NWO) and a MLDS grant (WO 11–72) of the Dutch Digestive Foundation. J. Fu is supported by a VENI grant (863.09.007). L. Franke received a Horizon Breakthrough grant from the Netherlands Genomics Initiative (93519031) and a VENI grant from NWO (ZonMW grant, 916.10.135). The research leading to these results has received funding from the European Community’s Health Seventh Framework Program (FP7/2007–2013) under grant agreement 259867. This study was financed in part by the SIA-raakPRO subsidy for project BioCOMP. Reprints: Rinse K. Weersma, MD, PhD, Department of Gastroenterology and Hepatology, University of Groningen, University Medical Centre Groningen, P.O. Box 30001, Groningen 9700 RB, the Netherlands (e-mail: [email protected]). Copyright © 2014 Crohn’s & Colitis Foundation of America, Inc. DOI 10.1097/MIB.0000000000000032 Published online 15 April 2014.
Inflamm Bowel Dis Volume 20, Number 6, June 2014
skin (erythema nodosum, pyoderma gangrenosum), the eyes (uveïtis, episcleritis), the liver (primary sclerosing cholangitis, PSC), or spine (ankylosing spondylitis, AS).1 The skin and eye diseases occur during active intestinal disease, whereas PSC and AS can also occur while IBD itself is in a quiescent state. Decreased bone mineral density (BMD), gallstones and kidney stones are considered to be mainly driven by the (metabolic) consequences of malabsorption and chronic use of steroids. Longstanding active disease predisposes to the development of colorectal carcinoma (CRC), necessitating frequent surveillance colonoscopies.2 Despite this knowledge, it remains unclear why the IBD phenotype includes so many EIM, in what way disease pathogenesis is overlapping between IBD and EIM and why certain patients develop a particular EIM and others do not. One possible answer to these questions is an overlap in genetic architecture; this would explain why these EIM co-occur so often with IBD and moreover variation in the genetic make-up in individuals could partly explain the variability in disease phenotype. It has been demonstrated before that many immune-mediated diseases share a genetic background, and because IBD and its EIM co-occur so frequently we hypothesize that they have a common genetic background.3 It has become clear that IBD is not caused by an abnormality in a single gene leading to a single uniform disease, but is a consequence of the perturbations of complex pathways www.ibdjournal.org |
987
van Sommeren et al
leading to multiple specific subphenotypes.1 Recently there has been tremendous progress in unraveling the genetic background of ulcerative colitis and Crohn’s disease, to date 163 independent genetic susceptibility loci have been identified. The identified single genetic risk factors have been shown to cooperate in disease relevant pathways.4 Similarly, genetic variants predisposing to many EIM have been established, of which some are shared with IBD. We hypothesize that, next to these overlapping genetic risk loci, disease pathways are shared between IBD and EIM. To understand how these genetic variants predispose to IBD-EIM, we performed a survey of overlapping genetic loci. For this we included the following EIM or complications of IBD: AS, PSC, decreased BMD, colorectal carcinoma, erythema nodosum, pyoderma gangrenosum, uveïtis, episcleritis, kidney stones, gallstones, and venous thrombosis. Next, we analyzed coregulation between all pairs of genes within IBD loci and EIM loci (with available and reliable genetic data), for protein–protein interactions (PPI) and cotranscription using large databases (including .80,000 human Affymetrix mRNA expression data sets). We show that the pathogenetic overlap between IBD and its EIM is partly driven by a shared genetic predisposition and extends beyond purely shared risk genes to distinct shared biological pathways.
MATERIALS AND METHODS Literature Search and Selection of Associated Loci Genetic association data for IBD was extracted from the most recent analyses.4 An extensive literature search was performed in Pubmed and the GWAS catalog (www.genome.gov/ gwastudies) till October 2011 to assess available data on genetic susceptibility of the EIM. For EIM, all performed GWAS and GWAS meta-analyses were included. We also included large candidate studies if limited GWAS data were available. From GWAS, we included single nucleotide polymorphisms (SNPs) with reported genome wide significance (in most studies defined as P , 5E-8). For candidate studies, we included SNPs with P , 1E-4. For PSC, we also included the GPBAR1 locus because this locus showed functional evidence.5 When there were only small candidate gene studies available or with conflicting results, we did not include the identified susceptibility loci for the pathway analyses. All well-established genetic loci in IBD and the EIM were included and genetic overlap was assessed as presented in Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456.
Mapping SNPs to Genes For the cotranscriptional and PPI analyses, all candidate genes in the loci were included. Therefore, we assessed which genes are located at associated loci in IBD and EIM. The diseaseassociated SNPs were linked to proximate candidate genes in linkage disequilibrium (LD) with them, using a previously described approach.6 We downloaded the recombination hotspot
988
| www.ibdjournal.org
Inflamm Bowel Dis Volume 20, Number 6, June 2014
and LD information from www.hapmap.org for CEU population (release 28) and genome build hg18. The information of gene positions was based on genome build hg18 and was downloaded from UCSC Genome Browser. For each of the associated SNPs, we first defined the disease locus as the region containing the SNPs with LD r2 . 0.5 to the associated SNP and then extended it to the nearest recombination hotspot. This region was further extended 100 kb on each side to include the potential regulatory regions for genes. If any transcript isoform of a gene overlaps with the defined disease locus, this gene was included as a candidate gene. Thus an ith IBD locus with n number of candidate genes was defined as IBDi {g1, g2, ., gn}; and a jth EIM locus with z number of candidate genes was defined as EIMj {ga, gb, ., gz}.
PPI Between IBD and EIM Loci We used the PPI database that was described by Rossin et al6 and Lage et al7 and extracted all the direct interactions between the candidate genes in IBD loci and the candidate genes in EIM loci, for example, between the given ith IBD loci IBDi {g1, g2, ., gn} and the given jth EIM loci EIMj {ga, gb, ., gz}. To assess the specificity of the interactions between loci, we further calculated P values empirically by testing interactions between the EIMj loci and the mimicked random IBDi locus with similar number of genes: if n # 10, the mimicked locus must contain the same n number of genes; if n . 10, the mimicked random locus could contain the number of genes within 10% variation (i.e., gene numbers within 0.9*n and 1.1*n). The random IBDi locus was mimicked 1000 times. We then scored, out of 1000 random loci, how many times we could observe at least 1 interaction between the EIMj locus and the random IBDi locus and calculated the empirical P value. The significance threshold was controlled at 0.05.
Cotranscriptional Interactions Between IBD and EIM Loci To assess the similarities in transcription between genes in IBD loci and EIM loci, we used a new gene cotranscriptional network based on gene expression (www.genenetwork.nl). First, to derive a regulatory model of the human transcriptome, we performed principal component analysis on microarray expression data on 3 species (homo sapiens, mus musculus, and rattus norvegicus) publicly available in Gene Expression Omnibus. We used data for samples hybridized to HG-U133 Plus 2.0, HG-U133A, MG-430 2.0, and RG-230 2.0 Affymetrix Genechip platforms and performed principal component analysis on the probe correlation matrix of each of the 4 platforms. We ended up with 777, 377, 677, and 375 robust principal components and eigenvectors, respectively; each component describing an underlying factor that regulates expression. We treated each of these components as being equally interesting: Although the first components from each of the 4 platforms explain most variation in expression, they overshadow other (subtle) relationships that also exist within these data, which have
Inflamm Bowel Dis Volume 20, Number 6, June 2014
Shared Biological Pathways in IBD and Its EIM
been captured by other components. Conventional coexpression analysis, which uses, e.g., Pearson’s correlations between the expression levels of pairs of genes and does not use principal component analysis, is likely to miss such subtle relationships. Here, we use a correlation measure that uses the 2200 components between pairs of genes and is able to identify such relationships. First, we mapped each probe set on each of the 4 platforms to a gene and averaged eigenvector coefficients over probe sets mapping to the same gene. Probes with no or unambiguous mapping (due to e.g., cross-hybridization) were excluded from further analysis. Second, we calculated for each pair of human genes, a Pearson product–moment correlation coefficient over the 2200 eigenvector coefficients of each platform that contained probe sets for the genes or, for the mouse and rat platforms, their most similar orthologs. Both the probe set and gene ortholog mapping information were downloaded from Ensembl Biomart (Ensembl release 65). The correlation coefficients were further converted to Z-scores to account for different numbers of available eigenvectors for pairs of genes because of missing orthologs. The Z-score for a pair of genes can be positive or negative and describes the similarity of their regulation. The genes (19,997 in total) form the nodes and the correlation coefficients forms the edges of the network. We used this cotranscriptional network to find significantly cotranscribed genomic regions implicated by SNPs associated with IBD or EIM. First, we determined regions of interest for each phenotype. For each SNP associated with a phenotype, we defined implicated genes by identifying furthest SNPs in both the 30 and 50 directions in LD with it (r2 . 0.5 based on CEU HapMap 2) and extending the region first to the nearest recombination hotspot and then an additional 100 kb in both directions. Each gene in the coregulation network overlapping with this region was considered implicated by the SNP and part of the locus. Overlapping loci were merged together. We then examined the potential cotranscription between the regions associated with IBD and those associated with each of the EIM by finding pairs of genes that show the strongest cotranscription in either direction in each pair of loci between the phenotypes. To ascertain whether genes in a locus were often cotranscribed with genes in other regions, we ran a permutation test with 1000 permutations randomly picking a region from the genome with a similar centimorgan range and a similar number of genes as the IBD locus and repeating the procedure with the random locus replacing the IBD locus. We then applied a 5% of false discovery rate to eliminate potential false-positive gene pairs.
RESULTS
Biological Interpretation of Interactions To identify pathways in which the interacting genes of IBD and the different EIM play roles, we used DAVID Bioinformatics database.8 DAVID uses several resources like GO terms, KEGG—the Kyoto encyclopedia of genes and genomes,9 and BioCarta (www.biocarta.com) to cluster genes together in pathways.
Genetic Overlap We first assessed genetic overlap between IBD and all EIM: AS, PSC, BMD, CRC, erythema nodosum, pyoderma gangrenosum, uveïtis, episcleritis, kidney stones, gall stones, and venous thrombosis. Most robust genetic data were available for AS, CRC, BMD, and PSC. For gallstones, kidney stones, and deep venous thrombosis, limited genetic data were available and only candidate gene studies were available, whereas for the skin EIM (erythema nodosum, pyoderma gangrenosum) and eye EIM (uveïtis, episcleritis), no genetic data were available (Table 1). Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456, shows all associated loci per EIM. The classical immune-driven diseases AS and PSC have the largest number of risk loci that are shared with IBD. In AS, 13 of 18 known risk loci are shared with IBD; for PSC this is 10 of 14. In CRC, 3 of 15 risk loci are shared with IBD and for BMD, this is 1 of 23. For deep venous thrombosis, 2 loci are identified, the traits gallstones and kidney stones both have 1 identified genetic risk locus. None of these loci are also associated with IBD. Figure 1 shows all the shared loci between any EIM and IBD per chromosome. Because the HLA and CARD9 loci are shared between IBD, AS, and PSC, there are 25 unique shared loci across all EIM. Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456 lists all established EIM loci.
PPI and Cotranscriptional Regulation We then searched for statistical significant PPIs and cotranscriptional interactions between IBD genes and EIM genes. From the PPI database, we identified 370 PPIs between IBD and EIM loci, excluding PPIs between shared IBD-EIM loci. The (extended) HLA locus is involved in 57 of 370 interacting loci. After 1000 permutations of testing PPIs between the tested EIM locus and a mimicked IBD locus 86/370 PPIs turned out to be significantly specific (Fig. 2). Interacting IBD-EIM loci can hold multiple PPIs between different genes in the loci. These 370 interacting loci hold
TABLE 1. Available Genetic Data for EIM and Complications of IBDs EIMs/Complications CC Decreased BMD AS PSC Gall stones Kidney stones Venous thrombosis Erythema nodosum Pyoderma gangrenosum Uveïtis Episcleritis
No. of Genetic Studies 910–18 919–27 428–31 432–35 136 137 138 0 0 0 0
www.ibdjournal.org |
989
van Sommeren et al
Inflamm Bowel Dis Volume 20, Number 6, June 2014
FIGURE 1. Chromosomes with risk loci for IBD and its EIM. Each dot represents the position of an associated locus on the chromosome. For shared risk loci, the candidate gene is depicted. Complete loci information can be found in Table, Supplemental Digital Content 1, http://links.lww.com/IBD/A456.
a total of 915 PPIs between IBD and EIM genes (441 of these PPIs involve a gene originating in the HLA locus). From the cotranscriptional network we created, we identified 10,890 significant cotranscribed genes between IBD and its EIM. After 1000 permutations, 446 turned out to be significantly specific (Fig., Supplemental Digital Content 2, http://links.lww.com/IBD/A457).
Pathway Annotation We then annotated the biological pathways in which the interacting genes of IBD and the different EIM play roles, by using the DAVID Bioinformatics database. To ensure the reliability of the observed pathways, an extensive literature search was performed. Because of the small amount of associated loci in gallstones, venous thrombosis and kidney stones, few interactions were found and no defined pathways were established. For AS, PSC,
990
| www.ibdjournal.org
CRC, and BMD, we find several interesting pathways where genes with interactions through PPI or cotranscription play roles. These are highlighted in (Fig. 3), and we discuss them in detail further on.
DISCUSSION We aimed to give an overview of the current genetic data on IBD and its EIM and provide additional downstream pathway analyses using publicly available and new locally developed databases. The data as shown here are first of all important to understand the pathogenetic mechanisms that lead to the co-occurrence of immune-mediated extraintestinal manifestations or the development of long-term complications. Second, in the future, clustering patients based on distinctive involved pathways may enable stratification of patients to predict development of EIM and investigate specific screening protocols.
Inflamm Bowel Dis Volume 20, Number 6, June 2014
Shared Biological Pathways in IBD and Its EIM
FIGURE 2. PPI between IBD loci and EIM loci. Each dot represents the position of a locus associated to IBD or one of the EIM on the chromosomes. Names of the candidate genes in the associated loci are presented on the outer side of the circle. The colored lines represent the PPIs that are significant after permutations between IBD genes and EIM genes; different colors are used for the different EIM. The gray lines represent PPIs that are not significant after permutations.
Colorectal Carcinoma The most prominent shared pathway between IBD and CRC is the regulation of the intrinsic epithelial barrier integrity. Central in this pathway is CDH1; encoding E-cadherin, which anchors cells
together in adherens junctions and is associated to both IBD and CRC.39 Other interacting genes influence epithelial barrier function by regulating the cellular actin cytoskeleton (RHPN2, FMN1, RHOA, ARPC2, CAPN10) or establishing attachment of the cytoskeleton to www.ibdjournal.org |
991
van Sommeren et al
Inflamm Bowel Dis Volume 20, Number 6, June 2014
FIGURE 3. Summary of shared biological pathways in IBD and its EIM. Each square represents an associated gene. An IBD gene with multiple interactions across the EIM is depicted multiple times. A blue line represents the most significant cotranscriptional interaction between 2 genes. PPIs significance after permutations are represented by black lines, nonsignificant PPIs by gray lines. Pathways are encircled, subpathways within a bigger pathway are depicted in gray circles.
the basement membrane (DAG1) or in the basement membrane (LAMB1, LAMA5).40 The TGF-b signaling pathway harbors many genes associated to CRC or IBD, illustrating the crucial regulatory role TGF-b plays in different relevant pathways including epithelial barrier function and the immune system.41 The regulatory functions of TGF-b are shared with Wnt signaling, which is an established pathway in CRC and has recently been associated with UC-related CRC.42
992
| www.ibdjournal.org
Bone Mineral Density Wnt signaling genes are abundantly present in BMDassociated loci (WNT16, CTNNB1, GPR177, LRP5) and have interactions with IBD genes like CDH1 and SMAD3. Wnt signaling has been shown to influence the RANK/RANKL/OPG (TNFSF11, TNFRSF11A, TNFRSF11B) pathway.43 TNFSF11 is associated with BMD and was previously identified as a CD
Inflamm Bowel Dis Volume 20, Number 6, June 2014
locus44; however, in the most recent IBD analyses,4 this locus could not be replicated. Therefore, the association of this locus with IBD remains questionable. TNFSF11 has interactions with its receptor TNFRSF11A and the inhibiting decoy receptor TNFRSF11B, which are both associated to BMD. This pathway is involved in both bone homeostasis as regulating T-cell— dendritic cell communications, dendritic cell survival, and lymph node organogenesis, crucial in IBD pathogenesis.45 We see many coregulation between genes in this pathway and genes that are involved in the adaptive immune system (e.g., PTPN22, CCR6). These findings highlight possibilities for e.g., drug repositioning for Denosumab (Amgen/GlaxoSmithKline), which targets TNFSF11 and is a marketed drug for the treatment of postmenopausal women at high risk of fracture with osteoporosis.46 The gene ITGB7, encoding integrin beta-7, which associates with alpha-4 to form integrin alpha-4/beta-7, resides in a BMD-associated locus and interacts with the IBD gene CDH1. Natalizumab (Biogen, Weston, MA), registered for use in Crohn’s disease, blocks homing of lymphocytes to vascular endothelial cells of the gastrointestinal tract through MADCAM1 on the endothelium and integrin alpha-4/beta-7 on lymphocytes.47 These findings highlight the importance of investigating pathways associated with the broader IBD-EIM phenotype.
Primary Sclerosing Cholangitis and Ankylosing Spondylitis Most striking in the analysis for PSC and AS is the large number of overlapping loci. Even if we focus on PSC and AS genes that are not shared with IBD, we find many interactions in pathways that are shared with IBD. PSC genes and their connected IBD genes play several roles in T-cell signaling, like T-cell apoptosis (UBASH3A, BCL2L11, FOXO1 and IRF8) and the JAK-STAT signaling pathway (SOCS1, JAK2, STAT3 and TYK2).48,49 For AS, the largest shared pathway is the T-cell apoptosis pathway, with many interactions with IBD genes. The AS-associated genes TAPBPL and NPEPPS function in the same process as the AS-IBD shared gene ERAP1, namely the process of protein antigen binding to the MHC-I molecules, which is likely to be crucial to both AS and IBD pathogenesis.50 Given the fast moving field of IBD genetics, we do not provide a definitive analysis. Very recently, the number of shared genes between IBD and other diseases has increased with the publication of several articles using a customized GWAS chip (Immunochip, Illumina, San Diego, CA) focussed on immunemediated diseases.4,51 At the time of this analyses, these data were not available but the number of shared pathways is expected to increase substantially. Large-scale analyses of completely phenotyped IBD cohorts (including reliable data on EIM) are necessary to further investigate whether patients with specific EIM are enriched for risk genes or risk pathways as presented here. In conclusion, we show that the pathogenetic overlap between IBD and its EIM or complications extends substantially beyond purely shared risk genes to shared biological pathways. We identified numerous statistically significant interactions
Shared Biological Pathways in IBD and Its EIM
clustering in several distinct biological pathways between IBD and the different EIMs. Hereby, we further highlight the genetic background as risk factor for IBD-EIM next to known mechanisms as malabsorption, chronic inflammation, or medication.
ACKNOWLEDGMENTS Author contributions: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma have contributed equally. Study design: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma; Data analysis and interpretation: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma; Manuscript preparation: S. van Sommeren, M. Janse, J. Karjalainen, L. Franke, J. Fu, and R. K. Weersma.
REFERENCES 1. Podolsky DK. Inflammatory bowel disease. N Engl J Med. 2002;347: 417–429. 2. Rothfuss KS, Stange EF, Herrlinger KR. Extraintestinal manifestations and complications in inflammatory bowel diseases. World J Gastroenterol. 2006;12:4819–4831. 3. Zhernakova A, van Diemen CC, Wijmenga C. Detecting shared pathogenesis from the shared genetics of immune-related diseases. Nat Rev Genet. 2009;10:43–55. 4. 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. 5. Hof JR, Keitel V, Laerdahl JK, et al. Mutational characterization of the bile acid receptor TGR5 in primary sclerosing cholangitis. PLoS One. 2010;5:e12403. 6. Rossin EJ, Lage K, Raychaudhuri S, et al. Proteins encoded in genomic regions associated with immune-mediated disease physically interact and suggest underlying biology. PLoS Genet. 2011;7:e1001273. 7. Lage K, Karlberg EO, Størling ZM, et al. A human phenome-interactome network of protein complexes implicated in genetic disorders. Nat Biotechnol. 2007;25:309–316. 8. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009;4:44–57. 9. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27–30. 10. Houlston RS, Cheadle J, Dobbins SE, et al. Meta-analysis of three genome-wide association studies identifies susceptibility loci for colorectal cancer at 1q41, 3q26.2, 12q13.13 and 20q13.33. Nat Genet. 2010;42: 973–977. 11. Cui R, Okada Y, Jang SG, et al. Common variant in 6q26-q27 is associated with distal colon cancer in an Asian population. Gut. 2011;60: 799–805. 12. Tomlinson IP, Webb E, Carvajal-Carmona L, et al. A genome-wide association study identifies colorectal cancer susceptibility loci on chromosomes 10p14 and 8q23.3. Nat Genet. 2008;40:623–630. 13. Houlston RS, Webb E, Broderick P, et al. Meta-analysis of genome-wide association data identifies four new susceptibility loci for colorectal cancer. Nat Genet. 2008;40:1426–1435. 14. Tomlinson IP, Webb E, Carvajal-Carmona L, et al. A genome-wide association scan of tag SNPs identifies a susceptibility variant for colorectal cancer at 8q24.21. Nat Genet. 2007;39:984–988. 15. Zanke BW, Greenwood CM, Rangrei J, et al. Genome-wide association scan identifies a colorectal cancer susceptibility locus on chromosome 8q24. Nat Genet. 2007;39:989–994. 16. Tenesa A, Farrington SM, Prendergast JG, et al. Genome-wide association scan identifies a colorectal cancer susceptibility locus on 11q23 and replicates risk loci at 8q24 and 18q21. Nat Genet. 2008;40:631–637. 17. Jaeger E, Webb E, Howarth K, et al. Common genetic variants at the CRAC1 (HMPS) locus on chromosome 15q13.3 influence colorectal cancer risk. Nat Genet. 2008;40:26–28. www.ibdjournal.org |
993
van Sommeren et al
18. Broderick P, Carvajal-Carmona L, Pittman AM, et al. A genome-wide association study shows that common alleles of SMAD7 influence colorectal cancer risk. Nat Genet. 2007;39:1315–1317. 19. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, et al. Multiple genetic loci for bone mineral density and fractures. N Engl J Med. 2008;358: 2355–2365. 20. Rivadeneira F, Styrkarsdottir U, Estrada K, et al. Twenty bone-mineraldensity loci identified by large-scale meta-analysis of genome-wide association studies. Nat Genet. 2009;41:1199–1206. 21. Cho YS, Go MJ, Kim YJ, et al. A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing eight quantitative traits. Nat Genet. 2009;41:527–534. 22. Richards JB, Rivadeneira F, Inouye M, et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet. 2008;371:1505–1512. 23. Guo Y, Zhang LS, Yang TL, et al. IL21R and PTH may underlie variation of femoral neck bone mineral density as revealed by a genome-wide association study. J Bone Miner Res. 2010;25:1042–1048. 24. Timpson NJ, Tobias JH, Richards JB, et al. Common variants in the region around Osterix are associated with bone mineral density and growth in childhood. Hum Mol Genet. 2009;18:1510–1517. 25. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, et al. New sequence variants associated with bone mineral density. Nat Genet. 2009;41:15–17. 26. Xiong DH, Liu XG, Guo YF, et al. Genome-wide association and follow-up replication studies identified ADAMTS18 and TGFBR3 as bone mass candidate genes in different ethnic groups. Am J Hum Genet. 2009;84:388–398. 27. Kung AW, Xiao SM, Cherny S, et al. Association of JAG1 with bone mineral density and osteoporotic fractures: a genome-wide association study and follow-up replication studies. Am J Hum Genet. 2010;86:229–239. 28. Evans DM, Spencer CC, Pointon JJ, et al. Interaction between ERAP1 and HLA-B27 in ankylosing spondylitis implicates peptide handling in the mechanism for HLA-B27 in disease susceptibility. Nat Genet. 2011;43:761–767. 29. Australo-Anglo-American Spondyloarthritis Consortium (TASC), Reveille JD, Sims AM, et al. Genome-wide association study of ankylosing spondylitis identifies non-MHC susceptibility loci. Nat Genet. 2010; 42:123–127. 30. Danoy P, Pryce K, Hadler J, et al. Association of variants at 1q32 and STAT3 with ankylosing spondylitis suggests genetic overlap with Crohn’s disease. PLoS Genet. 2010;6:e1001195. 31. Wellcome Trust Case Control Consortium, Burton PR, Clayton DG, et al. Association scan of 14,500 nonsynonymous SNPs in four diseases identifies autoimmunity variants. Nat Genet. 2007;39:1329–1337. 32. Folseraas T, Melum E, Raush P, et al. Extended analysis of a genomewide association study in primary sclerosing cholangitis detects multiple novel risk loci. J Hepatol. 2012;57:366–375. 33. Janse M, Lamberts LE, Franke L, et al. Three ulcerative colitis susceptibility loci are associated with primary sclerosing cholangitis and indicate a role for IL2, REL, and CARD9. Hepatology. 2011;53:1977–1985.
994
| www.ibdjournal.org
Inflamm Bowel Dis Volume 20, Number 6, June 2014
34. Melum E, Franke A, Schramm C, et al. Genome-wide association analysis in primary sclerosing cholangitis identifies two non-HLA susceptibility loci. Nat Genet. 2011;43:17–19. 35. Karlsen TH, Franke A, Melum E, et al. Genome-wide association analysis in primary sclerosing cholangitis. Gastroenterology. 2010;138: 1102–1111. 36. Buch S, Schafmayer C, Völzke H, et al. A genome-wide association scan identifies the hepatic cholesterol transporter ABCG8 as a susceptibility factor for human gallstone disease. Nat Genet. 2007;39:995–999. 37. Thorleifsson G, Holm H, Edvardsson V, et al. Sequence variants in the CLDN14 gene associate with kidney stones and bone mineral density. Nat Genet. 2009;41:926–930. 38. Trégouët DA, Heath S, Saut N, et al. Common susceptibility alleles are unlikely to contribute as strongly as the FV and ABO loci to VTE risk: results from a GWAS approach. Blood. 2009;113:5298–5303. 39. Van Roy F, Berx G. The cell-cell adhesion molecule E-cadherin. Cell Mol Life Sci. 2008;65:3756–3788. 40. Turner JR. Intestinal mucosal barrier function in health and disease. Nat Rev Immunol. 2009;9:799–809. 41. Massagué J, Blain SW, Lo RS. TGFbeta signaling in growth control, cancer, and heritable disorders. Cell. 2000;103:295–309. 42. Van Dekken H, Wink JC, Vissers KJ, et al. Wnt pathway-related gene expression during malignant progression in ulcerative colitis. Acta Histochem. 2007;109:266–272. 43. Häusler KD, Horwood NJ, Chuman Y, et al. Secreted frizzled-related protein-1 inhibits RANKL-dependent osteoclast formation. J Bone Miner Res. 2004;19:1873–1881. 44. Franke A, McGovern DP, Barrett JC, et al. Genome-wide meta-analysis increases to 71 the number of confirmed Crohn’s disease susceptibility loci. Nat Genet. 2010;42:1118–1125. 45. Tilg H, Moschen AR, Kaser A, et al. Gut, inflammation and osteoporosis: basic and clinical concepts. Gut. 2008;57:684–694. 46. Sanseau P, Agarwal P, Barnes MR, et al. Use of genome-wide association studies for drug repositioning. Nat Biotechnol. 2012;30:317–320. 47. Ghosh S, Goldin E, Gordon FH, et al. Natalizumab for active Crohn’s disease. N Engl J Med. 2003;348:24–32. 48. O’Shea JJ, Gadina M, Schreiber RD. Cytokine signaling in 2002: new surprises in the Jak/Stat pathway. Cell. 2002;109:S121–S131. 49. Hughes PD, Belz GT, Fortner KA, et al. Apoptosis regulators Fas and Bim cooperate in shutdown of chronic immune responses and prevention of autoimmunity. Immunity. 2008;28:197–205. 50. Chang SC, Momburg F, Bhutani N, et al. The ER aminopeptidase ERAP1, trims precursors to length of MHC class I peptides by a “molecular ruler” mechanism. Proc Natl Acad Sci U S A. 2005;102:17107– 17112. 51. Liu JZ, Hov JR, Folseraas T, et al. Dense genotyping of immune-related disease regions identifies nine new risk loci for primary sclerosing cholangitis. Nat Genet. 2013;45:670–675.
