867848 research-article20192019 TAG0010.1177/1756284819867848Therapeutic Advances in GastroenterologyM Chaparro, A Aterido Therapeutic Advances in Gastroenterology Original Research Functional rare variants influence the Ther Adv Gastroenterol 2019, Vol. 12: 1–14 clinical response to anti-TNF therapy in DOI: 10.1177/ https://doi.org/10.1177/1756284819867848 https://doi.org/10.1177/1756284819867848 1756284819867848 Crohn’s disease © The Author(s), 2019. Article reuse guidelines: sagepub.com/journals- permissions María Chaparro*,# , Adrià Aterido#, Iván Guerra, Marisa Iborra*, Jose Luis Cabriada, Luis Bujanda*, Carlos Taxonera , Valle García-Sánchez*, Ignacio Marín-Jiménez, Correspondence to: María Chaparro Manuel Barreiro-de Acosta, Isabel Vera, Maria Dolores Martín-Arranz, Borja Hernández- Gastroenterology Unit. Breijo, Francisco Mesonero, Laura Sempere, Fernando Gomollón*, Joaquín Hinojosa, Hospital Universitario de La Princesa, Instituto de Fernando Bermejo, Belén Beltrán*, Ainhoa Rodríguez-Pescador, Jesús María Banales*, Investigación Sanitaria David Olivares, Patricia Aguilar-Melero, Luis Menchén, Rocío Ferreiro-Iglesias, Princesa (IIS-IP), Universidad Autónoma de Isabel Blazquez Gómez, Beatriz Benitez García, Luis G Guijarro, Alicia C Marin*, Madrid, 62. 28006 Madrid, David Bernardo*, Sara Marsal, Antonio Julia** and Javier P Gisbert*,** on behalf Spain mariachs2005@gmail.com of the PREDICROHN Study Group from GETECCU Adrià Aterido Rheumatology Research Group, Vall d’Hebron Research Institute, Barcelona, Spain Abstract Department of Experimental Background: The effect of low-frequency functional variation on anti-tumor necrosis factor and Health Sciences, Universitat Pompeu Fabra, alpha (TNF) response in Crohn’s disease (CD) patients remains unexplored. The objective of Barcelona, Spain this study was to investigate the impact of functional rare variants in clinical response to anti- Iván Guerra TNF therapy in CD. Fernando Bermejo Gastroenterology Unit, Methods: CD anti-TNF naïve patients starting anti-TNF treatment due to active disease Hospital Universitario de Fuenlabrada, Instituto de [Crohn’s Disease Activity Index (CDAI > 150)] were included. The whole genome was Investigación de Hospital sequenced using the Illumina Hiseq4000 platform. Clinical response was defined as a CDAI La Paz (IdiPaz), Madrid, Spain score <150 at week 14 of anti-TNF treatment. Low-frequency variants were annotated and Marisa Iborra classified according to their damaging potential. The whole genome of CD patients was Belén Beltrán Gastroenterology Unit, screened to identify homozygous loss-of-function (LoF) variants. The TNF signaling pathway Hospital Universitario de was tested for overabundance of damaging variants using the SKAT-O method. Functional La Fe, Valencia, Spain Jose Luis Cabriada implication of the associated rare variation was evaluated using cell-type epigenetic Ainhoa Rodríguez-Pescador enrichment analyses. Gastroenterology Unit, Hospital Universitario de Results: A total of 41 consecutive CD patients were included; 3250 functional rare variants Galdakano, Vizcaya, Spain were identified (2682 damaging and 568 LoF variants). Two homozygous LoF mutations Luis Bujanda Jesús María Banales were found in HLA-B and HLA-DRB1 genes associated with lack of response and remission, Gastroenterology Unit, respectively. Genome-wide LoF variants were enriched in epigenetic marks specific for the Hospital Universitario de Donostia, Instituto gastrointestinal tissue (colon, p = 4.11e–4; duodenum, p = 0.011). The burden of damaging Biodonostia, UPV/EHU, Ikerbasque, Guipúzcoa, variation in the TNF signaling pathway was associated with response to anti-TNF therapy Spain (p = 0.016); damaging variants were enriched in epigenetic marks from CD8+ (p = 6.01e–4) Carlos Taxonera David Olivares and CD4+ (p = 0.032) T cells. Gastroenterology Unit, Conclusions: Functional rare variants are involved in the response to anti-TNF therapy in CD. Hospital Universitario Clínico San Carlos and Cell-type enrichment analysis suggests that the gut mucosa and CD8+ T cells are the main IdISSC, Madrid, Spain mediators of this response. Valle García-Sánchez Patricia Aguilar-Melero Gastroenterology Unit, Instituto Maimónides de Keywords: adalimumab, Crohn’s disease, infliximab, genes, tumor necrosis factor alpha, Investigación Biomédica whole-genome analysis de Córdoba (IMIBIC), Hospital Universitario Reina Sofía/Universidad de Received: 10 April 2019; revised manuscript accepted: 28 June 2019. Córdoba, Spain journals.sagepub.com/home/tag 1 Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage). Therapeutic Advances in Gastroenterology 12 Ignacio Marín-Jiménez Luis Menchén Introduction on complex phenotypes than common genetic Gastroenterology Unit, Anti-tumor necrosis factor alpha (TNF) drugs variants.14 From a clinical perspective, loss-of- Hospital Universitario Gregorio Marañón e IiSGM, have been shown to be effective for induction function (LoF) variants are one of the most inter- Madrid, Spain and maintenance of remission in Crohn’s dis- esting forms of rare functional genetic variation.15 Manuel Barreiro-de Acosta ease (CD); however, up to 30% of patients do Accordingly, in the pharmacogenomics field, the Rocío Ferreiro-Iglesias Gastroenterology Unit, not respond to these drugs.1–3 In 2014, the use analysis of LoF variants is starting to provide new Hospital Universitario of vedolizumab, which targets α4β7 integrins, insights into the genetic basis underlying the clini- Clínico de Santiago, Santiago de Compostela, was approved for CD. In 2016, ustekinumab, cal response to therapies in complex diseases. The Spain directed against the p40 subunit shared by inter- analysis of low-frequency variation could there- Isabel Vera leukins 12 and 23, was approved in patients fore shed new light on the biological mechanisms Isabel Blazquez Gómez Gastroenterology Unit, with CD, while the JAK-inhibitor tofacitinib has that are relevant for anti-TNF response in CD. Hospital Universitario just been approved by the European Agency of Puerta de Hierro Majadahonda, Madrid, Spain Medicines for ulcerative colitis (UC) patients, In the present study, we sequenced the whole Maria Dolores Martín- hence increasing the therapeutic arsenal against genome of CD patients treated with anti-TNF Arranz Beatriz Benitez García inflammatory bowel diseases (IBD). All these therapy. At the genome-wide level, we identified Gastroenterology Unit, drugs have high cost, so they pose a great eco- LoF variants affecting critical genes for the Hospital Universitario La Paz, Madrid, Spain nomic burden on health systems. At present, the immune response. Specific analysis of the TNF Borja Hernández-Breijo medical community does not have reliable crite- signaling pathway has also shown the significant Universidad de Alcalá, ria for selecting which patient will benefit from burden of damaging variation associated with Alcalá de Henares, Spain Immuno-Rheumatology any of the above-mentioned drugs. Thus, the treatment response. These findings support the Research Group. IdiPaz. variables (epidemiological, clinical, analytical, importance of functional rare variation for the Hospital Universitario La Paz, Madrid, Spain etc.) commonly used to predict patient response anti-TNF response in CD, and could help to Francisco Mesonero to biologic therapy have shown low utility. select those patients who will benefit most from Gastroenterology Unit, anti-TNF treatment. Hospital Universitario Ramón y Cajal, Madrid, Therefore, it is a priority to identify mechanisms Spain involved in the response to each drug, to suggest Laura Sempere more rational use of resources and for personal- Methods Gastroenterology Unit, Hospital Universitario ized medicine. Thus, achieving that goal will Alicante, Alicante, Spain allow us to indicate the most appropriate treat- Patients Fernando Gomollón Gastroenterology ment to each individual, hence avoiding adminis- A total of 41 patients diagnosed with CD based Unit, Hospital Clínico tering drugs to patients who will not respond (an on European Crohn’s and Colitis Organization Universitario, Lozano Blesa, IIS Aragón, Zaragoza, Spain inadequate use of resources and an unjustified criteria that started anti-TNF treatment were Joaquín Hinojosa risk of adverse effects). included.16 Patients came from the Predicrohn Gastroenterology Unit, study, which is a multicenter study aiming to Hospital Universitario Manises, Valencia, Spain There is compelling evidence that the clinical optimize treatment with anti-TNF drugs.17 The Luis G Guijarro response to anti-TNF therapy has a genetic com- trial included 150 CD patients naïve to anti-TNF Universidad de Alcalá, ponent.4 To date, several candidate-gene studies treatment; only patients with active disease, based Alcalá de Henares, Spain Alicia C Marin have investigated the contribution of common on the Crohn’s Disease Activity Index (CDAI) at David Bernardo variation to the heterogeneity of anti-TNF baseline, were eligible for the genetic approach. Javier P Gisbert Gastroenterology Unit, response in CD.5–9 However, the few associations All patients were naïve to anti-TNF therapy and Hospital Universitario de identified so far have not been consistently repli- received 160/80 mg adalimumab at weeks 0 and La Princesa, Instituto de Investigación Sanitaria cated.10 The study of the genetic basis of the clini- 2, and 40 mg every other week thereafter, or inf- Princesa (IIS-IP), cal response to anti-TNF therapy at the liximab 5 mg/kg at weeks 0, 2, 6, and 14, and Universidad Autónoma de Madrid, Spain genome-wide scale represents a new opportunity every 8 weeks thereafter. Sara Marsal to identify genetic factors underlying the anti- Antonio Julia TNF response in CD. Rheumatology Research Group, Vall d’Hebron Study design Research Institute, The advent of next generation sequencing (NGS) We conducted a prospective, multicenter cohort Barcelona, Spain *Centro de Investigación technologies like whole genome sequencing study. Treatment (infliximab or adalimumab) Biomédica en Red de (WGS) is enabling the analysis and identification was not assigned randomly but based on the cli- Enfermedades Hepáticas y Digestivas (CIBEREHD). of low-frequency variation associated with com- nician’s decision. Clinical response was evalu- #Equally contributed as plex traits,11,12 including the response to anti-TNF ated at week 14 by measuring the change in the First authors. agents.13 In particular, rare variants with a high CDAI score. Remission was defined as **Equally contributed as functional impact can display much larger effects CDAI⩽150 at week 14, and short-term partial Senior authors 2 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. Table 1. Characteristics of the study population. Characteristics Responders (n = 31) Nonresponders (n = 10) p-value Men (%) 13 (41.93) 8 (80.00) 0.066 Location (%) Ileal 9 (29.03) 5 (50.00) 0.27 Colonic 5 (16.13) 0 (0.00) 0.31 Ileocolonic 15 (48.39) 3 (30.00) 0.47 Behavior (%) Inflammatory 20 (64.52) 2 (20.00) 0.026 Stricturing 4 (12.90) 2 (20.00) 0.99 Fistulizing 5 (16.13) 5 (50.00) 0.044 Perianal disease (%) 7 (22.58) 1 (10.00) 0.65 Extraintestinal 12 (38.71) 2 (20.00) 0.45 manifestations (%) Smoking habit (%) 13 (41.94) 3 (30.00) 0.71 Steroids 5 (16.13) 1 (10.00) 0.98 Immunomodulators (%) Thiopurines 21 (67.74) 8 (80.00) 0.69 Methotrexate 2 (6.45) 0 (0.00) 0.99 Anti-TNF type (%) Adalimumab 13 (41.94) 3 (30.00) 0.71 Infliximab 18 (58.06) 7 (70.00) 0.71 Previous Surgery 11 (35.48) 6 (60.00) 0.27 response as a decrease in the CDAI score of at WGS of the study population least 70 points at week 14, without achieving WGS of the 41 CD patients included in the pre- remission. Patients that achieved remission, and sent study was performed using the Illumina those with partial response, were considered Hiseq4000 platform at Centre Nacional d’Anàlisi responders. Patients not meeting those criteria Genòmica (CNAG, Barcelona, Spain). All DNA were considered nonresponders. Concomitant samples were sequenced to an average coverage treatments were also recorded. Blood samples >20× using 150 bp paired-end reads. For each were obtained at baseline. The distribution of sample, we obtained >360 million reads. the categorical variables between responders and Duplicate reads were subsequently excluded from nonresponders to anti-TNF therapy was com- the analysis. The remaining reads were aligned to pared using the Fisher’s exact test (Table 1). the GRCh38/hg38 assembly of the human refer- ence genome using the Burrows Wheeler algo- The study was approved by the ethical review board rithm. Variant calling was performed according to of all participating centers. All patients signed an the best practices outlined in the Genome Analysis informed written consent to participate. ToolKit (GATK).18 journals.sagepub.com/home/tag 3 Therapeutic Advances in Gastroenterology 12 In the quality control (QC) analysis, genetic vari- the Exome Aggregation Consortium (ExAC) ants with a genotype quality ⩾50, a read depth Database,24 damaging variants were characterized ⩾20 and a read position rank sum ⩾–8 were according to their impact using the Sequence selected for the present study. From these, only Ontology tool.25 Briefly, damaging variants were genetic variants that showed a coverage depth ⩾2 classified into: start lost: codon variants that for the alternative allele were included change at least one base of the start codon; stop (n = 3,564,615 genetic variants). In order to dis- gained: sequence variants that change at least one card genetic variants that showed a different geno- base of a codon leading to a premature stop codon type between the forward and reverse strands, and shortening of the protein; stop lost: sequence genetic variation showing a Fisher’s strand bias variants that change at least one base of the termi- ⩾60 and a strand odds ratio ⩾3 were excluded nator codon resulting in an elongated transcript; from the study. In addition, in order to keep only synonymous: sequence variants that do not alter those genetic variants that were confidently the amino acid sequence encoded; missense: mapped, we selected only genetic variation that sequence variants that change one or more bases showed a mapping quality score ⩾40 and a map- resulting in a different amino acid sequence with- ping quality rank sum ⩾–10. After QC, only vari- out altering the length of the protein sequence; ants showing a minor allele frequency <0.05 were splice donor: splice variants that change the two included in downstream analyses. base pair region at the 5′ end of an intron; splice region: sequence variants that change the region In order to discard any confounding effect from of the splice site either within 1–3 bases of the population stratification in the CD patient cohort, exon or 3–8 bases of the intron; 5′ UTR: variants we performed principal component analysis mapping on the 5′ UTR region; 3′ UTR: variants (PCA) incorporating genome-wide association mapping on the 3′ UTR region; structural inter- study (GWAS) data from a cohort of 1558 healthy action: variants that impact the internal interac- controls from the Spanish population. These tions of the resulting polypeptide structure; individuals were collected and genotyped by lincRNA: variants mapping on a long intergenic the Immune-Mediated Inflammatory Disease noncoding RNA region; snRNA: variants map- (IMID) Consortium, as described previously.19 ping on a small nuclear RNA region; sense After selecting the single nucleotide polymor- intronic: variants mapping on a noncoding tran- phisms (SNPs) that were in common between script region located within an intronic region both samples, we evaluated the existence of pop- with no overlap with exonic sequences; nonstop ulation stratification using the EIGENSOFT decay: variants that might impact the translation (v4.2) software.20 Using the first 10 principal of mRNA molecules lacking a stop codon; and components (PCs) of variation over 10 iterations, nonsense-mediated decay: variants that might we did not detect any CD patient showing an out- impact the mRNA degradation. lier genetic background (Figure S1). According to the methodology implemented in SnpEff, genetic variants that correlate with the Annotation of genetic variation complete loss of function of the encoded tran- Low-frequency variants were annotated using the script were classified as LoF. These genetic vari- SnpEff software,21 and their damaging potential ants include: stop codon-introducing variants, was predicted using PolyPhen-2 and Sift.22,23 In start codon-introducing variants, splice site-dis- PolyPhen-2, the functional significance of allele rupting variants, insertion/deletion variants pre- replacement is predicted by a Naïve Bayes classi- dicted to disrupt the transcript reading frame, fier approach that also estimates the probability of and large deletions that remove either the first each variant to be classified as damaging when it exon of the gene or more than 50% of the protein- is nondamaging [false positive rate (FPR)]. In the coding sequence of the affected transcript.21 present study, low-frequency variants were Given that stop codon-introducing and frameshift defined as damaging if classified as possibly or variants are enriched toward the 3′ end of the probably damaging by both PolyPhen-2 (FPR gene and tend to show tolerance to truncation <10% and FPR <20% for probably and possibly close to the 5′ end, putative LoF variants identi- damaging, respectively) and Sift, as previously fied in the last 5% of the coding region were not described.13 Based on the annotation displayed in included in the present study. 4 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. For each variant predicted as damaging or LoF, epigenetic data from H3K4me3 was obtained we also estimated its putative impact at the pro- from the Roadmap Epigenomics project.30 tein level using the method implemented in SnpEff.26 According to this method, three catego- The statistical methodology implemented in the ries were defined: high impact, the variant is EPIGWAS software consists of five consecutive assumed to have a disruptive impact in the pro- steps. First, the LD between the associated vari- tein, probably causing protein truncation, loss of ants and those located within a distance window function or triggering nonsense mediated decay; of 500 kb is computed to define the loci to be moderate impact, the variant is nondisruptive and analyzed. Second, the regulatory activity of each might change the protein effectiveness; and low variant in each cell type is scored as the height of impact, the variant is assumed to be mostly harm- the nearest H3K4me3 peak, divided by the dis- less or unlikely to change the protein behavior. tance between the SNP and the summit of the peak. For each cell type and locus, the value of the variant with the highest score is selected as a SKAT-O rare variant burden test measure of the overlap between the locus and the The contribution of low-frequency damaging vari- cell type-specific regulation. Third, these scores ants from the TNF signaling pathway in the clini- are normalized using Euclidean methodology. cal response to anti-TNF therapy was investigated Fourth, the scores from all loci are summed using the SKAT-O set-based collapsing method within each cell type in order to compute the cell implemented in EPACTS software.27 This statisti- type regulatory score. In the last step, the statisti- cal approach was designed to efficiently unify both cal significance of cell-type-specific overlap was the burden and nonburden algorithms for rare obtained using a permutation-based approach variant analysis. While the former method col- (n = 10,000 permutations). In each permuta- lapses genetic variation into a single burden varia- tion, a matched set of SNPs is selected from a set ble, the latter aggregates the individual variant-score of independent SNPs from the HapMap Project test statistics resulting from the association analysis and the cell type specific regulatory score is com- into a kernel matrix. With this algorithm, the puted again for each cell type. The statistical sig- SKAT-O test has been found to maximize the sta- nificance of the enrichment analysis is finally tistical power of the analysis. We used the SKAT-O defined as the proportion of sets of SNPs matched test to compare the global burden of low-frequency with a cell type specific regulatory score exceed- variation at the TNF signaling pathway between ing the observed score for the associated variants responders and nonresponders to anti-TNF ther- of interest. apy. This statistical analysis was adjusted for the clinical variables that showed a significantly differ- ent distribution between responder and nonre- Results sponder patients (Table 1). The annotation for the TNF signaling pathway was obtained from the Study population Kyoto Encyclopedia of Genes and Genomes A total of 41 CD patients were included. The (KEGG) reference database (hsa04668, n = 108 main characteristics of the study cohort are sum- genes).28 marized in Table 1. Mean CDAI was 256 (stand- ard deviation = 57). Most of the patients had ileocolonic location and inflammatory behavior, Epigenetic enrichment analysis and over 40% had been operated on due to CD. In order to identify the relevant cell types for anti- Approximately 40% received adalimumab and TNF response, we performed an epigenetic 60% infliximab. At week 14, 25 patients (61%) enrichment analysis. First, we analyzed the achieved remission and 6 had partial response enrichment of genome-wide LoF variation (n = (14.6%), and therefore considered responders, 568 SNPs) on cell-type-specific H3K4me3 chro- and 10 (24.4%) were primary nonresponders. matin marks using the EPIGWAS software.29 When comparing the distribution of the different Second, we followed the same procedure to ana- epidemiological and clinical variables between lyze the enrichment of damaging variation at the responders and nonresponders to anti-TNF ther- TNF signaling pathway on cell-type-specific apy, we detected significant differences only in H3K4me3 chromatin marks. Cell type-specific the disease behavior (p < 0.05, Table 1). In journals.sagepub.com/home/tag 5 Therapeutic Advances in Gastroenterology 12 Table 2. Characterization of the damaging low- Genome-wide LoF variants associated with frequency genetic variation identified in the study anti-TNF therapy response population. We detected 568 low-frequency LoF variants Variant annotation Variants* associated with response to anti-TNF therapy, including 281 and 386 variants in adalimumab- Start lost 8 and infliximab-treated patients, respectively (n = 99 overlapping variants). A total of 275 Stop gained 2 and 384 low-frequency LoF variants were Stop lost 2 detected in patients with and without prior sur- gery, respectively (n = 92 overlapping variants). Synonymous 30 They were all heterozygous LoF except for two Missense 2590 variants that had homozygous LoF carriers. These variants mapped to the HLA-B (SNP Splice donor 2 rs41563412) and HLA-DRB1 (SNP rs1071752) genes. The unique homozygous carrier of the Splice region 89 damaging HLA-B rs41563412-GCA variant Structural interaction 50 (frameshift mutation, p.Ser33LeufsTer9) did not respond to anti-TNF therapy (i.e. adali- Long intergenic noncoding RNA 56 mumab) and was not involved in a surgical inter- Sense intronic 17 vention. Conversely, the three patients that carried one or two copies of the damaging HLA- Small nuclear RNA 12 DRB1 rs1071752-C variant (synonymous muta- tion, p.Gln125Gln) were all responders to 3’ UTR 479 anti-TNF therapy (i.e. infliximab), and only one 5’ UTR 86 of them was not involved in a surgical interven- tion. Both SNPs were found to be protein-cod- Nonsense-mediated decay 1 ing variants with a high functional impact at the Nonstop decay 9 protein level. The strongest damaging impact was detected in 10 LoF SNPs mapping at *Number of low-frequency damaging variants showing NECAP2, HEYL, THNSL2, NPHP3-ACAD11, the indicated annotation. Given that a single genetic CLDN18, LY6G6C, C6orf226, RAB2IL1 and variant can impact different genes, the annotated variants outnumber those predicted as damaging. EFCAB5 genes (Table 3). The SNP chr2:88173152 at THNSL2 gene was not anno- tated in the ExAC or dbSNP databases,31 sug- gesting this could be a new genetic variant. addition, there was no difference in CDAI score Using cell-type epigenetic enrichment analysis, we at baseline between responders and n ­ onresponders found that the genome-wide LoF variants associ- (259 versus 246, p = 0.5). ated with anti-TNF response were significantly enriched for epigenetic marks characteristic of the colonic (p = 4.11e–4) and duodenum mucosa Classification of low-frequency damaging (p = 0.011, Figure 1). In patients with prior sur- variants gery, the genome-wide LoF variants associated A total of 14,994,115 genetic variants were iden- with the overall anti-TNF response were signifi- tified by WGS, of which 1,738,956 were found cantly enriched in epigenetic marks from the to be low-frequency variants. Using PolyPhen-2 CD34+ hematopoietic precursor cells (p = 0.021), and Sift, 2682 low-frequency genetic variants whereas a significant enrichment in epigenetic were predicted to be damaging. In order to char- marks from the colonic (p = 4.10e–4) and duode- acterize the effect of the damaging variants, we num mucosa (p = 7.61e–3) was detected in classified them according to their damaging patients without prior surgery. When performing potential (Table 2). Overall, we found that 1.9% the cell-type epigenetic enrichment analysis strati- and 98.1% of the low-frequency damaging vari- fied by drug, the infliximab-associated LoF variants ants have a high and moderate/low impact at the showed also a significant enrichment in epigenetic protein level, respectively. marks from the colonic mucosa (p = 6.01e–4) as 6 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. Table 3. Characterization of the damaging loss-of-function variants identified in the study population. SNP Chr Coord A1 A2 Gene Variant Annotation CD (%) Resp Nresp type rs150581659 1 16447908 G A NECAP2 exonic stop gained 1 (2.4) 1 0 rs761330653 1 39639623 C T HEYL exonic start lost 1 (2.4) 1 0 rs764641613 1 39639624 A C HEYL exonic start lost 1 (2.4) 1 0 chr2:88173152 2 88173152 T C THNSL2 exonic start lost 1 (2.4) 0 1 rs41272317 3 132618633 C A NPHP3- intronic splice donor 1 (2.4) 0 1 ACAD11 rs138856042 3 137998869 A G CLDN18 exonic start lost 1 (2.4) 1 0 rs151314696 6 31719303 C T LY6G6C exonic start lost 1 (2.4) 1 0 rs61732354 6 42890786 C T C6orf226 exonic start lost 2 (4.8) 1 1 rs192561318 11 61920191 A G RAB3IL1 intronic start lost 1 (2.4) 1 0 rs79556405 17 29941799 G A EFCAB5 intronic start lost 4 (9.7) 2 2 A1, reference allele; A2, alternative allele that causes the LoF of the protein encoded by the indicated gene; CD, number of CD patients carrying the LoF variant; Chr, chromosome; Coord, SNP coordinates in GRCh38/hg38; Nresp, number of patients that carry the LoF variant and did not show a significant clinical response to anti-TNF therapy; Resp, number of patients that carry the LoF variant and responded to anti-TNF therapy; SNP, single nucleotide polymorphism. Figure 1. Enrichment of genome-wide loss-of-function variation in H3K4me3 histone marks from specific cell types. Bar plot showing the statistical significance of the overlap between genome-wide loss-of-function variation and H3K4me3 histone marks from the 34 cell types analyzed. All cell types are shown in the x axis, and those with H3K4me3 histone marks significantly enriched in genome-wide loss-of-function variation are represented with an asterisk (p < 0.05). journals.sagepub.com/home/tag 7 Therapeutic Advances in Gastroenterology 12 Figure 2. Enrichment of damaging variation at TNF signaling pathway in H3K4me3 histone marks from specific cell types. Bar plot showing the statistical significance of the overlap between damaging variation from the TNF signaling pathway and H3K4me3 histone marks from the 34 cell types analyzed. All cell types are shown in the x axis, and those with H3K4me3 histone marks significantly enriched in damaging variation from the TNF signaling pathway are represented with an asterisk (p < 0.05). TNF, tumor necrosis factor alpha. well as from CD34+ hematopoietic precursor cells The analysis of low-frequency genetic variants (p = 0.012). The adalimumab-associated LoF var- mapping to the TNF signaling pathway using the iants showed a significant enrichment in epigenetic SKAT-O test showed that the damaging burden marks from the colonic mucosa (p = 2.81e–3), is different between anti-TNF responder (partial colonic smooth muscle (p = 0.02), CD4+ T cells response or remission) and nonresponder patients (p = 0.034) and CD34+ hematopoietic precursor (p = 0.016). When stratifying the analysis by cells (p = 0.043). anti-TNF agent, we detected that the burden of damaging variation is significantly different between responders and nonresponders to inflixi- Association of damaging variation at the TNF mab therapy (p = 9.76e–3). No significant differ- signaling pathway with anti-TNF response ences were detected among adalimumab-treated A total of 19 damaging low-frequency genetic patients (p > 0.05). The damaging burden was variants were found to map to the TNF signaling found to be also significantly different between pathway (hsa04668). The analysis of this varia- responders and nonresponders to anti-TNF ther- tion using the SKAT-O test showed that the apy in patients without prior surgery (p = 0.011), damaging burden is different between anti-TNF but not among those patients that had not been responder (partial response or remission) and previously involved in a surgical intervention nonresponder patients (p = 0.016). The cell-type (p > 0.05). enrichment analysis showed an increase in epige- netic marks specific from immune cell types (Figure 2), principally CD8+ T cells (p = 6.01e– Discussion 4) but also CD4+ T cells (p = 0.032). Like in In this study, we sequenced the whole genome of the functional characterization of the LoF vari- CD patients to examine the hypothesis that func- ants, epigenetic enrichment analysis also indi- tional rare variants are associated with anti-TNF cated an association with gut tissue (colonic response. Using genome-wide and set-based anal- mucosa, p = 0.031). yses, we found evidence of LoF and damaging 8 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. variants influencing the anti-TNF response. Cell not be found. Of relevance, the LY6G6C gene type enrichment analysis suggests that these vari- maps to a cluster of leukocyte antigen genes ants affect the cytotoxic response of the adaptive located in the MHC-III.41 The MHC-III locus immune system (CD8+ naïve T-cells) and the also contains the TNF gene, indicating that this gastrointestinal mucosa. genomic region is likely to be important in regu- lating the inflammatory activity of immune-medi- A major finding from this study is that, among the ated diseases.42 Given the physical proximity between millions of variants of the human genome, the the LY6G6C and TNF genes (i.e. ~140 kb), our only two homozygous LoF SNPs were found to results suggest that variation at LY6G6C could map to HLA genes. These two variants, rs1071752 alter the expression of the TNF gene, and, conse- and rs41563412, are located in an exonic region quently, influence the anti-TNF response. of the HLA-DRB1 gene and an intronic region of the HLA-B gene, respectively. The proteins We also found that the LoF variant at THNSL2 encoded by the HLA-DRB1 and HLA-B genes gene (i.e. chr2:88173152 SNP) is not annotated play a central role in the adaptive immune system in the databases of reference for genetic varia- by presenting antigens to T-cells, including CD4+ tion.31 The THNSL2 gene encodes a threonine [in a major histocompatibility complex class II synthase-like protein. To date, however, its bio- (MHC)-II dependent manner as is the case of logical function in humans is still unknown. The HLA-DRB1] and CD8+ T-cells (in an MHC-I scarce information on the biology of the THNSL2 dependent manner as the HLA-B), being the gene derives from an in vitro study showing that adaptive immune response altered in CD THNSL2 is transcribed into a rare mRNA splice- patients.32–34 Although it has been shown that variant that leads to SOAFT formation, a human genetic factors underlying clinical phenotypes of T cell secreted cytokine that exacerbates the CD (i.e. other than anti-TNF response) might not inflammatory state of autoimmune diseases.43 In necessarily be shared with CD risk loci, the HLA our study cohort, the nonresponse phenotype of region itself is indeed one of the strongest risk loci the patient carrying the chr2:88173152 SNP sug- for CD, and multiple genetic association studies gests that this variant could have a functional have previously identified variation at HLA-DRB1 effect on the anti-TNF mechanism of action that and HLA-B genes showing a significant associa- compromises the treatment response. Future tar- tion with CD risk.35–38 To our knowledge, genetic geted sequencing studies on the THNSL2 gene variation at these two genes has not been previ- are warranted to confirm the involvement of the ously associated with anti-TNF response in CD. chr2:88173152 SNP TNF signaling in CD. Our results support that genetic variation at the HLA-B and HLA-DRB1 disease susceptibility loci Epigenetic mark enrichment analysis of the LoF could also modulate anti-TNF efficacy, and, variants indicates that the gut mucosa is an impor- therefore, these two genes could be two promising tant mediator of anti-TNF response in CD. biomarkers for future tailored TNF-blocking ther- Previous genetic studies have shown the impor- apy in CD patients. tance of the colonic mucosa in the regulation of the inflammatory processes that characterize the In the genome-wide screening for LoF variants, pathogenesis of CD.4 The gastrointestinal mucosa we have also identified functionally-relevant of CD patients is highly enriched in activated variants at nine genes, including NECAP2, ­ TNF-secreting macrophages.44–46 From a phar- HEYL, THNSL2, NPHP3-ACAD11, CLDN18, macological perspective, the anti-TNF therapy LY6G6C, C6orf226, RAB3IL1, and EFCAB5. has been shown to inhibit not only the gut inflam- CLDN18 encodes a tight junction protein from mation but also the inflammation-induced angio- the claudin family that regulates the integrity of genesis in the intestinal mucosa.47,48 Therefore, the intestinal epithelium.39 The transcriptomic our findings are consistent with the critical role of profile of claudins has been found to depend on the intestinal mucosa as the location where the disease activity, and to vary along the intestinal immune response is initiated in CD, and, there- mucosa.40 CLDN18 has been recently associated fore, as the key organ determining the clinical with the pathogenesis of IBD,35 where, by modu- response to anti-TNF therapy. lating the permeability of the epithelial barrier, it may allow the lamina propria immune system to In this study, we provide the first evidence of asso- encounter luminal antigens that otherwise would ciation between rare damaging variation in the journals.sagepub.com/home/tag 9 Therapeutic Advances in Gastroenterology 12 TNF signaling pathway and anti-TNF response. A index scores to ensure that the symptoms were due recent sequencing study in rheumatoid arthritis mainly to the activity of the disease. In fact, in clini- found that rare variants in this pathway do not con- cal practice, short-term response is commonly based tribute to treatment response.13 However, in this on patient symptoms. The sample size of the present latter study, only exomes were sequenced, and, study is low if compared with genome-wide studies thereby, important functional variation from other on common variants. However, compared with genomic regions was missed. Additionally, there is GWAS, here, we used WGS to identify all possible clear evidence that rheumatoid arthritis and CD rare variants. Subsequently, by focusing our analysis have very little genetic risk overlap. This also sug- only on rare variants with great functional impact, gests that different biological mechanisms operate we increase the power of our study. In this respect, through TNF signaling, and, consequently, drug this strategy is similar to that used to identify varia- efficacy is mediated by different groups of genes tions associated with rare Mendelian diseases, where and associated regulatory variants. a small number of patients have almost exclusive mutations. In this regard, our findings are not trivial The epigenetic enrichment analysis of the TNF at all, as among all genes, the only homozygous LoF signaling pathway indicated that CD8+ T-cells are homozygosity, a topmost damaging variation, important determinants of the efficacy of anti- occurred by chance only in HLA genes, which TNF agents. The inflamed mucosa of CD patients encode key antigen-presenting proteins and are carries large numbers of immune cells, including known risk factors in IBD. Finally, given our limited CD8+ T cells,49 which are the main cell type sample size, we restricted rare variation burden anal- responsible for destroying the intestinal epithe- ysis specifically to the anti-TNF pathway, thereby lium, leading to an enhanced exposure to the gut preventing a large penalty by testing and analyzing microbiota and the subsequent triggering of hundreds of other biological pathways. chronic inflammatory processes.50–52 As opposed to CD4+ T-cells, antigen presentation to CD8+ The analytical strategy implemented in this study T-cells is MHC-I restricted. Indeed, the is a distinct feature of the present work. In order rs41563412 variant located in an intronic region of to identify rare damaging variants underlying the the HLA-B gene is included within such an MHC-I anti-TNF response, we screened the whole family. Besides, the unique homozygous carrier of genome of anti-TNF-treated CD patients. the damaging HLA-B rs41563412-GCA variant Compared with array-based genome-wide stud- (frameshift mutation, p.Ser33LeufsTer9) did not ies, WGS efficiently captures rare variants that respond to anti-TNF therapy, suggesting a role of otherwise would have been missed.56 The analyti- this LoF SNP in modulating the CD8+ cytotoxic cal approach used for testing association of the immune response. The relation between CD8+ TNF signaling pathway is also radically different T-cells and the response to anti-TNF therapy has from single-marker tests. Single-marker tests are been also investigated in previous experimental underpowered to detect significant associations of studies.46 Using in vitro isolated CD8+ T cells, the low-frequency variants, unless very large sample administration of anti-TNF therapy was shown to sizes are used.57 The set-based collapsing analysis inhibit the proliferation of this T cell subtype both method used here, SKAT-O, groups low-fre- in the gut mucosa and in blood.53,54 Importantly, quency variants into gene sets in order to detect the transcriptomic profile of CD8+ T cells has significant overabundance of low-frequency vari- been associated with the incidence of being refrac- ation associated with complex traits at high statis- tory to disease treatment.55 Notably, in our study, tical power, despite using relatively small the relationship between CD8+ T-cells and anti- cohorts.27 Given the sample size of our study, we TNF response was restricted to the naïve, but not restricted the analysis to the key signaling path- the memory, subset of CD8+ T-cells, suggesting a way that is interfered with by anti-TNF drugs. primary defect on this cell fraction. Our findings Using this approach, we found a significant bur- are in line with these results, and support the pre- den of rare damaging variants associated with the dominant role of CD8+ T cells in the TNF- response to TNF alpha blockers in CD. dependent inflammatory activity in the gut. In conclusion, we report evidence that func- Our study has some limitations. First, the response tional low-frequency variation is associated with was categorized based on clinical activity indexes. the response to anti-TNF therapy in CD. These However, we included patients with high activity findings also provide new insights into the 10 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. underlying heterogeneity of CD, revealing the M Iborra has received funding for educational basis of TNF-dependent biological mechanisms. activities, research projects and scientific meet- The identification of this diversity at the molecu- ings from Takeda and Abbvie. lar level will be essential to allow the stratifica- JL Cabriada has served as consultant for and tion of patients, and, ultimately, help clinicians received research funding from MSD, Abbvie, choose the most appropriate drug. Pfizer and Kern Ph. Acknowledgments L Bujanda: none María Chaparro and Adria Aterido contributed equally as first authors. Antonio Julia and Javier C Taxonera has served as a speaker, a consultant Gisbert contributed equally as senior authors. and advisory member for, or has received research The authors thank Dr Simon Heath and Dr Raul funding from, MSD, Abbvie, Hospira, Pfizer, Tonda from the Centre Nacional d’Anàlisi Takeda, Ferring, Faes Farma, Shire Genòmica (CNAG, Barcelona, Spain) for their Pharmaceuticals, Dr Falk Pharma, Gebro Pharma. help in the WGS data processing. V García-Sánchez has served as a speaker, a con- sultant and advisory member for or has received Author contributions research funding from MSD, Abbvie, Hospira, María Chaparro and Javier P. Gisbert: study Kern Pharma, Takeda, Pfizer, Ferring, Faes Farma, design, data collection, data analysis, data inter- Shire Pharmaceuticals, Dr Falk Pharma, Janssen, pretation, writing the manuscript. Gebro Pharma and Otsuka Pharmaceutical. Antonio Julià and Adrià Aterido: Genetic analy- I Marín-Jiménez has served as a consultant, advi- sis, data analysis, data interpretation and manu- sory member, speaker, or has received research script writing. funding from Abbvie, Chiesi, Faes Farma, Falk- Borja Hernández-Breijo, Luis González Guijarro, Pharma, Ferring, Gebro Pharma, Hospira, Janssen, Alicia C. Marín and David Bernardo: Sample MSD, Otsuka Pharmaceutical, Pfizer, Shire, management. Takeda, Tillots, and UCB Pharma. Remaining authors: Study design, patient inclu- M Barreiro-de Acosta has served as a speaker, a sion, sample obtaining and data collection. consultant and advisory member for or has received research funding from MSD, Abbvie, Hospira, All authors approved the final version of the Pfizer, Kern Pharma, Biogen Takeda, Ferring, manuscript. Faes Farma, Shire Pharmaceuticals, Dr Falk María Chaparro and Javier P. Gisbert are the Pharma, Tillotts Pharma, Chiesi, Gebro Pharma, guarantors of the article. Otsuka Pharmaceutical and Vifor Pharma. I Vera has served as a speaker for AbbVie, Takeda Funding and Shire The author(s) disclosed receipt of the following MD Martín-Arranz has served as a speaker, a con- financial support for the research, authorship, sultant and advisory member for MSD; Abbvie, and/or publication of this article: This research Takeda, Janssen, Ferring, Faes Farma, Shire was funded by grants from the “Instituto de Salud Pharmaceutical, Tillots Pharma and Chiesi Carlos III” (FIS.12/02557 and PI13/00041). B.Hernández-Breijo: None Conflict of interest statement F.Mesonero: None M Chaparro has served as a speaker, or has received research or education funding from L Sempere has served as a speaker, a consultant MSD, Abbvie, Hospira, Pfizer, Takeda, Janssen, and advisory member for TAKEDA, ABBVIE, Ferring, Shire Pharmaceuticals, Dr Falk Pharma, KERN, MSD Y TILLOTTS. Tillotts Pharma. F Gomollón has received fees for lectures from I Guerra has served as a speaker for Pfizer and Janssen, Abbvie, Takeda, MSD and research grants Janssen and consultant and advisory member for (Group): MSD, Abbvie. Advisory Committees: Kern Pharma and MSD. None active at present. journals.sagepub.com/home/tag 11 Therapeutic Advances in Gastroenterology 12 J Hinojosa Abbvie, MSD, Takeda, Janssen, Kern, Supplemental material Pfizer-Hospira, Otsuka, Faes, Ferring, Biogen, Supplemental material for this article is available Shire. online. F Bermejo has served as a speaker, a consultant and advisory member for, or has received research funding from, MSD, Abbvie, Pfizer, Hospira, Takeda, Ferring, Faes Farma, Shire, Tillotts, References Chiesi and Gebro. 1. Hanauer SB, Feagan BG, Lichtenstein GR, et al. Maintenance infliximab for Crohn’s disease: the B Beltrán: None ACCENT I randomised trial. Lancet 2002; 359: 1541–1549. A Rodríguez Pescador 2. Present DH, Rutgeerts P, Targan S, et al. JM Banales: has served as a speaker and advisory Infliximab for the treatment of fistulas in patients member for OWL Metabolomics. with Crohn’s disease. N Engl J Med 1999; 340: 1398–1405. D Olivares: None 3. Hanauer SB, Sandborn WJ, Rutgeerts P, et al. P Aguilar-Melero: None Human anti-tumor necrosis factor monoclonal antibody (adalimumab) in Crohn’s disease: the L Menchén has served as a speaker, a consult- CLASSIC-I trial. Gastroenterology 2006; 130: ant and advisory member for or has received 323–333; quiz 591. research funding from MSD, Abbvie, Hospira, 4. Goyette P, Boucher G, Mallon D, et al. High- Pfizer, Takeda, Ferring, Faes Farma, Shire density mapping of the MHC identifies a shared Pharmaceuticals, Dr Falk Pharma, Tillotts and role for HLA-DRB1*01:03 in inflammatory General Electric. bowel diseases and heterozygous advantage in ulcerative colitis. Nat Genet 2015; 47: 172–179. R Ferreiro-Iglesias has received grants from Abbvie,MSD, Shire, Falk, Tillotts. 5. Baumgart DC and Sandborn WJ. Crohn’s disease. Lancet 2012; 380: 1590–1605. I Blazquez Gomez: None 6. Cottone M and Criscuoli V. Infliximab to treat Crohn’s disease: an update. Clin Exp Gastroenterol B Benítez García: None 2011; 4: 227–238. LG Guijarro: None 7. Reenaers C, Louis E and Belaiche J. Current directions of biologic therapies in inflammatory AC Marin: None bowel disease. Therap Adv Gastroenterol 2010; 3: D Bernardo: None 99–106. 8. Lopez-Hernandez R, Valdes M, Campillo JA, JP Gisbert has served as a speaker, a consultant et al. Genetic polymorphisms of tumour necrosis and advisory member for, or has received research factor alpha (TNF-alpha) promoter gene and funding from, MSD, Abbvie, Hospira, Pfizer, response to TNF-alpha inhibitors in Spanish Kern Pharma, Biogen, Takeda, Janssen, Roche, patients with inflammatory bowel disease. Int J Ferring, Faes Farma, Shire Pharmaceuticals, Dr Immunogenet 2014; 41: 63–68. Falk Pharma, Tillotts Pharma, Chiesi, Casen 9. Stokkers PC and Hommes DW. New cytokine Fleet, Gebro Pharma, Otsuka Pharmaceutical, therapeutics for inflammatory bowel disease. Vifor Pharma. Cytokine 2004; 28: 167–173. 10. Rufini S, Ciccacci C, Novelli G, et al. ORCID iDs Pharmacogenetics of inflammatory bowel disease: M Chaparro https://orcid.org/0000-0002- a focus on Crohn’s disease. Pharmacogenomics 9275-4242 2017; 18: 1095–1114. C Taxonera https://orcid.org/0000-0001- 11. Styrkarsdottir U, Helgason H, Sigurdsson A, 9166-7350 et al. Whole-genome sequencing identifies rare genotypes in COMP and CHADL associated JP Gisbert https://orcid.org/0000-0003-2090- with high risk of hip osteoarthritis. Nat Genet 3445 2017; 49: 801–805. 12 journals.sagepub.com/home/tag M Chaparro, A Aterido et al. 12. Tachmazidou I, Suveges D, Min JL, et al. Whole- 24. Lek M, Karczewski KJ, Minikel EV, et al. Genome sequencing coupled to imputation Analysis of protein-coding genetic variation in discovers genetic signals for anthropometric 60,706 humans. Nature 2016; 536: 285–291. traits. Am J Hum Genet 2017; 100: 865–884. 25. Eilbeck K, Lewis SE, Mungall CJ, et al. The 13. Cui J, Diogo D, Stahl EA, et al. Brief report: the sequence ontology: a tool for the unification of role of rare protein-coding variants in anti-tumor genome annotations. Genome Biol 2005; 6: R44. necrosis factor treatment response in rheumatoid 26. MacArthur DG, Balasubramanian S, Frankish arthritis. Arthritis Rheumatol 2017; 69: 735–741. A, et al. A systematic survey of loss-of-function 14. Bomba L, Walter K and Soranzo N. The impact variants in human protein-coding genes. Science of rare and low-frequency genetic variants in 2012; 335: 823–828. common disease. Genome Biol 2017; 18: 77. 27. Lee S, Emond MJ, Bamshad MJ, et al. Optimal 15. MacArthur DG and Tyler-Smith C. Loss-of- unified approach for rare-variant association function variants in the genomes of healthy testing with application to small-sample case- humans. Hum Mol Genet 2010; 19: R125–R130. control whole-exome sequencing studies. Am J Hum Genet 2012; 91: 224–237. 16. Gomollon F, Dignass A, Annese V, et al. 3rd European evidence-based consensus on the 28. Kanehisa M and Goto S. KEGG: Kyoto diagnosis and management of Crohn’s disease encyclopedia of genes and genomes. Nucleic Acids 2016: part 1: diagnosis and medical management. Res 2000; 28: 27–30. J Crohns Colitis 2017; 11: 3–25. 29. Trynka G, Sandor C, Han B, et al. Chromatin 17. Hernandez-Breijo B, Chaparro M, Cano- marks identify critical cell types for fine mapping Martinez D, et al. Standardization of the complex trait variants. Nat Genet 2013; 45: homogeneous mobility shift assay protocol 124–130. for evaluation of anti-infliximab antibodies. 30. Bernstein BE, Stamatoyannopoulos JA, Costello Application of the method to Crohn’s disease JF, et al. The NIH roadmap epigenomics patients treated with infliximab. Biochem mapping consortium. Nat Biotechnol 2010; 28: Pharmacol 2016; 122: 33–41. 1045–1048. 18. McKenna A, Hanna M, Banks E, et al. The 31. Sherry ST, Ward MH, Kholodov M, et al. genome analysis toolkit: a mapreduce framework dbSNP: the NCBI database of genetic variation. for analyzing next-generation DNA sequencing Nucleic Acids Res 2001; 29: 308–311. data. Genome Res 2010; 20: 1297–1303. 32. de Mattos BR, Garcia MP, Nogueira JB, et al. 19. Julia A, Tortosa R, Hernanz JM, et al. Risk Inflammatory bowel disease: an overview of variants for psoriasis vulgaris in a large case- immune mechanisms and biological treatments. control collection and association with clinical Mediators Inflamm 2015; 2015: 493012. subphenotypes. Hum Mol Genet 2012; 21: 4549–4557. 33. Roggenbuck D, Reinhold D, Baumgart DC, et al. Autoimmunity in Crohn’s disease-a putative 20. Price AL, Patterson NJ, Plenge RM, et al. stratification factor of the clinical phenotype. Adv Principal components analysis corrects for Clin Chem 2016; 77: 77–101. stratification in genome-wide association studies. Nat Genet 2006; 38: 904–909. 34. Matricon J, Barnich N and Ardid D. Immunopathogenesis of inflammatory bowel 21. Cingolani P, Platts A, Wang le L, et al. A program disease. Self Nonself 2010; 1: 299–309. for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the 35. Dotti I, Mora-Buch R, Ferrer-Picon E, genome of Drosophila melanogaster strain w1118; et al. Alterations in the epithelial stem cell iso-2; iso-3. Fly (Austin) 2012; 6: 80–92. compartment could contribute to permanent changes in the mucosa of patients with ulcerative 22. Adzhubei IA, Schmidt S, Peshkin L, et al. A colitis. Gut 2017; 66: 2069–2079. method and server for predicting damaging missense mutations. Nat Methods 2010; 7: 36. Arimura Y, Isshiki H, Onodera K, et al. 248–249. Characteristics of Japanese inflammatory bowel disease susceptibility loci. J Gastroenterol 2014; 23. Kumar P, Henikoff S and Ng PC. Predicting the 49: 1217–1230. effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat 37. Newman B, Silverberg MS, Gu X, et al. Protoc 2009; 4: 1073–1081. CARD15 and HLA DRB1 alleles influence journals.sagepub.com/home/tag 13 Therapeutic Advances in Gastroenterology 12 susceptibility and disease localization in Crohn’s CD40/CD40L pathway in the mucosal disease. Am J Gastroenterol 2004; 99: 306–315. microcirculation: a novel anti-inflammatory mechanism of infliximab in Crohn’s disease. 38. Silverberg MS, Mirea L, Bull SB, et al. A J Immunol 2006; 176: 2617–2624. population- and family-based study of Canadian families reveals association of HLA DRB1*0103 48. Rutella S, Fiorino G, Vetrano S, et al. Infliximab with colonic involvement in inflammatory bowel therapy inhibits inflammation-induced angiogenesis disease. Inflamm Bowel Dis 2003; 9: 1–9. in the mucosa of patients with Crohn’s disease. Am J Gastroenterol 2011; 106: 762–770. 39. Mitic LL, Van Itallie CM and Anderson JM. Molecular physiology and pathophysiology of 49. Monteleone G and Caprioli F. T-cell-directed tight junctions I. Tight junction structure and therapies in inflammatory bowel diseases. Clin Sci function: lessons from mutant animals and (Lond) 2010; 118: 707–715. proteins. Am J Physiol Gastrointest Liver Physiol 50. Nancey S, Holvoet S, Graber I, et al. CD8+ 2000; 279: G250–G254. cytotoxic T cells induce relapsing colitis in 40. Lameris AL, Huybers S, Kaukinen K, et al. normal mice. Gastroenterology 2006; 131: Expression profiling of claudins in the human 485–496. gastrointestinal tract in health and during 51. Westendorf AM, Fleissner D, Deppenmeier S, et al. inflammatory bowel disease. Scand J Gastroenterol Autoimmune-mediated intestinal inflammation- 2013; 48: 58–69. impact and regulation of antigen-specific CD8+ T 41. Mallya M, Campbell RD and Aguado B. cells. Gastroenterology 2006; 131: 510–524. Characterization of the five novel Ly-6 52. Steinhoff U, Brinkmann V, Klemm U, et al. superfamily members encoded in the MHC, Autoimmune intestinal pathology induced by and detection of cells expressing their potential hsp60-specific CD8 T cells. Immunity 1999; 11: ligands. Protein Sci 2006; 15: 2244–2256. 349–358. 42. Bradley JR. TNF-mediated inflammatory disease. 53. Dahlen R, Strid H, Lundgren A, et al. Infliximab J Pathol 2008; 214: 149–160. inhibits activation and effector functions of 43. Rifas L and Weitzmann MN. A novel T cell peripheral blood T cells in vitro from patients cytokine, secreted osteoclastogenic factor of with clinically active ulcerative colitis. Scand J activated T cells, induces osteoclast formation in Immunol 2013; 78: 275–284. a RANKL-independent manner. Arthritis Rheum 54. Baert FJ, D’Haens GR, Peeters M, et al. 2009; 60: 3324–3335. Tumor necrosis factor alpha antibody 44. Breese EJ, Michie CA, Nicholls SW, et al. Tumor (infliximab) therapy profoundly down-regulates necrosis factor alpha-producing cells in the the inflammation in Crohn’s ileocolitis. intestinal mucosa of children with inflammatory Gastroenterology 1999; 116: 22–28. bowel disease. Gastroenterology 1994; 106: 55. Lee JC, Lyons PA, McKinney EF, et al. Gene 1455–1466. expression profiling of CD8+ T cells predicts 45. Lala S, Ogura Y, Osborne C, et al. Crohn’s prognosis in patients with Crohn disease and disease and the NOD2 gene: a role for paneth ulcerative colitis. J Clin Invest 2011; 121: cells. Gastroenterology 2003; 125: 47–57. 4170–4179. 46. Billmeier U, Dieterich W, Neurath MF, et al. 56. Cirulli ET and Goldstein DB. Uncovering the Molecular mechanism of action of anti-tumor roles of rare variants in common disease through necrosis factor antibodies in inflammatory whole-genome sequencing. Nat Rev Genet 2010; bowel diseases. World J Gastroenterol 2016; 22: 11: 415–425. Visit SAGE journals online 9300–9313. journals.sagepub.com/ 57. Lee S, Abecasis GR, Boehnke M, et al. Rare- home/tag 47. Danese S, Sans M, Scaldaferri F, et al. variant association analysis: study designs and SAGE journals TNF-alpha blockade down-regulates the statistical tests. Am J Hum Genet 2014; 95: 5–23. 14 journals.sagepub.com/home/tag