University of Groningen Inflammatory Bowel Disease Visschedijk, Marijn

(1)

University of Groningen

Inflammatory Bowel Disease

Visschedijk, Marijn

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from

it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date:

2018

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Visschedijk, M. (2018). Inflammatory Bowel Disease: 'New genes, rare variants & moving towards clinical

practice'. Rijksuniversiteit Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

521405-L-sub01-bw-Visschedijk

Processed on: 23-7-2018 Processed on: 23-7-2018 Processed on: 23-7-2018

Processed on: 23-7-2018 PDF page: 95PDF page: 95PDF page: 95PDF page: 95

Rudi Alberts*, Marijn C Visschedijk*, Sören Mucha, David Ellinghaus, Annika Bergquist, Jonas Halvarsson, Severine Vermeire, Brian Juran, The International PSC Study Group, Tom H Karlsen, Ca Anderson, Andre Franke, Eleonora A.M. Festen, Rinse K Weersma.

Manuscript in preparation *These authors contributed equally

Low-frequency and rare DNA

sequence variants associated with

primary sclerosing cholangitis

susceptibility

(3)

521405-L-sub01-bw-Visschedijk

96

ABSTRACT

Primary sclerosing cholangitis (PSC) is a rare, progressive inflammatory disease of the bile duct. There is currently no medical treatment except liver transplantation and patients are often young. The etiopathogenesis is still largely unknown. The hypothesis of occurrence of PSC is a genetic susceptible host triggered by environmental factors. PSC is considered as a genetically complex disease. Several genome wide association studies have found over 24 genetic loci that are associated with the susceptibility of PSC. These loci occur commonly in the general population. It is known that low-frequency and rare genetic variants are more population specific than common variants and that they often have larger effect sizes (ie, high odds ratios) that affect protein function. To study whether low-frequency and rare genetic variants are associated with susceptibility of PSC, we here used Illumina HumanExome BeadChips to genotype a large international cohort of 1,243 PSC cases and 10,038 population matched controls. Association analyses revealed 90 rare genetic loci associated with PSC. Exome array genotyping results were validated using Sanger sequencing on the same samples as used in the discovery stage. We identified 5 gene-wide low-frequency variants associated with PSC. Identified variants in the discovery phase are currently being analyzed in four independent cohorts of 2401 PSC cases and 5088 population matched controls. Conclusions: We performed a large international exome array study on PSC susceptibility and identified several rare genetic loci that are associated with PSC.

Replication of these variants is ongoing, results are expected soon.

INTRODUCTION

Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease characterized by inflammation and fibrosis of both intrahepatic and extrahepatic bile ducts1_{. In}

the majority of patients the progressive fibrosis ultimately leads to biliary cirrhosis and eventually hepatic failure. Currently, there is no curative therapy available, and oftentimes PSC patients with hepatic failure requiring liver transplantation2,3_{. It is}

(4)

521405-L-sub01-bw-Visschedijk

97

known that most PSC patients also have a concomitant form of inflammatory bowel disease (IBD). Around 80 percent of PSC patient with IBD actually have ulcerative colitis (UC), whereas the remaining 20 percent have Crohn’s disease (CD) or unclassified IBD4,5_.

The etiology of PSC is characterized by the interplay between genetic susceptible host triggered by environmental factors and microbiome. Genome-wide association studies (GWAS) have greatly expanded the knowledge of the pathogeneses of many complex diseases. Three GWAS have identified eight risk loci for PSC susceptibility outside the HLA complex 6–8_{. An International PSC Study Group (IPSCSG) study genotyped over}

3,000 PSC patients with the Immunochip genotyping array and increased the number of non-HLA risk loci to sixteen9_{. Since then, a cross disease meta-analysis identified four}

additional PSC loci10_{and a recent GWAS study identified another four risk loci for PSC}11_,

bringing the total number of associated loci to 24.

Despite large-scale efforts of GWAS studies, a large part of PSC heritability is still unexplained. Part of the “hidden” heritability is thought to reside in low-frequency (1% - 5%) or rare (<1%) exonic variants with greater effect sizes, compared to the common variants that have been identified by regular GWAS or Immunochip studies. To study low frequency and rare variants, the HumanExome BeadChips (Illumina, Inc., San Diego, CA) was designed. This exome array is specifically designed as an intermediate between GWAS genotyping arrays, which only focus on relatively common variants, and exome sequencing of very large numbers of samples 12_{. The exome array aims to}

include coding variants seen several times in existing sequence datasets and is based on information on ~12,000 sequenced genomes and exomes at the time of construction. In total the exome array includes ~250,000 non-synonymous variants, splice variants and stop-altering variants. Several exome array studies have investigated the effect of low-frequency and rare variants on different phenotypes13–16_{. A study in insulin}

secretion identified low-frequency coding variants associated with fasting proinsulin concentrations within previously identified susceptibility loci and three new genes with low-frequency variants associated with fasting proinsulin or insulinogenic index showing the efficacy of the approach. Furthermore, by using the same exome array five novel low frequency variants were identified as associated with heart rate15_{and two}

low frequency variants associated with acute anterior uveitis16_{. However a study into}

myocardial infarction remained failed to identify positive results reflecting the difficulty in assessing rare variant associations in complex disease14_.

(5)

521405-L-sub01-bw-Visschedijk

98

The aim of the current study is to identify low frequency and rare exonic variants that are associated with PSC susceptibility. We performed exome genotyping in a large international cohort of 1,243 PSC cases and 10,038 population matched controls. Next, we validated our findings using sanger sequencing and aimed to confirm our findings in independent cohort of PSC cases and controls. Replication in independent cohorts is ongoing. In total we will have data of 3,644 PSC cases and 15,126 population-matched controls in this study.

MATERIALS AND METHODS

PSC patient samples and controls

PSC patients and population-matched controls were recruited from hospitals throughout the Netherlands, Germany, Norway, Belgium, Sweden, the UK and the US. Subject recruitment was approved by the ethics committees or institutional review boards of all participating centres. Written informed consent was obtained from all participants. PSC diagnosis was based on clinical, biochemical, cholangiographic and histological criteria, as formulated by the European Association for the Study of the Liver guidelines17_.

STAGE 1: DISCOVERY, POPULATION AND GENOTYPING

In Stage 1, a total of 1,243 PSC patients were genotyped using Illumina HumanExome-12 v1.1 BeadChip arrays. This cohort consisted of 614 PSC cases from Germany, 255 cases from the Netherlands and 374 cases from Norway. Population matched exome array data was obtained for a total of 10,038 population matched controls, among which 7,026 from Germany, 2,594 from the Netherlands and 418 from Norway (Supplementary Table S1). The population-matched controls were genotyped on the Illumina HumanExome-12 v1.0 array. The v1.0 and v1.1 exome arrays interrogate 247,870 respectively 242,901 exonic genetic variants (figure 1).

Because rare variants are per definition not often observed and to obtain optimal genotype calling, it is advised to perform genotype calling, i.e. assigning genotypes to individuals based on signal intensity plots, in as large a dataset as possible 18_{. For this we}

performed joint genotype calling in four batches for a total of 27,912 exome arrays that were used to study rare variants in PSC, Crohn’s disease, Atopic dermatitis, psoriasis, Coronary artery disease and Common Variable Immunodeficiency. Genotype calling

(6)

521405-L-sub01-bw-Visschedijk

99

was performed using GenomeStudio version 11.0. After genotype calling we extracted the PSC cases and controls and performed the quality control steps mentioned in the next paragraph. Next, we applied the zCall procedure 19_{on the same large set of 27,912}

samples This dataset was used for association analysis and consists of 1,110 PSC cases, 9,412 population matched controls and 142,319 genetic variants.

Sample quality control consisted of four steps. First, samples with a call rate below 98% were removed (n=441). Next, samples with reported sex that was inconsistent with the sex observed in the genotype data were reassigned the sex observed in the genotype data (n=1). Then, we iteratively removed duplicated individuals after calculating estimates of pairwise Identity By Descent using the --genome option in PLINK 20_{with a threshold}

of 0.18 (n=280). For each pair we removed the one with the lowest genotyping call rate. Finally, ethnic outliers were removed by calculating the first two principal components of all samples of the 1000 Genomes Project21_{and projecting them onto our cases and}

controls (n=28)22_.

Figure 1

Figure 1. This flowchart shows an overview of the study. The figure includes the different types

of experiments and the amount of PSC cases and healthy control samples used. The Discovery stage (stage 1) and the validation stage is completed. The red marked Replication stage (stage

(7)

521405-L-sub01-bw-Visschedijk

100

Variant quality control involved removing SNPs with a call rate <98% (n=10,662), removing SNPs with Hardy-Weinberg Equilibrium below 1.0 × 10-5 _{(n=913) and removing}

SNPs with a minor allele frequency < 0.00001 (n=98,027). This last group consists of only monomorphic SNPs.

Statistical analyses

Stage 1, the discovery stage, comprised of single marker association analysis using the MLME-LOCO method of the GCTA software 23_{. Genotype intensity plots of all}

identified associations (p-value < 5.0 × 10-05_{) were visually inspected and kept if overall}

signal intensities were high enough and cluster separation was deemed clear.

Genome-wide significant SNPs (P < 5.0 × 10-08_{) were selected for replication. Also, SNPs}

associated with a P-value < 0.05 and co-localizing with previously reported GWAS loci for PSC susceptibility were selected. Finally, we included suggestively (P < 5.0 × 10 -05_{) associated nonsense or splice site variants and variants having a deleterious effect}

according to PolyPhen2 24_{or SIFT}25_{. In total 90 SNPs were included for replication.}

Power calculations were performed using the statistical power calculator CaTS 26_{. Power}

was calculated for the two-stage design and we assumed a disease prevalence of 13 in 100,000.

Validation of exome array data

Most of the identified variants in the discovery stage are either low frequency or rare variants. For those variants we either observed individuals homozygous for the common allele, or heterozygous. To validate the positive findings of the discovery stage associations found using exome array data, we selected heterozygous cases and controls for 90 SNPs identified by the association analysis. We used 176 cases and 86 population matched controls. We designed primers for Sanger sequencing using Primer3web version 4.0.0 (http://bioinfo.ut.ee/primer3/). We used the Task setting “generic” and we set the Mispriming Library to “human”. The product size was set to 200-500 and for the rest default values were used. Primer sequences are reported in Supplementary Table S2.

(8)

521405-L-sub01-bw-Visschedijk

101

STAGE 2: REPLICATION, POPULATION AND GENOTYPING

(ONGOING)

In stage 2, we aim to replicate SNPs identified in Stage 1 in four independent cohorts. 87 out of the 90 variants fitted the design of the iPLEX Agena Bioscience assay (Agena Biosciences, San Diego, CA, USA) (Supplementary Table S3). We will genotype 87 genetic variants in 131 PSC Belgian cases and 1678 population matched controls, 248 Swedisch PSC cases and 990 population matched controls, 946 US PSC cases and 363 population matched controls and 1076 UK PSC cases and 1005 population matched controls. All Belgian, Swedish and US samples will be genotyped at the Institute of Clinical Molecular Biology in Kiel, Germany, using the iPLEX Agena Bioscience MassARRAY® platform. The UK samples will be genotyped at the Sanger Institute in Cambridge using the same Agena Bioscience assay design (Supplementary Table S3). To this dataset we will add another dataset of 1,051 exome arrays of population-matched controls from the US downloaded from the database of Genotypes and Phenotypes (dbGaP). This data originates from the Exome Chip Study of National Institute of Mental Health (NIMH) Controls and the dbGaP Study Accession is phs000630.v1.p1.

RESULTS

STAGE 1 Discovery, Exome array genotyping

After quality control, the dataset consists of 141,732 SNPs, 988 PSC cases and 8776 population-matched controls.

The majority of SNPs after QC is rare (107,023). Nonetheless, there were 25,604 common variants and 9,015 low frequency variants. In this study we had 80% power to detect variants with an odds ratio > 2.09 or > 1.43 for variants with a minor allele frequency of 1% and 5%, respectively (Figure 2). Figure 3 shows the amount of SNPs in several MAF ranges after quality control.

(9)

521405-L-sub01-bw-Visschedijk

102

Figure 2

Figure 2 Power to detect variants with a MAF of 1% or 5% using the sample sizes in our study

and using the two-stage design. Power calculations were performed using the statistical power calculator CaTS26_.

MAF: Minor Allele Frequencies

Figure 3

Figure 3 The distribution of SNPs in the exome array data for different MAF strata. This figure

shows the amount of SNPs on the exome array in different MAF strata, ranging from monomor-phic SNP to low frequency (1%-5%) and common SNPs (>5%).

(10)

521405-L-sub01-bw-Visschedijk

103 SNP selection

To identify genetic variants associated with PSC susceptibility we performed a case control association analysis using mixed linear models (Methods). Table 1 shows five genetic variants that were found to be genome-wide significant in the discovery phase.

The prioritization of other interesting variants was based on different criteria. We compared established PSC risk loci with the associations identified in this study. Of 23 established risk loci outside the MHC complex 7,9–11,27,28_{, five were}

directly genotyped by the exome array and four had a proxy SNP within 500 kb and in LD with the reported risk SNP (r2_{> 0.8). All nine SNPs were nominally}

significant in this study (Supplementary Table S4). Among them, two had P values below 1.0 × 10-04_{. One SNP, rs3197999, was}

genome-wide significant (P = 6.64 × 10-09_,

OR = 1.43, RAF cases 0.35, RAF controls 0.27). This SNP resides in the gene MST17_.

Finally, we included suggestively (P < 5.0 × 10-05_{) associated nonsense or splice site}

variants, in total eight SNPs. Additionally, 68 variants with a P < 5.0 × 10-05_{having a}

deleterious effect according to PolyPhen2

24_{or SIFT} 25_{. In total 90 SNPs were}

included for replication.

Table 1 Allelic association r esults SNP ID Chr P osition Hg19 _(bp) Gene Ref Alt E xonic F unction

Amino Acid Chang

e MAF cases MAF contr ols P-value rs146913158 7 23018030 FAM126A T C MISSENSE p. Tyr64Cys 0.00042 0 6.64 x 10 -13 rs192682820 22 46652892 PKDREJ C A MISSENSE p. V al2110Leu 0.00028 0.00011 2.46 x 10 -11 rs143898891 15 71305200 LRRC49 G A MISSENSE p.Ala556Thr 0.00028 0.000054 8.07 x 10 -11 rs144169203 2 150294223 LYPD6 A G ST AR T L OST p.Me t1? 0.00014 0 2.09 x 10 -08 rs145979886 4 13601977 BOD1L1 C A MISSENSE p. V al2183Phe 0.00088 0.00038 4.27 x 10 -08

Table 1 Allelic association analysis r

esults (linear mix

ed models) at g

enome wide significance ((P < 5.0 × 10

-08

) f

or the 142,319 SNPs in the disco

very cohort (1,243 PSC

patients v

ersus 10,038 population mat

ched contr

ols).

Bp; base position, Ref; r

ef

er

ence allele, Alt; alt

ernativ

e allele, MAF; minor allele fr

equency

.

(11)

521405-L-sub01-bw-Visschedijk

104

The impact of rare variants can be large because their effect sizes are larger than those of common variants 29_{. In our dataset we can indeed confirm that effect sizes are larger}

for rare variants as compared to common variants. Rare variants that are suggestively associated with PSC have an average absolute effect size of 0.44 (OR = 1.55) or higher, and Table 2 shows that when increasing the MAF range, the average absolute effect size gradually decreases until 0.04 (OR = 1.04) for common variant.

Table 2 Effect sizes and odds ratios for SNPs in different minor allele frequency ranges.

MAF range Average absolute

effect size Corresponding Odds Ratio All SNPs after QC (n=142,319) <0.01% 0.16 1.17 0.01% - 0.1% 0.083 1.09 0.1% - 1% 0.03 1.03 1% - 5% 0.011 1.01 > 5% 0.0056 1.01 SNPs associated to PSC <0.01% 0.91 2.48 with P<5.0×10-06 (n=1,125) 0.01% - 0.1% 0.52 1.68 0.1% - 1% 0.44 1.55 1% - 5% 0.12 1.13 > 5% 0.04 1.04

Table 2 Effect sizes and odds ratios for SNPs in different minor allele frequency ranges. This table shows the average absolute effect size calculated using the MLMA-LOCO method for all SNPs after QC and also for SNPs that are suggestively associated with PSC at a P value threshold of 5.0 × 10-06_.

MAF; minor allele frequency, QC; quality control.

Validation of exome array data

To validate the exome array genotypes, we performed Sanger sequencing on a subset of 90 SNPs of heterozygous case or control samples. For validation we used the same DNA samples that were used in the discovery stage. In total, we used samples of 176 cases and 86 healthy control and we performed 308 sequencing assays (Supplementary Table 2). After removing 18 assays with low sequence quality, we found a concordance between Sanger sequencing and exome array data of 97.2% (282 out of 290 assays were concordant) showing high reliability of the exomechip genotyping.

(12)

521405-L-sub01-bw-Visschedijk

105

DISCUSSION

Here we present a large international case control association study on PSC susceptibility focusing on low frequency and rare genetic variants residing in the exons of genes. We have used Illumina HumanExome BeadChips to discover genetic variants associated to PSC susceptibility in three European populations. Furthermore, we validated the associations using the same samples, using Sanger sequencing technique. Identified variants in the discovery phase are currently being analyzed for replication in four independent cohorts from other countries.

Several studies have shown that rare genetic variants often have larger effect sizes and often are more population specific than common genetic variants, due to the variants being younger and undergoing different selective pressures29,30_.

Unfortunately, the amount of identified variants is much lower then expected and probably explain less of the hidden heritability then previously anticipated and hoped for31_{.Rare variants identification has proven to be complex in the past years. Furthermore,}

correction for population stratification and imputation are more challenging than for common variants29_{. To achieve appropriate statistical power thousands of individuals}

are needed. Hence, for rare diseases like PSC, rare variant discovery is even more difficult. Still, the latest WGS detected a burden of very rare, damaging missense variants in known Crohn’s disease risk genes, suggesting that more comprehensive sequencing studies will continue to improve understanding of the biology of complex diseases32_.

Selecting a more specific genotype, ea driven by very early onset variants, helps to detect the important functionally important variants for the disease.

In the past years, a growing body of literature describes the evolving picture of PSC without IBD phenotype and the distinct IBD-PSC phenotype33_{. We think that this}

PSC phenotype could be an end stage disease of the inflammatory process of the bile ducts that could develop in different ways, e.g. phenocopy of different diseases. It is hypothesized that this specific PSC could be driven by rare functionally important genomic variants (eg, very early onset or familial) that eventually result in a PSC-like phenotype.

In our exome wide analyses we here identified five rare SNPs with a exome-wide significantly associated with PSC susceptibility. The first SNP (rs146913158, P = 6.64 x 10-13_{) is a missense variant in the gene FAM126A (Family With Sequence Similarity 126}

(13)

521405-L-sub01-bw-Visschedijk

106

Member A) and has a deleterious effect according to SIFT and is probably damaging according to PolyPhen2. The protein encoded by FAM126A may play a role in the beta-catenin/Lef signaling pathway and expression of this gene is downregulated by beta-catenin. The second SNP, rs192682820, is also a missense variant and resides in the PKDREJ gene. This intronless gene encodes a member of the polycystin protein family and is involved in human reproduction. rs143898891 is a missense variant in

LRRC49. This gene is part of the neuronal tubulin polyglutamylase complex. The fourth

SNP, rs144169203 on chromosome 2, is a start lost variant in the LYPD6 gene with high impact, is deleterious according to SIFT and probably damaging according to Polyphen.

LYPD6 (LY6/PLAUR Domain Containing 6) modulates nicotinic acetylcholine receptors

function in the brain and acts as a positive regulator of Wnt/beta-catenin signaling by similarity. The fifth SNP, rs145979886, is a missense variant in BOD1L1 (Biorientation of Chromosomes In Cell Division 1 Like 1), with deleterious effect according to SIFT and benign according to PolyPhen2 and is involved in DNA repair

We compared the PSC risk loci identified in this study with earlier reported risk loci. The strongest association found in our discovery stage, for SNP rs3197999 residing in the MST1 gene (P = 6.64 × 10-09_{, OR = 1.43, RAF cases/controls 0.347/0.272) corresponds}

to the strongest association found in Melum et al.7_{, the same SNP with a reported P}

value of P = 2.45 × 10-26_{. Also, the risk allele frequencies and the direction of effect are}

very similar.

The detection of variants with very low frequencies, some SNPs were only detected in 1 or 2 individuals, is suspicious for false-positive findings. This is the first study that performed a validation of the exome array by Sanger sequencing, which resulted in a rather good concordance between Sanger sequencing and exome array data of 97.2%. We can conclude that our genotyping method is valid.

We performed a large international study on low-frequency and rare genetic variants associated to PSC susceptibility. We identified five SNPs with an exome-wide significantly associated with PSC susceptibility and confirmed a variant in the MST1 gene, already known to be associated with PSC. Since genetic association studies need robust replication, specifically for rare variants to avoid spurious associations, we are in the process of replication genotyping of the identified 87 variants in an independent cohort of 2401 PSC cases and 5088 population matched controls. Results are expected soon.

(14)

521405-L-sub01-bw-Visschedijk

107

REFERENCES

1. Chapman, R. W. et al. Primary sclerosing cholangitis: a review of its clinical features, cholangiography, and hepatic histology. Gut 21, 870–877 (1980). 2. Broome, U. et al. Natural history and prognostic factors in 305 Swedish patients

with primary sclerosing cholangitis. Gut 38, 610–615 (1996).

3. Card, T. R., Solaymani-Dodaran, M. & West, J. Incidence and mortality of primary sclerosing cholangitis in the UK: a population-based cohort study. J

Hepatol 48, 939–944 (2008).

4. Fausa, O., Schrumpf, E. & Elgjo, K. Relationship of Inflammatory Bowel Disease and Primary Sclerosing Cholangitis. Seminars in Liver Disease 11, 31–39 (1991). 5. Weismüller, T. J. et al. Patient Age, Sex, and Inflammatory Bowel Disease

Phenotype Associate With Course of Primary Sclerosing Cholangitis.

Gastroenterology 152, 1975--1984.e8 (2017).

6. Karlsen, T. H. et al. Genome-wide association analysis in primary sclerosing cholangitis. Gastroenterology 138, 1102–1111 (2010).

7. Melum, E. et al. Genome-wide association analysis in primary sclerosing cholangitis identifies two non-HLA susceptibility loci. Nat. Genet. 43, 17–19 (2011).

8. Janse, M. et al. Three ulcerative colitis susceptibility loci are associated with primary sclerosing cholangitis and indicate a role for IL2, REL, and CARD9.

Hepatology 53, 1977–1985 (2011).

9. Liu, J. Z. et al. Dense genotyping of immune-related disease regions identifies nine new risk loci for primary sclerosing cholangitis. Nat. Genet. 45, 670–5 (2013). 10. Ellinghaus, D. et al. Analysis of five chronic inflammatory diseases identifies

27 new associations and highlights disease-specific patterns at shared loci. Nat.

Genet. 48, 510–8 (2016).

11. Ji, S. et al. Genome-wide association study of primary sclerosing cholangitis identifies new risk loci and quantifies the genetic relationship with inflammatory bowel disease. 49, (2017).

12. NHLBI Exome Sequencing Project. No Title.

13. Huyghe, J. R. et al. Exome array analysis identifies new loci and low-frequency variants influencing insulin processing and secretion. Nat. Genet. 45, 197–201 (2013).

14. Holmen, O. L. et al. No large-effect low-frequency coding variation found for myocardial infarction. Hum. Mol. Genet. 23, 4721–4728 (2014).

15. van den Berg, M. E. et al. Discovery of novel heart rate-associated loci using the Exome Chip. Hum. Mol. Genet. 00, 1–18 (2017).

(15)

521405-L-sub01-bw-Visschedijk

108

16. Robinson, P. C. et al. Exome-wide study of ankylosing spondylitis demonstrates additional shared genetic background with inflammatory bowel disease. npj

Genomic Med. 1, 16008 (2016).

17. Lennard-Jones, J. E. Classification of inflammatory bowel disease. Scand J

Gastroenterol 170, 16–19 (1989).

18. Grove, M. L. et al. Best Practices and Joint Calling of the HumanExome BeadChip: The CHARGE Consortium. PLoS One 8, (2013).

19. Goldstein, J. I. et al. Zcall: A rare variant caller for array-based genotyping.

Bioinformatics 28, 2543–2545 (2012).

20. Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81, 559–575 (2007).

21. The 1000 Genomes Project Consortium. A global reference for human genetic variation. Nature 526, 68–74 (2015).

22. Liu, J. Z. et al. Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations. Nat. Genet. 47, 979–989 (2015).

23. Yang, J., Zaitlen, N. A., Goddard, M. E., Visscher, P. M. & Price, A. L. Advantages and pitfalls in the application of mixed-model association methods. Nat. Genet. 46, 100–106 (2014).

24. Adzhubei, I. A. et al. A method and server for predicting damaging missense mutations. Nat. Methods 7, 248–249 (2010).

25. Kumar, P., Henikoff, S. & Ng, P. C. Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat. Protoc. 4, 1073–1081 (2009).

26. Skol, A. D., Scott, L. J., Abecasis, G. R. & Boehnke, M. Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association studies. Nat. Genet. 38, 209–213 (2006).

27. Folseraas, T. et al. Extended analysis of a genome-wide association study in primary sclerosing cholangitis detects multiple novel risk loci. J. Hepatol. 57, 366–375 (2012).

28. Ellinghaus, D. et al. Genome-wide association analysis in primary sclerosing cholangitis and ulcerative colitis identifies risk loci at GPR35 and TCF4.

Hepatology 58, 1074–1083 (2013).

29. Bomba, L., Walter, K. & Soranzo, N. The impact of rare and low-frequency genetic variants in common disease. Genome Biol. 18, 77 (2017).

30. Rivas, M. A. et al. Deep resequencing of GWAS loci identifies independent rare variants associated with inflammatory bowel disease. Nat. Genet. 43, 1066–73 (2011).

(16)

521405-L-sub01-bw-Visschedijk

109

31. Hunt, K. A. et al. Negligible impact of rare autoimmune-locus coding-region variants on missing heritability. Nature 498, 232–235 (2013).

32. Luo, Y. et al. Exploring the genetic architecture of inflammatory bowel disease by whole genome sequencing identifies association at ADCY7. bioRxiv 49, 058347 (2016).

33. Weersma, R. K. & Lindor, K. D. Shifting Paradigms: What Is the True Prevalence and Clinical Course of Primary Sclerosing Cholangitis? Gastroenterology 151, 590–593 (2016).