Genetic predictors of benign course of ulcerative colitis –a North American IBD Genetics Consortium study

Abstract

Background and aims

A subset of patients with ulcerative colitis (UC) have a benign course and an overall favorable prognosis. Early identification of these low risk patients may allow for a less aggressive therapeutic approach and possible reduction of therapy-associated risks. The aim of this project was to identify the genetic predictors of benign UC phenotype.

Methods

UC patients were selected from the NIDDK Inflammatory Bowel Disease Genetics Consortium (NIDDK IBDGC). Benign phenotype was defined as no need for immunomodulatory or biologic therapy, hospitalizations or colectomy. The association between benign ulcerative colitis phenotype and known loci linked to the risk of inflammatory bowel disease was evaluated. The results for 156 index SNPs from the known IBD loci were extracted for the main analysis. The association of the benign phenotype to a genetic burden score was also evaluated.

Results

None of the index SNPs from the IBD loci reached the predefined threshold of 1×10⁻⁴. In the exploratory analysis of the remaining Immunochip SNPs and imputed MHC data, 5 distinct suggestive association signals are identified (rs1697950, rs2523639, rs17836409, rs11742854 and rs75001121)

Conclusions

No SNPS from IBD susceptibility loci were found to be associated (at our predefined threshold of 1×10⁻⁴) with a benign UC disease course. The rs11742570 variant on chromosome 5 was the one with the greatest association to benign disease although the association did not reach the predefined significant threshold. Given the modest power our study, the findings suggested on the exploratory analysis merit extension to larger discovery cohorts.

Introduction

Ulcerative colitis (UC) is characterized by a lifelong intermittent course. Up to 30% of UC patients will require surgery within the first 10 years of disease¹. Up to 50% of the patients will require treatment with immunomodulators or biologics. However, for a subset of patients with UC, their disease is characterized by a mild long-term course not requiring use of immunomodulators nor surgery. In a large cohort of Danish UC patients, the percent of patients not requiring active treatment gradually decreased over the years, although 90% of the patients had a relapsing course after 25 years². Early identification of patients with aggressive disease course may allow us to prevent complications by early aggressive intervention. Conversely, patients identified early on in their disease as having a benign course can be spared intensive medical therapy, which has its inherent risks. Clinical predictors of disease course in UC have been extensively studied. Extensive endoscopic involvement, high ESR, Hb <10.5 g/dL, and younger age of onset were associated with colectomy in a recent Norwegian study³. In another cohort study, extensive disease and family history of IBD were associated with refractory disease⁴. In a study from Toronto, disease severity was associated with a diagnosis before age of 40, proximal disease, rate and time to proximal extension and prednisone use⁵.

Clinical predictors do not provide reliable data to allow early stratification of UC patients. Several studies attempted to identify genetic predictors of severe clinical course of UC. In a recent genome-wide association (GWAS) study, a contribution of the major histocompatibility complex and the TNFSF15 (TL1A) locus to the risk of severe UC was reported⁴. In other studies, polymorphisms in the Il1B gene⁶, ABCB1/MDR1⁷, HSP70⁸, HLA-DR2⁹. In a recent multinational study, Jostins et al (2012) reported the discovery of more than 160 genetic variants associated to IBD. The majority of these variants were associated with both Crohn’s disease and UC¹⁰. None of these variants was associated with severe UC phenotype in a recent Canadian cohort study⁵. A recent large cohort multinational study from the International Inflammatory Bowel Disease Genetics Consortium identified several SNPs in the MHC region (6p21) (rs3115674, rs3129891, rs9268832, rs482044, rs77005575), as well as one SNP in the NOD2 region (16q12) (rs2066847) to be associated with the risk of colectomy.¹¹

The aim of our study was to evaluate the clinical and genetic predictors of a benign disease behavior in a large and well-characterized multicenter cohort of UC patients.

Methods

Study population

We searched the patient database of the National Institute of Diabetes and Digestive and Kidney Diseases Inflammatory Bowel Disease Genetics Consortium (IBDGC) repository. This is a large database of patients with diagnosis of inflammatory bowel disease (IBD) recruited from the United States, Puerto Rico and Canada. Phenotyping was performed by clinicians with experience in managing patients with IBD and was conducted according to an established Phenotype Operating Manual. IBD phenotype classification using this NIDDK IBDGC protocol has been validated¹². Only patients with an established diagnosis of ulcerative colitis were included.

Disease course

Patient data was collected cross-sectionally as per date of inclusion in the database. Benign disease course was identified as: 1) no colectomy for uncontrolled disease 2) no UC-related hospitalizations 3) no use of immunomodulators and biologics through the course of follow-up ^5,13. Patients not fitting the aforementioned criteria were characterized as having a non-benign course. Patients with incomplete clinical data regarding medical treatment, hospitalizations or surgery were excluded from the analysis. We included patients with follow-up duration of ≥ 10 years. Patient with a duration of disease of < 10 were included if they required surgery, immunomodulator treatment or surgery since the initial diagnosis. Patients who underwent colectomy for colorectal cancer or dysplasia were also excluded.

Genotype association

The main objective was to evaluate association of the benign ulcerative colitis phenotype to the 163 loci identified for IBD ¹⁰. For this purpose, a first analysis was performed with benign vs non-benign as a dichotomous trait, using logistic regression. A secondary analysis was performed on an ordered phenotype of benign, severe and severe with colectomy. The model used was an ordinal logistic regression with proportional odds. For the dichotomous model, an analysis conditional on disease extent was also performed. The sample size was considered too small to perform a stratified analysis. The significance threshold for these analyses was set at 1×10⁻⁴ to account for the number of tests (156 index SNPs). Patients were genotyped using the Immunochip custom genotyping array that was designed to include all known variants from European individuals in the February 2010 release of the 1000 Genomes Project in 186 high-density regions known to be associated to one or more of 12 immune-mediated diseases, ¹⁰. The analyses performed were restricted to SNPs with minor allele frequency >5% (107,215) and imputed HLA (MHC) data¹⁴. Only loci with dense genotyping on the Immunochip were included. The results for the SNPs identified on the Immunochip (or the best proxy) for the known IBD loci were extracted for the main analysis. Results for the remaining SNPs and imputed MHC data¹⁵ were kept for an exploratory analysis.

A second objective was to evaluate association of the benign phenotype to a genetic burden score computed from the known (published) IBD loci. The score was computed from 107 SNPs associated to ulcerative colitis at P <1×10⁻⁴ in the analyses including all samples from the IIBDGC¹⁰. The score was computed as the sum of risk alleles, weighted by the natural logarithm of the effect.

In order to avoid population stratification issues, only patients of European descent were included in the analysis and two principal components were used in the analyses.

Data handling and risk score analyses were performed in R (v2.15.2). Genetic associations for dichotomous encoding were performed using plink, while ordinal associations were performed in Trinculo¹⁰.

Results

A total of 799 UC patients were included in the analysis following quality control; 130/799 (16.3%) had a benign disease course, and 669 (83.7%) – a severe course. Three hundred and seventy nine (47.4%) underwent colectomy. A complete study flowchart appears in figure 1. Thirteen additional patients underwent colectomy for dysplasia but were classified as having a severe phenotype on account of their medical treatment. The only disease characteristic that was significantly different in patients with benign course as compared to severe or colectomized patients was the disease extent on registration (table 1 and 2).

Figure 1.

Selection of patients for analysis

Table 1.

Clinical and demographic characteristics of the included patients

		Severe phenotype (N=669)		Benign phenotype (N=130)		p
		N	%	N	%
Gender	Male	321	48.0%	74	56.9%	0.062
	Female	348	52.0%	56	43.1%
Age at enrollment, years		44.2± 13.6			47.3±13.3
Age at diagnosis, years		29.4±12.1			30.2±12.5
Jewish ethnicity		101	15.1%	23	17.7%	0.42
Smoking status	Never smoked	384	57.4%	84	64.6%	0.33
	Current smoker	68	10.2%	16	12.3%
	Past smoker	209	31.2%	29	22.3%
Disease extent	Proctitis	19	2.8%	14	10.8%	0.0001
	Left- sided colitis	178	26.6%	62	47.7%
	Extensive colitis	423	63.2%	53	40.8%
Underwent colectomy	Total	379	56.7%	0	0.0%	0.0001
	For dysplasia	13	1.9%	0	0.0%

Table 2.

Comparison of clinical and demographic characteristics of patients with benign phenotype vs post-colectomy UC patients

		Severe phenotype with colectomy (N=379)		Benign phenotype (N=130)
		N	%	N	%	P
Gender	Male	181	47.8%	74	56.9%	0.071
	Female	198	52.2%	56	43.1%
Age at enrollment, years		44.8±12.4			47.3±13.3
Age at diagnosis, years		29.8±11.9			30.2±12.5
Jewish ethnicity		41	10.8%	23	17.7%	0.087
Smoking status	Never	212	55.9%	84	64.6%	0.105
	Current smoker	32	8.4%	16	12.3%
	Past smoker	129	34.0%	29	22.3%
Disease extent	Proctitis	5	1.3%	14	10.8%	0.0001
	Left-sided colitis	65	17.2%	62	47.7%
	Extensive colitis	263	69.4%	53	40.8%

Genetic associations

The samples were represented on principal components computed from the International Inflammatory Bowel Diseases Genetic Consortium (IIBDGC) European dataset^11,15.None of the first five components were significantly associated to the phenotype (P>0.1) (figure 2). We still included two principal components in the analyses.

Figure 2.

Population stratification analysis

None of the index SNPs from the IBD loci reached the predefined threshold of 1×10⁻⁴ in any of the analyses. The most associated variant in the dichotomous analysis was rs11742570 on chromosome 5, (P=0.0080, OR=1.46, 40,410 kbp). The most associated variant in the ordinal analysis was the same SNP (P=0.003, OR=1.34). No clear enrichment for association was found in the list of IBD index SNPs.

In the exploratory analysis of the remaining Immunochip SNPs and imputed MHC data, 26 SNPs (from 7 regions) reached a suggestive threshold of 1×10⁻⁴ in the dichotomous analysis (23 SNPs) or in the ordinal analysis (3 SNPs). Fifty four distinct association signals remain (22 SNPs). Analysis conditional on disease extent for the dichotomous analysis gave similar results (table 3). The risk score was not significantly associated to the benign phenotype (P=0.085 in the dichotomous model) (figure 3).

Table 3.

Suggestive associations

SNP Description							Dichotomous		Ordinal Model
SNP	CHR	kbp (HG19)	A1	A2	IBD Loci	MAF	OR	P	OR	P
rs16979501	5	617	A	G	Yes	0.42	0.55	5.6×10−5	0.77	1.2×10−2
rs11742854	5	674	G	A	Yes	0.15	2.40	2.4×10−2	2.72	5.8×10−5
rs25236392	6	31,344	A	G	Yes	0.084	2.49	1.2×10−5	1.49	3.0×10−2
rs75001121	15	79,255	A	G	No	0.08	0.36	3.9×10−3	0.43	8.9×10−5
rs17836409	19	55,184	A	G	No	0.10	2.19	8.8×10−5	1.35	4.2×10−2

¹Eight other associated SNPs in this region for the dichotomous analysis (all highly correlated)

²Nine other associated SNPs in the MHC region for the dichotomous analysis

Figure 3.

UC risk score analysis

Discussion

Our study attempted to evaluate the genetic predictors of disease course in UC using a large and well-identified study IIBGC-cohort. We did not demonstrate a significant association between the IBD loci previously identified by Jostins et al¹⁰. No significant association could also identified for the burden risk score comprised of the presence of the aforementioned loci. On exploratory analysis, several SNP located on chromosome 5, chromosome 6 (MHC region), 15 and 19, were identified.

Several previous studies attempted to establish an association between disease course and genetic markers in UC. The largest study to date, none of 163 IBD loci was associated with severe disease course⁵. In addition, no clear genetic associations could be identified for disease extent in UC¹¹.

In a previous work by Haritunians et al, SNPs within the MHC area were associated with the risk of medically refractory ulcerative colitis⁴, as with the results of our exploratory analysis. However, the MHC SNPs that demonstrated a significant association in that study did not show a similar association in our cohort. Additional studies suggested the impact of MHC area mutations on the risk of severity in UC^16,17. In an Australian study, an inverse correlation with risk of colectomy and proximal disease extent was demonstrated in carriers of the ATG16L1 T300A variant¹⁸. No correlation with the ATG16L1 mutations was observed in our study.

Our phenotypic definitions differed from the previous studies as we aimed to evaluate the associations with disease course defined not only by the need for colectomy, but also by requirement of immunomodulatory and biologic treatment. Our analysis included both the dichotomous and ordinal model addressing patients with severe phenotype with and without colectomy as separate phenotypes. We also applied stringent criteria requiring at least 10 years of follow-up prior to inclusion in order to meet the definitions of benign phenotype. Moreover, the previous studies aimed to identify the predictors of severe clinical course. In the current study, we aimed to identify the cohort of patients with benign disease course, as early identification of these patients may alleviate the need for more aggressive treatment and follow-up as frequently required in patients with more severe disease course.

Our study has some limitations. Importantly, our data collection was limited to the data included in the IBDGC clinical database. We were unable to accurately evaluate the history of corticosteroid treatment in our patients, however there is no clear definitions of what is considered mild or significant corticosteroid use. In addition, some patients were excluded from the analysis due to insufficient clinical data. The most important limitation of this study, however, is power, given the number of patients included. Another limitation is the coverage of the genome by the Immunochip, which is designed to deeply cover known loci associated to immune and inflammatory diseases, but is not genome-wide.

In summary, our study did not identify clear genetic associations with disease course in UC, confirming the multifactorial nature of this disease, with multiple factors impacting disease course and progression.

The suggested associations resulting from the exploratory analysis of our data merit validation if future studies. Given the modest power our study, the findings suggested on the exploratory analysis merit extension to larger discovery cohorts.