Background

Crohn's disease (CD) belongs to inflammatory bowel diseases (IBD) that are characterized by chronic, relapsing and recurrent inflammation of intestinal mucosa. The disease is thought to result from the action of environmental factors in genetically susceptible individuals. Three variants in the NOD2 [1, 2], IBD5 locus [3] and one variant in the IL23R [4] and in the ATG16L1 [5] have been independently confirmed to be associated with CD, including associations found previously in the Czech population [6, 7]. Recent studies, however, show that this list is far from being complete [5, 8–12].

The CTLA4 gene may also be considered as a plausible candidate for a genetic association with IBD. Its product, the cytotoxic T-lymphocyte-associated protein 4 (CTLA4) is a T-cell suppressor which plays an essential role in the function of the CD25(+)CD4(+) regulatory cells that control the process of intestinal inflammation [13, 14]. The CTLA4 gene maps within the 2q33 region that has been found to carry suggestive linkage significance for IBD [15]. The CTLA4 gene is associated with other immunopathologic diseases (type 1 diabetes, Graves' disease, Addison's disease, celiac disease, systemic lupus erythematosus, rheumatoid arthritis, vitiligo) [16]. Among the studied single nucleotide polymorphisms (SNPs), the CT60 (rs3087243) shows the most prominent associations, being followed by other three SNPs: JO31 (rs11571302), JO30 (rs7565213) and JO27-1 (rs11571297) [16]. A recent publication has shown evidence for association of another SNP within the CTLA4 with type 1 diabetes, the rs1427676 [17]. Previously published papers about genetic association with CD tested three variants in the CTLA4 gene: g.49A > G (rs231775), g.-318C > T (rs5742909) and the previously mentioned CT60, having found no association [18–21]. However, several works suggested that CTLA4 variants may influence the phenotype of CD [18, 19].

The aim of this study was to assess the association in the Czech population, using a set of markers previously shown to associate with other diseases.

Methods

Subjects

In a case-control design, 333 Czech patients were compared to 482 unrelated healthy Czech controls representing a general population sample from the same geographical region. We tested 137 pediatric-onset patients (71 boys, 66 girls) who developed CD under or at the age of 18 years and were diagnosed according to the Porto criteria [22], and 196 adult onset patients (77 males, 119 females) diagnosed according to endoscopic, radiological, histological and clinical criteria. Phenotypic classification was done according to the Montreal Classification [23]. The demographic and clinical characteristics of the patients are listed in Table 1 and Table 2. The control group included 482 individuals: 295 children, 187 adult; 311 males, 171 females; median age 12 years, interquartile range 7-34 years. The study was approved by the Ethics Committees of the authors' institutions, and a written informed consent was obtained from all participants or their guardians.

Table 1 Demographic characteristics of patients and control subjects
Full size table
Table 2 Clinical characteristics of patients
Full size table

Genotyping

Genomic DNA was extracted from peripheral blood leukocytes with a routine salting out method, or from salivary samples using Oragene DNA Self-Collection Kit according the manufacturer's protocol (DNA Genotek Inc., Ottawa, Ontario, Canada). One SNP proximal to CTLA4 (rs736611), one from within the gene (g.49G > A, rs231775), and four SNPs located distally from the coding part of CTLA4 gene (rs3087243 also called CT60; rs11571302 called JO31; rs11571297 called JO27-1; and rs1427676) were selected based on available literature and genotyped using the TaqMan SNP genotyping assays (TaqMan SNP Genotyping Assay by Applied Biosystems, Foster City, CA, USA). The assays were run on an ABI 7300 machine (Applied Biosystems, Foster City, CA, USA) and evaluated according to manufacturer's instructions. To ensure consistency between runs, samples of known genotypes were repeated in every analysis. For testing interactions with other associated genes, we used genotypes generated in previously published studies on this sample set [6, 7].

Statistical analysis

The Hardy-Weinberg equilibrium was checked by comparing observed to expected genotype frequencies in the control subjects, and tested using exact tests. Associations of particular SNPs with CD were evaluated using odds ratios (OR) with 95% confidence intervals (CI). Haplotype analysis was performed by estimating the haplotype frequencies by the expectation-maximization algorithm implemented in the R-project package 'haplo.stats' version 1.3.1. Association of haplotypes with the conditions was tested using log-linear modeling. Then, a potential interaction between the CTLA4 variants and other genetic factors associated with the autoimmune conditions were tested. The statistical analysis was performed using the R-project package 'SNPassoc' version 1.5-2 [24].

Results

Crude associations

The frequencies of the variants and respective OR are listed in Table 3. No crude associations with CD were found for the tested SNPs under the log-additive or dominant models. The genotype distributions in control subjects conformed to Hardy-Weinberg equilibrium in all SNPs (p > 0.20) except the rs1427676 (p = 0.014 in exact tests) which was therefore excluded from all further analyses.

Table 3 Distribution of genotypes of the studied CTLA4 polymorphisms1
Full size table

As the part of chromosome under the CTLA4 gene is divided into the several blocks [16] we performed a haplotype analysis using the five SNPs; no crude association with CD was observed (data not shown).

Interaction of the CTLA4 SNPs with variants in IL23R andNOD2

We then tested possible interactions between variants in the CTLA4 and polymorphisms in other genes previously associated with CD: NOD2 gene p.Leu1007fsX1008 (c.3020insC), IL23R gene rs11209026 (c.1142G > A) [6, 7], see Figure 1. This was done in dominant models using an R-project package 'SNPassoc' version 1.5-2 [24]. Significant interactions were observed between the three CTLA4 variants (CT60, JO31, JO27-1) and NOD2 p.Leu1007fsX1008, and the same variants in the CTLA4 and IL23R rs11209026.

Figure 1
figure 1

SNPs interaction in the dominant model. The upper triangle in matrix from this function contains the p values for the interaction (epistasis) log-likelihood ratio test (LRT). The diagonal contains the p values from LRT for the crude effect of each SNP. The lower triangle contains the p values from LRT comparing the two-SNP additive likelihood to the best of the single-SNP models [24]. P value of interactions: * P < 0.05; ** P < 0.01; *** P < 0.001.

Full size image

For a quantification of the CTLA4 association stratified by the above NOD2 and IL23R polymorphisms see Table 4 a Table 5: the minor alleles of the CT60, JO31 and JO27-1 within the CTLA4 modified the risk of Crohn's disease in the stratum of subjects carrying the frameshift insertion p.Leu1007fsX1008 in NOD2, while no perceivable effect of CTLA4 was found in the stratum of p.Leu1007fsX1008 wild-type homozygotes. Similarly, the three CTLA4 variants clearly, albeit moderately, decreased the risk of CD in the stratum of subjects carrying minor alleles of rs11209026 within the IL23R (G/A and A/A), while no effect was observed in the IL23R wild-type homozygotes. The effect was observable also for the haplotype consisting of the three minor alleles of the variants in tight linkage disequilibrium, the CT60 "A", JO31 "T", JO27-1 "G" haplotype.

Table 4 Stratified analysis of the effect conferred by the CTLA4 CT60, JO31 and JO27-1 variants
Full size table
Table 5 Stratified analysis of the effect conferred by the CTLA4 CT60, JO31 and JO27-1 variants
Full size table

Genotype - phenotype analysis

Using a case-only design, we tested whether the phenotypic characteristics of the patients are dependent on carriage status of the minor alleles at CTLA4 variants.

Table 6 shows logistic regression with the outcomes of clinical characteristics and the three CTLA4 variants and the haplotype as the predictors. The pediatric-onset patients differed to the adult-onset patients in their frequencies of the minor allele at the CT60 (74% versus 63%, p = 0.03). The CT60, JO31 and JO27-1 SNPs, as well as their "ATG" haplotype retained their associations with the age at diagnosis after adjustment to the effect of the NOD2 variant (p.Leu1007fsX1008). No difference between pediatric and adult-onset group was found for the other two variants in the CTLA4 gene, data not shown.

Table 6 Genotype-phenotype analysis1
Full size table

The CTLA4 variants were weakly associated with the ileal-only (L1) and ileocolonic involvement (L3) in a dominant manner, while no association was observed with any of the remaining clinical characteristics: localization in the upper gastrointestinal tract, the stricturing or penetrating behavior of the disease, perianal disease, extraintestinal manifestation, or the need for abdominal surgery (data not shown).

Discussion

The immunologic importance of the CTLA4 gene is in striking contrast to the lack of knowledge on the functional relevance of its numerous polymorphisms. Consequently, many groups have investigated various polymorphisms located within various regions of the gene. The first published study on CTLA4 variants in CD investigated the g.49A > G (rs231775) and g.-318C > T (rs5742909) in the Dutch and the Chinese populations, finding only an association with the age of onset [18]. Similarly, in a Hungarian work, no association of g.49A > G with CD was detected [21]. Since the work by Ueda et al [16] had been published on dissecting the association of CTLA4 with immunopathological diseases, further investigators focused on the CT60 polymorphism. This variant was studied in the Japanese [19] and the Spanish [20] populations, however no crude association with CD was detected. The G/G genotype of g.49A > G was associated with penetrating form of CD in the Japanese dataset [19]. No association within 2q33 chromosomal region has been found by genome-wide studies [4, 5, 8–12].

Thus, compelling evidence has been gathered against simple association of the disease itself with the polymorphisms of CTLA4. In line with these findings, we observed no crude association unless further genetic factors were taken into account. However, when CTLA4 was considered as a modifier of the effects conferred by the NOD2 and IL23R genes, possible interactions substantiated. Interactions in multifactorial immunopathological diseases are not infrequent: in CD, the interactions with the NOD2 gene were detected in the IBD5 locus [25], IBD6 locus [15], TNFA [26], DLG5 [27], ATG16L1[28], IL23R [29], TLR4 [30] and in CD14 [30]. The interaction was also found between IBD5 locus and IL23R [31] and between Toll-like receptor-9 polymorphisms and variants in NOD2 and IL23R [32].

The interactions we found for the CT60, JO31, JO27-1 variants of CTLA4 (or their haplotype) with the p.Leu1007fsX1008 variant of NOD2 may imply that the effect of the strongest risk variant within the NOD2 (p.Leu1007fsX1008) can be expressed better on the background of the common CTLA4 haplotype. This is suggestive of a complex pattern of gene-gene interaction that may merit pursuing further functional studies. Similarly, the risk haplotype of CTLA4 also interacts with the IL23R protective variants. This rather weak interaction can be also due to the relatively limited size of the dataset.

In addition to the modifying effect on the risk of the disease itself, we observed an association with the age at onset and the disease subphenotypes. Indeed, the impact of genetic factors in early-onset patients with CD seems to be stronger than in adult-onset patients (reviewed by de Ridder L et al. [33]). In our dataset, the age at diagnosis was associated with CT60, JO31 and JO27-1. Influence of CTLA4 variants on the age at diagnosis has been previously described by Xia et al, although with a different SNP (g.-318C > T) [18]. Moreover, their patients were divided into groups where 40 years of age was the cut off, not 18 years as in our study.

The CTLA4 was associated with the ileal and ileocolonic involvement in our case set: up to our knowledge, this is the first time when localization of CD is influenced by any variant within CTLA4. It should be however noted that these associations are weak, merit further investigation in other populations, and their clinical relevance can be only hardly envisaged. The ileal form of disease has been shown more common in adult-onset patients and more common in patients carrying minor variants of the NOD2 gene. A possible explanation of the association of CTLA4 with localization of disease might be found in the interaction between CTLA4 and NOD2 gene.

Similarly to Machida et al [19] we also tested whether the g.49A > G variant influences the occurrence of penetrating disease, but we were not able to confirm this association. However, the genetic background between Japanese and Czech populations differs markedly.

Conclusions

We present a study of genetic association of polymorphisms within the CTLA4 gene with CD and its subphenotypes, using a representative set of markers previously reported from other studies. We observed interactions of the CTLA4 haplotype with variants in NOD2 and IL23R genes, and detected an effect of three variants of the CTLA4 on the age at diagnosis and localization of the disease.