International Journal of Colorectal Disease

, Volume 22, Issue 4, pp 419–424

Investigation of association of the DLG5 gene with phenotypes of inflammatory bowel disease in the British population

Authors

  • Alexandra V. Pearce
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Sheila A. Fisher
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Natalie J. Prescott
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Clive M. Onnie
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Reenal Pattni
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Peter Green
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
  • Alastair Forbes
    • Institute for Digestive DiseasesUniversity College London Hospitals Trust
  • John Mansfield
    • Department of Gastroenterology & HepatologyUniversity of Newcastle upon Tyne
  • Jeremy Sanderson
    • Department of GastroenterologyGuy’s and St Thomas’ NHS Foundation Trust
  • Stefan Schreiber
    • Institut für Klinische MolekularbiologieChristian-Albrechts Universität
  • Cathryn M. Lewis
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
    • Department of Medical & Molecular GeneticsKing’s College London School of Medicine
Original Article

DOI: 10.1007/s00384-006-0151-4

Cite this article as:
Pearce, A.V., Fisher, S.A., Prescott, N.J. et al. Int J Colorectal Dis (2007) 22: 419. doi:10.1007/s00384-006-0151-4

Abstract

Background and aims

Mutations in the DLG5 gene are associated with an increased risk of inflammatory bowel disease (IBD) in some European populations. Initial investigation of a British IBD population showed evidence of association of one of three DLG5 variants, R30Q, in a family-based collection but not in a case-control cohort. We have now examined the association of the R30Q polymorphism in a large cohort of British IBD cases, tested for interaction between the DLG5 and CARD15 genes and assessed possible association of DLG5 with clinical features of Crohn’s disease (CD) and ulcerative colitis (UC).

Materials and methods

DLG5 R30Q and the CARD15 polymorphisms, Arg702Trp, Gly908Arg and Leu1007fs were genotyped in 1,148 IBD cases and 749 controls. DLG5 R30Q was also analysed in cases stratified by CARD15 genotype, disease subtype and smoking history.

Results/findings

No significant difference in frequencies of the R30Q variant was observed between IBD cases (9.9%) and controls (10.1%) or in cases analysed separately as CD and UC. There was also no significant difference in the frequency of R30Q between CD cases carrying risk-associated CARD15 alleles and those that did not. The frequency of R30Q was higher in CD cases with ileal disease than cases without (p=0.042) and higher in CD cases who had smoked than in nonsmokers (p=0.009).

Interpretation/conclusion

The R30Q variant in the DLG5 gene does not appear to be associated with an overall increase in the risk of disease in a British IBD cohort, but differences in its frequency in subgroups of CD patients warrant further investigation.

Keywords

Inflammatory bowel diseaseDLG5Association

Introduction

The contribution of genetic factors to the development of inflammatory bowel disease (IBD) is well established [1, 2], culminating in the discovery that mutations in NOD2/CARD15 are associated with susceptibility to Crohn’s disease (CD) [35]. However, genome scans for genetic linkage to IBD and meta-analyses have identified multiple loci that may contribute to the risk of CD or ulcerative colitis (UC) [1, 6, 7], suggesting a polygenic aetiology for IBD.

We first implicated the pericentromeric region of chromosome 10 in IBD susceptibility in the largest IBD genome-wide scan to date, utilising 356 affected sib pairs [8]. This locus was then refined in an extended European linkage cohort that led to the identification of a two-peak linkage curve. Strong evidence for association at the distal linkage peak at 10q22–10q23 was narrowed to the DLG5 gene, in which two of four common haplotypes showed significant association [9]. Haplotype A, tagged by single nucleotide polymorphism (SNP) DLG5_e26, was undertransmitted to individuals with IBD, whereas haplotype D, tagged by the variant 113A which results in the amino acid substitution arg30gln (R30Q), was overtransmitted to individuals with IBD. A rare variant, 4136C>A, causing the amino acid substitution P1371Q, was also significantly associated with IBD. These associations were replicated in a case-control analysis of German CD cases with odds ratios of 1.35, 1.62 and 1.51 for the DLG5_e26, R30Q and P1371Q SNPs, respectively [9]. Interestingly, there was also evidence of interaction between DLG5 and the CD susceptibility gene, CARD15, as stratification of the TDT data showed that R30Q was overtransmitted to individuals with CD who carried risk-associated variants of CARD15 but not to those who lacked these variants. As DLG5 encodes a scaffolding protein and has been implicated in cell growth and maintenance of cell shape, it was postulated that functional variants within it may affect the maintenance of epithelial cell integrity, and thus, IBD susceptibility.

In a subsequent study [10], we tested the DLG5 variants associated with IBD for replication of the association in British, Canadian and Italian populations that were independent of the original report. The DLG5_e26 SNP, representing the common haplotype A, was not associated with IBD in either the British, Canadian or Italian case-control cohorts. However, R30Q was associated with IBD in a combined analysis of the Canadian and Italian populations (p=0.003) but not in the British case-control sample. The rare P1371Q variant was not associated with IBD in the British cohort, but was not analysed in the Canadian and Italian samples. In the family-based analysis, the Canadian and British cohorts were combined for TDT analysis, and we detected a modest excess of transmission of the 30Q allele (SNP 113A) (combined p=0.018; UK, only p=0.017; Canadian, p=NS). The common haplotype A was not observed to be associated with IBD in any of the populations tested in the TDT. We suggested that the positive association seen for the common haplotype A in the original report [9] could have been partially a statistical fluctuation, and partially, a result of linkage disequilibrium with the R30Q association [10].

The partial replication of the association of the R30Q variant in DLG5 with IBD, described above, left unresolved the question of whether DLG5 was indeed a susceptibility factor for IBD in the British population. Given the potential significance of this finding for our understanding of the genetic components of susceptibility to IBD, we designed a follow-up study to address this question more fully in three ways. First, the number of British IBD cases was increased from 689 in the study of Daly et al. [10] to 1,148, and the controls, from 493 to 749, to increase the power of the study to detect a modest genetic effect. Secondly, CD cases were stratified by CARD15 genotype to investigate the possibility that the association might be restricted to cases carrying CARD15 risk alleles, as a result of an interaction between the two loci. Thirdly, IBD cases were stratified by disease behaviour and localization (CD) and maximal extent of disease (UC) to establish whether the R30Q association might be confined to a subset of clinical phenotypes. Finally, we also looked for a possible interaction between R30Q and smoking behaviour, as this is known to influence risk for CD and UC.

Materials and methods

Patient ascertainment

DNA was obtained from a British population including 630 patients with CD, 518 with UC and 749 controls, after ethical review and informed consent at Guy’s and St. Thomas’ Hospitals, St. Mark’s and King’s College Hospitals (London), and the Royal Victoria Infirmary, Freeman Road Hospital and North Tyneside Hospital (Newcastle). A subset of 689 of the 1,148 IBD cases was included in the study of Daly et al. as described [10]. Diagnoses were established by conventional criteria of clinical, radiological, endoscopic analysis and from histology reports [11]. Controls were recruited through the European Collection of Animal Cell Cultures (Wiltshire, England) and residents in Newcastle, England, as part of a regional genetic study. Clinical data were available for a subset of CD and UC patients. CD patients were assessed for the site of the disease (ileal, colonic, perianal and combinations of these), disease behaviour (inflammatory, stenosing and perforating). UC patients were assessed for maximal extent of disease (left-sided: extending up to the splenic flexure) or extensive disease (extending past the splenic flexure) at Guy’s, St. Thomas’ or St. Mark’s Hospitals. Where there were any discrepancies over the extent of the disease as determined by macroscopic and microscopic evidence, histological results were used. Age at onset of symptoms, smoking status and the presence of extra-intestinal manifestations such as joint disease, erythema nodosum, uveitis and liver disease (primary sclerosing cholangitis) were also recorded.

Genotyping

DLG5 R30Q: The SNP rs1248696 (113G→A, R30Q) was genotyped by Pyrosequencing on a PSQ HS96A Pyrosequencer (Pyrosequencing AB, Uppsala, Sweden) [12]. Each polymerase chain reaction (PCR) amplification mixture of 10 μl contained 2-mM MgCl2, 0.2-mM dinucleotide triphosphates (purchased from Amersham Biosciences UK LTD., Chalfont St. Giles, UK), 0.2-μM forward and reverse primers (forward: 5′-GCAGCTGAATGGAGAGGTTC-3′, reverse: 5′CGTGAATGCCAGATGAACAC-3′), 2U Promega Taq DNA polymerase and 10–50-ng genomic DNA. The forward primer was biotinylated at the 5′ end. The PCR reaction was started by denaturation at 92°C for 2 min, followed by 30 cycles of 20 s denaturation at 92°C, 30 s annealing at 62°C, 30 s extension at 72°C, with a final extension step at 72°C for 5 min. Samples were prepared by immobilization on streptavidin-coated Sepharose beads (Amersham Biosciences UK LTD., Chalfont St. Giles, UK) according to standard methods utilizing a 0.1-mM sequencing primer (5′-CCCTCCTCACTGACC-3′). The assay was verified by comparison with a subset of 96 samples that had been genotyped by DNA sequencing.

CARD15 variants: All individuals with CD were genotyped for the three known risk variants associated with CD in the CARD15 gene: R702W (rs2066844) and Leu1007fs (rs2066847) were genotyped by Pyrosequencing, and G908R (rs2066845) was genotyped by TaqMan assay. The assay conditions for the CARD15 variants are available from the authors.

Statistical analysis

Genotypes were tested for Hardy–Weinberg equilibrium. Allele and genotype frequencies in cases and controls were compared using chi-squared tests (Splus v6.0). Results are expressed as a percentage frequency of the R30Q minor allele A. Significance levels reported were not corrected for multiple testing. Median age at disease onset was compared between genotype subgroups, using Wilcoxon’s rank sum test.

Results

In our previous study of the three DLG5 SNPs in the British IBD population, we had found no evidence for the association of the haplotype A tagging SNP or the rare P1371Q variant with IBD, but the results for R30Q were equivocal, showing no association in British case-controls but positive association in the TDT analysis [10]. In the current study, we, therefore, focused on the analysis of R30Q. In our extended collection of 1,148 IBD cases and 749 controls, we found no significant difference in the frequency of the R30Q variant between cases (9.9%) and controls (10.1%) in the combined IBD group, OR=0.98 (95% CI: 0.79, 1.22), or when cases were analysed separately as CD and UC (Table 1). Under a dominant genetic model, as proposed by Stoll et al. [9], the frequency of individuals carrying the variant was lower in IBD cases (18.9%) than in controls (19.1%), OR=0.99 (95% CI: 0.78, 1.25).
Table 1

British case-control data for the R30Q variant (113G>A)

 

N

GG

AG

AA

Freq (A) (%)

Controls

749

606, 80.9%

135, 18.0%

8, 1.1%

10.1

CD

630

515, 81.7%

108, 17.1%

7, 1.1%

9.7

UC

518

416, 80.3%

98, 18.9%

4, 0.8%

10.2

IBD

1,148

931, 81.1%

206, 17.9%

11, 1.0%

9.9

N=total number of individuals; all p-values >0.05 when comparing frequencies to controls

Genotypes were in Hardy–Weinberg equilibrium in both cases and controls. Disease risks observed in the previous studies were OR=1.53 (95% CI: 1.16, 2.03) [9] and OR=1.96 (95% CI: 1.21, 3.16) [10] for each copy of the R30Q variant. Based on our observed frequency of 10.1% in controls, our combined case-control cohort provides 98% power (5% level of significance) to detect an effect of the size we previously detected in the German population [9]. The equivalent power for CD and UC is 94% and 92%, respectively.

CARD15 Stratification

We explored the reported interaction between the R30Q G113A polymorphism and the three CARD15 variants R702W, G908R and L1,007 fsC which have been consistently found to be associated with CD [4, 13, 14]. In contrast to the original report, [9] which found over transmission of R30Q only to CD cases who were positive for CARD15 risk variants (thus, implicating specific relevance of this variant to CD patients), we did not find a significant increase in frequency of the rare 113A allele in 194 CD cases carrying the risk-associated CARD15 alleles compared to 418 CD cases who did not, or when compared to controls (Table 2).
Table 2

Stratification of the R30Q variant in CD CARD15 mutation positive cases vs CD CARD15 mutation negative cases

CARD15 status

N

GG

AG

AA

Freq (A) (%)

Negative

418

344, 82.3%

70, 16.7%

4, 1.0%

9.3

Positive

194

154, 79.4%

37, 19.1%

3, 1.5%

11.1

CARD15 status is defined as positive (carrying at least one of the three common CARD15 risk alleles-R702W, G908R, Leu1007fs) or negative (carrying none of these three CARD15 risk alleles); N=total number of individuals

Genotype–phenotype analysis

A possible reason for the lack of replication could be that the R30Q variant is associated only with specific subtypes of CD and that the proportions of these subtypes varied in the different studies and populations. We, therefore, stratified cases by the characteristics of their clinical phenotype including site of disease and disease behaviour (CD), maximal extent of disease (UC) and smoking history. The genotype frequencies of the rare R30Q allele were not significantly different between most subgroups of CD (Table 3) or UC (Table 4). The R30Q variant was, however, more frequent in CD cases with ileal disease than in cases with no ileal disease (p=0.042), although it was not significantly different in either of these subgroups when they were compared to controls (p>0.05). There were also no significant differences in allele frequencies in relation to disease behaviour. Upon stratification of the cases by smoking history, we observed that the frequency of R30Q variant was significantly higher in CD cases who had ever smoked compared to CD cases who were nonsmokers (p=0.0091), but the R30Q frequencies in these groups were not significantly different from the controls (p>0.05). No significant differences in frequency were observed for UC cases who had ever smoked compared to those who were nonsmokers. Finally, there was no significant difference in the median age at disease onset between CD patients who carried the R30Q variant (22.0 years) and those who did not (22.5 years, p>0.9). The corresponding median ages in UC patients were 31.5 and 29.0 years, respectively (p=0.6).
Table 3

Frequency of R30Q variant in CD cases stratified by clinical phenotype

 

N

GG

AG

AA

Freq (A) (%)

Localisation

     

 Colitis—All

244

204

39

1

8.4

—Pure

107

91

16

0

7.5

—None

186

140

44

2

12.9

 Ileitis—All

264

203

58

3

12.1

—Pure

148

112

35

1

12.5

—None

166

141

25

0

7.5

 Perianal—All

96

78

17

1

9.9

—Pure

19

16

3

0

7.9

—None

334

266

66

2

10.5

Behaviour

     

 Stenosing—All

202

155

45

2

12.1

—Pure

108

84

23

1

11.6

—None

228

189

38

1

8.8

 Fistulating—All

177

138

37

2

11.6

—Pure

42

36

6

0

7.1

—None

253

206

46

1

9.5

 Inflammatory—All

173

142

30

1

9.2

—Pure

127

106

21

0

8.3

—None

257

202

53

2

11.1

Bowel resection—Yes

241

190

50

1

10.8

—No

189

154

33

2

9.8

Extra-intestinal manifestations

120

96

24

0

10.0

Smoking history—Yes

238

179

58

1

12.6

—No

176

154

20

2

6.8

N=total number of individuals; all p-values >0.05 when comparing frequencies to controls

Table 4

Frequency of R30Q variant in UC cases stratified by clinical phenotype

 

N

GG

AG

AA

Freq (A) (%)

Extent of disease

     

 Left-sided

67

54

12

1

10.4

 Extensive

221

177

43

1

10.2

Extra-intestinal manifestations

59

47

12

0

10.2

Smoking history—yes

113

94

17

2

9.3

—no

188

149

39

0

10.4

N=total number of individuals; all p-values >0.05 when comparing frequencies to controls

Discussion

The strong and well-replicated association of genetic variants in the DLG5 gene in a predominantly German IBD population and our replication of the association of the R30Q variant in a mixed Italian and Canadian sample, suggested that this finding might be widely replicated in other European populations. However, our initial studies of the R30Q variant in British patients and families with IBD were equivocal, with some evidence of replication in a family-based study but no replication in a case-control study design. We chose to address this inconsistency by analysis of R30Q in a highly powered sample of British CD and UC cases and controls and also by stratification of CD cases by CARD15 genotype and of CD and UC cases by clinical phenotype to examine the possibility that the association was restricted to some genotypic or phenotypic subgroups.

In our case-control study, the frequency of the R30Q variant was almost identical in the IBD group and in controls (OR: 0.98, 95% CI: 0.79–1.22) and in the separate phenotypes of CD and UC. The 95% confidence interval for the odds ratio overlaps with the original report (1.53, 95% CI: 1.16, 2.03) [9]. Our combined sample of 1,148 IBD cases and 749 controls provides 98% power to detect an effect of this size and 80% power to detect a significant difference in allele frequencies if the true odds ratio was at least 1.35.

Our investigation of a possible gene–gene interaction between the DLG5 and CARD15 loci was based on the report [9] of an increased frequency of the R30Q variant with CD in individuals carrying the risk-associated CARD15 alleles. Also, it might be difficult to detect association of a gene of modest effect with disease if it was confined to a genotypic subgroup at a second locus. An example of this is the interaction that we previously demonstrated between CARD15 and the IBD5 locus on chromosome 5q31, in which the association signal in Crohn’s disease was far stronger in the CARD15 mutation positive group than in the total sample of CD patients [15]. However, in this study, we found that the frequency of the R30Q variant was very similar in the CARD15 mutation positive and mutation negative groups, and was not significantly different to controls. This suggests that it is unlikely that an interaction exists between R30Q in DLG5 and the CARD15 variants in this British population. This is consistent with the lack of association of R30Q with CD in the unstratified sample.

The possible effect of phenotypic differences between the patient populations in this and other studies on the lack of replication of the R30Q association was investigated by analysis of the frequency of the R30Q variant in phenotypic subgroups. The only significant difference in frequency between these subgroups was between CD cases with ileal disease and those without. The higher frequency of R30Q in the ileal subgroup may be worthy of follow-up in other cohorts given the proposed interaction of DLG5 with CARD15 [9] and the strong association of CARD15 variants with ileal disease [1, 2]. As smoking has been shown to increase the risk of late onset CD and worsens its clinical course, [16] we looked for a possible gene–environment interaction between smoking and the R30Q genotype. The frequency of the R30Q variant was higher in CD cases with a smoking history than in CD cases who had never smoked, although the difference in frequency was not significant when comparing either of these two groups to the controls. It would be interesting to see whether such an interaction can be detected in CD populations in which an association with R30Q has been reported [9, 10]. Why such an interaction might exist is not yet clear. It is possible, for example, that the reported functional deficiency of macrophages from smokers in their ability to kill intracellular bacteria [16] in combination with an increased bacterial load that might result from the proposed defective barrier function of DLG5 variants [9] enhances susceptibility to CD.

There are a number of possible explanations for the lack of replication of the association of R30Q with IBD in a British population. There could be genetic heterogeneity, such that DLG5 does not contribute to the risk of IBD in the British population. There could also be allelic heterogeneity, where sequence variants in DLG5 other than R30Q are responsible for disease risk in the British. Both of these possibilities seem rather unlikely in the light of the very similar frequencies of CARD15 mutations and their associated risks in these two populations [13, 17]. However, a recent report does provide evidence of allelic heterogeneity in the Japanese population [18] as the R30Q variant is completely absent in the Japanese, but another SNP (DLG5_2) was weakly associated with CD. Another possibility is that the R30Q variant is not the causative mutation in DLG5 in the German population but is in strong linkage disequilibrium with the true disease causing allele and that this linkage disequilibrium is weaker in the British population, making the association with the surrogate marker more difficult to detect. The original DLG5 associations [9] seem unlikely to be false positives (type I error) as the R30Q association was replicated in a Canadian/Italian case-control sample [10]. Very recently, two other studies have failed to detect association of the DLG5 variants with disease in IBD populations from Scotland and Southern Germany [19, 20]. If the effect of DLG5 on disease risk is small in most populations, then a meta-analysis of the available studies may help to assess the contribution of this gene to susceptibility to inflammatory bowel disease.

Acknowledgements

This work was supported by the Wellcome Trust, the Guy’s and St. Thomas’ Charity, the European Commission Framework V program and CORE (UK). A. Pearce was supported by a studentship from King’s College London School of Medicine.

Copyright information

© Springer-Verlag 2006