There is now strong evidence implicating the enteric flora in the aetiopathogenesis of inflammatory bowel disease (IBD), and identification of CARD15 (NOD2) as a pattern recognition receptor (PRR) has given novel insights into host-bacteria interactions. CARD15 is implicated as the intracellular sensor of muramyl dipeptide, a highly conserved bacterial peptidoglycan motif, and raises the question of whether other PRRs are involved in the pathogenesis of Crohn’s disease (CD).

Toll-like receptor 4 (TLR4), in combination with CD14, LBP, and MD-2, acts as the PRR for the lipid A moiety of lipopolysaccharide, a major component of gram negative bacteria. Two common cosegregating polymorphisms of this gene have been described in humans, Asp299Gly and Thr399Ile. Asp299Gly has been associated with reduced bronchial responsiveness following lipopolysaccharide stimulation¹ although recent data have questioned the functional effect of this variant.²

We therefore note with interest the article of Franchimont and colleagues reporting the Asp299Gly frequency in a Belgian population with IBD (Gut 2004;53:987–92). Variant alleles were associated with both CD and ulcerative colitis (UC) in two cohorts and the allele was preferentially transmitted from carriers to affected subjects in a transmission disequilibrium test.

However, apparently contradictory data from elsewhere in Europe highlight the difficulties of interpretation of genetic association studies from single populations. Torok et al examined the presence of both the Asp299Gly and Thr399Ile polymorphisms in a smaller German IBD population.³ In contrast with the Belgian data, this group identified an association of Thr399Ile with UC but there was no association with Asp299Gly, and no association with CD. In addition, we have previously published data on Asp299Gly in 480 Scottish patients with IBD and found no association with either CD or UC.⁴

Why are these data sets discrepant? Issues relating to statistical power, population stratification in case control studies, and phenotypic heterogeneity within IBD may contribute but we suggest more detailed examination of these data (table 1) provides evidence for genetic heterogeneity between populations in Europe.

CD genotype frequencies were very similar in the Leuven and Edinburgh data sets, and not significantly different from the German CD results. However, allelic frequencies in healthy controls were significantly different in the Scottish population compared with European controls (8.8% v 4.6%; p = 0.008, odds ratio 1.47 (confidence interval 1.2–3.5)).

It is clearly relevant that a strong suggestion of heterogeneity between populations was given in the original description of Arbour and Lorenz¹ where control population allelic frequencies for Asp299Gly ranged from 3.3% to 7.9% in French and North American data sets. It is noteworthy that these allelic variants were absent in the Japanese population.⁵

Moreover, there is now compelling evidence for genetic heterogeneity between populations for CARD15 in both healthy controls and CD patients. Variant alleles are absent from Asian CD and control populations and exist at a lower frequency in African American CD patients and Ghanaian controls (carriage 1%).^6,7 Carriage frequencies in CD patients approaching 50% have been documented in Caucasian populations from North American and Central Europe. There is also striking evidence for heterogeneity within Europe, and evidence for a geographical North-South gradient in gene effect, as for the CFTR delta 508 mutation in cystic fibrosis. Lower CARD15 (NOD2) frequencies in CD have been reported from Scotland⁴ and Finland⁸ and are absent in a small Icelandic population.⁹

These data illustrate further the real difficulties in candidate gene analysis in complex diseases. It is likely that the contribution of individual genetic determinants will differ between populations. We suggest that further genetic (as well as functional) data are required before the exact contribution of inherited variants of the TLR4 gene can be confirmed.