Materials and methods

Results

Members of the 170 CD families were genotyped for 19 highly polymorphic markers, evenly spread on chromosome 12 and including the D12S83 locus that exhibited positive linkage and linkage disequilibrium in previous studies.10 ,12 The genetic map generated using these genotypes spanned 157 centiMorgan (cM), a genetic length slightly shorter than the previously estimated length of chromosome 12 (174 cM, see Dib and colleagues14). The positions of the markers and the inter-marker genetic distances were in agreement with the published maps. The markers were 8.7 cM apart on average (range 5.0–16.6 cM). As the data from our families were more extensive than previously published, we used the map distances estimated from our data in subsequent statistical analyses (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Linkage map of chromosome 12 markers. The genetic map was derived from genotyping data of 170 families with Crohn's disease and generated by the Crimap program. The markers used were selected from the ABI Prism linkage mapping set version 1.0 and from the Genethon database (http://www.genethon.fr).

Non-parametric two point linkage analyses performed on the 95 multiplex families did not provide evidence of linkage for any of the polymorphic markers tested (table 1). The observed allele sharing between affected siblings was maximal at D12S1685 on the short arm of chromosome 12 (observed value 0.59 compared with an expected value of 0.50). However, this locus was 70 cM from D12S83 and was thus far from the genetic region of interest. In the genetic region of interest, centred around the marker D12S83, the maximum value of identity by descent between affected siblings was no more than 0.51 (expected value 0.50; p=0.35) providing no evidence of linkage.

Non-parametric multipoint linkage analyses are shown in fig 2. The information statistic measures the proportion of the possible inheritance information extracted at each location on the map using the panel of polymorphic markers in the families studied (fig 2A). The information is about 80% on average, indicating that the markers used are highly informative. The multipoint linkage statistic did not indicate linkage at any point on the chromosome (maximum value at D12S86, NPL=1.52, p=0.07) (fig 2B). In the genetic region surrounding the D12S83 locus, the NPL values were negative indicating that sharing of alleles identical by descent was less than that expected in this region. Because negative results may arise by chance when statistical power is low, the power to detect linkage using these data was analysed. For a locus specific increased relative risk of λ_s=2, a value previously attributed to the chromosome 12 susceptibility locus,9 the entire genetic region of interest was excluded (fig 2C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Multipoint linkage analyses of chromosome 12 in 95 multiplex families with Crohn's disease (CD) using the Genehunter package. (A) Information content; (B) multipoint non-parametric linkage (NPL) statistics. The region implicated by Satsangi et al9containing the inflammatory bowel disease (IBD) susceptibility locus is indicated by a shaded box. (C) Exclusion map calculated for a hypothetical locus with a relative risk λ_s=2. The exclusion threshold of LOD score (−2) is indicated.

Finally, because linkage disequilibrium was reported between the IBD2 locus and D12S83 by three independent groups,10-12 a search for associations was also performed (table 1). Because multiplex and simplex families are informative in association studies, 75 additional simplex families were included in this analysis. Using all of the 170 CD families, it was not possible to demonstrate an association, using either the TDT or HRR test, for any typed chromosome 12 locus

As genetic heterogeneity has already been demonstrated in this disease, we also classified our families into different phenotypic subgroups and repeated the statistical analyses on each group. Neither linkage nor linkage disequilibrium was observed in any of the phenotypic subgroups, including: age at onset below or above 20 years; small intestine and/or colon involvement; or low or high severity during the disease course (data not shown).

Discussion

The study was performed to determine the contribution of a previously identified IBD susceptibility locus on chromosome 12 in an independent set of Caucasian families.

The initial study reporting linkage between IBD and a susceptibility locus on chromosome 12 used a total of 160 IBD families. In that study linkage was observed for IBD (p=0.0000003) as well as for UC (p=0.0025) and CD alone (p=0.0003).9 Even if these results do not argue for a genetic heterogeneity between the two disorders, given the phenotypic differences between CD and UC, genetic heterogeneity in susceptibility cannot be ruled out. For this reason, we decided to limit our investigation to the CD phenotype. In the original study, linkage for CD was initially found in a group of 67 CD families containing 81 affected sib pairs.9 Because the number of families required to replicate a positive linkage may be significantly greater than the number initially used to find linkage, we studied a nearly twofold larger sample of families (95 families corresponding to 157 affected relative pairs) than that used in the initial study. Our sample was comparable in size to those in other linkage replication studies (table 2). Whereas the initial investigation was performed on affected sibling pairs only, the present study also examined second and third degree relative pairs, which is more informative. Thus these arguments suggest that the panel of CD families studied here was sufficient to detect linkage if it existed in this sample.

The density of markers used in our study was sufficient to extract, on average, 80% of the inheritance information on the long arm of chromosome 12 using non-parametric multipoint analyses. Furthermore, there is no information gap in the region of interest. Thus it is unlikely that addition of new markers would reveal linkage on this region using the same sample. The linkage tests reported here are model free—that is, they do not require any assumption about the genetic model underlying the predisposition to CD. Thus the negative results we obtained are not likely to be due to incorrect parameterisation in the statistical calculation. Also, as identity by descent between affected relatives could be inferred unequivocally for most of the relative pairs, it is unlikely that allele frequency errors significantly affected the results.

We were unable to demonstrate linkage between CD and any locus mapped on chromosome 12. The linkage score was negative on most of the chromosome and reached its highest positive value of 1.52 (p=0.07) for a marker located on the short arm of chromosome 12, 70 cM away from the linkage peak reported by Satsangi and colleagues.9Considering that it does not correspond to a region with a previously reported linkage, the value of the statistical test (p=0.02) did not reach the threshold of significance widely accepted to demonstrate linkage in a genome wide screen (p<0.00002, see Kruglyak and Lander19).

In contrast, no linkage was found to the genetic region previously implicated in susceptibility to IBD. To test if this negative result could be attributed to lack of power, an exclusion map was generated. The exclusion map allowed us to conclude that, in our families, no locus with λ_s ⩾2 could be located within the genetic region proposed by Satsangi and colleagues.9 However, a gene with a smaller effect cannot be ruled out.

An association between IBD and the microsatellite marker D12S83 has also been reported.12 This observation prompted us to replicate this result. Familial association studies have the advantage over case control studies with unrelated controls of avoiding biases due to population stratification. They are thus considered as more reliable for association studies. However, using 170 CD families, we failed to reproduce the reported association in our family panel.

Our findings indicate that in our family panel the IBD locus mapped on chromosome 12 had at most a small effect. Several other studies have also failed to find linkage on the long arm of chromosome 12 (table 2). Only three of seven replication studies detected the IBD locus on chromosome 12. In contrast, four of five replication studies detected the IBD1 locus on chromosome 16.

Several explanations may be proposed to explain the discrepancies between studies. Differences in diagnostic criteria may play a role. In most studies, including ours, the diagnostic criteria used were widely accepted and have been used by numerous investigators. Furthermore, several authors used the same diagnostic criteria and obtained discordant results.7 ,19 It has been suggested that the chromosome 12 locus might be more important for UC than CD.19 However, in most studies reporting a positive linkage, the contribution of the CD families was at least as important as that of the UC families.7 ,9 ,10 Furthermore, when different phenotypes were analysed separately in the present study, we were unable to detect linkage or linkage disequilibrium. It is thus unlikely that the present negative results are related to major clinical differences.

Differences between statistical analyses could also be considered. However, the statistics used in the above mentioned studies were based on model free methods and were often identical. Thus statistical methods are unlikely to explain the observed discrepancies.

Ethnic heterogeneity may also result in discordant results. All of the published studies were performed in Caucasian families of European origin. They were often the result of a multicentre effort of recruitment in the USA,10 ,19 Canada21 or Europe (Curran et al,7 Satsangiet al,9 and the present study). However, ethnic differences cannot be ruled out. For example, it should be noted that the chromosome 12 locus was detected in two cohorts with a large proportion of families recruited in the UK and was less often identified in other geographic samples. It has also been suggested that the IBD susceptibility loci may be different in Ashkenazi Jews.3 But discordant results were also obtained in populations with comparable proportions of Ashkenazi Jews in North America.10 ,19-21

Linkage analyses on the chromosome 16 locus (IBD1) are often available in the family samples used for linkage studies on chromosome 12.3 ,4 ,7 ,20-22 With the exception of the Canadian set of families, all samples also provided evidence of linkage on chromosome 16 (including the present sample, data not shown). Thus linkage on chromosome 12 does not seem to be conditional on linkage on chromosome 16.

Finally, the discrepancy could be explained by the effect of differences in exposure to unknown environmental factors. Indeed, because IBD are multifactorial disorders, a susceptibility gene interacting with a specific environmental risk factor may be more easily discovered in an exposed population. In contrast, in a non-exposed population, the same gene may remain undetected because its contribution to the disease is small. Further investigations, stratifying the genetic analyses for environmental risk factors, are required to highlight this hypothesis.