Chromosome 7 haplotype segments (defined by flanking critical recombinants) that were shared among affecteds within the 16 families contributing most to the MOD score (family-based LOD scores . 0.1 at 59.3 cM under the recessive model that generated the MOD score). The single marker D7S484 at 59.3 cM—both copies of which are consistently shared among affecteds within families—is marked in grey. 

Context in source publication

Context 1

... linked), but linkage under genetic heterogeneity (allowing a proportion of families not to be linked) was suggested (P 1⁄4 0.005), with an estimated proportion of linked families of 0.30. Table II shows the association between phenotypic characteristics of the families in the combined data set and contribution to the MOD score under both the dominant and recessive models. For the dominant model that produced the MOD of 3.89 at 65.72 cM for the Oxford data alone, there was a highly significant difference in contribution to the LOD score between the two data sets at that location (Table II). There was an association with the number of relatives who had Stage B disease (equivalent to rAFS Stages III and IV), but this was entirely limited to the Oxford data set and mainly driven by the two families with four Stage B affecteds (data not shown). These two families contributed partial LOD scores of 0.55 and 0.46 to the overall MOD score under the dominant model. Limiting the linkage analysis to families with at least one Stage B affected, however, did not materially improve the combined LOD score. Moreover, even after adjusting for the number of affecteds with Stage B, there remained a significant difference in LOD score at this locus between the Oxford and Australian data sets (suggesting genetic heterogeneity between the two data sets, which was not explained by any of the collected phenotypes). Under the recessive model, the phenotypic characteristic most associated with an increased family-based LOD score appeared to be the presence of multiple relatives with Stage 723 B disease ( P , 0.001, adjusted for Oxford / Australian origin, number of affecteds, number of ‘Stage A’ affecteds, and polymorphic information content). This result was largely driven by one family in the Oxford data set (containing four women with Stage B and one woman with unknown disease stage), which contributed 1.66 to the overall MOD score. An increased contribution of families with multiple Stage B affecteds was further suggested when the recessive parametric analysis was limited to the 189 families who had at least one member with Stage B disease. This resulted in a virtually unchanged LOD score of 3.29 at 59.35 cM. Critical recombinant analysis was adopted to narrow down the linkage region observed for the Oxford data under the dominant model with reduced penetrance (see Materials and Methods). In this study, 32 out of the 52 families had a LOD score . 0.1, of which 14 had a LOD . 0.2. The haplotypes that were shared by most affecteds within the 14 families all included D7S484 and D7S510, a 17.9 Mb region encompassed by D7S516 and D7S519 (data not shown). Fine mapping of the families with recombinations at either side of these markers, by including markers separated by a maximum of 1 cM, reduced the suggested region of interest to the area between D7S484 724 and D7S2548, a 6.4 Mb interval containing 48 genes according to the latest assembly of the human genome (NCBI 35) (Wheeler et al. , 2002). For the 16 families in the combined data set contributing most to the MOD score under the recessive model (with family-based LODs . 0.1), the location of critical recombinants was also investigated in the haplotypes of affecteds. As expected—because the best fitting model was recessive with high penetrance—virtually all affecteds within each family shared two haplotypes spanning the linkage region (Figure 3). There was a single marker, D7S484 (the location of the linkage peak), which showed allele sharing among affecteds within (but not between) all of the 16 families. There was no association between disease status and D7S484 ( x 2 1⁄4 5.37, df 1⁄4 7, P 1⁄4 0.6), confirming that there was no increased allele sharing among affecteds between families. The linkage region surrounding D7S484 at 35.1 Mb was flanked by D7S516 at 28.0 Mb and D7S510 at 39.0 Mb, an 11 Mb region containing 96 genes. The results presented in this article provide significant evidence that a locus with near-Mendelian autosomal inheritance on chromosome 7p is linked to endometriosis. Traditionally, a LOD 3 has been considered as evidence of significant linkage; however, more recent recommendations suggest assessing significance by data simulation (Wiltshire et al. , 2002). Our simulations found that the original result in the Oxford data set for chromosome 7 was genome-wide significant and that adding the Australian data set upheld this significance, despite a substantially smaller albeit positive linkage signal in the latter data set. Endometriosis is a condition that affects millions of women world-wide in terms of pain and infertility, and yet little is known about its aetiology. Despite ample evidence of familial aggregation (Bischoff and Simpson, 2000; Zondervan et al. , 2001; Stefansson et al. , 2002), candidate gene studies have, to date, been largely unsuccessful in unco- vering any aetiological targets. To a large extent, this is probably due to inappropriate and underpowered study design (Zondervan et al. , 2002; Zondervan and Cardon, 2004), but the complexity of the underlying biological and pathological mechanisms provides an additional challenge in targeting promising candidate genes. Results from linkage analyses—although in themselves problematic because of the large numbers of families with multiple affecteds required for complex traits—could provide vital new clues in the search for predisposing genetic variants and thus underlying biology. For this reason, the Oxford and Australian research groups independently started recruiting families in the mid—1990s, culminating in the collaborative IES which comprised a uniquely large collection of nearly 1200 families with multiple endometriosis cases. At the start of the collaboration, the phenotypes of all affected members in the families of both data sets were reviewed by the two gynaecologists collaborating in the two studies (S. K. and D. T. O’C.) to ensure that phenotypic definitions were standardized. This highlighted initial differences in recruitment strategies between the two data sets. Oxford focused on families with Stage B (rAFS Stages III / IV) disease at the outset, before including families with Stage A disease (rAFS Stages I / II) later on, whereas the Australian study collected both types from the outset. In addition, even though phenotypic staging criteria were standardized between the two studies, unidentified differences related to practicalities in recruitment protocols and the willingness of women to volunteer for studies or their inclination to be a member of an endometriosis self-help group all could have resulted in phenotypic differences between the data sets. The relative effect of these phenotypic differences on the linkage results in the two studies is unknown. 725 The results of the present analysis of families with three or more affecteds strongly suggest the presence of one or more genetic variants of low frequency predisposing to endometriosis in the Oxford data set. This is supported by non-parametric as well as parametric linkage results (either under dominant or recessive models). In addition, there appears some evidence for a disease allele at a similar location in the Australian data set, but this result is only supported when a recessive disease model is applied. On the one hand, these results could be interpreted as referring to the same signal. Since the Oxford data contained more affected mother – child pairs than the Australian data set, this would have meant that the former data set was more likely to support a dominant segregation model. This difference in family composition could reflect differences in ascertainment as well as in diagnostic patterns at a population level. Moreover, discrimination between different sets of parameters as estimated by the MOD score method may be difficult and different sets of disease model parameters could provide similar MOD scores (Clerget-Darpoux et al. , 1986). On the other hand, the results could represent genuine genetic heterogeneity between the two data sets. This would imply that there are multiple disease variants with both recessive and dominant inheritance patterns in the region of interest on chromosome 7 in the Oxford data set, whereas only one or more variants with a recessive segregation pattern are present in the Australian data set. Alternatively, the predisposing ...