No evidence of linkage to HPC20 on chromosome 20q13 in hereditary prostate cancer
Recently, Berry et al. 1 found evidence for linkage to chromosome 20q13 on the basis of a genomewide search involving 162 North American families with Ն3 members affected with prostate cancer. The highest 2-point LOD score was 2.69 for marker D20S196 at a recombination fraction of 0.20. Under the assumption of heterogeneity, heterogeneity LOD score (HLOD) only reached 1.08 with an estimated 12% of the families demonstrating linkage. Nonparametric statistical analysis supported this evidence of linkage, with a maximum multipoint NPL score of 3.02 (p ϭ 0.002) at D20S887. This locus was called HPC20 for hereditary prostate cancer on chromosome 20. To confirm this localisation, we selected 3 markers located in the HPC20 candidate region and analysed 66 families with 3 or more prostate cancer cases. Sixty-four families were of European origin: 61 were French, 2 Spanish and 1 Italian. 2 The 2 other families originated from North Africa: 1 from Tunisia and 1 from Algeria. The average number of affected individuals per pedigree was 3.74 (range 3-7). The number of genotyped individuals was 225 including 163 affected men (mean 2.5 per family; range 2-5). The average age at diagnosis was 66.7 years (range 48 -85 years). The 3 selected markers D20S887, D20S196 and D20S857 spanned 5.7 cM in the candidate region of HPC20. Polymerase chain reaction (PCR) products, amplified with 1 fluorescently labeled primer, were resolved on an ABI-prism 377 automated fluorescent DNA sequencer (Perkin-Elmer, California) by electrophoresis in 5% denaturing acrylamide gels. Genotypes were determined with Genescan Analysis 3.1 and Genotyper 2.5 softwares (Perkin-Elmer). As in Berry et al., 1 linkage analyses were performed with both parametric and model-free methods. Parametric 2-point LOD scores were calculated with the package LINKAGE using FASTLINK 3.0 implementation. 3,4 Parametric and non parametric multipoint were computed by use of GENEHUNTER. 5 For the parametric analyses, 2 different models, previously described in Berthon et al., 6 were used. In brief, they all assume an autosomal dominant mode of inheritance with a disease allele frequency of 0.003. In these models, 4 age-dependent penetrance classes were established with a penetrance reaching 90% at 70 years of age. The 2 models differed by the phenocopy rate, which was set at 0.01 in all age classes for model M1 and 10% of the susceptible-genotype penetrance for model M2. Allele frequencies were estimated from the data set.
Considering the whole subset of families, no evidence of linkage was found. Two-point parametric LOD scores were negative for the 3 markers with model M1 and only reached 0.19 at a recombination fraction () of 0.2 with model M2 at D20S857. Multipoint parametric LOD scores were negative with both models. Assuming heterogeneity, maximum multipoint HLOD scores were 0.03 and 0.11 for models M1 and M2, respectively, with an estimated 4% and 18% of linked families (upper 95% CI of 30% and 80%, respectively). Model-inde-