Hereditary non-polyposis colorectal cancer (HNPCC): phenotype-genotype correlation between patients with and without identified mutation

Affected members of hereditary non-polyposis colorectal cancer (HNPCC) families develop colorectal cancer at an early age (mean 45 yr) and frequently get extracolonic cancers particularly in the uterus, urinary tract, and small intestine. They have a high risk of developing more than one primary colorectal cancer if not treated with subtotal colectomy at first operation and have more frequent right-sided colon cancers and less frequent rectum cancers, compared to patients with sporadic colorectal cancer. We have screened 31 families fulfilling the Amsterdam criteria and 54 families with a colorectal cancer clustering but not fulfilling the Amsterdam criteria for mutations in MLH1 and MSH2 by direct sequencing, and detected a mutation in 61% of the Amsterdam positive families but only in 15% of the Amsterdam negative families. Genotype-phenotype correlation was compared between 141 affected individuals with an identified mutation and 78 affected individuals from Amsterdam positive families in which a mutation was not identifiable in MLH1 or MSH2. In the affected persons with identified mutations, all expected phenotypic traits were represented, whereas affected persons in whom no mutation was detected fell into two clearly distinguishable subgroups. The minor subgroup, in which no mutation was identified, generally had the same characteristics as found in affected persons with identified mutations. The major subgroup differed significantly in clinical features and exhibited phenotypic traits similar to those found in late-onset families, including abundance of rectal cancer, few HNPCC-related cancers, lower frequency of multiple colorectal cancers, and later age at onset. Finally, for six missense mutations and one single codon deletion, the pathogenic potential was evaluated by domain localization, lod score calculation or segregation analysis when possible, and mutation-induced biochemical change. The results indicate that the majority of missense mutations are pathogenic, although further characterization by functional assays is necessary before implementation in predictive testing programs.