Exome Sequencing Identifies Biallelic MSH3 Germline Mutations as a Recessive Subtype of Colorectal Adenomatous Polyposis

Abstract

In ∼30% of families affected by colorectal adenomatous polyposis, no germline mutations have been identified in the previously implicated genes APC, MUTYH, POLE, POLD1, and NTHL1, although a hereditary etiology is likely. To uncover further genes with high-penetrance causative mutations, we performed exome sequencing of leukocyte DNA from 102 unrelated individuals with unexplained adenomatous polyposis. We identified two unrelated individuals with differing compound-heterozygous loss-of-function (LoF) germline mutations in the mismatch-repair gene MSH3. The impact of the MSH3 mutations (c.1148delA, c.2319-1G>A, c.2760delC, and c.3001-2A>C) was indicated at the RNA and protein levels. Analysis of the diseased individuals' tumor tissue demonstrated high microsatellite instability of di- and tetranucleotides (EMAST), and immunohistochemical staining illustrated a complete loss of nuclear MSH3 in normal and tumor tissue, confirming the LoF effect and causal relevance of the mutations. The pedigrees, genotypes, and frequency of MSH3 mutations in the general population are consistent with an autosomal-recessive mode of inheritance. Both index persons have an affected sibling carrying the same mutations. The tumor spectrum in these four persons comprised colorectal and duodenal adenomas, colorectal cancer, gastric cancer, and an early-onset astrocytoma. Additionally, we detected one unrelated individual with biallelic PMS2 germline mutations, representing constitutional mismatch-repair deficiency. Potentially causative variants in 14 more candidate genes identified in 26 other individuals require further workup. In the present study, we identified biallelic germline MSH3 mutations in individuals with a suspected hereditary tumor syndrome. Our data suggest that MSH3 mutations represent an additional recessive subtype of colorectal adenomatous polyposis.

Keywords: adenomatous polyposis; candidate genes; exome sequencing; familial colorectal cancer; hereditary tumor syndromes; massive parallel sequencing; mismatch repair.

Figure 1

Pedigrees of the Two Index Individuals with Biallelic MSH3 Germline Mutations Pedigrees of family 1275 (A) and 1661 (B). The index persons are indicated by arrows (see main text for details). Above the symbols, identifiers are given for affected individuals. The number on the upper right side of a symbol displays the age at death, or in living persons, the age at last contact. On the lower right, genotype and phenotype information is displayed. The numbers following a disease represent the age at first diagnosis. Abbreviations are as follows: ad, adenomas; CRC, colorectal carcinoma; CUP, cancer of unknown primary; duod, duodenal; GC, gastric cancer; LC, lung cancer; polyps, multiple colorectal adenomatous polyps; and yrs, years.

Figure 2

Germline Mutations in MSH3 All four MSH3 mutations identified in the present study were validated by Sanger sequencing of the respective exons with flanking intronic regions. Their protein-level effects are depicted below in an adapted lollipop plot of MSH3 (created with MutationMapper). A premature stop is symbolized by a red star, a purple bar denotes altered amino acids, and a red bar denotes lost amino acids. Regions with conserved functional residuals are highlighted. Abbreviations are as follows: DD, dimerization domain; DNA-Rc, DNA-recognition; and MutS_II-III, PFAM domains of the MutS family.

Figure 3

Transcript Analyses of MSH3 Splice-Site Mutations and Confirmation of Compound Heterozygosity (A) An agarose gel shows RT-PCR products obtained from mRNA of individual 1275.1 (lane 1) and control samples (lanes 2–3); primers are localized in exon 21 (forward) and exon 23 (reverse). In the affected individual’s sample, a shorter band of around 165 bp is visible in addition to the expected band of 295 bp. Sequencing of the shorter fragment demonstrated a loss of exon 22 caused by the c.3001−2A>C mutation. (B) An agarose gel shows RT-PCR products obtained from mRNA of individual 1661.1 (lane 1) and control samples (lanes 2–3); primers are localized in exon 16 (forward) and exon 18 (reverse). In the affected individual’s sample, a shorter band of around 162 bp is visible in addition to the expected band of 279 bp. Sequencing of the shorter fragment demonstrated a loss of exon 17 caused by the c.2319−1G>A mutation. (C) The mutation sites in family 1275 were amplified and sequenced from lymphocyte DNA of an unaffected sibling. Only one of the mutations was detectable in the heterozygous state, confirming that the two mutations are located on separate alleles. (D) Agarose gel shows RT-PCR products obtained from mRNA of individual 1661.1 (lane 1) and control samples (lanes 2–3); primers are localized in exon 16 (forward) and exon 21 (reverse) in order to capture the effects of the splice-site mutation and the frameshift mutation within the same amplicon. Sequencing the shorter band proved that the product of aberrant splicing does not carry frameshift mutation c.2760delC in exon 20, whereas the frameshift mutation is present in the longer band. This confirms that the two mutations are located on distinct transcripts and thus separate alleles (compound-heterozygous state).

Figure 4

IHC Staining of MSH3 in Tumor and Normal Tissue MSH3 was stained with rabbit polyclonal antibody targeting N-terminal human MSH3. (A) MSH3 was nearly undetectable in FFPE normal colon mucosa and colorectal adenoma samples from index person 1275.1 and a colorectal adenoma sample from index person 1661.1, who has a complete loss of nuclear MSH3. (B) In contrast, control samples, which were taken from FFPE normal mucosa of an independent subject with colon cancer and were processed in parallel, show a strong nuclear MSH3 staining. The same results were obtained in an examination of a second independent colorectal adenoma from both index persons (data not shown). Scale bars represent 100 μm.

Figure 5

Results of Microsatellite Analysis Five tetranucleotide repeat markers (D20S82, D2S443, D21S1436, D9S747, and UT5037) were examined in normal (N) and tumor (T) tissue. In each tumor sample, two or three out of five markers showed clear instability and thus demonstrated elevated microsatellite instability at selected tetranucleotide repeats (EMAST). The x axes denote peak positions related to fragment lengths. Please note that only x axes within tumor-normal pairs allow direct comparison. The y axes quantifying peak heights were generally not standardized. The results of both index persons were confirmed in a second colorectal adenoma (data not shown).