MSH3 and MLH3 encode non-canonical components of the DNA mismatch repair machinery implicated in the repair of insertion-deletion errors. Prior reports described colorectal cancer (CRC) and adenomas in individuals with bi-allelic germline MSH3 mutations and in a small number of bi-allelic MLH3 mutation carriers. The authors hypothesized that loss-of-function germline variants in MSH3 and MLH3 might act like other mismatch repair defects (for example constitutional mismatch repair deficiency or Lynch syndrome), producing CRC either from bi-allelic germline mutations or from heterozygous germline alleles that acquire somatic “second hits.”
The study examined nearly 12,000 CRC and multiple polyp cases and approximately 460,000 controls. Cancer genome sequencing was available for 2,023 patients, permitting direct analysis of somatic inactivation events, mutational burden and mutation signatures. The work reported here is a preprint and has not been peer reviewed; the authors explicitly caution against using these findings to change clinical practice.
Among the cases evaluated, one individual was identified with bi-allelic germline MSH3 mutations and another with bi-allelic MLH3 mutations. The authors report that the phenotypes of these bi-allelic carriers resemble some patients with constitutional mismatch repair deficiency (cMMRd).
The analysis found that heterozygous carriers of loss-of-function germline alleles had an increased risk of CRC. Specifically, MSH3 heterozygotes showed a 2.2-fold increased CRC risk (P = 6.6 × 10−5) and MLH3 heterozygotes a 1.6-fold increased risk (P = 0.028). The increased risk in heterozygotes was attributed to subsequent somatic inactivation of the remaining wild-type allele in tumours.
Sequencing of tumours revealed somatic “second hits” that inactivated the wild-type allele in heterozygous carriers, consistent with classical tumour suppressor behaviour. In some cases, a single somatic event inactivated both MSH3 and the nearby APC gene, suggesting that a shared structural or regional somatic event can simultaneously target a mismatch repair gene and a key colorectal tumour suppressor.
Tumours from heterozygotes that lacked detectable somatic second hits did not show hypermutation, supporting a model in which germline heterozygosity alone is insufficient to drive the hypermutant phenotype and elevated CRC risk without somatic inactivation.
All CRCs with MSH3 or MLH3 deficiency were reported to be microsatellite-stable yet hypermutant, indicating high overall mutation burden without classic microsatellite instability. There was a pronounced enrichment of deletions of two or more base pairs — an approximately 12-fold increase compared with controls — consistent with defective repair of insertion-deletion events.
A mutation signature labeled ID4 was usually present in tumours with MSH3 or MLH3 deficiency (P < 0.0001). The authors note that a causal association between specific mutations and ID4 has not previously been reported in human tumours. They also state that ID4 likely does not have a single aetiology and can result from deficiency of either MSH3 or MLH3.
The authors interpret the data to mean that heterozygous germline MSH3 and MLH3 alleles show incomplete penetrance but can increase CRC risk when a somatic second hit causes loss of wild-type function. The resultant tumour phenotype — microsatellite-stable but hypermutant with excess multi-base deletions and ID4 — is reported to phenotypically resemble PMS2-mutant Lynch syndrome in some respects.
As stated in the preprint, these findings are provisional and should not be used to guide clinical practice, patient management, health-related behavior or policy decisions until validated and peer reviewed.
Data sources included the National Genomic Research Library (Genomics England), UK Biobank, the All of Us Research Program and other cohorts (including CORGI and Swedish study participants). Access to many of these datasets is restricted and governed by the respective platforms' access and ethical policies; the paper provides guidance on how approved researchers may request access. Ethical approvals and committee references for the contributing studies are reported in the manuscript.
The research received funding from Cancer Research UK and the Wellcome Trust, among other sources acknowledged by the preprint platform. The authors declared no competing interests.
Note: all statements above are derived from the source preprint. Specific methodological details, per-patient data, variant lists and supplementary analyses are available in the original manuscript and supplementary materials; the preprint status means conclusions remain provisional pending peer review.