A novel panel of short mononucleotide repeats linked to informative polymorphisms enabling effective high volume low cost discrimination between mismatch repair deficient and proficient tumours

Abstract

Somatic mutations in mononucleotide repeats are commonly used to assess the mismatch repair status of tumours. Current tests focus on repeats with a length above 15bp, which tend to be somatically more unstable than shorter ones. These longer repeats also have a substantially higher PCR error rate, and tests that use capillary electrophoresis for fragment size analysis often require expert interpretation. In this communication, we present a panel of 17 short repeats (length 7-12bp) for sequence-based microsatellite instability (MSI) testing. Using a simple scoring procedure that incorporates the allelic distribution of the mutant repeats, and analysis of two cohort of tumours totalling 209 samples, we show that this panel is able to discriminate between MMR proficient and deficient tumours, even when constitutional DNA is not available. In the training cohort, the method achieved 100% concordance with fragment analysis, while in the testing cohort, 4 discordant samples were observed (corresponding to 97% concordance). Of these, 2 showed discrepancies between fragment analysis and immunohistochemistry and one was reclassified after re-testing using fragment analysis. These results indicate that our approach offers the option of a reliable, scalable routine test for MSI.

Fig 1. Example distributions of read frequencies.

Relative frequencies of reads classified according to length are shown for MNRs LR46 (an 8bp long poly-A tract) and LR44 (12 bp poly-A) in an MSS sample (169259) and MSI-H sample (U179H03T). The abscissa represents the deviation from the reference sequence length in bp.

Fig 2. Example representing allelic bias.

Allele specific read frequencies and sizes are shown for LR46 in two samples from a patient who is heterozygous for a flanking SNP (rs6040079). U029N = normal somatic tissue, U029T = microsatellite unstable tumour.

Fig 3. Assessing the ability of single MNRs to discriminate between MMR proficient and deficient samples.

Fig 4. Establishing analysis parameters for MSI test.

(A) Relative frequency of reads carrying a deletion in MSI-H and MSS samples for the MNR LR44. (B) Analysis of allelic bias at the MNR LR44 for MSI-H and MSS samples stratified according to the proportion of reads showing deletions. Y-axis is represented in log-scale. MSI-H = Microsatellite instability- high. MSS = Microsatellite stable.

Fig 5. Classification of MMR status in an independent dataset of 70 CRC samples.

Classification using only deletion frequency data (A), only allelic bias data (B), and both parameters combined (C). MSI-H = Microsatellite instability- high. MSS = Microsatellite stable.

Fig 6. MMR status classification of the training set.

MSI-H = Microsatellite instability- high. MSS = Microsatellite stable.

Fig 7. In silico analysis of mixtures.

The abscissa represents the proportion of MSS reads from an MSI-H/ MSS sample pair. The ordinate represents the fraction of the 1224 pairs that were classified as MSI-H for a given proportion.