Jaskowski et al1 have confirmed our previous paper2 reporting an association between BAT26 stability and germline MSH2 deletion. Among seven highly unstable tumours with MSH2 deletion spanning exon 5 they found normal BAT26 sequences in 4. Although their analysis was based on a small number of tumours, they observed this phenomenon in 57% of cases, a percentage very close to the 68% we obtained in the analysis of 19 tumour DNAs.2 It is reasonable to suppose that the molecular mechanism leading to somatic loss of BAT26 sequences, as we described previously,2 is responsible for the observed BAT26 stability in both subsets of MSH2 deleted tumours.

In our current dataset including 29 MLH1- and 49 MSH2-mutated tumours (27 of which retaining exon 5), only two additional cases displayed BAT26 stability; these were from two patients who were heterozygous for point mutations in MSH2 and MLH1, respectively. On the contrary, in the study by Jaskowski et al1 6/48 (12.5%) samples from patients with MSH2 germline mutations with exon 5 retention and only one out of 53 (1.8%) samples from MLH1 mutation carriers were found to be stable at BAT26. A different prevalence of BAT26 instability was also detected in a series of moderate–high-risk tumours, for which distinction between MLH1- and MSH2-related cases was based on immunohistochemistry alone. In the hereditary tumours such differential BAT26 stability was associated with an overall lower degree of instability in MSH2 compared to MLH1-related cancers (69.9 and 81.5% unstable markers, respectively).

On the other hand, extrapolating data from some previous studies on different series of unstable tumours with MLH1 and MSH2 defects and considering only the markers belonging to the reference NCI panel, the average percentage of instability was 92 and 79.4%,3 92.3 and 86.8%4, 80 and 72%,5 for MLH1 and MSH2 tumours, respectively. Nonetheless, despite the constant slight instability excess of MLH1 tumours, statistically significant differences could not be highlighted, even when these data were pooled (89.8 and 80.1%; P=0.08, Fisher's exact test).

We also analyzed in more detail our MSI data on a total of 78 unstable tumours with unequivocal test results (among which 61 were colorectal adenocarcinomas); in this series, we could not find a higher instability in the MLH1 group. In fact, despite the inclusion of 14 samples with germline MSH2 intragenic deletion and wild-type BAT26 sequences, the average number of unstable markers per tumour was 79.6% in 29 MLH1-deficient tumours and 83.6% in 49 MSH2-deficient tumours (P=0.38, Fisher's exact test). Similar figures were obtained when the analysis was restricted to the 23 MLH1- and 38 MSH2-mutated unstable colorectal carcinomas only (78.3 and 86.8%; P=0.1, Fisher's exact test).

Histopathological features and mutational spectra of the tumours analyzed could account for the discrepancies between ours and Jaskowski et al's1 data.

In conclusion, we welcomed additional evidences of the limited usefulness of BAT26 alone for the detection of MMR deficiency, and we read with interest this letter suggesting that MSH2 mutation carriers have increased BAT26 stability overall, compared to MLH1 mutation carriers. It is noteworthy that while our study evidenced a strong association between BAT26 stability and exon 5 MSH2 loss, Jaskowski et al's1 data showed that the absence of BAT26 instability is indicative of a generic MSH2 mutation. If further confirmed, this could represent an additional element to properly address mutational analyses. It is unclear whether the observed BAT26 stability is a consequence of a lower overall degree of instability in the MSH2-mutated tumours or it is a locus-specific molecular phenomenon connected with the presence of BAT26 in the MSH2 gene. It will be very interesting to investigate this issue in more detail, for instance by analyzing several mono- and di-nucleotides on additional larger series, to verify the proportion of markers that escape instability in MSH2-deficient tumours.