Gene Conversion Is a Frequent Mechanism of Inactivation of the Wild-Type Allele in Cancers from MLH1/MSH2 Deletion Carriers

Hereditary nonpolyposis colorectal cancer (HNPCC) is an autosomal dominantly inherited cancer predisposition syndrome caused by germ line mutations in DNA mismatch repair genes, predominantly MLH1 and MSH2, with large genomic rearrangements accounting for 5% to 20% of all mutations. Although crucial to the understanding of cancer initiation, little is known about the second, somatic hit in HNPCC tumorigenesis, commonly referred to as loss of heterozygosity. Here, we applied a recently developed method, multiplex ligation–dependent probe amplification, to study MLH1/MSH2 copy number changes in 16 unrelated Swiss HNPCC patients, whose cancers displayed microsatellite instability and loss of MLH1 or MSH2 expression, but in whom no germ line mutation could be detected by conventional screening. The aims of the study were (a) to determine the proportion of large genomic rearrangements among Swiss MLH1/MSH2 mutation carriers and (b) to investigate the frequency and nature of loss of heterozygosity as a second, somatic event, in tumors from MLH1/MSH2 germ line deletion carriers. Large genomic deletions were found to account for 4.3% and 10.7% of MLH1 and MSH2 mutations, respectively. Multiplex ligation–dependent probe amplification analysis of 18 cancer specimens from two independent sets of Swiss and Finnish MLH1/MSH2 deletion carriers revealed that somatic mutations identical to the ones in the germ line occur frequently in colorectal cancers (6 of 11; 55%) and are also present in extracolonic HNPCC-associated tumors. Chromosome-specific marker analysis implies that loss of the wild-type allele predominantly occurs through locus-restricted recombinational events, i.e., gene conversion, rather than mitotic recombination or deletion of the respective gene locus. (Cancer Res 2006; (66)2: 659-64)

Hereditary nonpolyposis colorectal cancer (HNPCC) is an autosomal dominantly inherited cancer predisposition syndrome characterized by the occurrence of early onset colorectal carcinoma (CRC) as well as a defined spectrum of extracolonic tumors, such as cancers of the endometrium and renal pelvis (1). HNPCC is caused by germ line mutations in DNA mismatch repair (MMR) genes, predominantly MLH1 and MSH2, with 5% to 20% of mutations being large genomic rearrangements missed by conventional mutation screening techniques (2–4). Recently, the multiplex ligation–dependent probe amplification (MLPA) method has been introduced to assess DNA copy number changes semiquantitatively (5). This method requires considerably less DNA (50-200 ng) than conventional Southern blotting (5-10 μg) and, moreover, the short recognition sequence of the probes allows us to determine copy number changes in partially degraded DNA, such as DNA from formalin-fixed cancer tissue. According to Knudson's “two-hit” hypothesis, a second, somatic mutation, inactivating the wild-type allele, commonly referred to as loss of heterozygosity (LOH), is required for tumorigenesis to start (6). Subsequent MMR deficiency leads to accumulation of replication errors, mainly at short repetitive DNA sequences in the tumor cells, and gives rise to the molecular hallmark of HNPCC, microsatellite instability (MSI). Although crucial to the understanding of cancer initiation, only scarce data are available on the nature of the second hit in HNPCC tumors (7). Here, we applied the MLPA technique on 16 unrelated Swiss HNPCC patients, whose cancers displayed MSI and loss of MLH1 or MSH2 expression, but in whom no germ line mutation could be detected by conventional DNA sequencing. The study aimed (a) to determine the proportion of large genomic rearrangements in our set of Swiss MLH1/MSH2 mutation carriers and (b) to investigate the frequency and nature of LOH as a second, somatic event in HNPCC tumorigenesis in cancers from MLH1/MSH2 germ line deletion carriers.

Sixteen unrelated Swiss patients referred to the division of Medical Genetics because of clinically suspected HNPCC syndrome were included in this study. The cancers from these patients had been found to display MSI and loss of MLH1 or MSH2 expression (Table 1). Because no pathogenic germ line mutation in MLH1 or MSH2 could be detected by conventional DNA sequencing, all patients were investigated for the presence of large genomic rearrangements in their germ line. Subsequently, the presence or absence of the identified MLH1/MSH2 germ line deletion was assessed in 18 cancers (seven cancers from known Finnish germ line deletion carriers; refs. 8, 9). Written informed consent was obtained from all patients included in the study.

DNA extraction from peripheral blood and tumor tissue. DNA from peripheral blood was isolated by using a salting-out procedure described by Miller et al. (10). Prior to DNA extraction from tumor tissue, histopathologic classification of H&E stained, formalin-fixed tissue blocks was carried out and a, representative portion of the tumor with an average tumor content of ≥70% was scraped off. DNA extraction was done according to the QIAamp tissue kit protocol (Qiagen, Basel, Switzerland).

Analysis of MSI and MMR protein expression. Based on the recommendations of the National Cancer Institute workshop on MSI, a panel of five microsatellite loci (BAT25, BAT26, D5S346, D17S250, and D2S123) was used to assess MSI (11). The presence or absence of four MMR proteins (MLH1, MSH2, MSH6, and PMS2) in the tumor was examined by standard immunohistochemical techniques (12).

Analysis of LOH. LOH, also referred to as “allelic loss”, was investigated using the following flanking polymorphic microsatellite markers: D3S1304, D3S1263, D3S2338, D3S1266, D3S1277, D3S1300, D3S1566, and D3S1278 for the MLH1 locus; and D2S168, D2S165, D2S367, D2S391, D2S337, D2S2110, D2S286, D2S2333, and D2S347 for the MSH2 locus. LOH was scored at any informative marker if the area under one allelic peak in the tumor was reduced by >50% relative to the other allele, after correcting for the relative peak areas in leukocyte-derived constitutional DNA (13). PCR conditions for LOH analysis are available from the authors upon request. PCR products were analyzed on an ABI PRISM 310 Genetic Analyzer (Applied Biosystems, Rotkreuz, Switzerland).

MLPA. For the detection of aberrant copy numbers in the MLH1 and MSH2 genes in constitutional, leukocyte-derived, and tumor DNA, the SALSA P003 MLH1/MSH2 test MLPA kit (MRC Holland, Amsterdam, the Netherlands) was used (5). The kit contains probes for the 16 exons of MSH2 and the 19 exons of MLH1 as well as seven probes located on different chromosomes as controls. DNA samples from two known germ line deletion carriers (MLH1 exon1_10del and MSH2 exon8_15del) as well as from 10 healthy probands were used to confirm the sensitivity and specificity of the method. Each mutation was confirmed on a second, independently drawn, blood sample from the respective patients. All reactions were carried out according to the manufacturer's protocol. Fragment analysis was done on an ABI 310 capillary sequencer and results were analyzed using the Genescan and Genotyper software (Applied Biosystems) to identify the specific amplicons representing the respective exons and control loci. Peak areas and heights were then exported to a Microsoft Excel spreadsheet and calculations were done according to the method described by Taylor et al. (14).⁴

Fragments with high SD (≥20%) were omitted from further analysis. An average dosage quotient close to 1 is expected for individuals with two copies, whereas values close to 0.5 indicate loss of one copy. In tumor-derived DNA samples, inevitably containing some degree of contaminating normal tissue, values ≤0.3 implied loss of both copies. MLPA results indicative of a germ line or a somatic deletion were independently confirmed in at least one additional, independent experiment as well as independently drawn blood samples, and, where available, cDNA was used to assess the individual break points. All apparently single exon deletions were screened by direct DNA sequencing to exclude sequence variations within the ligation-probe binding site which can mimic single exon deletions (15, 16).

Statistical analysis. Statistical comparison of patients' features, encompassing phenotypic characteristics (gender, age at diagnosis, etc.), and mutational status, was done using the χ² and Fisher's exact test for categorical variables, or Student's t test for continuous variables, with all of the probabilities reported as two-tailed P values, considering P < 0.05 to be statistically significant.

In this study, 16 unrelated Swiss HNPCC patients without identified pathogenic germ line mutation in MLH1 or MSH2 were screened for the presence of large genomic rearrangements using the MLPA assay (Table 1; Fig. 1). Three out of eight (38%) index patients whose CRCs showed high MSI and loss of the MSH2 protein were found to harbor three different genomic deletions in MSH2. Among the eight index patients whose tumors had lost MLH1 expression, two (25%) were found to carry large genomic rearrangements in the MLH1 gene. In view of the late age at diagnosis (≥68 years), as well as an inconspicuous family history in four of the MLH1 mutation-negative patients, it is likely that these cancers are actually sporadic in origin, due to hypermethylation of the MLH1 promoter rather than due to an inherited germ line mutation (17). With mRNA available for further study, the consequence of the MLH1 deletion encompassing exons 7 to 9 (family 1806) could be assessed. By cDNA sequencing, we found this genomic deletion to result in direct joining of exons 5 to 11 (c.454_884del) leading to a frameshift and a first premature stop codon 11 amino acids downstream. Taken together, large genomic deletions account for 10.7% (3 of 28) of MSH2 and for 4.3% (2 of 47) of MLH1 mutations in our set of mutation-positive Swiss HNPCC families (n = 75). The remaining mutation-negative patients are currently under intensive study for mutations in regulatory regions and/or epimutations in these MMR genes.

Overall, genomic deletion carriers were statistically significantly later diagnosed of CRC (median, 59 years; IQR, 19.5; n = 5) compared with index patients carrying “conventional” MLH1/MSH2 mutations (median, 40 years; IQR, 14.0; n = 70; P < 0.006). Four out of six (67%) deletion carriers (Table 2) had, in addition to their CRC, developed extracolonic cancers: patient 1806/3, an ovarian cancer at age 52; patient 1835/1, an endometrial cancer at age 61; patient 2227/1, a urothelial carcinoma at age 60 and an astrocytoma WHO grade 3 at age 67; and patient 2264/1, a duodenal and an endometrial cancer at age 41 and 48, respectively. Compared with other index mutation carriers in our HNPCC database, the frequency of extracolonic HNPCC-associated cancers was statistically significantly increased (67% versus 20%; P < 0.03). With regard to other phenotypic properties (e.g., site of CRC or tumor stage) no further significant differences were observed between the two groups. With one exception, i.e., family 2227, in which HNPCC-related cancers were confined to one generation only, all genomic deletion carriers met the Amsterdam criteria I (Table 1). Because the phenotypic observations might have an effect on counseling and management of MLH1/MSH2 germ line deletion carriers, it is important that these findings are confirmed on larger patient sets.

Following the identification of the MLH1/MSH2 germ line deletion carriers, we applied microsatellite marker analysis and the MLPA assay to the cancer specimens of these patients to gain further insight into the frequency and the nature of the second, somatic mutational event, commonly referred to as LOH, involved in HNPCC tumorigenesis. A total of 11 formalin-fixed cancers from six genomic deletion carriers (seven CRCs, one ovarian, one endometrial, one kidney cancer, and one astrocytoma) were available for investigation.

As depicted in Table 2A, MLPA analysis revealed that four (57%) out of seven CRCs, as well as one astrocytoma, actually harbor somatic deletions identical to the ones identified in the germ line (three MLH1 and two MSH2) and evidenced by an average decrease in gene dosage (normalized ratio) from 0.45 (±0.09 SD; germ line) to 0.18 (±0.07 SD; tumor; Fig. 2). One colorectal and one urothelial carcinoma showed loss of one exon only. No copy number changes were detected in the remaining four tumors (two CRCs, one ovarian, and one endometrial cancer).

In order to further substantiate the high frequency of cancers harboring a somatic deletion identical to the one in the germ line, seven additional tumors (four CRCs, two endometrial, and one stomach cancer) from Finnish HNPCC patients carrying an MSH2 or MLH1 germ line deletion were investigated (Table 2B). Two out of four CRCs were found to carry identical, biallelic deletions which were absent in the remaining cancers. Thus, although based on an arguably small number of cases, our findings from two independent sets of patients indicate that the occurrence of somatic deletions identical to the ones in the germ line are a frequent event in HNPCC-related colorectal tumorigenesis (6 out of 11 CRCs; 55%). Remarkably, none of the specimens from the Swiss or Finnish tumors showed evidence of large somatic deletions encompassing the entire, respective, gene locus.

With regard to extracolonic cancers (n = 7), only one tumor, a grade 3 astrocytoma (patient 2227/1), was found to carry an identical, biallelic deletion. Intriguingly, two other cancers (cecum and kidney) from this patient, with an exon 8 to 11 germ line deletion, harbored an identical single exon deletion (exon 11; Table 2A). A false-positive MLPA result could be excluded by directly sequencing the ligation-probe binding site for exon 11. Both introns 10 and 11 of MSH2 comprise several Alu repeats of the AluSx subfamily, which have been shown to be involved in genomic rearrangements of MMR genes (18).

In order to distinguish between the possible mechanisms leading to homozygosity of the germ line mutation in the tumor, i.e., loss of the chromosome harboring the wild-type allele followed by chromosomal reduplication, mitotic recombination, or gene conversion (19), we investigated eight highly polymorphic short tandem repeat markers flanking the gene loci on the respective chromosome (chromosome 2 for MSH2, chromosome 3 for MLH1). As depicted in Supplementary Table S3 (supporting online material), none of the tumors showed allelic loss at the markers flanking the respective gene locus. Therefore, loss of the chromosome carrying the wild-type allele followed by chromosomal reduplication as well as mitotic recombination per se, extending from the gene locus to the telomere, could be ruled out as the underlying mechanism responsible for loss of the wild-type allele. Hence, these findings indicate that rather a locus-restricted event, i.e., gene conversion, has occurred in all cancers which are homozygous for the germ line mutation.

This might, at least in part, be explained by the presence of short interspersed nuclear elements, particularly Alu repeats, which have been shown in several studies to be involved in germ line MLH1/MSH2 locus rearrangements (3, 8, 18, 20). The recurrent pattern of somatic deletions identical to the germ line indicate that it is possible that Alu-mediated gene conversion at the MLH1 and MSH2 loci is frequently occurring in HNPCC-associated tumors and that it is not restricted to CRC only, as evidenced by our findings on kidney and brain cancer specimens. Moreover, inactivation of the wild-type allele by gene conversion seems to be independent of the type of germ line mutation.⁵

If gene conversion is indeed the predominant mechanism which leads to inactivation of the wild-type allele, the LOH frequency at the MLH1 or MSH2 loci would be expected to be low. This is supported by data from Kruse et al., who observed LOH in only one out of nine skin tumors from eight unrelated Muir-Torre patients with MSH2 mutations (7). Clearly, future studies on larger sets of HNPCC patients carrying MLH1/MSH2 mutations are needed to conclusively establish the frequency of gene conversion events in colorectal as well as extracolonic cancers. These investigations should also help to elucidate the mechanistic role of Alu-mediated recombination in the generation of the second, somatic mutation in tumors from MMR gene mutation carriers.

In conclusion, large genomic deletions in MLH1/MSH2 were found to account for 4.3% and 10.7% of MLH1 and MSH2 mutations, respectively, in our set of Swiss HNPCC families. Deletion carriers were statistically significantly later diagnosed of CRC and exhibited more extracolonic cancers when compared with “conventional” MLH1/MSH2 mutation carriers. Analysis of cancer specimens from two independent sets of Swiss and Finnish MLH1/MSH2 deletion carriers revealed (a) that somatic deletions identical to the ones in the germ line occur frequently (55%) in CRCs and (b) that this type of inactivation of the wild-type allele is also present in extracolonic HNPCC-associated tumors. Chromosome-specific marker analysis implies that loss of the wild-type allele predominantly occurs through locus-restricted recombinational events, i.e., gene conversion, rather than mitotic recombination or deletion of the respective gene locus.

Grant support: Swiss National Science Foundation (no. 3200-067571; K. Heinimann) and the Swiss Cancer League/Oncosuisse (no. 01358-03-2003; K. Heinimann); the Sigrid Juselius Foundation, the Finnish Cancer Society, and the Academy of Finland (P. Peltomaki); and the Science Foundation of Helsinki University (A. Lindroos).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank all patients for their participation in this study as well as their respective doctors for contributing clinical information and tumor specimens. We also thank Sibylle Bertschin, Marianne Haeusler, Ritva Haider, and Thomas Woodtli for excellent technical assistance as well as Waltraut Friedl and Elisabeth Mangold for the MLH1/MSH2 deletion control samples.