Partial loss of heterozygosity events at the mutated gene in tumors from MLH1/MSH2 large genomic rearrangement carriers

Hereditary non-polyposis colorectal cancer, HNPCC (also known as Lynch syndrome), is a frequent, autosomal, dominantly-inherited predisposition mainly for early-onset of colorectal and endometrial cancers, and less frequently for other type of malignancies [1]. The clinical definition of HNPCC diagnosis is based largely on the Amsterdam criteria (AC), which were set up in 1991 with the purpose of selecting families for mutational screening of mismatch repair (MMR) genes, whose germline pathogenic mutations cause this condition [2, 3]. The less stringent Bethesda guidelines (BG), and later, the Revised Bethesda Guidelines (RBG) were established to identify additional HNPCC patients that do not fulfill the strict Amsterdam criteria [4]. The examination of tumor tissue of HNPCC-suspected patients for the presence of microsatellite instability (MSI) and/or alterations in the MMR protein expression patterns may be also useful in pre-selecting patients for mutational screening. Nevertheless, a precise distinction between familial and sporadic MMR deficient tumors before mutational screening still remains a complex issue [5] and not all of the clinico-genetical or tumor characteristics may be available for each case.

Identification of the predisposing germline mutation is important because it confirms the clinical diagnosis of HNPCC and enables targeted clinical surveillance, which significantly reduces cancer morbidity and mortality in Lynch syndrome families [6]. Taking into account the number of potentially mutated MMR genes (MLH1, MSH2, MSH6, PMS2, PMS1, MLH3) and different types of mutations, the screening approaches are very complex and involve many techniques. Despite the tremendous advances in molecular diagnosis of HNPCC that have been made since the discovery of genetic clues of this disease, the genetic etiology of many clinically defined HNPCC patients remains unsolved because of technical limitations and/or difficulties in confirming the pathogenicity of identified gene alterations [7].

Inherited large genomic rearrangements (LGRs) in the MMR genes, which are not detectable by commonly applied DNA sequencing, became a subject of numerous HNPCC studies in recent years. One of the goals of these studies was to determine the frequency of these mutations in HNPCC patients. Although direct comparisons of frequencies determined in the different studies are not possible due to variability in the way the cohorts were selected, it seems that in some populations, LGRs are more frequent than in others. LGRs are reported to comprise 10% to 55% of all MMR gene mutations, and most occur in major MMR genes; i.e. the MLH1 and MSH2 genes [8–17]. An exceptionally low frequency of LGRs in these two genes (<1.5%) was reported in a study of the Spanish population [18], though a higher frequency of LGRs was found in a Basque Country population [19]. The frequency of LGRs may be remarkably higher in certain populations due to founder effects [9, 10, 19–24]. There is no consensus on how to select patients for LGR-screening. In many studies, the patients who were previously screened by conventional screening/scanning methods (DGGE, HDA, DHPLC, DNA sequencing), and in whom no germline mutation could be found, are analyzed for the presence of LGR [25]. However, the examination of LGR prior to the laborious exon-by exon mutation scanning of the MMR genes has also been suggested and is feasible by applying simple and robust techniques, such as MLPA (multiplex ligation-probe dependent amplification) [18]. Knowing the frequency of LGR in the population can significantly influence the screening algorithms for patients at risk for HNPCC.

Contrary to the germline LGRs, somatic LGRs in the MMR genes, which represent potential MMR inactivating events referred to as second hits during tumorigenesis, are rarely studied. The heterozygous LGR region present in the germline DNA can be used as a marker to study loss of heterozygosity (LOH) in the corresponding tumor DNA. One study showed that tumors of the MLH1/MSH2 germline LGR carriers often display a somatic mutation identical to one that is present in the germline DNA [1]. Furthermore, LOH analyses of these tumors using microsatellite markers flanking the respective gene have revealed that loss of the wild-type allele predominantly occurs through gene conversion, rather than mitotic recombination or physical deletion of the respective gene locus. Although in general, little is known about gene-conversion events that occur in cancer [26], it is likely that the conversion tracts do not encompass a whole sequence of the MLH1 or MSH2 gene in the LOH-positive cases. These two genes carry a number of single nucleotide polymorphisms (SNPs) that can be potentially utilized to delineate the conversion tracts more precisely than is possible using the extragenic microsatellite markers. The SNP markers, though less heterozygous than the microsatellite markers, have been proven to be useful in detecting LOH at MMR loci [27, 28]. In their recent LOH study on the heterogeneously defined microsatellite unstable carcinomas, van Puijenbroek et al [29] used SNP arrays to assess LOH on a genome-wide level. Although, such commercial arrays are very attractive for LOH studies, they probably would not involve sufficient numbers of SNPs located within the MMR genes to enable a very detailed LOH study in the MMR gene of the LGR carriers.

To date, among 58 clinically well-defined Slovak HNPCC families, 21 clearly pathogenic MMR gene mutations have been identified by applying direct DNA genomic sequencing [30, 31]. As DNA sequencing fails to detect LGRs, it is possible that in some of these patients, an LGR is responsible for HNPCC. In the presented study, we used the MLPA technique to screen for LGRs in the MMR genes for the first time in the Slovak population. As a second step, we studied potential LOH events in those patients in whom the LGR was detected in the germline, by combining MLPA and DNA sequencing at the intragenic MLH1/MSH2-SNP markers.

The index cases of 37 families included in this study were analyzed for LGRs in the MMR genes using two MLPA kits, and 7 (19%) were found to have this type of alteration. Taking into account all mutations identified so far (28 mutations in total), including the mutations found in this study, the LGRs account for 25% of clearly pathogenic changes in MMR genes from Slovakian HNPCC families. The large deletion in the MLH1 gene, a del5-6, has not been reported in other populations and, as a single mutation, represents only 6.7% (1 of 15) of all MLH1 mutations, while LGRs in the MSH2 gene account for 50% (6 of 12) of all MSH2 mutation-positive cases [30–32] (one family with inherited MSH6 mutation is unpublished). Although large duplications in MMR genes seem to occur much less frequently than deletions [18, 39–43], we uncovered one duplication of exons 5 and 6 affecting the MSH2 gene in our relatively small cohort. A duplication of these two exons in the MSH2 gene has been reported [14].

Our results provide the first evidence that, like in many other studied populations, large genomic changes in the MLH1/MSH2 genes exist in Slovakian HNPCC families. In addition, this type of mutation is frequent, particularly in the MSH2 gene, implying that the mutation screening algorithm should begin with MLPA and not with DNA sequencing, especially in cases where the protein expression pattern of the tumor shows a loss of MSH2 protein or is unknown. For cases that lack detectable MLH1 protein expression in tumors analyzed by IHC, DNA sequencing of the MLH1 gene prior MLPA screening is more appropriate. LGR correlated strongly with MSI-H status; more than 83% tumors of LGR carriers displayed MSI-H. However, also high rate of AC-I patients with MSI-L tumors may be potentially LGR carriers (in our cohort ~33%). For patients with no FFPE tissue available for MSI analysis such as the case SK-25 in our study, MLPA assay may be considered if the patient fulfills strict AC criteria. Similarly, MMR gene expression results correlated with MLPA findings, indicative of MSH2 and MLH1 mutation.

Similar to what has been found in other studies, there was a high proportion 5/6 (83%) of deletions encompassing exon 1 among the MSH2 rearrangements [8, 11, 13, 41, 42, 44, 45]. Common origin, or founder effect, cannot be excluded for deletions of exons 1-3 detected in the patients, SK-21 and SK-28, and likewise of exons 1-7 identified in the patients, SK-14 and SK-25. The examination of haplotypes and/or breakpoints may help to determine whether this is the case. Unfortunately, the risk haplotype could not be traced in the families due to the small number of people available for examination. For deletions encompassing exon 1, determination of breakpoints by long range PCR is difficult given that 5' breakpoints may be located up to 200 kb upstream of the MSH2 transcription initiation site [44]. However, according to the preliminary results of the large on-going study to test a new methodological approach for breakpoint detection involving the LGR carriers of our work, the SK-14 and SK-25 cases do seem to share common breakpoints (Dr. Benno Röthlisberger, personal communication). For three out of five MSH2 deletion cases (SK-14, SK-21, and SK-25), we demonstrated that one breakpoint should lie in the TACSTD1 gene, which is upstream of the MSH2 gene. Current findings that some HNPCC patients carry aberrations exclusively in the TACSTD1 gene leading to the generation of TACSTD1/MSH2 fusion transcripts [46, 47] led to the creation of a new version of the P003 kit (B1 version), including TASCTD1 probes to enable simultaneous MLPA analysis of this gene and the major MMR genes (MLH1 and MSH2). Using the P008 kit (MSH6/PMS2) in our cohort, we detected copy number changes for the TASCTD1 gene only in conjunction with deletions of exon(s) in the MSH2 gene.

Although based on a small number of cases, our cumulative data from LOH analyses performed by various markers (heterozygous LGR region, SNPs, and microsatellites) indicate that partial LOH can take place at the mutated MLH1 or MSH2 loci in LGR carriers. We have adapted the term 'partial LOH' from LOH studies conducted on the chromosomal level, showing for instance partial LOH confined to one arm of a chromosome [48], and have used this term to signify concurrent retention of heterozygosity of at least one marker and loss of heterozygosity of at least one other marker within the gene. At this point, it is not clear whether partial LOH also occurs in HNPCC tumors of individuals carrying other types of MMR gene mutations, for instance point mutations. A varying frequency of LOH has been described in the literature for the MLH1 and MSH2 loci in the series of MMR gene mutation carriers. LOH at the MLH1 locus has been reported in 35-85% of all tumors with a germline mutation in the MLH1 gene [49–55]. LOH at the MSH2 locus has been described in 14-50% of all tumors with a germline MSH2 mutation [49, 51, 54, 56]. Given that partial LOH may be not detected when only a few markers are used, it is possible that the importance of LOH events at MMR genes in HNPCC tumors is still underestimated.

Schematic representations of partial LOHs in both genes are shown in Figure 5 and 6. The precise characterization of gene conversion events through the determination of initiating and terminating points is practically difficult; thus, annotation of the converted tract usually represents the minimal and maximal converted tracts (MinCT, MaxCT) [26]. In the tumors we examined, the overall heterozygosity of SNP markers was unexpectedly low, and thus allowed us to characterize converted tracts only partially. The microsatellite markers flanking the MLH1 or MSH2 gene that were tested in previous studies from our laboratory [32, 38] were also uninformative, due to the presence of instability in these MMR deficient tumors.

The underlying mechanism for partial LOH is unclear at the moment; however, it is likely that specific motifs surrounding DNA sequences involved in gene conversion [26] play a role in partial LOH. Such motifs include alternating purine and pyrimidine or polypurine and polypyrimidine tracts, minisatellite sequences, and chi-like sequences, all of which are present in the MLH1 and MSH2 genes. The only two chi sequences (5'-GCTGGTGG-3') present within the MLH1 gene, as well as the (TA)₁₂(T)₂₁ tract known as BAT-21, are coincidentally located in the region that we confined to the initiating point of the putative gene-conversion event in the SK-22 tumor (blue colored box in Figure 5). The regions of the MSH2 gene that we confined to the initiating points of the gene-conversion events in the SK-14 and SK-20 tumors are rich of polypyrimidine tracts (blue colored box in Figure 6). The Alu sequences, which are abundant in both MMR genes, are also considered to play a role in gene conversion; however, their action is attributed mainly to interlocus, rather than interallelic, gene-conversion events [26].

Understanding the nature of the second, i.e. somatic, hit in the respective MMR gene may be crucial for the understanding of cancer initiation in HNPCC patients. Interestingly, only in the case of SK-22 carrying an exon 5-6 del of the MLH1 gene, has the wild-type sequence been replaced by the mutated sequence encompassing a deletion of exons 5 and 6 and resulting in a homozygous alteration in the tumor. This LOH event should clearly lead to a complete inactivation of the MLH1 gene in the tumor which is in correlation with MSI and IHC findings. In the remaining two cases of MSH2 mutation, it is difficult to distinguish which allele served as the 'donor' (unaltered) and which as the 'acceptor' (altered). Since the replaced region does not harbor a genomic rearrangement observed in the germline, the role of such a 'second hit' in tumorigenesis remains unclear. Nevertheless, in one tumor (SK-20) that could be analyzed by MSI and IHC, presumably MMR inactivation had occurred as demonstrated by the presence of high level MSI (MSI-H) and loss of MSH2 expression. It cannot be excluded that the gene conversion-mediated sequence homogenization at several SNPs plays a role in tumorigenesis. Another explanation might be the presence of a 'third hit', which is difficult to fully assess in tumor-derived DNA. Clearly, future studies on larger sets of HNPCC patients with various types of mutations are needed to confirm the occurrence of partial LOH events in the MMR genes of HNPCC tumors and to elucidate the mechanism by which they arise and their roles in tumorigenesis.

In conclusion, given that LGRs of the MSH2 gene appear to be frequent in Slovakian HNPCC families, we recommend beginning mutational screening in this population with the MSH2/MLH1-MLPA kit. We have discovered partial LOH events at mutated genes in tumors from the MLH1/MSH2 large genomic rearrangement carriers. To the best of our knowledge, this finding is novel. For the fine mapping of conversion tracts in these genes, a large set of heterozygous markers is essential. Our results also extend the limited knowledge of the role of gene conversion in cancer.