Frameshift mutations in coding repeats of protein tyrosine phosphatase genes in colorectal tumors with microsatellite instability

Chromosomal instability (CIN) and aneuploidy are molecular features of most sporadic colorectal cancers (~85%) and may confer a worse prognosis [1‐3]. About 15% of colorectal cancers (CRC) show microsatellite instability (MSI) due to defective DNA mismatch repair (MMR; [4]). In hereditary non-polyposis colorectal cancer (HNPCC/Lynch syndrome; about 5% of all CRC cases) most of the tumors display this MSI phenotype [5]. As a common molecular theme, MMR-deficient MSI tumors of the colon and other organs accumulate numerous insertion/deletion mutations [6, 7] not only at non-coding but also at coding microsatellites (cMS) that cause translational frameshifts and abrogate normal protein function. Such frameshift protein derived neo-peptides can be highly immunogenic and are capable to induce cytotoxic T-cell-mediated killing of MSI-H tumor cells in vitro [8‐11]. Frameshift mutations in cMS sequences of a large number of candidate genes have been identified [12‐16] and mutations in some of them (TGFBR2, ACVR2, BAX; TCF-4) appear to provide a growth advantage to affected cells [17‐20]. Both, sporadic and HNPCC-associated colorectal MSI-H cancers, show distinct clinico-pathological characteristics that include frequent proximal site, diploidy, poor differentiation, less distant metastases, peritumoral lymphocytic infiltrate, comparably good prognosis, and altered chemoresponsiveness [6, 7, 21‐26]. Increasing evidence suggests that cMS mutations in a limited number of target genes appear to be selected for during MSI carcinogenesis and might account for some of these clinico-histopathological features.

Protein tyrosine phosphatases (PTPs) like their antagonizing protein tyrosine kinases are key regulators of signal transduction thereby assuring normal control of cellular growth and differentiation [27]. Alterations in the delicate balance between tyrosine phosphorylation and dephosphorylation contribute to the pathogenesis of different inherited or acquired human diseases including autoimmunity, diabetes, and cancer [27‐29]. Several studies indicate that mutations in PTP genes may be involved in colorectal carcinogenesis. For example, increased PTPRA mRNA levels have been observed in late stage colorectal tumors [30] and frequent overexpression of the human transmembrane-type PTP SAP-1 may occur relatively late in the adenoma-carcinoma sequence [31]. Expression profiling studies also suggested that PTPs appear to be involved in metastasis of colorectal cancer [32]. In a similar approach, differential expression of the human PTPN21 gene was observed when comparing MSI-H with microsatellite stable (MSS) colorectal cancer cell lines [33] and mutations in this gene were reported to occur in a subset of MSI-H colorectal carcinomas [34]. Additionally, a somatic mutation in the non-receptor PTP Shp2, encoded by the PTPN11 gene, has been detected in a single colon tumor with an increased frequency of somatic alterations, but without microsatellite instability [35]. Moreover, identification of the murine PTP gene Ptprj as a modifier locus conferring susceptibility to colorectal cancer also led to the detection of frequent deletions of the human PTPRJ gene in primary colon cancers [36]. Finally, systematic mutational analysis of the human PTP gene super family identified somatic mutations in six PTPs (PTPRF, PTPRG, PTPRT, PTPN3, PTPN13, PTPN14), affecting 26% of colorectal cancers [37].

However, whether coding mononucleotide repeats (cMNR) in PTP genes are specific targets of frameshift mutations in MMR-deficient colorectal tumors is still unknown. In the present study we identified 16 human PTP genes harboring coding region microsatellites and determined their mutation frequencies in MSI-H colorectal tumors. About 32% of MSI-H tumors showed frameshift mutations in any of these PTP genes. However, gene-specific cMNR mutation frequencies did not reach statistical significance according to our recently proposed model for Selective Target Gene prediction [16]. Hence, there is no significant statistical support for a common involvement of any of these PTP genes in MSI colon carcinogenesis.

The goal of this study was to determine type, frequency, and pattern of cMNR frameshift mutations in protein PTP genes in MSI-H colorectal cancer cell lines and primary colon tumors. This gene family was selected because PTPs are expected to exert tumor suppressive function and thus represent potential targets of inactivating mutations in DNA mismatch repair deficient tumors. Among the large number of PTP genes in the human genome we selected 16 PTP genes that contained coding region microsatellites of increased length (≥ 7 repetitive units). Our results provide evidence for frameshift mutations in six of these PTP genes (PTPN21, PTPRS, PTPN5, PTPN23, PTPRA, PTPRE) affecting about 32% of MSI-H colorectal cancers. This frequency differs slightly from a previous study that reported PTP mutations in 26% of unselected colorectal carcinomas [37]. Although most of the coding repeat-harboring PTP genes (13/16 candidates) analyzed by us have also been examined by Wang and co-workers, different sets of mutated PTP genes were identified in both studies. In particular, we observed tumor-specific genetic alterations in PTPN21, PTPN23, PTPN5, PTPRA, PTPRE and PTPRS whereas mutations reported by Wang et al. were confined to PTPRF, PTPRG, PTPRT, PTPN3, PTPN13, and PTPN14. This difference most probably can be explained by the different screening strategies used in these two studies. Wang et al. performed their initial mutational pre-screening on a set of 18 unselected colorectal tumors, This subset would be expected to include only 2 to 3 MSI tumors based on a general MSI frequency of 15% reported for unselected colorectal cancers. Since the type of mutation and the spectrum of affected genes is remarkably different among MSS and MSI-H tumors, the observed differences are not unexpected. However, there were also some consistent findings in both studies. For example, PTPN13 was examined by both groups but no cMNR mutations were observed in a cumulative number of 69 MSI-H CRC samples ([43] and data presented here). Interestingly, Wang et al. described two nonsense or frameshift mutations at the expected MNR position within PTPN13. Unfortunately, any information about the MSI status of these tumors is missing. So, our investigation represent a reasonable completion of a systematic analysis of human PTP genes in human colorectal cancers.

Recently, we proposed a statistical model that allows to predict positively or negatively selected target genes of MSI tumorigenesis in an organ-specific manner based on cMNR mutation frequencies [16]. According to the actual release of this model (http://www.seltarbase.org, release 200711), cMNRs of 7 or 8 repetitive units – in the absence of any biological selection pressure – are expected to show a mean somatic mutation rate of about 4% and 9% respectively. In order to qualify as positively selected MSI target genes, tumor-specific mutation frequencies for cMNRs of this length have to exceed 22% and 28%, respectively. The three most frequently mutated PTPs identified in the present study (PTPN21 [A₈], PTPRS [C₇], PTPN5 [C₇]) show mutation frequencies of 17%, 12%, and 6%, respectively. The final overall mutation frequency of PTPN21 A₈ (14.4%) is even slightly lower, when considering data from the literature [14]. Certainly, we did not extend our mutation search to the entire coding sequence of each PTP gene, and therefore, additional mutations outside the cMNR sequence cannot be excluded. Furthermore, detailed studies of individual tumor types have provided compelling evidence that mutations of different genes but within the same pathway can have similar functional effects, i.e. leading to its disruption and providing a growth advantage to affected cells [37], as it has been proposed for apoptotic genes like Fas, Apaf-1, and Bcl-10 [44]. In conclusion, although these gene-specific mutation rates are higher than the mean mutation frequencies for this length of repeats they do not reach statistical significance and prediction about positive Selective Target Genes in MSI-H tumorigenesis is not feasible. Therefore, human PTP genes do not seem to play a common role in MSI-H tumorigenesis. However, contribution to individual tumor development cannot be excluded. Furthermore, the observation of 2 mutations within a set of 17 MSI-H adenomas in PTPN21 argues for an early event in malignant transformation.

Despite some structural diversity, the six PTP proteins found to be mutated in the present study share a catalytically active phosphatase that resides in one (PTPN21, PTPN23, PTPN5, PTPRS) or two (PTPRA) C-terminal domains [45]. Notably, all cMNR frameshift mutation sites are located upstream or within the first half of these catalytic domains (see Figure 2) resulting in truncated proteins that are expected to show partial or complete loss of phosphatase activity. However, all PTP frameshift mutations only affected a single allele leaving the cMNR on the remaining allele intact. The presence of contaminating normal inflammatory cells in these tumors may well account for this observation. Alternatively, the remaining wildtype copy may be silenced by epigenetic mechanisms. The absence of biallelic mutations in the analyzed tumors could be due to a dominant negative fashion or affection of gene dosage [46, 47]. Shortened transcripts of murine PTPs lacking catalytic or interaction domains by alternative splicing can act in a dominant negative manner [48].

From our analysis of preneoplastic lesions we further conclude that PTPN21 frameshift mutations also occur in MSI-H colorectal adenomas albeit at lower frequency (12%) thereby indicating an early step during MSI tumorigenesis. At the protein level, PTPN21 is known to bind to and activate c-Src and Etk protein kinases [49, 50]. Interestingly, activated Etk has been reported to trigger apoptosis in breast cancer cells via Stat1 and p21 [51]. By analogy, we hypothesize that inactivation of PTPN21 by cMNR frameshift mutations in MSI-H colon cancer cells might lead to Etk inactivation and subsequent inhibition of apoptosis.

The work presented here shows a systematic investigation using an evidence based approach: cMNRs of human PTP genes with the highest chance for mutational events in MSI-H colorectal tumors were selected. Although we observed a similar overall mutation frequency in MSI colorectal cancers compared to the results reported by Wang et al. for unselected colorectal cancers, a different set of PTP genes appears to be mutated in MSI tumors. Human PTP genes identified in the present study seem to play no common role in the tumorigenesis of MSI-H tumors, supported by statistical considerations. However, within individual MSI-H carcinogenesis such mutations could have had impact to malignant transformation, but so far no functional data exist that could confirm this hypothesis.

Table 1: Coding microsatellites in candidate PTP genes examined in this study.

For each of the 16 PTPs is given the official HUGO identification (hugoID), the actual Ensembl entry ID at http://www.ensembl.org/Homo_sapiens (ENSG), the chromosomal localization (Chr.), the former EMBL accession number (Acc. no.), the type of tract (nucleotide and length, cMNR), the exact position of the tract in relation to the sequence of the EMBL entry (Pos.), the annealing temperature for PCR (Ta) as well as the primer sequences (5' -> 3', sense and antisense) used for fragment analysis of the respective tract.

Table 2: Allelic cMNR mutation status in MSI-H colorectal cancer cell lines.

Wild type: wt; Deletions of one or two mononucleotides: -1, -2; Insertions of one or two mononucleotides: +1, +2; Analysis failed or not evaluable: 0.

Table 3: Allelic cMNR mutation status in MSI-H colorectal carcinomas.

For abbreviations see legend of Table 2.

Table 4: PTP cMNR mutation frequencies in MSI-H colorectal cancer cell lines, tumors, and adenomas*.