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

Background 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. Increasing evidence suggests that mutations in PTP genes are associated with human malignancies. For example, mutational analysis of the tyrosine phosphatase (PTP) gene superfamily uncovered genetic alterations in about 26% of colorectal tumors. Since in these studies tumors have not been stratified according to genetic instability status we hypothesized that colorectal tumors characterized by high-level of microsatellite instability (MSI-H) might show an increased frequency of frameshift mutations in those PTP genes that harbor long mononucleotide repeats in their coding region (cMNR). Results Using bioinformatic analysis we identified 16 PTP candidate genes with long cMNRs that were examined for genetic alterations in 19 MSI-H colon cell lines, 54 MSI-H colorectal cancers, and 17 MSI-H colorectal adenomas. Frameshift mutations were identified only in 6 PTP genes, of which PTPN21 show the highest mutation frequency at all in MSI-H tumors (17%). Conclusion Although about 32% of MSI-H tumors showed at least one affected PTP gene, and cMNR mutation rates in PTPN21, PTPRS, and PTPN5 are higher than the mean mutation frequency of MNRs of the same length, mutations within PTP genes do not seem to play a common role in MSI tumorigenesis, since no cMNR mutation frequency reached statistical significance and therefore, failed prediction as a Positive Selective Target Gene.

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][28][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.

Cell Lines and Tumor Tissues
Most of the cell lines analyzed in the present study have been described previously [15]. Additional human colorectal cancer cell lines were obtained from Cell Line Services (CLS), Heidelberg, Germany (Colo94H, Colo205, HCT8, SW1116, SW403, T84) or kindly provided by Dr. M. Brattain, University of Texas, Health Science Center, San Antonio, TX (CBS, FET), Dr. J. Wilson, Case Western Reserve University, Cleveland, Ohio (Vaco5, Vaco6, Vaco432, Vaco457), or Dr. I. Fidler (KM12). Cells were grown in RPMI supplemented with 10% fetal calf serum (Life Technologies, Karlsruhe, Germany). Formalin-fixed and Paraffin-embedded tumor and matched normal mucosae samples were treated as described [38]. The MSI status of tumor cell lines (19 MSI-H, 17 MSS), primary colorectal tumors (54 MSI-H and matched normal mucosae), and colorectal adenomas (17 MSI-H, 6 MSS, and matched normal mucosae) has been determined using the NCI/ICG-HNPCC microsatellite marker panel [39] as previously described [38]. Whole blood DNA samples of 60 healthy donors served as additional MSS controls. Informed consent was obtained from all patients and blood donors.

Candidate cMNR Sequences in PTP Genes
Human PTP genes were sought in our cMNR database ( [15] accessible at http://www.seltarbase.org, MNR_ensembl) that covers all human candidate coding mononucleotide repeats in the entire human genome annotated in the Ensembl database (Ensembl 19.34b, 2004). Candidate cMNRs with repeat length of at least 7 repeat units were considered for further analysis.

cMNR Instability Analysis
Primer design, DNA fragment and data analysis was performed as described [40]. Primer sequences and anneal-ing temperatures are shown in Table 1 [see Additional file  1]. Frameshift mutations were verified by DNA sequence analysis.

cMNR instability of PTPs in MSI-H colorectal cancer cell lines
A candidate set of 16 PTP genes that contained cMNR sequences composed of at least 7 repeat units were retrieved from our cMNR database (  figure 1. The highest mutation frequencies were associated with the A 8 cMNRs in PTPN13 (22%) and PTPN21 (26%). All mutations occurred in a heterozygous state and predominantly represented single nucleotide deletions (minus 1 allele, n = 10). Two were single nucleotide insertions (plus 1 allele, PTPN13 and PTPN21). No larger shifts could be detected. Overall, the larger fraction of MSI-H colorectal cancer cell lines (63%; 12/19) showed frameshift mutations in any of the four affected PTP genes.

PTP cMNR Mutations in MSI-H Primary Colorectal Carcinomas
We next analyzed the same set of 16 PTP genes for cMNR frameshift mutations in 54 MSI-H colorectal carcinomas and adjacent normal mucosa. Frameshift mutations in Electropherograms of allelic shifts  Figure 1). In MSS mucosa samples such changes almost never occurred (1/703 analyses). Two genes, PTPN21 and PTPRS, were most frequently affected in primary colorectal tumors (16% and 12%, respectively). However, primary tumors lacked frameshift mutations in the A 8 repeat of PTPN13 and thus appeared to arise only in cultured MSI-H CRC cell lines. Since LS174T and LS180 are derived from the same primary tumor [41,42] and the overall mutation frequency is quite low it is very likely that this mutation already manifested within the primary tumor. Vice versa, mutations in three other PTPs (PTPN5, PTPRA, PTPRE) remained restricted to primary tumors and occurred at very low frequency (Tables 3 and 4 [see  Additional file 1]). Overall, frameshift mutations in any of these 6 PTP genes were observed in 32% of MSI-H tumors.
The most common type of mutation were single nucleotide deletions (16/21, 76%) and all frameshift mutations appeared to affect only one allele (heterozygous). Interestingly, all PTP mutations are located within or upstream of the phosphatase domains and hence are expected to impair or completely abrogate phosphatase activity ( Figure 2).

PTPN21 cMNR Mutations in MSI-H Colorectal Adenomas
Since mutations in the A 8 Figure 1). These results rather argue for an early event during MSI colorectal carcinogenesis.

Discussion
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 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 [A 8 ], PTPRS [C 7 ], PTPN5 [C 7 ]) show mutation frequencies of 17%, 12%, and 6%, respectively. The final overall mutation frequency of PTPN21 A 8 (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.

Conclusion
The work presented here shows a systematic investigation using an evidence based approach: SK carried out the molecular analyses, participated in the data analysis and interpretation, and drafted the manuscript, SMW participated in the design of the study, in the data analysis and interpretation, performed the statistical analysis, and contributed to manuscript writing, YPY and PB did the bioinformatic analysis and established the human cMNR database and the list of candidate genes, MvKD provided project supervision, JG provided the general concept, design of study, supervision and contributed to manuscript writing.
All authors read and approved the final manuscript. 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. 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. For abbreviations see legend of Table 2.