Mutational Analysis of the Tyrosine Phosphatome in Colorectal Cancers

See allHide authors and affiliations

Science  21 May 2004:
Vol. 304, Issue 5674, pp. 1164-1166
DOI: 10.1126/science.1096096


Tyrosine phosphorylation, regulated by protein tyrosine phosphatases (PTPs) and kinases (PTKs), is important in signaling pathways underlying tumorigenesis. A mutational analysis of the tyrosine phosphatase gene superfamily in human cancers identified 83 somatic mutations in six PTPs (PTPRF, PTPRG, PTPRT, PTPN3, PTPN13, PTPN14), affecting 26% of colorectal cancers and a smaller fraction of lung, breast, and gastric cancers. Fifteen mutations were nonsense, frameshift, or splice-site alterations predicted to result in truncated proteins lacking phosphatase activity. Five missense mutations in the most commonly altered PTP (PTPRT) were biochemically examined and found to reduce phosphatase activity. Expression of wild-type but not a mutant PTPRT in human cancer cells inhibited cell growth. These observations suggest that the mutated tyrosine phosphatases are tumor suppressor genes, regulating cellular pathways that may be amenable to therapeutic intervention.

Phosphorylation of tyrosine residues is a central feature of many cellular signaling pathways, including those affecting growth, differentiation, cell cycle regulation, apoptosis, and invasion (1, 2). This phosphorylation is coordinately controlled by protein tyrosine kinases (PTKs) and phosphatases (PTPs). Although a variety of PTK genes have been directly linked to tumorigenesis through somatic activating mutations (36), only a few PTP genes have been implicated in cancer (710). Moreover, it is not known how many or how frequently members of the PTP gene family are altered in any particular cancer type. We have systematically addressed these issues by comprehensive mutational analysis of the PTP gene superfamily in colorectal tumors.

The PTP gene superfamily is composed of three main families: (i) the classical PTPs, including the receptor PTPs (RPTPs) and the nonreceptor PTPs (NRPTPs); (ii) the dual specificity phosphatases (DSPs), which can dephosphorylate serine and threonine in addition to tyrosine residues; and (iii) the low molecular weight phosphatases (LMPs) (1). Using an approach similar to that described in a recent bioinformatic analysis of PTPs in the mouse genome (11), we employed a combination of Hidden Markov Models representing catalytic domains of members of the PTP superfamily to identify 53 classical PTPs (21 RPTPs and 32 NRPTPs), 33 DSPs, and one LMP in the human genome (12). This analysis revealed a set of genes representing all known human PTPs (13) as well as seven putative PTPs.

As an initial screen to evaluate whether these phosphatases are genetically altered in human cancer, we analyzed the coding exons of all 87 members of this gene superfamily in 18 colorectal cancers. A total of 1375 exons from all annotated RPTPs, NRPTPs, DSPs, and LMPs were extracted from genomic databases (12). These exons were amplified by polymerase chain reaction from cancer genomic DNA samples and directly sequenced with dye terminator chemistry (12). Whenever a presumptive mutation was identified, we attempted to determine whether it was somatically acquired (i.e., tumor specific) by examining the sequence of the gene in genomic DNA from normal tissue of the relevant patient.

From the 3.3 Mb of sequence information obtained, we identified six genes containing somatic mutations, including three members of the RPTP subfamily (PTPRF, PTPRG, and PTPRT) and three members of the NRPTP subfamily (PTPN3, PTPN13, and PTPN14). These six genes were then further analyzed for mutations in another 157 colorectal cancers. Through this strategy, we identified 77 mutations in the six genes, in aggregate affecting 26% of the colorectal tumors analyzed (table S1 and Fig. 1). Examination of these six genes in seven other tumor types identified PTPRT mutations in two of 11 (18%) lung cancers and two of 12 gastric cancers (17%), and PTPRF mutations in one of 11 (9%) lung cancers and one of 11 (9%) breast cancers. No mutations were identified in 12 pancreatic cancers, 12 ovarian cancers, 12 medulloblastomas, or 12 glioblastomas (table S1 and Fig. 1). In total, 83 nonsynonymous mutations were observed, all of which were somatic in the cancers that could be assessed (12).

Fig. 1.

Distribution of mutations in PTPRT, PTPN13, PTPN14, PTPRG, PTPRF, and PTPN3. Black arrows indicate location of missense mutations, red arrows indicate location of nonsense mutations or frameshifts, and boxes represent functional domains (B41, band 41; CA, carbonic anhydrase; FN3, fibronectin type III; IG, immunoglobulin; MAM, meprin/A5/PTPμ; PDZ, postsynaptic density, discs large, zonula occludans; PTPc, catalytic phosphatase domain). Black stars indicate PTPRT mutants evaluated for phosphatase activity (Fig. 2), and the red star indicates PTPRT mutant evaluated for effects on cell proliferation (Fig. 3).

Fifteen of the 83 mutations were nonsense, frameshift, or splice-site alterations, all of which were predicted to result in aberrant or truncated proteins. In 16 tumors both alleles of the phosphatase gene appeared to be mutated, a characteristic often associated with tumor suppressor genes. The majority of tumors with PTP gene mutations also contained mutations in KRAS or BRAF, and nine tumors contained alterations in previously reported tyrosine kinase genes (table S1). Thus, the mutant phosphatases identified in this study are likely to operate through cellular pathways distinct from those associated with previously identified mutant kinases.

Analysis of mutations in tumors is complicated by the fact that mutations can arise either as functional alterations affecting key genes underlying the neoplastic process or as nonfunctional “passenger” changes. The multiple waves of clonal expansion and selection that occur throughout tumorigenesis lead to fixation of any mutation that had previously occurred in any predecessor cell, regardless of whether the mutation was actually responsible for the clonal expansion. Two independent lines of evidence suggest that the sequence alterations we observed are functional. First, the ratio of nonsynonymous to synonymous mutations provides an indication of selection, as synonymous alterations usually do not exert a growth advantage. There were no somatic synonymous mutations detected in the colorectal cancers analyzed, resulting in a ratio of nonsynonymous to synonymous mutations of 77 to 0, much higher than the expected 2:1 ratio for nonselected passenger mutations (P < 1 × 10–6). Second, the prevalence of mutations in the coding regions of the analyzed genes was ∼19 per Mb of tumor DNA, similar to the prevalence of functional somatic alterations observed in other gene families [e.g., the tyrosine kinome (6)] and significantly higher than the prevalence of nonfunctional alterations previously observed in the cancer genomes (∼1 per Mb, P < 0.01) (14). These data support the idea that these mutations were the targets of selection during tumorigenesis.

The great majority of the nonsense and frameshift mutations (Fig. 1) would result in polypeptides devoid of C-terminal phosphatase catalytic domains, thereby leading to inactivation of the phosphatase. To evaluate whether tumor-specific point mutations alter phosphatase activity, we biochemically tested mutant versions of PTPRT, the most frequently mutated PTP in the superfamily. Mutations in both intracellular PTP domains (D1 and D2) were evaluated. His-tagged versions of the catalytic region of wild-type PTPRT, two D1 mutants (Q987K and N1128I), and three D2 mutants (R1212W, R1346L, and T1368M) were produced in bacteria and analyzed for phosphatase activity by using 6,8-difluoro-4-methylumbelliferyl phosphate (DiFMUP) as a substrate (Fig. 2) (12). All D1 and D2 mutants had reduced phosphatase activity compared with the wild-type protein (Fig. 2). The kinetic parameter Kcat was reduced in both D1 mutants, while Km was increased in all three D2 mutants, suggesting that mutations in the two domains may have different effects on enzymatic activity. Although it has been thought that the D2 domain is usually catalytically inactive (1), our results are consistent with recent studies that show that the D2 domain is important for phosphatase activity in some receptor phosphatases (15).

Fig. 2.

Evaluation of phosphatase activity of mutant PTPRT. (A) Saturation kinetics of wild-type and mutant PTPRT. His-tagged versions of PTPRT protein segments comprising the two catalytic domains containing wild-type (WT) and tumor-specific mutant sequences were expressed in bacteria and purified using nickelaffinity chromatography. Equal amounts of WT and mutant proteins were used to evaluate enzyme kinetics. The rate of hydrolysis of substrate (DiFMUP) is plotted against increasing substrate concentration. Data were fitted to the Michaelis-Menton equation, and the resulting kinetic parameters of WT and mutant proteins are indicated in (B).

These biochemical data on missense mutations, coupled with the large number of truncating mutations, suggested that PTPRT functions as a tumor suppressor gene. To determine whether PTPRT expression can inhibit tumor cell growth, we transfected wild-type PTPRT into HCT116 colorectal cancer cells (12). An identical expression vector containing an R632X mutant of PTPRT was used for comparison. Wild-type PTPRT potently inhibited cell growth in this assay, as seen by the substantial decrease in the number of neomycin-resistant colonies compared with the R632X mutant or with vector alone (Fig. 3, A and B). Similar results with wild-type and mutant PTPRT were also observed in DLD1 colorectal cancer cells.

Fig. 3.

PTPRT overexpression suppresses growth of human cancer cells. (A) HCT116 colorectal cancer cells were transfected with wild-type (WT) PTPRT construct, truncated R632X mutant PTPRT construct, or empty pCI-Neo vector. The photographs show colonies stained with crystal violet after 14 days of geneticin selection. (B) Number of resistant colonies (mean of two 25 cm2 flasks) for WT PTPRT, mutant PTPRT, and empty vector.

The combination of these genetic, biochemical, and cellular data suggest that PTPRT and the other identified phosphatases are likely to act as tumor suppressors. This is consistent with the function of other phosphatases implicated in tumorigenesis (7, 8, 16) and with the general role of phosphatases in inhibiting various growth-promoting signaling pathways (2). The absence of biallelic mutations in a subset of the analyzed tumors suggests that some alterations may act in a dominant-negative fashion or may affect gene dosage, mechanisms that have been previously involved in inactivation of other tumor suppressor genes (17, 18).

Little is known about the functional role of the tyrosine phosphatases discussed here. PTPN13 appears to be involved in apoptosis (19) and may be partly responsible for the antitumor effects of tamoxifen (20). Overexpression of PTPN3 inhibits growth of NIH/3T3 cells, possibly through interaction with valosin-containing protein (VCP/p97) (21). PTPN14 and PTPRF are thought to play a role in cell adhesion by regulating tyrosine phosphorylation of adherens junction proteins (22, 23). Because increased phosphorylation of adherens junctions has been shown to increase cell motility and migration (22, 24), mutational inactivation of these genes may be an important step in cancer cell invasion and metastasis. PTPRG maps to chromosome 3p14.2, a region frequently lost in lung, renal, and early-stage breast tumors, and is thought to be a target of the translocation at 3p14 in familial renal cell carcinoma (2527). However, no point mutations in PTPRG (28) or any of the other genes identified here have been previously described in any cancer. PTPRT is expressed in the developing central nervous system and in the adult cerebellum (29) and had not been thought to play a role in the growth or differentiation of other tissues. We have found that PTPRT is expressed in a variety of human tissues, including normal colon epithelium as well as cells derived from colorectal cancers (fig. S1).

Phosphatases affect signaling pathways that may be amenable to therapeutic intervention in cancer cells (2). Although reactivation of incapacitated phosphatases is likely to be pharmacologically challenging, identification of the corresponding kinases that phosphorylate substrates normally regulated by the mutant phosphatases could provide novel therapeutic targets. Like the analysis of genetic alterations in tyrosine kinases (6), the present study suggests the possibility of individualized therapy based on the mutant phosphatases present in specific tumors. Conceivably, this approach would have broad therapeutic use, because more than 50% of colorectal tumors analyzed to date have alterations in at least one member of the tyrosine phosphatome or kinome.

Supporting Online Material

Materials and Methods

Fig. S1

Tables S1 to S3


References and Notes

View Abstract

Navigate This Article