Microtubule Disruption and Tumor Suppression by Mitogen-Activated Protein Kinase Phosphatase 4

Ras activation is one of the most common alterations in human carcinogenesis (1). Ras mutation activates a signaling cascade that leads to the activation of a number of downstream effectors essential for carcinogenesis. One of these is the extracellular signal-regulated kinase (Erk), which is activated by mitogen-activated protein (MAP)/Erk kinase (MEK), a dual-specificity kinase that phosphorylates threonine and tyrosine residues in the TXY motif of Erk. The phosphorylated Erk is translocated from the cytosol to the nucleus, where it phosphorylates numerous substrates needed for cell cycle entry. Erk activation is essential for carcinogenesis (2), and constitutively activated Erk is found in a variety of human cancers (3). Constitutive activation of MEK is sufficient to transform mammalian cells (4). Furthermore, Erk inhibition by small-molecule inhibitors of MEK has been shown to result in tumor regression in vivo (5).

Normally, Erk activity is tightly regulated by MAP kinase (MAPK) phosphatases (MKP). MKPs are dual-specificity phosphatases that dephosphorylate the corresponding TXY residues phosphorylated by MEK (6). There are at least nine distinct mammalian MKPs, and these share a highly conserved COOH-terminal catalytic domain. The specificity of MKPs seems to be achieved through MAPK interaction with the less-conserved NH₂-terminal domains (7). The substrate specificity of MKPs may not be simply a biochemical issue. Other properties of MKPs, such as cell type–specific expression, cellular localization, and gene regulation, may contribute to the substrate specificity and function of MKPs (8). Many MKPs are induced by activated MAPKs, thus forming a negative feedback loop. MKP1 is the prototype of MKPs, first identified as an immediate early gene induced by mitogen and oxidative stress (9, 10). The Erk inhibition by MKP1 is sufficient to block DNA synthesis and cell cycle entry activated by oncogenic Ras (11). In addition to transcriptional induction, MKP1, MKP2, and MKP3 proteins are stabilized by Erk-mediated phosphorylation. Furthermore, the enzymatic activity of MKP1 and MKP3 is allosterically activated by Erk binding to their NH₂-terminal noncatalytic domain. MKP1 and MKP2 are localized in the nucleus, whereas MKP3 and MKP4 are mainly cytosolic, suggesting distinct roles to inactivate Erk in different cellular compartments. Multiple regulations of Erk by the different MKPs provide a feedback mechanism to fine-tune Erk activity for its biological action.

With the identification of MKPs as negative regulators of Erk, it was originally predicted that MKPs may function as tumor suppressors in carcinogenesis (11). However, analysis of MKP1 expression in human cancer suggests that MKP1 does not behave as a tumor suppressor. Overexpression of MKP1 is found in a variety of human cancers, including breast cancer (12), non–small cell lung cancer (13), and prostate cancer (14). Furthermore, down-regulation of MKP1 suppresses tumorigenicity of pancreatic cancer cells in vivo (15). Our studies of mouse epidermal carcinogenesis suggest that MKPs may have different roles at different stages of carcinogenesis. We found elevated MKP1 expression in the initiated keratinocyte (03C), followed by normalized MKP1 expression in their paired malignant cell derivative (03R; ref. 16). Cytosolic MKP3 protein, on the other hand, is overexpressed in mild dysplasia as well as in severe dysplasia/carcinoma in situ in pancreatic ducts, but MKP3 is underexpressed in poorly differentiated pancreatic carcinoma (17). The tumor suppressor activity of MKP3 has been shown by MKP3-mediated growth arrest and apoptosis in pancreatic cancer cells (17), as well as suppression of Ras-dependent tumorigenesis by a green fluorescent protein (GFP)–fused MKP3 protein (18). The identification of cytosolic MKP3 but not nuclear MKP1 as a tumor suppressor raises questions about the role of cytosolic Erk in carcinogenesis and about the effect of Erk inhibition by cytosolic MKPs.

MKP4 is another cytosolic MKP with MAPK selectivity Erk > p38 ≥ c-jun-NH₂-kinase (JNK) (19, 20). It shows closest sequence similarity to MKP3 (57% identity) and MKPx (also a cytosolic MKP, 61% identity) in the MKP family. Deletion of the MKP4 gene results in embryonic lethality due to placental insufficiency (21). In the current study, we identified MKP4 as a potential tumor suppressor in epidermal carcinogenesis. The loss of MKP4 expression is associated with malignant epidermal tumor lineages but not benign papilloma lineages from a common non-transformed keratinocyte progenitor clone. Reconstitution of MKP4 in tumor cells leads to microtubule disruption, G₂-M associated cell death, and tumor suppression in vivo. This suggests that the microtubule is the target of cytosolic MAPKs regulated by MKP4, having implications for the use of MAPK inhibitors that mimic the activities of MKP4 in human cancer therapy.

Cell culture and Affymetrix GeneChip analysis. Non-transformed epidermal cells (291) were grown in fibroblast-conditioned low calcium medium as described (22). The initiated cell lines (03C and 09C) as well as the tumorigenic cell lines (03R and 09R) were maintained in high-calcium EMEM. H1299 cells were grown in DMEM supplemented with 10% FCS. Total RNAs were isolated from each cell type performed in biological triplicate using TRIzol reagent according to the instructions of the manufacturer (Invitrogen). The Affymetrix GeneChip analysis was performed by the Gene Microarray Shared Resource at Oregon Health and Science University. The expression level of MKP4 was confirmed by reverse transcription-PCR (RT-PCR) with a set of MKP4-specific primers: ACCCACCTTCCTTCTTTACTACCC (forward) and TTCTACTCTGTTCCTGCCTTGCTC (reverse), which amplified a 220-bp cDNA fragment of mouse MKP4. A set of primers specific to hypoxanthine-guanine phosphoribosyltransferase (HPRT), CCTGCTGGATTACATTAAAGCACTG (forward) and GTCAAGGGCATATCCAACAACAAAC (reverse), was used in each RNA sample to amplify HPRT cDNA as a reference.

Antibody production and immunoblotting. The anti-MKP4 antibody was produced by immunizing rabbits with the synthetic peptide corresponding to the last 20 amino acids of murine MKP4, NH₂-Asp-Pro-Pro-Ser-Phe-Phe-Thr-Thr-Pro-Thr-Ser-Asp-Gly-Val-Phe-Glu-Leu-Asp-Pro-Thr-COOH. The specificity of the antisera was determined by immunoblotting analysis of MKP4 protein in 03R cells infected with MKP4 recombinant lentivirus. Antibodies for dual-phosphorylated Erk (T202/Y204), JNK (T183/Y185), and p38 (T180/Y182) were from Cell Signaling Technology. Antibody for MKP3 was from Santa Cruz Biotechnology.

Recombinant lentivirus and transduction. The mouse MKP4 cDNA digested with BamHI (blunted) and SmaI from MKP4 cDNA vector (ATCC no. MGC-6681) was subcloned into XhoI- (blunted) and SmaI-digested pSL35 lentiviral vector (obtained from Luigi Naldini, HSR-TIGET, Milan, Italy; ref. 23), which contains an internal ribosome entry site (IRES) sequence for coexpression of GFP. The mouse MKP3 cDNA was amplified from MKP3 cDNA vector (ATCC no. MGC-6625) and cloned into XhoI and EcoRI sites of pSL35 lentiviral vector. Recombinant lentiviruses were generated using the four-plasmid system by cotransfection of 293T cells with pSL3, which expresses the vesicular stomatitis virus G envelope protein; pSL4, which expresses the HIV-1 gag/pol genes; pSL5, which expresses the rev gene; and pSL35 containing MKP4 or MKP3. Lentivruses were harvested at 48 and 72 h after the transfection and concentrated by ultracentrifugation at 500,000 × g for 90 min. The recombinant virus titer was determined for use of minimal viral particles to achieve ≥90% infection of 03R cells.

Tetracycline-inducible MKP4 expression. Tet-responsive cell line (H1299-Tet) was generated by transfection of H1299 cell with pIRES-TetR vector constructed by cloning NheI/SmaI TetR fragment from pcDNA6/TR (Invitrogen) into pIRES2-enhanced green fluorescent protein (EGFP) vector (Clontech). The H1299-Tet clones were maintained in G418 medium and were selected by measuring luciferase activity induced by tetracycline (5 μg/mL) after transient transfection of pcDNA4/TO-Luciferase reporter. The treatment of tetracycline at 5 μg/mL had no effect on the growth of H1299-Tet clones (data not shown). Murine wild-type MKP4 cDNA was cloned into pcDNA4/TO vector (Invitrogen) and transfected into H1299-Tet cells with zeocin selection. The tetracycline-responsive clones (H1299/MKP4-Tet) were selected by Western blot analysis of MKP4 protein levels in response to tetracycline treatment.

Tubulin immunostaining. Cells were plated on collagen-coated coverslips. Before fixation, cells were permeabilized with 0.2% Triton X-100 for 10s to release free tubulin. The cells on the coverslips were fixed with 4% paraformaldehyde and stained with anti–α-tubulin (Sigma). The image was captured and processed with a Nikon confocal imaging system and a Leica imaging system.

Flow cytometry. Cells were harvested at 2, 4, and 6 days after MKP4 induction by tetracycline (1 μg/mL). The cells were fixed with 100% cold ethanol, treated with RNase, and stained with propidium iodide (50 μg/mL). The DNA content was analyzed by a Becton Dickinson FACSCalibur analyzer. All analyses were performed in duplicate.

In vivo tumorigenesis. Animal studies were performed according to institutional guidelines. Two million lentivirus-infected 03R cells were s.c. inoculated into the back of BALB/c neonates. Tumor sizes were compared between groups infected with GFP lentivirus versus MKP4 lentivirus (six mice each group). For tetracycline-inducible MKP4, 5 million H1299-Tet/MKP4 cells were s.c. injected into the flanks of 8-week-old female nude mice (Taconic). MKP4 expression was induced by administering doxycycline water (1 mg/mL) when visible tumor appeared (∼50 mm³). The tumor size was measured and calculated by the formula: 0.4 × (tumor length) × (tumor width)².

MKP4 loss is associated with malignancy. In an attempt to identify alterations that activate the Ras pathway in a non-ras mutation model of epidermal carcinogenesis (22), we profiled the gene expression of epidermal cells representing different carcinogenesis stages by means of Affymetrix GeneChip analysis (Fig. 1A). We selected candidate gene defects based on the criteria that it must be altered at the initiation stage (03C) and sustained or further altered in the malignant stage (03R). The selection was further narrowed down by comparing tissue-cultured 03R cells with 03R tumor tissue. The selection with these criteria led to the identification of MKP4, a negative regulator of MAPK. At the mRNA level, MKP4 was down-regulated in the initiated cells and was further lost in malignant tumor cells (Fig. 1B). To evaluate MKP4 expression at the protein level, we developed a MKP4 polyclonal antibody that recognizes the COOH terminus of MKP4. MKP4 loss was specific to squamous cell carcinoma (SCC)–producing (03C progressing to 03R) but not papillomagenic lineages (09C progressing to 09R; Fig. 1C). To address the broader significance of MKP4 loss as a potential initiation event and to test the hypothesis that its loss or inactivation is associated with malignancy, we examined tumors independently generated by the application of chemical or physical carcinogens in mice (Fig. 1D). MKP4 loss occurred in all (100%, n = 5) rapidly invasive SCC tested, induced by UVB light, but fewer than half (40%, n = 20) of 7,12-dimethylbenz(a)anthracene (DMBA)–12-O-tetradecanoylphorbol-13-acetate (TPA) chemically induced papillomas. The finding of MKP4 loss in a subset of the DMBA-TPA–induced tumors corresponds to the ∼50% malignant conversion rate of these papillomas in the clonal carcinogenesis model (24). MKP4 loss in SCC-producing lineages and UVB-induced SCC suggests that MKP4 loss is an initiation event that is associated with the progression of carcinogenesis to a more invasive and metastatic phenotype.

Tumor cell death by MKP4 reconstitution. We first determined whether the malignant 03R cells, as a result of their sporadic MKP4 expression loss, showed elevated Erk, JNK, and p38 activity relative to 291 cells, measured by immunoblotting with their corresponding dual phospho-specific antibodies (Fig. 2A,, first two lanes of each panel). The phosphorylated Erk, JNK, and p38 MAPKs were significantly reduced in 03R cells infected with MKP4 lentivirus (Fig. 2A , third lane relative to fourth lane of each panel), which is consistent with previous reports that MKP4 is a phosphatase with MAPK specificity for Erk > p38 ≥ JNK (19, 20).

To test whether MKP4 reconstitution could counteract tumor cell characteristics, we constructed a recombinant MKP4 lentivirus with a bicistronic cassette in which MKP4 and GFP are coexpressed through an IRES, allowing us to monitor phenotypes of infected cells. We adjusted viral titers to the minimum that achieved ≥90% infection of 03R cells. To test whether the MKP4 expressed from the lentivirus is functional, the phosphatase activity of MKP4 was evaluated by examining the phosphorylation status of endogenous MAPKs. Both MKP4 and control viruses infected more than 90% of 03R cells (comparing GFP expression throughout the cells in MKP4 (3 days) and Con; Fig. 2B). The MKP4 virus–infected cells progressively enlarged over time after the infection. The MKP4 virus–infected cells were selected again, gradually being replaced by the small subset of uninfected cells (GFP negative) as a result of cell death by MKP4 (Fig. 2B), such that by 9 days after the infection, nearly 90% of the infected cells were eliminated.

To test whether 03R tumor cell death can be induced by other MKP members of more restricted MAPK substrate specificity, we compared the activity of MKP4 with MKP3, a cytosolic MKP with MAPK selectivity predominantly to Erk. We generated a recombinant MKP3 lentivirus with MKP3 and GFP coexpression in the bicistronic cassette. MKP3 showed a similar Erk dephosphorylation, but had no effect on JNK and p38, as compared with MKP4 (Fig. 2C). Although MKP3 inactivated Erk as effectively as MKP4, MKP3 lentivirus-infected 03R cells showed no signs of enlargement or cell death, whereas nearly 90% of MKP4 lentivirus-infected 03R cells died by 9 days after the infection (Fig. 2D). Although both MKP3 and MKP4 are cytosolic MKPs with phosphatase activity to Erk, failure to induce 03R tumor cell death by MKP3 suggests that the inactivation of other MAPKs associated with MKP4 is required to cooperate with Erk inactivation to mediate the death of these highly invasive, metastatic tumor cells.

MKP4 induces G₂-M associated cell death but not G₁ growth arrest. Cellular proliferation is normally determined by extracellular signal(s)–mediated Erk activation that engages the expression of cyclin D1 and the assembly of the cyclin D–cyclin-dependent kinase 4 complex needed for cell cycle entry (25, 26), with small-molecule inhibitors of Erk such as CL-1401 blocking cell cycle progression at the G₁-S boundary (5). Therefore, we evaluated whether MKP4 reconstitution similarly blocked cell cycle progression in G₁. Because 03R cells exhibit strong homophilic cell adhesion, we expanded these studies to the human small cell lung carcinoma cell line H1299 cells, also characterized by undetectable MKP4 protein expression (data not shown), to evaluate the effect of MKP4 on cell cycle progression and tumor suppression. Induction of MKP4 expression was highly responsive to tetracycline in the H1299 stable lines (Fig. 3A). MKP4 induction resulted in dephosphorylated Erk, JNK, and p38, consistent with the phosphatase activity of MKP4. Upon MKP4 induction, H1299/MKP4-Tet cells became enlarged with decondensed chromatin, followed by cell death (Fig. 3B). In contrast to G₁ growth arrest, as expected from Erk inhibition (5), cells with MKP4 induction accumulated in the G₂-M phase of the cell cycle (Fig. 3C), corresponding to cell death (Fig. 3D). In conjunction with cell enlargement and decondensed chromatin in MKP4-expressing cells, the G₂-M associated cell death by MKP4 suggests that these cells are undergoing mitotic catastrophe, a general term applied to cell death occurring during mitosis as a result of DNA damage or defective spindle formation in tumor cells (27).

Microtubule disruption by MKP4. Because death occurred upon MKP4 induction without added DNA-damaging agents, we examined the microtubule structure in response to MKP4 induction to see if defective spindle formation is the cause of G₂-M associated cell death. Indirect immunofluorescence staining of polymerized microtubules with anti-tubulin antibody revealed microtubule disruption, as indicated by the collapsed microtubule structure in malignant 03R cells infected with MKP4 lentivirus (Fig. 4A,, top right), as compared with a typical network of polymerized mictotubules in cells infected with control lentivirus (Fig. 4A,, top left). Although the staining pattern of microtubule structure in H1299 cells was different from that in 03R tumor cells, the changes in microtubule structure was also observed in H1299 cells with MKP4 expression induced by tetracycline (Fig. 4B  bottom right compared with bottom left).

Tumor suppression by MKP4. We then explored MKP4 activity as a tumor suppressor by two approaches, one in a tumorigenicity assay and the other by the reconstitution of MKP4 by tetracycline induction of established tumors. To evaluate the effect of MKP4 on tumorigenesis in vivo, the 03R cells were infected with MKP4-containing virus and immediately inoculated s.c. to BALB/c neonates (Fig. 5A). Tumors reached an average of 72 mm³ in mice injected with 03R cells infected with GFP virus by day 22 after the inoculation compared with no tumors in six mice injected with cells infected with MKP4 virus. By 30 days, tumors in 6/6 GFP alone, control cell–injected mice averaged 319 mm³ compared with only 43 mm³ for the tumors in mice injected with the MKP4 virus. Immunostaining of the tumor sections revealed that GFP was negative in tumors from mice injected with 03R cells infected with virus coexpressing MKP4 and GFP (Fig. 5B,, right), indicating that these tumors were likely derived from the small fraction of uninfected 03R cells in the injected cell population. In contrast, tumors generated from cells infected with control lentivirus were GFP positive (Fig. 5B , left). The evidence from both the in vitro assays and in vivo tumorigenesis studies suggests that MKP4 has tumor-suppressive activities.

We next evaluated the ability of MKP4 to suppress growth in well-established tumors, by s.c. inoculating H1299/MKP4-Tet cells into the flanks of nude mice. MKP4 was induced when visible tumors appeared, by administering tetracycline to the mice in their drinking water 35 days after the inoculation. The tumors with MKP4 induction had significantly reduced tumor size compared with control tumors (Fig. 5C). Histology of tumor specimens from MKP4-induced mice showed distinct morphologic changes as compared with tumors from control mice (Fig. 5D). The tumor cells were evidently enlarged with decondensed nuclei, consistent with the morphology observed in cultured cells shown in Fig. 3B. Although the tumor growth curve suggests tumor stasis, the enlarged volume of the tumor cells after MKP4 induction suggests that MKP4 has a tumoricidal effect that contributes to tumor suppression. Taken together, the evidence from these in vitro assays and in vivo tumorigenesis studies suggests that MKP4 can act as a tumor suppressor, and that its activities could be exploited in developing novel approaches to molecular-targeted cancer therapy.

Growth arrest but not cell death by MKP4 in nontumorigenic keratinocytes. Ideally, therapies should be selective for tumor cells while leaving normal cells unaffected. To test whether normal cells are spared from the action of MKP4, we compared the effect of MKP4 on nontumorigenic keratinocyte precursor cells (291) versus their tumorigenic derivative cells (03R). 291 and 03R cells were infected with MKP4 lentivirus in which MKP4 and GFP are cotranslated from a bicistronic transcript, providing GFP as an indicator for MKP4 expression. The 291 and 03R cells showed similar infection efficiency and MKP4 expression as indicated from the level of GFP and MKP4 in both cell lines (Fig. 6A). As expected, the MKP4-positive 03R cells showed significant cell enlargement followed by cell death. At 9 d after the infection, MKP4-positive 03R cells were decreased more than 6-fold in number, and those that remained were significantly enlarged. The enlarged 03R cells gradually disappeared and were replaced with noninfected, GFP-negative cells. However, over the same time period after MKP4 expression, the nontumorigenic 291 cells did not show significant morphologic changes, and there was no significant reduction in number of MKP4-positive 291 cells (Fig. 6B and C). Thus, MKP4 expression caused cell death in the 03R malignant cells but not in phenotypically normal 291 keratinocytes. As growth arrest in normal cells can be reversible or consistent with adult tissue function, the differential effect of MKP4 expression in nontumorigenic cells versus tumorigenic cells is promising for selective cancer therapy designed to mimic the effect of MKP4.

The concept of the function of MKP as a tumor suppressor has been a matter of speculation ever since MKP1 was identified as a negative regulator of Erk (9, 28). Because MKPs are downstream in the Ras/Erk pathway, their inactivation should be essential for sustained Erk activation in carcinogenesis. However, the involvement of MKPs in carcinogenesis remains controversial. In this study, we show several features of MKP4 as a tumor suppressor. MKP4 was down-regulated in initiated keratinocytes and was lost in their malignant derivatives. Reconstitution of MKP4 expression in tumor cells triggered cellular enlargement, microtubule disruption, and G₂-M associated cell death, features of mitotic catastrophe. Furthermore, MKP4 reconstitution suppressed tumorigenicity in vivo in established human non–small cell lung carcinoma H1299 xenografts. The above lines of evidence strongly suggest that MKP4 is a tumor suppressor. Further confirmation of MKP4 tumor suppression will require prospective tumorigenesis studies using conditional MKP4 knock out mice, given that MKP4 knock-out is embryonically lethal (21).

These results are the first evidence for MKP4 as a tumor suppressor, including MKP4 loss in sporadic multistage carcinogenesis and tumor suppression in vivo, and support the original concepts of MKPs as negative regulators of oncogenic Erk activation and putative tumor suppressors. Furthermore, our studies provide evidence for a mechanism of action of tumor suppression by cytosolic MKP4 that involves the disruption of microtubules and cell death. The current data point toward MKP localization as a determinant for its contribution in carcinogenesis. Because both MKP1 and MKP2 are nuclear MKPs induced by activated Erk, they seem to be good candidates for forming negative feedback loops to inactivate Erk in the nucleus and opposing the role of nuclear Erk in activation of cell cycle entry. However, the evidence from MKP1 and MKP2 overexpression in human cancer argues against their role in tumor suppression. Although MKP3 showed robust anti-Erk activity without inducing cell death of the 03R tumor cells, the tumor suppression activity of MKP3 has been suggested from its ability to induce pancreatic carcinoma cell death (17) and fibroblast growth inhibition (18). Because MKP3 is an Erk-specific phosphatase, the different effects of MKP3 on different tumor cell types are likely determined by their dependence on Erk. Such a difference may explain why some tumor cells are resistant to Erk inhibition, whereas others are sensitive (29). Nevertheless, cytosolic MKP3 and MKP4 are the only MKPs reported to have tumor suppressor types of activities, indicating the importance of the activation of cytosolic MAPK(s) in carcinogenesis. Although activated MAPKs are generally localized in the nucleus for transcriptional regulation, tumor suppression by cytosolic MKP3 and MKP4 suggests that the inhibition of cytosolic MAPK activity is critical for tumor suppression. However, the contributions of cytosolic MAPKs in carcinogenesis have been largely unknown.

In this study, we found that the reconstitution of MKP4 leads to microtubule disruption, implicating the regulation of microtubule formation by MKP4 and its MAPK substrates. In fact, both Erk and JNK were originally identified as microtubule-associated protein kinases long before being recognized as mitogen-activated protein kinases (30, 31). Despite the fact that about half of activated Erk activity is associated with microtubules (32), the role of Erk activation in microtubule regulation is largely overlooked. Not only are microtubules structural proteins contributing to cell shape maintenance and cell polarity, but they also act as molecular motors for intracellular transport and mitosis. Microtubules undergo constant polymerization and depolymerization, termed as microtubule dynamic instability. This process is regulated through stabilizing factors such as microtubule-associated protein 2 (MAP2) and destabilizing factors such as stathmin or oncoprotein 18 (33). Microtubule poisons disrupt microtubules by perturbing either polymerization (colchicines) or depolymerization (taxol). In addition to Erk, other MAPKs, JNK, and p38 are also implicated in microtubule regulation. MAP2 is also a substrate of JNK1. Compromised microtubule integrity was observed in neuronal cells from JNK1-deficient mice (34). Stathmin, a microtubule-destabilizing factor, is phosphorylated by p38 (35). Stathmin phosphorylation inhibits its function as a destabilizing factor; thus, p38 also potentially contributes to microtubule polymerization. Microtubule disruption by MKP4, a phosphatase with specificity to Erk, JNK, and p38, suggests that the regulation of microtubule polymerization/depolymerization is a combined effect from different MAPKs.

Erk activation is generally considered to contribute to cell cycle entry by activating the expression of cyclins needed for the cell cycle, such as cyclin D1. However, cyclin D1 expression and cell cycle entry are independent of Erk activation in many malignant tumor cells (29). This is probably due to the inactivation of the Rb pathway in those tumor cells, for example, the loss of p16 in 03R cells (data not shown). Because both Rb inactivation and Ras activation are required for malignant transformation as suggested from the transformation of primary human mammary epithelial cells (36), sustained Erk activation by oncogenic Ras should contribute to activities other than the activation of cell cycle entry in tumor cells. Although the mechanisms of microtubule regulation by Erk remain largely unclear, perturbation of microtubule dynamic instability by MKP4 suggests the involvement of Erk in coordination of microtubule activity in cell cycle progression. We speculate that the regulation of Erk and other MAPK activities by MKP4 is a mechanism to coordinate DNA synthesis and microtubule formation to prevent premature cell division. MKP4 induction is particularly critical when JNK and p38 are activated under stress.

In this study, novel approaches to cancer therapy are suggested by our evidence for the mechanism of MKP4 action in tumor suppression through its ability to induce tumor cells to undergo mitotic catastrophe. The actively cycling tumor cells, being defective in the G₁-S and G₂-M DNA structure or spindle checkpoints, undergo mitotic catastrophe instead of growth arrest upon MKP4 reconstitution, providing a basis for cancer therapy strategies using combinations of MAPK inhibitors. Erk inhibition by small pharmacologic molecules has shown promising therapeutic value in animal models (5). However, Erk inhibition alone has been found to be ineffective in many tumor cells (29) and has failed in clinical trials for treatment of advanced human tumors (37). Overexpression of Erk-specific MKP3 in tumor cells results in growth arrest but not mitotic catastrophe (18) and, in this study, had no effect on 03R tumor cell growth and survival (Fig. 2D). In addition, Erk inhibition by UO126 has no effect on the growth and survival of 03R and H1299 cells (data not shown). These lines of evidence suggest that Erk inactivation is not sufficient to induce mitotic catastrophe in cancer, and that other activities associated with MKP4 are required. In addition to Erk, at least, JNK and p38 are inactivated by MKP4. Therefore, combinations of MAPK inhibitors and, possibly, specificity for MAPK isoforms warrant exploration aimed toward mimicking the activities of MKP4 as a more effective strategy in cancer therapy.

Grant support: NIH CA98893 and Oregon Health and Science University Cancer Institute CA69533.

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 G. Tim Bowden and Henry Hennings for providing mouse tumor specimens, Luigi Naldini for providing lentiviral packaging plasmids, Mandy Boyd for flow cytometry assistance, Carl Pelz and Rosalie Sears for Affymetrix GeneChip analysis, and Robin Dickinson and Stephen Keyse for providing MKP4 antibody, MKP3 cDNA, and helpful discussion. Y. Liu is the recipient of a Career Development Award from the Dermatology Foundation.