Map2k4 Functions as a Tumor Suppressor in Lung Adenocarcinoma and Inhibits Tumor Cell Invasion by Decreasing Peroxisome Proliferator-Activated Receptor γ2 Expression ^▿

Abstract

INTRODUCTION

A growing body of evidence from cellular and murine models indicates that tumor cells must acquire multiple somatic mutations in order to invade locally, disseminate hematogenously, and colonize at distant sites. Genomic sequencing studies have been performed on human tumor tissues in an effort to identify somatic mutations that might act in concert to promote tumorigenesis (11, 16, 39, 40, 44). The relevance of those mutations is, in many cases, obscure, because the mutated genes have no documented roles in cancer, are found at low prevalence in tumors, or have indeterminate effects on the functions of their gene products. Although bioinformatic tools have aided investigators in predicting whether specific mutations are “drivers” or “passengers” in tumor cells (39, 44), determining whether these mutations do, in fact, promote tumorigenesis and how they do so requires biochemical studies on the mutated gene products and biological studies on tumor cells that express them.

Elucidating the genetic mutations that drive the progression of lung cancer has been an area of particular interest, given that lung cancer is the most common cause of cancer-related death in Western countries. Mice that express Kras^G12D alleles inducibly, conditionally, or somatically develop lung adenocarcinomas with low invasive and metastatic potential (13, 17, 20, 21, 23). The subjection of Kras-mutant mice to a second oncogenic event leads to lung adenocarcinomas that arise earlier, grow faster, and metastasize widely. Proven cooperative events include inactivating mutations in tumor suppressor genes (Tsc1, Lkb1, or Pten), overexpression of Hif2α, and introduction of Tp53^R172H, a mutation that confers metastatic potential to tumors in mice and is found in patients with Li-Fraumeni syndrome and sporadic lung cancer (5, 19, 28, 33, 52). However, the incidence of metastatic disease in these double-mutant mice is, in many cases, far less than 100%, suggesting that additional genetic events are necessary for such an outcome.

Genomic studies to identify somatic mutations in the kinome of human cancer samples found that a total of 11 tumors (3% of the 356 tumors evaluated) had somatic mutations in MAP2K4 (11, 16, 40). These mutations are located primarily in the kinase domain of the MAP2K4 gene; include frameshift, nonsense, and missense mutations; and occur in colorectal cancer, non-small-cell lung cancer, melanoma, and ovarian cancer specimens. MAP2K4 encodes MKK4, a dual-specificity kinase that is activated by environmental stress, cytokines, and peptide growth factors (6). As a component of stress signaling pathways, MKK4 directly phosphorylates c-Jun N-terminal kinase (JNK) and p38 (47). MKK4 and its substrates can reportedly suppress or promote tumorigenesis and metastasis, leading investigators to conclude that MKK4 and its substrates play an important role in tumor progression and can function as either a tumor suppressor or an oncogene (18, 25, 26, 34, 45, 47). However, the spectrum of MAP2K4 somatic mutations has not been fully evaluated in MKK4 biochemical assays to delineate loss or gain of function, and downstream effectors of MKK4 that mediate cellular transformation have not been adequately explored.

Here we performed a comprehensive biochemical analysis of a panel of MAP2K4 somatic mutations reported in cases of human cancers and found that the majority represent loss-of-function mutations, and data mining of human cancer cell lines revealed that homozygous inactivating MAP2K4 mutations are accompanied by TP53 and KRAS mutations. Therefore, we posited that MKK4 functions as a tumor suppressor in lung adenocarcinomas driven by mutant Kras and Tp53 and tested this hypothesis in mice. The findings reported here support this hypothesis and revealed that MKK4 functions as a tumor suppressor partly by decreasing levels of peroxisomal proliferator-activated receptor γ (PPARγ), a nuclear receptor for fatty acids and eicosanoids that plays divergent roles in different tumor models (31, 32, 37, 41, 42).

MATERIALS AND METHODS

Plasmids and site-directed mutagenesis.

Human MKK4 cDNA (provided by Jia Le Dai, M. D. Anderson Cancer Center) was Flag-tagged at the N terminus and inserted into pLHCX vector (Clontech). MEKK1/pCMV5 (Zhimin Lu, M. D. Anderson Cancer Center), HA-ubiquitin/pcDNA3.1 (Edward Yeh, M. D. Anderson Cancer Center), and HA-JNK2 (APF)/SRα (Bing Su, Yale School of Medicine) were gifts. Murine Pparg2 cDNA (catalog no. 8895) was purchased (Addgene). Vectors expressing mouse Map2k4 (OriGene), Bgn, Vegfc, Satb1, Pparg (SA Biosciences), MAP2K4, and PPARG short hairpin RNAs (shRNAs) were purchased (Open Biosystems). To construct MAP2K4 mutants, a PCR-based site-directed mutagenesis strategy (36) was carried out using Flag-MKK4/pLHCX as a backbone. Site-specific primers are listed in Fig. 1A.

Open in a separate window

Fig. 1.

MAP2K4 mutations in human cancers and sequence alignment of MKK family members. (A) MAP2K4 mutations. Eleven MAP2K4 mutations found in human cancers analyzed herein are depicted schematically and in tabular form, denoting the cancer types associated with them (which are referenced) and the primers used to create these mutants by site-directed mutagenesis (mutation sites are underlined and in bold). The synthetic MAP2K4 dominant-active S257E; T261D (ED) mutant and dominant-negative K131R (KR) mutant used as controls in this study are indicated. *, nonsense mutation; fs, frameshift mutation. (B) Amino acid sequences in MKK family members were aligned using the ClustalW method. MAP2K4 mutation sites are labeled and boxed. Certain mutation sites (N234, S251, and P326) are well conserved across MKK family members.

RESULTS

Using computational tools, we examined the structural consequences of 11 MAP2K4 somatic mutations (point mutations and deletions) that are representative of the mutations reported thus far (11, 16, 40). The locations of mutations within the kinase domain are illustrated graphically in Fig. 1A. These mutations involve sites that are conserved among MKK family members (Fig. 1B). Mapping of the mutation sites on solved MKK4 crystal structures (35) revealed that several mutations involve residues in critical domain structures, including the ATP-binding pocket (N234I), activation loop (S251N), and αC-helix (R134Q and Q142L) (Fig. 2A). The predicted effects of these mutations on kinase function were further examined by using PolyPhen (2) and SIFT (38), which examine the functional consequence of a mutation on the basis of sequence conservation and physical properties of amino acids. Both of these tools classified several of the MAP2K4 mutations (S251N, N234I, P326L, V321M, and R154W) as functionally “damaging” and the others (R134Q, A279T, and Q142L) as “tolerated” (Fig. 2B). A third tool, CanPredict (24), which is designed to distinguish cancer-associated mutations from genetic polymorphisms and rare mutations, classified all of the “damaging” mutations and one “tolerated” mutation (Q142L) as “likely cancer-associated” and the remaining tolerated mutations (R134Q and A279T) as “likely non-cancer-associated” (Fig. 2B).

Open in a separate window

Fig. 2.

MAP2K4 mutants exhibit aberrant kinase activities. (A) MKK4 mutation sites were superimposed on an MKK4 X-ray crystal structure (Protein Data Bank; 3ALO) using PyMOL (http://www.pymol.org). Mutated amino acids and their locations are indicated with arrows and labeled in red. (B) Prediction algorithms were used to predict whether MAP2K4 missense mutations affect kinase activity (PolyPhen and SIFT) and are associated with cancer (CanPredict). (C) 293T cells were transfected with Flag-tagged MKK4 mutants and then immunoprecipitated (IP) with Flag antibody. In vitro kinase assays were performed using a kinase-dead mutant form (K55M) of JNK1 as a substrate. As controls, the synthetic MAP2K4 dominant-active S257E; T261D (ED) and dominant-negative K131R (KR) mutants were used. KA, kinase activity; WB, Western blot; total WB, WB of total lysate. (D) Quantification of kinase assays. The activity of each mutant was normalized to that of the wild type (WT), which was set at a value of 1. Graph values are presented as means ± standard deviations (SD) (n = 3), and P values were calculated with Student's t test (* or #, P < 0.01). (E) Response of mutants to stress. 293T cells were transfected with Flag-MKK4 alone and treated with UV (250 J/m² for 30 min) or with sorbitol (0.5 M for 3 h). Kinase assays performed as described above. (F) Response of mutants to upstream signal. 293T cells were cotransfected with Flag-MKK4 mutants and MEKK1 (an upstream kinase of MKK4), and then in vitro kinase assays were performed.

The same 11 MAP2K4 somatic mutants were created by performing site-directed mutagenesis using human MAP2K4 cDNA. The mutants were expressed in 293T cells, immunoprecipitated, and subjected to in vitro kinase assays using as a substrate a mutant JNK1 (JNK1-K55M) that is incapable of autophosphorylation. Relative to wild-type (WT) MKK4, eight of the mutations (V321M, P326L, R154W, S251N, N234I, W310*, R304*, and I295fs*23) were loss-of-function mutations, one (Q142L) was a gain-of-function mutation with activity similar to that of a synthetic constitutively active mutant (ED), and two (R134Q and A279T) were mutations with activities similar to that of wild-type MKK4 (Fig. 2C and D).

Given the importance of MKK4 in stress-activated pathways (47), we examined the effect of UV irradiation and sorbitol on kinase activity, focusing first on the missense mutants. The results of activation of Q142L, R134Q, and A279T were similar to those seen with the wild type, whereas R154W, P326L, S251N, and N234I had sharply attenuated responses (Fig. 2E). To determine whether the most inactive mutants (P326L, S251N, and N234I) were also refractory to an MKK4 kinase, these mutants were cotransfected with MEKK1 in 293T cells and subjected to MKK4 kinase assays. Indeed, P326L and S251N had attenuated responses to MEKK1, and N234I was not activated at all (Fig. 2F). This was not related to an inability to bind to JNK, as determined on the basis of findings from immunoprecipitation and Western blot assays (data not shown). Furthermore, when expressed in a green fluorescent protein (GFP)-tagged construct and transfected into PT67 cells, the mutants localized within the nucleus and cytoplasm in a pattern similar to that seen with wild-type MKK4 (data not shown), suggesting that their loss of function was not related to aberrant trafficking.

In 293T cells, the C-terminal truncation mutant proteins (W310*, R304*, and I295fs*23) were present at sharply reduced levels relative to those of the wild-type and missense mutant proteins (Fig. 3A). Findings from semiquantitative RT-PCR analysis performed on RNA samples from the transfectants (wild type and C-terminally truncated mutants) revealed no differences in the levels of exogenous MKK4 mRNA (Fig. 3B), raising the possibility that these mutants had reduced protein stability. Indeed, ³⁵S-methionine labeling studies revealed the C-terminally truncated mutants to have a shorter half-life than the wild-type protein (Fig. 3C and D). Furthermore, these mutants (W310*, R304*, and I295fs*23) were highly ubiquitinated (Fig. 3E), and the degradation of one of them, W310*, was attenuated by treatment with the proteasome inhibitor MG132 or clasto-lactacystin-β-lactone (Fig. 3F), suggesting that its degradation was mediated through ubiquitin-dependent pathways in a manner similar to that seen with wild-type MKK4 (3). Deletion of 200 C-terminal amino acids from wild-type MKK4 enhanced MKK4 ubiquitination (Fig. 3G), indicating that the C-terminal domains are required for protein stability.

Open in a separate window

Fig. 3.

C-terminally truncated MAP2K4 mutants are rapidly degraded. (A) 293T cells were transfected with MAP2K4 mutants and subjected to Western blotting using anti-MKK4 or antitubulin. The protein levels of truncated forms (ΔC-term) were very low compared with those of full-length forms. (B) Reverse transcriptase PCR of Flag-tagged MKK4 mRNA. There was no difference in mRNA expression levels between wild-type and C-terminally truncated MKK4. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (C) ³⁵S-Met labeling to measure protein stability, demonstrating the extinction of the truncated mutant forms by 4 h after the pulse. (D) Quantification of data in panel C by densitometric analysis. (E) In vivo ubiquitination assay. Immunoprecipitation and Western blotting of transfectants revealed that truncated forms were more highly ubiquitinated than were full-length forms. A total of 2-fold more lysates of the truncated mutants were loaded to compensate for their rapid degradation. (F) Proteasome inhibitors MG132 (50 μM) and β-lactone (clasto-lactacystin-β-lactone) (5 μM) attenuated the degradation of C-terminally truncated mutants based on ³⁵S-Met labeling analyses. (G) Deletion of 300 (300del) but not 100 (100del) or 200 (200del) C-terminal amino acids abrogated ubiquitination of MKK4. MKK4 C-terminal deletion constructs were transiently transfected into 293T cells, immunoprecipitated using anti-Flag antibodies, and subjected to Western blotting. The numbers of N-terminal amino acids remaining in the deletion constructs are indicated on right side of the panel.

We searched existing databases and found 17 (2.2% of 746 analyzed) human cancer cell lines that contain different inactivating MAP2K4 mutations (16 with homozygous deletions, C-terminal truncations, or point mutations and 1 with a heterozygous point mutation [Table 1 ]) and comprise a variety of epithelial tumor types, including lung cancer (4 cell lines). Therefore, we posited that MAP2K4 functions as a tumor suppressor in lung cancer and tested this hypothesis by inactivating Map2k4 in the bronchial epithelium of mice in the presence or absence of a potent oncogenic driver, Kras^G12D. Mice were interbred that have a Map2k4 allele with LoxP sites surrounding exon 6 within the kinase domain (MKK4^L/L) (46), a Kras allele with a G12D mutation under the control of a Lox-Stop-Lox transcriptional cassette (Kras^G12D) (20), and a Scgb1a1 allele in which the Cre recombinase has been knocked in and that is expressed specifically in Clara cells (19). Cohorts were generated that did not express Cre (WT; n = 40), expressed Kras^G12D alone (Kras^G12D; n = 34), were MKK4 deficient (MKK4^L/L; n = 38), or expressed Kras^G12D and were MKK4 deficient (Kras^G12D) (MKK4^L/L; n = 34). Mice were sacrificed either at the first sign of morbidity for Kaplan-Meier analysis or at 7 to 10 months of age for examination of tumor histologic features. Kras^G12D mice in this age range are expected to have incipient lung neoplasias of various histologic descriptions and rare lung adenocarcinomas (19).

Based on Kaplan-Meier analysis of the two cohorts, Map2k4 inactivation accelerated mortality (P = 0.0186; log rank test) (Fig. 4A). Lung tumors were present exclusively in the two cohorts that expressed Kras^G12D (examples illustrated in Fig. 4D). The conditional Kras allele was recombined in all of the tumors analyzed (Fig. 4B), and the conditional Map2k4 allele was recombined in tumors exclusively from the Kras^G12D;MKK4^L/L cohort (Fig. 4C). Although it induced no lung histologic abnormalities alone, Map2k4 inactivation increased the frequency and accelerated the growth of incipient lung neoplasias driven by mutant Kras (Fig. 4D and E). Relative to lung tumors in Kras^G12D mice, tumors in Kras^G12D; MKK4^L/L mice were more numerous (P = 0.030; two-tailed Student's t test) and involved a greater surface area per lung (P = 0.019; two-tailed Student's t test) (Fig. 4F and G). None of the tumor-bearing mice had distant metastatic disease. Microscopic analysis of lung tissues demonstrated a variety of histopathologic changes in both cohorts that were more numerous in the Kras^G12D; MKK4^L/L cohort (P = 0.048; two-tailed Fisher's exact test) (Table 2). In fact, 5 (31%) of 16 Kras^G12D mice had no lung histologic abnormalities at all, whereas all of the Kras^G12D; MKK4^L/L mice had abnormalities. We conclude that Map2k4 inactivation potentiated the development of incipient lung neoplasias initiated by oncogenic Kras, accelerated the growth of these tumors, and caused a more rapid demise of tumor-bearing mice.

Open in a separate window

Fig. 4.

MKK4 inactivation enhances lung tumorigenesis driven by mutant Kras. (A) Map2k4 inactivation shortened mouse survival times. The graph presents the results of Kaplan-Meier survival analysis of mice with the indicated genotypes. P values for comparison of the results obtained with Kras^G12D/+ and Kras^G12D/+; MKK4^L/L groups were calculated with the log rank (Mantel-Cox) test. (B) Recombination of Kras^LSL allele specifically in lung tumors. Kras^WT, wild-type Kras allele (∼622 bp); Kras^LSL, germ line conditional Kras allele (500 bp); Kras^G12D, recombined conditional Kras allele (∼650 bp). The PCR amplification strategy employed has been previously reported (20). (C) PCR confirmation of floxed allele recombination. A diagrammatic illustration of the PCR amplification strategy (top) used to detect the wild-type Map2k4 allele (MKK4⁺) and germ line allele (MKK4^L) and recombined (MKK4⁻) floxed Map2k4 allele by the use of forward (F) and reverse (R) primers is shown. The left gel demonstrates recombination of MKK4^L specifically in lung tissues of Kras^G12D/+; MKK4^L/L mice. The right gel demonstrates recombination of the MKK4^L allele specifically in lung tumors. The PCR amplification strategy used has been previously reported (46). (D) Histologic sections of lung tissues. Lung tissue sections from each cohort were stained with hematoxylin and eosin and photographed (main images). Magnifications of adenomas are shown in the insets. (E) Examples of lung tissues obtained at necropsy. Lungs and heart (H) were removed at necropsy. Photographs illustrate representative lung tissues that are wild type (WT) or have inactivated Map2k4 (MKK4^L/L) or express Kras^G12D or both. Lung tumors (arrows) are indicated. (F and G) Map2k4 inactivation promoted lung tumor growth. The numbers of lung tumors visible on the pleural surface were determined at necropsy and are expressed as the mean value per lung per cohort (F). Areas of tumor surface in lung tissue sections were measured using ImageJ (http://rsbweb.nih.gov/ij) and are expressed as the ratio of the sum of all tumor surface areas to lung surface area per cohort (G). Mean values, long horizontal lines; standard deviations, short horizontal lines.

To identify other oncogenic driver mutations in human cancer cell lines that have homozygous MAP2K4 loss, we mined the catalogue of COSMIC (Catalogue of Somatic Mutations in Cancer; http://www.sanger.ac.uk/cosmic) (Table 1). Coexisting mutations identified in the 17 cell lines with MAP2K4 mutations included TP53 (13 cell lines), CDKN2A (6 cell lines), RB1 (4 cell lines), BRAF1 (4 cell lines), SMAD4 (4 cell lines), KRAS (4 cell lines), and PTEN (1 cell line). Thus, MAP2K4 mutations coincided frequently with mutations in TP53 and its upstream activator CDKN2A. In four cell lines, KRAS mutations coincided with mutations in MAP2K4 and TP53. On the basis of these findings, we posited that MAP2K4 functions as a tumor suppressor in lung adenocarcinomas driven by mutant KRAS and TP53 and tested this hypothesis in lung adenocarcinoma cell lines derived from Kras^G12D;p53^R172HΔG mice, which develop metastatic lung adenocarcinomas owing to expression of mutant Kras and p53 (52). In syngeneic mice, these cell lines have different characteristics with respect to tumorigenicity and metastatic potential despite their uniform expression of mutant Kras and Tp53 (15), suggesting that additional genetic events are required for malignant progression.

In a panel of 13 murine lung adenocarcinoma cell lines, none was deficient in MKK4 expression (data not shown), but this cell line sample size was underpowered given that homozygous MAP2K4 deletions occur in only 3% of TP53-mutant human cancer cell lines (4). Therefore, we undertook a synthetic approach to examine the role of MKK4 in these cell lines. MKK4 (wild type, kinase dead, or constitutively active) was overexpressed in 344SQ (Fig. 5A), a highly invasive and metastatic lung adenocarcinoma cell line derived from Kras^LA1;p53^R172HΔG mice. Conversely, MKK4 was knocked down in 393P, a lung adenocarcinoma cell line derived from Kras^LA1;p53^R172HΔG mice that has low invasive activity in vitro and undetectable metastatic potential in syngeneic mice despite the presence of Kras and Tp53 mutations (15). Two 393P clones (393P-shY and 393P-shZ) were generated that were deficient with respect to MKK4 (Fig. 5D). Forced expression of wild-type or constitutively active (ED mutant) MKK4 minimally suppressed 344SQ cell proliferation (Fig. 5B) and caused a 60 to 70% suppression of cell invasion, whereas kinase-dead MKK4 (KR or N234I) did not have these effects (Fig. 5C). Conversely, MKK4-deficient 393P cells proliferated slightly faster than controls did at days 2 and 3 but not at later time points (Fig. 5E and F) and exhibited increased cell migration (3- to 4-fold) (Fig. 5G) and invasion (3- to 5-fold) (Fig. 5H) but no differences in colony formation in soft agar (data not shown). Similarly, MKK4 knockdown in a human lung cancer cell line (H2009) that has mutations in both KRAS (G12A) and TP53 (R273L) increased cell growth (Fig. 5J) and invasion (Fig. 5K).

Open in a separate window

Fig. 5.

MKK4 depletion enhances tumor cell migration and invasion. (A) Flag-tagged MKK4 (wild-type or mutant) vectors (vec) were stably transfected into 344SQ cells. (B) MKK4-transfected 344SQ cells were counted every day for 5 days under normal culture conditions. Wild-type MKK4 had no effect, but a constitutively active (ED) mutant slightly inhibited cellular growth. Results represent the mean values (±SD) obtained with quadruplicate wells. *, P < 0.05; **, P < 0.01 (compared with vector). (C) RAW 264.7- and MKK4-transfected 344SQ cells were seeded into the lower and upper chambers, respectively, of Matrigel-coated chambers for invasion assays. Invasive cells were photographed (images) and counted (bar graph). Results represent the mean values (±SD) obtained with triplicate wells. (D) 393P cells were transfected with each of four distinct MKK4 shRNAs (A, B, Y, or Z) and subjected to Western blotting, which was quantified densitometrically relative to the results obtained with a scrambled control, which was set at a value of 1. Two clones with the most prominent MKK4 depletion (shY and shZ) were selected for further experiments. (E and F) 393P-scr, shY, and shZ cells were seeded into 12-well plates, and cell numbers were counted every day for 5 days under normal culture conditions (E) or after 2 days under serum-free conditions (0%) (F), revealing a slight increase in cellular proliferation under normal conditions at days 2 and 3 but not at later time points. *, P < 0.05 compared with scr results. (G and H) MKK4 depletion enhanced 393P cell migration and invasion. RAW 264.7 and 393P cells were seeded into the lower and upper chambers, respectively, of Transwell plates for migration assays (G) and Matrigel-coated chambers for invasion assays (H) in the presence of mitomycin C (1 μg/ml). Cells were photographed (images) and counted (bar graphs). Results represent mean values (±SD) obtained with triplicate wells. (I) H2009 human lung cancer cells were stably transfected with each of five distinct human MKK4 shRNAs (sh1 to sh5). The two clones with the most prominent MKK4 depletion (sh3 and sh4) were selected for further experiments. (J) H2009 transfectants (scr, sh3, and sh4) were counted every day for 5 days under normal culture conditions, which revealed increased proliferation in clone sh3. Results represent mean values (±SD) obtained with triplicate wells. *, P < 0.05; **, P < 0.01 (compared with scr results). (K) H2009 transfectants were seeded into Matrigel-coated chambers for invasion assays. Invasive cells were photographed (images) and counted (bar graph). Results represent mean values (±SD) obtained with triplicate wells.

To examine whether tumor cell invasion is dependent on the presence of MKK4 kinase, 393P-shZ cells were stably transfected with wild-type Map2k4 or mutants with defined kinase activity (reduced, unchanged, or increased [Fig. 2D]) (Fig. 6A), and the transfectants were subjected to invasion assays. Adding back wild-type MKK4 suppressed 393P-shZ cell invasion (Fig. 6B and C), and the mutant MKK4 expression constructs reduced invasion in direct proportion to the kinase activities of the mutants (R = −0.070; P = 0.013 by one-tailed Spearman's rank correlation) (Fig. 6D), suggesting that MKK4 kinase activity was required. Similarly, forced expression of wild-type MKK4 in a human pancreatic cancer cell line with homozygous MAP2K4 loss (BxPC3) (Fig. 6E) had no detectable effect on cell proliferation in monolayer culture but suppressed invasion by 40% (Fig. 6F and G). However, in contrast to 393P cells, BxPC3 cells transfected with kinase-dead MKK4 (S251N) exhibited small but statistically significant increases in cell proliferation and invasion (Fig. 6F and G), suggesting a dominant-negative effect of kinase-dead MAP2K4 that is specific with respect to cell type.

Open in a separate window

Fig. 6.

MKK4 reintroduction abrogates invasive activity. (A) Flag-tagged MKK4 (wild-type or mutant) vectors were stably transfected into 393P-shZ cells. (B and C) 393P-shZ-MKK4 cells and RAW 264.7 cells were seeded into the upper and lower chambers, respectively, of Matrigel-coated wells, and then invasive cells were photographed (B) and counted (C). Results represent mean values (±SD) obtained with triplicate wells. *, P < 0.01 compared to empty vector. (D) Graph denoting the relationship between MKK4 kinase activity and invasion activity. According to linear regression analysis, most MKK4 mutants were within 95% confidence intervals (dotted curves; R² = 0.6737). (E) Flag-tagged MKK4 (wild-type or S251N mutant) vectors were stably transfected into BxPC3 cells. (F) Cell proliferation of MKK4-transfected BxPC3 cells was measured by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assays daily for 4 days. Wild-type MKK4 had no effect, whereas the S251N mutant increased cellular growth. Results represent mean values (±SD) obtained with quadruplicate wells. *, P < 0.01 compared with vector. (G) Wild-type MKK4 (WT) suppressed BxPC3 cell invasion in Matrigel-coated chambers, whereas the S251N mutant slightly increased invasion. Invasive cells were photographed and counted (bar graph). Results are expressed as mean values (±SD) obtained with triplicate wells. (H) Inhibitors of JNK (SP) or p38 (SB) blocked the phosphorylation of downstream targets but did not increase 393P cell invasiveactivity. 393P cells were stimulated with sorbitol (0.5 M for 3 h) to induce MAP kinase activation in the presence of SP600125 (SP; 10 μM), SB202190 (SB; 10 μM), or vehicle (dimethyl sulfoxide [DMSO]) and subjected to Western blot analysis (gels in left panel) to detect phosphorylated (p) or total proteins. 393P cells were seeded into Matrigel-coated chambers for invasion assays and treated with inhibitors or vehicle (DMSO). Invasive cells were photographed (images) and counted (bar graph). Results represent mean values (±SD) obtained with triplicate wells.

The tumorigenicity and metastatic activity of MKK4-deficient 393P cells were examined by subcutaneous injection into syngeneic, immunocompetent mice. Eight weeks following injection, necropsies were performed to determine the number of mice in each cohort bearing primary tumors and lung metastases. Although the levels of tumor incidence in mice injected subcutaneously with 393P-scr (9 of 15 mice), 393P-shY (8 of 9), and 393P-shZ (11 of 15) were similar, lung metastases in mice bearing 393P-shY tumors (5 of 8 mice) were significantly more frequent (P = 0.013; one-sided chi-square test) than they were in mice bearing 393P-scr tumors (1 of 9 mice), and a similar trend (P = 0.021) was observed in mice bearing 393P-shZ tumors (6 of 11 mice) (Table 3). 393P-shZ tumors reconstituted with wild-type MKK4 generated lung metastases in fewer mice (1 of 5) than did nonreconstituted tumors (6 of 8; P = 0.027) or kinase-dead S251N-reconstituted tumors (7 of 10, P = 0.034), and a similar trend was observed with kinase-dead N234I-reconstituted tumors (5 of 10; P = 0.13) (Table 3). Collectively, these findings suggest that MKK4 kinase activity suppressed the invasion and metastasis of Kras-Tp53-mutant lung adenocarcinoma cells.

We next sought to identify MKK4 substrates and transcriptional targets that regulate tumor cell invasion. Surprisingly, basal and sorbitol-induced phosphorylation of the two canonical substrates of MKK4 (JNK and p38) was only minimally reduced in MKK4-deficient cells (data not shown), a result that is potentially related to compensatory effects of other JNK and p38 kinases expressed in these cells (i.e., MKK7). Treatment of 393P-shZ cells with inhibitors of JNK (SP600125) or p38 (SB202190) decreased substrate phosphorylation but did not significantly change invasion (Fig. 6H). Collectively, these findings suggest that invasion was dependent on the presence of MKK4 kinase but was mediated through novel substrates. We next turned our attention to MKK4 transcriptional targets. Triplicate RNA samples from 393P-shZ and 393P-scr cells were transcriptionally profiled. Comparison of the resulting profiles demonstrated that, relative to controls, 393P-shZ cells differentially expressed 449 genes (319 up- and 130 downregulated; P < 0.01 [>1.5-fold change]) (Fig. 7 A), and 33 were confirmed by quantitative RT-PCR analysis (Fig. 7B). One of the genes with the most prominently increased (5.0-fold) expression was PPARγ, a transcription factor that functions as a nuclear receptor for fatty acids and eicosanoids and can reportedly promote or inhibit tumorigenesis (31, 32, 37, 41, 42). Pparg encodes two protein isoforms (γ1 and γ2), the mRNA levels of which increased 1.5- and 6.3-fold, respectively, following MKK4 knockdown in 393P cells (Fig. 7C), and a substantial increase in PPARγ2 protein induced by MKK4 depletion was confirmed by Western blot analysis (Fig. 7D). Actinomycin D treatment of 393P-shZ cells abrogated the increase in Pparg2 (Fig. 7E), and chromatin immunoprecipitation assays revealed increased binding of RNA polymerase II to the Pparg2 gene promoter in 393P-shZ cells (Fig. 7F), suggesting that MKK4 regulated Pparg2 expression through transcriptional mechanisms. Relative to their levels in 393P-scr cells, both spliced and nonspliced Pparg2 mRNA levels increased in 393P-shZ cells (Fig. 7G), arguing against posttranscriptional regulation of Pparg2 through mRNA splicing.

Open in a separate window

Fig. 7.

Expression profiling of 393P-shZ and 393P-scr cells. (A) Expression data matrix (“heat map”) of 574 RNA transcripts (i.e., Affymetrix probe sets, representing 449 unique genes) differentially expressed between 393P-shZ (MKK4-shRNA) and 393P-scr (control). Each row represents a gene, each column a profiled sample. The relative abundance of each gene in each group is represented using a yellow-blue color scale (blue represents low expression; yellow represents high expression). To the right of the data matrix, red, green, blue, and yellow bars denote the corresponding annotations of genes in the matrix presented using selected Gene Ontology (GO) terms. Specific genes of particular interest are indicated. (B) Quantitative RT-PCR (Q-PCR) confirmation of genes that were differentially expressed in the Affymetrix profiles. Results were normalized on the basis of ribosomal protein L32 levels. Mean values (±SD) were computed from triplicate samples and are expressed as the ratio of 393P-shZ to 393P-scr. *, P < 0.01 compared with scrambled shRNA transfectant. (C) Q-PCR analyses of PPARγ1 and PPARγ2 mRNAs in 393P-scr and 393P-shZ cells. Data represent mean values (±SD) obtained with triplicate samples. (D) Western blot of PPARγ2 in 393P-scr and 393P-shZ cells. Actin was included as a loading control. (E) Q-PCR analysis of PPARγ2 mRNA in 393P-scr and 393P-shZ cells treated with actinomycin D (Act.D; 2 μg/ml) for the indicated times. Results were normalized on the basis of ribosomal protein L32 mRNA values and are expressed as mean values (±SD) obtained with triplicate wells. (F) RNA polymerase II chromatin immunoprecipitation assays on 393P-scr and 393P-shY cells. Chromatin-RNA polymerase II complexes were immunoprecipitated, and then promoter regions of Pparg2 and Gapdh were amplified by quantitative PCR. Results were normalized on the basis of Gapdh values and are expressed as mean values (±SD) obtained with triplicate samples. (G) Q-PCR analysis of nonspliced (pre-PPARγ2) and spliced (PPARγ2) mRNA performed using PPARγ2-specific primers to amplify exon-intron and exon-exon products, respectively. Results were normalized on the basis of L32 mRNA values and are expressed as mean values (±SD) obtained with triplicate samples. (H) Q-PCR analyses of PPARγ2 mRNA in 393P-shZ cells and BxPC3 cells stably transfected with wild type (WT) or kinase-dead mutant (S251N) MKK4.

JNK phosphorylates PPARγ and inhibits PPARγ transcriptional activity (1). Therefore, we examined whether the increase in Pparg2 expression was dependent on the presence of MKK4 kinase and required JNK. Following reexpression of (wild-type or S251N) MKK4 in 393P-shZ cells, Pparg2 expression decreased in wild-type but not S251N transfectants (Fig. 7H). Similarly, Pparg2 levels decreased following reintroduction of wild-type but not kinase-dead MKK4 into human BxPC3 pancreatic cancer cells that had a homozygous MAP2K4 deletion (Fig. 7H). However, treatment of 393P cells with SP600125 or SB202190 did not increase the levels of expression of Pparg2 or Pgc-1α, a PPARγ target gene (19) (data not shown). Thus, MKK4 suppressed PPARγ expression through a kinase-dependent mechanism involving novel MKK4 substrates.

PPARγ has numerous transcriptional targets that mediate diverse biological functions. To examine the extent to which MKK4 deficiency activated a PPARγ-dependent transcriptional program, we mined a database that lists genes regulated by treatment with a PPARγ agonist in human lung cancer cells (Gene Expression Omnibus {"type":"entrez-geo","attrs":{"text":"GSE7035","term_id":"7035"}}GSE7035) and found that, of the 1,343 genes that exhibited increased expression, 35 overlapped with genes with increased expression in 393P-shZ cells (Table 4), which exceeded the expected overlap due to chance (23 genes; P < 0.01 [Fisher's exact test]). Furthermore, MKK4 knockdown in 393P cells increased the levels of Pgc-1α (Fig. 8 A). Thus, MKK4 depletion activated a PPARγ-dependent transcriptional program.

Open in a separate window

Fig. 8.

MKK4 suppresses tumor cell invasion through PPARγ. (A) Quantitative RT-PCR (Q-PCR) analysis of PGC-1α mRNA in 393P-scr and 393P-shZ cells. Data represent mean values (±SD) obtained with triplicate samples. (B) Q-PCR analyses of PPARγ2 (left) and PGC-1α (middle) and reporter assays (right) in cells transiently transfected with a PPAR-responsive element (PPRE) luciferase reporter plasmid. Results were normalized on the basis of renilla luciferase activity values and are expressed as mean values (±SD) obtained with triplicate samples. (C and D) Q-PCR analyses of PPARγ2 mRNA in 344SQ cells (C) or 393P-shZ cells (D) stably transfected with shRNAs against PPARγ (shP-A or shP-B) or scrambled control (scr). Results were normalized on the basis of L32 mRNA values and are expressed as mean values (±SD) obtained with triplicate samples. (E and F) Invasion assays on 344SQ cells (E) or 393P-shZ cells (F) stably transfected with PPARγ or scrambled shRNA. Invasive cells were photographed (images) and quantified (bar graphs). Results are expressed as mean values (±SD) obtained with triplicate wells. (G) Invasion assays on 393P-shZ cells treated with a PPARγ-selective antagonist (T0070907) or vehicle (DMSO). Invaded cells were photographed (images) and quantified (bar graph). Vehicle-treated cells are indicated as a “0” dose. Cellular toxicity after 24 h of treatment of 393P-shZ cells with T0070907 was examined by MTT assay (line graph). Vehicle-treated cells are indicated as a “0” dose. Results are expressed as mean values (±SD) obtained with triplicate wells. (H) Invasion assays on 393P cells stably transfected with PPARγ cDNA or empty (vec) expression vectors. Invasive cells were photographed (images) and quantified (left bar graph). Results are expressed as mean values (±SD) obtained with triplicate wells. Q-PCR analysis confirmed exogenous PPARγ mRNA expression (right bar graph). Results were normalized on the basis of L32 mRNA values and are expressed as mean values (±SD) obtained with triplicate samples. (I) Q-PCR analysis of human pancreatic cancer cells (BxPC3) stably transfected with one of four distinct shRNAs against PPARγ (shP1 to shP4) or scrambled control (scr). Results were normalized on the basis of L32 mRNA values and are expressed as mean values (±SD) obtained with triplicate samples. (J) Invasion assays on BxPC3 cells stably transfected with PPARγ shRNAs (shP3 and shP4) or scrambled (scr) shRNA. Invasive cells were photographed (images) and quantified (bar graphs). Results are expressed as mean values (±SD) obtained with triplicate wells.

Relative to 393P cells, 344SQ cells are more invasive (15) and exhibit higher Pparg2 and Pgc-1α expression and higher PPAR transcriptional activity in reporter assays (Fig. 8B). We posited that PPARγ promotes invasion and tested this hypothesis by introducing one of two distinct murine PPARγ shRNA constructs (shP-A or shP-B) into 393P-shZ cells and 344SQ cells, which resulted in basal PPARγ levels that were decreased by 48% in 344SQ cells (Fig. 8C) and abrogation of the MKK4 shRNA-induced increase in Pparg2 in 393P-shZ cells (Fig. 8D). As a comparison to the effects of PPARγ knockdown, shRNAs against other genes that exhibited increased expression with MKK4 depletion (biglycan, vascular endothelial growth factor c, and special AT-rich sequence-binding protein 1) were introduced into 393P-shZ cells. PPARγ knockdown decreased the invasion of 344SQ cells (Fig. 8E) and 393P-shZ cells (Fig. 8F), whereas the other shRNAs did not abrogate 393P-shZ cell invasion (data not shown). Treatment with T0070907, a PPARγ antagonist (29), blocked 393P-shZ cell invasion without affecting cell viability (Fig. 8G). Conversely, forced expression of PPARγ enhanced 393P cell invasion (Fig. 8H). To determine whether PPARγ promotes invasion in human cancer cells that have homozygous MAP2K4 loss, BxPC3 cells were stably transfected with one of four distinct human PPARγ shRNAs (shP1 to shP4) (Fig. 8I). Relative to scrambled shRNA controls, PPARγ-depleted BxPC3 cells (shP3 and shP4) exhibited reduced invasion (Fig. 8J).

DISCUSSION

Numerous studies have demonstrated MAP2K4 somatic mutations in cancer cells (18, 26, 34, 45, 47, 50), but the spectrum of MAP2K4 somatic mutations has not been fully evaluated in MKK4 biochemical assays to delineate loss or gain of function, and downstream effectors of MKK4 that mediate cellular transformation have not been adequately explored. Here we systematically characterized the biochemical properties of a large panel of cancer-associated MAP2K4 mutations, examined the consequences of Map2k4 inactivation in genetically engineered mouse models of human lung cancer driven by mutant Kras alone or in combination with mutant Tp53, and profiled transcriptional changes induced by MKK4 depletion in Kras/Tp53-mutant lung adenocarcinoma cells. Findings from these studies revealed that Map2k4 functions as a tumor suppressor in lung adenocarcinoma and inhibits tumor cell invasion by decreasing PPARγ2 expression.

Map2k4 inactivation increased the multiplicity and accelerated the growth of incipient lung neoplasias and promoted the invasion and metastasis of Kras/Tp53-mutant lung adenocarcinoma cells, leading us to conclude that MKK4 exerts tumor suppressor activity at both early and late stages of lung tumorigenesis. These findings add to a growing list of somatic mutations that cooperate with mutant Kras to promote lung adenocarcinoma development in mice (19, 22, 28, 33, 52). However, Map2k4 inactivation mitigates carcinogen-induced skin tumors in the same mouse strain (MKK4^L/L) used in this study (12), which may reflect tissue-specific roles of MKK4, differential signaling by mutant Ras family members (Kras and Hras in lung and skin carcinomas, respectively), or other factors.

Biochemical analysis of mutant MKK4 revealed that the MAP2K4 somatic mutations inactivated MKK4 through diverse mechanisms. The missense mutations targeted motifs required for kinase activation, including the ATP-binding pocket (N234I) and activation loop (S251N), and nonsense and frameshift mutations resulted in rapid degradation of MKK4 through the ubiquitin-proteasome complex, which was presumably engaged due to misfolding of the C-terminally truncated proteins, as has been reported for other mutant tumor suppressors (14, 43, 48). Although the biological relevance of degrading an inactive kinase is not immediately obvious, one potential consequence is the removal of proteins that would otherwise have a dominant-negative effect. The C-terminally truncated mutants lack a domain required for interactions with an upstream kinase, MEKK1, creating kinase-dead proteins that can compete with wild-type MKK4 for binding to MKK4 substrates and other proteins within the larger mitogen-activated protein (MAP) kinase complex. Given that these mutants are unstable, such a dominant-negative effect is unlikely.

Several findings reported here suggest that the tumor suppressor activity of MKK4 is kinase dependent. MAP2K4 somatic mutations cluster within the kinase domain, and most of the mutants had reduced kinase activity. Furthermore, the ability of MAP2K4 mutants to attenuate invasion correlated positively with their residual kinase activities. On the basis of these findings, we expected that canonical MKK4 substrates would function as key mediators of tumor cell invasion and metastasis. However, basal and sorbitol-induced JNK and p38 phosphorylation did not decrease in MKK4-depleted cells, and chemical antagonists of JNK (SP600125) and p38 (SB202190) did not enhance 393P cell invasion. These findings were surprising, given that MKKs are viewed as kinases dedicated primarily to the regulation of MAP kinases (9). The only noncanonical MKK4 substrate reported thus far is RXR, which MKK4 phosphorylates on Tyr249 in the ligand-binding domain, thereby inhibiting RXR transactivation (30). However, RXR knockdown did not decrease PPARγ expression levels in 393P-shZ cells (data not shown), arguing against RXR as the requisite MKK4 substrate. These findings warrant efforts to discover additional substrates through which MKK4 mediates tumor suppression.

To gain insight into candidate mediators of MKK4, we performed gene expression profiling studies and identified 449 genes from a broad range of Gene Ontology categories that increased or decreased in expression in MKK4-depleted tumor cells. Some of these genes play critical roles in the extracellular matrix (Bgn and Vcam1), lymphangiogenesis (Vegfc), inflammation and response to inflammatory cytokines (Il13ra1, Csf1, Ccr9, Cxcl11, and Cxcl7), transcriptional regulation (Satb1, Smarca2, and Id2), cell polarity (Inadl, Crb3, and Pard6b), and epithelium-to-mesenchyme transition (Ncam1, Vim, and Snail1). Although many of these genes could have contributed to the enhanced invasive and metastatic activity, we chose to investigate the role of PPARγ, owing to the degree to which its expression increased in MKK4-deficient cells and its reported role in other tumor models (7, 27, 37, 41). We found that PPARγ promoted tumor cell invasion and increased in abundance through an MKK4 kinase-dependent pathway that did not involve canonical MKK4 substrates. Collectively, these findings reinforce a growing body of evidence indicating that MAP2K4 somatic mutations contribute to tumor progression in a variety of epithelial tumor types (18, 49–51) and suggest that MKK4 exerts its tumor suppressor properties through noncanonical substrates that have diverse transcriptional targets.

ACKNOWLEDGMENTS

Footnotes

REFERENCES