Biochemical Journal

Research article

Expression and functional validation of new p38α transcriptional targets in tumorigenesis

Aneta Swat , Ignacio Dolado , Ana Igea , Gonzalo Gomez-Lopez , David G. Pisano , Ana Cuadrado , Angel R. Nebreda

Abstract

p38α MAPK (mitogen-activated protein kinase) plays an important tumour suppressor role, which is mediated by both its negative effect on cell proliferation and its pro-apoptotic activity. Surprisingly, most tumour suppressor mechanisms co-ordinated by p38α have been reported to occur at the post-translational level. This contrasts with the important role of p38α in the regulation of transcription and the profound changes in gene expression that normally occur during tumorigenesis. We have analysed whole-genome expression profiles of Ras-transformed wild-type and p38α-deficient cells and have identified 202 genes that are potentially regulated by p38α in transformed cells. Expression analysis has confirmed the regulation of these genes by p38α in tumours, and functional validation has identified several of them as probable mediators of the tumour suppressor effect of p38α on Ras-induced transformation. Interestingly, approx. 10% of the genes that are negatively regulated by p38α in transformed cells contribute to EGF (epidermal growth factor) receptor signalling. Our results suggest that inhibition of EGF receptor signalling by transcriptional targets of p38α is an important function of this signalling pathway in the context of tumour suppression.

  • epidermal growth factor receptor (EGFR)
  • gene expression
  • p38 mitogen-activated protein kinase (p38 MAPK)
  • Ras oncogene
  • tumorigenesis

INTRODUCTION

p38α MAPK (mitogen-activated protein kinase) is an ubiquitously expressed serine/threonine protein kinase, which is activated by many extracellular stimuli and co-ordinates numerous cellular processes [1,2]. p38α can also negatively regulate malignant cell transformation, which seems to be accounted for by its inhibitory role on cell-cycle progression, and its stimulatory effect on apoptosis and differentiation [3,4].

Intriguingly, despite the recognized importance of p38α in the regulation of gene expression [59], little is known about how this p38α function impinges on tumorigenesis. In fact, most of the known tumour suppressor mechanisms co-ordinated by p38α occur at the post-translational level. These include the post-translational regulation of ERK1/2 (extracellular-signal-regulated kinase 1/2) pathway activity, protein stability of cyclin D1, subcellular localization/stability of Cdc25 (cell division cycle 25) phosphatases, EGFR [EGF (epidermal growth factor) receptor] intracellular fate, and up-regulation of cell-cycle inhibitors such as p21Cip1 or p27Kip1 [3,4]. There are a few examples of the regulation of gene expression by p38α with potential relevance for malignant transformation, most notably the transcriptional control of cyclin D1 [10] or the cell-cycle inhibitor p16INK4a [11], and the up-regulation of Egfr mRNA in lung tumours of p38α-deficient mice [12].

We have investigated how transcriptional regulation contributes to the tumour suppressor effect of p38α by comparing whole-genome expression profiles of WT (wild-type) and p38α−/− (p38α-deficient) MEFs (mouse embryonic fibroblasts) expressing oncogenic H-RasG12V. Our previous work showed that p38α deficiency facilitates H-RasG12V-induced transformation in these cells [1315]. We have now identified a group of genes that are regulated by p38α in the context of malignant cell transformation, and have also functionally validated nine of these genes in anchorage-independent cell-growth assays. Interestingly, we have found that approx. 10% of the genes whose expression is negatively regulated by p38α can potentially enhance EGFR signalling. These results provide further evidence for a key role of EGFR pathway down-regulation in tumour suppression by p38α.

MATERIALS AND METHODS

Microarray procedures and data analysis

mRNA was extracted from three biological replicates of exponentially proliferating WT and p38α−/− MEFs expressing H-RasG12V using the Qiagen RNeasy kit according to the manufacturer's instructions. RNA quality was checked spectrophotometrically by its 260/280 nm absorbance ratio as well as by RT (reverse transcription)–PCR. RNAs were labelled with Cy5 (indodicarbocyanine) and Cy3 (indocarbocyanine), pooled, hybridized and processed with the Amersham Codelink 20K mouse platform (GeneCore Laboratory, EMBL, Heidelberg).

Microarray dataset normalization was performed via cyclic loess normalization. Differentially expressed genes were obtained by applying linear models with R limma package [16] (Bioconductor project, http://www.bioconductor.org) and multiple hypotheses testing was accounted for by adjusting the estimated significance level (P value) with the Benjamini and Hochberg FDR (false discovery rate) correction. Those genes with an FDR<0.05 and a fold change >1.5 (accounting for a total of 2100 genes) were selected as differentially expressed between H-RasG12V-expressing p38α−/− and WT MEFs and subsequently discriminated with a parametric one-way ANOVA test (P<0.01) with the GeneSpring 6.2 software (Silicon Genetics). This latter analysis allowed us to cut down the number of differentially expressed genes to a more workable set of 202 genes, which corresponded to those genes deregulated ±1.5-fold between H-RasG12V-expressing p38α−/− and WT MEFs within the 99% confidence level (P<0.01) and an FDR<0.05.

Gene functions were manually annotated on the basis of exhaustive searches in PubMed and Gene Ontology-based applications such as the FatiGO and Source Batch databases. The microarray data have been deposited in the GEO (Gene Expression Omnibus) database under accession number GSE26762.

GSEA (gene set enrichment analysis)

GSEA was performed with the EGFR-associated set of 19 genes identified in the present study compared with microarray data from [17] as a reference. Affymetrix arrays (GEO ID GSE9599) were normalized using RMA (Robust Multi-Array Average). Whole microarray genes were ranked based on limma moderated t statistic. After Kolmogorov–Smirnoff testing, the gene set was considered statistically significant if the FDR<0.25, a widely accepted cut-off for the identification of biologically relevant gene sets [18].

mRNA isolation and RT–PCR analysis

Total RNA was extracted from MEFs or mouse tumours [12,14] using the RNeasy Kit (Qiagen) according to the manufacturer's instructions. The RNA concentration was determined by measuring the absorbance at 260 nm and the quality was monitored by the ratio of absorbance at 260/280 nm and by 1% agarose gel electrophoresis. cDNA was prepared using MMLV (Moloney murine leukaemia virus) reverse transcriptase (Invitrogen) and used as the template for PCR amplification. PCR primers were designed with the Vector NTI 8 software (InforMax) and used at a final concentration of 160 nM (Supplementary Table S1 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). RT–PCR conditions were as follows: RT at 37 °C for 90 min, denaturation at 95 °C for 2 min, followed by 30–35 cycles of denaturation at 95 °C for 45 s, annealing at 55 °C for 45 s and elongation at 72 °C for 60 s. A final step of elongation at 72 °C for 10 min was performed. The products of the RT–PCR were analysed by 1% agarose gel electrophoresis, stained with ethidium bromide and imaged using the Gel Doc 2000 system (Bio-Rad).

qRT-PCR (quantitative real-time PCR) analysis was performed as described previously [15]. Primer sequences are listed in Supplementary Table S2 at http://www.BiochemJ.org/bj/434/bj4340549add.htm.

Cell culture and reagents

Primary WT and p38α−/− MEFs were immortalized using the 3T3 protocol, maintained as described previously [14] and used as the starting biological material in the experiments described in the present study. The term MEFs is used throughout the present paper to refer to spontaneously immortalized cells or derivatives of them (i.e. H-RasG12V transformed). MEFs expressing OHT (4-hydroxytamoxifen)-inducible H-RasG12V have been described previously [14]. HEK (human embryonic kidney)-293 human epithelial cells were maintained in the same medium as MEFs and used for transient transfection. HEK-293T [HEK-293 cells expressing the large T-antigen of SV40 (simian virus 40)] cells were used to generate retroviruses to stably infect MEFs and were provided by Dr M. Serrano (CNIO, Madrid, Spain).

Expression constructs and retroviral infections

pBabe-puro-H-RasG12V and pWZL-hygro-based constructs containing the Gstm2 (please note that information regarding the proteins encoded by all genes can be found in Tables 1 and 2), Cd9, Mmp3, Egln3, Pmp22, Aig1, Ctsh, Nck2 and Rtn4 mouse cDNAs were used for the generation of retroviruses and subsequent MEF transduction as described previously [14]. Cloning details, cDNA origin, amplification primers and restriction sites are provided in Supplementary Table S3 (at http://www.BiochemJ.org/bj/434/bj4340549add.htm). Myc-tagged versions of pWZL-hygro-Gstm2 and pWZL-hygro-Cd9 were prepared by site-directed mutagenesis on the non-tagged constructs (QuikChange®, Stratagene) and verified by sequencing. The retroviral construct MSCV-hygro-p38α [19] was used to re-express p38α in p38α−/− MEFs for the rescue experiments.

Immunoprecipitation

Control and H-RasG12V-expressing WT MEFs were transduced with Cd9 or an empty vector, and then EGFR was immunoprecipitated. Subconfluent 10-cm cell-culture plates were washed twice with ice-cold PBS, scraped on ice and harvested by centrifugation at 400 g and 4 °C for 10 min. The cell pellets were lysed in 100–600 μl of modified RIPA lysis buffer [50 mM Hepes (pH 7.4), 150 mM NaCl, 10% glycerol, 1.5 mM MgCl2, 1 mM EGTA, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 20 mM NaF, 0.1 mM sodium orthovanadate, 1 mM PMSF, 2.5 mM benzamidine, 2 μM microcystin and 10 μg/ml leupeptin and aprotinin]. Whole-cell extracts (500 μg) were immunoprecipitated with an anti-EGFR antibody (Santa Cruz Biotechnology) overnight at 4 °C. Immunoprecipitates were incubated with Protein G–Sepharose (Amersham) for 1 h at 4 °C, washed three times, analysed by SDS/PAGE and visualized with an anti-EGFR antibody (Santa Cruz Biotechnology).

Immunoblotting, transformation and tumorigenicity assays

Immunoblot analysis was performed as described previously [14] using the following primary antibodies: anti-c-Myc (9E10, Santa Cruz Biotechnology), anti-p38 MAPK (C20-G, Santa Cruz Biotechnology), anti-tubulin (clone DM1A, Sigma), anti-Ras (clone 18, Sigma), anti-phospho-ERK (9101, Cell Signaling Technology), anti-phospho-Akt (9271, Cell Signaling Technology), anti-EGFR (1005, sc-03, Santa Cruz Biotechnology), anti-phospho-EGFR (Tyr1068, Cell Signaling Technology), anti-Egln3 (Novus Biologicals) and anti-Nck2 (Upstate Biotechnology).

Cell proliferation {MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide]} and anchorage-independent growth (soft agar) assays were performed as described previously [14]. Briefly, soft agar assays were performed in duplicate with a top layer of 0.35% agar and a bottom layer of 0.5% agar. Cell colonies were stained with 0.1% Crystal Violet (Sigma) at the end of the experiment and counted if bigger than 0.2 mm. Colony numbers normally displayed an S.D. of ±500 colonies. Pictures of representative fields were taken at 40× magnification with a Leica MZ6 microscope coupled to an RS Photometrics camera.

Statistical analysis

Unless otherwise indicated, all results are expressed as means ± S.D. for at least three independent experiments. Statistical analysis was performed using a Student's t test with a statistically significant P value (P<0.01).

RESULTS

Gene expression profiling of H-RasG12V-transformed WT and p38α−/− MEFs

In order to identify new targets of p38α in the regulation of oncogene-induced transformation, RNA was extracted from biological triplicates of exponentially proliferating H-RasG12V-transformed WT and p38α−/− MEFs, labelled and analysed with the Amersham Codelink® mouse 20K platform. Statistical filtering, retaining genes with an increased or decreased expression greater than ±1.5-fold, an FDR<0.05 and an adjusted P value<0.01, revealed 202 genes that were significantly deregulated between WT and p38α−/−MEFs (Supplementary Table S4 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). Among them, 139 genes (69%) were up-regulated and 63 genes (31%) were down-regulated in H-RasG12V-p38α−/− MEFs (Figure 1). Genes were subsequently clustered according to their biological function using GO-based approaches and subcategorized (Figure 1 and Supplementary Table S5 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). A flowchart showing all of the individual steps undertaken is shown in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/434/bj4340549add.htm).

Figure 1 Functional classification of genes differentially expressed between H-RasG12V- p38α−/− and H-RasG12V-WT MEFs

Genetic deletion of p38α significantly affected the expression of 202 genes. The number of genes within each functional category, either up-regulated (black) or down-regulated (white) in H-RasG12V- p38α−/− MEFs is indicated as a percentage of the total number of genes.

Gene expression validation by semi-quantitative RT–PCR in cultured MEFs

To validate microarray gene expression differences, 31 out of the 202 genes potentially regulated by p38α were selected for semi-quantitative RT–PCR analysis, which was performed on new mRNA preparations extracted from the two transformed MEF lines growing exponentially. Gene selection was based on the different mRNA expression levels in H-RasG12V-p38α−/− compared with H-RasG12V-WT MEFs, as well as on the putative involvement of each gene in Ras-induced transformation or p38α signalling. Out of the 31 genes selected, 19 were up-regulated and the other 12 were down-regulated in H-RasG12V-p38α−/− cells (Supplementary Table S6 at http://www.BiochemJ.org/bj/434/bj4340549add.htm).

Two independent cellular systems were used to dissect the relative contribution of p38α and H-RasG12V to the expression of these genes. The first involved established WT and p38α−/− MEFs constitutively expressing H-RasG12V for at least 2 weeks [13,14], which allowed us to evaluate the steady-state gene expression levels in each cell line. The second consisted of WT and p38α−/− MEFs expressing an OHT-inducible form of H-RasG12V (ER-HRasG12V) [14]. This system allowed us to monitor gene expression in both cell lines in the absence of H-RasG12V signalling as well as early (24 h) after its activation with OHT (Figure 2A).

Figure 2 Semi-quantitative RT–PCR validation of genes differentially regulated by p38α and/or H-RasG12V

(A) Immunoblot analysis of cell extracts from ER-HRasG12V-expressing WT and p38α−/− MEFs stimulated with 1 μM OHT for 24 h. (B) WT and p38α−/− MEFs expressing ER-HRasG12V were treated with OHT, as indicated. Established H-RasG12V-expressing p38α−/− and WT MEFs were analysed in parallel. Asterisks indicate changes related to p38α deficiency, but independently of H-RasG12Vexpression. Up and Down indicate mRNAs up-regulated or down-regulated respectively in established H-RasG12V-p38α−/− MEFs according to the microarray analysis; FC, fold change in the microarray analysis. (C) Genes differentially expressed in MEFs by the combined effect of both p38α and H-RasG12V. Samples are as described in (B). Asterisks indicate changes caused mainly by H-RasG12V expression. (D) Changes in gene expression were confirmed by analysing WT, p38α−/− and p38α add-back p38α−/− MEFs. Immunoblot analysis is shown at the bottom. The experiments were performed at least three times with similar results.

We found that the majority of the genes analysed by semi-quantitative RT–PCR were deregulated between stable H-RasG12V-expressing WT and p38α−/− MEFs (Figures 2B and 2C, two right-hand lanes) in a way that mirrored the microarray results in more than 90% of the cases (note the microarray fold-change column to the right-hand side). However, the differential expression levels of Ralgds, Rtn4 and Ctsl in H-RasG12V-WT compared with H-RasG12V-p38α−/− MEFs were less clear than expected from the microarray results. This might be related to sensitivity limitations of the semi-quantitative RT–PCR technique, at least for Rtn4 (+2.2) and Ctsl (+2.4), since their moderate up-regulation was detectable by qRT-PCR (Figures 3 and 4A). Of note, we could not detect expression of Stac2, Rom1, Dpep1, Traf1 and Lzts2.

Figure 3 qRT-PCR analysis of gene expression in various tumour models

qRT-PCR was performed on RNA extracted from exponentially proliferating H-RasG12V-expressing WT and p38α−/− MEFs (light grey bars), subcutaneous xenografts derived from the same cells grown in nude mice for 2 weeks (white bars) or K-RasG12V-induced lung tumours in WT and p38α−/− mice (dark grey bars). qRT-PCR results were compared with the respective microarray fold-change value (black bars). Asterisks indicate fold changes greater than 28. All values, together with their corresponding S.D. values, are listed in Supplementary Table S7 at http://www.BiochemJ.org/bj/434/bj4340549add.htm. Loxl1, lysyl oxidase-like 1.

Figure 4 Transcriptional regulation of EGFR signalling by p38α in oncogenesis

(A) EGFR signalling-related genes were validated by qRT-PCR in exponentially proliferating H-RasG12V-p38α−/− compared with H-RasG12V-WT MEFs and fold changes (white bars) were plotted along with the microarray fold changes for each gene (black bars). (B) WT MEFs were transduced with H-RasG12V, Cd9 or empty vectors (−), as indicated, and whole-cell lysates were analysed by immunoblotting with the indicated antibodies either directly (right-hand panel) or after immunoprecipitation with anti-EGFR antibodies (left-hand panel). (C) qRT-PCR analysis of Egfr mRNA in exponentially proliferating H-RasG12V-expressing p38α−/− compared with WT MEFs (black bars), in tumour xenografts derived from the same MEFs and grown for 2 weeks in nude mice (grey bars) and in lung tumours induced by K-RasG12V expression in p38α−/− mice compared with WT mice (white bars).

Concerning the remaining 26 genes, analysis of ER-HRasG12V-expressing WT and p38α−/− MEFs revealed that the expression of half of them was dependent on p38α (Figure 2B, first and third lanes marked with asterisks), whereas the other half were additionally influenced by H-RasG12V signalling (Figure 2C, second and fourth lanes marked with asterisks). Interestingly, the effect of H-RasG12V on gene expression in the latter case was already clear 24 h after the onset of H-RasG12V signalling and was usually maintained during the transformation process in MEFs stably expressing H-RasG12V. However, the relative mRNA abundance of some genes in H-RasG12V-expressing WT compared with p38α−/− MEFs was attenuated (Ralgds, Ctsl) or enhanced (Cd9, Ctsh) with time, arguing for a kinetic component in the regulation of gene expression by p38α and H-RasG12V. This could in principle be due either to specific changes in p38α or H-RasG12V signalling or to non-specific alterations or selective pressure related to the transformed phenotype. For example Cxcl1, Nck2 and Tnfrsf23 showed opposite expression patterns in early transformed compared with established cell lines, whereas Rtn4 and Snfl1k showed opposite regulation in non-transformed compared with H-RasG12V-expressing cells. Interestingly, p38α reconstitution into p38α−/− MEFs rescued the expression levels of putative p38α-regulated genes to approximately WT levels in both non-transformed (Figure 2D) and transformed cells (Supplementary Figure S2 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). Taken together, the results suggest that at least the selected genes that were analysed are indeed regulated by p38α, although some of them can be additionally modulated by the cellular context (e.g. non-transformed compared with transformed cells, and early compared with late transformation stages).

Gene expression validation by qRT-PCR in cultured cells and mouse tumours

In order to evaluate the relevance of the observed gene expression differences between WT and p38α−/− MEFs for in vivo tumorigenesis, selected genes were further analysed by qRT-PCR in subcutaneous tumours obtained from the injection of H-RasG12V-expressing WT and p38α−/− MEFs in nude mice [14], as well as in lung tumours induced by K-RasG12V expression in WT and p38α−/− mice [12]. We found that all of the genes analysed were similarly deregulated in a p38α-dependent manner in transformed MEFs, MEF-derived subcutaneous tumours and K-RasG12V-induced lung tumours, in accordance with the original results obtained from the microarray analysis (Figure 3). Although the microarray-based profiling and qRT-PCR analysis did not result in identical quantitative changes, gene expression fold changes were qualitatively conserved in the different tumour models tested (Figure 3 and Supplementary Table S7 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). These results link the p38α-mediated modulation of particular mRNAs with the ability of p38α to regulate tumorigenesis.

Functional validation of genes deregulated between H-RasG12V-transformed WT and p38α−/− MEFs

Next, we selected nine out of the 26 genes mentioned above, according to their level of deregulation in H-RasG12V-expressing p38α−/− MEFs and their putative connection with the process of cell transformation, to perform functional studies (Table 1). Three genes encoded membrane (Aig1) or extracellular matrix (Mmp3, Ctsh) proteins and were selected due to their acute deregulation in H-RasG12V-expressing p38α−/− compared with WT MEFs. Little is known about Aig1 in the context of malignant transformation. In contrast, Mmp3 and Ctsh are known to be involved in tumour progression (Table 1). Four other genes encoded the tyrosine-kinase-signal-modulating proteins CD9 and Nck2, the proliferation-regulating protein Pmp22 and the pleiotropic protein Rtn4, and were all up-regulated in the more transformed H-RasG12V-p38α−/− MEFs, suggesting that enhanced expression of these genes might contribute to transformation. Accordingly, some of them have been described to be up-regulated in tumours and potentially associated with tumour progression (Table 1). Finally, two genes that were down-regulated in H-RasG12V-p38α−/− MEFs, namely Gstm2 and Egln3, were also considered good candidates to test functionally due to their reported roles in oxidative stress and apoptosis respectively (Table 1).

View this table:
Table 1 Genes differentially expressed between H-RasG12V-p38α−/− and H-RasG12V-WT MEFs that were selected for functional validation

We have previously reported that the absence of p38α enhances H-RasG12V-induced anchorage-independent growth [14]. Thus genes found to be up-regulated in H-RasG12V-p38α−/− MEFs were overexpressed in H-RasG12V-WT MEFs (Table 1) to test whether they could enhance the ability of transformed cells to grow in soft agar. Proliferation rates were also measured in parallel, and overexpression was confirmed by immunoblotting or RT–PCR (Supplementary Figure S3A at http://www.BiochemJ.org/bj/434/bj4340549add.htm), depending on the availability of antibodies. Interestingly, Mmp3, Cd9 and Rtn4 enhanced the ability of H-RasG12V-WT MEFs both to proliferate and to grow anchorage-independently (Table 1 and Supplementary Figures S3B, S3C and S4 at http://www.BiochemJ.org/bj/434/bj4340549add.htm). Similarly, Nck2 stimulated the proliferation of H-RasG12V-WT MEFs to p38α−/− levels, although it did not affect H- RasG12V-induced soft agar growth (Table 1 and Supplementary Figures S3B and S4). These results suggest that p38α might regulate different traits of the cancer cell (i.e. proliferation and anchorage-independent growth) by specifically targeting the expression of distinct genes in each case. Overall, it appears that the down-regulation of these four genes by p38α may contribute to its inhibitory effect on H-RasG12V-induced transformation of MEFs. In contrast, no conclusions could be drawn for Aig1 since we did not manage to express this gene (results not shown), whereas Pmp22 unexpectedly decreased, rather than enhanced, the transformed phenotype of H-RasG12V-WT MEFs (Supplementary Figures S3D and S4). This would agree with the pro-apoptotic and anti-proliferative functions ascribed to Pmp22 (Table 1), but raises the question of whether endogenous Pmp22 is actually functional in p38α−/− MEFs, since Pmp22 overexpression in these cells (where it is already up-regulated) attenuates their transformed phenotype as well (Supplementary Figure S3D).

In a complementary set of experiments, we overexpressed in H-RasG12V-p38α−/− MEFs three genes found to be up-regulated in H-RasG12V-WT cells (Table 1) and tested whether they could inhibit transformation. We found that Egln3 attenuated the proliferation rate and the ability to grow in soft agar of H-RasG12V-p38α−/− MEFs to approximately H-RasG12V-WT levels (Supplementary Figures S5A, S5B and 5C at http://www.BiochemJ.org/bj/434/bj4340549add.htm), in agreement with its lower expression levels in p38α−/− MEFs (Supplementary Table S6 at http://www.BiochemJ.org/bj/434/bj4340549add.htm) and its reported pro-apoptotic activity (Table 1). In contrast, Ctsh affected neither proliferation nor soft agar colony formation in H-RasG12V-p38α−/− MEFs (Table 1 and Supplementary Figures S5A and S5B). This is in agreement with the putative matrix remodelling function of Ctsh, which is likely to regulate tumour cell invasion rather than cell proliferation or anchorage-independent growth. Of note, Gstm2 did not significantly affect H-RasG12V-p38α−/− MEFs, but enhanced the proliferation (Supplementary Figure S5D) and soft agar growth [14] of H-RasG12V-WT MEFs. This apparent paradox could be explained if one considers that Gstm2 overexpression interferes with the activation and subsequent pro-apoptotic function of p38α in H-RasG12V-expressing cells [14]. Mechanistically, this involves Gstm2 binding to ASK1 (apoptosis signal-regulating kinase 1), an upstream activator of p38α, which in turn blocks the activation of the pathway by H-RasG12V-induced ROS (reactive oxygen species). Thus cells overexpressing H-RasG12V and Gstm2 bypass the inhibitory apoptotic response mediated by p38α, allowing the accumulation of high levels of ROS, which enhances malignant transformation. Since Gstm2 overexpression targets p38α activation, it is not surprising that it has no effect on H-RasG12V-p38α−/− MEFs.

Regulation of EGFR signalling by transcriptional targets of p38α in malignant transformation

Experiments using mouse models of cancer suggest that EGFR transcription could be targeted by p38α for tumorigenesis regulation in vivo [12], in agreement with studies using cultured cells [2022]. Moreover, EGFR activity is enhanced in p38α-deficient MEFs and contributes to the more transformed phenotype of H-RasG12V-p38α−/− MEFs [15]. We found a set of 19 genes that were deregulated between H-RasG12V-expressing WT and p38α−/− MEFs and were linked to EGFR signalling (Table 2). Most of these genes encode proteins that positively regulate EGFR signalling and were up-regulated in H-RasG12V-p38α−/− MEFs, indicating negative regulation by p38α. These include the plasma membrane protein CD9 that stimulates EGFR activity by facilitating the interaction of the receptor with its ligands, the mitogenic EGFR ligand Areg previously associated with malignant cell transformation, the transcription factor Klf5 that stimulates EGFR expression, the Src-family kinase Hck that participates in EGFR recycling and signal relay as well as in cell survival, and the tyrosine kinase substrate Eps8 that amplifies EGFR signalling following mitogenic stimulation (Table 2). Validation by qRT-PCR confirmed the differential expression of these genes in H-RasG12V-p38α−/− compared with H- RasG12V-WT MEFs (Figure 4A), as predicted from the microarray results.

View this table:
Table 2 EGFR pathway-modulating genes that are regulated by p38α

To strengthen the implication of the 19 gene dataset in the regulation of EGFR signalling, we performed GSEA. We found that our set of p38α-regulated genes aligned with the EGFR-associated gene signature (GEO ID GSE9599) of a study correlating the sensitivity of human pancreatic tumour cell lines to the EGFR inhibitor erlotinib with the expression of an EGFR-associated gene network that sustains high levels of EGFR activity [17]. Interestingly, GSEA revealed that our 19 gene cluster was significantly enriched in erlotinib-sensitive pancreatic tumours (FDR q value=0.165) (Supplementary Figure S6 at http://www.BiochemJ.org/bj/434/bj4340549add.htm) correlating with high EGFR activity. This result supports the importance of EGFR signalling inhibition by p38α in the regulation of cell transformation.

At the mechanistic level, we confirmed that Cd9 overexpression in H-RasG12V-WT MEFs not only enhanced their transformed phenotype to H-RasG12V-p38α−/− levels (Supplementary Figures S3B and S3C), but also activated EGFR kinase activity, as indicated by higher levels of phospho-EGFRY1068 (Figure 4B, left-hand panel) and its downstream signalling pathways ERK1/2 and Akt (Figure 4B, right-hand panel). This argues that p38α negatively regulates cell transformation, at least in part, by inhibiting EGFR via down-regulation of several genes that stimulate its activity or downstream signal relay. Of note, the interplay between EGFR and p38α in cell transformation might not be limited to a cluster of EGFR-regulating genes, but also involve direct regulation of the Egfr mRNA by p38α. We found that Egfr mRNA was up-regulated in the absence of p38α in three different cellular systems: H-RasG12V-transformed MEFs growing exponentially, subcutaneous xenografts derived from H-RasG12V-expressing MEFs and K-RasV12-induced lung tumours in mice (Figure 4C).

DISCUSSION

Cancer is a genetic disease characterized by the simultaneous disruption of several cellular traits required for organism homoeostasis. At the molecular level, the combination of deregulated cell proliferation and reduced apoptosis is thought to set the foundations for neoplastic progression. Accordingly, the ability of p38α to engage cell-cycle arrest and apoptosis has been shown to play an important tumour suppressor role [3,4]. p38α counteracts the transforming activity of oncogenic Ras in several cellular and animal models through various mechanisms, with most of them occurring at the post-translational level [11,12,14,23]. However, cancer progression is normally accompanied by profound alterations in gene expression, which prompted us to investigate transcriptional targets of p38α in the regulation of tumorigenesis. We have found a number of genes not previously connected to p38α signalling that are likely to mediate its negative effect on H-RasG12V-induced transformation. Intriguingly, our analysis has also identified various p38α-induced genes that are related to cell migration, invasion and angiogenesis, which may belong to a family of p38α effectors mediating some of its less characterized pro-oncogenic functions [24]. In fact, emerging evidence suggest that p38α may have oncogenic rather than tumour suppressive functions in some advanced human tumours [4]. Further studies are needed to understand the cellular contexts in which p38α can either suppress or promote tumour formation. In the present paper we report a set of p38α-regulated genes that are associated with tumour suppressor functions of p38α in different tumour models, and may be useful to identify human cancers where p38α negatively regulates tumour formation.

New p38α effectors in the regulation of cell survival and proliferation

Down-regulation of p38α enhances H-RasG12V-induced transformation of MEFs, as indicated by their increased proliferation and survival, refringent morphology, loss-of-contact inhibition and anchorage-independent growth [14]. Not surprisingly, we have found that genes stimulating cell proliferation and survival are negatively regulated by p38α, including Ralgds, a Ras effector that promotes cell survival [25], and Tnfrsf23, a putative TNF (tumour necrosis factor) receptor decoy that inhibits apoptosis [26]. Conversely, apoptosis-stimulating genes such as Egln3 [27] are up-regulated in the less-transformed H-RasG12V-WT MEFs, in line with the importance of the apoptotic activity of p38α for tumour suppression [14]. Of note, p38α seems to regulate Egln3 expression at both transcriptional and post-translational levels [28]. It is possible that other genes found in the present study are also post-transcriptionally regulated by p38α, for instance through the well-known p38α–MK2 (MAPK-activated protein kinase 2)-mediated stabilization of mRNAs containing AU-rich 3′-UTRs (untranslated regions) [29].

Potential membrane targets of p38α in cell transformation

Apart from survival and proliferation, p38α can also regulate other traits of the cancer cell, such as cell morphology, contact inhibition and anchorage-independent growth, which are likely to be regulated by signals relayed from plasma membrane proteins [14,15]. Previous attempts to identify proteins that were differentially expressed in the membranes of H-RasG12V-expressing WT and p38α−/− MEFs had limited success [13]. Interestingly, we now report several p38α-regulated genes encoding integral-to-membrane or membrane-associated proteins that may be involved in such processes (Supplementary Table S5). These include Ezr, encoding the cytoskeleton protein ezrin known to be important for the regulation of cell–cell contact inhibition and tumorigenesis [30], as well as various genes encoding transmembrane proteins, such as Tspan13 (tetraspanin 13), or Pcdh7 and Pcdh10 (protocadherins). Of note, certain protocadherins (i.e. Fat1) are emerging as important regulators of contact inhibition [31], a process where p38α plays a key role [15]. Future studies should reveal whether these proteins may link p38α signalling with the regulation of membrane-mediated cellular processes associated with malignant transformation.

The EGFR pathway as a transcriptional target of p38α in cell transformation

A striking finding of the present study has been the identification of 19 genes (Table 2) that impinge on EGFR signalling and whose expression seems to be co-ordinately regulated by p38α to inhibit malignant transformation. These include genes encoding proteins that enhance EGFR expression (Klf5), activation (Hck, Cd9, Prkca, Areg) and downstream signalling (Nck2, Limk2, Mmp3) or that regulate EGFR intracellular trafficking (Ctsl, Lamp2, Arfgap3, Cdc42ep5). Interestingly, we have shown that Cd9 overexpression not only stimulates EGFR signalling in H-RasG12V-WT cells, but also enhances transformation to H-RasG12V-p38α−/− levels. The ability of p38α to down-regulate EGFR signalling at various levels may also contribute to the inhibitory role of p38α in Ras-induced transformation, since Ras has been shown to require EGFR signalling for full transformation [32,33].

We have recently reported that p38α can inhibit EGFR at the post-translational level upon cell–cell contact [15], preventing cell overgrowth at high cellular densities. The relative contribution of transcriptional compared with post-translational regulation to the overall effect of p38α on EGFR signalling is unclear and may depend on the transformed state of the cell, the cell-cycle phase and the scope of the response induced by p38α. For instance, p38α regulates Egfr mRNA levels in proliferating H-RasG12V-transformed cells, but no differences were found in Egfr mRNA abundance between contact-inhibited, non-transformed WT and p38α−/− MEFs [15]. It is possible that the post-translational regulation of EGFR by p38α in contact inhibition is aimed at eliciting a fast response, whereas the long-term inhibition of H-RasG12V transformation by p38α requires tuning the steady-state expression levels of genes encoding proteins, which in turn impinge on EGFR signalling.

In summary, we have identified and validated in different models of tumorigenesis several p38α targets that are functionally implicated in transformation. A significant number of the genes that are down-regulated by p38α at the mRNA level encode positive regulators of EGFR signalling, suggesting that inhibition of this pathway probably represents an important tumour suppressor function of p38α.

AUTHOR CONTRIBUTION

Aneta Swat, Ignacio Dolado and Angel Nebreda designed the research; Aneta Swat performed most of the research with contributions from Ignacio Dolado, Ana Igea and Ana Cuadrado in some experiments; Gonzalo Gomez-Lopez and David Pisano performed the bioinformatics analysis; Aneta Swat, Ignacio Dolado and Angel Nebreda interpreted the results and wrote the paper.

FUNDING

This work was supported by the MICINN (Ministerio de Ciencia e Innovacion) [grant numbers BFU2007-60575, RD06/0020/0083], Fundacion Cientifica de la AECC and Fundacion La Caixa. A.S. acknowledges a FPU fellowship from the MICINN.

Acknowledgments

We thank Stephan Tenbaum and Juan Jose Ventura (CNIO, Madrid, Spain) for providing mouse lung tumour samples, and Vladimir Benes and Tommi Ivacevic (EMBL, Heidelberg, Germany) for help with microarray data processing.

Abbreviations: Cdc, cell division cycle; EGFR, EGF (epidermal growth factor) receptor; ERK, extracellular-signal-regulated kinase; FDR, false discovery rate; GEO, Gene Expression Omnibus; GSEA, gene set enrichment analysis; HEK, human embryonic kidney; MAPK, mitogen-activated protein kinase; MEF, mouse embryonic fibroblast; OHT, 4-hydroxytamoxifen; qRT-PCR, quantitative real-time PCR; ROS, reactive oxygen species; RT, reverse transcription; WT, wild-type

References

View Abstract