TEF-1 and C/EBPβ are major p38α MAPK-regulated transcription factors in proliferating cardiomyocytes

Abstract

p38 MAPKs (mitogen-activated protein kinases) play important roles in the regulation of cellular responses to environmental stress. Recently, this signalling pathway has also been implicated in the regulation of processes unrelated to stress, for example, in T lymphocytes and cardiomyocytes. In order to identify molecular targets responsible for the housekeeping functions of p38 MAPKs, we have analysed the differences in the transcriptomes of normally proliferating wild-type and p38α knockout immortalized embryonic cardiomyocytes. Interestingly, many potential components of the myocardium extracellular matrix were found to be upregulated in the absence of p38α. Further analysis of the microarray data identified TEF-1 (transcriptional enhancer factor-1), a known regulator of heart-specific gene expression, and C/EBPβ (CCAAT/enhancer-binding protein β), as the two transcription factors the binding sites of which were most enriched in the promoters of p38α-regulated genes. We have focused on the study of the extracellular matrix component COL1A1 (α1 chain of type I collagen) and found evidence for the involvement of both TEF-1 and C/EBPβ in the p38α-dependent inhibition of COL1A1 transcription. Our data therefore show that p38 MAPKs regulate TEF-1 and C/EBPβ transcriptional activity in the absence of environmental stress and suggests a role for p38α in the expression of extracellular matrix components that maintain organ architecture.

INTRODUCTION

The p38 MAPK (mitogen-activated protein kinase) pathway was originally described as a signalling cascade activated in response to stress conditions and pro-inflammatory cytokines. However, it has become clear that p38 MAPKs can also regulate cellular processes that are unrelated to inflammatory and environmental stress responses, such as proliferation, differentiation and the survival of several cell types (reviewed in [1]). Four p38 MAPK family members have been identified and termed p38α, β, γ and δ. p38α is ubiquitiously expressed and is thought to be the most abundant p38 MAPK in most tissues. Many types of proteins can be phosphorylated by p38 MAPKs, but transcription factors probably constitute the largest group of substrates [2]. The activity of several transcription factors, such as ATF (activating transcription factor)-1/2, CHOP [C/EBPα (CCAAT enhancer-binding protein)-homologous protein], p53, MEF (myocyte enhancer factor) 2A/C and C/EBPα, can be regulated by p38 MAPKs either via direct phosphorylation [3] or through downstream kinases such as MSK1 (mitogen- and stress-activated protein kinase 1) [4]. In addition, p38 MAPKs can also regulate gene expression at the post-transcriptional level [5].

The existence of shared substrates among p38 MAPK family members suggests that they may have some overlapping functions. However, the degree of redundancy between the different p38 MAPKs in the regulation of cellular processes, both under normal and stress conditions, is not clear yet. The generation of knockout mice has revealed that p38α plays an essential role in placental morphogenesis [6–8]. Furthermore, a role for the p38 MAPK pathway has been shown in the regulation of T-cell function [9], as well as in the development of or pathophysiology in diverse organs, including the heart [10].

Stimuli that promote hypertrophic growth of cardiomyocytes, such as pressure overload and endothelin-1 stimulation, induce the activation of the ERK (extracellular-signal-regulated kinase) and p38 MAPK pathways [11–13]. Osmotic, oxidative and ischaemia/reperfusion stressors also induce p38 MAPK activation in the heart [14]. Heart injury provokes a complex response that initiates the process of myocardium remodelling, which involves cardiomyocyte hypertrophy, fibrosis and cell death. This fibrosis is caused by the accumulation of the main components of the cardiac ECM (extracellular matrix), such as COL1 (collagen type 1) and COL3 (type 3) molecules, which play essential roles in the maintenance of myocardium integrity under normal conditions [15].

In order to better understand the role of p38 MAPKs in the control of heart pathophysiology and the regulation of cardiomyocyte function, we have investigated genes regulated by p38α in cardiomyocyte cell lines derived from wt (wild-type) and p38α^−/− (p38α null mutant) embryo hearts. In the present study, we provide evidence for the stress-independent control of gene expression by p38α in non-stimulated cells. We show that there is an increased expression of several components of the myocardium ECM in p38α^−/− cardiomyocytes, suggesting that p38α plays a role in the maintenance of myocardium structural integrity. Moreover, we have identified two p38α-regulated transcription factors, C/EBPβ and TEF-1 (transcriptional enhancer factor-1), which are involved in the control of ECM gene expression in cardiomyocytes. The involvement of TEF-1 in p38 MAPK-regulated gene expression has not been described to date.

MATERIALS AND METHODS

Cell lines

The generation of immortalized p38α^−/− and wt cardiomyocyte cell lines and culture conditions have been previously described [6,16]. The construction of MSCV (murine stem cell virus)-p38α and MSCV-p38α-D/A (with the mutation D168A) and the procedure for retroviral infections have been already reported [17]. Stably infected cell lines were selected and grown in culture medium supplemented with 100 μg/ml hygromycin (Sigma).

Microarray analysis

Total RNA was extracted with RNA Clean (Hybaid) and further purified with an RNeasy mini kit (Qiagen). Direct labelling of 20 μg of total RNA primed with oligo(dT) was performed in the presence of Cy-5- and Cy-3-dUTP (Amersham Pharmacia Biotech) using SuperScript II (Invitrogen). The Cy5- and Cy3-labelled cDNA samples were then mixed together and purified using a GFX DNA purification kit (Amersham Pharmacia Biotech). Before hybridization, samples were supplemented with 1.2 μg of salmon sperm DNA and 5 μg of poly(dA) and dried. The high-density glass microarray chips were prepared at the EMBL (European Molecular Biology Laboratory) core facility by spotting of the PCR products prepared from the NIA (National Institute on Ageing) 15K mouse cDNA set [18]. The microarray slides were pre-hybridized in 6×SSC, 0.5% SDS and 1% BSA at 42 °C for 1 h and incubated in boiled water for 2 min prior to hybridization. The cDNA samples were resuspended in 15 μl of hybridization buffer (50% formamide, 6×SSC, 0.5% SDS and 5% Denhardt's solution) and denatured by incubating at 95 °C for 2 min. Hybridization was carried out at 42 °C for 16 h. The slides were washed for 10 min each with 2×SSC, then with 0.5×SSC/0.1% SDS followed by 0.1×SSC/0.1% SDS at room temperature. Scanning was performed with GenePix 4000B (Axon). The data acquisition and initial data analysis was performed with GenePix Pro 3.0 software and the data tables were analysed further with the program Excel in order to obtain the gene list.

Quality control was performed to confirm scanner alignment and the absence of significant bubbles and scratches. Scatter plots were further used to eliminate unacceptable hybridization data. Hybridizations were performed in triplicate, using total RNA samples extracted from three independent batches of proliferating cells. In two hybridizations, the RNAs from p38α^−/− and wt cells were labelled with Cy-5- and Cy-3-dUTP respectively, whereas in the third hybridization, the dyes were swapped. The multiple spike-in RNA control templates were added to the sample upon direct labelling, and successful labelling and hybridization was confirmed in each hybridization [19]. Twelve housekeeping genes in the array gave an average ROM (ratio of medians) value of 1.30±0.26, which was confirmed to be acceptable. The GenePix Pro program calculates the normalization factor for each hybridization, based on the premise that the arithmetic mean of the ratios from every feature on the given array should be equal to 1. Normalization was therefore performed by multiplying this factor with the ROM value for each gene. The program also identifies elements that do not produce good alignment to the expected spotted area as ‘flagged’ spots, indicative of a low-quality hybridization of the specific genes. All flagged genes were removed from the list. Genes fulfilling all these criteria were sorted by ROM and those that showed a greater than ±1.5-fold change were selected from each hybridization. Finally, data from 3 independent hybridizations were compared and only those genes that were identified in all 3 experiments were selected for the final list. The microarray data have been deposited in the GEO database (under accession number GSE1568).

Northern blotting and nuclear run-on

Total RNA was prepared from proliferating cells with TRIzol reagent (Life Technology) and a sample of 20 μg was used for Northern blotting analysis as previously described [17]. The COL1A1 (α1 chain of COL1) and the COL3A1 (α3 chain of COL3) probes were prepared from the corresponding clones of the NIA mouse cDNA set used for the microarray. A fragment of the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) open reading frame (nt 134–600) was used for normalization. After autoradiography, the bands were excised and DNA quantified using a Fuji FLA 2000 PhosphorImager instrument. The experiments were performed at least twice.

The run-on analysis was performed as previously described [17]. The membranes were analysed and quantified using a PhosphorImager instrument.

Cell extracts and immunoblotting

Whole-cell extracts were prepared and analysed by immunoblotting as described previously [17,20]. The membranes were probed with the following antibodies according to the manufacturer's instructions: anti-p38α and anti-C/EBPβ (Santa Cruz Biotechnology), anti-TEF-1 (Transduction Laboratories), anti-COL1 (Calbiochem), anti-tubulin (Sigma), anti-phospho-ERK (Cell Signaling) and anti-MKK (MAPK kinase) 6 [20].

PRIMA (promoter integration in microarray analysis)

The PRIMA program identifies transcription factors the binding sites of which are significantly over-represented in particular promoters [21]. The fingerprint contains the number of hits (putative binding-sites) of the various transcription factors identified in the promoter. EXPANDER then executes the PRIMA program on a fixed set of pre-computed fingerprints, which are constructed as a set of about 15000 mouse promoter sequences, spanning from 1000 bp upstream to 200 bp downstream of the transcriptional start site [22]. In the present study, the scan was performed for each transcription factor motif in the TRANSFAC (version 5.4) database that corresponds to a human transcription factor. The set of fingerprints was prepared on mouse promoters (15000 promoters, Ensemble release 13.30) using a threshold that corresponds to a hit frequency of 10% in random sequences. We have used the locus ID of genes that were mostly retrieved from the NIA site (http://lgsun.grc.nia.nih.gov/cDNA/15k.html) and the rest were from Source (http://source.stanford.edu). In total, we used locus IDs for 9113 genes on the array and for 114 p38α-regulated genes.

Nuclear lysates and EMSA (electrophoretic mobility-shift assay)

Nuclear and cytosolic extracts to be used in assaying DNA binding activity and the amount of C/EBP protein were prepared as described previously [23]. For C/EBP EMSA, the following oligonucleotide probes were used: 5′-GATCGGACGTCACATTGCACAATCTTAATAAT-3′ (interleukin 6 C/EBP binding site), 5′-GGACGTCACACTACAAACTCTTAATAA-3′ (mutant interleukin 6 C/EBP binding site). Nuclear extracts (3 μg) were incubated for 5 min on ice in 20 μl of 10% glycerol, 60 mM KCl, 1 mM EDTA, 1 mM DTT (dithiothreitol), 2 μg of poly(dI-dC) (Boehringer Mannheim). The γ-³²P-labelled double-stranded probe [0.2 ng, (4–6)×10⁴ c.p.m.] was then added with or without a 100-fold molar excess of the competitor wt or mutant oligonucleotide. Where indicated, 1 μg of the anti-C/EBP antibody (Santa Cruz Biotechnology) or purified rabbit IgG (Sigma) was added before the probe and the mixture was incubated on ice for 1 h. The reactions were incubated at room temperature for 20 min and analysed on a 5% acrylamide/bisacrylamide (30:1) gel in 22.5 mM Tris/borate, 0.5 mM EDTA and 3.7 mM mercaptoethanol. Gels were dried and subjected to autoradiography.

For the TEF-1 EMSA, the M-CAT [muscle C, A and T sites, sequence related to CATTCC(A/T)] oligonucleotide probe 5′-GATCGGACGTCACATTGCACAATCTTAATAAT-3′ was used. Nuclear extracts (5 μg) were prepared as described previously [24] and incubated for 5 min on ice in 20 μl of 25 mM Tris (pH 7.5), 10% glycerol, 100 mM KCl, 2 mM EDTA, 1 mM DTT and 2 μg of poly(dI-dC) (Boehringer Mannheim). The γ-³²P-labelled double-stranded probe (0.2 ng, 10×10⁴ c.p.m.) was then added with or without a 100-fold molar excess of competitor wt oligonucleotide. For the supershift assay, 0.5 μg of anti-TEF-1 antibody (Transduction) or purified rabbit IgG was added before the probe and the mixture was incubated on ice for 30 min. The reactions were processed and analysed as described above for C/EBP.

ChIP (chromatin immunoprecipitation)

ChIP assays were performed using the protocol for the acetyl-histone H4 ChIP Assay kit (Upstate Biotechnology) using 4 ×10⁷ wt or p38α^−/− exponentially growing cells for each immunoprecipitation. Cells were cross-linked by the addition of formaldehyde (1% final concentration) to the medium for 30 min at 37 °C and the reaction was stopped by the addiction of glycine. Cells were washed and scraped in ice-cold PBS supplemented with protease inhibitors, collected by centrifugation and resuspended at 1×10⁸/ml in ChIP lysis buffer [1% SDS, 10 mM EDTA, 50 mM Tris/HCl (pH 8.0) and protease inhibitors]. After 10 min of incubation on ice, aliquots of 0.4 ml were sonicated. The lysates were then diluted ×10 with ChIP dilution buffer [0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris/HCl (pH 8.0) and 167 mM NaCl] and pre-cleared by incubation for 1 h at 4 °C with Protein A beads previously washed and resuspended in ChIP dilution buffer containing 0.4 mg/ml salmon sperm DNA. After pre-clearing, 4 μg of the anti-C/EBPβ or anti-TEF-1 antibodies was added. For wt cells, a control using purified rabbit IgG was included. After overnight incubation at 4 °C, the immunocomplexes were collected with Protein A beads and washed twice with both low-salt and high-salt buffers, LiCl wash buffer and finally TE [10 mM Tris/HCl (pH 7.5)/1 mM EDTA] 1×. The complexes were then eluted by 3 sequential incubations with 170 μl of freshly prepared elution buffer (1% SDS, 0.1 M NaHCO₃) for 15 min at room temperature. The cross-linking was reverted by the addition of 20 μl of 5 M NaCl and by heating at 65 °C for 5 h. The control DNAs were also incubated at 65 °C after the addition of 5 M NaCl and processed as described for the immunoprecipitated DNA. After Proteinase K digestion, the DNA was purified by phenol/chloroform extraction and precipitated. The pellets were resuspended in 20 μl and 1 μl was used for PCR using AmpliTaq (Perkin Elmer). The PCR primers (forward, 5′-GCCAGAGGTGCTGTCAC-3′; reverse, 5′-GGTGTGTCTGGCATGGCAG-3′) were designed based on the published mouse COL1A1 promoter sequence [25], to amplify a region conserved in the rat and human COL1A1 promoters.

RESULTS

The inactivation of p38α in cardiomyocytes changes the levels of mRNAs encoding ECM components

We have previously proposed that the low level of p38α basal activity present in normally proliferating cells could modulate the expression of specific genes, which may be related to house-keeping functions of the p38 MAPK pathway [17]. This idea prompted us to look for genes regulated by p38α under normal growing conditions. We have examined the gene expression profiles of wt and p38α^−/− cell lines derived from mouse embryo cardiomyocytes, which express heart-specific markers (results not shown and [6]).

Total RNA was prepared from three independent batches of proliferating p38α^−/− and wt cardiomyocyte cultures and hybridized on to high-density glass microarray chips containing the NIA 15K mouse cDNA clone set [18]. Strict measures were applied for the purpose of quality control (as described in the Materials and methods section) and the data were obtained from the three experiments that had the best possible conditions. The genes that showed a greater than ±1.5-fold change in all three hybridization experiments were included in the final list. The expression level of 54 genes was found to be upregulated, whereas 60 genes were downregulated in p38α^−/− cardiomyocytes under normal growing conditions (Figure 1; see Supplementary Tables 1 and 2 at http://www.BiochemJ.org/bj/396/bj3960163add.htm). Approx. 40% of the genes in the list were Riken or IMAGE clones that did not show homology with any known genes based on BLAST searches against DNA sequence databases. Analysis by Northern blotting of several genes that showed either large or small changes in their expression level confirmed the validity of the microarray results (results not shown).

Figure 1. Summary of the genes for which expression levels significantly changed in proliferating p38α^−/− versus wt cardiomyocytes.

Three independent p38α^−/− and wt cardiomyocyte cultures were analysed and only those genes that showed a greater than ±1.5-fold changes in all three experiments were selected. Gene functions were initially categorized based on the NIA mouse 15K cDNA gene ID list and modified when necessary. Note that many genes that are differentially regulated in the p38α^−/− cardiomyocytes encode matrix and structural proteins.

A statistical validation of the results was performed using the Onto-Express program [26], which associates Unigene clusters with GeneOntology terms relative to biochemical and molecular functions, biological processes, cellular components and roles, and then computes for each function the probability that the regulated genes could have been picked randomly from the population (i.e. the total number of genes in the microarray). This analysis revealed that the expression of ECM components was the cellular function potentially regulated by p38α with the best P value (7.0e⁵), which is also in agreement with the functional categorization based on the NIA mouse 15K cDNA gene ID list (Figure 1).

p38α downregulates collagen transcripts in cardiomyocytes

Several collagen transcripts were upregulated in the p38α^−/− cells (see Supplementary Table 1 at http://www.BiochemJ.org/bj/396/bj3960163add.htm). COL1 is a major component of bone, skin, tendons, blood vessel walls and connective tissues and represents approx. 85% of myocardium ECM proteins. The COL1A1 and COL1A2 genes, which encode the α1 and α2 chains of COL1 respectively, were upregulated to approx. equivalent levels in our microarray analysis, supporting the specificity of the result. Previous studies have shown that the p38 MAPK pathway is involved in the control of COL1A1 and COL1A2 gene expression upon stress treatment [27,28]. By Northern blotting we confirmed a 3–4-fold increase in the COL1A1 transcript content of p38α knockout cardiomyocytes (Figures 2A and 2B). COL3A1 mRNA expression was also upregulated in p38α^−/− cardiomyocytes (Figures 2B and 2C), which is in agreement with previous reports showing that the COL1A1 and COL3A1 genes frequently have the same expression pattern [29]. We focused on the COL1A1 gene for further experiments.

Figure 2. Up-regulation of COL1A1 and COL3A1 in p38α^−/− cardiomyocytes.

(A) and (C) Northern blotting analysis of RNAs prepared from wt and p38α^−/− cardiomyocytes. The membranes were hybridized with COL1A1, COL3A1 and GAPDH probes as indicated. (B) PhosphorImager quantification of COL1A1 and COL3A1 transcript levels (normalized to GAPDH values) in the indicated cell lines. The results were compiled from six or two independent experiments for COL1A1 and COL3A1 respectively.

COL1A1 mRNA levels are regulated by p38α under stress and normal growth conditions

To confirm that the upregulation of the COL1A1 transcript level was directly related to the absence of p38α, we re-introduced either wt p38α or the kinase-dead mutant p38α-D/A into p38α^−/− cardiomyocytes [17]. As shown in Figure 3(A), the presence of p38α decreased the COL1A1 mRNA content in p38α^−/− cardiomyocytes to approx. the same levels as in wt cells. However, the expression of p38α-D/A further increased the cellular level of COL1A1 mRNA. At the protein level, the amount of COL1A1 was also increased in the p38α^−/− cardiomyocytes and was decreased by expression of wt p38α, but not by p38α-D/A (Figure 3B, upper panel). As a control, MKK6 upregulation [17] was confirmed in the absence of p38α activity, either in p38α^−/− cardiomyocytes or in those transfected with p38α-D/A (Figure 3B, middle panel).

Figure 3. COL1A1 up-regulation in p38α^−/− cardiomyocytes is reversed by p38α expression.

(A) Cardiomyocytes were infected with a retrovirus expressing wt p38α or the kinase-dead mutant p38α-D/A. RNA was prepared from the infected cells and the COL1A1 and GAPDH transcript levels were determined by Northern blotting. PhosphorImager quantification (normalized to GAPDH) from four independent experiments is shown. (B) COL1 protein expression was analysed in the same cell lines by immunoblotting. The filter was re-probed with antibodies to MKK6, the expression of which is known to be negatively regulated by p38α activity, and tubulin as a loading control. The results are representative of three independent experiments.

The expression of the COL1A1 gene is tightly regulated and several transcription factors are known to bind to its promoter [30]. To analyse the mechanism of COL1A1 upregulation, we investigated the activation status of the ERK MAPK pathway in normally proliferating p38α^−/− cells, as this pathway can also regulate COL1A1 expression [31] and sometimes is negatively regulated by p38 MAPK [32]. In agreement with previous results [16], we found that the deletion of p38α increased the basal levels of phosphorylated and active ERK MAPKs (Figure 4A, upper panel). The re-introduction of wt p38α, but not the inactive form p38α-D/A, in p38α^−/− cardiomyocytes decreased ERK phosphorylation to the same level as in wt cells (Figure 4A, upper panel). These results suggested that ERK activation could potentially contribute to the upregulation of COL1A1 expression observed in p38α^−/− cardiomyocytes. To address this possibility, we used the ERK inhibitors PD98059 and U0126. Incubation of p38α^−/− cardiomyocytes with either of these inhibitors for up to 36 h did not affect the level of COL1A1 mRNA, as determined by Northern blotting (results not shown). It therefore seems that the main cause of an increased COL1A1 mRNA level in the p38α^−/− cardiomyocytes is p38α inactivation rather than the upregulation of ERK.

Figure 4. Increased basal ERK activity in the p38α^−/− cardiomyocytes and super-induction of COL1A1 upon TNFα treatment.

(A) Cell lysates were prepared from the indicated cardiomyocyte lines and analysed by immunoblotting. Antibodies that specifically recognize the phosphorylated and active forms of ERK1/2 were used. The total amount of p38α or ERK1/2 was also determined and the filter was re-probed with a tubulin antibody as the loading control. (B) Cardiomyocytes were starved overnight (0.5% foetal calf serum) and then treated with TNFα (20 ng/ml) for 6 h. Total RNA was prepared and analysed by Northern blotting. The Phosphorimager quantification data (normalized to GAPDH values) are shown at the bottom of the blots. The data are compiled from two independent experiments.

The pro-inflammatory cytokine TNF (tumour necrosis factor)α has a negative effect on collagen expression in cardiac fibroblasts [33]. Therefore we analysed the role of p38α in decreasing the level of COL1A1 mRNA in the cardiomyocyte lines stimulated with TNFα. For this purpose, cells were stimulated with a low concentration of TNFα (20 ng/ml) for 6 h. This mild treatment was necessary to avoid the induction of apoptosis [16]. We found that TNFα stimulation resulted in the inhibition of COL1A1 expression only in wt cardiomyocytes, consistent with the TNFα-induced activation of p38α, whereas COL1A1 transcription was further induced by TNFα stimulation in the p38α^−/− cardiomyocytes (Figure 4B).

Taken together these results show that, in cardiomyocytes, COL1A1 mRNA levels are negatively regulated by p38α in a kinase-dependent manner under both normal proliferative and stress conditions.

The p38 MAPK pathway negatively regulates COL1A1 promoter transcriptional activity

To characterize the molecular mechanism by which p38α regulated the COL1A1 mRNA expression level, we performed nuclear run-on experiments. Using nuclei purified from proliferating cardiomyocytes, transcription from the COL1A1 promoter was found to be about 2-fold higher in p38α^−/− than in wt cardiomyocytes (Figure 5A). This observation suggested a direct negative role for p38α in the regulation of the COL1A1 promoter's transcriptional activity. As COL1A1 has been reported to be post-transcriptionally regulated in some cell lines (reviewed in [34]), we also investigated this possibility. Cardiomyocytes were treated with actinomycin D for 3 and 6 h (longer actinomycin D treatment is toxic in these cells; C. Ambrosino unpublished work) and the COL1A1 mRNA level was determined by Northern blotting. We observed only minor differences, that were reproducible and are unlikely to be significant, in the stability of the COL1A1 transcript between wt and p38α^−/− cardiomyocytes (Figure 5B), therefore we concentrated on the transcriptional regulation of COL1A1 promoter.

Figure 5. COL1A1 gene expression is regulated by p38α at the transcriptional level.

(A) Proliferating cardiomyocytes were used for the run-on analysis with 5 μg of the indicated plasmids. The signals were visualized by autoradiography (upper panel) and quantified by PhosphorImager (arbitrary units). The data were normalized to GAPDH values and plotted at the bottom of the blots. (B) Cardiomyocytes were treated with actinomycin D (Act; 2.5 μg/ml) and RNA was prepared, at the indicated times, and analysed by Northern blotting (upper panel). The amount of COL1A1 mRNA remaining after the treatment (normalized to GAPDH values) is shown at the bottom of the blots.

Identification of the transcriptional network regulated by p38α in cardiomyocytes

To characterize the transcriptional network involved in gene expression regulated by p38α, we performed a PRIMA analysis on the microarray data set. PRIMA is a program that identifies transcription factors for which the binding sites are enriched in a defined set of promoters [21]. The analysis was performed by comparing the promoter regions of 9113 genes on the array (considered as a background set) either with those of the entire set of p38α-regulated genes, or with the sets of genes up-regulated or down-regulated in two different clusters. In the first case, we found that C/EBP had the best P value (2.46×10⁻⁴) among the transcription factors that were present in the promoter of differentially regulated genes, followed by TEF-1 and Nkx6-2 that had P values of 8.32×10⁻³ and 2.7×10⁻² respectively. In the cluster of up-regulated genes, TEF-1 had the largest value (P=1.36×10⁻⁴), followed by the transcription factors C/EBP (P=6.72×10⁻³), Octamer (P=9.71×10⁻³) and NF-AT (nuclear factor of activated T-cells) (P=1.89×10⁻²). In the cluster of down-regulated genes, only AP-1 (activator protein-1) was found with a P value of less than 0.005. Taken together, there was a clear enrichment of binding sites for TEF-1 and C/EBP in the promoters of p38α-regulated genes, suggesting an important role for these transcription factors in the regulation of gene expression by p38α in cardiomyocytes. Furthermore, the genes harbouring a TEF-1 binding site in their promoter were all found to be up-regulated in the p38α^−/− cells, whereas those with C/EBP binding sites included both up-regulated and down-regulated genes (Table 1). Interestingly, the only genes that had binding sites for both TEF-1 and C/EBP in their promoters were COL1A1, COL1A2, Tra1 and cathepsin D.

Table 1. Results of the PRIMA analysis showing genes for which promoters are enriched in binding sites for TEF-1 and C/EBP transcription factors.

The cut-off P values for TEF-1 and C/EBP were 0.000136 and 0.000431 respectively. GB, Gen Bank.

					Binding site
Gene name	Gene symbol	GB accession number	UniGene ID	Average fold-change	TEF-1	C/EBP
Cathepsin D	Ctsd	BG087382	Mm.231395	2.66	*	*
Procollagen, type I, alpha 1	Col1a1	BG073196	Mm.277735	5.10	*	*
Procollagen, type I, alpha 2	Col1a2	BG073735	Mm.277792	4.39	*	*
Tumour rejection antigen gp96	Tra1	BG078337	Mm.87773	2.94	*	*
High mobility group AT-hook 2	Hmga2	BG086396	Mm.157190	−1.92		*
Ladinin	Lad1	BG076566	Mm.36726	−2.62		*
Myeloid ecotropic viral integration site-related gene 1	Mrg1	BG067314	Mm.247566	−1.69		*
Plexin A4	Plxna4	BG086250	Mm.242174	−2.30		*
RIKEN cDNA 1110032D12 gene	1110032D12Rik	BG076581	Mm.325307	1.98		*
RIKEN cDNA 1110033O09 gene	1110033O09Rik	BG067483	Mm.21171	2.08		*
ROD1 regulator of differentiation 1 (Schizosaccharomyces pombe)	Rod1	BG063016	Mm.331640	−3.87		*
TBC1 domain family, member 13	Tbc1d13	BG072158	Mm.19172	1.70		*
Activated leucocyte cell adhesion molecule	Alcam	BG086024	Mm.288282	1.66	*
ADP-ribosylation factor 1	Arf1	BG082675	Mm.371546	1.66	*
ATPase, Na+/K+ transporting, beta 1 polypeptide	Atp1b1	AW544616	Mm.4550	2.19	*
Barren homolog (Drosophila)	Brrn1	BG085939	Mm.29786	3.01	*
Centaurin, delta 2	Centd2	BG087361	Mm.277687	2.00	*
Cullin 3	Cul3	BG085494	Mm.12665	1.58	*
Cyclin D3	Ccnd3	BG066310	Mm.246520	1.85	*
Decorin Dcn	D16Bwg1547e	BG088078	Mm.56769	2.73	*
DNA segment, Chr 16, Brigham & Women's Genetics 1547 expressed	D16Bwg1547e	BG063060	Mm.567	4.12	*
Immunity-related GTPase family, Q1	AF322649	BG069854	Mm.41108	2.19	*
Integral membrane protein 2B	Itm2b	BG088523	Mm.4266	1.89	*
Mannose-P-dolichol utilization defect 1	Mpdu1	BG083639	Mm.89579	2.81	*
Microsomal glutathione S-transferase 1	Mgst1	BG086330	Mm.14796	2.23	*
Pleiomorphic adenoma gene-like 1	Plagl1	BG085853	Mm.287857	3.60	*
Potassium intermediate/small conductance calcium-activated channel, subfamily N, member 4	Mpdu1	BG080543	Mm.9911	2.52	*
Procollagen, type III, alpha 1	Col3a1	BG085576	Mm.249555	4.25	*
RIKEN cDNA 1110059P08 gene	1110059P08Rik	C85340	Mm.227983	2.15	*
RIKEN cDNA 1190007F08 gene	1190007F08Rik	BG076322	Mm.260690	2.32	*
Secreted acidic cysteine rich glycoprotein	Sparc	BG078305	Mm.291442	3.25	*
Sema domain, immunoglobulin domain (Ig), short basic domain, secreted, (semaphorin) 3 F	Sema3f	BG074382	Mm.12903	2.15	*
Vinculin	Vcl	BG063694	Mm.279361	2.29	*

p38α regulates TEF-1 and C/EBPβ binding to the COL1A1 promoter

The regulation of COL1A1 gene expression is known to involve several transcription factors, including C/EBP [30]. We have also identified 3 potential TEF-1 binding sites in the COL1A1 promoter [25] at positions −603 bp to −598 bp, −540 bp to −535 bp and −359 bp to −354 bp (as described below; Figure 7A), but the possible regulation of COL1 expression by TEF-1 has never been investigated before.

Figure 7. Differential loading of C/EBPβ and TEF-1 on to the COL1A1 promoter in wt and p38α^−/− cardiomyocytes.

(A) Schematic representation of the COL1A1 promoter indicating the positions of transcription factor binding sites and the PCR primers (a and b). CBF, CCAAT-binding factor. (B) ChIP assays were performed using whole lysates prepared from 40×10⁶ cells: wt (lanes 1, 2, 8 and 11), p38α^−/− infected with MSCV-p38α (lanes 3, 9 and 12) or p38α^−/− infected with the MSCV vector alone (lanes 4, 10 and 13). The protein extracts were incubated with the anti-C/EBPβ antibody (lanes 1, 3 and 4), anti-TEF-1 antibody (lanes 8–10) or with the same amount of IgG, as a control (lanes 2, 11–13). DNA purified from one-tenth of total lysates was also used as a control (Input; lanes 5–7 and 14–16). The ChIP assay was repeated twice.

We analysed C/EBP and TEF-1 activity in wt and p38α^−/− cardiomyocytes by EMSA. We found that C/EBP DNA binding activity was decreased in cells lacking p38α (Figure 6A, lanes 1 and 7) and re-introduction of wt p38α, but not of catalytically inactive p38α-D/A, was able to restore C/EBP DNA binding activity to the same level as in wt cells (Figure 6A). By super-shift assay, we found that C/EBPβ was the most abundant component of the C/EBP protein complex and was specifically regulated in the cardiomyocyte lines (Figure 6A, *). The increase in C/EBP binding was probably due to direct regulation of its DNA binding activity, as the amount of the protein was very similar in wt and p38α^−/− cardiomyocytes (Figure 6B). No changes in the nuclear/cytoplasmic distribution of the two isoforms of C/EBPβ, p35 and p20, were observed (results not shown).

Figure 6. Regulation of C/EBPβ and TEF-1 DNA binding activity by p38α.

(A) C/EBP DNA binding activity was determined by band-shift analysis using nuclear extracts of the indicated cardiomyocyte lines. The specificity of the binding was determined by competition with an excess (100×) of the unlabelled oligonucleotide (C/EBP wt) or a mutated version (C/EBP μ). The C/EBPs and the non-specific (NS) complexes are indicated. Identification of C/EBP proteins present in the complexes was performed by super-shift analysis using the same nuclear extracts. The C/EBP proteins in the complex were identified by the addition of specific antibodies (1 μg) against C/EBPα (lanes 5, 11, 17 and 23) or C/EBPβ (lanes 6, 12, 18 and 24). The anti-C/EBPβ antibody recognizes both p35 and p20 proteins. An equal amount of purified IgG was used as a control (lanes 4, 10, 16, and 22). The super-shifted complexes are indicated by an asterisk. The results are representative of four independent experiments. (B) The nuclear levels of the p35 and p20 C/EBPβ proteins in the same nuclear extracts as used for EMSA were determined by Western blotting. (C) TEF-1 DNA binding activity was determined by EMSA and the specificity of the binding was established by competition with an unlabelled oligonucleotide (lanes 6–9). The participation of the TEF-1 protein in the M-CAT DNA binding complexes was verified by super-shift assay using an anti-TEF-1 specific antibody (lanes 14–17). An equal amount of purified IgG was used as a control (lanes 10–13). The TEF-1 EMSA assay was performed twice. (D) The same nuclear extracts used for EMSA were also analysed by Western blotting with an anti-TEF-1 antibody (left panel). The histogram represents the normalization of TEF-1 DNA binding activity, as determined by EMSA and super-shift assays, to the amount of nuclear TEF-1 detected by Western blotting. Quantifications were performed using the Image J program. AU, arbitrary units.

We also investigated TEF-1 activity and found more pronounced binding to the M-CAT probe in the p38α^−/− cardiomyocytes (Figures 6C and 6D). Moreover, upon re-expression of p38α in p38α^−/− cells, TEF-1 DNA binding activity reverted to the same level as in wt cells. By supershift assay, TEF-1 appears to be the most relevant component of such complexes among the TEF-1 family members. We confirmed that the level of TEF-1 expression was not very different between wt and p38α^−/− cell lines, supporting the idea that the p38 MAPK pathway regulates TEF-1 DNA binding activity, rather than its expression (Figure 6D).

Finally, we also investigated differences in the loading of C/EBPβ and TEF-1 on to the COL1A1 promoter between wt and p38α^−/− cardiomyocytes. ChIP assays, which provide a real-time estimation of DNA binding proteins loaded on to a particular promoter, were performed using exponentially growing wt and p38α^−/− cardiomyocytes. The primers were chosen in order to amplify the region of the COL1A1 promoter (−620 to −225) containing the different C/EBP and TEF-1 binding sites (Figure 7A). Consistent with the EMSAs, we detected more C/EPBβ loaded on to the COL1A1 promoter in the wt cells (Figure 7B, upper panel). The re-introduction of p38α into the p38α^−/− cardiomyocytes partially reverted C/EBPβ binding to the same level as that observed in wt cells (Figure 7B, lane 3). TEF-1 was also detected on the COL1A1 promoter but only in the p38α-lacking cells (Figure 7B, lower panel).

These results strongly suggest that the up-regulation of COL1A1 gene transcription in p38α^−/− cells is related to the ability of p38α to regulate the basal DNA binding activities of C/EBPβ and TEF-1.

DISCUSSION

The involvement of the p38 MAPK pathway in inflammatory and environmental stress responses has been extensively investigated in many different cellular contexts. However, the molecular targets responsible for the housekeeping functions of p38 MAPK are poorly characterized. By analysing global differences in the transcriptome between wt and p38α^−/− cardiomyocytes, we found changes in the expression of genes that are associated with several cellular processes (Figure 1), implying possible functions for the low-level activity of p38 MAPKs in non-stressed proliferating cells. A similar question has recently been addressed in lymphocyte cell lines treated with the p38 MAPK inhibitor SB203580 [35]. However, there was little overlap between the SB203580-regulated genes in lymphocytes and our data set of p38 MAPK-regulated genes in cardiomyocytes, with the exception of genes involved in cell cycle regulation, such as cyclin D3 and cullin 3, which are present in both cases. We also found that many of the main components of the myocardium ECM were overexpressed in p38α^−/− cardiomyocytes. As ECM components play a key role in heart development and function [15], our results indicate a potential role for p38α in the regulation of postnatal heart physiology. Therefore we focused our studies on the mechanism of COL1 regulation, which represents approx. 85% of the protein content of the myocardium ECM.

We found that a significant number of the genes differentially expressed in p38α^−/− cells were potentially regulated by the transcription factors C/EBP and TEF-1. Although C/EBP was known to be regulated by the p38 and ERK MAPK pathways, heart-specific transcriptional activity of TEF-1 has not been shown to be regulated by MAPK pathways to date. The TEF-1 gene family consists of 4 transcriptional activators, which regulate promoters containing the M-CAT sites in several tissues [36]. In cardiac muscle, M-CAT sites are required for the full activity of promoters induced by hypertrophic signals [37]. C/EBPs include 6 transcription factors in the leucine zipper family, which play key regulatory roles in cell proliferation and differentiation and can be regulated by MAPK phosphorylation (reviewed in [38]). C/EBPβ proteins have been proposed to be involved in the down-regulation of COL1A1 gene expression [39], the promoter of which contains three C/EBP binding sites (Figure 7A) [40].

In some cell types, down-regulation of p38 MAPKs has been shown to correlate with up-regulation of the ERK pathway [16,32]. While this manuscript was in preparation, a study has been published reporting the regulation of collagen expression by p38α in primary rat cardiac fibroblasts subjected to mechanical load. Using the inhibitor SB203580, these authors concluded that p38 MAPK negatively regulates collagen transcription by acting through ERK [41]. We also found a higher level of ERK basal activity in p38α^−/− cardiomyocytes, however, incubation with ERK inhibitors did not change the level of COL1A1 mRNA expression in p38α^−/− cells. Moreover, TNFα treatment promotes the inhibition of COL1A1 expression in wt cardiomyocytes, but increased COL1A1 expression in p38α^−/− cells; similar results were observed upon expression of the p38α-D/A mutant in the p38α^−/− background. Taken together, these results suggest a negative role for p38 MAPK, independent of ERK activity, in the regulation of COL1A1 in normally growing and cytokine-stimulated cardiomyocytes.

Our studies indicate that gene transcription is the primary mechanism of COL1A1 expression regulated by p38α in cardiomyocytes. The expression of COL1 genes is controlled by multiple regulatory modules, which are differentially required in various cell types. Such organization allows the system to respond to many environmental stimuli [42]. Several p38 MAPK-regulated transcription factors, such as AP-1, NFκB (nuclear factor κB) and C/EBP, have been implicated in basal, as well as cytokine-modulated COL1 gene expression (reviewed in [30]). C/EBPβ is induced during myocardium injury and can bind to the COL1A1 promoter [43]. Depending on the cell type and experimental conditions, the transcriptional activity of C/EBP may be regulated in a positive or negative manner through phosphorylation by different signal transduction pathways including ERK [44] and p38 MAPKs [45].We observed that less C/EBPβ was loaded on to the COL1A1 promoter in p38α^−/− cells, indicating that it could mediate the inhibition of COL1A1 transcription induced by p38α in normally proliferating cardiomyocytes. Conversely, both TEF-1 DNA binding activity and its loading on to the COL1A1 promoter were increased in the p38α^−/− cells. Therefore the up-regulation of COL1A1 gene expression in p38α^−/− cells may be related to the binding of the transcriptional activator TEF-1 and the simultaneous release of the transcriptional repressor C/EBPβ. The COL1A1 promoter also contains Sp1 sites that have been suggested to mediate TEF-1 binding to promoters that lack an M-CAT binding site [46]. Moreover, we have noticed that the C/EBP binding sites on the murine COL1A1 promoter overlap with sites recognized by STAT (signal transducer and activator of transcription) transcription factors, and p38α^−/− cardiomyocytes are known to contain increased amounts of nuclear phosphorylated STAT3 [16]. Further experiments will be necessary to investigate the potential contribution of all of these proteins to the upregulation of COL1A1 gene expression in p38α^−/− cells.

The generation of knockout mice lacking p38α or the upstream activators MKK6 and MKK3 has allowed us to investigate the role of this pathway in mammalian embryogenesis. The absence of p38α expression or activation resulted in mouse embryonic lethality due to a defect in placental morphogenesis, probably due to defective vascularization [6,7,47,48]. Collagens are the main components of blood vessel ECM, as well as of the chorionallantoic ECM that provides a scaffold for placental vascular growth. Therefore our results regarding the regulation of COL1 and COL3 gene expression by p38 MAPKs suggest that the placental vascularization defect in p38α^−/− mice could be related to impaired collagen accumulation. This is consistent with the proposed role of p38α in the vascular remodelling associated with angiogenesis, rather than with the processes involved in the formation of the primary capillary plexus [7]. The study of COL1 and COL3 expression in the placentas of p38α^−/− embryos will be necessary to confirm this possibility. Previous work has suggested a role for p38 MAPKs in cardiac morphogenesis and pathological conditions such as hypertrophy and fibrosis (reviewed by [49]). However, defects in cardiac development were not detected when the placental phenotype of p38α^−/− embryos was rescued by fusion with tetraploid wt embryos [6]. In the present study, we have shown that p38 MAPK controls the expression of several myocardium ECM components in normally proliferating cells. Our results are in agreement with previous work showing that the expression of transgenes for dominant negative forms of p38α or their activators MKK6 and MKK3 in mouse heart results in interstitial fibrosis, which leads to the modification of myocardium integrity and function [13]. Interestingly, cardiac-specific p38α knockout mice have recently been reported to display normal global cardiac structure and function, but these mice develop an abnormal response to pressure overload, which is accompanied by massive cardiac fibrosis [50]. In agreement with our results, the expression levels of COL1A1 and COL3A1 mRNAs after pressure overload were also significantly increased in the hearts of cardiac-specific p38α knockout mice compared with the levels in wt control mice [50].

In summary, we propose that the basal level of p38 MAPK activity in non-stimulated cardiomyocytes may be implicated in the inhibition of pro-fibrotic and pro-hypertrophic pathways, which are activated upon heart injury and play a key role in the repair response (reviewed in [49]). Our results also indicate that C/EBPβ and TEF-1 transcription factors are important mediators of p38 MAPK function in cardiomyocytes.