Endothelin-1 promotes cardiomyocyte hypertrophy by inducing changes in gene expression. Immediate early genes including Atf3 (activating transcription factor 3), Egr1 (early growth response 1) and Ptgs2 (prostaglandin-endoperoxide synthase 2) are rapi-dly and transiently up-regulated by endothelin-1 in cardiomyocytes. Atf3 regulates the expression of downstream genes and is implicated in negative feedback regulation of other immediate early genes. To identify Atf3-regulated genes, we knocked down Atf3 expression in cardiomyocytes exposed to endothelin-1 and used microarrays to interrogate the transcriptomic effects. The expression of 23 mRNAs (including Egr1 and Ptgs2) was enhanced and the expression of 25 mRNAs was inhibited by Atf3 knockdown. Using quantitative PCR, we determined that knockdown of Atf3 had little effect on up-regulation of Egr1 mRNA over 30 min, but abolished the subsequent decline, causing sustained Egr1 mRNA expression and enhanced protein expression. This resulted from direct binding of Atf3 to the Egr1 promoter. Mathematical modelling established that Atf3 can suffice to suppress Egr1 expression. Given the widespread co-regulation of Atf3 with Egr1, we suggest that the Atf3–Egr1 negative feedback loop is of general significance. Loss of Atf3 caused abnormal cardiomyocyte growth, presumably resulting from the dysregulation of target genes. The results of the present study therefore identify Atf3 as a nexus in cardiomyocyte hypertrophy required to facilitate the full and proper growth response.
Cardiomyocytes, the terminally differentiated contractile cells of the heart, undergo hypertrophy (i.e. increase in cell size) in response to GqPCR (Gq-protein-coupled receptor) agonists including ET-1 (endothelin-1) . ET-1 particularly activates the ERK1/2 (extracellular-signal-regulated kinase 1/2) cascade, a pathway that is associated with cardiomyocyte hypertrophy [1,2]. ERK1/2 signalling is strongly implicated in transcriptional regulation and elicits many longer term effects through changes in gene expression . Perhaps unsurprisingly therefore, ~70% of the changes in gene expression induced in neonatal rat cardiomyocytes over the first 4 h of stimulation with ET-1 requires ERK1/2 signalling [4–6]. However, activation of ERK1/2 in cardiomyocytes by ET-1 is rapid and transient , and this phase may be viewed as an initiating trigger of the longer term response. Consistent with this, exposure of cardiomyocytes to ET-1 results in multiphasic patterns of gene expression that presumably culminate in hypertrophic growth . Our previous studies identified a number of IEGs (immediate early genes; genes that are regulated by pre-existing transcription factors) that are particularly rapidly up-regulated in cardiomyocytes exposed to ET-1 including Atf (activating transcription factor) 3 .
Atf3 is a member of the ATF/CREB [CRE (cAMP-response element)-binding protein] family proteins which bind as homo- or hetero-dimers to consensus ATF/CRE sites in gene promoters to regulate transcription . Like other members of this family, full-length Atf3 possesses a bZIP (basic leucine zipper) domain for dimerization, although alternative splicing can produce Atf3 isoforms without a bZIP domain. The activity of some ATF/CREB proteins is regulated by phosphorylation, but Atf3 is largely regulated at the level of expression, being present at very low levels in quiescent cells and induced as an IEG by a range of extracellular stimuli (e.g. peptide growth factors, GqPCR agonists and cytokines) and cellular stresses (e.g. oxidative stress and ischaemia/reperfusion) in a variety of cell types. Indeed, it is difficult to identify a stimulus that does not induce Atf3 expression, suggesting that it plays a significant global role in regulating gene expression programmes. MAPKs (mitogen-activated protein kinases) including ERK1/2 are particularly implicated in promoting Atf3 mRNA expression, potentially acting through a variety of transcription factors (e.g. CREB, Atf2 and c-Myc) [9–11]. Several studies also implicate the Egr1 (early growth response 1) transcription factor in positive regulation of Atf3 transcription, probably resulting from its phosphorylation and activation by ERK1/2 [10,12,13].
Atf3 is generally viewed as a transcriptional repressor, particularly when present as homodimers. Thus Atf3 represses transcription of Gadd153 (growth arrest and DNA damage-inducible protein 153)/Chop10 [C/EBP (CCAAT/enhancher-binding protein) homologous protein 10]/DDIT3 (DNA-damage-inducible transcript 3 protein) and may also repress transcription from its own promoter to limit expression . A systems biology study of TLR (Toll-like receptor) 4 signalling to gene expression in macrophages further confirmed its role as a transcriptional repressor, demonstrating that Atf3 is induced by lipopolysaccharide and represses expression of IL (interleukin) 6 and IL12b mRNAs by antagonizing NF-κB (nuclear factor κB)-dependent stimulation of transcription . A similar system operates downstream from TLR9 , suggesting that negative feedback regulation of cytokine production by Atf3 is a feature of the innate immune response . However, Atf3 (potentially as heterodimers with other ATF/CREB proteins) promotes transcription of other genes such as proglucagon . Moreover, in the context of DNA damage in cancer cells, Atf3 may act as a positive regulator of gene expression, probably by enhancing p53 function . The dichotomous role of Atf3 and the variation in underlying mutations that cause cancer presumably account for the variation in opinion regarding the role of Atf3 in this disease .
As mentioned above, Atf3 is induced in cardiomyocytes by ET-1 as an IEG , but it is also up-regulated in these cells by, for example, doxorubicin , oxidative stress , insulin  and hypoxia . Transient ischaemia increases Atf3 expression in whole hearts [22,23], and cardiospecific overexpression of Atf3 in transgenic mice results in pathological features of cardiac hypertrophy/failure  (this is distinguishable from, though may encompass, cardiomyocyte hypertrophy ). Confusingly, cardiac hypertrophy/failure induced by pressure overload is exaggerated in Atf3-null mice . This probably results from enhanced signalling through ERK1/2 and other MAPKs, all of which are implicated in the development of cardiac pathology . Although these studies suggest that control of Atf3 expression is important in homoeostatic control of cardiac function, it is difficult to develop a mechanistic understanding of Atf3 function in these models with long-term manipulation of Atf3 expression. We used the cardiomyocyte model to investigate the role of Atf3 in negative feedback regulation of IEG expression and positive feed-forward regulation of second-phase genes. Adopting an antisense knockdown approach, we identified Egr1 as a prime target for Atf3 repression in response to ET-1. Notably, Egr1 has been associated with cardiac hypertrophy for many years [27–29], and the results of the present study shed further light on the transcriptional networks within which it operates.
Cardiomyocyte cultures and knockdown of Atf3 with adenoviruses for antisense Atf3
Ventricles from neonatal Sprague–Dawley rat hearts (Harlan) were dissociated by serial digestion and cultured as described previously . For immunoblotting, RNA studies or ChIP (chromatin immunoprecipitation), cardiomyocytes were plated in 15% (v/v) fetal bovine serum at a density of 4×106 cells/dish on 60 mm Primaria dishes pre-coated with sterile 1% (w/v) gelatin (Sigma–Aldrich). For immunostaining experiments, glass coverslips were placed in Primaria 35 mm culture dishes and coated with 1% (v/v) gelatin then laminin (0.2 mg/ml in PBS, 2 h). Coverslips were washed with PBS. Cardiomyocytes were plated at 1.5×106 cells/dish in serum-containing medium. After 18 h, cardiomyocytes were incubated in serum-free medium for 24 h then unstimulated (controls) or exposed to 100 nM ET-1 (Bachem) in the presence or absence of cycloheximide (Sigma–Aldrich), U0126 or PD184352 (Enzo Life Sciences). Agonists/inhibitors were prepared as 1000× stock solutions in water (ET-1, cycloheximide) or DMSO (U0126, PD184352) and added directly to the tissue culture medium.
Preparation of AdVs (adenoviruses)
Replication-deficient AdVs expressing full-length rat Atf3 antisense RNA (AS-Atf3) were prepared using the AdEasy™ XL Adenoviral Vector System (Stratagene). The coding sequence for Atf3 was isolated by PCR from rat cardiomyocyte cDNA using Pfu polymerase and primers designed to the 5′ start site (5′-ATGATGCTTCAACATCCAGGC-3′) and 3′ stop codon (5′-TTAGCTCTGCAATGTTCCTTC-3′) regions, and further amplified with primers that included sites for restriction enzymes HindIII (5′-CTTATCTAGAAGCTTATGATGCTTCAACATCCAGGC-3′) and KpnI (5′-TAGAGATCTGGTACCTTAGCTCTGCAATGTTCCTTC-3′) for insertion into the multiple cloning site of the pShuttle-CMV vector. Control samples were prepared with empty vector (No-AS). Constructs were sequenced using an ABI 3100 Genetic Analyser. Shuttle plasmids were linearized with PmeI and used to transform BJ5183-AD-1 cells. AdV plasmids from positive recombinants were expanded in XL10Gold cells, linearized with PacI and used to transform HEK (human embryonic kidney)-293 cells. AdVs were amplified through subsequent re-infection of HEK-293 cells and titres were determined using Adeno-X™ Rapid Titer kits (Clontech). For experiments with AdVs, cardiomyocytes were incubated in serum-free medium for 4 h, AdVs were added and the cells incubated for a further 48 h.
Cardiomyocytes were washed in ice-cold PBS and nuclear extracts prepared for immunoblotting as described previously . Proteins (20 μg) were separated by SDS/PAGE (10% or 12% gels) and transferred electrophoretically on to nitrocellulose membranes. Immunoblotting was performed as described previously . Primary antibodies against Atf3 (sc-188), Egr1 (sc-189) and Atf2 (sc-187) were from Santa Cruz Biotechnology and used at a 1/1000 dilution. Secondary antibodies conjugated to horseradish peroxidase were from Dako (1/5000 dilution). Bands were detected using ECL (enhanced chemiluminescence) Plus with an ImageQuant 350 digital imager (GE Healthcare). ImageQuant 7.0 software was used for densitometric analysis.
RNA preparation, microarrays and data analysis
Cardiomyocytes were uninfected or infected with empty AdVs or AdVs encoding AS-Atf3 and then unstimulated (controls) or exposed to 100 nM ET-1 (90 min). Uninfected cardiomyocytes were exposed to ET-1 for 30, 60 or 90 min. Total RNA was prepared and concentrations determined as described previously . To minimize the variation resulting from different cardiomyocyte preparations, equal amounts of RNA from three individual experiments were pooled to generate a single sample set. A total of three sets of pooled samples were prepared for hybridization to separate Affymetrix rat exon 1.0 ST arrays (i.e. three separate sets of samples were analysed for each condition, prepared from a total of nine myocyte preparations). Samples were prepared for hybridization using Genechip® WT Sense Target Labelling kits (Affymetrix). For all samples, except those from uninfected cells exposed to ET-1 for 60 min with their corresponding controls, the rRNA reduction protocol was modified and the RiboMinus Transcriptome Isolation Kit (Human/Mouse) (Invitrogen) was used according to the manufacturer's instructions with 7 μg of RNA. Hybridization to Affymetrix rat exon 1.0 ST arrays was performed at the CSC/IC Microarray Centre (Imperial College London, London, U.K.). For samples from uninfected cells exposed to ET-1 for 60 min with their corresponding controls, total RNA was provided to the European Arabidopsis Stock Centre at Nottingham (NASC) for preparation and hybridization to Affymetrix rat exon 1.0 arrays according to their protocols (http://affymetrix.arabidopsis.info). The results are available from ArrayExpress (accession numbers E-MEXP-3392, E-MEXP-3393 and E-MEXP-3394).
The data (.CEL files) were imported into GeneSpring 11.5 (Agilent Technologies) using the Extended annotations, with RMA16 normalization and normalization per gene to corresponding uninfected unstimulated controls within each sample set. Probesets were selected for analysis with baseline expression >50 in all of any of the conditions. To determine effects of AdV infection, probesets were selected with >1.5-fold change in unstimulated cells infected with No-AS AdVs relative to uninfected cells, applying an unpaired Student's t test with the Benjamini and Hochberg multiple testing correction [FDR (false discovery rate)<0.05)]. To identify transcripts regulated by Atf3, we selected those which were significantly changed (>1.5-fold; FDR<0.05) in uninfected cells in response to ET-1 at 30, 60 or 90 min (unpaired Student's t tests with the Benjamini and Hochberg multiple testing correction). For transcripts with <1.2-fold change in baseline expression with No-AS AdVs, subsequent analysis employed an interpretation with AdV-infected conditions, using statistical testing [one-way ANOVA with SNK (Student-Newman-Keuls) post-test and Benjamini and Hochberg multiple testing correction] and filtering on the basis of >1.5-fold difference between AS-Atf3 and No-AS infected cells. For transcripts with >1.2-fold change in baseline expression with No-AS AdVs, subsequent analysis employed an interpretation with all conditions, using statistical testing (one-way ANOVA with SNK post-test and the Benjamini and Hochberg multiple testing correction) to identify mRNAs with significant difference between cells infected with AS-Atf3 and No-AS AdVs or uninfected cells, but with no significant difference between uninfected cells and those infected with No-AS AdVs with exposure to ET-1 (90 min) and/or in unstimulated cells. Further filtering identified transcripts with >1.5-fold difference between cells infected with AS-Atf3 and those infected with No-AS AdVs or uninfected cells.
Cardiomyocytes were treated and total RNA extracted as for microarray analysis. cDNAs were synthesized using High Capacity cDNA Reverse Transcription Kits with random primers (Applied Biosystems). sqPCR [semi-qPCR (quantitative PCR)] and qPCR were performed as described previously  using primers listed in Supplementary Table S2 (at http://www.BiochemJ.org/bj/444/bj4440343add.htm). Values were normalized to Gapdh (glyceraldehyde-3-phosphate dehydrogenase) expression and then to control values.
Cardiomyocytes (16×106 cells per sample) were unstimulated or exposed to 100 nM ET-1 and fixed in 1% (v/v) formaldehyde (10 min). The reaction was terminated with 125 mM glycine (10 min). Cells were washed and harvested in ice-cold PBS containing protease/phosphatase inhibitors , collected by centrifugation (3000 g at 4°C for 5 min) and lysed (15 min at 4°C) in 50 mM Tris/HCl (pH 8.0), 2 mM EDTA, 0.1% Nonidet P40 and 10% (v/v) glycerol containing protease/phosphatase inhibitors. Nuclei were pelleted (3000 g at 4°C for 5 min), resuspended in 50 mM Tris/HCl (pH 8.0), 1% (w/v) SDS and 5 mM EDTA, and the DNA sheared to 200–800 bp fragments by sonication (5×30 s with 2 min recovery at 4°C). Following centrifugation (4000 g at 4°C for 5 min), the supernatants were retained and a sample reserved for input DNA. The remainder was diluted in 50 mM Tris/HCl (pH 8.0), 0.5% Nonidet P40, 200 mM NaCl and 0.5 mM EDTA. Samples were pre-cleared with Protein A–Sepharose then incubated without or with 0.01 mg of anti-Atf3 antibodies (Santa Cruz Biotechnology, catalogue number sc-188X) with rotation (overnight at 4°C). Protein A–Sepharose was added together with sonicated salmon sperm DNA (1 μg/ml) and samples were incubated (4°C for 2 h). Beads were pelleted by centrifugation (1000 g at 4°C for 3 min) and washed (4°C for 3 min) in 20 mM Tris/HCl (pH 8.0), 0.1% SDS, 1% (v/v) Nonidet P40, 2 mM EDTA and 500 mM NaCl followed by 10 mM Tris/HCl (pH 8.0) containing 1 mM EDTA. Immune complexes were eluted (10 min at 65°C) in 10 mM Tris/HCl (pH 8.0), 1 mM EDTA and 1% (w/v) SDS. Samples were centrifuged (200 g for 1 min) and the supernatants collected. Cross-links were reversed by incubation (overnight at 65°C) with 0.2 M NaCl. Samples were incubated (5 min at 4°C) with phenol/chloroform/3-methylbutan-1-ol (25:24:1, pH 8.0) and the phases separated by centrifugation (15300 g at 4°C for 10 min). DNA in the upper aqueous phase was precipitated with isopropanol (−80°C for 3 h), recovered by centrifugation (15300 g for 10 min at 4°C), washed [70% (v/v) ethanol] and resuspended in nuclease-free water for PCR. PCR reactions were performed using specific primers: Egr1 forward, 5′-ACTGCCGCTGTTCCAATACT-3′; Egr1 reverse, 5′-CGAATCGGCCTCTATTTCAA-3′; Ptgs2 forward, 5′-GCAGCAAGCACGTCAGACT-3′; Ptgs2 reverse, 5′-TAACCCGGAGAACCTTGCTT-3′; IL6 forward, 5′-TGCTCAAGTGCTGAGTCACT-3′; IL6 reverse, 5′-AGACTCATGGGAAAATCCCA-3′. The primers for an arbitrary downstream sequence ~2000 bp 3′ to the IL6 promoter were: forward, 5′-CACCTCTCCACCCTGACATT-3′ and reverse, 5′-CCAACTAGACAGCCCAGAGC-3′. PCR amplification conditions were: 34 cycles of denaturing at 95°C for 30 s, annealing at 60°C for 30 s and extension at 72°C for 45 s. PCR products were visualized on 2% (w/v) agarose gels with Sybr-Safe (Invitrogen) and the bands captured under UV illumination. Densitometric analysis was performed using Imagemaster 1D Prime, version 3.0 (GE Healthcare).
Cardiomyocytes were exposed to 100 nM ET-1, then washed with PBS and fixed in 4% (v/v) formaldehyde [10 min at room temperature (25°C)]. Immunostaining was performed as described previously  using mouse monoclonal primary antibodies against troponin T (1/40 dilution for 60 min; Stratech Scientific, catalogue number MS-295-P1) and Alexa Fluor® 488 anti-mouse secondary antibodies (1/200 dilution for 60 min). Coverslips were mounted using fluorescent mounting medium (Dako) and viewed using a Zeiss Axioskop fluorescence microscope. Images were captured using a digital camera (1600×1200 pixels resolution, ×1.4 zoom factor and 160×120 μm field dimension). Colour images were converted to greyscale using Adobe Photoshop.
Statistical analysis of microarray data used GeneSpring 11.5. Other analyses used GraphPad Prism 4.0.
Regulation of Atf3 expression in cardiomyocytes by ET-1 and identification of Atf3 target genes
ET-1 (100 nM) stimulated a rapid transient increase in expression of Atf3 mRNA in cardiomyocytes (Figure 1A). Expression was maximal within 30 min, declining rapidly thereafter. Cycloheximide (a protein synthesis inhibitor) did not inhibit the response, indicating that Atf3 was regulated as an IEG (Figure 1B), and up-regulation of Atf3 mRNA by ET-1 was inhibited by U0126 or PD184352, inhibitors of the ERK1/2 cascade (Figures 1B and 1C). Atf3 protein was also increased by ET-1 (maximal at 1 h) (Figure 1D). To suppress expression of Atf3, we generated AdVs encoding full-length Atf3 in the antisense orientation (AS-Atf3), using empty AdVs with no AS-Atf3 (No-AS) to control for effects of viral infection. Effective knockdown of Atf3 protein induced by ET-1 was obtained using 150 MOI (multiplicity of infection) AS-Atf3 AdVs with limited effect of No-AS AdVs and no effect on expression of Atf2 (Figures 1E–1G).
Atf3 mRNA and protein is increased in cardiomyocytes exposed to ET-1 and suppressed by AS-Atf3
Uninfected cardiomyocytes or cardiomyocytes infected with 150 MOI No-AS or AS-Atf3 AdVs were unstimulated or exposed to 100 nM ET-1 for 90 min (since expression of Atf3 protein was maximal at 1 h, we expected to detect downstream consequences by 90 min; Figure 1D) and mRNA expression profiling was performed (Affymetrix rat exon 1.0 ST microarrays). As expected, No-AS AdV infection alone significantly (>1.5-fold; FDR<0.05) affected expression of a subset of mRNAs (156 mRNAs up-regulated and 59 mRNAs down-regulated), many of which are associated with a viral response and/or cytokine/chemokine signalling in other cells (Figure 2 and Supplementary Table S3 at http://www.BiochemJ.org/bj/444/bj4440343add.htm).
Infection with AdVs modulates expression of a subset of cardiomyocyte mRNAs
To determine the effects of Atf3 in the ET-1-responsive myocyte transcriptome, probesets were selected with significant changes in expression (>1.5-fold, FDR<0.05) induced by ET-1 at 30, 60 or 90 min (334 up-regulated and 109 down-regulated mRNAs). From these, we identified 90 mRNAs that were up-regulated by ET-1 and affected by Atf3 knockdown (Figure 3A and Supplementary Tables S4–S6 at http://www.BiochemJ.org/bj/444/bj4440343add.htm). Of the up-regulated transcripts without a baseline effect of viral infection (75 mRNAs), AS-Atf3 enhanced the response of 15 mRNAs (clusters A1 and A2), inhibited the response of 18 mRNAs (clusters B1 and B2) and, for 42 mRNAs (cluster C), AS-Atf3 enhanced the response in unstimulated cells without further effect on the response to ET-1. Of the up-regulated mRNAs with a baseline effect (>1.2-fold change) of viral infection, the response to ET-1 was enhanced or inhibited by AS-Atf3 for eight (cluster D) or seven (cluster E) mRNAs respectively. With ET-1-down-regulated mRNAs (Figure 3B and Supplementary Table S7 at http://www.BiochemJ.org/bj/444/bj4440343add.htm), we identified only five mRNAs for which AS-Atf3 reduced the magnitude of the down-regulation by ET-1 (cluster F) and five mRNAs for which AS-Atf3 increased the magnitude of the response (cluster G). Having previously defined IEGs compared with downstream gene expression for mRNAs up-regulated by ET-1 (Affymetrix rat genome 230 2.0 arrays) , we cross-referenced the datasets (Supplementary Tables S4 and S6). A greater proportion of mRNAs in clusters A1, A2 and D (Atf3-knockdown-enhanced ET-1-induced expression) were IEGs rather than second-phase genes, consistent with negative feedback by Atf3. A greater proportion of mRNAs in clusters B1 and B2 (Atf3-knockdown-inhibited ET-1-induced expression) were second-phase genes rather than IEGs, suggesting that Atf3 plays a positive feed-forward role in these cases (genes in cluster E were not identified in the previous dataset ). Overall, we conclude that up-regulation of Atf3 by ET-1 in cardiomyocytes plays a significant role in modulating the response of other up-regulated genes and a minor role in regulating mRNAs that are down-regulated by ET-1.
Atf3 knockdown modulates the response of the cardiomyocyte transcriptome to ET-1
Negative regulation of Egr1 and Ptgs2 expression by Atf3
Microarray data were validated by qPCR, selecting IEGs with enhanced expression of the ET-1 response by AS-Atf3 (Egr1 and Ptgs2) and non-IEGs with reduced expression by AS-Atf3 [Areg (amphiregulin) and Dusp (dual specificity phosphatase) 5] for analysis. For Egr1 and Ptgs2, the relative increase in expression induced by ET-1 at 0.5 h was similar in cardiomyocytes expressing No-AS or AS-Atf3 (Figures 4A and 4B), indicating that Atf3 was not involved in this up-regulation phase. AS-Atf3 produced sustained expression of Egr1 mRNA (Figure 4A) and enhanced expression of Ptgs2 at subsequent times (Figure 4B), consistent with negative feedback of Atf3 on these two genes. In contrast, AS-Atf3 inhibited expression of Areg (Figure 4C) and Dusp5 (Figure 4D). The qPCR data (Figure 4, left-hand panels) are consistent with the microarray results (Figure 4, right-hand panels). Dusp1 (an IEG) and Il1rl1 (IL1 receptor-like 1; a non-IEG) were not significantly affected by AS-Atf3 in our microarray experiments and there was no difference in the relative changes in expression induced by ET-1 in cardiomyocytes expressing No-AS or AS-Atf3 (Figures 4E and 4F). Thus our selection criteria were not excessively stringent. Finally, although others have identified IL6 as a target for negative feedback by Atf3 in the context of inflammation , there was no difference in relative expression of IL6 mRNA in cardiomyocytes expressing No-AS or AS-Atf3 following stimulation with ET-1 at any time studied (Figure 4G). To validate the Atf3–Egr1 feedback loop further, we examined the effect of AS-Atf3 on Egr1 protein expression following induction with ET-1. AS-Atf3 (but not No-AS) inhibited expression of Atf3 for the 150 min period studied (Figures 5A and 5B). AS-Atf3 did not substantially increase the total amount of Egr1 protein at 60 min, but levels remained elevated up to ~105 min (Figures 5A and 5C). After this time Egr1 protein declined, but levels remained elevated up to at least 150 min.
qPCR validation of microarray data
AS-Atf3 expression enhances expression of Egr1 protein in cardiomyocytes exposed to ET-1
ChIP with antibodies against Atf3 was used to determine if Atf3 binds directly to the Egr1 and Ptgs2 promoters. The Egr1 promoter contains two CRE elements , and a CRE element potentially subject to negative regulation by Atf3 has been identified in the Ptgs2 promoter [33,34]. We used primers either side of these sequences for ChIP–PCR (Figures 6A and 6C). For both genes, the amount of ChIP–PCR product in unstimulated cells was similar to that of the no-antibody controls and stimulation with ET-1 (1 h) substantially increased the amount of ChIP–PCR product (Figures 6B and 6D). Thus Atf3 binds directly to Egr1 and Ptgs2 promoters to inhibit transcription. Interestingly, ChIP–PCR for the IL6 promoter (Figure 6E) demonstrated binding of Atf3 (Figure 6F) despite there being no effect of Atf3 knockdown on expression of IL6 mRNA (Figure 4G). This is discussed below, but potentially results from ET-1 signalling through ERK1/2 rather than NF-κB. The specificity of the ChIP–PCR experiments was demonstrated by PCR across an arbitrary sequence approximately 2000 bp downstream of the IL6 promoter (Figure 6G).
ChIP assays for binding of Atf3 to Egr1, Ptgs2 and IL6 promoters
Mathematical modelling of the Atf3–Egr1 negative feedback loop
Atf3 appears to be a dominant-negative regulatory factor in suppressing Egr1 transcription (Figure 4A), but we cannot eliminate the possibility of additional negative regulatory elements from the experimental data. We therefore developed a deterministic ordinary differential equation model for the Atf3–Egr1 feedback system (Figure 7A) to determine whether Atf3 expression can suffice to suppress Egr1 transcription. We assumed the following: Atf3 and Egr1 are co-regulated through ERK1/2 signalling (expression of each requires a similar degree of ERK1/2 signalling ); ERK1/2 promote expression by increasing transactivating activities of transcription factors pre-bound to both promoters (i.e. the transcription factor/promoter is viewed as a single entity regulated by ERK1/2 binding/phosphorylation); and it was not necessary to consider ERK1/2 or Atf3 signal termination (i.e. no attempt was made to switch off the signal). The model was informed initially by parameters derived in the present study and from previously published literature (see the Supplementary Online Data at http://www.BiochemJ.org/bj/444/bj4440343add.htm).
Mathematical modelling of the Atf3–Egr1 feedback system in the cardiomyocyte response to ET-1
We initially assumed that Atf3 suppression of Egr1 mRNA transcription was competitive with the positive ERK1/2 signal, but this did not give a good fit with the experimental data. An alternative model in which Atf3 eliminates the ERK1/2 signal (modelled by removing the ERK1/2 signal upon Atf3 binding; Figure 7A) gave a good qualitative fit. Adjusting the rate of mRNA synthesis to 275 bases/s (within the recently estimated range of 55 to >800 bases/s [35,36]) for calculation of the rates of transcription of Egr1 and Atf3 mRNAs, and adjusting parameters for association of phosphorylated ERK1/2 with Egr1 and Atf3 DNA (see the Supplementary Online Data) gave a very good fit to the experimental data in terms of magnitude of variation in Egr1 mRNA (~20-fold; Figure 4A), and a relatively good fit (qualitatively) to the suppression of Egr1 mRNA by Atf3. The best-fit model was obtained with a rate of Atf3 suppression of Egr1 transcription (k7) in the range 1×105(Ms)−1≤ k7≤6×105(Ms)−1 (Figure 7B). Modelling the system in the absence of Atf3 (to mimic Atf3 knockdown) resulted in a similar rate of production of Egr1 mRNA to ~0.5 h, with sustained expression thereafter (Figure 7B, centre left-hand panel), a profile that replicates the experimental data (Figure 4A, left-hand panel). Thus Atf3 alone can account for the suppression of Egr1 mRNA expression and probably does so by over-riding the positive ERK1/2 signal.
Effects of Atf3 on morphological changes induced in cardiomyocytes by ET-1
To determine the consequences of knockdown of Atf3 for ET-1-induced hypertrophy, we examined the effects of AS-Atf3 AdVs on myocyte morphology by immunostaining for troponin T, a component of the myofibrillar apparatus. As shown previously [37,38], uninfected and unstimulated cardiomyocytes appear small with disorganized myofilaments (Figure 8A). Stimulation with ET-1 increased cell size, increased the content/organization of myofilaments and promoted formation of cell–cell contacts (Figure 8B), classic features of hypertrophy. No-AS AdVs increased myofibrillar content to a small degree in unstimulated cells (Figure 8C), but had no overt effect on morphological changes induced by ET-1 (Figure 8D). Relative to No-AS AdVs, AS-Atf3 AdVs had no effect on unstimulated cells, but resulted in abnormal morphology following ET-1 stimulation with ‘clumping’ of cells and generation of elongated processes extending between them (Figures 8E–8H). It is unclear if this is an effect on hypertrophy itself or whether there is a loss of cell–cell or cell–matrix adhesion. Thus, although Atf3 is transiently up-regulated within the first hour of stimulation, it has a significant effect on long-term phenotypic changes.
Atf3 is required for cardiomyocyte hypertrophy induced by ET-1
Atf3 is emerging as an extremely important feedback regulator of transcription in general. Greatest emphasis has been placed on its role in inflammation and it is clear that Atf3 is essential for restraining the immune response [15,17]. However, although viewed as a stress-regulated and -regulatory gene, Atf3 is also up-regulated in many cells by growth stimuli including peptide growth factors and GqPCR agonists [5,39–41]. The genes that Atf3 regulates in a growth context are likely to differ from those that it regulates in a stress response. In the present study, we identified a number of genes that are regulated in cardiomyocytes in response to ET-1 for which Atf3 plays either a negative feedback or positive feed-forward role (Figure 3). Since loss of Atf3 results in a severely abnormal response to ET-1 (Figure 8), strict regulation of any or all of these targets must be important in ensuring an appropriate hypertrophic growth response. However, of all the potential targets, Egr1 emerged as a significant candidate and we provide substantial evidence that Atf3 is a transcriptional repressor for Egr1 in the ET-1 response. Thus Atf3 knockdown leads to sustained expression of Egr1 mRNA and enhanced expression of Egr1 protein, and Atf3 binds directly to the Egr1 promoter. Furthermore, mathematical modelling of the system established that Atf3 alone can suffice to suppress Egr1 mRNA expression.
Egr1 was among the first IEGs to be identified  and it is up-regulated in many cells in response to a wide range of growth stimuli . As such, it is implicated in a host of different systems including kidney differentiation , macrophage differentiation , the nervous system and learning/memory [46,47], the immune response , wound healing and tissue repair , prostate cancer  and acute lung injury . In the vascular system, Egr1 is associated with excessive smooth muscle cell proliferation resulting in in-stent restenosis following angioplasty . Egr1 is also implicated in cardiac/cardiomyocyte hypertrophy. Thus expression of Egr1 is increased in cardiomyocytes and the heart in response to a variety of hypertrophic stimuli [27,28], and loss of Egr1 attenuates the hypertrophic response . The importance of Egr1 in such a range of systems, together with the emergence of Atf3 as a negative feedback regulator and the co-regulation of the two genes in many situations, raises the possibility that the Atf3–Egr1 negative feedback loop is of broad general significance required to restrict expression of Egr1 and maintain a normal growth response.
Although Atf3 and Egr1 are co-regulated by a range of stimuli, their up-regulation is usually rapid and transient. As transcription factors, IEGs such as Atf3 and Egr1 are important in regulating downstream gene expression, and dysregulation leads to functional abnormalities and abnormal phenotypic responses. As demonstrated in the present study (Figure 8), cardiomyocyte hypertrophy cannot proceed normally in the absence of Atf3. This is not inconsistent with induction of cardiac hypertrophy in mice in vivo by overexpression of Atf3 . Atf3 knockout also enhances cardiac hypertrophy induced by pressure overload in mice, but, in this model, MAPK signalling (including ERK1/2) induced by pressure overload is significantly enhanced  and this alone could account for enhanced hypertrophy [2,53]. Intriguing questions relate to how Atf3 modulates MAPK signalling and if the up-regulation of MAPK signalling is a compensatory mechanism for loss of Atf3. Other studies indicate that Atf3 is important in regulating, for example, the immune response [15,17], liver function  and pancreatic β-cell function [55,56]. Because of this, it is important to identify the genes that Atf3 regulates.
At least two previous studies took a systems approach to identify Atf3-regulated genes, neither of which identified Egr1 as a target. The first focused on TLR4 signalling and the effects of lipopolysaccharide on the macrophage transcriptome . The strategy used Cytoscape (software that draws on published literature to identify protein–protein interactions) to predict Atf3-regulated genes from microarray data for subsequent experimental verification. Atf3 was particularly implicated in negative regulation of IL6. IL6 is up-regulated in cardiomyocytes by ET-1 as we have reported previously  and in the present study (Figure 4G). We have also previously reported preliminary data on the effects of AS-Atf3 AdV infection on IL6 mRNA expression. In those experiments, low viral titres and a low MOI (<15) were used to suppress the increase in expression of Atf3 by 10 nM ET-1 (rather than 100 nM ET-1 used in the present study). As in the present study, AS-Atf3 AdVs enhanced the increase in IL6 mRNA induced by ET-1 . Our interpretation was that Atf3 suppressed IL6 transcription, consistent with the report on the role of Atf3 in the TLR4 response . However, we know now that AdV infection has substantial effects on elements of the cardiomyocyte transcriptome and on IL6 in particular (Figure 2), and the data in the earlier report require re-interpretation. With submaximal virus and agonist concentrations (rather than the saturation levels used in the present study), slight differences between No-AS and AS-Atf3 viral titres were likely to be exaggerated. The No-AS AdV titre may have been slightly less than that of the AS-Atf3 AdVs and this could account for the significant effect of AS-Atf3 on IL6 mRNA expression in control cells over and above that induced by the empty virus. We assumed that this indicated a basal effect of Atf3 on IL6 mRNA expression. However, if the effect of AS-Atf3 in control cells is taken into account, ET-1 stimulation of IL6 mRNA was similar (~4-fold) in cardiomyocytes infected with either No-AS or AS-Atf3 AdVs. We now therefore consider that Atf3 knockdown did not influence the increase in expression of IL6 mRNA induced by ET-1. The lack of an effect of Atf3 knockdown on the increase in IL6 mRNA expression induced by ET-1 in contrast with the TLR4 response probably reflects a fundamental difference in the signalling mechanisms. TLR4 signals predominantly through NF-κB and suppression of IL6 expression by Atf3 results from its interaction with and inhibition of NF-κB . ET-1 signals predominantly through ERK1/2 to up-regulate expression of IL6 . With little/no signal through NF-κB, Atf3 is unlikely to inhibit transcription of IL6. However, Atf3 still binds to the IL6 promoter (Figure 6G). It is less clear why Egr1 was not identified as an Atf3-regulated gene in the TLR4 study . One reason may be that Cytoscape is informed only by published data. Hence, the study focused on the Atf3–NFκB interaction and all validations centred on the NF-κB response.
The second study identifying Atf3-regulated genes focused on DNA damage in cancer cells. In that study, 5984 potential Atf3-binding promoters were identified in HCT116 colon cancer cells subjected to DNA damage and 1493 were identified in LNCaP prostate cancer cells exhibiting enhanced Atf3 expression . In contrast with TLR4 signalling in macrophages , Atf3 plays a greater role in positive regulation of mRNA expression in these cancer cells, and there was no obvious effect of Atf3 knockdown on Egr1 expression. However, this paradigm differs radically from a growth response associated with IEG expression: Atf3 up-regulation is delayed (mRNA and protein expression are maximal at 6 and 12 h respectively) and potentially driven by p53  with many of the Atf3-regulated genes being classic p53 targets . It should also be considered that cancer cell lines were used rather than primary cells, and this may influence the signalling/gene expression response.
Since Atf3 and Egr1 are co-regulated in many systems, it is worth considering the extent to which the Atf3–Egr1 feedback mechanism operates. We expect it will apply in cardiomyocytes exposed to other stimuli that regulate gene expression primarily through MAPK signalling. We suggest that the Atf3–Egr1 negative feedback loop operates in other systems that also drive gene expression primarily through ERK1/2 signalling, but this awaits further investigation.
activating transcription factor
basic leucine zipper
dual specificity phosphatase
early growth response 1
extracellular-signal-regulated kinase 1/2
false discovery rate
human embryonic kidney
immediate early gene
IL1 receptor-like 1
mitogen-activated protein kinase
multiplicity of infection
nuclear factor κB
The present study was conceived and directed by Angela Clerk. Alejandro Giraldo and Oliver Barrett conducted the majority of the experimental work with equal contribution. Stephen Fuller constructed the vectors, produced the AdVs for the present study and, with Emre Amirak, assisted with parameterization of the model. Emre Amirak contributed microarray data and performed the qPCR experiments. The mathematical modelling was performed by Marcus Tindall with assistance from Bonhi Bhattacharya. Overall supervision of the present study was undertaken by Angela Clerk and Peter Sugden.
We thank Ross Bullivant and Joel Pearson for assistance in the initial phase of the mathematical modelling.
The work was funded by the British Heart Foundation [grant number PG/07/074/23445] and Fondation Leducq (Translatlantic Network of Excellence) [grant number CV05-02].
These authors contributed equally to this work.