The early growth response 1 (Egr1) gene is a transcription factor that acts as both a tumor suppressor and a tumor promoter. Egr1-null mouse embryo fibroblasts bypass replicative senescence and exhibit a loss of DNA damage response and an apparent immortal growth, suggesting loss of p53 functions. Stringent expression analysis revealed 266 transcripts with >2-fold differential expression in Egr1-null mouse embryo fibroblasts, including 143 known genes. Of the 143 genes, program-assisted searching revealed 66 informative genes linked to Egr1. All 66 genes could be placed on a single regulatory network consisting of three branch points of known Egr1 target genes: TGFβ1, IL6, and IGFI. Moreover, 19 additional genes that are known targets of p53 were identified, indicating that p53 is a fourth branch point. Electrophoretic mobility shift assay as well as chromatin immunoprecipitation confirmed that p53 is a direct target of Egr1. Because deficient p53 expression causes tumors in mice, we tested the role of Egr1 in a two-step skin carcinogenesis study (144 mice) that revealed a uniformly accelerated development of skin tumors in Egr1-null mice (P < 0.005). These studies reveal a new role for Egr1 as an in vivo tumor suppressor.
- path finding
- Egr1-regulatory pathway
- two-step carcinogenesis
The transcription factor early growth response 1 (Egr1) is a member of the immediate-early gene family and encodes a 59-kDa phosphoprotein observed at ∼80 kDa by electrophoresis. Egr1 is involved in the regulation of cell growth and differentiation in response to signals, such as mitogens, growth factors, and stress stimuli ( 1– 4). Egr1 has a COOH-terminal DNA-binding domain consisting of three zinc fingers that regulates transcription through GC-rich elements. The 9-bp DNA consensus-binding sequence GCG(G/T)GGGCG has been identified in the promoter of several growth-regulatory genes ( 5). Besides the highly conserved DNA-binding domain, Egr1 also contains a nuclear localization signal, two activator domains, and one repressor domain ( 6). In addition, Egr1 binds to regulatory proteins called NAB1 and NAB2 (nerve growth factor-I A-binding protein) that repress its transcriptional activity ( 7).
Analysis of certain human tumor cells and tissues indicate that Egr1 exhibits prominent tumor suppressor function ( 5, 8– 11). Many human tumor cell lines express little or no Egr1 in contrast to their normal counterparts ( 9– 12). Furthermore, Egr1 is decreased or undetectable in small cell lung tumors, human breast tumors ( 11, 13), and human gliomas ( 12). Reexpression of Egr1 in human tumor cells inhibits transformation. The mechanism of suppression involves the direct induction of TGF-β1 leading to an autocrine-mediated suppression of transformation ( 8), increased fibronectin, and plasminogen activator inhibitor ( 9). Egr1 also has been implicated in the regulation of p53 in human melanoma cells leading to apoptosis ( 14– 16), and the proapoptotic tumor suppressor gene PTEN also is directly regulated by Egr1 ( 17).
Recently, we have identified and established a new role of Egr1 as a “gatekeeper” of p53-dependent growth-regulatory mechanisms in replicative senescence and cell growth ( 18). This result was revealed by examination of primary mouse embryo fibroblasts (MEF) isolated from Egr1-null mice developed previously by Lee et al. ( 19) and Topilko et al. ( 20). Egr1-null cells from either strain express no Egr1 protein and much reduced p53 protein. These cells completely bypass replicative senescence in culture, thereby appearing to be immortal without passing through a “crisis” stage. Moreover, these cells fail to arrest following treatment with DNA-damaging agents, such as γ-irradiation ( 18) and 12-O-tetradecanoylphorbol-13-acetate (TPA; see below). Replicative senescence and the DNA damage response of the Egr1-null cells may be restored by infection with an Egr1-expressing retrovirus in contrast to p53-null MEFs. In the Egr1-null cells, p53 is not expressed and the p53 gene remains entirely free of mutation to high passage numbers. In contrast, the wild-type (WT) cells express normal levels of p53, arrest at low passage in normal tissue culture, and undergo crisis. WT cells that survive crisis inevitably exhibit mutations of the p53 gene. We speculated that Egr1 exerts a gatekeeper function over p53, thereby allowing the null cells to bypass replicative senescence and avoid the mutation association with the survival of crisis. However, the exact mechanism whereby Egr1 functions in the regulation of p53-dependent growth arrest is unknown, and there is no evidence that Egr1 mediates these roles in a whole organism.
Here, we examined gene expression in the genetically defined Egr1-deficient MEFs. These cells exhibit at least 266 significant differences in transcript expression compared with WT cells, many of which could be recognized as Egr1 target genes or downstream signal transduction mediators. Indeed, an extended regulatory network based on four “nodes” could be constructed. Three nodes are upstream of numerous recognized Egr1-regulated genes. However, a fourth node, and potential Egr1 target gene, p53, is upstream of at least 19 genes not previously recognized as Egr1-regulated genes but are well-recognized mediators of p53 function. Electrophoretic mobility shift assay (EMSA) and chromatin immunoprecipitation (ChIP) experiments directly confirmed the binding of the p53 promoter by Egr1 in cells both in vitro and in vivo. Moreover, TPA-induced skin tumors appeared rapidly in Egr1-null mice compared with WT and heterozygous littermates. These studies indicate that Egr1 regulates an extensive network of genes that include direct regulation of p53 at the level of RNA and protein expression. Our observations indicate that this mechanism may function in vitro and in vivo to control DNA damage–induced growth arrest. Loss of this mechanism in vivo may be related to increased skin tumor formation in Egr1-deficient mice.
Materials and Methods
Cells and cell culture. MEFs were prepared as described earlier ( 21) from 19-day-old embryos from Egr1 WT, Egr1-null, and Egr1 heterozygous mice kindly provided by Dr. J. Milbrandt (Washington University Medical School, St. Louis, MO) ( 19). Cultures of WT and Egr1-null MEFs were established and maintained in parallel for >60 passages. The predicted genotype and expression properties of the MEFs derived from Egr1-null and heterozygous mice were confirmed by PCR-based analysis of DNA and RNA and by Western analysis for protein expression as described previously ( 18).
Oligonucleotide microarray analysis. Total cellular RNA was isolated by using RNeasy kits (Qiagen Inc., Valencia, CA) and quantified. The total RNA quality was determined using the standard Affymetrix protocol. For hybridization, total RNA (10 μg) from WT or Egr1-null MEFs was reverse transcribed using an oligo(dT) primer harboring a T7 RNA polymerase promoter at the 5′ end (Genset, San Diego, CA). Following second-strand synthesis, biotinylated cRNA probes were produced. The probes were then fragmented and hybridized to Affymetrix MGU75Av2 array representing 12,566 mouse transcripts using the standard Affymetrix protocol (http://www.affymetrix.com). Quantitative analysis of the data from the microarrays was carried out using the MAS1 suite (http://www.affymetrix.com).
Two replicated RNA preparations from separate cultures of WT MEFs and two of Egr1-null MEFs were hybridized to four Affymetrix MGU75Av2 arrays leading to four different comparisons of data sets. In general, a gene was considered to be differentially expressed in Egr1-null MEFs when three criteria were met: (a) when “called present” (MAS1 suite), (b) when exhibiting a ≥2-fold change in net fluorescence relative to WT MEFs, and (c) when these criteria were satisfied by all four possible comparisons of the replicate WT and Egr1-null MEF data sets. The >2-fold change criterion was chosen in correspondence to the recommended cutoff for significant change by the manufacturer ( 22) and because it has been shown that difference of >2-fold change is detected >98% of the time by the Affymetrix assay ( 23).
Reconstruction of the early growth response 1 network. The significant differentially expressed genes from the expression analysis were selected for analysis. An algorithm was developed to convert the Affymetrix probe IDs to RefSeq IDs (NM numbers) using the LocusLink database. Of the 266 differentially expressed probes, 143 were identified as annotated genes in the LocusLink database. These known genes were then further analyzed for their potential connection to Egr1 or Egr1-regulated genes by examination of the literature using BiblioSphere (Genomatix Software GmbH, Munich, Germany; http://www.genomatix.de) for uncovering reported regulatory relationships between genes. BiblioSphere is a data mining tool intended to provide gene relationships from literature databases and genome-wide promoter analysis. Over 12 million Pubmed abstracts were scanned by this strategy using >500,000 gene names, synonyms, and the Genomatix proprietary semantic relation concepts. Relationships defined with this tool were confirmed by manual examination in Pubmed, and the confirmed related genes were entered into the Genomatix package to determine secondary relationships. Iteration led to a network relating all informative genes. An algorithm was developed to identify the human homologues for these selected genes from the LocusLink.
Quantitative real-time PCR. RNA expression levels of selected genes from the microarray results were quantified by quantitative real-time PCR (Q-PCR; Applied Biosystems ABI7900, Foster City, CA) as described previously ( 18). Genes validated by PCR were Bax, Fas, GADD45, p53, p21, TGFβ1, TGFβ1I, GRO1, IGFII, THSP2, and procollagen XI. Primers were selected for these genes with the aid of Primer Express software (Applied Biosystems) and are available on request. The results were normalized to the relative amounts of glyceraldehyde-3-phosphate dehydrogenase (GAPDH).
Chromatin immunoprecipitation assay. WT and null MEFs and human prostate cancer DU145 cells were treated with 1% formaldehyde for 20 minutes to cross-link protein to binding sites on DNA. Egr1-containing fragments were recovered by immunoprecipitation as described earlier ( 24). The ChIP-captured DNA was then screened for p53 promoter fragments by PCR analysis using the following p53 specific primers: (mouse) 5′-GTAGAGTAAGCCCCCGGAAG-3′ and 5′-GGTTACCGGGATTCAAAACA-3′ (the amplified fragments correspond to −925 and −424 region of the mouse p53 promoter from AF287146); (human) 5′-TGGGAGTTGTAGTCTGAACGCTTC-3′ and 5′-GAGAAGCTCAAAACTTTTAGCGCC-3′ (the amplified fragments correspond to −693 to +89 region of the human p53 promoter from X54156). Genomic DNA was used as a control for the amplification efficiency of each primer pair.
Electrophoretic mobility shift assay. WT and Egr1-null MEFs were grown in tissue culture to confluence, incubated in serum-free medium for 1 hour, and then supplemented with 20% fetal bovine serum for 1 hour. Cells were then lysed in hypotonic buffer (10 mmol/L HEPES, 25 mmol/L KCl, 1 mmol/L EDTA, 1 mmol/L EGTA) containing 1:100 protease inhibitor cocktail (Sigma, St. Louis, MO) on ice for 10 minutes. Samples were centrifuged at 1,500 rpm for 7 minutes and the pellets were incubated on ice for 30 minutes in nuclear extract buffer (50 mmol/L Tris-HCl, 420 mmol/L KCl, 5 mmol/L MgCl2, 0.1 mmol/L EDTA) with 1:100 protease inhibitor cocktail. The samples were centrifuged at 5,000 rpm for 30 minutes at 4°C and the supernatants were removed for protein concentration determination by the BCA method as per manufacturer's instructions (Pierce Biotechnology, Rockford, IL). These extracts were used as nuclear protein in the gel shift assays below. Double-stranded synthetic oligonucleotides (IDT DNA, Coralville, IA) corresponding to possible Egr1-binding sites as determined by transcription element search system were labeled with [γ-32P]ATP using T4 polynucleotide kinase or by [α-32P]dATP using the Klenow fragment as per manufacturer's instructions (Invitrogen, Carlsbad, CA). The five sites tested were probe A 5′-TTTCCCTTTCTCCCCCGCCCTCCCTTCA-3′ (nucleotides 886-913 from AF287146), probe B 5′-AAACCGTGGGGTTTGGGGGTGGGGCAGTG-3′, probe C 5′-AATGGAAGCTTGGCGGGCGGGATGAACGTT-3′, probe D 5′-AAATCCTGCGGGGCGGGGTGGCGGGGGGTT-3′, and probe E 5′-CTTTCTCCCCATCTCTCCCCCCTTCTTAA-3′ (nucleotides 850-876 from AF287146). Gel shift assays were done as described previously ( 9) with modifications as noted here. Briefly, nuclear protein (20 μg) and labeled DNA (104-105 cpm) were incubated in 20 μL binding buffer (25 mmol/L HEPES, 60 mmol/L KCl, 2 mmol/L MgCl2, 0.1 mmol/L EDTA, 0.5 mmol/L DTT, 100 μg/mL spermidine, 10% glycerol, 100 μg/mL bovine serum albumin) for 20 minutes at ambient temperature. Reaction mixtures containing anti-Egr1 (Santa Cruz Biotechnology, Santa Cruz, CA) were preincubated for 20 minutes at ambient temperature before the addition of probe. In some cases, an oligonucleotide (10 ng) with the consensus sequences for the binding of the transcription factor Sp1, 5′-ATTCGATCGGGGCGGGGCGAGC-3′ (Santa Cruz Biotechnology), was added to reduce background. Reactions were resolved by electrophoreses on “prerun” 4% nondenaturing polyacrylamide gels run in 0.5× Tris-borate EDTA, vacuum dried with heat, and exposed to film at −80°C.
Two-stage carcinogenesis. WT Egr1 heterozygous and Egr1-null mice of C57BL/6 background were produced from two Egr1 heterozygous breeding pairs [a generous gift from Dr. J. Milbrandt ( 19)]. A total of 144 mice ages 2 to 9 months were divided into 11 treatment groups and used for the two-step skin carcinogenesis assay as described previously ( 25). The dorsal surface of all mice was shaved over ∼2 × 2 cm surface area before the start of the experiment. Two days after shaving, a mutagen and tumor initiator, 7,12-dimethylbenz(a)anthracene (DMBA) in acetone, was applied at a concentration of 7.8 mmol/L (200 μL of 2.0 mg/mL) to the shaved area of the mice. After 7 days, the mice were treated thrice weekly with a tumor promoter TPA in ethanol at a concentration of 1 mmol/L (20 μL of 0.61 mg/mL) for 18 weeks. Control groups of at least 15 mice each included mice with no treatment, treatments with DMBA alone, TPA alone, or vehicle alone (acetone-only group, ethanol-only group) for each of the three genotypes (WT, heterozygous, and Egr1 null). The onset of the tumors and their growth was monitored for the next 18 weeks. After 18 weeks of TPA treatment, necropsy was done. Samples of tumors and representative organs were fixed in 10% buffered formalin and embedded in paraffin, and 5-μm-thick sections were prepared and stained (H&E) for histologic analysis.
Statistics. Pearson correlation coefficients with Bartlett's χ2 probabilities, regression coefficients, and t tests (unpaired, multiple variance) were carried out as implemented by Systat 4.0 (Evanston, IL) and Excel 2000 (Microsoft Corp., Redmond, WA).
Gene expression profile of normal cycling wild-type and early growth response 1–null mouse embryo fibroblasts reveals differential expression of growth and matrix-related genes. To identify genes regulated by Egr1, we compared the gene expression profiles of WT and Egr1-null MEFs using the mouse Affymetrix GeneChip (MGU74Av2) representing 12,566 genes. Two replicated RNA preparations from separate cultures of WT MEFs and Egr1-null MEFs were hybridized to four Affymetrix MGU75Av2 arrays leading to four different potential comparisons of data sets. In general, a gene was considered to be differentially expressed in Egr1-null MEFs (a) when “called present,” (b) when exhibiting a ≥2-fold change in net fluorescence relative to WT MEFs as recommended ( 23), and (c) when these criteria were observed for all four possible comparisons of the replicate WT and Egr1-null MEF data sets. Using these criteria, 266 genes were identified as differentially expressed. Of this group, 108 genes were down-regulated and 158 were up-regulated in Egr1-null MEFs compared with WT MEFs. Numerous known Egr1 target genes, such as TGFβ1 ( 5, 26), IGFII ( 27), and Fas antigen (CD95; ref. 28), were readily recognized (for a review, see ref. 4). Other notable genes are listed in Table 1 . Several of the differentially expressed genes are involved in growth differentiation and cell cycle control. In this regard, TGF-β1 mRNA expression was decreased 7.2-fold in Egr1-null MEFs and p21Cip1/Waf1 mRNA expression was decreased 2.5-fold. On the other hand, cyclin E and cyclin A2 mRNA expression was increased ∼2-fold in Egr1-null MEFs. These features suggest growth deregulation consistent with previous observation of very rapid and apparently immortal growth exhibited by primary Egr1-null cells from the time of explant ( 18).
To confirm the microarray gene expression results, the expression level of 12 differentially expressed genes listed in Table 1 were also tested by Q-PCR ( Fig. 1 ). The fold induction/repression calculated by Q-PCR assays produced a remarkable concordance with the corresponding ratio determined in the Affymetrix GeneChip analysis. In this regard, stimulatory or inhibitory effects of Egr1 gene ablation was confirmed in all cases examined. Indeed, calculation of the Pearson correlation coefficient of the Affymetrix results with real-time PCR results yielded a value of 0.94, which is significant (P ≤ 0.001). Confirmation of target gene expression was also examined by using an independent RNA preparation from WT MEFs and Egr1-null MEFs with a different passage number (data not shown). Thus, the sum of the results supports the reliability of the microarray gene expression data analysis for selecting genes exhibiting significantly changed RNA levels.
Derivation of network map relating early growth response 1–regulated genes. To determine whether there is a rational basis for the additional significant changes observed here, we looked for known relationships. Of the 266 significantly altered gene expression values between WT and Egr1-null MEFs, 151 expression values represent 143 genes identified by their annotation in the LocusLink database (Supplementary Table S1). The remaining 115 altered expression values represent either expressed sequence tags or genes that have not been annotated yet. Although several genes exhibiting significant change in RNA expression in MEFs were recognizable as directly regulated by Egr1 ( 4, 29), the vast majority of genes were of uncertain relationship to Egr1. We carried out alternate rounds of application of the software BiblioSphere followed by direct confirmation in Pubmed/Medline/Entrez to define putative relationships. Two genes that were linked by BiblioSphere were considered as related if they also were cited as coregulated or correlated in some fashion without regard to direction or mechanism of regulation. Those genes confirmed as correlated with Egr1 or Egr1 target genes (see Materials and Methods) were used as input to BiblioSphere Genomatix software for the successive cycle. Using this iterative process, 66 of the 143 genes were confirmed as previously observed in correlation with direct or indirect Egr1 regulation in one or more experimental studies of the mouse (references are given in Supplementary Table S1). The remaining 77 genes have no reported regulatory associations with Egr1 in the databases consulted. The 66 informative genes and associated references are summarized in Table 1 and Fig. 2 . When direction of induction was reported, this was preserved in the map.
The use of successive cycles revealed the number of putative intermediate regulatory steps relating any gene to Egr1 ( Fig. 3 ). Indeed, all 66 genes identified as informative Egr1-associated genes could be related by a single-branched regulatory network ( Fig. 3). The majority of Egr1-regulated genes are described as under the direct transcriptional regulation of Egr1 (target genes) or within one or two downstream regulatory steps of an Egr1 target gene ( Fig. 3). Two nodes, TGFβ1, a known direct target gene of Egr1 ( 30), and IL6, a known direct regulator of Egr1 ( 31, 32), are upstream regulators of >30 of these genes ( Fig. 2). IGFI constitutes a third node with five known subsequent target genes, including Egr1 itself. Several target genes noted here have been observed to be regulated in combination consistent with the regulatory pathway. Moreover, many of the expression changes observed here are also supported by individual experimental observations in the literature, suggesting that some of the regulatory mechanisms governing expression are common among several cells and tissues. Thus, virtually all 66 informative genes of the original set of 143 known genes defined as significantly differentially expressed between WT Egr1 and Egr1-null MEFs for which there is readily accessible literature may be placed on a single regulatory map consistent with published observations. The consistency between genes observed by expression analysis to be differentially regulated and the published observations provides support for the accuracy of the identification of genes whose expression is significantly altered in Egr1-null fibroblasts.
Early growth response 1 binds to the promoter of p53 in vitro and in vivo. One of the genes exhibiting a significant decrease in expression in Egr1-null fibroblasts is p53. The decrease was confirmed by Q-PCR ( Fig. 1). As further indication of the significance of this change, 18 of the 66 genes that make up the Egr1-regulatory map have no known direct regulatory link to Egr1 but all are well-recognized direct target genes of p53 ( Table 1). Thus, p53 is predicted to be a fourth major node among the Egr1-regulatory relationships observed here ( Fig. 3). This is consistent with previous studies in which it was inferred that p53 was required for the ability of Egr1 to regulate replicative senescence of MEFs in extended culture ( 18).
To test whether Egr1 is a direct regulator of p53, we examined the promoter of p53 for potential Egr1-binding sites. There are at least five potential Egr1-binding sites within 2 kb upstream of the ATG start site of mouse p53 gene that are consistent with the consensus sequence 3′-GCGGGGGCG-5′. To determine whether Egr1 bound to these sites, preliminary EMSAs were carried out using synthetic double-stranded deoxyoligonucleotides (dsDNA) and recombinant Egr1 for five potential Egr1-binding sites of the mouse p53 promoter (see Materials and Methods). Three sites were not able to bind Egr1 in parallel experiments, indicating that sites A and B were specific. Sites A and B exhibited strong binding and were examined in more detail.
Because genetically defined MEFs were available, nuclear extracts of both WT MEFs and Egr1-null MEFs as sources of endogenous Egr1-containing extracts and control extracts and their ability to bind dsDNA of each site were compared ( Fig. 4 ). In each case, extracts of WT MEF but not Egr1-null MEF formed a strong single band in EMSA assays. The band is similar to that formed with authentic Egr1 and is not mimicked by Sp1, another transcription factor that binds to GC-rich sequences (data not shown). Moreover, in both cases, the band was disrupted by inclusion of anti-Egr1 antisera, a known and specific property of anti-Egr1 on Egr1-DNA complexes ( 8, 9). The above result confirms Egr1 binding to p53 in vitro.
To test whether Egr1 and not other Egr1 family members are involved in binding to p53 and to test whether the interaction occurs under the conditions that prevail in living cells, we examined whether it was possible to recover Egr1 bound to chromatin in WT MEFs that had been cross-linked in the living state by addition of formaldehyde as described (refs. 17, 33, 34; Fig. 5 ). As a positive control, human prostate DU145 cells ( 29) were analyzed in parallel. The precipitated Egr1-DNA complexes were analyzed by PCR for the presence of p53 promoter sequences using primers specific for domains in the p53 promoter that had putative Egr1-binding sites (see Materials and Methods). Anti-Egr1-treated chromatin but not nonimmune control rabbit IgG precipitated DNA gave a readily detectable p53-specific band after PCR amplification ( Fig. 5), thus indicating the presence of Egr1-p53 promoter complex in vivo. No signal was obtained from Egr1 immunoprecipitates that were prepared using Egr1-null cells, further indicating that specific Egr1-p53 promoter complexes only occur in WT MEFs ( Fig. 5). Moreover, a PCR product similar to that from the WT MEF ChIP products was observed when human DU145 prostate carcinoma cells were used together with primers specific for the human p53 promoter. These results provide strong support for the conclusion that WT Egr1 indeed binds the p53 promoter in vivo.
The sum of observations above suggests that Egr1 binds directly to p53 in vitro and in vivo and that this effect may be responsible for the increase in p53 and its target genes in WT MEFs compared with Egr1-null cells.
Early onset of tumors in mice deficient in early growth response 1. The in vivo Egr1-binding properties provide a mechanistic basis for the strong up-regulation of p53 observed in WT MEFs and for the regulation of p53-dependent growth and DNA damage responses by Egr1 in WT MEFs in culture. It might be expected, therefore, that Egr1 plays a regulatory role over p53 in vivo and that Egr1-null mice would share phenotypic properties with p53-null mice, such as increased sensitivity to skin carcinogens ( 35– 37). To test the role of Egr1 in vivo, we carried out a two-step carcinogenesis assay in WT, heterozygous, and Egr1-null mice ( Fig. 6 ). In this assay, the shaved posterior skin of mice was treated once with the mutagen and tumor initiator DMBA (7.8 mmol/L) followed by treatment with the tumor promoter TPA (1.0 mmol/L) thrice weekly for 18 weeks for the three mouse genotypes (WT, heterozygous, and Egr1 null). Control groups treated by only one of the two agents or by the two vehicles of these agents were also included for each of the three genotypes.
Tumor formation and growth was monitored in the 11 groups of treated mice for 18 weeks following the treatment (144 mice in all). Only 3 of 11 groups of the treated mice, WT, heterozygous, and null treated with both DMBA and TPA, developed visible and palpable skin tumors ( Fig. 6A). The tumors of the WT group appeared with an average time of tumor onset of 12.5 weeks after the start of the protocol, which led to tumor development in >60% of the mice ( Fig. 6C). In contrast, Egr1-null mice exhibited a uniform and much earlier time of onset of 8 weeks with a similar prevalence ( Fig. 6C). Indeed, over half of the tumors appeared before any eruption of the skin or notable defect of the WT animals. The difference in the time of appearance is significant (P < 0.005). We examined the histologic appearance of the tumors ( Fig. 6B), which confirmed the presence of papilloma similar to those described for p53-null and p21-null mice ( 35– 37). No differences in the histologic appearance of any of the tumors were noted. Thus, these observations strongly indicate that the absence of Egr1 in genetically defined mice predisposes them to enhanced skin carcinogenesis.
Early growth response 1 is a growth suppressor in primary mouse embryo fibroblasts. In most human tumors, such as breast cancer, fibrosarcoma, and glioblastoma, Egr1 is described to be tumor suppressor gene ( 9– 11) that is required for maximal sensitivity to irradiation ( 17, 38). However, the tumor-suppressing role seems to be tissue specific, because recent studies implicated a tumor growth-promoting role of Egr1 in prostate cancer progression ( 27, 34, 39). Higher levels of Egr1 were found in prostate cancer ( 26, 40, 41) and Egr1 is growth promoting for vascular smooth muscle cells and for rat kidney tumor cells ( 42– 44). In our previous study, we therefore investigated the role of Egr1 by use of a defined genetic difference of primary MEFs from WT and Egr1-null mice generated by Lee et al. ( 19) and showed that deletion of Egr1 leads to a striking phenotype, including complete bypass of senescence and apparent immortal growth consistent with loss of a suppressor gene ( 18). To identify genes responsible for the growth-suppressing role of Egr1, we did gene expression analyses using mouse Affymetrix microarrays in WT and Egr1-null MEFs. Noteworthy, many of the genes down-regulated in Egr1-null MEFs are involved in growth inhibition or apoptosis, whereas genes that were up-regulated stimulate growth and play a role in matrix formation ( Table 1; Supplementary Table S1). In this regard, p53 and p53 marker genes, such as p21Cip1/Waf1 ( 45, 46), GADD45 ( 47, 48), Bax ( 49) and Fas ( 50, 51) were down-regulated in Egr1-null MEFs. In addition, TGF-β1, known as a potent growth inhibitor of epithelial proliferation ( 52), induced βig-h3, known as a growth inhibitor ( 53), serum amyloid A3, known as an inhibitor of proliferation ( 54), and cyclin G and thrombospondin 2, known as inhibitors of cell cycle progression ( 55, 56) were also down-regulated. On the other hand, genes, such as collagen type VI, known to mediate the three-dimensional organization of fibronectin in the extracellular matrix of cultured fibroblasts ( 57), FGF-inducible 15, known to be expressed during fibroblast growth factor-4–induced proliferation ( 58), cyclin E2, known to be aberrantly expressed in human cancer ( 59), or cyclin A2, a major regulator of the cell cycle progression required for progression to S phase ( 60), were up-regulated Egr1-null cells. These changes correlate well with the observed phenotype of Egr1-null MEFs of rapid growth on explant and a complete insensitivity to the growth arrest of “replicative senescence.” Thus, unstimulated low-passage MEFs seem to express many of the factors that regulate growth.
The regulatory map derived here explains not only the presence of genes as downstream events but also their existence in many reported interactions across nodes and as upstream regulators of Egr1. These relationships are shown in the extended map of Fig. 3.
Early growth response 1 regulates growth through the p53-MDM2-p19ARF pathway and transforming growth factor-β1. In our previous study, we proposed that Egr1 represents a novel upstream gatekeeper of the p53 tumor suppressor pathway and thereby has an important effect on the regulation of cellular growth ( 18). However, the exact mechanism of Egr1-dependent regulation of p53 remained unknown. In this study, we show that Egr1 directly binds the p53 promoter in vitro and in vivo using EMSA and ChIP experiments, respectively. We have observed that Egr1 binds to the p53 promoter at two distinct sites ( Fig. 4). This is in accordance with studies that showed that Egr1 transactivates the human p53 gene promoter of human melanoma cells ( 14, 15). In this case, however, the binding by Egr1 was associated with p53-mediated apoptosis stimulated by thapsigargin. Thus, the regulation of p53 in melanoma cells does not clearly provide a general explanation of the growth regulation and DNA damage response roles of Egr1 observed in MEFs. On the other hand, in WT MEFs, Egr1 is required for the normal level of expression of p53 mRNA and protein and for the expression of p53 target genes. Moreover, Egr1 is required for the p53-dependent development of replicative senescence of primary cultures ( 18), apoptosis in response to UV irradiation ( 17), and p53-dependent cell cycle arrest that follows DNA damage. Here, we find that WT cells that express Egr1 exhibit increased p53 protein and numerous p53 target gene products. Moreover, the mouse p53 promoter contains at least two putative Egr1-binding sites. These sites were confirmed by EMSA studies and by ChIP studies using antibodies to native Egr1. These results support the conclusion that in MEFs the nuclear regulation of the p53 promoter by Egr1 is an important mechanism regulating the level of p53 protein. This mechanism likely is the basis of the gatekeeper function ( 18) of Egr1 in cell cycle regulation in MEFs.
The expression analysis studies carried out here also revealed additional potential growth control mechanisms. First, in addition to p53, the TGF-β1 pathway might be another important component involved in Egr1-dependent growth control. TGF-β1 RNA is down-regulated in Egr1-null MEFs ( Table 1; Fig. 2) and up-regulated in Egr1-null MEFs infected with an Egr1 retrovirus ( 18). It has been shown that TGF-β1 is a direct target of Egr1 ( 3, 61). TGF-β1, in turn, may participate in the regulation of Egr1-dependent growth suppression through the induction of p21Cip1/Waf1 (ref. 62; Table 1) or by induction of p27Kip1 ( 63, 64), which represents an important mediator of the permanent cell cycle arrest ( 65). Second, we have observed recently that PTEN also is a direct target of Egr1 and is increased on reexpression of Egr1 ( 17). PTEN expression also results in an up-regulation of p27Kip1 ( 66). Intriguingly, p27Kip1 connects the TGF-β1- and p53-MDM2-p19ARF pathways by its ability to facilitate the assembly of the cyclin D-CDK4 complex ( 67), thereby inactivating pRb and releasing E2Fs, which in turn can induce p19ARF ( 68, 69).
Of these possibilities, the steady-state RNA expression differences for the 266 significant changes provide some indication of the mechanism operating in MEFs ( Fig. 3). Neither PTEN nor p27Kip1 is significantly changed, whereas the p53 and TGF-β1-regulated gene p21Cip1/Waf1 are significantly up-regulated. Moreover, the p21Cip1/Waf1-regulated gene, cyclin E, is significantly down-regulated. This configuration of results argues that growth regulation of MEFs is dominated by the p53-dependent control of CDK2-cyclin E complex.
Additional p53-regulated genes observed here, such as topoisomerase IIa, XPC/Rad53, and XPC stabilizer ( Fig. 3), may play a role in the p53-dependent DNA damage response characterized by cell cycle arrest and enhanced DNA repair ( 70– 72). The absence of the p53-dependent mechanisms provides an explanation for the lack of growth arrest and the lack of a DNA damage response by Egr1-null MEFs ( 18).
Early growth response 1–null mice exhibit enhanced tumorigenesis. We reasoned that, if Egr1 is a major regulator of the expression of p53, it might be expected that Egr1-null mice would share aspects of the phenotype of p53-null mice (i.e., a greatly increased rate of tumorigenesis for a variety of malignancies on aging or on challenge by skin carcinogenesis; refs. 35– 37). However, accelerated or spontaneous tumorigenesis is not a feature of Egr1-null mice ( 73, 19), which may be related to the known compensatory properties of other Egr1 family members ( 19, 74). The two-stage carcinogenesis experiment in the Egr1 WT and Egr1-null mice shows the early onset of tumors in Egr1-null mice, consistent with the expected phenotype of p53-null mice. The effect is striking in that the group of Egr1-null mice uniformly developed one or more palpable and visible skin tumors in over half the group before any mice of the WT group exhibit lesions. Histologically, all tumors appeared similar and exhibited the typical features of hyperplastic and hypertrophic epithelium with markedly increased epidermal thickness, prominent rete pegs, and florid hyperkeratosis. Infiltrative epithelial cells were not seen. The significantly increased time of onset of tumor formation is the first observation of increased tumor susceptibility in Egr1-null mice.
Interleukin-6 is a major node in mouse embryo fibroblasts and may explain the role of early growth response 1 in prostate cancer. Egr1 expression is regulated in part by the IL-6 signal transduction pathway ( 32, 75). The role of IL-6 in prostate cancer has drawn considerable attention ( 76). Autocrine-derived IL-6 is a potent growth factor for androgen-independent cells, promotes colony formation, and confers resistance to apoptosis. Circulating levels are elevated in metastatic prostate cancer, indicating a role in progression. Moreover, a variety of studies have linked Egr1 with an oncogenic role in prostate cancer ( 26, 27, 39– 41; reviewed in ref. 29). Egr1 mRNA and protein are elevated in prostate cancer in proportion to grade and stage and tumor progression in the TRAMP model of prostate cancer is slowed in hybrid Egr1-null TRAMP mice. Moreover, systemic treatment of TRAMP mice with antisense Egr1 reduces Egr1 expression and tumorigenicity ( 39). These mice lack functional p53 as do a significant fraction of human prostate cancers. Thus, one unifying although speculative explanation is that the role of IL-6 in human prostate cancer works in part by induction of Egr1, which functions as an oncogene in this setting.
Several additional reported Egr1-regulated pathways were not observed as significantly altered in Egr1-null cells. Thus, there is no representation of the inflammatory roles of Egr1, such as by differential regulation of thrombospondin, tissue factor, and the members of the platelet-derived growth factor family as has been observed in inflammatory settings ( 77, 78). This may be because the basal tissue culture conditions examined exclude proinflammatory responses. Thus, considerable additional characterization may be necessary to achieve complete understanding of Egr1.
Taken together, our results suggest that Egr1 exerts its growth control function by directly interacting with p53 promoter in MEFs leading to activation of the p53 pathway and furthermore establishes the role of Egr1 as a tumor suppressor in whole animals.
Grant support: Deutscher Akademischer Austauschdienst and AACR 2002 Scholar-in-Training award (A. Krones-Herzig), USPHS/NIH grants CA76173 (D. Mercola) and CA67888 (E.D. Adamson), and Department of Defense BCRP grant DAMD17-01-005 (E.D. Adamson).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank P. Charnay, N. Mackman, and J. Milbrandt for Egr1-null mice.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
A. Krones-Herzig and S. Mittal contributed equally to this work.
- Received October 27, 2004.
- Revision received February 25, 2005.
- Accepted April 9, 2005.
- ©2005 American Association for Cancer Research.