|Home | About | Journals | Submit | Contact Us | Français|
Cellular stress results in profound changes in RNA and protein synthesis. How cells integrate this intrinsic, p53-centered program with extracellular signals is largely unknown. We demonstrate that TGFβ1 signaling interferes with the stress response through coordinate transcriptional and translational repression of p53 levels, which reduces p53-activated transcription, and apoptosis in precancerous cells. Mechanistically, E2F4 binds constitutively to the TP53 gene and induces transcription. TGFβ1-activated Smads are recruited to a composite Smad/E2F4 element by an E2F4/p107 complex that switches to a Smad co-repressor, which represses TP53 transcription. TGFβ1 also causes dissociation of ribosomal protein RPL26 and elongation factor eEF1A from p53 mRNA, thereby reducing p53 mRNA association with polyribosomes and p53 translation. TGFβ1-signalling is dominant over stress-induced transcription and translation of p53 and prevents stress-imposed downregulation of Smad proteins. Thus, crosstalk between the TGFβ and p53 pathways defines a major node of regulation in the cellular stress response, enhancing drug resistance.
The cellular response to stress signals involves profound changes in RNA and protein synthesis whose net output directs specific cell fate decisions (Spriggs et al., 2010). The tumor suppressor protein p53 is the main transcription factor that orchestrates the stress program largely by inducing cell cycle arrest or apoptosis. p53 activity is controlled predominantly by protein stability and posttranslational modifications and is a frequent target of mutations during tumorigenesis (Levine et al., 2006; Vousden and Prives, 2009). Much less is understood about regulation of p53 biosynthesis, although p53 mRNA deregulation has been observed in human cancers (Saldana-Meyer and Recillas-Targa, 2011). Moreover, seminal studies have elegantly demonstrated that de novo p53 translation mediated by the 60S ribosomal protein RPL26, is required for efficient p53 accumulation to direct specific cell fate outcomes (Chen and Kastan, 2010; Schumacher et al., 2005; Takagi et al., 2005).
Transforming growth factor β (TGFβ) has a dual role in cancer by acting as a tumor suppressor through cell growth arrest and as a tumor facilitator at later stages (Massague, 2008). Central to TGFβ1 signaling is phosphorylation of Smad 2/3 transcription factors by the TGFβRI/TGFβRII receptor complex. Phosphorylated Smads assemble into heterotrimeric and -dimeric structures with Smad 4 and translocate into the nucleus as activated complexes. Interaction of Smad complexes with other DNA-binding proteins targets them to specific promoters where they activate or repress transcription (Massague et al., 2005). TGFβ1 signaling controls cell growth, invasiveness and the Epithelial to Mesenchymal Transition (EMT) through both activation or repression of transcription and translation of its target genes (Massague, 2008; Pardali and Moustakas, 2007) (Chaudhury et al., 2010; Hussey et al., 2011; Lin et al., 2010). Yet the strategies used by TGFβ1 to switch from a tumor suppressor to a cancer enhancer and the stage in which it occurs are largely undefined.
In unstressed cells various p53 family members can cooperate with TGFβ/Smad signaling to facilitate Xenopus mesoderm differentiation. Also, in certain mammalian cells that lack p63 and p73, p53 can enhance TGFβ-mediated growth arrest (Cordenonsi et al., 2003). Smads also associate with mutant p53 to deregulate p63-mediated transcription and enhance metastasis (Adorno et al., 2009). Thus, the influence between p53 family members and TGFβ may be significant in tumor biology. But whether TGFβ signaling directly intersects with the p53-induced stress response to impact cell fate decisions is poorly understood.
To address these issues, we examined the effects of TGFβ1 on the DNA damage response using non-tumorigenic, spontaneously immortalized human mammary epithelial cells. We found that TGFβ1-activated Smads attenuate the stress-induced p53 transcriptional program and protect damaged cells from apoptosis through coordinate transcriptional and translational repression of p53 protein levels. TGFβ1-mediated downregulation of p53 occurs in precancerous and some breast and lung cancer cells but not in patient-matched normal mammary cells, and confers apoptotic resistance to a variety of chemotherapeutic agents. Mechanistically, TGFβ signaling induces assembly of a Smad/E2F-4/p107 repressor complex on the TP53 gene which downregulates transcription and disrupts interaction between the ribosomal protein RPL26 and the elongation factor eEF1A with p53 mRNA to attenuate p53 translation. Our findings demonstrate an unexpected dominance of TGFβ signaling over the cellular stress response by its ability to simultaneously affect two central nodes of regulation: transcription and translation. These results reveal a tumor-enhancing role for TGFβ in which it facilitates the survival of damaged precancerous and malignant cells by impairing the pro-apoptotic actions of p53.
To examine the impact of TGFβ signaling on the p53-mediated DNA damage response, we used immortalized, non-tumorigenic human breast epithelial MCF10A cells. DNA damage in these cells by the anti-cancer drug Doxorubicin (DoxR) resulted in p53-dependent expression of p21, HDM2, and the pro-apoptotic gene PUMA (Figures 1A, 1B) and induced p53-dependent apoptosis (Figure S1). TGFβ1 also efficiently activated the Smad pathway in MCF10A cells (Figures 1A, 1DE, ,2C;2C; Figure S6; and data not shown). We then examined the effects of crosstalk between TGFβ and p53 signaling by simultaneously inducing both pathways with TGFβ1 and DoxR. Unexpectedly, we found that activation of p53 target genes p21 and HDM2 was reduced in the presence of TGFβ1 (Figure 1A). Since either DNA damage or TGFβ signaling can activate the p21 gene, the observed interference between these two pathways on p21 expression was surprising. We further investigated the effects of TGFβ signaling on the p53-mediated damage response by priming MCF10A cells with TGFβ1 24 hours before the addition of DoxR (Figure 1C). This protocol allows TGFβ1 signaling to exert both direct and indirect effects and may approximate a more physiological condition since sustained TGFβ activation of Smads is observed in both normal and neoplastic breast tissues, which we corroborated (Figures S2, S3). This approach also resulted in attenuation of the p53-mediated response (Figure 1D). Therefore, the priming protocol was chosen for all subsequent studies since it more closely recapitulated the in vivo microenvironment.
We next analyzed primary patient-matched sets of normal human mammary epithelial cells (HMECs) and its rare variant vHMEC subpopulation, which displays precancerous properties (Crawford et al., 2004; Romanov et al., 2001), obtained from cancer-free individuals undergoing reduction mammoplasty. Remarkably we found that TGFβ1 downregulated the p53 pathway in the precancerous, extended lifespan vHMECs but not in normal HMECs (Figure 1E). Interestingly, DNA damage of MCF10A cells impaired TGFβ1-induced expression of PAI-1 and MMP2 (Figure 1A) and downregulated cellular levels of Smads 2–3 and 4, which could be reversed by TGFβ1 stimulation (Figure S1). Together, these data support the notion of reciprocal interference between the stress response and TGFβ pathways and suggest that TGFβ1 can overcome the general stress-mediated shutdown of gene expression.
Chromatin immunoprecipitation studies revealed reduced association of the active form of RNA polymerase II (S5P-RNAPII) with p53 target promoters in TGFβ1-treated cells (Figure 2A). Correspondingly, a specific significant reduction of DNA damage-induced association of p53 with the p21, HDM2 and PUMA promoters in TGFβ1-treated cells was also detected (Figure 2B). Conversely, TGFβ1 strongly induced interaction of Smads 2 and 3 with the PAI-1 promoter, which is reciprocally reduced by DoxR (Figure 2C). Thus, TGFβ1 signaling results in insufficient recruitment of p53 and RNAP II complexes to target promoters causing a downregulation of p53-responsive genes upon stress.
We then measured total and stress-stabilized serine 15 phosphorylated (S15P)-p53 accumulation over time in DoxR-treated MCF10A cells in the presence and absence of TGFβ1. Surprisingly, we found that levels of both phosphorylated and total p53 were significantly reduced in cells exposed to TGFβ1, indicating a decrease in cellular concentrations of p53 (Figure 2D). A similar effect was obtained when MCF10A cells were stressed by UV-irradiation (Figure 2E). Strikingly, TGFβ1 also decreased DNA damage-induced wild type p53 levels in vHMECs but not in patient-matched normal HMECs (Figure 2F), suggesting that this phenotype is acquired at a very early stage in tumorigenic transformation.
This effect was found in mutant p53-expressing breast cancer cell lines MDA-MB-231 and MDA-MB-435 as well as in transformed (immortalized) bronchial epithelial cells and some wild type p53-expressing lung cancer cells (Figure 2G,H). Thus, TGFβ1-mediated downregulation of stress-induced p53 protein levels is independent of wild type p53 activity and is not restricted to breast tissues. These results suggest that TGFβ1 might attenuate both the wild type and mutant, DNA damage-induced p53 pathway in precancerous and tumorigenic human cells by reducing levels of bulk cellular and promoter-bound p53 protein, downregulating critical p53 target genes, and potentially compromising p53-directed cell fate choices.
Additionally, immunohistochemistry analyses of normal human breast tissues and breast carcinomas indicated that co-activation of TGFβ and p53 (either wt or mutant) pathways does not always exist. Moreover, Western analyses of freshly frozen lysates from another tumor set showed that activated Smad 2 might inversely correlate with both wild type and mutant p53 levels. This suggests that cooperation between TGFβ and apoptotic (p53-dependent or -independent) pathways might represent a potential node of deregulation in specific tumor contexts (Figures S2, S3).
We then addressed whether TGFβ1 can modify cell fate decisions upon DNA damage. Indeed, in the presence of TGFβ1, DoxR-induced cell death was significantly reduced as measured by propidium iodide DNA content analysis (Figure 3A). Consistently, TGFβ1 decreased both DoxR- and UVC-induced apoptosis (Figure 3B, left panel). Notably, the proliferative rates at the times of these measurements were similar for both TGFβ1- and vehicle-treated cells, as assessed by BrDU incorporation (Figure 3B, right panel). This suggested that impaired cell death mediated by TGFβ1 was independent of cell proliferative rates. Strikingly, TGFβ1 markedly abrogated cleavage of Poly(ADP-ribose) Polymerase (PARP)-1 into its inactive 89 kDa catalytic fragment when induced by 3 types of conventional chemotherapeutic drugs: Doxorubicin; 5-fluorouracil and Paclitaxel (Figure 3C, D and data not shown). PARP-1 participates in the TGFβ-induced transcriptional cycle by interacting with Smad complexes at TGFβ-regulated promoters (Lonn et al., 2010). Thus, our results support the existence of a positive feedback in which TGFβ1 may protect PARP-1 from cleavage to ensure homeostasis of the TGFβ1 pathway in damaged cells.
Interestingly, TGFβ1 treatment of vHMECs and MCF10A cells induces cell cycle arrest at G1 after prolonged incubation (48 hr) as determined by BrDU incorporation and cell cycle profiles (Figure 3D and data not shown). Conversely, MDA-231 and MDA-435 cells were refractory to growth arrest. Therefore, cell cycle arrest and protection from apoptosis resulting from TGFβ1 signaling appear as two independent events. These observations prompted us to test whether p53 is required for TGFβ1-mediated cell cycle arrest in MCF10A cells. We found that p53 knockdown in MCF10A cells, which express similar levels of p63 and p73 as control cells (Figure S1B), did not impede TGFβ1-induced growth arrest or p21 or PAI-1 mRNA activation (Figure 3F, G; data not shown). TGFβ1 also reduces DNA damage-induced G2/M cell cycle arrest at the expense of an increased number of cells arrested in G1, indicating that they can overcome the G2/M imposed arrest (Figure 3H). In addition, TGFβ1 increased the number of surviving cells after 48 hr exposure to DNA damage from 20% to 75%, suggesting that TGFβ1 signaling initiates a cell survival pathway (Figure 3I). Together, our results show that TGFβ1 protects precancerous, immortalized mammary epithelial cells from apoptosis and promotes a shift from the G2/M arrest towards a G1 cell cycle arrest in surviving cells, in part, by impairing p53 function.
TGFβ1 repressed p53 levels over a wide range of physiological and pathological concentrations (Figure S4A, B), but cannot repress exogenously expressed p53 (either wild type or HA-tagged) from a constitutively active promoter (Figure S4C, D). This argued against the possibility that TGFβ1 destabilizes p53 by posttranslational modulation. Moreover, both basal and DoxR-induced levels of HDM2 are also downregulated upon TGFβ1-treatment of MCF10A cells and vHMEC primary cultures (Figure S4E). Thus, the negative effect of TGFβ1 on p53 levels is independent of p53 protein degradation through increased HDM2 expression. Curiously, a previous report showed that TGFβ can activate the HDM2 gene in colon carcinoma HCT116 cells which were genetically modified to re-express the TGFβ receptor (Araki et al., 2010).
Interestingly, an analysis of p53 mRNA levels in human breast tumors revealed that the expression range of both wild type (wt) and mutant p53 is far greater than that found in normal breast tissues (Figure 4A, Table S2). Decreased abundance of p53 mRNA has been correlated with poor prognosis in aggressive breast carcinomas that express wt p53 (Miller et al., 2005). We hypothesized that TGFβ1 might attenuate the p53-mediated DNA damage response by restricting p53 mRNA abundance. p53 mRNA levels were reduced (3–4 fold) by TGFβ1 in MCF10A cells and in primary cultures of vHMECs from three different patients in the presence or absence of DNA damage (Figure 4B-D). In addition, mRNA stability experiments indicated that TGFβ1 does not promote p53 mRNA decay but slightly increases its stability (Figure S4F). The delay in TGFβ-mediated repression of p53 mRNA (Figure 4B) is consistent with the p53 mRNA half-life extending over 9 hours. Together, these results reveal that TGFβ1 potentially acts by repressing TP53 RNA synthesis rather than by facilitating mRNA or protein turnover, which was further confirmed by RT-qPCR using intron-specific primers on vHMEC primary cells (Figure S4G).
Interestingly, repression of p53 protein lasted much longer than p53 mRNA repression after a single addition of initial TGFβ1 (Figure 4E) and we speculated that TGFβ1 might downregulate p53 mRNA translation. A short pulse metabolic labeling experiment with 35S-methionine showed reduced levels of newly synthesized p53 in TGFβ1-treated cells (Figure 4F). Furthermore, ribosome analysis indicated that TGFβ1 treatment specifically displaced p53 mRNA from heavier to lighter poly/monosomes (Figures 4G, S4L and data not shown). TGFβ1 also prevented the redistribution to lighter poly/monosomes of p53 mRNA induced by DoxR (Figure S5A).
In addition, RNA Immunoprecipitation (RIP) experiments revealed that TGFβ1 specifically inhibits interaction of p53 mRNA with both the 60S ribosomal protein RPL26 and the eukaryotic elongation Factor 1A (eEF1A) in the presence or absence of DoxR (Figures 4H–I, S5B–E). RPL-26 interaction is essential for efficient translation of p53 mRNA in a CAP- and poly A-independent manner through specific pairing of both the 5’ and 3’ UTR (Chen and Kastan, 2010), while eEF1A delivers aminoacyl-tRNAs to the A-site of the ribosomes and its transient association with mRNAs also indicates an efficient translational elongation process.
Importantly, neither cellular levels of RPL26 and eEF1A nor their specific distribution across cytosolic fractions were affected by TGFβ1 (Figure S5A). Moreover, the distribution between cytoplasmic (ribosome-bound) and nuclear (ribosome-free) pools of RPL26 was also unaffected by DoxR (Figure S5E, F), in agreement with results by Kastan and colleagues (Chen and Kastan, 2010), or by TGFβ1. TGFβ1 prevented DoxR-induced accumulation of RPL26 in the 60S/80S ribosomes, indicating its dominance over the DNA damage response (Figure S5A). We consistently found that association of RPL26 and eEF1A with Smad 4 mRNA was dampened by DoxR-induced general translational repression (Spriggs et al., 2010) but restored in the presence of TGFβ1 (Figure 4H–I). Mechanistically, our data suggest that TGFβ1 represses p53 translation at the elongation stage, consistent with reports showing that TGFβ1 affects this step in translation (Lin et al., 2010).
Altogether, these results demonstrate that TGFβ1 abrogates the DNA damage response at the level of protein translation in at least two ways: by directly repressing p53 mRNA translation through interference with RPL26 and eEF1A binding and by relieving DoxR-induced repression of Smad mRNA translation.
Our findings also suggested that activated TGFβ1 signaling might repress TP53 gene expression through Smad proteins. We identified three potential Smad binding regions (SR1–3) in the TP53 gene containing shortly spaced SBEs (Figure 5A). The SR2 localizes in the regulatory region of the TP53 gene and contains an E2F-4 binding site that overlaps an imperfect SBE. Depletion of Smad 4 protein from MCF10A cells resulted in both increased p53 protein levels (Figure 5B) and abrogation of TGFβ1-induced TP53 gene repression (Figure 5C). Thus, TGFβ1 represses TP53 through the canonical Smad-dependent pathway.
ChIP assays showed that TGFβ1 induced a clear recruitment of phosphorylated P-Smad 2 to the TP53 gene (Figure 5D) with highest association near the SR2 region. P-Smad 2 remained bound for up to 6 hours with maximal occupancy at 2 hours (Figure S6A). In addition, Smad 3 was also recruited to similar locations with faster but weaker association than P-Smad 2 (Figures 5D, S6A). These results support the hypothesis that Smads contribute directly to transcriptional repression of the TP53 gene.
Additional ChIP experiments indicated that both Smad 2 and Smad 3 associate with the SR2 region upon TGFβ1 stimulation in unstressed cells and, importantly, in the presence of DoxR (Figure 5E), indicating that TGFβ1–induced repression of p53 is stable in the context of damaged cells and predominates during the stress response. Furthermore, association of Smad 2/3 proteins with the SR2 region of TP53 was accompanied by a simultaneous decrease in promoter recruitment of the initiating form of RNA Polymerase II (S5P-RNAP II) and abundance of the elongating form of RNAP II (S2P-RNAP II) throughout the coding region of the gene (Figure 5F). Moreover, TGFβ1-induced decrease of RNAP II interaction with the p53 promoter was also observed in the presence of DoxR. These results indicate that upon TGFβ1 signaling in both unstressed and stressed cells, Smad 2/3 complexes are recruited to the TP53 gene and repress transcription by reducing the assembly of transcriptional initiation complexes on the promoter. TGFβ1 also decreased association of RNAP II with the p53 promoter in MDA-MB-231 and HaCaT cells, which express mutant p53 proteins (Figure S6D, E).
Using high–resolution IP analysis with purified mononucleosomes (MnIP) we confirmed P-Smad 2 binding specifically within the SR2 sequence of the TP53 gene (Figure 6A). Next, we performed co-transfection assays with reporter plasmids containing fragments from the p53 promoter fused to a luciferase reporter gene (Figure 6B). As expected, TGFβ/Smad 4 co-transfection efficiently transactivated the PAI-1-derived SBE multimeric, activatable pBV-SBE4-luc control reporter (Figure 6C). By contrast, both p53 promoter constructs were downregulated by TGFβ/Smad 4 (Figure 6D). Deletion or mutation of the SR2 region in the p53 promoter (+80 to +150) abrogated TGFβ/Smad-mediated repression (Figures 6E, S6), indicating that the SR2 sequences are required for repression. Together, these results demonstrate that TP53 transcription is directly repressed by TGFβ1/Smads through the SR2 sequence within the p53 promoter.
The TGFβ-Inhibitory Element in the c-MYC promoter contains a composite sequence formed by an E2F-4 site and an imperfect SBE (GGCT) similar to SR2 in the TP53 gene. Smads 2/3 and 4 repress c-MYC transcription by associating with E2F-4 or −5 and Retinoblastoma-like (RBL)-1/p107 (thereafter p107) to form a complex in the cytoplasm that is recruited to the c-MYC promoter upon TGFβ stimulation (Chen et al., 2002). Indeed, endogenous E2F-4 and p107 specifically associate with the SR2 element in the p53 promoter before and after TGFβ1 stimulation (Figure 7A). This suggested that a conditional co-repressor complex was pre-assembled on the p53 promoter and responsive to subsequent TGFβ1 signaling. Next we depleted E2F-4 protein from MCF10A cells by shRNA expression (Figures 7B, S7A). Quantitative E2F-4 depletion (shRNAs A6, A10) not only significantly impaired TGFβ1-imposed repression of the p53 promoter but also downregulated basal p53 transcription in the absence of TGFβ1 (Figures 7C; S7B, C). Thus, E2F-4 normally functions as an activator of basal TP53 transcription but, upon TGFβ1 signaling, it switches to a Smad co-repressor to negatively regulate the p53 promoter.
We then analyzed interaction of Smad-containing complexes with the SR2 sequence in Smad 4- or E2F-4 depleted cells by ChIP. Loss of Smad 4 had no effect on E2F-4 binding to the p53 promoter (Figure 7D); however, both Smad 4 and E2F-4 were necessary to recruit P-Smad 2 to the SR2 region upon TGFβ1-induction (Figure 7E). Furthermore, E2F-4 was specifically required for recruitment of P-Smad 2 to the TP53 gene but not to the PAI-1 locus whereas Smad 4 was necessary for P-Smad 2 binding to PAI-1 (Figures 7E, S7D). Thus, association of E2F4 with the p53 promoter precedes Smad activation by TGFβ1 and does not require the presence of nuclear activated Smad complexes. These results indicate that E2F-4 is part of a pre-existing, transcription-engaged, Smad-independent complex with a dual role as either an activator or a TGFβ1-induced co-repressor and serves as an interface between Smads and TP53 gene activity.
Given that TGFβ1 impairs the stress response at both the transcriptional and translational levels, we examined whether this would result in enhanced survival of MCF10A cells treated with either DoxR or Paclitaxel. A comparison of the drug concentration required to inhibit cell growth by 50% (IC50) revealed that exposure of cells to TGFβ1 increased resistance to both DoxR (ΔIC50=2 fold) and Paclitaxel (ΔIC50=8 fold), resulting in a greater number of surviving damaged cells (Figure 7F, upper panels). Importantly, depletion of Smad 4 in MCF10A-C2 cells abolished the enhanced drug-resistance and cell survival to both drugs, indicating that Smad-mediated transcriptional downregulation of the stress response plays a central role in the ability of TGFβ1 to interfere with DNA damage-induced cell death. Together, our results indicate that TGFβ1/Smad signaling has a protective role against elimination of damaged cells, which is achieved, at least in part, through transcriptional and translational repression of the TP53 gene (Figure 7F, lower panels).
We show that cellular signalling by the pleiotropic cytokine TGFβ1 interferes with the stress response through coordinate transcriptional and translational repression of p53, which impacts cell fate decisions upon stress (Figure 7G). We propose that TGFβ/Smad signaling can attenuate many downstream events in the p53-mediated DNA damage response in both precancerous and tumorigenic cells that retain wild type p53.
Crosstalk between TGFβ and p53 plays a pivotal, albeit complex, role in tumorigenesis and tumor progression. We focused on the intersection between TGFβ and p53 specifically during DNA damage when p53 has a critical function in directing the stress response. Using human precancerous and tumorigenic mammary and lung cells, but not normal primary HMECs, we revealed a direct antagonistic crosstalk in which TGFβ protects cells from the p53-mediated DNA damage response. By repressing TP53 transcription and translation, TGFβ1 very effectively lowers cellular levels of p53 protein before and after stress, a process conserved in various precancerous and malignant (wild type and mutant p53) human cancer cells.
Certainly, the disparities that exist among published reports may reflect the complex nature of TGFβ and p53 pathway interactions that are influenced not only by tissue- and cell-stage specificity but also by the experimental paradigm used in each study. We found that TGFβ-p53 cooperation is absent in various cancer cell lines which corresponded with an inverse correlation between TGFβ signaling and expression of p53 and the apoptotic marker PUMA in subsets of human breast tumors.
Long-term functional abrogation of p53 is a hallmark of many cancers and mainly occurs by mutations in the coding sequence or degradation rather than by decreasing p53 protein synthesis. In contrast to HDM2/HDMX-mediated post-translational deregulation of p53 levels, transcriptional or translational control of TP53 gene expression as a mechanism to evade wild type p53 signaling in cancer is poorly understood. Several studies have reported the existence of potential regulatory elements in the TP53 gene promoter. Yet only a few have shown that transcriptional downregulation of p53 mRNA synthesis impedes p53-mediated responses in tumorigenesis (Saldana-Meyer and Recillas-Targa, 2011).
Elegant studies by Kastan and colleagues demonstrated the significance of RPL26 in mediating p53 mRNA translational induction in stressed cells by interacting with the (5’/3’)-UTRs of p53 mRNA (Chen and Kastan, 2010; Takagi et al., 2005). Translational induction of p53 is CAP- and poly A-independent, which suggests that it may occur at the level of elongation. This raised the possibility that control of p53 translation may be a potential target for malignant transformation, which had not been previously demonstrated. TGFβ1 is known to regulate mRNA translation during tumorigenesis, generally by repressing elongation while stimulating translation of select EMT-related genes (Chaudhury et al., 2010; Hussey et al., 2011; Lin et al., 2010). Indeed, our results reveal opposite effects of TGFβ1 on p53 and Smad4 interaction with RPL26 and the elongation factor eEF1A as well as accumulation of p53 mRNA in monosomes.
The ability of TGFβ1 to coordinately deregulate p53 by directly targeting both transcription and translation was unexpected. Importantly, TGFβ1-signalling is dominant over stress-activated transcription and translation of p53 (Spriggs et al., 2010; Takagi et al., 2005) while simultaneously preventing stress-imposed downregulation of Smad levels and PARP-1 cleavage.
The E2F4 binding site within the TP53 gene, which is also recognized by E2F1 and E2F6 (data not shown), constitutes a functionally composite E2F/Smad recognition sequence. Curiously, repression of the c-MYC promoter by TGFβ1 was shown to be mediated by a composite E2F/SBE sequence and require the interaction of Smads 3/4 with E2F4/5 and p107/DP1 complexes in the cytoplasm. These complexes only associate with the c-MYC promoter after TGFβ1-induction (Chen et al., 2002). Thus, our study has an intriguing difference in the mechanism of Smad-mediated repression of the TP53 gene compared to c-MYC. Although Smad recruitment to each promoter is dependent upon TGFβ signaling, the E2F4/p107 complex is already assembled on the TP53 promoter before stimulation unlike the situation with c-MYC. While E2F4 is considered to be a transcriptional repressor, it does contain a transactivation domain, as does E2F1–5 (Rowland and Bernards, 2006). We hypothesize that E2F4/p107 proteins form a transcriptionally permissive scaffold at the p53 promoter whose activity can be switched after associating with Smads 2/3 upon TGFβ1 signaling. This mature corepressor complex then triggers repression of TP53 transcription and subsequently impairs downstream p53-mediated events (Figure 7).
p53 family members, including ΔN /TA p63 isoforms, interact with Smads 2/3 and bind to a p53/p63-recognition sequence overlapping the Smad-Binding Element of the Activin-inducible Mix.2 promoter in mesoderm Xenopus cells. TGFβ/Smads can also activate their transcription program and cell cycle arrest in p53-null/mutated epithelial cells (Chen et al., 2002; Datto et al., 1995). Importantly, Cordenonsi and colleagues showed that p53 was only required for Activin or TGFβ to induce cell cycle arrest and gene expression if cells lacked p63 and p73, either naturally or through siRNAs. Notably, p63 is selectively expressed in basal epithelia in breast, prostate and bronchia and most stratified epithelial cells (Di Como et al., 2002). This might explain why TGFβ1 reduces proliferation of basal-like MCF10A cells which express both p63 and p73 even after p53-depletion (Figure S1), but fails to induce growth arrest in MEFs and hematopoietic progenitors from p53 −/− mice (Cordenonsi et al., 2003).
Mutant p53 proteins disrupt the balance between Smads and ΔNp63 /TAp63-mediated transcription to enhance the metastatic potential of TGFβ (Adorno et al., 2009). In studies using murine and Xenopus embryos, the only p53 isoform identified as a Smad 2 mesoderm co-inducer was the p53 splicing variant p53AS. Paradoxically, p53AS lacks the C-terminal domain that is rapidly acetylated by CBP/p300 upon DNA damage in human cells (Vousden and Prives, 2009). Altogether these data suggest that in human epithelial cells it is more likely that p53 family members such as p63, rather than p53 itself, normally cooperate with TGFβ signaling. It is possible that in mammary epithelial cells TGFβ might repress p53 while still cooperating with p63 to induce TGFβ-related phenotypes.
In breast cancers, analyses of a large cohort of tumors have shown that low p53 mRNA levels correlate with decreased p53 signaling, reduced therapeutic response, and poor prognoses (Miller et al., 2005). Our work also demonstrates that TP53 gene expression is deregulated in primary breast tumors. Mechanistically, we observed that TGFβ1/Smad signaling represses p53 mRNA levels in precancerous and metastatic human cancer cell lines in both unstressed and DNA damaged cells. p53 repression by TGFβ1 also occurs in immortalized-(non-tumorigenic) bronchial epithelial cells, highly metastatic NCI-H460 lung cancer cells (Figure 2H), and keratinocyte-derived HaCaT cells (Figure S6), which extends the implications of our findings beyond breast tumors. Interestingly, TGFβ1 signaling represses expression of both wild type and mutant TP53 genes (Figures 2, S6). Since mutant p53 proteins often acquire a gain-of-function (Goh et al., 2011), TGFβ might influence the spectrum of p53-dependent activities, both positive and negative, in tumors that contain such mutants.
Studies with genetically modified mice have shown that TGFβ1 signaling suppresses apoptosis while increasing the invasiveness and metastatic potential of ErbB2-expressing primary tumors (Muraoka-Cook et al., 2006; Muraoka et al., 2002). Moreover, blocking TGFβ signaling improved the effectiveness of chemotherapeutic drugs in cancer cells and murine models of breast cancer (Barcellos-Hoff and Akhurst, 2009; Biswas et al., 2007). Our data support these observations and provide a mechanistic basis by which to understand the inhibitory interaction between TGFβ and DNA damage in breast cancer cells both in vivo and in vitro [reviewed by (Barcellos-Hoff and Akhurst, 2009). In agreement with these reports, we found that TGFβ protects cells not only from DoxR but also from 5-Fluorouracil and Paclitaxel-induced cell death specifically though Smad 4-mediated complexes (Figure 3, Figure 7). During the course of our studies, TGFβ1 signaling was shown to mediate MED12–induced resistance to cisplatin and various tyrosine kinase receptor inhibitors in several cancer cell types (Huang et al., 2012). Although the mechanism was not elucidated in this report, we anticipate that our findings may be generally applicable to other biological contexts that display TGFβ1-mediated drug resistance.
Interestingly, although in certain cell types TGFβ can induce apoptosis, in most epithelial cells, including mammary cells, TGFβ induces cell cycle arrest and EMT rather than apoptosis (Massague, 2008; Pardali and Moustakas, 2007). Our study provides mechanistic insights to explain how cancer progression may be facilitated by the ability of TGFβ1 signaling to promote survival of damaged cells. Moreover, our data indicate that TGFβ can become a tumor promoter by impeding the p53 pathway very early in tumorigenesis (vHMECs, MCF10A) but not in normal cells (HMECs). On this basis, one may predict that breast tissue containing both normal HMECs and slightly abnormal vHMECs will undergo very different cell fate outcomes when exposed to stress in a TGFβ-rich microenvironment, with facilitated elimination of damaged normal cells but enhanced survival of abnormal cells by the mechanisms that we have deciphered. In this scenario, tissue heterogeneity can be understood in part by the distinct programmed stress responses of normal and abnormal cells in a TGFβ-rich or -poor microenvironment.
Primary cultures of HMECs and vHMECs were derived from healthy human donors by reduction mammoplasty. shRNA-expressing cells were generated by lentiviral transduction and selected with puromycin or GFP-based cell sorting. Puromycin was removed from the media 16–24 hours before treatment with TGFβ1 (Peprotech or inhouse purified) or vehicle buffer.
MCF10A cells were transfected using Fugene HD reagent (Roche) with luciferase reporter plasmids containing the p53 promoter, different amounts of the pLV-Smad-HA plasmid and a Renilla-Luciferase reporter. After 16 hours, cells were treated with TGFβ1 (5 ng/ml) for an additional 6–8 hours and luciferase activity was determined with the Dual-luciferase system (Promega).
ChIPs were performed essentially as described with minor modifications (Gomes et al., 2006). Mononucleosomes were obtained using the ChIP-IT-enzymatic kit (Active Motif) with minor modifications and IPs were performed as for ChIPs. For RIPs, extracts from 5 ×107 MCF10A cells were prepared and incubated with specific antibodies (Table S3) using the Magna RIP™ kit (Millipore).
Cell cycle distribution assays of MCF10A cells were performed using the Watson distribution model as described (Gomes et al., 2006). Apoptotic index assays were performed as in Gomes et al (2006) using Alexa Fluor350-labeled Annexin V (Molecular Probes). Cells were sorted in a Becton-Dickinson LSR instrument and analyzed by FlowJo software.
Unless specified otherwise, all data is presented as the mean values ± S.E.M. from at least 3 independent experiments; two-sided t tests assuming unequal variance were used to test the relationships between the means of data sets and p values indicate the probability of the means compared being equal with *=p<0.01 and **=p<0.001.
We gratefully acknowledge Dr. Martha Stampfer for providing variant HMECs; Drs. Inder Verma and Gustavo Tiscornia for p53 and shGT4 shRNA-coding plasmids and advice; Ms. Ruo Huang for expert technical assistance; Dr. Martin Widschwendter for human breast tumors and Emerson lab members for helpful discussions. This work was supported by NCI Grant U54CA143803; the Chambers Medical Foundation; and Cancer Center Grant P30 CA014195. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NCI or the NIH.
Author contributions: FLD designed the project scope, led the team, performed the experiments, and analyzed the data. PG and JZ performed the IHC experiments and data analyses. SVdR performed Western blots on tumor lysates. SKB contributed to Figures 2H, S1B, and S5F. CS purified RNA from frozen tumors. TDT provided tissues and advice. BME designed the project scope, analyzed the data, and supervised the work. FLD and BME wrote the paper.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Supplemental information includes data (7 supplemental figures and 2 tables), extended experimental procedures and supplemental references.