• Open Access

SnoN activates p53 directly to regulate aging and tumorigenesis

Authors


Kunxin Luo, Department of Molecular and Cell Biology, University of California, 16 Barker Hall, MC3204, Berkeley, CA 94720, USA. Tel.: +1 510 643 3183; fax: +1 510 642 7038; e-mail: kluo@berkeley.edu

Summary

We have identified SnoN as a direct activator of p53 to accelerate aging and inhibit tumorigenesis. SnoN has been shown previously to promote proliferation and transformation by antagonizing TGFβ signaling. We show that elimination of this TGFβ antagonistic activity of SnoN in vivo results in accelerated aging and resistance to tumorigenesis. The SnoN knockin mice display a shortened lifespan, decreased reproductivity, osteoporosis, reduced regenerative capacity, and other aging phenotypes, similar to that found in mice expressing an active p53. These activities of SnoN rely on the ability of SnoN to activate p53. SnoN can bind directly to p53 and compete with Mdm2 for binding to p53, preventing p53 ubiquitination and degradation and additionally facilitating p53 acetylation and phosphorylation. SnoN also binds to p53 on the promoter of p53 responsive genes to promote transcription activation. This activation of p53 by SnoN is necessary for its antitumorigenic and progeria activities in vivo because elimination of one copy of p53 reverses the aging phenotypes and accelerates tumorigenesis. Thus, we have revealed a novel function of SnoN in regulating aging and tumorigenesis by directly activating p53.

Introduction

p53 is activated by various stress signals to coordinate cell cycle arrest, apoptosis, senescence, and DNA repair processes (Vousden & Prives, 2009). While activation of p53 serves as an effective mechanism to reduce cancer susceptibility, it also compromises longevity by accelerating aging (Rodier et al., 2007; Donehower, 2009). Expression of a constitutively active p53 in mice protects the animal from tumorigenesis, at the same time accelerates aging (Tyner et al., 2002). Understanding the regulation of this pathway by intrinsic and extrinsic signals is therefore critical to address unsolved questions in senescence and cancer.

Regulation of p53 mainly occurs at post-transcriptional levels, in particular protein stability, and is primarily controlled by the Mdm2 E3 ubiqitin ligase that induces poly-ubiquitination and degradation of p53 (Kruse & Gu, 2009). Mdm2 may further repress p53 activity by forming a repressor complex with p53 on the promoter DNA. The stress signals induce stabilization of p53 by blocking the Mdm2/p53 interaction through either post-translational modification of p53 and Mdm2 or physical competition (Lu, 2010). Phosphorylation and acetylation of p53 can further activate its transcriptional activity and facilitate recruitment of other transcription co-activators (Oren et al., 2002; Kruse & Gu, 2009). Activation of p53 in response to cellular stresses also requires the promyelocytic leukemia (PML) protein. PML is essential for the formation of the PML nuclear bodies (NBs) that allow recruitment of diverse proteins to the nuclear domain (Dellaire et al., 2006; Bernardi & Pandolfi, 2007) and serve as a scaffold to increase the local concentrations of factors involved in p53 activation (Pearson et al., 2000; Pearson & Pelicci, 2001). Loss of PML allows mouse embryo fibroblasts (MEF) to bypass senescence owing to lack of p53 activation (Ferbeyre et al., 2000; de Stanchina et al., 2004).

We have recently identified SnoN as a novel upstream regulator of p53 to induce cellular senescence in response to oxidative stress. SnoN is a member of the Ski family of proteins (Nomura et al., 1989) and is ubiquitously expressed in all cell types and tissues (Deheuninck & Luo, 2009; Jahchan & Luo, 2010). Its expression can be induced during tissue morphogenesis (Jahchan et al., 2010), by injury (Mayoral et al., 2010) and upon stimulation with cytokines such as TGFβ (Zhu et al., 2005). SnoN is a potent negative regulator of TGFβ signaling by binding to the Smad proteins and repressing their transactivation activity (Deheuninck & Luo, 2009). This ability to antagonize the growth inhibitory responses to TGFβ is likely responsible for the pro-oncogenic activity of SnoN (Wu et al., 2002; He et al., 2003). Indeed, reducing SnoN in breast and lung cancer cell lines reversed transformation in vitro and tumor growth in vivo, and re-introduction of wild type (WT) but not the mutant SnoN lacking the Smad binding sites rescued this response (Zhu et al., 2007).

In an effort to uncover Smad-independent functions of SnoN, we generated a knockin mouse (SnoNm/m) expressing a mutant SnoN defective in binding to the Smad proteins in the original snoN locus. Using MEF isolated from the knockin mice, we have shown that high levels SnoN can bind to PML and be recruited to the PML NBs where it upregulates p53 expression, leading to premature senescence (Pan et al., 2009). Consistent with this ability of SnoN to activate the PML-p53 tumor suppressor pathway, overexpression of SnoN inhibited oncogene-induced cellular transformation of MEF cells and significantly blocked chemical carcinogen-induced carcinogenesis carcinogenesis due likely to the accumulation of senescent cells in the SnoNm/m mice (Pan et al., 2009).

Two key unresolved questions are as follows: (i) how does SnoN upregulate and activate p53 once it is recruited to the PML NBs? (ii) What is the physiological function of the SnoN activation of p53? As active p53 promotes aging, we predict that SnoN may also accelerate aging. In this report, we directly addressed these questions. Our studies have revealed a previously unidentified function of SnoN to promote premature aging. We have also determined the mechanism by which SnoN activates p53.

Results

The SnoN knockin mice display accelerated aging phenotypes

The SnoN knockin mouse (SnoNm/m) expresses a mutant SnoN that contains point mutations altering the R-Smad and Smad4 binding sites (Fig. 1A). Our earlier study has shown that this mutant SnoN protein (mSnoN) is defective in binding to the Smad proteins, and as a result of this, cells harboring this mSnoN display elevated Smad signaling activity as well as increased mSnoN protein levels (Pan et al., 2009). Using MEF isolated from the SnoNm/m mice, we have uncovered a novel Smad-independent function of SnoN in inducing premature senescence through modulating p53 (Pan et al., 2009). During the routine maintenance and analysis of these mice, we noticed that these animals were very sensitive to environmental stress and often had difficulties getting pregnant or repairing wounds. Some of these phenotypes are often associated with the aging process. We therefore asked whether these mice display accelerated aging and whether they contain more senescent cells in vivo. To do so, we measured lifespan of a cohort of 43 SnoNm/m mice and 34 WT (SnoN+/+) mice. About 20.9% of SnoNm/m mice died within the first year as compared with only 8.8% of SnoN+/+ mice (Fig. 1A). The median lifespan of SnoNm/m mice is around 75.3 weeks, about 25 weeks shorter than that of SnoN+/+ mice (Fig. 1A). This decrease in lifespan is similar to that displayed by mice expressing an active p53 (Tyner et al., 2002). In the first few months after birth, the SnoNm/m mice did not show apparent gross abnormalities in development. Visible premature aging symptoms in SnoNm/m mice, including growth retardation, gray hair appearance, body weight loss, and kyphosis, started to be detected after 6 month of age (Figs 1B and 2A). Both male and female SnoNm/m mice stopped gaining body mass around 6–12 months of age, resulting in smaller body size (Fig. 1B). SnoNm/m mice also displayed decreased reproductive activity. The number of pups born to the matings of SnoNm/m mice with WT mice was significantly lower than that born to the mating of WT parents (Fig. 1C). Interestingly, when the metabolic activities including the oxygen consumption, generation of CO2 and heat, and the general mobility were evaluated, no obvious difference was observed, suggesting that the accelerated aging observed in SnoNm/m mice does not involve changes in metabolic activities (data not shown).

Figure 1.

 The SnoNm/m mice display shortened lifespan. (A) Top: Schematic drawing of mSnoN. Bottom: Kaplan–Meier survival curve and the table below showed a shortened lifespan by the SnoNm/m mice. n = 34 for SnoN+/+ mice; n = 43 for SnoNm/m mice. P < 0.0072. (B) The SnoNm/m mice were smaller. Top: 18-month-old SnoN+/+ and SnoNm/m littermates are shown. Bottom: body weight analysis. The SnoNm/m mice stopped gaining weight after 6-month old (P < 0.0001). (C) The SnoNm/m mice exhibited reduced litter size. The mean litter size was analyzed during a 6-month period of breeding between WT parents (SnoN+/+ × SnoN+/+: five pairs) and WT/homozygous parents (SnoNm/m male × SnoN+/+ female: six pairs or SnoNm/m female × SnoN+/+ male: five pairs). The three groups were age-matched first time parents. Both male and female SnoNm/m mice displayed a reduced litter size (male: 4.4 P = 0.0060; female: 3.7, P = 0.0010) compared with SnoN+/+ controls (6.4).

Figure 2.

 The SnoNm/m mice exhibit osteoporosis and other aging phenotypes. (A) A 18-month-old SnoNm/m mouse displayed reduced muscle mass, decreased subcutaneous fat and kyphosis (dotted line, left panel). Whole-mount X-ray (middle) and MicroCT (top right) analyses detected decreased bone density in tail, limbs and spinal cord regions in SnoNm/m mice. Lower right panels: H and E stain of cross-sections of tibias from 18-month-old SnoN+/+ (left) and SnoNm/m (right) mice. Note the reduction in cortical bone thickness (c, arrow) and the absence of trabecular bone (t) in the tibia of the SnoNm/m mouse. P: epiphyseal plate. (B). H and E stain of dorsal skin from a 12-month-old SnoNm/m mouse (right) showed near absence of subcutaneous adipose layer when compared to that of the SnoN+/+ (left) mouse. d: dermis; a: subcutaneous adipose tissue. Data from three sets of H and E sections per mouse, and two pairs of mice for each age group were quantified (bottom). (C) Reduced hair regrowth in 12-month-old SnoNm/m mice. The shaved area is indicated by the red dotted line. Pictures were taken 30 days after shaving. Mean hair regrowth was scored with a transparent grid and show in the graph (P = 0.0229, n = 3). Data are presented as mean ± SD. (D) Left: satellite cells isolated from the injuried 12-month-old SnoNm/m mice expressed a higher level of p15, p16 and p27. RNA was extracted from satellite cells prepared from 6- or 12-month-old mice and subjected to RT–PCR analysis. Right: Myogenic differentiation of SnoNm/m satellite cells (m/m) or WT satellite cells (+/+) in culture was monitored by the expression of myogenic marker desmin (red), and myotube formation (phase contrast, middle) and quantified. Data represent means ± SEM from four images for each cell type in two different experiments (P < 0.001). (E) Cryostat sections from the skin of 18-month-old SnoN+/+ and SnoNm/m mice were stained for SA-β-gal. (F) Left: p53 levels in 4- or 18-month-old mice by Western blotting. p53 expression was normalized to tubulin and quantified (bottom). Right: Representative equatorial microscopic images of skin sections from 18-month-old mice stained with anti-p53. The number of p53 postive cells per field was quantified and shown below. Data represent means ± S.E.M from four images per sample in two different experiments (P < 0.001).

As there is no significant difference in bone density and structure between WT and SnoNm/m mice at 1 month of age (Fig. S1A), the marked kyphosis appeared in older SnoNm/m mice (Fig. 2A, left) suggested osteoporosis, an age-related bone degenerative disease common in both human and mouse (Watanabe & Hishiya, 2005). Indeed, when compared to age-matched WT mice, 18-month-old SnoNm/m mice exhibited an overall reduction in bone density as shown by X-ray analysis (Fig. 2A, middle), a severe loss of honey comb-like bone structure inside the vertebrae as indicated by high-resolution MicroCT scan (Fig. 2A, top right), and a reduced cortical bone thickness and absence of trabecular bone as demonstrated by Haematoxylin Eosin (H & E) stain of the tibias cross-sections (Fig. 2A, bottom right). In addition to this early onset of osteoporosis, the skin of SnoNm/m mice also showed signs of prematured aging. H & E staining of dorsal skin sections of 12-month-old SnoNm/m mice revealed a significant decrease in the thickness of subcutaneous adipose layer, but not the dermal layers, when compared to that of age-matched WT skin. These differences were not detected at 4 month of age, again suggesting an aging process (Fig. 2B). The SnoNm/m mice also showed a markedly reduced hair regrowth activities (Fig. 2C). Thus, the SnoN knockin mice exhibit premature aging in multiple organs.

Aged animals often exhibit reduced regenerative activities. We therefore performed muscle injury and regeneration assay in age-matched SnoN+/+ and SnoNm/m mice. Twelve-month-old mice were subjected to injury by cardiotoxin-1 injection. Satellite cells (muscle stem cells) were isolated, and their ability to differentiate into myoblasts in vitro were analyzed by the expression of myogenic markers Desmin and MyoD (Brack et al., 2008). Satellite cells from injuried SnoNm/m muscles expressed a higher level of p15, p16, and p27 than that in SnoN+/+ cells (Fig. 2D, left), but this difference was not detected in cells from uninjured mice (Fig. S1B). Consistent with this, these SnoNm/m satellite cells displayed moderately reduced proliferation (Fig. S1C). Importantly, the SnoNm/m satellite cells displayed a markedly reduced myogenic ability with much less Desmin-positive cells than the SnoN+/+ culture (Fig. 2D, right). The expression levels of Desmin and MyoD were also decreased significantly in cells differentiated from SnoNm/m satellite cells than that derived from the SnoN+/+ cells (Fig. S1D). This reduced proliferation and differentiation of the SnoNm/m muscle stem cells were only observed in aged mice but not in mice of 1- to 6-month of age (Fig. S1D). Thus, muscle stem/progenitor cells from the SnoNm/m have a decreased regenerative potential.

Consistent with this accelerated aging, an increased number of senescent cells were observed in skin sections from SnoNm/m mice as evidenced by the marked increase in the level of SA-β–Gal (Fig. 2E) and p19ARF (Fig. S1E), two markers of senescence. Similarly, kidney sections from 18-month-old SnoNm/m mice also contained more senescent and apoptotic cells (Fig. S1F and data not shown). Taken together, our analyses indicate that the SnoNm/m mice showed shortened lifespan and accelerated aging, most likely due to enhanced senescence and apoptosis.

SnoN binds to p53 directly

We have shown before that SnoNm/m MEF cells displayed elevated p53 expression and premature senescence (Pan et al., 2009). As hyperactive p53 has been shown to accelerate aging (Tyner et al., 2002), we asked whether tissues from the SnoNm/m mice may show heightened p53 expression. Indeed, p53 expression is elevated in skin (Fig. 2F) and kidney (Fig. S1G) of SnoNm/m mice when compared to age-matched WT mice by both Western blotting and immunohistochemical staining, and this increase is more pronounced in aged animals (Fig. 2F, left). In addition, these tissues also displayed elevated levels of SnoN protein (Fig. S1H), as expected owing to a lack of Smad-induced degradation. This elevated SnoN protein is likely to be responsible for the increased p53 expression in aging tissues.

In our earlier study, we have shown that high levels of SnoN can be recruited to the PML NBs to upregulate p53 expression in an Smad-independent manner (Pan et al., 2009). However, how SnoN upregulates p53 in PML NBs is not clear. We first tested the possibility that SnoN may physically interact with p53 to promote its stabilization and activation. To do this, Flag-tagged p53 was transfected into 293T cells together with HA-SnoN and isolated by immunoprecipitation with anti-Flag, and p53-bound SnoN was detected by Western blotting with anti-HA. As shown in Fig. 3A, SnoN can bind to p53 readily when overexpressed, and this interaction was not affected by the presence of the transfected PML. Consistent with this, mutant SnoN (Δ322–366) defective in binding to PML could bind to p53 as well as WT SnoN when overexpressed (Fig. 3B). In a glutathione S-transferase (GST) pull-down assay, recombinant SnoN purified from bacteria could be pulled down together with GST-p53 but not with GST alone (Fig. 3C), suggesting that the SnoN/p53 interaction is direct. To determine whether this interaction also occurs at the endogenous level, MEF cells were treated with H2O2 (to induce p53), and the SnoN/p53 interaction was examined by co-immunoprecipitation assay. As shown in Fig. 3D, endogenous p53 readily interacts with SnoN. In addition, PML could also be detected in the SnoN immune complex, as shown before (Pan et al., 2009). When the PML level was reduced by specific shRNA, p53 expression was decreased dramatically (Fig. 3D), confirming that PML is required for p53 stabilization. Blocking the proteasome-mediated degradation with MG132, a proteasome inhibitor, in these shPML-expressing cells stabilized p53 expression, and SnoN was found to complex with p53 (Fig. 3D). This result suggests that PML is mainly required for the stabilization of p53 and recruits both SnoN and p53 to the PML bodies, but is not directly involved in the SnoN/p53 interaction. In addition, binding of SnoN to PML is independent of p53. In p53−/− MEFs, SnoN colocalized with PML as efficiently as in WT cells (Fig. S2A), suggesting that binding of SnoN to PML and localization of SnoN to PML bodies are upstream of the p53/SnoN interaction.

Figure 3.

 SnoN directly binds to p53. (A) Flag-p53 and HA-SnoN were transfected together with or without His-PML into 293T cells. SnoN that associated with p53 was isolated by immunoprecipitation with anti-Flag and detected by Western blotting with anti-HA. (B) PML is not required for the SnoN/p53 interaction. Flag-tagged WT SnoN or SnoN(Δ322–366) defective in PML binding was cotransfected with HA-p53. The SnoN/p53 interaction was measured as described in (A). (C) Bacterial expressed recombinant GST-p53 or GST was incubated with recombinant SnoN (bottom) in a GST pull-down assay. GST-p53-bound SnoN was detected by Western blotting with anti-SnoN (top). The amounts of input GST-p53 and GST were shown by Coomassie staining (middle panel). (D) Endogenous SnoN and p53 interact in mouse embryo fibroblasts (MEF). Primary MEFs infected with a retrovirus expressing shPML or vector control were treated with 80 μm of H2O2 for 24 h (to induce p53) or 10 μm of MG132 or DMSO for 5 h as indicated. Endogenous p53 and PML associated with SnoN were isolated by immunoprecipitation with anti-SnoN and detected by Western blotting with anti-p53 or anti-PML. (E) Top: Schematic drawing of SnoN mutants. Bottom: Residues 255–258 in SnoN is required for binding to p53. Interaction between Flag-tagged WT or mutant SnoN with HA-p53 was analyzed as described before.

The interaction between SnoN and p53 is not unique to fibroblasts. In nonimmortalized human normal mammary epithelia cells (the 184 cells), SnoN could also bind to p53 (Fig. S2B). Again in the absence of PML, when p53 was stabilized with MG132, endogenous p53 bound to SnoN readily (Fig. S2B). Thus, SnoN can interact with p53 in both epithelial and fibroblast cells.

Mapping of p53 binding site in SnoN

We next mapped the residues or domains in SnoN that are required for binding to p53 using various SnoN mutants in a co-immunoprecipitation assay. p53 bound to the N-terminal portion SnoN (residue 1–366), but not to the C-terminal fragment (residues 367–684) (Fig. S2C). A small deletion of residues 255–258 in SnoN completely abolished the binding to p53 (Fig. 3E), but did not affect the binding of SnoN to Smads or PML (Fig. S2D,E) or the ability of SnoN to repress TGFβ-induced transcription (Fig. S2F). This suggests that residues required for binding to p53 did not overlap with those for binding to PML and Smads.

Binding of SnoN to p53 is required for stabilization of p53

Previously, we have shown that high levels of SnoN upregulate p53 expression, and this is critical for SnoN-induced senescence. To determine whether binding of p53 to SnoN causes stabilization of p53, a pulse-chase assay was performed to compare the half-life of SnoN-bound p53 with that of free p53. As reported before, in the absence of stress signals, p53 has a half-life shorter than 30 min (Fig. 4A). In contrast, the half-life of SnoN-bound p53 was significantly lengthened to at least 90 min. Overexpression of PML together with p53 and SnoN did not further stabilize p53. In the same cells, the half-life of SnoN was not affected by the expression of p53 or p53 plus PML (Fig. S3A). Consistent with the requirement of SnoN-p53 binding in p53 stabilization, the SnoN mutant (Δ255–258) defective in binding to p53 failed to stabilize p53 (Fig. 4B).

Figure 4.

 SnoN induces stabilization and activation of p53. (A) Pulse-chase assay. 293T cells transfected with Flag-p53 with or without HA-SnoN or HA-SnoN plus His-PML were pulse-labeled with 35S-express and chased for the indicated time. 35S-labeled p53 was isolated by immunoprecipiation and detected by autoradiography. SnoN-bound p53 showed a significant longer half-life. (B) Stabilization of p53 requires a direct SnoN/p53 interaction. Flag-p53 was cotransfected with WT or SnoN Δ255–258. The levels of p53 expression were measured by Western blotting with anti-Flag. (C) SnoN can compete with Mdm2 for binding to p53. 293T cells were transfected with a fixed amount of HA-p53 and Flag-Mdm2 and increasing amounts of Flag-SnoN or fixing amount of Flag-SnoN and increasing amount of Flag-Mdm2 as indicated. Mdm2 and SnoN that associated with p53 were isolated by immunoprecipitation with anti-HA (p53) and detected by Western blotting with anti-Flag. (D) Recombinant SnoN competed with Mdm2 for binding to GST-p53 in vitro. GST-p53 immobilized on glutathione sepharose was incubated with fixing amounts of recombinant Mdm2 and increasing amounts of recombinant SnoN. Mdm2 or SnoN that were bound to p53 was analyzed by Western blotting with anti-Mdm2 or anti-SnoN. The input amounts of recombinant SnoN and Mdm2 were visualized by Coomassie stain. (E) SnoN and Mdm2 may bind to the same residues in p53. Flag-tagged WT or mutant p53 (L14Q/F19S) was cotransfected with HA-SnoN into 293T cells, and SnoN/p53 interaction was analyzed by co-immunoprecipitation assay. (F) SnoN enhances p53 acetylation and phosphorylation. p53 bound to SnoN was isolated from primary mouse embryo fibroblasts (MEF) cells expressing Flag-SnoN by immunoprecipitation with anti-Flag, and free p53 was isolated from SnoN-depleted lysates by immunoprecipitation with anti-p53. Acetylation and phosphorylation of these p53 populations were examined by Western blotting with anti-acetyl-p53 (K373; K382) or anti-phosph-p53 (Ser15). (G) Acetylation or phosphorylation of p53 is not required for its interaction with SnoN. Flag-SnoN was transfected with WT or mutant p53 (8R and S15A). p53 that associated with SnoN was isolated by immunoprecipitation with anti-Flag and detected by Western blotting with anti-p53.

SnoN competes with Mdm2 for binding to p53

The half-life of p53 is tightly controlled by the Mdm2 E3 ubiquitin ligase. Dissociation of Mdm2 from p53 is essential for the stabilization and activation of p53 in response to cellular stress. To determine how SnoN binding stabilizes p53, we compared the association of Mdm2 with p53 in the presence or absence of SnoN. When cells were transfected with a fixed amount of Mdm2, p53 and increasing amounts of SnoN, Mdm2 was gradually competed away from p53 by the increased amount of SnoN (Fig. 4C). In the reverse experiment with a fixed amount of SnoN and increasing amounts of Mdm2, the SnoN-p53 complex was not affected by Mdm2 (Fig. 4C). This suggests that SnoN has a higher affinity for p53 than Mdm2. This is further confirmed in an in vitro binding experiment using bacterially expressed recombinant GST-p53, Mdm2, and SnoN. When GST-p53 and recombinant Mdm2 were incubated with increased amount of recombinant SnoN, the SnoN-p53 complex increased while the Mdm2-p53 complex was significantly decreased (Fig. 4D). Similarly, in the reverse experiment with a fixed amount of SnoN and increased amounts of Mdm2, Mdm2 had no effect on the SnoN-p53 complex (Fig. S3B).

SnoN may compete with Mdm2 for binding to p53 through two possible mechanisms. First, the SnoN binding site in p53 may overlap with that of Mdm2. To test this, a p53 mutant, L14Q/F19S, defective in binding to Mdm2 but still retaining the transcription activity was examined for its ability to bind to SnoN. Interestingly, this p53 mutant also failed to bind to SnoN (Fig. 4E), suggesting that SnoN may bind to the same site in p53 as Mdm2.

SnoN may also regulate the post-translational modification of p53, for example, acetylation or phosphorylation, which is known to reduce the affinity of p53 for Mdm2. To test this, SnoN-bound p53 was isolated from cells cotransfected with Flag-SnoN with anti-Flag, and free-p53 isolated from the SnoN-depleted cells with anti-p53. The acetylation and phosphorylation (Ser 15) of the SnoN-bound p53 were compared with that of free p53. As shown in Fig. 4F, SnoN-bound p53 clearly showed elevated acetylation and phosphorylation, suggesting that SnoN may also facilitate p53 modification. Although SnoN-bound p53 is acetylated and phoshporylated, these modifications are not required for binding of p53 to SnoN. The p53 (8R) mutant lacking all known acetylation Lysine residues and p53 (S15A) mutant both bound to SnoN to the same extent as WT p53 (Fig. 4G). Thus, SnoN may also bind to p53 to promote its acetylation and phosphorylation.

SnoN competes with Mdm2 for binding to the p53 responsive promoter DNA

Mdm2 not only regulates p53 degradation but also represses its transcriptional activity at the promoter DNA (Kruse & Gu, 2009). As SnoN competes with Mdm2 for binding to p53, it may also release the transcriptional repression by Mdm2 at the p53 responsive promoters. To test this, we examined recruitment of SnoN to the p53 responsive element in the p21CIP1 promoter by chromatin immunoprecipitation (ChIP) assay. The p21CIP1 promoter contains two p53 responsive elements located at −2.9 kb (RE1) and −1.9 kb (RE2) upstream of the transcription initiation site (el-Deiry et al., 1995). It also contains a TGFβ/Smad responsive element (SBE) at approximately −2.1 kb position (Seoane et al., 2004) (Fig. 5A). We examined binding of SnoN and Mdm2 to the RE1 region and also to the SBE region as a control. SnoN was found to bind to RE1, but only upon treatment of cells with H2O2 that induces p53 expression (Fig. 5A, right). This suggests that binding of SnoN to RE1 may be p53-dependent. In contrast, binding of SnoN to the SBE region was detected both in the presence and in the absence of p53 (Fig. 5A, right), confirming that this binding is p53 independent. To further confirm the p53-dependent nature of SnoN binding to RE1, the SnoN mutant (Δ255–258) defective in binding to p53 failed to bind to RE1 (Fig. 5B). Moreover, in p53−/− cells, even WT SnoN failed to be detected at the RE1 site (Fig. 5B). Thus, SnoN is recruited to the p53 responsive element through its association with p53.

Figure 5.

 SnoN is recruited to the p53 binding element in the p21 promoter. (A) Top: Schematic drawing of the p53 binding element, RE1 and Smad binding site, SBE in the p21 promoter. The positions of the predicted PCR fragments in the chromatin immunoprecipitation (ChIP) assay were shown. Bottom: Binding of SnoN and p53 to RE1 region is induced by oxidative stress. SnoN (right panels) or p53 (left) bound to the RE1 region was examined by the CHIP assay with anti-SnoN or anti-p53, respectively, in primary mouse embryo fibroblasts (MEF) cells before or after treatment with H2O2. The amount of RE1 fragment in total cell lysate was examined by PCR using the same set of primers as Input. A parallel ChIP assay with anti-IgG was used as a negative control. Lower right panel: SnoN bound to SBE in primary MEF cells before or after treatment with H2O2 was analyzed by ChIP using anti-SnoN. (B) Recruitment of SnoN to p53 responsive element (RE1) is dependent on its interaction with p53. Flag-tagged WT or SnoNΔ255-258 was transfected into WT or p53−/− MEF. The amount of Flag-SnoN bound to RE1 was analyzed by ChIP with anti-Flag. (C) SnoN prevents binding of Mdm2 to the p53 responsive element in the p21 promoter. Flag-Mdm2 were cotranfected into primary MEF with or without HA-SnoN. Binding of Mdm2 or SnoN to the RE1 region was analyzed by the CHIP assay using anti-Flag or anti-HA, respectively. (D) SnoN activates p53-dependent transcription by antagonizing Mdm2. p53−/− MEF was transfected with p21-luc together with p53, SnoN and/or Mdm2. Luciferase activities were measured 48 h after transfection. The expression levels of p53 in the reactions were detected by Western blotting.

To determine whether SnoN could release Mdm2 from the promoter DNA, binding of Flag-Mdm2 to RE1 was examined in the absence or in the presence of HA-SnoN. In the absence of SnoN, Mdm2 can be found to bind to RE1 (Fig. 5C). However, this binding was dramatically decreased when SnoN was present and bound to RE1 (Fig. 5C). This indicates that SnoN can displace Mdm2 from the promoter DNA of p53 responsive genes.

SnoN activates p53-dependent transcription

We next examined the effects of SnoN on p53-dependent transcription using a p21 promoter-driven luciferase reporter assay in p53−/− MEF cells. Expression of WT p53 significantly induced p21 transcription (Fig. 5D). As p21 promoter is also induced by TGFβ/Smad pathway through the SBE site, and SnoN can potentially repress p21 transcription through antagonizing the Smads, thus complicating the interpretation of our assay, the mSnoN mutant defective in Smad binding was employed for this experiment. Expression of mSnoN alone had no effect on p21 transcription in p53−/− cells. However, p53-dependent transcription was clearly enhanced by co-expression of mSnoN (Fig. 5D). As reported previously, Mdm2 inhibited p53-dependent transcription. Consistent with the ability of SnoN to block the Mdm2/p53 interaction, mSnoN could significantly rescue p53-dependent transcription in the presence of Mdm2 (Fig. 5D).

The SnoN/p53 interaction is required for premature senescence

We have shown before that high levels of SnoN induced premature senescence of MEF in a p53-dependent manner (Pan et al., 2009). To determine whether the interaction of SnoN and p53 is required for this premature senescence, SnoN Δ255–258 defective in binding to p53 was introduced into primary MEF cells at P3. When expressed at similar level as WT SnoN, SnoN Δ255–258 failed to stabilize p53 and did not induce senescence (Fig. 6A). This SnoN still binds to PML, indicating that interaction of SnoN with PML is necessary but not sufficient for activation of p53 and induction of senescence and that interaction of SnoN with p53 is critical for this response.

Figure 6.

 Interaction between SnoN and p53 is required for SnoN-induced senescence and inhibition of oncogene transformation. (A) Primary mouse embryo fibroblasts (MEF) cells at P3 were infected with retroviruses expressing Flag-tagged WT or mutant SnoN. Senescence of cells was measured by SA-β-gal staining. The levels of p53 and SnoN in these cells were analyzed by Western blotting. (B) The antitumorigenic activity of SnoN is dependent on the SnoN/p53 interaction. WT or p53−/− MEF cells at P3 were infected by retroviruses expressing the constitutively active Ras followed by infection again with retroviruses expressing WT or mutant SnoN or shSnoN. Doubly infected cells were subjected to the soft-agar assay, and the colonies were visualized and quantified. The levels of SnoN, p53, and Ras were detected by Western blotting. (C) p53−/− MEF cells at P3 were infected with retroviruses expressing WT or mutant SnoN and subjected to the soft-agar assay. (D) The two-step skin carcinogenesis protocol was applied to SnoNm/mp53+/−, SnoNm/m, p53+/−, and SnoN+/+ mice. The incidence (D), average number (E) and sizes (F) of tumors developed within the 30-week window were recorded and shown in the graphs (n = 10 mice for each group). (G) Tumors from the SnoNm/m mice displayed elevated p53. The expression of p53 and SnoN in tumor samples was assessed by Western blotting with anti-p53 or anti-SnoN. (H) Senescence is detected in tumors derived from the SnoNm/m mice. Tumors of 2 mm in diameter were collected from various mouse lines and stained for SA-β-gal. (I) The defective hair regrowth activity in SnoNm/m mice was rescued in 12-month-old SnoNm/mp53+/− mice. The shaved area is indicated by the red dotted line. Pictures were taken 30 days after shaving. Mean hair regrowth was scored with a transparent grid and show in the graph (n = 3). Data are presented as mean ± SD. (J) SnoNm/mp53+/− mice displayed a markedly increased bone thickness. Cross-sections of tibias from 12-month-old SnoNm/mp53+/− (left) and SnoNm/m (right) mice.

The anti-oncogenic activity of SnoN is dependent on its interaction with p53

Next, we determined whether the SnoN/p53 interaction is required for the anti-oncogenic function of SnoN in a MEF transformation assay. We have shown previously that high levels of SnoN inhibited oncogene-induced transformation of MEF, acting as a tumor suppressor (Pan et al., 2009). However, in the absence of p53, SnoN could function as an oncogene to induce transformation by antagonizing the growth inhibitory activity of TGFβ. To determine whether elimination of the SnoN/p53 interaction disrupts the anti-oncogenic activity of SnoN to allow its oncogenic activity to be manifested, we introduced SnoN Δ255–258 into WT MEF cells together with the active Ras. The control experiment confirmed that p53 activation was abolished when the interaction between SnoN and p53 or PML was disrupted (Fig. 6B, right). While WT SnoN could not induce transformation of MEF cells in conjunction with the active Ras, SnoN Δ255–258 readily transformed MEF together with Ras (Fig. 6B, left). Moreover, overexpression of SnoN Δ255–258 alone in p53−/− MEF was sufficient to transform these cells (Fig. 6C). Thus, disrupting the SnoN/p53 interaction converts SnoN into an oncogene, and this suggests that interaction of SnoN with p53 is critical for its anti-oncogenic function in cells.

p53 is required for the antitumorigenic and progeria activity of SnoN in vivo

We have shown before that SnoNm/m mice are resistant to chemical-induced carcinogenesis, due possibly to p53-dependent senescence triggered by high levels of the SnoN protein (Pan et al., 2009). If elevated p53 in the SnoNm/m mice is indeed responsible for the resistance to tumorigenesis, reducing p53 level in these mice should restore tumor growth. We therefore crossed the SnoNm/m mice with p53−/− mice to generate SnoNm/mp53+/− mice that expressed reduced level of p53. Mice were administered with one dose of DMBA followed by twice weekly treatment of TPA for 30 weeks, and development of papillomas was monitored. As shown before, the SnoNm/m mice displayed a greatly reduced ability to form papillomas when compared to SnoN+/+ or p53+/− mice (Fig. 6D–F). Interestingly, loss of one allele of p53 completely restored tumor growth. The SnoNm/mp53+/− mice started to develop tumors after 8-week treatment, and papilloma development was observed in all SnoNm/mp53+/− mice, similar to what was observed for SnoN+/+ or p53+/− mice (Fig. 6D,E). More importantly, unlike that from the SnoNm/m mice, tumors from the SnoNm/mp53+/− mice continued to grow to >10 mm in diameter (Fig. 6F). When tumors harvested from these mice were examined for the presence of senescence responses, no senescent signals could be detected, in clear contrast from those from SnoNm/m mice that exhibited significant numbers of senescent cells (Figs 6H and S4). Consistent with these, a high level of p53 was only detected in tumors from the SnoNm/m mice but not in those from SnoNm/mp53+/− or SnoN+/+ mice (Fig. 6G). These results indicate that p53 is required for the anti-oncogenic activity of SnoN in vivo. Interestingly, delection of one copy of p53 also rescued some of the premature aging phenotypes. The defective hair regrowth activity, reduced cortical bone thickness and trabecular bone structure as well as decreased body mass found in the SnoNm/m mice were all significantly rescued in the SnoNm/mp53+/− mice (Fig. 6I, J and data not shown). Thus, the ability of SnoN to activate p53 is a major contributor to its anti-oncogenic and progeria activity.

Discussion

We have established SnoN as a direct activator of p53 by competing with Mdm2 for binding to p53 both off and on promoter DNA, leading to p53 stabilization and activation. This ability of SnoN to activate p53 and induce cell senescence is not only responsible for its anti-tumorigenic activity, but also results in accelerated aging. Knockin mice expressing a mutant form of SnoN that accumulates to a high level are resistant to chemical carcinogen-induced tumorigenesis, more sensitive to environmental stress and display premature aging in a p53-dependent manner, consistent with the model that SnoN exerts its effects on tumor suppression and aging through activating p53. Thus, in adult cells, SnoN is not just an important negative regulator of TGFβ signaling, but plays a much broader role in co-ordinate the cellular stress responses by activating p53 to regulate cell cycle arrest, senescence, and apoptosis.

SnoN likely induces p53 activation in the PML NBs. PML NBs contain many enzymes responsible for p53 modification and threfore are considered a regulation center for p53 in various stress responses. It is required for p53 stabilization and activation (Pearson & Pelicci, 2001; Dellaire et al., 2006). Although not required for the direct SnoN/p53 interaction, PML NBs likely provide a scaffold to increase the local concentrations of SnoN and p53 to facilitate their interaction and for further post-transcriptional modifications and activation of p53 (Bernardi & Pandolfi, 2007). Once SnoN-bound p53 is fully activated, this SnoN-p53 complex is then released from the PML NBs and localize to the promoter regions of p53 target genes, where they can activate transcription.

Consistent with SnoN activtion of p53, the accelerated aging phenotype found in the SnoNm/m mice is very similar to that observed in mice expressing a hyperactive p53, including a similarly shortened longevity, osteoporosis, reduced subcutaneous adipose tissue, decrease in hair regrowth and wound healing, and reduced tolerance to stresses (Tyner et al., 2002; Maier et al., 2004). It is worth noting that several mouse strains expressing an extra copy of WT p53 do not display premature aging phenotypes (Garcia-Cao et al., 2002), suggesting that p53 activation is crucial for the aging activity. The apparent progeria phenotypes observed in the SnoNm/m mice therefore support our model that SnoN induces not only p53 stabilization but also p53 activation. Furthermore, elimination of one copy of p53 from the SnoNm/m mice abolished the cancer resistance and many premature aging phenotypes, again indicating that these phenotypes are dependent on the activity of p53. It should be noted that although one p53 allele was detected in the SnoNm/m p53+/− mice, the p53 protein level was not readily detected in tumors from these mice, suggesting that the remaining p53 allele might be inactivated in these tumors. This inactivation of the remaining p53 allele could contribute to the complete lack of cancer resistance in the SnoNm/m p53+/− mice.

Unlike the SnoNm/m mice, the SnoN−/− mice did not display most of the aging phenotypes, suggesting that the closely related Ski protein did not compensate for the lack of SnoN. In addition, as the SnoN−/− mice also display elevated Smad signaling, the lack of aging phenotypes indicates that the elevated Smad activity is not a major contributor to aging. This is further supported by in vitro senescence assays where reducing Smad2/3 had no effect on premature senescence of SnoNm/m MEF (Pan et al., 2009). However, we cannot completely rule out the possibility that Smad signaling may contribute to some of the phenotypes. p53 has been shown previously to partner with Smad2 to form transcription complexes on DNA, and this interaction is required for Smad-dependent developmental events in Xenopus embryos (Cordenonsi et al., 2003; Atfi & Baron, 2008). Although this cross talk has not been shown to affect p53 activity, we cannot exclude the possibility that some of the observed aging phenotypes of the SnoNm/m mice are the result of coordinated actions of active p53 and elevated Smad signaling.

Taken together, our study showed that SnoN may serve as an important anticancer defence mechanism at times of prolonged cellular stress and tissue injury by integrating various cellular stress signals to the p53 antitumorigenic pathway. These stress responses protect the organism against cancer but do so at the price of accelerated aging. An important future question is whether SnoN is also involved in the human aging process. Although no report implicating SnoN or its mutations in human aging syndromes has been published, given our result linking SnoN with oxidative stress-induced p53 activation, it is likely that SnoN may play a role in the natural human aging process.

Experimental procedures

Immunoprecipitation and Western blotting

Immunoprecipitation and Western blotting were carried out as described previously (Zhu et al., 2007). For pulse-chase assays, cells were pulsed with 0.4 mCi ml−1 35S-express (Roche Applied Science, Indianapolis, IN, USA) for 30 min and chased as described previously (Stroschein et al., 1999).

GST pull-down

About 1.5 μg GST-p53 immobilized on the glutathione Sepharose were incubated with 2 μg recombinant SnoN for 1 h at 4 °C. For competition experiments, 1.5 μg immobilized GST-p53 was incubated with indicated amounts of recombinant SnoN or Mdm2. p53-bound proteins were eluted by boiling in SDS and detected by Western blotting.

SA-β-gal staining and Immunofluorescence

The senescence detection kit (EMD Millipore, Billerica, MA, USA) was employed to stain the senescent cells and 6-μm-thick frozen tissue sections as described (Pan et al., 2009).

Immunofluorescence was carried out as described in Data S1

For BrdU assay, cells were incubated with medium containing 0.1 mg ml−1 Bromodeoxyuridine (5-bromo-2′-deoxyuridine,BrdU) (Roche Applied Science, Indianapolis, IN, USA) for 6 h and stained with anti-BrdU working solution (Roche Applied Science) followd by Alexa Fluor 488-conjugated goat anti-mouse-IgG (Invitrogen, Carlsbad, CA, USA).

Chromatin immunoprecipitation assay

Transcription factor-associated DNA fragments were isolated by ChIP as described previously (Zhu et al., 2005) using the primer pairs described in Data S1.

Skin carcinogenesis assay and Soft-agar assay

The two-step skin carcinogenesis assay and soft-agar assay were carried out as described previously (Pan et al., 2009).

Analysis of aging phenotypes

The SnoNm/m and SnoN+/+ littermates were monitored weekly for gross differences in appearance, weighted monthly from 0 to 18 months and various tissues collected at 4–18 months of age for histology analysis. For the hair regrowth assay, the lower dorsal surfaces were shaved, and the rate of hair regrowth was assessed 1 month later. Muscle injury and regeneration assay and in vitro differentiation of satellite cells were performed as described previously (Carlson et al., 2008). More details can be found in Data S1.

Statistical analysis

Paired and unpaired Student’s t-test and Mann–Whitney U-test were used for statistical analysis. Quantitative data are presented as the mean ± SEM. Values of P ≤ 0.05 were considered statistically significant.

Acknowledgments

We thank Drs. Astar Winoto, Xinbin Chen and Wei Gu for various p53 mutants and Mdm2 constructs; Dr. Sunita Ho, Sabra Djomehri and Mel Abulencia at UCSF for X-rays and microCT services; Dr. David Raulet for p53−/− mice; Drs. Hitoshi Nishimura and Dragana Cado for generating the SnoNm/m mice. We are grateful to Dr. Judy Campisi for discussions and helpful suggestions. This work was supported by NIH RO1 CA101891 and R01 DK090347 to K. Luo. Part of the work was also supported by grants CRIM RN1-00532 and NIH/NIA AG027252 to Irina Conboy. The authors declare no conflict of interest.

Author contributions

K.L designed research; D.P., Q.Z.and M.C. performed research; I.C. contributed new reagents/analytic tools; D.P., Q.Z. and K.L. analyzed data; and D.P., Q.Z. and K.L. wrote the paper.

Ancillary