Various p53 mutant proteins differently regulate the Ras circuit to induce a cancer-related gene signature

The development of tumors is characterized by the accumulation of mutations in tumor-suppressor genes and in proto-oncogenes. Two of the most frequently mutated genes in human cancer are the gene encoding the tumor-suppressor p53 (TP53) and the proto-oncogene Ras (Bos, 1989; Hollstein et al., 1991).

p53 is a transcription factor that accumulates and is activated in response to stress signals. Following its activation, p53 induces processes such as cell cycle arrest, programmed cell death and DNA repair, thereby, guarding the cell from malignancy (Levine, 1997). In most cancer cells the p53 pathway is defective, usually because of genetic mutations within the p53 sequence (Guimaraes and Hainaut, 2002). Most of the mutations in p53 are missense mutations that reside within the p53 DNA binding domain (DBD), causing an impaired binding of p53 to the DNA. The missense mutations can be divided into two rudimentary sub-groups according to their impact on the DBD folding. The first group consists of DNA-contact mutations (e.g. R248, R273), which affect amino acids that directly interact with DNA while retaining wild-type p53 conformation. The second group includes p53 conformational mutations (e.g. R175, H179), which alter the scaffold that orients the structure of the DNA-binding interface (Cho et al., 1994). The observation that mutant p53 is accumulated in tumors, raised the possibility that mutant p53 plays an active role in the process of malignant transformation. Indeed, it is now accepted that mutant p53 exerts its activity in either a dominant-negative manner, in which it inhibits the normal activity of the wild-type p53, or by a gain-of-function mechanism, in which mutant p53 undertakes new oncogenic activities. Several mechanisms were suggested to account for mutant p53 gain of function, such as transcriptional activation (e.g. MYC, MDR1, NFκB2), transcriptional repression (e.g. ATF3, CD-95 and MST1), unique interaction with specific DNA motives (MAR), epigenetic modification and interactions with other proteins (e.g. p63, p73, NFY) (Sigal and Rotter, 2000; Li and Prives, 2007; Lozano, 2007; Weisz et al., 2007a; Brosh and Rotter, 2009; Buganim and Rotter, 2009).

Another protein frequently manipulated by cancer cells is the Ras proto-oncogene, which becomes hyperactive because of a point mutation during tumor development. There are three pivotal Ras genes in the human genome: HRAS, KRAS and NRAS. In spite of their high similarity in sequence and function, H-Ras was found to be a more potent mediator of cellular transformation in fibroblast cells, whereas K-Ras and N-Ras are more active in epithelial cells and hematopoetic cells, respectively (Maher et al., 1995). Ras activation depends on its binding to GDP (the non-active state) or GTP, which enables its active conformation (Malumbres and Barbacid, 2003). Switching to the active state is controlled by the guanine nucleotide exchange factor (RasGEF) enzyme family, which facilitates the replacement of GDP by GTP. Shifting to the non-active state is regulated by the GTPase activating protein (RasGAP) enzyme family, which facilitates GTP hydrolysis. When mutated, Ras is less sensitive to regulation by RasGAPs and therefore remains constitutively active. Over-activated Ras causes uncontrolled proliferation, survival and tumor aggressiveness by induction of signaling pathways such as MAPK, PI3K and RALGDS (Downward, 2003).

Early studies in cancer research revealed that mutant p53 and Ras oncogene cooperate to induce cellular transformation (Eliyahu et al., 1984; Parada et al., 1984). Later studies reported several processes regulated by both p53 and Ras, such as cell motility, proliferation and survival (Boiko et al., 2006; Song et al., 2007; Xia and Land, 2007; Meylan et al., 2009). However, hitherto, a mechanism underlying the cooperation between the common forms of mutant p53 and activated Ras, has not been addressed. Because p53 is commonly mutated during carcinogenesis, and mutated forms of p53 further facilitate transformation using a wide range of mechanisms (Buganim and Rotter, 2009), we decided to focus on the role of the different mutant p53 categories in the crosstalk with Ras.

Thus, the main motivation of this work was to identify the specific molecular details that underlie the crosstalk between Ras and the various mutant p53 categories.

Recently, we established an in vitro transformation model and performed a wide genomic profiling to identify clusters of genes that are related to the different steps of transformation (Milyavsky et al., 2005; Buganim et al., 2010). In this analysis we identified the cancer-related gene signature (CGS) that was of particular interest because it was synergistically upregulated when H-Ras^V12 and WTp53 inactivating peptide (GSE56) were concomitantly expressed in the cells (Milyavsky et al., 2005; Buganim et al., 2010). The CGS mainly consists of secreted molecules that were shown to have pro-cancerous functions, and at least parts of them were shown to be induced by activated H-Ras (Sternlicht et al., 1999; Dhawan and Richmond, 2002; Mendes et al., 2005; Minn et al., 2005; Wang et al., 2006; Apte and Voronov, 2008). The CGS consists of chemokines [chemokine (C-X-C motif) ligand 1, 2 and 3 (CXCL1, CXCL2 and CXCL3)], interleukins (IL-1β, IL-6 and CSF2) and extra-cellular matrix (ECM)-related proteins [matrix metallopeptidase 3 (MMP-3), CLECSF2 and TREM-1]. For further investigation, we decided to focus on representative genes of this signature, mainly CXCL1, IL1B and MMP3, that represent the different sub-groups of the CGS. In this study we show that the different p53 mutants utilize different mechanisms to induce the CGS. Whereas the p53 DNA-contact mutations cooperate with the NFκB pathway, the Zn⁺² region conformational mutants augment H-Ras activity. Notably, the L3 loop conformational mutant p53^G245S did not seem to affect the examined pathways.

These findings extend our knowledge of the mutant p53–Ras axis in cancer and might be useful for rationale-based drug design.

Previously, we reported a gene cluster, the CGS, encoding mainly pro-cancerous secreted molecules, that is synergistically upregulated by p53 inactivation and oncogenic H-Ras^V12 expression (Milyavsky et al., 2005; Buganim et al., 2010).

Because inactivation of p53 in human tumors occurs predominantly by point mutations, and the fact that the various p53 mutants exhibit a wide range of mechanisms toward malignancy (Buganim and Rotter, 2009), we decided to examine the effect of various p53 mutants on the regulation of the CGS.

To that end, immortalized primary human embryonic lung fibroblasts (WI-38 cells) overexpressing the H-Ras^V12 oncogene (Ras) or its empty control vector (Con), were stably infected with one of the following vectors: an shp53 vector [vector expressing a short hairpin RNA (shRNA) against the wild-type p53 (TP53) sequence], an shCon control vector (vector expressing an shRNA against the mouse Noxa coding sequence) or with a vector expressing one of the five most frequent p53 mutant forms (p53^R175H, p53^H179R, p53^G245S, p53^R248Q and p53^R273H; Fig. 1A). These p53 mutations represent three principal categories: conformational mutations within the Zn²⁺ region (p53^R175H, p53^H179R), conformational mutation within the L3 loop region (p53^G245S) and DNA-contact mutations (p53^R248Q, p53^R273H).

Next, we compared the expression levels of the CGS in the established cell lines using three representative genes (CXCL1, IL1B, MMP3). In agreement with our previous data, we found that cells expressing the H-Ras^V12 oncogene along with p53 knock down (hereafter referred to as Ras/shp53 cells) upregulated expression of the CGS both at the mRNA and protein levels compared with their control counterparts (Con/shCon, Ras/shCon, Con/shp53; Fig. 1B,C). Interestingly, the different p53 mutant forms had distinct expression patterns. Whereas the Zn²⁺ region conformational mutant (p53^R175H, p53^H179R) cells upregulated the CGS to a similar level as observed in cells in which p53 was knocked down, the DNA-contact mutant (p53^R248Q and p53^R273H) cells upregulated the CGS to a much higher extent. Interestingly, the L3 loop region conformational mutant (p53^G245S) cells exhibited only a minor effect on the CGS expression (Fig. 1B,C).

To further strengthen this observation, we examined the mutant-p53-dependent expression of the CGS in six human-tumor-derived cell lines that endogenously express different mutant p53 types. Accordingly, following mutant p53 knock down, the representative CGS genes (CXCL1, IL1B, IL8) were downregulated only in cells expressing the p53 DNA-contact mutations (SW-620, SW-480, NCI-H322), whereas p53 knock down hardly affected the CGS expression in cells expressing the p53 conformational mutants (NCI-H23, Hs-578-T, SKBR-3; supplementary material Fig. S1). This result supports our hypothesis that the p53 conformational mutants give rise to similar CGS levels as p53 knock down, whereas the p53 DNA-contact mutations further induce the CGS.

Altogether, these data demonstrate unique characteristics and modes of action for each mutant p53 type in regulating the CGS.

Because CGS expression was dependent on H-Ras^V12, and in order to reveal the mechanisms by which the various p53 mutants induce the CGS, we examined whether H-Ras activity is directly affected by the various p53 mutants. Using the Ras activity pull-down assay [Ras binding domain (RBD) assay] (Vojtek et al., 1993) we have recently demonstrated that wild-type p53 inhibits the activity of the H-Ras^V12 oncogene by reducing its GTP-loading state, resulting in inhibition of CGS expression (Buganim et al., 2010). To measure the activity levels of H-Ras^V12 in the various mutant-p53-expressing cells, we performed the RBD assay on established WI-38 cells. Importantly, all H-Ras^V12-expressing cells exhibited comparable levels of H-Ras mRNA and protein (Fig. 2A,B). Notably, cells expressing high CGS levels (Ras/shp53, Ras/p53^R175H) also exhibited high H-Ras–GTP levels, whereas cells expressing low CGS levels (Ras/shCon, Ras/p53^G245S) exhibited low H-Ras–GTP levels (Fig. 2B). Supporting this observation, we found that the levels of activated H-Ras^V12 were similar in Ras/p53^H179R, Ras/p53^R175H and Ras/shp53 cells (supplementary material Fig. S2). Interestingly, Ras/p53^R248Q cells, which expressed the highest levels of CGS, exhibited only moderate H-Ras–GTP levels (Fig. 2B), suggesting that the DNA-contact p53 mutants use a different mechanism to augment CGS levels. Together, these results raise two interesting issues; the first one being the mechanism by which the different p53 mutant forms regulate H-Ras^V12 activity, and the second being the additional mechanism that is responsible for the super-elevation of the CGS in p53 DNA-contact mutants. These two issues will be addressed in the following results descriptions.

Recently, we suggested that the p53 target gene, BTG2, interacts with H-Ras^V12 and mediates the p53-dependent CGS suppression (Buganim et al., 2010). Additionally, it was shown that p53 mutants act in a dominant-negative manner (Shaulian et al., 1992). In order to elucidate the mechanism by which the p53 Zn²⁺ region conformational mutants induce H-Ras^V12 activity, we examined whether these p53 mutants could reduce BTG2 expression. Surprisingly, although Ras/shp53 cells reduced the expression levels of BTG2, the five mutant p53 forms, hardly affected its expression (Fig. 3A) although they accumulated in the cells (Fig. 1A). The same phenomenon was observed when we monitored the mRNA levels of a well-known p53 target, p21^WAF (also known as CDKN1A; supplementary material Fig. S3A). To exclude the possibility that the mutant p53 constructs retain some wild-type p53 activity, we introduced the three representative mutant p53 constructs into a p53-null cell line (H1299), and measured the mRNA levels of the p53 downstream targets BTG2 and p21^WAF. As expected, and in contrast to the wild-type p53, all mutant p53 forms did not affect the mRNA levels of the p53 targets (Fig. 3B; supplementary material Fig. S3B). By contrast, when cells were treated with cisplatinum, a well-known DNA-damage and p53-stabilizing agent, the induction of the p53 targets was attenuated by the various p53 mutants, excepts of p53^G245S (supplementary material Fig. S4), indicating that the various mutant p53 forms can attenuate p53 activity by a dominant-negative mechanism following DNA damage. Together, these results suggest that, in our system, although cells express ectopic mutant p53 proteins, the transcriptional activity of the endogenous wild-type p53, under basal condition, remains mostly intact.

These observations suggest that the various mutant p53 forms upregulated the CGS expression by different mechanism rather than inhibiting the expression of BTG2 or suppressing wild-type p53 transactivation capability. Because mutational analysis of p53 revealed that the amino acids Leu22 and Trp23 are essential for transactivation (Lin et al., 1995; Roemer and Mueller-Lantzsch, 1996), we decided to mutate these amino acids (L22Q, W23S) in the Zn⁺² region conformational mutants and to measure the CGS expression. Interestingly, similar induction of CGS was observed in both the intact and in the mutated Zn⁺² region conformational mutants (Fig. 3C). This indicates that the capability of the Zn⁺² region conformational mutants to induce the CGS expression by upregulating H-Ras activity is not dependent on the transactivation activity of the p53 mutants or the ability to reduce the expression of the wild-type p53-mediated target, BTG2.

In our recent paper we showed that p53 interacts with BTG2 and ATF3 proteins (Buganim et al., 2010). This observation together with the suggestion that mutant p53 can exert its oncogenic function through protein–protein interactions (Buganim and Rotter, 2009), raised the possibility that mutant p53 might interact with BTG2, thus squelching its function. To this end, we examined a possible interaction between BTG2 and the various mutant p53 categories. By using GST–BTG2-fused beads (or GST–Empty as a control), we pulled down and measured the levels of mutant p53 proteins that were bound to BTG2 in the various mutant-p53-expressing cells extracts. Interestingly, the Zn⁺² region conformational mutants (p53^R175H, p53^H179R) that caused the highest elevation in H-Ras activity (Fig. 2B) also bound most strongly with BTG2 (Fig. 4A). The DNA-contact mutants (p53^R248Q, p53^R273H) that moderately elevated H-Ras activity exhibited a weak interaction with BTG2. As expected, the L3 loop region conformational mutant (p53^G245S) that does not affect H-Ras activity also did not interact with BTG2 (Fig. 4A). These results might suggest that by binding to BTG2, the Zn⁺² region conformational mutants squelch its activity, which accounts for the elevated H-Ras activity, and induction of the CGS.

To examine the hypothesis that the interaction of mutant p53 with BTG2 perturbs BTG2 function, we measured the ability of H-Ras to interact with BTG2 in the different mutant-p53-expressing cells. By using the GST–BTG2 beads, we pulled down and measured the protein levels of H-Ras that were bound to BTG2 in representative cells extracts. As expected, the control cells Ras/shCon and Ras/shp53 exhibited strong BTG2–H-Ras interaction. Importantly, this interaction was abolished in cells expressing the Zn⁺² region conformational mutants. Cells expressing the DNA-contact mutants that exhibited a weak interaction with BTG2 (Fig. 4A) exhibited a reduced binding of BTG2 to H-Ras, whereas cells expressing the L3 loop region conformational mutant demonstrated a similar BTG2–H-Ras interaction as the control cells (Fig. 4B). These data suggest that by binding to BTG2, the Zn⁺² region conformational mutants withdraw BTG2 from the BTG2–H-Ras equation and thus augment H-Ras activity.

Our results imply that although p53 mutants do not interfere with wild-type p53 transactivation, they can affect its targets by a squelching mechanism.

The highest expression of the CGS was observed in cells expressing H-Ras^V12 in conjunction with the p53 DNA-contact mutants (p53^R248Q and p53^R273H) and was much higher than in p53 knock-down cells. This suggests that these p53 mutants possess a gain-of-function activity in inducing the CGS expression. However, these cells express only moderate levels of H-Ras–GTP, suggesting that the mechanism of the CGS induction by these p53 mutants does not exclusively involve H-Ras activation. Therefore, we sought to elucidate the mechanism that underlies the enhanced activity of the DNA-contact mutants in respect to the CGS expression. Because both wild-type p53 and mutant p53 were shown to be involved in the NFκB pathway (Mayo et al., 1997; Dreyfus et al., 2005; Scian et al., 2005; Weisz et al., 2007b; Fontemaggi et al., 2009; Meylan et al., 2009), and because we recently demonstrated that NFκB is involved in the induction of the CGS (Buganim et al., 2010) we postulated that NFκB might be the factor that mediates the activity of the DNA-contact p53 mutants. To this end, we knocked down p65, a member of the NFκB family, in the various WI-38 cell lines and measured its effect on mutant p53-mediated CGS expression. The capacity of the siRNA to knock down p65 levels was estimated by measuring its protein levels (Fig. 5A; supplementary material Fig. S5A). Notably, downregulation of p65 resulted in a substantial reduction in the mRNA levels of the CGS independently of p53 status (Fig. 5B; supplementary material Fig. S5B). Interestingly, we found that the super-induction of the CGS exerted solely by the DNA-contact p53 mutants (p53^R248Q and p53^R273H) was totally abolished by the knock down of p65 (Fig. 5B; supplementary material Fig. S5B). To further support the above results we examined the expression of CGS in three different mutant p53 bearing cancer cell lines (SW-620, SW-480, SKBR-3) in which p53 was stably knocked down (supplementary material Fig. S5C). Notably, silencing p65 in SW-620 (supplementary material Fig. S5D), which endogenously express the DNA-contact mutant p53^R273H, reduced the expression of the representative CGS genes, and abolished their increased expression cause by mutant p53^R273H (Fig. 6A). A complementary strategy to manipulate the NFκB pathway would be to activate it by TNF-α. To this end, SW-480, which harbors the DNA-contact mutant p53^R273H, and SKBR-3, which harbors the Zn⁺² region conformational mutant p53^R175H, were treated with TNF-α, and the mRNA levels of the representative CGS genes (IL8, CXCL1, CXCL2) were examined. In agreement with the above-described results, a substantial decrease in the level of CGS was observed in SW-480 cells when mutant p53 was knocked down in basal conditions. Following TNF-α treatment, the CGS expression levels were induced. Interestingly, the fold reduction observed following mutant p53^R273H knock down, under basal conditions, were diminished. This suggests that the mutant-p53^R273H-dependent induction of CGS expression under basal conditions is mediated by NFκB. (Fig. 6B). By contrast, no major differences in the levels of CGS were observed in SKBR-3 cells following mutant p53^R175H knock down under basal condition. Following TNF-α treatment, CGS expression was induced in shp53 and shCon cells to similar levels (Fig. 6C). Similarly to the SKBR-3 cells, when we examined NCI-H23 and Hs-578-T cells, harboring p53 conformational mutants (p53^M246I and p53^V157F, respectively), no substantial differences were observed in the levels of CGS when comparing shCon and shp53, both in basal condition and following TNF-α treatment (supplementary material Fig. S6). These results indicate that NFκB is a major regulator of the CGS expression mediated by the DNA-contact p53 mutants.

Recently, by using an in vitro stepwise well-controlled transformation model (Milyavsky et al., 2003) with a detailed genomic profiling (Milyavsky et al., 2005), we were able to uncover that loss of wild-type p53 synergizes with H-Ras^V12 to induce a unique CGS (Buganim et al., 2010). This agrees with a previous report suggesting that the combination of p53 loss and Ras activation synergistically induces the expression of pro-malignant genes (McMurray et al., 2008). In our previous study, we suggested that wild-type p53 represses the expression of CGS by at least two mechanisms. On the one hand, p53 transactivates its target BTG2, which binds to H-Ras^V12 and represses its activity, and on the other hand, following p53 stabilization, the p53 target ATF3 is activated, binds to the CGS gene promoters and represses their expression (Buganim et al., 2010).

Here, we investigate the role of the different p53 mutants in the induction of the CGS during tumor development.

When mutated, p53 either acts in a dominant-negative manner, in which it inhibits wild-type p53 tumor-suppressive functions, or acts in a gain-of-function manner, in which mutant p53 shows additional oncogenic activities, supporting tumorigenesis (Brosh and Rotter, 2009). Here we examined the mechanisms utilized by the different p53 mutants, and found that the DNA-contact mutants (p53^R248Q and p53^R273H), but not the p53 conformational mutants (p53^R175H and p53^H179R), induce the CGS in a gain-of-function manner. This was demonstrated by the higher CGS expression levels observed in cells expressing the p53 DNA-contact mutants, compared with the p53 knock down and p53 conformational mutants (Fig. 1B,C).

Supporting these data, the p53 DNA-contact mutants showed higher CGS induction than p53 conformational mutants, when endogenously expressed in human-cancer-derived cell lines (supplementary material Fig. S1). These observations are especially intriguing because they suggest that not all p53 mutants classes possess gain-of-function activity in CGS induction.

Furthermore, we demonstrate that different types of mutant p53 engage in different molecular pathways to facilitate the cooperation with oncogenic H-Ras^V12 in inducing the CGS (Fig. 7). The DNA-contact mutants (p53^R248Q and p53^R273H) cooperate with the NFκB-dependent pathway to further induce the CGS. Concordantly, it was demonstrated that mutant p53 can exert a gain-of-function activity by stimulating NFκB activity (Scian et al., 2005; Weisz et al., 2007b). Interestingly, we have found that the expression of the different p53 mutants did not reduce the expression levels of wild-type p53 gene targets, including BTG2, which we recently found to mediate the p53-dependent H-Ras activity inhibition. Nevertheless, we suggest that the Zn⁺² region conformational p53 mutants (p53^R175H and p53^H179R) interact with BTG2 and disrupt its binding to H-Ras, therefore enabling higher levels of active H-Ras and CGS. This proposed mechanism is supported by the observation that damaging the transactivation capability of mutant p53, by mutating the L22, W23 codons, did not affect the expression of the CGS, and agree with other studies that demonstrated protein–protein interaction between mutant p53 and other proteins (e.g. p63, p73, Mre11, NF-Y) (Buganim and Rotter, 2009) that leads to mutant p53 gain of function. By the ability to interact with the BTG2 protein, p53 mutants could interfere with p53 pathway, and inhibit its tumor-suppressive functions.

Most of p53 mutations reside within the DNA binding domain; however, they can be divided into several categories according to their structural conformation. Therefore, the differences in the mode of action between the various mutant p53 categories could be explained by their specific characteristics. For example, mutant p53^G245S, which in our system hardly elevates the expression levels of the CGS, partially retains some of the wild-type p53 characteristics such as stabilization and DNA binding affinity (Bullock and Fersht, 2001).

The three Ras family members, K-Ras, H-Ras and N-Ras, are highly homologous (85% sequence identity), and share similar functions (Downward, 2003). Moreover, in fibroblast cells, H-Ras is a more effective mediator of cellular transformation than the other Ras oncoproteins (Maher et al., 1995). Therefore, because our in vitro transformation model is based on human primary fibroblast cells, H-Ras^V12 was used. Yet, human-cancer-derived cell lines that harbor the K-Ras oncogene along with p53 DNA-contact mutants also showed higher CGS expression then cells expressing p53 conformational mutants (supplementary material Fig. S1). Therefore, the results obtained in a fibroblast in vitro transformation model, exploiting the H-Ras oncogene, are predictive also of Κ-Ras-oncogene-derived cancer.

In conclusion, we suggest that during the transformation process, when p53 is mutated, the expression of a cancer-related gene signature is induced, and the tumor-suppressive function of p53 is overridden. The different p53 mutations can augment these phenotypes by different means; either by cooperating with the NFκB protein or by squelching the activity of BTG2.

Li-Fraumeni syndrome is a familial disorder associated with a germline mutation in one of the p53 alleles. This predisposes the patients to a high incidence of cancer (Malkin et al., 1990). Because our model represents a situation whereby both wild-type and mutant p53 are present in the cells, as in Li-Fraumeni syndrome, it is tempting to speculate that high levels of mutant p53 can overcome the tumor-suppressive effect of the wild-type p53 protein, and thus, a p53 heterozygous state is sufficient to drive the cells toward tumorigenesis. This hypothesis is in accordance with Varley et al., demonstrate that 40% of Li-Fraumeni syndrome patients did not undergo loss of heterozygosity, and gave rise to tumors that were heterozygous for p53 (Varley et al., 1997). Interestingly, it should be mentioned that our established cells expressing the various p53 mutants (except of p53^R245S) had a tumorigenic phenotype, measured by tumor formation when injected into nude mice (data not shown). Both the WI-38 cells grown in vitro and tumors generated in vivo were heterozygous for p53, suggesting that these cells utilize other mechanisms for tumorigenesis, as in 40% of Li-Fraumeni syndrome patients.

Uncovering the nature of the multiple pathways involved in this important cellular junction and the capability to distinguish between the p53 mutant categories might provide important tools for tailor-made cancer therapy that is based on p53 status and its distinct regulatory pathways.

The following primary antibodies were used: anti-p53 (DO-1 and 1801) were kindly provided by David Lane (Ninewells Hospital and Medical School, Dundee, Scotland); anti-H-Ras, anti-p65 (Santa Cruz Biotechnology, Santa Cruz, CA); anti-GAPDH (Chemicon, Billerica, MA); anti- β-tubulin (Sigma, St. Louis, MO).

The pRetroSuper-p53 shRNA-Blast, the pRetroSuper-shmNOXA-Blast and the pBabe-H-Ras^V12-Hygro constructs were kindly provided by Doron Ginsberg (Bar-Ilan University, Israel). GST-BTG2 expression plasmids were constructed by cloning the BTG2 open reading frame (ORF) into the pGEX-GP1/3 vector that was kindly provided by Elior Peles (The Weizmann Institute, Israel). The various p53 vectors were constructed by sub-cloning the wild-type p53 ORF from the PLXSN-neo-p53 construct into the PWZL-Blast construct using BamHI restriction. Site-directed mutagenesis (Santa Cruz Biotechnology), was conducted on the pWZL-blast-p53 construct according to the manufacturer's instructions using specific primers to create the specific mutations. Supplementary material Table S2 shows all primer sequences.

Dimethyl sulfoxide (DMSO) (Sigma), cisplatinum (Pharmachemie, Petach Tikva, Israel), TNF-α (R&D systems, Minneapolis, MN), GelCode Blue staining reagent (Thermo Scientific, MA, USA), Ponceau (Sigma).

The immortalized primary human embryonic lung fibroblasts (WI-38) were established and maintained as described previously (Milyavsky et al., 2003). The colon carcinoma cell lines, SW-480 and SW-620, were maintained in DMEM supplemented with 10% fetal calf serum and antibiotics. The breast adenocarcinoma cell line, SKBR-3, was maintained in McCoy's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, 1 mM sodium pyruvate and antibiotics. The non-small lung carcinoma cell lines H1299, NCI-H322, NCI-H23 and breast cell line Hs-578-T were maintained in RPMI-1640 supplemented with 10% fetal calf serum and 2 mM L-glutamine. The infection procedures using ecotropic Phoenix-packaging cells were described previously (Milyavsky et al., 2003). The transfections into H1299 cells were conducted using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Knock down of p65 was conducted by transfection with specific oligonucleotides (Dharmacon, Lafayette, CO) using DharmaFECT3 reagent (Dharmacon). The analysis was performed 72 hours after transfection.

Total RNA was isolated using a NucleoSpin kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's protocol. A 2 µg aliquot of the total RNA was reverse transcribed using Bio-RT (Bio Lab, Jerusalem, Israel) and random hexamer primers. QRT-PCR was performed with an ABI 7300 instrument (Applied Biosystems, Carlsbad, CA) using Platinum SYBR Green qPCR SuperMix (Invitrogen). Specific primers were designed for the following genes: CXCL1, IL1B, MMP3, CXCL2, IL8, BTG2, p21^WAF, H-Ras and p65. cDNA levels were normalized to GAPDH (for primer sequences see supplementary material Table S1).

Cells (2×10⁵) were grown on six-well plates with serum-free MEM for 72 hours. Cell-conditioned media were collected and centrifuged at 20,800 g, at 4°C. CXCL1 and IL-1β proteins were detected by enzyme-linked immunosorbent assay (ELISA) using the Human GRO-α/CXCL1 or IL-1β immunoassay kits (R&D systems), according to the manufacturer's instructions. For determination of MMP3 protein levels, equal volumes of cell media were separated on gelatin-containing polyacrylamide gels. Then, gels were incubated overnight in developing buffer containing 0.01% Brij (Sigma), at 37°C, to activate the MMP enzymes. Gels were then stained with Coomassie Blue solution (0.1–0.5% Coomassie Brilliant Blue R250 in 5% acetic acid and 10% methanol) to detect areas of catalytic activity.

The glutathione S-transferase (GST)–Ras binding domain (RBD; GST–RBD) and the GST–BTG2 fused beads were prepared as described previously (Frangioni and Neel, 1993). Cells were lysed in cold RBD lysis buffer. Equal amounts of cell extracts were incubated with the GST–RBD, GST–BTG2 or GST–Empty beads, overnight. The proteins were eluted with sample buffer, separated on SDS-polyacrylamide gels, and analyzed. Western blotting was conducted as described by Milyavsky et al. (Milyavsky et al., 2003).