|Home | About | Journals | Submit | Contact Us | Français|
Studies on the role of TP53 mutation in breast cancer response to chemotherapy are conflicting. Here, we show that, contrary to dogma, MMTV-Wnt1 mammary tumors with mutant p53 exhibited a superior clinical response compared to tumors with wild-type p53. Doxorubicin-treated p53-mutant tumors failed to arrest proliferation leading to abnormal mitoses and cell death, while p53 wild-type tumors arrested, avoiding mitotic catastrophe. Senescent tumor cells persisted, secreting senescence-associated cytokines that exhibited autocrine/paracrine activity and mitogenic potential. Wild-type p53 still mediated arrest and inhibited drug response even in the context of a heterozygous p53 point mutation or absence of p21. Thus, we show wild-type p53 activity hinders chemotherapy response and demonstrate the need to reassess the paradigm for p53 in cancer therapy.
The tumor suppressor TP53 is mutated or inactivated in the majority of cancers (Soussi and Lozano, 2005). p53 exerts its effects by binding specific promoter sequences following cellular stress and activating transcription of genes involved in cell cycle arrest, senescence and apoptosis (Riley et al., 2008). DNA damage, such as that induced by radiation or chemotherapy drugs, is a potent activator of p53.
Classic studies using mouse models have demonstrated an in-vivo role for p53 in the induction of apoptosis following DNA damage (Jackson et al., 2011). Thymocytes from p53 null mice do not undergo apoptosis after radiation (Clarke et al., 1993; Lowe et al., 1993b), and the embryonic neural tube in p53 null mice is similarly resistant (Lang et al., 2004; Liu et al., 2004). p53 also contributes to response after exposure to DNA damaging drugs by inducing apoptosis in E1A/Ras transformed mouse embryo fibroblasts (Lowe et al., 1994; Lowe et al., 1993a).
Interestingly, studies examining the paradigm that wild-type p53 activity improves drug response are lacking in tumors arising from epithelial tissues (Brown and Attardi, 2005). This deficiency is exemplified in breast cancer. Some reports on the response of breast cancer with or without TP53 mutation to chemotherapy drugs have been inconclusive (Bonnefoi et al., 2011; Makris et al., 1995; Mathieu et al., 1995), while others show wild-type TP53 activity is beneficial to response (Aas et al., 1996; Berns et al., 2000; Kroger et al., 2006; Rahko et al., 2003). Intriguingly, still other reports have shown that TP53 mutant tumors respond better (Bertheau et al., 2002; Bertheau et al., 2007; Mathieu et al., 1995). It is unclear why these reports reached different conclusions, and, importantly, the notion that wild-type TP53 activity would hinder response, given its known apoptotic and tumor suppressive functions, is controversial and lacks a mechanistic explanation. Thus, it is desirable to have a controlled setting to examine if wild-type p53 activity is beneficial in breast cancer response, and if not, as some have suggested (Bertheau et al., 2008), then why?
In order to address the role of the p53 response in breast cancer, we bred MMTV-Wnt1 (Tsukamoto et al., 1988) mice into p53 wild-type as well as p53−/− (Jacks et al., 1994) and p53R172H (Lang et al., 2004) (“p53 mutant” herein) backgrounds. p53R172H is a structurally defective mutant that cannot bind DNA, and mice that are homozygous or heterozygous for this mutation are identical to p53−/− or p53+/− mice, respectively, in survival (Lang et al., 2004) and for results presented here. After mammary tumors formed, mice were treated with doxorubicin and monitored. We found that, despite heterogeneity of responses (Figure S1A), tumors from mice in the wild-type p53 background generally underwent minimal tumor regression, stabilized for several days, and quickly relapsed (Figure 1A, E).
We next examined p53 mutant MMTV-Wnt1 mammary tumors. In most human tumors, biallelic point mutations or deletions of p53 are relatively rare. More typically, a p53 point mutation is acquired in a single allele, and the other allele is subsequently retained or lost. Thus, we treated MMTV-Wnt1p53 heterozygous mutant mice (p53R172H/+ genotype) bearing tumors, of which approximately 40% undergo loss of heterozygosity (LOH) for the wild-type p53 allele (Donehower et al., 1995). Surprisingly, we found that tumors in p53R172H/+ MMTV-Wnt1 mice that lost the wild-type allele (rendering the tumor p53R172H/0 and functionally null) showed greater tumor regression and longer time to relapse (Figure 1B, E, Figure S1B).
Also to our surprise, we found that as long as tumors from p53R172H/+ mice retained the wild-type p53 allele, they exhibited a response typical of p53 wild-type tumors (Figure 1C, Figure S1C), despite the presence of a mutant allele with reported dominant negative activity (Lang et al., 2004; Willis et al., 2004). We also observed dramatic tumor volume reductions in MMTV-Wnt1 mammary tumors from treated p53 homozygous mutant mice (Figure 1D, Figure S1D), although these mice were typically harvested within 1 week following the final doxorubicin treatment (Figure 1D) due to gastrointestinal syndrome (Komarova et al., 2004). Use of the p53 homozygous mutant mice was also limited by the difficulty acquiring females in the cohort (Sah et al., 1995). In sum, although tumors of all genotypes eventually relapsed, treated tumors lacking all p53 activity, whether homozygous null or heterozygous mutant with LOH, had significantly greater decrease in tumor volume and significantly longer time to relapse compared to tumors retaining p53 activity (Figure 1E).
To assess the acute biochemical response to treatment, we examined p53 function in MMTV-Wnt1 tumors 24 hr after the final doxorubicin dose. p53 target genes p21, Puma and Ccng1 were induced 4 to 6-fold compared to untreated tumors (Figure 1F), demonstrating retention of wild-type p53 in MMTV-Wnt1 tumors from p53+/+ mice, as others have shown (Donehower et al., 1995). Predictably, p53 homozygous mutant tumors did not induce target genes (Figure 1F). Noxa levels, however, were elevated in some tumors in a p53-independent fashion. p53 heterozygous mutant MMTV-Wnt1 tumors that retained the wild-type allele also induced p53 target genes, in contrast to tumors with p53 LOH (Figure 1G).
To address the mechanism of response in p53 wild-type versus mutant tumors, we examined growth arrest and apoptosis in tumors 24 hr after doxorubicin treatment. Examining proliferation, we found that untreated tumors of all genotypes were highly positive for Ki67, a marker of cells outside G0 of the cell cycle. Following treatment, only p53 wild-type MMTV-Wnt1 tumors had large areas that were Ki67 negative or sparsely positive, demonstrating cell cycle exit, while p53 mutant tumors remained positive for Ki67 (Figure 2A). Further, tumors from parallel orthotopic transplants of a p53 wild-type MMTV-Wnt1 tumor ceased incorporating BrdU following treatment, while transplanted p53 mutant tumors continued to enter S-phase (Figure 2B). This failure to arrest in the presence of DNA damage resulted in aberrant mitoses, as evidenced by anaphase bridges, in treated p53 mutant tumors, while p53 wild-type tumors arrested and thus were not mitotic (Figure 2C, D, E). The p53 mutant MMTV-Wnt1 tumors also showed a significant increase in both cleaved caspase 3 and TUNEL positive cells after doxorubicin treatment (Figure 2F). p53 wild-type tumors, however, did not undergo apoptosis following treatment (Figure 2F). Taken together, our data show that despite induction of apoptotic genes such as Puma and Noxa, growth arrest, not apoptosis, was acutely induced in p53 wild-type tumors following doxorubicin treatment, and lack of arrest in mutant tumors resulted in aberrant mitoses, cell death and, ultimately, a superior clinical response.
To further address the mechanism responsible for the suboptimal response in p53 wild-type tumors, we harvested p53 wild-type or heterozygous MMTV-Wnt1 tumors 5–7 days post treatment (Figure S2A–C) and examined them for markers of senescence. We quantitatively determined that many markers of senescence (p21, Dcr2, p16, p15, Dec1 and PML) (Collado et al., 2005; te Poele et al., 2002) were elevated in many stably arrested MMTV-Wnt1 tumors that were p53 wild-type or heterozygous with retention of the wild-type allele, when compared to untreated tumors (Figure 3A, B, D). Treated and harvested p53 mutant tumors (Figure S2C) exhibited heterogeneity in expression levels of several senescence markers, although only Pml was consistently elevated to a degree similar to that observed in p53 wild-type tumors 5 days following treatment (Figure 3C). Dcr2 was induced slightly following treatment in p53R172H/0 tumors, possibly attributable to stromal cells within the tumor that still have the p53 wild-type allele. We used orthotopic transplants of p53 wild-type and mutant MMTV-Wnt1 tumors to further validate these findings. Parallel transplants of a p53 wild-type tumor showed increased expression of senescence genes p21 and Dcr2 in cohorts that were treated and harvested 2 or 5 days following the final treatment compared to untreated or relapsed tumors, while induction of Dec1 and Pml showed more variation among individual tumors (Figure 3E). A transplanted p53 mutant tumor did not express markers of senescence following treatment at levels near the p53 wild-type tumor, although, a low-level induction of p21 was observed, possibly attributable to p53 wild-type stromal cells from the recipient (Figure 3F). We next examined senescence associated β–galactosidase (SAβGal) staining in untreated versus treated MMTV-Wnt1 tumor transplants. We found clearly positive SAβGal staining in gross tumor specimens from treated p53 wild-type or p53R172H/+ tumors, but not untreated tumors or transplanted p53R172H/0 tumors that were treated (Figure S2D). Likewise, histological sections of only the treated p53 wild-type or p53R172H/+ tumors, but not p53R172H/0 tumors, were positive for SAβGal (Figure 3G and data not shown). In sum, we found that 3/3 different p53 wild-type tumors and 3/3 different p53R172H/+ tumor transplants were positive for SAβGal only after doxorubicin treatment, while histological sections of transplants from 2 different p53R172H/0 tumors were negative, and gross specimens showed only faint staining in <10% of the tumor surface area. Together with the analysis of mRNA markers in Figure 3A–C, our data show transplanted and spontaneous tumors that retain a wild-type p53 allele undergo cellular senescence following doxorubicin treatment.
p21 is a major mediator of p53 dependent cell cycle arrest and senescence, and was strongly induced in the arrested p53 wild-type tumors. Previous studies show that cells deficient for p21 fail to arrest following DNA damage and fail to undergo senescence (Brown et al., 1997; Brugarolas et al., 1995; Bunz et al., 1998; Chang et al., 1999; Deng et al., 1995; Waldman et al., 1995), similar to the treated p53 mutant MMTV-Wnt1 tumors in this study. Thus, we hypothesized that p21-null MMTV-Wnt1 tumors would respond similarly to p53 mutant tumors. To our surprise, we found that the p21-null MMTV-Wnt1 tumors exhibited on average a response intermediate between p53 wild-type and mutant tumors (Figure 4A, Figure S3A), with a significant subset showing a muted response similar to wild-type tumors (Figure 4A, upper chart) in addition to tumors that responded like p53 mutant tumors (Figure 4A, lower chart). However, all p21-null tumors examined were Ki67 positive (Figure 4B), suggesting these tumors failed to arrest in G0. To further examine the p21-null phenotypes, we used the amenable system of parallel orthotopic transplants of p21-null MMTV-Wnt1 tumors. We found examples of p21-null MMTV-Wnt1 tumors that arrested following treatment, ceased incorporating BrdU and halted mitotic activity (transplant #1, Figure 4C), as well as p21-null tumor transplants that failed to arrest, continued to incorporate BrdU, and underwent aberrant mitoses following treatment (Figure 4D, Figure S3B–C). In order to expand our examination of the arrest phenotype in p21 null MMTV-Wnt1 tumors beyond these parallel transplants, we quantitated mitotic activity in 10 spontaneous p21-null MMTV-Wnt1 tumors 24 hours following treatment to assess the fraction of these tumors that undergo arrest following treatment, compared to p53 wild-type and mutant tumors. We found that 6/10 tumors were essentially in mitotic arrest 24 hr following treatment, similar to 5/5 p53 wild-type treated tumors (Figure 4E). However, mitotic activity persisted in 4/10 treated tumors at a level similar to untreated tumors (Figure 4E). p53 mutant MMTV-Wnt1 tumors failed to arrest in all cases examined, as mitotic figures were abundant in all 8 treated spontaneous tumors. These data show p53 can mediate arrest in the absence of p21 in a subset of tumors, providing a plausible explanation for the observation that some p21-null tumors respond like p53 wild-type tumors that arrest, and some like mutant tumors that fail to arrest.
The above results show that some p21-null MMTV-Wnt1 tumors arrested mitotic activity and cease incorporating BrdU after treatment, but all treated tumors examined were Ki67 positive (Figure 4B). This suggests tumor cells were arrested somewhere in the cell cycle other than in G0. Indeed, FACS analysis revealed that orthotopic transplants of the BrdU negative p21-null tumor in Figure 4C arrested with 4n DNA content after treatment, consistent with a G2 arrest (Figure 5B) while the p21/p53 wild-type transplanted MMTV-Wnt1 tumor responded to treatment by a predominantly G1 arrest (Figure 5A). DNA content histograms of proliferating, relapsed tumors resembled those of proliferating untreated tumors (Figure 5A, B).
To address the mechanism of the G2 arrest in the two types of p21-null responses, we examined G2 regulators. The p21-null transplanted tumor that ceased DNA synthesis and mitosis (Figure 4C), had reduced levels of G2 regulators such as Cyclin B, Cdc2 and Stathmin1 (Figure 5C, upper chart, transplant #1). In contrast, p21-null tumors that were BrdU positive after treatment and had aberrant mitoses (Figure 4D) failed to downregulate genes that promote transition through mitosis (Figure 5C, middle chart, transplant #2, Figure S4), similar to a transplanted p53 mutant tumor (Figure 5C, lower chart). Thus, p53, in the absence of the cyclin dependent kinase inhibitor p21, can still direct a G2 cell cycle arrest in treated tumors, preventing the mitotic catastrophes associated with the superior response of p53 mutant tumors. While we were able to perform the extensive analysis using parallel orthotopic transplants of p21 null tumors in only a limited number of tumors, our finding that even the p21 null spontaneous tumors that lacked mitotic figures remained Ki67 positive, suggests these tumors likewise arrested outside of G0.
To further address the reason for early relapse in the arrested, senescent, p53 wild-type MMTV-Wnt1 tumors, we examined an array of cytokines and their receptors and cofactors that others have shown to be produced by normal cells made senescent after oncogenic stress (Acosta et al., 2008; Coppe et al., 2008; Kuilman et al., 2008; Wajapeyee et al., 2008). We found that our stably arrested, senescent MMTV-Wnt1 tumors, either p53 wild-type or heterozygous with retention of the wild-type allele (from Figure 3A, B) expressed elevated levels of a cytokine signaling network that included ligands and receptors (Figure 6A, B). Most tumors with mutant p53 did not have significantly elevated levels of cytokines following treatment, with the exception of Rantes and Tnfα, although some approached significance (Figure 6C).
The senescent, cytokine producing cells of the treated p53 wild-type tumors did not undergo apoptosis and the tumors did not lose significant volume as p53 mutant tumors (Figures 1–2). Therefore, we next tested whether the cytokines produced by the persistent senescent tumor cells could be detected in the sera of treated mice. Interestingly, we found no changes in serum levels of cytokines in treated p53 wild-type mice with and without MMTV-Wnt1 tumors at various stages before and following treatment (Figure S5A–D). This led us to test whether the expressed cytokines in treated tumors might be acting in a paracrine/autocrine manner. Stat transcription factors are activated by many of the cytokines elevated in treated, senescent, p53 wild-type tumors and are important mediators of cytokine signaling (Clevenger, 2004; Novakova et al.). Staining serial tumor sections, we found that untreated tumors were largely phospho-Stat3 negative, or showed only light nuclear staining, but were highly positive for the proliferation marker Ki67 (Figure S5E–F). Treated tumors, however, had regions of intense nuclear phospho-Stat3 staining that were Ki67 negative (Figure 6D and Figure S5G–H). Conversely, some regions in treated tumors were phospho-Stat3 negative, and Ki67 positive (Figure 6E). In fact, many tumors had adjacent regions in the same field of view with this inverse phospho-Stat3/Ki67 staining (Figure 6F–G; Figure S5I–J). Interestingly, we also observed small areas within treated tumors that contained a mixed population of phospho-Stat3 and Ki67 positive cells along the borders of the inversely staining areas (Figure 6F, red outline, Figure S5K–L). This suggested the possibility that cytokines secreted by senescent cells could induce proliferation in neighboring, non-senescent cells. Indeed, we found that Eotaxin and, to a lesser extent, Cxcl5 and Rantes, could induce proliferation in cultured MMTV-Wnt1 mammary tumor cells (Fig 6H, stimulation by known mitogens epidermal growth factor plus insulin is shown for comparison), suggesting a functional role for the cytokines produced by senescent cells in promoting tumor cell proliferation leading to clinical relapse.
To determine the role of p21 in the induction of the senescence associated cytokines, we examined treated, p21-null MMTV-Wnt1 tumors for expression of these genes. Treated spontaneous tumors did not show a clear induction of senescence markers, including cytokines and chemokines, when compared to untreated tumors (Figures S5M–N). It is possible that combining p21-null MMTV-Wnt1 tumors that arrest with those that continue proliferation following treatment could confound the interpretation of results in the spontaneous tumors. Therefore, we examined 3 parallel orthotopic transplants. We found that the p21-null tumor that arrested following treatment (transplant #1) induced some senescent markers and cytokines following treatment (Figure S5O–P). Of the 2 tumors that failed to arrest following treatment, we found that one failed to induce any senescent markers or cytokines, (Figure S5S–T) while the other strongly induced several senescence associated cytokines (Figure S5Q–R). The fact that some treated p21-null MMTV-Wnt1 tumors retain the ability to induce senescence associated cytokines following treatment could be a contributing factor in their early relapse.
We next investigated p53-p21 mediated response to doxorubicin in human breast cancer cells in culture. We found that MCF-7 and ZR-75.1 cells (TP53 wild-type) treated with doxorubicin were positive for SAβGal activity, and induced cytokines following treatment that led to phosphorylation of STAT3 (Figure 7A, B, C; Figure S6A, B, C). Both cell lines also induced p53 and p21, ceased incorporating BrdU and adopted a flattened morphology following doxorubicin treatment (Figure 7D, E, F, Figure S6D, E), as previously described (Jackson and Pereira-Smith, 2006). However, similar to the treated MMTV-Wnt1 tumors lacking functional p53, MCF-7 and ZR-75.1 cells with TP53 knockdown failed to arrest, adopted a rounded morphology consistent with cell death, and had fewer viable cells present following treatment than Si-control transfected, treated cells that adopted a senescent like phenotype (Figure 7D–G, Figure S6D–F). Also consistent with our results in p53 wild-type and mutant MMTV-Wnt1 tumors, both cell lines exhibited numerous aberrant mitoses, including anaphase bridges, and stark evidence of cell death such as micronuclei and condensed nuclei when treated after TP53 or p21 knockdown (Figure 7H, Figure S6G). The response of MCF-7 and ZR-75.1 cells with p21 knockdown was consistent with the subset of p21-null MMTV-Wnt1 tumors that continued to proliferate and enter mitosis following treatment. These data in breast cancer cell lines are similar to observations in other cell lines, such as the isogenic variants of the HCT-116 colon cancer cell line that lack p53 or p21 (Bunz et al., 1999; Waldman et al., 1997).
In this study, we show that a robust response of breast cancer to chemotherapy is highly dependent on the absence of p53-mediated arrest. Functional p53 activated a cell cycle arrest/senescence program, preventing mitosis in the presence of DNA strand breaks. Treated tumors with mutant p53 proceeded through the cell cycle and into aberrant mitoses. Another recent study observed that a p53 mutant xenografted cell line showed an increase in aberrant mitotic figures after chemotherapy treatment, while a p53 wild-type xenograft had increased SAβGal staining, however, outcome was not examined in this study (Varna et al., 2009).
Our data are consistent with retrospective human breast cancer studies showing that tumors with functional p53 respond worse to dose-dense doxorubicin based chemotherapy than tumors with non-functional p53 (Bertheau et al., 2002; Bertheau et al., 2007). The delivery of a high dose of DNA damaging agent is critical for this effect (Lehmann-Che et al.). In addition to differing dose regimens used in conflicting studies, none of these reports stratified responses by LOH status of TP53. Our data presented here show that determining LOH status is critical for predicting response. We found that MMTV-Wnt1p53R172H/+ tumors that retained the wild-type allele, despite the presence of the “dominant negative” point mutant (Lang et al., 2004; Willis et al., 2004), exhibited minimal tumor regression and had a transcriptional profile very similar to p53 wild-type MMTV-Wnt1 tumors, with acute expression of p53 targets and long term expression of senescence markers. This finding has important clinical implications for using TP53 status to risk-stratify patients with breast cancer, where the wild-type allele is often retained (Mazars et al., 1992). Thus, identifying a TP53 mutation without assessing LOH is a confounding factor likely responsible for the contradictory results observed in multiple studies analyzing the role of TP53 mutational status in breast cancer response (Aas et al., 1996; Berns et al., 2000; Bertheau et al., 2008; Bertheau et al., 2002; Bertheau et al., 2007; Kroger et al., 2006; Makris et al., 1995; Mathieu et al., 1995). Further, our results suggest that TP53 mutations that occur in basal-like breast cancers contribute to the relatively high rate of complete responses observed in this subset (Carey et al., 2007; Straver et al., 2010), and also imply these tumors are likely to undergo LOH. Conversely, in the luminal subtypes of breast cancer, which are mostly TP53 wild-type, p53 activity could contribute to the lower frequency of complete remissions (Carey et al., 2007; Straver et al., 2010), and in instances where a TP53 mutation is present in this subtype, the wild-type allele is likely to be retained. It will also be interesting to determine if the p53 mediated arrest phenotype we describe here is also predictive of poor chemotherapy response in other cancers.
Cells induced by doxorubicin to undergo senescence evidently persisted and likely contributed to the relapse. Our immunohistochemical analysis suggests the cytokines secreted by senescent tumors operate in an autocrine or paracrine fashion, reinforcing the dormant state (Kuilman et al., 2008). However, we also found Ki67 and phospho-Stat3 double positive regions in treated tumors that were nearing the time at which they typically relapse. Since cytokines promote the proliferation of non-senescent mammary tumor cells, this suggests neighboring, non-senescent cells within the tumor could be stimulated to proliferate by cytokines produced by senescent cells. Of note, the cytokines we tested for growth stimulation represent just 3 of many potentially mitogenic cytokines expressed in the senescent tumor cells. Also, it is not known what additive or synergistic growth effects the array of cytokines expressed by tumor cells would have within the natural environment of the tumor. Indeed, others have shown various cytokines elevated in our tumors to have pro-tumorigenic properties in different cellular contexts (Begley et al., 2008; Dhawan and Richmond, 2002; Pirianov and Colston, 2001; Takamori et al., 2000; Yang et al., 2006). Further, it is also possible that the cytokines promote other phenotypes such as metastasis and cell survival (Clevenger, 2004; Karnoub and Weinberg, 2006; Krtolica et al., 2001). The finding that some p53 mutant tumors also expressed selected cytokines following treatment underscores the importance of apoptosis in these tumors. Apoptosis likely eliminated the damaged p53 mutant tumor cells that expressed mitogenic cytokines, while cytokine producing cells in a p53 wild-type tumor persisted in a senescent state.
Our work also suggests that pharmacological inhibition of arrest and/or senescence, by inactivating p53 could improve chemotherapy response by redirecting cells toward apoptosis. However, it is important to note that complete loss of p21 in the MMTV-Wnt1 tumors was insufficient to bypass the p53-mediated arrest phenotype in a subset of tumors. p21-null MMTV-Wnt1 tumors showed a dichotomy of responses, resulting in an intermediate mean tumor regression following treatment. This is consistent with these tumors having both a “p53 wild-type like” response, where treated tumors do not enter S phase or mitosis, arresting in G2, and a “p53 mutant-like” response, where treated tumors continued through S phase and mitosis. This effect has not been observed in several in vitro studies of a p21 null cell line. Using isogenic variants of the colon cancer cell line HCT-116, Waldman et al. observed a superior response to radiation in p21-null xenografts, as compared to the parental cell line. These data, from a single xenografted cell line, are consistent with the fraction of tumors in our study that failed to arrest following treatment (Waldman et al., 1997).
In summary, we have modeled chemotherapy response in mice, showing p53 activity induces p21 dependent and independent growth arrest and cellular senescence instead of cell death, resulting in minimal tumor regression and early relapse. Bypassing senescence, via p53 deletion/mutation, initiated p53-independent cell death and improved tumor response. These results provide a compelling explanation for previous studies that showed improved patient response to anthracycline-based chemotherapy in TP53 mutant human breast tumors (Bertheau et al., 2002; Bertheau et al., 2007).
All experiments were approved by the M. D. Anderson Cancer Center Institutional Animal Care and Use Committee, Protocol ID# 079906634, and conformed to the guidelines of the United States Animal Welfare Act and the National Institutes of Health. MMTV-Wnt1, p53R172H/R172H, p53−/−, p21−/−mice have been described (Brugarolas et al., 1995; Jacks et al., 1994; Lang et al., 2004; Tsukamoto et al., 1988). Breeders were backcrossed into C57Bl6J background (Jackson Lab, Bar Harbor, Maine) until >90% C57Bl6J as determined by polymorphic allele analysis by the Research Animal Support Facility-Smithville, Genetic Services. Subsequent breedings produced MMTV-Wnt1 mice in p53 wild-type, p53R172H/R172H, p53−/−, p53R172H/+ and p21−/− backgrounds. Homozygous p53 mutant mice, particularly females, were not born at Mendelian ratio, consistent with other reports (Sah et al., 1995), and were thus difficult to acquire in the cohort. Mammary tumors formed in MMTV-Wnt1 mice in our study with median latency of 185 days. p53R172H/R172H, p53−/− and p53R172H/+ mice had median latencies of 108, 111 and 173 days, respectively. Mice were monitored frequently for tumor formation and tumors measured regularly using digital calipers [tumor volume in mm3 = (width2 × length)/2] (Bearss et al., 2000). Histologically, all of the MMTV-Wnt1 tumors were ductal carcinomas, generally of solid, cystic or mixed subtypes. p53 wild-type, p53R172H/+, p21−/− were all 68–79% mixed, while p53R172H/R172H and p53R172H/0 were 61% solid. When tumors reached a volume of ~500mm3 and were growing, 4mg/kg doxorubicin (Sigma) in PBS was injected intraperitoneally once daily for 5 consecutive days. Treatments were well tolerated in p53 wild-type and heterozygous mutant mice as well as p21−/− mice, with minimal (less than 10%) or no weight loss during or after treatment. p53 homozygous mutant mice, however, did show signs of toxicity, primarily weight loss ~4 days after treatment, likely due to GI syndrome, as previously described in p53 mutant mice (Komarova et al., 2004). This toxicity did not appear to contribute to tumor regression, as p53 heterozygous mutant mice with LOH in the tumor had no toxicity and had similar tumor regression as the homozygous mutant mice. At the defined end point for each mouse, tumors were harvested and portions were fixed in formalin for 48 hr and paraffin embedded (FFPE), and also flash-frozen for biochemical analysis. Loss of heterozygosity analysis was performed exactly as previously described (Post et al., 2010).
RNA was extracted from frozen, pulverized tumors using Trizol reagent (Invitrogen, Carlsbad, CA), subjected to DNAse treatment (Roche, Indianapolis, IN) and then reverse transcribed using a kit (GE Healthcare, Piscataway, NJ). Real time PCR using Sybr green (BioRad, Valencia, CA) was performed as previously described (Jackson and Pereira-Smith, 2006). Expression was normalized to Gapdh and verified with Rplp0. Primer sequences are available on request.
Cleaved caspase staining was performed as previously described (Post et al., 2010), and staining for phospho-Stat3 polyclonal antibody (1:100; Cell Signaling, Danvers, MA) and Ki67 (1:100; Leica, Newcastle Upon Tyne, UK, Ki67-MM1) were performed similarly. For phospho-Stat3 monoclonal, antigen retrieval was in Tris EDTA buffer, pH9. For detecting incorporation of 5-bromo-2'-deoxyuridine (BrdU), tumor bearing mice were injected with BrdU (Invitrogen, Carlsbad, CA) according to manufacturer's instructions, 24 and 4 hours before harvest, followed by standard fixation and processing. Denaturation was in 2N HCl for 90 min, followed by neutralization in 0.1M Na2B4O7, standard processing and then incubation with anti-BrdU (BD Immunocytometry Systems, San Jose, CA) at 1:40 for 1 hour (McGinley et al., 2000). Antigen detection for IHC was performed using a Vectastain kit and ABC (Vector Laboratories, Burlingame, CA). Images were acquired on a Nikon 80i microscope equipped with a Nikon DS-Fi1 color camera using the 10×/0.45 objective and Nikon Elements software. Some images were processed minimally in Adobe Photoshop only by histogram stretching and gamma adjustment. At least 4 random fields were manually counted for cleaved caspase 3 experiments. Number of positive cells was averaged for the 4 fields.
Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining was performed using the FragEL DNA fragmentation detection kit (Calbiochem, Darmstadt, Germany) and quantitated as above.
Senescence associated β-Galactosidase (SAβGal) assay was performed essentially as described (Dimri et al., 1995) for in vitro and in vivo experiments. Tumor segments ~1mm thick were fixed in 2% Formaldehyde, 0.2% Glutaraldehyde in PBS, stained for SAβGal, cut to 6μm sections and counterstained with Nuclear Fast Red (Vector labs, Burlingame, CA). These results were verified by staining of frozen sections in parallel.
Frozen pulverized tumors were lysed and separated by SDS PAGE as previously described (Pant et al., 2011). Antibodies and dilutions were: p21 (for mouse p21 detection), 1:1000, #556431, BD Pharmingen (San Jose, CA); Vinculin, 1:1000, #V-9131 (Sigma, St. Louis, MO); p21 (for human p21 detection), 1:200, SC6246, Santa Cruz Biotechnology, Inc (Santa Cruz, CA), mouse monoclonal p53, 1:200, OP09, EMD Biosciences (Darmstadt, Germany).
Two tailed student t-tests and ANOVA using Newman-Keuls post test were performed using GraphPad Prism software (La Jolla, CA).
H&E stained MMTV-Wnt1 tumor sections were scanned on a microscope at 400×, and cells in anaphase were photographed. Two different observers identified the presence or absence of bridges. For mitotic figures, ten random, 400×, high powered fields for each tumor were selected and mitotic figures were identified and counted by two different observers.
Tumors were harvested and processed as for transplants (above). After filtering, cells were washed in PBS, then fixed in 70% ethanol for at least 24 hours at −20°C. Propidium iodide staining was performed as previously described (Jackson and Pereira-Smith, 2006).
Primary MMTV-Wnt1 tumors were removed from euthanized mice, minced thoroughly with a scalpel blade, and then trypsinized for 10 minutes at 37°C. Trypsin was inactivated with DMEM plus 10% fetal calf serum, followed by passage through a 40μM filter. After PBS washing, cells were resuspended in Matrigel/PBS (BD Biosciences) at a concentration of 4×106/50μl. The 50μl solution was injected into each abdominal mammary fat pad of recipient C57Bl6 mice using a 30ga needle. Tumors were detectable at ~2 weeks, typically.
Approximately 1/3 of human breast cancers harbor mutations in the tumor suppressor gene TP53. The long held paradigm that wild-type p53 mediates apoptosis resulting in a favorable chemotherapy response is less clear in breast cancer, as many reports conflict, including some suggesting tumors harboring TP53 mutations respond more favorably. Here, we show that wild-type p53 activity is paradoxically detrimental to chemotherapy response because, unlike mutant p53 tumors, p53 wild-type tumors can avoid aberrant mitoses by undergoing arrest, which is followed by expression of cytokines in senescent cells that can stimulate cell proliferation and tumor relapse. Further, our data demonstrate that in order to accurately predict clinical response of TP53 mutated tumors, the status of the second allele must be determined.
The authors wish to thank MDACC Animal Support Facility-Smithville, the DNA Analysis Facility and the Flow Cytometry & Imaging Core Facility, each supported by NCI Grant CA16672, as well as the Department of Veterinary Medicine and Surgery and the Histology Core Research Lab. We also acknowledge Courtney Vallien for histological sectioning, Henry P. Adams for technical help with microscopy, and Archana Sidalaghatta Nagaraja for assistance with mouse necropsy and RNA preparation. JGJ was funded as an Odyssey Scholar by the Theodore N. Law Endowment for Scientific Achievement, a Dodie P. Hawn Fellowship in Genetics and by NIH grant U01DE019765-01. GL is supported by NIH grant CA34936.
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.
Author Contributions JGJ conceived, designed and performed all the experiments except where noted, and wrote the manuscript. VP and JGJ performed protein expression analysis. JGJ, QL and VP performed mRNA analysis. JGJ, LLC, QL, DG and AQC performed immunohistochemical analyses. TM performed the Bio-Plex cytokine array. JGJ, DG and AQC performed quantitative analysis of mitosis. YL developed cell lines. JGJ and OT performed senescence associated β-galactosidase studies. PY and JGJ determined loss of heterozygosity in tumors. AKE-N performed pathological analysis. AQC edited the manuscript. GL was the principal investigator for the study.