A p53/ARF-Dependent Anticancer Barrier Activates Senescence and Blocks Tumorigenesis without Impacting Apoptosis

Oncogene activation in otherwise normal cells can trigger the induction of apoptosis and senescence (1). Apoptosis is widely accepted to be one of the most critical safety mechanisms used by cells to protect against unbridled proliferation and malignant transformation (2). In accordance with it being a barrier to cancer, apoptosis usually subsides as precancerous lesions progress to full malignancy (3, 4). Furthermore, forced downregulation of genes that activate apoptosis accelerates the progression to cancer (reviewed in ref. 2).

Senescence is also detected in precancerous lesions in humans and animal models (5), and has also been reported to function as a physiologic barrier to the development of tumors of the hematopoietic system, lung, prostate, and skin (6–10). However, senescence has also been detected in some tumors, and senescent cells within a cancer have been found to aid tumor progression via releasing senescence-associated factors that instigate neighboring tumor cells (11–14). Therefore, it remains controversial whether senescence actually imposes a significant barrier to tumorigenesis.

We have reported mouse models of sporadic breast cancer by using intraductal injection of retrovirus to deliver oncogene into a small subset of mammary epithelial cells with an intact mammary gland (15–17). These models more closely mimic human breast cancer initiation than conventional transgenic and knockout (KO) models (18). Using retrovirus to introduce the gene encoding an activated version of ErbB2, a member of the EGFR family of tyrosine kinases that is commonly overactivated in human breast cancers, we detected robust apoptosis and senescence in the resulting early lesions (19, 20). As these lesions progress to frank tumors, apoptosis diminishes, but senescence remains resilient (19). Here, we use this mouse model to examine the role of senescence as a barrier to mammary tumorigenesis and show that loss of p53–ARF-dependent senescence is associated with rapid tumor induction by ErbB2 despite the erection of a robust apoptotic response, suggesting that senescence is, indeed, a critical barrier to tumor formation in the mammary gland.

The animal protocol used in this study was approved by the Institutional Animal Care and Use Committee of Baylor College of Medicine (Houston, TX). ARF-null mice (21) on B6.129 background were acquired from the NCI Mouse Repository (strain number 01XG7), then back-crossed two to three generations to FVB–wild-type (WT) mice before interbreeding for ARF-WT, -heterozygous, and -null females for experiments. Mouse genotype was determined using primers and PCR settings published by the NCI (http://mouse.ncifcrf.gov/available_details.asp?ID=01XG7). p16-null mice (22) on FVB.129 background were acquired from the NCI Mouse Repository (strain number 01XE4), then mated once to FVB-WT mice before generating p16-WT, -heterozygous, and -null females for experiments. Mouse genotype was determined using primers and PCR settings published by the NCI (http://mouse.ncifcrf.gov/available_details.asp?ID=01XE4). Tyrosinase-tagged p53-null mice (23) were graciously provided by Dr. Lawrence A. Donehower, Baylor College of Medicine (Houston, TX). Mice were crossed three generations to FVB-WT females to generate p53-WT and -null females for experiments. Mouse genotype was determined by color and confirmed with PCR, as described previously (23). All animals were euthanized according to NIH guidelines.

Constitutively activated ErbB2 (caErbB2) oncogene used to induce premalignant lesions and tumors was carried by either the lentiviral vector FUCGW (24) or the avian leucosis virus-derived RCAS (replication-competent ASLV long-terminal repeat with a Splice acceptor; ref. 25). Virus was prepared as described previously (15, 24). Viral particles were concentrated by ultracentrifugation at 27,000 rpm for 90 minutes, and then stored at −80°C until titration or intraductal injection. Viral titer was determined via limiting dilution transduction of either 293T (for lentivirus) or DF1 (for RCAS virus) cells. To preserve titer, intraductal injection was performed within 3 hours of thawing virus.

Virgin female mice ages 12- to 14-weeks-old (for ARF WT/KO and p16 WT/KO) or 6- to 10-weeks-old (for p53 WT/KO) were intraductally injected with 10⁶ IU of virus harboring the caErbB2 oncogene. Mammary glands were collected 2 to 3 weeks after injection, embedded in paraffin or frozen tissue matrix (OCT), and sectioned for analysis. Uninjected mammary glands were used as controls.

Staining and image capture were performed on 3-μm formalin-fixed and paraffin-embedded sections as described previously (15, 20). Primary antibodies used included mouse monoclonal antibodies against hemagglutinin tag (HA; MMS-101P, Covance; 1:500), p21 (cat#sc-6246 Santa Cruz Biotechnology; 1:200), gamma-H2AX (cat# 05-636 Millipore; 1:500), and pATM (cat#200-301-500 Rockland; 1:200), as well as rabbit polyclonal antibodies against cleaved caspase-3 (CC3, cat#9661S Cell Signaling Technology; 1:200), Ki67 (cat#NCL-Ki67P Novocastra; 1:200), phospho-histone 3 (pH3; cat#06-570 Millipore; 1:200), macroH2A (cat#ab37264 Abcam; 1:150), p16 (cat#sc-1207 Santa Cruz Biotechnology; 1:200), and p53 (cat#NCL-p53-CM5p Novacastra; 1:1,000). Cell counting was achieved using ImageJ software as well as Adobe Photoshop.

Formalin-fixed and paraffin-embedded 3-μm sections were prepared and stained using the ApopTag Red In Situ Apoptosis Detection Kit (cat#S7165; Chemicon) according to the manufacturer's instructions.

Mammary glands bearing premalignant lesions embedded in OCT were sectioned at 10 μm and stained for senescence-associated beta-galactosidase (β-galactosidase) activity as described previously (26–28). Briefly, frozen sections were fixed with glutaraldehyde, treated with X-gal, incubated for 12 to 16 hours until color developed, counterstained with hematoxylin, and then mounted.

Virgin female mice ages 12- to 14-weeks-old were intraductally injected into left and right #4 mammary glands with 10⁶ IU of RCAS virus harboring GFP. Mammary epithelial cells were isolated from injected glands 4 days after injection and analyzed for GFP by flow cytometry to determine the percentage of infected mammary epithelial cells. Uninjected mammary glands were used as GFP-negative controls.

Mice that were WT or KO for ARF were intraductally injected with 10⁶ IU of virus harboring the caErbB2 oncogene. Mammary glands were collected 2 weeks later, neutral-red stained (29), and wholemounted. Images of the wholemounted gland were captured using the Leica MZ16 F stereomicroscope (Leica) and Leica DFC300 FX Digital Color Camera (Leica). The number of lesions of diameters 200 to 299 μm, 300 to 399 μm, 400 to 499 μm, 500 to 599 μm, and ≥600 μm were quantified using ImageJ.

Virgin female mice ages 12- to 14-weeks-old were intraductally injected in one #4 mammary gland with 10⁶ IU of RCAS virus harboring the caErbB2 oncogene and monitored for tumor incidence by palpation of the mammary gland at least twice a week. Tumor latency was determined by recording the number of days after injection at which tumor was first palpable. Tumor growth was monitored by taking caliper and/or palpation measurements of up to three dimensions. Tumor volume was calculated using the formula 4/3 × Pi × (x/2)(y/2)(z/2), where x, y, and z are the three measured dimensions. The uninjected contralateral #4 mammary gland was used as control.

We have reported that intraductal injection of retrovirus carrying the gene encoding caErbB2, led to early lesions with elevated levels of pATM, γH2AX, and other markers of an active DNA damage response (DDR) pathway (19). Genetic ablation of ATM, with an accompanying decrease in p53 levels, diminished both apoptosis and senescence in these premalignant lesions (19), suggesting that an intact DDR (perhaps ATM–p53, specifically) is necessary for robust induction of both apoptosis and senescence.

To determine the effect of p53 loss on apoptosis and senescence, we generated premalignant lesions in WT and p53-null mice (Fig. 1A; ref. 23) by intraductally injecting female virgin mammary glands with a lentiviral vector (FUCGW) harboring caErbB2 (FUCGW-caErbB2; ref. 17). Injected mammary glands were collected 2 to 3 weeks after injection. The resultant premalignant lesions were evaluated for senescence by measuring levels of macroH2A, a histone variant that is enriched in senescence-associated heterochromatic foci (30), as well as expression of senescence-mediator p16 (31). We observed that loss of p53 led to a severe decrease in senescence response compared with WT (Fig. 1B and C), suggesting that p53 is required for senescence following ErbB2 activation in the mammary gland.

We next determined the effect of p53 on the apoptotic response via TUNEL (terminal deoxynucleotidyl transferase dUTP Nick-end labeling) detection of DNA fragmentation characteristic of dying cells and via presence of CC3, a critical player in programmed cell death. Surprisingly, loss of p53 did not impair the apoptotic response in premalignant lesions (Fig. 1D and E), which is contrary to the expected role of p53 in mediating apoptosis in early stages of tumorigenesis.

Like ATM, ARF is a key upstream regulator of p53 (reviewed in ref. 23). Although ARF has been reported to regulate senescence as well as apoptosis via p53 (32), its role in regulating apoptosis and senescence as well tumorigenesis in the mammary gland has not been rigorously tested (10, 33). We have reported that ARF is activated in mammary early lesions induced by caErbB2 (19). To determine whether the ARF tumor suppressor can mediate apoptosis and senescence following oncogene activation, we again generated premalignant lesions in WT and ARF-null mice (21) using FUCGW-caErbB2 (Fig. 2A), confirmed diminished p53 activity by assessing p21 levels (Supplementary Fig. S1A), and then evaluated lesions for senescence and apoptosis. As in p53-null mice, we observed a significant impairment of the senescence response in premalignant lesions arising in the ARF-null mammary epithelium (Fig. 2B–D) based on staining for macroH2A, SA-β-gal, and p16. Interestingly, there was no impairment of the induction of apoptosis (Fig. 2E and F). This impairment of senescence with preservation of the apoptosis response was also observed when caErbB2 was delivered via RCAS retrovirus (RCAS-caErB2; Supplementary Fig. S1A–S1E; ref. 15). Therefore, these data suggest that ARF, like p53, is required for full induction of senescence in early lesions but is dispensable for the apoptosis response. This phenocopying of p53 by ARF, given the existing knowledge of ARF as a p53 regulator (34), suggests that any role of ARF in mediating a barrier to mammary tumorigenesis is likely and primarily via p53 signaling.

Of note, although p16 has also been reported to be an alternative mediator of the senescence response (9, 22, 35, 36), we found that caErbB2-initiated premalignant lesions in p16-null mice (22) did not exhibit a diminished senescence response following oncogene activation (Supplementary Fig. S2), suggesting that pathways converging upon p53 (such as the ATM–p53 and ARF–p53 axes), rather than those regulating p16, play more critical roles in mediating the senescence response in the mammary cells that have suffered an oncogenic mutation.

Next, we determined whether senescence is truly a barrier to tumorigenesis by examining the effect of senescence loss on premalignant lesion advancement and tumor latency. Historically, this study has been difficult to conduct because loss of senescence and loss of apoptosis often occur concurrently, confounding interpretation of the results. However, we have presented above two mouse models in which, at least in the mammary gland, senescence is severely diminished but apoptosis is completely preserved following oncogene activation. The early lethality experienced by p53-null mice (due to the early and frequent development of lymphoma and other nonmammary cancers; ref. 37) precluded the use of these mice in long-term studies. Therefore, we elected to use the ARF-null model, which recapitulates the effect of p53 loss on apoptosis and senescence, but has a longer lifespan amenable to long-term tumor latency studies.

To determine whether loss of senescence is associated with failure of the mammary epithelium to eliminate premalignant lesions, we first quantified the premalignant lesion load following oncogene activation. After confirming that both mammary gland development and RCAS viral infection rates were comparable between WT and ARF-null mammary glands (Supplementary Fig. S3), we quantified lesion load of the entire wholemounted (Fig. 3A) mammary gland following RCAS-mediated caErbB2 delivery, and observed a significant increase in number of both total (>200 μm; Fig. 3B) and advanced (>600 μm; Fig. 3C) premalignant lesions in ARF-null mammary glands compared with WT. To ensure that the observed increase in premalignant lesion load was due to primarily to impairment of senescence, as opposed to loss of other functions of ARF, we examined the effect of ARF loss on proliferation rates and induction of autophagy, both of which have been reported to be at least partially regulated by ARF (38, 39). We found that both proliferation rates and levels of autophagy were comparable between WT and ARF-null early lesions (Supplementary Fig. S4). Taken together, these data suggest that ARF regulates a senescence-mediated barrier that functions to (i) prevent oncogene-activated mammary epithelial cells from forming premalignant lesions and (ii) subsequently impair the progression of these early lesions to advanced lesions.

To determine whether loss of senescence equates to the loss of a critical tumor barrier in the mammary gland, we carried out a tumor study in which mammary epithelia of ARF-null and WT mice were infected with RCAS-caErbB2 and then palpated for tumors. We found that loss of ARF led to significantly decreased tumor latency (Fig. 4A), strongly suggesting that oncogene-induced senescence truly inhibits tumorigenesis independently of apoptosis. ARF-null tumors also grew more rapidly (Fig. 4B). Together, these findings suggest that senescence plays a critical role as a barrier to tumor initiation and growth.

Previous studies have identified that the DDR pathways as critical for mediating apoptosis and senescence following oncogene activation (19, 40–45). A large volume of data point to the p53 tumor suppressor as an important mediator of apoptosis, cell-cycle arrest, and senescence under a variety of cellular circumstances (reveiwed in refs. 38–41). However, it has been recently reported (44), and we have confirmed (Supplementary Fig. S5A), that complete loss of p53 does not perturb the oncogene-induced DNA damage complex formation and upstream signaling. In this report, we show that p53 is critical for at least the senescence response to oncogene activation, and that the ARF–p53 axis likely works in concert with the ATM–p53 axis to execute a robust senescence response and inhibit tumorigenesis. In support of this, we find that loss of ARF leads to a compensatory increase in the DDR (Supplementary Fig. S4D), which is diminished in frank tumors (Supplementary Fig. S6B). These findings are in agreement with those of Evangelou and colleagues (45) and Gupta and colleagues (44), both of which suggest that the DDR and the ARF pathway interact cooperatively to erect barriers to tumorigenesis. However, in contrast with the suggestion by Gupta and colleagues (44) that the ARF pathway is required as a tumor barrier only in the context of impaired DDR, our studies indicate that ARF is a necessary tumor suppressor even in the context of an intact (and even elevated) DDR. On the basis of the observation that ARF is induced at a higher threshold of oncogenic stress than is the DDR (45), we hypothesize that impaired DDR leads to a more rapid accumulation of DNA damage and replicative stress, thereby triggering an early robust induction of ARF. It would be interesting to determine whether oncogene-activated DDR-WT cells are able to induce ARF to a similar degree as DDR-impaired cells, albeit in a delayed manner.

Our observation that p53-null lesions exhibit robust apoptosis is heavily contrary to conventional wisdom that places p53 as a central mediator of apoptosis, oncogene-induced or otherwise (46–49). It is possible that oncoprotein-induced cellular stresses (such as replicative and metabolic stress) can access multiple p53-independent pathways that culminate in apoptosis. On the basis of our findings, it seems that the senescence barrier, more than the apoptotic barrier, relies primarily on the ARF–p53 axis for full induction. Because of the abundance of evidence identifying p53 as a pivotal and multifaceted tumor suppressor, it would certainly be of keen interest to clearly delineate additional processes by which p53 plays its most decisive tumor-suppressive roles.

Our data revealed that loss of ARF led to an increased premalignant lesion load and more rapid tumor induction. These findings are in agreement with the documented role of ARF as tumor suppressor since the early 1990s (21, 50, 51). Previous attempts to determine the role of ARF in the mammary gland in vivo have yielded limited conclusions, presumably because ARF-null mice succumb to lymphoma before the incidence of mammary tumors using transgenic models (50, 52). Intraductally delivering an activated oncogene to the mammary epithelium allows us to generate premalignant lesions and tumors before the development of lymphoma in ARF-null mice. Incidentally, though it is not the primary aim of the study to focus on ARF, the set of data we have presented is the first to prove, using a completely in vivo model, that ARF functions as a bona fide mammary tumor suppressor.

Although ARF has been reported to regulate other processes in the cell, including autophagy and proliferation via ribosome biogenesis (34), the only difference we have been able to detect between WT and ARF-null mice is in the induction of senescence following oncogene activation, leading us to propose that it truly is this senescence difference that is responsible for the ultimate ARF-null phenotype of increased lesion load and shortened tumor latency. Furthermore, we found that senescence in WT tumors is reduced to levels comparable with that in ARF-null tumors (Supplementary Fig. S6A), suggesting that senescence in WT mice must have been disabled over the course of tumor progression, in line with the role of senescence as a barrier to tumorigenesis.

Mammary glands from ARF-null virgin mice have been reported to be similar to those of WT mice with the exception of possibly increased dilation of primary ducts and enhanced tertiary branches (53). Our wholemount staining confirmed similar ductal trees between ARF-null versus WT mice (Supplementary Fig. S3). It has also been reported that ductal proliferation in virgin mice was unaltered by ARF loss (53). Nevertheless, we cannot exclude the possibility that ARF loss skewed the mammary gland cell fate, resulting into a more susceptible state to transformation by ErbB2. ARF loss-induced reduction in cellular senescence (Fig. 2) and increase in transplantation potential of mammary epithelial cells (53) suggest an expanded stem cell population in ARF-null mammary glands. However, the mammary cell subtype that is most susceptible to tumor induction by ErbB2 seems to be committed luminal cells expressing the alveolar cell marker whey acidic protein (20, 54) or luminal progenitor cells (55), but not the less differentiated cells expressing keratin 6 (56) or cells with active Wnt signaling (57).

Our in vivo evidence that the induction of senescence at the premalignant lesion stage can greatly inhibit progression of premalignant mammary lesions has significant implications for cancer chemoprevention; our observations suggest that chemoprevention administered to patients bearing premalignant lesions may be more effective in halting or delaying tumor formation if both apoptosis and senescence are efficiently induced. However, because of increasing reports that within an established cancer senescence instigate tumor progression via a protumor secretory mechanism (12), it is essential to elucidate why senescent cells within an established cancer are tumor supportive whereas senescent cells within a premalignant lesion are tumor suppressive.

In summary, p53 and ARF are required for a robust senescence response in breast cells that have suffered an oncogenic mutation while they are unexpectedly dispensable to an apoptosis response. Furthermore, the ARF gene function provides a critical barrier to mammary tumorigenesis, most likely via induction of a robust senescence response following oncogene activation. In addition, our data strongly suggest that weakening of the senescence barrier at the premalignant stage promotes tumor formation.

No potential conflicts of interest were disclosed.

Conception and design: V.C. Sinha, Y. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V.C. Sinha, L. Qin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.C. Sinha, Y. Li

Writing, review, and/or revision of the manuscript: V.C. Sinha, Y. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Li

Study supervision: Y. Li

The authors thank Dr. Lawrence Donehower for advice in experimental design and data interpretation and for the p53 KO mouse line, and Drs. Jeffrey Rosen, Hua Dai, Jie Dong, and Svasti Haricharan for advice in experimental design and data interpretation.

This work was supported in part by funds from the Cancer Prevention Research Institute of Texas RP101499 (to V.C. Sinha); NIH CA124820 (to Y. Li) and U54CA149196 (to Y. Li; PI, Stephan Wong); from CDMRP BC085050 (to Y. Li) and BC073703 (to Y. Li); and from the Nancy Owens Memorial Foundation (to Y. Li); as well as by the resources from the Dan L. Duncan Cancer Center (P30CA125123) and the Sue and Lester Breast Center (P50-CA058183).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.