Introduction

p27^Kip1 and p21^Cip1 are members of the Cip/Kip family of Cdk inhibitors. They share a conserved N-terminal domain that binds to and inhibits multiple cyclin/Cdk complexes (reviewed in Nakayama and Nakayama, 1998; Sherr and Roberts, 1999). Other than a nuclear localization signal, the C terminus is not conserved and likely provides specific functions to each Cip/Kip member. p27 responds to intrinsic or extracellular antimitogenic signals by binding to and inhibiting cyclin/Cdks, and thereby blocking cell cycle progression (Nourse et al., 1994; Polyak et al., 1994). p27 can also regulate apoptosis, timing of differentiation, and migration, through yet to be defined mechanisms (Philipp-Staheli et al., 2001; Coqueret, 2003; McAllister et al., 2003). In general, p27 levels are high in quiescent cells and fall during the G1 to S cell cycle transition. p27 expression is regulated at multiple levels, although current data support a major role for post-translational regulation through the ubiquitin/proteasome pathway (Pagano et al., 1995; Carrano et al., 1999). Point mutations or deletions in the coding region of the gene encoding p27 (CDKN1B) are rare in human tumors, even those that have suffered loss of heterozygosity at chromosome 12p12–12p13.1, where the CDKN1B gene is localized (Bullrich et al., 1995; Pietenpol et al., 1995; Ponce-Castaneda et al., 1995). However, p27 protein expression is frequently reduced in a remarkably wide range of human tumor types, including lymphomas and carcinomas of the breast, colon, stomach, lung, bladder, prostate, ovary, and esophagus (reviewed in Lloyd et al., 1999; Slingerland and Pagano, 2000). In most instances, low p27 expression is seen in more aggressive tumors and correlates with poor patient survival. The basis for reduced p27 expression in tumors is poorly understood. Reduced p27 levels in colon carcinoma, non-small-cell lung cancer, glioma, and mantle cell lymphoma was attributed to increased rates of proteasome-mediated protein degradation (Esposito et al., 1997; Loda et al., 1997; Piva et al., 1999; Chiarle et al., 2000), while p27 mRNA levels were reduced in gastric tumors, suggesting transcriptional control (Yasui et al., 1997). Other tumor types, such as Barrett's associated esophageal cancer, ovarian cancer, and some colon and breast tumors, have decreased nuclear p27, but retain cytoplasmic p27 (Ciaparrone et al., 1998; Singh et al., 1998; Masciullo et al., 1999; Blain and Massague, 2002). In these cases, p27 may be sequestered in the cytosol, which effectively decreases its nuclear activity. There are likely multiple mechanisms to regulate p27 levels and activity.

Experiments in p27-deficient mice have established that p27 acts as a tumor suppressor. p27-/- mice are predisposed to spontaneous pituitary adenomas (Fero et al., 1996; Kiyokawa et al., 1996; Nakayama et al., 1996), as well as radiation- or ENU-induced tumors in multiple tissues, including adenomas and adenocarcinomas of the intestine and lung, granulosa cell tumors of the ovary, and uterine tumors of several histological subtypes (Fero et al., 1998). p27 heterozygous (+/-) mice show an intermediate susceptibility to the same tumor types. Retention of the intact wild-type allele of Cdkn1b in these tumors indicates that p27 is haploinsufficient for tumor suppression.

The tumor suppressor gene p53 is among the most commonly mutated genes in human cancer, again in a wide range of tumor types, including those of the colon, stomach, breast, lung, and lymphomas (Hollstein et al., 1994). These same tumor types also show frequent loss of p27 expression, so a significant percentage of tumors are likely to have mutations in p53, together with reduced p27 expression. However, it is not known whether these alterations interact in any way to influence tumor progression or patient outcome. p53-deficient mice spontaneously develop lymphomas with >75% incidence, as well as soft-tissue sarcomas (Donehower et al., 1992; Jacks et al., 1994), while untreated p27 null mice have a predisposition to pituitary adenomas (Fero et al., 1996). This nonoverlapping spectrum permits evaluation of potential interactions between these tumor suppressors in mice lacking both gene products.

p21 is a transcriptional target of p53 and is the major effector of p53-mediated G1 cell cycle checkpoint (Nakayama and Nakayama, 1998; Sherr and Roberts, 1999). Like p27, p21 binds to and inhibits G1 cyclin/Cdks, blocking cell cycle progression. p21 also has a PCNA binding domain in the C terminus, which may participate in the coordination of DNA synthesis and repair (Waga et al., 1994). Similar to p27, p21 is also reported to modulate differentiation and apoptosis (Skapek et al., 1995; Nakayama and Nakayama, 1998; El-Diery, 2001). p21 mutations are infrequent in human tumors (Shiohara et al., 1994), but unlike p27, p21 expression has not been consistently associated with prognosis or tumor grade. p21 knockout mice are apparently developmentally normal. Although their cells have an impaired G1 cell cycle arrest in response to DNA damage, the mice show only a modest susceptibility to tumor development (Brugarolas et al., 1995; Deng et al., 1995; Martin-Caballero et al., 2001). Nevertheless, the functional similarity to p27 and its role as an effector of p53 suggest that p21 may function as a tumor suppressor.

Results

To determine if alterations in p27 and p53 cooperate during tumor development, p27 knockout mice were interbred with p53 knockout mice to generate single and compound mutant mice, which were observed for tumor development. The median latency to morbidity and/or mortality from tumor burden in the compound mutant p53-/-p27-/- mice (125 days) and p53-/-p27+/- mice (139 days) was significantly reduced relative to the single mutant p53-/- (188 days) or p27-/- mice (>280 days) (Figure 1a). No p27-/- mice developed lethal tumors prior to 230 days of age, at which time nearly all mice of other genotypes had died. Simultaneous deletion of p27 and p53 had a greater effect on decreasing tumor latency than the additive effect of deletion of each gene alone (P=0.02). A synergistic interaction between the reduction of p27 and mutation in p53 was also observed with the deletion of just a single allele of p27 (P=0.03, Figure 1a). The mice in this experiment were of a mixed genetic background (see Materials and methods). As genetic background can modify tumor spectrum and latency (Harvey et al., 1993), the experiment was repeated with all mice on a uniform C57BL/6 background. The results were in close agreement with those of the first experiment (Figure 1b). The median age to morbidity from tumor burden was reduced from 185 days in p53-/- mice to 157 days in p53-/-p27+/- mice and 128 days in p53-/- p27-/- mice. Tumor latency in p27-/- animals was much longer, with a median age to morbidity of 419 days. There was again evidence of a synergistic effect of deletions at the two genes on median age to morbidity in both the p53-/- p27-/- and p53-/- p27+/- mice (P=0.04 and 0.01, respectively).

Figure 1.

Reduction of p27 reduces tumor latency in p53-deficient mice. Kaplan–Meier analysis of tumor-free survival of p53 and p27 compound mutant mice. (a) Mice of a mixed background (C57BL/6J/C3H/NIH). (b) Mice of a uniform C57BL/6J background. The difference between survival rates of p53-/- and p53-/-p27-/- mice was highly significant in both studies (see text)

Full figure and legend (38K)

In both the experiments, T-cell lymphoma (thymic, splenic, or disseminated) was the most common tumor type and cause of morbidity in p53-/- mice regardless of p27 genotype, with an incidence ranging from 57 to 80% (Figure 2). Sarcomas (rhabdomyosarcoma, osteosarcoma, hemangiosarcoma, and poorly differentiated) were second in frequency in p53-/- mice, again irrespective of p27 genotype. In addition to lymphomas and sarcomas, p53 p27 compound mutant mice developed a significantly greater number of tumor types not seen in p53 single mutants (P<0.01). These included sertoli cell tumors, adrenal tumors, carcinomas of the stomach and kidney, a glioblastoma, a neuroblastoma, and an angioma (Figure 3). Compound mutant mice also tended to develop multiple tumors per mouse; 6/70 did so vs 1/38 single mutant p53-/- mice. Several compound mutants also showed pituitary gland enlargement, indicative of the early stage of pituitary tumor development. The most common tumor type in the single mutant p27-/- mice was pituitary adenoma followed by lung adenoma and adenocarcinoma, pheochromocytoma, and harderian gland adenoma.

Figure 2.

Tumor spectrum of p53 p27 compound mutant mice. Pie chart analysis shows the percentage of each tumor type that was observed in each genotype. Data are combined from both experiments in Figure 1. Lymphomas were primarily T cell in origin. Sarcomas were of diverse histological subtypes and included lymphosarcoma, hemangiosarcoma, osteosarcoma, rhabdomyosarcoma, and poorly differentiated sarcomas. Note the predominance of lymphomas in all p53-deficient mice regardless of p27 genotype as well as the increase in tumors of diverse histotypes in p27-deficient mice

Full figure and legend (137K)

Figure 3.

Histology of tumors from p53 p27 mutant mice. (a) Neuroblastoma from a p27+/- p53-/- mouse. Note the tumor cells (t) invading the brain (br) (arrows). (b) Sertoli cell tumor from a p27+/-p53-/- mouse. The boxed area shows cells growing in a palisading arrangement. (c) Transitional cell carcinoma from a p27-/-p53-/- mouse. Note the tumor cells (t) invading the muscle layer (m) (arrows). (d) Glioblastoma mutiforme from a p27+/-p53-/- mouse shows giant multinucleate cells (arrows) and abnormal mitotic figures. (e) Pheochromocytoma from a p27-/- mouse. (f) Poorly differentiated carcinoma from the stomach from a p27-/- p53-/- mouse shows giant multinucleated cells. Original magnification in (a) 40, (b–f) 200, and (insets) 600

Full figure and legend (275K)

Thus, on either strain background, the tumor spectrum in p53 p27 compound mice more closely resembled that of single mutant p53 null mice, rather than p27 null mice. Deletion of p27 decreased the latency to morbidity from tumors that had mutations in p53, by approximately 25%. Decreased tumor latency of p53 mutant mice was also seen in p27 heterozygotes (P=0.01 in mixed background mice, P=0.02 in C57BL/6), consistent with previous results showing that tumor growth is highly sensitive to p27 gene dosage (Fero et al., 1998; Philipp-Staheli et al., 2002). Novel tumor types and multiple tumors per animal seen in the compound mutant mice indicate that reduction of p27 reveals additional tumor predisposition of p53 mutant mice.

To determine whether p27 expression was altered during mutant p53-driven tumorigenesis, we examined the levels of protein and mRNA transcript. Normal lymphocytes within the thymus showed strong nuclear p27 immunoreactivity (Figure 4c). In contrast, p27 staining in p53-/- lymphomas was weak or undetectable (Figure 4b). Likewise, sarcomas from p53-/- mice showed weak p27 immunoreactivity, particularly in the nucleus (Figure 4e). This marked decrease in p27 in tumors was confirmed by Western blot and densitometry analysis of tumor extracts. The p27 protein levels, in both nuclear and cytoplasmic fractions from all p53-/- thymic lymphomas analysed (n=8), were <10% that in normal thymus (Figure 5a). To determine if this reduction in p27 protein was due to reduced p27 mRNA transcript, we performed real-time PCR analysis from RNA isolated from normal thymus and lymphomas. The quantitative nature of the real-time analysis was demonstrated by 50% reduction in mRNA in normal kidney samples from p27+/- compared to p27+/+ mice (data not shown). The p27 mRNA levels in p53-/- tumors (wild type for p27) were reduced by an average of 39% (range 9–65%, n=15) relative to normal thymus (Figure 5b). To determine whether p27 mRNA levels in tumors reflected Cdkn1b gene dosage, we also examined p27+/- mice. The p27 mRNA levels in tumors from p27+/- mice (n=11) were on average 40% less than in tumors from p27+/+ mice (n=15), indicating that levels of mRNA in both normal and tumor tissue closely mirror Cdkn1b gene dosage (Figure 5b). The levels of p27 protein were reduced to a greater extent than was the mRNA, indicating that p27 reduction occurs at the transcriptional, but primarily at the post-transcriptional, level. In addition, levels of cyclin E were increased in tumors (Figure 5c). These alterations translated to increased Cdk2 activity.

Figure 4.

Reduced p27 expression in tumors. (a) H&E stained section of p53-/- lymphoma. The arrow points to the mitotic figure. (b) Immunostaining for p27 in a p53-/- lymphoma. Arrows point to the tumor cell negative for nuclear p27 staining. Note the occasional p27-positive cell within the tumor. (c) Immunostaining for p27 of normal p53-/- thymus shows strong nuclear staining, especially in the medulla. (d) H&E stained section of p53-/- sarcoma. (e) Immunostaining for p27 in p53-/- sarcoma. The arrow points to the tumor cell negative for nuclear p27 staining

Full figure and legend (350K)

Figure 5.

Reduced p27 occurs primarily at the post-transcriptional level. (a) Western blot analysis of nuclear (n) and cytoplasmic (c) extracts from normal thymus and thymic lymphomas from p53-/- mice was performed using a p27 antibody. p27-/- mice were used for the negative control. (b) Real-time PCR analysis of p27 mRNA extracted from normal thymus and lymphomas from p53-/-p27+/+ and p53-/-p27+/- mice. Each symbol represents the mean value calculated from two to three independent assays from individual samples. The values are normalized to levels in the normal thymus. The average p27 levels are shown by the horizontal bars. (c) Cdk2 activity is increased in lymphomas. Top row: Cdk2 immunoprecipitation of lysates from tissues was followed by histone H1 kinase assay. Rows 2–5: Western blot analysis of extracts using antibodies against Cdk2, cyclin D, cyclin E, and p27

Full figure and legend (119K)

p21^Cip1 is structurally and functionally related to p27 and is a key transcriptional target of p53. Both p27 and p53 have demonstrated marked tumor-suppressing activity, but in different tissues and contexts. Therefore, it was of interest to determine if tumor-suppressing activity could be detected for p21. Neither p21- nor p27-deficient mice show a marked predisposition to spontaneous tumorigenesis. However, p27-deficient mice are notably tumor prone following carcinogen challenge, in particular to the carcinogen and alkylating agent ENU (Fero et al., 1998; Philipp-Staheli et al., 2002). To provide a direct comparison, we challenged a cohort of p21-/-, +/-, and wild-type littermates with ENU. Mice of all p21 genotypes developed tumors at a similar rate and frequency with a median latency between 40 and 50 weeks of age (Figure 6a). The most common lethal tumor type in wild-type and p21+/- mice was lymphoma, with an incidence of 31 and 35%, respectively (Figure 6b). p21-/- mice developed significantly fewer lymphomas, with an incidence of only 10% (P=0.06 for p21-/- vs p21+/+ and P=0.02 for p21-/- vs p21+/-). When mice that developed lymphomas were excluded from the analysis, the latency to tumor development was slightly accelerated, by 10%, in p21-/- relative to wild-type mice. The incidence of tumors of the liver and soft-tissue sarcomas was increased in p21-/- mice compared to p21+/- or wild-type mice (Figure 6b). In the cohort of 27 p21-/- mice, five developed hepatocellular adenoma, three developed myxoma of the dermis, two hemangiosarcoma, two poorly differentiated sarcomas, and one transitional cell carcinoma of the bladder. One liver tumor and four sarcomas were detected out of 29 p21+/- and 26 p21 wild-type mice. In all, 100% of all mice regardless of p21 genotype developed lung tumors, which were identified as bronchoalveolar adenomas or adenocarcinomas. There was little or no difference in the average number of lung tumors per mouse (Figure 6b), tumor size, or histological grade between p21 genotypes (data not shown). Several intestinal adenomas and adenocarcinomas were seen in both p21-deficient and wild-type mice (Figure 6b).

Figure 6.

Tumor latency and spectrum in ENU-treated p21-deficient mice. (a) Kaplan–Meier analysis of tumor-free survival of p21+/+ (n=26), p21+/- (n=29), and p21-/- (n=27) mice. (b) Incidence and multiplicity of different tumor types from p21-deficient mice. 'Other' category includes mainly sarcomas, and a transitional cell carcinoma of the bladder as described in the text

Full figure and legend (107K)

Two previous studies also showed reduced T-cell lymphomas in p21-deficient mice, one in combination with germline mutation in Atm (Wang et al., 1997) and another in response to ionizing radiation (Martin-Caballero et al., 2001). Both the studies attributed this reduction to increased levels of apoptosis in the absence of p21. Likewise, we observed a two-fold increase in apoptotic index in lymphomas from p21-/- mice relative to wild-type littermates (data not shown), confirming that loss of p21 leads to increased apoptosis. Thus, p21-deficient mice had a reduced incidence of lymphomas and an increased incidence of liver tumors and sarcomas. These contrasting effects effectively canceled each other out, as overall tumor-free survival did not differ between p21 genotypes.

Discussion

Comparison of tumor predisposition between specific mouse mutants or combinations of mutants facilitates the analysis of functional pathways and hierarchical events that contribute to neoplasia. Here, we examined the interaction of p53 and p27 as well as the role of p21 in tumor suppression. p53 and p27 are both established major tumor suppressor genes, p53 by virtue of mutational inactivation in many human cancers and p27 by virtue of its loss of expression in many human tumors and knockout mouse studies. p53 loss contributes to neoplasia in multiple ways: through loss of multiple cell cycle checkpoints, deregulated apoptosis, and increased genetic instability (Vousden and Lu, 2002). p27 is inhibitory to the G1-S transition in response to extracellular signals and loss of p27 increases the proliferation and growth rate of tumors (Philipp-Staheli et al., 2001). Thus, both proteins function in different contexts to regulate the cell cycle and neoplasia. The distinct tissue predisposition spectra of p53 (lymphomas and sarcomas) compared to p27-deficient mice (tumors of both mesenchymal and epithelial origin) highlight the differences in the mode of action of these tumor suppressors that is embedded at the level of tissue type and cellular architecture. The unique tumor phenotypes of these two models permit evaluation of any potential interaction between loss of p53 and loss of p27. The results demonstrate significant synergism between loss of p53 and p27, in that tumors in compound mutants arose faster than the combined additive effect of single mutants. This indicates that distinct functions of both proteins are used in a cooperative manner to suppress tumor growth. The tumor spectrum in the p53 p27 compound mutants more closely resembled that of the single p53 mutants. Thus, lymphomas and sarcomas were the major tumor types in p53 null mice regardless of p27 genotype. Additional tumor types developed in p53 p27 compound mutant mice, including sertoli cell tumors, pheochromocytomas, brain tumors, and carcinomas. Pituitary neoplasia, seen in p27 null mice, did not appear to be accelerated in the absence of p53. Thus, loss of p53 did not modify tumor progression in a p27-/- setting but, in contrast, deletion of p27 clearly accelerated tumor development driven by mutation in p53. This suggests a hierachical relationship, with p27 as a modifier of mutant p53-driven tumorigenesis. As discussed below, this highly cooperative nature of p27 is seen in other settings; p27 modifies the growth of tumors with mutations in a number of cancer genes.

p27 expression is frequently reduced in human lymphomas and, with the exception of mantle cell lymphomas, reduced p27 correlates with increased proliferative fraction and patient survival (Sanchez-Beato et al., 1997; Erlanson et al., 1998; Quintanilla-Martinez et al., 1998; Chiarle et al., 2000; Moller, 2000). Here, we show that p27 expression is markedly downregulated in murine T-cell lymphomas and that experimental, germline reduction of p27 in the same tumor type accelerated tumor growth leading to earlier mortality. This indicates that p27 inhibits the latency and hence mortality from tumors, notably lymphomas, which harbor p53 mutations. Extrapolation of these results across species would imply that reduced p27 expression observed in human tumors that contain p53 mutations plays a causal role in the development of those tumors.

In the limited cases where it has been examined, the reduction of p27 in human tumors occurs primarily at the post-transcriptional level (Esposito et al., 1997; Loda et al., 1997; Piva et al., 1999; Chiarle et al., 2000). p27 is also downregulated in murine tumors, providing an experimental model to address the mechanism of p27 misregulation during autochthonous tumor development. We previously showed an element of genetic control of p27 expression in tumors, in that reduced p27 was seen in intestinal tumors with Apc mutations, but not in tumors with mutations in Smad3 (Philipp-Staheli et al., 2002). Thus, a reduction in p27 depends on the oncogenic pathway that is altered in the tumor. Here, we show a consistent and significant reduction in p27 expression in tumors (lymphomas and sarcomas) with p53 mutations. Comparison of RNA and protein levels in thymic lymphomas with a normal thymus indicates that the reduction occurs primarily at the post-transcriptional level, thus ruling out mutation or epigenetic silencing of the Cdkn1b gene that encodes p27.

Tumor development in p53 null mice was accelerated by loss of only a single allele of p27, further confirmation of the exquisite dosage sensitivity of tumor suppression by p27. Both p27 mRNA and protein levels in both normal and tumor tissue from p27+/- mice were about 50% that from p27 wild-type mice (this study and Fero et al., 1998; Philipp-Staheli et al., 2002). This indicates that p27 transcript levels reflect Cdkn1b gene dosage and that the haploinsufficiency of p27 stems from reduced transcript and hence protein levels.

The two observations, that p27 expression is greatly reduced in lymphomas from p53-/- mice and germline deletion of p27 significantly decreased the latency of lymphoma development, are at first paradoxical, but suggest two possibilities. One is that the residual low expression of p27 has significant suppressing activity, and germline deletion of p27 further reduces p27 and accelerates tumor growth. The second explanation is that p27 downregulation may not occur initially, but rather at some later stage of tumor development. High levels of p27, which occur in normal lymphocytes, may remain high during the early stages of tumorigenesis, thereby retarding clonal expansion of the initiated cells. Additional genetic or epigenetic events would be required to downregulate p27 at the protein level. In this scenario, germline deletion of p27 would accelerate proliferation and/or the acquisition of other gene mutations from the outset, resulting in earlier appearance of tumors.

In contrast to potent tumor suppression by p53 and p27, evidence for tumor suppression by p21 was equivocal. The overall tumor-free survival in response to the broad-spectrum mutagen and carcinogen ENU was not affected by p21 deficiency. The incidence of lymphomas was significantly reduced in ENU-treated p21-/- mice relative to similarly treated p21+/- or wild-type mice. Lymphomas from p21-deficient mice showed increased apoptosis compared to wild-type mice, indicating that p21 deficiency leads to increased apoptosis in autochthonous tumors and a delay in tumor development. This observation is not confined to this particular experimental setting as two other studies likewise showed reduced or delayed T-cell lymphomagenesis in p21-deficient mice, one in combination with a germline mutation in Atm (Wang et al., 1997) and another in response to ionizing radiation (Martin-Caballero et al., 2001). Tissue culture studies have also indicated a role for p21 in protection from apoptosis (Waldman et al., 1996). The observations of reduced apoptosis and increased lymphomagenesis in p53-deficient mice and increased apoptosis and reduced lymphomagenesis in p21-deficient mice argue that, at least in the lymphocyte lineage, deregulated apoptosis is a critical roadblock to neoplasia.

The incidence of liver tumors and sarcomas of diverse histotypes was modestly increased in ENU-treated p21-/- mice, indicating that p21 may contribute, albeit in a relatively minor way, to tumor suppression in these tissues. Other studies have examined tumor predisposition of p21-deficient mice in response to carcinogenic challenge, or in combination with defined genetic models of neoplasia. Philipp et al. (1999) showed no increase in the number, size, or growth rate of DMBA/TPA-induced squamous cell papillomas or progression to carcinomas in p21-deficient mice, but the carcinomas that did develop were more poorly differentiated spindle cell tumors. This suggests that the cell cycle regulation function of p21 does not play a major or direct role in tumor suppression but its impact on differentiation may play a more significant role. Irradiated p21-/- mice (1.75 Gy/week for 4 weeks) lived longer and developed fewer tumors than did wild-type mice, which was attributed to enhanced apoptosis in p21-deficient tissues (Martin-Caballero et al., 2001). Recently, Jackson et al. (2003) reported no difference in tumor-free survival of irradiated p21-deficient mice (4 Gy at 2 weeks of age), but a modest increase in tumors per mouse and metastasis. Different radiation doses between the two studies could account for these differences. Both results stand in contrast to the markedly enhanced susceptibility of both p53- and p27-deficient mice to radiation-induced carcinogenesis (Kemp et al., 1994; Fero et al., 1998). Jackson et al. (2002) reported increased lung tumor incidence in response to urethane at early but not late time points. The current study shows no major effect of p21 deficiency on ENU-induced lung tumor multiplicity, size, or histological progression in mice 6–12 months of age. Thus, p21 may have a role early in pulmonary adenomagenesis that diminishes with age. Germline deletion of p21 enhanced intestinal adenoma development in Apc mutant mice (Yang et al., 2001). In the current study, using ENU, GI tumor incidence was not increased in the p21-deficient cohort relative to wild-type mice, suggesting that intestinal tumor suppression by p21 is not robust and may be limited to the Apc pathway. p21 deficiency did not accelerate tumor latency or incidence in two transgene-induced breast tumor models, MMTV-wnt-1 and MMTV-myc, but did accelerate MMTV-ras induced tumors (Jones et al., 1999; Adnane et al., 2000; Bearss et al., 2002), suggesting that tumor suppression by p21 in this tissue may be pathway specific. Taken together, these results indicate that p21 behaves more as a modifier than as a direct suppressor of tumor development (Jones et al., 1999). The apparent lack of prominent tumor predisposition in p21-deficient mice suggests that G1 cell cycle checkpoint deficiency, such as occurs in cells from these mice, may not be a major contributing factor to neoplastic development.

In summary, although p53, p27, and p21 are direct or indirect inhibitors of cell cycle progression, they clearly play different roles in tumor suppression within the intact organism. These proteins have pleiotropic functions, the relative contribution of which can vary between tissues and tumor types. An intriguing quality of p27 is evident in its broad tissue effects and its consistent synergistic interactions with an array of both tumor suppressor genes and oncogenes that are central to human cancer. Decreased p27 leads to increased tumor development in combination with mutations or overexpression of H-ras (skin cancer) (Philipp et al., 1999), Apc (intestinal adenomas) (Philipp-Staheli et al., 2002), Rb (pituitary tumors) (Park et al., 1999), myc, (lymphomas) (Martins and Berns, 2002), cyclinD/CDK inhibitor p18/INK4c (some endocrine tumors) (Franklin et al., 2000), cyclin E (lymphomas) (Geisen et al., 2003), PTEN (prostate) (Di Cristofano et al., 2001), and p53 (lymphomas and sarcomas). These effects occur in multiple tissues of both epithelial and mesenchymal origin, and in homogeneous and heterogeneous genetic backgrounds. Generally, reduction of p27 leads to an increase in the rate of appearance of tumor types driven by the primary oncogenic lesion, and this leads to an increase in tumor-associated mortality. The fact that p27 inhibits tumor development in multiple tissues involving multiple genetic pathways and in a consistent dosage-dependent manner indicates that p27 is a nodal and highly sensitive suppressor of tumorigenesis. p27 may be pervasively used by multiple and varied oncogenic pathways as the ultimate antagonist to unrestrained cellular proliferation.

Materials and methods

Mice

C57BL/6 p53 knockout mice were originally obtained from Donehower et al. (1992) and genotyped by PCR as described (Timme and Thompson, 1994). The p53 knockout allele was backcrossed >10 times into the C3H/HeJ strain. C57BL/6 p27+/- mice were crossed to C3 H p53-/- mice to generate C57BL/6 C3 H F₁ p27+/- p53+/- mice. Also, C3 H p53+/- mice were crossed to NIH p27-/- mice (Philipp et al., 1999) to generate NIH C3 H F₁ p53+/- p27+/- mice. These offspring were bred through three consecutive rounds of breeding to generate mice of the appropriate genotypes on a mixed genetic background of 50% C3H and 50% C57BL/6 for the p53-/- mice (n=21) and 40% C3 H, 40% C57BL/6, 20% NIH for the remaining genotypes, which included, p27-/- (n=18), p53-/-p27-/- (n=16), and p53-/-p27+/- (n=15). To examine the interaction of p53 and p27 on a different and uniform genetic background, C57BL/6 p53+/- mice were crossed to C57BL/6 p27+/- mice to generate p53+/- p27+/- mice. These mice were intercrossed to generate experimental mice of the following genotypes: p53-/- (n=20), p27-/- (n=12), p53-/-p27-/- (n=6) and p53-/-p27+/- (n=23). These were observed for spontaneous tumor formation.

p21 knockout mice on a mixed 129/C57BL/6 background were generously provided by P Leder (Deng et al., 1995). The p21 knockout allele was genotyped and backcrossed >10 times to the NIH strain as described (Philipp et al., 1999). Mice of all three p21 genotypes (p21+/+, +/-, and -/-) were generated from p21+/- breeders and treated at 18–20 days of age with ENU (0.5 mol/g body weight, i.p.).

All mice were observed daily and killed when moribund, which included the following criteria: excessive loss of weight, difficulty breathing, excessive abdominal swelling, or when visible tumors reached >1 cm in size. Morbidity was scored as tumor-related if tumor burden was >1 cm³, if tumors resulted in compression of thoracic cavity, or hemorrhaging. All tumors were fixed in formalin and processed for H&E staining or frozen at -70°C for analysis of RNA or protein levels.

Histology and immunohistochemistry

Tumor histotype was scored using established criteria (Maronpot, 1999). p27 immunostaining was performed as described (Fero et al., 1998). For immunophenotyping tumors according to T- or B-cell origin, 6-m frozen sections of thymic lymphomas were fixed in acetone at -20°C overnight. Sections were stained using a standard two-step indirect fluorescent technique. The primary antibodies used were rat-anti-mouse CD3 (Serotec) and rat-anti-mouse CD45R/B220 (Pharmingen). An FITC-labeled anti-rat antibody (Caltag) was used to visualize bound primary antibody.

Western blot analysis

Nuclear and cytoplasmic protein extracts were prepared as described (Schreiber et al., 1989) with the following modifications. Pieces of tumors and normal tissues were minced with a razor blade and dissolved in buffer A (Schreiber et al., 1989), and further homogenized for 1 min on ice (PowerGen 125, Fisher Scientific). Buffers A and C both contained 1mM DTT, 0.4 mg/ml Pefablock, 25 mg/ml Aprotinin, 10 mg/ml Pepstatin, and 10 mg/ml Leupeptin (Boehringer Mannheim) to inhibit proteases and in addition buffer C contained 25% glycerol. Protein concentrations were standardized using the Bradford assay (BioRad) and equal loading was confirmed by Ponzo S staining of PVDF membranes after blotting (BioRad). Polyclonal antibodies directed against p27 (sc-528), cyclinE (sc-481), cyclinD1 (sc-755), and cdk2 (sc-163) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti--tubulin (clone B-5-1-2, Sigma) and anti-max antibodies (generous gift of Dr R Eisenmann) served as controls for the completeness of cellular fractionation (data not shown). Densitometric analysis of Western blots was performed using ImageQuant software for Macintosh, version 1.2.

Histone kinase assay

Protein extracts were prepared and stored in 50% glycerol. Protein (400 g) was incubated on ice in 500 l RIPA containing cdk2 antibody. Then, 30 l of washed Protein A–Sepharose beads was added and the reaction was rotated for 1 h at 4°C. Beads were washed twice in RIPA and once in a buffer containing 25 mM Tris HCl (pH 7.5), 70 mM NaCl, 10 mM MgCl₂, and 1 mM DTT. The beads were incubated with Histone H1 as a substrate and ³²P-ATP for 30 min at 37°C. The reaction was stopped by the addition of sample buffer, denatured for 5 min at 95°C, and the protein separated on a 12% SDS acrylamide gel run for 45 min at 150 V. The gel was washed, fixed, and exposed to an X-ray film to visualize bands.

Quantitative real-time PCR

The total RNA from lymphomas was isolated using the RNAeasy Midi kit (Qiagen). Approximately 0.12–0.24 g of solid tissue was grounded with mortar and pestle and lysed in guanidine isothiocyanate lysis buffer. The lysate was bound to a silica gel-based column membrane, washed, eluted in RNase-free water, and quantitated with a Quant spectrophotometer (Bio-Tek Instruments, Inc.). All reactions were performed with the Applied Biosystem division 7700 sequence detector (Foster City, CA, USA) and the Taqman detection probe. Amplification and detection of p27 and an internal control, GAPDH, was performed using the one-step RT–PCR master mix reagent set. Separate tubes were used for each reaction, each containing p27 forward 5'-AGGCTGGGTTAGCGGAGC- 3' and reverse 5'-GAACCGTCTGAAACATTTTCTTCTGT- 3' primers and probes for p27 5'-ACCTGCTGCAGAAGATTCTTCTTCGCAA-3' or GAPDH (Taqman rodent GAPDH control reagents). The reactions consisted of 50% PCR master mix, 2.5% RT reaction mix, 15 M p27, or 5 M GAPDH forward and reverse primers, 10 M taqman probe, and 50ng RNA in a final reaction volume of 50 l. The expression levels of p27 transcript in each sample were first normalized using the endogenous control, GAPDH according to the equation: Ct p27 - Ct GAPDH=Ct, where Ct is the threshold cycle number. The average Ct of all normal thymus samples was then calculated and Ct values of individual tumor samples were subtracted from this value resulting in a Ct for each tumor sample reflected in the equation: Ct normal tissue samples (averaged)–individual Ct tumor samples=Ct of individual tumor samples. We then applied this value to the equation 2^-Ct, which provided the percentage of p27 transcript present in the tumor samples as compared to normal tissue.

Statistical methods

Kaplan–Meier survival curves were used to display the time to tumor morbidity or mortality. Corresponding two-sample log-rank statistics were used to test for rate differences between genotype groups (Kalbfleish and Prentice, 1980). Fisher's exact test was used to test for the difference in the probability of developing tumors other than lymphomas and sarcomas between mice with compound deletions and those with single gene p53 deletions. A synergistic effect of deletions at p27 and p53 will result in tumor mortality times shorter than expected from the combined independent contributions of the corresponding single locus deletions. The P-values for a test of two-locus synergism were based on the simulated distribution of the median survival time, obtained with bootstrap methods under the null hypothesis of independent locus effects.

References

1. Adnane J, Jackson RJ, Nicosia SV, Cantor AB, Pledger WJ and Sebti SM. (2000). Oncogene, 19, 5338–5347. | Article | PubMed | ISI | ChemPort |
2. Bearss DJ, Lee RJ, Troyer DA, Pestell RG and Windle JJ. (2002). Cancer Res., 62, 2077–2084. | PubMed | ISI | ChemPort |
3. Blain SW and Massague J. (2002). Nat. Med., 8, 1076–1078. | Article | PubMed | ISI | ChemPort |
4. Brugarolas J, Chandrasekaran C, Gordon JI, Beach D, Jacks T and Hannon GJ. (1995). Nature, 377, 552–557. | Article | PubMed | ISI | ChemPort |
5. Bullrich F, MacLachlan TK, Sang N, Druck T, Veronese ML, Allen SL, Koff A, Heubner K, Croce CM and Giordano A. (1995). Cancer Res., 55, 1199–1205. | PubMed |
6. Carrano AC, Eytan E, Hershko A and Pagano M. (1999). Nat. Cell Biol., 1, 193–199. | Article | PubMed | ISI | ChemPort |
7. Chiarle R, Budel LM, Skolnik J, Frizzera G, Chilosi M, Corato A, Pizzolo G, Magidson J, Montagnoli A, Pagano M, Maes B, Wolf-Peeters C and Inghirami G. (2000). Blood, 95, 619–626. | PubMed | ISI | ChemPort |
8. Ciaparrone M, Yamamoto H, Yao Y, Sgambato A, Cattoretti G, Tomita N, Monden T, Rotterdam H and Weinstein IB. (1998). Cancer Res., 58, 114–122. | PubMed | ISI | ChemPort |
9. Coqueret O. (2003). Trends Cell Biol., 13, 65–70. | Article | PubMed | ISI | ChemPort |
10. Deng C, Zhang P, Harper JW, Elledge SJ and Leder P. (1995). Cell, 82, 675–684. | Article | PubMed | ISI | ChemPort |
11. Di Cristofano A, De Acetis M, Koff A, Cordon-Cardo C and Pandolfi PP. (2001). Nat. Genet., 27, 222–224. | Article | PubMed | ISI | ChemPort |
12. Donehower LA, Harvey M, Slagle BL, McArthur MJ, Montgomery Jr CA, Butel JS and Bradley A. (1992). Nature, 356, 215–221. | Article | PubMed | ISI | ChemPort |
13. El-Diery WS. (2001). Nat. Cell Biol., 3, 71–73.
14. Erlanson M, Portin C, Linderholm B, Lindh J, Roos G and Landberg G. (1998). Blood, 92, 770–777. | PubMed | ISI | ChemPort |
15. Esposito V, Baldi A, De Luca A, Groger AM, Loda M, Giordano GG, Caputi M, Baldi F, Pagano M and Giordano A. (1997). Cancer Res., 57, 3381–3385. | PubMed | ISI | ChemPort |
16. Fero ML, Randel E, Gurley KE, Roberts JM and Kemp CJ. (1998). Nature, 396, 177–180. | Article | PubMed | ISI | ChemPort |
17. Fero ML, Rivkin M, Tasch M, Porter P, Carow CE, Firpo E, Polyak K, Tsai LH, Broudy V, Perlmutter RM, Kaushansky K and Roberts JM. (1996). Cell, 85, 733–744. | Article | PubMed | ISI | ChemPort |
18. Franklin DS, Godfrey VL, O'Brien DA, Deng C and Xiong Y. (2000). Mol. Cell. Biol., 20, 6147–6158. | Article | PubMed | ISI | ChemPort |
19. Geisen C, Karsunky H, Yucel R and Moroy T. (2003). Oncogene, 22, 1724–1729. | Article | PubMed | ChemPort |
20. Harvey M, McArthur MJ, Montgomery Jr CA, Bradley A and Donehower LA. (1993). FASEB J., 7, 938–943. | PubMed | ISI | ChemPort |
21. Hollstein M, Rice K, Greenblatt MS, Soussi T, Fuchs R, Sorlie T, Hovig E, Smith-Sorensen B, Montesano R and Harris CC. (1994). Nucleic Acids Res., 22, 3551–3555. | PubMed | ISI | ChemPort |
22. Jacks T, Remington L, Williams BO, Schmitt EM, Halachmi S, Bronson RT and Weinberg RA. (1994). Curr. Biol., 4, 1–7. | Article | PubMed | ISI | ChemPort |
23. Jackson RJ, Adnane J, Coppola D, Cantor A, Sebti SM and Pledger WJ. (2002). Oncogene, 21, 8486–8497. | Article | PubMed | ISI | ChemPort |
24. Jackson RJ, Engelman RW, Coppola D, Cantor AB, Wharton W and Pledger WJ. (2003). Cancer Res., 63, 3021–3025. | PubMed | ISI | ChemPort |
25. Jones JM, Cui XS, Medina D and Donehower LA. (1999). Cell Growth Differ., 10, 213–222. | PubMed | ISI | ChemPort |
26. Kalbfleish JD and Prentice HM. (1980). The Statistical Analysis of Failure Time. DataWiley: New York.
27. Kemp CJ, Wheldon T and Balmain A. (1994). Nat. Genet., 8, 66–69. | Article | PubMed | ISI | ChemPort |
28. Kiyokawa H, Kineman RD, Manova-Todorova KO, Soares VC, Hoffman ES, Ono M, Khanam D, Hayday AC, Frohman LA and Koff A. (1996). Cell, 85, 721–732. | Article | PubMed | ISI | ChemPort |
29. Lloyd RV, Erickson LA, Jin L, Kulig E, Qian X, Cheville JC and Scheithauer BW. (1999). Am. J. Pathol., 154, 313–323. | PubMed | ISI | ChemPort |
30. Loda M, Cukor B, Tam SW, Lavin P, Fiorentino M, Draetta GF, Jessup JM and Pagano M. (1997). Nat. Med., 3, 231–234. | Article | PubMed | ISI | ChemPort |
31. Maronpot RR. 1999. Pathology of the Mouse, Maronpot RR, Boorman GA, Gaul U1 (eds). Cache River Press: Vienna, IL.
32. Martin-Caballero J, Flores JM, Garcia-Palencia P and Serrano M. (2001). Cancer Res., 61, 6234–6238. | PubMed | ISI | ChemPort |
33. Martins CP and Berns A. (2002). EMBO J., 21, 3739–3748. | Article | PubMed | ISI | ChemPort |
34. Masciullo V, Sgambato A, Pacilio C, Pucci B, Ferrandina G, Palazzo J, Carbone A, Cittadini A, Mancuso S, Scambia G and Giordano A. (1999). Cancer Res., 59, 3790–3794. | PubMed | ISI | ChemPort |
35. McAllister SS, Becker-Hapak M, Pintucci G, Pagano M and Dowdy SF. (2003). Mol. Cell. Biol., 23, 216–228. | Article | PubMed | ISI | ChemPort |
36. Moller MB. 2000. Leukemia Lymphoma, 39, 19–27. | PubMed |
37. Nakayama K, Ishida N, Shirane M, Inomata A, Inoue T, Shishido N, Horii I and Loh D. (1996). Cell, 85, 707–720. | Article | PubMed | ISI | ChemPort |
38. Nakayama K and Nakayama K. (1998). Bioessays, 20, 1020–1029. | Article | PubMed | ChemPort |
39. Nourse J, Firpo E, Flanagan WM, Coats S, Polyak K, Lee MH, Massague J, Crabtree GR and Roberts JM. (1994). Nature, 372, 570–573. | Article | PubMed | ISI | ChemPort |
40. Pagano M, Tam WS, Theodoras AM, Beer-Romero P, Sal GD, Chau V, Yew PR, Draetta GF and Rolf M. (1995). Science, 269, 682–685. | Article | PubMed | ISI | ChemPort |
41. Park MS, Rosai J, Nguyen HT, Capodieci P, Cordon-Cardo C and Koff A. (1999). Proc. Natl. Acad. Sci. USA, 96, 6382–6387. | Article | PubMed | ChemPort |
42. Philipp-Staheli J, Kim KH, Payne SR, Gurley KE, Liggitt D, Longton G and Kemp CJ. (2002). Cancer Cell, 1, 355–368. | Article | PubMed | ISI |
43. Philipp-Staheli J, Payne SR and Kemp CJ. (2001). Exp. Cell Res., 264, 148–168. | Article | PubMed | ISI | ChemPort |
44. Philipp J, Vo K, Gurley KE, Seidel K and Kemp CJ. (1999). Oncogene, 18, 4689–4698. | Article | PubMed | ISI | ChemPort |
45. Pietenpol JA, Bohlander SK, Sato Y, Papadopoulos N, Liu B, Friedman C, Trask BJ, Roberts JM, Kinzler KW and Rowley JD et al. (1995). Cancer Res., 55, 1206–1210. | PubMed | ISI | ChemPort |
46. Piva R, Cancelli I, Cavalla P, Bortolotto S, Dominguez J, Draetta GF and Schiffer D. (1999). J. Neuropathol. Exp. Neurol., 58, 691–696. | PubMed | ISI | ChemPort |
47. Polyak K, Kato JY, Solomon MJ, Sherr CJ, Massague J and Roberts JM. (1994). Genes Dev., 8, 9–22. | Article | PubMed | ISI | ChemPort |
48. Ponce-Castaneda MV, Lee MH, Latres E, Polyak K, Lacombe L, Montgomery K, Mathew S, Krauter K, Sheinfeld J and Massague J et al. (1995). Cancer Res., 55, 1211–1214. | PubMed | ChemPort |
49. Quintanilla-Martinez L, Thieblemont C, Fend F, Kumar S, Pinyol M, Campo E, Jaffe ES and Raffeld M. (1998). Am. J. Pathol., 153, 175–182. | PubMed | ISI | ChemPort |
50. Sanchez-Beato M, Saez AI, Martinez-Montero JC, Sol MM, Sanchez-Verde L, Villuendas R, Troncone G and Piris MA. (1997). Am. J. Pathol., 151, 151–160. | PubMed | ISI | ChemPort |
51. Schreiber E, Matthias P, Muller MM and Schaffner W. (1989). Nucleic Acids Res., 17, 6419. | Article | PubMed | ISI | ChemPort |
52. Sherr CJ and Roberts JM. (1999). Genes Dev., 13, 1501–1512. | PubMed | ISI | ChemPort |
53. Shiohara M, El-Deiry WS, Wada M, Nakamaki T, Takeuchi S, Yang R, Vogelstein B and Koeffler HP. (1994). Blood, 84, 3781–3784. | PubMed | ISI | ChemPort |
54. Singh SP, Lipman J, Goldman H, Ellis Jr FH, Aizenman L, Cangi MG, Signoretti S, Chiaur DS, Pagano M and Loda M. (1998). Cancer Res., 58, 1730–1735. | PubMed | ISI | ChemPort |
55. Skapek SX, Rhee J, Spicer DB and Lassar AB. (1995). Science, 267, 1022–1024. | Article | PubMed | ISI | ChemPort |
56. Slingerland J and Pagano M. (2000). J. Cell. Physiol., 183, 10–17. | Article | PubMed | ISI | ChemPort |
57. Timme TL and Thompson TC. (1994). Biotechniques, 17, 462–463.
58. Vousden KH and Lu X. (2002). Nat. Rev. Cancer, 2, 594–604. | Article | PubMed | ISI | ChemPort |
59. Waga S, Hannon GJ, Beach D and Stillman B. (1994). Nature, 369, 574–578. | Article | PubMed | ISI | ChemPort |
60. Waldman T, Lengauer C, Kinzler KW and Vogelstein B. (1996). Nature, 381, 713–716. | Article | PubMed | ISI | ChemPort |
61. Wang YA, Elson A and Leder P. (1997). Proc. Natl. Acad. Sci. USA, 94, 14590–14595. | Article | PubMed | ChemPort |
62. Yang WC, Mathew J, Velcich A, Edelmann W, Kucherlapati R, Lipkin M, Yang K and Augenlicht LH. (2001). Cancer Res., 61, 565–569. | PubMed | ISI | ChemPort |
63. Yasui W, Kudo Y, Semba S, Yokozaki H and Tahara E. (1997). Jpn. J. Cancer Res., 88, 625–629. | PubMed | ChemPort |

Acknowledgements

We thank Keri Fisher for technical assistance and Matthew Fero for valuable suggestions. This work was funded by research grants from the American Cancer Society, the NIH (CA70414), the NIEHS (ES011045), and the Life Possibilities Fund to CJK.