Evidence for a p27 tumor suppressive function independent of its role regulating cell proliferation in the prostate

Abstract

Reduced p27 levels correlate with poor prognosis in a wide spectrum of human tumors and can accelerate tumorigenesis in mouse tissues. To determine whether p27 deficiency can accelerate tumorigenesis in tissues with inactive Rb and p53 pathways, we examined the effect of p27 status on prostate tumorigenesis in mice expressing simian virus 40 large T antigen (LT). In p27-deficient mice expressing LT, tumors progressed from high-grade prostatic intraepithelial neoplasia to poorly differentiated carcinoma at a greatly accelerated rate. p27 deficiency could not collaborate with a mutant of LT that fails to inactivate the Rb pathway alone. Furthermore, p27 deficiency does not increase the proliferation index, reduce the apoptotic index, or affect the expression of E2F-dependent genes in cells expressing LT at any stage of the disease. Expression of LT alone leads to maximal proliferation, but p27 deficiency still increases the amount of cyclin A and cyclin-dependent kinase 2-associated kinase activity in tissues. Interestingly, this model recapitulates an important feature of the human disease, specifically a high frequency of allelic loss of chromosome 16q, which is syntenic to mouse chromosome 8. Loss of heterozygosity may accelerate the inactivation of other tumor suppressors, such as E-cadherin, which are located in this interval. These experiments provide direct physiological and causal evidence that p27 has tumor suppressive functions independent of its role regulating cell proliferation.

Amember of the CIP-KIP family of cyclin-dependent kinase (cdk) inhibitors, p27, has been reported to interact with cdk2, Grb2, and rho family GTPases, affecting cell proliferation, ras signaling, and cell migration, respectively (1, 2). Cdks promote cell-cycle progression, in part by phosphorylating Rb and the related pocket proteins, which affects their ability to regulate E2F-transcription factors, leading to altered expression of E2F-regulated genes, which are rate-limiting for S-phase entry. Mutations that render the Rb pathway functionally inactive are found in nearly all tumors (3). Rb is the gate-keeper of cell proliferation, and inactivation of this pathway, when combined with mutations in the Arf/53 pathway that disable the cytostatic or cytotoxic response, contribute to unregulated proliferation in cancer cells (4). Although a reduction in the number of cells positively staining for nuclear p27 is correlated with poor prognosis for a variety of tumor types (5–11), we have little understanding of how this may contribute to tumor progression.

Mice that are null for p27 (12, 13) or that lack the ability to bind cyclin–cdk complexes (14) (allele designated Δ51) have been used to determine the contribution p27 makes to tumor suppression. A reduction in p27 gene dosage increases the number of tumors and the rate at which they progress after challenge with carcinogens (15), oncogenes (16), or tumor suppressor loss (17–20). In some cases, there is a 2- to 3-fold increase in proliferation or decrease in apoptosis, but whether this is sufficient to account for the tumor promoting role of p27 deficiency is not clear (6). We decided to use mice in which the Rb and p53 pathways were specifically inactivated to ask whether there were other contributions that p27 deficiency might make to accelerate tumor progression.

Here we report that p27 deficiency markedly accelerates tumor progression in prostate tissues when both the Rb and p53 pathways are inactivated by large T antigen (LT). During the course of tumor progression the proliferation and apoptotic indices and the expression of E2F-regulated genes were similar in wild type and p27-deficient mice. p27 deficiency did increase the amount of cdk2 kinase activity in LT-expressing cells. Loss of heterozygosity (LOH) at the distal end of chromosome 8 was more frequent in tumors arising in p27-deficient animals compared with similarly staged tumors in wild-type animals; however, the overall frequency of chromosomal gains and losses as measured by FISH at any stage of the disease was equivalent. These experiments provide direct physiological and causal evidence for nonproliferative-related tumor suppressor functions for p27.

Methods

Transgenic Mice. Sequence encoding the FLAG epitope (DYKD-DDDK) was added by PCR mutagenesis in frame to the 5′ end of the simian virus 40 A ORF. The FLAG-tagged simian virus 40 A sequence was linked to the 0. 43-kb minimal rat probasin promoter (21). To create the K-transgene, the conserved LXCXE site was mutated to LFKLK by using a Gene Editor kit (Promega). Mice were genotyped by PCR of tail DNA by using primers and conditions described in Supporting Methods, which is published as supporting information on the PNAS web site. The p27^Δ51/Δ51 mouse was described in ref. 14.

Immunohistochemistry. Paraffin-embedded tissues were sectioned at 5- to 8-μM thickness. The antibodies that were used included 6 μg/ml BrdUrd (clone BMC9318, Roche); 1 μg/ml cleaved caspase 3 (catalog no. 9661, Cell Signaling Technology, Beverly, MA), 5 μg/ml phosphorylated histone H3 (catalog no. 06570/06571, Upstate Cell Signaling Solutions, Charlottesville, VA); 12 μg/ml FLAG (M5) (Sigma), 1 μg/ml p27 (catalog no. K25020, Becton Dickson), 1 μg/ml p53 (catalog no. SC-6243, Santa Cruz Biotechnology); 0.1 μg/ml cleaved poly(ADP-ribose) polymerase (catalog no. 9544, Cell Signaling Technology); 0.16 μg/ml Ki67 (catalog no. NCL-Ki67P, NovoCastra, Newcastle, U.K.), 2.5 μg/ml E-cadherin (catalog no. 610181, Becton Dickson); 1:2,000 chromogranin SP-1, (catalog no. 20085, Immunostar, Hudson, WI).

Extracts from Prostate Tissues. Extracts were prepared by homogenizing and sonicating snap-frozen tissues in 50 mM Tris·HCl, pH 7.4/50 mM NaCl/5 mM EDTA/0.5% Nonidet P-40, with protease inhibitors and clearing the lysate at 13, 000 × g for 10 min at 4°C.

Immunoblotting. Unless noted otherwise, all primary antibodies were from Santa Cruz Biotechnology and used at a concentration of 1:1,000. Secondary antibodies were from Amersham Pharmacia and used at a concentration of 1:10,000. Primary antibodies included α-cyclin E (M-20), α-cyclin A (C-19), α-cyclin D1 (72–13G), α-cyclin D2 (M-20), α-cyclin D3 (C-16), α-cdk2 (M-2), α-cdk4 (C-22), α-cdk6 (C-21), α-p21 (F5), α-p27 (C-19), α-proliferating cell nuclear antigen (PC-10), α-p53 (FL-393), α-p107 (C-18), α-skp2 (H-435), α-androgen receptor (N-20), α-FLAG (catalog no. F-7425, Sigma) at 1:3,000, α-smooth muscle actin (catalog no. A-5228, Sigma) at 1:3,000, α-extracellular signal-related kinase 1/2 (catalog no. 9102, Cell Signaling Technology), α-phospho-extracellular signal-related kinase 1/2 (catalog no. 9101, Cell Signaling Technology), α-AKT (catalog no. 9272, Cell Signaling Technology), α-phospho-AKT (S473) (catalog no. 9271, Cell Signaling Technology), α-eIF4E (catalog no. 9742, Cell Signaling Technology), α-phospho-p90 ribosomal S6 kinase (S380) (catalog no. 9341, Cell Signaling Technology), and α-phospho-c-Jun N-terminal kinase (T183/Y185) (catalog no. 9251, Cell Signaling Technology).

Kinase Assays and Oligonucleotide Array Expression Analysis. We incubated 40 μg of extract with 4 μg of α-cdk2 (M2), α-cyclin A (H432), or α-cyclin E (M20), and the kinase reaction carried out as described in ref. 22. We used 2 μg of total RNA extracted from snap-frozen tissues as a template for reverse-transcription with an oligo(dT)-T7 primer. The resulting cDNA was amplified and biotinylated. We hybridized 10 μg of biotinylated, fragmented cRNA to a MOE430A array (Affymetrix, Santa Clara, CA) for 16 h at 45°C, and then the product was stained and washed. Chips were scanned with a high-numerical aperture and flying objective lens in the GS3000 scanner (Affymetrix). The image was quantified by using microarray suite 5.1 (Affymetrix) with the default parameters for the statistical algorithm and all probe-set scaling with a target intensity of 500.

Comparative Genomic Hybridization Array Analysis. The whole-genome, 1-Mb-resolution clone set of mouse bacterial artificial chromosomes is described in ref. 23. Bacterial artificial chromosome DNA was amplified by ligation-mediated PCR. Purified PCR products were spotted in duplicate onto UltraGAPS-coated slides (Corning Life Sciences), and slides were UV-crosslinked (3,000 mV). Genomic DNA was extracted from tumor and liver of the same mouse. One microgram each of HaeIII-digested template and reference genomic DNA were labeled in parallel by random priming. After removal of unincorporated nucleotides with Sephadex G-50 columns, labeled template and reference DNAs were combined (110 μl each), and 135 μl of mouse cot-1 DNA at 1 μg/μl (Invitrogen) was added, and the mixture was concentrated to 10 μl. We then added 130 μl of SlideHyb hybridization buffer 3 (Ambion) containing 0.25 μg/μl of yeast tRNA (Invitrogen). Hybridization was at 37°C for 44 h. After washing steps in low and medium stringency buffers (Genomic Solutions, Ann Arbor, MI), slides were dipped in water and dried. Arrays were scanned by using an Axon 4000B scanner (Axon Instruments, Union City, CA). Images were analyzed by using genepix pro 4.1 (Axon Instruments). The data were normalized by looking at individual printing blocks. The median ratio for each block was calculated, and the spots were then transformed to a median ratio of 1 (23). Replicate spots were averaged, and, for each sample, two experiments were done with reversal of the dye labels. Log2 ratios of normalized signal intensity (Cy3/Cy5) were plotted together along individual chromosomes to exclude labeling artifacts. Thresholds for copy number gain and loss were set at 1.2 and 0.8, respectively.

FISH Analysis of Touch Preps. Prostate touch preps on glass slides were air-dried, fixed in -20°C 3:1 methanolz/glacial acetic acid for 20 min, air dried, then stored at -20°C. Slides were incubated with collagenase H in PBS, fixed in 2% formaldehyde in PBS containing 50 mM MgCl₂. Bacterial artificial chromosomes were purchased from Invitrogen: X chromosome, RP23–54C14 and RP23–109E24; chromosome 12, RP23–54G4 and RP23–41E22; and Y chromosome, CITB-590P11. The X chromosome was labeled by nick-translation with biotin, the Y chromosome was labeled with digoxigenin, and chromosome 12 was labeled with spectrum orange and spectrum green (Vysis, Downer's Grove, IL). FISH was performed by following standard procedures (24). Areas of overlapping cells were excluded from analysis.

Results

p27 Loss Accelerates Progression of Prostate Tumors Induced by Simian Virus 40 Large T Antigen. We generated transgenic mice by using the minimal rat probasin promoter to direct prostate-specific expression of the simian virus 40 early region encoding LT and small T antigens (ST). An N-terminal FLAG epitope tag was added to quantitate expression of transgenes with mutated LT sequences. Transgenes were expressed ubiquitously in glandular epithelial cells of the ventral, dorsal, and lateral lobes of the prostate, and no significant expression was seen in other tissues, consistent with previous reports (21).

We detected LT but not ST in tumor extracts (Fig. 1A). ST was detected in extracts from tumors arising in the transgenic adenocarcinoma of mouse prostate mouse, which uses a similar transgene lacking the FLAG tag (Fig. 1 A) (21). The FLAG tag disrupted ST expression in a rat fibroblast transformation assay as well (25) (data not shown). Thus, the transgenic mice described here express only LT and are not directly comparable to the TRAMP prostate model (21).

Fig. 1.

p27 deficiency collaborates with prostate-specific expression of LT to decrease survival. (A) Immunoblot. Extracts obtained from the prostates of 4-month-old TRAMP, W¹⁰, K³³, or nontransgenic (-) animals were resolved on 10% SDS-polyacrylamide gels, and the expression of LT, ST, FLAG, and proliferating cell nuclear antigen (PCNA) was determined. The genotypes of the animals are indicated above each lane, and the antibody used for blotting is indicated to the left of each autoradiogram. (B) Immunohistochemisty. Tissues obtained from 4-month-old animals were sectioned, and the expression of FLAG-tagged transgene and p53 was examined. Animals were labeled with BrdUrd to allow the detection of proliferating cells. (C) Kaplan–Meier survival analysis comparing W¹⁰ transgenic mice and nontransgenic mice differing by p27 status.

After back-crossing the Δ51 mutation six generations into the FVB/N background, we generated sibling pairs of W¹⁰p27^+/+, W¹⁰p27^Δ51/Δ51, p27^+/+, and p27^Δ51/Δ51 mice. As expected, p53 accumulated and BrdUrd staining increased in prostate cells expressing the W¹⁰ transgene (Fig. 1B). Morbid mice were killed and examined for prostate tumors. All animals were killed by 52 weeks of age. The survival of W¹⁰p27^Δ51/Δ51 mice was markedly reduced, with a t_1/2 of 29 weeks, compared with a t_1/2 of 51 weeks for W¹⁰p27^+/+ mice (P < 0.001, log-rank test) (Fig. 1C). At the time of necropsy, grossly visible prostate tumors were found in 16 of 19 W¹⁰p27^Δ51/Δ511 mice (16 of 16 older than 22 weeks) and 12 of 25 W¹⁰p27^+/+ mice (12 of 22 older than 22 weeks). Gross metastases to lung, liver, and lymph nodes were apparent in 2 of 12 W¹⁰p27^+/+ animals and 4 of 16 W¹⁰p27^Δ51/Δ51 animals. These findings indicate that p27 deficiency cooperates with LT to accelerate morbidity, suggesting that p27 loss might contribute independently of the Rb and p53 pathway.

p27 Loss Accelerates Progression to a Poorly Differentiated Cancer Phenotype. In mouse, prostate cancer progresses through a series of histologically well defined stages from high-grade prostatic intra-epithelial neoplasia (HGPIN) lesions to poorly differentiated carcinoma (PDCA) (26). We examined whether p27 status affected the rate at which tumors progressed through these stages (Fig. 2). At 2–4 months of age HGPIN was detected in all W¹⁰ mice examined, regardless of p27 status (Fig. 2). Remarkably, one of the 3-month-old W¹⁰p27^Δ51/Δ51 animals also had a large region of PDCA adjacent to HGPIN. By 6–8 months, well differentiated carcinoma (WDCA) was detected in 7 of 7 W¹⁰p27^+/+ mice but 9 of 10 W¹⁰p27^Δ51/Δ51 mice had already progressed as far as PDCA. By 10–12 months, all of the W¹⁰p27^Δ51/Δ51 mice had become morbid and had to be killed, but 4 of 13 mice in the W¹⁰p27^+/+cohort had WDCA, 1 of 13 had moderately differentiated carcinoma, and 8 of 13 had PDCA. In summary, p27 deficiency did not alter the rate at which HGPIN developed, but it did accelerate the rate at which these lesions progressed to PDCA.

Fig. 2.

Histologic progression is accelerated in p27-deficient mice. Hematoxylin- and eosin-stained images representing normal prostate tissue, HGPIN, WDCA, moderately differentiated carcinoma (MDCA), and PDCA are shown, along with a summary of histologic types as a function of age.

p27 Loss Accelerates Tumorigenesis Independently of Effects on Proliferation, Apoptosis, or E2F-Regulated Gene Expression. To determine whether the acceleration associated with p27 deficiency increased proliferation or decreased apoptosis, we measured these indices in equivalently staged lesions during the course of tumor progression by counting the numbers of cells positive for three immunohistochemical markers of proliferation (BrdUrd, Ki67, and phosphorylated histone H3) or two markers of apoptosis [cleaved caspase 3 and cleaved Poly(ADP-ribose) polymerase 1]. p27 deficiency increased proliferation indices 2-fold in nontransgenic mice (data not shown). In W¹⁰p27^+/+ and W¹⁰p27^Δ51/Δ51 transgenic mice, proliferation indices increased 10- to 20-fold. There were no differences related to p27 status in transgene-positive tissue at any stage of tumor progression (Fig. 6A, which is published as supporting information on the PNAS web site). Apoptotic rates were also equivalent in both backgrounds in all tumor stages (Fig. 6A), which suggested that p27 deficiency does not significantly enhance proliferation or decrease apoptosis in cells that have already sustained functional inactivation of the Rb and p53 pathways through expression of LT.

The E2F-dependent gene expression program is largely associated with S-phase entry, but there are also targets that maintain genome integrity (27). To ascertain whether p27 deficiency affected E2F-dependent gene regulation, we compared the expression of E2F-regulated genes in histologically matched tissues that differed with respect to p27 status (Fig. 7 and Data Set 1, which are published as supporting information on the PNAS web site). By using hierarchical clustering analysis focusing on the E2F-regulated genes described by Nevins and colleagues (28), we identified histologic subtype-specific expression signatures. However, in transgenic tissues, we did not observe even a single gene that was regulated differently based on p27 status. Thus, if E2F gene expression is used as a surrogate to assess Rb pathway inactivation, p27 deficiency did not lead to further inactivation of this pathway in transgenic tissues, again suggesting that the mechanism of accelerated tumor progression is likely independent of the Rb–E2F pathway.

p27 Loss Cannot Substitute for LT-Dependent Inactivation of Rb. p27 is an inhibitor of cdk2, which in turn can inactivate Rb and the other pocket proteins. To determine whether p27 loss can substitute for Rb pathway inactivation by LT, we made K³³ transgenic mice in which the conserved LXCXE sequence was mutated to LXKLK, which is predicted to disrupt Rb binding without affecting p53 binding (29). In extracts from prostates of 4-month-old mice, the level of expression of the W¹⁰ and K³³ transgenes was similar (Fig. 1A). p53 accumulated in prostate cells expressing the K³³ transgene, but BrdUrd staining did not increase (Fig. 1B). We found no evidence of PIN lesions in K³³p27^+/+ (n = 45) and K³³p27^Δ51/Δ51 (n = 35) animals up to 52 weeks of age, when all animals were killed. The survival of these animals was similar to that seen for their nontransgenic siblings. However, the proliferation index in K³³p27^Δ51/Δ51 mice was ≈3-fold higher than in K³³ transgenic mice (Fig. 6B). This level of proliferation was still low compared with that observed in mice expressing the W¹⁰ transgene and not sufficient to drive progression into the PIN stage. Thus, although p27 deficiency cannot substitute for LT-dependent inactivation of the Rb pathway (a requirement for tumorigenesis in this model), it can increase the number of proliferating cells.

Accelerated Tumor Progression in W¹⁰p27^Δ51/Δ51 Mice Correlated with Increased Cyclin A–cdk2 Kinase Activity. We measured the amount and activity of cell cycle-regulatory molecules that might be dys-regulated in tissues lacking functional p27. We found that the amount of cyclin A (Fig. 3A) and cyclin A kinase activity (Fig. 3B) increased in PDCA arising in W¹⁰p27^Δ51/Δ51 mice. Immunoblots used to detect a variety of other molecules linked to the Pten pathway (Akt, rsk, and eIF4E), the mitogen-activated protein kinase pathway, and their cell cycle effectors (p21, p107, and cyclin D3) did not reveal significant differences associated with p27 status and are shown in Fig. 8, which is published as supporting information on the PNAS web site.

Fig. 3.

Cyclin (cyc) A–cdk2 activity increases in p27-deficient animals expressing LT. (A) We resolved 40 μg of pooled extracts from three to four mice by 10% SDS/PAGE, and the expression of proteins was determined by immunoblot. (B) Cyclin A–cdk2 kinase activity increases in p27-deficient W¹⁰ WDCA and PDCA. Pooled extracts were immunoprecipitated with the antibodies indicated, and histone H1 phosphorylation was measured. Control immunoprecipitations with rabbit anti-mouse nonspecific antibodies were negative (data not shown).

The data suggest that p27 deficiency does not contribute to tumor progression by increasing proliferation in cells expressing LT. To test this possibility further, we synchronously induced the transgene by supplementing castrated animals with testosterone. Castration induces regression of the prostate gland and reduces LT expression and BrdUrd incorporation, all of which were reversed within 5 days after testosterone treatment (Fig. 9, which is published as supporting information on the PNAS web site). The number of cells actively engaged in S phase was then measured by BrdUrd incorporation, and extracts were prepared for immunoblotting and kinase assays at various times after testosterone addition. p27 deficiency or expression of LT alone accelerated the rate at which cells reentered the cell cycle after addition of testosterone (Fig. 4A). After the wave of proliferation peaking at the third day, the cells began to exit the cell cycle, and p27 deficiency did not kinetically affect this process (compare wild type and p27 deficient mice), but expression of LT did. Proliferation indices remained high in both W¹⁰p27^+/+ and W¹⁰p27^Δ51/Δ51 mice. cdk2 activity was much higher in p27 deficient tissue (Fig. 4B). LT does not affect the accumulation of cyclins A or E as cells reentered the cell cycle, but it did affect their down-regulation after the peak of proliferation on day 3 (Fig. 4C). In both tumors and the cell synchrony experiments, any proliferative contribution of p27 appeared to be masked by expression of LT, which suggests that LT expression alone can completely account for the contribution that increased proliferation would make to tumorigenesis in this model, a contribution through its ability to inactivate Rb.

Fig. 4.

p27 deficiency increases the amount of cdk2 activity after induction of LT. (A) Proliferation indices during castration–regeneration. Values were obtained from three to five animals at each time point, and the average is plotted. Each genotype is shown adjacent to each curve. (B) Cyclin (cyc)–cdk2 kinase activity is increased in W¹⁰p27^Δ51/Δ51 tissues. Individual mice were killed 5 days after testosterone supplementation, and the extracts pooled for immunoprecipitation with the antibodies are indicated above each lane. R αM, nonspecific rabbit anti-mouse antibody. The phosphorylation of histone H1 was measured. (C) Protein expression. Extracts were prepared from two to three animals and pooled before resolution of proteins by 10%SDS/PAGE, and the expression of p27, the transgene (FLAG), cyclin E, or cyclin A was measured by immunoblot. Above each set of autoradiograms, we indicated the genotypes of the animals used to prepare the pooled extracts. C, castrated; 2T, castrated and supplemented with testosterone for 2 days; 5T, castrated and supplemented with testosterone for 5 days; 10T, castrated and supplemented with testosterone for 10 days. Antibodies are indicated to the left of each lane. All of the experiments were repeated at least twice for each time point with independent animals.

p27 Loss Does Not Alter the Subtype Progression of Prostate Cancer. Neuroendocrine subtypes of prostate cancer are more aggressive and can be identified by the expression of synaptophasin and chromogranin A (30, 31). Only one of the seven tumors arising in the wild-type background was even weakly stained for chromogranin A (Fig. 10A, which is published as supporting information on the PNAS web site), which suggested that p27 deficiency did not promote the development of prostate cancer into a neuroendocrine subtype.

Nkx3.1 inactivation, Pten inactivation, Rb and p53 loss, and the loss of E-cadherin contribute to prostate cancer progression (32, 33). Loss of Nkx3.1 and Pten contribute early to the formation and advancement of PIN and are unlikely to be involved in this model. Rb and p53 inactivation are carried out by LT. However, loss of surface E-cadherin is a late event in human prostate cancer and can promote the transition from adenoma to carcinoma in mouse models (34). Thus, we looked at E-cadherin expression. E-cadherin was markedly reduced in all of the PDCA samples analyzed (n = 15) from W¹⁰p27^+/+ and W¹⁰p27^Δ51/Δ51 mice, although it was abundant in WDCA, HGPIN, and normal prostate tissues (Fig. 10B), which suggested that p27 deficiency does not replace the requirement to inactivate this tumor suppressor during tumor progression.

PDCA in p27^Δ51/Δ51 Animals Is Characterized by an Increased Frequency of Chromosomal Alterations. Our data indicated that p27 deficiency contributed to tumor progression independently of the effects on proliferation or apoptosis, and it did not alter the developmental progress of the disease; i.e., E-cadherin was still reduced. Thus, we began to think of this as a “rate” problem.

LOH accelerates tumor progression by increasing the frequency of inactivation of tumor suppressor loci by mutation or methylation. To assess whether LOH might underlie the more rapid transition to PDCA in p27 deficient tissues, we prepared DNA from four W¹⁰p27^+/+ and five W¹⁰p27^Δ51/Δ51 poorly differentiated prostate tumors and the liver from the same animals for comparative genomic hybridization analysis. We found a total of eight chromosomal alterations involving four of the five W¹⁰p27^Δ51/Δ51 PDCAs and only a single chromosomal abnormality in one of the W¹⁰p27^+/+ PDCAs (Fig. 5A and Table 1). This difference is even more striking when we consider that PDCAs arose in the p27^+/+ mice at 49–52 weeks of age, compared with 26–35 weeks of age in the p27^Δ51/Δ51 mice.

Fig. 5.

Increased LOH in p27-deficient mice. Chromosomal abnormalities detected by comparative genomic hybridization array analysis in PDCA from W¹⁰p27^+/+ and W¹⁰p27^Δ51/Δ51 mice. (A)(Top) A representative image for a PDCA arising in wild-type mice. (Middle) A representative image for a PDCA arising in p27-deficient mice. (Bottom) Gains and losses are noted in an expanded view of Middle.(B) The regions of chromosome 8 are mapped. Black (W¹⁰p27^Δ51/Δ51) and blue (W¹⁰p27^+/+) represent regions lost, and pink represents those gained.

Table 1. Summary of chromosomal abnormalities.

Genotype	Mouse	Age, weeks	Large-scale change
W10p27+/+	198-10	50	No
	541-10	52	No
	481-10	49	Yes
	520-10	51	No
W10p27Δ51/Δ51	590-10	35	Yes
	455-10	37	Yes
	580-10	35	Yes
	1789-10	26	Yes
	1293-10	26	No

Shown are the results detected by comparative genomic hybridization array analysis in PDCA from W¹⁰p27^+/+ and W¹⁰p27^Δ51/Δ51 mice.

These changes involved chromosomes 6 and 8. This technique only surveys the genome at low resolution and significant deletions/additions were required for a “positive” call. Selected genes that were lost in the overlapping region on chromosome 8 deleted in four of the nine tumors are shown in Fig. 11B, which is published as supporting information on the PNAS web site, and include the cadherin gene cluster, integrin β1, p130, and E2F-4. Mouse 8q is syntenic with human chromosome 16q and to a lesser extent with 10q, loss of which are among the most common genomic alterations in human prostate cancer (35). We did not systematically evaluate expression of all of the genes on chromosome 8 or 6, but no significant differences in expression of E-cadherin, integrin B1, p130, and E2F-4 were seen at any stage of the disease in p27 wild-type or mutant mice (Fig. 11). These findings suggest that p27 deficiency may accelerate the progression of tumors by allowing earlier inactivation of a late-stage tumor suppressor, such as E-cadherin.

p27 Loss Is Not Sufficient to Promote Genomic Instability. cdk2-associated kinase activity has been associated with changes in chromosome stability that result in a gain or loss of chromosomes (36, 37) and may be a precursor to LOH (38). Our data suggested that p27 deficiency might accelerate tumor progression by increasing the frequency at which LOH occurs. To assess this possibility we performed interphase FISH by using randomly selected probes for chromosomes 12, X, and Y on nuclei with tissues showing normal, HGPIN, WDCA, or PDCA histology (Table 2, which is published as supporting information on the PNAS web site). p27 deficiency did not increase the frequency of chromosome gain or loss over that seen in genetically and histologically matched controls, and we concluded that p27 deficiency did not directly increase genome instability over that provided by LT alone.

Discussion

In this study we have shown that the loss of the cyclin–cdk2 binding function of p27 accelerates prostate tumorigenesis even when the Rb and p53 pathways have been inactivated by expression of LT. We show that this mechanism is independent of any measurable effect of p27 loss collaborating with loss of Rb on the E2F program and cell proliferation. Of note, this model recapitulates important features of the human disease, specifically the high frequency of allelic loss of chromosome 16q, which is syntenic to mouse chromosome 8. We interpret our data to suggest that p27 deficiency accelerates tumor progression by allowing the rapid outgrowth of cells in which specific chromosomal alterations advantageous to the cancer occur as a consequence of proliferation driven by LT. Based on the data that reduced E-cadherin is a feature of PDCA, we suspect that the LOH events observed on chromosome 8 facilitate rapid inactivation of this locus and accelerate tumor progression.

Why would we observe an increase in the frequency of tumors that have suffered LOH in p27 deficient mice? It is clear that the type of chromosomal gains and losses reported for inappropriate cdk2 activity (36–38) do not appear to underlie the increased LOH. Mechanisms that account for LOH in epithelial tumors are not always clear and may reflect tumor and chromosome-specific features (39). It seems unlikely that p27 plays a direct role in the DNA damage checkpoints, although we cannot exclude this possibility, but an indirect contribution might be possible. The absence of p27 and, perhaps, inappropriately high levels of cdk2 activity might interfere with the ability of the proliferating prostatic epithelial cell to detect or resolve DNA damage. Because a number of DNA-damage repair proteins are substrates for cdks (40, 41), it seems reasonable that abnormally high cdk activity might affect recognition or repair. Alternatively, inappropriate elevation of cdk activity or the direct loss of p27 might attenuate any cytostatic/cytotoxic effects associated with LOH.

In summary, the experiments here provide physiologic and causal evidence for a nonproliferation-related tumor suppressor function for p27, and this function is important in controlling the transition of tumor cells to malignant disease.