E2f binding-deficient Rb1 protein suppresses prostate tumor progression in vivo

Abstract

Mutational inactivation of the RB1 tumor suppressor gene initiates retinoblastoma and other human cancers. RB1 protein (pRb) restrains cell proliferation by binding E2f transcription factors and repressing the expression of cell cycle target genes. It is presumed that loss of pRb/E2f interaction accounts for tumor initiation, but this has not been directly tested. RB1 mutation is a late event in other human cancers, suggesting a role in tumor progression as well as initiation. It is currently unknown whether RB1 mutation drives tumor progression and, if so, whether loss of pRb/E2f interaction is responsible. We have characterized tumorigenesis in mice expressing a mutant pRb that is specifically deficient in binding E2f. In endocrine tissue, the mutant pRb has no detectable effect on tumorigenesis. In contrast, it significantly delays progression to invasive and lethal prostate cancer. Tumor delay is associated with induction of a senescence response. We conclude that the pRb/E2f interaction is critical for preventing tumor initiation, but that pRb can use additional context-dependent mechanisms to restrain tumor progression.

Mutation of the RB1 tumor suppressor gene initiates the onset of human retinoblastoma and other human cancers (1). Similarly, Rb1 mutation in mice can initiate lethal tumorigenesis, typically with alterations in other genes like Trp53 (2–8). The best characterized mechanism underlying Rb1 protein (pRb) function is its physical interaction with E2f transcription factors (9, 10). When bound to E2fs, pRb blocks their ability to activate transcription and recruits repressive chromatin-modifying factors to actively silence gene expression (11). E2f target genes include those important for the cell cycle (12), and E2f activity is necessary for cell proliferation (13). Such observations have led to the hypothesis that pRb suppresses tumor initiation by binding E2f and blocking the elaboration of an E2f-dependent transcriptional program that drives cells to proliferate. This hypothesis has not been directly tested by specific disruption of the pRb/E2f interaction. RB1 mutation occurs late in the progression of other human cancers like prostate cancer (14–17), suggesting loss of RB1 may also drive progression of previously initiated neoplasia. As current mouse models cannot distinguish between effects of Rb1 loss on tumor initiation from possible effects on tumor progression, in vivo experimental evidence supporting this possibility is lacking.

To address these issues, we created a mutant Rb1 allele in the mouse encoding pRb specifically deficient for E2f binding (18). The mutant allele (Rb1⁶⁵⁴) encodes an arginine-to-tryptophan substitution at codon 654, a position within a hydrogen bond network specifying the structure of the pRb/E2f interaction surface (19). The amino acid substitution compromises E2f binding, but does not affect binding to other known pRb binding partners (18, 20, 21). An analogous germ line mutation has been identified in human hereditary retinoblastoma. Tumorigenesis associated with this mutation is characterized by reduced penetrance and expressivity (22). We have tested the tumor suppressor activity of this mutant Rb1 allele in pituitary, thyroid, and prostate tissue.

Results

Spontaneous Endocrine Tumorigenesis Is Similar in Rb1^−/+ and Rb1^654/+ Mice.

Mice heterozygous for an Rb1 null allele (Rb1^+/−) are prone to multiple endocrine neoplasia, including tumors of the pituitary and thyroid glands, that form upon spontaneous loss of the remaining wild-type Rb1 allele (23–26). We have monitored a cohort of Rb1^654/+ mice to test whether the mutant protein affects endocrine tumorigenesis. Survival of Rb1^654/+ mice is not significantly different from Rb1^−/+ mice (Fig. 1A). The distribution of tumors arising is also similar with all mice developing pituitary tumors (Rb1^−/+ 95% and Rb1^654/+ 87%), medullary thyroid tumors (Rb1^−/+ 71% and Rb1^654/+ 65%), or both (Rb1^−/+ 66% and Rb1^654/+ 55%). Tumors arising in Rb1^654/+ and Rb1^−/+ mice are not distinguishable by histology. Tumors from Rb1^654/+ mice expressed pRb exclusively from the Rb1⁶⁵⁴ allele, whereas tumors from Rb1^−/+ mice completely lacked pRb (Fig. 1 B and C).

Fig. 1.

Pituitary and thyroid tumorigenesis in Rb1-deficient mice. (A) The graph shows cumulative survival for Rb1-deficient mice of the indicated genotypes (sample size = n). Mice dying for reasons unrelated to tumorigenesis are censored from the data and indicated by cross hatches. Differences in cumulative survival between Rb1⁻ or Rb1⁶⁵⁴ mice, either in the presence or absence of E2f1, are not statistically significant by log rank test (P > 0.05). (B) Western blot analysis of pRb in tissue extracts from mice of the indicated genotypes. The letter B represents nontumor-bearing brain tissue and serves as a positive control. The letters T and P are extracts from thyroid or pituitary tumors, respectively. Hsp70 levels serve as a protein loading control. (C) RNA was extracted from the indicated tissues of Rb1^654/+ mice, amplified by RT-PCR, and sequenced. The chromatogram is shown for the sequence surrounding codon 654 (wild type, CGA; Rb1⁶⁵⁴, TGG).

As pRb⁶⁵⁴ expression had no detectable effect on endocrine tumorigenesis, disruption of the pRb/E2f interaction was apparently sufficient to initiate tumorigenesis. However, it remained possible that the R654W mutation disrupts additional functions beyond pRb/E2F interaction. If disruption of pRb/E2F interaction was the primary defect causing tumor initiation in Rb1^654/+ mice, then reducing E2f activity should lower the incidence of lethal endocrine tumors as previously demonstrated in Rb1^−/+ mice (27). Loss of E2f1 increased the survival of both Rb1^−/+ and Rb1^654/+ mice to a similar extent, with at least 80% of mice surviving to 600 d (Fig. 1A). As pRb⁶⁵⁴ had no detectable tumor suppressor activity and tumorigenesis was rescued by compound E2f1 loss, disruption of the pRb/E2f interaction and deregulation of E2f activity was the primary driver of tumor initiation in the affected tissues.

Rb1⁶⁵⁴ Retains Tumor Suppressor Activity in the Prostate.

Rb1^654/654 mice are embryonic lethal (18), precluding analysis of adult mice. However, prostate tissue can be rescued from Rb1 null embryonic urogenital sinus tissue by subrenal transplantation (28). Rescued Rb1^−/− prostate tissue is prone to hormone-induced carcinogenesis (29). Urogenital sinus transplanted from Rb1^654/− embryos is also able to differentiate into prostate tissue (Fig. S1), and we have compared the susceptibility of Rb1^−/− and Rb1^654/− rescued prostate tissue to hormone-induced carcinogenesis.

The majority of prostate epithelium in rescued grafts is normal, with localized regions of epithelial hyperplasia in about 75% of cases. Rb1 status does not affect the incidence of hyperplasia. Eight weeks of testosterone plus estradiol (T+E2) treatment increases the incidence of hyperplasia to 100% of grafts for all genotypes and also increases the extent of hyperplasia. Consistent with previous reports, mouse prostatic intraepithelial neoplasia (mPIN) is observed in 83% of treated Rb1^−/− grafts, but not in wild-type grafts. In contrast, mPIN lesions are present in only 13% of treated Rb1^654/− grafts (Table 1). Similar results are observed upon extending T+E2 treatment to 16 wk, the maximum length of treatment due to ensuing tissue atrophy.

Table 1.

Histopathology of rescued or recombined prostate tissue

Genotype	rUGM	Treatment (wk)	Hyperplasia	mPIN
Rb1+/+	—	—	7/9	0/9
Rb1−/−	—	—	9/12	0/12
Rb1654/−	—	—	7/10	0/10
Rb1+/+	—	T+E2 (8)	12/12	0/12
Rb1−/−	—	T+E2 (8)	24/24	20/24
Rb1654/−	—	T+E2 (8)	17/17	2/17
Rb1+/+	—	T+E2 (16)	10/10	0/10
Rb1−/−	—	T+E2 (16)	12/12	10/12
Rb1654/−	—	T+E2 (16)	15/15	2/15
Rb1+/+	+	—	8/14	0/14
Rb1−/−	+	—	15/15	0/15
Rb1654/−	+	—	9/14	0/14
Rb1+/+	+	T+E2 (8)	21/21	0/21
Rb1−/−	+	T+E2 (8)	19/19	11/19
Rb1654/−	+	T+E2 (8)	20/20	0/20

The fraction of grafts exhibiting the indicated histotypes was determined by examining sections throughout the entire graft at 90-μM intervals. rUGM indicates grafts recombined with wild-type rat urogenital mesenchyme.

To assess possible effects of Rb1-deficient mesenchyme on carcinogenesis, we have analyzed E+T2-induced carcinogenesis in grafts generated by recombining urogenital sinus tissue with wild-type rat urogenital mesenchyme. In recombined grafts, the stromal component of the rescued tissue is composed primarily of wild-type rat cells, whereas prostate epithelium is of mouse origin (28). Eight weeks of T+E2 treatment induces hyperplasia in all recombined grafts and mPIN in 58% of Rb1^−/− grafts. No mPIN is observed in recombined Rb1^654/− or wild-type grafts (Table 1). Thus wild-type rat mesenchyme generally suppresses carcinogenesis, but the difference between Rb1^654/− and Rb1^−/− prostate epithelium in sensitivity to T+E2-induced carcinogenesis remains. The data suggest pRb⁶⁵⁴ is able to suppress progression of hormone-induced hyperplasia to neoplasia.

The length of these experiments is limited, thus it is unclear whether mPIN lesions will progress to carcinoma. To bypass this limitation, we tested pRb⁶⁵⁴ tumor suppressor activity in an autochthonous mouse model of prostate cancer. Prostate-specific deletion of Rb1 and Trp53 causes metastatic carcinoma with neuroendocrine features (7, 30). We compared prostate tumorigenesis in Rb1^F/F:Trp53^F/F:PB-Cre4 (Rb1⁻) and Rb1^654/F:Trp53^F/F:PB-Cre4 (Rb1⁶⁵⁴) mice, where F represents floxed alleles and PB-Cre4 drives cre expression in the prostate epithelium (31). Affected Rb1⁻ prostate epithelial cells (PrE) will lack both pRb and p53, whereas Rb1⁶⁵⁴ cells will express pRb⁶⁵⁴. This was verified by PCR and immunostaining (Fig. S2). Survival of Rb1⁶⁵⁴ mice was significantly longer than Rb1⁻ mice, with 50% of Rb1⁶⁵⁴ mice alive at an age when all Rb1⁻ mice had succumbed to prostate cancer (Fig. 2A). Lethal prostate tumors did eventually arise in Rb1⁶⁵⁴ mice, and the histopathology of tumors arising in both genotypes was similar (Fig. S3A). Both genotypes showed evidence of metastasis to the liver (Rb1⁻ 57% and Rb1⁶⁵⁴ 77%), lung (Rb1⁻ 79% and Rb1⁶⁵⁴ 76%), and lymph node (Rb1⁻ 36% and Rb1⁶⁵⁴ 35%). Prostate tumors arising in Rb1⁶⁵⁴ mice expressed pRb exclusively from the Rb1⁶⁵⁴ allele (Fig. 2 B and C).

Fig. 2.

Prostate tumorigenesis in Rb1-deficient mice. (A) The graph displays cumulative survival for Rb1⁻ and Rb1⁶⁵⁴ mice (sample size = n). Mice dying for reasons unrelated to tumorigenesis have been censored from the data and are indicated by cross hatches. Survival of Rb1⁶⁵⁴ mice is significantly longer than Rb1⁻ mice by log rank test (P < 0.001). (B) Rb1 protein expression was analyzed in tissues from mice of the indicated genotypes by Western blotting. The letter B is normal brain tissue and serves as a positive control. LM, NM, and P represent liver metastasis, lymph node metastasis, and primary prostate tumor, respectively. Hsp70 serves as a protein loading control. (C) RNA was extracted from the indicated tissues of Rb1⁶⁵⁴ mice, amplified by RT-PCR, and sequenced. The chromatogram is shown for the sequence surrounding codon 654 (wild type, CGA; Rb1⁶⁵⁴, TGG).

Invasive prostate adenocarcinoma is detectable at 16 wk of age in Rb1⁻ mice, and tumors first develop from mPIN-like precursor lesions in the periurethral proximal prostate (30). We have graded the histopathology of prostate tissue from a cohort of mice at 14 through 22 wk of age to determine whether pRb⁶⁵⁴ affects early preneoplastic or mPIN lesions. Both Rb1⁶⁵⁴ and Rb1⁻ mice exhibit focal cellular atypia and mPIN lesions with comparable frequency at 16 wk (Table 2). The histology of mPIN arising in mice of both genotypes is similar (Fig. S3B). The mPIN lesions progress to invasive carcinoma in about 25% of Rb1⁻ mice at 16 wk, increasing to about 60% by 22 wk. Despite the similar incidence of focal cellular atypia and mPIN, no invasive carcinoma is detected in Rb1⁶⁵⁴ mice by 22 wk. We have examined older mice (n = 10 per genotype), and no carcinoma is detected in Rb1⁶⁵⁴ mice as old as 26 wk of age. These findings indicate that pRb⁶⁵⁴does not prevent initiation of preneoplastic and early neoplastic lesions, but does slow progression to invasive prostate carcinoma.

Table 2.

Histopathology of prostate tissue from Rb1-deficient mice

Genotype	Age (wk)	Normal	Focal atypia	mPIN	Carcinoma
Rb1−	14	1/3	0/3	2/3	0/3
Rb1654	14	3/3	0/3	0/3	0/3
Rb1−	16	0/11	3/11	5/11	3/11
Rb1654	16	1/10	4/10	5/10	0/10
Rb1−	18	0/6	0/6	4/6	2/6
Rb1654	18	2/4	0/4	2/4	0/4
Rb1−	20	0/6	0/6	4/6	2/6
Rb1654	20	2/9	6/9	1/9	0/9
Rb1−	22	0/5	0/5	2/5	3/5
Rb1654	22	0/5	1/5	4/5	0/5

Proximal prostate tissue from age-matched mice of the indicated genotype was dissected, and H&E stained sections throughout the entire gland at 50-μM intervals were examined. Each mouse was scored with the most advanced lesion detected. The fraction of mice with the indicated histopathology is shown.

pRb⁶⁵⁴ Tumor Suppressor Activity Is Associated with Cellular Senescence.

We isolated PrE cell cultures from both rescued and native tissue to explore cellular mechanisms underlying the differences in tumor progression observed. We successfully cultured PrE rescued from Rb1^+/+, Rb1^−/−, and Rb1^654/654 embryos. We also successfully cultured PrE from the proximal prostate of Rb1^F/F:Trp53^F/F:PB-Cre4 or Rb1^654/F:Trp53^F/F:PB-Cre4 adult mice. The identity and purity of the PrE cell cultures was verified by cytokeratin expression, androgen receptor expression, and genotyping of Trp53 and Rb1 (Fig. S4). Under optimal culture conditions, no consistent difference in proliferation of PrE cultures was observed that correlated with Rb1 status (Fig. S5). We did note that wild type or pRb⁶⁵⁴ expressing PrE plated at low density exhibited cells with an enlarged and flattened morphology reminiscent of cellular senescence.

To explore this finding, we assayed senescence-associated β-galactosidase activity (SA-bgal) in PrE cultures subjected to senescence-inducing stresses. PrE rescued from Rb1^+/+ or Rb1^654/654 embryos showed a significant percentage of SA-bgal positive cells when treated with H₂O₂ or when cultured without serum (Fig. 3 A and B). Little or no detectable SA-bgal staining was observed in PrE rescued from Rb1^−/− embryos. PrE isolated from Rb1^654/F:Trp53^F/F:PB-Cre4 mice, but not from Rb1^F/F:Trp53^F/F:PB-Cre4 mice, also exhibited SA-bgal staining upon H₂O₂ treatment (Fig. 3C). The intensity of SA-bgal staining was lower in PrE from Rb1^654/F:Trp53^F/F:PB-Cre4 mice compared with PrE rescued from Rb1^654/654 embryos. This might be due to the different Trp53 status or Rb1⁶⁵⁴ allele copy number in these cells. Consistent with induction of cellular senescence, the fraction of proliferating cells was significantly lower in H₂O₂-treated pRb⁶⁵⁴ expressing PrE than in PrE completely lacking pRb (Fig. 3D). These observations demonstrated that PrE expressing pRb or pRb⁶⁵⁴ were capable of undergoing a senescence response to H₂O₂ treatment or serum deprivation, whereas PrE completely lacking pRb were not.

Fig. 3.

Senescence in PrE cultured in vitro. (A) PrE rescued from embryos of the indicated genotypes were cultured in vitro under optimal growth conditions, in the absence of serum, or upon treatment with H₂O₂. Cells were stained for SA-bgal activity and representative images captured. (Scale bar, 25 μM.) (B) The fraction of PrE of the indicated genotypes staining positive for SA-bgal activity was quantitated with or without H₂O₂ treatment. The results show the mean and SD for three independent experiments. The percentage of SA-bgal positive cells is significantly higher in H₂O₂-treated Rb1^654/− (P < 0.001) or wild-type PrE (P < 0.001) than in treated Rb1^−/− PrE by Student's t test. (C) PrE isolated from proximal prostate tissue of Rb1⁶⁵⁴ or Rb1⁻ adult mice were treated with H₂O₂ and the fraction of SA-bgal positive cells counted. SA-bgal staining is not observed in untreated cells. The results show the mean and SD of three independent experiments. The percentage of SA-bgal positive cells is significantly higher in Rb1⁶⁵⁴ PrE than in Rb1⁻ PrE (P = 0.004 by Student's t test). (D) PrE from C with or without H₂O₂ treatment were pulse labeled with BrdU and the fraction of cells incorporating BrdU counted. The results show the mean and SD of three independent experiments. The percentage of BrdU-labeled cells is significantly lower in H₂O₂-treated Rb1⁶⁵⁴ PrE than in similarly treated Rb1⁻ PrE (P = 0.003 by Student's t test).

The ability of pRb⁶⁵⁴-expressing PrE to undergo senescence in vitro suggests the possibility that it also retains senescence-associated functions in vivo. To test this, we analyzed sections from both rescued and native prostate tissue for evidence of senescence. Little or no detectable SA-bgal staining was observed in proximal prostatic ducts from wild-type or Rb1⁻ mice, but SA-bgal staining was observed in some ducts from Rb1⁶⁵⁴ mice (Fig. 4A). Primary prostate tumors from Rb1⁶⁵⁴ mice also showed rare SA-bgal positive cells, but SA-bgal staining was not detectable in tumors from Rb1⁻ mice (Fig. 4B). SA-bgal staining was also observed in T+E2-treated prostate tissue rescued from Rb1^+/+ or Rb1^654/− embryos, but not from Rb1^−/− embryos (Fig. 4C).

Fig. 4.

Senescence in prostate tissue in vivo. (A) Frozen tissue sections showing proximal prostatic ducts from mice of the indicated genotypes were stained for SA-bgal activity and counterstained with eosin. Representative duct cross-sections are marked with the letter D. (Scale bar, 50 μm.) (B) Frozen sections of prostate tumors from Rb1⁶⁵⁴ or Rb1⁻ mice were stained for SA-bgal and counterstained with eosin. (Scale bar, 200 μm.) (C) Frozen sections of rescued prostate tissue of the indicated genotype treated with T+E2 were stained for SA-bgal activity. Representative ducts are marked with the letter D. (Scale bar, 50 μm.)

Another marker of cellular senescence is expression of the cyclin-dependent kinase inhibitor p16INK4a (p16). Immunostaining of tissue sections for p16 indicated that most cells within proximal mPIN lesions from 18- to 22-wk-old Rb1⁶⁵⁴ mice were positive for p16. In contrast, few cells within Rb1⁻ mPIN lesions showed detectable p16 staining (Fig. 5A). Consistent with induction of senescence, the fraction of proliferating cells was significantly lower in proximal mPIN lesions from Rb1⁶⁵⁴ mice compared with Rb1⁻ mice (Fig. 5B). Primary and metastatic tumors from Rb1⁶⁵⁴ mice also had a lower fraction of proliferating cells, although these smaller differences did not reach statistical significance. We also noted significant p16 staining in foci of atypical, preneoplastic cells in distal prostate epithelium from Rb1⁶⁵⁴ mice, whereas similar atypical cells from Rb1⁻ mice lacked detectable p16 staining (Fig. 5C).

Fig. 5.

Cell proliferation and p16 immunostaining in prostate tissue. (A) Proximal prostate tissue sections from an 18-wk-old Rb1⁻ mouse or a 22-wk-old Rb1⁶⁵⁴ mouse (both selected for the presence of mPIN lesions) as well as a wild-type mouse, were immunostained for p16 and counterstained with hematoxylin. P marks ducts with involved mPIN lesions; D marks a normal duct. (Scale bar, 50 μm.) (B) Consecutive sections from tissue in A were immunostained for pH3 and counterstained with hematoxylin. P marks ducts with involved mPIN lesions. The graph at Right depicts the percentage of pH3 positive cells in mPIN lesions, primary prostate tumors, or metastatic tumors from mice of the indicated genotype. The mean and SD are shown for 29 Rb1⁻ mPIN lesions, 15 Rb1⁶⁵⁴ lesions, two primary tumors for each genotype, and multiple metastatic tumors from two different mice for each genotype. The asterisk marks a statistically significant difference between genotypes (Student's t test, P < 0.05). (C) Prostate ventral lobe sections from mice of the indicated genotype were immunostained for p16 and counterstained with hematoxylin. The box outlines the area magnified in the Lower panel. The arrow indicates a representative atypical preneoplastic cell positively stained for p16. The arrowhead shows an atypical cell lacking p16 staining. (Scale bars, 50 μm.)

Regulation of E2f Target Genes in Prostate Tissue.

Whereas pRb⁶⁵⁴ is deficient in binding E2f in PrE (Fig. S6), low residual E2f binding may still repress expression of some E2f target genes. To test this, we have compared E2f target gene expression in Rb1⁶⁵⁴ or Rb1⁻ prostate tissue from 17-wk-old mice. Seventy-seven of 82 established E2f cell cycle or senescence-associated target genes tabulated by Bracken et al. (32) fail to show significant differences in relative expression between Rb1⁶⁵⁴ and Rb1⁻ prostate tissue (P ≤ 0.05, Student's t test). These genes include those uniquely repressed by pRb during oncogene-induced cellular senescence in vitro such as CCNE1, PCNA, DHFR, and MCM (33). Of the five genes that do show a difference in relative gene expression, four have increased rather than decreased expression in Rb1⁶⁵⁴ tissue, including Rb1 as expected. We have verified the microarray data for some of these relevant E2F target genes (Fig. 6). Thus we find no evidence that pRb⁶⁵⁴ is able to repress the expression of E2f cell cycle or senescence-associated target genes.

Fig. 6.

E2f target gene expression in prostate tissue. RNA was extracted from prostate tissue of the indicated genotype and the expression levels of the indicated genes determined by quantitative RT-PCR. The data represent the fold change in expression relative to a wild-type prostate tissue reference sample, normalized to β-actin RNA levels. The mean and SD of data from two mice for each genotype, each performed in triplicate, are shown. The asterisk marks a statistically significant difference between genotypes (Student's t test, P < 0.05).

Discussion

The importance of the pRb/E2f interaction for regulating gene expression and the cell cycle has been established by over two decades of research (34). It is assumed that this interaction mediates pRb tumor suppressor activity, yet this has not been directly tested in vivo. To fill this gap, we have characterized the tumor suppressor activity of pRb⁶⁵⁴, a mutant deficient in binding E2f. PRb⁶⁵⁴ has no detectable tumor suppressor activity in endocrine tissue. Further, compound loss of E2f1 significantly reduces the incidence of endocrine tumors in pRb⁶⁵⁴-expressing mice. These data provide direct in vivo evidence implicating disruption of the pRb/E2f interaction as the primary molecular defect causing tumor initiation in the affected tissues.

Surprisingly, pRb⁶⁵⁴ retains detectable tumor suppressor activity in two mouse models of prostate cancer. In both models, pRb⁶⁵⁴ fails to prevent initiation of preneoplastic or early neoplastic lesions. This is consistent with the R654W mutation fully disrupting the ability of pRb to prevent tumor initiation. However, the progression of these early lesions to invasive carcinoma and lethal disease is markedly slowed by pRb⁶⁵⁴. We conclude that the R654W mutation does not fully disrupt the ability of pRb to restrain prostate tumor progression. Whether all Rb1⁶⁵⁴ mice would eventually die from prostate cancer cannot be unequivocally determined as they succumb to pituitary and thyroid tumors beginning at ≈40 wk of age. These endocrine tumors arise at a rate similar to those in Rb1^−/+ or Rb1^654/+ mice, or Rb1^654/F:Trp53^F/F mice lacking the cre transgene. Thus tumorigenesis is initiated by spontaneous loss of the floxed wild-type Rb1 allele rather than cre-mediated recombination. This observation highlights within individual mice the difference in Rb1⁶⁵⁴ tumor suppressor activity in different tissues. This observation supports the emerging hypothesis that pRb uses multiple, context-dependent mechanisms to suppress tumorigenesis.

In both mouse models, suppression of prostate tumor progression by pRb⁶⁵⁴ is associated with cellular senescence. PrE expressing pRb⁶⁵⁴ or wild-type pRb, but not PrE lacking pRb, are also capable of undergoing senescence in vitro. The requirement for pRb in PrE senescence is in contrast to previous observations in murine fibroblasts and may be due to lack of compensatory up-regulation of p107 or p130 in pRb null PrE (Fig. S7). Deregulated E2f activity is a known trigger of cellular senescence (35), and pRb is known to be important for enforcing the senescence response (36). Our results suggest that loss of pRb/E2f interaction in Rb1⁶⁵⁴ or Rb1⁻ prostate epithelium triggers a senescence response. The response is not enforced in Rb1⁻ epithelium, permitting initiated tumor cells to progress. Tumor progression is slowed in Rb1⁶⁵⁴ epithelium because pRb⁶⁵⁴ retains sufficient function to enforce a senescence response.

The best characterized mechanism contributing to pRb-mediated cellular senescence involves pRb/E2f interaction and repression of E2f target genes (33, 37). However, we find no evidence that pRb⁶⁵⁴ is able to repress E2f cell cycle or senescence-associated target genes. We conclude that functions beyond pRb/E2F contribute to the senescence-associated suppression of tumor progression observed. Additional study will be required to identify these mechanisms. We note that despite continued expression of pRb⁶⁵⁴, advanced and metastatic prostate tumors from Rb1⁶⁵⁴ mice exhibit less cellular senescence than early lesions. Whatever the mechanisms responsible, evolving tumors have a means to bypass or abrogate them in a way that does not depend on loss of the Rb1⁶⁵⁴ allele.

The genetics of human retinoblastoma demonstrate that Rb1 loss plays an important role in tumor initiation. The results presented here support the hypothesis that disruption of the pRb/E2f interaction and deregulation of E2f activity is the primary molecular defect driving tumor initiation in the context of Rb1 mutation. In some human cancers, however, pRb loss occurs only at later stages of disease. A significant unresolved question is whether pRb loss drives the progression of previously initiated neoplasia and, if so, whether disruption of the pRb/E2f mechanism is responsible. By genetically separating pRb-mediated effects on tumor initiation from effects on tumor progression, the R654W Rb1 allele provides direct in vivo experimental evidence implicating Rb1 loss as a driver of tumor progression in addition to its effects on tumor initiation. Although the data do not address whether disruption of pRb/E2f affects tumor progression, they do suggest that additional mechanisms beyond pRb/E2f contribute to the ability of pRb to restrain tumor progression and suppress tumorigenesis. The observed pRb-mediated effects on tumor progression are tissue specific. Thus the impact of Rb1 alteration on tumorigenesis is context dependent, varying both on the type of defect and the cell where it occurs. This has important implications for our understanding of how pRb deficiency contributes to cancer. A more thorough appreciation of the variety of molecular mechanisms contributing to pRb-mediated tumor suppression will advance efforts to use this important regulatory network for the diagnosis and treatment of cancer.

Materials and Methods

Mice.

Derivation and genotyping of mice have been previously described (7, 18, 23). Experimental mice are on a mixed C57BL/6, 129/SVJae, and FVB genetic background. For survival analysis, mice were monitored daily, euthanized when moribund, and necropsy performed to verify cancer diagnosis. Survival analysis using the Kaplan–Meier method was performed with SPSS version 17. All animal work was performed under Roswell Park Cancer Institute or Vanderbilt University Medical Center institutional animal care and use committee approved protocols.

Tissue Rescue and Isolation of PrE Cultures.

Prostate tissue rescue and recombination using urogenital sinus tissue from embryos at 13 d of gestation were performed as previously described (29). Hormone treatment was by implantation of Silastic capsules filled with 25 mg of testosterone and 5 mg of 17β-estradiol (Sigma) at the time of grafting.

For isolation of prostate epithelial cell cultures, tissue grafts or native tissue dissected from the proximal prostate of 60-d-old mice were minced and plated onto collagen-coated culture dishes in DMEM (BioWhittaker) supplemented with 2.5% charcoal stripped FCS, 5 μg/mL of insulin/transferring/selenium (Collaborative Research), 10 μg/mL of bovine pituitary extract (Sigma), 10 μg/mL of epidermal growth factor (Collaborative Research), 1 μg/mL of cholera toxin (Sigma), 100 U/mL of penicillin G and streptomycin (BioWhittaker). Two weeks later, areas of epithelial cells were dissected and digested with collagenase, trypsinized, and replated. Epithelial cells were further purified from remaining fibroblasts as described (38). Senescence experiments were performed using a senescence detection kit (Calbiochem). For BrdU-labeling experiments, cultured cells were treated with 1.0 or 2.0 mM of H₂O₂ for 24 h with equivalent results, allowed to recover for 2 d, pulse labeled with BrdU for 3 h, and then fixed and stained using a cell proliferation assay kit as recommended (GE Healthcare).

Histology and Immunostaining.

Prostate tissue was fixed in 10% neutral buffered formalin or phosphate-buffered 4% paraformaldehyde, paraffin embedded, and serially sectioned at 5-μm thickness through the entire specimen. Every 10th section was H&E stained for assessment of histopathology. Pathology was verified by a pathologist (M.W.) and a veterinary pathologist (A.Y.N.). Primary antibodies used for immunostaining were antiserine 10 phosphorylated histone H3 (Upstate) and p16Ink4A (Santa Cruz Biotechnology). Stain was developed using biotinylated goat anti-rabbit secondary antibody and the ABC reagent as recommended (Vector Laboratories). For senescence associated β-galactosidase staining, tissue specimens were embedded in OCT and cryosections prepared at 5-μm depth. Sections were fixed and stained as recommended using a senescence detection kit (Calbiochem).

RNA and Protein Analysis.

RNA was extracted from prostate tissue using the RNeasy method (Qiagen). RNA was reversed transcribed for PCR amplification and DNA sequencing using the SuperScript first-strand synthesis system (Invitrogen). The Rb1 PCR primers used were 5′-GTGTAAATTCTGCTGCAAAT-3′ and 5′-GGTCCAAATGTCGGTCTCTC-3′. Two-color gene expression profiling, using 44K whole mouse genome oligo microarrays (Agilent Technologies), was performed as previously described (39). Prostate tissue RNA from two Rb1⁶⁵⁴ mice or two age-matched Rb1⁻ mice was individually profiled against prostate RNA pooled from three wild-type mice. The microarray data have been deposited in the GEO repository, accession no. GSE25615. Quantitative RT-PCR was performed using the Brilliant II 1-step high ROX master mix (Agilent Technologies). The 5′ nuclease quantitative PCR assays (Integrated DNA Technologies) were run using an ABI 7300 real-time PCR machine. Relative expression was calculated with Sequence Detection software, version 1.4, using the ddCt method. The endogenous control was β-actin. The sequences of quantitative PCR primers and probes are available upon request. Protein was extracted from tissue or cultured cells and analyzed by Western blotting as previously described (40). Blots were stained with primary antibodies against pRb (Pharmingen) or Hsp70 (Stressgen).