Prostate Cancer Induced by Loss of Apc Is Restrained by TGFβ Signaling

Recent work with mouse models of prostate cancer (CaP) has shown that inactivation of TGFβ signaling in prostate epithelium can cooperate with deletion of the Pten tumor suppressor to drive locally aggressive cancer and metastatic disease. Here, we show that inactivating the TGFβ pathway by deleting the gene encoding the TGFβ type II receptor (Tgfbr2) in combination with a deletion of the Apc tumor suppressor gene specifically in mouse prostate epithelium, results in the rapid onset of invasive CaP. Micro-metastases were observed in the lymph nodes and lungs of a proportion of the double mutant mice, whereas no metastases were observed in Apc single mutant mice. Prostate-specific Apc;Tgfbr2 mutants had a lower frequency of metastasis and survived significantly longer than Pten;Tgfbr2 double mutants. However, all Apc;Tgfbr2 mutants developed invasive cancer by 30 weeks of age, whereas invasive cancer was rarely observed in Apc single mutant animals, even by one year of age. Further comparison of the Pten and Apc models of CaP revealed additional differences, including adenosquamous carcinoma in the Apc;Tgfbr2 mutants that was not seen in the Pten model, and a lack of robust induction of the TGFβ pathway in Apc null prostate. In addition to causing high-grade prostate intra-epithelial neoplasia (HGPIN), deletion of either Pten or Apc induced senescence in affected prostate ducts, and this restraint was overcome by loss of Tgfbr2. In summary, this work demonstrates that TGFβ signaling restrains the progression of CaP induced by different tumor suppressor mutations, suggesting that TGFβ signaling exerts a general tumor suppressive effect in prostate.


Introduction
Prostate cancer (CaP) is one of the leading causes of cancer death in men, with the NCI predicting more than 230,000 cases and 29,000 deaths in the US in 2013 (www.cancer.gov/ cancertopics/types/prostate). Several signaling pathways are frequently disrupted in human CaP, either by genetic alterations or by changes in expression of key components of the pathway [1]. Mutation or deletion of the PTEN tumor suppressor gene is found in more than 30% of primary human prostate cancers and more than 60% of CaP metastases [2][3][4][5]. Loss of PTEN lipid phosphatase activity activates the PI3-kinase signaling pathway [6,7], including activation of the downstream Akt/PKB kinase, which is itself an oncogene [8][9][10]. Akt activation promotes cell survival by inhibiting apoptosis, and also regulates proliferation and cell size [11,12].
Transforming growth factor (TGF) b ligands assemble a signaling complex of type I and type II receptors, in which the type II receptor phosphorylates and activates the type I [13][14][15]. The activated type I receptor then phosphorylates receptor Smad (R-Smad) proteins, which are the key transcription factors that mediate TGFb family signaling. Smad2 and Smad3 are the primary R-Smads that respond to TGFb signaling. Phosphorylated R-Smads bind Smad4 and accumulate in the nucleus where they regulate gene expression [16]. In many cell types, including epithelial cells, TGFb signaling via Smad2/3 causes a G1 cell cycle arrest that prevents uncontrolled proliferation and plays a tumor suppressive role [17,18]. The TGFb signaling pathway is frequently disrupted by mutation or loss of expression of pathway components in human CaP [19][20][21]. Reduced expression of the TGFb type I and type II receptors (encoded by the TGFBR1 and TGFBR2 genes) is associated with increased Gleason score and decreased survival, and reduced SMAD4 expression is also found in advanced human CaP [19,[22][23][24].
The APC (Adenomatous Polyposis Coli) gene encodes a tumor suppressor that is inactivated in a large proportion of human colorectal cancers [25]. Loss of APC function activates downstream components of the canonical WNT signaling pathway [26,27]. APC functions as a scaffold that targets b-catenin for phosphorylation and subsequent proteasomal degradation. In the presence of WNT ligands, or a mutation in the APC gene, bcatenin is no longer degraded and can accumulate in both the cytosol and the nucleus, where it activates target gene expression primarily via interaction with the TCF/LEF DNA binding transcription factors [28]. Although the APC gene encodes a major tumor suppressor gene for colorectal cancer, and APC mutations have been found in some other cancers, inactivating mutations in the APC gene are rare in human CaP [29]. However, in the majority of cases of advanced human CaP b-catenin is found in the nucleus, suggesting that this pathway is frequently deregulated. Other mechanisms for activation of this pathway have been identified in human CaP, including methylation of the APC gene [30,31]. Activating mutations in b-catenin itself (encoded by CTNNB1), that prevent phosphorylation and targeting to the proteasome, have also been identified [32]. While de-regulation of this pathway appears to be quite frequent in human CaP, the importance of nuclear b-catenin in the initiation of human CaP and progression to castration resistant prostate cancer (CRPC) remains to be elucidated.
Recent work with genetically engineered mouse models has begun to shed light on the combinatorial effects of tumor suppressor mutations in prostate cancer progression. Prostatespecific deletions of the Tgfbr2 and Smad4 genes have both been tested in mice. Neither mutation alone is sufficient to initiate tumorigenesis, but when combined with a Pten deletion, inactivation of the TGFb pathway results in very rapid progression to locally invasive and metastatic disease [33,34]. Mouse models in which a stabilized b-catenin transgene was expressed in the prostate resulted in high-grade prostate intra-epithelial neoplasia (HGPIN) [35,36]. Deletion of the Apc gene specifically in mouse prostate epithelium also results in HGPIN with high penetrance, but this rarely progresses to invasive cancer, and metastases were not found [37]. Here we show that deletion of both the Tgfbr2 and Apc genes in mouse prostate epithelium results in rapid progression to invasive cancer, with metastases to lymph nodes and lungs in some cases. Additionally, we show that deletion of either Apc or Pten alone induces senescence in prostate ducts with HGPIN and that additional loss of the Tgfbr2 gene overcomes this senescence checkpoint. In summary, this work suggests that TGFb signaling limits progression from HGPIN to invasive prostate cancer, irrespective of the tumor initiating mutation. Thus TGFb signaling plays a key tumor suppressive role in prostate epithelium.

Results and Discussion
Lethal prostate cancer in Apc r/r ;Tgfbr2 r/r mutants Homozygous deletion of Apc in mouse prostate epithelium results in HGPIN with squamous differentiation in all prostatic lobes, although it rarely progresses to locally invasive cancer, and metastases have not been detected [37]. TGFb signaling restrains the progression of prostate cancer initiated by loss of the tumor suppressor Pten [33,34]. To test whether TGFb signaling plays a similar role in prostate when Apc is lost, we used mouse models to combine mutations in the Tgfbr2 and Apc genes. Conditional, loxPflanked alleles of both the Apc and Tgfbr2 genes were combined with the PbCre4 transgene, which is expressed only in prostate epithelium [38]. The recombined alleles, generated by prostate epithelium-specific expression of PbCre4, are referred to hereafter as Apc r/r and Tgfbr2 r/r . As a first test of whether combining the Apc and Tgfbr2 mutations promotes prostate cancer progression, we followed a cohort of Apc r/r and Apc r/r ;Tgfbr2 r/r mice to one year of age. Over this time course, none of the Apc r/r mice showed signs of distress, whereas all of the double mutants displayed a tumor burden that required euthanasia prior to 30 weeks of age ( Figure 1A). For comparison, we also analyzed a number of Pten r/r and Pten r/r ;Tgfbr2 r/r mice over the same period. All but one of the Pten single mutants survived to one year, whereas none of the double mutants survived beyond 16 weeks, consistent with our previous analysis ( Figure 1A and [33]). At 18 weeks of age, the prostates of Apc single null mutants were almost indistinguishable from the wild-type, whereas, prostatic enlargement was readily apparent in Pten nulls ( Figure 1B). By 52 weeks, the Apc null prostate tumors were clearly evident, but were still much smaller than those in Apc;Tgfbr2 double null mice at 16 weeks of age ( Figure 1B). Thus, deletion of the Tgfbr2 gene in the background of loss of either Pten or Apc had a highly significant effect on tumor growth, and reduced the median survival times to 82 days and 148 days respectively ( Figure 1A and C). There was also a significant survival difference between the two double mutant combinations that is likely due to the specific pathways affected by loss of Pten and Apc.
We previously found that a relatively high proportion (, 66%) of Pten r/r ;Tgfbr2 r/r mice developed metastases to the lumbar lymph nodes or lung [33]. In the present study, micro-metastases were found in the lumbar lymph nodes of 18% of the Apc r/r ;Tgfbr2 r/r mice, and in the lungs of 12% of the Apc r/r ;Tgfbr2 r/r mice. Thus, metastases in the Apc r/r ;Tgfbr2 r/r mice were significantly less frequent (p,0.01) than in the Pten r/r ;Tgfbr2 r/r mutants examined here, and in our previous analysis ( Figure 1C). We never observed metastases in Apc r/r single mutant mice, even by one year of age, and only one of fourteen Apc r/r mice analyzed at either 36 or 52 weeks of age showed signs of locally invasive cancer. These data clearly show that the combination of mutations in the Apc and Tgfbr2 genes can accelerate tumor progression to invasive and metastatic disease over that seen with loss of Apc alone.
Inactivation of TGFb signaling, either by deletion of Tgfbr2 or Smad4, accelerates the progression of Pten null CaP, and Tgfbr2 deletion has also been shown to cooperate with a constitutively active Akt transgene to drive invasive cancer [33,34]. Expression of a dominant negative TGFb type II receptor transgene in the prostate was able to increase the severity of tumors initiated by prostate specific expression of T antigen, and resulted in increased metasases [39]. In each case, progression to invasive and metastatic disease is accelerated by loss of TGFb signaling, even when the initiating mutation does not on its own result in metastases, as seen here with the Apc mutants. In summary, the results of these mouse models of CaP together with the data presented here suggest that TGFb signaling plays a major tumor suppressive role in CaP.
Adenosquamous carcinoma in Apc r/r ;Tgfbr2 r/r mutant prostate Given the differences in survival and in the gross appearance of the tumors from Apc r/r ;Tgfbr2 r/r and Pten r/r ;Tgfbr2 r/r mice we performed a more detailed comparison of these tumors, as well as of prostates from the single null mutants. As previously reported, histological analysis did not reveal any differences between the wild type and Tgfbr2 null prostates ( Figure 2 and [33]). Comparison of the Pten and Apc single mutants revealed HGPIN in prostates of both genotypes, with squamous differentiation in the Apc null that was not seen in the Pten null. Introduction of a Tgfbr2 mutation resulted in progression to poorly differentiated adenocarcinoma (PDA) in all Pten;Tgfbr2 double mutants. When combined with the Apc mutation, deletion of the Tgfbr2 gene in prostate resulted in progression to invasive cancer, although the majority of Apc r/r ;Tgfbr2 r/r tumors retained the squamous phenotype seen in Apc single nulls ( Figure 2). Note the prominent foci of keratin in the Apc null HGPIN (arrow, Figure 2) and adenosquamous differentiation indicated in the Apc r/r ;Tgfbr2 r/r tumor shown in Figure 2 (arrowheads). In addition to the histological appearance, increased expression of keratins 1 and 10 has been used as a marker of squamous differentiation in the Apc null prostate [40]. We, therefore, stained sections of wild type and mutant prostate for keratin 10. As shown in figure 3, strong staining for Krt10 was observed in cells underlying the prominent foci of keratin present in Apc r/r HGPIN, with little or no staining evident in wild type, Tgfbr2 null or Pten null prostates. In the Apc r/r ; Tgfbr2 r/r tumors large regions of Krt10 positive cells were present throughout the tumor, whereas in the Pten;Tgfbr2 double mutants only rare scattered Krt10 positive cells were seen ( Figure 3).
To better compare the frequencies of the phenotypes described above we compiled data on the predominant phenotypes in each genotype as scored in a blinded manner based on established mouse prostate phenotypes [41][42][43]. In Apc r/r mice most prostates analyzed at 8-24 weeks of age had HGPIN with squamous differentiation (adenosquamous HGPIN; Asq-HGPIN) and by 36-52 weeks all had this phenotype ( Figure 4A). Only one of the Apc nulls developed regions of locally micro-invasive cancer (at 36 weeks), and this was significantly less frequent (p,0.02) than the incidence of invasive cancer in Pten r/r mice in the 36-52 week age range ( Figure 4A). Comparison of the Tgfbr2 compound mutants revealed that all Pten r/r ;Tgfbr2 r/r mice had PDA without squamous differentiation, whereas all but one of the Apc r/r ;Tgfbr2 r/r mice euthanized for tumor burden had extensive adenosquamous carcinoma ( Figure 4B). We also compared phenotypes in mice with intermediate genotypes, in which one or other mutation was heterozygous (some of the Pten +/r ;Tgfbr2 r/r and Pten r/r ;Tgfbr2 +/r mice included in this phenotype summary were analyzed for survival in ref [33]). For this analysis, mice were analyzed when tumor burden became excessive, or when mice reached at least one year of age. This showed that the frequency of invasive cancer was significantly higher in Pten r/r ;Tgfbr2 +/r than in Apc r/r ;Tgfbr2 +/r mice (p,0.05), and in Pten +/r ;Tgfbr2 r/r than in Apc +/r ;Tgfbr2 r/r prostates (p,0.001), supporting the idea that loss of Pten results in a more aggressive phenotype than loss of Apc ( Figure 4B).
In summary, this analysis suggests that the phenotypes observed are quite specific to each prostate cancer model, and that within each model there is little variation in the type of tumor. For all combinations tested the Pten mutation results in a more aggressive, invasive cancer than Apc deletion. Interestingly, none of the Apc heterozygotes developed tumors, even in the presence of a homozygous deletion of Tgfbr2, whereas the majority of Pten heterozygotes developed PDA if they were null for Tgfbr2. This is consistent with inactivation of the remaining Pten allele, which has been suggested to be a major route by which Pten heterozygous mouse prostates develop a phenotype [44]. TGFb signaling appears to restrain the progression of HGPIN to both poorly differentiated adenocarcinoma and to adenosquamous carcinoma, with the difference in phenotype being dependent on the tumor initiating mutation. While nuclear accumulation of b-catenin is found in a high proportion of human CaP, there is no consensus as to how the APC/b-catenin pathway might be disrupted in human CaP. The squamous differentiation induced by loss of Apc is relatively rare in human CaP, although it is more common after androgen deprivation therapy [45]. The presence of adenosquamous carcinoma in Apc;Tgfbr2 double nulls suggests that loss of TGFb signaling does not prevent the squamous differentiation induced by loss of Apc. Similarly, the combination of a Pten mutation with a stabilized b-catenin transgene in prostate results in tumors with squamous differentiation, suggesting that this may be the dominant phenotype of early activation of b-catenin [40]. Perhaps the increased nuclear b-catenin seen in advanced human CaP is a later phenotype, or occurs to a lower level than seen with transgenic manipulation in mice, and thus cannot drive squamous differentiation. The number of animals for each is shown above. The p-values (log-rank test) are shown for Apc r/r ;Tgfbr2 r/r versus Pten r/r ;Tgfbr2 r/r , for Apc r/ r ;Tgfbr2 r/r versus Apc r/r , and for Pten r/r ;Tgfbr2 r/r versus Pten r/r . (B) Gross anatomy of the prostate tumors is shown for each at the following ages: wild type (18 weeks), Pten r/r (18 weeks), Pten r/r ;Tgfbr2 r/r (13 weeks), Apc r/r (18 and 52 weeks), and Apc r/r ;Tgfbr2 r/r (15 weeks). (C) The survival data for each of the double nulls is shown, along with a summary of metastases to lumbar lymph nodes (LN) and lung. The proportion of mice with metastases is significantly different, comparing the Pten r/r ;Tgfbr2 r/r and Apc r/r ;Tgfbr2 r/r groups in C (p,0.01). doi:10.1371/journal.pone.0092800.g001 Increased stroma and invasive cancer with Tgfbr2 deletion Examination of sections stained with H&E revealed evidence of locally invasive cancer, with increased stroma in both Apc r/ r ;Tgfbr2 r/r and Pten r/r ;Tgfbr2 r/r prostates. To visualize expansion of the stroma, we stained sections with Masson's Trichrome. In wild type and Tgfbr2 null prostates, the duct structure was similar, and there was minimal fibrous tissue between prostatic ducts. In areas of HGPIN in Pten r/r animals there were focal areas of increased fibrosis and in the Apc null significant fibrosis was seen between ducts with HGPIN ( Figure 5). In areas of invasive cancer in both the Apc r/r ;Tgfbr2 r/r and Pten r/r ;Tgfbr2 r/r prostates, increased fibrosis was present, and this was particularly evident in the Apc r/r ;Tgfbr2 r/r mice ( Figure 5).
We next examined the breakdown of duct structure in each of the two prostate cancer models. Immunostaining for Collagen IV, as a marker for basement membrane integrity, demonstrated that the basement membrane was intact in both wild type and Tgfbr2 null prostates. In the Apc and Pten mutants (at 36 and 21 weeks of age respectively), there was minimal basement membrane breakdown around ducts with HGPIN ( Figure 6). The ages of these mice are well beyond the median survival times of 21 and 12 weeks for each of the double mutants. In contrast, examination of double mutant animals that were euthanized due to tumor burden showed complete breakdown of the basement membrane in areas of invasive cancer ( Figure 6). To examine the invasive phenotype, we stained sections for Foxa1, as an epithelial marker, and for Sma to identify the stromal cells. In both the wild type and Tgfbr2 null, prostatic ducts were fully enclosed by Sma positive stromal cells.  Although there was clear evidence of HGPIN in the Apc and Pten null prostates, in most cases the ducts were still surrounded by Sma-positive stroma at ages significantly greater than the median survival times of the double mutants (Figure 7). In contrast, in Apc r/ r ;Tgfbr2 r/r and Pten r/r ;Tgfbr2 r/r animals, the distinct separation between stroma and ducts has clearly broken down, as duct structure is no longer evident upon transition to invasive cancer (Figure 7). Taken together, these analyses suggest that while there are some differences in the histologic type of cancer initiated by loss of Pten or Apc, the additional inactivation of the TGFb pathway causes rapid progression to locally invasive cancer.
To test for evidence of epithelial to mesenchymal transition (EMT) we examined expression of E-cadherin and vimentin. TGFb signaling is a well known driver of EMT and increased TGFb signaling frequently results in increased vimentin expression and a decrease in E-cadherin expression, which can contribute to the breakdown of epithelial cell junctions and an invasive phenotype [46]. As shown in Figure 8, E-cadherin is robustly expressed and present at the cell periphery in the majority of epithelial cells in the Apc and Pten single mutants. E-cadherin expression still appears largely normal in the Pten r/r ;Tgfbr2 r/r mice, despite the clear breakdown of duct structure. Some evidence of Ecadherin de-localization from the cell membrane was seen in the Apc r/r ;Tgfbr2 r/r prostate, together with a reduction in overall signal compared to the Apc r/r . However, even in the two double mutant prostates, epithelial cell junctions appear to be largely intact with clear E-cadherin staining (Figure 8). Vimentin staining revealed little change in in the epithelial cells in any of the mutants, although there was some increased vimentin expression in the stroma of Apc and Pten single null and Apc;Tgfbr2 double null prostates ( Figure 8). These data suggest that although the tumors in Pten;Tgfbr2 and Apc;Tgfbr2 double nulls are invasive and can metastasize, this is not accompanied by a large scale EMT phenotype. Given the role of TGFb signaling in driving EMT, it might be expected that the double null tumors would not have down-regulated E-cadherin and up-regulated vimentin, as they have lost a key component of the TGFb signal transduction pathway. These results do, however, leave open the question of how these tumors become metastatic. One possibility is that rare epithelial cells within the tumor have undergone EMT, presumably driven by a signal other than TGFb. Another intriguing possibility is that the double null epithelial cells undergo some form of collective invasion, in which small groups of epithelial cells become invasive and motile, while maintaining their cell junctions. This type of invasion has been attracting more interest as a potential driver of metastasis [47,48], and it will now be of interest to examine how the tumors examined here become invasive and metastatic.
Apc deletion does not activate TGFb signaling in prostate Deletion of Pten in mouse prostate initiates tumorigenesis and induces the activity of the TGFb pathway [33,34]. Similarly, expression of a constitutively active AKT1 transgene in prostate epithelium increases TGFb signaling [33], suggesting that Akt activation, which occurs downstream of Pten loss, is sufficient to activate this pathway. Given that deletion of the Tgfbr2 gene allowed for progression from HGPIN to invasive cancer in the Apc null prostate, we examined whether the TGFb pathway was induced in this model. Deletion of either Apc or Pten resulted in little change in overall b-catenin levels, whereas phospho-Akt levels were increased specifically in Pten null prostates ( Figure 9). To test whether the TGFb pathway was affected by Apc deletion, we first analyzed levels of the TGFb type II receptor and the intracellular mediator, Smad4. While both were significantly increased in the Pten null, there was no significant increase in either Smad4 or Tgfbr2 levels in Apc mutant prostates compared to those from wild type mice ( Figure 9). As an indication of pathway activation, we next analyzed levels of Smad2 phosphorylated at the carboxyl-terminal serines that are a substrate for type I TGFb receptors. Phospho-Smad2 was significantly increased in the Pten null but not in the Apc null prostates, indicating that pathway activation occurs with loss of Pten, but not with loss of Apc. The induction of TGFb signaling by Pten deletion could be driven by signal transduction events downstream of Akt activation, or could The phenotypes of Apc r/r and Pten r/r mice are shown, grouped by two age ranges: 8 to 24 weeks and 36 to 52 weeks. All animals analyzed had some prostate tumor phenotype, which is classified as Focal PIN, HGPIN, or HGPIN with adenosquamous differentiation (Asq-HGPIN). Animals with micro-invasive cancer and Asq-HGPIN or a mixture of HGPIN and PDA are grouped separately. Numbers of animals analyzed are shown above each column, and the distribution of phenotypes is shown as a percentage. The proportion of animals with any signs of invasive cancer at 36-52 weeks is significantly different between the two genotypes (p,0.02). (B) The numbers of mice with each phenotype are shown for animals with different combinations of mutations in both Apc and Tgfbr2 (above), or in both Pten and Tgfbr2 (below). For each genotype the numbers of animals with each phenotype are shown. Normal -no tumor phenotype evident. HGPIN and adenosqamous HGPIN -extensive HGPIN without evidence of invasion. PDA and adenosqamous carcinoma -extensive locally invasive cancer. All mice were euthanized at more than one year of age, unless they had to be sacrificed for tumor burden at a younger age. Significantly more animals with invasive cancer were observed in the Pten r/r ;Tgfbr2 +/r and Pten +/r ;Tgfbr2 r/r groups than in the Apc r/r ;Tgfbr2 +/r and Apc +/r ;Tgfbr2 r/r groups (14/14 vs 2/4; p,0.05, and 15/17 vs 0/6; p,0.001). One of the Apc r/r ;Tgfbr2 +/r mice (scored as PDA) had only small invasive foci. All others had extensive local invasion if scored as PDA or Asq-carcinoma. doi:10.1371/journal.pone.0092800.g004 be a consequence of the type of differentiation in this model -poor glandular differentiation in the Pten null rather than the squamous differentiation seen with Apc deletion. However, the clear cooperative effects of Apc and Tgfbr2 deletion suggest that even the low level of basal TGFb signaling present in the Apc mutant tumors is important for restraining cancer progression to locally invasive and metastatic disease.

Loss of TGFb signaling overcomes senescence
To examine the effects of the Tgfbr2 mutation on proliferation in Pten and Apc null tumors we examined expression of Cyclin D, increased levels of which correlate with advanced human prostate cancer [34]. As shown in Figure 10A, the number of epithelial cells with high levels of nuclear Cyclin D increased significantly in HGPIN in both the Pten and Apc single mutant prostates, with further significant increases in invasive cancer in each double mutant compared to the corresponding single mutant. We next examined expression of the CDK inhibitor, p27 (encoded by Cdkn1b) by immunofluorescence microscopy. For this analysis we co-stained for b-catenin to identify cells in which the Apc mutant phenotype was strongest. Expression of p27 increased in the Pten null and this increase was less pronounced in areas of invasive cancer in the Pten r/r ;Tgfbr2 r/r animals ( Figure 10B). This observation suggests that the transition to invasive cancer is concomitant with decreased p27 expression, consistent with our previous analysis [33]. In the Apc mutant, we observed a dramatic change in the localization of b-catenin from the cell periphery to a more diffuse expression pattern throughout the cell. This change in bcatenin expression was accompanied by a redistribution of p27 from the nucleus, as seen in the wild type prostate, to a diffuse signal throughout the cell ( Figure 10B). The change in p27 localization appeared to correlate with altered b-catenin expression, as small foci of cells in which b-catenin was still present at lower levels and only at the cell membrane retained nuclear p27 ( Figure 10B arrow). Interestingly, in the Apc r/r ;Tgfbr2 r/r mutants p27 levels decreased relative to those seen in the Apc single mutant, suggesting that loss of Tgfbr2 has similar effects on p27 in both models ( Figure 10B). These data are consistent with loss of TGFb signaling resulting in increased proliferation in both models of prostate cancer, although the effects of Pten and Apc deletions appear to be different with respect to p27 expression, and it is possible that the p27 re-localization in the Apc null prostate represents a first step in its inactivation. Phosphorylation of p27 at threonine 157 by AKT1 down-regulates human p27, but this phosphorylation site is not conserved in the mouse protein [49]. Additionally, phosphorylation of p27 on other sites that are conserved between mouse and human, results in redistribution to the cytoplasm and in some cases subsequent proteasomal   degradation [50]. However, we do not know if the effect of Apc deletion on p27 localization is due to phosphorylation of p27 itself.
A senescent phenotype can be induced in tumors either by activation of an oncogene, or by inactivation of a tumor suppressor gene [51]. Constitutive Akt activation in prostate epithelium induces cellular features of senescence, including SA-b-Gal (senescence-associated b-galactosidase) activity and increased p27 expression [52]. To test whether deletion of Pten and Apc induced senescence we analyzed the ventral prostates from different genotype mice for SA-b-Gal. Regions of HGPIN in both the Apc and Pten null prostates were positive for SA-b-Gal, whereas no staining was seen in the wild type prostate ( Figure 10C). Areas of invasive cancer in the two double mutants were devoid of SA-b-Gal staining, although isolated regions of HGPIN in the double mutants did still retain some SA-b-Gal signal, suggesting that the transition from HGPIN to an invasive phenotype is associated with overcoming senescence ( Figure 10C). These data suggest that deletion of either Apc or Pten in prostate epithelium initiates tumorigenesis, and also induces a senescent phenotype that can be overcome by deletion of the Tgfbr2 gene, which allows rapid progression to invasive cancer.
Although Apc deletion in prostate has been shown to generate HGPIN with only rare progression to locally invasive cancer [37], to our knowledge this is the first study to examine the combination of Apc deletion with another mutation in prostate epithelium. Expression of stabilized b-catenin in prostate epithelium was able to cooperate with Pten deletion, or with expression of an activated Ras transgene to accelerate the onset of locally invasive cancer [35,40]. However, either Pten deletion or Ras activation alone are sufficient to cause invasive CaP, albeit with slower kinetics than when combined with stabilized b-catenin. While stabilized bcatenin was able to accelerate the Pten null phenotype, deletion of b-catenin did not slow the progression of Pten null tumors to HGPIN, suggesting that it is not required for the early stages of tumorigenesis in this model [40]. We found that inactivation of TGFb signaling results in a dramatic acceleration of tumor progression initiated by Apc deletion. Despite the lack of an abnormal phenotype seen with loss of TGFb signaling in prostate Prostates of the indicated genotypes, selected to be representative of the most common phenotype of each, were analyzed by indirect immunofluorescence for E-cadherin (red) and Vimentin (green). Ages are as follows, wild type: 21 weeks, Pten r/r : 45 weeks, Apc r/r : 52 weeks, Tgfbr2 r/r : 70 weeks, Pten r/r ;Tgfbr2 r/r : 12 weeks, and Apc r/r ;Tgfbr2 r/r : 23 weeks. doi:10.1371/journal.pone.0092800.g008 epithelium, the combination of inactivating TGFb pathway mutations with either Pten or Apc deletion results in highly aggressive mouse models of CaP. Thus, TGFb signaling might restrain a relatively early step in the progression of these tumors, although it does not appear to affect tumor initiation.
It is possible that loss of TGFb signaling contributes to the transition from androgen-sensitive to CRPC. Androgen deprivation therapy is one of the major treatments for biologically significant human CaP, but tumors virtually always return and in general are more aggressive after becoming castration resistant. Previous work suggests that early castration of prostate-specific Apc null mice slows the progression of tumors in this model [37]. By 32 weeks of age regions of hyperplasia and metaplasia were still evident following castration at six weeks, whereas intact animals at this age display adenosqamous HGPIN. We previously showed that the Pten;Tgfbr2 null tumors were resistant to castration, suggesting that Tgfbr2 deletion speeds the progression to CRPC. However, in this model it is difficult to separate out the effects of Pten deletion from Tgfbr2 deletion, since HGPIN in Pten null mouse prostate has a limited response to castration. Therefore, the Apc model might present an opportunity to test whether inactivation of TGFb signaling contributes to progression to CRPC.
In summary, we show that loss of TGFb signaling in mouse prostate epithelium cooperates with loss of either the Apc or Pten tumor suppressor genes to drive invasive CaP, despite clear differences in the pathways activated and the tumor phenotypes. Loss of TGFb overcomes a restraint on tumor progression resulting in rapid onset of invasive and metastatic disease, further

Ethics statement
All animal procedures were approved by the Animal Care and Use Committee of the University of Virginia, which is fully accredited by the AAALAC.

Mice and DNA analysis
The loxP flanked Pten, Apc and Tgfbr2 alleles and the Pb-Cre4 allele have been described previously [38,[53][54][55]. Tgfbr2 and Apc mice, and the Pb-Cre4 transgenics were obtained from the NCI mouse repository. Conditional loxP flanked alleles each contain loxP flanked exons, which when recombined result in null alleles, and are referred to here as 'r' for recombined (null), or 'f' for the conditional loxP flanked (equivalent to wild type). All mouse lines were maintained on a mixed C57BL/6J x FVB strain background, as previously described [33]. Genomic DNA for PCR genotype analysis was purified from ear punch, at post-natal day 21 (P21), by HotShot [56], and genotypes were determined by PCR.

Histology, IHC and IF
Prostates were fixed in zinc-formalin, paraffin-embedded and stained with Hematoxylin and Eosin (H&E) or with Masson's Trichrome by standard techniques. Immunohistochemistry (IHC) and immunofluorescence (IF) analyses were performed as previously described [57][58][59][60]. Whole-mount prostate images were taken with a Leica MZ16 stereomicroscope and QImaging 5.0 RTV digital camera. IF images were captured on an Olympus BX51 microscope and DP70 digital camera, or on a Nikon Eclipse NI-U and captured with a DS-QI1 camera with NIS Elements software. Senescence-associated b-galactosidase (SA-b-G) staining Ventral prostates were fixed for 20 minutes in PBS + 4% paraformaldehyde, then washed 3 times in detergent wash (100 mM phosphate buffer pH 7.4, 2 mM MgCl 2 , 0.01% sodium deoxycholate, 0.02% NP-40). Whole lobes were stained overnight at 37uC in detergent wash with 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 1 mg/ml X-gal and 20 mM Tris pH 7.5, then post-fixed in 4% paraformaldehyde overnight at 4uC. They were then dehydrated through ethanol (70%, 80%, 90%, 95%, 100% 6 2), washed twice in xylene for 30 minutes, once in 50% xylene/50% wax for 1 hour, once in liquid wax for 1 hour, and then incubated overnight in liquid wax. Blocks were sectioned and counterstained with eosin. Images were captured on a Nikon Eclipse NI-U and a DS-RI1 camera.
Significantly increased protein expression (Figure 9) was tested for using a one-tailed Student's T test. Differences in Cyclin D expression ( Figure 10A) were tested using a two-tailed Student's T test.