Locally invasive, castrate-resistant prostate cancer in a Pten/Trp53 double knockout mouse model of prostate cancer monitored with non-invasive bioluminescent imaging

Here we have improved an existing mouse model of prostate cancer based on prostate-specific deletion of Pten and Trp53 by incorporating a Cre-activatable luciferase reporter. By coupling the deletion of those genes to the activation of a luciferase reporter, we were able to monitor tumor burden non-invasively over time. We show that, consistent with previous reports, deletion of both Pten and Trp53 on a C57BL/6 background accelerates tumor growth and results in both the loss of androgen receptor expression and castrate resistant tumors as compared with loss of Pten alone. Loss of Trp53 results in the development of sarcomatoid histology and the expression of markers of epithelial-to-mesenchymal transition Zeb1 and vimentin, with kinetics and penetrance dependent on whether one or both alleles of Trp53 were deleted. Homozygous deletion of Trp53 and Pten resulted in uniformly lethal disease by 25 weeks. While we were able to detect locally invasive disease in the peritoneal cavity in aggressive tumors from the double knockout mice, we were unable to detect lymphatic or hematogenous metastatic disease in lymph nodes or at distant sites.


Introduction
Prostate cancer is the most commonly diagnosed malignancy in men worldwide [1]. While the 5-year survival rate is high overall and has improved with prostate specific antigen (PSA) screening guidelines and new treatment modalities, it is still the second leading cause of cancer deaths for men in the United States [2]. Although most prostate cancers are found while the disease is still localized, for those with metastatic dissemination the 5-year survival drops to approximately 30% [2]. Two of the most commonly altered pathways in metastatic prostate cancer are PI3K-AKT and p53. The PI3K-AKT axis is hyperactive in prostate cancer through the deletion of its negative regulator and tumor suppressor gene, phosphatase and tensin homolog on chromosome 10 (PTEN). PTEN mutations occur at a frequency of approximately 17% in primary tumors and is increased to near 40% in metastatic tumors [3,4]. Mutations in and amplification of the PI3K catalytic subunits are also observed. TP53 mutations occur in approximately 7% of primary tumors and 50% of metastases [3,4]. Other mutations that occur at an elevated frequency in metastatic disease are in the retinoblastoma, MYC, and androgen receptor pathways [3,4].
The first genetically engineered mouse model for prostate cancer was the TRAMP model [5]. TRAMP mice are engineered for prostate-specific expression of both the large and small T-antigen, which results in the inhibition of Trp53 and Rb1 as well as the activation of AKT. While this model exhibits metastasis, its pathology more closely reflects rare high-grade, neuroendocrine-like disease [6,7]. In an attempt to build better models of prostate cancer, investigators have utilized prostate-specific, conditional gene knockout strategies [8].
Homozygous deletion of Pten using Cre recombinase controlled by a prostate-specific probasin promoter initially demonstrated lethal metastatic prostate cancer in mice [9]. However, subsequent efforts found indolent disease in prostate-specific Pten knockout mice extensively backcrossed onto the C57BL/6 background [10,11]. Chen, et al. showed that combining deletion of Trp53 with Pten on a C57BL/6 background unleashed a more aggressive, lethal phenotype with mice succumbing to bulky, locally invasive tumors [11]. Martin, et al. expanded on those findings and demonstrated that the prostate specific dual deletion of Trp53 and Pten resulted in rapid progression from prostate intraepithelial neoplasia to adenocarcinoma and ultimately sarcomatoid pathology, coupled with a progressive decrease in androgen receptor staining [12]. More recently, the Goodrich group has combined deletion of Rb1 with Pten and/ or p53 resulting in aggressive, metastatic disease including bone metastasis [13]. The addition of a Trp53 deletion conferred castrate-resistance to the Rb1; Pten knockout tumors [13]. However, with the deletion of Rb1 in this model, the resulting disease more closely resembled highgrade neuroendocrine prostate cancer.
We have previously described a prostate-specific Pten knockout mouse with a genetically engineered luciferase reporter that allowed quantitative, noninvasive monitoring of disease progression [10]. However, as mentioned above, this model did not develop metastatic disease, and castration resulted tumor regression with no relapse. Here we have combined loss of Trp53 with Pten with a luciferase reporter on an extensively backcrossed albino C57BL/6 background to facilitate bioluminescence imaging (BLI). We show that these mice rapidly develop fatal, bulky, locally invasive tumors. We also note the consistent development of sarcomatoid pathology and epithelial-to-mesenchymal transition (EMT)-like features in these tumors. Despite the presence of these features, we did not detect disseminated metastatic disease via BLI or histologic analysis, supporting previous findings with this model.

Mouse strains and genotyping
All animal procedures were performed under approval from the Institutional Animal Care and Use Committee at the University of Iowa, AP #1302028. Animals were housed in standard conditions following guidelines outlined in PHS Animal Welfare Assurance (D16-00009, A3021-01) and United States Department of Agriculture (USDA No. 42-R-0004). Mice were housed in Thoren individually ventilated cages that contained SoftZorb Enrichment Blend bedding (Northeastern Products Crop, Warrensburg, NY), had 24 hr access to 0.2 um filtered water via Edstrom automatic watering system, and fed Teklad 7913 irradiated diet (Harlan). We crossed homozygous albino C57BL/6J-Tyr c-2J /J ROSA26 LSL-Luc Pten fl/fl generated from our previous studies [10] mice with heterozygous B6.129P2-Trp53 tm1Brn /J (Jackson Laboratory), as well as with B6.Cg-Tg(Pbsn-cre)4Prb (Pb-Cre4 + ) mice obtained from the NIH Mouse Models of Human Cancer Consortium. The mice were selected for Pten fl/fl and either Trp53 fl/+ orTrp53 fl/fl . Once the desired alleles were obtained, mice were backcrossed to C57BL/6J-Tyr c-2J (Jackson Laboratory) mice for six generations, selecting for albino-coat color offspring. Mice were genotyped for Cre, floxed alleles of Pten, Trp53, and luciferase as described previously [14][15][16][17]. Three groups of males were generated yielding 15 wild type Trp53 mice with homozygous Pten deletion (WT), 17 mice heterozygous for Trp53 with homozygous Pten deletion (HET), and 16 double knockout mice (DKO) for a total of 48 mice. Additionally, 5 more DKO mice had undergone surgical castration to study the response to androgen deprivation. Mice were examined and imaged starting at five weeks of age. Mouse health was monitored 3 times weekly and mice were euthanized by gas CO 2 followed by cervical dislocation if body scores were 2 or less [18], or if they displayed signs of declining health such as decreased mobility. Three mice died before meeting experimental endpoints.

Bioluminescence imaging
All bioluminescence imaging (BLI) was performed using an Ami HTX imager from Spectral Instruments Imaging (Tuscon, AZ). Animals were anesthetized in a chamber with 3.5% isoflurane, and then 150 mg/kg D-luciferin (VivoGlo, Promega, Madison, WI) was injected intraperitoneally. Mice were then placed into the imager while maintaining anesthesia and an image was obtained five minutes after D-luciferin injection. Exposure time was five minutes per image. AMIView Imaging Software, version 1.5.0 was used to analyze the region of interest and measure photon flux (photons/sec/cm 2 /sr). Images were obtained on ex vivo necropsy tissues shortly after necropsy with an additional five-minute exposure time.

Mouse examinations
All mice were examined and weighed bi-weekly. Examination included palpation of the abdomen and pelvis and evaluation of overall health and hygiene. Mice were imaged bi-weekly. Mice were euthanized according to pre-determined endpoints, including 20% loss of starting body weight, decreased mobility, lack of grooming, or other morbidity. Homozygous Pten deleted, wild type Trp53 mice were euthanized alongside double knockout mice regardless of health status as an age-matched control. Additional cohort of Pten knockout mice were aged and euthanized alongside Trp53 heterozygous mice as age matched controls for those mice. Necropsy included removal of the primary prostate tumors, imaging of the remainder of the animal, and subsequent removal of any bioluminescent tissues. Tissues were fixed in 10% neutral buffered formalin at 4˚C for 24-48 hours and then dehydrated and embedded in paraffin.

Histopathology
Formalin-fixed, paraffin-embedded tissues were sectioned and stained using standard hematoxylin and eosin stains. Immunohistochemistry was performed on serial sections with the following antibodies: cytokeratin pan antibody cocktail (ThermoFisher, Waltham, MA; catalog #MA5-13203), androgen receptor (AR) (AbCam, Cambridge, UK; catalog #ab133273), vimentin (Cell Signaling Technology, Danver, MA; catalog #5741S), Zeb1 (obtained as a generous gift from the Richer lab at the University of Colorado) [19], and smooth muscle actin (SMA) (Bio-Rad, Hercules; CA catalog #MCA5781GA). Immunohistochemistry protocol included a heat-based antigen retrieval in 0.1 mM citrate buffer (pH 6.0), quench with 10% hydrogen peroxide, and blocking with 5% goat serum for all antibodies except for AR, which was blocked with 1.5% horse serum. The primary antibodies were then introduced at dilutions of 1:100 for cytokeratin, 1:250 for AR, 1:100 for vimentin, 1.5:1000 for Zeb1, and 1:500 for SMA. Slides were finally incubated in either horseradish peroxidase-conjugated (HRP) goat anti-rabbit (Jackson ImmunoResearch Laboratories, West Grove, PA; catalog #AB_2307391) or EnVision + HRP anti-mouse (Dako, Carpinteria, CA; catalog #K400111-2) secondary antibodies and developed using DAKO DAB. The slides were interpreted by a single pathologist with expertise in rodent models of prostate cancer who was blinded to genotype. Analysis for EMT status was determined by overlapping regions of Zeb1 and Vimentin positivity.

RNA extraction and RT-qPCR
RNA was isolated from 5μm thick whole-tissue FFPE sections as previously described [20]. Tissue sections were deparaffinized by washing twice with 1 mL of xylene, vortexed, and centrifuged (16,000 g x 5 min). Pellets were washed twice with 1 mL of 95% ethanol prior to RNA extraction. RNA extraction was performed using the RNeasy FFPE Kit (Qiagen, Valencia, CA; catalog #73504) according to the manufacturer's instructions. RNA integrity was determined using the 2100 Bioanalyzer (Agilent, Santa Clara, CA) and for RNA used in downstream applications the RIN was assessed to be � 1.8. Reverse transcription was performed on 750 ng of RNA with iScript cDNA Synthesis Kit (BioRad, Hercules, CA; catalog #). Quantitative PCR was performed using the PowerUp SYBR Green Master Mix (Applied Biosystems, Waltham, MA; catalog #A25780), according to manufacturer's guidance. Primers were designed using NCBI following the guidance on downstream applications from Qiagen. Expression of Foxa2 (GGAAATGAGAGGCTGAGTGGA; AAGTGTGGTGGCTACTTTTCCT), Syp (CGTTAAAGG GGGCACTACCA; TTCAGCCGAGGAGGAGTAGT), Msmb (CAATTCCACCCAGAGAACTGC; CCTGGGTTCTTCCGATCCAC), and Apof (CTCTCTGGCTGAAGTCGAACA; GGGTTGAGAA GTAGAACTGTGC) were compared to the rRNA control 18S (GAGGCCCTGTAATTGGAATG AG; GCAGCAACTTTAATATACGCTATTGG).

Statistical analysis
Statistical analyses were performed with GraphPad Prism software 8.0. and R 3.6.3. Growth rates were analyzed by response feature analysis through fitting the growth data for each mouse individually to an exponential growth model, obtaining estimates of the rate parameters via nonlinear least squares. The growth rates were then compared using a one-way ANOVA with Tukey's honest significant differences used to correct for multiple comparisons. Survival data was analyzed using the Kaplan-Meier method. Pathologic data was qualitatively characterized and analyzed using Chi-squared tests with a Bonferroni correction for multiple hypothesis testing. RNA expression data was compared using a T-test to compare the delta Ct values. Data for gene expression is presented as a fold-change (2 exponent of the delta delta Ct) with SEM error bars.

Prostate specific deletion of Trp53 and Pten leads to fast-growing, lethal, sarcomatoid tumors
To non-invasively monitor tumor burden, bioluminescence imaging (BLI) was performed on mice with Probasin-Cre mediated deletion of Pten and Trp53, coupled with the activation of a ROSA-LSL luciferase reporter. In this study, all mice had homozygous floxed Pten alleles, with cohorts having wildtype Trp53 (WT), heterozygous floxed/wildtype Trp53 (HET), or homozygous floxed Trp53 alleles (DKO). The longitudinal BLI imaging (Fig 1A and 1B) demonstrated that tumors grew faster in the DKO mice than those in the HET and WT group (S1A Fig).
After an initial rapid growth phase in all groups from 5-10 weeks, which reflects pubertal activation of Cre and growth of the prostate, WT mice showed little growth out to 50 weeks consistent with our previous studies [10]. For reference, mice in this strain lacking Cre expression do not show any BLI signal above background [17]. Notably, the BLI growth curves of the Pten-only deleted animals are remarkably similar as reported here and in our previous study despite the fact that they were measured in separate experiments conducted many years apart and on different imaging systems, here on an AMI HTX and previously on an IVIS 100 [10]. The DKO mice showed rapid tumor growth around 19 weeks of age that resulted in formation of readily palpable primary tumors and a median survival of 26.5 weeks (Fig 1B and 1C). These tumors were larger in comparison to the prostates of WT mice taken down at the same time-point (S1B and S1C Fig). In comparison to the DKO group, the HET group demonstrated slower disease progression with tumor growth not increasing until mice were around 29 weeks of age (Fig 1B). Moreover, the median survival for heterozygous Trp53 null mice was closer to that of WT (41 and 50 weeks, respectively) (Fig 1C).
We performed histopathological analysis of prostates collected from DKO, HET, and WT mice euthanized at 25 weeks of age, as well as HET and WT mice at~40 weeks (Fig 2A). At 25 weeks, 13/13 mice from the DKO group demonstrated sarcomatoid tumors as confirmed by negative staining with a pan-cytokeratin antibody, with most of the HET and WT showing high-grade mouse prostate intraepithelial neoplasia (HG mPIN) features. For the WT group, there was no difference in this distribution at 40 weeks; however, for the HET group, 6/6 tumors evaluated were sarcomatoid (Fig 2B). To rule out potential stromal expansion as the cause for the lack of cytokeratin staining in the mice presenting with sarcomatoid pathology, we stained for smooth muscle actin (SMA). In WT mice at 40 weeks of age, anti-SMA antibody stains the fibromuscular stroma surrounding glands with mPIN lesions. In contrast, in KO mice at 25 weeks of age or HET mice at 40 weeks of age, although there are some small, isolated patches of SMA staining, the bulk of these large tumors are SMA negative (Fig 2A).
Based on the histological characterization, we wished to determine if loss of Trp53 lead to elevated EMT-like features in the tumors. For this we stained tissue for the EMT markers Zeb1 and vimentin (Fig 3A). For tumors to be scored positive for EMT, we applied the strict criteria that regions of overlapping positivity on serial sections for both markers are present. From this analysis, we observed an increase in EMT-positive tumors in the DKO group at 25 weeks of age as compared to the WT group (Fig 3B). As evident in Fig 3A, there were examples of some tumors in both WT and HET mice that showed some positive Zeb1 staining, but not vimentin staining which we scored negative for EMT. We analyzed a smaller group of HET and WT tumors at 40 weeks of age and although not statistically significant, we observed EMT-positive tumors in HET mice, but not WT mice (S2 Fig). We also asked if the loss of Trp53 with Pten lead to an increase in neuroendocrine-like (NE) features. To achieve this we compared the expression of the prostate NE genes Foxa2 and Syp mRNA by RT-qPCR between 5 DKO mice and 5 WT mice that were euthanized at 25 weeks of age (Fig 3C) [21,22]. From this analysis we observed a non-statistically significant trend towards a moderate increase in Foxa2 and Syp expression, suggesting that there may be isolated regions of NE-like differentiation present in the DKO tumors, consistent with the heterogeneous expression of Syp observed previously in Pten;Trp53 knockout mice [12].

Pten/Trp53 double knockout mice develop locally invasive disease but not lymphatic or distant metastasis
Due to the aggressive nature of the tumors from the DKO group and the prevalence of PTEN and TP53 mutations in metastatic disease, we determined if there was metastatic disease in these mice. We evaluated residual tumor burden by BLI post-necropsy, after removal of the primary prostate tumor for 10 out of 16 mice (Fig 4A, 4D and 4F). This was necessary because the intense signal from the primary tumor could obscure weaker signals emanating from metastatic sites. We observed intraperitoneal spread in 8 out of 10 mice. We then histologically evaluated BLI signal-positive tissue to determine if there was metastatic disease (Fig 4B, 4C,

PLOS ONE
Pten/Trp53 double knockout mouse model of prostate cancer 4E and 4G). Upon histological evaluation of a mouse that appeared to have lumbar lymph node involvement (Fig 4A), we found that instead of intralymphatic growth, the tumor invaded into the local perilymphatic fat (Fig 4B and 4C). DKO mice (8/10) were found to have visceral BLI signal upon post-necropsy imaging (Fig 4D), but this seems more likely to be peritoneal seeding and local invasion than hematogenously-disseminated disease (Fig 4E). There was also evidence of aggressive, invasive disease with BLI-positive tumor invading into pelvic sidewall musculature (Fig 4F and 4G).

Pten/Trp53 double knockout results in loss of androgen receptor expression and castrate resistant tumors
While the DKO tumors were not metastatic, we wished to investigate the androgen sensitivity of this model. To assess this, we generated a cohort of five DKO mice which were surgically castrated at 10 weeks of age. After castration, the BLI signal noticeably decreased between 10 and 15 weeks of age (Fig 5A). Eight weeks after castration, the tumor burden started to increase in all mice, demonstrating castrate-resistant prostate cancer. Histological analysis from these tumors revealed that, like the intact DKO mice, they were all sarcomatoid (S3A  and S3B Fig). Due to the rapid and uniform onset of castrate-resistant growth, we determined if the tumors from intact mice demonstrated alterations in AR expression (Fig 5B and 5C). From this analysis, we found that~75% of DKO mice lacked AR protein expression at 25 weeks, but all of the HET and WT mice retained AR expression at this timepoint (Fig 5C). At 40 weeks,~50% HET mice and less than 25% of WT mice showed loss of AR expression. Consistent with the observation that tumors from the castrate DKO mice exhibited androgeninsensitive growth also lacked AR expression (S3C Fig). However, we did not observe a significant difference in the expression of the AR target genes Msmb and Apof when comparing DKO and WT mice at 25 weeks (Fig 5D) [23,24].

Discussion
Here we describe a Pb-Cre mediated Pten;Trp53 double knockout mouse model of prostate cancer with a bioluminescent reporter for non-invasive monitoring of tumor growth. Our model showed accelerated tumor growth and mortality of the DKO mice as compared to HET and WT mice. While at approximately 25 weeks of age the HET mice were more like the WT

PLOS ONE
Pten/Trp53 double knockout mouse model of prostate cancer knockout mice, analysis of tissue collected at necropsy demonstrated that by 40 weeks of age the HET mice had progressed and became similar to the DKO mice. Necropsy tissue from both the DKO and HET mice showed that tumors had sarcomatoid pathology. This is a rare, undifferentiated and mesenchymal prostate pathology (<1% of prostate cancer tumors) that is considered to be highly aggressive [25][26][27]. The loss of p53 function may be associated with the development of sarcomatoid tumors in patients. Case studies have reported strong nuclear accumulation of p53 in the sarcomatoid regions of the tumor, consistent with oncogenic dominant-negative or gain-of-function mutations in p53 [28], whereas adenocarcinoma typically lacks strong p53 staining [29,30]. In addition to the accumulation of p53, it has been demonstrated that the sarcomatoid portion of a prostate tumor lacks AR staining [31].
Sarcomatoid prostate cancer has been previously observed in Trp53;Pten prostate KO mice. In a report from Martin et. al. the authors noted that the progression of prostate tumors from Pb-Cre mediated knockout of Trp53;Pten resulted in a complete penetrance of spindle cell carcinoma or sarcomatoid pathology for mice between 26-30 weeks [12]. They also noted that the regions containing spindle cell carcinoma had low or undetectable AR expression [12]. Our results are consistent with these findings, as pathological analysis of necropsy tissue from our DKO mice displayed sarcomatoid features with most tumors lacking AR staining. However, our analysis of AR target genes Msmb and Apof did not demonstrate a significant difference in expression between the WT and DKO mice [23,24]. This could be due to not utilizing laser capture microdissection to isolate tissue for mRNA analysis and thus the AR signal is diluted in the WT group due to stromal contamination. Additionally, due to reciprocal feedback between PI3K and AR signaling it is possible that the AR transcriptional output in Pten knockout animals is reduced below the level necessary to detect meaningful differences in expression across different Trp53 genotypes, in spite of the differences in AR protein expression observed [32,33].
While we were unable to demonstrate a difference in AR target gene expression between WT and DKO mice at 25 weeks, we were able to demonstrate a correlation of AR protein loss with the sarcomatoid histopathology. Utilizing BLI, we were able to determine that the DKO mice were initially responsive to castration but developed castrate-resistant disease 5 weeks later, which has not been previously reported in primary tumors. The age for the onset of tumor growth in castrated DKO mice (18 weeks) is consistent with the age in which the Martin study observed sarcomatoid pathology [12]. This association of loss of AR staining and sarcomatoid pathology was also observed in the tissue from the HET mice at 40 weeks of age. These tumors were similar to the tumors from DKO mice, as all tumors presented as sarcomatoid and 50% did not stain for AR. Given the reduced AR protein expression in sarcomatoid tumors and that the castrate DKO mice all had sarcomatoid pathology, it is likely that the rapid onset of castrate-resistant tumor growth was due to the progression of the prostate cancer to a sarcomatoid pathology. While the development of sarcomatoid pathology under castrate levels of androgen is interesting, it is still unknown what events lead to the formation of this disease state.
Another interesting aspect of this model is the association between the sarcomatoid pathology and EMT. In both the Martin study and case reports from patients, the sarcomatoid regions stained strongly for vimentin expression [12,29]. Vimentin, along with Zeb1, are widely regarded as markers of EMT and have been implicated in the development of metastatic disease in many cancers, including prostate [34]. Our results show increased expression of both proteins in DKO tumors, and indeed we have shown that the expression of these proteins occurs within cells on adjacent serial sections if not the same cell. Interestingly EMT-like characteristics may result from the reduction of AR signaling in the sarcomatoid tumors. It has been demonstrated that androgen deprivation leads to elevated expression of EMT markers in both normal and cancerous prostate tissue [35]. Moreover, Zeb1 and AR may regulate each other through mutually antagonistic processes [35].
Despite the presence of EMT markers in the sarcomatoid tumors, neither in the earlier reports nor in our study were distant metastases detected [11,12]. Even with the addition of a bioluminescent reporter, we were able to detect only bioluminescent-positive tumor deposits that represented local invasion into peri-lymphatic fat as well as peritoneal seeding with tumor found on the pancreas. However, we cannot rule out the possibility that we did not detect micrometastatic disease in some organs which lacked sufficient BLI signal for detection because we did not exhaustively section metastatic end organs without BLI signal. In spite of the inability of BLI to detect metastatic disease in the model reported here, we believe there is utility in non-invasive, bioluminescent monitoring in the detection of distant, and potentially very small volume, metastatic disease. Moreover, the results from our study underscores the importance of validating BLI signal with histopathology.
For distant metastases to occur in the context of Trp53;Pten prostate-specific deletion, additional genes may need to be disrupted, such as Rb1 or Smad4 [13,36,37]. However, expression of EMT markers Zeb1 and vimentin suggests metastatic potential in this model, and certainly correlates with more aggressive disease including castrate resistance. Additionally, the trend for an increase in the expression of NE markers also suggests that this model would result in metastatic disease. There are at least two additional aspects of this model which may contribute to the lack of observed metastatic disease. First, because Trp53;Pten are deleted in the entire prostatic epithelium, the subsequent tumors in this model are rapidly fatal, likely from large tumor masses and urinary obstruction, which limits the time for progression of disease. To circumvent these problems, somatic deletion of Pten and Trp53 may provide a new direction. Indeed, a virally-introduced, Cre-mediated deletion of Pten and Trp53 in mouse prostates has reportedly produced reliable metastasis [37]. Second, the host strain of the model may influence metastatic potential [6]. It has been shown that normal prostate tissue from different strains have alterations in expression of genes such as MMP7 and prostate specific stem cell antigen [38]. Moreover, BALB/c mice have loss of p16 Ink4a , which could alter metastatic potential by dysregulation of the Rb pathway [39]. Evaluating the loss of Trp53;Pten function across a range of host strains may elucidate this possibility and lead to the identification of metastasis modifier genes.