AbstractPenile squamous cell carcinoma (PSCC) accounts for over 95% of penile malignancies and causes significant mortality and morbidity in developing countries. Molecular mechanisms and therapies of PSCC are understudied, owing to scarcity of laboratory models. Herein, we describe a genetically engineered mouse model of PSCC, by co-deletion of Smad4 and Apc in the androgen-responsive epithelium of the penis. Mouse PSCC fosters an immunosuppressive microenvironment with myeloid-derived suppressor cells (MDSCs) as a dominant population. Preclinical trials in the model demonstrate synergistic efficacy of immune checkpoint blockade with the MDSC-diminishing drugs cabozantinib or celecoxib. A critical clinical problem of PSCC is chemoresistance to cisplatin, which is induced by Pten deficiency on the backdrop of Smad4/Apc co-deletion. Drug screen studies informed by targeted proteomics identify a few potential therapeutic strategies for PSCC. Our studies have established what we believe to be essential resources for studying PSCC biology and developing therapeutic strategies. IntroductionPenile cancer can constitute up to 10% of male malignancies in some African, Asian and South American areas1. While it represents ~0.5% of all cancers among men in the United States and other developed countries, penile cancer often results in devastating disfigurement and only half of the patients survive beyond 5 years1. PSCC accounts for over 95% of penile malignancies1. Risk factors associated with PSCC incidence include the absence of neonatal circumcision, phimosis, chronic inflammation, poor hygiene, tobacco use, and infection with human papillomavirus (HPV)2. A recent large meta-analysis reports a pooled HPV prevalence of 50.8% in penile cancer3. HPV infection is more common in carcinomas in situ than in invasive forms4, and the presence of high鈥恟isk HPV confers a survival advantage in patients with PSCC5. It is possible that HPV-positive and HPV-negative PSCC develop through certain convergent mechanisms, which may be leveraged for therapies. Expression alterations for some genes have been documented in PSCC, such as RAS, MYC, EGFR, E-cadherin, COX2, IGF1R, and MMP2/96. However, the precise molecular mechanisms governing PSCC initiation and progression remains elusive, mainly due to the shortage of experimental tools. Recently, a number of PSCC cell lines have been developed7,8,9. However, the generation of genetically engineered mouse (GEM) models of PSCC has not been reported.Standard treatment regimens for penile cancer are limited to surgery, radiation, and chemotherapy1,10,11. The mainstay of systemic therapy for advanced PSCC is cisplatin-based chemotherapy, with average response rates between 15 to 55% and median overall survival (OS) ranging between 5 and 12 months1,10,11,12. Patients whose PSCC progresses or recurs after front-line cisplatin-based chemotherapy experience poor responses to salvage treatments (OS鈥?lt;鈥? months)13. Therefore, there is an urgent need to identify molecular mechanisms for chemoresistance of PSCC and seek other approaches for metastatic PSCC. Immunotherapy using immune checkpoint blockade (ICB), such as anti-CTLA4 and anti-PD1 antibodies, has revolutionized cancer treatment and has generated durable therapeutic responses in a significant subset of patients across a variety of cancer types. Currently clinical trials are recruiting patients to assess the role of ICB agents in penile cancer specifically or a rare tumor cohort that includes penile cancer. Examples include pembrolizumab for advanced PSCC (NCT02837042) and ipilimumab plus nivolumab for advanced rare tumors including penile cancer (NCT03333616, NCT02834013). Developing resources for PSCC such as GEM models will be critical to evaluate the efficacy of ICB in a meaningful preclinical setting and, perhaps more importantly, to predict potential resistance mechanisms for ICB and design corresponding combination therapy strategies.Here, we report the generation and characterization of two GEM models of PSCC through the manipulation of signaling pathways relevant to HPV carcinogenesis mechanisms and clinical evidence of PSCC. The murine PSCC displays strong immune gene signatures and infiltration, consistent with clinical evidence gathered by bioinformatics analysis and immunohistochemistry validation. Intratumoral immunosuppressive myeloid cell infiltration suggests the benefit of combining targeted therapy and immunotherapy to achieve maximal clinical efficacy. This illuminates a testable clinical trial hypothesis for combination therapy in the treatment of lethal PSCC.ResultsEstablishment of genetically engineered mouse model of PSCCAndrogen receptor (AR) is expressed in the epithelium of the developing human fetal penis14. We stained human primary penile tumor samples (n鈥?鈥?, Supplementary Table聽1) and observed pronounced nuclear AR expression in adjacent normal epithelium (5/5 cases) but not in the malignant area (8/8 cases, Fig.聽1a). There is also persistent AR expression in the rodent penis15, which we further confirmed in the penis of wild type adult male mice (Fig.聽1b). This result prompted us to hypothesize that a GEM model employing an AR-responsive Cre driver in conjunction with LoxP alleles of tumor suppressor genes with relevance to human penile cancer could form penile cancer in mice. The Cre-expressing line PB-Cre4 carries the Cre gene under the control of a strong AR-responsive promoter and is frequently used to manipulate gene expression in the mouse prostate16. To determine if PB-Cre4 drives recombination of mouse penile epithelium, we crossed PB-Cre4 to the fluorescence reporter allele mTmG17 and the penis clearly showed GFP expression in the epithelium (Fig.聽1c and Supplementary Fig.聽1a). AR+ nuclei overlapped with GFP+ epithelial cells (Supplementary Fig.聽1b). To examine genetic changes that may cause penile tumorigenesis, we analyzed several previously developed prostate tumor models that used PB-Cre4, including Pten18, Pten Smad419, and Pten p53 Smad420. Normal histology of the penile epithelium was observed for these mice even at the age when they succumbed to the prostate tumor burden (Supplementary Fig.聽1c), indicating loss of function for Pten, p53 or Smad4 (individually or in combination) is insufficient to drive penile tumorigenesis.Fig. 1: Smad4 and Apc co-deletion leads to penile squamous cell carcinoma in mice.a, b IHC stain for AR in human penile tumors (n鈥?鈥?) and normal mouse penis (n鈥?鈥?). Scale bar 100鈥壩糾. c, d Morphology and fluorescence for resected mouse penises of indicated genotype and age. Scale bar 2鈥塵m. e Penile prolapse free survival curves for mice of three genotypes with n indicated. ****P鈥?lt;鈥?.0001, log-rank test. f Morphology and fluorescence for resected mouse penis of SAm genotype at 6.4 months of age, with GFP+ tumor nodules clearly visible. Scale bar 2鈥塵m. g, h H E stain of penis and prostate lobes (anterior AP, dorsolateral DLP) of SA mouse at 6.8 month of age. White arrows denote keratin pearls. Scale bar 100鈥壩糾. i SA males at 8鈥?2 weeks underwent castration or mock surgical procedure. After 4 weeks, the mice were euthanized. The serum testosterone was measured with ELISA for the mice and age-matched wild type (WT) males were used as control (n鈥?鈥? for each group). j Representative images of the mouse penis at each week post-surgery. k, l Weight, representative image and H E staining of resected penises from WT control or SA mice 4 weeks post-surgery. Scale bar 10鈥塵m and 200鈥壩糾 (n鈥?鈥? for WT, n鈥?鈥? for mock and castration). In i, k, data represent mean鈥壜扁€塖D. #P鈥?gt;鈥?.05, ***P鈥?lt;鈥?.001, ****P鈥?lt;鈥?.0001, two-sided Student鈥檚 t test.Full size imageNext, we focused on two pathways, Wnt/Apc/尾-catenin and TGF尾/Smad pathways, based on their relevance to oncogenic HPV. Oncogenic HPV encodes two oncoproteins E6 and E7, which bind and facilitate degradation of p53 and Rb, respectively21. While E6 and E7 can immortalize cells in vitro, they are not sufficient for tumor development in vivo21, suggesting that additional mechanisms are necessary for tumor initiation. E6 binds to cellular partner E6AP to stabilize 尾-catenin and stimulate Wnt signaling22, and knockdown of E6 and E7 expression induces a substantial reduction of nuclear 尾-catenin and TCF transcriptional activity23. E7 blocks TGF-尾 induced transcription and growth inhibition by directly binding to Smad3 and Smad4 thus interfering with their interaction24. E6 has a similar activity25. To recapitulate these signaling effects from E6/E7, we reasoned that Wnt/尾-catenin activation and TGF尾/Smad pathway inactivation in the mouse penile epithelium might lead to penile tumorigenesis. We used conditional null alleles of Apc and Smad4 to test our hypothesis. PB-Cre4+ Smad4L/L mice had no prostate lesion20, nor did we observe them to show any abnormality of the penis (Supplementary Fig.聽1d). PB-Cre4+ ApcL/L mice developed prostate hyperplasia with squamous metaplasia as early as 7 months of age26, and the penis of these mice was normal at 3.5 months of age but displayed mild dysplasia with hyperkeratosis at 12.5 months of age (Fig.聽1d and Supplementary Fig.聽1d). The epidermal layer was moderately multiplied, yet no SCC pathology was found (Supplementary Fig.聽1d). Penile prolapse (paraphimosis) is a condition in which the penis no longer retracts back into the prepuce, and can be used as a sign to indicate penile tumor formation in mice. In total, 0/19 PB-Cre4+ Smad4L/L mice and 1/18 PB-Cre4+ ApcL/L mice exhibited penile prolapse (Fig.聽1e). The only prolapsed penis of a PB-Cre4+ ApcL/L mouse was caused by excessive keratinization. Remarkably, when Smad4 and Apc were co-deleted, PB-Cre4+ Smad4L/L ApcL/L mice (SA genotype) developed penile prolapse at 100% penetrance (Fig.聽1e). When crossed to mTmG allele (SAm genotype), GFP+ solid tumor nodules formed in mouse penis (Fig.聽1f). At the histological level, the tumors contain conspicuous keratin pearls, which are pathognomonic for SCC (Fig.聽1g). We noted that, by the time the penile tumors formed in these mice (median 17.2 weeks), the prostate remained normal (Fig.聽1h). Given the AR expression by normal penile epithelium (Fig.聽1b), we determined whether AR signaling was required to sustain SA penile tumor growth by performing mock procedure or surgical castration on 8鈥?2-week-old SA males and following the penile tumor progression for 4 weeks. Castration depleted serum testosterone level (Fig.聽1i), yet elicited undetectable effect on penile tumor growth (Fig.聽1j鈥搆), histology (Fig.聽1l), proliferation marker Ki67, or apoptosis marker cleaved caspase-3 (Supplementary Fig.聽1e). We conclude that AR signaling is dispensable for sustaining PSCC. In summary, the first GEM model of PSCC was developed by co-deleting Apc and Smad4 in the penile epithelium.Transcriptomic analysis of mouse penile cancerTo identify differentially expressed genes between normal and cancerous mouse penis, we performed RNA-seq on penile tumors from 5-month-old SA mice and penile epithelium from age-matched wild type (WT) mice. Significant genes were defined by using a cut-off of 0.01 on the Benjamini-Hochberg corrected P-value and an absolute log2(fold change) value of at least 2 (i.e. fold change鈥夆墺鈥?). We detected a total of 1660 significantly upregulated genes and 642 downregulated genes in SA compared with the WT penis (Fig.聽2a, Supplementary Fig.聽2a, and Supplementary Table聽2). To identify significantly enriched pathways by SA tumors, Ingenuity Pathway Analysis (IPA) was used to analyze the differentially expressed genes and identified 102 Ingenuity Canonical Pathways with P-value鈥?lt;鈥?.05 (Supplementary Table聽3). Of note, the top ranked two pathways are Granulocyte Adhesion and Diapedesis and Agranulocyte Adhesion and Diapedesis, because of upregulation of many cytokines and cytokine receptors suggestive of an inflammatory phenotype (Supplementary Table聽4). Matrix Metalloproteases and Eicosanoid Signaling were significantly enriched (Supplementary Tables聽3 and 4), underscoring the invasive SCC histology and supporting the inflammation of SA tumors. Both Wnt/尾-catenin Signaling and FGF Signaling were hyperactivated in SA tumors as a result of genetic loss of Apc and dramatically enhanced expression of FGF ligands, respectively (Supplementary Tables聽3 and 4). IPA can provide analyses to identify putative upstream regulators to account for gene expression changes, which listed 尾-catenin as the top putative upstream regulator (Supplementary Table聽5). Massive transcriptional regulation by 尾-catenin complex (Fig.聽2b) was expected because of Apc loss, with confirmed nuclear concentration in SA tumors compared with WT penis (Fig.聽2c). Enhanced 尾-catenin localization in the nucleus and cytoplasma was also validated in about one third of human PSCC samples using tissue array analysis (n鈥?鈥?91 total, Supplementary Figs.聽2b, c). Two of 尾-catenin transcriptional targets, Sox2 and Ptgs2, were upregulated by 137-fold and 9-fold respectively at the RNA level. Sox2 was shown to be a master regulator for cancer stem cells in skin SCC27. Ptgs2, better known as Cox2, converts arachidonic acid to prostaglandins, which in turn induce inflammatory reactions. Significant upregulation of Sox2 and Cox2 in SA tumors at the protein level was confirmed by immunohistochemistry (IHC) (Figs.聽2d, e). By a separate algorithm in IPA to estimate master regulators to explain transcriptomic changes and pathway enrichment, Cox2 was identified as the top regulator to account for cytokine upregulation and the phenotype Adhesion of immune cells (Fig.聽2f and Supplementary Table聽6). COX2 upregulation is highly relevant to PSCC as previously shown28. Western blot further confirmed upregulation of Cox2 and Sox2 in SA tumors relative to penile tissues from WT and PB-Cre4+ Smad4L/L mice (Fig.聽2g). At the protein level, we also detected upregulated cyclin D1 and phospho-Rb (Ser780) signals in SA tumors (Fig.聽2g). Intensified cyclin D1 expression is expected given its known regulation by Smad420 and 尾-catenin/LEF1 complex29. Cyclin D1 overexpression and correlated Ki67 expression was reported in PSCC30 and confirmed in our model (Fig.聽2h). Using a penile cancer cell line, SA1, established from an SA penile tumor, we generated shRNA-mediated Sox2 knockdown (Fig.聽2i), which attenuated subcutaneous tumor growth of SA1 (Fig.聽2j鈥搇). Together, transcriptomic profiling revealed extensive signaling changes associated with PSCC formation in mice and, in particular, signaling molecules involved in inflammation and immunity.Fig. 2: Transcriptomic analysis reveals activation of 尾-catenin signaling and inflammatory pathways in mouse PSCC.a Hierarchical clustering of differentially expressed genes between WT and SA mouse penile samples (n鈥?鈥? for each genotype). b Gene regulation network by 尾-catenin (i.e. CTNNB1), the top upstream regulator identified by IPA to account for differential gene expression changes in SA tumors compared with WT penis. Red and green colors indicate upregulation and downregulation, respectively. c鈥?b>e IHC stain for 尾-catenin, Sox2 and Cox2 in WT and SA penis, respectively. Scale bar 20鈥壩糾 for c, 100鈥壩糾 for d, 200鈥壩糾 for e. f Mechanistic network for the top ranked master regulator Ptgs2/Cox2 by IPA to illustrate its effect on cytokine expression regulation and function in immune cells. g Western blot showing differential protein expression in penile tissues from WT, PB-Cre4+ Smad4L/L (Smad4) and SA mice. h IHC stain for Ki67 in SA penile tumor. Scale bar 200鈥壩糾. i Sox2 expression silenced by two independent shRNA in SA1 cell line, detected by western blot. j Growth curves of subcutaneous tumors formed by control or Sox2 knockdown sublines of SA1 in nude mice (n鈥?鈥?). k, l Weight and gross images of subcutaneous tumors formed by control or Sox2 knockdown sublines of SA1 at endpoint (Day 83) (n鈥?鈥?). In j, k, data represent mean鈥壜扁€塖D. **P鈥?lt;鈥?.01, ****P鈥?lt;鈥?.0001, two-sided Student鈥檚 t test.Full size imageInfiltration of immunosuppressive myeloid cellsTranscriptional profiling suggests massive immune response in the SA penile tumors. To examine the immune infiltration, we dissociated penile tumors from 4鈥? month-old SA mice and normal penile epithelium from Cre- littermates, and catalogued the intratumoral immunocytes by mass cytometry (CyTOF) using a 35-antibody panel31. Consistent with the enrichment of various immune-related signaling pathways, there was a 7-fold increase of CD45+ immune cells (as fraction of all live cells) in mouse PSCC compared with normal penis (Fig.聽3a). When quantified as percentage of all immune cells, there were significant decreases of CD8+ T cells, NK cells, B cells and tumor-associated macrophages (TAMs), yet there was a dramatic increase from ~1% to ~50% of CD11b+ Gr1+ myeloid cells (Fig.聽3b, c, and Supplementary Fig.聽3). We isolated the CD11b+ Gr1+ cells from established SA tumors and demonstrated their strong ability to suppress T cell proliferation (Fig.聽3d). This result indicates that the intratumoral CD11b+ Gr1+ cells belong to myeloid-derived suppressor cells (MDSCs) that have been recognized as a major immunosuppressive population to induce T cell tolerance in solid tumors. Recent work from others and us indicated that PI3K signaling is crucial to mediate the immunosuppressive activity of myeloid cells31,32,33. In accordance with this notion, viSNE analysis of the CyTOF data showed stronger phospho-PI3K, phospho-mTOR and phospho-S6 signals in the CD11b+ Gr1+ population compared with other immune cells (Fig.聽3c). This result suggests that targeting PI3K signaling may be an effective approach to attenuate MDSC activities in the mouse PSCC and restore the response to immune checkpoint blockade (ICB) therapy. Given that the infiltrating T cells in the SA tumors are positive for PD1 (Fig.聽3e), it is rational to test whether co-targeting MDSCs synergizes with the ICB therapy.Fig. 3: Infiltration of CD11b+ Gr1+ immunosuppressive myeloid cells in mouse penile cancer.a Frequency of CD45+ immune cells of all live cells in penile tissues from Cre- littermates (control, n鈥?鈥?) and SA mice (n鈥?鈥?, 4鈥? months of age), with data from CyTOF. b Frequency of indicated immune cell subpopulations of all immune cells for the same sets of penile samples as in a. c Representative viSNE plots of indicated markers for the same sets of penile samples as in a. Note that the sparse dots and numerous dots for Cre- littermate or SA mouse, respectively, reflect the vast difference of the total intratumoral immune cell frequency of the two genotypes. d Activity of CD11b+ Gr1+ cells from SA tumors to suppress proliferation of normal spleen T cells stimulated with anti-CD3/CD28 antibodies. T cells were preloaded with CFSE whose signals declined as T cells divided. Higher MDSC: T cell ratio led to lower percentage of CFSElow T cells (n鈥?鈥?, 4, 6, and 3 for the ratios in order). e Representative viSNE plots of CD3e and PD1 (CD279) for SA tumors, showing the partial overlap of the two signals. In a, b, all data are shown with the line representing the median. In d, box plots visualize the five-number summary of a data set (minimum, lower quartile, median, upper quartile and maximum). *P鈥?lt;鈥?.05, **P鈥?lt;鈥?.01, ***P鈥?lt;鈥?.001, two-sided Student鈥檚 t test.Full size imageCombining targeted therapy and immunotherapyTo examine the efficacy of ICB in the SA model, we randomized SA males with established PSCC at 4鈥? month old to receive isotype IgG (control) or an anti-PD1/anti-CTLA4 antibody cocktail (ICB) at doses described recently31. The treatment lasted for one month and affected minimally on tumor weight (Fig.聽4a, b). To enhance ICB, we focused on two FDA-approved drugs, cabozantinib and celecoxib. Cabozantinib is a multi-targeted tyrosine kinase inhibitor and recently shown to block PI3K signaling in MDSCs to enable efficacy from ICB in metastatic prostate cancer31. Celecoxib is a selective COX2 inhibitor and highly relevant to our study, because Cox2 was upregulated in the SA tumors (Fig.聽2e) and identified as the top putative master regulator for inflammation-related phenotypes (Fig.聽2f). Cyclooxygenase-dependent tumor immune evasion is through myeloid cell reprograming and Cox2 blockade with celecoxib synergizes with anti-PD1 or dendritic cell immunotherapy in syngeneic models of a few cancer types34,35.Fig. 4: Combined targeted therapy and immunotherapy for spontaneous PSCC in mice.a Weight of resected penis from SA mice at the endpoint of the treatment course with single or combination therapies (n鈥?鈥?, 3, 5, 3, 3, and 4, respectively). b Representative images of penises for the SA mice before and after the 1-month treatment course. c, d Quantification of IHC staining for CD11b and Ly6G, markers of tumor infiltrating granulocytic MDSCs, using SA tumor samples treated differently (independent IHC images for CD11b: n鈥?鈥?, 5, 4, 9, 4, and 4, respectively; independent IHC images for Ly6G: n鈥?鈥?4, 10, 12, 12, 13, and 9, respectively). e Representative IHC images of Ly6G from the experiment in d. Scale bar 50鈥壩糾. f Quantification of IHC staining for Foxp3, marker of Tregs, using SA tumor samples treated differently (independent IHC images: n鈥?鈥?1, 6, 4, 3, 8, and 6, respectively). In a, box plots visualize the five-number summary of a data set (minimum, lower quartile, median, upper quartile and maximum). In c, d, and f, all data are shown with the line representing the median. #P鈥?gt;鈥?.05, *P鈥?lt;鈥?.05, **P鈥?lt;鈥?.01, ***P鈥?lt;鈥?.001, ****P鈥?lt;鈥?.0001, two-sided Student鈥檚 t test.Full size imageWhile the targeted therapy drugs, similar to ICB alone, caused marginal effect on penile tumor progression and endpoint weight, the combinations (cabozantinib plus ICB, celecoxib plus ICB) led to eradiation of most tumor nodules (Fig.聽4a, b). At the cellular level, cabozantinib or celecoxib was sufficient to significantly reduce the infiltration of myeloid cells positive for CD11b (Fig.聽4c and Supplementary Fig.聽4a) and Ly6G (Fig.聽4d, e), consistent with the reported effect from these drugs. On the other hand, ICB potently decreased the number of Foxp3+ regulatory T cells (Tregs) in the tumors (Fig.聽4f and Supplementary Fig.聽4b). Cabozantinib plus ICB, or celecoxib plus ICB treatment led to simultaneous reduction of both myeloid cells and Tregs (Fig.聽4c, d, and f), explaining the synergistic efficacy. Our results demonstrate that the SA model can serve as the platform to discover combination strategies for immunotherapy of spontaneous PSCC.Pten deletion confers resistance to cisplatinPI3K/Akt/mTOR pathway is relevant to both HPV-positive and HPV-negative PSCC. Infection by HPV 16, the most common high-risk HPV type in PSCC, induces PI3K/Akt/mTOR pathway which plays a critical role mediating the cellular entry of the virus36. Clinically, PIK3CA mutation is found to happen in 29% of PSCC cases37, and PTEN expression loss and PI3K/Akt/mTOR pathway activation is common in PSCC independently of HPV infection38,39. These findings prompted us to cross the SA mice with the conditional null allele of Pten to generate the PB-Cre4+ Smad4L/L ApcL/L PtenL/L mice (SAP genotype). Homozygous, but not heterozygous, loss of Pten dramatically accelerated the formation of penile tumors (Fig.聽5a). Median prolapse survival was 17.2 weeks and 8.1 weeks for SA and SAP mice, respectively. As expected, Pten loss in the penis augmented the phospho-Akt signals, shown in PB-Cre4+ PtenL/L mice and more pronounced in the SAP mice (Fig.聽5b). In contrast to the SA model, the SAP mice co-develop highly aggressive prostate tumors at a similar timeframe as the penile tumors. To address the question whether PSCC forms in SAP mice independently of the prostate tumors, we took two approaches. First, we performed radical prostatectomy on SAP male at 8 weeks of age and followed the penile tumor progression for 3 more weeks (Fig.聽5c). Without the prostate, SAP penile tumors continued to grow and at the endpoint stained strongly with Ki67 and weakly with cleaved caspase-3 (Fig.聽5d). Second, we established a PSCC cell line SAP1 from the SAP model and orthotopically injected into the penile and prostate tissues individually or simultaneously to evaluate whether orthotopic penile tumor growth was affected by orthotopic prostate tumors. The result showed robust injection-site-dependent orthotopic tumor formation (Fig.聽5e) and the orthotopic penile tumors remained salient SCC features such as keratin pearl and positivity for cytokeratin-5 (Fig.聽5f). Orthotopic injection of SA1 (the cell line derived from SA tumors) also generated PSCC, albeit with a longer latency (Supplementary Fig.聽5a). Therefore, these data rule out that the PSCC formed in the SAP model is due to distant effects or metastasis from the concomitantly formed prostate tumors.Fig. 5: SAP mice as a model to study chemoresistance of penile cancer.a Penile prolapse free survival curves for mice of three genotypes with n indicated. ****P鈥?lt;鈥?.0001, #P鈥?gt;鈥?.05, log-rank test. b Representative IHC images of phospho-Akt (Ser473). Scale bar 100鈥壩糾. c SAP mouse at 8 weeks old with all prostate lobes removed (anterior AP, dorsolateral DLP, ventral VP) and penile tumor followed for 3 weeks before sacrifice. d Representative H E and IHC images of Ki67 and cleaved caspase-3 (CC3) of the penile tumor from the prostate-removed SAP mouse (n鈥?鈥?). Scale bar 200鈥壩糾. e Orthotopic injection of SAP1 cells under the penile epithelium, or into the prostate gland or at both sites of 6-week Rag1鈭?鈭?/i> males (n鈥?鈥?). The mice were euthanized at Day 30 post-injection for organ weight measurement and imaging. Scale bar 10鈥塵m. (f) Representative penile tumor formed by penile injection of SAP1 with H E staining and IHC of cytokeratin-5 (n鈥?鈥?). Scale bar 10鈥塵m and 200鈥壩糾. g Weight of resected penis from SA and SAP mice at the endpoint of cisplatin treatment course (n鈥?鈥? for each group). h Representative images and H E staining of penises for the SA and SAP mice before and after the 1-month cisplatin treatment. Scale bars 5鈥塵m (bright field and GFP); 1鈥塵m (H E). In e, g, box plots visualize the five-number summary of a data set (minimum, lower quartile, median, upper quartile and maximum). #P鈥?lt;鈥?.05, **P鈥?lt;鈥?.01, ****P鈥?lt;鈥?.0001, two-sided Student鈥檚 t test.Full size imageThere are four general modes of resistance mechanisms for cisplatin40, and PI3K/Akt activation stands as an important post-target mechanism41,42. To evaluate the effect of Pten deletion on cisplatin response in penile cancer, we subjected SA and SAP mice with established penile tumors to cisplatin dosed at 10鈥塵g/kg weekly for a month. Cisplatin generated potent anti-tumor effect for SA mice, but affected SAP tumors minimally (Fig.聽5g, h). This anti-tumor effect on SA but not SAP models was likely the result of distinct sensitivity of the two models to cisplatin-induced apoptosis (Supplementary Fig.聽5b). Overall, the SAP model provides an essential tool to study cisplatin resistance.Targeted proteomics profiling and drug screenReverse-phase protein arrays (RPPA) is a technology for high-throughput protein activity measurement across many signaling pathways involved in cancer43. We performed RPPA profiling and observed a genotype-correlated protein expression pattern (Fig.聽6a and Supplementary Table聽7). Penises from wild type mice clustered together with those from PB-Cre+ ApcL/L and PB-Cre+ PtenL/L mice (no tumors), whereas SA and SAP tumors clustered in separate branches. Effect from cisplatin treatment (Fig.聽5c) was notable by the separation of treated tumor from untreated tumors in both genotypes. Despite the separation of SA and SAP samples in the clustering, there was a large overlap of differentially expressed protein species in SA and SAP tumors compared with WT penises (Fig.聽6b and Supplementary Table聽8). Most of the overlapped proteins were up- or down-regulated in the same direction in the two genotypes, with the exception of phospho-Akt (S473) which was upregulated in SAP tumors but downregulated in SA tumors (Fig.聽6b). Among the commonly upregulated proteins in SA and SAP tumors, proteins involved in metabolic regulation, cell motility and morphology, cell cycle and stemness were notable (Fig.聽6c). Interestingly, p16INK4a, whose overexpression is often found correlated with high-risk HPV infection44, was upregulated in both SA and SAP tumors (Fig.聽6c and Supplementary Fig.聽6a). The alternative product p19ARF (not included in RPPA) from the INK4a locus was also upregulated in both tumors (Supplementary Fig.聽6a). Another drastically overexpressed protein, CDH3 (i.e. P-Cadherin), was confirmed by IHC (Supplementary Fig.聽6b). Among the commonly underexpressed proteins (Fig.聽6d), AR was notable because of similar trend observed in clinical PSCC (Fig.聽1a). AR downregulation and exclusion from nucleus in SA and SAP tumors was evident by IHC analysis (Supplementary Fig.聽6c), consistent with our result showing AR signaling is dispensable for sustaining PSCC progression (Fig.聽1i-l). Using signals from the phospho-specific and total Akt antibodies, the calculated pS473/total-Akt ratio showed significant increase in SAP but not SA tumors (Fig.聽6e). This result was validated by western blot (Fig.聽6e) and consistent with Pten-loss induced PI3K/Akt pathway activation in SAP tumors.Fig. 6: RPPA profiling and drug screen using mouse penile cancer cells.a Unsupervised hierarchical clustering of penile samples from mice of denoted genotypes and treatment (cisplatin) based on normalized RPPA data. One-way ANOVA identified 117 differentially expressed proteins (P鈥?lt;鈥?.05), which were used for the clustering. b Venn diagram showing the relationship of differentially expressed proteins between SA vs. WT and SAP vs. WT comparisons, with data from RPPA. two-sided Student鈥檚 t test was used to select significant proteins for each comparison (P鈥?lt;鈥?.05). c Normalized RPPA signals of selected proteins upregulated in both SA and SAP tumors compared with WT penises, with protein names and functions shown (n鈥?鈥?, 8, and 4 for WT, SA and SAP, respectively). d Normalized RPPA signals of selected proteins downregulated in both SA and SAP tumors compared with WT penises (n鈥?鈥?, 8, and 4 for WT, SA and SAP, respectively). e Normalized RPPA signals and ratios for phospho-Akt (S473) and total Akt in WT, SA and SAP penises (n鈥?鈥?, 8 and 4 for WT, SA and SAP, respectively), validated by western blot of the tissues. f Dot plot comparing IC50 values of SA and SAP cell lines to 31 drugs, with three drugs highlighted. g, h Cell viability in vitro measured by MTT assay for indicated drugs, with IC50 noted (n鈥?鈥? for each concentration). i Western blot showing the upregulation of phospho-HER (Tyr1248 and Tyr1221) signaling in SAP tumors compared with SA tumors (n鈥?鈥? for each group). In c鈥?b>e, box plots visualize the five-number summary of a data set (minimum, lower quartile, median, upper quartile and maximum). In g, h, data represent mean鈥壜扁€塖D. #P鈥?gt;鈥?.05, *P鈥?lt;鈥?.05, **P鈥?lt;鈥?.01, ***P鈥?lt;鈥?.001, two-sided Student鈥檚 t test.Full size imageOverexpressed proteins in SA and SAP tumors provide the opportunity for the identification of potential therapeutic targets. We established 3 SA cell lines (SA1, SA2, SA3) and 2 SAP cell lines (SAP1, SAP2) from the two tumor models respectively. Compared with the tumor issues which contain heterogeneous cell populations, these cell lines showed cleaner pattern of respective gene knockouts (Supplementary Fig.聽6d, e). We used SA1 and SAP1 cell lines to perform a 42-compound dose titration assays with small molecules that target the overexpressed proteins and related pathways (Supplementary Table聽9). These cells showed medium to high sensitivity to 31 drugs (IC50鈥?lt;鈥?00鈥壩糓 for both lines), with SAP1 cells displaying comparable or higher IC50 for most drugs except for mubritinib (Fig.聽6f). In the validation assays, we confirmed the high sensitivity of both SA1 and SAP1 cells to staurosporine, a potent protein kinase C inhibitor with anti-tumor activity and low toxicity in a liposomally encapsulated form in preclinical models45. The higher sensitivity of SAP1 to mubritinib relative to SA1 (Fig.聽6h) is likely due to the on-target effect on the active HER2 signaling in SAP tumors (Fig.聽6i), because mubritinib is a selective HER2 tyrosine kinase inhibitor46. Together, the SA1 and SAP1 cell lines, in conjunction with the GEM models, provide the in vitro and in vivo models to develop therapeutic strategies for the treatment of lethal PSCC.Expression convergence of human and mouse PSCCIn order to determine if the spontaneous PSCC developed in SA and SAP mice resemble human PSCC at the molecular level, we compared the top 200 up- and down-regulated genes in human penile cancer47 with the mouse transcriptomic data. For the 172 out of the 200 human genes that have the mouse homologs, 108 (63%) and 89 (52%) showed significant difference in SAP and SA models, respectively (Fig.聽7a and Supplementary Table聽10), demonstrating significant convergence of gene expression pattern. IPA analysis of the Kuasne dataset identified several highly enriched pathways that were also enriched in SA tumors, in particular immune-related pathways such as Agranulocyte Adhesion and Diapedesis (P鈥?鈥?.1鈥壝椻€?0鈭?), Leukocyte Extravasation Signaling (P鈥?鈥?.003) and Eicosanoid Signaling (P鈥?鈥?.03) (Supplementary Table聽11). Wnt/尾-catenin Signaling was also highly enriched (Supplementary Fig.聽7, P鈥?鈥?.001), consistent with the enhanced cytoplasmic and nuclear 尾-catenin level (Supplementary Fig.聽2c). IPA identified potential upstream regulators that explain the transcriptional changes of human penile cancer, which included activated targets by pro-inflammatory factors such as TNF, CSF2, IFN纬, IL1B, and inhibited targets by TGF尾1 (Supplementary Table聽12). Interestingly, cytokines (IL1A, IL1B, TNF, CXCL1, CCL20) and MMPs (MMP7, 9, 10, 12, 13), upregulated in both human and mouse PSCC, showed no correlation with HPV status (Fig.聽7b). The strongly expressed inflammatory genes in both human and mouse PSCC prompted us to examine the immune cell markers in human penile cancer tissues. Compared with adjacent normal penile compartment, the malignant tissue was infiltrated extensively with T cells (CD3+) including the CTL subset (CD8+), as well as macrophages (CD68+) (Fig.聽7c). Recent reports showed frequent PD-L1 upregulation in PSCC48, which was confirmed in our small cohort (Fig.聽7c). We also validated that SOX2 and COX2 were upregulated in human PSCC (Fig.聽7c). Together, the resemblance of murine PSCC model with human PSCC underscores the clinical relevance of the mouse model.Fig. 7: Conserved gene expression pattern of human and mouse penile cancers.a Conserved expression pattern of most up- and down-regulated genes in human penile cancer relative to normal human penis as compared with mouse penile cancer (SA, SAP) relative to normal mouse penis (n鈥?鈥?, 3 and 2 for WT, SA and SAP, respectively). Human data were from GSE5795547, n鈥?鈥?9 (12 are HPV+, 25 are HPV鈭?/sup>, 2 with unknown HPV status, all referenced to normal glan tissues). b Normalized RNA levels of representative upregulated genes in human and mouse PSCC. For human, HPV+ and HPV鈭?/sup> cases were plotted separately, but showed no statistically significant differences (n鈥?鈥?2 for human HPV+, 25 for human HPV鈭?/sup>, 3 for mouse SA, and 2 for mouse SAP). All data points are shown with the line representing the median. c IHC images of selected immune cell markers and signaling molecules in human penile tumors. In all cases, more than half of the eight cases were positive for the marker being stained (n鈥?鈥?). Scale bar 100鈥壩糾.Full size imageDiscussionOur studies build essential resources for translational research in penile cancer. The GEM model of PSCC in our studies is based on the use of PB-Cre4 as the Cre driver in AR-expressed androgen-responsive epithelial cells of the mouse penis. Androgen-AR signaling has a fundamental role in masculinization during development, initiating formation of the prostate, penis and tissues of the male reproductive tract49. Transcriptional activity of AR is essential for normal penile growth, as inhibition of androgen in the neonatal phase induces micropenis in mice50. Our models indicate that the androgen-sensitive epithelial cells are the potential cells of origin for PSCC, although AR expression is unlikely to be essential for sustaining tumorigenesis based on its diminished level in established PSCC (Fig.聽1a). Given the variety of tumor suppressor genes and oncogenes manipulated to model prostate cancer using the PB-Cre4 driver, it is remarkable that concurrent inactivation of Smad4 and Apc were the only reported genetic changes sufficient to drive oncogenesis of mouse penile epithelium. APC loss and WNT pathway activation may also play a key tumorigenic role in human PSCC. Inactivation of APC through mutation, promoter methylation, and loss of heterozygosity is frequent in oral SCC51,52,53. WNT ligand and 尾-catenin downstream targets (WNT4, MMP7, CCND1, MYC) were found upregulated in PSCC54. It is worth noting that activation of WNT/尾-catenin targets may be achieved through various mechanisms. For example, loss of function of FAT1, caused by recurrent somatic mutations in various cancer types, leads to 尾-catenin nuclear localization and transcriptional activation of WNT downstream targets55. FAT1 mutation occurs in about 15% of penile cancer56 and may account for 尾-catenin activation in penile cancer. In the backdrop of Apc deletion, removal of the tumor suppressive function of TGF尾/SMAD (through Smad4 deletion) and installation of the tumor promoting function of PI3K/AKT/mTOR (through Pten deletion) independently promotes the formation of SCC of the mouse penis. In fact, Smad4 knockout in head and neck epithelia of mice led to development of head and neck SCC57, which together with our results suggest the general role of Smad4 as a gatekeeper gene for SCC. We like to note that our model should raise a concern to the many investigators who use PB-Cre4 to model prostate diseases like prostate cancer. Since this Cre driver is active in penile epithelium, attention should be made to whether the prostate phenotype might be influenced by the potentially concurrent phenotype of the penile tissue.Among the large number of downstream targets of 尾-catenin activation, we focused on Sox2 and Cox2, both of which are relevant to PSCC. SOX2 is frequently amplified in SCC, including 32% of penile tumors58. PSCC tissues express high level of COX2, PGE2 and EP receptors28. SOX2 and COX2 may function as the master regulators to orchestrate intrinsic and extrinsic mechanisms of penile malignancy in the mouse model, in a similar manner as the cooperation of Sox2 and inflammation-inducing Stat3 to transform foregut basal progenitor cells59. The essential role of Sox2 in controlling the self-renewal of tumor-initiating cells in SCC have been shown in the skin, head and neck, and lung27,60,61. In Apc-deficient PSCC, Sox2 may collaborate with 尾-catenin to regulate the pro-tumor transcriptome, such as upregulation of cyclin D1 as shown previously in breast cancer62. We have shown that Sox2 knockdown attenuated mouse PSCC growth (Fig.聽2i鈥搇). Future studies will uncover precise mechanisms of Sox2 in PSCC. The most striking feature of the infiltrating immunoprofile of the SA tumors is the massive CD11b+ Gr1+ MDSCs, which may result from the induction by COX2-mediated PGE2 production in the milieu, a phenomenon observed in both murine and human MDSCs63,64,65. The vulnerability of MDSCs to the inhibition of induction signals (e.g. COX2-prostaglandins) and survival signals (e.g. PI3K) provides opportunities to block these cells and foster a more hospitable environment for immunotherapy to work, as we demonstrated in the SA models (Fig.聽4).The rationale to treat metastatic PSCC with ICB certainly exists, with especially anti-PD-1 or anti-PD-L1 therapies66. PD-L1 is detected in 40鈥?0% of PSCC cases and is correlated with poor prognosis and the number of tumor-infiltrating lymphocytes48. In a case report, a patient with an HPV-negative PSCC treated with nivolumab experienced a partial response67. Currently, there are multiple ICB trials recruiting patients with PSCC (e.g. NCT02837042, NCT02721732, NCT03333616, NCT02834013, and NCT03391479). Our data suggest that better response may be achieved when ICB is combined with targeted therapeutics that eliminates activity from immunosuppressive cells. In a recent Phase I clinical trial on metastatic genitourinary malignancy using cabozantinib with nivolumab and ipilimumab (NCT02496208), manageable toxicity and promising antitumor activity was observed, including partial response for two out four PSCC patients68. An important question is whether HPV status should be a criterion for enlisting PSCC patients in ICB trials. Given that PD-L1 expression in PSCC is not correlated with HPV infection48, together with the observed inflammatory gene signature in both HPV-positive and negative clinical PSCC samples (Fig.聽7b), we argue for the concept of treating both HPV-positive and negative PSCC with combined ICB and MDSC-inhibiting drugs.The PSCC resources developed in our studies include two GEM models and related murine PSCC cell lines. The applications of these reagents could have an impact on developing effective targeted therapy and immunotherapeutic approaches for patients with PSCC in addition to defining pathways of resistance.MethodsAnimalsAll animal works performed in this study were approved by The University of Texas MD Anderson Cancer Center Institutional Animal Care and Use Committee and The University of Notre Dame Institutional Animal Care and Use Committee. Acquisition and maintenance of the genetic alleles PB-Cre4, PtenL/L, Smad4L/L, mTmGL/L and ApcL/L were recently described in our studies31,69. C57BL/6 and Rag1 knockout were purchased from Jackson Laboratories (stock# 000664 and 003145, respectively). NCr nude mice were purchased from Taconic (stock# NCRNU-M). Since penile cancer is a male disease, all the experimental cohorts used male mice. The mice were maintained in 12鈥塰/12鈥塰 light/dark with an ambient temperature of 72鈥?sup>oF and humidity of 35%. All animals were maintained in pathogen-free conditions and cared for in accordance with the International Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) policies and certification.Cell cultureLentivirus packaging cell line 293T (ATCC, CRL-3216) was cultured in DMEM, 10% fetal bovine serum (FBS), 100鈥塙/mL (1x) Penicillin-Streptomycin. Murine penile cancer cell lines SA1, SA2, SA3, SAP1 and SAP2 were established from individual established SA and SAP penile tumors. All the cells were cultured in DMEM, 10% fetal bovine serum (FBS), 100鈥塙/mL (1x) Penicillin-Streptomycin. The cell lines were of the male gender and confirmed free of mycoplasma with MycoAlert Mycoplasma Detection Kit (Lonza).Clinical samplesThe eight patient specimen cohort (see Supplementary Table聽1) was retrieved from banked archival tissue specimens among patients who had previously consented to have their incidental specimens banked for future research at the University of Texas M.D. Anderson Cancer Center utilizing an approved protocol. Specific blocks were selected based upon the abundance of tissue for additional studies (i.e., usually tumors greater than 1鈥塩m in size) and the patients were further characterized with respect to age, clinical stage, grade, histology, type of surgery, and Human papillomavirus status. De-identified specimens were provided to the research team under an institutional protocol (PA15-0138) for further characterization under waiver of consent.Histology, tissue array, immunostaining, and western blotTissues were fixed in 10% formalin overnight and embedded in paraffin. Hematoxylin eosin staining was performed using standard protocols on 5鈥壩糾 paraffin sections. Histologic evaluation of H E stained sections of mouse penile tumors was performed by two urological pathologists (P.T. P.R.).For immunohistochemistry (IHC), 5鈥壩糾 paraffin sections were deparaffinized, followed by pressure cooker boiling in antigen retrieval solution (10鈥塵M sodium citrate, pH鈥?鈥?). Subsequent steps followed manual instructions of VECTASTAIN Elite ABC-HRP Kit (Vector Laboratories) with DAB substrate (Vector Laboratories). Primary antibodies are listed in Key Resources Table. IHC slides were scanned with Pannoramic Digital Slide Scanner (3DHISTECH) and images were cropped from virtual slides in Pannoramic Viewer. For tissue array studies of 尾-catenin IHC of human normal penile tissue and PSCC, we used two commercially available tissue arrays, PE241 and PE2081, from US Biomax, Inc. PE241鈥塰ad 20 PSCC cases and PE2081 had 174 PSCC cases (3 had no tissues). Together these two arrays had seven adjacent normal tissues. The tissue array slides were deparaffinized and stained with 尾-catenin antibody (Cell Signaling Technology).For immunofluorescence (IF) staining, tissues were harvested freshly and cryopreserved in Optimal Cutting Temperature (OCT) Compound. Cryosections were blocked in 5% normal goat serum (Vector Laboratories), and incubated with primary antibodies (see Key Resources Table) overnight at 4鈥壜癈. Samples were then incubated with secondary antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 568 (Invitrogen) at RT for 1鈥塰. Slides were mounted with VECTASHIELD mounting medium with DAPI (Vector Laboratories). Stained slides were imaged using Nikon A1R Confocal Laser Microscope and images were managed with ImageJ software.For western blot, fresh tissues were soaked in ice-cold RIPA buffer (Boston BioProducts) supplemented with protease and phosphatase inhibitors (Roche), and homogenized with Fisherbrand 150 Handheld Homogenizer. Samples were centrifugation at 15,000脳g for 10鈥塵in at 4鈥壜癈. Proteins were resolved by SDS鈥揚AGE, transferred to nitrocellulose membrane and analyzed by immunoblot using primary antibodies listed in the Key Resources table. Anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase (Cell Signaling Technology) were used as secondary antibodies. Detection was performed with Clarity Max Western ECL Substrate (Bio-Rad) followed by exposure to Amersham ECL Hyperfilm (GE Healthcare).Reverse phase protein arrayReverse phase protein array (RPPA) experiment was performed at the RPPA core at MD Anderson Cancer Center. Total 27 freshly dissected mouse penile normal or tumor samples were probed with 299 antibodies. Briefly, tumor lysates were serially diluted two-fold for 5 dilutions (from undiluted to 1:16 dilution) and arrayed on nitrocellulose-coated slides in an 11鈥壝椻€?1 format. Samples were probed with antibodies by tyramide-based signal amplification approach and visualized by DAB colorimetric reaction. Slides were scanned on a flatbed scanner to produce 16-bit tiff image. Spots from tiff images were identified and the density was quantified by Array-Pro Analyzer. Relative protein levels for each sample were determined by interpolation of each dilution curves from the standard curve (supercurve) of the slide (antibody). These values (given as Log2 values) are defined as Supercurve Log2 (Raw) values. All the data points were normalized for protein loading and transformed to linear value and used for bar graphs. The antibody list, including vendor and catalog information, can be found on the website of the RPPA core at MD Anderson Cancer Center. For hierarchical clustering, normalized signals generated from rabbit and goat antibodies were analyzed using one-way ANOVA (alpha 0.05), and the resulted 117 significant proteins were clustered using average lineage method with Pearson correlation.Mass cytometry (CyTOF)The procedure for antibody staining and gating strategy was described as following70. Briefly, single cells from mouse penile tumors were prepared using the Mouse Tumor Dissociation kit (Miltenyl Biotec). All isolated cells were depleted of erythrocytes by hypotonic lysis, and blocked for Fc纬R using CD16/CD32 antibody (clone 2.4G2, BD Biosciences) and incubated with metal-conjugated antibodies targeting cell surface antigens for 30鈥塵inutes at room temperature. Cells were washed and incubated with 5鈥壩糓 Cell-ID Cisplatin-195Pt (Fluidigm) for 3鈥塵in for viability staining. Cells were washed and fixed following the steps of the Foxp3/Transcription Factor Staining Buffer Set (eBioscience). Fixed cells were stained with metal-conjugated antibodies targeting intracellular antigens for 1鈥塰our at room temperature. Cells were washed and incubated with Nucleic Acid Intercalator-Ir (Fluidigm) at 4oC overnight to stain all the nuclei. The samples were analyzed with CyTOF instrument (Fluidigm) in the Flow Cytometry and Cellular Imaging Core Facility at MD Anderson Cancer Center. CyTOF data were analyzed and visualized using Cytobank (http://www.cytobank.org).Therapeutic treatment in SA modelSA mice develop spontaneous penile tumors. Therefore, the mice were monitored daily for penile prolapse phenotype and, once the tumors reached 2鈥?鈥塵m in diameter, the mice were randomly assigned to one of the treatment arms of a preclinical trial. Cabozantinib (Selleck Chemicals, S1119) and celecoxib (LC Laboratories, C-1502) were orally administered at daily doses of 30鈥塵g/kg and 200鈥塵g/kg, respectively, daily on a Monday through Friday schedule. Anti-PD1 (clone RMP1-14, BioXcell, BE0146) and anti-CTLA4 (clone 9H10, BioXcell, BE0131) antibodies (or their respective isotype IgG controls) were intraperitoneally administered at 200鈥壩糶/injection three times/week. Cisplatin was intraperitoneally injected weekly at 10鈥塵g/kg (equivalent to 40鈥塵g/m2 in human). The duration of drug treatment was 4 weeks. At the endpoint, the penis from each treated mouse was resected and weighed, then fixed in formalin for histological analysis.Gene expression profiling and analysisThree 5-month-old SA mice were used for harvesting penile tumors. Age-matched WT (wild type) C57BL/6 mice were used for harvesting normal penile epithelium. Total RNA from fresh tissues was extracted using Direct-zol RNA Miniprep Kit (Zymo Research). The quantity and quality of each sample were measured using an Agilent 2100 Bioanalyzer. RNA sequencing was performed at Sequencing and Microarray Facility at MD Anderson Cancer Center using standard stranded RNA-seq protocol on an Illumina HiSeq 4000 (20 million pair-end sequence reads per sample).The raw reads were aligned to the mouse reference genome build mm10, using Tophat RNASeq alignment software. The aligned reads were verified for quality using FASTQC software. HTseq software was used to summarize the gene expression counts from Tophat alignment data after sorting the BAM files. The raw counts were normalized and differential expression analysis was performed using the DEseq2 package. Cluster analysis was performed with the log transformed normalized count Data. Heatmaps were used to depict the clustering and differential expression analysis results. Significant genes were defined by using a cut-off of 0.01 on the BH corrected P-value and an absolute log2(fold change) value of at least 2. Pathway analysis and upstream regulator analysis were performed with Ingenuity Pathway Analysis (QIAGEN). The raw and normalized data files are deposited by Basepair Tech to Gene Expression Omnibus (GEO) with series entry number GSE130052.To examine if the differentially expressed genes between mouse penile tumor and normal penile epithelium were also dysregulated in human penile cancer samples, we downloaded published transcriptomic dataset of human penile cancer GSE5795547 and identified the top 200 up- and down-regulated genes. Among these 200 genes, 172 have mouse homolog genes. Among these 172 genes, 108 (63%) and 89 (52%) genes showed significant difference in SAP and SA models (penile tumor compared with normal mouse penis in the RNA-seq data), respectively. The results were displayed as a clustering heatmap (Fig.聽7a). For a number of selected genes, human and mouse gene RNA levels (normalized to normal penile tissues) were plotted to demonstrate consistent cross-species upregulation of these genes (Fig.聽7b).T cell proliferation assay and suppression by MDSC cocultureBriefly, to isolate MDSCs from SA penile tumors, tumors were harvest and digested by Mouse Tumor Dissociation kit (Miltenyi Biotec) to obtain the single cells. Total MDSCs were isolated using Mouse Myeloid-Derived Suppressor Cell Isolation Kit (Miltenyi Biotec). Total spleen T cells were isolated from spleen of 2-month-old male C57BL/6 mice (Jackson Laboratory) using the mouse Pan T Cell Isolation Kit II (Miltenyi Biotec). To assess the suppressive activities of MDSCs on T cell proliferation, we first labeled the T cells with 5鈥壩糓 carboxyfluorescein diacetate succinimidyl ester (CFSE) (Invitrogen). The labeled T cells were stimulated in an antigen-nonspecific manner with anti-CD3 and anti-CD28 antibodies following eBioscience鈥檚 protocol. MDSCs were added at different ratios to T cells. CFSE intensity was quantified 72鈥塰 later with BD LSRFortessa Cell Analyzer. Viable CD3+ T cells with CFSE peaks located left of the highest peak (i.e. no proliferation, thus no decline of CFSE intensity) were counted as CFSElow (i.e. proliferative) cells.Sox2 knockdown and in vivo experimentsTo knockdown Sox2 with shRNA, scramble control SHC202 or Sox2-specific shRNA (Sigma-Aldrich, TRCN0000416106 as #2, TRCN0000420955 as #3) lentiviruses were packaged with 293T cells and used to infect SA1 cell line to generate puromycin-resistant stable sublines. The knockdown was confirmed by western blot. For in vivo tumorigenesis, 1鈥壝椻€?06 cells in 100鈥壩糽 PBS were injected subcutaneously in the dorsal skin of male 6-week-old NCr nude mice. Tumor dimensions were measured with a digital caliper and tumor volume was determined as 1/2 (length鈥壝椻€墂idth2). The mice were euthanized at the experimental endpoint, and tumors were dissected, weighted and photographed.Inhibitor screening using mouse penile cancer cell linesInhibitors were prepared as 10鈥塵M stock in DMSO or water according to manufacturer鈥檚 instructions. SA1 and SAP1 cell lines were seeded at 104 cells/well in 96-well plates in the absence or presence of inhibitors at 0.2, 2.0, and 20鈥壩糓 in triplicates. In all, 48鈥塰 later, 3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) was added to the wells at final concentration of 0.5鈥塵g/ml. The plates were gently shaken at 500鈥塺pm for 1鈥塵in, and incubated at 37鈥壜癈 for 4鈥塰. The medium was removed completely, and 100鈥壩糽 DMSO was added into each well. The plates were gently shaken at 1000鈥塺pm for 5鈥塵in to mix well the colored solution, and placed to microplate reader (Epoch 2, BioTek) to read 540鈥塶m and 690鈥塶m (reference). A confirmative second round of MTT assay was performed with a more refined range of inhibitor concentrations to determine IC50 more precisely.Penile orthotopic injectionMale Rag1 knockout males are anesthetized by inhalation of 2% isoflurane, then placed on the back with arms and legs extended. The urogenital region is prepared for surgery with alcohol swab. Sterile insulin syringe with 30-gauge needle is loaded with 50鈥壩糽 of 1鈥壝椻€?06 mouse penile cancer cells. With one hand keeping the glans penis stretched, from the ventral side of the glans penis, the other hand aims the needle to the superficial epithelial spine layer and keeps the bevel parallel with the glans axis so that the needle does not penetrate the glans. Cells are slowly injected with the formation of a small bump at the injection site as the indication of successful injection.Radical prostatectomyAnesthetized mouse is placed on its back with arms and legs extended. An operating stereomicroscope is used for the entire surgery. Access to the pelvis is via a median abdominal incision (2.5鈥塩m). The prostate ventral lobes and dorsolateral lobes are bluntly separated from the urethra using a cotton swab. The anterior prostate is dissected using the seminal vesicles as landmarks, and the prostate is then detached from the dorsal urethra.Surgical castrationThe mouse was anesthetized and both testes of the mouse were pushed down into the scrotal sacs, followed by ~1鈥塩m incision through the skin along the midline of the scrotal sac. Then a 5鈥塵m incision was made in the membrane on one side of the midline. The testis was pushed out, dissected away from the fat pad, and removed with scissors. After that, the fat pad was pushed back into the scrotal sac. Repeat above steps for the other testis. Buprenorphine was used for analgesia after all surgical procedures. If it is sham-operation, there is no removal of the testis.Prostate orthotopic injectionThe mouse was anesthetized and placed on its back with arms and legs extended. The lower abdomen wall of the mouse was wiped with 70% ethanol followed by betadine swab. The prostate anterior lobe was exposed with a low midline abdominal incision about 1鈥塩m, and 30-gauge needle was used to inject of 5鈥壝椻€?05 cells in 50鈥壩糽 PBS. The prostate was returned to the peritoneum. Buprenorphine was used for analgesia after all surgical procedures.Statistics and reproducibilityData were presented as mean鈥壜扁€塻tandard deviation (SD) unless indicated otherwise. Number of biological samples (n) for the experiments was denoted in figure legends. Results showing representative experiments (such as micrographs) were repeated independently at least three times with similar results. Unless indicated otherwise, two-sided Student鈥檚 t test assuming two-tailed distributions was used to calculate statistical significance between groups (no assumption was made that variance was similar between the groups that are being statistically compared). P鈥?lt;鈥?.05 was considered statistically significant. Data analysis was assisted with Graphpad Prism version 7, Cytobank 6.3.1, R version 3.3.3, and Ingenuity Pathway Analysis (QIAGEN).Reagents and resourcesAll the reagents and resources used and generated in this study are listed in Supplementary Table聽13.Reporting summaryFurther information on research design is available in the聽Nature Research Reporting Summary linked to this article. The RNAseq data generated in the study have been deposited and available for public access in the Gene Expression Omnibus (GEO) database under the accession code GSE130052. The GSE57955 data referenced during the study are available in a public repository from the GEO website. All the other data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon reasonable request. A reporting summary for this article is available as a聽Supplementary Information file. References1.Clark, P. E. et al. Penile cancer: clinical practice guidelines in oncology. J. Natl Compr. Cancer Netw. 11, 594鈥?15 (2013).CAS聽 Article聽Google Scholar聽 2.Daling, J. R. et al. Penile cancer: Importance of circumcision, human papillomavirus and smoking in in situ and invasive disease. Int. J. Cancer 116, 606鈥?16 (2005).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 3.Olesen, T. B. et al. Prevalence of human papillomavirus DNA and p16INK4a in penile cancer and penile intraepithelial neoplasia: a systematic review and meta-analysis. Lancet Oncol. 20, 145鈥?58 (2019).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 4.Ferr谩ndiz-Pulido, C., de Torres, I. Garc铆a-Patos, V. Penile squamous cell carcinoma. Actas Dermo-Sifiliogr.谩ficas (Engl. Ed.) 103, 478鈥?87 (2012). Google Scholar聽 5.Lont, A. P. et al. Presence of high-risk human papillomavirus DNA in penile carcinoma predicts favorable outcome in survival. Int. J. Cancer 119, 1078鈥?081 (2006).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 6.Kayes, O., Ahmed, H. U., Arya, M. Minhas, S. Molecular and genetic pathways in penile cancer. Lancet Oncol. 8, 420鈥?29 (2007).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 7.Naumann, C. M. et al. Establishment and characterization of primary cell lines of squamous cell carcinoma of the penis and its metastasis. J. Urol. 187, 2236鈥?242 (2012).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 8.Mu帽oz J. J. D. S., Kuasne H., Villacis R. A., Marchi F. A., Domingues M. A., Lopes A., Santos T. G., Rogatto S. R. A comprehensive characterization of cell cultures and xenografts derived from a human verrucous penile carcinoma. Tumour Biol. 37, 11375鈥?1384 (2016).9.Zhou, Q. H. et al. Molecular characterization and integrative genomic analysis of a panel of newly established penile cancer cell lines. Cell Death Dis. 9, 684 (2018).Article聽 CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 10.Agarwal, G., Gupta, S. Spiess, P. E. Novel targeted therapies for the treatment of penile cancer. Expert Opin. drug Discov. 9, 959鈥?68 (2014).CAS聽 Article聽Google Scholar聽 11.Pagliaro, L. C. et al. Neoadjuvant paclitaxel, ifosfamide, and cisplatin chemotherapy for metastatic penile cancer: a phase II study. J. Clin. Oncol. 28, 3851鈥?857 (2010).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 12.Pettaway, C. A., Pagliaro, L., Theodore, C. Haas, G. Treatment of visceral, unresectable, or bulky/unresectable regional metastases of penile cancer. Urology 76, S58鈥揝65 (2010).Article聽 PubMed聽 PubMed Central聽Google Scholar聽 13.Wang, J., Pettaway, C. A. Pagliaro, L. C. Treatment for metastatic penile cancer after first-line chemotherapy failure: analysis of response and survival outcomes. Urology 85, 1104鈥?110 (2015).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 14.Kim, K. S. et al. Expression of the androgen receptor and 5 alpha-reductase type 2 in the developing human fetal penis and urethra. Cell Tissue Res 307, 145鈥?53 (2002).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 15.Tietjen, D. N., Uramoto, G. Y., Tindall, D. J. Husmann, D. A. Characterization of penile androgen receptor expression in micropenis due to hypogonadotropic hypogonadism. J. Urol. 160, 1075鈥?078 (1998).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 16.Wu, X. W. J., Huang, J., Powell, W. C., Zhang, J., Matusik, R. J., Sangiorgi, F. O., Maxson, R. E., Sucov, H. M. Roy-Burman, P. Generation of a prostate epithelial cell-specific Cre transgenic mouse model for tissue-specific gene ablation. Mechanisms Dev. 101, 61鈥?9 (2001).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 17.Muzumdar, M. D., Tasic, B., Miyamichi, K., Li, N. Luo, L. A global double-fluorescent cre reporter mouse. Genesis 45, 593鈥?05 (2007).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 18.Wang, S. et al. Prostate-specific deletion of the murine Pten tumor suppressor gene leads to metastatic prostate cancer. Cancer Cell 4, 209鈥?21 (2003).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 19.Ding, Z. et al. Telomerase reactivation following telomere dysfunction yields murine prostate tumors with bone metastases. Cell 148, 896鈥?07 (2012).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 20.Ding, Z. et al. SMAD4-dependent barrier constrains prostate cancer growth and metastatic progression. Nature 470, 269鈥?73 (2011).ADS聽 CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 21.Tommasino, M. The human papillomavirus family and its role in carcinogenesis. Semin. Cancer Biol. 26, 13鈥?1 (2014).CAS聽 Article聽 PubMed聽Google Scholar聽 22.Lichtig, H. et al. HPV16 E6 augments Wnt signaling in an E6AP-dependent manner. Virology 396, 47鈥?8 (2010).CAS聽 Article聽 PubMed聽Google Scholar聽 23.Rampias, T. et al. Activation of Wnt signaling pathway by human papillomavirus E6 and E7 oncogenes in HPV16-positive oropharyngeal squamous carcinoma cells. Mol. Cancer Res. 8, 433鈥?43 (2010).CAS聽 Article聽 PubMed聽Google Scholar聽 24.Lee, D. K., Kim, B.-C., Kim, I. Y., Cho E-a, Satterwhite, D. J. Kim, S.-J. The human papilloma virus E7 oncoprotein inhibits transforming growth factor-尾 signaling by blocking binding of the smad complex to its target sequence. J. Biol. Chem. 277, 38557鈥?8564 (2002).CAS聽 Article聽 PubMed聽Google Scholar聽 25.Mendoza, J.-A., Jacob, Y., Cassonnet, P. Favre, M. Human papillomavirus type 5 E6 oncoprotein represses the transforming growth factor 尾 signaling pathway by binding to SMAD3. J. Virol. 80, 12420鈥?2424 (2006).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 26.Bruxvoort, K. J. et al. Inactivation of Apc in the mouse prostate causes prostate carcinoma. Cancer Res 67, 2490鈥?496 (2007).CAS聽 Article聽 PubMed聽Google Scholar聽 27.Boumahdi, S. et al. SOX2 controls tumour initiation and cancer stem-cell functions in squamous-cell carcinoma. Nature 511, 246鈥?50 (2014).ADS聽 CAS聽 Article聽 PubMed聽Google Scholar聽 28.Golijanin, D. et al. Cyclooxygenase-2 and microsomal prostaglandin E synthase-1 are overexpressed in squamous cell carcinoma of the penis. Clin. Cancer Res. 10, 1024鈥?031 (2004).CAS聽 Article聽 PubMed聽Google Scholar聽 29.Shtutman, M. et al. The cyclin D1 gene is a target of the 尾-catenin/LEF-1 pathway. Proc. Natl Acad. Sci. USA 96, 5522鈥?527 (1999).ADS聽 CAS聽 Article聽 PubMed聽Google Scholar聽 30.Papadopoulos, O. et al. Expression of cyclin D1 and Ki-67 in squamous cell carcinoma of the penis. Anticancer Res 27, 2167鈥?174 (2007).CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 31.Lu, X. et al. Effective combinatorial immunotherapy for castration-resistant prostate cancer. Nature 543, 728鈥?32 (2017).ADS聽 CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 32.De Henau, O. et al. Overcoming resistance to checkpoint blockade therapy by targeting PI3K纬 in myeloid cells. Nature 539, 443鈥?47 (2016).ADS聽 Article聽 CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 33.Kaneda, M. M. et al. PI3K纬 is a molecular switch that controls immune suppression. Nature 539, 437鈥?42 (2016).ADS聽 CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 34.Zelenay, S. et al. Cyclooxygenase-dependent tumor growth through evasion of immunity. Cell 162, 1257鈥?270 (2015).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 35.Veltman J. D., et al. COX-2 inhibition improves immunotherapy and is associated with decreased numbers of myeloid-derived suppressor cells in mesothelioma. Celecoxib influences MDSC function. BMC Cancer 10, 464 (2010).36.Surviladze Z., Sterk R. T., DeHaro S. A., Ozbun M. A. Cellular entry of human papillomavirus type 16 involves activation of the PI3K/Akt/mTOR pathway and inhibition of autophagy. J. Virol. 87, 2508鈥?517 (2012).37.Andersson, P., Kolaric, A., Windahl, T., Kirrander, P., S枚derkvist, P. Karlsson, M. G. PIK3CA, HRAS and KRAS gene mutations in human penile cancer. J. Urol. 179, 2030鈥?034 (2008).CAS聽 Article聽Google Scholar聽 38.Chaux, A. et al. Immunohistochemical expression of the mammalian target of rapamycin pathway in penile squamous cell carcinomas: a tissue microarray study of 112 cases. Histopathology 64, 863鈥?71 (2014).Article聽 PubMed聽 PubMed Central聽Google Scholar聽 39.Ferrandiz-Pulido, C. et al. mTOR signaling pathway in penile squamous cell carcinoma: pmTOR and peIF4E over expression correlate with aggressive tumor behavior. J. Urol. 190, 2288鈥?295 (2013).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 40.Galluzzi, L. et al. Molecular mechanisms of cisplatin resistance. Oncogene 31, 1869鈥?883 (2012).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 41.Lee, S., Choi, E. J., Jin, C. Kim, D. H. Activation of PI3K/Akt pathway by PTEN reduction and PIK3CA mRNA amplification contributes to cisplatin resistance in an ovarian cancer cell line. Gynecol. Oncol. 97, 26鈥?4 (2005).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 42.Liu, L. Z., Zhou, X. D., Qian, G., Shi, X., Fang, J. Jiang, B. H. AKT1 amplification regulates cisplatin resistance in human lung cancer cells through the mammalian target of rapamycin/p70S6K1 pathway. Cancer Res. 67, 6325鈥?332 (2007).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 43.Akbani, R. et al. Realizing the promise of reverse phase protein arrays for clinical, translational, and basic research: a workshop report: the RPPA (Reverse Phase Protein Array) society. Mol. Cell. Proteom. 13, 1625鈥?643 (2014).CAS聽 Article聽Google Scholar聽 44.Cubilla A. L., Lloveras B., Alejo M. Value of p16(INK)(4)(a) in the pathology of invasive penile squamous cell carcinomas: a report of 202 cases. Am. J. Surg. Pathol. 35, 253鈥?61 (2011).45.Mukthavaram, R. et al. High-efficiency liposomal encapsulation of a tyrosine kinase inhibitor leads to improved in vivo toxicity and tumor response profile. Int. J. Nanomed. 8, 3991鈥?006 (2013). Google Scholar聽 46.Nagasawa, J., Mizokami, A., Koshida, K., Yoshida, S., Naito, K. Namiki, M. Novel HER2 selective tyrosine kinase inhibitor, TAK-165, inhibits bladder, kidney and androgen-independent prostate cancer in vitro and in vivo. Int J. Urol. 13, 587鈥?92 (2006).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 47.Kuasne, H. et al. Genome-wide methylation and transcriptome analysis in penile carcinoma: uncovering new molecular markers. Clin. Epigenetics 7, 46 (2015).Article聽 CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 48.Udager, A. M. et al. Frequent PD-L1 expression in primary and metastatic penile squamous cell carcinoma: potential opportunities for immunotherapeutic approaches. Ann. Oncol. 27, 1706鈥?712 (2016).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 49.Matsushita, S. et al. Regulation of masculinization: androgen signalling for external genitalia development. Nat. Rev. Urol. 15, 358鈥?68 (2018).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 50.Zheng, Z., Armfield, B. A. Cohn, M. J. Timing of androgen receptor disruption and estrogen exposure underlies a spectrum of congenital penile anomalies. Proc. Natl Acad. Sci. USA 112, E7194鈥揈7203 (2015).ADS聽 CAS聽 Article聽 PubMed聽Google Scholar聽 51.Uzawa, K. et al. Abnormalities of the adenomatous polyposis coli gene in human oral squamous-cell carcinoma. Int. J. Cancer J. Int. du Cancer 58, 814鈥?17 (1994).CAS聽 Article聽Google Scholar聽 52.Supic, G., Kozomara, R., Brankovic-Magic, M., Jovic, N. Magic, Z. Gene hypermethylation in tumor tissue of advanced oral squamous cell carcinoma patients. Oral. Oncol. 45, 1051鈥?057 (2009).CAS聽 Article聽 PubMed聽Google Scholar聽 53.Perez-Sayans, M. et al. The role of the adenomatous polyposis coli (APC) in oral squamous cell carcinoma. Oral. Oncol. 48, 56鈥?0 (2012).CAS聽 Article聽Google Scholar聽 54.Arya, M. et al. Targets of Wnt/ss-catenin transcription in penile carcinoma. PLoS One 10, e0124395 (2015).Article聽 CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 55.Morris, L. G. et al. Recurrent somatic mutation of FAT1 in multiple human cancers leads to aberrant Wnt activation. Nat. Genet 45, 253鈥?61 (2013).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 56.Feber, A. et al. CSN1 somatic mutations in penile squamous cell carcinoma. Cancer Res. 76, 4720鈥?727 (2016).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 57.Bornstein, S. et al. Smad4 loss in mice causes spontaneous head and neck cancer with increased genomic instability and inflammation. J. Clin. Invest 119, 3408鈥?419 (2009).CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 58.Maier, S. et al. SOX2 amplification is a common event in squamous cell carcinomas of different organ sites. Hum. Pathol. 42, 1078鈥?088 (2011).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 59.Liu, K. et al. Sox2 cooperates with inflammation-mediated Stat3 activation in the malignant transformation of foregut basal progenitor cells. Cell Stem Cell 12, 304鈥?15 (2013).ADS聽 CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 60.Lee, S. H. et al. SOX2 regulates self-renewal and tumorigenicity of stem-like cells of head and neck squamous cell carcinoma. Br. J. Cancer 111, 2122鈥?130 (2014).Article聽 CAS聽 PubMed聽 PubMed Central聽Google Scholar聽 61.Ferone, G. et al. SOX2 is the determining oncogenic switch in promoting lung squamous cell carcinoma from different cells of origin. Cancer Cell 30, 519鈥?32 (2016).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 62.Chen, Y. et al. The molecular mechanism governing the oncogenic potential of SOX2 in breast cancer. J. Biol. Chem. 283, 17969鈥?7978 (2008).CAS聽 Article聽 PubMed聽Google Scholar聽 63.Fujita, M. et al. COX-2 blockade suppresses gliomagenesis by inhibiting myeloid-derived suppressor cells. Cancer Res. 71, 2664鈥?674 (2011).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 64.Obermajer, N., Muthuswamy, R., Lesnock, J., Edwards, R. P. Kalinski, P. Positive feedback between PGE2 and COX2 redirects the differentiation of human dendritic cells toward stable myeloid-derived suppressor cells. Blood 118, 5498鈥?505 (2011).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 65.Prima, V., Kaliberova, L. N., Kaliberov, S., Curiel, D. T. Kusmartsev, S. COX2/mPGES1/PGE2 pathway regulates PD-L1 expression in tumor-associated macrophages and myeloid-derived suppressor cells. Proc. Natl Acad. Sci. USA 114, 1117鈥?122 (2017).CAS聽 Article聽 PubMed聽Google Scholar聽 66.McGregor, B. Sonpavde, G. Immunotherapy for advanced penile cancer - rationale and potential. Nat. Rev. Urol. 15, 721鈥?23 (2018).Article聽 PubMed聽Google Scholar聽 67.Trafalis, D. T., Alifieris, C. E., Kalantzis, A., Verigos, K. E., Vergadis, C. Sauvage, S. Evidence for efficacy of treatment with the anti-PD-1 mab nivolumab in radiation and multichemorefractory advanced penile squamous cell carcinoma. J. Immunother. 41, 300鈥?05 (2018).CAS聽 Article聽 PubMed聽Google Scholar聽 68.Nadal R. M., et al. Results of phase I plus expansion cohorts of cabozantinib (Cabo) plus nivolumab (Nivo) and CaboNivo plus ipilimumab (Ipi) in patients (pts) with with metastatic urothelial carcinoma (mUC) and other genitourinary (GU) malignancies. J. Clin. Oncol. 36 (Suppl. 6S), abstr. 515 (2018).69.Boutin, A. T. et al. Oncogenic Kras drives invasion and maintains metastases in colorectal cancer. Genes Dev. 31, 370鈥?82 (2017).CAS聽 Article聽 PubMed聽 PubMed Central聽Google Scholar聽 70.Wang, G. et al. Targeting YAP-dependent MDSC infiltration impairs tumor progression. Cancer Disco. 6, 80鈥?5 (2016).Article聽 CAS聽Google Scholar聽 Download referencesAcknowledgementsWe thank the members of our laboratories for helpful discussions. The research reported in this publication was supported by Award Grant Number# CA096297/CA096300 from the National Cancer Institute of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. X.L. was partially supported from Grant Numbers KL2 TR002530 and UL1 TR002529 (A. Shekhar, PI) from the National Institutes of Health, National Center for Advancing Translational Sciences, Clinical and Translational Sciences Award, Susan G. Komen Grant # CCR18548293, Boler Family Foundation endowment at University of Notre Dame. T.H. was partly supported by a graduate fellowship from China Scholarship Council. J.C. is supported by an ASCO Conquer Cancer Foundation Young Investigator Award.Author informationAffiliationsDepartment of Biological Sciences, Boler-Parseghian Center for Rare and Neglected Diseases, Harper Cancer Research Institute, University of Notre Dame, Notre Dame, IN, 46556, USATianhe Huang,聽Xi Cheng,聽Yanting Luo,聽Xuemin Lu聽 聽Xin LuTumor Microenvironment and Metastasis Program, Indiana University Melvin and Bren Simon Cancer Center, Indianapolis, IN, 46202, USATianhe Huang,聽Xi Cheng聽 聽Xin LuDepartment of Oncology, The First Affiliated Hospital of Xi鈥檃n Jiaotong University, Xi鈥檃n, Shaanxi, 710061, ChinaTianhe HuangDepartment of General Surgery, , Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, 200025, Shanghai, ChinaXi ChengDepartment of Cancer Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAJad ChahoudDepartment of Urology, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAAhmed Sarhan聽 聽Curtis A. PettawayDepartment of Pathology, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAPheroze Tamboli聽 聽Priya RaoDepartment of Pathology/Lab Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAMing GuoDepartment of Bioinformatics Computational Biology, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAGaniraju ManyamDepartment of Environmental Health, University of Cincinnati, Cincinnati, OH, 45267, USALi ZhangDepartment of Biochemistry and Molecular Biology, The University of Texas Health Science Center at Houston McGovern Medical School, Houston, TX, 77030, USAYu Xiang聽 聽Leng HanDepartment of Cancer Biology, The University of Texas MD Anderson Cancer Center, Houston, TX, 77030, USAXiaoying Shang,聽Pingna Deng,聽Y. Alan Wang聽 聽Ronald A. DePinhoMass Spectrometry Core Facility, School of Life Sciences, Westlake University, Hangzhou, 310024, Zhejiang, ChinaShan FengDepartment of Pharmaceutical Sciences, School of Pharmacy, University of Puerto Rico, San Juan, PR, 00936, USAMagaly Martinez FerrerUniversity of Puerto Rico Comprehensive Cancer Center, Medical Sciences Campus, San Juan, PR, 00936, USAMagaly Martinez FerrerAuthorsTianhe HuangView author publicationsYou can also search for this author in PubMed聽Google ScholarXi ChengView author publicationsYou can also search for this author in PubMed聽Google ScholarJad ChahoudView author publicationsYou can also search for this author in PubMed聽Google ScholarAhmed SarhanView author publicationsYou can also search for this author in PubMed聽Google ScholarPheroze TamboliView author publicationsYou can also search for this author in PubMed聽Google ScholarPriya RaoView author publicationsYou can also search for this author in PubMed聽Google ScholarMing GuoView author publicationsYou can also search for this author in PubMed聽Google ScholarGaniraju ManyamView author publicationsYou can also search for this author in PubMed聽Google ScholarLi ZhangView author publicationsYou can also search for this author in PubMed聽Google ScholarYu XiangView author publicationsYou can also search for this author in PubMed聽Google ScholarLeng HanView author publicationsYou can also search for this author in PubMed聽Google ScholarXiaoying ShangView author publicationsYou can also search for this author in PubMed聽Google ScholarPingna DengView author publicationsYou can also search for this author in PubMed聽Google ScholarYanting LuoView author publicationsYou can also search for this author in PubMed聽Google ScholarXuemin LuView author publicationsYou can also search for this author in PubMed聽Google ScholarShan FengView author publicationsYou can also search for this author in PubMed聽Google ScholarMagaly Martinez FerrerView author publicationsYou can also search for this author in PubMed聽Google ScholarY. Alan WangView author publicationsYou can also search for this author in PubMed聽Google ScholarRonald A. DePinhoView author publicationsYou can also search for this author in PubMed聽Google ScholarCurtis A. PettawayView author publicationsYou can also search for this author in PubMed聽Google ScholarXin LuView author publicationsYou can also search for this author in PubMed聽Google ScholarContributionsConceptualization, X.L. and C.A.P.; methodology, X.L. and C.A.P.; investigation, T.H., X.C., J.C., A.S., X.S., P.D., Y.L., S.F., and X.L.; formal analysis, P.T., P.R., M.G., G.M., L. Z., Y. X., L. H. and Xuemin L.; resources, P.T., P.R. and C.A.P.; writing鈥攐riginal draft, X.L.; writing鈥攔eview and editing, X.L. and C.A.P.; funding acquisition, Y.A.W., M.M.F., C.A.P. and X.L. Supervision, Y.A.W., M.M.F., C.A.P. R.A.D., and X.L.Corresponding authorCorrespondence to Xin Lu.Ethics declarations Competing interests The authors declare no competing interests. Additional informationPeer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.Publisher鈥檚 note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary information Andrew W. Hahn, Jad Chahoud, Matthew T. Campbell, Daniel D. Karp, Jennifer Wang, Bettzy Stephen, Shi-Ming Tu, Curtis A. Pettaway Aung Naing Investigational New Drugs (2021) Carlos E. Stecca, Marie Alt, Di Maria Jiang, Peter Chung, Juanita M. Crook, Girish S. Kulkarni Srikala S. Sridhar Oncology and Therapy (2021) CommentsBy submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate. Sign up for the Nature Briefing newsletter 鈥?what matters in science, free to your inbox daily.