Penile Cancer-Derived Cells Molecularly Characterized as Models to Guide Targeted Therapies

Penile cancer (PeCa) is a common disease in poor and developing countries, showing high morbidity rates. Despite the recent progress in understanding the molecular events involved in PeCa, the lack of well-characterized in vitro models precludes new advances in anticancer drug development. Here we describe the establishment of five human primary penile cancer-derived cell cultures, including two epithelial and three cancer-associated fibroblast (CAF) cells. Using high-throughput genomic approaches, we found that the epithelial PeCa derived- cells recapitulate the molecular alterations of their primary tumors and present the same deregulated signaling pathways. The differentially expressed genes and proteins identified are components of key oncogenic pathways, including EGFR and PI3K/AKT/mTOR. We showed that epithelial PeCa derived cells presented a good response to cisplatin, a common therapeutic approach used in PeCa patients. The growth of a PeCa-derived cell overexpressing EGFR was inhibited by EGFR inhibitors (cetuximab, gefitinib, and erlotinib). We also identified CAF signature markers in three PeCa-derived cells with fibroblast-like morphology, indicating that those cells are suitable models for PeCa microenvironment studies. We thus demonstrate the utility of PeCa cell models to dissect mechanisms that promote penile carcinogenesis, which are useful models to evaluate therapeutic approaches for the disease.


Introduction
Penile cancer (PeCa) is an aggressive and mutilating disease that presents a high incidence in poor and developing countries [1,2]. According to two population-based cancer registry surveys, the survival of PeCa patients has not improved in Europe or the United States in the last three decades [3,4]. Although ongoing preclinical studies show promising results for more personalized therapy, the lack of improvement in survival is probably due to delayed diagnosis and lack of advances in curative standardized treatment options (reviewed in [5]).
The doubling time was calculated by seeding the cells at a density of 1.5 × 10 5 cells in a 6 cm 2 plate. Cell count was performed using Trypan blue (GIBCO, Carlsbad, CA, USA) dye exclusion in a Neubauer chamber every 24 h for eight days (the medium was replaced every three days). The doubling time was determined from the growth curves using the data from three independent experiments, each with three technical replicates.
The doubling time was calculated for all cell cultures ( Figure S1B). Cells 2, 3, 4, and 6 presented an average doubling time of 23.1 h, whereas cell 5 had a doubling time of 27.7 h ( Figure S1B).
The ability of all cell lines to migrate and invade was evaluated using transwell assays. Cell 4 presented the highest average score of migration (score 4), followed by cells 3, 6 (score 3), and 2 (score 2.5), with cell 5 showing the lowest score (score 0.5) ( Figure 1C,D). The invasive potential was high in cells 3, 4, and 6 (2 to 4 scores) and low in cell 2 (0 to 1 score), whereas cell 5 did not invade (0 score) ( Figures 1D and S1C). The ability to migrate and invade was not related to the morphology of the different cell lines.

Penile Cancer-Derived Cells Recapitulate the Molecular Profile of PeCa Primary Tissues
Genomic copy number alterations found in primary tumors were compared with their PeCa-derived cells (Cytoscan HD platform, Thermo Fisher Scientific, Waltham, MA, USA), revealing a high similarity level. The cells presented at least 80% matching single nucleotide polymorphisms with the primary tumor, confirming their parental origin ( Figure 2A). In this analysis, cells 3 and 6 presented 100% genomic similarities with their matched primary tumors. A lower similarity was observed between the tumor and its derived cell 5 (~80%). Supplementary Table S1 and Figure S3 show the genomic alterations identified in primary tumors and their derived cells.
Overall, the primary tumors and their derived cells presented a lower number of chromosomal imbalances, with the highest number of altered regions in cell 3 (44) and the lowest in cell 4 (8). Cells with epithelial morphology shared a higher number of common alterations with the matched primary tissues. Primary tumor 2 and its derived cell 2 presented 24 chromosomal imbalances, including a homozygous loss of 9p21.3 (CDKN2A and CDKN2B genes). Cell 3 and its primary tumor 3 presented 28 genomic alterations in common (Table S1), including gains of 4q12 (PDGFRA), an amplification of 11p15.5 (H19 and IGF2 genes), and a loss of 6p25.3 (DUSP22). A loss of 8p, encompassing the DLC1 gene, was identified only in cell 3.
We also detected similar genomic alterations in more than one derived tumor cell (Table S1). Among them, cells 2 and 3 shared a loss of 9p21.3 (CDKN2A and CDKN2B), all PeCa derived cells presented gains of 14q32.33, and cells 5 and 6 presented gains of 17q21.31 (KANSL1). A detailed description of the genomic alterations detected in our cells paired with their respective primary tumors is presented in Table S1 and Figure S3.
Targeted next-generation sequencing was performed in all PeCa-derived cells and three matching primary tumors (patients 2, 5, and 6) using a custom panel composed of 105 cancer-related genes. After filtering, we identified variants in all but one cell line (cell 3). A total of 18 single nucleotide variants -SNVs (nonsynonymous, stop gains, splice site, three and five prime untranslated regions) and two indels mapping to 17 genes were detected. The variants (missense, frameshift, and truncation) identified in the primary tumors and their matched derived cells are represented in Figure 2B and Table 2. All variants found in tumor 2 were also observed in cell 2 (two variants of PIK3CA and one of FGFR1). Tumor 5 and its derived cell 5 presented variants in nine different genes, among them ALK, BRCA2, NOTCH1, and PIK3CA. One variant in a CpG island, in the promoter region of the APC gene, was common to tumor 6 and its derived cell 6. Two variants were found exclusively in cell 2 and not in its tumor (MMP1 and STAT3), whereas two variants were found only in the tumor from which cell 6 was derived (ERBB2 and MLH3). The gene most frequently altered was PIK3CA (two variants in cell/tumor 2, one in cell/tumor 5). Table 2 summarizes the tNGS results found in the cells compared with the primary tumors.

Penile Cancer-Derived Cells Recapitulate the Molecular Profile of PeCa Primary Tissues
Genomic copy number alterations found in primary tumors were compared with their PeCa-derived cells (Cytoscan HD platform, Thermo Fisher Scientific, Waltham, MA, USA), revealing a high similarity level. The cells presented at least 80% matching single nucleotide polymorphisms with the primary tumor, confirming their parental origin (Figure 2A). In this analysis, cells 3 and 6 presented 100% genomic similarities with their matched primary tumors. A lower similarity was observed between the tumor and its derived cell 5 (~80%). Supplementary Table S1 and Figure S3 show the genomic alterations identified in primary tumors and their derived cells. Overall, the primary tumors and their derived cells presented a lower number of chromosomal imbalances, with the highest number of altered regions in cell 3 (44) and the lowest in cell 4 (8). Cells with epithelial morphology shared a higher number of common alterations with the matched primary tissues. Primary tumor 2 and its derived cell 2 presented 24 chromosomal imbalances, including a homozygous loss of 9p21.3 (CDKN2A and CDKN2B genes). Cell 3 and its primary tumor 3 presented 28 genomic alterations in common (Table S1), including gains of 4q12 (PDGFRA), an amplification of 11p15.5 (H19 and  To evaluate the actively translated mRNAs, we performed a polysome profiling analysis using a sucrose gradient, which involves the separation of mRNAs into fractions according to the number of bound ribosomes. The mRNA profile from the polysomal fraction of each cell was investigated using the Clariom™ D Assay platform (ThermoFisher, USA) ( Figure 3A,B). A heatmap containing the top 250 differentially expressed genes (DEG) showed three clusters. The first was composed of cell 3 (epithelial morphology), the second cluster of cell 2 (epithelial morphology), and the third cluster of cells 4, 5, and 6 (fibroblast-like morphology-CAFs). Penile cancer-derived cells 4, 5, and 6 presented overexpression of markers related to a CAF signature (MMP2, REAB3B, COL6A1, COL6A2, CTSK, THY1, PDGFRA, DCN, and fibroblast activation ACTA2) [31,32] ( Figure 3B). The fibroblast activation protein α gene (FAP) did not show increased expression. The top 15 DEG of cells 2 to 6 compared with cell 1 (normal foreskin) and the number of DEGs for each comparison are shown in Table 3.
A single-sample gene set enrichment analysis (ssGSEA) was performed, and the main pathways identified as dysregulated in the PeCa cells are depicted in Figure 3C,D. Using oncogenic signatures ( Figure 3C) and the Reactome ( Figure 3D) sub-collection of the MSigDB, we identified that EGFR, PI3K, and mTOR pathways were dysregulated in cell 2. Cell 2 also presented a high score for several cell junction signatures ( Figure 3D). Cell 3 expressed genes related to the regulation of apoptosis (BCL2L1 and BCL2) and pathways related to SRC and ERK activation. Cells 4, 5, and 6 presented pathways related to the CAF phenotype, including YAP and MYC signatures ( Figure 3C,D).
We also evaluated the protein profile of 304 pre-selected antibodies involved in pathways potentially dysregulated in cancer ( Figure S2). Figure 3E shows the main altered proteins or phosphorylated isotypes in PeCa cells. Potential therapeutic targets were Cells 2021, 10, 814 9 of 20 identified among the dysregulated proteins identified in the PeCa cells. Cell 2 presented increased expression of EGFR, HER3, and VEGFR2, which are described as targets for therapy and tested in clinical trials (NCT01728233-dacomitinib). Increased EGFR expression was observed in cell 2 ( Figure 3F). In addition, several proteins downstream to the EGFR receptor were dysregulated ( Figure 3E). Cell 2 also presented a decreased expression of p16INK4A, which correlates with the genomic deletion of the CDKN2A gene detected by CNA analysis (Figure S3A).
To evaluate the actively translated mRNAs, we performed a polysome profiling analysis using a sucrose gradient, which involves the separation of mRNAs into fractions according to the number of bound ribosomes. The mRNA profile from the polysomal fraction of each cell was investigated using the Clariom™ D Assay platform (ThermoFisher, USA) ( Figure 3A,B). A heatmap containing the top 250 differentially expressed genes (DEG) showed three clusters. The first was composed of cell 3 (epithelial morphology), the second cluster of cell 2 (epithelial morphology), and the third cluster of cells 4, 5, and 6 (fibroblast-like morphology-CAFs). Penile cancer-derived cells 4, 5, and 6 presented overexpression of markers related to a CAF signature (MMP2, REAB3B, COL6A1, COL6A2, CTSK, THY1, PDGFRA, DCN, and fibroblast activation ACTA2) [31,32] ( Figure  3B). The fibroblast activation protein α gene (FAP) did not show increased expression. The top 15 DEG of cells 2 to 6 compared with cell 1 (normal foreskin) and the number of DEGs for each comparison are shown in Table 3.   High levels of phosphorylated AKT were observed in cells 3 and 5. The epithelial cells marker, epiplakin (VHL-PPK1), was highly expressed in cells 2 and 3. A weak expression of epiplakin was observed in cell 4, whereas no expression of this marker was found in cells 5 or 6. Vimentin, a marker of mesenchymal cells, was expressed in cells 5 and 6 ( Figures 3E and S2).
The expression levels of PI3K/AKT/mTOR, EGFR, ERBB2, TP53, and CDKN2A proteins found in our cells were previously reported in PeCa (Table 4). In addition to EGFR and CDKN2A described above, we also detected alterations in targetable tyrosine kinase receptors, including ERBB2 and ERBB3, and several proteins downstream to this pathway, such as AKT, S6, and mTOR (Table 4, Figure 3E).

Identification of Potential Therapeutic Targets for PeCa and Chemo-Sensitivity Assays
We performed chemo-sensitivity assays in epithelial PeCa cells 2 and 3. Using the Ingenuity Pathway Analysis software (Qiagen, Valencia, CA, USA), we identified potential drugs that target the differentially expressed kinases and proteins that showed higher expression ( Table 5). The upstream regulator analysis of cell 2 revealed several genes predicted to be regulated by EGFR ( Figure 4A,B). EGFR was a core molecule in both mRNA ( Figure 4A) and protein analysis in cell 2 ( Figure 4B). Table 4. Summary of the studies in penile cancer that evaluated the expression (immunohistochemistry or immunofluorescence) of the same proteins presented in our RPPA (reverse-phase protein arrays) analysis. Based on the evidence that EGFR was dysregulated in cell 2 at mRNA and protein (RPPA and Western blot) levels ( Figure 3E,F), we tested the chemo-sensitivity of this cell using anti-EGFR inhibitors (cetuximab, gefitinib, and erlotinib). Cell 3 was used as a negative control since it did not show EGFR overexpression. We also tested cell viability in response to cisplatin (commonly used in the treatment of advanced PeCa) in cells 2 and 3. The IC50 and dose-response curve results showed that these cell lines were sensitive to cisplatin. However, only cell 2 (EGFR overexpression) was sensitive to anti-EGFR drugs ( Figure 4C).
Having shown that EGFR inhibition can block the proliferation of PeCa cells in vitro, we evaluated whether PeCa primary tissues overexpress the EGFR gene. Samples that presented overexpression of the EGFR signature most likely have activation of the EGFR pathway. We accessed EGFR mRNA-related signatures (MySigDB) in a set of 36 primary PeCa tissues previously published by our group (GSE57955) [35]. We identified a cluster of samples (~30% of the cases) showing EGFR mRNA signature overexpression ( Figure 5A,B). Several genes from the EGFR signature (downstream to EGFR) were positively correlated with EGFR expression ( Figure 5C).  Based on the evidence that EGFR was dysregulated in cell 2 at mRNA and protein (RPPA and Western blot) levels ( Figure 3E,F), we tested the chemo-sensitivity of this cell using anti-EGFR inhibitors (cetuximab, gefitinib, and erlotinib). Cell 3 was used as a negative control since it did not show EGFR overexpression. We also tested cell viability in response to cisplatin (commonly used in the treatment of advanced PeCa) in cells 2 and 3. The IC50 and dose-response curve results showed that these cell lines were sensitive to cisplatin. However, only cell 2 (EGFR overexpression) was sensitive to anti-EGFR drugs ( Figure 4C).
Having shown that EGFR inhibition can block the proliferation of PeCa cells in vitro,

Discussion
An in vitro culture of tumor cells is a valuable model for drug development and preclinical drug testing in oncology. At least in part, these cells maintained the molecular alterations of the parental tumor and thus can be used in functional studies [56,57]. However, establishing cancer cell lines from fresh tumor tissues represents a technical challenge [58]. These models also present certain limitations (such as the lack of tumor heterogeneity, the absence of components of the tumor microenvironment, and genotypic and phenotypic drift during culture) that must be taken into consideration [59].
The development of in vitro models of PeCa is hampered by the low incidence of this tumor type. In this study, we derived PeCa cells from fresh penile primary tumors. Two of the established PeCa-derived cells presented a typical polygonal epithelial cell morphology (cells 2 and 3), and three presented fibroblast-like morphology (cells 4, 5, and 6). These CAFs are components of the tumor microenvironment playing crucial roles in tumor progression and treatment response [60,61]. Interesting approaches were developed by Di Donato et al [23]., where the prostate tumor growth was stimulated by androgen in co-cultures using CAFs derived from patients with positive AR (androgen receptor) and cell lines that were either AR positive or negative [23]. Bladder cancer progression, migration, and invasion were also shown to be triggered by factors secreted by CAFs [22]. In both cases, inhibiting CAF signaling sufficed to decrease tumor aggressiveness [22,23].
Although several studies indicated the influence of the microenvironment in response to chemotherapy, we selected only cells showing epithelial features (cells 2 and 3) to evaluate the treatment response since the selected drugs target the tumor cells and not the microenvironment.
The primary tumor from patient 2 was positive for high-risk HPV, whereas the derived cell 2 was HPV negative. Attempts to reproduce HPV replication in standard cell culture have been unsuccessful, mostly because the replication process is linked to the differentiation of keratinocytes, and it is challenging to recreate the stratified structure of the epithelium in vitro [12,62].
The primary tumor that generated cell 2 was a mixed tumor (usual with few sarcomatoid differentiation areas). Sarcomatoid PeCas are rare and aggressive tumors (1-2% of PeCa) associated with metastasis and poor prognosis [63,64]. Using phase-contrast microscopy, we exclusively detected epithelial cells without spindle cells, expected in the sarcomatoid component ( Figure 1A,B). Indeed, microarray and RPPA analysis showed the expression of several epithelial markers in cell 2, including KRT5, KRT17, CDH1, DSG3 (Table 3), and epiplakin (VHL-PPK1) ( Figure S2). These results suggest that only the usual component of this mixed PeCa was selected in vitro.
Very few studies have evaluated the mRNA expression profile of PeCa cell lines [18,19]. To our knowledge, we were the first to evaluate the expression levels of ribosome-associated mRNAs (translatomic analysis) in PeCa-derived cells. Since these mRNAs are associated with several ribosomes, they are likely actively translated [65]. Cell 2 was the only PeCaderived cell that presented high EGFR overexpression at gene and protein levels. Gene set enrichment analysis ( Figure 3C,D) and upstream regulator analysis ( Figure 4A,B) also showed that several genes downstream to the EGFR pathway and EGFR interactors were dysregulated. Silva Amancio et al. [42] reported increased EGFR protein expression (3+: 67 of 139 cases) and its association with cancer recurrence and perineural invasion. Chaux et al. [66] found EGFR overexpression in 44% of the PeCas evaluated. These studies suggested that targeting the EGFR pathway would be an effective therapeutic strategy to treat a subset of cases. In contrast, Zhou et al. [19] identified EGFR protein overexpression (Western blot) in four of five PeCa cell lines and resistance to erlotinib and afatinib. Only one of these cell lines showed a minor sensitivity to cetuximab. The authors suggested that HRAS and PI3KCA alterations might be related to anti-EGFR therapy resistance. A chemo-sensitivity assay with a panel of drugs targeting EGFR (cetuximab, gefitinib, and erlotinib) showed that cell 2 was sensitive to anti-EGFR drugs ( Figure 4C).
Anti-EGFR agents have been used to treat PeCa patients, mainly as a salvage treatment after first-line chemotherapy failure [67][68][69]. Di Lorenzo et al. [67] reported that 50% of 28 patients (24 treated with cetuximab) were sensitive to anti-EGFR monoclonal antibodies [67]. PeCa patients treated with nimotuzumab after chemotherapy failure showed clinical response or stable disease [69]. Necchi et al. [70] reported that dacomitinib (pan-HER inhibitor) was active and well-tolerated in patients with advanced PeCa and may represent an option when combined chemotherapy cannot be administered. The evaluation of downstream effectors of EGFR signaling and how these mutations can affect therapy response over time should be taken into account. Cell 2 presented a good initial response against EGFR inhibitors, but the occurrence of the two PIK3CA pathogenic variants could be translated into the later acquisition of resistance to treatment, as described in metastatic colorectal cancer [71]. However, it was not possible to confirm this assumption because the patient was only treated by surgery and lost to follow-up. The effect of PIK3CA mutation and resistance to EGFR inhibitors in PeCa should be better investigated. Other receptors such as ERRB3, phospho-ERBB3, and VEGFR2 were also overexpressed in this cell line and represent potential therapeutic opportunities.
Several potentially targetable dysregulated oncogenic pathways have been described in PeCa [9,38,[72][73][74], including PI3K/Akt/mTOR, EGFR, and ERRB2, among others. Recently, clinical trials testing immunotherapy in PeCa patients have shown promising results [75]. Positive expression of PD-L1 was found in 48-60% of cases [75], indicating that a significant proportion of these patients do not respond to immune-checkpoint inhibitors. Genetic studies have shown that the PI3K/Akt/mTOR pathway is significantly altered in PeCa cases [9,38,73,74]. Here, cells 2 and 5 presented alterations involving the PI3K/Akt/mTOR signaling pathway axis. We identified several mechanisms involved in the dysregulation of this pathway in PeCa cells, including PIK3CA mutation and increased expression levels of mTOR, RICTOR, phosphorylated S6 (cell 2), and phosphorylated AKT (cell 5), among others. The PI3K/AKT/mTOR signaling pathway has been associated with cell proliferation, migration, metabolism, and survival [76]. This pathway was dysregulated in cell 5, suggesting that the microenvironment cells can also be a potential target for cancer therapy [77].
Previously we described a similar genomic profile of cell 3 at passages 1, 5, and 10, thus showing genomic stability after cell culture [10]. Here, we used translatomic and protein expression analyses to better characterize this cell line. The enrichment analysis using DEGs revealed increased expression levels of Hedgehog pathway genes. Hedgehog signaling pathway inhibitors have been developed for cancer treatment [78]. Cell 3 also presented the dysregulation of other targetable pathways such as SRC and ERK ( Figure 3C,D), widely investigated in several tumor types. This PeCa-derived cell presented a complex pattern of gene alteration that affected several oncogenic pathways. Thus, combined therapies targeting one or more dysregulated pathways are promising alternatives to treat PeCa, and clinical trials should be encouraged.
We showed that cell 2 and cell 3 were sensitive to cisplatin, but only cell 2 (EGFR overexpression) was sensitive to anti-EGFR drugs ( Figure 4C). As expected, cell 3 did not respond to EGFR inhibitors. The EGFR status is a crucial step to be investigated before treating patients with EGFR inhibitors. The in-depth molecular characterization used in our PeCa-derived cells can assist the selection of drugs to be tested in representative models of PeCa.
The genomic analysis performed in five penile cancer-derived cells revealed a high level of similarities with the primary tumors. The most relevant differentially expressed genes and proteins are components of critical oncogenic pathways supporting previous studies published in the literature. We identified proteins with the potential to be targeted in PeCa patients, including EGFR, VEGFR2, PIK3/AKT/mTOR, MAPK, and SRC. Considering that our PeCa-derived cells closely recapitulated the molecular features of their primary tumors and presented the same dysregulated pathways, they are excellent models to be used in preclinical tests. Moreover, the cancer-associated fibroblasts can be used for microenvironment-related studies.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/cells10040814/s1, Table S1: Genomic alterations shared in primary tumors and their derived PeCa cells, Figure S1: Characteristics of the penile cancer-derived cells; Figure S2: Heatmap of the reverse phase protein array (RPPA) showing the protein expression levels (in alphabetical order); Figure S3: Ideogram of primary tumors (thin lines in pink) and derived cell lines (thin lines in blue) showing the copy number alterations (CytoScan HD, Affymetrix).  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.