Comprehensive Genomic Profiling of Androgen-Receptor-Negative Canine Prostate Cancer

Canine carcinomas have been considered natural models for human diseases; however, the genomic profile of canine prostate cancers (PCs) has not been explored. In this study, 14 PC androgen-receptor-negative cases, 4 proliferative inflammatory atrophies (PIA), and 5 normal prostate tissues were investigated by array-based comparative genomic hybridization (aCGH). Copy number alterations (CNAs) were assessed using the Canine Genome CGH Microarray 4 × 44K (Agilent Technologies). Genes covered by recurrent CNAs were submitted to enrichment and cross-validation analysis. In addition, the expression levels of TP53, MDM2 and ZBTB4 were evaluated in an independent set of cases by qPCR. PC cases presented genomic complexity, while PIA samples had a small number of CNAs. Recurrent losses covering well-known tumor suppressor genes, such as ATM, BRCA1, CDH1, MEN1 and TP53, were found in PC. The in silico functional analysis showed several cancer-related genes associated with canonical pathways and interaction networks previously described in human PC. The MDM2, TP53, and ZBTB4 copy number alterations were translated into altered expression levels. A cross-validation analysis using The Cancer Genome Atlas (TCGA) database for human PC uncovered similarities between canine and human PCs. Androgen-receptor-negative canine PC is a complex disease characterized by high genomic instability, showing a set of genes with similar alterations to human cancer.


Introduction
Prostate cancer (PC) is the second most common cancer in men worldwide and its incidence is especially high in developed countries [1]. Among mammals, the dog is the only animal that naturally develops prostate cancer with aging [2,3]. Moreover, by sharing the same environment, dogs have been considered a natural model for human cancer [4]. The canine disease usually displays a more aggressive behavior and is associated with a high metastatic rate and poor prognosis [3]. Furthermore, the continuous basal cell layer observed in the human prostate is not found in canine normal prostates, and no specific anatomic zones are associated with canine PC and prostatic hyperplasia development [5][6][7][8][9]. At the immunophenotypical level, canine PC commonly present loss of androgen receptor (AR), NKX3.1, and PTEN expression [10,11]. The expression levels of these three proteins are associated with the aggressive behavior of canine PC, and resembles the clinical features of human hormone-refractory PC [7]. Human PC is, androgen dependent, and only more aggressive cancers show loss of NKX3.1, and PTEN expression [12,13]. Canine PC can develop from luminal, ductal, and urothelial cells [3]. Therefore, a morphological analysis and a panel of proteins (uroplakin III, CK8/18, CK7, PSA, and AR) should be used to differentiate luminal and urothelial origins in tumors with undifferentiated patterns [7,14].
Despite all the differences, there are also some similarities at pathological, clinical, and molecular levels between canine and human PC. The concomitant presence of prostatic intraepithelial neoplasia, proliferative inflammatory atrophy (PIA), and bone metastases as well as the deregulation of cancer-related transcripts/proteins, including MDM2, PTEN, TP53, CTNNB1, and CDH1, are common events in dog and human PCs [10,11,[15][16][17].
The first reports using comparative genomic hybridization (CGH) analysis to evaluate copy number alterations (CNAs) in dogs were performed in a glial tumor cell line and lymphomas [18,19]. The authors reported chromosomal imbalance similarities with human gliomas and lymphomas [18,19]. Subsequently, other neoplasms (transmissible venereal tumor, osteosarcoma, histiocytic sarcoma, and oral melanoma) were assessed through array-based CGH (aCGH), showing similar genomic alterations with the corresponding human tumors [20][21][22].
In canine PC, the cell line named Probasco, derived from a mixed breed dog with spontaneous tumors, presented genomic complexity, including large deletions, and duplications mapped in various chromosomes [23]. In this context, here, we report the genomic profile of canine pre-neoplastic prostate lesions and prostate carcinomas to better understand the molecular alterations involved in canine carcinogenic process. Moreover, three genes (TP53, ZBTB4, and MDM2) involved in regions with losses or gains were evaluated at the gene expression level. We also compared the genomic alterations in canine and human PCs. To our knowledge, this is the first study that has evaluated genomic imbalances in a cohort of dogs with spontaneous PC.
A total of 370 CNAs (98 gains, and 272 deletions) were found in 14 PC samples. Particularly, three cases (8C, 8D1, and 40DM) presented more than 60 CNAs. Almost all chromosomes had genomic alterations, except chromosome 19 ( Figure 1B). Losses of large segments were distributed along several chromosomes, particularly the chromosomes 2, 5, 6, 9, 10, 24, 26, and 28. Gains were also observed in numerous chromosomes but encompassing only small regions (details in Table S2). Moreover, 14 high copy gains (average log2 ratio > 2.0), and 12 homozygous deleted regions (average log2 ratio < −2.0) were observed in six and four cases, respectively. However, the majority of these CNAs covered no genes (Table S3). Recurrent alterations found in more than 20% of PC samples are depicted in Table S4. Likewise, 989 genes were affected by common CNAs (Table S5). Excluding the 221 microRNA genes, which involved one unique probe, 768 genes were evaluated using the Ingenuity Pathway Analysis (IPA) software. The in silico functional analysis revealed 655 of 768 (~85%) cancer-related genes (Table S6). Canonical pathway analysis identified several cancer-related pathways (Table S7), including molecular mechanisms involved with the disease (Figure 2A), and many specific signaling pathways in different types of cancer (e.g., colorectal, pancreatic, bladder, melanoma, breast, and prostate). Additionally, 5 of 25 networks were cancer-associated. Networks 2 ( Figure 2B), and 6 ( Figure 2C) are both associated with cell death and survival, organismal injury and abnormalities (Table S8).
The interspecies cross-validation analysis performed using human PC from The Cancer Genome Atlas (TCGA) database revealed 79 genes altered by genomic losses in both species (>20% of samples) ( Table 1). The synteny map, constructed with CNAs detected in more than 20% of canine PC samples, shows the genomic similarities with human PC (Figure 3).    The interspecies cross-validation analysis performed using human PC from The Cancer Genome Atlas (TCGA) database revealed 79 genes altered by genomic losses in both species (>20% of samples) ( Table 1). The synteny map, constructed with CNAs detected in more than 20% of canine PC samples, shows the genomic similarities with human PC (Figure 3).

Discussion
A comprehensive evaluation of canine cancer has the potential to uncover molecular mechanisms common to the human disease, including tumor progression, metastasis, and diagnostic, and prognostic markers. The presence of a high level of similarities opens avenues to identify combined or alternative therapies to better treat the disease in both species [24].
In a limited number of PIA cases studied, we detected a small number of CNAs, which revealed a low level of genomic instability. Among the seven genes exclusively identified in PIA, two (ADIPOQ, and BCL6) were previously described in human PC [25,26]. More interestingly, at least 11 , and TPI1) of 17 genes affected by CNAs in PIA, and also PC samples were associated with different aspects of human prostate carcinogenesis (searched in the PubMed database using as headlines the gene name and prostate cancer, consulted in February 2019). A previous report using gene expression analysis showed that TIMP downregulation was implicated in the progression of prostatic intraepithelial neoplasia to PC [27]. In dogs, PIA is poorly explored and mostly limited to the immunohistochemistry (IHC) of a few proteins [11,17]. To our knowledge, this is the first study showing CNAs in primary canine PC, and PIA. Our findings point to a set of common genomic alterations in PIA, and PC, suggesting their involvement at the early stages of the disease. In human PIA, and prostatic intraepithelial neoplasia (PIN, a well-established preneoplastic lesion), similar frequencies of 8p22 loss, and 8q24 gain were described [28]. Apparently, several imbalances of chromosome 8 occur during human prostatic carcinogenesis [29]. Our results point out the presence of common chromosomal imbalances in canine PIA, and PC. Considering that PIA occurs at a higher frequency than in PIN [15], dogs could represent an interesting spontaneous model to study prostate cancer progression.
In contrast to PIA, PC presented several genetic alterations, predominately large deletions. The enrichment analysis revealed a large number of genes showing copy number alterations related to carcinogenesis. To our knowledge, only one immortalized canine prostate cancer cell line (Probasco) was characterized by CNAs [23]. Although this cell line presented various CNAs, similar to our cases, the authors found mainly gains, particularly of chromosomes 4,6,8,10,11,14,15,17,18,20,23,24,26,35,37, and 38 [23]. However, no comparison was performed between the primary tumor and the cell line. Although cell lines are excellent models to study several molecular aspects of cancer, it has been intensively discussed that culture conditions, cell growth time in vitro and cell immortalization can modify the DNA, and specific cells could be selected [30,31].
In human PC, chromosomal imbalances are prevalent and significantly more frequent than somatic mutations [32,33]. Deletions affecting tumor suppressor genes, including BRCA1, BRCA2, CDH1, and TP53, are common in human PC [34][35][36]. Similarly, our patients presented losses in several well-known tumor suppressor genes such as ATM, BAX, BRCA1, CDH1, MEN1, and TP53. Four homozygous deletions harboring genes (LOC403555, and TNNT2) were identified in three different PC cases. Among them, LOC403555 has an uncertain function, and TNNT2 is related to contraction of cardiac muscle cells [37].

Discussion
A comprehensive evaluation of canine cancer has the potential to uncover molecular mechanisms common to the human disease, including tumor progression, metastasis, and diagnostic, and prognostic markers. The presence of a high level of similarities opens avenues to identify combined or alternative therapies to better treat the disease in both species [24].
In a limited number of PIA cases studied, we detected a small number of CNAs, which revealed a low level of genomic instability. Among the seven genes exclusively identified in PIA, two (ADIPOQ, and BCL6) were previously described in human PC [25,26]. More interestingly, at least 11 (AK1 silvia.regina.rogatto@rsyd.dk (S.R.R.) silvia.regina.rogatto@rsyd.dk (S.R.R.) silvia.regina.rogatto@rsyd.dk (S.R.R.) AKT2, BRD2, CD4, FAM83H, GNB3, MIR221, MIR338, SCN2B, TIMP1, and TPI1) of 17 genes affected by CNAs in PIA, and also PC samples were associated with different aspects of human prostate carcinogenesis (searched in the PubMed database using as headlines the gene name and prostate cancer, consulted in February 2019). A previous report using gene expression analysis showed that TIMP downregulation was implicated in the progression of prostatic intraepithelial neoplasia to PC [27]. In dogs, PIA is poorly explored and mostly limited to the immunohistochemistry (IHC) of a few proteins [11,17]. To our knowledge, this is the first study showing CNAs in primary canine PC, and PIA. Our findings point to a set of common genomic alterations in PIA, and PC, suggesting their involvement at the early stages of the disease. In human PIA, and prostatic intraepithelial neoplasia (PIN, a well-established preneoplastic lesion), similar frequencies of 8p22 loss, and 8q24 gain were described [28]. Apparently, several imbalances of chromosome 8 occur during human prostatic carcinogenesis [29]. Our results point out the presence of common chromosomal imbalances in canine PIA, and PC. Considering that PIA occurs at a higher frequency than in PIN [15], dogs could represent an interesting spontaneous model to study prostate cancer progression.
In contrast to PIA, PC presented several genetic alterations, predominately large deletions. The enrichment analysis revealed a large number of genes showing copy number alterations related to carcinogenesis. To our knowledge, only one immortalized canine prostate cancer cell line (Probasco) was characterized by CNAs [23]. Although this cell line presented various CNAs, similar to our cases, the authors found mainly gains, particularly of chromosomes 4, 6, 8, 10, 11, 14, 15, 17, 18, 20, 23, 24, 26, 35, 37, and 38 [23]. However, no comparison was performed between the primary tumor and the cell line. Although cell lines are excellent models to study several molecular aspects of cancer, it has been intensively discussed that culture conditions, cell growth time in vitro and cell immortalization can modify the DNA, and specific cells could be selected [30,31].
In human PC, chromosomal imbalances are prevalent and significantly more frequent than somatic mutations [32,33]. Deletions affecting tumor suppressor genes, including BRCA1, BRCA2, CDH1, and TP53, are common in human PC [34][35][36]. Similarly, our patients presented losses in several well-known tumor suppressor genes such as ATM, BAX, BRCA1, CDH1, MEN1, and TP53. Four homozygous deletions harboring genes (LOC403555, and TNNT2) were identified in three different PC cases. Among them, LOC403555 has an uncertain function, and TNNT2 is related to contraction of cardiac muscle cells [37].
Gains and amplifications have been associated with the activation of proto-oncogenes, which promote gain of function and contribute to cancer development or progression. We detected 14 high copy number gains and only 3 covered known genes (QK1, BRD2, and EMR4P). Among them, BRD2 has been described as overexpressed in human castration-resistant PC [38]. Nevertheless, 6 of our 14 canine PCs also had hemizygous deletion of BRD2. In human PC cases from TCGA, CNAs mapped at BRD2 are rare (8.3%; 41/492), being 18 gains, and 23 losses.
Interestingly, the synteny map showed similarities between recurrent losses detected in canine PCs with chromosomal losses described in human PCs, including 2q, 3p, 8p, 12p, 16q, 17p, and 21q.36. An interspecies cross-validation analysis using TCGA database, which currently contains 492 human PC cases with CNA information, revealed additional similarities between human and canine PC. At least 79 genes were altered by gains/losses in more than 20% of our cases and those from the TCGA. The inactivation of the clusterin gene (CLU), deleted in~55% of TCGA cases, was associated with neoplastic transformation and tumor progression, while downregulation of MX1 was involved in proliferation, progression and metastasis of PC cells [39][40][41]. Loss of the ZBTB4 gene, a transcriptional repressor gene that binds to the GC-rich promoter regions of their targets, increases genomic instability (aneuploidy) and promotes tumorigenesis in human neoplasia, including PC [42,43]. In addition, ZBTB4 overexpression is related to an increased survival time for PC-affected patients [44].
To better understand the relationship among the genes covered by CNAs in canine PC, we performed the canonical pathway and network analyses using the IPA software. The cancer-related canonical pathway called the molecular mechanisms of cancer revealed the involvement of genes reported as altered in several human cancers, including TP53, MDM2, ATM, and BRCA1. Additionally, network analysis revealed complex interactions among the genes. In particular, two networks directly related to cancer presented MDM2, ICAM1, and FTH1 as central nodes. The deregulation of both ICAM1, and FTH1 was previously associated with human prostate cancer [45,46]. The well-known MDM2 oncogene is an important negative regulator of TP53, commonly deleted in human PC. Loss of TP53 is involved with the transformation of prostate epithelium, and MDM2 overexpression was associated with PC growth and metastasis [47,48]. Of interest, ZBTB4 interacts both with MDM2, and TP53 through HIPK2, a tumor suppressor involved in human cancer, and deleted in our cases [49,50].
Genomic alterations in TP53, MDM2, and ZBTB4 were investigated at the transcriptional level in an independent cohort of PC patients. We found a significant down-expression of TP53, and ZBTB4 in PC cases compared with normal tissues. In addition, MDM2 was significantly overexpressed in cancer samples. Previously, we reported TP53 down-expression and increased MDM2 proteins, and transcript levels in a different cohort of animals with PC, thus confirming the results presented here [11]. Moreover, our data also suggested that these gene expression alterations might be, at least in part, explained by the presence of CNAs.
In addition to the similarities, we also observed differences between canine, and human PCs. Highly prevalent alterations in human PC, such as gains of the MYC oncogene and deletions of the tumor suppressors NKX3-1, PTEN, RB1, and CDKN1B, were absent or present in only one case [33,34]. While lacking studies of RB1, and CDKN1B in canine PC, expression of MYC, NKX3-1, and PTEN were assessed in our previous studies [11,35]. Using qPCR, IHC or Western blot, we reported MYC overexpression, and NKX3-1, and PTEN down-expression [11,35]. A plausible explanation is the sample bias or mechanisms other than CNAs involved in the MYC activation (translocation or mutation), and PTEN/NKX3-1 inactivation (point mutations or hypermethylation).
An interesting difference between canine and human PC relates to the AR. AR has been described as involved in the development, progression and survival of human PC cells [50,51]. In human PC, AR is an oncogenic driver and its amplification/overexpression has been suggested as a mechanism that leads to the appearance of castration-resistant PC, which hypersensitizes cancerous cells to low levels of androgens [52,53]. In our PC sample set, genomic loss involving the AR gene was observed in three cases and no gains or amplifications were detected. Canine PCs have been characterized by a lack of or low AR protein levels, and the animals are not responsive to castration as a therapeutic option [7,11,54,55]. Previous reports have suggested that the initial transformation of neoplastic canine prostate cells starts from an AR-negative clone, even in intact animals [7,11].
All canine PC AR-negative cases herein evaluated by CNAs presented a Gleason score of 6 and no metastasis. In addition, no information on AR protein expression in TCGA dataset was available, thus limiting a more detailed comparison. Although these factors are limitations of our study, we reported similar genomic alterations in both human and canine PC, addressing a gap in the literature. Further studies are needed to unravel genes and pathways involved in more aggressive canine PC, including in metastatic cases.

Ethics Statement
This study was performed in accordance to the National and International Recommendations for the Care and Use of Animals. All procedures were performed under the approval of the Animal Ethics Committee from Faculty of Veterinary Medicine and Animal Science, UNESP, Botucatu, SP, Brazil (Protocol #10/2007).

Patients
The total number of cases included in this study was 33 PCs, 4 PIAs, and 15 normal prostate tissues. Array-CGH assays were performed in 14 PC, 4 PIA, and 5 normal prostatic tissues. The necropsies were performed within an interval of 1 h from the animal's death, with the consent of the owners. The mixed breed dogs were intact and had a mean age of 9 years. Although they had no clinical features of PC or PIA, lesions were found in the histopathological analysis. The prostatic gland was collected, fixed in 10% formalin, routinely processed, embedded in paraffin and sectioned. The areas of interest were identified and the cells were obtained from macrodissected samples using a 16-gauge needle, as previously described [11]. Tumor areas with more than 85% of epithelial neoplastic cells were selected for DNA extraction ( Figure S1). The PIA cases evaluated in our study were diagnosed during necropsies and no evidence of PC was found in these animals. The histopathological characterizations of PIA, and PC samples were performed according to Palmieri et al. [15], and Lai et al. [56], respectively. All PC cases presented an infiltrative pattern with no measurable mass and the same tumor grade (Gleason score of 6). In this set of cases, no metastasis was found during the necropsies. Furthermore, all PC samples showed negativity for uroplakin III immunoexpression [7].

Chromosomal Imbalances Analysis (aCGH)
Genomic DNA was extracted from FFPE samples using the Qiagen DNeasy Blood & Tissue Kit (Quiagen, Hilden, Germany) according to the manufacturer's recommendations. A pool of genomic DNA obtained from five FFPE canine normal prostate tissues was used as reference.
Copy number alterations were assessed using the Canine Genome CGH Microarray 4 × 44K G2519F (ID-021193, Agilent Technologies, Santa Clara, CA, USA), according to the manufacturer's instructions. Data were extracted with Feature Extraction 10.1.1.1 software (Agilent), and the analyses were performed using Agilent Genomic Workbench Standard Edition 5.0.14 software (Statistical algorithm ADM-2, threshold of 4.0, fuzzy zero correction). Nonredundant CNAs called by at least three consecutives probes and a log2 ratio > 0.3 for gains and <−0.3 for losses were considered as significant. The ideograms showing the CNAs identified in each case were constructed using the PhenoGram online software (http://visualization.ritchielab.psu.edu/phenograms/document).
Synteny blocks between canine and human genomes were obtained through the Synteny Portal [57]. The SynBuilder tool was used to obtain synteny blocks between the canine canFam2 and human hg19 reference genomes, with a resolution of 150 Kbp. Regions overlapping CNAs in more than 20% of PC cases were identified with the CoNVaQ web tool [58]. The Circos plot showing the altered regions and the corresponding regions in humans was generated using the OmicCircos at R/Bioconductor package [59].
Enrichment analysis, based on human data, was performed with Ingenuity Pathway Analysis software (IPA v.2.2.1, Qiagen, http://www.qiagen.com/ingenuity/), considering the genes affected by CNAs in more than 20% of PC cases. In this analysis, the following parameters were used: core analysis, Ingenuity Knowledge Base (genes only) as a reference set, direct and indirect relationships, endogenous chemicals, 35 molecules per network and a maximum of 25 networks. Statistical significance was considered with p < 0.05 calculated using Fisher's exact test. The activation (orange lines in the network) or inhibition (blue lines in the network) was used with z-score × 2.0 or ≤ −2.0, respectively. The CNA information of 492 human PCs from TCGA database, stored in the cBioPortal for Cancer Genomics (http://www.cbioportal.org/, accessed in April 2018), was used to explore the similarities between canine and human PCs.

Gene Expression Analysis
A hematoxylin-eosin tissue section (PC, and normal tissue) slide to guide the tissue dissection was demarcated and isolated using a 40-gauge needle in the FFEP block. Total mRNA was extracted using the RecoverAll™ Total Nucleic Acid Kit (Ambion, Life Technologies, MA, USA), according to the manufacturer's instructions. cDNA synthesis was carried out as previously described [11]. The quantity and quality of RNA were evaluated using a NanoDrop ND-1000 Spectrophotometer v.3.0.1 (Labtrade, Wilmington, NC, USA), and 2100 Bioanalyser RNA 6000 Nano kit (Agilent Technologies, Palo Alto, CA, USA), respectively.
Three genes (TP53, MDM2, and ZBTB4) were selected and evaluated by qPCR based on their importance in carcinogenesis, network analysis and TCGA cross-validation analysis. TP53, and MDM2 are well-known cancer genes involved in the emergence of several human malignancies, including PC. TP53 has been implicated in several cancer-related canonical pathways, and MDM2 is a central node in the molecular network 2. In addition, ZBTB4 indirectly interacts with MDM2, and has also been related to prostate carcinogenesis. Moreover, TP53, and ZBTB4 deletions were cross-validated using TCGA database analysis.
The primer sequences of TP53, MDM2, and reference genes (ACRB, HPRT, and PRS5) were previously published [11,60]. The primer sequences of ZBTB4 was forward: 5 TTGGGATTTTGCCATTTCG 3 , and reverse: 5 GGGCAAGGAGGGCACAA 3 . qPCR for TP53, MDM2, ZBTB4, and the endogenous genes was conducted in a total volume of 10 µL containing Power SYBR Green PCR Master Mix (Applied Biosystems; Foster City, CA, USA), 1 µL of cDNA (1:10), and 0.3 µM of each primer. The reactions were performed in triplicate in 384-well plates using QuantStudio 12K Flex Thermal Cycler equipment (Applied Biosystems, Foster City, CA, USA). The PCR product specificity was determined by dissociation curve for all experiments. Relative gene expression was quantified using the 2 −∆∆CT method [61].

Statistical Analysis
Statistical analysis was performed using GraphPad Prism 5 v.5.0 (GraphPad Software Inc., La Jolla, CA, USA). The Kruskal-Wallis or Mann-Whitney U test was applied to compare TP53, MDM2, and ZBTB4 transcript levels between normal and PC samples. Statistical significance was set at p < 0.05.

Conclusions
To our knowledge, this is the first study to describe the genomic profile in a cohort of canine PIA, and PC samples. Our results revealed that PIA cases presented a small number of CNAs compared with PC. PC samples presented a large number of chromosomal imbalances, mainly deletions, which revealed a set of cancer-related genes involved in canonical pathways or networks previously described in human PC. Deletions of TP53, and ZBTB4, and gains of MDM2 were translated in down-and overexpression, respectively, at the transcriptional level. Furthermore, TCGA cross-validation analysis pointed out similarities between human and canine prostate carcinogenesis. Canine PC is an interesting model to use in comparative prostate cancer research; however, it is important to note the differences, including those reported for the AR gene.  Table S1: List of genomic alterations identified in four PIA cases, Table S2: List of genomic alterations identified in 14 PC cases, Table S3: List of high copy gains and homozygous deleted regions found in the 14 PC samples, Table S4: Recurrent CNAs in canine PCs (present in more than 20% of cases), Table S5: List of genes covered by CNAs in more than 20% of PC cases, Table S6: Cancer-related functions of genes evaluated by IPA software, Table S7: Cancer-related canonical pathways of genes evaluated by IPA software, Table S8: Gene interaction networks generated by IPA software. Acknowledgments: The authors would like to thank Deborah Mundin, and Sara M Silveira for their contribution in the aCGH assays.

Conflicts of Interest:
The authors declare no conflict of interest.