Multiple Genes in Human 20q13 Chromosomal Region Are Involved in an Advanced Prostate Cancer Xenograft¹

Prostate cancer is the most frequent noncutaneous malignancy and the second leading cause of cancer death among males in the Western world. Currently, no curative treatment for prostate cancer is available once the disease has progressed beyond the organ. The genetic events underlying prostate carcinogenesis and progression are by and large obscure, and most likely involve many genes. Classical cytogenetics, which became a powerful tool in the analysis of leukemia and solid tumors, has identified only a few consistent chromosomal changes in prostate cancer. This was partly because of their poor growth in culture (1). Previous molecular cytogenetic studies of prostate tumors using fluorescence in situ hybridization and CGH⁴ demonstrated chromosomal losses (at 1p, 6q, 8p, 10q, 13q, 16q, and 18q) and gains (at 1q, 2p, 7, 8q, 18q, and Xq) reviewed by Nupponen and Visakorpi (2). Putative target genes associated with some of these chromosomal aberrations were additionally identified, including amplifications of the AR (Xq12), MYC (8q24), EIF3S3 (8q23), and Cyclin-D1 (11q13) in a significant fraction of hormone-refractory prostate cancer patients (2, 3). Gene amplification in other cancers has been shown to be associated with adverse outcome, such as in neuroblastoma (MYCN) and breast cancer (HER-2/neu), and may thus have a prognostic value. However, in prostate cancer, the prognostic value of HER-2/neu amplification is still controversial. Additionally, specific mutations were described in patients, in the RNASEL (4) and ELAC2 (5) genes, and in prostate tumors tissues, in tumor suppressor genes such as PTEN, as reviewed by Ittmann (6) and KLF6 (7). Gene-expression studies have identified a number of genes involved in different stages of prostate carcinogenesis, but many of the genes that play a role in prostate cancer formation and progression are still unknown. With the advance in microarray technology, it became possible to define gene-expression profiles characteristic to specific tumor type and stage in many prostate tumors (8, 9, 10, 11, 12). In this study we report novel genetic alterations in a xenograft model of a locally advanced prostate cancer, using a combination of cytogenetic and microarray methods. Additional analysis of these genes in two large microarray expression databases revealed several genes in chromosome 20q13 locus that may play an important role in advanced stages of prostate cancer tumorigenesis.

WISH-PC14, a novel human prostatic adenocarcinoma xenograft, was established recently in SCID mice supplemented by testosterone.⁵ The tumor derived from a channel transurethral resection of the prostate of a late recurrent primary tumor, Gleason score 9 (4 + 5), after definitive radiation therapy. At the time of the channel transurethral resection of the prostate, the patient had metastases and was on hormone therapy. The WISH-PC14 xenograft is androgen dependent and secretes prostate-specific antigen. Fresh and frozen specimens from the prostate xenograft tissues were obtained.

Chromosome analysis was performed on primary short-term cultures derived from the WISH-PC14 xenograft. Short-term culture set-up, chromosome preparations, G band staining, and SKY analysis were done according to standard cytogenetic techniques, as described previously (13).

High molecular-weight genomic DNA was extracted from frozen tissue, using PUREGENE DNA Isolation kit (Gentra Systems, Inc., Minneapolis, MN), according to the manufacturer’s instructions. Total RNA was isolated from frozen tissue using TRI REAGENT (SIGMA, St. Louis, MO), according to the manufacturers instructions. Total RNA from two normal prostate tissues, used as controls, were purchased from Clontech Laboratories Inc. (Palo Alto, CA) and Ambion Inc. (Austin, TX).

Detection of gene amplifications was performed by array-based CGH using the AmpliOnc I microarray and the GenoSensor reader system (Vysis Inc., Downers Grove, IL) according to the manufacturer’s recommendations (Vysis), as described previously (13). Briefly, the AmpliOnc I microarray contains 59 different probes (P1, P1-derived artificial chromosome, or bacterial artificial chromosome clones) corresponding to 57 different genes. The list of the genes is available⁶ containing gene names, cytogenetic loci, and location on the grid.

Complementary DNA microarray analysis of gene expression was performed according to the manufacturer’s instructions. Briefly, biotin-labeled RNA fragments were hybridized to the high-density Human U95Av2 GeneChip Expression Arrays (Affymetrix Inc., Santa Clara, CA), which contain probes for ∼12,000 known genes. The hybridized array was stained with strepavidin phycoerythrin conjugate and scanned by the GeneArray Scanner at an excitation wavelength of 488 nm. The amount of light emitted at 570 nm is proportional to the bound target at each location on the probe array. A custom software (GeneChip Analysis Suite software) was used to analyze the intensity data and calculate a set of absolute metrics. A fold-change of each transcript represented on the probe array was calculated relative to the control. Overexpression was defined as significant change ≥2-fold.

Two published microarray expression databases were analyzed as described below. The MU database includes gene expression profiles from 14 localized prostate tumors, 20 metastatic tumors, and 14 BPH samples that were tested using the newly developed 10K array based on the Research Genetics Unigene microarray (8). The complete raw data, normalized to the median, is available.⁷ The Brigham and Women’s/DFCI database includes Affymetrix U95A array gene expression profiles from 52 locally confined tumors and from 50 paired nontumor samples (12). The complete raw data, normalized to the median, is available.⁸

The raw data for relevant genes was retrieved from each of these datasets, separately. To allocate possible differences in expression levels, the rank-sum tests Mann-Whitney (for comparison between two groups) and Kruskal-Wallis (for comparison between three groups) were applied. These tests are nonparametric and, therefore, do not presuppose any assumptions (14, 15), and use algorithms in which the gene score is treated as an ordinal variable. The entire sample population (tumors and controls) was ranked ordinarily, and the ranking scores for each group were summed up. The higher the difference in ranking scores, the better the differentiation between the two examined populations of samples. A P < 0.05 was considered as statistically significant.

G band and SKY analyses of the WISH-PC14 xenograft revealed a clone of tumor cells that contain multiple chromosomal rearrangements, gains, and losses (Fig. 1,A). This complex karyotype demonstrates hyperploidy with modal number range of 72–75 chromosomes. The cytogenetic rearrangements mainly involved chromosomes 3, 8, 11, 14, 17, 20, 22, and X. According to G banding and SKY analyses, the abnormalities were predominantly of chromosomes 8, 20, and X. It is of note, that all five derivatives of chromosome 20 present in the tumor demonstrated rearrangements within cytobands 20q12–13 (Fig. 1 B).

The normal range of test to reference (G/R) ratios was determined in a normal male control DNA for all of the genes on the array, as described previously (13) with minor modifications. The mean G/R ratio was 1.015 with a SD of 0.119. A value of the mean +2SD was set as cutoff level for normal gene copy number, and the normal range was therefore defined as 0.77–1.26. A significant increase in gene copy number (amplification) was considered as an average G/R ratio ≥1.4 in three independent experiments and corresponded to mean +3SD. A 1.4-fold amplification suggests three copies in a diploid cell. Results of array-based CGH of WISH-PC14 revealed amplification mainly at chromosomes 8q, 20q, and Xq. The AR gene on chromosome Xq11–12 and PTPN1 on 20q13.1–13.2 showed a 3.52 ± 0.64 and 3.25 ± 0.36 fold amplification, respectively, whereas MYC and MOS on 8q24.12–24.13 and 8q11 were amplified by 2.15 ± 0.29 and 1.79 ± 0.18 fold, respectively. Other genes such as N-RAS (1q13.2), PIK3CA (3q26.3), and CSE1L (20q13) showed lesser degree of amplification ranging from 1.40 ± 0.04 to 1.62 ± 0.17 fold (see also Table 1).

Gene expression profile was determined in two WISH-PC14 tumor samples derived from different passages of the xenograft and two normal human prostate tissue controls. Table 1 presents the levels of amplification and expression for genes detected by the array-based CGH analysis. These analyses revealed two groups of genes: The first included genes that were both amplified and overexpressed (MYC, PTPN1, and CSE1L; Table 1), and the second included genes that were amplified but their expression did not change (NRAS and AR). MYC amplification and overexpression in prostate tumors have been reported previously (2, 3, 16). Because both PTPN1 and CSE1L map to 20q13, also known as the HPC20 prostate cancer susceptibility locus (17), we asked whether there are other genes in this locus that are not amplified but are overexpressed. Therefore, we analyzed all of the genes located on chromosome 20q13 region that are presented in the Human U95Av2 GeneChip. This analysis revealed a third group of genes including MYBL2, ZNF217, and STK15 (Table 1; Fig. 2 A). Two additional genes that showed minor amplification in the CGH-arrays, PIK3CA and MOS, are not represented on the Affymetrix U95Av2 array and therefore were not included in this analysis.

The expression patterns of the genes included in the first (amplification and overexpression) and the third (overexpression) groups were additionally analyzed statistically in the MU and the DFCI microarray databases (Table 2). MYC, CSE1L, MYBL2, ZNF217, and STK15 were significantly (P < 0.05) overexpressed in prostate tumors, compared with the control tissues. MYC and CSE1L were significantly overexpressed in both databases, whereas MYBL2 was significantly overexpressed only in the MU database. ZNF217 was not represented at the MU database but was significantly overexpressed in the DFCI database. STK15 was significantly overexpressed in the MU database but its signals were absent in all samples of the DFCI database. Because localized tumors and metastases are listed separately in the MU database, we analyzed these genes to see whether their expression pattern can distinguish between metastases and tumors (Fig. 2 B). The sum rank of MYC and CSE1L clearly distinguishes between localized tumors and metastases from BPH (P < 0.05). However, the sum rank of MYBL2 and STK15 significantly distinguishes the metastases from the localized tumors and BPH. PTPN1 showed no significant differences of expression between tumor and control tissues.

Development and progression of prostate cancer to lethal, hormone-refractory, and metastatic disease involves multiple genetic changes. However, the genetic events associated with different stages of the disease are not fully understood. Here we applied a strategy based first on the analysis of an experimental model, a locally advanced prostate cancer xenograft, using a combination of cytogenetic, array-based CGH and expression analyses. This analysis allowed us to focus on genes, located on chromosome 20q13 region, that are significantly overexpressed in the WISH-PC14 tumor. As the second step, the expression pattern of these genes was confirmed in two large human prostate cancer databases. This may suggests a possible role for these genes in the transition from a local prostate tumor to metastasis.

Only a few consistent chromosomal changes were detected previously in prostate cancer. Loss of chromosome 20 was reported only once (1, 2).⁹ We describe here, for the first time, a locally advanced prostate tumor with rearrangements of all five of the chromosome 20s involving cytoband 20q13. We additionally showed that genes in this locus, PTPN1 and CSE1L, are both amplified and overexpressed in the tumor xenograft. The protein tyrosine phosphatase PTPN1 (PTP1B) involvement in carcinogenesis was suggested in esophageal cancer (18) and in several human breast cancer cell lines (19). However, PTPN1 was not shown to be overexpressed in the two prostate cancer microarray databases (Table 2).

In contrast, CSE1L was significantly overexpressed in both the MU and DFCI tumor samples. CSE1L is the human homologue of the yeast chromosome segregation gene, CSE1, which encodes a nuclear transport factor that plays a role in proliferation and apoptosis (20). CSE1L is a mitotic spindle checkpoint, which assures genomic stability during cell division. It is perturbed frequently in neoplasias of various origins (21, 22, 23). Furthermore, Wellmann et al. (24) demonstrated that CSE1L overexpression in hepatocellular carcinoma is correlated with the grade of tumor dedifferentiation.

We additionally focused on genes in the 20q13 region, and demonstrate for the first time the involvement, amplification, and overexpression of CSE1L in prostate cancer. Other genes in this locus, ZNF217, MYBL2, and STK15 were also overexpressed in the WISH-PC14 xenograft and in multiple prostate tumors reported in the MU and DFCI databases. ZNF217 encodes a putative C2H2 kruppel-like transcription factor and was suggested as one of two driver genes in a complex breast cancer amplicon (25). Its amplifications were also detected in ovarian cancers (26). Additionally, the ability of overexpressed ZNF217 to immortalize human mammary epithelial cells and to overcome breast cells senescence was shown to be associated with breast cancer progression (25, 27). MYBL2 is a transcription factor that plays an essential role during cell cycle progression, and its amplification was detected in various cancers (26, 28, 29, 30, 31). STK15 encodes a centrosome-associated kinase, and its amplification was also detected in several tumors and cell lines (26, 32). STK15 overexpression was significantly associated with chromosomal instability in breast cancers and was therefore suggested as an indicator of poor prognosis and resistance to hormone therapy (33).

In summary, we performed microarray analysis of a prostate cancer xenograft, derived from a locally advanced tumor, and analysis of two large microarray expression databases. Genes in the 20q13 region that are significantly overexpressed in prostate tumor samples were identified. Furthermore, MYBL2 and STK15 overexpression in the xenograft, as well as in prostate metastases directly derived from patients, allows clear distinction between localized tumors and metastases. Our data suggests that these genes may be involved in advanced stages of prostate cancer and may serve as markers for tumor progression.