Decreased Retinoid X Receptor-α Protein Expression in Basal Cells Occurs in the Early Stage of Human Prostate Cancer Development

Prostate cancer is the most prevalent solid tumor and the second leading cause of cancer mortality in males in the United States. For the year 2003, an estimated 220,900 American men will be newly diagnosed and 28,900 American men will die of prostate cancer (1). Vitamin A-related compounds, known as retinoids, have been demonstrated to play important roles in prostate biology and carcinogenesis. Epidemiological studies have revealed an inverse trend between serum vitamin A levels and subsequent incidence of prostate cancer (2, 3). Carcinomatous human prostates contain significantly less of the biologically active retinoic acid than normal prostate (4), which may result from aberrant retinol synthesis (5). In various models for human cancer, retinoids are powerful agents of cell differentiation (6). Retinoids have been especially effective in inhibiting tumor growth and progression in chemically induced mouse prostate cancer models (7–9) and in reducing the growth and tumorigenic potential of human prostate cancer cells lines (10, 11). Additional studies are investigating retinoids as agents of prostate cancer prevention and treatment (12–15).

Most actions of retinoids are thought to be mediated by its nuclear receptors, which function as ligand-activated transcription factors (reviewed in 16–18). There are two types of retinoid receptors, retinoic acid receptors (RARs) and retinoid X receptors (RXRs). The RARs and RXRs each have three subtypes: α, β, and γ. Each subtype has distinct and conserved sequences, a specific pattern of expression during embryonal development, and a different distribution in adult tissue. The subtypes are thought to regulate the expression of specific sets of genes (19, 20). Heterodimers of the RARs and RXRs bind to a specific DNA promoter sequence, termed the retinoic acid response element, and regulate gene transcription (21, 22). RXRs can also form homodimers and activate the retinoid X response element or form heterodimers with other members of the steroid receptor superfamily, thus serving in multiple regulatory functions in different signaling pathways (23).

Among the retinoid receptor subtypes, RXRα appears to have a critical functional role in the development of prostatic preneoplastic lesions. In the transgenic mouse model with the conditional disruption of the RXRα gene in the prostate epithelium, an age-dependent development of multifocal prostatic intraepithelial neoplasias (PIN) occurred, with high-grade PIN developing in some animals beginning at 10 months of age (24). The heterozygous RXRα mutant mice developed PIN, but in a temporally delayed manner, suggesting a dose dependence on RXRα in the epithelium (24). In human prostate cancer cell lines, RXRα protein was decreased relative to noncancer prostate cell lines, and a reduction of cell growth or increased susceptibility to apoptosis was demonstrated with increases in the level of RXRα in RXRα-transduced prostate cancer cells (25).

In light of the evidence of a functional role for RXRα in prostate tumor progression, we undertook an immunohistochemical (IHC) examination of a relatively large number of fresh frozen human prostate specimens and report our analysis of the distribution and semiquantitative expression levels of RXRα protein in noncarcinomatous prostate, PIN, adenocarcinoma (CA), and histopathologically “normal” glands distant and adjacent to CA. Our study provides a detailed characterization of RXRα protein expression in prostate cancer, which may elucidate its role in prostate neoplasia.

Patients were seen at Memorial Sloan-Kettering Cancer Center from April 1, 1994 to June 30, 1997. Those with newly confirmed diagnoses of prostate CA were enrolled consecutively for a hospital-based case-control study conducted under the guidelines of the Memorial Sloan-Kettering Cancer Center Protection of Human Subjects Institutional Review Board and with the written informed consent of the study participants. Prostate specimens (N = 111) were obtained from radical prostatectomy. The study population had a median age of 60 ± 6 years, was predominantly Caucasian, and was not subjected to hormones, drugs, or chemotherapy before surgery. Initial pathology reports, obtained after surgery or on discharge, found 23 very well differentiated (Gleason score 2–5), 78 well to moderately differentiated (Gleason score 6–7), and 10 poorly differentiated (Gleason score 8–10) CAs. Also included in our study were 27 normal noncarcinomatous prostates that were obtained from bladder cancer patients who underwent cystoprostatectomy. These prostates were pathologically confirmed to be free of cancer by H&E staining.

Immediately after surgical removal, prostates were separated into anterior, posterior, right, and left lobes. Between six and eight blocks containing portions of the lobes were embedded in OCT medium (Fisher Scientific, Pittsburgh, PA), freshly frozen, and stored at −70°C. The blocks of frozen tissues were shipped with dry ice to the Molecular Epidemiology Laboratory of the University of California-Los Angeles (UCLA) Jonsson Comprehensive Cancer Center, and the study was approved by the UCLA Institutional Review Board for Human Subjects. One to three blocks from each prostate were selected for histological analyses. Frozen tissues were cut with a cryostat into 7-μm thick sections. The first and last sections for each block of tissue were stained with H&E for histological diagnosis and the adjacent middle sections were used for IHC analysis. Two independent IHC analyses were conducted for each prostate specimen.

A RXRα affinity-purified rabbit polyclonal antibody that specifically recognizes human RXRα (Santa Cruz Biotechnology, Santa Cruz, CA) was used at 1:100 concentration. A Keratin-903 (K903) antibody (Enzo Diagnostics, Inc., Farmingdale, NY) mouse monoclonal antibody that is specific for high molecular weight cytokeratins 1, 5, 10, and 14 was used to identify epithelial basal cells (26). The K903 antibody was used at 1:500 concentration. Manual IHC was performed as follows. The frozen prostate sections were warmed and air-dried at room temperature for 30 min. The slides were immersed in acetone at 4°C for 10 min to fix the sections. Slides were rinsed in PBS, incubated in PBS-hydrogen peroxide (0.3% hydrogen peroxide) for 30 min, and then equilibrated in PBS-Tween (0.25% Tween). Sections were first blocked with 5% normal donkey serum (Sigma Chemical Co., St. Louis, MO) for 45 min at room temperature and then incubated at room temperature with primary antibody diluted in PBS-Tween to the specified concentration for 14–16 h. Secondary antibody [donkey anti-rabbit for RXRα or donkey anti-mouse for cytokeratin K903 (CK903); Sigma Chemical] at a 1:200 concentration was applied to the slides for 90 min at room temperature. Vecastain (Vector Laboratories, Inc., Burlingame, CA) was used as the conjugate and 3,3′-diaminobenzidine tetrahydrochloride (Sigma Chemical) was used as the chromagen for detection. The slides were counterstained with hematoxylin. The specificity of the IHC procedure was confirmed by using negative controls; tissues were incubated in either preimmune serum or by using the RXRα antibody preabsorbed with an excess of purified RXRα antigen (Santa Cruz Biotechnology).

The staining intensity was scored using a scale of 0–3 as follows: 0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining. RXRα staining in secretory epithelial cells was judged as either present or absent.

The frozen prostate sections were fixed and blocked as described above. The slides were simultaneously incubated with RXRα antibody (1:50) and CK903 antibody (1:150) for 14–16 h at room temperature. Slides were rinsed with PBS and then incubated with secondary antibodies consisting of donkey anti-rabbit labeled with FITC for RXRα (Jackson ImmunoResearch Laboratories, Wesgrove, PA) and donkey anti-mouse labeled with rhodamine (Jackson ImmunoResearch Laboratories) for CK903 for 3 h at room temperature. The signal was observed with confocal microscopy (Olympus AX 70, Melville, NY) and photographed using Kodak (Rochester, NY)Elite Chrome 100 film.

t test analyses were used to compare the mean expression levels of RXRα among the histologically distinct tissues and to compare mean expression levels of RXRα in tissues with Gleason scores 2–5 to tissues with higher Gleason scores. Two-sided t tests were used and P values were reported.

All of the 111 specimens from prostate cancer patients contained histologically benign (“normal”) glands. In addition, there were 27 PIN and 26 CA. Twenty-seven prostates from bladder cancer patients who underwent radical cystoprostatectomy were confirmed to be free of cancer by H&E and used as normal/normal in our IHC analysis of RXRα expression. In all of the prostate specimens, RXRα protein was detectable by IHC in stromal cells and included endothelial cells, fibroblasts, and vascular smooth muscles. Positive signal for RXRα in these stromal cells was uniformly strong, localized to the nucleus, and served as internal positive controls for RXRα IHC. No staining was observed in the negative controls, which included incubation of tissues in preimmune serum or preabsorption of RXRα antibody with an excess of purified RXRα antigen.

Representative images of H&E and IHC staining for RXRα are presented in Fig. 1, A and B, respectively. RXRα signal was uniformly strong in basal cells and weak in secretory epithelial cells. The intensity was much stronger in the nucleus versus in the cytoplasm. The expression of RXRα in basal cells was confirmed by colocalization of the RXRα signal with the basal cell marker K903 by double-label immunofluorescence under confocal microscopy as shown in Fig. 1, C and D. Table 1 shows the relative RXRα protein expression levels in basal cells of the histologically distinct areas (normal/normal, “normal” tissue distant or adjacent to CA, and PIN). Noncarcinomatous prostates had strong positive RXRα signal in the basal cells (expression level 3.0 ± 0). Weak expression of RXRα signal in the secretory epithelial cells as clearly shown in Fig. 1D was found in all (27 of 27) of the noncarcinomatous specimens, which is graphically presented in Fig. 2.

In the carcinomatous prostate specimens, “normal” glands that are distant (at least 1 cm away) from CA were identified by H&E (Fig. 3A). The presence of the basal cell marker K903 as shown in Fig. 3B confirmed the non-neoplastic nature of the glands. Figure 3C shows a representative IHC of RXRα that was positive in all (111 of 111) of the specimens. Similar to noncarcinomatous normal/normal prostates, RXRα protein signal was more intense in the basal cells than in the secretory epithelial cells. However, the mean expression level of basal cells in the distant “normal” tissue was 2.13 ± 0.44 as shown in Table 1 and was notably less than in normal/normal prostate (P < 0.001). Weak RXRα expression was noted in the secretory epithelium in 100% (111 of 111) of the “normal” cancerous specimens (Fig. 2).

Prostate glands adjacent (<1 cm) to CA were examined by H&E as shown in Fig. 3D and had a normal appearance in architecture with strong K903 expression (Fig. 3E). However, much weaker RXRα expression in the basal cells was noted as compared with the basal cells of the noncarcinomatous and “normal” distant glands (Fig. 3F). The mean RXRα expression level of the adjacent tissue was 1.25 ± 0.53, which is significantly lower (P < 0.001) than either normal/normal or “normal” distant tissue (Table 1). One of the CA adjacent specimens lacked visible secretory cell RXRα signal (Fig. 2). Thus, histologically “normal” adjacent tissues display reduced RXRα protein, which may indicate molecular field changes in the neoplastic process.

Focal regions of PIN were identified by H&E as shown in Fig. 4A. Figure 4B shows K903 signal that was positive in all of the PIN specimens, indicating the presence of basal cells. However, RXRα signal was absent in the basal cells in 52% (14 of 27) of the PIN specimens (Fig. 4C; Table 1). Double-label immunofluorescence under confocal microscopy clearly demonstrates K903 expression in basal cells as shown in Fig. 4D compared with the absence of RXRα in the basal cells shown in Fig. 4E. Among the PIN specimens that did have RXRα signal, most were positive in <20% of the basal cells. The mean basal cell expression of RXRα was significantly lower in PIN (0.56 ± 0.58) as compared with noncarcinomatous tissue or “normal” tissue from carcinomatous prostates (P < 0.001; Table 1). Weak RXRα expression in secretory epithelium was present in 48% of PIN samples (Fig. 2).

The CA lacks basal cells, which was observed by H&E in Fig. 4A as shown side-by-side with PIN. Figure 4B shows negative K903 staining in CA but positive K903 in basal cells of the PIN lesion. Figure 4C shows negative RXRα signal in both CA and PIN lesions. Signal for both RXRα and K903 was typically negative in CA as shown by double-label immunofluorescence under confocal microscopy (Fig. 4, D and E). However, 2 of 26 CA specimens (8%) showed weak expression of RXRα protein in the tumor cells (Fig. 2).

The mean signal levels of RXRα in the basal cells distant from CA were determined for specimens categorized into three groups of Gleason scores: 2–5, 6–7, and >7 as shown in Table 2. Tissues with Gleason scores of 2–5 corresponding to well-differentiated tumor had an average RXRα expression level of 2.35 ± 0.34 in distant “normal” glands. A lower RXRα expression level, 2.10 ± 0.44, was found in distant “normal” glands with moderately differentiated tumors with Gleason scores of 6 or 7. Poorly differentiated glands corresponding to Gleason scores of 8 or 9 had a significantly lower expression (1.83 ± 0.36) of RXRα in distant “normal” glands (P = 0.04) as compared with Gleason scores 2–5 tumor.

This study involved a systematic analysis of RXRα protein expression in human prostate carcinogenesis. Our major finding is that RXRα protein expression decreases in the basal cells in the field as neoplasia progresses from noncancerous normal (normal/normal) to “normal” cells in cancerous prostate and then to PIN. While RXRα signal was weakly present in non-neoplastic secretory epithelial cells, the expression is lost in epithelial cells of most PIN and CA areas. Furthermore, the level of RXRα protein varies among what was pathologically considered to be normal tissue (i.e., RXRα signal was significantly different among normal/normal, “normal” adjacent to CA, and “normal” distant to CA tissues). Moreover, we found a correlation between the decrease in RXRα protein expression in distant “normal” tissue from carcinomatous prostates and increasing Gleason score. That “normal” glands undergo a reduction in RXRα protein preceding histopathological changes indicates that control of RXRα may be an early event that extends to a field of affected tissue. These findings support the hypothesis that RXRα protein loss begins at an early stage of carcinogenesis and may represent a field disease marker for prostate malignancy.

Descriptions of the expression of RXRα in human prostate have been limited to a few published studies. A highly heterogeneous pattern of RXRα protein expression, with some areas of low or no staining, was reported in 13 CA specimens by Zhong et al. (25). Alfaro et al. (27) and Kikugawa et al. (28) have detected RXRα protein in samples from normal and carcinomatous prostates. However, artifactual antibody staining due to formalin fixation may have occurred in some studies because RXRα positivity in inflammatory cells was observed as well (29). Our study not only compared the expression in normal prostates and cancer but also carefully examined the expression in the distant and adjacent fields including PIN. We found that RXRα protein expression was absent in basal and luminal secretory epithelial cells in 52% of PIN specimens, a finding highly consistent with the results obtained using prostate-specific RXRα null mouse model (24). In addition, a careful examination of expression of RXRα in basal versus epithelial cells was performed by side-by-side IHC labeling with K903 (29), a basal cell marker, as well as double-immunofluorescence labeling of K903 and RXRα followed by confocal microscopic examination.

Our analysis showed that the basal cell expression of RXRα in PIN was significantly reduced compared with expression in basal cells of normal glands. That the loss of RXRα signal in PIN was a result of reduced RXRα expression in basal cells and not due to loss of basal cells was confirmed by positive K903 signal in PIN. The reduced RXRα signal in PIN basal cells supports the hypothesis that PIN is a precursor to prostate cancer and that loss of RXRα protein may indicate a premalignant alteration to basal cells along the pathway to neoplasia. It is generally accepted that the basal cells act as reserve cells for the production of secretory cells (30). Prostate CA, by definition, should not contain basal cells (31). Because most expression of RXRα appeared to be in the basal cells and only weakly in the epithelial cells, one should be cautious in the interpretation of findings in the cancer area. Although we also observed a decreased expression of RXRα in secretory epithelial cells from normal to PIN to cancer and 92% (24 of 26) of the cancer areas showed a total absence of the staining, a more sensitive and quantitative type of analysis will be needed to confirm the observation.

In recent years, several genes in the retinoid signaling pathway have been identified as critical factors during prostate cancer initiation and progression (32). The mRNA expression levels of RARβ and RXRβ have been shown to be significantly less in patients with prostate cancer compared with normal patients, and notably, RARβ mRNA expression was lost in adjacent, benign prostatic glands (33). There is strong evidence that the critical RXR in prostate biology appears to be RXRα. Mice lacking both RXRβ and RXRγ are normal in terms of prostate morphology (34), whereas prostate epithelium in mice with the conditional disruption of the RXRα gene displays increased proliferation and induction of PIN, which is considered to be a preneoplastic lesion of CA (24). The underlying mechanism responsible for decreased RXRα in prostate epithelium is not clear and is not the subject of this study. Aberrant promoter methylation has been identified as an epigenetic mechanism for transcriptional silencing (35); however, Lotan et al. (33) observed RXRα transcript signal in prostate cancer. Although RXRα transcript is present in prostate cancer, the levels of RXRα protein may be regulated. For example, RXRα protein is degraded by ubiquitin-mediated proteolysis that occurs on binding of RXR agonist ligands or when heterodimeric partners (i.e., RAR) become activated (36). Decreased RXRα would result in deficient RAR/RXRα heterodimers and RXR/RXRα homodimers. Thus, in the model in which RXR functions as a transcriptionally active partner (37), this could result in functional cellular retinoid deficiency. Moreover, to mediate multiple signaling pathways in the prostate, RXRα may partner with other nuclear receptors, such as the peroxisomal proliferator-activated receptor-γ and vitamin D receptor, the ligands of which have been shown to inhibit prostatic cell growth (17, 38).

The importance of the RXRs is underscored by the effectiveness of RXR-specific ligands to suppress tumorigenesis. RXR-specific ligands have been considered as promising agents for the prevention of prostate cancer and are generally less toxic than RAR-selective retinoids (39). The RXR-selective retinoid, SR11246, was able to inhibit the clonal growth of prostate cancer cells (11). In the LNCaP human prostate cancer cell line, 13-cis-retinoic acid inhibited the tumorigenic potential and expression of prostate-specific antigen, and there was also a significant increase in the expression of RXRα mRNA after 13-cis-retinoic acid treatment as compared with untreated cells (10). To assess the potential for retinoids as a therapeutic modality, the RAR/RXR content of the tumors is important, and specific determination of the relative levels of RXR is a prerequisite for the use of RXR-selective ligands. IHC using biopsy specimens from tumors and preneoplastic lesions is a simple and reliable procedure to identify patients with RXRα as well as those patients who are most likely to benefit from retinoid treatments.

In conclusion, our results show that suppressed RXRα protein expression in basal cells occurs early in prostate carcinogenesis and that RXRα undergoes a statistically significant decline during the progression from normal tissue to PIN and on to primary CA, supporting its role as a field disease marker. Further investigation into its potential as an intermediate end point marker is warranted. If the loss of RXRα protein is proven to be a premalignant feature, it may be an important marker for use in retinoid-based chemoprevention trials to evaluate core biopsy material to determine who should take retinoid therapies. Alternatively, it may be potentially informative to evaluate changes in receptor protein expression in conjunction with clinical outcomes in those who take retinoid-based therapies but fail to show response to retinoids.

We thank Sharon Sampogna for assistance with immunohistochemistry and Dr. Angelica Olcott for critical review of the manuscript.