Genetic and Epigenetic SLC18A2 Silencing in Prostate Cancer Is an Independent Adverse Predictor of Biochemical Recurrence after Radical Prostatectomy

Purpose: This study investigates SLC18A2 (vesicular monoamine transporter 2) expression in prostate adenocarcinoma and examines its potential as a predictive marker for prostate cancer patient outcome after radical prostatectomy.

Experimental Design: Expression and single nucleotide polymorphism microarray analyses identified SLC18A2 as both down-regulated and subject to common loss-of-heterozygosity in prostate cancer. Down-regulated SLC18A2 expression was validated on tissue microarrays containing benign and malignant prostate specimens from an independent patient group (n = 738). Furthermore, SLC18A2 immunoreactivity in radical prostatectomy tumor specimens (n = 506) was correlated to clinicopathologic characteristics and recurrence-free survival. The possibility of SLC18A2 silencing by aberrant DNA methylation in prostate cancer cells was investigated by bisulfite sequencing.

Results: Tissue microarray analysis revealed markedly lower cytoplasmic SLC18A2 staining in cancer compared with nonmalignant prostate tissue samples, confirming RNA expression profiling results. Furthermore, multivariate analysis identified cytoplasmic SLC18A2 immunoreactivity as a novel predictor of biochemical recurrence following prostatectomy (hazard ratio, 0.485; 95% confidence interval, 0.333-0.709; P < 0.001) independent of prostate-specific antigen, Gleason score, tumor stage, and surgical margin status. SLC18A2 showed loss-of-heterozygosity in 23% of the tumors and was densely hypermethylated in 15 of 17 (88%) prostate cancer samples plus 6 of 6 prostate cancer cell lines. In contrast, SLC18A2 was unmethylated in 4 of 4 adjacent nonmalignant prostate and 3 of 5 benign prostatic hyperplasia tissue samples, whereas 2 of 5 benign prostatic hyperplasia samples had monoallelic hypermethylation. Methylation and histone deacetylase inhibitory agents rescued SLC18A2 expression in three prostate cancer cell lines.

Conclusions:SLC18A2 silencing by DNA hypermethylation and/or allelic loss is a frequent event in prostate cancer and a novel independent predictor of biochemical recurrence after prostatectomy.

Prostate adenocarcinoma is a major cause of cancer morbidity and mortality in the western world. It is typically diagnosed based on increased serum prostate-specific antigen (PSA) combined with histopathologic inspection of needle biopsies covering <1% of the prostate volume. This method, however, is associated with significant false-negative rates and does not distinguish well between clinically indolent and aggressive tumors (1). The use of PSA testing for prostate cancer detection has increased the incidence of diagnosis as well as shifted detection to earlier and theoretically curable stages. Indicators currently used for outcome prediction following intended curative radical prostatectomy of primary prostate cancer are PSA, Gleason score, pathologic stage, and surgical margin status (2), but additional markers are needed to improve stratification of low-, medium-, and high-risk patients.

Prostate cancer development and progression is characterized by the accumulation of genetic and epigenic alterations. Epigenetic changes seem to generally occur at earlier stages of carcinogenesis and may be more common and consistent (3). Accordingly, mapping of epigenetic alterations could be particularly valuable for biomarker discovery. Several types of epigenetic changes have been reported for prostate cancer including DNA hypomethylation, loss of imprinting, and altered histone modification patterns. The best described epigenetic alteration in prostate carcinogenesis, however, is DNA hypermethylation of specific CpG islands located near gene promoters as reported for numerous genes including GSTP1, APC, and RASSF1A (for recent reviews, see refs. 1, 4). CpG island hypermethylation has been closely linked to gene silencing (5).

SLC18A2 encodes vesicular monoamine transporter 2, an integral membrane protein of secretory vesicles with predominant expression in neurons, neuroendocrine cells, and amine-handling hematopoietic cell types (6). It transports monoamines (dopamine, serotonin, epinephrine, norepinephrine, and histamine) from the cytosol into vesicles for storage and/or exocytotic release, for example, during neurotransmission or autocrine/paracrine factor release (6). SLC18A2 is also expressed in certain endocrine tumors, including neuroendocrine prostate tumors (7), but rarely in nonendocrine cancers of the same tissues (refs. 8, 9 and this study). Although murine knockout models have disclosed important biological roles of SLC18A2 in the nervous system (10–12), its possible function in normal and malignant prostate biology remains unknown. However, several of the monoamines that are substrates for SLC18A2-mediated transport have been shown to influence growth (13–15), proliferation (16), migration (17, 18), or apoptosis (19) of prostate cancer cells in vitro and in vivo.

In this study, SLC18A2 was identified as a new common target gene for CpG island hypermethylation and loss-of-heterozygosity (LOH) in prostate cancer. By bisulfite sequencing, SLC18A2 was found to be hypermethylated in ∼90% of prostate cancers, indicating that it is one of the most frequently methylated genes in this malignancy. Using prostate cancer tissue microarrays representing >700 patients from Denmark and Switzerland, SLC18A2 was significantly down-regulated in adenocarcinoma compared with nonmalignant prostate tissues, thereby confirming for the first time, at the SLC18A2 protein level, microarray expression profiling data from our group (20) and several others (21–25). Moreover, loss of cytoplasmic SLC18A2 immunoreactivity was identified as a novel adverse predictor of biochemical recurrence after radical prostatectomy, which was independent of the well-established predictive factors preoperative PSA, Gleason score, tumor stage, and surgical margin status.

Prostate cancer tissue microarray. Tissue microarrays contained 738 formalin-fixed, paraffin-embedded prostate tissues (from Institute of Surgical Pathology, University of Zurich, and Institute of Pathology, Aarhus University Hospital), including 506 consecutive radical prostatectomy specimens, 41 hormone-refractory prostate cancer samples, 36 lymph node metastases, 28 distant metastases (bone, lung, and urinary bladder), 65 benign prostatic hyperplasias (BPH), 19 nonmalignant prostate tissues from patients with prostate cancer, 15 high-grade prostatic intraepithelial neoplasias, and 28 primary tumor specimens from patients with metastatic prostate cancer. Before tissue microarray construction, H&E-stained slides of all specimens were evaluated by experienced pathologists (P.J.W., H.M., and P.O.) to identify representative areas, and tumor stage and Gleason score were reassigned according to International Union Against Cancer and WHO/International Society of Urological Pathology criteria (26). Clinical follow-up data were available for 464 (91.7%) prostatectomy patients with a median follow-up period of 72 months (range, 1-167 months). The local scientific ethics committees in both countries approved the study. Tissue microarray construction has been described previously in details (27); only blocks 1 and 3 of the published Danish tissue microarray (28) were used.

Immunohistochemistry. TEG buffer was used for epitope demasking. Tissue microarrays were stained with a polyclonal SLC18A2 (AB1767; Chemicon) antibody diluted 1:300 in TBS with 1% bovine serum albumin. The anti-rabbit EnVision+ System (DakoCytomation) with horseradish peroxidase-labeled polymer and DAB solution (Kem-En-Tec) was used for secondary staining. For negative controls, primary antibody was omitted. Blinded scoring for cytoplasmic and nuclear SLC18A2 immunoreactivity (0, no; 1+, weak; 2+, moderate; 3+, strong) was done by a trained pathologist (B.P.U.) and a scientist (N.T.) with extensive experience in prostate histology. In cases of disagreement, cores were reevaluated to obtain consensus. κ statistics showed good interobserver agreement (0.73 for cytoplasmic and 0.9 for nuclear SLC18A2 staining). Lost specimens and cores without epithelial cell content were excluded from analysis. AB1767 antibody specificity was validated by staining of a human multitissue array (T8235713-5; BioChain Institute), which showed the expected SLC18A2 expression patterns (data not shown).

Western blotting. Protein extracts from cultured cells and fresh-frozen BPH specimens prepared in RIPA lysis buffer were analyzed on 12% NuPAGE polyacrylamide gels (Invitrogen), blotted to polyvinylidene fluoride membranes (Immobilon-P Transfer Membrane; Millipore), and blocked with 5% skimmed milk in PBS with 0.1% Tween 20. Primary antibody (AB1767) was diluted 1:500 and secondary antibody (horseradish peroxidase-conjugated swine anti-rabbit immunoglobulin; DakoCytomation) was diluted 1:5,000 in PBS plus 0.5% Tween 20. For visualization, we used the ECL Plus WB Detection System (Amersham Biosciences). β-Actin was used as internal control (29). Lysate from COS7 cells transfected with an expression vector for human SLC18A2 (gift from Dr. Arnold Ruoho, University of Wisconsin) was used as positive control. This vector encodes a glycosylation mutant of SLC18A2 with HA and Flag/His tags (30). As negative control, COS7 cells were mock transfected with empty vector (pcDNA3.1-Flag/His; Invitrogen).

Cell culture and epigenetic drug treatment. All cell lines were grown in RPMI 1640 with l-glutamine (Life Technologies/Invitrogen) supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin. LNCaP, PC3, DU145, and 22rv1 cells were obtained from the American Type Culture Collection, BPH-1 cells from German Collection of Microorganisms and Cell Cultures, and PNT1A cells from European Collection of Cell Cultures. PSK-1 cells were kindly provided by Dr. Adrie van Boekhoven (University of Colorado), VCaP and DuCaP cells by Dr. Kenneth Pienta (University of Michigan), and H69 and H69A cells by Dr. Marie Stampe Ostenfeld (Danish Cancer Society). Cell lines were treated with 1 μmol/L 5-aza-2′-deoxycytidine (5-aza; Sigma) for 48 h and allowed 5 days of recovery in complete medium. Four cell lines were given a combination of 1 μmol/L 5-aza (48 h treatment + 5 days of recovery) and 1 mmol/L 4-phenylbutyric acid (Sigma; continuously for 7 days). Mock-treated cells were given buffer only. Growth medium was changed daily and cells were harvested on day 7. All experiments were done in duplicate and repeated two to three times.

RNA preparation, cDNA synthesis, and quantitative reverse transcription-PCR. Total RNA from cultured cells and prostate tissue samples was isolated with the RNeasy MinElute Cleanup Kit (Qiagen), as described elsewhere (20). First-strand cDNA synthesis was done with SuperScript II Reverse Transcriptase (Invitrogen) using oligo(dT) priming. SLC18A2 expression was measured with TaqMan Gene Expression Assay Hs00161858_m1 and TaqMan Universal PCR Master Mix on a real-time ABI PRISM 7500 Sequence Detection System (all from Applied Biosystems). For normalization, UBC expression was measured using primers 5′-GATTTGGGTCGCGGTTCTT-3′ plus 5′-TGCCTTGACATTCTCGATGGT-3′ and SYBR Green PCR Master Mix (Applied Biosystems). All reactions were run in triplicates.

Bisulfite sequencing. Genomic DNA from prostate cell lines and carefully selected 20 μm sections of fresh-frozen Tissue-Tek-embedded BPH, adenocarcinoma, and adjacent nonmalignant prostate tissue samples was isolated using the Puregene DNA Purification Kit (Gentra Systems) with proteinase K treatment (100 units, 30 min, 37C) as described previously (29). Laser microdissection was done as described in ref. 31. DNA was bisulfite converted using the MethylEasy DNA Bisulfite Modification Kit (Human Molecular Signaling). The SLC18A2 promoter CpG island was PCR amplified with TEMPase Hot Start DNA Polymerase (Ampliqon) using primers 5′-TTTTAGGTTTGGGTTTTTAAGGTATT-3′ and 5′-AACTCTAAAAACCTCCCTACCTCCCTAC-3′. Gel-purified (illustra GFX PCR DNA and Gel Band Purification Kit; GE Healthcare) amplicons were subcloned into the pCR4-TOPO vector (Invitrogen) and several clones were sequenced using M13 forward and reverse primers. Clinicopathologic data from patients undergoing methylation analysis are compiled in Supplementary Table S2.

Statistical analyses of tissue microarray data. SPSS version 16.0 (SPSS) was used for statistical analyses. P values < 0.05 were considered significant. Contingency table analysis, two-sided Fisher's exact tests, and χ² tests were used to study statistical associations between clinicopathologic and immunohistochemical data. Time to PSA recurrence (cutoff ≥ 0.1 ng/mL) and time to death were selected as endpoints. For survival analysis, only patients with primary prostate cancer undergoing radical prostatectomy were included (clinicopathologic data in Supplementary Table S3). Recurrence-free survival (RFS) and overall survival (OS) curves were calculated by the Kaplan-Meier method and evaluated by two-sided log-rank statistics. For RFS analysis, patients were censored at the time of their last tumor-free clinical follow-up visit; patients not reaching PSA nadir (<0.1 ng/ml) postoperatively were censored at PSA recurrence. For OS analysis, patients were censored at the time of their last clinical follow-up visit. A stepwise multivariable Cox regression model was adjusted, testing the independent prognostic relevance of SLC18A2 immunoreactivity. The limit for reverse selection was P = 0.01. The proportionality assumption for all variables was assessed with log-negative-log survival distribution functions.

Candidate gene selection. To identify novel candidate genes down-regulated in prostate cancer, we reanalyzed microarray expression profiling (20) and single nucleotide polymorphism (SNP) array (31) data sets generated earlier in our group. Supplementary Table S1 lists the top 20 most significantly down-regulated genes in prostate cancer versus adjacent nonmalignant prostate tissue samples based on Affymetrix exon array analysis. Most of these genes have also been found down-regulated in prostate cancer by traditional 3′ array analysis, confirming the usefulness of exon arrays for transcript-level expression analysis. Two novel down-regulated genes were identified: DCHS2 and VIT. Three of the 20 genes were located at common (>20%) LOH regions mapped by Affymetrix 50K SNP array analysis: CTSB at 8p22 (50% LOH), ALOX15B at 17p13.1 (23% LOH), and SLC18A2 at 10q25 (23% LOH), indicating that these genes are selectively lost in prostate cancer cells. Whereas possible roles in prostate cancer have been reported for CTSB (32) and ALOX15B (33), this remains unexamined for SLC18A2, which we selected for further investigation. SLC18A2 expression levels determined by exon array analysis were successfully validated by quantitative reverse transcription-PCR (Fig. 1A). Our findings corroborate expression profiling results for SLC18A2 in prostate cancer from several other studies using distinct microarray platforms (21–25).

Tissue microarray analysis. SLC18A2 protein expression in nonmalignant and prostate cancer tissue samples was investigated by immunohistochemical analysis of two tissue microarrays containing 738 specimens from Danish and Swiss patients with benign or malignant prostatic disease. The specificity of the commercial SLC18A2 antibody has been thoroughly validated by several users (7–9) and was further tested here by Western blotting analysis. Three immunoreactive bands of the expected sizes, corresponding to glycosylated, partially glycosylated, and unglycosylated SLC18A2 protein (34), were detected in BPH tissue samples (Fig. 1B), which by real-time reverse transcription-PCR analysis expressed SLC18A2 at high levels (data not shown). The antibody also detected ectopically expressed SLC18A2 protein in transfected COS7 cells (Fig. 1B). No other significant bands were detected in any of the samples, indicating high SLC18A2 antibody specificity. After staining of the tissue microarrays, 664 of 738 (90%) cores could be evaluated for cytoplasmic and nuclear SLC18A2 immunoreactivity, respectively (staining patterns are summarized in Fig. 1C and D), using a scale from 0 to 3 (score 0, negative; score 1+, weak; score 2+, moderate; score 3+, strong staining; Fig. 2A-D). The remaining cores (10%) were either lost or excluded from analysis due to poor technical quality or high stromal cell content.

Moderate/strong cytoplasmic SLC18A2 staining of secretory epithelial cells was observed in most (≥80%) BPH (Fig. 2E) and adjacent nonmalignant samples (Fig. 1C), whereas most (>90%) cancer samples showed no/weak cytoplasmic staining (Fig. 2F and G; P < 0.001, χ² test). Thus, SLC18A2 is down-regulated at the protein level in prostate cancer, confirming RNA expression profiling results. Cytoplasmic immunoreactivity was oriented toward the glandular lumen (Fig. 2H), consistent with SLC18A2 localization in secretory vesicle membranes with its immunoreactive peptide extruding into the cytoplasm (35). Cytoplasmic SLC18A2 expression was higher in high-grade prostatic intraepithelial neoplasia lesions (Figs. 1C and 2I) than in prostate cancer samples. However, 5 of 12 high-grade prostatic intraepithelial neoplasia samples showed no/weak cytoplasmic staining (Fig. 2J and K), suggesting that SLC18A2 loss in some cases occur already in precancerous lesions. No differences in cytoplasmic SLC18A2 staining patterns were observed in localized versus metastatic prostate cancer, hormone-refractory prostate cancer, or metastases (Fig. 1C), consistent with SLC18A2 loss being a relatively early event. Cytoplasmic SLC18A2 staining was also detected in prostate neuroendocrine cells (Fig. 2L) known to secrete serotonine (15).

Nuclear SLC18A2 immunoreactivity was observed in some specimens, most predominantly in metastatic disease types (Figs. 1D and 2M and N; P < 0.001, χ² test). Aberrant protein glycosylation, as shown for other proteins in prostate cancer (36), has been associated with changes in intracellular SLC18A2 trafficking (37). The nuclear staining is therefore likely to represent mislocated SLC18A2 protein.

SLC18A2 is hypermethylated in prostate cancer. Genomic bisulfite sequencing of a CpG island at the 5′ end of SLC18A2 (Fig. 3A) was used to investigate if SLC18A2 silencing in prostate cancer cells involves DNA hypermethylation. Bisulfite sequencing revealed dense (52-88%) hypermethylation of SLC18A2 in 5 of 5 prostate adenocarcinoma cell lines and in PSK-1 prostate small cell carcinoma cells as well as low-density (17%) methylation in BPH-1 cells. Essentially no methylation was seen in H69 small cell lung carcinoma cells (Fig. 3B), one of only a few cell lines known to express SLC18A2 (9). Real-time PCR analysis showed ≥5-fold lower SLC18A2 expression in 9 of 9 prostate cell lines compared with H69 (Fig. 3C), indicating that SLC18A2 silencing is mediated by CpG island hypermethylation in cultured prostate cancer cell lines.

In clinical samples, SLC18A2 was virtually unmethylated in 4 of 4 adjacent nonmalignant prostate glands and in 3 of 5 BPH samples, whereas the remaining 2 of 5 BPHs had dense monoallelic hypermethylation (Fig. 4A). Dense SLC18A2 hypermethylation was detected in 15 of 17 (88%) prostate cancer samples, whereas 2 of 17 (12%) cancer samples were moderately methylated (PC-21 and PC-45; Fig. 4B). Immunohistochemistry data available for 10 of the cancer samples showed no/weak cytoplasmic SLC18A2 staining in 9 cases, which also had SLC18A2 densely hypermethylated, whereas one tumor (PC-21) showed marked cytoplasmic staining and low-density methylation (Supplementary Table S2; Fig. 4B). This strongly indicates that SLC18A2 is epigenetically silenced in prostate cancer by frequent CpG island hypermethylation. Intriguingly, two tumors (PC-23 and PC-510) displayed dense SLC18A2 hypermethylation and marked nuclear SLC18A2 staining. This inconsistency, however, could be due to tumor tissue heterogeneity. The other tumor with low-density methylation (PC-45) was not analyzed by immunohistochemistry.

Scarcity of unmethylated clones in most tumors indicated that hypermethylation generally affected both alleles of SLC18A2 as confirmed for three heterozygous tumors without LOH (PC-01, PC-09, and PC-20; Fig. 4B). Biallelic hypermethylation was also seen in a crudely dissected tumor with LOH (PC-41), suggesting that SLC18A2 hypermethylation preceded allelic loss in this case. Furthermore, laser microdissection of eight tumors proved that SLC18A2 hypermethylation was present in prostate adenocarcinoma cells and not caused by contaminating inflammatory or stromal cells.

Induction of SLC18A2 expression by drug treatment. Treatment with DNA methylation inhibitor 5-aza induced SLC18A2 expression in PSK-1 and H69 cells (Fig. 3C). This was accompanied by CpG island demethylation in PSK-1 cells from 69% to 11% but most probably indirect in H69 cells (Supplementary Fig. S1). Such indirect induction was not observed in the derived adherent line H69A, where SLC18A2 was unmethylated despite low expression (Fig. 3B and C). In small cell lung carcinoma cells, SLC18A2 expression therefore seems to be regulated by other mechanisms than DNA methylation of its promoter associated CpG island. Treatment of four prostate cell lines with 5-aza plus histone deacetylase inhibitor 4-phenylbutyric acid induced SLC18A2 expression markedly in DU145 and PC3 adenocarcinoma cells and less strongly in BPH-1 cells (Fig. 3C). SLC18A2 induction was associated with CpG island demethylation (52% to 9%) in PC3 cells, whereas methylation intensities remained unchanged in BPH-1 cells (Supplementary Fig. S1). Low-density DNA methylation in BPH-1 cells therefore appeared to be permissive for moderate induction of SLC18A2 possibly through histone acetylation changes caused by 4-phenylbutyric acid. The combined drug treatment scheme did not effectively demethylate (Supplementary Fig. S1) or induce expression of SLC18A2 in LNCaP cells (Fig. 3C).

Univariate and multivariate survival analyses. To investigate the predictive value of SLC18A2 immunoreactivity for patient outcome, we performed PSA-based RFS and OS analyses for 506 patients with primary prostate cancer undergoing radical prostatectomy (158 from Denmark and 348 from Switzerland represented on the tissue microarrays; see Supplementary Table S3 for clinicopathologic data). By univariate analysis, short RFS times were significantly associated with the established indicators high PSA, high Gleason score, high T stage, positive nodal, and positive surgical margin status (Table 1; Supplementary Fig. S2), strongly indicating that this cohort is representative. Fisher's exact test showed no clear associations between SLC18A2 immunoreactivity (positive versus negative) in prostatectomy tumor specimens and any of these indicators (Supplementary Table S4). However, when Gleason scores were instead compared with SLC18A2 intensity scores, absence of strong/moderate cytoplasmic staining was significantly associated with higher Gleason score (P = 0.029, χ² test; Supplementary Fig. S4).

Using log-rank statistics, negative cytoplasmic SLC18A2 immunoreactivity was significantly associated with PSA recurrence (Fig. 5A). Estimated mean RFS time for patients negative (score 0) for cytoplasmic SLC18A2 was 72 months (95% confidence interval, 60-84 months) compared with 115 months (95% confidence interval, 106-125 months) for positive (scores 1+ to 3+) cases (P < 0.001). The difference was also significant between patients with negative and weak (score 1+) cytoplasmic staining (P < 0.001). We note that negative cytoplasmic staining was related to shorter RFS times in patients with Gleason 7 tumors (P = 0.009; Fig. 5B), a large histologic subgroup for which outcome prediction is particularly difficult. This was also the case, when leaving out Gleason 7 cases (P < 0.001; not shown). Nuclear SLC18A2 staining was not associated with RFS (P = 0.416; Fig. 5B). Likewise, no significant associations were found between OS and cytoplasmic (P = 0.399) or nuclear (P = 0.072) SLC18A2 staining (Fig. 5C). Short OS was significantly associated only with high Gleason score (P = 0.001) and high tumor stage (P = 0.023; Table 1; Supplementary Fig. S3).

Using a global multivariate Cox regression model, high Gleason score (P = 0.026), high tumor stage (P = 0.003), positive surgical margins (P = 0.002), high preoperative PSA (P = 0.003), and negative cytoplasmic SLC18A2 staining (P = 0.001) were significantly associated with shorter RFS times (Table 1). After reverse selection, the same five variables remained in the model, strongly indicating that negative cytoplasmic SLC18A2 immunoreactivity is an independent adverse risk factor for prostate cancer recurrence. Our concomitant identification of four already established risk factors for recurrence supports the validity of this finding. The hazard ratio for biochemical recurrence in patients with positive cytoplasmic SLC18A2 staining was 0.485 (95% confidence interval, 0.333-0.709). When evaluated as a test for prediction of PSA recurrence within 5 years, cytoplasmic SLC18A2 immunohistochemical analysis (negative or positive) had a sensitivity of 0.48, specificity of 0.82, positive predictive value of 0.69, and negative predictive value of 0.67.

Whereas several studies have investigated the function of vesicular monoamine transporters in neurons and neuroendocrine cells, this study is to our knowledge the first to describe a possible role for monoamine transporters in human prostate adenocarcinoma development and progression. Using a bioinformatic approach combining whole-genome expression and SNP microarray data, we found that SLC18A2 was both significantly down-regulated and a common target for LOH in prostate adenocarcinomas. Reduced SLC18A2 expression in cancer compared with nonmalignant prostate tissues, as seen in several microarray expression profiling studies (21–25), was validated at the protein level by tissue microarray analysis of a large (>700) independent patient group. Furthermore, in a series of 506 unselected radical prostatectomy patients from two European countries, cytoplasmic SLC18A2 staining (positive versus negative) was found by multivariate analysis to be at least as strong an independent predictor of RFS as the established indicators: preoperative serum PSA, Gleason score, pathologic T stage, and surgical margin status.

Although this needs confirmation in future studies, our results indicate that a simple and inexpensive immunohistochemical analysis for cytoplasmic SLC18A2 staining in prostate tumor specimens may add significant prognostic value to routinely used predictors for patient outcome and thereby possibly could improve patient stratification, facilitate individualized treatment, and ultimately increase survival for patients with prostate cancer. Our experimental approach used radical prostatectomy specimens for immunohistochemical analysis; however, a similar test using prostate biopsies may have potential utility in clinical management of prostate cancer, where, for example, loss of cytoplasmic SLC18A2 staining could argue against watchful waiting and instead favor immediate intervention by radical prostatectomy or other treatments such as radiation or cryotherapy. Further studies should elucidate this aspect.

SLC18A2 is localized in the membrane of intracellular vesicles in neurons and neuroendocrine cells (6), consistent with cytoplasmic SLC18A2 staining in prostate secretory epithelial cells and prostate neuroendocrine cells. There are no previous reports of nuclear SLC18A2 staining, as found here to be associated with metastatic prostate cancer but not with outcome after prostatectomy. However, SLC18A2 has been visualized in the smooth endoplasmic reticulum of neurons (38), a compartment directly connected to the nuclear membrane. Also, aberrant protein glycosylation, as described for MUC1 in prostate cancer (36), is linked to altered intracellular SLC18A2 trafficking in neuroendocrine and neuronal cells (37, 39). Accordingly, the nuclear SLC18A2 immunoreactivity seen in a subset of advanced prostatic disease specimens most likely reflects mislocated SLC18A2 protein. Furthermore, several lines of evidence indicate that the SLC18A2 antibody used here has very high specificity. It has been extensively tested in several studies, where it was shown to specifically recognize SLC18A2 and not the most closely related protein SLC18A1 (vesicular monoamine transporter 1; refs. 8, 34, 37, 40). The 19-amino acid immunogenic sequence from SLC18A2 also has no significant homology with other proteins in the BLAST database.⁶

In the present study, SLC18A2 antibody specificity was further confirmed by Western blotting and immunohistochemical analysis of relevant tissue samples with and without known SLC18A2 expression.

The sample set used in this study did not allow a systematic comparison between SLC18A2 hypermethylation and LOH, although hypermethylation appeared to be the predominant mode of down-regulation compared with copy number loss (88% versus 23% in clinical tumor samples). In total, SLC18A2 was hypermethylated in 91% (20 of 22) of prostate adenocarcinoma cell lines and tumors investigated, indicating that SLC18A2 is one of the most frequent target of aberrant CpG island methylation in prostate cancer described to date. Such a high incidence has earlier been reported for only a few genes in prostate cancer, including GSTP1, APC, and 14-3-3σ (4).

The detection of SLC18A2 hypermethylation on both the frequent allele (not affected by LOH) and the rare allele (affected by LOH) in a crudely dissected tumor with LOH (PC-41) may suggest that hypermethylation of SLC18A2 precedes allelic loss and hence that LOH at 10q25 primarily is driven by other gene(s) than SLC18A2. The fact that negative but not weak cytoplasmic SLC18A2 staining was associated with increased recurrence risk indicates that complete SLC18A2 silencing may provide a survival advantage to prostate cancer cells. Consistent with this model, SLC18A2 hypermethylation generally affected both alleles in tumors without LOH. Detection of monoallelic hypermethylation in 2 of 5 BPHs and 17% methylation in BPH-1 cells suggests that SLC18A2 might also play a role for development of BPH, as proposed for other genes hypermethylated in both prostate cancer and BPH but unmethylated in normal prostate tissues (RASSF1A, 14-3-3σ, and MDR1; ref. 4). We observed an inverse correlation between SLC18A2 expression and hypermethylation levels, and silencing of SLC18A2 was reversed by 5-aza and 4-phenylbutyric acid treatment in PSK-1, DU145, and PC3 cells, suggesting that it could be a new candidate target for epigenetic anticancer therapy.

In mammalian cells, transport of monoamines is regulated by vesicular monoamine transporters 1 (SLC18A1) and 2 (SLC18A2), which have distinct but overlapping functions and expression patterns (6). As opposed to rodent prostate epithelial cells that express exclusively SLC18A1 (41), both transporters are expressed in normal human prostate epithelial cells. However, only SLC18A2 shows differential expression in nonmalignant versus prostate cancer tissue samples (own results; ref. 20 and Oncomine database),⁷

suggesting that it is the main monoamine vesicular transporter with a possible role in prostate cancer. In addition to their roles in neurobiology, monoamines are known to regulate prostate, breast, and colon cancer cell behavior (13–19), but the potential function of SLC18A2 and/or SLC18A1 in these almost exclusively protumorigenic processes remains to be investigated.

Serotonin, secreted by neuroendocrine prostate cells, stimulates proliferation of prostate cancer cells and xenografts, whereas serotonin receptor antagonists and serotonin uptake inhibitors reduce growth (14, 15, 42). Down-regulation of SLC18A2 in prostate cancer cells may prevent sequestering of serotonin into storage vesicles and thereby increase sensitivity to mitogenic serotonin released from neighboring neuroendocrine cells. This paracrine growth signal could be further boosted by epinephrine-induced neuroendocrine differentiation and thereby increased serotonin production as reported for prostate cancer cells in vitro (43). Similarly, loss of SLC18A2 might sensitize prostate cancer cell to the promigratory effects of norepinephrine released from sympathetic nerves (17, 18) and/or antiapoptotic stimuli from circulating epinephrine (19), both mediated via β₂-adrenergic receptors. Histamine, secreted by infiltrating mast cells, augments β₂-adrenergic receptor-mediated responses in prostate cancer cells in vitro independent of the H1 receptor (44) but has also been shown to inhibit proliferation via the H1 receptor (16). Finally, perineural invasion is regarded as a key mechanism for prostate cancer spreading, where neurons in the prostate seem to be a source for tumor-promoting paracrine factors, leading to reduced apoptosis and increased proliferation of nearby prostate cancer cells (45). Based on this, it is possible that SLC18A2 controls the level of proproliferative, antiapoptotic, and/or promigratory monoamines available in the normal prostate to maintain homeostasis. To our knowledge, this study is the first to indicate a possible tumor suppressor function for a vesicular monoamine transporter in prostate cancer and to suggest a role for SLC18A2 inactivation in prostate cancer progression.

No potential conflicts of interest were disclosed.

We thank Dr. Arnold Ruoho for kindly providing the SLC18A2 expression vector, Dr. Marie Stampe Ostenfeld for H69 and H69A cells, Dr. Kenneth Pienta for DuCaP and VCaP cells, and Dr. Adrie van Boekhoven for PSK-1 cells. We also thank Dr. Gerda Egger for helpful advice on 4-phenylbutyric acid treatment. Furthermore, Pamela Celis, Karen Bihl, and Susanne Bruun are acknowledged for excellent technical assistance.