PERSISTENT PROSTATE-SPECIFIC ANTIGEN EXPRESSION AFTER NEOADJUVANT ANDROGEN DEPLETION: AN EARLY PREDICTOR OF RELAPSE OR INCOMPLETE ANDROGEN SUPPRESSION

Abstract

Objectives.

To analyze post-androgen depletion (AD) primary tumors to identify markers of treatment failure because AD does not reduce the probability of prostate-specific antigen (PSA) failure after prostatectomy.

Methods.

Tumors removed by radical prostatectomy after 3 months of AD from 21 patients were analyzed for gene expression using oligonucleotide arrays. Differences between patients with and without relapse were identified using a conservative significance criteria of a threefold change and delta 0.68, ensuring a false discovery rate of less than 11%.

Results.

At 50 months of follow-up, 7 of the 18 evaluable patients developed a biochemical recurrence. Gleason grade, pretherapy PSA level, T stage, and margin status were similar between the two groups. Patients with recurrence had greater post-AD PSA levels than those without recurrence (0.87 versus 0.19 ng/mL; P = 0.042). Gene expression analysis revealed 35 probe sets overexpressed in tumors from patients who relapsed. Among the highest ranked probe sets were PSA and other androgen-responsive genes. Serum PSA values during AD revealed similar findings. After 40 days of AD, the PSA level in those without recurrence was 1.21 ng/mL versus 4.5 ng/mL in those with recurrence (P = 0.0034). Immunohistochemistry of post-AD tumors also demonstrated a high PSA staining intensity in many tumors that recurred relative to those that didn’t.

Conclusions.

The results of our study show that early recurrence is associated with expression of androgen-responsive genes. Surprisingly, these could be identified as early as 3 months after the initiation of AD therapy. Whether this represents a failure to abrogate androgen receptor mediated signaling with androgen depletion or early reactivation of signaling is under study.

Androgen deprivation (AD) slows prostate cancer growth by reducing signaling through the androgen receptor (AR). Although it has been hypothesized that AD before surgery may improve outcomes compared with surgery alone, trials designed to test this hypothesis have been disappointing, because no study of AD has shown a delay in the interval to prostate-specific antigen (PSA) relapse, clinical metastasis, or prolongation of life. Furthermore, reducing serum testosterone to castrate levels rarely eradicates the tumor completely.^1-3 Although significant differences in global gene expression exist between localized and metastatic disease, AR signaling is a principal effector of tumor growth and survival throughout all disease states. AR signaling in an androgen-depleted environment is, therefore, a hallmark of a castration-resistant tumor phenotype.⁴

After prostatectomy, a rising PSA level signifies disease recurrence. Subsequent outcomes vary but have correlated with the onset and rate of PSA rise.⁵ Because PSA gene transcription is regulated by the AR, the re-emergence of PSA after surgery suggests the presence of a tumor with a functional AR signaling axis.

The present study used molecular profiling of past androgen depletion primary prostate cancers and clinical follow-up to explore whether an analysis of the residual tumors, insofar as they represent a subset of the primary tumor that has survived castration, can reveal factors predictive of outcome. Our findings suggest that a reduced response to AD, as reflected in persistent expression of androgen responsive genes, is associated with a greater likelihood of recurrence.

MATERIAL AND METHODS

Patient Tumor Samples

Tissues were obtained from prostatectomy specimens performed at Memorial Sloan-Kettering Cancer Center. The primary tumors removed from 21 patients (Table I) who had received 3 months of AD (subcutaneous goserelin monthly and daily oral flutamide for 3 months before prostatectomy) and the tumors from 23 patients who had received no AD were analyzed. The Memorial Sloan-Kettering Cancer Center Institutional Review Board approved all studies.

TABLE I.

Patient characteristics

Characteristic	No Relapse (n = 11)	Relapse (n = 7)	P Value
Age (yr)
Median	59	61
Range	48–71	54–69
PSA at diagnosis (ng/mL)
Median	7.80	13.0	0.72
Range	0.8–52.5	4.3–16
PSA nadir before prostatectomy (ng/mL)
Median	0.10	0.53	0.04
Range	0–0.26	0–2.14
Gleason sum (n)
median	7	7
6	4	0
7	6	6
8	1	1
T stage (n)
T2	8	3
T3	1	3
T4	1	0
TxN1	1	1
Extracapsular extension	4 (36)	4 (57)
Seminal vesicle invasion	1 (9)	2 (29)
Positive surgical margins	4 (36)	1 (14)

Data in parentheses are percentages.

Gene Expression Analysis

Total RNA was extracted from frozen tissues by homogenization in guanidinium isothiocyanate-based buffer (TRIzol, Invitrogen, Carlsbad, Calif), purified using RNAeasy (Qiagen, Valencia, Calif), and evaluated for integrity by denaturing agarose gel. Complementary DNA was synthesized from total RNA using a T7-promoter-tagged oligo-dT primer. The RNA target was synthesized by in vitro transcription and labeled with biotinylated nucleotides (Enzo Biochem, Farmingdale, NY). The labeled target was assessed by hybridization to Test2 arrays (Affymetrix, Santa Clara, Calif) and detected with phycoerythrin-streptavidin (Molecular Probes, Eugene, Ore) amplified with anti-streptavidin antibody (Vector, Burlingame, Calif). Gene expression analysis was performed using Affymetrix U95 human gene arrays with 63,175 features for individual gene/expressed sequence tag (EST) clusters using protocols recommended by the manufacturer. The U95 set consists of five distinct microarrays (A through E), each containing probes for approximately 12,000 gene/EST transcripts. Values on each array were multiplicatively scaled to an average expression of 2500 across the central 96% of all genes on the array.

Statistical Analysis

Scanned image files were visually inspected for artifacts and analyzed using Microarray Suite, version 4.0 (Affymetrix). Gene expression analysis was performed using the significance analysis of microarrays software platform. Significance criteria consisted of a threefold change and delta of 0.68, which ensured a false discovery rate of 10.7%, which accounts for the small sample size and generates a list of genes large enough for meaningful analysis. The data sets used for hierarchical clustering were normalized by standardizing each gene and sample (array) to a mean of 0 and variance of 1. Hierarchical clustering and display were performed using Cluster and TreeView software (28). Genes corresponding to probe sets were identified through the Affymetrix database (http://www.affymetrix.com/index.affx). Comparisons of the PSA measurements of groups was performed using Student’s two-tailed t test. Correlations between gene expression and serum PSA levels were calculated using the formula of Pearson.

Immunohistochemistry

Immunohistochemical detection of PSA (1:2000, Biogenex, San Ramon, Calif) was done with the streptavidin-biotin-peroxidase method using formalin-fixed, paraffin-embedded tissue and microwave antigen retrieval (20) from multitissue blocks containing paraffin-embedded tissue corresponding to the samples used in this analysis.

RESULTS

Clinical Features and Response of Serum PSA to AD

Follow-up data were available for 18 of the 21 patients who underwent radical prostatectomy after neoadjuvant AD and had gene expression profiling. Three patients were lost to follow-up. Of the 18 patients, 7 (39%) had experienced a recurrence in the form of a detectable and rising serum PSA level. The median time to recurrence was 36 months (range 10 to 57). The median time of follow-up for patients without recurrence was 52 months.

The rate of decline in serum PSA levels during AD was compared between those without and with relapse. The serum PSA level was measured at baseline and at two intervals before prostatectomy (Fig. 1). The pretherapy median PSA level was 7.75 ng/mL (range 0.8 to 52.5) and 13.0 ng/mL (range 4.3 to 16; P = 0.72), and the median PSA level at 40 and 42 days after treatment was 1.21 and 4.5 ng/mL, for those without and with relapse, respectively (P = 0.034). At 67 and 79 days, the corresponding PSA values were 0.19 and 0.87 ng/mL (P = 0.042). At this point, 6 (60%) of 10 patients without relapse had experienced a PSA decline of 90% or greater compared with 1 of 8 patients with relapse (12.5%, P = 0.011). The interval from the final serum PSA level until prostatectomy varied because the timing of surgery was at the discretion of the urologist. The total time between the initiation of AD and surgery was 99 days and 135 days for those without versus those with recurrence, respectively (P = 0.007). The 90% PSA reduction rate at this point was 10 (100%) of 10 for those without recurrence versus 6 (75%) of 8 for those with recurrence (P = 0.048).

FIGURE 1.

Serum PSA during neoadjuvant AD. Serum PSA measured at baseline (Pre-Rx) and at intervals (PSA 2 after median of 40 and 42 days and PSA 3 obtained at 67 and 79 days for those without and with relapse, respectively) before prostatectomy.

Gene/EST Sequences Expressed in Association with Relapse

A revised analysis of gene expression was performed comparing those without and with recurrence. The criteria for differential expression included a threefold change between groups and a delta of 0.68, which defined the false discovery rate as 10.7%. This analysis revealed 35 of 12,496 significantly different probe sets (Fig. 2). The genes/ESTs most significantly overexpressed in the tumors of patients with subsequent relapse included many that are known to be androgen regulated such as KLK3 (PSA), FOLH1 (prostate-specific membrane antigen), KLK2, prostatic acid phosphatase, prostate-specific transglutaminase, KLK1, pyrroline 5-carboxylate reductase, and cyclic guanosine monophosphate phosphodiesterase. Patients who were treated with AD and did not experience serologic relapse had a low overall expression of PSA (Fig. 3). In contrast, those who were treated with AD and did relapse had consistently greater PSA mRNA levels. No difference was found in the AR mRNA levels between those with and without relapse (2514 versus 2767, P = 0.50).

FIGURE 2.

Genes highly expressed in patients with relapse. Gene expression level of most highly differentially expressed genes comparing those with versus those without relapse. Expression values range from highest (blue) to lowest (red). Individual patients represented along top as belonging to relapsing group (Rel X) versus nonrelapsing (NRX).

FIGURE 3.

PSA/KLK3 gene expression correlation with tissue expression of PSA by immunohistochemistry. Values on Y axis represent intensity of expression on Affymetrix U95A chip for KLK3/PSA. Pictures represent immunohistochemical analysis of PSA expression in paraffin-embedded tumors corresponding to marked gene expression levels.

The expression of serum PSA correlated with PSA gene expression and staining intensity by immunohistochemistry. A moderate positive correlation (correlation coefficient 0.4) was observed between the serum PSA measured after 2 months of AD and the PSA mRNA measured by gene array. PSA immunostaining in tumors was more intense in those with high levels of gene expression. The expression intensity for PSA in those with relapse was similar to that of those who had not received neoadjuvant AD (Fig. 3). AR mRNA and immunostaining was not significantly different between those with and without relapse and untreated controls.

COMMENT

This analysis of gene expression in post-AD primary tumors has demonstrated that the residual tumors of patients who subsequently experience relapse had greater expression levels of a variety of androgen-regulated genes, including several forms of PSA. These differences were observed, despite a lack of a significant difference between the groups with respect to pretherapy PSA, T stage, Gleason grade, or total content of tumor analyzed; that they were observed after only 3 months of AD was surprisingly early.

Previous work from our group has demonstrated that AD leads to dramatic changes in the expression of many androgen-responsive genes.⁴ The present data have suggested that the magnitude of the change in these genes correlates with likelihood of PSA relapse. Although little is known about the subsequent clinical outcomes of this cohort of patients (eg, metastasis rate, survival), these findings suggest that AR activity in residual tumors after exposure to androgen ablation may help identify a more aggressive disease phenotype. By extension, these data support the hypothesis that resistance to the effects of androgen ablation may be an early, intrinsic trait of some tumors.

Kitagawa et al.⁶ recently reported that the absence of a significant histologic “treatment effect” in primary prostate tumors after AD corresponded with a greater likelihood of relapse. Our findings have supplemented this observation and imply that the biologic basis of the failure of neoadjuvant AD is, in fact, a failure of AD to downregulate the activity of the AR in all patients.⁶

The differences in gene expression between those with and without relapse were evident, even though those with relapse were treated with neoadjuvant AD for a significantly longer period than those without relapse (135 versus 99 days, respectively, P = 0.007). A potential confounding factor, therefore, may be that many men, particularly those with relapse, went more than 1 month without AD in the group with relapse. This may explain why patients who waited longer for surgery had greater PSA expression in the prostate, but would not explain the greater rate of relapse in this group. Alternatively, rather than confound these data, this fact may support the hypothesis that persistent androgen-responsive gene expression is intrinsic to the resistant tumors and that a longer duration of androgen ablation might not be associated with a greater likelihood of AR suppression, although definitive data on this question are lacking.

Furthermore, the gene expression patterns seen in the primary tumors paralleled differences in serum PSA, despite the presence of low serum PSA values overall. It is also significant that probe sets corresponding to PSA and PSA isoforms were the most differentially expressed among all 12,496 probe sets studied (Fig. 2).

Although seemingly small, marginally statistically significant differences in serum PSA measurements during neoadjuvant AD were observed between the two groups. Although the assay is sensitive to minor fluctuations at low levels, a PSA level of less than 1 ng/mL is generally considered to reflect a response to AD therapy, with low levels of PSA expression attributed to constitutive production by nonneoplastic prostate. These data were also significant when comparing the proportion of those with eventual relapse who achieved a 90% reduction in PSA versus those who did not. That these differences were apparent in as few as 4 months was surprising and shows the prognostic significance of a posttherapy PSA level after AD (note the significant differences at 40 days in the two groups in Fig. 1). An additional explanation for this has recently been offered by Titus and colleagues,⁷ who have shown that disease recurrence after AD is associated with the presence of greater levels of androgens within the prostate. Application of this method may be useful in future studies of neoadjuvant AD in patients undergoing radiotherapy, for which this practice is in widespread use.

Current thinking has suggested that multiple pathways contribute to castration-resistant growth.^8,9 The unknown biologic question is whether these findings are representative of persistent AR activation within a subpopulation of the primary tumor, because of a reactivated AR, indicating rapid development of androgen-independent tumor cell survival, or because of failure to suppress the AR-signaling axis with combined androgen blockade in this group of patients. The gene/EST sequences seen in the present analysis also did not suggest profiles consistent with metastatic phenotype¹⁰ (eg, increases in markers of cellular adhesion, angiogenesis). They did show that the residual cells present within the prostate harbor a signature for continued androgen/AR signaling.

CONCLUSIONS

Castration resistance is the primary clinical manifestation of progression of, and death from, prostate cancer. A critical gap in our knowledge of the development of this clinical phenotype is not only how, but when, it develops. The results of this study have demonstrated a correlation between the serum PSA level and expression of gene sequences corresponding to this protein and have shown that persistent AR signaling correlates with early relapse. Future work should seek to distinguish whether failure to suppress or early reactivation of the AR underlies castration resistance. A clinical trial combining the depletion of androgen receptor ligand with receptor targeting is ongoing.