Purpose: Bcl-2 is antiapoptotic, and its overexpression has been associated with resistance to androgen deprivation and poor outcome in some patients treated with radiotherapy. Bax is proapoptotic, regulating Bcl-2 through heterodimer formation. In a prior study, Bcl-2 and Bax were not related to outcome in locally advanced patients treated with radiotherapy or short-term androgen deprivation + radiotherapy (STAD+RT) on another Radiation Therapy Oncology Group trial (86-10). A follow-up investigation was carried out here in more contemporary high-risk men treated on Radiation Therapy Oncology Group 92-02 with STAD+RT or long-term AD+RT (LTAD+RT).
Experimental Design: Adequate tissue was available to be analyzed immunohistochemically in 502 patients for Bcl-2 and 343 patients for Bax. Univariate and multivariate analyses by Cox proportional hazards models were applied to end points of failure.
Results: Bcl-2 was positive in 45.6% cases, and Bax expression altered in 53.9% cases. Abnormal Bcl-2 was not related to any of the failure end points tested. Altered Bax expression was significantly associated with any failure (P = 0.023) and marginally with biochemical failure (P = 0.085). The combination of negative Bcl-2/normal Bax expression seemed more robust, being significantly related to reduced biochemical failure (P = 0.036) and any failure (P = 0.046). The predictive value of negative Bcl-2/normal Bax was most pronounced in those who received STAD+RT, as opposed to LTAD+RT.
Conclusions: Normal Bax expression was associated with significantly more favorable outcome. The combination of negative Bcl-2 and normal Bax was more consistently significant, particularly when STAD+RT was the treatment administered. These data suggest that LTAD+RT should be used when either Bcl-2 or Bax is abnormally expressed.
- Androgen Deprivation
- Radiation Therapy
The Bcl-2 family proteins consist of members with opposing functions, interacting with one another to achieve a balance between cell survival and apoptosis signals. Bcl-2, a pro-survival member found in the basal epithelium of the prostate, has been extensively studied in prostate cancer. Androgen deprivation treatment causes an increase in Bcl-2 expression (1–4), and the overexpression of Bcl-2 is associated with the development of androgen independence (1, 5, 6). Our in vitro studies also show Bcl-2 suppression by antisense Bcl-2 sensitizes prostate cancer cells to radiation (7).
Conversely, Bax is proapoptotic (8, 9). Bax expression is primarily in the secretory epithelial cells of the prostate and thus more susceptible to apoptosis caused by androgen deprivation (10). In vivo studies show that Bax up-regulation temporarily induces apoptosis in the setting of androgen deprivation (11). Indeed, it has even been shown that in androgen-independent tumors that are androgen receptor positive, reintroduction of androgen reduces tumor growth via Bax-mediated apoptosis (12). The influence of these proteins on persistent localized prostate cancer characterizes them as highly suitable candidates for prognostic markers of the disease.
Previous analyses of Bcl-2 and Bax expression by our group and others involved men treated with definitive radiotherapy alone (13, 14) and short-term androgen deprivation + radiotherapy (STAD+RT; ref. 14). In a study including patients with a range of risk factors (e.g., T1-T3), abnormal expression of both proteins was associated with treatment failure (13). In another cohort of more locally advanced patients [Radiation Therapy Oncology Group (RTOG) trial 86-10], neither Bcl-2 nor Bax was predictive of outcome (14); prostate-specific antigen (PSA) data were not available in this older cohort. Both studies were also limited in sample size.
The present study investigates the value of Bcl-2 and Bax expression in a more contemporary high-risk group of men who were enrolled in the phase III randomized trial (RTOG 92-02). RTOG 92-02 builds on the prior RTOG experience, comparing STAD+RT with long-term AD+RT (LTAD+RT).
Materials and Methods
Study population. RTOG protocol 92-02 has been previously described (15, 16). There were 1,518 assessable patients. Of these, 502 pretreatment diagnostic needle-core biopsies or transurethral resection tumor specimens were acquired and adequately stained for Bcl-2 (and 343 for Bax) expression. Sixty additional cases stained for Bax showed high staining levels with excessive background and were unsuitable for analysis. The slides stained for Bax were also those cut closer to the end of the blocks, and 99 cases were found to be lacking in tissue or adequate tumor.
The median age of the Bcl-2 study cohort was 70 years (range, 43-88 years), median initial pretreatment PSA (iPSA) was 20.7 ng/mL (range, 0.9-219.7 ng/mL), and 283 (56.4%) patients had T3-T4 disease. Median follow-up for patients alive in the Bcl-2 cohort was 10.4 years. The Bax study group had a median age of 71 years (range, 49-86 years) and median iPSA of 21.7 ng/mL (range, 0.8-151.1 ng/mL), and 179 (52.2%) patients had T3-T4 disease. Median follow-up for patients alive in the Bax cohort was 10.5 years.
Immunohistochemical technique. The immunohistochemical protocol was detailed previously (14). The primary antibodies used were Bcl-2 (clone 124, DAKO Corp.; 1:100 dilution) and Bax (clone 2D2, Zymed Laboratories, Inc.; 1:200 dilution). The labeled streptavidin-biotin (LSAB) immunohistochemical method was used (DAKO LSAB 2 kit) followed by a diaminobenzidine chromagen and a hematoxylin counterstain (DAKO) for visualization. All staining was done on an autostainer (DakoCytomation).
The investigators had no knowledge of patient outcome. The first investigator (L-Y.K.) independently examined the slides for cytoplasmic staining intensity under light microscopy. A tissue microarray control consisting of four plugs of prostate cancer tissue, each expressing a different intensity (0-3), was referenced to rate the intensity of each case. A cluster of >20 positive tumor cells was considered Bcl-2 overexpression. When two or more intensities were encountered, the highest intensity was recorded. Normal prostate epithelium also served as an internal control for Bcl-2 staining. Underexpression or overexpression of Bax relative to the staining intensities of neighboring normal prostate epithelial cells were considered “altered Bax expression” (13, 14). Both the prostate cancer tissue microarray control and a separate human lymph node tissue were used as Bax staining controls. All the cases were then reviewed with an oncologic surgical pathologist with a special interest in urological pathology (T.A-S.). Both investigators sat at a multi-headed microscope and discussed each case. Minor variations in the grading of staining intensity between the investigators was seen in 130 Bcl-2 and 170 Bax cases. These differences would not have altered the final classification of Bcl-2–positive (any staining) or altered Bax staining.
Definition of end points. Two definitions were considered for biochemical failure: the American Society of Therapeutic Radiology and Oncology (ASTRO) consensus definition (ASTRO BF), three consecutive increases in PSA or a posttreatment PSA nadir > 4.0 ng/mL (17) and the Phoenix definition (BF Phoenix), PSA > nadir + 2 ng/mL after treatment (18). Local failure was assessed by palpation and defined as tumor growth of 25% or local persistent disease beyond 18 months. Distant metastasis was defined as clinical evidence of distant disease by any method. The failure events for any failure were defined as all occurrences (local failure, regional nodal failure, distant failure, or ASTRO BF) except death. Cause-specific mortality was death certified as due to prostate cancer, death due to treatment complications, death from unknown causes with active malignancy (clinical disease relapse), or death from another cancer with documented bone metastases attributed to prostate cancer before the appearance of the second independent cancer. Finally, overall mortality was defined as death due to any cause.
Statistical analysis. All pretreatment characteristics were dichotomized. Age was dichotomized by the median age in the entire cohort (<70 versus ≥70 years), iPSA and T stage by their stratification groupings in RTOG 92-02 (≤30 versus >30 ng/mL; T2b/T2c versus T3/T4), and Gleason score as 2 to 6 versus 7 to 10. The χ2 test and Cox proportional hazards model (19) were applied to assess if the missing Bcl-2 and Bax data were dependent on pretreatment characteristics, treatment arm, or outcome. Actuarial estimates for overall mortality were based on Kaplan-Meier methods (20), and the cumulative incidence method was used to estimate cause-specific mortality, distant metastasis, local failure, biochemical failure, and any failure (21).
Bcl-2 was overexpressed in 229 (45.6%) of the cases studied. Altered Bax was seen in 185 (53.9%) cases: 26 (7.6%) cases had Bax underexpression, and 159 (46.4%) cases had Bax overexpression. In prior studies, either type of abnormal Bax expression was associated with an adverse outcome (13, 14); thus, they were considered together here.
Comparisons were made between cases that were assessed for Bcl-2 (n = 502) and those that were not (n = 1,016). There were no significant differences in age, iPSA, or T stage. Those with Bcl-2 data had a greater proportion of Gleason score 7 to 10 (65.5% versus 55.9%; P = 0.0005), and more cases were assigned to the LTAD+RT arm (53.6% versus 47.9%; P = 0.0382). Likewise for Bax, the distributions of patients by pretreatment characteristics were similar, except a greater proportion in the Bax cohort had Gleason score 7 to 10 (64% versus 57.7%; P = 0.0423), and more cases were assigned to LTAD+RT (55.4% versus 48.2%; P = 0.0186) compared with the cases in which Bax staining was not available. There were no statistical differences between the biomarker (Bcl-2 or Bax) available and unavailable groups in terms of outcome by all the end points (overall mortality, cause-specific mortality, distant metastasis, local failure, biochemical failure, or any failure) tested.
Table 1 shows the distribution of patients by Bcl-2 and Bax staining, pretreatment prognostic factors, and assigned treatment. Bcl-2 overexpression was significantly related to Gleason score 7 to 10 disease (70.7% versus 61.2%; P = 0.0305), whereas altered Bax expression and the combined results were not related to any pretreatment factor (age, Gleason score, iPSA, and T stage) or assigned treatment.
In the univariate analyses, Bcl-2 positivity was significantly related to overall mortality [relative risk (RR), 1.31; 95% confidence interval (95% CI), 1.03-1.67; P = 0.0307] and distant metastasis (RR, 1.51; 95% CI, 1.02-2.23; P = 0.0387). However, in multivariate analyses, Bcl-2 overexpression was not independently related to these or any of the other end points tested, although the RRs for failure were higher for overall mortality, cause-specific mortality, distant metastasis, and local failure (Table 2 ). In univariate analyses, altered Bax expression was not significantly associated with any of the end points tested, although there was a trend for any failure (RR, 1.30; 95% CI, 0.97-1.73; P = 0.0806). In the multivariate analyses (Table 3 ), abnormal Bax expression was significantly associated with any failure (RR, 1.43; 95% CI, 1.05-1.95; P = 0.0226) and marginally with biochemical failure using the ASTRO definition (RR, 1.37; 95% CI, 0.96-1.97; P = 0.0851).
Because both Bcl-2 and Bax seem to contribute to the apoptotic response of prostate tumor cells to radiation and androgen deprivation (22–24), the two biomarkers were examined together. We first divided the patients into three groups: negative Bcl-2 and normal Bax expression (negative Bcl-2/normal Bax) versus either positive Bcl-2 or abnormal Bax expression versus positive Bcl-2/abnormal Bax expression. The first group was segregated from the other two. The results displayed summarize the comparisons of negative Bcl-2/normal Bax versus the other possible combinations of positive Bcl-2 and/or altered Bax expression. In univariate analyses, negative Bcl-2/normal Bax was significantly related to both any failure (RR, 1.44; 95% CI, 1.01-2.03; P = 0.0408) and ASTRO BF (RR, 1.62; 95% CI, 1.06-2.48; P = 0.0242). In the multivariate analyses (Table 4 ), the relationships remained significant for any failure (RR, 1.45; 95% CI, 1.01-2.10; P = 0.0457) and biochemical failure (RR, 1.60; 95% CI, 1.03-2.49; P = 0.0359). The failure curves for these end points are displayed in Fig. 1 . These results seemed more robust than the relationships seen with Bax alone.
Finally, the relationship of Bcl-2/Bax expression to length of androgen deprivation therapy (assigned treated in the protocol) was tested. Table 5 shows that the 5-year failure rates (biochemical failure and any failure) were much greater for patients who had positive Bcl-2 and/or altered Bax tumors when STAD+RT was used. Although there was a trend for worse outcome in patients with negative Bcl-2/normal Bax tumors after STAD+RT, compared with LTAD+RT, the association was much less apparent. In the setting of LTAD+RT, positive Bcl-2 and/or altered Bax expression was not related to increased failure; the failure rates were about the same as for patients with negative Bcl-2/normal Bax expression.
The predictive value of Bcl-2 and Bax in relation to clinical outcome has been investigated in various correlative studies on prostate cancer, and the results have been contradictory or inconclusive (13, 14, 22, 24–27). The reasons for this may include variations in clinical characteristics of the study cohorts, different scoring methods, and low statistical power. The overwhelming evidence for the involvement of Bcl-2 and Bax in prostate cancer recurrence (22, 24, 27) and progression to androgen independence (1, 5, 6, 28–30) highlight the need for further studies of these biomarkers. The current investigation is, to our knowledge, the largest multi-institutional study of Bcl-2 and Bax expression in contemporary high-risk prostate cancer patients.
The study population described herein was derived from patients treated on RTOG protocol 92-02 with STAD+RT or LTAD+RT. There were some slight differences in patient characteristics between those in the study cohort (Bcl-2 and/or Bax available) and the remaining biomarker unavailable patients. However, there were no differences in patient outcome between the two groups using a number of end points.
The approach used was modeled after that described in two prior reports from our group (13, 14). We initially reported that both Bcl-2 overexpression (16% of cases) and abnormal (underexpression or overexpression) Bax expression (23% of cases) were significant predictors of biochemical failure in men treated with external beam radiotherapy alone at a single institution in the PSA era. The associations of Bcl-2 and Bax expression were independent of iPSA, Gleason score, T stage, and Ki-67 staining. The best outcome was seen in those whose tumors had a negative Bcl-2/normal Bax pattern.
In the second study, Bcl-2 overexpression (26% of cases) and altered Bax expression (47% of cases) were measured in pretreatment diagnostic tissue samples from RTOG 86-10, a randomized trial comparing radiotherapy alone with STAD+RT. The STAD+RT treatment in this protocol was identical to the STAD+RT treatment in the current analysis of patients in RTOG 92-02. Neither Bcl-2 nor Bax was predictive of patient outcome. We did not combine Bcl-2 and Bax expression in the analysis of 86-10 patients because there was no suggestion of a relationship with either marker when tested individually. The patients in this trial were locally advanced and were treated in the pre-PSA era. The patients in RTOG 92-02 represent more contemporary high-risk patients.
In both prior studies of Bcl-2 and Bax expression from our group, the number of cases examined was in the 100 to 120 range. Biomarker studies are typically underpowered. An important feature of the analysis described here is that it involves much greater numbers and, consequently, power to show significance. Bcl-2 overexpression was seen in 46% and abnormal Bax expression in 54%. These percentages are higher than in our prior studies, possibly reflecting better preservation of antigens. It should be noted that the levels of Bcl-2 overexpression and abnormal Bax expression in samples from RTOG 92-02 are the same as for prospectively analyzed samples in another single institution cohort of men treated with external beam radiotherapy (31).
In multivariate analyses of cases from RTOG 92-02, Bcl-2 overexpression was not associated with any outcome measure, whereas altered Bax expression was significantly associated with any failure and was borderline significant for ASTRO BF. The most consistent relationship was seen when Bcl-2 and Bax expression were considered together. These data are concordant with our first analysis (13) and other reports of combining Bcl-2 and Bax (22). Some have advocated the Bcl-2/Bax ratio as being important (32–35), but in our hands, Bax overexpression has been a determinant of adverse outcome, and the Bcl-2/Bax ratio is not applicable.
We also investigated the influence of Bcl-2 and Bax expression on patient outcome when stratified by length of androgen deprivation. There was an apparent benefit in giving LTAD+RT to patients with positive Bcl-2 and/or abnormal Bax expression, with similar more favorable results to those who had negative Bcl-2/normal Bax (Table 5). The greatest risk of failure was seen in men who received STAD+RT and who had positive Bcl-2 and/or abnormal Bax.
In conclusion, Bcl-2 and Bax expression were most predictive of outcome when combined, although at this point, the associations seen were limited to ASTRO BF and any failure. The data suggest that men with tumors overexpressing Bcl-2 and/or altered Bax expression should receive LTAD+RT as opposed to STAD+RT. The RTOG translational research group has been investigating a series of potential biomarkers in tissue from patients in protocol 92-02 including Ki-67, p53, MDM2, p16 (16, 36, 37), and now Bcl-2 and Bax. The aim is to ultimately establish a consensus biomarker panel for risk stratification of high-risk prostate cancer patients. The data described suggest that Bcl-2 and Bax should be included, after which the panel will be validated on an independent patient cohort.
Grant support: National Cancer Institute grants CA-006927, CA-101984-01, and CA-109556; Varian Medical Systems, Palo Alto, CA; and Pennsylvania Department of Health.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: The contents are solely the responsibility of the authors and do not necessarily represent the official views of the supporting organizations.
- Received December 14, 2006.
- Revision received February 22, 2007.
- Accepted March 12, 2007.