The prognostic value of expression of HIF1α, EGFR and VEGF-A, in localized prostate cancer for intermediate- and high-risk patients treated with radiation therapy with or without androgen deprivation therapy

Purpose Androgens stimulate the production of hypoxia-inducible factor (HIF1α) and ultimately vascular endothelial growth factor (VEGF-A). Additionally, epithelial growth factor (EGF) mediates HIF1α production. Carbonic anhydrase IX (CAIX) expression is associated with tumor cell hypoxia in a variety of malignancies. This study assesses the prognostic relation between HIF1α, VEGF-A, EGF Receptor and CAIX expression by immunochemistry in diagnostic samples of patients with intermediate- and high-risk localized prostate cancer treated with radiation therapy, with or without androgen deprivation therapy (ADT). Materials and methods Between 1994 and 2004, 103 prostate cancer patients (mean age, 68.7 ± 6.2), with prostate cancer (mean PSA, 13.3 ± 3.7), were treated with radiation therapy (RT, median dose, 74 Gy). Fifty seven (55.3%) patients received ADT (median duration, 6 months; range, 0 – 24). Median follow-up was 97.6 months (range, 5.9 – 206.8). Results Higher EGFR expression was significantly (p = 0.04) correlated with higher Gleason scores. On univariate analysis, HIF1α nuclear expression was a significant (p = 0.02) prognostic factor for biological progression-free survival (bPFS). A trend towards significance (p = 0.05) was observed with EGFR expression and bPFS. On multivariate analysis, low HIF1α nuclear (p = 0.01) and high EGFR (p = 0.04) expression remained significant adverse prognostic factors. Conclusions Our study suggests that high nuclear expression of HIF1α and low EGFR expression in diagnostic biopsies of prostate cancer patients treated with RT ± ADT is associated with a good prognosis.

Results: Higher EGFR expression was significantly (p = 0.04) correlated with higher Gleason scores. On univariate analysis, HIF1α nuclear expression was a significant (p = 0.02) prognostic factor for biological progression-free survival (bPFS). A trend towards significance (p = 0.05) was observed with EGFR expression and bPFS. On multivariate analysis, low HIF1α nuclear (p = 0.01) and high EGFR (p = 0.04) expression remained significant adverse prognostic factors. Conclusions: Our study suggests that high nuclear expression of HIF1α and low EGFR expression in diagnostic biopsies of prostate cancer patients treated with RT ± ADT is associated with a good prognosis.

Background
It has been recognized that cancer-stromal cell interactions is a major player of malignant behavior in cancer. More specifically, hypoxia may trigger vascular endothelial growth factor (VEGF) expression via the transcription complex of hypoxia-inducible factor HIF1α. Hypoxia and the consequential angiogenesis may play a major role in prostate cancer progression, as VEGF and HIF1α is increased in prostate cancer, when compared to benign prostatic hypertrophy [1,2]. Additionally, a direct link between androgen receptors and pro-angiogenic factors may exist, as HIF1α, via epithelial growth factor (EGF), expression is increased with androgens [3] and decreased in prostatectomy specimen treated with pre-operative androgen deprivation therapy (ADT) [4]. The carbonic anhydrase IX (CAIX) gene is a target of HIF1α and is up-regulated in hypoxia [5]. Likewise, it has been shown that androgen deprivation in cell culture decreases VEGF mRNA expression [6] and castration in rodents' decreases VEGF expression in androgen sensitive xenografts [7]. The prognostic significance of EGFreceptor (EGFR), HIF1α and VEGF-A in prostate cancer is somewhat disputed. Some series have shown a negative outcome in patients with pre-treatment VEGF expression [8], whilst other studies have suggested an improved outcome in patients undergoing radical prostatectomy [4].
Given the lack of strong prognostic evidence regarding these markers in patients who may benefit from dose escalation RT, we analyzed tumor expression of EGFR, HIF 1α, VEGF-A and CAIX with respect to freedom from biochemical progression in men with intermediateand high-risk prostate cancer treated with RT, with or without ADT.

Methods
Between 1994 and 2004, 103 patients with clinically localized (according to the 2002 TNM American Joint Committee on Cancer staging system: cT1c -cT2b) or locally advanced (cT3a -cT4) non-metastatic prostate cancer with diagnostic samples with histological proven adenocarcinoma were indentified in our institutional database. All patients underwent bone scintigraphy and endo-rectal MRI staging prior to radiation therapy (RT). All patients were classified as having intermediate-risk disease (n = 53; 51.5%) or high risk (n = 50; 48.5%), as defined by d' Amico et al. criteria (Table 1). Six months ADT was administered to 57 (55.3%) patients (Table 1). RT (mean dose, 75.1 ± 2.8 Gy; Table 1) was delivered concomitantly after 3 months of ADT for those receiving ADT or immediately for those not receiving ADT. Patients were followed with 6-monthly PSA tests. This study was approved by the institutional ethic committee (NAC 08-076R) and complied to the Helsinki declaration. Prior to study initiation, written, informed consent to perform this analysis was obtained from all patients. The mean duration of the follow up time was 96.4 ± 33.7 months. No patients were lost to follow-up.

Immunochemistry
All tissues obtained by prostate biopsy or transurethral resection of the prostate were formalin-fixed or Duboscq-Brazil-fixed and paraffin embedded. The hematoxylin-eosin stained sections were reviewed to confirm the diagnosis and only sections showing typical Gleason score were selected. For immunohistochemistry (IHC), section 4 μm from one representative block of each patient were deparaffinized, rehydrated, and then submitted to IHC analysis as follow.
HIF1α After boiling with a pressure cooker in Tris-EDTA pH:9.0 buffer for 3 min, sections were incubated in DAKO autostainer with the monoclonal mouse HIF1α antibody clone H1alpha67 (NB100-123, NOVUS Biologicals) diluted 1/1000 and stained with CSA-II-Biotin-free Tyramide Signal Amplification System (K1497, DAKO). Renal clear cell carcinoma served as a positive control. Primary antibody was substituted with mouse IgG2b for negative control.
VEGF-A After boiling with a pressure cooker in citrate pH:6.0 buffer for 3 min, sections were incubated in DAKO autostainer with the monoclonal mouse VEGF-A antibody clone VG1 (18-7328, ZYMED Laboratories) diluted 1/50 and stained with stained with EnVision anti mouse/rabbit (K5007, DAKO). Renal clear cell carcinoma served as a positive control. Primary antibody was substituted with mouse IgG1 for negative control.
CAIX After boiling with a pressure cooker in citrate pH:6.0 buffer for 3 min, sections were incubated with the polyclonal rabbit CA-IX antibody (NB100-417, NOVUS Biologicals) diluted 1/1500 and stained with En-Vision anti mouse/rabbit (K5007, DAKO). Renal clear cell carcinoma served as a positive control. Primary antibody was substituted with nonimmune rabbit immunoglobulin (DAKO) for negative control. Visualization of the primary antibody was achieved using diaminodenzine as chromogen and section were lightly counterstained with hematoxylin.

Quantification
The percentage and intensity of positively nuclear and intensity of cytoplasmic staining in tumor cells were evaluated. HIF1α expression was assessed in tumor cells using a modified previously published semiquantitative scoring [9]. The immunohistochemical results for HIF1α were classified as follow for nuclear and cytoplasmic percentage staining: 0, no staining; 1, less than 1% of cells; 2, 1-10%; 3, 10-50%; 4, more than 50%; for nuclear and cytoplasmic intensity staining [8]: 0, no staining; 1, weak staining; 2, moderate staining; 3, strong staining. The percentage and intensity nuclear and cytoplasmic intensity scores were added together to give a final immunoreactive score (IRS) of 0 to 10. HIF1α was categorized as low HIF1α = ≤ 50% cells staining and high HIF1α = > 50% cells staining. VEGF-A expression was assessed in tumor cells using a previously published semiquantitative scoring in prostate tissue [10]. The percentage of positively tumor cells was evaluated and the VEGF-A staining intensity was assessed. The percentage and intensity scores were added together to give a final immunoreactive score (IRS) of 0 to 8. VEGF-A IRS scores were categorized as low VEGF-A = IRS score < 5, high VEGF-A = IRS score >5.
EGFR expression was assessed in tumor cells using a previously published semiquantitative scoring in prostate tissue [11]. EGFR expression was assessed in tumor cells and only membranous EGFR staining was considered. The percentage of positive tumor cells was estimated as follow: 0, no membranous staining, 1, <30% of cells; 2, 30-50% of cells; 3, >50% of cells. The staining intensity was scored as follow: 0, no staining; 1, weak staining; 2, moderate staining; 3, strong staining. Tumor were subsequently categorized as negative (no membranous staining), strongly positive (>50% with moderate intensity or >30% of cells with strong intensity) or weakly positive if not reaching the criteria selected above.
CA-IX expression was assessed in tumor cells using a previously published score [12]. Only membranous CA-IX staining was considered. Immunostaining of >10% of tumor cells was necessary to be positive. Diagnostic biopsies were assessed by a two prostatecancer histopathologist blinded to patient outcome.

Statistical considerations
Biochemical progression-free survival (bPFS), cancerspecific (CSS) and overall survival (OS) were calculated from the date of RT using Kaplan-Meier estimates. The events were death (all causes of death included) for OS, death from prostate cancer for CSS and biochemical PSA failure or death for bPFS. Biochemical failure was defined by use of the Houston criteria. Patients free from biochemical failure were censored on the date of their last PSA test. Proportions were compared using the Chisquare test for values > 5 and Fisher's exact test for values ≤ 5. Differences between groups were assessed using the log-rank test. The log-rank test was used to compare different survival functions according to the HIF1α, EGFR and VEGF-A expression. Multivariate cause-specific Cox models that accounted for competing risks were fit separately for prostate cancer patients. The proportional hazards assumption was tested using scaled Schoenfeld residuals, with visual inspection of the log minus log plots. Predictors included PSA, Gleason, age, HIF1α, EGFR and VEGF-A expression. All statistical tests were two sided, with alpha levels lower than .05 considered statistically significant. Univariate analysis of the parameters in relation to biochemical control is detailed in Table 3. High expression of HIF1α was associated with a significant increase in bPFS (p = 0.019; Table 3 (Table 3). On multivariate analysis, low HIF1α (p = 0.01) and high EGFR (p = 0.04) expression remained significant adverse prognostic factors (Table 4).

Results
ADT administration did not improve the biological outcome of patients with low or high HIF1α expression (Table 5).

Discussion
In our study, approximately three quarters of all prostate tumors strongly expressed HIF1α, and VEGF-A (Table 2), a number that is similar to the figures reported by others in prostate cancer [8]. Unlike locally-advanced prostate cancer, strong EGFR expression was observed in one quarter of the studied patients, as reported by other investigators in early stage prostate cancer [13].
High expression of HIF1α was unexpectedly associated with an improved biochemical survival (Table 3; Figure 1). These findings contrast with published results suggesting that low expression of this transcription factor is associated with better clonogenic survival in breast cancer cell lines under hypoxia [14] or cervical cancer  PSA prostatic-specific antigen, ADT anti-androgen deprivation therapy. treated with RT [15], but published reports on these cancers are not unequivocal [16]. In a head and neck (H&N) series, high expression of HIF1α in 79 surgically treated patients with squamous cell carcinoma was significantly associated with improved disease-free and overall survival in multivariate analysis [17]. Likewise, HIF1α expression was assessed in 85 patients with early stage T 1−2 H&N squamous cell carcinoma treated with surgery alone by IHC on tissue micro arrays [18]. High expression of HIF1α was associated with an improved 5year disease-free and overall survival in multivariate analysis. Of note, the transcription of the HIF1α-subunit is regulated by two synergistic mechanisms. First, PHD enzymes catalyze the hydroxylation of two prolin residues in the oxygen degradation-dependant domain of this subunit. Consequently, HIF1α will be recognized by von Hippel-Lindau protein that will allow degradation by the proteosome. The second mechanism involves Factor Inhibiting Hypoxia-inducible factor 1 (FIH-1). Under normoxic conditions, FIH-1 hydroxylates an aspirigine residue in the C-terminal portion of the two HIF1α isoforms. This modification prevents the interaction of HIF1α C-terminal domain with the transcriptional co-activator p300, thus decreasing HIF1α transcriptional activity and increasing HIF1α expression. Interestingly, nuclear FIH-1 was associated with a favorable biochemical survival in a recent prostatectomy series [19]. These data suggests that HIF1α may be associated with a better biochemical outcome in prostate cancer patients, although the r-value of the FIH-1/HIF1α correlation was not given by the English authors [19]. Noteworthy, HIF1α was not prognostic in prostate cancer patients included in a French dose escalation study [20] and an US prostatectomy series [4] using IHC and gene expression, respectively. The reasons for these discrepant observations are unclear. Notwithstanding the issue of hypoxia and radio-resistance, the potential phenotypic aggressiveness of HIF1α-negative tumor cells has been documented in series [21]. HIF1α is known to play a role in promoting tumor cell's apoptosis. In a series of embryonic stem cells model the proliferation HIF1α +/+ knockout cells was reduced or delayed in hypoxic conditions [21]. Conversely, growth of HIF1α −/− knockout embryonic cells was not retarded but was increased, possibly because of decreased hypoxia-induced apoptosis and increased stress-induced proliferation [21]. It may well be that these HIF1α-negative cells may loses their ability to undergo apoptosis, at a distance from blood vessels, reducing thus their critical dependence on vascular supply. Our results contrast radically with those published by the Royal Marsden group (RMH) [8]. In this study, the diagnostic biopsies of 308 localized prostate cancer patients were entered into two sequential dose-escalation trials (64 Gy vs. 74 Gy) with ADT. The same biochemical failure definition was used in both studies. Patients in the RMH study has somehow more favorable characteristics when compared to those in the present study (T 2 59% vs. 23%; Gleason < 7, 74% vs. 58%; median PSA, 11.5 vs. 13.1 ng/ml). Possible explanation for these contradictory findings may include imbalances between the two cohorts (the biochemical progression rate was 38% vs. 28% in the RMH and present series, respectively) or the immunoreactivity assessment methodology. In the RMH study, HIF1α was assessed in terms of cytoplasmic staining. We found HIF1α-nuclear only expression in our study for this nuclear transcription factor (Table 2). Unlike the RMH staining methodology, we used double IHC staining method, as detailed by Vaughan et al. [22]. Alternatively, more advanced prostate tumors may express differentially HIF1α. HIF1α mRNA gene expression was significantly unregulated in blood samples of localized prostate cancer patients, when compared to individuals with no malignancies or those with more advanced tumors in a recent prospective study [23]. Strong EGFR expression was associated (p = 0.05) with a decrease in bPFS (Table 3; Figure 1). The prognostic relevance of EGFR expression was also observed in a recent Italian series of prostate cancer patients [13]. The observed median time to biochemical failure in this series was 104 and 30 months in EGFR <50% and ≥50% tumors, respectively (HR, 2.5; p = 0.02). EGFR expression may have a role in the development of prostate cancer [24]. EGFR is down regulated at the transcriptional level by androgens in normal prostate tissue but up-regulated in prostate malignancy, especially in androgen-independent prostate cancer. Di Lorenzo et al., reporting on 76 patients with androgen-dependent and -independent prostate cancer, observed 41%, 76% and 100% EGFR expression in radical prostatectomy,  hormone-sensitive and hormone-refractory metastatic patients, respectively [24]. We have observed a significant association between EGFR expression and higher Gleason scores (p = 0.04). These results may also be in keeping with other series [24]. It remains to be demonstrated if EGFR therapeutic targeting may optimize patient outcome [25]. EGFR prognostication needs to be more fully assessed in the framework of prospective studies. The expression of the angiogenic factor VEGF-A, a soluble growth factor acting as a specific endothelial mitogen, and its receptor may be an important factor in the prostate carcinogenesis. In our series, high-VEGF-A expression was usually not associated with biochemical failure (Table 3, Figure 1). The absence of a significant correlation observed in our study may be due to the small sample size, that may have limited the statistical power to detect associations between VEGF-A expression and biochemical outcome, or to the diffuse and multifocal IHC-expression pattern in prostate cancer that may render the quantification of this glycoprotein somewhat difficult [26,27].
CAIX is normally expressed in epithelial cells of the intestines and stomach but may be expressed when tumor cell hypoxia occurs in malignancies. It is expressed in carcinomas derived from cells not expressing this membrane-bound glycoprotein, such as those observed in lung, breast or kidney and may be associated with a negative prognosis in these tumours. Prostate cancer cell line may express CAIX in strong hypoxic conditions. In our series, none of the prostate cancer cells expressed CAIX in diagnostic samples and could thus not be considered strongly hypoxic ( Table 2). The NB100-417 antibody, used in this series, has been associated with false positive but not false negative IHC results. As such, prostate cancer cells in core our histological samples were not hypoxic.
We could not demonstrate an impact on biochemical outcome in patients with unfavorable (i.e. tumors low HIF1α and/or high EGFR immunoreactivity) tumors treated with RT ± ADT (Table 5). Androgen deprivation improves tumor oxygenation and may thus increase the efficacy of RT in patients with unfavorable prognosis. Small patient numbers complicate the analysis of these findings. The number of patients in the low-HIF1α group receiving or not receiving ADT was 10 and 11, respectively (data not shown). In our series, a better 8-year bPFS was observed with low HIF1α immunoreactivity treated with ADT when compared to RT alone. Future efforts should be directed toward the understanding of the role of these parameters in selecting treatment for intermediate-and high-risk prostate cancer patients in the frame of prospective studies.
This study has potential limitations inherent in all retrospective analyses, including uncontrolled patients selection into the different treatment groups. Major limitations of this study include but are not limited to the IHC evaluation in a limited sample of diagnostic tissue that may not reflect the intrapatient heterogeneity of tissue marker expression and the limited overall number of patients. To our knowledge, the present report is however the first to report a positive association between biochemical outcome and high-HIF1α immunoreactivity in intermediate-and high-risk prostate cancer patients treated with RT.

Conclusions
In summary, HIF1α and VEGF-A was frequently expressed in prostate cancer cells. HIF1α possibly non-hypoxia related expression in diagnostic biopsies was associated with an improved biochemical survival. EGFR immunoreactivity was associated with poor outcome.

Competing interests
The authors declare that they have no competing interests.
Author's contribution DCW and RM were responsible for the primary concept and the design of the study; DCW and JCT, performed the data capture and analysis. DCW drafted the manuscript; DCW and CC performed the statistical analysis; DCW and JCT reviewed patient data; all authors revised the manuscript. All authors have read and approved the final manuscript.