Predictive value of combined analysis of pro-NPY and ERG in localized prostate cancer

This study aimed to investigate if combined analysis of pro‐Neuropeptide Y (NPY) and ERG expression in tumor tissue are associated with biochemical failure (BF), castration‐based treatment, castration‐resistant prostate cancer (CRPC), and prostate cancer (PCa)‐specific death for men undergoing radical prostatectomy (RP) for PCa. This study included 315 patients, who underwent RP from 2002 to 2005. Both pro‐NPY and ERG expression were analyzed using immunohistochemistry and were scored as low or high and negative or positive, respectively. Risk of BF, castration‐based treatment, CRPC, and PCa‐specific death were analyzed with multiple cause‐specific Cox regression analyses and stratified cumulative incidences using competing risk assessment. Median follow‐up was 13.0 years (95% CI: 12.7–13.2). In total, 85.7% were pro‐NPY high and 14.3% were pro‐NPY low. The combined analyses of pro‐NPY and ERG expression was not associated with risk of BF (p = 0.7), castration‐based treatment (p = 0.8), CRPC (p = 0.4) or PCa‐specific death (p = 0.5). In the multiple cause‐specific Cox regression analysis, pro‐NPY high and ERG positivity was not associated with BF (HR: 1.02; 95% CI 0.6–1.7; p = 0.94). In conclusion the combination of pro‐NPY and ERG expression did not show association with risk of BF, castration‐based treatment, CRPC, and PCa‐specific death following RP.

prostate cancer; predictive biomarkers; pro‐NPY; ERG; radical prostatectomy

Prostate cancer (PCa) has a distinct tumor biology ranging from an indolent to an aggressive course [1] [2] [3] . Due to the increased awareness of the disease and the possibility of early diagnosis by prostate‐specific antigen (PSA) measurement, most men are currently diagnosed with localized PCa, suitable for curatively intended treatment. However, early detection and treatment can lead to overtreatment of many patients with indolent tumors [4] [5] . In men with localized PCa, treatment strategies rely on a combination of clinical, histological, and biochemical parameters [6] [7] [8] . Applying biomarkers reflecting tumor biology might refine prognostication and reduce the risk of overtreatment of men with localized PCa.

Gene fusions are known to be important in the initial steps of PCa tumorigenesis, and approximately 50% of men with PCa carry the gene fusion of TMPRSS2 and ERG leading to ERG protein expression. Several studies have indicated distinct molecular mechanisms in ERG negative vs ERG positive tumors, and ERG expression in combination with other biomarkers might be a good predictor for PCa aggressiveness [9] [10] [11] . One such biomarker is the peptidic neurotransmitter neuropeptide Y (NPY) [12] . Recent studies has focused on NPY as an important factor for mitogenic and angiogenic activity [13] , and it has previously been shown that NPY and its precursor pro‐NPY are overexpressed in high‐risk PCa [12] [14] . In a recent study, we demonstrated an association between high pro‐NPY expression and PCa‐specific death in patients with low‐risk cancers managed on watchful waiting [12] . Moreover, the combination of ERG and pro‐NPY expression had independent prognostic value in terms of PCa‐specific death.

Whether pro‐NPY alone or in combination with ERG could be applied as a prognostic biomarker for PCa‐related outcomes following radical prostatectomy (RP) is to the best of our knowledge not known. The aim of the present study was to analyze if a combined analysis of pro‐NPY and ERG expression in tumor tissue together with standard clinicopathological variables has predictive value in terms of risk of biochemical failure (BF), castration‐based treatment, castration‐resistant prostate cancer (CRPC), and PCa‐specific death for men undergoing RP.

The study included a consecutive series of men (n = 336), with clinically localized PCa who underwent RP with curative intent from January 1st 2002 until December 31st 2005 at the Department of Urology, Rigshospitalet, Copenhagen, Denmark (Fig. A). In 2016, a TMA was generated using formalin‐fixed paraffin‐embedded tissue samples. For each patient, two experienced pathologists identified representative malignant and non‐malignant tissue areas and reclassified the RP Gleason score (GS) according to the International Society of Urological Pathology (ISUP) 2005 Gleason grading system [15] . Seventeen patients were excluded from the study as there was not enough malignant tissue for TMA construction. Two 1‐mm cores from each of the representative areas were mounted on a total of 44 TMA blocks using an ATA‐27 automated tissue microarrayer (Beecher Instruments, Sun Prairie, Wisconsin, USA). The study is approved by the Danish National Committee on Health Research Ethics for the Capital Region (Journal no.: H‐6‐2014‐111) and The Danish Data Protection Agency (file#2006‐1‐6256).

Clinical and pathological information was collected from medical records. Data included preoperative information and pathological findings following RP as previously described in detail [16] . Data on BF, development of distant metastases, CRPC, date and cause of death, and PCa‐specific treatment were updated in July 2017. BF was defined as PSA ≥ 0.2 μg/L. Four patients had missing follow‐up PSA values (none Danish citizens) and were excluded from the analyses of BF. CRPC was defined as S‐testosterone in castration level (<1.7 nmol/L) and either biochemical progression (PSA increase > 50% in two measurements) or radiological progression (two or more new lesions) [17] . In case of missing data, CRPC was defined as start of any treatment approved for CRPC. Seven patients had missing follow‐up data regarding start of castration based‐treatment, time of CRPC, or time of death and were excluded from these analyses.

Freshly cut 2.5 μm sections of each TMA block were used for immunohistochemistry (IHC) staining for pro‐NPY (1:200 dilution; anti‐pro‐NPY (HPA044572) rabbit polyclonal antibody, Atlas Antibodies) and ERG (anti‐ERG (EPR3864) rabbit monoclonal primary antibody, Roche Ventana). Positive controls for pro‐NPY and ERG staining consisted of sections of brain tissue and colon, liver, and tonsil tissue, respectively. Moreover, nerve bundles and endothelial cells were used as internal controls for pro‐NPY and ERG staining, respectively. Separate HE‐stained TMA sections were validated with regard to the presence of cancer in each core by two expert uropathologists (BGT + JE).

All stained slides were digitalized using the Hamamatsu Nano Zoomer XR at a magnification equivalent to 20x. The images were evaluated using the Hamamatsu NDP.view v2.6.13 viewing software. After agreement of the scoring of pro‐NPY and ERG with an expert uropathologist (BGT), whole‐field inspection of pro‐NPY and ERG immunoreactivity for each core on the TMA was assessed by one observer (GK). Four patients had no malignant tissue available for analyses due to progressive slicing and were excluded (Fig. A).

For each patient, cytoplasmatic pro‐NPY immunoreactivity in malignant prostate epithelial cells was evaluated and scored for the strongest immunoreactivity present and categorized as negative, weak, moderate, or strong (Fig. B). Nuclear ERG immunoreactivity in malignant prostate epithelial cells was evaluated and scored as negative if all cores were negative, and positive if any of the cores showed positive immunoreactivity (Fig. C). Assessment of the IHC staining was blinded to study end points.

Associations between pro‐NPY expression and clinicopathological variables were analyzed using χ2‐test or Fisher's exact test for categorical variables and the Mann–Whitney U test for continuous variables. The median time of follow‐up was calculated using the reverse Kaplan–Meier method [18] . Follow‐up for BF was calculated until the latest PSA measurement, whereas time to castration‐based treatment, CRPC, and death was calculated until the latest follow‐up date.

Cumulative incidences of study end‐points were analyzed using the Aalen–Johansen method for competing risks. Death before BF, start of castration‐based treatment, or CRPC was treated as competing events when analyzing risk of BF, castration‐based treatment, and CRPC, respectively. Moreover, other cause mortality was treated as a competing event when analyzing risk of PCa‐specific death. Gray's test was used to assess differences in the cumulative incidences between biomarker subgroups [19] .

Univariate and multivariate cause‐specific Cox proportional hazard regression models were performed for risk of BF, castration‐based treatment, CRPC, and PCa‐specific death, with results presented as hazard ratios (HR) and 95% confidence intervals (CI). The analyses included age at RP, log2‐transformed PSA, pathological tumor (pT)‐stage, N‐stage, RP GS, margin status, ERG, and pro‐NPY expression.

All tests were two‐sided and p < 0.05 was considered to be statistically significant. All statistical analyses were performed using SPSS (software version 22; IBM) or R (R Development Core Team, Vienna, Austria).

A total of 315 patients had malignant tissue available for analysis of pro‐NPY and ERG expression (Table ). The median number of malignant cores was 6 (Inter quartile range [IQR]: 5–8) per patient. The median PSA was 10.0 μg/L (IQR: 6.8–15.0). A total of 64.4% had pT2 PCa, and 35.6% had pT3 PCa. Most patients (91.1%) had RP GS ≤ 7 (Table ). The median time of PSA follow‐up after RP was 11.6 years (95% CI: 11.1–12.2), while the median time of follow‐up regarding castration‐based treatment, CRPC, and PCa‐specific death was 13.0 years (95% CI: 12.7–13.2).

Baseline characteristic

1 IQR, interquartile range; PPB, percent positive biopsies; PSA, prostate‐specific antigen. *Mann–Whitney U test, **Fisher's exact test, ***Chi‐square test.

IHC analyses of malignant tissue revealed that 270 patients (85.7%) had strong or intermediate pro‐NPY immunoreactivity (pro‐NPY high) while 45 patients (14.3%) had weak or negative pro‐NPY immunoreactivity (pro‐NPY low). Furthermore, 195 (62.7%) were ERG positive and 116 (37.3%) were ERG negative (Table ).We found a significant association between pro‐NPY expression and RP GS (p < 0.0001), and ERG expression (p < 0.0001). There was no statistically significant relation between pro‐NPY expression and the other clinicopathological variables (Table ).

The 10‐year cumulative incidence of BF was 48.7% (95% CI: 33.7–63.7) in the pro‐NPY low group compared to 44.1% (95% CI: 37.9–50.4) in the pro‐NPY high group (p = 0.48) (Fig. A). High pro‐NPY expression was not found associated with BF in either the univariate or the multivariate cause‐specific Cox regression analyses. No association between ERG expression and time to BF (p = 0.95) was found (Fig. B).

Addition of ERG expression did not add any predictive value in terms of predicting BF. The 10‐year cumulative incidence for BF was 48.6% (95% CI: 33.2–64.0) in the ERG negative and pro‐NPY low group compared to 41.8% (95% CI: 34.6–49.0) in the ERG positive and pro‐NPY high group (p = 0.7) (Fig. C). In the univariate and multivariate cause‐specific Cox regression analyses the combination of ERG positivity and pro‐NPY high expression was not found associated with BF (Table ). Stratification on pT‐stage (pT2 vs pT3) and RP GS (<6 vs 3 + 4, vs >4 + 3) did not change the predictive value of ERG and pro‐NPY expression (data not shown). Furthermore, none of the combinations of ERG and pro‐NPY expression was found associated with a risk of BF.

Uni‐ and multivariate cause‐specific Cox proportional hazard of biochemical failure

2 CI, Confidence interval; HR, hazard ratio; PSA, prostate‐specific antigen; REF, reference; RP, radical prostatectomy.

Overall, the 10‐year cumulative incidences of castration‐based treatment, CRPC, and PCa‐specific death were 10.7%, 6.4%, and 3.0%, respectively. No association between pro‐NPY expression and time to start of castration‐based treatment (p = 0.65), CRPC (p = 0.25), and PCa‐specific death (p = 0.31) was found.

The 10‐year cumulative incidence of PCa‐specific death was 10.5% (95% CI: 0.8–20.3) in the ERG negative and pro‐NPY low group compared to 2.2% (95% CI: 0.1–4.3) in the ERG positive and pro‐NPY high group (p = 0.53) (Fig. F). The combination of ERG positivity and high pro‐NPY expression was not found associated with PCa‐specific death (HR: 0.5; 95% CI: 0.2–1.4; p = 0.18). Furthermore, we did not find a difference in time to start of castration‐based treatment or risk of CRPC between any of the biomarker subgroups (Tables and , Fig. D,E). Moreover, none of the combinations of ERG and pro‐NPY immunoreactivity was found associated with a risk of castration‐based treatment, CRPC, or PCa‐specific death.

Univariate cause‐specific Cox proportional hazard of castration‐based treatment

3 CI, Confidence interval; HR, hazard ratio; PSA, prostate‐specific antigen; REF, reference; RP, radical prostatectomy.

Univariate cause‐specific Cox proportional hazard of castration‐resistant prostate cancer

4 CI, Confidence interval; HR, hazard ratio; PSA, prostate‐specific antigen; REF, reference; RP, radical prostatectomy.

Neuropeptide Y is known to be important in cardiovascular regulation, control of food intake, and regulatory activities of the neuroendocrine axis [20] , whereas the association between NPY expression and development or progression of cancer remains controversial [13] . NPY is synthesized as pre‐pro‐NPY, which is first cleaved to pro‐NPY, and then cleaved into the two chains: the mature NPY and C‐terminal flanking peptide of neuropeptide Y (CPON) [12] . NPY and pro‐NPY have been found associated with development of different endocrine‐related cancers [21] , including neural crest‐derived tumors [22] , breast cancer [23] [24] , ovarian cancer [25] and PCa [12] [14] [26] . Even though studies have failed to observe significant changes in proliferation of PCa cells in response to NPY stimulation [27] [28] , it has been suggested that NPY regulate the tumor microenvironment and have a systemic effect on androgen levels [21] [22] [29] . Furthermore, proteome profiling of tumor tissue and plasma from PCa patients has identified pro‐NPY as a potential biomarker for PCa aggressiveness [12] [14] .

To the best of our knowledge, the present study is the largest and most comprehensive analysis of pro‐NPY as a predictive biomarker for BF, clinically progression, and PCa‐specific death following RP. We examined whether pro‐NPY expression was associated with other known predictive parameters and found an association with RP GS and ERG expression, respectively. However, pro‐NPY alone and in combination with ERG expression was not predictive for BF, castration‐based treatment, CRPC, or PCa‐specific death. Nonetheless, we found that the 10‐year incidence of PCa specific death was increased by 4.8‐fold in the ERG negative and pro‐NPY low group when compared with the ERG positive and pro‐NPY high group, even though it was not statistically significant. Therefore, it must be speculated if the study is underpowered to show a real difference.

In a previous study, we demonstrated that 32–40% of the patients in two historical watchful waiting cohorts including a total of 752 patients had a high pro‐NPY expression [12] . Furthermore, we found that high pro‐NPY expression was significantly associated with PCa‐specific death in patients harboring low GS tumors, and patients with high pro‐NPY and positive ERG expression had a significantly increased risk of PCa‐specific death independent of GS. In a cohort of 122 patients, we found that high pro‐NPY was not associated with increased risk of BF following RP. The large difference in prevalence of pro‐NPY expression between the watchful waiting and RP cohorts may be explained by patient selection and of the included tissue, as the intratumoral heterogeneity can hamper the reliability of the biomarker. The fact that pro‐NPY has a prognostic value in patients managed by watchful waiting and no predictive value in patients following curative treatment may be due to removal of the aggressive component by RP. In a previous study including 190 patients treated with RP for clinically localized PCa and with a median follow‐up time of 81 months, mature NPY immunoreactivity was found to have independent predictive value for BF [26] indicating that NPY is a better predictive biomarker for BF than pro‐NPY. However, these results have yet to be validated in an independent cohort. In addition, that study only included a small number of patients with a short follow‐up time.

Several studies have previously shown an association between ERG expression and PCa‐specific death for patients followed on watchful waiting [30] [31] [32] and disease progression for patients on active surveillance [33] . In contrast, there seems to be no association between ERG expression and BF or PCa‐specific death following RP [34] . In accordance with the literature, we found no association between ERG expression and BF, castration‐based treatment, CRPC, and PCa‐specific death.

Nonetheless, it can be speculated that some biomarkers have prognostic value in patients managed by conservative treatment while the same biomarker has no or only limited predictive value following curative intended treatment. Therefore, it must be hypothesized if pro‐NPY in combination with ERG expression could have a prognostic role in men followed on active surveillance. Furthermore, as pro‐NPY is associated with regulatory activities of the neuroendocrine axis, one could speculate that pro‐NPY holds predictive value for patients with neuroendocrine tumors in the prostate.

Despite research and the unmet clinical need, there are currently no IHC‐based predictive tissue biomarkers in daily clinical practice. One plausible reason is the missing consensus of validation and standardization of the IHC technique and selection of antibodies. Furthermore, quantitative of IHC‐stained sections can be difficult and is known to be associated with interobserver variation, which limits reproducibility [35] . Another issue is the lack of validation studies which is needed to confirm the performance of the biomarker across study populations [36] .

There are some limitations to the present study, besides those related to its retrospective nature. First, the negative findings may partly be the result of a selected patient cohort with predominantly low‐grade PCa with low risk of progression; however, these patients are comparable to most contemporary RP cohorts. Second, we used ERG immunoreactivity as a proxy for ERG gene fusion status, but previous studies have shown that this is consistent with true ERG rearrangement status in PCa [37] [38] . Third, only a small fraction of the tumor tissue in the RP specimen was evaluated for ERG and pro‐NPY expression, and intratumoral heterogeneity of PCa and sampling bias during the TMA construction can hamper the reliability of biomarker assays in TMAs. However, two consecutive sections from the TMAs were used for ERG and pro‐NPY expression analyses respectively; thus the protein expression was examined in the same tumor cells. Finally, we only had a small number of patients who progress to CRPC or died from PCa. Strengths of the present study are that this is the first large‐scale evaluation of pro‐NPY as a predictive biomarker following RP and the study cohort is a consecutive series of men who underwent RP with curative intent at a single institution with long follow‐up and few lost‐to‐follow‐up. Further, although the cohort is historical, the indication for RP has not changed and the RP GS was updated by two experienced pathologists according to the ISUP 2005 Gleason grading system [15] . Moreover, we analyzed BF as well as more clinically relevant endpoints such as metastatic progression and PCa‐specific death, and finally, the pro‐NPY‐derived tryptic peptide correspond to a portion of the protein C‐terminal end, which is proteolytically removed to generate mature NPY [13] [20] [39] . Therefore, in line with our previous paper [12] , we used an antibody that binds to the C‐terminal domain of pro‐NPY, while others have used antibodies for mature NPY [26] .

In conclusion, pro‐NPY expression is frequent in PCa. We found no association between combined pro‐NPY and ERG expression and the risk of BF, castration‐based treatment, CRPC, or PCa‐specific death. Our findings suggest that pro‐NPY alone or in combination with ERG expression does not hold potential as a biomarker of progression following RP.

PHOTO (COLOR): (A) Flow chart. (B) IHC pro‐NPY staining in representative prostate cancer samples showing negative, weak, moderate, and strong immunoreactivity (IR). (C) IHC ERG staining in representative prostate cancer samples showing negative and positive staining.

PHOTO (COLOR): The cumulative incidence of (A–C) biochemical failure (BF), (D) castration‐based treatment, (E) castration‐resistant prostate cancer (CRPC), and (F) prostate cancer (PCa)‐specific death following radical prostatatectomy (RP). Comepeting event are (A–C) death without BF, (D) death without castration, (E) death without CRPC, and (F) death from other causes. Patients are stratified according to (A) pro‐NPY expression, (B) ERG expression, and (C–F) ERG & pro‐NPY expression at RP. The p values for Gray's test are added.