Negative LC3b immunoreactivity in cancer cells is an independent prognostic predictor of prostate cancer specific death

Background Autophagy is a catabolic cellular process used for degradation of cytoplasmic organelles and preservation of cell viability. In this study we aimed to analyse the level of autophagy markers in benign and malignant prostate tissue and to evaluate the prognostic properties for patients with prostate cancer (PCa). Results LC3b expression was significantly upregulated in PCa, especially in metastatic and castration-resistant PCa samples compared to benign prostate tissue (p<0.001). Evaluation of expression in malignant radical prostatectomy specimens revealed an inverse association with preoperative serum PSA levels (p=0.02) and Gleason Score (p=0.07). LC3b immunoreactivity was identified as a novel predictor of PCa specific death after radical prostatectomy, independent of Gleason score, tumour stage, and surgical margin status in a multivariable cox regression analysis (hazard ratio 0.09, 95% confidence interval 0.01-0.69, p=0.021). A significant association of ATG-5 and Beclin 1 with LC3b expression could be noticed (p<0.001), but no link with other clincopathologic parameters was observed. Materials and Methods A Tissue microarray containing 468 formalin-fixed, paraffin-embedded prostate tissue cores was stained immunohistochemically for major autophagy proteins LC3b, ATG5 and Beclin 1. Immunoreactivity was semiquantitatively scored and correlated with pathologic and clinical parameters, including tumour stage, Gleason score, preoperative PSA level, biochemical recurrence rate and survival. The median clinical follow-up was 132 months. Conclusion LC3b was significantly overexpressed in malignant compared to benign prostate tissue. However, positive LC3b immunoreactivity in PCa, as a marker of increased autophagy, was independently associated with a reduced disease-specific mortality.


INTRODUCTION
When prostate cancer (PCa) is localized in the prostate and considered significant, the treatment of choice is prostatectomy or radiation [1]. In advanced and metastatic PCa androgen ablation is the standard of care to palliate symptoms and postpone cancer related complications [2]. However, after short-term remission, surviving tumour cells reappear as castration-resistant PCa (CRPC) and lead to death within months or few years [3]. Therefore, extensive efforts are made to offer new therapeutic options to patients with CRPC including development of new anti-androgens and combination therapies. One promising approach has been the pharmacological inhibition of autophagy combined with androgen ablation or novel antitumour agents [4]. Autophagy is a catabolic cellular mechanism involving degradation of unnecessary or dysfunctional cellular components through autophagosomes allowing maintenance of homeostasis and ensuring cell survival under stress conditions. [5] The process of autophagy is regulated by autophagy-related proteins (ATGs) [6]. The key elements necessary for autophagosome formation are two ubiquitin like conjugation systems: (1) the ATG12-ATG5 and (2) the microtubule-associated protein 1 light chain 3 (LC3)-phosphatidylethanolamine systems [7]. In mammalian cells, three types of LC3 were reported; A, B and C, with LC3b expression being the most valid marker of autophagosome formation [8] and therefore one of the most widely used in situ techniques of autophagy measurement in benign and malignant tissue [9].
Whether autophagy acts as a promoter or a suppressor during tumorigenesis seems to be context and organ-specific. Improvement of the cytotoxic effect of chemotherapy drugs or resensitization of chemoresistent tumour cells through inhibition of the protective role of autophagy has been reported in several studies [10]. Particularly in PCa, targeting autophagy is a promising new therapeutic modality [11][12][13][14][15][16][17]. However, the relevance of autophagy-related protein expression in treatment-naive PCa is largely unknown.
In the current study, we analyse the expression profile of the three main autophagy markers involved in autophagosome formation, LC3b, ATG5 and Beclin 1 in benign prostate tissue, localized PCa, CRPC and PCa metastases, respectively. Autophagy marker expression is further correlated with clinicopathological parameters including biochemical recurrence free survival (BCRFS), overall-(OS) and disease-specific survival (DSS).
Regarding all tissue samples, positive LC3b immunostaining was highly associated with positive Beclin 1 and ATG5 immunostaining (p<0.001, Table 2). Negative immunoreactivity of Beclin 1 and ATG5 was associated in 89% and 96% of the cases with a lack of LC3b staining, respectively (p<0.001). However, a positive immunostaining for Beclin 1 and ATG5 was related to a LC3b positive staining in only 41% and 47% of the cases, respectively (p<0.001).
Clinicopathologic characteristics of RP patients were correlated with LC3b, Beclin 1 and ATG5 expression (

DISCUSSION
This is the first study investigating expression levels of the established markers of autophagy LC3b, Beclin 1 and ATG5 and their association with the solid clinical endpoint of survival in a representative cohort of PCa patients. LC3b, a major marker of autophagy, was significantly upregulated in adenocarcinomas of the prostate, in metastatic and in CRPC. Within the group of patients with localized PCa, who underwent radical prostatectomy, the absence of LC3b immunoreactivity was identified as a novel predictor of PCa-specific death after radical prostatectomy, which was independent of the well-established predictive factors Gleason score, tumour stage, and surgical margin status.
Autophagy not only plays a central role in maintenance of cellular homeostasis, but also is essential in the process of malignant transformation [18]. The pro-survival and pro-death role of autophagy can differ, depending on several different factors such as tissue type, stimulating factors and cellular environment [19]. Accumulating evidence support the thesis, that autophagy is a tumour suppressor and inhibits tumour development by detaching damaged organelles and promoting cell death of cancerous cells [20]. Particularly, downregulation of ATG5 contributed to tumorigenesis of earlystage cutaneous melanoma in a study by Liu et al. [21].    Deletions of Beclin 1 have been described in specimens of human breast, ovarian and prostate tumours [22,23].
On the other hand, other studies reported that increased autophagy can support cancer development by maintaining the stability of intracellular environment and acts as a protective mechanism against apoptosis and external death stimuli including anti-cancer drugs [24]. Evidence for a role of both reduced and increased autophagy in cancer cells was also found in the present study; while higher expression of autophagy proteins were observed in malignant tissue, lack of autophagy proteins was associated with worse clinicopathological parameters and with an increased tumour specific mortality. Our findings are in contrast with a recently published report investigating the immunohistochemical expression of LC3b and Beclin 1 in 96 PCa specimens [25]. Giatromanolaki et al. found a significant association of high Gleason scores and high tumour stages with high LC3b and Beclin 1 expression levels. However, this report exclusively included node-negative PCa of which 68% had a Gleason score of ≤6. Only 31 specimens with a Gleason score ≥7 and 19 with an extracapsular extension were evaluated compared to 214 with Gleason score ≥7 (including 51 Gleason 8-10) and 81 with a tumour stage ≥pT3 in the present study. Furthermore, due to the pooling of moderately differentiated tumours of Gleason 7 with the poorly differentiated of Gleason 8-10 it remains unclear, how many genuine high risk tumours were stained in this study cohort. No clinical data regarding preoperative PSA values, patterns of biochemical recurrences or survival Note: Bold face reprensenting P < 0.05. Liu et al. just recently evaluated the prognostic impact of 5 autophagy markers on the clinical endpoint of BCR after radical prostatectomy and reported no association for LC3b in agreement with our results. Yet, an association of positive UNC-51-like kinase 1 (ULK1) reactivity with BCR was reported after a relatively short median follow-up of 51 months [26]. In the present study, no association of autophagy with BCR rates could be observed. However, after a median follow-up of over 10 years we could show that a positive LC3b immunoreactivity had a positive and distinct impact on overall and disease-specific survival. This impact remained significant in a multivariable analysis with a HR of 0.09 (95% CI 0.01-0.69) for dying of PCa if the staining for LC3b showed an expression at the time of prostatectomy. LC3b immunoreactivity was an even more powerful predictor of PCa specific death than the well-established parameters Gleason score and tumour stage. However, the low number of prostate cancer specific deaths lead to a wide 95% CI in our model with a significant influence on the point estimator (0.09). Considering HRs of other biomarkers, the real HR is expected to be slightly higher (95% CI 0.01-0.69) without having an impact on the conclusions drawn from our analysis.
Patients who had received ADT prior to surgery were excluded from the primary analysis in order to minimize the confounding effect of a neoadjuvant treatment on protein expression. However, this population can be used to provide an insight on the effect of ADT on autophagy. Recently, several studies have investigated the role of autophagy as an escape mechanism of PCa exposed to antitumour agents, hormonal ablation and radiation therapy [4,27,28]. An inhibition of autophagy was able to increase the antitumour effect of these agents [12,14]. Of note, inhibition of autophagy in androgen sensitive cell lines (e.g. LNCaP) under hormonal ablation therapy showed reduced cell viability, suggesting the protective role of autophagy in androgen withdrawal [16,17]. We, therefore, expected an up-regulation of autophagy in hormonal naïve patients treated with androgen withdrawal prior to surgery. Our findings demonstrated the opposite; autophagy was significantly reduced in our cohort of patients (11.5% vs. 36.1%, p<0.001) compared to the control group. However, preoperative ADT was mainly given to patients with high-risk disease and, therefore, this subgroup of patients differed in its perioperative parameters significantly from the control group.
Based on the methods of this investigation, no conclusions can be drawn regarding the causal relation of these findings. However, our results provide a new insight into the role of autophagy in PCa patients and indicate that the regulation of autophagy in PCa depends on the stage of the tumour. Increased autophagy levels may be a sign of starvation and metabolic stress in well differentiated tumours due to poorer selective advantage. In more aggressive tumours, cells may have activated other mechanisms to deal with cellular stress and nutrition deficiency making them independent from autophagy proteins. Therefore, low or no autophagy might be a sign of malignant cells not struggling with survival. Further investigations will be necessary evaluating the actual functional role of autophagy in different stages of PCa.

Patients and specimen characteristics
Tissue microarrays (TMA) contained 468 formalin-fixed, paraffin-embedded prostate tissues and were constructed as previously described [29]. Specimens were collected between 1993 and 2007 from the Institute of Surgical Pathology, University Hospital of Zurich, Switzerland. The TMA included a series of 348 consecutive (non-selected) malignant radical prostatectomy (RP) specimens, 29 CRPC samples, 18 lymph node metastases, 28 distant metastases (bone, lung, urinary bladder), and 45 benign prostatic hyperplasia samples. A neoadjuvant androgen deprivation therapy (NADT) had been established preoperatively in 56 men with primary PCa. Haematoxylin and eosin-stained slides of all specimens were re-evaluated by two experienced pathologists (P.J.W., H.M.) to identify representative areas. Tumour stage and Gleason score of the Zurich cohort were assigned according to the International Union Against Cancer and World Health Organization/ International Society of Urological Pathology criteria [30]. In total, clinical follow-up data were available for 317 of 348 prostatectomy patients (91.1%). After RP, follow-up of patients was conducted by periodic measurement of the serum PSA. Median follow-up was 132 months (range 1-252). The study was approved by the local scientific ethics committees (KEK-ZH-No.2008-0025).  The specificity of the commercial antibodies has been thoroughly validated in former studies [31][32][33][34].

Antibody specificity test
The specificity of the anti-LC3b (rabbit polyclonal; Abcam plc; no. ab48394, dilution 1:200) antibody was tested using LnCaP prostate cancer cells (American Type Culture Collection, Manassas, VA). LnCaP cells were cultured in the presence of pharmacological modulator of autophagy, 5 mM 3-MA (Sigma-Aldrich, Buchs, Switzerland), an inhibitor commonly used to block autophagy [35] and 2 µM rapamycin (LuBioScience, Luzern, Switzerland) an inducer of autophagy for 7 days on the slide flasks (Nunc flasks, cat: 170920). Cells were immunostained with anti-LC3b (1:200) for 24h at 4°C. The slides were incubated with secondary antibody Cy3conjugated sheep anti-rabbit antibody (Sigma, 1:200) at room temperature for 1 h. The slides were counter-stained with DAPI (4',6-diamidino-2-phenylindole, Sigma, 1:200) and analysed with a Leica fluorescence microscope (40x). The specificity of the anti-LC3b antibody could be confirmed by detecting no staining in the negative control, a minimal staining in cells inhibited by 3MA, a modest staining in untreated cells (basal autophagy level) and the typical LC3b punctuation in cells induced by rapamycin (Supplementary Figure 1).

Immunohistochemical readout
One surgical pathologist (N.J.R.) and one urologist experienced in TMA readouts (A.M.) performed a blinded evaluation of the slides without knowledge of clinical data. In case of a discrepancy the core was assessed by a senior pathologist (P.J.W.). Causes of noninterpretable results included lack of target tissue, presence of necrosis, or crush artefact. Since autophagy proteins showed a homogeneous expression pattern, staining intensity (exclusively cytoplasmic staining pattern was evaluated) was assigned using a semiquantitive, four-tired score: negative (0), weak (1+), moderate (2+) or strong (3+). In cases with heterogeneity, the predominant staining intensity (>80%) was counted. Searching for cut-offs in an unbiased way is a major problem in immunohistochemical studies dealing with a continuous readout. The median LC3b immunoreactivity in prostatectomy cases with malignant tissue (weak) was chosen as cut-off. Accordingly, positive LC3b, ATG5 and Beclin 1 immunoreactivity was defined as moderate or strong staining of target cells.

Statistical analyses
SPSS version 22.0 (IBM Corporation, Armonk, NY, USA) was used for statistical analyses. To study statistical associations between clinicopathologic and immunohistochemical data, contingency table analysis with 2-sided Fisher's exact tests were used for categorical and the student t-test for continuous variables. Descriptive statistical analyses for patients who had received any hormonal manipulation preoperatively (NADT) were performed separately. Only patients who were treated by radical prostatectomy were considered for survival analysis. The outcome measures were biochemical recurrence-free survival (BCRFS), overall survival (OS) and disease-specific survival (DSS). Biochemical recurrence (BCR) was defined as PSA value ≥0.1 ng/ mL with subsequent confirmation after reaching the PSA nadir of 0.1 ng/ml postoperatively. Patients not reaching this nadir threshold postoperatively were excluded from BCRFS analysis. Univariate Cox regression analysis was performed for BCRFS and DSS. Investigated variables were age, Gleason score, preoperative PSA level, pathologic tumour stage, surgical margin status, immunoreactivity for LC3b, ATG5 and Beclin 1. A stepwise multivariable Cox regression model was adjusted, testing the independent prognostic relevance of LC3b immunoreactivity for DSS. BCRFS, OS and DSS curves were calculated using the Kaplan-Meier method with significance evaluated by 2-sided log-rank statistics. All p-values <0.05 were considered statistically significant.