Expression of PD-1 and CTLA-4 Are Negative Prognostic Markers in Renal Cell Carcinoma

Immuno-oncological therapy with checkpoint inhibition (CI) has become a new standard treatment in metastatic renal cell carcinoma (RCC), but the prognostic value of the expression of CI therapy target molecules is still controversial. 342 unselected consecutive RCC tumor samples were analyzed regarding their PD-1, PD-L1, and CTLA-4 expression by immunohistochemistry (IHC). The prognostic values for cancer-specific survival (CSS) and overall survival (OS) were analyzed for those not exposed to CI therapy. The expression of PD-1 in tumor-infiltrating mononuclear cells (TIMC) and PD-L1 in tumor cells was detected in 9.4% and 12.3%, respectively (Immune reactive score (IRS) > 0). Furthermore, PD-L1 expression in TIMC (IRS > 0) and CTLA-4 expression in TIMC (>1% positive cells) was detected in 4.8% and 6.3%. PD-1 expression and CTLA-4 expression were significantly associated with a worse OS and CSS in log rank survival analysis and univariate Cox regression analysis. CTLA-4 expression is a prognostic marker that is independently associated with a worse outcome in multivariate Cox regression analysis in the whole cohort (OS: p = 0.013; CSS: p = 0.048) as well as in a non-metastatic subgroup analysis (OS: p = 0.028; CSS: p = 0.022). Patients with combined CTLA-4 expression and PD-1-expression are at highest risk in OS and CSS. In RCC patients, PD-1 expression in TIMC and CTLA-4 expression in TIMC are associated with a worse OS and CSS. The combination of PD-1 expression in TIMC and CTLA-4 expression in TIMC might identify high risk patients. This is, to our knowledge, the first description of CTLA-4 expression to be a prognostic marker in RCC.


Experimental Section
Consecutive and unselected tumor specimens from 453 patients undergoing radical or partial nephrectomy for RCC between 1998 and 2011 at the Department of Urology and Pediatric Urology at the University Hospital Erlangen were collected. No tissue samples of renal metastases from renal cell carcinomas or from carcinomas of other origin were included. Tumors already metastasized at the time of surgery are referred to as primary metastatic diseases; tumors metastasized during follow-up are referred to as secondary metastatic diseases. Tumors have been reevaluated independently by two experienced pathologists (AH, FE), and the histological subtype was reclassified according to the UICC 2010 TNM tumor staging system. Construction of the tissue microarray (TMA) has been described in detail previously [34]. All patients, beginning from 2008, gave informed consent. For samples before the 2008 Ethic Commission in Erlangen, all patients waived the need for informed individual consent. The study is based on the approvals of the Ethic Commissions of the University Hospital Erlangen (No.3755). The study was carried out according to the latest version of the Declaration of Helsinki and is approved by the institutional ethics committee.
The expression of PD-1, PD-L1, CD3, and CTLA-4 was investigated by immunohistochemistry (IHC) on 3 µm sections from formalin-fixed paraffin-embedded TMA tissue blocks. For CTLA4-staining, we used a mouse monoclonal anti-CTLA4/CD152 antibody (clone BSB-88, BSB2883, dilution 1:50, BioSB, Santa Barbara, CA, USA). The CTLA4/CD152 antibody was validated on a multi tissue TMA containing colon, breast, small intestine, adrenal, uterus, prostate, ovary, liver, tonsil, salivary gland, brain, heart, renal, appendix, skin, nerve, lung, testis, placenta, spleen, pancreas, endometrium, stomach, and parotid tissue. The CTLA4/CD152 antibody demonstrated specific staining in lymphatic tissue. Therefore, lymphatic tissue in a normal tonsil was used as a positive control. Upon evaluation of the CTLA4-stained TMAs, negative control slides without the addition of the primary antibody were added as internal negative controls. For enhancement and visualization, we used an EnVision + System, HRP (Dako, Agilent Technologies GmbH & CoKG, Hamburg, Germany) according to the manufacturer's instructions. PD1, PD-L1, and CD3 stainings were performed on a Ventana Benchmark ULTRA staining system using standard protocols and a mouse monoclonal anti-PD1 antibody (clone NAT105, Ventana 760-4895, ready to use, Cell Marque TM , Merck KGaA, Darmstadt, Germany), a rabbit monoclonal anti-PD-L1antibody (clone28-8, ab205921, 1:200, Abcam plc, Cambridge, UK), and a rabbit monoclonal anti-CD3-antibody (clone SP7, RBG024, 1:150, Zytomed-Systems GmbH, Berlin, Germany), respectively. The secondary reaction was performed using a Ventana ultraView Universal DAB Detection Kit (Ventana Medical Systems, Tucson, AZ, USA) for anti-PD1 and anti-CD3 and a Ventana OptiView DAB IHC Detection Kit (Ventana Medical Systems, Tucson, AZ, USA) for anti PD-L1. All procedures were performed according to the manufacturer's instructions. We used hematoxylin for counterstaining. IHC-staining was semi quantitatively assessed using the immunoreactivity score (IRS) [35]. The intensity of the staining was classified in 4 categories (no staining (0), weak staining (1), moderate staining (2), and strong staining (3)) and multiplied by the categorized proportion of positive cells (no (0), <10% (1), 10-50% (2), 51-80% (3) and >80% (4)). An IRS = 0 was considered as negative, and an IRS > 0 was counted as positive. PD1 staining was analyzed on TIMC. PD-L1 staining on tumor cells and on TIMC was assessed separately. As no cut-off value for CLTA-4 expression in IHC in RCC is defined, and CTLA-4 expression has been described in about 1% of TIMC in RCC [30], we defined a cut off at ≥2% CTLA-4 positive cells in tumor tissue infiltrating TIMC for defining a tumor as CTLA-4 positive. CD3 positive cells were counted in 4 high power fields, and the number of CD3 positive cells per high power field was recorded. Thirty additional controls revealed no PD1-, PD-L1, or CTLA-4 staining in corresponding normal renal tissue distant to RCC. All stained TMAs were assessed independently by two experienced pathologists (AH, FE), both blinded for clinical data. In cases for which the results were inconsistent, the pathologists worked to reach a consensus.
After the exclusion of missing tissue, missing tumors, and incomplete survival data, 342 tumor-representative specimens could be assessed. There were no differences in the clinical and histopathological features between evaluable cases and the entire tumor cohort. Statistical analyses was performed using IBM SPSS Statistics 21 (IBM-Corporation Germany GmbH, Ehningen, Germany). The nonparametric correlation was assessed by the two-sided Spearman Rho test. Survival analysis was done with the log-rank test, and Kaplan-Meier survival curves were drawn. The association of marker expression and survival was assessed by univariate and multivariate Cox regression analysis. Differences were regarded statistically significant at p < 0.05.
Individual participant data that underlie the results reported in this article will be shared after de-identification in the supplementary files. Data will be available immediately following publication for 5 years.

Log Rank Test
A survival analysis showed a significantly longer estimated OS in patients with PD-1 negative TIMC (Table 3)

Log Rank Test
A survival analysis showed a significantly longer estimated OS in patients with PD-1 negative TIMC (Table 3). The estimated mean OS advantage was 109.7 vs. 55.8 months (p = 0.002) (Figure 2). A similar trend was detected in estimated mean cancer specific survival (CSS), although this did not reach statistical significance (142.1 vs. 84.5 months, p = 0.072) (Figure 3)

Multivariate Cox Regression Analysis
After controlling for the univariately significant parameters age, gender, tumor grade, tumor stage, and ECOG performance status, CTLA-4 expression in TIMC remains independently prognostic for OS (HR = 2.8, p = 0.013) and CSS (HR = 3.7, p = 0.048) in the whole cohort (Table 6), as well as in the ccRCC-only subgroup analysis (HR = 4.1, OS p = 0.006; CSS HR = 8.2, p = 0.003) ( Table 7) and in the primary non-metastatic subgroup analysis (OS HR = 3.4, p = 0.028; CSS HR = 7.4, p = 0.022) ( Table 8). PD-1 expression in TIMC alone and in combination with CTLA-4 expression in TIMC is not an independent prognostic factor in multivariate Cox regression analysis.

Discussion
As part of the anti-tumor immune response, naive T cells are activated by the presentation of tumor antigens from dendritic cells and other antigen-presenting cells via Major Histocompatibility Complex (MHC) molecules. In this immunological checkpoint, costimulatory signals for the activation or anergy of the effector T cell are crucial. Two of these immunosuppressive costimulatory pathways are the PD-1/PD-L1 interaction and the activation of CTLA-4 via B7-1 or B7-2. Recently, the combination of checkpoint inhibitors ipilimumab (anti-CTLA-4 antibody) and nivolumab (anti-PD-1 antibody) in the phase III Checkmate 214 study (NCT02210117) in advanced RCC demonstrated the statistically significant improvement of overall response rate (ORR) compared to the standard of care with sunitinib in first line therapy in intermediate and poor risk patients [36]. Though checkpoint inhibitors targeting these pathways are changing therapy in metastatic renal cell carcinoma [12,36], the prognostic value of PD-1, PD-L1, and CTLA-4 expression in localized renal cell carcinoma remains unclear.
PD-1 on the cell surface of activated T cells is immunosuppressive when it is activated in peripheral tissue by tumor cells via PD-L1 or PD-L2 expression. In our analysis, PD-1 expression was detected in TIMC in about 9.4% of our tumor specimens, with significantly higher expression in high grade tumors and primary metastatic diseases. Previous studies on the prognostic value of PD-1 expression in TIMC in ccRCC are contradictory [15,16,37]. In our cohort, we found a significant association with OS and CSS by univariate Cox regression analysis, but, after an adjustment to tumor stage, tumor grade, gender, age, and ECOG performance status, multivariate Cox regression analysis revealed that it is not an independent prognostic factor.
PD-1 expression on TIMC and PD-L1 expression in tumor cells are considered to be markers of reduced T cell function in the tumor microenvironment [14]. An association of PD-L1 expression and a poor prognosis is described in ccRCC as well as in other histological subtypes [17,18,20,21,23,37], and recently published data show an association with a worse OS and progression free survival (PFS) in high risk non metastatic diseases [38]. In our analysis, PD-L1 expression is significantly correlated with high grade diseases and non-metastatic diseases and lower in secondary metastatic diseases. However, no association with OS or CSS could be detected.
CTLA-4 is the second target of checkpoint inhibition therapies in renal cell carcinoma. Antibodies to CTLA-4 were the first checkpoint inhibitors with anti-tumor activity [39,40]. CTLA-4 expression was significantly correlated with primary metastatic diseases and associated with a a reduced OS and CSS in the whole cohort, as well as in a ccRCC-only subgroup. Though associated with primary metastatic diseases, CTLA-4 expression in a non-metastatic disease subgroup is still significantly associated with a worse OS and CSS. Multivariate analysis revealed CTLA-4 expression as an independent prognostic factor after an adjustment to tumor grade, tumor stage, age, gender, and ECOG performance status. This is, to our knowledge, the first description of CTLA-4 expression to be a prognostic marker in RCC.
As there is significant cross correlation, the synchronous expression of PD-1 in TIMC and of CTLA-4 in TIMC identifies patients with a worse outcome. The estimated median OS is more than six years shorter in this subgroup (29.8 vs. 108.8 months, p = 0.001) in the whole cohort and more than seven years shorter (32.0 vs. 116.2 months, p < 0.001) in the primary non-metastatic subgroup. The simultaneous expression of PD-1 in TIMC and CTLA-4 in TIMC seems to be a predictor for rapid disease progression and death, even in localized diseases. As we showed that the combination of PD-1 and CTLA-4 expression is correlated to metastatic diseases, we hypothesize that the expression of these checkpoint molecules may be an early marker for micro metastatic diseases undetectable by cross-sectional imaging. Local therapy alone therefore is not sufficient for long term tumor control, and adjuvant therapy may be considered.
In metastatic diseases, PD-L1 expression in tumor cells or TIMC is the most studied biomarker for the prediction of a response to PD-1/PD-L1 CI therapy [33,41,42]. Response rates are better in PD-L1 positive tumors, but, as there are relevant response rates in PD-L1 negative subgroups, PD-L1 expression still is regarded as a prognostic but not predictive marker and, therefore, cannot be recommended for therapy allocation [43,44]. Major problems in the development of predictive biomarkers for CI therapy are the dynamic expression, the heterogeneity within the primary tumor, and the low correlation between the primary and metastatic sites [45][46][47]. In primary non-metastatic diseases, the whole tumor could be assessed after a resection directly prior to allocation to adjuvant therapy. The expression of key molecules for immune escape mechanisms could therefore be more representative for the immune status of the whole disease. In our view, inhibition of tumor specific immune escape mechanisms in high risk patients at a non-metastatic stage could become an important part of RCC therapy. Large scale trials on adjuvant CI therapies in RCC with antibodies targeting PD-1 and CTLA-4 (NCT03138512; NCT03142334; NCT03288532; NCT03024996) are initiated or ongoing. These studies contain large biomarker arms and may reveal predictive markers in the near future [33]. Exploratory analyses in metastatic renal cell carcinoma already show that gene expression analyses allow the classification into new subgroups which may have an association to responses across different treatments [48].
In reflection of an analysis by Choueri et al., we have evaluated the influence of the localization (left vs. right), but neither OS (p = 0.165) nor CSS (p = 0.10) demonstrated a significant correlation [49]. Thus, we did not consider localization as a separate factor in our multivariate analysis. The limitations of our study are the restriction to IHC expression analysis and the non-interventional retrospective design. There is no reliable information on treatment after a tumor resection, which may bias survival analysis. We were able to show that CTLA-4 expression is an independent marker for high risk of early cancer specific death after local therapy, but evidence for efficiency of CTLA-4 targeting therapy in an adjuvant setting is still missing. Furthermore, our analysis does not contain a comprehensive assessment of the tumor immune status, and the regulation of anti-tumor cytotoxicity in RCC is much more heterogeneous and diverse [50,51]. It would also be desirable to validate the results of our study in an independent cohort.
Another limitation is the relatively small number of patients with non-clear cell histology. Though PD-L1 expression is associated with papillary RCC, no correlation to OS or CSS was detected. For chromophobe RCC, no association of PD-1 and PD-L1 expression with OS was reported previously in a different chromophobe RCC-only cohort by Erlmeier and colleagues [52].
In conclusion, to our knowledge, this is the first description of CTLA-4 expression as an independent prognostic marker for OS and CSS in RCC. It is of special note that the combination of PD-1 and CTLA-4 expression identifies high risk patients with poor OS and CSS.