Orchestration and Prognostic Significance of Immune Checkpoints in the Microenvironment of Primary and Metastatic Renal Cell Cancer

A solid tumor is an intricate and dynamic ecosystem containing tumor and immune cells, fibroblasts, blood, and lymphatic vessels (1). The density and composition of the immune microenvironment is heterogeneous among patients and tumor types, and it is becoming a robust tool to predict postsurgical recurrence and death (1, 2). In the vast majority of cancers, tumor infiltration by CD8⁺ memory cytotoxic T cells and Th1 cells is associated with good clinical outcome (1). In addition, it has been reported that an organized local immune reaction, characterized by the presence of mature dendritic cells (DC) localized in tumor-associated tertiary lymphoid structures (TLS), is necessary to orchestrate this cytotoxic and Th1 immune contexture and is associated with good clinical outcome (3–8) and response to therapeutic vaccines (9, 10).

However, several recent findings challenge this concept of the unequivocal relation of effector memory CD8⁺ T-cell infiltration with a good clinical outcome in cancer. First, in diffuse large B-cell lymphoma (11), Hodgkin lymphoma (12), and clear cell renal cell carcinoma (ccRCC; ref. 13) high densities of tumor-infiltrating CD8⁺ T cells have been associated with poor prognosis in some studies. Second, a recent publication from our group showed that the infiltration of lung metastases from ccRCC by CD8⁺ T cells was correlated with poor overall survival (OS) whereas it was a factor of good prognosis in lung metastases from colorectal carcinoma (14). Finally, in non–small cell lung cancer (NSCLC) in which CD8⁺ T cells densities correlate with good prognosis, a low density of TLS-DC associated with high CD8⁺ T-cell infiltration identifies a group of patients with high risk of death, suggesting a functional impairment of intratumoral CD8⁺ T cells in this situation (4).

Immune checkpoints on infiltrating T cells are key regulators of immune escape in cancer. In primary ccRCC, several studies have shown that the expression of the inhibitory receptor PD-1 on the immune cells (15, 16) or its ligand PD-L1 on tumor cells (17–19) is associated with a poor clinical outcome. Interestingly, antibodies that block the PD-1 axis have yielded a 20% to 30% response rate in metastatic ccRCC (20, 21) that seems to be related to the expression of PD-L1 by the tumor cells (20, 22). Another inhibitory molecule that has gained recent attention is the lymphocyte activation gene-3 (LAG-3), which is coexpressed with PD-1 on CD8⁺ tumor-infiltrating lymphocytes in melanoma (23), and together with PD-1 synergistically regulates T-cell function (24). We expected immune checkpoints high ccRCC patients to have a worse prognosis than the immune checkpoint low.

In these newly described scenarios in which CD8⁺ T-cell infiltration correlates with poor prognosis, it is important to define the combination of immune-based biomarkers that will predict patients' prognosis and further guide immunotherapeutic approaches. This study aims to investigate the expression and prognostic significance of immune checkpoint receptors and paired ligands on primary and metastatic ccRCC in relation with TLS-DC and T-cell densities.

A cohort of 135 primary ccRCC human tumors and another of 51 ccRCC lung metastases were collected. The primary cases derived from specimens of radical nephrectomy operated between 1999 and 2003 at the hospital Necker-Enfants Malades (Paris, France). The ccRCC lung metastases cohort resected at the Hotel Dieu hospital (Paris, France) or Hôpital Européen George Pompidou (HEGP, Paris, France) between 1992 and 2010 was already described in ref. (14). This research was conducted according to the recommendations outlined in the Helsinki declaration and approved by the medical ethics boards of all participating institutions, and with the agreement of the Ile-de-France II ethics committee (no. 2012-0612). The demographic characteristics of the cohorts are depicted in Supplementary Table S1 (14).

In addition, expression data for 496 primary ccRCC samples with complete follow-up were downloaded from The Cancer Genome Atlas' (TCGA) KIRC study, using version 2 of the normalized RNA sequencing data. Corresponding clinical data (updated on 2013-04-06) were downloaded from (25).

The original histologic diagnosis was confirmed on archival hematoxylin and eosin–stained slides, and histopathologic features such as histologic subtype (26), tumor size, regional lymph node invasion, distant metastases at surgery, Fuhrman nuclear grade (27), and sarcomatoïd features were collected. All tumors were pathologically staged according to the TNM (tumor–node–metastasis) classification (28). The duration of follow-up was calculated from the date of the surgery (nephrectomy or metastasectomy) to the date of cancer progression, last follow-up or death.

Serial 5-μm formalin-fixed paraffin-embedded (FFPE) tissue sections from primary and metastatic ccRCC were stained using autostainerPlus Link 48 (Dako). Antigen retrieval and deparaffinization were carried out on a PT-Link (Dako) using the EnVision FLEX Target Retrieval Solutions (Dako). The antibodies used in this study for IHC and immunofluorescence (IF) are listed in Supplementary Table S2. IF-stained slides were scanned after secondary antibody incubation and mounting. For the IHC staining, peroxidase activity was detected using 3-amino-9-ethylcarbazole substrate or Novared and alkaline phosphatase using alkaline phosphatase substrate III (Vector Laboratories).

Tests of the specificity and sensitivity of PD-1, PD-L1, PD-L2 and LAG-3 monoclonal antibodies for IHC experiments were performed using generated FFPE cell pellets from transfected 300.19 cells (for PD-1, PD-L1, and PD-L2; ref. 29) and CHO cells (for hLAG-3), whereas parental untransfected cells served as negative controls (Supplementary Fig. S1; Costim Pharmaceuticals). Multiple organs (n = 32), human TMA (FDA999a, US Biomax), and malignant cancer tissues (n = 10) from different oncology indications were also tested using the above mentioned protocols. Normal human FFPE tonsil sections for PD-1, LAG-3, PD-L2, and normal placenta for PD-L1 were used as positive controls (Supplementary Fig. S1).

Stained slides were scanned with a Nanozoomer (Hamamatsu) and analyzed with Calopix software (Tribvn). For quantification purposes, tissues were divided into Invasive Margin (IM) and Tumor Center (TC) as previously described (30), and the density of positive cells was calculated in the whole tumor region (IM and TC). Because of the small size (or absence) of TC region in the majority of the metastases, the analysis on this cohort was done not discriminating between the two regions. The percentages of tumor cells stained positive for PD-L1 and PD-L2 were quantified by two independent reviewers (A. Lupo and N.A. Giraldo) without prior knowledge of patient outcome, and the tumors above 5% tumor cell expression were considered as positive in accordance with studies in other types of cancer (20, 31).

Comparisons among the demographic and pathologic features, immune marker densities, and PD-L1 and PD-L2 expressions were evaluated by using χ², Fisher exact, and Wilcoxon rank-sum tests. Association of variables to prognosis was assessed using the Kaplan–Meier method, univariate and multivariate Cox regression analyses. To segregate patients in two groups based on numerical variables (cell density or gene expression), Log-rank P values of each possible cutoff were computed. The cutoff that minimized the P value of a log-rank test for DFS was retained, and the corresponding P value was corrected using the method published by Altman and colleagues (32). These cutoffs were later used to segregate patients into two groups, and their associated Kaplan–Meier curves are displayed throughout the figures. To further confirm and validate the prognosis association of the cell densities, we determined the median and third quartile cutoff and calculated the corresponding univariate Cox-regression P values (Supplementary Table S3). Only those variables univariately associated with prognosis were included in the multivariate Cox regression analysis. The duration of follow-up was calculated from the date of nephrectomy or metastasectomy to the date of death or last follow-up.

The density of CD8⁺ T cells in the IM of the primary tumors was heterogeneous in the cohort of 135 primary ccRCC. On the basis of the Optimal P value cutoff for DFS (OPv; 630 cells/mm²) we found that the CD8^High (n = 41/135, 30%) group had a shorter disease-free survival (DFS, P = 0.0001, Fig. 1A) and OS (P = 0.001, Fig. 1A). The density of CD8⁺ T cells in the TC had no prognostic impact (Supplementary Table S4). The density of CD8⁺ T cells in the IM correlated with the Fuhrman grade (Supplementary Fig. S2), but not with any other pathologic variables, including TNM (data not shown).

An independent cohort of lung metastases of ccRCC, in which the negative impact of high densities of CD8⁺T cells on OS had been previously described using semiquantitative counting techniques (14), was reanalyzed using a quantitative approach on 51 cases. The OPv for CD8⁺ T cells density was 490 cells/mm², and the CD8^High (n = 14/51, 27%) group displayed a shorter OS (P = 0.001, Fig. 1B). This result confirms our previous observations (14).

From TCGA public database of 496 primary ccRCC, we analyzed the expression of 844 immune-related genes, from which we extracted data concerning seven genes expressed in a Th1- and CD8⁺ T-cell–oriented response according to Galon and colleagues (30). We found that the expression of most of the genes associated with this cell signature correlated with poor prognosis: CD8A P = 0.04, TBX21 P = 0.03, IRF1 P = 0.01, GZMB P = 4.4 × 10⁻⁵, and IFNG P = 3.17 × 10⁻⁷ that displayed the lowest P value (Fig. 1C). On the basis of the OPv for IFNG, patients that had high quantities of intratumoral mRNA for this gene displayed a shorter OS (P = 0.006, Fig. 1D).

To investigate the impact of immune checkpoint molecules on the negative prognostic impact of the CD8^High group, we analyzed the protein expression of PD-1, LAG-3, PD-L1, and PD-L2 molecules in this group of tumors (we could analyze 40/41 tumors) and a randomly matched group of the same size (n = 40/94) coming from the CD8^Low.

On the basis of the OPv cutoff, 15 tumors of 80 were considered as PD-1^High in the IM and they displayed a shorter DFS (P = 0.0005, Fig. 2A) and OS (P = 0.03, Fig. 2A). The density of PD-1⁺ cells in the TC was not significantly associated with prognosis (Supplementary Table S4). The Fuhrman grade was associated with the PD-1⁺ cell density in the IM (Supplementary Fig. S2).

Of the 80 patients, 9 were considered as LAG-3^High (subdivided by the OPv cutoff) in the IM and they displayed a shorter DFS (P = 0.02, Fig. 2A), and did not reach significance for the OS (P = 0.07, Fig. 2A). The density of LAG-3⁺ cells in the TC was not significantly associated with patients DFS or OS (Supplementary Table S4). Representative pictures of the IHC staining of highly and poorly PD-1– and LAG-3–infiltrated lesions are shown in Supplementary Fig. S3A and S3B, respectively.

On the basis of the 5% cutoff generally used in clinical trials with anti-checkpoint antibodies (20), 22 of 80 (27%) and 27 of 80 (34%) patients were PD-L1⁺ and PD-L2⁺, respectively (Supplementary Fig. S3C and S3D); among them, 9 tumors were double positive. Patients with PD-L1⁺ tumors had a shorter OS than those with less than 5% positive tumor cells (P = 0.02, Fig. 2A); a trend toward shorter DFS was also found (P = 0.06, Fig. 2A). Univariately, patients with PD-L1⁺ tumors were 2.9 times more likely to die in the 5 years post-nephrectomy than patients with less than 5% PD-L1⁺ tumor cells [OS risk ratio, 2.87; 95% confidence interval (CI), 1.2–7.1; P = 0.02, Supplementary Table S4]. Patients with PD-L2⁺ tumors displayed a shorter OS compared with those with less than 5% positive tumor cells (P = 0.005, Fig. 2A), but no impact on the DFS was found (P = 0.13, Fig. 2A). Univariately, patients with PD-L2⁺ tumors were 3.4 times more likely to die than those with negative tumor cells (risk ratio, 3.4; 95% CI, 1.4–8.5; P = 0.005, Supplementary Table S4).

Patients with tumors exhibiting both high densities of PD-1⁺ lymphocytes in the IM and PD-L1⁺ and/or PD-L2⁺ tumor cells (n = 11) had the worst prognosis, as assessed by DFS and OS (DFS, P = 1.3 × 10⁻⁵; OS, P = 0.005; Fig. 2B). Patients that met these criteria had 6.1 times more risk to progress after resection than patients who did not (risk ratio, 6.08; 95% CI, 2.4–15.0; P < 0.001, Supplementary Table S4). Strikingly, 91% of the patients having a PD1^High infiltrate and tumor cells expressing PD-L1 and/or PD-L2 progressed in the subsequent 5 years of surgery versus 36% in the negative group (complete 5-years follow-up n = 70, Fisher exact test P = 0.001, Fig. 2C). In the TCGA public datasets of 496 primary ccRCC, the gene expression of PD-1 (PDCD1), LAG-3, and PD-L2 (PDCD1LG) was associated with shorter OS (P = 0.03, P = 0.0001, and P = 0.0003, respectively) whereas PD-L1 (CD274) was not (P = 0.67; Fig. 2D).

It has been reported that IFNγ can induce PD-L1 expression on tumor cells (27). We found a significant positive correlation between the gene expression of IFNG with PD-L1 (R = 0.13, P = 0.004) and with PD-L2 (R = 0.42, P = 2.2 × 10⁻¹⁶) in the TCGA cohort (Fig. 3A). In addition, a significant positive correlation was found between the densities of PD-1⁺ and CD8⁺ T cells (R = 0.31, P = 0.004), LAG-3⁺ and CD8⁺ T cells (R = 0.42, P = 0.001) and PD-1⁺ and LAG-3⁺ cells (R = 0.55, P < 0.0001) in the IM of the 80 primary ccRCC (Fig. 3B and 3C). An even stronger correlation was found at the gene-expression level between PD-1 and LAG-3 (R = 0.81, P = 0.002), PD-1 and CD8A (R = 0.95, P < 0.001), and LAG-3 and CD8A (R = 0.96, P < 0.001) in the TCGA cohort of 496 primary tumors (Fig. 3B and C). In accordance with all these observations, we also detected the presence of triple-positive CD8⁺/PD-1⁺/LAG-3⁺ cells in 6 of 7 cases from the CD8^High group of primary tumors by IF on paraffin sections (representative pictures are displayed in Fig. 3D).

Metastatic ccRCC has been one cancer in which antibodies inhibiting the PD-1 axis have induced remarkable tumor regression in some patients (20), highlighting the necessity to define prognostic and predictive immune-based biomarkers in this disease. Of the 51 patients with ccRCC lung metastasis, 13 were considered as PD-1^High, and the latter displayed a shorter OS (P = 0.008, Fig. 4A). Seven of 51 were considered as LAG-3^High, and again they displayed a shorter OS (P = 0.048, Fig. 4A).

On the basis of a 5% cutoff, 5 of 51 (10%), and 15 of 51 (29%) metastases were PD-L1⁺ or PD-L2⁺, respectively; of them, 3 were double positive. Whereas patients with PD-L1⁺ metastases did not have a significantly worse clinical outcome (P = 0.12, Fig. 4A), the expression of PD-L2 on tumor cells was associated with a shorter OS (P = 0.03, Fig. 4A). Univariately, patients with PD-L2⁺ metastasis were 2.2 times more likely to die compared with patients with PD-L2- ones (risk ratio, 2.17; 95% CI, 1.03–4.35; P = 0.04, Supplementary Table S4).

Patients with high densities of PD-1⁺ lymphocytes and PD-L1⁺ and/or PD-L2⁺ tumor cells in their ccRCC lung metastases (n = 12) had the worst prognosis as assessed by OS (P = 0.003, Fig. 4B). The patients that met these criteria had 3.1 times more risk to die after metastasectomy than patients who did not (risk ratio, 3.1; 95% CI, 1.28–6.66; P = 0.003, Supplementary Table S4). Strikingly, 100% of the patients having a PD1^High infiltrate and metastases simultaneously expressing one or both of its ligands (PD-L1 or PD-L2) died in the subsequent 5 years after surgery versus 57% in the negative group (Fisher tests, P value = 0.004, Fig. 4C).

A significant and strong correlation was found between densities of PD-1⁺ and CD8⁺ T cells (R = 0.51, P = 0.0001), LAG-3⁺ and CD8⁺ T cells (R = 0.4, P = 0.004; Fig. 4D) and PD-1⁺ and LAG-3⁺ cells (R = 0.71, P < 0.0001).

In conclusion, the combined analysis of the expression of PD-1, PD-L1, and PD-L2 identified a group of patients with a high risk of progression and death in primary and in an independent cohort of metastatic ccRCC.

It has been shown that TLS-DC orchestrate intratumoral CD8⁺ cytotoxic T cells and Th1 responses in NSCLC (3, 4). In the primary ccRCC cohort (n = 135), DC-Lamp⁺ cells were detected within TLS (TLS-DC) in the IM. They coexpressed CD83 and high amounts of MHC Class II, and were localized in the vicinity of PNAd⁺ high endothelial venules (HEV; Fig. 5A) as described in other tumor types (3). There was a trend where high densities of TLS-DC were associated with longer DFS (OPv = 0.09), but not with OS (Supplementary Table S4). TLS-DC densities in the TC were not associated with prognosis. Interestingly, high densities of TLS-DC (based on the OPv cutoff) in the IM identified a group of patients with good prognosis among the CD8^High group for both DFS (P = 0.001, Fig. 5A) and OS (P = 0.03, Fig. 5A).

However, in contrast with NSCLC where most of the DC-Lamp⁺ cells are localized within TLS (4), in ccRCC a large percentage (79% ± 20%) of DC-Lamp⁺ cells with DC morphology was found isolated and outside TLS (NTLS-DC, Fig. 5B). They colocalized with CD34⁺ blood vessels, but not with PNAd⁺ HEV, expressed low amounts of MHC class II and were CD83 negative (Fig. 5B). The majority of NTLS-DC was localized in the IM (71 ± 14%) and their high densities (based on the OPv cut-off) were associated with poor clinical outcome in primary ccRCC patients (DFS P = 0.006, OS P = 5.1 × 10⁻⁵, Fig. 5B).

The opposite influences of TLS-DC and NTLS-DC on prognosis prompted us to explore their relationships with the expression of PD-1 and its ligands in the CD8^High group of patients. A negative correlation was found between the densities of TLS-DC and PD-1⁺ cells (r = −0.23, P = 0.04, Fig. 5C). Tumors containing PD-L1⁺ and/or PD-L2⁺ tumor cells exhibited less TLS-DC (P = 0.014; Supplementary Fig. S4), but similar densities of CD8⁺ T cells (P = 0.96). In contrast, the density of NTLS-DC was associated with the tumor expression of PD-L1 (OR, 6.54; 95% CI, 1.23–45.45; P = 0.012) and PD-L2 (OR, 2.7; 95% CI, 1.2–18.1; P = 0.04). Most of the patients that were PD-1^High and PD-L1⁺ and/or L2⁺ (8 of 11) were also CD8^High and TLS-DC^Lo in primary ccRCC (OR, 15.02; 95% CI, 1.66–72.7; P = 0.004), and the presence of one or both patterns correlated with shorter DFS (P < 0.0001) and OS (P = 0.001, Fig. 5D).

To define the independent prognostic significance of the previously mentioned immune profiles (CD8^High and TLS-DC^Low or PD-1^High and PD-L1^+and/orL2⁺), we included them in a multivariate Cox regression analysis with the other significant prognostic clinical variables (TNM and Furhman grade). The strongest independent poor prognostic factors in primary ccRCC for DFS were to have a CD8^High and TLS-DC^Low (P = 0.001, Table 1) or PD-1^High and PD-L1^+and/orPD-L2⁺ (P = 0.03, Table 1) immune profile. For the metastatic cohort, we found that the strongest independent worst prognostic variable for OS was being CD8^High (P = 0.004, Table 1) or having a tumor with PD-1^High and PD-L1⁺ and/or PD-L2⁺ immune profile (P = 0.02, Table 1).

Compared with other neoplasia, the immune microenvironment in ccRCC has not yet been studied in detail. However, it is of paramount importance to understand its regulation in view of the paradoxical correlation of CD8⁺ T-cell infiltration with poor prognosis (13, 14) and the recent advances of immunotherapy with anti-checkpoint antibodies (anti PD-1 and anti PD-L1; refs. 20, 21).

ccRCC has been described as a proinflammatory neoplasia where tumor cells produce several cytokines (such as VEGF, IL6 and TGFβ; refs. 33–35) that may lead to the recruitment and activation of polyclonal CD8⁺ T cells (36–39). Our data suggest that these recruited CD8⁺ T cells could only be locally educated when high densities of TLS-DC are present, and in only in these cases their density correlates with favorable prognosis. ccRCC is the first tumor type where a large proportion of DC-Lamp⁺ cells outside TLS structures have been observed (NTLS-DC). Interestingly, we found that these cells do not express activation and costimulatory markers, and they are probably recruited directly from the blood into the tumor stroma—contrary to the usual path that DC-Lamp⁺ cells follow in other type of tumors (5, 40). Accordingly, several studies have shown that the ccRCC microenvironment can induce a dysfunctional DC maturation, a downregulation of costimulatory molecules and tolerogenicity (34, 41, 42), whereas DC in the TLS are likely to be protected from these effects (43). Altogether, our results suggest that the particular proinflammatory milieu initiated by tumor cells induces the recruitment of CD8⁺ T cells that—due to the low number of fully functional mature DC present in specialized T-cell–priming sites, or the presence of DC with suppressive phenotype—are not able to mount an effective antitumor immune response, but rather express exhaustion/inhibition molecules.

This work reinforces the concept that T-cell exhaustion/inhibition (44) plays an important role in ccRCC pathogenesis. It has been described that the density of PD-1⁺ cells (15, 16) and the tumor expression of PD-L1 (17–19) in primary ccRCC are associated with a poor clinical outcome. In this study, we confirm these findings and extend them to ccRCC lung metastases. This information is highly relevant because metastatic ccRCC-treated patients have shown one of the highest objective durable response rates to PD-1 blockade (approximately 30%; ref. 20) and many efforts are being dedicated to define theranostic tools in this pathology. Furthermore, we describe for the first time the prognostic significance of PD-L2. This molecule seems to be expressed in a higher proportion of tumors than PD-L1, and up to 30% of them might express it solely according to our results. This might be of clinical relevance because—although in two different nonrandomized cohorts—the anti–PD-L1 treatment alone seems to have a lower response rate than the anti–PD-1 (12% and 27%, respectively; refs. 20, 21). Furthermore, there are PD-L1–negative tumors that respond to anti–PD-1 treatment (22), suggesting that indeed there are other molecules beside PD-L1 implicated in the PD-1 inhibition axis of ccRCC. Few publications have reported PD-L2 mRNA or protein expression in other tumors, including primary mediastinal large B-cell lymphoma (29), NSCLC (45), ovarian (46), and esophageal (47), where it has shown a limited impact of patients' prognosis.

To our knowledge, this is the first report on the poor prognostic impact associated with high densities of LAG-3⁺ cells in human tumors. Furthermore, we provide clear evidence that the expression of PD-1 and LAG-3 is highly correlated in ccRCC. Some studies on mouse models have shown a synergistic effect of the inhibition of both pathways in boosting antitumor immune response (24). Therefore, our data support the rationale of dual blockade of these molecules in ccRCC.

Although recent works have emphasized that PD-L1 is preferentially upregulated by IFNγ (31, 48), whereas PD-L2 is regulated by IL4 (49), the weak association of PD-L1 mRNA with IFNG might suggest an important role for posttranscriptional regulation of PD-L1 expression in ccRCC as for other aggressive tumors (50). Furthermore, it is highly suggestive that IFNγ can also induce PD-L2 upregulation in tumors, as in immune cells (51), and supports the rationale to use therapeutic antibodies targeting this ligand. Taken together, our results suggest that the expression of these molecules is related with a chronic inflammatory and highly suppressive process that is unselectively recruiting CD8⁺ and NTLS-DC cells from the circulation, and overall is associated with a poor prognosis.

Another characteristic of ccRCC is the lack of prognostic significance of the immune cells in the TC. Although CD8⁺, PD-1⁺, LAG-3⁺, and NTLS-DC densities in the IM of the primary tumors were associated with poor prognosis, they had no prognostic significance when present in the TC. In colorectal carcinoma, both regions are important to define the best immune score, which correlates with patient's longer survival (30). Whether this dichotomy between ccRCC and colorectal carcinoma reflects different tumor tissue organization or relates to other factors of the tumor microenvironment remains to be clarified.

Recent unsupervised gene clustering of stage IV primary ccRCC showed that tumors with high inflammatory immune infiltrate (approximately 15%) also have a high expression of PDCD1 (PD-1) and its ligands, and correlate with the worst prognosis (52). Indeed, this inflammatory/proangiogenic profile (probably originated from the ccRCC tumor cells) found in ccRCC primary tumors and in ccRCC lung metastases (14), is not found in colorectal carcinoma lung metastases (14), highly supporting that it may contribute to local immunosuppression process, the absence of fully mature DC, and high expression of immune checkpoints (53).

In summary, we identify a subset of primary and metastatic ccRCC patients characterized by (i) an extensive CD8⁺ T-cell infiltrate, (ii) expression of immune checkpoints, and (iii) the absence of fully functional DC, which is associated with poor clinical outcome. Our results highlight novel independent prognostic factors in ccRCC based on the concomitant quantification of densities of DC, CD8⁺, PD-1⁺, and LAG-3⁺ lymphocytes in addition to PD-L1/PD-L2 expression by tumor cells. These immune profiles should guide the selection of suitable patients to receive immunotherapies and need to be further validated in larger and independent cohorts. Because this choice depends on both the extent of CD8⁺ T-cell infiltration and the maturation and localization of DC, it invites revision of the idea that intratumoral CD8⁺ T cells are always associated with favorable prognosis in human tumors.

Y. Vano reports receiving other research grants from Novartis and is a consultant/advisory board member for Novartis and Pfizer. G.J. Freeman has ownership interest (including patents) in Amplimmune, Boehringer-Ingelheim, Bristol-Myers Squibb, EMD Serono, Merck, Novartis, and Roche and is a consultant/advisory board member for Novartis and Roche. S.M. Oudard reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Janssen, Novartis, Pfizer, Roche, and Sanofi. W-H. Fridman is a consultant/advisory board member for Costim. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N.A. Giraldo, G.J. Freeman, S. Oudard, W.H. Fridman, C. Sautès-Fridman

Development of methodology: N.A. Giraldo, G. Skliris, L. Lacroix, G.J. Freeman, M.-C. Dieu-Nosjean, S. Oudard

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.A. Giraldo, F. Pagès, V. Verkarre, Y. Vano, N. Saint-Aubert, L. Lacroix, I. Natario, A. Lupo, M. Alifano, D. Damotte, A. Cazes, S. Oudard

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.A. Giraldo, E. Becht, A. Lupo, M.-C. Dieu-Nosjean, S. Oudard, W.H. Fridman, C. Sautès-Fridman

Writing, review, and/or revision of the manuscript: N.A. Giraldo, E. Becht, F. Pagès, G. Skliris, Y. Vano, A. Mejean, F. Triebel, G.J. Freeman, M.-C. Dieu-Nosjean, S. Oudard, W.H. Fridman, C. Sautès-Fridman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.A. Giraldo, F. Pagès, V. Verkarre, A. Mejean, L. Lacroix, D. Damotte, F. Triebel, S. Oudard, W.H. Fridman, C. Sautès-Fridman

Study supervision: N.A. Giraldo, W.H. Fridman, C. Sautès-Fridman

The authors thank Romain Remark for his help in the case/block selection and collection of clinical data from the metastatic cohort, Jose Balcaceres (Association pour la recherche sur les Thérapeutiques en Cancérologie) for his support in the collection of the clinical data of the primary cohort, Patricia Bonjour for cutting the paraffin-embedded blocks, Dr. Marc Riquet for the case selection of the metastatic cohort and all members of the teams 13 and 15 in the Cordeliers Research Center for their valuable discussions.

This work was supported by the “Institut National de la Santé et de la Recherche Médicale,” the University Paris-Descartes, the University Pierre et Marie Curie, the Institut National du Cancer (2011-1-PLBIO-06-INSERM 6-1 and PLBIO09-088-IDF-KROEMER), CARPEM (CAncer Research for PErsonalized Medicine), Labex Immuno-Oncology (LAXE62_9UMS872 FRIDMAN), Universidad de los Andes School of Medicine (to N.A. Giraldo), Colciencias (to N.A. Giraldo), and NIH (P50CA101942; to G.J. Freeman).

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.