Purpose:

A minority of patients currently respond to single-agent immune-checkpoint blockade (ICB), and strategies to increase response rates are urgently needed. AXL is a receptor tyrosine kinase commonly associated with drug resistance and poor prognosis in many cancer types, including in clear-cell renal cell carcinoma (ccRCC). Recent experimental cues in breast, pancreatic, and lung cancer models have linked AXL with immune suppression and resistance to antitumor immunity. However, its role in intrinsic and acquired resistance to ICB remains largely unexplored.

Experimental Design:

In this study, tumoral expression of AXL was examined in ccRCC specimens from 316 patients who were metastatic receiving the PD-1 inhibitor nivolumab in the GETUG AFU 26 NIVOREN trial after failure of antiangiogenic therapy. We assessed associations between AXL and patient outcomes following PD-1 blockade, as well as the relationship with various markers, including PD-L1; VEGFA; the immune markers CD3, CD8, CD163, and CD20; and the mutational status of the tumor-suppressor gene von Hippel-Lindau (VHL).

Results:

Our results show that high AXL-expression level in tumor cells is associated with lower response rates and a trend to shorter progression-free survival following anti–PD-1 treatment. AXL expression was strongly associated with tumor–PD-L1 expression, especially in tumors with VHL inactivation. Moreover, patients with tumors displaying concomitant PD-L1 expression and high AXL expression had the worst overall survival.

Conclusions:

Our findings propose AXL as candidate factor of resistance to PD-1 blockade, and provide compelling support for screening both AXL and PD-L1 expression in the management of advanced ccRCC.

See related commentary by Hahn et al., p. 6619

Translational Relevance

This is the first study to demonstrate in a sizable cohort of cases that high tumor-AXL expression may affect outcomes of patients with advanced clear-cell renal cell carcinoma (ccRCC) treated with anti-PD1 therapy. This study also reveals AXL expression is associated with tumor–PD-L1 expression and that co-occurrence of PD-L1 positivity and high AXL expression coincides with reduced overall survival in patients treated with PD-1 blockade. This finding may aid in the selection of treatment for advanced renal cell carcinoma (RCC) patients.

Kidney cancer is responsible for about 175,000 deaths each year (1). The most common type of kidney cancer is clear-cell renal cell carcinoma (ccRCC) accounting for 70% to 80% of cases. Treatment with immune checkpoint blockade (ICB) has resulted in striking clinical benefits and changes in the management of patients with advanced ccRCC since anti-PD1 (nivolumab) therapy has been approved as second-line treatment for recurrent metastatic diseases, as well as the anti–PD-1 (nivolumab) plus anti–CTLA-4 (ipilimumab) combination approved as first-line treatment for patients with intermediate and high-risk metastatic RCC followed by the combination of PD-1 inhibitor with VEGFR tyrosine kinase inhibitors (TKI) in any risk group in front line (2–5). Cell-surface receptor PD-1 is especially expressed by T lymphocytes, which can prevent T cells from killing tumor-target cells upon binding to its ligands PD-L1 and PD-L2, dampening the immune response. Therefore, PD-1 blockade using specific neutralizing antibodies is able to inhibit interaction of PD-1 with its ligands, which may at least partially restore their activity and boost antitumor T-cell immunity. Nevertheless, only a small proportion of patients benefits from these treatments as shown by the low response rate (RR) on one hand, and high rates of disease progression on the other hand, even in initial responders. This illustrates primary and acquired resistances to PD-1 blockade. In an era of progress where anti–PD-1-based combination approaches are expected to increase RR and prolong treatment efficacy (6, 7), it is urgent to better define the underlying biology of resistance and response to PD-1 blockade. Thus, new biomarkers and targets that can be easily and routinely inspected in pathologic sections might facilitate treatment decisions and give new therapeutic options for metastatic ccRCC (m-ccRCC).

Notably, ccRCC has specific features: (i) In contrast to most tumor types, ccRCC can be highly infiltrated by CD8+ T cells, but this infiltration is often associated with poor prognosis (8); (ii) Renal cell carcinoma (RCC) has relatively low mutational load and this parameter does not appear to be a good predictor of response to PD-1 blockade (5); (iii) controversies also exist regarding the utility of PD-L1 expression to predict response. Indeed, PD-L1 expression is associated with poor prognosis in RCC, but its value to predict response to PD-1 blockade is still a matter of debate (5, 9, 10). Furthermore, genetic alterations in PBRM1 and BAP1 genes as well as focal loss in chromosome 10q23.31 have demonstrated interesting properties to estimate patients who are likely to benefit from ICB (11, 12). The von Hippel-Lindau (VHL) tumor-suppressor gene is also altered in a majority of RCC cases. Defects in VHL gene include germline or somatic mutations, loss of heterozygosity through chromosomal loss of 3p, and promoter hypermethylation (13). It is often considered that two “hits” must occur to inactivate VHL. Thus, a combination of these alterations can cause biallelic inactivation of VHL in 60% to 90% of RCC cases leading to a complete loss-of-function (LOF) of VHL gene in some cases, whereas in other cases, such function likely remains intact. VHL LOF in tumor cells results in chronic accumulation of hypoxia-inducible factors (HIF) in the cells entering a pseudohypoxic state likely contributing to tumorigenesis, angiogenesis, and changes in the tumor microenvironment (TME), in particular via induction of the HIF-target gene, VEGFA. Recent studies reported clinical benefits for combination of VEGFR TKIs with ICB (5). However, the impact of the VHL/HIF/VEGF pathway, beyond antiangiogenic therapy, in regulating refractory responses to ICB remains unclear.

AXL is member of the TYRO3-AXL-MER (TAM) family of receptor tyrosine kinases, which can regulate tumor-cell survival, proliferation, migration, angiogenesis, and interactions with TME (14). AXL has been associated with drug-resistance and poor prognosis in many cancer types including in RCC (14). Notably, AXL expression in human RCC has been associated with high risk of relapse after surgery (15), worse clinical outcome in patients who are metastatic treated with antiangiogenic agents, as well as epithelial–mesenchymal transition (EMT) features and fibrosis (16). Recent evidence in breast-, pancreatic-, and lung-cancer models have also linked AXL with immune suppression and resistance to antitumor immunity (17–20). Yet, the role of AXL in resistance to anti-PD1 therapy remains unknown.

In this study, we aimed to elucidate the potential impact of AXL expression on clinical outcomes of patients with advanced RCC following anti-PD1 treatment.

Patients and samples

The renal-cancer samples have been collected as part of the NIVOREN GETUG-AFU 26 phase II trial assessing the activity and safety of nivolumab (anti-PD1) in 729 patients with m-ccRCC who failed VEGF-directed therapies (ClinicalTrials.gov identifier: NCT03013335). This trial was conducted by UNICANCER, approved by the human research ethics committee of UNICANCER and conducted in accordance with guidelines from the International Conference on Harmonization and the Declaration of Helsinki. Patients were enrolled across 27 institutions between February 12, 2016 and July 27, 2017. Safety activity data was reported previously (21). The follow-up was censored on January 2019.

As part of a biomarker cohort, 324 archived formalin-fixed, paraffin-embedded (FFPE) tumor-tissue specimens were retrospectively analyzed by IHC for immune-cell profiling (CD3, CD8, PD-L1, CD20, CD163, VEGFA, and AXL). Specimens were mostly obtained prior antiangiogenic therapy as first- or second-line regimens (comprising Sunitinib, 83.3%; Axitinib, 27.5%; Pazopanib, 23.8%; Sorafenib, 6.5%; Bevacizumab, 3.4%; Cabozantinib, 0.3%), or mTOR inhibition (Everolimus 11%). Cases with paraffin block showing less than 40% of tumor or more than 40% of necrosis, or corresponding to another histologic subtype (TFE3 translocation RCC, chromophobe RCC, or papillary RCC) were excluded.

Alterations in VHL gene

VHL-inactivation status was determined by PCR, FISH, and promoter-methylation analysis. Evaluation of VHL point mutations was evaluated by gene sequencing (22). The three exons and exon–intron junctions of the VHL gene were sequenced on DNA extracted from tumor samples, as described previously (23). Search for deletion of the VHL genomic region was performed by FISH using ZytoLight SPEC VHL/CEN3 Dual Color Probe (Zytovision) reviewed by F. Dugay. Promoter hypermethylation of VHL was assessed by methylation-specific multiplex ligation-dependent probe amplification as described (23). Presumably two hits are needed for VHL inactivation to occur. Therefore, tumors were classified in 2 groups. When biallelic alteration was evidenced, the case was classified in a group with VHL inactivation, also referred to as VHL-/-. Another group consisted of noninactivated cases, with no alteration or one allelic alteration, referred to as VHL +/+; ±. A few cases were excluded from this analysis when biallelic or monoallelic alterations could not be ascertained because of noninterpretable results, or when the mutation detected was simply defined as possibly damaging with no clear loss of VHL function.

mRNA extraction and analysis of RNA-sequencing datasets from patients with RCC

RNA-sequencing (RNA-seq) data were available for 83 FFPE cases of the Nivoren cohort (24). Briefly, archived FFPE samples were retrieved before anti–PD-1 treatment, and upon review of hematoxylin and eosin (H&E) slides, the tissue blocks with the highest tumoral content were selected for further processing. Blank slides were produced with 5 μmol/L thickness and macrodissected to only include tumoral tissue, using a total surface area of 50 to 1800 mm2. RNA was extracted using the Maxwell RSC RNA FFPE kit (Promega) according to the manufacturer's instructions. Subsequently cDNA libraries were prepared using the Forward QuantSeq 3′ mRNA-Seq Library Prep Kit for Illumina (Lexogen) according to the manufacturer's instructions with 5 μL of RNA as input and 16 PCR cycles. cDNA concentrations and fragment length were measured with the QubitTM dsDNA HS assay (Thermofisher) and Bioanalyzer HS DNA electrophoresis (Agilent). Illumina cBOT was used for clonal-cluster generation and RNA-seq was performed using the HiSeq 4000 kit (Illumina) according to the manufacturer's instructions. For subsequent survival analysis, samples were classified as AXL high (defined by AXL mRNA expression ≥ highest 1/3) or AXL Low (defined by AXL mRNA expression ≤ lowest 2/3) in the dataset. Progression-free survival (PFS) was estimated using the Kaplan–Meier method to explore the impact of AXL in the different subsets. For association studies, VHL status was determined as above. AXL immunostaining also served to classify tumors as AXLhigh or AXLneg/low, and in this case, RNA-seq data was interrogated to evaluate changes in the expression of known gene-expression signatures (25–27).

CheckMate RNA-seq data, clinical, and genomic information were interrogated from the Braun and colleagues study (11) for 181 advanced ccRCC treated with anti-PD1 in the CheckMate 009, 010, and CheckMate 025. The CheckMate cohorts consist mostly of previously-treated antiangiogenic-refractory patients. Experimental details including whole-exome and RNA-seq analyses have been reported previously (11). Among the samples with both RNA and genomic information, 93% exhibited loss of chromosome 3p carrying VHL gene, and 64% were reported with a point mutation in VHL. Cases with at least one wild-type (WT) copy of VHL were considered as noninactivated VHL cases, whereas cases with at least two alterations were considered as inactivated. Samples were classified as above with AXLhigh (defined by AXL-mRNA expression ≥ highest 1/3) and AXLlow cases (defined by AXL-mRNA expression ≤ lowest 2/3). PFS and overall survival (OS) estimates using the Kaplan–Meier method were generated to explore the impact of AXL in all patients (n = 181), VHL-inactivated (n = 78), or VHL-noninactivated (n = 44) subgroups.

IHC

All tumor specimens were centrally reviewed by a genitourinary pathologist (N. Rioux-Leclercq). For each case, 1 representative archived FFPE-tissue block was used for IHC. Primary antibodies to stain for PD-L1 (clone 22C3; Dako), CD8 (clone C8/144B; Dako), VEGF (clone SP28; Spring Bioscience Corp), CD163 (clone 10D6; Diagnostic BioSystem), CD20 (clone L26; Agilent), CD3 (polyclonal rabbit antibody; Agilent) were used according to standard protocols. For PD-L1, AXL, CD8, and VEGF, slides were independently evaluated by a genitourinary pathologist (J. Adam, N. Rioux-Leclercq) blinded to the patients' treatment outcomes. For AXL IHC assessment, epitope retrieval was performed in a citrate buffer pH 7.3 (Diapath) during 30 minutes in a water bath set to 98°C. Endogenous peroxidase activity was inactivated with 3% hydrogen peroxide for 10 minutes. Blocking was performed with IHC/ISH Super Blocking (Leica biosystems) for 1 hour at 37°C. The primary antibody (clone C89E7; Cell Signaling Technology), was incubated overnight. Omission of the primary antibody was used as negative control. A multiplicative H-score (0–300) was calculated as the product of intensity scores and the percentage of stained tumor cells. Representative stainings are shown in Supplementary Fig. S1. To note, AXL staining could also be detected in various nontumor cells, in particular immune cells with morphology evocative of myeloid cells [i.e., monocytes/macrophages/dendritic cel (DC)] (Supplementary Fig. S1). A score of 0 to 3 was given based on the abundance of AXL+ immune cells.

Patients were considered with Negative (H-score 0), Low ([0;50]), or High (>50) AXL tumor expression. H-score greater than 50 was used as cutoff to discriminate between AXLhigh and AXLneg/low expression. For PD-L1, VEGF, and CD8, stainings were performed using a Discovery XT research instrument (Roche) using standardized protocols. For PD-L1 (anti–PD-L1, clone 22C3; Dako). The percentage of positive tumor cells was evaluated. Percentage ≥ 1 was used as cutoff to discriminate between PD-L1 TC neg and PD-L1 TC+ expression. For assessment of PDL1 staining in tumor-infiltrating lymphocytes (TIL), a manual semiquantitative three-point scale was employed with score 0 (no PDL1+ TILs), score 1 (rare and focal PDL1+ TILs), score 2 (several foci of PDL1+ TILS), score 3 (diffuse and numerous PDL1+ TILs). For subsequent analysis, the cutoff ≥ 1 was used to discriminate PD-L1neg and PD-L1+ in TILs. CD8 positivity (anti-CD8, clone C8/144B; Dako) in TILs was scored according to the proposal of the International ImmunoOncology Biomarkers Working Group (28), using a manual semiquantitative four-point scale: score 0 (no CD8+ TILs or very rare CD8+ TILS), score 1 (rare diffuse or focal CD8+ TILs), score 2 (diffuse numerous CD8+ TILs), and score 3 (diffuse and numerous CD8+ TILs with some aggregates of TILs). Anti-VEGF (clone SP28; Spring Bioscience Corp) was used and the percentage of positive tumor cells assessed. For CD3 (polyclonal rabbit antibody; Agilent), CD20 (clone L26; Agilent), and CD163 (clone 10D6; Diagnostic BioSystem) immunostaining, tissues were deparaffinized and rehydrated by successive baths of Clearene (Leica Biosystems, 3803600E) and ethanol gradient (100%, 90%, 70%, and 50%). Antigen retrieval was performed with a PT-link (Dako) at pH 6.1 (EnVision FLEX Target Retrieval Solution, Low pH 50x concentrated Citrate buffer; Dako, K8005) by an incubation of 30 minutes at 97°C. Fifteen minutes of Endogenous Peroxidase block (H2O2 3% Gifrer, 1060351) and 10 minutes of Protein Block (Dako, X0909) reagents were used for alkaline-phosphatase and FcR blocking. The primary and secondary antibodies were incubated for 30 minutes respectively. All these markers were processed on an automated immunostainer (AutostainerPlusLink 48, Dako). Secondary antibodies were revealed by Permanent HRP Green (Zytomed Systems, ZUC070–100) for CD3 (10 minutes), High-Def red IHC chromogen (AP; Enzo, ADI-950–140–0030) for CD20 (5 minutes) and AEC Substrate Kit, Peroxidase HRP: 3-amino-9-ethylcarbazole (Vector Laboratories, SK-4200) for CD163 (23 minutes). The nuclei were counterstained with hematoxylin (Dako, S3301) for 5 to 7 minutes and slides were scanned with a Nanozoomer (Hamamatsu) after mounting with Ecomount (Biocare, EM897L). For CD20 and CD3, and glycergel Mounting Medium (Dako, C056330–2) for CD163. The density of positive cells/mm² was quantified in the Tumor Core (TC) by 2 independent analysts (A. Bougouin, G. Lacroix) using Halo10 software (Indica Labs).

Statistical analysis

Qualitative data are reported by frequency and proportion. The χ2 and Fisher exact tests were used to identify associations between biomarkers, status, or groups as categorical variables. Median follow-up is calculated using the reverse Kaplan–Meier method and presented associated with its 95% confidence interval (CI). PFS was defined as the time between treatment initiation and first progression or death, and OS was defined as the time between treatment initiation and death or date of final contact for patients alive. OS and PFS were estimated using the Kaplan–Meier method and were described in terms of median or specific time-point estimation in each subgroup, along with the associated two-sided 95% CI for the estimates and were compared with the log–rank test. A multivariate Cox regression model was used to estimate the HR and 95% CI and adjusted for the following baseline characteristics: International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) risk groups, age, sex, and number of previous systemic therapies. Objective response rate (ORR) as per central radiology review using RECIST version 1.1 has been previously described (21). Quantitative data are reported by means and range. Nonparametric Mann–Whitney U or Kruskal–Wallis test were applied. A value of P < 0.05 was considered statistically significant, and all p values were two-sided. Data analyses were carried out using GraphPad (GraphPad Prism), Excel (Microsoft Corp), and SAS 9.4 (SAS Institute Inc).

AXL expression is associated with worse PFS and resistance to anti–PD-1 therapy

AXL tumor-cell staining was detectable in 54.4% of cases (172/316) and AXL expression was considered with high expression (H-score >50) in 23.7% of cases (75/316). Population characteristics of patients with AXLhigh versus AXLneg/low tumors were comparable in terms of gender, age at diagnosis, metastatic features, radiotherapy, IMDC, or Eastern Cooperative Oncology Group (ECOG) Performance Status (Table 1). Median follow-up was 23.5 months (95% CI, 22.6 to 24.0) and median PFS was 4.5 months (95% CI, 2.9 to 5.2). ORR was 25% for the evaluable cases (78/312). In the AXLhigh group of patients (H-score > 50), median PFS was shorter (2.8 months; 95% CI, 2.5 to 4.8) than in the AXLneg/low group (4.6 months; 95% CI, 3.5 to 5.3; P = 0.11). HR for disease progression was 1.26 (95% CI, 0.95 to 1.68), and 12-month PFS rates were 21.3% (12.9–31.2) for AXLhigh compared with 28.6% (23.0–34.4) for the AXLneg/low group (Fig. 1A). On multivariate analysis adjusting for age, sex, IMDC, and number of previous therapy lines (2 vs. >2), high AXL expression was associated with a trend toward reduced PFS (HR = 1.28; 95% CI, 0.96 to 1.71; P = 0.098; Supplementary Table S1). The ORR in patients with AXLhigh tumors was 15.5% (11/71) versus 27.8% (65/234) in patients with AXLneg/low tumors, indicating an inverse association between AXL expression and response to anti–PD-1 (Fig. 1A, P = 0.036). For OS, univariate and multivariate analyses indicated no statistical difference (Fig. 1A; Supplementary Table S2). Additional analysis showed that patients whose tumors exhibited AXLlow or AXLneg profiles had similar clinical outcomes and characteristics (Supplementary Fig. S2A; Supplementary Table S3). No association was observed between AXL staining in immune cells and OS, PFS, or ORR (not shown).

Table 1.

Patient characteristics.

AXL tumor-cell score
(0;50)>50All patients
Variablesn = 241n = 75n = 316Tests
Age (years) 
 N 241 75 316 t test 
 Mean (std) 62.6 (10.9) 61.9 (10.4) 62.4 (10.7) 0.608 
 Median (min; max) 64.0 (22.0; 87.0) 62.0 (41.0; 86.0) 64.0 (22.0; 87.0)  
Gender 
 Male 194 (80.5%) 63 (84.0%) 257 (81.3%) χ2 
 Female 47 (19.5%) 12 (16.0%) 59 (18.7%) P = 0.497 
IMDC group 
 Favorable 42 (17.4%) 12 (16.0%) 54 (17.1%) χ2 
 Intermediate 142 (58.9%) 49 (65.3%) 191 (60.4%) P = 0.580 
 Poor 57 (23.7%) 14 (18.7%) 71 (22.5%)  
ECOG 
 Missing 11  
 0 or 1 199 (85.4%) 63 (87.5%) 262 (85.9%) χ2 
 2 or 3 34 (14.6%) 9 (12.5%) 43 (14.1%) P = 0.656 
M stage 
 Missing 14 19  
 0 83 (36.6%) 28 (40.0%) 111 (37.4%) χ2 
 1 59 (26.0%) 20 (28.6%) 79 (26.6%) P = 0.657 
 X 85 (37.4%) 22 (31.4%) 107 (36.0%)  
Fuhrman 
 Missing data  
 I, II 69 (29.5%) 13 (17.8%) 82 (26.7%) χ2 
 III, IV 165 (70.5%) 60 (82.2%) 225 (73.3%) P = 0.049 
Brain metastasis 
 Missing 16 19  
 No 197 (87.6%) 65 (90.3%) 262 (88.2%) χ2 
 Yes 28 (12.4%) 7 (9.7%) 35 (11.8%) P = 0.533 
Surgery 241 (100%) 75 (100%) 316 (100%)  
Nephrectomy 
 No 9 (3.7%) 3 (4.0%) 12 (3.8%) χ2 
 Yes 232 (96.3%) 72 (96.0%) 304 (96.2%) P = 0.916 
Radiotherapy 
 No 148 (61.4%) 47 (62.7%) 195 (61.7%) χ2 
 Yes 93 (38.6%) 28 (37.3%) 121 (38.3%) P = 0.845 
AXL tumor-cell score
(0;50)>50All patients
Variablesn = 241n = 75n = 316Tests
Age (years) 
 N 241 75 316 t test 
 Mean (std) 62.6 (10.9) 61.9 (10.4) 62.4 (10.7) 0.608 
 Median (min; max) 64.0 (22.0; 87.0) 62.0 (41.0; 86.0) 64.0 (22.0; 87.0)  
Gender 
 Male 194 (80.5%) 63 (84.0%) 257 (81.3%) χ2 
 Female 47 (19.5%) 12 (16.0%) 59 (18.7%) P = 0.497 
IMDC group 
 Favorable 42 (17.4%) 12 (16.0%) 54 (17.1%) χ2 
 Intermediate 142 (58.9%) 49 (65.3%) 191 (60.4%) P = 0.580 
 Poor 57 (23.7%) 14 (18.7%) 71 (22.5%)  
ECOG 
 Missing 11  
 0 or 1 199 (85.4%) 63 (87.5%) 262 (85.9%) χ2 
 2 or 3 34 (14.6%) 9 (12.5%) 43 (14.1%) P = 0.656 
M stage 
 Missing 14 19  
 0 83 (36.6%) 28 (40.0%) 111 (37.4%) χ2 
 1 59 (26.0%) 20 (28.6%) 79 (26.6%) P = 0.657 
 X 85 (37.4%) 22 (31.4%) 107 (36.0%)  
Fuhrman 
 Missing data  
 I, II 69 (29.5%) 13 (17.8%) 82 (26.7%) χ2 
 III, IV 165 (70.5%) 60 (82.2%) 225 (73.3%) P = 0.049 
Brain metastasis 
 Missing 16 19  
 No 197 (87.6%) 65 (90.3%) 262 (88.2%) χ2 
 Yes 28 (12.4%) 7 (9.7%) 35 (11.8%) P = 0.533 
Surgery 241 (100%) 75 (100%) 316 (100%)  
Nephrectomy 
 No 9 (3.7%) 3 (4.0%) 12 (3.8%) χ2 
 Yes 232 (96.3%) 72 (96.0%) 304 (96.2%) P = 0.916 
Radiotherapy 
 No 148 (61.4%) 47 (62.7%) 195 (61.7%) χ2 
 Yes 93 (38.6%) 28 (37.3%) 121 (38.3%) P = 0.845 

Abbreviations: max, maximum; min, minimum.

Figure 1.

High AXL expression is associated with a trend to shorter PFS and lower ORR in patients with advanced ccRCC treated with nivolumab. A, Kaplan–Meier survival curves for PFS and OS of patients with RCC when stratified according to AXL H-score into 2 groups, negative/Low, ≤50 vs. High >50. Corresponding ORR, HRs, median PFS (mPFS), and survival estimates at 12 months posttreatment are shown. mo, months. B, Kaplan–Meier PFS and OS curves, and outcome measures deriving from RNA-seq data available for a subset of cases. Nivolumab-treated patients with RCC divided into 2 groups based on a cutoff ≥ highest 1/3 (High) vs. ≤ lowest 2/3 (Low) AXL mRNA expression to assess the impact of AXL on survival outcomes. C, Kaplan–Meier PFS and OS curves derived from CheckMate RNA-seq data available from the study of Braun (11). Nivolumab-treated patients with RCC were considered and divided into AXLhigh or AXLlow expression groups as in (B).

Figure 1.

High AXL expression is associated with a trend to shorter PFS and lower ORR in patients with advanced ccRCC treated with nivolumab. A, Kaplan–Meier survival curves for PFS and OS of patients with RCC when stratified according to AXL H-score into 2 groups, negative/Low, ≤50 vs. High >50. Corresponding ORR, HRs, median PFS (mPFS), and survival estimates at 12 months posttreatment are shown. mo, months. B, Kaplan–Meier PFS and OS curves, and outcome measures deriving from RNA-seq data available for a subset of cases. Nivolumab-treated patients with RCC divided into 2 groups based on a cutoff ≥ highest 1/3 (High) vs. ≤ lowest 2/3 (Low) AXL mRNA expression to assess the impact of AXL on survival outcomes. C, Kaplan–Meier PFS and OS curves derived from CheckMate RNA-seq data available from the study of Braun (11). Nivolumab-treated patients with RCC were considered and divided into AXLhigh or AXLlow expression groups as in (B).

Close modal

To further explore the link between AXL expression and clinical outcomes, we exploited the RNA-seq data available for a subset of the cases (n = 79). Despite the small number of cases, we recapitulated our findings. A higher AXL expression in the specimens tended to be associated with reduced 12-month PFS (P = 0.15) and OS rates (P = 0.63; Fig. 1B). We then interrogated expression data from the Braun and colleagues study (11). This cohort included patients with advanced RCC enrolled in CheckMate (CM-009, CM-010, and CM-025 trials) who were treated with an mTOR inhibitor (Everolimus), or nivolumab after progression on at least one prior antiangiogenic treatment (11). The potential effect of high AXL expression on survival and ORR was examined in the anti–PD-1–treated arm. High AXL expression was associated with worsened OS under anti–PD-1 (HR for death, 1.46; 95% CI, 1.00 to 2.22, P < 0.05; Fig. 1C). Median OS was 19.8 months in patients with tumors expressing high levels of AXL mRNA versus 35.1 months in cases displaying reduced levels of AXL mRNA. A trend was observed towards shorter PFS under anti–PD-1 blockade and lower ORR in tumors with high AXL mRNA expression (19%, 11/58 cases) compared with tumors with low AXL expression (24.56%, 28/114), although the differences could not reach statistical significance.

Together, these findings suggest that AXL, especially when highly expressed in cancer cells, may be predictor of tumors with reduced response rate and worse progression following PD-1 blockade.

AXL expression is associated with increased tumor expression of PD-L1

We then explored the TME composition of these tumors by investigating associations between AXL and various markers as assessed by IHC. This included VEGFA, PD-L1, and markers of the most common immune-cell populations CD3, CD8 (T cells), CD163 (macrophages), and CD20 (B-cells). PD-L1 tumor-cell staining was detected in 23.9% (76/317) of evaluable ccRCC cases and PD-L1 on TILs in 58,6% (186/317) of cases. We then stratified cases according to AXL expression and PD-L1 expression status in tumor cells, and tested putative associations between the two markers (Fig. 2A). A significant enrichment of PD-L1 positivity within tumors with AXLhigh profile was observed (P = 0.0031). Moreover, the mean percentage of PD-L1–positive tumor cells was increased in the AXLhigh group compared with tumors with AXLneg/low profiles (P = 0.0019; Fig. 2A; Supplementary Fig. S2B). In contrast, no significant associations were observed between AXL and PD-L1 TIL expression (not shown, p > 0.99). AXL expression was associated with a greater proportion of tumor cells positive for VEGF (P = 0.002; Fig. 2B). We observed a trend toward a higher level of tumor-infiltrating CD8+ T cells in AXLhigh versus AXLneg/low cases, and no association between AXL and CD3, CD163, CD20 densities in the tumor core (Fig. 2B and C; Supplementary Fig. S2B). RNA-seq data revealed a trend toward greater inflammation (P = 0.072), immune suppression (P = 0.105), and monocytic-lineage (P = 0.052) signatures in AXLhigh tumors (Fig. 2D).

Figure 2.

AXL is associated with PD-L1 expression. A, Distribution of PD-L1 (% of cases and % of positive cells) in tumors with AXLneg/low vs. AXLhigh profiles. B, Distribution of VEGF, CD8, CD3, CD163, CD20 expression in AXLNeg/Low vs. AXLhigh cases. C, A heatmap is shown to illustrate the biomarker composition across the tumor specimens according to AXL groups. D, RNA-seq–derived expression of various gene signatures in AXLneg/low vs. high cases. Violin plots are shown with box representing the 25th to 75th percentiles and central line representing the median (Mann–Whitney test).

Figure 2.

AXL is associated with PD-L1 expression. A, Distribution of PD-L1 (% of cases and % of positive cells) in tumors with AXLneg/low vs. AXLhigh profiles. B, Distribution of VEGF, CD8, CD3, CD163, CD20 expression in AXLNeg/Low vs. AXLhigh cases. C, A heatmap is shown to illustrate the biomarker composition across the tumor specimens according to AXL groups. D, RNA-seq–derived expression of various gene signatures in AXLneg/low vs. high cases. Violin plots are shown with box representing the 25th to 75th percentiles and central line representing the median (Mann–Whitney test).

Close modal

High AXL combined with PD-L1 expression is associated with poor survival in patients treated with PD-1 blockade

We set out to determine the extent to which combining AXL and PD-L1 markers would impact on OS under anti–PD-1 treatment (Fig. 3). PD-L1–positive staining in tumor-cell alone (using cutoff 0;0.5 vs. > = 1) associated with worse 12-month survival rate (70%; 95% CI 58.6–78.8 vs. 77.3; 95% CI, 71.5–82.1; with HR for death 1.51; log–rank P = 0.02; Fig. 3A). No association was found with PFS (HR = 1.10, P = 0.5) or the ORR (28.6% in PD-L1+ vs. 23.9% in PD-L1neg subsets; Supplementary Fig. S3A). PD-L1 staining in TILs had limited impact on OS (12-month survival rate 73.3%, 95% CI, 66.4–79 vs. 78.7, 95% CI, 70.7–84.8; HR to death 1.34; log-rank P = 0.09), no impact on PFS (P = 0.16), nor on the ORR (25.1% in PD-L1+ vs. 25% in PD-L1neg; Fig. 3B; Supplementary Fig. S3B).

Figure 3.

High AXL expression plus PD-L1 TC positivity is associated with worse OS in patients treated with nivolumab. A, Kaplan–Meier estimates for OS according to PD-L1 TC status (pos/neg), or AXL expression and PD-L1 TC status (four biomarker subgroups). B, Kaplan–Meier estimates for OS according to PD-L1 TIL status (pos/neg), or AXL expression and PD-L1 TIL status (4 biomarker subgroups). pos, positive; neg, negative.

Figure 3.

High AXL expression plus PD-L1 TC positivity is associated with worse OS in patients treated with nivolumab. A, Kaplan–Meier estimates for OS according to PD-L1 TC status (pos/neg), or AXL expression and PD-L1 TC status (four biomarker subgroups). B, Kaplan–Meier estimates for OS according to PD-L1 TIL status (pos/neg), or AXL expression and PD-L1 TIL status (4 biomarker subgroups). pos, positive; neg, negative.

Close modal

The combination of the two markers AXL and PD-L1 on tumor cells had significant impact on OS since the group of patients whose tumors exhibited both high levels of AXL and PD-L1 tumor-cell positivity had the lowest survival rate at 12 months 60.7% (40.4–76.0%), HR to death 1.98 (1.17–3.35), log–rank P = 0.046, compared with the 3 other groups (AXLneg/low/PD-L1+; AXLneg/low/PD-L1neg; AXLhigh/PD-L1neg; Fig. 3A). On multivariate analysis adjusting for age, sex, IMDC risk groups, and number of previous therapy lines (≤2 vs. >2), high AXL combined with PD-L1 expression was associated with a trend to reduced OS (HR = 2.01, 95% CI, 1.18–3.44, P = 0.084; Supplementary Table S5). We further interrogated the prognostic value of combining AXL and PD-L1 TC markers specifically in patients with IMDC intermediate-risk/poor-risk RCC. In this setting, a AXLhigh/PD-L1+ TC profile was associated with the lowest OS rate at 12 months 56.0% (34.8–72.7%) and HR to death 2.05 (1.21–3.50), P = 0.053 (Supplementary Fig. S4).

There was no significant difference between ORR across the different groups (P = 0.123), and AXL/PD-L1 TC combination had little if no further influence on PFS compared with AXL alone (Supplementary Fig. S3A). However, it is noteworthy that the combination of high AXL expression with PD-L1 staining in TILs displayed notable features, with patients in the AXL-high/PD-L1+ TIL group showing the worst clinical outcomes particularly in terms of ORR (12%, P = 0.04) and PFS, with relatively low PFS rate at 12 months (15.9%, 7.0–28.1%, P = 0.14; Supplementary Fig. S3B).

We examined the repartition of the different markers in each subgroup, namely, PD-L1 tumor, PD-L1 TIL, VEGF, CD3, CD8, CD163, and CD20. Patients whose tumors were positive for PD-L1 had increased immune-cell infiltration. Hence, both AXLhigh and AXLneg/low cases within the PD-L1+ subset demonstrated high immune-cell infiltration compared with PD-L1neg groups (Fig. 4A and B).

Figure 4.

Biomarker distribution across groups stratified based on PD-L1 TC and AXL expression. A, Bar graphs comparing the biomarker composition in the corresponding subgroups. Data are presented as means. Errors bars are SEM Mann–Whitney test. B, A heatmap showing the biomarker distribution across the tumor specimens in the indicated subgroups.

Figure 4.

Biomarker distribution across groups stratified based on PD-L1 TC and AXL expression. A, Bar graphs comparing the biomarker composition in the corresponding subgroups. Data are presented as means. Errors bars are SEM Mann–Whitney test. B, A heatmap showing the biomarker distribution across the tumor specimens in the indicated subgroups.

Close modal

These data indicate that AXL and PD-L1 staining can identify a particular group of patients with poor outcomes when receiving anti–PD-1 treatment.

VHL status does not interfere with clinical outcomes or AXL expression in advanced ccRCC treated with nivolumab

Since VHL is the most common genetic alteration in RCC, we characterized VHL alterations in the ccRCC cases of the NIVOREN cohort. Point mutations were found in 76.7% (231/301), LOH in 84.1% (254/302), and promoter hypermethylation in 19.1% (27/141) of cases. VHL status could be ascertained in 260 cases. Seventy-four percent of cases (193/260) were found to have a biallelic alteration of VHL which supposes LOF of the gene. Among the remaining cases, 7.7% (20/260) presented with no VHL alteration (+/+), and 18% (47/260) had one VHL alteration (±) implying one remaining functional copy. No significant differences in ORR (χ2P = 0.40), PFS (HR 1.00, 0.73–1.37, P = 0.98), or OS (0.92, 0.61–1.38, P = 0.68) were observed between patients with biallelic VHL inactivation (VHLneg/neg) and those presenting with at least one unaltered VHL copy (VHL+/+ and VHL+/neg; Supplementary Fig. S5A and S5B).

AXL expression was homogeneously distributed across cases with or without biallelic VHL alterations excluding the possibility that AXL expression is dependent on VHL status (Supplementary Fig. S5C). When stratified based on VHL status, a trend was found towards increased PD-L1 positivity in TILs within tumor specimens harboring a VHLneg/neg profile (Supplementary Fig. S5C), whereas VHL+/+/VHL+/neg tumors tended to be enriched with PD-L1+ tumor cells (Supplementary Fig. S5C) in line with a previous study (22). Analysis of the Nivoren RNA-seq dataset (24) indicated that VHL+/+/VHL+/neg tumors expressed more PD-L1 (P = 0.01) and VHL mRNAs (P = 0.03; Supplementary Fig. S5D). As expected, VHLneg/neg tumors exhibited increased angiogenesis (P = 0.02) and hypoxia signatures (P = 0.02). VHL+/+/VHL+/neg tumors displayed a higher VEGF tumor expression, suggesting a VHL/HIF-independent regulation of VEGF in this group, as previously proposed (29; Supplementary Fig. S6A).Together, this data support that despite similar clinical outcomes under PD-1 blockade, VHLneg/neg, and VHL+/+/VHL+/neg tumors are biologically different and associated with a distinct microenvironment.

Association between AXL, PD-L1, and survival according to VHL status

We then questioned whether AXL staining would yield comparable prognostic and biologic insights in patients with complete inactivation of VHL compared with those with a noninactivated status. We categorized cases into 4 groups based on AXL expression more than 50 and the VHL status. Patients whose tumors exhibited high AXL expression and VHLneg/neg appeared to have the worst median PFS and 12-month PFS compared with cases with AXLneg/low expression within VHLneg/neg subset, though the overall difference was not significant in the log–rank tests comparing the 4 groups (Fig. 5A). In the VHLneg/neg subset, a separation in the PFS curves was evident between AXLhigh versus AXLneg/low groups after approximately 5 months and continued to show differences in time as supported by the 12-month PFS rates of 19.6% versus 29.9%, respectively (Fig. 5A). In contrast, in the VHL+/+/VHL+/neg tumors, there were no signs of such differences. Of note, when interrogating RNA-seq data for a subset of patients, a trend in the same direction was observed toward a negative impact for AXL on PFS in the inactivated VHL subset but not in the noninactivated VHL subset (Fig. 5B). We then conducted a similar analysis using the Checkmate RNA dataset and generated Kaplan–Meier curves for PFS and OS according to tumor-VHL status. AXLhigh expression associated with worsened survival (PFS and OS) compared with AXLneg/low in the inactivated VHL group (n = 78; P = 0.05), whereas no difference was observed in patients conserving at least one WT copy of VHL in their tumor (n = 44; P = 0.49; Fig. 5C). Therefore, the influence of AXL on clinical outcomes was mainly apparent in a VHLneg/neg tumors. The results also suggest that both OS and PFS could be impacted by high AXL expression in this setting.

Figure 5.

AXL association with PD-L1 and outcomes in VHLneg/neg and VHL+/+/VHL+/neg tumors. A, Kaplan–Meier estimates for PFS in patients divided into 4 groups based on AXL expression and VHL status. In the VHLneg/neg subset, patients with tumors expressing high levels of AXL (red line) trended to worse PFS compared with those with AXLNull/Low expression (gray line). B, Kaplan–Meier PFS curves generated after categorization using RNA-seq data on a subset of cases in VHL-inactivated (VHLneg/neg) and VHL-noninactivated (VHLneg/+/VHL+/+) cases. C, Kaplan–Meier estimates for PFS and OS as in B, but using the CheckMate dataset (11). P values from log–rank tests are shown. D, AXL expression associated with PD-L1 TC expression especially in the VHLneg/neg tumors. Distribution of PD-L1 (% of cases and % of positive cells) across the tumor specimens according to the indicated subgroups. Data are presented as means. Errors bars are SEM. χ2 and Mann–Whitney tests were applied.

Figure 5.

AXL association with PD-L1 and outcomes in VHLneg/neg and VHL+/+/VHL+/neg tumors. A, Kaplan–Meier estimates for PFS in patients divided into 4 groups based on AXL expression and VHL status. In the VHLneg/neg subset, patients with tumors expressing high levels of AXL (red line) trended to worse PFS compared with those with AXLNull/Low expression (gray line). B, Kaplan–Meier PFS curves generated after categorization using RNA-seq data on a subset of cases in VHL-inactivated (VHLneg/neg) and VHL-noninactivated (VHLneg/+/VHL+/+) cases. C, Kaplan–Meier estimates for PFS and OS as in B, but using the CheckMate dataset (11). P values from log–rank tests are shown. D, AXL expression associated with PD-L1 TC expression especially in the VHLneg/neg tumors. Distribution of PD-L1 (% of cases and % of positive cells) across the tumor specimens according to the indicated subgroups. Data are presented as means. Errors bars are SEM. χ2 and Mann–Whitney tests were applied.

Close modal

Finally, we sought to examine the putative interaction between AXL and PD-L1 according to tumoral VHL status. The enrichment for PD-L1 positivity in AXLhigh cases was restricted to VHLneg/neg tumors (P = 0.0003) and accompanied by a significant increase in the percentage of PD-L1+ tumor cells in AXLhighVHLneg/neg cases (Fig. 5D). In VHL+/+/VHL+/neg cases, PD-L1 staining was detected, but the proportion of positive cases was identical in both the AXLneg/low and AXLhigh groups (p > 0.99). In contrast, the relationship between AXL with VEGF was perceptible in both VHLneg/neg and VHL+/+/VHL+/neg subsets (Supplementary Fig. S6B). We observed no particular effects of AXL on the immune infiltrate (CD8, CD3, CD163, CD20) in the VHLneg/neg and VHL+/+/VHL+/neg specimens(Supplementary Fig. S6B and S6C).

Immunotherapy approaches are increasingly used in clinical practice. Yet many challenges remain to improve efficacy and RR. Most patients exhibit primary or acquired resistance to immunotherapy for reasons that remain unclear. The development of markers predictive for response or resistance is warranted. While recent studies have already shown promising results of immunotherapies as first-line therapy, it will be critical to provide the best appropriate combinatorial therapy for each patient, and this implies the development of tools to capture the unique microenvironment of their tumor.

In this study, patients with ccRCC tumors exhibiting high expression levels of AXL had poorer clinical response (PFS and ORR) after Nivolumab treatment, but not altered OS. The median follow-up in this study was 23.5 months. Outcome measures derive from a still-active multicenter phase II study. Extended follow-up should better inform on the impact of AXL on OS. A final analysis at a 4-year minimum follow-up is planned. The findings obtained from the analysis of the CheckMate dataset are concordant with a role for AXL on OS.

AXL expression was strongly associated with PD-L1 TC positivity on tumor cells, and patients with tumors displaying concomitant PD-L1 TC expression and high AXL expression had the worst OS. We found no significant differences for PFS. Such discordance between PFS and OS should be interpreted with caution since prior work indicates that PFS may not be an ideal predictor of OS in the setting of anti–PD-1 monotherapy (9).

The absence of correlation between PD-L1 expression and ICB therapy-response in this cohort is interesting. It could be due to possible interference between factors such as AXL and PD-L1 expressed by tumor cells or TILs, or discrepancies in PD-L1 expression in different tumor sites. Intriguingly, patients with tumors exhibiting high AXL TC expression and PD-L1 positivity in TILs demonstrated a much lower response rate, suggesting that AXL might play a substantial role in tumors infiltrated with PD-L1–expressing TILs. The combination of AXL and PD-L1 biomarkers could identify various subgroups of patients displaying high or reduced tumor-infiltrating immune cells. The findings support the hypothesis that AXL may be one factor interfering with the predictive value of PD-L1, though it's not clear yet whether such effects could act directly on tumor cells, TILs or indirectly via the TME components, a question that should be addressed in future investigations.

It will be important to explore the connections between AXL, microenvironment components, and therapeutic outcomes in the context of other treatment regimens, namely the anti–PD-1/anti-CTLA4 combination emerging as a standard-of-care first-line therapy option (10, 30), anti–PDL1–based combinations used to treat patients with advanced RCCs (5, 31, 32), as well as anti–PD-1–based combinations integrating agents targeting AXL such as cabozantinib, expected to target VEGFR2, MET, RET, and AXL (5, 33). Recent data have demonstrated activity of cabozantinib after ICB in m-ccRCC (34). This treatment might enable to overcome the deleterious effects of AXL.

Although AXL expression was not significantly associated with tumor infiltration by macrophages, B or T cells, a broader assessment of the TME and immune checkpoints is required. Our RNA-seq analysis suggested potential interactions between AXL expression and increased inflammation, immune suppression and monocytic-lineage signatures. Such analysis likely suffered from the small number of cases analyzed. Studies using sizeable cohorts of patients will be needed to validate these findings. Multiplex IHC or IF approaches would be needed to fully characterize tumor and immune-cell infiltrates in this cohort, dissect intratumor heterogeneity, and test potential associations with therapeutic outcomes.

Our analysis is retrospective and has other potential limitations. A number of cases with FFPE block showing less than 40% of tumor or more than 40% of necrosis were excluded. The archival tumor specimens were often obtained before antiangiogenic therapy or at an early-stage disease. This is in line with routine clinical practice for treatment decision-making. However, a longitudinal survey with a multiple sampling scheme may be necessary to fully appreciate the TME composition and monitor the influence of previous treatments. Diseases with specific features (e.g, Sarcomatoid, low-grade, IMDC favorable-risk groups) also need further investigations, as the current study was limited by the number of cases in these categories. As reported by Boysen and colleagues (35), we noted a trend toward a greater sarcomatoid differentiation in AXL-positive cases (n = 25, P = 0.09) as well as an enrichment of high-grade tumors (Fuhrman III and IV) in cases with high AXL expression.

Interestingly, AXL association with PD-L1 and worse PFS was observed in tumors harboring VHLneg/neg genetic profile, but it has been unclear if this also occurs in those with VHL+/+/VHL+/neg status where the number of cases was lower. It is postulated that the TME from VHLneg/neg tumors differs from those displaying VHL+/+ or VHL+/neg. It would be interesting to see if among the identified subgroups, some are enriched for genomic alterations reported with values for predicting anti–PD-1 responsiveness including BAP1 and PBRM1 (11). This could explain, in part, some of the demonstrated effects on PFS, OS, or ORR. Previous studies have demonstrated that loss of PBRM1 functions can accentuate the transcriptional response of HIF in VHLneg/neg RCC (36). We postulate that a molecular link exists between AXL and loss of PBRM1 in VHLneg/neg RCC tumors.

Moreover, tumor cells may have intrinsic characteristics which likely affect the TME. AXL expression has been associated with EMT, epithelial–mesenchymal plasticity, and phenotypic changes of tumor cells in various cancer types (37). Compelling evidence also exists that differentiation of tumor cells to a more mesenchymal phenotype can greatly influence the TME via increased secretion of immunosuppressive substances such as TGF-beta, VEGF, PD-L1, or IL8 (38–43). In a lung-cancer study, Thompson and colleagues observed that non–small cell lung cancer (NSCLC) tumors displaying a more mesenchymal phenotype had reduced clinical responses to PD-1 blockade. From this work, the investigators derived an EMT/inflammation-based gene-expression signature for predicting clinical response of patients with metastatic NSCLC (44). In a cohort of patients with metastatic urothelial cancer treated with nivolumab, Wang and colleagues found that particularly in patients with T-cell infiltrated tumors, higher EMT/stromal-related gene expression was associated with lower RRs and poor outcomes (45). Moreover, Chakravarthy and colleagues showed that an extracellular matrix (ECM) transcriptional program involving TGF-β signalling and EMT-related genes is associated with worse prognosis in patients with melanoma and bladder cancer treated with anti–PD-1 (46). In melanoma, Hugo and colleagues identified a gene-expression signature associated with innate PD-1 resistance (IPRES) of metastatic melanoma which involves the upregulation of EMT-related factors, including AXL, immunosuppressive cytokines, hypoxia, and proangiogenic factors (47).

In our survey, AXL staining in immune cells had no prognostic value. However, because the panel of markers analyzed was limited, we cannot exclude the possibility that a particular immune-cell subpopulation expressing AXL contributes to anti–PD-1 responsiveness. Others have reported evidence of associations between AXL-expressing immune-cell populations, and the establishment of an immunosuppressed TME (48–50). Future studies are warranted to elucidate the interplay between pseudo-hypoxia status, epithelial–mesenchymal plasticity, AXL, microenvironmental changes, and their impact on anti–PD-1 responsiveness.

In summary, we identified associations between expression of AXL, PD-L1 expression, and VHL status in advanced ccRCC which could contribute to disease progression under ICB therapy. Furthermore, we demonstrated that the combined assessment of AXL and PD-L1 expression may be useful to predict outcomes and discriminate patients who are more likely to be resistant or to benefit from anti–PD-1 immunotherapy.

N. Rioux-Leclercq reports personal fees from Ipsen and BMS outside the submitted work. J. Adam reports personal fees from MSD and AstraZeneca outside the submitted work. J.B. Lorens reports personal fees from BerGenBio ASA outside the submitted work; in addition, J.B. Lorens has a patent application pending to BerGenBio ASA. G. Gausdal is an employee at BerGenBio ASA, which has supported this work financially. N. Chaput reports personal fees from AstraZeneca and grants from Sanofi and GSK outside the submitted work. B. Beuselinck reports grants from BMS and UNICANCER during the conduct of the study as well as other support from BMS, MSD, AstraZeneca, Ipsen and Merck outside the submitted work. Y. Vano reports personal fees from BMS, MSD, Pfizer, Ipsen, Merck, Novartis, Janssen, Roche and Sanofi outside the submitted work. L. Albiges reports grants from BMS during the conduct of the study as well as other support from Pfizer, Novartis, BMS, Ipsen, Astellas, MSD, Merck & Co, Janssen, AstraZeneca, and Eisai outside the submitted work. No disclosures were reported by the other authors.

S. Terry: Conceptualization, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C. Dalban: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. N. Rioux-Leclercq: Resources, data curation, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. J. Adam: Validation, investigation, methodology, writing–review and editing. M. Meylan: Data curation, software, formal analysis, investigation, visualization, methodology, writing–review and editing. S. Buart: Investigation, methodology. A. Bougoüin: Formal analysis, investigation, methodology, writing–review and editing. A. Lespagnol: Formal analysis, investigation, methodology, writing–review and editing. F. Dugay: Investigation, methodology, writing–review and editing. I.C. Moreno: Formal analysis, investigation. G. Lacroix: Investigation, methodology. J.B. Lorens: Conceptualization, writing–review and editing. G. Gausdal: Writing–review and editing. W.H. Fridman: Supervision, writing–review and editing. F. Mami-Chouaib: Resources, writing–review and editing. N. Chaput: Writing–review and editing. B. Beuselinck: Resources, formal analysis, investigation, writing–review and editing. S. Chabaud: Formal analysis, supervision, validation. J. Barros-Monteiro: Data curation, formal analysis, investigation, project administration. Y. Vano: Supervision, investigation. B. Escudier: Conceptualization, writing–review and editing. C. Sautès-Fridman: Resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–review and editing. L. Albiges: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing. S. Chouaib: Conceptualization, supervision, funding acquisition, investigation, writing–review and editing.

This work was supported by INSERM, Université Paris-Saclay, Cancéropôle Ile-de-France, Sorbonne Université, Université de Paris, la Ligue Contre le Cancer (EL2015.LNCC/SaC, to S. Chouaib), l'Institut National du Cancer and Direction générale de l'offre de soins (grant INCa-DGOS-PRT-K16–181, GETUG-AFU26-Nivoren translational program, to L. Albiges), CARPEM program of the Sites Intégrés de Recherche sur le Cancer (SIRIC, to C. Sautès-Fridman), and the association pour la recherche en thérapeutiques innovantes en cancérologie (ARTIC). M. Meylan received a PhD fellowship from ARTIC. We thank Florence Tantot (Unicancer) for participating in the translational research program; Sophie Ferlicot (Hôpital de Bicêtre, Paris-Sud 11) for providing material in preliminary studies; and Laurence Cornevin, Roselyne Viel, and Pascale Bellaud (Plateforme H2P2, Biosit, Université de Rennes, Rennes, France) for outstanding technical assistance.

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.

1.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
. 
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
2.
Rini
BI
,
Plimack
ER
,
Stus
V
,
Gafanov
R
,
Hawkins
R
,
Nosov
D
, et al
Pembrolizumab plus axitinib versus sunitinib for advanced renal-cell carcinoma
.
N Engl J Med
2019
;
380
:
1116
27
.
3.
Motzer
R
,
Alekseev
B
,
Rha
S-Y
,
Porta
C
,
Eto
M
,
Powles
T
, et al
Lenvatinib plus Pembrolizumab or everolimus for advanced renal cell carcinoma
.
N Engl J Med
2021
;
384
:
1289
300
.
4.
Choueiri
TK
,
Powles
T
,
Burotto
M
,
Escudier
B
,
Bourlon
MT
,
Zurawski
B
, et al
Nivolumab plus cabozantinib versus sunitinib for advanced renal-cell carcinoma
.
N Engl J Med
2021
;
384
:
829
41
.
5.
Xu
W
,
Atkins
MB
,
McDermott
DF
. 
Checkpoint inhibitor immunotherapy in kidney cancer
.
Nat Rev Urol
2020
;
17
:
137
50
.
6.
Sharma
P
,
Hu-Lieskovan
S
,
Wargo
JA
,
Ribas
A
. 
Primary, adaptive, and acquired resistance to cancer immunotherapy
.
Cell
2017
;
168
:
707
23
.
7.
Upadhaya
S
,
Neftelino
ST
,
Hodge
JP
,
Oliva
C
,
Campbell
JR
,
Yu
JX
. 
Combinations take centre stage in PD1/PDL1 inhibitor clinical trials
.
Nat Rev Drug Discov
2021
;
20
:
168
9
.
8.
Giraldo
NA
,
Becht
E
,
Pages
F
,
Skliris
G
,
Verkarre
V
,
Vano
Y
, et al
Orchestration and prognostic significance of immune checkpoints in the microenvironment of primary and metastatic renal cell cancer
.
Clin Cancer Res
2015
;
21
:
3031
40
.
9.
Motzer
RJ
,
Escudier
B
,
McDermott
DF
,
George
S
,
Hammers
HJ
,
Srinivas
S
, et al
Nivolumab versus Everolimus in advanced renal-cell carcinoma
.
N Engl J Med
2015
;
373
:
1803
13
.
10.
Motzer
RJ
,
Tannir
NM
,
McDermott
DF
,
Arén Frontera
O
,
Melichar
B
,
Choueiri
TK
, et al
Nivolumab plus Ipilimumab versus Sunitinib in advanced renal-cell carcinoma
.
N Engl J Med
2018
;
378
:
1277
90
.
11.
Braun
DA
,
Hou
Y
,
Bakouny
Z
,
Ficial
M
,
Sant' Angelo
M
,
Forman
J
, et al
Interplay of somatic alterations and immune infiltration modulates response to PD-1 blockade in advanced clear cell renal cell carcinoma
.
Nat Med
2020
;
26
:
909
18
.
12.
Miao
D
,
Margolis
CA
,
Gao
W
,
Voss
MH
,
Li
W
,
Martini
DJ
, et al
Genomic correlates of response to immune checkpoint therapies in clear cell renal cell carcinoma
.
Science
2018
;
359
:
801
6
.
13.
Hsieh
JJ
,
Le
VH
,
Oyama
T
,
Ricketts
CJ
,
Ho
TH
,
Cheng
EH
. 
Chromosome 3p loss-orchestrated VHL, HIF, and epigenetic deregulation in clear cell renal cell carcinoma
.
J Clin Oncol
2018
;
36
:
JCO2018792549
.
14.
Schoumacher
M
,
Burbridge
M
. 
Key roles of AXL and MER receptor tyrosine kinases in resistance to multiple anticancer therapies
.
Curr Oncol Rep
2017
;
19
:
19
.
15.
Peterfi
L
,
Bjercke
T
,
Yusenko
MV
,
Kovacs
G
,
Banyai
D
. 
Cytoplasmic expression of AXL Is associated with high risk of postoperative relapse of conventional renal cell carcinoma
.
Anticancer Res
2020
;
40
:
3485
9
.
16.
Landolt
L
,
Eikrem
Ø
,
Strauss
P
,
Scherer
A
,
Lovett
DH
,
Beisland
C
, et al
Clear cell renal cell carcinoma is linked to epithelial-to-mesenchymal transition and to fibrosis
.
Physiol Rep
2017
;
5
:
e13305
.
17.
Terry
S
,
Abdou
A
,
Engelsen
AST
,
Buart
S
,
Dessen
P
,
Corgnac
S
, et al
AXL targeting overcomes human lung cancer cell resistance to NK- and CTL-mediated cytotoxicity
.
Cancer Immunol Res
2019
;
7
:
1789
802
.
18.
Guo
Z
,
Li
Y
,
Zhang
D
,
Ma
J
. 
Axl inhibition induces the antitumor immune response which can be further potentiated by PD-1 blockade in the mouse cancer models
.
Oncotarget
2017
;
8
:
89761
74
.
19.
Aguilera
TA
,
Rafat
M
,
Castellini
L
,
Shehade
H
,
Kariolis
MS
,
Hui
AB
, et al
Reprogramming the immunological microenvironment through radiation and targeting Axl
.
Nat Commun
2016
;
7
:
13898
.
20.
Ludwig
KF
,
Du
W
,
Sorrelle
NB
,
Wnuk-Lipinska
K
,
Topalovski
M
,
Toombs
JE
, et al
Small-molecule inhibition of Axl targets tumor immune suppression and enhances chemotherapy in pancreatic cancer
.
Cancer Res
2018
;
78
:
246
55
.
21.
Flippot
R
,
Dalban
C
,
Laguerre
B
,
Borchiellini
D
,
Gravis
G
,
Négrier
S
, et al
Safety and efficacy of nivolumab in brain metastases from renal cell carcinoma: results of the GETUG-AFU 26 NIVOREN multicenter phase II study
.
J Clin Oncol
2019
;
37
:
2008
16
.
22.
Messai
Y
,
Gad
S
,
Noman
MZ
,
Le Teuff
G
,
Couve
S
,
Janji
B
, et al
Renal cell carcinoma programmed death-ligand 1, a new direct target of hypoxia-inducible factor-2 alpha, is regulated by von Hippel-Lindau gene mutation status
.
Eur Urol
2016
;
70
:
623
32
.
23.
Kammerer-Jacquet
SF
,
Crouzet
L
,
Brunot
A
,
Dagher
J
,
Pladys
A
,
Edeline
J
, et al
Independent association of PD-L1 expression with noninactivated VHL clear cell renal cell carcinoma-A finding with therapeutic potential
.
International journal of cancer.
2017
;
140
:
142
8
.
24.
Meylan
M
,
Beuselinck
B
,
Dalban
C
,
Vano
Y
,
Rioux Leclercq
N
,
Sautès-Fridman
C
, et al
700O kidney ccRCC immune classification (KIC) enhances the predictive value of T effector (Teff) and angiogenesis (Angio) signatures in response to nivolumab (N)
.
Ann Oncol
2020
;
31
:
S553
.
25.
McDermott
DF
,
Huseni
MA
,
Atkins
MB
,
Motzer
RJ
,
Rini
BI
,
Escudier
B
, et al
Clinical activity and molecular correlates of response to atezolizumab alone or in combination with bevacizumab versus sunitinib in renal cell carcinoma
.
Nat Med
2018
;
24
:
749
57
.
26.
Becht
E
,
Giraldo
NA
,
Lacroix
L
,
Buttard
B
,
Elarouci
N
,
Petitprez
F
, et al
Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression
.
Genome Biol
2016
;
17
:
218
.
27.
Buffa
FM
,
Harris
AL
,
West
CM
,
Miller
CJ
. 
Large meta-analysis of multiple cancers reveals a common, compact and highly prognostic hypoxia metagene
.
Br J Cancer
2010
;
102
:
428
35
.
28.
Hendry
S
,
Salgado
R
,
Gevaert
T
,
Russell
PA
,
John
T
,
Thapa
B
, et al
Assessing tumor-infiltrating lymphocytes in solid tumors: a practical review for pathologists and proposal for a standardized method from the international immunooncology biomarkers working group: part 1: assessing the host immune response, TILs in invasive breast carcinoma and ductal carcinoma in situ, metastatic tumor deposits and areas for further research
.
Adv Anat Pathol
2017
;
24
:
235
51
.
29.
Patard
J-J
,
Rioux-Leclercq
N
,
Masson
D
,
Zerrouki
S
,
Jouan
F
,
Collet
N
, et al
Absence of VHL gene alteration and high VEGF expression are associated with tumour aggressiveness and poor survival of renal-cell carcinoma
.
Br J Cancer
2009
;
101
:
1417
24
.
30.
Tannir
NM
,
Signoretti
S
,
Choueiri
TK
,
McDermott
DF
,
Motzer
RJ
,
Flaifel
A
, et al
Efficacy and safety of nivolumab plus ipilimumab versus sunitinib in first-line treatment of patients with advanced sarcomatoid renal cell carcinoma
.
Clin Cancer Res
2021
;
27
:
78
86
.
31.
Motzer
RJ
,
Robbins
PB
,
Powles
T
,
Albiges
L
,
Haanen
JB
,
Larkin
J
, et al
Avelumab plus axitinib versus sunitinib in advanced renal cell carcinoma: biomarker analysis of the phase 3 JAVELIN Renal 101 trial
.
Nat Med
2020
;
26
:
1733
41
.
32.
Motzer
RJ
,
Penkov
K
,
Haanen
J
,
Rini
B
,
Albiges
L
,
Campbell
MT
, et al
Avelumab plus axitinib versus sunitinib for advanced renal-cell carcinoma
.
N Engl J Med
2019
;
380
:
1103
15
.
33.
Choueiri
TK
,
Kaelin
WG
. 
Targeting the HIF2-VEGF axis in renal cell carcinoma
.
Nat Med
2020
;
26
:
1519
30
.
34.
McGregor
BA
,
Lalani
A-KA
,
Xie
W
,
Steinharter
JA
,
E Bakouny
Z
,
Martini
DJ
, et al
Activity of cabozantinib after immune checkpoint blockade in metastatic clear-cell renal cell carcinoma
.
Eur J Cancer
2020
;
135
:
203
10
.
35.
Boysen
G
,
Bausch-Fluck
D
,
Thoma
CR
,
Nowicka
AM
,
Stiehl
DP
,
Cima
I
, et al
Identification and functional characterization of pVHL-dependent cell surface proteins in renal cell carcinoma
.
Neoplasia
2012
;
14
:
535
46
.
36.
Gao
W
,
Li
W
,
Xiao
T
,
Liu
XS
,
Kaelin
WG
. 
Inactivation of the PBRM1 tumor suppressor gene amplifies the HIF-response in VHL−/− clear cell renal carcinoma
.
Proc Natl Acad Sci USA.
2017
;
114
:
1027
32
.
37.
Lotsberg
ML
,
Rayford
A
,
Thiery
JP
,
Belleggia
G
,
Peters
SD
,
Lorens
JB
, et al
Decoding cancer's camouflage: epithelial-mesenchymal plasticity in resistance to immune checkpoint blockade
. Cancer Drug Resist 
2020
;
3
:
832
53
.
38.
Terry
S
,
Savagner
P
,
Ortiz-Cuaran
S
,
Mahjoubi
L
,
Saintigny
P
,
Thiery
J-P
, et al
New insights into the role of EMT in tumor immune escape
.
Mol Oncol
2017
;
11
:
824
46
.
39.
Lou
Y
,
Diao
L
,
Cuentas
ER
,
Denning
WL
,
Chen
L
,
Fan
YH
, et al
Epithelial-mesenchymal transition is associated with a distinct tumor microenvironment including elevation of inflammatory signals and multiple immune checkpoints in lung adenocarcinoma
.
Clin Cancer Res
2016
;
22
:
3630
42
.
40.
Mak
MP
,
Tong
P
,
Diao
L
,
Cardnell
RJ
,
Gibbons
DL
,
William
WN
, et al
A patient-derived, pan-cancer EMT signature identifies global molecular alterations and immune target enrichment following epithelial-to-mesenchymal transition
.
Clin Cancer Res
2016
;
22
:
609
20
.
41.
Dongre
A
,
Rashidian
M
,
Reinhardt
F
,
Bagnato
A
,
Keckesova
Z
,
Ploegh
HL
, et al
Epithelial-to-mesenchymal transition contributes to immunosuppression in breast carcinomas
.
Cancer Res
2017
;
77
:
3982
9
.
42.
Chen
L
,
Gibbons
DL
,
Goswami
S
,
Cortez
MA
,
Ahn
YH
,
Byers
LA
, et al
Metastasis is regulated via microRNA-200/ZEB1 axis control of tumour cell PD-L1 expression and intratumoral immunosuppression
.
Nat Commun
2014
;
5
:
5241
.
43.
Horn
LA
,
Fousek
K
,
Palena
C
. 
Tumor plasticity and resistance to immunotherapy
.
Trends Cancer
2020
;
6
:
432
41
.
44.
Thompson
JC
,
Hwang
W-T
,
Davis
C
,
Deshpande
C
,
Jeffries
S
,
Rajpurohit
Y
, et al
Gene signatures of tumor inflammation and epithelial-to-mesenchymal transition (EMT) predict responses to immune checkpoint blockade in lung cancer with high accuracy
.
Lung Cancer
2020
;
139
:
1
8
.
45.
Wang
L
,
Saci
A
,
Szabo
PM
,
Chasalow
SD
,
Castillo-Martin
M
,
Domingo-Domenech
J
, et al
EMT- and stroma-related gene expression and resistance to PD-1 blockade in urothelial cancer
.
Nat Commun
2018
;
9
:
3503
.
46.
Chakravarthy
A
,
Khan
L
,
Bensler
NP
,
Bose
P
,
De Carvalho
DD
. 
TGF-β-associated extracellular matrix genes link cancer-associated fibroblasts to immune evasion and immunotherapy failure
.
Nat Commun
2018
;
9
:
4692
.
47.
Hugo
W
,
Zaretsky
JM
,
Sun
L
,
Song
C
,
Moreno
BH
,
Hu-Lieskovan
S
, et al
Genomic and transcriptomic features of response to anti-PD-1 therapy in metastatic melanoma
.
Cell
2016
;
165
:
35
44
.
48.
Rothlin
CV
,
Ghosh
S
,
Zuniga
EI
,
Oldstone
MBA
,
Lemke
G
. 
TAM receptors are pleiotropic inhibitors of the innate immune response
.
Cell
2007
;
131
:
1124
36
.
49.
Maier
B
,
Leader
AM
,
Chen
ST
,
Tung
N
,
Chang
C
,
LeBerichel
J
, et al
A conserved dendritic-cell regulatory program limits antitumour immunity
.
Nature
2020
;
580
:
257
62
.
50.
Tanaka
M
,
Siemann
DW
. 
Gas6/Axl signaling pathway in the tumor immune microenvironment
.
Cancers
2020
;
12
:
1850
.