Purpose:

Combined axitinib/pembrolizumab is approved for advanced renal cell carcinoma (aRCC). This exploratory analysis examined associations between angiogenic and immune-related biomarkers and outcomes following axitinib/pembrolizumab treatment.

Patients and Methods:

Prospectively defined retrospective correlative exploratory analyses tested biospecimens from 52 treatment-naïve patients receiving axitinib and pembrolizumab (starting doses 5 mg twice daily and 2 mg/kg respectively, every 3 weeks). Tumor tissue, serum, and whole blood samples were collected at baseline, at cycle 2 day 1 (C2D1), and end of treatment (EOT) for blood-based samples. Clinical outcomes were objective response rate (ORR) and progression-free survival (PFS).

Results:

Higher baseline tumor levels of CD8 showed a trend toward longer PFS (HR 0.4; P = 0.091). Higher baseline serum levels of CXCL10 (P = 0.0197) and CEACAM1 (P = 0.085) showed a trend toward better ORR and longer PFS, respectively. Patients for whom IL6 was not detected at baseline had longer PFS versus patients for whom it was detected (HR 0.4; P = 0.028). At C2D1 and/or EOT, mainly immune-related biomarkers showed any association with better outcomes. The genes CA9 (P = 0.084), HIF1A (P = 0.064), and IFNG (P = 0.073) showed trending associations with ORR, and AKT3 (P = 0.0145), DDX58 (P = 0.0726), GZMA (P = 0.0666), LCN2 (NGAL; P = 0.0267), and PTPN11 (P = 0.0287) with PFS.

Conclusions:

With combined axitinib/pembrolizumab treatment in patients with aRCC, mostly immune-related biomarkers are associated with better treatment outcomes. This exploratory analysis has identified some candidate biomarkers to consider in future prospective testing.

Translational Relevance

Efforts continue to identify biomarkers that may confer treatment sensitivity or resistance for advanced renal cell carcinoma (aRCC). Recent studies have demonstrated good efficacy and safety of combined axitinib, a tyrosine kinase inhibitor, and pembrolizumab, an immune checkpoint inhibitor, in patients with aRCC. In this exploratory analysis of 52 treatment-naïve patients with aRCC, we explored specific angiogenic or immune-related biomarkers and their potential association with treatment outcomes. Baseline levels of some immune-related biomarkers were associated with a trend toward better outcomes. Similarly, levels of some immune-related biomarkers at the end of treatment were associated with treatment benefit. Gene expression analysis was broadly supportive of these findings, suggesting that immune/inflammation-related genes may be a major component driving treatment outcomes in addition to angiogenesis-related genes and other components of tumor biology. These results identify possible candidate biomarkers that may be explored further in a large, independent cohort.

Aberrant angiogenesis and immunosuppression are key features of the biology and pathophysiology of renal cell carcinoma (RCC; refs. 1–4). As a result, dominant proangiogenic and proimmunomodulatory mechanisms are the focus of treatment development in RCC. Treatment approaches based on these efforts include the tyrosine kinase inhibitors (TKI) and monoclonal antibodies (mAb) that target VEGF or its receptors (VEGFR; refs. 5–10), and the immune checkpoint inhibitors that target elements of the immunomodulatory pathways (11–14).

Axitinib is a potent and selective inhibitor of VEGFRs (15) that demonstrated clinical activity and an acceptable safety profile in the treatment of patients with metastatic RCC (9, 16). Pembrolizumab is an antiprogrammed cell death 1 (anti–PD-1) checkpoint inhibitor (17), and early phase II trial results show encouraging efficacy and acceptable tolerability of pembrolizumab as first-line therapy in patients with advanced RCC (aRCC; refs. 12, 18). Following the favorable results from a phase Ib dose-finding and dose-expansion study in patients with untreated aRCC (19), a phase III randomized trial of combined axitinib and pembrolizumab (axitinib/pembrolizumab) versus sunitinib was initiated which showed benefits in the intent-to-treat population based on all outcome measures (20). On the basis of the phase III trial, in 2019, the FDA approved axitinib/pembrolizumab for the treatment of clear cell RCC. Here, we present exploratory correlative analyses of the phase Ib study. In this phase Ib study, the objective response rate (ORR) was 73.1% and median progression-free survival (mPFS) was 20.9 months, while 65% of patients reported grade ≥3 treatment-related adverse events (AE; ref. 19).

Efforts continue to identify biomarkers that may be used to select appropriate patients with RCC for specific treatments, predict treatment efficacy, or circumvent resistance to treatment. This research is predicated on the assumption that detectable associations exist among functional tumor, angiogenic, and/or immunomodulatory proteins and patient responses that are a consequence of the mechanisms of action of the agents, and require certain host or tumor attributes that vary from patient to patient. Several previous studies have identified angiogenic and immunomodulatory biomarkers that are associated with outcomes in patients with RCC, including outcomes to treatment with antiangiogenic or antiimmunomodulatory therapies (21–25). More recently, biomarkers for combined antiangiogenic/antiimmunomodulatory treatment have been assessed. The phase III JAVELIN Renal 101 trial evaluated a combination of avelumab, an inhibitor of the immune checkpoint molecule programmed cell death-ligand 1 (PD-L1), and axitinib (26). The phase II IMmotion150 and phase III IMmotion151 studies examined combined atezolizumab, an inhibitor of PD-L1, and bevacizumab, an anti-VEGF mAb (27, 28). In these studies, biomarker analysis identified molecular characteristics that predicted treatment outcomes both within and among treatment arms (27, 29, 30).

To date, in the studies of axitinib/pembrolizumab treatment, the only biomarker assessed was tumor expression of PD-L1 (19, 20). Other biomarkers that may indicate differential sensitivity or resistance to axitinib/pembrolizumab treatment have not been reported. The tumor tissue, serum, and whole blood samples obtained from patients during the phase Ib study provide an opportunity to identify candidate biomarkers that may be associated with outcomes following axitinib/pembrolizumab treatment, with the caveat that this is a single-arm study in a small sample size and the need for later validation is required. The objective of this exploratory analysis was to therefore evaluate candidate biomarkers related to angiogenic and immunomodulatory pathways that may indicate sensitivity or resistance to combined axitinib/pembrolizumab treatment in patients with previously untreated aRCC.

Study design

The design of the nonrandomized, open-label, dose-finding, and dose-expansion phase Ib study is comprehensively described elsewhere (ClinicalTrial.gov identifier NCT02133742; ref. 19). Briefly, 52 patients (11 in the dose-finding phase and 41 in the dose-expansion phase) were ages ≥18 years with histologically or cytologically confirmed previously untreated and predominantly clear cell aRCC. Patients had not received prior systemic therapy for their advanced disease. Eligible patients received axitinib at a starting dose of 5 mg orally twice daily with a maximum permitted dose of 10 mg twice daily, together with pembrolizumab at a dose of 2 mg/kg intravenously every 3 weeks, the approved dose at the time of study conduct. The study protocol, amendments, and informed consent forms were reviewed and approved by the institutional review board or independent ethics committee of each participating study center, and the study was conducted in accordance with the International Council for Harmonisation Good Clinical Practice guidelines and the Declaration of Helsinki. All local regulatory requirements were followed and all patients provided written informed consent before any screening procedures were conducted. For this analysis, the dose-finding and dose-expansion cohorts were combined and analyzed together (Supplementary Fig. S1). The analysis population consisted of all enrolled patients who received ≥1 dose of axitinib/pembrolizumab and who had ≥1 baseline biomarker assessment for a given biospecimen and analysis.

Selection of angiogenic and immunomodulatory biomarkers

Candidate angiogenic protein biomarkers for evaluation were selected on the basis of their general involvement in angiogenesis or specific involvement in the VEGF pathway [angiopoietin-1 (Ang-1), angiopoietin-2 (Ang-2), VEGF, VEGFR-1, VEGFR-2, and VEGFR-3]. Immune and inflammation–related candidate protein biomarkers for evaluation were selected on the basis of their involvement in lymphocyte function (PD-L1, CD8, and CD68), their involvement in the immune or inflammatory responses [carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1), E-selectin, IFNγ-induced protein 10 (CXCL10), IL2, IL6, IL8, IL10, monokine-induced by IFNγ (CXCL-9), T cell–specific protein RANTES (RANTES), TGFβ, and neutrophil gelatinase–associated lipocalin (NGAL)], as well as T-cell receptor (TCR) repertoire dynamics. Other target protein biomarkers potentially involved in other tumor biology-related mechanisms were cMET, GRO-α, hepatocyte growth factor (HGF), matrix metalloproteinase-9 (MMP-9), osteopontin, stromal cell–derived factor-1 (SDF-1), and tissue inhibitor of metalloproteinase-1 (TIMP-1). Candidate gene biomarkers were also selected with respect to their angiogenesis-related, immune and inflammation–related, and other tumor biology-related activity. The 89 candidate genes are listed in Supplementary Table S1, in addition to 12 housekeeping genes.

Tumor tissue IHC analysis

Tumor tissue samples, either archival or from de novo biopsy, were collected at baseline from 39 of 52 (75.0%) patients; 23 patients provided archival samples only, 4 patients provided de novo samples only, and 12 patients provided archival and de novo samples. The tumor tissue specimens were initially tested for PD-L1 status utilizing the mouse mAb 22C3 as a potential diagnostic (Dako, Agilent Technologies), the results of which were published previously (19). The remaining tumor tissue was then submitted for additional IHC analysis. Analysis of formalin-fixed, paraffin-embedded tumor samples was performed in accordance with procedures developed, validated, and documented at Mosaic Laboratories. The tumor tissue biomarkers assessed were PD-L1, CD8, and CD68. Because of limited tissue availability, analysis of tumor biomarkers was prioritized for PD-L1 > CD8 > CD68. PD-L1 was measured using the rabbit mAb E1L3N (Cell Signaling Technology), which was different to the previously used mouse mAb 22C3. CD8 was measured using the mouse mAb C8/144B (Dako, Agilent Technologies) and CD68 using the mouse mAb KP1 (Dako, Agilent Technologies). Each assay was designed and validated as a fit-for-purpose laboratory developed test. Pathology review was conducted within a Good Clinical Laboratory Practice-like environment in a College of American Pathologists/Clinical Laboratory Improvement Amendments–certified facility.

Analysis of serum-based proteins

Serum samples for protein analysis were collected at baseline, cycle 2 day 1 (C2D1), and end of treatment (EOT). The EOT samples were available for a subgroup of patients who had discontinued from the study due to disease progression (n = 17) or due to AEs (n = 18). The serum-based proteins analyzed were Ang-1, Ang-2, CEACAM1, cMET, E-selectin, GRO-α, HGF, CXCL10, IL2, IL6, IL8, IL10, MMP-9, CXCL9, NGAL, osteopontin, SDF-1, RANTES, TGFβ, TIMP-1, VEGF, VEGFR-1, VEGFR-2, and VEGFR-3. Analysis was performed at Myriad RBM Inc using a quantitative multiplexed method to determine the concentrations of the defined proteins. Multiplexed cocktails of biotinylated reporter antibodies for each multiplex were added to the sample and developed using an excess streptavidin–phycoerythrin solution, which was then analyzed on a Luminex instrument. The resulting data stream was interpreted using proprietary data analysis software (Plate Reader) developed at Myriad RBM Inc. The least detectable dose for each protein was determined as the mean + 3 STDs of 20 blank readings and appropriate dilutions were made to ensure a quantitative measurement within the limits of the assay. The lower limit of quantification (LLOQ) was determined as the concentration of each protein where the measurement demonstrated a coefficient of variation of 30%, that is, it was the lowest protein concentration that could be measured with a precision better than or equal to 30%.

TCR variable β chain sequencing

For analysis of TCR repertoire dynamics, whole blood samples were obtained at baseline, C2D1, and EOT, and immunosequencing of the CDR3 regions of TCRβ chains was performed using the ImmunoSEQ Assay (Adaptive Biotechnologies). Extracted genomic DNA from whole blood specimens was amplified in a bias-controlled multiplex PCR followed by high-throughput sequencing. Sequences were collapsed and filtered to identify and quantitate the absolute abundance of each unique TCRβ CDR3 region for further analysis, as described previously (31–33). For statistical analysis of sequencing results, clonality was defined as 1-Peilou evenness (34) and was calculated on productive rearrangements by

formula

where pi is the proportional abundance of rearrangement i; and N is the total number of rearrangements. Clonality values range from 0 to 1 and describe the shape of the frequency distribution. Values approaching 0 indicate a very even distribution of frequencies whereas values approaching 1 indicate an increasingly asymmetric distribution in which a few clones are present at high frequencies.

Gene expression analysis

The EdgeSeq Oncology Biomarker Panel (HTG Molecular Diagnostics) was used for tumor tissue mRNA profiling. Sample preparation was conducted following the laboratory processes and manufacturer protocols and sequencing was performed on a NextSeq 500 sequencer (Illumina). Gene expression data were quantile normalized and log2 transformed (HTG Molecular Diagnostics).

Analysis of biomarker endpoints

For PD-L1, CD8, and CD68, the percentage of positive tumor cells (% positive cells) was captured, and the density of positive cells (cells/mm2) was also captured for CD8 and CD68. For PD-L1, patients were categorized as either negative or positive based on the proportion of positive tumor cells (negative, <1%; positive, ≥1%; ref. 35). For serum-based biomarkers, summary statistics of baseline values and the ratio of posttreatment (i.e., C2D1 and EOT) to baseline values were captured. Biomarker results were correlated with the antitumor efficacy endpoints of ORR and PFS. Summaries of levels of tumor tissue, serum, and gene biomarkers at baseline and/or ratio of posttreatment values to baseline values versus ORR category (best overall response) were made, with comparison between responders [complete response (CR) + partial response (PR)] versus nonresponders [stable disease (SD) + progressive disease (PD) + indeterminate response (IR)] made using Wilcoxon rank-sum test. PFS was summarized using the Kaplan–Meier method after stratification by <median or ≥median baseline concentration values, and if there were too few events in either group to meaningfully interpret the Kaplan–Meier analysis neither P value nor HR were reported. TCR repertoire was assessed by measuring richness, a measure of the number of different species in a repertoire, using the Daley–Smith estimate (36). Tumor tissue and serum biomarker levels, and TCR repertoire dynamics, at baseline were also compared when patients were stratified by duration of PFS (<9 months vs. >20 months). These cutoffs were retrospectively selected on the basis of the standard mPFS of sunitinib of approximately 11 months (37) to segregate those patients with no or limited benefits in PFS (i.e., < mPFS of sunitinib) and those with the greatest benefits in PFS (i.e., approximately >2 × mPFS of sunitinib). For those serum biomarkers that were below the LLOQ, analyses by ORR, by PFS, and when patients were stratified by duration of PFS (<9 months vs. >20 months) were conducted comparing when samples were detected and when they were not detected. Unadjusted and adjusted P values were reported as appropriate, and the adjusted P value obtained by calculating FDRs using Benjamini and Hochberg's step-up method (38). All statistical tests were univariate and significance was declared if the adjusted P value was < 0.05.

Study samples

Fifty-two treatment-naïve patients (11 in the dose-finding phase and 41 in the dose-expansion phase) received combined axitinib and pembrolizumab at starting doses of 5 mg twice daily and 2 mg/kg every 3 weeks, respectively, and were included in the analysis. The number of available samples for each analysis is shown in Supplementary Fig. S1.

Tumor tissue biomarkers

For PD-L1, the range of % positive cells was 0% to 80%. Overall, 27 (69.2%) patients were PD-L1 negative (<1%) and 12 (30.8%) were PD-L1 positive (≥1%). For CD8, the median (range) % positive cells was 13.2% (0.7%–39.7%) and the median (range) density of positive cells was 492.3 (25.0–2,107.5) cells/mm2. For CD68, the median (range) % positive cells was 20.4% (5.8%–65.4%) and the median (range) density of positive cells was 567.0 (244.0–1,615.0) cells/mm2.

No significant associations were observed among baseline levels of PD-L1, CD8, or CD68 and ORR (Table 1). Moreover, when patients were stratified by baseline PD-L1 levels (negative or positive), there was no association between the proportion of patients in each group and ORR (Supplementary Table S2). Similarly, no significant associations were observed among baseline levels of PD-L1 (negative or positive), CD8, or CD68 and PFS (Table 2). However, patients whose CD8 levels were ≥median at baseline had a numerically longer PFS than patients whose CD8 levels were <median [HR, 0.4; 95% confidence interval (CI), 0.2–1.2; P = 0.091; Fig. 1A].

Table 1.

Comparison of tumor tissue biomarker levels at baseline versus tumor response category.

Biologic profileTumor response categoryNMean (STD)%CVMedianPaAdjusted Pb
% Positive cells (%) 
PD-L1 CR 3.3 (5.8) 173.2 0.2449 0.5803 
 PR 27 2.5 (5.2) 204.7   
 SD —   
 PD —   
 IR 40.0 (56.6) 141.4 40.0   
CD8 CR 11.2 (7.1) 63.9 10.4 0.1223 0.5803 
 PR 26 18.4 (12.2) 66.2 18.5   
 SD 8.7 (4.7) 54.4 11.0   
 PD 8.4 — 8.4   
 IR 18.3 (2.3) 12.5 18.3   
CD68 CR 13.2 (4.8) 36.4 13.9 0.4017 0.5803 
 PR 13 26.3 (15.3) 58.2 25.0   
 SD 14.7 (2.1) 14.5 14.7   
 PD 22.5 — 22.5   
 IR — — —   
Density of positive cells (cells/mm2
CD8 CR 499.3 (246.9) 49.4 412.0 0.3194 0.5803 
 PR 26 763.5 (600.9) 78.7 638.5   
 SD 358.5 (199.5) 55.6 404.0   
 PD 373.0 — 373.0   
 IR 751.5 (188.8) 25.1 751.5   
CD68 CR 579.0 (432.9) 74.8 385.0 0.6149 0.6661 
 PR 13 699.0 (356.2) 51.0 577.0   
 SD 471.0 (48.8) 10.4 471.0   
 PD 821.0 — 821.0   
 IR — — —   
Biologic profileTumor response categoryNMean (STD)%CVMedianPaAdjusted Pb
% Positive cells (%) 
PD-L1 CR 3.3 (5.8) 173.2 0.2449 0.5803 
 PR 27 2.5 (5.2) 204.7   
 SD —   
 PD —   
 IR 40.0 (56.6) 141.4 40.0   
CD8 CR 11.2 (7.1) 63.9 10.4 0.1223 0.5803 
 PR 26 18.4 (12.2) 66.2 18.5   
 SD 8.7 (4.7) 54.4 11.0   
 PD 8.4 — 8.4   
 IR 18.3 (2.3) 12.5 18.3   
CD68 CR 13.2 (4.8) 36.4 13.9 0.4017 0.5803 
 PR 13 26.3 (15.3) 58.2 25.0   
 SD 14.7 (2.1) 14.5 14.7   
 PD 22.5 — 22.5   
 IR — — —   
Density of positive cells (cells/mm2
CD8 CR 499.3 (246.9) 49.4 412.0 0.3194 0.5803 
 PR 26 763.5 (600.9) 78.7 638.5   
 SD 358.5 (199.5) 55.6 404.0   
 PD 373.0 — 373.0   
 IR 751.5 (188.8) 25.1 751.5   
CD68 CR 579.0 (432.9) 74.8 385.0 0.6149 0.6661 
 PR 13 699.0 (356.2) 51.0 577.0   
 SD 471.0 (48.8) 10.4 471.0   
 PD 821.0 — 821.0   
 IR — — —   

Abbreviations: CR, complete response; CV, coefficient of variation; IR, indeterminate response; PD, progressive disease; PR, partial response; SD, stable disease; STD, standard deviation.

aUnadjusted P value (two-sided Wilcoxon rank-sum test) comparing responders (CR+PR) versus nonresponders (SD+PD+IR).

bAdjusted P value obtained by calculating the FDR using Benjamini and Hochberg's step-up method.

Table 2.

Comparison of PFS stratified by median tumor tissue biomarker level (% positive cells) at baseline.

Biologic profileNmPFS (95% CI), monthsHR (95% CI)Pa
PD-L1 
 Negative (<1%) 27 22.1 (15.1–26.2) 0.6 (0.2–1.8) 0.377 
 Positive (≥1%) 12 NR (7.1–NR)   
CD8 
 <Median 19 15.4 (8.2–NR) 0.4 (0.2–1.2) 0.091 
 ≥Median 19 26.2 (20.7–NR)   
CD68 
 <Median 23.2 (15.4–NR) ND ND 
 ≥Median 10 26.2 (2.7–NR)   
Biologic profileNmPFS (95% CI), monthsHR (95% CI)Pa
PD-L1 
 Negative (<1%) 27 22.1 (15.1–26.2) 0.6 (0.2–1.8) 0.377 
 Positive (≥1%) 12 NR (7.1–NR)   
CD8 
 <Median 19 15.4 (8.2–NR) 0.4 (0.2–1.2) 0.091 
 ≥Median 19 26.2 (20.7–NR)   
CD68 
 <Median 23.2 (15.4–NR) ND ND 
 ≥Median 10 26.2 (2.7–NR)   

Abbreviations: ND, not determined; NR, not reached.

aP value based on Kaplan–Meier analysis. HRs and P values were not displayed when n ≤ 10 in either group.

Figure 1.

PFS by baseline tumor level of CD8 (A) and by baseline serum level of CEACAM1 (B). Data in A are for % positive cells. NR, not reached.

Figure 1.

PFS by baseline tumor level of CD8 (A) and by baseline serum level of CEACAM1 (B). Data in A are for % positive cells. NR, not reached.

Close modal

When patients were stratified by PFS (<9 months vs. >20 months), the mean % positive cells and cell density at baseline were similar irrespective of PFS duration (Supplementary Table S3). In addition, the proportions of patients who were PD-L1 negative or positive were similar and not associated with PFS (Supplementary Table S4).

Serum-based protein biomarkers

For IL2 and VEGFR-1, 98.0%–100% of concentrations at each collection time point were below the LLOQ. Moreover, the majority of concentrations were below the LLOQ for IL6 (68.6%–74.5%), IL8 (34.3%–60.8%), IL10 (74.3%–90.2%), and MMP-9 (89.8%–91.4%). For the remaining serum-based proteins, the concentration at each collection time point, and ratios of C2D1 and EOT to baseline, are shown in Supplementary Table S5.

For the majority of serum-based proteins, there was no association between concentrations at baseline, C2D1, or EOT, or the ratios of C2D1 or EOT to baseline, and ORR. For CXCL10, there was a trend toward better ORR with higher baseline concentrations (unadjusted P value = 0.0197; Table 3). There were also trends toward better ORR and lower EOT concentrations of CEACAM1, GRO-α, HGF, and TIMP-1, that were supported by the associations between EOT:baseline values and ORR (Table 3).

Table 3.

Comparison of serum-based protein biomarker concentrations at baseline, and ratio of EOT:baseline values versus tumor response category.

Time pointAnalyteTumor response categoryNMean (STD)%CVMedianPaAdjusted Pb
Baseline CXCL10 (pg/mL) CR 362.8 (325.7) 89.8 179.0 0.0197 0.2235 
  PR 32 183.7 (108.1) 58.8 151.0   
  SD 133.2 (113.1) 84.9 77.5   
  PD 97.0 (44.0) 45.4 80.0   
  IR 133.7 (31.9) 23.9 121.0   
EOT CEACAM1 (ng/mL) CR 26.7 (6.1) 22.9 28.0 0.0026 0.0784 
  PR 24 26.2 (7.6) 29.0 25.5   
  SD 37.6 (8.2) 21.8 36.0   
  PD 56.0  56.0   
  IR 34.0 (9.9) 29.1 34.0   
 GRO-α (pg/mL) CR 118.3 (67.6) 57.1 114.0 0.0495 0.3410 
  PR 24 153.8 (79.2) 51.5 135.0   
  SD 230.8 (95.9) 41.6 220.0   
  PD 321.0  321.0   
  IR 168.0 (142.8) 85.0 168.0   
 HGF (ng/mL) CR 13.4 (10.1) 75.3 9.1 0.0112 0.1566 
  PR 24 10.8 (9.6) 89.0 8.4   
  SD 18.1 (11.5) 63.7 15.0   
  PD 30.0  30.0   
  IR 13.0 (1.4) 10.9 13.0   
 TIMP-1 (ng/mL) CR 119.0 (24.3) 20.5 107.0 0.0044 0.0784 
  PR 24 146.7 (58.3) 39.7 130.5   
  SD 212.2 (130.6) 61.6 154.0   
  PD 423.0  423.0   
  IR 386.5 (222.7) 57.6 386.5   
EOT:Baseline CEACAM1 CR 1.0 (0.3) 25.8 1.0 0.0493 0.3410 
  PR 24 1.1 (0.3) 27.7 1.0   
  SD 1.2 (0.2) 17.6 1.3   
  PD 1.8  1.8   
  IR 2.4 (1.8) 75.8 2.4   
 GRO-α CR 0.8 (0.2) 31.5 0.8 0.0392 0.3264 
  PR 24 1.0 (0.4) 43.2 1.1   
  SD 1.3 (0.3) 20.6 1.3   
  PD 0.8  0.8   
  IR 1.6 (0.4) 23.0 1.6   
 HGF CR 1.6 (0.8) 51.0 1.2 0.0216 0.2245 
  PR 24 1.1 (0.6) 56.2 0.9   
  SD 1.5 (0.2) 15.5 1.5   
  PD 1.8  1.8   
  IR 1.5 (0.3) 20.5 1.5   
 TIMP-1 CR 0.9 (0.5) 54.8 0.6 0.0034 0.0784 
  PR 24 1.0 (0.3) 28.2 1.0   
  SD 1.4 (0.5) 33.0 1.3   
  PD 1.3  1.3   
  IR 2.7 (2.2) 78.9 2.7   
Time pointAnalyteTumor response categoryNMean (STD)%CVMedianPaAdjusted Pb
Baseline CXCL10 (pg/mL) CR 362.8 (325.7) 89.8 179.0 0.0197 0.2235 
  PR 32 183.7 (108.1) 58.8 151.0   
  SD 133.2 (113.1) 84.9 77.5   
  PD 97.0 (44.0) 45.4 80.0   
  IR 133.7 (31.9) 23.9 121.0   
EOT CEACAM1 (ng/mL) CR 26.7 (6.1) 22.9 28.0 0.0026 0.0784 
  PR 24 26.2 (7.6) 29.0 25.5   
  SD 37.6 (8.2) 21.8 36.0   
  PD 56.0  56.0   
  IR 34.0 (9.9) 29.1 34.0   
 GRO-α (pg/mL) CR 118.3 (67.6) 57.1 114.0 0.0495 0.3410 
  PR 24 153.8 (79.2) 51.5 135.0   
  SD 230.8 (95.9) 41.6 220.0   
  PD 321.0  321.0   
  IR 168.0 (142.8) 85.0 168.0   
 HGF (ng/mL) CR 13.4 (10.1) 75.3 9.1 0.0112 0.1566 
  PR 24 10.8 (9.6) 89.0 8.4   
  SD 18.1 (11.5) 63.7 15.0   
  PD 30.0  30.0   
  IR 13.0 (1.4) 10.9 13.0   
 TIMP-1 (ng/mL) CR 119.0 (24.3) 20.5 107.0 0.0044 0.0784 
  PR 24 146.7 (58.3) 39.7 130.5   
  SD 212.2 (130.6) 61.6 154.0   
  PD 423.0  423.0   
  IR 386.5 (222.7) 57.6 386.5   
EOT:Baseline CEACAM1 CR 1.0 (0.3) 25.8 1.0 0.0493 0.3410 
  PR 24 1.1 (0.3) 27.7 1.0   
  SD 1.2 (0.2) 17.6 1.3   
  PD 1.8  1.8   
  IR 2.4 (1.8) 75.8 2.4   
 GRO-α CR 0.8 (0.2) 31.5 0.8 0.0392 0.3264 
  PR 24 1.0 (0.4) 43.2 1.1   
  SD 1.3 (0.3) 20.6 1.3   
  PD 0.8  0.8   
  IR 1.6 (0.4) 23.0 1.6   
 HGF CR 1.6 (0.8) 51.0 1.2 0.0216 0.2245 
  PR 24 1.1 (0.6) 56.2 0.9   
  SD 1.5 (0.2) 15.5 1.5   
  PD 1.8  1.8   
  IR 1.5 (0.3) 20.5 1.5   
 TIMP-1 CR 0.9 (0.5) 54.8 0.6 0.0034 0.0784 
  PR 24 1.0 (0.3) 28.2 1.0   
  SD 1.4 (0.5) 33.0 1.3   
  PD 1.3  1.3   
  IR 2.7 (2.2) 78.9 2.7   

Note: Data are reported for biomarkers where the two-sided P value was < 0.05.

Abbreviations: CR, complete response; CV, coefficient of variation; IR, indeterminate response; PD, progressive disease; PR, partial response; SD, stable disease; STD, standard deviation.

aUnadjusted P value (two-sided Wilcoxon rank-sum test) comparing responders (CR+PR) versus nonresponders (SD+PD+IR).

bAdjusted P value obtained by calculating the FDR using Benjamini and Hochberg's step-up method.

For Ang-1, Ang-2, E-selectin, CXCL10, NGAL, osteopontin, VEGF, VEGFR-2, and VEGFR-3, there was no association between concentrations at baseline, C2D1, or EOT, or the ratios of C2D1 or EOT to baseline, and PFS. At baseline, CEACAM1 levels ≥median were associated with a trend toward better PFS (Table 4; Fig. 1B). C2D1, GRO-α and HGF levels <median were associated with better PFS, and levels of RANTES, TGFβ, and TIMP-1 <median were associated with a trend toward better PFS (Table 4). At EOT, HGF and TIMP-1 levels <median were associated with better PFS, and levels of GRO-α, RANTES, and TGFβ <median were associated with a trend toward better PFS (Table 4). For the C2D1:baseline ratio, levels of CEACAM1 <median were associated with better PFS and levels of CXCL9 <median were associated with a trend toward better PFS (Table 4). Finally, for the EOT:baseline ratio, levels of cMET and SDF-1 <median were associated with a trend toward better PFS (Table 4).

Table 4.

Comparison of PFS stratified by median analyte levels of serum-based protein biomarkers at time point and ratio of time points to baseline.

< Median analyte value> Median analyte value
Time pointAnalyteMedian analyte valueNmPFS (95% CI), monthsNmPFS (95% CI), monthsPaHR (95% CI)
Baseline CEACAM1 26.0 ng/mL 24 18.0 (9.9–23.5) 27 NR (15.4–NR) 0.085 0.5 (0.2–1.1) 
C2D1 GRO-α 147.0 pg/mL 24 NR (18.0–NR) 25 18.0 (8.2–23.5) 0.034 2.5 (1.0–6.1) 
 HGF 9.6 ng/mL 24 NR (26.2–NR) 25 15.4 (8.2–20.9) <0.001 7.6 (2.6–22.6) 
 RANTES 11.0 ng/mL 23 NR (18.0–NR) 26 20.7 (8.2–26.2) 0.056 2.3 (1.0–5.7) 
 TGFβ 25,350 pg/mL 24 NR (15.2–NR) 25 20.7 (6.9–26.2) 0.068 2.2 (0.9–5.3) 
 TIMP-1 150.0 ng/mL 24 NR (18.0–NR) 25 20.7 (9.9–23.2) 0.088 2.1 (0.9–4.9) 
C2D1:Baseline CEACAM1 1.2 24 NR (15.2–NR) 25 20.7 (10.9–23.5) 0.024 2.8 (1.1–7.3) 
 CXCL9 2.3 24 NR (18.0–NR) 25 20.9 (7.1–26.2) 0.058 2.3 (0.9–5.5) 
EOT GRO-α 153.0 pg/mL 17 26.2 (9.9–NR) 18 18.0 (7.1–20.7) 0.078 2.4 (0.9–6.4) 
 HGF 9.1 17 26.2 (7.1–NR) 18 15.4 (8.2–18.0) 0.006 5.0 (1.4–17.1) 
 RANTES 11.0 ng/mL 17 22.1 (9.9–NR) 18 18.0 (7.1–26.2) 0.093 2.3 (0.8–6.3) 
 TGFβ 21,550 pg/mL 17 23.5 (10.9–NR) 18 18.0 (6.9–22.1) 0.072 2.4 (0.9–6.3) 
 TIMP-1 140.0 ng/mL 17 23.5 (20.7–NR) 18 15.1 (6.9–18.0) 0.014 3.3 (1.2–9.1) 
EOT:Baseline cMET 1.0 17 22.1 (15.4–NR) 18 15.1 (6.9–26.2) 0.067 2.4 (0.9–6.4) 
 SDF-1 1.0 17 22.1 (15.4–NR) 18 15.1 (6.9–NR) 0.055 2.6 (0.9–7.4) 
< Median analyte value> Median analyte value
Time pointAnalyteMedian analyte valueNmPFS (95% CI), monthsNmPFS (95% CI), monthsPaHR (95% CI)
Baseline CEACAM1 26.0 ng/mL 24 18.0 (9.9–23.5) 27 NR (15.4–NR) 0.085 0.5 (0.2–1.1) 
C2D1 GRO-α 147.0 pg/mL 24 NR (18.0–NR) 25 18.0 (8.2–23.5) 0.034 2.5 (1.0–6.1) 
 HGF 9.6 ng/mL 24 NR (26.2–NR) 25 15.4 (8.2–20.9) <0.001 7.6 (2.6–22.6) 
 RANTES 11.0 ng/mL 23 NR (18.0–NR) 26 20.7 (8.2–26.2) 0.056 2.3 (1.0–5.7) 
 TGFβ 25,350 pg/mL 24 NR (15.2–NR) 25 20.7 (6.9–26.2) 0.068 2.2 (0.9–5.3) 
 TIMP-1 150.0 ng/mL 24 NR (18.0–NR) 25 20.7 (9.9–23.2) 0.088 2.1 (0.9–4.9) 
C2D1:Baseline CEACAM1 1.2 24 NR (15.2–NR) 25 20.7 (10.9–23.5) 0.024 2.8 (1.1–7.3) 
 CXCL9 2.3 24 NR (18.0–NR) 25 20.9 (7.1–26.2) 0.058 2.3 (0.9–5.5) 
EOT GRO-α 153.0 pg/mL 17 26.2 (9.9–NR) 18 18.0 (7.1–20.7) 0.078 2.4 (0.9–6.4) 
 HGF 9.1 17 26.2 (7.1–NR) 18 15.4 (8.2–18.0) 0.006 5.0 (1.4–17.1) 
 RANTES 11.0 ng/mL 17 22.1 (9.9–NR) 18 18.0 (7.1–26.2) 0.093 2.3 (0.8–6.3) 
 TGFβ 21,550 pg/mL 17 23.5 (10.9–NR) 18 18.0 (6.9–22.1) 0.072 2.4 (0.9–6.3) 
 TIMP-1 140.0 ng/mL 17 23.5 (20.7–NR) 18 15.1 (6.9–18.0) 0.014 3.3 (1.2–9.1) 
EOT:Baseline cMET 1.0 17 22.1 (15.4–NR) 18 15.1 (6.9–26.2) 0.067 2.4 (0.9–6.4) 
 SDF-1 1.0 17 22.1 (15.4–NR) 18 15.1 (6.9–NR) 0.055 2.6 (0.9–7.4) 

Note: Data are presented for biomarkers where the two-sided P value was < 0.1.

Abbreviation: NR, not reached.

aBased on two-sided log-rank test.

When patients were stratified by PFS (<9 months vs. >20 months), mean baseline levels of all serum biomarkers were similar irrespective of PFS duration, although there was a trend toward lower mean concentrations of NGAL in patients with PFS >20 months (486.3 ± 171.4 ng/mL, n = 20) versus in patients with PFS <9 months (596.8 ± 247.6 ng/mL, n = 18; unadjusted P value = 0.0722). In comparison, at C2D1, patients with PFS >20 months had lower levels of Ang-2, NGAL, and RANTES than patients with PFS <9 months, and at EOT, patients with PFS >20 months had lower levels of Ang-1, CEACAM1, GRO-α, HGF, NGAL, osteopontin, RANTES, TGFβ, and TIMP-1 than patients with PFS <9 months (Table 5). For the C2D1:baseline ratio, patients with PFS >20 months had a lower ratio for CXCL10 and CXCL9 than patients with PFS <9 months, and for the EOT:baseline ratio, patients with PFS >20 months had a lower ratio for GRO-α, osteopontin, SDF-1, TIMP-1, and VEGF than patients with PFS <9 months (Table 5).

Table 5.

Comparison of levels of serum-based protein biomarkers at time points and ratio of time points to baseline values in patients stratified by PFS (<9 months vs. >20 months).

Time pointAnalytePFS (months)NMean (STD)%CVMedianPaAdjusted Pb
C2D1 Ang-2 (ng/mL) <9 16 4.9 (2.5) 50.7 4.8 0.0369 0.2309 
  >20 20 3.4 (2.3) 67.7 3.0   
 NGAL (ng/mL) <9 16 766.9 (243.5) 31.8 768.0 0.0036 0.0894 
  >20 20 544.2 (191.6) 35.2 535.0   
 RANTES (ng/mL) <9 16 17.4 (10.9) 62.4 13.5 0.0313 0.2226 
  >20 20 9.2 (4.0) 43.2 10.0   
C2D1:Baseline CXCL10 <9 16 3.6 (2.0) 55.8 3.2 0.0292 0.2226 
  >20 20 2.2 (1.0) 45.9 2.1   
 CXCL9 <9 16 4.4 (2.6) 59.0 4.0 0.0177 0.1844 
  >20 20 2.5 (1.4) 55.0 2.3   
EOT Ang-1 (ng/mL) <9 13 40.2 (20.1) 49.9 38.0 0.0174 0.1844 
  >20 11 22.9 (12.1) 52.9 23.0   
 CEACAM1 (ng/mL) <9 13 34.1 (10.1) 29.7 32.0 0.0421 0.2508 
  >20 11 24.4 (8.1) 33.1 25.0   
 GRO-α (pg/mL) <9 13 210.6 (99.6) 47.3 220.0 0.0277 0.2226 
  >20 11 119.4 (72.4) 60.6 104.0   
 HGF (ng/mL) <9 13 13.9 (9.4) 68.2 12.0 0.0368 0.2309 
  >20 11 7.1 (3.1) 43.2 7.2   
 NGAL (ng/mL) <9 13 895.2 (545.9) 61.0 761.0 0.0127 0.1844 
  >20 11 512.2 (188.0) 36.7 559.0   
 Osteopontin (ng/mL) <9 13 15.8 (13.9) 88.5 8.7 0.0077 0.1367 
  >20 11 6.3 (3.3) 52.3 6.0   
 RANTES (ng/mL) <9 13 15.6 (10.5) 67.4 13.0 0.0221 0.2123 
  >20 11 6.6 (5.9) 90.1 4.8   
 TGFβ (pg/mL) <9 13 30,227 (17,574) 58.1 24,150 0.0175 0.1844 
  >20 11 16,291 (7,745) 47.6 16,200   
 TIMP-1 (ng/mL) <9 13 242.7 (139.3) 57.4 172.0 0.0001 0.0164 
  >20 11 110.5 (19.3) 17.5 112.0   
EOT:Baseline GRO-α <9 13 1.3 (0.4) 27.3 1.2 0.0321 0.2226 
  >20 11 0.9 (0.4) 48.8 1.0   
 Osteopontin <9 13 2.0 (1.7) 83.8 1.2 0.0014 0.0600 
  >20 11 0.6 (0.4) 73.1 0.5   
 SDF-1 <9 13 1.2 (0.2) 18.2 1.2 0.0065 0.1348 
  >20 11 1.0 (0.2) 21.3 1.0   
 TIMP-1 <9 13 1.5 (0.9) 58.6 1.3 0.0031 0.0894 
  >20 11 0.9 (0.3) 33.1 0.8   
 VEGF <9 13 1.5 (0.6) 38.7 1.2 0.0009 0.0539 
  >20 11 0.9 (0.3) 30.5 0.9   
Time pointAnalytePFS (months)NMean (STD)%CVMedianPaAdjusted Pb
C2D1 Ang-2 (ng/mL) <9 16 4.9 (2.5) 50.7 4.8 0.0369 0.2309 
  >20 20 3.4 (2.3) 67.7 3.0   
 NGAL (ng/mL) <9 16 766.9 (243.5) 31.8 768.0 0.0036 0.0894 
  >20 20 544.2 (191.6) 35.2 535.0   
 RANTES (ng/mL) <9 16 17.4 (10.9) 62.4 13.5 0.0313 0.2226 
  >20 20 9.2 (4.0) 43.2 10.0   
C2D1:Baseline CXCL10 <9 16 3.6 (2.0) 55.8 3.2 0.0292 0.2226 
  >20 20 2.2 (1.0) 45.9 2.1   
 CXCL9 <9 16 4.4 (2.6) 59.0 4.0 0.0177 0.1844 
  >20 20 2.5 (1.4) 55.0 2.3   
EOT Ang-1 (ng/mL) <9 13 40.2 (20.1) 49.9 38.0 0.0174 0.1844 
  >20 11 22.9 (12.1) 52.9 23.0   
 CEACAM1 (ng/mL) <9 13 34.1 (10.1) 29.7 32.0 0.0421 0.2508 
  >20 11 24.4 (8.1) 33.1 25.0   
 GRO-α (pg/mL) <9 13 210.6 (99.6) 47.3 220.0 0.0277 0.2226 
  >20 11 119.4 (72.4) 60.6 104.0   
 HGF (ng/mL) <9 13 13.9 (9.4) 68.2 12.0 0.0368 0.2309 
  >20 11 7.1 (3.1) 43.2 7.2   
 NGAL (ng/mL) <9 13 895.2 (545.9) 61.0 761.0 0.0127 0.1844 
  >20 11 512.2 (188.0) 36.7 559.0   
 Osteopontin (ng/mL) <9 13 15.8 (13.9) 88.5 8.7 0.0077 0.1367 
  >20 11 6.3 (3.3) 52.3 6.0   
 RANTES (ng/mL) <9 13 15.6 (10.5) 67.4 13.0 0.0221 0.2123 
  >20 11 6.6 (5.9) 90.1 4.8   
 TGFβ (pg/mL) <9 13 30,227 (17,574) 58.1 24,150 0.0175 0.1844 
  >20 11 16,291 (7,745) 47.6 16,200   
 TIMP-1 (ng/mL) <9 13 242.7 (139.3) 57.4 172.0 0.0001 0.0164 
  >20 11 110.5 (19.3) 17.5 112.0   
EOT:Baseline GRO-α <9 13 1.3 (0.4) 27.3 1.2 0.0321 0.2226 
  >20 11 0.9 (0.4) 48.8 1.0   
 Osteopontin <9 13 2.0 (1.7) 83.8 1.2 0.0014 0.0600 
  >20 11 0.6 (0.4) 73.1 0.5   
 SDF-1 <9 13 1.2 (0.2) 18.2 1.2 0.0065 0.1348 
  >20 11 1.0 (0.2) 21.3 1.0   
 TIMP-1 <9 13 1.5 (0.9) 58.6 1.3 0.0031 0.0894 
  >20 11 0.9 (0.3) 33.1 0.8   
 VEGF <9 13 1.5 (0.6) 38.7 1.2 0.0009 0.0539 
  >20 11 0.9 (0.3) 30.5 0.9   

Note: Data are reported for biomarkers where the two-sided P value was < 0.05.

Abbreviations: CV, coefficient of variation; STD, standard deviation.

aUnadjusted P value (two-sided Wilcoxon rank-sum test) comparing PFS <9 months versus PFS >20 months.

bAdjusted P value obtained by calculating the FDR using Benjamini and Hochberg's step-up method.

For IL2, IL6, IL8, IL10, MMP-9, and VEGFR-1, the majority of concentrations at each time point were below the LLOQ. For each of these biomarkers, the relationship to ORR and PFS, and comparison when stratified by duration of PFS, was analyzed when samples of each biomarker were detected and not detected. There were similar numbers and proportions of patients with detectable and undetectable samples at baseline for all of these biomarkers. For IL8 and IL10, samples in which these were not detected at EOT were associated with better ORR than samples in which they were detected (Supplementary Table S6), and this was also true for IL10 at C2D1. For the other biomarkers, there was no association between whether it was detectable or nondetectable and ORR (Supplementary Table S6). For IL6, patients for whom it was not detected at baseline had a longer PFS compared with patients for whom it was detected, but at C2D1 and EOT, there were no such associations (Supplementary Table S7). For the other biomarkers, there was no association between detectable or nondetectable samples and PFS at any time point (Supplementary Table S7). When patients were stratified by PFS (<9 months vs. >20 months), patients for whom IL10 was undetectable at C2D1 had a longer PFS. Similarly, undetectable IL8 at EOT was associated with a longer PFS, while detectable MMP-9 at EOT was also associated with a longer PFS (Supplementary Table S8).

TCR repertoire dynamics

Assessment of the Daley–Smith richness estimate at baseline demonstrated no association with best overall response categories (unadjusted P value = 0.2196). In comparison, greater Daley–Smith richness at C2D1 or EOT was associated with a trend toward better tumor response (unadjusted P values = 0.0866 for C2D1 and 0.0324 for EOT). When patients were stratified by PFS (<9 months vs. >20 months), Daley–Smith estimates of richness were similar at baseline (P = 0.9888), C2D1 (P = 0.7475), and EOT (P = 0.7237).

Gene expression analysis

Expression analysis of a subset of 89 genes from the EdgeSeq Oncology Biomarker Panel that included genes involved in the regulation of angiogenesis, immune modulation, and other tumor biology in RCC was performed in baseline tumor tissue from 14 patients. Following stratification by tumor response category (responders vs. nonresponders), the following gene expressions showed trends of association: CA9 (unadjusted P value = 0.084), HIF1A (unadjusted P value = 0.064), and IFNG (unadjusted P value = 0.073; see Supplementary Table S9 for results for the full list of genes). The correlation of the individual gene expressions by stratified median cut-points for each mRNA expression level with PFS was also explored (Supplementary Table S10). The following genes showed trends of association: AKT3 (unadjusted P value = 0.0145), DDX58 (unadjusted P value = 0.0726), GZMA (unadjusted P value = 0.0666), LCN2 (NGAL; unadjusted P value = 0.0267), and PTPN11 (unadjusted P value = 0.0287). A combined 12-gene signature was derived using the Euclidean method for hierarchical cluster analysis on the standardized median values of the biomarker values, and tested to predict PFS (Supplementary Fig. S2A and S2B). The signature comprised the following 12 genes: AKT3, DNMT1, IL2RG, PTPN11, MTCP1, TNFSF10, CCR3, PDCD1, TBX21, IDO1, NOS3, and IFNG. On the basis of this 12-gene signature, after calculation of the median z-score value from the average z-score values of all the genes for an individual's tumor, patients whose tumor had a lower expression of the gene signature had a longer PFS than those patients whose tumor had a higher expression (P = 0.0055). A smaller 5-gene angiogenesis signature (CD34, ESM1, FLT1, KDR, and VEGFA) did not demonstrate any trend of association with PFS after stratification by the median z-score (P = 0.4424; Supplementary Fig. S2C).

In treatment-naïve patients with aRCC who received combined axitinib/pembrolizumab, there were no blood-based or tissue-based biomarkers that clearly predicted for treatment outcomes. Higher baseline tumor levels of CD8, higher baseline serum levels of CXCL10 and CEACAM1, and lower baseline serum levels of NGAL and nondetectable serum levels of IL6, were associated with a trend toward better ORR and/or longer PFS. For NGAL, there was also a trend toward lower baseline serum levels in patients with PFS >20 months compared with patients with PFS <9 months. These exploratory biomarkers are all components of immunomodulatory pathways, and it is notable that no angiogenic biomarkers at baseline, including specific VEGF-related biomarkers, exhibited any association with treatment outcomes. For the genomic analysis, there was a trend for expression levels of 8 genes to be associated with outcomes, the majority either immune and inflammation–related, or related to other tumor biology. A separately derived 12-gene signature helped to identify patients with longer PFS; patients whose tumor had lower expression of the 12-gene signature had longer PFS than those with higher expression. Again, the majority of genes in the 12-gene signature were immune and inflammation–related or related to other tumor biology. Thus, in agreement with the tumor tissue and serum protein analyses, it seems that immune/inflammation-related genes may be a major component driving treatment outcomes in addition to angiogenesis and other components of tumor biology.

Biomarker analysis of other VEGF pathway-targeting and checkpoint inhibitor treatment combinations have been performed, notably axitinib/avelumab in JAVELIN Renal 101 (30) and bevacizumab/atezolizumab in IMmotion 150 and IMmotion 151 (27, 29). All of these studies also incorporated a sunitinib treatment arm. In JAVELIN Renal 101, as observed in this study, PD-L1 expression did not distinguish PFS benefit in the axitinib/avelumab arm, while patients whose tumors contained a greater number of CD8+ cells had extended PFS (30). Patients with high angiogenic gene expression signatures (GES) at baseline had significantly improved PFS in the sunitinib arm but not the axitinib/avelumab arm, whereas patients with high effector T-cell (Teff) and T cell–inflamed GES at baseline had longer PFS in the axitinib/avelumab arm versus the sunitinib arm. This should be expected because the major difference between treatment arms is the addition of the immune checkpoint inhibitor avelumab. Patients in the axitinib/avelumab arm whose tumors at baseline were positive for novel immune-related GES that incorporated pathway indicators for T- and natural killer–cell activation and IFNγ, had longer PFS than those with negative tumors. Consistent with these findings, the results presented in this study showed that the small 5-gene angiogenesis signature did not distinguish PFS benefit of the axitinib/pembrolizumab treatment. In comparison, the 12-gene signature, which included markers of TCR signaling, T-cell activation, proliferation, differentiation, and chemokines, and was therefore similar to the 26-gene signature identified in JAVELIN Renal 101, segregated a group of patients greatly benefiting from combined treatment. In summary, the angiogenic markers matter for VEGF-only treatment, but only T-effector markers mattered for the combination treatment in distinguishing it from antiangiogenic therapy.

The results from JAVELIN Renal 101, IMmotion 150, and IMmotion 151 appear to show clear associations between immunomodulatory biomarkers and positive treatment outcomes with combined anti-VEGF pathway/checkpoint inhibitor treatment. In comparison, angiogenic biomarkers seem to be more associated with relatively better outcomes with anti-VEGF (i.e., sunitinib) treatment only. The results from these analyses suggest that for combined axitinib/pembrolizumab treatment, assessment of immunomodulatory-related biologic profiles would best inform efficacy outcomes. This hypothesis is supported by the findings reported here, where immunomodulatory-related markers are associated with better treatment outcomes, while a similar trend is not observed for the angiogenic-related markers. Thus, further evaluation of immunomodulatory-related biomarkers appears to be a suitable strategy to identify predictive biomarkers for RCC treatment with combined axitinib/pembrolizumab, and perhaps TKI/checkpoint inhibitor combinations in general.

This study had several limitations. This was a small phase Ib single-arm study, leading to a small sample size, and the original study was not designed to assess associations between biologic profiles and treatment outcomes, therefore making it difficult to draw firm conclusions. Because of limited tissue availability, analysis of tumor biomarkers was prioritized PD-L1 > CD8 > CD68, and as a result, there were few available samples for analysis of CD68. The low expression of PD-L1 in the overall sample increases the variance in the estimated difference of the proportions, which made it more difficult to achieve a significant association with outcomes. Although adjusted P values were obtained by calculating the FDR using Benjamini and Hochberg's step-up method, these results should be considered exploratory, and confirmatory studies are required to validate the findings. Finally, median overall survival was not reached and the great majority of patients benefited from treatment, with ORR of 73.1% and tumor shrinkage in 90% of patients (19).

In this hypothesis-generating exploratory analysis of a broad selection of candidate biomarkers, no markers were identified that clearly predict outcomes with axitinib/pembrolizumab. However, there are multiple possibilities that could be explored in further prospective analysis in an independent cohort with large sample size, such as tumor CD8 expression, blood levels of certain chemokines (e.g., CXCL10), and tumor gene expression levels of the 12-gene signature (AKT3, DNMT1, IL2RG, PTPN11, MTCP1, TNFSF10, CCR3, PDCD1, TBX21, IDO1, NOS3, and IFNG). The putative markers identified here could also be candidates for further exploration in other TKI/immune checkpoint inhibitor trials once the datasets are available. The markers identified in this study were largely immunomodulatory-related markers, appearing to be consistent with previous experience of TKI/checkpoint inhibitor combinations. Identifying candidate biomarkers is further complicated because with combined TKI/checkpoint inhibitor treatment it is difficult to know which is the dominant antitumor mode of action and what is the most appropriate outcome to assess. However, continued efforts are required to identify suitable biomarkers so that patients who would most benefit from combined axitinib/pembrolizumab can receive appropriate treatment, and to understand mechanisms of resistance.

J.-F. Martini reports personal fees from Pfizer Inc (employment) outside the submitted work; and reports (i) a patent for Cancer Treatment, WO 2015/162532 (owned by Pfizer Inc.), which is a method of treatment, around the rationale to combine an antiangiogenic agent with a checkpoint inhibitor, and (ii) a patent for Combination of a PD-1 Antagonist and a VEGFR Inhibitor for Treating Cancer, WO 2015/119930 (co-owned by Pfizer Inc. and Merck Sharp & Dohme Corp.), which more specifically covers the potential use and benefit of combining a VEGFR inhibitor such as axitinib with an anti–PD-1 antibody such as pembrolizumab, both unlicensed. E.R. Plimack reports grants from Pfizer (contracted payment to run clinical trial) during the conduct of the study; grants from Astellas, BMS, Genentech, Merck, AstraZeneca, Peloton (contracted payment to run clinical trial); personal fees from BMS, Clovis, Exelexis, Flatiron, Genentech, Incyte, Janssen, Merck, Seattle Genetics (scientific advisory), Pfizer, AstraZeneca, Infinity Pharma (data safety monitoring) outside the submitted work; a patent for U.S. Patent Application No.: 14/588,503 pending “Methods for Screening Muscle Invasive Bladder Cancer Patients for Neoadjuvant Chemotherapy Responsiveness. Filed 1/2/2015,” owned by FCCC; and honoraria for CME-certified presentations from the following entities: AUA, Clinical Care Options, Fox Chase Cancer Center, Georgetown, ASCO, Medscape, Mount Sinai Ichan School of Medicine, NCCN, Omniprex, OncLive, PER, prIME Oncology, Research to Practice, Spire Learning, University of Pennsylvania, Thomas Jefferson University, and the University of Michigan. T.K. Choueiri reports grants, personal fees, nonfinancial support, and other from Merck (trials, advisory board, consultancy) and Pfizer (trials, advisory boards, consultantcy) during the conduct of the study; grants, personal fees, nonfinancial support, and other from BMS (advisory boards, clinical trials, honorarium), EMD (advisory boards, clinical trials, honorarium), Novartis (advisory boards, clinical trials, honorarium), Exelixis (advisory boards, clinical trials, honorarium), Roche (advisory boards, clinical trials, honorarium), and Peleton (advisory boards, clinical trials, honorarium) outside the submitted work; and research (institutional and personal) from AstraZeneca, Alexion, Bayer, Bristol-Myers Squibb/ER Squibb and Sons LLC, Cerulean, Eisai, Foundation Medicine Inc., Exelixis, Ipsen, Tracon, Genentech, Roche, Roche Products Limited, F. Hoffmann-La Roche, GlaxoSmithKline, Lilly, Merck, Novartis, Peloton, Pfizer, Prometheus Labs, Corvus, Calithera, Analysis Group, Sanofi/Aventis, Takeda; honoraria from AstraZeneca, Alexion, Sanofi/Aventis, Bayer, Bristol-Myers Squibb/ER Squibb and Sons LLC, Cerulean, Eisai, Foundation Medicine Inc., Exelixis, Genentech, Roche, Roche Products Limited, F. Hoffmann-La Roche, GlaxoSmithKline, Merck, Novartis, Peloton, Pfizer, EMD Serono, Prometheus Labs, Corvus, Ipsen, Up-to-Date, NCCN, Analysis Group, Michael J. Hennessy (MJH) Associates, Inc (Healthcare Communications Company with several brands such as OnClive, PeerView, and PER), Research to Practice, L-path, Kidney Cancer, Clinical Care Options, Platform Q, Navinata Healthcare, Harborside Press, American Society of Medical Oncology, New England Journal of Medicine, Lancet Oncology, Heron Therapeutics, Lilly Oncology; consulting or advisory role at AstraZeneca, Alexion, Sanofi/Aventis, Bayer, Bristol-Myers Squibb/ER Squibb and Sons LLC, Cerulean, Eisai, Foundation Medicine Inc., Exelixis, Genentech, Heron Therapeutics, Lilly, Roche, GlaxoSmithKline, Merck, Novartis, Peloton, Pfizer, EMD Serono, Prometheus Labs, Corvus, Ipsen, Up-to-Date, NCCN, Analysis Group, Pionyr, Tempest, Lilly Ventures; stock ownership in Pionyr, Tempest; other present or past leadership roles as Director of GU Oncology Division at Dana-Farber and past President of Medical Staff at Dana-Farber, member of NCCN Kidney Panel and the GU Steering Committee, past chairman of the Kidney Cancer Association Medical and Scientific Steering Committee, KidneyCan advisory board, Kidney Cancer Research Summit co-chair (2019); patent owned by DFCI related to biomarkers of IO and circulating cfmethDNA.; travel, accommodations, expenses, in relation to consulting, advisory roles, or honoraria; medical writing and editorial assistance support may have been funded by communications companies funded by pharmaceutical companies (ClinicalThinking, Envision Pharma Group, Fishawack Group of Companies, Health Interactions, Parexel, Oxford PharmaGenesis, and others); the institution (Dana-Farber Cancer Institute) may have received additional independent funding of drug companies or/and royalties potentially involved in research around the subject matter; CV provided upon request for scope of clinical practice and research; mentored several non-US citizens on research projects with potential funding (in part) from non-US sources/Foreign Components; other support in part by the Dana-Farber/Harvard Cancer Center Kidney SPORE and Program, the Kohlberg Chair at Harvard Medical School, the Trust Family, Michael Brigham, and Loker Pinard Funds for Kidney Cancer Research at DFCI, and various National Cancer Institute (NCI), Department of Defense (DOD), foundations and industry grants. D.F. McDermott reports personal fees from Merck (consulting) and personal fees from Pfizer (consulting) during the conduct of the study; personal fees from BMS (consulting) and from Exelixis (consulting) outside the submitted work. I. Puzanov reports personal fees from Amgen and Merck outside the submitted work. M.N. Fishman reports other from Pfizer (contract to the employer for the trial) during the conduct of the study and has a patent 10300245 issued to none (immunotherapy and radiation). D.C. Cho reports personal fees from Nektar Therapeutics, HUYA, PureTech, Pfizer, and Torque outside the submitted work. U. Vaishampayan reports grants, personal fees, and nonfinancial support from Pfizer Inc (manuscript preparation and study costs to organization) during the conduct of the study; grants and personal fees from Exelixis, Bayer, Astellas, and Sanofi outside the submitted work. B. Rosbrook reports other from Pfizer Inc (empoyee of Pfizer and holds stocks) during the conduct of the study; other from Pfizer Inc (employee of Pfizer and holds stocks) outside the submitted work. K.C. Fernandez reports other from Pfizer Inc (employee of Pfizer and holds stocks) during the conduct of the study; other from Pfizer Inc (employee of Pfizer and holds stocks) outside the submitted work. J.C. Tarazi reports other from Pfizer Inc. (employed by the sponsor) during the conduct of the study; other from Pfizer Inc. (stocks) outside the submitted work; patent WO 2015/119930 for Combination of a PD-1 Antagonist and a VEGFR Inhibitor for Treating Cancer pending, owned by Pfizer Inc. and Merck Sharp & Dohme Corp., and patent WO 2015/162532 on cancer treatment, owned by Pfizer Inc. S. George reports grants and personal fees from BMS (institutional grant), Bayer (institutional grant), Pfizer (institutional grant), Corvus (institutional grant), Seattle Genetics/Astellas (institutional grant), Eisai (institutional grant), and Merck (institutional grant); personal fees from Genentech, Sanofi/Genzyme, EMD Serono, Exelixis; and grants from Agensys (institutional grant), Novartis (institutional grant), Calithera (institutional grant), and Immunomedics (institutional grant) outside the submitted work. M.B. Atkins reports grants from Pfizer (study sponsor) during the conduct of the study; grants and personal fees from Pfizer (advisory board, research support to institution), Merck (advisory board, research support to institution), BMS (advisory board, research support to institution), and Genentech-Roche (advisory board, research support to institution); grants from AstraZeneca (research support to institution), Calathera (research support to institution), DOD (to institution); and personal fees from Pyxis Oncology (advisory board), Exelixis (advisory board), Eisai (advisory board), Novartis (advisory board), Werewolf (advisory board), Agenus (consultant), and Leads BioPharma (advisory board) outside the submitted work. No other potential conflicts of interest were disclosed.

J.-F. Martini: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. E.R. Plimack: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. T.K. Choueiri: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. D.F. McDermott: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. I. Puzanov: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. M.N. Fishman: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. D.C. Cho: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. U. Vaishampayan: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. B. Rosbrook: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. K.C. Fernandez: Acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. J.C. Tarazi: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. S. George: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript. M.B. Atkins: Conception and design; development and methodology; acquisition of data; analysis and interpretation of data; writing, review and/or revision of manuscript.

We thank the participating patients and their families, as well as the investigators, subinvestigators, research nurses, study coordinators, and operations staff. We would also like to thank Amber Donahue, PhD, for her support with the biomarker analyses for this study, and Keith Ching, PhD, and Xinmeng Jasmine Mu, PhD, for their computational biology support. This study was sponsored by Pfizer in collaboration with Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc, Kenilworth, NJ, USA. The authors acknowledge the contribution of Merck & Co., Inc., Kenilworth, NJ, USA and Pfizer study team members. Medical writing support was provided by David Cope, PhD, of Engage Scientific Solutions and funded by Pfizer.

Upon request, and subject to certain criteria, conditions, and exceptions (see https://www.pfizer.com/science/clinical-trials/trial-data-and-results for more information), Pfizer will provide access to individual deidentified participant data from Pfizer-sponsored global interventional clinical studies conducted for medicines, vaccines, and medical devices: (i) for indications that have been approved in the United States and/or European Union or (ii) in programs that have been terminated (i.e., development for all indications has been discontinued). Pfizer will also consider requests for the protocol, data dictionary, and statistical analysis plan. Data may be requested from Pfizer trials 24 months after study completion. The deidentified participant data will be made available to researchers whose proposals meet the research criteria and other conditions, and for which an exception does not apply, via a secure portal. To gain access, data requestors must enter into a data access agreement with Pfizer.

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.
Choueiri
TK
,
Motzer
RJ
. 
Systemic therapy for metastatic renal-cell carcinoma
.
N Engl J Med
2017
;
376
:
354
66
.
2.
Dutcher
JP
. 
Update on the biology and management of renal cell carcinoma
.
J Investig Med
2019
;
67
:
1
10
.
3.
Nabi
S
,
Kessler
ER
,
Bernard
B
,
Flaig
TW
,
Lam
ET
. 
Renal cell carcinoma: a review of biology and pathophysiology
.
F1000Res
2018
;
7
:
307
.
4.
Sanchez
DJ
,
Simon
MC
. 
Genetic and metabolic hallmarks of clear cell renal cell carcinoma
.
Biochim Biophys Acta Rev Cancer
2018
;
1870
:
23
31
.
5.
Escudier
B
,
Pluzanska
A
,
Koralewski
P
,
Ravaud
A
,
Bracarda
S
,
Szczylik
C
, et al
Bevacizumab plus interferon alfa-2a for treatment of metastatic renal cell carcinoma: a randomised, double-blind phase III trial
.
Lancet
2007
;
370
:
2103
11
.
6.
Motzer
RJ
,
Hutson
TE
,
Tomczak
P
,
Michaelson
MD
,
Bukowski
RM
,
Oudard
S
, et al
Overall survival and updated results for sunitinib compared with interferon alfa in patients with metastatic renal cell carcinoma
.
J Clin Oncol
2009
;
27
:
3584
90
.
7.
Rini
BI
,
Halabi
S
,
Rosenberg
JE
,
Stadler
WM
,
Vaena
DA
,
Archer
L
, et al
Phase III trial of bevacizumab plus interferon alfa versus interferon alfa monotherapy in patients with metastatic renal cell carcinoma: final results of CALGB 90206
.
J Clin Oncol
2010
;
28
:
2137
43
.
8.
Sternberg
CN
,
Hawkins
RE
,
Wagstaff
J
,
Salman
P
,
Mardiak
J
,
Barrios
CH
, et al
A randomised, double-blind phase III study of pazopanib in patients with advanced and/or metastatic renal cell carcinoma: final overall survival results and safety update
.
Eur J Cancer
2013
;
49
:
1287
96
.
9.
Hutson
TE
,
Lesovoy
V
,
Al-Shukri
S
,
Stus
VP
,
Lipatov
ON
,
Bair
AH
, et al
Axitinib versus sorafenib as first-line therapy in patients with metastatic renal-cell carcinoma: a randomised open-label phase 3 trial
.
Lancet Oncol
2013
;
14
:
1287
94
.
10.
Motzer
RJ
,
Hutson
TE
,
McCann
L
,
Deen
K
,
Choueiri
TK
. 
Overall survival in renal-cell carcinoma with pazopanib versus sunitinib
.
N Engl J Med
2014
;
370
:
1769
70
.
11.
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
.
12.
McDermott
DF
,
Lee
JL
,
Szczylik
C
,
Donskov
F
,
Malik
J
,
Alekseev
B
, et al
Pembrolizumab monotherapy as first-line therapy in advanced clear cell renal cell carcinoma (accRCC): results from cohort A of KEYNOTE-427
.
J Clin Oncol
36
:
15s
, 
2018
(
suppl; abstr 4500
).
13.
Motzer
RJ
,
Tannir
NM
,
McDermott
DF
,
Aren 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
.
14.
Lalani
AA
,
McGregor
BA
,
Albiges
L
,
Choueiri
TK
,
Motzer
R
,
Powles
T
, et al
Systemic treatment of metastatic clear cell renal cell carcinoma in 2018: current paradigms, use of immunotherapy, and future directions
.
Eur Urol
2019
;
75
:
100
10
.
15.
Bellesoeur
A
,
Carton
E
,
Alexandre
J
,
Goldwasser
F
,
Huillard
O
. 
Axitinib in the treatment of renal cell carcinoma: design, development, and place in therapy
.
Drug Des Devel Ther
2017
;
11
:
2801
11
.
16.
Hutson
TE
,
Al-Shukri
S
,
Stus
VP
,
Lipatov
ON
,
Shparyk
Y
,
Bair
AH
, et al
Axitinib versus sorafenib in first-line metastatic renal cell carcinoma: overall survival from a randomized phase III trial
.
Clin Genitourin Cancer
2017
;
15
:
72
6
.
17.
Khoja
L
,
Butler
MO
,
Kang
SP
,
Ebbinghaus
S
,
Joshua
AM
. 
Pembrolizumab
.
J Immunother Cancer
2015
;
3
:
36
.
18.
Donskov
F
,
McDermott
DF
,
Lee
JL
,
Szczylik
C
,
Malik
J
,
Alekseev
BY
, et al
KEYNOTE-427 Cohort A: pembrolizumab monotherapy as first-line therapy in advanced clear cell renal cell carcinoma (ccRCC)
.
Ann Oncol
2018
;
29
:
mdy283.080
.
19.
Atkins
MB
,
Plimack
ER
,
Puzanov
I
,
Fishman
MN
,
McDermott
DF
,
Cho
DC
, et al
Axitinib in combination with pembrolizumab in patients with advanced renal cell cancer: a non-randomised, open-label, dose-finding, and dose-expansion phase 1b trial
.
Lancet Oncol
2018
;
19
:
405
15
.
20.
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
.
21.
Nakano
O
,
Sato
M
,
Naito
Y
,
Suzuki
K
,
Orikasa
S
,
Aizawa
M
, et al
Proliferative activity of intratumoral CD8(+) T-lymphocytes as a prognostic factor in human renal cell carcinoma: clinicopathologic demonstration of antitumor immunity
.
Cancer Res
2001
;
61
:
5132
6
.
22.
Bromwich
EJ
,
McArdle
PA
,
Canna
K
,
McMillan
DC
,
McNicol
AM
,
Brown
M
, et al
The relationship between T-lymphocyte infiltration, stage, tumour grade and survival in patients undergoing curative surgery for renal cell cancer
.
Br J Cancer
2003
;
89
:
1906
8
.
23.
Siddiqui
SA
,
Frigola
X
,
Bonne-Annee
S
,
Mercader
M
,
Kuntz
SM
,
Krambeck
AE
, et al
Tumor-infiltrating Foxp3-CD4+CD25+ T cells predict poor survival in renal cell carcinoma
.
Clin Cancer Res
2007
;
13
:
2075
81
.
24.
Choueiri
TK
,
Figueroa
DJ
,
Fay
AP
,
Signoretti
S
,
Liu
Y
,
Gagnon
R
, et al
Correlation of PD-L1 tumor expression and treatment outcomes in patients with renal cell carcinoma receiving sunitinib or pazopanib: results from COMPARZ, a randomized controlled trial
.
Clin Cancer Res
2015
;
21
:
1071
7
.
25.
Liu
XD
,
Hoang
A
,
Zhou
L
,
Kalra
S
,
Yetil
A
,
Sun
M
, et al
Resistance to antiangiogenic therapy is associated with an immunosuppressive tumor microenvironment in metastatic renal cell carcinoma
.
Cancer Immunol Res
2015
;
3
:
1017
29
.
26.
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
.
27.
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
.
28.
Rini
BI
,
Powles
T
,
Atkins
MB
,
Escudier
B
,
McDermott
DF
,
Suarez
C
, et al
Atezolizumab plus bevacizumab versus sunitinib in patients with previously untreated metastatic renal cell carcinoma (IMmotion151): a multicentre, open-label, phase 3, randomised controlled trial
.
Lancet
2019
;
393
:
2404
15
.
29.
Rini
BI
,
Huseni
M
,
Atkins
MB
,
McDermott
DF
,
Powles
TB
,
Escudier
B
, et al
Molecular correlates differentiate response to atezolizumab (atezo) + bevacizumab (bev) vs sunitinib (sun): Results from a phase III study (IMmotion151) in untreated metastatic renal cell carcinoma (mRCC)
.
Ann Oncol
2018
;
29
:
724
5
.
30.
Choueiri
TK
,
Albiges
L
,
Haanen
J
,
Larkin
JMG
,
Uemura
M
,
Pal
SK
, et al
Biomarker analyses from JAVELIN Renal 101: avelumab + axitinib (A+Ax) versus sunitinib (S) in advanced renal cell carcinoma (aRCC)
.
J Clin Oncol
37
:
15s
, 
2019
(
suppl; abstr 101
).
31.
Robins
HS
,
Desmaris
C
,
Matthis
J
,
Livingston
R
,
Andriesen
J
,
Reijonen
H
, et al
Ultra-sensitive detection of rare T cell clones
.
J Immunol Methods
2012
;
375
:
14
9
.
32.
Snyder
A
,
Nathanson
T
,
Funt
SA
,
Ahuja
A
,
Buros Novik
J
,
Hellmann
MD
, et al
Contribution of systemic and somatic factors to clinical response and resistance to PD-L1 blockade in urothelial cancer: an exploratory multi-omic analysis
.
PLoS Med
2017
;
14
:
e1002309
.
33.
Robins
HS
,
Campregher
PV
,
Srivastava
SK
,
Wacher
A
,
Turtle
CJ
,
Kahsai
O
, et al
Comprehensive assessment of T-cell receptor beta-chain diversity in alphabeta T cells
.
Blood
2009
;
114
:
4099
107
.
34.
Kirsch
I
,
Vignali
M
,
Robins
H
. 
T-cell receptor profiling in cancer
.
Mol Oncol
2015
;
9
:
2063
70
.
35.
Herbst
RS
,
Baas
P
,
Kim
DW
,
Felip
E
,
Perez-Gracia
JL
,
Han
JY
, et al
Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial
.
Lancet
2016
;
387
:
1540
50
.
36.
Daley
T
,
Smith
AD
. 
Predicting the molecular complexity of sequencing libraries
.
Nat Methods
2013
;
10
:
325
7
.
37.
Motzer
RJ
,
Hutson
TE
,
Tomczak
P
,
Michaelson
MD
,
Bukowski
RM
,
Rixe
O
, et al
Sunitinib versus interferon alfa in metastatic renal-cell carcinoma
.
N Engl J Med
2007
;
356
:
115
24
.
38.
Benjamini
Y
,
Hochberg
Y
. 
Controlling the false discovery rate: a practical and powerful approach to multiple testing
.
J R Stat Soc Series B Methodol
1995
;
57
:
289
300
.