Immunomodulatory Activity of Nivolumab in Metastatic Renal Cell Carcinoma

Abstract

Purpose: Nivolumab, an anti-PD-1 immune checkpoint inhibitor, improved overall survival versus everolimus in a phase 3 trial of previously treated patients with metastatic renal cell carcinoma (mRCC). We investigated immunomodulatory activity of nivolumab in a hypothesis-generating prospective mRCC trial.

Experimental design: Nivolumab was administered intravenously every 3 weeks at 0.3, 2, or 10 mg/kg to previously treated patients and 10 mg/kg to treatment-naïve patients with mRCC. Baseline and on-treatment biopsies and blood were obtained. Clinical activity, tumor-associated lymphocytes, PD-L1 expression (Dako immunohistochemistry; ≥5% vs. <5% tumor membrane staining), tumor gene expression (Affymetrix U219), serum chemokines, and safety were assessed.

Results: In 91 treated patients, median overall survival [95% confidence interval (CI)] was 16.4 months [10.1 to not reached (NR)] for nivolumab 0.3 mg/kg, NR for 2 mg/kg, 25.2 months (12.0 to NR) for 10 mg/kg, and NR for treatment-naïve patients. Median percent change from baseline in tumor-associated lymphocytes was 69% (CD3⁺), 180% (CD4⁺), and 117% (CD8⁺). Of 56 baseline biopsies, 32% had ≥5% PD-L1 expression, and there was no consistent change from baseline to on-treatment biopsies. Transcriptional changes in tumors on treatment included upregulation of IFNγ-stimulated genes (e.g., CXCL9). Median increases in chemokine levels from baseline to C2D8 were 101% (CXCL9) and 37% (CXCL10) in peripheral blood. No new safety signals were identified.

Conclusions: Immunomodulatory effects of PD-1 inhibition were demonstrated through multiple lines of evidence across nivolumab doses. Biomarker changes from baseline reflect nivolumab pharmacodynamics in the tumor microenvironment. These data may inform potential combinations. Clin Cancer Res; 22(22); 5461-71. ©2016 AACR.

Figure 1

Evaluation of tumor-associated lymphocytes in tumor biopsies obtained at baseline and at C2D8 of nivolumab treatment. A. Immunohistochemistry for CD3⁺, CD4⁺, and CD8⁺ cells. Scale bars are denoted on the images. Top two panels: CD3 (red), CD8 (brown); bottom two panels: CD3 (brown), CD4 (purple), FoxP3 (nuclei, red). B. Change from baseline in percentage of cells that are CD3⁺, CD4⁺, or CD8⁺ in tumor biopsies. Data are included for patients with immunohistochemistry data at both baseline and C2D8 in all treatment groups combined (N = 36). C. Expression levels for genes CD3D (915_at), CD8A (925_at), and CD4 (920_at) in tumor biopsies. Values presented are least squares means of the (log-2) robust multi-array intensity for the treatment group and time point indicated. Error bars indicate 95% confidence intervals estimated from the extended linear model.

Figure 2

Change from baseline tumor gene expression for immune lineage-specific transcripts and 24-hour change from baseline in peripheral blood immune-specific transcripts. A. Change of the expression level in tumor biopsies of the 43 regulated transcripts (>1.3 fold, P < 0.01) that are specifically associated with either the lymphoid or myeloid immune lineage. Within the lymphoid lineage, the 10 transcripts indicated are specific to T cells. Data are included from the 42 patients with measures at both time points, separated by their previous treatment status. Genes labeled in orange are members of interferon-regulated transcription modules collated by the BRi2 consortium (34). Markers of immune cytolytic activity are labeled in green. B. Change of 30 transcripts in peripheral blood associated with immune lineages and significantly regulated (≥1.2-fold, P < 0.01) in all treatment groups at cycle 1 day 2. Data are included from the 70 patients with measures at both time points, separated by treatment group. Genes labeled in orange are members of interferon-regulated transcription modules collated by the BRi2 consortium (34).

Figure 3

Effect of nivolumab on chemokine markers. A. Fold change from baseline at cycle 2 day 8 vs. baseline in serum concentrations of CXCL9 and CXCL10 in all treatment groups (N = 83). Both axes are on the log base-2 scale. B. Scatter plot matrix of fold changes from baseline (log base-2 scale) in CXCL9 and CXCL10 among 83 patients who had serum data at baseline and C2D8. The diagonal panels give kernel density estimates and histograms summarizing the univariate distributions of CXCL9 and CXCL10 individually. C. Gene expression levels for CXCL9 (4283_at) and CXCL10 (3627_at) in fresh tumor tissue samples. Values presented are least squares means of the (log-2) robust multi-array intensity for the treatment group and time point indicated. Error bars indicate 95% confidence intervals estimated from the extended linear model. D. Gene expression levels for CXCL9 (4283_at) and CXCL10 (3627_at) in biopsies obtained at cycle 2 day 8 versus serum concentrations of CXCL9 and CXCL10 in the same patient at cycle 2 day 8 (N = 54). Both axes are on the log base-2 scale. Shaded area represents 95% confidence interval estimated from a linear model.