TNF blockade uncouples toxicity from antitumor efficacy induced with CD40 chemoimmunotherapy

Abstract

Agonist CD40 antibodies are under clinical development in combination with chemotherapy as an approach to prime for antitumor T cell immunity. However, treatment with anti-CD40 is commonly accompanied by both systemic cytokine release and liver transaminase elevations, which together account for the most common dose-limiting toxicities. Moreover, anti-CD40 treatment increases the potential for chemotherapy-induced hepatotoxicity. Here, we report a mechanistic link between cytokine release and hepatotoxicity induced by anti-CD40 when combined with chemotherapy and show that toxicity can be suppressed without impairing therapeutic efficacy. We demonstrate in mice and humans that anti-CD40 triggers transient hepatotoxicity marked by increased serum transaminase levels. In doing so, anti-CD40 sensitizes the liver to drug-induced toxicity. Unexpectedly, this biology is not blocked by the depletion of multiple myeloid cell subsets, including macrophages, inflammatory monocytes, and granulocytes. Transcriptional profiling of the liver after anti-CD40 revealed activation of multiple cytokine pathways including TNF and IL-6. Neutralization of TNF, but not IL-6, prevented sensitization of the liver to hepatotoxicity induced with anti-CD40 in combination with chemotherapy without impacting antitumor efficacy. Our findings reveal a clinically feasible approach to mitigate toxicity without impairing efficacy in the use of agonist CD40 antibodies for cancer immunotherapy.

Keywords: Cancer immunotherapy; Cytokines; Immunology.

Figure 1. Systemic CD40 activation sensitizes the liver to chemotherapy-induced hepatotoxicity.

(A) Treatment schema. Patients with chemotherapy-naive, surgically incurable PDAC received gemcitabine (1000 mg/m²) infused on days 1, 8, and 15 of each 28-day cycle, with CP-870,893 administered once on day 3 of each cycle. (B) ALT and (C) AST serum levels in patients treated as shown in A. n = 22 patients. One-way ANOVA with comparison to baseline (Pre) was performed. C2D1, cycle 2 day 1. (D) Study schema for E–G. Shown are (E) ALT serum levels and (F) number of lesions/mm² in the liver detected on the day of analysis (shown in parentheses) after the indicated treatment. (F) Kruskal-Wallis with Dunn’s multiple comparisons test was performed. (G) Mouse weight over time after treatment (indicated by arrows). (H) Study schema for I–K. Shown are (I) ALT serum levels and (J) number of lesions/mm² in the liver detected on day 2 after αCD40 treatment. (J) Brown-Forsythe and Welch’s 1-way ANOVA test with Dunnett’s T3 multiple-comparison test was performed. (K) Mouse weight over time. (L) Study schema for M–O. Shown are (M) ALT serum levels and (N) number of lesions/mm² in the liver detected on day 2 after gemcitabine treatment. (N) Kruskal-Wallis with Dunn’s multiple-comparison test was performed. (O) Mouse weight over time. For D–O, n = 8 mice per group. Data are representative of ≥ 3 experimental replicates in control and αCD40→2d→Gem treated groups, ≥ 1 experimental replicate for all other groups. For G, K, and O, data shown are mean ± SEM with significance tested on day 2, and ordinary 1-way ANOVA with Dunnett’s multiple-comparison tests were performed. All other data shown are mean ± SD. For E, I, and M, red lines indicate upper range of the 95% CI for normal serum level of ALT derived from all experiments in the manuscript, and Kruskal-Wallis with Dunn’s multiple-comparison test was performed. Gem, gemcitabine; αCD40, clone FGK45; AST, aspartate aminotransferase; ALT, alanine aminotransferase. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 2. Myeloid cells are dispensable for hepatotoxicity triggered by chemoimmunotherapy.

For A–C, mice (n = 3–6) were treated with control or αCD40, and liver was collected 2 days later for analysis. Shown are heatmaps from QuantSeq 3′ mRNA sequencing for (A) chemotaxis/acute inflammatory response-associated genes and (B) cell-specific markers. Heatmaps were generated from normalized FPKM values. (C) Quantification of cells expressing Ly6G and MPO, and clusters of cells (>900 mm²) expressing F4/80 in liver tissue detected by IHC. Mann-Whitney U tests were performed. n = 1 experimental replicate. For D and E, mice (n = 8 per group) were treated with αCD40 on day –2 and Gem on day 0. Analysis was performed on day 2. Myeloid-depleting agents were given as follows: CEL (days –4, –1); αLy6C (days –3, –2, –1, 0); αLy6G (days –3, –2, –1, 0); and αCSF1R (days –4, –2, 0). (D) ALT serum levels detected on day 2 after gemcitabine. Significance was tested with Kruskal-Wallis with Dunn’s multiple-comparison test. Significant comparisons with control are shown. n = 2 experimental replicates for control, αCD40→Gem, and αCD40→Gem+αCSF1R. n = 1 experimental replicate for all other groups. Red line indicates the upper range of the 95% CI for normal serum level of ALT derived from all experiments in the manuscript. (E) Mouse weight over time. Significance was tested with ordinary 1-way ANOVA with Dunnett’s multiple-comparison for weight on day 2. Data shown are mean ± SEM. n = 4 experimental replicates for control, αCD40→Gem, and αCD40→Gem+αCSFR. n = 2 experimental replicates for αCD40→Gem+CEL and αCD40→Gem+CEL+αLy6C. n = 1 experimental replicate for αCD40→Gem+αLy6G. All other data shown are mean ± SD. Gem, gemcitabine; αCD40, clone FGK45; CEL, clodronate-encapsulated liposomes; ALT, alanine aminotransferase; FPKM, fragments per kilobase of transcript per million mapped reads. **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 3. TNF is necessary for hepatotoxicity produced with chemoimmunotherapy.

(A) RNA was extracted from bulk liver tissue of control- (Ctrl) or αCD40-treated mice 2 days after treatment. Gene expression for Stat1, Stat2, Stat3, Il6, and Tnf displayed as FPKM detected using QuantSeq 3′ mRNA sequencing. n = 3–6 mice/group, 1 experimental replicate. (B) Quantification by IHC of phosphorylated (p-) STAT1, p-STAT3, and p–NF-κBp65 protein expression in liver tissues collected 2 days after treatment with αCD40 compared with control. (C) Representative images and (D) quantification of Tnf expression detected by RNA-ISH in the liver 2 days after αCD40 treatment compared with control. Positive and negative controls for RNA-ISH are shown. Scale bars: 50 μm. Insets wre generated by zooming in on indicated 50 x 50 μm regions. (A–D) n = 8 mice/group, 1 experimental replicate. Mann-Whitney U tests were performed. (E) Study schema for F–H. n = 8 mice/group, 2 experimental replicates. (F) Mouse weight pretreatment and posttreatment on day 2. Paired 2-tailed t tests were performed. (G) Number of lesions/mm² in the liver detected by H&E stain. (H) ALT serum levels on day 4. Red line indicates upper range of 95% CI for normal serum level of ALT derived from all experiments in the manuscript. (I) IFN-γ serum levels detected 24 hours after treatment with αCD40 in control, NSG, and Rag2^–/– mice. (J) TNF serum levels detected 1 day after treatment with αCD40 (compared with control). Anti–IFN-γ and isotype control (IgG1) were given on days –1 and 0. αCD40 and isotype control (IgG2a) were given on day 0. (I and J) n = 6–8 mice per group, 2 experimental replicates. Significance was tested using (G and I) Kruskal-Wallis with Dunn’s multiple-comparison test and (H and J) ordinary 1-way ANOVA with Dunnett’s multiple-comparison test. In G–J, comparisons with control and αCD40→Gem (G and H) or αCD40 (I and J) are shown. Data shown are mean ± SD. Gem, gemcitabine; αCD40, clone FGK45; ALT, alanine aminotransferase; FPKM, fragments per kilobase of transcript per million mapped reads. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 4. TNF blockade inhibits hepatotoxicity due to chemoimmunotherapy without affecting treatment efficacy.

(A) Study schema used in B–D. Shown are (B) mean mouse weights over time, (C) overall survival, and (D) mean tumor growth curves. Statistical significance was determined using the following tests: in B, ordinary 1-way ANOVA with Dunnett’s multiple-comparison test was performed on weights on day 2; in C, Mantel-Cox test was used; and in D, Kruskal-Wallis with Dunn’s multiple-comparison test was performed on tumor volume on day 44. n = 10 mice/group. Data are representative of n = 2 experimental replicates. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.