Background: Inhibition of immune-checkpoint targets including PD1 is clinically effective in a variety of cancers. However, only a subset of patients respond and complete response remains uncommon. To understand the mechanisms of response and resistance, recent studies have focused on neoantigens, copy-number alterations, and transcriptional signatures of tumor tissues collected from patients treated with immune-checkpoint inhibitors. Given the known role of metabolites in modulating immunity, we sought to understand how individual patients' metabolic activities adapt to PD1 immune checkpoint blockade and how they associate with therapeutic benefits. Methods: We profiled 106-202 metabolites in pre- and multiple on-treatment patient serum samples from three independent immunotherapy trials using liquid chromatography-mass spectrometry. These metabolites are involved in the metabolism of amino acids, nucleotides, nitrogen, and lipids, among others. Our study consisted of two Phase 1 trials (CA209-038 and -009) which included 78 patients with advanced melanoma and 91 patients with metastatic renal cell carcinoma (RCC) treated with nivolumab. RNASeq was performed on matched pre- and on-treatment tumor biopsies from the melanoma cohort. To investigate the generalizability of our results, we also performed metabolomics and serum specimens from a large randomized Phase 3 trial (CheckMate 025) with 743 RCC patients, among which 394 received nivolumab and 349 received everolimus. Results: During treatment with nivolumab, kynurenine, a product of tryptophan catabolism and the IDO/TDO genes, was the most significantly changed metabolite at week 4 and at week 6 compared to pre-treatment levels among melanoma patients (37% and 34% increase on average, q=1 × 10-10 and q=1 × 10-8 respectively, paired t-test). By using Kyn/Trp (Kynurenine/Tryptophan) as a metric indicating tryptophan-kynurenine conversion, we found that this ratio falls in a range spanning approximately 8-fold among these patients, suggesting prominent individual-to-individual differences. Specifically, 78% patients had any increases and 26.5% patients had increases above 50% at week 4. We confirmed this significant kynurenine up-regulation following nivolumab treatment in RCC patients in the phase 1 and phase 3 trials. In particular, the phase 3 cohort showed 23% and 24% increase on average at week 4 and week 8 respectively (q=1 × 10-10 and q=1 × 10-12, paired t-test). Additionally, 69.4% and 8.2% patients had Kyn/Trp increases above zero and 50% respectively at week 4. Notably, patients receiving everolimus control treatment had a decrease in kynurenine (q=1 × 10-5, t-test). Kynurenine is synthesized during tryptophan catabolism by indoleamine-2,3-dioxygenase (IDO) or tryptophan dioxygenase (TDO), and has been shown to suppress anti-tumor immune responses. To explore whether the circulating Kyn/Trp correlates with immune-suppression in the tumor microenvironment, we analyzed RNAseq data of matched tumor biopsies from the CA209-038 melanoma trial. We found a significant correlation between the Kyn/Trp ratio and PD-L1 expression, 4 weeks after starting nivolumab treatment (Pearson, p=0.01). We also discovered a correlation between Kyn/Trp and IDO1 but not TDO mRNA levels at the same time point. In contrast, Kyn/Trp was not associated with prior anti-CTLA4 treatment, or tumor mutational load. Moreover, compared to all other metabolites, increases of the Kyn/Trp ratio in the melanoma cohort (week 4 vs baseline) were consistently associated with greater risks for death (p=1.2*10-4, HR=2.71, 95% CI, 1.63-4.51). In particular, patients with a >50% increase in Kyn/Trp had a median OS of 15.7 months while those with decreases had a median survival time of > 38 months (p = 6.0*10-5, log-rank test). In contrast, the baseline Kyn/Trp ratio did not significantly associate with the melanoma patients' overall survival. To confirm this result, the association between Kyn/Trp ratios and overall survival in the larger phase 3 trial in RCC (CheckMate 025) was evaluated using serum samples collected at different time points. We found that at baseline, higher Kyn/Trp ratios associated with shorter overall survival both for the nivolumab- and the everolimus-treated patients (p = 3.6*10-4, HR=1.79, 95% CI, 1.30-2.47; p = 1.7*10-5, HR=2.06, 95% CI, 1.48-2.85; Cox model). However, at week 4, Kyn/Trp significantly associated with overall survival only in the nivolumab arm (p = 4.7*10-4, HR=2.81, 95% CI, 1.57-5.01; Cox model) but not in the everolimus arm (p = 0.53, HR=0.76, 95% CI, 0.32-1.78; Cox model). For nivolumab-treated RCC patients, those with a >50% increases of Kyn/Trp had a median survival of 16.7 months while those with any Kyn/Trp decreases had a median survival of 31.3 months (p = 4.3*10-4, log-rank test).Conclusions: We identified increased tryptophan to kynurenine conversion in response to PD1 blockade in a subset of melanoma and RCC patients. By using independent cohorts, we showed that Kyn/Trp temporal alterations robustly correlated with overall survival of patients receiving nivolumab. Our findings illustrate that checkpoint blockade in combination with IDO/TDO inhibitors might only benefit a selected group of patients with checkpoint-inhibition-triggered kynurenine pathway activation. Given the lack of improved therapeutic outcomes with PD1 and selective IDO1 inhibition among unselected patient populations in a recent phase 3 trial, our findings highlight the need and feasibility of patient stratification by monitoring serum Kyn/Trp alterations, show that kynurenine signaling may still be a relevant therapeutic target and more generally point to the relevance of metabolic adaptations in cancer immunotherapy. Our findings highlight the need and feasibility of patient stratification by monitoring serum Kyn/Trp alterations and point to the relevance of metabolic adaptations in cancer immunotherapy.

Citation Format: Haoxin Li, Kevin Bullock, Carino Gurjao, David Braun, Sachet A. Shukla, Dominick Bosse, Aly-Khan A. Lalani, Shuba Gopal, Chelsea Jin, Christine Horak, Megan Wind-Rotolo, Sabina Signoretti, David F. McDermott, Gordon J. Freeman, Eliezer M. Van Allen, Stuart L. Schreiber, Frank Stephen Hodi, William R. Sellers, Levi A. Garraway, Clary B. Clish, Toni K. Choueiri, Marios Giannakis. Metabolomic adaptations and correlates of survival to immune checkpoint blockade [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr NG13.