Ketogenic diet and ketone bodies enhance the anticancer effects of PD-1 blockade

Abstract

Limited experimental evidence bridges nutrition and cancer immunosurveillance. Here, we show that ketogenic diet (KD) - or its principal ketone body, 3-hydroxybutyrate (3HB), most specifically in intermittent scheduling - induced T cell-dependent tumor growth retardation of aggressive tumor models. In conditions in which anti-PD-1 alone or in combination with anti-CTLA-4 failed to reduce tumor growth in mice receiving a standard diet, KD, or oral supplementation of 3HB reestablished therapeutic responses. Supplementation of KD with sucrose (which breaks ketogenesis, abolishing 3HB production) or with a pharmacological antagonist of the 3HB receptor GPR109A abolished the antitumor effects. Mechanistically, 3HB prevented the immune checkpoint blockade-linked upregulation of PD-L1 on myeloid cells, while favoring the expansion of CXCR3+ T cells. KD induced compositional changes of the gut microbiota, with distinct species such as Eisenbergiella massiliensis commonly emerging in mice and humans subjected to carbohydrate-low diet interventions and highly correlating with serum concentrations of 3HB. Altogether, these results demonstrate that KD induces a 3HB-mediated antineoplastic effect that relies on T cell-mediated cancer immunosurveillance.

Keywords: Cancer; Immunotherapy; Metabolism; Mouse models; Oncology.

Conflict of interest statement

Conflict of interest: LZ, RD, and GK are founders of EverImmune.

Figures

Figure 1. Ketogenic diet decreases melanoma and renal cell tumor growth.

(A–B) Growth kinetics of RET melanoma in C57BL/6J mice. The minimal tumorigenic dose of RET melanoma cells was inoculated s.c. the day of starting the dietary intervention. Normal diet (ND) versus ketogenic diet (KD). Tumor size was monitored 3 times a week for 12 days. Each curve represents the tumor kinetics of 1 animal (A), and the mean ± SEM of tumor size was calculated for both nutritional interventions (B). The graphs depict the concatenation of 5 independent experiments involving 5–6 mice/group (n = 29–30 mice). (C) Monitoring of RENCA progression using bioluminescence imaging of luciferase activity in 4 representative mice among 15 BALB/c mice fed ND versus KD. (D) The ratio of luminescence at day 15 (D15) versus day 7 after orthotopic tumor injection was calculated for each diet (D15, IVIS measurement; D7, day of randomization). All experiments were composed of 5–7 mice/group and were performed at least twice, yielding similar results (n = 14 for ND and n = 12 for KD). Dedicated software (https://kroemerlab.shinyapps.io/TumGrowth/) (A and B) and Student’s t test. *P < 0.05 (D).

Figure 2. Ketone bodies accumulate in tissues of ketogenic diet fed mice.

(A) Follow up of mouse weight over time until sacrifice in tumor bearers fed normal diet (ND) versus ketogenic diet (KD) (n = 30 for each groups). (B and C) Heatmap of the nonsupervised hierarchical clustering highlighting differences in the metabolic profiling of C57BL/6J mice fed ND (n = 20) versus KD (n = 21) in plasma (B) and PCA based on Bray-Curtis Dissimilarity Index showing significant compositional differences across diet types (C) (PERMANOVA, P = 0.001). Arrows indicate 3-hydroxybutyrate (3HB) and acetoacetate positions in the heatmap. (D and E) Concentrations of acetoacetate and 3HB in 3 compartments of tumor bearers. A representative experiment involving 5 mice/group is depicted out of 2 metabolomics profiles yielding similar results (n = 3 in ND and n = 5 in KD). (F) Spearman correlations between levels of 3HB in plasma and RET tumor size at day 12 indicating r and P values for the KD group. Not significant for the ND group. ANOVA (A) and Student’s t test statistical analyses of mean ± SEM. *P < 0.05 (D and E).

Figure 3. Ketone body 3-hydroxybutyrate (3HB) is necessary and sufficient to account for the anticancer effects of ketogenic diet.

(A and B) Similar experimental setting as in Figure 1, A and B, comparing RET tumor size at day 12 of each nutritional intervention (normal diet [ND], ketogenic diet [KD], 3-hydroxybutyrate per os [3HBpo] versus 3-hydroxybutyrate i.p. [3HBip]) for all 12–15 mouse tumors from 3 independent experiments (A) and the complete regression rates for all the experiments performed (n, number of independent experiments encompassing 5–6 mice/group, SEM of percentage of tumor free mice across 3–7 experiments) (B). (C) Pharmacological inhibition of GPR109A (versus PBS as control) using i.p. daily administration of mepenzolate bromide (C21H26BrNO3). Tumor sizes at day 12 are depicted for each group. Data from 2–3 experiments involving 5–6 mice/group are depicted (n = 20 for ND, KD, and 3HBpo without mepenzolate and n = 24 for ND, KD, 3HBpo + mepenzolate. (D–F) Effects of sucrose supplementation on the antitumor (D) and metabolic effects (E and F) mediated by KD. Id. as in Figure 1, A and B, showing RET tumor sizes at day 12 after various dietary interventions (D) and the 2 ketone body (KB) plasma metabolites as in Figure 2, D and E (E and F). Results from 1 representative experiment out of 2 are depicted (n = 30 for group without sucrose and n = 12 with sucrose groups). Global comparison using Kruskall-Wallis test, with a post hoc multiple comparison using Dunn’s test (A–D); Student’s t test (E and F) (*P < 0.05, **P < 0.01, ***P < 0.001).

Figure 4. Ketogenic diet shifts the microbiota composition.

(A) PCoA representing the differences in β diversity of fecal microbiota between dietary interventions normal diet (ND) versus Ketogenic diet (KD) at day 12 in RET–tumor bearing mice. ANOSIM and PERMANOVA defines the separation of the groups; the P value defines the significance of such separation after 999 permutations of the samples. (B) For each principal coordinate axis (PCo1 and PCo2), the collected variance, the Pearson r coefficient and the corresponding P value are shown. Partial Least Squares Discriminant Analysis (PLS-DA) plot of the variance between KD- and ND-fed tumor-bearing mice. LEfSe plot of species at day 12 discriminating ND- from KD-fed tumor-bearing mice, ordered by their LDA score. (C) Detailed relative abundance of distinct species. Refer to Supplemental Figure 4 for additional taxa. (D) Effects of broad-spectrum antibiotics on the percentages of tumor-free mice after ND or KD or 3-hydroxybutyrate per os (3HBpo) after RET inoculation. Data from 2–3 experiments involving 5–6 mice/group are depicted (n = 12 for ND group and n = 21 for KD group). Student t test or ANOVA. **P < 0.01.***P < 0.001.

Figure 5. Correlations between microbial species and ketone bodies in mice and humans.

(A–C) Mouse data. Alignment of the taxa relevant in our mouse preclinical studies and the plasma ketone bodies (KB) and Spearman correlations between 2 parameters. The significant ones are indicated with an asterisk (P < 0.05), and their raw data are presented in B and C (each dot represents 1 mouse). (D) Human data. Heatmap showing a correlation matrix between estimates of carbohydrate intake from food questionnaire (FQ) and KB monitoring in plasma in > 1000 individuals from the PREDICT-1 study. The taxa highly significant in our mouse models are highlighted. Pearson correlation was used.

Figure 6. T cell–dependent effects of ketogenic diet and synergy with immune checkpoint blockade.

(A) Effects of T cell depletion on tumor sizes at day 12 of diet interventions. Tumor sizes in all diet-fed (normal diet [ND], ketogenic diet [KD], or 3-hydroxybutyrate per os [3HBpo]) groups in the presence of i.p. injections of anti-CD4 and anti-CD8 depleting Abs prior to tumor inoculation. A representative experiment out of 2 is depicted, both yielding similar conclusions; each dot represents 1 mouse, and each group contained 5–6 mice. Mann Whitney U statistical analyses was used. *P < 0.05. (B) Experimental setting for combinatorial regimen. (C and D) Effects of diet interventions on tumor size (C) at day 9 after RET inoculation and diet intervention, after only 1 i.p. administration of cICB (anti–CTLA-4 plus anti–PD-1 mAbs) and overall survival (D). Results of 4 concatenated experiments are depicted; each dot represents 1 mouse (n = 30 for ND and KD groups; n = 12 for 3HBpo and 3HBip). (C) Fold increase of tumor sizes in cICB groups fed with ND, KD, or 3HBpo or 3-hydroxybutyrate i.p. (3HBip) (normalized to ND groups). (D) Overall survival appreciated with Kaplan Meier curves for 12 animals/group after a complete treatment combining continuous diet plus cICB, according to B. (E) Flow cytometry analyses of Tc1 (defined as CXCR3+CD8+ T splenocytes) at day 9 of the combinatorial regimen; each dot represents 1 spleen. Results from 2 pooled experiments are shown. (F) Effects of sucrose supplementation on the synergy between diets and cICB against RET tumor progression. Individual tumor sizes at day 19 of 3 concatenated experiments involving 6 mice/group are shown for each diet intervention; each dot represents 1 tumor/mouse (n = 27 for groups without sucrose, n = 28 for groups on sucrose). Student’s t test (A); global comparison using Kruskall-Wallis test and post-hoc multiple comparisons using Dunn’s test (C and E–F). *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 7. Intermittent 3HB scheduling affects systemic expression of T cell inhibitory receptor and their ligands.

(A and B) Experimental setting for intemittent diet and combination immune checkpoint blockers (cICB) in the RET model and Kaplan Meier curves of overall survival for 12 animals/group, according to therapeutic scheme comparing continuous (Cont) versus intermittent (On/off) diet interventions. (C) Flow cytometric analyses of PD-1 and CTLA-4 expression in splenic CD8+ T lymphocytes at day 17, after diet (ketogenic diet [KD] or 3-hydroxybutyrate [3HB] Cont or On/off) and systemic cICB combinatorial regimen in RENCA bearing BALB/c mice. Refer to Supplemental Figure 5, A and B, for Tim-3, Lag-3, and 4-1BB in all splenic T cells. (D and E) Flow cytometry analysis of MFI for the membrane expression of costimulatory (CD86 in D) or inhibitory (PD-L1 in E) on macrophages (D) and other myeloid subsets from the spleens at day 17, after complete diet (3HB On/off) and systemic cICB combinatorial regimen in RENCA-bearing BALB/c mice. The results from 2–3 experiments involving 6 mice/group are depicted; each dot represents 1 spleen. (F and G) In vitro effects of increasing dosing of 3HB onto a DC cell line stimulated or not with rIFN-γ. Flow cytometry determination of MHC II/I-Ab (F) and PD-L1 expression (G) as mean fluorescence intensity (MFI) on the surface expression at 48 hours of stimulation. One representative experiment out of 4 is depicted, yielding similar conclusions (n = 12 or each groups). Student’s t test (C), global comparison using Kruskall-Wallis test with post hoc multiple comparisons using Dunn’s test (*P < 0.05) (D and E). Global comparison using ANOVA with post hoc multiple comparisons using Dunnett’s test (*P < 0.05) (F and G). *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 8. Efficacy of intermittent 3HB in circumventing primary resistance to PD-1 blockade in an orthotopic cancer model.

(A–F) Extension of the comparison between continuous (Cont) versus intermittent (On/off) feeding with KD or 3HBpo in 2 tumor models (RENCA [A], TC-1 [B–D]) treated with various therapeutic mAbs (refer to Supplemental Figure 6, A and B, for experimental setting designs). Cross-sectional RENCA tumor burden monitored by bioluminescence imaging of luciferase activity (ratio of luminescence at D8 versus D0 [D8: day of randomization]) (A) are depicted, gathering 2 independent experiments containing 6 mice/group. All experiments were performed twice. Monitoring of TC-1 progression using bioluminescence imaging (B) and percentages of TC-1_luc tumor–free animals overtime monitored by bioluminescence imaging of luciferase activity (C) and overall survival (D) estimated by Kaplan Meier curves for 2 experiments of 6 mice/group pooled together. (E and F) Tumor growth curves represented as mean ± SEM of tumor sizes for each group after s.c. rechallenge of TC-1_luc tumor–free mice (from B–D) with the MTD of TC-1_luc (E) or irrelevant MCA205 (F) (refer to Supplemental Figure 6C for experimental setting designs). Experiments were performed twice (n = 12 per groups). Global comparison for the log10-normalized data using ANOVA with post hoc multiple comparisons using Sidak’s test (*P < 0.05) (A) or dedicated software (https://kroemerlab.shinyapps.io/TumGrowth/) (C–F). *P < 0.05, **P < 0.01, ***P < 0.001.