Randomized Placebo-Controlled, Biomarker-Stratified Phase Ib Microbiome Modulation in Melanoma: Impact of Antibiotic Preconditioning on Microbiome and Immunity

Abstract

Gut-microbiota modulation shows promise in improving immune-checkpoint blockade (ICB) response; however, precision biomarker-driven, placebo-controlled trials are lacking. We performed a multicenter, randomized placebo-controlled, biomarker-stratified phase I trial in patients with ICB-naïve metastatic melanoma using SER-401, an orally delivered Firmicutesenriched spore formulation. Fecal microbiota signatures were characterized at baseline; patients were stratified by high versus low Ruminococcaceae abundance prior to randomization to the SER-401 arm (oral vancomycin-preconditioning/SER-401 alone/nivolumab + SER-401), versus the placebo arm [placebo antibiotic/placebo microbiome modulation (PMM)/nivolumab + PMM (NCT03817125)]. Analysis of 14 accrued patients demonstrated that treatment with SER-401 + nivolumab was safe, with an overall response rate of 25% in the SER-401 arm and 67% in the placebo arm (though the study was underpowered related to poor accrual during the COVID-19 pandemic). Translational analyses demonstrated that vancomycin preconditioning was associated with the disruption of the gut microbiota and impaired immunity, with incomplete recovery at ICB administration (particularly in patients with high baseline Ruminococcaceae). These results have important implications for future microbiome modulation trials. Significance: This first-of-its-kind, placebo-controlled, randomized biomarker-driven microbiome modulation trial demonstrated that vancomycin + SER-401 and anti-PD-1 are safe in melanoma patients. Although limited by poor accrual during the pandemic, important insights were gained via translational analyses, suggesting that antibiotic preconditioning and interventional drug dosing regimens should be carefully considered when designing such trials.

Figure 1.

Efficacy of vancomycin preconditioning followed by SER-401 and nivolumab versus placebo and nivolumab in advanced melanoma patients. A, Schema shows the performed trial design (NCT03817125) incorporated placebo-controlled arms of microbiome intervention with the vanco + SER-401 arm. All arms were designed to receive preconditioning with either vancomycin or placebo antibiotic, followed by a SER-401 lead-in (vs. placebo), then maintenance microbiome intervention (or placebo) concurrent with nivolumab treatment for all arms. B, Best change after trial period; asterisk indicates confirmed PR/CR by RECIST 1.1. C, Swimmer plot showing time on treatment demonstrates the response and survival characteristics by the patient. Rightward arrow denotes continued response at the time of study conclusion. D and E, OS (D) and PFS (E) are displayed using Kaplan–Meier curves between the two arms. PR: partial response; CR: complete response; Abx: Antibiotics; OS: overall survival; PFS: progression-free survival. P value denotes the log-rank test.

Figure 2.

Gut microbiome assessment after treatment with vancomycin preconditioning, followed by SER-401 and nivolumab (vs. placebo control). D-7 time point is after vancomycin but before SER-401 or nivolumab, whereas C1D1 is after SER-401 lead-in but before nivolumab initiation. A, Ruminococcaceae family relative abundance, displayed by percent change compared with baseline abundance, shows a decrease in abundance at D-7 with subsequent recovery. Wilcoxon signed-rank test performed against baseline in each arm. B, Stacked bar plots of taxa (by family) present in the gut microbiome of individual patients over time. C, Relative abundance of the Ruminococcaceae family over time split by prospective analysis of baseline Ruminococcaceae abundance. Colors indicate the best objective response by RECIST criteria. D and E, Imputed pathway analysis of whole metagenomic sequencing of stool sorted by previously published MetaCyc pathways associated with either beneficial or adverse impact in the setting of immunotherapy (Simpson et al.; ref. 16). In the SER-401 arm at D-7 (after preconditioning but before microbiome intervention), there were lower levels of beneficial pathways and higher levels of adverse pathways; these differences were not present at baseline and did not persist at C1D1 (after initial microbiome intervention but before nivolumab). Wilcoxon rank sum test performed between arms. Abx: antibiotic treatment (vancomycin in SER-401 arm; placebo antibiotic in the placebo arm).

Figure 3.

Changes in intratumoral and systemic inflammation after vancomycin preconditioning followed by SER-401 and nivolumab (vs. placebo control). A, There was a detectable increase in tumor-infiltrating CD8 cells at C2D1 compared with baseline, with an increased percentage in a pooled analysis (P = 0.06, Mann–Whitney test). B, In patients with available matched baseline and on-treatment samples (n = 3), there was an increase of intratumoral CD8 T-cell infiltration on treatment. Representative IHC images are shown, 1 per arm. C, Comparison of circulating immune cell populations representing change from baseline to D-7 in the SER-401 arm versus the placebo arm by dimensional flow cytometry. Left, Red horizontal line on the volcano plot indicates statistical significance (P < 0.01). Right, Three immune cell populations (1: PD-1⁺TIGIT⁺ Treg; 2: PD-1⁺ Treg; 3: FOXP3⁻ PD-1⁺ effector cells) from volcano plot over time as mean fold change over baseline. P values represent statistical significance as compared with the baseline by the Mann–Whitney test. D, Comparison of circulating proteins between SER-401 and placebo arms from baseline to D-7 (left) and baseline to C1D1 (right). The red horizontal line on the volcano plot indicates statistical significance (P < 0.05).