2570 Background: A number of studies have demonstrated that the gut microbiome of responders to immune checkpoint inhibitors (ICI) is compositionally different compared to that of non-responders. However, differences in study design, patient cohorts and bioinformatic analyses make it challenging to identify bacterial species consistently associated with response to ICI across different cohorts and cancer types. Methods: We leveraged the statistical power of mega- and meta-analyses to identify bacterial species consistently associated with response to ICI using data from three published fecal metagenomic studies (Gopalakrishnan et al., Science 2018; Matson et al., Science 2018; Routy et al., Science 2018). Metagenomic data was uniformly processed and analyzed using Metaphlan v2.0. We conducted a two-part modelling approach of bacterial species present in at least 20% of samples to account for both prevalence and relative abundance differences between responders/non-responders. Results: A total of 190 patients (n = 103 responders; n = 87 non-responders) were included from the three studies. Data from Routy et al., was analyzed as subsets based on tumor type for a total of 4 analyzed cohorts. We identified five species including Bacteroides thetaiotaomicron, Clostridium bolteae, Holdemania filiformis, Clostridiaceae bacterium JC118 and Escherichia coli that were concordantly significantly different between responders and non-responders using both meta- and mega-analyses. B. thetaiotaomicron and Clostridium bolteae relative abundance (RA) were independently predictive of non-response to immunotherapy when data sets were combined and analyzed using mega-analyses (AUC 0.59 95% CI 0.51-0.68 and AUC 0.61 95% CI 0.52-0.69, respectively). Conclusions: Despite inter-cohort heterogeneity in tumor type, treatment regimens, and sequencing modalities, meta- and mega analysis of published metagenomic studies identified generalizable bacterial species associated with ICI response or lack thereof. B. thetaiotaomicron and C. bolteae were predictors of non-response to ICI suggesting the clinical potential of narrow spectrum anti-biotics targeting non-response associated bacterial species to improve outcomes in ICI recipients.
Clostridium bolteae
