Prospective correlation between the patient microbiome with response to and development of immune-mediated adverse effects to immunotherapy in lung cancer

Abstract

Background: Though the gut microbiome has been associated with efficacy of immunotherapy (ICI) in certain cancers, similar findings have not been identified for microbiomes from other body sites and their correlation to treatment response and immune related adverse events (irAEs) in lung cancer (LC) patients receiving ICIs.

Methods: We designed a prospective cohort study conducted from 2018 to 2020 at a single-center academic institution to assess for correlations between the microbiome in various body sites with treatment response and development of irAEs in LC patients treated with ICIs. Patients must have had measurable disease, ECOG 0-2, and good organ function to be included. Data was collected for analysis from January 2019 to October 2020. Patients with histopathologically confirmed, advanced/metastatic LC planned to undergo immunotherapy-based treatment were enrolled between September 2018 and June 2019. Nasal, buccal and gut microbiome samples were obtained prior to initiation of immunotherapy +/- chemotherapy, at development of adverse events (irAEs), and at improvement of irAEs to grade 1 or less.

Results: Thirty-seven patients were enrolled, and 34 patients were evaluable for this report. 32 healthy controls (HC) from the same geographic region were included to compare baseline gut microbiota. Compared to HC, LC gut microbiota exhibited significantly lower α-diversity. The gut microbiome of patients who did not suffer irAEs were found to have relative enrichment of Bifidobacterium (p = 0.001) and Desulfovibrio (p = 0.0002). Responders to combined chemoimmunotherapy exhibited increased Clostridiales (p = 0.018) but reduced Rikenellaceae (p = 0.016). In responders to chemoimmunotherapy we also observed enrichment of Finegoldia in nasal microbiome, and increased Megasphaera but reduced Actinobacillus in buccal samples. Longitudinal samples exhibited a trend of α-diversity and certain microbial changes during the development and resolution of irAEs.

Conclusions: This pilot study identifies significant differences in the gut microbiome between HC and LC patients, and their correlation to treatment response and irAEs in LC. In addition, it suggests potential predictive utility in nasal and buccal microbiomes, warranting further validation with a larger cohort and mechanistic dissection using preclinical models.

Trial registration: ClinicalTrials.gov, NCT03688347 . Retrospectively registered 09/28/2018.

Keywords: Adverse effects; Immune checkpoint; Immunotherapy; Lung cancer; Microbiome; Response; Toxicity.

Conflict of interest statement

The authors of this manuscript declare no conflicts of interest.

© 2021. The Author(s).

Figures

Fig. 1

Description of patient cohorts and study schema. (a) Study schema showing the collection of microbiome samples from three separate body sites. Samples undergo 16S rRNA amplicon sequencing followed by taxonomic profiling. The resulting data is correlated to clinical outcomes such as response to ICI therapy or development of AEs. (b) An abbreviated demographics chart summarizing notable disease and patient characteristics of LC contributors. For granular individual characteristics, please refer to Supplemental Table 1. (c) Breakdown of patients belonging to each study cohort. ICI = immune checkpoint inhibitor. Checkpoint inhibitor status (number of patients enrolled – number of fecal samples that were unable to be analyzed or not submitted – number of nasal/buccal samples that were unable to be analyzed or not submitted

Fig. 2

Flowchart of patients enrolled in the study. Breakdown of number of patients enrolled, were deemed ineligible for the study, as well as number of samples provided for each stage of analysis

Fig. 3

Baseline microbiome composition. Bar and heatmap plots comparing baseline gut, nasal and buccal microbiomes in LC compared to HC. (a)Left: Bar graph showing relative ratios of phyla constitution in LC and HC samples. Right: Box plot showing a statistically significant decrease in α-diversity when comparing LC to HC patients (p = 9.36 × 10− 04). (b)Left: 2-dimensional PLS-DA graph identifying notable differences in genus expression when comparing LC vs HC samples. Right: Heatmap showing genus level expression in LC patients (upper half) compared with HC (lower half). There is a notable difference at the genus level. (c) Principal coordinate analysis (PCoA) comparing the beta-diversity of buccal, nasal and gut microbiome in LC patients

Fig. 4

Response to immunotherapy. Microbiome changes notable in responders to ICI. Normalized data is presented in log-adjusted relative abundances. Left panels show PLS-DA graphs from nasal, buccal and gut sites all showing a microbiome separation when comparing ICI responders vs nonresponders. (a) Taxonomic profiling of nasal samples identified notable enrichment in Finegoldia, of phylum Firmicutes, in responders to ICI (p = 0.0005). (b) Buccal analysis of ICI responders show enrichment in Megasphaera of phylum Firmicutes (p = 8.6 × 10− 03) and decrease in Actinobacillus of phylum Proteobacteria (p = 9.7 × 10− 03). (c) In the fecal samples of ICI responders, Clostridiales was enriched (phylum Firmicutes, p = 0.017875) and Rikenellaceae decreased (phylum Bacteroidetes, p = 0.016013)

Fig. 5

Toxicity analysis. (a) PLS-DA analysis showing significant microbiome differences between LC patients who experienced toxicities and those who did not using different grouping methods of irAE severities, e.g. grade 0 vs. grade 1 + 2 + 3 + 4; grade 0 + 1 vs. grade 2 + 3 + 4; and grade 0 vs. 1 + 2 vs. 3 + 4. (b) Normalized abundances of Bifidobacterium (phylum Actinobacteria) and Desulfovibrio (phylum Proteobacteria) showed enrichment in both bacteria in patients who developed less irAEs. All differences were statistically significant irrespective of categorization of AE severity

Fig. 6

Longitudinal changes in microbiome with development and resolution of toxicities. (a) Analysis identified five patients who had submitted multiple samples during development and resolution of irAE. JZLC-24 and JZLC-6 did not have stool samples available for analysis but did submit all three sets of nasal and buccal samples. On the x-axis, sample collections are listed: V1, prior to initiation of immunotherapy; V2; at onset of toxicity, and V3, at resolution of irAE. The y-axis denotes the logarithmic (base 10) relative change in α -diversity compared to the previous visit, trended by the line plots, overlaid. Across nearly all sets of microbiome samples, a drop in microbiome α-diversity is observed at onset of irAE. At resolution of irAE to grade 1 severity or better, a third set of samples exhibit a trend toward either slowed rate of decrease in α-diversity or a reversal altogether toward baseline. (b) A consistent trend in increase of Staphylococcus at onset of toxicity and decrease with resolution of toxicity was also observed in the nasal samples (left), with a similar trend in buccal samples (right). (c)Megasphaera, a bacterium belonging to Firmicutes, was previously identified as being enriched in responders to immunotherapy. Here, it is also shown to decrease in buccal samples of patients who developed irAEs, then increasing in abundance with resolution of toxicity