Fluctuations in Gut Microbiome Composition During Immune Checkpoint Inhibitor Therapy

Abstract

Background: Immune checkpoint inhibitors (ICIs) such as programmed cell death protein-1 (PD-1) inhibitors or PD-1 ligand-1 (PD-L1) inhibitors have led to remarkable improvement in outcomes of non-small cell lung cancer (NSCLC). Unfortunately, the significant benefits of ICI therapy are frequently limited by resistance to treatment and adverse effects, and the predictive value of pre-treatment tumor tissue PD-L1 expression is limited. Development of less invasive biomarkers that could identify responders and non-responders in early on-treatment could markedly improve the treatment regimen. Accumulating evidence suggests that baseline gut microbiota profile is associated with response to PD-1/PD-L1 blockade therapy. However, change in the gut microbiome composition during PD-1/PD-L1 blockade therapy and its relation to response remain unclear.

Methods: Here, we analyzed pre- and on-treatment fecal samples from five NSCLC patients receiving anti-PD-1 immunotherapy, alone or in tandem with chemotherapy, and performed 16S rRNA sequencing.

Results: The overall alpha diversity of the baseline gut microbiome was similar between three responders and two non-responders. While the gut microbiome composition remained stable overall during treatment (R2 = 0.145), responders showed significant changes in microbiome diversity between pre- and on-treatment samples during anti-PD-1 therapy compared to non-responders (P = 0.0274). Within the diverse microbiota, responders showed decreases in the abundance of genera Odoribacter, Gordonibacter, Candidatus Stoquefichus, Escherichia-Shigella, and Collinsella, and increase in abundance of Clostridium sensu stricto 1. In contrast, non-responders demonstrated on-treatment increases in genera Prevotella, Porphyromonas, Streptococcus, and Escherichia-Shigella, and decrease in abundance of Akkermansia.

Conclusions: This pilot study identified a substantial change in gut microbiome diversity between pre- and on-treatment samples in NSCLC patients responding to anti-PD-1 therapy compared to non-responders. Our findings highlight the potential utility of gut microbiota dynamics as a noninvasive biomarker to predict response to PD-1/PD-L1 blockade therapy for a wide variety of malignancies, which sets a path for future investigation in larger prospective studies.

Keywords: Biomarker; Gut microbiome; Immunotherapy; Non-small cell lung cancer; PD-1.

Figure 1

Pre- and on-treatment imaging study and circulating biomarker performance in non-small cell lung cancer (NSCLC) patients treated with anti-PD-1 therapy. (a) Contrast-enhanced cross sectional imaging obtained prior to and during treatment in three patients. Expression of CX3CR1 in peripheral blood CD8⁺ T cells (b) and % change of CX3CR1⁺ in CD8⁺ T cells from baseline (CX3CR1 score) (c) at different time points as indicated. PD-1: programmed cell death protein-1; CX3CR1: CX3C chemokine receptor 1.

Figure 2

Fluctuations in gut microbiome in non-small cell lung cancer (NSCLC) patients during anti-PD-1 therapy. (a) Alpha diversities of baseline gut microbiome in responders (R) and non-responders (NR). First column: observed diversity reflects the total number of unique organisms. Second column: Chao1 diversity reflects total richness, weighted towards rare species. Third column: Shannon index reflects both richness and evenness of each sample. Fourth column: Simpson index reflects richness, weighted toward common species. (b) Beta-diversity using Bray-Curtis dissimilarity coupled with multidimensional scaling depicting pre- and post-treatment with anti-PD-1 immunotherapy. The first three pairwise principal components were displayed. P value was estimated from PERMANOVA using Bray-Curtis dissimilarity implemented by vegan R package (v2.5.6). Pt: patient; PD-1: programmed cell death protein-1; PERMANOVA permutational multivariate analysis of variance.

Figure 3

Substantial change in microbial profiles in NSCLC patients responding to anti-PD-1 therapy. (a, b) Significantly abundant genera found in pre- and on-treatment microbial composition between responders (a) and non-responders (b). Heatmap demonstrating significant differences in pre- and post-treatment microbial composition for each subject (n = 5); each pair of blue and yellow columns represents one subject. A variance-stabilization transformation (implemented by DESeq2) was used for the taxa abundance values. Darker shades represent higher differential abundance. (c) Genus composition pre- and on-treatment for all subjects. Pt: patient; NSCLC: non-small cell lung cancer; PD-1: programmed cell death protein-1.