Influence of the Gut Microbiome on Clinical Outcomes in the CPIT-001 Trial, a Phase Ib-IIA Study of Combined Checkpoint Inhibition with Nivolumab and Ipilimumab after Autologous Hematopoietic Stem Cell Transplantation in Patients at High-Risk for Post-Transplant Recurrence

Background: Autologous stem cell transplantation (ASCT) remains a standard of care for patients with high-risk and recurrent diffuse large B cell lymphoma (DCLBL), T cell lymphoma (TCL) and multiple myeloma (MM). The discovery that differences in the composition of the gut microbiome contribute to the heterogeneity of response to chemotherapies or immunotherapies has prompted us to investigate the influence of the gut microbiome on progression-free survival (PFS) and overall survival in these patient populations treated with combined nivolumab (Nivo) and ipilimumab (Ipi) in the post-ASCT consolidation setting.

Methods: Longitudinal analysis of the gut microbiome was performed on stool samples collected from 26 patients (6 de novo DLBCL, 5 recurrent DLBCL, 1 de novo high-risk T cell lymphoma 2 recurrent TCL, 7 transplant-naïve high-risk MM, 4 recurrent MM) treated with Nivo + Ipi, 14-28 days post-ASCT. Stool samples were collected from patients prior to conditioning and ASCT (baseline) and at engraftment (within 72 hrs of absolute neutrophil count ≥ 500/mcL). Stool samples were then collected serially on Weeks 1, 4, 7, 12, 18 and 26 of treatment, 9, 12 and 18 months post-ASCT and at relapse of disease. The composition of the gut microbiome was identified using 16S rRNA sequencing of the V4 amplicon and bioinformatic analysis of α- and β-diversity metrics was performed using the Second Genome Microbiome Discovery Platform. Patients were classified as early relapsers, (n=5), PFS <6 months (Short-PFS) versus patients (n=21), with PFS, ≥ 6 months (Long-PFS).

Results: An average of 267 operational taxonomic units (OTUs) was identified per patient sample with a Shannon diversity index of 2.85. The top phyla in DLBCL, TCL and MM included Firmicutes, Bacteroides, Actinobacteria, Proteobacteria, Verrucomicrobia and Euryarchaeota and most samples were dominated by Firmicutes. The top families included Lachnospiraceae, Ruminococcaceae, Bacteroidaceae, Enterococcaceae, Erysipelotrichaceae, Bifidobacteriaceae, Enterobacteriaceae and Streptococcaceae.

The Shannon diversity index was reduced at engraftment (2.26) compared to at time of conditioning (3.15) and full recovery of diversity was attained at week 18 of treatment (3.15). At conditioning, Firmicutes, followed by Actinobacteria, and Bacteroides dominated the microbiomes in all indications and at the family level, Lachnospiraceae followed by Ruminococcaceae were major components of the microbiome. The microbiome, at the phylum and family levels, also varied considerably between conditioning and at relapse. While significant shifts in microbiome β-diversity appeared to be associated with indication and timepoint (PERMANOVA, P<0.001 and P<0.01 respectively), the samples did not separate in the weighted ordination, unweighted ordination and hierarchical clustering analyses by indication or timepoint.

A significant change of the relative abundance of the Bacteroidetes phylum at conditioning, engraftment, prior to consolidation and on-treatment (Week 7) was observed in the Short-PFS group (Friedman Test, 0.0638, P<0.02). A shift from high to low abundance of Bacteroidetes was observed in the Short-PFS group at conditioning and during engraftment prior to consolidation therapy (Kruskal-Wallis test, P<0.025 and P<0.01 respectively). A repeated measures Poisson regression model assessed patient's relative risk (RR) of response to treatment based on the relative abundance of specific gut microbiota. Relative abundances were categorized by tertile into low, medium, and high abundance groups. The regression model estimated the RR of response for medium and high abundance compared to low abundance groups. Compared to patients with a low abundance of Verrucomicrobia, those with medium and high abundance showed a 2% (RR= 1.022; p=0.078) and 4% (RR=1.035; p=0.078) trend toward predictive capability of patients falling into the Long-PFS group respectively.

Conclusions: Our interim taxonomic analyses of 26 patients suggest that the taxonomic identity of the microbiomes were largely similar across indications and timepoints, but the relative abundances of the phylotypes differed between samples. While our results suggest that Bacteroidetes and Verrucomicrobia at conditioning and engraftment may correlate with the likelihood of Short-PFS versus Long-PFS, our results need to be validated in ongoing cohort expansion studies.