Gut microbiome composition as predictor of the efficacy of adding atezolizumab to firstline FOLFOXIRI plus bevacizumab in metastatic colorectal cancer: A translational analysis of the AtezoTRIBE study

3534 Background: Gut microbiome has emerged as a biomarker of clinical benefit to immune-checkpoint inhibitors (ICI) but no data are available in metastatic colorectal cancer (mCRC). The AtezoTRIBE study demonstrated that the addition of atezolizumab (atezo) to FOLFOXIRI plus bevacizumab (bev) prolongs progression-free survival (PFS), but this benefit is limited for patients with proficient mismatch repair (pMMR) tumors. Here, we aimed at investigating the potential predictive role of microbiome in identifying mCRC patients able to achieve benefit from ICI. Methods: AtezoTRIBE was a phase II trial in which 218 mCRC patients, unselected for MMR status, were randomized 1:2 to receive first-line FOLFOXIRI/bev (arm A) or FOLFOXIRI/bev/atezo (arm B). Stools were prospectively collected. Metagenomic (MG) data from whole genome sequencing (WGS) at level of species genome bins (SGBs) were analysed by linear models corrected for clinical and tumor-related parameters and fold-ratios. We defined as responders (R) those patients who experienced a PFS ≥ 12 months. Results: Stool samples were collected at baseline for 171 (78%) patients (55 in arm A and 116 in arm B) but only 163 were available for MG. Patients with deficient MMR (dMMR) tumors (N = 10) showed significantly lower MG diversity than pMMR ones (N = 148) harboring oral bacteria and pathobionts whose intrinsic immunogenicity has not been demonstrated. Regarding pMMR mCRC patients, baseline microbiome composition was not significantly different according to the treatment arm. The microbiome diversity was not significantly different between R and not-R in both arms, but specific immunogenic SGBs (Lachnospiraceae family members) were over-represented in R treated in arm B. Veillonellaceae and pathobionts were associated with poor prognosis and/or differential benefit from the addition of atezo. Fusobacterium nucleatum was associated with a poor prognosis, also in arm B. Conclusions: This is the largest prospective analysis showing that SGBs may be useful as a biomarker of potential benefit or detrimental effect from atezo in pMMR mCRC patients. Our results prompt the design of microbiota-centered diagnostic tests to identify pMMR mCRC patients more likely to benefit from ICI-based therapeutic strategies. Clinical trial information: NCT03721653 .