AUTHORS
Kuo-Kai Chin, M.D., Anqi Dai, MS2, Kenton Wu3, Zoe Katsamakis4, Mirae Baichoo, PhD5, Emily Walsh, PhD6, Mary-Jane Lombardo, PhD7, Elizabeth Halvorsen, PhD8, Karen Zhao, M.D.9, Lindsay Rutte10, Sean Sykes11, Eytan M Stein I, MD12, Aaron D Goldberg, MD, PhD13, Jonathan Peled, MD/PhD14
AFFILIATIONS
1 Memorial Sloan Kettering Cancer Center, New York, New York, United States
2 Memorial Sloan Kettering Cancer Center, New York, New York, United States
3 Memorial Sloan Kettering Cancer Center, New York, New York, United States
4 Memorial Sloan Kettering Cancer Center, New York, New York, United States
5 Memorial Sloan Kettering Cancer Center, New York, New York, United States
6 Seres Therapeutics, Cambridge, Massachusetts, United States
7 Seres Therapeutics, Cambridge, Massachusetts, United States
8 Seres Therapeutics, Cambridge, Massachusetts, United States
9 Seres Therapeutics, Cambridge, Massachusetts, United States
10 Seres Therapeutics, Cambridge, Massachusetts, United States
11 Seres Therapeutics, Cambridge, Massachusetts, United States
12 Memorial Sloan Kettering Cancer Center, New York, New York, United States
13 Memorial Sloan Kettering Cancer Center, New York, New York, United States
14 Memorial Sloan Kettering Cancer Center, New York, New York, United States
ABSTRACT
Background: Gut-microbiome disruption predicts adverse outcomes in patients with AML undergoing alloHCT, including bloodstream infections and increased mortality. Prior studies reported associations between microbial community domination and bloodstream infections during high-intensity frontline AML chemotherapy. The clinical significance of microbiome dynamics during frontline low-intensity therapy, which has seen increasing use with regimens such as azacitidine and venetoclax, remains largely unexplored.
Aim: To compare gut microbiome trajectories during high-intensity and low-intensity AML therapy and evaluate associations with infectious outcomes.
Methods: We prospectively collected fecal samples from patients with newly diagnosed AML that received high or low-intensity therapy at our center (2023-2025). Fecal microbiome composition was profiled by whole metagenomic shotgun sequencing (mean 12M reads/sample).
Results: Sixty-three patients contributed 164 samples (high-intensity n=32; low-intensity n=31). Mean age was 65 years; low-intensity patients were older (73.0 vs 54.8 years, p<0.001). Response rates did not differ significantly between groups (CR/CRi: high-intensity n=23, 72%; low-intensity n=18, 58%; p=0.3).
Recipients of high-intensity had more fevers, (n=32, 100% vs n=15, 48%), febrile neutropenia (FN; n=26, 81% vs n=12, 39%), and documented infections (composite of bloodstream infection and stool PCR positivity, n=20, 63% vs n=6, 19%) (all p<0.001). There were no significant differences in neutropenia incidence (n=32, 100% vs n=30, 97%, p=0.5), and neutropenic days (median 16.5 vs 21, p=0.5). High-intensity patients were hospitalized for a median of 35 days in the first 50 days of treatment compared to 20 days for low-intensity patients (p<0.001).
At baseline (D-14 to D3 of induction), α-diversity did not significantly differ between high-intensity and low-intensity groups (inverse Simpson index, p=0.2). Burden of neutropenia was most pronounced between days 10 and 21 in both groups and was used to define treatment nadir. High-intensity therapy was associated with a 63.5% decline in α-diversity from baseline (p=0.0031), whereas low-intensity showed minimal change (+2.4%) from baseline in GEE models adjusting for repeated measures and rolling 14-day antibiotic exposure. For high-intensity patients, significant declines in α-diversity from baseline were also seen in the late window (D22-42: -73.4%, p<0.001) and across the extended window (D10-42: -65.0%, p<0.001), indicating persistent microbiome disruption beyond nadir. For low-intensity patients, α-diversity change from baseline to the late and extended windows were small (D22-42: -3.5%, D10-42: -3.8%). Phase-by-intensity interaction terms were significant at nadir (p=0.045), late (p=0.0018), and extended (p=0.014), indicating significantly smaller phase-related declines with low-intensity.
Global β-diversity did not differ between high-intensity and low-intensity groups by at baseline (Bray-Curtis dissimilarity, PERMANOVA p=0.79) and β-dispersion were not significantly different (p=0.059). At nadir, β diversity differed by treatment intensity (PERMANOVA p=0.032, R2=0.052) with greater displacement in high-intensity patients (median distance to centroid 0.641 vs 0.540 in low-intensity, p=0.004).
Microbiotae domination events (defined as any taxon found at relative abundance ≥30%) were numerically more frequent in high-intensity patients (n=14, 45% vs n=7, 25%, OR 0.41, p=0.17). The most common dominant taxa were the genera Enterococcus (n=8 sample-weeks, 5 high-intensity and 3 low-intensity) and Streptococcus (n=5, all high-intensity), as well as the phylum Bacteroidetes (n=5, all high-intensity). Domination was associated with increased odds of FN (OR 3.6, p=0.036) and FN days (median difference 5 days, p=0.016), though these associations were not significant after adjusting for treatment intensity (OR 2.9, p=0.116; 1.3 days, p=0.25).
Conclusion: During frontline AML therapy, low-intensity regimens are associated with preserved α-diversity and less β-diversity displacement at nadir compared with high-intensity regimens, but microbiome disruption persists, including microbial domination by potential bloodstream infection pathogens. Microbial domination was associated with febrile neutropenia, supporting investigation of microbiome-targeted strategies to mitigate toxicity during AML therapy.
DISCLOSURES
Kuo-Kai Chin, M.D.: Seres Therapeutics - Research Funding; Syndax Pharmaceuticals - Research Funding; Servier Laboratories - Consultancy (Includes expert testimony); Syndax Pharmaceuticals - Consultancy (Includes expert testimony)