AUTHORS
Jinglei Yu, MBBS, Yuqi Lv2, Mingming Zhang, PhD3, Shan Fu4, Jingjing Feng5, He Huang, MD, PhD6, Yongxian HU, MD7, Guoqing Wei8
AFFILIATIONS
1 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, China, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
2 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
3 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
4 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
5 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
6 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
7 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
8 The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
ABSTRACT
Background:
CD19 CAR-T therapy induces high remission rates in R/R B-ALL, but CRS remains a major toxicity. Antibiotic exposure and gut dysbiosis have been linked to CAR-T outcomes, but longitudinal microbiome dynamics, metabolic remodeling, and their relationships with CRS severity remain unclear. We integrated serial gut metagenomics, paired fecal/serum metabolomics and lipidomics, and functional models to investigate microbiota-metabolite-immune relationships in CAR-T-associated CRS.
Methods:
This prospective single-center study enrolled 66 patients with R/R B-ALL receiving CD19 CAR-T therapy. CRS was graded using the 2019 ASTCT criteria; grade ≥3 was defined as severe. Fecal samples were collected before FC lymphodepletion (FCb), after FC before infusion (FCa), before CRS onset (CRSb), at peak CRS (CRSp), and during recovery (CRSr). A total of 286 specimens were collected for shotgun metagenomic sequencing. Paired fecal and serum samples underwent untargeted metabolomic and lipidomic profiling.
Unsupervised species-level clustering defined enterotypes. Analyses included multivariable logistic regression, timepoint-specific enterotype-by-antibiotic interactions, bootstrap-corrected ROC curves, and mediation models. Antibiotic exposure was any antibiotic use within 4 weeks before infusion. Omics-wide comparisons were FDR-adjusted; targeted analyses used two-sided P values. LPC(16:0) was tested in a macrophage/CAR-T/Raji-GFP-Luc co-culture and in Raji-bearing SCID-beige mice given LPC(16:0) by oral gavage (10 mg/kg daily) and/or intraperitoneal injection (5 mg/kg every 2 days; n=5/group).
Results:
CRS occurred in 64/66 (97.0%), including grade 3 CRS in 19 (28.8%), with no grade ≥4 events. Three enterotypes were identified: ET1 (high-diversity, commensal-rich), ET2 (low-diversity, E. faecium-dominant), and ET3 (intermediate-diversity, Bacteroidetes-dominant). Enterotype trajectories differed by CRS severity (overall χ²=10.98, P=0.012), with grade 3 CRS more frequent with transition to or persistence of ET2 than transition from ET2 to ET1/ET3. At CRSp, ET2 was associated with higher odds of grade 3 CRS versus ET1 in univariable analysis (OR, 5.16; 95% CI, 1.48-20.83; P=0.0207). An ET2-by-antibiotic interaction was observed (OR, 5.28; 95% CI, 1.35-22.30; P=0.019). Among antibiotic-exposed patients, ET2 was associated with 8-fold higher odds of grade 3 CRS (95% CI, 2.74-26.20; P=0.0003). Associations of E. faecium abundance with grade 3 CRS were absent at FCb/FCa, emerged at CRSb (adjusted OR, 1.53; 95% CI, 1.04-2.40; P=0.043), and strengthened at CRSp (adjusted OR, 1.76; 95% CI, 1.12-3.11; P=0.025). Pre-CRS models including enterotype, antibiotic interaction, age, BMI, and baseline marrow blasts achieved bootstrap-corrected AUCs of 0.836 at FCb, 0.851 at FCa, and 0.877 at CRSb.
FDR-adjusted analyses linked ET2 to increased fecal glycerophospholipids and LPCs, decreased fecal ceramides, and enrichment of serum inflammatory lipids and glycerophospholipids. At CRSp, fecal LPC(16:0) correlated with E. faecium abundance (Spearman r=0.52, P=0.0018) and ET2 assignment (r=0.38, P=0.045). At FCa, CRSb, and CRSp, fecal LPC(16:0) was higher in patients with grade 3 CRS than in those with grade 0-2 CRS (all P<0.001). At CRSp, adjusted mediation analyses estimated that LPC(16:0) accounted for 21.7% of the ET2-grade 3 CRS association (P<0.01) and 26.1% and 18.4% of the associations of E. faecium with IFN-γ (P<0.01) and IL-6 (P<0.05), respectively.
In co-culture, macrophages increased IL-6, IL-1β, MCP-1, and G-CSF release; LPC(16:0) further increased these cytokines, TNF-α, and IFN-γ. In Raji-bearing mice, combined oral and intraperitoneal LPC(16:0) produced the greatest weight loss, hyperthermia, clinical scores ≥3, serum cytokine elevations (murine IL-6, MCP-1, and IL-1β; human IFN-γ), and M1-predominant myeloid remodeling. Two of 5 mice died after 48 hours versus none in other groups.
Conclusion: In CD19 CAR-T-treated R/R B-ALL, an E. faecium-dominant, low-diversity enterotype was associated with grade 3 CRS, particularly among patients with preinfusion antibiotic exposure. LPC(16:0) was elevated in grade 3 CRS, showed partial statistical mediation of the ET2-CRS and E. faecium-cytokine associations, and augmented myeloid inflammation in vitro and in vivo. These findings support a potential gut dysbiosis-lipid remodeling-myeloid inflammation axis in CAR-T-associated CRS and warrant independent clinical and mechanistic validation.
DISCLOSURES
Jinglei Yu, MBBS: No relevant conflicts of interest to declare.