Microbial metabolism and CAR-T cell fitness: a metabolome-centered translational framework
This narrative review proposes that microbiota-derived metabolites — SCFAs, indoles, and inosine — modulate CAR-T cell fitness, with associative clinical evidence indicating that pre-infusion broad-spectrum antibiotics favor inferior outcomes (HR 2.54 for OS, 95% CI 1.62–3.97).
| Population | Adults with relapsed/refractory hematologic malignancies (LBCL, B-ALL, multiple myeloma) receiving anti-CD19 or anti-BCMA CAR-T therapy |
|---|---|
| Intervention | Pre-infusion exposure to broad-spectrum antibiotics (piperacillin-tazobactam, carbapenems, broad-spectrum cephalosporins) and/or modulation of microbial metabolites (SCFAs, indoles, inosine) |
| Comparator | No high-risk antibiotic exposure or preserved microbiota |
| Outcome | Overall survival (OS) with P-I-M exposure; Progression-free survival (PFS) with antibiotic exposure; ICANS incidence with antibiotic exposure; CAR-T memory phenotype and exhaustion with butyrate (experimental); PFS and CAR-T cytotoxicity with inosine |
Summary of findings
| Outcome | Effect | 95% CI | Certainty | Clinical relevance | Notes |
|---|---|---|---|---|---|
| Overall survival (OS) with P-I-M exposure | HR 2.54, 95% CI 1.62-3.97 | — | Low | — | 1 studies |
| Progression-free survival (PFS) with antibiotic exposure | HR 1.50 (Prasad); HR 2.19 (Stein-Thoeringer); 95% CI not fully reported across both | — | Low | — | 2 studies |
| ICANS incidence with antibiotic exposure | direction reported without full HR/CI in this review | — | Low | — | 1 studies |
| CAR-T memory phenotype and exhaustion with butyrate (experimental) | directional only, in the quantitative effect size reported | — | Very low | — | 1 studies |
| PFS and CAR-T cytotoxicity with inosine | exploratory association; in vitro A2A-dependent cytotoxicity increase, in the HR/CI reported | — | Very low | — | 1 studies |
Context
Heterogeneous CAR-T responses in hematologic malignancies drive the search for modifiable host determinants. The gut microbiota has emerged as a potentially modulable extratumoral factor, with observational studies linking broad-spectrum antibiotic exposure to shorter survival and greater neurotoxicity. Most available studies remain associative and taxonomy-centered, without establishing metabolic causality.
What the study showed
In Smith et al. (n=228), P-I-M exposure in the 4 weeks before infusion was independently associated with shorter OS (HR 2.54; 95% CI 1.62–3.97; P<0.001) and higher ICANS incidence. Prasad et al. (n=422) found antibiotic exposure associated with worse PFS in the LBCL subgroup (HR 1.50; P=0.013), but exposed patients had greater tumor burden and systemic inflammation — substantial confounding by indication. Stein-Thoeringer et al. (n=172) reported HR 2.19 for PFS with high-risk antibiotics, with no independent OS association after multivariable adjustment. Direct CAR-T evidence for butyrate showed promotion of memory phenotypes and reduced exhaustion markers in experimental models; inosine was exploratorily associated with longer PFS and increased CAR-T cytotoxicity in vitro in an A2A-dependent manner.
How it was done
Integrative narrative review without a registered PRISMA protocol, synthesizing retrospective observational clinical studies (cohorts of 140–422 patients), translational studies, and preclinical evidence. No formal meta-analysis or data pooling. Included clinical studies used pre-infusion exposure windows of 3–6 weeks with variable follow-up.
Effect magnitude
The most robust effect reported is HR 2.54 (95% CI 1.62–3.97) for OS with P-I-M exposure in multivariable analysis (Smith et al., n=228); the remaining HRs (1.50 and 2.19) derive from cohorts with incompletely controlled confounding by indication.
Risk of bias
Narrative design without formal risk-of-bias assessment (no RoB 2 or ROBINS-I applied); all clinical studies are retrospective and observational, with substantial confounding by indication. Extrapolation from conventional T-cell mechanisms to CAR-T cells is explicitly acknowledged as speculative by the authors. No metagenomic sequencing or integrated metabolomic data in included clinical studies.
What this study does NOT prove
This study does not prove causality between specific microbiota or metabolites and CAR-T outcomes in humans; it does not demonstrate that microbial modulation improves OS or PFS. Antibiotic-outcome associations at least partially reflect confounding by indication and cannot be attributed exclusively to induced dysbiosis.
In clinical practice
Clinicians should record and stratify pre-infusion antibiotic exposure — especially P-I-M and carbapenems — as a risk variable in CAR-T outcome analyses. Current evidence does not support modification of antimicrobial protocols based on this framework; clinically indicated antibiotics must not be withheld. Monitoring of microbial diversity and fecal metabolome may be considered within prospective clinical research settings.
Limitations
Narrative design without formal risk-of-bias assessment (no RoB 2 or ROBINS-I applied); all clinical studies are retrospective and observational, with substantial confounding by indication. Extrapolation from conventional T-cell mechanisms to CAR-T cells is explicitly acknowledged as speculative by the authors. No metagenomic sequencing or integrated metabolomic data in included clinical studies.
What is still missing
Prospective clinical trials testing microbiota-directed interventions (probiotics, fecal microbiota transplantation, SCFA supplementation) integrated into CAR-T manufacturing, with functional endpoints of cell persistence and serum metabolomics as primary outcomes.
Technical appendix
Version history
- 1.0 · 2026-10-15 — Auto-generated under Evidence Standard v1.0