INTRODUCTION The therapeutic capacity of CAR-T cells is shaped by context-dependent factors, including their ability to traffic, expand, and exert effector functions—processes influenced by cellular metabolism and the broader metabolic environment. As such, adipose tissue quantity and distribution have been linked to survival outcomes and cytokine release syndrome (CRS), yet the mediators of these associations remain elusive. Here, we investigated how adipose tissue–associated serum metabolites affect CAR-T therapy outcomes through integrated body composition and metabolomic analyses.

METHODS In a retrospective discovery cohort of 54 lymphoma patients (50 LBCL, 4 MCL) receiving CD19-directed CAR-T therapy, we quantified visceral (VAT) and subcutaneous adipose tissue (SAT) using CT imaging and profiled 1,039 fasting serum metabolites at days 0, 3–5, and 14 using targeted mass spectrometry (158 longitudinal samples). Key metabolites were tested in a validation cohort of 17 patients (10 LBCL, 3 MCL, 4 MM) treated with CD19- or BCMA-directed CAR-T. To explore mechanisms, we performed single-cell RNA sequencing (scRNA-seq) on murine bone marrow immune cells co-cultured with either adipocytes or fibroblasts.

RESULTS Using adipose tissue-informed feature selection, we identified 76 baseline metabolites at baseline significantly associated with progression-free survival (PFS) in multivariable Cox models adjusted for sex, costimulatory domain, and tumor burden. Five lipid classes emerged: diacylglycerols (HR 1.66, 95%CI 1.15–2.38, p=0.006) and phosphatidylethanolamines (PEAs; HR 1.37, 95%CI 1.22–1.53, p<0.001) were linked to inferior PFS, while acylcarnitines (ACs; HR 0.81, 95%CI 0.72–0.90, p<0.001), sphingomyelins (HR 0.74, 95%CI 0.61–0.90, p=0.002), and plasmalogens (HR 0.78, 95%CI 0.62–0.99, p=0.04) were associated with improved outcomes. Beneficial lipids were primarily VAT-derived, whereas PEAs were SAT-driven and negatively associated with VAT.

At day 3–5, PEAs (OR 2.03, 95%CI 1.44–2.86, p<0.001) and ACs (OR 1.72, 95%CI 1.24–2.41, p=0.001) were positively associated with CRS grade ≥2, while lysophosphatidylcholines (LPCs; OR 0.38, 95%CI 0.29–0.47, p<0.001) were inversely associated. Stratification by median levels of representative metabolites confirmed these trends: elevated levels of AC-(10:0) (p=0.003) and PEA-(38:06) (p=0.01) were associated with a higher incidence, whereas LPC-(20:1) (p<0.0001) was linked to a lower incidence. Notably, ACs remained VAT-associated throughout treatment, while PEAs shifted from SAT- to VAT-association at the time of CRS.

Longitudinal analysis revealed that short-chain ACs rose early and remained elevated, especially in VAT-high patients. Long-chain PEAs showed transient spikes at CRS onset only in VAT-high individuals, while LPCs rose more steeply in VAT-low patients.

We confirmed key associations between adiposity, serum lipids, and CRS severity in the validation cohort. While patients with higher VAT levels had an increased incidence of CRS grade ≥2 (p=0.04), multivariable regression confirmed associations between elevated ACs and CRS (p=0.02), and inverse associations for LPCs (p=0.01). An “AC signature score,” assigning one point for each of four long-chain acylcarnitine species with above-median serum levels, effectively stratified CRS risk: over 50% of patients with scores of 3-4 developed CRS grade ≥2, while none of the patients with a score of 0 experienced CRS beyond grade 1.

Ultimately, scRNA-seq on murine bone marrow cells cultured with either adipocytes or fibroblasts showed a shift towards pro-inflammatory transcriptional programs in adipocyte co-cultured immune cells. In T cells (n=1,588 cells), 79 genes were differentially expressed (73 upregulated) with enrichment for cytokine signaling, and humoral immune response. Among macrophages (n=6,585), 1,968 genes were differentially expressed (1,219 upregulated) with an enrichment for lipid sensing, chemokine signaling, cytokine production, and inflammatory regulation.

CONCLUSIONS These findings identify dynamic adipose tissue–associated lipid profiles as modulators of CAR-T therapy outcomes, with acylcarnitines and phosphatidylethanolamines emerging as candidate metabolic mediators. Consistent associations across two independent cohorts, supported by transcriptomic evidence from co-culture models, provide a framework for integrating host metabolism into CAR-T cell biology and clinical risk stratification.

This content is only available as a PDF.
Sign in via your Institution