CCL8/CCL13-Producing Tumor-Associated Macrophages Linked to Poor Outcomes after CAR T Cell Therapy for LBCL

Introduction

Autologous T cells engineered to express a chimeric antigen receptor (CAR) have transformed the standard of care for patients with relapsed/refractory large B cell lymphoma (r/r LBCL), but >50% of patients progress following therapy. To elucidate the underlying mechanisms of resistance in the tumor microenvironment (TME), we conducted a comprehensive analysis of the LBCL TME following CAR T cell infusion using multi-omic single-cell sequencing and interactome analysis tools.

Methods

We studied 18 consecutive LBCL patients and one cutaneous B-lineage acute lymphoblastic leukemia (B-ALL) patient treated with a CD19-CAR (axi-cel; n = 12) or a bispecific CD19/CD22-CAR T cell therapy in an ongoing investigator-initiated trial (NCT03233854; n = 7). On-treatment fine needle aspirate biopsies were obtained from all patients at the time of early CAR T cell expansion (CD19-CAR, day 2; CD19/CD22-CAR, day 7). Live cells were sorted from fresh biopsies and analyzed using single-cell sequencing for transcriptome and a panel of surface proteins using CITE-seq, yielding 31,808 high-quality cells.

Results

We found that an intratumoral CAR T cell exhaustion program characterized by high expression of IRF4 and IRF8 was associated with progression at 3 months, whereas expression of JUN and FOS was linked to durable remission. Transcription factor activity analysis using SCENIC corroborated these results.

To reconstruct cell-cell crosstalk in the post-CAR LBCL TME in the context of response or progression at 30 days, 3 months, or 6 months, we employed REMI, CellPhoneDB, and CellChat algorithms. In progressing patients, our analyses implicated interactions of CCL8, CCL13, and CCL18 secreted by myeloid cells with their cognate receptors on myeloid, conventional, or regulatory T cells. These interactions were diminished or absent in responding patients.

Using CIBERSORTx deconvolution analysis of 3 publicly available baseline bulk gene expression datasets from CAR-treated patients (axi-cel, n = 242), we validated that high CCL8, CCL13, or CCL18 levels in myeloid cells are linked to poor outcomes after CAR T cell therapy for LBCL. We extended our investigation to LBCL patient outcomes following standard of care (SOC) CHOP or R-CHOP therapies. Deconvolution analysis of 3 independent baseline LBCL cohorts comprising 310 CHOP-treated and 233 R-CHOP-treated patients showed that high CCL8, CCL13, or CCL18 levels in myeloid cells are also linked to poor outcomes after SOC therapy for LBCL.

To probe the underlying biological mechanisms, we performed differential gene expression and pathway enrichment analyses. We observed that CCL8 and CCL13 were co-expressed in a subset of tumor-associated macrophages (TAMs) enriched for both CD206 and CD209, whereas CCL18 was expressed in APOE+ TAMs. The CCL8/CCL13+ TAM subset showed high proliferative capacity and interferon signaling, and was linked to poor outcomes following both SOC and CAR T cell therapies in all available data, suggesting a common mechanism of treatment resistance.

Conclusion

In this study, we present the first on-treatment CAR T TME singe-cell sequencing dataset and the resulting LBCL TME interactomes. Our analysis raises the prospect that TAMs expressing CCL8 and CCL13 act as a barrier to therapeutic efficacy in both CAR T and SOC therapies for LBCL. Our study presents deep characterization of the LBCL TME in the context of CAR T cell immunotherapy and nominates TME-mediated resistance mechanisms that could be modulated to enhance efficacy of CAR T cell therapies for r/r LBCL and could generalize to other immunotherapies and malignancies beyond LBCL.