Single-cell genomics links targeted functional manipulations to efficacy-associated chromatin signatures in CAR T cells

Technical Summary: Single-cell genomics links targeted functional manipulations to efficacy-associated chromatin signatures in CAR T cells

Problem Statement
Chimeric antigen receptor (CAR) T cell therapies have demonstrated remarkable efficacy in certain hematologic malignancies, yet clinical outcomes remain inconsistent across different targets and patient populations. This heterogeneity is driven by variability in autologous products and the T cell pool itself. While preclinical studies show that modifying CAR architecture or T cell manufacturing conditions (e.g., cytokine milieu) can improve efficacy, linking these specific pre-transfer manipulations to the long-term in vivo fates of T cells has been challenging. Traditional pooled screening methods often rely on in vitro readouts of abundance or transcriptomics, which provide only a snapshot of cellular status and may fail to predict long-term differentiation programs or persistence. Furthermore, comparing T cells in separate in vivo hosts introduces confounding variables such as disparate tumor burdens and antigen exposure.

Methodology
The authors adapted the "Spear-ATAC" (single-cell perturbations with an accessibility read-out) framework to track the chromatin states of pooled CAR T cells using genetically encoded DNA barcodes.

• Vector Design: They engineered murine and human CAR constructs to include a 16bp DNA barcode flanked by Nextera sequencing adapters, positioned downstream of a reporter (mThy1.1).
• Pooled Screening: Distinctly barcoded CAR variants were transduced into primary murine or human T cells. These cells were then pooled and subjected to single-cell ATAC-seq (scATAC-seq).
• Experimental Models:
  • Murine Model: Primary CD8+ T cells were transduced with barcoded CARs and expanded under three IL-2 conditions: High (100 IU/mL), Low (10 IU/mL), and a "Swap" condition (High to Low) to mimic the IL-2 deprivation experienced post-transfer.
  • Human Model: Primary human T cells were transduced with a panel of CD22-CARs varying in antigen-binding affinity (Standard vs. High) and linker length (Short vs. Long), which modulates tonic 4-1BB signaling. These were profiled in vitro (Pre-Transfer) and after adoptive transfer into NSG mice bearing CD22-low leukemia (Post-Transfer).
• Data Analysis: The pipeline involved rigorous filtering to distinguish species (mouse vs. human) and barcode identity based on read mapping and specificity thresholds. Downstream analysis utilized ArchR for dimensionality reduction (Iterative LSI, UMAP), ChromVAR for inferring transcription factor (TF) activity, and gene scoring to link chromatin accessibility to regulatory programs.

Key Results

1. Method Validation: The study established robust in silico filtering parameters (specificity >0.95, >100 barcode reads) to accurately assign cells to their respective barcodes and species, minimizing multiplets and misassignment in mixed "barnyard" pools.
2. IL-2 Concentration Dictates TF Programs: In murine CAR T cells, IL-2 concentration during expansion drove distinct and reproducible chromatin accessibility profiles.
   • High IL-2 (Hi) was associated with amplified STAT5 activity.
   • Low IL-2 (Lo) and the Swap condition showed distinct Il2 gene scores and variable TF activities (e.g., FOSL/JUND, TCF/LEF, RUNX).
   • The "Swap" condition revealed that some programs were transient (reverting upon IL-2 reduction) while others remained stable, highlighting the plasticity of IL-2-dependent transcriptional states.
3. Adoptive Transfer Drives Antigen-Independent Programs: In human CAR T cells, the transition from in vitro culture to the in vivo xenograft environment caused a major shift in chromatin accessibility, independent of antigen engagement. This shift was characterized by a drop in IL-2 signaling (reduced STAT5 activity) and an increase in IL-2 production signatures, particularly in CD4+ cells, suggesting a helper role for CD4 CAR T cells in vivo.
4. CAR Architecture Determines Efficacy-Associated Signatures:
   • Linker Length: Short linkers (SL), which induce strong tonic 4-1BB signaling, drove altered effector differentiation programs in both CD4 and CD8 subsets. Specifically, SL CARs (SALL, SASL) were associated with reduced T-box (TBX/EOMES) activity and increased NFκB or nuclear receptor (NR) activity, correlating with lower anti-tumor efficacy.
   • Affinity and Linker: High-affinity, long-linker (HALL) CARs, which minimize tonic signaling, promoted robust CD8 effector programming (high T-box activity) and closely mimicked the chromatin signatures of highly efficacious CD19-CARs.
   • Cluster Analysis: Unsupervised clustering separated cells based on efficacy. High-efficacy CARs (CD19, HALL) clustered together with strong effector differentiation signatures (T-box, ZNF), while low-efficacy or bystander CARs (CD33, SALL) clustered with memory-like or resting signatures (TCF7L/LEF).
5. Mechanistic Insight into Toxicity: The study noted that the SASL architecture (Short Linker, Standard Affinity), which is associated with severe clinical toxicity (IEC-HS), drove distinct transcriptional programs characterized by altered effector differentiation (e.g., reduced PRF1 accessibility), suggesting a mechanistic link between tonic signaling-induced chromatin states and toxicity.

Significance and Claims
The paper claims to establish a methodology for linking targeted functional manipulations (culture conditions, CAR design) directly to single-cell chromatin profiles in vitro and in vivo.

• Decoding Fate: By using genetic barcoding with scATAC-seq, the authors demonstrate the ability to track how specific modifications dictate stable versus transient transcriptional programs and long-term T cell fates.
• Optimization Strategy: The data suggests that tuning IL-2 concentrations during manufacturing can program specific TF activities, and that engineering CARs to minimize detrimental tonic signaling (via long linkers) can restore physiologic effector differentiation programs necessary for controlling antigen-low tumors.
• Clinical Relevance: The findings provide a chromatin-level explanation for why certain CAR architectures (like HALL) are more effective than others (like SL variants) and offer potential insights into the mechanisms of severe toxicities like IEC-HS, proposing that modulating tonic signaling could improve the therapeutic window of CAR T cell therapies.

The authors conclude that this approach provides a roadmap for decoding the chromatin-level determinants of CAR T cell function, potentially enabling the rational design of next-generation immunotherapies with enhanced persistence and efficacy.