The tumor microenvironment (TME) imposes structural and metabolic barriers that impair CAR-T cell function. Hypoxia is a central factor in this resistance, yet conventional in vitro cytotoxicity assays lack the spatial resolution to quantify CAR-T performance across oxygen gradients.

To develop an advanced heterotypic 2D co-culture model of CD19⁺ lymphomas that mimics radial oxygen and metabolic gradients using the “coverslip hypoxia” approach, enabling automated, spatially resolved quantification of CAR-T activity via automated quantitative fluorescence microscopy - high-content screening (HCS). This configuration allows simultaneous assessment of hypoxic and normoxic niches within the same well, supporting high-throughput screening of pharmacological or genetic strategies in 96-well plates.

HS-5 stromal cells (10⁴/well) were seeded and, the next day, co-cultured with CD19⁺ Raji or CD19⁻ K562 tumor cells (10⁴/well, eFluor 670-labeled) and either anti-CD19 CAR-T or non-transduced T cells (2 × 10⁴/well, CellTrace Violet-labeled) under four conditions, in triplicate. Hypoxia was induced with 5 mm glass coverslips (CS), with wells without coverslips (No-CS) as normoxic controls. Whole-well images (5 × 5 fields, 10× Obj.) were acquired on an ImageXpress MICRO XLS HCS system, stitched, and processed in CellProfiler to define four spatially distinct regions: CS_InnerCore (severe hypoxia at CS center), CS_OuterCore (moderate hypoxia), CS_Periphery (transition at CS edge), Outside (normoxia beyond CS). Cy5 and DAPI channels enabled segmentation of tumor and T cells, respectively, assigning each to its oxygen-defined compartment. Mean SYTOX Green intensity (FITC channel) identified dead cells, while eccentricity (ranging from 0, round, to 1, elongated) served as a T cell activation metric.

In Raji co-cultures, CAR-T cells induced marked tumor death after 24h, highest in Outside (∼40%) and No-CS (∼60%) regions, but minimal in hypoxic regions (< 10%). Non-transduced T cells showed negligible cytotoxicity, with only modest increases (∼20%) in Outside and No-CS regions. In CD19⁻ K562 co-cultures, tumor death was negligible across all regions (< 5–10% in normoxia), confirming antigen specificity. Effector death was greater in normoxia (∼25–40% in Outside/No-CS) versus hypoxia (∼5–10%), suggesting low oxygen protects both targets and effectors. Morphometric analysis showed CAR-T with Raji targets had increased eccentricity (up to ∼0.6 at 24h), with a hypoxia-to-normoxia gradient, indicating greater motility under oxygenated conditions. CAR-T with K562 showed no significant eccentricity changes, confirming antigen-specific activation. Non-transduced T cells maintained lower eccentricity (∼0.45–0.5) in all conditions.

These findings show that hypoxia gradients selectively impair CAR-T cytotoxicity, survival, and motility, while preserving antigen specificity. The assay integrates physiologically relevant oxygen gradients with automated, multiparametric image analysis, enabling real-time, region-specific quantification of CAR-T function across parallel conditions. Unlike bulk readouts, this platform captures microenvironment-driven functional heterogeneity and directly compares matched hypoxic vs. normoxic zones, making it a powerful preclinical screening tool to identify strategies restoring CAR-T efficacy in the hostile stromal–hypoxic TME. This study was financed, in part, by the São Paulo Research Foundation (FAPESP), Brasil. Process Number 2022/12856-6.