Matched Healthy Lung Organoids Reveal Dose-Dependent CAR-T Selectivity
Lung tumor and matched healthy epithelial organoids exposed target- and effector-ratio-dependent CAR-T responses. Some conditions injured both tissues, while antigen abundance and T-cell activation alone did not reliably establish selective tumor killing.
Testing the Tumor Beside Its Healthy Counterpart
CAR-T target selection must consider healthy tissue as well as the tumor. A Charité-led study in Nature Biomedical Engineering established lung tumor organoids alongside matched healthy epithelial organoids and tested CAR T cells against both. The comparison exposed conditions in which a seemingly active treatment also killed healthy cultures.
Six of twelve fresh or short-frozen samples yielded long-term tumor cultures maintained for at least six months. Four were overtaken by healthy epithelial cells, and two declined after short-term growth. Eight additional samples frozen for more than six months failed to yield tumor organoids. Genomic analyses included eight tumor lines from six patients, reflecting more than one culture in some cases.
The epithelial cultures retained approximately 75% of differential DNA methylation patterns, while showing culture selection and some subclonal changes. Stromal and immune components were absent until T cells were added. Matched healthy organoids were often intermediate or resistant to chemotherapy, so they did not provide a convincing healthy-tissue toxicity control for every drug class.
Target Expression Required Tissue and Culture Checks
The investigators compared standard chemotherapy responses in selected cases and profiled potential CAR targets. Some culture-associated changes were consequential: cMET was overexpressed in organoids without matching expression in the original tissue, while MUC1 was high in healthy and tumor cultures. These observations limited the suitability of those antigens for selective targeting in the system.
Further experiments used HER2- and PD-L1-directed CAR T cells from healthy donors. They were not autologous patient T cells. The engineering strategy removed endogenous T-cell-receptor expression, helping to limit alloreactivity in the coculture. CARs against CD19 or EGFRvIII supplied antigen-related controls.
Activation and Killing Were Measured Separately
T-cell activation and cytokine release were assessed after 12–16 hours at an effector-to-target ratio of 1:1. A separate live-imaging assay followed intact living-organoid area every three hours for up to 30 hours at ratios of 5:1, 1:1 and 1:5. This design distinguished T-cell engagement from loss of the target structure.
High effector ratios caused substantial killing in both tumor and healthy organoids. In two patient-derived pairs, HER2 CAR T cells killed healthy and tumor cultures similarly, and those pairs were excluded from further selectivity analysis. Other pairs, including models 9 and 12, showed greater tumor than healthy-organoid killing. In model 9, HER2-specific killing was 53% at the intermediate ratio and 14.6% at the lower ratio.
Activation and cytokine release sometimes occurred with limited organoid killing. A positive activation assay therefore did not identify every effective cytotoxic condition. The matched healthy cultures also showed why increasing the effector ratio could trade apparent potency for reduced selectivity.
Antigen Abundance Did Not Fully Explain the Response
HER2 abundance had a positive but nonsignificant association with maximum killing: Spearman’s correlation was 0.771, with P = 0.103. Model 12 showed strong killing despite a relatively low HER2 staining score. Cytokine and protein profiles differed between responsive models, including interleukin-related signals, but these associations were not tested as causal resistance mechanisms.
PD-L1-directed CAR T cells showed selective activity in model 12 within the main comparison. Two later patient cases tested feasibility within approximately three months of sampling: one showed selective responses to both CAR targets, while the other showed little killing. These experiments demonstrate variation in model response rather than clinical CAR-T response prediction.
A Selectivity Assay with Remaining Biological Gaps
Paired tumor and healthy cultures let the investigators compare a target’s functional consequences in related epithelial tissues. They also revealed limits of antigen staining, activation measurements and high effector ratios as stand-alone selection criteria. Healthy-donor T cells, epithelial-only organoids, incomplete culture establishment and short assay duration constrain the inference. No matched clinical CAR-T outcomes were available to validate treatment selection or predict toxicity in patients.