Single-Cell Tumor–Normal Analysis Proposes Combinatorial CAR Strategies

Chimeric antigen receptor (CAR)–expressing immune cells are a promising strategy for cancer therapy, particularly for hematologic malignancies. However, the efficacy of CAR therapies in solid tumors remains limited by difficulties in identifying antigens expressed by most cells within a tumor but that are largely absent in normal tissue. To expand beyond the specificity achieved with a single antigen–targeted CAR, Kwon, Kang, and colleagues explored the gene expression of individual tumor versus normal cells to highlight pairs of candidate genes for use in the design of Boolean logic–gated CARs, which would target cells specifically expressing Gene 1 AND Gene 2, Gene 1 OR Gene 2, or Gene 1 but NOT Gene 2. Integration of single-cell RNA sequencing data of roughly 1.4 million tumor, tumor-infiltrating normal, and reference normal cells enabled the generation of a single-cell tumor–normal expression atlas of 412 patient tumors, representing 17 cancer types and 12 reference organs. Focusing on cell-surface genes, a random forest machine learning algorithm classified cells as tumor or normal depending on gene expression patterns across individual cells, and candidate genes were ranked based on their contribution to the discrimination between tumor and normal. A convolutional neural network was applied to pairwise combinations of the top 100 genes to identify combinatorial AND, OR, and NOT logic gating strategies that encompass a large proportion of tumor cells with minimal off-targets, with the folate receptor 1 (FOLR1) protein emerging as a top candidate for multiple gating strategies in ovarian cancer, including EPCAM-and-FOLR1, FOLR1-or-CD24, and FOLR1-not-CD52. Candidate combinatorial pair expression was confirmed at the single-cell epitope level in patient samples and could distinguish between tumor and normal in multiple cancer contexts including colorectal, breast, and non–small cell lung cancer, for which CLDN3-and-CLDN4 gating seemed particularly promising. Overall, this study creates a single-cell tumor–normal atlas that not only informs the design of logic gate–based CAR therapy but also may aid in the prediction of off-tumor toxicity.

Kwon J, Kang J, Jo A, Seo K, An D, Baykan MY, et al. Single-cell mapping of combinatorial target antigens for CAR switches using logic gates. Nat Biotechnol 2023 Feb 16 [Epub ahead of print].

Note:Research Watch is written by Cancer Discovery editorial staff. Readers are encouraged to consult the original articles for full details. For more Research Watch, visit Cancer Discovery online at http://cancerdiscovery.aacrjournals.org/CDNews.