Pairing a weak glycan-targeting CAR with CD19 revives its cancer-killing power

CD19 CAR T cell therapy is highly effective against B cell cancers, but it has limits. It targets healthy B cells too, which raises infection risk. Cytokine release syndrome and neurotoxicity are also concerns, and a single target leaves tumors able to escape by losing it. Researchers are therefore looking for complementary targets, including tumor-associated sugars. These are hard to target because antibodies against carbohydrates are usually IgM with little affinity maturation.

This study built a CAR from HH2, an IgM antibody that recognizes CD75s, an N-glycan capped with α2,6-linked sialic acid made by the enzyme ST6Gal1. HH2 stained tumor cells from follicular, mantle cell, marginal zone/CLL, and many DLBCL patient samples, as well as most lymphoma cell lines. Staining disappeared after neuraminidase treatment or ST6GAL1 knockout, confirming that HH2 binds the sialylated sugar. Healthy B cells expressed CD75s at a lower level than malignant cells.

A second-generation HH2CAR killed CD75s-positive lymphoma cells in vitro, and this killing was lost in ST6GAL1-knockout cells. However, it triggered little degranulation and only modest cytokine release. In a Burkitt lymphoma mouse model it gave brief tumor control and no survival benefit. Tumor cells recovered from mice had markedly lower CD75s than cultured cells, and HH2CAR T cells could not kill primary B cells, which also have dim CD75s. This points to strong dependence on antigen density.

The team then used a combinatorial design with two CARs on one vector. HH2 carried CD3ζ (signal 1) and CD19 carried 4-1BB (signal 2). The 4-1BB version was chosen over CD28 variants, which performed similarly in killing but showed a slight manufacturing disadvantage. HH2/CD19CAR T cells showed better degranulation, functional avidity, and cytokine production than HH2CAR, though less than CD19CAR, and killed MAVER-1 cells that HH2CAR could not. ST6GAL1-knockout cells were less susceptible to the combination than to CD19CAR, and cells lacking both targets were spared, suggesting better target discrimination. In a JeKo-1 mantle cell lymphoma mouse model, HH2/CD19CAR matched CD19CAR in tumor control and survival (over 40 days), while HH2CAR gave only a slight benefit.

These results suggest that low-affinity CARs, usually discarded, can be useful in logic-gated combinations. Caveats: in vitro, healthy primary B cells were still eliminated by HH2/CD19CAR, so B cell aplasia is not avoided. The in vivo data come from small xenograft cohorts, and some assays used few donors. The efficacy of the combination was comparable to, not better than, CD19CAR alone.