CD24 CAR T-cell design narrows in on two scFvs with the cleanest tumor-specific profile
CD24 has moved into view as a candidate antigen for CAR T-cell therapy because it is present across a wide range of solid and hematologic malignancies and can persist after antigen escape from CD19-directed treatment. In this brief research report, investigators combined public expression data, a single-cell relapse analysis, and a head-to-head functional screen of four second-generation 4-1BB CAR constructs to ask a practical question: which CD24 binder actually behaves like a usable CAR scaffold once transplanted into T cells? The answer was not uniform. Two clones, SN3 and GB7011–05, emerged as the most specific candidates, while the other binders carried liabilities that would matter immediately in a therapeutic setting.
The logic is straightforward. CAR efficacy does not depend on antigen choice alone; it depends on how the single-chain variable fragment shapes epitope recognition, tonic signaling, and target discrimination once embedded in the receptor architecture. A binder that looks acceptable in antibody format can behave very differently after fusion to a hinge, transmembrane domain, and intracellular signaling cassette. Here, the CD24-directed constructs all shared the same second-generation 4-1BB/CD3ζ backbone, so the variable scFv was the main experimental lever. That design makes the comparison unusually clean.
CD24 expression spans tumor types and survives at least one CD19-negative relapse
Public transcriptomic data showed elevated CD24 expression in several cancers relative to normal tissue, including bladder urothelial carcinoma, breast invasive carcinoma, cholangiocarcinoma, kidney renal papillary cell carcinoma, liver hepatocellular carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, stomach adenocarcinoma, and uterine corpus endometrial carcinoma. The same analysis also showed CD24 in normal tissues, which immediately frames the antigen as broadly expressed rather than tumor-restricted. For CAR design, that matters because broad expression widens the therapeutic opportunity and the toxicity concern at the same time.
A single-cell dataset from a bone marrow sample taken before and after CD19 CAR T-cell therapy added a second layer of relevance. In a B-cell acute lymphoblastic leukemia relapse that had lost CD19, CD24 remained detectable. Even without over-reading a single case, the biological implication is plain: CD24 may still be available after classical antigen escape, making it attractive either as a salvage target or as part of a combinatorial strategy designed to reduce the odds that leukemic clones can simply shift phenotypes and survive.
One CAR format, four binders, and sharply different biologic behavior
The study generated four CD24 CARs from publicly available antibody sequences: H2L3, SWA11, SN3, and GB7011–05. All four were built into a matched second-generation platform with a CD8α signal peptide, a CD8-derived transmembrane domain, 4-1BB costimulation, and CD3ζ signaling. In other words, the investigators removed most of the usual engineering noise and kept the focus on the binder itself.
That choice exposed differences that matter at the bedside. H2L3-derived CAR T cells showed high basal IFN-γ production, a classic warning sign of tonic activation. Constitutive cytokine release can reflect spontaneous receptor clustering or excessive sensitivity, both of which can exhaust T cells before they ever meet a tumor. SWA11-derived CAR T cells were even more problematic in a different way: they displayed cytotoxicity against cell lines with undetectable CD24 expression. A CAR that kills without the antigen it is supposed to recognize raises immediate specificity concerns, especially for a target like CD24 that is already present on some normal tissues.
That is the sharpest safety signal in the paper.
SN3 and GB7011–05 keep specificity without the obvious functional liabilities
By contrast, the SN3 and GB7011–05 CARs behaved as the most promising pair in the screen. Both were described as CD24-specific and phenotypically equivalent, which makes them attractive starting points for further preclinical development. The report does not claim they are ready for clinical translation, and it should not. What it does show is more precise and more useful: among the four tested binders, these two preserved the desired antigen dependence without the obvious basal activation seen with H2L3 or the off-target cytotoxicity associated with SWA11.
That distinction is especially important for CD24 because the antigen is not tumor-exclusive. A workable CD24 CAR will need enough sensitivity to respond to malignant cells, but not so much background activity that it begins to erode healthy epithelium or B-cell compartments. In a target space like this, the cleanest binder is not necessarily the one with the strongest apparent activation signal; it is the one that maintains a disciplined response profile when the receptor is fully wired into the T-cell signaling machinery.
Why glyco-epitope recognition may matter for CD24 targeting
The four binders were not equivalent in what they recognized. H2L3, SWA11, and GB7011–05 were directed against the protein core of CD24, while SN3 recognized a glyco-epitope. That detail is more than biochemical ornamentation. CD24 is heavily glycosylated, and glycosylation varies by cell type and malignant state. A glyco-epitope may therefore help the CAR distinguish tumor-associated CD24 from healthy counterparts that present a different surface architecture. If so, the glycan context could become as important as the polypeptide backbone itself in determining selectivity.
The report also explored predicted physicochemical features of the scFvs, including solubility and positive charge patch scores, but the functional readout remains the main story. Predictions are useful when they help explain why one construct behaves better than another, yet the decisive observation here came from the matched cellular assays: CD24-directed CARs are not interchangeable, and binder choice directly changes specificity and baseline activation.
A translational path that still depends on toxicity control
For hematologic disease, CD24 remains especially intriguing because it may persist after loss of CD19. For solid tumors, the attraction is broader but more difficult to operationalize because the same antigen that is enriched on malignant tissue can also appear on healthy cells. That duality is the central design problem in this field, and this report addresses it in a practical way by narrowing the candidate list to two more defensible binders.
The next steps are obvious from the data, even if they are not yet answered by it: deeper testing against normal tissues, tighter functional studies in primary tumor material, and direct comparison of these binders under conditions that better mimic in vivo antigen density and tissue context. The present screen does not establish therapeutic safety, persistence, or antitumor potency in animals or patients. It does, however, identify two scFvs that merit continued development while making clear that at least two other CD24 binders carry liabilities that should not be ignored.
Clinical development of CD24 CAR T-cell therapy will rise or fall on antigen selectivity, and this study shows how quickly that selectivity can collapse when the wrong binder is embedded in an otherwise well-built CAR framework.
- • Frontiers in Immunology DOI 10.3389/fimmu.2026.1989845: https://doi.org/10.3389/fimmu.2026.1989845