CAR-T Cells Engineered to Target GPNMB Selectively Destroy Senescent Blood Vessel Cells

Vascular aging is one of the earliest and most consequential manifestations of organismal aging, with senescent endothelial cells (ECs) releasing a pro-inflammatory secretome — the senescence-associated secretory phenotype (SASP) — that promotes arterial stiffness, atherosclerosis, and systemic cardiometabolic dysfunction. Existing senolytic drugs (navitoclax, dasatinib/quercetin) lack cell-type specificity and carry off-target toxicity risks. This study, from Beijing Anzhen Hospital at Capital Medical University, asked whether CAR-T cell immunotherapy — already proven in hematologic cancers and explored for cardiac fibrosis — could be repurposed to selectively eliminate senescent ECs using a surface antigen unique to aged vessels.

The team first mined three independent GEO transcriptomic datasets (GSE195517, GSE163251, GSE37091) and found GPNMB mRNA consistently elevated in senescent versus young HUVECs across all three datasets. They then built two independent cellular senescence models: replicative senescence (passage 30 vs. passage 5 HUVECs) and doxorubicin-induced premature senescence (10 nM Dox, 48 h treatment). Both models confirmed the senescent state via SA-β-galactosidase staining, flattened cell morphology, and significant upregulation of both GPNMB and p16INK4A at mRNA (qRT-PCR) and protein (western blot) levels (p < 0.05 to p < 0.001). Flow cytometry further verified increased GPNMB surface expression on senescent HUVECs, confirming it as a cell-surface targetable antigen rather than merely an intracellular marker.

The GPNMB-CAR construct was a second-generation design: an anti-GPNMB scFv (derived from monoclonal antibody CR011) linked to a CD8α hinge and transmembrane domain, a CD137 (4-1BB) costimulatory domain, and CD3ζ signaling domain, all delivered via lentiviral transduction into primary human T cells isolated from six healthy donors (3 male, 3 female, aged 25–40). Transduction efficiency was confirmed by GFP co-expression and flow cytometry detection of GPNMB scFv on T cell surfaces. T cells were expanded with IL-2, IL-7, and IL-15 to clinically relevant numbers over 96 hours before use.

In the cytotoxicity assays, GPNMB-CAR-T cells were co-cultured with senescent or young HUVECs at effector-to-target (E:T) ratios of 1:1, 2.5:1, and 5:1 for 6, 12, and 24 hours. LDH release assays showed dose- and time-dependent killing of senescent HUVECs by GPNMB-CAR-T cells, while young HUVECs were largely spared — demonstrating the critical selectivity that chemical senolytics lack. IFN-γ concentrations in co-culture supernatants at 24 hours were significantly elevated in the GPNMB-CAR-T condition versus controls, and CD69 upregulation on CAR-T cells confirmed antigen-driven T cell activation rather than nonspecific bystander killing. The same selective killing pattern was replicated in the doxorubicin-induced senescence model.

The most translationally significant result was the Matrigel plug experiment. Senescent HUVECs (10⁶ cells) were mixed with or without an equal number of GPNMB-CAR-T cells in Matrigel, then injected subcutaneously into immunodeficient NSG mice. After 3 days, plugs were excised and sectioned. Anti-human CD31 immunofluorescence staining showed that GPNMB-CAR-T cells successfully eliminated the senescent HUVECs within the matrix, while control T cells did not. This semi-in vivo model is a meaningful step beyond purely in vitro killing assays, demonstrating that engineered CAR-T cells can operate within a three-dimensional extracellular matrix environment mimicking tissue architecture. The authors acknowledge that fully in vivo efficacy and safety testing in aged animal models remains the critical next step before any clinical translation.