An integrative meningioma surfaceome atlas and a transferable framework for CAR-T target nomination

Meningioma is the most common primary intracranial tumor, yet recurrent and high-grade disease has few durable systemic options, and, unlike hematological malignancies, no systematically filtered cell-surface antigen resource exists to guide chimeric antigen receptor (CAR) T development. We address this gap not by nominating a trial-ready target but by building two reusable resources: an integrative meningioma surfaceome atlas and a transferable, gate-based framework that separates transcriptomic discovery from CAR-T target nomination. Analyzing single-cell RNA sequencing from eight meningioma patients (GSE183655; 80,220 quality-controlled cells), patient-aware differential expression followed by a curated surfaceome filter (UniProt plasma-membrane localization; extracellular domain ≥20 aa; exclusion of ribosomal, nuclear, cytoplasmic, and secreted proteins) reduced 83 scored candidates to 13 bona fide plasma-membrane (Tier A) antigens, removing prominent false positives such as the nuclear proteins FOXC1 and MGMT, the cytoplasmic enzyme lactate dehydrogenase A (LDHA), and the secreted protein PTGDS. Because expression ranking establishes discovery but not CAR-T suitability, we then applied explicit nomination gates: a systemic-safety score integrating whole-body GTEx transcriptomics with Human Protein Atlas protein-level evidence; a membrane-topology tractability gate distinguishing antibody-accessible single-pass and GPI-anchored ectodomains from multi-pass transporters and tail-anchored proteins; and a safety × coverage metric. Under stringent thresholds (combined safety ≥0.55, tumor-cell detection fraction ≥50%, safety × coverage ≥0.30), no Tier A antigen qualified. Ubiquitously expressed antigens were correctly demoted, including the tetraspanin CD9 (53.2% tumor-cell coverage but pan-tissue protein expression and high plasma abundance), and the highest-ranked transporter SLC7A2 was demoted by multi-pass topology (largest extracellular loop 38 aa) despite favorable transcriptomic safety. Leukemia inhibitory factor receptor (LIFR), a single-pass type I receptor with a contiguous 789-aa ectodomain, was the only antibody-tractable nominee for experimental follow-up yet remained liability laden (combined safety 0.41; safety × coverage 0.135) and is not a validated target. Cross-dataset expression overlap (93.3%) reflected mRNA reproducibility, not CAR-T validity or clinical validation, and generic-antibody docking was exploratory only. All findings are transcriptomic and require surface-level confirmation. This work contributes a reproducible meningioma surfaceome atlas and a transferable nomination framework and shows that rigorous gating leaves no single antigen with a favorable safety × coverage × tractability profile, motivating bispecific and logic-gated CAR-T strategies.