Abstract We previously demonstrated efficacy for bivalent chimeric antigen receptor (CAR) T cells targeting EGFR epitope 806 and IL13Ra2 (CART-EGFR-IL13Rα2 cells) in patients with rGBM. However, tumor assessment on magnetic resonance imaging (MRI) can be difficult following immunotherapy for brain tumors, and biomarkers to predict response to intracerebroventricular (ICV) CAR T cell therapy for GBM have not been developed. Given that TERTp mutations C228T and C250T are detected in the vast majority of GBM tumors and are not subject to spatial or temporal heterogeneity, we hypothesized that TERTp mutant copies in CSF ctDNA would increase rapidly as a result of tumor killing and be associated with response to treatment. METHODS: We adapted a previously published droplet digital PCR (ddPCR) assay for high-sensitivity detection of TERT promoter (TERTp) mutations C228T and C250T for longitudinal CSF samples obtained via Ommaya reservoir in patients treated in our phase 1 clinical trial of CART-EGFR-IL13ra2 cells. After cell-free DNA extraction, the TERTp locus was pre-amplified, and 7-deaza-dGTP (7dG) utilized to mitigate the deleterious effects of CpG-rich regions on mutation detection. Negative controls included CSF samples from patients with non-cancer neurologic conditions. Clinical tumor tissue next generation sequencing was used to identify the TERTp mutation. RESULTS: To assess sensitivity, we first applied the assay to CSF samples obtained intra-operatively from 15 GBM patients who received standard of care chemotherapy and radiation. The known TERTp mutations were detected in 13 patients (86. 7%). TERTp mutations were undetectable in the CSF of 9 negative control patients with multiple sclerosis. Among 15 Day 0 (D0, pre-infusion) CSF samples obtained from patients with rGBM who received CART-EGFR-IL13Rα2 therapy, the known TERTp mutations were detected in 15 (100. 0%). Among CART patients, D0 mutant copies/mL were correlated with MRI-determined D0 bidimensional tumor measurements (Spearman R=0. 6879, P=0. 0231). Patients with D0 mutant copies/mL below median had longer PFS (log-rank P=0. 0380). Next, we analyzed 113 longitudinal CSF samples collected for 15 CART patients over the first month post-CART infusion. ctDNA levels peaked on D7 after infusion, with TERTp mutant copies/mL significantly higher than at D0 (P=0. 0479). A D7 increase in mutant copies was negatively correlated with tumor size at D1 and at one month (Spearman R=-0. 7000, P=0. 0204 and Spearman R=-0. 5714, P=0. 1511, respectively). All patients with a durable response of 4 months had increased mutant copies within 7 days of CART infusion. For one patient with imaging showing significant tumor enlargement at 1 month followed by a decrease in tumor volume at the 2-month scan, D7 change in mutant copies/mL was +1407. 5, the second highest increase of all patients, suggesting tumor killing and reinforcing MRI findings consistent with pseudoprogression. CONCLUSION: High-sensitivity ddPCR-based mutation detection in CSF may have clinical utility for prediction of response to ICV CART therapy. Citation Format: Sophia G. Giliberto, Rahul Chowdhury, Melinda R. Yin, Dominique G. Ballinger, Siri C. Dandu, Jacob E. Till, Adam S. Corner, Francisco A. Bizouarn, Donald M. O'Rourke, Zev A. Binder, Stephen J. Bagley, Erica L. Carpenter. Early cerebrospinal fluid (CSF) circulating tumor DNA (ctDNA) dynamics following intracerebroventricular CAR T cell therapy predict response to treatment of recurrent glioblastoma (rGBM) abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (6Suppl): Abstract nr A051.
Building similarity graph...
Analyzing shared references across papers
Loading...
Giliberto et al. (Mon,) studied this question.
www.synapsesocial.com/papers/69c37adcb34aaaeb1a67cd3f — DOI: https://doi.org/10.1158/1538-7445.brain26-a051
Sophia G. Giliberto
Rahul Roy Chowdhury
Melinda Yin
Cancer Research
University of Pennsylvania
Bio-Rad (United States)
Building similarity graph...
Analyzing shared references across papers
Loading...