Moolayadukkam, S. | bioRxiv Science

bioRxiv2026

Transcription Factor-Mediated Reprogramming of Cancer-Associated Fibroblasts Reveals Targetable Vulnerabilities in Solid Tumors

Cancer-associated fibroblasts (CAFs) contribute to immune exclusion and therapy resistance in solid tumors, limiting the efficacy of chimeric antigen receptor (CAR) T cell and immune cell therapy. To overcome this, we developed a transcription factor (TF)-based strategy to reprogram prostate-derived CAFs (pCAFs) into normal fibroblast-like cells (NFs). We prioritized TFs enriched in quiescent stellate cells--Vitamin D receptor (VDR), Peroxisome Proliferator-Activated Receptor gamma (PPAR{gamma}), and p53--and selected VDR for proof-of-concept studies. Lentiviral VDR expression in pCAFs produced VDR-reprogrammed NFs (VDR-rpNFs) with reduced CAF markers, increased ATP, and suppressed TGF-{beta} and IL6, indicating phenotypic and metabolic reversion. In both in vitro 3D co-cultures and in vivo, VDR-rpNFs disrupted tumor architecture, enhanced CAR T cell infiltration, and reduced necrosis. PPAR{gamma}- and p53-rpNFs showed similar reprogramming effects. These results suggest TF-guided fibroblast reprogramming as a viable strategy to remodel the tumor microenvironment and improve CAR T cell efficacy in solid tumors.

cancer biology↗

bioRxiv2025

AI-Driven and 3D-Bioprinted New Approach Methodology (NAM) Identifies NEO100 as Potent Ultrasound-Activated Therapeutic for Primary and Metastatic Brain Tumors

Primary and metastatic brain tumors are among the deadliest and treatment-resistant cancers, mainly because of their inherent resistance to chemoradiation and limited drug delivery across the blood-brain barrier (BBB). Identifying molecules that can cross the BBB and serve as sonosensitizers is crucial for developing noninvasive, targeted therapies such as sonodynamic therapy (SDT). To overcome the limitations of traditional low-throughput screening, a New Approach Methodology (NAM) was developed, starting with AI-driven molecular discovery. A positive-unlabeled neural network, trained on over 200 molecular descriptors, was used to predict compounds likely to respond to focused ultrasound (FUS), penetrate the BBB, and mediate therapeutic sonodynamic activity. While AI helps prioritize potent candidates, effective SDT also depends on clinically scalable FUS delivery. Low-intensity FUS enables noninvasive activation of small molecules in the brain and has been validated in several clinical settings. Therefore, AI-guided predictions were combined with ultrasound-based sonodynamic testing using parameters relevant to human therapy. A major challenge in SDT development is the lack of rapid, physiologically relevant tumor models, as traditional 3D organoids take weeks to mature, delaying validation. To address this, a magnetic bioprinting platform was used to produce uniform 3D tumor spheroids within hours, enabling high-throughput screening of SDT. These spheroids replicate key microenvironmental gradients, supporting reliable ultrasound-driven cytotoxicity testing. Through this combined NAM pipeline, AI identified NEO100--an ultrapure, pharmaceutical-grade perillyl alcohol currently in Phase 2a clinical trials--as a promising sonosensitizer candidate. These predictions were validated in rapidly 3D bioprinted tumor models representing glioblastoma, Group 3 pediatric medulloblastoma, meningioma, and breast-to-brain and lung-to-brain metastases. In all tumor types, ultrasound activation significantly increased NEO100s cytotoxicity. Given its established safety profile in humans and ability to cross the BBB, NEO100 demonstrates how integrating AI-based molecular discovery, accelerated 3D bioprinting, and clinically relevant ultrasound parameters can rapidly advance precision sonodynamic therapies for various primary and metastatic brain cancers.

cancer biology↗