Citations

ABSTRACT Purpose Layer fMRI is an increasingly utilized technique that provides insights into the laminar organization of brain activity. However, both blood‐oxygen‐level‐dependent (BOLD) fMRI and vascular space occupancy data (VASO) have certain limitations, such as bias towards larger cortical veins in BOLD fMRI and high specific absorption rate in VASO. This study aims to explore the feasibility of whole‐brain laminar functional quantitative susceptibility mapping (fQSM) compared to laminar BOLD fMRI and VASO at ultra‐high field. Methods Data were acquired using 3D EPI techniques. Complex data were denoised with NORDIC and susceptibility maps were computed using 3D path‐based unwrapping, the variable‐kernel sophisticated harmonic artifact reduction, as well as the streaking artifact reduction for QSM algorithms. To assess layer‐specific activation, 20 layers were segmented in the somatosensory and motor cortices, obtained from a finger tapping paradigm, and further averaged into six anatomical cortical layers. The magnitude of signal change and z ‐scores were compared across layers for each technique. Results fQSM showed the largest activation‐dependent mean susceptibility decrease in Layers II/III in M1 and Layers I/II in S1 with up to −1.3 ppb while BOLD showed the strongest mean signal increase in Layer I. Our data suggest that fQSM demonstrates less bias towards activation in superficial layers compared to BOLD. Moreover, activation‐based susceptibility change was comparable to VASO data. Conclusion Studying whole‐brain, layer‐dependent activation with submillimeter fQSM is feasible and reduces bias towards venous drainage effects on the cortical surface compared to BOLD, thereby enabling better localization of laminar activation.

ABSTRACT Extracellular vesicles (EVs) are increasingly recognized as mediators of intercellular communication in cancer. We previously demonstrated that re‐expression of the leucine‐rich domain (LRD) of neurofibromin (NF1) suppresses glioblastoma (GBM) invasion and limits microglia/macrophage infiltration into the tumor microenvironment (TME). Given the central role of tumor‐associated microglia/macrophages (TAMs) in GBM progression, we investigated whether NF1‐LRD‐containing EVs (NF1‐LRD‐EVs) could modulate TAM function and remodel the TME. Our results showed that NF1‐LRD‐EVs attenuated microglia and macrophage recruitment in migration assays, consistent with reduced microglia/macrophage recruitment observed in vivo . In parallel, treatment with NF1‐LRD‐EVs enhanced phagocytic activity of both microglia and iPSC‐derived macrophages, accompanied by induction of pro‐inflammatory cytokines TNF‐α, IL‐6, and IL‐1β, and downregulation of immunosuppressive mediators such as Arginase‐1 and IL‐10. Mechanistically, NF1‐LRD‐EVs induced ADAM17‐associated inflammatory signaling, accompanied by NF‐κB activation. Pharmacological inhibition of ADAM17 reduced TNF‐α release and attenuated NF‐κB activation, supporting a role for ADAM17‐dependent signaling in amplifying this inflammatory response. Together, these findings show that NF1‐LRD‐EVs reprogram TAMs toward a pro‐inflammatory phenotype and modulate the GBM immune microenvironment. These results provide a framework for understanding NF1‐LRD‐EV‐mediated immune regulation and support further investigation of EV‐mediated immune modulation of the TME.

Abstract Anthracycline-induced cardiotoxicity (AIC) remains a dose-limiting complication of cancer treatment, often leading to irreversible heart failure, yet effective cardioprotective therapies remain limited. Graph-driven drug repurposing offers a rapid pathway to identify cardioprotective therapies, but different computational models often generate divergent candidate lists that exceed experimental validation capacity. We applied a novel "Union-then-Filter" framework to prioritize therapeutics for AIC using AI-based approaches and evaluated candidates in zebrafish models. Three state-of-the-art graph learning models (TxGNN, CompGCN, and RLR) were trained on the PrimeKG to generate an initial candidate pool. A Large Language Model (LLM) agent automated evidence synthesis across 4,286 PubMed abstracts for the top 50 drugs from each model. Candidates were stratified using dual decision criteria and filtered against existing clinical trials. The pipeline narrowed 150 initial candidates to four novel candidates: Cysteine, Dasatinib, Tranilast, and Tretinoin. Prospective in vivo validation was subsequently conducted using an adult AIC (aAIC) zebrafish model. Phenotypic screening via echocardiography revealed that Cysteine, Tranilast, and Tretinoin significantly restored ejection fraction (EF) (overall one-way ANOVA: \(\:F\left(\text{4,95}\right)=47.66,\:\:p<0.0001\); Tukey's post hoc: \(\:\:\:p<0.01\) for all three treatments vs. DOX+ control) and normalized cardiac remodeling markers in doxorubicin-treated subjects (Kruskal-Wallis: \(\:{\chi\:}^{2}\left(4\right)=17.21,\:\:p=0.0018\); Dunn’s post hoc: \(\:\:\:p<0.01\) for all three comparisons vs. DOX control). This study demonstrates the translational utility of an LLM-integrated AI pipeline for drug repurposing and highlights zebrafish AIC as an efficient bridge between in silico prediction and in vivo phenotypic validation in cardio-oncology.