Esophageal squamous cell carcinoma (ESCC) carries a high risk of recurrence and metastasis and is associated with poor prognosis. Chemotherapy remains essential for ESCC treatment. Oxaliplatin (OXA), a third-generation platinum drug, offers advantages in cancer therapy; however, its lack of tumor targeting and high toxicity to normal cells limit clinical efficacy. To address this, we developed photothermal-assisted nanoparticles coated with N1-type neutrophil membranes engineered to express IL-21 on their surface and co-loaded with OXA and BMT-BBT (a photosensitizer) (OXA-BMT@ICVs) to enhance tumor killing. Results demonstrated effective uptake of OXA-BMT@ICVs by ESCC cells (KYSE-150) in vitro. This uptake induced immunogenic cell death (ICD) via activation of oxidative stress and mitochondrial dysfunction, further triggering the cGAS-STING signaling pathway. In vivo studies revealed that, combined with photothermal therapy (PTT), OXA-BMT@ICVs accumulated within tumor tissue, enabling targeted release of OXA and BMT-BBT. The combined action of the cytokine IL-21 and the chemotherapeutic drug activated immune responses, promoting dendritic cell (DC) maturation and CD8+ T cell infiltration. This enhanced inflammatory responses and immune-mediated killing while activating the cGAS-STING pathway, thereby inducing ICD. The combination of OXA-BMT@ICVs and PTT significantly inhibited tumor cell proliferation, promoted apoptosis, and suppressed tumor growth in ESCC-bearing mice. This strategy activates potent immune responses, reduces systemic toxicity, exhibits good biocompatibility and safety, and represents a promising approach for ESCC clinical treatment.

Background: Accurate assessment of pathologic complete response (pCR) after neoadjuvant immunochemotherapy (NICT) is crucial to implement active surveillance or tailor therapeutic strategies for esophageal squamous cell carcinoma (ESCC), while reliable non-invasive methods for pCR prediction are lacking. We aimed to evaluate the potential of integrating circulating tumor DNA (ctDNA) and PET/CT for predicting pCR to NICT for ESCC. Methods: A total of 123 eligible patients were enrolled, including 68 patients from our prospective clinical trial (ChiCTR2000028900) and a real-world study (NCT04822103) that formed the discovery cohort, as well as 55 patients from another clinical trial (ChiCTR2100051763) comprising the validation cohort. Blood samples for ctDNA sequencing and PET/CT metrics were collected before and after NICT. Results: The ctDNA status and PET/CT parameters at the post-NICT stage rather than the pre-NICT stage significantly differentiated pCR from non-pCR patients. ctDNA and PET/CT synergistically enhanced the prediction of pCR from perspectives of sensitivity and specificity, respectively. The model integrating ctDNA concentration and mean standardized uptake value (SUVmean) demonstrated area under curves (AUCs) of 0.860 in the discovery cohort and 0.798 in the validation cohort for pCR prediction and stratified patients into high- and low-risk groups with differential survival prospects. The key gene modules converged on TP53 as the core mutation for pCR prediction, among which those located in the exon regions contributed the most to its predictive capacity. The model constructed based on TP53 mutation and SUVmean differentiated pCR from non-pCR with comparable performance to the model based on PET/CT and the overall ctDNA concentration. Conclusion: The combination of post-treatment TP53-centric ctDNA and PET/CT synergistically enhances the prediction of pCR following NICT in ESCC patients, indicating the potential to inform clinical decision-making for these patients.

Drug repositioning (DR) presents a cost-effective strategy for drug development by identifying novel therapeutic applications for existing drugs. Current computational approaches remain constrained by their inability to synergize localized substructure patterns with global network semantics, leading to overreliance on data augmentation to mitigate latent drug–disease association (DDA) information gaps. To address these limitations, we present multi-view stacked graph convolutional network (MVSGDR), a novel DR framework featuring three technical innovations: (i) multi-view stacked module that enables depth-wise feature enhancement through hierarchical aggregation of multi-hop neighborhood interactions across distinct graph convolutional layers; (ii) bi-level subgraph transformer module that decomposes DDAs into METIS (a graph partitioning tool) informative subgraphs for breadth-wise analysis of external and internal subgraph drug–disease relationships; and (iii) negative sampling balancing strategy that mitigates sample imbalance through negative sample synthesis. Extensive 10-fold cross-validation experiments across four benchmark datasets confirm MVSGDR’s superior performance, demonstrating its statistically significant improvements over existing methods. Moreover, case studies further validate MVSGDR’s potential utility through identification of previously unreported DDAs with supporting literature evidence.