A pathology-to-single-cell framework links SPP1+ M2 macrophages with NETs-prone high-risk niches in HCC

BACKGROUND: Neutrophil extracellular traps (NETs) facilitate hepatocellular carcinoma (HCC) progression, but the upstream cellular organizers and the histopathological correlates of NETosis-prone niches remain poorly defined. We aimed to build a pathology-to-single-cell framework to (i) quantify NETs-associated risk from routine whole-slide images (WSI) and (ii) nominate candidate organizer cells and mechanisms, focusing on SPP1+ M2 macrophages. METHODS: We integrated WSI, bulk transcriptomics, single-cell RNA-seq, spatial transcriptomics, and germline association data. NETs activity was quantified using gene-set scoring and aligned with WSI-derived morphology features (e.g. texture, stromal boundary disruption, and tumor–stroma interface complexity) to train and validate a prognostic model using regularized feature selection and survival machine learning. A 26,928-cell single-cell atlas was constructed to annotate macrophage states and neutrophil subsets, followed by trajectory inference to model macrophage polarization dynamics. Cell–cell communication was evaluated by ligand–receptor co-expression networks, and pathway/metabolic programs were inferred using multi-method enrichment and flux estimation. Spatial co-localization and neighborhood effects were assessed by deconvolution and spatial interaction modeling. Finally, network-based target prioritization was performed to highlight druggable nodes and candidate intervention pathways. RESULTS: The WSI-derived NETs risk score stratified overall survival in training and validation cohorts (HR ≈ 8.48 and ≈6.44; AUC > 0.78). Multi-omics integration prioritized SPP1 as a central hub. SPP1+ M2 macrophages and NETs+ neutrophils preferentially localized at the tumor–stroma interface, where inferred communication converged on OPN(SPP1)–CD44/integrins and ICAM1–β2-integrin axes with downstream FAK–PI3K–Akt and NF-κB/MAPK signaling. Metabolic inference suggested a shared hypoxia/glycolysis–lactate–glutathione/antioxidant program in SPP1+ M2 macrophages mirrored by NETs+ neutrophils, consistent with a NETosis-permissive niche. Germline mapping indicated an antagonistic association pattern for the SPP1+ M2 program. Network prioritization highlighted SRC, AKT1, and CEBPB and pathways including ECM–receptor interaction, FAK/PI3K–Akt, MAPK, and TNF/IL-17. CONCLUSIONS: Our framework links routine pathology morphology to NETs activity and nominates SPP1+ M2 macrophages as candidate organizers of NETosis-prone high-risk niches in HCC. Importantly, the proposed macrophage–NETs axis is supported by convergent multi-omics/spatial evidence but remains primarily associative and inference-based, not definitive causation. Future work should include functional validation (e.g. macrophage–neutrophil co-culture NETosis assays, SPP1/CD44/integrin or ICAM1–ITGB2 perturbation, and in vivo depletion/blockade studies), prospective multi-center evaluation of the WSI risk score, and testing whether this axis generalizes beyond HCC or is liver-specific. These results provide a quantitative pathology-based risk stratification approach and a set of mechanistically plausible, targetable signaling–metabolic nodes for therapeutic exploration.