CXCR4/LITAF as predictive biomarkers for nimotuzumab plus AG therapy in pancreatic cancer with liver metastasis: Integrative single-cell and spatial analysis

Background: Pancreatic cancer with liver metastasis (PCLM) accounts for most advanced pancreatic cancer cases and carries poor prognosis. While Gemcitabine plus Nab-paclitaxel (AG) remains the NCCN-recommended regimen for fit patients, outcomes remain suboptimal. Previous studies suggested that adding Nimotuzumab may improve conversion to resection, but the molecular determinants of response are unclear (2025 AACR# CT167, NCT06405685) . This study integrates clinical and correlative analyses to identify predictive biomarkers for Nimotuzumab + AG therapy. Methods: Treatment-naïve PCLM patients received Nimotuzumab (400 mg iv, qw) plus AG (Gemcitabine 1000 mg/m² and Nab-paclitaxel 125 mg/m², d1, d8, q3w). Primary endpoints were objective response rate (ORR) and disease control rate (DCR); secondary endpoints included conversion and R0 resection rates. In parallel, single-cell RNA-seq data from paired primary and metastatic lesions (GEO) were analyzed to identify PCLM-specific subpopulations via hdWGCNA and machine-learning algorithms (SVM, LASSO, random forest). For clinical validation, immunofluorescence staining of CXCR4 and LITAF was performed on PCLM tumor tissues. Spatial quantification was performed using an automated CK-pan-guided epithelial segmentation. Intensities were normalized and a hotspot-weighted aggregation produced a combined CXCR4/LITAF score per patient, reflecting co-activation. Results: Among 23 evaluable PCLM patients, the ORR was 17.4%, DCR 78.3%, and 43.5% achieved conversion surgery, consistent with previous findings. Single-cell analysis identified 22 cell clusters, including a distinct PCLM-specific subpopulation enriched in hepatic metastases. CXCR4 and LITAF emerged as central hub genes positively correlated with the abundance of these metastatic-specific cells and with poorer clinical outcomes. Imaging-based spatial quantification showed that patients with higher CXCR4/LITAF combined scores exhibited significantly worse treatment responses (progressive disease, PD). The automated pipeline-from epithelial segmentation (CK-pan-guided) to combined molecular scoring-achieved robust reproducibility across samples and accurately stratified responders versus non-responders. Preliminary analyses suggest that KRAS-mutant tumors may exhibit enhanced CXCR4/LITAF co-activation, contributing to therapeutic heterogeneity. Conclusions: Nimotuzumab plus AG demonstrates favorable disease control and conversion potential in PCLM. Integrative single-cell and spatial analyses identify CXCR4 and LITAF as candidate biomarkers linked to treatment response. The CXCR4/LITAF combined spatial score offers a quantitative imaging tool for predicting response to Nimotuzumab + AG in metastatic pancreatic cancer.