Purpose <p>Cancer stem cells (CSCs) serve as critical drivers of cancer relapse, metastasis and drug resistance, and are closely associated with poor prognosis. Increasing evidence has highlighted the regulatory effects of CSCs on immune cells such as macrophages in the tumor microenvironment (TME). Therefore, it is imperative to thoroughly study the specific mechanisms by which cancer stemness traits modulate macrophages.</p> Materials and methods <p>Transcriptomic and clinical data of STAD were retrieved from The Cancer Genome Atlas (TCGA). A prognostic Lasso-Cox model was constructed based on CSC-related genes using the R package ‘glmnet’. Pathway enrichment analysis was performed using the ‘clusterProfiler’ package. Composition estimation of infiltrating immune cells in STAD tissues was conducted by CIBERSORTx. Western blotting and flow cytometry were used to detect the expression of CSC-related and macrophage polarization-related markers.</p> Results <p>Through analysis of the TCGA-STAD dataset, 35 CSC-related genes were upregulated in tumor tissues and associated with shorter survival, whereas 4 CSC-related genes were downregulated and correlated with longer survival. These 39 genes were used to construct the prognostic model, from which an optimal 17-gene CSC-related signature was derived and used to stratify patients into high-risk and low-risk groups. The high-risk group was related to poorer prognosis than the low-risk group in both training and testing cohorts. Within this model, CXCR4 exhibited the highest regression coefficient and was significantly associated with poor prognosis in STAD patients. Furthermore, CXCR4 inhibition significantly attenuated CSC-like properties in STAD cells and reduced expression of the CSC markers CD44 and CD24. Immune infiltration analysis revealed that the proportion of M2 macrophages was significantly increased, while M1 macrophages were decreased in the high-risk group. Moreover, CXCR4 expression was positively correlated with hypoxia-inducible factors and glycolysis regulators, and CXCR4 facilitated M2 macrophage polarization and migration by regulating lactate secretion.</p> Conclusion <p>We established and validated a CSC-related prognostic model that was closely associated with macrophage polarization and clinical outcomes in STAD. CXCR4 was identified as the key gene in this model, which unregulated lactate secretion to drive M2 macrophage polarization and maintain CSC properties in STAD, and may serve as a putative regulatory factor in STAD progression. The CXCR4 inhibitor AMD3100 exerted significant anti-tumor effects in vitro, suggesting that CXCR4 may serve as a promising prognostic biomarker and a candidate target for STAD.</p>