Integrative multi-omics analyses unravel the immunological implication and prognostic significance of CXCL12 in breast cancer

Breast cancer accounts for about 30% of female cancers with significant mortality, underscoring the need for improved diagnostic and therapeutic strategies. Not all patients benefit from targeted or immune checkpoint blockade therapies due to primary or acquired resistance, emphasizing the importance of identifying mechanisms and biomarkers that determine treatment response. Chemokines orchestrate leukocyte migration via GPCR and atypical receptors, impacting survival, proliferation, angiogenesis, organogenesis, and cancer progression. CXCL12 (SDF-1), a homeostatic chemokine acting through CXCR4 and ACKRs, regulates hematopoiesis and immune cell trafficking and can also have inflammatory roles. In breast cancer, higher CXCL12 levels have been associated with improved disease-free and overall survival, potentially by limiting metastasis, though its roles remain context-dependent and complex. This study set out to systematically assess CXCL12 expression, its spatial and cellular context in the tumor microenvironment, identify related biomarkers, and construct a prognostic model to improve patient stratification and therapeutic decision-making.

The paper discusses chemokine biology, distinguishing classical GPCR chemokine receptors and atypical chemokine receptors, and their roles in homeostasis and inflammation. CXCL12 (initially PBGF/SDF-1) is essential for embryogenesis, lymphopoiesis, and hematopoietic cell migration. It signals via CXCR4 and atypical receptors (ACKR1/3), and through binding to glycosaminoglycans. The CXCL12–CXCR4 axis retains immune cells in bone marrow, influences tumor-associated macrophage differentiation, and supports CD4+ T cell survival via PI3K/MAPK activation. Prior studies and meta-analyses show that high CXCL12 expression correlates with better OS and DFS in breast cancer, but worsens outcome in other cancers (e.g., pancreatic, ovarian), indicating context dependency. Preclinical data suggest epigenetic silencing of CXCL12 augments metastatic potential, and CXCL12-driven myeloid recruitment contributes to immunosuppression and resistance to immunotherapy in certain breast cancer settings. These findings motivate a comprehensive, multi-omics interrogation of CXCL12’s role in breast cancer.

- Datasets: TCGA breast cancer RNA-seq and clinical data (cBioPortal); GEO datasets GSE42568 and GSE183947 for tumor vs. normal expression; somatic mutation data from GDC (MAF format) analyzed with maftools to compute TMB. Validation cohorts: GSE19615 (n=115), GSE21653 (n=252), GSE61304 (n=58). - Immunohistochemistry: Tissue microarray of 157 breast cancer specimens from Renmin Hospital of Wuhan University; standard deparaffinization, antigen retrieval, peroxidase block, BSA block, CXCL12 primary antibody (Abcam ab9797, 1:200, overnight 4°C), HRP-conjugated secondary, DAB chromogen, hematoxylin counterstain; negative and positive controls included; scoring by two pathologists (percentage positive cells via ImageJ). - WGCNA: Constructed co-expression network from breast cancer expression data to identify modules associated with CXCL12. Soft-threshold chosen to approximate scale-free topology; minimum module size 50; eight modules generated; module–trait correlations identified a CXCL12-associated red module (402 genes). - Prognostic model construction: Intersected gene lists across TCGA and GEO datasets. Univariate Cox regression identified survival-associated genes; LASSO Cox regression selected 11 genes to build the CXCL12-related prognostic signature. Risk score computed as sum of (coefficient × expression) for the 11 genes. Patients split into high/low-risk by median risk score in training (TCGA); performance assessed via time-dependent ROC (survivalROC) and validated in independent cohorts using same coefficients and cutoff. - Functional analyses: DEGs by DESeq2 (adjusted p<0.05, |log2FC|>1). GSEA (clusterProfiler) with FDR<0.05. GSVA for 16 recurrent cancer cell state signatures. - Immune analyses: Immune cell infiltration estimated with TIMER2.0/CIBERSORT (22 immune cell types). Correlated risk scores with immunomodulators and checkpoint genes. - Single-cell analyses: scRNA-seq with paired bulk RNA-seq from 24 breast tumors (GSE176078). QC: cells with <200 genes or >15% mitochondrial genes removed; genes detected in <3 cells excluded. Batch correction via fastMNN (Seurat-Wrappers). PCA/UMAP with top 15 PCs; cluster annotation via canonical markers. Myeloid sub-clustering into monocytes, macrophages, cDCs, pDCs; M2-like macrophage signature scoring. - snRNA-seq and spatial transcriptomics: Sample-paired nuclei isolation and spatial profiling following prior work. Normalization via SCTransform; CAF localization using top 50 CAF marker genes (from snRNA-seq) and ssGSEA; mapping CXCL12 and CAF signatures onto spatial spots. - Drug response and immunotherapy: oncoPredict to estimate IC50 for chemotherapeutics; Wilcoxon tests for high/low-risk comparisons. Immunotherapy validation using IMvigor210 (urothelial carcinoma) cohort to relate risk score to survival under ICB. - Survival modeling: Univariate and multivariate Cox regression for clinical covariates and risk score; construction of nomogram (rms package) for 1-, 3-, 5-year OS; performance via C-index, time-dependent ROC, and calibration plots. - Statistics: R v4.1.3; significance at p≤0.05.

- CXCL12 expression was significantly reduced in breast tumor tissues compared to normal across TCGA and GEO datasets; lower expression associated with older age, higher stage, and varied across PAM50 subtypes (highest in normal-like; lowest in basal-like). - Prognostic value: Higher CXCL12 expression correlated with superior overall survival (OS) and recurrence-free survival (RFS) in the METABRIC cohort; validated by IHC on an independent TMA showing higher CXCL12 linked to better disease-free survival (DFS). - Spatial/cellular context: snRNA-seq and spatial transcriptomics showed CXCL12 predominantly expressed in cancer-associated fibroblasts (CAFs), with spatial co-localization; also detected in subsets of pericytes and endothelial cells. CXCR4 expression was higher in immune cells, implying stromal–immune CXCL12–CXCR4 interactions. - WGCNA identified a CXCL12-associated module (red) containing 402 genes. From these, 11 genes formed a prognostic signature: VAT1L, TMEM92, SDC1, RORB, PCSK9, NRN1, NACAD, JPH3, GJA1, BMP8B, ADAMTS2. - Risk model performance: In TCGA (training), high-risk patients had worse survival; time-dependent AUCs at 1/2/3/5 years were 0.69/0.73/0.73/0.70. In GSE19615 (validation), AUCs were 0.85/0.75/0.80/0.82 with significantly worse OS in high-risk group; additional validation in GSE21653 and GSE61304 confirmed prognostic separation. - Functional programs: High-risk tumors enriched for myogenesis, TNFα/NF-κB signaling, estrogen response; low-risk enriched for E2F targets, G2M checkpoint, MTORC1 signaling, MYC targets. High-risk showed higher cycling, hypoxia, mesenchymal and pEMT cancer cell state scores; low-risk showed higher AC-like and basal states. - Genomics: High-risk group had higher mutation frequencies in TP53, TTN, KMT2C; low-risk had higher CDH1 mutations. TMB was higher in high-risk and positively correlated with the risk score. Combined TMB and risk score improved prognostic stratification. - Immune microenvironment: High-risk tumors had higher M2-like macrophage infiltration and higher inhibitory checkpoint markers (e.g., HAVCR2, PDCD1LG2), while low-risk had more CD8+ T cells and activated NK cells. scRNA-seq confirmed higher myeloid/macrophage fractions and elevated M2-like macrophage signature in high-risk samples. - Drug sensitivity: Risk score correlated with predicted IC50 for multiple agents. Some compounds (e.g., leflunomide, pevonedistat, sabutoclax, telomerase inhibitor IV) correlated positively with risk score; others (e.g., BMS-536924, foretinib, PRT062607) indicated higher sensitivity in high-risk. FDA-approved drugs such as docetaxel, mitoxantrone, palbociclib, vincristine, vinorelbine, and fulvestrant showed higher predicted IC50 in high-risk versus low-risk patients. - Immunotherapy: In the IMvigor210 cohort, higher risk score associated with superior survival under immunotherapy, consistent with higher TMB. - Nomogram: Combining age, tumor stage, metastasis stage, and risk score improved prediction of 1-, 3-, 5-year OS with AUCs of 0.827, 0.827, and 0.781 and good calibration.

The study addressed how CXCL12 expression and its related gene network shape breast cancer prognosis and the tumor immune microenvironment. Despite lower tumor CXCL12 expression compared to normal tissue, higher CXCL12 levels were associated with improved outcomes, aligning with evidence that CXCL12 may reduce metastasis in breast cancer. Spatial and single-cell analyses localized CXCL12 to CAFs and subsets of vascular cells, while its receptor CXCR4 was prominent in immune cells, underscoring stromal–immune crosstalk via the CXCL12–CXCR4 axis. The derived 11-gene CXCL12-related signature stratified patients consistently across cohorts and reflected underlying tumor biology: high-risk tumors exhibited proliferative and mesenchymal programs, increased TMB, and immunosuppressive microenvironments dominated by M2-like macrophages, which likely contribute to poorer outcomes. Conversely, low-risk tumors had more cytotoxic immune infiltration. The risk score also predicted therapeutic responses, including differential sensitivity to chemotherapeutics and potential benefit from immunotherapy when combined with TMB, suggesting clinical utility for treatment selection. Overall, the findings integrate molecular, spatial, and immune features to elucidate the immunological implications of CXCL12 in breast cancer and provide a robust prognostic tool.

This work comprehensively characterizes CXCL12’s role in breast cancer using integrative multi-omics and constructs a validated 11-gene CXCL12-related prognostic signature. The model effectively stratifies patients into distinct risk groups with differences in survival, clinical features, pathway activity, mutational burden, immune infiltration (notably M2-like macrophages), and predicted treatment responses. A combined nomogram incorporating clinical factors and the risk score enhances survival prediction. These results inform biological understanding of CXCL12 in the tumor microenvironment and support using the signature for prognosis and therapeutic decision-making. Future research should include mechanistic in vivo/in vitro validation and prospective multicenter clinical trials to confirm predictive utility and guide personalized treatment.

The study is retrospective and relies on publicly available datasets and computational inferences, which may introduce biases. Mechanistic roles of CXCL12 and the identified genes were not experimentally validated in vivo or in vitro. Immunotherapy response was assessed using an external urothelial carcinoma cohort (IMvigor210) rather than breast cancer immunotherapy datasets, limiting generalizability. Prospective, multicenter trials with larger sample sizes and functional studies are needed to validate and extend these findings.