RNA-seq profiling of macrophage polarization in THP-1 cells co-cultured with GPNMB-knockdown TNBC spheroids

Metastasis remains the leading cause of cancer-related mortality, driven by complex interactions within the tumor microenvironment (TME). Tumor-associated macrophages (TAMs) play a pivotal role in metastatic progression, yet molecular diversity and upstream regulators remain poorly defined. Glycoprotein nonmetastatic melanoma protein B (GPNMB), overexpressed in subsets of tumors including triple-negative breast cancer (TNBC), is implicated in epithelial-mesenchymal transition (EMT) and cancer stemness. Recent single-cell RNA-seq studies have identified GPNMB as a marker of immunosuppressive TAMs associated with poor prognosis, but its mechanistic role in TAM polarization in TNBC has remained unclear. Co-culturing monocytic cells with three-dimensional sphere-forming TNBC cells induces their conversion into GPNMB?Siglec-9? tumor-associated macrophages (TAMs). Tumor-derived GPNMB promotes monocyte-to-TAM polarization by inducing secondary GPNMB expression in monocytes, establishing a feed-forward amplification loop. Knockdown of GPNMB in TNBC cells significantly inhibits multiple immunosuppressive TAM subtypes, including Siglec-9? TAM and EMT-associated TAM populations, as inferred from scRNA-seq and validated in patient tumors using bulk RNA-seq deconvolution. Distinct sialylation patterns were identified: tumor-derived GPNMB exhibited a2,3-sialylation, whereas macrophage-derived GPNMB exhibited a2,6-sialylation, enabling differential Siglec-9 recognition. Elevated GPNMB and Siglec-9 expression correlated with poor prognosis in TNBC patient cohorts. Importantly, dual inhibition of Siglec-E (murine Siglec-9 ortholog) and PD-1 suppressed IL-6-dependent EMT, reduced tumor stemness, and significantly limited lung metastasis in vivo. The GPNMB–Siglec-9 axis thus represents a critical glyco-immunological checkpoint driving TAM-mediated metastasis, providing a promising therapeutic target in TNBC. Overall design: To investigate the role of GPNMB expressed in TNBCs in macrophage M1/M2 polarization, we established a 3D co-culture system with TNBC spheres and performed RNA sequencing analysis. TNBC cells (Hs578T) were treated with GPNMB and control siRNAs 24 hours prior to sphere formation. Three days later, monocyte-like cells (THP-1) were added directly to the culture of TNBC cell spheres. The optimal co-culture duration was determined to be three days. On the harvest day, the entire co-culture medium was passed through a cell filter to separate the cancer spheres, which accumulated on the filter cap, allowing for the collection and analysis of RNA from the macrophages exclusively. We conducted a comparative gene expression profiling analysis of RNA-seq data from three distinct cell conditions: (1) "THP-1 control," co-cultured with Hs578T cells transfected with control siRNA; (2) "THP-1 siRNA1," co-cultured with Hs578T cells with GPNMB knockdown using siRNA1; and (3) "THP-1 siRNA2," co-cultured with Hs578T cells with GPNMB knockdown using siRNA2. All data were normalized against the gene expression profile of THP-1 cells cultured alone, without co-culture. We reanalyzed a subset of single-cell RNA-seq samples (GSM5188373, GSM5188379, GSM5188386, GSM5188388, GSM5188393) from GSE169246, which represent tumor-derived myeloid cell populations from triple-negative breast cancer (TNBC) patients. The original libraries were generated using the 10x Genomics Chromium Single Cell 5? platform and processed with Cell Ranger v3.0.0. Raw UMI count matrices were imported into Seurat v4.3 (R v4.1.2), and quality filtering was applied to exclude cells with fewer than 400 or more than 8,000 detected genes, or with >10% mitochondrial gene content. Normalization was performed using SCTransform, followed by principal component analysis (PCA) and batch correction using Harmony v1.0. Myeloid cell subsets were identified based on canonical macrophage markers (CD68, CD14, CSF1R, MERTK, CD163, HLA-DRA, VCAN, LYZ), and tumor-associated macrophages (TAMs) were further subclassified using a curated 8-subtype framework including IFN-responsive, regulatory, inflammatory, lipid-associated, angiogenic, resident-like, proliferative, and GPNMB?/SIGLEC9? TAM populations. Cluster-level marker genes were identified using the Wilcoxon rank-sum test, and highly specific, non-overlapping markers were selected to generate a custom TAM signature matrix for digital cytometry.