Integrated Single-Cell Analysis Identifies IL1RAP as a Master Regulator of TAMs and a Prognostic Biomarker in Breast Cancer

TL;DR

This study identifies IL1RAP as a key regulator of tumor-associated macrophages in breast cancer, suggesting it could be a new target for cancer therapy.

Contribution

The paper introduces IL1RAP as a novel regulator of macrophage recruitment and M2 polarization in breast cancer.

Findings

IL1RAP expression in macrophages correlates with immune suppression and poor prognosis in breast cancer.

IL1RAP deficiency reduces macrophage migration and M2 polarization, as shown by decreased M2 markers like Arg1 and Mrc1.

Abstract

The recruitment and polarization of tumor-associated macrophages (TAMs) play a pivotal role in shaping the immunosuppressive tumor microenvironment in breast cancer. Interleukin-1 receptor accessory protein (IL1RAP), a critical co-receptor for IL-1 family cytokines, is emerging as a potential regulator of macrophage function, though its specific role in TAM biology remains to be explained. In this study, we investigated the impact of IL1RAP on macrophage recruitment and M2-like polarization. Initial bioinformatics analysis of public databases revealed a significant correlation between elevated IL1RAP expression in macrophages and signatures of immune suppression and poor prognosis in breast cancer. To functionally validate these findings, we performed IL1RAP knockdown in a murine macrophage cell line. Our results demonstrated that IL1RAP deficiency markedly impaired the migratory capacity of macrophages towards classic chemotactic stimuli. Furthermore, under M2-polarizing conditions, IL1RAP-knockdown macrophages exhibited a significantly attenuated M2 phenotype, as evidenced by the decreased expression of canonical M2 markers (e.g., Arg1, Mrc1) and reduced functional outputs. Collectively, our integrated approach combining bioinformatics and in vitro experimentation identifies IL1RAP as a novel regulator that potentiates both the recruitment and the M2 polarization of macrophages. These findings suggest that targeting the IL1RAP pathway could represent a promising therapeutic strategy for reprogramming the tumor-immune microenvironment by limiting pro-tumoral macrophage infiltration and polarization.

Figures

Full text

1. Introduction

Breast cancer remains the most frequently diagnosed malignant tumor and a great cause of cancer-related deaths in women worldwide [1]. Although significant improvement has been made in targeted therapy and early detection [2], tumor recurrence, metastasis and therapeutic resistance pose persistent clinical challenges, especially in invasive subtypes such as triple-negative breast cancer [3]. The biological complexity and clinical heterogeneity of breast tumors are not only determined by cancer cells, but also deeply affected by the tumor microenvironment (TME) [4]. In this complex ecosystem, the continuous interference of different immune cell groups with tumor cells was a decisive factor promoting tumor malignancy, immune escape and therapeutic resistance [5]. Therefore, a deep insight to the mechanisms of key immune regulators in the TME is crucial to identify new therapeutic vulnerabilities.

The TAM is one of the most abundant and functionally plastic immune cells in the breast TME [6], and patients with high density usually have poor prognosis and faced with therapeutic resistance [7]. TAMs mainly come from circulating mononuclear cells recruited by tumor-derived chemokines (such as CSF1), and are educated by local signals to adapt to different functional states [8]. Although the classic M1/M2 dichotomy provides a basic framework, it is increasingly seen as an oversimplification of the complex and continuous spectrum of macrophage phenotypes observed in human cancer [9]. In established breast tumors, TAMs show an M2-like, pro-tumoral polarization, facilitating tumor growth, angiogenesis, extracellular matrix remodeling, and the suppression of adaptive immunity [10]. This polarization is based on metabolic reprogramming, engaging metabolic pathways such as arginine metabolism to sustain their immunosuppressive effector functions [11]. This critical role makes the molecular drivers of macrophage recruitment and polarization attractive therapeutic targets [12]. Given the strong association of TAM density and poor clinical prognosis in breast cancer [13], accurate molecular mechanisms controlling its recruitment and polarization are crucial for the development of new immunomodulation strategies [14]. Notably, the IL-1 cytokines are important molecules in immunity and inflammation response [9], and are powerful regulators of immune cell function. The signaling of this family widely requires the common co-receptor IL1RAP for multiple members (including IL-1, IL-33, and IL-36) [15], presenting a promising but underexplored node that could potentially orchestrate both the recruitment and phenotypic skewing of TAMs in the breast TME.

However, while the mechanisms driving TAM recruitment and polarization are gradually unraveled, the exact role of the IL-1 family signaling pathway in shaping TAM function in the breast TME is still underdefined [16]. Considering its important role in the TME, the expression of IL1RAP and its mediated signaling are particularly active in macrophages [17], indicating that it may be a key node in regulating macrophage behavior. Moreover, in the breast TME, continuous IL-1 family signals are presumed to be key factors in maintaining tumor-promoting inflammation and driving immunosuppressive networks [18]. For example, the IL-33/ST2 axis is associated with the promoted progression of breast cancer [19]. Yet, it is unknown whether and how IL1RAP specifically governs the recruitment and M2 polarization of macrophages to reshape the breast TME. Therefore, clarifying the function of IL1RAP in breast cancer TAMs would not only provide a novel perspective on TAM regulation but may also reveal a powerful therapeutic target which can reprogram the immunosuppressive TME and potentially improve the therapeutic efficacy of immunotherapies.

In order to specifically investigate this knowledge gap, we conducted an integrated study combining bioinformatic analyses with experimental validation. First of all, we used transcriptome data obtained from the public cohort to systematically determine the relationship between the IL1RAP expression level and myeloid cell dynamics in breast cancer, with a focus on TNBC, and scRNA sequence data to analyze its cell type specific correlation. Secondly, to establish direct functional causality, we employed an in vitro system with iBMM murine macrophage cell lines with stable IL1RAP knockdown. This approach allowed us to definitively test the hypothesis that IL1RAP expression intrinsically regulates the key pro-tumoral functions of macrophages, including their recruitment capacity, propensity for M2-like polarization, and subsequent tumor-promoting activities. Our results demonstrate that IL1RAP is a significant regulator of these processes, providing novel mechanistic insight into how IL-1 family signaling shapes the immunosuppressive myeloid compartment in breast cancer.

2. Results

2.1. Single-Cell Transcriptomic Profiling of Primary Breast Cancer Reveals Expression Patterns of the IL-1 Receptor Family

To systematically characterize the TME of primary breast cancer and investigate the expression landscape of the interleukin-1 receptor (IL-1R) family, we integrated and analyzed single-cell RNA sequencing (scRNA-seq) data from public cohorts. The unsupervised clustering of all cells produced a comprehensive cellular atlas, and they are visualized independently using both t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) (Figure 1A). We identified the major cell subsets in the TME, including epithelial cells, diverse immune cells, and stromal components. The identity of each cluster was rigorously annotated by visualizing the expression distribution of canonical marker genes: EPCAM for epithelial cells, NKG7 for NK/T cells, LYZ for myeloid cells, CD79A for B cells, DCN for fibroblasts, and CLDN5 for endothelial cells (Figure 1B). The differentially expressed genes were exhibited by volcano plot and highlighted the most significantly variable transcripts, underscoring the distinct molecular signatures defining each compartment (Figure 1C). Subsequently, we generated a gene expression heatmap using the SCP package, coupled with enrichment analyses (GO_BP, KEGG, and WikiPathways) for each cell type, providing a detailed functional annotation of the identified populations (Figure 1D).

Next, we specifically studied the expression pattern of the IL-1 receptor family in this cellular landscape. The scRNA-seq analysis revealed the unique expression patterns of different cell types of the family members. In particular, we observed that only IL1R2 and IL1RAP exhibit particularly high expression levels in myeloid cells (Figure 1E). To assess the potential clinical relevance of these two receptors, we conducted survival analysis using the KM-Plotter database. Notably, high expression levels of IL1RAP were significantly related to poorer patient prognosis in breast cancer cohorts, suggesting a critical role for this signaling axis in tumor progression, while the expression of IL1R2 has no impact on the prognosis of breast cancer patients (Figure 1F). This integrated analysis establishes a foundational map of the breast cancer TME and determined that IL1RAP is a factor of interest associated with specific cellular subsets and unfavorable clinical outcomes.

2.2. IL1RAP Expression Delineates a Distinct Functional and Metabolic State in Breast Cancer Myeloid Cells

In order to investigate the specific role of IL1RAP in the TME, we performed high-resolution sub-clustering, which defined more granular cell subtypes (celltype_minor), as visualized in both t-SNE and UMAP embeddings (Figure 2A). The expression of IL1RAP was then mapped across this refined cellular landscape. Spatial visualization on the t-SNE plot confirmed its specific expression pattern (Figure 2B), which was further quantified using bubble plots. These analyses consistently demonstrated that IL1RAP expression was predominantly restricted to and highly enriched within the myeloid cell compartment, with negligible expression in lymphoid or other stromal populations (Figure 2C,D).

To functionally characterize the impact of IL1RAP expression, we stratified myeloid cells into IL1RAP-positive (IL1RAP+) and IL1RAP-negative (IL1RAP−) subsets. GO_BP enrichment analysis of differentially expressed genes between these two groups revealed significant differences. Pathways associated with immune response activation, cytokine-mediated signaling, and cell migration were significantly enriched in the IL1RAP+ myeloid subset, underscoring a more immunomodulatory and active phenotype (Figure 2E). Furthermore, we applied Gene Set Variation Analysis (GSVA) across all major cell types and highlighted a broad spectrum of pathway activity differences. The results show that IL1RAP+ myeloid cells display a unique enrichment profile compared to other cells in the TME (Figure 2F).

Generally, our focused investigation into metabolic reprogramming through a hallmark of both cancer cells and immune cells in the TME revealed profound differences [20]. The heatmap analysis of specific metabolic pathways showed that IL1RAP+ myeloid cells possess a distinct metabolic signature (Figure 2G). This signature included enrichment in pathways such as glycolysis and lipid metabolism (Figure 2H), as well as amino acid and nucleotide metabolism (Figure 2I). Collectively, these data indicate that IL1RAP expression is a marker for myeloid lineage and defines a specific myeloid subpopulation, with a unique, pro-tumoral functional phenotype. This population is characterized by altered immunoregulatory capacity and a broad rewiring of the core metabolic pathways.

2.3. IL1RAP Expression Correlates with an M2-Polarized Phenotype and Enhanced Pro-Tumorigenic Cell–Cell Communication

In order to evaluate the broader immunomodulatory impact of IL1RAP, we first analyzed bulk transcriptomic data from the TCGA-BRCA cohort. Patients with high IL1RAP expression exhibited significantly elevated levels of multiple immune checkpoint genes, indicating an association with a more immunosuppressive TME (Figure 3A). Given the central role of macrophages in immune regulation [21], we next investigated the relationship between IL1RAP and macrophage polarization states. Analysis using the CIBERSORTx-ABS database displayed a stronger correlation between IL1RAP abundance and M2 macrophage scores than M1 macrophage across samples, indicating that IL1RAP+ macrophages tend to polarize towards the M2 phenotype (Figure 3B).

In order to validate this association at single-cell resolution, we compared the polarization scores between IL1RAP+ and IL1RAP− myeloid cells within our scRNA-seq dataset. Consistent with the bulk analysis, the M2 score of IL1RAP+ myeloid cells is significantly higher than that of IL1RAP−, while the M1 score shows limited difference (Figure 3C). This inclined polarization has been further confirmed at the gene expression level. The bubble plot visualization confirmed that canonical M2-associated genes were predominantly upregulated in the IL1RAP+ subset, while key M1 markers were not differently expressed in the two subsets (Figure 3D).

We then employed CellChat analysis to infer how IL1RAP+ myeloid subsets influence intercellular signaling networks. Reconstructed cell–cell communication networks revealed that IL1RAP+ myeloid cells were among the most active cell populations, engaging in both more and stronger intensity of interactions compared to other subsets in the TME (Figure 3E). Moreover, differential interaction analysis quantified that IL1RAP+ myeloid cells especially showed enhanced communication with diverse cell types, including cancer cells and endothelial cells (Figure 3F,G). Furthermore, the examination of outgoing and incoming signaling pathways highlighted that IL1RAP+ myeloid cells were particularly prominent in sending signals via pathways like VEGF, GALECTIN, and MK. These key oncogenic and immunomodulatory pathways position IL1RAP+ myeloid cells as central signaling hubs in the tumor ecosystem (Figure 3H). Collectively, these results demonstrate that IL1RAP expression marks a myeloid subpopulation with a strong M2-polarized phenotype and a heightened capacity to drive pro-tumorigenic cross-talk within the breast cancer microenvironment.

2.4. Il1rap Knockdown Impairs the Migratory Capacity and M2 Polarization Potential of Macrophages In Vitro

To directly investigate the functional consequences of IL1RAP loss, we established a stable Il1rap-knockdown model using the iBMM murine macrophage cell line via lentiviral-mediated shRNA (shIl1rap), with a scrambled shRNA (shSCR) serving as the control. Transcriptome sequencing of shIl1rap and shSCR cells was performed to uncover globally altered pathways. GO enrichment analysis of differentially expressed genes exhibited significant downregulation of biological processes related to cell motility and response to chemokine (Figure 4A). Correspondingly, KEGG pathway analysis highlighted the suppression of pathways associated with cytokine–cytokine receptor interaction and cell adhesion (Figure 4B). GSEA further confirmed these findings, showing negative enrichment for terms such as “myelination” (GO-based) and “nitrogen metabolism” (KEGG-based) in shIl1rap cells (Figure 4C,D), corroborating a broad disruption of activation and metabolic states. The spectrum of genes significantly altered following IL1RAP knockdown were shown through the volcano plot (Figure 4E).

Guided by the transcriptomic prediction of impaired motility, we functionally validated the migratory capacity of iBMM using a Transwell assay. After knockdown of Il1rap, the number of iBMM cells migrating towards Py8119 cells in the lower chamber was significantly reduced compared to the shSCR (Figure 4F). We next tested the impact on M2-like polarization. Western blot analysis confirmed efficient Il1rap protein knockdown and demonstrated a concomitant decrease in the expression level of Arginase-1 (Arg1), a canonical M2 marker (Figure 4G). This reduction in M2 polarization was further validated at the transcriptional level; qRT-PCR analysis showed that shIl1rap cells exhibited significantly lower expression levels of critical marker genes of M2 macrophages (including Arg1, Mrc1, Mgl2, Irf4, Fizz1, and Ym1) [22] compared to control cells (Figure 4H). Finally, flow cytometric analysis quantitatively confirmed that the proportions of Arg1-positive and Mrc1 (CD206)-positive cells were markedly diminished in the shIl1rap population (Figure 4I). Collectively, these integrated multi-omics and functional data establish that Il1rap is intrinsically required for both the migratory ability and the M2 polarization program of macrophages.

2.5. Co-Culture with Il1rap-Knockdown Macrophages Attenuates the Pro-Tumorigenic Capacities of Breast Cancer Cells

To determine whether Il1rap-mediated regulation of macrophage function consequentially impacts tumor cell behavior, we performed an in vitro co-culture system. iBMM shSCR or shIl1rap macrophages were co-cultured with murine Py8119 breast cancer cells for 96 h, after which the tumor cells were sorted for subsequent functional analyses (see Section 4).

The tumor-promoting capacity of the conditioned cancer cells was first assessed by clonogenic assay. Py8119 cells co-cultured with shIl1rap macrophages formed significantly fewer and smaller colonies compared to those co-cultured with control shSCR macrophages, indicating a reduction in the long-term proliferative potential and clonogenicity (Figure 5A). This compromised growth was further corroborated by a CCK-8 assay, which showed a decreased proliferative rate in Py8119 cells following co-culture with Il1rap-deficient macrophages (Figure 5B).

We next evaluated the effect on tumor cell motility and invasiveness, which are key drivers of metastasis. Transwell assays demonstrated that the migratory (Figure 5C) and invasive (Figure 5D) capacities of Py8119 cells were substantially impaired after co-culture with shIl1rap macrophages relative to the control. This finding correlates with a scratch wound healing assay, where Py8119 cells conditioned by shIl1rap macrophages exhibited a markedly slower rate of gap closure (Figure 5E).

Given the association between stem-like properties, therapy resistance, and aggressive disease, we examined the expression of core stemness genes in the co-cultured Py8119 cells. qRT-PCR analysis revealed a significant downregulation of key stemness markers (e.g., Sox2, Nanog) in tumor cells that had interacted with Il1rap-knockdown macrophages (Figure 5F).

In summary, these data demonstrate that the intrinsic loss of Il1rap in macrophages not only alters their own phenotype but also functionally impairs their ability to support critical pro-tumorigenic properties in neighboring breast cancer cells, including proliferation, motility, invasion, and the maintenance of stem-like characteristics.

2.6. A Prognostic Model Based on IL1RAP-Associated Features Demonstrates Robust Predictive Power and Clinical Applicability

To translate our biological findings into a clinically relevant tool, we developed and validated a prognostic signature based on genes co-expressed with IL1RAP in the breast cancer microenvironment. Using the TCGA-BRCA cohort, we applied the least absolute shrinkage and selection operator (LASSO) Cox regression algorithm to identify the most predictive features from an initial gene set, visualizing the coefficient profiles (Figure 6A) and performing cross-validation to determine the optimal penalization parameter (lambda) that minimized the partial likelihood deviance (Figure 6B).

The predictive performance of the resulting risk model was rigorously evaluated. Time-dependent receiver operating characteristic (ROC) curves demonstrated excellent predictive accuracy for patient survival, with high area under the curve (AUC) values in the entire TCGA cohort (Figure 6C), as well as in the randomly split training set (Figure 6D) and independent testing set (Figure 6E). The model’s robustness was then confirmed in the external validation cohort GSE20685, maintaining strong prognostic discrimination (Figure 6F).

For potential clinical application, we constructed a nomogram that integrates the gene signature-based risk score with key clinical parameters (e.g., age, stage) to provide an individualized probability estimate of survival (Figure 6G). The calibration curve of this nomogram showed close agreement between the predicted and observed survival outcomes, indicating good reliability (Figure 6H). ROC analysis confirmed that the combined nomogram offered superior predictive power compared to using clinical features alone (Figure 6I). The concordance index (C-index) was calculated to quantify the model’s discriminative ability, and a bar plot illustrated the contribution of each constituent gene to the overall risk score (Figure 6J). Furthermore, the analysis of restricted mean survival time (RMST) revealed a clear, graded decrease in survival expectancy as the risk score percentile increased, underscoring the model’s continuous predictive value (Figure 6K).

Finally, to benchmark our model against existing prognostic signatures in breast cancer, we conducted a comprehensive comparative analysis. We evaluated the performance of six previously published signatures (Chen, Hou, Li, Tian, Zhong, and Zhu) alongside our model. Comparative ROC curve analysis at critical time points (e.g., 3, 5 years) consistently showed that our IL1RAP-associated signature achieved higher or comparable AUC values (Figure 7A–F). More importantly, Kaplan–Meier survival analysis based on each model’s risk stratification demonstrated that our signature effectively identified patient groups with significantly distinct clinical outcomes, with performance metrics meeting or exceeding those of the established models (Figure 7G–L). This comparative validation establishes our IL1RAP-centric model as a robust and competitive tool for prognostic assessment in breast cancer.

3. Discussion

In this study, our investigation outlines the pivotal role of IL1RAP in orchestrating the pro-tumoral functions of TAMs in breast cancer. We found that IL1RAP is highly enriched in specific myeloid subsets in TME. Moreover, IL1RAP expression correlates with poor patient prognosis and is related to a distinct cellular state characterized by M2 polarization and metabolic reprogramming. Genetic ablation of IL1RAP impaired macrophage migration and M2 polarization, and also reduced the capacity of macrophages to enhance the proliferative, invasive, and stem-like properties of co-cultured breast cancer cells. In addition, an IL1RAP-associated gene signature showed reliable prognostic value. Overall, these findings position IL1RAP as a critical regulatory node linking myeloid cell functionality to tumor progression, with clear translational implications.

Our work directly solves the significant gap in understanding the specific contribution of IL-1 family signaling in breast cancer myeloid biology. Although individual ligands such as IL-1β or IL-33 have been implicated in cancer-associated inflammation [23,24], the function of the common co-receptor IL1RAP was still unexplored. The bioinformatic predictions along with the experimental validation of impaired Transwell migration indicates that IL1RAP is an active driver of macrophage recruitment. This aligns with the recognized role of IL-1 signaling in promoting chemokine secretion and integrin activation [25], suggesting that IL1RAP may act as a central amplifier for such pathways in the TME.

Another important part of our study is the metabolic reprogramming characteristic of IL1RAP expressing myeloid cells. Our single-cell data show significant enrichment in pathways like glycolysis, lipid metabolism, and arginine metabolism in this population. This metabolic signature is highly consistent with an M2-like, pro-tumoral activation state [26]. Enhanced glycolysis supports the activation of macrophages’ biological energy and biosynthetic needs [27]. Also, the altered lipid metabolism promotes the production of immunosuppressive mediators like prostaglandins [28]. The shift in arginine metabolism is a hallmark of M2 polarization and directly fuels tumor cell proliferation and tissue remodeling [29]. Our in vitro finding that IL1RAP knockdown reduces Arg1 expression correlates with this critical metabolic switch. We propose that IL1RAP signaling may coordinate a broad metabolic redistributing to stabilize the M2 phenotype and enhance the tumor-promoting capacity of TAMs.

The clinical relevance of our study is displayed by the association between elevated IL1RAP expression and poor survival outcomes across independent datasets. In order to translate the biological insight into a clinically applicable tool, we developed and validated a prognostic signature based on IL1RAP-associated genes. This model demonstrated excellent predictive accuracy for patient survival in the training set, an independent internal test set, and an external validation cohort (GSE20685). Its clinical utility is highlighted by a well-calibrated nomogram which integrates the risk score with key clinical parameters to assess individualized survival probability. Direct comparison with six previously published breast cancer prognostic signatures (Chen, Hou, Li, Tian, Zhong, Zhu) further established that our model performs comparably or superiorly in both ROC analysis and survival stratification. These analyses collectively indicate that IL1RAP not only represents a promising therapeutic target but also that its associated gene expression profile constitutes a robust and independently validated prognostic biomarker, which could potentially aid in future risk stratification and the personalized management of breast cancer patients.

Several limitations of the present study warrant acknowledgment and guide future directions. First, while our in vitro models establish causality, the precise downstream signaling axis (e.g., whether primarily mediated by IL-1β, IL-33, or IL-36) and the upstream regulators of IL1RAP in TAMs remain to be fully elucidated. Second, the in vivo relevance of our co-culture findings requires validation in immunocompetent murine breast cancer models to account for the full complexity of the tumor-immune ecosystem.

4. Methods and Materials

4.1. Data Obtaining and Single-Cell RNA Sequencing Data Processing

The TCGAbiolinks R package 5 was used to obtain bulk RNA sequencing data and corresponding clinical information for the TCGA-BRCA cohort from The Cancer Genome Atlas, ensuring both expression profiles and complete survival data are available. Publicly available single-cell RNA sequencing (scRNA-seq) data were sourced from the GEO dataset GSE176078 [30]. The Seurat R package (v5.3.1) was used to perform data processing [31,32,33,34]. The NormalizeData function was applied for data normalization, and FindVariableFeatures was used to identify highly variable features (nfeatures = 3000). Principal component analysis (PCA) was used to achieve dimensionality reduction. The t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) algorithms were employed for the visualization of cell clusters [35]. The expressions of canonical marker genes were checked to annotate cell clusters. FindAllMarkers function was used to conduct differential expression analyses in order to identify the marker genes of specific cell types.

4.2. Functional Enrichment Analysis

The SCP R package (v0.5.6) was used to explore the biological functions of specific cell subsets, by performing Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and WikiPathways enrichment analyses. Gene Set Variation Analysis (GSVA, v2.0.5) was utilized to evaluate the enrichment of MSigDB Hallmark gene sets across different cell subpopulations [36]. Metabolic features of the subpopulations were characterized using the scMetabolism package (v0.2.1) [37,38].

4.3. Analysis of Intercellular Communication

Cell–cell communication between different cell types was investigated using the CellChat R package (v1.6.1) and its built-in database CellChatDB.human [39].

4.4. Immune Infiltration and Correlation Analysi

The polarization state of macrophages was assessed by calculating M1 and M2 signature scores based on gene sets derived from the published literature [40]. Immune infiltration scores were quantified using CIBERSORT [41,42] and CIBERSORT-ABS [43]. The association between IL1RAP expression levels and immune cell infiltration was examined through Pearson correlation analysis. Additionally, 19 inhibitory immune checkpoints with therapeutic potential were collected from prior studies for comparison between high-risk and low-risk patient groups.

4.5. Construction of a Prognostic Signature

Differentially expressed genes (DEGs) of IL1RAP+ and IL1RAP− myeloid cell subsets were identified using DESeq2 package (v1.46.0) via a pseudobulk approach [44]. The TCGA-BRCA cohort was split and presented as a training set and an internal testing set at random (1:1 ratio). In the training set, univariate Cox regression analysis was first used to filter the genes (p < 0.05), followed by variable selection using the least absolute shrinkage and selection operator (LASSO) regression. An optimal prognostic signature was subsequently constructed via a multivariate Cox proportional hazards model with bidirectional stepwise selection. The glmnet package (v4.1-8) was applied to define the critical genes and corresponding regression coefficients [45]. The formula Risk Score = Σ (Coefficient_i_ × Expression_i_) was used to calculate the risk score for each patient, where “i” represents each signature gene. Patients were stratified into high-risk and low-risk groups using the median risk score from the training set as the cutoff. The survminer (v0.5.0) and survivalROC (v1.0.3.1) packages were used to generate Kaplan–Meier survival analysis and time-dependent receiver operating characteristic (ROC) curves. The concordance index (C-index) was computed with the survcomp package (v1.52.0). A nomogram integrating clinical features and the risk score was constructed using the rms package (v8.1-0) to predict 1-, 3-, and 5-year overall survival probabilities for TCGA-BRCA patients.

4.6. Cell Lines

The HEK293T, macrophage cell line iBMM and murine breast cancer cell line Py8119 were purchased from the American Type Culture Collection (ATCC) (Manassas, VA, USA). STR authentication confirmed cell line identities. All cells are regularly tested for being mycoplasma-free. Py8119, iBMM and HEK293T cells were all cultured in DMEM medium (Gibco, New York, NY, USA, cat# C11995500BT) with 10% FBS and 1% penicillin-streptomycin. All the cells were cultured at 37 °C in a 5% CO_2_ incubator.

4.7. Plasmid Construction and Lentivirus Infection

The sequences of IL1RAP shRNA were downloaded on the Sigma-Aldrich shRNA website. The five core lentiviral vectors were cloned into the pLKO.1 mcherry lentiviral vector. After that, the core vectors with the packaging plasmids psPAX2 (Addgene, Cambridge, MA, USA, cat#12260) and pMD2.G (Addgene, Cambridge, MA, USA, cat#12259) were introduced to HEK293T cells at a ratio of 5:3:2 to produce lentiviruses. After two days of transfection, the 0.45 μm filters were applied to filter the collected lentiviruses in the supernatants. The lentiviruses were then added to tumor cells, accompanied with polybrene (10 μg/mL) (Beyotime, Shanghai, China, cat#C0351) for 12 h. The infected cells were selected 3 times for mcherry positive subsets using a MoFlo Astrios (Beckman Coulter, Brea, CA, USA). The plasmid construction primers are listed in Table S1.

4.8. Colony Formation Assay

After 4 days of co-culture, Py8119 cells (300 cells) were sorted and seeded in 6-well plates and then cultured for 10 days. The plate with colonies were fixed by methanol for 30 min, and then stained with 0.1% crystal violet (Sangon Biotech, Shanghai, China, cat#A600331). The numbers of colonies were counted through the entire field.

4.9. Transwell Migration and iBMM Recruitment Assays

A measurement of 200 μL serum-free DMEM was used to suspend Py8119 cells after co-culture (5 × 10^4^ cells/well). The cells were then seeded to the upper chamber with an 8 μm-pore size membrane (Corning, NY, USA, cat#353097) in a 24-well plate. A measurement of 600 μL DMEM medium with 20% FBS fill the lower chamber of the plate. For the test of iBMM recruitment assay, Py8119 (5 × 10^4^ cells/well) was first seeded in the lower chamber in DMEM with FBS. After 24 h, the iBMMs (5 × 10^4^ cells/well) were then seeded on the upper chamber in serum-free DMEM. After 16 h of the migration, non-migrated Py8119 or iBMM that remained at the upper side of the chamber were cleaned with a cotton swab. Migrated cells at the lower surface of the chamber were fixed and stained as described, and then photographed and counted.

4.10. Quantitative RT-qPCR

TRIzol (Invitrogen, Carlsbad, MA, USA, cat# 15596018CN) was used to extract total RNA. A Nanodrop instrument was used to quantify the samples. The RNA samples were then reverse transcribed into cDNA using the PrimeScript RT reagent Kit with gDNA Eraser (Takara, Kusatsu, Japan, cat#RR047A). qRT-PCR was conducted using TB Green Advantage qPCR Premix (Takara, Kusatsu, Japan, cat#639676) on the Real-Time PCR System (ThermoFisher, Waltham, MA, USA). Specific primer pairs were used to recognize the targeted genes. The primers used above are listed in Table S2. The mRNA relative expression level was normalized to the TBP, and comparative Ct method (2^ΔΔCT^) was used to get quantitative measures.

4.11. RNA Sequencing

iBMMs (5 × 10^6^) were collected with TRIzol reagent for RNA sequencing (RNA-Seq). RNA-seq libraries were prepared using the VAHTSUniversalV8 RNA-seq Library Prep Kit for Illumina and subjected to quality control using a Qubit4.0 (Thermofisher, Waltham, MA, USA), and then a Novaseq (Illumina, San Diego, CA, USA) was used to sequence. KEGG and GO databases were used to conduct enrichment pathway analysis. The RNA-seq data of iBMM-shSCR and iBMM-shIL1RAP have been deposited at NCBI and can be searched by the accession codes PRJNA1403188.

4.12. Western Blotting

The RIPA buffer (Sangon Biotech, Shanghai, China, cat#C500005-0100) was used to extract the total protein of iBMMs, supplemented with Halt^TM^ Protease and Phosphatase Inhibitor Cocktail (ThermoFisher, Waltham, MA, USA). The protein concentrations of the samples were determined by the BCA protein assay kit (Solarbio, Beijing, China, cat#PC0020) and then normalized with 5× Native Gel Sample Loading Buffer (NCM biotech, Suzhou, China). After being boiled for 10 min at 100 °C, average amounts of protein were separated using 10% SDS-PAGE gel electrophoresis and transferred to PVDF membrane in Trans-Blot Turbo transfer system (Bio-Rad, Hercules, CA, USA), and then blocked with PBS containing 5% non-fat milk powder and 0.1% Tween (PBST). After being blocked, the primary antibodies were used to incubate the membranes at 4 °C overnight. After being washed by PBST, the membranes were incubated with HRP-conjugated secondary antibodies (Anti-Rabbit; Abcam, Cambridge, UK) at room temperature for 1 h, and then washed by PBST 3 times again. Membranes were then incubated with Pierce ECL Western Blotting Substrate (ThermoFisher, Waltham, MA, USA) and imaged using ChemiDoc^TM^ XRS+ System (Bio-Rad, Hercules, CA, USA). The antibodies used are listed as below: IL1RAP (IIL1RAP PTG, Rosemont, IL, USA, cat#30966-1-AP, 1:100), Arginase-1 (ARG1, CST, Danvers, MA, USA, cat#93668S, 1:1000).

4.13. Tumor Cells and iBMMs Co-Culture

The co-culture system plated tumor cells and iBMMs in a 10 cm plate at a ratio of 4:1. After 4 d, the tumor cells were selected by flow cytometry and then lysed to conduct subsequent analysis.

4.14. Flow Cytometry

iBMMs were collected and incubated for 20 min with Rat Anti-Mouse CD16/CD32 antibody (BD Biosciences, San Jose, CA, USA) to block the fc receptors, and then treated with antibody against Arg1 overnight at 4 °C. After washing with PBS, samples were incubated with AF647-labeled Goat Anti-Rabbit IgG (Beyotime, Shanghai, China, cat#A0468) and anti-CD206-FITC (Biolegend, San Diego, CA, USA, cat# 141703) for 30 min. Beckman Coulter CytoFlex S system was used to analyze the stained samples.

4.15. Statistical Analysis

All bioinformatic statistical analyses were conducted with R software (v4.4.1). Comparisons between two groups were conducted using the Wilcoxon rank-sum test or Student’s t-test, as appropriate. The differences in survival were analyzed using the log-rank test. A two-sided p-value < 0.05 was considered statistically significant.

All of the in vitro experiments were independently repeated three times. GraphPad Prism V9.0 software was used to conduct statistical analyses. Comparisons between two groups were conducted using unpaired Student’s t-tests or Mann–Whitney tests. Statistical significance is defined as below: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001; ns, no significance.