Abstract Background Prostate cancer (PCa) is characterized by high incidence and recurrence rates, presenting as an immune ‘cold’ tumor that exhibits a poor response to immunotherapy. The mechanisms underlying immune suppression and evasion within the tumor microenvironment (TME) of PCa remain inadequately understood. Methods A comprehensive analysis of the immune environment in PCa was conducted using combined single-cell and spatial transcriptomic approaches, encompassing samples from healthy tissue, adjacent normal tissue, and localized tumors. Cell abundance and polarization state analyses were performed to identify pivotal cellular populations. Spatial deconvolution techniques were employed to elucidate cell composition within its spatial context. Additionally, cell niche and spatial colocalization analyses were conducted to evaluate potential cellular interactions. Immune response enrichment analysis was utilized to assess cellular response states. In vivo and in vitro experiments were conducted to validate hypotheses. Results Data indicated a prevalent immunosuppressive state among CD8 T cells, accompanied by variations in cell abundance. Macrophages emerged as key regulators in recruiting CD8+ effector T cells and regulatory T cells (Tregs) into the TME, mediated by the CXCL12/CXCR4 axis. A spatial proximity relationship was established between CD8+ effector T cells and Tregs, suggesting Tregs directly influence CD8+ T cell function. Immune cell state analysis revealed interleukin-2 (IL-2) as a critical cytokine in reshaping the immune microenvironment, with Tregs competitively depleting IL-2 and mediating IL-2/STAT5 signaling to induce CD8+ effector T cell exhaustion. Treatment with CXCR4 inhibitor and IL-2 demonstrated significant antitumor effects and reversed immune dysfunction in both in vivo and in vitro experiments, with combined treatment exhibiting superior efficacy. Conclusion These findings elucidate the role of macrophages in mediating the CXCL12/CXCR4 axis to aggregate CD8+ effector T cells and Tregs, thereby influencing the TME. Furthermore, Tregs competitively deplete IL-2 and mediate IL-2/STAT5 signaling, leading to CD8+ effector T cells exhaustion and the establishment of an immunosuppressive microenvironment. Keywords: prostate cancer, tumor immune microenvironment, CD8+ T cell, regulatory T cells, CXCL12/CXCR4 axis, IL-2/STAT5 signaling 1. Introduction Prostate cancer (PCa) is one of the most prevalent malignancies among men globally, ranked as the second most common type and the fifth leading cause of cancer-related mortality in males ([37]1). Despite advances in treatment modalities, including androgen deprivation therapy (ADT), targeted therapies, and chemotherapy, the impact on overall cure rates remains limited, alongside a notably high rate of biochemical recurrence ([38]2). Although ADT effectively controls early-stage hormone-sensitive prostate cancer (HSPC), approximately 25%-30% of patients progress to castration-resistant prostate cancer (CRPC) within five years, further evolving into metastatic CRPC (mCRPC) ([39]3, [40]4). Targeted therapies have shown some promise in prolonging overall survival (OS) and progression-free survival (PFS); however, effective responses are observed in only a subset of PCa patients ([41]5). Consequently, immunotherapy has emerged as a promising approach in cancer management, aiming to identify common response characteristics within the tumor microenvironment (TME). PCa is often characterized as an immune ‘cold’ tumor, exhibiting immunosuppressive properties due to a paucity of immune cells within the TME, which contributes to its poor response to immunotherapeutic strategies ([42]6). Consequently, checkpoint blockade therapies, such as anti-PD-1/PD-L1 and anti-CTLA-4 agents, are currently not preferred treatments in this context, as they demonstrate limited efficacy. The complex immune cellular landscape within the TME of PCa remains poorly understood, necessitating deeper exploration of the cellular interactions and variations in immune populations that play pivotal roles in determining the effectiveness of immunotherapies ([43]7). Thus, there is an urgent need to elucidate the immune cell states and interactions within the TME, with the goal of transitioning the immune response from ‘cold’ to ‘hot’ ([44]8). Identifying a common immunosuppressive mechanism to reshape the TME could enhance the efficacy of immunotherapy and expand the benefits to more PCa patients. Recent advancements in high-throughput sequencing technologies, including single-cell and spatial transcriptomics, have been rapidly integrated into PCa research to elucidate underlying cellular mechanisms within the TME at high resolution ([45]8, [46]9). In this study, single-cell RNA sequencing data obtained from improved tissue dissociation and library preparation techniques were utilized to retain and preserve a diverse array of immune cells, providing a comprehensive depiction of the TME landscape in PCa. Additionally, paired spatial transcriptome data from Slide-seq V2 was employed to contextualize the spatial distribution of TME populations, particularly focusing on the proximity of key immune cells. Through the integrated analysis of single-cell and spatial transcriptomic data, this study aims to reveal the characteristics of the immune microenvironment and enhance our understanding of the modulatory mechanisms underlying the immunosuppressive TME in PCa. Our observations indicate (1): CD8 T cell subpopulations frequently exhibit signs of exhaustion (2); there is an elevation in macrophage proportions and their interaction with lymphocytes within tumor regions (3); macrophages mediate the CXCL12/CXCR4 axis, facilitating the recruitment of regulatory T cells (Tregs) and CD8+ effector T cells to exert anti-tumor effects (4); spatial colocalization between Tregs and CD8+ effector T cells has been identified (5); the activation state of immune cells is confirmed, with IL-2 cytokines playing a crucial role in shifting the immune microenvironment from ‘cold’ to ‘hot’ (6); Tregs compete for IL-2, inducing CD8+ T cell exhaustion and promoting tumor progression. This careful and insightful dissection offers novel perspectives on the molecular mechanisms and cellular crosstalk within the TME, potentially guiding the development of new therapeutic targets and improving the efficacy of immunotherapy for PCa patients. 2. Materials and methods 2.1. Comprehensive single-cell data processing After excluding samples with low sequencing saturation, a total of 36 scRNA-seq datasets, encompassing 149,200 cells, were analyzed. These datasets included 4 from healthy prostate tissues, 14 from adjacent normal tissues paired with localized prostate cancer (PCa) samples, and 18 from localized PCa tissues. Enhanced library construction strategies were employed to capture a greater diversity of immune cell populations within the tissue, including healthy, adjacent normal, and tumor single-cell RNA sequencing (scRNA-seq) data ([47]10). The ‘Seurat’ R package (4.3.2 version) was used to conduct downstream analysis. Rigorous quality control measures were implemented: cells expressing < 400 genes or yielding raw counts < 800 were excluded from further analysis. The R packages ‘doubletFinder’ and ‘decontX’ were utilized to identify potential doublet cells and contaminants from environmental RNA ([48]11, [49]12). Subsequently, the top 2,000 highly variable genes were selected for principal component analysis (PCA). To mitigate batch effects, the ‘harmony’ algorithm was employed, and cell clusters were constructed at a resolution of 0.6, based on the K-nearest neighbors (KNN) graph. Cell annotation was performed according to marker genes identified in the original research ([50]10). 2.2. Tumor cell identification To accurately identify tumor cells and their epithelial subtypes, all epithelial cells were processed using the same methodology described previously. Different epithelial subtypes were annotated based on classical marker genes and associated enriched pathways. The ‘infercnvpy’ software, a Python implementation of the ‘infercnv’ R package, was employed to detect malignant cells within the epithelial population. Healthy epithelial cells served as reference cells, with other parameters set to default values. Based on copy number variation (CNV) scores, the Leiden algorithm was utilized to cluster the tumor cells. Additionally, the ‘monocle2’ and ‘PAGA’ packages were used to investigate the differentiation trajectory of the cell populations. 2.3. Analysis of key immune cell abundance and state To further ascertain the abundance and state of lymphocytes, the observed/expected (O/E) ratio was employed to explore the tissue preferences of each lymphocyte subtype population. To assess the