High-dimensional deconstruction of ovarian cancer at single-cell precision reveals HEBP2 that reshape the TIME and drive carboplatin resistance.

BiKang Yang: Investigation, Formal analysis, Data curation, Conceptualization. Miao Dai: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources.

All authors have agreed to publish this manuscript.

This work was supported by the 10.13039/501100001809 National Natural Science Foundation of China (Grant no. 82203410), Scientific Research Project of Hunan Provincial Health Commission (Grant no. B202305017803).

In this study, we conducted a thorough analysis of single-cell transcriptomic data from ovarian cancer (datasets GSE165897 , GSE158937 , and GSE154600 ) available in the GEO database. Our primary objective was to delineate the cellular composition and molecular features of ovarian cancer. The initial step involved the exclusion of cells with excessive mitochondrial gene content. Subsequently, we employed dimensionality reduction clustering, utilizing an approach that integrates highly variable gene (HVG) analysis with PCA, as illustrated in Supplementary Fig. 1(A), (B). This methodology facilitated an assessment of the influence of cell cycle genes on clustering, shown in Supplementary Fig. 1(C), and enabled us to explore correlations involving mitochondrial-related genes, the number of features (nfeatures), erythrocyte-associated genes, and counts (ncounts), as presented in Supplementary Fig. 1(D)–(F). Furthermore, we detailed the metrics for nfeatures, ncounts, mitochondrial genes, erythrocyte genes, and cell cycle-related scores for each sample, as documented in Supplementary Fig. 1(G). Following rigorous quality control and the removal of batch effects, we successfully isolated a total of 104,054 cells. The clustree method was applied to illustrate the impact of varying clustering resolutions on cellular grouping, depicted in a dendrogram format ( Fig. 1 (A)). Post-batch effect mitigation, the t-SNE and UMAP plots revealed a homogeneous cell distribution, confirming the effective elimination of sample bias ( Fig. 1 (B)). Employing marker genes from existing literature, we identified three predominant cell clusters: cancer, stromal, and immune cells ( Fig. 1 (C)) [26] . The distinct proportionality of these cell types across various lesions highlighted the tumor's heterogeneity and lesion consistency ( Fig. 1 (D)). Lastly, we pinpointed the cell types in each cluster based on their specific high gene expression profiles ( Fig. 1 (E)). Fig. 1 Precision transcriptomic analysis of single cells in ovarian cancer. (A). Resolution-based clustering of all cells. (B). t-SNE and UMAP plots for individual ovarian cancer samples. (C). t-SNE and UMAP visualizations of 11 principal cell types in ovarian cancer. (D). Distribution of cell clusters across samples. (E). Gene expression profiles for each cell cluster, with dot size indicating the proportion of cells expressing a specific marker gene and color intensity reflecting average gene expression level. Fig. 1 Precision transcriptomic analysis of single cells in ovarian cancer. (A). Resolution-based clustering of all cells. (B). t-SNE and UMAP plots for individual ovarian cancer samples. (C). t-SNE and UMAP visualizations of 11 principal cell types in ovarian cancer. (D). Distribution of cell clusters across samples. (E). Gene expression profiles for each cell cluster, with dot size indicating the proportion of cells expressing a specific marker gene and color intensity reflecting average gene expression level. In this study, six distinct subgroups of ovarian cancer cells were identified. The umap visualization indicated diverse spatial distributions among these subgroups, implying that they may possess unique biological properties ( Fig. 2 (A)). Subsequent analyses revealed that subgroups of cancerous epithelial cells, categorized by clinical type, demonstrated significantly varied transcriptomic profiles, further underscoring the heterogeneity within ovarian cancer cells ( Fig. 2 (B)). Additionally, a notable variance in the proportions of these subgroups was observed across different ovarian cancer samples. This finding highlights the substantial heterogeneity in ovarian cancer cell composition, thereby shedding light on the varied nature of the disease in different patients ( Fig. 2 (C)). Fig. 2 Heterogeneity in ovarian cancer. (A). UMAP analysis identifying six main cancerous epithelial cell subgroups in ovarian cancer. (B). Heatmap showing differential expression of genes (DEGs) in ovarian cancer epithelial cell subgroups. (C). Distribution of six ovarian cancer cell clusters across various samples. Fig. 2 Heterogeneity in ovarian cancer. (A). UMAP analysis identifying six main cancerous epithelial cell subgroups in ovarian cancer. (B). Heatmap showing differential expression of genes (DEGs) in ovarian cancer epithelial cell subgroups. (C). Distribution of six ovarian cancer cell clusters across various samples. Immune cells are crucial in recognizing and eliminating tumor cells. They detect and attack abnormally proliferating cells, thus impeding tumor development. However, tumor cells may enhance their survival by modifying immune cell functions or evading immune detection, underscoring the importance of investigating immune cell mechanisms in tumor resistance. In our study, we identified nine distinct immune cell subgroups: B-cells, Dendritic Cells (DC), Innate Lymphoid Cells (ILC), macrophages, Mast-cells, plasmacytoid Dendritic Cells (pDC), Plasma-cells, and T-cells, as illustrated in Fig. 3 (A). Utilizing a UMAP heatmap to display marker genes, we observed their exceptional ability to discriminate between cell types ( Fig. 3 (B)). The gene expression profiles and biological functions of these immune cells vary significantly, as shown in Fig. 3 (C). Notably, the proportion of these cell subgroups varies substantially among patients, reflecting the marked heterogeneity of immune cells ( Fig. 3 (D)). Fig. 3 Detailed analysis of immune cells in the ovarian cancer microenvironment. (A). UMAP plots delineating different immune cell types. (B). Density plots for typical immune cell marker genes. (C). Heatmap of DEGs in immune cell subclusters. (D). Proportion of each immune cell subgroup in ovarian cancer samples. Fig. 3 Detailed analysis of immune cells in the ovarian cancer microenvironment. (A). UMAP plots delineating different immune cell types. (B). Density plots for typical immune cell marker genes. (C). Heatmap of DEGs in immune cell subclusters. (D). Proportion of each immune cell subgroup in ovarian cancer samples. Stromal cells play a pivotal role in the tumor microenvironment by providing essential support and nutrients, significantly influencing tumor growth, dissemination, and metastasis. Investigating the interactions between stromal and tumor cells is vital for understanding tumor progression and devising novel therapeutic approaches. An optimal resolution for clustering was determined by evaluating the impact of various resolutions ( Fig. 4 (A)). Analysis of stromal cells using UMAP plots revealed a two-dimensional cell distribution, suggesting successful batch effect mitigation, as evidenced by the absence of a clear relationship between samples. This study identified three distinct stromal subgroups: Cancer-Associated Fibroblasts (CAF), Endothelial, and Mesothelial cells, and displayed the chosen resolution for subgroup identification ( Fig. 4 (B)). Notably, stromal cells from patients with chemotherapy-resistant ovarian cancer exhibited marked transcriptomic expression differences compared to those from chemotherapy-sensitive patients ( Fig. 4 (C)). Additionally, ridge plots and marker gene analysis demonstrated that these genes effectively discriminate between cell clusters ( Fig. 4 (D)). The majority of stromal cells infiltrating the tumor microenvironment were CAFs, aligning with prior research. Nevertheless, considerable variability was observed in the infiltrating cells across different samples, highlighting the heterogeneity and complexity of tumors ( Fig. 4 (E)). Fig. 4 In-depth study of stromal cells in the ovarian cancer microenvironment. (A). Cell clustering at varying resolutions. (B). UMAP plots illustrating samples, cell types, and clusters. (C). Volcano plot highlighting differentially expressed genes in stromal cells, with upregulated genes in CR-OC indicated in red. (D). Peak plots for key immune cell marker genes. (E). Distribution of each immune cell subgroup in ovarian cancer samples. Fig. 4 In-depth study of stromal cells in the ovarian cancer microenvironment. (A). Cell clustering at varying resolutions. (B). UMAP plots illustrating samples, cell types, and clusters. (C). Volcano plot highlighting differentially expressed genes in stromal cells, with upregulated genes in CR-OC indicated in red. (D). Peak plots for key immune cell marker genes. (E). Distribution of each immune cell subgroup in ovarian cancer samples. In the Tumor Microenvironment (TME) of ovarian cancer, intensive communication between various cell types, including tumor cells, immune cells, and stromal cells, significantly affects the response to chemotherapy drugs. The interactions among these cells may promote chemotherapeutic drug resistance in tumor cells, thereby reducing the effectiveness of treatment and impacting patient prognosis. In ovarian cancer, we observed that the quantity and intensity of communication between cellular subgroups are particularly rich ( Fig. 5 (A), Supplementary Figs. 2 and 3). Further exploring the outgoing and incoming signals of these cell subgroups, we identified the primary incoming and outgoing signals as MIF, SPP1, MK, and CXCL ( Fig. 5 (B)). At the same time, we visualized the dominant senders (sources) and receivers (targets) in 2D space, with the main senders being CAF, Mesothelial, EOC cells, and the main receivers being DC, B-cells, and macrophages ( Fig. 5 (B)). Concurrently, we also discovered that malignant epithelial cells in ovarian cancer can influence other cell subgroups through various signaling pathways (Supplementary Fig. 4). Fig. 5 Inter-subgroup cell communication. (A). Analysis of communication strength and frequency among cell subgroups. (B). Key patterns of outgoing and incoming signaling. (C). 2D representation of primary communication emitters and receivers. Fig. 5 Inter-subgroup cell communication. (A). Analysis of communication strength and frequency among cell subgroups. (B). Key patterns of outgoing and incoming signaling. (C). 2D representation of primary communication emitters and receivers. In contemporary research, which focuses on the intricate communication among specific cell subgroups and individual pathways, a critical question emerges: How do we integrate the functions of various cell groups and signaling pathways? To address this, we employed CellChat for the identification of overarching communication patterns, utilizing pattern recognition methods. Through the application of selectK, we discerned the number of distinct output and input patterns ( Fig. 6 (A), (E)), subsequently classifying both cell subgroups and communication patterns into several primary categories (input and output, n = 3 each) ( Fig. 6 (B), (F)). River plots were utilized to visualize the inferred potential output patterns in association with cell groups and signaling pathways ( Fig. 6 (C), (G)). Moreover, we conducted pattern recognition analysis to examine minute alterations in the outgoing and incoming signals across all key pathways ( Fig. 6 (D), (H)). Our findings highlight the pivotal role of ovarian cancer cells, T cells, and macrophages in shaping the global communication landscape. Fig. 6 Systematic analysis of intercellular communication global patterns. (A) Estimation of output pattern numbers via Cophenetic and Silhouette values. (B) Determination of cell subgroups and associated signal patterns in output. (C) Alluvial diagram representation of secretory cells' outgoing signal patterns. (D) Dot plots detailing variations in outgoing signals across major pathways. (E) Estimation of input pattern numbers via Cophenetic and Silhouette values. (F) Determination of cell subgroups and associated signal patterns in input. (G) Alluvial diagram representation of secretory cells' incoming signal patterns. (H) Dot plots detailing variations in incoming signals across major pathways. Fig. 6 Systematic analysis of intercellular communication global patterns. (A) Estimation of output pattern numbers via Cophenetic and Silhouette values. (B) Determination of cell subgroups and associated signal patterns in output. (C) Alluvial diagram representation of secretory cells' outgoing signal patterns. (D) Dot plots detailing variations in outgoing signals across major pathways. (E) Estimation of input pattern numbers via Cophenetic and Silhouette values. (F) Determination of cell subgroups and associated signal patterns in input. (G) Alluvial diagram representation of secretory cells' incoming signal patterns. (H) Dot plots detailing variations in incoming signals across major pathways. CR-OC typically exhibit a less favorable prognosis compared to those with Ovarian Cancer-Chemosensitive (OC-CS). Extensive research underscores the substantial influence of intercellular communication on disease outcomes. Consequently, analyzing cell communication disparities between CR-OC and OC-CS is vital for comprehending the emergence of drug resistance in CR-OC. Both patient groups demonstrate a wide range of cellular communication (see Fig. 7 (A)), characterized by notable differences in the quantity and intensity of interactions within their respective cellular networks. Heatmaps were employed to elucidate these differences, particularly within subcellular group communications, revealing a pronounced escalation in macrophage interactions in CR-OC cases (refer to Fig. 7 (B)). To delve deeper into CR-OC pathogenesis, an investigation into the variance of signaling pathways among subcellular groups was conducted. This analysis identified several pathways (e.g., MIF, VISFATIN, SPP1, CALCR, BAFF, EGF) as being markedly overexpressed in CR-OC (see Fig. 7 (C)). Intriguingly, numerous studies correlate these highly expressed pathways with macrophage activity and fatty acid metabolism, hinting at their critical role in CR-OC [28] , [29] , [30] , [31] , [32] , [33] . Remarkably, macrophages in CR-OC patients demonstrate a significant surge in both input and output signals, thereby implying their contribution to the development of drug resistance in ovarian cancer (refer to Fig. 7 (D)). Additionally, a comparison of specific signaling pathways and ligand-receptor interactions, which regulate communication probability in both OC-CS and CR-OC, revealed distinct differences in incoming, outgoing, overall signals, and communication probabilities between the two groups (see Fig. 7 (G)–(I) and Supplementary Fig. 5). Fig. 7 Comparative analysis of cell communication in CR-OC vs. OC-CS patients. (A) Differential cell interaction quantification between CR-OC and OC-CS. (B) Heatmap illustrating interaction number and intensity disparities. (C) Comparative overview of information flow in signaling pathways. (D) 2D comparison of primary sources and targets in CR-OC and OC-CS. (G) Analysis of outgoing signals specific to each cell group. (H) Analysis of incoming signals specific to each cell group. (I) Overall signal comparison across each cell group. Fig. 7 Comparative analysis of cell communication in CR-OC vs. OC-CS patients. (A) Differential cell interaction quantification between CR-OC and OC-CS. (B) Heatmap illustrating interaction number and intensity disparities. (C) Comparative overview of information flow in signaling pathways. (D) 2D comparison of primary sources and targets in CR-OC and OC-CS. (G) Analysis of outgoing signals specific to each cell group. (H) Analysis of incoming signals specific to each cell group. (I) Overall signal comparison across each cell group. Research in cell communication suggests that the metabolism of fatty acids plays a crucial role in determining ovarian cancer's resistance to treatment. Understanding the interplay among fatty acid metabolism, the immune microenvironment, and intercellular communication is crucial for comprehending cancer progression and chemotherapy resistance. Fatty acid metabolism is pivotal not only in providing energy for tumor cell growth but also in influencing cell membrane composition and function, which in turn modifies cell signaling and substance transport. Furthermore, fatty acids, as signaling molecules, regulate the tumor microenvironment, impacting immune cell function and promoting tumor immune evasion. The interaction between tumor cells and adjacent cells, including immune and stromal cells, critically affects tumor behavior, potentially enhancing tumor growth, spread, and chemotherapy resistance. Therefore, a thorough investigation of these interconnected mechanisms is vital for developing more effective cancer treatments. Ovarian cancer cells were categorized into chemotherapy-sensitive and chemotherapy-resistant groups, with differential analysis identifying 137 genes associated with chemotherapy resistance in the latter group (Supplementary Fig. 6(A), (B)). Additionally, dividing these cells into high and low fatty acid metabolism groups revealed 854 fatty acid metabolism-related genes (Supplementary Fig. 6(C), (D)). An overlap between chemotherapy resistance and fatty acid metabolism-related genes identified 56 significant genes (FAM-CR genes) ( Fig. 8 (A)). Subsequent analysis in the GEO database confirmed the prognostic value of HEBP2 among these genes (Supplementary Fig. 7). Fig. 8 HEBP2′s regulatory impact in ovarian cancer. (A) Venn diagram of genes linked to high fatty acid metabolism and drug resistance. (B) UMAP plot of HEBP2 expression. (C) Volcano plot contrasting genes with varying HEBP2 expressions. (D) Gene Ontology (GO) analysis of differentially expressed genes. (E) Gene Set Variation Analysis (GSVA) of differential genes. (F) Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differential genes. (G) Gene Set Enrichment Analysis (GSEA) of differential genes. Fig. 8 HEBP2′s regulatory impact in ovarian cancer. (A) Venn diagram of genes linked to high fatty acid metabolism and drug resistance. (B) UMAP plot of HEBP2 expression. (C) Volcano plot contrasting genes with varying HEBP2 expressions. (D) Gene Ontology (GO) analysis of differentially expressed genes. (E) Gene Set Variation Analysis (GSVA) of differential genes. (F) Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differential genes. (G) Gene Set Enrichment Analysis (GSEA) of differential genes. To assess HEBP2′s effect on ovarian cancer's oncogenic epithelial cells, we categorized these cells based on HEBP2 expression levels ( Fig. 8 (B)) and analyzed differential gene expression using a volcano plot ( Fig. 8 (C)). Gene Ontology (GO) analysis of these differential genes highlighted their involvement in cellular adhesion regulation, immune response activation, and other crucial biological processes ( Fig. 8 (D)). Gene Set Variation Analysis (GSVA) revealed that HEBP2-high cells exhibit upregulation in pathways critical for tumor growth, proliferation, apoptosis, signal transduction, metabolic changes, and microenvironmental intercellular interactions ( Fig. 8 (E)). Additionally, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis linked these changes to fatty acid metabolism, immune system regulation, and cancer-promoting pathways ( Fig. 8 (F)). Gene Set Enrichment Analysis (GSEA) further showed significant correlations between HEBP2 expression, M2 macrophage activation, and T cell differentiation ( Fig. 8 (G)). These findings suggest that HEBP2 plays a crucial role in ovarian cancer, particularly in remodeling the tumor immune microenvironment, cell stress regulation, immune system processes, and fatty acid metabolism. Extensive research corroborates the involvement of these processes in the immune microenvironment's remodeling in ovarian cancer. Thus, exploring HEBP2′s role in this context holds significant potential for advancing ovarian cancer diagnosis and treatment. To investigate the association between HEBP2 and the Tumor Microenvironment (TME) in ovarian cancer, we employed cibersortX to analyze the cellular composition of the TME using pre-annotated single-cell RNA (scRNA) data. Our analysis revealed significant heterogeneity in the immune cell infiltration within the ovarian cancer TME across patients, as illustrated in Fig. 9 (A). Predominant cellular constituents in the overall TME included cancer cells, stromal cells, T cells, and macrophages (refer to Fig. 9 (B)). Further, we observed a notable correlation between HEBP2 and most cell types infiltrating the ovarian cancer TME, with a positive association specifically with cancer cells (see Fig. 9 (C)). These findings imply that HEBP2 may contribute to adverse prognoses in ovarian cancer patients by altering the TME. Consequently, elucidating the regulatory impact of HEBP2 on immune cells and its potential role in modulating the immune microenvironment is crucial for understanding immune suppression and recognition in ovarian cancer. Fig. 9 HEBP2′s role in reshaping the tumor microenvironment (TME) in ovarian cancer. (A) Assessment of immune infiltration per sample. (B) Analysis of cellular composition proportions in samples. (C) Correlation study between HEBP2 and bone marrow microenvironment cells. Fig. 9 HEBP2′s role in reshaping the tumor microenvironment (TME) in ovarian cancer. (A) Assessment of immune infiltration per sample. (B) Analysis of cellular composition proportions in samples. (C) Correlation study between HEBP2 and bone marrow microenvironment cells. The intrinsic cytological and molecular biological properties of ovarian cancer predispose it to developing resistance against specific treatments. To investigate the influence of HEBP2 on the drug sensitivity of ovarian cancer, we manipulated gene expression in ovarian cancer tissues. We then analyzed the TCGA and GSE140082 datasets using the OncoPredict algorithm to compute scores for each sample. The half-maximal inhibitory concentration (IC50) values of prevalent drugs were evaluated using the OncoPredict package. Spearman correlation analysis revealed that elevated HEBP2 expression is associated with increased IC50 values for various drugs ( Fig. 10 (A), (B)), implying that HEBP2 may contribute to multi-drug resistance in ovarian cancer. Concurrently, molecular docking studies were conducted to elucidate the interaction between HEBP2 and primary ovarian cancer drugs. These studies demonstrated a significant affinity of HEBP2 with carboplatin, doxorubicin, and docetaxel ( Fig. 10 (C)), suggesting that HEBP2 is a promising target for future ovarian cancer therapies. Fig. 10 Impact of HEBP2 upregulation on drug resistance in ovarian cancer. (A) Correlation between HEBP2 levels and IC50 values for various drugs (GEO database). (B) Correlation between HEBP2 levels and IC50 values for various drugs (TCGA database). (C) Molecular docking analysis of HEBP2 interaction with primary ovarian cancer drugs. Fig. 10 Impact of HEBP2 upregulation on drug resistance in ovarian cancer. (A) Correlation between HEBP2 levels and IC50 values for various drugs (GEO database). (B) Correlation between HEBP2 levels and IC50 values for various drugs (TCGA database). (C) Molecular docking analysis of HEBP2 interaction with primary ovarian cancer drugs. In ovarian cancer, macrophages are often reprogrammed into M2 macrophages within the tumor microenvironment, aiding cancer growth and metastasis. These cells contribute to a tumor-friendly microenvironment by secreting pro-inflammatory and pro-angiogenic factors. M2-type macrophages also dampen the immune response, facilitating immune evasion by ovarian cancer cells. Additionally, they play a role in maintaining and remodeling the tumor matrix, thus exacerbating ovarian cancer's aggressiveness and drug resistance. Our research in cell communication highlighted macrophages' pivotal role in CR-OC, leading to their isolation for detailed analysis. Post dimensionality reduction, clustering, and batch effect elimination, we examined the effects of various resolutions on cell clustering (Supplementary Fig. 8(A)), selected an optimal resolution, and categorized macrophages into M0_TAM, M1_TAM, and M2_TAM (Supplementary Fig. 8(B)). Uniform cell distribution across samples confirmed successful batch effect removal (Supplementary Fig. 8(C)). M1_TAM predominantly expresses IFIT1, while M2_TAM expresses MAF, CD163, and MRC1, and M0_TAM shows minimal specific marker expression. These markers distinctly classify the three macrophage types (Supplementary Fig. 8(D)). Distinct expression profiles and biological functions of macrophages in various states are evident in Supplementary Fig. 8(E). Trajectory analysis of macrophages in ovarian cancer lesions, utilizing Monocle 2 and Monocle 3 algorithms, revealed their spatiotemporal evolution. Marker genes indicate MAF, CD163, and HEBP2 predominantly in M2_TAM ( Fig. 11 (A)). Employing Monocle 3, we estimated macrophage evolution and visualized it on the UMAP plot, illustrating cell clusters, types, and differentiation stages ( Fig. 11 (B)–(D)). Monocle 2 was also used for further analysis of macrophage subtypes, HEBP2 expression, and differentiation stages ( Fig. 11 (E)–(G)). Both algorithms suggest M0_TAM differentiates into M1_TAM and M2_TAM, primarily following the M2_TAM developmental pathway. The gene kinetics in macrophage evolution is notable, with HEBP2 exhibiting a consistent upward trend, correlating with M2_TAM cell accumulation ( Fig. 11 (J)). 3D reconstructions more clearly demonstrate the differentiation from M0_TAM to M1_TAM and M1_TAM to M2_TAM, featuring a significant rise in HEBP2 during M2_TAM differentiation ( Fig. 11 (H), (I)). Single-cell trajectory branches emerge from divergent gene expression programs, with macrophages at branching point 1 differentiating into numerous M1_TAM or M2_TAM cells, and a notable increase in HEBP2 during M2_TAM differentiation ( Fig. 11 (K)). Fig. 11 Role of HEBP2 in altered macrophage evolution. (A) UMAP plot illustrating classic markers of macrophage subgroups. (B-D) Monocle 3 trajectory plots depicting macrophage subcluster differentiation paths: (B) Trunk and nodes, (C) Trunk with cell annotations, (D) Trunk with time progression. (E-G) Monocle 2 trajectory plots of macrophage subcluster dynamics: (E) Subcluster changes, (F) Time progression, (G) HEBP2 expression. (H, I) 3D trajectory plots of macrophage subcluster dynamics (H) and HEBP2 expression (I). (J, K) Heatmap hierarchical clustering for developmental timing and subgroup-specific marker genes: (J) Overall subgroups, (K) Differential expression in subgroups at point-1. Fig. 11 Role of HEBP2 in altered macrophage evolution. (A) UMAP plot illustrating classic markers of macrophage subgroups. (B-D) Monocle 3 trajectory plots depicting macrophage subcluster differentiation paths: (B) Trunk and nodes, (C) Trunk with cell annotations, (D) Trunk with time progression. (E-G) Monocle 2 trajectory plots of macrophage subcluster dynamics: (E) Subcluster changes, (F) Time progression, (G) HEBP2 expression. (H, I) 3D trajectory plots of macrophage subcluster dynamics (H) and HEBP2 expression (I). (J, K) Heatmap hierarchical clustering for developmental timing and subgroup-specific marker genes: (J) Overall subgroups, (K) Differential expression in subgroups at point-1. We successfully developed carboplatin-resistant ovarian cancer cell lines OVCAR4/R and Ovsaho/R using a gradient increment method. The sensitivity of these resistant lines to carboplatin was evaluated using the CCK-8 assay. The half-maximal inhibitory concentration (IC50) values for OVCAR4/R and Ovsaho/R were 23 μM and 43 μM, respectively, compared to 3.4 μM and 4.5 μM for their parental cells, OVCAR4 and Ovsaho. This indicates the successful establishment of carboplatin resistance (Supplementary Fig. 9(A)). Western blot analysis showed significantly reduced HEBP2 protein expression in OVCAR4 and Ovsaho compared to their resistant counterparts, OVCAR4/R and Ovsaho/R ( Figs. 12 (A), S9(B)). To investigate HEBP2′s role in proliferative processes in vitro, we constructed three specific shRNAs targeting HEBP2 (si-HEBP2#1, si-HEBP2#2, and si-HEBP2#3). CCK-8 and cell colony formation assays revealed that HEBP2 knockdown decreased its expression in resistant cells and inhibited cell proliferation ( Figs. 12 (B)–(D), S9(C)). Flow cytometry analysis of cell apoptosis post-HEBP2 knockdown demonstrated that si-HEBP2#2 significantly increased apoptosis ( Fig. 12 (E), Supplementary Fig. 9(D)). In summary, these findings suggest that HEBP2 facilitates the growth and proliferation of ovarian cancer cells. Fig. 12 HEBP2 expression in ovarian cancer cell lines and its role in drug resistance. (A) Western blot analysis of HEBP2 in various cell lines. (B) Verification of HEBP2 knockdown in OVCAR4/R and Ovsaho/R cells. Functional assays in drug-resistant cells: CCK-8 assay (C), colony formation (D), and apoptosis detection via flow cytometry (E). Data are presented as mean ± SD ( n = 3), with ** p < 0.01 indicating significant intergroup differences. Fig. 12 HEBP2 expression in ovarian cancer cell lines and its role in drug resistance. (A) Western blot analysis of HEBP2 in various cell lines. (B) Verification of HEBP2 knockdown in OVCAR4/R and Ovsaho/R cells. Functional assays in drug-resistant cells: CCK-8 assay (C), colony formation (D), and apoptosis detection via flow cytometry (E). Data are presented as mean ± SD ( n = 3), with ** p < 0.01 indicating significant intergroup differences. With increasing post-tumor inoculation duration, there was a marked growth in tumor volume in the shRNA-NC group mice. Conversely, mice in the HEBP2-shRNA group exhibited significantly smaller tumor volumes than those in the shRNA-NC group ( Figs. 13 (A), (B)). In vivo studies demonstrated a substantially reduced tumor mass in the HEBP2-shRNA group compared to the shRNA-NC group ( P < 0.001) ( Fig. 13 (C)). Immunofluorescence analysis revealed a notably lower green fluorescence intensity in the HEBP2-shRNA group's tumor tissues, suggesting a reduced HEBP2 protein expression rate relative to the shRNA-NC group ( P < 0.05) ( Fig. 13 (D)). Histological examination via HE staining of mouse tumor tissues indicated slower tumor growth in the HEBP2-shRNA group compared to the shRNA-NC group ( Fig. 13 (E)). TUNEL staining results showed a significant increase in TUNEL-positive cells (red fluorescence), indicative of increased apoptosis, in the HEBP2-shRNA group ( P < 0.05), as opposed to fewer TUNEL-positive cells in the shRNA-NC group. These findings preliminarily validate that HEBP2 inhibition enhances growth and apoptosis in subcutaneous ovarian cancer ( Fig. 13 ). Fig. 13 In vivo validation of HEBP2′s oncogenic role. (A) Dissected mouse tumor specimens. (B, C) Tumor volume and weight measurements in different mouse groups. (D) Immunofluorescence for HEBP2 in tumor tissues. (E, F) Hematoxylin and eosin (HE) staining and HEBP2 assay of tumor tissues. Fig. 13 In vivo validation of HEBP2′s oncogenic role. (A) Dissected mouse tumor specimens. (B, C) Tumor volume and weight measurements in different mouse groups. (D) Immunofluorescence for HEBP2 in tumor tissues. (E, F) Hematoxylin and eosin (HE) staining and HEBP2 assay of tumor tissues.

We sourced single-cell sequencing datasets for ovarian cancer from the Gene Expression Omnibus (GEO). Three scRNA-seq datasets ( GSE154600 , GSE158937 and GSE165897 ) were selected based on their corresponding clinical information. Furthermore, bulk RNA-seq clinical samples, equipped with survival time data, were also retrieved from GEO. Datasets TCGA and GSE140082 were identified to fulfill our criteria. For the processing, quality assessment, dimensional reduction, and clustering of the single-cell RNA sequencing (scRNA-seq) data, the Seurat package for R (version 4.1.3) was utilized. Cells were excluded from analysis if their gene count fell outside the boundaries of 300 to 9000 or if the proportion of mitochondrial gene expression was over 20 %, indicating low quality. Following the exclusion criteria, normalization of the remaining cells was performed using the ``lognormalize'' method, which applies a global scaling factor. Subsequently, we utilized the FindVariableFeatures function to pinpoint highly variable genes, which were then used in the Principal Component Analysis (PCA). The first 10 significant principal components were chosen for dimensionality reduction, and the Seurat FindClusters function facilitated cell type clustering at a resolution of 0.20. Subsequent to this process, dimensionality reduction and visualization were further accomplished through the implementation of UMAP and tSNE techniques. The signature genes of each cluster were identified using the FindAllMarkers function. The clustering tree (clustree) visually delineates the inter-relationships of clusters across different resolutions. Using the cell annotation outcomes, we employed ClusterGVis for heatmap representation and enrichment analysis of subgroup differential genes. Marker gene visualization was achieved using the Nebulosa package, scCustomize package . For the assessment of cell-to-cell signaling, we employed CellChat (version 1.5.0). The initial step involved the creation of a CellChat object, which was derived from the RNA expression matrix and cell metadata through the createCellChat function. Subsequently, our analysis focused on the combined expression profiles of ligand-receptor (L-R) pairs, incorporating aspects such as ``Secreted Signaling,'' interactions with the ``Extracellular Matrix-Receptor,'' and ``Cell-Cell Contact.'' The computeCommunProb function was used to estimate the probabilities of communication, which aided in the construction of the cellular interaction network. In an effort to elucidate overarching signaling trends, the selectK function was employed with nPatterns set at 3 for inbound and 4 for outbound signaling pathways. For tracing the pseudotemporal progression of macrophage subsets, we utilized the Monocle2 and Monocle3 packages within R. The construction of these single-cell developmental pathways was based on the expression patterns of genes differentially expressed among clusters. The dimensionality of the data was reduced via the DDRTree algorithm. The visualization of the cellular trajectories, which reflect cell differentiation stages, was achieved using Monocle's plot_cell_trajectory function. The metabolic status of fatty acids in each EOC cell was calculated using scMetabolism, and EOC cells were divided into high or low fatty acid metabolism cells based on the median value. Through differential analysis, genes related to fatty acid metabolism were identified. Using the R package `survival', we conducted survival analyses for each gene. Using established cutoffs, patients were segregated into cohorts with either high or low gene expression for analysis. Survival evaluations were subsequently carried out within these cohorts. The distinction in survival distributions was determined by the Log-Rank test, with P -values beneath the 0.05 threshold regarded as statistically significant. When identifying differentially expressed genes (DEGs) in different samples of the same cell type or at varying levels of gene expression, harnessing the Wilcox method, was applied for comparative analysis and the discernment of DEGs. The selection parameters set forth stipulated (1) a minimum absolute average log2 fold change threshold of 1.2, and (2) a statistical significance criterion with a P -value below 0.05. Subsequent to the differentiation among cell types, the clusterProfiler tool was employed for the enrichment analysis, drawing upon data from the Kyoto Encyclopedia of Genes and Genomes (KEGG). Pathways from KEGG that exhibited a P -value under 0.05 were earmarked for further investigation to explicate the predominant biological processes associated with the DEGs. Gene set enrichment analyses, including GSEA and GSVA, were conducted using dedicated R package resources. Infiltration scores for each cellular subtype within the GEO clinical cohort were inferred utilizing the CibersortX instrument alongside the average expression levels of characteristic genes. The correlation of HEBP2 expression with various cell clusters was evaluated via Spearman's rank correlation method. The prediction of drug responsiveness in ovarian cancer cases was conducted using the OncoPredict software from the R package suite. This tool integrates gene expression data with the drug sensitivity metrics, specifically IC50 values, drawing upon the Genomics of Drug Sensitivity in Cancer (GDSC) database as well as the cancer cell line profiles from the CCLE provided by the Broad Institute. An assortment of 198 therapeutic agents was assessed, and Spearman's rank correlation method was utilized to determine the relationship between HEBP2 expression levels and the drugs' IC50. A correlation coefficient's absolute value exceeding 0.2 combined with a false discovery rate (FDR) below 0.05 was considered statistically significant. Based on prior research [27] , OVCAR and Ovsaho cells were initially cultured in DMEM supplemented with 10 % FBS and 1 % penicillin/streptomycin, with subculturing occurring every 3–4 days. For the establishment of drug-resistant cell lines, the IC50 was used as the starting concentration for induction. Upon achieving stable growth, the concentration was incrementally doubled, and the IC50 was monitored to assess cell viability during carboplatin treatment. This entire process was conducted in accordance with the protocols established in earlier studies. Cells were harvested and subsequently lysed using 1 ml of TRIzol reagent, followed by the addition of 200 μl of chloroform to extract total RNA. The RNA was then reverse-transcribed using a Takara kit from Japan, adhering to the prescribed ratios. The reaction conditions included an incubation at 37 °C for 15 min and a denaturation step at 85 °C for 5 s. The resulting cDNA served as a template for PCR, with primers detailed in Supplementary Table 1. To ensure stability, all samples and reagents were kept on ice throughout the procedure and shielded from light. Prior to the assay, the cDNA was diluted between three to five-fold. The amplification methodology included an initial melting step at 95 °C for half a minute, followed by annealing at 60 °C for a brief 5-s duration, and then synthesis at the same temperature of 60 °C for 34 s. This sequence was performed over 40 cycles and was succeeded by an analysis of the dissociation curve, involving a 15-s interval at 95 °C, a minute at 60 °C, and another 15 s at 95 °C. Samples were processed in a set of three replicates. Quantification of the target genes' relative expression was carried out employing the comparative 2-ΔΔCt technique. Cells undergoing exponential growth were plated in 6-well plates. Transfection was initiated when cells attained a confluency of 70–80 %, using the Lip3000 transfection agent according to the manufacturer's instructions. At six hours following transfection, the medium was substituted with fresh standard culture medium, and the cells were incubated for another 24 h. Further analyses were conducted on these harvested cells. Details on the si-HEBP2 sequence are available in Supplementary Table 1. OVCAR/R and Ovsaho/R cell lines were plated at 1 × 10 4 cells/ml density in 96-well plates, allocating 100 µl of the cell mixture per well. Both cell populations were transfected with si-NC or si-HEBP2, setting up three replicates per variant. At 24-h intervals up to 72 h, a 10 µl volume of CCK-8 reagent was introduced to the medium, and the cells were subsequently incubated for 4 h. Absorbance at OD450 was recorded using a spectrophotometric microplate reader. The obtained data were processed with GraphPad Prism 7, applying dose-response curve analysis. The extent of apoptosis was determined using an Annexin V-FITC/PI apoptosis detection kit. Post-transfection cells, both si-NC and si-HEBP2 groups, were washed with PBS, followed by a rinse with 1 × Binding Buffer. After centrifugation, the supernatant was discarded, and the cells were suspended again in 100 µL of Binding Buffer. Subsequently, 10 μL of Annexin V-FITC was added to the cells, which were then incubated for 15 min in the dark at room temperature. Post additional washing and centrifugation, 5 μL of PI stain was applied, and cells were resuspended in 200 μL of 1 × Binding Buffer. The apoptosis rates were then gauged using flow cytometry. The colony formation ability was assessed by establishing a 0.6 % agar base layer within 6-well plates, pouring 0.5 ml per well and allowing it to set at room temperature. For further incubation, plates were placed in a 37°C and 5 % CO 2 incubator. OVCAR/R and Ovsaho/R cells, in their exponential growth phase and post si-NC or si-HEBP2 transfection, were spun down at 1000 rpm for 5 min. These cells were then resuspended in IMDM medium, counted, and diluted to a target density of 1000 cells in 0.5 ml, which was layered atop the base agar. Cultivation proceeded at 37°C with 5 % CO 2 for a duration of 10 to 14 days, until colonies became visible. Upon colony development, the process was halted and the colonies were assessed using an inverted microscope. Colony counts were performed to determine the rate of colony formation, with triplicates maintained for consistency. The crystal structure of the target protein was sourced from the SWISS-MODEL protein database and saved in PDB format. Drug structures in sdf format were obtained from PubChem and converted to mol2 format using Open Babel GUI software. Protein modifications were made using PyMOL and AutoDock4, which included the removal of water molecules, addition of nonpolar hydrogen, charge calculations, and identification of rotatable bonds in the ligand molecule, with the final structures saved as PDBQT files. Docking parameters were set based on receptor and ligand dimensions. Molecular docking was executed with AutoDock4, and visualization was achieved with PyMOL. The binding energy threshold was set at −2.0 kcal/mol. We divided six 4-week-old female NOD-scid IL2rγnull (NSG) mice into two groups: a knockdown group and a control group, with three mice in each. We injected Ovsaho/R cells, with stable knockdown of HEBP2 (sh-HEBP2; see Supplementary Table 1 for sequence) and control cells (sh-NC), subcutaneously into the NSG mice (1 × 10 7 cells in 0.1 mL PBS). Carboplatin was administered intraperitoneally to the respective groups at a dosage of 50 mg/kg once daily over a span of seven days. Tumor sizes were recorded every two weeks, with the volume being determined by the formula: volume = length × width 2 / 2. On the 21st day, the mice were humanely sacrificed, following which the tumors were excised and their weights were ascertained. We processed paraffin-embedded tumor tissues through xylene and a series of alcohols for dewaxing and rehydration, followed by staining with hematoxylin and eosin to visualize nuclei and cytoplasm, respectively. Following the staining procedure, the tissue sections were subjected to a graded alcohol series for dehydration, then clarified with xylene, and finally embedded with neutral balsam. Fluorescence microscopy was utilized for the examination and capture of images of these sections. We detected apoptosis in the breast cancer tumor tissues using an in situ fluorescein-labeled TUNEL system. We prepared cryosections from the mouse tumor tissues, which, after equilibration to room temperature, were washed and incubated with a TdT and dUTP mixture for one hour in a humidified chamber at 37 °C. Following incubation, sections were rinsed and mounted with an anti-fade solution and examined by fluorescence microscopy. To prepare for immunofluorescence, tissue sections set in paraffin were warmed to 70 °C for a duration of one hour, followed by a sequential dewaxing and rehydration process employing ethanol and PBS. For antigen exposure, we utilized a citric acid solution in an autoclave. Post cooling, we thoroughly rinsed the sections, applied a blocking solution, and then proceeded with an overnight incubation at 4 °C with the primary antibody against HEBP2 at a dilution of 1:1000. Subsequently, the sections were cleansed and exposed to a secondary antibody conjugated with Alexa Fluor 555 at a dilution of 1:600 for a period of 20 min at an ambient temperature. After the final rinse, we used a DAPI-incorporated mounting medium for the sections before their visualization with a fluorescence microscope. Cell lysates were prepared using RIPA buffer (Beyotime), and protein concentrations were determined using a BCA assay (Beyotime). Proteins, in equal quantities, were resolved by 10 % SDS-PAGE and subsequently transferred onto PVDF membranes. These membranes were blocked with 5 % skim milk and incubated with primary antibodies at 4 °C on a shaker. Following this, membranes were incubated with HRP-conjugated secondary antibodies (Proteintech) at room temperature. Post-washing, signal detection was performed using the UVP ChemStudio system (Ultraviolet Products, USA), and data were analyzed with VisionWorks software (Analytik Jena, Germany). The primary antibodies employed were anti-HEBP2 (1:1000, Proteintech), anti-BCL2 (1:200, Abcam), anti-BAX (1:1000, Abcam), and anti-GAPDH (1:10,000, Proteintech). For the statistical examination, SPSS software (version 23.0) was utilized. Group comparisons of data adhering to a normal distribution were conducted using the t -test for independent samples. Conversely, the Wilcoxon rank-sum test was employed for the evaluation of non-normally distributed variables. A P -value threshold of less than 0.05 was set to denote statistical significance.

The high incidence and poor prognosis of ovarian cancer presents substantial challenges in gynecological oncology, with variations in tumor behavior and genetic makeup. Among these, Type II epithelial ovarian cancers, particularly high-grade serous carcinomas that constitute approximately 75 % of cases, are markedly aggressive with poor outcomes, often characterized by p53 and BRCA mutations. Conversely, Type I tumors present a more indolent course and genetic stability, frequently originating from conditions such as endometriosis [34] . This poor prognosis is closely associated with the tumor's immune microenvironment and its resistance to chemotherapy. In ovarian cancer, the diversity and complexity of the immune microenvironment profoundly influence chemotherapy responses and resistance [ 35 , 36 ]. Reflecting on our journey through this research, we recognize the untapped potential of single-cell sequencing technology. It not only deepens our understanding of the tumor microenvironment but also heralds a new era of personalized medicine in oncology. The advent of single-cell sequencing technology has enabled a more nuanced understanding of this microenvironment. Through the analysis of gene expression at the single-cell level, this technology reveals the cellular heterogeneity within tumors with remarkable precision. In our study, we employed multiple single-cell databases to conduct a comprehensive analysis of ovarian cancer, categorizing cells into various types such as stromal cells, immune cells, and tumor cells. Our analysis identified significant differences in cellular communication patterns between chemotherapy-resistant and chemotherapy-sensitive patients within the tumor microenvironment, which might be pivotal in the varied responses to chemotherapy. Our analysis opens the door to novel therapeutic targets, offering hope for more effective treatments. Our findings underscore the substantial individual variations in the cellular makeup of the tumor microenvironment among ovarian cancer patients, with potential implications for treatment outcomes. Consequently, leveraging single-cell sequencing technology to investigate these disparities is essential for developing personalized and more efficacious treatment strategies, potentially improving chemotherapy sensitivity and identifying novel therapeutic targets. Looking ahead, we envision a future where such technologies become commonplace in clinical settings, guiding the development of targeted therapies. In biomedicine, In biomedicine, drug resistance in ovarian cancer presents a formidable challenge, markedly impacting treatment outcomes and patient survival [ 37 , 38 ]. Acknowledging the knowledge gaps, we are motivated to explore the molecular underpinnings of drug resistance. Our study sheds light on macrophages' role but also emphasizes the need for further research to fully understand and combat this challenge. Consequently, tackling drug resistance and prolonging patient life are central objectives in clinical research. Recent studies highlight the pivotal role of macrophages in the pathogenesis, progression, and drug resistance development of ovarian cancer [ 39 , 40 ]. Macrophages secrete diverse factors such as inflammatory cytokines (TNF-α, interleukins), chemokines, vascular endothelial growth factor (VEGF), and other growth factors, fostering a tumor microenvironment conducive to cancer growth and metastasis [41] , [42] , [43] . Intriguingly, macrophages may also dampen immune responses by expressing immunosuppressive molecules like PD-L1, thereby impairing the body's immune surveillance and facilitating tumor cell survival and proliferation, which contributes to increased drug resistance [44] . Moreover, evidence indicates that BRCA deficiency triggers a STING-dependent innate immune response, leading to the production of type I interferon and pro-inflammatory cytokines. It has been further observed that PARP inhibition can inactivate GSK3 and increase PD-L1 expression in a dose-dependent manner, thereby suppressing T-cell activation and enhancing cancer cell apoptosis. This insight supports the therapeutic strategy of combining PARP inhibitors with immunotherapies, such as anti-CTLA-4 and PD-1/PD-L1, especially in cancers with BRCA1/2 or HR deficiency, which are hypothesized to have a higher neo-antigen load and thus, a more potent anti-tumor immune response [45] . The single-cell sequencing results of this study warrant further investigation into the molecular mechanisms underlying chemotherapy resistance in ovarian cancer. As researchers, we are on a quest to unravel these complexities, striving to translate our findings into clinical benefits. Our signal flow analysis identified heightened activity in several key signaling pathways—MIF, VISFATIN, SPP1, CALCR, BAFF, and EGF—in chemotherapy-resistant cells. A thorough literature review revealed a strong association of these pathways with macrophage functions and metabolism of fatty acid [ 28 , 30 , [46] , [47] , [48] ]. Most notably, a two-dimensional analysis of chemotherapy-resistant cells showed a marked increase in macrophage-related incoming and outgoing signals, underscoring the potential critical role of macrophages in the drug resistance mechanism of ovarian cancer. Fatty acid metabolism significantly contributes to tumor drug resistance by supporting the survival and proliferation of cancer cells. It alters the tumor immune microenvironment and supplies an alternative energy source to cancer cells [ 49 , 50 ]. Furthermore, fatty acid metabolism can activate key signaling pathways, such as the PI3K/Akt/mTOR pathway, essential for cell growth, survival, and drug resistance development. Moreover, it is crucial to note that the upregulated PI3K pathway plays a significant role in maintaining genomic stability by being involved in various DNA replication and cell cycle regulation processes. Inhibition of the PI3K pathway may lead to genomic instability and mitotic catastrophe by reducing the activity of the spindle assembly checkpoint protein Aurora kinase B, which in turn increases the occurrence of lagging chromosomes during prometaphase. This highlights the complex interplay between metabolic pathways and cellular mechanisms of genomic integrity [51] . These mechanisms enhance tumor cell energy supply and may interfere with apoptotic processes, increasing cancer cells' resistance to anticancer treatments. Therefore, targeting fatty acid metabolism pathways in research and therapy could offer new strategies to combat tumor drug resistance. In the next five years, we anticipate significant advances in understanding and treating ovarian cancer. By integrating single-cell RNA sequencing (scRNA-seq) and bulk sequencing (Bulk-seq) data, we identified HEBP2, a gene associated with high fatty acid metabolism, overexpressed in chemotherapy-resistant ovarian cancer and linked to patient survival prediction. Using bioinformatics tools like GO, KEGG, GSVA, and GSEA, we elucidated HEBP2′s roles in ovarian cancer, including tumor progression, immune evasion, cell signaling, and drug tolerance induction. Sensitivity and affinity analyses of chemotherapy drugs revealed HEBP2′s role in increasing half-maximal inhibitory concentration (IC50) values and binding affinity with frontline chemotherapy drugs. In vitro and in vivo experiments showed HEBP2′s ability to inhibit apoptosis and promote proliferation in ovarian cancer cells, including drug-resistant strains and mouse tumor models, suggesting HEBP2′s critical role in ovarian cancer's carboplatin chemotherapy tolerance. Notably, macrophages, key immune response cells, can recognize and kill tumor cells. However, in the complex ovarian cancer immune microenvironment (TME), tumor-associated macrophages often exhibit an M2 polarized state, leading to reduced effector functions and proliferation, thus fostering ovarian cancer recurrence and drug resistance. Our time-series analysis further revealed that HEBP2 is predominantly expressed in pro-cancer macrophages (M2_TAM), indicating its potential role in promoting the transition from the initial M0_TAM state to the M2_TAM state. Our work not only contributes to the scientific community's knowledge but also paves the way for innovative treatments that could significantly impact patient care. While our study provides valuable insights into the tumor microenvironment and potential therapeutic targets in ovarian cancer, we acknowledge several limitations. Firstly, the complexity of the tumor microenvironment may not be fully captured by our single-cell sequencing approach, as spatial relationships and dynamic changes over time are not accounted for. Secondly, our findings are based on available databases and samples, which may not represent the full spectrum of ovarian cancer heterogeneity. Thirdly, the translation of our findings into clinical practice requires validation through clinical trials, which are beyond the scope of our current research. Lastly, while we highlight the potential of targeting fatty acid metabolism and macrophage signaling pathways, the development of such therapies involves complex considerations, including safety, efficacy, and patient-specific factors. In conclusion, our study embodies the intersection of cutting-edge technology and patient-centered research. By embracing the complexity of ovarian cancer and utilizing advanced sequencing techniques, we have begun to chart a course towards more personalized and effective treatments. The potential of our findings to influence future research and therapy is immense, and we are committed to continuing this important work, driven by the hope of improving outcomes for patients worldwide.

In this study, we found that HEBP2 potentially plays a pivotal role in ovarian cancer resistance by augmenting the proliferative capacity of the cancer cells and diminishing their sensitivity to carboplatin. Notably, HEBP2 appears to facilitate a functional transition in macrophages, leading to a preferential shift towards an M2_TAM phenotype, potentially disrupting the immune homeostasis within the ovarian cancer microenvironment. However, the mechanism of resistance in ovarian cancer cells is multifaceted, involving numerous pathways and molecular interactions. Therefore, a more comprehensive understanding of HEBP2′s role in fostering resistance is imperative to identify novel therapeutic targets aimed at counteracting this resistance. This investigation will lay a crucial theoretical foundation for the formulation of safer and more efficacious treatment modalities.

Ovarian cancer is a highly lethal malignancy within gynecology, characterized by a concerning mortality rate. Recently, its incidence has been rising steadily, garnering increased attention from the medical community [1] . The use of prenatal ultrasonography has uncovered more asymptomatic ovarian masses, around 5 % of which are malignant, affecting cancer incidence. Current practice advises surgical intervention for masses over 6 cm or those showing symptoms [2] . Despite advancements in clinical management through conventional surgery, chemotherapy, radiation, and targeted therapy, these treatments often have limited efficacy due to ovarian cancer's intricate pathogenesis [ 3 , 4 ]. This complexity arises from the interplay of various molecular and cellular pathways, complicating both the disease and its treatment. Recent shifts towards systemic therapies underscore the critical role of predictive biomarkers in ovarian cancer management. The evolution from traditional chemotherapy to the integration of targeted treatments has been marked by significant milestones, highlighting the need for ongoing research into the dynamic therapeutic landscape of ovarian cancer. Studies such as the MOUSEION series provide pivotal insights into this arena, demonstrating the influence of gender and performance status on the efficacy of immunotherapies [ 5 , 6 ]. These findings advocate for a personalized approach to treatment, emphasizing the importance of biomarkers in optimizing therapeutic outcomes [7] .Consequently, an in-depth understanding of ovarian cancer's molecular mechanisms and the identification of novel molecular targets are imperative for developing more effective therapeutic approaches. Recent advancements in proteomics technologies, notably mass spectrometry and protein array analysis, have significantly contributed to the elucidation of the complex molecular signaling pathways and the proteome of ovarian cancer. These technologies enable a more detailed understanding of the tumor's proteomic landscape and its adaptive responses to various therapies, thereby identifying novel molecular targets. As a result, leveraging proteomics analysis can pave the way for the development of innovative therapeutic strategies, potentially mitigating drug resistance and enhancing patient survival rates [8] . The Tumor Immune Microenvironment (TIME) is a complex, dynamic network comprising tumor cells, immune cells (T cells, B cells, natural killer cells, macrophages, dendritic cells), inflammatory cells, vascular endothelial cells, fibroblasts, and extracellular matrix components [9] , [10] , [11] . TIME plays a pivotal role in tumor initiation, progression, and metastasis by facilitating immune evasion, inflammatory responses, immune suppression, angiogenesis, and the creation of physical barriers [ 12 , 13 ]. Recent advances in immunotherapy have marked significant milestones in treating various cancers, including ovarian cancer [14] . However, the immunosuppressive nature of the Tumor Microenvironment (TME) can impede immune cell functionality, thereby reducing the effectiveness of immunotherapy. Consequently, an in-depth understanding of the intricate interactions between tumor cells and their microenvironment, as well as their impact on the immune response to cancer, is vital for refining immunotherapy approaches and improving therapeutic outcomes [15] , [16] , [17] . The development of resistance to platinum-based chemotherapeutics presents a substantial challenge in ovarian cancer treatment. Chemotherapy-resistant ovarian cancer (CR-OC) patients exhibit a significantly poorer prognosis compared to those with chemotherapy-sensitive ovarian cancer (CS-OC), a disparity largely attributable to the tumor immune microenvironment (TIME) and intricate intercellular interactions. Within the TIME, an array of cells, including tumor cells, immune cells, and others, engage in complex communication through cytokines, chemical signals, exosomes, and direct cell-to-cell contact. These interactions facilitate not only the survival and proliferation of tumor cells but also augment their resistance to chemotherapeutic agents [18] , [19] , [20] , [21] . For instance, tumor cells may release specific exosomes that alter the behavior of adjacent cells, contributing to the establishment of a tumor-friendly microenvironment and enhancing tumor growth and chemotherapy resistance [ 22 , 23 ]. Consequently, a comprehensive examination of the diverse mechanisms within the ovarian cancer immune cell network, both promoting and inhibiting the immune response, is essential to address the challenge of immune resistance. The occurrence and development of ovarian cancer are closely related to specific subgroups of immune cells, rather than a single type of immune cell. However, current understanding of the composition and characteristics of the ovarian cancer immune microenvironment is still limited. Single-cell sequencing technology, as a cutting-edge tool for studying the heterogeneity and diversity of tumor cells, can detect transcriptional activity at the single-cell level, revealing the functional state of cells. By applying single-cell sequencing technology, we can more deeply explore the composition and function of different immune cell subgroups in the ovarian cancer tumor microenvironment, thereby deepening our understanding of the ovarian cancer immune microenvironment, identifying new therapeutic targets and mechanisms of drug resistance. This not only optimizes the treatment efficacy for ovarian cancer patients and improves survival rates, but also has significant implications for scientific research and clinical practice. In this investigation, single-cell transcriptomics datasets ( GSE165897 , GSE158937 , and GSE154600 ) were employed to examine variations in cellular composition between ovarian cancer patients responsive and resistant to chemotherapy [24] , [25] , [26] . The purpose of this analysis was to thoroughly investigate immune cell remodeling and tumor heterogeneity in ovarian cancer. By concentrating on specific cell types, a comprehensive examination of ovarian cancer resistance was carried out, thereby enriching and enhancing the detailed study of the immune microenvironment and drug resistance in ovarian cancer.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.