Multi-omics profiling reveals an immunosuppressive plasma cell subset within tertiary lymphoid structures in cervical cancer Authors: Qingyu Huang, Fuhao Wang, Xiaohui Li, Wenxue Zou, Zhou Huang, Xia Li, Wenhui Yang, Tianyu Lei, Shengqin Yue, Qian Wang, Qi An, Bicheng Zhang, Cong Wang, Chao Liu, Qinyong Hu Published: 2025 ABSTRACT AbstractBackgroundCervical cancer (CC) is a leading cause of cancer-related deaths in women, and understanding the tumor immune microenvironment is crucial for identifying novel biomarkers and therapeutic targets. While T cells have been extensively studied in oncology, the role of B cells in CC remains poorly understood.MethodsIn this study, we generated and integrated multi-omics data, including single-cell RNA sequencing, single-nucleus RNA sequencing, spatial transcriptomics, bulk RNA sequencing, and multiplex immunofluorescence to investigate the composition and transcriptomic states of B cells in CC.ResultsInitially, we analyzed plasma cells, classifying them into IgA+ and IgG+ subtypes. We found that the proportion of IgA+ plasma cells was lower in cancer samples, whereas IgG+ plasma cells were higher. IgA+ plasma cells exhibited protein folding and B cell activation characteristics, while IgG+ plasma cells were associated with apoptosis and immune suppression. Subsequently, we focused on two specific plasma cell populations, MANF_PC and HSPA1B_PC. MANF_PC, which were predominantly enriched in normal tissues, were linked to enhanced antibody synthesis and better prognosis, while HSPA1B_PC, primarily found in cancerous tissues, were associated with apoptotic pathways and poor clinical outcomes. Importantly, HSPA1B_PC are associated with tertiary lymphoid structures (TLS), where they show significant spatial colocalization with immunosuppressive T cell subsets, including Tregs and Th17 cells within TLS.ConclusionsWe identified distinct plasma cell subsets in CC and found that HSPA1B_PC are enriched in tumors, associated with immunosuppressive T cells in TLS. These findings suggest that HSPA1B_PC contribute to an immunosuppressive microenvironment in CC, highlighting their potential as therapeutic targets.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00262-025-04268-w.Keywords: Cervical cancer, B cell, Multi-omics, Single-cell RNA sequencing, Tertiary lymphoid structure Introduction Cervical cancer (CC) ranks as the fourth most prevalent cancer in women and the sixth leading cause of cancer-related deaths globally [1, 2]. While advances in treatment have improved outcomes, many patients with advanced CC still face poor prognoses due to recurrence and metastasis [1, 2]. A deeper understanding of the tumor immune microenvironment is critical for identifying novel biomarkers and therapeutic targets to improve treatment efficacy. While research on anti-tumor immunity has focused mainly on tumor-infiltrating T cells, studies on B cells in oncology are limited. Human B cells are categorized into several distinct subsets, including naive B cells, memory B cells, plasma cells, and regulatory B cells [3]. Emerging evidence suggests that B cells do not act in isolation but interact with other immune cells within the tumor [4]. Plasma cells produce tumor-specific antibodies that form antigen–antibody complexes, which dendritic cells (DCs) can internalize [4]. This mechanism enhances antigen presentation to T cells, thereby promoting an effective anti-tumor immune response [4]. Beyond their individual roles, B cells are thought to influence the spatial organization and activity of other immune cells, particularly T cells, within the tumor microenvironment (TME) [5]. The spatial proximity between B cells and T cells may reflect functional interactions, contributing to the formation of tertiary lymphoid structures or modulating local immune responses [5]. Understanding these spatial relationships is crucial for elucidating how B cells shape the immune landscape in CC. In this study, we employed a multi-omics approach, integrating single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, and bulk RNA sequencing, to comprehensively analyze the heterogeneity of B cells in CC. By examining their transcriptional profiles, proportional relationships, and spatial colocalization with T cells, we aim to uncover novel insights into the roles of B cells in the TME. Materials and methods Human sample and data collection A total of 15,876 cells from 65 samples were analyzed. The discovery cohort included 17 samples, from which 7903 B cells were isolated. The validation cohort comprised 48 samples, from which 7973 cells were isolated. The discovery cohort included 17 samples, comprising 7903 B cells. Of these, 2168 B cells (27.4%) were derived from three newly enrolled patients with HPV+ CC (T1, T2, and T3), diagnosed with stage IIIC cervical adenocarcinoma (T1), stage IB2 cervical adenocarcinoma (T2), and stage IB3 cervical squamous cell carcinoma (T3) according to the International Federation of Gynecology and Obstetrics standards. The remaining 5735 B cells (72.6%) were from previously studied CC and normal cervical tissues [6–8]. This study was approved by the Ethics Committee of Shandong Cancer Hospital (SDTHEC201906009) and adhered to the Declaration of Helsinki. The validation cohort consisted of 48 untreated tumor samples, comprising 7973 cells, from publicly available datasets, including GSE197461, GSE208653, GSE208654, GSE236738, SCP1950, HRA004971, HRA001742, HRA003491, CNP0002535, GSE168652, and S-BSST1035. Sample preparation The preparation of single-cell suspensions followed protocols outlined in prior research [6–8]. Mouse anti-human monoclonal antibodies against CD19 (clone HIB19) and CD45 (clone 2D1) from BioLegend were used for cell labeling. Cells were incubated with these antibodies for 15 min at 4 °C, followed by two washes and resuspension in 100 μl of sorting buffer. Subsequently, cells were stained with a 7-AAD viability staining solution (eBioscience). A BD FACSAria III cell sorter was employed to isolate live B cells (CD45+ CD19+) from the tumor tissues for subsequent scRNA-seq analysis. Multiplex immunofluorescence labeling for HSPA1B_PC identification Multiplex immunofluorescence was utilized to identify the HSPA1B_PC in the TME using specific cellular markers. HSPA1B_PC were identified by staining for HSP70, IGHG1, and CD38 markers. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). Initially, CC tissue sections were treated with primary antibodies against HSP70 (ab181606, Abcam), IGHG1 (PA5-75428, Invitrogen), and CD38 (PA5-95840, Invitrogen). This was followed by secondary antibodies conjugated with horseradish peroxidase, with signal enhancement achieved via the tyramide signal amplification (TSA) method. Each TSA phase was preceded by microwave heating for antibody removal before the next labeling step. After all antigens were labeled, DAPI was used for nuclear counterstaining. The slides were subsequently scanned using the Mantra system (PerkinElmer) to capture fluorescence spectra from 420 to 720 nm at 20-nm intervals with consistent exposure settings. The spectral images were analyzed using 3DHISTECH software (Hungary). scRNA‐seq and data preprocessing Dimension reduction and unsupervised clustering of single-cell RNA data were performed using Seurat (version 4.9.9) [9]. Potential doublets were identified and excluded with DoubletFinder (version 2.0.3). Per-cell QC metrics were computed in Seurat. We quantified the mitochondrial RNA fraction (percent.mt) using genes with the prefix ^MT-Distributions were inspected per sample (grouped by orig.ident) using violin plots to set data-driven cutoffs. Cells were retained if they met all of the following criteria: 450 ≤ detected genes ≤ 4500, UMIs ≤ 40,000, and percent.mt ≤ 20%. Quality control thresholds were visualized using violin plots (Figure S1). Following quality control, the expression matrix was normalized using the LogNormalize method with a scale factor of 10,000. Cell type annotation and B cell and T cell isolation Batch effects were mitigated using the Harmony package’s harmony algorithm (version 1.1.0). Principal component analysis (PCA) was applied to highly variable genes, and dimensionality reduction was achieved using uniform manifold approximation and projection (UMAP). The “Find All Markers” function identified signature genes for each cluster, and clusters were annotated based on lineage-specific gene expression. A second round of clustering and dimension reduction was performed, specifically targeting B cells and T cells. Differential analysis and pathway enrichment Gene set functionality and activity were examined through gene ontology (GO) enrichment analysis using ClusterProfiler (version 4.6.2), focusing on pathways with a significance level below 0.05. The cancer genome atlas (TCGA) analysis Bulk RNA-seq and clinical data for 187 patients with CC undergoing radiotherapy were obtained from UCSC Xena (http://xenabrowser.net/hub). Subtype-specific signatures for MANF_PC and HSPA1B_PC were defined by the top five upregulated DEGs identified from scRNA-seq. Sample-wise enrichment scores were computed using single-sample gene set enrichment analysis (ssGSEA). Patients were divided into high- and low-score groups using the optimal cutoff identified by the “surv_cutpoint” function from the survminer (0.5.0). Survival differences were evaluated by Kaplan–Meier curves with log-rank tests. Temporal analysis In this study, B cells were temporally aligned based on trajectories corresponding to fundamental biological processes such as cell differentiation, enabling a comprehensive analysis of the cellular time course. Pseudotime analysis was conducted using the monocle2 R package (version 1.6.1) [10, 11]. Spatial transcriptome data analysis For the public spatial transcriptomics (ST) datasets [12], data were processed and filtered with the Scanpy package (v1.8.2), applying quality control to exclude low-quality cells based on criteria such as low gene count and high mitochondrial gene expression. The data were normalized using total-count normalization and log transformed for further analysis. PCA was performed to reduce dimensionality, and the top principal components were retained for constructing a nearest neighbor graph. Gene set scoring was carried out using the scanpy.tl.score_genes function to calculate the average expression of predefined gene sets for each cell. Results Single-cell transcriptome landscape of B cell in CC To investigate the heterogeneity of B cells in CC, we performed a comprehensive analysis by integrating scRNA-seq, single-nucleus RNA sequencing, spatial transcriptomics, RNA-seq, and multiplex immunofluorescence (mIF) (Fig. 1A). The discovery cohort included 17 samples (12 tumor and 5 normal) from our own data. The validation cohort consisted of 48 tumor samples. Two additional tumor samples were analyzed using spatial transcriptomics. In our discovery cohort, nine subclusters were identified: CD37 + B cells (CD37_B), GPR34 + B cells (GPR34_B), IFIT3 + B cells (IFIT3_B), TTN + B cells (TTN_B), IGKC + plasma cells (IGKC_PC), HSPA1B + plasma cells (HSPA1B_PC), MANF + plasma cells (MANF_PC), IGHA1 + plasma cells (IGHA1_PC), and IGLC2 + plasma cells (IGLC2_PC, Figs. 1B, C, S2A). TTN_B were identified as naive B cells due to the lack of CD27 expression and high expression of MS4A1 (Figs. 1D, S2B). CD37_B, GPR34_B, and IFIT3_B were classified as memory B cells, distinguished by their expression of CD27 and MS4A1 (Figs. 1D, S2B). Identification of distinct B cell clusters in cervical cancer. A Schematic overview of the study design and scRNA-seq process. B UMAP visualization of B cells, color coded by cell type. C UMAP plots displaying the expression of selected marker genes for identifying B cell clusters, with color intensity indicating normalized gene expression. D Bubble plot showing the proportions and average expression levels of selected marker genes for B cell and plasma cell clusters Differential enrichment and characteristic insights of IgA and IgG plasma cells The proportions and characteristics of major B cell populations in CC and normal cervical tissues were analyzed (Figs. 2A, B, S3). Memory B cells and naive B cells showed no significant differences in their proportions between normal and cancerous tissues (Fig. 2C). However, the proportion of IgA+ plasma cells was significantly lower in cancer samples, while the proportion of IgG+ plasma cells was notably higher (Fig. 2C). GO enrichment analysis indicated that IgA+ plasma cells were enriched in pathways related to endoplasmic reticulum stress response, protein folding, B cell activation, and the positive regulation of B cell activation (Fig. 2D). In contrast, IgG+ plasma cells were enriched in pathways associated with oxidative stress response, endoplasmic reticulum stress response, B cell activation, and apoptotic signaling regulation (Fig. 2D). Differentially expressed transcription factors revealed that both IgA and IgG plasma cells in cancer expressed high levels of the mature plasma cell gene XBP1 (Fig. 2E). In cancer, GO analysis showed that IgA+ plasma cells exhibited upregulation in immune activation pathways, including the positive regulation of B cell activation, immunoglobulin-mediated immune response, defense response to viruses, antigen receptor-mediated signaling, and complement activation (Fig. 2F). Conversely, IgG+ plasma cells in cancer demonstrated upregulation in pathways associated with immunosuppression and apoptosis, such as the negative regulation of immune response and the positive regulation of cysteine-type endopeptidase activity involved in apoptotic processes (Fig. 2F). Differential enrichment and functional insights of IgA and IgG plasma cells. A Distribution of memory B cells (Mem B), IgG plasma cells (IgG+ PC), IgA plasma cells (IgA+ PC), and naive B cells across tumor and normal samples. B Cell compositions by tissue type, color coded by cell types. C Box plots showing the proportions of major clusters in normal and cancer samples. D GO enrichment analysis of IgA+ plasma cells and IgG+ plasma cells. E Dot plots of differentially expressed transcription factors in IgA+ and IgG+ plasma cells between normal and cancer. F GO analysis of upregulated pathways in IgA+ and IgG+ plasma cells in cancer Multi-omics analysis reveals the heterogeneity of plasma cells and identifies HSPA1B_PC in CC To further dissect the characteristics of plasma cell subsets, a detailed analysis focused on five distinct plasma cell populations was conducted. Notably, IGKC_PC, HSPA1B_PC, and IGLC2_PC were significantly elevated in cancerous tissues, whereas MANF_PC were more prevalent in normal tissues (Fig. 3A, C). HSPA1B_PC cells demonstrated characteristic expression of HSP70 family proteins and notable expression of IGHG genes (Fig. 3D). We further confirmed the presence of HSPA1B_PC in CC tissues by employing multiplex immunofluorescence to assess the expression of HSP70, IGHG1, and CD38, which are protein products of the key marker genes identified earlier (Fig. 3E). We also observed the presence of this population in the validation cohort (Figs. 3F, S4A, S4B). Additionally, spatial transcriptomics analysis demonstrated the colocalization of HSP family genes with IgG plasma cell signature genes, further validating the existence of this subset within the TME (Figure S4C). Mature states of two plasma cell clusters. A Distribution of nine B cell clusters across cervical cancer and normal tissue samples. B Cell compositions by tissue type, color-coded by cell types. C Proportions of nine B cell clusters in cancerous and normal tissues. D Heatmap showing the expression levels of key genes in different B cell and plasma cell subsets. E Representative multiplex immunofluorescence image showing labeling of CD38 (brown), IGHG1 (green), HSP70 (red), and DAPI (blue) in HSPA1B_PC within CC tissue. F tSNE visualization showing the presence of HSPA1B_PC in the validation cohort. The left one panel shows the clustering of plasma cells into five groups (PC1–PC4), and the right panel highlights the higher expression of HSPA1A, HSPA1B, MZB1, and IGHG4 in the PC1 cluster. G Heatmap illustrating key gene expression patterns across five plasma cell clusters Gene expression analysis highlights two plasma cell subsets Key gene expression patterns associated with human plasma cell maturation and functional states were examined. HSPA1B_PC and MANF_PC exhibited upregulation in genes related to class-switch recombination machinery, indicating active engagement in isotype switching and suggesting an advanced state of immune response readiness (Fig. 3G). These clusters also showed heightened gene activity linked to B cell receptor activity and signaling, essential for the activation and regulation of immune functions (Fig. 3G). Additionally, they were distinguished by upregulated expression of genes involved in IgH locus binding, suggesting roles in antigen specificity and antibody diversity (Fig. 3G). Further, these cells exhibited increased expression of genes associated with endoplasmic reticulum activity and stress responses, reflecting the intensive protein synthesis and folding required for robust antibody production (Fig. 3G). Metabolic pathways were also emphasized in these plasma cell subsets, underscoring the metabolic demands of high antibody production (Fig. 3G). In summary, the detailed analysis of plasma cell subsets in CC reveals distinct functional roles for MANF+ and HSPA1B+ plasma cells, characterized by unique gene expression profiles and pathway enrichments. MANF_PC showed advanced differentiation and antibody synthesis correlating with enhanced radiotherapy efficacy The heterogeneity of IgA plasma cells in CC was further investigated, with a focus on the MANF_PC group, which is enriched in normal tissues. GO analysis revealed that MANF_PC showed upregulation in pathways related to endoplasmic reticulum stress response, protein folding, positive regulation of lymphocyte activation, and response to unfolded protein (Fig. 4A). This pattern indicates a heightened state of antibody synthesis in these cells. Subsequent analysis of bulk RNA-seq data from patients with CC who underwent radiotherapy demonstrated that MANF_PC enrichment correlated with better prognosis (p = 0.018, Fig. 4B). Pseudotime analysis of memory B cells and IgA PC revealed that MANF_PC are at the terminal stage of differentiation, characterized by the downregulation of B cell-specific genes PAX5 and BACH2 and the upregulation of antibody-secreting cell-specific genes IRF4, XBP1, and PRDM1 (Fig. 4C, D, S5). Further analysis of upregulated pathways in MANF_PC indicated distinct functional roles in cancerous and normal tissues. In cancerous tissues, MANF_PC were involved in immune activation pathways, while in normal tissues, they participated in protein folding and response to topologically incorrect proteins (Fig. 4E). MANF_PC exhibit advanced differentiation and antibody synthesis linking to improved radiotherapy prognosis. A GO analysis of upregulated pathways in MANF+ plasma cells (MANF_PC). B Kaplan–Meier survival analysis of TCGA CC patients who underwent radiotherapy, stratified by high and low MANF_PC levels. The p value of the two-sided log-rank test is shown. C Pseudotime analysis of IgA+ plasma cells, colored by cell cluster. D Pseudo-heatmap displaying dynamic changes in gene and transcription factor expression along the pseudotime. E GO analysis of upregulated pathways in MANF+ plasma cells in cancerous and normal tissues HSPA1B_PC exhibit apoptosis characteristic and associated with poor radiotherapy outcomes The HSPA1B_PC group, predominantly found in cancerous tissues, was further investigated. GO enrichment analysis indicated significant involvement of HSPA1B_PC in pathways related to oxidative stress response, endoplasmic reticulum stress response, unfolded protein response, cellular response to unfolded proteins, intrinsic apoptotic signaling pathway, and regulation of apoptotic signaling (Fig. 5A). Survival analysis showed that elevated levels of HSPA1B_PC correlated with poorer overall survival in patients with CC receiving radiotherapy (p = 0.064, Fig. 5B). Pseudotime analysis identified HSPA1B_PC as terminally differentiated plasma cells, suggesting their maturity (Fig. 5C). Gene expression profiles reinforced this, with upregulated genes linked to terminal differentiation (Fig. 5D). Notably, transcription factors and genes like XBP1 were highly expressed in these cells (Fig. 5D). Additionally, pro-apoptotic gene expression in HSPA1B_PC was examined in both normal and cancerous tissues. In cancerous tissues, HSPA1B_PC exhibited upregulation of several pro-apoptotic genes, including BCL2L11, BAX, BID, and BIM, indicating an enhanced apoptotic response (Fig. 5E). In conclusion, our analysis showed that HSPA1B_PC upregulate apoptosis-related pathways and are linked to a worse prognosis in patients with CC undergoing radiotherapy. These cells, identified as terminally differentiated mature plasma cells, displayed distinct pro-apoptotic gene expression profiles in cancerous tissues, highlighting their potential role in regulating apoptosis within the TME. HSPA1B_PC exhibit apoptosis-related pathways and associated with poor radiotherapy outcomes in cervical cancer. A GO analysis of upregulated pathways in HSPA1B+ plasma cells (HSPA1B_PC). B Kaplan–Meier survival analysis of TCGA CC patients who underwent radiotherapy, stratified by high and low HSPA1B_PC levels. The p value of the two-sided log-rank test is shown. C Pseudotime analysis of IgG+ plasma cells, colored by cell cluster. D Pseudo-heatmap displaying dynamic changes in gene and transcription factor expression along the pseudotime. E Expression of pro-apoptotic genes in HSPA1B_PC within both normal and cancerous tissues HSPA1B_PC correlate with immunosuppressive T cell subsets Tertiary lymphoid structures (TLSs) are commonly composed of B cells and T cells, but the spatial organization and relationship between TLSs and HSPA1B_PC remain unclear. We observed that HSPA1B_PC highly express TLS-associated markers (Fig. 6A), which were validated in the spatial transcriptomics and bulk RNA-seq data (Figs. 6B, C, S6A), suggesting their potential involvement in TLSs. Next, we explored HSPA1B_PC’s potential relationship with T cells within the TME. Using scRNA-seq data, T cells were categorized into several subsets based on their gene expression profiles (Figs. 6D, E, S7A). We observed significant proportional correlations between HSPA1B_PC and specific T cell subsets. The proportions of Th17 cells and Tregs were enriched in cancer tissues and positively correlated with HSPA1B_PC abundance (Figs. 6F, G, S7B). Notably, these T cell subsets have been implicated in immunosuppressive mechanisms and are associated with poor prognosis in CC [6]. This relationship suggests a potential immunosuppressive microenvironment facilitated by these interactions. To validate these findings, spatial transcriptomics analysis was performed, revealing the spatial colocalization of HSPA1B_PC with Th17 cells and Tregs within TLSs (Fig. 6H). To elucidate the mechanisms underlying the spatial colocalization of HSPA1B_PC with Tregs and Th17 cells, we conducted ligand–receptor interaction analysis focusing on potential crosstalk between HSPA1B_PC and Tregs/Th17 cells. This analysis revealed significantly stronger communication probabilities in HSPA1B_PC. Notably, the SPP1–CD44 pair showed high interaction strength from HSPA1B_PC to both Tregs and Th17 cells (Figure S8). In summary, our findings suggest that HSPA1B_PC may involve in forming immunosuppressive TLSs in the TME. The colocalization of HSPA1B_PC with Th17 cells and Tregs within these structures highlights their role in promoting an immunosuppressive environment, which may contribute to tumor progression and poor prognosis. Spatial colocalization between HSPA1B_PC and immunosuppressive T cell subsets. A Violin plot showing the TLS score across B cell subsets. B Scatter plot depicting the relationship between HSPA1B_PC and TLS scores in spatial data. C Scatter plot showing HSPA1B_PC signature genes positively correlated with TLS signature genes in bulk RNA-seq data. D UMAP plots displaying T cell clusters, color coded by cell type. Teff, effector T cell; Treg, regulatory T cell; Tn, naïve T cell; Tex, exhausted T cell; Th17, T helper cell 17; Tem, effective memory T cell. E Heatmap showing the top 5 differentially expressed genes in T cell subsets. F Correlation heatmap showing the relationship between T cell subsets and HSPA1B_PC. G Scatter plots showing the correlation between HSPA1B_PC expression and Treg (left) and Th17 (right) scores in the TCGA dataset. H Spatial feature plots illustrating the distribution of HSPA1B_PC, Treg, Th17, and TLS scores in cervical cancer samples To deepen the mechanistic insights into the immunosuppressive role of HSPA1B_PC compared to MANF_PC, we performed additional analyses focusing on the expression of chemokines, cytokines, and immune checkpoints between HSPA1B_PC and MANF_PC. HSPA1B_PC exhibited higher expression of CXCL chemokines family such as CXCL2, CXCL3, CXCL8, CXCL16, and CXCL17 relative to MANF_PC (Supplementary Figure S9A). Cytokine profiling revealed that HSPA1B_PC had elevated expression of TGFB3, IL1B, and TNFSF13B (Figure S9B). Immune checkpoint analysis showed that HSPA1B_PC clustered with higher expression for inhibitory molecules such as IDO1, HAVCR2, and CD274 compared to MANF_PC (Figure S9C). We further compared ligand–receptor interactions of HSPA1B_PC and MANF_PC with B cells. Both subsets showed enrichment of the MIF—(CD74 + CXCR4) and MIF—(CD74 + CD44) signaling axes (Figure S10A, B). To assess potential interactions with DCs, key organizers of TLSs, we analyzed cDC1, cDC2, LAMP3⁺ DC, and pDC subsets defined by canonical markers (Figure S11A–C). Consistently, both plasma cell subsets exhibited similar enrichment of the MIF–(CD74 + CXCR4/CD44) pathways with DCs (Figure S11D). Taken together, these findings indicate that plasma cell–B/DC interactions are largely conserved. Additionally, HSPA1B_PC showed elevated hypoxia score (Figure S12). This hypoxic signature aligns with findings from a pan-cancer T cell atlas, where a stress response state T cell characterized by heat shock proteins co-occurs with hypoxia and contributes to immunosuppression [13]. Discussion The role of B cells in solid tumors remains unclear. Depending on their phenotype, antibody production, and interactions within the TME, B cells can exhibit both tumor-promoting and tumor-suppressing effects [14]. By integrating multi-omics analyses, we identified the differential enrichment of MANF_PC and HSPA1B_PC in normal versus cancerous tissues indicates their divergent roles in tumor immunology. MANF_PC, predominant in normal tissue, is linked to IgA secretion and mucosal immunity, aligning with the cervical mucosa’s role in CC pathogenesis [15]. High expression of genes related to endoplasmic reticulum stress, protein folding, and XBP1 activation highlights their advanced differentiation and readiness for antibody secretion [16, 17]. Despite prior associations of IgA with poor prognosis and tumor-promoting effects [14], our findings suggest that MANF_PC may confer a protective effect, as higher levels correlated with favorable outcomes in CC. In contrast, HSPA1B_PC, enriched in cancerous tissue, are linked to IgG secretion and apoptosis pathways. Elevated levels of HSPA1B_PC were associated with poor survival in CC patients, suggesting a more aggressive disease phenotype and treatment resistance. This aligns with reports that stress-associated B cells expressing HSPA1B foster an immunosuppressive environment, contributing to worse outcomes [18]. IgA⁺ plasma cells exhibited enrichment for protein folding/UPR, B cell activation, and defense response to viruses [19, 20]. The protein folding signature is consistent with the IRE1α–XBP1-dependent expansion of the ER and secretory machinery required for high-throughput antibody production in plasma cells, indicating a functional secretory state rather than generic stress [19, 20]. In HPV-driven CC, enrichment of antiviral programs in IgA⁺ cells is biologically plausible within a mucosal tumor bed, where humoral responses against HPV antigens have been documented [21]. Together with a B cell activation/TLS-like program—previously associated with favorable outcomes in CC—these findings support a protective role for IgA-dominant plasma cell responses in this context. In contrast, IgG⁺ and particularly HSPA1B_PC were enriched for apoptotic signaling, oxidative stress response, and negative regulation of immune response. We interpret this axis as a chronic proteotoxic/oxidative stress program in a subset of terminally differentiated plasma cells. HSP70 family induction under oxidative stress can preserve cell viability and has been linked to therapy resistance [22, 23]. In parallel, sustained oxidative and ER stress in the TME fosters PD-L1 upregulation and immunoregulatory outputs, offering a mechanistic basis for the immunosuppressive phenotype and worse prognosis [24]. Pseudotime analysis confirmed that HSPA1B_PC are terminally differentiated, with upregulation of pro-apoptotic genes, highlighting their involvement in apoptotic regulation within the TME. While IgG is generally linked to good prognosis [3], in patients with renal cell carcinoma receiving immune checkpoint inhibitor therapy, treatment response and progression-free survival are associated with IgG-stained tumor cells [25]. However, a poor prognosis in patients with CC exhibiting the presence of HSPA1B_PC was observed. The distinct pro-apoptotic gene expression profiles of HSPA1B_PC in cancer tissues highlight their potential as therapeutic targets for modulating apoptosis and improving patient outcomes in CC. TLSs play a crucial role in shaping the immune microenvironment and enhancing anti-tumor immune responses. However, TLSs also exhibit heterogeneity depending on the factors of TME [5]. Our study revealed a significant correlation and spatial colocalization between HSPA1B_PC and immunosuppressive T cell subsets, including Th17 cells and Tregs, within the TLSs. These T cell populations are associated with immunosuppression and poor prognosis in CC [6]. The colocalization suggests potential functional interactions, where HSPA1B_PC may contribute to the recruitment or activation of these T cell subsets, further promoting an immunosuppressive TLSs. It has shown that the presence of Treg cells in TLS may have a negative impact on the anti-tumor ability of TLS [26]. The close association between HSPA1B_PC and these T cells highlights their role in shaping the immune landscape and potentially influencing treatment resistance. Targeting these interactions could represent a novel strategy to improve immune responses and clinical outcomes in CC. Our results revealed that HSPA1B_PC are terminally differentiated plasma cells with pronounced upregulation of apoptosis-related pathways. Whether this pro-apoptotic program is a cause or consequence of the immunosuppressive TME remains an important question. The pro-apoptotic phenotype of HSPA1B_PC appears to be a consequence of the harsh conditions within the TME. HSPA1B_PC are enriched for pathways related to oxidative and endoplasmic reticulum stress and show an elevated hypoxia score. It is well established that the TME is characterized by hypoxia, nutrient deprivation, and high levels of reactive oxygen species, which collectively impose significant proteotoxic and metabolic stress on resident cells [27]. The high expression of HSP70 family proteins in HSPA1B_PC is a classic cellular response to such stressors, aimed at preserving cell viability [22, 23]. However, sustained and overwhelming stress can push cells past a survival threshold, triggering intrinsic apoptotic pathways [28]. Therefore, the TME likely acts as the primary driver that induces this stressed, pro-apoptotic cellular state in a subset of terminally differentiated plasma cells. Secondly, HSPA1B_PC are not merely passively driven toward apoptosis; rather, they actively execute a program that suppresses anti-tumor immunity. Our study uncovered several potential mechanisms by which these cells influence neighboring immune populations. Notably, HSPA1B_PC exhibited high expression of TGFB3, a potent immunosuppressive cytokine known to enhance Treg function while dampening cytotoxic T cell response [29]. In addition, HSPA1B_PC upregulated inhibitory molecules such as CD274 (PD-L1) and HAVCR2 (TIM-3), suggesting the capacity to directly suppress activated T cells via cell–cell contact and promote T cell exhaustion [30]. Furthermore, our analysis identified strong interactions between SPP1 secreted by HSPA1B_PC and CD44 expressed on Tregs and Th17 cells, pointing to a specific mechanism for recruiting and modulating these immunosuppressive T cell subsets. The SPP1–CD44 axis has been established as an immune checkpoint that impairs T cell activation and facilitates immune evasion [31]. Collectively, these findings support a model in which HSPA1B_PC are both products of a hostile TME and active drivers of immunosuppression. Their dual role highlights the complexity of plasma cell biology in cancer. TLSs are frequently described as favorable prognostic features and as sites of active anti-tumor immunity; emerging evidence indicates that their functional roles are heterogeneous and context dependent [5]. In our study, TLSs enriched for HSPA1B_PC appeared to differ from classical protective TLSs. HSPA1B_PC showed higher expression of CXCL chemokines and immunoregulatory cytokines, together with increased levels of inhibitory ligands such as PD-L1 and TIM-3. In addition, HSPA1B_PC displayed a higher hypoxia score, and hypoxia has been proposed to reprogram immune responses within TLSs by promoting checkpoint expression and T cell dysfunction [32]. Thus, rather than uniformly enhancing anti-tumor immunity, TLSs enriched with HSPA1B_PC may represent a distinct, potentially immunosuppressive configuration. Such TLSs could be characterized by disproportionate infiltration of Tregs and Th17 cells, a cytokine and chemokine milieu skewed toward immunoregulation, increased expression of inhibitory checkpoints, and adaptation to hypoxic stress. This study provides novel insights into B cell functionality within the CC microenvironment but has limitations. Although the integration of scRNA-seq, spatial transcriptomics, and RNA-seq provided a comprehensive view of B cell heterogeneity and their interactions in CC, the conclusions regarding the immunosuppressive function of HSPA1B_PC are primarily based on computational inference from high-throughput transcriptomic and spatial data rather than direct functional validation. We did not perform mechanistic experiments, such as in vitro co-culture assays or in vivo depletion studies, to experimentally confirm the interactions between HSPA1B_PC and immunosuppressive T cell subsets. Additionally, although spatial transcriptomics enabled colocalization analysis, direct evidence of molecular interactions, such as ligand–receptor signaling, remains to be established. Finally, further studies are necessary to confirm the clinical relevance of these findings, particularly the prognostic implications of B cell subsets and their spatial interactions. Despite these limitations, our data on ligand–receptor pairs and expression of immunosuppressive molecules provide a strong rationale for the proposed mechanisms, underscoring the need for future functional validation to definitively substantiate them. Conclusions This study provides a comprehensive multi-omics analysis of B cell heterogeneity in CC, identifying two key plasma cell subsets, MANF_PC and HSPA1B_PC, with distinct functional roles. MANF_PC, enriched in normal tissues, are associated with enhanced antibody synthesis and better prognosis, while HSPA1B_PC, predominantly found in cancerous tissues, correlate with immunosuppressive T cell subsets and poor clinical outcomes. Spatial transcriptomics further validated the colocalization of HSPA1B_PC with immunosuppressive T cell populations, highlighting their potential role in shaping an immunosuppressive TME. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 11187 KB) Acknowledgements The authors have nothing to report. Abbreviations Cervical cancer Multiplex immunofluorescence Plasma cell Single-cell RNA sequencing Tertiary lymphoid structures Tumor microenvironment Author contributions Q.Y.H., C.L., C.W., and B.C.Z. conceived the project, designed the study, and interpreted the results. Q.Y.H., X.H.L., W.X.Z, and Z.H. contributed to sample collection and clinical data collection. F.H.W. and W.H.Y. performed the data analysis and prepared the figures. Q.Y.H., X.L., X.H.L., W.X.Z, T.Y.L., and S.Q.Y. wrote the manuscript. Q.W. and Q.A. checked and embellished the figures. Q.Y.H., C.L., C.W., and B.C.Z. jointly supervised this work. All authors reviewed and approved the final manuscript. Funding This work was supported by the following grants: National Natural Science Foundation of China (81670144, 82403773), Postdoctoral Fellowship Program of CPSF (GZB20230041, 2024M760140, 2025T180600), The Medical Education Research Project of the Medical Education Branch of Chinese Medical Association and Medical Education Professional Committee of Chinese Association of Higher Education (2020B-N15414), Teaching research project of Medical Faculty of Wuhan University in 2021 (2021026), Special Project of Science and Technology Cooperation between Hubei Province and Chinese Academy of Sciences (42000021817T300000050), and Hubei Province Big Health Industry Development Special Fund (2019916000001). Data availability Data is provided within the manuscript or supplementary information files. Any additional information required to reanalyze the data reported in this paper is available from the correspondence author upon request. Declarations Conflict of interest The authors declare no competing interests. Consent for publication All authors gave their consent for publication. Ethics approval and consent to participate This study received approval from the Ethics Committee of Shandong Cancer Hospital (SDTHEC201906009) and adhered to the Declaration of Helsinki. All patients provided written informed consent. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Qingyu Huang, Fuhao Wang, Xiaohui Li, Wenxue Zou, and Zhou Huang have contributed equally. Contributor Information Bicheng Zhang, Email: bichengzhang@hotmail.com. Cong Wang, Email: wangcongsdu@163.com. Chao Liu, Email: charles.liu@whu.edu.cn. Qinyong Hu, Email: 2925829838@qq.com. References Associated Data Supplementary Materials Supplementary file1 (DOCX 11187 KB) Data Availability Statement Data is provided within the manuscript or supplementary information files. Any additional information required to reanalyze the data reported in this paper is available from the correspondence author upon request.