Integrating Post-Translational Modifications and Stemness Landscapes: A Multi-Omics Framework for Secondary Precision Intervention in Triple-Negative Breast Cancer

Integrating Post-Translational Modifications and Stemness Landscapes: A Multi-Omics Framework for Secondary Precision Intervention in Triple-Negative Breast Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Integrating Post-Translational Modifications and Stemness Landscapes: A Multi-Omics Framework for Secondary Precision Intervention in Triple-Negative Breast Cancer Shenao Qu, Jinshuang Zhu, Haozhe Huang, Haoqi Yan, Zhanghang Li, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9353905/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Purpose To address the critical unmet need for effective maintenance therapies in triple-negative breast cancer (TNBC) patients with residual disease following neoadjuvant therapy, we investigated the interplay between post-translational modifications (PTMs) and cancer stemness—a regulatory axis that remains poorly characterized in TNBC. This study aims to uncover novel therapeutic vulnerabilities and establish a precision stratification tool to improve outcomes in TNBC patients. Methods We performed an integrated multi-omics analysis to characterize the molecular crosstalk between PTM signaling and stemness features within TNBC. This led to the identification of distinct molecular subtypes based on their regulatory and immunological landscapes. A prognostic scoring system, termed Stemness-PTM (SPT), was subsequently developed and validated in independent cohorts to quantify these biological features and predict therapeutic response. Results Our profiling delineated two distinct TNBC subtypes: an immunologically active 'C1' subtype with a specific dependence on the CDK4/6 signaling axis, and a biologically aggressive 'C2' subtype. Despite its complex regulatory environment, the C1 subtype demonstrated marked sensitivity to CDK4/6 inhibitors, such as palbociclib. The SPT prognostic score effectively stratified patient risk and outcomes across independent validation cohorts, converting this molecular signature to clinical prognosis. Conclusion This study establishes a mechanistic rationale for a "secondary precision intervention" in high-risk TNBC. By positioning CDK4/6 inhibitors as a targeted maintenance therapy for patients identified by the SPT scoring system, our findings offer a translatable strategy to achieve sustained disease control and improve survival outcomes. Triple-negative breast cancer Post-translational modifications Molecular subtype CDK4/6 inhibitors Maintenance therapy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction TNBC represents the most recalcitrant subtype of breast malignancy, defined by a lack of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression[1]. Despite the integration of immune checkpoint inhibitors with neoadjuvant chemotherapy—exemplified by the KEYNOTE-355 and KEYNOTE-522 regimens—as the new standard of care[2], the clinical management of TNBC remains a formidable challenge. A significant proportion of patients fail to achieve a pathological complete response (pCR), and the absence of targeted maintenance therapies for residual disease leaves these patients vulnerable to rapid recurrence and metastasis[3, 4]. Consequently, deciphering the molecular heterogeneity of TNBC to identify actionable vulnerabilities for maintenance strategies represents an urgent unmet clinical need. The therapeutic resistance and aggressive phenotype of TNBC are increasingly attributed to the synergistic interplay between proteomic plasticity and cellular stemness. PTMs—ranging from phosphorylation and acetylation to ubiquitination—dramatically expand the functional diversity of the proteome, enabling rapid adaptation to therapeutic stress and remodeling of the tumor microenvironment (TME)[5-11]. Concurrently, cancer stem cells (CSCs) serve as the apex of intratumoral heterogeneity. Through intricate crosstalk with the TME and immune evasion mechanisms—such as the downregulation of antigen presentation and the induction of an immunosuppressive stroma—CSCs orchestrate therapy resistance and tumor relapse[12-14]. While recent evidence suggests that aberrant PTMs can reprogram stemness traits to sustain malignancy[15, 16], the precise molecular architecture governing the crosstalk between PTM machinery and stemness maintenance remains poorly understood. Specifically, how the co-regulation of post-translational modification-related genes (PTMRGs) and stemness-regulating genes (SRGs) dictates the immune landscape and therapeutic sensitivity in TNBC constitutes a significant knowledge gap. Addressing this "black box" offers a compelling opportunity: by dissecting these regulatory networks, we may uncover novel biomarkers for stratification and targets for precision intervention[17-20]. In this study, we employed an integrated multi-omics systems biology approach to delineate the crosstalk between PTM regulation and stemness in TNBC. We identified two distinct molecular subtypes—C1 and C2—characterized by divergent immunological profiles and therapeutic vulnerabilities. We demonstrate that the C1 subtype, despite its stemness features, exhibits an immunologically active microenvironment and specific dependency on the CDK4/6 signaling axis. Building on these insights, we developed and validated a SPT prognostic scoring model. Furthermore, we propose a "secondary precision intervention" framework, wherein SPT-guided stratification identifies candidates for targeted maintenance therapy—such as CDK4/6 inhibitors—thereby offering a novel paradigm to sustain disease control and improve survival outcomes in this challenging patient population. Materials and Methods Data Acquisition and Preprocessing Within The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/), only TNBC specimens exhibiting over one month of follow-up were retained. Single-cell RNA-seq data (GSE161529[21]; 33,538 cells from 8 TNBC patients) were acquired from the Gene Expression Omnibus database (GEO database, https://www.ncbi.nlm.nih.gov/gds/). SRGs (n = 925) were sourced from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb). PTMRGs (n = 808) were compiled through extensive literature review based on prior publications. Transcriptomic profiles and clinical annotations for 320 TNBC cases, serving as the training cohort, were derived from METABRIC (https://ega-archive.org/studies/EGAS00000000083); Validation cohorts comprised publicly available datasets: TCGA-TNBC (n = 113), GSE58812[22] (n = 107), and GSE37751[23] (n = 14). Integrated identifiers were mapped to official gene symbols. Only protein-coding genes were analyzed. Utilizing univariate cox regression, 862 differentially expressed genes (DEGs) linked to overall survival (OS) were identified within the METABRIC cohort. Associations among 48 selected SRGs and PTMRGs were examined using Spearman correlation. Prognostic Gene Identification and Molecular Subtyping Patients were stratified into distinct clusters via unsupervised consensus clustering ("ConsensusClusterPlus" R package). To verify cluster robustness, the procedure underwent 1000 iterations. Peak intra-group consensus determined the optimal cluster count (k = 2), yielding subtypes C1 (n = 153) and C2 (n = 167). Mechanistic insights were pursued through Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, and Gene Set Enrichment Analysis (GSEA). The 48 SRGs and PTMRGs underwent GO and GSEA. Kaplan-Meier survival analysis revealed significant differences among these subtypes, while the heatmap integrated survival distributions with evaluations of clinical feature correlations. Differential Expression Analysis and Mechanistic Investigation Expression profiles distinguishing subtypes C1 and C2 were generated using the “limma” R package. DEGs were defined by False discovery rate (FDR)-adjusted p-value 1. DEGs identified between C1 and C2 were subjected to GO, KEGG, and GSEA. Tumor Immune Microenvironment Characterization CIBERSORT quantified tumor-infiltrating immune cell populations. Immunophenotype score (IPS), incorporating MHC molecules, effector cells (EC), suppressor cells (SC), and immune checkpoints (CP), was computed. The anticancer immune status and immune cell proportions across the seven-step cancer-immunity cycle (CIC) were analyzed and visualized using the Tracking Tumor Immunophenotype platform (TIP; http://biocc.hrbmu.edu.cn/TIP/)[24] based on RNA-seq data. Immunotherapy response likelihood was estimated via the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm (http://tide.dfci.harvard.edu)[25]. The immune landscapes of subtypes C1 and C2 were evaluated using these computational approaches. Prognostic Model Construction and Validation Least absolute shrinkage and selection operator (LASSO) regression was applied to the 320 TNBC cases derived from the METABRIC database, which identified the top 20 genes from the initial 48. Subsequently, 10 distinct machine learning algorithms (StepCox, Ridge, plsRcox, Lasso, CoxBoost, Enet, GBMs, SVMs, SuperPC, RSF) generated 117 algorithm combinations ("Mime1" R package). The "ML.Dev.Prog.Sig" function calculated the concordance index (C-index) for each combination; the model achieving the highest C-index was selected, incorporating 10 genes to formulate the SPT prognostic scoring system. Univariate Cox regression validated all 10 SPT genes. Based on SPT scores, patients were dichotomized into high-risk and low-risk categories. Prognostic performance was appraised using the risk score distribution, Kaplan-Meier survival analysis and time-dependent receiver operating characteristic (ROC) curves (AUCs for 1 year) across training dataset and 3 validation datasets. Genetic Alterations Analysis and Drug Response Prediction Tumor mutational burden (TMB) within subgroups was calculated with the "maftools" R package. Sensitivity profiles for 12 targeted therapeutics were forecasted using the Genomics of Drug Sensitivity in Cancer (GDSC) database via the R package "oncoPredict". Correlations between the top 10 prognostic genes and predicted drug sensitivity were further assessed. Reagents and Chemicals Carboplatin (Cat. No. HY-17393) and Palbociclib (Cat. No. HY-50767) were purchased from MedChemExpress (MCE, Monmouth Junction, NJ, USA). The purity of all test compounds was >99%. Stock solutions were prepared in dimethyl sulfoxide (DMSO; Macklin, Shanghai, China) and stored at 4°C protected from light. The Cell Counting Kit-8 (CCK-8) was obtained from TransGen Biotech (TransDetect® Cell Counting Kit, FC101-04, Beijing, China). Cell Lines and Culture Conditions The human breast cancer cell lines MDA-MB-231 and MDA-MB-468 were provided by GemPharmatech (Chengdu, China). Cells were cultured in their respective standard media supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. The cultures were maintained in a humidified incubator at 37°C with 5% CO 2 . All experiments were performed using cells in the logarithmic growth phase. Cell Viability Assay The in vitro antitumor activity of the test compounds was evaluated using the CCK-8 assay. Briefly, MDA-MB-231 and MDA-MB-468 cells were harvested and seeded into 96-well plates at a density of 5 times 10^3 cells/well in 150 muL of culture medium. The plates were incubated at 37°C for 24 h to allow for cell attachment. Following the initial incubation, the culture medium was replaced with serum-free medium containing serial dilutions of the test compounds or the vehicle control (DMSO). The treatment concentrations were set as follows: Carboplatin: 0.08–8 uM for MDA-MB-231 cells; 0.2–20 uM for MDA-MB-468 cells. Palbociclib: 0.5–50 uM for both cell lines. Five replicate wells were set for each concentration group (n=5). After 48 h of drug exposure, 10 uL of CCK-8 reagent was added to each well, and the plates were incubated for an additional period (typically 1–4 h) at 37°C according to the manufacturer’s instructions. The absorbance was measured using a microplate reader (iMark™ Microplate Absorbance Reader, BIO-RAD, Hercules, CA, USA). Single-cell Transcriptome Analysis Raw single-cell RNA sequencing (scRNA-seq) data underwent quality control and normalization ("Seurat" R package)[26]. Cells meeting the following criteria were retained: >1000 UMIs, expression of 200-6000 unique genes, and <20% mitochondrial gene contribution. Cell clusters were visualized via t-distributed stochastic neighbor embedding (t-SNE). Cluster annotation leveraged canonical marker genes and cluster-specific DEGs. Malignant epithelial cells were discerned using inferCNV, employing B cells/endothelial cells as reference. SPT scores assigned samples to high-risk (n=4) or low-risk (n=4) groups; cell type proportions were contrasted between these groups. Cellular communication networks were inferred with CellChat. Predictive Nomogram Construction A nomogram integrating survival time, status, age, stage, SPT score, and grade was built ("rms" R package) to visualize TNBC patient prognosis. Nomogram performance was gauged using Harrell's C-index, calibration curves, and decision curve analysis (DCA). Spatial transcriptome analysis Spatial transcriptomic profiling was performed using the 10x Genomics Visium platform, with scRNA-seq data serving as the reference. Quality control of the scRNA-seq data was conducted by filtering cells based on gene detection counts, unique molecular identifier (UMI) abundance, and mitochondrial gene expression proportion. Cellular deconvolution of each Visium spot was carried out using the SPOTlight package to estimate the relative abundances of annotated cell types based on the scRNA-seq reference.In parallel, cell type enrichment scores were computed with the Cottrazm package. Specifically, the top 25 most specifically expressed genes for each cell type were selected from the scRNA-seq reference to construct cell type–specific gene signatures. Enrichment scores for each spatial spot were then calculated using the get_enrichment_matrix and enrichment_analysis functions.Based on malignant cell scores derived from deconvolution, spatial spots were categorized into three microregion types: malignant (score > 0) and normal (score = 0). Differential gene expression between malignant and normal microregions was assessed using the Wilcoxon rank-sum test. The dominant cell type for each spot, defined as the cell type with the highest inferred abundance, was visualized with Seurat’s SpatialDimPlot. Spatial expression patterns of candidate genes (PCGF1, KDM5B, and TK1) were displayed using Seurat’s SpatialFeaturePlot. Furthermore, Spearman correlation analysis was employed to evaluate relationships between cell type abundances and candidate gene expression, as well as intercellular interactions across all spatial spots. Correlation results were visualized using the "linkET " R package. The spatial transcriptomics data used in this study (sample ID: Sample_093C) were obtained from a publicly available database (GSE210616, GSM6433589)[27]. Statistical analysis Analyses employed GraphPad Prism (8.4.3), Python (3.10) and R software (4.2.2). Two-group comparisons used paired two-tailed Student's t-tests or Mann-Whitney-Wilcoxon tests. Multi-group comparisons utilized ANOVA or Kruskal-Wallis rank-sum tests. Clinical characteristic associations were tested via chi-square. Statistical significance was defined as p < 0.05. In the drug sensitivity assay, data are presented as mean ± standard deviation (Mean ± SD). The half-maximal inhibitory concentration (IC₅₀) was calculated using nonlinear regression analysis. Depending on the distribution characteristics of the data, between-group differences were assessed for statistical significance using Student's t‑test or analysis of variance (ANOVA). A P‑value < 0.05 was considered statistically significant. Results Identification of SRGs and PTMRGs in TNBC To systematically investigate the interplay between stemness and post-translational modifications, we developed a multi-omics analytical framework (Supplementary Figure 1). By intersecting 925 stemness-related genes (SRGs) and 862 post-translational modification-related genes (PTMRGs) with survival-associated genes from the METABRIC cohort (n = 320), we identified a core set of 48 prognostic genes, comprising 23 SRGs and 25 PTMRGs (Figure 1A). These genes exhibited extensive correlations and intricate intergenic interactions (Figure 1B). Using unsupervised consensus clustering based on this gene set, we stratified the TNBC cohort into two distinct molecular subtypes: Cluster 1 (C1, n = 153) and Cluster 2 (C2, n = 167), with k = 2 providing optimal clustering stability (Figure 1C; Supplementary Figures 2A–C). Functional enrichment analysis suggested that these signature genes are involved in key biological processes, including ubiquitin-like protein transferase activity and cytoskeletal organization (Figures 1D–E). Survival analysis revealed that the C1 subtype was associated with significantly superior overall survival (OS) compared to C2 (HR = 2.24, 95% CI: 1.63–3.08, P < 0.0001; Figure 1F). Clinical profiling showed that the C1 subtype was enriched for early-stage (Stage I) patients and correlated with favorable survival outcomes (Table 1; Figure 1G). Moreover, Gene Set Enrichment Analysis (GSEA) uncovered marked differences in immune-related signaling pathways between the two subtypes, pointing to intrinsic variations in their tumor microenvironments (Figure 1H; Supplementary Figures 2D–E). In summary, we successfully defined two TNBC molecular subtypes based on stemness and PTM-related genes that exhibit significant differences in clinical characteristics, prognostic outcomes, and immune pathway activity. Schematic overview of the multi-omics framework integrating post-translational modifications and stemness landscapes in triple negative breast cancer. Sequencing data (DNA/RNA) are used to identify stemness-related genes and PTM-related genes, yielding 48 candidate genes. Consensus clustering of TNBC samples classifies two distinct molecular subtypes: C1 and C2. The C1 subtype is characterized by high PD-L1 expression, CD8+ T cell exhaustion, high tumor mutational burden, active PTM signaling, and high expression of the CDK4/6 signaling axis. The C2 subtype presents an immune-desert or stem-like phenotype, with tumor mutational burden correlating with M2 macrophage or MDSC infiltration, and exhibits higher expression of stemness markers (PCGF1, KDM5B, TK1) in cancer cells. The SPT score model is constructed using LASSO regression screening (20 genes) and machine learning (10-gene model) for risk scoring (high/low risk). Spatial transcriptomics and single-cell validation reveal high PCGF1 expression in malignant regions and PCGF1 expression in immune cell regions. Clinical decision for secondary precision intervention: low-risk C1 subtype receives CDK4/6 inhibitors (e.g., palbociclib), while high-risk C2 subtype receives novel combination therapies, leading to improved survival. TNBC: triple negative breast cancer; PTM: post-translational modification; TMB: tumor mutational burden; MDSC: myeloid-derived suppressor cells; SPT: Stemness-PTM score. Table 1. Baseline characteristics of TNBC patients in the METABRIC cohort, stratified by clusters C1 and C2. cCharacteristic C1 N = 109 1 C2 N = 114 1 p-value 2 Age 0.14 50 95 (87%) 91 (80%) Type 0.083 Invasive Ductal Carcinoma 95 (87%) 103 (90%) Invasive Lobular Carcinoma 5 (4.6%) 9 (7.9%) Invasive Mixed Mucinous Carcinoma 1 (0.9%) 0 (0%) Mixed Ductal and Lobular Carcinoma 8 (7.3%) 2 (1.8%) Menopausal State 0.5 Post 70 (64%) 78 (68%) Pre 39 (36%) 36 (32%) Stage 0.2 I 38 (35%) 27 (24%) II 61 (56%) 72 (63%) III 10 (9.2%) 15 (13%) Grade 0.5 I 3 (2.8%) 1 (0.9%) II 14 (13%) 12 (11%) III 92 (84%) 101 (89%) OS Status <0.001 Alive 63 (58%) 36 (32%) Dead 46 (42%) 78 (68%) RFS Status <0.001 Not recurred 79 (72%) 51 (45%) Recurred 30 (28%) 63 (55%) Chemotherapy 55 (50%) 59 (52%) 0.8 Hormone Therapy 34 (31%) 31 (27%) 0.5 Radio Therapy 82 (75%) 84 (74%) 0.8 1 n (% ) 2 Pearson's Chi-squared test; Fisher's exact test Immune landscape of distinct molecular subtypes in TNBC Given the significant survival disparities observed between the C1 and C2 subtypes, we subsequently conducted an in-depth analysis of the tumor immune microenvironment (TIME) to uncover the immunological basis driving these divergent clinical outcomes. CIBERSORT analysis revealed that the C1 subtype was distinguished by robust infiltration of adaptive immune cells, particularly CD8⁺ T cells, whereas the C2 subtype exhibited an immune-desert phenotype (Figure 2A). To assess immunogenicity, we applied the Immunophenoscore (IPS) algorithm, which showed that C1 subtype had elevated scores for MHC molecules and effector cells (EC), consistent with a "hot" tumor phenotype (Figure 2B–C). Interestingly, although the aggregate immune checkpoint (CP) score derived from IPS was lower in C1, gene-level analysis revealed a marked upregulation of Programmed death-ligand 1 (PD-L1) in this subtype (Figure 2D). This apparent divergence suggests that immune evasion in C1 is not driven by broad inhibitory checkpoint upregulation—which would elevate the CP score—but rather by a specific adaptive resistance mechanism mediated predominantly through the PD-L1 axis. Further dissection of the cancer-immunity cycle (CIC) demonstrated that C1 exhibited enhanced activity across key steps, including antigen release and T-cell recognition (Figure 2E–F). However, cytotoxic killing efficiency remained disproportionately low relative to the degree of immune infiltration, supporting the hypothesis of PD-L1-mediated functional T-cell inhibition. TIDE analysis corroborated that although C1 tumors exhibited lower overall immune evasion potential (reflected by lower TIDE scores), they were characterized by T-cell dysfunction (Figure 2G). Moreover, while the C1 subtype showed higher immunotherapy response rates (Figure 2H) and CTL levels (Figure 2I), the C2 subtype demonstrated greater immune benefit ratios (Figure 2J), with its microenvironment dominated by immune exclusion and immunosuppressive populations such as M2 macrophages and MDSCs (Figure 2K-L). These findings collectively delineate two starkly different immune landscapes: the C1 subtype presents a PD-L1-driven immune-active but exhausted phenotype, while the C2 subtype is characterized by immune exclusion and suppression. Construction and validation of the SPT signature score Following the identification of biological differences between the molecular subtypes, we sought to develop a clinically actionable quantitative tool to precisely assess mortality risk in TNBC patients by integrating key SRGs and PTMRGs. Firstly, we reduced the gene set to 20 candidates using LASSO regression (Figures 3A-B), then evaluated 117 algorithm combinations using integrated machine learning across a training set and three independent validation cohorts (TCGA, GSE58812, GSE37751), ultimately finalizing the SPT prognostic scoring system composed of 10 genes (Figure 3C). Patients were stratified into high- and low-risk groups based on their SPT scores. The predictive risk distribution was visualized in Figure 3D-G, where the bottom heatmap revealed the distinct expression patterns of the 10 core genes that distinguished high-risk from low-risk individuals. For instance, genes such as DMX2 showed decreased expression in the high-risk group, whereas DPM2 exhibited increased expression in the high-risk group. The model demonstrated robust prognostic discrimination, with high-risk patients exhibiting significantly worse OS across all four cohorts (METABRIC: HR = 1.96, P < 0.0001; TCGA: HR = 3.04, P = 0.028; GSE58812: HR = 2.59, P = 0.011; GSE37751: HR = 5.25, P = 0.035; Figures 4H–K). Time-dependent ROC analysis confirmed the model's predictive stability, yielding 1-year AUC ranging from 0.75 to 0.92 across datasets (Figures 4L), validating the SPT score as a reliable prognostic instrument. This SPT score integrated key SRGs and PTMRGs (Figure 4M). Extensive validation across multiple independent cohorts consistently demonstrates that the SPT risk score system is highly robust and serves as an effective tool for predicting long-term survival outcomes in TNBC patients. Differential responses of C1 and C2 to targeted and chemotherapy drugs To link molecular subtypes to therapeutic vulnerabilities, we analyzed somatic mutational landscapes. The C1 subtype exhibited a significantly higher TMB compared to C2. While TP53 mutations were ubiquitous in both subtypes (~80%), other drivers such as MLL2 showed lower mutation frequencies (Figures 4A–B). Pharmacogenomic profiling revealed that the transcriptomic differences between C1 and C2 correlated with distinct drug sensitivity patterns (Figure 4C). We mapped subtype-specific gene signatures to drug response data (Figure 4D) and identified the CDK4/6 inhibitor palbociclib as a candidate agent. To validate this in vitro , MDA-MB-468 and MDA-MB-231 cell lines were selected as representative models for the C1 and C2 subtypes, respectively. Drug sensitivity assays confirmed that the C1-like MDA-MB-468 cells were significantly more sensitive to palbociclib compared to the C2-like MDA-MB-231 cells (Figures 4E–F), establishing a preclinical rationale for targeting the CDK4/6 axis in the C1 population.The results of the in vitro experiments confirm that the C1 subtype is highly sensitive to CDK4/6 inhibitors, providing prospective experimental evidence for implementing precision maintenance therapy. ScRNA-seq analysis of immune cell infiltration To further validate these findings, we analyzed the single-cell RNA sequencing dataset GSE161529, which comprised 33,538 cells from eight triple-negative breast cancer (TNBC) patients after quality control and filtering. Unsupervised clustering categorized these cells into six major populations: B cells, T/NK cells, myeloid cells, macrophages, cancer-associated fibroblasts (CAFs), and epithelial cells (Figures 5A–C). To distinguish benign from malignant epithelial cells, we extracted epithelial cells along with B cells and endothelial cells as reference groups, and applied infercnv to assess the degree of malignant transformation. This analysis further subdivided epithelial cells into normal mammary epithelial cells and tumor cells (Figures 5D–F). Using pseudo-bulk analysis, we computed the SPT score for each sample and stratified the eight TNBC cases into high-risk (n = 4) and low-risk (n = 4) groups. The high-risk group exhibited a significantly higher proportion of tumor cells compared to the low-risk group (Figure 5G). Cell-cell communication analysis indicated that both the number and strength of inferred interactions were significantly greater in the low-risk group, suggesting more active intercellular crosstalk (Figure 5H-I). Furthermore, the interaction network analysis revealed distinct communication patterns between the two groups: the high-risk group exhibited more complex and intensive interactions involving normal epithelial cells, CAFs, and endothelial cells, which may underlie its adverse prognosis. In contrast, the low-risk group showed enhanced signaling among tumor cells, CAFs, and endothelial cells (Figure 5J-K). Finally, we analyzed the information flow of ligand–receptor pairs and observed differential activity between risk groups. Genes such as IFN-β, TIGIT, and NECTIN displayed higher information flow in the high-risk group, while ICAM, VEGF, and SELE were more active in the low-risk group (Figure 5L), suggesting their potential roles in shaping distinct tumor microenvironments associated with prognosis. These results indicate that high-risk tumor lacks global immune signal exchange and instead rely on inhibitory pathways like TIGIT to maintain their immune-excluded and treatment-resistant phenotype. Construction of a prognostic nomogram To assess the predictive performance of our model, we conducted a comparative evaluation against established TNBC models. Based on the concordance index (C-index) across multiple datasets (METABRIC, TCGA, GSE58812, and GSE37751), our model achieved the highest predictive accuracy in the METABRIC cohort and was ranked among the top models in the independent validation sets, confirming its superior and robust prognostic value. (Figure 6A). To facilitate clinical translation, we developed two nomograms for prognostic prediction. The first nomogram was constructed for patients in the METABRIC TNBC cohort based on clinical variables including age, stage, risk score, and grade, along with 1-, 3-, and 5-year survival probabilities (Figure 6B). The other nomogram was established for the entire METABRIC cohort, incorporating the same clinical variables and extending the prediction to 1-, 3-, 5-, and 10-year survival probabilities (Figure 6C). The calibration curve demonstrated excellent agreement between the nomogram-predicted and observed outcomes, indicating high predictive accuracy (Figure 6D). Furthermore, decision curve analysis (DCA) showed that the nomogram provided a greater net clinical benefit than the use of clinical features alone at both 3.5 and 5 years (Figures 6E–F). Strongly support the potential of our model as a reliable predictive tool with clinical applicability. Single-gene analysis and spatial transcriptomic analysis To investigate the potential mechanisms through which the SPT score predicts TNBC prognosis, we first performed single-gene expression and survival analysis across the C1 and C2 subtypes (Supplement Figure 3). Based on these findings, we selected three key genes—PCGF1, TK1, and KDM5B—for further protein-protein interaction analysis. In TNBC, expression levels of PCGF1, TK1, and KDM5B were consistently lower in the C1 subtype but elevated in C2 (Figure 7A–C). Kaplan–Meier survival analysis demonstrated that high expression of each gene was significantly associated with poorer overall survival: PCGF1 (HR = 1.59, 95% CI: 1.17–2.17, p = 0.0027; Figure 7D), KDM5B (HR = 1.52, 95% CI: 1.12–2.07, p = 0.0068; Figure 7E), and TK1 (HR = 1.48, 95% CI: 1.09–2.01, p = 0.012; Figure 7F). Functional annotation revealed distinct biological roles for these genes: PCGF1-associated proteins were primarily involved in immune-related regulation, extracellular matrix organization, cellular structure, and receptor–ligand binding (Figure 7G); KDM5B-related proteins were mainly enriched in cell proliferation (Figure 7H); and TK1-interacting partners were implicated in cell proliferation, DNA synthesis, and repair pathways (Figure 7I). Furthermore, immunohistochemistry (IHC) data from TCGA confirmed that both PCGF1 and KDM5B protein levels were significantly upregulated in breast tumor tissues compared with adjacent normal controls (Figures 7G–O). The construction of the nomogram confirms the predictive value of the model in clinical practice, while spatial localization analysis identifies PCGF1 , TK1 , and KDM5B as key spatial regulators driving the immune-excluded phenotype (Figures 7P–R). It not only explains the biological basis of the SPT score from a mechanistic perspective, but also provides a theoretical basis for its use as a prognostic predictor and potential therapeutic target. Discussion Deciphering the molecular heterogeneity of TNBC is pivotal for overcoming the therapeutic impasse characterized by drug resistance and high recurrence rates. In this study, we integrated the regulatory landscapes of PTMs and cancer stemness—two fundamental drivers of tumor plasticity—to delineate a novel classification framework. Our multi-omics stratification revealed two biologically distinct subtypes: an immunologically active but dysfunctional "C1" subtype and a stem-like, immune-desert "C2" subtype. This dichotomy not only clarifies the intrinsic variability in TNBC prognosis but also establishes a mechanistic rationale for a "secondary precision intervention" strategy, specifically positioning CDK4/6 inhibitors as a potential maintenance therapy for the C1 population. A central finding of our work is the paradoxical immune landscape of the C1 subtype. While characterized by robust infiltration of cytotoxic T lymphocytes (CTLs) and a high TMB—features typically associated with favorable immunotherapeutic responses—the C1 microenvironment exhibits profound functional exhaustion. Our analysis of the "Cancer-Immunity Cycle" indicates that while T cell recruitment is efficient, the terminal effector phase is compromised. This aligns with the "hot but exhausted" phenotype, where high expression of immune checkpoints (e.g., PD-L1) and chronic antigen stimulation drive T cells into a dysfunctional state [28, 29]. Mechanistically, we postulate that aberrant PTMs, such as glycosylation of PD-L1, may stabilize checkpoint proteins to subvert CTL cytotoxicity[30, 31], thereby creating a microenvironment that is infiltrated yet immunosuppressed. Conversely, the C2 subtype recapitulates an "immune-excluded" or "desert" phenotype, driven by stemness-associated crosstalk that actively repels immune infiltration, consistent with the aggressive, therapy-resistant nature of CSCs[32]. This stemness-associated crosstalk is mediated by dynamic interactions with the TME, characterized by hypoxia, inflammatory signals, and extracellular matrix components, ultimately contributing to therapy resistance and tumor relapse [33, 34]. Furthermore, conventional therapies such as chemotherapy and radiation, which predominantly target rapidly proliferating tumor cells, are ineffective at eliminating CSCs[35]. The delineation of distinct therapeutic vulnerabilities between these subtypes represents a significant translational advance. We demonstrate that the C1 subtype, represented by the MDA-MB-468 model, exhibits specific sensitivity to CDK4/6 inhibitors (e.g., palbociclib). This finding challenges the conventional restriction of CDK4/6 inhibitors to luminal breast cancer and supports emerging evidence of their efficacy in specific TNBC niches [36]. Mechanistically, the sensitivity of C1 tumors likely stems from a dual dependency: a reliance on rapid cell cycle progression driven by high mutational load, rendering them susceptible to cell cycle blockade, and an underlying immune plasticity. Notably, CDK4/6 inhibitors possess distinct immunomodulatory properties capable of stimulating antigen presentation and reducing regulatory T cell proliferation [37], which may reinvigorate the "hot but exhausted" T cell phenotype we observed in C1 tumors. Post-neoadjuvant therapy, this molecular stratification framework provides a compelling rationale to identify candidates—specifically those harboring C1-like features—who may benefit from CDK4/6 inhibitor-based maintenance regimens, thereby sustaining disease control where standard chemotherapy reaches its limit[38]. At the molecular level, our SPT prognostic signature elucidates the genetic architecture underpinning these phenotypes. The signature integrates key regulators such as PCGF1 [39], TK1 [40], and KDM5B [41], connecting epigenetic remodeling with cell cycle kinetics. For instance, PCGF1 (a component of the PRC1 complex) and KDM5B (a histone demethylase) are implicated in maintaining the plastic, undifferentiated state of CSCs while simultaneously orchestrating chromatin accessibility for immune-evasive gene expression [42, 43]. The spatial transcriptomic co-localization of these genes with malignant clusters further underscores their role in shaping the tumor-intrinsic signaling that dictates the surrounding immune conte xt[44, 45]. For instance, TK1—a key node within our signature—has been identified in previous studies as a potential therapeutic target in Treg-mediated immunosuppression. Moreover, the immunomodulatory factors MUC1-C and CXCL9, which are known to influence T cell exhaustion and macrophage differentiation within the TME, may represent downstream effectors or interacting partners in this regulatory network. Despite these promising insights, our study has limitations inherent to in silico analyses. While we cross-validated our findings across multiple independent cohorts (METABRIC, TCGA, GEO) to ensure robustness, the retrospective nature of these datasets introduces potential selection bias. Additionally, while our pharmacogenomic predictions were validated in cell line models, the complex interplay between PTMs and the immune microenvironment requires further elucidation in immunocompetent in vivo models. Future studies should focus on mapping the precise PTM sites governing the identified SRGs and validating the synergistic efficacy of CDK4/6 inhibitors with immunotherapy in preclinical trials. In conclusion, our systems biology approach unveils a novel TNBC taxonomy governed by the crosstalk between PTMs and stemness. By defining a subtype characterized by immune exhaustion and CDK4/6 dependency, we provide a blueprint for precision maintenance therapy. The SPT scoring system serves as a clinically translatable tool to stratify patients, offering a new avenue to dismantle the lethal heterogeneity of TNBC and improve long-term survival. Conclusion In summary, we classified patients with TNBC into distinct molecular subtypes based on SRGs and PTMRGs, and elucidated the associated immunological and genetic profiles of each subtype. Furthermore, we developed a SPT scoring model capable of predicting both patient prognosis and response to immunotherapy. This SPT score has been validated across multiple independent datasets and shows promising potential for guiding individualized and precise clinical management of TNBC. Abbreviations TNBC: Triple-negative breast cancer; ER: Estrogen receptor; PR: Progesterone receptor; HER2: Human epidermal growth factor receptor 2; PD-L1: Programmed death-ligand 1; PTMs: Post-translational modifications; TME: The tumor microenvironment; CSCs: Cancer stem cells; TCGA: The Cancer Genome Atlas Program; GEO: Gene Expression Omnibus; SRGs: Stemness-associated genes; PTMRGs: Post-translational modification-related genes; DEGs: Differentially expressed genes; OS: Overall survival; FDR: False discovery rate; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; GSEA: Gene Set Enrichment Analysis; IPS: Immune Proportion Score; EC: Effector cell; SC: Suppressor cell; CP: Checkpoint; CIC: Cancer-immunity cycle; TIP: Tumor Immunophenotype; TIDE: Tumor Immune Dysfunction and Exclusion; TMB: Tumor mutational burden; GDSC: Genomics of Drug Sensitivity in Cancer; LASSO: The least absolute shrinkage and selection operator; ROC: Receiver operating characteristic; scRNA-seq: single-cell RNA-seq; t-SNE: t-distributed stochastic neighbor embedding; GSEA: Gene set enrichment analysis; CTL: Cytotoxic T lymphocyte; TAM: Tumor-associated macrophages; MDSCs: Myeloid-derived suppressor cells; HRs: Hazard ratios; AUC: Area under curve; CAFs: Cancer-associated fibroblasts; DCA: Decision curve analysis; IHC: Immunohistochemistry; ICI: Immune checkpoint inhibitors; TIME: Tumor Immune Microenvironment; SPT: Stemness-PTM Declarations Authorship contribution statement Huimin Zhang : Conceptualization, Writing original draft, Data curation, Funding acquisition. Shenao Qu : Writing original draft, Methodology, Data curation, Validation. Jinshuang Zhu : Methodology, Data curation. Haozhe Huang : Methodology, Data analysis. Haoqi Yan and Zhanghang Li : Data curation. Ju Zhang and Wenya Xue : Validation. Funding This study was funded by the Key Research and Development Project of Shaanxi Province (2025GH-YBXM-069). Competing interests Not applicable. Ethical Approval and Consent to participate Not applicable. Data availability statement Publicly available datasets were analyzed in this study. These data can be found in the following databases: TCGA (https://portal.gdc.cancer.gov/), GEO (https://www.ncbi.nlm.nih.gov/), and METABRIC (https://ega-archive.org/). Mechanistic cartoons were created with BioRender (https://biorender.com/). Additional Declarations No competing interests reported. Supplementary Files supply1.png Supply2.tif Supply3.tif Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 15 May, 2026 Reviewers invited by journal 28 Apr, 2026 Editor assigned by journal 14 Apr, 2026 Submission checks completed at journal 14 Apr, 2026 First submitted to journal 08 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9353905","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":631791628,"identity":"2eed3300-befa-4852-ad70-e3854dc6bdc9","order_by":0,"name":"Shenao Qu","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Shenao","middleName":"","lastName":"Qu","suffix":""},{"id":631791629,"identity":"6a0f9267-1d75-4a2a-9889-5fc469c96cb6","order_by":1,"name":"Jinshuang Zhu","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Jinshuang","middleName":"","lastName":"Zhu","suffix":""},{"id":631791630,"identity":"09490547-b37b-4e03-a18a-69176baf7756","order_by":2,"name":"Haozhe Huang","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Haozhe","middleName":"","lastName":"Huang","suffix":""},{"id":631791634,"identity":"81a8c79f-a6c9-4d19-93dc-6f92980ec459","order_by":3,"name":"Haoqi Yan","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Haoqi","middleName":"","lastName":"Yan","suffix":""},{"id":631791635,"identity":"88755e38-b6a1-4e3f-9b46-c3938e3b68f0","order_by":4,"name":"Zhanghang Li","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Zhanghang","middleName":"","lastName":"Li","suffix":""},{"id":631791636,"identity":"5f57eb13-0bea-48b3-9261-e7cd3d3fb827","order_by":5,"name":"Wenya Xue","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Wenya","middleName":"","lastName":"Xue","suffix":""},{"id":631791643,"identity":"e939e052-8ac1-4c08-ab7e-3dc7a2ff9b2f","order_by":6,"name":"Ju Zhang","email":"","orcid":"","institution":"Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Ju","middleName":"","lastName":"Zhang","suffix":""},{"id":631791644,"identity":"ffd4260a-c465-4992-8e55-05ece00dac52","order_by":7,"name":"Huimin Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+UlEQVRIiWNgGAWjYBACAwbGBoYEMAMIPjCwgUUliNbCOIOBTYIILUgMZh6oarxazCWSWzc83MFgb85+9vBr2za+OoMDzAdv8zDY5eHSYjkjse1G4hkGZsuevDTrnDNsEgYH2JKteRiSi3E67AZISxsDm8GBHDPjnAqQFh4zaR6GA4kNBLTwGJx/Y2ZsYQDSwv+NKC0SBjdyjB8zQGxhw6/lzEOQFgkDgxtvzBh7zrBJzjzMZmw5xyAZt5bj6c9u/myzsTc4n2P84WfbMX6+480Pb7ypsMOpBQrAEQGKxmPA2AEbhV89DDB/YGCoIU7pKBgFo2AUjCgAAFhwVPhCZirtAAAAAElFTkSuQmCC","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":true,"prefix":"","firstName":"Huimin","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2026-04-08 08:25:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9353905/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9353905/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108978691,"identity":"618ed428-8fb9-416e-a474-c8c8f380ae9f","added_by":"auto","created_at":"2026-05-11 11:47:26","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":44804720,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIdentification of molecular subtypes based on SRGs and PTMRGs. \u003c/strong\u003e(A) Venn diagrams showing the intersection of OS-associated genes with SRGs (green) and PTMRGs (red). (B) Correlation heatmap of the 48 identified candidate genes. (C) Consensus clustering matrix (k = 2) identifying two distinct molecular subtypes (C1 and C2). (D, E) GO enrichment analysis of the 48 signature genes. (F) Kaplan-Meier survival analysis comparing OS between C1 and C2 subtypes (Log-rank test). (G) Heatmap displaying the expression landscape of the 48 genes and their association with clinicopathological features. (H) Ridge plot of GSEA highlighting differentially enriched hallmark pathways between C1 and C2. OS: overall survival; SRGs: Stemness-associated genes; PTMRGs: Post-translational modification-related genes; GO: Gene Ontology; GSEA: Gene Set Enrichment Analysis\u003c/p\u003e","description":"","filename":"FIGURE1.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/ceb121f6006c1d2d1a4a18ee.png"},{"id":108978711,"identity":"80bc9159-d63d-445d-ba8e-5797e3e50501","added_by":"auto","created_at":"2026-05-11 11:47:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":34584369,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eImmune landscape characterization of TNBC subtypes.\u003c/strong\u003e (A) Boxplot of relative immune cell infiltration levels estimated by CIBERSORT. (B) Violin plots comparing IPS components: MHC molecules, EC, SC, and CP. (C) Comparison of aggregate IPS scores between C1 and C2. (D) Comparison of PD-L1 expression based on TIDE. (E)Radar plot depicting the activity of the cancer-immunity cycle steps. (F) Summary of enrichment scores for each step of the cancer-immunity cycle. (G) TIDE analysis showing scores for TIDE, dysfunction, and exclusion. (H–J) Predicted immunotherapy response rates (G), CTL levels (H), and immune benefit ratios (I) based on TIDE. (K–L) Comparison of M2 macrophage scores (K) and MDSC scores (L). P values derived from Wilcoxon rank-sum test or Chi-square test; **P \u0026lt; 0.01, ***P \u0026lt; 0.001, ****P \u0026lt; 0.0001. IPS: Immunophenoscore; EC: Effector Cells; SC: Suppressor Cells; CP: Checkpoints; TIDE: Tumor Immune Dysfunction and Exclusion; CTL; cytotoxic T lymphocyte\u003c/p\u003e","description":"","filename":"FIGURE2.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/ccf064b2bf16c6ce6d121586.png"},{"id":108979751,"identity":"99b74666-e3b1-4cf0-aa12-54ee06c8e704","added_by":"auto","created_at":"2026-05-11 12:01:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":28589720,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConstruction and validation of the SPT prognostic signature.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A, B) LASSO regression analysis showing coefficient profiles (A) and partial likelihood deviance (B). (C) Performance heatmap of 117 machine learning algorithms evaluated to optimize the model. (D–G) Risk score distribution, survival status, and heatmaps of signature gene expression in METABRIC (D), TCGA (E), GSE58812 (F), and GSE37751 (G) cohorts. (H–K) Kaplan-Meier OS curves for high- vs. low-risk groups in METABRIC (H), TCGA (I), GSE58812 (J), and GSE37751 (K) cohorts. (L) Time-dependent ROC curves for 1-year survival predictions. (M) Forest plot of the multivariate Cox regression coefficients for the final 10-gene signature. LASSO: The least absolute shrinkage and selection operator; OS: overall survival; ROC: Receiver operating characteristic\u003c/p\u003e","description":"","filename":"FIGURE3.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/5ffb2c3d50c8e82fe3d950e5.png"},{"id":108978712,"identity":"129841f7-1eec-42dd-94f5-68d029865ba4","added_by":"auto","created_at":"2026-05-11 11:47:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":41684659,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePharmacogenomic profiling and preclinical validation. \u003c/strong\u003e(A, B) Oncoplots summarizing the somatic mutation landscape and TMB in C1 (A) and C2 (B) subtypes. (C) Differential drug sensitivity predicted for C1 and C2 subtypes (IC50 values). (D) Correlation matrix between subtype-specific gene expression and drug response. (E, F) In vitro dose-response curves for Carboplatin (E) and Palbociclib (F) in MDA-MB-468 (C1 model) and MDA-MB-231 (C2 model) cell lines. Data represent mean ± SD. TMB: tumor mutational burden\u003c/p\u003e","description":"","filename":"FIGURE4.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/53a34e1bf6b1eb45b2fe2b56.png"},{"id":108978725,"identity":"e4eea46b-0e48-41ea-a555-d7f37118d47f","added_by":"auto","created_at":"2026-05-11 11:48:07","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":33937232,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-cell resolution of the tumor microenvironment. \u003c/strong\u003e(A–C) t-SNE visualization of 33,538 cells from the GSE161529 dataset, colored by Harmony clusters (A), patient origin (B), and cell type (C). (D) Heatmap of inferCNV scores distinguishing malignant (tumor) from non-malignant cells. (E) t-SNE plot highlighting normal vs. tumor cells. (F) t-SNE plot colored by CNV score. (G) Proportion of cell types in high-risk vs. low-risk groups defined by SPT score. (H, I) Bar charts showing the number (H) and strength (I) of inferred cell-cell interactions. (J, K) Circle plots visualizing cell-cell communication networks (J) and differential interaction strength (K). (L) Information flow of signaling pathways enriched in high-risk (red) vs. low-risk (green) groups. \u0026nbsp;t-SNE: t-distributed Stochastic neighbor embedding; CNV: copy number variation; SPT: Stemness-PTM\u003c/p\u003e","description":"","filename":"FIGURE5.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/d8f26344f66fa61b59a10f5d.png"},{"id":108978704,"identity":"94603193-4983-471a-ac14-22ef788db6f8","added_by":"auto","created_at":"2026-05-11 11:47:40","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":43084020,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDevelopment of a clinical prognostic nomogram. \u003c/strong\u003e(A) Comparison of C-index values between the SPT model and established clinical signatures across four cohorts. (B, C) Nomograms predicting 1-, 3-, and 5-year OS in TNBC patients (B) and up to 10-year OS in the full breast cancer cohort (C). (D) Calibration curve assessing the agreement between predicted and observed survival. (E, F) DCA demonstrating the net clinical benefit of the nomogram at 3.5 (E) and 5 (F) years. C-index: concordance index; SPT: Stemness-PTM; OS: overall survival; TNBC: triple negative breast cancer; DCA: Decision Curve Analysis\u003c/p\u003e","description":"","filename":"FIGURE6.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/b22089e2ebcd00c143a59f71.png"},{"id":108979603,"identity":"ab204557-c2ff-4b54-ba9a-0fa0fa53cbb1","added_by":"auto","created_at":"2026-05-11 12:00:14","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":52200952,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpatial transcriptomic and functional analysis of key genes. \u003c/strong\u003e(A–C) Ridge plots showing expression of \u003cem\u003ePCGF1\u003c/em\u003e, \u003cem\u003eKDM5B\u003c/em\u003e, and \u003cem\u003eTK1\u003c/em\u003e in C1 vs. C2 subtypes. (G–L) Spatial transcriptomic visualization of: (G) Tissue architecture (H\u0026amp;E); (H) Malignant vs. normal microregions; (I) Immune cell distribution; and (J–L) Spatial expression of \u003cem\u003ePCGF1\u003c/em\u003e, \u003cem\u003eKDM5B\u003c/em\u003e, and \u003cem\u003eTK1\u003c/em\u003e. (M–O) Quantification of gene expression in malignant vs. normal spots. ***P \u0026lt; 0.001. (P–R) Network plots illustrating correlations between gene expression and immune cell abundance across spatial spots. Positive (red) and negative (blue) correlations are shown.\u003c/p\u003e","description":"","filename":"FIGURE7.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/fce34d2b7cebf3460887165d.png"},{"id":108945802,"identity":"f374a4bb-c5c6-42e0-885b-957804af0c97","added_by":"auto","created_at":"2026-05-11 06:16:59","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":252530,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/9d1b818e-274f-430b-a592-16af161fdc05.pdf"},{"id":108978714,"identity":"021fff65-5e50-41f4-a528-66c024ed6f6e","added_by":"auto","created_at":"2026-05-11 11:47:55","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5317911,"visible":true,"origin":"","legend":"","description":"","filename":"supply1.png","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/050721e21a668d74dafa9de2.png"},{"id":108979606,"identity":"9f6406d9-91e9-46f4-b262-4747a3116e87","added_by":"auto","created_at":"2026-05-11 12:00:15","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":146931380,"visible":true,"origin":"","legend":"","description":"","filename":"Supply2.tif","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/55cf48c1aa2b4dc85b678a5b.tif"},{"id":108978728,"identity":"682e5d2d-c62b-4b26-8de2-08ebcf9d5cb8","added_by":"auto","created_at":"2026-05-11 11:48:09","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":40676920,"visible":true,"origin":"","legend":"","description":"","filename":"Supply3.tif","url":"https://assets-eu.researchsquare.com/files/rs-9353905/v1/ec2c3e2c53edc9ba446e30b0.tif"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integrating Post-Translational Modifications and Stemness Landscapes: A Multi-Omics Framework for Secondary Precision Intervention in Triple-Negative Breast Cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTNBC represents the most recalcitrant subtype of breast malignancy, defined by a lack of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression[1]. Despite the integration of immune checkpoint inhibitors with neoadjuvant chemotherapy—exemplified by the KEYNOTE-355 and KEYNOTE-522 regimens—as the new standard of care[2], the clinical management of TNBC remains a formidable challenge. A significant proportion of patients fail to achieve a pathological complete response (pCR), and the absence of targeted maintenance therapies for residual disease leaves these patients vulnerable to rapid recurrence and metastasis[3, 4]. Consequently, deciphering the molecular heterogeneity of TNBC to identify actionable vulnerabilities for maintenance strategies represents an urgent unmet clinical need.\u003c/p\u003e\n\u003cp\u003eThe therapeutic resistance and aggressive phenotype of TNBC are increasingly attributed to the synergistic interplay between proteomic plasticity and cellular stemness. PTMs—ranging from phosphorylation and acetylation to ubiquitination—dramatically expand the functional diversity of the proteome, enabling rapid adaptation to therapeutic stress and remodeling of the tumor microenvironment (TME)[5-11]. Concurrently, cancer stem cells (CSCs) serve as the apex of intratumoral heterogeneity. Through intricate crosstalk with the TME and immune evasion mechanisms—such as the downregulation of antigen presentation and the induction of an immunosuppressive stroma—CSCs orchestrate therapy resistance and tumor relapse[12-14].\u003c/p\u003e\n\u003cp\u003eWhile recent evidence suggests that aberrant PTMs can reprogram stemness traits to sustain malignancy[15, 16], the precise molecular architecture governing the crosstalk between PTM machinery and stemness maintenance remains poorly understood. Specifically, how the co-regulation of post-translational modification-related genes (PTMRGs) and stemness-regulating genes (SRGs) dictates the immune landscape and therapeutic sensitivity in TNBC constitutes a significant knowledge gap. Addressing this \"black box\" offers a compelling opportunity: by dissecting these regulatory networks, we may uncover novel biomarkers for stratification and targets for precision intervention[17-20].\u003c/p\u003e\n\u003cp\u003eIn this study, we employed an integrated multi-omics systems biology approach to delineate the crosstalk between PTM regulation and stemness in TNBC. We identified two distinct molecular subtypes—C1 and C2—characterized by divergent immunological profiles and therapeutic vulnerabilities. We demonstrate that the C1 subtype, despite its stemness features, exhibits an immunologically active microenvironment and specific dependency on the CDK4/6 signaling axis. Building on these insights, we developed and validated a SPT prognostic scoring model. Furthermore, we propose a \"secondary precision intervention\" framework, wherein SPT-guided stratification identifies candidates for targeted maintenance therapy—such as CDK4/6 inhibitors—thereby offering a novel paradigm to sustain disease control and improve survival outcomes in this challenging patient population.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003eData Acquisition and Preprocessing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWithin The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/), only TNBC specimens exhibiting over one month of follow-up were retained. Single-cell RNA-seq data (GSE161529[21]; 33,538 cells from 8 TNBC patients) were acquired from the\u0026nbsp;Gene Expression Omnibus\u0026nbsp;database (GEO database, https://www.ncbi.nlm.nih.gov/gds/). SRGs (n = 925) were sourced from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb). PTMRGs (n = 808) were compiled through extensive literature review based on prior publications. Transcriptomic profiles and clinical annotations for 320 TNBC cases, serving as the training cohort, were derived from METABRIC (https://ega-archive.org/studies/EGAS00000000083); Validation cohorts comprised publicly available datasets: TCGA-TNBC (n = 113), GSE58812[22]\u0026nbsp;(n = 107), and GSE37751[23]\u0026nbsp;(n = 14). Integrated identifiers were mapped to official gene symbols.\u0026nbsp;Only protein-coding genes were analyzed. Utilizing univariate cox regression, 862 differentially expressed genes (DEGs) linked to overall survival (OS) were identified within the METABRIC cohort. Associations among 48 selected SRGs and PTMRGs were examined using Spearman correlation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrognostic Gene Identification and Molecular Subtyping\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients were stratified into distinct clusters via unsupervised consensus clustering (\u0026quot;ConsensusClusterPlus\u0026quot; R package). To verify cluster robustness, the procedure underwent 1000 iterations. Peak intra-group consensus determined the optimal cluster count (k = 2), yielding subtypes C1 (n = 153) and C2 (n = 167). Mechanistic insights were pursued through Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, and Gene Set Enrichment Analysis (GSEA). The 48 SRGs and PTMRGs underwent GO and GSEA. Kaplan-Meier survival analysis revealed significant differences among these subtypes, while the heatmap integrated survival distributions with evaluations of clinical feature correlations.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential Expression Analysis and Mechanistic Investigation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eExpression profiles distinguishing subtypes C1 and C2 were generated using the\u0026nbsp;\u0026ldquo;limma\u0026rdquo;\u0026nbsp;R package. DEGs were defined by False discovery rate (FDR)-adjusted p-value \u0026lt; 0.05 and absolute log2 fold-change \u0026gt; 1. DEGs identified between C1 and C2 were subjected to GO, KEGG, and GSEA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTumor Immune Microenvironment Characterization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCIBERSORT quantified tumor-infiltrating immune cell populations. Immunophenotype score (IPS), incorporating MHC molecules, effector cells (EC), suppressor cells (SC), and immune checkpoints (CP), was computed. The anticancer immune status and immune cell proportions across the seven-step cancer-immunity cycle (CIC) were analyzed and visualized using the Tracking Tumor Immunophenotype platform (TIP; http://biocc.hrbmu.edu.cn/TIP/)[24] based on RNA-seq data. Immunotherapy response likelihood was estimated via the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm (http://tide.dfci.harvard.edu)[25]. The immune landscapes of subtypes C1 and C2 were evaluated using these computational approaches.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrognostic Model Construction and Validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLeast absolute shrinkage and selection operator (LASSO) regression was applied to the 320 TNBC cases derived from the METABRIC database, which identified the top 20 genes from the initial 48. Subsequently, 10 distinct machine learning algorithms (StepCox, Ridge, plsRcox, Lasso, CoxBoost, Enet, GBMs, SVMs, SuperPC, RSF) generated 117 algorithm combinations (\u0026quot;Mime1\u0026quot; R package). The \u0026quot;ML.Dev.Prog.Sig\u0026quot; function calculated the concordance index (C-index) for each combination; the model achieving the highest C-index was selected, incorporating 10 genes to formulate the SPT prognostic scoring system. Univariate Cox regression validated all 10 SPT genes. Based on SPT scores, patients were dichotomized into high-risk and low-risk categories. Prognostic performance was appraised using the risk score distribution, Kaplan-Meier survival analysis and time-dependent receiver operating characteristic (ROC) curves (AUCs for 1 year) across training dataset and 3 validation datasets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGenetic Alterations Analysis and Drug Response Prediction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTumor mutational burden (TMB) within subgroups was calculated with the \u0026quot;maftools\u0026quot; R package. Sensitivity profiles for 12 targeted therapeutics were forecasted using the Genomics of Drug Sensitivity in Cancer (GDSC) database via the R package \u0026quot;oncoPredict\u0026quot;. Correlations between the top 10 prognostic genes and predicted drug sensitivity were further assessed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReagents and Chemicals\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCarboplatin (Cat. No. HY-17393) and Palbociclib (Cat. No. HY-50767) were purchased from MedChemExpress (MCE, Monmouth Junction, NJ, USA). The purity of all test compounds was \u0026gt;99%. Stock solutions were prepared in dimethyl sulfoxide (DMSO; Macklin, Shanghai, China) and stored at 4\u0026deg;C protected from light. The Cell Counting Kit-8 (CCK-8) was obtained from TransGen Biotech (TransDetect\u0026reg; Cell Counting Kit, FC101-04, Beijing, China).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell Lines and Culture Conditions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe human breast cancer cell lines MDA-MB-231 and MDA-MB-468 were provided by GemPharmatech (Chengdu, China). Cells were cultured in their respective standard media supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. The cultures were maintained in a humidified incubator at 37\u0026deg;C with 5% CO\u003csub\u003e2\u003c/sub\u003e. All experiments were performed using cells in the logarithmic growth phase.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell Viability Assay\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe \u003cem\u003ein vitro\u003c/em\u003e antitumor activity of the test compounds was evaluated using the CCK-8 assay. Briefly, MDA-MB-231 and MDA-MB-468 cells were harvested and seeded into 96-well plates at a density of 5 times 10^3 cells/well in 150 muL of culture medium. The plates were incubated at 37\u0026deg;C for 24 h to allow for cell attachment.\u003c/p\u003e\n\u003cp\u003eFollowing the initial incubation, the culture medium was replaced with serum-free medium containing serial dilutions of the test compounds or the vehicle control (DMSO). The treatment concentrations were set as follows:\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eCarboplatin: 0.08\u0026ndash;8 uM for MDA-MB-231 cells; 0.2\u0026ndash;20 uM for MDA-MB-468 cells.\u003c/li\u003e\n \u003cli\u003ePalbociclib: 0.5\u0026ndash;50 uM for both cell lines.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eFive replicate wells were set for each concentration group (n=5). After 48 h of drug exposure, 10 uL of CCK-8 reagent was added to each well, and the plates were incubated for an additional period (typically 1\u0026ndash;4 h) at 37\u0026deg;C according to the manufacturer\u0026rsquo;s instructions. The absorbance was measured using a microplate reader (iMark\u0026trade; Microplate Absorbance Reader, BIO-RAD, Hercules, CA, USA).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell Transcriptome Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRaw single-cell RNA sequencing (scRNA-seq) data underwent quality control and normalization (\u0026quot;Seurat\u0026quot; R package)[26].\u0026nbsp;Cells meeting the following criteria were retained: \u0026gt;1000 UMIs, expression of 200-6000 unique genes, and \u0026lt;20% mitochondrial gene contribution. Cell clusters were visualized via t-distributed stochastic neighbor embedding (t-SNE). Cluster annotation leveraged canonical marker genes and cluster-specific DEGs. Malignant epithelial cells were discerned using inferCNV, employing B cells/endothelial cells as reference. SPT scores assigned samples to high-risk (n=4) or low-risk (n=4) groups; cell type proportions were contrasted between these groups. Cellular communication networks were inferred with CellChat.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePredictive Nomogram Construction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA nomogram integrating survival time, status, age, stage, SPT score, and grade was built (\u0026quot;rms\u0026quot; R package) to visualize TNBC patient prognosis. Nomogram performance was gauged using Harrell\u0026apos;s C-index, calibration curves, and decision curve analysis (DCA).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial transcriptome analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpatial transcriptomic profiling was performed using the 10x Genomics Visium platform, with scRNA-seq data serving as the reference. Quality control of the scRNA-seq data was conducted by filtering cells based on gene detection counts, unique molecular identifier (UMI) abundance, and mitochondrial gene expression proportion. Cellular deconvolution of each Visium spot was carried out using the SPOTlight package to estimate the relative abundances of annotated cell types based on the scRNA-seq reference.In parallel, cell type enrichment scores were computed with the Cottrazm package. Specifically, the top 25 most specifically expressed genes for each cell type were selected from the scRNA-seq reference to construct cell type\u0026ndash;specific gene signatures. Enrichment scores for each spatial spot were then calculated using the\u0026nbsp;get_enrichment_matrix\u0026nbsp;and\u0026nbsp;enrichment_analysis\u0026nbsp;functions.Based on malignant cell scores derived from deconvolution, spatial spots were categorized into three microregion types: malignant (score \u0026gt; 0) and normal (score = 0). Differential gene expression between malignant and normal microregions was assessed using the Wilcoxon rank-sum test. The dominant cell type for each spot, defined as the cell type with the highest inferred abundance, was visualized with Seurat\u0026rsquo;s\u0026nbsp;SpatialDimPlot. Spatial expression patterns of candidate genes (PCGF1, KDM5B, and TK1) were displayed using Seurat\u0026rsquo;s\u0026nbsp;SpatialFeaturePlot. Furthermore, Spearman correlation analysis was employed to evaluate relationships between cell type abundances and candidate gene expression, as well as intercellular interactions across all spatial spots. Correlation results were visualized using the \u0026quot;linkET \u0026quot; R package. The spatial transcriptomics data used in this study (sample ID: Sample_093C) were obtained from a publicly available database (GSE210616, GSM6433589)[27].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAnalyses employed GraphPad Prism (8.4.3), Python (3.10) and R software (4.2.2). Two-group comparisons used paired two-tailed Student\u0026apos;s t-tests or Mann-Whitney-Wilcoxon tests. Multi-group comparisons utilized ANOVA or Kruskal-Wallis rank-sum tests. Clinical characteristic associations were tested via chi-square. Statistical significance was defined as p \u0026lt; 0.05. In the drug sensitivity assay, data are presented as mean \u0026plusmn; standard deviation (Mean \u0026plusmn; SD). The half-maximal inhibitory concentration (IC₅₀) was calculated using nonlinear regression analysis. Depending on the distribution characteristics of the data, between-group differences were assessed for statistical significance using Student\u0026apos;s t‑test or analysis of variance (ANOVA). A P‑value \u0026lt; 0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eIdentification of SRGs and PTMRGs in TNBC\u0026nbsp;\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo systematically investigate the interplay between stemness and post-translational modifications, we developed a multi-omics analytical framework (Supplementary Figure 1). By intersecting 925 stemness-related genes (SRGs) and 862 post-translational modification-related genes (PTMRGs) with survival-associated genes from the METABRIC cohort (n = 320), we identified a core set of 48 prognostic genes, comprising 23 SRGs and 25 PTMRGs (Figure 1A). These genes exhibited extensive correlations and intricate intergenic interactions (Figure 1B). Using unsupervised consensus clustering based on this gene set, we stratified the TNBC cohort into two distinct molecular subtypes: Cluster 1 (C1, n = 153) and Cluster 2 (C2, n = 167), with k = 2 providing optimal clustering stability (Figure 1C; Supplementary Figures 2A\u0026ndash;C). Functional enrichment analysis suggested that these signature genes are involved in key biological processes, including ubiquitin-like protein transferase activity and cytoskeletal organization (Figures 1D\u0026ndash;E). Survival analysis revealed that the C1 subtype was associated with significantly superior overall survival (OS) compared to C2 (HR = 2.24, 95% CI: 1.63\u0026ndash;3.08, P \u0026lt; 0.0001; Figure 1F). Clinical profiling showed that the C1 subtype was enriched for early-stage (Stage I) patients and correlated with favorable survival outcomes (Table 1; Figure 1G). Moreover, Gene Set Enrichment Analysis (GSEA) uncovered marked differences in immune-related signaling pathways between the two subtypes, pointing to intrinsic variations in their tumor microenvironments (Figure 1H; Supplementary Figures 2D\u0026ndash;E). In summary, we successfully defined two TNBC molecular subtypes based on stemness and PTM-related genes that exhibit significant differences in clinical characteristics, prognostic outcomes, and immune pathway activity.\u003c/p\u003e\n\u003cp\u003eSchematic overview of the multi-omics framework integrating post-translational modifications and stemness landscapes in triple negative breast cancer. Sequencing data (DNA/RNA) are used to identify stemness-related genes and PTM-related genes, yielding 48 candidate genes. Consensus clustering of TNBC samples classifies two distinct molecular subtypes: C1 and C2. The C1 subtype is characterized by high PD-L1 expression, CD8+ T cell exhaustion, high tumor mutational burden, active PTM signaling, and high expression of the CDK4/6 signaling axis. The C2 subtype presents an immune-desert or stem-like phenotype, with tumor mutational burden correlating with M2 macrophage or MDSC infiltration, and exhibits higher expression of stemness markers (PCGF1, KDM5B, TK1) in cancer cells. The SPT score model is constructed using LASSO regression screening (20 genes) and machine learning (10-gene model) for risk scoring (high/low risk). Spatial transcriptomics and single-cell validation reveal high PCGF1 expression in malignant regions and PCGF1 expression in immune cell regions. Clinical decision for secondary precision intervention: low-risk C1 subtype receives CDK4/6 inhibitors (e.g., palbociclib), while high-risk C2 subtype receives novel combination therapies, leading to improved survival. TNBC: triple negative breast cancer; PTM: post-translational modification; TMB: tumor mutational burden; MDSC: myeloid-derived suppressor cells; SPT: Stemness-PTM score.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Baseline characteristics of TNBC patients in the METABRIC cohort, stratified by clusters C1 and C2.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ecCharacteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eC1 N = 109\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eC2 N = 114\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003ep-value\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026lt;=50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e14 (13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e23 (20%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026gt;50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e95 (87%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e91 (80%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eType\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.083\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Invasive Ductal Carcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e95 (87%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e103 (90%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Invasive Lobular Carcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e5 (4.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9 (7.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Invasive Mixed Mucinous Carcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1 (0.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Mixed Ductal and Lobular Carcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e8 (7.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2 (1.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMenopausal State\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Post\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e70 (64%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e78 (68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Pre\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e39 (36%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e36 (32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eStage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; I\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e38 (35%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e27 (24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; II\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e61 (56%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e72 (63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; III\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e10 (9.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e15 (13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eGrade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; I\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3 (2.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1 (0.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; II\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e14 (13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e12 (11%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; III\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e92 (84%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e101 (89%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eOS Status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Alive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e63 (58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e36 (32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Dead\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e46 (42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e78 (68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eRFS Status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Not recurred\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e79 (72%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e51 (45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; Recurred\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e30 (28%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e63 (55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eChemotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e55 (50%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e59 (52%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHormone Therapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e34 (31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e31 (27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eRadio Therapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e82 (75%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e84 (74%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003csup\u003e1\u003c/sup\u003e\u003c/strong\u003e\u003cstrong\u003en (%\u003c/strong\u003e)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003csup\u003e2\u003c/sup\u003e\u003c/strong\u003e\u003cstrong\u003ePearson\u0026apos;s Chi-squared test; Fisher\u0026apos;s exact test \u0026nbsp;\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmune landscape of distinct molecular subtypes in TNBC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGiven the significant survival disparities observed between the C1 and C2 subtypes, we subsequently conducted an in-depth analysis of the tumor immune microenvironment (TIME) to uncover the immunological basis driving these divergent clinical outcomes. CIBERSORT analysis revealed that the C1 subtype was distinguished by robust infiltration of adaptive immune cells, particularly CD8⁺ T cells, whereas the C2 subtype exhibited an immune-desert phenotype (Figure 2A). To assess immunogenicity, we applied the Immunophenoscore (IPS) algorithm, which showed that C1 subtype had elevated scores for MHC molecules and effector cells (EC), consistent with a \u0026quot;hot\u0026quot; tumor phenotype (Figure 2B\u0026ndash;C).\u003c/p\u003e\n\u003cp\u003eInterestingly, although the aggregate immune checkpoint (CP) score derived from IPS was lower in C1, gene-level analysis revealed a marked upregulation of Programmed death-ligand 1 (PD-L1) in this subtype (Figure 2D). This apparent divergence suggests that immune evasion in C1 is not driven by broad inhibitory checkpoint upregulation\u0026mdash;which would elevate the CP score\u0026mdash;but rather by a specific adaptive resistance mechanism mediated predominantly through the PD-L1 axis. Further dissection of the cancer-immunity cycle (CIC) demonstrated that C1 exhibited enhanced activity across key steps, including antigen release and T-cell recognition (Figure 2E\u0026ndash;F). However, cytotoxic killing efficiency remained disproportionately low relative to the degree of immune infiltration, supporting the hypothesis of PD-L1-mediated functional T-cell inhibition. TIDE analysis corroborated that although C1 tumors exhibited lower overall immune evasion potential (reflected by lower TIDE scores), they were characterized by T-cell dysfunction (Figure 2G). Moreover, while the C1 subtype showed higher immunotherapy response rates (Figure 2H) and CTL levels (Figure 2I), the C2 subtype demonstrated greater immune benefit ratios (Figure 2J), with its microenvironment dominated by immune exclusion and immunosuppressive populations such as M2 macrophages and MDSCs (Figure 2K-L). These findings collectively delineate two starkly different immune landscapes: the C1 subtype presents a PD-L1-driven immune-active but exhausted phenotype, while the C2 subtype is characterized by immune exclusion and suppression.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstruction and validation of the SPT signature score\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing the identification of biological differences between the molecular subtypes, we sought to develop a clinically actionable quantitative tool to precisely assess mortality risk in TNBC patients by integrating key SRGs and PTMRGs.\u0026nbsp;Firstly, we reduced the gene set to 20 candidates using LASSO regression (Figures 3A-B), then evaluated 117 algorithm combinations using integrated machine learning across a training set and three independent validation cohorts (TCGA, GSE58812, GSE37751), ultimately finalizing the SPT prognostic scoring system composed of 10 genes (Figure 3C). Patients were stratified into high- and low-risk groups based on their SPT scores. The predictive risk distribution was visualized in Figure 3D-G, where the bottom heatmap revealed the distinct expression patterns of the 10 core genes that distinguished high-risk from low-risk individuals. For instance, genes such as DMX2 showed decreased expression in the high-risk group, whereas DPM2 exhibited increased expression in the high-risk group.\u003c/p\u003e\n\u003cp\u003eThe model demonstrated robust prognostic discrimination, with high-risk patients exhibiting significantly worse OS across all four cohorts (METABRIC: HR = 1.96, P \u0026lt; 0.0001; TCGA: HR = 3.04, P = 0.028; GSE58812: HR = 2.59, P = 0.011; GSE37751: HR = 5.25, P = 0.035; Figures 4H\u0026ndash;K). Time-dependent ROC analysis confirmed the model\u0026apos;s predictive stability, yielding 1-year AUC ranging from 0.75 to 0.92 across datasets (Figures 4L), validating the SPT score as a reliable prognostic instrument. This SPT score integrated key SRGs and PTMRGs (Figure 4M). Extensive validation across multiple independent cohorts consistently demonstrates that the SPT risk score system is highly robust and serves as an effective tool for predicting long-term survival outcomes in TNBC patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential responses of C1 and C2 to targeted and chemotherapy drugs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo link molecular subtypes to therapeutic vulnerabilities, we analyzed somatic mutational landscapes. The C1 subtype exhibited a significantly higher TMB compared to C2. While \u003cem\u003eTP53\u003c/em\u003e mutations were ubiquitous in both subtypes (~80%), other drivers such as \u003cem\u003eMLL2\u003c/em\u003e showed lower mutation frequencies (Figures 4A\u0026ndash;B). Pharmacogenomic profiling revealed that the transcriptomic differences between C1 and C2 correlated with distinct drug sensitivity patterns (Figure 4C).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;We mapped subtype-specific gene signatures to drug response data (Figure 4D) and identified the CDK4/6 inhibitor palbociclib as a candidate agent. To validate this \u003cem\u003ein vitro\u003c/em\u003e, MDA-MB-468 and MDA-MB-231 cell lines were selected as representative models for the C1 and C2 subtypes, respectively. Drug sensitivity assays confirmed that the C1-like MDA-MB-468 cells were significantly more sensitive to palbociclib compared to the C2-like MDA-MB-231 cells (Figures 4E\u0026ndash;F), establishing a preclinical rationale for targeting the CDK4/6 axis in the C1 population.The results of the in vitro experiments confirm that the C1 subtype is highly sensitive to CDK4/6 inhibitors, providing prospective experimental evidence for implementing precision maintenance therapy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eScRNA-seq analysis of immune cell infiltration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo further validate these findings, we analyzed the single-cell RNA sequencing dataset GSE161529, which comprised 33,538 cells from eight triple-negative breast cancer (TNBC) patients after quality control and filtering. Unsupervised clustering categorized these cells into six major populations: B cells, T/NK cells, myeloid cells, macrophages, cancer-associated fibroblasts (CAFs), and epithelial cells (Figures 5A\u0026ndash;C). To distinguish benign from malignant epithelial cells, we extracted epithelial cells along with B cells and endothelial cells as reference groups, and applied infercnv to assess the degree of malignant transformation. This analysis further subdivided epithelial cells into normal mammary epithelial cells and tumor cells (Figures 5D\u0026ndash;F).\u003c/p\u003e\n\u003cp\u003eUsing pseudo-bulk analysis, we computed the SPT score for each sample and stratified the eight TNBC cases into high-risk (n = 4) and low-risk (n = 4) groups. The high-risk group exhibited a significantly higher proportion of tumor cells compared to the low-risk group (Figure 5G).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCell-cell communication analysis indicated that both the number and strength of inferred interactions were significantly greater in the low-risk group, suggesting more active intercellular crosstalk (Figure 5H-I). Furthermore, the interaction network analysis revealed distinct communication patterns between the two groups: the high-risk group exhibited more complex and intensive interactions involving normal epithelial cells, CAFs, and endothelial cells, which may underlie its adverse prognosis. In contrast, the low-risk group showed enhanced signaling among tumor cells, CAFs, and endothelial cells (Figure 5J-K).\u003c/p\u003e\n\u003cp\u003eFinally, we analyzed the information flow of ligand\u0026ndash;receptor pairs and observed differential activity between risk groups. Genes such as IFN-\u0026beta;, TIGIT, and NECTIN displayed higher information flow in the high-risk group, while ICAM, VEGF, and SELE were more active in the low-risk group (Figure 5L), suggesting their potential roles in shaping distinct tumor microenvironments associated with prognosis. These results indicate that high-risk tumor lacks global immune signal exchange and instead rely on inhibitory pathways like TIGIT to maintain their immune-excluded and treatment-resistant phenotype.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstruction of a prognostic nomogram\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess the predictive performance of our model, we conducted a comparative evaluation against established TNBC models. Based on the concordance index (C-index) across multiple datasets (METABRIC, TCGA, GSE58812, and GSE37751), our model achieved the highest predictive accuracy in the METABRIC cohort and was ranked among the top models in the independent validation sets, confirming its superior and robust prognostic value. (Figure 6A).\u003c/p\u003e\n\u003cp\u003eTo facilitate clinical translation, we developed two nomograms for prognostic prediction. The first nomogram was constructed for patients in the METABRIC TNBC cohort based on clinical variables including age, stage, risk score, and grade, along with 1-, 3-, and 5-year survival probabilities (Figure 6B). The other nomogram was established for the entire METABRIC cohort, incorporating the same clinical variables and extending the prediction to 1-, 3-, 5-, and 10-year survival probabilities (Figure 6C). The calibration curve demonstrated excellent agreement between the nomogram-predicted and observed outcomes, indicating high predictive accuracy (Figure 6D). Furthermore, decision curve analysis (DCA) showed that the nomogram provided a greater net clinical benefit than the use of clinical features alone at both 3.5 and 5 years (Figures 6E\u0026ndash;F). Strongly support the potential of our model as a reliable predictive tool with clinical applicability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-gene analysis and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003espatial transcriptomic analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate the potential mechanisms through which the SPT score predicts TNBC prognosis, we first performed single-gene expression and survival analysis across the C1 and C2 subtypes (Supplement Figure 3). Based on these findings, we selected three key genes\u0026mdash;PCGF1, TK1, and KDM5B\u0026mdash;for further protein-protein interaction analysis. In TNBC, expression levels of PCGF1, TK1, and KDM5B were consistently lower in the C1 subtype but elevated in C2 (Figure 7A\u0026ndash;C). Kaplan\u0026ndash;Meier survival analysis demonstrated that high expression of each gene was significantly associated with poorer overall survival: PCGF1 (HR = 1.59, 95% CI: 1.17\u0026ndash;2.17, p = 0.0027; Figure 7D), KDM5B (HR = 1.52, 95% CI: 1.12\u0026ndash;2.07, p = 0.0068; Figure 7E), and TK1 (HR = 1.48, 95% CI: 1.09\u0026ndash;2.01, p = 0.012; Figure 7F). Functional annotation revealed distinct biological roles for these genes: PCGF1-associated proteins were primarily involved in immune-related regulation, extracellular matrix organization, cellular structure, and receptor\u0026ndash;ligand binding (Figure 7G); KDM5B-related proteins were mainly enriched in cell proliferation (Figure 7H); and TK1-interacting partners were implicated in cell proliferation, DNA synthesis, and repair pathways (Figure 7I).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFurthermore, immunohistochemistry (IHC) data from TCGA confirmed that both PCGF1 and KDM5B protein levels were significantly upregulated in breast tumor tissues compared with adjacent normal controls (Figures 7G\u0026ndash;O). The construction of the nomogram confirms the predictive value of the model in clinical practice, while spatial localization analysis identifies \u003cem\u003ePCGF1\u003c/em\u003e, \u003cem\u003eTK1\u003c/em\u003e, and \u003cem\u003eKDM5B\u003c/em\u003e as key spatial regulators driving the immune-excluded phenotype (Figures 7P\u0026ndash;R). It not only explains the biological basis of the SPT score from a mechanistic perspective, but also provides a theoretical basis for its use as a prognostic predictor and potential therapeutic target.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eDeciphering the molecular heterogeneity of TNBC is pivotal for overcoming the therapeutic impasse characterized by drug resistance and high recurrence rates. In this study, we integrated the regulatory landscapes of PTMs and cancer stemness\u0026mdash;two fundamental drivers of tumor plasticity\u0026mdash;to delineate a novel classification framework. Our multi-omics stratification revealed two biologically distinct subtypes: an immunologically active but dysfunctional \u0026quot;C1\u0026quot; subtype and a stem-like, immune-desert \u0026quot;C2\u0026quot; subtype. This dichotomy not only clarifies the intrinsic variability in TNBC prognosis but also establishes a mechanistic rationale for a \u0026quot;secondary precision intervention\u0026quot; strategy, specifically positioning CDK4/6 inhibitors as a potential maintenance therapy for the C1 population.\u003c/p\u003e\n\u003cp\u003eA central finding of our work is the paradoxical immune landscape of the C1 subtype. While characterized by robust infiltration of cytotoxic T lymphocytes (CTLs) and a high TMB\u0026mdash;features typically associated with favorable immunotherapeutic responses\u0026mdash;the C1 microenvironment exhibits profound functional exhaustion. Our analysis of the \u0026quot;Cancer-Immunity Cycle\u0026quot; indicates that while T cell recruitment is efficient, the terminal effector phase is compromised. This aligns with the \u0026quot;hot but exhausted\u0026quot; phenotype, where high expression of immune checkpoints (e.g., PD-L1) and chronic antigen stimulation drive T cells into a dysfunctional state [28, 29]. Mechanistically, we postulate that aberrant PTMs, such as glycosylation of PD-L1, may stabilize checkpoint proteins to subvert CTL cytotoxicity[30, 31], thereby creating a microenvironment that is infiltrated yet immunosuppressed. Conversely, the C2 subtype recapitulates an \u0026quot;immune-excluded\u0026quot; or \u0026quot;desert\u0026quot; phenotype, driven by stemness-associated crosstalk that actively repels immune infiltration, consistent with the aggressive, therapy-resistant nature of CSCs[32]. This stemness-associated crosstalk is mediated by dynamic interactions with the TME, characterized by hypoxia, inflammatory signals, and extracellular matrix components, ultimately contributing to therapy resistance and tumor relapse [33, 34]. Furthermore, conventional therapies such as chemotherapy and radiation, which predominantly target rapidly proliferating tumor cells, are ineffective at eliminating CSCs[35].\u003c/p\u003e\n\u003cp\u003eThe delineation of distinct therapeutic vulnerabilities between these subtypes represents a significant translational advance. We demonstrate that the C1 subtype, represented by the MDA-MB-468 model, exhibits specific sensitivity to CDK4/6 inhibitors (e.g., palbociclib). This finding challenges the conventional restriction of CDK4/6 inhibitors to luminal breast cancer and supports emerging evidence of their efficacy in specific TNBC niches [36]. Mechanistically, the sensitivity of C1 tumors likely stems from a dual dependency: a reliance on rapid cell cycle progression driven by high mutational load, rendering them susceptible to cell cycle blockade, and an underlying immune plasticity. Notably, CDK4/6 inhibitors possess distinct immunomodulatory properties capable of stimulating antigen presentation and reducing regulatory T cell proliferation [37], which may reinvigorate the \u0026quot;hot but exhausted\u0026quot; T cell phenotype we observed in C1 tumors. Post-neoadjuvant therapy, this molecular stratification framework provides a compelling rationale to identify candidates\u0026mdash;specifically those harboring C1-like features\u0026mdash;who may benefit from CDK4/6 inhibitor-based maintenance regimens, thereby sustaining disease control where standard chemotherapy reaches its limit[38].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAt the molecular level, our SPT prognostic signature elucidates the genetic architecture underpinning these phenotypes. The signature integrates key regulators such as PCGF1\u003ca id=\"_anchor_3\" onmouseover=\"msoCommentShow('_anchor_3','_com_3')\" onmouseout=\"msoCommentHide('_com_3')\" href=\"#_msocom_3\" language=\"JavaScript\" name=\"_msoanchor_3\"\u003e\u003c/a\u003e [39], TK1\u003ca id=\"_anchor_4\" onmouseover=\"msoCommentShow('_anchor_4','_com_4')\" onmouseout=\"msoCommentHide('_com_4')\" href=\"#_msocom_4\" language=\"JavaScript\" name=\"_msoanchor_4\"\u003e\u003c/a\u003e [40], and KDM5B\u003ca id=\"_anchor_5\" onmouseover=\"msoCommentShow('_anchor_5','_com_5')\" onmouseout=\"msoCommentHide('_com_5')\" href=\"#_msocom_5\" language=\"JavaScript\" name=\"_msoanchor_5\"\u003e\u003c/a\u003e [41], connecting epigenetic remodeling with cell cycle kinetics. For instance, PCGF1 (a component of the PRC1 complex) and KDM5B (a histone demethylase) are implicated in maintaining the plastic, undifferentiated state of CSCs while simultaneously orchestrating chromatin accessibility for immune-evasive gene expression [42, 43]. The spatial transcriptomic co-localization of these genes with malignant clusters further underscores their role in shaping the tumor-intrinsic signaling that dictates the surrounding immune conte\u003ca id=\"_anchor_6\" onmouseover=\"msoCommentShow('_anchor_6','_com_6')\" onmouseout=\"msoCommentHide('_com_6')\" href=\"#_msocom_6\" language=\"JavaScript\" name=\"_msoanchor_6\"\u003e\u003c/a\u003e xt[44, 45]. For instance, TK1\u0026mdash;a key node within our signature\u0026mdash;has been identified in previous studies as a potential therapeutic target in Treg-mediated immunosuppression. Moreover, the immunomodulatory factors MUC1-C and CXCL9, which are known to influence T cell exhaustion and macrophage differentiation within the TME, may represent downstream effectors or interacting partners in this regulatory network.\u003c/p\u003e\n\u003cp\u003eDespite these promising insights, our study has limitations inherent to in silico analyses. While we cross-validated our findings across multiple independent cohorts (METABRIC, TCGA, GEO) to ensure robustness, the retrospective nature of these datasets introduces potential selection bias. Additionally, while our pharmacogenomic predictions were validated in cell line models, the complex interplay between PTMs and the immune microenvironment requires further elucidation in immunocompetent in vivo models. Future studies should focus on mapping the precise PTM sites governing the identified SRGs and validating the synergistic efficacy of CDK4/6 inhibitors with immunotherapy in preclinical trials.\u003c/p\u003e\n\u003cp\u003eIn conclusion, our systems biology approach unveils a novel TNBC taxonomy governed by the crosstalk between PTMs and stemness. By defining a subtype characterized by immune exhaustion and CDK4/6 dependency, we provide a blueprint for precision maintenance therapy. The SPT scoring system serves as a clinically translatable tool to stratify patients, offering a new avenue to dismantle the lethal heterogeneity of TNBC and improve long-term survival.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, we classified patients with TNBC into distinct molecular subtypes based on SRGs and PTMRGs, and elucidated the associated immunological and genetic profiles of each subtype. Furthermore, we developed a SPT scoring model capable of predicting both patient prognosis and response to immunotherapy. This SPT score has been validated across multiple independent datasets and shows promising potential for guiding individualized and precise clinical management of TNBC.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eTNBC: Triple-negative breast cancer; ER: Estrogen receptor; PR: Progesterone receptor; HER2: Human epidermal growth factor receptor 2; PD-L1: Programmed death-ligand 1; PTMs: Post-translational modifications; TME: The tumor microenvironment; CSCs: Cancer stem cells; TCGA: The Cancer Genome Atlas Program; GEO: Gene Expression Omnibus; SRGs: Stemness-associated genes; PTMRGs: Post-translational modification-related genes; DEGs: Differentially expressed genes; OS: Overall survival; FDR: False discovery rate; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; GSEA: Gene Set Enrichment Analysis; IPS: Immune Proportion Score; EC: Effector cell; SC: Suppressor cell; CP: Checkpoint; CIC: Cancer-immunity cycle; TIP: Tumor Immunophenotype; TIDE: Tumor Immune Dysfunction and Exclusion; TMB: Tumor mutational burden; GDSC: Genomics of Drug Sensitivity in Cancer; LASSO: The least absolute shrinkage and selection operator; ROC: Receiver operating characteristic; scRNA-seq: single-cell RNA-seq; t-SNE: t-distributed stochastic neighbor embedding; GSEA: Gene set enrichment analysis; CTL: Cytotoxic T lymphocyte; TAM: Tumor-associated macrophages; MDSCs: Myeloid-derived suppressor cells; HRs: Hazard ratios; AUC: Area under curve; CAFs: Cancer-associated fibroblasts; DCA: Decision curve analysis; IHC: Immunohistochemistry; ICI: Immune checkpoint inhibitors; TIME: Tumor Immune Microenvironment; SPT: Stemness-PTM\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthorship contribution statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHuimin Zhang\u003c/strong\u003e: Conceptualization, Writing original draft, Data curation, Funding acquisition.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eShenao Qu\u003c/strong\u003e: Writing original draft, Methodology, Data curation, Validation. \u003cstrong\u003eJinshuang Zhu\u003c/strong\u003e: Methodology, Data curation. \u003cstrong\u003eHaozhe Huang\u003c/strong\u003e: Methodology, Data analysis. \u003cstrong\u003eHaoqi Yan and Zhanghang Li\u003c/strong\u003e: Data curation. \u003cstrong\u003eJu Zhang and Wenya Xue\u003c/strong\u003e: Validation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was funded by the Key Research and Development Project of Shaanxi Province (2025GH-YBXM-069).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to participate\u003c/strong\u003e Not applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePublicly available datasets were analyzed in this study. These data can be found in the following databases: TCGA (https://portal.gdc.cancer.gov/), GEO (https://www.ncbi.nlm.nih.gov/), and METABRIC (https://ega-archive.org/). Mechanistic cartoons were created with BioRender (https://biorender.com/).\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"breast-cancer-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"brcr","sideBox":"Learn more about [Breast Cancer Research](http://breast-cancer-research.biomedcentral.com)","snPcode":"13058","submissionUrl":"https://submission.nature.com/new-submission/13058/3","title":"Breast Cancer Research","twitterHandle":"@BCRJournal","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Triple-negative breast cancer, Post-translational modifications, Molecular subtype, CDK4/6 inhibitors, Maintenance therapy","lastPublishedDoi":"10.21203/rs.3.rs-9353905/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9353905/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e\u003cbr\u003e\nTo address the critical unmet need for effective maintenance therapies in triple-negative breast cancer (TNBC) patients with residual disease following neoadjuvant therapy, we investigated the interplay between post-translational modifications (PTMs) and cancer stemness—a regulatory axis that remains poorly characterized in TNBC. This study aims to uncover novel therapeutic vulnerabilities and establish a precision stratification tool to improve outcomes in TNBC patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003cbr\u003e\nWe performed an integrated multi-omics analysis to characterize the molecular crosstalk between PTM signaling and stemness features within TNBC. This led to the identification of distinct molecular subtypes based on their regulatory and immunological landscapes. A prognostic scoring system, termed Stemness-PTM (SPT), was subsequently developed and validated in independent cohorts to quantify these biological features and predict therapeutic response.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003cbr\u003e\nOur profiling delineated two distinct TNBC subtypes: an immunologically active 'C1' subtype with a specific dependence on the CDK4/6 signaling axis, and a biologically aggressive 'C2' subtype. Despite its complex regulatory environment, the C1 subtype demonstrated marked sensitivity to CDK4/6 inhibitors, such as palbociclib. The SPT prognostic score effectively stratified patient risk and outcomes across independent validation cohorts, converting this molecular signature to clinical prognosis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003cbr\u003e\nThis study establishes a mechanistic rationale for a \"secondary precision intervention\" in high-risk TNBC. By positioning CDK4/6 inhibitors as a targeted maintenance therapy for patients identified by the SPT scoring system, our findings offer a translatable strategy to achieve sustained disease control and improve survival outcomes.\u003c/p\u003e","manuscriptTitle":"Integrating Post-Translational Modifications and Stemness Landscapes: A Multi-Omics Framework for Secondary Precision Intervention in Triple-Negative Breast Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-11 06:16:52","doi":"10.21203/rs.3.rs-9353905/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"278837449381608976570388298515940361699","date":"2026-05-15T13:37:46+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-28T18:04:51+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-14T14:14:26+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-14T09:34:55+00:00","index":"","fulltext":""},{"type":"submitted","content":"Breast Cancer Research","date":"2026-04-08T08:13:55+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"breast-cancer-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"brcr","sideBox":"Learn more about [Breast Cancer Research](http://breast-cancer-research.biomedcentral.com)","snPcode":"13058","submissionUrl":"https://submission.nature.com/new-submission/13058/3","title":"Breast Cancer Research","twitterHandle":"@BCRJournal","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d97b9f34-2a5a-4926-8556-4ab06b62465f","owner":[],"postedDate":"May 11th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"278837449381608976570388298515940361699","date":"2026-05-15T13:37:46+00:00","index":17,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-11T06:16:53+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-11 06:16:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9353905","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9353905","identity":"rs-9353905","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}