Introduction

Female breast cancer (BC) has become the most common cancer worldwide, with an estimated 2.3 million new cases diagnosed in 2020, of which more than two-thirds are classified as hormone receptor-positive (HR+) and human epidermal growth factor receptor 2-negative (HER2-) subtype1,2. According to SEER Cancer Statistic, the average five-year survival rate for all HR+/HER2- BC patients is 94.8% but only 34.0% for advanced and metastatic HR+/HER2- BC3. HR+/HER2- BC has a longer duration of recurrence risk compared to other subtypes, ranging from 13 to 41% over a 20-years period after stopping endocrine therapy at 5 years4.

Endocrine therapy is a well-established treatment for HR+/HER2- BC patients. However, approximately 25-50% of patients respond poorly to endocrine therapy due to de novo resistance, potentially stemming from the limited contribution of estrogen receptor-α positive (ERα+) clones in the primary tumor5,6. Moreover, acquired resistance is inevitable over time, with the clinical benefit rate decreasing to around 30% for second or greater lines of endocrine therapy7. Cyclin-dependent kinases 4 and 6 (CDK4/6) inhibitors have shown efficacy in halting the progression of cancer cells in G1 phase of the cell cycle, which are crucial for preventing cancer cells from undergoing further cell division and proliferation by inhibiting the activity of CDK4/6-cyclin D complexes8. Since CDK4/6 inhibitors, such as Palbociclib, Ribociclib, and Abemaciclib, have been recommended as an initial treatment option for HR+/HER2- advanced BC in combination with an endocrine therapy aromatase inhibitor (AI) or Fulvestrant9,10,11,12,13the addition of CDK4/6 inhibitors to standard endocrine therapy have changed the treatment approach of HR+/HER2- BC patients. Unfortunately, despite the efficacy of CDK4/6 inhibitors in combination therapy, there are instances where intrinsic and acquired resistance mechanisms can limit the efficacy, ultimately lead to disease progression in approximately 12% of cases14,15.

Numerous studies have been dedicated to unraveling the resistance mechanisms linked to CDK4/6 inhibitors. One of the identified resistance mechanisms appears to involve the activation of the cyclin D-CDK4/6-Rb-E2F cell cycle-related genes axis16. The intricacies of cell division hinge on a meticulously regulated process governed by a complex network of regulatory mechanisms and numerous cell cycle factors, ensuring the timely and precise duplication and segregation of genomic DNA17,18,19. Cell cycle regulatory pathways, driven by indispensable mutations, such as TP53, CCNE1, CDKN2A and RB1, play a crucial role in the development process of various types of cancers20,21,22. Despite these insights, the mechanisms leading to CDK inhibitor resistance and their subsequent impact on prognosis remain unclear. Additionally, the identification of HR+/HER- BC patients who stand to benefit from CDK4/6 inhibitors proves challenging due to the inherent heterogeneity of the disease.

Although cell cycle factors have been utilized to develop prognosis models for various cancer types, including colon cancer, bladder cancer and lung adenocarcinoma23,24their predictive value for prognosis in HR+/HER2- BC has remained uncertain. To address this gap, our study strategically leveraged cell cycle factors to develop and validate a HR+/HER2- BC Prognostic Signature (HBPS), utilizing comprehensive data from TCGA and GEO datasets. Furthermore, our analysis delves into the intricate associations between the HBPS score and biological signatures. We further conducted in vitro experiments to validate the relationships between key gene expression, cellular aggressiveness, and immune microenvironment. These findings establish a coherent framework for understanding the prognostic landscape in this specific breast cancer subtype.

Materials and methods

Data download

A cohort of 421 h+/HER2- BC patients sourced from the TCGA dataset served as the training set for this study. Validation involved two additional datasets: 2236 h+/HER2- BC patients from the GSE96058 dataset and 1369 h+/HER2- BC patients from the METABRIC dataset. Comprehensive RNA sequencing (RNA-Seq), whole-exon sequencing (WES), and clinical data for HR+/HER2- BC patients were obtained from TCGA database and UCSC Xena database. The detailed clinical data are listed in Supplementary Table S1. RNA sequencing and survival information for GSE96058 and METABRIC datasets were retrieved from https://www.ncbi.nlm.nih.gov/geo and https://www.cbioportal.org, respectively25,26. Gene expression and drug response sensitivity (IC50) for 196 compounds across 51 BC cell lines were sourced from the Genomics of Drug Sensitivity in Cancer database (GDSC, www.cancerrxgene.org). A collection of 1,874 cell cycle-related genes extracted from a previously published literature23is detailed in Supplementary Table S2, along with the functional annotations for these genes. Furthermore, the 50 Hallmark pathway gene sets were downloaded from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb/).

Signature construction and prognostic analysis

Differential cell cycle-related genes between tumor and normal tissues were identified based on log2-transformed TCGA RNA expression (FPKM) data using the R package Limma. Univariate Cox regression analysis was conducted for the differentially expressed cell cycle factors, with a significance threshold set at P < 0.01 to discern statistically significant prognostic factors. Subsequently, multivariate Cox analysis was conducted to calculate coefficients for the identified prognostic factors. Patients’ scores were calculated based on their gene expression levels and the corresponding regression coefficients. Utilizing the median HR+/HER2- BC Prognostic Signature (HBPS) score as a cutoff, patients were stratified into high and low score groups. The R package SurvMINER (https://CRAN.R-project.org/package=survminer) was employed to analyze the differential prognosis between the two groups. Additionally, the R packages timeROC (https://CRAN.R-project.org/package=timeROC) and Survival (https://CRAN.R-project.org/package=survival) were utilized to calculate the area under the curve (AUC) of the time-dependent receiver operating characteristics (ROC) curve and the concordance index (C-index), respectively. A prognostic nomogram was developed based on the results of the multivariate analysis. To assess its predictive accuracy, calibration curves and the C-index were employed.

Biological mechanisms analysis

To investigate the biological mechanism underlying the prognostic differences between the high and low score groups, further analysis was performed on the TCGA dataset. We employed single sample gene set enrichment analysis (ssGSEA) using the R package GSVA (https://bioconductor.org/packages/release/bioc/html/GSVA.html)27. The normalized enrichment scores (NES) for the 50 Hallmark pathways were calculated. Spearman correlation was performed to assess the relationship between the HBPS score and the NES for each pathway. Subsequently, Hallmark pathways exhibiting a statistically significant correlation with the score (P ≤ 0.05) were identified. Furthermore, to assess the degree of pathway enrichment between the two groups, we conducted Gene Ontology Biological Process (GOBP) and Kyoto Encyclopedia of Genes and Genome (KEGG) pathway enrichment analyses using the R package clusterProfiler. To gain insights into immune-related aspects, immune cell infiltration was calculated using the CIBERSORT28 and ESTIMATE29 algorithms implemented in the R package IOBR. Comparative analyses were then conducted to elucidate differences in the immune microenvironment between the two groups.

Clinical samples and multiplexed immunohistochemistry staining

Breast cancer tissue microarray (TMA, catalog HBre-Duc159Sur-01) was obtained from Outdo BioTech (Shanghai, China), and ethical approval was granted by the Clinical Research Ethics Committee in Outdo Biotech. The 63 h + HER2- tumor samples represent the total number of instances after excluding the samples that were lost throughout several staining procedures. All clinic-pathological were provided by Outdo Biotech.

Paraffin-embedded TMA was submitted for multiplexed immunohistochemistry (mIHC) staining. Standard operating procedures were employed for mIHC staining30. The primary antibodies used for mIHC staining were as follows: anti-CDKN2C (1:2000, catalog ab192239, Abcam) and anti-CD8 (1:1000 dilution, catalog 85336, CST). The HALO software was used to evaluate the number of CDKN2C and CD8 positive cells.

Cell lines and western blotting analysis

Human HR + HER2- breast cell lines MCF-7 (catalog KGG3332-1) and T47D (catalog KGG3219-1) were purchased from KeyGEN (Nanjing, China). MDA-MB-231 cells were cultured in DMEM media and T47D cells were cultured in RPMI-1640 media. All media were added with 10% FBS at 37 °C with 5% CO2. All human cell lines were authenticated using short tandem repeat profiling and all assays were conducted with mycoplasma-free. The CDKN2C overexpression plasmid was synthesized by KeyGEN (Nanjing, China).

Breast cancer cells were plated in maintained in 6-well plates. Total proteins of cells were harvested using lysis buffer. Then, SDS-PAGE and Western blotting analysis were conducted referring to standardized protocols. The primary antibodies used as follows: CDKN2C (1:1000 dilution, catalog ab192239, Abcam) and GAPDH (1:2000 dilution, Cat. 60004-1-Ig, ProteinTech). Proteins levels were standardized to GAPDH.

In vitro assays for cellular functions

In order to assess cell proliferation levels, cancer cells were seeded in six-well plates at a density of 500 cells per well and cultured at 37 °C for 1–2 weeks. At the end of the incubation, colonies were stained with crystal violet solution for 30 min. Wells were rinsed with water followed by air drying and the colonies were counted. Each measurement was performed in triplicate. Also, the expression of PCNA (PCNA antibody, 1:1000, catalog 10205-2-AP, ProteinTech), a proliferation marker, was also checked by western blotting analysis. To assess cell cycle, a cell cycle assay kit was used (catalog KGA9101, KeyGEN, Nanjing, China). In order to assess cell migration and invasion levels, transwell chambers (catalog 3422, Corning) were used; with or without Matrigel (Corning) coating depending on the desired assay. Cancer cells (5 × 104) in 200µL of serum-free medium were seeded in the upper chamber while 600µL of medium containing 10% FBS was added to the lower chamber. In addition, cancer cells were cultivated to confluence in 96-well plates before being seeded for study of wound healing. After the culture insert was removed, the cell monolayers were damaged, and any remaining cell debris was washed away with PBS. The migratory area was determined by subtracting the wound closure edge from 0 to 24 h.

Statistical analysis

All statistical analyses were performed using R Version 4.1.2. The log-rank test was utilized to compare differences in overall survival (OS) between two groups. Spearman’s correlation coefficient was used to measure the correlation between two groups of variables. The Wilcoxon test and Kruskal-Wallis test were applied for comparisons involving continuous and ordered classification variables, respectively. Fisher’s exact test was used for comparisons involving disordered classification variables. A significance level of P ≤ 0.05 was deemed statistically significant for all analyses.

Results

Identification of differentially expressed cell cycle-related genes in HR+/HER2- BC

The workflow chart for this study is presented in Fig. 1A. To explore the expression signature of cell cycle-related genes, we compared breast tumor and normal tissues obtained from the TCGA dataset. As depicted in Fig. 1B, a total of 4400 differentially expressed genes (DEGs) were identified (|log2(fold change) | > 0.585, FDR < 0.05, P ≤ 0.05). Among these DEGs, 469 were cell cycle-related genes, comprising 301 up-regulated and 168 down-regulated genes (Supplementary Table S3). The expression profiles of these 469 DEGs are displayed in Fig. 1C.

Fig. 1
figure 1

Differential cell cycle-related gene expression analysis in the TCGA dataset. (A) Flowchart illustrating the bioinformatics analysis process of this study; (B) Dysregulated cell cycle-related genes identified between tumor and normal tissues; (C) RNA expression profile of significantly differentially expressed cell cycle-related genes in tumor and normal tissues.

Full size image

Construction and validation of the prognostic model

We then proceeded to analyze the association between these 469 DEGs and patient survival in the TCGA training set. Three genes were identified to have a significant association with OS in HR+/HER2- BC (Fig. 2A). Differential expression analysis demonstrated downregulation of BRINP1 and CDKN2C, along with upregulation of SYCP2 in tumor versus normal control tissues (Fig. 2B). Figure 2 C illustrates that expression of CDKN2C (hazard ration [HR] = 0.41, 95% confidence interval [CI], 0.21–0.80) and BRINP1 (HR = 0.40, 95% CI, 0.21–0.79) correlates with better OS, while expression of SYCP2 correlates with poorer OS (HR = 2.62, 95% CI, 1.31–5.25). Subsequently, a three-gene based HR+/HER2- BC Prognostic Signature (HBPS) was established using the regression coefficients obtained from the multivariate Cox regression analysis (Supplementary Figure S1). The risk score calculation is expressed as follows: HBPS Score_os = (expression of CDKN2C) * (-0.7356) + (expression of BRINP1) *(-0.5094) + (expression of SYCP2) * 0.1942. Patients from the TCGA training set were stratified into high and low score groups using the median cutoff value (Supplementary Table S1). The heatmap visualizes that BRINP1 and CDKN2C were downregulated, while SYCP2 was upregulated in the high score group (P < 0.001) (Fig. 2D). Survival analysis demonstrated that patients with a high HBPS score had significantly worse OS compared to those with a low score (HR = 4.02; 95% CI, 1.93–8.35), P < 0.001, Fig. 2E). This association remained significant after adjusting for age and PR status (HR = 3.57; 95% CI, 1.70–7.50) (Table 1). Additionally, time-dependent ROC analyses were performed to assess the predictive accuracy of the HBPS score. The AUCs at 3 years and 5 years for HBPS score prediction were 0.733 (95% CI, 0.619–0.847) and 0.793 (95% CI, 0.704–0.883), respectively (Supplementary Figure S2A). In comparison, the AUCs for the TNM stage were 0.656 (95% CI, 0.523–0.785) at 3 years and 0.585 (95% CI, 0.473–0.697) at 5 years (Supplementary Figure S2B). Although there was no significant difference in AUC between the HBPS score and TNM stage at 3 years (P = 0.301), a significant difference was observed at 5 years (P = 0.001).

Fig. 2
figure 2

Identification and validation of the prognostic signature. (A) Hazard ratios (HR) and p-values from Cox proportional hazards analyses for all differentially expressed cell cycle-related genes; (B) Comparison of the expression of three prognostic-related genes between HR+/HER2 BC tumor samples and normal control samples in the TCGA dataset; (C) Kaplan-Meier (KM) OS curves for CDKN2C, BRINP1, and SYCP2 genes in the TCGA dataset; (D) RNA expression profile of three prognostic genes between two risk groups in the TCGA dataset; (E) KM OS curves for patients with high and low HBPS scores in the TCGA dataset; (F,G) KM OS curves for patients with high and low HBPS scores in the GSE96058 and METABRIC validation datasets.

Full size image
Table 1 Independent prognostic value of HBPS score in the TCGA dataset.
Full size table

To validate the prognostic capability of the HBPS score, patients from two independent validation sets were categorized into high score and low score groups based on their respective median cutoff values. Similar prognostic results were observed for the HBPS score in both the GSE96058 and METABRIC validation sets (P < 0.001) (Fig. 2F-G). Furthermore, in the multivariate Cox regression analysis, the HBPS score was identified as an independent prognostic factor for HR+/HER2- BC, with detailed results available in Supplementary Table S4.

Establishment of a prognostic nomogram in combining HBPS score and clinical features

To facilitate the clinical applicability of the prediction model, we integrated clinical features and HBPS score from the TCGA training set. A multivariable Cox regression analysis was conducted to construct a prognostic nomogram (Fig. 3A). The nomogram exhibited a C-index of 0.81 for OS prediction, indicating a relatively satisfactory predictive performance. The calibration curves further illustrated the good consistency between predicted and observed OS at 1-year, 3-year, and 5-year intervals (Fig. 3B).

Fig. 3
figure 3

Nomogram for predicting OS based on HBPS scores and clinical features in the TCGA dataset. (A) Nomogram illustrating the prediction of OS at 1-year, 3-year, and 5-year time points; (B) Calibration plot showcasing the performance evaluation of the nomogram. The gray dot line represents perfect calibration, where predicted probabilities match observed outcomes exactly. The colored dots are calculated by bootstrapping (resample: 1000) and represent the performance of the nomogram. The blue crosses and the vertical lines represent actual survival and 95% CIs measured by Kaplan-Meier analysis.

Full size image

Comparison of clinical features in two HBPS risk groups

We analyzed the correlations between HBPS scores and clinical characteristics of patients from the TCGA dataset. Significant differences in HBPS scores were observed based on age, menopausal status, and HRD scores. Specifically, HBPS scores were higher in patients over 50 years old, post-menopausal, and HRD-negative individuals (Fig. 4).

Fig. 4
figure 4

Correlation of HBPS score with clinical features in the TCGA dataset. A comparison of risk score difference between different clinical features including age, menopause status, HRD status, TNM stage, and progesterone receptor (PR) status. Wilcoxon test was conducted. NS: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001, ****: P ≤ 0.0001.

Full size image

The association between HBPS score and biological signatures

Spearman correlation analysis revealed that CDKN2C, BRINP1, and SYCP2 exhibited significant but modest correlations with several well-established cell cycle-related genes from the G2M checkpoint and E2F targets (Supplementary Figure S3), and the overall correlations strengths were not particularly strong (Supplementary Table S5). We further investigated whether the three-gene signature was associated with broader biological processes beyond cell cycle activity. Based on the TCGA dataset, ssGSEA analysis revealed significant differences in 30 Hallmark signaling pathways between the high and low score groups (Supplementary Figure S4). The high HBPS score group shows negative correlation with signaling pathways such as HALLMARK_KRAS_SIGNALING_DN and HALLMARK_IL2-STAT5 signaling (Fig. 5A). Furthermore, GSEA analysis using KEGG and GOBP showcased significant downregulation of multiple key immune-related pathways in high HBPS score patients, including the chemokine signaling pathway, JAK-STAT signaling pathway, cell chemotaxis, interleukin-6 production, and T cell activation (Fig. 5B-C, Supplementary Table S6-S7). The results of immune microenvironment analysis showed that low HBPS scores were associated with significantly higher levels of infiltration in B cells naïve, T cells CD8, T cells gamma delta, NK cells activated, and Macrophages M1. Conversely, high HBPS scores were associated with significantly higher levels of infiltration in T cells CD4 memory resting, T cells follicular helper, Monocytes, Macrophages M0, and Mast cells resting (Fig. 5D). Additionally, the high HBPS score group exhibited higher tumor purity levels (Fig. 5E) and demonstrated an immunosuppressive status, indicated by significantly decreased expression levels of checkpoint genes such as CD40LG, PDCD1, and SIRPA (Fig. 5F).

Fig. 5
figure 5

Association between HBPS score and biological signatures in the TCGA dataset. (A) Beeswarm and boxplots of hallmark pathways with significant differences between the two groups (showing a portion); (B) GSEA of the KEGG signaling pathway, demonstrating significantly down-regulated pathways in the high HBPS score group: Chemokine signaling pathway (NES = -1.671, p < 0.001), JAK_STAT signaling pathway (NES = -1.486, p = 0.001); (C) GSEA of the GOBP signaling pathway, demonstrating significantly down-regulated pathways in the high HBPS score group: Cell_chemotaxis (NES = -1.822, p < 0.001), Interleukin_6_production (NES = -1.783, p < 0.001), and T_cell_activation (NES =-1.739, p < 0.001). the Kanehisa laboratory have happily provided permission. (D) Differences in immune cell infiltration; (E) Differences in tumor purity; (F) Differences in expression of common immune checkpoints. NS: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001, ****: P ≤ 0.0001.

Full size image

The association between HBPS score and predicted drug sensitivity

As part of exploratory analyses, we stratified 51 breast cancer cell lines from GDSC based on HBPS scores and performed differential analysis, revealing significant differences in sensitivity to 10 different drugs between the high- and low-score groups. With the exception of AZD5582, an IAP antagonist that induces apoptosis, which exhibited increased sensitivity in the high-score group, the remaining nine drugs showed greater sensitivity in the low-score group (Supplementary Figure S5A).

Furthermore, we analyzed the differential response to immunotherapy between the two groups using a published metastatic urothelial cancer cohort, which includes RNA-seq and anti-PD-L1 treatment data31. Consistent with the immunosuppressive features observed in the high HBPS score group, patients who did not respond to anti-PD-L1 treatment generally had higher HBPS scores (Supplementary Figure S5B).

Validation of the expression and cellular roles of CDKN2C

To further investigate the clinical values of three HBPS score-related genes, we analyzed their expression in tumor and para-tumor tissues. The results showed that CDKN2C was lowly expressed in HR + HER2- tumors compared with non-HR + HER2- tumors and para-tumor tissues (Fig. 6A). We acquired a TMA cohort for verification in order to further confirm the results mentioned above. The results exhibited that CDKN2C was downregulated in HR + HER2- tumors compared with para-tumor tissues, and mainly located in tumor cells (Fig. 6B-C). In addition, CDKN2C+ tumor cells were positively correlated with CD8+ T cells (Fig. 6D-E). Moreover, in vitro assays indicated that overexpression of CDKN2C notably inhibited MCF-7 and T47D cells proliferation and induced G2/M phase arrest (Fig. 7A-D). Furthermore, overexpression of CDKN2C significantly suppressed cell migration, and invasion (Fig. 8A-B). To conclude, CDKN2C deficiency was associated with immuno-cold tumor microenvironment and enhanced HR + HER2- breast cancer cells aggressiveness.

Fig. 6
figure 6

Expression of CDKN2C in HR+/HER2- breast cancer. (A) Expression of CDKN2C, BRINP1, and SYCP2 in para-tumor, HR + HER2-, and other tumors. NS: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001.  (B) Representative images uncovering CDKN2C expression in HR + HER2- tumor and para-tumor tissues in the in-house cohort. (C) Quantitative analysis of CDKN2C in HR + HER2- tumor and para-tumor tissues. *: P ≤ 0.05.  (D,E) Representative images uncovering Co-expression of CDKN2C and CD8 in HR + HER2- tumor tissues in the in-house cohort and correlation analysis.

Full size image
Fig. 7
figure 7

Cellular roles of CDKN2C in HR + HER2- breast cancer cells. (A) The overexpression efficiency of CDKN2C in HR + HER2- breast cancer cells was assessed by Western blotting. NS: P > 0.05, ***: P ≤ 0.001.    (B) The proliferative capacity of control and CDKN2C-overpressed cells was examined by clone formation assay. **: P ≤ 0.01, ***: P ≤ 0.001. (C) Cell cycle analysis of control and CDKN2C-overpressed cells was examined by flow cytometry assay. (D) The expression of PCNA in control and CDKN2C-overpressed cells was examined by Western blotting.

Full size image
Fig. 8
figure 8

The effects of CDKN2C on migration and invasion of tumor cells. (A) The migratory capacity of control and CDKN2C-overpressed cells was examined by wound healing assay. *: P ≤ 0.05.  (B) The migratory and invasive capacity of control and CDKN2C-overpressed cells were examined by Boyden chamber assay. *: P ≤ 0.05, **: P ≤ 0.01.

Full size image

Discussion

In this study, we developed a HBPS prognostic model for HR+/HER2- BC based on the TCGA dataset and validated it in two independent datasets. The results suggested that patients with high HBPS scores tended to have worse outcomes compared to those with low HBPS scores. Compared to the low HBPS score group, the high-score group demonstrated significant downregulation in the HALLMARK_KRAS_SIGNALING_DN, HALLMARK_IL2_STAT5_SIGNALING, and other immune-modulatory pathways. Additionally, the high HBPS score group exhibited higher tumor purity and a stronger immunosuppressive tumor microenvironment. Furthermore, mIHC staining showed a positive correlation between CDKN2C-positive tumor cells and CD8 + T cells. In vitro assays revealed that overexpressing CDKN2C led to reduced proliferation and invasion in cell lines.

The three genes that constitute the HBPS are involved in cell cycle activities and may also impact the immune microenvironment and invasiveness. The CDKN2C gene encodes a cyclin-dependent kinase inhibitor (CDKi) that interacts with CDK4 or CDK6, preventing CDK kinase activation and thereby controlling the G1 phase of the cell cycle. CDKN2C has been reported to have independent prognostic significance in various types of cancer and is associated with the immune microenvironment32. Additionally, CDKN2C deficiency was found to promote growth and invasion in lung adenocarcinoma33. The BRINP1 gene encodes a protein that inhibits cell proliferation by negatively regulating the transition from the G1 phase to the S phase and is involved in a non-classical apoptotic form of cell death. Hypermethylation of BRINP1 has been observed in bladder cancer34. As for the SYCP2 gene, it encodes a protein that is a major component of synaptic complexes and may bind to DNA in the scaffold attachment region. Overexpression of SYCP2 has been found to be associated with poor prognosis in BC and is involved in DNA methylation and immune infiltration pathways35. Similarly, in this study, increased expression of SYCP2 was found to be positively correlated with the risk score. We observed that the mutation prevalence of the genes CDKN2C, BRINP1, and SYCP2 among HR+/HER2- breast cancer patients in the TCGA dataset is very low (approximately 1.9% overall). The differential expression levels of these genes could possibly be attributed to epigenetic heterogeneity or modulation by specific signaling pathways.

In this study, we observed that patients with high HBPS scores had a significant downregulation of the HALLMARK_KRAS_SIGNALING_DN Gene Set (genes down-regulated by KRAS activation). It is well known that the activation of the KRAS pathway triggers multiple downstream signaling pathways, including the MAPK pathway and the PI3K pathway, which play crucial roles in promoting cell survival and proliferation36. Additionally, CDK4/6 can be affected by KRAS through multiple signaling pathways, and the activation of the KRAS pathway may be one of the mechanisms contributing to resistance to CDK4/6 inhibitors37. Furthermore, the activation of the KRAS pathway also plays an important role in the formation of an immunosuppressive tumor microenvironment36. These observations support that patients with high HBPS scores tend to suffer a worse prognosis.

We also noticed that patients with low HBPS scores have upregulated immune-regulatory pathways, such as the IL-2-STAT5 and JAK-STAT pathways. Previous studies have shown that IL-2 regulates PD-1 gene expression by activating the STAT5 signaling pathway, and enhances the sensitivity of CD8 + T cells to PD-1-expressing cells, potentially increasing the cytotoxic capability of T cells against tumor cells38. Moreover, we found that low HBPS scores were positively correlated with both higher infiltration levels of CD8 + immune cells and increased expression of immune checkpoints. Additionally, in vitro experiments confirmed that overexpression of the CDKN2C gene in the HBPS was positively associated with CD8 + T cells and reduced tumor invasiveness. These findings may explain the improved clinical prognosis observed in the low-score group. Notably, three hallmark pathways were significantly upregulated in the high-score group: HALLMARK_ESTROGEN_RESPONSE_EARLY, HALLMARK_PROTEIN_SECRETION, and HALLMARK_ESTROGEN_RESPONSE_LATE. Prior research has indicated that HALLMARK_ESTROGEN_RESPONSE_LATE is significantly downregulated in melanoma patients responding to immunotherapy39. This suggests that the high HBPS score group may exhibit reduced efficacy to immunotherapeutic interventions. This finding aligns with the differential response to immunotherapy observed between the two groups in the metastatic urothelial carcinoma cohort.

In conclusion, in this study, we constructed an effective prognostic model for HR+/HER2- BC based on the expression of cell cycle-related genes. Moreover, the identified broader biological differences between the two groups shed light on the underlying factors contributing to the variations in prognosis. These findings not only contribute to a better understanding of HR+/HER2- BC but also hold promise for personalized medication and addressing drug resistance challenges. Further researches are warranted to validate the clinical application value of the HBPS score.