Association of CA12 with calcification in ER+/HER-2 negative breast cancer

Abstract

Background

Calcification in breast cancer is closely associated with ER+/HER2- breast cancer, and identifying key molecular markers is crucial for early diagnosis and treatment.

Methods

We integrated bioinformatics analyses of data from the TCGA and GEO databases with in vitro experiments on MCF-7 and MDA-MB-231 cell lines to explore the role of CA12 in breast cancer calcification.

Results

Our analysis identified 132 calcification-related genes, with CA12 highlighted as a significant biomarker. In vitro experiments showed that osteogenic induction promoted calcific nodule formation in ER-positive MCF-7 cells with concomitant CA12 upregulation, though this increase did not reach statistical significance; notably, ER-negative MDA-MB-231 cells exhibited comparable CA12 upregulation without mineralization. We propose integrating CA12 detection with imaging to boost diagnostic accuracy for early ER+/HER2- breast cancer, enabling timely endocrine therapy and improving prognosis.

Conclusion

This study offers new insights for early breast cancer diagnosis and individualized treatment, along with proposing CA12 as a potential biomarker for breast cancer calcification.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05123-9.

Introduction

Breast cancer is one of the most common malignancies among women worldwide. The development of diagnostic and therapeutic strategies for breast cancer often relies on a variety of clinical and radiological features. Among these, breast microcalcifications serve as a crucial radiological biomarker, playing an essential role in the early diagnosis, prognostic evaluation, and treatment decision-making for breast cancer. With the advancement of imaging technologies and in-depth research in molecular biology, our understanding of breast microcalcifications has evolved from a mere radiological feature to a comprehensive understanding of its molecular mechanisms and clinical significance.

Early clinical observations have linked breast microcalcifications to poor breast cancer prognosis. In a retrospective study, Pálka et al. showed that certain microcalcification types (e.g., casting microcalcifications) are significantly associated with early recurrence and higher mortality in breast cancer patients [1]. In radiological diagnosis, the morphological features of microcalcifications (including size, shape, and distribution) have long been key diagnostic indicators. Recently, advanced imaging techniques such as digital breast tomosynthesis (DBT) have enabled more accurate detection and assessment of breast microcalcifications [2]. However, traditional radiological analysis still has clinical limitations [3]. To address this, researchers have developed various computer-aided diagnosis (CAD) technologies, such as deep learning algorithms [4] and a new assessment system based on the regression detectability index (RDI) [5], which have significantly improved the sensitivity and specificity of microcalcification detection. Recently, Scott et al. refined the microcalcification morphological classification system, providing a more precise standard for clinical diagnosis [6].

Regarding molecular mechanisms, scientists have gradually uncovered the biological basis of microcalcification formation. In 2016, Sharma et al. first systematically explored the molecular characteristics of pathological microcalcifications, revealing that their formation involves specific gene expression profiles and is closely related to non-coding RNA regulatory networks [7]. Subsequent studies have found that microcalcifications may be associated with mutations in breast cancer susceptibility genes (e.g., BRCA1/2) [8]. Notably, Li et al. found that in hormone receptor-positive (HR+)/HER2-negative (HER2-) breast cancer, microcalcification frequency is significantly increased and is closely related to worse clinical prognosis [9]. These findings deepen our understanding of breast cancer heterogeneity and offer new avenues for personalized treatment strategies.

Overall, research on breast microcalcifications has evolved from basic imaging descriptions to in-depth exploration of their molecular mechanisms and clinical implications. Existing evidence shows breast microcalcifications aren’t just key diagnostic markers for breast cancer, but may also be a biological feature of specific molecular subtypes, especially the ER+/HER2- type. Future studies should clarify microcalcification’s molecular mechanisms and assess its value as a therapeutic target or prognostic biomarker, aiding precise breast cancer diagnosis, individualized treatment, and improved patient outcomes.

Method

Data collection

Gene information related to breast malignancies with calcification was collected from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/), including 109 samples of breast malignancies with or without calcification in GSE146558 (94 ER+/HER2- and 15 ER-/HER2-). Gene expression and clinical data were downloaded from the TCGA-BRCA dataset, which includes 238 ER- and 808 ER+ breast tumor samples. After data processing, R (version 4.4.1) was used to select 64 ER+/HER2- and 30 ER-/HER2- tumor samples measured by immunohistochemistry (IHC).

Identification of differentially expressed genes

Identification of Differentially Expressed Genes (DEGs)The raw gene expression matrices downloaded were normalized and processed using R software (version 4.4.1). The R package “limma” was employed to identify DEGs from both the GSE146558 and TCGA datasets. In the GSE146558 dataset, DEGs were identified with an adjusted P value of < 0.05 and |log2 fold change (FC)| ≥ 1.0, while in the TCGA dataset, DEGs were identified with an adjusted P value of < 0.05 and |log2 FC| ≥ 1.0. Volcano plots and differential gene clustering heatmaps were generated using R software.

Construction and module identification of weighted gene co-expression networks

Weighted Gene Co-expression Network Analysis (WGCNA) Network Construction and Module IdentificationA co-expression network was constructed using the “WGCNA” package in R, with a soft threshold applied to ensure scale-free network topology. Hierarchical clustering trees were utilized to identify gene modules, and modules with robust connectivity were generated using hierarchical clustering based on the topological overlap matrix (TOM). The correlation between disease status and modules was assessed using Pearson correlation coefficients, and the module with the highest disease correlation was selected for subsequent analysis. Genes within this module were used for further analyses. Hub genes were identified based on module membership (MM > 0.5) and gene significance (GS > 0.5), representing strong co-expression within the module and significant correlation with clinical traits, respectively.

Identification of shared genes and pathway enrichment analysis

Identification of Shared Genes and Pathway EnrichmentCommon genes identified by both DEG analysis and WGCNA were displayed using Venn diagrams. The “clusterProfiler” and “org.Hs.eg.db” packages were used to investigate the pathways and functions associated with these genes through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses.

Screening strategy for hub genes

Hub Gene IdentificationTo identify hub genes with diagnostic significance, three machine learning algorithms were employed: Least Absolute Shrinkage and Selection Operator (LASSO) regression, Random Forest (RF) algorithm, and Support Vector Machine Recursive Feature Elimination (SVM-RFE). LASSO was performed using the “glmnet” package to prevent overfitting. The RF model was constructed using the “randomForest” package. SVM-RFE was conducted using the “rsm” package with a 10-fold cross-validation strategy. The top 6 genes were selected based on their relevance ranking. Subsequently, the intersection of genes obtained from the three machine learning methods using the “VennDiagram” package was used to evaluate the diagnostic value of ER status in ER+/HER2- breast malignancies with calcification.

Cell culture system and experimental reagent preparation

Cell Culture and ReagentsHuman breast cancer cell lines MCF-7 and MDA-MB-231 were obtained from ATCC. These cell lines were cultured in DMEM medium (C11995500BT, Gibco) supplemented with 10% fetal bovine serum (F801-500, BDBIO, China) and 1% penicillin/streptomycin (C3420-0100, BI). For mineralization assays, cells were seeded in 24-well tissue culture plates and grown to 70–80% confluence. The culture medium was then replaced with fresh complete medium containing 5% fetal bovine serum (control group) or osteogenic mixture (OC: 10 mM β-glycerophosphate + 50 µg/mL ascorbic acid, DEX: 100 nM dexamethasone) (experimental group). Cells were cultured under these conditions for 7 days, with half the medium in each well replaced every two days. After 7 days, mineralization was assessed using Alizarin Red S staining.Three independent biological replicates were performed for each experimental condition.

Alizarin red staining

For Alizarin Red staining, cells in each well were fixed with 1 mL of 4% paraformaldehyde on ice for 15 min. After fixation, cells were rinsed once with distilled water and then incubated with 1 mL of Alizarin Red S solution (Servicebio, China) at room temperature for 30 min to detect calcium release. Following incubation, cells were washed twice with distilled water, and the stained nodules were captured using ZEN 3.1 imaging software.

Western blotting

Total protein was extracted from MCF-7 and MDA-MB-231 cells subjected to osteogenic induction or cultured only in complete medium using RIPA lysis buffer (Servicebio, China). Protein concentrations were measured using an enhanced BCA assay kit (Servicebio, China). Equal amounts of protein (30 µg) were separated by 15% SDS-PAGE and transferred to PVDF membranes (Servicebio, China). The membranes were blocked with commercial blocking solution (NCM Biotech, China) and then incubated overnight at 4 °C with primary antibodies against CA12 (1:1000, PK07922, Abmart, China) and GAPDH (1:4000, GB15004-100, Servicebio, China). After three washes with TBST, the membranes were incubated with HRP-conjugated secondary antibodies at room temperature for 1 h, followed by three additional washes with TBST. Protein bands were detected using ECL reagent (Servicebio, China) in a Bio-Rad system and quantified using ImageJ software (NIH, USA).All Western blot analyses were conducted with three independent biological replicates to ensure reproducibility of the results.

Clinical data collection

A total of 446 patients diagnosed with breast cancer and treated with surgery at the Breast Clinic of the Second Affiliated Hospital of Xi’an Jiaotong University from 2020 to 2024 were included. The inclusion criteria were: (1)female gender; (2)first-time diagnosis of breast cancer; (3)presence of breast calcification on mammography; (4)treatment with calcification biopsy and surgery. The exclusion criteria were: (1)pregnancy or breastfeeding; (2)incomplete clinical data. Among them, 292 were ER-positive and 154 were ER-negative, with 114 having calcification and 332 without. The study was approved by the Ethics Committee of the Second Affiliated Hospital of Xi’an Jiaotong University.

Statistical analysis

All statistical analyses were performed using R software (version 4.4.1). For differential expression analysis, genes with adjusted P value < 0.05 and |log2 fold change| ≥ 1 were considered significantly differentially expressed. In WGCNA, Pearson correlation was used to assess module-trait relationships, and P values were adjusted for multiple testing using Benjamini-Hochberg correction. Module membership (MM) and gene significance (GS) were calculated to identify core genes with |MM| ≥ 0.5 and |GS| ≥ 0.5.

For machine learning, LASSO regression employed 10-fold cross-validation to select the optimal tuning parameter. Random forest analysis ranked gene importance based on mean decrease accuracy, and SVM-RFE was performed with 10-fold cross-validation to minimize feature selection error.

For clinical data, categorical variables are presented as counts (n) and percentages (%). The chi-square (χ²) test was used to compare the distribution of ER status between calcification and non-calcification groups, with P < 0.05 considered statistically significant.For in vitro experiments, all assays were performed with three independent replicates. Data are presented as mean ± standard deviation (SD). Comparisons between two groups were conducted using unpaired t-test or one-way ANOVA as appropriate. Statistical significance was set at P < 0.05.

For quantification of Alizarin Red staining, images were captured under identical lighting conditions using ZEN 3.1 imaging software and converted to RGB format. The red channel was extracted, and color threshold was applied to select positively stained areas. The percentage of positive staining area (%Area) was calculated for each field, with three wells per group.

Result

Clinical data collection and analysis results

In this study, a total of 446 patients with breast cancer were included, comprising 292 ER-positive and 154 ER-negative cases. Among them, 114 cases had calcification, while 332 cases did not. All patients underwent mammography and biopsy localization, and the calcified lesions were surgically removed for molecular typing. The results showed a significant difference in the distribution of ER status between the calcification and non-calcification groups [OR = 1.77(1.10–2.85), P = 0.018], suggesting a potential association between calcification and ER status(Table 1).

Table 1.

Association between ER status and calcificationin malignant tumors

Variable	ER-status			Statistics
	ER-positive	ER-positive	Total	OR (95%Cl)	P
Calcification				1.77(1.10–2.85)	0.018*
With calcification	85	29	114
Without calcification	207	125	332
Total	292	154	446

^*P < 0.05

DEGs analysis in TCGA and GEO

Gene expression data for estrogen receptor-positive (ER+) and estrogen receptor-negative (ER-) breast cancer (BC) were obtained from The Cancer Genome Atlas (TCGA) database. Additionally, data related to ER+ breast cancer with or without calcification from the Gene Expression Omnibus (GEO) database (GSE146558) were acquired and normalized. In the TCGA dataset, a total of 1303 differentially expressed genes (DEGs) were identified, comprising 901 upregulated and 402 downregulated genes. In the GSE146558 cohort, compared with ER+ breast cancer without calcification, 260 DEGs were identified in calcified tissues, including 158 upregulated and 102 downregulated genes. The DEGs from both TCGA and GSE146558 were displayed in heatmaps and used for further analysis(Fig. 1).

Fig. 1.

Differential gene screening in public databases: a, b Volcano plot & heatmap of DEGs from TCGA; c, d Volcano plot & heatmap of DEGs from GEO(GSE146558)

Construction of co-expression network using weighted gene co-expression network analysis (WGCNA)

To identify co-expressed gene modules associated with calcification in ER-positive breast cancer, we constructed a weighted gene co-expression network using the TCGA breast cancer dataset. Samples with outliers were first removed through sample clustering(Fig. 2a and f), followed by selection of an appropriate soft-thresholding power to achieve scale-free topology(Fig. 2b and g). Genes were then clustered based on dissimilarity measure (1-TOM)(Fig. 2c and h), and assigned to distinct co-expression modules(Fig. 2d and i). In WGCNA, module names are randomly assigned colors used solely to distinguish different co-expression gene clusters and carry no inherent biological meaning. Our primary focus was on the correlation between these modules and clinical traits, specifically ER and HER2 status.

Fig. 2.

a–f WGCNA construction via TCGA database. a Outlier elimination and sample screening. b Soft-threshold selection. c Dendrogram of all expressed genes clustered by dissimilarity (1-TOM). d Heatmap of module eigengene correlations. e Scatter plot of Cox regression degrees and p-values in the turquoise module (x-axis: regression degree; y-axis: gene significance; each dot: a gene). f–j WGCNA construction utilizing GSE146558 database. f Outlier removal and sample filtration. g Determination of soft-thresholding power. h Dendrogram of all expressed genes clustered based on dissimilarity measure (1-TOM). i Heatmap depicting correlations among module eigengenes. j Scatter plot of Cox regression degrees and p-values within the grey module (x-axis: regression degree; y-axis: gene significance; each point represents a gene)

Based on the breast cancer (BC) gene expression data from The Cancer Genome Atlas (TCGA) database, the analysis revealed that several gene modules were significantly associated with estrogen receptor (ER) and human epidermal growth factor receptor 2 (HER2) status. Among them, the turquoise module (MEturquoise) showed the strongest positive correlation with ER status (r = 0.78, P = 1 × 10⁻¹⁸³), while its association with HER2 status was weak, indicating that the genes in this module may play an essential role in ER-positive breast cancer, independent of HER2 status(Fig. 2e). This module contained 1,441 genes, with highly connected genes including ESR1, PGR, CA12 and GATA3 (Supplementary Table S1, module membership shown in Fig. 2e). In contrast, other modules such as the brown, blue, green, yellow, and red modules exhibited no significant correlation with either ER or HER2 status (|r| < 0.2, P > 0.05), suggesting that the genes within these modules are not directly associated with hormone receptor status and may be involved in other biological processes.

Similarly, using the breast cancer gene expression data from the Gene Expression Omnibus (GEO) database, we identified several co-expression modules significantly associated with ER/HER2 status through WGCNA. As noted above, the module colors are arbitrary labels assigned by the algorithm to distinguish distinct gene clusters. The grey module (MEgrey) exhibited the strongest positive correlation with ER status (r = 0.75, P = 2 × 10⁻¹⁷), suggesting that the genes in this module may play a crucial role in ER-positive breast cancer with calcification(Fig. 2j). This module contained 2,243 genes, with highly connected genes including ESR1, GATA3, FOXA1, and CA12 (Supplementary Table S2, gene significance shown in Fig. 2j). Other modules in the GEO analysis, including the brown, blue, green, and red modules, did not show significant associations with ER/HER2 status (|r| < 0.2, P > 0.05), further supporting the specificity of the grey module in relation to ER-positive breast cancer.

By calculating gene significance (GS) and module membership (MM), we further validated the association of these modules’ genes with clinical traits and screened out core genes (|MM| ≥ 0.5 and |GS| ≥ 0.5) for subsequent analysis. All analyses were conducted using the standard WGCNA workflow, with network construction meeting the scale-free topology criterion (R² ≥ 0.85), and results were corrected for multiple hypothesis testing(Fig. 2).

Identification and analysis of hub genes

To identify hub genes associated with estrogen receptor-positive (ER+) breast cancer with calcification, we performed an intersection analysis of differentially expressed genes (DEGs) and module core genes derived from The Cancer Genome Atlas (TCGA) database and Gene Expression Omnibus (GEO) datasets. A total of 132 genes related to breast cancer calcification were identified through this analysis(Supplementary Table S3). Subsequent biological enrichment analysis was conducted on these genes. Gene Ontology (GO) functional annotation revealed that these genes are primarily involved in epithelial morphogenesis, mammary epithelial development, female sex differentiation, and gland morphogenesis. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis showed significant enrichment in estrogen signaling pathways, neuroactive ligand-receptor interactions, parathyroid hormone synthesis, secretion and action, and proteoglycan signaling pathways in cancer(Fig. 3).

Fig. 3.

Hub Gene Selection and Enrichment Analysis. a Venn Diagram of Shared Genes Between TCGA and GSE146558 Datasets. b GO Biological Process Enrichment Analysis of Shared Genes. c KEGG Pathway Analysis of Shared Genes

Machine learning-based selection of differentially expressed genes

To further evaluate the prognostic significance of genes associated with breast cancer calcification, we employed a series of machine-learning techniques to analyze the 132 hub genes. Random forest analysis ranked gene importance (Fig. 4c) and identified 28 genes with high predictive and diagnostic value for breast cancer calcification, including ESR1, GATA3, AGR3, AGR2, TBC1D9, CA12, FSIP1, and TFF1. LASSO regression was performed to select optimal features (Fig. 4a and b), identifying 17 genes closely associated with ER-positive breast cancer with calcification, including SLC7A2, CA12, ESR1, ZNF516, VGLL1, PADI2, PPP1R3C, BBOX1, FSIP1, TFF1, KLHDC9, AGR3, SOX11, FDCSP, ABAT, GREB1, and GATA3-AS1. Additionally, the Support Vector Machine-Recursive Feature Elimination (SVM-RFE) model was applied to further refine feature selection (Fig. 4d), identifying 15 key genes, including ESR1, AGR3, GATA3, TBC1D9, CA12, THSD4, FSIP1, ERBB4, ANXA9, DNALI1, GATA3-AS1, TFF1, FOXA1, GREB1, and SLC16A6 (Supplementary Table S4).

Fig. 4.

Machine learning-based screening of differentially expressed genes.a-b LASSO regression analysis of 28 hub genes. a Cross-validation for tuning parameter selection(dashed line marks optimal lambda). b LASSO coefficient profiles of candidate genes(each line represents one gene). c Random Forest analysis of 28 hub genes showing gene importance ranking. d SVM-RFE model for 28 hub genes with cross-validation(optimal feature number indicated). e Venn diagram integrating the results of three machine learning methods, identifying ESR1, CA12, FSIP1, AGR3, TFF1, and GREB1 as essential factors in ER+/HER2- breast cancer with calcification

By integrating the results from these three machine learning approaches, six genes—ESR1, CA12, FSIP1, AGR3, TFF1, and GREB1—were consistently identified as key genes associated with ER-positive breast cancer with calcification, as shown in the Venn diagram (Fig. 4e). Therefore, this study focused on these six candidate genes for subsequent analyses.

Mineralization capacity and CA12 expression in breast cancer cell lines

This study aimed to screen human breast cancer cell lines for their mineralization capacity. MDA-MB-231 and MCF-7 cells were cultured for 7 days in either control medium or medium supplemented with an osteogenic induction cocktail plus dexamethasone (OC + Dex). Mineralization capacity was assessed using Alizarin Red S staining. The results showed that under these experimental conditions, MDA-MB-231 cells did not exhibit mineralization capacity (Fig. 5a, b). In contrast, MCF-7 cells cultured in OC + Dex medium for 7 days displayed prominent orange-red calcific nodule deposits (Fig. 5d), with the boxed areas indicating representative calcific nodules, while no calcification was observed in MCF-7 cells cultured in control medium (Fig. 5c). Quantitative analysis further confirmed that OC + Dex treatment significantly increased the mineralization area in MCF-7 cells compared to the control group (****p < 0.001, Fig. 5e). This result suggests that estrogen receptor-positive MCF-7 cells possess the potential to induce calcification upon stimulation with osteogenic promoters.(****p < 0.001, Fig. 5e).

Fig. 5.

a–d Alizarin Red staining was performed to detect the formation of mineralized nodules (4× objective). a Control MDA-MB−231 cells. b OC + DEX-induced MDA-MB−231 cells. c Control MCF−7 cells. d OC + DEX-induced MCF−7 cells. e Quantitative analysis of Alizarin Red staining in MCF−7 cells. Data were obtained from three independent experiments. ****p < 0.001.f. Western blot analysis of CA12 protein expression in MCF−7 and MDA-MB−231 cells. Left panel shows representative protein bands, and right panel shows the quantitative analysis of CA12/GAPDH expression. ns indicates no statistical significance

Furthermore, we examined CA12 protein expression levels in MDA-MB-231 and MCF-7 cells from both the control and experimental groups. Western blot analysis showed an upward trend in CA12 expression in both MCF-7 and MDA-MB-231 cells under OC + Dex culture conditions. However, statistical analysis based on three independent biological replicates indicated that this increase did not reach statistical significance in either cell line compared to their respective controls (n.s., Fig. 5f). Of note, the upregulation of CA12 appeared more pronounced in MDA-MB-231 cells than in MCF-7 cells in the blot images, although its basal expression level was lower. Despite the increase in CA12 protein in MDA-MB-231 cells, no mineralized nodules were observed in this cell line.

Discussion

In recent years, there have been remarkable advancements in the early diagnosis and treatment of breast cancer. One of the key methods in this progress is mammography. The 2019 ESMO guidelines once again emphasized that breast cancer diagnosis should be based on bilateral mammography, and the definitive diagnosis relies on image-guided core needle biopsy [10]. Although some studies indicate that mammography has certain limitations compared to MRI [11], it still holds an irreplaceable position in the screening of breast microcalcifications due to its simplicity [12].

Calcification is one of the important signs in mammography. The distribution pattern of calcifications in breast cancer is closely related to clinical features such as the degree of invasiveness and lymph node metastasis. Highly suspicious calcifications in breast cancer may be associated with a higher risk of recurrence and have significant diagnostic value [13]. The presence of a calcification site may indicate the disruption of the microenvironment homeostasis in normal breast tissue [14], suggesting abnormal proliferation, regulation, and apoptosis of breast cancer cells [15], which may have a potential impact on tumor formation and progression [16]. Calcification may be closely related to breast cancer subtype differentiation [17].

The correlation between ER status and calcification in breast cancer has been corroborated by numerous studies. The incidence rate of malignant calcification in ER-positive tumors is markedly higher than that in ER-negative tumors, a disparity potentially attributable to the promotion of osteoblast-like differentiation by estrogen signaling. Li et al. discovered that patients with HR+/HER2- breast cancer exhibiting highly suspicious calcification demonstrated genomic instability and cell cycle activation, suggesting potential benefits from CDK4/6 inhibitor therapy [18]. Furthermore, the interaction between Orai1α and SPCA2 in ER-positive cells modulates calcium influx, thereby influencing microcalcification formation [19]. These findings indicate that ER status is not only clinically associated with calcification but may also participate in its molecular mechanisms through the regulation of calcium metabolism genes. The identification of CA12 as an ER-related gene in this study offers a novel perspective for elucidating this mechanism.

According to previous studies and clinical data, we speculate that breast cancer with calcifications is more likely to be ER+/HER2-, which provides an imaging basis for selecting endocrine therapy for early breast cancer. However, in current research, calcification detection relies on imaging, and there is a lack of reliable biomarkers to directly link calcification with ER status. This may lead to lagging treatment strategies for some patients and restrict the implementation of early precision therapy. Therefore, this study uses bioinformatics and in vitro experiments to investigate the essential molecules of ER-positive breast cancer with calcifications. The study aims to enhance early detection, improve treatment strategies, and promote timely prevention.

This study first systematically explored the expression pattern of CA12 in calcified ER + breast cancer and its potential as a biomarker. Bioinformatics analysis identified 132 calcification-related genes, mainly enriched in epithelial morphogenesis, mammary epithelial development, gland morphogenesis, and the estrogen signaling pathway. Notably, among these genes, ESR1, CA12, FSIP1, AGR3, TFF1, and GREB1 are candidate genes of ER+/HER2- breast cancer with calcification. Co-linearity assessment and clinical value investigation demonstrated that ESR1 and CA12 are closely related to ER-positive breast cancer with calcification.

ESR1 is an essential factor of HR+/HER2- breast cancer, which undergoes activating mutations leading to ligand-independent ERα activation, which is a major mechanism of aromatase inhibitor resistance [20]. These mutations, directly linked to the ER signaling pathway and nearly exclusive to ER+ tumors, can be dynamically monitored via liquid biopsy (ddPCR/NGS of ctDNA) [21], guiding treatment optimization such as switching to SERDs or combining CDK4/6 inhibitors [22]. ESR1 mutations directly affect ER function and are a central feature of endocrine therapy resistance [23].

Meanwhile, CA12 also is an essential transmembrane enzyme, which regulates pH homeostasis and ion metabolism, playing an important role in physiological and pathological processes [24]. In cancers, high CA12 expression promotes tumor progression, metastasis [25], and immune evasion [26] by acidifying the tumor microenvironment, which may suppress anti-tumor immunity and correlate with treatment resistance. Moreover, CA12 synergizes with CFTR in microenvironment remodeling of pancreatic cancer [27]. Its expression correlates significantly with patient prognosis, making it a potential biomarker and therapeutic target.

The MCF-7 and MDA-MB-231 cell lines were selected based on their representative molecular characteristics. MCF-7 is a classic ER-positive/HER2-negative breast cancer cell line that retains estrogen responsiveness and an epithelial phenotype, while MDA-MB-231, as a representative triple-negative breast cancer cell line, exhibits a mesenchymal phenotype and strong invasive capacity. This experimental design aimed to elucidate the differential responses of distinct molecular subtypes to the osteogenic microenvironment.

Notably, osteogenic induction promoted an upward trend in CA12 expression in both cell lines, yet mineralized nodule formation was observed exclusively in MCF-7 cells. This finding is consistent with the study by Alarmo et al. [28], which demonstrated that under equivalent osteogenic induction conditions, MCF-7 cells possessing an epithelial phenotype and osteogenic differentiation potential exhibited calcific deposition, whereas MDA-MB-231 cells with a mesenchymal phenotype displayed enhanced migratory and invasive capabilities.

MCF-7 cells developed calcified nodules upon osteogenic induction, whereas MDA-MB-231 cells did not, suggesting that calcification capacity is estrogen receptor ER-dependent. The expression of the CA12 protein exhibited an upward trend in both cell lines, and although this increase did not achieve statistical significance, it closely aligned with the clinical association between CA12 and calcification identified through bioinformatics analysis. This alignment implies that CA12 may serve as a concomitant biomarker for calcification in ER-positive breast cancer. The discrepancy between protein levels and phenotypic alterations may stem from post-translational regulatory mechanisms or limitations in detection sensitivity. Notably, despite the upregulation of CA12 in MDA-MB-231 cells, these cells did not exhibit mineralization, further reinforcing the notion that the functional relationship between CA12 and calcification is contingent upon an ER-positive cellular context. These findings offer preliminary experimental evidence supporting the specific role of CA12 in calcification within ER+/HER2-negative breast cancer, with the underlying causal mechanisms necessitating further validation through functional manipulation experiments.

The present study has several limitations. First, the bioinformatics analysis revealed a statistical association between CA12 and calcification rather than a direct causal relationship, and the underlying molecular regulatory mechanisms require further elucidation through subsequent functional experiments. Second, although the in vitro osteogenic induction model recapitulated essential features of the mineralized microenvironment, it cannot fully replicate the complex in vivo context of tumor-associated microcalcifications, which could be addressed in future studies incorporating animal models. Third, due to the limitations of publicly available dataset sizes, multiple algorithmic strategies were employed to identify consistently reproducible essential molecules and enhance the robustness of the results, although external validation in independent cohorts is still required. Fourth, the clinical correlation analysis was based on retrospective data; additionally, the lack of statistical significance in CA12 protein upregulation may reflect technical or biological factors, without contradicting the strong clinical association observed.

Conclusion

In summary, through bioinformatics screening and in vitro experimental validation, this study identifies CA12 as an essential molecule associated with calcification in ER-positive breast cancer. The results demonstrate that osteogenic stimulation induces calcific nodule formation in MCF-7 cells but not in MDA-MB-231 cells, suggesting that the response of breast cancer cells to the osteogenic microenvironment is dependent on molecular subtype. Although CA12 expression showed an upward trend in both cell lines, calcification occurred exclusively in ER-positive MCF-7 cells, indicating that the association between CA12 and calcification may be contingent upon an ER-positive cellular context.

This study provides new insights into the molecular mechanisms underlying breast cancer microcalcifications and suggests that CA12 may serve as a potential auxiliary marker for ER-positive breast cancer with calcification, offering a molecular complement to imaging-based assessment. Further studies are warranted to elucidate the direct regulatory mechanisms of CA12 in the calcification process and to validate its feasibility and translational value as a biomarker in clinical samples.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Zhu Y and Liu D conceived and designed the study and completed the article. Zhu Y and Du Q carried out experiments and the analysis of all experimental data. Ma Y, Xu N and Luo Y collected and analyzed clinical data. Zhu Y and Du Q were responsible for creating figures and tables of the study and edited the manuscript. Liu D scrutinized the whole process of this study and edited the initial manuscript. All authors contributed to the article and approved the submitted version.

Funding

This work was funded by: Natural Science Basic Research Program of Shaanxi Province(2025JC-YBMS-1090).Project of the Xi'an Jiaotong University Education Foundation(XJYG2025-SFJJ026).The Science and Technology Project of Shaanxi Province (2022SF-496). The Funds of the Second Affiliated Hospital of Xi’an Jiaotong University for Scientists. No.RC (GG201807). The Funds of the Second Affiliated Hospital of Xi’an Jiaotong University. No.YJ (ZYTS)2019012/No.2020YJ(ZYTS)226. The Xinrui Cancer Research Support Program (cphcf-2022-231). The Medical Research Developing Funds (KM228009). The Beijing Ronghe Medical Development Foundation (2025HXKT064). The China Health & Medical Development Foundation (2025HXKT068/2024HXK072).

Data availability

All data in the public dataset comes from the TCGA public database (https://portal.gdc.cancer.gov) GEO Public Database (https://www.ncbi.nlm.nih.gov/geo/) All of them are publicly available.The datasets generated and/or analysed during the current study are not publicly available due to the large volume and complexity of raw data files, as well as the need for further internal validation, but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Xi’an Jiaotong University (approval number: 2025-lunshen-068). It retrospectively collected 446 mammographic images of breast cancer patients from the same hospital. Only existing clinical data were used: no patients were contacted, no specimens were taken, and no personal information was disclosed. The risk was judged to be no more than minimal, and the Committee waived the requirement for informed consent without compromising patient rights or welfare. Since all data were de-identified and retrospectively analyzed, informed consent was not required, as confirmed by the Ethics Committee. The data-collection protocol was reviewed and certified to comply fully with the ethical standards of the Declaration of Helsinki. All public databases used are publicly accessible. Informed consent was waived by the Ethics Committee of the Second Affiliated Hospital of Xi'an Jiaotong University due to the retrospective nature of the study and the use of de-identified clinical data. For data obtained from public databases (TCGA and GEO), informed consent was obtained from participants in the original studies, in accordance with their respective institutional review board approvals and data access policies.

Consent for publication

All authors have reviewed and approved the final version of the manuscript and agree to its submission for publication.

Competing interests

The authors declare no competing interests.