Breast cancer proteomics datasets (TCGA Cancer Proteome Study of Breast Tissue) and clinical data were downloaded from the CPTAC website (https://proteomics.cancer.gov/programs/cptac). A total of 105 TCGA breast cancer samples were analyzed using iTRAQ (isobaric tags for relative and absolute quantification) protein quantification methods. The samples included four breast cancer subtypes (luminal A, luminal B, basal-like, and HER2-enriched). After excluding patients with incomplete clinical pathological data, this study enrolled 101 patients for subsequent analysis. The scRNA-seq (single-cell RNA-seq) data (1,534 cells in six fresh triple-negative breast cancer tumors) of GSE118389 (10) were also downloaded from GEO (Gene Expression Omnibus) (https://www.ncbi.nlm.nih.gov/geo/). Gene expression and clinical data of 1,904 breast cancer patients from the METABRIC (Molecular Taxonomy of Breast Cancer International Consortium) breast studies were downloaded from the cBioPortal (https://www.cbioportal.org/).

The batch correction of the CPTAC datasets was performed using ‘Limma’ package (version 3.44.3) and ‘impute’ package version 1.62.0 of R language (version 4.0.0). The R language was used to analyze differentially expressed proteins in the non-TNBC (n=77) and TNBC (n=24) groups. Differentially expressed protein cutoff criteria were a FDR (false discovery rate) corrected p-value<0.01, and logFC > 0.

The ‘clusterProfiler’ package (version 3.16.0), ‘org.Hs.eg.db’ package (version 3.11.4), ‘enrichplot’ package (version 1.8.1), ‘digest’ package (version 0.6.25), ‘GOplot’ package (version 1.0.2) and ‘ggplot2’ package (version 3.3.1) of R language were used to perform GO analysis to expound the BPs (biological processes), MFs (molecular functions), CCs (cellular components) and KEGG analysis.

The correlation between OS (overall survival) and candidate genes or proteins was identified through Kaplan-Meier survival analysis using ‘survminer’ package of R language. A total of 101 breast cancer patients and clinical data from CPTAC, and 1904 breast cancer patients and clinical data from METABRIC, were selected for survival analysis.

The ‘survivalROC’ package (version 1.0.3) of R language was used to calculate the ROC (receiver operating characteristic) curve of candidate proteins, the selecting criteria being AUC > 0.7. The candidate genes were predicted using the ROC curve of TNBC with the ‘pROC’ package (version 1.16.2) of R language.

To identify the cells that significantly express OBFC2A, scRNA-seq data of 1,534 cells in six fresh triple-negative breast cancer tumors from GSE118389 were analyzed using the Serat package and the SingleR package in the R language. The specific steps of the scRNA-seq data analysis were carried out as described earlier (11).

The Pearson correlation coefficient was employed to evaluate the correlation between OBFC2A mRNA expression levels and the level of immune cell infiltration. TIMER gene modules were utilized to investigate OBFC2A expression in diverse tumors and the relationship between OBFC2A expression and the abundance of immune infiltrates. The infiltration data of B cells, CD4 + T cells, CD8 + T cells, dendritic cells, macrophages, and neutrophils can be downloaded from the TIMER database (https://cistrome.shinyapps.io/timer/).

The breast cancer cell lines MCF10A, BT549, MDA-MB-231, MCF7, SK-BR-3, YCCB1, ZR75-1, and T47D were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). MCF10A culture was performed as described by Lomoriello et al. (12), and the rest of the cell lines were cultured in 90% RPMI-1640 (Gibco, USA) + 10% fetal bovine serum (Biological Industries, Israel) at 37°C in an atmosphere of 5% CO₂.

All cells were lysed in RIPA buffer to isolate total proteins. After extracting proteins and measuring the concentration with the BCA protein kit (Beyotime, China), equal amounts of denatured proteins (40 µg) were used for western blotting as previously described (13). The same amount of denatured protein (40 μg) and 10% SDS-PAGE were used for electrophoresis and then transferred to the PVDF membrane. After blocking in 5% non-fat milk for 1h at RT (room temperature), the membrane was incubated with the primary antibody overnight at 4˚C, and then incubated with the second antibody (anti-rabbit IgG or anti-mouse IgG) at RT for 1 h. The Fusion FX7 Spectra multifunction imaging system was used to detect bands. The following primary antibodies were used: OBFC2A (16719-1-AP, Proteintech Group, USA) and GAPDH (GB11002, Servicebio, Hubei, China).

The shRNA targeting OBFC2A (sh-OBFC2A) was constructed by TSINGKEBIO (Tsingkebio Co., Ltd, Beijing, China). The lentivirus shRNA sequence was as follows: 5-GATCGTGCAAAGTAGCAGATA-3′. The second generation lentivirus packaging system was used for generate lentivirus. According to the instructions, Lipofectamine 2000 was used to transfect the lentiviral vector (20 μg, PxpAx2: PMD2g: pLVX = 3: 1: 4) into HEK293T cells at 37°C for 48h, and the lentivirus was generated by transfecting the packaging plasmid with calcium phosphate. When MDA-MB-231 cell density reached 30-50%, the cells were infected with lentivirus at 37°C according to the protocol provided by the manufacturer. At 48 h after transfection, the stable cell lines were selected by using puromycin for 48-72 h. Stable cell lines were selected using 2 µg/ml puromycin. Stable cell lines were used in subsequent cell biology experiments.

The colony formation assay was performed as previously described (14). The cells of each group were inoculated into 12-well plates with 1000 cells in each well and three multiple holes in each group, which were cultured in an incubator for one week. Cells were then fixed with 4% paraformaldehyde for 15 min and stained with crystal violet (C0121, Beyotime) for 15 min.

Wound healing assays were performed as described previously (15). The cells were inoculated into 6-well plates to form a confluent monolayer at 37°C for 24 h. One wound per well was conducted with a 10-ul tip and washed twice with PBS. The cells were then replaced with serum-free medium, and live cell imaging was performed at 0 and 24 h.

Cells were washed twice with PBS and fixed in 70% cold ethanol at 4°C for 30 min. The cells were rewashed with PBS and incubated with 100 μl RNase A (0.1 mg/ml) and 2 μl PI (propidium iodide) for 30 min at 37°C in the dark. Cell cycle distribution was analyzed via flow cytometry (BD CellQuest Pro, version 5.1) using a FACSCalibur (BD Biosciences).

Approximately 1 × 10⁶ cells in each group were washed with PBS and suspended in 100 μL of binding buffer. After incubation with Annexin V-FITC (20 μg/ml) and PI (50 μg/ml) (both Sigma-Aldrich; Merck KGaA) at RT for 30 min, apoptosis was detected by FACSCalibur flow cytometry (BD Biosciences). Total cell apoptosis was calculated as the percentage of early + late apoptotic cells.

The analysis was performed using R language and SPSS software (version 19.0). Student’s t-tests were used to conduct differential comparisons of two groups. Kruskal-Wallis’s test was used to compare multiple independent samples. P value <0.05 was considered to be statistically significant.

A total of 1565 differentially expressed proteins were identified from CPTAC ( Table S1 ). Figure S1 showed the PPI (protein-protein interaction network) of differentially expressed genes. Finally, six candidate proteins (DLAT, ELAVL2, KARS, OBFC2A, RAVER2, SSBP1) were selected by survival analysis filtering from CPTAC proteomics datasets ( Figures S2 , S3 ). Only OBFC2A was validated in METABRIC dataset. As shown in Figure 1B , OBFC2A mRNA expression was higher in grade III (p < 0.05). To explore the expression distribution of OBFC2A, the expression of OBFC2A in different subtypes was performed. The results showed that the expression levels of OBFC2A were higher in the TNBC group than in the other subtypes, in both CPTAC and METABRIC ( Figures 1A, C ). To further confirm this finding, the ROC was evaluated for OBFC2A expression and TNBC subtype of all breast cancers. The results showed that AUC (the areas under the curve) was 81.5% (TNBC) in the METABRIC ( Figure 1J ).

To explore the relationship between OBFC2A expression and overall survival, the overall survival of each group was analyzed. The results showed that the OS of patients with higher OBFC2A expression was shorter than that of patients with lower OBFC2A expression in CPTAC and METABRIC ( Figures 1D–F ; p < 0.05). To further confirm this finding, the ROC was evaluated for OBFC2A expression and OS of all grade breast cancers. The results showed that the AUC were 79% (three years) and 95.9% (five years) in the CPTAC ( Figure 1I ). To explore OBFC2A was an independent prognostic factor, Univariate ( Figure 1G ) and Multivariate ( Figure 1H ) Cox regression analyses were performed in CPTAC and METABRIC. Multivariate Cox regression analysis showed that OBFC2A was independently associated with OS in CPTAC and METABRIC, respectively (HR = 5.309, 1.200-23.493, P=0.028; HR = 3.658, 1.881-7.114, p < 0.0001) ( Figure 1H and Table 1 ).

To explore the biological function associated with OBFC2A expression in breast cancer, Pearson correlation analysis between OBFC2A expression and other genes in whole-genome profiling of 1904 patients in METABRIC was performed. The results showed that 236 genes were positively correlated with OBFC2A expression (R > 0.5) ( Figure 2A ). GO and KEGG pathway enrichment analyses of the above genes were performed. The results were listed in Figures 2B, C , wherein 3 MFs, 3 BPs, and 3 CCs for 236 genes were observed including immune receptor activity, MHC protein complex binding, MHC class II receptor; T cell activation, regulation of T cell activation, positive regulation of cell activation; external side of plasma membrane, MHC protein complex, MHC class II protein complex. As for KEGG analysis, natural killer cell mediated cytotoxicity, cell adhesion molecules (CAMs), and hematopoietic cell lineage were included. In the t-SNE analysis, 14 cell clusters were found, and the scatter plot showed that OBFC2A was not evenly expressed among the cell clusters ( Figures 2D, E ). Moreover, the scatter diagram and bubble diagram showed that OBFC2A and its related genes were specifically expressed in macrophages ( Figure 2F ). Since OBFC2A was possibly related to tumor immunity, TIMER database was used to analyze the correlation between OBFC2A and immune cell subtype infiltration. As shown in Figure 2G , the correlation coefficients between macrophages, B cells, CD8 + T cells, CD4 + T cells, DCs, and neutrophils, and OBFC2A were 0.329, 0.384, 0.431, 0.516, 0.592, and 0.614, respectively (P < 0.001).

It was further confirmed that OBFC2A was highly expressed in TNBC cell lines (BT549, MDA-MB-231, and YCCB1) by verifying the protein expression in all breast cancer cell lines ( Figures 3A, C , P<0.01). To determine the role of OBFC2A in the development of breast cancer, the function of OBFC2A in breast cancer cells was investigated. A lentiviral transfection system was used to knockdown the protein expression of OBFC2A in the MDA-MB-231 cell line ( Figures 3B, D ). Clone formation assays and wound healing assays showed that knockdown of OBFC2A inhibited the proliferation and migration of TNBC cells ( Figures 3E–H ). The results of flow cytometry showed that silencing OBFC2A promoted apoptosis and blocked cell cycle in G1 phase ( Figures 3I–L ).