A total of 121 BRCA samples and 49 matched normal tissues were collected from the Department of Breast Surgery, Xiangya Hospital, Central South University (Changsha, China). Patients with BRCA were aged 31–76 years, with a median age of 63 years. The samples were collected from female patients with BRCA who were diagnosed according to the postoperative pathological report between January 2016 and December 2018 after surgical resection. Cases of invasive ductal carcinoma, intraductal carcinoma and mucinous adenocarcinoma were included in the present study. Cases with multiple primary cancers and a postoperative positive margin were excluded. The postoperative immunohistochemical (IHC) staining results of the patients for the markers estrogen receptor (ER), progesterone receptor (PR), HER2 and Ki67 were recorded in detail. Clinical and pathological data, including age, histopathological grade, number of axillary lymph node dissection, and metastasis and site of metastasis, were collected (Table I). The patients were all female and the median age was 63 years. The present study was approved by the Ethics Committee of Xiangya Hospital (Changsha, China). Written informed consent was obtained from all patients.

The present study first evaluated the expression levels of STARD4 in 57 BRCA cell lines by analyzing data in the Cancer Cell Line Encyclopedia (CCLE) database (https://portals.broadinstitute.org/ccle) (20,21). The expression levels of STARD4 in T1, T2, T3 and T4 stage BRCA samples were analyzed using the Betastasis database (http://www.betastasis.com). The expression levels of STARD4 in different types of BRCA were analyzed using the Tumor Immune Estimation Resource (TIMER) database (https://cistrome.shinyapps.io/timer/), which is based on The Cancer Genome Atlas RNA-sequence data (https://portal.gdc.cancer.gov/). TIMER provides six major analytic modules that allow users to interactively explore the associations between immune infiltrates and a wide spectrum of factors, including gene expression, clinical outcomes, somatic mutations and somatic copy number alterations (22). The prognostic value of STARD4 was evaluated using Kaplan-Meier Plotter (www.kmplot.com). To analyze the association between STARD4 levels and distant metastasis-free survival (DMFS) time, patient samples were divided into two groups (low and high expression) based on median mRNA levels with a hazard ratio with 95% confidence intervals and log-rank P-values (23). Log-rank P-values <0.05 were considered statistically significant.

Tissue samples were fixed in 10% (v/v) formaldehyde at room temperature for 24 h, embedded in paraffin and cut into 5-µm sections. All IHC staining assays used in the present study were conducted as previously described (24). The biotin-labeled IgG (cat. no. Sp-9001; 1:1,000; OriGene Technologies, Inc.) was used as secondary antibody for 30 min at room temperature. Dehydration was subsequently performed and sample sections were sealed using cover glasses. Five fields were randomly selected and imaged under an optical microscope (BX51T-PHD-J11; Olympus Corporation). The images were analyzed using Image-Pro Plus 6.0 (Media Cybernetics, Inc.). The staining was assessed independently by two pathologists in a blinded manner and all discrepancies were resolved by consensus.

STARD4 immunohistochemical staining was performed using a polyclonal antibody against STARD4 (dilution, 1:200; cat. no. ab202060; Abcam). The staining intensity was divided into the following categories: 0 points, negative (−); 1–2 points, weakly positive (+); 3–4 points, positive (++); 6–9 points, strongly positive (+++). The percentage of positively-stained cells was scored as follows: 0, 0–5%; 1, 5–25%; 2, 25–50%; 3, 50–100%.

ER and PR immunohistochemical staining was performed using a mouse monoclonal anti-human ER antibody (clone, 1D5; Dako; Agilent Technologies, Inc.) and a mouse monoclonal anti-human PR antibody (clone, 636; Dako; Agilent Technologies, Inc.) at a dilution of 1:100. The cut-off value for a positive result was positive staining for ER and PR in ≥1% of tumor cells in 10 selected tumor sub-regions (25). The results were recorded as the percentage of positively-stained nuclei, and the intensity was graded between 0 and 3+ as follows: 0 (negative result), positive staining in <1% of the tumor cells; 1+, mildly distinct, positive staining in ≤25% of the tumor cells; 2+, moderately distinct, positive staining in 25–50% of the tumor cells; and 3+, strong, positive staining in >50% of the tumor cells.

HER2 immunohistochemical staining was performed using the HercepTest™ assay (Dako; Agilent Technologies, Inc.) according to manufacturer's protocols. The expression levels of HER2 were initially determined by immunohistochemistry and graded between 0 and 3+ as follows: 0 (negative result), absence or presence of HER2 in <10% of the tumor cells; 1+ (negative result), membranous, weak and discontinuous staining in >10% of the tumor cells; 2+ (questionable result), membranous, low/moderate and continuous staining in >10% of the tumor cells, or membranous, intense and continuous staining in ≤30% of the tumor cells; and 3+ (positive result), membranous, intense and continuous staining in >30% of the tumor cells. Samples with HER2 scores of 2+ were confirmed to be HER2-negative or HER2-positive using fluorescence in situ hybridization analysis as previously described (26).

Ki67 immunohistochemical staining was performed using a mouse monoclonal anti-human Ki67 antibody (dilution, 1:100; clone, MIB-1; Dako; Agilent Technologies, Inc.). At least three fields in particular staining ‘hot-spots’ were selected in order to represent the spectrum of staining observed upon the initial overview of the entire section. The cancer cells in the three micrographs were manually counted (500–1,000 cells were counted), and the percentage of positively-stained cancer cells was considered to be the Ki67 score (27).

HCC1937 and MDA-MB-231 cells were obtained from The Cell Bank of Type Culture Collection of the Chinese Academy of Sciences and cultured in low-glucose DMEM (Gibco; Thermo Fisher Scientific, Inc) with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin-streptomycin (Gibco; Thermo Fisher Scientific, Inc.) in a 37°C incubator with 5% CO₂. Briefly, recombinant lentiviral vectors were obtained Shanghai Genechem Co., Ltd. A total of 6 µg shRNA plasmids and shRNA control were transfected to knockdown target expression levels using Lipofectamine 3000 reagent (Invitrogen; Thermo Fisher Scientific, Inc.), according to standard molecular techniques. At 48 h after transfection, the transfection efficiency was determined using reverse transcription-quantitative (RT-q)PCR and western blotting. The short hairpin RNA (sh)STARD4 sequence was 5′-CCGGTCCTATACTGTGGGCTATAAACTCGAGTTTATAGCCCACAGTATAGGATTTTTG-3′. The sh-cAMP responsive element binding protein 1 (CREB1) sequence was 5′-CCGGTCCCAACAAATGACAGTTCAACTCGAGTTGAACTGTCATTTGTTGGGATTTTTG-3′. The shcontrol sequence was 5′-CCGGTCTTCTCCGAACGTGTCACGTCTCGAGACGTGACACGTTCGGAGAAGATTTTTG-3′. Lentivirus construction and infection were performed as previously described (28).

For the cell proliferation assay, 2,000 HCC1937 or MDA-MB-231 cells/well were seeded into 96-well plates. The optical density was detected at 450 nm using a microplate reader after adding 10 µl Cell Counting Kit-8 (CCK-8; Dojindo Molecular Technologies, Inc.) reagent to each well at each time point (day 1, 2, 3, 4 and 5) according to the manufacturer's protocol.

The cell counting assay was performed using the Celigo image cytometer (Nexcelom Bioscience LLC). For cell cytometry system Celigo^® analysis, HCC1937 and MDA-MB-231 cells transfected with shSTARD4 or negative control were seeded into 96-well plates (1,500 cells/well) and cultured for 5 days. The cell images were captured and analyzed using the using a fluorescence microscope system (Celigo^®; Nexcelom Bioscience LLC). Each experiment was performed three times.

The apoptosis of HCC1937 and MDA-MB-231 cells was analyzed using a FACS Calibur flow cytometer (BD Biosciences) with a FITC Annexin V Apoptosis Detection kit (BD Biosciences. Cell apoptosis data were analyzed using FCS Express software (version 3.0; De Novo Software). Each experiment was performed three times.

The cell migration assay was performed as previously described using a Transwell assay (29). For the wound healing assay, HCC1937 and MDA-MB-231 cells were seeded into 6-well plates in duplicate. Once the confluency of the cells reached ~80%, the monolayer of cells was wounded using a 10-µl plastic pipette tip and then the scratched cells were rinsed with PBS and cultured with serum-free DMEM for 24 h. The wound was imaged at 0, 12 and 24 h using a fluorescence microscope (CK40-F200; Olympus Corporation) and then analyzed using ImageJ software (version 1.62; National Institutes of Health). The following formula was used: Wound closure rate (%)=migrated cell surface area/total surface area ×100. Each experiment was performed three times.

Total RNA was extracted from BRCA cell lines using the TRIzol reagent (Thermo Fisher Scientific, Inc.) and was transcribed to cDNA using the RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific, Inc.). Reverse transcription was conducted at 37°C for 1 h, followed by 85°C for 5 sec according to the manufacturer's protocol. RT-qPCR was performed using the QuantiTect SYBR-Green PCR kit (Roche Diagnostics) as previously described (30). The primer sequences are listed in Table SI. The 2^−ΔΔCq method was used to calculate the relative expression levels of the target genes (31). GAPDH was used for normalization. The reaction conditions were as follows: 95°C for 10 min, followed by 95°C for 30 sec, 60°C for 30 sec and 72°C for 30 sec (39 cycles), and finally 72°C for 10 min.

A total of 3×10⁶ STARD4-knockdown or control HCC1937 cells were suspended in PBS and injected subcutaneously into the dorsal skin of 6-week-old female BALB/c nude mice (10 mice/group) with a median weight of 20 g (Shanghai SLAC Laboratory Animal Co., Ltd.). The protocol for the animal experiments was approved by the Shanghai Medical Experimental Animal Care Commission (approval no. ShCI-14-008). Mice were housed under specific pathogen-free conditions at 25°C with a 12-h light/dark cycle, and free access to food and water. The maintenance conditions for the mice were as follows: Humidity, 40–60%; ventilation, 15 times/h. STARD4 knockdown and control luciferase-expressing HCC1937 cells were injected into the mice and monitored using an in vivo Imaging System (IVIS; PerkinElmer, Inc.). The animals were sacrificed when the tumors reached 1 cm in diameter. The mice were sacrificed 7 weeks after injection using CO₂ (with a flow rate of CO₂ for euthanasia displacing ≤30% of the chamber volume/min). The volume and weight of all nude mice were measured after 4 weeks. Tumor growth was calculated using the following formula: Volume = 0.5 × length × width².

Total protein was extracted from cells with RIPA buffer (Beyotime Institute of Biotechnology). The BCA assay kit (Boster Biological Technology) was used to measure the total protein concentration. Protein (30 µg/lane) was subsequently separated by 12% SDS-PAGE, followed by transfer to PVDF membranes. Subsequently, the membranes were blocked with 3% BSA (Sangon Biotech Co., Ltd.)/TBS with 0.05% Tween-20 (TBST; Sangon Biotech Co., Ltd.) for 2 h at room temperature and incubated with primary antibodies at 4°C overnight. The PVDF membranes were rinsed three times for 5 min each with TBST and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. The antibodies are listed in Table SII. The total protein levels were assessed using enhanced chemiluminescence reagents (Thermo Fisher Scientific, Inc.). ImageJ 6.0 software (National Institutes of Health) was used to analyze the images.

Total RNA was extracted using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.) and quantified by the NanoDrop ND-2000 (Thermo Fisher Scientific, Inc.). The RNA integrity was assessed using the Agilent Bioanalyzer 2100 (Agilent Technologies, Inc). Whole-genome gene expression levels were examined using the Gene Chip Prime View Human Gene Expression Array (Thermo Fisher Scientific, Inc.). Sample labeling, microarray hybridization, washing and gene normalization were performed according to the manufacturer's protocols. RNA integrity was also examined using an Agilent Bioanalyzer 2100 (cat. no. G2938A; Agilent Technologies, Inc.). To obtain biotin-tagged cDNA, total RNA was subsequently amplified, labeled and purified using a WT PLUS Reagent kit (cat. no. 902280; Affymetrix; Thermo Fisher Scientific, Inc.). Array hybridization was performed using an Affymetrix GeneChip Human Gene 2.0 ST Array (Affymetrix; Thermo Fisher Scientific, Inc.) and Hybridization Oven 645 (cat. no. 00-0331-220V; Affymetrix; Thermo Fisher Scientific, Inc.). The Gene Chip was subsequently washed using a Hybridization, Wash and Stain kit (cat. no. 900720; Affymetrix; Thermo Fisher Scientific, Inc.) in a Fluidics Station 450 (cat. no. 00-0079, Affymetrix; Thermo Fisher Scientific, Inc.). A GeneChip Scanner 3000 (cat. no. 00-00213; Affymetrix; Thermo Fisher Scientific, Inc.) was used to scan the results, which were analyzed by Command Console Software 4.0 (Affymetrix; Thermo Fisher Scientific, Inc.) to summarize probe cell intensity data, namely, the CEL files with default settings. Following this, CEL files were normalized according to gene and exon levels using Expression Console Software 4.0 (Affymetrix; Thermo Fisher Scientific, Inc.). All procedures, including array hybridization and scanning, were independently performed according to a standard protocol (32) for microarray experiments (n=3).

Differentially expressed mRNAs were subsequently identified using fold change values. The threshold for upregulated and downregulated genes was ≥1.5. Genes that were differentially expressed were identified based on the following criteria: P<0.05 and absolute fold change >1.5.

All statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software, Inc.). All data are presented as the mean ± SEM. All assays were conducted at least three times. The unpaired Student's t-test, Mann-Whitney U test and Wilcoxon matched-pairs signed rank test were used to analyze the differences between two groups according to the test condition. One-way ANOVA with Tukey's post hoc test was used to analyze the differences among multiple groups. Survival curves were estimated using the Kaplan-Meier method and the log-rank test. P<0.05 was considered to indicate a statistically significant difference.

In order to evaluate the expression levels of STARD4 in BRCA, 121 BRCA samples and 49 adjacent normal tissue samples were collected from patients who underwent surgical resection. The clinicopathological characteristics and the expression levels of STARD4 in BRCA are shown in Table I. STARD4 was expressed in the nuclear regions in both BRCA and normal breast tissues (Fig. 1A-D). The comparison of the STARD4 levels between BRCA and normal samples demonstrated that STARD4 protein levels were significantly upregulated in 121 BRCA samples compared with in 49 adjacent normal tissues (Fig. 1E). The BRCA levels were upregulated in 49 paired BRCA samples (Fig. 1F).

The associations between STARD4 protein expression and the clinicopathological features of patients with BRCA are summarized in Table II. Positive protein expression of STARD4 was associated with advanced histological stage, metastatic status and ER expression. However, it was not observed to be significantly associated with age, T stage, M stage, N stage, PR and HER-2 expression. These results demonstrated that STARD4 was significantly associated with ER expression, but not PR expression. When analyzing the association between ER and STARD4 using the TIMER database, a negative correlation was observed between the two genes in all types of BRCA (Fig. S1A), HER2-positive BRCA (Fig. S1B) and luminal BRCA (Fig. S1C). However, STARD4 expression was positively correlated with ER expression in basal BRCA (Fig. S1D).

Furthermore, online public datasets were analyzed. When analyzing the Betastasis database, the data indicated that STARD4 expression was upregulated in T1, T2, T3 and T4 stage BRCA samples compared with in normal breast samples (Fig. 1G). However, STARD4 expression was not significantly different among the different stages. Furthermore, the TCGA dataset was analyzed to detect STARD4 levels in different types of BRCA. STARD4 expression was upregulated in basal and Her2-positive BRCA tissues, whereas it was not upregulated in luminal BRCA samples compared with in normal breast tissues (Fig. 1H). The analysis demonstrated that STARD4 expression may be associated with BRCA progression and may serve as an early diagnostic biomarker. However, the dysregulation of STARD4 could not be used for the prediction of BRCA stage.

The association between STARD4 levels and DMFS time was determined using the Kaplan-Meier survival Plotter database. The median expression of STARD4 in all BRCA cases was selected as a cut-off to divide the BRCA samples into STARD4-high and -low groups. Kaplan-Meier survival curve analysis indicated that higher expression levels of STARD4 were significantly associated with shorter DMFS time in patients with BRCA (Fig. 2A).

Furthermore, the association between STARD4 expression and DMFS time was examined in patients with BRCA according to their clinicopathological characteristics, including ER, PR and HER2 status. The results indicated that high STARD4 expression was associated with poor DMFS rates compared with low STARD4 expression in patients with ER-negative (Fig. 2B), PR-negative (Fig. 2C) and HER2-positive (Fig. 2D) BRCA. However, dysregulation of STARD4 was not significantly associated with DMFS time in patients with ER-positive (Fig. 2E), PR-positive (Fig. 2F) and HER2-negative (Fig. 2G) BRCA. Furthermore, the present study evaluated the prognostic utility of STARD4 based on subtype (luminal and HER2). As presented in Fig. S2, although a potential tendency that higher expression levels of STARD4 were associated with shorter DMFS time in luminal A, luminal B and HER2-positive BRCA was observed, no significant association between STARD4 expression and DMFS time was observed in groups of patients based on subtype (luminal and HER2).

Upregulation of STARD4 in BRCA suggested that STARD4 may serve as an oncogene in BRCA. Therefore, loss-of-functions assays were used to detect the effects of STARD4 on cell proliferation. The present study first evaluated the expression levels of STARD4 in 57 BRCA cell lines by analyzing the CCLE database (20,21). The results demonstrated that STARD4 was highly expressed in multiple BRCA cell lines, including MDA-MB-231 and HCC1937 cells, which do not have the highest STARD4 levels; however, they are widely used in triple-negative BRCA research (Fig. 3A). Following transfection of HCC1937 (Fig. 3B and C) and MDA-MB-231 (Fig. 3D and E) cells with shSTARD4 lentivirus, the mRNA and protein levels of STARD4 in BRCA cells were reduced compared with those of the control group.

Subsequently, the effects of STARD4 on BRCA proliferation were assessed using a Celigo cell counting assay for 5 days. Knockdown of STARD4 reduced cell proliferation in HCC1937 (Fig. 3F) and MDA-MB-231 (Fig. 3H) cells compared with in control cells. In addition, a CCK-8 assay revealed that knockdown of STARD4 suppressed the HCC1937 (Fig. 3G) and MDA-MB-231 (Fig. 3I) cell proliferation rate at day 3, 4 and 5.

The effect of STARD4 on the cell cycle progression of BRCA cells was investigated. Flow cytometry demonstrated that knockdown of STARD4 expression in BRCA cells inhibited the cell cycle progression at the G₀/G₁ phase (Fig. 4A-D). The proportion of cells arrested at the G₁ phase was significantly higher in shSTARD4-transfected cells compared with that noted in the control cells. However, the percentage of cells arrested at the S phase was significantly lower in the shSTARD4-transfected cells compared with that in the control cells.

The effects of STARD4 on cell apoptosis were also investigated using flow cytometry. STARD4 knockdown induced apoptosis in BRCA cells compared with the control group. The apoptotic rates of STARD4-silenced HCC1937 and MDA-MB-231 cells were higher than those of the corresponding control groups (Fig. 4E-H). These results indicated that STARD4 may promote cell proliferation by reducing cell apoptosis.

The ability of STARD4 to affect cell migration was examined using Transwell and wound healing assays. The results of the Transwell assay demonstrated that the migratory activity of HCC1937 and MDA-MB-231 cells was significantly attenuated following transfection with STARD4 shRNA compared with that of the negative control cells (by 59 and 99%, respectively; Fig. 5A-D). Similar results were observed in the wound healing assay (Fig. 5E-H).

The effects of STARD4 knockdown on tumor growth were analyzed in vivo using a nude mouse model. The luciferase-expressing MDA-MB-231 cells were constructed to implant tumors in live animals. Using an IVIS system, the data indicated that luciferase signaling in the STARD4 knockdown group was decreased compared with that in the control group (Fig. 6A and B). It was revealed that the tumor volume in the STARD4 knockdown group was lower than that in the negative control group (Fig. 6C and D).

To understand the mechanisms of STARD4 in regulating BRCA progression, a gene expression microarray was used to identify targets of STARD4 in MDA-MB-231 cells. A total of 772 genes, including 350 upregulated and 422 downregulated genes, were identified as downstream targets of STARD4 in BRCA (Fig. 7A; Table SIII).

In order to validate the downstream targets of STARD4 identified by microarray analysis, the expression levels of the potential targets were detected following silencing of STARD4. A total of 10 upregulated (RPS15A, PLA2G12A, CDKN1B, ITGA5, FKBP1A, PIK3R1, HDAC6, RPSA, CDK6 and RAP1A; Fig. 7B) and 21 downregulated genes (RPS2, JMJD7.PLA2G4B, CDK1, EIF4A2, ELF1, CCNE2, RHOJ, HMOX1, PRKCE, CREB1, PLD6, PAK2, GRB2, PRKAG2, RAPGEF1, DDIT4, CCNE1, H3F3B, PRKCH, CDK2 and RHEB; Fig. 7C) were selected for further validation. The results indicated that knockdown of STARD4 increased the expression levels of CDKN1B and FKBP1A (Fig. 7D), but decreased the expression levels of PRKCH, RAPGEF1, p21 (RAC1) activated kinase 2 (PAK2), CDK2, CREB1, E74 like ETS transcription factor 1 (ELF1), PLD6, CCNE1, CCNE2, PRKAG2 and HMOX1 (Fig. 7E). Furthermore, western blot analysis indicated that the protein expression levels of ELF1, CREB1 and PAK2 were reduced in the STARD4 knockdown group compared with those in the control group (Fig. 7F). However, the protein levels of PKC epsilon, and Histone H3.3 were not changed following STARD4 knockdown (Fig. 7F).

The present study focused on CREB1, which showed the most significant decreased expression following STARD4 knockdown. When analyzing the correlation between CREB1 and STARD4 using the TIMER database, a significant positive correlation between the two genes was observed in all types of BRCA (Fig. 8A), HER2 positive BRCA (Fig. 8B), luminal BRCA (Fig. 8C) and basal BRCA (Fig. 8D).

Loss-of-functions assays were used to detect the effects of CREB1 on cell proliferation and migration. The infection efficiency and knockdown efficiency are shown in Fig. 8E and F. As shown in Fig. 8G, knockdown of CREB1 significantly reduced cell proliferation in MDA-MB-231 cells compared with in control cells at day 3, 4 and 5. The results of the Transwell assay demonstrated that the migratory activity of MDA-MB-231 cells transfected with CREB1 shRNA was significantly attenuated compared with that of the negative control cells (Fig. 8H and I).