The ER-positive breast cancer MCF-7 cell line, TNBC MDA-MB-231 and BT549 cell lines, and a normal human mammary epithelial MCF-10A cell line were purchased from the American Type Culture Collection. All cell lines were cultured in a cell incubator (model no. thermo3111; Thermo Fisher Scientific, Inc.) at 37°C with 5% CO₂. The MDA-MB-231 and MCF-7 cells were cultured in DMEM (cat. no. L110KJ; BasalMedia, Inc.) supplemented with 10% fetal bovine serum (FBS; cat. no. 16000-044; Gibco; Thermo Fisher Scientific, Inc.) BT549 cells were cultured in DMEM supplemented with 10% FBS and 0.023 U/ml insulin. MCF-10A cells were cultured in DMEM/F12 (cat. no. L310KJ; BasalMedia, Inc.) supplemented with 5% donor horse serum (cat. no. 26050088; Gibco; Thermo Fisher Scientific, Inc.) and certain additives (20 ng/ml EGF, 10 µg/ml insulin and 10 ng/ml cholera toxin). Belinostat (HDAC6 inhibitor) and 17-AAG (HSP90 inhibitor) were purchased from Target molecule Corp. and prepared in DMSO as 50 mM stock solutions. The solutions were diluted in cell culture medium immediately prior to use.

Four different cell lines (MDA-MB-231, BT549, MCF-7 and MCF-10A) were seeded in 96-well flat bottom plates at a density of 3,000 cells/well and treated with 200 µM to 0.78 nM (double gradient dilution) 17-AAG or Belinostat at 37°C in triplicate 12 h later. The viability of cells was measured using the CellTiter-Glo Luminescent Cell Viability assay (Promega Corporation) at 72 h after treatment. The optical density of each well was recorded using a microplate reader (PerkinElmer Envision multimode plate reader; PerkinElmer, Inc.). A range of drug concentrations was added at their fixed ratio based on their respective individual IC₅₀ values at 72 h (51). Using the CompuSyn software (52) (CompuSyn v1.0; ComboSyn, Inc.), combination indexes (CIs) were calculated according to the cell viability at corresponding concentrations. Briefly, the CI value indicates the interaction of co-administrated compounds: CI<1 indicates synergism effect, CI=1 indicates additive effect and CI>1 indicates antagonism effect (53).

For the apoptosis analysis, cells were seeded in 6-well flat bottom plates at a density of 1.5×10⁵ cells/well and treated with single 17-AAG (0.5 and 1.0 µM), single Belinostat (0.1 and 0.2 µM), or the combination of 17-AAG and Belinostat at 37°C for 12 h. After incubation for 72 h, cells were harvested in iced PBS buffer. Apoptosis was evaluated using the Annexin V-FITC Apoptosis Detection kit (Vazyme Biotech Co., Ltd.). Briefly, collected cells were re-suspended in 100 µl binding buffer and incubated with 5 µl propidium iodide (PI) and 5 µl Annexin V-FITC for 15 min in the dark at room temperature. Before analyzing apoptosis percentage using the FACSCalibur flow cytometry system (BD Biosciences), 400 µl binding buffer was added to the samples to stop the staining procedure. The apoptosis data were analyzed using FlowJo™ software (FlowJo 7.6.1; BD Biosciences), with ≥10,000 cells for each sample. For the cell cycle analysis by PI staining, cells were seeded and harvested as for the apoptosis assay. Each sample was fixed and re-suspended in 1 ml iced 70% ethanol at −20°C overnight. After washing twice with iced PBS buffer, samples were stained with PI in the dark at room temperature for 30 min. The PI stain was included in the Cell Cycle and Apoptosis Analysis kit (Yeasen). Subsequently, cell cycle analysis was performed using FACScan flow cytometry (FACS Canto II; BD Biosciences) by red fluorescence at an excitation wavelength of 488 nm, and the data were analyzed using ModFit software (ModFit LT v3.3; BD Biosciences).

Total RNA was extracted from MDA-MB-231 cells using the RNA Isolation kit (Beyotime Institute of Biotechnology). Each sample of 300 ng total RNA was reverse transcribed to cDNA using the HiScript II Q RT SuperMix for qPCR kit (Vazyme Biotech Co., Ltd.). The reverse transcription temperature protocol was as follows: 50°C for 15 min, followed by 80°C for 5 sec. The qPCR procedures were performed on the ViiA 7 Real-Time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.) using the ChamQ SYBR qPCR Master Mix (Low ROX Premixed) kit (Vazyme Biotech Co., Ltd.). The thermocycling program for the RT-qPCR reactions was as follow: Pre-denaturation at 95°C for 10 min, 40 cycles of denaturation at 95°C for 15 sec and annealing at 60°C for 30 sec, followed by extension at 72°C for 1 min. Subsequently, the expression values of mRNA were calculated using the 2^−ΔΔCq method (54), with GAPDH as a reference for normalization, and represented as fold change. Primers were synthesized by Sangon Biotech Co., Ltd. and are listed in Table I.

Cells were seeded in 6-well flat plates and split in 1X SDS sample loading buffer (250 mM Tris HCl pH 6.8, 10% SDS, 30% glycerol, 5% β-mercaptoethanol and 0.02% bromophenol blue). Protein concentrations were determined using the BCA Protein assay (Thermo Fisher Scientific, Inc.). Samples (25 µg/lane) were resolved by 6, 10 or 12% SDS-PAGE and transferred to 0.22-µm nitrocellulose membranes. Following blocking with 5% milk at room temperature for 1 h, the nitrocellulose membranes were washed four times for 15 min with TBS with 0.1% Tween-20 (TBST) buffer and incubated with primary antibodies (dilution, 1:1,000) at 4°C overnight. Subsequently, the membranes were washed four times with TBST buffer and incubated with horseradish peroxidase (HRP)-conjugated Goat Anti-Mouse lgG (cat. no. D110087; Sangon Biotech Co., Ltd.) and HRP-conjugated Goat Anti-Rabbit lgG (cat. no. D110058; Sangon Biotech Co., Ltd.) secondary antibodies (dilution, 1:10,000; BBI Life Sciences Corporation) for 1 h at room temperature. The bands were visualized using an enhanced chemiluminescence kit (Thermo Fisher Scientific, Inc.) with a ChemiScope 3400 mini imaging system (Clinx Science Instrument Co., Ltd.). Densitometry was performed for each group using ImageJ 1.50b software (National Institutes of Health). All primary antibodies used are listed in Table II.

To characterize the genomic impact of single and combined compound treatment, RNA-seq data was collected using 1.5×10⁵ MDA-MB-231 cells treated with DMSO as a control, 17-AAG (1.0 µM), Belinostat (0.2 µM) and the combination of 17-AAG and Belinostat at 37°C for 72 h. After 24 h, total RNA was isolated and purified using DNaseI (Takara Bio, Inc.) and Dynabeads Oligo (dT) 25 (Thermo Fisher Scientific, Inc.). Subsequently, purified RNA (100 ng) was used for cDNA library construction, using the NEBNext Ultra™ RNA Library Prep kit for Illumina (cat. no. E7530L; New England BioLabs, Inc.). Sequencing data was collected on an Illumina HiSeq 2500 instrument. Subsequently, paired-end reads were processed using the Tophat2 v2.1.1 software package (55), with the GRCh38/hg18 Ensembl transcript set as a reference. Following transcript assembly using the Cufflinks v2.2.1 software package (56), differentially expressed genes (|log₂fold-change| >0.5 and P<0.05) were identified using Cuffdiff 2 (57). From the list of differentially expressed genes, each gene was checked in the Gene database of the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/gene/). According to the relevant abstract description of genes in NCBI, only the most migration-related or metastasis-related genes were selected for the final heatmap. The Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.kegg.jp/) was used for pathway analysis. Using the GeneAnswers v3.0 package (58) of the Bioconductor project, the P-values of involved KEGG pathways were calculated with a threshold value of 0.1, based on all differentially expressed genes. Finally, the heatmap and pathways histogram were plotted using the ggplot2 v2.1.0 package in R (http://www.rdocumentation.org/packages/gglot2/versions/2.1.0). The raw sequencing data and processed expression tables have been deposited to the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/; accession no. GSE129944).

MDA-MB-231 cells were seeded in 6-well plates at a density of 7.5×10⁴ cells/ml and permitted to adhere overnight. After cells were treated with 0.2 µM Belinostat, 1 µM 17-AAG or the combination at 37°C for 3 days, cells were stained with 1 µM DCFH-DA for 30 min in the dark at 37°C. Subsequently, cells were washed three times with PBS and subjected to FACScan flow cytometry (FACS Canto II; BD Biosciences). Green fluorescence FITC (FL1) was detected at 488 nm (excitation wavelength) and 530 nm (emission wavelength). FlowJo™ software (FlowJo 7.6.1; BD Biosciences) was used to calculate and analyze the data. Fluorescence microscopy (magnification, ×40) was used to capture images with fixed exposure time.

For the wound healing assay, coordinates were marked on the 35-mm dish, and MDA-MB-231 cells were inoculated at a density of 1.5×10⁶ cells/well. When the confluence was ~100%, cells were rinsed twice with PBS, and serum-free DMEM was added for cell starvation. After cells were starved for 24 h, three marks were scratched on the dish covered with cells with a 200-µl pipette tip. Subsequently, the dish was washed twice with PBS to remove cell debris, and serum-free DMEM containing 0.2 µM Belinostat, 1 µM 17-AAG or the combination of both was added after scratching. Images were captured under an inverted light microscope (magnification, ×40) at different time points. The cell scratch area at each coordinate point was recorded. The relative migration area was calculated, and average values were taken for comparison. The area of cells was calculated for each group using ImageJ 1.50b software (National Institutes of Health) The wound-closing procedure was observed for 36 h, and images were captured at 0 and 36 h, respectively.

For the cell migration assay, serum-free medium was added to the Transwell (cat. no. 353097; Corning Inc.) and outer chamber, and the membrane of the chamber was hydrated 1 h later. A total of 1.5×10⁵ MDA-MB-231 cells were plated into 35-mm dishes containing serum-free DMEM. Following starvation at 37°C for 12–24 h to further remove the effect of serum, cells were digested, centrifuged at 300 × g at room temperature for 5 min and the media were discarded. Subsequently, cells were washed once or twice with PBS, and re-suspended with the serum-free culture medium containing 0.1% BSA (Sigma-Aldrich; Merck KGaA) The density of cells was adjusted to 1×10⁶ cells/ml. A total of 1.0×10⁵ MDA-MB-231 cells were added to the upper Transwell chamber containing 0.2 µM Belinostat, 1 µM 17-AAG or combination of both, and 600 µl DMEM containing 10% FBS was added to the lower chamber of the 24-well plate. Placed on a 24-well plate, the surface of the Transwell compartment membrane was carefully observed for bubble formation. Following incubation at 37°C for 24 h, Transwell cells were washed twice with PBS, and a cotton swab was used to remove cells from the upper chamber. Subsequently, cells were fixed with 90% ethanol at room temperature for 15 min, and the membrane was dried and stained with 0.1% crystal violet at room temperature for 30 min. Finally, the Transwell membrane was cut off and placed on a glass slide for microscopic observation, and six locations were randomly selected for imaging by means of an inverted light microscope (magnification, ×40). The number of cells was calculated for each group using ImageJ 1.50b software (National Institutes of Health). For the cell invasion assay, the procedure was similar to the cell migration assay, except that the Transwell membranes were pre-coated with 50 mg/ml Matrigel (BD Biosciences) at 37°C for 1 h.

TEA domain family members (TEADs) siRNAs were synthesized by Shanghai GenePharma Co., Ltd. Cells were seeded in a 6-well plate to be 60–80% confluent. A total of 1.5 µl siRNA (20 µM) and 9 µl Lipofectamine^® RNAiMAX reagent (cat. no. 13778150; Invitrogen; Thermo Fisher Scientific, Inc.) were mixed with 150 µl Opti MEM medium (cat. no. 31985070; Gibco; Thermo Fisher Scientific, Inc.). The final concentration of siRNAs was 25 pM. Next, diluted siRNA was added to diluted Lipofectamine RNAiMAX reagent and cultured 5 min at room temperature. siRNA-lipid complex was added to cells for 6–8 h at 37°C and cells were re-cultured in complete medium for 48 h at 37°C. Subsequent experiments were performed after the cells were transfected for 48 h. Control siRNA was used as a negative control. siRNAs used for siRNA transfection are listed in Table III.

All data are presented as the mean ± SEM. All experiments, including western blotting, were performed in duplicate or triplicate. Western blotting semi-quantification was implemented by ImageJ 1.50b (National Institutes of Health) Statistical analyses were carried out using the GraphPad 5.0 software (GraphPad Software, Inc.). Comparisons between two groups were implemented using the unpaired two-tailed Student's t-test. Comparisons of three groups were executed by multiple t-tests with the Holm-Sidak method, as recommended by the GraphPad software. P<0.05 was considered to indicate a statistically significant difference.

To confirm whether TNBC cell lines have relatively high expression levels of both HSP90 and HDAC6, the mRNA expression and protein abundance levels of HSP90 and HDAC6 were detected in four breast cell lines. It was identified that the mRNA expression levels of HDAC6 were upregulated in TNBC cell lines, but for HSP90, the two subunits, HSP90AA1 and HSP90AB1, exhibited no significant differences across the four cell lines examined (Fig. 1A and B). Based on the protein expression levels presented in Fig. 1C, HSP90 exhibited similar abundance in the cell lines, except in BT549 where its expression was slightly decreased. Statistical analysis demonstrated that HSP90 protein levels did not no differ across the four cell lines (Fig. 1D). However, HDAC6 expression was upregulated in TNBC cells which was consistent with the results for mRNA expression. Overall, the results indicated that TNBC cell lines maintain relatively high expression levels of HDAC6, whereas the expression levels of HSP90 were almost consistent. It may be speculated that the combination of HSP90 inhibitor and HDAC6 inhibitor could have an improved therapeutic effect in TNBC cell lines. As aforementioned, reciprocal effects exist between HSP90 and HDAC6 (38). The present study aimed to determine whether combined treatment with HSP90 inhibitor and HDAC6 inhibitor may achieve enhanced efficacy compared with treatment with either inhibitor alone.

To determine whether synergistic effects exist between the HSP90 inhibitor 17-AAG and HDAC6 inhibitor Belinostat, the IC₅₀ value and cell viability of these two drugs were measured to generate a concentration-inhibition matrix for combined treatment (Fig. 2A and B). Based on a previous study (51), the effects of the combination were calculated using the CI, because it has been recognized that the fixed dose ratio of two drugs based on individual IC₅₀ values can achieve a rational CI. MDA-MB-231, MCF-10A, MCF-7 and BT549 cells were treated with three different concentrations of 17-AAG and Belinostat. The effective concentration ratio between 17-AAG and Belinostat was ~5, 0.023, 1.8 and 0.0008, respectively (Fig. 2A and B). In MCF-10A and BT549 cells, the CI values were >1 at three groups of concentrations selected according to effective concentration ratio between 17-AAG and Belinostat (Fig. 2C). This indicated that the two drugs have an antagonistic effect. The antagonistic effect may be caused by high sensitivity to 17-AAG of MCF-10A and BT549 cells. BT549 cells exhibited higher sensitivity to 17-AAG than Belinostat (Fig. 2A and B), which could not be observed and concluded from Fig. 1. The mechanism of the dominant effect of 17-AAG in BT549 cells requires further study; however, the present study aimed to explore the combination effect of the HSP90-HDAC6 axis. The CI values were <1 at different concentrations in MDA-MB-231 and MCF7 cells. However, considering the IC₅₀ value of 17-AAG in MCF-7, it was revealed that the difference in cell inhibition rate was not obvious at the near drug concentration (Fig. 2C). Therefore, when setting the fixed molar ratio of the two drugs, the selected three concentration points would lead to a Fa value higher than that of MDA-MB-231. These results indicated that 17-AAG and Belinostat exhibited the best synergistic effect in MDA-MB-231 cells compared with in the other three breast cell lines. Three concentration groups in MDA-MB-231 cells were established for 17-AAG and Belinostat, including 0.50 µM 17-AAG and 0.10 µM Belinostat, 0.80 µM 17-AAG and 0.16 µM Belinostat, and 1.00 µM 17-AAG and 0.20 µM Belinostat (Fig. 2D).

In addition to combination index, the combination of 17-AAG and Belinostat also led to enhanced cell cycle arrest and apoptosis in MDA-MB-231 cells. Two combination groups with low CI values (0.1 µM Belinostat with 0.5 µM 17-AAG, and 0.2 µM Belinostat with 1.0 µM 17-AAG) were selected to treat MDA-MB-231 cells for 3 days. The relative apoptosis rate was not significantly increased in cells treated with 17-AAG or Belinostat (Fig. 3A and C). The combination of 17-AAG and Belinostat led to a marked increase in apoptosis, and the total apoptosis rate increased with the elevation of concentrations (Fig. 3A and C). Additionally, a higher proportion of cells were arrested in the G₂ phase following the combined treatment with 17-AAG and Belinostat compared with the single treatment of 17-AAG or Belinostat alone (Fig. 3B), and the rate of G₂ phase arrest in the combined treatment group was higher than that in the single treatment groups (Fig. 3D). Consistent with the flow cytometry results, western blotting results demonstrated that the levels of cleaved poly (ADP-ribose) polymerase and caspase 3 were upregulated in the combination group. In addition, the protein abundance of CDK1 phosphorylation (p-CDK1) was upregulated in the combination group and in the single 17-AAG (1 µM) group, indicating increased inactivation of CDK1. Furthermore, the protein abundance of cyclin B1 was downregulated in the combination group, indicating an increased proportion of G₂/M phase arrest.

According to a previous study (59), reciprocal interactions may exist between HSP90 and HDAC6. To determine whether Belinostat affected HSP90 expression, and whether 17-AAG affected the protein abundance of HDAC6, qPCR and western blotting were performed. Subsequently, it was demonstrated that Belinostat downregulated the transcription levels of HSP90 (HSP90AA1 and HSP90AB1; Fig. 4A). However, Belinostat did not reduce the HSP90 protein level (Fig. 4B). Compared with the treatment with Belinostat alone, the acetylation of HSP90 was sharply enhanced in the combination group, indicating that 17-AAG amplified the acetylation effect of Belinostat on HSP90. Although Belinostat induced higher acetylation of α-tubulin than 17-AAG, this effect was not further enhanced in the combination group. Similarly, the protein abundance of HDAC6 was decreased following treatment with 17-AAG alone, whereas this effect was not further enhanced in the combination group (Fig. 4B).

To determine the specific genes and pathways that are responsible for the synergistic effect of the combination of 17-AAG and Belinostat, RNA-seq was carried out using MDA-MB-231 cells treated with 17-AAG alone, Belinostat alone, and the combination of 17-AAG and Belinostat. The IC₅₀ ratio of Belinostat and 17-AAG was ~1 vs. 5, as shown in Fig. 2A and B, and the concentration used for RNA-seq was selected according to the previously calculated CI value, apoptosis and cell cycle experiments. When the dose for 17-AAG was 1.0 µM and that for Belinostat was 0.2 µM, cell viability and apoptosis data exhibited an improved combination effect. Overall, this dosage was selected for RNA-Seq exploration. The results demonstrated that the numbers of differentially expressed genes were 3,576, 1,517 and 874, for the treatment of 17-AAG alone compared with the DMSO group, Belinostat alone compared with the DMSO group, and the combination group compared with the DMSO group, respectively. The numbers of shared and unique differentially expressed genes for each treatment group were presented as a Venn diagram (Fig. 5A). In terms of pathways analysis, the P-values for each significantly enriched KEGG pathway were plotted as a histogram for each treatment group, and the length of each bar is proportional to the P-value of each KEGG pathway. the ‘cell cycle’ pathway was more significantly enriched than the ‘TGF-beta signaling pathway’ and the ‘Focal adhesion’ pathway, for both the treatment of 17-AAG alone, or Belinostat alone (Fig. 5B and C). However, ‘Focal adhesion’, ‘TGF-beta signaling pathway’ and ‘Regulation of actin cytoskeleton’ were the most significantly enriched KEGG pathways in the combination group (Fig. 5D). For treatment with 17-AAG alone, the P-values for ‘Focal adhesion’, ‘TGF-beta signaling pathway’ and ‘Regulation of actin cytoskeleton’ were 1.9×10⁻³, 2.1×10⁻³ and 2.9×10⁻³, respectively. For treatment with Belinostat alone, the P-values for ‘Focal adhesion’ and ‘TGF-beta signaling pathway’ were 0.014 and 0.022, respectively. However, in the combination group, the P-values for ‘Focal Adhesion’, ‘TGF-beta signaling pathway’ and ‘Regulation of actin cytoskeleton’ were 6.8×10⁻⁵, 2.9×10⁻⁴ and 4.7×10⁻⁴, respectively. In addition, considering the over-representation of the ‘Oxidative phosphorylation’ pathway in MDA-MB-231 cells following with treatment with Belinostat alone, the accumulation ROS in different groups was examined. It was identified that the ROS level was significantly higher in the combination group (Fig. S1A). In addition, several genes that regulate relevant signaling pathways were also significantly affected in the combination group, including EGFR, COX5B and UBA52 (Fig. S1B). Consistent with the pathway analysis, dozens of migration-associated genes were identified in the list of differentially expressed genes. Some of them were observed in the shared list of all the treatment groups, including the treatment with 17-AAG alone, Belinostat alone and the combination group (Fig. 5E), whereas some of them were observed only in the list of the combination group (Fig. 5F). As these pathways and genes mainly mediate the migration and invasion of cancer cells, these findings indicated that the combination of 17-AAG and Belinostat may have enhanced the inhibition of the migration and invasion of MDA-MB-231 cells compared with treatment with 17-AAG or Belinostat alone.

To confirm whether the combination of 17-AAG and Belinostat may have inhibitory effects on migration and invasion, MDA-MB-231 cells were treated with single or combined compounds and subjected to Transwell and wound-healing assays. The doses for 17-AAG and for Belinostat were consistent with those for RNA-seq. Compared with the treatment of 17-AAG alone, or Belinostat alone, the numbers of migrating and invasive cells were decreased following the combined treatment with 17-AAG and Belinostat for 24 h (Fig. 6A and B). Similarly, in the wounding healing assay, it was identified that the migration of cells was significantly suppressed following the combined treatment with 17-AAG and Belinostat for 24 h (Fig. 6C). According to the RNA-seq data analysis, relevant signaling pathways and genes mediating cell migration and invasion were investigated (60,61). TEAD1 is one of the differentially expressed genes shared by the single and combined treatment groups (Fig. 5E). Additionally, significant downregulation was observed for the TEAD family proteins in the combination group (Fig. 6D). Consistent with RNA sequencing results (Fig. 5E), western blotting results demonstrated that the abundance of TEAD family proteins was decreased in the combination group (Fig. 6E). Additionally, the phosphorylation of YAP and MLC was increased in the combination group. Subsequently, siRNA was used to knockdown TEAD genes in MDA-MB-231 cells. The knockdown efficiency of each siRNA was >80% compared with the normal group (Fig. S2A). Additionally, the protein expression levels of TEADs were decreased (Fig. S2B). As shown in the results of the cell migration assay (Fig. 6F), TEAD1 exhibited a greater effect on migration of MDA-MB-231 cells than other TEADs, whereas TEAD3 had no effect on cell migration.