All of the samples used in the present study were de-identified samples from the hospital's biobank added with the informed consent of the patients, but the requirement of informed consent to include them in the present study was waived by the ethics committee. Primary breast tumor specimens had been collected from female patients diagnosed with invasive breast ductal carcinoma and treated with surgical resection at Kaohsiung Medical University Hospital (Kaohsiung, Taiwan) between January 2010 and January 2017. The inclusion criteria were that the patients had no history of cancer and were not simultaneously diagnosed with any other type of cancer, and the exclusion criteria were a diagnosis with benign breast conditions, ductal carcinoma in situ, microinvasive carcinoma and rare histological tumor types. The follow-up of the patients after treatment was up to 70 months (median, 41 months). Clinical data used in this study were obtained from the hospital's cancer registry with protocols approved by the Institutional Review Board of Kaohsiung Medical University Hospital [approval nos. KMUH-IRB-20130031 and KMUHIRB-E(I)-20180136] and conducted in accordance with the Declaration of Helsinki.

Paraffin-embedded and formalin-fixed tumor sections prepared from the samples taken from surgically treated patients with breast cancer were immunostained with anti-RAD51 antibodies by the Bond-Max automated IHC stainer (Leica Microsystems GmbH) according to the manufacturer's instructions and a previous study by our group (18). The mouse monoclonal antibody against human RAD51 (clone 14B4) for IHC staining was obtained from GeneTex. The images of IHC staining were captured with an Eclipse E600 microscope (Nikon Corp.), followed by quantitation of the IHC staining with categorical scores according to the percentage of positively stained cells (score 1, ≤25%; score 2, 26-50%; score 3, 51-75%; score 4, ≥76%), where samples with scores 1 and 2 were further categorized as low RAD51 expression, and those with scores 3 and 4 were categorized as high RAD51 expression for dichotomized comparisons (18,19). The IHC scoring was evaluated independently by two trained investigators and confirmed by a pathologist.

The human breast cancer cell lines MDA-MB-231 and MCF-7 were purchased from the Bioresource Collection and Research Center (Hsinchu, Taiwan) and maintained in DMEM (Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (Biological Industries) and 1% penicillin-streptomycin-amphotericin B solution (Thermo Fisher Scientific, Inc.). The genotypes of these human cell lines were authenticated by short tandem repeat profiling (TopGen Biotechnology).

Overexpression of RAD51 was conducted by transfecting the breast cancer cell lines with prepackaged lentiviral particles carrying pReceiver-Lv105 vector that expresses full-length human RAD51 (accession no. NM_002875.4) or empty pReceiver-Lv105 vector (GeneCopoeia) as a control. Knockdown of RAD51 was conducted by transfecting the breast cancer cell lines with prepackaged lentiviral particles carrying pLKO.1-puro vector that expresses short hairpin RNA (shRNA) targeting human RAD51 (5′-CGC CCT TTA CAG AAC AGA CTA-3′) or targeting firefly luciferase (5′-GCG GTT GCC AAG AGG TTC CAT-3′; National RNAi Core Facility, Academia Sinica) as a control. The prepackaged lentiviral particles were added to the corresponding cells in their cell culture medium containing 8 µg/ml polybrene (Sigma-Aldrich; Merck KGaA). After 48 h of transfection, 2 µg/ml puromycin (Sigma-Aldrich; Merck KGaA) was added for selection. Surviving cells were continuously maintained in the cell culture medium containing 2 µg/ml puromycin until further experiments.

MDA-MB-231 or MCF-7 cells were seeded in 96-well plates (5×10³ cells/well) overnight and cultured for 72 h prior to the addition of XTT solution (Sigma-Aldrich; Merck KGaA) according to the manufacturer's instructions. Cell viability was assessed by measuring the optical density (OD) at the main wavelength at 475 nm subtracted by the reference wavelength at 660 nm. To determine the half-maximal inhibitory concentration (IC₅₀), the cells were treated with a series of different concentrations of cisplatin (min. to max., 0-160 µM) for 72 h prior to the addition of the XTT solution followed by readout of the OD as described above. The IC₅₀ was calculated using GraphPad Prism 8 (Dotmatics).

MDA-MB-231 or MCF-7 cells were seeded in 6-well plates (2×10⁵ cells/well) under normal cell culture conditions. At 80% confluence, the cells were collected and fixed with 75% ethanol (Sigma-Aldrich; Merck KGaA) at -20°C overnight, followed by staining the DNA contents with propidium iodide/RNase Staining Buffer (BD Biosciences) for 30 min at 37°C. These cells were then measured with a Cytomics FC 500 flow cytometer (Beckman Coulter), and the distribution of the cell cycle phases was analyzed by WinMDI 2.8 (http://www.cyto.purdue.edu/flowcyt/software.htm).

MDA-MB-231 or MCF-7 cells were plated in Transwell inserts with 8 µm pores (2×10⁴ cells/insert in 500 µl serum-free cell culture medium per well) of the 24-well plate setting (Corning, Inc.), with the bottom wells containing complete cell culture medium. After 24 h under normal cell culture conditions, the cells remaining on the upper side of the insert were removed with cotton swabs, while those that had migrated to the underside of the insert were fixed with 4% formaldehyde (Sigma-Aldrich; Merck KGaA) for 30 min at room temperature, followed by staining with 0.05% crystal violet (Sigma-Aldrich: Merck KGaA) for 15 min at room temperature. The images of crystal violet staining were captured with an Eclipse Ti Series microscope (Nikon Corp.) and analyzed by ImageJ version 1.53e [National Institutes of Health (NIH)].

MDA-MB-231 cells were plated onto 8-well Nunc Lab-Tek glass chamber slides (5×10³ cells/well; Thermo Fisher Scientific, Inc.) that were pre-coated with Cy3-conjugated gelatin (Merck Millipore) according to the procedures described in the QCM Gelatin Invadopodia Assay (20). After 24 h under normal cell culture conditions, the cells were fixed with 4% formaldehyde (Sigma-Aldrich; Merck KGaA) for 30 min at room temperature, followed by concurrent staining with 2 µg/ml FITC-conjugated phalloidin and 1 µg/ml DAPI for 20 min at room temperature to label filamentous actin (F-actin) and the cell nucleus, respectively. The images of each well were captured by a Zeiss Axioscope fluorescence microscope (Carl Zeiss GmbH) and gelatin degradation was quantified as the degradation area of gelatin normalized to the number of cells using ImageJ (NIH).

Protein was extracted using RIPA buffer and the concentration of the protein was measured with a BCA protein assay kit (Pierce™ BCA Protein Assay Kit; ThermoFisher). Equal amounts of total protein extracted from MDA-MB-231 or MCF-7 cells (30 µg/sample) were separated by one-dimensional 10% SDS-PAGE and subsequently transferred to PVDF membranes (PALL Life Sciences) using the Mini-PROTEAN and Trans-Blot systems (Bio-Rad Laboratories, Inc.). The membranes were blocked with 5% non-fat milk (Anchor) for 1 h at room temperature, followed by sequential incubation with primary antibodies overnight at 4°C and species-matched horseradish peroxidase-conjugated secondary antibodies (Thermo Fisher Scientific, Inc.) for 1 h at room temperature. The signal of immunoreactive proteins was developed in the presence of chemiluminescent reagents (Thermo Fisher Scientific, Inc.) and acquired by the ChemiDoc XRS+ imaging system (Bio-Rad Laboratories, Inc.). The primary antibodies used for western blot were as follows: Mouse monoclonal antibodies against human RAD51 (cat. no. 100469; 1:5,000 dilution; GeneTex), F-actin (cat. no. 205; 1:500 dilution; Abcam) or Lamin A/C (cat. no. 4777; 1:2,000 dilution; Cell Signaling); and rabbit polyclonal antibodies against human RAD51, GAPDH (cat. no. 100118; 1:60,000 dilution; GeneTex) or α-tubulin (cat. no. 112141; 1:5,000 dilution; GeneTex).

The cytoplasmic and nuclear protein fractions of MDA-MB-231 cells were extracted using the Subcellular Protein Fractionation Kit (Thermo Fisher Scientific, Inc.). Equal amounts of each subcellular fraction (500 µg/sample) were added to 100 ml Protein A Mag Sepharose (Sigma-Aldrich; Merck KGaA) conjugated with 5 µg of the RAD51 antibodies or F-actin antibodies (as specified above) overnight at 4°C on a rotary tube mixer (Thermo Fisher Scientific, Inc.). The immunoprecipitated protein complex was then separated from the sepharose by boiling for 10 min in Laemmli sample buffer (Bio-Rad Laboratories, Inc.) and subjected to one-dimensional 10% SDS-PAGE for western blot analysis according to the above-mentioned western blot protocol or in-gel digestion of the proteins.

In-gel digestion and protein identification by LC-MS/MS were conducted following previously reported procedures (21). The protein band identified from one-dimensional 10% SDS-PAGE by Coomassie Brilliant Blue R-250 staining (Thermo Fisher Scientific, Inc.) was sliced and de-stained with 25 mM ammonium bicarbonate (Sigma-Aldrich; Merck KGaA) in 50% acetonitrile (Sigma-Aldrich; Merck KGaA) for 1 h at room temperature. The sliced gel was then dehydrated in 100% acetonitrile (Sigma-Aldrich; Merck KGaA) for 5 min and vacuum-dried for 30 min at room temperature. In-gel digestion was conducted in the presence of 0.5 µg trypsin (Sigma-Aldrich; Merck KGaA) dissolved in 25 mM ammonium bicarbonate (Sigma-Aldrich; Merck KGaA) for 16 h at 37°C. The digested peptide fragments were extracted with 50 µl of a mixture containing 50% acetonitrile (Sigma-Aldrich; Merck KGaA) and 0.1% trifluoroacetic acid (Sigma-Aldrich; Merck KGaA). The extracted peptides were then dissolved in a mixture of 0.1% formic acid (Sigma-Aldrich; Merck KGaA) and 50% acetonitrile (Sigma-Aldrich; Merck KGaA), and analyzed by LC-MS/MS at the Center for Research Resources and Development, Kaohsiung Medical University (Kaohsiung, Taiwan). Protein matching was conducted by a database search for the results of LC-MS/MS spectra with the aid of Mascot (version 2.2; Matrix Science) (https://www.matrixscience.com) (22).

MDA-MB-231 cells grown on 8-well Nunc Lab-Tek II chamber slides (Thermo Fisher Scientific, Inc.) were fixed with 4% paraformaldehyde (Sigma-Aldrich; Merck KGaA) for 30 min at room temperature, followed by incubation of the cells with a permeabilization/blocking buffer containing 0.5% Triton-X 100 (Sigma-Aldrich; Merck KGaA) and 1% bovine serum albumin (Sigma-Aldrich; Merck KGaA) in PBS for 1 h at room temperature. The cells were then sequentially incubated with rabbit polyclonal antibodies against human RAD51 (cat. no. 100469; 1:200 dilution; GeneTex) overnight at 4°C and Alexa Fluor 555-conjugated donkey polyclonal antibodies against rabbit IgG (cat. no. A32794; 1:500 dilution; Thermo Fisher Scientific, Inc.) for 1 h at room temperature, followed by co-staining with 2 µg/ml FITC-conjugated phalloidin (for F-actin) (cat. no. A12379; Thermo Fisher Scientific, Inc.) and 1 µg/ml DAPI (AAT Bioquest) (for the cell nuclei) for 20 min at room temperature. After the cells were mounted onto coverslips with Mounting Media (Thermo Fisher Scientific, Inc.), images of the cells were captured by Zeiss Axioscope fluorescence microscope (Carl Zeiss GmbH) and analyzed by ImageJ (NIH).

The statistics in this study were analyzed by SPSS (version 25; IBM Corp.) or GraphPad Prism 8 (Dotmatics). Kaplan-Meier survival analysis was conducted to determine disease-free and overall survival of the patients, and differences between survival curves were assessed with the log-rank test. The association between RAD51 expression and clinicopathological parameters was assessed by chi-square test. Hazard ratio (HR) and 95% confidence interval (CI) were derived from Cox regression models. For the in vitro studies, the data were presented as the mean ± standard deviation from three independent experiments unless otherwise indicated, and the difference between experimental and control groups was assessed by Student's t-test. P<0.05 was considered to indicate a statistically significant difference.

To investigate the association of cytoplasmic (cyto) and nuclear (nu) RAD51 expression with cancer progression, primary tumors obtained from patients with surgically treated breast cancer (age range, 28-86 years; mean ± SD, 51.54±11.42 years) were assessed by IHC (Fig. 1A), and categorical scores were assigned according to the expression levels of RAD51. In general, a heterogeneous expression pattern of RAD51 was observed across the tumor tissues (Fig. 1A). The initial screening was conducted via classification of RAD51 expression into four groups as reported previously (17), including cyto^low/nu^high, cyto^high/nu^high, cyto^low/nu^low and cyto^high/nu^low of RAD51 (n=148). Using this analysis, consensus results of the association between subcellular RAD51 expression and breast cancer progression were observed, such as tumor size (P=0.022; Table SI) and disease-free survival of patients with breast cancer (P=0.007; Fig. S1A). In addition to these, the current results revealed that cancer stage and lymph node (LN) metastasis (P=0.001 for both; Table SI) and overall survival of the patients (P=0.025; Fig. S1B) were associated with subcellular RAD51 expression. Overall, significant differences in patient survival, along with other clinicopathological parameters, including tumor size, stage and LN metastasis, were observed among the four patient groups according to their status of subcellular RAD51 expression.

To further compare the roles of differential RAD51 expression in the cytoplasm and in the cell nuclei of breast tumors, independent analyses were performed for patients who were dichotomously grouped into low vs. high expression of cytoplasmic RAD51 (n=224) and nuclear RAD51 (n=180). It was found that the patient group with high cytoplasmic RAD51 expression was associated with increased cancer stage (P=0.003), grade (P=0.035), tumor size (P=0.036) and LN metastasis (P=0.021; Table I). However, the association was the inverse in the patient group with high nuclear RAD51 expression, as it was observed that the patient group with high nuclear RAD51 expression was associated with decreased cancer stage (P<0.001), tumor size (P=0.005) and LN metastasis (P=0.001; Table II). In the survival analysis, the patient group with high cytoplasmic RAD51 expression was associated with poorer disease-free survival (P=0.002; Fig. 1B) and overall survival (P=0.011; Fig. 1C) during a postoperative follow-up period up to 72 months, whereas the patient group with high nuclear RAD51 expression conversely showed positive association with disease-free survival (P=0.003; Fig. 1D) and overall survival (P=0.008; Fig. 1E) compared to their corresponding low RAD51 expression groups. These results suggest the utility of subcellular RAD51 in breast tumor tissues as prognostic markers, with the notion that high cytoplasmic RAD51 expression was adversely associated with breast cancer progression, such as increased cancer stage, tumor size and LN metastasis, along with reduced patient survival; however, high nuclear RAD51 expression had the inverse effect.

The univariate and multivariate Cox regression models were subsequently applied to analyze the utility of parameters of clinicopathologic characteristics and differential RAD51 expression as potential risk factors for patient survival. In the univariate analysis, the parameters showing a strong association with disease-free survival of the patients included cancer stage (II vs. I, P=0.035; III vs. I, P=0.001) and cytoplasmic RAD51 expression (high vs. low, P=0.004) (Table III). In the multivariate analysis, the cancer stage (III vs. I, P=0.012) remained significantly associated with disease-free survival of the patients (Table III). In the univariate analysis for overall survival, the parameters of cancer stage (II vs. I, P=0.026; III vs. I, P=0.006) and cytoplasmic RAD51 expression (high vs. low, P=0.015) also showed a strong association with patient survival, and the association of cancer stage (II vs. I, P=0.048; III vs. I, P=0.017) with overall survival remained significant in the multivariate analysis (Table IV). In these Cox regression models, the association between nuclear RAD51 expression and patient survival was not significant (Tables III and IV), suggesting that cytoplasmic RAD51 expression, rather than nuclear RAD51 expression, represented a more influential risk regarding survival of patients with breast cancer. In addition to these, it was observed that the histologic grade, which is frequently associated with patient outcomes, showed a trend of increasing hazard ratios but did not reach a statistically significant level in the univariate and multivariate analyses (Tables III and IV); this discrepancy may be resolved by further investigation with a larger magnitude of sample size.

Resistance to chemotherapy has been one of the main obstacles in breast cancer management, yet clinical markers to effectively predict responses to chemotherapy are still under exploration (23,24). Therefore, in the present study, the effects of subcellular RAD51 expression on patient survival with regard to chemotherapy was further investigated in these cohorts. As presented in Fig. 2A and C, it was found that high cytoplasmic RAD51 expression was associated with poorer disease-free survival (P=0.010) and overall survival (P=0.040) in the patient group receiving adjuvant chemotherapy, while the association of cytoplasmic RAD51 expression with patient survival was not significant in the patient group without chemotherapy (Fig. 2B and D). An opposite trend was observed in the analysis for nuclear RAD51 expression, where high nuclear RAD51 expression was in favor of disease-free survival (P=0.006) and overall survival (P=0.040) in the patient group receiving chemotherapy (Fig. 2E and G), while the significance of association was not observed in the patient group without chemotherapy (Fig. 2F and H). These results indirectly suggest the potential of a different prognostic role of subcellular RAD51 expression in consideration of patients with breast cancer receiving adjuvant chemotherapy.

The roles of differential RAD51 expression on breast cancer cell behaviors were investigated via the following in vitro assays. Two widely studied human breast cancer cell lines, MDA-MB-231 (basal type) and MCF-7 (luminal type A) (25), were transfected with lentiviral particles that contain full-length RAD51 for gene overexpression or shRNA targeting RAD51 for gene knockdown (Fig. 3A). The XTT assay to determine cancer cell growth revealed that overexpression of RAD51 enhanced the proliferative ability of these breast cancer cells (Fig. 3B), whereas knockdown of RAD51 reduced their proliferative ability compared to their corresponding controls (Fig. 3C). The cell cycle distribution was further evaluated by flow cytometry, which showed that both MDA-MB-231 and MCF-7 cells with overexpression of RAD51 had an increased proportion of the G2/M phases (Fig. 3D). On the contrary, knockdown of RAD51 led to a reduced proportion of the G2/M phases in these breast cancer cells (Fig. 3E). Taken together, the results suggest that elevation of RAD51 expression promoted breast cancer cell growth, which potentially resulted from the progression of the cell cycle into the G2/M phases.

To investigate whether the differential RAD51 expression is involved in the resistance of breast cancer cells to chemotherapy, the effect of RAD51 overexpression or knockdown on cell growth of the breast cancer cells that were treated with cisplatin, a first-line therapeutics administered in breast and various tumors, was assessed (26). An increased shift of the IC₅₀ value of cisplatin was observed in both RAD51-overexpressing MDA-MB-231 cells (Fig. 4A) and MCF-7 cells (Fig. 4B), whereas knockdown of RAD51 resulted in a reduced IC₅₀ of cisplatin in these breast cancer cells (Fig. 4C and D). The shift of the IC₅₀ for cisplatin treatment further suggests that the differential RAD51 expression may be involved in the regulation of cancer cell growth during chemotherapy, with the notion that resistance to cisplatin treatment occurred in RAD51-overexpressing breast cancer cells.

While the effect of cisplatin treatment on RAD51 subcellular localization has not been previously reported, to the best of our knowledge, it was reported that curcumin induces DNA damage and leads to RAD51 migrating to the nucleus (27). Previous studies also showed that in HeLa and HCT116 cells, RAD51 redistributed from the cytoplasm to the nucleus after exposure to ionizing radiation (28,29). In the current study, it was shown that RAD51 overexpression led to increased cytoplasmic and nuclear RAD51 expression and more surviving cells (chemoresistance) among MDA-MB-231 and MCF-7 cells after cisplatin treatment (Figs. S2 and 4). On the other hand, RAD51 knockdown led to decreased surviving cells (chemosensitivity) in MDA-MB-231 cells after cisplatin treatment. All of these data together suggest that RAD51 expression levels are positively associated chemoresistance and radioresistance in breast cancer cells.

The effect of RAD51 expression on the in vitro metastatic ability of breast cancer cells was investigated by Transwell migration and invadopodial invasion assays. As presented in Fig. 5A, MDA-MB-231 and MCF-7 cells with overexpression of RAD51 exhibited an increased ability of cell migration. On the contrary, knockdown of RAD51 decreased the cell migration ability of these breast cancer cells (Fig. 5B). In addition to cell migration, increased ability of invadopodial cell invasion was observed in MDA-MB-231 cells with overexpression of RAD51, whereas knockdown of RAD51 repressed the cell invasion ability (Fig. 5C). To further explore the biological activity associated with subcellular RAD51 expression, the cytoplasmic and nuclear proteins from MDA-MB-231 cells with overexpression of RAD51, or those from control cells, were fractionated prior to immunoprecipitation with anti-RAD51 antibodies. The fractionation result showed that endogenous RAD51 is expressed in both the cytoplasm and nuclei of MDA-MB-231 cells and overexpression increased RAD51 expression in both cytoplasm and nuclei (Fig. S2A). After immunoprecipitation with anti-RAD51 antibodies, a clearly stained band was shown above and near the molecular weight of 40 kDa in the cytoplasmic fraction of RAD51-overexpressing MDA-MB-231 cells, while there was no clear band in the nuclear fraction of RAD51-overexpressing MDA-MB-231 cells or control cells (Fig. S2B), suggesting the presence of a substantial protein interaction between cytoplasmic RAD51 and the co-immunoprecipitated protein. Identification of the protein sequences for this particular band by LC-MS/MS revealed that it matched the sequences of β-actin (Fig. S2C), the major cytoplasmic isoform of the actin family expressed in non-muscle cells (30). As β-actin is the essential subunit of globular actin that forms F-actin in control of various types of cell growth and motility (30-32), the potential of the protein interaction between RAD51 and F-actin was further examined using breast cancer cell models in the present study. As indicated in Fig. 5D, overexpression of RAD51 in MDA-MB-231 cells did not affect the endogenous amount of F-actin expression. However, immunoprecipitation with anti-F-actin antibodies revealed that RAD51 was more abundantly co-immunoprecipitated in RAD51-overexpressing MDA-MB-231 cells compared to control cells (Fig. 5E). The increased protein interaction between RAD51 and F-actin was also evidenced by immunofluorescence, where RAD51-overexpressing MDA-MB-231 cells showed higher abundance of co-localization between RAD51 and F-actin compared to control cells (Fig. 5F). These results together suggest that alteration of RAD51 expression was associated with the migration and invasion ability of breast cancer cells, and the interaction between cytoplasmic RAD51 and actin filaments was potentially an association factor on these biological activities, which warrants further investigation.