A circulating microRNA-based diagnostic model for breast cancer, in which the miR-139-3p/RPA2 axis regulates the sensitivity to DNA-damaging agents

Introduction

Breast cancer (BC), which accounts for 11.7% of all new cancer diagnoses, has now surpassed lung cancer as the most commonly diagnosed worldwide (1). Delayed diagnosis and advanced BC are associated with a worse prognosis; thus, early diagnosis is a key determining factor in reducing mortality (2).

Mammography and ultrasonography are powerful screening methods for the early detection of BC. However, challenges persist, including a lower sensitivity in young women with dense breast tissue, concerns about radiation exposure, false-positive results, and the inability to monitor between examinations (3,4). Biopsy is the gold standard for diagnosing BC and provides information on the molecular pathology of the tumor. However, puncture biopsy is highly invasive, difficult to perform, and subject to tumor heterogeneity. This poses significant challenges to the treatment of BC. These challenges have fueled interest in liquid biopsy, which has become an important component of molecular diagnosis (5,6). Liquid biopsy analyzes body fluids such as blood and urine to non-invasively assess the general molecular characteristics of tumors. In clinical practice, serum tumor markers such as carbohydrate antigen 15-3 (CA15-3) and carcinoembryonic antigen (CEA) are highly specific but extremely insensitive in BC detection (7,8). Thus, more sensitive and specific novel non-invasive diagnostic biomarkers need to be identified.

MicroRNAs (miRNAs or miRs) are endogenous, small (between 17 and 25 nucleotides), non-coding RNAs that can decrease the messenger RNA (mRNA) levels of their targets, mainly by targeting the corresponding prime untranslated region (3’-UTR). Studies have shown that miRNAs can either promote or suppress BC, thereby regulating proliferation, metastasis, and drug sensitivity (9-11). Cancers often have aberrant miRNA expression profiles in tissues and circulation (12). Circulating miRNAs often bind to proteins or lipids or are encapsulated in vesicles, remaining highly stable in human blood and unaffected by endogenous ribonuclease (RNase) activity (13,14). This characteristic makes circulating miRNAs excellent blood-based tumor biomarkers for cancer diagnosis and prognosis, and potential therapeutic targets.

A number of miRNAs in circulation have been suggested as diagnostic biomarkers of BC (15-18). Research has shown that miR-21, miR-202, and miR-19b levels are increased in the plasma of BC patients (19,20). Additionally, other studies have identified combinations of miRNAs that have high diagnostic accuracy (21-23). However, there is little overlap among the miRNAs investigated in these studies. This may be due to the small sample sizes of these studies, the limited number of miRNAs, the use of different detection methods, and the lack of validation in clinical practice.

Therefore, this study aimed to identify key circulating miRNAs and develop a robust diagnostic model for BC using large-scale datasets and machine learning. Furthermore, by functionally characterizing one of the prioritized miRNAs, we sought to explore its role in regulating DNA damage sensitivity. Understanding this mechanism is critical not only for deciphering the biological basis of the miRNA’s association with patient prognosis, but also for assessing its potential as a predictive biomarker for therapy response and as a novel therapeutic target to sensitize tumors to DNA-damaging agents, thereby bridging diagnostic discovery with actionable therapeutic insights. We present this article in accordance with the TRIPOD, ARRIVE and MDAR reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2592/rc).

Methods

Data acquisition and process

Four public BC datasets were analyzed in this study: three serum datasets (GSE211692, GSE106817, and GSE73002) and one tissue dataset [i.e., The Cancer Gene Atlas (TCGA)-Breast Invasive Carcinoma (BRCA)]. The GSE211692, GSE106817, and GSE73002 datasets were acquired from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds). TCGA-BRCA dataset was obtained from the Genomic Data Commons (GDC) database (https://portal.gdc.cancer.gov/).

TCGA-BRCA dataset included the clinical and miRNA sequence information of 1,103 BC samples, together with 104 samples of adjacent normal tissue. The GSE211692 dataset comprised the clinical information and serum miRNA array data of 675 BC patients and 5,643 non-cancer controls, and was used for the differential expression analysis. The GSE106817 dataset comprised the clinical information and serum miRNA array data of 115 BC patients and 2,759 non-cancer controls, and was used for the differential expression analysis. Due to the imbalanced distribution of samples between the normal and BC groups in the large serum miRNA dataset GSE211692, the “ROSE” R package with random undersampling was used to process the dataset to obtain a suitable training set (normal: n=643, BC: n=675). The same method was applied to the GSE106817 dataset to obtain the testing set (normal: n=117; BC, n=115). The GSE73002 dataset comprised the data of 900 BC patients and 1,500 non-cancer controls (after removing the missing values), and served as the validation set for evaluating the model’s diagnostic efficacy.

Data analysis

Differential expression analyses of the GSE211692, GSE106817, and TCGA-BRCA datasets were performed using the “limma” package in R with a false discovery rate (FDR) <0.05 and |log₂fold change| >1 (24). The differentially expressed miRNAs were identified by intersecting the upregulated and downregulated miRNAs from the three datasets. Venn diagrams were plotted in R using the package “Venn”. Receiver operating characteristic (ROC) curves were created using the R package “pROC” (25). Correlation coefficients were estimated using the “cor” R function and the Pearson method. The “rms” package was used to create a probability prediction graph (nomogram) and perform calibration. For the survival analysis of TCGA-BRCA cohort, starBase (https://rnasysu.com/encori/) was used to generate the log-rank P value. For the survival analysis of METABRIC, Kaplan-Meier plotter (http://kmplot.com/analysis) was used to generate the log-rank P value. The miRNA pathway analysis was performed using Tumor-miRNA-Pathway (http://bioinfo.life.hust.edu.cn/miR_path).

Machine-learning algorithms

The least absolute shrinkage and selection operator (LASSO), random forest (RF), and support vector machine-recursive feature elimination (SVM-RFE) machine-learning algorithms were used to select features with higher diagnostic accuracy in the training set. The “glmnet” R package was used to implement the LASSO logistic regression, and “lambda.1se” was chosen as the optimal lambda. The RF analysis was conducted using the “randomForest” R package. The RF model outputted the importance of each feature. The “e1071” R package was used to apply the SVM-RFE algorithms, and obtain the optimal variable number and average rank.

Patient cohort and samples

In total, 102 serum samples were obtained from 55 BC patients and 47 healthy controls from the Tianjin Medical University Cancer Institute and Hospital between August 2022 and December 2022. Patients with newly diagnosed BC who had not undergone surgery, radiotherapy, or chemotherapy prior to specimen collection were included in the analysis, as were healthy controls without BC or any history of any malignant tumors. The fresh blood was processed by centrifugation (2,000 ×g, 15 min, 4 ℃) within 24 hours. Hemolyzed blood samples were excluded. The supernatants were stored in Eppendorf tubes at −80 ℃. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Ethical approval was obtained from the Tianjin Medical University Cancer Institute and Hospital Ethical Review Board (No. bc2021276). Written informed consent was obtained from each patient.

RNA extraction and quantification

The total RNA was extracted from the serum samples using a HiPure Liquid RNA Mini Kit (Magen, Guangzhou, China) according to the instructions provided. Briefly, a 250-µL serum sample was lysed in 750 µL of MagZol LS reagent. The total RNA was extracted and then eluted into 30 µL of RNase-free water using the HiPure Viral Mini Column, following the provided protocol, and stored at −80 ℃.

The “miRNA Design tool (version 1.01)” was used to design the miRNA stem-loop reverse transcription primers and forward primers. U6 served as the internal reference. The sequences of the primers are provided in Table S1. A specific miRNA reverse transcription reaction to generate complementary DNA (cDNA) was undertaken using the miRNA 1^st Strand cDNA Synthesis Kit (by stem-loop) (Vazyme, Nanjing, China). The cycling reaction conditions were 25 ℃ for 5 min, 55 ℃ for 15 min, and 85 ℃ for 5 min. The cDNA was diluted 10-fold, and miRNA-specific forward primers and universal reverse primers were used for subsequent real-time quantitative polymerase chain reaction (RT-qPCR). RT-qPCR was conducted using a Roche LightCycler 480 and miRNA Universal SYBR qPCR Master Mix (Vazyme). The reaction conditions of RT-qPCR were as follows: preincubation at 95 ℃ for 5 min, and amplification for 40 cycles at 95 ℃ for 10 s and 60 ℃ for 30 s. The relative expression of the miRNAs was calculated using ΔCq (ΔCq = Cq miRNA − Cq U6), and the fold change was determined using the 2^−ΔΔCq method (ΔΔCq = ΔCq sample − mean ΔCq control). The fold change was subsequently used to analyze diagnostic efficacy and explore correlations with clinical characteristics.

Total RNA was extracted from the cells and tissues with TRIzol, and the mRNA levels were analyzed by RT-qPCR. One microgram of the diluted RNA was reverse-transcribed into cDNA using a 1^st Strand cDNA Synthesis Kit (Transgene, Beijing, China). The cDNA product was diluted 40-fold for RT-qPCR, and RT-qPCR was performed using the SYBR Green Master Mix (Vazyme). The reaction conditions of RT-qPCR were preincubation at 95 ℃ for 10 min, amplification for 40 cycles at 95 ℃ for 30 s, and 60 ℃ for 30 s. ACTB was selected as an internal reference for normalization using the 2^−ΔΔCq method.

Cell lines and cell culture

MDA-MB-231, MCF-7, U2OS, and HEK293T cells were purchased from the American Type Culture Collection (ATCC) and cultured in accordance with the accompanying instructions. The cells were certified mycoplasma-free. The cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37 ℃ and 5% CO₂.

DNA and RNA transfection

The miRNA mimics and their negative controls (NCs) were purchased from GenePharma (Suzhou, China). Their sequences are provided in Table S2. The RPA2 overexpression plasmid was synthesized by General Biosystems (Anhui, China). MiRNA transfection was conducted using the Lipofectamine RNAiMAX Reagent (Invitrogen, Waltham, USA) at a final concentration of 50 nM. Plasmid transfection was performed using Lipo2000 (Invitrogen) in accordance with the provided instructions. The cells were collected 48–96 hours after transfection according to the experimental purposes.

Lentivirus preparation and infection

The lentiviral vectors encoding RPA2 were co-transfected with a packaging plasmid mixture (GeneCopoeia, Rockville, USA) into the HEK293T cells using the EndoFectin Lenti transfection reagent (GeneCopoeia) in accordance with the provided instructions. After transfection for 48 hours, the supernatants with the packaged lentiviruses were harvested by centrifugation and filtering. The lentivirus-infected cells were selected using puromycin.

Cell viability assays

The cells (3,000–3,500/well) were inoculated in 96-well plates and grown for 24 hours, after which the medium was exchanged for a medium containing Cisplatin or Olaparib (Selleck, Houston, USA) in a concentration gradient. After 72 hours, the MTS reagent (Promega, Madison, USA) was included in each well at a ratio of 1:5 (MTS: media). After 1 hour, absorbances at 490 nm were read in a microplate reader (Bio-Tek, Madison, USA). The cell viability was expressed relative to that of the control-treated cells. Each drug concentration was tested in triplicate.

Colony-forming assays

Colony-forming assays were used to evaluate the sensitivity of the BC cells to X-ray irradiation (IR). The MCF-7 and MDA-MB-231 cells were inoculated in 6-well plates at 1,000 and 600 cells/well, respectively, and grown for 24 hours, after which they were exposed to X-ray IR from an X-ray generator (RS2000 PRO, RadSource Corporation, Suwanee, USA) at a dose rate of 1 Gy/50 s. The cells were cultured for another 10–12 days; the culture medium was changed every 3 days.

Western blotting

Western blotting was conducted using standard methods. The following antibodies were used: anti-gH2AX (Millipore, Massachusetts, USA, 05-636), anti-β-actin (Servicebio, Wuhan, China, GB12001), anti-RPA2 (ProteinTech, Wuhan, China, 10412-1-AP).

Immunofluorescence

The cells were seeded into 12-well plates with glass coverslips (BD Biosciences, New Jersey, USA). After treatment, the coverslips were fixed in 4% paraformaldehyde for 8 min, permeabilized in 0.2% Triton X-100 for 4 min, and blocked in 1% bovine serum albumin (BSA) for 30 min. The slides were incubated with the primary antibodies (anti-gH2AX, Millipore, Massachusetts, USA, 05-636; anti-RAD51, Abcam, Cambridge, UK, ab133534) overnight at 4 ℃, followed by incubation with a fluorescent secondary antibody for 1 hour at room temperature, and subsequent counterstaining with 4’,6-diamidino-2-phenylindole (DAPI; Thermos, Massachusetts, USA). Confocal images of the scanned areas were acquired using a Zeiss LSM900 microscope (Oberkochen, Germany) with a ×63 oil objective.

Dual-luciferase assays

Recombinant pmirGLO-RPA2-3’UTR-wt and corresponding mutant-type pmirGLO-RPA2-3’UTR-mut plasmids were synthesized by GenePharma. The HEK293T cells were inoculated in 12-well plates, and wild-type or mutant reporter plasmids (250 ng) were co-transfected with either miR-139-3p or NC mimics (50 nM) using Lipo2000. After 48 hours, the cells were evaluated using the Dualucif Firefly and Renilla Assay Kit (UElandy, Suzhou, China).

Homologous recombination (HR) and non-homologous end joining (NHEJ) reporter assays

The HR repair and NHEJ efficiencies were examined using HR repair and NHEJ reporter cells as described previously (26). Green fluorescent protein (GFP)-positive cells were scored through fluorescence-activated cell sorting (FACS) analysis using a BD FACSVerse (BD Biosciences) and a minimum of 10,000 counted cells.

Xenograft tumor experiment

The mice we purchased were sourced from SPF (Beijing) Biotechnology Co., Ltd. (Beijing, China). The MDA-MB-231 BC cells (3×10⁶) were transplanted into the mammary fat pads of 6-week-old female BALB/c nude mice. Ten days after inoculation, the mice were randomly allocated to four groups: NC, miR-139-3p, NC + cisplatin, and miR-139-3p + cisplatin. Cisplatin [2 mg/kg in phosphate-buffered saline (PBS)] was administered by intraperitoneal injection once every 3 days for a total of 5 doses (over a 15-day treatment course) to the groups receiving cisplatin. The mice in the control group received an equivalent volume of the solvent via the same route. Control or miR-139-3p agomiR [5 nmol in diethyl pyrocarbonate (DEPC) water] was peritumorally injected once every 3 days for a total of 5 doses (over a 15-day treatment course). The nude mice were anesthetized with 2% isoflurane inhalation by injecting tumor cells into the mammary fat pad and administering drugs via intraperitoneal and peritumoral injections. Five mice were analyzed per group. The tumor volume (V = length × width² × 0.5) of each mouse was assessed every 3 days after the first administration. On the third day after the completion of the treatment, the mice were sacrificed and the tumors were harvested. The mice were euthanized through CO₂ inhalation following the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals: 2020 Edition. The animal experiments were approved by the Laboratory Animal Management Committee of Tianjin Medical University Cancer Institute and Hospital (No. PMIF-2021082). The standard operating procedures involving animals were conducted in accordance with the Institutional Animal Care and Use Committee Guidebook. A protocol was prepared before the study without registration.

Statistical analyses

The statistical analyses were performed using GraphPad Prism software (version 8.0, GraphPad, San Diego, CA, USA). Comparisons among the biological data were performed using the Student’s t-test or one-way analysis of variance. Each biological experiment was performed in triplicate, and the data are presented as the mean ± standard deviation. P<0.05 was considered significant for all analyses.

Results

Differentially expressed circulating miRNAs in BC patients

The differential expression analysis of circulating miRNAs was performed using two large-scale circulating miRNA datasets. In the GSE106817 dataset (normal: n=2,759; BC: n=115), 421 upregulated and 512 downregulated miRNAs were identified in the BC patients (Figure 1A). Similarly, in the GSE211692 dataset (normal: n=5,643, BC: n=675), 464 upregulated and 594 downregulated miRNAs were identified in the BC patients (Figure 1B). The analysis of the tissue miRNA expression profiles from the TCGA-BRCA dataset (normal: n=104, BC: n=1,103) revealed 85 upregulated and 107 downregulated miRNAs (Figure 1C). After overlapping the differential analysis results from the three datasets, 7 common downregulated miRNAs and 22 common upregulated miRNAs were identified.

Figure 1 Screening of the key circulating miRNAs for breast cancer diagnosis. (A-C) Volcano maps of the differential analysis in the (A) GSE106817, (B) GSE211692, and (C) TCGA-BRCA datasets. (D) LASSO regression coefficient profiles of the 29 variables in the training dataset. (E) Cross-validation curve for tuning the LASSO regression (parameter λ). (F) Cumulative error rates by random forest classification. (G) Random forest variable importance for diagnosis. (H) The curve of SVM feature numbers with error rates. (I) Overlap of the results of the three machine-learning algorithms. CV, cross-validation; FDR, false discovery rate; FC, fold change; LASSO, least absolute shrinkage and selection operator; miRNA, microRNA; RFE, recursive feature elimination; SVM, support vector machine; TCGA, The Cancer Gene Atlas.

To further identify the key circulating miRNAs involved in BC diagnosis, three machine-learning algorithms were applied to the training set. The LASSO regression analysis identified 15 miRNAs with non-zero coefficients among the 29 candidate miRNAs (Figure 1D,1E). The RF algorithm was used to assess the association between the error rate and number of trees and calculate the importance of each circulating miRNA (Figure 1F,1G). The results of the SVM-RFE analysis indicated that the error was minimized and the accuracy was the highest when considering the top four variables in the ranking (Figure 1H and Figure S1). Collectively, the three machine-learning algorithms identified the following four circulating miRNAs as optimal features for BC diagnosis: miR-139-3p, miR-134-3p, miR-629-3p, and miR-191-3p (Figure 1I). As the heatmap shows (Figure S2), the four key miRNAs could effectively discriminate between the control group and the BC group.

Construction and validation of a circulating miRNA-based BC diagnostic model

The correlation analyses of these four miRNAs showed that only miR-629-3p and miR-191-5p exhibited moderately positive correlations with BC (r=0.25) (Figure 2A). To evaluate the independent diagnostic performance of the four miRNAs selected by the machine-learning algorithms, an ROC curve analysis was performed to determine the area under the curve (AUC) for each miRNA for predicting BC. Our findings indicated that these four miRNAs exhibited strong diagnostic performance for BC in the training and testing datasets. The AUCs and corresponding 95% confidence intervals (CIs) for miR-191-5p, miR-629-3p, miR-134-3p, and miR-139-3p in the training set were 0.91 (0.89–0.92), 0.83 (0.81–0.86), 0.91 (0.90–0.93), and 0.93 (0.92–0.95), respectively (Figure 2B). In the testing set, the corresponding AUCs and 95% CIs were 0.89 (0.85–0.94), 0.86 (0.82–0.91), 0.88 (0.83–0.93), and 0.90 (0.85–0.94), respectively (Figure 2C).

Figure 2 Establishment and validation of a circulating miRNA-based diagnostic model for breast cancer. (A) Correlation heatmap of the four circulating miRNAs in the training set. (B) AUC forest plot of the four key miRNAs in the training set. (C) AUC forest plot of the four key miRNAs in the testing set. (D) Nomogram for predicting breast cancer. (E-G) ROC curves (left) and calibration curves (right) of the model in the (E) training, (F) testing, and (G) validation datasets. AUC, area under the curve; CI, confidence interval; miRNA, microRNA; ROC, receiver operating characteristic.

These four circulating miRNAs were used to establish a predictive model using logistic regression. The parameters used for model construction are shown in Table S3. The model was visualized using a nomogram for clinical application (Figure 2D). We then tested the efficacy of the nomogram in the validation cohort using a large-scale external circulating miRNA GSE73002 dataset (normal: n=1,500, BC: n=900). This nomogram model achieved high AUCs in the training (0.99, 95% CI: 0.99–1.00), testing (0.95, 95% CI: 0.92–0.98), and validation (0.85, 95% CI: 0.83–0.86) datasets (Figure 2E-2G). In addition, the calibration curve of the model demonstrated close alignment with the diagonal line across all three cohorts, indicating high accuracy (Figure 2E-2G).

Clinical cohort validation of the diagnostic model

To verify these findings in clinical samples, a smaller cohort comprising 102 clinical samples was collected, including sera from 47 healthy individuals and 55 BC patients. The clinicopathological characteristics of the cohort are summarized in Table S4. RT-qPCR was used to examine the relative expression of the four miRNAs in the serum. We observed the differential expression of these miRNAs between the BC patients and healthy individuals (Table 1). Consistent with our earlier findings, miR-139-3p and miR-134-3p were downregulated in the sera of BC patients, while miR-629-3p and miR-191-5p were upregulated (Figure 3A-3D and Table 1). The four miRNAs (miR-139-3p, miR-134-3p, miR-629-3p, and miR-191-5p) had higher AUCs for predicting BC of 0.73, 0.65, 0.75, and 0.82, respectively (Figure 3E) than the clinically routine analyzed tissue polypeptide specific (TPS) antigen, CA15-3, CEA, and CA-125 (which had AUCs of 0.70, 0.68, 0.64, and 0.55, respectively) (Figure 3F).

Figure 3 Validation of the diagnostic model in a clinical cohort. (A-D) Relative expression of (A) miR-139-3p, (B) miR-134-3p, (C) miR-629-3p, and (D) miR-191-5p in the serum of healthy individuals and BC patients. (E) ROC curve analysis of the four key circulating miRNAs in the clinical cohort. (F) ROC curve analysis of the four BC biomarkers in the clinical cohort. (G) ROC curve of the miRNA-based diagnostic model in the clinical cohort. (H) Calibration curve of the miRNA-based diagnostic model in the clinical cohort. **, P<0.01; ***, P<0.001. AUC, area under the curve; BC, breast cancer; CA-125, carbohydrate antigen 125; CA15-3, carbohydrate antigen 15-3; CEA, carcinoembryonic antigen; miRNA, microRNA; ROC, receiver operating characteristic; TPS, tissue polypeptide specific.

The circulating miRNA-based model was thus effective in differentiating between BC and normal sera with an AUC of 0.87 (0.80–0.94) (Figure 3G,3H). Furthermore, this model exhibited robust diagnostic performance across multiple datasets. Specifically, its sensitivity values were 0.966, 0.970, 0.821 and 0.800 in the training set, testing set, validation set and clinical cohort, respectively; the specificity values were 0.974, 0.963, 0.745 and 0.851, respectively; and the accuracy values were 0.970, 0.967, 0.774 and 0.824, respectively (Table 2).

The expression of the four miRNAs in the tissue and prognosis analysis

The levels of the four miRNAs in tumor tissues and paired non-tumor tissues were then evaluated using data from TCGA. The results revealed that the expression trends of these four miRNAs in the paired tissue samples were consistent with their circulating levels. Specifically, miR-139-3p and miR-134-3p were observed to be downregulated, while miR-629-3p and miR-191-5p were observed to be upregulated in the BC tissues (Figure 4A-4D). Among them, miR-139-3p was significantly downregulated in BC (Figure 4A).

Figure 4 Expression of the four miRNAs in paired tissues, and the survival analysis of miR-139-3p. The expression differences in TCGA-BRCA paired samples of (A) miR-139-3p, (B) miR-134-3p, (C) miR-629-3p, and (D) miR-191-5p. (E-H) Overall survival curves according to the miR-139-3p expression levels (E) in the entire METABRIC cohort, (F) among patients who did not receive treatment, (G) among patients who did not receive chemotherapy, and (H) among patients who received chemotherapy. BRCA, Breast Invasive Carcinoma; HR, hazard ratio; miRNA, microRNA; RMP, RNA matched pairs; TCGA, The Cancer Gene Atlas.

We further conducted a prognostic analysis in the METABRIC dataset, which had a large sample size and detailed treatment data. In terms of the entire cohort, we found that patients with low miR-139-3p levels had a shorter overall survival [hazard ratio (HR) =0.76 (95% CI: 0.62–0.93), P=0.007, Figure 4E]. Further, the subgroup survival analysis revealed that there was no significant difference in the patients who received no treatment (P=0.01, Figure 4F) and those who did not receive chemotherapy (P=0.10, Figure 4G); however, among the patients who received chemotherapy, those with low miR-139-3p had a worse prognosis [HR =0.6 (95% CI: 0.44–0.81), P<0.001, Figure 4H]. This was not observed in the subgroup survival analysis of miR-134-3p and miR-629-3p (Figure S3). A survival analysis of miR-191-5p was not conducted due to the lack of data in the METABRIC dataset. We then sought to understand the potential mechanisms by which miR-139-3p exerts its effects in the subgroup of patients who received chemotherapy.

MiR-139-3p increases the sensitivity of BC cells to agents causing DNA damage

To explore the potential mechanisms through which miR-139-3p affects the prognosis of patients who receive chemotherapy, we first performed a pathway analysis. We found that miR-139-3p was involved in multiple DNA damage response (DDR) pathways, while BRCA1, BRCA2, and ATR, as well as ATM signaling, were linked to cancer susceptibility (Figure 5A). Therefore, we hypothesized that miR-139-3p enhances sensitivity to chemotherapy by modulating the DDR pathway.

Figure 5 miR-139-3p increases the sensitivity of BC cells to DNA-damaging agents. (A) Pathway analysis results of miR-139-3p obtained by the tumor-miRNA-pathway. (B,C) IR sensitivity colony-forming assays in (B) MDA-MB-231 and (C) MCF-7 cells transfected with NC or miR-139-3p mimics and stained with crystal violet. 1× magnification. (D,E) MTS assays of MDA-MB-231 cells treated with (D) cisplatin and (E) olaparib. (F,G) MTS assays of MCF-7 cells treated with (F) cisplatin and (G) olaparib. (H) MDA-MB-231 BC cells were exposed to 4 Gy of IR, and the dynamic levels of γH2AX foci in the NC group and the miR-139-3p overexpression group were detected by immunofluorescence assays within 24 hours. The gH2AX foci numbers were counted from a minimum of 150 cells. Scale bar: 10 mm. (I) BC cells were exposed to 4 Gy of IR, and the irradiated cells were harvested at the indicated time points after IR. The dynamic protein levels of γH2AX in the NC and miR-139-3p-overexpression groups were detected by western blot. ns, non-significant; *, P<0.05; **, P<0.01; ***, P<0.001. BC, breast cancer; NC, negative control; gH2AX, phosphorylated H2AX; IR, irradiation.

We overexpressed miR-139-3p in BC cells and evaluated its effect on the sensitivity of DNA-damaging agents, including cisplatin, poly (ADP-ribose) polymerase inhibitor (PARPi), and IR. The colony formation assays indicated that miR-139-3p significantly enhanced the sensitivity of BC cells to IR (Figure 5B,5C). The MTS assays showed that the overexpression of miR-139-3p enhanced the sensitivity of BC cells to cisplatin in the MDA-MB-231 cell line (Figure 5D). Additionally, miR-139-3p overexpression increased the sensitivity of BC cells to olaparib (Figure 5E), and similar results were observed in the MCF-7 cell line (Figure 5F,5G).

As phosphorylated H2AX (gH2AX) serves as a biomarker for DNA damage, we conducted immunofluorescence staining on BC cells after IR and examined gH2AX foci. We found that miR-139-3p overexpression led to a notable elevation of IR-induced γH2AX levels at 4 hours after IR, and a delayed recovery at 8–24 hours post-IR (Figure 5H). Western blotting also revealed an increase in gH2AX levels at 4 hours after IR in the miR-139-3p-transfected BC cells, and inhibited gH2AX recovery (8–24 hours after IR) (Figure 5I). These findings suggest that miR-139-3p increases the sensitivity of BC cells to DNA damage and suppresses damage repair.

MiR-139-3p targets RPA2 to regulate HR repair and sensitivity to DNA damage

To examine the possible mechanism underlying the biological function of miR-139-3p, bioinformatics was used to predict the top 1,000 target genes, among which five were associated with the DNA double-strand break (DSB) response (Figure 6A). We overexpressed miR-139-3p in the BC cells and examined alterations in the mRNA levels of the five target genes. Notably, we observed significant reductions in RPA2 in the MDA-MB-231 and MCF-7 cells (Figure 6B). Western blotting revealed that miR-139-3p modulated RPA2 protein levels (Figure 6C). To confirm that RPA2 was targeted directly by miR-139-3p, the wild-type and mutant 3’UTRs of the RPA2 gene were cloned into the pmirGLO luciferase reporter (Figure S4). The dual-luciferase assays revealed that miR-139-3p mimics lowered the relative luciferase activity of the wild-type RPA2 3’UTR plasmid but not the mutant plasmid, which suggests that RPA2 is targeted directly by miR-139-3p (Figure 6D).

Figure 6 MiR-139-3p targets RPA2 to enhance DNA damage sensitivity in breast cancer cells. (A) Venn diagram of the miR-139-3p target genes and DSB response genes. (B) The relative messenger RNA level of five potential targets in the miR-139-3p in MDA-MB-231 and MCF-7 cells transfected with NC or miR-139-3p mimics. (C) Western blot of RPA2 in the cells transfected with NC or miR-139-3p mimics. (D) The relative Firefly/Renilla luciferase activity in the 293T cells transfected with the luciferase vector containing wt or mut RPA2 3’UTR luciferase plasmids, and NC or miR-139-3p mimics. (E) Stable overexpression of RPA2-3Flag in BC cells. (F,G) IR colony formation ability of the (F) MDA-MB-231 and (G) MCF-7 cells stably overexpressing RPA2 and transfected with miR-139-3p mimics and stained with crystal violet. 1× magnification. (H,I) Cisplatin sensitivity MTS assays of the (H) MDA-MB-231 and (I) MCF-7 cells in different groups. (J) Schematic of HR repair reporter assays (upper). The HR repair efficiency of the miR-139-3p overexpression group and the co-overexpression of the RPA2 group were measured by flow cytometry, with siBRCA1 as a positive control. (K) Schematic of NHEJ reporter assays (upper). The NHEJ efficiency of the miR-139-3p overexpression group and the co-overexpression of the RPA2 group were measured by flow cytometry, with si53BP1 as a positive control. (L) gH2AX foci immunostained for gH2AX (green) and counterstained with DAPI (blue) at 12 hours post-IR in the MDA-MB-231 cells. The gH2AX foci numbers were counted from a minimum of 150 cells. Scale bar: 10 μm. (M) RAD51 foci in the S phase immunostained for RAD51 (red) and counterstained with DAPI (blue) at 12 hours post-IR in the MDA-MB-231 cells. RAD51 foci numbers were counted from a minimum of 150 cells. Scale bar: 10 mm. ns, non-significant; *, P<0.05; **, P<0.01; ***, P<0.001. DAPI, 4’,6-diamidino-2-phenylindole; DSB, double-strand break; GEP, gene editing plasmid; GFP, green fluorescent protein; gH2AX, phosphorylated H2AX; HR, homologous recombination; iGFP, internal GFP reporter; IR, irradiation; mut, mutant type; NC, negative control; NHEJ, non-homologous end joining; wt, wild type; UTR, untranslated region.

As RPA2 is an integral component of the DDR pathway, we investigated if miR-139-3p regulates the DDR by modulating RPA2 expression. BC cells stably overexpressing RPA2 were established and transfected with miR-139-3p. A shift in the position of RPA2 was observed on the blot owing to the triple FLAG (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) tag fused to the C-terminus of the recombinant protein in the expression vector (Figure 6E). The colony formation assays showed that the miR-139-3p mimics enhanced BC cell sensitivity to IR, but the co-overexpression of RPA2 partially rescued the synergistic effect of miR-139-3p and IR (Figure 6F,6G). Similarly, the MTS assays showed that the co-overexpression of RPA2 restored BC cell sensitivity to cisplatin induced by miR-139-3p overexpression (Figure 6H,6I).

Given the crucial role of RPA2 in HR, we examined the effect of miR-139-3p and RPA2 on HR efficiency by flow cytometry. HR reporter assays showed that RPA2 overexpression could partially rescue the HR dysfunction caused by miR-139-3p (Figure 6J). However, our results further indicated that miR-139-3p did not influence the efficiency of NHEJ (Figure 6K). The immunofluorescence of gH2AX showed that relative to the NC-transfected cells, the cells transfected with miR-139-3p mimics exhibited a significant delay of gH2AX foci recovery at 12 hours after IR, and this phenomenon was attenuated in the cells overexpressing RPA2 (Figure 6L). We also examined the loading of RAD51, a key recombinase in the HR. The results revealed that following transfection with the miR-139-3p mimics, the cells showed a notable decrease in RAD51 foci during the S phase, while RPA2 co-overexpression restored the levels of the RAD51 foci (Figure 6M). The above experiments indicate that miR-139-3p suppresses HR and enhances sensitivity to DNA-damaging agents by directly targeting RPA2.

The combination of miR-139-3p and cisplatin considerably suppresses tumor growth

As in vivo verification of the ability of miR-139-3p to regulate cisplatin sensitivity, the MDA-MB-231 BC cell line was used to construct a xenograft nude mouse model, and the animal experiments were designed as shown in Figure 7A. Treatment with cisplatin alone mildly inhibited the growth of tumors, while the use of miR-139-3p agomiR alone did not significantly suppress tumor growth. The combination of miR-139-3p agomiR and cisplatin markedly reduced the growth of BC xenografts compared with either alone (Figure 7B,7C). The immunohistochemistry (IHC) results indicated that the miR-139-3p agomiR markedly decreased the protein levels of RPA2 in vivo (Figure 7D). Additionally, miR-139-3p agomiR and cisplatin alone decreased the levels of Ki-67, a marker of proliferation. However, the combination of miR-139-3p agomiR and cisplatin resulted in a more significant reduction in the Ki-67 levels (Figure 7E).

Figure 7 The efficacy of the combination of miR-139-3p and cisplatin in vivo. (A) Schematic diagram of the construction and treatment of the xenograft model in the BALB/c nude mice. (B) Representative images of nude mice with xenograft tumors in each group. (C) Growth curve of xenograft tumors in nude mice from different groups. (D) IHC staining and analysis of the RPA2 protein in xenograft tumor. 20× magnification. (E) IHC staining and analysis of the Ki-67 protein in xenograft tumors. 1× magnification. *, P<0.05; **, P<0.01; ***, P<0.001. IHC, immunohistochemistry; NC, negative control; PBS, phosphate-buffered saline.

Discussion

Previous studies have shown the importance of circulating miRNAs in the diagnosis of BC. However, investigations into circulating miRNAs have always been based on small cohorts. Although the AUC values reported in previous studies were higher than 0.9, there was no or only small-scale external validation in these studies (27,28). Further, due to the cost of research, many studies have only screened the diagnostic performance of miRNAs associated with BC, and the entire circulating miRNA profile has not been comprehensively explored (22,29,30). A recent multicenter study developed and validated an eight-miRNA signature, which had AUC values of 0.918 and 0.915 in two validation cohorts. However, due to the limitations of RT-qPCR, the study only screened 324 miRNAs, and might have excluded other important diagnostic miRNAs (30).

To address these issues, we screened comprehensive miRNA profiles from multiple large-scale miRNA datasets. We then harnessed the advantages of machine-learning algorithms to handle the complexity of large datasets, enabling the identification of the most generalizable key miRNAs. The ROC curve analysis confirmed the high independent diagnostic efficacy of each miRNA. Conversely, commonly used clinical biomarkers for BC, such as CA15-3, TPS, CEA, and CA-125, had AUCs only slightly higher than 0.5 in the clinical cohort. Previous studies have indicated that these biomarkers lack sensitivity for diagnosing BC (31,32), which is consistent with our findings. Thus, reliable circulating biomarkers urgently need to be established to supplement diagnosis in clinical settings. The high sensitivity and specificity of our model suggest its potential efficacy in improving BC diagnosis, thereby reducing unnecessary diagnostic procedures and mortality rates. Our model maintained its diagnostic efficacy in a prospective clinical research cohort (with a sample size of 102 cases), and its AUC was 0.869. Although the observed AUC (0.869) here is slightly lower than the values in the large discovery dataset—this is an expected weakening phenomenon when transitioning to real-world clinical samples—it is still significantly superior to the conventional serum biomarkers (CA15-3, CEA, TPS, CA-125; AUC ranging from 0.55 to 0.70) measured in the same batch of patients. This relative advantage highlights the potential value of the microRNA markers we proposed in clinical applications.

All four miRNAs used in the model have been documented to be associated with diagnosis, metastasis, and other pathological features in BC. Multiple studies have reported the utility of increased circulating miR-629-3p and miR-191-5p as liquid biopsy biomarkers for BC (33-35). Additionally, decreased circulating miR-139-3p and increased circulating miR-629-3p have been linked to axillary lymph node metastasis in early stage BC (35,36). MiR-139-3p was also found to be a biomarker of BC in tissue samples (37). MiR-134-3p, a tumor suppressor in breast and ovarian cancers, has been reported to have diagnostic potential in BC (38). The identification of these miRNAs in multiple reports indicates their broad applicability and accuracy as potential biomarkers for BC.

The prognostic analysis of the key miRNAs indicated that increased miR-139-3p levels were related to a better prognosis in the entire cohort. This survival advantage was only observed in the chemotherapy subgroup and was not observed in the untreated or treated without chemotherapy subgroups. This phenomenon sparked our interest; therefore, we conducted an analysis of the miR-139-3p-related pathways and found that it was closely related to DNA damage repair. Although it is known to be a tumor suppressor in various cancer types, this is the first investigation of the function of miR-139-3p in DNA damage repair in BC. Collectively, the finding that high miR-139-3p levels were associated with better prognosis specifically in chemotherapy-treated patients, coupled with our subsequent experimental evidence, suggests a potential link between this circulating biomarker and tumor treatment response. This implies that measuring serum miR-139-3p could serve a dual purpose: as a diagnostic biomarker and as a predictive companion biomarker to help identify patients more likely to benefit from chemotherapy or PARPi.

We hypothesized that miR-139-3p affects the prognosis of patients with BC who undergo chemotherapy by influencing the DDR. Our subsequent cellular experiments supported this hypothesis, demonstrating that miR-139-3p markedly increased the sensitivity of BC cells to IR, cisplatin, and PARPi. In the animal experiments, miR-139-3p was also observed to potentiate the anti-tumor effect of cisplatin. Additionally, the DNA damage marker gH2AX was significantly increased and its recovery was delayed after IR in the BC cells in which miR-139-3p was overexpressed. This indicates that in addition to increasing DNA damage, miR-139-3p also delays the repair process following DNA damage.

Due to the significant effects of DSB on the genetic information and survival capacity of cells, the accuracy and efficiency of the repair process are crucial for maintaining cell function and survival. Thus, we screened for the target genes of miR-139-3p related to the DSB response, and we identified RPA2 as a significant target through prediction and validation. As an important part of the RPA complex, RPA2 plays a crucial role in HR. RPA binds to single-stranded DNA generated during the process, preventing it from being cleaved by nucleases. Subsequently, RPA is replaced by the recombinase RAD51, ensuring the proper progression of HR (39,40). A previous study found that high levels of RPA were associated with increased treatment resistance (41), and targeting RPA enhanced the synthetic lethality of PARPi and BRCA1 defects (42). The downregulation of HR reporter efficiency and the reduction in RAD51 foci indicate that miR-139-3p induces HR defects in HR-proficient BC cells. HR-deficient cells may use error-prone pathways for DNA repair pathways, destabilizing the genome and increasing sensitivity to DNA-damaging agents (43). Targeting RPA2 may be an important reason why the overexpression of miR-139-3p in BC exhibits high sensitivity to cisplatin and olaparib. We directly correlated the clinical observation results with our functional data. The overexpression of miR-139-3p enhanced the sensitivity of breast cancer cells to cisplatin, PARPi and radiation. This was achieved by directly targeting RPA2 to inhibit HR repair. This led to persistent DNA damage (increased γH2AX levels) and HR defects (reduced RAD51 subunit aggregation). Therefore, tumors with higher intrinsic miR-139-3p levels may be more susceptible to the cytotoxic effects of chemotherapy, resulting in a better therapeutic response and ultimately improving survival rates.

However, most miRNAs use multiple targets to simultaneously regulate a particular pathway. For example, another mature sequence of miR-139, miR-139-5p, has been shown to regulate DNA damage repair by targeting TOP2A, POLQ, and RAD54L (44). Our results also showed that the overexpression of RPA2 only partially restored the effects of miR-139-3p on HR and DNA damage. Thus, miR-139-3p may have additional target effects that contribute to the DDR. MiR-139-3p has been shown to target CHK1 in lung cancer to regulate DNA damage (45). We confirmed this in the BC cell lines and found that the downregulation of CHK1 by miR-139-3p was extremely mild (data not provided), while the downregulation of RPA2 was highly significant. Therefore, other targets of miR-139-3p in BC should be explored.

Other inhibitors targeting key proteins in the HR pathways, such as ATR, have shown promising results in clinical trials (46). The development of these inhibitors is important for reversing chemotherapy and PARPi resistance, as well as for expanding their applicability. RPA2 needs further verification in animal studies and clinical trials to assess its feasibility and safety as a therapeutic target. As regulatory factors in these processes, miRNAs have opened new avenues for treatment. Evidence suggests that miRNAs can be used as single agents or in combination with synthetic lethality-based therapies for cancer (47-50). Thus, targeting the miR-139-3p/RPA2 axis has significant translational potential, as it can induce “BRCAness” in BRCA-proficient cells to increase their sensitivity to DNA-damaging agents and expand the application range of PARPi. Compared with the existing clinical standards, this model outperforms serum markers (such as CA15-3/CEA) in terms of diagnostic accuracy, and can provide molecular-level dynamic information that is difficult to achieve through imaging and tissue biopsy in a minimally invasive and repeatable liquid biopsy form. More importantly, its key component miR-139-3p has a mechanism association with treatment sensitivity, making this model have the potential for both diagnostic assistance and efficacy prediction, and can serve as a powerful supplementary tool for optimizing the existing clinical workflow. After further verification, our four microRNA markers are expected to serve as a non-invasive blood testing method, complementing existing screening methods and capable of dynamically monitoring treatment effects. The miR-139-3p/RPA2 axis provides a biomarker candidate with predictive value, used to identify patients who may benefit from DNA damage therapy (chemotherapy/PARPi). Additionally, this axis itself is a targetable pathway, which can be used to develop new therapeutic drugs (such as miR mimics or RPA2 inhibitors) to overcome drug resistance. To advance the clinical translation of this model, the subsequent key steps include conducting large-scale multicenter validation and detection standardization, in-depth exploration of other action targets of miR-139-3p and its delivery strategies, and ultimately establishing a direct correlation between circulating miR-139-3p levels and tumor HR status and patient prognosis through clinical research.

Limitations

This study had some limitations. First, the screening and validation were conducted exclusively within an Asian population, necessitating further verification of its extrapolation. Second, we only investigated the differential miRNA profiles between BC and non-cancerous individuals; a comparison of the miRNA profiles of other cancer types was not conducted. Third, the sample size of the prospective clinical cohort in this study (n=102) is relatively small, which may limit the statistical power of the validation.

Conclusions

We established and validated a robust and clinically applicable diagnostic model for BC based on four circulating miRNAs. Among them, miR-139-3p enhanced the sensitivity of BC to DNA-damaging agents by targeting RPA2 and represents a potential target in the DDR pathway.

Acknowledgments

The authors would like to thank Professor Lei Shi from Tianjin Medical University for providing the experimental platform and partial technical guidance for this experiment. They would also like to thank Editage (www.editage.cn) for the English language editing.

Footnote

Reporting Checklist: The authors have completed the TRIPOD, ARRIVE and MDAR reporting checklists. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2592/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2592/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2592/prf

Funding: This work was supported by the National Natural Science Fund (Nos. 82203407, 82372699, and 82403769).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2592/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Ethical approval was obtained from the Tianjin Medical University Cancer Institute and Hospital Ethical Review Board (No. bc2021276). Written informed consent was obtained from each patient. The animal experiments were approved by the Laboratory Animal Management Committee of Tianjin Medical University Cancer Institute and Hospital (No. PMIF-2021082). The standard operating procedures involving animals were conducted in accordance with the Institutional Animal Care and Use Committee Guidebook.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Obeagu EI, Obeagu GU. Breast cancer: A review of risk factors and diagnosis. Medicine (Baltimore) 2024;103:e36905. [Crossref] [PubMed]
3. Ahmad J, Akram S, Jaffar A, et al. Deep learning empowered breast cancer diagnosis: Advancements in detection and classification. PLoS One 2024;19:e0304757. [Crossref] [PubMed]
4. Darbandi MR, Darbandi M, Darbandi S, et al. Artificial intelligence breakthroughs in pioneering early diagnosis and precision treatment of breast cancer: A multimethod study. Eur J Cancer 2024;209:114227. [Crossref] [PubMed]
5. Fina E, Vitale E, De Summa S, et al. Liquid biopsy for guiding breast cancer immunotherapy. Immunotherapy 2025;17:369-83. [Crossref] [PubMed]
6. Venkataraman J, Crook T, Mokbel K. Liquid biopsy in breast cancer: a practical guide for surgeons. Gland Surg 2025;14:754-60. [Crossref] [PubMed]
7. Zou P, Li K, Qian Y, et al. The predictive prognostic value of serum CEA, CA153, HE4 combined with a variety of immune inflammatory indicators in breast cancer. Afr Health Sci 2024;24:224-32. [Crossref] [PubMed]
8. Park BW, Oh JW, Kim JH, et al. Preoperative CA 15-3 and CEA serum levels as predictor for breast cancer outcomes. Ann Oncol 2008;19:675-81. [Crossref] [PubMed]
9. Zhao X, Zhuang Y, Wang B, et al. The miR-34a-5p-c-MYC-CHK1/CHK2 Axis Counteracts Cancer Stem Cell-Like Properties and Enhances Radiosensitivity in Hepatocellular Cancer Through Repression of the DNA Damage Response. Radiat Res 2023;199:48-60. [Crossref] [PubMed]
10. Bertoli G, Cava C, Castiglioni I. MicroRNAs: New Biomarkers for Diagnosis, Prognosis, Therapy Prediction and Therapeutic Tools for Breast Cancer. Theranostics 2015;5:1122-43. [Crossref] [PubMed]
11. Petri BJ, Klinge CM. Regulation of breast cancer metastasis signaling by miRNAs. Cancer Metastasis Rev 2020;39:837-86. [Crossref] [PubMed]
12. Alizadeh M, Ghasemi H, Bazhan D, et al. MicroRNAs in disease States. Clin Chim Acta 2025;569:120187. [Crossref] [PubMed]
13. Mitchell PS, Parkin RK, Kroh EM, et al. Circulating microRNAs as stable blood-based markers for cancer detection. Proc Natl Acad Sci U S A 2008;105:10513-8. [Crossref] [PubMed]
14. Vickers KC, Palmisano BT, Shoucri BM, et al. MicroRNAs are transported in plasma and delivered to recipient cells by high-density lipoproteins. Nat Cell Biol 2011;13:423-33. [Crossref] [PubMed]
15. Hassanin AAI, Ramos KS. Circulating Exosomal miRNA Profiles in Non-Small Cell Lung Cancers. Cells 2024;13:1562. [Crossref] [PubMed]
16. Bayraktar R, Fontana B, Calin GA, et al. miRNA Biology in Chronic Lymphocytic Leukemia. Semin Hematol 2024;61:181-93. [Crossref] [PubMed]
17. Afzal M, Greco F, Quinzi F, et al. The Effect of Physical Activity/Exercise on miRNA Expression and Function in Non-Communicable Diseases-A Systematic Review. Int J Mol Sci 2024;25:6813. [Crossref] [PubMed]
18. Sun Q, Lei X, Yang X. CircRNAs as upstream regulators of miRNA//HMGA2 axis in human cancer. Pharmacol Ther 2024;263:108711. [Crossref] [PubMed]
19. Zhao Q, Shen L, Lü J, et al. A circulating miR-19b-based model in diagnosis of human breast cancer. Front Mol Biosci 2022;9:980841. [Crossref] [PubMed]
20. Kim J, Park S, Hwang D, et al. Diagnostic Value of Circulating miR-202 in Early-Stage Breast Cancer in South Korea. Medicina (Kaunas) 2020;56:340. [Crossref] [PubMed]
21. Itani MM, Nassar FJ, Tfayli AH, et al. A Signature of Four Circulating microRNAs as Potential Biomarkers for Diagnosing Early-Stage Breast Cancer. Int J Mol Sci 2021;22:6121. [Crossref] [PubMed]
22. Adam-Artigues A, Garrido-Cano I, Carbonell-Asins JA, et al. Identification of a Two-MicroRNA Signature in Plasma as a Novel Biomarker for Very Early Diagnosis of Breast Cancer. Cancers (Basel) 2021;13:2848. [Crossref] [PubMed]
23. Feliciano A, González L, Garcia-Mayea Y, et al. Five microRNAs in Serum Are Able to Differentiate Breast Cancer Patients From Healthy Individuals. Front Oncol 2020;10:586268. [Crossref] [PubMed]
24. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
25. Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77. [Crossref] [PubMed]
26. Seluanov A, Mao Z, Gorbunova V. Analysis of DNA double-strand break (DSB) repair in mammalian cells. J Vis Exp 2010;2002. [Crossref] [PubMed]
27. Chan M, Liaw CS, Ji SM, et al. Identification of circulating microRNA signatures for breast cancer detection. Clin Cancer Res 2013;19:4477-87. [Crossref] [PubMed]
28. Shimomura A, Shiino S, Kawauchi J, et al. Novel combination of serum microRNA for detecting breast cancer in the early stage. Cancer Sci 2016;107:326-34. [Crossref] [PubMed]
29. Swellam M, Ramadan A, El-Hussieny EA, et al. Clinical significance of blood-based miRNAs as diagnostic and prognostic nucleic acid markers in breast cancer: Comparative to conventional tumor markers. J Cell Biochem 2019;120:12321-30. [Crossref] [PubMed]
30. Zou R, Loke SY, Tang YC, et al. Development and validation of a circulating microRNA panel for the early detection of breast cancer. Br J Cancer 2022;126:472-81. [Crossref] [PubMed]
31. Duffy MJ, Evoy D, McDermott EW. CA 15-3: uses and limitation as a biomarker for breast cancer. Clin Chim Acta 2010;411:1869-74. [Crossref] [PubMed]
32. Zhang J, Wei Q, Dong D, et al. The role of TPS, CA125, CA15-3 and CEA in prediction of distant metastasis of breast cancer. Clin Chim Acta 2021;523:19-25. [Crossref] [PubMed]
33. Wang S, Li L, Yang M, et al. Identification of Three Circulating MicroRNAs in Plasma as Clinical Biomarkers for Breast Cancer Detection. J Clin Med 2022;12:322. [Crossref] [PubMed]
34. Ashirbekov Y, Abaildayev A, Omarbayeva N, et al. Combination of circulating miR-145-5p/miR-191-5p as biomarker for breast cancer detection. PeerJ 2020;8:e10494. [Crossref] [PubMed]
35. Shiino S, Matsuzaki J, Shimomura A, et al. Serum miRNA-based Prediction of Axillary Lymph Node Metastasis in Breast Cancer. Clin Cancer Res 2019;25:1817-27. [Crossref] [PubMed]
36. Escuin D, López-Vilaró L, Mora J, et al. Circulating microRNAs in Early Breast Cancer Patients and Its Association With Lymph Node Metastases. Front Oncol 2021;11:627811. [Crossref] [PubMed]
37. Zhang W, Xu J, Wang K, et al. miR 139 3p suppresses the invasion and migration properties of breast cancer cells by targeting RAB1A. Oncol Rep 2019;42:1699-708. [Crossref] [PubMed]
38. Chao TY, Kordaß T, Osen W, et al. SOX9 is a target of miR-134-3p and miR-224-3p in breast cancer cell lines. Mol Cell Biochem 2023;478:305-15. [Crossref] [PubMed]
39. Chen H, Lisby M, Symington LS. RPA coordinates DNA end resection and prevents formation of DNA hairpins. Mol Cell 2013;50:589-600. [Crossref] [PubMed]
40. Huang RX, Zhou PK. DNA damage response signaling pathways and targets for radiotherapy sensitization in cancer. Signal Transduct Target Ther 2020;5:60. [Crossref] [PubMed]
41. Toledo LI, Altmeyer M, Rask MB, et al. ATR prohibits replication catastrophe by preventing global exhaustion of RPA. Cell 2013;155:1088-103. [Crossref] [PubMed]
42. Cong K, Peng M, Kousholt AN, et al. Replication gaps are a key determinant of PARP inhibitor synthetic lethality with BRCA deficiency. Mol Cell 2021;81:3128-3144.e7. [Crossref] [PubMed]
43. Byrum AK, Vindigni A, Mosammaparast N. Defining and Modulating 'BRCAness'. Trends Cell Biol 2019;29:740-51. [Crossref] [PubMed]
44. Pajic M, Froio D, Daly S, et al. miR-139-5p Modulates Radiotherapy Resistance in Breast Cancer by Repressing Multiple Gene Networks of DNA Repair and ROS Defense. Cancer Res 2018;78:501-15. [Crossref] [PubMed]
45. Zheng X, Zhang Y, Wu S, et al. MiR-139-3p Targets CHEK1 Modulating DNA Repair and Cell Viability in Lung Squamous Carcinoma Cells. Mol Biotechnol 2022;64:832-40. [Crossref] [PubMed]
46. Yap TA, O'Carrigan B, Penney MS, et al. Phase I Trial of First-in-Class ATR Inhibitor M6620 (VX-970) as Monotherapy or in Combination With Carboplatin in Patients With Advanced Solid Tumors. J Clin Oncol 2020;38:3195-204. [Crossref] [PubMed]
47. Mansour WY, Bogdanova NV, Kasten-Pisula U, et al. Aberrant overexpression of miR-421 downregulates ATM and leads to a pronounced DSB repair defect and clinical hypersensitivity in SKX squamous cell carcinoma. Radiother Oncol 2013;106:147-54. [Crossref] [PubMed]
48. Yin Y, Xu L, Chang Y, et al. N-Myc promotes therapeutic resistance development of neuroendocrine prostate cancer by differentially regulating miR-421/ATM pathway. Mol Cancer 2019;18:11. [Crossref] [PubMed]
49. Petrovic N, Davidovic R, Bajic V, et al. MicroRNA in breast cancer: The association with BRCA1/2. Cancer Biomark 2017;19:119-28. [Crossref] [PubMed]
50. Ge Y, Yan Z, Du M, et al. A summary of our serial study: mechanism of invasion and metastasis in nasopharyngeal carcinoma. Holist Integ Oncol 2023;2:1-5.

(English Language Editor: L. Huleatt)