The RNA-seq data from normal mammary tissues were acquired from the GTEx (Consortium, 2013), while RNA-seq data from breast cancer and adjacent (adj.) normal tissues were obtained from TCGA-BRCA. Data acquisition was performed using the UCSC Xena (http://xena.ucsc.edu/) (Goldman et al., 2020). PAM50 subtypes (determined by RNA-seq data), tumor gene-level copy number alterations (CNAs, delete germline copy number variation), survival data, including progression-free survival (PFS), and disease-specific survival (DSS), were also extracted.

Kaplan-Meier Plotter (http://kmplot.com) was used to check recurrence-free survival (RFS) data integrated from microarray data of 1809 breast cancer patients (Gyorffy et al., 2010). Kaplan-Meier survival curves were generated by the median or optimal cutoff of RPA3 expression.

The association between RPA3 expression and the functional states of breast cancer cells at the single-cell level was assessed using the CancerSEA platform (http://biocc.hrbmu.edu.cn/CancerSEA/home.jsp), which provides an analytic strategy to determine the correlation between gene expression and 14 functional states of cancer cells (Yuan et al., 2019). These states were estimated according to the gene-expression profile at the single-cell level, using the signatures from Gene Ontology, MSigDB, Cyclebase, HCMDB and StemMapper. The state activity scores were calculated using the Gene Set Variation Analysis (GSVA) (Yuan et al., 2019). The tumor cell data from basal-like (N = 130) and HER2⁺ (N = 89) cases in one previous single-cell RNA-seq dataset (GSE75688) (Chung et al., 2017) were retrieved for analysis.

TF gene list was collected from the JASPAR database (http://jaspar.genereg.net/) (Stormo, 2013). The promoter segment of RPA3 was acquired by checking the promoter clone of human RPA3 (#HPRM44615, Genome = hg38; chr7-:7719949-7718330; TSS = 7718607) in Genecopoeia. The promoter segment was scanned in JASPAR to identify potential YY1 binding sites by setting the relative profile score threshold to 90%.

Basal-like representative MDA-MB-231 and HER2⁺ representative SK-BR-3 human breast cancer cell lines were purchased from the Cell Resource Center, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (Beijing, China). These two cell lines were maintained in Dulbecco’s modified eagle medium (DMEM) (Lonza, Walkersville, MD, USA) at 37°C in a 90% humidified incubator, with 5% CO₂.

Small interfering RNA (siRNA) and scramble controls were synthesized by General Biosystem (Chuzhou, Anhui, China), with the following sequences: RPA3 siRNA (#1, 5’-CCGGCAUGCUAGCUCAAUUTT-3’; #2, 5’-GCAUGCUAGCUCAAUUCAUTT-3’, #3, 5’-GCCACCAUCUUGUGUACAUTT-3’) and scramble: 5’-GCUAUGCUCGCAUAUCACUTT-3’; YY1 siRNA (#1, 5’-CCAAACAACUGGCAGAAUUTT-3’; #2, 5’-GCUCCAAGAACAAUAGCUUTT-3’, #3, 5’-CCCAAACAACUGGCAGAAUTT-3’) and scramble: 5’-CACACGACUGACAAGCAUATT-3’. Traditional forward transfection was conducted. Briefly, cells were seeded into a six-well plate at a density of 1 × 10⁶ cells per well for 24 h at 37°C. Then, the cells were transfected with the siRNAs (50 nM), using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions.

Total RNA was extracted from cells using the High Pure RNA Isolation Kit (Roche Applied Science, Mannheim, Germany). Then, total RNA was reversely transcribed into cDNA using the first-strand cDNA synthesis kit (Roche Applied Science) according to the manufacturer’s instructions. Real-time qPCR was then performed as described previously (Tuo et al., 2015). The sense and antisense primers for gene amplification were: RPA3, 5’-AAGCCTGTCTGCTTCGTAGGGA-3’ and 5’-CGGTTACTCTTCCAACCACTTCC-3’, YY1, 5’-GGAGGAATACCTGGCATTGACC-3’ and 5’-CCCTGAACATCTTTGTGCAGCC-3’. The relative expression level of mRNA was evaluated by using the 2^−ΔΔCt method. Gene expression levels were normalized to ACTB.

Total proteins were extracted using RIPA lysis buffer (Beyotime, China) with protease inhibitors cocktail (Roche Diagnostics, Basel, Switzerland). Then, equal amounts of protein lysates were run on 10% SDS-PAGE, and the separated bands were transferred to polyvinylidene fluoride membrane (PVDF; EMD Millipore, Billerica, MA, USA). The membranes were blocked with 5% non-fat dried milk for 1 h at room temperature, and then were incubated with primary antibodies against RPA3 (1:1000, ab97436, Abcam, Cambridge, MA, USA), YY1 (1:5000, ab245365, Abcam), γ-H2AX (1:5000, ab11174, Abcam) and β-Actin (1:5000, ab179467, Abcam) overnight at 4°C, followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibody for 1 h at room temperature. The protein bands were visualized using ECL detection reagent (Millipore, Billerica, MA, USA).

Clonogenic survival assay was conducted following one previous protocol (Franken et al., 2006). Cells transfected with RPA3 siRNA or scrambled control were seeded into 6-well plates. After adhesion, they were irradiated at defined doses (0, 2, 4, 6, or 8 Gy), using a Rad Source R2000 X-ray irradiator (1.1 Gy/min., 160 kV, 25 mA, 0.3 mm copper filters, Rad Source Tech, Suwanee, GA, USA). After 7–14 days of incubation, the cultures were fixed with 100% methanol and then stained with 1% crystal violet. Colonies containing >50 cells were counted by microscopic inspection. The plating efficiency (PE) of un-irradiated cells (0 Gy) was calculated by the formula: PE = number of colonies counted/number of cells plated. The surviving fraction (SF) of the irradiated cells was calculated using the formula: SF = number of colonies formed after treatment/number of cells seeded × PE. A linear-quadratic model was utilized to generate survival curves using the following equation (Bodgi and Foray, 2016): Y = e^{(−1 × (A×X+B×X2))}, in which Y is the fraction survival, and X is the dose. A equals (−1)-times the initial slope, and the initial value of B equals (−0.1)-times the initial slope.

MDA-MB-231 and SK-BR-3 cells with or without inhibition of endogenous RPA3 were seeded in 24-well plates and exposed to 6 Gy of irradiation. 24 h later, the cells were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 and blocked using 10% goat serum. Then, the cells were incubated with the primary antibody against γ-H2AX (ab11174, Abcam). After that, the cells were incubated with a secondary antibody conjugated to fluorescein isothiocyanate. DAPI was used for nuclear staining. γ-H2AX foci were visualized under a fluorescence microscope (Olympus IX71, Tokyo, Japan). Five random fields were examined to estimate the number of foci per cell for each coverslip.

ChIP was performed using the Chromatin Immunoprecipitation Kit (17-295, Merck Millipore, Boston, MA, USA) according to the recommended protocol. Briefly, formaldehyde was used to cross-link the proteins to the DNA. The lysates of MDA-MB-231 and SK-BR-3 cells were sonicated to shear DNA to an average fragment size of 200–1000 bp. Then, samples were pre-cleaned with Protein A Agarose/Salmon Sperm DNA (50% Slurry) (Catalog #16-157C) and subsequently incubated with anti-YY1 (ab245365, Abcam) or IgG antibodies overnight at 4°C with rotation. Immunoprecipitated DNA was collected by adding the previously mentioned Protein A beads. Then, the samples were washed, and DNA levels were measured by qPCR. Three primer sets were designed (Tab. 1), among which two sets covered potential YY1 binding sites (−998 ~ −882/−884 ~ −808) (primer set 1 and 2, Tab. 1) and one set of primers not covering YY1 binding site (−22 ~ +73) (set 3, Tab. 1).

Full length and truncated 5’ flanking sequences of RPA3 promoter, including −900/+277, −800/+277, and −150/+277, were cloned into the pGL3-basic plasmid (Promega, Madison, WI, USA). MDA-MB-231 cells were seeded in 24-well plates (2 × 10⁵ cells per well). 24 h later, the cells were transfected with either 1 μg recombinant vectors with different fragments of RPA3 promoter together with 50 nM YY1 siRNA or scramble control, using Lipofectamine 2000 (Invitrogen). 0.05 μg of the pRL-CMV vector was co-transfected to normalize the transfection efficiency. 24 h later, the cells were lysed for measuring luciferase activity, using a dual-specific luciferase assay kit (#E1910, Promega).

Welch’s unequal variances t-test was performed for two-group comparison. A Log-rank test was performed for comparing Kaplan-Meier survival curves. Pearson’s correlation r was calculated for correlation assessment. p < 0.05 was considered statistically significant.

RNA-seq data from GTEx-normal mammary tissue and TCGA-BRCA showed that normal mammary tissue owned the lowest RPA3 expression (Fig. 1A). Tumor adj. normal and all PAM50 subtypes of breast tumors presented elevated RPA3 expression (Fig. 1A). IHC staining in the HPA confirmed RPA3 expression at the protein level in breast cancer tissues (Fig. 1B). Then, we assessed the survival difference between patients with high (top 50%) and low (bottom 50%) RPA3 expression in luminal A, luminal B, HER2⁺, and basal-like subgroups, respectively, using survival data from TCGA-BRCA. The log-rank test indicated that in patients with HER2⁺ tumors, the high RPA3 expression group had significantly worse PFS and DSS (p < 0.05, Figs. 1E and 1I). The high RPA3 expression basal-like tumor group was also associated with significantly shorter PFS (p = 0.024) and tended to have worse DSS (p = 0.057) (Figs. 1F and 1J). However, no survival difference was observed in luminal A and B cases by median RPA3 stratification (Figs. 1C, 1D, 1G and 1H).

To validate the survival difference, we also checked survival data in the Kaplan-Meier Plotter. Under both median and optimal RPA3 expression cutoff models, high RPA3 expression HER2⁺ cases had significantly worse RFS (Suppl. Figs. 1A and 1B). In basal-like cases, high RPA3 expression-associated worse RFS was confirmed under the optimal cutoff model (Suppl. Fig. 1D), but not in the median expression model (Suppl. Fig. 1C).

Using cellular functional states assessed by the CancerSea, we analyzed the correlation between RPA3 expression and 14 functional states of basal-like (N = 89) and HER2⁺ tumor cells (N = 130) in GSE75688. Correlation analysis indicated that RPA3 expression was positively correlated with cell cycle progression, DNA damage, and DNA repair in both basal-like and HER2⁺ tumor cells (Pearson’s r ≥ 0.2, Fig. 2A).

MDA-MB-231 and SK-BR-3 cells were transfected with RPA3 siRNA (Figs. 2B and 2C). Compared with the scramble group, the survival fractions of siRPA3 transfected MDA-MB-231 (Fig. 2D) and SK-BR-3 (Fig. 2E) cells were dramatically decreased after irradiation. Besides, siRPA3 treatment also increased the number of γ-H2AX foci (Figs. 2F–2H) and γ-H2AX expression 24 h after irradiation in both MDA-MB-231 and SK-BR-3 cells (Figs. 2I and 2J). These results suggested that RPA3 inhibition enhanced the radiosensitivity of basal-like and HER2⁺ breast cancer cells.

To explore the mechanisms underlying RPA3 dysregulation, we assessed the correlation between the expression of transcriptional factors (TFs) in the JASPAR database (N = 669) and RPA3 in basal-like and HER2⁺ tumors, respectively (Fig. 3A). Correlation analysis (Suppl. Tab. 1) indicated that among the 669 TFs, only YY1 was moderately and positively correlated (|Pearson’s r| ≥ 0.4) with RPA3 expression in both basal-like and HER2⁺ tumor tissue (Suppl. Tab. 1, Figs. 3B and 3C). Via scanning the promoter sequence of RPA3, we found five high potential YY1 binding sites (Fig. 3D). YY1 depletion in MDA-MB-231 and SK-BR-3 cells resulted in significantly decreased RPA3 expression at the mRNA and protein levels (Figs. 3E–3G). Concerning CHIP-qPCR findings, the two amplicons covering YY1 binding sites, but not of the amplicon without YY1 bind site, were significantly enriched upon anti-YY1 immunoprecipitation in both breast cancer cells (Figs. 3H–3I). Dual-luciferase reporter assay showed that the luciferase construct with the integrated promoter sequence had the strongest luciferase activity (Fig. 3J). Truncating the binding sites significantly reduced the luciferase activity (Fig. 3J). YY1 inhibition also substantially decreased the intensity of luciferase expression (Fig. 3J).

RPA3/YY1 expression and their CNAs in basal-like and HER2⁺ breast cancer cases in TCGA are shown in Fig. 4A. Correlation analysis indicated that in basal-like cases, the expression of RPA3 and YY1 was moderately correlated (Pearson's r ≥ 0.4) with their gene-level copy number (Figs. 4B and 4D). In HER2⁺ cases, a moderate positive correlation was observed between YY1 expression and its copy number (Fig. 4E). In comparison, no significant correlation was found between RPA3 expression and its copy number (Fig. 4C).