Animal experiments

Female NOD/SCID mice (4–5 weeks old, weighing 18–22 g, Hunan SJA Laboratory Animal Co., Ltd, Changsha, China) were housed for 1 week in specific pathogen-free environment under constant humidity (45–50%) and temperature (25–27 °C) with 12-h day/night cycle daily. The mice were fasted for 12 h before drug administration and were allowed to eat and drink freely at other time. Animal experimental operations were ratified by the animal ethics committee of Shenzhen Beike Biotechnology Research Institute.

MDA-MB-231 cells (5 × 106/0.2 mL) were injected in situ into the inguinal fat pad of mice (n = 8) to construct a mouse model of TNBC. The mice were sacrificed 6 weeks after injection, while the TNBC tissues were dissected and removed, and mouse normal (n = 8) breast tissues were taken as control. Tissues were preserved at − 80 °C for RNA-Seq.

In the subsequent animal experiments, mice were randomly treated with lentiviral empty vector, lentivirus overexpressing Fbxo22 (Flag-Fbxo22), and/or KDM5A lentivirus (HA-KDM5A), with 8 mice in each treatment. Lentiviral-infected MDA-MB-231 cells (5 × 106/0.2 mL) were injected in situ into inguinal fat pads of mice. After 1 week, the tumor width and length of mice under different treatment were checked once a week, followed by calculation of tumor volume. After 7 weeks, the mice were killed and the tumor tissues were excised, weighed, and photographed with a camera.

For the lung metastasis model, cells (2 × 106/0.2 mL) were injected into the tail vein of the mice. Mice were killed 6 weeks after injection and lung tissues were removed by autopsy and metastatic nodules were counted by microscopy. The presence of micrometastases in the lung tissues was examined by H&E staining.

RNA-Seq

Mammary tissues from TNBC mice (n = 3) and normal mice (n= 3) were collected for RNA-Seq (Illumina HiSeq) at the UW Cystic Fibrosis Genomics Core Center (Miyazaki et al. 2019). Sequencing library was generated using the NEBNext® Ultra™ RNA Library Prep Kit for Illumina® (NEB, Beverly, MA). Quality control was then completed employing RNA-SeQC v1.1.8 and the HTSeq-counts v0.7.2 was applied for counting. The R language limma package was utilized to identify the differentially expressed genes between TNBC and controls with |logFoldChange|> 1 andp-value < 0.05 as cutoff values.

Bioinformatics analysis

Correlation between genes and prognosis of TNBC patients was analyzed by ExSurv based on TCGA database breast cancer samples (with p < 0.05 as screening criteria). ROC curves were obtained by ROC Plotter based on TNBC patient prognostic data to evaluate key factors and their efficacy of prognosis in TNBC patients. Factors were considered to have good predictive performance with AUC > 0.5 and p < 0.05. The ubiquitination loci of the factors were obtained via Ubibrowser. The H3K4me3 methylation sites of the factors were obtained via UCSC browser. Pearson correlation coefficients were measured using the R package psych.

Cell culture

The human breast epithelial cell line MCF-10A (CRL-10317); the luminal-type breast cancer cell lines MCF-7 (HTB-22), T47D (HTB-133), and ZR-75–1 (CRL-1500); and the TNBC cell lines MDA-MB-231 (CRM-HTB-26), MDA-MB-468 (HTB-132), Hs578T (HTB-126), and HEK293T (CRL-3216) were all procured from ATCC (Manassas, VA). MCF-10A, MCF-7, T47D, ZR-75–1, Hs578T, and HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium appended to 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin in a 37 °C thermostatic incubator (BB15, Thermo, Madison, MA) with 5% CO2, while MDA-MB-231 and MDA-MB-468 cells were cultured in L-15 medium (Gibco) appended to 10% FBS in a 37 °C thermostatic incubator without CO2.

Lentiviral infection

Lentiviral vectors overexpressing Fbxo22 and KDM5A were prepared as previously described (Sun et al. 2018). The lentiviral packaging vectors sh-Fbxo22 (5′-GGAATTGTAGTGACTCCAATG-3′) and sh-NC were procured from GenePharma (Shanghai, China), while the sh-p16 (5′-CCCTAAGCGCACATTCATGT-3′, TRCN0000010483) and sh-KDM5A (5′-CCAGACTTACAGGGACACTTA-3′, TRCN0000014629) were procured from Sigma-Aldrich (St.Louis, MO). The lentivirus titer was 109 TU/mL.

Cell viability, clone formation, cell migration, and invasion assays

Breast cancer cells were seeded in 96-well plates (8 × 103cells per well) and cell viability was tested by CCK-8 (Sigma-Aldrich). For clone formation assay, cells were seeded into 6-well plates (2000 cells/well) evenly. The cells were cultured for a fortnight under normal growth conditions; the dispersed single cells may divide and proliferate and eventually form aggregated cell clusters, each of which was a “colony” (Kabakov and Gabai 2018). Cells were fixed in 4% paraformaldehyde and stained with 0.5% crystal violet staining solution (0.5% w/v, Solarbio, Beijing, China) for 15 min, followed by counting using an inverted microscope. Migration and invasion (with 50 μL of Matrigel, 354,234, BD Biosciences, San Diego, CA) assays were completed in differently treated breast cancer cells (Torres et al. 2019).

Immunofluorescence staining

Cells were seeded into fluorescent culture dishes and incubated overnight to allow adhesion. The irradiation beam was used to locally induce DNA damage within the cells by irradiating the cells at 30 Gy at 4 Gy/min with a γ-ray emitter (cobalt 60), and the medium was renewed immediately after irradiation. Cells were then fixed in 4% formaldehyde for 10 min and then permeabilized with PBS containing 0.2% Triton X-100 for 3 min. Cells were blocked with 10% goat serum for 40 min and incubated with the primary antibodies of rabbit anti-γH2AX (1:50, AP0687, ABclonal, Boston, MA) overnight at 4 °C. Subsequently, cells were then incubated with fluorescent secondary antibodies of Cy3 Goat Anti-Rabbit IgG (H + L) (1:50, AS007, ABclonal) at ambient temperature for 2 h. The nuclei were stained with 4′,6-diamidino-2-phenylindole (10,236,276,001, Roche, Shanghai, China) and the slides were observed under a fluorescent microscope (XSP-BM22AY, Shanghai Optical Instrument Factory, Shanghai, China) for observation.

Flow cytometry

For cell cycle analysis, cells were digested with trypsin, washed with pre-cooled PBS, resuspended in pre-chilled ethanol, and incubated overnight at − 20 °C. The following day, cells were stained with PI (Invitrogen, Eugene, OR) followed by cell cycle assessment employing a flow cytometer (FACS Calibur, BD Biosciences).

Cell apoptosis was detected using a membrane-linked protein V-FITC/PI double staining kit (70-AP101-100, MultiSciences, Hangzhou, China) employing a flow cytometry.

Co-IP

To immunoprecipitate the exogenous proteins, the indicated plasmids (Flag-Fbxo22 and HA-KDM5A) were transfected into HEK293T cells. Cell lysis was carried out in IP lysis buffer (P0013, Beyotime, Shanghai, China) containing protease and phosphatase inhibitors. Cell lysis buffer was obtained by centrifugation at 12,000 g for 20 min at 4 °C. The cell lysis buffer containing 200 μg protein was then incubated with anti-flag antibody (1:50, F3165, Sigma-Aldrich) or anti-HA antibody (1:50, #3724, Cell Signaling, Hercules, CA) for 4 h at 4 °C.

To immunoprecipitate endogenous proteins, the same method as above-described was performed. Antibodies of lysates, KDM5A (1:100, ab70892, Abcam, UK), and IgG (1:50, #3900, Cell Signaling), at a concentration of 1 μg/mg, were added to the cell lysate and incubated overnight at 4 °C. The antibody-protein complexes were then captured with protein A/G Sepharose microbeads (Santa Cruz, CA). The complexes were then subjected to immunoblotting with mouse anti-Fbxo22 (1:1000, sc-100736, Santa Cruz) and mouse anti-KDM5A (1:1000, ab78322, Abcam).

Ubiquitination assay

The indicated plasmids (Flag-Fbxo22, HA-KDM5A, and V5-Ubiquitin) were transfected into HEK293T cells. Prior to collection, cells were treated as previously described (Zhang et al. 2019). The obtained lysates were assayed and immunoprecipitated with protein A/G agarose (Sigma) pre-conjugated with the indicated antibody anti-HA (1:50, #3724, Cell Signaling). In addition, to prevent detection of ubiquitination of the E3 ligase itself and proteins associated with KDM5A, ubiquitination assays were also performed under denaturing conditions. Transfected cells were lysed in lysis buffer and incubated with SDS-PAGE sample buffer at 100 °C for 8 min, followed by incubation with HA antibody for IP and Immunoblotting. Antibodies used for immunoblotting were anti-flag (1:50, F3165, Sigma), anti-HA (1:50, #3724, Cell Signaling), and anti-V5 (1:50, 13,202, Cell Signaling).

Additionally, the ubiquitination level of KDM5A in TNBC cell lines was also examined. The obtained lysates were immunoprecipitated with protein A/G agarose (Sigma) pre-conjugated with the antibody of anti-KDM5A (1:100, ab70892, Abcam) or anti-IgG (1:50, #3900, Cell Signaling). The expression of the relevant proteins was then detected by immunoblotting using the antibodies of mouse anti-KDM5A (1:1000, ab78322, Abcam) and anti-ubiquitin (1:1000, 04–263, Millipore).

Protein half-life detection

TNBC cells were treated with cycloheximide (CHX, Sigma; 10 μM) for different times (0, 0.5, 1, and 2 h) to block protein synthesis. Each group of proteins was then extracted and protein levels of KDM5A were assessed using immunoblotting.

ChIP

ChIP was implemented employing the diluted sonicated lysates. Incubation of cells with the antibodies of KDM5A (1:100, ab70892, Abcam, UK), H3K4me3 (A2357, 1:50, ABclonal), and IgG (#3900, 1:50, Cell Signaling) overnight at 4 °C was carried out, with IgG antibody as a NC. The recovered and purified DNA fragments were used as amplification templates. Immunoprecipitated p16 was tested by RT-qPCR using iQ SYBR Green Supermix (BioRad, Hercules, CA), in which p16 gene promoter primer sequences were designed and provided by Sangon Biotech (Shanghai, China).

RT-qPCR

Trizol-extracted total RNA reagent (15,596,026, Invitrogen) was reversely transcribed to cDNA employing a PrimeScript RT reagent Kit (RR047A, Takara, Japan). Fast SYBR Green PCR kit (Applied Biosystems, Waltham, MA) and ABI PRISM 7500 RT-PCR system (Applied Biosystems) were utilized for RT-qPCR. The relative gene expression was assayed utilizing the 2−ΔΔCt method and normalized to GAPDH. The primer design is summarized in Table S1.

Immunoblotting

The protein extracts from breast cancer tissues and cells were electro-separated and transferred to PVDF membrane which was then incubated with primary antibodies of mouse anti-KDM5A (1:1000, ab78322, Abcam, UK), rabbit anti-p16 (1:1000, A0262, ABclonal), mouse anti-Fbxo22 (1:1000, sc-100736, Santa Cruz), rabbit anti-E-cadherin (1:1000, #3195, Cell Signaling), rabbit anti-N cadherin (1:1000, #13,116, Cell Signaling), rabbit anti-Vimentin (1:1000, #5741, Cell Signaling), rabbit anti-H3K4me3 (A2357, 1:1000, ABclonal), and rabbit anti-GAPDH (1:1000, #2118, Cell Signaling, internal reference) overnight at 4 °C. HRP-labeled goat anti-mouse (1:10,000, BA1050, Boster, Wuhan, China) or goat anti-rabbit IgG (1:10,000, BA1054, Boster) secondary antibodies were incubated with the membrane for 1 h at ambient temperature. The films were exposed in an Amersham Imager 600 (UK). Gray-scale analysis was then performed using ImageJ.

H&E staining

Paraffin-embedded Sects. (5 μm thickness) were first conventionally dewaxed and hydrated by gradient alcohol, and then stained with hematoxylin solution (Solarbio) for 2 min, followed by eosin solution staining for 1 min. Observation of histomorphological changes were completed under a light microscope (XP-330, ShBingyu, Shanghai, China).

Immunohistochemical staining

The protein level of KDM5A, p16, and Fbxo22 was detected using streptavidin peroxidase labeled by immunohistochemical peroxidase. Paraffin-embedded specimens of breast cancer tissues were serially sectioned (5 μm thickness) and immune-stained with primary antibodies of rabbit anti-KDM5A (1:1000, ab78322, Abcam, UK), rabbit anti-p16 (1:100, A0262, ABclonal), and rabbit anti-Fbxo22 (13,606–1-AP, Protein-tech, Wuhan, China) overnight at 4 °C. The samples were incubated at 37 °C for 20 min with biotin-labeled goat anti-rabbit secondary antibody (BA1003, Boster), followed by incubation of 50 μL of streptomyces anti-biotin–peroxidase solution for 10 min at ambient temperature. Color development was completed with 3,3′-diaminobenzidine and microscopic observation for the specimens was performed with IgG used as a NC. Protein-positive cells were identified by the brownish-yellow color of normal positive cells, and positive staining was statistically analyzed using ImageJ.

Statistical analysis

All data were processed using SPSS 22.0 statistical software (IBM, USA) and graphPad Prism 8.0. Measurement data were described as mean ± standard deviation and unpaired t-tests were used for comparisons between two groups, while ANOVA for comparisons among multiple groups. The tumor volume at different time points was assayed by repeated measure ANOVA. p < 0.05 indicated as statistically significant.