All breast cancer cells were obtained from American Type Culture Collection (Manassas, VA, USA) and maintained in Dulbecco's modified Eagle's medium (DMEM; Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) containing 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.) and 1% PenStrep (100 U/ml penicillin and 100 µg/ml streptomycin; Invitrogen; Thermo Fisher Scientific, Inc.), in an incubator at 37°C with 5% CO₂. MDA-MB-231 or MDA-MB-157 cells were plated onto tissue culture plates 24 h prior to transfection. Transient transfection of siUSP21 (siUSP21-1: 5′-GCUAGAAGAACCUGAGUUA-3′; siUSP21-2: 5′-GAGCUGUCUUCCAGAAAUA-3′) or siControl (5′-GUACUUGUACUCCAGCUUUGUG-3′) (Invitrogen; Thermo Fisher Scientific, Inc.) at a final concentration of 50 nM was accomplished with Lipofectmine^® 2000 reagent according to the manufacturer's protocol (Invitrogen; Thermo Fisher Scientific, Inc.).

For scratch wound-healing assays, 48 h following siRNA transfection, 1.5×10⁵ cells were seeded into six-well plates and serum starved for 24 h. Cells were wounded by scratching with a pipette tip and cultured in medium containing 0.5 g/ml mitomycin-C (Sigma-Aldrich; EMD Millipore, Billerica, MA, USA) to block cell division. Cells were photographed at 0, 5 and 10 h timepoints. The distance of migration was calculated using ImageJ software (version 1.46; imagej.nih.gov/ij/). Matrigel invasion assay was performed using a Corning^® BioCoat™ Matrigel Invasion Chamber (Corning Incorportated, Corning, NY, USA). Tumor cells (1×10⁵) treated with control or USP21 siRNA in 200 µl serum-free DMEM were placed in the upper chamber. The lower chamber was filled with 600 µl conditioned medium (DMEM medium containing 1% FBS for 24 h) as chemoattractant. After 24 h of incubation at 37°C, the cells on the upper surface of the filter were removed with a cotton swab. The cells that had invaded the Matrigel and reached the lower surface of the filter were fixed in methanol, stained with hematoxylin and eosin, and counted under magnification, ×400. A total of five fields were randomly selected and the number of invasive cells was counted.

Cells were lysed with radioimmunoprecipitation assay buffer (Sigma-Aldrich; EMD Millipore) as previously described (22–24). Cells were centrifuged at 4°C for 10 min at 16,000 × g. Protein concentrations were determined by the Bradford assay (25). Aliquots containing 20 µg of total protein were separated by 10% sodium dodecyl-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. Blots were probed with primary antibodies against USP21 (1:1,000; goat polyclonal; sc-79305; Santa Cruz Biotechnology, Inc., Dallas, TX, USA) and β-actin (1:5,000, mouse monoclonal; A5316; Sigma-Aldrich; EMD Millipore). Appropriate secondary antibodies (1:3,000; rabbit anti-goat; ab6741; 1:5,000; rabbit anti-mouse; ab97046; Abcam, Cambridge, MA, USA) conjugated to horseradish peroxidase and enhanced chemiluminescence (GE Healthcare Life Sciences, Chalfont, UK) were used to detect the bound primary antibodies. Co-IP was performed with cell lysate (500 µg) incubated with USP21 (1:1,000; goat polyclonal; sc-79305; Santa Cruz Biotechnology, Inc.), relA (1:1,000; rabbit polyclonal; sc-372; Santa Cruz Biotechnology, Inc.) or non-specific-IgG antibodies (1:1,000; rabbit IgG, monoclonal; ab172730; Abcam) using µMACS™ Protein A/G MicroBeads and MACS^® Separation Columns according to the manufacturer's protocol (Miltenyi Biotec, Auburn, USA).

As previously described (26–28), total RNA was isolated from cells using the RNeasy Mini kit (Qiagen, Inc., Valencia, CA, USA). cDNA was prepared from 1 µg of RNA using the iScript cDNA Synthesis kit (Bio-Rad Laboratories, Inc., Hercules, CA, USA). RT-qPCR was performed using a GeneAmp Gold RNA PCR Core kit (Applied Biosystems; Thermo Fisher Scientific, Inc.) and an iCycler iQ™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.), where each 20 µl reaction included 1% cDNA preparation, 0.5 µM primers and 10 µl SYBR Green (Bio-Rad Laboratories, Inc.). Primer sequences are presented in Table I. Expression of glyceraldehyde-3-phosphate dehydrogenase was used to normalize the gene expression level. The relative difference in the expression level was calculated using the ΔΔCq method (29). The data presented are representative of three independent biological repeats each assayed in triplicate and show the relative expression levels.

MDA-MB-231 cells were plated onto a 6-well plate at 70% confluence and transfected with siControl or siUSP21. A total of 48 h subsequent to transfection, the cells were washed with phosphate-buffered saline, and total RNA was obtained using the RNeasy Mini kit (Qiagen, Inc.) and treated with 1 unit of DNase I (Qiagen, Inc.). Expression profiles were generated by hybridizing 10 µg of total RNA to GeneChip^® Human Genome U133 Plus2.0 Gene Chips (Affymetrix, Inc., Santa Clara, CA, USA) according to the Affymetrix Eukaryote One-cycle protocol (30). Briefly, 5–10 µg of total RNA were used to generate biotinylated cDNA, which was fragmented and hybridized to a chip for 16 h at 45°C in a GeneChip Hybridization Oven 640 (Affymetrix, Inc.). Arrays were then washed and stained on a GeneChip Fluidics Station 450 (Affymetrix, Inc.) and subsequently scanned on a GeneChip Scanner 3000 7G (Affymetrix, Inc.) to obtain fluorescence intensities. To eliminate data with low reliability, genes whose expression was regarded as absent in these cell lines as a result of software analysis were excluded. Following identification of the differentially expressed genes, DAVID online software (david.ncifcrf.gov/) was used to perform Kyoto Encyclopedia of Genes and Genomes pathway significant enrichment analysis of 1,007 differentially expressed genes associated with USP21, and the six signaling pathways with the smallest P-values were selected. For TCGA analysis, data including 1,105 breast invasive carcinoma samples from 1,098 patients were obtained from the TCGA website using the cBioPortal tool (www.cbioportal.org/).

Data are presented as the mean ± standard deviation. Statistical analyses were performed using SPSS version 19.0 software (IBM SPSS, Armonk, NY, USA). Statistically significant differences were determined by Student's t-test. P<0.05 was considered to indicate a statistically significant difference.

USP family proteins have a significant role in multiple signaling and cell regulatory networks in breast cancer (9). To characterize the extent of changes in all 50 USP genes in breast cancer, the present study generated an alteration-summary for these genes in breast cancer using the cBioPortal tool for Cancer Genomics (31,32) (Fig. 1A). The results of the present study demonstrated that USP21 was altered in 39% (375/971 samples) of patients with breast invasive carcinomas, indicating it is the most altered gene among all the USP gene family members. In all the deregulation situations, including gene copy number amplification and mRNA expression alteration, of USP21, 13.6% (132/971) of the patients displayed copy number amplification, while 37.8% (367/971) of the patients showed mRNA upregulation (Fig. 1B), suggesting a role for USP21 in breast cancer.

To investigate the expression of USP21 in breast cancer cell lines in vitro, the present study analyzed a panel of 10 breast cancer cell lines including luminal, basal and non-cancerous breast epithelial lines. Notably, it was observed that the expression of USP21 was enriched in triple negative cell lines, MDA-MB-231 and MDA-MB-157 (Fig. 1C). Therefore, it appears that USP21 is required for the cancerous ability of TNBC cells.

To directly investigate the contribution of USP21 to breast tumorigenesis, the present study knocked down USP21 protein using two specific siRNAs in MDA-MB-231 and MDA-MB-157 cells. Western blot analysis confirmed the knockdown of USP21 under these conditions (Fig. 2A). It was initially investigated whether USP21 is crucial to the proliferation of these cells. Cells transfected with control or USP21-siRNAs were cultured for up to 7 days. The difference in cell proliferation was measured by viable cell counting from day 1 to day 7. Knockdown of USP21 reduced the proliferation of MDA-MB-157 and MDA-MB-231 cells 5 days after siRNA transfection (Fig. 2B and C).

To determine the effects of USP21 on cell migration, a wound-healing assay was performed following USP21 silencing. The scratch wounds were almost identical sizes in each experimental group at 0 h; however, knockdown of USP21 using two different siRNAs markedly decreased the migration ability of MDA-MB-231 and MDA-MB-157 cells at 5 and 10 h (Fig. 3A). The present study measured the distance between the migrating frontlines and calculated the rate of wound closure. It was observed that the USP21 silenced groups were less effective than the control group in terms of the healing process (Fig. 3A).

To investigate the role of USP21 in TNBC cell invasion, the present study measured the invasive ability using invasion assays. Consistent with the migration assay results, downregulation of USP21 in MDA-MB-231 and MDA-MB-157 cells by siRNA knockdown resulted in a marked reduction in the cell invasive capability compared to the control group (Fig. 3B). These results suggested that USP21 is involved in TNBC cancer growth, migration and invasion.

To further investigate USP21-mediated gene expression changes, the cDNA from USP21-knockdown and control MDA-MB-231 cells was subjected to microarray analyses. A total of 1,007 genes in USP21 siRNA treated cells exhibited a differential expression more than double that of the control sample. A total of 309 genes were upregulated in the USP21 knockdown sample compared with the control, while 698 genes demonstrated decreased expression in the USP21 siRNA treated sample. Furthermore, it was observed that the major signaling pathways of the differentially expressed genes were associated with the following molecular pathways: NOD-like receptor signaling, TGF-β signaling pathway, RNA degradation, small cell lung cancer, pathways in cancer and extracellular matrix-receptor interaction. It was observed that USP21 depletion markedly attenuated genes specifically associated with the NOD-like receptor signaling pathway [interleukin (IL)6, NLR family, pyrin domain containing (NLRP)3, IκBα and chemokine (C-X-C motif) ligand (CXCL)8] and stimulated genes associated with the TGF-β signaling pathway (bone morphogenetic protein (BMP)4, BMP type IB receptor, ID1, ID2, ID3, ID4 and TGF-β receptor 1) (Fig. 4A and B). To confirm the microarray results, 8 genes were selected (USP21, IL8, caspase recruitment domain family member 8 (CARD8), chemokine (C-C motif) ligand 2 (CCL2), IL6, CXCL1, NLRP3 and IκBα) and their expression was measured with RT-qPCR on the same sample used for microarray studies. All of these genes exhibited moderate to high expression and demonstrated high concordance between microarray and PCR data (Fig. 4C).

As USP21 has been reported to be a histone H2A deubiquitinase that initiates transcriptional activity (12), the present study screened the association of USP21 with NOD-like receptor associated transcription factors, including NF-κB, activator protein 1 or interferon regulatory factors. Notably, it was observed that relA, an essential subunit of the transcriptionally active NF-κB dimer, may be ‘pulled down’ with USP21 using a co-IP assay (Fig. 4D). Thus, these results indicate that USP21 may associate with NF-κB transcription factors.