A Novel Small-Molecule Aurora Kinase Inhibitor Attenuates Breast Tumor–Initiating Cells and Overcomes Drug Resistance

Breast cancer is one of the most common malignancies and is a leading cause of mortality in women (1). Although treatment strategies that combine surgery with adjuvant chemotherapy improve survival rates, a significant portion of patients eventually acquire resistance to chemotherapeutic agents. Chemoresistance can be a consequence of cell intrinsic genetic changes, such as reprogramming of metabolic pathways, upregulation of drug efflux pumps, activation of detoxifying enzymes, or apoptotic defects (2, 3). In addition, chemoresistance can also result from cell extrinsic factors, such as cytokines and growth factors (4, 5). The intrinsic or extrinsic mechanisms of chemoresistance have been demonstrated to be a major cause of cancer treatment failure. Thus, strategies to overcome chemoresistance are urgently needed.

The cancer stem cell (CSC) hypothesis offers new insight into the mechanism of chemoresistance. The CSC hypothesis states that a small population of cancer cells, also termed tumor-initiating cells (TIC), has undifferentiated phenotypes, an increased capacity for self-renewal, and the ability to form new tumors (6–8). The characteristics that differentiate TICs from other tumor cells are high expression of levels of drug efflux transporter proteins and detoxification enzymes (9, 10). These molecules are responsible for protecting TICs from drug damage and for producing chemoresistant clones. Therefore, eradicating TICs could be a therapeutic avenue for overcoming chemoresistance.

Aurora-A (AurA) belongs to the Aurora family of serine/threonine kinases, which are central for mitotic progression (11). Small-molecule kinase inhibitors of AurA have attracted considerable interest due to their overexpression in various tumor types, and because of their function as an oncogene depending on the kinase activity (12, 13). Recent studies suggest that the kinase activity of AurA is responsible for chemoresistance as it overrides cell-cycle checkpoint (14–17). AurA also plays an important role in the tumorigenicity and chemoresistance of colorectal TICs (18). However, whether the kinase activity of AurA could be a potential anti-TIC target for overcoming chemoresistance remains unclear.

In this study, we explored the potential anti-TIC role of the novel small-molecule Aurora kinase inhibitor AKI603. We demonstrated that AKI603 inhibited AurA kinase and xenograft tumor growth. Interestingly, epirubicin-resistant cells displayed enhanced AurA kinase activity and were enriched in the CD24^Low/CD44^High TIC population. Targeting AurA kinase activity with AKI603 potently suppressed the epirubicin-induced accumulation of CD24^Low/CD44^High TICs. Thus, we presented a novel mechanism by which the inhibition of AurA kinase can overcome chemoresistance by abolishing the accumulation of TICs. This mechanism has important implications for cancer therapy.

AKI603 was designed and synthesized by our laboratory. The synthetic route is shown in Supplementary Fig. S1A. 2,4,6-Trichloropyrimidine was treated <1> with 3-amino-5-methylpyrazole <2>, which afforded the C4-substituted pyrimidine <3> regioselectively. Next intermediate 5 was generated via nucleophilic substitution at the C2-position of the pyrimidine core in the presence of 4-nitroaniline <4>. Microwave-assisted substitution at the C6-position of the pyrimidine ring with 1-methylpiperazine <6> yielded AKI603.

AKI603 was dissolved in dimethyl sulfoxide (DMSO) to a stock concentration of 100 mmol/L and was stored at −20°C. Epirubicin was dissolved in H₂O to a stock concentration of 100 mmol/L and was stored at −20°C. The human breast cancer cell lines BT549, MCF-7, Sk-br-3, MDA-MB-231, MDA-MB-453, and MDA-MB-468 were obtained from the American Type Culture Collection (ATCC). The cell lines were authenticated at ATCC before purchase by their standard short tandem repeat DNA typing methodology. Epirubicin-resistant MCF-7 (MCF-7-Epi) cells have been described previously (19), and were authenticated by the standard short tandem repeat DNA typing methodology. SUM149 cells were kindly provided by Prof. Zhi-Min Shao (Department of Medical Oncology, Cancer Hospital of Fudan University, Shanghai Medical College, Shanghai, China), and were authenticated by the standard short tandem repeat DNA typing methodology. Each cell line was cultured in its standard medium as recommended by ATCC. MCF-7-Epi cells were cultured in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS; Hyclone). SUM149 cells were cultured in F-12 Hams (Gibco) supplemented with 5% FBS (Hyclone), 5 μg/mL insulin (Sigma), and 1 μg/mL hydrocortisone (Sigma). All cell lines used in this study were obtained in 2013, and were used within 6 months of receipt for this study.

The kinase activity of AurA was measured using the Caliper Mobility Shift Assay. Briefly, 3-fold dilutions of a 10 μmol/L stock of AKI603 (a total of 10 concentrations) were tested. AKI603, AurA, the FAM-labeled peptide, and ATP were added in a 384-well format, and the plate was incubated at 28°C for 1 hour. The reaction was stopped by the addition of a stop buffer and the data were collected with a LabChip EZ reader (PerkinElmer).

The cells were treated with the indicated concentrations of AKI603 for 48 hours, collected, and fixed in ice-cold alcohol (75%). After an overnight incubation at 4°C, the cells were collected by centrifugation, and were resuspended in a propidium iodide (PI; Sigma) staining solution (PI, 50 μg/mL; RNase A, 100 μg/mL; Triton X-100, 0.2% in PBS; and 1 mL/sample) at a concentration of 1.0 × 10⁶ cells/mL. After PI staining, the quantification of the cell-cycle distribution was carried out using a flow cytometer (Beckman) equipped with the Multicycle software (Phoenix).

To evaluate CD24 and CD44 expression, 2 × 10⁵ cells were plated in dishes (Φ = 6 cm) and were cultured with the indicated concentrations of AKI603 and/or epirubicin for the indicated times. The cells were harvested and incubated with anti-human CD24 phycoerythrin (PE; eBioscience) and anti-human CD44 FITC (eBioscience) at 4°C for 0.5 hours in the dark. The cells were then resuspended in 1 mL of PBS and subjected to flow cytometric analysis.

To purify the CD24^Low/CD44^High and CD24^High/CD44^High populations, 1 × 10⁷ exponentially growing cells were harvested and incubated with anti-human CD24 PerCP-eFluor 710 (eBioscience) and anti-human CD44 FITC (eBioscience) at 4°C for 0.5 hours in the dark. Next, the cells were resuspended in 3 mL of PBS and subjected to sorting.

The cells incubated with 0.6 μmol/L AKI603 for 24 hours were fixed in paraformaldehyde (PFA) for 15 minutes at 4°C and were permeabilized in 0.5% TritonX-100 in PBS at room temperature for 10 minutes. The cells were then incubated with 1% BSA for 1 hour at room temperature to block nonspecific binding before the addition of the primary antibody. The slides were incubated with the primary antibodies to AurA (Upstate) and α-Tubulin (Sigma) at room temperature for 1 hour, followed by an Alexa Fluor 488- or 546–conjugated secondary antibody (Invitrogen). After counterstaining with DAPI (1 μg/mL; Sigma), the cells were visualized using a confocal microscope (×630; Olympus).

The cells treated with indicated concentration of AKI603 for 48 hours were harvested and were lysed in RIPA buffer. The protein concentrations were determined by the Bradford method using BSA (Sigma) as the standard. Equal amounts of cell extract were subjected to electrophoresis in SDS–polyacrylamide gel and were transferred to nitrocellulose membrane (Millipore). The membranes were blocked and then incubated with GAPDH (Ambion), β-actin (Santa Cruz Biotechnology), pAurA (Thr288; Cell Signaling Technology), pAurA/B (Cell Signaling Technology), AurA (Upstate), AurB (Cell Signaling Technology), HA-tag (Sigma), β-catenin (Cell Signaling Technology), c-Myc (Santa Cruz Biotechnology), Sox2 (Epitomics), Oct4 (Epitomics), and Nanog (Epitomics) primary antibodies at 4°C overnight. Next, the membranes were incubated for 1 hour at room temperature with the appropriate secondary antibodies. Antibody binding was detected with an enhanced chemiluminescence kit (Pierce).

The mammosphere formation assay was performed as previously described (20). A single-cell suspension was obtained by trypsinization. Clumped cells were excluded with a 40-μm sieve, and the suspension was analyzed microscopically for single cellularity. The percentage of clumped cells was <5%. Single cells were plated in ultralow attachment 6-well plates at a low density of 1,000 viable cells/mL. The cells were maintained in DMEM/F12 (Gibco) supplemented with B27 (Invitrogen), 20 ng/mL EGF (Sigma), 20 ng/mL basic fibroblast growth factor (bFGF; BD Biosciences), and 4 μg/mL heparin (Sigma) for 10 days. The mammospheres were photographed using inverted microscope (×100; Olympus). The diameters of the mammospheres were calculated with the Image pro plus 6.0 software (Media Cybernetics).

The cells were plated at 1 × 10³ cells per well and were treated with the indicated concentrations of AKI603 and/or epirubicin for 6 days. The cells were then fixed in 2% PFA and stained with a 0.05% crystal violet solution (Sigma). The number of colonies was counted.

Adherent cells were treated with the indicated concentrations of AKI603 or epirubicin, collected, and resuspended in the binding buffer. Annexin-V–FITC and PI were added to the cells according to the protocol contained in the Annexin V–FITC Apoptosis Detection Kit (EMD Biosciences). The samples were then incubated for 15 minutes in the dark at 4°C and subjected to flow cytometric analysis.

The data acquired from the MTT assay were used to evaluate the combination index (CI). The CI was analyzed with the CompuSyn software using the average fraction of cells that responded to each drug (21). CI values of less than 0.8, between 0.8 and 1.2, and more than 1.2 were defined as synergistic, additive, and antagonistic, respectively (22).

Eight-chambered RS glass slides (BD Falcon) were precoated with 40 μL/well of Matrigel (BD Biosciences). The Matrigel was solidified at 37°C for 30 minutes. Next, the cells were suspended in growth medium containing a final concentration of 2% Matrigel and plated at a density of 2,500 cells per well. The cells were fed with growth medium containing 2% Matrigel every 4 days. After 10 days of culture, the cells were photographed using an inverted microscope. The diameters were calculated with Image pro plus 6.0 software.

After obtaining informed consent, breast cancer specimens were collected from patients undergoing surgery in accordance with the Institutional Review Boards of the Sun Yat-sen University Cancer Center (Guangzhou, China). Breast lesions were classified by histologic diagnostic assessment and were sampled by pathologists. The lesions were mechanically and enzymatically dissociated to yield clumps of epithelial cells, termed organoids, by incubation at 37°C for 2 hours in a 1:1 solution of collagenase I (3 mg/mL):hyaluronidase (100 U/mL; Sigma). After filtration through a 40-μm pore filter and washing with PBS, the organoids from the tumor tissue were trypsin dissociated to single cells for subsequent experiments.

MCF-7-Epi cells (7.5 × 10⁶) were injected into the right flank of 4-week-old female nude mice (n = 5). The mice were not supplemented with estrogen, as previous studies showed that drug-resistant MCF-7 cells do not require estrogen signaling to form tumors in nude mice (23, 24). On day 30, the mice (tumor size, ∼50 mm³) were randomly distributed into two groups that were treated intragastrically every day with either a vehicle control or with 50 mg/kg AKI603 dissolved in PEG300. The tumor volumes (A × B²/2; A being the greatest diameter and B being the diameter perpendicular to A) were measured by calipers. Other indicators of general health, such as body weight, feeding behavior, and motor activity, of each animal were also monitored. After administering the drug or vehicle for 14 days, the mice were fed for another 14 days. At day 59, the mice were sacrificed and the tumor xenografts were immediately dissected, weighed, stored, and fixed. All animal procedures were approved by the Institutional Animal Care and Use Committee of the Sun Yat-sen University Cancer Center.

Statistical analyses were performed using the SPSS software, version 16.0 (SPSS Inc.) or with GraphPad Prism 5.0 (GraphPad Software, Inc.). The Kruskal–Wallis test, followed by a Dunn multiple comparison test, was used to perform a statistical comparison with regard to mammospheres size distribution. The unpaired Student t test was used to perform a statistical comparison between two groups. The ANOVA test, followed by the least significant difference test, was used when performing multiple comparison. The level of significance was set at P < 0.05.

The chemical structure of AKI603 was shown in Fig. 1A. To assess whether AKI603 inhibited AurA kinase activity, an in vitro kinase activity assay was performed. As shown in Fig. 1B, the IC₅₀ of AKI603 was determined to be 12.3 ± 1.5 nmol/L. To evaluate the effect of AKI603 on breast cancer cell proliferation, eight breast cancer cell lines were treated with AKI603 for 48 hours, and the rate of cell proliferation was determined by an MTT assay. The clinicopathologic features of the eight cell lines have been previously summarized (25). As shown in Supplementary Fig. S1B, all the tested breast cancer cell lines were sensitive to AKI603 and showed typical sigmoidal log (dose)–response curves. A cell counting assay was also performed to confirm the MTT results for the MDA-MB-453 and MDA-MB-468 cells (Supplementary Fig. S1C). The in vitro IC₅₀s, as measured by sigmoidal curve fitting for each cell line, are summarized in Supplementary Fig. S1D. The IC₅₀s that were acquired by the MTT assay were consistent with the cell counting experiments. Next, we examined whether AKI603 inhibited the kinase activity of AurA in breast cancer cells, by examining the phosphorylation of AurA Thr288. As shown in Fig. 1C, treatment with 0.6 μmol/L or higher concentrations of AKI603 for 48 hours significantly inhibited Thr288 phosphorylation (detected with pAurA) in all the cell lines tested. The total AurA protein levels were marginally reduced or not changed. Moreover, we evaluated the inhibitory activity of AKI603 against AurB by detecting Thr232 phosphorylation in SUM149 and BT549 cells. We found that AKI603 also inhibited AurB kinase activity (Fig. 1C). The inhibitory activity of AKI603 toward AurB kinase was lower than that toward AurA kinase as determined by Western blotting (Fig. 1C).

Because of the crucial role of AurA in mitosis, we determined whether inhibition of AurA kinase activity by AKI603 could block the cell cycle in SUM149, BT549, MCF-7, MCF-7-Epi, Sk-br-3, and MDA-MB-231 cells. After treatment with 0.6 μmol/L AKI603, SUM149 cells were fixed and stained with the indicated antibodies, and the cells in mitosis were analyzed. As shown in Fig. 1D, mitotic cells in the control group displayed mainly bipolar spindle (Fig. 1D, white arrow in the left). AurA (Fig. 1D, white arrow in the left, green) localized at the region of the centrosomes and mitotic spindle. AKI603 treatment (Fig. 1D, white arrow in the right) increased monopolar spindle formation compared with the control group (7.40% ± 0.02% vs. 3.47% ± 0.01%; P = 0.025; Fig. 1E). A similar phenomenon was observed in BT549, MCF-7, MCF-7-Epi, Sk-br-3, and MDA-MB-231 cells (Supplementary Fig. S2). For cell-cycle profile analysis, SUM149 cells were treated with 0.3 or 0.6 μmol/L AKI603 for 48 hours, and the DNA content was analyzed after PI staining. As shown in Fig. 1F, with increasing concentrations of AKI603 from 0 to 0.3 μmol/L and 0.6 μmol/L, the percentage of cells in G₂–M phase (4N) was increased from 20.9% ± 0.6% to 36.6% ± 6.2% and 45.2% ± 1.5%, respectively. Furthermore, the percentage of cells that contained more than 4N DNA was increased from 0.2% ± 0.2% to 14.0% ± 1.9% and 20.5% ± 3.6%, respectively. Similar results were obtained with BT549, MCF-7, MCF-7-Epi, Sk-br-3, and MDA-MB-231 cells (Supplementary Fig. S3). These results suggested that the inhibition of cell proliferation that was induced upon AKI603 treatment is associated with cell-cycle blockage.

MCF-7-Epi cells were used to investigate the potential role of AurA kinase activity in the regulation of TICs and drug resistance. An MTT assay showed that MCF-7-Epi cells (IC₅₀ = 7.45 ± 2.01 μmol/L) exhibited significant resistance toward epirubicin compared with wild-type MCF-7 cells (IC₅₀ = 0.38 ± 0.09 μmol/L; Fig. 2A). To confirm these results, a cell counting assay was performed. The results were consistent with those from the MTT assay (Fig. 2B). The kinase activity of AurA was significantly increased in MCF-7-Epi cells compared with wild-type MCF-7 cells, suggesting that AurA kinase activity might be important for maintaining resistance to epirubicin. Self-renewal genes, including β-catenin, c-Myc, Sox2, Oct4, and Nanog, play key roles in maintaining stemness (26–33) and mediating chemoresistance (34–36). As shown in Fig. 2C, the expression of β-catenin, c-Myc, Sox2, Oct4, and Nanog was also increased in MCF-7-Epi cells compared with wild-type MCF-7 cells. Interestingly, flow cytometric analysis showed that MCF-7-Epi cells contained a higher percentage of CD24^Low/CD44^High TICs (26.6% ± 3.5679%) than their wild-type counterparts (0.67% ± 0.21%; P = 0.0002; Fig. 2D). Indeed, our data indicated that the CD24^Low/CD44^High TIC population was more resistant to conventional chemotherapeutic agents than the CD24^High/CD44^High population (Supplementary Fig. S4). These results suggested that targeting the kinase activity of AurA might be a potential strategy to deplete the CD24^Low/CD44^High TIC population in breast cancer. The cytotoxic effects of genotoxic agents, such as epirubicin, might also be enhanced.

We next asked whether AKI603 targeted the CD24^Low/CD44^High TIC population in breast cancer cells. To evaluate whether AKI603 induced apoptosis, MCF-7-Epi cells were collected for flow cytometry analysis of AnnexinV/PI staining. Our results showed that MCF-7-Epi cells treated with 0.6 μmol/L of AKI603 did not undergo obvious apoptosis (Supplementary Fig. S5A). To determine whether AKI603 has an anti-TIC effect, we first examined the short-term effects of AKI603 in the CD24^Low/CD44^High TIC population. Results showed that a short-term (<6 days) treatment of MCF-7-Epi cells with AKI603 did not show an obvious anti-TIC effect (Supplementary Fig. S5B). We further tested the long-term effect of AKI603 in the CD24^Low/CD44^High TIC population. MCF-7-Epi cells were treated with 0.6 μmol/L AKI603 or DMSO for 6 and 9 days, and the adherent cells were collected for flow cytometric analysis of CD24/CD44 expression. As shown in Fig. 3A, before treatment (at day 1), both the DMSO control group (Fig. 3B; CD24^Low/CD44^High TIC population = 25.1% ± 1.6%) and the AKI603 treatment group (Fig. 3B; CD24^Low/CD44^High TIC population = 23.8% ± 4.0%) contained comparable proportions of CD24^Low/CD44^High cells (P = 0.6483). The CD24^Low/CD44^High TIC population was gradually depleted by treatment with AKI603 (Fig. 3A, bottom) at day 7 (Fig. 3B; CD24^Low/CD44^High TIC population = 12.3% ± 2.0%; P = 0.0081) and day 10 (Fig. 3B; CD24^Low/CD44^High TIC population = 2.4% ± 2.3%; P = 0.0006), whereas the CD24^Low/CD44^High TIC population did not significantly change in the DMSO control group (Fig. 3A, top and Fig. 3B; day 7, CD24^Low/CD44^High TIC population = 25.4% ± 4.2%; day 10, CD24^Low/CD44^High TIC population = 25.4% ± 3.3%). Similar results were obtained with SUM149 (Supplementary Fig. S5C) and BT549 cells (Supplementary Fig. S5D). We further evaluated whether the CD24^Low/CD44^High TICs from primary patient samples were sensitive to AKI603. We found that AKI603 also effectively depleted the CD24^Low/CD44^High TIC population of primary breast cancer cells (Supplementary Fig. S5E). Interestingly, the CD24^Low/CD44^High population of normal primary breast cells was insensitive to AKI603 (Supplementary Fig. S5E), indicating that AKI603 may exhibit low toxicity. To confirm that the disruption of AurA kinase activity resulted in anti-TIC effects in breast cancer cells, kinase-dead mutants of AurA (K162M or D274N; refs. 37, 38) were expressed in SUM149 cells (Fig. 3C). Indeed, the overexpression of either K162M or D274N reduced the CD24^Low/CD44^High TIC population in SUM149 cells (Fig. 3D). These effects were similar to the anti-TIC effects induced by AKI603.

In addition, we examined the effects of AKI603 on mammosphere formation in breast cancer cells. To this end, MCF-7-Epi cells were maintained in mammosphere culture medium containing DMSO, 0.6, or 1.2 μmol/L AKI603, and the mammosphere diameters and numbers were measured. As shown in Fig. 3E and F, compared with controls, the mammosphere sizes were smaller when cultured with 0.6 μmol/L (P < 0.001; Fig. 3E and F) and 1.2 μmol/L AKI603 (P < 0.001; Fig. 3E and F). The number of large mammospheres (Φ > 60 μm) in the 0.6 μmol/L (65 ± 18 mammospheres per 1,000 cells; P = 0.0001) and 1.2 μmol/L (27 ± 8 mammospheres per 1,000 cells; P = 0.0005) AKI603 treatment groups was lower than in the DMSO control group (268 ± 29 large mammospheres per 1,000 cells; Fig. 3G). Similar results were also observed in SUM149 (Supplementary Fig. S5F–S5H) and BT549 cells (Supplementary Fig. S5I–S5K).

Next, we investigated the expression of several self-renewal genes in breast cancer cells after AKI603 treatment. Results showed that 0.6 μmol/L concentration of AKI603 significantly inhibited AurA activity in all tested breast cancer cell lines (Fig. 1C). Meanwhile, AKI603 suppressed the expression levels of the self-renewal genes, including β-catenin, c-Myc, Sox2, Oct4, and Nanog in a dose-dependent manner (Fig. 3H and Supplementary Fig. S6A).

We next explored the combinational effects of epirubicin and AKI603 in breast cancer cells. The long-term (6 days) growth-inhibitory effects of low doses of AKI603 were determined with an MTT assay. The dose that inhibited cell proliferation by 20% was chosen for further investigations on the anti-TICs effects. As shown in Supplementary Fig. S6B, in SUM149 cells, 0.078 μmol/L of AKI603 or 0.078 μmol/L epirubicin alone inhibited cell proliferation by approximately 20% after 6 days of treatment in SUM149 cells. Flow cytometric analysis showed that incubation with a control solvent did not significantly change the ratio of the CD24^Low/CD44^High TIC population at days 1, 7, and 10 (35.2% ± 2.24%, 34.7% ± 2.70%, and 33.3% ± 2.16%, respectively; Fig. 4A and B). Interestingly, epirubicin alone resulted in an enrichment of the CD24^Low/CD44^High TIC population and the percentages at days 1, 7, and 10 were 34.3% ± 2.70%, 43.1% ± 2.46%, and 60.6% ± 10.10%, respectively (Fig. 4A and B). For AKI603 treatment, the CD24^Low/CD44^High TIC population at days 1, 7, and 10 were 34.16% ± 2.10%, 10.0% ± 2.03%, and 6.0% ± 1.72%, respectively (Fig. 4A and B). Crucially, when SUM149 cells were treated with epirubicin and AKI603 together, the epirubicin-induced enrichment of the CD24^Low/CD44^High TIC population was abolished. Under these conditions, the CD24^Low/CD44^High TIC population at days 1, 7, and 10 were 34.2% ± 2.86%, 11.8% ± 2.61%, and 4.1% ± 1.02%, respectively (Fig. 4A and B). Similar findings were observed in MCF-7 cells (Supplementary Fig. S6C and S6D).

In SUM149 cells, the expression of AurA, pAurA, and self-renewal genes was also analyzed by Western blot analysis. As shown in Fig. 4C, the expression levels of tested markers were relatively stable in the control groups (Fig. 4C, compare lanes 2–3 with lane 1). However, in the epirubicin-treated cells (Fig. 4C, compare lanes 5–6 with lane 4), pAurA and Nanog were significantly increased, c-Myc and Oct4 were decreased, and AurA, β-catenin, and Sox2 expression levels did not change significantly. The elevated expression of pAurA and Nanog suggested that these proteins might be responsible for the epirubicin resistance and the enriched the CD24^Low/CD44^High TIC population. In the AKI603-treated cells (Fig. 4C, compare lanes 8–9 with lane 7), the expression levels of pAurA as well as that of β-catenin, c-Myc, Sox2, Nanog, and Oct4 were significantly suppressed. Importantly, for the cells exposed to the epirubicin/AKI603 combination, AKI603 reversed the epirubicin-induced upregulation of pAurA and Nanog. Furthermore, the combination of epirubicin and AKI603 resulted in a greater inhibition of β-catenin, c-Myc, Sox2, and Oct4 expression (Fig. 4C, compare lanes 10–12 with lane 4–6).

Moreover, we evaluated the combinational effects of AKI603 and epirubicin on mammosphere formation using SUM149 cells. As shown in Fig. 4D and E, epirubicin alone (P < 0.001) or AKI603 alone (P < 0.001) slightly reduced mammosphere size compared with the control group. When the SUM149 cells were treated with epirubicin and AKI603 simultaneously, mammosphere formation was significantly blocked (P < 0.001; Fig. 4D and E). As shown in Fig. 4F, AKI603 decreased the formation of large mammospheres (125 ± 35 mammospheres per 1,000 cells; P = 0.0319) compared with the control group (187 ± 30 mammosphere per 1,000 cells), whereas epirubicin did not significantly downregulate the formation of large mammosphere (182 ± 21 mammospheres per 1,000 cells; P = 0.4336). However, the combination of epirubicin and AKI603 potently inhibited the formation of large mammospheres (11 ± 2 mammospheres per 1,000 cells; P = 0.0106).

Chemotherapy with high doses of chemotherapeutic reagents often causes side effects. Therefore, new strategies to allow for treatment with lower dose chemotherapy are urgently needed. Synergistic analysis was performed to evaluate the interactions between AKI603 and epirubicin in breast cancer cell lines. The experimental setting of AKI603 and epirubicin treatment was the same as in Supplementary Fig. S6B, where AKI603 and epirubicin were combined in a fixed ratio (1:1). The results showed that the combination therapy resulted in a greater growth inhibition of SUM149 cells (Fig. 5A, labeled in green; IC₅₀ = 0.089 ± 0.02 μmol/L) than was achieved with either AKI603 (Fig. 5A, labeled in red; IC₅₀ = 0.41 ± 0.09 μmol/L) or epirubicin alone (Fig. 5A, labeled in blue; IC₅₀ = 0.41 ± 0.15 μmol/L). This result, which was achieved using an MTT assay, was confirmed by cell counting (Fig. 5B). The average fraction of cells that responded to each drug was subjected to CI analysis (21). As shown in Fig. 5C and D, AKI603 and epirubicin acted synergistically to inhibit SUM149 cell proliferation. Similar findings were observed for BT549 (Supplementary Fig. S6E and S6F) and MCF-7 cells (Supplementary Fig. S6G and S6H). To confirm these results, a drug concentration of 0.078 μmol/L was used in a colony formation assay. As shown in Fig. 5E and F, in SUM149 cells, AKI603 (78.8% ± 9.2%; P = 0.0161) and epirubicin (73.8% ± 5.7%; P = 0.0013) alone marginally inhibited colony formation compared with the control groups, whereas the combination (10.3% ± 3.3%; P < 0.001) substantially suppressed colony formation compared with the control group. Furthermore, a three-dimensional (3D) culture model was used to mimic the in vivo tissue environment. Compared with in vivo animal models, the 3D culture model is a more purified system with high reproducibility, and experimental results can be recorded in real time (39). Thus, 3D culture model was used to confirm the enhanced efficacy of the epirubicin/AKI603 drug combination. Indeed, AKI603 enhanced the cytotoxicity of epirubicin in this 3D culture model (Fig. 5G and H). Our results suggested that the inhibition of AurA by AKI603 provided a potential strategy by which the concentrations of genotoxic chemotherapeutic agents can be lowered in the treatment of cancer.

To evaluate the in vivo anticancer effects of AKI603, we used a xenograft model. Nude mice bearing MCF-7-Epi xenograft tumors were treated with AKI603 (50 mg/kg, every day) by intragastric administration for 14 days. As shown in Fig. 6A and B, the tumor size in the AKI603-treated group (172 ± 69 mm³; P = 0.0136) was smaller than that in the control group (367 ± 119 mm³), indicating that the growth of xenograft tumors was significantly inhibited by AKI603. Consistently, the tumor weight in the AKI603-treated group (0.1121 ± 0.0973 g; P = 0.0159) was significantly lower than that in the control group (0.3829 ± 0.1807 g; Fig. 6C). During the experimental period, the mice in the AKI603-treated group showed a slight decrease in body weight (Fig. 6D), indicating the low toxicity of intragastrically administered AKI603.

In this study, we evaluated the anti-TIC effects of a novel small-molecule Aurora kinase inhibitor AKI603. We showed that AKI603 effectively inhibited AurA kinase activity and xenograft tumor growth (Figs. 1B and C and Fig. 6). Epirubicin treatment induced the upregulation of AurA kinase activity (Figs. 2C and 4C) and an enrichment in the CD24^Low/CD44^High TIC population (Fig. 4A and B and Supplementary Fig. S6C and S6D). Surprisingly, AKI603 abolished this epirubicin-induced CD24^Low/CD44^High TIC population enrichment (Fig. 4A and B and Supplementary Fig. S6C and S6D) and overcame epirubicin resistance (Figs. 4D–F and Fig. 5).

The development of chemoresistance is a major hindrance to the effective treatment of cancer. The mechanism of chemoresistance is complicated because of numerous factors contributing to the regulation of drug sensitivity, such as accelerated drug efflux, drug inactivation, DNA-damage repair, and the evasion of apoptosis (40). Traditional chemotherapies target rapidly dividing cancer cells and effectively lower the tumor burden. However, a small fraction of cancer cells, termed TICs, remain quiescent in the G₀-phase and express high levels of drug efflux transporter proteins and detoxification enzymes (9, 10). Our data suggested that chemoresistance might result from an enrichment in TICs in response to chemotherapeutics (Fig. 2 and Fig. 4A–C). Over a period of chemotherapeutic treatment, TICs survive and can subsequently form a new tumor with a chemoresistant phenotype. Consistent with this notion, previous reports show that CD34⁺/CD38⁻ leukemic precursors exhibit increased resistance to daunorubicin in comparison with their CD34⁺/CD38⁺ counterparts. The CD34⁺/CD38⁻ progenitors also express higher expression levels of multidrug resistance genes and have lower levels of Fas/Fas-L and Fas-induced apoptosis compared with CD34⁺/CD38⁺ blasts (41). Moreover, the treatment of KG1a leukemic cells with 50 μg/mL of 5-fluorouracil (5-FU) for 4 days enriches the CD34⁺/CD38⁻ subpopulation more than 10-fold. These cells show higher levels of ABCG2 and an increase in side population compared with untreated cells (42). Indeed, the CD24^Low/CD44^High TIC population displayed higher drug resistance compared with the CD24^High/CD44^High population in breast cancer (Supplementary Fig. S4). These results from our group and others suggested that TICs could be targeted in an effort to overcome chemoresistance.

AurA regulates multiple critical mitotic processes (11, 43), including cell fate determination. In Drosophila bearing AurA with a mutation in the catalytic domain, Numb is distributed uniformly around the cell cortex during mitosis and is separated equally into the two daughter cells. This uniform distribution of Numb leads to cell fate changes in the sensory organ precursor lineage (44). Drosophila bearing another mutation of AurA had an increased number of brain neuroblast numbers (45). Most recently, a study shows that AurA is necessary for embryonic stem cell maintenance. AurA phosphorylated Ser212 and Ser312 of p53 to control ectodermal and mesodermal differentiation (46). Although a role for AurA kinase activity in the control of cell fate determination in normal stem cells has been well documented, it is not known if it is similarly involved in cancer. Our data showed that the inhibition of AurA activity by AKI603 reduced the CD24^Low/CD44^High TIC population and decreased mammosphere formation and self-renewal gene expression (Fig. 3). Consistently, the overexpression of kinase-dead AurA had similar inhibitory effect on the CD24^Low/CD44^High TIC population in SUM149 cells (Fig. 3D), suggesting that AurA kinase activity was important for maintaining the TIC population. Interestingly, our data showed that AKI603 also inhibited AurB kinase activity. However, the inhibitory effect of AKI603 on AurB was lower than that on AurA. These data indicated that AurB might also be involved in maintaining the TIC population.

Consistent with our finding that the inhibition of AurA increased cell sensitivity to conventional chemotherapeutic drugs (Fig. 5), previous studies indicate that the kinase activity of AurA is critical for overriding cell-cycle checkpoints in cancer, and that it is therefore responsible for the chemoresistance that occurs upon AurA overexpression (14–17). In addition, a recent report demonstrates that the Aurora kinase inhibitor CCT129202 increases the sensitivity of ABCB1/ABCG2–overexpressing cells and side population cells to chemotherapeutic drugs by inhibiting the function of drug efflux pumps (47). In this study, we showed that AKI603 displayed novel anti-TIC effects with long-term treatment (≥6 days; Figs. 3A and 4A). Short-term treatments with AKI603 did not have obvious anti-TIC effects (Supplementary Fig. S5B). Our study showed that the cells treated with AKI603 contained less TIC population and did not undergo obvious apoptosis (Supplementary Fig. S5A). These results suggested that AKI603 might cause a shift from a TIC population to a non-TIC population instead of inducing TIC death. Thus, short-term treatment (2 days) with AKI603 primarily had effects on non-TICs. Importantly, we observed that a long-term treatment with AKI603 indeed inhibited the epirubicin-induced enrichment of TICs and had synergistic effects with epirubicin on proliferation (Figs. 4A and 5A). Thus, our data also provided a new drug combinational strategy to reduce the dose-related side effects of chemotherapeutics for clinical therapy.

One of the most important characteristics of stem cells is their capacity for self-renewal. Numerous factors, such as Wnt/β-catenin, c-Myc, Sox2, Oct4, and Nanog, play key roles in regulating the balance between the self-renewal and differentiation of stem cells (26–29). For example, Sox2 overexpression increases mammosphere formation, whereas the suppression of Sox2 suppresses mammosphere formation and delays tumor progression in xenograft tumor initiation models (32). In addition, the overexpression of Oct4 promotes the dedifferentiation of melanoma cells to tumor-initiating–like cells, whereas the knockdown of Oct4 in dedifferentiated cells lead to a loss of TIC phenotypes (31). Moreover, Nanog regulates self-renewal of TICs through the insulin-like growth factor pathway in human hepatocellular carcinoma (33). In agreement with these studies, our data suggested that epirubicin significantly stimulated the kinase activity of AurA (Figs. 2C and Fig. 4C). We also found that the protein levels of self-renewal gene, β-catenin, c-Myc, Sox2, Oct4, and Nanog, were upregulated in epirubicin-resistant cells (Fig. 2C). AKI603 repressed AurA activity and decreased β-catenin, c-Myc, Sox2, Oct4, and Nanog expression in a dose-dependent manner (Fig. 3H and Supplementary Fig. S6A). These results implied that AurA kinase activity was related to self-renewal signaling. Further studies should focus on identifying the mechanism by which AurA kinase activity regulates self-renewal gene expression.

In summary, this study unveiled a potential anti-TIC function for the potential Aurora kinase inhibitor AKI603. We also presented a novel mechanism that inhibition of AurA kinase by AKI603 overcame chemoresistance through abolishing TICs in breast cancer, suggesting a novel strategy in cancer treatment.

No potential conflicts of interest were disclosed.

Conception and design: F.-M. Zheng, G. Lu, Q. Liu

Development of methodology: F.-M. Zheng, Z.-J. Long

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F.-M. Zheng, Z.-J. Long, Z.-J. Hou, X.-J. Lai, G. Lu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.-M. Zheng, Z.-J. Long, Z.-J. Hou, J.-L. Xia, E.W.-F. Lam, Q. Liu

Writing, review, and/or revision of the manuscript: F.-M. Zheng, Z.-J. Long, L.-Z. Xu, J.-W. Liu, X. Wang, M. Kamran, M. Yan, S.-J. Shao, E.W.-F. Lam, S.-W. Wang, G. Lu, Q. Liu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F.-M. Zheng, Y. Luo, G. Lu, Q. Liu

Study supervision: E.W.-F. Lam, Q. Liu

The authors thank Yan Zhang, Cai-Feng Yue, Wei Zhang, Na Yang, Yuan-Min Qian, Jia-Jun Xie, Bin He, Jia-Liang Zhang, Rui Gao, Bi-Cheng Wang, and other members of Liu laboratory for their critical comments and technical support.

This work was supported by the National Basic Research Program of China (973 Program; No. 2012CB967000 to Q. Liu), the National Natural Science Foundation of China (No. 81130040 to Q. Liu, No.81000217 to Z.-J. Long, and No. 81201547 to M. Yan), Innovative Research Team in University of Ministry of Education of China (No. IRT13049 to Q. Liu), and the Science and Technology Foundation of Guangzhou (No. 2012J2200077 to Z.-J. Long).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.