Bromodomain and Extraterminal Protein Inhibition Blocks Growth of Triple-negative Breast Cancers through the Suppression of Aurora Kinases

Abstract

Bromodomain and extraterminal (BET) proteins are epigenetic “readers” that recognize acetylated histones and mark areas of the genome for transcription. BRD4, a BET family member protein, has been implicated in a number of types of cancer, and BET protein inhibitors (BETi) are efficacious in many preclinical cancer models. However, the drivers of response to BETi vary depending on tumor type, and little is known regarding the target genes conveying BETi activity in triple-negative breast cancer (TNBC). Here, we show that BETi repress growth of multiple in vitro and in vivo models of TNBC by inducing two terminal responses: apoptosis and senescence. Unlike in other cancers, response to BETi in TNBC is not dependent upon suppression of MYC. Instead, both end points are preceded by the appearance of polyploid cells caused by the suppression of Aurora kinases A and B (AURKA/B), which are critical mediators of mitosis. In addition, AURKA/B inhibitors phenocopy the effects of BETi. These results indicate that Aurora kinases play an important role in the growth suppressive activity of BETi in TNBC. Elucidating the mechanism of response to BETi in TNBC should 1) facilitate the prediction of how distinct TNBC tumors will respond to BETi and 2) inform the rational design of drug combination therapies.

Introduction

Triple-negative breast cancer (TNBC)³ is a heterogeneous disease that comprises ∼15–20% of all breast cancers that can be further subdivided into “basal” and “claudin low” subtypes. As a whole, TNBC is highly aggressive, lacks effective targeted therapies, and conveys poor clinical outcome (1, 2). This is in part due to the lack of estrogen, progesterone, and HER2 receptors. It is therefore critical to develop new targeted treatment strategies that will decrease aggressiveness and improve patient outcomes.

The bromodomain and extraterminal (BET) family of epigenetic readers is a potential therapeutic target in many cancers (3). This family is comprised of four proteins (BRD2, BRD3, BRD4, and BRDT) characterized by two bromodomains that bind acetylated lysines of histone tails and transcription factors (4). The most well characterized BET protein, BRD4, regulates gene transcription and is critical for cell cycle progression and mitosis (5, 6). In cancer, it is enriched at superenhancers that drive the expression of critical oncogenes (7, 8). As a result, inhibition of BRD4 has been shown to selectively suppress key oncogenic drivers (7, 8).

BET inhibitors (BETi) compete with acetylated lysines for binding to the bromodomain pockets of BET proteins, including BRD4 (9, 10). These inhibitors are effective in mouse models of various diseases, including cancer, inflammation, HIV, and heart failure (3). In cancer, they have been proposed to function by disrupting superenhancers that drive expression of key oncogenes such as MYC. This leads to growth arrest, and, in some cases, apoptosis (7, 8, 11). To date, eight BETi are being investigated for safety and efficacy in diverse cancers in early phase clinical trials.

The BET protein gene BRD4 is amplified or overexpressed in ∼17% of basal type breast cancers, the largest subclass of TNBC (12), suggesting that this protein may be a particularly useful target in TNBC. Recent studies have examined the therapeutic impact of BETi in models of TNBC (13–15) and also uncovered mechanism(s) of resistance (16). However, the utility of BETi in different subtypes of TNBC and the genes that control BETi activity in these subtypes remains unknown. Here, we show that BET inhibition halts the growth of multiple TNBC cell lines, independent of their subtype classification as basal and claudin low (17) or according to the six subtypes recently described by Lehmann et al. (18). We also show that BETi inhibit growth of TNBC tumors in multiple mouse models representing different subtypes. Mechanistically, BETi suppress expression of the critical mitosis factors Aurora kinase A and B through direct loss of BRD4 binding to the Aurora kinase promoters, and this is followed by the generation of multinucleated cells that then either senesce or die via apoptosis. Aurora kinase inhibitors also induce polyploidy and apoptosis/senescence in TNBC cells and thus phenocopy BETi. Together, these data reveal that suppression of Aurora kinases mediates the activity of BETi in both the basal and claudin low subtypes of TNBC.

Results

BET Inhibition Blocks Growth of Diverse TNBC Cells without Consistently Down-regulating MYC

Multiple studies have shown that BETi suppress growth of TNBC cells without specific consideration of the distinct TNBC subtypes (13–16). We validated the efficacy of BETi in TNBC cells by treating six cell lines representing the two predominant subtypes of TNBC, basal (HCC1143, MDA-MB-468, and HCC70) and claudin low (MDA-MB-231, BT549, and HCC38), with the prototypical BETi, JQ1. We also assessed BETi responsiveness of MDA-MB-453 cells, which are triple-negative but express androgen receptor and are thus often classified within the luminal breast cancer subtype. As is typical for TNBC, TP53 (p53-encoding gene) is mutated to transcriptionally inactive isoforms in all seven cell lines tested (19). Within 72 h, JQ1 inhibited growth of all seven cell lines in a dose-dependent manner, independent of TNBC subtype (Fig. 1, A–C). Qualitatively similar effects as JQ1 were observed with two additional BETi, I-BET151, and I-BET762 (Fig. 1, D and E), indicating JQ1-induced growth suppression was due to inhibition of BET protein function and not off target effects.

FIGURE 1.

BET inhibition blocks growth of TNBC cells without consistently down-regulating MYC. A–C, relative fold change in cell number (compared with initial plating density) for seven TNBC cell lines representing the basal (A), claudin low (B), and luminal androgen receptor (C) subtypes of TNBC. The cells were treated with increasing concentrations of JQ1, and viable cells were counted after 72 h. D and E, growth curves of TNBC cells treated with increasing doses of I-BET151 (D) or I-BET762 (E). Viable cells were counted after 72 h. F and G, RT-qPCR (F) and Western blotting analysis (G) of MYC expression levels in TNBC cell lines treated for 24 h with vehicle or 500 nm JQ1. The values on the Western blot are relative to the vehicle-treated 1143 sample following normalization to β-actin. For all graphs, the data are means ± S.D. *, p < 0.05 compared with vehicle.

In numerous cancer types, the effects of BETi are due to their suppression of MYC expression; often MYC transcription is significantly decreased with BET protein inhibition, and overexpression of MYC reverses the effects of these drugs (20, 21). To determine whether MYC repression is necessary for the BETi response in TNBC, we treated the seven TNBC cell lines with JQ1 for 24 h and assessed the response of MYC mRNA and protein expression. Although BETi suppressed growth of all seven cell lines, the impact of JQ1 on MYC mRNA and protein was highly variable (Fig. 1, F and G), in some cases having no effect, and did not correlate with TNBC subtype or the GI₅₀ for JQ1 (Table 1). For example, MDA-MB-231 and HCC38 cells are among the most growth-suppressed, yet MYC protein expression is unchanged in response to JQ1 treatment. These data indicate that suppression of MYC is not essential for BETi to inhibit the growth of TNBC cells. This result contrasts with luminal breast cancer models where growth inhibition by BETi is in part mediated by suppression of MYC (22).

TABLE 1.

TNBC subtypes and responses to BET inhibition

Cell line	Gray classificationa	Pietenpol classificationa	p53 statusb	GI50c	MYC expressiond	Senescencee	Apoptosise
				nm
HCC1143	Basal A (basal)	Basal-like 1	Inactivating mutation, missense	800	0.33	++	+
MDA-MB-468	Basal A (basal)	Basal-like 1	Inactivating mutation, missense	470	0.41		++
HCC70	Basal A (basal)	Basal-like 2	Inactivating mutation, missense	930	0.31		++
MDA-MB-231	BasalB (claudin-low)	Mesenchymal stem-like	Inactivating mutation, missense	270	0.81	++	+
HCC38	BasalB (claudin-low)	Basal-like 1	Inactivating mutation, missense	240	1.13		+++
BT549	BasalB (claudin-low)	Mesenchymal	Inactivating mutation, missense	2000	0.97		++
MDA-MB-453	Luminal	Luminal Androgen Receptor	Inactivating mutation, deletion	220	0.60		+++

^a Seven TNBC cell lines organized according to subtype based on two classification systems: Gray (35) and Pietenpol (18).

^b “Inactivating mutation” is defined as a mutation that lacks the ability to regulate the transcription of known p53 target genes, such as CDKN1A.

^c GI₅₀ of JQ1 after 72 h of treatment.

^d Relative MYC expression after 24 h of treatment with 500 nm JQ1 compared with vehicle, measured by quantitative RT-PCR.

^e Responses to prolonged (up to 8 days) JQ1 treatment. ++ and +++ represent the time to and extent of response to BET inhibition.

Sustained BET Inhibition Induces Apoptosis and Senescence in TNBC Cells

To determine whether the growth suppression of TNBC cells by BETi was due to an induction of apoptosis, we treated four cell lines (MDA-MB-468, HCC1143, MDA-MB-231, and BT549; two basal and two claudin low, respectively) with vehicle or JQ1 for 72 h and stained them for pyknotic nuclei with Hoechst. JQ1 increased the number of apoptotic cells in all four cell lines (Fig. 2A). However, two of the lines (MDA-MB-468 and BT549) had a more pronounced apoptotic response than the other two. Notably, the extent of apoptosis was again independent of TNBC subtype; both a claudin low and a basal line were highly apoptotic. Surprisingly, MDA-MB-231 and HCC1143 cells exhibited greater growth suppression than BT549 cells (Fig. 1, A and B), yet they displayed fewer apoptotic cells. This suggested that these two cell lines may undergo a different cellular response in addition to apoptosis. Consistent with this possibility, expression of p21, a protein involved in the induction of senescence, was greatly elevated in MDA-MB-231 and HCC1143 cells within 4 days of JQ1 treatment (Fig. 2B). In contrast, the two cell lines displaying a greater degree of JQ1-induced apoptosis (MDA-MB-468 and BT549) only modestly increased p21 levels. The increase in p21 suggested that MDA-MB-231 and HCC1143 cells may undergo senescence in response to BETi. Supporting this postulate, within 8 days of JQ1 treatment, MDA-MB-231 and HCC1143 cells became flattened, had a greater cytoplasmic to nuclear ratio, and stained positively for senescence-associated β-galactosidase (SA-βgal) activity (Fig. 2, C and D), all of which are hallmarks of senescence. Within 14 days, most JQ1-treated MDA-MB-231 and HCC1143 cells expressed SA-βgal (Fig. 2, C and D), indicating the majority of cells had undergone senescence. Colony forming assays with MDA-MB-231 and HCC1143 cells revealed that JQ1-induced senescence is irreversible (Fig. 2, E and F). Thus, sustained BETi exposure induces two terminal responses: apoptosis and senescence. These responses are independent of TNBC subtype, impact on MYC expression, and the GI₅₀ of JQ1 (Table 1).

FIGURE 2.

Sustained BET inhibition induces apoptosis and senescence in TNBC cells. A, the indicated TNBC cell lines were treated with vehicle or JQ1 (250, 500, or 1000 nm) for 72 h and stained with Hoechst. The percentage of pyknotic nuclei was then determined. The data are means ± S.E. *, p < 0.05 compared with vehicle. B, Western blots for p21 and β-actin in TNBC cell lines treated for 24 h with vehicle or 500 nm JQ1. C and D, top panels, representative phase images (10×) of MDA-MB-231 (C) and HCC1143 (D) cells treated with vehicle or JQ1 for 8 days. Middle and bottom panels, representative images (4×) of MDA-MB-231 (C) and HCC1143 (D) cells treated with vehicle or JQ1 for 8 days (middle panels) or 14 days (bottom panels) and stained for SA-βgal activity. E and F, representative images of plates of MDA-MB-231 (E) and HCC1143 (F) cells examined for their ability to form colonies after removal of JQ1. Following 14 days of treatment with vehicle or JQ1, 500 (231) or 1000 (1143) live cells were seeded and grown in complete medium in the absence of JQ1 for 11 additional days. The resulting colonies were stained with crystal violet. Triplicate wells are shown for each concentration of JQ1. (B), basal; (C), claudin low.

BET Inhibition Abrogates TNBC Tumor Growth

To determine whether the distinct cell fates (senescence versus apoptosis) would translate to differential tumor responses in vivo, we compared the impact of BETi on tumor growth using two xenograft mouse models. First, mice harboring palpable orthotopic xenografts of one of the cell lines that primarily undergoes senescence (MDA-MB-231) were treated with vehicle or JQ1 for 28 days. JQ1 suppressed the growth of these tumors (Fig. 3A), and at the conclusion of this study, the average final tumor size was smaller in JQ1-treated mice compared with those in vehicle-treated animals (p < 0.01; Fig. 3B). JQ1 also suppressed the incidence of liver metastasis (p < 0.05; Fig. 3C) compared with vehicle-treated mice. A similar study was performed using a cell line that primarily undergoes apoptosis in response to BETi (MDA-MB-468). These tumors had a more robust response to JQ1 treatment than those derived from MDA-MB-231 cells. Within 5 days, MDA-MB-468 tumors treated with JQ1 began to regress (Fig. 3D). Using RECIST criteria (23), JQ1 treatment resulted in one mouse having progressive disease, five having partial regression, and the remainder having stable disease after 35 days (Fig. 3E). Thus, in vivo models mimicked our in vitro observations, because tumors formed from a predominantly apoptotic cell line partially regressed in response to BETi, whereas BETi suppressed the growth but did not elicit regression of tumors generated from the cell line that senesces. In both tumor models, JQ1 regulation of MYC expression followed a similar pattern as the in vitro studies (Fig. 3, F and G), providing in vivo evidence that at least in MDA-MB-231 tumors, MYC repression is unnecessary to mount a growth inhibitory response to BETi. Lastly, to assess the impact of BETi in a more clinically relevant model, we evaluated BETi efficacy in a patient-derived xenograft (PDX) model, PDX BCM-4013 (24), classified as basal by PAM50 (25). Similar to MDA-MB-231 tumors, JQ1 blunted growth of tumors formed from this PDX (Fig. 3H).

FIGURE 3.

BET inhibition abrogates tumor growth. A, NOD/scid/γ mice with palpable MDA-MB-231 tumors were treated with vehicle or JQ1. The graph shows relative tumor size over 28 days of treatment. The data are means ± S.E. B, Waterfall plot showing the percentage of change in tumor size for each mouse at the end of the study compared with the first day of treatment. Each bar represents an individual tumor. C, after 28 days of treatment, surface liver macrometastases were counted. Each dot represents the number of liver metastases within an individual mouse. D, NOD/scid/γ mice with palpable MDA-MB-468 tumors were treated with vehicle or JQ1. The graph shows relative tumor size over 35 days of treatment. The dashed line indicates initial tumor size. The data are means ± S.E. E, waterfall plot showing the percentage of change in size for each MDA-MB-468 tumor at the end of the study. RECIST values (23) are indicated by horizontal lines and shown to the right of the graph. PD, progressive disease; SD, stable disease; PR, partial regression. F, RT-qPCR analysis of MYC expression in vehicle and JQ1-treated tumors (MDA-MB-231 tumors n = 5/group, MDA-MB-468 tumors n = 10/group). The data are means ± S.D. *, p < 0.05 compared with vehicle; comparing vehicle- and JQ1-treated MDA-MB-231 tumors, p = 0.21. G, quantitation of Western blotting analysis of MYC expression levels in MDA-MB-231 tumors treated with vehicle or JQ1 for 28 days. MYC protein expression was normalized to β-actin. The values are means ± S.D. H, relative tumor size of PDX BCM-4013 tumors treated with vehicle or JQ1 (n = 10/group). The data are means ± S.E. (B), basal; (C), claudin low.

As previously reported for multiple other mouse models, there was no change in body weight for mice with MDA-MB-231 or PDX tumors that were treated with JQ1 for 28–30 days compared with vehicle-treated mice (Fig. 4A). In contrast, JQ1-treated mice with MDA-MB-468 tumors did weigh less than their vehicle-treated counterparts at 32 days post-treatment (Fig. 4A). However, this is likely due to extensive tumor growth in the vehicle-treated group that increased the overall weights of these mice. Further demonstrating selectivity of drug response in breast tumors compared with normal tissues, non-tumor-bearing female mice treated with vehicle or JQ1 displayed no differences in mammary gland architecture, proliferation (phospho-histone H3 staining), or apoptosis (TUNEL staining) (Fig. 4, B and C, and data not shown). Thus, BETi lack toxic side effects in the mammary gland. Together, these data reveal that BET protein inhibition selectively suppresses tumor growth in numerous in vivo models of TNBC, as well as metastatic progression with minimal to no toxicity.

FIGURE 4.

BETi lack toxic side effects and do not impact normal adult mouse mammary gland morphology or proliferation. A, mice from all three tumor studies were weighed once per week. The data are means ± S.E. B, representative whole mounts of inguinal mammary glands isolated from FVB/N adult female mice treated with vehicle or JQ1 for 1 week. The glands were stained with Carmine alum. C, quantitation of phospho-histone H3-positive cells in sections of mammary glands from mice treated with vehicle or JQ1 for 1 week. The data are means ± S.E. *, p < 0.05 compared with vehicle, for all graphs. (B), basal; (C), claudin low.

Aurora Kinases Are Downstream Targets of BETi

To begin to assess the mechanism(s) by which BETi induce apoptosis and senescence in TNBC cells, cellular morphology was examined for several days after exposure to JQ1. This revealed that the cells become multinucleated regardless of whether they undergo apoptosis or senescence, or their particular TNBC subtype, and multinucleation occurred prior to the induction of either cell fate (Fig. 5A). Within 8 days, nearly half of JQ1-treated MDA-MB-231 cells were tetraploid (40.5 ± 5.9%) (Fig. 5B). These data suggest that BETi disrupt mitosis and/or cytokinesis, both of which are known to induce apoptosis and/or senescence depending on the cell line being examined.

FIGURE 5.

Aurora kinases are downstream targets of BETi. A, left panel, representative images (20×) of the indicated cell lines treated with vehicle or 500 nm JQ1 for 96 h and stained with DAPI (blue, nuclei) and Texas Red-X phalloidin (red, actin cytoskeleton). Insets show examples of multinucleated cells. Arrows indicate multinucleated cells. Bars, 50 μm. Right panel, quantitation of multinucleated cells. The data are means ± S.D. B, MDA-MB-231 cells were treated for 8 days, stained with propidium iodide, and analyzed by flow cytometry. The data are means ± S.E. (p < 0.05 compared with vehicle for both 2N and 4N populations). C, RT-qPCR analysis of AURKA and AURKB expression in the indicated TNBC cell lines treated with vehicle or 500 nm JQ1 for 24 h. The data are means ± S.D. D, Western blotting analysis of AURKA and AURKB expression in four TNBC cell lines treated with vehicle or 500 nm JQ1 for 24 h. The values on the Western blot are relative to untreated samples per cell line following normalization to β-actin. E, representative gene-specific ChIP-PCR analysis of MDA-MB-231 cells assessing binding of BRD4 to AURKA and AURKB. CDKN1A was used as a control that does not lose BRD4 binding with JQ1 treatment. Negative primers were designed to areas outside the promoter region for each gene. The data are means ± S.D. F, RT-qPCR analysis of AURKA and AURKB expression in MDA-MB-231 and MDA-MB-468 tumors from mice treated for 28 (MDA-MB-231 cells) or 35 (MDA-MB-468 cells) days with vehicle or JQ1. The data are means ± S.D. *, p < 0.05 compared with vehicle, for all graphs. (B), basal; (C), claudin low.

It is well established that deregulation of Aurora kinases A or B (AURKA or AURKB) induces polyploidy (26–28). Both kinases play critical roles in mitosis (29), and a previous study reported that BRD4 stimulates transcription of AURKB (30), although it was not shown whether this was a direct effect. This led us to determine whether Aurora kinases may mediate BETi response in TNBC by first examining the expression of Aurora kinases following JQ1 treatment. Both AURKA and AURKB mRNA and protein were down-regulated within 24 h of JQ1 treatment in four TNBC cell lines, regardless of whether the cell lines ultimately senesce or apoptose in response to the drug (Fig. 5, C and D). JQ1 also disrupted the binding of BRD4 to AURKA and AURKB but did not reduce binding to CDKN1A, the gene encoding p21 (Fig. 5E). Both Aurora kinase transcripts were also suppressed in JQ1-treated xenografted tumors (Fig. 5F). The effects of JQ1 on AURKA and AURKB expression and binding of BRD4 to their respective genes were again independent of TNBC subtype. Together, these data indicate that Aurora kinases are direct targets of BET inhibition through disruption of BRD4 binding to their respective genes.

Aurora kinases are precisely regulated throughout the cell cycle. Overexpression or silencing of these proteins elicits mitotic dysfunction, precluding restoring their cyclic expression to assess their involvement in the BETi response (31, 32). To circumvent this limitation, we used a selective AURKB inhibitor, AZD1152 (AZD) (33), to determine whether loss of AURKB can phenocopy the effects of BETi. Indeed, 100 nm AZD induced multinucleation in MDA-MB-231 and MDA-MB-468 cells (Fig. 6A) similar to that observed with JQ1. Moreover, 72 h of treatment with 100 nm AZD caused the different cell lines to primarily undergo the same differential cell fates of apoptosis or senescence in a manner that corresponded to their response to JQ1. Specifically, a greater proportion of MDA-MB-468 and BT549 cells underwent apoptosis compared with MDA-MB-231 and HCC1143 cells (Fig. 6B). MDA-MB-231 and HCC1143 cells, on the other hand, permanently senesced, as evidenced by increased cytoplasmic/nuclear ratio, enhanced SA-βgal activity, and decreased colony formation (Fig. 6, C and D). Thus, inhibition of AURKB elicited the same cellular responses as JQ1 (Fig. 6E). We also tested whether an AURKA inhibitor, MLN8237 (34), phenocopies BETi and found it generated similar responses as JQ1 in the MDA-MB-231 and MDA-MB-468 cell lines (Fig. 6, A, F, and G). Together, these data indicate that suppression of AURKA and/or AURKB activity causes the same effects as BET inhibition, specifically polyploidy, apoptosis, and senescence, and that the particular cellular fate (i.e. senescence or apoptosis) was consistent within each cell line regardless of whether aurora kinases or BET proteins were inhibited.

FIGURE 6.

Aurora kinase inhibitors phenocopy BETi. A, representative images (20×) of MDA-MB-231 and MDA-MB-468 cells treated with vehicle, 100 nm AZD1152 (AZD), or 250 nm MLN8237 (MLN) for 96 h. The cells were stained with DAPI (blue, nuclei) and Texas Red-X phalloidin (red, actin cytoskeleton). Arrows indicate multinucleated cells. Bars, 50 μm. B, the indicated TNBC cell lines were treated with vehicle, 1000 nm JQ1, or 100 nm AZD for 72 h and stained with Hoechst. The percentages of pyknotic nuclei were then determined. The data are means ± S.E. C, representative images (4×) of MDA-MB-231 (top panels) and HCC1143 (bottom panels) cells treated with vehicle, 1000 nm JQ1, or 100 nm AZD for 8 days and stained for SA-βgal activity. D, representative image of colony forming assays using MDA-MB-231 cells treated for 14 days with vehicle or 100 nm AZD. E, summary of the predominant responses of seven TNBC cell lines to JQ1 and AZD treatment. F, MDA-MB-231 and MDA-MB-468 cells were treated with vehicle, 1000 nm JQ1, or 250 nm MLN8237 for 72 h and stained with Hoechst. The data are means ± S.E. G, representative images (4×) of MDA-MB-231 cells treated with vehicle or 250 nm MLN8237 for 8 days and stained for SA-βgal activity. *, p < 0.05 compared with vehicle, for all graphs. (B), basal; (C), claudin low.

Discussion

TNBC is the most aggressive subtype of breast cancer due, in part, to its lack of effective targeted therapies. Although BETi have previously been shown to reduce invasiveness of TNBC cells in vitro (14) and inhibit tumor growth in xenograft models (13, 14, 16), the utility in different subtypes of TNBC and the mechanism by which BETi elicit their effects on growth have not been previously established. Here, we report that BET inhibition results in growth suppression of TNBC cell lines independent of their intrinsic subtype, including claudin low and basal subtypes (35, 36), as well as five of the six more recently defined TNBC subtypes described by Lehman et al. (18). Although suppression of MYC expression is essential for the effects of BETi in other cancers (20–22), this is not the case in a subset of TNBC cell lines, because we and others (16) have found that BETi-mediated growth inhibition occurs independently of MYC down-regulation. We also report that BETi reduce tumor growth in three TNBC xenograft models and decrease the incidence of liver metastasis in a highly aggressive model while having no impact on the normal mammary gland. To our knowledge, these data are the first to demonstrate that BETi suppress distal metastasis in vivo for any tumor type. Through the use of diverse in vitro and in vivo models of TNBC, our data together with the recently published study by Polyak and co-workers (16) provide clear preclinical evidence of efficacy of BETi in TNBC models, supporting the future assessment of these drugs in clinical trials of patients with this spectrum of disease. Notably, BETi induce minimal to no toxicity in these animal models, suggesting high selectivity for cancers rather than normal tissues. This selectivity may be due to the differential responsiveness of normal and cancer cells to disruption of mitotic and cytokinetic events. Although normal cells maintain an ability to arrest to ensure accurate cell division, cancer cells lack these mechanisms and ultimately abort aberrant cell division by either dying or activating a senescence program (37).

Although BETi have been shown to inhibit growth or induce cell death of many tumor models, the mechanism(s) underlying either response to these drugs has not been fully explored. As indicated above, some tumor types are highly reliant on MYC, and BETi have been shown to suppress this gene in these models. However, in TNBC, the mechanism of action of these drugs appears more complex and likely involves the suppression of AURKA and AURKB and the resulting induction of multinucleation. AURKB has previously been suggested to be a target of BRD4, because its expression is reduced upon BRD4 silencing (30, 38, 39). This serine/threonine kinase is essential for cytokinesis, phosphorylation of histone H3, and appropriate spindle attachment to the kinetochore during metaphase (40). Overexpression or loss of AURKB induces multinucleation followed by either apoptosis or senescence in other cellular models (26–28, 41). Another family member, AURKA, also plays a critical role in mitosis by mediating centrosome maturation and duplication, bipolar spindle assembly, alignment of chromosomes, and cytokinesis (40). Here, we show BETi rapidly down-regulate both AURKA and AURKB, and this occurs with the reduction of BRD4 binding to their respective promoters. Furthermore, use of selective AURKA and AURKB inhibitors revealed that blocking the activity of either protein phenocopies the polyploidy, apoptosis, and senescence phenotypes induced by BETi in TNBC cells. Notably, the cell-specific fates that occur in response to AURKA/AURKB inhibitors mimicked those of BETi. Together, these data indicate that the mechanism of action of BETi in TNBC involves direct suppression of the AURKA and AURKB genes. What remains unclear is the cell-specific mechanism underlying the choice to undergo apoptosis or senescence in response to BET or Aurora kinase inhibitors. Thus far, we have found that the extent to which cells undergo JQ1-induced apoptosis and/or senescence is independent of TNBC subtype, changes in MYC expression, the extent of AURKA/AURKB suppression, or the JQ1 GI₅₀. It will now be important to identify the pathways driving the choice of cells to undergo BETi-induced senescence versus apoptosis to reveal biomarkers of therapeutic response, as well as to identify approaches that ensure cell death in tumors treated with BETi. It will also be important to determine whether cell lines that are representative of other types of cancer, including luminal breast cancer, also respond to BETi by suppressing AURKA/AURKB and altering ploidy or whether this effect is specific to TNBC.

In summary, inhibition of BET proteins in diverse forms of TNBC reduces the expression of AURKA and AURKB, critical factors for normal cell division. This results in growth arrest and polyploidy. Cells respond to the ploidy defects by either undergoing senescence or cell death. In two mouse models of TNBC, BETi suppresses tumor growth while it induces regression in a third. These responses are consistent with the ability of BETi to induce senescence or apoptosis in a cell line-dependent manner. Lastly, BETi reduce liver metastasis in a highly aggressive TNBC model. Together, these data reveal that Aurora kinases play a key role in the response of TNBC to BETi and provide preclinical evidence supporting the future use of BETi in diverse subtypes of TNBC to suppress tumor growth and metastasis. Moreover, they suggest that changes in AURKA and/or AURKB expression may serve as valuable biomarkers to predict therapeutic response.

Experimental Procedures

Cell Culture and Reagents

The cell lines were obtained from the American Type Culture Collection. The cells were maintained at 37 °C with 5% CO₂. MDA-MB-231, HCC38, MDA-MB-468, HCC1143, BT549, and HCC70 cells were grown in RPMI 1640 supplemented with 10% FBS. For BT549 cells, 0.023 IU/ml insulin was added to the media. MDA-MB-453 cells were grown in DMEM supplemented with 10% FBS. JQ1, I-BET151 (MedChem Express), I-BET762 (Cayman Chemical), AZD1152 (AdooQ Bioscience), and MLN8237 (AdooQ Bioscience) were dissolved in DMSO. For growth curves, the cells were treated with the indicated drugs for 72 h and trypsinized. Viable cells were identified by trypan blue exclusion and counted on a hemocytometer.

Flow Cytometry

Cell cycle analyses were performed as previously described (42) with the following modification: the cells were harvested, fixed in ice-cold 70% ethanol, suspended in propidium iodide/RNase solution, and analyzed using Attune NxT Flow Cytometer (Thermo Fisher). During analysis, gating was performed to remove doublets from the results.

RNA Analysis

RNA was isolated using TRIzol reagent (Ambion; 15596018) and treated with DNase I (Ambion; AM1906). Reverse transcription was performed using SuperScript II reverse transcriptase (Invitrogen; 18064-014). Quantitative real time PCR was performed on an Applied Biosystems Step One Plus real time PCR system using the following TaqMan Gene Expression Assays (Thermo Fisher): 1) AURKA (Hs01582072_m1), 2) AURKB (Hs00945858_g1), 3) MYC (Hs00153408_m1), and 4) GAPDH (Hs02758991_g1).

Western Blotting Analysis

Western blotting analysis was performed as previously described (43) using the following primary and secondary antibodies: c-Myc (Cell Signaling; 9402), p21 (Cell Signaling; 2947), AURKA (Cell Signaling; 4718), AURKB (Cell Signaling; 3094), β-actin (Sigma; A1978), goat anti-rabbit IgG-HRP (Santa Cruz; sc-2054), and goat anti-mouse IgG-HRP (Santa Cruz; sc-2005). The blots were developed using Pierce ECL Western blotting substrate (Thermo Scientific). The bands were quantified using ImageJ software.

Senescence-associated β-Galactosidase Activity Stain

The cells were fixed in PBS-buffered 2% formaldehyde and 0.2% glutaraldehyde for 5 min at room temperature. The cells were then washed three times in PBS and stained (5 mm potassium ferricyanide, 5 mm potassium ferrocyanide, 2 mm MgCl₂, 150 mm NaCl, 30 mm citric acid/phosphate buffer, pH 6, 1 mg/ml X-gal; Invitrogen, 15520) for 12–16 h at 37 °C.

Colony Forming Assay

The cells were treated with vehicle or JQ1 for 14 days and then harvested. For each treatment group, 500 (MDA-MB-231) or 1000 (HCC1143) viable cells were seeded in three 35-mm plates and evenly dispersed. The cells were then grown in drug-free media for 11 days. The colonies were fixed and stained (0.05% crystal violet, 1% formaldehyde, 1% methanol, and 10% 10× PBS) for 20 min at room temperature.

Mouse Xenograft Studies

All in vivo experiments were performed with approval from the Institutional Animal Care and Use Committee at Case Western Reserve University, which is certified by the American Association of Accreditation for Laboratory Animal Care. The mice were housed in microisolator units, given standard sterile chow and water ad libitum, and maintained on a 12-h light/dark cycle. Xenografts were implanted into the two inguinal mammary glands of adult (2–4 month old) female NOD/scid/γ mice. Tumor size was measured twice per week using calipers. Mouse weight was measured once per week to assess toxicity. Upon completion of treatments, the tumors were removed and processed for sectioning, RT-qPCR, and Western blotting analysis.

For MDA-MB-231 xenografts, mice with tumors 120 ± 50 mm³ in size were randomized into two treatment groups of five mice each, vehicle (1:1 propylene glycol:water) or JQ1 (50 mg/kg) i.p. daily, and tumor size was measured with calipers for 28 days. At this time, livers were harvested, and surface macrometastases were counted. For MDA-MB-468 xenografts, mice with tumors 90 ± 20 mm³ in size were randomized into 2 treatment groups of 10 mice each vehicle or JQ1 (50 mg/kg) i.p. daily, and tumor size was measured with calipers for 32 days. For the patient-derived xenograft, BCM-4013, mice with tumors 260 ± 150 mm³ in size were randomized into 2 treatment groups of 10 mice each, vehicle or JQ1 (50 mg/kg) i.p. daily, and tumors were measured for 30 days.

Analysis of Mammary Gland Morphology, Apoptosis, and Proliferation

To assess the impact of BETi on normal mammary glands, 10 adult female FVB/N mice were treated with either vehicle or JQ1 (50 mg/kg i.p. daily) for 1 week. For each mouse, both inguinal mammary glands were collected and either processed for whole mounts or sectioned for TUNEL and phospho-histone H3 staining. For whole mounts, mammary glands were isolated, fixed, and processed as previously described (43). For analysis of apoptosis and proliferation, mammary glands were collected, fixed in 4% paraformaldehyde for 6 h, paraffin-embedded, and sectioned by the Case Western Reserve University Tissue Procurement and Histology Core Facility. TUNEL staining was performed according to the manufacturer's instructions (ApopTag Plus peroxidase in situ apoptosis detection kit; Millipore). For phospho-histone H3 staining, mammary gland sections were cleared and hydrated by washing twice each with xylene, 100% ethanol, and 95% ethanol, and once with PBS. Antigen retrieval was accomplished using a Biocare Medical decloaking chamber at 125°F for 10 min in 10 mm citrate buffer (pH 6). The sections were then washed three times with PBS, and peroxidase was blocked using Dako EnVision Plus kit (Dako; K4011) supplemented with 15 μl/ml goat serum. Phospho-histone H3 (Ser¹⁰) antibody (Cell Signaling; 9701) diluted in PBST with 5% BSA and 15 μl/ml goat serum was added, and sections were incubated overnight at 4 °C. After two PBS washes, the slides were stained and developed with Dako Envision + System-HRP (Dako; K4011). Specifically, the secondary antibody was incubated for 1.5 h at room temperature. The sections were then washed twice with PBS and counterstained with Hematoxylin (Gill's III; CS402-1D) and Scott's bluing solution. Sections were dehydrated prior to mounting by washing once each with 95% ethanol and100% ethanol and twice with xylene.

Nuclear Morphology

The cells were seeded onto sterile glass coverslips and treated with vehicle (DMSO), JQ1, AZD1152, or MLN8237 for 4 (MDA-MB-468) or 8 (MDA-MB-231 and HCC1143) days. The cells were fixed with 3.7% formaldehyde, 1× PBS, permeabilized with 0.1% Triton X-100, 1× PBS, and blocked with 1% BSA, 1× PBS. F-actin was labeled with Texas Red-X phalloidin (Invitrogen; T7471). The nuclei were counterstained with Vectashield hard set mounting medium with DAPI (Vector Labs; H-1500). To identify cells undergoing apoptosis, the cells were stained with 10 μm Hoechst 33342 (Thermo Scientific; 62249) for 10 min at 37 °C. The number of cells with and without pyknotic nuclei were counted, and the percentage of apoptotic cells was calculated.

Gene-specific Chromatin Immunoprecipitation

ChIP-PCR was performed as previously described (44). MDA-MB-231 cells were treated for 24 h with vehicle or 500 nm JQ1. Chromatin was immunoprecipitated with a BRD4-specific antibody (Bethyl Laboratories; A301-985A50) or control mouse IgG (Sigma, I5281). The following promoter-specific primer sequences were used: CDKN1A, forward 5′-GCCTCCCTCCATCCCTATG-3′ and reverse 5′-CAGCCCAAGGACAAAATAGC-3′; AURKA, forward 5′-AGGACAAGGGCCTTCTTAGG-3′ and reverse 5′-TAGTGGGTGGGGAGACAGAC-3′; and AURKB, forward 5′-AGCCGTGAGAAGCAGAGAAA-3′ and reverse 5′-ATTGGGGCTAGTGTGCTGAC-3′. As a negative control, the following gene-specific primer sequences designed against regions outside of the promoter region were used: CDKN1A, forward 5′-GCAGCAGGGGAGGAAAGTAT-3′ and reverse 5′-CCCCATGCTGTTCTCGTAAC-3′; AURKA, forward 5′-ATCTCTGGCACAGAATTCCAG-3′ and reverse 5′-TTTGTCTGGTTTCTCCACTGT-3′; and AURKB, forward 5′-TATATCCCAAAGCCCCAGAG-3′ and reverse 5′-ATGTCCCCAGTGAACTCCAA-3′.

Statistics

Statistical analyses were performed using two-tailed Student's t test (in vitro data) and Mann-Whitney U test (in vivo data). p values less than 0.05 were considered statistically significant. All in vitro data are represented as mean values from three independent experiments performed in triplicate.

Author Contributions

J. M. S. and R. A. K. designed the study and wrote the manuscript with input from all other authors. J. M. S., S. S. G., K. L. W. B., L. C. V., J. L. Y., B. W., E. K. R., and D. D. S. performed the experiments and analyzed the data. J. E. B. provided JQ1. M. D. L. and J. C. C. provided the patient-derived xenograft model. R. A. K. supervised the study.