Introduction

Aurora kinases comprise three mammalian serine/threonine kinases (Aurora A, B, C) that play a critical role in mitosis and cell division. The primary role of Aurora A relates to centrosome maturation and mitotic spindle assembly. Aurora B is essential for chromosome condensation, spindle attachment, and cytokinesis. The function of Aurora C is predominantly restricted to meiosis [1, 2]. Aurora kinases are frequently overexpressed in cancer, including breast cancer, and play an important role in oncogenesis [3, 4].

Inhibition of Aurora A leads to mitotic spindle abnormalities and accumulation of cells in mitosis, whereas inhibition of Aurora B leads to chromosome alignment defects. These effects then lead to failed cytokinesis or endoreduplication, induction of polyploidy, and eventually cell death [5–9]. In preclinical models an increase in polyploid tumor cells is specifically correlated with loss of p53 function [10–12]. Cells that lack functional p53 have an increased ability to reenter the cell cycle [13–15]. p53-dependent arrest of tetraploid cells is sometimes referred to as the “G1 tetraploidy checkpoint” and its existence remains controversial [16]. This checkpoint is hypothesized to play a role in removal of tetraploid cells either via irreversible G1 cell cycle arrest or by apoptosis [16–18].

More than a dozen small-molecule Aurora kinase inhibitors have been tested in clinical trials (reviewed in [19, 20]). They differ in their specificity and potency against the three Aurora kinase family members. To date, no consistent molecular predictors of response to Aurora kinase inhibitors have been defined [21–23]. Consequently, a subset of breast cancer patients most likely to benefit from treatment has yet to be identified.

AMG 900 is a novel, potent, orally bioavailable, and highly selective pan-Aurora kinase inhibitor currently being evaluated in Phase I clinical trials [24]. AMG 900 was shown to inhibit the enzyme activity of all three Aurora kinase family members with IC50 values ≤ 5 nmol/L [25]. AMG 900 has been previously shown to be active in multidrug-resistant tumor cell lines and to have a considerable activity in several xenograft models, including breast cancer [25]. We hypothesized that a distinct molecular subgroup of breast cancers may be more likely to respond to AMG 900. To identify this subgroup, we evaluated response to AMG 900 across a large panel of well-characterized breast cancer cell lines.

Methods

Cell lines, cell culture, and reagents

The cell line panel included 41 breast cancer and 3 immortalized breast epithelial cell lines representing the known molecular subgroups of breast cancer and was described in detail previously [26–28].

AMG 900 was obtained from Amgen Inc. (Thousand Oaks, CA) and diluted in DMSO.

TP53 mutation analysis

TP53 mutation detection in DNA extracted from the breast cell lines was performed using two different methods in parallel; the arrayed primer extension (APEX; Asper Biothech) [29] and the temporal temperature gradient gel electrophoresis (TTGE) [30]. The APEX TP53 assay is a microarray constructed of oligo-nucleotides to capture TP53 mutations in exon 2–9 by primer extension and fluorescent-labeled terminator nucleotides. The mutation of interest is characterized directly using APEX. The TTGE is a pre-screening method covering exon 2–11 based on a mutation sensitive denaturation of DNA using a combination of chemicals and temperature. Gene alterations detected by TTGE are uncovered as aberrant migrating bands in gel electrophoresis, and need subsequent characterization. Samples showing diverging results from the two methods were validated by sequencing, using the ABI 377 DNA Sequencer (Applied Biosystems, Foster City, CA).

Detailed information on cell lines, proliferation assay, microarray analysis, flow cytometry analysis of cell cycle and apoptosis, Western blots, and statistical analysis have been described previously [26, 27] and can be found in Online Resource Supplementary Methods.

Results

AMG 900 has potent anti-proliferative effects in breast cancer cell lines regardless of subtype

AMG 900 inhibited growth very effectively in all of the breast cancer cell lines in our panel, even at low nanomolar concentrations. The IC50 values for all but one cell line (UACC-812; IC50 = 15.2 nM) were <10 nM which is well below the clinically achievable plasma concentrations (unpublished data obtained from Amgen). This potency made it difficult to stratify the cell lines as sensitive or resistant to AMG 900 based on IC50 as the response metric. Alternatively, we categorized the cell lines into two groups: highly sensitive versus less sensitive. This binary classification was made using a cutoff of 10 % lethality (defined as a decrease in cell number from baseline) at 10 nM of AMG 900. Those cell lines that had greater than 10 % lethality were classified in the highly sensitive group, and, conversely, those cell lines that were below 10 % were classified as less sensitive. Based on these response criteria, 25 cell lines were classified as highly sensitive and 19 cell lines as less sensitive to AMG 900 (Table 1).

Table 1 The calculated response parameters for each cell line and the response classification
Full size table

The most common clinically relevant breast cancer biomarkers including HER2 amplification status [31] and ER (estrogen receptor) status and subtypes (luminal or non-luminal subtypes) were analyzed for an association with response, but no association was found to be statistically significant (Supplementary Table S1).

Association analysis of Aurora kinase copy number variations (CNVs) and expression levels with response to AMG 900

We investigated the relationship between Aurora kinase levels and response to AMG 900 in our cell line panel. AURKA gene amplification was highly prevalent in our panel of cell lines. 13 of our cell lines have greater than 2-fold amplification (log2(ratio) > 1) of the AURKA gene (Table 2). Cell lines with AURKA amplification were more likely to be classified as highly sensitive to AMG 900 but this association failed to reach the cutoff for statistical significance (RR = 2.53 (95 % CI 0.8–8.0), p = 0.09) (Table 3). AURKB and AURKC CNVs were not observed in our panel.

Table 2 Aurora kinases and response to AMG 900
Full size table
Table 3 Statistical analysis of predictors of response to AMG 900
Full size table

AURKA, AURKB, and AURKC baseline mRNA expression levels by microarray did not associate with response to AMG 900 (Tables 2, 3). Similarly, Aurora A and Aurora B baseline protein levels measured by Western blot did not associate with response to AMG 900 in our panel (Table 3, Supplementary Table S2; Supplementary Figure S1).

p53 dysfunction predicts for response to AMG 900

Given the interaction of p53 and Aurora kinases at regulating cell cycle progression, we investigated the relationship between p53 dysfunction and response to AMG 900 in our cell line panel. TP53 somatic mutations found in our genotyping analysis were cross-referenced against functional data derived from yeast transactivation assays performed on common mutations, as listed in the IARC p53 database [32] (http://www-p53.iarc.fr/) (Supplementary Table S3). Mutations whose functional status was not listed were assumed to be loss of function (LOF). We observed a higher frequency of TP53 LOF mutants among highly sensitive cell lines (22 of 25; 88.0 %) compared to less sensitive cell lines (9 of 19; 47.3 %) (Table 4). This association was found to be statistically significant in our analysis (RR = 1.86, 95 % CI 1.1–3.1, p = 0.004) (Table 3).

Table 4 p53/p21 and response to AMG 900
Full size table

We also investigated the role of p21 in response to AMG 900 due to the tight functional association between p21 and p53. Baseline p21 protein levels in the entire cell line panel were measured by Western blot (Table 4; Fig. 1; Supplementary Table S4). High p21 protein expression correlated with classification of p53 functional status (Pearson r = 0.58, p value < 0.0001). Consequently, cell lines with low p21 protein levels were also far more likely to be classified as highly sensitive to AMG 900 (RR = 2.28, 95 % CI 1.4–5.8, p = 0.0004) (Table 3).

Fig. 1
figure 1

p21 baseline protein levels. p21 baseline protein levels were measured by Western blots on the cell line panel as described in Supplementary Methods. Cell lines were grouped based on their sensitivity to AMG 900 as described in Table 1; a highly sensitive, b less sensitive cell lines. Highly sensitive cell lines have lower p21 protein expression compared to less sensitive cell lines. α-tubulin was used as a loading control; plus asterisk common control (MCF-7); densitometry data available in Supplementary Table S4

Full size image

To verify that AMG 900 induces a p53/p21 response, a p53/p21 protein expression time course experiment was performed in a subset of cell lines with either wild-type (wt) or mutated TP53 and variable sensitivity to AMG 900. Significant increases in p53 and p21 expression were observed 24 and 48 h after treatment with AMG 900. These effects were more pronounced in cell lines with wild type p53 (Fig. 2).

Fig. 2
figure 2

p53 and p21 time course analysis with AMG 900. p53/p21 protein expression time course Western blot experiment was performed in a subset of a p53-functional cell lines with either wt TP53 (MCF-7, KPL-1, CAL-51) or TP53 mutation that does not lead to LOF (HCC-2218), or b cell lines with TP53-LOF mutations (HCC-38, MDA-MB-361, HCC-1187, SK-BR-3, T-47D). Cells were treated with 10 or 100 nM AMG 900 for 10 min to 48 h. Significant increases in p53 and p21 expression were observed 24 and 48 h after treatment with AMG 900. These effects were more pronounced in cell lines with functional p53. α-tubulin was used as a loading control

Full size image

AMG 900 effects on cell cycle and apoptosis

To investigate the mechanism of response to AMG 900, its effect on cell cycle was analyzed by flow cytometry in a subset of cell lines (n = 18) with variable sensitivity to AMG 900 and p53 functional status. After a 24- and 48-h treatment with 10 nM of AMG 900 we observed 4 N DNA accumulation in all tested cell lines except for UACC-812. After a 48-h treatment, polyploidy was observed in 11 of 18 cell lines (Fig. 3). Polyploidy was achieved in nearly all cell lines tested (17 of 18 cell lines) after 5 days of treatment. The onset of polyploidy was independent from the sensitivity to AMG 900 or p53 functional status (data not shown).

Fig. 3
figure 3

AMG 900 induced ≥4 N DNA accumulation. The effects of AMG 900 on cell cycle were determined by flow cytometry in a subset of cell lines with a variable sensitivity to AMG 900 and p53 functionality. Cells were treated with 10 nM of AMG 900 for 24 h to 5 days. Examples of cell lines where AMG 900 caused a ≥4 N DNA accumulation, b 4 N DNA accumulation

Full size image

The effects of AMG 900 on apoptosis were analyzed by flow cytometry in a subset of cell lines (n = 10) with variable sensitivity to AMG 900 and p53 functional status (Fig. 4). Cells were treated with 10 nM of AMG 900 for 5 days. Highly sensitive cell lines with TP53 LOF mutations showed a strong induction of apoptosis. The apoptotic effects were also present, albeit at lesser degree, in the highly sensitive wt TP53 MDA-MB-134 cell line and in less sensitive cell lines with TP53 LOF mutations. AMG 900 caused minimal or no changes in apoptosis in less sensitive wt TP53 cell lines.

Fig. 4
figure 4

The effects of AMG 900 on apoptosis. The effects of AMG 900 on apoptosis were analyzed by flow cytometry in a subset of cell lines with variable sensitivity to AMG 900 and p53 functionality. Cells were treated with 10 nM of AMG 900 for 5 days. a Highly sensitive TP53-LOF cell lines showed highest amounts of AMG 900-induced apoptosis (Annexin-V positive cells). b Highly sensitive p53-functional MDA-MB-134 cell line and c less sensitive TP53-LOF cell lines showed less induction of apoptosis. d Less sensitive p53-functional cell lines show a minimal or no induction of apoptosis after treatment with AMG 900

Full size image

Discussion

In this study, we have evaluated the anti-proliferative activity of AMG 900, a novel pan-Aurora kinase inhibitor, on a panel of 41 cell lines representing various subtypes of breast cancer and three immortalized breast epithelial cell lines. We found that AMG 900 is highly active in breast cancer cell lines at concentrations achievable in clinical settings, with IC50 values below 10 nM in all but one cell line. Due to the difficulty in stratifying cell lines based on IC50 values alone, other outcome measures were considered for use in classifying response to AMG 900. Lethality is a less commonly used outcome measure than IC50 in cell proliferation assays. Percentage lethality was calculated at each concentration and can be interpreted as the percent of cell death from baseline after the 5-day treatment. We believe that  % lethality may provide a more accurate quantification of cytotoxic effects as opposed to IC50, which is better suited for quantification of cytostatic or growth inhibitory effects. A subset of the cell lines in our panel (25 of 44 cell lines) showed significant lethality (>10 %) at 10 nM of AMG 900 and were classified as highly sensitive. The cell lines with no significant lethality were classified as less sensitive (19 of 44 cell lines). In this study, the 10 nM cutoff point was chosen because, at this concentration, AMG 900 inhibits all three Aurora kinases and has minimal off-target effects [25]. Additionally, the 10 nM concentration is readily achievable in human plasma following treatment with AMG 900. In the current study, AMG 900 was more potent than in a previous report [25]. These differences are likely explained by the different methodologies used between the studies for calculating growth inhibition.

Breast cancers represent a genetically and phenotypically heterogeneous group of tumors. This heterogeneity is well represented in our panel of breast cancer cell lines. Despite the fact that several Aurora kinase inhibitors have already been tested in clinical trials in breast cancer, no subpopulation of patients has been identified that would benefit [21–23]. Preclinical studies with Aurora kinase inhibitors that differed in target specificity using large panels of breast cancer cell lines did not lead to a definite conclusion as far as which subtype associates with response. In vitro, response to GSK1070916 (Aurora kinase B/C inhibitor) was shown to be associated with a “claudin-low” subtype (which overlaps with our post-EMT group, Supplementary Table 1), whereas VX-680 (pan-Aurora kinase inhibitor) response was associated with cell lines that lacked HER2 amplification [21]. ENMD-2076 (Aurora A/B/angiogenic kinase inhibitor) and AS703569 (pan-Aurora kinase inhibitor) showed more activity in triple-negative breast cancer cell lines [22, 23]. In this study, we did not see a significant enrichment for any of the most commonly used clinical subgroups (HER2± , ER±, luminal/non-luminal) in our subset of cell lines classified as highly sensitive to AMG 900.

Aurora kinase inhibitors were initially developed upon the observation that Aurora kinase expression is commonly upregulated in many cancers [3, 4]. We therefore investigated the association between the expression levels of Aurora kinases and response to AMG 900. Cell lines with AURKA amplification as measured by aCGH were more likely to be classified as highly sensitive to AMG 900, but this association did not reach statistical significance (p = 0.09). AURKA baseline mRNA levels by microarrays and Aurora A baseline protein levels did not associate with response to AMG 900. No cell lines had observed copy number changes for the AURKB or AURKC genes. AURKB and AURKC baseline mRNA levels and Aurora B baseline protein levels (Aurora C baseline protein levels were not measured in this study) did not associate with response to AMG 900. These data indicate that neither gene copy number changes nor expression of Aurora kinases A/B/C are good predictive markers of response to AMG 900 in breast cancer cell lines.

TP53 mutations are found in approximately 30 % of breast cancers [33]. However, the phenotype of these mutations is variable and much uncertainty remains as to whether p53/p21 function correlates with sensitivity or resistance to chemotherapeutic agents [34–36]. Several studies have made conflicting observations regarding p53 status and response to Aurora kinase inhibitors. Some investigators have observed that Aurora kinase inhibitors have increased apoptotic or anti-proliferative activity in p53-deficient cells [37–39], while others have found that these effects were more pronounced in cell lines with wt p53 [40–42], and yet others, that the effects of Aurora kinase inhibition are independent of p53 [43–47].

Using a carefully characterized database of p53 function in our panel of breast cancer cell lines, we observed that TP53 LOF mutations were significantly more frequent in cell lines classified as highly sensitive to AMG 900 (p = 0.004; in vitro sensitivity = 0.88, specificity = 0.62). In addition to p53, we found that low protein expression of p21, a downstream target of p53, associated with increased sensitivity to AMG 900 (p = 0.0004). We found that baseline p21 protein expression correlated with p53 functional status in our panel. However, we did not have a sample large enough to investigate the independent effects of p21 outside of p53, so it remains unclear whether low p21 expression is an independent predictor of response to AMG 900 or simply a surrogate biomarker for p53 dysfunction.

Analyses of cell cycle and apoptosis by flow cytometry were conducted to elucidate the mechanisms of AMG 900 activity. We observed polyploidy in a vast majority of tested cell lines within 5 days of treatment with AMG 900. The onset of polyploidy was independent from sensitivity to AMG 900 or p53 functional status. The presence of polyploidy after the treatment with AMG 900 in both wt TP53 and p53-deficient cell lines is in agreement with previous studies with pan-Aurora kinase inhibitors (VX-680, danusertib) or Aurora B inhibitor ZM447439 [12, 48, 49]. However, induction of apoptosis was observed in every TP53 mutant, AMG 900-highly sensitive cell line we tested. Apoptosis induction was less common in less sensitive, wild-type TP53 cell lines. These findings of an increased apoptosis in TP53 LOF cell lines corroborate the results of the study conducted by Gizatullin et al. [39], who found that cell lines with a compromised p53/p21 postmitotic checkpoint function are more likely to undergo apoptosis after treatment with VX-680 than cells with intact checkpoint function. Furthermore, Kaestner et al. [9] have shown that the p53-dependent postmitotic G1 checkpoint was not required for the induction of apoptosis after treatment with a selective Aurora B inhibitor ZM447439 but was required after treatment with Aurora kinase A inhibitor MLN8054.

In this study, we did not attempt to determine the specific causal effects of inhibition of individual Aurora kinases on cell cycle regulation and apoptosis. Payton et al. [25] have shown previously that AMG 900 effectively blocks the autophosphorylation of Aurora A and B, as well as the phosphorylation of histone H3, a proximal substrate of Aurora B. Pan-Aurora kinase inhibitors generally induce cellular phenotypes that are compliant with perturbed Aurora B function [50]. Similar phenotype findings were previously reported for AMG 900, as evidenced by aborted cell division without a prolonged mitotic arrest [25].

A time-course Western blot experiment was conducted in a subset of cell lines with variable sensitivity to AMG 900 and p53/p21 status. Significant increases in p53 protein expression accompanied with an induction of p21 protein expression were observed 24–48 h post-treatment with AMG 900 in cell lines with functional p53. Conversely, these effects on p53/p21 expression were not as prominent in cell lines with dysfunctional p53. The up-regulation of p53 and p21 following Aurora kinase inhibition has been reported previously [8, 25, 38, 51], and further confirms the functional interactions between Aurora kinases and the p53/p21 complex.

These data lead us to hypothesize that the functional p53/p21 complex may play a role in preventing cells with AMG 900-induced polyploidy from undergoing apoptosis. This mechanism could explain the differential response to treatment observed in our proliferation experiments, where TP53 LOF mutant cell lines displayed more lethality at 10 nM AMG 900 as well as more induction of apoptosis. We further hypothesize that the nearly ubiquitous induction of polyploidy by AMG 900 may underlay the consistent and strong growth inhibition we observed (as evidenced by the low IC50 values).

In summary, we have found that AMG 900 is highly active in breast cancer cell lines and identified biomarkers that predict for response to this compound in vitro. Specifically, we have described the novel observation that there is a strong association between p53/p21 functional status and response to Aurora kinase inhibition in breast cancer. This observation requires clinical validation incorporating TP53 somatic mutation analysis and/or p21 expression to identify those patients most likely to benefit from treatment.