The Aurora Kinase Inhibitor VX-680 Induces Endoreduplication and Apoptosis Preferentially in Cells with Compromised p53-Dependent Postmitotic Checkpoint Function

The Auroras comprise a family of serine/threonine kinases that are essential for mitotic progression. The mammalian kinases, Aurora A, Aurora B, and Aurora C, share similar catalytic domains, but show different subcellular localization (1). Aurora A localizes to duplicated centrosomes and the spindle poles and participates in processes required for assembly of the spindle apparatus, including centrosome maturation and separation (2). In contrast, Aurora B is a chromosome passenger protein, localized to centromeres during prometaphase and relocated to the spindle midphase during anaphase onset (3). Aurora B is required for the phosphorylation of histone H3, chromosome segregation, and cytokinesis. Aurora C has also been shown to be a chromosome passenger protein that can complement Aurora B function and is required for cytokinesis (4, 5).

Aurora kinases are strongly associated with cancer. Overexpression of Aurora A transforms Rat1 and NIH/3T3 cells and gives rise to genetically unstable aneuploid cells with multiple centrosomes (6). Elevated expression has been detected in a high percentage of colon, breast, ovarian, gastric, and pancreatic tumors, and in a subset of these tumors, the AURKA locus (20q13) is amplified (7–10). Aurora B and Aurora C are also expressed at high levels in both primary tumors and cell lines (10). Levels of Aurora B have been shown to increase as a function of Duke stage in primary colon cancers (11). The expression of Aurora B often parallels that of Aurora A (7, 12).

To date, a growing number of inhibitors of Aurora kinases have been described (12) including Hesperadin (13), ZM447439 (14), and VX-680 (15). ZM447439 is an ATP-competitive inhibitor of Aurora A and Aurora B, whereas VX-680 inhibits all three family members (K_i's of 0.6, 18 and 4.6 nmol/L for Aurora A, B, and C, respectively). VX-680 treatment results in cells with high levels of cyclin B1 and 4N DNA content 8 to 12 hours after release from a G₁-S block, indicating that cells can enter mitosis. The peak of cyclin B1 expression is slightly later in treated cells than in control cells, suggesting a delay in cell cycle progression. However, in VX-680-treated and control cultures, cyclin B1 expression subsequently decreases, indicating that mitotic exit is not blocked. After the completion of mitosis, control cells divide to yield daughters with 2N DNA content. In contrast, VX-680 induces the accumulation of cells arrested in a pseudo-G₁ state with a 4N DNA content or the accumulation of cells with >4N DNA content, the latter population representing cells that exit mitosis and subsequently proceed through S phase in the absence of cell division. Continued proliferation in the presence of aberrant mitosis and failed cytokinesis presumably results in cell death (15). ZM447439 and Hesperadin-treated cells both enter and exit mitosis with normal kinetics, demonstrating no effect on cell cycle progression, but similarly result in the accumulation of cells with 4N or >4N DNA content. The defective cytokinesis, as well as the inhibition of phosphorylation of histone H3 by all three drugs, suggests that the cellular effects of these agents might be largely mediated by the disruption of Aurora B function (12).

Whether cells arrest with 4N DNA content in pseudo-G₁ or endoreduplicate with the accumulation of >4N DNA content likely depends on the integrity of the p53-dependent postmitotic checkpoint (16, 17). This checkpoint has been reported to induce the G₁ arrest of tetraploid cells that arise following mitotic errors including spindle disruption and cytokinetic failure, and involves p53-dependent induction of p21^Waf1/Cip1, resulting in the suppression of cdk2 activity (18, 19). There is current controversy over whether this postmitotic checkpoint is truly a “tetraploid checkpoint,” because in some experimental systems, it is not triggered by tetraploidy, cytokinetic failure, or aberrant centrosome number (20–22). Nonetheless, p53 is frequently activated in tetraploid cells. In this way, p53 adds to the spindle checkpoint to prevent the propagation of cells with an abnormal genome. Consistent with the role of p53 in constraining endoreduplication following Aurora inhibition, the expression of dominant-negative p53 in U2OS osteosarcoma cells exaggerated the polyploidy induced by ZM447439 (14).

However, endoreduplication has been reported in response to ZM447439 and VX-680 in p53 wild-type cells (14, 15), suggesting that p53 status alone does not determine the response to Aurora inhibition. Furthermore, it has not yet been clarified whether cell death after exposure to Aurora kinase inhibitors depends on the absence of p53. To examine these issues further, we have explored the role of p53 in the response to VX-680. Here, we show that disruption of p53 function in wild-type A549 non–small cell lung cancer (NSCLC) cells causes increased endoreduplication after VX-680 treatment, followed by induction of apoptosis. However, in some p53 wild-type cells, including RKO colon carcinoma cells, endoreduplication is poorly constrained and is associated with cell death. In these cells, induction of p21^Waf1/Cip1 is delayed and suppression of cyclin E-cdk2 kinase activity and retinoblastoma (Rb) protein phosphorylation is incomplete. Furthermore, in A549 cells expressing short interfering RNA (siRNA) pools targeting p21^Waf1/Cip1, polyploidy is substantially increased, confirming the critical role of this cdk inhibitor in limiting endoreduplication. These data suggest that the overall integrity of the p53-p21^Waf1/Cip1 pathway in the G₁ postmitotic checkpoint governs the response to VX-680, and that cells with compromised checkpoint function are most likely to undergo apoptosis.

Cell culture. A549, Calu-1, NCI-H1299, SK-LU-1 and Calu-6 NSCLC, MCF-7 breast carcinoma, HeLa cervical carcinoma, RKO and RKO-E6 colon carcinoma, and U2OS osteosarcoma cell lines were obtained from the American Type Culture Collection and cultivated in the recommended medium. A549, MCF-7, RKO, and U2OS cells express wild-type p53. In Calu-1 and NCI-H1299 cells, p53 is deleted; SK-LU-1 and Calu-6 cells carry p53 mutations. A549 control and A549-E6 cells, generated via retroviral infection with LXSN or LXSN-HPV16E6, have been previously characterized and were maintained in medium supplemented with 400 μg/mL of G418 (23). In all instances, A549 parental cells and A549 LXSN cells behaved identically in response to VX-680.

Drug treatment. A 10 mmol/L stock of VX-680 was prepared in DMSO and stored at −20°C. Subconfluent cultures (5 × 10⁵-10⁶ cells per 10 cm dish) were seeded 24 hours prior to drug treatment. To assess the suppression of phosphorylation of histone H3 by VX-680, cells were first treated with 0.4 μg/mL of nocodazole (Sigma-Aldrich, Co., St. Louis, MO) for 16 hours followed by the addition of DMSO or drug for 3 hours (14).

Fluorescence-activated cell sorting analysis, bromodeoxyuridine analysis, and detection of apoptosis by flow cytometry. For cell cycle analysis, cells were analyzed for DNA content by flow cytometry using ModFit (Verity Software House, Topsham, ME) and CellQuest software (BD Biosciences, Franklin Lakes, NJ); the latter was used for quantification of polyploidy. For bromodeoxyuridine (BrdUrd) analysis, the In situ Cell Proliferation Kit, FLUOS (Roche Applied Science, Indianapolis, IN) was used according to the manufacturer's protocol. Briefly, cells were pulse-labeled with 5-bromo-2′-deoxyuridine 1 hour before fixation, denaturation, and incubation with FITC-conjugated anti-BrdUrd antibody. Cells were stained with propidium iodide and analyzed by two-color flow cytometry. For apoptosis assays, a fluorescein apoptosis detection kit was used for terminal deoxynucleotidyltransferase (TdT)–mediated nick end labeling (TUNEL) assay (Promega, Madison, WI), as previously described (23). Following formaldehyde and ethanol fixation, pooled adherent and nonadherent cells were incubated with fluorescein-12 dUTP in the absence or presence of TdT, stained with propidium iodide and analyzed for DNA content and apoptosis using two-color flow cytometry. Apoptosis was quantified as the percentage of cells shifting to fluorescein-positivity in the presence of TdT, and was confirmed by Western blotting with an anti–cleaved PARP antibody (Cell Signaling Technology, Beverly, MA).

Protein extraction and Western blot analysis. Whole cell extracts were prepared in NP40 lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, and 1% NP40], containing protease and phosphatase inhibitors (Calbiochem, San Diego, CA). In order to analyze the Rb protein, nuclear extraction was done using a nuclear extraction kit (Pierce, Rockford, IL) according to the manufacturer's protocol. For analysis of phosphohistone H3, cells were resuspended in radioimmunoprecipitation assay buffer [50 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 0.5% sodium deoxycholate, 1% NP40, and 0.1% SDS] and briefly sonicated twice. Protein concentration was determined using Bradford assay (Bio-Rad, Richmond, CA). Typically, 20 μg of cellular protein was subjected to SDS-PAGE and subsequent Western blotting using standard procedures, with the following primary antibodies: anti-p53, anti-p21^Waf1/Cip1, anti-Rb 4H1, anti-cyclin B1 V152 (all from Cell Signaling Technology); anti-β-actin (Sigma-Aldrich); anti-Ku70 (BD Transduction Laboratories, San Diego, CA); anti-cyclin E HE12 (Santa Cruz Biotechnology, Santa Cruz, CA); anti-Aurora A and anti-Aurora B (BD Biosciences PharMingen, San Jose, CA); and anti-phosphohistone H3 (ser10, clone 3H10; Upstate Biotechnology, Lake Placid, NY). Where indicated, bands were quantified using AlphaEase FC Software (Alpha Innotech Corp., San Leandro, CA).

Immunoprecipitation and histone H1 kinase assays. Cell lysates were prepared using protease and phosphatase inhibitors (Calbiochem) and subjected to immunoprecipitation with either anti-cyclin E HE111 or anti-cyclin B1 H433 antibodies (Santa Cruz Biotechnology). Immunoprecipitates were used to direct phosphorylation of exogenous histone H1 (Roche Applied Science) in the presence of 20 μmol/L of ATP and 10 μCi of [γ-³³P]ATP. Phosphorylated histone H1 was detected following SDS-PAGE, transfer to nitrocellulose, and exposure to X-ray film. Bands were quantified using AlphaEase FC Software (Alpha Innotech).

siRNA transfection. “Smart-pool” nonspecific oligos and those containing siRNA duplexes targeting Aurora A, Aurora B, p53, and p21^Waf1/Cip1 were obtained from Dharmacon, Inc. (Lafayette, CO). Cells were plated at 0.75 × 10⁵ in six-well plates. Twenty-four hours later, transfection was done with 2 to 4 μL of OligofectAMINE (Invitrogen, Carlsbad, CA) using 50 nmol/L of the appropriate siRNA pools. In the analysis of combined Aurora A and B depletion, cells were collected 72 hours after transfection for flow cytometry and protein analyses. Forty-eight hours after transfection with p53 or p21^Waf1/Cip1 siRNAs, cells were replated and subsequently treated with VX-680 for 48 or 72 hours followed by flow cytometry and protein analyses.

Generation of NCI-H1299 cells inducibly expressing p21^Waf1/Cip1. The coding sequence of p21^Waf1/Cip1, derived from the plasmid containing the human cDNA clone (OriGene Technologies, Rockville, MD), was subcloned into pTRE-Hyg. The resulting construct was transfected into NCI-H1299 TetOn cells using Fugene 6 (Roche Applied Science). Clones were selected in 250 μg/mL of hygromycin (Invitrogen) and tested for inducible expression of p21^Waf1/Cip1 in the presence of doxycycline (Sigma-Aldrich). NCI-H1299 TetOn cells and the TRE2-Hyg plasmid were provided by Dr. Andrew Phillips (Medical College of Georgia, Augusta, GA).

Statistical analysis. Statistical analysis was done with the two-tailed, unpaired Student's t test. P < 0.05 was considered statistically significant.

VX-680-induced endoreduplication is enhanced in the absence of p53 function. As shown in Fig. 1A, treatment of A549 p53 wild-type NSCLC cells with VX-680 induces the accumulation of cells with 4N and 8N DNA content, similar to that described with ZM447439 (14). The predominant effect of VX-680 at 24 hours is the accumulation of cells with 4N DNA content. However, at 48 hours, endoreduplication occurs in only a minority of cells, with ∼20% demonstrating 8N DNA content (Fig. 2B). No further increase in the degree of polyploidy was observed up to 120 hours of drug exposure. In contrast, in A549 cells expressing the HPV16-E6 oncoprotein, in which p53 function is abrogated, endoreduplication is significantly enhanced at the 48-hour time point, with >50% of the population demonstrating 8N DNA content. In addition, endoreduplication continues with time, as evidenced by the appearance of cells with 16N DNA content at 72 hours and beyond (Fig. 1A and B). This is reflected in incorporation of BrdUrd in which pulse-labeling was done between 71 and 72 hours after VX-680 exposure; in A549 cells, there is a marked diminution in DNA synthesis, whereas in A549-E6 cells, DNA synthesis clearly continues, consistent with continued endoreduplication (Fig. 1C).

Enhanced endoreduplication induced by VX-680 in p53-negative cells is accompanied by loss of viability. Over time, in A549-E6 cells, cell death occurs, as shown by viable cell counts at the 72- and 96-hour time points (Fig. 1D). This is accompanied by the appearance of a sub-G₁ peak in the DNA content analysis (Fig. 1A), confirmed to represent apoptosis by TUNEL assay (Fig. 1E and F) and by assessment of PARP cleavage (Fig. 1G). In TUNEL assays, 8N and 16N most readily shifted to fluorescein positivity, suggesting that endoreduplicating cells were the most likely to undergo apoptosis (Fig. 1F). In contrast, in control cells, viability is maintained, and sub-G₁ DNA content, significantly increased TUNEL positivity, and PARP cleavage are not detected.

These results were confirmed in several additional cell lines. MCF-7 cells also express wild-type 53 and behave similarly to A549 cells after VX-680 treatment. Over a 72-hour exposure, most cells are arrested with 4N DNA content, and only a small proportion show 8N DNA content (Fig. 2A and B). In contrast, substantial endoreduplication is detected 24 to 48 hours after VX-680 treatment in cells in which p53 is deleted (Calu-1 and NCI-H1299), mutated (SK-LU-1 or Calu-6), or in which p53 is inactivated by HPV16E6 oncoprotein expression (HeLa). By 72 hours, 16N DNA content is apparent. These results are reflected in combined siRNA-mediated codepletion of Aurora A and Aurora B. In MCF-7 cells, this manipulation increases the 4N DNA content without inducing endoreduplication; in HeLa cells, the degree of increase of 4N DNA content is greater and endoreduplication is detected (Fig. 2C).

In MCF-7 cells, apoptosis is minimal, as detected by TUNEL assay (Fig. 2F) and confirmed by the absence of PARP cleavage (Fig. 2D). NCI-H1299 and HeLa cells show a greater proportion with sub-G₁ DNA content and DNA fragmentation than the other cell lines at the 72-hour time point (Fig. 2A,, E, and F), indicating that loss of viability occurs more rapidly in some cell types than others. Nonetheless, evidence of PARP cleavage can be shown in other p53-negative cell lines as well, including Calu-1 (Fig. 2D).

Endoreduplication and apoptosis are readily detected in some p53 wild-type cell lines. RKO and U2OS cells also express wild-type p53. In contrast to the behavior of A549 and MCF-7 cells, in these cell lines, VX-680 causes progressive endoreduplication until 72 hours (with only a slight further increase at 96 hours) that is then followed by apoptosis (Fig. 3A). Similarly, siRNA-mediated depletion of Aurora A and Aurora B together from U2OS cells results in the appearance of cells with 8N and sub-G₁ DNA content (Fig. 3B). Abrogation of p53 function still enhances the degree of endoreduplication and cell death, as shown by the analysis of RKO-E6 cells (Fig. 3C and D). Nonetheless, clearly not all cell lines expressing wild-type p53 respond to Aurora kinase inhibition with predominant 4N arrest and limited apoptosis.

Delayed p21^Waf1/Cip1 induction and persistent cyclin E-cdk2 activity with Rb phosphorylation occurs in response to VX-680 in RKO and U2OS cells. In order to ascertain why endoreduplication was more pronounced in RKO and U2OS cells than in A549 and MCF-7 cells, we examined components of the p53-dependent postmitotic checkpoint after 24 and 48 hours of exposure to VX-680. Figure 4A shows the induction of p53 after VX-680 treatment by 24 hours in all four of the cell lines, a response that persists at the 48-hour time point as well. In addition, at 24 hours, induction of p21^Waf1/Cip1 is robust in A549 and MCF-7 cells. In contrast, the induction of p21^Waf1/Cip1 is delayed in RKO and U2OS cells; at the 24-hour time point, there is only a slight increase in p21^Waf/Cip1 in RKO cells, whereas no increase is detected in U2OS cells until the 48-hour time point. The delay in p21^Waf1/Cip1 induction in RKO cells, compared with A549 cells, was confirmed in a more detailed time course over the first 48 hours of VX-680 exposure (Fig. 4D).

Consistent with differences in the timing of induction of p21^Waf1/Cip1 in these cell lines, the kinetics and degree of inhibition of cyclin E-cdk2 activity also differed. As shown in Fig. 4B, cyclin E-dependent kinase was reduced by ∼80% at 24 hours and 98% at 48 hours in A549 cells, the time point after which further endoreduplication is not detected. In contrast, cyclin E-dependent kinase activity is decreased by only 27% at 24 hours and 73% at 48 hours in RKO cells, accounting for continued endoreduplication observed at 72 hours in these cells. At the 72-hour time point, cyclin E-dependent kinase activity still persists in RKO cells, decreased by 93% (data not shown); between 72 and 96 hours, there is continued slight progression of endoreduplication (Fig. 3). These differences are also reflected in the degree of Rb phosphorylation after VX-680 exposure (Fig. 4A). In A549 cells and MCF-7 cells, Rb is largely dephosphorylated by 24 hours. In RKO and U2OS cells, phosphorylated Rb is detected at both the 24- and 48-hour time points, consistent with continued S phase entrance and endoreduplication. Finally, consistent with the arrest of cells with 4N DNA content at the G₁-S boundary (pseudo-G₁ arrest), A549 cells show persistent cyclin E expression but decreased cyclin B expression after VX-680 exposure, reflected in a substantial diminution in cyclin B-dependent kinase activity (Fig. 4D and E). In contrast, in endoreduplicating RKO cells, cyclin B expression and cyclin B-dependent kinase activity are not compromised.

Induction of p21^Waf1/Cip1 correlates with suppression of endoreduplication in a concentration-dependent manner. As additional evidence that the amount of p21^Waf/Cip1 induced after exposure to VX-680 correlates with the degree of endoreduplication that occurs, A549 cells were treated with a range of VX-680 concentrations. As shown in Fig. 5A, at 24 hours, there is a concentration-dependent increase in the amount of p21^Waf1/Cip1 induced. Full suppression of phosphohistone H3 phosphorylation in these cells is seen at VX-680 concentrations >200 nmol/L. At 48 hours after treatment with >200 nmol/L of VX-680, endoreduplication is constrained in a concentration-dependent manner, correlating with increased induction of p21^Waf1/Cip1 at higher concentrations (Fig. 5B,, bottom left). The decrease in cells with 8N DNA content is accompanied by a reciprocal increase in the percentage of cells with 4N DNA content, indicating more potent arrest in pseudo-G₁ at these concentrations. Similar data were obtained at 72 hours (data not shown). In contrast, the degree of endoreduplication plateaus and remains maximal irrespective of VX-680 concentration in A549-E6 cells (Fig. 5B , bottom right).

siRNA targeted depletion of p53 and p21^Waf1/Cip1 from A549 cells increases endoreduplication in response to VX-680. To further confirm the role of the p53-p21^Waf1/Cip1 pathway in constraining endoreduplication in response to VX-680, we used siRNA pools to transiently deplete these proteins from A549 cells. In the absence of p53, there is progressive endoreduplication in response to 48 or 72 hours of VX-680 exposure, as was seen in A549 cells expressing the HPV16E6 oncoprotein (Fig. 6A and B). Endoreduplication after exposure to VX-680 for 48 hours is also substantial after depletion of p21^Waf1/Cip1 (Fig. 6A and B). This effect persists at the 72-hour time point. In the context of this experiment, progressive endoreduplication between 48 and 72 hours occurred but was slight, perhaps because of nonspecific induction of p53 after introduction of the siRNA targeting p21^Waf1/Cip1, limiting the depletion of p21^Waf1/Cip1 achieved over time (Fig. 6C).

Induction of p21^Waf1/Cip1 in p53-negative NCI-H1299 cells limits endoreduplication and cell death. Several NCI-H1299 p21^Waf1/Cip1-inducible clones were generated. Results for a representative clone are shown in Fig. 6D and E. In untreated cells, there is a strong induction of p21^Waf1/Cip1 in the presence of doxycycline (which is not detected in its absence). Interestingly, VX-680 treatment itself causes a small degree of p53-independent induction of p21^Waf1/Cip1 in these cells. Nonetheless, the amount of p21^Waf1/Cip1 induced is insufficient to prevent endoreduplication after 48 hours of drug exposure, with the appearance of a substantial percentage of cells with either 8N or 16N DNA content. In contrast, in the presence of doxycycline, the degree of endoreduplication is sharply reduced at the 48-hour time point. Quantification of cells with 8N or 16N DNA content in the absence or presence of doxycycline after 48 hours of VX-680 treatment is shown for this clone and for one additional clone in Fig. 6F. In the absence of doxycycline, cells with sub-G₁ DNA content are detected by 72 hours, consistent with the induction of apoptosis; the percentage of cells with sub-G₁ DNA content is substantially reduced in the presence of doxycycline (Fig. 6E).

Aurora kinase inhibitors, including Hesperadin, ZM447439, and VX-680, all induce similar effects on cell division in vitro. The predominant outcome of exposure to these agents is not inhibition of cell cycle progression. Rather, following genome replication, drug-treated cells are able to enter and exit an aberrant mitosis, with suppression of spindle checkpoint function. Because of failed cell division, tetraploidy results (12).

Following defective cytokinesis, the effect of longer exposure to Aurora kinase inhibition is cell line–dependent. Some cells undergo arrest in a pseudo-G₁ state, evidenced by a decrease in cyclin B in cells maintaining 4N DNA content, as shown in A549 cells (Fig. 4D). Other cells undergo additional rounds of DNA synthesis and with continued defective cytokinesis, develop polyploidy. Whether there is arrest in a tetraploid state or further endoreduplication most likely depends on the integrity of the postmitotic checkpoint (16, 17). This checkpoint is p53-dependent, although there is currently controversy over whether tetraploidy truly triggers this checkpoint and signals p53 activation (21). Nonetheless, induction of this checkpoint causes stable pseudo-G₁ arrest, as evidenced by the limited endoreduplication that occurs in A549 and MCF-7 cells beyond 48 hours, even with persistent exposure to VX-680 up to 120 hours. Consistent with the role of p53 in determining the degree of endoreduplication after Aurora kinase inhibition, expression of the HPV16E6 oncoprotein or siRNA pools targeting p53 in A549 cells results in a significantly greater degree of endoreduplication than that observed in parental cells.

However, the degree of endoreduplication in response to VX-680 is not governed by the presence of p53 alone. Some cell lines that express wild-type p53 still show substantial endoreduplication, including RKO and U2OS cells. We have shown that in these cell lines, there is a delay in the induction of p21^Waf1/Cip1. Because of this delay, at earlier time points (e.g., at 48 hours), the amount of p21^Waf1/Cip1 induced is insufficient to fully suppress cyclin E-cdk2 activity and Rb phosphorylation, permitting continued S phase entry and endoreduplication. In general, once maximal induction of p21^Waf1/Cip1 has occurred (24 hours in A549 and MCF-7 cells and 48 hours in RKO and U2OS cells), endoreduplication continues to progress for an additional 24 hours before there is >90% suppression of cyclin E-cdk2 activity with either no further increase or only minimal increase in polyploidy. Of note, the abrogation of p53 function in RKO cells permits progressive endoreduplication beyond 72 hours, whereas the degree of polyploidy remains minimally increased beyond this time point in parental cells (Fig. 3).

The importance of p21^Waf1/Cip1 in contributing to the constraint of endoreduplication was shown in experiments in which its level was modulated. In A549 cells, the amount of p21^Waf1/Cip1 induced by VX-680 is concentration-dependent. In the experiment shown in Fig. 5, the degree of endoreduplication observed is most likely the result of a balance between the potency of Aurora inhibition and the amount of p21^Waf1/Cip1 induced. At low concentrations (<200 nmol/L), there is probably insufficient p21^Waf1/Cip1 to fully restrict endoreduplication, and accumulation of cells with 8N DNA content increases with increasing concentrations of VX-680 and increasing degrees of Aurora inhibition. Above 200 nmol/L, Aurora inhibition is more complete, after which the amount of p21^Waf1/Cip1 continually increases, translating to a concentration-dependent limitation of endoreduplication. Further confirmation of the importance of p21^Waf1/cip1 in regulating the response to VX-680 was evidenced by the significant increase in endoreduplication observed in A549 cells when its expression was reduced by siRNA targeting and the reduced degree of endoreduplication that occurred when it was expressed in p53-negative NCI-H1299 cells.

It is not yet clear why the response of the p53-p21^Waf1/Cip1 pathway to VX-680 is not as robust in some cell lines. It is possible that there are other components of the p53 response after VX-680 exposure that must be intact for full postmitotic checkpoint function. These components might overlap with those of p53-dependent G₁ checkpoints induced after DNA damage or oncogene activation, including those that contribute to p53 stabilization such as chk2-mediated phosphorylation (24) or induction of p14^ARF (25, 26). Other components might affect p53 activation, such as the ATM-dependent association of p53 with 14-3-3 proteins (27). The degree of p53 stability and activation likely contribute to the speed and abundance of p21^Waf1/cip1 that is ultimately induced. For example, U2OS cells have been reported to have low endogenous levels of chk2, and its ectopic expression enhances p53 stabilization and G₁ arrest in response to genotoxic stress (24). In this regard, use of VX-680 could help probe the role of proteins such as chk2 or p14^ARF in the postmitotic checkpoint.

In the cell lines examined, the overall integrity of the p53-dependent postmitotic checkpoint governed not only the degree of endoreduplication but also the viability of cells exposed to VX-680. Flow cytometry-based TUNEL assays routinely showed fluorescein shift of polyploid cells (Figs. 1F and 2E; data not shown), suggesting that mitotic catastrophe (28) and DNA fragmentation following endoreduplication results in cells with variable DNA content, including sub-G₁ DNA content.

In some cell lines, endoreduplication proceeded longer, prior to the onset of cell death, whereas in other cell lines, the induction of apoptosis was faster. As shown in Fig. 2, Calu-1 cells continue to endoreduplicate with a substantial 16N peak at 72 hours; in HeLa cells, death is more abrupt, so that there is a smaller 16N peak and more cells with sub-G₁ DNA content. HeLa cells were among the most sensitive of the cell lines we studied. Importantly, these cells lack both functional Rb and p53. Rb also plays an important role in the postmitotic checkpoint (16, 29–31), and it will be of interest to more rigorously study whether loss of both Rb and p53 synergize in the induction of endoreduplication and eventual cell death.

As cells with a compromised p53-p21^Waf1/Cip1 pathway are more likely to die in response to VX-680, there is promise that Aurora inhibition will be selectively cytotoxic to malignant cells. Further work will be necessary to confirm that both normal fibroblasts and normal epithelial cells undergo pseudo-G₁ arrest without cell death in response to VX-680. In addition, short exposures to the drug may result in reversible effects in nontransformed cells and progressive endoreduplication and commitment to cell death in malignant cells. It is encouraging that there was little VX-680-induced toxicity in nude mice or rats harboring colon carcinoma or leukemia xenografts treated at concentrations that induced tumor growth inhibition and regression (15).

Because VX-680 inhibits all three Auroras, it is not yet clear which kinase is the primary target for antitumor activity. The effects of VX-680 most closely mirror those described for the siRNA-mediated depletion of Aurora B or combined depletion of Aurora A and Aurora B (32, 33). In both cases, there is an override of the spindle checkpoint, completion of and exit from an aberrant mitosis, with defective cytokinesis leading to polyploidy. In addition, in VX-680 cells, inhibition of tumor growth is paralleled by a reduction in histone H3 phosphorylation, an event attributed to Aurora B inhibition.

In contrast, the spindle checkpoint is effectively engaged following siRNA-mediated depletion of Aurora A alone in HeLa and U2OS cells (34). Therefore, it is unlikely that Aurora A inhibition alone could account for the effects of VX-680. Nonetheless, Aurora A depletion does delay mitotic entry, which has been described for VX-680, and causes failure of multiple mitotic events including separation of centriole pairs, misalignment of chromosomes on the metaphase plate, and incomplete cytokinesis (2). Therefore, it is possible that the combined inhibition of Aurora A and Aurora B by VX-680 is critical for its antitumor effects.

The effects of VX-680 on cell lines possessing variable degrees of postmitotic checkpoint competency were reflected with siRNA-mediated codepletion of Aurora A and Aurora B in our experiments. In MCF-7 cells, combined depletion or depletion of Aurora A alone induced comparable increases in G₂-M DNA content without endoreduplication (Fig. 2C; data not shown); depletion of Aurora B alone caused a slight increase in G₂-M DNA content, again without endoreduplication (data not shown). In contrast, in postmitotic checkpoint–compromised U2OS cells, increases in G₂-M and 8N DNA content after combined Aurora A and Aurora B depletion were generally more apparent than with depletion of either alone (data not shown), as previously reported (33). Similarly, combined depletion reflects the effects of VX-680 in HeLa cells; endoreduplication after combined depletion in these cells is also greater than after depletion of Aurora B alone (33). However, the effects of combined depletion were overall not as dramatic as those induced by drug. This could be related to the incomplete suppression of expression by the siRNA duplexes. Alternatively, VX-680 also targets Aurora C, the activity of which should not be affected in siRNA-depleted cells. It has recently been shown that Aurora C can complement Aurora B function in mitotic cells, and that depletion of Aurora C also leads to polyploidy (4, 5), so that Aurora C inhibition may also be contributing to the antitumor effects of VX-680.

VX-680 has recently entered clinical trials, and an increasing number of compounds will likely join this novel antineoplastic class. If the potential for selective antitumor cytotoxicity is met, further assessment of the critical Auroras necessary to inhibit for antitumor activity, as well as their interacting proteins, will guide new drug development. Finally, our data suggest that further dissection of the components of the postmitotic G₁ checkpoint may help select patient populations most likely to respond to Aurora kinase inhibition.

Grant support: Research Fellowship funded by Vertex Pharmaceuticals and Merck & Co., Inc., the Friends of Dana-Farber Cancer Institute (F. Gizatullin), NIH grant R01 CA90687 (G.I. Shapiro), and the Dana-Farber Harvard Cancer Center Specialized Program of Research Excellence in Lung Cancer P20 CA90573 (M. Loda and G.I. Shapiro).

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.

Kathryn Folz-Donahue and Laura Prickett of the Dana-Farber Flow Cytometry Core provided technical assistance. We thank our colleagues at Merck & Co., Inc. and Vertex Pharmaceuticals, Inc. for helpful discussions, including Carolyn Buser-Doepner, Paul Deutsch, Jamie Freedman, Steven Averbuch, Giulio Draetta, and Robert Fram.