Introduction
One of the hallmarks of transformed/malignant cells is their limitless proliferation capacity and defective cell cycle checkpoints that, when functional, operate to detect errors in replication processes and direct cells into apoptosis [
1,
2]. Thus, interfering with mitosis has proven to be a successful cancer treatment strategy [
3]. Several components of the mitotic machinery have been identified as potential therapeutic targets, and antimitotic agents are already crucial in the chemotherapy of both adult and childhood malignancies. For instance, the microtubule-targeting
Vinca alkaloids are a central component of curative regimens for many childhood solid tumors and leukemias. Other appealing targets include mitotic kinesins [
3,
4], centromere components required for chromosome alignment and spindle complex formation [
5], as well as Polo-like kinases and the Aurora kinases [
6].
The Aurora serine/threonine protein kinases are a family of three kinases (Aurora A, Aurora B, and Aurora C) with different tissue and temporal expression profiles that play key roles in mitosis and meiosis, defects in which can lead to abnormal mitotic events and apoptosis induction [
7]. The essential nature of Aurora kinase A is highlighted by the fact that genetically engineered null mice are embryonic lethal (dying at the blastocyst stage) [
8]. Aurora kinase A activity is also required for centrosome duplication and separation, microtubule–kinetochore attachment, spindle checkpoint formation, cytokinesis [
9,
10], the G2/M transition [
11], and phosphorylation of Polo-like kinase 1 [
12]. Further, Aurora kinase A has been implicated as an oncogenic driver in human cancers [
13]. Aurora kinase A has been found to be overexpressed in cancer cells, and the AURKA gene locus is amplified in selected adult tumors [
14]. However, limited information on the role of Aurora kinase A in pediatric cancers is available.
Aurora kinase inhibitors are the focus of several pharmaceutical development programs. Aurora kinase inhibitors with different specificities and activities as well as pharmacodynamic markers are currently being assessed, and some are already well advanced in clinical trials (reviewed in [
15]). Most of these inhibitors show a broad range of activity, with AZD-1152 being an example of a selective Aurora kinase B inhibitor and MLN8054 (or its derived compound, MLN8237, used in the present study) an example of a selective Aurora kinase A inhibitor. The effects of Aurora kinase A inhibition are multiple, as corresponds to the varied nature of its substrates, and include abnormal spindle pole formation, proliferation reduction (with G2-M arrest), and polyploidy [
16], followed by apoptosis induction. The latter could involve signaling mediated by p53, as Aurora kinase A has been shown to modify the phosphorylation status of p53 and histone H3 [
17,
18] and to interact with the MYCN protein, limiting p53 ubiquitination and degradation by the proteasome in neuroblastoma cell lines [
19,
20]. Although p53 is frequently non-functional in cancer cells, inhibition of Aurora kinase A by MLN8054 can lead to p73-dependent apoptosis in p53-deficient cells [
21]. Aurora kinase A has also been reported to influence cell survival through the Akt pathway and by interfering with IkBα [
22].
The primary focus of the Pediatric Preclinical Testing Program (PPTP) is to identify novel agents that have significant antitumor activity against models of childhood solid tumors and acute lymphoblastic leukemia (ALL) as one source of data to use in prioritizing clinical development of such agents in the pediatric setting. The PPTP has reported the single-agent evaluation of activity of the Aurora kinase A inhibitor MLN8237 against its panels of in vitro cell lines and in vivo xenograft models [
23]. Both the neuroblastoma and ALL panels were particularly sensitive to the single-agent treatment. In fact, this Aurora kinase A inhibitor is the only drug out of more than 20 tested with preferential activity against the neuroblastoma panel. Despite these encouraging results, issues of how responsiveness relates to drug exposure in mice and humans, the dose range over which MLN8237 exerts significant antitumor activity, and the correlation of sensitivity to Aurora kinase A expression remain unanswered. Here, we report the in vitro activity of MLN8237 against an extended panel of neuroblastoma and Ewing sarcoma cell lines, and we report in vivo dose–response efficacy studies focusing on neuroblastoma and pediatric ALL xenografts, as well as assessment of pharmacokinetic, pharmacodynamic, and molecular parameters associated with these responses.
Materials and methods
In vitro testing
In vitro testing was performed using DIMSCAN, a semiautomatic fluorescence-based digital image microscopy system that quantifies viable (using fluorescein diacetate [FDA]) cells in tissue culture multiwell plates [
24]. Cells were incubated in the presence of MLN8237 for 96 h at concentrations from 1 nM to 10 μM and analyzed as previously described [
25]. Two measures of sensitivity were used; the absolute IC
50, defined as the drug concentration inhibiting growth by 50% compared to controls, and the relative IC
50 (previously termed EC
50), defined as the drug concentration yielding 50% of the maximum inhibitory effect.
Cell lines for in vitro testing
The cell lines used in this study were obtained from the originator of the cell line or the Deutsche Sammlung von Mikroorganismen unde Zellkulturen (Braunschweig, Germany) or the American Type Culture Collection (ATCC, VA, U.S.A.) and were maintained in culture according to the corresponding initial report. All lines underwent DNA genotyping as described [
26]. Short tandem repeat (STR) assay was used to verify each line against the Children’s Oncology Group STR database (this can be found in
http://www.COGcell.org).
In vivo tumor growth inhibition studies
CB17SC
scid
−/− female mice (Taconic Farms, Germantown NY) were used to propagate subcutaneously implanted kidney/rhabdoid tumors, sarcomas (osteosarcoma, rhabdomyosarcoma), and neuroblastoma tumors as previously described [
27‐
29]. Human leukemia cells were propagated by intravenous inoculation in female non-obese diabetic (NOD)/
scid
−/− mice as described previously [
30]. Details of these tumor panels can be obtained at
http://pptp.nchresearch.org/documents.html. Female mice were used irrespective of the patient gender from which the original tumor was derived. All mice were maintained under barrier conditions, and experiments were conducted using protocols and conditions approved by the institutional animal care and use committee of the appropriate consortium member. Ten mice (solid tumor models) and 8 mice (ALL models) were used in each control or treatment group. Tumor volumes (cm
3) [solid tumor xenografts] or percentages of human CD45-positive cells out of the total leukocyte population in peripheral blood [hCD45, ALL xenografts] were determined as previously described [
31]. An event was defined for the solid tumors as a quadrupling of tumor volume from the tumor volume at start of treatment, and for the ALL models when the proportion of hCD45 reached 25%. Event-free survival (EFS) was estimated for individual mice as the time required from treatment initiation to reach the defined event threshold.
Determination of response
Responses were assessed using three activity measures as previously described [
31]. For all the solid tumors on an individual basis, progressive disease (PD) was defined as <50% regression from initial volume during the study period and >25% increase in initial volume at the end of study period. Stable disease (SD) was defined as <50% regression from initial volume during the study period and ≤25% increase in initial volume at the end of the study. Partial response (PR) was defined as a tumor volume regression ≥50% for at least one time point but with measurable tumor (≥0.10 cm
3). Complete response (CR) was defined as a disappearance of measurable tumor mass (<0.10 cm
3) for at least one time point. A complete response was considered maintained (MCR) if the tumor volume was <0.10 cm
3 at the end of the study period. Similarly, for the ALL xenografts, PD was defined as a continuous increase in hCD45 that reached event before the end of the 42-day monitoring period, SD as hCD45 that did not decrease to under 1% and did not reach the event threshold, PR was defined as a decrease in hCD45 to under 1% at only one time point, CR when the hCD45 stayed under 1% for two consecutive measures, and MCR when the hCD45 stayed under 1% for the last 3 weeks of the monitoring period. Tumor growth delay (TGD) values were calculated based on the numbers of days to event. For each individual mouse that had PD and had an event in the treatment groups, a TGD value was calculated by dividing the time to event for that mouse by the median time to event in the respective control group. Median times to event were estimated based on the Kaplan–Meier event-free survival distribution. For treatment groups only, if the tumor response was PD, then the score was further classified into PD1 or PD2 based on the TGD value: if the TGD value ≤1.5, that mouse was considered PD1 while if the TGD value >1.5, the mouse was considered PD2. Mice that had PD but did not have an event at the end of the study were coded as SD. A Median Group Response was obtained for each cohort based on the median of the scores of the individual mice. An in-depth description of the analysis methods is included in the
Supplemental Response Definitions section. The three scores reflecting tumor regression (PR, CR and MCR) were considered Objective Responses (ORs).
Millennium Pharmaceuticals, Inc., through the Cancer Therapy Evaluation Program (NCI), provided MLN8237 to the Pediatric Preclinical Testing Program. MLN8237 was dissolved in DMSO and diluted in culture media for in vitro tests or was suspended in 10% 2-hydroxypropyl-β-cyclodextrin and 1% sodium bicarbonate in water and administered to mice via oral gavage twice daily for 5 days repeated each week for a total of 6 weeks for the solid tumor xenografts, and for 3 weeks for the leukemia xenografts at doses of 20.8 mg/kg (the MTD), 10.4, 5.2, and 2.6 mg/kg.
Statistical methods
The exact log-rank test, as implemented using Proc StatXact for SAS®, was used to compare event-free survival (EFS) distributions between treatment and control groups. P values were two-sided and were not adjusted for multiple comparisons given the exploratory nature of the studies. The Mann–Whitney test was used to test for differences in the median IC50 values for the Ewing sarcoma and neuroblastoma cell lines.
Pharmacokinetic analysis
Blood was obtained by cardiac puncture 0, 0.5, 1, 2, 4, 6, 8, 12, and 24 h after dosing with MLN8237 (10 mg/kg) from 3 mice per time point. Plasma samples were prepared and analyzed for MLN8237 concentrations by an LC/MS/MS assay according to previously described methods for MLN8054 [
16].
Pharmacodynamic analysis
Accumulation of mitotic cells was used as a pharmacodynamic measure of Aurora kinase A inhibition in NB-1771 tumor-bearing animals dosed with 20.8 mg/kg MLN8237. Tumors were collected from animals at 0, 2, 6, 8, 12, and 24 h following MLN8237 dosing from 3 mice per time point and were formalin-fixed and paraffin-embedded. Tumor sections (5 μm) were stained for two independent mitotic markers, MPM2 (Mitotic protein monoclonal #2) and histone H3 phosphorylated on serine 10 (pHistH3) using the Discovery® XT (Ventana Medical Systems, AZ, U.S.A.) automated slide stainer. Sections were deparaffinized with EZ prep™ solution, and antigen retrieval was completed with Cell Conditioning 1 solution, CC1 (Ventana Medical Systems). The sections were incubated for 60 min at room temperature with mouse MPM-2 antibody (1:100, Upstate Biotechnology, MA, U.S.A.) and rabbit anti-phospho-histone H3 polyclonal (1:25, Upstate Biotechnology). Biotin-conjugated anti-mouse antibody (Animal Research Kit, DAKO, CA, U.S.A.) was included to amplify the MPM2 signal. Conjugated fluorophores, including Alexa Fluor 488-conjugated streptavidin (1:100, Molecular Probes, OR, U.S.A.) or Rhodamine-Red-X-AffiniPure goat anti-rabbit IgG (1:25, Jackson ImmunoResearch, PA, U.S.A.), were incubated for 60 min at room temperature. Slides were washed in PBS and mounted with DAPI Vectashield Hard Set Mounting Medium (Vector Laboratories, CA, U.S.A.). Images were acquired using a Canon E300 microscope (Canon, U.S.A.) with an automated stage. Five images from each slide were captured using a 40× PlanFluor objective (Nikon Instruments, U.S.A.) and analyzed on the MetaMorph® image processing software (Universal Imaging Corp., PA, U.S.A.) that used a custom image processing application module. Mitotic indices were determined as the percentage of total cells that were positive for either pHisH3 or MPM2 staining.
Copy number analysis
Copy number analysis was performed using the Affymetrix Genome-Wide SNP Array 6.0. DNA (500 ng) from each sample was processed according to the manufacturer’s instructions. SNP 6.0 data were processed from CEL files to extract raw signal intensity values using dChip [
32] PM-only model-based expression analysis. The signal data were then normalized using a reference-based normalization algorithm [
33]. For each marker in each array, the log2 ratio of tumor versus the median signal obtained from 90 reference samples from St. Jude Children’s Research Hospital was calculated. Then, the segmentation algorithm [
34] implemented in the DNAcopy package from Bioconductor [
35] was applied to the above log2 ratio data to identify copy number alterations for each tumor sample. Copy number gains and losses were defined by genomic segments with log2 ratios >0.2 or <0.2, respectively.
Correlation analysis of gene expression and genomic copy number variation
Gene expression data were obtained previously using the Affymetrix HG-U133 Plus 2.0 array [
36]. To estimate the variance in gene expression attributed to underlying copy number variation, a linear regression model was fitted to compare SNP data (segment log2 ratio) against expression data (log2 signal). For each probe set on the HG-U133 Plus 2.0 array, the correlation coefficient was calculated using each segment falling within a genomic window of ±5 kb up/downstream of the annotated gene. Because multiple segments and probe sets can arise within a given gene boundary, the segment and probe set with the highest correlation (R
2) value were selected for subsequent analysis.
Discussion
The main goal of the PPTP is to prioritize drugs being developed predominantly for adult cancer treatment for expedited clinical trials in children with relapsed/refractory cancers. MLN8237, which has 200-fold specificity for Aurora kinase A inhibition versus Aurora kinase B [
16], showed high-level activity at its MTD in its initial PPTP evaluation; therefore, it was critical to validate and extend these previous results. This was done by evaluating MLN8237 against an extensive number of Ewing sarcoma and neuroblastoma cancer lines in vitro, and by assessing its activity in vivo against neuroblastoma and ALL xenografts across a range of doses with pharmacokinetic and pharmacodynamic correlation.
Aurora kinase inhibitors have to date shown only modest clinical activity against solid tumors in adults, although more pronounced activity has been reported in leukemia patients [
14]. There are limited data available to support Aurora kinase A as a relevant molecular target in pediatric cancers besides the report by Shang et al. [
38] and the PPTP’s previous report of MLN8237 Stage 1 testing [
23]. In this latter publication, high levels of activity were obtained against several solid tumor models and against ALL xenografts of both T and B lineage. The most intriguing set of results was that MLN8237 performed more impressively than other investigational drugs, and even established drugs, against the neuroblastoma panel as a single agent at its MTD.
The Aurora kinases play critical roles in cell division, and alteration of their expression and function has been associated with oncogenesis. Knockdown of Aurora kinase A using RNA interference results in mitotic spindle defects, mitotic delay, and apoptosis in human cells [
39], while overexpression leads to transformation of normal cells [
40]. Also, Aurora kinase A is amplified or overexpressed in some adult cancers [
41‐
43], which supports its potential exploitation as a cancer therapeutic target [
14]. Similarly, the overexpression of Aurora kinase A has been postulated as predictive of susceptibility to inhibition of the specific kinase activity. Thus, Ewing sarcomas, with genetic alterations that enhance Aurora kinase A expression [
37], should have higher sensitivity than the lower expressing neuroblastoma or ALL panels. The results presented in this study confirm our previous results of high-level activity for MLN8237 against neuroblastoma and ALL xenografts, which express markedly lower Aurora kinase A levels compared to other PPTP xenografts [
23], thereby calling into question the premise that overexpression of Aurora kinase A is associated with more effective cell kill upon kinase inhibition. Although the Ewing sarcoma xenografts had slightly increased expression of AURKA compared to the median for all xenografts, our study did not confirm enhances in sensitivity to MLN8237 in vitro or in vivo. Indeed, the gene copy number analysis for AURKA appears to support an inverse relationship between Aurora kinase A expression and sensitivity. Increased copy number was present in half of the rhabdomyosarcomas and in 14 of the solid tumors. Loss of copy number was detected in 7 solid tumors and ALL-17. Further, the correlation between gene expression variation and copy number variation was strong, placing this locus in the top 1.6% of all genes tested. Although there is no absolute relationship between copy number variation and tumor sensitivity, of the 14 solid tumors with increased copy number, there were only two that showed sensitivity to MLN8237 (1 PR, 1 SD). In contrast, of the eight models demonstrating decreased copy number, there were five sensitive models (3 MCR, 1 CR, and 1 SD).
The in vitro activity of MLN8237 against the Ewing sarcoma and neuroblastoma extended panels (
n = 11 and
n = 17, respectively) is consistent with the PPTP’s Stage 1 results for MLN8237, which showed median relative and absolute IC
50 values against all of the cell lines in the PPTP in vitro panel of 49 and 61 nM, respectively [
23]. The larger number of Ewing and neuroblastoma cell lines described in this report compared to those studied in Stage 1 testing allowed detection of significantly lower IC
50 values for the neuroblastoma cell lines compared to the Ewing sarcoma cell lines. Further, one Ewing sarcoma cell line (CHLA-56) was resistant to MLN8237 (IC
50 > 10 μM). The identification of this highly resistant cell line places it as a valuable tool for identifying resistance mechanisms and warrants further investigation. Recently, a functional Aurora kinase A mutation (T217D) that renders the kinase impervious to MLN8054 and MLN8237 inhibition has been reported [
44] and points to a mechanism of resistance independent from levels of expression.
The efficacy of MLN8237 treatment in vivo at its MTD was confirmed against the xenograft panel included in this report. Out of 10 xenografts also evaluated in the previous report, only one (OS-1) was scored more than one response category apart from its previous score [
23]. We have confirmed the high level of activity of MLN8237 against xenograft models of neuroblastoma and ALL, when administered as a single agent at its MTD. This further demonstrates the potential relevance of Aurora kinase A inhibition for neuroblastoma cancer treatment. However, the efficacy of MLN8237 (as indicated by the Median Group Response) was reduced or lost for most of the solid tumor models with dose reduction (Fig.
1). Thus, at 0.5MTD, only two xenografts exhibited an objective response, and at 0.25MTD, only one xenograft was classified as PR. In contrast, the dose–response relationship for the ALL xenografts was not as steep, with all three models exhibiting objective responses at 0.5MTD and only one not reaching an objective response upon further reduction to 0.25MTD.
Data for the pharmacokinetics of MLN8237 in patients have recently been presented [
45]. In patients receiving 50 mg BID, the
C
max and AUC
0–24 h were 1.3 and 40 μM h, respectively. At the recommended phase 2 dose of 50 mg BID for 7 days, average trough concentrations exceeded 1 μM, the efficacious concentration estimated in previous preclinical work. In mice receiving MLN8237 at 10 mg/kg, the
C
max and AUC
0–24 h were 16 and 39μM h, respectively, with the 12 h level being 1.2 μM. Thus, results presented here suggest that drug exposures achievable in patients may induce responses in only the most sensitive of tumors and that dose intensity and scheduling may be critical as a minority of the solid tumor models exhibited objective responses at this level of drug exposure.
When comparing the plasma exposure of MLN8237 to the pharmacodynamic response, the peak of pharmacodynamic activity (~8–12 h) was delayed relative to the peak plasma exposure (0.5 h). This is consistent with previous observations using the Aurora kinase A inhibitor MLN8054 in a colon tumor xenograft [
16] and is likely due to the time it takes for a sufficient number of cells to transit the cell cycle and accumulate in mitosis subsequent to Aurora kinase A inhibition as well as to the time during which MLN8237 drug levels are above a threshold level required for Aurora kinase A inhibition. The comparable mitotic indices estimated using MPM2 and pHistH3 as mitotic markers are consistent with specific inhibition of Aurora kinase A by MLN8237 in vivo, as histone H3 is phosphorylated by Aurora kinase B [
46].
A likely critical step in the development of MLN8237 for use in the treatment of pediatric cancers is the development of effective drug combinations. The limited activity observed at reduced doses of MLN8237 as a single agent against most solid tumor xenografts may be overcome if synergistic interactions with other drugs can be identified. Combinations of MLN8237 with established drugs against in vivo models of pediatric solid tumors and ALL are under evaluation by the PPTP.
The cumulative evidence of anti-tumor activity observed in preclinical testing together with the results presented here provides strong rationale for expeditious evaluation of MLN8237 in the childhood cancer setting. A pediatric phase 1/2 trial was opened in the Children’s Oncology Group Phase 1 Consortium during 2008. As results from that clinical trial emerge, it will be crucial to correlate the observed anti-tumor activities with pharmacokinetic measurements to assess whether drug levels are in the range associated with substantial preclinical activity.
Acknowledgments
This work was supported by NO1-CM-42216, NO1-CM91001-03, CA21765, and CA108786 from the National Cancer Institute and used MLN8237 supplied by Millennium Pharmaceuticals, Inc. Sherry Ansher, Catherine A. Billups, Joshua Courtright, Mila Dolotin, Edward Favours, Henry S. Friedman, Debbie Payne-Turner, Charles Stopford, Chandra Tucker, Joe Zeidner, Ellen Zhang, and Jian Zhang contributed to this work in addition to the authors. Children’s Cancer Institute Australia for Medical Research is affiliated with the University of New South Wales and Sydney Children’s Hospital.