Patupilone is an epothilone in advanced clinical development that has shown promising efficacy in heavily pretreated patients. This study aimed at characterizing the mechanisms of patupilone activity in resistant patients. To this end, we generated patupilone-resistant cells using two cellular models, the first characterized by high chemosensitivity and low class III β-tubulin (TUBB3) expression (A2780), and the second by low chemosensitivity and high TUBB3 expression (OVCAR-3). The obtained cell lines were named EPO3 and OVCAR-EPO, respectively. The same selection procedure was done in A2780 cells to generate a paclitaxel-resistant cell line (TAX50). Factors of resistance are expected to increase in the drug-resistant cell lines, whereas factors of drug sensitivity will be down-regulated. Using this approach, we found up-regulation of TUBB3 in TAX50, but not EPO3, cells, showing that TUBB3 mediates the resistance to paclitaxel but not to patupilone. Moreover, TUBB3 was a factor of patupilone sensitivity because OVCAR-EPO cells exhibited a dramatic reduction of TUBB3 and a concomitant sensitization to hypoxia and cisplatin-based chemotherapy. To identify the mechanisms underlying patupilone resistance, tubulin genes were sequenced, thereby revealing that a prominent mechanism of drug resistance is represented by point mutations in class I β-tubulin. Overall, these results suggest that paclitaxel and patupilone have nonoverlapping mechanisms of resistance, thus allowing the use of patupilone for those patients relapsing after paclitaxel-based chemotherapy. Furthermore, patupilone represents a promising first-line option for the treatment of high-risk ovarian cancer patients, who exhibit high TUBB3 levels and poor response to standard paclitaxel-platin chemotherapy. [Cancer Res 2008;68(24):10197–204]
- class III β-tubulin
- Drug resistance
Microtubule-targeting agents are widely used for cancer treatment. Microtubule-targeting agents are classically divided into drugs acting as inhibitors and those as enhancers of tubulin polymerization, although it is likely that at the low drug concentrations obtainable in patients, both groups act through inhibition of microtubule dynamics ( 1). The taxanes paclitaxel and docetaxel are widely used in treatments of solid tumors. They are two natural compounds that enhance tubulin polymerization and share the same framework (10-deacetylbaccatin III, obtained from Taxus spp. throughout the world) and are approved for clinical use in ovary, lung, breast, head and neck, gastric, and prostate cancers. Whereas an acceptable response rate is obtained in some patients, which has led to their clinical approval, the major problem in their clinical use is drug resistance. Unfortunately, often remission lasts only a few months or even weeks, and on relapse, patients do not respond further. Efforts to develop novel taxanes with an improved activity profile in resistant patients have not been successful ( 2). One reason for this failure is that the clinical development of novel taxanes has been hampered by poor knowledge of the drug resistance mechanisms in resistant patients. In fact, the development of novel taxanes has been mainly guided by selecting lead compounds for the activity against P-glycoprotein (P-gp), the gene product of the ABCB1 (MDR) gene. This has resulted in the clinical development of a series of taxanes able to potently inhibit the P-gp function, but their response rates in phase II clinical studies were disappointingly lower than expected, showing that this resistance mechanism, which is easily observable in cultured cell lines ( 3), does not actually operate in patients with solid tumors ( 2). Screening for additional compounds directed against P-gp–expressing cells revealed another class of natural compounds, the epothilones ( 4), which are produced by the myxobacterium Sorangium cellulosum and partially share the same pharmacophore of taxanes ( 5). However, in contrast to P-gp–targeting taxanes, epothilones remain effective when used in resistant patients, suggesting that the clinical activity depends on factors other than P-gp. For this reason, the first epothilone (Ixabepilone) has been recently approved by the Food and Drug Administration for treating aggressive metastatic or locally advanced breast cancer that is no longer responsive to currently available chemotherapies ( 6). Patupilone (epothilone B, EPO906) is another epothilone in advanced phase III clinical development. Preclinical data have confirmed that it also retains activity in patients relapsing from or resistant to a taxane-containing chemotherapy ( 7). This study aimed to identify the molecular basis for patupilone activity in resistant cells. The results show that the mechanisms underlying patupilone resistance are significantly different from those responsible for taxane resistance, thereby providing a molecular basis for identifying the subset of patients who are potentially responsive to this novel therapy and the background to a sequential combination of the two drugs in distinct clinical subsets.
Materials and Methods
Drugs. Paclitaxel was kindly provided by Indena, and patupilone by Novartis. Drugs were diluted in absolute DMSO. These solutions was further diluted at each experimental day to achieve a 0.1% final DMSO concentration. All the other chemicals were purchased from Sigma if not otherwise specified.
Cell cultures. A2780 and OVCAR-3 human ovarian cancer cells were purchased from the European Collection of Cell Cultures (ECACC). TAX50, PTX, and TC1 cells are A2780 cells resistant to paclitaxel and are described elsewhere ( 8, 9). Culture media were selected according to ECACC suggestions. Patupilone-resistant cells were obtained through stepwise increase of drug concentration. The obtained patupilone-resistant cell lines do not depend on drug presence for growth. For normoxia, cultures were incubated at 37°C in a fully humidified atmosphere of 5% CO2/95% air. For hypoxia, experiments were done using a special chamber (Billups-Rotenberg, Inc.) filled with 100% nitrogen. After removal of the supernatant, live attached cells were counted blindly by two independent observers using a Bürker cell.
Growth experiments with cytotoxic agents. Cells were seeded in black 96-well flat-bottomed plates (Perkin-Elmer Life Sciences). After 24 h, media were replaced and, after one washing, media containing the drugs were added. A dose-response logarithmic curve was established in quadruplicates for each plate, starting from 0.1 to 10,000 nmol/L. Each experiment was done thrice. After 72 h of culture in the presence of the tested compounds, plates were harvested and the number of viable cells was estimated using the ATPlite kit (Perkin-Elmer Life Science) and the automated luminometer Topcount (Perkin-Elmer Life Science). The kit was used according to the manufacturer's suggestions. For each drug/cell line, a dose-response curve was plotted and the half maximal inhibiting concentration (IC50) values were then calculated by fitting the dose-effect curve data obtained in the three independent experiments with the sigmoid-Emax model using nonlinear regression, weighted by the reciprocal of the square of the predicted effect.
Real-time quantitative PCR. Total RNA was obtained from cultured cells using RNeasy plus kit (Qiagen) according to the manufacturer's directions. cDNA was prepared starting from 1 μg of total RNA using the iScript cDNA Synthesis Kit (Bio-Rad) according to the manufacturer's instructions. Amplifications were carried out using specific primers and the iQ SYBR Green Supermix (Bio-Rad) in a final volume of 25 μL, starting with a 3-min template denaturation step at 95°C followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. Standard curves were generated using a serial dilution of the initial amount of control cDNA to determine the range of template concentrations, which showed a good linearity and efficiency for the different reactions. Melt curves of the reaction products were also generated to assess the specificity of the measured fluorescence. Samples were run in triplicate and the mean of threshold cycles (Ct) for each specimen was used to obtain the fold change of gene expression level using the following equation: Fold change = 2 − (Ct), where Ct = Ct specific gene − Ct GAPDH, and (Ct) = Ct specimen − Ct control (i.e., A2780). A fold change equal to 1 represents a sample with an expression level equal to the control cell line. This operation was done using the Excel spreadsheet RelQuant (Bio-Rad).
Western blot analysis. Total cellular proteins were obtained lysing the cells with radioimmunoprecipitation assay buffer [50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L sodium chloride, 1% NP40, 0.5% sodium deoxycholate] in the presence of proteases and phosphatases inhibitors. Protein concentration was determined by spectrophotometer and proteins were run on SDS denaturing gels. Each protein sample (50 μg) was separated on a 10% SDS polyacrylamide gel and electroblotted onto polyvinylidene difluoride membranes (Millipore Co.). After electroblotting, the membranes were incubated in 5% nonfat dry milk in 1× TBS/Tween 20 (0.1 mol/L Trizma base, 0.15 mol/L NaCl, 0.05% Tween 20, pH 7.4) for blocking and then with the following antibodies: anti–class III β-tubulin polyclonal antibody (Covance), anti–α-tubulin monoclonal antibody (Calbiochem), and anti–β-actin monoclonal antibody (Sigma). After incubation with secondary horseradish peroxidase–conjugated antibodies (Bio-Rad), visualization was done with the enhanced chemiluminescence plus system (Amersham Biosciences) using a Versadoc imaging system (Bio-Rad). For the assessment of the specific β-tubulin isotypes, we have used and validated a panel of polyclonal anti–β-tubulin isotypes, which were in-house generated with the following specific reactivity: anti–class I (TUBB), anti–class II (TUBB2A and TUBB2B), and anti–class V (TUBB6); all the antibodies were validated with a protocol adapted from Hiser and colleagues ( 10), and the validation dot blot is reported in Supplementary Fig. S1. The antibody anti–class IV was purchased from Genetex.
Genomic DNA isolation and bisulfite treatment of DNA. The methods for the assessment of the methylation status of the 3′ flanking region of TUBB3 human gene have been previously reported ( 11). Briefly, genomic DNA was isolated from cell lines using Wizard Genomic purification kit (Promega) following the manufacturer's instructions. To distinguish methylated CpG from unmethylated CpG in hypoxia-inducible factor-1α (HIF-1α) binding site of human TUBB3 3′ flanking region, 1 μg of DNA was treated with sodium bisulfite using CpGenome DNA modification kit (Serologicals Corp.). After this treatment, unmethylated cytosine is converted to uracil, whereas methylated cytosine remains unmodified.
Amplification, cloning, and sequencing of bisulfite-treated DNA. PCR primers to amplify the TUBB3 3′-untranslated region were 5′-ggttttgttttttttatttaggttatg-3′ (forward) and 5′-aatatctaacaacaataaatttattaaatatccc-3′ (reverse). The PCR product was cloned into the TA cloning kit (Invitrogen) and clones were sequenced using the Big Dye Terminator v3.1 labeling kit (Applied Biosystem) and M13 primers.
β-Tubulin isotypes and patupilone resistance. Drug resistance to microtubule-targeting agents is often associated to TUBB3 up-regulation ( 12, 13). To assess if TUBB3 is a factor of drug resistance for patupilone, we exposed A2780 cells to patupilone and paclitaxel (as a positive control). A2780 cells were chosen for their chemosensitivity profile and low levels of TUBB3. Quantitative PCR was used to measure the TUBB3 level at each step increase in drug concentration. The levels of TUBB3 increased 5-fold in cells grown in the presence of 50 nmol/L paclitaxel (TAX50), but this was not detected in A2780 cells selected with patupilone, hereafter referred to as EPO3x cells, where x corresponds to the patupilone dose used for the selection. Data are shown in Fig. 1A , and the main features of the obtained cell lines are reported in Table 1 . Indeed, during the selection with patupilone, the expression of TUBB3 actually diminished, which suggests that TUBB3 affects the resistance to paclitaxel but not to patupilone. To further determine the correlation between patupilone resistance and TUBB3, we repeated the same protocol in OVCAR-3 cells. This cell line was chosen for its native drug resistance (with respect to A2780 cells) and high expression levels of TUBB3. The cell lines obtained with this selection procedure are named here as OVCAR-EPOx, where x again corresponds to the patupilone dose used for the selection ( Table 1). Also in this case, quantitative PCR was used to monitor TUBB3 during the selection process. The results showed ( Fig. 1A) that, beginning from the first selection step with patupilone (2.5 nmol/L), TUBB3 levels markedly decreased (by up to 30-fold), suggesting that cells with elevated levels of TUBB3 are more sensitive to patupilone. Due to the importance of this finding, the selection step was independently repeated four times up to a patupilone dose of 5 nmol/L, to elucidate anecdotal variations in the TUBB3 content. Again, we found consistent changes in the TUBB3 content, with a down-regulation similar to that evident in Fig. 1A (data not shown).
The above phenomenon was analyzed at the protein level by measuring TUBB3 expression with Western blotting in patupilone-resistant cell lines obtained at the highest drug doses (7.5 and 10 nmol/L for EPO3 and OVCAR-EPO cells, respectively) and in A2780 cells made stably resistant to paclitaxel (PTX and TC1). Also at the protein level, no increase in TUBB3 was evident in EPO3-7.5 cells, whereas, compared with the parental cell line, TUBB3 was dramatically reduced in OVCAR-EPO10 cells ( Fig. 1B).
To assess if changes in TUBB3 were associated with the additional modulation of other β-tubulin isotypes, class I (TUBB), class II (TUBB2A and TUBB2B), class IV (TUBB4 and TUBB2C), and class V (TUBB6) were assessed by quantitative PCR and Western blotting ( Fig. 2 ). Class VI (TUBB1) was not assessed because it is not detectable in ovarian cancer cells. The panel of used antibodies is validated in Supplementary data. Comparing quantitative PCR results and Western blots revealed only a decrease in TUBB6 in OVCAR-EPO10 cells at the mRNA and protein levels, suggesting that the decreases in TUBB3 and TUBB6 are correlated in the OVCAR-3 system.
Efflux pumps and patupilone resistance. Because patupilone is reportedly active in cells expressing efflux pumps belonging to the ATP binding cassette (ABC) superfamily ( 7), quantitative PCR was also used to measure the expressions of ABCB1 (P-gp), ABCC1 (MRP), and ABCG2 (BCRP) in EPO3-7.5 cells; OVCAR-EPO10 cells; and their parental counterparts ( Fig. 3 ). Rather than an overexpression, we found that the ABCC1 and ABCG2 levels were consistently down-regulated in the resistant cell lines, showing that these proteins do not contribute to patupilone resistance. On the other hand, the ABCB1 gene was up-regulated (12-fold) in EPO3-7.5 cells, reaching the expression level measured in parental OVCAR-3 cells. However, in OVCAR-EPO cells expression of ABCB1 was consistently down-regulated, suggesting that the fluctuations in the level of ABCB1 mRNA in EPO-3 cells are anecdotal and not related to the process of patupilone resistance. In keeping with this interpretation, when the analysis was repeated in Western blotting ( Fig. 3A), the ABCB1 protein was below the detection threshold in all cell lines except for TAX50 (the cell line exposed stepwise to up to 50 nmol/L paclitaxel). These findings suggest that efflux pumps belonging to the ABC superfamily do not contribute to patupilone resistance because they are not detectably up-regulated in resistant cells at the protein level.
Tubulin point mutations in patupilone-resistant cells. Previous reports have associated drug resistance to microtubule-targeting agents with mutations in α-tubulin and β-tubulin genes ( 14, 15). This issue was investigated by sequencing TUBB and TUBB3 as well as the most-expressed α-tubulin genes in patupilone-resistant cells and their parental counterparts. A heterozygous mutation was found at position 298 in EPO3-7.5 cells (N298[N,S]), whereas the homozygous R282Q mutation was evident in OVCAR-EPO10 cells ( Fig. 4A ). No mutations were detectable in TUBB3, whereas two silent polymorphisms were found in each of TUBA1A and TUBA1B. A heterozygous polymorphism at position 45 in TUBA1C (G45[G,E]) was detected in all the cell lines. To correlate the presence of point mutations with the microtubule polymerization activity, patupilone-resistant cells and their parental counterparts were treated for 30 minutes with 5 and 10 nmol/L patupilone, and Western blotting was used to assess the amount of polymerized and soluble tubulin after ultracentrifuging the samples ( Fig. 4B and C). The polymerization did not differ between EPO3-7.5 and A2780 cells for 5 nmol/L patupilone, but was slightly lower in EPO3-7.5 cells for 10 nmol/L patupilone. It is evident that the R282Q mutation in OVCAR-EPO10 cells strongly affects drug-induced tubulin polymerization because polymerization was lower with both 5 and 10 nmol/L patupilone. However, we cannot exclude that other microtubule-associated factors different from the R282Q mutation could be involved in the reduced patupilone-driven tubulin polymerization observed in OVCAR-EPO cell lines.
Patupilone resistance and sensitivity to hypoxia and chemotherapy. Previous studies have shown that hypoxia is able to induce the expression of TUBB3, and that this enhanced expression is associated with methylation of a specific enhancer positioned in the 3′ flanking region of the human TUBB3 gene ( 11). Interestingly, once methylated, this enhancer makes the expression of TUBB3 constitutive and not further inducible by hypoxia ( 11). Therefore, we investigated the effects of 24, 48, and 72 hours of hypoxia in patupilone-resistant cells and their parental counterparts ( Fig. 5A ). The sensitivity to hypoxia did not differ between A2780 and EPO3-7.5 cells. OVCAR-3 cells, exhibiting high levels of TUBB3, were more resistant to hypoxia, but this was inhibited in OVCAR-EPO10 cells. TUBB3 expression was also monitored in these cell lines during hypoxia. Exposure to hypoxia for 48 hours increased the levels of TUBB3 in OVCAR-EPO10 cells but not in parental OVCAR-3 cells ( Fig. 5B). This prompted us to investigate if the decrease in TUBB3 expression and the restoration of TUBB3 inducibility by hypoxia were linked to a modulation in the methylation of the enhancer at the 3′ flanking region of the TUBB3 human gene. The methylation status of this enhancer was assessed in the four cell lines ( Fig. 5C). The methylation did not differ between A2780 and EPO3-7.5 cells, whereas in OVCAR-EPO10 cells there was a dramatic reduction in the number of cells with a methylated enhancer, thereby explaining the mechanism underlying the reduced basal levels of the protein as well as TUBB3 induction in hypoxia. These findings also prompted us to assess if these cells are more sensitive to conventional chemotherapy ( Fig. 5D). In fact, cisplatin is the most effective treatment for ovarian cancer, and increased TUBB3 expression has also been associated with resistance to DNA-damaging agents such as cisplatin ( 16). In support to this interpretation, in OVCAR-EPO7.5 and OVCAR-EPO10 the reduction of TUBB3 was paralleled to a chemosensitization to cisplatin. Moreover, patupilone resistance was not linked to paclitaxel resistance because the sensitivity to paclitaxel was slightly enhanced in OVCAR-EPO7.5 and did not change in OVCAR-EPO10. These findings point out that the combination cisplatin/carboplatin with patupilone could be promising in patients expressing high-levels of TUBB3 and prone to drug resistance.
Improving therapies for drug-resistant patients requires characterization of the mechanism(s) underlying chemotherapy failures. Although drug resistance is a multifactorial process, the mechanisms that have already been validated in clinical studies are attractive as targets for novel therapies aimed at high-risk patients. Among these factors, a prominent role has been assumed by the ectopic expression of TUBB3, as reviewed recently ( 12). Under normal conditions, this protein is expressed at low levels in many epithelia, such as the ovary and lung. However, under stress conditions characterized by a low nutrient supply and oxygen level, TUBB3 is transcriptionally induced through engagement of HIF-1α binding at the 3′ flanking region of the gene ( 11). Once adapted to hypoxia, the enhancer at the 3′ flanking region of the gene is methylated and TUBB3 is no longer inducible, remaining constitutively expressed ( 11). Cells with high levels of TUBB3 are more resistant to hypoxia and chemotherapy. Characterization of TUBB3 function suggests that this protein is not only a marker but also a mediator of cell survival under stress conditions ( 17).
Therefore, similar to other targets of HIF-1α ( 18), in tumors that have progressed in a compelling hypoxic microenvironment, strong TUBB3 expression after the first surgery will be the hallmark of aggressive behavior and will increase the risk of resistance to conventional chemotherapy ( 12, 19). For this reason, high expression levels of TUBB3, as evident in around 20% of ovarian cancer patients, are associated with an unfavorable outcome even when treated, having a mean survival of 24 months ( 20), which is close to that reported in patients treated only with palliative cares without chemotherapy ( 21). In other words, the current therapies do not have any significant effect on the clinical course in these patients. Similar trends have been reported in lung ( 22), breast ( 23), pancreas ( 24), and stomach ( 25) cancers, thereby implicating TUBB3 in drug resistance in several types of solid tumors.
The results here obtained with patupilone provide new hope for these patients with an unfavorable prognosis. Our method was based on the generation of drug-resistant cells in which the antitumor effects of patupilone are inhibited through overexpression and down-regulation of those genes acting as factors of resistance and sensitivity, respectively. We adopted two cellular models, characterized by low and high TUBB3 expression, as well as low and high sensitivity to hypoxia and chemotherapy. This approach revealed that TUBB3 does not affect resistance to patupilone because A2780 cells with a low TUBB3 expression and high sensitivities to hypoxia and chemotherapy do not overexpress this protein during the development of resistance. This behavior is different from that seen with paclitaxel, whose resistance is easily achieved through TUBB3 overexpression. More importantly, TUBB3 expression is indicative of sensitivity to this drug because OVCAR-EPO cells from the first step of drug selection exhibited a dramatic reduction of TUBB3. This phenomenon cannot be ascribed to a random clonal selection with the drug because we repeated the selection with patupilone four times in OVCAR-3 cells and always found a consistent down-regulation of TUBB3 expression. This phenotypic conversion is associated with sensitization to hypoxia and cisplatin. OVCAR-3 cells, which exhibit high constitutive (not inducible) TUBB3 expression and resistance to hypoxia (and cisplatin), acquire a phenotype very similar to that of drug-sensitive A2780 cells. Moreover, the enhancer at the 3′ flanking region of the gene becomes demethylated, and on hypoxia, the protein is inducible. Along with this effect in hypoxia, patupilone-resistant cells exhibited mild hypersensitivity to cisplatin, making the combination of patupilone with cisplatin or carboplatin particularly attractive.
In contrast to observations in paclitaxel, whose resistant phenotype as described above is frequently associated with TUBB3 overexpression ( 8), tubulin point mutation occurs in cell lines made resistant to patupilone. Such mutations have been extensively found with epothilones in previous studies, including also the R282Q mutation ( 14), which is mapped directly in the drug binding site ( 26). No clinical studies have elucidated the frequency of point mutations in patients treated with epothilones. On the other hand, this resistance mechanism is not detectable in patients treated with paclitaxel ( 27– 29). It therefore seems that with replacement of the TUBB3-driven resistance mainly occurring in paclitaxel-resistant cells (and patients), patupilone therapy fails, at least in part, through point mutations in the drug binding pocket, thereby inducing a diminished susceptibility to drug-induced microtubule polymerization. The nonoverlapping mechanism of drug resistance opens the prospect of the sequential use of patupilone in patients relapsing from taxane-including chemotherapy and could explain the good response rate that patupilone has achieved in clinical trials.
Many studies have focused on P-gp as a central mediator of multidrug resistance. Cytotoxic taxanes, which also act as potent P-gp inhibitors, have not achieved the desired outcomes in phase II clinical studies ( 2). Clinical trials have also shown that combining P-gp inhibitors with chemotherapy does not have the desired effect in the treatment of solid tumors ( 30). These unfavorable results indirectly show that the P-gp role in drug resistance is marginal at the bedside. This is supported by our previous experience of finding no correlation between P-gp staining and drug resistance in ovarian cancer patients ( 29). The clinical efficacy of patupilone was originally mainly assigned to its activity against P-gp–expressing cells ( 4). As the main change to this viewpoint, in the present study we have identified patupilone as a targeted chemotherapeutic drug that has the ability to specifically kill TUBB3-expressing cells. This is supported by the dramatic reduction evident in OVCAR-EPO cells and by the reappearance of cells in which TUBB3 is newly inducible by hypoxia. In a modeling study, Magnani and colleagues ( 26) predicted that TUBB3 would be targeted more specifically by patupilone than by paclitaxel. Therefore, this study is in line with this model.
In conclusion, this study shows that patupilone is endowed with unique biological properties that do not overlap those of traditional taxanes. Although originally selected for activity against P-gp–expressing cells, this is not the mechanism underlying the drug activity found in heavily pretreated patients. Moreover, our findings could initiate a novel phase of preclinical development of these agents. Several epothilones have been synthesized, but promising leads were selected for their activity against P-gp–positive cells, and hence it is possible that compounds with intrinsic minor toxicity and better TUBB3 targeting have been discarded in favor of high activity in P-gp–expressing cellular systems. Now that the present study has identified TUBB3 as a main target of patupilone, a better approach might be to look again at these compounds to obtain a less toxic lead that might be more easily combinable with other agents. Combining patupilone with cisplatin or carboplatin represents a promising advance for the treatment of high-risk patients characterized by high TUBB3 expression levels, who are at a high risk of rapid progression during standard conventional chemotherapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: This study was supported by a grant from the Italian Ministry of University and Research (MIUR grant no. 4210011).
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received June 6, 2008.
- Revision received August 25, 2008.
- Accepted October 3, 2008.
- ©2008 American Association for Cancer Research.