Two of the hallmark features of melanoma are its development as a result of chronic UV radiation exposure and the limited efficacy of cisplatin in the disease treatment. Both of these DNA-damaging agents result in large helix-distorting DNA damage that is recognized and repaired by nucleotide excision repair (NER). The aim of this study was to examine the expression of NER gene transcripts, p53, and p21 in melanoma cell lines treated with cisplatin compared with melanocytes. Basal expression of all genes was greater in the melanoma cell lines compared with melanocytes. Global genome repair (GGR) transcripts showed significantly decreased relative expression (RE) in melanoma cell lines 24 hours after cisplatin treatment. The basal RE of p53 was significantly higher in the melanoma cell lines compared with the melanocytes. However, induction of p53 was only significant in the melanocytes at 6 and 24 hours after cisplatin treatment. Inhibition of p53 expression significantly decreased the expression of all the GGR transcripts in melanocytes at 6 and 24 hours after cisplatin treatment. Although the RE levels were lower with p53 inhibition, the induction of the GGR genes was very similar to that in the control melanocytes and increased significantly across the time points. The findings from this study revealed reduced GGR transcript levels in melanoma cells 24 hours after cisplatin treatment. Our findings suggest a possible mechanistic explanation for the limited efficacy of cisplatin treatment and the possible role of UV light in melanoma. Cancer Res; 70(20); 7918–26. ©2010 AACR.
Two of the hallmark features of melanoma are its development as a result of chronic repeated exposure to UV radiation (1, 2) and the limited efficacy of the DNA-damaging agent cisplatin in the disease treatment (3). Both of these agents result in large helix-distorting DNA damage that is recognized and repaired by the DNA repair pathway nucleotide excision repair (NER). NER is a versatile DNA repair system that eradicates UV light–induced lesions, such as cyclobutane pyrimidine dimers and pyrimidine (4–6) pyrimidine photoproducts, as well as lesions induced by many chemical compounds including cisplatin (4–6). The importance of the NER pathway is very evident in the rare autosomal recessive disease xeroderma pigmentosum (XP). Patients with XP have a drastically diminished NER capacity; therefore, DNA damage accumulates rapidly after UV light exposure, resulting in up to a 1,000-fold increase in development of skin cancers on sunlight-exposed areas of skin (7). Despite this, the role of NER in the development of melanoma in the general population as a result of UV radiation has not been thoroughly investigated. To date, only weak evidence has been reported on the genetic association of NER-related genes with melanoma (8–11).
Another example of DNA damage recognized and repaired by NER is intra- and inter-strand cross-links induced by DNA-damaging or cross-linking agents. The most commonly used example of a DNA-damaging agent is cisplatin. Despite the high level of efficacy of cisplatin treatment for many types of malignancies, several factors such as cell resistance and adverse drug reactions have undermined its therapeutic potential. In melanoma, the efficacy of cisplatin is limited (12), which is in contrast to the case of most germ-cell tumors having cure rates of more than 90% (13). This contrast in efficacy is yet to be explained, and in recent years, many studies have focused on elucidating the anticancer mechanism of cisplatin. NER is now known to be a vital component of this process; however, studies investigating the role of NER in cisplatin efficacy and resistance are yet to be conducted in melanoma.
NER consists of approximately 30 proteins that remove helix-distorting lesions through four steps: (a) recognition of the DNA lesion, (b) opening of a bubble around the lesion, (c) incision of the DNA upstream and downstream of the lesion by endonucleases, and (d) DNA resynthesis and ligation (14). There are two damage recognition arms of the NER pathway: global genome repair (GGR) and transcription-coupled repair (TCR). GGR encompasses the noncoding parts of the genome, silent genes, and the nontranscribed strand of active genes (15). TCR ensures that the transcribed strand of active genes is repaired with higher priority than the rest of the genome by using RNA polymerase II as a lesion sensor (15). Once the damage is recognized through one of these processes, the remainder of the repair process follows a convergent pathway (7).
The tumor suppressor protein p53 has recently been implicated in the transactivation of the GGR components of the NER pathway following DNA-damaging events (16, 17). p53 plays an important role in many cellular responses to DNA damage, including cell cycle arrest and apoptosis. Recent evidence suggests that p53 is essential for both basal and induced GGR response to certain types of UV- and chemical-induced DNA damage (18–21). The role of p53 in regulating GGR seems to be mediated in part by its ability to transactivate the gene transcripts encoding the GGR proteins DDB2 and XPC.
In contrast with other tumor types, melanoma has a relatively infrequent p53 mutation rate of only around 9% (reviewed in refs. 22, 23). In addition, p53 protein and mRNA expression have been reported in melanoma, with reports showing increases in p53 with tumor progression (24–27). The presence of wild-type p53 expression in melanoma indicates that other events that bypass p53 must be occurring for tumor development and/or progression to occur. One candidate event could be the lack of transactivation of GGR in response to DNA damage. Given the limited knowledge available detailing the role of NER in melanoma, the aim of this study was to determine the level of NER activity and its relationship with p53 in response to cisplatin-induced DNA damage in melanocytes and melanoma cell lines.
Materials and Methods
Cell lines and cisplatin treatment
One melanocyte, three primary melanoma (MM200, IgR3, and Me4405), and two metastatic melanoma (Mel-RM and Sk-mel-28) cell lines were used for this study. The derivation of MM200, IgR3, Me4405, Mel-RM, and Sk-mel-28 melanoma cell lines has been described previously (28–31). Melanocytes were purchased from Cascade Biologics at the commencement of this study. DNA for cell line authentication was extracted from all the cell lines while cultured for this study. Individual cell line authentication was confirmed using the AmpFlSTR Identifiler PCR Amplification Kit from Applied Biosystems and GeneMarker V1.91 software. A panel of 16 markers were tested, and each cell line had a distinct individual set of markers present.
All of the melanoma cell lines were cultured in DMEM (5% FCS) and the melanocytes were cultured in Medium 154 (Cascade Biologics). All cell lines were maintained in exponential growth at 37°C and 5% CO2. Cells were treated with 10 μg/mL cisplatin (Pharmacia Upjohn) as previously described (28) and were harvested before and at 6 and 24 hours after treatment for gene expression analysis.
Inhibition of p53 by short hairpin RNA
Short hairpin RNA (shRNA) sequences to p53 or a control were expressed in the pSIH1-H1-copGFP (Copepod green fluorescent protein) shRNA expression vector (Systems Biosciences). The p53-directed shRNA sequence corresponds to nucleotides 1026–1044 (GenBank accession no. NM_000546; ref. 32). The control shRNA sequence 5′-TTAGAGGCGAGCAAGACTA-3′ showed no homology to any known human transcript.
Stable transduction of melanoma cell lines.
Lentiviruses were produced in HEK293T cells using the pSIH1-H1-copGFP shRNA expression vector (Systems Biosciences) encased in viral capsid encoded by three packaging plasmids as described previously (33). Viruses were concentrated as described previously (34). Viral titres were determined using 1 × 105 U2OS cells per well in six-well plates, which were transduced with serial dilutions of the concentrated viral stocks in the presence of 8 μg/mL polybrene (Sigma). Cells were harvested 48 hours after transduction and analyzed by flow cytometry for copGFP expression, and viral titre was calculated.
To generate a p53-silenced stable melanocyte cell line, melanocytes were transduced at a multiplicity of infection of 10 with either a virus encoding p53 shRNA or a control shRNA that has no homology to any human gene. Cells were transduced twice at 3-day interval. The efficiency of transduction was monitored with coexpression of copGFP and was consistently >95%. All cell lines tested negative for the presence of replicative competent virus using the Retrotek HIV-1 p24 antigen ELISA kit (ZeptoMetrix Corporation).
Sample preparation and real-time PCR
Total RNA was extracted from all of the cell lines after cisplatin treatment using the SV Total RNA Isolation System Kit (Promega). The total RNA was quantified using the Quant-iT RiboGreen RNA Assay Kit (Invitrogen) and a fluorometer (BMG Labtech). To ensure consistency across all the samples being tested, a standardized amount (500 ng) of total RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems), and a 1:20 dilution of the resultant cDNA was used in triplicate for each sample. Relative gene expression was measured in triplicate and normalized to β-actin (ΔCt) using TaqMan gene expression assays (Applied Biosystems) and the 7500 real-time PCR system (Applied Biosystems) for the following gene transcripts: DDB2 (XPE; Hs00172068_m1), DDB1 (Hs_00172410_m1), XPC (Hs01104206_m1), ERCC8 (CSA; Hs01122123_m1), ERCC6 (CSB; Hs00972920_m1), XPA (Hs00166045_m1), RPA1 (Hs00161419_m1), RPA2 (Hs00358315_m1), ERCC3 (XPB; Hs01554450_m1), ERCC2 (XPD; Hs00361161), ERCC5 (XPG; Hs00164482), ERCC4 (XPF; Hs00193342), ERCC1 (Hs01012161_m1), p53 (Hs00153340_m1), and p21 (Hs00355782_m1). To ensure that β-actin itself did not change between cell lines and/or treatments, the ratio of β-actin to a second reference gene, GAPDH, was measured. The average β-actin/GAPDH ratio was 1.02 ± 0.04 across all the individual cell lines and treatment time points.
Relative gene expression was calculated using 2−ΔCt, and unpaired two-tailed t tests (P < 0.05) were used to identify significantly altered expression in melanoma compared with melanocytes as described previously (35).
Western blot analysis
Protein extraction, separation by SDS-PAGE, and Western blot analysis of cell lines were done as described previously (28). The mouse monoclonal antibody Bp53-12, used for the detection of p53, was purchased from Upstate. The mouse monoclonal antibody for the housekeeping gene GAPDH was purchased from Ambion.
Global genome repair
The relative expression (RE) of the GGR gene transcripts XPC, DDB1, and DDB2 (XPE) was measured at 0, 6, and 24 hours after cisplatin treatment in melanoma and melanocyte cell lines. The RE of all three GGR transcripts was higher at the basal level and at 6 hours after treatment in the melanoma cell lines, although not significantly different when compared with melanocytes (Fig. 1). The expression of these transcripts significantly increased in a time-dependent manner following cisplatin treatment in melanocytes. No cisplatin-dependent increase of these transcripts was observed in melanoma cell lines (Fig. 1). The induction of XPC, DDB1, and DDB2 (XPE) relative to the basal expression level was measured at 6 and 24 hours after treatment. At 24 hours after cisplatin treatment, there was a 38.77-fold (P = 0.0003), 12.27-fold (P = 0.0005), and 81.46-fold (P = 0.0002) increase in RE in the melanocytes, respectively (Fig. 2). In comparison, a 1.33-fold decrease in expression was observed for DDB1 in the melanoma cell lines and only a small fold increase in RE was observed for XPC (3.68) and DDB2 (1.84), all of which were not significant (Fig. 2).
Clearly, a significantly lower RE level of the GGR genes was observed in the melanoma cell lines 24 hours after cisplatin treatment (Fig. 1). In addition, a significant increase in induction of these transcripts was observed in the melanocytes but not in the melanoma cell lines (Fig. 2).
The RE of the TCR gene transcripts ERCC6 (CSB) and ERCC8 (CSA) was very low in all of the melanoma and melanocyte cell lines at all time points. However, the RE for ERCC6 (CSB) was significantly lower (P = 0.005) in the melanoma cell lines (1.34) at the 24-hour time point compared with the melanocyte cell line (6.52). For ERCC8 (CSA), RE was significantly higher in the melanoma cell lines (0.99) than in the melanocyte cell line (0.14) at the 6-hour time point (P = 0.04). No significant differences in RE between the melanoma and melanocyte cell lines were observed for ERCC6 (CSB) or ERCC8 (CSA) at all other time points. The relatively low level of expression of all the TCR transcripts at all time points suggests limited involvement of TCR in response to cisplatin-induced DNA damage.
Nucleotide excision repair
The NER-related genes XPA, RPA1, RPA2, ERCC1, ERCC2 (XPD), ERCC3 (XPB), ERCC4 (XPF), and ERCC5 (XPG) showed an overall trend of higher basal expression in melanoma cell lines compared with the melanocyte cell line (Table 1). Specifically, the basal RE of RPA1, RPA2, ERCC2 (XPD), ERCC3 (XPB), and ERCC4 (XPF) was significantly greater in the melanoma cell lines. In addition, RPA1 and RPA2 RE levels were significantly greater in the melanoma cell lines compared with the melanocyte cell line 6 hours after cisplatin treatment, but the expression levels were similar at 24 hours. In contrast, ERCC3 (XPB), ERCC4 (XPF), and ERCC5 (XPG) RE levels were significantly lower in the melanoma cell lines than in the melanocyte cell line 24 hours after cisplatin treatment. In the melanoma cell lines, ERCC2 (XPD) and XPA were slightly increased in RE from 0 to 24 hours after cisplatin treatment; however, this was not significantly different from the RE observed in the melanocytes at the 24-hour time point (Table 1).
p53 is currently thought to transactivate expression of the GGR genes DDB2 and XPC in response to DNA damage. The RE of p53 was measured in all the cell lines after 0, 6, and 24 hours of cisplatin treatment. The RE of p53 was significantly higher in the melanoma cell lines compared with the melanocytes at 0 and 6 hours after cisplatin treatment but not at 24 hours (Fig. 3). Although the transcript levels of p53 were higher in the melanoma cell lines at 0 and 6 hours, the induction of p53 was of borderline significance at 24 hours (3.91-fold increase, P = 0.05). Conversely, induction of p53 was significant in the melanocytes at 6 hours (1.6-fold increase, P = 0.0001) and 24 hours (19.57-fold increase, P = 0.007) after cisplatin treatment (Fig. 3).
Increased expression of p21 is indicative of p53 activation in response to DNA damage. The RE of p21 was measured in all cell lines after 0, 6, and 24 hours of cisplatin treatment. The RE of p21 seemed to be higher in melanoma cell lines at the 0- and 6-hour time points, but it was not statistically significant. Conversely, the RE of p21 was significantly lower in the melanoma cell lines compared with the melanocyte cell line after 24 hours of cisplatin treatment (Fig. 3). Although transcript levels of p21 were higher in the melanoma cell lines at 0 and 6 hours, the induction of p21 was not significantly elevated by 24 hours compared with the melanocyte cell line. Induction of p21 was observed in the melanocyte cell line at 6 hours (9.85-fold increase, P = 0.00004) and 24 hours (142.44-fold increase, P = 0.002) after cisplatin treatment (Fig. 3).
GGR transcript expression with shRNA inhibition of p53
To determine if the GGR transcript level changes in the melanoma cell lines were related to the absence of p53, knockdown experiments were undertaken using the melanocyte cell line.
The transcript level of all the GGR genes, XPC, DDB1, and DDB2, was significantly lower in the p53 shRNA melanocyte cell line compared with the control melanocyte cell line at 6 and 24 hours after cisplatin treatment. The basal transcript levels, however, were not significantly different (Fig. 4). Although the RE levels were lower in the p53 shRNA melanocyte cell line, the induction of the GGR genes was very similar to that in the control melanocyte cell line and increased significantly across the time points (Fig. 5).
shRNA inhibition of p53 and its effect on p21 transcript expression
shRNA was used to inhibit the transcript expression of p53 in the melanocyte cell line. To confirm inhibition of p53 transcription, p53 and p21 transcript expression was measured in the control shRNA melanocyte cell line and the p53 shRNA melanocyte cell line. The RE of p53 was 8.12-fold lower (P = 0.01) in the p53 shRNA melanocyte cell line than in the control melanocyte cell line. Similarly, the RE of p21 was 2.33-fold lower (P = 0.07) in the p53 shRNA melanocyte cell line than in the control melanocyte cell line. In addition, Western blots were used to confirm the inhibition of p53 protein expression in the p53 shRNA melanocyte cell line (Supplementary Data).
Despite overwhelming evidence that one of the primary causal agents of melanoma is chronic repeated exposure to UV radiation, its exact relationship with the disease remains to be fully elucidated. Cisplatin is a common DNA-damaging agent that is used in the treatment of many types of cancer; the efficacy of cisplatin in the treatment of melanoma, however, is very limited (12). Cisplatin binds to DNA, forming helix-distorting intra- and inter-strand cross-links (36, 37), which must be removed before either transcription or replication. The removal and repair of large helix-distorting DNA damage induced by both UV light and DNA-damaging agents such as cisplatin is orchestrated by NER. The aim of the present study was to examine the changes in gene expression of the key constituents of the NER pathway in melanoma cell lines treated with cisplatin compared with a control melanocyte cell line and to examine the relationship to the cell cycle checkpoint control gene p53.
GGR operates throughout the entire genome and is a crucial step in the initial recognition of DNA damage (38). The intra- and inter-strand cross-links caused by cisplatin are recognized by the GGR component XPC; then, the DDB1/DDB2 complex is recruited to bind specifically to the large helix-distorting DNA adducts (38). Thereafter, the repair process proceeds through the rest of the NER pathway. Previous studies have identified a strong correlation between reduced XPC mRNA and protein levels and increased resistance of cancer cells to cisplatin treatment (39–41). In this study, the genes involved in the GGR pathway, XPC, DDB1, and DDB2 (XPE), were shown to have no increase in RE in the melanoma cell lines after cisplatin treatment. In contrast, a significant increase in RE 24 hours after cisplatin treatment was seen in the melanocyte cell line. These results show that following cisplatin treatment, the GGR genes are poorly induced in melanoma cell lines compared with melanocytes.
The most important mechanism for the anticancer action of cisplatin is recognition of cisplatin-induced DNA damage, which triggers cell death. This process is known as DNA damage–mediated apoptotic cell death (40). The cisplatin treatment–mediated p53 response and activation of caspase-3, both key mechanisms involved in triggering apoptosis, are significantly reduced in XPC-defective cell lines, which suggests that XPC plays a critical role in initiating apoptosis mediated by cisplatin-induced DNA damage (40, 41).
In addition to the key role that XPC plays in DNA damage–mediated apoptosis after cisplatin treatment, DDB2 also has a role in this process. DDB2-deficient cells exhibit enhanced resistance to cell growth inhibition and apoptosis induced by cisplatin (42, 43), and DDB2 expression in cisplatin-resistant ovarian cancer cell lines is lower than that in their cisplatin-sensitive parental cells (42). Overexpression of DDB2 sensitizes cells to cisplatin-induced cytotoxicity and apoptosis through activation of the caspase pathway and downregulation of antiapoptotic Bcl-2 protein (42).
What we have observed in the melanoma cell lines is a failure to upregulate key damage recognition genes, including XPC and DDB2, in response to cisplatin-induced DNA damage. The failure to elicit the appropriate response of the GGR pathway after cisplatin treatment suggests that the normal relationship between DNA damage recognition and subsequent apoptosis is no longer functionally accomplishable in these melanoma cell lines. The low expression of XPC and DDB2 may therefore play a key role in the resistance of melanoma to cisplatin.
Basal expression levels of the GGR genes XPC, DDB1, and DDB2 (XPE) seemed to be greater in the melanoma cell lines compared with the melanocytes. In addition, the NER genes RPA1, RPA2, ERCC3 (XPB), ERCC2 (XPD), and ERCC4 (XPF) showed significantly higher basal-level expression in the melanoma cell lines compared with the melanocytes. Increased basal expression of several NER genes has been previously related to cisplatin resistance (44). Specifically, ovarian tumors resistant to cisplatin treatment have shown high levels of expression of XPA, ERCC1, XPF, and ERCC3 (XPB). Conversely, testicular cancers display low expression of XPA, ERCC1, and XPF and are highly responsive to cisplatin (reviewed in ref. 44). Although only a subset of the NER genes previously reported to have higher basal expression in relation to cisplatin resistance were identified in this study, overall, increased basal-level expression of NER genes seems to be associated with cisplatin resistance. Taken together with the results of this study, this could provide some explanation about why cisplatin is ineffective in melanoma treatment.
p53 is a critical regulator of NER both at the basal level and in the induction of GGR following UV light–induced and chemically induced DNA damage (18–21, 45). The role of p53 in regulating GGR seems to be mediated, in part, by its ability to transactivate gene transcripts encoding the GGR proteins DDB2 and XPC. In this study, a key downstream p53 target gene, p21, showed a similar trend to p53, indicating that altered expression of p53 was indeed affecting downstream signaling. We confirmed that induced p53 and GGR gene transcript levels occur in concert in melanocytes 24 hours after cisplatin treatment. In melanoma cell lines studied herein, the relationship is not as clear. The basal level of p53 is significantly higher in melanoma cell lines relative to the basal levels of the majority of the GGR genes. Furthermore, the RE of p53 after 24 hours of cisplatin treatment is remarkably similar in melanoma and melanocyte cell lines, but the RE of the GGR genes in the melanoma cell lines is very low. The most likely explanation for this occurrence is the difference in the induction of p53 after cisplatin treatment in the melanoma cell lines compared with the melanocytes. There is a highly significant increase in p53 induction at the 6- and 24-hour time points for the melanocytes, but the melanoma cell lines fail to respond as rapidly and p53 is only differentially expressed after 24 hours (P = 0.05). This suggests that it is the induction of p53 that is required for GGR activation, rather than just a high constitutive level of p53 expression.
The current understanding of the relationship between p53 and GGR relies on the hypothesis that an increase in p53 results in transactivation of GGR transcript expression (16, 17). To investigate this notion further, shRNA was used to inhibit p53 expression in the melanocyte cell line used for this study. Once reduced p53 expression was confirmed by Western blot and direct measurement of p53 transcript levels and the consequent reduction in p21 mRNA, the expression of the three GGR transcripts was investigated. The RE levels of the three GGR genes, XPC, DDB1, and DDB2 (XPE), was significantly lower in the p53 shRNA melanocytes at all time points after cisplatin treatment. However, the induction of these transcripts was not altered by the inhibition of p53, as all three GGR genes showed significant fold-change increases in expression at 24 hours after cisplatin treatment, thereby showing their similarity to the control melanocytes. This result confirms that a high level of p53 expression is not the only requirement for GGR activation. In addition, the induction of GGR in the melanocytes despite the inhibition of p53 suggests that p53 may not be responsible for the lack of GGR induction in the melanoma cell lines.
In 2006, a study by Yang and colleagues identified a number of p53 target genes that were significantly downregulated in melanomas compared with melanocytes (46). One of their major findings was the association of reduced DDB2 activity and melanoma development (46). In addition, the expression of other p53 target genes was also altered, including the cell cycle regulator CDKN1A (p21CIP), which was confirmed by Kaufmann and colleagues (47). Taken together, the findings of these studies and those of the current study reveal that reduced DDB2 activity is potentially implicated in melanoma development and/or progression and that a number of other p53 targets may have altered expression in melanoma.
In summary, this study has revealed that melanoma cell lines treated with cisplatin have lower expression of the GGR genes XPC, DDB1, and DDB2 (XPE) when compared with melanocytes after 24 hours. Interestingly, the basal expression levels of p53 and several NER repair genes in melanoma cell lines were greater than those observed in the melanocyte cell line, yet on stimulation by the DNA-damaging agent cisplatin, only the melanocyte cell line was capable of response. This was confirmed by studies aimed at inhibiting p53 expression in the melanocytes, which showed a reduction in the expression, but not induction, of the GGR transcripts.
Low mRNA (40, 41) and protein (39) expression of GGR genes in relation to cisplatin resistance have been previously reported. Although the present study has only reported mRNA expression of NER genes, the results are in accordance with the previous studies and suggest that GGR deficiency may play a role in cisplatin resistance in melanoma. Further studies confirming the differences in NER expression after cisplatin treatment at the protein level would further support this finding.
These findings provide evidence of the possible biological mechanisms involved in the reduced efficacy of cisplatin in melanoma. Notwithstanding, the DNA damage caused by cisplatin is similar to UV radiation, as both result in large helix-distorting DNA damage that is recognized by the NER pathway. Therefore, the findings of this study point toward a biological mechanism involved in melanoma resistance to cisplatin and in melanoma development as a result of UV exposure.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Lyndee L. Scurr and Helen Rizos from Westmead Institute for Cancer Research, University of Sydney at Westmead Millennium Institute, NSW, Australia, for the p53 shRNA melanocyte cell lines and Amanda Croft for the Western blot analysis.
Grant Support: The Hunter Medical Research Institute and The University of Newcastle Centre for Information Based Medicine. N.A. Bowden is supported by the National Health and Medical Research Council, NBN Telethon Children's Cancer Research Group, and The Hunter Medical Research Institute. K.A. Ashton is supported by Mr. and Mrs. Bogner and the Hunter Medical Research Institute.
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 January 15, 2010.
- Revision received July 12, 2010.
- Accepted July 27, 2010.
- ©2010 American Association for Cancer Research.