The DNA N-Glycosylase MED1 Exhibits Preference for Halogenated Pyrimidines and Is Involved in the Cytotoxicity of 5-Iododeoxyuridine David P

Research Article

The DNA N-Glycosylase MED1 Exhibits Preference for Halogenated Pyrimidines and Is Involved in the Cytotoxicity of 5-Iododeoxyuridine David P. Turner,1 Salvatore Cortellino,1 Jane E. Schupp,2 Elena Caretti,1 Tamalette Loh,2 Timothy J. Kinsella,2 and Alfonso Bellacosa1

1Human Genetics Program, Fox Chase Cancer Center, Philadelphia, Pennsylvania and 2Department of Radiation Oncology, Case Comprehensive Cancer Center, Cleveland, Ohio

Abstract enzymes that add a methyl group to the 5 position of the cytosine pyrimidine ring within a cytosine-phosphate-guanine (CpG) The base excision repair protein MED1 (also known as MBD4), dinucleotide (ref. 2 and references therein). Given the importance an interactor with the mismatch repair protein MLH1, has of CpG methylation, it is somewhat surprising to find that CpG a central role in the maintenance of genomic stabilitywith sequences are potential mutational hotspots within the genome. It dual functions in DNA damage response and repair. MED1 has been estimated that approximately one third of all human acts as a thymine and uracil DNA N-glycosylase on T:G and inherited mutations and cancer-associated mutations occur at CpG U:G mismatches that occur at cytosine-phosphate-guanine sites, including 50% of p53 mutations in Li-Fraumeni syndrome (CpG) methylation sites due to spontaneous deamination of and sporadic colorectal cancer (3) and 50% of hereditary non- 5-methylcytosine and cytosine, respectively. To elucidate the polyposis colorectal cancer mutations (4). Furthermore, aberrant mechanisms that underlie sequence discrimination byMED1, methylation patterns at the promoter region of tumor suppressor we did single-turnover kinetics with the isolated, recombinant genes are associated with the onset of carcinogenesis through gene glycosylase domain of MED1. Quantification of MED1 sub- transcriptional repression (5, 6). strate hierarchyconfirmed MED1 preference for mismatches The increased prevalence of mutations at CpG dinucleotides is within a CpG context and showed preference for hemi- attributed to the increased tendency of the methylation product, methylated base mismatches. Furthermore, the k values st 5-methylcytosine (5-meC), to spontaneously deaminate to thymine. obtained with the uracil analogues 5-fluorouracil and 5- This produces a T:G mismatch, which, if not repaired, results iodouracil were over 20- to 30-fold higher than those obtained in a C:G to T:A transition mutation and the consequent loss of with uracil, indicating substantiallyhigher affinityfor a methylation site (7, 8). Oxidative damage of cytosine and halogenated bases. A 5-iodouracil precursor is the halogenat- 5-methylcytosine may lead to G:C to A:T transitions at CpG sites ed nucleotide 5-iododeoxyuridine (5IdU), a cytotoxic and via the formation of uracil- and thymine glycol, 5-hydroxyuracil, radiosensitizing agent. Cultures of mouse embryo fibroblasts and 5-hydroxycytosine, and subsequent incorporation of adenine (MEF) with different Med1 genotype derived from mice with by bypass polymerases opposite these oxidized bases (9). targeted inactivation of the gene were evaluated for sensi- In keeping with the importance of methylation consensus tivityto 5IdU. The results revealed that Med1-null MEFs sequences, cells have evolved specialized DNA repair processes are more sensitive to 5IdU than wild-type MEFs in both to help protect their integrity within the genome from hydro- 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide lytic deamination events that generate T:G and U:G mismatches and colonyformation assays.Furthermore, high-performance from 5-meC:G and C:G pairs, respectively. In eukaryotes, the base liquid chromatographyanalysesrevealed that Med1-null excision repair proteins thymine-DNA glycosylase and MED1 (also cells exhibit increased levels of 5IdU in their DNA due to known as MBD4) both repair these mismatches that occur at increased incorporation or reduced removal. These findings CpG methylation sites by removing the mismatched T and U with establish MED1 as a bona fide repair activityfor the removal their DNA N-glycosylase activity (10–15). DNA N-glycosylases of halogenated bases and indicate that MED1 mayplay hydrolyze the N-glycosylic bond between the target base and the a significant role in 5IdU cytotoxicity. (Cancer Res 2006; deoxyribose sugar, resulting in a free base and an apurinic/ 66(15): 7686-93) apyrimidinic site; the latter is further processed by apurine/ apyrimidine endonucleases, which cleave the phosphodiester bond Introduction in preparation for the gap filling functions of downstream repair Cytosine methylation is the most common epigenetic modifica- proteins (16–20). In addition to the COOH-terminal glycosylase tion in vertebrate genomes and is essential for the normal domain, MED1 contains an NH2-terminal 5-methylcytosine binding development and functioning of organisms through its control of domain (MBD) and a central linker region of unknown function gene activity and epigenetic inheritance (reviewed in ref. 1). (12–14, 21, 22). Methylation is mediated by a family of DNA methyltransferases, When in a CpG context, TDG and MED1 have glycosylase activity against thymine glycol paired with guanine (23), suggesting a protective role against oxidative damage as well. MED1 also has Note: D.P. Turner and S. Cortellino contributed equally to this work. activity against the halogenated pyrimidines 5-fluorouracil (5-FU; Present address for D.P. Turner: Department of Pathology, Hollings Cancer Center, ref. 13), 5-chlorouracil, and 5-bromouracil (24). Finally, MED1 has Medical University of South Carolina, Charleston, SC 29425. 4 Requests for reprints: Alfonso Bellacosa, Fox Chase Cancer Center, 333 Cottman weak activity against 3,N -ethenocytosine (14). Avenue, Philadelphia, PA 19111. Phone: 215-728-4012; Fax: 215-214-1623; E-mail: CpG sequences are also susceptible to the damaging effects of [email protected]. 6 O6-me I2006 American Association for Cancer Research. alkylation. O -Methylguanine ( G), the most prominent alkylat- doi:10.1158/0008-5472.CAN-05-4488 ing lesion, also has the potential to form at CpG sequences through

Cancer Res 2006; 66: (15). August 1, 2006 7686 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2006 American Association for Cancer Research. MED1 Exhibits Preference for Halogenated Pyrimidines such agents as N-nitroso compounds that act via a unimolecular stepwise gradient of the above wash buffer containing increasing concentrations of imidazole (i.e., 25, 50, 100, and 200 mmol/L respectively). Fractions (SN1) nucleophilic substitution reaction (25, 26). If unrepaired, O6-meG pairs with thymine in the next round of DNA replica- were analyzed using SDS-PAGE with Coomassie staining and the fraction tion and ultimately results in cell cycle arrest and apoptosis containing the MED1 protein was dialyzed against gel filtration buffer [20 mmol/L HEPES (pH 7.6), 50 mmol/L KCl, 1 mmol/L EDTA, 1 mmol/L (27, 28). Similar processing is reported following treatment with DTT, 1 mmol/L benzamidine, 1 mmol/L phenylmethylsulfonyl fluoride]. The 6-thioguanine (29, 30). Recently, MED1 has also been assigned a sample was then applied to a pre-equilibrated Hi-prep 26/60 Sephacryl S-100 functional role in the activation of DNA damage checkpoints in gel filtration column (Amersham Biosciences). Protein was eluted at a flow response to treatment with alkylating agents, such as N-methyl- rate of 0.4 mL/min and collected in 2-mL fractions. Fractions were then N¶-nitro-N-nitrosoguanidine, as well as platinum compounds, 5-FU, analyzed using SDS-PAGE with Coomassie staining and the fractions and irinotecan (31). In the same study, MED1 was shown to display containing the MED1 protein were pooled and concentrated using Ultrafree glycosylase activity on thymine paired with O6-meG (31). centrifugal filter devices (Millipore, Bedford, MA). The protein concentration Given the importance of the biochemical roles of the MED1 of the glycosylase domain of MED1 was determined by absorbance e Â 4 À1 À1 protein and the fact that frameshift mutations of the MED1 gene spectroscopy at 280 nm using the 280 of 5.60 10 (mol/L) cm (34). have been identified in human colorectal, gastric, endometrial, and After the addition of 10% glycerol, samples were flash-frozen and stored at À70jC. MED1 glycosylase domain point mutations were introduced into the pancreatic cancer, MED1 seems to be not only a major guardian of active site via overlap mutagenesis (35) or via the QuickChange mutagenesis methylation site fidelity but also a candidate tumor suppressor kit (Stratagene) and overexpressed and purified in the same manner. gene (21, 32, 33). Oligodeoxynucleotide synthesis and substrate preparation. Oligo- Despite several studies on MED1 repair activity, there has been deoxynucleotides were synthesized using standard phosphoramidite little quantitative assessment of sequence preference or context. chemistry on an automated DNA synthesizer (Applied Biosystems, Foster We previously conducted a kinetic analysis of MED1 and found City, CA) and purified by PAGE as previously described (13, 36). 5-FU- and that its thymine and uracil glycosylase reaction follows single- 3,N 4-ethenocytosine-containing oligonucleotides were purchased from turnover, pre-steady-state burst kinetics; this was due not to GenSet (La Jolla, CA) whereas 5IU-containing oligonucleotides were pur- inactivation of the enzyme after a single reaction cycle but rather chased from Qiagen (Valencia, CA); these oligos were purified as above. to tight binding to the apurinic/apyrimidinic site reaction product Double-strand oligonucleotide substrates were prepared by annealing complementary, gel-purified single-strand 37-mer oligonucleotides as follows: (13). However, our previous kinetic data were based on computer- 5¶-CAATCCTAGCTGACAZGATGTGGCCAATGGCATGACT-3¶;3¶-TTAGGAT- generated simulations and were limited in the variety of substrates CGACTGTGYTACACCGGTTACCGTACTGAG-5¶, where Z is G (GpG context, analyzed (13). To help elucidate the mechanisms that underlie in which case the 5¶ G pairs with C), or C and 5-meC (unmethylated and sequence discrimination by MED1, we have done single-turnover hemimethylated CpG context); and Y is C (no mismatch control), T, U, or kinetics with the isolated glycosylase domain of MED1 to set 5-FU, as required. For the alkylation damage and A:5-FU substrates, the benchmarks for future comparisons. For these experiments, we mismatched G in the top strand was replaced by O6-meG and A, respectively. used oligonucleotide substrates that contained a variety of lesions For the 5IU experiments, double-strand oligonucleotide substrates were both within a hemimethylated and unmethylated CpG sequence prepared by annealing complementary, gel-purified single-strand 34-mer ¶ context and in a non-CpG context. oligonucleotides as follows: 5 CAATCCTAGCTGACACGATGTGGCCAAT- ¶ ¶ ¶ The results revealed a striking preference of the enzyme for GGCATG 3 ;3 GTTAGGATCGACTGTGYTACACCGGTTACCGTAC 5 , where Y is C (no-mismatch control), T, or 5IU. For the studies of A:5IU substrates, halogenated pyrimidines opposite guanine, including 5-iodouracil the mismatched G in the top strand was replaced by A. (5IU). For this reason, we next examined whether MED1 may play a For both the 37- and 34-mers, before annealing, the bottom ssDNA was significant role in the cytotoxicity of the halogenated nucleotide 5¶ labeled with [g-32P]ATP and T4 polynucleotide kinase (37), and salts and 5-iododeoxyuridine (5IdU), a 5IU precursor. For these experiments, excess label were removed using a nucleotide removal kit (Qiagen). we used the isogenic system represented by mouse embryo DNA N-glycosylase assay. DNA N-glycosylase assays were carried out fibroblasts (MEF) with different Med1 genotype derived from mice in 20 AL of reaction buffer [20 mmol/L HEPES (pH 7.9), 1 mmol/L DTT, with targeted inactivation of the Med1 gene (31). 1 mmol/L EDTA] supplemented with 1 mg/mL bovine serum albumin. Double-stranded oligodeoxynucleotide substrate (5 nmol/L) was incubated with purified wild-type or mutant MED1 glycosylase domain protein Materials and Methods (100 nmol/L) at 37jC for 30 minutes and subsequently treated with NaOH Protein overexpression and purification. The human MED1 protein (100 mmol/L) for 20 minutes at 85jC to break the apurinic/apyrimidinic or its glycosylase domain (amino acids 427-580 of the full MED1 protein) site. After the addition of denaturing gel loading buffer (formamide Â with an NH2-terminal 6 histidine tag were cloned in the pET28b containing 8 mol/L urea, 0.5 mol/L EDTA, 1 mg/mL bromophenol blue, and expression vector (Novagen, La Jolla, CA; ref. 22). Protein expression was 1 mg/mL xylene cyanol), substrate and product were resolved by denaturing carried out in Escherichia coli BL21 Codon Plus (DE3)-RIL cells (Stratagene, PAGE (15% gels containing 8 mol/L urea) in Tris-borate EDTA buffer

La Jolla, CA) grown to an A600 absorbance of between 0.4 and 0.6 with [89 mmol/L Tris-borate (pH 8.2), 1 mmol/L EDTA] at 30 W for 40 minutes. shaking at 37jC in Luria-Bertani broth containing kanamycin (30 Ag/mL). Manual single-turnover repair assay. For single-turnover conditions, Protein overexpression was induced by the addition of isopropyl-L-thio- the glycosylase assay was typically carried out in 80 AL of reaction buffer B-D-galactopyranoside to a final concentration of 1 mmol/L and further [20 mmol/L HEPES (pH 7.9), 1 mmol/L DTT, 1 mmol/L EDTA] supplemented incubation of the culture at 37jC with shaking for 3 hours. Cells were with 1 mg/mL bovine serum albumin. Purified MED1 glycosylase domain harvested by centrifugation (2,000 Â g for 20 minutes) and resuspended (1.6 Amol/L) was incubated with 5¶-radiolabeled oligodeoxynucleotide in lysis buffer containing 10 mmol/L potassium phosphate (pH 8.0), (80 nmol/L) at 37jC, and 5-AL samples were taken at regular time 500 mmol/L NaCl, 10 mmol/L imidazole, and Complete EDTA-free protease intervals, added to 2 AL of 250 mmol/L NaOH, and immediately frozen on inhibitors (Roche, Mannheim, Germany). Cells were sonicated on ice dry ice. Once all the samples had been collected, they were incubated at and the soluble fraction was isolated by centrifugation (10,000 Â g for 85jC for 20 minutes to cleave the apurinic/apyrimidinic site before the 40 minutes). The lysate was then applied by gravity flow to a 2-mL pre- addition of denaturing gel loading buffer. Substrate and product were equilibrated Ni2+-chelating sepharose column (Amersham, Piscataway, NJ) resolved by denaturing PAGE (15% gels), conducted as described above. and washed with 20 mL of wash buffer (10 mmol/L potassium phosphate Quench flow single-turnover repair assay. When hydrolysis rates were pH 8.0, 500 mmol/L NaCl, 10 mmol/L imidazole). Protein was eluted via a too fast for a manual assay, we used the quench flow assay to obtain the www.aacrjournals.org 7687 Cancer Res 2006; 66: (15). August 1, 2006

timed samples required for accurate k st measurements. The quench flow for single-turnover was confirmed by the fact that any increase in assay was carried out using a KimTek RQF-3 chemical quench flow protein concentration (i.e., to 2 or 5 Amol/L) failed to show an apparatus following previously published protocols (38). Briefly, assays increase of the hydrolysis rates (results not shown). were carried out under the same conditions outlined for the manual assay Initial single-turnover hydrolysis rates (k ) were obtained for with 100 mmol/L NaOH used as the quenching solution. Samples were st the G:T mismatch in three sequence contexts: hemimethylated immediately frozen on dry ice and subsequently analyzed as described above for the manual assay. and unmethylated CpG and non-CpG (Fig. 1). As summarized in Single-turnover data analysis. Following electrophoresis, radioactivity Table 1, the MED1 glycosylase domain showed the highest cleav- of each band, representing the amounts of substrate and product, was age rate for a G:T mismatch in which the mispaired G was quantified with a Fuji BAS-2500 bioimaging analyzer using the accompa- positioned within a hemimethylated CpG site, producing a kst À nying software package (Image Gauge 4.0). The counts present in each value of 0.859 F 0.01 min 1 (Fig. 1A and B). Removal of the individual band were calculated as a percentage of the total counts in the methylated context (Fig. 1C) resulted in a 15% reduction in activity À lane. Data from the single-turnover experiments were fitted to a first-order (0.736 F 0.01 min 1). This finding refines earlier experiments rate equation using nonlinear least-squares analysis with the software using modeled data, which showed no significant difference in GraFit. Manual and quench flow single-turnover rates were calculated as the MED1 activity when the mismatched G was in a methylated or mean of three and two separate experiments, respectively, with the error unmethylated CpG context (13). On the other hand, in agreement bars representing F1 SD. À À Cell lines. MEFs with different Med1 genotype (Med1+/+ and Med1 / ) with previous reports (12, 13), removal of the CpG context of the were obtained from day 12.5 embryos as previously described (31). To mismatch resulted in an almost 50% reduction in activity, F À1 minimize genetic variability, MEF cultures from littermate embryos were producing a kst value of 0.483 0.08 min (Fig. 1D). This used. MEFs were cultured in DMEM supplemented with 15% FCS, 1 mmol/L confirms the designation of MED1 as a guardian of CpG fidelity. sodium pyruvate, 2 mmol/L glutamine, 10 units/mL penicillin, and 10 Ag/mL The results obtained with the G:T mismatches were mirrored streptomycin (31). when T was replaced with U (Table 1; Fig. 2). Whereas the MED1 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide glycosylase domain showed a 2.5- to 3-fold increase in uracil versus assay. MEFs were seeded in 96-well plates (4,000 per well) and, after thymine glycosylase activity irrespective of the sequence context, 24-hour incubation, were treated with 5IdU at the indicated concentrations removal of the methyl group again produced a significant 28% for 72 hours. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide À reduction in activity (from 2.638 F 0.14 to 1.908 F 0.25 min 1; (MTT; 5 mg/mL solution in PBS) was added and the plates were incubated for 2 hours at 37jC. Cells were lysed in 20% SDS, 50% N,N-dimethylforma- Fig. 2B and C); as for G:T mismatches, removal of the CpG context F À1 mide, 2.5% glacial acetic acid, 2.5% HCl pH 4.7 (39). After overnight resulted in a 50% reduction in repair efficiency (1.266 0.05 min ; incubation at 37jC, plates were analyzed in an ELISA reader (570 nm). Fig. 2D). Surviving fraction was computed with respect to vehicle-treated cells. Recently, MED1 has not only been shown to remove thymine Colonyassay. MEFs were plated in 10-cm dishes (2 Â 103 per dish) when paired with the alkylation product O6-meG but has also been and, after 18-hour incubation, were treated with 5IdU at the indicated implicated in the DNA damage response to alkylating agents and concentrations for 48 hours. After removing the 5IdU-containing med- other DNA-damaging drugs (31). We therefore obtained hydrolysis j ium, cells were washed with PBS and incubated at 37 C for 7 days in rates for the glycosylase activity of MED1 against T when opposite complete DMEM. Colonies (>50 cells) were stained with crystal blue O6-meG in both a CpG (Fig. 3A) and non-CpG context (Fig. 3B). and counted. When in a CpG context, thymine was removed at a higher 5IdU incorporation analysis by high-performance liquid chroma- À1 Â 5 rate (0.653 F 0.03 min ) than in a non-CpG context (0.188 F tography. MEFs were plated in 10-cm dishes (5 10 per dish) and, after À1 4 16 hours, were treated with 5IdU concentrations ranging from 2 to 0.02 min ). Kinetic values for substrates containing 3,N - 30 Amol/L for 24 hours. Medium was removed after the 24-hour pulse with ethenocytosine were unobtainable due to a low yield of product 5IdU, and fresh DMEM with 15% fetal bovine serum and supplements was over time. added to plates. Time points were taken at time of 5IdU removal (time 0), MED1 activityon halogenated bases. Previously, MED1 has and 6, 12, and 24 hours after 5IdU removal. Time points were taken by been shown to have significant activity against the halogenated trypsinizing cells, centrifuging them to pellet, and washing twice with PBS uracil analogues 5-FU (13), 5-chlorouracil, and 5-bromouracil (24). À j in a similar manner before freezing the dry pellet at 80 C. DNA was Interestingly, in our study, initial assays using 5-FU as a substrate extracted from cells; alkaline RNA hydrolysis was done; and DNA was then produced very rapid hydrolysis rates, which could not be measured digested with DNase I, phosphodiesterase, and alkaline phosphatase, manually. Therefore, a quenched flow apparatus was used to obtain filtered, and then analyzed by high-performance liquid chromatography using authentic standards to determine nanomolar quantities of thymidine a catalytic rate for this substrate. In these experiments, the and 5IdU as previously described (40). difference in hydrolysis rates between the different sequence contexts was more defined and showed a higher rate of hydrolysis for a mismatch situated within a hemimethylated CpG context. Results Hydrolysis of a hemimethylated CpG site containing a G:5-FU Single-turnover hydrolysis reveals MED1 preference for mismatch (Fig. 4A) produced a 20-fold increase in activity substrates in a methylated CpG context. To date, catalytic rates compared with unhalogenated uracil (59.62 F 3.79 and 2.638 F À for the MED1 protein have only been assigned using modeled data 0.14 min 1, respectively). Removal of the methyl group reduced À resulting from the limited hydrolysis of substrate (13). To obtain activity by almost 60% (26.85 F 9.12 min 1) and positioning the more accurate rates from real-time experiments, we did a single- mismatch within a non-CpG context reduced activity by almost À turnover analysis that measures the slowest step of the catalytic 80% (10.49 F 2.2 min 1). The wild-type full-length MED1 protein cycle, up to and including hydrolysis, but does not include the also exhibits preference for the halogenated uracil over the product release step. These experiments were conducted with the nonhalogenated uracil (Fig. 4B). recombinant MED1 glycosylase domain due to the ease of Recently, the halogenated nucleotide, 5IdU, and its oral prodrug purification and higher activity than the full-length protein. The 5-iodo-2-pyrimidinone-2¶-deoxyribose have been employed as condition in which our hydrolysis assay meets the criteria required preclinical and clinical radiation sensitizers in human cancers

Cancer Res 2006; 66: (15). August 1, 2006 7688 www.aacrjournals.org

Figure 1. MED1 glycosylase domain hydrolysis of oligonucleotides containing a T:G mismatch. A, denaturing PAGE of the substrate and product produced on incubation with the MED1 glycosylase domain over time. B, single-turnover analysis of the glycosylase activity of MED1 on a T:G mismatch in a hemimethylated CpG context. C, single-turnover analysis of the glycosylase activity of MED1 on a T:G mismatch in an unmethylated CpG context. D, single-turnover analysis of the glycosylase activity of MED1 on T:G mismatch in a non-CpG context. Representative of three independent experiments for each substrate.

(41, 42). 5IdU is sequentially phosphorylated intracellularly to Given the high affinity MED1 shows for 5-FU, we repeated the 5IdUTP and competes with dTTP for incorporation (43). The extent hydrolysis assays with an oligodeoxynucleotide substrate contain- of radiosensitization correlates with the percent of thymine ing a single 5IU:G mismatch in a CpG context. Incubation of replacement in DNA by 5IU (44). Based on prior studies on muta- this substrate with the MED1 glycosylase domain produced a genesis in E phage and E. coli (45, 46), 5IdU can be incorporated single-turnover kst value similar to that observed with its 5-FU À into DNA to form either the pair A:5IU or the mispair G:5IU. The counterpart (i.e., 35.44 F 4.12 min 1 compared with 26.85 F À former is the result of incorporation of 5IdU opposite A during 9.12 min 1;Fig.4C). As the iodine atom at the 5 position of DNA synthesis. On the other hand, the mispair G:5IU may originate the pyrimidine ring has a similar van der Waals radius as the from misincorporation of 5IdU opposite G during DNA synthesis or, methyl group in thymine (43), the single-turnover rates for alternatively, from misincorporation of G opposite 5IdU. 5IU:G are >40-fold compared with the hemimethylated and unmethylated T:G mismatch. Whereas we detected elevated activity of MED1 on the G:5IU mispair, a very low activity was detected on an A:5IU pair (Fig. 4D). Table 1. Single-turnover rates for all the double-stranded Interestingly, similar results (i.e., G:5IU mispair processed with oligonucleotide substrates analyzed greater affinity than a G:T mispair and lack of recognition of the A:5IU mispair) were reported by our group with the mismatch À1 Mismatch context CpG context kst (min ) repair recognition complex hMutSa (47). We also confirmed (13) that MED1 has undetectable activity on an A:5-FU substrate T:G Hemimethylated 0.859 F 0.01 (Fig. 4D). F Unmethylated 0.736 0.01 Sensitivityof Med1-null cells to 5IdU. The 20- to 30-fold higher F Non-CpG 0.483 0.08 k values observed for MED1 with oligodeoxynucleotide sub- U:G Hemimethylated 2.638 F 0.14 cat strates containing the 5IU:G mismatch versus U:G and T:G Unmethylated 1.908 F 0.25 Non-CpG 1.266 F 0.05 mismatches suggest that MED1 is a major repair activity for 5IU T:O6-meG Unmethylated 0.653 F 0.03 that may play a significant role in 5IdU cytotoxicity. To functionally Non-CpG 0.188 F 0.02 test this hypothesis, we evaluated the sensitivity of cultures of 5-FU:G Hemimethylated 59.62 F 3.79 MEFs with different Med1 genotype derived from mice with Unmethylated 26.85 F 9.12 targeted inactivation of the Med1 gene (31). In this isogenic system, Non-CpG 10.49 F 2.2 cell cultures differ only for their wild-type or null Med1 status. 5IU:G Hemimethylated 67.86 F 0.13 Because 5IdU cytotoxicity is related to proliferation and DNA F Unmethylated 35.44 4.12 synthesis, we confirmed that wild-type and Med1-null MEFs exhibit similar cell cycle profile and proliferation rate (31). www.aacrjournals.org 7689 Cancer Res 2006; 66: (15). August 1, 2006

Figure 2. MED1 glycosylase domain hydrolysis of oligonucleotides containing a U:G mismatch. A, denaturing PAGE of the substrate and product produced on incubation with the MED1 glycosylase domain over time. B, single-turnover analysis of the glycosylase activity of MED1 on a U:G mismatch in a hemimethylated CpG context. C, single-turnover analysis of the glycosylase activity of MED1 on a U:G mismatch in an unmethylated CpG context. D, single-turnover analysis of the glycosylase activity of MED1 on U:G mismatch in a non-CpG context. Representative of three independent experiments for each substrate.

MEF cultures were treated with increasing doses of 5IdU for 3 Increased 5IdU incorporation in Med1-null cells. We next days and their surviving fraction was measured with the MTT determined whether the increased cytotoxicity and reduced colony À À assay. Med1 / MEFs displayed reduced survival in comparison formation of Med1-null cells might be related to augmented 5IdU with wild-type cultures (Fig. 5A). To corroborate these findings, we incorporation in their DNA. To this end, we conducted high- measured the colony-forming potential of wild-type and Med1-null performance liquid chromatography analyses on DNA samples À À MEFs: 5IdU cytotoxicity that would affect sensitive cells, preventing extracted from Med1+/+ and Med1 / MEFs after a 24-hour cell growth and colony formation. MEFs of different genotype were treatment with 5IdU (2-30 Amol/L). The results indicated that À À treated with increasing doses of 5IdU for 48 hours and then grown Med1 / MEFs incorporated 5IdU in their DNA at higher levels in drug-free medium for 7 days to allow the formation of colonies. than Med1+/+ MEFs, both at time of 5IdU removal and at 6, 12, and In comparison with wild-type MEFs, Med1-null MEFs showed a 24 hours after 5IdU removal (Fig. 6). These data suggest that the dramatic reduction in colony number (Fig. 5B) and size (data not increased sensitivity of Med1-null cells to 5IdU is associated with shown). augmented incorporation or reduced removal of 5IdU.

Figure 3. MED1 glycosylase domain hydrolysis of oligonucleotides containing a T:O6-meG mismatch. A, single-turnover analysis of the glycosylase activity of MED1 on a T:O6-meG mismatch in a hemimethylated CpG context. B, single-turnover analysis of the glycosylase activity of MED1 on a T:O6-meG mismatch in a non-CpG context. Representative of three independent experiments for each substrate.

Cancer Res 2006; 66: (15). August 1, 2006 7690 www.aacrjournals.org

Figure 4. MED1 glycosylase domain hydrolysis of oligonucleotides containing a halogenated U:G mismatch. A, single-turnover analysis of the glycosylase activity of MED1 on a 5-FU:G mismatch in a hemimethylated CpG context. B, wild-type full-length MED1 also exhibits preference for halogenated uracil. Denaturing PAGE of the substrate and product produced on incubation of wild-type full-length MED1 with 5-FU:G and U:G mismatches in an unmethylated CpG context over time. Percent product generation is indicated. C, single-turnover analysis of the glycosylase activity of MED1 on a 5IU:G mismatch in an unmethylated context. D, very low and undetectable activity of MED1 on 5IU:A (34-mer) and 5-FU:A (37-mer) pairs, respectively. Denaturing PAGE of the substrate and product produced on incubation for 20 minutes of wild-type full-length MED1 (fl) and its glycosylase domain (gd), or no enzyme (À), with the indicated halo-U:A pairs and halo-U:G mismatches (positive controls). Only one of two independent experiments is shown for each substrate.

Discussion significance and wide substrate spectrum of this enzyme, little Recent publications provide compelling evidence that the MED1 quantitative assessment of its sequence preference or context has glycosylase is a multifunctional protein involved not only in the been examined. The only known kinetic data available identify the excision repair of mismatched bases but also in the cellular response preference of MED1 for G:T and G:U mismatches that occur at CpG to DNA damage (31, 48, 49). MED1 recognizes a series of DNA lesions methylation sites (13). The same study failed to show any preference that result from a wide range of DNA damage, including deami- for the methylation status of the CpG site containing the lesion. This nation, oxidation, and alkylation. However, despite the growing is contrary to the data outlined in the present article. Whereas a

Figure 5. Med1 status affects sensitivity to 5IdU. A, survival analysis conducted with the MTT assay. Wild-type and Med1-null MEFs were treated with the indicated doses of 5IdU for 72 hours and the surviving fraction was detected by measuring the absorbance of metabolized MTT. The sensitivity of cells to 5IdU was tested by colony assay formation. Wild-type and Med1-null MEFs were treated with the indicated doses of 5IdU for 48 hours and then allowed to form colonies in drug-free medium for 7 days. Colonies were stained with crystal blue and their number was plotted with respect to untreated control cultures. Representative of an isogenic pair of cell lines (wild-type and Med1-deficient); similar results were obtained with several independent wild-type and Med1-null cell lines.

www.aacrjournals.org 7691 Cancer Res 2006; 66: (15). August 1, 2006

5-FU or 5IU produce single-turnover values more than 20- and 35-fold higher than those obtained for normal unhalogenated uracil and thymine, respectively. The reasons for the very efficient removal of halogenated uracil residues by MED1 are currently unknown. From a biochemical perspective, the G:5IU mismatch would be most similar to a G:T mismatch. However, the van der Waals radius of the iodine in the 5 position is slightly larger than the methyl group of thymidine, which may cause greater distortion of the DNA helix (43), thus facilitating preferential removal of the halo-U by the glycosylase domain of MED1. Modeling studies indicate that the mode of action of MED1 is analogous to that of OGG1 and AlkA family members that have a hydrophobic patch in line with the catalytic aspartic residue (52), thus conferring a more promiscuous substrate specificity (53). The elevated activity of the MED1 catalytic domain for 5IU and 5-FU may be explained by the high electro- negativity and hydrophobicity of the iodine and fluorine atoms (47, 54) and their high affinity for the hydrophobic cavity of the Figure 6. 5IdU DNA levels in Med1-null and wild-type MEFs. Wild-type and MED1 active site. An alternative possibility may be related to Med1-null MEFs were treated with the indicated doses of 5IdU for 24 hours and electron withdrawing effects of the halogen that facilitate the levels of 5IdU and thymidine in DNA were determined by high-performance liquid chromatography analysis in triplicate samples at time of 5IdU removal glycosylase reaction. (time 0), and 6, 12, and 24 hours after 5IdU removal. Results are plotted 5IdU is a radiosensitizing agent that is incorporated in the DNA as percent of 5IdU incorporation. as 5IU (42, 43, 47) and the extent of radiosensitization is related to distinct preference for a lesion situated within a CpG context was thymine replacement by 5IU (44). The results reported here reveal that, unlike wild-type cultures, MED1-null MEFs are dramatically confirmed, there was also a significant preference for a mismatch sensitive to cell killing by 5IdU in both MTT and colony formation situated within a CpG context that was hemimethylated over an assays. In keeping with this observation, we detected increased unmethylated CpG context. This preference was observed with all of incorporation of 5IdU in the genomic DNA of MED1-null MEFs in the relevant substrates analyzed. These contradicting data may be comparison with wild-type MEFs. These in vivo experiments explained by the fact that in the initial investigation, only 30% to 50% validate the 20- to 30-fold higher k values observed in vitro in the of the substrate was converted into product and therefore required cat single-turnover assay on halogenated versus nonhalogenated a kinetic simulation program to assign kinetic values (13). In the pyrimidine substrates (Table 1). Taken together, these findings present study, virtually all of the substrate was turned into product, establish MED1 as a bona fide repair activity involved in the allowing a direct real-time measurement of catalytic turnover. removal of halogenated bases and in the cytotoxicity of 5IdU. Interestingly, the ratios between catalytic rates for mismatches in To date, the only known molecular determinant of response to hemimethylated versus unmethylated CpG contexts were similar f the radiation-sensitizing agent 5IdU was mismatch repair status with different substrates ( 1.2, 1.4, 2.2, and 1.9 for thymine, uracil, (55). The observations reported here indicate that MED1 status 5-FU, and 5IU, respectively; see Table 1). This suggests that also has a significant effect on 5IdU cytotoxicity. Previous work recognition of the top strand opposite the mismatched pyrimidine using methoxyamine, a small-molecule inhibitor of base excision imparts a defined and specific effect on catalysis independent of repair, had suggested an involvement of this repair system in 5IdU the nature of the mismatched pyrimidine itself. Along the same cytotoxicity and radiosensitization (40). Our results indicate that lines, the ratio between unmethylated and non-CpG contexts is MED1 is one of the base excision repair activities involved in 5IdU f similar for T:G and U:G ( 1.5 for both substrates; see Table 1) but cytotoxicity. The results reported here may have important f O6-me is increased to 3.5 for T: G (see Table 1), a likely consequence implications in the oncology clinic as they predict that the 6 of altered hydrogen bonding due to methylation of the O position previously described tumors with MED1 inactivation (56, 57) may of the top strand guanine. It should be emphasized that the kinetic be highly sensitive to the cytotoxic and radiosensitizing activity of values obtained here may not necessarily be applicable to all 5IdU. Future experiments will address the role of MED1 in 5IdU sequences, as recently shown for the N-methylpurine-DNA radiosensitization of mammalian cells. glycosylase (50). However, similar results were obtained with longer (63-mer) substrates (data not shown). Acknowledgments The MED1 preference for hemimethylated lesions mirrors that Received 12/20/2005; revised 3/29/2006; accepted 5/17/2006. observed for the bacterial VSR protein, which also displays a Grant support: NIH grants CA78412, CA06927, CA50595, and CA84578; an preference for a hemimethylated mismatch within the cytosine appropriation from the Commonwealth of Pennsylvania to the Fox Chase Cancer Center; and fellowship from the Board of Associates of Fox Chase Cancer Center methylation sequence of the dcm-methyltransferase (51). As both (D.P. Turner). proteins prefer mismatches that occur at the sites of methylation, it The costs of publication of this article were defrayed in part by the payment of page is reasonable to assume that a hemimethylated context would be charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. their physiologic substrate and, therefore, the present data add This article is dedicated to the memory of our friend and colleague Elizabeth further credence to the assignation of MED1 as a guardian of (Betty) Travaglini, who was among the first to investigate the kinetics of incorporation of complementary and non-complementary nucleotides by DNA polymerase, and was methylation site fidelity (21, 32, 33). exemplary in placing the highest value upon friendship, tolerance and science above Importantly, by using the sensitive single-turnover assay, we were all, before fighting with dignity and serenity her battle against cancer. able to quantify the high activity of MED1 on mismatched, We thank D. Markham, Y. Matsumoto, and T. Yeung for critical reading of the manuscript; R. Sonlin for expert secretarial assistance; and the Fannie E. Rippel halogenated uracil residues, confirming results previously obtained Biotechnology Facility, the DNA Sequencing Facility, and the Cell Culture Facility at the by our laboratory and others (13, 24). Substrates containing either Fox Chase Cancer Center.

Cancer Res 2006; 66: (15). August 1, 2006 7692 www.aacrjournals.org

References repair: biological consequences. Int J Radiat Biol 1994; for measuring cell growth/cell kill. J Immunol Methods 66:579–89. 1989;119:203–10. 1. Ito Y, Jones PA, Vogt PK, et al., editors. DNA 21. Bellacosa A. Role of MED1 (MBD4) Gene in DNA 40. Taverna P, Hwang HS, Schupp JE, et al. Inhibition of methylation and cancer. Vol. 249. New York: Springer- repair and human cancer. J Cell Physiol 2001;187:137–44. base excision repair potentiates iododeoxyuridine- Verlag; 2000. 22. Bellacosa A, Cicchillitti L, Schepis F, et al. MED1, a induced cytotoxicity and radiosensitization. Cancer 2. El-Osta A. DNMT cooperativity—the developing links novel human methyl-CpG-binding endonuclease, inter- Res 2003;63:838–46. between methylation, chromatin structure and cancer. acts with the DNA mismatch repair protein MLH1. Proc 41. Seo Y, Yan T, Schupp JE, Colussi V, Taylor KL, Kinsella Bioessays 2003;25:1071–84. Natl Acad Sci U S A 1999;96:3969–74. TJ. Differential radiosensitization in DNA mismatch 3. Kleihues P, Schauble B, zur Hausen A, Esteve J, Ohgaki 23. Yoon JH, Iwai S, O’Connor TR, Pfeifer GP. Human repair-proficient and -deficient human colon cancer H. Tumors associated with p53 germline mutations: a thymine DNA glycosylase (TDG) and methyl-CpG- xenografts with 5-iodo-2-pyrimidinone-2¶-deoxyribose. synopsis of 91 families. Am J Pathol 1997;150:1–13. binding protein 4 (MBD4) excise thymine glycol (Tg) Clin Cancer Res 2004;10:7520–8. 4. Greenblatt MS, Bennett WP, Hollstein M, Harris CC. from a Tg:G mispair. Nucleic Acids Res 2003;31:5399–404. 42. Schulz CA, Mehta MP, Badie B, et al. Continuous Mutations in the p53 tumor suppressor gene: clues to 24. Valinluck V, Liu P, Kang JI, Jr., Burdzy A, Sowers LC. 28-day iododeoxyuridine infusion and hyperfractionated cancer etiology and molecular pathogenesis. Cancer Res 5-halogenated pyrimidine lesions within a CpG sequence accelerated radiotherapy for malignant glioma: a phase I 1994;54:4855–78. context mimic 5-methylcytosine by enhancing the binding clinical study. Int J Radiat Oncol Biol Phys 2004;59: 5. Pfeifer GP, Tang M-S, Denissenko MF. Mutation of the methyl-CpG-binding domain of methyl-CpG- 1107–15. hotspots and DNA methylation. In: Ito Y, Jones PA, Vogt binding protein 2 (MeCP2). Nucleic Acids Res 2005;33: 43. Kinsella TJ. An approach to the radiosensitization of PK, et al., editors. DNA methylation and cancer. New 3057–64. human tumors. Cancer J Sci Am 1996;2:184. York: Springer-Verlag; 2000. p. 1–19. 25. Spratt TE, Levy DE. Structure of the hydrogen 44. Miller EM, Fowler JF, Kinsella TJ. Linear-quadratic 6. Zingg JM, Jones PA. Genetic and epigenetic aspects of bonding complex of O6-methylguanine with cytosine analysis of radiosensitization by halogenated pyrimi- DNA methylation on genome expression, evolution, and thymine during DNA replication. Nucleic Acids Res dines. II. Radiosensitization of human colon cancer cells mutation and carcinogenesis. Carcinogenesis 1997;18: 1997;25:3354–61. by bromodeoxyuridine. Radiat Res 1992;131:90–7. 869–82. 26. Grover PL. Chemical carcinogens and DNA. Vol. 1. 45. Hutchinson F, Stein J. Mutagenesis of lambda phage: 7. Lutsenko E, Bhagwat AS. The role of the Escherichia Boca Raton (FL): CRC Press; 1979. p. 112–59. 5-bromouracil and hydroxylamine. Mol Gen Genet 1977; 4 coli mug protein in the removal of uracil and 3,N - 27. Karran P, Bignami M. Mismatch repair and cancer. 152:29–36. ethenocytosine from DNA. J Biol Chem 1999;274: In: Smith PJ, Jones CJ, editors. DNA recombination and 46. Rydberg B. Bromouracil mutagenesis in Escherichia 31034–8. repair. New York: Oxford Univ. Press; 1999. p. 66–159. coli evidence for involvement of mismatch repair. Mol 8. Duncan BK, Miller JH. Mutagenic deamination of 28. Bellacosa A. Functional interactions and signaling Gen Genet 1977;152:19–28. cytosine residues in DNA. Nature 1980;287:560–1. properties of mammalian DNA mismatch repair pro- 47. Berry SE, Loh T, Yan T, Kinsella TJ. Role of MutSa in 9. Wallace SS. Biological consequences of free radical- teins. Cell Death Differ 2001;8:1076–92. the recognition of iododeoxyuridine in DNA. Cancer Res damaged DNA bases. Free Radic Biol Med 2002;33:1–14. 29. YanT,BerrySE,DesaiAB,KinsellaTJ.DNA 2003;63:5490–5. 10. Neddermann P, Jiricny J. The purification of a mismatch repair (MMR) mediates 6-thioguanine geno- 48. Sansom OJ, Zabkiewicz J, Bishop SM, Guy J, Bird A, mismatch-specific thymine-DNA glycosylase from HeLa toxicity by introducing single-strand breaks to signal a Clarke AR. MBD4 deficiency reduces the apoptotic cells. J Biol Chem 1993;268:21218–24. G2-M arrest in MMR-proficient RKO cells. Clin Cancer response to DNA-damaging agents in the murine small 11. Neddermann P, Jiricny J. Efficient removal of uracil Res 2003;9:2327–34. intestine. Oncogene 2003;22:7130–6. from GÁU mispairs by the mismatch-specific thymine 30. Yan T, Desai AB, Jacobberger JW, Sramkoski RM, Loh 49. Parsons BL. MED1: a central molecule for mainte- DNA glycosylase from HeLa cells. Proc Natl Acad Sci T, Kinsella TJ. CHK1 and CHK2 are differentially nance of genome integrity and response to DNA U S A 1994;91:1642–6. involved in mismatch repair-mediated 6-thioguanine- damage. Proc Natl Acad Sci U S A 2003;100:14601–2. 12. Hendrich B, Hardeland U, Ng HH, Jiricny J, Bird A. induced cell cycle checkpoint responses. Mol Cancer 50. Xia L, Zheng L, Lee HW, et al. Human 3-methyl- The thymine glycosylase MBD4 can bind to the product Ther 2004;3:1147–57. adenine-DNA glycosylase: effect of sequence context on of deamination at methylated CpG sites. Nature 1999; 31. Cortellino S, Turner D, Masciullo V, et al. The base excision, association with PCNA, and stimulation by AP 401:301–4. excision repair enzyme MED1 mediates DNA damage endonuclease. J Mol Biol 2005;346:1259–74. 13. Petronzelli F, Riccio A, Markham GD, et al. Biphasic response to antitumor drugs and is associated with 51. Turner DP, Connolly BA. Interaction of the E. coli kinetics of the human DNA repair protein MED1 mismatch repair system integrity. Proc Natl Acad Sci DNA G:T-mismatch endonuclease (vsr protein) with (MBD4), a mismatch-specific DNA N-glycosylase. J Biol U S A 2003;100:15071–6. oligonucleotides containing its target sequence. J Mol Chem 2000;275:32422–9. 32. Millar CB, Guy J, Sansom OJ, et al. Enhanced CpG Biol 2000;304:765–78. 14. Petronzelli F, Riccio A, Markham GD, et al. Investi- mutability and tumorigenesis in MBD4-deficient mice. 52. Wu P, Qiu C, Sohail A, Zhang X, Bhagwat AS, Cheng gation of the substrate spectrum of the human Science 2002;297:403–5. X. Mismatch repair in methylated DNA. Structure and mismatch-specific DNA N-glycosylase MED1 (MBD4): 33. Wong E, Yang K, Kuraguchi M, et al. Mbd4 activity of the mismatch-specific thymine glycosylase Fundamental role of the catalytic domain. J Cell Physiol inactivation increases Cright-arrowT transition muta- domain of methyl-CpG-binding protein MBD4. J Biol 2000;185:473–80. tions and promotes gastrointestinal tumor formation. Chem 2003;278:5285–91. 15. Neddermann P, Gallinari P, Lettieri T, et al. Cloning Proc Natl Acad Sci U S A 2002;99:14937–42. 53. Stivers JT, Jiang YL. A mechanistic perspective on the and expression of human G/T mismatch-specific 34. Stoscheck CM. Quantitation of protein. Methods chemistry of DNA repair glycosylases. Chem Rev 2003; thymine-DNA glycosylase. J Biol Chem 1996;271: Enzymol 1990;182:50–68. 103:2729–59. 12767–74. 35. Ho SN, Hunt HD, Horton RM, Pullen JK, Pease LR. 54. Coll M, Saal D, Frederick CA, Aymami J, Rich A, Wang 16. Scha¨rerOD, Jiricny J. Recent progress in the biology, Site-directed mutagenesis by overlap extension using AH. Effects of 5-fluorouracil/guanine wobble base pairs chemistry and structural biology of DNA glycosylases. the polymerase chain reaction. Gene 1989;77:51–9. in Z-DNA: molecular and crystal structure of Bioessays 2001;23:270–81. 36. Oleykowski CA, Bronson Mullins CR, Godwin AK, d(CGCGFG). Nucleic Acids Res 1989;17:911–23. 17. David SS, Williams SD. Chemistry of glycosylases and Yeung AT. Mutation detection using a novel plant 55. Berry SE, Kinsella TJ. Targeting DNA mismatch endonucleases involved in base-excision reair. Chem endonuclease. Nucleic Acids Res 1998;26:4597–602. repair for radiosensitization. Semin Radiat Oncol 2001; Rev 1998;98:1221–61. 37. Sambrook J, Fritsch EF, Maniatis T. Molecular 11:300–15. 18. Hickson ID. General introduction to DNA base cloning: a laboratory manual. Cold Spring Harbor 56. Bader S, Walker M, Hendrich B, et al. Somatic excision repair. In: Hickson ID, editor. Base excision (NY): Cold Spring Harbor Laboratory Press; 1989. frameshift mutations in the MBD4 gene of sporadic repair of DNA damage. Austin (TX): Landes Bioscience 38. Connolly BA, Liu HH, Parry D, Engler LE, Kurpiewski colon cancers with mismatch repair deficiency. Onco- (Springer-Verlag); 1997. p. 1–5. MR, Jen-Jacobson L. Assay of restriction endonucleases gene 1999;18:8044–7. 19. Krokan HE, Nilsen H, Skorpen F, Otterlei M, using oligonucleotides. Methods Mol Biol 2001;148: 57. Riccio A, Aaltonen LA, Godwin AK, et al. The DNA Slupphaug G. Base excision repair of DNA in mamma- 465–90. repair gene MBD4 (MED1) is mutated in human lian cells. FEBS Lett 2000;476:73–7. 39. Hansen MB, Nielsen SE, Berg K. Re-examination and carcinomas with microsatellite instability. Nat Genet 20. Wallace SS. DNA damages processed by base excision further development of a precise and rapid dye method 1999;23:266–8.

www.aacrjournals.org 7693 Cancer Res 2006; 66: (15). August 1, 2006

Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2006 American Association for Cancer Research. The DNA N-Glycosylase MED1 Exhibits Preference for Halogenated Pyrimidines and Is Involved in the Cytotoxicity of 5-Iododeoxyuridine

David P. Turner, Salvatore Cortellino, Jane E. Schupp, et al.

Cancer Res 2006;66:7686-7693.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/66/15/7686

Cited articles This article cites 51 articles, 17 of which you can access for free at: http://cancerres.aacrjournals.org/content/66/15/7686.full#ref-list-1

Citing articles This article has been cited by 4 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/66/15/7686.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/66/15/7686. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.