Thiopurine Methyltransferase Predicts the Extent of Cytotoxicty and DNA Damage in Astroglial Cells after Thioguanine Exposure

Thiopurine methyltransferase (Tpmt) is the primary enzyme responsible for deactivating thiopurine drugs. Thiopurine drugs (i.e., thioguanine [TG], mercaptopurine, azathioprine) are commonly used for the treatment of cancer, organ transplant, and autoimmune disorders. Chronic thiopurine therapy has been linked to the development of brain cancer (most commonly astrocytomas), and Tpmt status has been associated with this risk. Therefore, we investigated whether the level of Tpmt protein activity could predict TG-associated cytotoxicity and DNA damage in astrocytic cells. We found that TG induced cytotoxicity in a dose-dependent manner in Tpmt+/+, Tpmt+/− and Tpmt−/− primary mouse astrocytes and that a low Tpmt phenotype predicted significantly higher sensitivity to TG than did a high Tpmt phenotype. We also found that TG exposure induced significantly more DNA damage in the form of single strand breaks (SSBs) and double strand breaks (DSBs) in primary astrocytes with low Tpmt versus high Tpmt. More interestingly, we found that Tpmt+/− astrocytes had the highest degree of cytotoxicity and genotoxicity (i.e., IC50, SSBs and DSBs) after TG exposure. We then used human glioma cell lines as model astroglial cells to represent high (T98) and low (A172) Tpmt expressers and found that A172 had the highest degree of cytoxicity and SSBs after TG exposure. When we over-expressed Tpmt in the A172 cell line, we found that TG IC50 was significantly higher and SSB's were significantly lower as compared to mock transfected cells. This study shows that low Tpmt can lead to greater sensitivity to thiopurine therapy in astroglial cells. When Tpmt deactivation at the germ-line is considered, this study also suggests that heterozygosity may be subject to the greatest genotoxic effects of thiopurine therapy.


Introduction
Thiopurine methyltransferase (Tpmt) is a cytoplasmic enzyme responsible for catalyzing the S-methylation of aromatic and heterocyclic compounds and is the primary deactivating enzyme for thiopurine drugs [1]. Thiopurines are antimetabolite pro-drugs that require intracellular conversion to the active metabolites to exert their cytotoxic effects ( Fig. 1) [2]. Alternatively, thiopurine activation is in competition with deactivation pathways primarily involving Tpmt. Tpmt has been studied extensively because of inter-patient variability in Tpmt protein expression owing to a few common genetic polymorphisms [3,4,5]. Deactivating polymorphisms in Tpmt can result in a trimodal population distribution in enzymatic activity. Inheritance of two functional alleles results in high protein activity (,90% of individuals), heterozygosity leads to intermediate activity (,10% of individuals), and inheritance of two dysfunctional alleles results in practically no protein activity (,1% of individuals) [6,7]. Importantly, low Tpmt has been associated with life-threatening toxicities including cancer in patients receiving thiopurine therapy [8,9,10,11,12].
Over the past several decades, thiopurines [e.g. thioguanine (TG), mercaptopurine, and azathioprine] have been commonly used in the treatment of cancer, autoimmune disorders, and in organ transplant. In each of these patient populations there have been a number of reports of cancer development after thiopurine therapy to include acute myeloid leukemias, lymphomas, and cancers of the skin and brain [8,13,14,15,16,17,18,19,20,21,22,23,24]; although, the relevance of Tpmt status is not clear. However, in a study by Relling et al it was found that when patients were treated with antimetabolites (i.e. mercaptopurine, methotrexate) and prophylactic cranial irradiation as part of their antileukemic therapy, whether they developed a brain tumor depended on their Tpmt status [8,24]. Patients who had a low Tpmt phenotype were more likely to develop a brain tumor as a late complication than those who had high Tpmt (42.9% versus 15.8% 10-year cumulative incidence, respectively) [8].
Tpmt is constitutively expressed in several tissues including blood, kidney, liver, and brain [25,26,27,28,29]. The association of Tpmt status with brain tumor risk after thiopurine exposure prompts the question of whether Tpmt genotypes can predict thiopurine drug phenotypes in astroglial cells. The relevance of Tpmt status to thiopurine-associated phenotypes in important brain cell populations has not been studied. Indeed, the first step to address the question of whether Tpmt plays any role in brain cancer risk after thiopurines is to determine whether Tpmt phenotypes in the brain are similar to those in other tissues. Thiopurine drug-associated genotoxicity has been linked to mutagenesis and transformative events [30,31,32]. We hypothesize that Tpmt deficiency can lead to greater genotoxicity (i.e. DNA damage) and cytotoxicity in astroglial cells after thiopurine exposure. To address this hypothesis we used primary astrocytes isolated from transgenic mice of each Tpmt genotype and performed in vitro studies to compare thiopurine-induced cytotoxicity and DNA damage between Tpmt genotypes. We also used established human glioma cell lines as model astroglial cells to validate the findings observed in primary mouse astrocytes.
Primary cultures were established from mouse cortices of each Tpmt genotype and astrocyte purity was confirmed by GFAP (glial fibrillary acid protein, an astrocyte marker) fluorescent staining (Fig. 2a). We found that there was no significant difference in growth rates between cultures of different Tpmt genotypes (Fig. 2b). The mean doubling rates for Tpmt +/+ , Tpmt +/2 and Tpmt 2/2 cultures were 33.9, 34.2, and 29.6 hours, respectively (p = 0.89). Next, we measured the level of Tpmt protein activity using a well-established radiochemical assay that measures the conversion of 6-mercaptopurine to 6-methylmercaptopurine using [14C] S-adenosyl-Lmethionine (SAM) as a methyl donor [34]. In agreement with previous studies comparing protein activity in hematological cells of different Tpmt genotypes [5,29,35], protein activities in astrocytes were significantly different and were predicted by Tpmt genotype (Fig. 2c). Tpmt +/+ , Tpmt +/2 , and Tpmt 2/2 primary astrocytes displayed high (7.6 unit/10 9 mg protein), intermediate (3.9 unit/ 10 9 mg protein), and low (0.25 unit/10 9 mg protein) levels of protein activity, respectively (p = 0.018). Based on these data, we next determined whether Tpmt could modulate the cytotoxic effects of TG in primary astrocytes.
Comparison of thioguanine-associated cytotoxicity in Tpmt +/+ , Tpmt +/2 , and Tpmt 2/2 primary astrocyte cultures To determine whether Tpmt could predict thiopurine cytotoxicity in primary astrocytes, the MTT colorimetric assay was Figure 1. Thiopurine drug metabolism pathway. Azathioprine (AZA) is a prodrug of mercaptopurine (MP). Thioguanine (TG) and MP can be converted by hypoxanthine phosphoribosyl transferase (HPRT) to thioguanine nucleotide (TGNs) metabolites. MP can also be converted to the methylthioinosine monophosphate (meTIMP) by thiopurine methyltransferase (Tpmt) which inhibits purine synthesis; however, thioguanine bypasses the conversion to this metabolite. Thiopurines can be converted to inactive metabolites [i.e. methyl-mercaptopurine (meMP); methylthioguanine (meTG); and methyl-thioguanine nucleotides (meTGNs)] by Tpmt. TGN metabolites are incorporated into DNA and RNA leading to cell death. However, DNA incorporation is believed to be the primary mode of cytotoxicty. doi:10.1371/journal.pone.0029163.g001 employed to compare TG sensitivity between astrocytes of each Tpmt genotype. This assay measures cell viability by quantitating the extent at which metabolically active cells can reduce tetrazolium salt (MTT) to form purple formazan crystals. Astrocytes were exposed to escalating concentrations of TG (0.125, 1.25, 12.5, 25, 50, and 100 mM) for five days and then subjected to the MTT assay. Cell viability data at each TG concentration was used to calculate IC 50 values. As expected, TG treatment resulted in cytotoxicity in a dose dependant manner (Figure 3a). Cell viability was significantly lower in Tpmt +/2 and Tpmt 2/2 cultures at four concentrations of TG (12.5, 25, 50, and 100 mM) when compared to Tpmt +/+ cultures (p,0.04). Interestingly, Tpmt +/2 but not Tpmt 2/2 astrocytes were significantly more sensitive at 1.25 mM of TG than Tpmt +/+ astrocytes (p = 0.004); and this finding was reflected in corresponding IC 50 values. There was a 2.9 fold reduction in IC 50 in Tpmt +/2 (3.5 mM TG) versus a 1.2 fold reduction in Tpmt 2/2 (8.6 mM TG) when compared Tpmt +/+ astrocytes (10.3 mM TG); and the Tpmt +/2 and Tpmt 2/2 IC 50 values were significantly different from Tpmt +/+ (p,0.05) (Fig. 3b). Studies have shown that genotoxicity (i.e. DNA strand breaks) correlates with cytotoxicity after TG exposure [33,36]. Hence, we investigated whether the degree of TG-induced DNA strand breaks was associated with the level of cytotoxicity in primary astrocytes.
Comparison of the extent of DNA damage between Tpmt +/+ , Tpmt +/2 , and Tpmt 2/2 primary astrocytes To address the question of whether the level of TG-induced cytotoxicity is correlated with DNA damage, we employed the alkaline comet assay. This assay allows for the quantitation of DNA damage [i.e. single strand breaks (SSBs), double strand breaks (DSBs) and alkali labile sites] by using fluorescent showing high astrocyte purity determined by GFAP (green) and propidium iodide (red) fluorescent staining. (b) There was no significant difference in growth rates between astrocyte cultures of each Tpmt genotype (p = 0.89). (c) The level of Tpmt protein activity was significantly different between the three Tpmt genotypes (p = 0.018); and levels were comparable to Nalm6 and CEM control lymphoblastoid cell lines. Bars represent the means; whiskers represent the standard deviation (6); * p,0.05. doi:10.1371/journal.pone.0029163.g002 microscopy to visualize the extent of DNA migration from the nucleus of individual cells embedded in agarose [37]. We exposed astrocytes to 5 mM of TG for 72 hours. IC50 data was used to select a dose that would allow us to capture damage in each genotype. 72 hour time-point was selected to allow at least two cell doublings for TG incorporation and to capture cell damage prior to cell death. We observed a significant difference in DNA damage between the three genotypes as measured by mean comet tail moment. Both Tpmt +/2 and Tpmt 2/2 astrocytes had a significantly greater degree of DNA damage than Tpmt +/+ astrocytes (mean comet tail moment = 2.64, 1.26, and 0.4, respectively; p = 0.034) (Fig. 3c). In agreement with MTT studies, Tpmt +/2 astrocytes were more sensitive to TG, as evidenced by a greater degree of DNA damage, than were Tpmt 2/2 astrocytes. Although the difference in DNA damage between Tpmt +/2 and Tpmt 2/2 astrocytes was not significantly different (p = 0.083). Tpmt genotype were exposed to thioguanine (TG) and cell viability was determined using the MTT assay. Tpmt +/2 and Tpmt 2/2 astrocytes were significantly more sensitive to TG (12.5-100 mM) than Tpmt +/+ (p,0.04). (b) IC 50 values were significantly different between the three genotypes (p = 0.027) and Tpmt +/2 astrocytes were the most sensitive. (c) The alkaline comet assay was used to compare TG-induced DNA damage between genotypes. Tpmt +/2 and Tpmt 2/2 astrocytes had significantly higher mean comet tail moments than Tpmt +/+ astrocytes (p = 0.033). (d) Immunoflourescence staining for cH2AX confirms IC 50 and comet data. Tpmt +/2 astrocytes had greater damage than Tpmt +/+ (p = 0.03). Below are representative images of cH2AX staining for control treated (left) and TG treated (right) astrocytes. cH2AX foci stained in green; nuclei stained in red (Zeiss LSM 710 Laser Scanning Microscopes, 4006magnification). Bars represent the means; whiskers represent the standard deviation (6); * p,0.05. doi:10.1371/journal.pone.0029163.g003 There was no difference in DNA damage between Tpmt genotypes of control treated astrocytes.
Studies have shown that the predominant DNA lesions caused by TG are SSBs although some DSB lesions are formed [33]. Therefore, we then compared cH2AX (an established DSB marker) foci staining after TG exposure to determine the extent of the DNA lesions that are in the form of DSBs and whether this damage is different between Tpmt genotypes. Astrocytes were seeded on glass coverslips and exposed to either 0 or 5 mM of TG. We found that Tpmt +/2 astrocytes had the greatest degree of foci staining (19.4%) compared to Tpmt +/+ (9.3%), and Tpmt 2/2 (7.22%) (Fig. 3d). In agreement with comet data, Tpmt +/2 was the most sensitive genotype. There was a significant difference between Tpmt +/2 and Tpmt +/+ (p = 0.03) but no significant difference was observed between Tpmt 2/2 and Tpmt +/+ cells.

Comparison of thioguanine phenotypes between human glioma cell lines with different Tpmt protein activities
To determine whether our findings could be recapitulated in human cell lines, we used established human glioma cell lines as model astroglia and conducted MTT and comet experiments. We first identified cell lines with high and low Tpmt activity and compared TG-induced cytotoxicity and DNA damage between them. First, we measured Tpmt protein activity in four human glioma cell lines (T98, A172, U87, and LN18) to identify a high and low Tpmt expresser. We chose T98 (10.48 unit/10 9 mg protein) and A172 (6.15 unit/10 9 mg protein) as the high and low expressers, respectively (Fig. 4a). We then subjected T98 and A172 to the MTT assay to measure TG-induced cytotoxicity. Both cell lines were exposed to escalating concentrations of TG (0.15, 0.75, 1.5, 6.25, 12.5 and 25 mM) for five days. The cell viability was determined for each concentration and these data were used to calculate IC 50 values. As predicted by Tpmt activity and in agreement with MTT studies performed in primary astrocytes, A172 had significantly lower cell viabilities at each TG concentration as compared to T98 (p,0.05) (Fig. 4b). The higher TG sensitivity of A172 was also reflected in the IC 50 values. A172 displayed a 6.9 fold lower IC 50 (1.2 mM) when compared to T98 (8.3 mM) (Fig. 4c).
Next, we employed the comet assay to compare DNA damage between T98 and A172 after exposing cells to 3 mM TG for 72 hours. In agreement with primary astrocyte data, a low level of Tpmt activity led to significantly more DNA damage compared to high activity (mean comet tail moment: A172 = 1, T98 = 0.3; p = 0.02) (Fig. 4d). DNA damage in control treated cells was not significantly different between the two cell lines.

Comparison of thioguanine phenotypes between A172 mock and A172 Tpmt isogenic cell lines
To further validate our findings, TG phenotypes were compared using isogenic cell lines. A172 was the lowest Tpmt expresser, hence we over-expressed Tpmt in this cell line and compared TG-induced cytotoxicity and DNA damage between the mock transduced (A172 mock ) and Tpmt transduced (A172 Tpmt ) cell lines. We performed growth rate studies and found that there were no differences in doubling times between A172 Tpmt and A172 mock (28.5 and 29.6 hours, respectively; p = 0.38). The level of Tpmt protein activity in A172 Tpmt was significantly higher as compared to A172 mock (151.23 unit/10 9 mg protein versus 9.47 unit/10 9 mg protein, respectively; p,0.05) (Fig. 5a). When A172 mock and A172 Tpmt were exposed to escalating concentrations of TG for five days we found that A172 mock had significantly lower cell viabilities at 0.75, 1.5, 6.25, 12.5 and 25 mM of TG as compared to A172 Tpmt (p,0.005) (Fig. 5b). These findings were reflected in the A172 Tpmt IC 50 value that was approximately twofold higher than A172 mock (3.9 mM versus 1.9 mM, respectively; p,0.05) (Fig. 5c).
Next, we exposed A172 Tpmt and A172 mock to 5 mM of TG for 72 hours and compared the extent of DNA damage using the comet assay. We selected 5 mM based on the IC50 value of A172 Tpmt . As expected, A172 mock had a significantly greater degree of DNA damage when compared to A172 Tpmt (mean comet tail moment: A172 mock = 0.7, A172 Tpmt = 0.24; p = 0.03) (Fig. 5d). DNA damage in control treated cells was not significantly different between the two cell lines (p = 0.35). These data support our findings observed in both the primary mouse astrocytes and the glioma cell lines (T98 and wild-type A172) that when Tpmt protein activity is low a greater degree of cytotoxicity and DNA damage can occur after TG exposure.

Discussion
Tpmt is a cytoplasmic enzyme responsible for deactivating thiopurine drugs and a low Tpmt phenotype can lead to life-threatening hematological toxicities when thiopurines are given. Indeed, chronic administration of thiopurines has been linked to an increased risk of cancer including cancers of the brain [8,15,17,22,24]. Unfortunately, not all patient populations that are treated with thiopurines are evaluated for their Tpmt status prior to initiating thiopurine therapy; hence, the importance of Tpmt to thiopurine-associated cancer risk is not clear. This fact coupled with the fact that Tpmt deficiency is relatively rare across populations, makes it difficult to assess the importance of Tpmt status to thiopurine therapy-related cancer risk. However, in a report by Relling et al whether patients developed a brain tumor as a late complication post-antileukemic therapy depended on their Tpmt status [24]. In this report, patients who were determined to be at high risk for treatment relapse received prophylactic cranial irradiation concurrent with antimetabolite therapy; and those patients with a low Tpmt had a surprisingly high incidence of brain tumors as a late complication. Indeed, ionizing radiation in itself is a risk factor for brain tumorigenesis. However, in this report, patients with low Tpmt had a 42.9% incidence of brain tumors compared to 15.8% for those with high Tpmt.
Astrocytomas are the most common brain tumor type that has been reported in patients after thiopurine therapy [8,15,22,24]. The importance of Tpmt status to thiopurine-associated phenotypes in astroglial cells has never been studied. We assessed Tpmt and TG phenotypes in astroglial cells using primary mouse astrocytes of each Tpmt genotype and human glioma cells of different Tpmt phenotypes. In this study we have shown that TG induces cytotoxicity in a dose-dependent manner in astroglial cells; and Figure 5. Comparison of Tpmt-associated phenotypes in A172 isogenic cells. Tpmt protein activity was measured in A172 mock and A172 Tpmt cell lines. (a) Tpmt protein activity was significantly higher in A172 Tpmt (p,0.05); additionally, (b) A172 Tpmt showed a significantly higher resistance to TG-induced toxicity when compared to A172 mock as determined by the MTT cell viability data (p,0.005) and (c) by IC 50 values (p,0.05). (d) The alkaline comet assay was used to compare TG-induced DNA damage. A172 mock displayed a significantly higher DNA damage than A172 Tpmt (p = 0.03) as measured by mean comet tail moment. Bars represent the means; whiskers represent the standard deviation (6). * p,0.05. doi:10.1371/journal.pone.0029163.g005 cells with a low Tpmt phenotype (Tpmt +/2 and Tpmt 2/2 astrocytes and A172 mock ) were significantly more sensitive. We also showed that TG exposure induced more DNA damage in the form of SSBs in astroglial cells with low Tpmt than in cells with high Tpmt. DNA damage in the form of DSBs was significantly different between astroglial cells with low and high Tpmt. SSBs have been shown to be the primary DNA lesion induced by thiopurine drugs [33].
Interestingly, we found that low but not the complete absence of Tpmt sensitized primary astrocytes to a higher degree of cytotoxicity and genotoxicity (i.e. IC 50 , SSBs and DSBs) after TG exposure. This discordance could possibly be explained by the induction of a compensatory mechanism that deactivates thiopurines with the complete loss of Tpmt. Another possible explanation could be a defect in the mismatch repair (MMR) system's ability to recognize DNA-TG nucleotide lesions. The MMR system is responsible for recognizing DNA mis-pairing caused by DNA-TG nucleotides and the subsequent signaling for cell death [38,39,40,41]. The latter possibility suggests that astroglial cells with a defective Tpmt could have DNA lesions persisting in the genome that could potentially lead to mutagenesis and cancer. Indeed, some studies suggest that heterozygous children develop brain tumors and other cancers more often than children with mutant Tpmt. However, the frequency of heterozygous or mutant Tpmt is very low making it is difficult to make any definitive conclusions [24,42]. It would be interesting to investigate whether Tpmt status influences the extent of TG nucleotide incorporation into DNA and damage recognition by the MMR system as well as whether ionizing irradiation alone or combined with thiopurines results in similar findings.
In summary, chronic thiopurine therapy has been associated with the development of brain cancer and Tpmt status has been linked to this risk. We found that Tpmt is an important factor for thiopurine drug-induced cytotoxicity and genotoxicity in astroglial cells. Even though glioma cells can exhibit a number of genetic alternations that could contribute to the phenotypes observed, our study shows that Tpmt is indeed an important determinant of thiopurine toxicity. Low Tpmt can increase the level of TG-induced genotoxic damage in both normal and in neoplastic astroglial cells. Indeed, the observations that glioma cells can have variable degrees of Tpmt protein activity and that low protein activity can cause cells to be more sensitive to TG poses the question of whether thiopurines may potentially have a therapeutic role in low Tpmt expressing glial tumors. In vitro studies have shown that thiopurine drugs are associated with mutagenic events in a variety of cell lines [30,36]. It is possible that the DNA damage caused by low Tpmt function can ultimately contribute to transformative events over time. One study found that Tpmt was not an important factor for cancer risk after thiopurine therapy [42]; however in this study only Tpmt genotype was determined which may have excluded patients with discordance between Tpmt genotype and phenotype. Individuals who are wild-type for Tpmt can display a wide variation in their Tpmt phenotype [43]. Further studies are needed to understand the importance of Tpmt in the brain when thiopurines are given. Whether theses findings can be replicated in vivo and whether factors important to DNA repair mechanisms are responsible for the TG phenotypes described in this study should also be investigated.

Animals
The Tpmt knockout mice were a generous gift of Dr. Mary Relling (St. Jude Children's Research Hospital) and the generation of these mice has been described previously [35]. The Tpmt +/2 , and Tpmt 2/2 mice were indistinguishable from Tpmt +/+ mice by appearance, life span, and organ histology (i.e. brain, liver, spleen, intestine, thymus, and lymph nodes). Animals were maintained in accordance to NIH guidelines for the care and use of laboratory animals and were housed in an AAALAC accredited facility. The University of Tennessee Health Science Center Institutional Animal Care and Use Committee approved all animal procedures.

Primary astrocyte cultures and cell lines
Astrocyte cultures were established from 2-5 day old pups of each Tpmt genotype. The DNA extracted from tails snips was used to confirm Tpmt genotype. A dissecting microscope was used to remove the meninges and hippocampus and the cortices were mechanically dissociated. Subsequently, cultures were established and maintained in Dulbecco's modified Eagle's medium/Ham's F-12 50/50 mix (Cellgro) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 IU/ml penicillin, 100 mg/ml streptomycin and 20 ng/ml epidermal growth factor (Millipore) and grown in Primaria flasks, (BD biosciences). Cells were refed with 20 ng/ml epidermal growth factor after three days of incubation, and the media was changed after five days. At passage 2, three individual primary cultures of the same genotype were pooled and seeded for experiments. Only cultures with greater than 80% astrocyte purity as determined by GFAP immunohistochemistry were included in experiments.

Lentiviral transduction
Human Tpmt cDNA was obtained by digesting pCMV6-Tpmt plasmid (Origene) at BamHI and EcoRI sites. Tpmt cDNA was then amplified by PCR, purified through agarose gel electrophoresis, and subcloned into BamHI and EcoRI sites of pLenti-CMVpgk-puro vector (Viral vector core, UTHSC). The plenti-Tpmtpgk-puro and plenti-GFP-pgk-puro lentivial vectors were produced by packaging in 293T cells [44]. The lentiviral vectors were used to transduce the A172 cell line in the presence of 6 mg/mL polybrene (Sigma). Transduction efficiency was monitored by GFP expression using fluorescent microscope. The mock-(A172 mock ) and Tpmt-transduced (A172 Tpmt ) cell lines were seeded for experiments at comparable passage numbers and Tpmt activity was measured prior to the experiments.

Cell proliferation comparison
The doubling rates for primary astrocyte cultures were determined and used as an index of cell proliferation. At passage two, 75,000 cells/well were seeded in duplicate 6-well plates. Every 24 hours the cells from one well of each plate were harvested using 0.05% Trypsin-EDTA (GIBCO) and viable cells were counted by trypan blue exclusion for four consecutive days. Doubling rates were calculated using the following formula:

TPMT activity assay
Primary mouse astrocyte and human glioma cell lines were sonicated using a probe sonicator (Masonix XL-2000 series) prior to performing the Tpmt activity assay. The level of Tpmt activity was determined using a pre-established non-chelated radiochemical assay [26,34] with slight modifications. Tpmt activity was quantitated by measuring the extent of conversion of 6mercaptopurine (Sigma) to radioactively labeled 6-methylmercaptopurine with [14C] S-adenosyl-L-methionine (SAM) (Perkin-Elmer). The protein concentration of each lysate was measured prior to performing the assay and was used to calculate Tpmt activity. The mean activities for each genotype were calculated using data from triplicate inter-day studies of three separate pooled cultures per Tpmt genotype.

MTT assay
Cell viability was measured using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma). Briefly, 1000 cells were seeded per well in 96-well plates. Plates were pre-coated with poly-L-ornithine (Sigma) and Laminin (Invitrogen) for primary astrocytes to facilitate cell attachment. A fifteen millimolar stock solution of TG (Sigma) was prepared by dissolving TG in 0.1 N NaOH. Cells were allowed to attach over night and TG or vehicle control was added to the appropriate wells. After five days of drug exposure, MTT was added and allowed to incubate for 3 hours. Afterwards, the media was carefully aspirated and the purple formazan crystals were dissolved in 100 ml DMSO (Fisher). Absorbance was measured at wavelength of 570/690 nm using a FLx800 fluorescence microplate reader (BioTek Instruments, Inc.).

Alkaline comet assay
Techniques used to perform the comet assay were adopted from the laboratory of Dr. Peter Mckinnon (St. Jude Children's Research Hospital) [45]. Briefly, 50,000 primary astrocyte or 30,000 human astroglial cells were seeded in Primaria 24-well plates (BD Biosciences) or in Falcon 24-well plates (BD Biosciences), respectively. Cells were allowed to attach overnight and then exposed to TG for three days. After three days of drug treatment, cells were harvested and counted to prepare a 3610 5 cells/ml suspension in PBS. The cell suspension was mixed with 1.2% ultra-pure low melting point agarose (Invitrogen) and casted onto chilled fully frosted glass slides (Fisher) pre-coated with 0.6% agarose (Bio-Rad). The cells were then lysed for 1.5 hours at 4uC, in lysis buffer (100 mM EDTA, 2.5 M NaCl, 10 mM Tris, 1.3% triton X-100 and 3.3% DMSO). The slides were washed twice with distilled water and incubated for 45 minutes at 4uC in electrophoresis buffer (1 mM EDTA, 50 mM NaOH and 1% DMSO). Electrophoresis was conducted at 12 V and ,90 mA at 4uC for 25 minutes. Slides were incubated for one hour in 400 mM Tri-HCl neutralization buffer followed by a 20 minute incubation in SYBR green I (Sigma). Images were captured using a Nikon eclipse TE300 fluorescent microscope. All experiments were replicated using two different sets of pooled cultures. The extent of DNA damage was expressed as the comet tail moment (the amount and distribution of DNA in the tail). The comet tail moment was measured in a minimum of 60 cells using Cometscore version 1.5 (Tri Tek corp).

Immunohistochemistry
Primary astrocytes were grown on poly-D-lysine chamber slides (BD Biocoat) for GFAP staining (to confirm astrocyte purity) or on glass coverslips and treated with 0 or 5 mM of TG for cH2AX staining. Cells were fixed with 4% PFA/PBS for 10 minutes, permeabilized with 0.5% Triton X-100/PBS for 5 minutes. Cells were then washed with PBS and blocked with 3% BSA/PBS for one hour. Cells were then immunostained with either anti-GFAP (1:500 in PBS, Sigma) for two hours or anti-cH2AX (1:1000 in 3%BSA, Milipore) for one hour. Subsequently, cells were washed with PBS and incubated for one hour with in Cy2-conjugated donkey antimouse secondary antibody (Jackson ImmunoResearch, 1:250 in PBS containing 2% donkey serum and 0.1% Triton X-100) for GFAP or with donkey anti-mouse secondary antibody labeled with Alexa 488 (In vitrogen, 1:800 in 3% BSA) for one hour. Finally, vectashield mounting media (Vectorlabs) containing propidium iodide was used to stain nuclei. GFAP slides were visualized and images were captured using Nikon eclipse TE300 fluorescent microscope. Images were captured for cH2AX stained slides using a LSM-710 confocal microscope (Zeiss). Images were analyzed for cH2AX foci by counting a minimum of 100 cells per genotype. Cells containing fewer than five foci were excluded to normalize for baseline staining observed in control treated cells. Cells with ''panstaining'' were considered to have more than 5 foci per cell. Hence, the percent of cells with cH2AX foci staining represents the percent of cells with greater than five foci per cell.

Statistical analysis
Differences in cytotoxicity and DNA damage among all three groups were compared using Kruskal-Wallis analysis of ranks. Mann-Whitney U test was used when only two groups were compared. Significance was declared when p,0.05. All statistical analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC).