Isolation and characterization of thymitaq (AG337) and 5-fluoro-2-deoxyuridylate-resistant mutants of human thymidylate synthase from ethyl methanesulfonate-exposed human sarcoma HT1080 cells.

Thymidylate synthase plays an essential role in the synthesis of DNA. Recently, several new and specific thymidylate synthase inhibitors that occupy the folate binding site, including Tomudex(R), BW1843U89, and Thymitaq, have demonstrated therapeutic activity in patients with advanced cancer. In order to find drug-resistant forms of human thymidylate synthase for gene therapy applications, human sarcoma HT1080 cells were exposed to ethyl methanesulfonate and Thymitaq selection. Thymitaq-resistant clonal derived sublines were established, and analysis indicated that both gene amplification and point mutations contributed to drug resistance. Eight mutant cDNAs that were identified from Thymitaq-resistant sublines were generated by site-directed mutagenesis and transfected into thymidylate synthase-negative cells. Only K47E, D49G, or G52S mutants retain enzyme activity. Moreover, cytotoxicity studies demonstrated that D49G and G52S transfected cells, besides displaying resistance to Thymitaq with IC50 values 40- and 12-fold greater than wild-type enzyme transfected cells, respectively, also lead to fluorodeoxyuridine resistance (26- and 97-fold in IC50 values, respectively) but not to Tomudex or BW1843U89. Characterization of the purified altered enzymes obtained from expression in Escherichia coli is consistent with the cell growth inhibition results. We postulate that the D49G or G52S mutation leads to the structural perturbation of the highly conserved Arg50 loop, decreasing the binding of thymidylate synthase to the inhibitors, Thymitaq and fluorodeoxyuridylate.

the reductive methylation of the substrate, 2Ј-deoxyuridine 5Ј-monophosphate (dUMP) by transfer of a methylene group from the cofactor, 5,10-methylene-5,6,7,8-tetrahydrofolate (CH 2 H 4 folate), to generate 2Ј-deoxythymidine 5Ј-monophosphate (dTMP) and 7,8-dihydrofolate. Human TS has been sequenced (2), purified (3,4), and crystallized (5). As an attractive target for anti-cancer drug design, since the 1950s, many TS analogues of both the substrate, dUMP, and the cofactor, CH 2 H 4 folate, have been synthesized and tested as potential anti-cancer therapeutics. Until recently, 5-fluorouracil and fluorodeoxyuridine (FdUrd) were the sole TS-targeted drugs approved for clinical application. In vivo, 5-fluorouracil and FdUrd are metabolized to 5-fluoro-2-deoxyuridylate (FdUMP), a compound that subsequently occupies the pyrimidine binding site forming a ternary complex with TS and the folate cofactor, resulting in inhibition of enzyme function. The recent determination of the three-dimensional structure of human TS has allowed the design of highly specific inhibitors, leading to the emergence of novel folate analogues, such as Tomudex (ZD1694), BW1843U89, and Thymitaq (AG337) (Fig. 1) (6). These promising compounds have entered clinical trials in recent years (7).
Previous studies have attempted to correlate enzyme structure and function using mutagenesis. To date, a large number of mutations have been made in Lactobacillus. casei and Escherichia coli TS (1,8,9). In most reports, the procedures included the following two steps: generated mutants were first screened for their enzyme activity by genetic complementation in a TS-deficient E. coli host in the absence of thymine, and kinetic characterizations in vitro were subsequently performed for functional mutants. A few mutants of human TS and their expressed enzymes in mammalian cells have been studied (10 -14). The Y33H human TS mutant, the only mutation in TS reported to be related to TS-directed drug resistance, was discovered in a human colon tumor cell line and conferred approximately a 3-4-fold resistance to FdUrd (10 -12). This mutation was reported to affect the catalytic properties of human TS enzyme, showing an 8-fold decrease in k cat but without significant change in the K m values for both dUMP and CH 2 H 4 folate between the mutant and wild-type TS.
As point mutations in human TS leading to the generation of antifolate-resistant genes have not yet been reported, an important goal of this work was to obtain human TS mutants conferring resistance to novel antifolates with minimal changes * This work was supported by USPHS Grant CA08010 from the National Institutes of Health. 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.
in the catalytic activity of the altered enzymes. Such mutants would be of much interest both for understanding structurefunction for human TS and for their potential applications in gene therapy by protecting hematopoietic progenitors from TS inhibitor toxicity.
In the present study, we generated TS mutants from human sarcoma HT1080 cells following ethyl methanesulfonate (EMS) exposure and Thymitaq selection. Clonal sublines containing these mutations were identified by DNA-SSCP and sequencing. These mutant cDNAs were cloned into mammalian expression vectors and transfected into mouse TS-negative cells for cytotoxicity assays. The relevant enzymes were expressed in E. coli purified and characterized by kinetic studies in vitro. Cell Lines and Culture Conditions-Human fibrosarcoma HT1080 cells were obtained from the American Type Culture Collection (Rockville, MD). Stock cultures of the parental cell line HT1080 and resistant sublines were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 g/ml streptomycin, and 100 units/ml penicillin. The TS-negative cell line FSthy21, a kind gift of Dr. T. Seno (9), was originally established from mouse FM3A cells. FSthy21 cells were grown in Eagle's minimum essential medium supplemented with 10% dialyzed fetal bovine serum, 1 M 5-CHOtetrahydrofolate, and 10 M thymidine. Prototrophic transformant clones, derived from FSthy21 cells by transfection of wild-type or mutant human TS cDNAs, were cultured in the same medium as mentioned above without thymidine and reduced folate supplements.

Materials-Ethyl
Chemical Mutagenesis and Drug Selection-In order to determine the optimum concentration of EMS and Thymitaq for random mutagenesis experiments, we first carried out the cytotoxicity assay of human fibrosarcoma HT1080 cells with EMS or Thymitaq. IC 50 values and the minimal concentration of EMS or Thymitaq that resulted in no colony formation in HT1080 cells were obtained by plots of colonies surviving versus various EMS or Thymitaq concentrations.
HT1080 cells (50 ml, total 4 ϫ 10 8 cells) growing in logarithmic phase were exposed to EMS (400 g/ml) for 18 h. The cells were washed and then incubated for an additional 3 days in EMS-free medium to allow phenotypic expression. EMS-treated cells were subcultured at 6 ϫ 10 7 cells/100-mm dishes in 15 ml of medium and then grown in the presence of 40 M Thymitaq for 14 days. Surviving clones obtained from EMS exposure and Thymitaq selection were isolated with a ring cylinder and expanded into stable resistant sublines. Control cells were treated with Thymitaq without EMS exposure (16,17).
Single-stranded Conformation Polymorphism (SSCP) Analysis-RNA from the HT1080 cells and resistant sublines was isolated, and first-strand cDNAs were synthesized by reverse transcriptase-PCR. For DNA-SSCP analysis, TS fragments (150 -260 bp) were obtained by PCR amplification using the cDNAs as the template and 6 pairs of TSspecific primers (the sequences of oligonucleotide primers used in various experiments such as PCR amplification, DNA-SSCP, and sequence analysis are described in Table I). The reaction mixture containing 1 Ci of [␣-32 P]dCTP and a small volume (3 l) of final PCR products was subsequently mixed with 10 l of loading buffer containing 96% formamide. Samples were denatured at 94°C for 3 min and chilled on ice for at least 5 min, and 2 l was loaded onto a 6 -8% non-denaturing polyacrylamide gel with 10% glycerol. Gels were electrophoresed at 10 watts for 6 -8 h at 4°C, using 0.5ϫ Tris borate/EDTA buffer. The separated single strand DNA fragments were visualized by autoradiography.
An alternative method (nonisotopic SSCP) was also utilized in which the single strand DNA bands are detected by ethidium bromide staining instead of autoradiography. At least 40 ng (20 l) of amplified DNA was denatured by addition of 1 g of 0.5 M NaOH, 10 mM EDTA at 42°C for 5 min. Before loading, 1 l of formamide containing 0.5% bromphenol blue and 0.5% xylene cyanol were added. Non-denaturing gels (1.5 mm thick, 6 -8% polyacrylamide) with 5% glycerol were made in a standard vertical gel apparatus. Gels, using 0.5ϫ TBE as running buffer, were electrophoresed at 15 V/cm for 4 h, and the temperature was maintained at 4°C by circulating cold water. Finally, SSCP gels were neutralized and stained in 0.5ϫ TBE containing 0.5 g/ml ethidium bromide for visualization of band.
DNA Sequence Analysis for Mutant Detection-The reverse transcriptase-PCR fragments exhibiting abnormal mobility on SSCP gels were subsequently subcloned into pCR-Script TM SK(ϩ) vector, which permits the efficient cloning of PCR fragments with a high yield and sequencing by two commercially designed primers (M3 (Ϫ20) and M13 reverse). Sequence analysis was performed by the dideoxy chain termination method with ␣-35 S-dATP using modified T7 DNA polymerase according to the manufacture's instructions (Amersham Pharmacia Biotech).
Southern, Northern, and Western Blot Analysis-The human TS probe, which was 32 P-labeled by random primer DNA labeling kit (Boehringer Mannheim), was a 950-bp gel-purified fragment of human TS cDNA cleaved from the pET-17ϫb(hTS) plasmid with NdeI and HindIII restriction enzymes (15). As a loading control, a ␥-actin cDNA probe was used for Southern blotting and ribosomal phosphoprotein 36B4 was for Northern blotting.
To detect the expressed levels of human TS protein, Western blots were probed with a rabbit anti-human TS polyclonal antibody (a generous gift of Dr. Frank Maley) and incubated with goat anti-rabbit IgG antibody. Finally, the protein bands were visualized by the ECL method using a kit from Amersham Pharmacia Biotech.
Site-directed Mutagenesis-The vector pcDNA3 under T7 promoter control was used to obtain a mammalian expression system for human TS. The construction of the plasmid containing recombinant human TS cDNA was performed by digestion of pET-17ϫb(hTS) with NdeI and HindIII to generate a 950-bp fragment containing a full cDNA sequence of human TS. After treatment with T4 DNA polymerase to create blunt ends, the 950-bp DNA fragment was ligated to the unique EcoRV site of pcDNA3 vector to produce the pcDNA3-hTS plasmid. The correct size (6.4 kilobase pairs) and orientation was confirmed by restriction mapping and sequencing.
The Transformer TM site-directed mutagenesis kit was used to obtain point mutations in human TS. The mutagenic and selective primers used for site-directed mutagenesis experiments are described in Table  II. A selection primer that contains a unique KspI restriction site instead of the unique SmaI restriction site on the pcDNA3 vector is shown last, and the other 8 primers were designed to generate mutants of the human TS gene at the targeted site with 1 to 3 nucleotide changes. Mutant human TS cDNA was obtained by annealing one mutagenic primer and one selection primer to the single strand pcDNA3-hTS plasmid. The experimental procedures followed the manufacturer's instructions. The resulting colonies were initially screened using KspI digestion that only cuts newly synthesized plasmids. Plasmids with correct point mutations in human TS were examined further by restriction mapping and DNA sequencing.

Transfection of Mouse TS-negative Cells with Wild-type and Mutant
Human TS cDNAs-The mouse TS-negative FSthy21 cells were used as host for DNA transfections that were performed using a DOTAP transfection kit (Boehringer Mannheim). Three days after transfection, cells were placed in selective media, which lacked thymidine and reduced folate. Surviving cells having the ability to grow in the absence of thymidine were cloned in soft agarose.
Growth Inhibition Assay for Transfected Cells-Cloned logarithmically growing suspension mouse TS-negative FSthy21 cells transfected with either wild-type or variant human TS cDNA were seeded in 96well plates at 2,000 cells/well in 180 l of complete medium. Two hours later, drug (Tomudex, Thymitaq, BW1843U89, or FdUrd) was added, and the cells were grown in drug-containing medium for an additional 7 days. Cell viability was measured by the Alamar Blue TM assay. To the above 96-well cultured cells, 25 l of Alamar Blue (10% of incubation volume) was added according to the manufacturer's instructions. The 96-well plates were then incubated at 37°C for 4 h. Viable cells induce chemical reduction of the media which results in a change in color from blue to red. The intensity of red color (and fluorescence) is proportional to the number of viable cells. After incubation, fluorescence was read at 530 -560 nm excitation wavelength and 590 nm emission wavelength by an automated plate reader (model EL340; Bio-Tek). Drug concentrations needed to reduce cell growth by 50% (IC 50 values) were determined graphically by plotting cell growth versus inhibitor concentrations.
TS Protein Expression Vector Construction and Purification-Human TS variants (G52S and D49G) selected for enzyme kinetic characterization were recloned into the protein expression vector pET-17ϫb. DNA fragments carrying the entire mutant human TS cDNA were amplified using the corresponding pcDNA3-hTS vector as the template and two primers designed to contain appropriate restriction site for cloning. The 5Ј-primer (5Ј GGAATTCTGCAGCATATGCTTGTTGC 3Ј) contains a created NdeI restriction site, and the 3Ј-primer (5Ј CTAGAT-GCATGCTCGAGCGGCCGCC 3Ј) has an XhoI site. After PCR amplification, the reaction mixture was digested with NdeI and XhoI enzymes, and the restricted DNA fragment (950 bp) was inserted into the corresponding sites of pET-17b vector. The correct construction of mutant pET-17b(hTS) was verified by restriction mapping and sequence analysis.
The protein expression pET-17ϫb(hTS) plasmid (wild-type or mutant TS) was used to transform E. coli strain BL21(DE3). Bacterial cells were grown at 30°C in 1 liter of tryptone phosphate medium (2% bacto-tryptone, 1.5% yeast extract, 0.2% sodium phosphate (dibasic), 0.1% potassium phosphate (monobasic), 0.8% sodium chloride, and 0.2% glucose) supplemented with 100 g/ml ampicillin (18). The wildtype and the two mutant TSs were induced with 1 mM isopropyl-␤thiogalactose for 5 h at which point the enzyme was induced to about 10 -15% of the soluble protein. The purification procedures were basically carried out as described by Ciesla et al. (19) for rat TS purification, which included streptomycin treatment of crude extract, followed by ammonium sulfate precipitation, ion-exchange cellulose DE-52, and phenyl-Sepharose CL-4B chromatography. A minor modification was made by eluting TS protein from a phenyl-Sepharose CL-4B column with a decreasing linear gradient of ammonium sulfate from 0.8 mM to 0 instead of 0.8 to 0.4 mM. Finally, TS fractions were precipitated with solid ammonium sulfate and stored at Ϫ80°C until use. TS purity was demonstrated by 12% SDS-PAGE.
Michaelis constants (K m ) for CH 2 H 4 folate and dUMP were determined from initial velocity measurements, which were obtained by measuring the change in A 340 with a Shimadzu UV-2101 PC spectrophotometer. For determination of K m of CH 2 H 4 folate, the concentration of dUMP was fixed at 500 M, whereas CH 2 H 4 folate concentrations were varied between 5 and 600 M. For determination of K m of dUMP, CH 2 H 4 folate was present at a concentration of 600 M, and dUMP was varied between 1 and 300 M. Steady-state kinetic parameters were subsequently obtained by a nonlinear least squares fit of the data to the Michaelis-Menten equation using a computer program. k cat (s Ϫ1 ) values were obtained by dividing V max (mol/min/mg protein) by the estimated concentration of enzyme (mol) used in the reaction.
Inhibition constants (K i ) were determined from the steady-state inhibition reaction rates for mixtures of enzyme, dUMP, CH 2 H 4 folate, and inhibitor. A high fixed CH 2 H 4 folate concentration (600 -800 M) and variable antifolate concentrations were used to measure the inhibition produced by Tomudex, Thymitaq, and BW1843U89. A constant dUMP concentration (500 M) was used, and FdUMP concentrations were varied to measure inhibition by FdUMP.

RESULTS
Random Mutagenesis-Based on EMS and Thymitaq cytotoxicity assays, a concentration of 400 g/ml EMS that resulted in 80% inhibition of colony formation and 40 M Thymitaq, approximately 20-fold higher than IC 50 value for this compound, were chosen for selection of resistant clones. Following EMS exposure and Thymitaq selection, a relatively large number of resistant clones (41 per 4 ϫ 10 8 cells) were generated in comparison with no EMS pretreatment but selected in Thymitaq (1 per 10 8 cells), indicating that EMS increased the frequency of surviving Thymitaq-resistant clones by approximately 10-fold. All 41 of these clones from EMS and Thymitaq treatment and 1 clone from a control experiment without EMS were expanded to establish stable Thymitaq-resistant cell lines, which were grown in the presence of 40 M Thymitaq.
DNA-SSCP Analysis-To detect putative point mutations leading to drug resistance, all 41 EMS-exposed Thymitaq-re- sistant cell lines and one control-resistant cell line (HT1080-A, without EMS treatment) were investigated by SSCP analysis. After total RNA isolation and reverse transcription, fragments that include most of the entire coding sequence of human TS gene except nucleotides 1-69 and 915-939 were amplified by 6 pairs of primers. DNA-SSCP screening demonstrated that shifted bands in addition to normal migrating bands were observed on SSCP gels for nine EMS-exposed Thymitaq-resistant cell lines (HT1080 -1b, -2b, -1c, -2c, -1d, -2d, -1e, -2e, and -6e), indicating the possible presence of mutations in these fragments (Fig. 2).
The Expressed Levels of Human TS in Resistant Cells-Western blotting revealed that some resistant sublines (HT1080-A, -1b, -2b, -2c, -2d, -2e, and -6e) demonstrated elevation of protein levels, and modest increases in TS genomic DNA were also seen in these sublines by Southern blotting. The corresponding mRNA levels were observed to increase in only HT1080 -2c, -2d, and -2e sublines as determined by Northern blotting, using ribosomal phosphoprotein 36B4 mRNA as controls. In contrast, some sublines (HT1080 -1c, -1d, and -1e) did not amplify the TS gene and nor did they overexpress TS mRNA and protein (Fig. 3).
Mutations Identified by Sequencing Analysis-The fragments produced by PCR amplifications with abnormal migration on SSCP gels were cloned into pCR-Script vector for sequencing. More than 20 mutations including some silent mutations in the 9 Thymitaq-resistant cell lines were identified, which are Q214R, A228T, and K266I mutations from HT1080-1b; I40T and D49N from HT1080-2b; T51A from HT1080-1d; R25H and F59L from HT1080-2d; G52S from HT1080-1e; D49G, D130G, and T234M from HT1080-2e; and K47E, R50C, Y65C, V79M, and V84A from HT1080-6e (the G52S mutation as an example is shown in Fig. 4). The results combining Western and SSCP analyses indicated that human TS protein overexpression and mutations in the TS coding region were both present in the same cell lines. No mutations were detected in the untreated HT1080 cell line or in the HT1080-A line that was obtained by Thymitaq selection without EMS exposure.

Construction of Mutations in Human TS by Site-directed Mutagenesis-
To determine whether only single point mutations result in changes in Thymitaq binding to TS, we expressed mutant TS genes with only single point mutations. The human TS expression vector pcDNA3-hTS, which contains the entire coding sequence of the human TS gene with minor modifications of the N-terminal nucleotide codon, was utilized for these studies (15). Eight TS mutants, including K47E, D49N, D49G, R50C, T51A, G52S, F59L, and Q214R detected in the resistant subline analysis, were generated by site-directed mutagenesis using pcDNA3-hTS as a template. These mutations were selected from the 20 or more mutations found, based on their occurrence in highly conserved regions of this enzyme (Table III). Sequencing the coding regions demonstrated that the expected substitution had been introduced and that no other alteration had occurred (data not shown).
Rescue of TS-negative Cells by Transfection of Various Human TS cDNAs-Transfection of the eight human TS variants

FIG. 2. Autoradiogram of SSCP analysis of human TS gene mutations in AG337-resistant cells.
Amplified DNA fragments by reverse transcriptase-PCR corresponding to region A were denatured by heating, and electrophoresis was performed in 8% polyacrylamide gel containing 5% glycerol at constant 30 watts at 4°C. Lane 1, control HT1080 cells; lanes 2-5, AG337-resistant cells. Fragments with a mobility shift in addition to wild-type bands are observed in lanes 4 and 5, suggesting mutations in human TS gene from nucleotides 70 to 187 for HT1080/1b and HT1080/2b. Both were hybridized with a [ 32 P]dCTP-labeled human TS cDNA. C, a control for RNA loading, using ribosomal phosphoprotein 36B4 mRNA. Northern blot membrane was stripped and rehybridized with a [ 32 P]dCTP-labeled cDNA probe for the 36B4 mRNA. D, Western blot analysis for human TS proteins, which were probed by a rabbit polyclonal antibody against human TS. Lane 1, HT1080 cells. Lane 2, resistant cell line HT1080-A, which was selected by AG337 without EMS pretreatment. Lanes 3-11 represent nine different resistant sublines from EMS treatment, followed by AG337 selection. as generated above into TS-negative mouse cells was used to demonstrate whether or not the altered TS protein had sufficient catalytic activity to permit normal growth in the absence of thymidine. Mouse TS-negative cells (FSthy21) were transfected with a plasmid (pcDNA3-hTS) encoding wild-type or each of the mutant human TS enzymes by standard DOTAP transfection procedures. The results of these experiments showed that three TS variants (K47E, D49G, and G52S) as well as wild-type TS were able to complement growth of TS-negative cells in selective medium lacking thymidine. In contrast, cells transfected with the other five human TS variants (D49N, R50C, T51A, F59L, and Q214R) did not survive (data not shown). For each surviving transfectant, at least 12 individual clones were isolated. Those clones were then expanded to cell lines by growth in normal media.
Growth Sensitivity to Antifolates and FdUrd-Before determining growth sensitivity to Tomudex, Thymitaq, BW1843U89, and FdUrd, the levels of human TS protein for the transfectants were measured by Western blot analyses. For each of the human TS wild-type or variants, three cell clones were randomly selected. The three mutant clones (K47E-2, D49G-1, and G52S-3) expressed levels of TS protein similar to that of a control clone (wt-1) transfected with the wild-type vector (data not shown). These four clones were chosen for determination of cytotoxicity of drugs as described below.
The effect of the three antifolates and FdUrd on the growth of transfectants that stably express various TS proteins was evaluated by cytotoxicity studies. Cell growth was measured by the Alamar Blue assay. The IC 50 values of Tomudex, Thymitaq, BW1843U89, and FdUrd are presented in Table IV. D49G and G52S transfectants conferred resistance to Thymitaq (40-and 12-fold, respectively) and to FdUrd (26-and 97-fold, respectively) as compared with the wild type transfected cell line. In contrast, no marked changes were observed for Tomudex and BW1843U89 cytotoxicity compared with the wild type transfected cell line.
To determine whether different levels of TS expression in the transfectants contributed to drug sensitivity, multiple clones of wild-type transfections were analyzed for TS levels and only slight changes of drug sensitivity were observed in cytotoxicity studies (data not shown). These results suggested that the changes in drug sensitivity of different clones transfected with the same vector are roughly proportional to their expressed TS protein levels. Thus a major reason for the observed changes in IC 50 values may be attributed to the mutation in TS rather than to its level of expression.
Subcloning and Expression of Human TS in E. coli-Based on the drug sensitivity results presented above, TS mutants D49G and G52S were selected for enzyme kinetic studies. The D49G and G52S human TS cDNA were inserted into protein expression vectors pET-17ϫb. The pET system provided high yields of soluble protein in a derivative of the E. coli strain BL21(DE3). The activity of the enzyme was monitored and found to be highest 5 h after the addition of isopropyl-␤-thiogalactose. E. coli extracts were analyzed by SDS-PAGE and revealed an intensely staining band at a molecular mass of about 36 kDa, absent from extracts of the host E. coli cells. This new protein band was estimated to represent about 10 -20% of the total soluble protein in the extract. The crude extracts from bacterial cells transformed by mutant TS vectors had similar high levels of the altered proteins, comparable to wild-type enzyme.
Purification of Mutant TS Proteins-A procedure previously used to purify rat TS, using sequential ion-exchange/phenyl-Sepharose chromatography, was adopted for purification of wild-type and mutant human TS proteins (19), modified in that the human TS was eluted from phenyl-Sepharose using a linear gradient of ammonium sulfate of 0.8 M to 0 instead of 0.8 to 0.4 M employed for rat proteins, which is similar to that described in a recently published paper (15). After purification, a single major component on SDS-PAGE gel migrating with an apparent molecular weight of human TS protein was observed for the wild-type and mutant enzymes. Purity was estimated to be greater than 80% as determined by densitometric scanning.
Kinetic Properties of Mutant Enzymes-To obtain information about the catalytic and ligand-binding properties of these TS variants, the kinetic parameters V max and K m values for substrate and cofactor and K i values for inhibitors were evaluated. The K m values for CH 2 H 4 folate and dUMP were not significantly different between the wild-type and G52S mutant forms, whereas the catalytic efficiency (k cat ) of G52S was even higher than the k cat of wild-type TS ( Table V). The K m for dUMP and CH 2 H 4 folate of the D49G variant was increased 3-fold over wild-type TS, and D49G mutants showed diminished catalytic activity with k cat values 3-fold lower than wildtype TS.
The K i values of wild-type and mutant TS proteins for Tomudex, BW1843U89, and Thymitaq were determined at a high concentration (over 8-fold K m of TS variants) of CH 2 H 4 folate. The K i values of FdUMP were determined at a fixed concentration (500 M) of dUMP. As expected, the K i value for FdUMP of the G52S mutant was 20-fold greater than the K i of wild-type TS, consistent with cytotoxicity results. The K i value of the D49G mutant for FdUMP was only 5.4-fold higher than the wild-type. By comparison, inhibition of G52S and D49G mutants by the folate inhibitors of TS was 3-6-fold less effective than against wild-type TS (see Table V).
In addition, to address the question about possible inhibitory effects of the unnatural stereoisomer of CH 2 H 4 folate upon the kinetic analyses of the enzyme, we prepared the natural isomer (6R)-l-CH 2 H 4 folate following the procedure described by Bruice and Santi (22). No significant difference between the K m values of CH 2 H 4 folate for wild-type and mutant TS enzyme was observed using a racemic mixture or the pure natural isomer of cofactor (data not shown), indicating that the unnatural isomer does not inhibit significantly this enzyme activity, as was previously demonstrated for mouse TS (23).

DISCUSSION
Strategies-As expected, random mutagenesis by exposure of human sarcoma HT1080 cells to an alkylating agent (EMS) and selection with Thymitaq resulted in the generation of a large number of resistant colonies, as compared with Thymitaq selection without EMS pretreatment. Thymitaq was employed as a selective drug in as much as Thymitaq is not a substrate for folylpolyglutamate synthetase and enters cells by passive diffusion, thus eliminating resistance mechanisms such as an altered reduced folate transporter and a decrease in folylpolyglutamate synthetase activity. Therefore, the expected major causes of resistance in the Thymitaq-resistant clones are overexpression and/or an alteration of the target human TS enzyme. The advantage of this approach is that mutations caused by EMS could occur anywhere in the entire TS gene and may not be limited to specific regions of the gene as in cassette or site-directed mutagenesis. However, as it is almost impossible to obtain more than one mutation in a nucleotide codon by random mutagenesis, some desirable amino acid substitutions are excluded by this approach.
Human TS Mutants Identified in EMS-exposed Cells-The presence of point mutations in EMS-exposed Thymitaq-resistant cells was determined by SSCP analysis using 6 pairs of primers to span the cDNA for TS; each amplified fragment was 150 -260 bp, providing an 80 -90% range of sensitivity (24). By screening most of the entire coding sequence of human TS gene from Thymitaq-resistant HT1080 sublines, shifted bands in addition to normal migrating bands were observed on SSCP gels, indicating the presence of wild-type and mutant TS genes in the sample. Previous studies demonstrated that polymorphisms could be detected when mutant DNA comprised as little as 3% of the total gene copies in a PCR mixture (25). However, SSCP cannot discriminate between pre-existing mutations and those mutations introduced by Taq polymerase errors during early-stage amplification.
TS fragments showing abnormal migration on SSCP gels were sequenced, and more than 20 mutations of human TS were identified in nine Thymitaq-resistant sublines, indicating that some cell lines contain more than one mutation in TS. This phenomenon was observed in other EMS-treated cell lines (17). We assumed that some or most of those mutations were generated by EMS exposure, based on following evidence: 1) none of the mutations were detected in parent HT1080 and EMSuntreated HT1080-A cells; and 2) most amino acid substitutions were found in highly conserved positions (Table III), suggesting that these variants likely led to drug resistance. We

Drug-resistant Mutants of Human Thymidylate Synthase
chose eight mutations from 20 or more found for further investigation. Based on knowledge of the crystal structure of the enzyme, F59L and Q214R mutations were chosen since these amino acids are important for ligand binding and structural stability. Additionally, six other mutations (K47E, D49N, D49G, R50C, T51A, and G52S) located in the Arg 50 loop were chosen, as this loop becomes more ordered by movement and reorientation upon ligand binding (26). These eight TS mutants were expressed in mouse TS-negative cells to examine directly the ability of these mutations to allow growth in the absence of thymidine and confer resistance to Thymitaq. Three of these cDNAs allowed growth in the absence of thymidine, and cytotoxicity studies showed that D49G and G52S TS variants displayed resistance to the selective drug Thymitaq, providing additional proof that these mutations were involved in the Thymitaq-resistant phenotype.
The Arg 50 Loop and Drug Resistance-Of interest, of the 20 point mutations identified from random mutagenesis, six occurred in the highly conserved Arg 50 loop (amino acids 47-52). This loop connects elements of protein secondary structure, an ␣-helix A (residues 30 -43) near the N terminus, and a ␤-sheet i (residues 54 -66). Comparison of nearly identical crystal structures between the native unbound and dUMP-bound TS revealed that the only difference is that the mobile Arg 50 loop has less than 1.0-Å movement and undergoes reorientation upon Arg 50 binding to the phosphate moiety of dUMP, steps necessary to accept the incoming folate molecule. Once the ternary complex is formed, the carboxylate of the C-terminal residue and N-1 of CH 2 H 4 folate form hydrogen-bond networks with Arg 50 through fixed H 2 O molecules. This flexible Arg 50 residue seems to be a bridge linking the enzyme C terminus, substrate, and cofactor (or antifolates) together. The movement of the Arg 50 residue is accompanied by adjustment and reorientation of its neighbor residues, indicating that the entire Arg 50 loop undergoes relocation and encompasses new interactions. For example, the hydrophobic atoms of Thr 51 has contacts with the buried Val 313 side chain after movement (6,24,(27)(28)(29)(30).
Besides structural studies, the residues in the conserved Arg 50 loop have been intensively studied by mutagenesis of E. coli and L. casei TS (1). The corresponding amino acids representing Arg 50 , Asp 49 , and Thr 51 in these proteins are quite sensitive to substitutions by other amino acids. Surprisingly, contrary to these neighboring residues, Gly 52 , which undergoes an apparent reorientation upon the formation of the ternary TS complex, accepts any mutations in E. coli without loss of activity (31). The R50C mutant of human TS is catalytically inactive, and both T51A and D49N replacements are not tolerated, consistent with comparably altered L. casei and E. coli TS. However, three other point mutations (K47E, D49G, and G52S) retain TS catalytic function. Cytotoxicity assays also showed that expression of D49G and G52S human mutant protein in mouse TS-negative cells confers resistance to Thymitaq, with IC 50 values 40-and 12-fold greater than cells expressing wildtype TS. These mutant transfected cell lines also display resistance to FdUrd (26-and 97-fold, respectively) but interestingly not to Tomudex or BW1843U89. These data indicate that Thymitaq binds differently to TS than Tomudex and BW1843U89. The fact that the K47E mutant did not show a difference in binding to the four drugs tested, as compared with the wild-type TS, may be related to its position, which is relatively distant from Arg 50 .
Based on the above results and structural information of the Arg 50 loop, we postulate that the basis for reduced Thymitaq and FdUMP binding observed for D49S and G52S mutants results from impaired movement of the Arg 50 loop and result-ing interaction of the loop with nucleotide or folate molecule. This is especially true for the Arg 50 residue which is involved in a hydrogen bond network with folate, dUMP, and enzyme C terminus. More drastic changes such as those in mutants R50C, D49N, and T51A result in inactive TS enzymes, whereas small structural perturbations of Arg 50 loop caused by mutants K47E, D49G, and G52S may be compensated for by the local adjustment of neighboring residues, which still maintain contacts with ligands in the new position. The modification of the loop Arg 50 causes changes in binding affinity leading to drug resistance, related not only to folate but also to nucleotide binding.
Kinetic Studies of D49G and G52S Human TS Mutants-The kinetic characterization studies of the highly purified altered enzymes obtained from an E. coli expression system are consistent with the cell growth inhibition results. The product of K i ⅐k cat /K m has been suggested as a comparative index to represent both catalytic efficiency and drug inhibition (32). The values of K i ⅐k cat /K m for wild type and G52S via FdUMP are 0.34 and 8.3 (s Ϫ1 ϫ 10 3 ), respectively, which correlates with the resistance displayed by the G52S TS transfectants to FdUrd over the wild-type TS transfectants in culture.
Future Applications in Gene Therapy-Treatment with intense dosages of chemotherapeutic agents, including fluorodeoxyuridine and antifolates, may increase the curability of sensitive tumors (33). However, this approach is often limited due to myelosuppression. To overcome bone marrow toxicity from chemotherapy, transfer of drug resistance genes into hematopoietic progenitors is a promising approach. Introduction of drug resistance genes such as mutant dihydrofolate reductase, the multiple drug resistance gene (MDR-1), glutathione transferase, alkyl transferase, cytidine deaminase, and aldehyde dehydrogenase into murine hematopoietic cells has been shown to improve chemotherapy tolerance in vitro and in vivo (34 -38). The newly identified human TS mutants (D49G and G52S), because of their desirable properties that include catalytic function (k cat ) and resistance to FdUrd and Thymitaq, are excellent candidates for gene transfer studies. Preliminary transfection experiments of mutant human TS cDNA retroviral constructs have demonstrated that the G52S mutant functions in a dominant manner to protect murine bone marrow cells from FdUrd toxicity as compared with wild-type TS. 2