Creation and Characterization of 5-Fluorodeoxyuridine-resistant Arg⁵⁰ Loop Mutants of Human Thymidylate Synthase¹

TS³(EC 2.1.1.45) is an essential S-phase enzyme that catalyzes the formation of dTMP from dUMP. It provides the only de novosource of dTMP and is therefore indispensable for cell division. Inactivation of TS results in decreased production of TMP, cessation of DNA synthesis, and ultimately thymineless death (1). For these reasons, TS has been a key target for the design and use of chemotherapeutic agents (2, 3, 4). Fluoropyrimidine-based analogues, such as 5-FU and 5-FdUR, are used extensively for the treatment of colonic, breast, and ovarian carcinomas. Intracellularly,the fluoropyrimidines are metabolized to 5-FdUMP, which forms a stable inhibitory complex with thymidylate synthase and the cosubstrate CH₂H₄-folate.

Several 5-FdUR and antifolate-resistant TS enzymes have been created in the last 2 years (5, 6, 7, 8, 9). Recently, three novel human drug resistant TS mutants (D49G, G52S, and K47E) were identified via EMS mutagenesis of human sarcoma HT1080 cells (5). The mutations are located in the highly conserved Arg⁵⁰-loop, a loop that forms a bridge linking the enzyme COOH-terminus, cofactor, and cosubstrate together and undergoes reorientation upon binding the phosphate moiety of dUMP to accept the incoming folate molecule (10, 11, 12). The resultant enzymes demonstrate a high degree of resistance to both 5-FdUR and the antifolate Thymitaq (AG337).

In the present study, we generated human TS mutants of the Arg⁵⁰-loop and residue Tyr 33 through random mutagenesis. These mutants were selected for their ability to confer growth of E. coli in the presence of 5-FdUR. Those able to survive at the highest dosages were expressed in E. coli,purified, and characterized by kinetic studies in vitro. Our results indicate that the Arg⁵⁰-loop is highly mutable, and many of the mutants are highly resistant to 5-FdUR.

The sensitivity of normal human cells, particularly bone marrow cells,is a major limiting factor in chemotherapy by 5-FU. Gene therapy holds promise in boosting the resistance of these cells to the cytotoxic effects of chemotherapeutic agents. The potential for bone marrow protection has been strengthened by gene transfer experiments in relevant mammalian cells, such as hematopoietic stem cells, and, in some cases, by success in animals (13, 14). Our goal is to create novel TS enzymes for ex vivo gene therapy that could reduce the myelosuppression observed after treatment with 5-FU.

CH₂H₄-folate was obtained as a racemic mixture from Schircks Labs (Jona, Switzerland). ABI Prism Dye Terminator Cycle Sequencing kits for fluorescent sequencing were the products of Perkin-Elmer (Branchburg, NJ). E. coli DNA Pol I was from New England Biolabs (Beverly, MA). Pfu DNA Polymerase was from Stratagene (La Jolla, CA). Plasmid DNA was isolated using the Maxiprep and Miniprep kits from Qiagen (Chatsworth, CA) or the PERFECTprep Plasmid DNA kit from 5 Prime-3 Prime, Inc. (Boulder,CO). 5-FdUR, 5-FdUMP, dUMP, N-tris[hydroxymethyl]methyl-2-aminoethane-sulfonic acid,and all of the other reagents were from Sigma Chemical Co. 6-[³H]FdUMP (22.0 Ci/mmol) was obtained from Moravek Biochemicals (Brea, CA). Protein assay dye reagent concentrate was from Bio-Rad (Hercules, CA). E. coli NM522 (Stratagene,La Jolla, CA) was used for cloning and library construction. E. coli χ2913recA (ΔthyA572, recA56),kindly provided by Dr. Daniel Santi (University of California, San Francisco, CA), is tetracycline resistant and was used in all of the complementation studies and in the purification of plasmid-encoded TS. Unless otherwise stated, all of the DNA oligodeoxyribonucleotides were from Operon Technologies (Alameda, CA).

Plasmid pGCHTS-TAA, from Dr. Daniel Santi, contains the wild-type TS cDNA in a pUC vector background and has been described previously (10, 15). A nonfunctional Arg⁵⁰-TS stuffer vector that spans the residues targeted for randomization was created by replacing the TS open reading frame between nucleotides 94 (NcoI site) and 183(SphI site) with a 920-bp fragment derived from the pET34-LIC vector (Novagen, Madison, WI). The DNA insert was prepared by digestion of pET34-LIC with NcoI and SphI,purified using the Qiaquick PCR spin kit (Qiagen, Chatsworth, CA), and ligated into the NcoI-SphI-digested TS vector.

The TS random library was constructed by annealing two single-stranded DNA oligodeoxyribonucleotides, both containing randomized nucleotide segments. Random oligomer TS-R⁵⁰-Nco is a 67-mer that corresponds to the sense nucleotides 71–137 and contains a NcoI site (nucleotide 94) for cloning. It is 5′-d(GCGTCCGCCCCATGGTGAACTGCAGTACCTGGGGCAGATCCAACACATCCTCCGCTGCGGCGTCAGG)-3′,and it contains partially random nucleotides (underlined) corresponding to residue 33 of TS. Oligomer TS-R⁵⁰-Sph is a 67-mer that corresponds to the antisense nucleotides 126–192 and contains a SphI site (nucleotide 183) for cloning. Its sequence is 5′-d(AGCGCGCCTGCATGCCGAATACCGACAGGGTGCCGGTGCCCGTGCGGTCGTCCTTCCTGACGCCGCA)-3′,and it contains partially randomized nucleotides (underlined)corresponding to the Arg⁵⁰-loop residues 47–52. All of the partially randomized nucleotides were designed to contain 80% of the wild-type base and 20% of the remaining three nucleotides. These oligodeoxyribonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA). Oligomers TS-R50-Nde and TS-R50-Sph were annealed in 50 μl of 200 mm Tris-HCl (pH 7.5), 100 mm MgCl₂, 250 mm NaCl by incubation at 80°C for 5 min,followed by 55°C for 15 min, at 37°C for 15 min, and at room temperature for 15 min. The partial oligonucleotide duplex was extended in a 40-μl reaction mixture that contained 10 mm Tris-HCl (pH 7.5), 5 mmMgCl₂, 7.5 mm DTT, 250μ m deoxynucleotide triphosphates, and 5 units Klenow fragment of E. coli DNA Pol I for 2 h at 37°C. The extended DNA was purified using the Qiaquick PCR Purification Kit,digested with NcoI (New England Biolabs) and SphI(New England Biolabs), and purified by phenol extraction and ethanol precipitation.

The purified partially random oligonucleotides were used as inserts for construction of the human TS plasmid library. The nonfunctional “stuffer” insert was removed by digestion with SphI and NcoI, and the resulting 3.6-kb fragment was ligated with a 5:1 molar excess of the 122-bp restricted random insert using T4 DNA ligase (Life Technologies, Inc.). The ligation mixture was directly transformed (Bio-Rad Genepulser; 2 kV, 25 microfarads, 400 ohm) into fresh electrocompetent NM522 cells(Stratagene) in 20 separate transformations using 2 μl of the ligation mixture and 100 μl NM522 E. coli. After combining the transformation reactions, the size of the library that contained the TS plasmid was determined by plating an aliquot transformation mixture on media that contained carbenicillin (50 μg/ml; Island Scientific, Bainbridge Island, WA). The remainder of the library was amplified by growing the transformed NM522 cells overnight in 2 × YT media in the presence of carbenicillin,and the plasmid was harvested. Electrocompetent χ2913(TS⁻) cells were transformed with the plasmid library, pooled, plated to confirm adequate transformation efficiency,grown overnight in 50 μg/ml carbenicillin, 12.5 μg/ml tetracyline,and 50 μg/ml thymidine, and stored in aliquots at −80°C in 10%glycerol. The extent of randomization was verified by sequencing plasmid DNA isolated from 24 clones of NM522 E. coli cells that contained the random library. DNA was isolated using a Qiagen Miniprep kit, and sequencing reactions were conducted with the ABI Prism Dye Terminator Cycle Sequencing Kit with AmpliTaq DNA polymerase using the sequencing primer DLPCRTS3R,5′-d(AAAAAAAAACCATGTCTCCGGATCTCTGGTAC)-3′.

For selection of active TS, thawed χ2913 cells that contained the random library were inoculated (1:100) into 1 × YT medium that contained 50 μg/ml thymidine, 50 μg/ml carbenicillin,and 10 μg/ml tetracycline, and grown at 37°C until the absorbance at 600 nm attained a value of 0.8 to 1.0. Aliquots of 1 ml of the exponentially growing cells were pelleted and resuspended in M9 salts,plated on minimal medium that contained appropriate antibiotics, and incubated at 37°C for 36 h. Plasmid DNA was isolated from 66 surviving colonies, and the corresponding random region was sequenced.

To select for library members that are resistant to killing with 5-FdUR, transfectants were plated on the above described minimal medium that contained increasing amounts of 5-FdUR (0–175 nm5-FdUR) and incubated at 37°C for 36 to 48 h. Colonies formed on 5-FdUR that contained agar plates were isolated, and the plasmid from each clone was retransformed into fresh χ2913 E. coli to confirm the drug-resistant phenotype. Each retransformed bacterium was then subjected to the same selection procedure. Plasmid DNA from cells that survived 175 nm FdUR (n = 70), which is lethal to E. coli harboring the wild-type TS, was sequenced.

From a fresh overnight culture, χ2913 cells that contained the wild-type or mutant forms of TS were grown in 30 ml 2 × YT medium that contained carbenicillin. After attaining an absorbance at 600 nm of 0.8, cells were harvested by centrifugation, resuspended in 50 mm Tris-HCl (pH 7.5), 1 mm EDTA, 200 mm NaCl, 10% glycerol, and 200 μg/ml lysozyme,aliquoted, and stored at −80°C. Frozen cells were thawed and lysed on ice for 3 h, centrifuged (27,000 × g), and the pellet discarded. Quantitation of active TS dimer was determined by [³H]-5-FdUMP binding. Crude extracts (50 μl) were incubated with 600 mmCH₂H₄-folate and 300 nm [³H]-5-FdUMP in a standard reaction (500 ml) that contained 50 mmN-tris[hydroxymethyl]methyl-2-aminoethane-sulfonic acid(pH 7.4), 20 mm MgCl₂, 6.5 mm formaldehyde, 1 mm EDTA,and 150 mm 2-mercaptoethanol. The specific activity of [³H]-5-FdUMP was 18.6 Ci/mmol. After incubation for 1 h at room temperature, 125 μl 50%trichloroacetic acid was added, and the mixture was centrifuged at 13,000 × g for 5 min. The pellet was washed four times with 10% trichloroacetic acid, resuspended in a mixture of 2 m NaOH/50% ethanol, and added to 5 ml Scinti-Verse scintillation fluid (Fischer), and the radioactivity was quantitated. For calculations, we assumed 1.7 mol FdUMP bound to 1 mol of TS (7, 8). The assay was repeated using one-half of the amount of crude extract to ensure[³H]-5-FdUMP saturation. Reactions were conducted in duplicate. No counts above background were detected in a control reaction incubated in the absence of cell extract. Quantitation of purified wild-type TS was conducted as a control and was in agreement with the concentration obtained by the dye-binding procedure of Bradford. Total protein in the crude extracts was quantitated using the Bradford assay.

To construct a plasmid expressing the mutant TS enzymes linked to a 6X-His polypeptide, the TS mutants were digested with MroI and NcoI and ligated into a digested pHis-TS-WT construct. pHis-TS-WT was constructed via PCR amplification of TS, creation of a SalI restriction endonuclease site, digestion with NdeI and SalI, and ligation into the similarly digested vector pHis (a modified pUC12 vector provided by Amnon Hizi,Tel Aviv University, Tel Aviv, Israel). Details of its construction have been described previously (7). Cloning procedures were confirmed by restriction analysis and DNA sequencing. The TS-6X-His fusion proteins were purified by a one-step Ni²⁺ affinity chelation chromatographic procedure using resin and buffers (His-Bind resin and buffer kit; Novagen)according to a previously described protocol (7) modified from that of the supplier. An overnight culture of approximately 125 ml yielded from 400 to 600 μg of purified TS. After SDS polyacrylamide gel analysis and dialysis, the concentration of purified TS was determined by using the Bradford assay.

TS activity was monitored spectrophotometrically by the increase in absorbance at 340 nm that occurs concomitant with the production of H₂-folate (Δε = 6400 m⁻¹ cm⁻¹;Ref. 16, 17). The standard reaction buffers and methodology have been described previously (7). When the concentration of dUMP was varied, a high concentration of(6R,S)-CH₂H₄-folate(600–2400 μm) was added; when CH₂H₄-folate was varied,the concentration of dUMP was 500 μm. CH₂H₄-folate was added to initiate the reaction at 25°C. Michaelis constants(K_m) for CH₂H₄-folate and dUMP were determined from initial velocity measurements, obtained on a Perkin-Elmer Lamba Bio 20 UV/Vis spectrophotometer. Steady state kinetic parameters were subsequently obtained by a nonlinear least squares fit of the data to the Michaelis-Menten equation using Kalidegraph 3.0 software (Abelbeck Software, Reading, PA). k_cat values were obtained by dividing the specific activity (V_max/mg protein) by the molecular weight of enzyme.

K_is for 5-FdUMP were obtained from the steady-state inhibition reaction rates for mixtures of enzyme, dUMP,CH₂H₄-folate, and inhibitor at 25°C. A constant dUMP concentration of 500μ m was used, and FdUMP concentrations were varied to measure inhibition by FdUMP. To initiate reactions,CH₂H₄-folate was added to a final concentration of 600 μm for all of the mutants with the exception of K47Q;D48E, where 1200μ mCH₂H₄-folate was used.

A library of 1.5 × 10⁶ human TS mutants was created by random sequence mutagenesis. The randomized oligonucleotides were designed with a bias such that each of the seven randomized codons contains on average 80% of the wild-type nucleotide at each nucleotide position and 20% of the remaining three nucleotides.

Before selection, plasmid DNA was isolated from 24 transformed clones and sequenced. The number of substitutions/clone is presented in Table 1, and the types of substitutions are tabulated in Fig. 1,A. An average of 5.1 nucleotide changes and 2.8 amino acid changes/clone were detected in the nonselected library (Table 1). Amino acid substitutions in the nonselected clones were approximately evenly distributed among the seven residues encoded by the randomized nucleotides (Fig. 1 A). The least number of substitutions were observed at G52. Four of the 24 nonselected clones analyzed (17%)contained frameshift mutations (all were deletions), and three contained termination codons. On the basis of the number and frequency of random substitutions, we calculated that the probability of obtaining a wild-type nucleotide sequence in the nonselected library is approximately 1%. The likelihood of obtaining a wild-type protein is higher because of the degeneracy of the nucleotide code and is approximately 4.5%. Therefore, it is not surprising that no wild-type molecules were detected among the 24 sequenced nonselected clones.

We isolated active TS enzymes from the large plasmid libraries using a positive genetic complementation (7). The wild-type TS construct is able to rescue the TS⁻ E. coli phenotype and form colonies on minimal medium (containing no thymidine), yet E. coli expressing an inactive TS(containing a “stuffer” insert) will not grow in the absence of thymidine. Approximately 1% of the members of the random library, or 15,000 clones, were able to complement the TS⁻phenotype and presumably encode and express active TS. Unlike the nonselected library, plasmid DNA from the active mutant library was without any nonsense or frameshift mutations. The average number of nucleotide and amino acid substitutions in the active library were 3.4 and 1.7, respectively (Table 1). Unlike the nonselected library, 7 of the 66 clones produced wild-type enzyme (10%). As statistically expected, each of the DNAs that encoded the wild-type enzymes was unique as indicated by the presence of one or more silent nucleotide changes. Among protein sequences, the double mutant D48Y;G52C was detected twice. Most of the targeted residues retained a high level of mutability with the exception of Tyr 33, Arg 50, and, to a lesser degree, T51. Only two alterations at Y33 were observed, Y33C (in the context of the triple mutant Y33C;D48N;G52A), and Y33S (as Y33S;D48E). Arg 50 was able to be substituted by proline, asparagine, serine, and tryptophan. In each clone that harbored a mutation at Arg 50, at least one other amino acid substitution was detected within the Arg⁵⁰-loop. Threonine 51 was altered in only seven clones, twice to alanine and five times to serine (Fig. 1 B).

To isolate members from the random enzyme library that exhibit enhanced resistance to 5-FdUMP, positive genetic selection was again used by plating on minimal medium agar plates that contained gradients of 5-FdUR. Previously, we reported (7, 8) that the survival of the E. coli harboring wild-type TS is only modestly reduced to 90% at 75 nm 5-FdUR; however,it precipitously declines to 0.1% at 100 nm of the analogue. Consistent with these results, no clones harboring wild-type TS enzyme formed colonies on media that contained >100 nm 5-FdUR (Fig. 2). At 100 nm 5-FdUR, approximately 60% of the random library formed colonies. Library survival decreased linearly, to 40% at 150 nm 5-FdUR and 8% at 175 nm 5-FdUR.

The mutability of residues retained a trend similar to that observed in the active mutant library, with one exception. Thr 51, which was mutated only in 7 of the 66 active clones (11%), is altered in 30(43%) of the 5-FdUR-selected clones. It has now become the third most commonly mutated residue. Serine is by far the most common substitution of this residue, present in 18 of the 30 mutations detected among the 5-FdUR-resistant clones.

Whereas wild-type TS represented 10% of the active mutant library, no wild-type enzymes were detected among the 70 5-FdUR survivors. Seven mutants were detected twice in the library (D48E, D48E;D49G, D48E;T51S,T51S;G52S, D48F;T51S;G52C, D48Y;T51S;G52A and K47E). One mutant, K47N,was detected three times. In all of the cases but two, the clones isolated were unique, as indicated by silent changes (nucleotide substitutions that did not alter the protein sequence) in other nucleotides within the random region. Thus these 70 mutants isolated from the 5-FdUR resistant clones, which harbored anywhere from one to three amino acid changes, were independently created and selected among 1.5 million original mutants.

It is apparent from Fig. 1 C that many alterations lead to drug resistance. Mutants harboring a single amino acid change included:K47Q, K47E, D48E, G52S, and G52P. Interestingly, T51S was not detected as a single mutant despite being the single most common change seen in the library. However, the alteration T51S was often detected in concert with the mutation G52S. Examples include T51S;G52S, D48G;T51S;G52S, and D49E;T51S;G52S. T51S;G52S was detected twice, with unique silent nucleotide changes, in the library. As was observed among active mutants, changes at Arg 50 and Tyr 33 were rare and never occurred in the absence of other alterations. The only mutation at Y33 in the drug-selected clones was to tryptophan in the clone D48E;Y33W.

Several mutants were manually screened at increasing doses of 5-FdUR (Fig. 2). The previously identified drug-resistant mutant A197V;L198I;C199F was used as a positive control (7). This mutant, which was found to have a K_dfor 5-FdUMP 20 times greater than that of wild type, formed colonies at dosages up to but not above 150 nm, consistent with previous observations (Fig. 2,A; Ref.7, 8). The single mutants D48E and G52S formed colonies in media that contained as much as 200 nm 5-FdUR(Fig. 2,A). The mutant G52C was identified in the active mutant pool. Because of the published importance of alterations at G52 in conferring drug resistance (5), this single mutant was manually screened and demonstrated drug resistance up to 175 nm 5-FdUR (Fig. 2,A). Interestingly,with one exception, the multiple mutants demonstrated improved colony-forming ability compared with the single mutants. K47Q;D48E formed colonies at up to 250 nm 5-FdUR;D48E;T51S;G52C at up to 300 nm; and lastly T51S;G52S grew at up to 400 nm (Fig. 2 B). The exception is the single mutant K47E, also identified by Tong et al. (5), which formed colonies at 400 nm.

To rule out that TS overexpression in the transfected E. coli contributed to the observed drug resistance, multiple clones of the wild-type and mutant-transfected χ2913 cells were analyzed using a [³H]-FdUMP binding assay(18). The wild-type and single mutant G52S demonstrated the highest level of expression (6.3 and 5.1 ng TS/μg total protein),four times that of the double and triple mutants (range 1.3–1.9 ng TS/μg total protein). However, these two proteins demonstrated the lowest survival against 5-FdUR in E. coli. These results suggest that the major reason for the observed improvement in survival at dosages of 5-FdUR clearly lethal to the wild-type TS can be attributed to the mutation in TS rather than to its level of expression.

Because augmented survival in E. coli could potentially result from mechanisms unrelated to altered enzymology, we purified the wild-type and four mutant enzymes as NH₂-terminal fusions with a histidine tag. After purification, a single major component on SDS polyacrylamide gel migrating with an apparent molecular weight of human TS monomer(M_r 36,000) was observed for the wild-type and mutant enzymes. Purity was estimated to be >80%.

To obtain information about the catalytic and ligand-binding properties of these TS variants, the V_max and K_m values for the substrate and cofactor and K_i values for the inhibitor 5-FdUMP were determined (Table 2). Kinetic parameters of the single mutant G52S were consistent with the findings of Tong et al. (5). All of the mutants were catalytically efficient, with the lowest k_cat measuring approximately half that of the wild-type enzyme.

As expected, the K_i values for FdUMP of all of the mutant enzymes were greater than that of the wild type,consistent with the E. coli survival data. The K_i value of the G52S single mutant was 18-fold higher than wild type, consistent with that observed by others(5). By comparison, the K_i values of the double mutant T51S;G52S, which demonstrated colony-forming ability in E. coli at much higher concentrations of 5-FdUR, were increased modestly to 26 times that of wild type. The mutant K47Q;D48E demonstrated 75-fold increase in K_iover wild type. The highest level of drug resistance was seen in the triple mutant D48E;T51S;G52C. This mutant demonstrated a high degree of drug resistance in E. coli (approximately 4-fold improvement in maximal tolerable dose), only a modestly increased K_m for the normal substrate dUMP(6-fold), and an observed K_i for FdUMP 100-fold higher than that of wild-type TS.

The three loop residues Asp 48, Asp 49, and Gly 52 tolerated many substitutions. In fact, by combining the substitutions found in the active library with the 5-FdUR resistant library, 14 of the possible 20 amino acids were tested at D48. D49 and G52S were each altered to 10 other residues. Interestingly, whereas D48 and G52 are not conserved throughout evolution, and mutagenesis studies have indicated that G52 can tolerate any substitution without significant loss of activity (19), D49 is absolutely conserved. Yet in our studies, all of the three residues tolerated polar, charged, or hydrophobic alterations. The lack of conservation in D48 and G52 is likely a reflection of their exterior location in the loop. Although we would have expected alterations at D49 to be more restrictive, it has been a common finding that highly conserved residues of TS are often tolerant of amino acid mutations in the laboratory (20).

In contrast, relatively few changes were observed in evolutionarily highly conserved residues Tyr 33, Arg 50, and Thr 51. The only alterations detected at Tyr 33 were to cysteine and serine in the active clones and to tryptophan (as D48E;Y33W) in the drug-resistant clone. Although substitutions at Tyr 33 have not been extensively studied, the mutant Y33H results in an enzyme with an 8-fold decrease in k_cat, a K_d value for the inhibitor 5-FdUMP of four times that of the wild type, and a near normal affinity for dUMP and the cosubstrate (21, 22, 23). Residue 33, a Tyr in all of the species of TS except Lactobacillus Lactis, is located within the α-helix A and forms hydrogen bonds with residues of the α-helix J which forms a wall of the active site cavity(24, 25). Our study did not confirm the Y33H alteration nor find other substitutions at this residue that were highly associated with drug resistance. Given the role of Arg 50 in maintaining the bound structure, we find it remarkable that even a modest breadth of residue substitutions were detected (Pro, Ser, Asn,Trp, Gly). None of these were single mutations, preventing us from ruling out the effects of complementary changes in the loop. However R50P, R50G, and R50S have been constructed as single mutants via site-directed mutagenesis and determined to be active via E. coli complementation (4, 26).

Although it was not found as a single substitution in the drug-resistant library, T51S was the most common substitution observed. It occurred synergistically with many changes, notably G52S. Also,survival data indicate that clones tested with the T51S alteration appear to demonstrate colony-forming ability in E. coli at the highest doses of 5-FdUR. After binding of the substrates, it has been observed that reorientation of the Arg⁵⁰-loop allows for the interaction of hydrophobic atoms of Thr 51 to contact the buried V313 side chain (10, 11, 12, 19, 27). Mutation to serine could alter the conformation of this absolutely required terminal valine or nearby residues.

Remarkably, it appears that mutants that contain multiple amino acid changes survive at higher dosages of 5-FdUR than their corresponding single mutants. For example, G52S and D48E both formed colonies up to 200 nm, and G52C to 175 nm, but the double mutants T51S;G52S and D48E;T51S;G52C formed colonies at 400 and 300 nm 5-FdUR, respectively (Fig. 2 A and B). This is in agreement with the observation that the average number of amino acid changes increases in the drug-resistant library (2.2 amino acid alterations/clone) relative to the active mutant library (1.7 amino acid changes/clone; Table 1). The increase in the average number of amino acid changes is not solely attributable to elimination of the wild-type enzyme; the relative proportion of mutants with three and four changes is greater than that of the active mutant library (data not shown).

In our sequencing of 70 drug-resistant mutants, we did not detect the single mutant D49G, identified by Tong et al.(5). However, this mutation was detected in concert with other alterations. All of the mutants analyzed survived not only at concentrations of 5-FdUR prohibitive for the growth of wild-type TS, as expected, but at concentrations higher than the previously reported active site mutant that we identified A197V;L198I;C199F(7). Thus these novel mutants appear in our hands to be the most highly 5-FdUR-resistant TS enzymes to date. The observed small number of colonies at high concentrations of 5-FdUR (Fig. 2) may be attributable to local perturbations in concentration of 5-FdUR caused by adjacent nondividing cells or scavenging of thymidine from the adjacent cells in plates.

Binding studies that used [³H]-FdUMP demonstrated that the increased survival was not because of preferential overexpression of the mutant TS forms. Expression of G52S was approximately equivalent to that of the wild-type TS, whereas enzyme levels of the three remaining mutants analyzed were approximately 4-fold less than wild type. This could be attributable to decreased intrinsic stability and preferential degradation of mutants with two to three amino acid changes as compared with the wild-type or single mutant G52S.

On the basis of both robust survival in E. coli and the substitutions encoded, we selected three mutants (T51S;G52S,D48E;T51S;G52C, and K47Q;D48E) alongside the wild type and the previously identified G52S, for purification and kinetic studies. Each of these mutant TSs were catalytically active, with no mutant displaying more than a 50% decrease in k_cat. K_m values for dUMP for all of the mutants were approximately five times the wild type, with the exception of K47Q;D48E (K_m ∼ 20 times wild type). CH₂H₄-folate affinity appeared to follow the same pattern as dUMP affinity. G52S demonstrated a K_m for CH₂H₄-folate not significantly different from wild type, consistent with that observed by Tong et al. (5). However, the same alteration in combination with T51S (T51S;G52S) had a 10-fold increase in K_m for the cosubstrate. The triple mutant, with alterations at these two positions alongside D48E(D48E;T51S;G52C), demonstrated a K_m of almost 20-fold wild type. Lastly, K47Q;D48E demonstrated a 30-fold increase in K_m for CH₂H₄-folate (Table 2).

Kinetic inhibition studies confirmed the E. coli genetic selection studies in that all of the enzymes were 5-FdUR resistant compared with the wild-type TS. In accord with the E. colidata but to a larger degree, K47Q;D48E demonstrated a 75-fold increase in K_i, and the triple mutant D48E;T51S;G52C displayed the greatest resistance with a K_i for 5-FdUMP 100-fold greater than that of the wild type. In some cases, K_i has been shown to not fully correlate with IC₅₀ (5, 6). However,as our enzymes have been selected in a biological system, we expect the observed survival advantage to extend to mammalian cells. Work is currently underway to determine survival in a mammalian system.

For many years, the only known drug-resistant human TS enzyme was the mutant Y33H, discussed previously (20). In the last 2 years, other drug-resistant TSs that contain amino acid substitutions in many different regions of the protein have been identified. We have previously identified several 5-FdUR-resistant TSs by creating random substitutions in a stretch of 13 residues near the active site(7, 8). By subjecting highly conserved residues important in cofactor binding to site-directed mutagenesis, I108A was found to be resistant to the antifolates Raltitrexed (Tomudex; ZD1694) and Thymitaq, and F225W was found resistant to the antifolate BW1843U89 and 5-FdUR (6). Recently, a cell line adapted to increasing concentrations of 5-FdUR was determined to encode a P303L mutant that,although metabolically unstable, was nonetheless able to confer resistance to transfected cells against FdUR, Raltitrexed, Thymitaq,and BW1843U89 (9).

Studies that used EMS mutagenesis followed by selection in human HT1080 cells with the antifolate Thymitaq have identified three mutants that demonstrate resistance to Thymitaq and 5-FdUR (5). All three of these identified mutants contain substitutions in the essential and conserved Arg⁵⁰-loop. These mutants, all of which harbor a single amino acid alteration (K47E,D49G, or G52S), confer a high degree of drug resistance to mammalian cells in culture (5). The mutant G52S conferred almost a 100-fold increase in IC₅₀ against 5-FdUR, whereas D49G demonstrated a 40-fold increase in IC₅₀ for Thymitaq while still retaining resistance to 5-FdUR. Inhibition studies with 5-FdUMP and three antifolate inhibitors indicated that D49G and G52S demonstrated an increase in the K_i for FdUMP of 5.4- and 20-fold,respectively, while retaining a K_i for Thymitaq approximately 6-fold that of wild type. Interestingly, despite a 5-fold increase in IC₅₀ for Thymitaq and 5-FdUR, K47E did not demonstrate variant kinetics for any of the inhibitors tested (5).

Because EMS mutagenesis is statistically unlikely to test synergistic effects of multiple mutations in one polypeptide or to allow multiple nucleotide substitutions within a particular codon, we view the mutants identified by Tong et al. (5) as prototype, or first-generation, drug-resistant TSs that can guide the discovery of mutants with yet greater resistance. Because 5-FdUMP is structurally similar to the natural substrate dUMP, it is difficult to predict how single amino acid substitutions or multiple substitutions could restrict the binding of 5-FdUMP without affecting binding of dUMP. Random oligonucleotide mutagenesis provides a combinatorial alternative that can examine a large amount of sequence space and create altered enzymes without requiring detailed knowledge about amino acid interactions or effects of specific alterations. Reselection of the single mutants discovered by Tong et al. (Ref.5; G52S and K47E) demonstrates the validity of our E. coli assay in several ways. Firstly, this rapid E. coli assay was effective in identifying previously detected 5-FdUR-resistant TSs among an extremely large plasmid library. Secondly, the types of alterations seen as a whole in the resistant mutants appear to mimic those seen via selection in human cells. The fact that the identical mutants (G52S, K47E) could be selected via either 5-FdUR or a folate-based inhibitor (Thymitaq) can be explained likely by the central role of the Arg⁵⁰-loop in coordination with both of the corresponding substrates.

The decrease in affinity to 5-FdUMP in mutants with near normal catalytic activity suggests these mutants may be suitable for use as drug-resistant genes in gene therapy applications. Although effective, the use of 5-FU as a chemotherapeutic agent has been limited by toxicity to bone marrow, gastrointestinal, and other tissues(28). The introduction and expression of mutants of human TS that can function in the presence of systemic 5-FU chemotherapeutic treatment could protect normal cells from cytotoxicity or allow augmentation of the maximally tolerated dose of 5-FU. Additionally, a better understanding of the interactions of thymidylate synthase with 5-FdUMP will be useful in the targeted drug design of more clinically effective pyrimidine or antifolate inhibitors.

We thank Dr. Joseph Bertino and members of his laboratory for guidance, detailed kinetic information, and stimulating discussion.