The folate pathway is a target for resistance to the drug para-aminosalicylic acid (PAS) in mycobacteria



The increasing rate of multidrug-resistant tuberculosis has led to more use of second-line antibiotics such as para-aminosalicylic acid (PAS). The mode of action of PAS remains unclear, and mechanisms of resistance to this drug are undefined. We have isolated PAS-resistant transposon mutants of Mycobacterium bovis BCG with insertions in the thymidylate synthase (thyA) gene, a critical determinant of intracellular folate levels. BCG thyA mutants have reduced thymidylate synthase activity and are resistant to known inhibitors of the folate pathway. We also find that mutations in thyA are associated with clinical PAS resistance. We have identified PAS-resistant Mycobacterium tuberculosis isolates from infected patients, which harbour mutations in thyA and show reduced activity of the encoded enzyme. Thus, PAS acts in the folate pathway, and thyA mutations probably represent a mechanism of developing resistance not only to PAS but also to other drugs that target folate metabolism.


Para-amino salicylic acid (PAS) was one of the first antibiotics found to be effective in the treatment of tuberculosis in the 1940s (Lehmann, 1946). PAS potentiates the activity of isoniazid and streptomycin and was widely used in combination chemotherapy against Mycobacterium tuberculosis (Offe, 1988). However, PAS caused gastrointestinal toxicity leading to poor patient compliance (Iwainsky, 1988). As more easily tolerated antibiotics became available, PAS usage diminished considerably. Two developments have, however, led to the re-emergence of PAS in antitubercular therapy. First, a new formulation of the drug has fewer gastrointestinal side-effects (Iwainsky, 1988). Secondly, the appearance of a widespread epidemic of multidrug-resistant tuberculosis has necessitated the use of alternatives to the first-line antimycobacterial agents that are used routinely in short-course chemotherapy regimens (Dye et al., 2002). As PAS treatment has been relatively uncommon, even highly drug-resistant strains appear to have limited resistance to this agent. Thus, PAS has become one of the principle second-line antibiotics for the treatment of multidrug-resistant tuberculosis (Mitnick et al., 2003).

Despite the long history of PAS usage, its mechanism of action has remained unclear. The structural similarity between PAS and sulphonamides led to the suggestion that it might compete with para-amino benzoic acid (PABA) for dihydropteroate synthase (DHPS), a key enzyme in folate biosynthesis (Fig. 1A). However, unlike sulphonamides, PAS appears to be a poor inhibitor of DHPS in vitro, raising the possibility that it may have a different target (Nopponpunth et al., 1999).

Figure 1.

Mycobacterium bovis BCG strains with insertions in thyA are resistant to PAS.
A. Folate biosynthesis. Thymidylate synthase catalyses the conversion of dUMP to dTMP, which is required for DNA synthesis. This methylation reaction uses 5,10-methylene (6R) THF as a reductant and generates DHF. DHF is converted back to THF by DHFR. As a result, the intracellular stores of THF derivatives, which serve as one-carbon donors in diverse metabolic processes, are effectively replenished. PABA, para-aminobenzoic acid; DHPS, dihydropteroate synthase; DHP, dihydropteroate; DHF, dihydrofolate; DHFR, dihydrofolate reductase; THF, tetrahydrofolate; dUMP, deoxyuridine 5′-monophosphate; dTMP, deoxythymidine 5′-monophosphate.
B. The thyA locus in M. tuberculosis showing the folate binding and dUMP binding sites based on homology with the E. coli gene. Transposon insertion sites for PAS1, PAS2 and PAS5 are indicated.

In order to elucidate the mode of action of PAS, we sought to identify mycobacterial genes conferring resistance to this drug. Disruption of specific genes has been shown to result in drug resistance in mycobacteria. For example, most clinical resistance to the antibiotic isoniazid is associated with loss of function of the catalase-encoding gene katG. Similarly, transposon insertions in katG produce isoniazid resistance in Mycobacterium bovis BCG (Sassetti et al., 2001; Somoskovi et al., 2001). We therefore reasoned that transposon insertions might also give rise to PAS resistance in mycobacteria.

Here, we show that a class of PAS-resistant transposon mutants of M. bovis BCG harbours insertions in the thymidylate synthase (thyA) gene, which encodes a critical enzyme involved in determining intracellular folate levels. These mutants have reduced thymidylate synthase activity and are resistant to known inhibitors of the folate pathway. We also find that some PAS-resistant clinical isolates of M. tuberculosis have mutations in thyA that result in decreased enzyme activity. These data suggest that PAS acts as a folate antagonist and that mutations in thyA are implicated in clinical PAS resistance.


Transposon insertions in thyA produce resistance to PAS in M. bovis BCG

To isolate PAS-resistant mutants, we plated a library of ≈ 105mariner transposon insertion mutants in M. bovis BCG on to solid medium containing 1 µg ml−1 PAS and selected for growth. Within the transposon library, PAS-resistant colonies were present at a frequency 560-fold higher than in wild-type BCG. The insertions were cloned and sequenced, and three out of 12 were found to lie within the thyA gene, which is predicted to encode the enzyme thymidylate synthase. Thymidylate synthase catalyses the reductive methylation of deoxyuridine 5′-monophosphate (dUMP) to yield deoxythymidine 5′-monophosphate (dTMP), which is necessary for DNA synthesis (Green et al., 1996) (Fig. 1A). This enzyme is highly conserved across species, especially in the regions involving dUMP and folate binding (reviewed by Finer-Moore et al., 2003). As shown in Fig. 1B, all insertions were near the 3′ end of the gene. thyA mutants (designated PAS1, PAS2, PAS5) were able to grow in the presence of high concentrations of PAS (up to 27 µg ml−1), while PAS inhibited the growth of wild-type BCG at 0.3–1 µg ml−1.

Mutants harbouring insertions in thyA did not exhibit growth defects in vitro when grown in Middlebrook 7H9 broth. Interestingly, these mutants also grew normally in Proskauer–Beck medium, a minimal medium that does not contain thymine or thymidine. The addition of thymine to this medium had no effect on growth rate (data not shown). Thus, unlike thyA mutants of Escherichia coli and Salmonella typhimurium (Green et al., 1996; Kok et al., 2001), thyA transposon mutants of M. bovis BCG do not appear to require exogenous thymine for growth and are thus not auxotrophs.

thyA transposon mutants have reduced thymidylate synthase activity

In order to determine whether insertions in thyA resulted in decreased enzyme activity, we indirectly assayed thymidylate synthase in BCG extracts. Thymidylate synthase activity requires the cofactor 5,10-methylene tetrahydrofolate (5,10-methylene THF), which functions as both a reductant and a one-carbon donor in the methylation reaction that results in dTMP formation from dUMP (Finer-Moore et al., 2003). We monitored spectrophotometrically the conversion of 5,10-methylene THF to dihydrofolate (DHF) in the presence of extracts and dUMP substrate over time. Mutants containing insertions in thyA all had reduced activity compared with wild-type BCG (Fig. 2).

Figure 2.

Reduced thymidylate synthase activity in thyA mutants. Extracts from wild-type (WT) and M. bovis BCG thyA mutants (PAS1, PAS2 and PAS5) were assayed for thymidylate synthase activity as described in Experimental procedures. Data are representative of four independent experiments.

To test whether the reduction in thymidylate synthase activity resulted from the thyA-encoded thymidylate synthase, we used a thyAE. coli strain. thyAE. coli are thymine auxotrophs and require exogenous thymine from supplemented media (Neuhard and Kelln, 1996). The wild-type and disrupted thyA genes were expressed in E. coli under the control of a promoter with activity that is inducible with arabinose. thyAE. coli are unable to grow on medium lacking thymine (Fig. 3). In agreement with a previous report (Matsumoto et al., 1998), expression of wild-type M. bovis BCG thyA restored growth of thyAE. coli in minimal medium lacking thymine (Fig. 3). In contrast, expression of the disrupted thyA genes from PAS-resistant BCG (PAS2 and PAS5) failed to complement thyAE. coli. Moreover, although E. coli expressing the wild-type gene showed enzyme activity comparable to that of wild-type E. coli, thyAE. coli expressing the mutant genes had no measurable thymidylate synthase activity (data not shown).

Figure 3.

Growth of thyAE. coli upon expression of wild-type and mutant thyA genes from M. bovis BCG. thyA genes from wild type and thyA transposon mutants were cloned into pBAD24 and transformed into thyA+ and thyAE. coli as described in Experimental procedures. Growth on M9 minimal medium alone and M9 + 0.2% arabinose is shown. 1 and 5 (pBAD24 vector), 2 and 6 (wild-type thyA from BCG), 3 and 7 (mut2 thyA from PAS2), 4 and 8 (mut5 thyA from PAS5). 1–4 are in thyA+E. coli, 5–8 are in thyAE. coli.

thyA mutants are resistant to other antifolates

The folate biosynthesis pathway in bacteria generates folate cofactors that are required for DNA and RNA synthesis and essential one-carbon addition reactions. DHF is converted to THF by dihydrofolate reductase (DHFR) (Fig. 1A). As THF is consumed largely by thymidylate synthase activity, inhibition of thymidylate synthase results in increased availability of THF for one-carbon addition reactions (Green et al., 1996). Thus, thyA mutants in many bacteria are relatively resistant to inhibitors of DHFR (Huovinen et al., 1995). We tested the sensitivity of thyA mutant E. coli expressing M. bovis BCG thyA to the DHFR inhibitor trimethoprim. Although thyA-deficient E. coli were resistant to trimethoprim, thyAE. coli expressing wild-type thyA from BCG became sensitive (Fig. 4A). However, bacteria expressing transposon-disrupted thyA genes remained resistant; their growth in the presence of trimethoprim was comparable to that of thyA mutant E. coli. All constructs in the thyA+E. coli background remained sensitive to trimethoprim. We obtained comparable results for susceptibility to trimethoprim by assaying growth on plates using Kirby–Bauer sensitivity tests (data not shown).

Figure 4.

Effect of antifolate inhibitors on thyA mutants.
A. Effect of trimethoprim (5 µg ml−1) on growth of thyA
+ and thyAE. coli strains each transformed with vector (pBAD24), wild-type thyA from BCG, mutant thyA from PAS2 (mut2) and PAS5 (mut5). Growth (OD600) after 2.5, 4, 7.5 and 22 h of incubation with trimethoprim is shown.
B. Effect of DHFR inhibitor 8710 (50 µg ml−1) on PAS-sensitive (wild-type BCG) and PAS-resistant thyA mutants (PAS1, PAS2, PAS5). Growth (OD600) on days 0, 2 and 6 after addition of inhibitor is shown. Data are representative of three independent experiments.

Trimethoprim does not inhibit the growth of mycobacteria. However, several lipophilic trimethoprim analogues have been synthesized recently that antagonize mycobacterial DHFR and successfully inhibit bacterial growth (Suling et al., 1998). We found that one such compound, 8710, was active against wild-type M. bovis BCG. In contrast, the thyA mutants PAS1, PAS2 and PAS5 were able to grow in the presence of 8710 (Fig. 4B). Measurements of growth in broth correlate with growth on plates in these experiments. Thus, similar to PAS, resistance to known inhibitors of folate biosynthesis appears to be linked to compromised thymidylate synthase function in mycobacteria.

Clinical M. tuberculosis isolates resistant to PAS have mutations in thyA

As we found that mutations in mycobacterial thyA do not result in auxotrophy, we wished to determine whether thyA mutants were represented in PAS-resistant M. tuberculosis strains from infected patients. We obtained PAS-resistant M. tuberculosis strains, each isolated from patients in different locations in Russia. We amplified and sequenced the entire thyA gene and 250 bp of sequence upstream of the translational start site from each of eight PAS-resistant and 10 PAS-sensitive clinical isolates of M. tuberculosis. Interestingly, three of the eight PAS-resistant strains harboured mutations in thyA(Fig. 5A). Two isolates had single point mutations that resulted in substitution of threonine with alanine at amino acid residue 202 of the predicted protein, while the third had a point mutation that resulted in arginine to glycine substitution at amino acid 222. All 10 PAS-sensitive isolates contained no thyA mutations. In addition, we sequenced 250 bp of sequence upstream of the translational start site and the open reading frame (ORF) of the thyX gene, which encodes an alternate thymidylate synthase (Myllykallio et al., 2002). All eight PAS-resistant and the 10 PAS-sensitive clinical isolates had wild-type thyX sequence (data not shown), indicating that mutations in thyX did not account for PAS resistance in these strains.

Figure 5.

PAS-resistant clinical isolates of M. tuberculosis have mutations in thyA.
A. Diagram representing location of mutations in the thyA gene of M. tuberculosis. Sequences of PAS-resistant strains (R1, R2, R3) containing mutations are shown in comparison with the reference strain H37Rv.
B. Mutant thyA alleles fail to confer sensitivity to trimethoprim. Growth (OD600) of thyAE. coli strains transformed with vector (pBAD24), the thyA gene from PAS-sensitive (S1, S2) and PAS-resistant (R1, R2, R3) strains after 2.5, 4, 7.5 and 22 h incubation with trimethoprim. Data are representative of three independent experiments.
C. Mutant thyA alleles show reduced ability to restore PAS sensitivity to thyA transposon mutants compared with wild-type thyA. Growth of strains over 6 days was assessed in the absence of drug or in the presence of 0.3, 1.0, 3.0, 9.0 and 27.0 µg ml−1 of PAS as described in Experimental procedures. BCG was transformed with pMV261 vector alone (BCG/V) or with wild-type thyA from PAS-sensitive M. tuberculosis (BCG/WT thyA). PAS2 was transformed with vector (PAS2/V), wild-type thyA (PAS2/WT thyA) or thyA from PAS-resistant M. tuberculosis strain R1 (PAS2/R1 thyA) or R2 (PAS2/R2 thyA). Growth of each strain in the absence of PAS is represented as 100. Relative OD600 represents growth in the presence of each concentration of PAS relative to the OD600 in the absence of drug (×100). Asterisk indicates P < 0.01 by Student's t-test. Data are representative of three independent experiments.

To assess the activity of these altered thyA alleles, we cloned and expressed the genes in E. coli as described in Fig. 3. All the M. tuberculosis thyA alleles allowed the growth of thyAE. coli when expressed in the presence of arabinose, and extracts from these strains showed normal levels of thymidylate synthase activity (data not shown). However, only E. coli expressing the mutant thyA alleles remained relatively resistant to trimethoprim (Fig. 5B). These data indicate that the thymidylate synthase encoded by the mutant alleles has reduced activity in vivo compared with wild type, implying that decreased thymidylate synthase activity can confer clinical PAS resistance.

In order to determine the effect of overexpressing wild-type ThyA in a thyA-deficient mutant, we cloned the thyA gene from M. tuberculosis and expressed it in the thyA transposon mutant PAS2 and in wild-type BCG. We then assessed the growth of these transformants in the presence of PAS. Overexpression of wild-type thyA restored sensitivity to PAS in a thyA transposon mutant (PAS2/thyA) (Fig. 5C). BCG expressing heterologous ThyA (BCG/thyA) appeared to be more sensitive to PAS than the control (BCG/V) (Fig. 5C). We also overexpressed thyA alleles from the clinical mutant strains R1 and R2 in PAS2. Interestingly, the mutant thyA alleles showed significantly decreased ability to confer susceptibility to PAS compared with wild-type thyA (Fig. 5C). These data thus suggest that the thyA mutations in the clinical isolates contribute to PAS resistance.


The generation of folate derivatives is crucial for bacterial metabolism, and the folate pathway is thus thought to be an important target for the development of new antibiotics and antimetabolites. The renewed interest in using PAS as a second-line drug for treating multidrug-resistant tuberculosis prompted us to study how resistance to this drug may occur. Here, we show that mutations in thyA result in PAS resistance in both experimental systems and clinical isolates, strongly suggesting that this antibiotic targets a component of folate biosynthesis. PAS is a structural analogue of PABA and has therefore been thought to be an inhibitor of DHPS (Iwainsky, 1988) (Fig. 1A). However, PAS fails potently to inhibit purified DHPS in vitro (Nopponpunth et al., 1999). The inability to demonstrate inhibition in vitro, however, does not prove that DHPS is not the in vivo target. It is possible that PAS accumulates to high levels within bacterial cells, leading to inhibition of DHPS activity in vivo. Alternatively, analogous to the mode of action of other antituberculosis drugs such as isoniazid and pyrazinamide, PAS might be converted to a more potent ‘active’ form inside bacterial cells (Somoskovi et al., 2001).

thyA mutants in several bacteria are resistant to folate antagonists. As thymidylate synthase is a major consumer of reduced folate in bacteria, when enzyme activity declines or is inhibited, utilization of THF derivatives is also decreased. Thus, more reduced folate becomes available for other essential one-carbon addition reactions leading to bacterial survival (Green et al., 1996). However, thyA mutations do not appear to constitute a clinically relevant mechanism of resistance in most pathogens, as mutations in thyA frequently lead to thymine auxotrophy. For example, Salmonella typhimurium lacking thyA are unable to survive in macrophages and are attenuated for virulence in mice (Kok et al., 2001). Shigella flexneri thyA auxotrophs are also impaired in intracellular growth and intercellular spreading and show reduced virulence (Cersini et al., 1998). These data suggest that there is inadequate thymine in the host to support the growth of intracellular bacteria lacking thymidylate synthase. Thymidylate synthase thus appears to be required for virulence in many bacteria, which may explain why most trimethoprim-resistant clinical isolates do not have mutations in thyA but in the gene encoding DHFR (Huovinen et al., 1995). We cannot rule out the possibility that resistance to PAS in mycobacteria results from additional mutations leading to the overproduction of DHFR. However, as overexpression of thymidylate synthase in the M. bovis BCG thyA mutant abrogates PAS resistance (Fig. 5), this possibility appears to be unlikely. Although we have identified mutations in thyA in a subset of PAS-resistant clinical isolates, resistance to this antibiotic may also result from mutations that lead to altered expression or activity of other components of the folate pathway. We did not observe any other transposon mutations in genes encoding products involved in folate metabolism. This may be, in part, because, unlike thyA, these genes are indispensable for survival.

How are thyA mutant mycobacteria able to survive without thymine supplementation? Recently, Myllykallio et al. (2002) reported a distinct thymidylate synthase encoded by thyX, which has negligible homology to thyA. The thyA class of enzymes is found in a large number of bacterial species and many eukaryotes, whereas thyX-encoded thymidylate synthases are mostly present in microbial genomes lacking thyA and DHFR. Several Mycobacterium and Corynebacterium species are unusual in that they possess thyA and thyX as well as a DHFR encoded by dfrA (Myllykallio et al., 2002). Thus, mycobacteria are probably capable of synthesizing thymidine using the thyX-encoded thymidylate synthase despite disruptions in thyA, which could account for their thymidine prototrophy. In fact, the residual consumption of THF seen in PAS-resistant BCG (Fig. 2) might be accounted for by ThyX activity. In order to remain virulent, bacteria may require at least some activity from both thymidylate synthases. This might represent the different utilization of coenzymes. The ThyA/DHFR systems use NADPH for reduction, whereas the thyX-encoded enzyme appears to use flavin nucleotides (Myllykallio et al., 2002). It is possible that the abundance of these coenzymes varies during different growth stages and, thus, both enzyme activities are required. Interestingly, recent data from our laboratory indicate that insertion mutations in thyX in either M. bovis BCG or M. tuberculosis do not appear to be compatible with growth in vitro, suggesting that thyX mutants are essential for in vitro growth (Sassetti et al., 2003). Further, all PAS-resistant and PAS-sensitive clinical isolates had wild-type thyX sequence (data not shown), indicating that mutations in thyX did not account for PAS resistance in these strains.

With the alarming rise of multidrug-resistant tuberculosis, there has been an increasing interest in developing new chemotherapeutics and resurrecting older ones (Dye et al., 2002). Thus, drugs such as PAS are being used increasingly frequently, and several new antifolate agents are under investigation. Our data suggest that, in mycobacteria, thyA mutations represent a pathway for the development of resistance not only to PAS but also to other drugs that act in the folate biosynthesis pathway.

Experimental procedures

Bacterial strains

Mycobacterium bovis BCG strains Pasteur were grown in Middlebrook 7H9 broth (Difco) supplemented with albumin, dextrose (Difco), 0.5% glycerol, 0.05% Tween-80. The isogenic strains E. coli V55 (thyA+) and V55K6 (thyA) were a gift from J. Blum (Harvard Medical School, Boston, MA, USA). E. coli strains were cultured in LB medium or in minimal M9 medium with 1× M9 salts (Bio101), 10 mM CaCl2, 2 mM MgSO4, 0.4% glycerol, 0.4 mM threonine, 0.005% leucine and 0.001% thiamine. Expression was induced with 0.2% arabinose. When required, medium was supplemented with 0.01% thymine. M. tuberculosis clinical isolates were selected from the collection at State Laboratory Institute, Boston, MA, USA, and tested for antibiotic resistance by standard methods (Metchock et al., 1999).

Plasmids and mutants

The construction of the transposon library in M. bovis BCG has been described previously (Sassetti et al., 2001). Approximately 105 colonies from the library were plated onto Middlebrook 7H10 (Difco) plates with 20 µg ml−1 kanamycin and 1 µg ml−1 PAS (Sigma). pBAD24 was a gift from J. Beckwith (Harvard Medical School, Boston, MA, USA).

Cloning of thy

The thyA gene from M. bovis BCG was amplified by polymerase chain reaction (PCR) from genomic DNA using the primers 5′-ATATCATGATGACGCCATACGAGGACCTGC-3′ (creating a BspHI site) and 5′-ACATGCATGCGCGCCTG GTCCTCGGGCAAG-3′ (creating an SphI site). The resulting amplicon was digested with BspHI and SphI and inserted into the NcoI and SphI sites of pBAD24. The mutant thyA genes from PAS2 and PAS5 were amplified using the same primers but, as BspHI cuts within the transposon, the 5′ truncated gene and some transposon sequence were cloned into pBAD24. Wild-type and mutant thyA genes from PAS-sensitive and PAS-resistant M. tuberculosis were amplified using the above primers, digested with BspHI and SphI and inserted into the NcoI and SphI sites of pBAD24. Each construct was transformed into thyA+ and thyAE. coli strains. To construct overexpressing strains, the thyA gene from PAS-sensitive M. tuberculosis or PAS-resistant clinical isolates was amplified by PCR from genomic DNA using the primers 5′- GGCTGGCCACGTGTCAATCG-3′ (creating an MscI site) and 5′-CCGAGCCGGATCCATTGTCGTCG-3′ (creating a BamHI site) and cloned into the BalI and BamHI sites of pMV261 (hyg). The resulting plasmid was transformed into M. bovis BCG and plated on to 7H10 plates containing 50 µg ml−1 hygromycin.

Complementation of thyA E. coli

Wild-type and mutant thyA cloned into pBAD24 were transformed into E. coli V55 thyA+ and V55K6 thyA strains and selected on LB medium containing 50 µg ml−1 ampicillin and 0.01% thymine. Single colonies were picked and plated on to minimal media (M9) alone, M9 + 0.01% thymine, M9 + 0.2% arabinose or M9 + 0.01% thymine + 0.2% arabinose.

Sensitivity to trimethoprim

The various pBAD24-based constructs in thyA+ and thyAE. coli strains were grown overnight in Mueller–Hinton broth containing 50 µg ml−1 ampicillin and 0.01% thymine. All strains were normalized to OD600 of 0.1, and 100 µl of each was added to microtitre wells of a 96-well plate and grown in the presence of 0.2% arabinose for 1 h on a shaker at 37°C. Trimethoprim (Sigma) was then added at 5 µg ml−1 to a total volume of 200 µl per well. Cells were monitored for growth by recording the OD600 at 22 h after the addition of trimethoprim.

Sensitivity to PAS and DHFR inhibitor 8710

Wild-type BCG and thyA mutants were grown in Middlebrook 7H9 broth to mid-log phase and then normalized to OD600 of 0.2. The strains were diluted to OD600 of 0.2, and 150 µl of each culture was grown in microtitre wells (in triplicate) in the presence of no drug or 0.03, 0.1, 0.33, 1.0, 3.0, 9.0 and 27.0 µg ml−1 of PAS for 6 days. The microtitre plate was placed in a shaking humidified chamber at 37°C. O.D600 was recorded each day using a plate reader after carefully mixing the cultures by pipetting 15 times each to avoid clumping. For assaying sensitivity to the DHFR inhibitor 8710 (a gift from R. Reynolds, Southern Research Institute, Birmingham, AL, USA), inhibitor was added at 50 µg ml−1, and the OD600 was measured at day 0, 2 and 6 after addition.

Thymidylate synthase activity

Preparation of extracts.

Wild-type BCG, PAS1, PAS2 and PAS5 strains were grown in 100 ml volumes in roller bottles to an OD600 of 0.8. Cells were centrifuged and the pellets resuspended in 2 ml of 1× TS buffer (25 mM TES, 0.5 mM EDTA, 12.5 mM MgCl2, 3.25 mM formaldehyde, 37.5 mM β-mercaptoethanol, pH 7.4) and placed on ice. Each sample was sonicated four times using a probe sonicator (Branson 250 sonifier) at 1 min intervals. Extracts were centrifuged at 4°C for 15 min at 13 000 r.p.m. The protein content of each supernatant was determined using the Bradford assay (Pierce). Extracts were diluted to a standard protein concentration with 1× TS buffer. Thymidylate synthase activity was measured by mixing 1 mg of extract with 100 µM 5,10-methylene (6R) THF (a gift from R. Moser, Eprova AG Research Institute, Switzerland). The reaction was initiated by the addition of 50 mM dUMP (Sigma). Conversion of 5,10-methylene THF to DHF in the presence of cell extracts and dUMP was monitored every minute at OD340 for 45 min at room temperature (Wahba and Friedkin, 1961). Rate of conversion to DHF was calculated from the slope and normalized to 100% for wild type.


We are grateful to Dr Robert Reynolds, Southern Research Institute, Birmingham, Al, USA, for the DHFR inhibitor 8710, and Dr R. Moser, Eprova AG Research Institute, Switzerland, for kindly providing the reagent 5,10-methylene (6R) THF. We thank Dr Jonathan Blum, Harvard Medical School, for E. coli thyA strains and helpful suggestions. We also thank Dr J.-L. Sankale for help with sequence analysis, and Drs Samantha Sampson, Adrie Steyn, Shruti Jain and Susanne Szabo for their valuable comments and suggestions. This work was supported in part by the Heiser programme of the New York Community Trust (J.R.) and NIH grants AI48704 and AI51929 (E.J.R.).