Essential Protein-Protein Interactions between Plasmodium falciparum Thymidylate Synthase and Dihydrofolate Reductase Domains*

In Plasmodium falciparum, dihydrofolate reductase and thymidylate synthase activities are conferred by a single 70-kDa bifunctional polypeptide (DHFR-TS, dihydrofolate reductase-thymidylate synthase) which assembles into a functional 140-kDa homodimer. In mammals, the two enzymes are smaller distinct molecules encoded on different genes. A 27-kDa amino domain of malarial DHFR-TS is sufficient to provide DHFR activity, but the structural requirements for TS function have not been established. Although the 3′-end of DHFR-TS has high homology to TS sequences from other species, expression of this protein fragment failed to yield active TS enzyme, and it failed to complement TS− Escherichia coli. Unexpectedly, even partial 5′-deletion of full-length DHFR-TS gene abolished TS function on the 3′-end. Thus, it was hypothesized that the amino end of the bifunctional parasite protein plays an important role in TS function. When the 27-kDa amino domain (DHFR) was provided in trans, a previously inactive 40-kDa carboxyl-domain from malarial DHFR-TS regained its TS function. Physical characterization of the “split enzymes” revealed that the 27- and the 40-kDa fragments of DHFR-TS had reassembled into a 140-kDa hybrid complex. Thus, in malarial DHFR-TS, there are physical interactions between the DHFR domain and the TS domain, and these interactions are necessary to obtain a catalytically active TS. Interference with these essential protein-protein interactions could lead to new selective strategies to treat malaria resistant to traditional DHFR-TS inhibitors.

Malaria, caused by the protozoan parasite Plasmodium falciparum, affects about 300 million individuals and causes about 2 million death per year (1). Traditional antimalarial agents such as chloroquine are ineffective in many regions of the world due to drug resistance (2,3). In addition, there is mounting evidence that highly drug-resistant parasite clones acquire resistance to new antimalarial agents at an enhanced rate (4,5). Identification of host-parasite biochemical differences that can lead to selective chemotherapy is more important than ever before.
In malaria as well as many other cell types, chemotherapy targeted at dihydrofolate reductase (DHFR) 1 and thymidylate synthase (TS) has proven to be highly effective (6,7). Even partial inhibition of DHFR or TS can lead to DNA strand fragmentation and cell death (8 -10). In malaria pharmacology, these two enzymes are of particular interest because traditional drugs such as pyrimethamine and cycloguanil are known to inhibit parasite DHFR-TS (11)(12)(13). In recent years, however, the effectiveness of these inhibitors has been compromised by malarial parasite strains expressing mutant forms of DHFR-TS (14 -16). One obvious way to continue selective killing of malarial parasites is to identify new folate analogs directed at DHFR-or TS-active sites that are effective against drug-resistant parasites (17)(18)(19)(20).
Beyond active site-directed strategies, there may be nontraditional opportunities to inhibit malarial DHFR-TS with selectivity. All protozoan parasites, including Plasmodia, have a single bifunctional protein that codes for both DHFR and TS activity (21)(22)(23)(24). In sharp contrast, mammalian cells (as well as bacteriophage, bacteria, and yeast) have separate genes that code for small monofunctional DHFR and TS proteins (25,26). This difference in organization of two well established drug targets represents a dramatic difference in host-parasite biochemistry. Yet, it has not been clear whether the bifunctional status of DHFR-TS can play a role in selective chemotherapy. First, all indications are that the active sites of DHFR and TS in these bifunctional proteins operate independently of each other. Inhibition of DHFR with methotrexate does not influence TS activity, and inhibition of TS function with 5-fluoro-2Ј-deoxyuridylate does not effect DHFR activity (27). The structure of Leishmania DHFR-TS suggests that the active sites of DHFR-TS are about 40 Å apart (28). Designing a "doublebarrel," parasite-specific bifunctional DHFR-TS inhibitor is not a practical goal at present. Second, kinetic measurements suggest that Leishmania and Toxoplasma TS may channel as many as 8 out of 10 dihydrofolate (DHF) molecules directly from the TS site to the DHFR site (27,29,30). Indeed, the Leishmania DHFR-TS structure indicates that this direct channeling may be attributed to an electropotential gradient in the bifunctional protein that leads from the TS folate-binding site to the DHFR folate-binding site (28,29). However, it is highly unlikely that blocking such channeling will lead to selective antimalarial chemotherapy since malarial DHFR-TS active sites do not have any trouble delivering or accepting substrates through bulk solvent in vitro (31,32) or in vivo (33).
In this study, we demonstrate that species-specific interactions between malarial DHFR domain and TS domain are essential for functioning of these enzymes. Inhibition of such protein-protein interactions may offer a powerful new strategy * This work was supported in part by United States Public Health Service Grants AI26912 and AI40956 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  for selective chemotherapy against malaria and possibly other protozoan parasites.
Plasmids-Plasmid pKK233.2 with a P. falciparum DHFR-TS coding sequence of clone K1 was obtained from John Hyde (University of Manchester) and maintained as described previously (34). Plasmids pET23d(ϩ) was from from Novagen (Madison, WI). This plasmid was cut with NcoI and HindIII and modified to express different domains of P. falciparum DHFR-TS (Fig. 1). pET-DHFR was designed to express amino acids 1-230 of the P. falciparum DHFR-TS coding region (32,34); pET-TS31 expressed amino acids 325-607 of the the P. falciparum DHFR-TS coding region; pET-TS40 coded for amino acids 249 -607 of the P. falciparum DHFR-TS coding region, and pET-DHFR-TS contained the full-length coding region (Fig. 1).
Antibody to P. falciparum DHFR Domain-The antigen, malarial DHFR domain, was expressed and purified on a methotrexate-affinity column as described previously (32). Rabbit polyclonal antibodies to this pure protein were raised by Josman Laboratories (Napa, CA). P. falciparum DHFR-specific antibody was purified by affinity chromatography using P. falciparum DHFR domain attached to Aminolink Plus from Pierce.
Genetic Complementation Assay for TS Function-The thymidylate synthase-deficient E. coli strain Rue10 was lysogenized with DE3 to provide T7 RNA polymerase function to this host (36). The host cells were maintained in LB medium in the presence of leucine (40 g/ml), proline (200 g/ml), and thymidine (100 g/ml). When testing for TS function in pET plasmids bearing malarial DHFR-TS gene, or portions of this gene, 50 ng of plasmid was electroporated into about 4 ϫ 10 7 electrocompetent Rue10(DE3) cells. The cells were spread on two sets of ampicillin agar plates, those with thymidine and those lacking thymidine.
Genetic Complementation Assay for Malarial DHFR Function-Appropriate pET plasmids were electroporated into BL21(DE3) cells as above. These cells were spread on two sets of ampicillin agar plates, those with trimethoprim (50 g/ml) and those without. Bacterial DHFR is sensitive to trimethoprim but malarial DHFR is not (34).
Exonuclease Treatment of DNA Fragment Coding for Malarial DHFR-TS-The 5Ј-deletion experiment involved the following four steps prior to testing for TS function: (i) generation of a DNA strand with malarial DHFR-TS sequence that was susceptible to exonuclease on the 5Ј-end but not the 3Ј-end; (ii) digestion of the long fragment with exonuclease for varying amounts of time to generate a series of short DNA fragments of varying length; (iii) preparation of a recipient vector with an AUG start codon; and (iv) ligation of the short partially digested fragments into the recipient vector.
Initially, 12 g of a pET23d(ϩ) plasmid with full-length malarial DHFR-TS insert was digested with 24 units of PvuII and 24 units of PstI in a total volume of 40 l for 6 h. The purified 3900-base pair fragment containing the DHFR-TS sequence was susceptible to exonuclease III on the PvuII end but not the PstI end. To initiate digestion on the 5Ј-end, 6 g of DNA was placed in 60 l of ExoIII buffer (Erase-abase kit, Promega, Madison, WI). A 2.5-l aliquot was removed at time 0. Digestion was started with 400 units of exonuclease III delivered in 2.3-l volume. Under these conditions, digestion occurred at a rate of 80 -90 base pairs per min at 22-25°C. Aliquotes of 2.5 l were removed every 2.5 min for 60 min. These were treated with S1 nuclease and blunt-ended with Klenow as described (Erase-a-base kit, Promega, Madison, WI). A cohesive end on the the 3Ј-end of the DHFR-TS coding region was generated by treating each of the 2.5-min samples with HindIII.
A recipient plasmid for expressing the exonuclease-treated DNA was generated by first cutting 5 g of pET23d(ϩ) with 6 units of NcoI. The NcoI-digested plasmid was blunt-ended with Klenow (10 units, room temperature, 14 min) and then digested with HindIII (24 units, 37°C, 4 h) to generate a cohesive site on the other end of the vector. Approximately 760 ng of exonuclease-digested DNA (5 l) was ligated to 312 ng of recipient vector. The ligation products from each 2.5-min time point were then transformed into electrocompetent Rue10(DE3) cells and plated without thymidine to measure TS function, or the ligated DNA was transformed into electrocompetent BL21(DE3) cells and plated on trimethoprim plates to measure DHFR function (see above).
Co-transformation of E. coli Rue10(DE3) with Two Plasmids Coding for a DHFR Fragment and a TS Fragment-In the first step, pET-DHFR was transformed into Rue10(DE3) and selected for ampicillin resistance. These Rue10(DE3)pET-DHFR were made electrocompetent, transformed with pET-TS40, and selected for growth on medium lacking thymidine. Subsequently, it was found that double transformants could also be obtained by first transforming cells with pET-TS40 (selection on ampicillin plates) and then transforming them with pET-DHFR (selection on medium without thymidine).
Physical Characterization of P. falciparum TS40 ϩ DHFR Split Protein-E. coli Rue10(DE3), co-transformed with pET-DHFR and pET-TS40, were freshly grown to an absorbance of 1.0 at 600 nm. The cells were collected by centrifugation and resuspended in 50 mM Tris-HCl, pH 8, 2 mM EDTA, 0.1% Triton X-100, and 0.1 mg/ml lysozyme (final lysis volume was 1/10 of the original culture volume). The cells were incubated at 30°C for 15 min, sonicated, and then stored in 1-ml aliquots at Ϫ20°C.
To estimate the native size of the DHFR and TS activities by gel filtration chromatography, 200 l of cell lysate from above was applied to a 1 ϫ 50 cm Superdex 200 column (Amersham Pharmacia Biotech). The buffer for gel filtration was 50 mM Tris-HCl, pH 8, 2 mM EDTA, 10% glycerol, 0.1 M NaCl, 2 mM dithiothreitol. The flow rate was 0.25 ml/min, and 0.5-ml fractions were collected.
The split protein was purified to homogeneity using methotrexate affinity chromatography. A total of 10 ml of cell lysate from the doubletransformed E. coli Rue10(DE3) was applied to a 2-ml methotrexateagarose column pre-equilibrated with 50 mM Tris-HCl, pH 8, 2 mM EDTA, 10% glycerol, 0.5 M NaCl, and 2 mM dithiothreitol. The column was washed with 20 ml of equilibration buffer and then the active fractions were eluted with the same buffer plus 4 mM dihydrofolate. One-ml fractions were collected.
Assay of TS Activity and Malarial DHFR Protein-After each chromatography step, malarial TS activity was monitored by assaying 10 l of each fraction using a 5-[ 3 H]-2Јdeoxyuridylate tritium release assay (34). Malarial DHFR activity was monitored by Western blotting of each fraction. A 10-l sample from each tube was applied to a well of a 10% SDS-polyacrylamide gel electrophoresis system. After separation, the protein was transferred to a polyvinylidene difluoride membrane. Malarial DHFR was detected using affinity purified polyclonal antibodies to malarial DHFR domain (see above), goat anti-rabbit secondary antibodies, and a Western-Star chemiluminescent detection system (Tropix, Inc., Bedford, MA). This assay easily detected 1 ng of malarial DHFR.

RESULTS AND DISCUSSION
The Carboxyl End of Malarial DHFR-TS Is Insufficient for TS Function-The amino-terminal DHFR domain of malarial DHFR-TS can be expressed in functional form (32,34,37,38), but all attempts to express the carboxyl end TS domain in a catalytically active form have been unsuccessful. A genetic system was used to assay for TS function from the malarial DHFR-TS gene. E. coli cells lacking TS function (strain Rue10) normally proliferate in the presence of thymidine but not in the absence of thymidine (39). Lysates prepared from Rue10 cells have DHFR enzyme activity but not TS activity (data not shown). When Rue10 cells are transformed with an expression plasmid coding for full-length malarial DHFR-TS gene (Fig.  1A), E. coli Rue10 proliferate even in the absence of thymidine (Fig. 1B). Transformation with plasmids coding for a minimal 30-kDa "TS domain" (Fig. 1A) or a 40-kDa protein coding for most of the joining region and all of the TS domain (Fig. 1A) failed to rescue Rue10 from thymidine dependence (Fig. 1B). These experiments suggested that the carboxyl end TS domain, defined by sequence homology to other monofunctional TS proteins, was not sufficient to express a functional TS protein.
Partial 5Ј-Deletion of Malarial DHFR-TS Abolishes TS Function-In order to define the minimum 5Ј-end of the functional TS domain, full-length malarial DHFR-TS gene was subjected to unidirectional exonuclease III treatment for varying durations ( Fig. 2A). The resulting fragments were blunt-ended and cloned into a recipient expression vector equipped with its own blunt-ended AUG start codon. The ligation mixes from each time point were transformed into E. coli Rue10(DE3) and scored for TS function (colonies growing without thymidine) and for malarial DHFR function (colonies growing in the presence of trimethoprim). The results showed the following: (i) DHFR and TS function were lost simultaneously (Fig. 2B); (ii) every colony that survived on plates without thymidine, when replica plated, had both malarial DHFR function and TS function (data not shown); and (iii) DNA sequencing of inserts from every TS ϩ E. coli showed an intact malarial DHFR-TS gene (data not shown). Control experiments confirmed that the repeated recovery of just the full-length DHFR-TS was not due to inability to digest the gene or high processivity of the exonuclease. In the early part of the 5Ј-digestion experiment, degradation of the malarial DHFR-TS gene could be visualized directly on agarose gel (data not shown). Intermediate sized fragments of appropriate length were observed when the DHFR-TS had been treated with exonuclease for 5-20 min (see schematic Fig. 2A). Additionally, selection of transformants in the presence of thymidine allowed recovery of inserts with incomplete DHFR-TS fragments (data not shown). These experiments were consistent with the hypothesis that the "carboxyl end TS domain" in malarial parasites was not adequate to obtain a functional TS enzyme. Furthermore, the exonuclease experiment suggested that the amino end of malarial DHFR-TS was necessary for TS function.
DHFR Domain Can Activate TS Domain in Trans-To test directly the contributions of malarial DHFR domain toward TS function, E. coli Rue10 transformed with the 40-kDa carboxyl TS domain were supplemented with the 27-kDa DHFR domain in trans and selected on medium lacking thymidine. The results showed that simultaneous transformation of these ThyA Ϫ E. coli with an expression vector coding for the 40-kDa TS domain and an expression vector coding for the amino end DHFR fully rescued the TS defect (Fig. 3). The two plasmids could be introduced in either order (data not shown). Polymer- ase chain reaction analysis demonstrated that TS function was truly restored by genes acting in trans. There were two distinct plasmids in the rescued E. coli, and there was no evidence for a new bifunctional gene created through recombination between the plasmids (data not shown). Direct biochemical TS assay on cell lysates showed that, as expected, only the double transformed E. coli that proliferated without thymidine expressed active TS enzyme (data not shown). These experiments demonstrated unambiguously that the amino-terminal DHFR domain of malarial DHFR-TS was necessary for the carboxylterminal TS sequence to gain TS function. Since the TS31 domain could not be rescued by the DHFR domain (data not shown), the "joining region" must make significant contacts with the malarial DHFR domain. The requirement for DHFR domain was species-specific since endogenous E. coli DHFR could not activate malarial TS sequence expressed in the bacterial host.
Physical Associations within the Split Protein-There were two general ways in which the malarial DHFR domain could be activating TS function in the TS domain: (i) the DHFR domain could have acted as a transient chaperon during the folding of the TS sequence, or (ii) the DHFR domain could have formed a stable association with the carboxyl-terminal TS. To determine if there was stable association between the DHFR and TS domains of the split protein, the physical characteristics of the active proteins from the double transformed E. coli were studied. First, cell lysate from E. coli Rue10(DE3) pET-DHFR, pET-TS40, was subjected to gel filtration chromatography. TS activity and malarial DHFR emerged as a single superimposing peak (Fig. 4A). The size of the TS ϩ DHFR complex was 140,000 daltons. This was consistent with the hypothesis that the DHFR and TS fragments had assembled into a tetramer made up of two 40-kDa TS units and two 27-kDa DHFR units, thereby reestablishing the natural relationship between the domains. All reasonable alternative structures are inconsistent with the gel filtration data. A free DHFR domain with a mass of 27 kDa was not seen on the gel filtration column. We also did not see a DHFR-free obligate TS dimer with a mass of 80 kDa.
Since the TS ϩ DHFR split protein formed a complex stable enough to survive gel filtration chromatography, we chose to purify the complex to homogeneity and study its physical characteristics. Cell lysates from the double-transformed E. coli Rue10(DE3) were passed through a methotrexate affinity column, which is known to bind malarial DHFR domain (32). All unbound protein was eluted with 500 mM salt, and specifically bound protein was eluted with 4 mM dihydrofolate (Fig. 4B). Both the TS activity and the malarial DHFR activities eluted simultaneously in the 4 mM dihydrofolate fractions. The active fractions were concentrated and subjected to SDS-polyacrylamide gel electrophoresis (Fig. 4C). Coomassie Blue staining revealed two bands, one with a mass of 40 kDa and one with a mass of 27 kDa. Careful densitometry revealed that the 27-and the 40-kDa bands were present in a stoichiometric ratio of 1 to 0.97, after compensating for size differences. The larger 40-kDa band could be labeling with radioactive 5-fluoro-2Ј-deoxyuridy- late as one would predict for the catalytically active TS domain (Fig. 4C). The smaller band, 27-kDa, was shown to be malarial DHFR domain by Western blotting (Fig. 4C). The final purified material had a DHFR-specific activity of 39.2 mol/min/mg of total protein. Assuming a k cat of about 1800/min for DHFR, this also was consistent with a protein having two 27-kDa DHFR subunits attached to two 40-kDa TS subunits. In sum, the physical characteristics of the purified split protein support the gel filtration studies and are consistent with the model where two DHFR and two TS domains assemble to form a 140-kDa tetramer. These experiments show that the interactions between the DHFR domain and the TS domain that lead to activation of TS function are not transient; the split malarial DHFR domain and the 40-kDa domain reassemble to form a stable complex that reflects the natural state of the bifunctional protein (Fig. 5).
The present study uses a combination of genetic and biochemical approaches to demonstrate that species-specific interactions between malarial DHFR domain and TS domain are essential for TS function.
Although a malarial DHFR-TS structure is not available, the crystal structure of Leishmania major DHFR-TS bifunctional protein offers one obvious clue on how the amino end of the DHFR domain may act as a determinant of TS function. The Leishmania enzyme has a 22-amino acid extension on the amino terminus of the DHFR domain which encircles the attached TS domain (28). At the very least, this polypeptide extension from the amino end of the DHFR domain appears necessary to stabilize the tertiary structure of the TS domain, but it is also conceivable that some residues from this extended polypeptide contribute directly to the formation of the TSactive site. Whereas the malarial DHFR has a shorter extension on the amino end (21), it is conceivable that it too directly interacts with residues of the TS domain and is necessary to form a catalytically active TS enzyme. Of course, it is also possible that, unlike the Leishmania bifunctional protein, the amino end of the malarial DHFR domain is necessary primarily to stabilize the DHFR tertiary structure and this, indirectly, influences correct folding and activity of the malarial TS domain.
Regardless of the mechanistic details underlying activation of the TS domain by the amino end of the DHFR domain, inhibition of these species-specific protein-protein interactions may offer a powerful new strategy for selective chemotherapy against malaria and possibly other protozoan parasites. Small molecular weight peptidomimetics that disrupt the interactions between the domains but fail to activate TS are expected to inhibit malarial TS with selectivity because human TS does not require "activation" by an accessory protein. There is precedence for drug development inspired by species-specific protein-protein interactions. For instance, virus-specific interactions between the two subunits of ribonucleotide reductase from herpes simplex virus led to selective peptidomimetic inhibitors that were effective at nanomolar concentrations (40). In addition, inhibition of protein-protein interactions is being considered in many systems including dimerization of HIV proteases, assembly of herpes simplex virus DNA polymerase subunits, interactions involving G-proteins and signal transduction machinery, and binding of adhesion molecules to surface proteins of platelets (41).
The present E. coli genetic assay for malarial TS function is expected to play a powerful role in future mutagenesis experiments that will help establish contact sites between the two malarial protein domains. In addition, the present genetic system will also serve well in high throughput screens to identify selective inhibitors of malarial TS function. Finally, the present "two-hybrid" genetic system provides a novel means for maintaining two plasmids indefinitely without any selectable drug marker.