Structure, Vol. 12, 1705–1717, September, 2004, 2004 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.str.2004.07.011
Crystal Structure of 7,8-Dihydropteroate Synthase from Bacillus anthracis: Mechanism and Novel Inhibitor Design Kerim Babaoglu,1,2 Jianjun Qi,2 Richard E. Lee,2,* and Stephen W. White1,3,* 1 Department of Structural Biology St. Jude Children’s Research Hospital Memphis, Tennessee 38105 2 Department of Pharmaceutical Sciences 3 Department of Molecular Sciences University of Tennessee Health Science Center Memphis, Tennessee 38163
Summary Dihydropterate synthase (DHPS) is the target for the sulfonamide class of antibiotics, but increasing resistance has encouraged the development of new therapeutic agents against this enzyme. One approach is to identify molecules that occupy the pterin binding pocket which is distinct from the pABA binding pocket that binds sulfonamides. Toward this goal, we present five crystal structures of DHPS from Bacillus anthracis, a well-documented bioterrorism agent. Three DHPS structures are already known, but our B. anthracis structures provide new insights into the enzyme mechanism. We show how an arginine side chain mimics the pterin ring in binding within the pterin binding pocket. The structures of two substrate analog complexes and the first structure of a DHPS-product complex offer new insights into the catalytic mechanism and the architecture of the pABA binding pocket. Finally, as an initial step in the development of pterinbased inhibitors, we present the structure of DHPS complexed with 5-nitro-6-methylamino-isocytosine. Introduction The rapid spread of HIV, the emergence of new pathogenic organisms such as SARS, and the expanding problem of drug-resistant bacterial infections, notably MDR Mycobacterium tuberculosis and vancomycinresistant Staphylococcus aureus, have forced a reevaluation of our current risk from infectious agents (http:// www.niaid.nih.gov/factsheets/antimicro.htm). In addition, the threat of bioterrorism using agents such as weaponized Bacillus anthracis and Yersinia pestis highlight the need for continuing research in infectious diseases and the search for new therapeutic agents (Greenfield and Bronze, 2003). One approach to the development of new antimicrobial agents is to identify novel targets that are essential for the viability of the infectious agent, but which are nonessential or absent in higher organisms. This approach is now being widely adopted as an increasing number of genomes from pathogenic organisms are becoming available. However, despite major investments in genomic technologies, relatively few promising drug candidates have *Correspondence:
[email protected] (S.W.W.); relee@ utmem.edu (R.E.L.)
been developed using this approach. A second approach is to revisit well-validated targets for which antimicrobial agents have already been successfully developed, and to design new agents using high-resolution structural information and structure-assisted drug discovery techniques. One such well-characterized target is the bacterial folate pathway, and more specifically, the enzyme dihydropterate synthase (DHPS) that catalyzes a key step in the pathway. All organisms require tetrahydrofolate for metabolic reactions that involve one-carbon transfer, but only prokaryotes, some lower eukaryotes, and plants contain the enzymes for the complete folate biosynthetic pathway. Higher eukaryotes instead rely on high-affinity transport systems for the uptake of folate or its reduced derivatives that are present in their diet (Matherly, 2001). Almost all eubacteria, with the exception of Lactobacilli and E. faecalis, lack transport systems for folate (Kumar et al., 1987) and must synthesize tetrahydrofolate de novo. DHPS is encoded by the folP gene, and it catalyzes the formation of the folate intermediate 7,8-dihydropteroate from p-aminobenzoic acid (pABA) and 6-hydroxymethyl-7,8-dihydropterin-pyrophosphate (DHPP) (Figure 1). DHPS has been a bona fide drug target for almost 70 years (Domagk, 1935) and is the target for the highly successful sulfonamide compounds (sulfa drugs) that act as competitive inhibitors and dead-end substrate analogs of pABA (Brown, 1962; Roland et al., 1979; Anand, 1996). A variety of sulfa drugs with differing pharmacokinetic properties have been developed to treat respiratory infections, meningococcal disease, urinary tract infections, infections by streptococci and staphylococci, leprosy, toxoplasmosis, and malaria by chloroquinine-resistant Plasmodium falciparum (Anand, 1996). In addition, they are still the drugs of choice used in combination with the dihydrofolate reductase inhibitor trimethoprim for treating Pneumocystis carinii infections in immunocompromised HIV and cancer patients (Hughes, 1988). Drug resistance has emerged as an important factor that severely limits the current clinical use of sulfa-based inhibitors of DHPS (Sko¨ld, 2000, 2001). Mutations in the folP gene that confer sulfonamide resistance have now been characterized in clinical isolates of many pathogenic organisms including Streptococcus pneumoniae (Haasum et al., 2001), Streptococcus pyogenes (Swedberg et al., 1998), Neisseria meningitidis (Ferme´r and Swedberg, 1997), Haemophilus influenzae (Enne et al., 2002), Mycobacterium leprae (Kai et al., 1999), Campylobacter jejuni (Gibreel and Sko¨ld, 1999), Plasmodium falciparum (Wang et al., 1997), Toxoplasma gondii (Aspinall et al., 2002), and Pneumocystis carinii (Ma and Kovacs, 2001). Crystal structures of DHPS have been determined from three organisms, Escherichia coli (Achari et al., 1997), Staphylococcus aureus (Hampele et al., 1997), and Mycobacterium tuberculosis (Baca et al., 2000). Although poorly characterized, the pABA binding site is apparently formed from residues that reside in flexible loop regions, and the mutations that confer sulfa-drug
Structure 1706
Figure 1. The Chemical Structures of DHPS Binding Small Molecules Shown at the top are the natural DHPS substrates and product in the reaction scheme. Shown below are the other small molecules described in this study. DHPP, 6-hydroxymethyl-7,8-dihydropterin-pyrophosphate; pABA, para-aminobenzoic acid; PtPP, 6-hydroxymethyl-pterine-pyrophosphate; PtP, 6-hydroxymethyl-pterine-monophosphate; MANIC, 6-methylamino-5-nitroisocytosine; SMX, sulfamethoxazole.
resistance are almost exclusively contained within these loops (Baca et al., 2000). It appears that the mutations are able to disrupt the binding of pABA drug analogs without seriously compromising the catalytic viability or the structural integrity of the protein. Surprisingly, despite the structure determination of three DHPS molecules and a variety of substrate complexes, the catalytic mechanism of the enzyme is still largely unresolved. It is clear that the mechanism involves conformational changes at the active site involving the surrounding loop regions, but the nature of these movements and the identity and roles of the active site residues within the loops have yet to be characterized. However, the same is not true for the binding site of the pterin substrate that is physically distinct from the pABA binding site, and deeply buried in a highly conserved region of the protein. There have been few attempts to develop lead inhibitory compounds directed against the pterin binding site even though such compounds have been shown to be effective inhibitors (Lever et al., 1985, 1986). To address these issues, we have determined the crystal structure of DHPS from B. anthracis. Here we present the structure of the unliganded enzyme as well as the structures of four complexes with small molecules that probe the two substrate binding pockets in the enzyme. We demonstrate how one of the loops may control the progression of the catalytic cycle and also stabilize the unliganded form of the enzyme. We have also been successful in determining the structure of a product analog complex that provides the clearest view to date of the mode of binding of pABA and which allows us to formulate a model of the transition state complex. Finally, we show how a nonphosphorylated pterinbased inhibitor of DHPS is able to specifically bind within the pterin binding pocket. Results Quality of the Structures We present here five crystal structures of B. anthracis DHPS. Analyses of the structures using PROCHECK
(Laskowski et al., 1993) showed that each has excellent stereochemistry and final R factors that are appropriate for the various resolutions. Pertinent statistics are shown in Tables 1 and 2. Our “best” structure is that of DHPS-MANIC, which has been determined at 1.8 A˚ with Rwork and Rtest values of 0.182 and 0.209, respectively, and this serves as a basis by which to judge the accuracy of the other structures. There are no significant differences apart from the conformations of the loops and associated residues that encompass the active site and the substrate binding regions. The asymmetric unit contains two monomers of the DHPS molecule, and each generates a biological dimer by the action of crystallographic 2-fold axes. Therefore, each structure that we have determined provides two independent views of the DHPS dimer for analysis and comparison. There is no electron density for the N-terminal methionine, the N-terminal 6⫻His tag, or the final three C-terminal residues in any of the structures. Description of the Overall Structure The overall fold of the B. anthracis DHPS molecule is very similar to those of the three orthologs that have been analyzed from E. coli (Achari et al., 1997), S. aureus (Hampele et al., 1997), and M. tuberculosis (Baca et al., 2000). Excluding loop structures that display variable conformations as described below, the core structure of 200 residues can be superimposed with rmsd values of 1.1, 1.25, and 1.8 A˚, respectively. The basic fold of the protein (Figures 2A and 2B) is a classic (/␣)8 TIM barrel (Banner et al., 1975) in which repeating /␣ units create an eight-stranded parallel -barrel surrounded by eight ␣ helices. Shown in Figure 3 are the aligned sequences of the four DHPS orthologs for which there are crystal structures. In our DHPS structure from B. anthracis, the  strands of the barrel (1–8) comprise residues 21–28, 58–62, 98–101, 118–121, 142–145, 180– 185, 213–217, and 252–255, and the associated eight ␣ helices (␣1–␣8) comprise residues 38–55, 75–94, 105– 115, 130–139, 155–174, 193–210, 234–249, and 258–273. Helix ␣6 within the motif is actually made up of two
Structure of DHPS from B. anthracis 1707
Table 1. Statistics of Data Collection Parameter (Unit)
Unliganded
PtPP
PtP
Pteroic Acid
MANIC
Space group Unit cell dimensions (A˚) (a, b, c) Resolution range (A˚)a Rsyma,b I/a Completeness (%) Redundancy Reflections Unique reflections
P6222 97.4, 97.4, 263.6
P6222 98.2, 98.2, 262.6
P6222 97.4, 97.4, 262.9
P6222 97.5, 97.5, 263.2
P6222 98.4, 98.4, 263.2
30.0–2.0 (2.07–2.0) 0.079 (0.293) 26.1 (3.4) 97.4 (94.8) 7.6 (4.0) 361,909 47,517
50.0–2.5 (2.59–2.5) 0.059 (0.297) 42.7 (3.7) 96.9 (81.7) 12.9 (7.7) 337,727 26,269
50.0–2.75 (2.85–2.75) 0.057 (0.324) 57.3 (7.5) 98.4 (99.9) 11.5 (12) 228,333 19,781
81.7–2.15 (2.21–2.15) 0.066 (0.577) 18.3 (2.0) 95.1 (94.4) 8.0 (4.2) 310,524 38,829
81.7–1.83 (1.88–1.83) 0.071 (0.273) 23.6(3.9) 93.8 (68.9) 11.1 (4.8) 699,369 63,046
a b
Values in parentheses refer to the highest resolution shell. Rsym ⫽ ⌺⌺|Ii ⫺ Im|/⌺⌺Ii, where Ii is the intensity of the measured reflection and Im is the mean intensity of all symmetry-related reflections
separate helices, with residues 193–202 forming a typical ␣ helix and residues 203–210 forming a 310 helix. There are two additional secondary structure elements worth noting. A short two-turn ␣ helix (␣-Loop7; residues 220–228) is inserted between the  strand and the ␣ helix of the seventh ␣/ TIM barrel element. Also, there is a short  ribbon structure with a terminal -hairpin close to the N terminus (residues 7–14) that serves as the “floor” at the N-terminal “pole” of the -barrel. The two  strands (N1 and N2) have alternating polar and hydrophobic/aromatic residues, with the former exposed and the latter facing the hydrophobic interior of the -barrel. This -ribbon appears to play a role in stabilizing the TIM barrel motif, but it is not a conserved feature of the DHPS fold. It is present in the E. coli structure but absent from the M. tuberculosis and S. aureus structures. Similar to the three orthologous structures, the B. anthracis DHPS molecule is organized as a dimer (Figure 2). The dimerization surface is relatively small and involves a reciprocal interaction of helices ␣7 and ␣8. The interaction is primarily driven by hydrophobic interactions, but the actual component residues are not well conserved. In the B. anthracis structure, the interface is characterized by a grouping of sulfur atoms and involves Ala240, Cys243, Leu244, Met264, and Met268 (Figure 2C). In addition, Arg263 extends across the dimer inter-
face to form salt bridge interactions with Glu233⬘ and Glu236⬘ (the prime indicates the opposite monomer). It has been noted that the alignment of the two monomers in the dimer is somewhat variable (Baca et al., 2000), and this probably results from the poor conservation of this interface. As noted in the three orthologous structures (Achari et al., 1997; Hampele et al., 1997; Baca et al., 2000), the connections between the ␣/ TIM barrel elements at the N-terminal pole of the -barrel are very short, but the connections within the ␣/ elements at the C-terminal pole are more extensive and convoluted. In TIM barrel enzymes, the active site is invariably located at the C-terminal pole of the -barrel and created by the surrounding loops. The same is true for DHPS, but the orthologs display significant variations in the loop structures, and there is debate as to their correct functional conformations. A full description of the loop conformations in each of our five structures, and their interactions with the bound small molecules at the active site are presented in detail below. The Native DHPS Structure Loop1 (residues 29–37 between the  strand and the ␣ helix of the first ␣/ element) is highly conserved and is completely visible in monomer A but not in monomer B where there is no electron density for residues 28–33.
Table 2. Statistics of Refinement Parameter
Unliganded
PtPP
PtP
Pteroic Acid
MANIC
Resolution range included (A˚) No. of reflections in working set No. of reflections in test set (5%) No. of protein atoms in ASU No. of cofactor atoms in ASU No. of water molecules in ASU Rwork Rtest Rmsd from ideal stereochemistry Bond lengths (A˚) Bond angles (⬚) Mean B overall (A˚2) Ramachandran plot Most favored region (%) Additionally allowed region (%) Generously allowed (%) Disallowed (%)
87.7–2.00 45,056 2,422 4,188 60 325 0.213 0.255
87.7–2.49 24,847 1,311 4,067 93 162 0.211 0.254
87.7–2.75 18,715 1,004 4,095 76 74 0.226 0.278
87.7–2.15 36,872 1,954 4,080 96 164 0.230 0.279
87.7–1.83 59,753 3,220 4,111 76 399 0.182 0.209
0.022 1.786 26.78
0.020 2.119 21.89
0.021 1.760 31.86
0.018 1.586 37.46
0.024 1.788 19.72
92.4 7.2 0.4 0.0
90.9 8.7 0.4 0.0
91.3 8.7 0.0 0.0
92.6 7.4 0.0 0.0
94.0 5.8 0.2 0.0
Structure 1708
Figure 2. The Structure of Bacillus anthracis DHPS (A and B) Orthogonal views to emphasize the TIM-barrel structure of the protein. ␣ helices are colored orange,  strands and coil regions are colored yellow, and the three additional secondary structural elements outside the TIM-barrel fold are shown in cyan. There are two monomers of DHPS in the crystal asymmetric unit, and the structure shown is the one that contains no disordered loop regions. Extended loops 1 and 2 are labeled, together with the N- and C termini. The location of the pterin binding pocket is denoted by the ball-and-stick representation of Arg68 which is inserted into the pocket in the unliganded form of the enzyme. Helix ␣7 that mediates the dimer interface is labeled. (C) A close up of the dimer interface showing the interacting ␣ helices ␣7, ␣8, ␣7⬘, and ␣8⬘ (the prime refers to the opposite monomer) and the contributing residues (see text for details). The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997).
The structure of loop1 is organized around a salt bridge interaction involving Arg82, Glu41, and Glu65, and is further stabilized by hydrogen-bonding interactions between the side chains of Asp31, Ser34, and Ser38. Despite the high conservation of loop1, its conformation is extremely variable when compared to the three other available DHPS structures. Loop1 is apparently highly mobile, and this would explain why it is not visible in our monomer B structure and also in one of the monomers of the S. aureus structure. Analysis of the crystal packing in our structure suggests that the conformation of loop1 in monomer A is a crystallization artifact and not functionally relevant. Specifically, the highly conserved Phe33 is completely exposed and occupies a cavity on the surface of monomer B from an adjacent molecule in the crystal lattice. Loop2 (residues 63–74) is also conserved and visible in both monomers A and B. It adopts a bent conformation (Figures 2A and 2B) in which the side chain of Arg68 penetrates into the TIM barrel where the terminal guanidinium group forms a -stacking interaction with its counterpart from Arg254 (Figure 4). The distance between the faces of the paired guanidinium groups is 3.5 A˚. Also visible in the electron density map are two water molecules that are involved in bridging hydrogenbonded interactions between the terminal guanidinium nitrogens of Arg68 and the side chains of Asn120 and Asp184. Additional stabilizing interactions include a salt bridge between Glu65 and Arg82, and hydrogen-bonding interactions between the NZ nitrogen of Lys104 and the main chain carbonyl oxygen of Val74, and between the NH2 nitrogen of Arg68 and the OD1 oxygen of Asp101. The amide nitrogen of Gly70 also interacts via two water molecules with the side chain of Asn27. Loop2 is more clearly defined in monomer A than in monomer B, but the conformations in both are essentially identical. This conformation for loop2 has not been observed in any of the three orthologous DHPS protein structures, and we believe that it has a key functional significance (see below).
Loops 3–8 are visible in both monomers of the asymmetric unit and have identical conformations in both, as well as in all of the four complexes. Loop3 is a 3 residue tight turn (residues 102–104) and has very similar conformations in all four of the known structures. The conserved Lys104 has a structural role, forming a salt bridge to Glu79 (conserved) and a number of hydrogen-bonding interactions. Loop4 (residues 122–129) is relatively short and nonconserved and has a slightly variable conformation in the orthologous structures. In the B. anthracis structure, loop4 is anchored by a salt bridge between Lys126, Asp163, and Asp166. Loop5 (residues 146–154) is somewhat shorter than those of E. coli and M. tuberculosis, and similar in length to that of S. aureus. Despite being poorly conserved with a variable structure, this loop is invariably organized around a structural histidine (His146 in B. anthracis) that forms multiple hydrogen bonds, and a buried tyrosine residue (Tyr153 in B. anthracis) that is present in the E. coli and M. tuberculosis proteins but not in the S. aureus protein. Loop5 is further stabilized in our structure by a salt bridge involving Arg148, Asp159, and Asp163. Loop6 (residues 186–192) is highly conserved both in sequence and structure and forms one side of the pterin binding pocket (see below). The N terminus of loop7 (residues 218–233) is also close to the active site and highly conserved, but the remainder of the loop is poorly conserved. It has a similar conformation in all four orthologs and contains the additional ␣-Loop7 helix. The M. tuberculosis molecule has an additional 6 residues immediately after this ␣ helix. Two common features of loop7 are a conserved structural arginine (Arg219 in B. anthracis) and a conserved phenylalanine (Phe222 in B. anthracis replaced by Met221 in E. coli) at the interface with the adjacent loop6. Finally, loop8 (residues 256–257) is another tight 2 residue turn conserved in sequence and structure, and contains key residues: Asp257 that forms a key salt bridge with Arg219, and His256 that is a component of the active site (see below). A total of 12 large spheres of electron density are
Structure of DHPS from B. anthracis 1709
Figure 3. Sequence Alignment of Four DHPS Orthologs Shown is a CLUSTALW sequence alignment of B. anthracis DHPS with the other three orthologs for which crystal structures have been previously determined. The secondary structural elements above the sequences are based on the B. anthracis structure, and these are colored according to the cartoon structure shown in Figure 2. Loops 1–8 described in the text are labeled L1–L8. Residues boxed in solid blue are identical, and residues boxed in white are conserved. The red dots correspond to known sites of mutation that result in sulfonamide resistance. Ba, Bacillus anthracis; Sa, Staphylococcus aureus; Ec, Escherichia coli; Mt, Mycobacterium tuberculosis.
present within the dimers, 6 each in identical locations within the two monomers. Based on their shapes, behavior during refinement and the crystallization conditions in the presence of 1.3 M Li2SO4, these were interpreted as sulfate ions. As described below, one of these sulfate ions is located adjacent to the pterin binding pocket and has a functional significance. This sulfate ion is
coordinated by hydrogen bonds involving the side chains of Asn27, Arg254, and His256 (Figure 4). The DHPS-PtPP and DHPS-PtP Structures PtPP (6-hydroxymethyl-pterine-pyrophosphate) is the oxidized version of the natural DHPS substrate DHPP, and PtP (6-hydroxymethyl-pterine-monophosphate) Figure 4. A Novel Conformation for Loop2 in Bacillus anthracis DHPS This stereoview shows how loop2 in the unliganded enzyme adopts a bent conformation that allows the side chain of Arg68 to be inserted into the pterin binding pocket. Arg68 stacks onto Arg254 and forms indirect hydrogen-bonding interactions with residues Asp184 and Asn120 through two water molecules. The overall ribbon structure of DHPS is colored in gray, helix ␣-Loop7 is colored in cyan, and loop2 is colored in green. Shown in ball-and-stick representation are the key residues at the DHPS active site. Note the sulfate ion that occupies the conserved anion binding site adjacent to the pterin binding pocket. The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997).
Structure 1710
Figure 5. Structures of the Bacillus anthracis DHPS-PtPP and DHPS-PtP Complexes PtPP and PtP are oxidized analogs of the natural substrate DHPP that contain two and one phosphates, respectively. The overall DHPS ribbon structure is colored in gray and ␣-Loop7 is colored in cyan. PtPP and PtP are shown in stick representation, and key DHPS residues are shown as ball-and-stick. (A) The DHPS-PtPP complex. The pterin ring binds at the center of the TIM barrel and makes hydrogen-bonding interactions with Asn120, Asp184, and Lys220. The -phosphate of PtPP is bound within the anion binding pocket and tightly held by hydrogenbonding and charge interactions with Arg254 and His256. Note that Asn27 has rotated away to interact with another bound sulfate ion. There are no discernable contacts between the protein and the ␣-phosphate. (B) The DHPS-PtP complex. The pterin ring is bound in the identical position as seen in the PtPP complex, and the lone ␣-phosphate has moved down to occupy the anion binding pocket. However, although the ␣-phosphate can contact with Arg254, it can no longer reach His256 and its binding within the pocket is suboptimal. The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997).
lacks the -phosphate moiety of PtPP (Figure 1). PtPP is a substrate analog that differs from the actual substrate, DHPP, by the addition of a stabilizing double bond to the pterin B ring. PtPP binds within the pterin binding pocket that has previously been identified in all three of the known orthologous structures (Achari et al., 1997; Hampele et al., 1997; Baca et al., 2000). Overall, the position and conformation of PtPP in our structure is identical to that reported by Hampele and coworkers. However, we reveal for the first time that the pterin ring of PtPP displaces the guanidinium group of Arg68 that occupies the pocket in the unliganded structure, and loop2 thereby becomes disordered. Specifically, residues 67–75 in monomer A and 66–75 in monomer B are no longer visible in the electron density. The pterin ring is deeply buried in a cleft at the center of the TIM barrel directly beneath the flexible loop regions and is involved in a number of interactions (Figure 5A). Hydrogen-bonding interactions involve three side chains, Asn120, Asp184, and Lys220, and a conserved structural water molecule that straddles 6 and 7 and the OD2 oxygen of Asp184. The ND2 and OD1 atoms of Asn120 access the N1 ring nitrogen and the exocyclic amine nitrogen, respectively. The OD1 oxygen of Asp184 interacts with the exocyclic amine nitrogen and the N3 ring nitrogen. The NZ nitrogen of Lys220 bonds with the ketone oxygen on atom C4 as well as the N5 ring nitrogen. Finally, the structural water molecule forms a second interaction with the ketone oxygen on atom C4. Of the two faces of the pterin ring, one face packs against Phe189 and Met145, while the opposite face stacks on the guanidinium group of Arg254. This arginine is firmly held in place by a salt bridge to Asp101. Other residues that line the pterin binding pocket include Ile122, Ile143, and Leu214. All of the residues that specifically interact with the pterin ring are totally conserved (Figure 3). Loop6 provides a significant part of the pocket, and conserved residues Pro185, Gly186, Leu187, and Gly188 (Figure 3) assume a conformation that allows the specific interactions made by Asp184 and Phe189.
Of the two phosphate groups in PtPP, the ␣-phosphate interacts weakly with the protein while the -phosphate occupies the tight sulfate binding site described above that is directly adjacent to the pterin binding pocket (Figure 4). It forms essentially identical interactions with Arg254 and His256, but the interaction with Asn27 is lost as this residue and the neighboring Val28 within loop1 rotate around by ⵑ180⬚ and interact with a different bound sulfate molecule (Figure 5A). In the structure of the S. aureus DHPS-PtPP complex reported by Hampele and coworkers, Asn27 forms part of a divalent metal ion coordination shell that includes the pyrophosphate group. An equivalent divalent metal ion is not present in our structure, and the rotation of Asn27 and Val28 is therefore likely to be a crystal artifact. The DHPS-PtP complex has a very similar structure to that of DHPS-PtPP (Figure 5B). The pterin rings are virtually superimposable, but the single ␣-phosphate has shifted down to occupy the sulfate/-phosphate binding site described above. However, the ␣-phosphate is shifted relative to the sulfate/-phosphate position, and the hydrogen-bonding interactions with Arg254 and His256 are suboptimal. Baca and coworkers also analyzed the structure of a DHPS-PtP complex in their M. tuberculosis study. The position and conformation of PtP are similar to those described here, but a divalent cation is coordinated to the ␣-phosphate, and this is not present in our structure. This difference is also probably related to the rotation of the coordinating side chain of Asn27. The DHPS-Product Analog Complex A key structure that we have determined, and one that has not been previously reported, is the complex between DHPS and pteroic acid (Figure 6). Pteroic acid is a product analog that differs from the actual product, 7,8-dihydropteroate, by the addition of a stabilizing double bond to the pterin B ring (Figure 1). In both monomers of the asymmetric unit, there is a single product molecule bound in the active site, and residues 65–73 in
Structure of DHPS from B. anthracis 1711
Figure 6. The Structure of Pteroic Acid Bound to Bacillus anthracis DHPS Pteroic acid is an oxidized analog of the normal DHPS product, 7,8-dihydropteroate. (A) Stereoview of the complex. The overall DHPS ribbon structure is colored in gray and helix ␣-Loop7 is colored in cyan. Pteroic acid is shown in stick representation, and key DHPS residues are shown as ball-and-stick. The pterin ring binds within the pterin binding pocket in exactly the same manner as seen in the PtPP and PtP complexes (Figure 5). The region of the pteroic acid that corresponds to the pABA moiety of the product sits against the hydrophobic acyl chain of Lys220. The nitrogen that corresponds to the free amine in pABA is opposite the phosphate binding site. The terminal carboxylic acid is hydrogen bonded by the backbone nitrogen and side chain oxygen of Ser221 located at the N-terminal end of helix ␣-Loop7. The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997). (B) Final 2Fo ⫺ Fc electron density map, contoured at 1 , calculated around the pteroic acid and the adjacent sulfate ion. The figure was produced using BOBSCRIPT (Esnouf, 1999) and rendered with RASTER3D (Merritt and Bacon, 1997).
loop2 have become disordered. As seen by comparing Figures 5 and 6, the pterin moiety of the pteroic acid is bound in exactly the same manner as the pterin in the PtPP and PtP complexes, with the same hydrogenbonding pattern involving Asn120, Asp184, and Lys220, and the -stacking interaction with Arg254. The completely novel aspect of the product structure is the location of the pABA moiety of the pteroic acid. The benzene ring of the pteroic acid sits against the hydrophobic acyl region of the side chain of Lys220 which has moved into the appropriate location compared to the unliganded structure to mediate this interaction. The terminal carboxylate group forms hydrogen-bonding interactions with the OG oxygen and amide nitrogen of Ser221 located at the N terminus of helix ␣-Loop7. No chargecharge interactions are made with the terminal carboxylate group on the product analog. The binding of pteroic acid does not result in any additional ordering of loops 1 and 2 since their position and visibility are very similar to those seen in the PtPP structure. The DHPS-MANIC Complex The molecule 6-(methylamino)-5-nitroisocytosine (MANIC) was originally discovered in 1985 following a search to identify inhibitors of DHPS that access the pterin rather than the pABA binding site. It was shown to have a IC50 of 2.8 M against the E. coli DHPS enzyme (Lever, et al., 1985), and we are using this compound as a starting point to develop novel pterin-based inhibitors of DHPS. We have resynthesized this compound (see Experimental Procedures) and determined the structure of the complex with B. anthracis DHPS to 1.8 A˚ (Figure 7). The conformations of loops 1 and 2 are very similar to those seen in the PtPP structure. However, the DHPS-MANIC structure is at a significantly higher resolution than the other four structures and more of the loops are visible in the electron density. Specifically, loop1 in monomer B is typically not visible but, with the exception of the three most distal residues, this loop is resolved in this
complex. In common with the PtP and PtPP complex structures, loop2 is disordered because MANIC displaces the terminal guanidinium moiety of Arg68 from the pterin binding pocket. However, in monomer A, only 4 residues (67–70) are not resolved. MANIC binds within the pterin pocket in a similar manner to PtPP and PtP using a combination of -stacking, van der Waals, and hydrogen-bonding interactions (Figure 7A). The aromatic ring -stacks onto the terminal guanidinum group of Arg254, and the MANIC triplet of nitrogens at the 1/2/3 positions are in the same position and engage similar hydrogen bond interactions as the equivalent triplet from the pterin substrates. Specifically, the exocyclic amine at position 2 forms two hydrogen bonds with the OD1 oxygens of Asn120 and Asp184, and the N1 interacts with the ND2 nitrogen of Asn120. Also, the hydroxyl oxygen at position 4, equivalent to the ketone in PtPP, forms a hydrogen bond with the structural water molecule that bridges 6 and 7. MANIC also makes two hydrogen bonds between its paired hydroxyl and nitro oxygens and the NZ nitrogen of Lys220, and these correspond to the interactions made with the ketone at position 4 and the acceptor nitrogen at the 5 position in the substrate. Discussion The Conformation and Roles of Loops 1 and 2 In the four crystal structures of DHPS that have now been determined, loops 1 and 2 adopt variable conformations, or are missing altogether, despite having the most highly conserved primary structures in the protein (Figure 3). These loops may only become fully ordered upon the binding of both substrates, which has yet to be observed in any crystal structure. Clearly, no description of the DHPS mechanism can be complete without an understanding of the roles of these two loops. As regards loop1, we can offer no new structural insights into why it is so conserved and the possible roles of its
Structure 1712
Figure 7. The Structure of the Bacillus anthracis DHPS-MANIC Complex MANIC (6-methylamino-5-nitroisocytosine) is a micromolar pterin-based inhibitor of DHPS. (A) Stereoview of the complex. The overall ribbon structure is colored in gray and helix ␣-Loop7 is colored in cyan. MANIC is shown in stick representation, and key DHPS residues are shown as ball-and-stick. The inhibitor binds in the pterin binding pocket of the enzyme, stacks on Arg254, and participates in the typical hydrogen-bonding interactions with Asn120, Asp184, and Lys220. The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997). (B) Final 2Fo ⫺ Fc electron density map, contoured at 1 , calculated around MANIC. The figure was produced using BOBSCRIPT (Esnouf, 1999) and rendered with RASTER3D (Merritt and Bacon, 1997). (C) The extended methyl group of MANIC fills a small pocket in the pterin binding site that results in a superior steric fit for the inhibitor compared to the substrate. The figure was made using the Sybyl modeling package from Tripos Inc. (St. Louis, MO).
residues in the catalytic mechanism. In our structure it is either missing from the electron density or is some distance from the active site involved in crystal contacts with an adjacent molecule. Based on the M. tuberculosis structure, it can apparently fold in such a way as to reach over and contribute to the pABA binding site (Baca et al., 2000), perhaps supplying a conserved aspartate (Asp35 in B. anthracis) to the catalytic mechanism. This is supported by the observation that Phe33, which mediates a crystal contact in our structure, is mutated in multiple forms of sulfa drug-resistant DHPS (Baca et al., 2000) and, therefore, likely to be in the vicinity of the pABA binding pocket. The conformation of loop2 that we observe in B. anthracis DHPS is unique among the known DHPS structures. In the absence of a bound pterin substrate, our structure reveals that the side chain of conserved Arg68 occupies the pterin pocket. Although this conformation has not been observed in any of the three orthologous structures, it exists in both monomers of our asymmetric unit and is unlikely to be an artifact. Strikingly, the combination of the guanidinium group and the two associated water molecules resembles the structure of the pterin ring in terms of its size and shape and the interactions made to the surrounding residues (compare Figures 4 and 5). We suggest that the enzyme has taken advantage of the structural similarities between the pterin ring and the guanidinium moiety to produce a novel device to stabilize the unliganded enzyme and to control the catalytic mechanism. Structurally, since the pterin binding cleft creates a cavity within the core of the -barrel structure, its occupation by the guanidinium moiety would be expected to impart added stability to the unliganded enzyme. Functionally, the role of Arg68 might be to ensure that pABA can only bind if the pterin substrate is already bound at the active site. The binding of pABA within ordered loops 1 and 2 would almost
certainly occlude the pterin binding pocket, and a prebound pterin ring would appear to be essential to the mechanism. Indeed, this order to the catalytic mechanism has been experimentally observed (Vinnicombe and Derrick, 1999), and we have never been able to observe a DHPS-pABA complex despite many crystallization trials. Our structure provides the first rational explanation for why the loop is so conserved (Figure 3). The bent conformation is apparently facilitated by two glycines, Gly63 and Gly64, at the base of the loop, and a -hairpin structure involving Pro69 and Gly70 at the tip of the loop. Also, the conserved Glu65 within the loop makes a stabilizing salt bridge interaction with the conserved Arg82. The Pterin and Phosphate Binding Pockets Two adjacent and conserved pockets at the active site are required to bind the DHPP. The first pocket is within the -barrel core of the enzyme and has been well characterized in previous studies as the cleft that binds the pterin ring. The second has been previously identified as a phosphate binding pocket. Its size, shape, and charge characteristics apparently generate a high-affinity site for large anions since it is invariably occupied by a sulfate ion in the absence of phosphate in the available crystal structures, including the ones reported here. Although there is no question that the second pocket is present to bind one of the two phosphates in the DHPP, there is a question as to which of the two it is. The structures of the E. coli and S. aureus complexes that contain DHPP and PtPP, respectively, suggest that this site binds the -phosphate (Achari et al., 1997; Hampele et al., 1997). However, based on the structure of the M. tuberculosis complex with bound PtP, it was suggested that these conserved residues interact with the ␣-phosphate (Baca et al., 2000). Our structures of the DHPS-PtPP and DHPS-PtP complexes suggest that the
Structure of DHPS from B. anthracis 1713
-phosphate scenario is correct. Comparison of our two structures reveals that either phosphate can occupy the site without compromising the binding of the pterin ring in its pocket, but given the choice in the DHPS-PtPP complex, it is the -phosphate that enters the pocket. Also, whereas the -phosphate can occupy the pocket in exactly the same fashion as the free sulfate ion, the ␣-phosphate is unable to do so and binds less optimally. The Divalent Cation Binding Site Loop1 contains a conserved asparagine residue (Asn27 in the B. anthracis structure) close to the active site, and it has been suggested that its role during catalysis is to bind the Mg2⫹ ion that typically coordinates adjacent oxygen atoms on a bound pyrophosphate moiety (Baca et al., 2000). This arrangement has never been formally observed, but the structures of the S. aureus and M. tuberculosis enzymes in complex with PtPP and PtP, respectively, provide strong support. In the former, a Mn2⫹ cation coordinates the OD1 oxygen of the asparagine, the -phosphate, and a water molecule that, in turn, is hydrogen bonded to the ␣-phosphate. In the latter, a Mg2⫹ cation coordinates the OD1 oxygen of the asparagine and the single ␣-phosphate. Thus far, despite many crystallization trials, we have not been able to experimentally observe a divalent cation in any of our structures. However, in our DHPS structure without a bound pterin substrate, the side chain of Asn27 is appropriately positioned to interact with a cation (Figure 4). In the four complex structures, the side chain of Asn27 is rotated away from the putative cation binding site toward a second bound sulfate (demonstrated in Figure 5), but we imagine that this is an artifact due to the absence of the cation. The Product Structure and the pABA Binding Site Our structure of DHPS in complex with pteroic acid, a product analog, provides the first view of the pABA binding site. It reveals that the pABA moiety sits against the acyl chain of Lys220, makes an edge-to-face interaction with conserved Phe189, and reaches across to make hydrogen-bonding interactions with the N terminus of the conserved helix ␣-Loop7 (Figure 6). This interaction with pABA now provides a functional explanation for the presence of this conserved, additional secondary structural element. A structurally characterized DHPS complex that resembles our pteroic acid structure is from the E. coli enzyme and contains DHPP and sulfanilamide (Achari et al., 1997). This structure shows the pABA analog, sulfanilamide, essentially occupying the same pocket, but it is rotated by 90⬚ compared to our pABA structure and slightly shifted away from ␣-Loop7 (Figure 8). Similar to pABA, one surface of the sulfanilamide ring sits against the side chain of the lysine equivalent to our Lys220, but a key difference is that the opposite surface abuts the side chain of the arginine equivalent to our Arg68, and the guanidinium group is close enough to establish a charge-charge interaction. Although Arg68 is not visible in our structure, the E. coli and S. aureus structures suggest that it creates the second side of the pABA binding pocket when ejected from the pterin binding pocket. The E. coli structure also
suggests that the terminal guanidinium group of Arg68 forms a salt bridge interaction with the terminal carboxylate group of pABA. SAR studies on clinically important sulfonamides have shown that the addition of electron-deficient substituents to the nitrogen of the sulfonamide moiety, such as an oxazole ring (see sulfamethoxazole in Figure 1) results in improved inhibition of the enzyme (Anand 1996). Presumably, these groups lower the pKa of the sulfonamide proton, increase the net negative charge in this region of the drug, and facilitate a charge-charge interaction with Arg68 that normally exists with the terminal carboxylate of pABA. We suggest that point mutations which result in sulfonamide resistance without seriously affecting the catalytic activity with pABA are likely to interact with these added electron-deficient groups. The majority of these mutations occur within loops 1 and 2 (Figure 3), and these have yet to be structurally characterized. However, one mutation site is at the N terminus of helix ␣-Loop7 (Figure 3) and corresponds to Ser221 that we have shown to interact with the terminal carboxylate group of pABA (Figure 6). Curiously, a serine at this position is atypical in DHPS orthologs and has been shown to generate sulfonamide resistance in N. meningitidis (Bennett and Cafferkey, 2003) and S. pyogenes (Jonsson et al., 2003). Therefore, the key interaction of the pABA carboxylate group with this residue appears to be with the main chain amide nitrogen rather than with the side chain. However, as shown in Figure 6, the side chain is ideally positioned to interact with the binding site of the electron-deficient substituents on the sulfonamides and thereby mediate resistance.
The Catalytic Mechanism of DHPS By superimposing our PtPP and pteroic acid structures, it is possible to generate a model of the DHPS transition state. This is shown in Figure 8. Although the model does not include a divalent cation and key regions of loops 1 and 2 such as Arg68, it does address an important point raised by Baca and coworkers. They noted (Baca et al., 2000) that the position and orientation of the sulfanilamide inhibitor in the E. coli complex is inconsistent with it being a true representation of the actual catalytic complex that presumably proceeds by an SN2 mechanism. Based on the structure of their PtP complex, they suggested that the occupation of the tight anion binding pocket by the ␣-phosphate rather than by the -phosphate is a more attractive scenario since the carbon-oxygen bond that is severed during catalysis would be more appropriately aligned for nucleophilic attack. We have confirmed that the -phosphate preferentially occupies the anion binding pocket when PtPP and presumably DHPP bind, and have further shown that the sulfanilamide is rotated by 90⬚ compared to the pABA moiety of pteroic acid. Although this rotation significantly improves the geometry for an SN2 mechanism, it is still not optimal. One possibility is that the ␣-phosphate does displace the -phosphate in the anion binding site during the formation of the transition state, and this would both optimize the SN2 geometry and impart strain into the bond to be cleaved. In the meantime, combining our new structural data
Structure 1714
Figure 8. Transition State Model for the DHPS Catalysis (A and B) The model was developed by overlaying the PtPP and product analog complex structures shown in Figures 5 and 6. Also shown in a transparent format is the sulfanilamide molecule that was visualized in an earlier study of the E. coli DHPS (Achari et al., 1997). Note that the pABA moiety of the pteroic acid is rotated by ⵑ90⬚ compared to the sulfanilamide molecule in the binding site. This rotation improves the orientation of the attacking nitrogen for an SN2 nucleophilic attack compared to the orientation of the sulfanilamide. The figure was produced using MOLSCRIPT (Kraulis, 1991) and rendered with RASTER3D (Merritt and Bacon, 1997).
with data from the previous studies, we suggest that the DHPS catalytic mechanism proceeds by four steps. Initially, in the unliganded enzyme, the pterin binding pocket at the center of the -barrel is occupied by the side chain of Arg68 from loop2 that mimics the pterin ring and orients the loop so as to prevent the binding of pABA. The first step involves the binding of DHPP in the ordered reaction (Vinnicombe and Derrick, 1999). During this binding, the -phosphate enters the anion binding pocket, the divalent cation binding site involving Asn27 becomes occupied by the Mg2⫹ ion that coordinates the pyrophosphate, and loop2 undergoes a conformational change that removes the arginine side chain from the pterin binding pocket. By comparing the loop2 conformations present in our structure with those in E. coli and S. aureus (it is not visible in any of the M. tuberculosis structures), it is possible to envisage how a conformational change might take place, moving from our observed structure, through an intermediate conformation seen in the S. aureus structure to the location that is seen in the E. coli structure. The second step creates the pABA binding pocket. As part of the pterin interaction, the side chain of Lys220 swings into a position that creates a platform for the pABA substrate and exposes the N terminus of helix ␣-Loop7 for additional hydrogen-bonding interactions. The pABA binding pocket is completed by movements of loops 1 and 2, and Arg68 in its new conformation now interacts with pABA via van der Waals and charge interactions. The third step involves the nucleophilic SN2 attack by the nitrogen of pABA on the methylene carbon attached at the C6 position of DHPP, and pyrophosphate is lost. This reaction is shown in Figure 8B. The pterin pocket is lined with residues that include a lysine and an arginine (Lys220 and Arg254 in B. anthracis) that presumably polarize the pterin ring system and facilitate this nucleophilic attack. The divalent cation coordinated to the pyrophosphate acts to make it a better leaving group during the SN2 reaction. Other conserved residues in loops 1 and 2 such as Asp35 may contribute to the catalytic mechanism, but we have not been able to visualize them to date in our structures. Also, as discussed above, there may be a switching of the phosphates in the anion binding site to facilitate the SN2 reaction. During the fourth and final step, loop2 moves back into its original
location, and the reinsertion of Arg68 into the pterin binding pocket displaces the product from the active site. The displacement may be facilitated by the loss of pyrophosphate from the pterin ring during catalysis. The pterin and phosphate binding pockets appear to work in concert since we have been unsuccessful in binding pterin without attached phosphates into our structure. Inhibitors that Access the Pterin Binding Pocket One aspect of DHPS that is clear and consistent in all four orthologous structures is the location of the pterin binding pocket and the interactions that the pterin moiety makes with the conserved residues that line the pocket. Based on the many DHPS sequences that are available, this is a highly conserved interaction that provides a firm structural basis for our notion that small molecules can be designed that access this site and be developed into novel, broad-spectrum therapeutic agents. Our DHPS-MANIC structure is the first that demonstrates how small molecules can be designed to access the pterin pocket (Figure 7). As noted earlier, we have not been successful in binding pterin without attached phosphates into the pocket, but we can identify a number of chemical features of MANIC that appear to contribute to its relatively high binding affinity in the absence of a phosphate moiety. Like pterin, MANIC forms aromatic -stacking interactions with the guanidinium group of Arg254 and is involved in similar hydrogen-bonding interactions within the pterin binding pocket. The one exception is a loss of the hydrogen bond between the N3 nitrogen and the OD1 oxygen of Asp184 due to the enolized nature of MANIC that causes the N3 nitrogen to no longer be protonated. However, due to the shorter distances and more optimal angles, the hydrogen bonds made by the NZ nitrogen of Lys220 with the hydroxyl and nitro oxygens of MANIC are considerably stronger than the corresponding hydrogen bonds made with the ketone oxygen and the N5 ring nitrogen of the substrate. Also, the electronegativity of the NO2 group would result in a favorable partial charge interaction with the electropositive Lys220. The two hydrogen bonds formed by the exocyclic amine group of MANIC would also be stronger than the equivalent bonds made by the amine group of the substrate due to the electron withdrawing properties of
Structure of DHPS from B. anthracis 1715
the para NO2 group that would increase the net positive charge on the hydrogens. The final feature of MANIC that appears to augment its binding affinity is the methyl amine group at the 6 position. This group is equivalent to the C7 carbon in the substrate, but it is rotated 180⬚ from its position in the substrate. This new positioning generates a superior steric fit for the pocket that would be expected to contribute to the binding affinity of the molecule by increasing the overall van der Waals contacts (Figure7C). In support of this, Lever Jr. and coworkers in their original SAR studies on this series showed that with increasing size of the substituent group at this position, the Ki decreased rapidly (Lever, et al., 1985). In our structure, the methyl group is tucked into a small pocket comprising Asp61, Asp101, Asn120, and Ile122, and larger groups could not fit without perturbing the structure. Experimental Procedures Cloning of the folP Gene The anthrax folP gene encoding DHPS was obtained by PCR amplification using genomic DNA obtained from the Sterne 34F2 strain of Bacillus anthracis. The following primers, with inserted restriction sites in italics, were designed using the sequence from the A2102 strain deposited in the NCBI database (gi:21397375): 5⬘ NdeI: CAT ATGAAGTGGGATTATGATTTGCGC and 3⬘ BamHI: GGATCCTTACT TTACCCCTTACCAATCATCGC. Three reaction mixtures were set up using the High Fidelity supermix kit (Invitrogen, Carlsbad, CA) and the protocol suggested in the product literature. The PCR reaction consisted of an initial heating step for 2 min at 94⬚C followed by 25 cycles of 30 s at 94⬚C, annealing at 53⬚C for 30 s, and extension at 72⬚C for 2 min. The PCR products were analyzed on a 1% agarose gel stained with ethidium bromide to identify the amplified species of the appropriate size (846 bp). The PCR products were purified, subcloned using the TOPO TA cloning kit (Invitrogen, Carlsbad, CA) into the pCR 2.1 TOPO vector, and finally sequenced to ensure that the restriction sites were present and that no point mutations had been introduced during the PCR reactions. Primer synthesis and DNA sequencing were performed by the Hartwell Center at St. Jude Children’s Research Hospital. Protein Expression and Purification The folP gene was subcloned from the TOPO TA vector into the pET28a expression plasmid (Novagen, Madison, WI) that incorporates a 6⫻His Tag at the N terminus of the expressed protein. The plasmid was then transformed into competent BL21 (DE3) Codon Plus (RIL) E. coli cells (Stratagene, La Jolla, CA) for expression. A 50 ml overnight culture was grown from a single colony of cells in LB containing 50 g/ml kanamycin and 50 g/ml chloramphenicol, and this culture was used to inoculate 6 liters of the same medium. This was grown at 37⬚C until the OD600 reached ⵑ0.6, at which point isopropyl--Dthiogalactopyranoside (IPTG) was added to a final concentration of 1 mM. Cells were then allowed to grow in the presence of IPTG for a further 3 hr and then harvested by centrifugation. The cell pellet was washed with 10 ml of lysis buffer (50 mM Tris, pH 8.0, 0.5 M NaCl, 1 mM imidazole, 0.1 mM phenylmethylsulfonyl fluoride [PMSF], 2% glycerol) and then resuspended in 40 ml of lysis buffer. Cells were lysed by sonication, and the cell debris was cleared by centrifugation at 20,000 ⫻ g for 30 min. Purification of DHPS required three steps of Fast Performance Liquid Chromatography (FPLC), Ni2⫹ chelation affinity chromatography, gel filtration, and HQ sepharose. The cell lysate was first applied to a 5 ml HiTrap Chelating HP column (Amersham Biosciences, Piscataway, NJ) and washed with ten column volumes of lysis buffer. The protein eluted during a gradient of 0–0.3 M imidazole, and the appropriate fractions were identified by SDS-PAGE. The pooled DHPS was concentrated to 3 ml, applied to a HiLoad 16/60 Superdex 75 column (Amersham Biosciences), and eluted with running buffer
(20 mM Tris-HCl, pH 8, 0.2 M NaCl, 2 mM dithiothreitol [DTT] and 2 mM ethylenediaminetetraacetic acid [EDTA]) at a flow rate of 0.5 ml/min. SDS-PAGE revealed one remaining contaminant, so the pooled sample was diluted 50:50 with running buffer containing no NaCl and applied to a 25 ml HQ sepharose column (Amersham Biosciences). The column was washed with five column volumes of 20 mM Tris-HCl, pH 8, 50 mM NaCl, 2 mM DTT, and 2 mM EDTA and eluted with a gradient of 0.05–1.0 M NaCl. The final purified protein was estimated to be ⬎95% pure by SDS-PAGE analysis. Crystallization and Data Collection The purified enzyme was dialyzed against 20 mM Tris, pH 8.0, 1 mM DTT, 100 mM NaCl and then concentrated with an Amicon stirred cell to a final concentration of 15 mg/ml. Using the hanging drop vapor diffusion method together with the Hampton Crystal Screen kits (Aliso Viejo, CA), a number of initial crystallization conditions were identified by mixing equal volumes (3 l) of protein and well solution. These conditions were subsequently refined, and large well-diffracting crystals could eventually be reproducibly grown from 1.3 M Li2SO4, 0.1 M Bis-Tris propane, pH 9.0. Hexagonal rodshaped crystals appeared within 48 hr at 18⬚C. The crystals are in space group P6222 with cell dimensions a ⫽ b ⫽ 97.4 A˚, c ⫽ 263.6 A˚. Consideration of the Matthews coefficient suggested that two molecules of DHPS should be present in the asymmetric unit, and this was subsequently confirmed. For the structural studies of the four DHPS complexes, the compounds were directly added to pregrown crystals in the form of a saturated solution in the crystallization buffer, and allowed to soak for 48 hr. The four compounds were 6-hydroxymethylpterin pyrophosphate (PtPP) (Schircks Laboratories, Jona, Switzerland), 6-hydroxymethylpterin monophosphate (PtP) (Schircks Laboratories), pteroic acid (PA) (Sigma, St. Louis, MO), and 6-(methylamino)-5-nitroisocytosine (MANIC) (synthesized, see below). All X-ray diffraction data were collected under cryogenic conditions. Crystals were briefly immersed in a cryoprotectant mixture of 50% Paratone-N and 50% mineral oil and directly flash-frozen in liquid nitrogen. An initial native data set to 2.8 A˚ resolution was collected in-house on a Bruker PROTEUM R CCD detector system mounted on a Bruker FR591 rotating anode, and integrated and processed using the SAINT and PROSCALE data processing packages (Bruker AXS). Subsequently, all data from native and compound-soaked crystals were collected at the Southeast Regional Collaborative Access Team (SER-CAT) synchrotron beamlines 22ID and 22-BM at the Advanced Photon Source (Argonne National Laboratory). All crystals were frozen and prescreened in-house before synchrotron data collection, and preliminary low-resolution electron density maps were calculated to confirm the presence of the soaked compounds before moving to high resolution. Anisotropy in the diffraction pattern around the long c-axis limited the effective resolution of the data sets to 2.0, 2.5, 2.8, 2.2, and 1.8 A˚ for the DHPS, DHPS-PtPP, DHPS-PtP, DHPS-PA, and DHPS-MANIC crystals, respectively. The DHPS, DHPS-PtPP, and DHPS-PtP synchrotron data were collected on a MAR165 CCD detector and processed and merged using HKL2000 (Otwinowski and Minor, 1997). The DHPS-PA and DHPS-MANIC data were collected on a MAR225 CCD detector and integrated with MOSFLM (Leslie, 1992), and scaled and merged using Scala (Evans, 1997). All relevant data collection statistics are shown in Table 1. Structure Determination The structure of DHPS from B. anthracis was determined by molecular replacement using the CNS crystallographic package (Bru¨nger et al., 1998). Of the available three DHPS structures in the protein structure database, the S. aureus DHPS had the highest sequence similarity (46% identity), and monomer A of that structure (PDB accession code 1AD4) was used as the initial search model. A solution was obtained using the data collected on the home source, and a preliminary model of the dimer in the asymmetric unit was built and refined by simulated annealing. The model was significantly improved using the higher resolution synchrotron data following several iterative rounds of rebuilding using the program O (Jones et al., 1991), and simulated annealing and temperature factor refinement using CNS. In addition, several rounds of maximum likelihood
Structure 1716
refinement using the CCP4 program REFMAC (Murshudov et al., 1997) with the TLS option were performed. For the TLS refinement, each monomer was defined as a separate TLS group, and in final rounds of refinement, the bound waters and ions were treated as a third TLS group. During initial stages of refinement, NCS restraints were applied to the two monomers in the asymmetric unit, but these were removed during the later stages. To solve the four complexes, rigid body refinement using REFMAC was used to position the native DHPS molecule in the slightly altered unit cells, and iterative rounds of rebuilding using the program O, simulated annealing, and finally REFMAC were used to complete the structures. In all the structures, water molecules were picked automatically using ARP-WARP (Lamzin and Wilson, 1993) and subsequently confirmed by visual inspection. Several large peaks were interpreted and successfully refined as sulfate ions. Finally, to verify ambiguous regions of the electron density and to check the models, a number of simulated annealing omit maps were calculated during the latter stages of the refinement process. The quality of the structures was verified using PROCHECK (Laskowski et al., 1993), and the final refinement statistics are shown in Table 2.
Synthesis of 6-(Methylamino)-5-Nitroisocytosine 6-(methylamino)-5-nitroisocytosine (MANIC) was prepared in a twostep synthesis using standard reaction conditions. Starting materials and reaction solvents were purchased from Sigma-Aldrich (Milwaukee, WI). Step A: Synthesis of 2-Amino-6-Chloro-5-Nitropyrimidin4(3H)-One A solution of 12 ml of concentrated sulfuric acid and 12 ml of 90% nitric acid was cooled to 0⬚C in an ice-salt bath. Then, 2-amino-6chloro-4-(3H)-pyrimidinone (2 g, 13.79 mmol) was added in small portions so that the reaction temperature did not exceed 10⬚C. After finishing the addition, the mixture was stirred for 3 hr. Then, the reaction mixture was poured to ice water and stirred for about 1 hr below 5⬚C. The precipitated solid was collected by filtration, washed with water and methanol, and dried to give 2-amino-6-chloro-5nitropyrimidin-4(3H)-one a yellow solid (2.40 g, yield 91.6%). 1H NMR (300 MHz, DMSO-d6) 12.15(s, 1H), 8.55(s, 1H), 7.15(s, 1H) MS(ESI) 189 (M⫺⫺1). Step B: Synthesis of 2-Amino-6-(Methylamino)5-Nitropyrimidin-4(3H)-One or 6-(Methylamino)5-Nitroisocytosine To a solution of 7 ml of 2 M methylamine in MeOH and 5 ml of water was added 2-amino-6-chloro-4-(3H)-pyrimidinone (500 mg, 2.63 mmol) at room temperature, and the reaction mixture was then heated to 60⬚C and stirred for 3 hr. The reaction solution was then cooled and filtered, and the filtrate concentrated. The residue was diluted with about 5 ml of water and filtered. The solid was washed with water and MeOH, and air dried to give 2-amino-6-(methylamino)-5-nitropyrimidin-4(3H)-one or 6-(methylamino)-5-nitroisocytosine a yellow solid (585 mg, yield 17.5%). 1H NMR (300 MHz, DMSO-d6) 10.55(s, 1H), 9.50(s, 1H), 6.40(brs, 2H), 2.97(d,J ⫽ 6 Hz, 3H); 13CNMR (300 MHz, DMSO-d6)159.5, 156.3, 154.0, 110.5, 28.3; MS(ESI) 186(M⫹⫹1).
Acknowledgments This work was supported by NIH grant AI60953, Cancer Center (CORE) Support Grant CA21765, and the American Syrian Lebanese Associated Charities. Synchrotron diffraction data were collected at the Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Special acknowledgment goes to the SER-CAT beamline staff for the collection and processing of the DHPS-PtPP and DHPS-PtP data on the 22-BM beamline. SER-CAT supporting institutions may be found at www.ser.anl.gov/new/index.html. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science, under Contract No. W-31-109-Eng-38. We thank Philip Hanna for invaluable help in obtaining the folP gene from Bacillus anthracis.
Received: May 13, 2004 Revised: June 24, 2004 Accepted: July 8, 2004 Published: September 7, 2004 References Achari, A., Somers, D.O., Champness, J.N., Bryant, P.K., Rosemond, J., and Stammers, D.K. (1997). Crystal structure of the anti-bacterial sulfonamide drug target dihydropteroate synthase. Nat. Struct. Biol. 4, 490–497. Anand, N. (1996). Sulfonamides and sulfones. In Burger’s Medicinal Chemistry and Drug Discovery, Fifth Edition, Volume 2: Therapeutic Agents, M.E. Wolff, ed. (New York: John Wiley & Sons, Inc.), pp. 527–573. Aspinall, T.V., Joynson, D.H., Guy, E., Hyde, J.E., and Sims, P.F. (2002). The molecular basis of sulfonamide resistance in Toxoplasma gondii and implications for the clinical management of toxoplasmosis. J. Infect. Dis. 185, 1637–1643. Baca, A.M., Sirawaraporn, R., Turley, S., Sirawaraporn, W., and Hol, W.G.J. (2000). Crystal structure of mycobacterium tuberculosis 6-hydroxymethyl-7,8-dihydropteroate synthase in complex with pterin monophosphate: new insight into the enzymatic mechanism and sulfa-drug action. J. Mol. Biol. 302, 1193–1212. Banner, D.W., Bloomer, A.C., Petsko, G.A., and Phillips, D.C. (1975). Structure of chicken muscle triose phosphate isomerase determined crystallographically at 2.5 A˚ resolution using amino acid sequence data. Nature 255, 609–614. Bennett, D.E., and Cafferkey, M.T. (2003). PCR and restriction endonuclease assay for detection of a novel mutation associated with sulfonamide resistance in Neisseria meningitides. Antimicrob. Agents Chemother. 47, 3336–3338. Brown, G.M. (1962). The biosynthesis of folic acid. Inhibition by sulfonamides. J. Biol. Chem. 237, 536–540. Bru¨nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D54, 905–921. Domagk, G. (1935). Ein beitrag zur chemotherapie der bakteriellen infektionen. Dtsch. Med. Wschr. 7, 250–253. Enne, V.I., King, A., Livermore, D.M., and Hall, L.M.C. (2002). Sulfonamide resistance in Haemophilus influenzae mediated by acquisition of sul2 or a short insertion in chromosomal folP. Antimicrob. Agents Chemother. 46, 1934–1939. Esnouf, R.M. (1999). Further additions to MolScript version 1.4, including reading and contouring of electron-density maps. Acta Crystallogr. D Biol. Crystallogr. 55, 938–940. Evans, P.R. (1997). SCALA. Joint CCP4 and ESF-EAMCB Newsletter on Protein Crystallography 33, 22–24. Ferme´r, C., and Swedberg, G. (1997). Adaptation to sulfonamide resistance in Neisseria meningitidis may have required compensatory changes to retain enzyme function: kinetic analysis of dihydropteroate synthases from N. meningitidis expressed in a knockout mutant of Escherichia coli. J. Bacteriol. 179, 831–837. Gibreel, A., and Sko¨ld, O. (1999). Sulfonamide resistance in clinical isolates of Campylobacter jejuni: mutational changes in the chromosomal dihydropteroate synthase. Antimicrob. Agents Chemother. 43, 2156–2160. Greenfield, R.A., and Bronze, M.S. (2003). Prevention and treatment of bacterial diseases caused by bacterial bioterrorism threat agents. Drug Discov. Today 8, 881–888. Haasum, Y., Stro¨m, K., Wehelie, R., Luna, V., Roberts, M.C., Maskell, J.P., Hall, L.M., and Swedberg, G. (2001). Amino acid repetitions in the dihydropteroate synthase of Streptococcus pneumoniae lead to sulfonamide resistance with limited effects on substrate Km. Antimicrob. Agents Chemother. 45, 805–809. Hampele, I.C., D’Arcy, A., Dale, G.E., Kostrewa, D., Nielsen, J., Oefner, C. Page, M.G., Schonfeld, H.J., Stuber, D., and Then, R.L.
Structure of DHPS from B. anthracis 1717
(1997). Structure and function of the dihydropteroate synthase from Staphylococcus aureus. J. Mol. Biol. 268, 21–31. Hughes, W.T. (1988). Comparison of dosages, intervals, and drugs in prevention of Pneumonocystis carinii pneumonia. Antimicrob. Agents Chemother. 32, 623–625. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved method for binding protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119. Jonsson, M., Strom, K., and Swedberg, G. (2003). Mutations and horizontal transmission have contributed to sulfonamide resistance in Streptococcus pyogenes. Microb. Drug Resist. 9, 147–153. Kai, M., Matsuoka, M., Nakata, N., Maeda, S., Gidoh, M., Maeda, Y., Hashimoto, K., Kobayashi, K., and Kashiwabara, Y. (1999). Diaminodiphenylsulfone resistance of Mycobacterium leprae due to mutations in the dihydropteroate synthase gene. FEMS Microbiol. Lett. 177, 231–235. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946–950. Kumar, H.P., Tsuji, J.M., and Henderson, G.B. (1987). Folate transport in Lactobacillus salivarius: characterization of the transport mechanism and purification and properties of the binding component. J. Biol. Chem. 262, 7171–7179. Lamzin, V.S., and Wilson, K.S. (1993). Automated refinement of protein models. Acta Crystallogr. D Biol. Crystallogr. 49, 129–149. Laskowski, R.A., McArthur, M.W., Moss, D.S., and Thornton, J.M. (1993). PROCHECK: a program to check the quality of protein structures. J. Appl. Crystallogr. 26, 282–291. Leslie, A.G.W. (1992). Recent changes to the MOSFLM package for processing film and image plate data. Joint CCP4 and ESF-EAMCB Newsletter on Protein Crystallography 26. Lever, O.W., Jr., Bell, L.N., McGuire, H.M., and Ferone, R. (1985). Monocyclic pteridine analogues. Inhibition of Escherichia coli dihydropteroate synthase by 6-amino-5-nitrosoisocytosines. J. Med. Chem. 28, 1870–1874. Lever, O.W., Jr., Bell, L.N., Hyman, C., McGuire, H.M., and Ferone, R. (1986). Inhibitors of dihydropteroate synthase: substituent effects in the side-chain aromatic ring of 6-[[3-(aryloxy)propyl]amino]-5nitrosoisocytosines and synthesis and inhibitory potency of bridged 5-nitrosoisocytosine-p-aminobenzoic acid analogues. J. Med. Chem. 29, 665–670. Ma, L., and Kovacs, J.A. (2001). Genetic analysis of multiple loci suggest that mutations in the Pneumocystis carinii f. sp. Hominis dihydropteroate synthase gene arose independently in multiple strains. Antimicrob. Agents Chemother. 45, 3213–3215. Matherly, L.H. (2001). Molecular and cellular biology of the human reduced folate carrier. Prog. Nucleic Acid Res. Mol. Biol. 67, 131–162. Merritt, E.A., and Bacon, D.J. (1997). Raster 3D: photorealistic molecular graphics. Methods Enzymol. 277, 505–524. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. Roland, S., Ferone, R., Harvey, R.J., Styles, V.L., and Morrison, R.W. (1979). The characteristics and significance of sulfonamides as substrates for Escherichia coli dihydropteroate synthase. J. Biol. Chem. 254, 10337–10345. Sko¨ld, O. (2000). Sulfonamide resistance mechanisms and trends. Drug Resist. Updat. 3, 155–160. Sko¨ld, O. (2001). Resistance to trimethoprim and sulfonamides. Vet. Res. 32, 261–273. Swedberg, G., Ringertz, S., and Sko¨ld, O. (1998). Sulfonamide resistance in Streptococcus pyogenes is associated with differences in amino acid sequence of its chromosomal dihydropteroate synthase. Antimicrob. Agents Chemother. 42, 1062–1067.
Vinnicombe, H.G., and Derrick, J.P. (1999). Dihydropteroate synthase from Streptococcus pneumoniae: characterization of substrate binding order and sulfonamide inhibition. Biochem. Biophys. Res. Commun. 258, 752–7. Wang, P., Lee, C.S., Bayoumi, R., Djimde, A., Dourmbo, O., Swedberg, G., Dao, L.D., Mshinda, H., Tanner, M., Watkins, W.M., et al. (1997). Resistance to antifolates in Plasmodium falciparum monitored by sequence analysis of dihydropteroate synthase and dihydrofolate reductase alleles in a large number of field samples of diverse origins. Mol. Biochem. Parasitol. 89, 161–177. Accession Numbers The structure factors and refined coordinates have been deposited in the PDB with accession codes 1TWS (unliganded), 1TWW (PtPP complex), 1TWZ (PtP complex), 1TX0 (Pteroic acid complex), and 1TX2 (MANIC complex).