Crystal Structure of Toxoplasma gondii Porphobilinogen Synthase

Abstract

Porphobilinogen synthase (PBGS) is essential for heme biosynthesis, but the enzyme of the protozoan parasite Toxoplasma gondii (TgPBGS) differs from that of its human host in several important respects, including subcellular localization, metal ion dependence, and quaternary structural dynamics. We have solved the crystal structure of TgPBGS, which contains an octamer in the crystallographic asymmetric unit. Crystallized in the presence of substrate, each active site contains one molecule of the product porphobilinogen. Unlike prior structures containing a substrate-derived heterocycle directly bound to an active site zinc ion, the product-bound TgPBGS active site contains neither zinc nor magnesium, placing in question the common notion that all PBGS enzymes require an active site metal ion. Unlike human PBGS, the TgPBGS octamer contains magnesium ions at the intersections between pro-octamer dimers, which are presumed to function in allosteric regulation. TgPBGS includes N- and C-terminal regions that differ considerably from previously solved crystal structures. In particular, the C-terminal extension found in all apicomplexan PBGS enzymes forms an intersubunit β-sheet, stabilizing a pro-octamer dimer and preventing formation of hexamers that can form in human PBGS. The TgPBGS structure suggests strategies for the development of parasite-selective PBGS inhibitors.

Introduction

The myriad forms of cyclic and linear tetrapyrroles, including heme, chlorophyll, siroheme, cobalamine (B₁₂), and the phycobilins, are essential for energy production/utilization in virtually all life forms. Tetrapyrrole biosynthesis pathways are complex and phylogenetically variable, suggesting the potential for species-selective control (1). The universal portion of this pathway consists of only three reactions, the first of which involves biosynthesis of the monopyrrole porphobilinogen from two molecules of 5-aminolevulinic acid (ALA)³ (Fig. 1), catalyzed by porphobilinogen synthase (PBGS; EC 4.2.1.24; also known as 5-aminolevulinic acid dehydratase or ALAD). The photoreactive and toxic nature of tetrapyrrole biosynthesis intermediates mandates close regulation of this pathway, and various mechanisms are used for control. We recently have described an unusual mechanism of allosteric regulation involving transition between a high activity PBGS octamer and a low activity hexamer (2–4). In plants, an allosteric magnesium ion favors octamer formation by binding at a subunit interface not present in the hexamer (2). The term “morpheeins” has been used to describe oligomers that can disassemble, change shape in the dissociated state, and reassemble to a structurally and functionally distinct oligomer (5).

FIGURE 1.

The PBGS-catalyzed reaction. PBGS catalyzes the condensation of two molecules of ALA to form porphobilinogen and two water molecules. A-side ALA (thick lines) becomes the acetyl-containing half of porphobilinogen; P-side ALA (thin lines) becomes the propionyl-containing half.

Fig. 2 illustrates the quaternary structure dynamics of plant and mammalian PBGS, using the human structure by way of example. “Hugging” and “detached” dimers (pathway A) are observed as the asymmetric crystallographic units for wild type and mutant forms of human PBGS (PDB codes 1E51 and 1PV8, respectively (2, 6)). The structure of the physiologically relevant dimer, and the pathway of hexamer-octamer conversion, remains unclear, however, as biochemical and biophysical tools demonstrate that interconversion between octameric and hexameric oligomers involves a conformationally flexible dimer (4, 7). Analysis of wild-type human PBGS establishes that the hexamer is a pH-dependent component of the quaternary structure equilibrium and suggests that oligomer dissociation proceeds via the pro-octamer and pro-hexamer dimers (Fig. 2, pathway B), rather than the crystallographically defined hugging and detached dimers (4). Sequence differences between the PBGS enzymes of different species may be manifested as differences in protein structure and quaternary dynamics, making it essential to understand the structural changes associated with this equilibrium if we are to develop species-selective small molecules (8) that are able to modulate PBGS activity by altering oligomeric equilibria.

FIGURE 2.

The quaternary structure dynamic of some PBGS. Human PBGS crystal structures (PDB codes 1E51 and 1PV8) are used to illustrate the alternate assemblies the protein can adopt. The alternate dissociation pathways denote possible mechanisms for PBGS oligomer dissociation to dimers. The hugging and detached dimers (pathway A) are the asymmetric crystallographic units of the crystal structures. Based on the hydrophobicity and hydrophilicity of surfaces exposed upon dissociation, the pro-octamer and pro-hexamer dimers (pathway B) of human PBGS have been proposed to be the physiologically relevant dimers (4). The “twist” that describes the transition between the pro-octamer and pro-hexamer dimers is illustrated by the enlarged sections that denote ϕψ angles at residue 23 (asterisk).

The protozoan parasite Toxoplasma gondii is of particular concern in immunocompromised patients and pregnant women (9). As in related parasites, including the Plasmodium species responsible for malaria, T. gondii PBGS (TgPBGS) is located within the apicoplast (a nonphotosynthetic plastid organelle acquired by secondary endosymbiosis (10, 11)) and resembles plant/algal PBGS in its primary structure (see Fig. 3). The protein contains all amino acids thought to be involved in binding both catalytic and allosteric magnesium ions (12). We have previously shown that TgPBGS exhibits a quaternary structure equilibrium dominated by the octameric form (13); dimeric TgPBGS is also observed, but there is no evidence for hexamer formation, unlike plant and human PBGS. In addition, the TgPBGS octamer is not destabilized by magnesium chelation (13), which is unlike in other magnesium-binding PBGS proteins; removal of magnesium causes the plant protein to favor the hexamer, whereas the Pseudomonas PBGS purifies as a dimer if magnesium is omitted (2–3). TgPBGS contains an unusual C-terminal sequence extending ∼13 amino acids beyond most other PBGS proteins (see Fig. 3), and removal of this C-terminal extension destabilizes the octamer (13) to favor a low specific activity dimer (∼4% wild-type activity).

FIGURE 3.

Alignment of PBGS protein sequences highlighting key features essential for structure and function. This alignment omits hundreds of species of PBGS that do not contain the C-terminal extension. Amino acid numbering (residues 1–356) pertains to the TgPBGS construct used for the current crystal structure. Negative numbers correspond to the extended N-terminal that include low complexity region of TgPBGS, which is not included in the expression construct. N-terminal portions of Pisum sativum and Plasmodium falciparum PBGS are shown with negative numbering but not aligned with each other or TgPBGS. Alignment was done using the ClustalW2 program (43). Species abbreviations are as follows: Tgo, T. gondii; Pfa, P. falciparum; Psa, P. sativum (pea); Vch, V. cholerae; Pae, P. aeruginosa; Cvi, C. vibrioforme; Eco, E. coli; Sce, yeast; Hsa, human. The underlined species are ones for which there is a crystal structure, including T. gondii (this work). The extended N and C termini of TgPBGS are shown in blue. The cysteine-rich region that binds an active site zinc ion is in red, and the cysteines that coordinate zinc are highlighted in yellow. PBGS that do not use an active site zinc lack the cysteine residues but contain aspartic acid residues (highlighted in gray), which were proposed to coordinate magnesium (34). The residues involved in allosteric magnesium ion coordination are highlighted in red. The active site lid residues are green. Conserved active site residues are highlighted in purple. The asterisk denotes amino acid positions, which contain identical residues (shown in boldface) in all proteins aligned here.

We present the crystal structure of TgPBGS, which is distinct from nearly 40 previously solved PBGS structures. Distinctions are seen in 1) the asymmetric crystallographic unit, 2) the structures of both the N and C termini, and 3) the active site ligands. The TgPBGS structure explains how the C-terminal extension characteristic of apicomplexan PBGS enzymes stabilizes the octameric structure through intersubunit domain swapping interactions and argues that the biologically relevant PBGS dimer is different from the dimer observed as the crystallographic asymmetric unit in other structures.

EXPERIMENTAL PROCEDURES

Materials

TgPBGS was expressed and purified as reported (13). Protein is derived from the TgPBGS3 splice variant, which excludes the N-terminal targeting signals (amino acids 1–147) and a low complexity sequence (amino acids 148–301) whose function is not known. It includes amino acids 303–658 of TgPBGS3, which for the purpose of this paper and the crystal structure are numbered 1–356. The ADDit^TM screen was obtained from Emerald BioSystems.

Crystallization

An initial crystallization hit was obtained in 0.2 m potassium thiocyanate, 20% PEG 3350. These C-centered orthorhombic crystals, which diffracted to ∼4 Å at a synchrotron, were then optimized with the ADDit screen. The final crystallant consisted of 0.18 m potassium thiocyanate, 18% PEG 3350, and 3% w/v diaminooctane. The protein solution, consisting of TgPBGS at 6.3 mg/ml, in 450 mm Tris-HCl, pH 7.0, in the presence of 1 mm MgCl₂, was incubated with 5 mm 5-aminolevulinic acid prior to crystallization. Crystals were obtained by sitting drop vapor diffusion at 289 K. Crystals that diffracted to 2.5 Å were obtained, cryoprotected with 25% ethylene glycol, and cryo-vitrified for shipping.

Data Collection and Processing

Intensity data from crystals of TgPBGS in complex with in situ evolved product porphobilinogen were collected under cryogenic conditions at 100 K using synchrotron radiation at beamline 5.0.2 of the Advanced Light Source, as part of the Collaborative Crystallography program. The x-ray wavelength was set to 1.00 Å. The oscillation range per image was set to 1.0°. The data sets were processed using XDS and XSCALE (14).

Crystal Structure Solution

The structure was solved by molecular replacement with the Phaser MR program from the CCP4 software suite. The starting model was prepared from Pseudomonas aeruginosa PBGS (PDB code 1B4K) by treating the entire dimer with the alignment tool CHAINSAW (15–17).

Crystal structure refinement involved iterative manual rebuilding in Coot and maximum likelihood refinement using the program Refmac5, with medium noncrystallographic symmetry restraints between the eight monomers (18–19). A Translation/Libration/Screw model with two groups per monomer was determined with the TLS server (20), which was employed for more accurate modeling of thermal motion. After six iterations of rebuilding and refinement, the electron density at the active site was deemed clear enough to build the enzyme reaction product porphobilinogen at the active sites of all eight monomers. Rebuilding and refinement were iterated for seven additional cycles. The stereochemical quality of the final model was verified using MolProbity (21).

TgPBGS Kinetic Analysis

Kinetic analysis was carried out as described previously (13). The assays were carried out in duplicate at a final pH of 8.0 for 30 min at 37 °C. Assay composition included 10 μg/ml TgPBGS, 0.1 m bis-tris propane, 10 mm ALA-HCl, with and without 10 mm MgCl₂. Kinetic data were fit to simple single and double hyperbolic equations using SigmaPlot (version 10).

RESULTS

The TgPBGS structure, mechanism, and metal ions are described in context with the PBGS family of enzymes, which is highly homologous throughout much of the protein, but significantly variant in the N and C termini (Fig. 3).

TgPBGS Crystal Structure Solution

The TgPBGS crystal belongs to the primitive orthorhombic space group P2₁2₁2 and diffracted to a resolution of 2.5 Å. The crystal had unit-cell parameters a = 177.11, b = 187.17, and c = 95.86. The data set was 98.6% complete, with an R_merge of 11.5%. The I/σ (I) ratio in the last resolution bin was 2.3. Statistics of data collection parameters are summarized in Table 1. TgPBGS crystal structure determination follows that of yeast, Escherichia coli, Pseudomonas aeruginosa, Chlorobium vibrioforme, human, and mouse PBGS proteins, which are all homo-oligomeric enzymes, with a common subunit structure minimally consisting of an αβ-barrel domain and an N-terminal arm domain (22). In all known PBGS proteins, the N-terminal arm is essential for oligomer assembly. In the case of TgPBGS, the oligomer is an octamer whose subunits are denoted A through H, which together establish the asymmetric crystallographic unit (Fig. 4a). In addition, TgPBGS contains fully formed porphobilinogen in all eight active sites.

TABLE 1.

Data collection, phasing, and refinement statistics

R_merge = (Σ(ABS(I(h,i) − I(h))))/(Σ(I(h,i))). r.m.s.d., root mean square deviation.

PDB code	3OBK
Space group	P21212
Space group number	18
Unit-cell parameters
a (Å)	177.11
b (Å)	187.17
c (Å)	95.86
α	90°
β	90°
γ	90°
Resolution limits (Å) (last shell)	49.93-2.5
Rwork overall	0.177
Rfree overall	0.232
Mean Biso	23.2
r.m.s.d. in bondlengths (Å)	0.014
r.m.s.d. in angles	1.497°
Rmerge (IOBS) (%) (last shell)	11.5 (53.8)
I/σ(I) overall (last shell)	11.8 (2.3)
Completeness overall (%)	98.6 (90.3)
Model details
Total non-hydrogen atoms	22,298
Protein atoms	21,476
Ligand atoms (PBG)	136
Ethylene glycol	76
Metal atoms	8
Chlorine atoms	16
Water O atoms	586
Ramachandran outliers	8/2800 (0.3%)
Ramachandran favored	2717/2800 (97.0%)

FIGURE 4.

The TgPBGS structure. a, two orthogonal orientations of the TgPBGS octamer showing each subunit in a different color, with the subunits labeled A–H. On the left, all subunits are shown as spheres. On the right, subunits A, D, E, and G are shown as ribbons to illustrate the intersubunit location of each N-terminal arm (N temini are labeled where visible). b, using the same orientation and coloring as the right side of part a, three subunits (A, B, and C) are illustrated in ribbon representation. Porphobilinogen is shown as sticks in each active site, colored as the subunit. For each subunit, the N-terminal arm and the C-terminal tail are in a darker shade. The N-terminal arm of the magenta subunit wraps around the αβ-barrel domain of the gold subunit, constituting the hugging dimer (see Fig. 2). The N-terminal arm of the magenta subunit lies against the internal face of the αβ-barrel domain of the cyan subunit, constituting the pro-octamer dimer (see Fig. 2). The C-terminal tails of the pro-octamer dimer form an intersubunit β-sheet (tail swap), emphasized in c. c, using the same orientation as in part b, and pale colors for the rest of subunits A and C, the tail swap is highlighted (circled in black). d, details of the C-terminal tail portion of the pro-octamer dimer illustrate additional intersubunit interactions between the C-terminal β-strand of the magenta subunit and both the N-terminal arm and αβ-barrel region of the cyan subunit. Reciprocal interactions are not shown. e, the eight subunits of the TgPBGS crystallographic asymmetric unit are overlaid using wireframe representation and colored by thermal factors (red is hot or mobile, and blue is cold or fixed). Porphobilinogen at the active site is shown in spheres.

N-terminal Arm of TgPBGS

As seen in other PBGS structures, the N-terminal arm of one subunit of the TgPBGS oligomer wraps around the side of the αβ-barrel of another subunit and nestles into the base of the αβ-barrel of a third subunit (Fig. 4b). Structural superposition of PBGS from different species (TgPBGS, PDB code 3OBK; yeast PBGS, PDB code 1OHL; P. aeruginosa PBGS, PDB code 2WOQ; and C. vibrioform PBGS, PDB code 2C1H) (Fig. 5) reveals excellent superposition in the region of the αβ-barrel domain and the first helix-turn-helix of the N-terminal arm but significant variation in the most N-terminal region of the protein (residues 5–24 of TgPBGS). Structural conservation through the αβ-barrel domain and the adjacent helix-turn-helix portion of the N-terminal arm (TgPBGS residues 25–342) is consistent with the sequence conservation (Fig. 3). However, the conformation of the N-terminal 24 amino acids of TgPBGS is different from that seen in any previous PBGS crystal structure, most specifically yeast PBGS, for which there is also significant structural information N-terminal to the conserved domains (Fig. 5). Interpretation of the significance of the N-terminal arm positioning in TgPBGS is confounded by a lack of information on the low complexity region of sequence, which is present in the gene, and presumably also present in the physiologically relevant mature protein, but not present in the expression construct used for this work. The likely trajectory for this missing portion of the protein is outside the octamer (see N termini in Fig. 4a). Like most other PBGS crystal structures, the most N-terminal residues of TgPBGS (residues 1–4 of each subunit) are disordered. Because of the importance of the intersubunit location of the N-terminal arm, phylogenetic sequence differences therein dictate differences in the strength of the interactions between subunits and factors into the kinetics and thermodynamics of the quaternary structure equilibrium shown in Fig. 2.

FIGURE 5.

A comparison of the structure of PBGS from various species. Two images are provided to illustrate the considerable variation observed in the conformation of PBGS N and C termini. Structures are overlaid for a single subunit from TgPBGS (PDB code 3OBK, green), yeast PBGS (PDB code 1OHL, yellow), C. vibrioforme PBGS (PDB code 2CLH, cyan), and P. aeruginosa PBGS (PDB code 2WOQ, magenta). Where visible, the N and C termini of TgPBGS are labeled; cylindrical arrows are provided adjacent to one helix to facilitate orientation.

C-terminal Extension (Tail) of TgPBGS

One unique aspect of the TgPBGS crystal structure is a C-terminal extension of ∼13 amino acids, which is present in only a small number of known PBGS sequences but interestingly in all known apicomplexan PBGS sequences. The sequence of this C-terminal extension is conserved poorly across the phylum but conserved highly within individual subclasses, i.e. the different Plasmodium species (Haemosporidia) all have the same extension sequence, which differs completely from the conserved sequence in coccidian parasites including Toxoplasma (supplemental Fig. 1). TgPBGS provides the first structure for the C-terminal extension. From the perspective of a TgPBGS monomer, as the N-terminal region is called the arm, we denote the C-terminal extension as the tail. Fig. 4c illustrates how the C-terminal tail also sits at an interface of TgPBGS subunits. In most PBGS crystal structures, the few C-terminal structured amino acids are α-helical, followed by a small number of unstructured residues. In TgPBGS, the first few amino acids of the C-terminal tail show a continuation of this common α-helix, followed by a loop which ends as a β-strand (amino acids 353–356). This β-strand structure is involved in the formation of an intersubunit anti-parallel β-sheet with the neighboring subunit such that the subunit pair forms the pro-octamer dimer (Fig. 4c and Fig. 2). This interaction is denoted as the tail-swap and is consistent with our previously published findings that the C-terminal tail is essential for octamer formation (13). Despite significant intersubunit interactions seen in many other PBGS octamers, the tail swap seen in TgPBGS is the first example of an intersubunit β-sheet. Nearly all other intersubunit interactions occur in helical or coiled regions and all involve amino acid side chains. Because β-sheet formation involves only backbone hydrogen bonds, lack of sequence conservation in this region does not preclude similar β-sheet formation in other PBGS that have a C-terminal tail.

TgPBGS Octamer

Although all PBGS proteins are believed to be capable of forming an octamer, TgPBGS is the first crystal structure to contain a full octamer in the asymmetric unit. The crystallographic asymmetric units of octameric yeast PBGS (PDB codes 1AW5, 1H7N, 1H7R, 1QNV, 1EB3, 1HO7, 1OHL, 1W31, 1GJP, 1H7P, 1QML, and 1YLV) are all monomers; those of E. coli PBGS (PDB codes 1B4E, 1I8J, 1L6S, and 1L6Y) are monomers or dimers; those of P. aeruginosa PBGS (PDB codes 1B4K, 1GZG, 2WOQ, 2C13, 2C14, 2C15, 2C16, 2C18, and 2C19) are dimers except for one, which is a monomer; that of human (PDB code 1E51) is a dimer; those of C. vibrioform PBGS (PDB codes 1W1Z and 2C1H) are dimers; and those of mouse PBGS (PDB codes 2Z0I and 2Z1B) are dimers or tetramers (23). With eight individual subunits in the TgPBGS crystallographic asymmetric unit, we are provided the opportunity to compare the subunits of the entire oligomer and to potentially witness variations in side chains or backbone conformations that reflect motions available to the protein. In fact, superposition of the eight subunits of TgPBGS shows that the backbone atoms are essentially superimposable (intersubunit Cα root mean square deviations are in the range of 0.152–0.170 Å), and there is some minor variation in the side chains (intersubunit root mean square deviations for all atoms are in the range of 0.182–0.197 Å) (Fig. 4e). Unlike some PBGS structures, which show significant variation in thermal factors between the subunits of a dimeric asymmetric crystallographic unit, the thermal factors of TgPBGS are consistent between all subunits (Fig. 4e).

The assembly of the TgPBGS octamer can be described in terms of the N-terminal arm-hugging interaction and the C-terminal tail-swapping interaction (Fig. 4, b and c). Subunit A swaps tails with C; C hugs D; D swaps tails with H; H hugs G; G swaps tails with F; F hugs E; E swaps tails with B; and B hugs A, thus forming the complete octamer with these two types of subunit interactions, as illustrated by the paper model (supplemental Fig. 2). Although these interactions are sufficient to describe TgPBGS assembly, each tail-swapping interaction of TgPBGS is accompanied by an interaction where the N-terminal arm of one subunit nestles into the base of the αβ-barrel of the partner subunit. For those PBGS that do not have the tail-swap interaction, this arm-to-base of barrel interaction is necessary for assembly of either the octamer or the hexamer (Fig. 2). Experimentally eradicating this interaction in human PBGS by the mutation W19A resulted in a dimeric protein that could make neither octamer nor hexamer (24).

Comparison of the human pro-octamer dimer (Fig. 2, which is derived from adjacent asymmetric units of PDB code 1E51) and the TgPBGS pro-octamer dimer (Fig. 4c) shows that the C-terminal tail-swapping of TgPBGS provides significant additional stability to the pro-octamer dimer. The C-terminal tail also provides stabilization for the pro-octamer dimer by interacting with the N-terminal arm (at amino acids 41, 43, and 44) as well as the αβ-barrel (at amino acids 315 and 319) of the partner subunit (Fig. 4d). The C-terminal tail does not provide intersubunit interactions outside that of the pro-octamer dimer. Pro-octamer dimer stabilizing interactions facilitated by the C-terminal tail prevent the conformational change required for formation of the pro-hexamer dimer and hence prevent hexamer assembly (Fig. 2). If the pro-octamer dimer is physiologically relevant, the inability to traverse from pro-octamer to pro-hexamer conformations explains why TgPBGS does not sample the hexameric assembly. If the hugging dimer is relevant physiologically, we cannot make a structural argument for why TgPBGS does not sample the hexameric assembly. Thus, we conclude that the pro-octamer dimer, rather than the hugging dimer is the physiologically relevant dimer of TgPBGS, which is seen as a low percentage of the oligomeric equilibrium in the purified protein (13). The mutant form of TgPBGS lacking the C-terminal 13 amino acids, and hence the entire β-strand structure, fails to form octamers and remains as dimers that are unstable and prone to precipitation. These dimers most likely exist as the physiologically irrelevant hugging dimer.

TgPBGS Structure as It Relates to PBGS-catalyzed Reaction

Like most αβ-barrel enzymes, the PBGS active site is in a pocket near the center of the αβ-barrel where eight loops connect the β-strands to the α-helices (25). Characteristically, there is an active site lid that gates entry and exit of substrates and products. In the human PBGS structure (PDB code 1E51) and the P. aeruginosa PBGS structure (PDB code 1B4K), the active site lid region is ordered in subunit A and disordered in subunit B of the dimeric asymmetric crystallographic units. By homology with these and other PBGS crystal structures that contain disordered lids, the active site lid of TgPBGS is roughly between residues 222–240 (Fig. 3). PBGS catalysis occurs in a stepwise fashion (26, 27). When the lid is open, P-side ALA (Fig. 1) binds first and forms a Schiff base intermediate to Lys-267, the water formed diffuses away and then A-side ALA binds. There is substantial evidence that the PBGS-catalyzed reaction also may involve Schiff base formation between A-side ALA and Lys-213 (e.g. Refs. 28–30). Interactions between the A-side carboxylate and active site lid residues secure a closed lid conformation, after which product is formed. However, data on the exact order of bond making and bond breaking involved in product formation remain unresolved. The TgPBGS crystal structure is the first to contain a fully formed porphobilinogen product in the active site (Fig. 6a). There are no water molecules within 4 Å of enzyme-bound porphobilinogen, but several within a 10 Å radius. The numbers of active site waters modeled in subunits A–H vary between 3 and 11, though most active sites are associated with between five and eight water molecules (data not shown). Despite this variation, the heterocyclic product porphobilinogen is present in all eight active sites, thus showing no evidence for the half of the sites reactivity that is apparent in some other PBGS structures, notably that of human (PDB codes 1E51 and 1PV8) and P. aeruginosa (PDB code 1B4K), which show ligand occupancy in one-half of the subunits (2, 6, 31). The root mean square deviation between the eight TgPBGS active sites, within 10 Å of porphobilinogen, is very small and provides no significant information on the molecular motions involved in catalysis. Previous structures from yeast and human PBGS (PDB codes 1AW5 and 1E51), which showed a porphobilinogen-like heterocyclic active site ligand, include evidence for the persistence of the bond between C4 derived from P-side ALA and an active site lysine, shown for yeast PBGS in Fig. 6b. Each of these structures shows the amino group of the product-like molecule as a direct ligand to an active site zinc ion, which is not present in TgPBGS (Figs. 3 and 6a). One obstacle to a uniform view of the PBGS-catalyzed reaction mechanism is the likely mechanistic difference between those PBGS that use an active site zinc, and those, like TgPBGS, which do not.

FIGURE 6.

A comparison of PBGS active site structures. a, the TgPBGS active site is illustrated with heteroatoms colored cpk; the carbons of porphobilinogen are green; active site lid residues are labeled with an asterisk. b, the yeast PBGS active site (PDB code 10HL) is colored as in a with the carbons of the heterocyclic reaction intermediate colored dark yellow, and the active site zinc shown as a maroon sphere; active site lid residues are labeled with an asterisk.

The closed-lid conformation is an important aspect of the PBGS catalyzed reaction, presumably due to the necessity for deprotonating the amino group of P-side ALA for formation of the carbinolamine intermediate that must form at C4 of A-side ALA (Fig. 1). Lid closure following A-side ALA binding in TgPBGS is facilitated in part by hydrogen-bonding interactions between the A-side ALA carboxylic acid group and residues on the active site lid (Arg-223 (two bonds), Lys-236, and Gln-240) (Fig. 6a). There is an interesting phylogenetic variation at the lid residue corresponding to Lys-236, which is an arginine in all PBGS that use an active site zinc, even though the zinc does not directly interact with any active site lid residues. The current structure supports a prior hypothesis that the cognate amino acid (Lys or Arg) interacts with the A-side ALA carboxylate (compare Fig. 6, a and b) (12). The closed conformation of the active site lid also provides intersubunit interactions that are present in the octamer but not present in the pro-octamer dimer (or the hexamer, were it to exist). These interactions are between residues of the active site lid of one subunit and the αβ-barrel of a neighboring subunit (Fig. 7). For this reason, addition of substrate to a mixture of PBGS assemblies promotes formation of the octamer (32, 33). In TgPBGS, Arg-27 of an N-terminal arm of one subunit forms a hydrogen bond to Tyr-239 of the active site lid of a neighboring subunit; similarly Arg-24 H-bonds to Thr-238 (Fig. 7). These lid-to-arm interactions can be viewed as occurring within the two subunits of one hugging dimer or alternatively as between two adjacent pro-octamer dimers of the octamer. Thermal factors for both backbone and side chain atoms indicate that the TgPBGS active site lid in the product-bound state is no more mobile than other surface components of this structure.

FIGURE 7.

The allosteric magnesium ion-binding site. a, two allosteric magnesium ions (green spheres and red arrows) are shown at the intersection of two pro-octamer dimers (green-blue and yellow-gray). b, a crossed-eye stereo image illustrating the details of the allosteric magnesium binding site indicated by the larger red arrow in a. Carbons are colored according to their subunit (blue or yellow). Dark blue spheres are the first coordination sphere water molecules. Light blue spheres are additional water molecules. Also illustrated are the nearby intersubunit interactions between the active site lid of the yellow subunit and the N-terminal arm of the blue subunit.

Metal Ions of PBGS

PBGS is a metalloenzyme with both active site and allosteric metal ion-binding sites, all of which have been variously interpreted on the presumption that there is some commonality of need for these metals among PBGS throughout evolution. A working model divides the PBGS into four classes depending on whether or not a PBGS uses zinc at the active site and whether or not it uses magnesium at an allosteric site (12). The active site zinc ion binds to a cysteine rich cluster (Figs. 3 and 6b) and the spatially distinct allosteric magnesium binds to one aspartic acid (Glu-252 of TgPBGS) and five water molecules with second sphere interactions involving a neighboring subunit (Fig. 7). For those PBGS that do not have the cysteine rich site, there is instead an aspartic acid-rich site (Fig. 3) (34). Those PBGS that do not have the allosteric magnesium site have an arginine residue whose guanidinium group is spatially equivalent to the magnesium ion, and which has been shown to stabilize the octamer (24). TgPBGS does not have the cysteine-rich site but has the allosteric magnesium binding site (as defined by comparative sequence information); this is characteristic of PBGS from eukaryotes in the phyla Alveolata, Rhodophyta, Crytophyta, Glaucocystophyceae, Stramenopiles, and Verdiplantae and a subset of bacteria (12).

Active Site Metal Ions of PBGS

The role of the active site zinc ion is in facilitating the binding and reactivity of A-side ALA (35). For those PBGS that use an active site zinc, this metal ion is still present at the end of the reaction, as is seen in the crystal structures of yeast and human PBGS (PDB codes 1OHL and 1E51) that contain the covalent heterocyclic product-like intermediate (Fig. 6b). There is a common understanding that those PBGS that do not use the catalytic zinc may instead use a catalytic magnesium ion. Some kinetic data, including data we recently published for TgPBGS, support this notion (13). However, the TgPBGS crystal structure, with the enzyme-bound product, does not contain magnesium at the active site (Fig. 6a). The presumed “specificity switch” in active site metal ion binding is based on the replacement of the cysteine-rich zinc-binding region of some PBGS by an aspartic acid-rich putative magnesium-binding region (Fig. 3) (34). It is well documented that three cysteines of this region are essential for binding a catalytic zinc ion, which was first shown in the crystal structures of yeast PBGS (e.g. Fig. 6b). However, some other PBGS structures from yeast, human, and mouse PBGS show these cysteines to be disordered, and zinc is not observed. The human PBGS octamer structure (PDB code 1E51) contains zinc in the subunits that contain the substrate-derived heterocycle, but not in the other subunits. This suggests that the active site zinc ion may bind and release during the PBGS-catalyzed reaction, thus acting more like a co-substrate with the A-side ALA (36). The cognate cluster of aspartic acid residues are present in PBGS crystal structures from the bacterial species P. aeruginosa (PDB codes 1B4K, 1GZG, 2C13, 2C14, 2C15, 2C16, 2C18, 2C19, and 2WOQ) and Chlorobium vibrioforme (PDB codes 1W1Z and 2C1H) and the current eukaryotic TgPBGS structure (PDB code 3OBK). Of the nine P. aeruginosa PBGS structures containing the wild-type sequence in this region, most do not show magnesium at the active site, consistent with kinetic data that suggest that basal P. aeruginosa PBGS activity is metal ion-independent (37). Exceptions include a sulfone-containing inhibitor where the sulfone moiety contributes to active site magnesium binding (PDB code 2C18) and a structure with the inhibitor aleramycin (PDB code 2WOQ), where the modeled magnesium has an unusual pentacoordinate geometry and is not liganded to the inhibitor. In these structures, the aspartic acid residues, which are spatially equivalent to the aforementioned cysteines, participate in magnesium coordination. Neither of the C. vibrioforme PBGS structures shows an active site metal ion. The current TgPBGS structure is the least ambiguous of those that contain the aspartic acid-rich sequence because TgPBGS contains the natural product porphobilinogen bound at each of the eight enzyme active sites. In this case, the aspartic acids of this region are nestled close to the product, leaving no room for an intervening metal ion (Fig. 6a). Because there is no active site magnesium present at the end of the reaction, the open question is whether an active site magnesium is present at any point in the reaction, as is suggested from our published kinetic data (13).

Allosteric Metal Ion of PBGS

The allosteric magnesium ion-binding site of PBGS is present in all archaea, nearly all bacteria, photosynthetic eukarya, and some nonphotosynthetic eukarya but not in metazoa or fungi (12). All PBGS crystal structures from the species P. aeruginosa, C. vibrioforme, and E. coli contain an octahedrally coordinated metal ion, almost always interpreted as magnesium. This magnesium is most closely associated with the αβ-barrel of one subunit and contains second and third sphere ligands that come from the N-terminal arm of a neighboring subunit, thus bridging two adjacent pro-octamer dimers (shown in Fig. 7a for TgPBGS). In all cases, the first coordination sphere ligands to this metal ion include one glutamic acid, derived from the αβ-barrel, and five water molecules, as illustrated for TgPBGS (Fig. 7b). Experimental data on PBGS from P. aeruginosa, E. coli, and the green plant pea (for which there is no crystal structure) demonstrate that removal of this magnesium favors octamer disassembly and formation of smaller assemblies (2–3, 32). For TgPBGS, removal of magnesium does not favor octamer disassembly (13). Nevertheless, considering both first and extended sphere ligands, the allosteric magnesium ion of TgPBGS bridges adjacent pro-octamer dimers and would thus be predicted to facilitate the assembly of octamers from pro-octamer dimers during the folding and assembly of TgPBGS. Note that regardless or whether or not magnesium is required for octamer formation, it is appropriate to define the magnesium as allosteric because it binds at a site distant from the active site, but changes the kinetic parameters of the enzyme (38, 39).

Effect of Magnesium on TgPBGS Kinetic Parameters

We have previously described how the kinetic behavior of an equilibrium of alternate quaternary structure assemblies, also called morpheein forms, with different K_m values is best fit to a double hyperbolic equation reflecting a mixture of kinetically distinct species catalyzing the same reaction (40). Most PBGS under optimal conditions that support the octamer show a K_m value in the range of 150 μm. This characteristic K_m reflects binding of the second substrate, which is the A-side ALA, whose binding is stabilized by interactions with the active site lid (Figs. 1 and 6). In turn, the closed conformation of the active site lid allows interactions between lid residues and a neighboring N-terminal arm; for TgPBGS, these are between Tyr-239 and Arg-27 and between Thr-238 and Arg-24 (Fig. 7b). These interactions, which are part of the hugging interaction, are possible in the octamer, but would not be possible in the hexamer or in isolated pro-octamer or pro-hexamer dimers (Fig. 2). In keeping with this explanation, the hexameric human PBGS (e.g. variant F12L), and E. coli PBGS in the absence of the allosteric magnesium ions exhibit K_m values in the range of 2–20 mm (2, 41). In other words, without the octamer-specific interactions that help stabilize the closed conformation of the active site lid, the lid cannot close to stabilize A-side ALA binding, and the K_m value is larger. Human PBGS, at pH values that reflect a mixture of octamer and hexamer, exhibit double hyperbolic kinetics reflecting both K_m values (4). Consistent with this rationale, the crystal structure of the human PBGS hexamer (PDB code 1PV8) contains disordered active site lids in both subunits of the asymmetric dimer, even though one subunit contains a substrate derived putative reaction intermediate (2).

We have applied these criteria to interpret the kinetics of TgPBGS and looked for evidence of both low and high K_m values. Under optimal conditions at pH 8, in the presence of 10 mm magnesium, and varying substrate from 3 μm to 10 mm, TgPBGS kinetics fit to a simple hyperbolic equation revealing a K_m of 176 ± 9 μm and a V_max of 22.7 ± 0.3 μmoles product per hour per milligram protein. This is consistent with previously published kinetic data and our prior deduction that the magnesium-containing form of TgPBGS in solution is an octamer (13). In the absence of added magnesium, the kinetic data also fit well to a hyperbolic equation yielding a K_m of 481 ± 31 μm and a V_max of 11.7 ± 0.2 μmol product per hour per milligram protein (R² = 0.9960). Although a marginally improved fit is obtained to a double hyperbolic where the derived K_m values are similar to the observed K_m for the human PBGS octamer (closed active site lid) and human PBGS hexamer (mobile active site lid), the minor improvement in the fit is insufficient to argue for multiple kinetically competent TgPBGS assemblies. Instead, the increased K_m of the octamer in the absence of magnesium may simply reflect a looser octameric assembly, one not stabilized by the allosteric magnesium ions, which would require a higher substrate concentration to achieve active site lid closure.

DISCUSSION

In recent years, although our understanding of apicomplexan biology from a genomic perspective has increased by leaps and bounds, functional characterization of apicomplexan proteins, especially at the level of protein structure, has been slow in comparison. Nevertheless, structural characterization of parasite proteins is of immense value for development of novel therapeutics. This has led to collaborative structural genomics efforts, and this study is a result of one such effort. Following up from our earlier biochemical characterization of TgPBGS (13), a parasite plastid-associated enzyme that is essential for heme biosynthesis, we now present its crystal structure highlighting distinct features, which distinguish it from human PBGS. Moreover, structure-function studies on TgPBGS will provide insights on the structure and function of PBGS from other apicomplexan parasites, all of which are very similar in sequence to TgPBGS. One outstanding feature of the PBGS family of enzymes is an unusual quaternary structure dynamic, which involves a reversible interconversion between high activity multimers and low activity multimers. This morpheein equilibrium for PBGS (Fig. 2) can be modulated by binding small molecules to allosteric sites (33, 42). Allosteric small molecule stabilization of the hexameric assembly of PBGS from the green plant pea and from humans has been accomplished using a computational docking approach wherein the target was a hexamer-specific surface cavity. The phylogenetically variable nature of this allosteric site resulted in the discovery of species-selective hexamer-stabilizing inhibitors called morphlocks. Species-selective modulation of TgPBGS activity is a long term goal, and the current work lays the foundation for such studies.

The TgPBGS crystal structure, particularly the unique structure of the C-terminal tail swap, provides a structural explanation for why TgPBGS forms the octamer but not the hexamer. Nevertheless, kinetic properties such as a protein concentration-dependent specific activity, which we have previously demonstrated for TgPBGS, argues that the protein exists as an equilibrium of low activity smaller multimers and high activity larger multimers (13). Previous work demonstrated that the low activity smaller multimer is a dimer and the high activity larger multimer is an octamer, but the structure of the dimer remained undetermined. The current work argues strongly for the pro-octamer dimer as the physiologically relevant low activity multimer of TgPBGS. This pro-octamer dimer, with its phylogenetically variable surface composition, is a potential target for future studies directed at finding species selective TgPBGS inhibitors that will function by binding to the dimer and preventing formation of the active octamer.

The C-terminal tail swap that occurs between partner subunits of a pro-octamer dimer in TgPBGS prevents the conformational rotation required for switching from a pro-octamer to pro-hexamer dimer. This observation predicts that other PBGS proteins that contain a C-terminal tail would not sample the hexameric assembly. These are PBGS from all apicomplexan parasites, including the Plasmodium species that cause malaria and the bacterial pathogen that causes cholera (Fig. 3 and supplemental Fig. 1). Consistent with this prediction, we have determined that PBGS from Vibrio cholerae does not sample the hexameric assembly.⁴ Thus, the TgPBGS structure provides the first template for preparation of a complete homology model of PBGS from the Vibrio and Plasmodium species. Although docking to a protein homology model is less likely to yield allosteric effector molecules relative to docking to a protein crystal structure, our previous success at finding morphlocks used a protein structure model (pea PBGS) in one case and an x-ray crystal structure (human PBGS) in the other (33, 42). Thus, the significant structural differences among these essential enzymes provide exciting possibilities for the development of probes of biological function and perhaps even for targeting tetrapyrrole biosynthesis for antimicrobial drug discovery.