Structure of a bacterial pyridoxal 5′-phosphate synthase complex

Abstract

Vitamin B₆ is an essential metabolic cofactor that has more functions in humans than any other single nutrient. Its de novo biosynthesis occurs through two mutually exclusive pathways that are absent in animals. The predominant pathway found in most prokaryotes, fungi, and plants has only recently been discovered. It is distinguished by a glutamine amidotransferase, which is remarkable in that it alone can synthesize the cofactor form, pyridoxal 5′-phosphate (PLP), directly from a triose and a pentose saccharide and glutamine. Here we report the 3D structure of the PLP synthase complex with substrate glutamine bound as well as those of the individual synthase and glutaminase subunits Pdx1 and Pdx2, respectively. The complex is made up of 24 protein units assembled like a cogwheel, a dodecameric Pdx1 to which 12 Pdx2 subunits attach. In contrast to the architecture of previously determined glutamine amidotransferases, macromolecular assembly is directed by an N-terminal α-helix on the synthase. Interaction with the synthase subunit leads to glutaminase activation, resulting in formation of an oxyanion hole, a prerequisite for catalysis. Mutagenesis permitted identification of the remote glutaminase and synthase catalytic centers and led us to propose a mechanism whereby ammonia shuttles between these active sites through a methionine-rich hydrophobic tunnel.

The term vitamin B₆ refers to the compounds pyridoxal, pyridoxine, pyridoxamine, and their phosphorylated derivatives. The vitamin is required for the catalytic function of numerous metabolic enzymes as well as for the upkeep of the nervous and immune systems in animals and has also recently been shown to be a potent antioxidant (1, 2). Two mutually exclusive de novo pathways exist, referred to as deoxyxylulose 5-phosphate (DXP)-dependent and DXP-independent (3). The DXP-dependent pathway is well characterized (4, 5), involving six proteins that catalyze the formation of pyridoxine 5′-phosphate, but is found only in a small subset of the γ-division of eubacteria (6). The DXP-independent pathway is much more predominant, being found in archaea, most eubacteria, fungi, apicomplexa, plants, and even some metazoans, but is less well characterized (2, 6, 7). The pathway is distinguished by the involvement of a glutamine amidotransferase (GATase), known as pyridoxal 5′-phosphate (PLP) synthase (8, 9).

GATases play a central role in metabolism, because they are responsible for the incorporation of nitrogen into amino acids, amino sugars, purine and pyrimidine nucleotides, coenzymes, and antibiotics (10). In general, they consist of a glutaminase domain where glutamine is hydrolyzed, yielding ammonia for use in a subsequent reaction specific for each GATase by the so-called synthase domain. The synthase subunit of PLP synthase, Pdx1, directly synthesizes PLP from either of the pentose sugars, ribose 5-phosphate or ribulose 5-phosphate, and either of the triose sugars, glyceraldehyde 3-phosphate or dihydroxyacetone phosphate, in the presence of the glutaminase subunit Pdx2 and glutamine. Although this pathway is of fundamental importance, its absence from animals may have implications for the discovery of new therapeutic agents (11), and it would greatly benefit from knowledge of the architecture of the enzyme complex. Recently, the 3D structures of the individual Pdx1 (12) and Pdx2 (13, 14) subunits have become available. Based on these structures, two independent hypotheses for the assembled PLP synthase have been proposed. One model (13) proposes an architecture akin to that of imidazole glycerol phosphate synthase (ImGP; refs. 15 and 16), where the synthase and glutaminase active sites are remote from each other. Implicit in this model is that there is a tunnel to channel the nascent ammonia between the disparate active sites. The second model (12) was proposed based on the structure of autonomous Pdx1 from Geobacillus stearothermophilus (named PdxS), where evidence for an ammonia tunnel could not be identified; thus, adjacent glutaminase and synthase active sites were predicted. In this work, we report the structure of the PLP synthase complex, in addition to those of the autonomous enzymes from Bacillus subtilis. This knowledge, in combination with biochemical and biophysical approaches, allows us to provide insight into the complicated mode of regulation and catalysis by an enzyme complex that synthesizes one of nature's most versatile cofactors.

Results

3D Structure of PLP Synthase.

The determination of the 3D structure of PLP synthase necessitates stabilization of the enzyme complex, because the interaction of Pdx1 and Pdx2 is believed to be transient in nature (9), complicating its analysis. Pdx2 is a class I glutaminase categorized by the catalytic triad Cys-79, His-170, and Glu-172 (10). During studies probing the catalytic site, we introduced an H170N mutation (17) into B. subtilis Pdx2 (previously designated YaaE), which resulted in glutaminase activity indistinguishable from the background hydrolysis (Fig. 1a Left) with a concomitant loss in Pdx1 (previously designated YaaD) PLP synthesis activity (Fig. 1a Right). A native PAGE mobility-shift assay as described in ref. 9 was used to probe the nature of the Pdx1:Pdx2 complex in the presence of glutamine (Fig. 1b Lower). In this assay, the formation of the PLP synthase complex can be observed by a decrease in the mobility of Pdx1 (i.e., a mobility shift) due to its interaction with Pdx2. Moreover, Pdx2 can be immunodetected (9) in the shifted band of Pdx1 (Fig. 1b Upper). Although Pdx2 interacts with Pdx1 using wild-type proteins (lane 2), immunodetection revealed a much higher presence of Pdx2 when mutant protein Pdx2 H170N is used (lane 3). A pronounced mobility shift is observed in the native PAGE (compare lanes 1 and 3), with a concurrent loss in the intensity of Pdx2 alone (compare lanes 7 and 3). These features indicated the assembly of a saturatable complex between Pdx1 and Pdx2 H170N and set the stage for crystallization trials of the enzyme complex.

Fig. 1.

Analysis of the B. subtilis Pdx1:Pdx2 H170N:glutamine ternary complex and autonomous Pdx1. (a Left) Glutaminase activity of Pdx2 WT (black), Pdx2 H170N (red) in the presence of Pdx1 WT, and Pdx2 WT in the presence of Pdx1 Δ1–18 (green). (Right) PLP synthase activity of Pdx1 in the presence of Pdx2 WT (black) or Pdx2 H170N (red) and of Pdx1 Δ1–18 in the presence of Pdx2 WT (green), using glutamine (solid lines) or ammonium sulfate (dashed lines) as the nitrogen source. (b) Analysis of complex formation by native PAGE (9). (Lower) Coomassie blue staining. Lane 1, Pdx1; lane 2, Pdx1 and Pdx2, lane 3, Pdx1 and Pdx2 H170N; lane 4, Pdx1 Δ1–18; lane 5, Pdx1 Δ1–18 and Pdx2 H170N; lane 6, Pdx2 WT; lane 7, Pdx2 H170N. (Upper) Western blot. The region of the gel where Pdx1 migrates was probed using a Pdx2-specific antibody (9). Glutamine (10 mM) was present in every case, and 10 pmol of each protein was loaded per lane. (c and d) The Pdx1:Pdx2 H170N:glutamine ternary complex shown in two orientations turned by 90°. Twelve Pdx2 glutaminase subunits attach to a double-ring-like core formed by 12 Pdx1 subunits. The Pdx1 synthase subunits are colored blue and green, and the Pdx2 glutaminase subunits are in orange and red. A single heterodimer is highlighted in dark blue (Pdx1) and red (Pdx2), and the position of bound glutamine is shown in stick representation. The N-terminal β-strand, βN, and α-helix, αN, as well as helix α2′ are highlighted in green. Approximate sizes are given in angstroms (10⁻¹⁰ m). (e) In the structure of the autonomous synthase subunit, Pdx1, the dodecameric assembly of protein subunits is identical to that in the ternary complex but lacking the N-terminal β-strand, βN, as well as helix α2′. Two hexameric rings lie on top of each other (pastel and light blue) and are interconnected by helices α6, α6′, and α6′′ that point into dents of the adjacent ring. Two monomers are highlighted in green and blue in each ring.

The structure of the Pdx1:Pdx2 H170N:glutamine ternary complex has been determined at 2.1-Å resolution [R_work 15.4%, R_free = 20.5%; see supporting information (SI) Table 2]. The current refined structure contains a total of 5,568 amino acids or 42,000 atoms. Twelve Pdx1 synthase subunits form a double hexameric ring, to which 12 Pdx2 glutaminase subunits attach like the cogs of a cogwheel, interacting only with the Pdx1 subunits but not with each other (Fig. 1 c and d; a scheme depicting the complex assembly is shown in SI Fig. 5). The core of the complex, made up of Pdx1, is essentially unchanged from the structure of the autonomous protein (Fig. 1e), also determined at 2.1-Å resolution (R_work 14.4%; R_free 19.4%; see SI Table 2). Pdx1 folds as a (βα)₈-barrel (18), whose axis is oriented nearly perpendicular to the ring perimeter. A bulge created by the helix α8′′ at the C terminus running parallel to helix α8 establishes the contact between adjacent (βα)₈-barrels. The hexameric rings interdigitate in the dodecameric structure. This is accomplished by elongation of helix α6 followed by two inserted helices α6′ and α6′′, which protrude from the N-terminal face of each barrel on one hexameric ring into a corresponding dent in the other (Fig. 1e). We used analytical ultracentrifugation to assess PLP synthase in solution. The molecular masses were determined from the sedimentation velocity data by computing the molar mass distribution c(M) (19, 20). On its own, Pdx1 is present as two species with molecular masses of 186 ± 8 and 370 ± 11 kDa, corresponding to the hexameric and dodecameric forms, respectively (Table 1 and SI Fig. 6a, dotted line). The observed behavior is similar to the one described for PdxS (12). The data indicate that the dodecamer is the dominant species in solution (K_D = 3.7 μM; SI Fig. 6c). When Pdx1 was coexpressed and copurified with Pdx2 H170N, followed by gel filtration in the presence of glutamine, only one species is observed by analytical ultracentrifugation (SI Fig. 6a, dashed line). The molecular mass corresponds to 576 ± 20 kDa and approximates that of a fully occupied Pdx1 dodecameric complex, i.e., with 12 monomers of Pdx2 attached (calculated mass, 636.7 kDa). The same result is obtained when the proteins are purified separately and then mixed in presence of glutamine (SI Fig. 6b, dashed line).

Table 1.

Comparison of measured/calculated sedimentation coefficients

Species	Sexp, S*	Scalc, S†	Mexp, kDa‡	Mcalc, kDa§	f/fo¶
(Pdx1)6	8.7 ± 0.3	8.1	186 ± 8	189.7	1.63
(Pdx1)12	13.7 ± 0.3	13.9	370 ± 11	379.0	1.29
(Pdx1)12-(Pdx2)12‖	17.2 ± 0.8	18.6	576 ± 20	636.7	1.38

*The sedimentation coefficient determined under standard conditions (20°C, H₂O) from the c(s) distribution by using SEDFIT (19). Standard deviations were determined by averaging results from single fits.

^†Calculated sedimentation coefficients using HYDROPRO and 3.3-Å bead size (35).

^‡Molecular mass derived from molar mass distributions c(M).

^§Calculated molecular mass.

^¶Ratio of the friction coefficient (f) to the friction coefficient (f_o) of a sphere.

^‖Pdx1:Pdx2 H170N in the presence of glutamine.

Structural Changes in the Pdx1 Synthase Subunit upon Complex Formation.

The comparison of autonomous Pdx1 and the ternary complex highlights structural changes that occur upon complex assembly, namely formation of an extra N-terminal β-strand βN, an ordering of the N-terminal α-helix αN, and formation of an extra α-helix α2′ (Figs. 1 and 2). The N terminus (residues 2–6) is involved in β-completion with β7/β8 of Pdx2 (Fig. 2). αN after β-strand βN is present in both the ternary complex and the structure of autonomous Pdx1, in contrast to the structure of PdxS (12), where it is absent. In autonomous Pdx1, αN (residues 7–17) has above-average B values, and a number of side chains are disordered. In the ternary complex, however, it is at the center of the interface and is well ordered. Importantly, αN covers the glutaminase active site and is held in place by two areas of polar interaction at either end of this helix, separated by a hydrophobic patch. One polar contact involves the side chain of Arg-8, which becomes ordered to interact with Glu-117 on β7 in Pdx2. Glu-117 also interacts with the side-chain hydroxyl of Thr-6 in Pdx1 (Fig. 2b). The second polar contact involves Lys-18 of Pdx1, which forms a salt bridge to Glu-15 in Pdx2. These two interactions allow for the precise alignment of αN, whereas the side chains of Met-13 and -16 in Pdx1 form an ordered hydrophobic interface with Ile-134 of Pdx2. In the structure of autonomous Pdx1, Arg-8, Met-13, Glu-15, and Met-16 are disordered, and Gln-17 and Lys-18 are in a different orientation. The side chain of Tyr-100 in the turn between α3 and β4 points toward the Met-13/Ala-14 peptide bond in the autonomous Pdx1, whereas in the complex, it forms a hydrogen bond with the carbonyl oxygen of Lys-10 (Fig. 2c). The functional implication of reorienting Gln-17 and Tyr-100 is discussed below in light of the proposed mechanism for the transport of ammonia between the remote catalytic sites.

Fig. 2.

Three major elements become ordered in Pdx1 (blue) upon complex formation with Pdx2 (orange). These are β-strand βN and helix α2′ that form only in the complex as well as α-helix αN (all highlighted in green). (a) αN covers the glutaminase active site, whereas β-strand βN is involved in β completion with β-sheet β7/β8 of Pdx2. The Pdx2 substrate, glutamine, is shown in stick representation, as is the catalytic residue Cys-79 of Pdx2. (b) A comparison with the structure of autonomous Pdx1 (gray) highlights how residues of Pdx1 are ordered and reoriented on complex formation. αN is held in place by hydrogen-bonding networks around side chains Thr-6 and Arg-8 on one side and Lys-18 on the other. (c) View as in b but rotated by ≈180° around the vertical axis, with bound glutamine. The three amino acid side chains of Arg-8, Met-13, and Met-16 become ordered in the complex, whereas Tyr-100 and Gln-17 reorient.

αN, not present in any of the other six class I GATase structures determined to date (21), is instrumental for complex formation. To corroborate its importance, we constructed a deletion mutant of Pdx1 missing the first 18 amino acids, Pdx1 Δ1–18. We then performed sedimentation velocity ultracentrifugation with Pdx1 Δ1–18 and found it does not interact with Pdx2 H170N in the presence of glutamine (SI Fig. 6b, dotted line), whereas native Pdx1 does (SI Fig. 6b, dashed line). Furthermore, native PAGE analysis of complex formation with Pdx2 H170N, as described above with the Pdx1 WT protein, indicated that Pdx1 Δ1–18 could not interact with Pdx2 (Fig. 1b, lane 5). The Pdx1 Δ1–18 mutant was neither able to activate the glutaminase activity of Pdx2 (Fig. 1a Left) nor synthesize PLP in the presence of Pdx2 and glutamine (Fig. 1a Right). However, PLP synthesis was observed if ammonium sulfate was used as the nitrogen source (9), indicating an intact synthase domain (Fig. 1a Right).

Activation of the Glutaminase Subunit.

Because the glutaminase activity of Pdx2 strictly depends on Pdx1 (9), a comparison of the Pdx1:Pdx2:glutamine ternary complex with the autonomous Pdx2 structure can be used to explain glutaminase activation. Pdx2 is a three-layered αβα sandwich with Rossmann topology (10, 13, 14). An important functional segment in Pdx2 is the oxyanion region, between the so-called oxyanion strand β3 and helix α3 (Fig. 3, ochre). The peptide nitrogens of Gly-47 and Ala-80 are expected to form the oxyanion hole, an element necessary to stabilize the transient negative charge that occurs on the glutamine amide oxygen during catalysis (22, 23). The current high-resolution structure of Pdx2 at 1.7 Å (Table 2) reveals details not seen in the previously published 2.5-Å structure (13). In particular, Gly-47 is not correctly poised for oxyanion stabilization, because the peptide nitrogen is pointing away from the glutaminase active site (Fig. 3 a and c). In the ternary complex, interaction with αN of Pdx1 and the presence of glutamine induce a reorganization that leads to the formation of the oxyanion hole (Fig. 3 b and d). This conformation is stabilized by a reorientation of Gln-10, resulting in the loss of the hydrogen bond to the peptide nitrogen of Gly-47 and formation of a new hydrogen bond with the peptide carbonyl of Lys-18 in Pdx1 (Fig. 3 a and b). Concomitantly, the Gly-46/Gly-47 peptide undergoes a peptide flip, such that the peptide nitrogen of Gly-47 now points toward the amide oxygen of the substrate glutamine in the complex (compare Fig. 3 c and d). Thus, oxyanion hole formation is most likely a key regulatory step, analogous to that described for Ser and Cys proteases (24, 25). The conformation seen in the autonomous Pdx2 structure would then represent an enzymatic “ground state,” similar to that observed for Plasmodium falciparum Pdx2 (14), as has also been observed in Thermotoga maritima ImGP (26), where the carbonyl oxygen points toward the substrate-binding site.

Fig. 3.

Activation of glutaminase subunit by a peptide flip in the oxyanion region (ochre). (a) Glutaminase active site in the structure of autonomous Pdx2 showing the catalytic triad, Cys-79, His-170, and Glu-172. Peptide Gly-46–Gly-47 is not poised for catalysis, because the carbonyl of Gly-46 points toward the glutamine-binding site, stabilized by the hydrogen-bonding network involving Gln-10. (b) In the ternary complex with bound glutamine, however, Gly-46–Gly-47 has undergone a peptide flip, such that the oxyanion hole is now in place (orange dotted lines). The peptide is fixed in position by hydrogen-bonding contacts with Gln-10, which is stabilized by Lys-18 of Pdx1. c and d show a larger region as in a and b, respectively, turned by 90°. In the structure of the ternary complex, Ser-75 and Asp-99 of Pdx1 (blue) mediate a reorganization of the substrate-binding region, including residues Glu-48, Arg-106, Asn-107, Asp-114, and Arg-135 of Pdx2. Observation of the ternary complex was facilitated by mutation of His-170 to Asn in Pdx2.

Aside from oxyanion hole formation, a number of changes surrounding the head group of bound glutamine occur in the ternary complex. Residues Glu-48, Arg-106, Asn-107, and Arg-135 in Pdx2 are seen in different orientations compared with the free enzyme, stabilized by interactions with the α2/β3 loop of Pdx1 (Fig. 3d). In particular, we observed that both the carbonyl and the side-chain hydroxyl group of Ser-75 in Pdx1 participate in hydrogen-bonding networks with residues around the active site of Pdx2, either directly or through water molecules. Furthermore, in all GATases examined so far, glutaminase and synthase activity is tightly coupled, mediated by binding of substrate in the synthase domain (21). Coupling has been observed for example in ImGP, and it has recently been proposed that it is mainly attributed to an interdomain salt bridge, which influences the precise alignment of catalytic residues in the active site of the glutaminase (21). This direct coupling, despite the observed structural divergence in synthase subunit architecture, was proposed to be a general feature of class I GATases. However, the recent biochemical characterization of PLP synthase has shown that the synthase substrates do not enhance the glutaminase half reaction (9). Moreover, we now observe that this specific interdomain contact is absent. Interestingly, though, the intramolecular salt bridge between Arg-135 and Asp-114 in autonomous Pdx2 is displaced by intermolecular salt bridges between R111, R135 of Pdx2, and Asp-99 of the α3/β4 loop of Pdx1 (Fig. 3 c and d). This contact may be of significance with respect to communication among the PLP synthase domains.

Discussion

Vitamin B₆ was discovered in the 1940s and has gained immense significance in the interim because of the vital role it plays in all organisms. As a consequence of the genomic era, it became clear only recently that the well studied Escherichia coli pathway is an exception, and most other organisms that synthesize the vitamin do so through an entirely different route (2, 7). In this context, the architecture of the synthesizing complex presented here is highly significant not only for knowledge of the machinery that synthesizes one of our essential nutrients but also for the development of potential antimicrobial or antiparasitic agents due to the absence of the pathway from animals (11, 14, 27, 28). Conversely, although the importance, ubiquity, and chemically distinct nature of GATases have been recognized, PLP synthase has properties clearly distinct from those of the canonical GATase and is the first reported GATase in such a higher oligomeric state (12). Furthermore, it displays a previously undescribed mode of interaction between glutaminase and synthase subunits, mediated by helix αN. Comparison with other class I GATases reveals that the presence of helix αN leads to an incongruent interaction between synthase and glutaminase subunits. PLP synthase and ImGP synthase (15, 16), for example, would require a rotation of 147° accompanied by a movement of 3.5 Å to match the glutaminase subunits of the two enzyme complexes, based on a superposition of the two synthase subunits (SI Fig. 7). Thus, the initial model (13) of the PLP synthase complex based on ImGP could not correctly predict the particular interface that is formed between Pdx1 and Pdx2. The importance of αN in Pdx1 is further emphasized by the crucial role it plays in controlling the activity of Pdx2, in that while directing the mode of interaction of both subunits, it covers the glutaminase active site, thus sequestering the substrate glutamine, and orchestrates glutaminase activation by triggering formation of the oxyanion hole, a prerequisite for catalysis. This phenomenon is clearly demonstrated through the availability of autonomous structures and that of the ternary complex.

PLP synthase is exceptional in that it catalyzes pentose and triose isomerizations, imine formation, amine addition, and ring formation, all in a single enzymatic system (8, 9). Two potential phosphate-binding sites that could be involved in substrate recognition are identified on the synthase subunit Pdx1, and they are referred to as P1 and P2 (Fig. 4a). In the structure of Pdx1, buffer component chloride ions were observed in these positions, whereas in PdxS (12), sulfate ions were found. P1 is on the C-terminal side of the (βα)₈ barrel and comprises a characteristic loop apparently designed for phosphate binding (12) involving the amino acids Gly-Thr-Gly (residues 153–155, GTG loop; yellow in Fig. 4a). Moreover, we have confirmed that residues Lys-81, Asp-24, and Asp-102, in the vicinity of P1, are necessary for catalysis, because mutation to Ala resulted in an enzyme without detectable PLP synthase activity (data not shown). These data together establish the Pdx1 active site around P1 on the catalytic face of the (βα)₈-barrel, as is consistent with placement of active sites for all structurally known enzymes of this fold (18). Furthermore, it is interesting that residues 47–56, which form the small helical segment α2′, are observed only in the ternary complex (Figs. 1c and 2a). This finding might indicate a “priming” of the synthase active site mediated by a signal from the glutaminase subunit.

Fig. 4.

Pdx1 active site and putative ammonia tunnel. (a) Buffer component Cl⁻ ions were observed in the sites marked as P1 and P2 in autonomous Pdx1. The binding site P1 at the C-terminal face of the barrel is characterized by a loop containing the amino acids Gly-Thr-Gly (GTG-loop in yellow) and is in close proximity to the catalytic center, consisting of D24 and K81. Residues H115, R137, and R138 from one subunit and K187 from an adjacent subunit form the P2-binding site. The catalytic residue Lys-149 is shown between P1 and P2. (b) Putative ammonia tunnel probed with a solvent radius of 1.2 Å (green) and 1.45 Å (blue). The cavity seen above the glutaminase active site is visible only in the ternary complex and is created by movement of Tyr-100 and Gln-17 of Pdx1 (compare Fig. 2c). A number of conserved hydrophobic residues line the proposed tunnel (M43, V126, M143, F210, and F233). In our model, we propose that the Mets in particular could change conformation to create a transient passage.

The second site, P2, is at the interface of the hexameric rings. It is characterized by residues His-115, Arg-137, and Arg-138, in combination with Lys-187 from a second Pdx1 monomer (Fig. 4a). A shallow groove lined with charged residues connects P1 and P2 over a distance of 21 Å. Lys-149 (8, 12), at the center of this groove, is pointing toward P2, and by mutation to Ala, we confirmed its catalytic role, because the mutant protein is devoid of PLP synthesis activity (data not shown). The remoteness between glutaminase and synthase active sites raises the question of how the labile ammonia is transferred between them. An elegant way of how this can be accomplished is by channeling through the synthase domain. It has already been shown for ImGP that the core of a (βα)₈-barrel can be used for the transfer of ammonia (15, 16, 33). However, a continuous channel could not be identified in the structure of either autonomous Pdx1 or the ternary complex. Indeed, one of the models proposed for the assembly of the PLP synthase complex was put forward on the basis that in the PdxS structure (12) no obvious channel could be observed, which in turn resulted in the erroneous prediction of adjacent active sites.

Comparative analysis of the 3D structures presented here allows us now to propose a model for ammonia transfer (Fig. 4b). First, as stated above, Gln-17 and Tyr-100 from the synthase subunit Pdx1 change their position in the ternary complex; αN is implied in organizing this structural switch (Fig. 2c). This movement creates a cavity inside the β-barrel, visible when probed with a solvent radius of 1.45 Å (Fig. 4b, blue map). It is plausible that this cavity is part of a passage/tunnel that could connect the two active sites over a distance of ≈22 Å. Among the hydrophobic residues in the core of the Pdx1 β-barrel, a number of Mets are observed (Fig. 4b). In this context, it has been documented that Met residues display a significant conformational plasticity [ref. 29; examples are the M domain of the signal recognition particle (30) or heat-shock proteins (31)]. Thus, we propose here that in Pdx1, consecutive movement of Met-13, -79, -145, and -43 (Fig. 4b) would allow ammonia to penetrate through the core of Pdx1. An additional cavity becomes visible, extending from the Pdx1 active site, when Pdx1 is probed with a solvent radius, of 1.2 Å (Fig. 4b, green map). This passage would have to widen to allow for passage of ammonia, requiring some plasticity of the surrounding residues [as is assumed to happen in the E. coli ammonium transporter with its narrow tunnel passage of only 0.6 Å (32)]. Ammonia would surface in close proximity to the catalytic residues Lys-81 and Asp-102 (Fig. 4). The discontinuity of the proposed channel in the current structure purports its transient nature, the opening of which is likely triggered by substrate binding. If the current proposal holds true, PLP synthase provides another example with a mechanism different from ImGP (15, 16, 33) for the transfer of ammonia, using the universal triosephosphate isomerase barrel fold (TIM). Although the presence of a transient tunnel in the core of the (βα)₈-barrel might represent a destabilizing feature, the oligomeric state of Pdx1 could be an adequate ersatz, because oligomerization provides additional conformational stability (34).

Materials and Methods

Generation of Pdx1 and Pdx2 Constructs and Protein Expression and Purification.

The constructs used in this study are described in ref. 9 and in SI Supporting Text. Expression and protein purification were carried out as described in ref. 9. Native Pdx2 was purified in a two-step protocol using a Source Q strong anion exchanger followed by gel filtration. For the PLP synthase complex, Pdx2-His₆ H170N and native Pdx1 were coexpressed in E. coli BL21 (DE3) cells (Stratagene, La Jolla, CA). Purification was carried out using a two-step protocol of Ni-affinity chromatography and gel filtration in the presence of 10 mM glutamine. Complete information on purification of all proteins is in SI Supporting Text.

Biochemical and Biophysical Protein Assays.

The glutaminase activity assay was performed as described in ref. 9 using a 4 μM concentration of either the wild-type or mutated Pdx1 and Pdx2 proteins. PLP synthase activity was also determined as described in ref. 9, with the exception that 20 μM each of the proteins was used and 0.5 mM ribose-5-phosphate and 1.0 mM dl-glyceraldehyde-3-phosphate were used as substrates. For native PAGE mobility-shift assays, equimolar concentrations of the respective Pdx1 and Pdx2 proteins were preincubated in the presence of glutamine (10 mM) for 15 min at 37°C followed by electrophoresis in 12.5% gels. Western blot analysis was performed as described in ref. 9.

The association states of Pdx1 and Pdx1/2 complexes in solution were determined by hydrodynamic analysis in analytical ultracentrifugation experiments using a Beckman Optima XL-A ultracentrifuge equipped with absorbance optics and an An60 Ti rotor (Beckman Coulter, Fullerton, CA). Sedimentation velocity studies were carried out at 35,000 rpm at 20°C using protein with an absorbance at 280 nm of 1. Partial specific volumes v̄, extinction coefficients at 280 nm, buffer density ρ, and viscosity η were calculated by using the program SEDNTERP, Version 1.05 (J. Philo, D. Hayes, and T. Laue, www.jphilo.mailway.com/download.htm). Calculated partial specific volumes v̄ for both Pdx2 and Pdx1 are 0.742 ml·g⁻¹; the calculated extinction coefficient is 10,930 M⁻¹·cm⁻¹ for Pdx1 and 2,980 M⁻¹·cm⁻¹ for Pdx2. Centrifugation was conducted in 20 mM Tris·Cl, pH 7.5, and 200 mM KCl with ρ = 1.0084 g·ml⁻¹ and η = 1.0045 mPa for WT Pdx1 and in 20 mM Tris·Cl, pH 8.0, and 300 mM KCl with ρ = 1.01291 g·ml⁻¹ and η = 1.0031 mPa for Pdx1 Δ1–18 and Pdx1/2 complexes. Theoretical hydrodynamic parameters were determined from the crystal structure coordinates by creating models composed of spherical beads of varying size with the program HYDROPRO on a Linux-PC using the subroutine Hydropro5amsd (35). The hexamer–dodecamer equilibrium was fitted to the velocity data according to

X-Ray Crystallographic Structure Determination.

The three 3D structures determined in this study were crystallized by using hanging-drop vapor diffusion experiments. Crystals were flash-frozen by using ethylene glycol as cryoprotectant. Data were collected at 100°K at the European Synchrotron Radiation Facility in Grenoble, France, and integrated and scaled with the HKL software (36). All additional data manipulation was carried out with the CCP4 suite of programs (37). The structures were determined by molecular replacement by using the program MOLREP (38) with the coordinate sets of 1ZNN (12) for Pdx1 and of 1R9G (13) for Pdx2. Iterative model building and refinement were carried out with the programs O (39) and REFMAC5 (40) cycled with ARP (41), using TLS refinement in the final cycles. Large crystals of up to 1 × 0.2 × 0.2 mm³ could be grown from Pdx2 at a concentration of ≈20 mg·ml⁻¹ using 12.4% PEG 8000/6% ethylene glycol/100 mM cacodylic acid, pH 6.2, as crystallization buffer. Pdx1-His₆ was concentrated to ≈20 mg·ml⁻¹ for crystallization in 100 mM Tris·Cl, pH 7.8/15–20% ethanol/200 mM MgCl₂. Hexagon-shaped disks of a size of 150 × 150 × 30 μm³ diffracted to 2.6 Å. Orthorhombic crystals occasionally grew under the same conditions and diffracted to 2.2 Å. A Pdx1 dodecamer extends over two crystallographic asymmetric units, with the crystallographic twofold relating two hexamers. The Pdx1:Pdx2 H170N:glutamine ternary complex was concentrated to ≈24 mg·ml⁻¹ for crystallization in 15–17% PEG 4000 buffered with 200 mM triammonium citrate at pH 7.0. Plate-like crystals of a size of 100 × 200 × 50 μm³ diffracted to 2.1 Å. After rigid-body refinement of the initial molecular replacement solution using Pdx1 electron-density maps showed the position and orientation of Pdx2 molecules, and these were docked manually by using O (39). All structural diagrams were drawn by using PyMOL (ref. 42; www.pymol.org).

Abbreviations

GATase

glutamine amidotransferase

ImGP

imidazole glycerol phosphate synthase

PLP

pyridoxal 5′-phosphate.

Note Added in Proof.

While this manuscript was in press, we learned that the x-ray structure of PLP synthase from T. maritima has been accepted for publication (43).