Structure of Aminodeoxychorismate Synthase from Stenotrophomonas maltophilia

Abstract

PabB, aminodeoxychorismate synthase, is the chorismic acid binding component of the heterodimeric PabAB complex that converts chorismic acid to 4-amino-4-deoxychorismate, a precursor of p-aminobenzoate and folic acid in microorganisms. The second component, a glutamine amidotransferase subunit, PabA, generates ammonia that is channeled to the PabB active site where it attacks the C4 carbon of a chorismate derived intermediate that is covalently bound, through C2, to an active site lysine residue. The presence of a PIKGT motif was, until recently, believed to be discriminate PabB enzymes from the closely related enzyme anthranilate synthase, which typically contains a PIAGT active site motif and does not form a covalent enzyme-substrate intermediate with chorismate. A subclass of PabB enzymes that employ an alternative mechanism requiring two equivalents of ammonia from glutamine and that feature a noncovalently bound 2-amino-2-deoxyisochorismate intermediate was recently identified. Here we report the 2.25 Å crystal structure of PabB from the emerging pathogen Stenotrophomonas maltophilia. It is the first reported structure of a PabB that features the PIAGT motif. Surprisingly, no dedicated pabA is evident in the genome of S. maltophilia suggesting that another cellular amidotransferase is able to fulfill the role of PabA in this organism. Evaluation of the ammonia-dependent aminodeoxychorismate synthase activity of S. maltophilia PabB alone revealed that it is virtually inactive. However, in the presence of a heterologous PabA surrogate, typical levels of activity were observed using either glutamine or ammonia as the nitrogen source. Additionally, the structure suggests that a key segment of the polypeptide can remodel itself to interact with a nonspecialized or shared amidotransferase partner in vivo. The structure and mass spectral analysis further suggest that S. maltophilia PabB, like Escherichia coli PabB, binds tryptophan in a vestigial regulatory site. The observation that the binding site is unoccupied in the crystal structure, however, suggests the affinity may be low relative to E. coli PabB.

Stenotrophomonas maltophilia is a complex, versatile, and ubiquitous aerobic Gram-negative bacterium (1). S. maltophilia is abundant in soil and endophytic strains have been isolated from many plants including key crops such as rice and wheat (2). While its association with plants is often beneficial and no phytopathogenic strains have been identified, S. maltophilia is a significant emerging human pathogen (1, 3). Typically a nosocomial concern, S. maltophilia infections are often associated with cystic fibrosis (CF)¹, severe immunosuppression, and cancer related hematologic disorders (4–7). Mortality rates attributable to S. maltophilia bacteremia range from 26 percent to over 60 percent (1, 8). The role of S. maltophilia in acute and long-term CF related lung disease remains somewhat controversial. It is clear that S. maltophilia is increasingly common in the CF lung flora; however, studies have reached disparate conclusions on the question of whether the organism negatively impacts lung function (5, 9, 10). Recent data, however, has shown that S. maltophilia may enhance the persistence of Pseudomonas aeruginosa infections via interspecies signaling and that S. maltophilia is likely a contributor in its own right to the long-term inflammatory damage to the lungs of CF patients (5, 11).

Much of the concern surrounding the emergence of S. maltophilia stems from its intrinsic antibiotic resistance. β-lactamase production, sophisticated multidrug efflux systems, biofilm formation, and other factors render many antibiotics ineffective against S. maltophilia (1, 12, 13). Trimethoprim-Sulfamethoxazole (TMP-SMX), an agent often disfavored due to toxicity and allergic concerns, is considered the primary therapeutic option for S. maltophilia infections (14, 15). The recent isolation of pandrug-resistant S. maltophilia exemplifies the severe threat posed by drug resistant Gram-negative bacteria (16).

TMP-SMX targets two enzymes in the folate pathway, dihydropteroate synthase and dihydrofolate reductase (17). The effectiveness of TMP-SMX illustrates that disabling folate production in S. maltophilia remains, in principle, a viable strategy. Our familiarity with the folate pathway and a curious mechanistic observation prompted us to investigate the details surrounding the synthesis of the folate precursor p-aminobenzoate in S. maltophilia (18, 19). The conversion of chorismic acid to p-aminobenzoate in bacteria typically involves three enzymes, PabA, -B, and -C. PabB catalyzes the amination of chorismate yielding 4-amino-4-deoxychorismate (ADC) and is a member of a family of structurally similar chorismate-utilizing enzymes that also includes anthranilate synthase (TrpE), isochorismate synthase, and salicylate synthase. PabA is an amidotransferase that supplies ammonia from glutamine hydrolysis and PabC is a pyridoxal phosphate-dependant enzyme that catalyzes the elimination of pyruvate from ADC, forming p-aminobenzoate (18). PabA and PabB must interact, at least transiently, for glutamine-dependant catalysis to occur (18). PabB, however, can use exogenous ammonia and does not strictly require PabA in order to produce ADC. Mechanistically, PabB enzymes are typically characterized by an active site lysine that attacks C2 of chorismate displacing the 4-OH group yielding a transient covalent intermediate (18, 20). The active site lysine is then displaced by nucleophilic attack of ammonia at C4 of the enzyme bound intermediate (Figure 1). Recently, however, Schadt et al. described a noncovalent mechanism involving a 2-amino-2-deoxyisochorismate (ADIC) intermediate that was unappreciated but is apparently characteristic of PabB from several organisms including Bacillus subtilis, the bioterror agent B. anthracis, and the emerging pathogen S. maltophilia (19). Instead of an active site lysine carrying out the initial attack on chorismate, in these enzymes, it is another ammonia molecule that attacks C2 of chorismate and displaces the 4-OH group. An alanine occupies the position analogous to that of catalytic lysine in the canonical PabB enzymes.

Figure 1.

Two schemes for the conversion of chorismate to ADC and the incorporation of p-aminobenzoate into folates. (A) PabB enzymes such as E. coli PabB that contain the ²⁷²PIKGT²⁷⁶ motif catalyze the formation of a covalent C2-lysyl intermediate that is released by subsequent attack of ammonia at C4. (B) Less common PabB enzymes such as SmPabB that contain a PIAGT motif and require two equivalents of ammonia and produce a noncovalently bound ADIC intermediate. In both cases, PabC converts ADC to p-aminobenzoate in a PLP-dependent elimination reaction. p-aminobenzoate is then incorporated into the folate precursor dihydropteroic acid.

In order to evaluate what structural differences accompany this alternative mechanism and to provide a starting point for structure based drug design efforts aimed at targeting an additional step in folate biosynthesis in S. maltophilia, we have conducted a structural, functional, and bioinformatic evaluation of S. maltophilia PabB (SmPabB). Here we report the 2.25 Å crystal structure of SmPabB. An unexpected trimer of dimers featuring a reorganized PabA interface region forms the basis of an unusually high solvent content crystal lattice. We confirm, using LC/MS that, like the B. subtilis PabB, SmPabB produces ADC via a noncovalent ADIC intermediate. Evaluation of the activity of the enzyme in light of the absence of a dedicated pabA reveals that SmPabB is virtually inactive alone but forms a high affinity complex with E. coli PabA suggesting the observed plasticity in the PabA interface region of SmPabB may promote in vivo accommodation of alternative or nonspecialized amidotransferase partners.

Experimental Procedures

Protein Expression and Purification

A synthetic construct containing S. maltophilia pabB, optimized for overexpression in E. coli, was obtained commercially (Genscript) and subcloned into the expression vector pET28a (Novagen). SmPabB was overexpressed in E. coli strain BL21(DE3) using the autoinduction method of Studier (21). Cells were grown initially at 37°C until the culture density reached an optical density of ~0.3 at 600 nm at which time the temperature of the shaking incubator was reduced to 20°C. The bacterial cells were then harvested by centrifugation after 14-16 h and stored at - 20° C until needed. SmPabB was purified by cobalt ion affinity chromatography essentially as directed by the resin manufacturer (Pierce). Human α-thrombin (Haematologic Technologies) was used to remove the portion of the fusion tag encoded by the pET-28a vector. Thrombin was then removed by passing the mixture over benzamidine agarose. The cleaved fusion tag was removed by a second passage over the cobalt resin. SmPabB judged pure by SDS-PAGE analysis was concentrated to ~13 mg/mL, dialyzed against 25 mM Tris, 1 mM DTT, 1 mM EDTA (pH 7.8) and stored in 0.2 mL aliquots at -80°C. SmPabB shows no particular tendency to aggregate and no noticeable loss of activity or precipitation were observed upon thawing after storage at -80 °C. Expression and purification procedures for E. coli PabA and PabC are described in the supporting information. E. coli PabB was purified as previously described (18).

Mass Spectral Identification of Intermediates and Products in SmPabB Catalyzed Reactions

LC/MS was used to confirm that SmPabB exhibited the predicted activity previously described for the B. subtilis PabB (19). ADIC, ADC, and p-aminobenzoate were positively identified using authentic standards of each compound (22). Reactions containing SmPabB and chorismate were subjected to liquid chromatography using a Waters Acquity UPLC system and a Waters HSS T3 2.1 × 100 mm UPLC column. The binary solvent system used for separation consisted of 0.1% formic acid (solvent A) and 0.1% formic acid in acetonitrile (solvent B). The chromatographic method time was five minutes and the following gradient profile was used: 0-1.5min, 5% B; 1.5-4.5min, a concave gradient (Waters type 7) to 95% B; 4.5-5.0 min, 95% B. The flow rate was 0.4 mL/min. Under these conditions the elution times were: ADC, 1.02 min; ADIC, 1.8 min, p-aminobenzoate, 3.8 min. MS identification was by means of an attached Waters LCT Premiere ESI-TOF mass spectrometer operating in positive ion mode.

Measurement of Enzymatic Activity

The ADC synthase activity of SmPabB was followed using a discontinuous HPLC based assay based on detection of p-aminobenzoate in reactions also containing recombinant PabC (ADC lyase) and in some cases, PabA, both from E. coli. ADC produced by PabB or PabA-PabB mixtures was converted to p-aminobenzoate and pyruvate by excess PabC. Reactions were performed at 25 °C in 20 mM HEPES, 5 mM MgCl₂, 0.1 mM DTT, 0.1 M glutamine, (pH 8.0). When exogenous ammonia was used as the ammonia source, glutamine was omitted from the buffer and 0.1M NH₄(SO4)₂ was included. Additional details for specific assays are given either in the corresponding figure or table legend, or in the supporting information.

Determination of the PabA7B Dissociation Constants by Fluorescence Titration

Steady-state fluorescence emission spectra from 300-400 nm (λ_ex = 290 nm) were obtained at 25 °C in 50 mM Hepes, 5 mM MgCl₂ (pH 8.0) using a Varian Eclipse fluorescence spectrometer. The K_d for the PabA-SmPabB complex was determined from the observed increase in tryptophan fluorescence as a function of increasing PabA concentration. The change in fluorescence at 340 nm was plotted as a function of PabA concentration and the data were fit to a quadratic plus linear function (Equation 1) where A represents the titration amplitude, S represents the offset from zero, [L] the PabA concentration, n the stoichiometry, and [P] the monomer concentration of SmPabB.

$$\begin{array}{cc}\hfill \Delta F=S+\frac{A^{\ast }\sqrt{{(K_d\left[P\right]+n+\left[L\right])}^2-4n\left[P\right]\left[L\right]}}{2n\left[P\right]}+S\left[L\right]\hfill & \hfill \hfill \\ \hfill \hfill & \hfill ⍰\hfill \end{array}$$ (1)

Molecular Weight from Laser Light Scattering

The solution molecular mass of SmPabB was determined by a combination of laser light scattering and interferometric refractometry. 20 μL samples were injected onto a Shodex KW-802.5 gel filtration column attached to an Agilent 1100 series HPLC. Light scattering and concentration measurements were made using a DAWN EOS 18-angle light scattering detector (Wyatt Technology) and an OPTILAB DSP interferometric refractometer (Wyatt Technology). The column was equilibrated in 20 mM KH₂PO₄, 100 mM KCl, 0.2 mM EDTA, and 0.02% sodium azide (pH 6.5). A refractive index increment of 0.185 mL/g was used to estimate concentrations for molecular weight estimates, and bovine serum albumin was used as an isotropic scatterer for detector normalization to compensate for slight differences in the electronic gain of the 18 detectors. Molecular mass calculations were performed using ASTRA software (Wyatt Technology).

Crystallization

Initial crystallization screens of PabB were performed at ambient temperature using the sitting drop vapor diffusion method with commercially available screening kits. The protein solution was typically ~13 mg/mL prior to screening. Equal volumes of protein solution and reservoir solution, typically 2-4 μL total volume, were mixed for each experiment. Initial screening hits were optimized until, ultimately, a mixture containing 0.05 M MgCl₂·8H₂O, 0.1 M HEPES, 12-15% w/v polyethylene glycol monomethylether 550, 0.1% CHAPS (pH 7.8) yielded the best results. Bipyramidal shaped (~0.1× 0.2× 0.4 mm) crystals were typically obtained after ~2 days and were cryopreserved in liquid nitrogen after a brief exposure to artificial mother liquor containing 30% polyethylene glycol monomethylether 550.

Collection and Processing of Crystallographic Data

Diffraction data for SmPabB was collected at National Synchrotron Light Source, Brookhaven National Laboratory on beamline X29 using a wavelength of 1.0809 Å. Crystals belonged to the P6₃22 space group with one monomer per asymmetric unit and they diffracted to 2.25 Å. Diffraction data were scaled and processed using HKL2000 (23). Data collection statistics are presented in Table 1.

Table 1.

Data collection and Refinement Statistics

	SmPabB
PDB Code	4GRH
space group	P6322
cell parameters (a, b, c) (Å)	161.3, 161.3, 116.1
Resolution (highest shell), (Å)	46.60-2.25 (2.33-2.25)
no. of measured intensities	1,804,918
no. of unique reflections	42,573
redundancy	42.4
completeness ( %) (last shell)	100.0 (100.0)
I/σ(I) (last shell)	20.1 (4.5)
Rmerge % (last shell)a	6.2 (70.9)
refinement statistics
resolution range (Å)	46.60-2.25
no. of reflections used	40,399
no. of protein atoms	3416
no. of waters	226
R-factorb	0.195
R free c	0.240
rmsd bond lengths (Å)	0.019
rmsd bond angles (deg)	1.98
average B-factors for main chain/side chain (Å2)	45.86/49.80

^aR_merge = Σ |I – (I)|/ΣI, where I is the intensity of an observed reflection and (I) is the average intensity of multiple observations.

^bR = Σ∥F_obs| – |F_cal∥/Σ|F_obs|.

^cR_free = Σ∥F_obs| – |F_cal∥/Σ|F_obs|, where F_obs is from a test set of reflections (5 % of the total) that are not used in structural refinement.

Structural Solution and Refinement

The structure of SmPabB was solved by molecular replacement using the program Phaser (24). The search model was PabB E. coli (PDB code 1K0G). The initial model was rebuilt using PHENIX auto builder to match the SmPabB sequence (25). The refinement procedure included periodic examinations of residues and subsequent refitting using the graphics program COOT (26). The model was completed and refined by using COOT for expanding the model and REFMAC5 for refining between building sessions (27). Solvent molecules were identified and included as the refinement progressed. The final model includes one magnesium ions, two CHAPS detergent molecules, and four PEG fragments, one of which is in the exposed active site. No density was observed for residues 286-297 or for residues 453-454 at the C-terminus. Refinement parameters and statistics are summarized in Table 1. The refined coordinates have been deposited in the Protein Data Bank as entry 4GRH.

Assay for Tryptophan Binding

LC/MS was used to assess whether purified SmPabB preparations contained noncovalently bound tryptophan. A 50 μL sample of SmPabB identical to those used for crystallization experiments was thawed and diluted tenfold with an aqueous mixture of 50% acetonitrile and 1% formic acid. After passage through a 10 kilodalton cutoff filter by centrifugation, 2 μL aliquots were analyzed by LC/MS. Similar analyses were conducted using E. coli PabB, E. coli MenF, and the ADIC synthase PhzE from Pseudomonas aeruginosa, all of which were previously purified in our lab (18, 28). Purification of PhzE described in supporting information.

Isothermal Titration Calorimetry

The interaction between SmPabB and tryptophan was analyzed using a Microcal iTC200 calorimeter. The titrant syringe contained 1 μM tryptophan and the cell contained 100 μM SmPabB. The cell was maintained at 25°C. An initial 0.5 μL injection and 34 1 μL injections were made. The spacing between injections was 180 s. Data were analyzed using Origin 7 software as described in the legend to figure S3.

Results and Discussion

Bioinformatic Analysis of p7Aminobenzoate Biosynthesis in S. maltophilia

While a few bacteria can acquire folates from their environment, most bacteria must make their own folates in order to survive and grow (29). The effectiveness of TMP-SMX against S. maltophilia illustrates that folate production is a required function in this organism. S. maltophilia K279a, a clinical isolate, has been sequenced and much of the folate biosynthetic machinery, including pabB (Smlt1032), is readily identifiable (12). Confusion, however, surrounds the identity of PabA and PabC (ADC lyase) in S. maltophilia. There is apparently no dedicated pabA gene in the S. maltophilia K279a genome raising the question of how the organism makes ADC and p-aminobenzoate for folate production. Furthermore, the gene adjacent to pabB is erroneously annotated as pabC (Smlt1034) in S. maltophila. The misidentification of Smlt1034 homologs as ADC lyase has, in fact, become entrenched throughout bioinformatic databases. The PDB, for example, features the structure of an Smlt1034 homolog from E. coli (2R1F; 46% identical to Smlt1034) that was deposited as “predicted ADC lyase from E. coli” despite there being no evidence to support the annotation and despite the fact that the structure of the true PabC from E. coli has been in the PDB for nearly a decade (1I2K and others) and bears no resemblance to 2R1F. While the identity of the S. maltophilia PabC remains uncertain, we have identified Smlt1373 as a candidate. It is 27% identical (E = 0.005) to the legitimate PabC from P. aeruginosa (PDB code: 2Y4R) but not detectably similar (using the same BLAST algorithm parameters) to the E. coli PabC (1I2K). Despite the incongruous BLAST results, many of the key PabC active site residues can be identified in Smlt1373 including K140, the residue that forms the Schiff base linkage with the required PLP cofactor.

In terms of the identity of PabA, one possibility is that S. maltophilia uses TrpG (Smlt4311), its anthranilate synthase glutamine amidotransferase subunit, in a second role as the ammonia donor for p-aminobenzoate biosynthesis. A similar dual function amidotransferase has been described previously in Bacillus (30). Interestingly, the putative dual function amidotransferase in B. subtilis (BSU00750) is encoded adjacent to pabB rather than being located in the trp operon. Alternative and less likely possibilities are that SmPabB functions with an unrecognizable PabA partner or has a novel ability to generate ammonia from glutamine without aid of another polypeptide. There are example of fused ADC synthases, both PabAB and PabBC fusions are known, however, at just 49 kDa SmPabB does not appear to be among these.

In order to conduct a timely analysis of SmPabB to complement the reporting of the structure, we elected to use PabA and PabC from E. coli in assays of ADC and p-aminobenzoate production by SmPabB.

Overall Structure of SmPabB and Crystal Packing

SmPabB is a 454 amino acid, ~49.5 kDa protein that exhibits the complex α/β fold characteristic of the family of chorismate utilizing enzymes that, in addition to PabB (Figure 2A), includes anthranilate synthase (TrpE), salicylate synthase, and isochorismate synthase (18, 28, 31). The monomer is best described as a sandwich of two β-sheets with α-helices on the outside (Figures 2A, 4). Crystals of SmPabB grown during this study had an unusually high solvent content of 71.7% and a corresponding Matthews coefficient of 4.36 ų/Da. Figures 2B-D illustrate the arrangement of SmPabB monomers in the crystal lattice. Each unit cell contains 12 monomers arranged as two trimers of dimers. These hexameric units interact in a manner that yields a large pore or solvent channel in the crystal lattice (Figure 2B) that contributes to the unusually high solvent content of the crystal. The arrangement of subunits that the lattice is built upon is shown in figure 2D and features a continuous β-sheet between two SmPabB monomers. The segment of SmPabB involved in this interaction (residues 286-297) corresponds to a key, normally α-helical segment, that is often disordered in the absence of substrate and/or the glutaminase subunit, PabA (18). Figures 3 and S1 illustrate the dramatically different orientations of this flexible segment that is found in all PabB-like chorismate utilizing enzymes. In the TrpE-G complex, the segment is a well defined helix that forms part of the interface with the TrpG glutaminase subunit and contributes key elements of the active site notably two of the four acidic residues that coordinate the active site magnesium ion (28, 31). The corresponding segment of SmPabB, in the structure presented here, appears to have been recruited and remodeled as a key element of the continuous β-sheet between two SmPabB monomers that forms during the crystallization process (Figure 2D). This behavior may be emblematic of an adaptable interface region that even when unstructured, does not cause irreversible aggregation or destabilization of the overall polypeptide. In addition, it may suggest an ability to cope with nonspecialized amidotransferase partners in the apparent absence of a dedicated PabA subunit. Despite the unusual crystal packing observed, however, laser light scattering experiments conducted on SmPabB confirmed that it, like E. coli PabB, is a monomer in solution with an apparent mass of 48.6 ± 0.9 kDa.

Figure 2.

Overall structure of the SmPabB monomer and illustration of the crystal packing leading to the anomalous Matthews coefficient. (A) Ribbon diagram of SmPabB colored by secondary structure. (B) Crystal packing observed for the SmPabB structure illustrating the large solvent channels present in the crystals. (C) Alternate view of the packing within the hexagonal unit cell. (D) Each unit cell contains two ‘trimers of dimers’. The dimers associate via a continuous β-sheet formed from structural elements that are believed to be disordered in the absence of substrates and/or the amidotransferase subunit PabA.

Figure 4.

Cartoon representation of the SmPabB monomer shown in stereo. A model of tryptophan illustrates the location of the putative vestigial Trp binding site. The long loop that may be associated with Trp binding and that is common to SmPabB and E. coli PabB but not C. hutchinsonii PabB is shown in dark blue. The approximate location of the catalytic Mg²⁺ ion required for catalysis is shown. Two of the four Mg²⁺ coordinating residues are observed to be in position (E435 and E438). The other two metal ligands (E298 and E301) are on the remodeled β-strand recruited for the unusual crystal packing. Also shown in yellow is the key PIAGT motif featuring A274 that occupies the position analogous to K274 in canonical PabB enzymes. The Trp, formate, and catalytic Mg²⁺ ion binding sites were modeled by superimposing SmPabB and either E. coli PabB (1K0E), Serratia marcesens TrpE (1I7Q), or E. coli MenF (3BZN).

Figure 3.

SmPabB is partially disordered in the absence of substrate and PabA. The disordered region forms not only part of the active site but also key elements of the PabA-B interface. Cartoon diagrams of the TrpEG complex (1I1Q; TrpG shown in red, TrpE shown in gray except residues 319-376 shown in orange; upper panel) and SmPabB (blue except residues 267-305 shown in yellow; lower panel) illustrate that residues ~280-305 of SmPabB correspond to the vertical portion of the orange kinked helical segment seen in TrpE and other chorismate-utilizing enzymes. In the SmPabB structure these residues are remodeled as part of the intersubunit β-sheet structure shown in Figure 2D.

Ligand Binding Sites

PabB-type chorismate utilizing enzymes all have an active site that binds chorismate and some enzymes, notably TrpE and at least one PabB, have an additional binding site that accommodates tryptophan. In the case of TrpE, tryptophan binding is part of a feedback regulatory mechanism that limits the overproduction of tryptophan. In PabB, which is thought to have evolved from TrpE (18), it is believed the site is a nonfunctional vestigial binding site. The locations of the two potential ligand binding sites present in each monomer of SmPabB are illustrated in Figure 4. The chorismate binding site is composed of amino acid residues contributed both by the β-sheet core and by two key α-helices, one of which is the segment remodeled as a β-strand (residues 300-309) and the other is the long C-terminal helix (residues 432-452). A more detailed description of the active site is provided below.

The putative vestigial tryptophan binding site, first identified in E. coli PabB, appears to be present in SmPabB as well, though it is unoccupied in the structure described here. We previously reported that tryptophan bound to the vestigial site in E. coli PabB either had evolved to serve a structural role or was simply bound so tightly that it could not be readily removed (18). In anthranilate synthase, tryptophan binding to an analogous site regulates the activity of the enzyme by stabilizing an active site conformation that is not conducive to catalysis (31). In all cases, the tryptophan binding site is located some 15 Å from the active site and it is composed exclusively of residues from the β-sheet sandwich and the associated loops between the β-strands (Figure 4).

Similarity to Known PabB Structures

SmPabB is just the third ADC synthase (PabB) structure to be solved and the first that uses the noncovalent intermediate mechanism described by Schadt et al. (19). The results of conducting automated structural alignments of SmPabB with the known PabB structures and with other relevant structural homologs are shown in Table 2 (32). The final model of SmPabB consists of 439 of the 454 residues in the native sequence. The larger of the two disordered internal regions that could not be modeled, residues 286-297, is a helical segment that is also partially or entirely disordered in the two other known PabB structures, from E. coli (1K0E, 1K0G; two crystal forms (18)) and from the cellulose degrading Gram-negative bacteria Cytophaga hutchinsonii (3H9M). Based on analysis of and comparison to heterodimeric anthranilate synthase (TrpEG) structures (31, 33), this helix is predicted to be part of the PabAB interface and likely becomes ordered upon PabAB complex formation. The shorter disordered segment (residues 37-39) is located near the putative tryptophan binding site and on a solvent exposed loop. Finally, no density was observed for the two C-terminal residues of SmPabB.

Table 2.

Results^a of Automated Structural Alignments of SmPabB with Selected Homologs

Homolog	Cα RMSD (Å)	% identity of sequences	residues aligned
PabB (1KOG)	2.8	26	395
PabB (3H9M)	2.4	19	343
TrpE (1I7Q)	3.3	27	392
MenF (3BZN)	3.9	19	342
SS (2FN0)	3.4	23	348
EntC (3HWO)	2.8	18	300
PhzE (3R75)	2.9	21	325

^aThe software DALI (32) was used to perform the alignments of SmPabB with E. coli PabB (1K0G), C. hutchinsonii PabB (3H9M), Salmonella typhimurium TrpE (1I7Q), E. coli MenF (3BZN), Yersinia enterocolitica salicylate synthase (SS; 2FN0), E. coli EntC (3HWO), and P. aeruginosa PhzE (3R75).

While some disorder is present in the structures of both E. coli PabB and SmPabB, it is clear that all major structural elements are conserved between these two proteins. The C. hutchinsonii PabB was a structural genomics deposition to the PDB has not been described by any publication. Interestingly, two elements of that structure differ from the SmPabB and E. coli PabB structures. The C. hutchinsonii PabB lacks two β-strands that form one end of one of the long β-sheets of each monomer. In SmPabB this corresponds to residues 60-70 and E. coli PabB, residues 63-78. While the significance of this is unclear, we had previously noted that these strands are also absent in the archeabacterial Sulfolobus TrpE structure (1QDL) but are present in the known eubacterial TrpE structures ((34)1I1Q, 1I1S). Additionally, SmPabB and E. coli PabB both feature a long loop region (Figure 4; residues ~90-105) that forms a partial flap over the vestigial tryptophan binding site identified in E. coli PabB. The corresponding loop in C. hutchinsonii PabB is only four residues (77-81). Tryptophan was not included in the C. hutchinsonii PabB model but it is unknown whether the shorter loop plays a major role in tryptophan binding and the data presented here and elsewhere suggest other factors are also important (18, 35). The missing strands and the shortened loop together account for the shorter length of C. hutchinsonii PabB (426 residues) when compared to SmPabB (454 residues) and E. coli PabB (453 residues).

Active Site

The SmPabB chorismate binding site is highly distorted and disordered. Ligand-bound structures of anthranilate synthase and isochorismate synthase have revealed that a ligand-Mg²⁺ complex is formed that ties together several secondary structural elements and promotes a closed active site conformation (31, 36). Four acidic residues in particular, on two parallel helical segments, tie the active site together by interacting with the active site Mg²⁺ ion. In E. coli PabB, three Glu residues (302, 436, and 439) and Asp299 coordinate the Mg²⁺ ion either directly or via water molecules. In SmPabB the four corresponding residues are all Glu residues (298, 301, 435, and 438). Two of these residues, however, Glu298 and Glu301 are located on the remodeled helix-turned-sheet element that forms the intersubunit β-sheet in the SmPab structure (Figures 2D, 3). This arrangement places these residues some 15 Ǻ from their expected location in a closed enzyme-substrate complex. In addition to close analogs or actual products, the known structures of E. coli PabB and isochorismate synthase illustrate that even seemingly poor substrates mimics such as formate and phosphate ions can promote a closed or partially closed conformation (18, 28). Kinetic data presented below suggest, however, that the interaction of SmPabB with an amidotransferase partner also has a significant, perhaps greater effect on active site organization. Based on an evaluation of TrpEG complexes, the disordered regions of SmPabB are likely part of the PabAB interface in addition to contributing two of the four Glu residues involved in Mg²⁺ coordination (Figures 3, 4).

One magnesium ion was identified in the SmPabB structure but it is located outside of the active site and its location suggests it has no physiologic relevance. The octahedral coordination involves five water molecules and the carboxylate groups of Glu284 and Glu331. The latter interacts indirectly via a water molecule while Glu284 interacts directly with the magnesium ion.

The ²⁷³PIAGT²⁷⁷ motif that differentiates SmPabB mechanistically from E. coli PabB and other canonical PabB enzymes is engaged in several interactions that hold it in a distorted conformation in the absence of substrate (Figure 5). Ile274, in particular, is engaged in two interactions that hold the motif in its observed location. The backbone amide of Ile274 is 2.5 Å from the side chain carboxyl of Glu258 and the side chain of Ile274 is sandwiched between the phenyl ring of Phe241 and the nonpolar portion of the Agr259. These interactions are likely of no functional significance but rather reflect favorable contacts that form in the absence of substrate and/or PabA. Following Thr277, the chain meanders in an unstructured fashion until the chain becomes disordered to the extent that no density was visible for residues 286-297.

Figure 5.

Cartoon representation illustrating the interactions anchoring Ile273 and the ²⁷²PIAGT²⁷⁷ motif in a kinked conformation away from the substrate binding site. Ile273 interacts with Glu258 via a hydrogen bond and with the side chains of Phe241 and Arg259. Isochorismate, benzoic acid, and pyruvate from EntC (3HWO) and TrpE (1I7Q) are shown to illustrate the location of the substrate binding site of SmPabB.

Tryptophan Binding and the Vestigial Trp Binding Site of SmPabB

Upon solving the structure of E. coli PabB, we surprisingly observed tryptophan bound to a vestigial site analogous to the regulatory binding site in TrpE, a tryptophan biosynthetic enzyme (18). Since E. coli PabB remained the only known PabB structure for some time, it was difficult to assess whether tryptophan binding is common among PabB enzymes or what its importance might be. The structures of SmPabB and C. hutchinsonii PabB have now afforded an opportunity to further examine whether tryptophan binding is typical among PabB enzymes. Tryptophan was ultimately not found in the SmPabB structure, however, a comparison with the E. coli PabB structure revealed significant similarity in putative binding site. Interestingly, examination of the analogous site in C. hutchinsonii PabB revealed that the side chain of Phe46 is wedged into the analogous site of that protein, likely preventing tryptophan from binding. A similar observation was made in the structure of the ADIC synthase PhzE where Trp184 occupies the site and likely blocks exogenous tryptophan binding (37). While SmPabB does have a tryptophan residue (Trp43) that corresponds to Trp184 of PhzE and Phe46 of C. hutchinsonii PabB its orientation is more consistent with Phe46 of E. coli PabB, which does not block tryptophan binding. Thus, while tryptophan was not observed bound to SmPabB nothing appeared to rule out the possibility that it could bind.

Three experiments were pursued to evaluate whether SmPabB binds tryptophan despite its absence in the crystal structure. First, we used LC/MS to examine whether purified solutions of SmPabB contained tryptophan. Figure S2 illustrates that tryptophan is clearly present in samples of SmPabB identical to those used for crystallization experiments. We conducted a similar analysis on three SmPabB homologs: E. coli PabB, E. coli MenF (isochorismate synthase), and Pseudomonas aeruginosa PhzE (ADIC synthase). The latter two were not believed to bind tryptophan based on crystallographic data (18, 28, 35, 37). The results show that E. coli PabB samples, as expected, contained tryptophan while the MenF and PhzE samples did not (Figure S2). Second, we used isothermal titration calorimetry to show that SmPabB and tryptophan form a complex characterized by a dissociation constant of ~11.7 μM (Figure S3). Deviation from the expected 1:1stoichiometry and further analysis by LC/MS confirmed that extensive dialysis did not remove all tryptophan prior to the calorimetric analysis. Finally, co-crystallization and soaking experiments were conducted in order to determine whether higher tryptophan concentrations could yield a tryptophan-bound SmPabB structure. Soaking experiments have thus far yielded promising but inconclusive results as data sets appear to show obvious density consistent with tryptophan bound in the expected location, however, no data set extends beyond ~3 Å and they suffer other issues that prevent them from refining satisfactorily.

Interestingly, it appears SmPabB and TrpE have roughly the same affinity for tryptophan (38). As with E. coli PabB, however, addition of tryptophan had no effect on the activity of SmPabB.

Enzymatic Activity of SmPabB

Reactions containing SmPabB, E. coli PabA, chorismate, and glutamine were analyzed by LC/MS to determine whether SmPabB, like B. subtilis PabB, catalyzed the formation of ADC via an ADIC intermediate. Figure 6 illustrates that both ADIC and ADC were formed as predicted. Qualitative evaluation of the chromatographic data suggested that the conversion of ADIC to ADC was potentially rate limiting since ADIC appears to accumulate. However, further investigation showed that ADIC is released into solution prior to being rebound and converted to ADC in a second SmPabB-catalyzed reaction. Competition from the initially high amount of chorismate present also favors accumulation of ADIC that becomes unbound. In order to confirm that ADIC was free in solution, we included, in some assays, catalytic amounts of the enzyme PhzD from P. aeruginosa, which acts very efficiently on ADIC, in SmPabB containing reactions (22). This resulted in accumulation of the PhzD product 2,3-dihydro-3-hydroxyanthranilate (DHHA) and confirmed that SmPabB does not effectively sequester the ADIC intermediate (Figure S4).

Figure 6.

Extracted ion mass chromatograms identifying ADIC as an intermediate in the SmPabB catalyzed conversion of chorismate to ADC. ADIC is present in reactions containing SmPabB but not those containing E. coli PabB (offset trace).

Kinetic Characterization of SmPabB

Analysis of the kinetics of the glutamine and PabA-dependent conversion of chorismate to ADC revealed that SmPabB behaved much like the B. subtilis PabB (Figures S5, S6; (19)). SmPabB has a somewhat lower K_m for chorismate, 240 μM vs 410 μM for the B. subtilis enzyme, but both of these enzymes have K_m values for chorismate that are markedly higher than what has been described for the E. coli PabB (~13 μM (20)).

Analysis of ADC formation using purified ADIC to examine only the second half of the reaction, revealed that the second step is actually somewhat faster (k_cat ~ 0.02 s^-1) than the overall reaction (k_cat ~ 0.007 s^-1) and that initial accumulations of ADIC (Figure 6) are likely a consequence of the second step being slow at very low initial ADIC concentrations and high initial chorismate concentrations.

Effect of PabA on the Activity of SmPabB

Assays containing only SmPabB, chorismate ammonia, and PabC indicated that SmPabB was a very poor catalyst. When assays were conducted with 500 μM chorismate, the specific activity was 2 × 10^-4 s^-1. Addition of equimolar E. coli PabA, however, significantly enhanced the reaction rate, even in the absence of glutamine as the ammonia donor. When assayed again at 500 μM chorismate, the specific activity increased 20-fold to 0.004 s^-1 in the presence of PabA.

We first attempted to characterize the association of SmPabB and PabA by examining the dependence of the reaction rate on the ratio of PabA:SmPabB. The results suggested that the association was likely very strong. However, we were unable to drop the enzyme concentrations in these assays far enough to reliably estimate the binding constant due the sensitivity level of the assay system. Therefore we examined the PabA:SmPabB interaction directly by fluorescence titration. The results are shown in Figure 7 and reveal that E. coli PabA and SmPabB form a surprisingly tight complex characterized by a ~40 nM dissociation constant. This result supports the possibility that SmPabB has an adaptable interface region that may accommodate a variety of nonspecialized amidotransferase partners. A similar analysis of the interaction of E. coli PabA and PabB revealed a dissociation constant of 109 ± 58 nM.

Figure 7.

Fluorescence titration experiments suggest SmPabB and E. coli PabA form a high affinity complex. Steady State emission spectra of SmPabB alone and in the presence of increasing amounts of PabA were recorded at 25 °C using an excitation wavelength of 285 nm. The change in fluorescence emission at 340 nm was evaluated as a function of PabA concentration by fitting the data to a quadratic function (Equation 1). Data shown above are corrected for the linear increase in fluorescence due to PabA alone. The dissociation constant determined by this method was (3.9 ± 3.1) × 10^-8 M.

Crystallographic evidence currently available for chorismate utilizing enzymes suggests that an occupied active can lead to structures that better reflect an active conformation. The structure of the enterobactin specific isochorismate synthase (EntC), for example, in complex with the product isochorismate features a well ordered, closed active site (36). Similarly, anthranilate synthase structures are well ordered in the presence of ligands (31). No structures of PabB have yet been solved with substrates or products present in the active site, however a simple formate ion found in the E. coli PabB active site does significantly reduce disorder relative to the completely unliganded structure (18). The formate is likely a mimic of the pyruvyl portion of chorismate as it binds in the same location and orientation as pyruvate in anthranilate synthase and the pyruvyl portion of isochorismate in the EntC structure. It is worth noting that the formate bound structure of E. coli PabB was obtained from crystals grown from solutions containing 2 M sodium formate a condition which may have favored occupancy of even a poor substrate mimic.

One explanation for the difficulty in obtaining substrate or product bound PabB structures is that chorismate may simply have little affinity for PabB in the absence of a PabA partner. This seems plausible in light of the disordered elements observed in all PabB structures solved to date. It may be that substrate alone cannot effectively promote the formation of an ordered PabB active site. In the case of SmPabB, evidence suggests that PabA:SmPabB complex formation reduces SmPabB disorder based on the enhanced catalytic performance of the enzyme in the presence of PabA.

Comparison of E. coli PabA and S. maltophilia TrpG

In order to better understand why E. coli PabA was able to effectively partner with SmPabB, we compared the residues from each that are expected to form the PabB interface. Using PDBSum (39) we identified 26 residues from S. marcescens TrpG that contact the TrpE subunit in the anthranilate synthase complex structure 1I7Q (31). We then identified the corresponding residues in E. coli PabA and S. maltophilia TrpG by examining sequence alignments. Nearly 81 percent (21/26) of these residues are identical between E. coli PabA and S. maltophilia TrpG. Two others are conservative substitutions and just three likely interface residues are not conserved. Cys20, Glu21, and Lys29 of E. coli PabA correspond to S. maltophilia TrpG residues Gln20, Thr21, and Val29. These observations, along with the overall identity (61%) between the two proteins likely also contribute to the observed ability of E. coli PabA to serve SmPabB effectively. Additionally, the kinetic data showing that the SmPabB:E. coli PabA complex has activity equal to the B. subtilis PabAB (19) complex suggests that E. coli PabA may be maximizing the abilities of SmPabB. The much lower sequence identity (Table 2) between SmPabB, E. coli PabB, and TrpE coupled with the disorder observed in all PabB structures makes conducting a similar analysis of the PabA interface of PabB difficult.

Conclusion

The structure of SmPabB, the chorismate-binding subunit of the heterodimeric ADC synthase complex from the S. maltophilia folate pathway, has been solved. It is just the third reported PabB structure and it is the first described structure of a PabB that uses the alternative, and until recently unrecognized mechanism involving a noncovalent ADIC intermediate (19). The structure is notable for the very high solvent content of the crystals and the unusual trimer-of-dimers structure that the normally monomeric SmPabB adapts in the crystals form (Figure 2). Key elements of the typically disordered PabA interface region of SmPabB are reorganized to form a PabB:PabB interface that is a key part of the crystal lattice.

Surprisingly, S. maltophila does not appear to encode a dedicated PabA and may rely on the product of trpG, which is found adjacent to trpE, to serve as the amidotransferase partner for both enzymes. Evaluation of the activity of SmPabB revealed that the enzyme is a very poor catalyst in the presence of only chorismate and exogenous ammonia (from ammonium sulfate). Inclusion of a surrogate PabA (from E. coli), however, stimulated activity 20-fold even in the absence of glutamine suggesting that an amidotransferase partner can stabilize a more productive conformation of SmPabB even if it is not supplying ammonia to the reaction via glutamine hydrolysis. Fluorescence titration experiments indicate that SmPabB and E. coli PabA from a very stable complex. Together, the data are consistent with a scenario where SmPabB alone has minimal activity due to inherent disorder in key elements of the active site. Substrate binding alone is not enough to stabilize a productive conformation. Only when an amidotransferase partner is present would SmPabB produce ADC for conversion to p-aminobenzoate by PabC.

Evaluation of the regulatory controls on the transcription of both pabB and the trp operon components in S. maltophilia may reveal how the organism is able to regulate the production of both folate and tryptophan precursors using a single amidotransferase enzyme. The situation appears different than in Bacillus since the apparent bifunctional amidotransferase is not located in the folate operon in S. maltophila (as it is in B. subtilis), but rather adjacent to trpE. In order to obtain a SmPabB structure with a well organized active site, it seems clear based on the three available PabB structures that a PabAB complex structure will ultimately be necessary.