Proton transfer dynamics of GART: The pH-dependent catalytic mechanism examined by electrostatic calculations

Abstract

The enzyme glycinamide ribonucleotide transformylase (GART) catalyzes the transfer of a formyl group from formyl tetrahydrofolate (fTHF) to glycinamide ribonucleotide (GAR), a process that is pH-dependent with pK_a of ∼8. Experimental studies of pH-rate profiles of wild-type and site-directed mutants of GART have led to the proposal that His108, Asp144, and GAR are involved in catalysis, with His108 being an acid catalyst, while forming a salt bridge with Asp144, and GAR being a nucleophile to attack the formyl group of fTHF. This model implied a protonated histidine with pK_a of 9.7 and a neutral GAR with pK_a of 6.8. These proposed unusual pK_as have led us to investigate the electrostatic environment of the active site of GART. We have used Poisson-Boltzmann-based electrostatic methods to calculate the pK_as of all ionizable groups, using the crystallographic structure of a ternary complex of GART involving the pseudosubstrate 5-deaza-5,6,7,8-THF (5dTHF) and substrate GAR. Theoretical mutation and deletion analogs have been constructed to elucidate pairwise electrostatic interactions between key ionizable sites within the catalytic site. Also, a construct of a more realistic catalytic site including a reconstructed pseudocofactor with an attached formyl group, in an environment with optimal local van der Waals interactions (locally minimized) that imitates closely the catalytic reactants, has been used for pK_a calculations. Strong electrostatic coupling among catalytic residues His108, Asp144, and substrate GAR was observed, which is extremely sensitive to the initial protonation and imidazole ring flip state of His108 and small structural changes. We show that a proton can be exchanged between GAR and His108, depending on their relative geometry and their distance to Asp144, and when the proton is attached on His108, catalysis could be possible. Using the formylated locally minimized construct of GART, a high pK_a for His108 was calculated, indicating a protonated histidine, and a low pK_a for GAR(NH₂) was calculated, indicating that GAR is in neutral form. Our results are in qualitative agreement with the current mechanistic picture of the catalytic process of GART deduced from the experimental data, but they do not reproduce the absolute magnitude of the pK_as extracted from fits of k_cat-pH profiles, possibly because the static time-averaged crystallographic structure does not describe adequately the dynamic nature of the catalytic site during binding and catalysis. In addition, a strong effect on the pK_a of GAR(NH₂) is produced by the theoretical mutations of His108Ala and Asp144Ala, which is not in agreement with the observed insensitivity of the pK_a of GAR(NH₂) modeled from the experimental data using similar mutations. Finally, we show that important three-way electrostatic interactions between highly conserved His137, with His108 and Asp144, are responsible for stabilizing the electrostatic microenvironment of the catalytic site. In conclusion, our data suggest that further detailed computational and experimental work is necessary.

The de novo biosynthetic pathway of purines comprises a cascade of catalytic reactions consisting of 11 steps and involving 11 different enzymes that convert α-D-ribose-5-phosphate to ionosine monophosphate (Garrett and Grishnan 1995). Two steps of the purine biosynthetic pathway, which involve the enzymes glycinamide ribonucleotide transformylase (GART) and 5-aminoimidazole-4-carboxyamide ribonucleotide transformylase (AICART), are dependent on folate derivatives. These enzymes are good targets for development of aolate therapeutic agents for cancer chemotherapy, autoimmune diseases, and microbial infections. In the fourth step of the purine biosynthetic pathway, GART catalyzes the transfer of a formyl group from N¹⁰-formyl-tetrahydrofolate (fTHF) to glycinamide ribonucleotide (GAR) to form formyl-GAR (fGAR).

GART is active at high pH in monomeric form (Mullen and Jennings 1996) with catalytic activity in the pH range of 6–10 (Inglese et al. 1990a; Shim and Benkovic 1999). Shim and Benkovic (1999) have studied the pH dependence of the kinetic parameters of Escherichia coli GART complexed with GAR and cofactor fDDF (10-formyl-5,8-dideazafolate), which strongly resembles fTHF. They have proposed a model of the catalytic tetrahedral transition state, based on experimental measurements of k_cat-pH and k_cat/K_m-pH profiles of wild-type and site-directed mutants His108Ala, Asp144Ala of GART, which suggested that a neutral GAR and a protonated His108 are necessary in the pH range of 6–10 for catalysis to take place. This model accounts for pK_as of ∼6.8 and 9.7 for GAR(NH₂) and His108, respectively (Shim and Benkovic 1999). In addition, the ionization states of His108 and GAR(NH₂) are stabilized by the presence of Asp144 and Asn106, which are also located in the catalytic site (Almassy et al. 1992; Chen et al. 1992; Shim and Benkovic 1999).

Figure 1 ▶ shows different representations of the high-pH ternary complex of GART (Almassy et al. 1992; PDB access code 1cde). Figure 1A ▶ shows a ribbon model of GART with a stick model of substrate GAR and cofactor 5dTHF. GART consists of 10 β-strands, six (at high pH) or five (at low pH) α-helices, one short 3₁₀ helix, and five ideied β-turns (Almassy et al. 1992; Chen et al. 1992). Strands β1–β7 form a 90°-twisted β-sheet (β1–β4, parallel; β5–β7 anti-parallel), and strands β8–β10 form a short antiparallel terminal β-sheet. Strands β1–β4 are flanked by two pairs of α-helices (α2, α3; α1, α4) and strands β5–β6 flank the long α-helix α6 (Fig. 1A ▶). Figure 1B ▶ shows substrate GAR and pseudocofactor 5dTHF alone in the same orientation as in Figure 1A ▶. The enzyme active site is located in a shallow surface cavity, closed by a loop-helix (Fig. 1A ▶, C in yellow) and a second loop (Fig. 1A, C in blue ▶), hereafter called the activation and binding loop, respectively. The active site contains critical ionizable residues for catalysis, His108 and Asp144 (Fig. 1A ▶).

Fig. 1.

(A) Ribbon model of GART (ternary complex, PDB code 1cde) with substrates in stick representation. Catalytic site assembly is formed by the activation loop-helix (in yellow) and the binding loop (in blue). Catalytically important residues His108 and Asp144 are shown in red and blue, respectively. (B) GAR and 5dTHF alone with notation of ionizable groups and atoms discussed in text. In the active complex, the formyl group is attached to 5dTHF N₁₀ and is transferred to GAR N₂₄ or GAR(NH₂) as referred to in text during catalysis. (C) A different orientation of a tube model of GART in which important catalytic site residues are shown: Asn106 (in cyan), His108 (in red), Tyr115 (in magenta), His119 and His121 (in yellow), His132 and His137 (in green), and Asp144 (in blue). The short C-terminal β-sheet (residues 188–209) has been deleted for clarity.

GART catalysis is coupled to two major structural changes, one associated with pH and another associated with binding of the substrate and the cofactor. Each of these structural changes involves a different loop in the proximity of the active site. The first loop, the activation loop, is a long flexible loop at low pH (< 7), which undergoes a transition, and part of it forms, reversibly, an α-helix at high pH (> 7; Fig. 1A, C in yellow ▶). This long loop is located between strands β5–β6 (residues 111–131; Fig. 1A, C in yellow ▶). The disorder of this loop at low pH is shown by lack of or reduced electron density in the X-ray structures (Almassy et al. 1992; Chen et al. 1992; Klein et al. 1995). In the ordered form (α-helix formed at residues 120–127, Fig. 1A, C in yellow ▶), the activation helix-loop entity caps the active site to aid catalysis (Almassy et al. 1992; Su et al. 1998; Greasley et al. 1999). It has been suggested that the closure of the activation loop is involved in the binding of the substrate (Shim and Benkovic 1999). Two histidines are present in the activation loop-helix, His121 and His119 (Fig. 1C in yellow ▶), and they play significant role in the formation of structure at high pH (Morikis et al. 2001). The second flexible loop, the binding loop, between strands β6 and β7 (residues 140–145, in blue; Fig. 1A,C ▶), becomes reoriented by turning toward the active site on substrate-cofactor binding (Almassy et al. 1992; Chen et al. 1992; Klein et al. 1995). The conformational change of this loop brings to interaction distance the catalytically critical residues His108 (in the active site; Fig. 1A,C ▶) and Asp144 (in the binding loop; Fig. 1A,C ▶). Two more histidines are in the active site, His132 and His137 (Fig. 1C ▶ in green), located in strand β6 between the activation and binding loop (Fig. 1C ▶). Active site residues Asn106, His108, Asp144, and His137 (Fig. 1C ▶) are highly conserved (Almassy et al. 1992; Mullen 2000).

Using the large amount of background work on GART (Inglese et al. 1990a,b; Almassy et al. 1992; Chen et al. 1992; Klein et al. 1995; Warren et al. 1996; Mullen and Jennings 1996, 1998; Shim and Benkovic 1998, 1999; Su et al. 1998), investigators are now focusing on the design of new inhibitors (Greasley et al. 1999); however, the proposed catalytic mechanism of GART (Shim and Benkovic 1999) is based solely on experimental data. There is a lack of supporting theoretical evidence that would amplify the existing background and provide a physical basis for understanding GART catalysis. In addition, theoretical and computational tools may be critical in the design of more efficient GART inhibitors.

In the present study we performed electrostatic calculations based on the solution of the linearized Poisson-Boltzmann equation (for review, see Davis and McCammon 1990; Honig and Nichols 1995; Antosiewicz et al. 1996^; Ullmann and Knapp 1999) and the multiple-site titration model (Gilson 1993) to predict pK_a values (Bashford and Karplus 1990; Yang and Honig 1993; Antosiewicz et al. 1994) of all ionizable groups in GART. We first discuss the initial protonation and conformational states of histidines for the preparation of the calculations. Second, we probe the importance of residues involved in catalysis by discussing the output of several calculations using the crystallographic structure of the ternary complex (Almassy et al. 1992) with critical theoretical mutations and theoretical deletions and variable initial conditions. Third, we discuss repeated pK_a calculations using a GART construct with a reconstructed, more realistic pseudocofactor, which contained the missing formyl group and was locally minimized to accommodate the addition of the formyl group. Finally, we compare our findings with recent experimental results (Shim and Benkovic 1999). We chose the high-pH ternary complex structure (PDB code 1cde; Almassy et al. 1992) for these calculations, because it represents closely the reactants of the actual complex during catalysis. The main difference is that this structure contains unformylated cofactor 5dTHF instead of formylated THF (fTHF), and a smaller difference is that nitrogen N₅ of the bicyclic ring of the cofactor is replaced by carbon.

Results

Initial protonation states of histidines

Histidines possess two distinct protonation sites at the N_δ1- and N_ɛ2-atoms that create an ambiguity in selecting the initial position of hydrogen in a neutral histidine. In addition, X-ray crystallography cannot always distinguish the κ₂ torsion angle that positions the histidine ring in a certain orientation from its flipped orientation by 180°. Similar ambiguities are also possible for the κ₁ and κ₂ torsion angles of asparagines and glutamines, respectively. This is because of similar electron densities of nitrogen, carbon, and oxygen atoms (Nielsen et al. 1999). Although in cases of isolated or solvent-exposed histidines the initial hydrogen position and flip state selections may not be important, in cases of histidines in the active site or interactive histidines, these selections become significant and can dramatically alter the results of the electrostatic calculations. The implementation of a global hydrogen-bonding network optimization method within the program WHAT IF v. 99 (Hooft et al. 1996; Nielsen et al. 1999; Vriend 1990) resolves these ambiguities.

In the case of the ternary complex of GART, the most stable position of the side chain of catalytically important residue His108, as provided by the global hydro-p gen-bonding optimization network, is at a 180°-flipped geometry, compared to its initial orientation in crystallographic structure 1cde (Almassy et al. 1992). Indeed, the His108 ring conformations of other GART crystallographic structures, 1gar (Chen et al. 1992), 2gar (Su et al. 1998), and 3gar (Su et al. 1998), which are not altered by the global hydrogen-bonding optimization network algorithm, are in agreement with a flipped ring conformation of structure 1cde.

Table 1 shows four different pK_a calculations using the crystal structure of the ternary complex, 1cde, with His108 in its original ring orientation and a 180° flipped-ring orientation, each with a hydrogen atom at N_ɛ2 and N_δ1. The four structures with different His108 states will be called hereafter 1cdeH108(N_ɛ2H), 1cdeH108(N_ɛ2H)/flip, 1cdeH108(N_δ1H), 1cdeH108(N_δ1H)/flip, respectively (Table 1). The pK_as of all 10 histidine residues, Asp144, GAR, and 5dTHF are tabulated. The pK_a of His108 fluctuates between values 7.8 and 3.9. Likewise, fluctuations of 3.6 and 0.4 pK_a units are found for catalytically important groups of GAR(NH₂) and His137, respectively. The rest of the ionizable groups, including catalytically important residue Asp144, show no pK_a change or insignificant change within the four calculations (Table 1). Figure 2A ▶ shows a detail of the catalytic site, including GAR, 5dTHF, His108, Asp144, and His137, in which two sets of three-way interactions are observed, His108-Asp144-GAR(NH₂) and His108-His137-Asp144, respectively. Figure 2B ▶ shows the proximity (in Å) of the key ionizable catalytic residues and Asn106. It should be noted that the global hydrogen-bonding network optimization algorithm does not produce flip of the ring of His137, which is stabilized in the catalytic site through a hydrogen bond to Asn106 (Fig. 2 ▶). Excluding the minor pK_a fluctuation of His137 (Table 1), it appears that the pK_as of His108 and GAR(NH₂) are closely dependent in a see-saw-type relation. When the pK_a of His108 is high, the pK_a of GAR(NH₂) is low (Table 1, columns 2, 4, and 5); and when the pK_a of GAR(NH₂) is high, the pK_a of His108 is low (Table 1, column 3). Both ionizable groups are located in an unfavorable desolvated environment, and in addition, they are in unfavorable coulombic interaction proximity with each other. This results in lowering the pK_as of His108 and GAR(NH₂) from their model pK_a; however, favorable coulombic interaction with Asp144 (Fig. 2 ▶; Table 1) counterbalances the unfavorable situation by increasing the pK_as of His108 and GAR(NH₂) (Table 1). A visual examination of the hydrogen bonds formed by His108 and GAR(NH₂) indicates that a flipped His108 with protonation at N_ɛ2 (Table 1, column 3) forms two hydrogen bonds: His108(N_ɛ2)-Tyr115(O) and His108(N_δ1)-GAR(NH₂), providing the most stable environment for the unfavorable interaction of His108 and GAR(NH₂) and for the catalytic site. The other three cases (Table 1, columns 2, 4, and 5) show unstable environments for His108-GAR(NH₂). More specifically: (1) When His108 is in its original ring position (as in PDB structure 1cde) with a hydrogen at N_ɛ2, only one possible hydrogen bond between His108(N_ɛ2)-Gly117(O) is found. (2) When His108 is in its original ring position with a hydrogen at N_δ1, there is no hydrogen bond formation for the histidine ring. (3) When His108 is in its flipped orientation with a hydrogen at N_δ1, only a His108(N_δ1)-GAR(NH₂) hydrogen bond is formed.

Table 1.

Effect of His108 κ2-torsion angle flip and His108 initial protonation state on the pKas of key residues/groups for catalysis, using crystallographic structure 1cde

	1cdeH108(Nɛ2H)	1cdeH108(Nɛ2H)/flip	1cdeH108(Nδ1H)	1cdeH108(Nδ1H)/flip
	Initial hydrogen at Nɛ2 of His108		Initial hydrogen at Nδ1 of His108
	κ2(H108) = 0°	κ2(H108) = 180°	κ2(H108) = 0°	κ2(H108) = 180°
Ionizable residue/groupa	pKa (apparent)
HIS54	7.2	7.2	7.2	7.2
HIS73	6.6	6.6	6.6	6.6
HIS99	6.0	6.0	6.0	6.0
HIS108	6.9	3.9	7.7	7.8
HIS119	5.7	5.7	5.7	5.7
HIS121	7.0	6.9	6.9	6.9
HIS132	0.4	0.5	0.4	0.4
HIS137	2.0	2.4	2.1	2.0
ASP144	−2.3	−2.4	−2.3	−2.4
HIS174	6.8	6.8	6.8	6.8
HIS192	6.3	6.3	6.3	6.3
GAR(NH2)	5.8	8.6	5.3	5.0
GAR PO42−	2.7	2.6	2.7	2.7
GAR PO4−	<−5	<−5	<−5	<−5
5dTHF HN3-C4=O	10.9	10.9	10.9	10.8
5dTHF CγOO−	2.7	2.7	2.7	2.7
5dTHF CαOO−	2.6	2.6	2.6	2.6

^a All 10 histidines, Asp144, and ionizable groups of GAR and 5dTHF are shown. Ionizable groups that show pK_a shifts >0.4 pK_a units, within the four calculations, are in bold face.

Fig. 2.

(A) Catalytic site showing the GAR, 5dTHF, His108, Asp144, and His137 interaction. (B) Schematic of interaction with distances (in Å) between heavy atoms of ionizable groups of His108, Asp144, His137, GAR(NH₂), and 5dTHF(N₁₀). Solid and dotted lines are distances < or > 6 Å, respectively. Values in parentheses correspond to distances involving His108 in its flipped orientation (see text). Interactions with catalytic residue Asn106 are also shown (in blue).

Although the results of column 3 in Table 1 reflect a more realistic local catalytic site structure, all of the above observations (summarized in Table 1 and Fig. 2 ▶) suggest that there is a very sensitive structural and electrostatic interdependence between His108 and GAR(NH₂). Small conformational changes tend to stabilize-destabilize the catalytic site and can swing the see-saw effect of pK_a toward a protonated His108 (with high pK_a)-unprotonated GAR(NH₂) (with low pK_a), or vice versa.

Figure 3 ▶ shows the theoretical titration curves for interacting catalytic ionizable sites His108, His137, Asp144, and GAR(NH₂). None of the titration curves of Figure 3 ▶ possesses a clean sigmoidal shape with a sharp transition centered at the pK_a value, defined as the pH of charge +0.5 or −0.5. Deviation from the typical shape of titration curve is caused by strong electrostatic interaction between two ionizable sites, as has been shown in other systems before (Bashford and Gerwert 1992; Yang et al. 1993) or among three-way electrostatic interaction as shown in certain cases here. The plots in Figure 3A ▶ were generated using the structure of the ternary complex 1cdeH108(N_ɛ2H), with unflipped His108 and initial hydrogen at N_ɛ2. The titration curves of His108 and GAR(NH₂) have similar line shapes with pK_as of 6.9 and 5.8, respectively (Table 1). A second inflection point at pH ∼2 in the titration curve of His108 (Fig. 3A ▶) can be attributed to unfavorable coulombic interaction with His137 (Fig. 2 ▶), which has a pK_a of 2.0 (Table 1). A similar second inflection point in the titration curve of GAR(NH₂) (Fig. 3A ▶) can be attributed to a relay effect of interaction with His137 through His108 (Fig. 2 ▶). The His137-His108 and His137-Asp144 unfavorable and favorable coulombic interaction, respectively (Fig. 2 ▶), is shown in the nontypical shape of the titration curve of His137 (Fig. 3A ▶). Likewise, the favorable coulombic interaction of Asp144 with His108 and His137 (Fig. 2 ▶) is shown in the slightly distorted shape of the titration curve of Asp144. Similar titration curves are observed in the case of 1cdeH108(N_δ1H), with unflipped His108 and initial hydrogen at N_δ1 (Fig. 3C ▶).

Fig. 3.

Theoretical titration curves for interacting catalytic site ionizable sites, His108 (squares), His137 (diamonds), Asp144 (triangles), and GAR(NH₂) (circles) of structures: (A) 1cdeH108(N_ɛ2H) with unflipped His108 (κ₂(H108) = 0°) and initial hydrogen at N_ɛ2, (B) 1cdeH108(N_ɛ2H)/flip with flipped His108 (κ₂(H108) = 180°) and initial hydrogen at N_ɛ2, (C) 1cdeH108(N_δ1H) with unflipped His108 (κ₂(H108) = 0°) and initial hydrogen at N_δ1, and (D) 1cdeH108(N_δ1H)/flip with flipped His108 (κ₂(H108) = 180°) and initial hydrogen at N_δ1. Quoted pK_as in Tables 1 and 2 correspond to pH values of charge +0.5 or −0.5.

In the case of structure 1cdeH108(N_ɛ2H)/flip, which includes the most favorable state in terms of hydrogen bonding (vide supra) of a flipped His108 and initial hydrogen at N_ɛ2, the theoretical titration curves of His108 and GAR(NH₂) (Fig. 3B ▶) are complex. This complexity applies to a lesser extent on His137 and Asp144 (Fig 3B ▶). Comparing Figure 3B ▶ with Figures 3A, 3C, and 3D ▶, the see-saw effect in the pK_as of His108 and GAR(NH₂) can be seen, as discussed above. It appears as the titration curves of each site in Figure 3B ▶ is a weighted superposition of the titration curves of all four sites, His108, His137, Asp144, and GAR(NH₂), indicating the direct and relay four-way electrostatic interactions of these residues that stabilize the electrostatic microenvironment of the catalytic site and prepare the enzyme for catalysis. Similar observation can be made in the case of the titration curves of structure 1cdeH108(N_δ1H)/flip, with flipped His108 and initial hydrogen at N_δ1 (Fig. 3D ▶), except that the see-saw effect has brought the pK_a of His108 to a higher value than the pK_a of GAR(NH₂), probably demonstrating the loss of a shared proton through hydrogen bonding, which was the case in structure 1cdeH108(N_ɛ2H)/flip.

Construction of theoretical mutations and deletions

There are two opposite effects that may be present when a charged residue is found within the low dielectric environment of the interior of the protein. First, desolvation of charged residues is energetically unfavorable, and it is accompanied by pK_a changes that favor the neutral form. For example, the pK_a of a desolvated basic residue shifts to lower value and the pK_a of an acidic residue shifts to a higher value. Second, unfavorable coulombic interactions of residues of the same charge are also accompanied by pK_a changes that favor the neutral form but to a lesser extent than desolvation. For example, the pK_a of interacting basic residues shifts to lower pK_a values and the pK_a of interacting acidic residues shifts to higher values. Third, favorable coulombic interactions of charged residues with opposite charge favor the ionized form. For example, the pK_a of the basic residue shifts to higher value and the pK_a of the acidic residue shifts to lower value.

Besides real effects caused by desolvation and coulombic interactions, the accuracy of pK_a values from electrostatic calculations may deviate from experimental data because of a number of reasons, e.g., the use of static crystallographic structures, which do not account for the dynamic nature of proteins, the solvation model used, the choice of protein dielectric constant, the selection of the atomic partial charges or the selection of the atom of the charged group to add the positive or negative unit charge, and the choice of van der Waals radii. In the case of GART, the presence of the inactive unformylated pseudocofactor 5dTHF, instead of formylated THF in the crystallographic structure, may also alter the electrostatic properties of the catalytic site. However, comparative studies of pK_a values from theoretically altered forms of crystallographic structure of GART (e.g., theoretical mutations, deletions, formyl addition, and local minimization) provide reliable insight into the electrostatic properties of the catalytic site, as shown in our study.

In our effort to elucidate pairwise electrostatic interactions in the catalytic site of GART, we constructed theoretical mutants using the crystallographic coordinates of the ternary complex of GART: structures 1cdeH108(N_ɛ2H) and 1cdeH108(N_ɛ2H)/flip. Specifically, we probed the conformational coupling of GAR and 5dTHF with catalytically important and highly conserved ionizable residues His108, Asp144, and His137 (Fig. 2 ▶). First, we performed calculations on the wild-type form of GART with His108 protonated at N_ɛ2 in both its unflipped and flipped position. Then we studied the effect of the theoretical mutation His108Ala in the electrostatic properties of residues in the catalytic site. Next we studied the theoretical mutant Asp144Ala with His108 in its unflipped and flipped position (Table 2A). We repeated the whole set of calculations after deleting substrate GAR and cofactor 5dTHF (Table 2B). These theoretical constructs are based on the real 1cde crystallographic structure under the assumption that no other structural changes occur during the His108, Asp144 theoretical mutations or deletions of GAR and 5dTHF. In reality this may not be the case in all constructs; indeed, we know that when substrate and cofactor are not present, the interaction between His108 and Asp144 is lost because of the reorientation of the binding loop. However, the theoretical constructs are used here as an aid to isolate the electrostatic codependence of His108, Asp144, GAR, and 5dTHF. Significant variations in pK_a values within the five calculations in each of Tables 2A and 2B are observed for six residues/groups: His108, His121, His132, His137, Asp144, and GAR(NH₂). We examined the nature of the electrostatic interactions that are responsible for the calculated pK_a values of these residues/groups.

Table 2A.

Effect of His108Ala (H108A), Asp144Ala (D144A) theoretical mutations and His108 κ2-torsion angle flip on the pKas of key residues/groups for catalysis, using crystallographic structures 1cdeH108(Nɛ2H) and 1cdeH108(Nɛ2H)/flipa

	1cdeH108(Nɛ2H)			1cdeH108(Nɛ2H)/flip
	κ2(H108) = 0°			κ2(H108) = 180°
	WTc	H108A	D144A	D144A	WT
Ionizable residue/groupb	pKa (apparent)
HIS54	7.2	7.2	7.2	7.1	7.2
HIS73	6.6	6.6	6.6	6.6	6.6
ARG90d	15.9	15.9	15.8	15.8	15.9
HIS99	6.0	6.0	6.0	5.9	6.0
HIS108	6.9		2.6	−1.2	3.9
HIS119	5.7	5.8	5.7	5.6	5.7
HIS121	7.0	7.0	7.0	7.0	6.9
HIS132	0.4	1.0	0.4	0.7	0.5
HIS137	2.0	3.5	−0.6	0.2	2.4
ASP144	−2.3	−0.5			−2.4
GLU173	1.7	2.4	2.1	2.5	1.9
HIS174	6.8	6.8	6.8	6.8	6.8
TYR177d	15.7	15.7	15.2	15.3	15.8
HIS192	6.3	6.3	6.3	6.3	6.3
GAR(NH2)	5.8	8.1	4.7	6.8	8.6
GAR PO42−	2.7	2.6	2.6	2.6	2.6
GAR PO4−	<−5	<−5	<−5	<−5	<−5
5dTHF HN3-C4=O	10.9	10.8	9.8	9.8	10.9
5dTHF CγOO−	2.7	2.7	2.6	2.6	2.7
5dTHF CαOO−	2.6	2.7	2.4	2.5	2.6

^a The side chain of His108 possesses initial hydrogen at N_ɛ2.

^b Besides all 10 histidines, Asp144, GAR, and 5dTHF, residues with relative pK_a shifts >1 pK_a unit within the 10 calculations of Tables 2A and 2B, are shown. Significant pK_a perturbations because of mutations or His108 flip state are in bold face. pK_a perturbations attributed to substrate binding (Tables 2A, 2B) are in italics.

^c WT stands for wild type and refers to the parent structure.

^d The pK_as of Arg90 and Tyr177 are very high to be of significant biological relevance, but they are listed here to show their interaction with substrate/cofactor, possibly important for binding.

Table 2B.

Effect of His108Ala (H108A), Asp144Ala (D144A) theoretical mutations, deletions of GAR, and 5dTHF and His108 κ2-torsion angle flip on the pKas of key residues for catalysis, using crystallographic structures 1cdeH108(Nɛ2H) and 1cdeH108(Nɛ2H)/flipa

	1cdeH108(Nɛ2H)			1cdeH108(Nɛ2H)/flip
	κ2(H108) = 0°			κ2(H108) = 180°
	WTc	H108A	D144A	D144A	WT
Ionizable residueb	pKa (apparent)
HIS52	7.1	7.1	7.1	7.1	7.1
HIS73	6.6	6.6	6.6	6.6	6.6
ARG90d	14.3	14.4	14.2	14.2	14.4
HIS99	6.0	6.0	6.0	6.0	6.0
HIS108	7.0		4.5	4.9	7.4
HIS119	5.2	5.4	5.2	5.2	5.2
HIS121	5.6	6.1	5.7	5.8	5.7
HIS132	0.7	1.4	0.6	0.7	0.8
HIS137	4.7	5.7	2.5	2.3	4.6
ASP144	0.7	2.4			0.5
GLU173	0.4	0.7	0.3	0.3	0.4
HIS174	6.1	6.1	6.1	6.1	6.1
TYR177d	14.4	14.4	13.9	14.0	14.5
HIS192	6.3	6.3	6.3	6.3	6.3

^a The side chain of His108 possesses initial hydrogen at N_ɛ2.

^b Besides all 10 histidines, Asp144, GAR, and 5dTHF, residues with relative pK_a shifts >1 pK_a unit within the 10 calculations of Tables 2A and 2B, are shown. Significant pK_a perturbations because of mutations or His108 flip state are in bold face. pK_a perturbations attributed to substrate binding (Tables 2A, 2B) are in italics.

^c WT stands for wild type and refers to the parent structure.

^d The pK_as of Arg90 and Tyr177 are very high to be of significant biological relevance, but they are listed here to show their interaction with substrate/cofactor, possibly important for binding.

The reorientation of the binding loop at high pH brings Asp144 into the catalytic site where it stabilizes substrate GAR (through NH₂ group) as well as His108 and His137 (Fig. 2 ▶). This is evident in the pK_a calculations, in which Asp144 always shows a significantly lower pK_a than its model value (Table 2A). In the absence of His108, the pK_a value of Asp144 increases (Table 2A), and in the absence of both His108 and GAR(NH₂), its pK_a further increases, but still remains lower than its model pK_a value, caused by the presence of His137 (Table 2B).

Likewise, the pK_a of His108 is lowered in the Asp144Ala mutant because of removal of favorable coulombic interaction (Table 2A), but is further increased when in addition to the absence of Asp144, unfavorable coulombic interaction with GAR(NH₂) is removed (Table 2B). The pK_a of His108 in the theoretical construct, which involves Asp144Ala mutation and GAR/5dTHF deletion, is 4.5 (Table 2B), which is attributed to the combined effect of desolvation and unfavorable coulombic interactions with His137. Of interest is the swing of the see-saw effect (vide supra), which is responsible for reversing the magnitude of the pK_as of His108 and GAR(NH₂), depending on the presence or absence of Asp144 (Table 2A, columns 2 and 4). Thus, Asp144 plays an important role in the relative protonation state of the pair His108-GAR(NH₂).

pK_a calculations with a realistic catalytic site and comparison with experimental results

We focused our final efforts in a construct of a more realistic active site by including the 5dTHF formyl group (HC=O) in structure 1cde. When we added a formyl group at the N₁₀ atom of the pseudocofactor 5dTHF by generating coordinates with proper geometry and hybridization, the mere presence of partial charges in formyl's C, O, and H atoms produced variations of 0.2–1.8 units in the pK_a of GAR(NH₂), depending on the orientation of the formyl group. However, these calculations were not performed on a sterically optimal catalytic site, because the added formyl group produced several clashes within the congested catalytic site, with GAR(NH₂) involving formyl's C, O, and H atoms and GAR's hydrogens of NH₂ and C₂₃ (carbon preceding GAR's NH₂; Fig. 1B ▶). It was apparent that a local minimization was necessary to create a less congested and more realistic catalytic site. Local minimization was performed by allowing movement of only the formylated pseudocofactor (5dfTHF) and substrate GAR, which were involved in the van der Waals clashes, while keeping the rest of the protein structure fixed. Local minimization was performed using the two alternative orientations of the formyl group, which accounted for swapping the C=O and C-H bonds; however, the local minimization was successful with only one of these orientations in optimizing clashes within the catalytic site.

The local minimization introduced a translation to the nitrogen atom of the amide group of GAR by 2.5 Å, which did not alter significantly GAR's relative orientation to His108, but it was sufficient to move GAR outside the hydrogen bond distance with His108. When we performed the global hydrogen-bonding network optimization of WHAT IF (Nielsen et al. 1999) on the formylated locally minimized structure, called hereafter 1cde-min5dfTHF-H108(N_ɛ2H)/flip, the side chain of His108 was flipped back to its original position while still retaining the hydrogen at N_ɛ2. We called this final and more realistic structure 1cde-min5dfTHF-H108(N_ɛ2H), which contained the reconstructed formylated pseudocofactor 5dfTHF, minimized coordinates for 5dfTHF and GAR, and most stable His108 state (unflipped with N_ɛ2H). We repeated the pK_a calculations using this final structure, 1cde-min5dfTHF-H108(N_ɛ2H). For comparison, we repeated the pK_a calculations using its parent structure, 1cde-min5dfTHF-His108(N_ɛ2H)/flip, which contained the reconstructed formylated pseudocofactor 5dfTHF, minimized coordinates for 5dfTHF and GAR, and a flipped His108 with hydrogen at N_ɛ2.

The results of the pK_a calculations of formylated locally minimized structure 1cde-min5dfTHF-H108(N_ɛ2H) are summarized in Table 3. Also, we have included in Table 3 the results of the pK_a calculations using the structure 1cde-min5dfTHF-H108(N_ɛ2H)/flip for comparison. The magnitudes of the pK_as of His108 and GAR(NH₂) are reversed when using either structure 1cde-min5dfTHF-H108(N_ɛ2H) or 1cde-min5dfTHF-H108(N_ɛ2H)/flip (Table 3), compared to the output of the calculation using structure 1cdeH108(N_ɛ2H)/flip) (Table 1). The more realistic catalytic site of the optimal minimized structure 1cde-min5dfTHF-H108(N_ɛ2H) has produced an elevated pK_a for His108 of 7.2 and a lower pK_a for GAR(NH₂) of 3.6 (Table 3), compared to pK_as of 3.9 and 8.6 for His108 and GAR(NH₂), respectively, of the optimal unminimized structure 1cdeH108(N_ɛ2H)/flip (Table 1).

Table 3.

pKas of key ionizable residues/groups of formylated locally minimized 1cde GART structure with His108 in its original (unflipped, κ2(H108) = 0°) state, structure 1cde-min5dfTHF-H108(Nɛ2H); and in its flipped (κ2(H108) = 180°) state, structure 1cde-min5dfTHF-H108(Nɛ2)/flip

		1cde-min5dfTHF-H108(Nɛ2H) κ2(H108) = 0°	1cde-min5dfTHF-H108(Nɛ2H)/flip κ2(H108) = 180°
Ionizable residue/groupa,b	pKa (model)	pKa (apparent)
HIS54	6.3	7.2	7.2
HIS73	6.3	6.6	6.6
ARG90	12.0	16.8	16.8
HIS99	6.3	6.0	6.0
HIS108c	6.3	7.2	7.2
HIS119	6.3	5.6	5.6
HIS121	6.3	7.0	6.9
HIS132	6.3	0.7	0.8
HIS137	6.3	3.5	4.0
ASP144	4.0	−2.0	−2.1
GLU173	4.4	−0.4	−0.3
HIS174	6.3	6.5	6.5
HIS192	6.3	6.4	6.4
GAR(NH2)	8.0	3.6	5.3
GAR PO42−	4.5	5.5	5.5
GAR PO4−	1.1	−4.2	−4.2
5dTHF HN3-C4=O	10.3	10.6	10.6
5dTHF CγOO−	4.4	2.9	2.9
5dTHF CαOO−	3.5	2.1	2.1

^a GART structure 1cde with added formyl group in pseudocofactor was locally minimized for GAR and 5dfTHF, whereas the conformation of the rest of the enzyme was kept fixed.

^b Besides all 10 histidines, Asp144, ionizable groups of GAR and 5dTHF, residues with relative pK_a shifts >0.4 pK_a units compared to the corresponding calculations of unminimized GART structures, are shown. Significant pK_a perturbations attributed to the local minimization of the formylated locally minimized structures, are in bold face.

^c Initial hydrogen of His108 is at position N_ɛ2.

In addition, theoretical mutations of His108Ala and Asp144Ala of structure 1cde-min5dfTHF-H108(N_ɛ2H) have been constructed and pK_a values have been calculated (data not shown). The pK_a of GAR(NH₂) increases from 3.6 (Table 3) to 7.7 in the His108Ala mutant and to 4.2 in the Asp144Ala mutant of structure 1cde-min5dfTHF-H108(N_ɛ2H). The results of these calculations show the sensitivity of the pK_a of GAR(NH₂) depending on mutation of His108 and Asp144 in agreement with the results summarized in Table 2A for the unminimized structures.

In agreement with Figure 3 ▶, the complex electrostatic interdependence among His108-His137-Asp144-GAR(NH₂), in the formylated locally minimized structures, is maintained and is shown by the nonsigmoidal shapes of the theoretical titration curves of these ionizable sites in Figure 4 ▶. A weighted superposition of the titration curves of His108, His137, Asp144, and GAR(NH₂) is more obvious in Figure 4B ▶. Figure 5 ▶ presents titration curves of ionizable sites Lys114, Tyr115, His119, His121, Arg122, Glu126, His132, GAR(PO₄⁻, PO₄⁼), and 5dfTHF(HN₃-C₄=O, C_αOO⁻, C_γOO⁻), which are located in, or in proximity to, the catalytic site to show the quality of the data. Only the titration curves of His121 and His132 deviate from the typical sharp sigmoidal transition, which reflects their strong coulombic interaction (Morikis et al. 2001).

Fig. 4.

Theoretical titration curves of interacting catalytic site ionizable sites, His108 (squares), His137 (diamonds), Asp144 (triangles), and GAR(NH₂) (circles) of formylated locally minimized 1cde GART structure with (A) His108 in its original (unflipped, κ₂(H108) = 0°) state, structure 1cde-min5dfTHF-His108(N_ɛ2H), and (B) in its flipped (κ₂(H108) = 180°) state, structure 1cde-min5dfTHF-His108(N_ɛ2H)/flip. In both structures, His108 has an initial hydrogen at N_ɛ2 position. Quoted pK_as in Table 3 corresponds to pH values of charge +0.5 or −0.5.

Fig. 5.

This figure shows the quality of titration curves from calculations using the formylated locally minimized GART structure 1cde-min5dfTHF-H108(N_ɛ2H). Plots for selected residues in, or in the proximity of, the catalytic site are shown. Panel (A) shows titration curves of positive ionizable sites, Lys114 (solid squares), His119 (solid diamonds), His121 (solid triangles), Arg122 (solid circles), and His132 (open squares), and 5dfTHF(N₃) (open circles). Panel (B) shows titration curves of negative ionizable sites, Tyr115 (solid squares), Glu126 (solid diamonds), GAR(PO₄⁻) (solid triangles), GAR(PO₄⁼) (solid circles), 5dfTHF(C_γOO⁻) (open squares), and 5dfTHF(C_αOO⁻) (open circles).

Overall, we have shown that His108, His137, Asp144, and GAR(NH₂) show significant pK_a variations within the series of calculations we have performed, and they also show significant differences from their model pK_a values. These variations depend on the local electrostatic microenvironment of the catalytic site.

Discussion

The origin and properties of His108-GAR(NH₂) electrostatic coupling

To investigate the physical properties of the pK_a see-saw effect, we discuss the significance of (1) the proximity of GAR(NH₂) and His108, (2) GAR and His108 desolvation and coulombic interactions with Asp144 and His137, and (3) the model pK_as of GAR and His108, as follows:

(1) We have found a dependence of the see-saw effect and the distance between GAR(N) and the site of addition of +1 charge in His108. This can be translated as the distance between GAR and His108. In the case of His108 with an initial hydrogen at N_ɛ2 and flipped by 180°, in which the distance of the site of addition of unit charge on His108 (N_δ1) from GAR(N) is the shortest, i.e., ∼3.2 Å (Fig. 2 ▶), the see-saw effect brought the pK_a of GAR(NH₂) higher than the pK_a of His108 (Table 1). In all the other three cases in which the distances of the site of unit charge addition from GAR(N) were ∼5.3 Å (initial hydrogen at N_ɛ2, unflipped His108, charge added at N_δ1), ∼4.1 Å (initial hydrogen at N_δ1, unflipped His108, charge added at N_ɛ2), ∼5.3 Å (initial hydrogen at N_δ1, flipped His108, charge added at N_ɛ2), the see-saw effect brought the pK_a of His108 higher than the pK_a of GAR(NH₂) (Table 1; Fig. 2 ▶). We have shown with theoretical mutations that the presence of Asp144 was also relevant for the pK_a values of GAR(NH₂) and His108. The distance between Asp144 C_γ to the four possible geometric sites of charge addition in His108, discussed above, was very similar, i.e., ∼3.8, ∼4.1, ∼4.0 and ∼4.2 Å, respectively, and did not introduce further complexity to our arguments.

(2) We have shown with the theoretical mutations/deletions that the absence of GAR from the catalytic site introduced an increase of the pK_a of His108 from 3.9 to 7.4 (Tables 2A, 2B, last columns). This pK_a reflects a combined effect of desolvation and coulombic interactions with Asp144 and His137. On the other hand, the absence of His108 only caused a reduction of the pK_a of GAR(NH₂) by 0.5 pK_a units (Table 2A, columns 3 and 6). In a similar control calculation in which GAR was present but GAR(NH₂) was not treated as an ionizable site, the pK_a of His108 increased from 3.9 to 8.1 (data not shown). The treatment of GAR(NH₂) as nonionizable site also introduced a direct effect in the pK_a of Asp144 (increased from −2.4 to −0.9) and a relay effect on His137 (increased from 2.4 to 3.5), which is in agreement with the results of Tables 2A and 2B.

(3) We also performed control calculations in a hypothetical situation in which His108 and GAR(NH₂) had the same propensity for a hydrogen atom, which would be reflected in possessing the same model pK_as. In the case in which His108 had the same model pK_a of 8.2 as GAR(NH₂), the final pK_as were 10.1 and 5.2 for His108 and GAR(NH₂), respectively. In the case in which GAR(NH₂) had the same model pK_a of 6.3 as His108, the final pK_as were 8.1 and −0.6 for His108 and GAR(NH₂), respectively. Last, in the case in which His108 and GAR(NH₂) had the same model pK_a of 7.15, which is the mean value of their model pK_as, their calculated pK_as were 9.1 and 1.2 for His108 and GAR(NH₂), respectively.

In summary, we established that the electrostatic microenvironment of the catalytic site, in the absence of GAR, favors an elevated pK_a value for His108, whereas the electrostatic microenvironment of the catalytic site, in the absence of His108, leaves the pK_a of GAR(NH₂) relatively unchanged. This means that His108 has a propensity for high pK_a within the catalytic site. We also showed that GAR(NH₂) has a propensity for high apparent pK_a because of its high model pK_a of 8.2. In addition, we showed that variation of the distance of GAR(NH₂)-His108 ring is responsible for shifting the balance of the see-saw effect. The distances of Asp144 to GAR(NH₂) and to His108 are critical for the see-saw effect, and the presence of His137 is essential for stabilization of Asp144.

Comparison of calculations with experimental results

A catalytic mechanism of GART was first proposed by Klein et al. (1995) and subsequently modified by Su et al. (1998). Shim and Benkovic (1999) have conducted experimental kinetic studies to measure k_cat-pH, k_cat/K_m(GAR)-pH, and k_cat/K_m(fDDF)-pH profiles of wild-type, His108Ala, Asp144Ala, and His121Gln GART, and have suggested a current catalytic mechanism for GART involving a tetrahedral transition state and a water molecule (Fig. 6 ▶). His108 is proposed to be in a protonated form stabilized through a salt bridge to Asp144. The formyl group of fTHF is hydrogen bonded to the protonated His108 and Asn106. GAR, in neutral form, acts as a nucleophile in attacking the formyl group to form the tetrahedral transition state. A water molecule, is also in the vicinity forming hydrogen bonds to GAR(NH₂), 5dTHF(N₁₀), and Asp144 (Fig. 6 ▶) to assist catalysis and the breakdown of the transition state (Shim and Benkovic 1999). This model is consistent with the involvement of two ionizable groups in catalysis with pK_a values of 9.7, assigned to protonated His108, and 6.6–6.9, assigned to GAR(NH₂). These pK_a values are extracted from the bell-shaped k_cat-pH profile of wild-type GART. The k_cat-pH profile remained unchanged in the acidic side but was lost at the basic side in the His108Ala and Asp144Ala mutants, implicating the pair of His108-Asp144 as being responsible for the high pK_a of 9.7. The assignment of the acidic part of the k_cat-pH profile to GAR(NH₂) is consistent with the fact that GAR should be in its neutral form to act as a nucleophile to attack and detach the formyl group of the cofactor (Shim and Benkovic 1999). Saturation experiments of enzyme-GAR complex and free substrate fDDF produced k_cat/K_m(fDDF)-pH curves similar to k_cat-pH profiles of the wild type with a bell shape that was not lost in the His108Ala and Asp144Ala mutants. The pK_as were 6.3–7.0 and 9.4, and they were assigned to GAR(NH₂) and to HN₃-C₄=O of free fDDF, respectively (Shim and Benkovic 1999). Finally, saturation experiments of enzyme-fDDF complex and free GAR produced bell-shaped curves in all wild-type, His108Ala, and Asp144Ala mutants. Their pK_as were 7.6–8.0 and 8.6–8.8, which were assigned to the NH₂ group of free GAR and the activation loop (Shim and Benkovic 1999).

Fig. 6.

The tetrahedral transition state of GART (in blue) proposed by Shim and Benkovic (1999) and adjusted for the flip state of His108. According to this model, the protonated His108 forms hydrogen bonds with Asp144 and formyl group oxygen, which is also stabilized by a hydrogen bond with Asn106. A water molecule (in red) is involved to assist a proton transfer from GAR to N₁₀ of tetrahydrofolate. We have also included His137 (in magenta), which forms a hydrogen bond with Asn106. Hydrogen bonds are depicted with dashed lines.

Our pK_a calculations on the ternary complex (GART + GAR + 5dTHF), which best resembles the catalytic system of the experimental studies (GART + GAR + fDDF) or the actual catalytic system (GART + GAR + THF), do not show an elevated pK_a for His108 to 9.7; however, they do show the see-saw effect for His108-GAR(NH₂) in which the pK_a of His108 increases when the pK_a of GAR(NH₂) decreases and vice versa, depending on the position of the initial hydrogen and flip state of His108 (vide supra, Table 1).

A superficial inspection of the entries in Table 1 points out that columns 4 (unflipped His108, initial proton at N_δ1) and 5 (flipped His108, initial proton at N_δ1) and to a lesser extent, column 2 (unflipped His108, initial proton at N_ɛ2), are in qualitative agreement with the measured k_cat-pH profiles by Shim and Benkovic (1999), but they do not generate the absolute magnitude of the experimentally modeled pK_as of His108 and GAR(NH₂). However, this qualitative agreement appears to be fortuitous because none of these three His108 initial protonation/flip states is optimal. A visual inspection of the structure and the output of global hydrogen bond optimization by WHAT IF suggest that His108 is found in the most stable hydrogen-bonding environment only when it initially possesses a hydrogen at N_ɛ2, and it is flipped by 180°. The calculation of GART pK_as with this His108 state produced the results in column 3 of Table 1. Yet, the output of this calculation is in disagreement with the results of Shim and Benkovic (1999), because the see-saw effect has brought the pK_a of GAR(NH₂) higher than the pK_a of His108.

The final pK_a values of 7.2 and 3.6 for His108 and GAR(NH₂), respectively, from the calculations using the locally minimized structure, are also in qualitative agreement with the experimentally measured k_cat-pH profiles (Shim and Benkovic 1999), despite the fact that they do not reproduce the absolute magnitude of the pK_as of 9.7 and 6.8 that were produced from the fits of the k_cat-pH profile curves (Shim and Benkovic 1999). This disagreement is probably because of the fact that we are using static time-averaged crystallographic structures, which do not take into account the dynamic nature of the protein. Nevertheless, taking into account the fact that GAR(NH₂) is favored to possess a high pK_a because of its model pK_a of 8.2 (vide supra), the significant lowering of its pK_a to 3.6 in our calculations shows that GAR is found in an unusual electrostatic environment and involved in nontrivial interactions with His108 and Asp144, which are further stabilized through interactions with His137 (Fig. 2 ▶; Table 2). Likewise, the electrostatic environment of the catalytic site favors a high pK_a (compared to its model pK_a) for His108 (vide supra). If we could hypothesize a bell-shaped k_cat-pH profile with pK_a of the acidic leg of 3.6 for GAR(NH₂) and pK_a of the basic leg of 7.2 for His108, our results would be in agreement with the fact that substrate GAR needs to be in neutral form to act as a nucleophile for catalysis, and His108 needs to be in protonated form to act as acid catalyst, as suggested by Shim and Benkovic (1999).

In our attempt to theoretically explain the experimental data by constructing the mutations used in the pH-rate profiles, His108Ala and Asp144Ala, we found a disagreement with the experimental data. Specifically, we found that the pK_a of GAR(NH₂) can fluctuate between 0.5 and 2.3 units in the His108Ala theoretical mutant, and between 1.1 and 1.8 units in the Asp144Ala theoretical mutant (Table 2A). In addition, increases in the pK_a of GAR(NH₂) of 4.1 and 0.6 units have been observed for the His108Ala and Asp144Ala theoretical mutants, respectively, of the optimal minimized structure of GART (vide supra). These observations are in agreement with the strong His108-GAR(NH₂) and Asp144-GAR(NH₂) interaction, as discussed in Results. However, when comparing the measured k_cat-pH profiles of wild-type GART and His108Ala and Asp144Ala mutants, the modeled pK_a value of GAR(NH₂) remained very robust with a value of about 6.8 in all three cases (Shim and Benkovic 1999). This is in disagreement with our proposed strong electrostatic coupling among His108-GAR(NH₂)-Asp144. It is possible that when making a site-directed mutation, several other conformational changes occur that are not modeled by our theoretical mutant constructs. It can be argued, though, in the case of GART, that these changes cannot be large enough to abolish substrate binding, which has been observed experimentally for the His108Ala and Asp144Ala mutants (Shim and Benkovic 1999). Also, in the case in which Asp144 is oriented away from the catalytic site (and His108) in the high-pH substrate-free structure of E70A GART (Su et al. 1998), the arrangement of the rest of the catalytic site is nearly identical to substrate-bound structure. This suggests that even in the absence of Asp144, the strong electrostatic dependence of His108-GAR(NH₂) should remain intact, an effect that is not seen in the modeled experimental data (Shim and Benkovic 1999).

In summary, although our calculations do not reproduce the absolute pK_a values that were proposed by fitting experimental pH-rate profiles, they are in qualitative agreement of the current catalytic mechanism of GART. However, our pK_a studies using theoretical mutants His108Ala and Asp144Ala significantly affect the pK_a of GAR(NH₂), an effect that is not accounted by the modeling of the experimental mutation data (Shim and Benkovic 1999). Further theoretical work involving a molecular dynamics simulation of GART with formylated pseudocofactor, including a water molecule in the catalytic site as proposed by Shim and Benkovic (1999), is necessary for a more accurate representation of the catalytic site.

Conclusions

We have shown that the initial protonation and flip state of catalytic residue His108 is responsible for significant variations in the pK_a values of His108 and substrate ionizable site GAR(NH₂). We have observed a sensitive structural and electrostatic interdependence between His108 and GAR(NH₂). A proton can be exchanged between His108 and GAR(NH₂) based on the structural and electrostatic microenvironment of the catalytic site. A see-saw effect has been observed, which depends on the location of the proton and is responsible for reversing the magnitudes of the pK_as of His108 and GAR(NH₂). When the pK_a of His108 is elevated, compared to the pK_a of GAR(NH₂), His108 is protonated and GAR(NH₂) is neutral. In contrast, when the pK_a of GAR(NH₂) is elevated, compared to the pK_a of His108, His108 is neutral and GAR(NH₂) is protonated. We have constructed theoretical mutation and deletion analogs, which have aided us in elucidating pairwise electrostatic interactions between key ionizable sites within the catalytic site. The His108-GAR(NH₂) interaction is finely modulated by the presence of negatively charged Asp144, which stabilizes both sites. More complexity is introduced in the stability of the catalytic site and in preparation for catalysis by the presence of a neutral His137, which strongly interacts with both Asp144 and His108, and conserved catalytic residue Asn106. A relay electrostatic interaction of His137 with GAR(NH₂) is present through Asp144 and His108. Small conformational changes tend to stabilize-destabilize the catalytic site and can swing the see-saw effect of pK_a toward a protonated His108 (with high pK_a)-unprotonated GAR(NH₂) (with low pK_a), or vice versa. We have established that His108 has a higher propensity for a proton (high pK_a) than GAR(NH₂), within the catalytic site environment, and GAR(NH₂) has a higher propensity for a proton (high pK_a) than His108 because of its high model pK_a.

We have shown that the distance GAR(NH₂)-His108 in the crystallographic structure, which contains the unformylated pseudocofactor 5dTHF, does not represent a realistic situation. We performed our final pK_a calculations after we added the missing formyl group in the pseudocofactor and locally minimized the crystallographic structure 1cde for 5dfTHF and GAR, while leaving the coordinates of the rest of the enzyme structure fixed. The final pK_a values of 7.2 and 3.6 for His108 and GAR(NH₂), respectively, are in qualitative agreement with the catalytic mechanism proposed by modeling the experimentally measured k_cat-pH profiles (Shim and Benkovic 1999), despite the fact that they do not reproduce the absolute magnitude of the pK_as of 9.7 and ∼6.8 that were produced from the fits of the k_cat-pH profile curves (Shim and Benkovic 1999). In addition, our theoretical mutations His108Ala and Asp144Ala have shown the presence of a strong electrostatic coupling between His108-GAR(NH₂) and Asp144-GAR(NH₂), which is reflected in the significant variations in the pK_a of GAR(NH₂) on mutation. However, this effect is not present in the modeled pK_a value of GAR(NH₂) from the experimental pH-rate profiles of His108Ala and Asp144Ala mutants. These disagreements can possibly be attributed to the fact that the time-averaged crystallographic structure used in the calculations is missing the dynamic nature of the enzyme, which can be described more accurately as a fluctuating ensemble of different states. In conclusion, our results support the current catalytic model of Shim and Benkovic (1999), according to which GAR should be in neutral form to act as a nucleophile to detach the formyl group from the cofactor, whereas His108 is acting as an acid catalyst, but further theoretical and experimental work is necessary to resolve ambiguities.

Materials and methods

The program UHBD (Madura et al. 1994, 1995) was used for Poisson-Boltzmann continuum electrostatic calculations as discussed in Morikis et al. (2001). The partial charge and van der Waals radii parameter set PARSE (Sitkoff et al. 1994) for the 20 amino acids was used in the calculations. Additional entries for substrate GAR and pseudocofactor 5dTHF were generated, consistent with PARSE rules. Changes from the neutral to the charged state of ionizable sites of substrate GAR and pseudocofactor 5dTHF were made by adding a positive unit charge at N₃ of 5dTHF and the amino group of GAR, and a negative unit charge at C_γ and C_α of 5dTHF, and the phosphate group of GAR. There are no coordinates for the carboxy-terminal residue of crystallographic structure 1cde of GART (Almassy et al. 1992), so we did not include an ionizable carboxyl terminus in the calculations. The experimental model pK_a for the substrate GAR were 8.2 for the amino group and 6.1 and 1.0 for the two ionization states of the phosphate group. Only one pK_a value (6.1) was reported for the doubly ionizable PO₄²⁻ group of GAR (Hartman et al. 1956). The second value of 1.0 was estimated by subtracting from 6.1 the difference of the pK_a values of phosphate group in solution for the equilibria H₃PO₄ |Zu H₂PO₄⁻ and H₂PO₄⁻ ↔ HPO₄²⁻, which are 2.1 and 7.2, respectively (Brey 1978). These values are consistent with proton dissociation constants of nucleotides (Garrett and Grishnan 1995). The experimental pK_a values for the cofactor 5dTHF were 10.5 for the N₃ group, 4.8 for the γ-carboxy and 3.5 for the α-carboxy, in analogy to 5,6,7,8-tetrahydrofolate (Kallen and Jencks 1966) or similarly, to 7,8-dihydrofolate (Poe 1977). The experimental pK_a values of GAR and 5dTHF were adjusted in the calculation to account for intramolecular electrostatic interactions at their fixed conformations in structure 1cde, as it was indicated by the output of pK_a calculations of free GAR and free 5dTHF. The adjusted pK_a values were 8.0 for NH₂ and 4.5 and 1.1 for the phosphate of GAR, and 10.3, 4.4, 3.5 for N₃, γ-carboxy and α-carboxy, respectively, of 5dTHF. The N₁₀ group of folate-free 5dTHF was not treated as an ionizable site in the final calculations because of its very low pK_a value of −1.3. Comparisons with initial calculations, in which the N₁₀ group was taken explicitly into account, were essentially identical.

Theoretical mutations His108Ala and Asp144Ala, and theoretical deletions for GAR and 5dTHF, were constructed in crystallographic structure 1cde (Almassy et al. 1992), using the program WHAT IF v. 99 (Vriend 1990). Addition of hydrogen atoms in GART structure 1cde and its theoretical mutation and deletion analogs was achieved by performing optimization of the hydrogen-bonding network (Nielsen et al. 1999; Hooft et al. 1996) using the program WHAT IF (Vriend 1990). This type of addition of hydrogen atoms in crystallographic structures is particularly important in the case of certain histidine residues in which the initial protonation state (at the N_δ1 or N_ɛ2) can significantly affect the computed pK_a values. This method was clearly an improvement compared to local optimization methods based on hydrogen bonding, solvent accessibility, and neighboring with other charged residues, which were also tested in the initial stages of the calculations. The initial protonation states for histidine residues in structure 1cde, determined by WHAT IF, were His54(N_ɛ2H), His73(N_ɛ2H), His99(N_ɛ2H), His108(N_ɛ2H), His119(N_ɛ2H), His121(N_δ1H), His132(N_δ1H), His137(N_δ1H), His174(N_ɛ2H), and His192(N_ɛ2H). During the hydrogen-bonding network optimization, the program WHAT IF also corrected for the side-chain orientation of nonoptimal histidine, asparagine, and glutamine residues (Nielsen et al. 1999). The following side-chain flips were performed by WHAT IF during the hydrogen-bonding network optimization in GART structure 1cde: Asn13, Asn36, and His174 in substrate-free form, and in addition to these three flips, His108 in substrate-bound form. Addition of hydrogens to side-chain carboxylates (O_δ2 for Asp and O_ɛ2 for Glu) was made using a special version of WHAT IF provided to us by J. Nielsen and G. Vriend (pers. comm.). Addition of hydrogens to substrate GAR and cofactor 5dTHF was made with WHAT IF and/or manually, using standard hybridization, with the program QUANTA (MSI). The program PRODGR (van Aalten et al. 1996) of the Dundee WWW server was used to convert the crystallographic coordinates of GAR and 5dTHF from GART structure 1cde (Almassy et al. 1992) to WHAT IF topology format. The program QUANTA was used to add the formyl group at N₁₀ atom of pseudocofactor 5dTHF to generate formylated pseudocofactor (5dfTHF). Local minimization for GAR and 5dfTHF of structure 1cde was performed with the program CNS (Brunger et al. 1998) using covalent geometry and van der Waals energy terms. GAR and 5dfTHF coordinates, including hydrogen atoms, were used to generate CNS-compatible topology and parameter files using the program XPLO2D (Kleywegt and Jones 1998) of the Uppsala HIC-Up Hetero-compound Information Centre WWW server. The initial protonation states for the 10 histidine residues in the formylated locally minimized structure 1cde, determined by WHAT IF, were as in the original 1cde structure (vide supra). Correction of side-chain flips were made for residues Asn13, Asn36, and His174, but not for His108, in the formylated locally minimized structure 1cde, using the global hydrogen-bonding network optimization of WHAT IF. Nomenclature compatibility among UHBD, WHAT IF, QUANTA, and CNS and preparation for UHBD runs, was achieved by a series of home-made PERL scripts.

Molecular graphics presented in this paper were created with the programs MOLMOL (Koradi et al. 1996) and Swiss PDB Viewer (Guex and Peitsch 1997).

Electronic supplemental material

A table with the proximity of key residues for catalysis and tables with the complete lists of pK_a values derived by calculations described in Tables 1, 2A, 2B, and 3 of the main text are included as supplemental material.

Abbreviations

• GAR, glycinamide ribonucleotide

• GART, GAR transformylase

• fGAR, formyl-GAR

• THF, tetrahydrofolate

• fTHF, N¹⁰-formyl-THF

• 5dTHF, 5-deaza-5,6,7,8-THF

• 5dfTHF, formylated 5dTHF

• fDDF, N¹⁰-formyl-5,8-dideazafolate

• PDB, Protein Data Bank