Native-state conformational dynamics of GART: A regulatory pH-dependent coil–helix transition examined by electrostatic calculations

Abstract

Glycinamide ribonucleotide transformylase (GART) undergoes a pH-dependent coil–helix transition with pK_a ∼ 7. An α-helix is formed at high pH spanning 8 residues of a 21-residue-long loop, comprising the segment Thr120–His121–Arg122–Gln123–Ala124–Leu125–Glu126–Asn127. To understand the electrostatic nature of this loop–helix, called the activation loop–helix, which leads to the formation and stability of the α-helix, pK_a values of all ionizable residues of GART have been calculated, using Poisson–Boltzmann electrostatic calculations and crystallographic data. Crystallographic structures of high and low pH E70A GART have been used in our analysis. Low pK_a values of 5.3, 5.3, 3.9, 1.7, and 4.7 have been calculated for five functionally important histidines, His108, His119, His121, His132, and His137, respectively, using the high pH E70A GART structure. Ten theoretical single and double mutants of the high pH E70A structure have been constructed to idey pairwise interactions of ionizable residues, which have aided in elucidating the multiplicity of electrostatic interactions of the activation loop–helix, and the impact of the activation helix on the catalytic site. Based on our pK_a calculations and structural data, we propose that: (1) His121 forms a molecular switch for the coil–helix transition of the activation helix, depending on its protonation state; (2) a strong electrostatic interaction between His132 and His121 is observed, which can be of stabilizing or destabilizing nature for the activation helix, depending on the relative orientation and protonation states of the rings of His121 and His132; (3) electrostatic interactions involving His119 and Arg122 play a role in the stability of the activation helix; and (4) the activation helix contains the helix-promoting sequence Arg122–Gln123–Ala124–Leu125–Glu126, but its alignment relative to the N and C termini of the helix is not optimal, and is possibly of a destabilizing nature. Finally, we provide electrostatic evidence that the formation and closure of the activation helix create a hydrophobic environment for catalytic-site residue His108, to facilitate catalysis.

Proteins are dynamic molecules that can be described as ensembles of states under physiological conditions. Conformational fluctuations between different states may be involved in regulating enzyme activity and/or protein– protein recognition events. Subtle changes in solution conditions may alter the conformational space available to a protein and thus its activity profile. Electrostatic interactions among charged groups play significant roles in local or global stabilization or destabilization when solvent charges are varied with pH. In several cases, protein pH-dependent conformational changes involve histidine residues. Histidines, which have a pK_a of 6.3 in free form in solution, are likely to participate in structural changes because of small variations around the physiological pH. A well characterized system, with structural, spectroscopic, and theoretical methods, is myoglobin, which undergoes pH-induced native-state local heme-pocket conformational change in its carbon monoxide form (Ansari et al. 1987; Morikis et al. 1989; Yang and Phillips 1996) which is distinct from pH-induced global unfolding (Hughson et al. 1990; Sage et al. 1991; Jennings and Wright 1993); both pH-dependent processes involve histidine residues (Morikis et al. 1989; Sage et al. 1991; Bashford et al. 1993; Yang and Honig 1994).

The enzyme glycinamide ribonucleotide transformylase (GART) participates in the de novo biosynthetic pathway of purines (Garrett and Grishnan 1995). GART undergoes two major conformational changes as a function of pH, a coil–helix transition (Almassy et al. 1992; Su et al. 1998) and a dimer–monomer transition (Almassy et al. 1992; Mullen and Jennings 1996; Su et al. 1998), both with pK_a ∼ 7. Specifically, an 8-residue segment (residues 120–127) of a 21-residue-long loop (residues 111–131), called hereafter the activation loop, becomes an α-helix at high pH, called hereafter the activation helix (Fig. 1 ▶). At high pH the activation loop–helix caps and shields the active site from solvent, to facilitate catalysis (Almassy et al. 1992 ; Greasley et al. 1999). In addition to the coil–helix transition, GART converts from a dimeric form at low pH to a monomeric form at high pH (Almassy et al. 1992; Mullen and Jennings 1996; Su et al. 1998). The two pH-dependent conformational changes appear to be independent because they occur at different surfaces of the enzyme; however, cooperativity cannot be excluded at present.

Fig. 1.

The relative topology of the active site of GART is shown, with the activation loop–helix and the binding loop. The ternary complex GART structure (PDB code 1cde) has been used to make the figure. The important catalytic-site histidine His108 with substrate GAR and pseudocofactor 5dTHF is drawn using an atomic color code (carbon in white, nitrogen in blue, oxygen in red, and phosphorus in yellow). The activation loop–helix and its histidines, His119 and His121, are drawn in yellow. The binding loop and its catalytically important residue Asp144 are drawn in blue. Histidines His132 and His137 located in strand β6, which connects the activation loop–helix with the binding loop, are drawn in green.

GART catalyzes the transfer of a formyl group from N¹⁰-formyl-tetrahydrofolate (N¹⁰-fTHF) to glycinamide ribonucleotide (GAR) to form formyl-GAR (fGAR). The presence of a folate derivative in catalysis makes GART an ideal target for rational design of antineoplastic or antimicrobial inhibitors (Greasley et al. 1999; Shim and Benkovic 1998,1999). A third conformational change in GART, which is pH-independent but substrate-dependent, reorients a mobile exposed loop (residues 140–145), called hereafter the binding loop, toward the catalytic site on substrate binding (Fig. 1 ▶; Almassy et al. 1992^; Chen et al. 1992^; Klein et al. 1995^; Su et al. 1998).

Figure 2 ▶ shows GART representations using four different crystallographic structures (references in legend to Fig. 2 ▶). The top panels (Fig. 2A ▶,B) are structures with bound substrates, and the bottom panels (Fig. 2C ▶,D) are structures of the E70A mutant. The structure in Figure 2A ▶ is one-half of a dimer; all other structures (Fig. 2B ▶–D) are from monomeric enzyme forms. The left panels (Fig. 2A,C) are low pH, inactive structures, and the right panels (Fig. 2B,D) are high pH, active structures. The flexible activation loop–helix of residues 111–131 and the mobile binding loop of residues 140–145, are depicted in yellow and blue, respectively, in Figure 2 ▶. In high pH structures (Fig. 2B,D ▶) the activation helix is formed, whereas in the low pH structures (Fig. 2A,C ▶) the activation loop is preserved, independent of the presence of substrate. In the substrate-bound structures (Fig. 2A,B), the binding loop has turned inward toward the catalytic site, independent of pH, but in the substrate-free structures (Fig. 2C,D), the binding loop is found in different orientations. The pHs of the crystal structures of Figure 2 ▶ are 6.3 (Fig. 2A ▶), 7.5 (Fig. 2B ▶), 3.5 (Fig. 2C ▶), and 7.5 (Fig. 2D ▶).

Fig. 2.

Different crystallographic structures of GART are used in this figure. In the top row are the structures with bound substrates (not shown), and in the bottom row are the substrate-free structures. (A,C) Structures at low pH; (B,D) structures at high pH. The pH-dependent coil–helix transition (activation loop–helix) is shown in yellow, and the substrate-dependent loop reorientation (binding loop) is shown in blue. Structures are shown at the same orientation after fitting the coordinates of the backbone C_α atoms. (A) Low pH ternary complex (PDB code 1gar; Klein et al. 1995). This is the low pH dimeric form, but only one monomer is shown here. The pseudosubstrate is deleted for clarity of the comparison. (B) High pH ternary complex (PDB code 1cde; Almassy et al. 1992). Substrate and pseudocofactor are deleted for clarity. (C) Low pH E70A mutant (PDB code 2gar; Su et al. 1998). This is a low pH monomeric form. (D) High pH E70A mutant (PDB code 3gar; Su et al. 1998). The apparent smaller length of the activation loop in the low pH structures (A,C) is attributable to the lack of crystallographic coordinates, a fact that is exhibited by lack of electron density because of flexibility. In all panels the C-terminal β-sheet (residues 188–209) has been deleted for clarity.

The activation loop–helix contains two histidines, His121 (in the helix segment) and His119 (in the loop segment, Fig. 1 ▶). A third histidine, His132, is the N-terminal residue of the β-strand (β6), adjacent to the C terminus of the activation loop (Fig. 1 ▶). Two more histidines, His108 and His137, both highly conserved, are located in the active site and in the proximity of the activation loop of GART (Fig. 1 ▶). His108 is located in the β-strand (β5) preceding the N terminus of the activation loop, and is directly involved in catalysis by stabilizing the tetrahedral transition-state complex (Shim and Benkovic 1999). His137 is located in the same β-strand as His132 and stabilizes catalytic residues His108 and Asp144 (Morikis et al. 2001). The binding loop contains the highly conserved residue Asp144, which is located in the active site of GART when substrates are bound, and stabilizes the tetrahedral transition state complex, while interacting with active site His108 and His137. The binding loop is adjacent to strand β6, which contains His132 and His137.

The pH-dependence of enzymatic catalysis, the two pH-dependent conformational changes, the presence of highly charged patches on the surface of GART, and the presence of 10 histidines (5 of which are in or in proximity to the active site), all point to the use of electrostatic modeling to understand the basic properties of GART structure– function.

In another work, we present pK_a calculations based on the solution of the linearized Poisson–Boltzmann equation, to understand the electrostatic nature of the catalytic site of GART (Morikis et al. 2001). Here we present pK_a calculations to probe the relation between the electrostatic properties and the coil–helix conformational transition of the activation loop–helix of GART. Two Escherichia coli GART structures of the E70A mutant, at high and low pH (Su et al. 1998), have been used for the pK_a calculations. The E70A GART is monomeric at both high and low pH. The mutation of Glu70, which is located at the dimer interface and forms intermonomer hydrogen bonds, disrupts dimerization at low pH (Mullen and Jennings 1998; Su et al. 1998). Missing coordinates (owing to lack of electron density) from the low pH E70A crystal structure of GART have been reconstructed, using the unfolding method of Elcock (1999), for the present calculations. Calculation of pK_a values (Bashford and Karplus 1990; Yang et al. 1993; Antosiewicz et al. 1994) of GART has been performed using Poisson–Boltzmann-type electrostatic calculations (for review, see Davis and McCammon 1990; Honig and Nichols 1995; Antosiewicz et al. 1996; Ullmann and Knapp 1999).

Although studies on helix–coil transitions have been typically performed on free peptides in solutions (for which significant literature has been created since the original theoretical work of Zimm and Bragg 1959), the activation helix of GART provides an opportunity for a study of a helix formation in a folded protein. We discuss the role of the side chains His119, His121, Arg122, Glu126, and His132 in the formation and stability of the activation helix. We also discuss the effect of the activation-helix formation on the catalytic site, especially the catalytically important active-site residues His108, His137, and the catalytically important binding-loop residue Asp144. A number of single and double theoretical mutants are used to elucidate pairwise interactions among ionizable residues.

Results

Table 1 compares the pK_as of key residues of GART from calculations using the coordinates of the two E70A GART structures, at high pH (PDB code 3gar; Su et al. 1998) and low pH (PDB code 2gar; Su et al. 1998). The definition of key residues includes all 10 histidines, the binding-loop residue Asp144 involved in catalysis, ionizable residues Arg122 and Glu126 of the activation helix, and ionizable residues that show a difference >1 pK_a unit within the two calculations. Residues showing pK_a differences when comparing the high and low pH structures (Table 1) are involved in pH-dependent conformational changes. The high pK_as (>9) of lysines, arginines, and tyrosines (Table 1) are not biologically significant, but they still show points of electrostatic interactions of residues involved in pH-dependent local conformational changes.

Table 1.

pKas of ionizable residues of high and low pH E70A GART structures

Ionizable residuea	Model pKab	High pH (3gar)c	Low pH (2gar)d
His54	6.3	5.9	5.8
His73	6.3	6.7	6.8
His99	6.3	6.4	6.4
His108	6.3	5.3	6.8
Lys114	10.4	11.2	(11.5)e
Tyr115	9.6	13.9	(9.9)
His119	6.3	5.3	(5.3)
His121	6.3	3.9	(6.0)
Arg122	12.0	13.0	(12.4)
Glu126	4.4	4.6	(4.4)
Asp129	4.0	3.3	(3.0)
Glu130	4.4	5.4	(4.7)
Glu131	4.4	3.5	(4.1)
His132	6.3	1.7	7.1
His137	6.3	4.7	6.3
Asp141	4.0	3.8	2.1
Asp144	4.0	2.7	3.6
Asp160	4.0	4.8	1.8
Asp163	4.0	1.8	2.8
Asp164	4.0	2.8	3.8
Glu173	4.4	0.0	1.6
His174	6.3	5.7	6.1
His192	6.3	5.9	6.3

^a All ten histidines, Asp144, ionizable residues of the activation loop–helix, and residues with relative pK_a shifts of 1 unit or more within the two calculations are tabulated.

^b Model pK_a of free amino acid in solution.

^c High pH E70A structure (PDB code 3gar; Su et al. 1998).

^d Low pH E70A structure (PDB code 2gar; Su et al. 1998) was used with reconstructed missing coordinates in the activation loop (see Materials and Methods).

^e Values in parentheses correspond to residues of missing and reconstructed coordinates.

Significant multiple-residue pK_a variations between the high and low pH E70A GART structures occur mainly in three loops, one β-strand, and one α-helix. Specifically, pK_a variations are found in the activation loop–helix (Leu111–Glu131), the binding loop (Thr140–Gly146), strand β6 that connects the activation and binding loops, flexible loop segment Phe157–Ile165, and residue Glu173 in helix α6 (Table 1). Note that all three loops are more flexible than any other regions of GART (with the exception of the C-terminal residues Ala210–Glu212, which are missing coordinates in the crystal structures, and are not included in our analysis). The binding loop and the flexible loop Phe157–Ile165 are regions with low electron density and high temperature factors in the crystal structures (Su et al. 1998). Because of the lack of electron density in the crystal structure, the coordinates of the activation loop of the low pH E70A structure were reconstructed from the coordinates of the high pH E70A structure, which were subsequently unfolded using the unfolding method of Elcock (1999; see Methodology). The method of Elcock uses molecular mechanics to unfold the coordinates of a known folded structure. This method provides a native-like unfolded state with expanded size, which maintains an overall shape similar to the folded state, but with only a small number of residual residue–residue interactions, while being much more solvent-exposed. The unfolding method of Elcock has been previously applied in stability studies of six proteins that were well characterized experimentally, and provided a good theoretical and experimental agreement (Elcock 1999). The method has also been shown to be more successful than a fully extended model in the case of Barnase (see Figs. 2 ▶ and 11 in Elcock 1999, where propagation of errors dependent on variable steps of the unfolding method is also shown). Given the lack of electron density, the activation loop of GART could be more realistically described as a distribution of states. However, the fact that the calculated pK_a values of ionizable residues His121, Arg122, and Glu126 of the segment that undergoes the coil–helix transition are close (within 5%, 3%, and 0% respectively) to their model pK_a values (Table 1) indicates that the single state generated by the method of Elcock has provided a successful model of the missing coordinates in the unstructured solvent-exposed local environment in the low pH E70A GART structure.

Most notable in Table 1 is a large drop of the pK_a of His132 from 7.1 in the low pH structure to 1.7 in the high pH structure, which means that His132 is protonated at pH < 1.7 and neutral at high pH > 1.7. Also, His121 shows a very low pK_a value of 3.9 in the high pH structure, compared with pK_a of 6.0 in the low pH structure (Table 1), which means that His121 is neutral at pH > 3.9 and protonated at pH < 3.9. These observations indicate involvement of His121 and His132 in pH-dependent conformational processes. Further examination of the variations of pK_as of Table 1 indicates that residues His108, Tyr115, Arg122, Glu130, His137, Asp141, Asp144, Asp160, Asp163, Asp164, and Glu173 are also involved in pH-dependent conformational changes.

Figure 3 ▶ presents the differences of the calculated apparent and model pK_a values for each ionizable residue of the high and low pH E70A GART structures. Residues with significant differences (>2 pK_a units) are labeled in the figure. These differences reflect unusual electrostatic and solvation environments for the residues involved. The positive differences of basic residues indicate favorable Coulombic interactions with acidic residues. The positive differences of acidic residues can be caused either by desolvation or by Coulombic interactions with other acidic residues or both, all of which are energetically unfavorable. Likewise, the negative differences of basic residues can be caused either by desolvation or by unfavorable Coulombic interactions with other basic residues or both. Furthermore, the negative differences of acidic residues indicate favorable Coulombic interactions with basic residues.

Fig. 3.

Plot of the difference of the calculated apparent and experimental model pK_as for all ionizable residues of GART against residue number. Calculations were performed using the high and low pH E70A GART structures 3gar (filled squares, solid lines) and 2gar (filled diamonds, dotted lines), respectively. Residues with pK_a differences >2 units from their model pK_a values are labeled.

It is known that protein conformation changes with pH. In the case of E70A GART the main conformational change between low and high pH involves the activation loop–helix. The highest pH in which a coiled activation loop is present in a known GART structure is 6.8 (structure 1grc, Chen et al. 1992; structure 1cdd, Almassy et al. 1992). The lowest pH in which an activation loop–helix is present in a known GART structure is 7.5 (structure 1cde, Almassy et al. 1992; structure 3gar, Su et al. 1998). Given these data, the coil–helix transition occurs in the pH range 6.9–7.4. In our pK_a calculations we have used two E70A GART crystallographic structures at extreme pHs of 3.5 (structure 2gar) and 7.5 (structure 3gar), rather than generating direct titration state information at a continuous range of pHs. At low pH values, the electrostatic properties of GART are best described by the results of the calculations using the structure 2gar. Likewise at high pH values, the electrostatic properties of GART are best described by the calculations using the structure 3gar. A combination of both calculations best describes the electrostatic properties of GART at intermediate pHs. The comparative analysis of the results has been used as an aid in ideication of the residues that are involved in pH-dependent conformational changes with emphasis on the formation and stabilization of the activation helix.

We proceed in our analysis using the high pH E70A GART structure, focusing on His121, His132, His119, His108, His137, Asp144, Arg122, and Glu126. His121 is located in the activation loop–helix; His132 is found in the proximity of the activation loop at both pHs and in the proximity of the activation helix at high pH; His119 is located in the activation loop; His108 is located in the catalytic site; His137 is found in the catalytic site at high pH; Asp144 is found in the binding loop; and ionizable residues Arg122 and Glu126 are located in the activation helix. The topology of these residues and/or their pK_a values (Fig. 1 ▶; Table 1) indicate that they are involved in electrostatic interactions during the activation-helix formation. To elucidate the effects of pairwise electrostatic interactions involving residues of the activation loop–helix, we have constructed single theoretical GART mutants His108Ala, His119Ala, His121Ala, His132Ala, His137Ala, Arg122Ala, Glu126Ala, Asp144Ala, and double theoretical GART mutants His121Ala/His132Ala and Arg122Ala/Glu126Ala, using the high pH E70A GART structure. The availability of substrate-free crystallographic structures at both high and low pH (helix on and off, respectively) made the E70A GART structure the ideal choice for our calculations. The choice of the E70A GART structures decouples pH effects caused by the formation of the activation helix from pH effects caused by substrate binding, or a combination of both, thus reducing the degree of complication of the problem. This is possible because the formation of the activation helix takes place independent of the presence of substrates (Su et al. 1998). The possibility of involvement of the substrates in the stability of the activation helix is addressed elsewhere. The presence of the Glu70Ala mutation, which is far away from the catalytic site and the activation loop–helix, does not affect the generality of our results.

Table 2A summarizes the effects of theoretical alanine replacements of His119, His121, and His132 in the pK_a values of key GART residues. Important pK_a variations from the parent structure are bold and underlined. The electrostatic interdependence of the side chains of His121 and His132 is clearly shown in Table 2A. In the absence of His132, the pK_a of His121 is raised from 3.9 to the model pK_a value of 6.3; and in the absence of His121, the pK_a of His132 is raised from 1.7 to 3.5. This observation indicates that the low pK_a of His132 in the parent structure is owing to a combined effect of desolvation and unfavorable Coulombic interaction with His121. Likewise, the low pK_a of His121 in the parent structure appears to be only caused by interaction with His132 but not because of desolvation. In addition, the mild increase of the pK_a value of His121 from 3.9 in the parent structure to 4.6 in the His119Ala mutant (Table 2A) implicates His119 in the stability of His121, through some type of relay long range interaction (>8 Å). A similar effect of the His121Ala mutant on His119 is not observed, presumably because of the higher degree of desolvation of His119. Solvent accessibility calculation using the high pH structure shows ∼40%, ∼10%, and ∼1% exposure to solvent for His119, His121, and His132, respectively. An interesting interaction among Glu173–His121–His132 is also observed (Table 2A).

Table 2.

pKas of helix-forming and stabilizing ionizable residues, using the high pH E70A GART structure a

A. Activation-helix histidine theoretical mutants
Ionizable residueb	Model pKac	Parent structured	His119Ala	His121Ala	His132Ala	His121Ala/His132Ala
His108	6.3	5.3	5.4	5.5	5.2	5.7
His119	6.3	5.3		5.3	5.1	5.3
His121	6.3	3.9	4.6e		6.3
Arg122	12.0	13.0	13.0	13.0	13.0	13.0
Glu126	4.4	4.6	4.8	4.6	4.6	4.7
His132	6.3	1.7	1.8	3.5
His137	6.3	4.7	4.7	4.7	4.6	4.7
Asp144	4.0	2.7	2.8	2.7	2.7	2.8
Glu162	4.4	4.0	4.4	4.2	3.8	4.4
Glu173	4.4	0.0	0.1	0.7	0.4	1.2

B. Activation-helix nonhistidine theoretical mutants
Ionizable residueb	Model pKac	Parent structured	Arg122Ala	Glu126Ala	Arg122Ala/Glu126Ala
His108	6.3	5.3	5.3	5.3	5.3
His119	6.3	5.3	6.2e	5.1	5.9
His121	6.3	3.9	4.1	3.9	4.1
Arg122	12.0	13.0		12.7
Glu126	4.4	4.6	4.8
His132	6.3	1.7	1.8	1.7	1.8
His137	6.3	4.7	4.7	4.6	4.7
Asp144	4.0	2.7	2.7	2.7	2.7
Glu162	4.4	4.0	4.5	3.9	4.4

C. Catalytic-site histidine theoretical mutants
Ionizable residueb	Model pKac	Parent structured	His108Ala	His137Ala
a High pH E70A structure (PDB code 3gar; Su et al. 1998).
b Active-site residues His108 and His137; activation loop–helix residues His119, His121, Arg122, and Glu126, His132: binding loop residue Asp144; and residues with relative pKa shifts >0.4 pKa unit from the parent structure (3gar) are shown.
c Model pKa of free amino acid in solution.
d Parent structure pKas from Table 1.
e Important pKa variations from the parent structure are bold and underlined.
His108	6.3	5.3		6.0e
His119	6.3	5.3	5.3	5.2
His121	6.3	3.9	4.5	3.9
Arg122	12.0	13.0	13.1	13.1
Glu126	4.4	4.6	4.7	4.6
His132	6.3	1.7	2.3	1.9
His137	6.3	4.7	5.3
Asp144	4.0	2.7	2.9	3.1
Tyr177	9.6	13.9	13.8	13.3

Table 2B summarizes the effect of theoretical alanine replacements of the remaining two charged residues of the activation helix, Arg122 and Glu126, in the pK_a values of key GART residues. Important pK_a variations from the parent structure are bold and underlined. The electrostatic interdependence of the side chains of His119 with Arg122 is clearly shown in Table 2B. In the absence of Arg122 the pK_a of His119 is raised from 5.3 in the parent structure to 6.2. Also, in the Arg122Ala/Glu126Ala mutant, the pK_a of His119 is raised to 5.9. However, this effect is not clearly shown on Arg122 in the His119Ala mutant (Table 2A).

Table 3.

Side-chain flips of His, Asn, and Gln residues, performed by the global hydrogen-bond network optimization using WHAT IF (Vriend 1990)

Structure	Flips
3gara	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
3gar-His108Alab	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
3gar-His119Ala	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
3gar-His121Ala	His54, Asn106, Asn127, His137, Asn194, Gln206
3gar-Arg122Ala	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
3gar-Glu126Ala	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
3gar-His132Ala	His54, Asn106, Asn127, His137, Asn194, Gln206
3gar-His137Ala	His54, His121, Asn127, Asn194, Gln206
3gar-Asp144Ala	His54, Asn106, His121, Asn127, His137, Gln206
3gar-His121Ala/His132Ala	His54, Asn106, Asn127, His137, Asn194, Gln206
3gar-Arg122Ala/Glu126Ala	His54, Asn106, His121, Asn127, His137, Asn194, Gln206
2garc	Asn24, Asn36, His54, His174

^a High pH E70A structure (Su et al. 1998).

^b Theoretical mutants were constructed using WHAT IF (Vriend 1990; see Materials and Methods).

^c Low pH E70A structure was used with reconstructed missing coordinates in the activation loop (see Materials and Methods).

Table 2C summarizes the effect of theoretical alanine replacements of catalytic-site histidines His108 and His137. Important pK_a variations from the parent structure are bold and underlined. An electrostatic effect of His108 on both activation loop–helix histidines His121 and His132 is observed (Table 2C). Specifically, the His108Ala mutation produces an increase of the pK_a of His121 from 3.9 to 4.5, and of the pK_a of His132 from 1.7 to 2.3, despite the fact that there is no direct spatial contact between His108–His121 (distance > 8 Å), or His108–His132 (distance > 7 Å). The upshift of the pK_as of His121 and His132 can be attributed to reduction of the overall positive charge of the catalytic site or to a more desolvated environment, in the absence of His108. However, the reverse mutations of His121Ala or His132Ala do not produce similar shifts in the pK_a of His108 (Table 2A). The electrostatic interdependence of catalytic residues His108 and His137 is obvious in Table 2C, as shown in a study of the catalytic mechanism of GART (Morikis et al. 2001). In addition, a small effect of His137 on Tyr177 is observed (Table 2C).

Figure 4A ▶ presents the calculated theoretical pH titration curves for the five critical histidines, His108, His119, His121, His132, and His137, using the high pH E70A GART structure. Interestingly, the titration curve for His132 forms a double sigmoidal, with inflection points at the calculated pK_a of 1.7 and at the pH of ∼5. Also, the titration curve of His121 does not show as sharp a transition as the titration curve of exposed His119, which has a pK_a of 5.1–5.3 and a clean symmetric shape in all calculations presented in Figure 4 ▶. Similar double sigmoidal shapes with close inflection points are seen with His108 and His137, but with a milder effect (Fig. 4A ▶). Figure 4B ▶ shows the effect of a His121Ala theoretical mutation on the titration curve of His132, supporting the interacting His121–His132 dyad argument. The calculated pK_a of His132 has been raised (also seen in Table 2A), but the second inflection point at pH ∼ 5 still remains. Figure 4C ▶ shows the effect of a His132Ala theoretical mutation on the titration curve of His121, which shifts to a higher pK_a value but retains an unusual shape, probably indicating a double sigmoidal character. Figure 4D ▶ shows the long range effect of a His119Ala mutation on the pK_a of the titration curve of His121, which is up-shifted compared to the titration curve of His121 of the parent structure (Fig. 4A ▶), as reported in Table 2A. Comparison of the titration curves of interacting catalytic-site histidines His108 and His137 (Morikis et al. 2001) with the titration curve of His119 also reveals distorted shape but not as obvious as for His132 and His121. Also, the indirect interaction of His108–His121 and His108–His132, discussed above, is shown by the changes in shape and pK_a of the titration curves of His121 and His132, in the His108Ala mutant (Fig. 4E,F). Finally, the effect of His108 on the shape and pK_a of the titration curve of His137, and vice versa, is shown in Figure 4, E and F. ▶

Fig. 4.

Titration curves for the five significant histidines discussed in text, His108, His119, His121, His132, and His137. The plots have been generated using (A) the high pH E70A GART structure 3gar and (B) high pH E70A GART theoretical mutants His121Ala, (C) His132Ala, (D) His119Ala, (E) His108Ala, and (F) His137Ala.

For comparison, we have plotted the calculated titration curves of residues that have abnormal calculated pK_as, using the high pH structure (Fig. 5 ▶). Nonhistidine residues labeled in Figure 3 ▶ and residues from Table 2 that are not discussed above in the context of the activation-loop formation and stability, have been selected and plotted in Figure 5 ▶. The shapes of the calculated titration curves (Fig. 5 ▶) are typical and representative of most ionizable residues of GART. Only Arg90 and Glu162 deviate slightly from this rule (Fig. 5 ▶). It should be noted that we have generated the theoretical titration curves using a single high pH structure of E70A GART, in which the selected residues occupy a single conformation. These titrations have been useful in elucidating strong electrostatic coupling among ionizable sites of the given structure.

Fig. 5.

Quality of the titration curves for representative residues of the high pH E70A GART structure, which show abnormal calculated pK_as, with the exception of histidines. (A) Positive residues; (B) negative residues. The titration curve for Arg90 shows a distorted line shape, and the titration curve of Glu162 does not appear to have as sharp a pH transition as other residues in the plot, which have very similar sharp, single sigmoidal shapes.

The following discussion of our results addresses the following questions: (1) Is His121 an internal molecular switch for the formation of the activation helix? (2) What is the role of His132 or the His132–His121 interaction in the activation-helix formation or stability? (3) Is exposed His119 involved in the stability of the activation helix? (4) Is the side-chain interaction between the charged residues Arg122–Glu126 important for stabilization of the activation helix? (5) Is substrate binding influenced by the formation of the activation helix?

Discussion

The significance of His119 and His121 in the activation-helix formation

Although we understand that the electrostatic nature of the binding-loop closure is caused by the interaction of Asp144 with the substrate(s) (Morikis et al. 2001), the cause of the formation of the activation helix at high pH has not been obvious. In search for an electrostatic molecular switch that activates the pH-dependent coil–helix transition of the activation loop–helix, we looked at the ionizable residues in the region. Because the coil–helix transition occurs at pH ∼ 7, naturally we focused on histidine residues, which have model pK_as of 6.3. Possible candidates are His121, the second residue of the activation helix; His119, a residue preceding the N-terminal residue (Thr120) of the activation helix; and His132 in close proximity with His121.

Examination of a helical wheel for the activation-helix residues 120–127 (Fig. 6 ▶) reveals a potential amphipathic helix with a surface comprising Ile120–Ala124–His121–Leu125, which can be hydrophobic at high pH when a neutral His121 is present. Likewise, at low pH with the presence of a charged His121, the amphipathicity is disrupted and the activation helix can become unwound. The pK_a value of 3.9 of His121 in the high pH structure of E70A GART (Table 1) supports the presence of a neutral His121 at pH > 7, which is favorable to formation of the activation helix at high pH. The increase of the pK_a value of His121, from 3.9 in the high pH E70A GART to 6.0 in the low pH E70A GART (Table 1), shows the trend for a protonated His121 as the pH of the protein decreases.

Fig. 6.

Helical wheel representation of the activation helix, Thr120–Asn127, which undergoes a pH-dependent coil–helix transition. Residues of hydrophobic character are shown in green, positively and negatively charged residues are shown in blue and red, respectively, polar residues are shown in yellow, and His121 is shown in gray. We propose that His121 is the molecular switch for helix formation, with charge on favoring the coiled state and charge off favoring the helical state.

Other charged residues in the activation helix are Arg122 and Glu126, which have neighboring side chains in the helical wheel (Fig. 6 ▶) and of opposite charge, probably stabilizing each other (see below). Residues Arg122–Glu126–Gln123–Asn127 form the polar surface of the amphipathic activation helix (Fig. 6 ▶). Examination of the crystal structures indicates that there is no direct side-chain interaction of His121 with Arg122 or Glu126.

Another histidine in the activation loop, preceding the N terminus of the activation helix, is His119. Solvent accessibility calculations show that His119 is solvent-exposed, compared with His121, at both high and low pH. But our pK_a calculations indicate that there may be a slight interaction with His121 (Table 2A). Also, the side chain of His119 is in close proximity and could be capable of forming a hydrogen bond with the side chain Arg122, but only after altering the His119 χ₂ torsion angle. As mentioned in Results, the pK_a of His119 is very robust, with a value of 5.1–5.3 in all calculations described in Tables 1 and 2, with the only exceptions being the Arg122Ala and Arg122Ala/Glu126Ala theoretical mutants of E70A GART, where the pK_a jumps to 6.2 and 5.9, respectively. It appears that the lower than model pK_a value of His119 is caused by unfavorable electrostatic interactions with Arg122 and His121 (Tables 1 and 2).

A histidine at the N-capping position (preceding the N terminal) is not a favorable residue for an α-helix (Aurora and Rose 1998). Examination of the crystallographic structures of GART does not show the presence of a hydrogen bond of the His119 ring with the backbone of a helix residue at position 2 or 3, as is the case with N-capping residues. It is likely that His119 plays a role in the activation coil–helix transition, through side-chain interactions with residues 2 and 3 of the helix, His121 and Arg122, despite its exposure to solvent.

Perhaps the most important feature of our pK_a calculations is the electrostatic interaction between His121 and His132 (Table 2A; Fig. 4 ▶). The role of His132 in the formation and/or stability of the activation helix is intriguing, and will be discussed next.

The role of the His132–His121 interaction in activation helix stability

The dramatically low pK_a value of 1.7 of His132 in the high pH structure indicates that His132 is found in a buried environment when the activation helix is formed (Table 1); whereas the close to neutral pK_a value of His132 at low pH indicates solvent exposure (Table 1). This observation indicates that the formation and closure of the activation loop–helix in the high pH structures creates more desolvated environment for His132. In addition, the low pK_a of His132 indicates possible unfavorable Coulombic interactions with basic residues. Our pK_a calculations and examination of the crystal structures have revealed that the nearby basic residue is His121 (see above).

Evidence of the strong electrostatic interaction of the pair His121–His132 is the double sigmoidal character of their titration curves (Fig. 4 ▶). Such deviations from sigmoidal titration curves have been observed in the past in bacteriorhodopsin (Bashford and Gerwert 1992). In this case, shape complementarity of a pair of titration curves was present, which was attributed to proton sharing between two ionizable sites (Bashford and Gerwert 1992). Subsequent theoretical modeling has shown that the double sigmoidal character of titration curves can be due to strong electrostatic interaction of two charged residues in close proximity (Yang et al. 1993). In the case of the ternary complex of GART, with bound pseudocofactor and substrate, we have observed triple sigmoidal character titration curves of the strong three-way interaction of catalytic groups of His108–Asp144–GAR (NH₂), as we report elsewhere (Morikis et al. 2001).

Figure 7 ▶ combines the topology of histidines of the activation loop–helix His119 and His121 in relation to interacting His132, the catalytic-site histidines His108 and His137, and the binding-loop/catalytic-site residue Asp144. Four crystallographic structures have been used in Figure 7 ▶, the high pH ternary complex structure (Fig. 7 ▶, red; 1cde, Almassy et al. 1992), the high pH E70A structure (Fig. 7 ▶, blue; 3gar, Su et al. 1998), the low pH binary complex structure (Fig. 7 ▶, yellow; 1gar, Klein et al. 1995), and the low pH E70A structure (Fig. 7 ▶, green; 2gar, Su et al. 1998). A significant conformational change, caused by pH or the presence of the activation helix, involves His132, which is found in different orientation in the high pH structures (where the activation helix is formed), compared with the low pH structures (with the coiled activation loop). This conformational change involves mainly the side chain and, to a lesser extent, the backbone of His132.

Fig. 7.

Comparison of relative orientation of ionizable residues involved in catalysis (His108, His137, Asp144) and in activation loop–helix formation (His119, His121, His132). The high pH ternary complex (1cde, red), high pH E70A mutant (3gar, blue), low pH E70A mutant (2gar, green), and low pH binary complex (1gar, yellow) are used. His119 and His121 for the low pH structures (yellow and green) are not shown because of missing coordinates in the crystallographic structures.

Hydrogen-bond formation between His121 and His132 is possible at high pH (Fig. 7 ▶). Interestingly, the most stable side-chain orientation of His121, given by the global hydrogen-bonding network optimization algorithm of WHAT IF, is at a flipped χ₂ torsion angle by 180° in the high pH E70A structure. This is different from the orientation of His121 in the high pH ternary complex, in which a similar side-chain flip is not favored by the global hydrogen-bonding network optimization algorithm of WHAT IF (Morikis et al. 2001). The difference in the most stable side-chain orientation of His121, between the ternary complex and the high pH E70A structures of GART, may reflect local structural changes induced by the substrate–cofactor system, which indirectly involve the activation helix in catalysis, as it has been previously suggested ( Shim and Benkovic 1999;Morikis et al. 2001).

In view of our results, we propose that the molecular switch for the activation helix of GART is His121. In addition, His132 is important for stabilization or destabilization of the activation helix through strong electrostatic interaction with His121. The relative ring orientation and relative protonation states of His121 and His132 determine the activation helix state of stability. In an enzyme like GART, in which the unfolded and folded forms of the activation loop–helix are equally important for function, destabilization is as significant as stabilization to facilitate the coil–helix transition.

The role of the pair Arg122–Glu126 in stabilization of the activation helix

Further examination of the structure and sequence of the activation helix, Ile120–His121–Arg122–Gln123–Ala124–Leu125–Glu126–Asn127, reveals the possibility for a salt bridge between the side chains of Arg122 and Glu126. The formation of the Arg122–Glu126 salt bridge is strongly favored by the separation of Arg122 and Glu126 by three amino acids. Contribution of electrostatic effects on helix stabilization, such as helix dipole and ion pair interactions, together with intrinsic amino acid propensities, has been the subject of extensive experimental studies (e.g., Scholtz and Baldwin 1992). Highly helical peptide constructs of de novo design that consist of repeated five-turn residue units of the type (Glu–X–X–X–Arg)_n, where n is the number of units and X is amino acids with a high propensity for helix formation (preferably alanines), are thought to be stabilized by a salt bridge between the first residue (Glu or Asp) and the fifth residue (Arg or Lys), or vice versa (Marqusee and Baldwin 1987; Merutka and Stellwagen 1991; Merutka et al. 1993). The central core of the activation helix of GART (Arg122–Gln123–Ala124–Leu125–Glu126) highly resembles such a salt-bridge-forming, five-residue block, with end points a positively and a negatively charged amino acid. In the case of the high pH structure, a distance of 4.3 Å is observed for Arg122(N_ɛ)—Glu126(O_ɛ2), which is indicative of the presence of a salt bridge. However, there is no effect of the Arg122 mutation on Glu126 and vice versa (Table 2B), which makes questionable the argument for an Arg122–Glu126 salt-bridge stabilization of the activation helix. Also, the order of the five-residue block is not favorable for helix stabilization. Optimally, the negatively charged residue (Glu or Asp) is aligned with the N terminus of the helix, and the positively charged residue (Arg or Lys) is aligned with the C terminus, in antiparallel arrangement with the helix dipole (Marqusee and Baldwin 1987).

Effect of the activation helix on the important catalytic residues His108, His137, and Asp144

We have also examined the electrostatic effect of the activation helix on the important catalytic site residues His108, His137, and Asp144. Table 1 shows that in the high pH structure, His137 has a pK_a of 4.7, which is significantly lower than its pK_a in the low pH structure or the model pK_a of free histidine. However, the pK_a of His137 in the high pH E70A structure (3gar) does not show variation in any of our activation helix theoretical mutant GART constructs (Table 2A,B). These results indicate that the formation of the activation helix (and closure of the catalytic site) creates a desolvated environment for His137, which favors a His137—Asn106 hydrogen bond (Morikis et al. 2001), but there is no direct interaction of His137 with important ionizable residues involved in the activation loop–helix conformation and stability, that is, His119, His121, Arg122, Glu126, and His132. Solvent accessibility calculation using the high pH structure (3gar) shows 6% exposure to solvent for His137. As mentioned earlier, the small effect of the His108Ala and His137Ala theoretical mutants on the titration curve of His132 is probably attributable to the positive charge reduction and decongestion of the catalytic site.

According to Figure 7 ▶, there is no conformational variation for the side chain of His137 in the high pH structures (3gar, 1cde) and in substrate bound structures (1cde, 1gar), but the side chain of His137 is found in an orientation away from the active site in the low pH, substrate-free structure (2gar). A slight variation in the side chain of catalytically important residue His108 is observed (Fig. 7 ▶) and has been attributed to the organization of the active site to accommodate the substrates and interactions with the substrates (Morikis et al. 2001). The variation of the side chain of Asp144 in the four structures of Figure 5 ▶ is caused by the different orientation of the binding loop in the substrate structures versus substrate-free structures.

Table 1 shows a lower pK_a value for His108 in the high pH E70A structure compared with the low pH E70A structure, or the model pK_a value of free histidine. This indicates the introduction of some hydrophobicity on His108 from the formation and closure of the activation helix. However, the pK_a of His108 is not as low as that of His137. There is no variation in the pK_a value of important catalytic residue His108, with the exception of a 0.4 pH unit shift between the His121Ala/His132Ala double mutant and the parent E70A structure. This increase of the pK_a of His108 (Tables 1, 2A) could be caused by a lower overall positive charge in the vicinity of the catalytic site in the double histidine mutant, but not because of direct side-chain interactions.

There is no variation in the pK_a value of Asp144 in any of the mutants of Table 2A,B. This is because of the orientation of the binding loop away from the catalytic site and activation loop–helix in the high pH E70A substrate-free GART structure. However, a direct electrostatic interaction between His108 and Asp144 in substrate-bound high pH ternary complex structure 1cde is present and is attributed to side-chain proximity when the binding loop, and Asp144, is reoriented toward the active site because of substrate binding (Morikis et al. 2001).

It has been proposed by the crystallographic studies that the formation of the activation helix prepares a hydrophobic environment for the catalytic site (Su et al. 1998). It has also been proposed by Shim and Benkovic (1999) that there is coupling between the activation helix and catalysis, based on k_cat–pH and k_cat/K_m(GAR)–pH profiles of a ternary complex of GART with GAR and 10-formyl-5,8,-dideazafolate (fDDF) and His121Gln, His108Ala, and Asp144Ala mutants. Our pK_a studies show that the formation and closure of the activation helix, which provides a solvent-free environment for the catalytic site, is the electrostatic medium of this coupling. Additional pK_a calculations involving a crystallographic structure with the presence of substrates are presented elsewhere (Morikis et al. 2001).

Other residues with abnormal pK_as

The following side-chain hydrogen bonds account for the abnormal pK_as of the residues labeled in Figure 3 ▶ but not discussed in text: Asp68—Arg64 (high pH), Asp68—Arg90 (low pH), Asp76—Tyr100 (high pH), Asp76—His99 (high and low pH), Tyr115—Asp129 (high pH), and Asp160—Arg168 (low pH).

In addition to the Tyr115(OH)—Asp129(O_δ1) hydrogen bond, Tyr115 forms a hydrogen bond between its carbonyl group and the side chain of His108. Also, Tyr115 forms a hydrogen bond with the side chain of Asn127 in the high pH structure, which is favored at a flipped side-chain orientation by the global hydrogen-bonding network of WHAT IF. Tyr115 is located in the hydrophobic environment of the catalytic site, with <10% calculated solvent accessibility at high pH, where it stabilizes or is stabilized by catalytic residue His108 and by the C-terminal residue of the activation helix Asn127. The desolvated environment explains the high pK_a of Tyr115 (Table 1).

In addition, residues Tyr67, Asp68, Tyr100, and Tyr177 are found in hydrophobic environments, with <5% calculated solvent accessibility, at high and low pH. Also, Arg90 and Glu173 have <10% solvent accessibility, at both high and low pH structures, and Asp60 has <10% solvent accessibility at high pH (14% at high pH).

Finally, residues Asp144 and Asp160 (in 2gar) suffer from crystal packing interactions, which may contribute to their pK_as (Su et al. 1998).

Conclusions

A cascade of electrostatic interactions is involved in the coupled pH-dependent processes of the coil-to-helix transition of the activation loop and catalysis. We focused our analysis on the mechanism of the activation-helix formation and its impact on catalytic-site residues. Our pK_a calculations indicate that His121 provides the molecular switch for the formation of the activation α-helix, depending on its charge state; a charged (protonated) His121 favors a coiled conformation and a neutral His121 favors a helical conformation. The side chain of His132 acts as a stabilization or destabilization factor for the activation helix through a strong electrostatic interaction with His121. The relative orientation of the side chains of His121 and His132 and their relative protonation states are the media of this strong interaction. Exposed His119 is not a suitable helix-capping residue, but may be involved in the activation helix–coil transition through side-chain interactions with Arg122. A weak, long range interaction between His119 and His121 may also be present. In addition, the center of the activation helix consists of the helix promoting sequence Arg122–Gln123–Ala124–Leu125–Glu126; however, the alignment of the positive and negative charges of Arg122 and Glu126, respectively, relative to the N and C termini of the activation helix is not optimal and possibly destabilizes the activation helix. An Arg122–Glu126 interaction is indicated by examination of the crystallographic structures, but electrostatic interaction indicative of a salt bridge between Arg122 and Glu126 is not observed in the pK_a calculations. Finally, we have presented electrostatic evidence that the partial ordering and closure of the activation loop acts as a gate on the shallow active site to create a hydrophobic environment for catalytically important residues His108, Tyr115, and His137, and to prepare the active site for catalysis. Based on our theoretical mutation results, we suggest that experimental crystallographic data on His121Ala and His132Ala GART mutants at high pH are needed to show clearly if His121 alone is responsible for the activation-helix formation, and the significance of the His121–His132 interaction.

Materials and methods

The method of Antosiewicz et al. (1994) was used to calculate pK_a values. This method involves the energetics of transfer of a single free ionizable group from the solvent into a specific site in the protein. The transfer is influenced by two electrostatic effects, the ionizable group desolvation and its Coulombic interaction with charges of other groups inside the protein. The electrostatic interactions of the ionizable group are implemented in the linearized Poisson–Boltzmann equation, which contains a desolvation energy term and a Coulombic interaction term that involves the interactions of the group with all partial charges in the protein when all other ionizable groups are in their neutral state. The solution of the finite difference Poisson–Boltzmann equation per ionizable group in the protein and isolated in solution, in its charged and neutral form, is required to calculate electrostatic potentials, which are then converted to ionization free energies.

First, an intrinsic pK_a is calculated for each ionizable residue, which does not account for interactions with other ionized residues, given by

where pK_a(mod) is the model pK_a for the residue alone in solution, from experimental data (Nozaki and Tanford 1967); z is −1 or +1 for a negatively or positively charged residue, respectively; ΔG^P is the difference in the ionization free energy of the charged and neutral forms of the residue in the protein and ΔG^S is the difference in the ionization free energy of the charged and neutral forms of the residue isolated in solution. Then the interactions among all ionizable sites, in their ionized states, are calculated using the multiple-site titration model or clustering method of Gilson (1993) to produce an apparent pK_a for each ionizable site. Unless otherwise noted, the pK_as reported here are apparent pK_as.

The program UHBD (Madura et al. 1994, 1995) was used to perform Poisson–Boltzmann continuum electrostatic calculations using dielectric smoothing (Davis and McCammon 1990; Gilson et al. 1993) at the protein–solvent boundary, and finite-difference focusing methods (Gilson and Honig 1988a,b). The parameter set PARSE (Sitkoff et al. 1994), which contains charges and van der Waals radii of the 20 amino acids, was used for the calculations. The dielectric constants of the solvent and protein were set to 78.4 and 20.0, respectively. The relatively high dielectric constant used for the protein interior is juied in Antosiewicz et al. (1994, 1996), where it was found to produce better agreement with experimental results. All calculations were performed at a temperature of 298 K and ionic strength for the solvent corresponding to 100 mM. For dielectric boundary smoothing, an ion exclusion layer around the surface of the protein was defined using a probe with a 2.0-Å radius. Focusing grids of 2.5, 1.25, 0.5 and 0.25 Å were used.

Changes from the neutral to the charged state of ionizable residues were made by adding a +1 charge to the positively charged amino acids and a −1 charge to the negatively charged amino acids. Positive unit charges were added at the following atoms: N_ζ of Lys, C_ζ of Arg, and N_δ or N_ɛ of His (depending on the position of the initial hydrogen in the neutral form). Negative unit charges were added at the following atoms: C_γ of Asp and C_δ of Glu, O_η of Tyr, and S_γ of Cys. The experimental model pK_a values used were 12.0 for Arg, 10.4 for Lys, 9.6 for Tyr, 8.3 for Cys, 6.3 for His, 4.4 for Glu, 4.0 for Asp, and 7.5 for the N terminus. Because there are no coordinates for the C-terminus residue of GART in any of the crystal structures used, we did not include an ionizable C terminus in the calculations.

Calculations were performed with PDB files 3gar (high pH, E70A mutant; 1.9-Å resolution; Su et al. 1998) and several theoretical single and double mutants of 3gar, and 2gar (low pH, E70A mutant; 1.8-Å resolution; Su et al. 1998). Solvents (water or phosphate molecules) present in structures 3gar and 2gar were not included in the calculations to facilitate comparison with parallel pK_a calculations using the structure of PDB file 1cde (ternary complex of GART), which does not have solvent molecules (Morikis et al. 2001). Theoretical mutants were constructed using the program WHAT IF (version 99; Vriend 1990). The constructed theoretical mutants of the high pH E70A structure were His108Ala, His119Ala, His121Ala, Arg122Ala, Glu126Ala, His132Ala, His137Ala, Asp144Ala, His121Ala/His132Ala, and Arg122Ala/Glu126Ala.

The addition of hydrogen atoms in the crystallographic structures was accomplished using the program WHAT IF (version 99; Vriend 1990^; Nielsen et al. 1999). The program WHAT IF was also used to perform a global hydrogen-bonding network optimization (Hooft et al. 1996; Nielsen et al. 1999), which corrected for the side-chain orientation of nonoptimal histidines, asparagines, and glutamines and established the initial hydrogen position at the N_ɛ2 or N_δ1 atom of histidines (Nielsen et al. 1999). Side-chain flips of His, Asn, and Gln residues performed by WHAT IF are summarized in Table 3. In structure 3gar and all its theoretical mutants (where applicable), His54, His137, His174, and His192 had a hydrogen atom at position N_δ1 and the rest at position N_ɛ2. In structure 2gar, His54 and His137 had a hydrogen atom at N_δ1 and the rest at N_ɛ2. 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.). Nomenclature compatibility between WHAT IF and UHBD and preparation for UHBD runs was achieved by a series of home-made PERL scripts.

The missing coordinates (caused by lack of electron density, residues Leu111–Glu131) of the activation loop in the low pH crystal structure of 2gar were generated by superimposing the structure of 2gar with the high pH structure 3gar, followed by transferring the coordinates from 3gar. The superposition was achieved by fitting the coordinates of the backbone C_α atoms for the residues for which coordinates were present. The end points of the transferred coordinates were connected to existing coordinates of 2gar by local 2-residue minimization using the program CHARMM (Brooks et al. 1983). To locally unfold the activation loop of structure 2gar the method of Elcock (1999) was employed using the program CHARMM and the CHARMM22 force field (MacKerell et al. 1995). To randomize the activation loop structure, the position of the energy minimum for van der Waals interactions between nonbonded atoms was aicially increased up to 6 Å in steps of 1 Å, followed by 50 steps of steepest descent and 250 steps of conjugate gradient energy minimization (Elcock 1999).

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

Abbreviations

• GAR, glycinamide ribonucleotide

• GART, GAR transformulase

• PDB, Protein Data Bank