Binding of cations in Bacillus subtilis phosphoribosyldiphosphate synthetase and their role in catalysis

Abstract

The binding sites for the two cations essential for the catalytic function of 5-phospho-d-ribosyl α-1-diphosphate (PRPP) synthases have been identified from the structure of the Bacillus subtilis phosphoribosyldiphosphate synthetase (PRPPsase) with bound Cd²⁺. The structure determined from X-ray diffraction data to 2.8-Å resolution reveals the same hexameric arrangement of the subunits that was observed in the complexes of the enzyme with the activator sulfate and the allosteric inhibitor ADP. Two cation binding sites were localized in each of the two domains of the subunits that compose the hexamer; each domain of the subunit has an associated cation. In addition to the bound Cd²⁺, the Cd²⁺-PRPPsase structure contains a sulfate ion in the regulatory site, a sulfate ion at the ribose-5-phosphate binding site, and an AMP moiety at the ATP binding site. Comparison of the Cd²⁺-PRPPsase to the structures of the PRPPsase complexed with sulfate and mADP reveals the structural rearrangement induced by the binding of the free cation, which is essential for the initiation of the reaction. The comparison to the cPRPP complex of glutamine phosphoribosylpyrophosphate amidotransferase from Escherichia coli, a type I phosphoribosyltransferase, provided information about the binding of PRPP. This strongly indicates that the binding of both substrates must lead to a stabilized conformation of the loop region, which remains unresolved in the known PRPPsase complex structures.

Phosphoribosyldiphosphate synthetase (PRPPsase, EC 2.7.6.1) catalyzes the transfer of an intact diphosphate (PP) group from ATP to ribose-5-phosphate (R-5-P), which results in the formation of AMP and 5-phospho-d-ribosyl-α-1-diphosphate (PRPP; Khorana et al. 1958), as shown in Figure 1 ▶. PRPP is an essential precursor for purine and pyrimidine nucleotides, both in the de novo synthesis and in the salvage pathway as well as in the synthesis of pyridine nucleotide coenzymes (Jensen 1983). The key position of PRPP in the metabolism of the cell explains why the activity of PRPPsase is highly regulated. Besides competitive inhibition at the substrate binding sites, most PRPPsases are regulated in an allosteric manner, in which ADP generally acts as the most potent inhibitor (Switzer and Sogin 1973; Becker et al. 1975; Gibson et al. 1982; Arnvig et al. 1990). Inorganic phosphate (P_i) is an activator of the bacterial and mammal PRPPsases, but it has not been investigated if this activation involves an allosteric mechanism (Switzer 1969; Roth et al. 1974; Meyer and Becker 1977; Hove-Jensen et al. 1986; Arnvig et al. 1990).

Fig. 1.

The reaction catalyzed by phosphoribosyldiphosphate synthase (PRPPsase).

The MgATP complex is the true substrate for the PRPPsase, and in addition, the enzyme requires a free Mg²⁺ ion as an activator (Switzer 1969Switzer 1971; Fox and Kelly 1972; Roth et al. 1974^; Hove-Jensen et al. 1986^; Arnvig et al. 1990^; Willemöes et al. 1996). Other divalent cations can serve as substitutes for Mg²⁺ but result in lower activity (Switzer 1969; Roth et al. 1974; Gibson et al. 1982; Arnvig et al. 1990; Willemöes et al. 1996). Substitution with Cd²⁺ gave a decrease in V_app accompanied by a decrease in K_M for both substrates (Willemöes et al. 1996). In another study in which the free Mg²⁺ ion was treated as a pseudosubstrate in kinetic studies of the PRPPsase from Escherichia coli (Willemöes and Hove-Jensen 1997), it was shown that the free Mg²⁺ must bind before both substrates. It was also suggested that the inhibitory effect of ADP binding at the regulatory site of the native enzyme results from competition with the binding of the free Mg²⁺ ion. This is in contrast to the situation in which ADP binds to the complex of PRPPsase with Mg²⁺, P_i, MgATP, and R-5-P, in which the inhibition is noncompetitive. A recent steady-state kinetic study of the E. coli PRPPsase revealed two possible routes of substrate binding: a fast pathway in which P_i is bound before Mg²⁺ and a slow route in which Mg²⁺ is bound in the first step (Willemoës, et al. 2000). Kinetic studies of three E. coli mutants indicate that the aspartic residues Asp133, Asp223, and Asp224 (in Bacillus subtilis) are part of the cation site (Bower et al. 1989; Willemöes et al. 1996). Strong indications also exist that both the free Mg²⁺ ion and the ATP-bound Mg²⁺ bridge the binding of the ATP molecule to the enzyme. In this complex, one of the cations is bound by α-phosphate and the other by the β- and γ-phosphates of ATP (Li et al. 1978; Gibson and Switzer 1980).

Structures of PRPPsase from B. subtilis are known in complex with mADP and SO₄²⁻(Eriksen et al. 2000), ADP-PRPPsase, and SO₄²⁻-PRPPsase. Both contain the PRPPsase as a closely packed homohexamer, and many of the residues that are part of the contacts between the six subunits in the hexamer are highly conserved. The subunits fold into two domains of similar topology resembling the type I phosphoribosyltransferases (PRTase; Eriksen et al. 2000). In addition to their ability to bind PRPP, the PRTases are related to the PRPPsases by a short sequence of high identity called the PRPP-binding fingerprint motif (Hove-Jensen et al. 1986). In the ADP-PRPPsase structure, ADP is bound both at the regulatory site and the ATP substrate binding site. It has been reported that sulfate can mimic the activation of phosphate (Eriksen et al. 2000), and in the SO₄²⁻-PRPPsase structure, a sulfate ion is bound at the regulatory binding site for ADP. Another sulfate ion located at the binding site for the 5`-phosphate of R-5-P is referred to as the 5`-sulfate. Comparisons of these two structures show that the activator SO₄²⁻ and the inhibitor ADP compete for a common regulatory binding site. Neither of the two structures reveals the presence of Mg²⁺ ions, because the Mg²⁺ present during the crystallizations was either chelated by citrate ions in the case of the ADP-PRPPsase or precipitated as the very insoluble MgNH₄PO₄ before crystallization of the SO₄²⁻-PRPPsase.

Here we report the structure of the Cd²⁺ complex with the PRPPsase of the B. subtilis enzyme, Cd²⁺-PRPPsase. This structure represents a new ligation state of the enzyme, with two Cd²⁺ ions bound in the active site. Furthermore, this structure contains the 5`-sulfate ion, a sulfate ion in the regulatory site, and a nucleotide in the ATP binding site. The characteristics of the three known PRPPsase structures are summarized in Table 1. Unfortunately, none of the known structures are R-5-P bound. To delineate the R-5-P binding site, the Cd²⁺-PRPPsase structure is compared with the structure of a type I PRTase, the glutamine PRPP amidotransferase (GPATase) from E. coli complexed with cPRPP, a PRPP analog (Krahn et al. 1997). The position of the cPRPP molecule in the GPATase structure is used as a paradigm for the R-5-P binding, and a particularly good structural agreement between the two structures is found in the close proximity of the PRPP binding site.

Table 1.

Characteristics of the three PRPPsase structures

Structure	Cd2+-PRPPsase	SO42−-PRPPsase	ADP-PRPPsase
Residues
Subunit 1	8–105, 108–198, 208–315	8–105, 108–198, 208–316	8–198, 208–280, 285–315
Subunit 2	8–105, 108–198, 208–315	6–198, 208–315	8–198, 208–280, 285–315
Resolution	2.8Å	2.3Å	2.2Å
Active site
ATPa site	AMP (mATP/ADP)		AMP (ADP)
R-5-P site	SO42−	SO42−
Mg2+ site	Cd2+
Flexible loop
Subunit 1	Open	Open	Open
Subunit 2	Open	Closed	Open
Regulatory site	SO42−	SO42−	mADP
Allosteric state	I	I	I

^a The nucleotide content in the crystallization mixture is shown in parentheses.

Results and Discussion

The Cd²⁺-PRPPsase structure

The Cd²⁺-PRPPsase structure was determined using a crystal identical to the ones used for the determination of the SO₄²⁻-PRPPsase but was soaked in CdCl₂. Surprisingly, soaking with Cd²⁺ resulted in not only the binding of two Cd²⁺ per subunit but also binding of nucleotide at the active site, where the latter is accompanied by structural rearrangements of surrounding loop regions (Fig. 2 ▶). Refinements showed that the Cd²⁺ sites are not completely occupied (Table 2) and that their coordination sphere is incomplete. It is expected that water molecules complete the coordination sphere of the Cd²⁺, but these remain unresolved from the higher electron density around Cd²⁺ at 2.8-Å resolution. The arrangement of and interaction between subunits in the unit cell of the Cd²⁺-PRPPsase structure is, however, identical to what was found for other PRPPsase structures, in which 12 molecules in the unit cell form two closely packed hexamers, burying 30% of the monomer surface. In contrast, only weak interactions between the hexamers in the unit cell are observed. On the basis of these observations, it must be concluded, as it was for the earlier structures, that the hexamer is the physiological unit for PRPPsase at this state of ligation. In the hexamer, the six subunits align in a 32-point group symmetry, giving a propeller shape with the N-terminal domain closest to the center. They are related by a crystallographic threefold axis going through the center of the propeller, with noncrystallographic pseudo twofold axes in the plane of the propeller.

Fig. 2.

The quaternary arrangement and fold of the Cd²⁺-PRPPsase subunits. Three subunits corresponding to half of the hexamer are shown. Subunits B and D are related by the crystallographic threefold axis (perpendicular to the plane), and subunit A is related to B and D by in-plane no-crystallographic two symmetry. To illustrate the fold, subunit A is drawn in color. The Cd²⁺-sheet (gray) surrounded by four α-helices (blue). The core is flanked at one side by a two-stranded antiparallel sheet known as the flag region (green). The N-terminal domain, which harbors a 3₁₀ helix (red) and the flexible loop Pro95-Ile112 (yellow), is closely packed against the N-terminal domain of the neighboring subunits. The C-terminal domain contains the PP binding (Pro173-Gly177 in magenta) and the R-5-P binding (Asp223-Thr231 in cyan) loops. All ligands (two Cd²⁺, two SO₄²⁻, and an AMP moiety per subunit) are included (red). This three-subunit arrangement is the smallest fraction of the hexamers that is needed to complete the regulatory site, which is located in a cavity created by the N-terminal domains. It is occupied by a sulfate ion (red) positioned in front of the 3₁₀ helix (red) belonging to subunit A. The active site is found in the cleft between the two domains of one subunit. Focusing on the active site of subunit A (located in front), the C-terminal Cd²⁺ is bound between the PP binding loop (magenta) and the opposing R-5-P binding loop (cyan), which also binds the 5`-sulfate (red). The nucleotide (red) in the active site is bound by residues from two neighboring subunits, and residues from the N-terminal flag region of subunit D are involved in the binding of the AMP moiety (red) in the active site of subunit A (and vice versa). The N-terminal Cd²⁺ (red) is located between the nucleotide and the N-terminal domain of subunit A.

Table 2.

Refinement summary

Resolution range (Å)	30–2.8
Rwork (%)	20.0
RFree (%)	27.3
Non-hydrogen atoms
PRPPsase	4597
AMP	46
SO42−	20
Cd2+	4
Water	47
RMS deviation from ideality
Bond lengths (Å)	0.015
Bond angles (°)	1.948
Average B-factors (Å2)
Main chain	26.7
Side chain	28.0
SO42−	37.3
Cd2+	30.9
AMP	46.9
Water	20.6
RMS deviation in B-factors (Å2)
Main chain atoms	1.4
All atoms	1.9

The structural features of the subunits are illustrated in Figure 3 ▶. They are comprised of two domains similar in fold and almost related by a twofold symmetry. Each domain is composed of a five-stranded parallel β-sheet surrounded by two α-helices on each side and flanked at one edge by a flag region that consists of a small antiparallel sheet. Some loop regions are distinguished in the comparison of the two domains. The magenta PP binding loop from Pro173 to Gly177 and the cyan R-5-P binding loop from Asp223 to Thr231 got their names from the structural comparison to the PRTases, which reveals the corresponding loops to be involved in the binding of the PP and the R-5-P portion of PRPP, respectively. The long yellow loop from Pro95 to Ile112 is termed the flexible loop. In the SO₄²⁻-PRPPsase complex, the loop appeared in different conformations in the two subunits of the asymmetric unit, an open and closed conformation, respectively. In the Cd²⁺-PRPPsase structure, the open conformation is clearly favored over the closed conformation in both crystallographically distinct molecules.

Fig. 3.

Close up of the active site. The AMP moiety (gray) and the two Cd²⁺ (black) are shown with omit maps contoured at the 3σ level. The following residues are included: Arg101, Gln102, Asp103, and Arg104 of the flexible loop and His135 corner (upper left, yellow); Asp174, His175, Gly176, and Gly177 of the PP binding loop and Arg180 (lower left, magenta); Asp223, Asp224, Ile225, Ile226, Asp227, Thr228, Ala229, Gly230, and Thr231 (lower right, cyan) of the R-5-P binding loop and the 5`-sulfate (lower right, green); and Phe40, Ser41, Asp42, Gly43, and Glu44 (upper right, orange) of the N-terminal flag from a neighboring subunit. All nitrogen and oxygen atom are drawn in blue and red, respectively.

Only small differences between the three structures are noticed, and it is an effect of the lower resolution that the Cd²⁺-PRPPsase structure contains a smaller number of water molecules than the other two structures. The structure of Cd²⁺-PRPPsase appears more related to the SO₄²⁻-PRPPsase structure as judged by overall superposition of the dimer in the asymmetric unit from the three structures. The root mean square is 0.3 Å and 0.4 Å, including 583 and 573 C_α atoms closer than 1 Å, respectively, from the comparisons to the SO₄²⁻-PRPPsase and the ADP-PRPPsase structures. Assuming that the largest deviations between the structures reflect the more flexible parts of the structure, we find highest flexibility in the N-terminal flag region and in the flexible loop, which can be rationalized in terms of differences in ligand binding.

The active site of the Cd²⁺-PRPPsase structure

Figure 2 ▶ illustrates how the active and regulatory sites are shared between the subunits. It has been shown that the activator P_i and the inhibitor ADP compete for the same binding site (Eriksen et al. 2000). In the Cd²⁺-PRPPsase structure, a sulfate ion close to the center in Figure 2 ▶ is bound at the regulatory site. It displays the same interactions with residues from three subunits as the sulfate ion at the same position in the SO₄²⁻-PRPPsase complex (Eriksen et al. 2000). The second sulfate ion is also found in the same position as the 5`-phosphate group of R-5-P as in the SO₄²⁻-PRPPsase structure.

The Cd²⁺ binding sites found in the active site localized between the two domains of one subunit (Figs. 2, 3 ▶ ▶). Each Cd²⁺ is exclusively associated with one domain. The C-terminal Cd²⁺ ion, which is seen closest to the bottom in Figure 3 ▶, connects Asp174 of the PP binding loop to Asp223 of the R-5-P binding loop. Both aspartic acid side-chains function as monodentate ligands, and they are among the completely conserved residues of the PRPPsases. The distances to the carboxylate oxygen atoms are rather long, which could be the result of fitting the much larger Cd²⁺ ion at a site that has evolved to bind Mg²⁺. The second Cd²⁺ bound by the N-terminal domain links NE2 of His135 to the O1P of α-phosphate of the bound nucleotide. The metal ligand bonds are shorter (2 Å) at this site. His135 is one of the few completely conserved residues in the PRPP synthetases, and it has been proposed as an active site residue by both mutational studies and chemical modification studies (Harlow et al. 1990; Busch 1996).

Only the AMP part of a mATP or an ADP molecule could be found at the ATP binding site. This was also the situation in the ADP-PRPPsase structure, and the AMP moiety is bound in a very similar manner in the two structures. The adenine is stacked in a pocket between Arg101 from the flexible loop and the Phe40 in the N-terminal flag region from adjacent subunit, as seen in Figure 3 ▶. N6 of the adenine donates protons to the carboxylate group of Asp42 and Glu44. Some minor differences are observed in the binding of the AMP molecules in the two active sites of the asymmetric unit. As in the ADP-PRPPsase structure, only the adenine part of the nucleotide is firmly bound by the enzyme. Two salt bridges, present in all PRPPsase structures from His135 to Asp224 and from Arg104 to Asp227, link the positive residues of the ATP binding site to the negative residues of the R-5-P binding site.

Changes in the active site with the Cd²⁺ binding

Comparison of the two structures containing the activating sulfate ions, that is, the Cd²⁺-PRPPsase and the SO₄²⁻-PRPPsase structures, provides direct information on the structural changes induced by the Cd²⁺ binding. The binding of ATP and the different conformations of the flexible loop represent the most dramatic differences. In SO₄²⁻-PRPPsase, the ATP binding site was unoccupied, and two distinct conformations of the flexible loop were recognized, stabilized by small differences in the crystal packing. In the open conformation, the binding pocket for the adenine of ATP is left open, allowing association with ATP, whereas the side-chain of Arg104 occupies the adenine binding pocket in the closed conformation. In the Cd²⁺-PRPPsase structure, the presence of divalent cations stabilizes the open conformation in both subunits. The two conformational states of the flexible loop might offer a structural explanation to the cation activation, which must initiate the binding of substrates. But it cannot be concluded from the structure whether it is the binding of a free ion or the Cd²⁺-ATP complex that causes the stabilization of the open conformation. In other words, is it the stabilization of the open conformation that increases the affinity for ATP or is it the binding of Cd²⁺-ATP complex that stabilizes the open conformation? The ADP-PRPPsase structure shows that it is possible for a nucleotide to bind to the ATP binding site in the absence of cations, but here the open conformation is clearly stabilized by the presence of ADP in the regulatory site.

There is hardly any change in the position of the two aspartic acid residues that bind the cation to the C-terminal domain. The distance between the carboxylate groups is, however, decreased slightly (0.5 Å) on the binding of Cd²⁺. With the C-terminal cation binding site empty, the negative charges of the two aspartic acid residues Asp174 and Asp223 are partly neutralized by a hydrogen bonding network to Arg180 through a single water molecule. When this site is occupied by Cd²⁺, the arginine side-chain has moved away and is no longer in contact with Asp174 or Asp223, and hydrogen bonds link Asp133 to Arg180. Arg180 is a highly conserved residue positioned in the helix preceding the C-terminal flag, where it can influence the conformation of the flag region. A superposition of the two domains reveals that the position of the flag regions relative to the core is related to the orientation of the preceding helix.

The PRPP binding site

PRTases bind PRPP and catalyze the transfer of the R-5-P moiety to an acceptor substrate. They are divided into the two subclasses, type I and type II. The type I enzymes share a common fold for the domain that harbors the PRPP-binding fingerprint motif (Smith 1995), which is similar to the fold found in the two domains of PRPPsase. The core region of the typical PRTase fold comprises four parallel β-strands surrounded by two-α helices on each side, when it is compared to the fold of the PRPPsase C-terminal domain, so the PRPP-binding fingerprint motifs are aligned (see Fig. 5 ▶). One PRTase, uracil-PRTase (UPRTase), contains the fifth β-strand of the central core, which is also found in PRPPsase (data not shown). The residues of the PRPP-binding fingerprint motif form a very characteristic loop (R-5-P binding loop) easily recognized in both the PRPPsase and PRTase structures. Figure 4 ▶ shows a superposition of the C-terminal domain of the Cd²⁺ complex structure superimposed on the C-terminal and domain containing the R-5-P loop in the GPATase structure (Krahn et al. 1997). Although the match is close to perfect for the backbone of the R-5-P binding loop, it is not so good for the neighboring PP binding loop. It is noteworthy that the PRPPsase does not contain the cis peptide in the PP binding loop conserved in the PRTases. This difference in backbone conformation could be of functional importance, and a factor that discriminates between PRPP as a substrate and as a product. The PRTases do not contain a negative residue at the position of Asp174, which indicates that this negatively charged residue serves to repel the negative PRPP ion from the PRPPsase. The cPRPP molecule and the Mn²⁺ from the GPATase structure are included in the close up of the active site shown in Figure 5 ▶, using the same superposition of the PRPPsase and GPATase as in Figure 4 ▶. The Mn²⁺ ion is found within 1 Å of the C-terminal Cd²⁺ in the Cd²⁺-PRPPsase structure, which further justifies the structural analogy because the superposition did not include these atoms. A similar observation is made for the 5`-sulfate ion and the 5`-phosphate of cPRPP. Two water molecules as well as the O1`, O2`, and O3` of the ribose ring and an oxygen on the β-phosphate of cPRPP constitute the entire coordination sphere of the Mn<sup>2+</sup> ion, which forms no bonds to the GPATase. This indicates that the C-terminal cation in the PRPPsase structure is part of the R-5-P binding site and that the cation at some stage during the reaction makes contact to the PP group, leaving the PRPPsase in a complex with PRPP. The side-chains of the first two aspartates in the PRPP-binding fingerprint motif corresponding to Asp223 and Asp224 in the PRPPsase are both part of the R-5-P binding site in the GPATase, where the analogs of Asp223 and Asp224 accept the hydrogens of the O3` and O2`, respectively. The largest deviation in the R-5-P binding loop of the two structures is found at the position of Asp224, where the displacement of this carboxylate group toward His135 in the PRPPsase influences both the side-chain and backbone of Asp224. From the comparison of the two structures, it is not obvious that Asp224 should be able to form hydrogen bonds to His135 and R-5-P at the same time.

Fig. 5.

Superposition of the Mn²⁺cPRPP complex from the GPATase structure on the active site of the Cd²⁺-PRPPsase. The active site is shown in the orientation that was used in Fig. 3 ▶, and all residues and ligands from Fig. 3 ▶ are included. In addition, the Mn²⁺cPRPP complex (black) originating from the GPATase structure (Krahn et al. 1997) is included, using superimposition shown in Fig. 4 ▶.

Fig. 4.

The superposition of the C-terminal domain (151–293) of the Cd²⁺-PRPPsase complex structure and the PRTase domain (272–438) from the GPATase structure in complex with cPRPP (Krahn et al. 1997): the PRPPsase (red) and the GPATase (blue). The superposition is made to give the best fit of the R-5-P binding loops in the two structures.

The above structural predictions of PRPP binding in the PRPPsase are agreement with previous enzymological results. Kinetic studies of the site-directed E. coli mutants corresponding to Asp223Glu, and Asp224Ala in B. subtilis, had indicated that these residues are involved in the binding of R-5-P in the PRPPsase (Willemöes et al. 1996). It is obvious from the Cd²⁺ complex structure that these mutations, which reduce the negative charge, will have a drastic impact on the electrostatic potential of the C-terminal cation binding site affecting cation binding. The kinetic studies also indicated the Asp corresponding to Asp133 in B. subtilis is involved in cation binding. This is not supported by the Cd²⁺-PRPPsase structure as the Asp133 carboxylate group is located >10 Å from both the N-terminal and the C-terminal cation binding sites and 15 Å from the α-phosphate of the nearest nucleotide. Again, however, the substitution has an indirect influence on the electrostatic potential at the C-terminal cation binding site through the interactions with Arg180 discussed earlier.

Implications for the catalytic mechanism

The differences noticed in the structures of the SO₄²⁻- and the Cd²⁺-PRPPsases support the idea that binding of a free cation can initiate the catalysis of the B. subtilis PRPPsase, similar to how it was found in the kinetic studies of PRPPsase from E. coli (Willemoës et al 2000). The structural results do not, however, identify to which cation site the initiating Mg²⁺ binds. Nevertheless, it is relevant to describe the different characteristics of the two binding sites. At the C-terminal site, Cd²⁺ is bound exclusively by the enzyme by the excessive negative charge from three neighboring aspartic acid side-chains, whereas the negative charge at the N-terminal site partly arises from the presence of the nucleotide. Based on kinetic studies of the substitution inert Co(III)tetraamine-β,γ-phosphate-ATP, which serves as a substrate in the presence of excess cations (Li et al. 1978), and measurements of K_act for the activation of Mg²⁺ for the E. coli mutants corresponding to Asp223Phe and Asp224Ala in B. subtilis (Willemöes et al. 1996), it was suggested that binding of the initiating cation occurs at the N-terminal binding site. From the analysis of the structure, it appears plausible that the initial rearrangement, which must be induced by the binding of the free cation, is the stabilization of the flexible loop in the open conformation, so that the substrate ATP can bind, but it is not obvious how binding of a cation can stabilize the open conformation of the flexible loop. His135 is a ligand to the N-terminal cation and is in very close proximity to the C-terminal binding site through the hydrogen bond with Asp224 and at the same time within hydrogen bond distance of Asp103 in the open conformation of the flexible loop. This indicates that His135 plays a key role in signal transduction from the cation binding site to the flexible loop.

It was earlier proposed that PP group is transferred in a SN₂ reaction, which proceeds via a direct nucleophilic attack of O1` of R-5-P on β-phosphorous atom of ATP (Miller et al. 1975). The very short distance from β-phosphorous atom of the cPRPP molecule to the methylene group of the mAMP supports this idea. In this case, the O1` must be activated by the removal of its proton. Based on the structure, it is not certain which of the groups effects this activation, because the active site is incomplete lacking residues from the C-terminal flag (Eriksen et al. 2000). From the structural results presented here, it is obvious that the C-terminal cation plays an important part in this activation, because there are good reasons to believe that it actually binds to the O1` group of R-5-P.

Materials and methods

Crystallization

The crystal was grown as reported earlier for the quaternary complex (Bentsen et al. 1996). After formation, it was transferred to a phosphate free drop, which was very similar in composition to the equilibrated growing conditions. This drop contained 1.8 M (NH₄)₂SO₄, 0.1 M Tris (pH 7.5), 2% PEG 400, 0.5% BOG, 1.8 mM mATP, 2.0 mM ADP, and 4.4 mM R-5-P. MgCl₂ was not included in the new conditions, but 5 mM of CdCl₂ was added instead. The crystal was left to soak in the Cd²⁺-containing drop for 1 mo.

Data collection and processing

The crystal was mounted in a glass capillary, and X-ray diffraction data were collected with an R-AXIS II imaging plate detector system mounted on a Rigaku Rotaflex RU 200 rotation copper anode operating at 50 kV and 180 mA using a graphite monochromator and a 0.5-mm collimator. The crystal to detector distance was 100.0 mm, and 2θ = 0. A total of 50 frames, each covering a 2°-oscillation range, were recorded with 20-min exposure time. The crystal was cooled to 1°C during data collection. The HKL-package was used for autoindexing and refinement of cell parameters and settings to obtain integrated intensities (Gewirth 1994). Data reduction was performed, and structure factors were derived with the programs ROTOVATA, AGROVATA, and TRUNCATE from the CCP4 suite (Collaborative Computational Project 1994). A total of 85,189 reflections were averaged to give 19,806 unique reflections in a 98%-complete data set, which included data from 30.0 to 2.8 Å. The R_merge was 13.6%. In the last shell from 3.0 to 2.8Å, the data was 96.2% complete, with I/σ(I) = 1.9 and R_merge = 40.0%.

The Cd²⁺ complex of PRPPsase crystallizes in the same space group as the other complexes of PRPPsase (Eriksen et al. 2000), namely P6₃, and the unit cell dimensions a = 115.6 Å and c = 107.7 Å are almost identical to the parameters obtained for the other investigated crystals. The asymmetric unit contains two molecules related by a pseudo twofold axis.

Model building and refinement

The structure of the original crystals before the soaking in CdCl₂ was solved using multiple isomorphous replacement with four mercury derivatives (Eriksen et al. 2000). Initial phases for the Cd²⁺ crystals were determined in a rigid-body refinement, using only the structurally well defined core of the PRPPsase dimer from the original crystals as a starting model. The search model included residues 8–101, 109–198, and 208–315 of both subunits in the asymmetric unit. The resulting F_obs-F_calc map at a 3.5σ contour level clearly showed the presence of two Cd²⁺ and two SO₄²⁻ ions per PRPPsase subunit. The residues Gln102-Lys105 and Ser108 of both subunits were easily traceable from the difference map. The resulting model was refined using simulated annealing, including data from 8.0 to 2.8Å. From the difference density, it was obvious that none of the four Cd²⁺ sites were fully occupied. Therefore, the occupancy of the different sites was individually adjusted in several refinement steps. Only an overall B factor was refined in these steps. No signal >2σ or <−2σ level was observed in the difference density at the Cd²⁺ ion binding sites, when the adjusted occupancies were used in the phase calculation. In all the following refinement steps, the occupancy of Cd²⁺ sites was kept constant, whereas the individual B factors were refined though strongly restrained. Bulk solvent correction was used in the later steps, including the data from 30.0 to 2.8Å. Residual density in both ATP binding sites of the asymmetric unit, which matched the mAMP part of ATP, became clear at a 3σ level, as a consequence of the bulk solvent correction. The R_free decreased by 0.5% when two mAMP molecules were included in the dimeric model.

The two molecules in the asymmetric unit were refined using partial noncrystallographic symmetry restraints. The regions that appeared to be influenced by differences in their crystalline environment were kept unrestrained, because the restraints resulted in less well defined electron density. These were Ser6-Leu9, Lys22-Ser52, Ala100-Arg109, His149-Met151, Tyr159-Asp167, Arg187-Lys189, and Glu300-Ser315. Loosing of the restraints did not lead to gross differences between the two molecules in the asymmetric unit of the Cd²⁺-PRPPsase structure, as superposition of the two structures, including 286 C_α atoms with a maximal distance of 1 Å, has a root mean square of 0.3 Å.

The refinement was performed with X-PLOR (Brünger 1992). The final model accounted for the residues Asn8-Lys105, Ser108-Arg198, and Met208-Phe315 of both subunits; 47 water molecules; 4 SO₄²⁻ ions; 4 Cd²⁺ ions; and 2 mAMP molecules. The final refinement statistics are listed in Table 2.

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC section 1734 solely to indicate this fact.

Abbreviations

• PRPPsase, phosphoribosyldiphosphate synthetase

• PP, diphosphate

• R-5-P, ribose 5-phosphate

• PRPP, 5-phospho-d-ribosyl α-1-diphosphate

• PRTase, phosphoribosyltransferase

• GPATase, glutamine PRPP amidotransferase

• mATP, α,β-methylene ATP

• mADP, α,β-methylene ADP

• mAMP, α-methylene AMP

• cPRPP, cyclopentyl analog of 5-phosphoribosyl-1-diphosphate