NMR Studies of the Nucleotide Conformation and the Arrangement of Substrates and Activators on Phosphoribosylpyrophosphate Synthetase*

The paramagnetic effects of Mn2+, an essential acti- vator bound at the active site of phosphoribosylpyrophosphate synthetase, on the longitudinal relaxation rates of the phosphorus atoms of the bound substrates tetraaminecobalt(II1)-P,y-phosphate-ATP and ribose 5'- phosphate, the substrate analog tetraammineco-balt(III)-/3,y-phosphate-a,~-methylene-ATP, the prod- ucts AMP and adenosine 5'-0-(thiophosphate) and the activator Pi have been used to calculate Mn2'-phospho- rus distances. These data, together with the measured distances from Mn2+ to the protons of the substrate tetraamminecobalt(II1)-P,y-phosphate-ATP were used to determine the preferred conformation of the enzyme-bound nucleotide. The metal-phosphorus distances (4.6 to 6.5 A) indicate second sphere complexes of the nucleoside triphosphate substrate and the AMP and aden- osine 5'-O-(l-thiotriphosphate) products. The conformation of the enzyme-bound nucleotide is somewhat folded with a torsional angle at the glycosidic bond (x = 62 f 5") which differs by at least 20" from that found free in solution. The enzyme-nucleotide interaction that causes such a conformational change, as found with other ATP-utilizing enzymes, correlates with the high base specificity of these enzymes.

The paramagnetic effects of Mn2+, an essential activator bound at the active site of phosphoribosylpyrophosphate synthetase, on the longitudinal relaxation rates of the phosphorus atoms of the bound substrates tetraaminecobalt(II1)-P,y-phosphate-ATP and ribose 5'phosphate, the substrate analog tetraamminecobalt(III)-/3,y-phosphate-a,~-methylene-ATP, the products AMP and adenosine 5'-0-(thiophosphate) and the activator Pi have been used to calculate Mn2'-phosphorus distances. These data, together with the measured distances from Mn2+ to the protons of the substrate tetraamminecobalt(II1)-P,y-phosphate-ATP were used to determine the preferred conformation of the enzymebound nucleotide. The metal-phosphorus distances (4.6 to 6.5 A) indicate second sphere complexes of the nucleoside triphosphate substrate and the AMP and adenosine 5'-O-(l-thiotriphosphate) products. The conformation of the enzyme-bound nucleotide is somewhat folded with a torsional angle at the glycosidic bond (x = 62 f 5") which differs by at least 20" from that found free in solution. The enzyme-nucleotide interaction that causes such a conformational change, as found with other ATP-utilizing enzymes, correlates with the high base specificity of these enzymes. Upon binding of ribose 5"phosphate to phosphoribosylpyrophosphate synthetase, no movement of the adenosine portion of the bound nucleotide occurs, as indicated by the absence of changes in the Mnz+-proton distances, but a significant conformational change occurs in the olyphosphate chain as manifested in 1.2 A and 0.6 1 increases in the distances from Mn2+ to the P,, and P,< atoms, respectively. The exchange rates of the nucleoside triphosphate substrates (lo4 s-') and the nucleotide products (10' s-') from their respective enzyme*MnZ+ complexes indicate that they are kinetically competent to participate in catalysis since kcat -22 s". The essential activator Pi is found to bind not only at the active site 6.3 A from the bound Mn2+, but also at the 5-P site of ribose 5"phosphate and at the P, site of tetraamine-PP-Rib-P' synthetase, which catalyzes a pyrophosphoryl transfer from ATP to the a-1-OH group of ribose-5-P, is a member of that class of ATP-utilizing enzymes which requires two divalent cations per active site (1-3). Previous studies with PP-Rib-P synthetase have shown that the A isomer of the substitution-inert P,y-bidentate CO(NH;$)~ATP is a slow substrate for the enzyme, with a maximal velocity 5% of that found with MgATP. Activity was detected only in the presence of an added divalent cation, which was found to bind directly to the enzyme (3). Hence, one of the two metal ions required for activity is on the nucleotide substrate, while the other is located elsewhere on the enzyme. Kinetic studies with ATPaS using Mg2+ or Cd" as activators (4) have suggested that one of the two metal activators, possibly the enzymebound metal, interacts directly with t,he a-phosphoryl group of the nucleotide sometime during catalysis.
The availability of substitution-inert complexes of ATP with trivalent cations allows one to substitute a paramagnetic cation probe selectively a t either the nuucleotide-metal site or the enzyme-metal site and to determine distances from the probe to various atoms of the substrate molecules by NMR relaxation measurements. In a previous paper, studies of this kind were described that used PP-Rib-P synthetase and the a,P,y-tridentate Cr(II1) (H2O)BATP complex (5). In the present work, Mn" was used as the paramagnetic probe and binding of Mn2+ to ATP was prevented by use of P,y-CO(III)(NH:~)~ATP or its analogs. The results have been used to determine the spatial arrangement and conformation of enzyme-bound substrates and of enzyme-bound activators. NMR spectra at 145.8 MHz (pH 8.2) of ATP, Co(NHa)4ATP, AMPCPP, and Co(NH&AMPCPP are shown in Fig. 1. Chemical-shift and spin-coupling data are given in Table 1. Introduction of the methylene group between the a-and P-phosphates of ATP causes a marked change in the chemical shifts of P, and Po (ie. 30.0 and 28.2 ppm downfield shifts, respectively) and in their spin-coupling constants (i.e. Jab decreases by 11.2 Hz, while Jpu increases by 4.3 Hz). Upon formation of the P,y-bidentate CoB+. tetraammine complexes, P, and P, shift downfield by 8.5 to 18.8 ppm and show small changes in the coupling constants. The different absolute configuration about the P-phosphorus in the Co3+tetraammine complexes gives rise to chemical shift differences between the P,, Pii, and P, resonances of the two stereoisomers (A, A) which are, respectively, 20, 10, and t 2 Hz for C O ( N H~)~A T P , and 16, 14, and 6 Hz for Co(NH3)aAMPCPP.
Studies of Binary Mn2+. Ligand Complexes-As a control for the enzyme studies, the binary complexes of Mn'+ with AMP, AMPaS, Rib-5-P, and Pi were investigated. The disso-P . pr l i ciation constants of these complexes were determined by EPR and are given in Table 11. The effects of Mn2' on the 31P longitudinal and transverse relaxation rates of the phosphorus nuclei in the above complexes were measured at varying Mn2+ concentrations. Calculated bound state relaxation rates are given in Table 11. Since the affinity of PP-Rib-P synthetase for Mn'+ is high at saturating levels of Pi ( K I , = 1.9 p~, Ref. 3), under the present experimental conditions the enzyme, when present, bound a major fraction of the metal ion. The above dissociation constant and the parameters in Table I1 were used to correct the experimental "P relaxation rates in TABLE I 3'P chemical shifts a n d coupling constants of ATP, AMPCPP, a n d their Co3+. tetraamine complexes These shifts and constants were measured at 145.8 MHz, pH 8.2, 25°C. The uncertainty in the chemical shifts is 0.01 ppm and in the coupling constants is 0.3 Hz. Relative to 85% H3P04 external reference. Positive values are upfield.

TABLE I1
Dissociation constants of Mn2+ complexes a n d bound-state 31P relaxation rates a t 145.8 MHz The uncertainty in the dissociation constants is &30% and in the bound state relaxation rates +15%. pH 7.0, T = 15°C.

and Arrangement on
PP-Rib-P Synthetase mixtures of Mn", enzyme, and ligands for the small contribution due to the formation of the weaker binary Mn'+. ligand complexes. In the case of CO(NHB).+ATP and C O ( N H~)~A M P C P P , these contributions were found to be negligible. In the case of AMP, the correction was -lo%, while with AMP& and Rib-5-P the corrections were 130%. The maximal errors in the distances calculated from these relaxation rates (see below) due to uncertainties in the binding constants for Mn2+ are estimated to be less than 5%.
The l/fTl, values in Table I1 reflect structural differences in the binary Mn".nucleotide complexes. For example, we have previously shown (12) that formation of the weak Mn2+. CO(NH,)~ATP complex involves interaction between the Mn" and the adenine ring only. The 40-fold greater effect of Mn'+ on the 31P of AMP reflects direct phosphoryl coordination. The 2.7-fold smaller effect of Mn" on l/fTl, of AMPaS as compared to AMP may be due to an 18% greater Mn2+ to phosphorus distance in this complex due to the presence of the sulfur atom.' These complexes were not studied further.  Table V. Addition of 10.6 m~ Rib-5-P to the experimental solution resulted in insignificant changes in the proton relaxation rates of C O ( N H~)~A T P .
Although a M n -E .
Co(NH:d.+ATP -Rib-5-P complex constitutes a reactive system (3), under the experimental conditions less than 20% of Co(NH:I)4ATP was estimated to be consumed by the enzyme in the short time required for this measurement.
Determination of Correlation Times and Internuclear Distances-The longitudinal relaxation rates can be used to calculate distances from the enzyme-bound Mn'+ to the proton and phosphorus nuclei of the other enzyme-bound ligands using the relations (9): where TM is the mean lifetime of the ligand in the paramagnetic environment, WI and OS are the nuclear and electron Larmor the paramagnetic contribution to the transverse relaxation rate within a given complex sets a lower limit on the dissociation rate ( 1 /~~) of the ligand from that complex. Comparing the l/fT1, and l/fF6TZp values (Tables I11 and IV), it is apparent that the longitudinal relaxation rates in the complexes studied are not limited by chemical exchange and may therefore be used for distance calculations. However the large T1,,/TLp ratios for ."P (Tables I11 and IV) indicate significant contact contributions to 1/fT2, which can occur with phosphorus even in second sphere complexes (9), rendering such ratios inappropriate for calculating dipolar correlation times. The correlation times that govern the paramagnetic relaxation rates were determined (9) by measuring the field (frequency) dependence of the longitudinal "P relaxation rates of the enzyme.Mn'+ complexes of Co(NH:<).+ATP, AMP, and P, at 40.5 and 145.8 MHz.
These measurements yielded an average ratio TI,, (145.8 MHz)/T,,, (40.5 MHz) = 1.4 f 0.1, with no systematic variation between the complexes studied (data not shown) suggesting very similar T~ values for both of the complexes. The correlation times of the enzyrne.Mn"-C~(NH~)~ATP complex in the absence and presence of Rib-5-P were also measured by the frequency dependence of the 'H longitudinal relaxation rate of Co(NH3),ATP, yielding a ratio of TI, (250 MHz)/TI,, (100 MHz) = 2.2 f 0.2 for both complexes (data not shown). This finding indicates that the binding of Rib-5-P to the enzyme. Mn2+.Co(NH3).+ATP complex does not alter T~. From these ratios, taking into account the two extreme limiting cases that T~ is either independent of or maximally dependent on frequency, an average correlation time was calculated to be 0.6 f 0.3 ns at Ho = 23.5 kG, 1.5 f 0.7 ns at 58.7 kG, and 2.7 f 1.8 ns a t 84.6 k c . T h e above three complexes which were used for T= measurements represent the various types of complexes examined in these studies.
The errors in the distances (510%) take into account the errors in both T~ and in l/fT1,. Distances calculated with these correlation times (Tables I11 and V) are significantly greater than the value (3 A) expected for direct phosphoryl coordination by the enzyme-bound Mn", but are appropriate for second sphere complexes (4.5 to 6.0 A) (11).
Conformation of the Enzyme-bound Nucleotide-The distances between the enzyme-bound Mni+ and the phosphorus and proton nuclei of Co(NH&ATP (Tables I11 and V) were used together with a computer search program to determine the preferred conformation of the enzyme-bound nucleotide. The search procedure has been described elsewhere (9, 10). Two sets of solutions, each unique to 5 to 15" about the six rotatable bonds of Co(NH3)4ATP, were obtained, which yielded Mn"-nuclei distances within the experimental distances and their errors, and less than 0.35% van de Waals atomic overlap. Within their accuracies, the two solutions were indistinguishable except for the torsional angle at the glycosidic bond, which was anti (x = 62 f 5") in one solution and syn ( X = 272 f 5') in the second. Since an anti conformation (x = 15 to 44") is usually found in adenine nucleosides and nucleotides in solution (12) and also in the crystalline state (13), this conformation (Fig. 4) is considered more likely. The conformation of the bound nucleotide thus differs significantly from that found free in solution.
The binding of Rib-5-P to the Mn-E .Co(NH&ATP complex was found to have no significant effect on the distances from Mn2+ to the protons of CO(NHJ)~ATP (see above) suggesting little or no change in the conformation of the adenosine moiety of the nucleotide. However, significant increases were found in the distances between Mn'+ and P, and P,, of the enzyme-bound C O ( N H~)~A M P C P P upon the binding of Rib-5-P (Table 111)   triphosphate chain. In order to explore the nature of this conformational change in greater detail, the computer search procedure was used to create a minimally altered nucleotide conformation induced by the binding of Rib-5-P. In the analysis, two assumptions were made. (a) The conformation of the adenosine moiety is the same for enzyme-bound C O ( N H :~)~A T P and Co(NH:,)4AMPCPP both in the absence and presence of Rib-5-P. This assumption permitted the use of the Mn2+ to proton distances obtained with CO(NH:~)~ATP (Table V) together with the Mn'+ to phosphorus distances obtained with Co(NH:J4AMPCPP (Table 111). ( b ) Under the experimental conditions, the P, resonance of Co(NH.h-AMPCPP was obscured by the stronger Pi resonance and the P, to Mn'+ distance could not be measured. In order to obtain the minimal alteration of the triphosphate conformation induced by the binding of Rib-5-P, the ratio of distances from    Mn2+ to the Po and P, of the enzyme-bound nucleotide was assumed to be constant (1.08, Table 111). With this assumption, the distances from Mn" to P, of CO(NH.~)~AMPCPP were estimated to be 5.4 -+: 0.6 p\ and 6.0 -+: 0.6 A in the absence and presence of Rib-5-P, respectively. Fig. 5 shows the minimal conformation change of the enzyme-bound nucleotide substrate that could be induced by the binding of the second substrate, Rib-5-P. It should be pointed out that, in view of the assumptions made, the actual conformational change may well be greater.  (3), which in this case is of the same order of magnitude as the true dissociation constant (3-51, we estimate k,, values of 21 X 10' M" s-' for both of these complexes. In the case of AMP and AMP& larger limiting values of k,,ff are observed ( 2 2 X IO5 s"). However, from these and from their KI values (which were calculated from the KI""" (Table VIII) using the relation Ki"'*" = KI"p" (1 + [ATP]/frl,)", which assumes simple competition of these products with ATP), similar k,, values of 2 1 X IO' M" s-' and 20.6 X 10" M" s" are estimated for AMP and AMPaS, respectively. Hence, all nucleotide substrates and analogs appear to have similar k,, values. The nucleotide exchange rates measured by NMR are  as inhibitors using Mg2+ and Cd2+ as activators were previously reported (4). In order to relate such studies to the present work in which Mn2+ was used, these kinetic measurements were repeated, under similar conditions, with Mn2+ as the activator (Tables VI1 and VIII). With ATP as substrate, the rate observed with Mn2+ as activator is intermediate between those found with Mgz+ and Cd'+ (Table  VII). While Mgz+ activates only with the A-isomer of ATPaS and Cd2+ activates weakly with both isomers, Mn9+ also activates with both isomers but a t higher rates than Cd" (Table VII). The concentration of AMP which is required for 50% inhibition is essentially independent of the activating metal (Table VIII). However, AMP& shows a slightly higher affinity (factor of -2) in the presence of Mnz+ relative to Mg", and a much higher affinity when Cd"" is the activator metal (Table VIII).
Binding Sites of P,-The presence of Pi (220 mM) in PP-Rib-P synthetase solutions is essential both for activity and to prevent irreversible denaturation of the enzyme (1). The longitudinal relaxation rates (1/T;,) of Pi (Fig. 2 H ) show significant decreases when substrates with greater numbers of phosphoryl groups are added. Thus, l / T i p decreases when NTP replaces NMP or when Rib-5-P is also bound (Table  IV). While such effects could in principle result from changes in the affinity for P, or for Mn2' or from changes in the 7, values, these mechanisms are unlikely in the present case in view of the saturating level of P, and the high levels of enzyme used in all of the studies, and the directly measured T,. values which showed no significant changes. Hence, these observations led to the hypothesis that in addition to an activating site on the enzyme, Pi may also bind at two substrate sites, i.e.

Complex
No. of P, the 5-P site of Rib-5-P and the P, site of ATP, when these sites are unoccupied. Indeed, the relaxation data could be well fit with such a 3-site model for Pi (Table VI). Accordingly, P, is observed to bind to the enzyme at a site 6.3 A from the enzyme-bound Mn2+, when Rib-5-P and ATP are both present." Making the reasonable assumption that this site is always occupied, the l/T,!, data can be explained by two additional sites for Pi, one at 5.0 A when Rib-5-P is omitted and another at 5.6 A when ATP is replaced by AMP. Within experimental error, these distances agree with those found from Mn2+ to the Rib-5-P site (5.3 A) and to the P, site of ATP (5.3 A) (Table III), consistent with the overlap of these P, sites with the respective substrate sites.

DISCUSSION
The present results establish that the metal-ATP substrate, upon binding to PP-Rib-P synthetase, undergoes a small but significant conformational change in the adenine-ribose region, resulting in a t least a 20" increase in the torsional angle at the glycosidic bond to a higher anti conformation (x = 62 l?r 5") (Fig. 4). Another possibility, consistent with t.he measured distances, but far less likely on energetic grounds, is a Actually, there may be n such activating sites equidistant from the enzyme-bound Mn2+ with an average distance of 6.3 X n'l" A, or additional sites 2 8.2 A from the Mn2+, which would be undetected hy NMR under the experimental conditions. 230" change in x to a syn conformation. Conformational changes of 40 to 50" about the glycosidic bond have also been found by NMR to accompany the binding of nucleotide substrates to protein kinase (12), DNA polymerase (14), and RNA polymerase (15) enzymes which, like PP-Rib-P synthetase, show a high substrate base specificity. Pyruvate kinase, an enzyme with low nucleotide base specificity (16), produces no significant change in x when ATP is bound (17). Hence, those enzyme-nucleotide interactions that are strong enough to alter x may also be responsible for high nucleotide base specificities of enzymes (18).
Upon binding of Rib-5-P to PP-Rib-P synthetase, no further change occurs in the adenine-ribose portion as indicated by the absence of changes in the Mn'+-proton distances, but a significant conformational change occurs in the polyphosphate chain of the bound metal-ATP substrate as manifested by 1.2 A and 0.6 A increases in the distances from the bound Mn2+ to the P,, and Plr atoms, respectively (Table 111, Fig. 5). This conformational change occurs only when a triphosphate chain is present on the nucleotide, since no corresponding change in distance is observed with AMP& upon the binding of Rib-5-P ( Table 111). The conformational change may better align the P,~-O-P, portion of the bound ATP for nucleophilic attack with inversion at PI{ by the a-1-OH of Rib-5-P (3,5), A conformational change when both ATP and Ribd-P are bound is also suggested from the fact that AMPCPP binds appreciably more tightly to PP-Rib-P synthetase when Rib-5-P is p r e~e n t .~ T h e reaction mechanism of Fig. 6 has been drawn to be consistent with the conformation of ATP in the Mn. E . Co(NH3).,ATP. Rib-5-P complex, with previously determined distances from the nucleotide-bound metal to the C-1 protons and 5-P atoms of Rib-5-P (5), and with distances from both metals to the essential P, activator. Two additional constraints that have been incorporated into the mechanism (Fig. 6) are an appropriate (180") alignment between the entering and leaving groups for inversion about P, (3) and the minimal distance (3.8 A) previously measured between the entering oxygen and the attacked Pi< of the ATP (5). The leaving AMP is not coordinated to the enzyme-bound Mnr+ activator but rather forms a predominantly second sphere complex, as do all of the phosphoryl groups of the metal-ATP substrate and its analog ( Table 111). The -1 A smaller Mn"-P,, distance in the enzyme. Mn2+ -AMP complex, as compared with this distance in all of the other enzyme. Mns+nucleotide complexes studied, suggests either a more closely packed second sphere AMP complex, or the averaging of a small amount (15%) of an inner sphere AMP complex with a distance of 2.9 A from the Mn'+. Similar second sphere distances were found on pyruvate kinase from the essential enzyme-bound divalent cation activator to the phosphorus atoms of the metal-ATP (17) and P-enolpyruvate substrates (19). Direct coordination of a metal-ATP substrate by an additional enzyme-bound divalent cation (on protein kinase) results in marked inhibition of the enzyme-catalyzed reaction (12).
The finding that Cd'+, but not Mg", activates with the Bisomer of ATPaS, while both Cd2+ and Mg2+ activate with the A-isomer, were interpreted to indicate a coordination of the P,y-bidentate metal ATP substrate by the enzyme-bound divalent cation (4). This view was plausible since P,y-bidentate Co(NH3)ATP (A isomer) was found to be a substrate (3) while a,P,y-tridentate CrATP was at least 10-fold less active, indicating that a coordination by the nucleotide-bound metal was not essential but actually inhibitory (5). The present NMR studies, which fail to detect coordination at P, by the enzyme-bound Mn" with either the substrate or product forms of the nucleotide (including AMPaS), argue strongly against such an interaction with Mn'+ in any of the complexes examined, but can be rationalized with the kinetic data in several ways. Labile a,P,y coordination of ATP by Cd'+ or Mn", although inhibitory, could partially relieve the complete inhibition caused by an incorrectly oriented bulky thiol group.2 A second alternative is that the enzyme-bound Cd2+, due to its extremely high affinity for sulfur, may allow s o n e activity with the B-isomer of ATPaS due to direct coordination of sulfur as previously proposed (4), but Mn'+, which activates more effectively with this isomer, may do so via a second sphere complex, as detected by NMR. Direct coordination between Cd2+ only and the thio-nucleotide is consistent with the finding that the KI"pp of AMP is largely independent of the nature of the divalent cation activator (Mg2+ -Mn'+ -CdZ+), while the K:pp of AMPaS is significantly smaller with Cd2+ (Mg2+ z Mn'+ > > Cd2+) (Table VII), and with the NMR studies (Table 111) which show only second sphere nucleotide complexes of the enzyme-bound Mn2+. Thirdly, transient aphosphoryl coordination by the enzyme-bound (or nucleotidebound) metal could occur only in the transition state before bond breaking to the leaving AMP group is complete. Finally, Co(NHd4ATP, although a substrate, may interact somewhat differently with the enzyme-bound cation than ATP or ATPaS. Further kinetic studies with stable metal complexes of ATPaS may distinguish among these possibilities. A self-consistent location for the activating phosphate ion, which is required by all known PP-Rib-P synthetases, has been found, based on distances from two paramagnetic reference points, the enzyme-bound (Table VI) and the nucleotidebound metal activators (5). The location of P, near the bound substrates ( Fig. 6) further supports our previous suggestion that it may be directly involved in catalysis, possibly functioning as the general base which deprotonates the a-1-OH group of Rib-5-P (5). In addition to binding at this site in all complexes, Pi also appears to bind weakly at the 5-P site of Rib-5-P and at the P, site of ATP on the enzyme when these sites are not occupied by the substrates. Independent evidence for the binding of approximately three phosphate anions at the active site of PP-Rib-P synthetase has previously been obtained in kinetic studies of the protection by P, of the enzyme against inactivation by permanganate oxidation of an essential thiol group (20). The dependence of Pi protection on the P, concentration raised to the 2.7th power led to the suggestion that "as many as three P, anions may associate with the active site to protect it against KMn04" (20). Binding of P, at these sites on PP-Rib-P synthetase, and on a number of other ATP utilizing enzymes, may constitute a physiological mechanism for stabilization of such enzymes, and by simple competition, may regulate their affinity for phosphorylated substrates in cells.