Cystathionine β-Synthase (CBS) Domain-containing Pyrophosphatase as a Target for Diadenosine Polyphosphates in Bacteria*

Background: Many soluble pyrophosphatases contain two regulatory nucleotide-binding CBS domains with or without an intercalating DRTGG domain. Results: Linear P1,Pn-diadenosine 5′-polyphosphates (ApnAs, n = 3–6) bind with nanomolar affinity to and activate DRTGG domain-containing pyrophosphatases; Ap3A binds cooperatively. Conclusion: Nucleotide-regulated pyrophosphatases may represent receptors for ApnAs in bacteria. Significance: The results suggest a novel regulatory pathway in some bacteria, involving ApnAs as messengers. Among numerous proteins containing pairs of regulatory cystathionine β-synthase (CBS) domains, family II pyrophosphatases (CBS-PPases) are unique in that they generally contain an additional DRTGG domain between the CBS domains. Adenine nucleotides bind to the CBS domains in CBS-PPases in a positively cooperative manner, resulting in enzyme inhibition (AMP or ADP) or activation (ATP). Here we show that linear P1,Pn-diadenosine 5′-polyphosphates (ApnAs, where n is the number of phosphate residues) bind with nanomolar affinity to DRTGG domain-containing CBS-PPases of Desulfitobacterium hafniense, Clostridium novyi, and Clostridium perfringens and increase their activity up to 30-, 5-, and 7-fold, respectively. Ap4A, Ap5A, and Ap6A bound noncooperatively and with similarly high affinities to CBS-PPases, whereas Ap3A bound in a positively cooperative manner and with lower affinity, like mononucleotides. All ApnAs abolished kinetic cooperativity (non-Michaelian behavior) of CBS-PPases. The enthalpy change and binding stoichiometry, as determined by isothermal calorimetry, were ∼10 kcal/mol nucleotide and 1 mol/mol enzyme dimer for Ap4A and Ap5A but 5.5 kcal/mol and 2 mol/mol for Ap3A, AMP, ADP, and ATP, suggesting different binding modes for the two nucleotide groups. In contrast, Eggerthella lenta and Moorella thermoacetica CBS-PPases, which contain no DRTGG domain, were not affected by ApnAs and showed no enthalpy change, indicating the importance of the DTRGG domain for ApnA binding. These findings suggest that ApnAs can control CBS-PPase activity and hence affect pyrophosphate level and biosynthetic activity in bacteria.

Ap n As are involved in many other processes, including neurotransmission (15), apoptosis (16), and analgesia (17). Of note, Ap 4 A is used in hypoxia therapy in humans (18).
Understanding the roles of Ap n As requires knowledge of their target proteins. Using a radioactive photocrosslinking Ap 4 A analog, Johnstone and Farr (19) detected 12 Ap 4 A-binding proteins in E. coli extract, some of which were identified as heat shock proteins based on their electrophoretic mobilities. Guo et al. (20) and Azhar et al. (21) used pulldown assays with immobilized Ap 4 A analogs followed by mass-spectral analysis to identify, respectively, 6 and 13 binding proteins in E. coli. The three protein sets obtained in these studies partially overlapped. Few Ap n A protein complexes have been subjected to biophysical and mechanistic studies. Apart from cases where Ap n As act as substrates or products of their metabolizing enzymes, the chaperone GroEL binds Ap 4 A with a dissociation constant of 10 M; the complex exhibits increased ATPase and chaperoning activities (11). Human 5Ј-nucleotidase II is allosterically activated by Ap n As (n ϭ 4 -6), which bind with dissociation constants of 60 -80 M (22).
Inorganic pyrophosphatases (PPases; EC 3.6.1.1), the major PP i -metabolizing enzymes in all types of organisms, belong to three nonhomologous families (23). Family II PPases, found in bacteria and archaea, are homodimeric Mn 2ϩ -or Co 2ϩ -metalloenzymes that additionally require Mg 2ϩ for catalysis (24). A quarter of the more than 500 putative family II PPase sequences contain a regulatory insert comprising a pair of cystathionine ␤-synthase (CBS) domains (Bateman module (25)) within one of the two catalytic domains. Regulatory CBS domains are found in proteins in all kingdoms of life and generally bind adenine nucleotides as regulating molecules (26 -28); mutations in CBS domains of human proteins are associated with hereditary diseases (29,30). Interestingly, only in CBS-PPases (but not all of them), are the CBS domains intercalated by another (DRTGG) domain. CBS-PPases are activated by ATP and inhibited by AMP and ADP (31,32). Both catalysis and regulation involve marked positive cooperativity, which is Mg 2ϩ -dependent (32).
The structure of the isolated dimeric regulatory insert of Clostridium perfringens PPase (cpPPase) obtained for crystals grown in the presence of 0.25 mM Ap 4 A contains an AP 4 A molecule bound by two CBS domain pairs at the subunit interface (33), raising the possibility that Ap 4 A may be a physiological ligand of CBS-PPases. Preliminary activity measurements (33,34) suggested that Ap 4 A activates cpPPase. Here we show that all Ap n As bind with nanomolar affinities to three DRTGG domain-containing CBS-PPases and modulate their catalytic activity and cooperative behavior. Our data thus identify a new type of ligand for CBS domains and an important target of Ap n As in the protein world.

Experimental Procedures
Enzymes and Reagents-Genes for CBS-PPases from Desulfitobacterium hafniense (dhPPase), Clostridium novyi (cnPPase), C. perfringens (cpPPase), Eggerthella lenta (elPPase), and Moorella thermoacetica (mtPPase) were expressed in E. coli, and the produced CBS-PPases were purified as described previously (32)(33)(34). Inactive aggregates were sepa-rated from soluble active proteins during size exclusion chromatography. The final products were at least 95% pure as estimated by SDS-PAGE using a Phast system with 8 -25% gradient gels (GE Healthcare). Protein concentrations were determined with a Nanodrop spectrophotometer (Thermo Scientific) using A 280 0.1% of 0.478 for dhPPase, 0.548 for cnPPase, 0.426 for cpPPase, 0.493 for elPPase, and 0.48 for mtPPase, as calculated from their amino acid compositions with ProtParam. Molar concentrations were calculated based on subunit molecular masses of 60.4, 63.6, 60.8, 52.5, and 48.1 kDa, respectively. All enzyme concentrations are given in terms of the dimer. P 1 ,P n -diadenosine 5Ј-polyphosphates (Ap n As) with n ϭ 3-5 were from Sigma; Ap 6 A was from Jena Bioscience. All Ap n As were Ն97% pure, and Ap 3 A was essentially free of other Ap n As, according to the manufacturer analyses (HPLC). The concentrations of stock nucleotide solutions were calibrated by measuring absorbance in the ultraviolet region (⑀ 259 ϭ 31,800 M Ϫ1 ⅐cm Ϫ1 for dinucleotides and 15,900 M Ϫ1 ⅐cm Ϫ1 for mononucleotides).
Kinetic Assays-The activity assay medium contained 5 mM MgCl 2 , 140 M PP i (yielding 50 M MgPP i complex) and 0.1 M TES-KOH (pH 7.2), except where specified otherwise. In measurements done at higher Mg 2ϩ concentrations, buffer concentration was decreased appropriately to maintain constant ionic strength. The reaction was initiated by adding enzyme, and P i accumulation caused by PP i hydrolysis was continuously recorded for 2-3 min at 25°C using an automated P i analyzer (35). Initial velocities of PPi hydrolysis were typically estimated graphically from the slopes of the tangents to the initial portion of hydrolysis time courses recorded with the P i analyzer.
Isothermal Calorimetry-A VP-iTC calorimeter (MicroCal Ltd.) was used. Enzyme and nucleotide solutions were made in 0.1 M MOPS/KOH (pH 7.2) buffer containing 2 mM MgCl 2 , 0.1 mM CoCl 2 , and 150 mM KCl. Titrations were performed at 25°C by successive 10-l injections of 0.1-10 mM mononucleotide or 33 M dinucleotide solution into 2 ml of CBS-PPase solution (2.5-5 M in terms of the dimer); the interval between injections was 5 min. All samples were degassed before the experiment. Binding isotherms were corrected by subtracting the ligand dilution isotherms, determined by titrating nucleotide solutions into the buffer.
Calculations and Data Analysis-The values of the apparent dissociation constants for the magnesium complexes of PP i used to maintain required free Mg 2ϩ ion and MgPP i complex concentrations at pH 7.2 were 112 M for MgPP i and 2.84 mM for Mg 2 PP i (36). Nonlinear least square fittings were performed using the program Scientist (Micromath). The dependence of hydrolysis rate on nucleotide concentration ([N]) was fit to Equation where v 0 and v N are activities of free and nucleotide-saturated enzyme, respectively, and K N1 and K N2 are the macroscopic dissociation constants describing successive binding of nucleotide to two regulatory sites per enzyme molecule. Cooperative kinetics of substrate (MgPP i ) hydrolysis were analyzed with which assumes different Michaelis constants (K m1 and K m2 ) and equal k cat values for the two active sites in the dimer.
[E] 0 and [S] are total enzyme and substrate concentrations, respectively. The corresponding binding schemes and details of the fitting procedure were described previously (32).
The dependences of K N1 , K N2 , K m1 , and K m2 on Mg 2ϩ (M) concentration were fit to Equation 3, where (K L ) 0 and (K L ) M are the limiting values of the respective K N or K m at 0 and infinite Mg 2ϩ concentrations, and K m is the metal binding constant.
Alternatively, rate dependences on substrate and nucleotide concentrations were fit to a Hill-type Equation 4, where L is S or N, v L is the rate at infinite [L], and h is the Hill coefficient. The value of v 0 was set to 0 when L was substrate, and the value of h was set to unity for noncooperative binding.
Isothermal titration calorimetry (ITC) data were analyzed with a MicroCal ITC subroutine in Origin 7.0 software using a single-binding site model. Thermodynamic parameters were calculated from the standard relationship, ⌬G ϭ RT lnK N ϭ ⌬H Ϫ T⌬S. Fig. 1 shows the concentration dependences of the effects of four Ap n As with n ϭ 3-6 on the activities of three CBS-PPases measured at fixed substrate (MgPP i ) and Mg 2ϩ concentrations (50 M and 5 mM, respectively). Nanomolar concentrations of Ap n As caused marked activation in all cases, except that Ap 3 A was effective with cpPPase at micromolar concentrations.

Effects of Ap n As on CBS-PPases at a Fixed Mg 2ϩ Concentration-
Analyses of the dependences shown in Fig. 1 and of similar dependences measured at different substrate concentrations (1 and 300 M) were initially done using Equation 4. The value of the Hill coefficient was indistinguishable from unity (1 Ϯ 0.05) at all substrate concentrations for Ap n As with n ϭ 4 -6. In contrast, Ap 3 A bound cooperatively (h ϭ 1.4 -1.7) at all substrate concentrations. Accordingly, the data for Ap 3 A were analyzed with Equations 1 and 4 in their general forms, whereas Equation 4 with h ϭ 1 was used for the other Ap n As. The parameter values derived from this analysis are summarized in Tables 1 and 2.
The values of the activation factor (v N /v 0 ) and their trends with changing polyphosphate length and substrate concentration were similar for the three enzymes. The value of v N /v 0 was greater at low than at high substrate concentrations. In the presence of 300 M substrate, which is in excess of the respective Michaelis constants (32), v N /v 0 approached a value of ϳ2 in all cases.
The apparent binding affinities of the nucleotides could be compared on the basis of the average binding constant ( ͱ K N1 K N2 ) for Ap 3 A and respective K N values for the other dinucleotides. As Tables 1 and 2 make clear, the binding affinity estimated at 50 M substrate was markedly lower for Ap 3 A compared with other dinucleotides for all CBS-PPases. Increasing n did not affect dhPPase affinity, but it did slightly increase cnPPase affinity and decrease cpPPase affinity. Increasing substrate concentration had opposite effects on the affinity of Ap 3 A and Ap 4 A for dhPPase and cpPPase (increased) and cnPPase (decreased). Of note, cpPPase exhibited much lower affinity for all dinucleotides compared with other CBS-PPases.
Surprisingly, neither dinucleotide at a concentration up to 10 M affected activities of elPPase or mtPPase measured with 50 M substrate. These CBS-PPases differ from those described above by having no DRTGG domain in their regulatory regions, which are formed by only two CBS domains. Moreover, 10 M Ap 4 A did not affect the concentration dependence of ADP inhibition of elPPase or mtPPase (data not shown), indicating that the dinucleotide is unable to interact with the ADP-binding site.
Dependence of CBS-PPase Activation on Mg 2ϩ Concentration-Given that cooperativity in CBS-PPases is Mg 2ϩdependent (32), measurements analogous to those illustrated in Fig. 1 were conducted for two representative dinucleotides, Ap 3 A and Ap 4 A, over a 0.05-20 mM Mg 2ϩ concentration range; substrate concentration was fixed at 50 M. The results of these experiments ( Fig. 2) indicated that Ap 3 A bound with positive cooperativity and Ap 4 A bound noncooperatively to all CBS-PPases at all Mg 2ϩ concentrations. In only one case (dhPPase with Ap 3 A), the degree of cooperativity, as characterized by the values of h and the ratio K N2 /K N1 , showed a pronounced dependence on [Mg 2ϩ ] because of the opposite effects of Mg 2ϩ on K N1 and K N2 (Fig. 2). In all other cases, K N1 and K N2 changed in the same direction to approximately the same degree and, consequently, without a marked effect on cooperativity. Of note, the ratio K N2 /K N1 equals 4 in the case of noncooperative binding and is less than 4 for positively cooperative binding (37).
In most cases (except for dhPPase with Ap 4 A), Mg 2ϩ modulated dinucleotide binding, with the direction of the effect depending on both the nature of the nucleotide and the CBS- The lines show the best fits of Equations 1 or 4 (see text for details). Activity without nucleotides (220, 350, and 800 s Ϫ1 for dhPPase, cnPPase and cpPPase, respectively) was taken as unity. dh, dhPPase; cn, cnPPase; cp, cpPPase.
PPase origin (Figs. 2 and 3). Mg 2ϩ generally stimulated Ap 3 A binding, except for dhPPase, where it exerted the opposite effect on K N1 (Fig. 2). Mg 2ϩ exhibited a full range of effects on Ap 4 A binding (Fig. 2): stimulation (cpPPase), suppression (cnPPase), and no effect (dhPPase). The effect of Mg 2ϩ on dinucleotide binding could be described by Equation 3, yielding the parameter values summarized in Table 3. The values of K m governing the Mg 2ϩ effects were in the millimolar range and were similar for both steps of Ap 3 A binding and Ap 4 A binding for a given CBS-PPase.
The degree of activation (v N /v 0 ) of dhPPase and cnPPase by Ap 3 A and Ap 4 A demonstrated no or only small variations with Mg 2ϩ concentration (Fig. 2). In contrast, activation of cpPPase showed a bell-shaped dependence (Ap 3 A) or markedly decreased (Ap 4 A) with increasing Mg 2ϩ concentration.
Analysis of CBS-PPase Activation in Terms of Michaelis-Menten Parameters-As previously reported, the rate of MgPP i hydrolysis by CBS-PPases does not obey Michaelis-Menten kinetics, requiring the use of a more complex equation with two Michaelis constants (32). Their ratio, K m2 /K m1 , was less than 4, and the Hill coefficient was greater than 1, indicating positive kinetic cooperativity. Surprisingly, Ap 3 A and Ap 4 A completely abolished or markedly suppressed the kinetic cooperativity in dhPPase, cnPPase and cpPPase, as indicated by a Hill coefficient with a value close to 1 (Table 4 and Fig. 3). That the h value is greater than 1 for cpPPase in the presence of Ap 3 A may reflect incomplete saturation of this enzyme by the dinucleotide, which binds much more weakly to cpPPase compared with the other CBS-PPases, especially at low substrate concentrations ( Table 1).
The kinetics of activation by Ap 4 A was investigated over a range of Mg 2ϩ concentrations (Fig. 3). The results showed that 10 M activator increased k cat , decreased the Michaelis constant, and abolished kinetic cooperativity. Again the largest effects were observed with cpPPase, which was therefore explored in greater detail.
The effects of four Ap n As on the Mg 2ϩ concentration dependence of k cat for cpPPase were qualitatively similar (Fig. 4). Mg 2ϩ induced a transition from low to high activity over a narrow range of concentrations, requiring a term with [Mg 2ϩ ] 2 in the corresponding equation (see Fig. 3 legend) (32). All four activators increased the limiting value of k cat at infinite [Mg 2ϩ ] (k cat,M ) and decreased the Mg 2ϩ binding constant (K m ) ϳ2-fold (Table 5). Most surprisingly, Ap n A binding conferred catalytic activity to the otherwise inactive cpPPase at low [Mg 2ϩ ] (see k cat,0 values in Table 5). The activity of Ap 4 A-activated cpPPase in these conditions approached its maximum activity observed at high [Mg 2ϩ ] in the absence of Ap 4 A (Fig. 4). Fig. 5 illustrates the concentration dependence of cpPPase activation by Ap 4 A in the presence of 0.5 mM Mg 2ϩ , analyzed in terms of k cat and K m values. The value of k cat increased ϳ7.5fold (from 240 Ϯ 100 to 1800 Ϯ 100 s Ϫ1 ), K m1 decreased ϳ18fold (from 70 Ϯ 10 to 4 Ϯ 1 M), and K m 2 changed insignificantly with increasing Ap 4 A concentration from 0 to 5 M. The Ap 4 A binding constant estimated from k cat and K m1 dependences was 0.04 Ϯ 0.01 and 1.7 Ϯ 1.0 M, respectively. Because k cat and K m dependences report on Ap 4 A binding to substratefree enzyme and enzyme-substrate complex, respectively, a likely implication is that Ap 4 A and the first bound substrate molecule mutually stabilize binding of each other to cpPPase 20 -40-fold.
Thermodynamics and Stoichiometry of Nucleotide Binding CBS-PPases-Using ITC allowed the direct measurement of changes in free energy (⌬G), enthalpy (⌬H), and entropic free energy (T⌬S) components of nucleotide binding to CBS-PPases. A typical titration profile is shown in Fig. 6A. The results of similar titrations performed with different CBS-PPases and nucleotides are summarized in Fig. 6B and Table 6.
One important result was that titrations of the DRTGG domain-lacking elPPase or mtPPase with up to 10 M Ap 4 A or  The value of the Hill coefficient was indistinguishable from unity in all cases. a v N and v 0 are activities extrapolated to infinite concentration of the variable nucleotide and measured in the absence of any nucleotide, respectively.

Enzyme/dinucleotide
b This parameter is equivalent to ͱ K N1 K N2 in Table 1. NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46

Nucleotide-regulated Soluble Pyrophosphatases
Ap 3 A produced no ITC signal, consistent with the inability of the dinucleotides to activate these CBS-PPases and modulate their inhibition by ADP. Because the lack of effect on activity did not rule out the possibility of a "silent" binding, the ITC data, which report on a different aspect of the binding reaction, provided an important support for the lack of complex formation between the DRTGG domain-lacking CBS-PPases and Ap n As. This interpretation was supported by parallel measurements employing AMP, ADP, and ATP (Table 6), which produced similar enthalpy changes in the cases, where previous measurements (32) revealed effects on activity, but no or reduced enthalpy change (elPPase with AMP and cnPPase with ADP, respectively), where no effect on activity was observed (32). Together, these findings suggest that modulation of activity and heat production are coupled phenomena and that the DRTGG domain is required for tight binding of diadenosine polyphosphates, but not monoadenosine phosphates, to CBS-PPases. The inability of elPPase to bind AMP is not associated with the absence of the DRTGG domain because another DRTGG domain-lacking CBS-PPase, mtPPase, is inhibited by AMP and hence binds it (31). Another important finding was that ⌬H, as calculated per mole of nucleotide, was nearly two times greater for Ap 4 A and Ap 5 A than for Ap 3 A and the mononucleotides in the titrations with the DRTGG domain-containing CBS-PPases. This effect correlated with a 2-fold lower binding stoichiometry for Ap 4 A and Ap 5 A compared with that for mononucleotides and Ap 3 A.
Because of the very tight binding, K N and, accordingly, T⌬S values could not be estimated with adequate precision in most Ap n A titrations. Where K N (and hence ⌬G) values were available, the free energy change of nucleotide binding was dominated by ⌬H, with a significant contribution from T⌬S, likely because of a hydrophobic effect. The K N values derived from ITC measurements are in a fair agreement with those obtained from nucleotide effects on activity (see Ref. 32 for mononucle-

FIGURE 2. Mg 2؉ concentration dependence of CBS-PPase activation by Ap 3 A (left panel) and Ap 4 A (right panel).
The panels show (from top to bottom) the activation factor K N1 (E) and K N2 (F) values and Hill coefficients. The K N1 and K N2 lines show the best fits to Equation 3. The horizontal dotted lines (h ϭ 1) mark the boundary between positive and negative cooperativity. dh, dhPPase; cn, cnPPase; cp, cpPPase. otides and Table 1 for Ap 3 A). It should be noted that ITC measurements can hardly distinguish positive binding cooperativity and yield an average ⌬H value for all binding sites.

Discussion
CBS domains, found in many proteins, are known for their ability to bind adenine nucleotides and in this way regulate activities of their carrier proteins. The list of regulating adenine nucleotides includes AMP, ADP, ATP, S-adenosyl methionine, NADH, and analogs of AMP and ATP (27). Examples of less common CBS domain ligands include Mg 2ϩ (38), DNA, and RNA (39,40). We earlier reported that crystals of the isolated dimeric regulatory region of cpPPase grown in the presence of Ap 4 A contains one Ap 4 A molecule per dimer bridging two

TABLE 4 Kinetic parameters for PP i hydrolysis in the presence of 50 M Ap 3 A and 5 mM Mg 2؉ estimated with Equation 2
The values in parentheses refer to parameter values previously measured in the absence of Ap 3 A (32).    pairs of CBS domains, whereas each CBS domain pair binds an AMP molecule (33). We also found that Ap 4 A induces a significant opening of the interface compared with the AMP-bound form. The results reported above extend these earlier findings by showing that (a) Ap n As with n ϭ 3-6 bind three CBS-PPases with nanomolar affinity and activate them in vitro; (b) Ap n A binding is only observed in CBS-PPases that have an intercalating DRTGG domain in the regulatory region; and (c) unlike common adenine nucleotides, long chain Ap n As (n Ͼ 3) abolish or markedly reduce kinetic cooperativity (non-Michaelian behavior) in CBS-PPases. The unique features of Ap n A complexes of CBS-PPases compared with those of their complexes with mononucleotides and complexes of other CBS proteins with their regulating ligands are described below. Notably, Ap n As have not been reported as ligands for any other CBS protein.
Based on their binding properties, Ap n As can be divided into two groups. Ap 3 A bound to CBS-PPases cooperatively and with lower affinity, as characterized by either K N1 and K N2 or their average value͑ ͱ K N1 K N2 ) ( Table 1) . The other dinucleotides (n ϭ 4 -6) bound noncooperatively and with a higher affinity that did not depend significantly on the n value ( Table 2). The affinities of Ap n As with n ϭ 4 -6 for CBS-PPases surpassed that of adenine mononucleotides (32) by 2-3 orders of magnitude. Such high affinities are unprecedented among other CBS proteins, which generally bind their nucleotide ligands in the millimolar range. The difference in the binding affinities of the two Ap n A groups was most pronounced with cpPPase, amounting to 3 orders of magnitude. As previously demonstrated (33), Ap 4 A interacts through both of its adenine moieties with two CBS domain pairs of different subunits in cpPPase. Such an arrangement is also likely with Ap 5 A and Ap 6 A, consistent with their similar ⌬H values and binding stoichiometries, determined from ITC measurements ( Table 6). In contrast, ⌬H for Ap 3 A was half that of Ap 5 A and Ap 6 A, and the binding stoichiometry was 2-fold higher, similar to values for mononucleotides (Table 6). These observations likely indicate that Ap 3 A predominantly binds CBS-PPases through only one adenine moiety.
The binding affinities of Ap n As showed a complex dependence on substrate and metal cofactor concentrations. At a constant Mg 2ϩ concentration, substrate increased the binding affinities of dhPPase and cpPPase for all Ap n As but exerted an opposite effect on cnPPase (Tables 1 and 2). Accordingly, Ap 3 A (Table 4) and Ap 4 A (Fig. 3) decreased the average Michaelis constant ( ͱ K m1 K m2 and K m ). The effect of Ap 3 A on ͱ K m1 K m2 for cpPPase measured in the presence of 5 mM Mg 2ϩ was quite modest, but keeping in mind the bell-shaped dependence of ͱ K m1 K m2 on [Mg 2ϩ ] for this enzyme in the absence of adenine  nucleotides (Fig. 3), one would expect, by analogy, greater effects of Ap 3 A at low [Mg 2ϩ ]. However, the most striking effect of Ap n As on substrate binding was abolition of kinetic cooperativity. This effect was observed with both Ap 3 A and Ap 4 A, representing the two dinucleotide groups and might be explained by two different mechanisms. First, the effectors may disrupt the communication between active sites, allowing them to function independently. Alternatively, the dinucleotides may induce asymmetry in the enzyme dimer such that only one active site operates in the dimer (ultimate negative cooperativity). Determining the three-dimensional structure of the enzyme with bound dinucleotide would make it possible to discriminate between these alternative explanations.
Mg 2ϩ effects on nucleotide binding also varied depending on the enzyme (Fig. 2) and differed from those observed with adenine mononucleotides (32). With Ap 3 A, values of K N1 and K N2 for dhPPase changed in different directions, decreasing the degree of cooperativity at low [Mg 2ϩ ] ( Fig. 2A). No bound Mg 2ϩ ion was observed in the structure of the regulatory region of cpPPase (33), suggesting that the modulatory Mg 2ϩ resides in the active site. Notable in this regard, three Mg 2ϩ ions per active site participate in catalysis among homologous nonregulated family II PPases (24,41). The effects of Mg 2ϩ on nucleotide binding may, in part, be a consequence of its effects on substrate binding, because these measurements were carried out at a nonsaturating substrate concentration (50 M).
Both Ap 3 A and Ap 4 A activated CBS-PPases under the conditions tested because of favorable changes in both k cat and the average Michaelis constant ( ͱ K m1 K m2 ) ( Table 4 and Fig. 3).
Accordingly, the degree of activation was greater at low substrate concentrations (Table 1) and varied from severalfold to several ten-fold. The largest effects were observed with cpPPase. Based on its k cat and K m values (Fig. 3), this enzyme is predicted to be activated by Ap 4 A in the presence of 1 mM Mg 2ϩ by a factor of ϳ51 and ϳ19 at substrate concentrations of 1 and 10 M, respectively. At low [Mg 2ϩ ], the activating effect of Ap n A is dominated by k cat , especially with cpPPase ( Fig. 4 and Table 5). In this enzyme, k cat is strongly Mg 2ϩ -dependent and Ap n A markedly released this dependence by allowing catalysis in the enzyme with a vacant Mg 2ϩ site and by somewhat increasing its affinity for Mg 2ϩ (Table 5). In this respect, Ap n As partially substitute for Mg 2ϩ as an enzyme activator.
Qualitatively similar activating effects on CBS-PPases were previously observed with ATP (31, 32), although ATP effects were smaller in size and required much higher effector concentrations. A further difference is that ATP bound cooperatively, like Ap 3 A. The effects of ATP and Ap 3 A are thus similar in many aspects. As noted above, activator binding induces significant opening of the CBS domain interface (33). Such opening can be achieved upon binding of a single molecule of Ap 4 A or a longer dinucleotide that binds to both subunits of CBS-PPase through two adenine moieties. Structure modeling of the cpPPase regulatory region indicated that the polyphosphate chain of Ap 3 A is too short for this binding mode. In this case, and with ATP, interface opening apparently results from repulsion between two molecules of the effector bound to different subunits.
The requirement for an intercalating DRTGG domain for Ap n A binding to CBS domains provides another interpretive challenge. In the structure of the regulatory region and the modeled structure of the whole cpPPase, both the DRTGG domain and CBS domain pairs participate in forming the subunit interface (33). DRTGG domain-containing CBS-PPases apparently have a larger binding cavity for the regulating ligands or increased flexibility of the CBS domains at the expense of their smaller contribution to the subunit contact area, allowing them to accommodate more bulky Ap n A molecules. This interpretation is supported by data showing that the DRTGG domain-deficient elPPase (32) and mtPPase (31) bind ATP with an affinity 1-2 orders of magnitude lower than that of the less bulky AMP and ADP. In contrast, no such discrimination is observed in DRTGG domain-containing CBS-PPases (32). Notably, the primary structures of the CBS domains in DRTGG domain-deficient CBS-PPases (Fig. 7) do not contain specific mutations that would disallow their binding of Ap n As. Despite a generally low degree of residue conservation in CBS domains, all residues involved in nucleotide binding are found in at least one of the DRTGG domain-deficient CBS-PPases. Based on these considerations, Ap n As are not expected to bind with comparable affinity to the numerous other CBS proteins that lack a DRTGG or other intercalating domain.
Ap n A binding is expected to significantly change CBS-PPase activity in vivo, particularly under low energy conditions, when the concentration of the alternative activator, ATP, is low. Although basal intracellular levels of Ap n As are 4 orders of magnitude lower than those of adenine mononucleotides, Ap n A concentrations can rise by 2 orders of magnitude under stress conditions (42,43). Also taking into consideration their extraordinarily high affinity, Ap n As could be expected to efficiently compete with mononucleotides for CBS-PPase binding in these circumstances. An increase in CBS-PPase activity is expected to decrease the concentration of PP i and thus release PP i -mediated inhibition of numerous biosynthetic reactions in which PP i is produced as a by-product (44). That the affinity of CBS-PPases for Ap n As markedly surpasses that of all known Ap n A-binding proteins suggests that this enzyme is a dominant target through which Ap n As fulfill their stress response-related functions in bacteria.