Isothermal titration calorimetry reveals a zinc ion as an atomic switch in the diadenosine polyphosphates.

Diadenosine polyphosphates (diadenosine 5',5'''-P(1),P(n)-polyphosphate (Ap(n)A)) are 5'-5'''-phosphate-bridged dinucleosides that have been proposed to act as signaling molecules in a variety of biological systems. Isothermal titration calorimetry was used to measure the affinities of a variety of metal cations for ATP, diadenosine 5',5'''-P(1),P(3)-triphosphate (Ap(3)A), diadenosine 5',5'''-P(1),P(4)-tetraphosphate (Ap(4)A), and diadenosine 5',5'''-P(1),P(5)-pentaphosphate (Ap(5)A). The binding of Mg(2+), Ca(2+), and Mn(2+) to ATP is shown to take place with the beta,gamma-phosphates (primary site) and be endothermic in character. The binding of Ni(2+), Cd(2+), and Zn(2+) to ATP is found to take place at both the primary site and at a secondary site identified as N-7 of the adenine ring. Binding to this second site is exothermic in character. Generally, the binding of metal cations to diadenosine polyphosphates involves a similar primary site to ATP. No exothermic binding events are identified. Critically, the binding of Zn(2+) to diadenosine polyphosphates proves to be exceptional. This appears to involve a very high affinity association involving the N-7 atoms of both adenine rings in each Ap(n)A, as well as the more usual endothermic association with the phosphate chain. The high affinity association is also endothermic in character. A combination of NMR and CD evidence is provided in support of the calorimetry data demonstrating chemical shift changes and base stacking disruptions entirely consistent with N-7 bridging interactions. N-7 bridging interactions are entirely reversible, as demonstrated by EDTA titration. Considering the effects of Zn(2+) on a wide variety of dinucleoside polyphosphate-metabolizing enzymes, we examine the possibility of Zn(2+) acting as an atomic switch to control the biological function of the diadenosine polyphosphates.

The proposal that diadenosine 5Ј,5ٞ-P 1 ,P n -polyphosphates (Ap n A) 1 are signaling molecules has had a long, controversial history since their original intracellular discovery over 30 years ago (1). Since then, dinucleoside polyphosphates have been found intracellularly in both prokaryotic and eukaryotic organisms, and have been shown to be specific purinoreceptor ligands extracellularly in systems ranging from synaptic trans-mission to vasocontrol. The younger field of extracellular functioning has arguably proved more successful in defining a role for the diadenosine polyphosphates (2), for in spite of their intracellular ubiquity, definitive proof of physiological roles remains elusive, although many such roles have been proposed. The range of roles in which intracellular diadenosine polyphosphates have been implicated includes the control of the proliferative status of mammalian cells (3), prokaryotic stress (4,5), DNA repair (6), the timing of cell division (7), and as substrates of the tumor suppressor fragile histidine triad protein (EC 3.6.1.29) (8). Extracellularly, they have been shown as important signaling molecules (9,10), inducers of nitric oxide release (11), important neurotransmitters (12), and as vasocontrollers (13). Recent reviews on intracellular functioning (14), extracellular functioning (2), anabolic enzymes (15), and catabolic enzymes (16) provide an excellent overview of the area.
To investigate the physiological functioning of this class of molecules, it is important to have an understanding of their structure and how varying conditions of metal cations, pH, and temperature may be controls of dinucleoside polyphosphate function. In particular, we have been interested in the role of Zn 2ϩ in strongly augmenting the aminoacyl-tRNA synthetase synthesis of adenyl dinucleoside polyphosphates (17)(18)(19) and the role of the ion in the heat shock response. Most of the class II aminoacyl-tRNA synthetases are stimulated by Zn 2ϩ in Ap n A synthesis, yet many do not have clear zinc-binding motifs. The aminoacyl-tRNA synthetases that do have zinc-binding motifs use the zinc for amino acid discrimination (20) without much of an effect on Ap 4 A synthesis. Furthermore, hydrolysis of the diadenosine polyphosphates has in general been shown to be inhibited by Zn 2ϩ , although stimulated by Mg 2ϩ , Mn 2ϩ , and Ca 2ϩ . A list of the effects of zinc on some Ap n A-metabolizing enzymes is shown in Table I. Such considerations of the links between Zn 2ϩ and Ap n A have been examined before (21), but in this work we test the hypothesis that such communication is enabled by the direct interaction between the diadenosine polyphosphates and the divalent zinc cation itself.
Previously, the most informative studies of dinucleoside polyphosphate structure have come from CD (34 -36), NMR (37)(38)(39), and x-ray crystallography (40). Generally, from these studies it has been shown that Ap 4 A can exist in three major conformations as follows: stacked, folded unstacked, and open. However, as the folded unstacked form is only stable below pH 5 and at high temperature (Ͼ80°C), under physiological conditions it has been assumed that a two-state equilibrium exists between the open form and the base stacked form (37). The structure is also affected by metal ions. NMR shows a single Mg 2ϩ to stabilize the ring-stacked conformation by interacting symmetrically with both ␤-phosphates (38). CD studies have revealed that Zn 2ϩ destabilizes the ring-stacked conformation (35), and cis-platin has been reported to bind to Ap 4 A in the form of an N-7 to N-7 chelate (41). Although Zn 2ϩ binding reduces base stacking, the conformational change has not been clearly defined by previous studies using CD nor has the strength of binding been resolved.
As a new probe into the nature of the polyphosphate-metal interaction, we have used isothermal titration calorimetry (ITC) to elucidate the binding of a range of metal ions to Ap 3 A, Ap 4 A, Ap 5 A, and to ATP. This technique has the advantage of being able to characterize fully the thermodynamics of the binding events and to clearly show the stoichiometry of the stronger binding events. We have also used NMR and CD to complement and confirm our calorimetric data. As such, ITC provides a powerful new handle on the understanding of metal ion control of dinucleoside polyphosphate conformation.

EXPERIMENTAL PROCEDURES
Materials-Ammonium salts of dinucleoside polyphosphates were purchased from Sigma. Metal chloride salts were purchased from Sigma and were of spectral grade. Milli-Q water (Millipore) was used throughout. The sodium cations of the dinucleoside polyphosphates were exchanged for protons using a Dowex cation exchange column and then polyphosphates were freeze-dried. Purities of the compounds were checked on an anion exchange column (MonoQ, Amersham Biosciences) using a 0 -1 M NaCl gradient in 20 mM Tris, pH 8.0, on a fast protein liquid chromatography system with an ultraviolet monitor (Amersham Biosciences) set at 260 nm. ATP was estimated by fast protein liquid chromatography peak integration to be 95% pure, Ap 3 A to be 95% pure, Ap 4 A to be 98% pure, and Ap 5 A to be 95% pure. Concentrations of polyphosphates were measured using the following extinction coefficients at 260 nm: Ap 3 A 25,100 M Ϫ1 cm Ϫ1 , Ap 4 A 27,100 M Ϫ1 cm Ϫ1 , and Ap 5 A 27,800 M Ϫ1 cm Ϫ1 (35).
Isothermal Calorimetry-Isothermal titration calorimetry (ITC) measurements were carried out with a VP-ITC calorimeter (Microcal, Northampton, MA) interfaced with a computer. Stock solutions of nucleoside polyphosphates (50 mM) and of divalent metal ions with chloride counter ions (1 M) were prepared, and then solutions for ITC were prepared by dilution into 20 mM Tris, pH 8.0, to the required concentration. All solutions were centrifuged for 5 min at 13,000 rpm and then thoroughly degassed under vacuum for 5 min with gentle stirring immediately before use. Each metal chloride solution (typically 10 mM) in a 250-l injection syringe was titrated into a given polyphosphate solution (typically 0.5 mM) placed in the sample cell of 1.56-ml volume. The reference cell, which acts only as a thermal reference to the sample cell, was filled with water. Titrations typically consisted of a preliminary 1-l injection followed by 25 10-l injections of 20 s duration each, with 3 min spacing between injections. Other titrations consisted of a preliminary 1-l injection followed by 50 5-l injections of 10 s duration each, with 3 min spacing between injections. All experiments were performed at 20°C. Data from the first 1-l injection, which were unreliable due to diffusion between the syringe and cell solutions, were discarded before analysis.
The raw data correspond to the power required to be supplied by the calorimeter to maintain a constant temperature during and following injection. Integration of power with respect to time gives the enthalpy change as a result of injection. Prior to analysis, data were corrected for the heat of dilution of the metal chlorides into the Tris buffer without polyphosphates. Binding isotherms thus obtained were fitted to the Origin models by Marquardt nonlinear least squares analysis (Origin 5.0, Microcal).
Nuclear Magnetic Resonance-1 H NMR spectra were measured at 400 MHz in 20 mM Tris, pH 8.0, at 20°C in 80% D 2 O.
Circular Dichroism-CD scans were carried out on a Jasco J-715 CD spectrometer in a 0.5-cm path length cuvette between 320 and 220 nm, a 5 nm bandwidth, 10 mdeg sensitivity, and a 4-s response at 10 nm/min. Measurements were carried out in 20 mM Tris, pH 8.0, at 20°C. Volume changes on titration were less than 1%.

RESULTS
Calorimetric Titration Curves-We carried out ITC experiments at pH 8.0, in 20 mM Tris buffer at 20°C. The binding of divalent metal cations to nucleoside polyphosphates was generally found to be endothermic, the upward peaks indicating heat uptake on injection of divalent cation solution into a given polyphosphate solution. A typical binding profile is shown for Mg 2ϩ binding to ATP (Fig. 1). The binding of Mn 2ϩ and Ca 2ϩ to ATP gave similar results. Titration analyses of Ni 2ϩ , Cd 2ϩ , and Zn 2ϩ binding to ATP showed not only this primary endothermic binding event but a second, weaker exothermic binding event (Fig. 2). We presume the stronger endothermic binding event correlates with the interaction between metal cations and phosphate chain, and the weaker exothermic event with the interaction between metal cations and the N-7 atom of the adenine ring. The exothermic association of metal ions with the N-7 atom of ATP has been reported previously (42).
The binding of cations to Ap 4 A was found to be weaker than for ATP, but general trends were similar although weaker exothermic binding events were not observed. Titration experiments involving Mg 2ϩ , Ca 2ϩ , Mn 2ϩ , Ni 2ϩ , and Cd 2ϩ all showed a single endothermic binding event (Fig. 3). In contrast, Zn 2ϩ titrations revealed two binding events, both endothermic in character (Fig. 4). The strength of binding (K a in range 10 5 M Ϫ1 ) associated with the second event correlates with the interaction between metal cations and phosphate chain (as described above) (Table II). The strength of binding (K a in range 10 7 M Ϫ1 ) of the first event is altogether stronger and was thought to relate to Zn 2ϩ binding to alternative sites such as the N-7 atoms of the two adenine rings.
Metal cation binding to Ap 3 A was similar to Ap 4 A if generally weaker, and in the case of Ni 2ϩ no appreciable binding was seen to Ap 3 A at all. The interaction of Zn 2ϩ with Ap 3 A showed a very similar pattern to Ap 4 A. Titration with Zn 2ϩ revealed two endothermic binding events, the first being stronger than the second. By extrapolation from Ap 4 A, the first event was thought to correlate with Zn 2ϩ /N-7 atom interactions and the second to Zn 2ϩ /phosphate chain interactions. Whereas a single metal cation was clearly shown to bind to the phosphate chain of Ap 3 A, Ap 4 A, and ATP, two such cations were found to associate with the phosphate chain of Ap 5 A. However, once again, titrations with Zn 2ϩ revealed a very strong endothermic binding event preceding the second endothermic event characteristic of Zn 2ϩ binding with the phosphate chain (Fig. 5). All thermodynamic data for these titration experiments is shown in Table II.
NMR and CD Spectroscopy-The effects of Mg 2ϩ and Zn 2ϩ on Ap 4 A were compared by 1 H NMR and CD spectroscopy to obtain more evidence to support N-7 atom interactions. Fig. 6 shows the NMR spectrum of unliganded Ap 4 A between 5 and 8 ppm; the signal at 8.37 ppm corresponds to H-8, at 8.12 to H-2, and the doublet at 6.02 to H-1Ј (37). The effect of Zn 2ϩ and Mg 2ϩ on the chemical shifts of these signals is shown in Table  III. Both Mg 2ϩ and Zn 2ϩ caused small changes to the shifts of the H-2 and H-1Ј protons, consistent with minor changes in proton environment as a result of metal cation binding to the phosphate chain. However, the presence of Zn 2ϩ , unlike Mg 2ϩ , also provoked a downfield shift of 0.08 ppm in the chemical shift of H-8. Such a shift may be caused by either the withdrawal of electron density from the purine ring and/or partial base destacking. Base stacking is well known to induce upfield chemical shifts of the H-8 proton relative to an unstacked ATP standard (43). In either event, both effects are consistent with Zn 2ϩ forming an N-7 bridging interaction between two adenine rings analogous to the N-7 to N-7 chelate interaction reported for cis-platin (41). The NMR spectrum of Ap 4 A suggested that the binding of metal cation occured without loss in molecular symmetry.
The effects of Mg 2ϩ and Zn 2ϩ on the CD spectrum of Ap 4 A are shown in Fig. 7 and are in agreement with previous work (35). Unliganded Ap 4 A showed a reverse Cotton effect centered on 266 nm, with a peak at 254 nm and a trough at 279 nm. This is indicative of the reverse base-stacking characteristic of the diadenosine polyphosphates. The binding of Mg 2ϩ caused an approximate 15% reduction in the Cotton effect, and a slight red shift in the center of absorption. The binding of Zn 2ϩ caused a more radical change in the overall symmetry of the Cotton effect and a greater reduction in the trough at 279 nm, characteristic of partial base destacking, entirely consistent with the interpretation of the NMR data (see above).
The potential reversibility (switch characteristics) of Zn 2ϩ binding was investigated by a CD titration binding experiment. Ap 4 A was titrated with Zn 2ϩ , and changes in the depth of the trough at 279 nm were observed as a function of Zn 2ϩ concentration (Fig. 8). The concentration of the polyphosphate was significantly above the dissociation constant; therefore, simple saturation thermodynamics were observed. These titration results compare directly with the results of the ITC experiment with Zn 2ϩ and Ap 4 A, shown in Fig. 4. The first 22% reduction in ellipticity appears to correlate well with the first endothermic binding event and the second 13% reduction with the second endothermic binding event. We have argued above that the first endothermic binding event corresponds to Zn 2ϩ forming an N-7 bridging interaction between two adenine rings. Such an interaction should be accompanied by extensive base destacking (see above) consistent with the 22% reduction in ellipticity during the first stage of the CD titration. The second endothermic binding event has been ascribed to interactions between Zn 2ϩ and phosphate chain, interactions that should disrupt base stacking less, therefore consistent with the much smaller 13% reduction in ellipticity during the second stage of the CD titration. The results of titrations with Mg 2ϩ and Ap 4 A support this interpretation. ITC experiments showed a single endothermic binding event presumed to correlate with Mg 2ϩ / phosphate chain interactions, whereas a corresponding 13% reduction in Ap 4 A ellipticity at 279 nm was observed as a result of Mg 2ϩ in a CD titration experiment. Therefore, a 13% reduction in Ap 4 A ellipticity does indeed appear to correlate with binding interactions between metal cations and phosphate chain.    Table III. The ellipticity changes induced by Zn 2ϩ were completely reversed by EDTA titration (Fig. 8). Stepwise sequestration of Zn 2ϩ occurred with complete recovery of ellipticity at titration end point, consistent with complete regeneration of unliganded Ap 4 A. Hence, the association of Zn 2ϩ with Ap 4 A, and by extrapolation with other diadenosine polyphosphates, is completely reversible as are the changes in the physical state of Ap 4 A, provided that Zn 2ϩ levels may be adequately controlled. DISCUSSION We have characterized the binding of a variety of metal ions to Ap 3 A, Ap 4 A, Ap 5 A, and ATP by ITC. The advantage of this technique is that the enthalpy change ⌬H, the binding constant, and the stoichiometry may be calculated in a single experiment. From these values, the Gibbs free energy change, ⌬G, and the entropy change, ⌬S, may also be calculated, and hence all thermodynamic parameters may be defined.
As a control, the binding of metal cations to ATP was studied by using titration calorimetry as reported previously (44). All the divalent cations displayed a primary endothermic binding event, whereas the heavier cations also showed a secondary exothermic binding event. Such interactions are well documented with ATP (42). The primary binding event may be attributed to binding of the divalent metal cation to the ␤-␥phosphates, and the secondary event may be attributed to the interaction with N-7 of the adenine base.
Titrations of diadenosine polyphosphates with most metal cations revealed one main endothermic binding event whose strength declined in the order Ap 5 A (strongest) Ͼ ATP Ͼ Ap 4 A Ͼ Ap 3 A (weakest). This appears to correlate with metal cation/phosphate chain interactions as seen with ATP. No secondary exothermic binding events were observed in contrast to ATP. Ap 5 A appears to bind two metal ions on the phosphate chain, unlike Ap 4 A, Ap 3 A, and ATP, which otherwise appear to bind only one. Zn 2ϩ was found to bind exceptionally with all the diadenosine polyphosphates studied. In all cases, titrations of diadenosine polyphosphates with Zn 2ϩ revealed two endothermic binding events, the first strong and the second weak. All the data interlock to support the view that the second binding event corresponds to routine Zn 2ϩ /phosphate chain interactions, whereas the first relates to some form of N-7 bridging interaction between two adenine rings, an interaction that leads to significant base destacking (Fig. 9).
Although this binding interaction is strong, it is also clearly reversible provided that strong Zn 2ϩ -chelating agents are available. Such a reversible, strong binding interaction between Zn 2ϩ and the diadenosine polyphosphates may well explain many of the effects of Zn 2ϩ on the Ap n A-metabolizing enzymes shown in Table I. N-7-N-7 bridging interactions by Zn 2ϩ involved in an enzyme-bound adenylate could easily provide the necessary means to lower the energy of the transition state leading to Ap n A synthesis, thereby offering a clear explanation for the ability of Zn 2ϩ to enhance the catalytic rate of these anabolic enzymes. Indeed, observation of the enzymebound structure of lysyl adenylate reveals the adenine N-7 directed toward a possible Zn 2ϩ -binding site that might bridge to a second adenine during Ap 4 A synthesis (45). Conversely, Zn 2ϩ may well inhibit catabolic Ap n A hydrolase enzymes because Zn 2ϩ -Ap n A complexes are so stable as to be intractable substrates. For instance, base destacking and other conformational distortions caused by Zn 2ϩ binding may render Ap n As

Zinc as a Switch in Diadenosine Polyphosphates
unrecognizable as Ap n A hydrolase substrates. Alternatively, Ap n A hydrolases may be able to bind Zn 2ϩ -Ap n A complexes in the active site but lack the necessary binding energy to overcome the conformational effects of the Zn 2ϩ and achieve Ap n A conformations appropriate for hydrolysis to take place.
The total zinc concentration of Escherichia coli has recently been reported as 0.2 mM; however, there is no cytosolic pool of Zn 2ϩ as all Zn 2ϩ is bound to molecules within the cell (46). Such over-capacity for the binding of Zn 2ϩ suggests tight control of Zn 2ϩ levels; it is unlikely that cellular zinc chemistry is under simple thermodynamic control. It may be possible that the diadenosine polyphosphates act in a protective role to sequester zinc under stress conditions. Metallothioneins, which act as the major intracellular store of Zn 2ϩ , have been implicated in the protection of cells from stress (47), yet the exact roles of this class of proteins remain unclear (48). It is possible that there is biochemical cross-talk between the metallothionein and diadenosine polyphosphate stress responses, whereby Zn 2ϩ release from the metallothioneins under stress leads to diadenosine polyphosphate synthesis and concomitant inhibition of diadenosine polyphosphate hydrolysis. The resultant increased diadenosine polyphosphate concentrations may then be involved in adaptation to stress by binding to several heat shock proteins including GroEL, DnaK, and ClpB (49,50). After stress has passed, diadenosine polyphosphate-accessible Zn 2ϩ will return to normal levels, and the diadenosine polyphosphate hydrolases may be reactivated. Such biochemical cross-talk may also occur extracellularly with regard to the catabolic enzymes (51). In this regard, we propose Zn 2ϩ as a conserved atomic switch controlling synthesis and biological activity of diadenosine polyphosphates.