Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

ATP and related adenine nucleotides present in cardiac cells have an important function as regulators of RyR channel gating. ATP can induce long open events and a high open probability (P_o) in the presence of micromolar cytosolic Ca²⁺ (Kermode et al., 1998). High levels of adenine nucleotides and Ca²⁺ tend to inactivate the channel (Kermode et al., 1998; Ching et al., 1999); therefore, the effects of this group of ligands are complex and are likely to shape both the activation and inactivation processes of intracellular Ca²⁺ release in cardiac cells. Understanding the underlying molecular nature of the binding of adenine nucleotides to RyR channels and the structural features of ATP that produce open RyR channels is therefore important for a greater understanding of how RyR channels are regulated during the process of excitation-contraction coupling. In a preliminary CoMFA study, we previously correlated the structure of ATP and other adenine-based ligands with the ability to modulate the gating of native sheep cardiac RyR incorporated into planar phospholipid bilayers (Chan et al., 2000). With millimolar luminal [Ca²⁺] and a maintained cytosolic [Ca²⁺] of 10 µM, ATP induced P_o levels of approximately 0.9. The maximum P_o levels produced by the other adenine-based ligands investigated (ADP, AMP, adenosine, adenine) were much lower. CoMFA demonstrated a high correlation between ligand structure and maximum P_o induced. Although it is now well established that the purine ring is important for agonist activity (Morii and Tonomura, 1983; Meissner, 1984; Chan et al., 2000), we demonstrated that the phosphate groups are essential to the high efficacy of ATP. Our model predicts that the charge produced by the phosphate groups is the single most important factor that enables bound ATP molecules to activate the cardiac RyR channel to a greater extent than ligands with fewer phosphate groups (for example, ADP and AMP). To investigate the significance of the charge on the phosphate groups of ATP, we examined how reducing charge in this region would affect the gating of RyR channels. AMP-PCP is a nonhydrolysable analog of ATP in which a methyl group substitutes for the oxygen between the - and -phosphate groups. The methyl substitution leads to a decrease in negative charge around the phosphate groups. We would predict a diminished ability of such a ligand to fully open the channel in the presence of 10 µM cytosolic Ca²⁺. This prediction, however, runs counter to accepted ideas about the effectiveness of AMP-PCP as an activator of RyR channels in that it is reported to be at least as effective as ATP (Smith et al., 1985). In fact, the stable nature of AMP-PCP has led to the routine use of this analog as a replacement for ATP in functional studies of RyR (for example, see Xu et al., 1996; Fruen et al., 2002).

Our single-channel experiments show that AMP-PCP is, as predicted from our initial model, a partial agonist at the ATP sites on the cardiac RyR, thus confirming the validity of our model. Importantly, we also show that AMP-PCP has a higher affinity for RyR than ATP, demonstrating that the affinity and efficacy of ligands at the ATP sites on RyR are dependent on different structural characteristics. Using the new information obtained from observing the effects of AMP-PCP on channel gating, we now present a more detailed description of the correlations between adenine nucleotide structure and the ability to open (efficacy) the sheep cardiac RyR channel. For the first time, we describe the structural characteristics of adenine nucleotides that are associated with high affinity for the ATP sites on the cardiac RyR. Electrostatic interactions are critically involved in the process of nucleotide binding to RyR, although steric factors also play a role. Use of comparative similarity index analysis (CoMSIA) demonstrates that although changes in hydrogen-bond acceptor ability of the adenine nucleotides are strongly correlated with changes in affinity of the ligand for RyR, it does not seem to influence the subsequent ability of the ligand to increase P_o.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Preparation of SR Membrane Vesicles and Planar Lipid Bilayer Methods. Heavy SR membrane vesicles were prepared from sheep hearts as described previously (Sitsapesan et al., 1991) and rapidly frozen and stored in liquid nitrogen. Vesicles were fused with planar phosphatidylethanolamine lipid bilayers as described previously (Sitsapesan et al., 1991). The vesicles were fused in a fixed orientation such that the cis chamber corresponded to the cytosolic space and the trans chamber to the SR lumen. The trans chamber was held at ground and the cis chamber held at potentials relative to ground. After fusion, the cis chamber was perfused with 250 mM HEPES, 125 mM Tris, and 10 µM free Ca²⁺ buffered with EGTA and CaCl₂, pH 7.2. The trans chamber was perfused with 250 mM glutamic acid and 10 mM HEPES and the pH was brought to 7.2 with Ca(OH)₂ (free [Ca²⁺], ~50 mM). Experiments were performed at room temperature (22 ± 2°C). The free [Ca²⁺] and pH of the solutions were determined at 22°C using a calcium plastic membrane half-cell (93-20; Thermo Orion, Beverly, MA) and epoxy body ROSS combination pH electrode (81-55; Thermo Orion) as described previously (Sitsapesan et al., 1991). Additions of ATP and AMP-PCP were made to the cis chamber and the free [Ca²⁺] was maintained at 10 µM at all ligand concentrations throughout all experiments.

Data Acquisition and Analysis. Single channel recordings were displayed on an oscilloscope and recorded on digital audio tape. Steady-state recordings were carried out at 0 mV. At this holding potential, Ca²⁺ current flows in the lumen-to-cytosol direction. The current recordings were filtered at 0.5 kHz (3 dB) and digitized at 2 kHz using Satori (Intracel, Cambridge). Channel open probability (P_o) and the lifetimes of the open and closed events were determined over 3 min of recording using the method of 50% threshold analysis (Colquhoun and Sigworth, 1983). When more than one channel was incorporated into the bilayer, average P_o was calculated according to the formula: Average P_o = [T_open1 + 2(T_open2) + 3(T_open3)... . + N(T_{open n})]/NT_total where T_open1, T_open2, and T_open3 are the times in the first, second, and third open channel levels, respectively, T_total is the total recording time, and N is the number of channels in the bilayer. The number of channels in the bilayer was determined by adding EMD 41000 (a caffeine analog that maximally activates the channels) at the end of each experiment. Lifetime analysis was carried out only when a single channel incorporated into the bilayer. Events <1 ms in duration were not fully resolved and were excluded from lifetime analysis. Lifetimes accumulated from 3-min steady-state recordings were stored in sequential files and displayed in noncumulative histograms. Individual lifetimes were fitted to a probability density function by the method of maximum likelihood (Colquhoun and Sigworth, 1983) according to the equation: f(t) = a₁(1/₁)exp(t/₁) +... + a_n(1/_n)exp(t/_n) with areas a and time constants . A missed-events correction was applied as described by Colquhoun and Sigworth (1983). A likelihood ratio test (Blatz and Magleby, 1986) was used to compare fits to up to four exponentials by testing twice the difference in log_e (likelihood) against the ² distribution at the 1% level. Single-channel current amplitudes were measured from digitized data using manually controlled cursors. All P_o values are quoted as mean ± S.E.M., where n  4. For n = 3, S.D. is given. The AMP-PCP dependence on P_o was characterized using the equation: P_o = P_max(1/(1 + EC₅₀/[AMP-PCP]^n_H1)) (1  (1 + (IC₅₀/[AMP-PCP])^n_H2)) where P_max equals the maximum P_o attained, EC₅₀ and IC₅₀ equal the half-maximum concentrations required for activation and inhibition, respectively, and n_H1 and n_H2 are the Hill coefficients of activation and inhibition, respectively.

CoMFA. A CoMFA is a mathematical expression of the correlation between the chemical structure of a collection of compounds and the experimentally determined biological activities of these compounds. (Cramer et al., 1988). Briefly, the Lennard-Jones (steric) potentials and electrostatic potentials are measured at the points in a three-dimensional array around each member of a series of compounds. Changes in these potentials are correlated with changes in the biological properties of the molecules using the method of partial least squares. We used SYBYL 6.6 (Tripos Associates, St. Louis, MO) and the Tripos force field for all of the calculations reported in this communication. The structures of ATP and analogs were correlated with the ability of these molecules to maximally activate the sheep cardiac RyR in the presence of 10 µM cytosolic Ca²⁺ and with the EC₅₀ values for each compound. Briefly, the conformation of ATP was determined by extended molecular dynamics of ATP in adenylate kinase (Protein Data Bank Accession code, 1AKY (Abele and Schulz, 1995)). The alignments of the chemical structures are slightly changed from the previous report. The other compounds were built on the ATP framework and aligned to ATP by rigid-body field fitting. Cross-validation was used to test for predictive ability of the model. A cross-validated correlation coefficient of 1.0 indicates no deviation between predicted and measured biological properties; a value of 0.0 indicates a random correlation between structure and biological activity. A cross-validated correlation coefficient of 0.3 indicates a probability of less than 5% that the relationship between structure and activity is attributable to chance (Clark et al., 1990; Clark and Cramer, 1993). If the cross-validated correlation coefficients are sufficiently high, partial least-squares is used to build a model using all of the tested compounds. This final correlation coefficient can be interpreted analogously to a conventional correlation coefficient.

CoMSIA. CoMSIA, like CoMFA, is a mathematical expression correlating the chemical structures and biological activities of a set of compounds (Klebe, 1998). In CoMSIA, changes in physicochemical properties, such as hydrophobicity or hydrogen bond potential, can be compared with changes in biological activity.

Materials. ATP, adenosine, and adenine were 99% pure (Sigma, Poole, UK). EMD 41000 was a gift from Merck (64271 Darmstadt, Germany). Solutions were prepared using MilliQ de-ionized water (Millipore, Harrow, UK) and filtered through a Millipore membrane filter (pore size, 0.45 µm) before use. Other chemicals were AnalaR or the best equivalent grade from BDH or Sigma (Poole, UK).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

The Effects of AMP-PCP on RyR Channel Gating. Figure 1 illustrates how AMP-PCP activates single sheep cardiac RyR channels incorporated into planar lipid bilayers. RyR channel openings are typically brief when activated solely by 10 µM cytosolic Ca²⁺ as demonstrated in the top trace. AMP-PCP caused a dose-dependent increase in P_o, but even the most effective concentrations (500 µM-2 mM) could not fully activate the channels. Inspection of the traces shows that AMP-PCP produced large increases in the frequency of channel opening. At the maximum level of activation (illustrated in Fig. 1 and occurring at approximately 1 mM AMP-PCP), very brief open and closed events occurred, giving rise to a flickery appearance of gating. This flickery type of gating is characteristic of agents that activate the channel via the ATP sites (Kermode et al., 1998). ATP, itself, also causes flickery gating but at the most effective concentrations produces long events in addition to bursts of short open and closed events. Equivalent long open events were not observed with AMP-PCP under the controlled cytosolic [Ca²⁺] and pH of the experiments. Figure 2 illustrates the typical gating behavior of a sheep cardiac RyR channel activated by 1 mM ATP. Consecutive single-channel recordings demonstrate the spontaneous changes in gating from high P_o levels characterized by long open times to lower P_o levels characterized by brief open and closed events. Over 3 min of recording, P_o averaged 0.89. This is approximately double the average maximum P_o obtained with AMP-PCP under the same experimental conditions (see Fig. 1). No measurable change in single-channel conductance was observed after addition of AMP-PCP to the cytosolic chamber (results not shown).

View larger version (44K): [in this window] [in a new window]   Fig. 1.   Effects of AMP-PCP on current fluctuations through a typical single sheep cardiac RyR. A, the channel is activated by 10 µM cytosolic Ca2+. B, 1 mM AMP-PCP has been added to the cis chamber. The free cytosolic [Ca2+] is maintained at 10 µM. O and C indicate the open and closed channel levels, respectively. The figure illustrates an optimal concentration of AMP-PCP under these experimental conditions.

View larger version (40K): [in this window] [in a new window]   Fig. 2.   Effects of ATP on RyR channel gating. The variation in gating observed over time in a typical single cardiac RyR channel activated by 10 µM cytosolic Ca2+ alone and 1 mM ATP (in the presence of 10 µM free cytosolic Ca2+) is shown. Sequential single-channel recordings of the channel are shown. Measurement of Po over 3 min of continuous recording gives an average Po of 0.89. ATP elicits long open events, as shown in traces A and B, but the channel can also revert to flickery gating characterized by brief opening and closing events as shown in traces C and D. In trace E, some flickery gating can be observed interspersed with longer open events. O and C indicate the open and closed channel levels, respectively.

View larger version (13K): [in this window] [in a new window]   Fig. 3.   Relationship between Po and AMP-PCP concentration in the presence of 10 µM free cytosolic Ca2+. S.E.M. for n  4 is shown.

How Does AMP-PCP Activation of RyR Compare with the Effects of Other Adenine Nucleotides? In Fig. 4, we compare the affinity, efficacy, and Hill coefficient for the activation of the cardiac RyR by AMP-PCP and other adenine nucleotides (Chan et al., 2000). The properties of AMP seem to deviate significantly from those of the other ligands shown. AMP has a very low affinity for RyR, a low ability to increase P_o, and also activates the channel with a Hill coefficient <1. ADP, although it has a low affinity and shows no evidence for a positively co-operative action, still has a greater ability to open the channel if present in sufficient quantities (10 mM). In the presence of 10 µM cytosolic Ca²⁺, AMP-PCP is not able to produce the same high level of activation as ATP because it does not induce long enough open events. AMP-PCP, however, has the highest affinity for RyR and the highest Hill coefficient of all the adenine nucleotides studied.

View larger version (24K): [in this window] [in a new window]   Fig. 4.   Comparison of the actions of AMP-PCP with ATP, ADP, AMP, adenosine, and adenine on cardiac RyR channel gating in the presence of 10 µM free cytosolic Ca2+. Top, maximum Po that can be achieved with each ligand. Center, comparison of EC50 values. Bottom, Hill coefficient for RyR channel activation for each ligand. The dose ranges used to obtain the above information were: AMP-PCP, 10 µM-10 mM; ATP, 10 µM-20 mM; ADP, 100 µM-50 mM; AMP, 100 µM-10 mM; adenosine, 10 µM-5 mM; adenine, 20 µM-5 mM. For AMP-PCP, ATP, ADP, AMP, adenosine, and adenine, maximum Po was observed at 1, 2, 20, 20, 1, and 2 mM, respectively.

CoMFA of P_o. The CoMFA values of P_o described in this article are similar to those obtained earlier with the more limited data set (Chan et al., 2000). In the previous report, electrostatic factors accounted for 64% of the observed correlation between structure and P_o. Steric factors accounted for the remaining 36%. In this report, we have used P_o to calculate the increased free energy required to raise the P_o above that seen in the presence of Ca²⁺ and zero ATP analog. This is calculated as G = (1.987 cal/mol K) (296 K) [ln((P_o in the presence of ATP analog)/(P_o in the presence of Ca²⁺ only)). The bilayer experiments were done at 23°C (296 K). The cross-validated correlation coefficient (a measure of predictive ability) for P_o is 0.307 with a final correlation coefficient (r²) of 0.994.

View larger version (42K): [in this window] [in a new window]   Fig. 5.   CoMFA of Po. The wire frames show the regions in space where changes in structure are most strongly correlated with changes in Po. A ball-and-stick representation of ATP is shown as a guide to the eye. Top, gray contours illustrate the regions where increased negative charge is correlated with increased Po; black contours illustrate the regions where increased positive charge is correlated with increased Po. The black contours reflect the increasing Po because the highly charged terminal phosphate is located further from the adenyl group. Bottom, gray contours detail where increased steric bulk is correlated with increased Po, and black contours show where increased steric bulk is correlated with decreased Po

CoMSIA of P_o. Using CoMSIA, no correlations were found between changes in steric bulk, hydrogen bonding, or hydrophobicity and the ability to modulate P_o. None of these properties, when tested alone, was sufficient to explain the differences in biological activities. The lack of correlation was seen regardless of inclusion of AMP-PCP in the test group of compounds. Only electrostatic CoMSIA showed strong correlations between changes in structure and changes in P_o (q² = 0.624, r² = 0.992). Of the properties tested, only the electrostatic field is sufficient by itself to predict P_o.

CoMFA of EC₅₀. We have expressed EC₅₀ values as the apparent free energy of binding. This is calculated as G = (1.987 cal/mol K)(296 K)[ln (EC₅₀)]. The cross-validated correlation coefficient (q²) for EC₅₀ is 0.583 (r² = 1.00). As for P_o, most of the correlation (66%) is between changes in electrostatic field and changes in EC₅₀. The remainder of the correlation is with changes in steric bulk. Figure 6 illustrates the locations in which changes in structure are most highly correlated with changes in EC₅₀, and compares the regions in which steric and electrostatic factors play the greatest role.

View larger version (45K): [in this window] [in a new window]   Fig. 6.   CoMFA of EC50. Regions in space where changes in structure are most strongly correlated with changes in affinity are shown. Top, gray contours illustrate the regions where increased negative charge is correlated with a decrease in EC50, and black contours show the regions where increased positive charge is correlated with decreased EC50. Bottom, gray contours detail where increased steric bulk is correlated with a decrease in EC50, and black contours show where increased steric bulk is correlated with increased EC50.

CoMSIA of EC₅₀. In contrast to the correlation with P_o, additional analysis by CoMSIA found a correlation between hydrogen-bond acceptor ability and changes in EC₅₀ (cross-validated correlation coefficient, 0.330; final correlation coefficient, 1.0). Figure 7 details the regions in which hydrogen-bond accepting properties are correlated most strongly with the affinity of the adenine nucleotides for RyR. As was the case with P_o, a correlation was seen with changes in electrostatic CoMSIA but not hydrogen-bond donating, hydrophobic potential, or steric bulk. The results demonstrate that at least part of the difference between the ligand-receptor interactions leading to changes in P_o and EC₅₀ is the importance of hydrogen-bond accepting ability of the modulators.

View larger version (34K): [in this window] [in a new window]   Fig. 7.   CoMSIA of hydrogen bond acceptor ability and EC50. The wire frames show the regions in space where changes in structure are most strongly correlated with changes in hydrogen bond acceptor ability. Hydrogen bond acceptor potential is itself sufficient to make a reasonably reliable prediction of the EC50 of the compounds tested. The black wire frames are those areas in which increasing hydrogen bond accepting potential is highly correlated with decreasing EC50. The gray wire frames are those areas in which increasing hydrogen bond accepting potential is highly associated with increasing EC50. The position of the contours reflects the decrease in EC50 as hydrogen bond acceptors are moved outward from the nucleoside.

Structure-Activity Relationships and Energetic Considerations. By correlating the single-channel effects of ATP and related ligands with their structural characteristics, we can distinguish between the structural properties required to produce strong binding to RyR and those required to drive the channel into the open state. Interactions between a ligand and a receptor can be divided into two broad types: A) those that exist in the ligand-receptor complex only and B) those that relate the relative solvation of the ligand by the receptor to that of the bulk medium (including counter ions, detergents, and so forth). In this report, the properties of AMP-PCP as an agonist at ATP sites on RyR have been analyzed in the context of other nucleotides and analogs. The properties compared include the maximum P_o, the concentration of ligand required to produce half-maximum P_o (EC₅₀), the Hill coefficient, and the mean open and closed lifetimes. Of these, the P_o and any altered open and closed life times are properties of the ligand-receptor complex only. These quantities are often obtained at saturating levels of ligand (e.g., P_o). They depend not on affinity of ligand for the binding site but on complementary interactions in the ligand-receptor complex that can provide energy to alter the equilibrium between open and closed states or to lower the energy barrier between open and closed states. In the following analysis, the focus will be on the energetic relationships measured by P_o and EC₅₀.

View this table: [in this window] [in a new window]   TABLE 1 Anatomy of the ATP-RyR interaction: distribution energy calculated from the experimental EC50 and the maximum increase in Po The imaginary compounds are indicated by the missing parts of the linear formula. The first row is the complete ATP molecule. Row two contains an imaginary molecule obtained by omitting the terminal ( ) PO3; the atoms remaining have exactly the same position and charge as in ATP. Row 2 is similar, but not identical to, the ADP used in the experimental data. Row 3 is obtained from ATP (row 1) by removal of the  PO2 group. This is similar (but not identical) to an AMP plus a phosphate. In row 4, the  PO2 group is removed. This is similar to an adenosine plus a pyrophosphate. In row 5, the ribose is omitted. In row 6, the adenine base is omitted. The values in columns 2 and 3 are calculated with the indicated atoms missing while the remaining atoms retain the charge, position, and other properties found in ATP. The experimentally determined values are shown in parentheses for comparison with the predicted values of ATP.

View this table: [in this window] [in a new window]   TABLE 2 Comparison of free energy of binding to free energy used to increase Po The free energies were calculated from the experimentally observed increase in Po (G column) or from the experimentally measured EC50 for increased Po (G column). See the text for the formulae used. The total binding energy is estimated by summing the G and G columns. This calculation neglects the contribution of binding energy to lowering transition state energies and interaction energies manifested in the Hill coefficient. The percentage of binding energy converted to increasing the Po is estimated by dividing the value in the G column by the value in the G(total) column and multiplying by 100.

Distribution of Energies Responsible for ATP Modulation of P_o and EC₅₀. Addition of ligands such as ATP increases the value of P_o above that seen with Ca²⁺ alone. This difference in P_o can be related to the additional free energy (G) required to push the equilibrium toward the open states of the RyR. In the case of ATP, the shift in P_o requires the input of 3459 cal/mol of free energy (Table 1). We used CoMFA to show quantitatively the distribution of this energy among the components of the most effective ligand, ATP. To do this, we have created imaginary molecules by removing fragments of the ATP and used CoMFA to predict the properties of these constructs. In the first column of Table 1, the missing elements are indicated by a space. The indicated fragments of the molecule remaining are left with the same charge and atomic positions as the intact ATP. In columns 2 and 3, we have calculated the free energies of these virtual molecules.

Conversion of Binding Energy to Modulation Energy. The total binding energy of ATP (Table 2, column 4; 8415 cal/mol) can be estimated from the sum of the experimental energies from the EC₅₀ (4956 cal/mol) and the incremental increase in the P_o (3459 cal/mol, see Table 2). Forty percent of the total binding energy is used to modulate channel function. In the case of ADP, the total binding energy is 6276 cal/mol as estimated from the sum of the experimental energies from the EC₅₀ (3958 cal/mol) and the incremental increase in P_o (2318 cal/mol). Again, about 40% of the binding energy is used to increase the open probability of the channel. Therefore, the difference between the P_o induced by ATP and ADP seems to be caused by the difference in total energy available.

Interaction Energy. As stated previously, the energy to modulate P_o and mean open and closed times comes from the energy of nucleotide-receptor interactions. Unfortunately, at this time, no atomic level model of the nucleotide-binding site of the RyR exists. However, nucleotide-binding sites are generally considered to have a common structural motif. To investigate the effects of nucleotide analog structure on ligand-receptor interactions, we used molecular dynamics to estimate changes in interaction energy using adenylate kinase (Protein Data Base code 1AKY) as a surrogate for the nucleotide-binding site on the RyR. The interaction energy of ADP is 80% of that of ATP (molecular dynamics data not shown). This value compares well with those in the paragraph above. The incremental increase in P_o induced by ADP (2318 cal/mol) is 67% of that of ATP (Table 2, column 2). The total binding energy of ADP (6276 cal/mol) is 74% of that of ATP (Table 2, column 4). The binding energy of ADP estimated from the EC₅₀ value (3958 cal/mol) is 80% of that of ATP (Table 2, column 3). In the present study, therefore, a surrogate nucleotide-binding site reasonably parallels the experimentally observed nucleotide-RyR interactions.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Our preliminary CoMFA model (Chan et al., 2000) predicted that, in the presence of 10 µM cytosolic Ca²⁺, AMP-PCP would be significantly less effective than ATP as an activator of the cardiac RyR channel. Our results demonstrated that the maximum P_o achieved by AMP-PCP was approximately half that obtained with ATP (Figs. 1 and 3). Although these results provided conformation of the validity of our model, they were still surprising in light of previous reports of the effectiveness of AMP-PCP to activate RyR channels. The literature contains no hint that AMP-PCP would lack efficacy, in fact quite the reverse. There are, however, a number of possible reasons to explain these discrepancies. First, in early reports, only a single dose of AMP-PCP had been compared with a single dose of ATP (Smith et al., 1985) and subsequent investigators did not examine the full dose-response relationship of these ligands. Second, because the effects of adenine nucleotides are strongly Ca²⁺-dependent, slight differences in free [Ca²⁺] could lead to large changes in AMP-PCP-induced effects. Third, it is possible that AMP-PCP has a higher efficacy at the skeletal isoform of RyR than at the cardiac isoform. Finally, we have demonstrated in the present study that AMP-PCP has a higher affinity for the cardiac RyR than ATP. The relatively high affinity of AMP-PCP for RyR will have masked its low efficacy in studies such as [³H]ryanodine binding studies or Ca²⁺ flux experiments. This type of experiment is often designed to produce the maximum possible measurable response; therefore, relatively low concentrations of a nucleotide can produce the maximum level of [³H]ryanodine binding or the maximum increase in the rate of Ca²⁺ efflux from SR membrane vesicles. Thus, partial agonists and full agonists will induce the same maximum level of [³H]ryanodine binding although a high-affinity ligand may seem to be more effective at opening RyR at low concentrations. It is impossible, therefore, to distinguish between the efficacy and affinity of a ligand under these conditions.

In this study, not only did we find that AMP-PCP has a higher affinity for the cardiac RyR than ATP, we also demonstrated that AMP-PCP activates the channel with a higher degree of positive cooperativity. The Hill coefficient for AMP-PCP was 2.1, whereas for ATP, the Hill coefficient was 1.5 (Kermode et al., 1998). An increase in cytosolic [Ca²⁺] produced a marked increase in the Hill slope for activation of the cardiac RyR by AMP (Ching et al., 1999). If this is true of adenine nucleotide channel activation in general, then the effects of AMP-PCP may be much more sensitive than ATP to changes in cytosolic [Ca²⁺] because of its apparently greater degree of positive cooperativity. These subtle differences in the characteristics of AMP-PCP and ATP may produce marked differences in RyR channel gating. Use of AMP-PCP as a nonhydrolysable "physiological replacement of ATP" may not, therefore, produce reliable results.

The structure-activity relationship reported here demonstrates that the interactions between ATP and related compounds are complex, consistent with a mechanism whereby at least part of the binding energy is converted into conformational changes in the receptor. Adenine triphosphate is the most effective ligand because, at least in part, it has the highest binding energy and because it converts the greatest percentage of its binding energy into increased P_o. Another interesting observation is the relationship of the nucleotide base with P_o, EC₅₀ and ligand specificity. The data in Table 1 imply that, in the context of ATP, the base contributes an unimportant amount of energy to the increase in P_o or to binding (EC₅₀), yet the RyR is specific for the adenine fragment. Replacement of the adenine with a guanine completely destroys the ability of the nucleotide to modulate RyR. This fact suggests a tight complementarity at the base sub-binding site such that the additional steric bulk (and/or polar interaction) at the 2-positon of the base blocks nucleotide binding. In addition, the unligated adenine binds with reasonable strength and promotes a significant increase in P_o (Table 2). We suggest that interactions between adenine and receptor are weakened in the context of ATP because of binding-induced conformational changes in receptor, ligand, or both. To maximize interactions between receptor and the triphosphate, interactions between receptor and base are relaxed (but not to the point of loss of specificity). We suggest that the ligands bound to the nucleotide site do not serve a permissive role (for example, shielding unfavorable interactions); instead, they are active participants in the functional state of the RyR and different ligands induce or stabilize unique conformers of the RyR. The conformers are manifested not in the magnitude of the conductance but in the equilibrium between open and close channel states and the energy barrier between them. To make an analogy with enzyme-catalyzed reactions, the bound ligand may stabilize one or more transition states between conducting and nonconducting forms of the channel and the ligand may bind somewhat more tightly to the conducting form of the RyR.

Hill coefficients derived from the experimental data indicate that there may be more that one nucleotide binding site per RyR channel. Because RyR is a homotetramer, there may be four nucleotide binding sites per tetramer. However, the three-dimensional structure-activity relationships described in the present study are based on EC₅₀ values and maximum P_o levels induced. The relationships are therefore not dependent upon assumptions about the number of binding sites or interactions between sites and cannot be used to explain any cooperativity observed. An important future investigation would be to examine the effect of nucleotide structure on the cooperativity of adenine-nucleotide-induced stimulation of P_o.

With increasing knowledge of the effects of adenine nucleotides on RyR gating, the more obvious the high efficacy of ATP becomes and the greater our certainty that the -phosphate group of ATP plays a unique role in allowing ATP to induce long open states and high P_o values. These results prompt the question of whether some of the effects of ATP could be caused by phosphorylation of the channel. Indeed, a recently published report suggests that part of the effects of ATP on the skeletal isoform of RyR results from phosphorylation of the channel by CaMKII (Dulhunty et al., 2001). However, we have no evidence that this is the case in our experiments. Unlike the report by Dulhunty et al. (2001), the effects of ATP that we observe are completely reversible; after 25-30 min in the presence of cytosolic ATP (1-5 mM), perfusion of the cytosolic chamber completely reverses the effects of ATP. Moreover, we have no evidence for any time-dependent increases in P_o, as would be expected to occur if phosphorylation of RyR were producing a change in gating. Compelling evidence that phosphorylation is not producing our ATP-dependent changes in gating comes from the CoMFA itself. Cross-validation of the molecules used in the CoMFA analysis demonstrates that ATP is not an outlier, as would be expected if some of the effects of ATP were caused by phosphorylation. We therefore have strong evidence that our reported effects of ATP on the gating of cardiac RyR channels are caused by ligand-receptor interactions that do not involve phosphorylation. Possibly, the differences between our results and those of Dulhunty et al. (2001) reflect differences in the effects of ATP on skeletal and cardiac channels. Alternatively, the differences may be caused by variations in the preparation of the heavy SR membrane vesicles, which could alter the phosphorylation state of RyR or the attachment of associated proteins to RyR.

In this study, the effects of adenine nucleotides have been examined in the presence of only one other modulator of RyR channel activity, cytosolic Ca²⁺, and the concentration of Ca²⁺ has been maintained at 10 µM so that the effects of the nucleotides rather than the effects of changes in [Ca²⁺] can be monitored. During EC-coupling, the RyR will experience cyclical changes in cytosolic [Ca²⁺] from 0.1 to >10 µM, so it is important that future studies investigate the ability of adenine nucleotides to modulate RyR activity at a range of cytosolic [Ca²⁺]. Moreover, approximately 0.5 mM Mg²⁺ is present in cardiac cells and it not known how the effects of the Mg²⁺-bound form of ATP differ from those of the Ca²⁺-bound form. Future experiments must address this by comparing the actions of nucleotides in the presence and absence of physiological levels of Mg²⁺, although this is complicated because Mg²⁺ itself also modulates RyR P_o directly.

In summary, we have demonstrated the reliability of our CoMFA model to predict the ability of ATP analogs to open the cardiac RyR channel. We also report on a new CoMFA model that describes the correlation between adenine nucleotide structure and affinity for RyR. The structural features responsible for high affinity differ from those responsible for high efficacy. The results of this study provide insight into the molecular mechanisms by which adenine nucleotides open the cardiac RyR and should provide a greater understanding of how adenine nucleotides regulate SR Ca²⁺ release in cardiac cells.