Introduction

Top Summary Introduction Procedures Results Discussion References

Within many subtypes of G protein-coupled receptors, molecular modeling studies predict that there is a high degree of amino acid identity around the primary neurotransmitter or hormone binding site (Trumpp-Kallmeyer et al., 1992). Thus, the development of drugs with subtype selectivity can be a daunting task, and within the muscarinic receptor field, for example, it has proved to be very difficult to design antagonists selective for one subtype over the other subtypes and so far impossible to discover subtype-selective agonists (Caulfield and Birdsall, 1998; Tucek and Proska, 1995). Within the adrenergic field, a limited number of compounds are known with 10-100-fold higher affinity for certain subtypes (Bylund et al., 1994). However, also in this field, there remains a need for the development of further and more subtype-selective drugs (Ruffolo et al., 1995).

It thus would be desirable to have an alternative target for the development of such drugs, and such an alternative could be an allosteric site. The potential usefulness of such a site for drug design is emphasized by the success of the benzodiazepines, which act at an allosteric site on the -aminobutyric acid_A receptors, enhancing the response to -aminobutyric acid (Barker et al., 1986).

A schematic representation of allosterism is shown in Fig. 1. In this model, the binding of an allosteric agent X to the allosteric site modulates the binding of a ligand L at the primary binding site. Depending on the magnitude and direction of the changes in the equilibrium and kinetic binding constants, the affinity of L at the primary site will either increase ( > 1, positive cooperativity), decrease ( < 1, negative cooperativity), or remain unchanged ( = 1, neutral cooperativity). This opens the possibility for the modulation of the action of the natural agonist of a receptor in a subtype-specific manner if, for example, the agonist activity is enhanced at one subtype (positive cooperativity) but inhibited at other subtypes (negative cooperativity) (Lazareno and Birdsall, 1995).

View larger version (13K): [in this window] [in a new window]   Fig. 1.   Schematic representation of the ternary complex allosteric model. In this scheme, radioligand L and allosteric agent X bind to two separate sites on the receptor R. KL and KX are the affinity constants for L and X, respectively;  is the cooperativity factor of X with L; and k1 and k2 are the dissociation rate constants for L dissociating from RL and XRL, respectively.

Allosteric interactions have been best characterized with the muscarinic family (Stockton et al., 1983; Lee and El-Fakahany, 1991a; Leppik et al., 1994; Lazareno and Birdsall, 1995; Tucek and Proska, 1995), and it is of interest to note that many of the somewhat subtype-selective compounds discovered to date have also been reported to act at the allosteric site. These include the m2 selective compounds gallamine (Stockton et al., 1983), himbacine (Lee and El-Fakahany, 1990; Matsui et al., 1995), and methoctramine (Lee and El-Fakahany, 1991b; Matsui et al., 1995), thus highlighting the potential value of the allosteric site in generating subtype selectivity. In regard to the enhancement of agonist activity, it has recently been found that the alkaloid brucine allosterically enhances the action of acetylcholine at only the m1 muscarinic receptor subtype, whereas N-chloromethyl-brucine allosterically enhances acetylcholine action at only the m3 muscarinic receptor subtype (Birdsall et al., 1997).

Although allosterism has been best characterized with the muscarinic receptors, it has been studied only to a limited extent at other G protein-coupled receptors. Therefore, it is not known whether allosterism is a general phenomenon of this class of receptors and hence of general value in drug discovery. The adrenergic receptors are another important therapeutic area, and it has been reported that amiloride analogues increase antagonist dissociation rates from the ₂-adrenergic (Howard et al., 1987, Jagadeesh et al., 1990), _2A-adrenergic (Nunnari et al., 1987), and _2B-adrenergic (Wilson et al., 1991) receptors. It was postulated that the amilorides were acting via an allosteric site. However, all of the above workers reported that the amilorides caused a decrease in the number of antagonist binding sites (B_max), a finding that is not compatible with the allosteric model.

Our work was undertaken to explore whether the modulation by amiloride analogues (Fig. 2) of antagonist binding at the human _2A-adrenergic receptor is compatible with the ternary complex allosteric model of Fig. 1. A preliminary account of some of these data has been published previously (Leppik et al., 1997).

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Structural formulae of amiloride and of its analogues examined in this report. BZA contains a benzyl substituent attached to the guanidinium moiety of amiloride, whereas the other analogues contain 5-N-alkyl substituents.

	Experimental Procedures

Top Summary Introduction Procedures Results Discussion References

Materials. [³H]Yohimbine (70-90 Ci/mmol) and [³H]rauwolscine (70-90 Ci/mmol) were from DuPont NEN (Hounslow, Middlesex, UK). [³H]RX821002 (40-70 Ci/mmol) was from Amersham International (Little Chalfont, Buckinghamshire, UK). Amiloride HCl, DMA HCl, HMA, and phentolamine HCl were from Sigma Chemical (Poole, Dorset, UK). BZA, EPA, and MBA were from RBI (Poole, Dorset, UK). Tissue culture reagents were from GIBCO BRL (Paisley, UK).

Cell culture and membrane preparation. The CHO cell line stably expressing the human _2A-adrenergic receptor (Kurose and Lefkowitz, 1994) was generously provided by Professor Robert J. Lefkowitz (Howard Hughes Medical Institute, Duke University Medical Center, Durham, NC). The cell line was grown in -minimum essential medium supplemented with 10% newborn calf serum, 2 mM L-glutamine, 50 IU/ml penicillin, and 50 µg/ml streptomycin, at 37° in 5% CO₂. For membrane preparations, the cell line was grown in 24.5 × 24.5-cm² plates until almost confluent; then, the cell layer was washed twice with phosphate-buffered saline, harvested in cold buffer 1 (20 mM Na-HEPES, pH 7.4, 10 mM EDTA), homogenized with a Polytron homogenizer (Brinkmann Instruments, Westbury, NY) (setting 6, 10 sec), and centrifuged at 40,000 × g for 15 min. The pellet was resuspended by vortexing in buffer 2 (20 mM Na-HEPES, pH 7.4, 0.1 mM EDTA) and then recentrifuged. The pellet was again resuspended in buffer 2 by passage through a 23-gauge needle and then a 27-gauge needle, diluted to 1 mg of protein/ml, and then stored at 70°. Protein concentrations were determined by the method of Bradford (1976) using bovine serum albumin as the standard.

Radioligand binding assays. For saturation experiments, membranes (10 µg of protein) were incubated with increasing concentrations (0.1-10 nM) of radioligand in duplicate, in a final volume of 1 ml of assay buffer (20 mM Na-HEPES, pH 7.4, 100 mM NaCl, 10 mM MgCl₂), at 30° for 60 min or 20° for 120 min, or with increasing concentrations (0.2-50 nM) of radioligand in duplicate, in a final volume of 0.4 ml of assay buffer at 30° for 60 min. Nonspecific binding was defined as the binding retained on the filter and membranes in the presence of 20 µM phentolamine. Where appropriate, amiloride analogues or solvents were added to both total and nonspecific binding assay tubes to control for effects on binding. Bound and free ligands were separated by rapid filtration under vacuum through GF/B glass-fiber filters (Whatman, Maidstone, Kent, UK), using a Brandell cell harvester (Semat, St. Albans, Hertfordshire, UK). The filters were washed three times with cold 20 mM sodium phosphate buffer, pH 7.4, transferred to scintillation vials, scintillation cocktail (Beckman, Palo Alto, CA) was added, and the filters were soaked overnight and counted.

Data analysis. Data were all fitted by nonlinear regression analyses, using the Grafit curve-fitting software (Erithacus Software, Staines, Middlesex, UK). This procedure allows the use of two or more independent variables (e.g., time and concentration), which was necessary for many of the analyses reported in this report.

(1)

	Results

Top Summary Introduction Procedures Results Discussion References

The equilibrium antagonist binding properties of the _2A-adrenergic receptor. As a necessary prelude to the allosteric binding studies, the equilibrium binding properties of three ³H-labeled antagonists at the human _2A-adrenergic receptor, permanently expressed in a CHO cell line (Kurose and Lefkowitz, 1994), were characterized. The nonspecific binding was defined as the residual binding measured in the presence of 20 µM phentolamine. In all instances, the saturation curves for the specific binding of the ³H-antagonists were compatible with the presence of a uniform population of binding sites. The experiments were carried out at both 20° and 30°, the two temperatures used in the dissociation studies. The K_d values calculated for the antagonists (0.7-3 nM, Table 1) were comparable with those reported in previous studies (Fraser et al., 1989; Halme et al., 1995). [³H]RX821002 had a 2-3-fold higher affinity than [³H]rauwolscine or [³H]yohimbine, and all three radioligands had a 50-100% higher affinity at 20° than at 30°. The B_max estimates in these experiments ranged between 16 and 18 pmol/mg of protein and were independent of radioligand or assay temperature.

View this table: [in this window] [in a new window]   TABLE 1 Affinity and Bmax values for the binding of three antagonists at the human 2A-adrenergic receptor Values are mean ± range/2 from two saturation binding assays, which were incubated at either 30° for 60 min or 20° for 120 min. The membranes were from a CHO cell line permanently expressing the cloned human 2A-adrenergic receptor (Kurose and Lefkowitz, 1994). The KL value (=1/Kd) is the affinity of the antagonists at the 2A-adrenergic receptor (Fig. 1).

Effect of amiloride analogues on the equilibrium binding of antagonists. Because it had been reported that 0.1 and 0.3 mM amiloride decreased both antagonist affinity and B_max in equilibrium binding studies (Howard et al., 1987; Nunnari et al., 1987; Jagadeesh et al., 1990), the effect of amiloride on [³H]yohimbine and [³H]rauwolscine binding was reexamined. The affinities of the radioligands were decreased in the presence of 0.1 and 0.3 mM amiloride (Fig. 3, Table 2). In contrast to the previous studies, however, the calculated B_max values for either radioligand (Table 2) were not significantly affected (p > 0.1, Student's paired t test) by the presence of the amiloride.

View larger version (27K): [in this window] [in a new window]   Fig. 3.   Scatchard plots of [3H]yohimbine and [3H]rauwolscine binding in the absence and presence of amiloride. Individual data points are shown. Membranes were incubated with increasing concentrations of radioligand at 30° for 60 min in the absence or presence of the stated concentrations of amiloride. Nonspecific binding was measured in the presence of 20 µM phentolamine.

View this table: [in this window] [in a new window]   TABLE 2 Effect of amiloride on the binding of [3H]yohimbine or [3H]rauwolscine to the human 2A-adrenergic receptor Values are mean ± standard error from three saturation binding assays, in which membranes containing the 2A-adrenergic receptor were incubated at 30° for 60 min with increasing concentrations of radioligand (0.2-50 nM) in a final volume of 0.4 ml of assay buffer. The Kd value (=1/KL, Fig. 1) is the dissociation constant of the antagonists in the absence or presence of amiloride at the 2A-adrenergic receptor.

View this table: [in this window] [in a new window]   TABLE 3 Affinity of the amiloride compounds at the 2A-adrenergic receptor The log affinities were determined in competition equilibrium experiments versus [3H]yohimbine, performed at either 30° for 60 min or 20° for 120 min. The data were fitted with eq. 1, with the slope factor set at 1; then, the derived log apparent affinity constants were converted to the log affinity constants logKi using the Cheng-Prusoff correction. Values are mean ± standard error from n experiments.

Antagonist dissociation at the _2A-adrenergic receptor. The dissociation data for all three radioligands at 20° and 30° fitted well to a monoexponential function. At 20°, RX821002, despite having the highest affinity, had a faster off-rate than either yohimbine or rauwolscine (0.083 ± 0.002 versus 0.034 ± 0.001 and 0.023 ± 0.001 min¹, respectively; see Table 4). At 30°, the dissociation rates were ~3.5-fold (yohimbine and rauwolscine) or 4.5-fold (RX821002) faster than those at 20° (data not shown).

View this table: [in this window] [in a new window]   TABLE 4 Effect of amiloride analogues on antagonist dissociation from the 2A-adrenergic receptor The experiments were performed at 20°, as described in the legend of Fig. 4. k1 and k2 are the dissociation constants for the dissociation of the antagonist from the unoccupied and the amiloride-occupied receptor, respectively (Fig. 1). The logobs is the difference between the log affinity of the amiloride analogue at the antagonist-occupied receptor (logKX) and the log affinity at the unoccupied receptor (logKi, Table 3). Values are mean ± standard error of n experiments.

Effect of individual amilorides on antagonist dissociation. To determine the affinities of the amilorides at the antagonist-occupied _2A receptor, their affects on [³H]yohimbine dissociation were examined. [³H]Yohimbine was chosen as the main radioligand for this study because it had a slower dissociation rate than [³H]RX821002 (above) and a lower nonspecific binding (as defined by 20 µM phentolamine) than [³H]rauwolscine. Amiloride, in a concentration-dependent manner, increased the dissociation rate (Fig. 4a). The dissociation curves remained monoexponential, and hence the data could be fitted to eq. 10 (Appendix). The simultaneous analysis of all the data gave an excellent fit, with the maximum increase in the [³H]yohimbine dissociation rate caused by amiloride (k₂/k₁) calculated to be 2.0-fold at 20° and the log affinity of amiloride at the yohimbine-occupied receptor to be 2.12 ± 0.02, equivalent to a dissociation constant (1/K_X) of 7.7 ± 0.4 mM (Table 4). The estimates of k₁ agreed well with the dissociation constant derived from studies on [³H]yohimbine dissociation alone (Table 4).

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Dissociation of [3H]yohimbine at 20° in the absence or presence of various concentrations of amiloride or EPA. Membranes were first preequilibrated with [3H]yohimbine; then, the dissociation commenced by mixing with phentolamine and various concentrations of the amiloride to be tested. Individual data points are shown. The data from each experiment were simultaneously fitted to eq. 10 (see Appendix), with time and amiloride analogue concentration as independent variables.

2

2

3

5

Competitive interactions between amiloride analogues as detected by their effects on [³H]yohimbine dissociation. Despite the excellent agreement between the experimental data and the predictions of the ternary complex allosteric model, the high concentrations of amilorides required to show kinetic effects raised the possibility that the kinetic effects could be the consequence of nonspecific perturbations of the structure of the receptor or membrane, which serendipitously fitted the model.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Dissociation of [3H]yohimbine at 20° in the absence or presence of various concentrations of HMA with and without amiloride or HMA with and without DMA. Membranes were first preequilibrated with [3H]yohimbine, then the dissociation commenced by mixing with phentolamine and various concentrations of the amilorides to be tested. Individual data points are shown. The data from each experiment were simultaneously fitted to eq. 9 (see Appendix), with time and amiloride analogue concentrations as independent variables.

View this table: [in this window] [in a new window]   TABLE 5 Effect of amiloride competition on yohimbine dissociation The experiments were performed at 20°, as described in the legends of Figs. 5 and 6. k1, k2, and k3 are the dissociation constants for the dissociation of [3H]yohimbine from either the unoccupied receptor or from the receptor occupied by amiloride X or Y, respectively, and logKX and logKY are the log affinities of amiloride X and amiloride Y, respectively, at the yohimbine-occupied receptor (Appendix). The logobs is the difference between logKY and the affinity of the amiloride analogue at the unoccupied receptor (logKi, Table 3). Values are mean ± standard error of n experiments.

2

3

3

1

View larger version (17K): [in this window] [in a new window]   Fig. 6.   Dissociation of [3H]yohimbine at 20° in the absence or presence of various concentrations of HMA with and without BZA. The experiments were performed as described in the legend of Fig. 5. Individual data points are shown. The data from each experiment were simultaneously fitted to eq. 9 (see Appendix), with time and HMA and BZA concentrations as independent variables.

	Discussion

Top Summary Introduction Procedures Results Discussion References

We explored whether there is a well-defined allosteric site on the human _2A adrenergic receptor, at which amiloride analogues could act to modulate antagonist binding to the primary binding site. This question has been investigated qualitatively in previous studies (Howard et al., 1987; Nunnari et al., 1987; Jagadeesh et al., 1990). However, no specific model was suggested for these data, and they do not fit the ternary complex allosteric model (Fig. 1) because the parent amiloride apparently decreased the B_max of the ³H-antagonist used to label the receptors. Our studies used cloned human _2A-adrenergic receptors expressed in a CHO cell line. In membrane preparations from this line, radiolabeled antagonists were found to bind to a uniform population of sites, which were expressed at a high density and displayed the appropriate pharmacology (Table 1, and data not shown). In our experiments, it was found that amiloride, at the concentrations used by others, did not significantly affect the calculated B_max values for either [³H]rauwolscine or [³H]yohimbine binding (Fig. 3, Table 2). The reason for the different result is not known.

Allosteric interactions between two ligands at a receptor can be explored and quantified by both equilibrium and kinetic binding studies (Stockton et al., 1983; Lee and El-Fakahany, 1991a; Leppik et al., 1994; Lazareno and Birdsall, 1995). Analyses of equilibrium binding data allow the estimation of both ligand affinities K_L and K_X and the cooperativity factor  (Fig. 1). However, a major limitation of equilibrium studies is that the analysis is difficult with high negative cooperativities (<0.01), the situation in the current study, because it then is not technically feasible to distinguish between the competition of two ligands at the one site and strong negative cooperativity between two sites.

In the kinetic approach, the effect of a test ligand on the dissociation rate of a radioligand is examined. Analysis of data from these experiments allows the estimation of the affinity of an allosteric ligand at the allosteric site (K_X, Fig. 1) when the primary site is occupied by the radioligand. In the current study, all of the amilorides tested were found to increase the rate of [³H]yohimbine dissociation from the _2A receptor (Fig. 4, Table 4). For each amiloride analogue concentration tested, the dissociation is monoexponential. The data from each experiment (except for those with BZA, which had a very small effect on dissociation rate) could be well fitted (Fig. 4) by eq. 10 (Appendix), showing that the data are compatible with the ternary complex allosteric model. From the fits, the calculated maximal increases of the amilorides on [³H]yohimbine dissociation ranged up to 160-fold. With such large increases in off-rate, the experiments are feasible technically only if the dissociation rate of the amilorides from the yohimbine-occupied receptor is very rapid, so as soon as the washing begins of the membranes trapped on the filters with cold buffer, the amiloride analogue rapidly dissociates, leaving only yohimbine occupying the receptor. The facts that the data are well-fitted and that the individual dissociation curves intersect at approximately the same point at zero time support this scenario.

The effect of the variation in the structures of the amilorides on the [³H]yohimbine dissociation rate is striking. Amiloride itself causes only a 2-fold increase in dissociation rate, and the addition of two methyl groups to the 5-amino function, to give DMA (Fig. 2), increases the yohimbine dissociation rate only a further 3-fold (Table 4). However, the addition of an additional three or four carbons, to give MBA, EPA, or HMA, causes a 50-80-fold increase in the effect caused by amiloride. The position of substitution also is important because the addition of a benzyl moiety to the other side of the molecule, to give BZA, slightly reduces the effect (from 2.0- to 1.4-fold, Tables 4 and 5). It can be speculated that within the allosteric site, the pocket around the 5-amino function is only large enough to accommodate a dimethylamino function, and larger groups cause a change in receptor conformation, leading to the dramatic increase in the yohimbine dissociation rate. The guanidinium function, on the other hand, may occupy a pocket large enough to accommodate a benzyl functionality, so the addition of such a grouping to the amiloride structure has only a minor effect on the yohimbine dissociation rate.

Because the increase caused by BZA is so slight, it could be quantified only in a competition dissociation assay (Fig. 6, Table 5). In these assays, the facts that the fits are so good, that the off-rate increases caused by HMA are either decreased or increased by the competing amiloride analogue in line with predictions, and that the parameters derived from the fits (Table 5) are in good agreement with those derived from dissociation assays performed with individual amilorides (Table 4) show that the data are compatible with the amilorides acting at the same site and rule out the possibility that the amilorides are acting in a nonspecific fashion (e.g., by perturbation of the membrane).

The calculated log affinity values of the 5-N-alkyl amilorides at the yohimbine-occupied receptor were found to range from 1.75 (1/K_X = 17.6 mM) for EPA to 2.47 (1/K_X = 3.4 mM) for HMA (Table 4). Thus, the spread in affinities of the 5-N-alkyl amilorides at the yohimbine-occupied receptor (5-fold) is far smaller than the range in affinities at the unoccupied receptor (145-fold) (Table 3). In addition, the increase in affinity of the amiloride analogues at the [³H]yohimbine-occupied receptor is not correlated with the increase in size of the 5-N-alkyl side chains, again in contrast to the situation with the unoccupied receptor (Table 3). This suggests either that the presence of an antagonist such as yohimbine at the primary binding site has a marked effect on the conformation of the allosteric site or that, for the unoccupied receptor, the amilorides are capable of binding at both the primary and allosteric binding sites. In the latter case, the affinity values derived from competition studies may reflect, partially or totally, binding at the primary site. The mathematics of equilibrium competition between [³H]yohimbine and the amilorides indicate that these two possibilities cannot be distinguished. This situation is in contrast to that found with the D₂ dopamine receptor, where "steep" competition curves between MBA and [³H]spiperone were better fitted with an equation derived from a model allowing MBA to interact with both the allosteric site and the primary site, with the MBA/MBA allosteric interaction exhibiting strong positive cooperativity (Hoare and Strange, 1996).

It also should be noted that the apparent association rate constant of [³H]yohimbine to the receptor/amiloride analogue complex (K_L × k₂) is approximately constant (0.8-2.5 × 10⁵ M¹ min¹) and ~100-fold slower than its association rate constant to the unliganded receptor (K_L × k₁).

The magnitude of the effects on the yohimbine dissociation rate is not correlated with the affinity of the amilorides at the yohimbine-occupied receptor. Thus, HMA and EPA, which have the largest effects on the yohimbine off-rate, have the highest and lowest affinities, respectively, of the 5-N-alkyl amilorides at the yohimbine-occupied receptor (Table 4). BZA, on the other hand, has the smallest off-rate effect but the highest affinity of the amilorides tested (Table 5). Thus, it is incorrect to use the magnitude of the effect on antagonist dissociation rate as a measure of the relative affinities of amiloride analogues at the liganded _2A-adrenergic receptor, as has been done in some previous studies.

The observed log cooperativity value (log_obs) is the difference between the log affinities of the amiloride analogue at the unoccupied and the antagonist occupied receptor. The calculated _obs values for the yohimbine/amiloride analogue interactions (Table 4) show a 65-fold variation. If the amilorides can only interact with the allosteric site, then the _obs value would correspond to the  depicted in Fig. 1.

In principle, the magnitude of an allosteric interaction is dependent on both of the ligands involved. The modulation by DMA and HMA of [³H]rauwolscine, [³H]RX821002, and [³H]yohimbine dissociation (Table 4) showed only small variations in the observed cooperativities between the amilorides and antagonists and small differences in the increases in dissociation rates caused by the amilorides This is different from the situation found for the M₂ muscarinic receptor in rat heart, where variation in antagonist structure produced a 10-fold variation in the negative cooperativity with gallamine (Stockton et al., 1983).

Thus, the results obtained in the current work are compatible with the existence of a discrete allosteric site on the _2A-adrenergic receptor, to which the amiloride analogues can bind and allosterically modulate antagonist binding at the primary site. The very different structure-binding relationships displayed by the amilorides at the unoccupied versus the antagonist-occupied receptor could argue either that the amilorides also interact with the primary binding site at the _2A-adrenergic receptor or that antagonist occupation at the primary site induces a substantial conformational change in the allosteric site that decreases the affinity of the amilorides by occluding a specific hydrophobic interaction.