Allosteric modulation of G protein‐coupled receptors by amiloride and its derivatives. Perspectives for drug discovery?

Abstract The function of G protein‐coupled receptors (GPCRs) can be modulated by compounds that bind to other sites than the endogenous orthosteric binding site, so‐called allosteric sites. Structure elucidation of a number of GPCRs has revealed the presence of a sodium ion bound in a conserved allosteric site. The small molecule amiloride and analogs thereof have been proposed to bind in this same sodium ion site. Hence, this review seeks to summarize and reflect on the current knowledge of allosteric effects by amiloride and its analogs on GPCRs. Amiloride is known to modulate adenosine, adrenergic, dopamine, chemokine, muscarinic, serotonin, gonadotropin‐releasing hormone, GABAB, and taste receptors. Amiloride analogs with lipophilic substituents tend to be more potent modulators than amiloride itself. Adenosine, α‐adrenergic and dopamine receptors are most strongly modulated by amiloride analogs. In addition, for a few GPCRs, more than one binding site for amiloride has been postulated. Interestingly, the nature of the allosteric effect of amiloride and derivatives varies considerably between GPCRs, with both negative and positive allosteric modulation occurring. Since the sodium ion binding site is strongly conserved among class A GPCRs it is to be expected that amiloride also binds to class A GPCRs not evaluated yet. Investigating this typical amiloride‐GPCR interaction further may yield general insight in the allosteric mechanisms of GPCR ligand binding and function, and possibly provide new opportunities for drug discovery.

"pre-crystal structure" research. 19 It is also the most conserved residue of the sodium ion site amongst GPCRs. The high conservation of the sodium ion pocket amongst class A GPCRs makes it probable that more structures with sodium ions bound in this site will emerge. There is little if any conservation present in the other GPCR classes, which makes it improbable that such a sodium ion binding site exists in these GPCRs.
Amiloride is primarily known as a potassium-sparing diuretic drug, acting through the blockade of renal epithelial sodium channels. 20 Amiloride and its analogs have also been found to bind to the sodium ion site of several GPCRs, modulating orthosteric ligand binding. 21 The negatively charged carboxylate of sodium ion site residue Asp 2.50 supposedly interacts with the positively charged guanidinium group present in all amilorides. The binding of amilorides into the sodium ion site of class A GPCRs renders these compounds potential pharmacological tools to probe molecular mechanisms of GPCR allosteric modulation. The chemical structures of amiloride and its analogs discussed in this review are depicted in Figures 1,2. Effects of the amilorides are represented in Table 1 categorized per GPCR and orthosteric ligands used. Most of the receptors in Table 1 are discussed in the main text.

| ADENOSINE RECEPTORS
Adenosine receptors have been studied extensively, and as a result, many orthosteric 22 and allosteric 23 24 Since the effects of amiloride binding to adenosine receptors appeared to be closely tied to sodium ion interactions, it was necessary to investigate and exclude the involvement of Na + /H + exchange proteins (one of the main targets of amiloride) in these interactions. 21 In this study, Garritsen et al 21  Gao and IJzerman 25 found that amiloride analogs benzamil, HMA, MCGMA, MIBA, and phenamil increased the dissociation rate of the antagonist [ 3 H]ZM-241,385 at the rat A 2A receptor, and that they were more potent than amiloride itself ( Figure 3). However, the affinity (defined by radioligand displacement in equilibrium) and the allosteric potency (defined by the concentration-dependent effect on the radioligand dissociation rate) did not correlate. This indicated a mixed competitive (ie, mutually exclusive displacement) and noncompetitive behavior of amilorides, in which amilorides and orthosteric ligands bind to the receptor at the same time, whereas amiloride influences the orthosteric ligand's dissociation rate. The amiloride analogs HMA and MIBA, with a lipophilic moiety on the 5′-position, proved to be the most potent compounds in increasing the dissociation rate of the orthosteric ligand, whereas they had equal affinities to benzamil and phenamil in displacing it. In contrast to the effect of amilorides, sodium ions decreased the dissociation rate of [ 3 H]ZM-241,385. Still, sodium ions and HMA appeared to compete for the same allosteric site.
In a study by Gao et al 26    Amilorides decreased the dissociation rate of agonist [ 125 I]-AB-MECA at the rat adenosine A 3 receptor, revealing that amilorides can also act as positive allosteric modulators depending on the radiolabeled probe used. 26 Furthermore the amilorides exhibited selectivity for the different adenosine receptor subtypes. Amiloride and 5-(N,N-dimethyl)amiloride (DMA) were more potent at the A 1 receptor in accelerating antagonist dissociation, whereas HMA was the most potent at the A 2A receptor and to a lesser extent at the A 3 receptor.
Solving the crystal structure of the adenosine A 2A receptor at a resolution of 1.8 Å provided a sufficiently high resolution to detect a sodium ion bound in its allosteric binding site for the first time ( Figure 4A). 13 The amino acids interacting with the sodium ion in this site are highly conserved amongst other GPCRs which confirmed previous studies in which modulation by sodium ions was tied to the same amino acids for different GPCRs. 12 The most conserved amino acid is a negatively charged aspartic acid (Asp52 2.50 ) which interacts directly with the positively charged sodium ion by means of a salt bridge. In molecular dynamics simulations, Gutiérrez-de-Terán et al 27 observed that the interaction of the sodium ion with Asp52 2.50 is highly stable in the receptor's inactive conformation. The presence of the ion also avoids rotamer changes in two other highly conserved residues,

| 697
In conclusion, the effects of amiloride and derivatives have been most extensively studied on adenosine A 2A receptors, through a number of orthogonal approaches. They all hint in the same direction, that is the amilorides compete with sodium ions at the allosteric sodium ion binding site in which Asp 2.50 is the central amino acid. The evidence for other GPCRs is less exhaustive but suggests similar conclusions, which will be discussed below.

| ADRENERGIC RECEPTORS
One of the first indications that amiloride inhibited the binding of orthosteric ligands at αand β-adrenergic receptors were found in 1987 by Howard et al, 36 which was followed by many studies with amiloride and its analogs at a number of adrenergic receptor subtypes. At the human α 1A -adrenergic receptor amiloride and its analogs benzamil, DMA, 5-(N-ethyl-N-isopropyl)-amiloride (EIA), MIBA, and HMA increased the dissociation rate of antagonist [ 3 H]prazosin, and the analogs with bulky lipophilic 5′-moieties were more potent in doing so. 37,38 Amiloride itself was characterized as an allosteric modulator acting at one allosteric site, but all the amiloride analogs appeared to bind to two different allosteric sites. The authors speculated that these allosteric sites could be present on one receptor or on a receptor dimer, but could not further confirm this. 37 The allosteric interaction by amilorides was seemingly in contradiction with previous results at rat and mouse α 1 -adrenergic receptors in which amiloride only showed a competitive interaction with antagonist [ 3 H]prazosin binding but did not influence its dissociation rate. 36 At the α 2B subtype, however, amilorides both increased and decreased the dissociation rate of antagonists. The 5′-substituted amilorides EIA and MIBA increased the dissociation rate of [ 3 H]rauwolscine binding, whereas the guanidino-substituted amiloride CBDMB decreased it. 44 The interaction of amiloride with β-adrenergic receptors has only been studied by Howard et al in 1987. At both the β 1and β 2 -adrenergic receptors amiloride displaced the antagonist [ 125 I]iodocyanopindolol competitively, because their binding was mutually exclusive. 36 Addition of sodium ions did not compete with amiloride binding, and it was concluded that amiloride did bind to the orthosteric site rather than to an allosteric sodium ion site.
Despite the lack of modulation of β-adrenergic receptors by sodium ions and amiloride, a sodium ion site was found in the crystal structure of the β 1 -adrenergic receptor. 14 The amino acids forming the sodium ion sites of the β 1adrenergic and the adenosine A 2A receptor are the most similar of the solved GPCR crystal structures with such a site. 12 That makes the difference in modulation by sodium ions and amilorides between these receptors remarkable and it is probably due to differences in the overall architecture of the two receptors.

| CHEMOKINE RECEPTORS
Amiloride interactions with the chemokine receptor family have only been studied by Zweemer et al 45  into Ala. Mutation of Trp256 6.48 even completely abolished HMA's allosteric effect, which is in contrast to the observed increase of HMA's affinity by the same mutation in adenosine receptors as discussed above. 32 Amino acid His297 7.45 is different from most class A GPCRs which usually harbor an Asn at the same position, but is conserved amongst chemokine receptors. The binding of HMA in CCR2s sodium ion binding site indicates that amiloride binding allows for a certain variation in the amino acids that constitute this binding cavity.

| DOPAMINE RECEPTORS
The general trend amongst the dopamine receptor subtypes is an increase of the dissociation rate of orthosteric ligands by amiloride and its analogs, as found in a comprehensive study of the effect of amiloride, benzamil, and MIBA. 50   dopamine receptors, and of these amilorides HMA was the most potent amiloride (Figure 7). Agonists were modulated similarly as antagonists by amilorides at the rat D 2 and D 3 dopamine receptors, because amiloride, DMA, and MIBA decreased the potency of the agonist dopamine in inducing receptor activation in functional assays. 50,54 At the D 4 receptor the allosteric effect of amiloride and its analogs was too small to be measured accurately, but an increase in antagonist [ 3 H]spiperone dissociation rate was still detected. As amilorides still inhibited binding of the orthosteric ligand the displacement was more competitive in nature. 50 The amino acids forming the sodium ion site in the dopamine receptors are conserved as well. into Asn decreased MIBA affinity, 58 indicating that amilorides bind in the sodium ion binding site as well. It may be assumed that amilorides also bind in the sodium ion binding site of the other dopamine receptors, but this has not been confirmed yet.

| GONADOTROPIN-RELEASING HORMONE RECEPTOR
The gonadotropin-releasing hormone (GnRH) receptor, also known as luteinizing hormone-releasing hormone receptor, is targeted by various drugs in the market for the treatment of sex-hormone-dependent diseases such as breast or prostate cancer. 59,60 These drugs are mostly peptidic agonists and antagonists that need to be administered by subcutaneous or intramuscular injections. The development of small-molecule ligands that may replace these peptidic ligands is therefore desirable. 61 Earlier results had indicated allosteric modulation of GnRH-stimulated luteinizing hormone release by sodium ions and amilorides. 62 In that light, the allosteric effects of amilorides on the GnRH receptor were investigated by Heitman et al 63  and M 3 receptors. 21 In rat trachea amiloride inhibited muscarinic M 3 receptor-mediated smooth muscle contraction 64 by the endogenous agonist acetylcholine, by an insurmountable noncompetitive interaction as its efficacy (E max ) was reduced. 65 In rat parotic acini, which express the muscarinic M 3 receptor, 66 amiloride inhibited binding of the muscarinic receptor antagonist [ 3 H]N-methylscopolamine in a competitive manner. 67 In the recent, relatively low-resolution crystal structures of the muscarinic M 2 and M 3 receptors sodium ion binding was not detected, [68][69][70] but the amino acids making up the sodium ion site are perfectly conserved when compared to adenosine and adrenergic receptors, 12 making amiloride binding to this site likely. In a recent molecular dynamics study sodium ion binding to (deprotonated), Asp 2.50 in the muscarinic M 3 receptor was suggested, keeping the receptor in an inactive state. 71 Along a similar vein, the egress pathway of a sodium ion from Asp 2.50 in the muscarinic M 2 receptor into the cytosol was also simulated in molecular dynamics calculations. 72

| SEROTONIN RECEPTORS
Amiloride and analogs have been found to inhibit orthosteric ligand binding to serotonin receptors. Benzamil inhibited agonist [ 3 H]8-OH-DPAT binding at the rat 5-HT 1A receptor. 21 Amiloride and EIA inhibited agonist [ 3 H]5carboxamidotryptamine binding at the human 5-HT 1B receptor. 73 In functional assays at the same receptor,  76 This high-resolution structure enabled the authors to inspect the putative sodium ion binding site around Asp100 2.50 , better than in an earlier crystal structure of this receptor. 77 Interestingly, and somewhat at odds with this review, the authors identified two water molecules rather than a sodium ion in the vicinity of this aspartic acid residue.
Triggered by this absence they performed additional radioligand binding studies in which no effects were observed from the addition of sodium ions or amiloride derivatives, whereas such effects were found in a control experiment the authors performed on the hA 2A R.
The receptors discussed above all belong to the class A family of GPCRs. Finally, we should like to discuss the evidence, admittedly limited and inconclusive, of amiloride interaction with two class C receptors.

| GABA B RECEPTORS
The GABA B receptor is activated by γ-aminobutyric acid (GABA) and it's derivative, baclofen (β-4-chlorophenyl-GABA). This receptor is coupled to potassium and calcium channels through G i /G o proteins. 78 Ong and Kerr explored the interaction of amiloride and its analogs with baclofen-induced depression of spontaneous discharges in rat isolated neocortical slices in Mg 2+ -free medium. The effect of baclofen (10 µM) was blocked by amiloride (200 µM), which increased the frequency of discharges and slightly reduced their amplitude when applied alone.
These effects persisted upon wash-out and baclofen remained ineffective on the discharges until 30 to 60 minutes after a switch to amiloride-free medium. Analogs of amiloride, DMA and MIBA, showed a similar mode of action, whereas they were at least twice as potent than amiloride in preventing the effect of baclofen on neocortical spontaneous discharges. DMA alone increased the discharge frequency and slightly reduced the amplitude in a concentration of 100 µM. Analogs lacking the guanidine moiety were ineffective. The authors explicitly stated, however, that an indirect effect of the amilorides via functional antagonism of coactivated adenosine A 1 receptors cannot be ruled out. 79

| T1R2/T1R3 RECEPTORS
The heterodimeric T1R2 and T1R3 taste receptor acts as a sweet taste sensor with multiple binding sites for sweeteners. 80 Amiloride (3 mM) were found to significantly reduce the responses to sweeteners such as sugar, artificial sweeteners, and sweet protein. Moreover, response inhibition of 1 mM aspartame by amiloride was observed in a concentration-dependent manner with an IC 50 value of 0.87 ± 0.20 mM. A study of the specificity towards the response mediated by the human sweet taste receptors showed that the suppression of receptor activity by amiloride is specific for hT1R2/hT1R3. Inhibitory effects of lactisole, a known hT1R2/hT1R3 inhibitor, and amiloride on the cellular response to aspartame were examined in cells expressing hT1R3 mutants (hT1R2/ hT1R3-A733V and hT1R2/hT1R3-F778A). Lactisole was less active on the mutants, whereas amiloride did not show such a differential effect. These results suggest that the binding site of amiloride is distinct from that of lactisole. 81 Amiloride inhibited the response of perillartine as a sweet activator on hT1R2/T1R3, T1R2, and T1R2heptahelical domain (HD). Molecular modeling suggested that perillartine and amiloride occupy the same binding pocket on the extracellular side of the hT1R2-HD. 82

| FUTURE DIRECTIONS FOR DRUG DISCOVERY
It is increasingly realized that GPCRs have multiple binding sites that may influence each other in allosteric ways.
The surge in crystal structures over the last decade has taught that ligands, including marketed drugs and clinical candidates, may have very different binding sites indeed. From this review, it has become obvious that the sodium ion binding site is yet another receptor domain to tune the ligand response, and that amiloride and its derivatives are prototypic small molecules that intervene with that site.
Does this offer options for future drug discovery? One might argue that the generic nature of the site and the evolutionary conservation of the amino acids aligning it are a drawback rather than an opportunity. In that view amilorides are another class of chemical probes that serve to unveil the complexities of GPCR functioning. A recent development, however, may prove this hypothesis wrong.
The crystal structure of the leukotriene B 4 (LTB 4 ) receptor BLT1 in complex with antagonist/inverse agonist BIIL260 has recently been reported. 83  to alanine markedly reduced the affinity of BIIL260 for the receptor providing also pharmacological evidence for the BIIL260's binding to the sodium ion binding site. Furthermore, benzamidine itself, as well as NaCl, served as negative allosteric modulators of radiolabeled agonist ([ 3 H]LTB 4 ) binding ( Figure 8B), suggesting their capability of forcing the receptor in an inactive state. 83 The chemical resemblance of amiloride's guanidine moiety and benzamidine might be a good starting point to further study the effects of amiloride and its analogs on the BLT1 receptor.

| CONCLUDING REMARKS
This review summarizes the current knowledge of the allosteric effects of amiloride and its analogs on GPCRs.
Allosteric effects of amilorides have been found on class A GPCRs (adenosine receptors, α-adrenergic receptors, the CCR2 chemokine receptor, dopaminergic receptors, the gonadotropin-releasing hormone receptor, the histamine H 1 receptor, muscarinic receptors, opioid receptors, and serotonin receptors), and, less convincingly, on class C receptors (GABA B and T1R2/3 receptors).
Amiloride and its analogs seem to follow a few general "rules" in their activity on these receptors. The propensity of amilorides to bind to the well-conserved sodium ion site amongst GPCRs may explain these common behaviors. For most receptors, amiloride analogs with bulky lipophilic moieties on the 5′-position have greater affinity and potency than the unsubstituted parent compound. This has not been explained fully, but it is clear that In contrast with these general "rules," differences in the affinities, potencies, and modulatory behaviors of amilorides can be quite outspoken, even between receptors where the sodium ion site harbors the same amino acids (i.e. adenosine, adrenergic, dopamine, and muscarinic receptors). To appreciate these differences it is important to discern between the different properties by which the allosteric effect of amilorides on orthosteric ligand binding may be described. In Table 1 we collected values for the different amilorides, of their affinity in displacing orthosteric ligands (IC 50 or K i ), their (allosteric) effect on the dissociation of orthosteric ligand (k off /k off (control) ), and their potency for these dissociation effects (EC 50 ). This information also helps to understand whether the interaction of a particular amiloride with an orthosteric ligand is competitive or noncompetitive. If amiloride inhibits orthosteric ligand binding but does not affect its dissociation rate, the binding is mutually exclusive and the interaction is defined as competitive. If the dissociation rate is changed though, both the orthosteric ligand and amiloride can bind to the receptor at the same time and the interaction is deemed noncompetitive. Another way to confirm a noncompetitive interaction is by showing insurmountability of the inhibiting effect in radioligand saturation (B max decrease) or functional assays (E max decrease), as discussed for the chemokine CCR2, muscarinic M 3 , and gonadotropin-releasing hormone receptor. However, these assays have been conducted far less than dissociation assays in amiloride research so we did not include these in Table 1.
In some cases, amilorides behave only as purely competitive inhibitors, whereas in other cases they behave as noncompetitive negative modulators, and a mixed behavior has also been observed. For some receptors the cause for mixed competitive/noncompetitive behavior was explained by a tendency of amilorides to bind both orthosteric and allosteric sites, but also in these cases the observed effect may be caused by binding in the sodium ion site only, where the competitive "fraction" of the allosteric effect is caused by either an overlap of binding with the orthosteric site or a conformational change of the receptor by amiloride binding. The latter option is quite likely from the structural evidence provided by the recently elucidated crystal structures.
At some of the discussed receptors, the modulatory effect by amilorides is probe-dependent, which has been described in other cases of allosteric modulation as well. 87,88 Amilorides act as positive allosteric modulators for agonist binding and as negative modulators for antagonists at the α 2A -adrenergic and adenosine A 3 receptors. Thus, in some cases, amilorides may also influence receptor signaling after agonist activation with consequences for effector bias or functional selectivity, for instance between G protein and β-arrestin signaling. 89,90 This has, however, not been demonstrated yet. At the α 2B -adrenergic receptor different amilorides even exhibit both positive and negative modulatory effects on the same orthosteric probe. Some of the differences in affinity and modulatory effect may be caused by differences in the sodium ion site itself, but the substantial conservation of the sodium ion site residues amongst GPCRs makes it more likely that these differences are caused by variations in receptor conformations.
Clinical application of amilorides targeting GPCRs is not self-evident due to their micromolar affinities and lack of selectivity. However, it may be feasible to synthesize amiloride analogs with variations on the 5′-position to improve their affinity and selectivity for GPCRs. In that sense, the recent structure elucidation of the BLT1/ leukotriene B 4 receptor in complex with BILL260 ( Figure 8) is noteworthy. BIIL260 is a selective, high-affinity antagonist for this receptor, occupying both the sodium ion and the orthosteric binding site. With the ongoing expansion of the crystal structure pool of GPCRs, further study and knowledge of the mechanisms of amiloride modulation will help in understanding and appreciating the allosteric mechanism in GPCR functioning and may pave the way for the design of antagonists forcing the receptor in a deeply inactive state.