The selectivity of α‐adrenoceptor agonists for the human α1A, α1B, and α1D‐adrenoceptors

Abstract Highly selective drugs offer a way to minimize side‐effects. For agonist ligands, this could be through highly selective affinity or highly selective efficacy, but this requires careful measurements of intrinsic efficacy. The α1‐adrenoceptors are important clinical targets, and α1‐agonists are used to manage hypotension, sedation, attention deficit hypersensitivity disorder (ADHD), and nasal decongestion. With 100 years of drug development, there are many structurally different compounds with which to study agonist selectivity. This study examined 62 α‐agonists at the three human α1‐adrenoceptor (α1A, α1B, and α1D) stably expressed in CHO cells. Affinity was measured using whole‐cell 3H‐prazosin binding, while functional responses were measured for calcium mobilization, ERK1/2‐phosphorylation, and cAMP accumulation. Efficacy ratios were used to rank compounds in order of intrinsic efficacy. Adrenaline, noradrenaline, and phenylephrine were highly efficacious α1‐agonists at all three receptor subtypes. A61603 was the most selective agonist and its very high α1A‐selectivity was due to selective α1A‐affinity (>660‐fold). There was no evidence of Gq‐calcium versus ERK‐phosphorylation biased signaling at the α1A, α1B, or α1D‐adrenoceptors. There was little evidence for α1A calcium versus cAMP biased signaling, although there were suggestions of calcium versus cAMP bias the α1B‐adrenoceptor. Comparisons of the rank order of ligand intrinsic efficacy suggest little evidence for selective intrinsic efficacy between the compounds, with perhaps the exception of dobutamine which may have some α1D‐selective efficacy. There seems plenty of scope to develop affinity selective and intrinsic efficacy selective drugs for the α1‐adrenoceptors in future.


| INTRODUC TI ON
Highly selective drugs are a prime goal in drug development because high-target receptor selectivity is expected to maximize clinical effectiveness while minimizing side-effects. 1 For antagonist drugs, this solely involves evaluating the affinity (ability of the ligand to bind to the receptor). However, for agonists, there are two properties that need to be evaluated: affinity and efficacy (ability of the receptorligand complex to induce a response). 1-4 A highly potent agonist could achieve this potency through high affinity or through high efficacy. Thus agonists can be highly selective due to highly selective affinity, or highly selective efficacy (where the compound could bind to several different receptors, but only activate one) or a mixture of both. 2,5 Agonist efficacy depends on several factors. Tissue and assaydependent factors include receptor number, receptor-effector coupling efficiency, effector response measured, assay response window, and any desensitization that occurs within the timeframe of the measurement. This makes direct comparisons of potency (EC 50 ) impossible across systems. Ligand/receptor factors are the innate ability of a certain receptor-ligand complex to induce a response and depend upon the chemical interaction between ligand and receptor. This, termed "intrinsic efficacy", is a measure of efficacy at the molecular/single ligand-receptor level 1,6 and is a more accurate measure of true ligand efficacy than either potency or maximal response. 7 A good way to compare the intrinsic efficacy of ligands is to remove tissue/assay factors and measure responses from individual receptor subtypes for many agonists in parallel in a null background.
Thus there are many structurally different α-adrenoceptor agonists with which to study agonist selectivity and determine how that is achieved.
In addition to α1-adrenoceptor-Gq-PLC-calcium signaling, the α1-adrenoceptors have also been shown to stimulate other signaling cascades. 14,22,23 Some recent studies have suggested that biased signaling can occur via the α1A-adrenoceptor. Isoprenaline was thought to have α1A-cAMP biased signaling. 24 Oxymetazoline was initially thought to have ERK1/2-phosphorylation bias. 25 It was later confirmed that the "biased" responses were occurring via a different receptor although phenylephrine and methoxamine ERK1/2phosphorylation bias and A61603 cAMP bias were proposed. 26 However the best way to determine whether a certain ligand is indeed an outlier inducing biased-signaling is to examine many ligands in parallel rather than just a few. 1 Many α1-agonist studies examine only a few ligands, study just one receptor, or use receptors from different species, making comparing intrinsic efficacy difficult (e.g. [24][25][26][27]. The aim of this study was to examine the selectivity of a large range of agonists for the human α1A, α1B, and α1D-adrenoceptors, with specific aims to identify whether agonists were selective due to selective affinity or selective intrinsic efficacy. Additionally, as several different agonist responses were examined, ligands with bias toward one signaling cascade over another would also be identified.

| 3 H-prazosin whole-cell radioligand binding
Cells were grown to confluence in white-sided 96-well view plates and whole-cell binding studies were conducted as previously described 28 in a total well volume of 200 µl per well. Cells were incubated with 3 H-prazosin and competing ligand in 200 μl for 2 h in serum-free media (sfm) at 37°C and plates counted using a Topcount (2 min per well) after a minimum of 6 h in the dark at room temperature. Total binding and non-specific binding (tamsulosin 10 μM for α1A and α1B, and 100 µM for α1D-see 28 for full data and explanation) were determined in every plate. 3

| Intracellular free calcium mobilization
Cells were grown to confluence in black-sided 96-well view plates, and calcium measurements were made using a

| 3 H-cAMP accumulation
Cells were grown to confluence in clear-sided 48-well plates and 3 H-cAMP accumulation was measured as previously described. 5 Following a 3 H-adenine load, cells were washed and incubated in sfm containing 1 mM IBMX (500 µl per well). Agonist (in 5 µl) was added and the cells were incubated for 5 h at 37°C. Basal and response to 10 µM forskolin were determined in every plate. Where used to examine Gi-coupled responses, basal cAMP was augmented by 10 μM forskolin and inhibition of this forskolin-induced response was examined. In these cases, forskolin was added to the wells 10 min after the addition of agonist. The assay was terminated with 50 µl concentrated HCl per well and 3 H-nucleotides separated by column chromatography. 5

| Data analysis
All pharmacological data were plotted using Graphpad Prism7.

| Whole-cell binding
The affinity of 3 H-prazosin has previously been determined from saturation binding in these cell lines. 28 The affinity of competing ligands was determined from a one-site sigmoidal response curve where the IC 50 is the concentration required to inhibit 50% of the specific binding of the 3 H-prazosin, A is the concentration of the competing ligand and NS is the non-specific binding (Equation 1).
The affinity (K D value) of the competing ligand was then calculated from the IC 50 using the Cheng-Prusoff equation (Equation 2) where [ 3 H-prazosin] is the concentration of 3 H-prazosin in that experiment and K D 3 H-prazosin is the K D value of the radioligand.

| Functional experiments
Agonist responses were usually best described by a one-site sig- quote for ERK1/2-phosphorylation is that of the initial stimulatory part of the response.

| Efficacy ratios
Efficacy ratios were calculated by dividing the K D value by the EC 50 value for each ligand as per method of Furchgott. 6 (1) % uninhibited binding = 100 − (100 × A) (A + IC 50 ) + NS.
(2)     H-prazosin specific binding at the α1D receptor was best described by a two-component inhibition curve.
Here the K D value for the first component (higher affinity) and second component (lower affinity) is given with the % of the response at the first component. For further details and example graphs see Proudman et al., 2020.

| Determination of ligand affinity from 3 Hprazosin whole-cell binding
The affinity (K D ) for 3 H-prazosin has previously been determined in these cell lines as 0.71, 0.87, and 1.90 nM for the α1A, α1B, and α1D-adrenoceptor, respectively, with receptor expression levels of 1152fmol/mg protein, 4350fmol/mg protein, and 417fmol/mg protein, respectively. 28 The α1D-adrenoceptor is the full-length receptor and is associated with lower levels of expression than either α1A or α1B-adrenoceptor expression, or an N-terminal truncated α1D-adrenoceptor. [30][31][32][33] As expected therefore, the window of specific binding was smaller in the CHO-α1D cells than the CHO-α1A  Figure 1). As expected, many agonists had relatively low affinity for the α1-adrenoceptors (Table 1, Figure 1). A61603 was the most selective agonist with an α1A-adrenoceptor selective binding affinity of over 660-fold (Table 1, Figure 1).

| Free intracellular calcium mobilization
As all three α1-adrenoceptors are primarily Gq-coupled receptors, intracellular calcium mobilization was studied. In CHO-α1A cells, adrenaline stimulated an increase in intracellular calcium (log EC 50 = −9.09) that was 58.9% that of the response to 10 µM ionomycin (Table 2, Figure 2). This gave adrenaline an efficacy ratio of 4.00 making it the ligand with the greatest intrinsic efficacy at the α1A-adrenoceptor (Table 2). A similar pattern was seen in CHO-α1B and CHO-α1D cells (Tables 3 and 4, respectively).

| ERK1/2-phosphorylation
Adrenaline stimulated an increase in ERK1/2-phosphorylation in CHO α1A cells that was best described by a two-component response. After an initial increase in ERK1/2-phosphorylation (log EC 50 (5) log efficacy ratio = log K D log EC 50 . Abbreviations: ND, not determined.

CHO-
a These compounds had a bi-phasic response. Log EC 50 and % PDBu given for initial stimulatory part of response. b These compounds stimulate ERK1/2-phosphorylation in parent CHO cells, see Supplementary data Table S1, Figure S1: however, the responses to oxymetazoline, xylometazoline, and labetolol are more than 10-fold more potent than the responses on the untransfected cells, so are likely to be α1-adrenoceptor mediated.

| 3 H-cAMP accumulation
Adrenaline stimulated an increase in 3 H-cAMP accumulation in CHO-α1A cells (log EC 50 −5.63) that was 164% of the response seen to 10 µM forskolin ( Figure 4, Table 2). This response is significantly rightshifted when compared with the stimulatory adrenaline-induced calcium mobilization and ERK1/2-phosphorylation responses in these cells. To look for Gi-mediated inhibition of cAMP, the ability of ligands to inhibit forskolin-stimulated cAMP was examined.
In CHO-α1A cells, adrenaline did not inhibit cAMP accumulation (suggesting no Gi-coupled response, Figure 5, Table 2). However, the stimulatory response was still seen and if anything, augmented, most likely as a result of forskolin augmentation of the Gs-coupled response (as seen in [34,35]). Responses were also observed in the CHO-α1B and CHO-α1D cells (Figures 4 and 5, Tables 3 and 4).

| Responses in parent CHO cells without the transfected receptors
There were no measurable intracellular calcium mobilization dose responses in response to any of the agonists in the parent (untransfected) CHO cells (Table S1). A few compounds had a higher than basal stimulation at the highest concentration only and are given in Table S1. Oxymetazoline, xylometazoline, dihydroergotamine, lisuride, labetalol, and CGP 12177 stimulated ERK1/2phosphorylation responses in the parent CHO cells (Table S1,  Table 1

| Correlation plots
In order to examine for any evidence of bias signaling, the log EC 50 values for calcium mobilization were correlated with those for ERK1/2-phosphorylation ( Figure 6A-C). This suggests little evidence for biased signaling between these two responses at any of the α1adrenoceptor subtypes. To examine for potential calcium-cAMP-bias, a similar plot was constructed for calcium versus cAMP accumulation.
Here, data are plotted for the augmented cAMP accumulation in the Finally returning to a major aim of the study -to look for any evidence of intrinsic efficacy selectivity -the efficacy ratios for calcium release were compared for α1A and α1B (Figure 6g) and α1A and α1D (Figure 6h). Here, dobutamine was the ligand furthest from the line of best fit suggesting it has some α1D-selective efficacy relative to that seen at the α1A or α1B-adrenoceptors.

| DISCUSS ION
This study compared the binding affinity and functional responses of 62 compounds at the human α1A, α1B, α1D-adrenoceptors. α1A and α1B-adrenoceptors are present in human heart. 13 Although These compounds stimulate ERK1/2-phosphorylation in parent CHO cells, see Supplementary data Table S1, Figure S1.
c Xylometazoline caused a decrease in forskolin-stimulated cAMP accumulation. The data given are log IC 50 and % inhibition of forskolin-stimulated cAMP.

TA B L E 3 (Continued)
TA B L E 4 Log K D values from 3 H-prazosin whole-cell binding (from Table 1  α1A and α1D-adrenoceptors are important for vasoconstriction, the role of the α1B-adrenoceptor (also present in blood vessels) is less certain. 10,14,36,37 Interestingly, the affinity of adrenaline and noradrenaline was substantially lower for the α1B-adrenoceptor than for α1A or α1D-adrenoceptors. Adrenaline and noradrenaline had high intrinsic efficacy, with adrenaline being marginally higher at each receptor (in keeping with the slightly more potent adrenaline vs. noradrenaline responses observed by [21,25,38]). Phenylephrine  (Tables 1-4), however, both oxymetazoline and xylometazoline had α1A-adrenoceptor selective affinity. A degree of α1A selective affinity of these two compounds has also been previously reported. 21,25,26,38,48 Although clear agonist responses were seen with oxymetazo- Abbreviation: ND, not determined.
a These compounds stimulate ERK1/2-phosphorylation in parent CHO cells, see Supplementary data Table S1, Figure S1. b Oxymetazoline and xylometazoline cause a decrease in forskolin-stimulated cAMP accumulation. The data given are log IC 50 and % inhibition of forskolin-stimulated cAMP as both compounds caused a decrease in cAMP accumulation. involving the extracellular end of transmembrane 4. 52 A "low" affinity state of the α1A-adrenoceptor has been previously proposed (α1L), initially reported as having a lower prazosin affinity 10 and ref- erences therein) but also seen with affinity measurements in functional assays. 11 Further studies are required to determine whether the low potency of these agonists are occurring at a lower affinity α1A-secondary agonist conformation, akin to that of the α1and α2adrenoceptors, and whether this has any relationship of this to the "α1L"-adrenoceptor. induce phospholipase C or inositol phosphate responses. They concluded that their calcium response was a non-Gq-coupled event, and thus isoprenaline was an ERK versus Gq-biased ligand. Evans et al. 25 and da Silva et al. 26 report phenylephrine and methoxamine as having ERK versus Gq-calcium bias. In our study, phenylephrine, methoxamine, and isoprenaline have different intrinsic efficacies, but no calcium versus ERK-phosphorylation bias. It is possible that the ERK1/2-phosphorylation in our study could be downstream from the calcium response (as suggested by [22]).
This lower potency Gs-coupling is similar to that seen at the adenosine A1 receptor 55 and may represent a lower agonist affinity for the Gs-coupled conformation of the α1-adrenoceptors than for the Gq-coupled conformation. This was not always the case for α1B-see

below.
There was no inhibition of forskolin-stimulated cAMP in CHO-α1A or CHO-α1B cells, suggesting no evidence for Gi receptor coupling. In fact, forskolin further increased the cAMP stimulatory responses, in keeping with forskolin-induced enhancement of GPCR-Gs-adenylyl cyclase coupling (proposed by [35] and [34]), and da Silva et al. 26 who were not able to measure a oxymetazoline-cAMP response, but observed an oxymetazloline response in the presence of 1 μM forskolin. In CHO-α1D cells, an inhibitory cAMP response was seen with oxymetazoline and xylometazoline, similar to that seen in parent CHO cells, suggesting that this was not α1D-receptor mediated. Thus oxymetazoline and xylometazoline cause non-α1-adrenoceptor-mediated responses in CHO cells that decrease cAMP and stimulate significant ERK-phosphorylation, very much in keeping with the CHO Gi-coupled 5HT-1B receptor proposed by da Silva et al. 26 The stimulatory response seen in CHO-α1A and CHO-α1B cells is likely receptor-mediated due to the higher level of transfected α-adrenoceptors in these cell lines.
There was a good correlation between calcium mobilization and cAMP stimulation in CHO-α1A-cells suggesting little calcium versus cAMP biased signaling. However, the correlation plot for the α1Badrenoceptor shows substantially more scatter with adrenaline, noradrenaline, and α-methylnorepinephrine having substantially more potent calcium than cAMP responses, whereas naphazoline, dexmedetomidine, medetomidine, allyphenyline, detmonidine, guanabenz, and dobutamine had more potent cAMP responses than calcium.
There may therefore be some bias signaling with respect to calcium and cAMP pathways via the α1B-adrenoceptor. was followed by a decrease at higher agonist concentrations ( Figure 3).
This appears to be an efficacy driven phenomena because these six ligands had the highest intrinsic efficacy as determined from the calcium release assay. This phenomena was also seen with adrenaline and noradrenaline in CHO-α1B cells, but not in CHO-α1D cells (lower receptor expression) where all responses were smaller relative to the PDBu response. Interestingly, 22 proposed that α1A-induced cAMP stimulation could have a negative effect on ERK1/2-phosphorylation.
Thus the Gs-coupled cAMP stimulation, which only occurs at higher agonist concentrations, could be the explanation for the decrease in ERK1/2-phosphorylation seen at higher agonist concentrations.
Finally, the intrinsic efficacy of ligands was examined.
Although direct EC 50 comparisons are not possible across cell lines, the rank order of intrinsic efficacies are either as presented in Tables 2-4 or pictorially from correlation plots ( Figure 6).
There was a good correlation for the intrinsic efficacy of agonists at these receptors, suggesting little intrinsic activity selectivity.
The ligand with the most selective intrinsic efficacy was dobutamine (ranked 4 th in the α1D table, and furthest from the line of best fit, Figure 6). Dobutamine stimulated a response with similar affinity, potency, and intrinsic efficacy to that of noradrenaline in the CHO-α1D cells, but despite a similar affinity, did not stimulate any measurable calcium or ERK1/2-phosphorylation CHO-α1B response and only a mid-table response intrinsic efficacy response in the CHO-α1A cells. Dobutamine has previously been shown to have affinity for α1-adrenoceptors. 56 However, A61603 apart, given the lack of selectivity of most α1-adrenoceptor agonists, there seems plenty of scope to develop both affinityselective and intrinsic efficacy-selective agonist drugs for the α1-adrenoceptors in future.

ACK N OWLED G M ENTS
Supported by a Medical Research Council MICA award (MR/ M00032X/1). We thank June McCulloch for technical assistance.

D I SCLOS U R E
JGB is on the Scientific Advisory Board for CuraSen Therapeutics.
The majority of the data in this study predates that appointment.

AUTH O R CO NTR I B UTI O N S
JGB designed the research study. JGB and RGWP performed the research. JGB analyzed the data. JGB wrote the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from the authors: The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.