The affinity and selectivity of α‐adrenoceptor antagonists, antidepressants and antipsychotics for the human α2A, α2B, and α2C‐adrenoceptors and comparison with human α1 and β‐adrenoceptors

Abstract α2‐Adrenoceptors, subdivided into α2A, α2B, and α2C subtypes and expressed in heart, blood vessels, kidney, platelets and brain, are important for blood pressure, sedation, analgesia, and platelet aggregation. Brain α2C‐adrenoceptor blockade has also been suggested to be beneficial for antipsychotic action. However, comparing α2‐adrenoceptor subtype affinity is difficult due to significant species and methodology differences in published studies. Here, 3H‐rauwolscine whole cell binding was used to determine the affinity and selectivity of 99 α‐antagonists (including antidepressants and antipsychotics) in CHO cells expressing human α2A, α2B, or α2C‐adrenoceptors, using an identical method to β and α1‐adrenoceptor measurements, thus allowing direct human receptor comparisons. Yohimbine, RX821002, RS79948, and atipamezole are high affinity non‐selective α2‐antagonists. BRL44408 was the most α2A‐selective antagonist, although its α1A‐affinity (81 nM) is only 9‐fold greater than its α2C‐affinity. MK‐912 is the highest‐affinity, most α2C‐selective antagonist (0.15 nM α2C‐affinity) although its α2C‐selectivity is only 13‐fold greater than at α2A. There are no truely α2B‐selective antagonists. A few α‐ligands with significant β‐affinity were detected, for example, naftopidil where its clinical α1A‐affinity is only 3‐fold greater than off‐target β2‐affinity. Antidepressants (except mirtazapine) and first‐generation antipsychotics have higher α1A than α2‐adrenoceptor affinity but poor β‐affinity. Second‐generation antipsychotics varied widely in their α2‐adrenoceptor affinity. Risperidone (9 nM) and paliperidone (14 nM) have the highest α2C‐adrenoceptor affinity however this is only 5‐fold selective over α2A, and both have a higher affinity for α1A (2 nM and 4 nM, respectively). So, despite a century of yohimbine use, and decades of α2‐subtype studies, there remains plenty of scope to develop α2‐subtype selective antagonists.


| INTRODUC TI ON
The α2-antagonist yohimbine, obtained from the African Corynanthe yohimbe tree (Pausinystalia johimbe), has been in clinical use as an aphrodisiac for over a century. 1,2 It has been used for erectile dysfunction and increases many sexual behaviours through central (CNS) α2effects and potential local effects as α2A, α2B, and α2C-adrenoceptors are expressed in human corpus cavernosum, 1,2 and can indeed bind yohimbine from tree bark. 3 The α2-antagonist idazoxan, developed in 1970s, is selective for α2 over α1-adrenoceptors, but also binds to other imidazoline binding sites which limits its usefulness in tissue or animal studies. 4,5 This led to the development of RX821002, a 2-methyl congener of idazoxan, in the 1980s which retained high α2-adrenoceptor affinity but without imidazoline receptor affinity (although 5-HT receptor interactions still occur 6,7 ).
α2-Adrenoceptors are subdivided into α2A, α2B, and α2Csubtypes. With receptors being present in the heart, blood vessels, and kidney, 8 α2-adrenoceptors are important in blood pressure control (an interplay between α1, α2, and β-adrenoceptors) and including central and peripheral α2-effects. In addition, many α2adrenoceptors present in the brain also have clinical roles in anaesthesia and psychiatric treatments 9 with both pre-and post-synaptic effects on neurotransmission. [10][11][12][13] α2A-adrenoceptors are widely expressed and are important for blood pressure, sedation, analgesia, platelet aggregation, and hypothermia. 14,15 In the brain, 90% of all α2-adrenoceptors are of the α2A subtype and they are highly expressed in the prefrontal cortex where activation increases cognitive function. 16,17 α2A-adrenoceptor antagonism may be important in sepsis (administration of the α2A-antagonist BRL44408 reduced pro-inflammatory cytokines, TNFα and IL-6 and increased survival in a rat model of sepsis 18 ) and potentially clinically relevant α2A-mirtazapine-induced reversal of analgesia. 19 The roles of the α2B-adrenoceptors are less clear. α2B-adrenoceptors are involved in blood pressure control (activation causes a hypertensive response related to renal salt balance. 14 The expression and effects of the α2Badrenoceptors appear very minor in the brain. 17 The α2C-adrenoceptor is involved in catecholamine release in adrenal chromaffin cells 15 and in the brain process of startle and stress responses. 14 α2C-adrenoceptors form 10% of all brain adrenoceptors but appear particularly prevalent in the striatum and hippocampus. 16 For certain antipsychotics (e.g., clozapine), α2C-antagonsim, in addition to dopamine D2 blockade, is thought to be beneficial in the management of schizophrenia 12,13,17 and α2C-antagonism may be helpful in improving cognition in dementia. 12 However, a lack of subtype selective α2-adrenoceptor ligands has impaired understanding and knowledge of α2-subtype expression and α2subtype function, with much information coming from knockout mice, with subtype adaptation problems that this brings. [12][13][14][15]17,20 Determining the affinity and selectivity between different α2adrenoceptor antagonists has been difficult due to significant variability both within individual, and between different existing studies.
In addition, substantial differences are reported for affinity measurements of single ligands at single subtypes. Reports of prazosin affinity at human α2A-adrenoceptors range 50-fold, from 300 nM 21,28 to a few thousand nM, 23,24,29 to 16000 nM. 6 Differences in affinity have also been attributed to technique. A 5-fold difference in 3 Hrauwolscine affinity, and 4-fold difference in 3 H-RX821002 and 3 H-atipamezole affinity was found with different buffers. 30 Thus, previously reported differences in affinity are likely to be due to several explanations: species is very important but techniques (cloned receptor vs. whole tissue, membrane vs. whole cell, different buffers) are also important and make direct comparison of studies difficult.
This study therefore measured the affinity and selectivity of a wide range of α-antagonists (including antidepressants and antipsychotics) in living CHO cells expressing the human α2A, α2B, or α2Cadrenoceptor. Furthermore, as these measurements were determined using an identical technique in human β1 and β2-adrenoceptors (included here, and 31,32 ) and α1-adrenoceptors, 33 this study explores the affinity and selectivity of ligands across the human adrenoceptors commonly targeted for cardiovascular, urological and CNS effects.

| Materials
All compounds, together with the supplier and catalogue number are given in alphabetical order in Supplementary Data Table 1

| Cell lines
CHO-K1 (RIDD: CVCL_0214) were stably transfected with the human α2A-adrenoceptor, human α2B-adrenoceptor or human α2C-adrenoceptor DNA (DNAs from Guthrie DNA Resource Centre) using Lipofectaime and Optimem according to the manufacturer's instructions. Following 3 weeks selection using resistance to neomycin (at 1 mg/ml), single clones from each transfection were isolated by dilution cloning. Thus stable cell lines CHO-α2A, CHO-α2B, and CHO-α2C were created. CHO lines stable expressing the human β1 or β2-adrenoceptor were also used. 31

| Cell culture
CHO cells were grown in Dulbecco's modified Eagle's medium nutrient mix F12 (DMEM/F12) containing 10% foetal calf serum and 2mM L-glutamine in a 37°C humidified 5% CO 2 : 95% air atmosphere. Cells were seeded into white-sided, clear bottomed 96-well view plates and grown to confluence. Cells were always grown in the absence of any antibiotics. Mycoplasma contamination has intermittently been monitored within the laboratory (negative) but cell lines were not tested routinely with each experiment.

| 3 H-rauwolscine and 3 H-RX821002 whole cell saturation binding
The K D value for both radioligands was determined in each cell line by saturation binding. The radioligands were diluted to twice the final concentration in serum-free media (sfm, DMEM/F12 containing 2 mM L-glutamine). Media was removed from each well and replaced with either 100 µl sfm (total binding) or 100 µl, 20 µM RX821002 (when 3 H-rauwolscine used) or 20 μM yohimbine (when 3 H-RX821002 used) in sfm to determine non-specific binding. 100µl radioligand was then added to the wells (quadruplicates per condition =1 in 2 dilution in well), and the plates incubated at 37°C (humidified 5% CO 2 : 95% air atmosphere) for 2 h. After 2 h, the cells were washed twice by the addition and removal of 2×200 µl cold (4°C) phosphate-buffered saline. A white base was applied to the plate to convert the wells into white-sided/white-bottomed plates, 100μl Microscint 20 was added to each well and a transparent top seal applied to the plates. Plates were left at room temperature in the dark for at least 6 h before being counted on a Topcount (PerkinElmer, 2-min count per well).

| 3 H-rauwolscine, 3 H-RX821002, and 3 H-CGP12177 whole cell competition binding
Affinity was assessed using the whole cell binding method of. 31 Ligands were diluted in sfm to twice their final concentration. Media was removed from each well and 100µl ligand added to triplicate wells. This was immediately followed by the addition of 100µl radioligand (diluted in sfm) and the cells incubated for 2 h at 37°C (5%  These concentrations were excluded from the analysis. Total binding (6 wells/plate) and non-specific binding (6 wells/plate (determined by the presence of 10µM yohimbine or 10µM RX821002 in sfm) was defined in every plate.
Given the two-component inhibition of 3 H-prazosin binding seen with dibenamine and phenoxybenzamine at the α1-adrenoceptors, sodium thiosulphate, which reacts with the ethyleniminium ions, was used in dibenamine and phenoxybenzamine experiments, in excess, as in Ref. 33 Thus all studies in human β, α1, and α2-adrenoceptors have been conducted in intact living mammalian cells using the same method.
The only differences between the experiments are the radioligand, the ligand used to define non-specific binding and the transfected receptor. As all experiments were conducted in living cells, physiological levels of intracellular endogenous GTP will always have been present and potentially are therefore more akin to how drugs bind in people, rather than studies conducted in membrane preparations.
There is theoretically a potential difference in affinity measurement if compounds have a different intrinsic efficacy for different receptor subtypes. Thus, if one compound is a partial agonist at one receptor subtype but an inverse agonist at another, a different receptor state is induced upon binding to the receptor. This may therefore affect how the compound and radioligand compete for the receptor, which in turn could theoretically affect affinity measurements. As this study was aimed at studying antagonists, this effect is likely to be minimal.

| Data analysis
Saturation curves for specific radioligand binding were plotted using the following equation in GraphPad Prism 7: where B max is the maximum specific binding, K D is the dissociation constant of the radioligand and [ 3 H-radioligand] is the concentration of the radioligand.
In all cases where a K D value is stated, increasing concentrations of the competing ligand fully inhibited the specific binding of the radioligand (unless otherwise annotated in the tables). The following equation was then fitted to the data using Graphpad Prism 7 and the IC 50 was then determined as the concentration required to inhibit 50% of the specific binding.
where [A] is the concentration of the competing ligand and IC 50 is the concentration at which half of the specific binding of radioligand that has been inhibited.
From the IC 50 value, the known concentration of radioligand and the known radioligand K D for at each receptor, a K D (concentration at which half the receptors are bound by the competing ligand) value was calculated using the Cheng-Prusoff equation: In some cases, the maximum concentration of competing ligand was not able to inhibit all of the specific binding. Where no inhibition of radioligand binding was seen, even with a maximum concentration of competing ligand possible, "no binding" is given in the tables. Where the inhibition produced by the maximum concentration of the competing ligand was 50% or less, an IC 50 could not be determined and thus a K D value not calculated. This is shown in the tables as IC 50 >top concentration used (i.e., IC 50 >100µM means that 100µM inhibited some but less than 50% of the specific binding). In cases where the competing ligand caused a substantial (greater than 50%, but not 100%) inhibition of specific binding, an IC 50 value was determined by extrapolating the curve to non-specific levels and assuming that a greater concentration would have resulted in 100% inhibition. These values are given as apparent K D values in the tables.
Nomenclature of Targets and Ligands.
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guide topha rmaco logy.
org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, 34

| Affinity and selectivity of ligands at α2-adrenoceptors
The affinity and selectivity of a large range of α-adrenoceptor antagonists was evaluated ( Figure 1; Table 2). It is clear that  Table 2) in a manner similar to that seen in the α1-adrenoceptors. 33 The responses in CHO-α2A cells and for dibenamine in CHO-α2C cells were too low affinity for a second component to be clearly determined. Dibenamine and phenoxybenzamine both contain a nitrogen mustard group, which cyclises to form ethyleniminium ions. 38 Sodium thiosulphate reacts with the ethyleniminium ions preventing them interacting with α-adrenoceptors. 38 Preincubation with sodium thiosulphate abolished the higher affinity components and reduced the affinity of both ligands at all three receptors a follows: dibenamine −4.59 ± 0.08 n=5, −4.64 ± 0.07 n=5, and −4.64 ± 0.11 n = 5 for α2A, α2B, and α2C, respectively; and for phenoxybenzamine −4.71 ± 0.13 n = 5, −4.86 ± 0.08 n = 5, and −4.96 ± 0.10 n = 5 for α2A, α2B, and α2C, respectively and are therefore similar to the second component response. The higher affinity K D values in Table 2 are therefore highly likely to be the affinity of the ligand interacting with the receptor (as in 33 ).
Given the more recent suggestions of α2C affinity being important for antipsychotic drug actions, the affinity and selectivity of antidepressants (Table 3) and antipsychotics ( Figure 3; Table 4) were examined.

| Affinity and selectivity of ligands at β1 and β2-adrenoceptors
Given that drug interactions at α1, α2, β1, and β2-adrenoceptors affect blood pressure control, and that the affinity of these ligand has been assessed in comparative assays in α1 and α2 receptors, the affinity of ligands was also evaluated in CHO cells stably expressing the human β1 or β2-adrenoceptor using 3 H-CGP12177 whole cell binding ( Figure 3; Table 5).

| DISCUSS ION
One aim of this study was to determine the selectivity of a range of ligands at the human α2-adrenoceptors and this study con-

| Selectivity across α1, α2 and β-adrenoceptors
Given that the affinity values determined in this study were using an identical technique to affinity values determined in the human α1 and β1 and β2-adrenoceptors (the only difference was transfected receptor, radioligand and ligand used for non-specific binding), a second aim of this study was to compare affinities between the human adrenoceptors (α2, β1, and β2 reported here, α1A, α1B, and α1D-adrenoceptor subtypes from 33 and β1, β2, and β3 from. 31,32 ).
The findings of these studies are therefore discussed as a whole, in comparison with other literature findings.   Silodosin (used for benign prostatic hyperplasia BPH) and naftopidil (used especially in Japan for BPH and ureteral stone expulsion, 45 ) have significant β2-adrenoceptor affinity (~30 nM).
Silodosin is highly α1A-selective (0.25 nM) giving a >100-fold selectivity window compared to the other adrenoceptors. Naftopidil, however is not selective, with α1A and β2 affinities only 3-fold apart and thus potentially increasing the risk of bronchospasm in TA B L E 3 Log K D values of antidepressants binding to the human α2A, α2B and α2C-adrenoceptors. Values represent mean ±s.e.mean of n separate experiments. Selectivity ratios are also given, where a ratio of 1 demonstrates no selectivity for a given receptor subtype over another. Thus, clompiramine has 2.5-fold higher affinity for the α2B than the α2A-adrenoceptor. Compounds are arranged in order of α2A-selectivity. Note: app = apparent affinity The maximum concentration of competing ligand inhibited most but not all of specific binding. An IC 50 was determined by extrapolating the curve assuming that all specific binding would be inhibited if a higher concentration of competing ligand were possible.

| Antidepressants and antipsychotics
Given the considerable CNS expression of α2A and α2Cadrenoceptors, and that many antidepressants and antipsychotics have high α1A-affinity, a third aim of this study was to compare the affinity of antidepressants and antipsychotics across the adrenoceptors.
The antidepressants generally had poor α2-adrenoceptor affinity, considerably lower affinity than that seen for the tricyclic antidepressant affinities at the α1A-adrenoceptor. The antidepressant mirtazapine is a slight outsider with the highest α2-affinity of the antidepressants studied here, and higher than α1A-affinity. It has been associated with antinociceptive properties attributed to α2adrenoceptors in mice. 19,51 Mirtazepine (α2A-affinity 158 nM) and α2C 110 nM), had similar affinity to the α2-antagonist idazoxan and similar values to those obtained in human α2A receptors (79-126 nM) in, 51 who also reported lower affinity at human α1 and unmeasurable affinity at human β1 or β2-adrenoceptors. Of note, 51 also reported similar values for mirtazapine for human and rat receptors, TA B L E 4 Log K D values of antipsychotics binding to the human α2A, α2B, and α2C-adrenoceptors. Values represent mean ±s.e.mean of n separate experiments. Selectivity ratios are also given where a ratio of 1 demonstrates no selectivity for a given receptor subtype over another. Compounds are arranged in order of α2A-selectivity. Note: app = apparent affinity. The maximum concentration of competing ligand inhibited most but not all of specific binding. An IC 50 was determined by extrapolating the curve assuming that all specific binding would be inhibited if a higher concentration of competing ligand were possible. ep = early plateau, the competing ligand did not fully inhibit specific binding and the inhibition curve reached a plateau of maximal inhibition of binding. The specific binding inhibited by pimozide was 79.1 ± 6.0% at α2A. Interestingly, many tricyclic antidepressants had a slight α2Bselectivity, something not seen with most α-ligands (Table 2), with the most potent (amitriptyline) having an α2B-affinity (76 nM) only 10-fold lower than that at the α1A-adrenoceptor. Vortioxetine was the only antidepressant with any significant β-adrenoceptor affinity and the only to have β-adrenoceptor affinity greater than αadrenoceptor affinity (178 nM for the β2-adrenoceptor).

| Conclusion
This study, using identical methods to previous α1 and βadrenoceptor studies, allows comparison of ligand affinity, and thus selectivity, between the α and β-adrenoceptor subtypes. Overall, there is huge variation in the literature for the affinity of α2 ligands (more so than for α1 or β), and for which species differences appear to play a significant role, but technique may also be important. Whilst selective antagonists exist for α1A, α1D, β1, and β2-adrenoceptor, there are few selective α2-adrenoceptor ligands and for those that do exist (BRL44408 for α1A and MK-912 for α2C) only have small windows of selectivity. Antidepressants (with the exception of mirtazapine) and first-generation antipsychotics have higher α1A than α2-adrenoceptor affinity. Second-generation antipsychotic varied widely in their α2-adrenoceptor affinity, however, this study does not lend much support for an important role for an α2C-selective action for certain antipsychotics. Clearly, however, even after a century of yohimbine use, there remains plenty of scope to develop selective α2-antagonists.

E TH I C S S TATEM ENT
No animals, human tissue, human volunteers, or patients were used in this study.

ACK N OWLED G EM ENTS
The

JGB has been on the Scientific Advisory Board for CuraSen
Therapeutics since 2019.

AUTH O R CO NTR I B UTI O N
JGB designed the research study. RGWP, JA, and JGB 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
Further information and requests for data and reagents should be directed to and will be fulfilled by the corresponding author, Jillian Baker. Please contact jillian.baker@nottingham.ac.uk.

O RCI D
Jillian G. Baker https://orcid.org/0000-0003-2371-8202 F I G U R E 4 Plot of log K D values showing the relative selectivity and affinity for the single most selective ligand at each receptor. Thus SNAP5089 is the most α1A-selective ligand and the length of the line represents the selectivity for α1A over the next closest adrenoceptor affinity. Terazosin, although the "most" α1B-selective ligand has no selectivity. The selectivity of the three most selective α2 ligands is considerably less than that for α1A, α1D, β1, or β2. Compounds within the black circles represent compounds where the log K D is greater than the −3, −4, or −5 stated but included here to demonstrate attempts were made measurement. Data for α1-adrenoceptors are from. 33 β3 data are included for CGP20712A and ICI118551 from 31