Involvement of the first transmembrane segment of human α₂-adrenoceptors in the subtype-selective binding of chlorpromazine, spiperone and spiroxatrine

Abstract

BACKGROUND AND PURPOSE

Some large antagonist ligands (ARC239, chlorpromazine, prazosin, spiperone, spiroxatrine) bind to the human α_2A-adrenoceptor with 10- to 100-fold lower affinity than to the α_2B- and α_2C-adrenoceptor subtypes. Previous mutagenesis studies have not explained this subtype selectivity.

EXPERIMENTAL APPROACH

The possible involvement of the extracellular amino terminus and transmembrane domain 1 (TM1) in subtype selectivity was elucidated with eight chimaeric receptors: six where TM1 and the N-terminus were exchanged between the α₂-adrenoceptor subtypes and two where only TM1 was exchanged. Receptors were expressed in CHO cells and tested for ligand binding with nine chemically diverse antagonist ligands. For purposes of interpretation, molecular models of the three human α₂-adrenoceptors were constructed based on the β₂-adrenoceptor crystal structure.

KEY RESULTS

The affinities of three antagonists (spiperone, spiroxatrine and chlorpromazine) were significantly improved by TM1 substitutions of the α_2A-adrenoceptor, but reciprocal effects were not seen for chimaeric receptors based on α_2B- and α_2C-adrenoceptors. Molecular docking of these ligands suggested that binding occurs in the orthosteric ligand binding pocket.

CONCLUSIONS AND IMPLICATIONS

TM1 is involved in determining the low affinity of some antagonist ligands at the human α_2A-adrenoceptor. The exact mechanism is not known, but the position of TM1 at a large distance from the binding pocket indicates that TM1 does not participate in specific side-chain interactions with amino acids within the binding pocket of the receptor or with ligands bound therein. Instead, molecular models suggest that TM1 has indirect conformational effects related to the charge distribution or overall shape of the binding pocket.

Introduction

α₂-Adrenoceptors regulate a wide range of physiological processes in the CNS and in the target organs of peripheral sympathetic innervation (Link et al., 1996; MacMillan et al., 1996; Altman et al., 1999; Hein et al., 1999; Brede et al., 2002). They are therefore potential drug targets with many clinical applications (e.g. in the treatment of elevated blood pressure and intra-ocular pressure, in alleviation of withdrawal symptoms from opioid and alcohol abuse and as anaesthetic adjuvants in intensive care and surgical settings) (Ruffolo and Hieble, 1994; MacDonald et al., 1997). Humans and other mammals have three α₂-adrenoceptor subtypes (α_2A, α_2B and α_2C) encoded by three distinct intronless genes. Among the three receptor subtypes, 175 of 450–461 amino acid residues are conserved (see Figure S1 of the Supporting information). The extent of conservation is greater within the seven transmembrane (TM) domains, and 29 of the 33 amino acids of TM2–TM7 that constitute the membrane-embedded ligand-accessible surface of the three human α₂-adrenoceptors are identical (Xhaard et al., 2005). This structural similarity is reflected in the highly similar ligand binding properties of the three receptor subtypes (Ruuskanen et al., 2005; Xhaard et al., 2005). Nonetheless, their tissue distributions and mechanisms of regulation differ (Eason et al., 1994; Scheinin et al., 1994; MacDonald and Scheinin, 1995; MacDonald et al., 1997; Saunders and Limbird, 1999; Mansouri et al., 2001). All three human α₂-adrenoceptor subtypes bind the endogenous catecholamines adrenaline and noradrenaline with quite similar affinities, while differences have been found in the binding affinities of some synthetic ligands. Current clinically employed α₂-adrenoceptor drugs (e.g. clonidine and dexmedetomidine), however, show only marginal subtype selectivity, which limits their therapeutic usefulness.

Until the year 2007, the structure of bovine rhodopsin (Palczewski et al., 2000) was the closest available structure to serve as a template for adrenoceptor models, but rhodopsin shares only low sequence identity (∼21%) with the adrenoceptors. The structures of two close homologues of the α₂-adrenoceptors, the human β₂-adrenoceptor (PDB code 2RH1, Cherezov et al., 2007; Rosenbaum et al., 2007) and the turkey β₁-adrenoceptor (2VT4, Warne et al., 2008), have now been reported. They share, in their TM regions, on average 37–43% identical aligned amino acids with the human α₂-adrenoceptor subtypes (Xhaard et al., 2006). The crystal structures that have been construed for these GPCRs, rhodopsin and β-adrenoceptors, demonstrate that they share the same overall design, having a core structure composed of seven α-helical TM segments. They have a ligand binding cavity mainly defined by amino acid side chains from TM2-TM7, thus placing the ligand binding site within the cell membrane (Cherezov et al., 2007; Rosenbaum et al., 2007). The second extracellular loop (XL2) folds as a β-hairpin and forms a lid over the binding cavity. A disulphide bridge that connects XL2 (Cxl2.50)1 to TM3 (C3.25) in the structures of rhodopsin and the β-adrenoceptors, as well as in the human A_2A adenosine receptor (PDB code 3EML, Jaakola et al., 2008), is highly conserved among all rhodopsin-like GPCRs. In the β-adrenoceptors, XL2 is partly folded as an α-helix and is further stabilized by one additional disulphide bond that is only found in the β-adrenoceptor structures, but not in rhodopsin or in, for example α₂-adrenoceptors (Cherezov et al., 2007; Weis and Kobilka, 2008).

In all types of adrenoceptors (α₁-, α₂- and β-adrenoceptors), a conserved aspartic acid residue located in TM3 (D3.32) provides an anchoring point for the positively charged amine group of agonist ligands (Strader et al., 1987; Wang et al., 1991; Ruffolo and Hieble, 1994; Hein and Kobilka, 1995). S5.42, S/C5.43 and S5.46 in TM5 are involved in hydrogen bond formation with the catechol hydroxyls of catecholamines (e.g. Strader et al., 1989; Peltonen et al., 2003; Xhaard et al., 2006). We have previously mapped the structural determinants of agonist and antagonist binding specificity in the α₂-adrenoceptor subtypes. We employed site-directed mutagenesis, sets of chemically related ligands and sulphydryl-reactive covalently binding ligands in combination with molecular modelling using receptor models based on the bovine rhodopsin structure (Marjamäki et al., 1999; Ruuskanen et al., 2005; Xhaard et al., 2005; Laurila et al., 2007). The main binding site differences among the α₂-adrenoceptor subtypes are located in TM5 and in XL2, where for compounds structurally similar to yohimbine, two residues in XL2 and Cys/Ser²⁰¹ in TM5 (position 5.43) were shown to be major determinants of subtype and species selectivity of ligand binding (Cockcroft et al., 2000; Laurila et al., 2007). However, there is one experimental observation that this model cannot account for; ‘bulky’ antagonists with an extended, branched and relatively rigid structure, and occupying more volume in comparison to other antagonist ligands, such as ARC239, prazosin, spiperone and spiroxatrine, bind to the human α_2A-adrenoceptor with 10- to 100-fold lower affinity than to the α_2B and α_2C subtypes, and differences in TM5 and XL2 do not explain these affinity differences (Cockcroft et al., 2000; Frang et al., 2001; Laurila et al., 2007).

Since the orthosteric binding cavity of α₂-adrenoceptors has already been quite extensively explored by targeted mutagenesis, we decided to search for determinants of antagonist binding specificity from other regions of the protein structures, in particular TM1, which is the most variable of the seven TMs, with only ∼30% sequence identity shared across the α₂-adrenoceptor subtypes in comparison with 60–80% for the other TMs. We made eight chimaeric α₂-adrenoceptor constructs where TM1 was exchanged either with or without the preceding N-terminal sequence. We tested these constructs with nine antagonist ligands. As a result, we showed that TM1 is involved in defining the specific pharmacological profile of the human α_2A-adrenoceptor. The exact mechanism of this influence is yet to be characterized, but molecular modelling based on the crystal structure of the human β₂-adrenoceptor (Cherezov et al., 2007; Rosenbaum et al., 2007) suggests that indirect effects are likely to be involved.

Methods

Drugs and other materials

cDNAs encoding the human α_2A-, α_2B- and α_2C-adrenoceptors were originally provided by Dr BK Kobilka (Stanford University, CA). Primers for DNA construction and sequencing were obtained from TAG Copenhagen A/S (Copenhagen, Denmark). [³H]-RX821002 (1,4-[6,7(n)-³H]benzodioxan-2-methoxy-2-yl)-2-imidazoline), specific radioactivity 58 Ci·mmol⁻¹, was purchased from Amersham Pharmacia Biotech (Buckinghamshire, UK). ARC239 was purchased from Tocris (Bristol, UK). Atipamezole was a gift from Orion Pharma (Turku, Finland). Clozapine, chlorpromazine, idazoxan, oxymetazoline, prazosin, spiperone and spiroxatrine were purchased from Sigma-Aldrich (St. Louis, MO). Fetal bovine serum was from Autogen Bioclear UK Ltd. (Wiltshire, UK). Cell culture reagents were supplied by Life Technologies, Inc. (Rockville, MD). Geneticin (G418 disulphate salt solution), hygromycin B, penicillin, streptomycin, trypsin and bovine serum albumin were purchased from Sigma-Aldrich. Other reagents were of analytical or reagent grade and were purchased from commercial suppliers.

DNA constructs and expression vectors

Eight receptor chimaeras were constructed (Figure 1). The wild-type cDNAs were cloned into pREP4 (α_2A and α_2B) or pcDNA3 (α_2C) expression vectors (Invitrogen Life Technologies, Inc., Rockville, MD). Receptor chimaeras were built utilizing a conserved threonine–serine site at the intracellular end of TM1. Chimaeras based on the α_2A- and α_2B-adrenoceptors were constructed using PCR and two pairs of primers designed for each subtype: a forward primer from the N-terminus of the receptor and a reverse primer until the end of TM1 (containing an artificial SpeI site; recognition sequence ACTAGT, coding for TS); a forward primer from the start of the first intracellular loop (also containing an artificial SpeI site) and a reverse primer until the C-terminus. In addition, the N- and C-terminal primers contained restriction sites for subcloning into the pREP4 expression vector. Fragments were PCR-amplified, digested and ligated into pREP4 for expression in mammalian cells.

Figure 1.

Schematic presentation of chimaeric receptor constructs. Sequences of the wild-type human α_2A-adrenoceptor, wild-type human α_2B-adrenoceptor and wild-type human α_2C-adrenoceptor are shown. Receptors are named based on the origin of the sequence of the N-terminal segment, the TM1 domain and the body of the receptor [e.g. (BBA) for the human α_2A-adrenoceptor containing the N-terminal segment and TM1 of the human α_2B-adrenoceptor].

Chimaeras based on the α_2C-adrenoceptor were constructed using the Gene Editor™in vitro Site-Directed Mutagenesis System (Promega, Madison, WI); the nucleotide sequence encoding the threonine and serine residues at the end of TM1 was mutated to contain an SpeI site. Subsequently, the α_2C-adrenoceptor-based chimaeras were constructed using the artificial SpeI site and PCR-amplified TM1 (and N-terminal) fragments of the α_2A- and α_2B-adrenoceptors, and the digested fragments were ligated and subcloned into the pcDNA3.1(+) expression vector.

The α_2A- and α_2C-adrenoceptor chimaeras with only TM1-region substitutions were constructed with the GeneEditor™in vitro Site-Directed Mutagenesis System and the previous chimaeras as templates; the nucleotide sequence coding for the conserved tyrosine-serine pair at the end of the N-terminus was mutated to contain an NheI site (recognition sequence GCTAGC, coding AS). Subsequently, using this artificial NheI site, the N-terminal fragments were digested, isolated from agarose gel and ligated to the appropriate α_2A- and α_2C-adrenoceptor-based chimaeras. Thereafter, the artificial NheI sites were restored to contain the original tyrosine–serine pair.

Cell culture and transfections

All plasmid DNA constructs were sequenced with vector- and gene-specific primers in order to confirm the successful construction of the desired chimaeras and to exclude undesired mutations. Adherent CHO cells (K1 strain) (American Type Culture Collection, Manassas, VA) were transfected and cultured as reported previously (Laurila et al., 2007). Briefly, cells were cultured in α-MEM medium (α-Minimum Essential Medium, Gibco™, Invitrogen, Carlsbad, CA) supplemented with 26 mM NaHCO₃, 50 IU·mL⁻¹ penicillin, 50 µg·mL⁻¹ streptomycin and 5% heat-inactivated fetal bovine serum supplemented with the appropriate selection antibiotic. Cells were grown in a humidified 5% CO₂ atmosphere at 37°C. Stable cell lines expressing the wild-type cDNAs encoding human α_2A-, α_2B- and α_2C-adrenoceptor subtypes were produced previously by Pohjanoksa et al. (1997) using the expression vector pMAMneo (Clontech, Palo Alto, CA) that contains a neomysin analogue (Geneticin®, G418) resistance gene. The chimaeric pREP4-based receptor constructs were transfected into CHO cells using the commercial Lipofectamine 2000 kit (Invitrogen), whereas the pcDNA3-based expression constructs were transfected with the FuGENE® HD Transfection Reagent kit (Roche Applied Science, Indianapolis, IN). Transfection mixtures were prepared in serum- and antibiotic-free α-MEM and contained 1–5 µg of plasmid DNA. After 24 h incubation, the pREP4-based transfected cell cultures were expanded using hygromycin B (600 µg·mL⁻¹), and the pcDNA3-based transfections were expanded using G418 (800 µg·mL⁻¹) for selection. The resulting cell cultures were screened for their capacity to bind the α₂-adrenoceptor antagonist radioligand [³H]-RX821002 (see below). Cell cultures with the highest levels of receptor expression were chosen for further experiments and were maintained in 200 µg·mL⁻¹ of the appropriate selection antibiotic.

Cells transfected with the α_2C-adrenoceptor-based plasmid construct containing TM1 and the N-terminus from the α_2B-adrenoceptor failed to show [³H]-RX821002 binding in spite of several transfection experiments with several batches of plasmid DNA. Total RNA was isolated from transfected, antibiotic-resistant CHO cells and converted to cDNA with the DyNAmo™ cDNA Synthesis kit (FinnZymes, Espoo, Finland). PCR amplification was performed with gene-specific primers (see the Supporting information, Figure S2), and control reactions were run using RNA from non-transfected cells.

Membrane preparations

Cell membranes were prepared as described previously (Pohjanoksa et al., 1997) and stored at −78°C. Protein concentrations were determined with the method of Bradford (1976) with BSA as a reference.

Saturation binding assays

Receptor expression levels were determined with saturation binding experiments as described previously (Halme et al., 1995) with [³H]-RX821002 as radioligand. Cell membranes (5–10 µg protein) were incubated with serial dilutions (0.0625–8 nM) of the radioligand. Bound radioactivity was measured with a Wallac 1410 liquid scintillation counter (Perkin-Elmer Wallac, Turku, Finland). Non-specific binding was defined in parallel tubes with an excess of phentolamine (10 µM). Specific binding was defined as the difference between total and non-specific binding. Saturation binding experiments were performed in triplicate and repeated at least three times for each receptor type. Receptor densities (B_max) and equilibrium dissociation constants (K_d) were calculated with GraphPad Prism 4.0 software (San Diego, CA).

Competition binding assays

Competition binding assays were implemented using a MultiScreen Vacuum Manifold system (Millipore Corporation, Bedford, MA) with Millipore MultiScreen MSFBN 96-well glass fibre filtration plates. The experiments were performed in a total assay volume of 180 µL (in 50 mM potassium phosphate buffer, pH 7.4) using [³H]-RX821002 at concentrations close to its K_d for each receptor variant, six to eight serial dilutions of the competitor ligands and membrane preparations containing 2–10 µg of protein per sample. Non-specific binding was determined in parallel wells in the presence of 100 µM oxymetazoline. Bound radioactivity was measured with a Wallac 1450 MicroBeta scintillation counter.

Nine antagonist ligands were chosen for the competition binding assays on the basis of their chemical diversity: four ‘bulky’ antagonists (ARC239, prazosin, spiperone, spiroxatrine), two imidazoles (atipamezole, idazoxan), two tricyclic antipsychotic drugs (chlorpromazine, clozapine) and the yohimbine stereoisomer rauwolscine. The apparent affinity (apparent K_i) of each ligand was determined using nonlinear regression analysis (GraphPad Prism), assuming one-site binding. For conversion of IC₅₀ estimates to K_i values, the Cheng–Prusoff equation was applied (Cheng and Prusoff, 1973). The statistical significance of differences between the chimaeras and the corresponding wild-type receptors was evaluated with unpaired t-tests.

Sequence alignments, molecular modelling, docking simulations

Molecular models of the α₂-adrenoceptor subtypes were constructed using the software Modeller 8v2 (Sali and Blundell, 1993) as described previously (Xhaard et al., 2005), but using the 2.4 Åβ₂-adrenoceptor structure (PDB code 2RH1; Cherezov et al., 2007) as the template. First, pairwise sequence alignments of the human β₂-adrenoceptor sequence with the amino acid sequences of each of the three human α₂-adrenoceptor subtypes were constructed using Malign (Johnson and Overington, 1993). Regions that were too dissimilar to be aligned (i.e. the N-terminal segment and the third intracellular loop) were deleted from the alignments and were not included in the models. During the modelling procedure, the sequence alignments were modified in order to ensure that the conformation of the second extracellular loop (XL2), as built by Modeller 8v2, was relaxed and free of any undesirable atomic contacts. A multiple sequence alignment that contains the three pairwise alignments used for modelling is shown in Figure 2. Ten models were constructed for each of the receptor subtypes that varied in the conformations adopted by their side chains and represented the space accessible to the software Modeller (see Xhaard et al., 2005).

Figure 2.

Multiple sequence alignment of the three human α₂-adrenoceptor subtypes together with the human β₂-adrenoceptor sequence as found in the PDB structure 2RH1. α-helices in the human β₂-adrenoceptor are shown as cylinders above the alignment. Stars (*) indicate amino acids that are located less than 4.5 Å away from the ligands bound to each structure (carazolol in the β₂-adrenoceptor) and indicate the location of the ligand binding sites. Positions conserved amongst the aligned sequences are shaded in grey (dark grey: four receptors; light grey: three receptors). Regions with small letters could not be aligned reliably. The third intracellular loop is not shown for clarity.

For molecular docking simulations, we selected only the three ‘bulky’ ligands for which exchange of TM1 had an experimental effect (spiperone, spiroxatrine and chlorpromazine), as well as rauwolscine for which comprehensive mutagenesis-derived information is available (Laurila et al., 2007). The three-dimensional models of the four ligands were prepared as reported previously (Xhaard et al., 2005), with protonation of their central nitrogen atom. Docking was performed using Gold v4.0 (Jones et al., 1997), a programme that allows flexibility in the ligands. Docking simulations were performed with 10 protein models for each of the three receptor subtypes in order to approximate the flexibility of the binding site. Individually for each model and each ligand, 10 docking runs were allowed. The top-ranking structure was further considered. The centre of the search space was taken as an oxygen atom of the conserved D3.32 with a radius of 10 Å around it. The models were ranked in comparison with carazolol bound to the β₂-adrenoceptor crystal structure (Cherezov et al., 2007), by computing a shape and electrostatic overlap index using the programme ShaEP (Vainio et al., 2009).

Results

The N-terminus and TM1 are highly variable regions

We constructed eight chimaeric α₂-adrenoceptors, initially exchanging between the human α₂-adrenoceptor subtypes both the N-terminus and TM1 (subsequently referred to as N_ter-TM1), and later only TM1. Seven of the constructs were successfully expressed and then tested along with the three wild-type receptors against a panel of nine antagonist ligands. TM1 is much more variable than the other six TMs, which is shown in the alignment of the α₂-adrenoceptor subtype sequences presented in Figure 2: of the 29 residues aligned with TM1 of the human β₂-adrenoceptor, in the α₂-adrenoceptors 17 differ between subtypes A and B, 18 differ between A and C and 14 differ between B and C. As seen in the sequence alignment, the length of the N-terminus (until the cut-off used in this study, between Y1.28 and S1.29) differs notably between the α₂-adrenoceptor subtypes, consisting of 28, 7 and 46 amino acids in subtypes A, B and C, respectively. Thus, any structural model of the N_ter-TM1 region is highly speculative, since reliable templates to model the N-terminal segment of each receptor subtype are lacking.

The following eight chimaeras were generated and expressed in CHO cells: each of the subtypes A, B and C with the N_ter-TM1 segments of the other two subtypes (six constructs), and the A and C subtypes with TM1 of the C and A subtypes (two constructs). These exchanges took advantage of a conserved tyrosine–serine pair at the end of the N-terminal region and a conserved threonine–serine pair at the end of TM1.

Expression of receptor chimaeras

All chimaeras, with the exception of the α_2C-adrenoceptor-based construct that contained N_ter-TM1 from the α_2B-adrenoceptor, were successfully expressed in CHO cells and displayed only relatively small differences in their affinity for the radioligand [³H]-RX821002 in comparison with the corresponding wild-type receptor (Table 1). For the wild-type receptors, [³H]-RX821002 showed lower affinity at the human α_2B-adrenoceptor subtype as compared with the A and C subtypes; this 5- to 10-fold affinity difference agrees with previous findings (Deupree et al., 1996; Uhlén et al., 1998). After introduction of the N_ter-TM1 region of the A or C subtype into the α_2B-adrenoceptor subtype, [³H]-RX821002 was bound with threefold higher affinity to both α_2B-based chimaeras in comparison with the wild-type receptor, shifting the affinity towards that of the wild-type A and C subtypes. Cells transfected with the α_2A-based construct containing N_ter-TM1 of the B subtype showed a reciprocal effect, with an almost threefold decrease in affinity relative to the A subtype (i.e. a shift towards that of the wild-type B subtype). With the α_2C-adrenoceptor-based construct containing N_ter-TM1 from the α_2B-adrenoceptor, binding of [³H]-RX821002 was not detectable, even if the successful transfection of the chimaeric cDNA was confirmed by reverse transcription PCR (results not shown).

Table 1.

Binding affinities of [³H]RX821002 and receptor densities in recombinant CHO cell lines (means ± SEM)

Receptor	Kd (nM)	Bmax (fmol·mg−1)
α2A AAA	0.45 ± 0.04	6 700 ± 100
α2A BBA	1.2 ± 0.7	400 ± 260
α2A CCA	0.39 ± 0.06	5 600 ± 500
α2A ACA	0.27 ± 0.02*	19 000 ± 3000
α2B BBB	4.6 ± 1.0	1 500 ± 300
α2B AAB	1.4 ± 0.1*	1 600 ± 100
α2B CCB	1.6 ± 0.2*	6 100 ± 500
α2C CCC	0.87 ± 0.06	1 800 ± 0
α2C AAC	0.71 ± 0.04	470 ± 10
α2C BBC	n.d.	n.d.
α2C CAC	0.55 ± 0.15	710 ± 80

Receptors are named based on the exchanged domains (see Figure 1). Statistical significance of differences between the chimaeric receptors and the respective wild-type receptors were tested with Student's t-test:

^*P < 0.05.

Abbreviations: A (α_2A); B (α_2B); C (α_2C); n.d., not determined.

Characterization of antagonist binding profiles

The binding affinities of the antagonist ligands ARC239, prazosin, spiperone, spiroxatrine, chlorpromazine, atipamezole, idazoxan, clozapine and rauwolscine at the different receptor chimaeras and wild-type α₂-adrenoceptors are shown in Table 2. Five ligands (ARC239, prazosin, spiperone, spiroxatrine and chlorpromazine) showed significantly lower affinity for the A subtype in comparison to the B and C subtypes, as evidenced by 10- to 100-fold differences in the binding affinities. The other antagonists did not show significant subtype selectivity between the wild-type receptors and were therefore used as controls.

Table 2.

Competition binding affinities of different ligands obtained with [³H]RX821002 at wild-type and chimaeric α₂-adrenoceptors expressed in CHO cells

			Human α2A
Ligand	WT (AAA)	nH	BBA	nH	CCA	nH	ACA	nH
ARC239	1600 (1000–2600)	0.89	450 (160–1400)	0.66	820 (600–1110)	0.72	760 (480–1200)	1.02
Prazosin	1500 (800–3100)	0.84	1400 (490–2400)	1.16	820 (600–1100)	0.94	910 (600–1500)	0.89
Spiperone	1200 (870–1600)	0.94	870 (450–1800)	0.93	560 (440–730)a	1.03	500 (320–770)aa	0.86
Spiroxatrine	550 (240–1300)	1.33	69 (34–140)aa	0.92	71 (39–130)aa	1.08	91 (48–180)aa	1.51
Chlorpromazine	600 (430–870)	0.82	160 (83–310)aaa	0.89	120 (56–260)aaa	0.68	160 (100–230)aaa	1.07
Clozapine	89 (54–150)	0.91	67 (32–150)	0.69	34 (17–67)	0.58	51 (35–77)	1.13
Atipamezole	2.1 (1.5–3.1)	0.89	2.0 (0.99–4.3)	1.00	1.2 (0.95–1.5)	0.82	1.1 (0.77–1.5)	0.97
Idazoxan	22 (15–31)	0.80	15 (2.7–45)	0.97	17 (13–24)	0.91	13 (8.3–22)	0.85
Rauwolscine	1.8 (0.78–4.5)	0.53	4.4 (2.0–9.8)	0.66	0.85 (0.40–2.1)	1.02	1.6 (0.82–3.1)	0.71

			Human α2B
Ligand	WT (BBB)	nH	AAB	nH	CCB	nH
ARC239	150 (40–570)bbb	0.74	240 (120–480)	0.72	98 (48–220)	0.74
Prazosin	47 (17–140)bb	0.79	200 (120–360)	0.82	170 (89–310)	0.89
Spiperone	12 (3.7–38)bbb	0.51	26 (11–65)	1.65	14 (6.0–38)	0.67
Spiroxatrine	2.4 (1.2–5.0)bbb/cc	0.64	6.8 (1.8–28)	2.03	1.6 (0.64–5.9)	0.91
Chlorpromazine	43 (20–100)bbb/cc	0.64	53 (20–200)	1.84	46 (18–120)	0.70
Clozapine	12 (5.0–28)bbb	0.93	18 (8.7–45)	1.09	6.7 (1.9–23)	0.64
Atipamezole	2.7 (0.56–14)	0.85	3.0 (1.7–5.6)	0.86	1.7 (0.80–4.0)	0.65
Idazoxan	24 (18–34)	1.08	37 (29–48)	0.95	30 (19–51)	0.88
Rauwolscine	1.1 (0.7–1.8)	0.97	1.2 (0.66–2.4)	1.25	2.7 (1.1–6.9)	1.02

			Human α2C
Ligand	WT (CCC)	nH	AAC	nH	BBC	CAC	nH
ARC239	130 (66–260)ddd	1.15	180 (86–380)	0.89	n.d.	77 (34–180)	0.65
Prazosin	45 (21–81)dd	0.67	110 (50–230)	0.78	n.d.	120 (55–260)	0.56
Spiperone	29 (9.2–81)ddd	0.66	26 (10–75)	0.63	n.d.	39 (18–89)	0.79
Spiroxatrine	13 (5.9–28)ddd	1.13	12 (5.8–27)	1.05	n.d.	18 (8.4–31)	0.68
Chlorpromazine	260 (150–350)dd	0.81	260 (110–720)	0.42	n.d.	230 (130–400)	0.80
Clozapine	6.5 (2.5–18)ddd	0.75	5.1 (2.6–8.2)	0.57	n.d.	1.8 (0.2–17)	0.33
Atipamezole	4.1 (2.0–9.0)	0.92	5.3 (4.8–8.3)	0.82	n.d.	10 (5.0–23)	1.16
Idazoxan	71 (20–251)	0.71	48 (21–110)	0.73	n.d.	18 (12–34)	0.89
Rauwolscine	0.47 (0.21–1.2)	0.83	0.27 (0.12–0.63)	0.78	n.d.	0.39 (0.20–0.83)	1.14

The apparent K_i (nM) and their 95% confidence intervals from three to six independent experiments, analysed using a one-site competition model. The pseudo-Hill slope (n_H) was obtained using a variable slope model. Receptors are named based on the exchanged domains (see Figure 1). Statistical significances are shown with symbols (one) P < 0.05; (two) P < 0.01; (three) P < 0.001. (a) chimaera versus wild-type α_2A; (b) wild-type α_2A versus wild-type α_2B; (c) wild-type α_2B versus wild-type α_2C; (d) wild-type α_2A versus wild-type α_2C.

In the chimaeric receptor based on the α_2A-adrenoceptor where N_ter-TM1 of the B subtype had been inserted, the binding affinities were significantly improved for two ligands (spiroxatrine and chlorpromazine), and in the case of the insertion of N_ter-TM1 from the C subtype into the α_2A-adrenoceptor, the binding affinities of three ligands (spiperone, spiroxatrine and chlorpromazine) were significantly improved in comparison with the corresponding wild-type receptor. Similar tendencies were also seen for ARC239 and prazosin, but they failed to reach statistical significance. In order to see which region of N_ter-TM1 was responsible for the improved affinity, we created an α_2A-adrenoceptor construct where only TM1 had been imported from the C subtype and the original N-terminus of the α_2A-adrenoceptor was retained. The binding affinities did not differ between this TM1-only chimaera and the previous N_ter-TM1 chimaera, pointing to the involvement of TM1 in the observed binding affinity changes.

In chimaeras based on α_2B- and α_2C-adrenoceptors and containing N_ter-TM1 from the A subtype, or the α_2C-adrenoceptor chimaera with only TM1 of the A subtype, no significant changes in ligand affinities were observed. Insertion of N_ter-TM1 from the C subtype into the α_2B-adrenoceptor did not result in notable differences in the binding affinities.

The β₂-adrenoceptor structure as template for α₂-adrenoceptor models

Our previous α₂-adrenoceptor models (Ruuskanen et al., 2004; Xhaard et al., 2005; 2006; Laurila et al., 2007) were based on the X-ray structure of bovine rhodopsin (Palczewski et al., 2000), but these proteins share only 20–23% sequence identity in their TM regions. The recently reported structures of the turkey β₁-adrenoceptor and the human β₂-adrenoceptor, unlike rhodopsin, share with the α₂-adrenoceptors both higher levels of sequence identity – 37–43% in the TM regions – and higher similarity of some specific amino acid residues shown by other methods to be involved in the recognition of catecholamines (Xhaard et al., 2006). The β₂-adrenoceptor structure (2.4 Å resolution; Cherezov et al., 2007), closely related in evolution to the α₂-adrenoceptors, is therefore likely to allow construction of more reliable molecular models of the α₂-adrenoceptors than the bovine rhodopsin structure.

The TM regions of the β₂-adrenoceptor structure are, nonetheless, well conserved in comparison with the rhodopsin structure: the Cα atoms of the TM helices of the β₂-adrenoceptor structure can be superimposed on those of rhodopsin with a root mean square deviation (RMSD) of 3.9 Å (Cherezov et al., 2007). All of the amino acid positions that were predicted to face the binding cavity in the rhodopsin-based α₂-adrenoceptor models are indeed positioned there in the β₂-adrenoceptor-based ones, too (Figure 3). For example, the key determinants of agonist binding, D3.32, W6.48, S5.43 and S5.46, are well positioned to interact with the protonated nitrogen, aromatic ring, and catecholic hydroxyls of catecholamine agonists. In the β₂-adrenoceptor structure, along the TM1-TM5 axis, TM5 is located closer to the centre of the binding cavity by about 2 Å in comparison to the rhodopsin structure, and this is reflected in our current models of the α₂-adrenoceptors. The distance between the Cα carbon of residue 3.32 and the Cα carbon of residue 5.42 is 12 Å in the β₂-adrenoceptor, ∼12.5 Å in our α₂-adrenoceptor models, and 14.5 Å in rhodopsin. The distance along the axis of the cavity that runs from TM3 to TM6 is slightly longer in the β₂-adrenoceptor structure than in rhodopsin, leading to a wider cavity. The distance between the Cα carbons of residues 3.32 and 6.52 is 12.5 Å in the rhodopsin structure and 13.5 Å in the β₂-adrenoceptor structure and the α₂-adrenoceptor models. Also, the TM1–TM5 distance is about 23.5 Å in rhodopsin and 26.5 Å in the adrenoceptors (Figure 3A).

Figure 3.

Comparison of molecular models of the α_2A-adrenoceptor subtype based on either bovine rhodopsin (grey) or the β₂-adrenoceptor (green) as template. (A) Amino acids of the TM helices were globally well positioned in the rhodopsin-based model. For clarity, not all amino acids that face the binding cavity are shown. Extracellular loop 2 forms the top of the binding pocket and is quite different in rhodopsin-based models (grey) and β₂-adrenoceptor based models (green) (B and C). (B) The general structure of XL2, viewed from the plane of the membrane. (C) Close-up view of positions xl2.48–xl2.54. In (B and C), the view has been tilted by about 70° compared with A.

The cysteine at position xl2.50 has similar roles in the β₂-adrenoceptor and in rhodopsin, forming a disulphide bond with the cysteine at position 3.50 and constraining XL2 to fold on top of the binding cavity. In the β₂- and β₁-adrenoceptor structures (Cherezov et al., 2007; Warne et al., 2008), XL2 forms a hairpin that contains an α-helical segment (residues H150 to A158) and a second internal disulphide bridge between C184 (xl2.43) and C190 (xl2.49) that is not present in the α₂-adrenoceptor subtypes or in rhodopsin. In the β₂-adrenoceptor structure, the segment of XL2 that lines the binding cavity (residues xl2.50 to xl2.53) is located clearly deeper by about 4 Å towards the centre of the binding cavity in comparison to rhodopsin (Figure 3B). Consequently, this segment is also shifted in the α₂-adrenoceptor models versus rhodopsin away from TM5 and towards TM1: e.g. the Cα carbons at positions xl2.52 and Cxl2.50 are shifted towards the extracellular surface by 4.0 Å and 5.8 Å, and towards TM1 by 2.5 Å (Figure 3C).

The differences among the three α₂-adrenoceptor subtypes in the amino acids facing the binding cavity that we have previously reported from the rhodopsin-based models (Laurila et al., 2007) are also found in the present β₂-adrenoceptor-based models: i.e. C/S/C variation at position 5.43 and variation of the portion of XL2 lying on top of the binding cavity, composed of residues xl2.49-xl2.52 (RCEI, QCKL, and QCGL in the A, B and C subtypes, respectively). Nonetheless, as a consequence of the shift of XL2 towards TM1, position xl2.49 is less exposed than previously thought, whereas position xl2.54 (D/Q/D) now also contributes to the binding cavity surface.

Docking of antagonists to the α₂-adrenoceptor model structures

Automated docking was used to investigate possible binding modes at the α₂-adrenoceptors of three antagonists (spiperone, spiroxatrine and chlorpromazine) that showed significant affinity differences in experiments with the chimaeric receptor models (Figures 4 and 5). Rauwolscine was also included in these binding simulations since it has been extensively studied in previous work (Laurila et al., 2007). Docking was performed with each of the ten model structures built for each of the three α₂-adrenoceptor subtypes in order to account for receptor side-chain flexibility (Xhaard et al., 2005). The docked molecules have molecular masses ranging from 354 to 395 g·mol⁻¹, and they all have a centrally located protonated nitrogen atom that has been proposed to form an ion pair with a side-chain oxygen of D3.32.

Figure 4.

Example of the diversity of ligand positions obtained by automated docking with the programme Gold 4.0, and their distance to TM1 (A) viewed through the membrane plane and (B) rotated by about 70°. In green, all first-ranking positions obtained for chlorpromazine are docked to the ten models of the α_2A-adrenoceptor subtype. A similar diversity of positions is observed for the α_2B- and α_2C-adrenoceptor subtypes (not shown). In yellow, the equivalent location of carazolol is shown in the β₂-adrenoceptor structure. When the positively charged nitrogen of chlorpromazine is at ion pair distance of the negatively charged side-chain oxygen of D3.32, a good superimposition of carazolol is found but the docked molecules do not reach TM1 at the left of the figure (blue). Furthermore, direct contact to TM1 is blocked by side chains from TM2 and TM7 (grey).

Figure 5.

Close-up view of the ‘best’ positions obtained by automated docking of rauwolscine (A, B), chlorpromazine (C, D), spiroxatrine (E, F) and spiperone (G, H) to the α_2A-adrenoceptor model. The poses were ranked according to the shape and electrostatic overlap on the most similar ligand in the β₂-adrenoceptor, either carazolol (rauwolscine, chlorpromazine) or timolol (spiroxatrine, spiperone), using the programme ShaEP (Vainio et al., 2009; see Supporting information). Left panels (A, C, E, G) show the binding site (grey) together with the location of either carazolol or timolol (yellow volumes). Right panels (B, D, F, H) show the actual superimposition of ligands (yellow and green) together with the side chains of D3.32, S5.42, C5.43 and Exl2.51. For spiroxatrine and spiperone (E–H), two alternative orientations were found as described in Xhaard et al. 2005. These binding modes are shown in Figure S3.

It is apparent from the docking simulations that all the antagonists investigated, when in contact with D3.32 within the agonist binding site, are too far away from TM1 to form direct contacts with it (Figure 4). Furthermore, access to TM1 from the cavity where agonists are known to bind is occluded by amino acid side chains from TM2 (2.58, 2.61 and 2.65) and TM7 (7.36, 7.40 and 7.43). Another observation from the docking simulations is that there was significant variation in the possible binding modes that were suggested by the docking software (Figure 4). Nonetheless, not all of these binding modes formed the required interaction with D3.32, and many were shifted away from the binding cavity identified in the β₂-adrenoceptor. These different binding modes are not supported by the similarities present in the ligands (compare, e.g. the three-dimensional structures of carazolol and chlorpromazine; see Figure 5) and by the effects of mutations of TM5 on the affinity of rauwolscine (Laurila et al., 2007). Thus, in order to select the most parsimonious binding modes (Figure 5), we used as a criterion the overlap of our ligands with either carazolol or timolol bound to the β₂-adrenoceptor, computed by the programme ShaEP (Vainio et al., 2009). As a result, extremely close superimposition with the poses based on the X-ray structure of the β₂-adrenoceptor was found for the conformations suggested by automated docking of our ligands (Figure 5; see the Supporting information for more detail on docking results). We did not observe subtype-specific clusters of binding modes, which very likely reflects the highly similar binding cavities present in the three α₂-adrenoceptor subtypes and should not be interpreted to imply faults in the theoretical model structures.

Discussion and conclusions

Previously, using site-directed mutagenesis of human and murine α_2A-adrenoceptors in combination with molecular models based on the X-ray structure of bovine rhodopsin, we have investigated the effects of mutations in TM5 and XL2 (Laurila et al., 2007). We found that single amino acid substitutions at positions 5.43 and xl2.49-xl2.51 could reverse the inter-species affinity differences of yohimbine, rauwolscine and RS79948-197. The species-specific binding affinities of antagonists with more extended molecular structures were not affected by these mutations, indicating that other residues were influencing their binding affinities. Based on this observation, and by the much lesser sequence conservation of TM1 compared with the other TMs, TM1 was suspected of having an influence on the binding affinity of some large antagonist ligands.

Involvement of TM1 in ligand binding

We prepared eight chimaeric α₂-adrenoceptor constructs: six in the first stage of the study where TM1 and the preceding N-terminal sequence were exchanged between the subtypes, and in the second stage of the study, two chimaeras where only TM1 was exchanged in order to discriminate between any effects due to the N-terminus and TM1. The result was a significant gain of binding affinity for the α_2A-adrenoceptor subtype as regards three of the five ligands that had relatively low affinity for this receptor subtype; for the remaining two antagonists, a similar but statistically non-significant trend was observed. No reciprocal effects were seen in the receptor chimaeras based on the α_2B- and α_2C-adrenoceptors, where the ligand binding affinities were only marginally affected by the substitutions. In the second stage of the study, we showed that the effects seen for the constructs based on the α_2A-adrenoceptor could be ascribed to TM1, and not to the N-terminal segment.

There is some previous experimental evidence from other amine GPCRs for the involvement of TM1 in ligand binding. For the histamine H₁-receptor, two amino acids in TM1, L1.35 and L1.39, have been proposed to participate in the binding of dimeric histaprodifen (Strasser et al., 2008a). For a closer homologue of the α₂-adrenoceptors, the dopamine D₂ receptor, the substituted cysteine accessibility method has been used to demonstrate that six amino acids in TM1 (A1.36, G1.49, N1.50, V1.53, A1.56 and V1.57) are at least transiently accessible to ligands (Shi et al., 2001), whereas in rhodopsin, two amino acids in TM1 (G1.49 and N1.50) were predicted to be accessible in the binding pocket (Shi et al., 2001). Thus, in the α₂-adrenoceptors, TM1 could also be accessible to ligands that enter the binding cavity, even if simultaneous binding to TM3 and TM1 now appears unlikely.

Docking site and accessibility of TM1

Most available experimental evidence suggests that antagonists bind at least in part within the same pocket where agonists bind in α₂-adrenoceptors and related receptors: ligands that form ion-pairs with D3.32 need to fit into an equivalent of the orthosteric pocket close to TM5. The β₁- and β₂-adrenoceptor structures were determined in complex with three ligands (carazolol, cyanopindolol and timolol) relatively similar to ours that all bind within the orthosteric pocket (Cherezov et al., 2007; Hanson et al., 2008; Warne et al., 2008). Phenoxybenzamine, which binds covalently to α₁- and α₂-adrenoceptors, has been shown to attach within the orthosteric site (Frang et al., 2001; Xhaard et al., 2005). For yohimbine analogues, contacts with TM5, TM7 and xl2.49-xl2.51 have been experimentally demonstrated (Suryanarayana et al., 1991; Laurila et al., 2007); direct interactions with TM1 are not likely since TM1 is so far away from TM5. Other experimental evidence based on site-directed mutagenesis also points to the orthosteric pocket as the binding location of antagonists in both α₁-adrenoceptors and 5-HT receptors that are structurally related to α₂-adrenoceptors (Zhao et al., 1996; Wurch et al., 1998).

The presently employed α₂-adrenoceptor models were constructed based on sequence identity of 37–43% with the β₂-adrenoceptor template and should be more accurate than our previous models based on bovine rhodopsin. Nonetheless, one should be aware that molecular receptor models may have inaccuracies resulting from the evolutionary divergence between the modelled structure and the template, as well as errors introduced during the modelling procedure. Generally, molecular models of GPCRs are relatively accurate in the TM regions whereas the extracellular and cytoplasmic loop domains are more difficult to model (Michino et al., 2009). Furthermore, molecular models represent only static views of receptors that are flexible entities in cells. Nonetheless, in the present case, the β₂-adrenoceptor that was used as the model template was bound to the partial inverse agonist carazolol (Cherezov et al., 2007) that, in light of current knowledge, should be optimal for building reasonable molecular models of the α₂-adrenoceptors in complex with antagonist ligands.

As such, our molecular models suggest very limited exposure of TM1 to the binding pocket, as TM1 is distant from the orthosteric binding cavity. Since the TM1–TM5 distance now is longer than in previous rhodopsin-based models (Figure 3) (26.5 Å in the β₂-adrenoceptor vs. 23.5 Å in rhodopsin), small-molecule ligands that occupy the agonist (orthosteric) site close to TM5 cannot simultaneously be in contact with TM1. Furthermore, access to TM1 is blocked by side-chain atoms from TM2 and TM7 (Figure 4). We propose that the β₂-adrenoceptor structure should be used to re-assess other docking simulation studies that have reported TM1 binding in amine receptors in addition to contacts with the conserved D3.32 in TM3. Such predictions include, for example, three residues (Y1.39, L1.42 and I1.43) in TM1 of the dopamine D₂ receptor that were proposed to bind spiperone (Boeckler et al., 2005) and three aliphatic residues in TM1 that were suggested to contribute to the binding cavity in a chemogenomic analysis of the amine GPCR family (Surgand et al., 2006).

Molecular docking to β₂-adrenoceptor-based models

That rauwolscine binds in the orthosteric pocket of α₂-adrenoceptors has by now been confirmed by several site-directed mutagenesis studies. In addition to ion-pairing with 3.32, direct contacts have been suggested at positions 7.39 and 5.43. Experimental results from substitution of F7.39 with an asparagine and of C5.43 with a serine show increased binding affinity of yohimbine at the human α_2A-adrenoceptor (Suryanarayana et al., 1991; Cockcroft et al., 2000). In addition, we have previously shown that substitutions at xl2.49 and xl2.51 significantly and reciprocally affect the binding of rauwolscine (Laurila et al., 2007). In our new molecular models based on the β₂-adrenoceptor, together with xl2.50 and xl2.52, position xl2.51 forms the aromatic ‘lid’ of the binding pocket. When comparing the automated docking of the extended structures of chlorpromazine, spiperone and spiroxatrine to that of the more curved rauwolscine, only rauwolscine comes into close contact with the side chain of xl2.51 (Figure 5). This supports the suggested binding modes as it provides a structural explanation for our previously observed effects of the side chains of XL2 to the binding of yohimbine but not chlorpromazine, spiperone and spiroxatrine.

The side chain of xl2.51 is, however, facing away from the binding cavity. In the β₂-adrenoceptor structure, the carboxyl group of D192 (xl2.51) forms a salt bridge with a positively charged lysine (K7.32) in TM7. The amino acids in the α₂-adrenoceptors at equivalent positions are in the α_2A-adrenoceptor Exl2.51 and R7.32, in the α_2B-adrenoceptor Kxl2.51 and H7.32 and two glycines at positions xl2.51 and 7.32 in the α_2C-adrenoceptor. Ligands, by affecting this interaction, may have allosteric effects, as recently reported for the β₂-adrenoceptor (Bokoch et al., 2010). An important earlier study that used the rhodopsin activation mechanism (Robinson et al., 1992) as a starting point had, furthermore, shown that exchange of the positively charged K7.32 to A (neutral) or E (negatively charged) in the α_1B-adrenoceptor increased the binding affinity of noradrenaline and adrenaline, suggesting this ion-pair interaction to be important for the resting receptor conformation (Porter et al., 1996). Thus, the indirect effects that we observed in this study, as well as the effects of XL2 mutations on the binding of yohimbine analogues (Laurila et al., 2007), may be mediated by this ionic interaction between XL2 and TM7 in the α_2A-adrenoceptor subtype. This positively charged arginine (R7.32) in TM7 of the human α_2A-adrenoceptor has so far not been subjected to mutagenesis experiments.

An effect on binding does not necessarily imply direct atomic contact

Our experimental results show that TM1 contributes to the binding properties of some large antagonist ligands at the α_2A-adrenoceptor, but reciprocal effects were not seen for α_2B- and α_2C-adrenoceptors. Binding affinity is, however, determined by more factors than the direct interactions of the ligand and the receptor (approximated by the enthalpic component), as it also includes the dynamics of different segments of the receptor and the loss of flexibility upon ligand binding (approximated by the entropic component) (Wittmann et al., 2009). Other factors that have been discussed include the presence of charged groups even outside of the pocket that affect the pK_a and hence the protonation state of compounds; the charged side chains do indeed differ between the α₂-adrenoceptor subtypes in the vicinity of the binding pocket. Side chains along TM1 may influence the packing of side chains on other TMs that pack against it and hence may affect the overall shape of the binding pocket. Thus, some regions outside of the orthosteric binding pocket may very well be involved in ligand binding specificity without direct contacts and direct effects on the binding pose. For example, from a recent structural analysis of the histamine H₁ receptor it has been proposed that its N-terminus may contribute indirectly to the pharmacological properties of certain types of ligands by indirect interactions with the second extracellular loop (Strasser et al., 2008b). Also, three TM6 amino acids facing away from the cavity, and several residues in the third intracellular loop of the melanocortin receptors MC₁ and MC₄ were, in a mutagenesis study, identified to account for most of the subtype difference in the affinity for α-MSH (Prusis et al., 2006).

Concluding remarks

We have experimentally demonstrated that TM1 is involved in determining the subtype-specific antagonist binding profile of the α_2A-adrenoceptor subtype. Our improved α₂-adrenoceptor models based on the β₂-adrenoceptor structural template suggest that ligands cannot bind simultaneously in the agonist binding pocket between TM3 and TM5 and be in direct contact with TM1. Furthermore, molecular modelling of the α₂-adrenoceptors suggests very similar binding cavities in the A, B and C subtypes. Such highly similar cavities would be expected to accommodate ligands in a similar manner. Thus, the pharmacological effects of the mutations introduced in this study are best explained by indirect effects. A better understanding of such effects may provide input into the design of highly subtype-selective new drugs.

Glossary

Abbreviations

B_max

receptor density

TM

transmembrane (domain)

XL2

second extracellular loop

Conflicts of interest

MS has received speaker's fees and consultation fees from Orion Corporation, a pharmaceutical company with an active α₂-adrenoceptor programme. He has also contract research relationships with Orion Corporation. JOR has received speaker's fees from Orion Corporation.

Supporting information

Additional Supporting Information may be found in the online version of this article:

Figure S1 Multiple sequence alignment of the three human α2-adrenoceptor subtypes indicating the predicted secondary structure and boundaries of the TM helices as derived from bovine rhodopsin and the β2-adrenoceptor X-ray structures. A multiple sequence alignment of the α2-adrenoceptors with β2-adrenoceptor and rhodopsin was first constructed to build the figure (not shown). The secondary structures are as reported in the PDB files and are displayed on top of the alignments for the β2-adrenoceptor (PDB 2RH1) and rhodopsin (PDB 1U19): cylinders indicate α-helical domains and arrows β-strands. The amino acids of the orthosteric binding site are indicated with an (*) above the respective secondary structure elements and were defined as having at least one atom within 5 Å of the ligand, either carazolol or retinal. Amino acids conserved among the three α2-adrenoceptor subtypes are highlighted in dark grey. Amino acids forming the third intracellular loop are highlighted in light gray.

Figure S2 Top. Ranking of docked poses using a shape similarity index computed by the programme ShaEP (Vainio et al., 2009). Similarity is measured between chlorpromazine, rauwolscine, spiroxatrine and spiperone docked to the 10 molecular models of each α2-adrenoceptor subtype, and carazolol (chlorpromazine and rauwolscine) or timolol (spiroxatrine and spiperone) bound to the β2-adrenoceptor. The TM regions of the β2-adrenoceptor were first superimposed with the equivalent region of each of the α2-adrenoceptor models in order to position carazolol or timolol into the α2-adrenoceptor models. The ‘best’ scores were assigned to poses with the highest molecular overlap on timolol or carazolol (see the dark green and yellow molecules in Figure 4 of the article). The maximum similarity ranges between 0 (no overlap) and, since the ligands differ, a value less than 1: 0.70 (maximum similarity computed between rauwolscine and carazolol when the superposition is optimized by the ShaEP genetic algorithm), 0.80 (chlorpromazine-carazolol), 0.67 (spiroxatrine-timolol) and 0.66 (spiperone-timolol). Only the similarity in molecular shape is reported, as the electrostatic similarity is difficult to interpret. Bottom. For comparison, the equivalent docking scores as computed by Goldscore are also given and show that there is no correlation between the Goldscore and the ranking according to ShaEP.

Figure S3 Alternative binding modes observed for spiroxatrine (A, B) and spiperone (C, D) docked into the α2A-adrenoceptor model. Left panels (A, C) show the binding site (grey) together with the location of timolol (yellow volumes). Right panels (B, D) show the actual superimposition of ligands (yellow and grey) together with the side chains of D3.32, S5.42, C5.43 and Exl2.51. These orientations complement Figure 5.

Table S1 A list of primers used for mutagenesis and RT-PCR. Nucleotides that encode an artificial restriction enzyme recognition site or mutation are highlighted with colour: KpnI (blue), SpeI (red), HindIII (green) and mutation (bold black)

Table S2 Numerical values used to draw Figure S2

Table S3 Key distances between functional groups of docked chlorpromazine, rauwolscine, spiroxatrine or spiperone as compared with carazolol or timolol bound to the β2-adrenoceptor when the TM region of the β2-adrenoceptor is superimposed to the equivalent region of each of the α2-adrenoceptor models. The functional groups compared are annotated as follows: N⁺/N⁺, distances measured between the protonated amine of carazolol or timolol and the protonated amine of chlorpromazine, rauwolscine or spiroxatrine; N⁺/OH, distances measured between the −OH group of carazolol or timolol and the protonated amine of chlorpromazine, rauwolscine or spiroxatrine; -N⁼/S or -N⁼/OH distances measured between the cyclic nitrogen atom of carazolol and the sulphur atom of chlorpromazine or the hydroxyl group of rauwolscine; −S^-/N distances between the sulphur atom of timolol and the cyclic nitrogen of spiroxatrine and spiperone. Distances are only given for the three most similar complexes, as computed by ShaEP

Table S4 Key distances between functional groups of docked ligands and specified amino acids of the α2-adrenoceptor models. For D3.32, distances are measured from the closest side-chain carboxylate oxygen of D3.32 to the protonated amine of ligands; for S5.42, distances are measured from the side-chain hydroxyl to the −OH of rauwolscine or the sulphur of chlorpromazine or the ring nitrogen of spiroxatrine or spiperone; otherwise, the distances are shown for the closest approaching atom between the amino acid and the ligand. Distances are only given for the three most similar complexes as computed by ShaEP

Table S5 Distances (Å) between D3.32 and the protonated amine of chlorpromazine, rauwolscine, spiroxatrine and spiperone. Models are ranked according to their ShaEP scores

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.