Azi-medetomidine: Synthesis and Characterization of a Novel α2 Adrenergic Photoaffinity Ligand

Abstract

Agonists at the α₂ adrenergic receptor produce sedation, increase focus, provide analgesia, and induce centrally-mediated hypotension and bradycardia, yet neither their dynamic interactions with adrenergic receptors nor their modulation of neuronal circuit activity is completely understood. Photoaffinity ligands of α₂ adrenergic agonists have the potential both to capture discrete moments of ligand-receptor interactions and to prolong naturalistic drug effects in discrete regions of tissue in vivo. We present here the synthesis and characterization of a novel α₂ adrenergic agonist photolabel based on the imidazole medetomidine called azi-medetomidine. Azi-medetomidine shares protein association characteristics with its parent compound in experimental model systems and by molecular dynamics simulation of interactions with the α_2A adrenergic receptor. Azi-medetomidine acts as an agonist at α_2A adrenergic receptors, and produces hypnosis in Xenopus laevis tadpoles. Azi-medetomidine competes with the α₂ agonist clonidine at α_2A adrenergic receptors, which is potentiated by photolabeling, and azi-medetomidine labels moieties on the α_2A adrenergic receptor as determined by mass spectrometry in a manner consistent with a simulated model. This novel α₂ adrenergic agonist photolabel can serve as a powerful tool for in vitro and in vivo investigations of adrenergic signaling.

Graphical Abstract

Introduction

Photoaffinity ligands have been used with great success to reveal binding locations, configurations, and even identify previously unknown molecular binding targets by making permanent the natively ephemeral ligand-receptor interaction.^1-3 A second use for anesthetic photolabels recently has been proposed– interrogating neural circuits in vivo by restricting drug to a specific region of tissue through covalent binding.⁴ Taken together, these uses of photolabels provide a novel approach to a multi-tiered investigation of drug actions ranging from molecular targets, to circuit-level effects, and culminating in behavioral consequences.

A₂ adrenergic agonists produce anxiolysis, hypnosis, and analgesia without respiratory depression through activation of endogenous recovery sleep circuits.^5,6 The most commonly clinically-used hypnotic of this class is dexmedetomidine, the dextrorotary isomer of medetomidine. The desirable hypnotic features of α₂ adrenergic agonists are primarily mediated by the α_2A adrenergic receptor (α_2A-R), as demonstrated by resistance in mice lacking a functional version of the gene encoding the α_2A-R, Adra2a.^7,8 Hypnosis stemming from α₂ adrenergic agonist hypnosis is dependent specifically upon postsynaptic α_2A-R, as restoration of Adra2a in presynaptic adrenergic neurons in Adra2a knockout mice fails to restore α₂ agonist hypnotic potency. Mice with α_2A-R rescue in adrenergic neurons (where α_2A-R is present predominantly presynaptically) but not on non-adrenergic neurons (where all α_2A-R is post-synaptic) were resistant to the hypnotic effects of α₂ agonists (indistinguishable from global Adra2a knockout mice) despite the fact that α₂ agonists are able to hyperpolarize their adrenergic neurons.⁷ Postsynaptic α_2A adrenergic receptors are widely distributed among various neuronal populations throughout the brain, and the exact α_2A-adrenergic-receptor-expressing neuronal population directly responsible for α₂ agonist hypnosis awaits identification.

The α_2A adrenergic receptor is a G protein-coupled receptor in the rhodopsin family and the molecular target of agents treating hypertension, attention-deficit disorder, and symptoms of opioid withdrawal. Current evidence, based on site mutagenesis and irreversible alkylating agonists, suggests that both the native ligand, norepinephrine, and other agonists bind to α_2A-R in the water-containing central cavity. The same experimental approaches predict that agonists functionally interact with moieties on transmembrane loops 3, 5, 6, and 7 that all face the central cavity.^9-13 While these previous investigations have provided valuable insight into potential mechanisms for agonist activity at α_2A-R, they suffer a similar limitation in that they do not represent a standard agonist interacting with the native protein. Alkylating irreversible agonists may induce structural changes in the receptor that do not occur during normal ligand-receptor interactions. Likewise, site-directed mutagenesis of a protein based upon inferences derived from the crystal structure of related receptors might produce unintentional and unrecognized changes to the tertiary structure that fundamentally alter receptor-ligand interactions.

Herein we describe the development of a novel imidazole α₂ photoaffinity ligand agonist, azi-medetomidine. Our goal in synthesizing and characterizing this novel compound was twofold: to produce an α₂ adrenergic agonist that is mechanistically similar to clinically-used drugs that could facilitate investigations into ligand-receptor binding, as well as future in vivo photolabeling to explore the neural circuits governing the behavioral actions of α₂ adrenergic agonists.

Results and Discussion

Synthesis of azi-medetomidine (1).

As a photolabel analog of medetomidine we employed 5-(1-(3-(3-(trifluoromethyl)-3H-diazirine-3-yl)phenyl)ethyl)-1H-imidazole (azi-medetomidine, 1, figure 1). Compared with medetomidine, azi-medetomidine replaces one of the methyl groups on the aromatic ring with the photoreactive trifluoromethyl diazirine group while also removing one of the methyl groups. This strategy was used to minimize possible steric interactions between the diazirine group and the removed methyl group, as well as for synthetic convenience. Preparation of azi-medetomidine was accomplished using a modification of a recent procedure for the efficient synthesis of medetomidine.¹⁴ Treatment of the previously reported¹⁵ alcohol 2 (Figure 1, bottom) with a mixture of (N-(trimethylsilyl)imidazole and titanium tetrachloride under reflux in methylene chloride produced azi-medetomidine (azi-medetomidine, 1, Figure 1) in 40% yield after workup and purification. It is noteworthy that the trifluoromethyldiazirine group survives intact under these strongly acidic conditions.

Figure 1.

(Top) Azi-medetomidine and its parent compound, the α₂ agonist medetomidine. (Bottom) Treatment of the previously described alcohol 2 with (N-(trimethylsilyl)imidazole and titanium tetrachloride produced azi-medetomidine (1) in 40% purified yield.

Physicochemical characterization and validation of azi-medetomidine(1) as photoaffinity label.

The parent compound of azi-medetomidine, medetomidine, displays stereoselective activity at central α₂ receptors, with only the dextro enantiomer producing the known physiologic effects associated with α₂ agonists.¹⁶ Using high performance liquid chromatography with an α₁-glycoprotein analytical column, and using ultraviolet–visible spectroscopy to evaluate absorbance peaks at 220 nm and 360 nm, we established purity of azi-medetomidine at greater than 98% (Supplemental Figure 2). As measure using isothermal titration calorimetry (ITC), the dissociation constants from the model protein human serum albumin (hSA) were comparable between dexmedetomidine and azi-medetomidine (Figure 2A). Azi-medetomidine displays a characteristic ultraviolet (UV) absorbance peak between 320 and 400 nm associated with its diazirine group. UV light exposure using a Rayonet RPR-3500 lamp on the compound in aqueous solution caused a progressive elimination of that peak, a result of photolysis, with a single-exponential-decay-modelled half-life of 73.55 seconds (Figure 2B).

Figure 2.

Physiochemical characterization of azi-medetomidine. (A) Isothermal Titration Calorimetry of azi-medetomidine and dexmedetomidine with hSA. 1 shows a similar, though slightly higher affinity for the model protein hSA as its parent compound, dexmedetomidine. (B) Diazirine Peak Photolysis of Azi-medetomidine with Ultraviolet Light. Increasing duration of ultraviolet (UV) light exposure results in the progressive decrease of the diazirine peak centered at ~355 nm. The black line represents absorbance prior to UV exposure, grey lines represent absorbance after 3, 30, 90, and 150 seconds of UV light, and the blue line is absorbance after 300 seconds of UV exposure. (C) Photoadduction sites of Azi-medetomidine (AziMED) on hSA, 1 labels four regions on hSA that are known drug/fatty-acid binding locations. (D) Photoadduction (as in C) in the presence of dexmedetomidine in excess causes two of those binding sites to be protected, suggesting dexmedetomidine and azi-medetomidine share those binding sites on hSA.

Model protein association of azi-medetomidine (1) and dexmedetomidine protection characterized by photoaffinity labeling.

To confirm that azi-medetomidine acts like an anesthetic photolabel, in that it is capable of photolabeling proteins with which it interacts and that competition with excess dexmedetomidine will protect specific sites from labeling, we performed UV light photoadduction of protein with or without dexmedetomidine. Dexmedetomidine is a highly protein bound drug with >90% bound to albumin in plasma, we therefore employed hSA as a suitable model to compare binding and investigate photolabeling capabilities. We identified photolabeled residues in hSA using mass spectrometry, achieving 93.8% sequence coverage (supplemental tables 1-3.) A total of 5 residues were photolabeled by azi-medetomidine within hSA, all within previously characterized drug- and fatty-acid-binding pockets, Drug Site I (T239), Drug Site 2 (I388), Drug Site 3 (H146), Fatty-Acid Site 2 (S65) and Fatty-Acid Site 5 (S579) (Figure 2C,D).¹⁷ The azi-medetomidine mass shift in y7 (1093.6 m/z) with the corresponding b10 ion and the mass shift in y2 (512.3 m/z) strongly suggested photolabeling of I 388 within Drug site 2. Photolabeling of S65 in Fatty Acid Site 2 was indicated by the total mass shift of the detected peptide and b2 (453.21 m/z). An excess of dexmedetomidine prevented photoadduction within those two sites likely indicating dexmedetomidine-specific protection of Drug Site 2 and Fatty-Acid Site 2 (Figure 2D). Of note, adduction within Fatty-Acid Site 5 only occurred in the additional presence of dexmedetomidine, which may indicate structural changes in the presence of dexmedetomidine, exposing S579. In total, azi-medetomidine photolabeled amino acids detectable by mass spectrometry and shared binding sites with its parent analogue.

Competitive binding and effects of photoaffinity labeling of azi-medetomidine (1) at α_2A -R.

We sought to assess the ability of azi-medetomidine to competitively bind at the hypnotic target of α₂ agonists, the α_2A adrenergic receptor, and to quantify effects of photoaffinity labeling on that competitive binding. For these purposes, we employed a competition binding assay using the radiolabeled α₂-agonist [³H]clonidine and recombinant human α_2A -R. Azi-medetomidine produced a dose-dependent decrease in radioactivity signal well-fit by a single-site binding model (Figure 3) suggesting that azi-medetomidine shares the binding site on the α_2A -R with clonidine. Exposure to 350 nM light, inducing photoadduction, did not affect the Kd for azi-medetomidine, but significantly reduced the amount of bound clonidine at every concentration of azi-medetomidine (Figure 3.)

Figure 3.

Radioligand competition assay. Radioactivity signal from ³H-clonidine decreased with increasing azi-medetomidine in a manner well fit by a single-site binding model (R2=0.88.) Kd with the Cheng–Prusoff correction was 2.10 nM without UV and 3.68 with UV exposure. Despite an insignificant Kd shift, significantly less ³H-clonidine is bound at all concentrations with UV exposure. Lighter shaded outer lines represent 95% CI for the curve.

Functional activation of α_2A-R by azi-medetomidine (1).

To confirm that azi-medetomidine is an agonist at α_2A-R, we used recombinant human α_2A -R in a GTPγS binding assay– an established measure of G-protein coupled receptor activation. UK14,304, a specific α_2A adrenergic agonist, and dexmedetomidine both showed activation consistent with previous reports,¹⁸ with EC₅₀ of 8.64 and 8.85 (−log[M]), respectively. Azi-medetomidine also demonstrated α_2A -R activation, though with decreased potency (EC₅₀ 6.53, Figure 4A). In the presence of the selective α₂ antagonist RX 821,002, UK14,304 saw a nearly 3 log increase in its EC₅₀, while the antagonist eliminated all α_2A -R activity associated with agonism by azi-medetomidine (Figure 4B).

Figure 4.

Ligand Activation of ADRA2A measured by GTPγS. Azi-medetomidine (1, aziMED) acts as an agonist at the human α_2A adrenergic receptor, like dexmedetomidine and the specific α_2A agonist UK 14,304. α_2A activation by azi-medetomidine is ablated in the presence of 1 μM RX 821002, a specific α_2A antagonist, while it significantly right-shifts induced activity by the α_2A specific agonist UK 14,304. Dotted lines represent 95% confidence intervals for the curve fit.

In vivo hypnotic activity of azi-medetomidine (1) in Xenopus laevis.

Dexmedetomidine produces hypnosis that has been shown to be dependent on the presence of α_2A -R, and reversible with the α₂ antagonist atipamezole.¹⁹ As an in vivo measure of α₂ adrenergic activity, we exposed X. laevis tadpoles to varying concentrations of aqueous azi-medetomidine alone or in the presence of 10 μM atipamezole and tested for loss of spontaneous movement. The hypnotic potency of azi-medetomidine was slightly less than its parent compound at an EC₅₀ of 4.65 μM (dexmedetomidine’s EC₅₀ was 0.86 μM) and, similarly to dexmedetomidine, the EC₅₀ was right-shifted by a factor of ~15 in the presence of the α₂ antagonist atipamezole (figure 5.)

Figure 5.

Hypnotic response in Xenopus tadpoles to Azi-medetomidine (AziMED) and Dexmedetomidine (Dex). Xenopus tadpoles were exposed to azi-medetomidine and dexmedetomidine alone and in the presence of 10 μM of the α₂ antagonist atipamezole (ATI). Azi-medetomidine is slightly less potent than dexmedetomidine, but equally antagonized by atipamezole.

Photoaffinity labelling of human α_2A adrenergic receptor (ADRA2A) with azi-medetomidine (1).

To investigate azi-medetomidine’s interaction with the region of the ADRA2A where imidazole agonists are thought to bind, we photoadducted azi-medetomidine using UV light on recombinant ADRA2A (with a His-SUMO tag) alone and in the presence of excess dexmedetomidine. (Note that for ease of comparison, we will refer to amino acids numbers using the native sequence, excluding the His-SUMO tag.) Similar to our approach in identifying bound moieties on hSA, we used mass spectrometry, achieving up to 85% coverage of the ADRA2A (Table 1.) Of note, some of the regions of ADRA2A that we were unable to resolve using mass spectrometry lay in the predicted agonist binding pocket. Mass spectrometry revealed an adducted moiety in the predicted extracellular region of TM2. The 12 amino acid sequence of the peptide and the amino acid modified by azi-medetomidine was identified from the spectrum (Figure 6). The azi-medetomidine mass shift present in y12 (385.14 m/z) and he lack of modification in b10 (1041.63 m/z) and b11 (1112.767 m/z) narrows the adduction site to N93 (Figure 6, location in predicted structure illustrated in Figure 7A.) However, it worth to note that there is some ambiguity in ion assignment regarding the being present on N93 or A92. In the presence of excess of dexmedetomidine during photolabeling, the azi-medetomidine modification was not identified within this region of ADRA2A suggesting specificity of the site for the pharmacologically active chiral species.

Table 1.

Coverage map for SUMO-His₁₀-ADRA2A photoadducted in the presence of 3 μM azi-medetomidine mass spectrometry analysis. Covered amino acids in bold, SUMO-His₁₀ tag in blue. 85.1% coverage.

MAHHHHHHMS	DSEVNQEAKP	EVKPEVKPET	HINLKVSDGS	SEIFFKIKKT	TPLRRLMEAF	AKRQGKEMDS
LRFLYDGIRI	QADQTPEDLD	MEDNDIIEAH	REQIGGGSHH	HHHHHHHHLV	PRGSRTMGSL	QPDAGNASWN
GTEAPGGGAR	ATPYSLQVTL	TLVCLAGLLM	LLTVFGNVLV	IIAVFTSRAL	KAPQNLFLVS	LASADILVAT
LVIPFSLANE	VMGYWYFGKA	WCEIYLALDV	LFCTSSIVHL	CAISLDRYWS	ITQAIEYNLK	RTPRRIKAII
ITVWVISAVI	SFPPLISIEK	KGGGGGPQPA	EPRCEINDQK	WYVISSCIGS	FFAPCLIMIL	VYVRIYQIAK
RRTRVPPSRR	GPDAVAAPPG	GTERRPNGLG	PERSAGPGGA	EAEPLPTQLN	GAPGEPAPAG	PRDTDALDLE
ESSSSDHAER	PPGPRRPERG	PRGKGKARAS	QVKPGDSLPR	RGPGATGIGT	PAAGPGEERV	GAAKASRWRG
RQNREKRFTF	VLAVVIGVFV	VCWFPFFFTY	TLTAVGCSVP	RTLFKFFFWF	GYCNSSLNPV	IYTIFNHDFR
RAFKKILCRG	DRKRIV

Figure 6.

Mass spectrometry of ADRA2A photoadducted in the presence 3 μM azi-medetomidine. Spectrum (A) and fragment table (B.) Detected identified a, b (red) and z, y (blue) ions are colored red and blue, respectively. Residues detected with a modification are noted, and the one modified by mAziMED (N93) additionally noted in bold and underlined.

Figure 7.

Representative positions of two azi-dexmedetomidine molecules during MD simulation with ADRA2A. Photoadducted amino acid (N93) depicted in purple. ADRA2A sections not resolved by mass spectrometry in red. (B) Distance of diazirine nitrogen from N93 during the course of simulation, which settles into a position roughly 8 Å from the labeled amino acid.

Binding site characterization using molecular dynamics simulations.

The orthosteric binding site was identified by docking dexmedetomidine, and the two enantiomers of azi-medetomidine, azi-dexmedetomidine and azi-levomedetomidine, in the ADRA2A homology model with Autodock Vina 1.1.2.²⁰ No single definitive pose was identified for any of these ligands, and the predicted poses were predominantly were adjacent to TM 1, 2 and 7. This suggests that in context of available experimental data in the literature and that fact that docking was done using a homology model, extensive refinement is necessary (identified poses depicted in Supplemental Figure 3). We conducted MD simulations to better refine the predicted site. Ligands were parameterized as in Methods, and we used FEP MD to calculate the logP for dexmedetomidine to be 2.85 and for azi-dexmedetomidine to be 3.24 (see Supplemental Figure 4 for free energy curves). For all three ligands, we placed a single ligand molecule in the orthosteric binding site of the homology model, placed in a POPC:cholesterol lipid bilayer as described in Methods. All-atom equilibrium MD simulations were run for 1 microsecond for dexmedetomidine and azi-dexmedetomidine, and 300 ns for azi-levomedetomidine. (The root-mean-square deviation of the backbone atoms over simulation time is plotted in Supplemental Figure 5.) A representative configuration is depicted in Figure 8A. To determine which residues in ADRA2A were most important in ligand binding, we identified those residues which were most commonly within 5 Å of any part of the ligand (Table 2).

Figure 8.

(A) Representative positions of azi-dexmedetomidine (thin) and dexmedetomidine (thick) during MD simulation with ADRA2A, and their proximetry to D113 (orange side chain.) (B, C) Distance of ligand imidazole nitrogen N1 to D113 side-chain oxygens OD1 and OD2, for both dexmedetomidine (B) and azi-dexmedetomidine (C), over the course of equilibrium MD simulations. Note that in both cases, the imidazole ring settles into a configuration proximal to the D113 side chain. Regions of ADRA2A not resolved by mass spectrometry are depicted in red.

Table 2.

Residues within 5.0 Å of bound ligand [Dex: dexmedetomidine, Azi-Dex: azi-dexmedetomidine] in greater than 50% of equilibrium MD trajectory frames. Ballesteros-Weinstein (B-W) numbering is shown in the first column.

B-W	Dex	Azi-Dex
3.29		L110
3.32	D113	D113
3.33	V114	V114
3.36	C117	C117
	N191	N191
5.36	K194
5.39	V197
5.40	I198	I198
5.43	C201	C201
5.44	I202	I202
5.46	T395

We found that both azi-dexmedetomidine and dexmedetomidine interacted with similar but non-identical subsets of residues in ADRA2A. The binding site involves moieties of TM3 and TM5. In particular, the imidazole ring in both dexmedetomidine and azi-dexmedetomidine remained near the side chain oxygens of D113 (Figure 8B,C) which is implicated in adrenergic signaling. In contrast, the levoenantiomer, azi-levomedetomidine, was placed in the orthosteric pocket, but had high rotational mobility and did not remain stable in the site, escaping by 60 ns simulation time.

We also considered the possibility that two azi-medetomidine molecules would bind in the receptor. We placed a second azi-dexmedetomidine in a hypothesized location and orientation, in a pocket formed by TM1, 2, and 7, with the diazirine group proximal to the photolabeled residue and guided by steric constraints. The resulting ~500 ns of production simulation showed that the first ligand remained proximal to D113 and the diazirine group of the second ligand settled within 8 Å from the beta carbon of the photolabeled residue (Figure 7B). We also conducted the same simulation but with azi-levomedetomidine as the second ligand, but found that it escaped the site within 35 ns simulation time.

Based on these results, azi-medetomidine (1), is predicted to be both a powerful tool for illuminating protein binding sites of imidazole α₂ adrenergic agonists as well as an anesthetic probe for investigating relevant neuronal circuits of α₂ adrenergic hypnosis in vivo. This potential dual utility is evident in the compound’s demonstrated α₂-dependent hypnotic activity in vivo and its shared binding properties with the common α₂ agonist dexmedetomidine.

The modifications to add the photolabile diazirine ring slightly altered the solubility, affinities, and α_2A-R activity in our novel compound compared to the parent compound medetomidine. While we had expected that addition of the diazirine at the para position on the benzene ring while maintaining the ortho methyl group might best recapitulate medetomidine’s α₂ agonism, preliminary demonstrated toxicity in Xenopus tadpoles of a synthesized compound using that approach led us to instead pursue the equally hypnotically potent azi-medetomidine that did not display any frank toxicity. While the calculated log P of 3.24 of azi-medetomidine is slightly more hydrophobic than dexmedetomidine’s at 2.85, azi-medetomidine’s interactions with the model protein hSA suggest the minor alteration in solubility has little effect on its associations. Dexmedetomidine shows only slightly higher affinity for hSA by ITC than azi-medetomidine and shares hSA binding sites with azi-medetomidine, as demonstrated by its ability to protect two of the drug/fatty-acid affinity sites on hSA from photoadduction. Drug Site 2 on hSA, labelled by azi-medetomidine and protected by excess dexmedetomidine, has similarly been labeled by two other anesthetic photolabels, halothane and azi-propofol.^21,22 Azi-medetomidine shared other binding pockets with these two anesthetic photolabels: Drug Site 1 with halothane, and Faty Acid Site 5 with azi-propofol. It is significant to note that some of the azi-medetomidine-photolabeled sites were not specifically protected by an excess of parent drug, including a site labelled only with excess parent drug in Fatty Acid Binding Site 5. We interpret this result as a likely minor increasing in binding promiscuity as a consequence of chemical modifications required for introducing the diazirine as well as the wide range dynamic ligand binding characteristic of albumins.

The decreased potency of azi-medetomidine at the adrenergic α_2A-R in comparison to its parent compound is not unexpected given substitutions on the aromatic ring of imidazoline/imidazole α₂ agonists significantly alter adrenergic receptor affinity.^23,24 As optical orientation of the chiral methyl group significantly affects potency of medetomidine derivatives, we hypothesize that the dextrorotary enantiomer of azi-medetomidine is the active enantiomer at the adrenergic α_2A-R, though we were unable to optically purify sufficient compound to test this theory. Though less than that of its parent compound, the measured potency of azi-medetomidine is nevertheless greater than that of the commonly-used α₂ adrenergic agonist, xylazine,¹⁸ and thus a reasonable substrate for investigation of α₂ adrenergic receptor binding and hypnosis.

The data suggest that azi-medetomidine not only binds within agonist pocket of ADRA2A, displacing the α₂ agonist clonidine in the radioligand competition assay, but itself acts as an agonist within that pocket, as demonstrated by receptor activation seen in the GTPγS assay. The decrease in clonidine binding at all concentrations of azi-medetomidine with UV exposure without a change in Kd reflects the less-than-100% efficiency of photolabeling– there were fewer available (unlabelled) sites, and those sites that were not labelled had unchanged affinity.

Our MD simulations suggest that dexmedetomidine interacts with transmembrane regions (TM) 3, 5, and 6 of ADRA2A, while the dextrorotary isomer of azi-medetomidine interacts with moieties on TM 3, 5, and the extracellular region between 4 and 5 (Table 2). Of note this includes D113, which is a characteristic part of the binding pocket of adrenergic agonists in other receptors. This supports and is supported by previous characterizations of TM 3 and 5 as involved in α_2A activation, determined using targeted receptor mutation and irreversible agonists. Azi-dexmedetomidine thus appears to interact with similar relevant areas of the critical TM regions as dexmedetomidine. Azi-dexmedetomidine additionally interacts with the extracellular loop between TM 4 and TM 5, though this interaction is of unclear significance. Unfortunately, using mass spectrometry we were unable to resolve key portions of TM3 – which includes D113 – and TM6 in the predicted region of azi-medetomidine interaction with recombinant α_2A-R. This can be the result of adduction increasing the hydrophobicity of a peptide, which would be expected to reduce its elution from the liquid chromatography column and its ability to ionize in the mass spectrometer, and has been reported for nucleotides.²⁵ However, because alternative mechanisms are possible, we cannot conclude with certainty that inability to detect these regions is due to photoadduction.

Our simulations additionally suggest, however, that we may not have consistently labeled a single moiety. The most stable conformation of the ligand and receptor has the diazirine facing into the central pore and most closely interacting with the highly mobile extracellular region between TM4 and TM5. We considered the possibility that the photolabeled residue represented a second bound photoaffinity ligand, and evaluated this idea by conducting a simulation with azi-dexmedetomidine plus a second bound ligand, finding that azi-dexmedetomidine but not azi-levomedetomidine stably remained in a position with the diazirine group approximately 8 Å from the photolabeled residue. Conversion of the photoaffinity group into a reactive carbine by irradiation may have provided a potential energy well that led to photoadduction. This result provides a hypothesis explaining the photolabeling result. In addition, we cannot exclude the possibility from simulations alone that there may be a stable configuration for azi-levomedetomidine, either as a first or second ligand. But given that the parent compound levomedetomidine is not clinically significant, we leave evaluation of these possibilities to future work.

Observed differences in hypnotic potency in tadpoles and human α₂ induced activity of azi-medetomidine and other imidazole and imidazoline α₂ agonists may be due to either inter-class receptor differences, or to interclass differences in CNS catecholamine architecture. Our observed EC₅₀ for dexmedetomidine of ~1.4 nM for human α_2A-R is consistent with the range of plasma concentrations producing hypnotic effects in humans. The observed behavioral EC₅₀ for dexmedetomidine in Xenopus tadpoles of 0.8 μm is consistent to previously published dexmedetomidine hypnotic EC₅₀ values in Xenopus.^26,27 This suggests that the relative potency of azi-medetomidine to dexmedetomidine is comparable for both the receptor activity assay and the behavioral assay. α₂ adrenergic hypnosis is mediated through α_2A-R, and other known off-targets of α₂ agonists, such as imidazoline receptors, do not contribute to hypnotic effects. While there is an overall identity alignment between human α_2A-R and the Xenopus homologue is 66%, the transmembrane regions forming the putative binding pocket are more highly conserved (supplemental table 8).²⁸ Similarly conserved non-mammalian α₂ receptors have shown responsiveness to α₂ agonists comparable to mammalian receptors,²⁹ which suggests that Xenopus α_2A adrenergic receptors should similarly be activated. The central adrenergic cell group many have posited as the functional effector of α₂ agonists, the locus coeruleus,³⁰ is both present and similarly interconnected with other CNS regions in amphibians and mammals.^31,32 Other mammalian brainstem adrenergic cell groups, classified as A3-A5, are not present in more primitive vertebrates, including amphibians and teleost fish.³³ Dexmedetomidine has a similarly decreased relative potency in teleost fish,³⁴ suggesting a possible role for those adrenergic cell groups present in mammals, but not in amphibians or fish, in α₂ adrenergic hypnosis. Non-mammalian organisms nevertheless remain a viable model system for investigation of α₂ adrenergic sedation and hypnosis, given the similarity in induced behavioral phenotype among mammals, amphibians, and fish, and the specificity of the sedation-hypnosis phenotype to α₂ agonism as indicated by reversibility with atipamezole.

The simulated MD interactions between dexmedetomidine and ADRA2A and between azi-dexmedetomidine and ADRA2A provide possible insight into their divergent potencies. In the simulations, dexmedetomidine interacts most highly with seven residues that are fully conserved between human and Xenopus versions of the receptor. While azi-medetomidine interacts with many of those same residues, it has additional significant interactions with VAL197 and CYS201, which are altered to isoleucine and serine in the Xenopus α_2A-R homologue. Though the former substitution is less likely to produce significant structure-function changes to the protein itself, both could alter ligand-substrate interaction. These differences in residue interactions between dexmedetomidine and azi-medetomidine may be sufficient to explain the variable difference in potency between agents with regards to hypnosis accompanying X. laevis versus human α_2A-R activation.

A second, unlikely, possibility for the observed increased hypnotic potency of azi-medetomidine in tadpoles over the GTPγS assay with ADRA2A could theoretically result from low-level photoadduction by ambient light. Infinitely increasing the ligand dissociation time as occurs with photoadduction would increase the functional concentration of the compound, making the ligand covalently adducted to its receptor appear more potent than would be expected from its free concentration. The diazirine on azi-medetomidine, however, is stable under ambient indoor lighting conditions. Moreover, both the receptor activation and assays of hypnosis were performed in identical lighting environments. It is therefore unlikely that any significant differential photoadduction occurred.

Photoaffinity binding using imidazole α₂ agonists such as azi-medetomidine may have limitations in discovering generalizable motifs of agonist binding to ADRA2A. It remains unclear whether potential sites discovered using azi-medetomidine will apply to binding by phenylethylate or oxaloazepine adrenergic agonists, with their significant structural differences. Photoaffinity ligand derivatives of those two α₂ agonist classes might serve to answer such questions. Such diverse photoaffinity ligand agonists could additionally confirm whether the known tertiary structural arrangements induced by imidazole agonists were in fact universal to agonist receptor activation of α_2A -R.

Future investigations using azi-medetomidine should make use of both its in vitro ability to identify its binding to proteins transducing its therapeutic and off-target effects and its in vivo potential to regionally photoadduct in the intact central nervous system to interrogate adrenergic circuits.⁴ The similar structure and binding profile of azi-medetomidine to its parent compound allows for the former, while the stability of the diazirine and lack of frank systemic toxicity of azi-medetomidine permits the latter. In vitro applications will help characterize imidazole and imidazoline association with known interacting proteins including adrenergic receptors, imidazole receptors, hyperpolarization-activated cyclic-nucleotide-gated channels, and monoamine oxidases. In vivo applications should assist in illuminating circuits involved in sedation, hypnosis, anxiolysis, attention, nociception, and centrally- versus peripherally-induced hypotension. The compound presented here offers substantial opportunity to investigate adrenergic agonist associations and functions and should prove to be a useful tool for revealing adrenergic signaling on a molecular and neuronal-circuit scale.

Methods

General synthetic procedures.

Reagents and solvents were all acquired from commercial sources. ¹H and ¹³C NMR spectra were obtained on a Bruker DMX 500 MHz nuclear magnetic resonance spectrometer and ¹⁹F NMR spectra were obtained on a Bruker DMX 360 MHz nuclear magnetic resonance spectrometer. Spectra for 5-(1-(3-(3-(trifluoromethyl)-3H-diazirin-3-yl)phenyl)ethyl)-1H-imidazole (Azi-medetomidine, 1) are reported in the supporting information. Accurate mass measurement analyses were conducted on either a Waters GCT Premier, time-of-flight, GCMS with electron ionization (EI), or an LCT Premier XE, time-of-flight, LCMS with electrospray ionization (ESI). Samples were taken up in a suitable solvent for analysis. The signals were mass measured against an internal lock mass reference of perfluorotributylamine (PFTBA) for EI-GCMS, and leucine enkephalin for ESI-LCMS. Waters Masslynx software calibrates the instruments, and reports measurements, by use of neutral atomic masses. The mass of the electron is not included.

Synthesis of 5-(1-(3-(3-(trifluoromethyl)-3H-diazirin-3-yl)phenyl)ethyl)-1H-imidazole (Azi-medetomidine, 1)

To a 25-mL rbf with a magnetic stir bar was added 3.63 g (25.9 mmol) N-(trimethylsilyl)imidazole and 5.5 mL methylene chloride under argon atmosphere. A solution of 2.53 mL (4.34 g, 23.2 mmol) titanium tetrachloride in 3.7 mL methylene chloride was added dropwise to the stirred solution over the course of 15 min turning the mixture red. A solution of 0.76 g (3.3 mmol) of 2 was added in one portion, and the mixture was heated to reflux overnight under an argon atmosphere. In the morning the mixture was cooled and was carefully quenched with 150 mL of water with vigorous stirring for 30 min. The mixture was extracted with three 50 mL portions of CH₂Cl₂ that were discarded. The aqueous layer was made basic (pH > 12) by dropwise addition of 150 mL of 1N sodium hydroxide producing a cloudy white precipitate. The basic aqueous layer was extracted with CH₂Cl₂ (4 × 100 mL), and the combined organic layers were washed with H₂O (3 × 100 mL) and then were evaporated. The residue was purified by silica gel column chromatography using 5% methanol:CH₂Cl₂. After evaporation of the solvent, 0.40 g (40 %) of 1 was obtained as a clear, light yellow gum. ¹H-NMR (500 MHz, CDCl₃): 10.42 (1H, bs), 7.38 (1H, bs), 7.28 (1H, t, J=7.8Hz), 7.21 (1H, bd, J=7.8Hz), 6.91 (1H, bs), 6.68 (1H, bs), 4.02 (1H, q, J = 7.1 Hz), 1.55 (3H, d, J = 7.2 Hz) ppm. ¹³C-NMR (90 MHz, CDCl₃):146.52, 141.32, 134.65, 129.21, 129.01, 128.87, 125.35, 124.61, 122.10 (quartet, J = 275 Hz), 115.86, 38.17, 28.42 (quartet, J = 40 Hz), 21.32 ppm. ¹⁹F-NMR (340 MHz, CDCl₃): −65.20 ppm. HRMS m/z calculated for C₁₃H₁₂F₃N₄ (M + H)⁺ 281.1014, found 281.1022.

Physicochemical characterization.

The purity and enantiomeric composition of azi-medetomidine (1) was determined using high performance liquid chromatography (HPLC) with α₁-glycoprotein analytical column. An isocratic gradient (11.9:88.1, acetonitrile: 30 mM phosphate buffer (pH 7.0), v:v) with a 0.8 mL/min flow at ambient temperature (21–22 °C) was applied, and azi-medetomidine was monitored for UV–vis absorbance at 220 and 360 nm. The retention times for the enantiomers were 24.4 min and 30.2 min with a purity of >98% and negligible enantiomeric excess (racemic). The UV spectrum and extinction coefficient (Σ_357nm = 315/M) of azi-medetomidine were gathered from known concentrations of ligand in metabolic solutions and monitoring of the diazirine absorption using a Varian Cary 300 Bio UV-visible spectrophotometer. The photoactivation of the diazirine was determined by the reduction of the diazirine absorption peak over the course of exposure to 350 nm light (Rayonet RPR-3500 lamp) < 3cm from the light source.

Isothermal titration calorimetry of dexmedetomidine and azi-medetomidine (1) with human serum albumin.

ITC measurements were carried out using a MicroCal VP-ITC instrument (Northampton, MA). AziMed were prepared as saturated solutions by adding excess ligand to PBS buffer, followed by vortexing and sonication and then filtration through 0.2 μm PTFE syringe filters. Concentrations were then measured by UV absorbance. 20 μM of HSA in PBS buffer (150 mM NaCl, 20 mM sodium phosphate pH 7.4) was added into the calorimeter sample cell. 0.22 mM azi-medetomidine was titrated into the sample cell. The reference cell contained ddH2O. An initial 2 μL volume over 4 s was titrated followed by 15 μL volumes over 30 s thereafter with 300 s intervals between each injection. The measurements were carried out at 30 °C and data were corrected for heats of dilution using titrations of ligand into buffer, buffer into protein, and buffer into buffer. The enthalpy data were fit to a model using a single set of independent sites using Origin 7.0 (MicroCal, Inc.). Triplicate experiments were performed the resultant Ka value (Kd=29 uM+/−9) was averaged.

Photoaffinity labeling of human serum albumin for protein microsequencing.

A final concentration 5 μM of azi-medetomidine with or without a final concentration of 0.5 mM dexmedetomidine was added to 0.2 mg/mL of hSA in phosphate buffered saline (pH 7.4). The sample was equilibrated on ice in the dark for 5 min prior to being exposed to 350 nm (RPR-3000 Rayonet lamp) in 1-mm path length quartz cuvettes at through a WG295 295nm glass filter (Newport Corporation) for 25 min.

Human serum albumin in-solution protein digestion.

25μg of photolabeled hSA was added to 4X volume of chilled acetone and stored at −20 °C overnight. The protein precipitate was pelleted by centrifugation at 16, 000 × g at 4 °C. The pellet was washed once with chilled acetone and air-dried for 5-10 min. The pellet was resuspended in 20 μL 0.2% (w/v%) ProteaseMAX™ Surfactant (Promega) in 50 mM NH4HCO3. Samples were diluted to 93.5 μL with 50 mM NH4HCO3. Following, 1 μL 0.5 M dithiothreitol was added and samples were incubated at 56 °C for 30 min. 2.7 uL of 0.55 M iodoacetamide was then added and protein samples were incubated at room temperature in the dark for 45 min. After, 1 μL of 1% (w/v%) ProteaseMax ™ Surfactant was added and sequencing grade-modified trypsin (Promega) was added to a final 1:20 protease: protein ratio (w:w). Proteins were digested overnight at 37°C. Acetic acid was then added to 0.5% (v/v%) or until pH < 2 and were incubated at ambient temperature for 10 min. To remove insoluble debris, samples were centrifuged at 16, 000 × g for 20 min at ambient temperature before desalting using C18 stage tips prepared in house. Samples were dried by speed vac and resuspended in 0.1% formic acid immediately prior to mass spectrometry analysis.

Photoaffinity labeling of human α_2A adrenergic receptor (ADRA2A) for protein microsequencing.

Azi-medetomidine to a final concentration 3 μM with or without a final concentration of 0.3 mM dexmedetomidine was added to 0.8 μg of SUMO-His10-ADRA2A (>85% purity) in 20 mM Tris-HCl, 150 mM NaCl (pH 8.0) containing 0.05% (DDM) and 10% glycerol (Cusabio, MD). His10-SUMO-ADRA2A was heterologously expressed within E. coli using previous methods. The sample was equilibrated on ice in the dark for 5 min prior to being exposed to 350 nm RPR-3000 Rayonet lamp in 1-mm path length quartz cuvettes at through a WG295 295nm glass filter (Newport Corporation) for 25 min.

In-Gel Protein Digestion.

Photolabeled samples were underwent dialysis and buffer exchange using 10kDa MWCO Amicon Ultra Centrifugal Filters (Millipore). SDS loading buffer was added to the sample containing a final concentration 100 mM DTT, samples were vortexed vigorously then incubated at room temperature for 45 min before the entire sample was separated by SDS-PAGE. Resulting gels were stained with Coomassie Blue G250 (BioRad). Gels were destained and washed with ddH2O; identified protein band between ~65- 70kDa, corresponding to His10-SUMO-ADRA2A was excised. Excised bands were destained, dehydrated and dried by speed vac before proteins were reduced by incubation at 56° C for 30 min in 5 mM DTT and 50 mM NH4HCO3. The DTT solution was removed and proteins were then alkylated by the addition of 55 mM iodoacetamide in 50 mM NH4HCO3 and incubation at room temperature for 45 min in the dark. Bands were dehydrated and dried by speed vac before resuspension in 100 μL 0.2% ProteaseMAX surfactant and 50 mM NH4HCO3 solution containing trypsin at a 1:20 protease:protein (w:w) ratio. Proteins were digested overnight at 37°C. After, samples were diluted to 200 μL with final concentration of 100 mM NH4HCO3 and 0.02% ProteaseMAX Surfactant prior to the addition of sequencing grade chymotrypsin (Promega) to a final 1:20 protease:protein (w:w) ratio. Proteins were digested overnight at 37°C. To increase hydrophobic peptide retrieval from the gel, multiple peptide extractions were performed. First the initial peptide digest solution was removed and 100 μL 30% acetylnitrile and 5% AcOH in ddH2O (v/v%) was added. Samples were sonicated for 20 min. The second peptide extraction was removed before 100 μL 70% acetylnitrile (ACN) and 5% acetic acid in ddH2O (v/v%) was added. The extractions were then pooled with the initial peptide digest. Samples were sonicated for 20 min. All peptide digests were pooled and dried by speed vac before resuspension in 0.5% acetic acid and further acidified until the pH < 2. Samples were sonicated for 10 min prior to centrifugation at 16, 000 × g for 20 min to remove insoluble debris. Samples were desalted using C18 stage tips prepared in house. Samples were dried by speed vac and resuspended in 0.1% formic acid immediately prior to mass spectrometry analysis.

Mass spectrometry.

Mass spectrometry analysis of samples was performed similar as previous reported.³⁵ Desalted peptides were analyzed on an Orbitrap EliteTM Hybrid Ion Trap-Orbitrap Mass Spectrometer (MS) coupled to an Easy-nanoLC 1000 system with a flow rate of 300 nL/min. Peptides were eluted with 100 min with linear gradients of ACN in 0.1% formic acid in water (v/v%) starting from 2% to 40% (85 min), then 40% to 85% (5 min) and finally 85% (10 min). Data dependent acquisition mode was applied with a dynamic exclusion of 45 s. In every 3 s cycle, one full MS scan was collected with a scan range of 350 to 1500 m/z, a resolution of 60K and a maximum injection time was 50 ms and automatic gain (AG) control of 500000. The MS2 scans were followed from the most intense parent ions. Ions were filtered with charge 2-5 with an isolation window of 1.5 m/z in quadruple isolation mode. Ions were fragmented using collision induced dissociation with collision energy of 35%. Ion trap detection was used with normal scan range mode and rapid ion trap scan rate. AG was set to be 10000 with a maximal injection time of 100 ms.

Mass spectrometry analysis.

Analysis was performed similar to as previously reported.³⁵ Spectral analysis was conducted using Thermo Proteome Discoverer 2.0 (Thermo Scientific) and the Mascot Daemon search engine using a customized database containing hSA protein sequence from the UniProt database (UniProtKB ID:P02768) or the SUMO-His10-ADRA2A sequence amended to an E. coli database generated using reviewed sequences from the UniProt database. All analyses included dynamic oxidation of methionine (+15.9949 m/z) as well as static alkylation of cysteine (+57.0215 m/z; iodoacetamide alkylation). Photolabeled samples were run with the additional dynamic azi-medetomidine (+252.2350 m/z) modification. A mass variation tolerance of 10 ppm for MS and 0.5 Da for MS/MS was used. The in-solution trypsin digests were searched with trypsin specificity and a maximum of 3 missed cleavages allowed. The in-gel sequential trypsin/chymotrypsin digests were searched without enzyme specification and both searched were set with a false discovery rate of 0.01%. Samples were conducted in triplicate and samples containing no photoaffinity ligand were treated similarly to control for false positive detection of photoaffinity ligand modifications.

Radioligand Competitive Binding Assay.

Competitive binding was performed with a fixed amount of [3H]-clonidine (15nM) in the presence of different concentrations of Azi-Med (0.1 nM-500 nM). α_2A adrenergic receptor membrane preparation (5ug/well, Millipore) were mixed with 15 nm [3H]-clonidine (American Radiolabeled Chemicals, MO) and Azi-Med (0.1-500 nM) in binding buffer (50 mM Tris 7.4, 5 mM MgCl2, 0.1% BSA) in a non-binding 96-well plate, and incubated for 2 hrs at RT. Prior to the reaction, an 96-well plate was coated with 0.33% polyethyleneimine for 30 min, then washed with 50mM Tris 7.4, 0.1% BSA. The binding reaction was done with or without exposure to 350-nm light for 30min (Rayonet RPR-3000 Lamp, Southern New England Ultraviolet Co.; Branford, CT, USA). Then the reaction was stopped by vacuum filtration, and washed 3 times (1 mL per well per wash) with ice cold wash buffer (50mM Tris 7.4, 0.1% BSA). The filters were soaked in 10ml scintillation cocktail and counted for 3min after the delay of 6 hrs.

Hypnotic activity Xenopus laevis tadpole bioassay.

Behavioral activity was initially determined in albino X. laevis tadpoles (stages 45–47 [Nasco, WI]) as described previously.^36,37 Briefly tadpoles were incubated in Petri dishes (10 tadpoles per dish; 21-22 °C) with varying concentration of dexmedetomidine or azi-medetomidine (1) with or without 10 μM atipamezole dissolved in artificial pond water (E3 medium: ddH₂O containing 5 mm NaCl, 0.17 mM KCl, 10 mM HEPES, 0.33 mM MgSO₄•7H₂O, 0.33 mM CaCl₂•6H₂O), containing < 0.01% (v/v%) Dimethyl sulfoxide vehicle, for 30 min. Hypnosis was defined as the percentage of tadpoles that did not demonstrate spontaneous movement over the course of a 30 s period. Tadpoles were transferred to fresh pond water and observed for signs of toxicity. Each concentration exposure was performed in triplicate and expressed as mean +/− standard error of the mean. All animal care and experimental procedures involving X. laevis tadpoles were carried out according to a protocol approved by the IACUC of the University of Pennsylvania.

[³⁵S]GTPγS Binding Assay.

In a 96 well filter plate (Millipore Sigma, Burlington MA), each well was washed with 100 μL of TMN buffer (50 mM Tris pH 7.6, 10 mM MgCl₂, 10 μM GDP). 5 μg of human α2A adrenergic receptor membrane preparation (Milipore Sigma, Burlington MA) in TMND buffer (50 mM Tris pH 7.6, 10 mM MgCl₂, 10 μM GDP, 1 mM dithiothreitol) was added in each well along with additional TMND buffer, non-radiolabeled GTPγS to a final concentration of 2.5 nM, or RX821002 to a final concentration of 1 μM in TMND buffer. After 5 minutes [³⁵S]GTPγS in TMND to a final well concentration of 0.25 nM is added, followed by agonists (1, dexmedetomidine, or UK-14,304) in TMND. The plate was covered and incubated with mild agitation for one hour at room temperature. The samples were filtered through the plate using a vacuum pump, and the wells washed and filtered four times with 200 μL TMN buffer. Filters were dried, removed, placed in scintillation vials with 20 mL scintillation fluid and total counts obtained. Counts were normalized to background and nonspecific binding (based on blockade with non-radiolabeled GTPγS, as per Neil and colleagues.)³⁸

Simulation Methods

Homology modeling.

Since no crystal structure of ADRA2A is available, we prepared a homology model using newly available active-state beta-adrenergic receptor structures (PDB: 6MXT, 3SN6, 3P0G, 6H7J)^39-42MODELLER 9.21 software was used.⁴³Corresponding residues among these structures were identified by structural alignment, weighting each structure equally, and the ADRA2A sequence was then aligned to this. This allowed for the generation of a consensus homology model. The disulfide bond between TM3 and the extracellular loop was explicitly included due to the proximity of the cysteines. We took the best-scoring of 100 generated models.

Derivation of molecular mechanics ligand parameters for dexmedetomidine and azi-dexmedetomidine.

The molecular mechanics parameters for dexmedetomidine and both isomers of azi-medetomidine were derived from the CGenFF force field parameter set for small molecules by combining the parameters for the relevant functional groups. Missing dihedral parameters were fit to reproduce the energy surface from quantum mechanics (QM) torsional energy scans, including a 2D torsion scan to capture the interaction of phenyl and imidazole rings resulting from their relative rotation, with which we fit a CMAP correction map.⁴⁴ Since parameters for the diazirine group were not available in the literature, we chose Lennard-Jones and electrostatics parameters to reproduce the QM interaction energy surface of this group with a water molecule in multiple positions in an arc around the group, a zero point energy adjustment was applied. All QM calculations were conducted using Gaussian09 software (Gaussian Inc., Wallingford, CT) at the B3LYP/6-31G** level of theory and basis set.

Molecular dynamics simulations.

We conducted molecular dynamics (MD) simulations of each ligand embedded in the GPCR orthosteric binding pocket of our homology model of ADRA2A. Each single-ligand system was oriented such that the imidazole ring of the ligand was proximal to D113, and a second ligand, when present, was oriented such that the diazirine group was oriented proximal to the photolabeled residue. We evaluated azi-dexmedetomidine, which is the dextrorotary enantiomer of azi-medetomidine, in the orthosteric site, and both this and the levorotary enantiomer in a secondary site adjacent to the photolabeled residue. The receptor was embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine:cholesterol 70:30 lipid bilayer and the system was solvated with TIP3P water with 0.15 M NaCl using the CHARMM-GUI service.^45,46. The CHARMM36 force field⁴⁷ and NAMD 2.12 simulation software⁴⁸ were used. Periodic boundary conditions with particle mesh Ewald summation of long-range electrostatics was used. The system was equilibrated according to the CHARMM-GUI protocol and production equilibrium MD simulation was run for 10 nanoseconds using the isothermic-isobaric ensemble with Langevin thermostat and barostat at 303.15 K. Further equilibrium MD simulations were continued from that point.

Octanol-water partition coefficient (log P).

This was calculated using free energy perturbation (FEP) MD. The solvation free energy of ligand to protein was calculated as the sum of the energy of two stages: 1) decoupling of the ligand from the protein into vacuum and 2) recoupling into water. 30 FEP windows were used with 2 ns production simulation for each window. Window sizes were progressively decreased as lambda approached 1 in order to improve convergence of the calculated energy.

We predicted the log P by calculating, using FEP, the energies of decoupling a single ligand molecule from a box of 1-octanol or water. Because free energy is a state function, the first quantity minus the second is equivalent to the energy of dragging the ligand from the octanol to water. Including a correction of these energies to 1 M ligand concentration (i.e. one ligand molecule per 1660 ų), the gas constant, temperature, and a correction for the base of the logarithm yields the free energy of this transition and therefore the equilibrium constant:

$$\log P=\frac{\Delta G_{\mathrm{oct},1M}-\Delta G_{\mathrm{wat},1M}}{2.303RT}$$

where ΔG_oct,1M and ΔG_wat,1M are the free energies of decoupling a ligand molecule from 1-octanol and water respectively at a ligand concentration of 1 M, R is the gas constant, and T is temperature, and 2.303 is a constant arising from logarithm base conversion.

Abbreviations Used:

α_2A-R

alpha 2A adrenergic receptor

Å

Angstrom

AG

automatic gain

B-W

Ballesteros-Weinstein

CMAP

2D dihedral energy grid correction map

FEP

free energy perturbation

GDP

guanosine diphosphate

GTPγS

guanosine 5'-O-[gamma-thio]triphosphate

HPLC

high pressure liquid chromatography

hSA

human serum albumin

ITC

isothermal titration calorimetry

logP

octanol-water partition coefficient

MD

molecular dynamics

MHz

megahertz

MS

mass spectrometer

NMR

nuclear magnetic resonance

QM

quantum mechanics

TM

transmembrane region

UV

ultraviolet