1521-0111/90/6/703–714$25.00 http://dx.doi.org/10.1124/mol.116.105007MOLECULAR PHARMACOLOGY Mol Pharmacol 90:703–714, December 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

Extracellular Loop 2 of the Adenosine A1 Receptor Has a KeyRole in Orthosteric Ligand Affinity and Agonist Efficacy s

Anh T. N. Nguyen, Jo-Anne Baltos, Trayder Thomas, Toan D. Nguyen, Laura López Muñoz,Karen J. Gregory, Paul J. White, Patrick M. Sexton, Arthur Christopoulos,and Lauren T. MayMonash Institute of Pharmaceutical Sciences (A.T.N.N., J.-A.B., T.T., L.L.M, K.J.G, P.J.W, P.M.S, A.C., L.T.M), Monashe-Research Centre (T.D.N), and Department of Pharmacology (A.T.N.N, J.-A.B., K.J.G., P.M.S., A.C., L.T.M), Monash University,Parkville, Victoria, Australia

Received May 2, 2016; accepted September 27, 2016

ABSTRACTThe adenosine A1 G protein–coupled receptor (A1AR) is animportant therapeutic target implicated in a wide range ofcardiovascular and neuronal disorders. Although it is wellestablished that the A1AR orthosteric site is located within thereceptor’s transmembrane (TM) bundle, prior studies haveimplicated extracellular loop 2 (ECL2) as having a significantrole in contributing to orthosteric ligand affinity and signaling forvarious G protein–coupled receptors (GPCRs). We thus per-formed extensive alanine scanning mutagenesis of A1AR-ECL2to explore the role of this domain on A1AR orthosteric ligandpharmacology. Using quantitative analytical approaches andmolecular modeling, we identified ECL2 residues that interact

either directly or indirectly with orthosteric agonists and antag-onists. Discrete mutations proximal to a conserved ECL2-TM3disulfide bond selectively affected orthosteric ligand affinity,whereas a cluster of five residues near the TM4-ECL2 junctureinfluenced orthosteric agonist efficacy. A combination of liganddocking,molecular dynamics simulations, andmutagenesis resultssuggested that the orthosteric agonist 59-N-ethylcarboxamidoa-denosine binds transiently to an extracellular vestibule formed byECL2 and the top of TM5 and TM7, prior to entry into the canonicalTM bundle orthosteric site. Collectively, this study highlights a keyrole for ECL2 in A1AR orthosteric ligand binding and receptoractivation.

IntroductionThe adenosine receptor (AR) family of G protein–coupled

receptors (GPCRs) consists of four subtypes: A1AR, A2AAR,A2BAR, and A3AR. A1AR is widely distributed within the body,with high levels found in the central nervous system andperipheral organs (Palmer and Stiles, 1995; Fredholm et al.,2001; Yaar et al., 2005). A1AR preferentially couples to Gi/o

proteins and A1AR stimulation can reduce cardiac and renalischemia reperfusion injury, atrial fibrillation, and neuro-pathic pain (Fredholm et al., 2001). Furthermore, A1ARantagonists may be beneficial as potassium-saving diuretics

(Jacobson and Gao, 2006; Müller and Jacobson, 2011) andcognition enhancers (Hess, 2001). Consequently, A1AR is anattractive therapeutic target, and a detailed understanding ofligand binding and function is imperative for this clinicallyrelevant target.Structures of the human A2AAR cocrystallized with various

agonists and antagonists have recently been solved, providingdeep insight into the orthosteric ligand binding region(Jaakola et al., 2008; Doré et al., 2011; Lebon et al., 2011,2015; Xu et al., 2011; Congreve et al., 2012; Hino et al., 2012;Liu et al., 2012). These crystal structures suggest that bothhelical transmembrane (TM) domains and extracellular loops(ECLs) contribute to the geography of the orthosteric pocket.In particular, an integral role for the helical region of ECL2was highlighted, with Phe168 and Glu169 in this loop forminga p-stacking interaction and hydrogen bonding, respectively,with both antagonist and agonists.Previous homology modeling and mutagenesis studies

suggested that the A1AR orthosteric binding pocket is locatedwithin the TM bundle, with key roles attributed to residuesL883.33, T913.36, Q923.37, and T2777.42 (Townsend-Nicholson

This research was supported by the National Health and Medical ResearchCouncil of Australia (NHMRC) [Program Grant APP1055134 and ProjectGrants APP1084487 and APP1084246]. A.T.N.N. is a recipient of anAustralian Endeavour Scholarship and Fellowship. L.T.M. is a recipient ofan Australian Research Council Discovery Early Career Researcher Award.A.C. is an NHMRC Senior Principal Research Fellow, P.M.S. is an NHMRCPrincipal Research Fellow, and K.J.G. is an NHMRC Overseas BiomedicalPostdoctoral Training Fellow.

dx.doi.org/10.1124/mol.116.105007.s This article has supplemental material available at molpharm.aspetjournals.

org.

ABBREVIATIONS: AR, adenosine receptor; CHO, Chinese hamster ovary; DMEM, Dulbecco’s modified Eagle’s medium; DPCPX, 8-cyclopentyl-1,3-dipropylxanthine; ECL, extracellular loop; GPCR, G protein–coupled receptor; HA, human influenza hemagglutinin; MD, molecular dynamics;NECA, 59-N-ethylcarbozamidoadenosine; SLV 320, trans-4-[(2-phenyl-7H-pyrrolo[2,3-d]pyrimidin-4-yl)amino]cyclohexanol; [3H]DPCPX,8-cyclopentyl-1,3-dipropylxanthine [dipropyl-2,3-3H(N)]; TM, transmembrane; UK432097, 6-[2,2-di(phenyl)ethylamino]-9-[(2R,3R,4S,5S)-5-(ethylcarbamoyl)-3,4-dihydroxyoxolan-2-yl]-N-[2-[(1-pyridin-2-ylpiperidin-4-yl)carbamoylamino]ethyl]purine-2-carboxamide; WT, wildtype; ZM241385, 4-(2-[7-amino-2-(2-furyl)-[1,2,4]triazolo-[2,3-a][1,3,5]triazin-5-ylamino]ethyl)-phenol.

703

http://molpharm.aspetjournals.org/content/suppl/2016/09/28/mol.116.105007.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

and Schofield, 1994; Rivkees et al., 1999). However, ECL2, aregion with high sequence variability across the AR subtypes,has also been implicated in A1AR orthosteric agonist andantagonist pharmacology (Olah et al., 1994). Similarly, anemerging picture of the contribution of ECL2 in orthostericligand recognition, and even the ability of G protein–coupledreceptors (GPCRs) to transition between multiple functionalstates, has beenmore broadly suggested for other AR subtypesand family A GPCRs, including the A2AAR, A3AR, cannabi-noid receptor 1,M2 andM3muscarinic acetylcholine receptors,rhodopsin, b-adrenoceptors, dopamine receptors, chemokinereceptor 4, and V1a vasopressin receptor (Olah et al., 1994;Howl and Wheatley, 1996; Kim et al., 1996; Gao et al., 2002;Shi and Javitch, 2004; Avlani et al., 2007; Scarselli et al., 2007;Gregory et al., 2010, 2012; Sabbadin et al., 2015). In partic-ular, ECL2 residues in the area adjacent to the highlyconserved disulfide-bonding cysteine appear to play a key rolein agonist binding and function at family A GPCRs (Wheatleyet al., 2012). At the human A2AAR, glutamate residues inECL2 have also been suggested to be involved in orthostericligand recognition (Kim et al., 1996). Furthermore, a recentseminal study using unbiased long-timescale molecular dy-namics (MD) simulations highlighted a role for ECL2 in theentry and egress of b-adrenoceptor orthosteric ligands fromthe binding pocket. Interestingly, in that study, the trajectoryof ligands into the orthosteric pocket involved a metastablestate in an extracellular vestibule, delineated by ECL2 andECL3, prior to entry into the binding pocket within the helicalbundle (Dror et al., 2011). It is possible that this paradigmextends to other receptors and that the residence time a ligandspends in thismetastable state prior to engaging the canonicalorthosteric site may have implications for receptor activationdynamics and potential allosteric targeting.Therefore, this study aimed to establish the role of A1AR-

ECL2 in orthosteric ligand binding and function using acombination of mutagenesis, quantitative analytical pharma-cology, andmolecular modeling. Complete alanine scanning ofECL2 residues was performed and the effects on orthostericligand affinity and efficacy were quantified using radioligandbinding and cAMP accumulation assays. Interpretation ofmutagenesis data was facilitated by ligand docking into A1ARhomology models based on agonist- and antagonist-boundA2AAR crystal structures followed by MD simulations. Liganddocking poses were evaluated by identification and muta-tional validation of novel ligand–receptor interactions withinA1AR. Collectively, our results demonstrate a key role forECL2 in orthosteric ligand binding and A1AR activation. Wepropose that ECL2, together with the top of TM5 and TM7,forms an extracellular vestibule that is recognized by 59-N-ethylcarboxamidoadenosine (NECA) and facilitates the tran-sition of this agonist into the orthosteric pocket within the TMbundle. This vestibule may also represent a target for recogni-tion of A1AR by allosteric ligands.

Materials and MethodsMaterials

Chinese hamster ovary (CHO) Flp-In cells, Dulbecco’s modifiedEagle’s medium (DMEM), Lipofectamine 2000, and hygromycin B(Hygrogold) were purchased from Invitrogen (Carlsbad, CA). Fetalbovine serumwaspurchased fromThermoTrace (Melbourne,Australia).The QuikChange II site-directed mutagenesis kit was purchased from

Agilent (La Jolla, CA). AlphaScreen reagents, OptiPhase Supermixscintillation cocktail, and 8-cyclopentyl-1,3-dipropylxanthine, [dipropyl-2,3-3H(N) ([3H]DPCPX; specific activity, 120 Ci/mmol) were obtainedfrom PerkinElmer Life Sciences (Waltham, MA). Primers were pur-chased from GeneWorks (Hindmarsh, Australia). All other reagentswere purchased fromSigma-Aldrich (St. Louis,MO) or TocrisBioscience(Bristol, UK) and were of analytical grade.

Receptor Mutagenesis

Mutations were introduced into the coding sequence of the humanA1AR containing an amino N-terminal triple human influenzahemagglutinin (HA) epitope tag (3xHA-A1AR) in pcDNA3.11 vectorfrom the Missouri S&T cDNA Resource Center (Rolla, MO). Thesequence was amplified by PCR using the following primers: 59-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCGCCACCACTTA-AGCTTGGTACCACCTG-39 (N-terminal forward primer includingattB, Kozak sequence, and N-terminal of pcDNA3.1 site) and 59-GGGGACCACTTTGTACAAGAAAGCTGGGTCCTAGTCATCAGGC-CTCTCTTCTGG-39 (C-terminal reverse primer including thecomplementary attB sequence and complementary C-terminal ofA1AR). The 3xHA-A1AR coding sequence was subsequently clonedinto the Gateway entry vector pDONR201 using BP Clonase enzymemix (Invitrogen). Alanine amino acid substitutions were performedusing oligonucleotides from GeneWorks (Supplemental Table 1) andQuikChange II site-directed mutagenesis kits (Stratagene, La Jolla,CA) according to the manufacturer’s instructions. Sequences of re-ceptor clones were confirmed by automated sequencing as describedpreviously (May et al., 2007). Receptor clones were then subclonedinto the Gateway destination vector pEF5/FRT/V5-DEST using LRClonase enzyme mix (Invitrogen).

Transfections and Cell Culture

DNA constructs in the pEF5/FRT/V5-DEST destination vector werestably transfected into the Flp-In-CHO cell line as described pre-viously (May et al., 2007). Briefly, Flp-In-CHO cells were transfectedin serum and antibiotic-free DMEM with 0.7 mg pEF5/FRT/V5-DESTvector containing the wild-type (WT) or mutant 3xHA-A1AR con-structs and 7 mg pOG44 vector using Lipofectamine 2000 (30 ml per25-cm2 flask). Flp-In-CHO cells stably expressing human A1AR(untagged) were generated previously (Valant et al., 2014). Cellswere selected and maintained in DMEM containing 10% fetal bovineserum and 600 mg/ml hygromycin B in a humidified environment at37°C in 5% CO2.

Radioligand Binding Assay

Cells were seeded at 40,000 cells per well into a transparent 96-wellplate and incubated in a humidified environment at 37°C in 5% CO2.After 8 hours, cells were washed with serum-free DMEM andmaintained in serum-free DMEM for approximately 18 hours at37°C in 5% CO2 before assaying. [3H]DPCPX whole-cell saturationbinding assays on Flp-In-CHO cells stably expressingA1AR constructswere performed at 4°C for 3 hours in a final volume of 100 ml HEPESbuffer (145 mM NaCl, 10 mM D-glucose, 5 mM KCl, 1 mM MgSO4,10 mM HEPES, 1.3 mM CaCl2, and 15 mM NaHCO3, pH 7.4) inthe absence or presence of increasing concentrations of [3H]DPCPX(0.1–10 nM). [3H]DPCPX whole-cell competition binding assayswere performed under the same conditions with approximately1 nM [3H]DPCPX and increasing concentrations of NECA (0.3 nMto 30 mM) or 8-cyclopentyl-1,3-dipropylxanthine (DPCPX; 0.1 nM to10mM).Nonspecific bindingwas defined in the presence of 1mMSLV320(trans-4-[(2-phenyl-7H-pyrrolo[2,3-d]pyrimidin-4-yl)amino]cyclohexanol),a selective A1AR antagonist (Kalk et al., 2007). Assays were terminatedby washing twice with 100 ml cold phosphate-buffered saline per well,followed by the addition of 100 ml OptiPhase Supermix scintillationcocktail, and bound radioactivity wasmeasured using aMicroBeta2 platecounter (PerkinElmer Life Sciences).

704 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

cAMP Accumulation Assay

Inhibition of forskolin-stimulated cAMP accumulation assays wereperformed as described previously (Baltos et al., 2016). Briefly, cellswere seeded at 20,000 cells per well into a transparent 96-well plateand incubated in a humidified incubator at 37°C in 5% CO2 overnight.Media were then replaced with stimulation buffer (140 mM NaCl,5.4 mM KCl, 0.8 mM MgSO4, 0.2 mM Na2HPO4, 0.44 mM KH2PO4,1.3 mM CaCl2, 5.6 mM D-glucose, 5 mM HEPES, 0.1% bovine serumalbumin, and 10 mM rolipram, pH 7.45) and incubated at 37°C for30 minutes. Concentration-response assays were performed by sub-sequent incubation of NECA (10 pM to 10 mM) and 3 mM forskolin foran additional 30 minutes at 37°C. Detection of cAMP was performedusing AlphaScreen cAMP assay kits and fluorescence was measuredwith an EnVision plate reader (both from PerkinElmer Life Sciences)using standard AlphaScreen settings. Agonist concentration-responsecurves were normalized to the response mediated by 3 mM forskolin(0%) or buffer (100%) alone.

Data Analysis

All data were analyzed using GraphPad Prism software (version6.03; GraphPad Software Inc., La Jolla, CA). Relative receptorexpression (Bmax) and [3H]DPCPX equilibrium dissociation constants(KA) were determined from saturation binding assays using thefollowing equation (eq. 1):

Y5Bmax½A�½A�1KD

1NS½A� (1)

whereY is radioligand binding,Bmax is the total receptor density, [A] isthe radioligand concentration, KD is the equilibrium dissociationconstant of the radioligand, andNS is nonspecific radioligand binding.

[3H]DPCPX competition binding curves were fitted to a one-siteinhibitory mass action equation (eq. 2):

Y5Bottom1Top2Bottom

1110ðX2LogIC50Þ (2)

where Y is the specific binding and IC50 is the concentration of ligandthat displaces 50% of the radioligand. Top and bottom represent themaximal and minimal binding in the presence of the competitiveligand, respectively. Equilibrium dissociation constant (KI) valueswere subsequently derived from inhibition binding experiments usingthe Cheng and Prusoff (1973) equation (eq. 3):

KI 5IC50

11 ½A�KD

(3)

NECA concentration-response curves were fitted to the following formof the operational model of agonism (Black and Leff, 1983) (eq. 4):

Y5Basal1Em 2Basal

11 ½ðKA 1 ½A�Þ=ðt � ½A�Þ�n (4)

where Em is the maximal system response, A is the agonist concen-tration, t is an index of coupling efficiency (efficacy) of the agonistand is defined as RT/KE [where RT is the total concentration ofreceptors (i.e., Bmax) and KE is the concentration of agonist-receptor complex that yields half the Em], and n is the slope ofthe transducer function that links occupancy to response. Thelatter parameter was constrained to be shared across all data sets,because an extra sum of squares (F test) determined that curvefitting was not improved by allowing different transducer slopes fordifferent receptor mutations. KA values were constrained to theirrespective KI values estimated from [3H]DPCPX competitionbinding studies. To account for effects of the expression level ofeach of the different mutant receptors on the observed efficacy ofeach agonist, the Bmax values determined from saturation bindingwere used to normalize the t values derived from the operational

model analysis (Gregory et al., 2010); these values are reported as“corrected t values” (tc).

Statistical analysis was performed using a t test, or a one-wayanalysis of variance with a Dunnett post test as indicated was used todetermine statistical differences (P , 0.05) where appropriate.

Molecular Modeling

Homology Modeling of A1AR. The human A1AR and A2AARshare approximately 40% sequence identity; therefore, A2AR crystalstructures are acceptable templates for A1AR homologymodels to gaina better understanding of ligand interactions at A1AR. The sequenceof the human A1AR was retrieved from the Swiss-Prot database.ClustalX software (Thompson et al., 1997; The Conway Institute ofBiomolecular and Biomedical Research, University College Dublin,Dublin, Ireland) was used to align the A1AR sequence with thecrystal structures of the antagonist- and agonist-bound humanA2AAR [Protein Data Bank identifiers 3EML (Jaakola et al., 2008)and 3QAK (Xu et al., 2011) for the inactive and partially activestates, respectively]. Three-dimensional structures of A1AR wereconstructed with MODELLER software (version 9.12, Andrej Sali,University of California San Franscisco, CA; Eswar et al., 2007). Theconserved disulfide bond between residue C803.25 at the top of TM3and C169ECL2 and the ECL3 intraloop disulfide bond betweenC260ECL3 and C263ECL3 present in the template structure werebuilt and maintained as a constraint for geometric optimization. Thebest structure was selected based on the MODELLER DiscreteOptimized Protein Energy assessment score and visual inspection.Structures obtained were optimized using the Duan force field (Duanet al., 2003).

Docking of AR Ligands. Docking of DPCPX and NECA wasperformed using ICM software (version 3.8.0;Molsoft LLC, SanDiego,CA). Potential binding sites were predicted using the ICM PocketFinder algorithm (An et al., 2005; Abagyan and Kufareva, 2009). Thedocking search boundaries were defined by a box of 30 Å� 30 Å� 30 Å(the region in which 0.5-Å grid energymapswere generated). This sizeis large enough to encompass the potential orthosteric binding pocketand the extracellular region of the receptor. Docking thoroughness,representing the length of the docking simulation, used defaultsettings. Ligand binding modes were ranked according to ICM score,which is inversely proportional to the number of favorable intermo-lecular interactions between the docked ligand and the receptor.The docking of each ligand was repeated five times and the confor-mation with the lowest ICM score was selected for subsequent MDsimulations.

MD Simulations. MD simulations of the final complex werecarried out with the NAMD2.10 software package (Theoretical andComputational Biophysics Group, Beckman Institute for AdvancedScience and Technology, University of Illinois at Urbana-Champaign,IL; Phillips et al., 2005) using the three-site rigid water TIP3P model,CHARMM27 (MacKerell et al., 1998, 2004), and CGenFF (version3.0.1; Vanommeslaeghe et al., 2010; Yu et al., 2012) force fields asdescribed previously (Shonberg et al., 2013; http://www.ks.uiuc.edu/Training/Tutorials/science/membrane/mem-tutorial.pdf). The particlemesh Ewald (Essmann et al., 1995) method was used to evaluateelectrostatic interactions. Each system contains an A1AR receptor, theligand, a lipid bilayer containing approximately 220 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine molecules generated using themembrane plugin of the VMD software (version 1.9.2, Theoreticaland Computational Biophysics Group, Beckman Institute for Ad-vanced Science and Technology, University of Illinois at Urbana-Champaign, IL) (Humphrey et al., 1996), and approximately 15,800water molecules. Sodium and chloride ions were added to neutralizethe system, with extra NaCl added to reach a final concentration of150 mM (approximately 35 sodium ions, approximately 48 chlorideions). Two energy minimization and equilibration periods wereperformed prior to the MD production run. The first period involvedenergy minimization and a 15-ns equilibration of lipid tails. The

Role of ECL2 on A1 Receptor Pharmacology 705

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

second period involved energy minimization and a 5-ns equilibrationof the whole system, with 10 kcal/mol per Å22 harmonic-positionrestraints applied to all heavy atoms of the protein and ligands.Subsequently, the production run of the final unconstrained systemwas performed for 40 ns using a 2-fs integration time step in theconstant temperature, constant pressure ensemble (310 K, 1 bar).Coordinates were written to the output trajectory file every 10 ps.VMD software (version 1.9.2) was used for the visualization andanalysis of the residue-ligand contacts through the course of eachsimulation using in-house scripts. The percentage of total MD timethat residues were involved in hydrogen bonding and/or found within3.5 Å of DPCPX or NECA heavy atoms was quantified for each MDsimulation. All figures were generated using PyMOL software(Schrödinger, Cambridge, MA).

Results3xHA-A1AR Has Comparable Pharmacology to the HumanA1AR

In this study, the role of ECL2 residues on agonist bindingand function was investigated using human A1AR constructswith an N-terminal triple HA epitope tag (3xHA-A1AR).Previous studies on other GPCRs have shown that theaddition of an N-terminal HA tag generally does not interfere

with receptor pharmacology (Sromek and Harden, 1998;Leach et al., 2011; Schiedel et al., 2011). To ensure that thepharmacology of 3xHA-A1AR was equivalent to the untaggedhuman A1AR, we compared agonist and antagonist affinityestimated between the two constructs. In homologous compe-tition binding studies, the N-terminal 3xHA tag significantlyreduced receptor expression (Bmax values of 5343 6 234.6 forA1AR and 2022 6 155 for 3xHA-A1AR; n 5 3–6; P , 0.05,t test); however, DPCPX affinity was unaffected (pKI values of8.846 0.08 for A1AR and 8.966 0.11 for 3xHA-A1AR; n5 3–6;P . 0.05, t test). In heterologous competition binding studies,NECA affinity was also unaffected (pKI values of 6.48 6 0.01for A1AR and 6.596 0.04 for 3xHA-A1AR; n5 3–44; P. 0.05,t test). Therefore, despite having decreased cell surface expres-sion, the N-terminal 3xHA epitope–tagged human A1AR has

Fig. 1. (A) Diagram highlighting alanine mutations within the humanA1AR. Single alanine substitutions are highlighted in red, whereas doubleand triple mutations are blue. (B) Receptor expression (Bmax) of WT andmutant 3xHA-A1ARs stably expressed in Flp-In-CHO cells. Bmax valuesare expressed as a percentage ofWT and represent themean6 S.E.M. of atleast three experiments performed in duplicate. *P, 0.05 (compared withWT; one-way analysis of variance, Dunnett post hoc test). ND, notdetermined because no specific radioligand binding was detected.

Fig. 2. Radioligand binding and cAMP accumulation data for Flp-In-CHO cells stably expressing WT or selected mutant 3xHA-A1ARs. (A)[3H]DPCPX whole-cell saturation binding. (B) [3H]DPCPX whole-cellcompetition binding. (C) cAMP accumulation. Data are means 6 S.E.M.from at least three experiments conducted in duplicate. Where errorbars are not shown, they lie within the dimensions of the symbol. CPM,counts per minute; FSK, forskolin.

706 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

equivalent affinity with the orthosteric ligands to the untaggedhuman A1AR.

Cell Surface Expression of Human 3xHA-A1AR ECL2Alanine Mutants

The generalized numbering scheme proposed by Ballesterosand Weinstein (1995) was used for TM residues. Alaninescanning mutations were performed within ECL2 of A1AR,with the exception of C169ECL2. Residues C803.25 andC169ECL2 are predicted to form a conserved disulphide bond,and alanine substitution of C169ECL2 was shown previously toabolish [3H]DPCPX-specific binding (Scholl and Wells, 2000;Martinelli and Tuccinardi, 2008). Single alanine mutationswere performed, with the exception of S150ECL2A 1V151ECL2A (SV150ECL2AA) and M162ECL2A 1 G163ECL2A 1E164ECL2A (MGE162ECL2AAA), for which the double andtriple mutations were similarly expressed (Fig. 1A).Cell surface expression of each mutant 3xHA-A1AR stably

expressed in Flp-In-CHO cells was determined from whole-cell [3H]DPCPX saturation binding (Figs. 1B and 2A). No[3H]DPCPX-specific binding could be detected at theF171ECL2Amutation; however, subsequent detection of the 3xHA tag withELISA showed that cell surface expression of the F171ECL2A

receptorwas not significantly different from that of theWTA1AR(P . 0.05, unpaired t test; Supplemental Fig. 1). Orthostericligand affinity is predicted to be significantly influenced by theF171ECL2A mutation, because the corresponding residue atA2AAR (F168ECL2) was previously shown to form ap-stackinginteraction with the antagonist ZM241385 [4-(2-[7-amino-2-(2-furyl)-[1,2,4]triazolo- [2,3-a][1,3,5]triazin-5-ylamino]ethyl)-phenol] and agonistNECA inA2AARcrystal structures (Jaakolaet al., 2008; Liu et al., 2012). As such, the lack of [3H]DPCPX-specific binding observed for F171ECL2A is likely attributable toa significant decrease in [3H]DPCPX affinity. Of the remaining25 mutant 3xHA-A1ARs investigated, 16 mutant 3xHA-A1ARFlp-In-CHO cell lines had significantly reduced cell surfacereceptor expression compared with theWT 3xHA-A1AR Flp-In-CHO cell line (Fig. 1B; Supplemental Table 2).

ECL2 Alanine Substitutions Influence A1AR OrthostericLigand Affinity

The affinity of the orthosteric antagonist [3H]DPCPX andorthosteric agonist NECA was determined from whole-cellsaturation and competition binding assays, respectively(Fig. 2, A and B; Table 1). Compared with the WT 3xHA-A1AR, alanine substitution of multiple residues (N159ECL2,

TABLE 1Orthosteric whole-cell equilibrium binding and agonist operational efficacy parameters at WT andmutant 3xHA-A1ARs stably expressed in Flp-In-CHO cellsValues represent means 6 S.E.M. from the indicated number of individual experiments (in parentheses) performed induplicate.

Mutation[3H]DPCPX NECA

pKD pKI Logtca

WT 9.17 6 0.04 (16) 6.59 6 0.04 (44) 1.78 6 0.09 (20)F144A 9.23 6 0.06 (5) 6.62 6 0.13 (10) 1.58 6 0.05 (6)G145A 9.00 6 0.02 (5) 6.38 6 0.08 (9) 1.46 6 0.02 (6)W146A 8.97 6 0.12 (5) 6.20 6 0.17 (6)* 1.63 6 0.06 (6)N147A 9.08 6 0.09 (5) 6.40 6 0.14 (6) 1.63 6 0.06 (5)N148A 8.98 6 0.08 (4) 6.70 6 0.09 (8) 1.21 6 0.09 (4)*L149A 9.35 6 0.10 (6) 6.94 6 0.10 (8)* 0.86 6 0.35 (3)*SV150AA 8.81 6 0.14 (3) 6.37 6 0.13 (10) 2.15 6 0.31 (3)E153A 9.17 6 0.09 (5) 6.58 6 0.07 (11) 1.34 6 0.04 (6)*R154A 9.06 6 0.07 (5) 6.52 6 0.07 (13) 1.16 6 0.22 (5)*W156A 9.07 6 0.07 (5) 6.5 6 0.09 (12) 0.89 6 0.19 (6)*N159A 8.56 6 0.11 (9)* 5.95 6 0.06 (12)* 1.64 6 0.08 (6)G160A 9.30 6 0.12 (5) 6.80 6 0.10 (10) 1.46 6 0.12 (5)S161A 8.66 6 0.13 (6)* 6.10 6 0.10 (8)* 1.73 6 0.09 (5)MGE162AAA 8.82 6 0.08 (3) 6.43 6 0.08 (9) 2.17 6 0.23 (3)P165A 9.05 6 0.08 (3) 6.45 6 0.08 (10) 1.72 6 0.05 (7)V166A 9.05 6 0.11 (3) 6.45 6 0.10 (12) 1.72 6 0.06 (7)I167A 8.42 6 0.13 (6)* 5.74 6 0.06 (12)* 1.45 6 0.04 (7)K168A 9.12 6 0.06 (3) 6.43 6 0.04 (12) 1.69 6 0.09 (7)E170A 9.24 6 0.15 (3) 6.22 6 0.06 (12)* 1.56 6 0.05 (7)F171A NA NA NAE172A 8.66 6 0.20 (3)* 5.84 6 0.08 (11)* 1.57 6 0.14 (9)K173A 9.21 6 0.02 (3) 6.42 6 0.08 (11) 1.56 6 0.05 (7)V174A 9.12 6 0.13 (3) 6.46 6 0.10 (8) 1.50 6 0.03 (7)I175A 8.45 6 0.10 (5)* 5.56 6 0.08 (7)* 1.45 6 0.06 (5)S176A 9.08 6 0.14 (3) 6.38 6 0.06 (13) 1.54 6 0.03 (7)M177A 8.39 6 0.17 (6)* 5.81 6 0.05 (11)* 1.52 6 0.05 (7)K265A (ECL3) 8.98 6 0.09 (4) 6.34 6 0.04 (6) 1.18 6 0.15 (6)*T270A (7.35) 8.83 6 0.04 (3) 5.76 6 0.05 (13)* 1.48 6 0.06 (12)*

pKD and pKI are the negative logarithms of the radioligand [3H]DPCPX and NECA equilibrium dissociation constants,respectively.

aLogtc values were corrected for changes in receptor expression by fold-normalization to the Bmax value of the WTreceptor. A transducer slope of 1.17 was shared across all data sets, because the fit was not significantly improved byallowing separate transducer slope values for the different constructs (as determined by an extra sum of squares, F test).

Logtc is logarithm of the operational efficacy parameter for NECA determined using the operational model of agonism,in which NECA pKI values were constrained to those determined from the radioligand binding assays.

*P , 0.05 (significantly different from WT value; one-way analysis of variance with the Dunnett post hoc test).NA, not applicable (no competition with the radioligand or no inhibition of cAMP).

Role of ECL2 on A1 Receptor Pharmacology 707

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

S161ECL2, I167ECL2, E172ECL2, I175ECL2, and M177ECL2)located in the middle to end of ECL2 significantly decreased[3H]DPCPX affinity (pKD) (Fig. 3A). The same residues had asimilar or greater effect on NECA affinity (pKI) (Fig. 3B),suggesting that these ECL2 residues facilitate, through director indirect interactions, the binding of [3H]DPCPX and NECAat A1AR. Three additional ECL2 mutations (W146ECL2A,L149ECL2A, and E170ECL2A) had a modest effect on NECAaffinity (Fig. 3B; Table1).

ECL2 Alanine Substitutions Influence A1AR OrthostericAgonist Efficacy

To investigate the influence of ECL2 on A1AR agonistefficacy, NECA-mediated inhibition of forskolin-stimulatedcAMP accumulation was quantified in WT and mutant 3xHA-A1AR Flp-In-CHO cell lines (Fig. 2C). The NECA-mediatedinhibition of cAMP accumulation was comparable in theabsence and presence of adenosine deaminase, suggestingminimal influence of endogenous adenosine (unpublisheddata). Furthermore, we recently demonstrated that the Flp-In-CHO cell line stably expressing the WT A1AR was notassociated with measureable constitutive activity (Vecchioet al., 2016). Similarly, comparison of the cAMP accumulationstimulated by forskolin (3 mM) showed no evidence for A1ARconstitutive activity in Flp-In-CHO cells stably expressingmutant A1ARs (unpublished data). An operational measure of

efficacy (logt), corrected for relative receptor expression levels,was used to quantify efficacy changes at mutant 3xHA-A1ARscompared with the WT 3xHA-A1AR. Negligible NECA-mediated inhibition of cAMP accumulation was observedin the Flp-In-CHO cell line expressing the 3xHA-A1AR(F171ECL2A) (data not shown). Interestingly, the cluster ofseven residues at the C-terminal end of ECL2 that signifi-cantly reduced NECA affinity had no effect on NECA efficacy(logtc). However, NECA efficacy was significantly reduced atfive of the mutant 3xHA-A1ARs (Fig. 4; Table 1). Of note,NECA efficacy was significantly reduced upon alanine sub-stitution of N148ECL2, E153ECL2, R154ECL2, and W156ECL2,mutations that had no significant effect on NECA affinity. Incontrast with the effect of L149ECL2A in improving NECAaffinity, the same mutation decreased NECA efficacy.

Ligand Docking and MD Simulations

Inactive and active-like homology models of human A1ARwere generated based on the crystal structures of humanA2AAR (Protein Data Bank identifiers 3EML and 3QAK,respectively), occupied by antagonist ZM241385 and agonistUK432097 (6-[2,2-di(phenyl)ethylamino]-9-[(2R,3R,4S,5S)-5-(ethylcarbamoyl)-3,4-dihydroxyoxolan-2-yl]-N-[2-[(1-pyridin-2-ylpiperidin-4-yl)carbamoylamino]ethyl]purine-2-carboxamide)(Jaakola et al., 2008; Xu et al., 2011). Ligand docking and MDsimulations were performed using our inactive and active-likeA1AR homology models to assist with the interpretation ofmutagenesis findings. The ICM PocketFinder algorithmidentified three potential binding pockets within our A1ARhomology models. DPCPX and NECA were docked into thepocket located within the TM bundle; however, the size of thedocking box was adjusted to be large enough to encompassthe extracellular region of the receptor. Docking simulations ofDPCPX resulted in a single cluster within the TM bundle,where DPCPX formed a p-stacking interaction with F171ECL2

(Fig. 5A); this key interaction was maintained during a 40-nsMD simulation (Table 2). Furthermore, the key residue(N2546.55) involved in hydrogen-bonding interactions in bothantagonist- and agonist-bound A2AAR crystal structures,formed a hydrogen bond with DPCPX during the MD simula-tion (Table 2). E172ECL2, another key residue involved in theformation of the DPCPX binding pocket (i.e., found within3.5 Å of DPCPX during the MD simulation), also decreased[3H]DPCPX affinity when mutated to alanine (Table 2).

Fig. 3. The change in affinity of DPCPX (ΔpKD; A) and NECA (ΔpKI; B) atmutant 3xHA-A1ARs relative toWT. Data represent the means6 S.E.M ofat least three experiments performed in duplicate. *P , 0.05 (comparedwith WT; one-way analysis of variance, Dunnett post hoc test).

Fig. 4. The change in signaling efficacy of NECA (Δlogtc) at mutant 3xHA-A1ARs relative to WT. Data represent the means6 S.E.M of at least threeexperiments performed in duplicate. *P , 0.05 (compared with WT; one-way analysis of variance, Dunnett post hoc test).

708 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

The root-mean-square deviation of DPCPX compared with itsfinal pose was calculated. Over the course of the simulation,DPCPX remained relatively stable after 12 ns (SupplementalFig. 2).Interestingly, molecular docking of NECA revealed two

main clusters among the top 40 docking poses, one withinthe TM bundle (site 1) and the other within the extracellularregion (site 2) (Fig. 5B). Similar findings were also observedwhen we repeated the molecular docking using the endoge-nous agonist adenosine instead of NECA (SupplementalFig. 3). The NECA poses found within the TM bundle (site 1)were located within the predicted A1AR orthosteric agonistsite. Within site 1, poses could be further divided into two

subclusters. One subcluster had a similar orientation of theadenine ring to that found within the NECA-bound A2AARcrystal structure and was relatively stable during our 40-nsMD simulation (Supplemental Fig. 2). During this simulation,NECA predominantly formed hydrogen-bond interactionswith the residues T913.36, N2546.55, and T2777.42. F171ECL2

was the only ECL2 residue located within 3.5 Å of NECAduring thisMD simulation (Table 2). The alternative cluster ofNECA site 1 poses was not considered for MD simulations,because the adenine ring was inverted relative to the positionof NECA bound within the A2AAR crystal structure.NECA poses that docked within the A1AR extracellular

region (site 2) formed hydrogen bonds with the three ECLsand residues at the top of TM6 and TM7. A 40-ns MDsimulation was performed on the highest-ranked (lowestICM score) pose within the extracellular vestibule. Within thisMD simulation, NECA was relatively stable and formedhydrogen bonds with a number of residues, includingE170ECL2, E172ECL2, N2546.55, K265ECL3, P266ECL3, andT2707.35 (Table 2; Supplemental Fig. 2). ECL2 residueslocated within 3.5 Å of NECA during this MD simulation wereE170ECL2, F171ECL2, E172ECL2, and M177ECL2 (Table 2).These findings are consistent with mutagenesis data thatsuggest each of these residues are involved in the binding ofNECA to A1AR (Table 1).A main pose of each ligand, which formed key interactions

as determined by the interaction frequency analysis of theMDtrajectory, was obtained from each 40-nsMD simulation (Figs.6 and 7; Supplemental Material Files 1–3). The purine ringsfor the DPCPX and NECA (site 1) poses overlap, stabilized bya hydrogen bond with N2546.55 and p-stacking interactionswith F171ECL2 (Supplemental Fig. 4). Residues within theextracellular vestibule, top of TM6 (L2506.51, L2536.54, andT2576.58), and E172ECL2 surrounded the DPCPX cyclopentylring, whereas the NECA ribose moiety made contacts withresidues deeper within the TM bundle (V873.32, L883.33,Q923.37, M1805.40, andW2476.48). NECAwas further anchoredby a predicted hydrogen bond with T913.36 (Fig. 7, A and B).Therefore, the residues predicted to form the NECA site1 binding pocket involve conserved residues between A1ARand A2AAR, V873.32, L883.33, T913.36, Q923.37, M1805.40,W2476.48, L2506.51, andN2546.55. These residues are predictedto be involved in conformational rearrangements of TM3,TM5, TM6, and TM7 upon receptor activation and facilitatethe binding of orthosteric agonists (Supplemental Fig. 4)(Lebon et al., 2011; Xu et al., 2011). The main NECA bindingpose within site 2 was predicted to be stabilized by hydrogen-bond interactions between the adenine ring and E170ECL2,E172ECL2, S267ECL3, and T2707.35 as well as a hydrogen bondbetween the 29-OH group of the ribose ring and K265ECL3 (Fig.7, C and D). The NECA site 2 binding pocket was also closelypacked by F171ECL2, M177ECL2, and T2576.58 side chains.Interestingly, the NECA (site 2) and DPCPX poses were nottopographically distinct and were instead predicted to par-tially overlap.

Alanine Substitution of Predicted Binding Pocket Residuesfrom Molecular Modeling Affects NECA Affinity and/orEfficacy

Results from the NECA site 1 MD simulations showed thatT913.36, F171ECL2, andW2476.48 located within 3.5 Å of NECA

Fig. 5. Docking of DPCPX andNECA into A1AR homologymodels. (A) Thetop 20 DPCPX (sticks; carbon, nitrogen, oxygen, and polar hydrogen atomsare colored in cyan, blue, red, andwhite, respectively) poses docked into theinactiveA1ARhomologymodel (gray ribbons). F171ECL2, predicted to formakeyp-stacking interaction with DPCPX, is shown as blue sticks. (B) The top40 NECA (sticks; carbon, nitrogen, oxygen, and polar hydrogen atoms arecolored in green, blue, red, and white, respectively) poses docked into theactive-like A1AR homology model (light blue ribbons). Residues predictedto be involved in hydrogen-bond and p-stacking interactions are shown asblue sticks.

Role of ECL2 on A1 Receptor Pharmacology 709

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

heavy atoms for at least 60% of the total MD simulation time.However, the NECA site 2 MD simulation showed thatE172ECL2, K265ECL3, and T2707.35 located within 3.5 Å ofNECA heavy atoms for at least 60% of the total MD simula-tion time. The corresponding conserved residues T913.36,F171ECL2, and E172ECL2 directly interacted with NECAwithin the A2AAR crystal structure. Furthermore, the in-volvement of TM residue T913.36 on A1AR agonist binding wasconfirmed previously through mutagenesis studies (Rivkeeset al., 1999). The conserved tryptophan at position 6.48 waspreviously suggested to be involved in the activation processof GPCRs (Shi et al., 2002; Kobilka, 2007; Stoddart et al.,2014). Given the known role of T913.36 and W2476.48 in A1ARorthosteric binding, these residues were not investigatedfurther in our study. However, the influence of the predictedNECA site 2 residues, K265ECL3 and T2707.35, on agonistbinding or function has not yet been investigated at the humanA1AR. Therefore, we investigated the influence of single-point alanine substitutions of K265ECL3A and T2707.35A onA1AR pharmacology. Consistent with DPCPX binding withinthe TM bundle, neither mutation significantly influencedreceptor expression (Fig. 8A) or DPCPX affinity (Fig. 8B). Incontrast, alanine substitution of T2707.35 caused a signifi-cant decrease in NECA affinity (Fig. 8C). Furthermore, bothmutations significantly decreased NECA efficacy (Fig. 8D).These data suggest that residues in the extracellular regionof A1AR play an important role in both agonist binding andfunction.

DiscussionA1AR represents a potential therapeutic target for a

number of conditions. A detailed understanding of thestructure-function relationships at A1AR can facilitate therational design of more potent and selective A1AR ligands.Since a high-resolution structure of A1AR bound to differentclasses of ligand is not yet available, less direct methods suchas site-directed mutagenesis and molecular modeling thatcan probe the structure-function properties underlying A1ARorthosteric ligands remain extremely valuable. Given thatECL2 has been suggested to have a critical role in the bindingand function of orthosteric ligands at a range of GPCRs(Wheatley et al., 2012), we performed alanine scanningmutagenesis of this region combined with molecular model-ing to identify the role of this domain on orthosteric ligandpharmacology. Moreover, key residues predicted from MD tobe involved in orthosteric agonist function were mutated,and the effects of the mutations validated the importance ofthese residues. Our A1AR-ECL2 mutagenesis data are ingood agreement with the binding modes of DPCPX andNECA from molecular modeling, identifying a key role forECL2 in A1AR orthosteric ligand binding and function.However, our data also suggest that the orthosteric agonistNECA recognizes a binding site in an extracellular vestibulethat may represent a transition pose prior to entry into thecanonical orthosteric site. This vestibular region may repre-sent part of an allosteric binding site for A1AR modulators,

TABLE 2A1AR residues that interact with DPCPX or NECA through the course of each MD simulationValues represent the percentage of the MD simulation time (40 ns) that the residue and ligand heavy atoms are within 3.5 Å of one another.

Residue DPCPXNECA

Site 1 Site 2

V622.56 — 5.7 —

I692.63 14.1 — —N70ECL1 8.9 — 5.4V873.32 54.0 26.6 —

L883.33 — 20.3 —T913.36 14.7 99.6 (L-RS: 44.4) —Q923.37 — 22.3 —

E170ECL2 — — 46 (L-RS: 19.7)F171ECL2 81.5 65.9 11.4E172ECL2 6.9 16.5 64.8M177ECL2 — — 25.6M1805.40 38.7 40.3 —

N1845.43 — 37.9 —W2476.48 34.2 63.9 —L2506.51 22.9 9.0 —

H2516.52 — 9.1 —L2536.54 8.9 — —

N2546.55 90.8 (L-RS: 38.7; RS-L: 14.4) 57.6 (L-RS: 12.9; RS-L: 10.2) 27.8T2576.58 51.8 — 15.1H264ECL3 — — 32.4K265ECL3 — — 98.1 (L-RM: 39.4; RM-L: 13.2)P266ECL3 — — 98.1S267ECL3 — — 57.5T2707.35 6.8 — 76.4 (L-RS: 5.8; RS-L: 14.1)Y2717.36 — — 44.9I2747.39 18.6 29.8 —T2777.42 — 42.9 (L-RS: 1.2; RS-L:4.5) —H2787.43 — 12.1 —

Residues were considered to interact with the ligand if they were in contact with the ligand for .5% of the simulation time. Hydrogen bondinteractions (in parentheses) were defined with the donor acceptor distance, 3.5 Å, and an angle cutoff of 20° and are displayed as a percentage ofMD simulation time. All values were calculated for one frame per 10 ps using VMD software (version 1.9.2). L-RM, hydrogen bond interactionbetween ligand (donor) and residue main chain (acceptor); L-RS, hydrogen bond interaction between ligand (donor) and residue side chain(acceptor); RM-L, hydrogen bond interaction between residue main chain (donor) and ligand (acceptor); RS-L, hydrogen bond interaction betweenresidue side chain (donor) and ligand (acceptor).

710 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

and this is explored in greater detail in our accompanyingarticle.Previous A1AR site-directed mutagenesis studies predomi-

nantly focused on TM domains, with residues in TM1, TM2,TM3,TM4,TM5, andTM7 implicated in thebinding of orthostericagonists and antagonists (Olah et al., 1994; Townsend-Nicholson

and Schofield, 1994; Barbhaiya et al., 1996; Rivkees et al., 1999;Fredholm et al., 2001). Until now, the involvement of ECL2 onorthosteric ligand binding and function has not been systemati-cally investigated atA1AR.However, a critical role for this domaininorthosteric pharmacologyhasbeen identified frommutagenesisstudies on otherGPCRs, such as theA3AR, dopamineD2 receptor,M2 and M3 muscarinic receptors, V1a vasopressin receptor, andglucagon-like peptide-1 receptor (Howl and Wheatley, 1996; Gaoet al., 2002; Shi andJavitch, 2004;Duong et al., 2005;Avlani et al.,2007; Scarselli et al., 2007; Koole et al., 2012). It is suggested thatorthosteric ligand binding to an unoccupied GPCR, which has amore exposedECL2, can induce a conformational change, causingECL2 to close over the entrance to the binding crevice and therebystabilize the ligand within the orthosteric pocket (Banères et al.,2005; Unal et al., 2010). Supporting this notion, ECL2 contributesto the orthosteric pocket in A2AAR crystal structures. The shorthelical domain within A2AAR-ECL2 is oriented toward the TMcore region and interacts directlywith agonists and antagonists inthe orthosteric site (Jaakola et al., 2008; Doré et al., 2011; Lebonet al., 2011, 2015; Xu et al., 2011;Congreve et al., 2012;Hino et al.,2012; Liu et al., 2012). Molecular modeling performed withinour study also suggests that the short helical domain withinA1AR-ECL2 forms the upper region of the principal orthostericpocket. The main pose from MD simulations performed onDPCPX and NECA docked within the TM bundle found twoECL2 residues, F171 and E172, located within 3.5 Å of theorthosteric ligand (Figs. 6 and 7). A decrease in [3H]DPCPX andNECA affinity was observed for E172ECL2A and no [3H]DPCPXbinding was observed at the F171ECL2A mutation, supportingthe role of the residues in DPCPX binding at A1AR,.Docking of DPCPX and NECA into homology models based

on inactive andactive-likeA2AARcrystal structures, respectively,

Fig. 7. Predicted binding mode of NECA (green sticks) at thepartially active A1AR homology model (light blue ribbons)located in the orthosteric site (site 1; A and B) and extracellularvestibule (site 2; C and D). Residues that significantly influ-enced NECA affinity (A and C) or efficacy (B and D) are shownas yellow (,5-fold change), orange (between 5- and 10-foldchange), and red (.10-fold change) sticks. Key residues pre-dicted to form hydrogen-bond interactions with NECA areshown as blue sticks.

Fig. 6. Predicted binding mode of DPCPX (cyan sticks) at the inactiveA1AR homology model (gray ribbons). Residues that significantly influ-enced DPCPX affinity are shown as yellow (,5-fold change) and orange(between 5- and 10-fold change) sticks. Key residues predicted to formhydrogen-bond (N2546.55) or p-stacking (F171ECL2) interactions withDPCPX are shown as blue sticks.

Role of ECL2 on A1 Receptor Pharmacology 711

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

revealed overlapping binding poses, with NECA (bound to site1) making additional contacts with residues deeper within theTM bundle than DPCPX. Previous studies showed that TM3residues differentially affected A1AR orthosteric antagonistversus agonist binding. Mutation of L883.33A caused a sub-stantial reduction of both agonist and antagonist affinity atA1AR, whereas V873.32A was shown to decrease antagonistaffinity alone and Q923.37A was shown to perturb agonistaffinity only (Rivkees et al., 1999). In the main poses afterMD simulation, V873.32 and L883.33A were predicted to havehydrophobic interactions with both DPCPX and NECA,whereas Q923.37 only interacted with NECA (site 1) (Supple-mental Fig. 4). Therefore, the key interactions predicted fromour molecular modeling are in agreement with previousmutagenesis studies and further support a deeper orthostericbinding pose for NECA (site 1) relative to the antagonist,DPCPX. It should be noted, however, that the single identifiedstable pose of the antagonist DPCPX is higher placed, suchthat it overlaps with both NECA binding clusters [i.e., withinthe orthosteric agonist pocket (site 1) and the extracellularvestibule (site 2) to antagonize agonist access].ECL2 has also been suggested to facilitate ligand access

from the extracellular space into the orthosteric TM bindingcavity. For instance, in unbiased long-timescale MD simula-tions, b-adrenoceptor orthosteric ligands recognized an in-termediate binding site within the extracellular vestibuleduring the transition from the extracellular space into theTM bundle (Dror et al., 2011). Furthermore, the importance ofECL2 flexibility and conformational integrity was highlightedin a study on the M2 muscarinic acetylcholine receptor, inwhich a reduction in ECL2 flexibility due to the introduction ofa disulfide bond between ECL2 and TM7 caused a significantdecrease in orthosteric and allosteric ligand affinity (Avlaniet al., 2007). Our docking of NECA in the active-like A1AR

homology model showed that 31 of the top 40 poses recognizeda common site within the extracellular vestibule, supportingthe role of ECL2 in facilitating ligand entry into the orthos-teric site. The main pose from the MD simulation performedon NECA docked within this site had hydrogen-bondinginteractions between the ligand and E170ECL2, E172ECL2,K265ECL3, S267ECL3, and T2707.35. ECL2 residues locatedwithin 3.5 Å of this NECA pose were E170ECL2, F171ECL2,E172ECL2, and M177ECL2. Two of these residues, F171ECL2

and E172ECL2, are also involved in the formation of theorthosteric pocket within the TM bundle. Nonetheless, theexistence of the site 2 binding pocket is supported by ECL2alanine scanning mutagenesis, which showed a significantdecrease in NECA affinity at the A1ARmutations E170ECL2A,E172ECL2A, and M177ECL2A. Furthermore, alanine mutationof residues predicted from ourMD simulations to interact withNECA at this extracellular site, K265ECL3 and T2707.35,caused a significant decease in NECA affinity and/or efficacy.Of note, b2-adrenoceptor residues that form key interactionswith orthosteric ligands within the extracellular vestibule,Y3087.35 and F193ECL2 (Dror et al., 2011), correspond toT2707.35 and F171ECL2 identified within our study. Initialrecognition within the extracellular vestibule (site 2) may actas a “selectivity filter” for GPCRs. The relatively open natureof the extracellular vestibule may lead to site 2 being morepopulated by a range of ligands; however, only those that formthe appropriate metastable poses will eventually transitioninto site 1. In contrast, ligands that do not form thesemetastable interactions within site 2 will leave this site anddissociate from the receptor. Collectively, these data addweight to the notion that the ECL2 of numerous class AGPCRs contributes to an extracellular vestibule that can bindsmall molecules, including orthosteric agonists as well asallosteric modulators.Within this study, alanine mutation of a number of residues

in the middle of ECL2, which are predicted from molecularmodeling to be distinct from the orthosteric site, caused asignificant decrease in orthosteric ligand affinity and/orefficacy. This is perhaps unsurprising, given the flexiblenature of ECL2 and therefore the spectrum of conformationsit likely assumes. Furthermore, ECL2 has not been welldefined within A2AAR crystal structures, with many residuesnot resolved within this region. However, ECL2 residuessuggested from alanine scanning mutagenesis to be involvedin orthosteric ligand binding may influence the energeticsof the unbound receptor conformations and/or facilitatethe transition of ligands from the extracellular space intothe orthosteric site. Indeed, we identified a cluster of fourresidues at the TM4 end of ECL2 that perturbed NECAefficacy in the absence of an effect on affinity, indicating thatECL2 is involved in the stabilization of A1AR active states.These findings are in line with previous studies of otherfamily A GPCRs in which mutations within ECL2 perturbedorthosteric agonist efficacy (May et al., 2007; Scarselli et al.,2007; Gregory et al., 2010, 2012; Wifling et al., 2015). At anumber of GPCRs, including the M2 muscarinic acetylcholinereceptor and the glucagon-like peptide-1 receptor, ECL2differentially influences orthosteric agonist efficacy depend-ing upon the signaling pathway assessed (Gregory et al., 2010,2012; Koole et al., 2012); as such, ECL2 is thought to play a keyrole in the stabilization of different active states that contrib-ute to the phenomenon of biased agonism. In the future, it will

Fig. 8. The influence of alanine substitution on residues predicted fromMD simulations to be involved in site 2 NECA binding. (A) Receptorexpression (Bmax) of WT and mutant 3xHA-A1ARs stably expressed in Flp-In-CHO cells. Bmax values are expressed as a percentage of WT. Thechange in [3H]DPCPX affinity (B), NECA affinity (C), and NECA efficacy(D) is shown for mutant 3xHA-A1ARs relative to WT. Data represent themeans 6 S.E.M of three experiments performed in duplicate. *P , 0.05(compared with WT; one-way analysis of variance, Dunnett post hoc test).

712 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

be pertinent to explore the influence of these ECL2 residues onthe bias profile of A1AR agonists.Informed by our iterative approach that combined site-

directed mutagenesis and molecular modeling with rigorousanalysis delineating effects of select A1AR residues on orthos-teric affinity and efficacy, we have demonstrated a key role forECL2 in A1AR agonist and antagonist binding and function.Our analysis indicated that agonist ligand affinity and efficacydeterminants could be clearly delineated into two majorclusters. In addition to making direct contact with ligandsbound to the TM bundle, we propose that ECL2 forms anextracellular vestibule that is recognized by, and facilitatesthe binding of orthosteric agonists. It would be of interest todetermine the extent to which this vestibule also contributesto the actions of A1AR allosteric ligands.

Acknowledgments

The authors thank Thomas Coudrat for assistance with molecularmodeling.

Authorship Contributions

Participated in research design: A.N.N. Nguyen, Christopoulos,May.

Conducted experiments: A.N.N. Nguyen, Baltos, Muñoz, Gregory.Performed data analysis: A.N.N. Nguyen, Thomas, T.D. Nguyen.Wrote or contributed to the writing of the manuscript: A.N.N.

Nguyen, Gregory, White, Sexton, Christopoulos, May.

References

Abagyan R and Kufareva I (2009) The flexible pocketome engine for structural che-mogenomics. Methods Mol Biol 575:249–279.

An J, Totrov M, and Abagyan R (2005) Pocketome via comprehensive identificationand classification of ligand binding envelopes. Mol Cell Proteomics 4:752–761.

Avlani VA, Gregory KJ, Morton CJ, Parker MW, Sexton PM, and Christopoulos A(2007) Critical role for the second extracellular loop in the binding of bothorthosteric and allosteric G protein-coupled receptor ligands. J Biol Chem 282:25677–25686.

Ballesteros J and Weinstein H (1995) Integrated methods for the construction ofthree dimensional models and computational probing of structure function rela-tions in G-protein-coupled receptors. Methods Neurosci 25:366–428.

Baltos JA, Gregory KJ, White PJ, Sexton PM, Christopoulos A, and May LT (2016)Quantification of adenosine A(1) receptor biased agonism: implications for drugdiscovery. Biochem Pharmacol 99:101–112.

Banères JL, Mesnier D, Martin A, Joubert L, Dumuis A, and Bockaert J (2005)Molecular characterization of a purified 5-HT4 receptor: a structural basis for drugefficacy. J Biol Chem 280:20253–20260.

Barbhaiya H, McClain R, Ijzerman A, and Rivkees SA (1996) Site-directed muta-genesis of the human A1 adenosine receptor: influences of acidic and hydroxyresidues in the first four transmembrane domains on ligand binding. MolPharmacol 50:1635–1642.

Black JW and Leff P (1983) Operational models of pharmacological agonism. Proc RSoc Lond B Biol Sci 220:141–162.

Cheng Y and Prusoff WH (1973) Relationship between the inhibition constant (K1)and the concentration of inhibitor which causes 50 per cent inhibition (I50) of anenzymatic reaction. Biochem Pharmacol 22:3099–3108.

Congreve M, Andrews SP, Doré AS, Hollenstein K, Hurrell E, Langmead CJ, MasonJS, Ng IW, Tehan B, Zhukov A, et al. (2012) Discovery of 1,2,4-triazine derivativesas adenosine A(2A) antagonists using structure based drug design. J Med Chem 55:1898–1903.

Doré AS, Robertson N, Errey JC, Ng I, Hollenstein K, Tehan B, Hurrell E, Bennett K,Congreve M, Magnani F, et al. (2011) Structure of the adenosine A(2A) receptor incomplex with ZM241385 and the xanthines XAC and caffeine. Structure 19:1283–1293.

Dror RO, Pan AC, Arlow DH, Borhani DW, Maragakis P, Shan Y, Xu H, and ShawDE (2011) Pathway and mechanism of drug binding to G-protein-coupled receptors.Proc Natl Acad Sci USA 108:13118–13123.

Duan Y, Wu C, Chowdhury S, Lee MC, Xiong G, Zhang W, Yang R, Cieplak P, Luo R,Lee T, et al. (2003) A point-charge force field for molecular mechanics simulationsof proteins based on condensed-phase quantum mechanical calculations. J ComputChem 24:1999–2012.

Duong HT, Gao ZG, and Jacobson KA (2005) Nucleoside modification and concertedmutagenesis of the human A3 adenosine receptor to probe interactions betweenthe 2-position of adenosine analogs and Gln167 in the second extracellular loop.Nucleosides Nucleotides Nucleic Acids 24:1507–1517.

Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, and Pedersen LG (1995) Asmooth particle mesh Ewald method. J Chem Phys 103:8577–8593.

Eswar N, Webb B, Marti-Renom MA, Madhusudhan MS, Eramian D, Shen MY,Pieper U, and �Sali A (2007) Comparative protein structure modeling usingMODELLER. Curr Protoc Protein Sci Chapter 2:Unit 2.9.

Fredholm BB, IJzerman AP, Jacobson KA, Klotz KN, and Linden J (2001) In-ternational Union of Pharmacology. XXV. Nomenclature and classification ofadenosine receptors. Pharmacol Rev 53:527–552.

Gao ZG, Chen A, Barak D, Kim SK, Müller CE, and Jacobson KA (2002) Iden-tification by site-directed mutagenesis of residues involved in ligand recogni-tion and activation of the human A3 adenosine receptor. J Biol Chem 277:19056–19063.

Gregory KJ, Hall NE, Tobin AB, Sexton PM, and Christopoulos A (2010) Identifi-cation of orthosteric and allosteric site mutations in M2 muscarinic acetylcholinereceptors that contribute to ligand-selective signaling bias. J Biol Chem 285:7459–7474.

Gregory KJ, Sexton PM, Tobin AB, and Christopoulos A (2012) Stimulus bias pro-vides evidence for conformational constraints in the structure of a G protein-coupled receptor. J Biol Chem 287:37066–37077.

Hess S (2001) Recent advances in adenosine receptor antagonist research. ExpertOpin Ther Pat 11:1533–1561.

Hino T, Arakawa T, Iwanari H, Yurugi-Kobayashi T, Ikeda-Suno C, Nakada-NakuraY, Kusano-Arai O, Weyand S, Shimamura T, Nomura N, et al. (2012) G-protein-coupled receptor inactivation by an allosteric inverse-agonist antibody.Nature 482:237–240.

Howl J and Wheatley M (1996) Molecular recognition of peptide and non-peptideligands by the extracellular domains of neurohypophysial hormone receptors.Biochem J 317:577–582.

Humphrey W, Dalke A, and Schulten K (1996) VMD: visual molecular dynamics.J Mol Graph 14:33–38, 27–28.

Jaakola VP, Griffith MT, Hanson MA, Cherezov V, Chien EYT, Lane JR, IJzermanAP, and Stevens RC (2008) The 2.6 angstrom crystal structure of a human A2Aadenosine receptor bound to an antagonist. Science 322:1211–1217.

Jacobson KA and Gao Z-G (2006) Adenosine receptors as therapeutic targets.Nat RevDrug Discov 5:247–264.

Kalk P, Eggert B, Relle K, Godes M, Heiden S, Sharkovska Y, Fischer Y, Ziegler D,Bielenberg GW, and Hocher B (2007) The adenosine A1 receptor antagonistSLV320 reduces myocardial fibrosis in rats with 5/6 nephrectomy without affectingblood pressure. Br J Pharmacol 151:1025–1032.

Kim J, Jiang Q, Glashofer M, Yehle S, Wess J, and Jacobson KA (1996) Glutamateresidues in the second extracellular loop of the human A2a adenosine receptor arerequired for ligand recognition. Mol Pharmacol 49:683–691.

Kobilka BK (2007) G protein coupled receptor structure and activation. BiochimBiophys Acta 1768:794–807.

Koole C, Wootten D, Simms J, Miller LJ, Christopoulos A, and Sexton PM (2012)Second extracellular loop of human glucagon-like peptide-1 receptor (GLP-1R) hasa critical role in GLP-1 peptide binding and receptor activation. J Biol Chem 287:3642–3658.

Leach K, Davey AE, Felder CC, Sexton PM, and Christopoulos A (2011) The role oftransmembrane domain 3 in the actions of orthosteric, allosteric, and atypicalagonists of the M4 muscarinic acetylcholine receptor. Mol Pharmacol 79:855–865.

Lebon G, Edwards PC, Leslie AGW, and Tate CG (2015) Molecular determinants ofCGS21680 binding to the human adenosine A2A receptor. Mol Pharmacol 87:907–915.

Lebon G, Warne T, Edwards PC, Bennett K, Langmead CJ, Leslie AGW, and Tate CG(2011) Agonist-bound adenosine A2A receptor structures reveal common features ofGPCR activation. Nature 474:521–525.

Liu W, Chun E, Thompson AA, Chubukov P, Xu F, Katritch V, Han GW, Roth CB,Heitman LH, IJzerman AP, et al. (2012) Structural basis for allosteric regulation ofGPCRs by sodium ions. Science 337:232–236.

MacKerell AD, Bashford D, Bellott M, Dunbrack RL, Evanseck JD, Field MJ, FischerS, Gao J, Guo H, Ha S, et al. (1998) All-atom empirical potential for molecularmodeling and dynamics studies of proteins. J Phys Chem B 102:3586–3616.

MacKerell AD, Jr, Feig M, and Brooks CL, 3rd (2004) Extending the treatment ofbackbone energetics in protein force fields: limitations of gas-phase quantum me-chanics in reproducing protein conformational distributions in molecular dynamicssimulations. J Comput Chem 25:1400–1415.

Martinelli A and Tuccinardi T (2008) Molecular modeling of adenosine receptors: newresults and trends. Med Res Rev 28:247–277.

May LT, Avlani VA, Langmead CJ, Herdon HJ, Wood MD, Sexton PM,and Christopoulos A (2007) Structure-function studies of allosteric agonism at M2muscarinic acetylcholine receptors. Mol Pharmacol 72:463–476.

Müller CE and Jacobson KA (2011) Recent developments in adenosine receptor li-gands and their potential as novel drugs. Biochim Biophys Act 1808:1290–1308.

Olah ME, Jacobson KA, and Stiles GL (1994) Role of the second extracellular loop ofadenosine receptors in agonist and antagonist binding. Analysis of chimeric A1/A3adenosine receptors. J Biol Chem 269:24692–24698.

Palmer TM and Stiles GL (1995) Adenosine receptors. Neuropharmacology 34:683–694.

Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, Villa E, Chipot C, SkeelRD, Kalé L, and Schulten K (2005) Scalable molecular dynamics with NAMD.J Comput Chem 26:1781–1802.

Rivkees SA, Barbhaiya H, and IJzerman AP (1999) Identification of the adeninebinding site of the human A1 adenosine receptor. J Biol Chem 274:3617–3621.

Sabbadin D, Ciancetta A, Deganutti G, Cuzzolin A, and Moro S (2015) Exploringthe recognition pathway at the human A2A adenosine receptor of the endogenousagonist adenosine using supervised molecular dynamics. MedChemComm 6:1081–1085.

Scarselli M, Li B, Kim SK, and Wess J (2007) Multiple residues in the second ex-tracellular loop are critical for M3 muscarinic acetylcholine receptor activation.J Biol Chem 282:7385–7396.

Schiedel AC, Hinz S, Thimm D, Sherbiny F, Borrmann T, Maass A, and Müller CE(2011) The four cysteine residues in the second extracellular loop of the humanadenosine A2B receptor: role in ligand binding and receptor function. BiochemPharmacol 82:389–399.

Role of ECL2 on A1 Receptor Pharmacology 713

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from

Scholl DJ and Wells JN (2000) Serine and alanine mutagenesis of the nine nativecysteine residues of the human A(1) adenosine receptor. Biochem Pharmacol 60:1647–1654.

Shi L and Javitch JA (2004) The second extracellular loop of the dopamine D2 re-ceptor lines the binding-site crevice. Proc Natl Acad Sci USA 101:440–445.

Shi L, Liapakis G, Xu R, Guarnieri F, Ballesteros JA, and Javitch JA (2002) Beta2adrenergic receptor activation. Modulation of the proline kink in transmembrane6 by a rotamer toggle switch. J Biol Chem 277:40989–40996.

Shonberg J, Herenbrink CK, López L, Christopoulos A, Scammells PJ, Capuano B,and Lane JR (2013) A structure-activity analysis of biased agonism at the dopamineD2 receptor. J Med Chem 56:9199–9221.

Sromek SM and Harden TK (1998) Agonist-induced internalization of the P2Y2 re-ceptor. Mol Pharmacol 54:485–494.

Stoddart LA, Kellam B, Briddon SJ, and Hill SJ (2014) Effect of a toggle switch mu-tation in TM6 of the human adenosine A₃ receptor on Gi protein-dependent signallingand Gi-independent receptor internalization. Br J Pharmacol 171:3827–3844.

Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, and Higgins DG (1997) TheCLUSTAL_X windows interface: flexible strategies for multiple sequence align-ment aided by quality analysis tools. Nucleic Acids Res 25:4876–4882.

Townsend-Nicholson A and Schofield PR (1994) A threonine residue in the seventhtransmembrane domain of the human A1 adenosine receptor mediates specificagonist binding. J Biol Chem 269:2373–2376.

Unal H, Jagannathan R, Bhat MB, and Karnik SS (2010) Ligand-specific confor-mation of extracellular loop-2 in the angiotensin II type 1 receptor. J Biol Chem285:16341–16350.

Valant C, May LT, Aurelio L, Chuo CH, White PJ, Baltos JA, Sexton PM, ScammellsPJ, and Christopoulos A (2014) Separation of on-target efficacy from adverse ef-fects through rational design of a bitopic adenosine receptor agonist. Proc NatlAcad Sci USA 111:4614–4619.

Vanommeslaeghe K, Hatcher E, Acharya C, Kundu S, Zhong S, Shim J, Darian E,Guvench O, Lopes P, Vorobyov I, et al. (2010) CHARMM general force field: a forcefield for drug-like molecules compatible with the CHARMM all-atom additivebiological force fields. J Comput Chem 31:671–690.

Vecchio EA, Tan CY, Gregory KJ, Christopoulos A, White PJ, and May LT (2016)Ligand-independent adenosine A2B receptor constitutive activity as a promoter ofprostate cancer cell proliferation. J Pharmacol Exp Ther 357:36–44.

Wheatley M, Wootten D, Conner MT, Simms J, Kendrick R, Logan RT, Poyner DR,and Barwell J (2012) Lifting the lid on GPCRs: the role of extracellular loops. Br JPharmacol 165:1688–1703.

Wifling D, Bernhardt G, Dove S, and Buschauer A (2015) The extracellular loop2 (ECL2) of the human histamine H4 receptor substantially contributes to ligandbinding and constitutive activity. PLoS One 10:e0117185.

Xu F, Wu H, Katritch V, Han GW, Jacobson KA, Gao ZG, Cherezov V, and StevensRC (2011) Structure of an agonist-bound human A2A adenosine receptor. Science332:322–327.

Yaar R, Jones MR, Chen JF, and Ravid K (2005) Animal models for the study ofadenosine receptor function. J Cell Physiol 202:9–20.

Yu W, He X, Vanommeslaeghe K, and MacKerell AD, Jr (2012) Extension of theCHARMM General Force Field to sulfonyl-containing compounds and its utility inbiomolecular simulations. J Comput Chem 33:2451–2468.

Address correspondence to: Arthur Christopoulos or Lauren T. May, DrugDiscovery Biology and Department of Pharmacology, Monash Institute ofPharmaceutical Sciences, Monash University, 399 Royal Parade, Parkville,VIC 3052, Australia. E-mail: arthur.christopoulos@monash.edu or lauren.may@monash.edu

714 Nguyen et al.

at ASPE

T Journals on February 2, 2021

molpharm

.aspetjournals.orgD

ownloaded from