Importance of the second extracellular loop for melatonin MT₁ receptor function and absence of melatonin binding in GPR50

Abstract

Background and Purpose

Recent crystal structures of GPCRs have emphasized the previously unappreciated role of the second extracellular (E2) loop in ligand binding and gating and receptor activation. Here, we have assessed the role of the E2 loop in the activation of the melatonin MT₁ receptor and in the inactivation of the closely related orphan receptor GPR50.

Experimental Approach

Chimeric MT₁‐GPR50 receptors were generated and functionally analysed in terms of 2‐[¹²⁵I]iodomelatonin binding, G_i/cAMP signalling and β‐arrestin2 recruitment. We also used computational molecular dynamics (MD) simulations.

Key Results

MD simulations of 300 ns revealed (i) the tight hairpin structure of the E2 loop of the MT₁ receptor (ii) the most suitable features for melatonin binding in MT₁ receptors and (iii) major predicted rearrangements upon MT₁ receptor activation, stabilizing interaction networks between Phe179 or Gln181 in the E2 loop and transmembrane helixes 5 and 6. Functional assays confirmed these predictions, because reciprocal replacement of MT₁ and GPR50 residues/domains led to the predicted loss‐ and gain‐of‐melatonin action of MT₁ receptors and GPR50 respectively.

Conclusions and Implications

Our work demonstrated the crucial role of the E2 loop for MT₁ receptor and GPR50 function by proposing a model in which the E2 loop is important in stabilizing active MT₁ receptor conformations and by showing how evolutionary processes appear to have selected for modifications in the E2 loop in order to make GPR50 unresponsive to melatonin.

Linked Articles

This article is part of a themed section on Recent Developments in Research of Melatonin and its Potential Therapeutic Applications. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v175.16/issuetoc

Abbreviations

DPPC

dipalmitoylphosphatidylcholine

E2

second extracellular

ER

extracellular region

I2(3)

second (third) intracellular loop

MD

molecular dynamics

RMSD

root mean square deviation

TM

transmembrane helices

Introduction

The functional relationship between the three members of the human http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=39 sub‐family, http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=287 and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=288 and the orphan http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=107, also called melatonin‐related receptor, is not well understood (Jockers et al., 2016). Whereas MT₁ and MT₂ receptors bind melatonin with sub‐nanomolar affinity and are the target of several clinically approved drugs (Liu et al., 2016), GPR50 does not bind http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=224 despite its 50% sequence identity with MT₁ and MT₂ receptors. Intriguingly, GPR50 is only found in mammals, and synteny analysis indicates that GPR50 is the mammalian orthologue of Mel1c, a melatonin binding receptor found in lower vertebrates (Dufourny et al., 2008). This prompted Dufourny et al. to postulate that GPR50 lost its capacity to bind to melatonin during evolution, but the molecular basis for this remains unknown. Despite this loss‐of‐function phenotype, GPR50 is evolutionarily conserved and sequence variants have been associated with mental disorders (Thomson et al., 2005) and altered lipid metabolism in humans (Bhattacharyya et al., 2006). In addition, GPR50 seems to have evolved towards a regulatory function, as GPR50 inhibits MT₁ receptor function through heteromerization and disruption of MT₁ receptor signalling by the large C‐terminal region of GPR50 (Levoye et al., 2006). The physiological importance of GPR50 is further demonstrated by the metabolic phenotype of GPR50 ^−/− mice (Ivanova et al., 2008; Bechtold et al., 2012).

At a structural level, previous investigations had shown that chimeric receptors in which I2‐TM4‐E2 or I3‐TM6 regions of MT₁ receptors were replaced by the equivalent GPR50 sequences abolished http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1344 (2‐[¹²⁵I]MLT) binding and subsequent cAMP signalling. In return, chimeras in which the I2‐TM4‐E2‐I3‐TM6 region of MT₁ receptors was exchanged by its GPR50 counterpart (GPR50[I2‐TM4‐E2‐I3‐TM6‐MT₁]) showed partial restoration of 2‐[¹²⁵I]MLT binding and cAMP signalling (Conway et al., 2000). Within TM6 of the MT₁ receptor, Gly258 was identified as a critical residue, because replacement by Thr, the corresponding residue in GPR50, severely affected high‐affinity 2‐[¹²⁵I]MLT binding and cAMP signalling (Conway et al., 2000; Gubitz and Reppert, 2000). However, these initial studies did not consider the potential participation of the flanking loops introduced into the GPR50[I2‐TM4‐E2‐I3‐TM6‐MT₁] chimera in the restoration of melatonin function in GPR50. Based on recent crystal structures, the second extracellular (E2) loop is of particular interest as this loop has been shown to be of functional importance for class A GPCRs (Woolley and Conner, 2016). The E2 loop was found to control access of ligands towards the orthosteric binding site [http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=317&familyId=50&familyType=GPCR(Granier et al., 2012), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=318&familyId=50&familyType=GPCR(Wu et al., 2012), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=319&familyId=50&familyType=GPCR(Manglik et al., 2012) opioid and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=71&familyId=14&familyType=GPCR (Wu et al., 2010), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=62&familyId=14&familyType=GPCR (Tan et al., 2013) chemokine receptors], be part of the orthosteric binding site [http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=262&familyId=33&familyType=GPCR (Shimamura et al., 2011) and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=219&familyId=21&familyType=GPCR (Shihoya et al., 2016) receptors], be the determinant of subtype selectivity [http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2&familyId=1&familyType=GPCR (Wang et al., 2013), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=7&familyId=1&familyType=GPCR (Wacker et al., 2017), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=8&familyId=1&familyType=GPCR (Renault et al., 2010) and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=18&familyId=3&familyType=GPCR/http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=19&familyId=3&familyType=GPCR adenosine (Glukhova et al., 2017) receptors] and be the main part of an allosteric binding site in the A_1A adenosine (Glukhova et al., 2017) and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=14&familyId=2&familyType=GPCR (Kruse et al., 2013). From a functional point of view, the E2 loop was shown to be related to the biased agonism of some ligands through cAMP, ERK1/2 or Ca²⁺ signalling pathways (http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=28&familyId=4&familyType=GPCR (Shukla et al., 2014), protease‐activated receptor1 (Soto et al., 2015), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=13&familyId=2&familyType=GPCR (Keov et al., 2014) and M₂ (Gregory et al., 2010) muscarinic receptors) and contributes to constitutive activity [http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=265&familyId=33&familyType=GPCR (Wifling et al., 2015), http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=32&familyId=5&familyType=GPCR (Klco et al., 2005) and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=282&familyId=38&familyType=GPCR/http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=283&familyId=38&familyType=GPCR receptors (Holst and Schwartz, 2003)].

Here, we have investigated, specifically, the role of the E2 loop in MT₁ receptors and GPR50, by exploring the functional behaviour of MT₁‐GPR50 chimeric receptors in terms of ligand binding and signal transduction, assisted by computational structural receptor models.

Methods

Cell culture and transfection

Human embryonic kidney 293T (HEK293) cells were cultured in DMEM medium supplemented with 10% fetal bovine serum and penicillin/streptomycin. Cells were transfected with polyethylenimine (PEI) (Polysciences Inc.; Niles, IL, USA), Xtremegene 9 (Roche Diagnostics, Basel, Switzerland) or by electroporation (GenePulserXcell, Bio Rad; Hercules, CA, USA) according to the supplier's instructions.

DNA constructs

The pcDNA3‐CMV vectors expressing the human wild type MT₁ receptor and GPR50 were described previously (Levoye et al., 2006; Maurice et al., 2010). All MT₁ and GPR50 chimera have been generated by chemical synthesis and verified by sequencing. The different domains were defined as follows: N‐terminal of GPR50 (Met1‐Pro25), C‐terminal of MT₁ receptors (Ser312‐Val350), TM6 of MT₁ (Asp233^6.30‐Ala262^6.59) and GPR50 (Glu232^6.30‐Ala261^6.59), E2 of MT₁ (Arg164^E2a(C‐13)‐Ser184^E2b(C + 7)), GPR50 (Tyr165^E2a(C‐13)‐Asn185^E2b(C + 7)), E2b of MT₁ (Ile174^E2b(C‐3)‐Ser184^E2b(C + 7)) and GPR50 (Thr175^E2b(C‐3)‐Asn185^E2b(C + 7)). Domains corresponding to the E2 loop of M₂ (Gly167^E2a(C‐9)‐Ala184^E2b(C + 8)) and M₃ (Val210^E2a(C‐10)‐Glu227^E2b(C + 7)) muscarinic acetylcholine receptors have also been synthesized, as they show size and hydrophobicity properties, similar to those of the E2 loop of MT₁.

Radioligand binding experiments

Membranes from HEK293T cells transiently expressing the different receptor constructs were prepared as previously described (Ayoub et al., 2004). 2‐[¹²⁵I]iodomelatonin saturation binding experiments were performed in the range of 1–1000 pM, and specific binding was defined as binding displaced by 10 μM http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1343. Assays were carried out in duplicates for 120 min at 37°C, followed by rapid filtration through GF/F glass fibre filters (Whatman; Clifton, NJ, USA). Filter‐retained radioactivity was determined with a γ‐counter (LB2111; Berthold Technologies, Bad Wildbad, Germany).

Homogeneous time‐resolved fluorescence (HTRF)‐based cAMP assay

cAMP levels were determined in HEK293 cells transiently expressing wild‐type MT₁, GPR50 or chimeric receptors by HTRF using the ‘cAMP femto2’ kit (Cisbio, Bagnols‐sur‐Cèze, France) according to the manufacturer's instructions. Cells were incubated with 2 μM forskolin and different concentrations of melatonin at room temperature for 30 min. Data were fitted by non‐linear regression to determine E_max and EC₅₀ values and normalized to the forskolin‐induced response (as 100%) and melatonin‐induced maximum inhibition (as 0%) using GraphPad Prism software.

BRET‐based β‐arrestin2 recruitment assay

For the ß‐arrestin2 recruitment assays, cells were transfected with plasmids coding for different receptors and the Rluc8‐ß‐arrestin2‐YPet biosensor. Forty‐eight hours post‐transfection cells were washed once with PBS, and Coelenterazine h substrate was added at a final concentration of 5 μM and left for 10 min at room temperature. Cells were then stimulated by different concentrations of melatonin for 5 min (Kamal et al., 2009). BRET readings were made, using a lumino/fluorometer (Mithras; Berthold Technologies, France) allowing sequential integration of luminescence with three filter settings (Rluc filter, 485 ± 10 nm; YFP (and YPet) filter, 530 ± 12.5 nm, RGFP filter, 515 ± 10 nm). Emission signals at 530 or 515 nm were divided by emission signals at 485 nm. The difference between this emission ratio obtained with donor and acceptor fused or co‐expressed and that obtained with the donor protein expressed alone was defined as the BRET ratio. Data were fitted by non‐linear regression to determine E_max and EC₅₀ values using GraphPad Prism software.

SDS‐PAGE/Western blot

Receptor expression in transfected cells was assessed by SDS‐PAGE/Western blot assays, as previously described (Cecon et al., 2015). Briefly, cells were collected into Laemmli lysis buffer, and denatured proteins were resolved in 10% SDS‐PAGE gels, immune‐blotted with antibody against the FLAG tag (present in all constructs) (1:1000 dilution; Sigma‐Aldrich) and revealed using the Odyssey LI‐COR infrared fluorescent system (LICOR Biosciences, Lincoln, NE, USA). Blots were quantified using the ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Homology modelling of MT1, mutMT1 receptors and GPR50

The sequences of MT₁, hGPR50, bovine rhodopsin, turkey β₁‐adrenoceptor, http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=29&familyId=4&familyType=GPCR and hA_2A receptors corresponding to P48039, Q13585, P02699, P07700, P07550 and P29274 Uniprot IDs, respectively, were aligned using ClustalΩ (Sievers et al., 2011). Only sequences corresponding to crystal structures of bovine rhodopsin (1U19 pdb entry) (Okada et al., 2004), turkey β₁‐adrenoceptor (2VT4) (Warne et al., 2008), hβ₂‐adrenoceptor (2RH1) (Cherezov et al., 2007) and A_2A receptor (A2AAR) (3EML) (Jaakola et al., 2008) were aligned. Based on the size of E2, I3 and E3 variable regions, rhodopsin was the closest template. That is why we used a full‐sequence alignment with rhodopsin, adrenoceptors and adenosine receptors only in conserved regions (I1, I2, E1 and TM domains), whereas the coordinates of variable E2, I3 and E3 regions were inherited only from the rhodopsin template. This initial alignment was manually refined using Chimera (Pettersen et al., 2004) to adjust some of the gaps in the E2, I3 and E3 loop regions and align conserved cysteine residues (Cys100^3.25, Cys177 of MT₁, Cys178 of GPR50) (Supporting Information Figure S1). Apo‐receptor structures for MT₁, mutMT₁ (chimeric receptor containing the last 11 C‐terminal amino acids) (Ile174^E2b(C‐3)–Ser184^E2b(C + 7)) of the E2 of the MT₁ receptor replaced by the corresponding region of GPR50 (MT₁[E2b‐GPR50]) and GPR50 systems were modelled with Modeller program (Webb and Sali, 2014). Disulfide bridges were added between Cys3.25 (Cys100^MT1R and Cys101^GPR50) and central cysteine residue of E2 (Cys177^MT1R and Cys178^GPR50). All models were subjected to thorough molecular dynamics (MD) optimization. The five best‐scoring models were inspected visually, mapped to Ramachandran diagrams and the most suitable model of each complex was selected in terms of low score and structure of the loops. Melatonin‐bound states of MT₁, mutMT₁ receptors and GPR50 resulted from the best docking pose of melatonin, by GoldScore scoring function, into a sphere described by a radius of 14 Å around the centroid of the only solvent accessible surface under the extracellular region, using GOLD suite v5.2 (Jones et al., 1995) within the Hermes v1.6 GUI (CCDC©). The goal was to get suitable complex as inputs for molecular dynamics which was the core method to reach very low energy states of each holo‐receptor.

MD simulations

Protein models were embedded into a triclinic lipid bilayer composed of 368 dipalmitoylphosphatidylcholine (DPPC) molecules and was hydrated into a layer of approximately 30 Å on each side. After addition of chloride ions to achieve charge neutrality, the final system contains a total of roughly 77 000 atoms. In a second step, the inflateGRO method (Kandt et al., 2007) was used to optimize the packing of lipids around the protein. This protocol aims to inflate and then iteratively increase, the density of lipids up to an optimal value of area per lipid observed on to an equilibrated DPPC membrane (Nagle, 1993). The system was then equilibrated restraining the α‐carbon trace in the NVT (constant N particles, volume and temperature ensemble) then NPT (constant N particles, pressure and temperature ensemble) ensemble before unrestrained MD simulations (3 × 300 ns) in the NPT ensemble, using Langevin dynamics to control temperature at 323 K, and at a time step of 2 fs, while constraining all bonds between hydrogen and heavy atoms. MD simulations were carried out with GROMACS 5.0 (Pronk et al., 2013) using the c36 CHARMM forcefield (Best et al., 2012). The GROMACS topology of melatonin was generated from PRODRG webserver (Schüttelkopf and van Aalten, 2004). Analysis of trajectories was made by GROMACS routines [calculation of root mean square deviation (RMSD) and intermolecular areas, detection of intermolecular hydrogen bonds and interatomic contacts], and their visualization was performed with visual molecular dynamics (Humphrey et al., 1996).

Materials

Melatonin was purchased from Sigma‐Aldrich (St. Louis, MO, USA) and 2‐[¹²⁵I]iodomelatonin from PerkinElmer (Waltham, MA, USA).

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Southan et al., 2016), and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 (Alexander et al., 2015).

Results

Essential role of E2 loop for MT₁ receptor function

To study the role of the E2 loop in MT₁ receptor function, we generated a series of chimeric receptors between the melatonin binding‐competent MT₁ receptor and the melatonin binding incompetent GPR50 and studied their functional properties in HEK293 cells (see Figure 1A for the sequence alignment of E2 loops for GPCRs referred to in this study). Both wild‐type receptors showed the expected properties, MT₁ receptors bound 2‐[¹²⁵I]MLT with high affinity (K_d 233 ± 20 pM), activated the G_i/cAMP pathway and recruited ß‐arrestin2, and GPR50 was completely inactive in all assays (Figure 1B; Table 1). The chimeric receptor containing the predicted E2 loop (residues 164–185) of the MT₁ receptor replaced by the entire corresponding region of GPR50 (MT₁[E2‐GPR50]) or only the C‐terminal 11 amino acids (MT₁[E2b‐GPR50], see below) was inactive in all three assays (Figure 1B; Table 1). Replacement of the N‐terminal domain of the MT₁ receptor by its GPR50 counterpart (MT₁[Nter‐GPR50]) had only a modest effect on its function (Table 1; Supporting Information Figure S2A) supporting the specificity of the effect observed with the two E2 loop chimeras. Expression of equivalent levels of MT₁ wild type, MT₁[E2‐GPR50], MT₁[E2b‐GPR50] and MT₁[Nter‐GPR50] chimera is shown by Western blotting in cell lysates (Figure 1C, D).

Figure 1.

Pharmacological investigations of wild‐type and chimeric MT₁/GPR50 receptors. (A) Sequence alignment of E2 loops of GPCRs referred to in this study. Sequence identity appears in shades of blue. (B) Melatonin‐induced activation of the G_i/cAMP pathway and recruitment of ß‐arrestin2 (see also Table 1 and Supporting Information Figure S2D). (C) Representative Western blot detection of receptor expression levels in cell lysates. (D) Quantification of receptor expression. Data are expressed as mean ± SEM of at least five independent experiments (n is shown in the bars for each group). (E) Melatonin‐induced activation of the G_i/cAMP pathway in mutated MT₁ and MT₁/GPR50 chimera (see also Table 2 and Supporting Information Figure S2E).

Table 1.

In vitro activity of chimeric MT₁/GPR50 mutants

Receptor structure	2‐[125I]MLT		Effect on cAMP production			ß‐arrestin recruitment
	Kd (pM)	n	pEC50	Maximal effect (% of Fsk‐induced response)	n	pEC50	Maximal effect (% of MT1 response)	n
MT1	233 ± 20	39	9.88 ± 0.20	70.9 ± 1.9	6	8.66 ± 0.17	100.0 ± 2.9	6
MT1[E2‐GPR50]	No binding	5	No effect	–	5	No activation	–	7
MT1[E2b‐GPR50]	No binding	5	No effect	–	3	No activation	–	6
MT1[Nter‐GPR50]	230 ± 44	3	8.56 ± 0.20	78.5 ± 1.9	6	9.82 ± 0.53	60.4 ± 7.0*	3
GPR50	No binding	5	No effect	–	3	No activation	–	10
GPR50[E2‐MT1]	No binding	6	9.50 ± 0.29	81.2 ± 2.0	6	8.00 ± 0.16	109.8 ± 6.3	7
GPR50[E2‐M2]	No binding	3	No effect	–	6	–	–	–
GPR50[E2‐M3]	No binding	3	No effect	–	5	–	–	–
GPR50[TM6‐MT1]	No binding	7	10.03 ± 0.44	87.7 ± 2.1*	4	8.56 ± 0.33	72.5 ± 7.5*	7
GPR50[E2‐, TM6‐MT1]	228 ± 18	5	9.66 ± 0.27	82.3 ± 1.6*	5	8.74 ± 0.20	103.8 ± 6.2	10
GPR50[E2‐, TM6‐, Cter‐MT1]	96 ± 15	3	9.39 ± 0.49	84.1 ± 2.6	5	8.91 ± 0.17	91.0 ± 5.9	5

Concentration–response curves were analysed by non‐linear regression. Binding affinity for 2‐[¹²⁵I]‐MLT was measured and is expressed as mean K_d ± SEM (pM). The effect of melatonin on cAMP and β‐arrestin pathways is expressed as pEC₅₀ ± SEM (M), while the maximal effect is expressed as a percentage of the maximal effect induced by forskolin (Fsk) or by melatonin on MT₁ receptor (=100%, β‐arrestin). Data are mean of at least three independent experiments; each of them performed using at least eight different ligand concentrations.

^* P < 0.05, significantly different from MT₁; two‐tailed Student's t‐test.

Homology modelling of the human MT₁ receptor in its inactive form indicated a tight hairpin structure of the MT₁ E2 loop and a separation into two distinct parts of the hairpin, an N‐terminal E2a part composed of 10 amino acids (Arg164^E2a(C‐13)–Arg173^E2a(C‐4)) and a C‐terminal E2b part composed of 11 amino acids (Ile174^E2b(C‐3)–Ser184^E2b(C + 7)) labelled relative to the reference Cys177 position (Figure 2). Based on the Fredriksson's phylogenetic tree (Fredriksson et al., 2003), rhodopsin, β₁‐adrenoceptor, β₂‐adrenoceptor and A_2A adenosine receptor crystal structure templates turned out to be the closest homologous templates of the three MT₁, mutMT₁ (MT₁[E2b‐GPR50]) and GPR50 sequences, particularly within the α‐class of rhodopsin‐like family (Supporting Information Figure S1). For the modelling of the E2 loop, rhodopsin represented the most suitable template. In contrast to the α‐helical organization of the E2b part in β‐adrenoceptors or the additional 3‐residues deletion in the E2a part of the A_2A receptor, the extended conformation of rhodopsin E2b facilitates the location of the 5‐residue deletion in the coil of the hairpin keeping the spatial constraint of the conserved Cys177 for disulfide bridge formation with Cys100^3.26 without any impairing constraint for TM4‐E2 and E2‐TM5 junctions (Supporting Information Figure S1). In addition to this shortened length of E2b, the spatial constraint due to the predicted disulfide bridge between Cys177 in E2b and Cys100^3.26 contributes to a tight hairpin structure of the MT₁ E2 loop. A similar structural organization of the E2 loop into two structurally distinct elements has been previously proposed for other GPCRs (Nygaard et al., 2009). The three selected MT₁, mutMT₁ and GPR50 homology models were very close to each other with a RMSD lower than 1 Å from the α‐carbon pairwise alignment of 283 homologous amino acids (Supporting Information Figure S1). Nevertheless, at this level, before MD simulations, we observed a first slight difference in the folding of E2b between MT₁, mutMT₁ and GPR50 apo‐models. Replacement of the E2b element of the MT₁ receptor by its GPR50 counterpart (MT₁[E2b‐GPR50]) abolished receptor function, thus experimentally validating the importance of the E2b element (Table 1, Figure 1B). Inspection of the sequence of the E2b element revealed five residues that are homologous between MT₁ and MT₂ receptors but differ in the GPR50 sequence (Ile/Thr^E2b(C‐3), Thr/Ile^E2b(C + 1), Gln/Tyr^E2b(C + 4), Ser/Asn^E2b(C + 7) and Ser/Asn^E2b(C + 7)) (Figure 1A). Replacement of Ile174^E2b(C‐3) and Thr178^E2b(C + 1) by their GPR50 counterparts (MT₁[I174T, T178I]) did not impair MT₁ receptor function, whereas replacement of Gln181^E2b(C + 4), Ser182^E2b(C + 5) and Ser184^E2b(C + 7) (MT₁[Q181Y, S182L, S184N]) severely diminished its function (Figure 1E, Table 2, Supporting Information Figure S2D). Similar results were obtained with single or double mutants (MT₁[Q181Y], MT₁[S182L,S184N], MT₁[S184N]) of these residues (Figure 1E, Table 2 and Supporting Information Figure S2D). To further evaluate the effects of Gln181^E2b(C + 4), Ser182^E2b(C + 5) and Ser184^E2b(C + 7) residues on MT₁ receptor function, we placed them back in the MT₁[E2b‐GPR50] chimera and monitored for restoration of function in this loss‐of‐function background (MT₁[E2b‐GPR50]Y181Q, L183S,N185S, N185S). Interestingly, responses to melatonin were restored in Y181Q and L183S,N185S mutants (Figure 1E, Table 2 and Supporting Information Figure S2D). Collectively, our results show the importance of the E2b element of the MT₁ receptor and particularly the C‐terminal region of the E2b element close to TM5 for 2‐[¹²⁵I]MLT binding and MT₁ receptor activation.

Figure 2.

Essential residues for the binding of melatonin from MD simulations. (A) The distributions show the percentage of the 3000 frames, covering 300 ns MD simulations that form hydrogen bonds or hydrophobic contacts between melatonin and each residue detected at least once in contact, using MT₁, mutMT₁ or GPR50 receptors. The frequency of each intermolecular receptor‐melatonin interaction was deduced from triplicates for MT₁‐MLT and GPR50‐MLT and from duplicates for mutMT₁‐MLT (please see Figure S3 for more details). For comparison, residues are identified with Ballesteros numbering with the exception for E2 residues, identified by their relative position to the conserved cysteine residue. (B) The most representative illustration of intermolecular interactions with melatonin (MLT) in MT₁ (blue), mutMT₁ (green) and GPR50 (red) holo‐receptor models are represented from a frame chosen at a time displaying most of highlighted residues in some lateral and extracellular views on the left and the right panels respectively. E2a and E2b segments of E2 loop are illustrated as yellow and red tubes respectively.

Table 2.

Consequences of amino acid substitutions in E2 loop of MT₁ and mutMT₁ on receptor activity

	Receptor structure	2‐[125I]MLT		Effect on cAMP production
		Kd (pM)	n	pEC50	Maximal effect (% of Fsk‐induced response)	n
	MT1	233 ± 20	39	9.88 ± 0.20	70.9 + 1.9	6
	MT1[Q181Y, S182L, S184N]	510 ± 74	7	No effect	–	4
	MT1[Q181Y]	640 ± 71	13	10.06 ± 0.47	90.9 ± 1.4*	5
	MT1[S184N]	No saturation	7	No effect	–	4
	MT1[Q181Y, S184N]	536 ± 51	5	9.91 ± 0.42	90.6 + 1.3*	4
	MT1[S182L, S184N]	602 ± 41	5	8.31 ± 0.53	93.5 ± 1.6*	7
	MT1[I174T, T178I]	60 ± 15	4	9.90 ± 0.18	74.7 ± 1.5	7
	MT1[E2b‐GPR50]Y182Q	196 ± 13	5	9.35 ± 0.36	84.7 ± 1.8*	5
	MT1[E2b‐GPR50]L183S, N185S	183 ± 15	6	8.00 ± 0.30	84.1 ± 2.5*	6
	MT1[E2b‐GPR50]N185S	No binding	5	No effect	–	3

Concentration–response curves were analysed by non‐linear regression. Binding affinity for 2‐[¹²⁵I]‐MLT was measured and is expressed as mean K_d ± SEM (pM). The effect of melatonin on cAMP pathway is expressed as pEC₅₀ ± SEM (M), while the maximal effect is expressed as a percentage of the maximal effect induced by forskolin (Fsk). Data are mean of at least four independent experiments; each of them performed using at least eight different ligand concentrations.

^* P < 0.05, significantly different from MT₁ ; two‐tailed Student's t‐test.

Exchange of the E2 loop of GPR50 for its MT₁ counterpart is sufficient to promote melatonin‐induced activation of GPR50

To better understand the role of the E2 loop in MT₁ receptor function, we introduced the latter into GPR50 (GPR50[E2‐MT₁]) in an attempt to induce sensitivity to melatonin in GPR50. This chimera was indeed fully functional in terms of coupling to the G_i/cAMP pathway and β‐arrestin2 recruitment (Figure 1B, Table 1 and Supporting Information Figure S2D). High‐affinity 2‐[¹²⁵I]MLT binding was not measurable, porbably due to the lower sensitivity of the binding assay as compared to the functional assays, which are based on signal amplification. Replacement of the E2 loop of GPR50 by the E2 loop of the muscarinic M₂ or M₃ receptors (GPR50[E2‐M₂] and GPR50[E2‐M₃]), two receptors with E2 loops with similar length, hydrophobicity and positioning of the Cys residue, compared with the MT₁ receptor, did not induce melatonin responsiveness (Figure 1A, Table 1 and Supporting Information Figure S2C, D). This confirms that the specific and positive effect of the E2 loop of the MT₁ receptor can be transferred to GPR50 and excludes a possible inhibitory effect of the E2 loop of GPR50 on GPR50 function.

Putative role of the E2 loop into the orthosteric binding site of melatonin

3D homology models of free and melatonin‐bound MT₁, mutMT₁ (MT₁‐[E2b‐GPR50]) receptors and GPR50 were built, refined and analysed in order to identify differential molecular features of the E2 loops that could explain such contrasting pharmacological profiles. The first hypothesis to be investigated was the potential participation of the E2 loop in the direct binding of melatonin. These investigations have been based on the putative orthosteric binding sites that are supposed to be the most voluminous solvent accessible cavity close to the extracellular region. All three receptor models revealed the presence of a comparable cavity. Three prototypical holoreceptor models (MT₁, mutMT₁ and GPR50) were built resulting from the top‐scored docking pose of the most populated binding mode. Specific interactions between receptors and melatonin were then studied after triplicated 300 ns MD simulations of energy relaxation. Only one of the nine simulations, concerning the mutMT₁ receptor system, was not taken into account as it showed the migration of melatonin into the lipid bilayer. Monitoring the displacement of melatonin during MD simulations showed its rapid stabilization in holo‐MT₁ and holo‐mutMT₁. On the contrary, although the fluctuations of melatonin appear to be flat by the end of the simulations because of the holo‐mutMT1R0 artefact, a significant instability of melatonin still remained until 300 ns in holo‐GPR50 (Supporting Information Figure S3). This finding would argue for more stable MT₁‐melatonin and mutMT₁‐melatonin, than GPR50‐melatonin, complexes. From a thermodynamic point of view, we aimed at linking the binding affinity of melatonin in all holo‐receptors to the frequency of detected intermolecular interactions during MD simulations. Although there are as many residues involved in both hydrogen bonding and hydrophobic interactions in both MT₁ and GPR50 receptors, interactions are significantly more frequent in MT₁ receptors than in GPR50 (Figure 2A). Melatonin fits between TM 3, 4, 5 and 6 under the E2 loop. Consistent with the aromatic nature of the indole core of melatonin, most of the interactions detected were hydrophobic. The three most frequent intermolecular interactions observed between melatonin and the MT₁ receptor involved residues Val111^3.36 (35%), Tyr170^E2a(C‐7) (33%) and Phe196^5.47 (30%) (Figure 2B), whereas the relative frequencies of these amino acids in mutMT₁ and GPR50 were lower than 22% and 17% respectively. To a lesser extent, Tyr187^5.39 and Phe256^6.53 complete the hydrophobic binding pocket, whereas Asn255^6.52 makes the only significant polar interaction with the amide extremity of melatonin in the MT₁ receptor, but not in mutMT₁ or GPR50 (Figure 2B). His195^5.461 has already been identified as part of the melatonin binding pocket of MT₁ (Kokkola et al., 2003) and MT₂ receptors (Conway et al., 1997) and, more generally, to be important for ligand stabilization or receptor activation of GPCRs (Almaula et al., 1996). Collectively, 3D homology modelling suggested that the E2 loop forms an extracellular cap over the melatonin binding pocket within the TM bundle but is unlikely to participate directly in melatonin binding to MT₁ receptors.

Flexibility and folding of the E2 loop in MT₁, mutMT₁ and GPR50 models

Based on the weak fluctuation of the E2 loop observed during MD simulations, we anticipate that this region undergoes a conformational restriction as a hairpin tensor, a feature that might be different between the three receptor models (Figure 3). In contrast to the impaired mutMT₁ and GPR50 models, most of the E2b residues in the MT₁ model are folded in regular units which are conserved during MD simulation of apo/holo‐MT₁ models (Supporting Information Document S4). In comparison with apo‐GPR50 simulation, there are significantly less regular folded units in holo‐GPR50. Binding of melatonin in holo‐GPR50 seems thus to alter the E2 structure rather than stabilize it, as in the holo‐MT₁ simulation. All these findings argue in favour of a specific folding of the E2 loop which is stabilized in functional MT₁ receptors, but not in impaired apo‐mutMT₁ or GPR50 systems. Nevertheless, although the E2b elements are structurally rigid during simulations of mutMT₁ and GPR50 systems, the global folding of their E2 loop seems clearly different. The E2a element of apo‐GPR50 tolerates an additional flexibility, due to the unstable flanking TM4, which thus shifts major folding units as ‘bend’ features towards this N‐terminal region of E2. On the other hand, apo‐mutMT₁ exhibits a rigid E2a element due to a higher stability of the flanking TM4, whereas the exchange of the MT₁ E2b for its GPR50 counterpart induces a significant flexibility. This could be considered as an anomalous local instability, probably due to an addition of 20 atoms within this packed folding of the E2b element, which induce a spatial restraint forcing a singular ‘bend’ folding unit from 220 ns until the end of the apo‐mutMT₁ MD simulation. All together, these data show that the E2 loop exhibits a common hairpin folding, however with important individual differences for each receptor model studied. This loop is rigid and stable in MT₁ receptors and GPR50 but much more regularly folded in MT₁ receptors, whereas the E2b element is unstable in mutMT₁ receptors. Although the E2 loop seems to be positioned on top of the melatonin binding pocket, it does not seem to be flexible enough to act as a dynamic gate to regulate melatonin accessibility to the binding pocket, by itself.

Figure 3.

Structural folding of the E2 loop in MT₁, mutMT₁ and GPR50 receptors. 3D folding was appreciated during MD simulations in term of root mean square fluctuation expressed by scaled B‐factor in nm² (A) and by the assignment of secondary structures (B) by the DSSP method (Kabsch and Sander, 1983) focusing onto the studied E2 loop and its flanking regions (140–220). A detailed analysis of the E2 loop folding is available in Supporting Information Document S4.

Interaction of the E2 loop with TM domains in MT₁, mutMT₁ and GPR50 models

We then evaluated whether differential interactions of the E2 loop with the TM domain or other extracellular regions (ERs) could explain the difference between active (MT₁) and inactive (mutMT₁ and GPR50) receptor models (Figure 4). Most intramolecular interactions of the E2 loop detected during MD simulations were apolar contacts. The E2 loop of the functionally inactive apo‐mutMT₁ and apo‐GPR50 preferentially interacts with ER (Figure 4A), whereas the E2 loop of the functionally active apo‐ and holo‐MT₁ preferentially interacts with the TM domain (Figure 4B). These hydrogen bonds exist particularly between Gln181^E2b(C + 4) (Figure 4D) and Thr188^5.40, observed in 40% of all conformations sampled from holo‐MT₁ MD simulations (Figure 5A). Experimental data showed that mutation of Gln181 into its GPR50 Tyr counterpart (MT₁[Q181Y]) had only a minimal effect on melatonin binding and receptor signalling (Figure 1E, Table 2 and Supporting Information Figure S2E). Conversely, introduction of a single Tyr181 residue in the functionally inactive MT₁[E2b‐GPR50] chimera was able to restore function to the MT₁[E2b‐GPR50]‐Y182Q chimera (Figure 1E, Table 2 and Supporting Information Figure S2E) supporting its important role in MT₁ receptor function. During MD simulation of the MT₁ model, we also detected significant interactions between TM6 and the E2 loop. In terms of polar interactions, one hydrogen bond was found between Tyr170^E2a(C‐7) and Asn255^6.52 in 25% of the sampled conformations. Although not observed simultaneously, Asn255^6.52 may also interact with melatonin (see above). Four additional apolar contacts were observed between Phe179^E2b(C + 2) (Figure 4D) and Ile257^6.54 (60%) with alternative interactions with neighbouring Leu254^6.51, Gly258^6.55 and Val261^6.58. The importance of Gly258^6.55 for MT₁ receptor function has been previously shown in 2‐[¹²⁵I]MLT binding and cAMP inhibition experiments (Conway et al., 2000; Gubitz and Reppert, 2000). Another favourable aromatic contact was frequently identified between Phe179^E2b(C + 2) and Tyr281^7.39 (residue conserved in GPR50) in the holo‐MT₁ model.

Figure 4.

Numbers of intramolecular interactions involving the E2 loop. The most conserved interactions were identified if their frequency was greater than 50% for apolar contacts or 20% for hydrogen bonds during MD simulations. The illustration focuses on nature of apolar or polar interactions from E2a or E2b moieties with ER (A) or TM (B). In (C) and (D), the individual residues of the E2 loop in each receptor model participating in the interactions with ER are shown in (C) or with TM in (D).

Figure 5.

Clusters of intramolecular interactions involving the E2 loop. (A) holo‐MT₁, (B) apo‐mutMT₁ and (C) apo‐GPR50. Cylinders represent TM domains, whereas loops are represented by ribbons. Peptidic backbones of E2a and E2b are illustrated in yellow and red respectively. In each illustration, domains interacting with the E2 loop appear as opaque, whereas the remaining structure is transparent.

In the two inactive models (mutMT₁ and GPR50) the E2 loop showed more interactions with the ER (Figures 4C and 5B, C), in particular van der Waals or aromatic interactions are observed between the E2b part (Tyr175/176^C‐2, Ile178/179^C + 1, Phe179/180^C + 2 and Tyr181/182^C + 4) and the N‐terminal region (Met1, Gly5, Leu8, Leu16, Arg17 and Arg22 in the apo‐mutMT₁ model and Met1, Gly2, Pro3, Val7, Pro8, Pro10, Tyr11, Gly12, Ile14, Gly15, Leu18, Pro19 and Tyr23 in the apo‐GPR50). The same type of interactions was also observed between the E3 loop (Pro264 and Lys265 of GPR50; Pro265 of mutMT₁) and Phe179/180^E2b(C + 2) or Tyr181/182^E2b(C + 4). Consequently, although Phe179^E2b(C + 2) is conserved between MT₁ and GPR50 receptors, its interactions are different. Its central position in the E2 loop makes the orientation of its side chain dependent on the E2b sequence and folding. As a result, our data converge towards a differential network of interactions established by E2b loop positions C + 2 and C + 4 in functional, as distinct from impaired, systems. In the inactive mutMT₁ and GPR50 systems, Phe179/180^E2b(C + 2) and Tyr181/182^E2b(C + 4) of E2b are predicted to favour a large amount of hydrophobic interactions with the N‐terminus and E3 loop of the ER, whereas the analogous residues in the functional holo‐MT₁ system (Phe179^E2b(C + 2) and Gln181^E2b(C + 4)) are engaged in more buried interactions with TM5, TM6 and TM7 (Figure 5).

In vitro evidence for an E2b‐TM6 unit supporting MT₁ receptor activity

Molecular modelling indicated preferential interaction of the E2b element with TM5, TM6 and TM7 in the melatonin‐bound holo‐MT₁ model (Figure 5A). Replacement of TM5 and TM7 by its GPR50 counterpart (MT₁[TM5‐GPR50] and MT₁[TM7‐GPR50]) is known not to affect MT₁ receptor function (Conway et al., 2000; Gubitz and Reppert, 2000). This observation is compatible with our holo‐MT₁ model as Thr188^5.40 and Tyr281^7.39, the residues interacting respectively with Gln181^E2b(C + 4) and Phe179^E2b(C + 2) of the E2b element, are conserved between MT₁ receptors and GPR50. In contrast, a similar exchange of TM6 (MT₁[TM6‐GPR50]) generated an inactive receptor indicating its functional importance (Conway et al., 2000; Gubitz and Reppert, 2000). To verify the relevance of TM6 more directly, we transferred TM6 of the MT₁ receptor into the GPR50 backbone (GPR50[TM6‐MT₁]) to see whether we could make GPR50 responsive to melatonin, which was indeed the case for the G_i/cAMP pathway activation and ß‐arrestin 2 recruitment (Table 1 and Supporting Information Figure S2A, D). As found with the GPR50[E2‐MT₁] chimera, high‐affinity 2‐[¹²⁵I]MLT binding was not measurable. Chimera combining the E2 loop and TM6 of the MT₁ receptor further confirmed these results and also restored measurable 2‐[¹²⁵I]MLT binding (GPR50[E2‐, TM6‐MT₁], GPR50[E2‐, TM6‐, Cter‐MT₁]) (Table 1 and Supporting Information Figure S2A, D).

Based on our molecular modelling results, Gln181^E2b(C + 4) and Phe179^E2b(C + 2) in the E2 loop seemed interesting candidates to interact in the MT₁ receptor with TM5 (Thr188^5.40) and TM6 (Ile257^6.54) respectively. The absence of these interactions in models of the inactive MT₁[E2b‐GPR50] chimera is in agreement with this hypothesis. To further validate these predictions experimentally, we performed functional assays with MT₁ receptors carrying mutations in TM6. Gly258^6.55 has been previously identified to be important for MT₁ receptor function, in particular in combination with Ala252^6.49 (MT₁[G258T], MT₁[A252C, G258T]) (Conway et al., 2000; Gubitz and Reppert, 2000). Indeed, our own results confirmed a partial and complete loss of function for MT₁[G258T] and MT₁[A252C, G258T], respectively (Table 3 and Supporting Information Figure S2B). Loss of signalling function in the Q181Y/G258T double mutant confirmed the importance of G258 for MT₁ receptor function. The substitution I257L alone (MT₁[I257L]) showed 2.5 times lower affinity for 2‐[¹²⁵I]MLT binding and no signalling capacity (Table 3, Supporting Information Figure S2B). To further study the relationship of Gln181^E2b(C + 4) with Ile257^6.54 and Gly258^6.55 in TM6, we evaluated the potential of these three residues to impair the function of the Y182Q‐dependent E2b loop MT₁[E2b‐GPR50]Y182Q recovery mutant. Whereas MT₁ receptor function was completely lost in the G258T mutant (MT₁[E2b‐GPR50]Y182Q, G258T), it was not affected by introducing I257L (MT₁[E2b‐GPR50]Y182Q, I257L) further supporting the privileged relationship between Q181 in the E2b loop and G258 in TM6 (Table 3 and Supporting Information Figure S2B).

Table 3.

In vitro assay of effects of mono‐ or di‐substitutions in the E2 loop and TM6 on MT₁ and mutMT₁ function

	Receptor structure	2‐[125I]MLT		Effect on cAMP production
		Kd (pM)	n	pEC50	Maximal effect (% of Fsk‐induced response)	n
	MT1	233 ± 20	39	9.88 ± 0.20	70.9 + 1.9	6
	MT1[G258T]	215 ± 47	4	No effect	–	3
	MT1[A252C, G258T]	No binding	6	No effect	–	4
	MT1 [I257L]	690 ± 130	3	No effect	–	3
	MT1[Q181Y, G258T]	386 ± 55	4	No effect	–	3
	MT1[Q181Y, I257L]	290 ± 45	5	No effect	–	3
	MT1[E2b‐GPR50]Y182Q, G258T	No binding	3	No effect	–	3
	MT1[E2b‐GPR50]Y182Q, I257L	245 ± 30	4	8.93 + 0.35	80.2 ± 2.8	3

Concentration–response curves were analysed by non‐linear regression. Binding affinity for 2‐[¹²⁵I]‐MLT was measured and is expressed as mean K_d ± SEM (pM). The effect of melatonin on cAMP pathway is expressed as pEC₅₀ ± SEM (M), while the maximal effect is expressed as a percentage of the maximal effect induced by forskolin (Fsk). Data are mean of at least three independent experiments; each of them performed using at least eight different ligand concentrations.

Discussion

The MT₁ receptor and GPR50 constitute an interesting case as both receptors show high sequence homology yet are functionally very different, as MT₁ receptors bind melatonin with high affinity and GPR50 receptors appear to have lost this capacity in mammals. Here, we present evidence that differences in the primary and secondary structure of the E2 loop explains the ability of MT₁ receptors to bind melatonin and the inability of GPR50 to do so. According to MD simulations, the E2 loop of MT₁ receptors and GPR50 receptors tends to fold as a hairpin composed by two distinct strands that we identified as E2a and E2b, respectively, flanking TM4 and TM5. Nevertheless, there is a major difference in the structure of the E2 loop of MT₁ and GPR50 receptors. In contrast to GPR50, the E2 loop of MT₁ receptors exhibits more regular folding units like a β‐hairpin (antiparallel β‐sheet) or B‐bridge‐hairpin (a unique hydrogen bond). These hairpin structures form an extracellular cap over the melatonin binding pocket, which is located within the TM bundle. Intriguingly, GPR50 exhibits a putative ligand binding cavity similar to MT₁ receptors in terms of volume and location. Consequently, even though the interactions between GPR50 and melatonin are more scattered in this virtual model and occur with a lower frequency than in the holo‐MT₁ model, particularly because of an irregular folding of the TM4 region flanking the E2a element of the E2 loop, it is not enough to explain the total loss of melatonin binding.

Besides, we also observed that the E2 loop did not participate significantly in the orthosteric binding of melatonin with only one residue of E2a, Tyr170^E2a(C‐7), stacking its indole scaffold by aromatic contacts, but which is identical in GPR50. Although exchange of the E2b region between MT₁ and GPR50 receptors makes this region unstable during MD simulations, the chimeric mutMT₁ receptor also contains an appropriate binding site to fit melatonin. For all these reasons, we suspect that ligand accessibility controlled by the E2 loop is the major differential event promoting melatonin binding to MT₁ receptors and preventing access to the orthosteric binding pocket of GPR50. According to our model, the E2 loop of MT₁ receptors caps the melatonin binding pocket and thus decreases its accessibility (Figure 6). The known crystal structures of class‐A GPCRs with similar closed capping of the binding site by the extracellular region containing a β‐hairpin E2 loop are rhodopsin (Palczewski et al., 2000) and the http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=275&familyId=135&familyType=GPCR (Hanson et al., 2012). The hydrophobic nature of their ligands was suggested to force these ligands to access their binding site by diffusing through the membrane (Hurst et al., 2010). Biophysical studies have shown that melatonin distributes rapidly in the aqueous and lipid phase indicating that melatonin could approach the receptor from both phases (Shida et al., 1994; Costa et al., 1997; Yu et al., 2016). Performing several exploratory computational experiments of restrained MD simulations steering the melatonin off the binding site, we highlighted a putative accessibility pathway between extracellular extremities of TM5 and TM6 (Supporting Information Electronic Material S5). Interestingly, the E2b element is involved in this pathway by interactions between the amide group of melatonin and Gln181^E2b(C + 4) found to be essential in silico to form E2‐TM5 units in the holo‐MT₁ model but also experimentally in vitro to restore function to a chimeric mutMT₁ receptor, if Tyr181 is replaced by Gln (MT₁[E2b‐GPR50]‐Y182Q) (Figure 1E, Table 2 and Supporting Information Figure S2E). However, this computational method is so biased by high force constants that such a ligand accessibility pathway could also be found in GPR50, even though it would be theoretically impossible at a thermodynamic level. It only suggests that, combined with the demonstrated weak flexibility of the E2 loop, some structural features of the E2b element in GPR50 and the related chimeric mutMT₁ would be enough to disrupt such a putative accessibility pathway for melatonin. Intrinsically, the E2b element of GPR50 contains 20 atoms more than that of the MT₁ receptor. It represents a non‐negligible, 12% subset of the total 164 atoms of the E2b element of the MT₁ receptor and seems to affect strongly the folding of the E2 loop and neighbouring ERs or TM domains. Our MD simulations indicate that this larger sterical hindrance of E2b in GPR50 could explain the preferential intramolecular interactions of the E2b with ER, rather than with the TM domain, as observed for MT₁ receptors. Indeed, MD simulations have shown a stabilization of E2b‐TM units by conserved interactions between Gln181^E2b(C + 4) and Thr188^5.40 and between Phe179^E2b(C + 2) and TM6 (Ile254^6.51, Leu257^6.54 and Gly258^6.55) and TM7 (Tyr281^7.39). On the other hand, although Phe179^E2b(C + 2) is conserved between MT₁ receptors and GPR50, its orientation is modified by more steric surrounding variable side chains and provokes an extracellular hydrophobic cluster with N‐terminal and E3 regions in the GPR50 model. Interestingly, Phe179^E2b(C + 2) and Gln181^E2b(C + 4) are conserved in Mel1c, the orthologue of GPR50 in lower vertebrates that binds melatonin with high affinity (Figure 1A).

Figure 6.

Surface of the MT₁ receptor from the extracellular (A) and lateral (B) view. E2a and E2b appear as yellow and red surfaces respectively, whereas the remaining receptor surface is in cyan. Melatonin is illustrated as a green space‐filling model.

Consequently, the E2 loop is likely to play a role in the conformational changes associated with MT₁ receptor activation, rather than in melatonin binding. Similar to rhodopsin (Ahuja et al., 2009), participation of the E2 loop of MT₁ receptors and, particularly the E2b element, in the stabilization of TM 5, 6 and 7 of the active melatonin‐bound form of the receptor could favour the conformational changes necessary for signal transmission to the cytoplasmic receptor surface. In the case of GPR50, the steric hindrance of E2b tends to favour interactions with other extracellular region, accentuating the volume of the extracellular cap and occluding the putative access pathway to the ligand binding pocket between extracellular extremities of TM5 and TM6.

The possible clinical relevance of this study resides in the identification of the E2 loop as the first site of interest for allosteric modulation of MT₁ receptor activity. Compounds targeting the E2 loop of MT₁ receptors are expected to behave as allosteric modulators. Currently, no allosteric modulators have been identified for these receptors. Such compounds are of particular interest as they would not alter the natural circadian activation pattern of melatonin receptor activity which is essential for a chronobiotic hormone, such as melatonin.

In conclusion, in silico molecular modelling and MD simulations and in vitro functional studies indicate that the E2 loop of MT₁ receptors is critical for its function by regulating access to the melatonin binding pocket and/or favouring the conformational changes necessary for signal transmission across the membrane to the cytoplasmic receptor surface. In contrast, the E2 loop of GPR50 is unable to fulfil this function(s) providing a first explanation why GPR50 does not bind melatonin despite the presence of a comparable ligand binding pocket.

Author contributions

N.C., N.R., J.L.G., F.C., A.G., P.D. and R.J. conceptualized the study; N.C., J.L.G., E.C., A.S.J. and K.T. carried out the biochemical investigations; N.R. and X.L. performed molecular modelling and MD simulations; N.R. and R.J. wrote the original draft; N.R., J.L.G., E.C. and R.J. did the writing, reviewing and editing; P.D., P.C. and R.J. acquired funding.

Conflict of interest

The authors declare no conflicts of interest.

Declaration of transparency and scientific rigour

This http://onlinelibrary.wiley.com/doi/10.1111/bph.13405/abstract acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research recommended by funding agencies, publishers and other organisations engaged with supporting research.

Supporting information

Figure S1 Sequence homology of melatonin receptors. According to GRAFS phylogenetic tree (Fredriksson et al., 2003, Mol Pharmaco), the closest sequences of crystal templates from melatonin MT₁ (MTNR1A) and MT₂ (MTNR1B) receptors and GPR50 are rhodopsin (RHO), β‐adrenoceptors (ADRB1 and ADRB2) and A2A (ADORA2A) adenosine receptors (A). Sequence identity and homology rates were computed for the whole or transmembrane region (B) whereas the multiple sequence alignment used by Modeller for homology modelling of MT1 receptor was manually adjusted after an automatic alignment (clustalΩ) in order to shift gaps outward the helical transmembrane regions (C). Panel D shows a superimposition of selected apo‐models of MT1, mutMT1 and GPR50 models.

Figure S2 Signalling and expression of wild type and chimeric MT₁/GPR50 receptors. (A) Melatonin‐induced activation of the G_i/cAMP pathway and recruitment of ß‐arrestin2 of receptors summarized in Table 1. (B,C) Melatonin‐induced activation of the G_i/cAMP pathway of receptors summarized in Table 3. (D‐F) Representative Western blot detection of receptor expression levels in cell lysates (D, E, F chimera from Tables 1, 2, 3, respectively). Western blots are representative of at least 3 independent experiments.

Figure S3 Structural deviation of melatonin during MD simulation of holoMT1, holoMT1mut and holoGPR50 receptors. The structural deviation (nm) is computed in term of RMSD to the reference t0 frame of both 300 ns MD simulations after fitting the α‐carbon trace in each of three replicated experiments.

Document S4 Detailed analysis of E2 folding in MT1, mutMT1 and GPR50 receptor models during MD simulations.

Electronic Material S5 Motion of a possible ligand accessibility pathway from supervised MD simulation steering the melatonin off the binding site of MT₁.

Clement, N. , Renault, N. , Guillaume, J.‐L. , Cecon, E. , Journé, A.‐S. , Laurent, X. , Tadagaki, K. , Cogé, F. , Gohier, A. , Delagrange, P. , Chavatte, P. , and Jockers, R. (2018) Importance of the second extracellular loop for melatonin MT₁ receptor function and absence of melatonin binding in GPR50. British Journal of Pharmacology, 175: 3281–3297. 10.1111/bph.14029.