Update on melatonin receptors: IUPHAR Review 20

MT receptor ligands

MT₁ and MT₂ receptors share a high degree of sequence homology and bind both the natural ligand, melatonin, with high affinity. Important progress has been made in the development of synthetic MT receptor antagonists and agonists and subtype-selective ligands by diversifying the chemical scaffolds. Indeed, MT receptor ligands from different structural classes show distinct structure–activity relationships on native and recombinant MT receptors (Dubocovich et al., 1997; Dubocovich et al., 2010; Zlotos, 2012; Zlotos et al., 2014). The methoxy group and the acetamido side chain of melatonin determine the intrinsic activity and the binding affinity, respectively, at both hMT₁ and hMT₂ receptors (Dubocovich et al., 1997; Browning et al., 2000; Audinot et al., 2003). Replacement of the amide methyl group by ethyl and propyl substituents enhances affinity (Sugden et al., 1995). Exchange of the indole ring by various aromatic scaffolds maintains high binding and agonist potency.

Substitutions at the two-position with a halogen or a phenyl group generate agonists with ~10-fold increased binding affinity. The majority of non-selective MT₁–MT₂ receptor ligands, including drugs used in humans, for example, ramelteon [Rozerem®, (Kato et al., 2005; Mini et al., 2007; Rawashdeh et al., 2011)], agomelatine [Valdoxan®, (de Bodinat et al., 2010)] and tasimelteon [Hetlioz®, (Rajaratnam et al., 2009; Lavedan et al., 2015)], are agonists (Figure 1). Ramelteon and tasimelteon are MT receptor selective, while agomelatine is also an antagonist at the 5-HT_2C receptors, a pharmacological property believed to contribute to its antidepressant action. The therapeutic effects of approved drugs acting on hMT₁ and/or hMT₂ receptors as agonists were recently reviewed (Liu et al., 2016). Other non-selective MT receptor agonists include 6-chloromelatonin, 6-hydroxymelatonin, 2-iodomelatonin, GR 196429 (Dubocovich et al., 1997; Browning et al., 2000; Audinot et al., 2003), UCM 793 (Rivara et al., 2007) and 2-methoxy-α,β-didehydro-agomelatine (Morellato et al., 2013). This latter ligand shows the highest affinity for hMT₁ (K_i = 0.03 nM) and hMT₂ (K_i = 0.07 nM) receptors and ~3500-fold greater potency than melatonin in the melanophore aggregation assay. TIK 301 (Mulchahey et al., 2004) acts also as an antagonist at the 5-HT_2C and 5-HT_2B receptors (Landolt and Wehrle, 2009). 5-HEAT has a unique pharmacological profile acting as a full agonist at the hMT₁ receptor and antagonist at the hMT₂ receptor (Nonno et al., 2000). EFPPEA, a high-affinity hMT₁ (K_i = 0.062 nM) and hMT₂ receptor (K_i = 0.420 nM) agonist, decreases the percentage of wakefulness and increases the percentage of slow wave sleep in cats (Koike et al., 2011). The competitive MT receptor antagonist luzindole lacks the methoxy group, which led to the suggestion that this group is necessary for intrinsic activity (Dubocovich, 1988). Similarly, S22153 acts as a partial agonist (Audinot et al., 2003). Luzindole, with a 15- to 25-fold higher affinity for hMT₂ than for hMT₁ receptors, is widely used for pharmacological characterization of functional MT receptors (Dubocovich et al., 1997; Dubocovich et al., 1998; Browning et al., 2000; Dubocovich et al., 2010).

A ligand is considered selective for a specific receptor type when its affinity or potency is at least 100-times higher than that for the other(s) receptor types in the family (Dubocovich et al., 2010). This concept holds true for in vitro studies where ligand concentrations can be easily adjusted. However, ligand selectivity might be more difficult to reach in vivo, in the body fluids reaching the receptors. Depending on ligand dose and pharmacokinetics, concentrations could easily raise to levels activating both receptors (e.g. MT₁ and MT₂ receptors). This is particularly of concern for melatonin and synthetic MT receptor ligands activating receptors at picomolar concentrations (Dubocovich et al., 1997; Browning et al., 2000; Audinot et al., 2003). Therefore, caution should be taken when interpreting selective MT receptor activation in vivo using MT receptor-selective ligands, unless pharmacological selectivity or lack of it, is confirmed by KO models with deletion of each receptor type.

Numerous ligands with high selectivity for hMT₂ over hMT₁ receptors have been identified (Zlotos et al., 2014). MT₂ receptors possess a lipophilic pocket close to the N1–C2 binding region of melatonin, which is absent in MT₁ receptors (Rivara et al., 2005). Accordingly, most MT₂ receptor-selective ligands bear a flexible bulky hydrophobic substituent in a position equivalent to C2 or N1 of melatonin (Figure 2). The tetrahydroquinoline analogue UCM1014 is the most potent MT₂ receptor-selective ligand reported to date. It shows picomolar binding affinity (K_i = 0.001 nM) at hMT₂ receptors, >10 000-fold selectivity over hMT₁ receptors and full agonist profile in the GTPγS test (Spadoni et al., 2015). Other agonists with approximately 800-fold hMT₂ receptor selectivity are BOMPPA (Hu et al., 2010; Heckman et al., 2011; Chan et al., 2013; Hu et al., 2013) and CIFEA (Koike et al., 2011). In imprinting control region mice, CIFEA reentrainment effects to a new light/dark cycle indicate the involvement of MT receptors in the regulation of chronobiotic activity (Koike et al., 2011). The dose of CIFEAA used in this study most likely reached micromolar concentrations, which would activate both MT₁ and MT₂ receptors, precluding any conclusion about the specific receptor type involved in the regulation of chronobiotic processes. Similarly, doses of the MT₂ receptor-selective antagonist 4P-PDOT (90 μg/mouse s.c.) used to block the melatonin-mediated phase advance of circadian activity rhythms in mice (Dubocovich et al., 1998) may have resulted in micromolar circulating 4P-PDOT concentrations hence blocking both MT₁ and MT₂ receptors.

Two moderately selective MT₂ receptor ligands, the agonist IIK7 (Faust et al., 2000) and the partial agonist UCM 765 (Rivara et al., 2007), have been used to examine the role of each MT receptor type in the modulation of sleep architecture. UCM 765 promoted non-rapid eye movement (NREM) sleep in rodents, and this effect was blocked by the MT₂ receptor antagonist 4P-PDOT (Ochoa-Sanchez et al., 2011). In contrast, the non-selective MT₁–MT₂ receptor agonist UCM793 decreased sleep onset without having an effect on NREM sleep maintenance suggesting that dual MT₁ and MT₂ receptor agonistic activity accounts for the effect on sleep onset, whereas selectivity for MT₂ receptors has an additional effect on NREM sleep maintenance. IIK7 was also reported to reduce NREM sleep onset latency and transiently increase the time spent in NREM sleep in rats without altering rapid eye movement (REM) sleep latency or the amount of REM sleep (Fisher and Sugden, 2009).

Among the hMT₂ receptor-selective partial agonists GR 128107, 5-methoxyluzindole, S 24014, S 24773 (Dubocovich et al., 1997; Audinot et al., 2003) and isoamyl agomelatine, the latter shows the highest affinity (K_i = 0.01 nM) and selectivity (7200-fold) (Ettaoussi et al., 2012). 4P-PDOT, an hMT₂ receptor-selective antagonist with 300- to 1500-fold higher affinity for hMT₂ receptors, is still considered the gold standard for pharmacological characterization of MT receptors (Dubocovich et al., 1997). Other MT₂ receptor-selective antagonists, such as K185 (Sugden et al., 1999; Faust et al., 2000), UCM 454 (Rivara et al., 2005) and 2-(indolin-1yl) melatonin (Zlotos et al., 2009), display ~100-fold higher affinity for hMT₂ receptors. For (hydroxymethyl)phenyl agomelatine, the affinity for hMT₂ receptors is 750-times higher than for hMT₁ receptors (Poissonnier-Durieux et al., 2008).

Discovery of MT₁ receptor-selective ligands remains a challenge, and only few compounds with preference for hMT₁ receptors have been reported (Zlotos et al., 2014). Ligands preferentially binding to these receptors reach maximally 100-fold selectivity, and, when investigated, this selectivity is significantly reduced in functional in vitro studies (Figure 3). A common structural feature conferring MT₁ receptor selectivity is a bulky, hydrophobic ether replacing the methoxy group. The first hMT₁ receptor-selective agents were obtained by connecting two agomelatine units via their ether oxygen by (CH₂)₃- and (CH₂)₄-linker to give S 26131 (antagonist) and S 26284 (partial agonist), both displaying ~100-fold selectivity (Audinot et al., 2003; Descamps-Francois et al., 2003). A similar approach led to the UCM 793 dimer with 100-fold hMT₁ receptor selectivity and partial agonist activity (Spadoni et al., 2011). Monomeric ligands, such as CBOBNEA (Mesangeau et al., 2010), and AAE M PBP amine (Rivara et al., 2012) are partial agonists showing similar ~100-fold selectivity for hMT₁ receptors. N-acetyl-O-phenoxypropyl serotonin is a full agonist obtained by exchange of the methoxy group of melatonin with an O(CH₂)₃OPh moiety. Although it shows only 10-fold binding preference toward hMT₁ receptors, its MT₁–MT₂ receptor binding ratio and hMT₁ receptor affinity were higher than that for the MT₁ receptor-selective reference compound S 26131 that was retested under the same experimental conditions (Markl et al., 2011). A 140-fold MT₁ receptor selectivity could be attained by introduction of two fluorine atoms into the N-acetyl group of agomelatine. The resulting difluoroagomelatine shows high hMT₁ receptor binding (K_i = 0.03 nM) and is a non-selective MT₁–MT₂ receptor full agonist (Ettaoussi et al., 2012). Very recently, tetrafluoro S26131, the difluoroacetamide analogue of S 26131, has been reported to show higher affinity and selectivity toward hMT₁ receptors than the parent ligand (Zlotos et al., 2015).

In summary, while numerous ligands selective for the MT₂ receptor subtype are available, discovery of ligands with at least >100-fold selectivity for MT₁ receptors remains a challenging task. None of the MT₁ receptor-selective ligands has been tested in vivo. Future progress on the elucidation of the structure of MT receptors will hopefully foster the discovery of such ligands. The MT₁–MT₂ non-selective receptor antagonist luzindole and the MT₂ receptor-selective antagonists 4P-PDOT are still considered the gold standards for pharmacological characterization of MT receptors.

Radioligands – update

Radioactive- and fluorescent-labelled ligands are indispensable tools for the pharmacological characterization of GPCRs. A major breakthrough in the field of MT receptor research was the labelling of 2-iodomelatonin with ¹²⁵I at carbon 2 resulting in a high-affinity radioligand with high specific activity (Vakkuri et al., 1984) for use in the localization (Vanecek, 1988) and pharmacological characterization of MT receptors in tissues (Dubocovich and Takahashi, 1987). The radioligand, 2-[¹²⁵I]iodomelatonin (2-[¹²⁵I]-IMLT), has been extensively used as a high-affinity radioligand for both MT₁ and MT₂ receptors, and was until recently the only available radioligand for the characterization and localization of melatonin binding sites in native tissues (Figure 4). Studies performed with [³H]-melatonin ([³H]-MLT) established the pharmacological profile of the human recombinant MT₁ and MT₂ receptors, as being identical to that established using 2-[¹²⁵I]-IMLT as a radioligand. However, due to the rather low specific activity of this [³H]-MLT, its use to characterize and/or localize melatonin sites in tissues with low MT receptor density is limited (Browning et al., 2000).

Three new iodinated radioligands have been recently characterized for use in the pharmacological characterization and localization of MT receptors (Figure 4). These radioligands are as follows: SD6 (N-[2-(5-methoxy-1Hindol-3-yl)ethyl]iodoacetamide), S70254 (2-iodo-N-2-[5-methoxy-2-(naphthalen-1-yl)-1H-pyrrolo[3,2-b]pyridine-3-yl])acetamide and DIV880 2-(2-[(2-iodo-4,5-dimethoxyphenyl)methyl]-4,5-dimethoxy phenyl) (Legros et al., 2013; Legros et al., 2016). [¹²⁵I]-SD6 has a similar pharmacological profile to that of 2-[¹²⁵I]-IMLT with the same affinity for MT₁ and MT₂ receptors. On the contrary, the two other radioligands [¹²⁵I]-S70254 and [¹²⁵I]-DIV880 show selectivity for MT₂ receptors with pK_d values of 9.6 and 9.7, respectively, in the absence of any specific binding to MT₁ receptors. All radioligands are agonists, either partial agonists ([¹²⁵I]-S70254, [¹²⁵I]-DIV880) or full agonists (2-[¹²⁵I]-IMLT, [¹²⁵I]-SD6, [³H]-MLT), which means that their K_d values depend not only on the affinity of the ligand for the receptor but also on the activation of the G protein in the ternary Ligand-Receptor-G protein complex.

The extensive pharmacological characterization of these three new radioligands in comparison with 2-[¹²⁵I]-IMLT and [³H]-MLT on membrane preparations from CHO-K1 cell lines stably expressing hMT₁ or hMT₂ receptors showed that [¹²⁵I]-S70254 and [¹²⁵I]-DIV880 mainly differ from 2-[¹²⁵I]-IMLT in their dissociation kinetics, which are faster for [¹²⁵I]-S70254 and [¹²⁵I]-DIV880 than for 2-[¹²⁵I]-IMLT (Legros et al., 2016). Interestingly, [¹²⁵I]-SD6 labelled only approximately half of the binding sites detected with 2-[¹²⁵I]-IMLT in cells expressing hMT₁ receptors while comparable amounts were detected in cells expressing hMT₂ receptors (Legros et al., 2013). This suggests the existence of different receptor subpopulations for hMT₁ receptors of which [¹²⁵I]-SD6 labels a more restricted number than 2-[¹²⁵I]-IMLT. In contrast, for hMT₂ receptors, similar subpopulations would be detectable by both radioligands. The nature of these receptor subpopulations is currently unknown but could be related to the differential engagement of hMT₁ receptors into complexes with different G proteins or ß-arrestins following the binding of these agonistic radioligands. [¹²⁵I]-SD6 detected as 2-[¹²⁵I]-IMLT binding sites in sheep retinal membranes, while the MT₂ receptor-specific ligands [¹²⁵I]-S70254 and [¹²⁵I]-DIV880 failed to do so.

The MT₂ receptor-specific [¹²⁵I]-S70254 was successfully used for autoradiography studies in rat and sheep brain and retina slices (Legros et al., 2016). A similar labelling pattern to 2-[¹²⁵I]-IMLT (detecting MT₁ and MT₂ receptors) was observed in several areas but also distinct labelling in others. Absence of labelling by [¹²⁵I]-S70254 in regions that are labelled by 2-[¹²⁵I]-IMLT can be explained by low (undetectable) MT₂ receptor levels. Absence of 2-[¹²⁵I]-IMLT labelling in regions labelled by [¹²⁵I]-S70254 could be due to the detection of different receptor complexes (see above).

Altogether, the new radioligands considerably expand the repertoire of pharmacological tools for MT receptors with the development of MT₂ receptor-specific radioligands and radioligands detecting distinct receptor populations revealing a previously unrecognized diversity. The availability of a radiolabelled antagonist would largely contribute in a better characterization of these different populations. Further advances can be expected from the development of fluorescent-labelled ligands.

Structural perspectives for MT receptors

Currently, crystal structures of MT₁ and MT₂ receptors are not available. Despite a sequence identity lower than 30% between MT receptors and the closest crystallized GPCRs, several three-dimensional (3D) models have produced some structural hypotheses for binding of (non)selective MT₁ and/or MT₂ receptor agonists (Table 1). According to site-directed mutagenesis data, most of these models corroborate the importance of both serine residues 3.35 and 3.39 in MT₁ as well as His5.46 in both MT₁ and MT₂ receptors. Although His5.46 seems to be an anchoring residue for polar interactions with the methoxy or amide group of melatonin, only a few models of MT₁ receptors display a direct participation of serine residues in melatonin binding (Chugunov et al., 2006; Farce et al., 2008), which could be otherwise involved in an essential bending of helix 3 for binding site plasticity. Models take also into account several receptor–ligand interactions with amino acids conserved within GPCRs and known to play a role in aromatic switch activation (F5.47, W6.48).

Such homology modelling methods make predictions of flexible receptor regions difficult. Although not directly proven for MT receptors, E2 and I3 loops are known to be key features for ligand accessibility and G protein binding of GPCRs. Moreover, amino acid sequences of MT receptors show several singularities like the presence of a ^3.49NRY^3.51 motif instead of the classical ^3.49DRY^3.51 motif of other rhodopsin-like GPCRs. Another specificity is the replacement of the proline by an alanine residue in the conserved ^7.49NPXXY^7.53 motif. Buried in the vicinity of the cytoplasmic surface, these marked differences are likely to affect receptor activation and/or signalling specificity of MT receptors rather than the ligand binding process.

Whereas 3D models of MT receptors were up to now dedicated to the discovery and optimization of new efficient drugs, the next generation of 3D models should be expanded toward larger, multimeric systems and not be restricted to receptor monomers. Computation of the energy landscape of GPCRs by enhanced molecular dynamics simulations, together with NMR and X-ray studies, provided valuable molecular insights on the dynamics of ligand recognition, receptor activation and oligomerization (Johnston and Filizola, 2014). Depicting free energy landscapes of MT receptors should address biasing molecular dynamics simulations from the inactive apoform transiting toward the active trimeric L-R-Gi or L-R-arrestin forms of receptors (Figure 5). As ligands modulate these free energy landscapes (Provasi et al., 2011; Dror et al., 2013), in silico optimization of new efficient ligand structures could be explored by predicting its functional selectivity through arrestin or G_i-mediated pathways. These approaches also open the way for the exploration of homodimers and heterodimers, particularly MT₁/MT₂ and MT₁/GPR50 receptor complexes, as discussed in the following section.

MT receptor dimers

MT receptors are part of dynamic signalling complexes that contain proteins involved in receptor biosynthesis, export, signalling, desensitization, internalization and cytoskeleton modulation (Daulat et al., 2007; Maurice et al., 2008) (IntAct database, http://www.ebi.ac.uk/intact/search/do/search?searchString=pubid:26514267). The core of these complexes is often composed of receptor dimers, either homodimers of the same receptor or heterodimers composed of two different receptors (see Maurice et al., 2011 and Ferre et al., 2014). Initial observations have been made in 2002 in transfected HEK293 cells demonstrating the capacity of MT₁ and MT₂ receptors to form homodimers and heterodimers (Ayoub et al., 2002) with MT₁/MT₂ heterodimers showing a pharmacological profile distinct from MT₂ homodimers (Ayoub et al., 2004). Shortly after, MT₁ and MT₂ receptors were reported to form heterodimers with the orphan GPR50, which completely abolished the function of MT₁ receptors in MT₁/GPR50 heterodimers (Levoye et al., 2006a). Sporadic reports on Western blots of endogenously expressed MT receptors in chicken astrocyte cultures (Adachi et al., 2002) and Xenopus tectal cells (Prada et al., 2005) further indicated the possible existence of MT receptor homodimers. However, the physiological relevance of these dimers remained largely unclear (Levoye et al., 2006b) until 2013 when compelling in vivo evidence for the functional significance of MT₁/MT₂ heterodimers was obtained. In retinal photoreceptor cells, melatonin enhances the light sensitivity during the night. The phenotype of MT₁ receptor KO (MT₁^−/−) and MT₂ receptor KOs (MT₂^−/−) mice, the use of type-selective ligands and overexpression of a dominant negative form of MT₂ receptors in photoreceptor cells of transgenic mice indicated the exclusive involvement of MT₁/MT₂ heterodimers in this physiological effect of melatonin (Baba et al., 2013). Interestingly, this effect was dependent on the activation of the G_q/PLC pathway by melatonin, an observation that could be confirmed in vitro in cells co-expressing MT₁ and MT₂ receptors.

Whether MT receptor heterodimers could become novel drug targets remains an open question. Recent evidence on the antidepressant agomelatine suggests this possibility (Kamal et al., 2015). Previous studies showed that agomelatine is a high-affinity agonist for MT₁ and MT₂ receptors and an antagonist with moderate affinity for 5-HT_2C receptors (Audinot et al., 2003; Millan et al., 2003). Of note, the antidepressant effect of agomelatine involves both pathways in a synergistic manner (Racagni et al., 2011). Formation of MT₂/5-HT_2C heterodimers was demonstrated in transfected HEK293 cells, and these heterodimers are targeted by agomelatine (Kamal et al., 2015). Agomelatine behaved as a biased ligand, activating the G_i/cAMP pathway and antagonizing the G_q/PLC pathway. Whether the MT₂/5-HT_2C heterodimer participates in the antidepressant effect of agomelatine remains to be shown.

Formation of receptor dimers offers the possibility to design dimeric ligands targeting receptor dimers. Several dimeric ligands with two identical pharmacophores have been synthesized for MT receptors and their binding properties have been determined (Audinot et al., 2003; Descamps-Francois et al., 2003; Mesangeau et al., 2010; Spadoni et al., 2011; Journe et al., 2014). Binding of the two pharmacophores of these dimeric ligands to the two protomers of the same receptor dimer has been only shown in one study using a BRET approach (Journe et al., 2014). Compounds linked through 22–24 atom spacers were able to bind to MT₁ and MT₂ receptor protomers in pre-existing homodimers and heterodimers and to induce conformational changes detected by BRET. Induction of receptor dimerization was not observed. The functional properties of these compounds remain to be studied. Taken together, the existence and physiological relevance of MT receptor dimers are increasingly recognized, but its functional role and pharmacological exploitation are still ongoing.

Genetic variants and mutants of MT receptors

The existence of many rare variants in the human population was discovered in recent genome sequencing programmes. The 1000 human genome project detected 38 million variants (Abecasis et al., 2012) and 172 variants, including 46 non-synonymous variants, has been identified on average per GPCR in a population of 14 002 individuals (Nelson et al., 2012; Karamitri and Jockers, 2014). Numerous variants have been identified in the MTNR1A and MTNR1B genes, encoding MT₁ and MT₂ receptors respectively. Here, only non-synonymous variants, modifying the amino acid sequence of the receptors, will be considered (Figure 6). Variants with altered receptor function can potentially participate in disease development. Ebisawa et al. (1999) were searching for variants in MTNR1A and MTNR1B genes in patients with circadian disorders. Two non-synonymous variants were identified in the MTNR1A gene (R54W, A157V) that were threefold and twofold more frequent in people with non-24 h sleep–wake syndrome (Table 2) (Ebisawa et al., 1999). Due to the small sample size (N = 22), statistical significance was not reached.

Alteration of melatonin synthesis has been reported in autism spectrum disorders (ASDs) triggering the search for variants in MTNR1A and MTNR1B genes in 295 patients with ASD, 362 controls and 284 individuals from the human genome diversity panel (Chaste et al., 2010). Six non-synonymous mutations were identified for MTNR1A and 10 for MTNR1B (Tables 2 and 3). The majority of these mutants showed altered receptor function. Particularly deleterious mutants were MT₁-I49N, which is devoid of any melatonin binding and cell surface expression, and MT₁-G166E and MT₁-I212T, which showed severely impaired cell surface expression and biased behaviour toward the ERK1/2 pathway. No significant difference in the prevalence of these mutations was found indicating that they do not represent major risk factors for ASD.

Four non-synonymous mutations were identified for MTNR1A and four for MTNR1B in a cohort of 101 individuals with attention-deficit/hyperactivity disorder (ADHD) (Tables 2 and 3); however, none of them was enriched in ADHD individuals as compared with the general population (Chaste et al., 2011). The MT₁-Y170X nonsense mutation was only detected in one ADHD patient and introduced a premature STOP codon resulting in complete loss of receptor function.

MT receptor variants have been most extensively sought in studies focused on type 2 diabetes (T2D) based on the discovery of several frequent polymorphisms associated with increased fasting plasma glucose (FPG) and T2D risk close to the MTNR1B gene in genome-wide association studies (Bouatia-Naji et al., 2009; Prokopenko et al., 2009). Sequencing of the coding region of the MTNR1B gene revealed six non-synonymous variants (G24E, L60R, V124I, R138C, R231H and K243R) of which none was associated with T2D risk. The common 24E variant was associated with increased body mass and decreased FPG (Andersson et al., 2010), an observation that was not replicated in a later study (Bonnefond et al., 2012). Whereas only subtle changes in the capacity of G24E and V124I to activate a GαΔ6qi4myr chimeric G protein, the L60R variant was completely inactive in transfected COS cells. A more extensive sequencing study discovered 40 non-synonymous variants in the coding region of MTNR1B (Tables 2-4) (Bonnefond et al., 2012) of which 36 very rare mutants associated with T2D risk. Functional analysis of the 40 variants revealed intact cell surface expression for all variants. There was complete loss of melatonin binding in four very rare cases (A42P, L60R, P95L and Y308S) and partially and severely blunted signalling (Gαqi9 chimera and ERK1/2 activation) in one rare case (R138C) and nine very rare cases (W22L, A52T, A74T, R138H, R138L, L166I, R222H, R330W and I353T). Carriers of the 13 very rare loss-of-function variants showed increased T2D risk establishing a functional link between MTNR1B and T2D (see Karamitri et al., 2013).

In conclusion, the genetic variability of the MTNR1B gene in terms of non-synonymous variants has now been well defined and an association of very rare variants with T2D risk established. Less is known about the variability of the MTNR1A gene in terms of non-synonymous variants.

MT receptor mouse models – update

MT₁^−/− mice were created in the late 1990s followed by the generation of MT₂^−/− mice in 2003 (Liu et al., 1997; Jin et al., 2003). Studies using these mice have provided important insights on the role that MT receptors play in the modulation of many different biological functions. In MT₁^−/− mice, but not in MT₂^−/− mice, the inhibitory effect of melatonin on neuronal activity in the suprachiasmatic nucleus (SCN) is impaired suggesting the involvement of MT₁ receptors. In contrast, in SCN slices from MT₁^−/− mice melatonin (1–10 pM) phase shifts the peak of circadian rhythms of neuronal firing by approximately 3 h suggesting the involvement of MT₂ receptors (Liu et al., 1997; Dubocovich et al., 2005). Blockade of this effect using the MT₂ receptor-selective 4P-PDOT antagonist confirmed the latter conclusion shaping a pathway where MT₂ phase shifts the peak of neuronal firing through PLC-PKC signalling pathway (Mc Arthur et al., 1997; Hunt et al., 2001; Dubocovich et al., 2005). Liu et al. (1997) reported that the phase shift of neuronal firing rhythms induced by 2-iodomelatonin (10 pM) was of smaller magnitude in the SCN slice from MT₁^−/− than in wild type (WT) mice suggesting a role for the MT₁ receptors in this response (see detailed discussion in Dubocovich, 2007). Together, these findings suggest a potential role for both MT₁ and MT₂ receptors in the phase shift of circadian rhythms of neuronal firing in the SCN slice in vitro. The use of MT₁^−/− mice demonstrated that the MT₁ receptor is required for the melatonin-meditated phase shift of the onset of overt circadian rhythm of locomotor activity (Dubocovich et al., 2005). An independent study demonstrated that C3H/HeN mice (melatonin-proficient) entrained faster to a phase advance of dark onset than the C57BL/6J mice (melatonin-deficient), suggesting a facilitating role of endogenous melatonin on circadian reentrainment (Pfeffer et al., 2012). However, we should note that faster entrainment could also result from genetic differences between the two mouse strains rather than different endogenous melatonin levels (Adamah-Biassi et al., 2013). In a mouse strain producing endogenous melatonin, the faster entrainment to an abrupt advance of dark onset persisted in MT₁^−/− C3H/HeN mice but was lost in MT₂^−/− and double KOs (MT₁^−/−/MT₂^−/−) suggesting again the involvement of MT₂ receptors. This apparent contradiction could be explained by the activation of MT₂ and MT₁ receptors by endogenous and exogenous melatonin, respectively, at different periods of sensitivity (subjective night vs. subjective day, respectively). Changes in efficacy could also result from desensitization and/or internalization of MT receptors in response to exposure to physiological and supraphysiological melatonin concentrations as demonstrated by the phase shift of the peak of neuronal firing in the SCN by physiological levels of melatonin, which involved the desensitization of MT₂ receptors (Gerdin et al., 2004).

MT receptor KO mice have been also used to elucidate the role played by these receptors in the regulation of the sleep/wake cycle. In MT₂^−/− NREM, sleep is decreased during the light phase (i.e. during the time that mice normally sleep), whereas MT₁^−/− mice showed an increase in the amount of NREM sleep during the dark phase (i.e. during active phase) (Ochoa-Sanchez et al., 2011). Further analysis of the data indicated that MT₁ receptor signalling is implicated in the modulation of the daily rhythm of REM sleep (Ochoa-Sanchez et al., 2011). An additional study in which double KOs (MT₁^−/−/MT₂^−/−) were used indicated that removal of both receptors induced an increase in wakefulness and a reduction in REM sleep (Comai et al., 2013). Hence, these data seem to indicate that removal of MT receptors may affect wakefulness rather than sleep.

The effect of MT receptor removal has been also investigated in the mouse retina, where these receptors are widely distributed (Baba et al., 2009; Baba et al., 2013). Removal of either receptor has profound effects on photoreceptors function as it abolishes the daily rhythms in the scotopic and photopic electroretinogram (Baba et al., 2009; Alcantara-Contreras et al., 2011; Sengupta et al., 2011). Such a result also indicates that MT₁ and MT₂ receptors form heterodimers in mouse photoreceptors (Baba et al., 2013). Further studies have also demonstrated that removal of MT receptors in addition affects the viability of the photoreceptors and retinal ganglion cells during aging (Baba et al., 2009; Alcantara-Contreras et al., 2011; Gianesini et al., 2016) as well as corneal biology (Baba et al., 2015).

As mentioned before, recent studies have also implicated MT receptors in the pathogenesis of T2D in humans (Bouatia-Naji et al., 2009; Lyssenko et al., 2009; Bonnefond et al., 2012). Thus, a few studies used MT receptor KO mice to determine the mechanisms by which these receptors contribute to regulation of glucose homeostasis and insulin sensitivity (Stumpf et al., 2008; Muhlbauer et al., 2009; Contreras-Alcantara et al., 2010). Mice lacking MT₁ receptors exhibit higher mean blood glucose levels than controls (Muhlbauer et al., 2009) and tend to be more glucose intolerant and insulin resistant than WT and MT₂^−/− mice (Contreras-Alcantara et al., 2010). Furthermore, removal of MT₁ or MT₂ receptors abolishes the daily rhythm in blood glucose levels (Owino et al., 2016).

Finally, it is important to mention that although the reproductive system of mice is not sensitive to photoperiod, the development of MT receptor KO mice provided an important tool for dissecting the mechanisms by which melatonin regulates reproduction in photoperiodic species. For example, MT₁ receptor signalling controls the rhythmic expression of the clock gene Period 1 in the pituitary gland (von Gall et al., 2002), and further studies have shown that the rhythmic expression of several other clock genes (Per1, Per 2, Bmal1 and Cry 1) in the mouse Pars tuberalis depends on MT₁ receptor signalling as well (Jilg et al., 2005). MT₁ receptor signalling has been also reported to be crucial for the photoperiodic response of gene expression in the ependymal cell layer and thus for the photoperiodic regulation of gonadal activity (Sheynzon and Korf, 2006; Yasuo et al., 2009). Finally, we should mention that a recent study reported that MT₁ receptor signalling plays a key role in photoperiodic programming of serotonergic neurons as well as depression- and anxiety-related behaviours in mice (Green et al., 2015).

In conclusion, studies in the last 20 years using MT receptor KO mice have greatly helped to understand the role(s) played by these receptors in the regulation of many physiological functions, and they have provided important insights on the mechanisms by which melatonin signalling affects these functions.

Functional role of MT receptors in physiology and pathophysiology

MT receptors are involved in many physiological processes that will however not all be covered by this review but can be consulted in other reviews (Dubocovich et al., 2010; Tosini et al., 2012; Karamitri et al., 2013; Tosini et al., 2014; Johnston and Skene, 2015). Here, we have focused our attention on two major processes, the immune system and the CNS. Important progress has been made recently in both fields, and links to diseases have been established justifying a review of our current knowledge on these aspects. Finally, we will make a critical assessment of reports of receptor-independent effects of melatonin, such as its binding to additional binding sites and the intrinsic antioxidant and free radical scavenger properties of this hormone.

MT receptors in the immune system

The role of melatonin as a player in immunity, first proposed by Berman in 1926, is now well accepted (Carrillo-Vico et al., 2013). Several reports have demonstrated that melatonin produced by the either pineal gland or immune cells can regulate the activation of an immune response. Melatonin derived from activated human lymphocytes induces the synthesis of IL-2 and IL-2 receptors (Carrillo et al., 2004; Carrillo-Vico et al., 2013). Luzindole and targeted deletion of the MNTR1A gene (Lardone et al., 2006; Lardone et al., 2010) block the effect of lymphocyte-derived melatonin. Interestingly, daily rhythms of plasma melatonin and IL-2 are transiently lost in non-infectious human inflammatory conditions, and the recovery of the IL-2 rhythm follows the restoration of the daily melatonin rhythm (Pontes et al., 2007). In addition, the daily and seasonal variation of melatonin production contributes to the seasonality of some diseases. In multiple sclerosis (MS), melatonin blocks the differentiation of Th17 cells and boosts the generation of protective type 1 regulatory T-cells by an MT₁ receptor-dependent mechanism, resulting in the seasonal variation of MS symptoms (Farez et al., 2015). Seasonality of regular immunity is also related to changes in the melatonin system (Weil et al., 2015). In the spleen of several species, extended light exposure decreases MT₁ receptor expression (Maestroni, 1993; Lahiri and Haldar, 2009; Yadav and Haldar, 2013). In healthy conditions, rolling and adhesion of neutrophils to the endothelial cell layer are inhibited by activation of MT₂ receptors and ligands binding to the putative MT3 binding site, respectively (Lotufo et al., 2001). In contrast, other effects of melatonin such as the inhibition of transcription factors that mediate acute inflammation induced by LPS (Tamura et al., 2010) or N-formyl-l-methionyl-l-leucyl-l-phenylalanine (fMLP) (Cernysiov et al., 2015) were not blocked by luzindole suggesting a mode of action independent of MT receptors.

MT receptors also play an important role in promoting engulfing of bacteria, fungi and parasites. Melatonin facilitates the invasion of erythrocytes by Plasmodium falciparum (Hotta et al., 2000), the invasion of macrophages by Leishmania amazonensis (Laranjeira-Silva et al., 2015) and the phagocytosis of zymosan by colostrum polymorphonuclear and mononuclear cells (Pires-Lapa et al., 2013) and the RAW 264.7 macrophage cell line (Muxel et al., 2012). The entrance of different microorganisms in polymorphonuclear and mononuclear cells, including colostral and lineage-established cell lineages, is blocked by luzindole. Indeed, parasites, bacteria and fungi activate the NF-κB pathway in these two cell models resulting in the expression of arylalkyl-N-acetyltransferase and the synthesis of melatonin. Luzindole and 4P-PDOT blocked the expression of dectin-1, a protein that is important for phagocytosis, suggesting the participation of MT₂ receptors in this effect (Muxel et al., 2012; Pires-Lapa et al., 2013; Muxel et al., 2016). Thus, the evaluation of binding parameters and functional states of MT receptors in immune-competent cells needs to consider the masking effect of on-demand synthesized melatonin.

Although complex, the role of melatonin on the immune system is now beginning to be understood. MT₁ and MT₂ receptor types appear to play different roles, with MT₁ receptors as the main target in acquired immune response and MT₂ receptors as the target for innate immune responses.

MT receptors in the CNS

MT receptors are widely expressed throughout the CNS and are particularly well characterized in the SCN of the hypothalamus, where they are known to inhibit neuronal firing and mediate the phase shifting effect of melatonin on circadian rhythms (see above). In addition to its chronobiotic effect, melatonin participates in the modulation of neuronal functions, neurodevelopment at early and late stages (Kong et al., 2008; Chen et al., 2014) and affects brain structures underlying sleep regulation (Ochoa-Sanchez et al., 2011), drug-related learning (Wang et al., 2005; Savaskan et al., 2006) and reward (Hutchinson et al., 2012; Clough et al., 2014). MT receptors mediate the melatonin-induced increase in dendrite length, thickness and complexity of hippocampal neurons, as these effects were partially blocked by luzindole (Dominguez-Alonso et al., 2015). Similarly, melatonin-induced differentiation and maturation of adult neural stem cells were almost abolished in the presence of luzindole (de la Fuente Revenga et al., 2015). A recent study using MT₂^−/− mice showed that MT₂ receptors were essential for axogenesis and for the formation of functional synapses (Liu et al., 2015). MT₂ receptors were also involved in melatonin-induced protection against oxidative stress and memory impairment in a mouse model of ageing (Shin et al., 2015). Recent advances in the understanding of presynaptic MT receptors and their role in neurodegenerative diseases are discussed in the following sections.

Melatonin as antioxidant and free radical scavenger

The IUPHAR classifies only clearly identified pharmacological targets in mammals. However, some effects of melatonin persist even in the absence of MT₁ and MT₂ receptors or upon complete pharmacological blockade of MT receptors, indicating the existence of MT receptor-independent mechanisms, which are still not fully understood. In addition, MT receptor-dependent and -independent mechanisms can participate simultaneously, as demonstrated by O'Neal-Moffitt et al. (2015) regarding the antioxidant and pro-cognitive effects of melatonin on AD mice models, for example. Two main mechanisms have been put forward to explain the antioxidant and free radical properties of melatonin: these are melatonin binding to the MT3 binding site (Nosjean et al., 2000; Dubocovich et al., 2003) and to the cytosolic enzyme quinone reductase 2 (QR2) (Nosjean et al., 2000; Dubocovich et al., 2003), and melatonin scavenging of free radicals, as this hormone has been suggested to be an electron donor (see Tan et al., 2015). Binding of melatonin to intracellular targets is readily achieved, due to the hydrophilic nature of this indolamine. Melatonin binds with nanomolar affinity to MT3/QR2 binding sites but shows a pharmacological profile distinct from MT₁ and MT₂ receptors. The order of affinities for the MT3 binding site is 2-iodomelatonin > N-acetyl-serotonin > melatonin (Dubocovich, 1995; Nosjean et al., 2000), the order for MT₁ and MT₂ receptors is 2-iodomelatonin > melatonin >>>>N-acetyl-serotonin. MCA-NAT (5-methoxycarbonylamino-N-acetyltryptamine), prazosin and N-acetyltryptamine are selective ligands for the membrane MT₃ binding site (Dubocovich, 1995; Molinari et al., 1996; Nosjean et al., 2000). Nosjean et al. (2000) showed that a cytosolic binding site identified as QR2 has the pharmacological characteristics of the membrane MT3 binding site. QR2 is a cytosolic flavin adenine dinucleotide (FAD)-dependent flavoprotein that reduces menadione and other quinones by using N-ribosyl- and N-alkyldihydronicotinamides as the co-substrates (Liao and Williams-Ashman, 1961), thus acting as a detoxifying enzyme to increase the antioxidant defence (Jockers et al., 2008). There are still open questions as to whether the melatonin binding site on QR2 corresponds to the MT3 binding site, in particular regarding those sites that are membrane-associated.

Several physiological effects of melatonin such as inhibition of leukocyte adhesion to rat endothelial cell layers were mimicked by MT3 agonists (Lotufo et al., 2001). Similar observations were made for the expression of adhesion molecules by granulocytes (Cernysiov et al., 2015), the increase in dopamine levels in chick retina (Sampaio Lde et al., 2014) and the reduction of intraocular pressure (IOP) in rabbits (Alarma-Estrany et al., 2009). However, it has been questioned whether the functional effects of MCA-NAT are indeed mediated by QR2, as the lack of QR2 did not prevent the MCA-NAT-induced reduction on IOP, and overexpression of QR2 did not promote receptor-like responses (Vincent et al., 2010). In addition, MCA-NAT turned out to be a partial agonist for MT₁ and MT₂ receptors at submicromolar concentrations suggesting the possibility that some of the effects of MCA-NAT might be mediated by MT₁ and/or MT₂ receptors (Vincent et al., 2010).

Melatonin and its metabolites, with or without open ring structures, have been described as potent electron donors. Cyclic-3-hydroxymelatonin, N1-acetyl-5-methoxykynuramine (secondary metabolite) (AMK, tertiary metabolite) and N-acetyl-N-formyl-5-methoxykynuramine (AFMK, quaternary metabolite) scavenge free radicals neutralizing reactive oxygen and nitrogen species (Ressmeyer et al., 2003; Tan et al., 2007; Zavala-Oseguera et al., 2014). Hence, one melatonin molecule and its associated metabolites could scavenge a large number of reactive species, and thus, the overall antioxidant capacity of melatonin is believed to be greater than that of other well-known antioxidants, such as vitamin C and vitamin E, under in vitro or in vivo conditions (Gitto et al., 2001; Sharma and Haldar, 2006; Ortiz et al., 2013). However, the ability of melatonin in reducing oxidative stress does not only rely on donating electrons. Indeed, by acting on MT₁ and MT₂ receptors, low pM and low nM concentrations of melatonin increased the expression or activity of enzymes such as superoxide dismutase, catalase and glutathione peroxidase, which are involved in oxygen detoxification (Rosen et al., 2009). Thus, depending on the dose of exogenous or endogenous melatonin, receptor-dependent or -independent mechanisms may be involved. A further complexity in the interplay between receptor-dependent and -independent processes could arise from the fact that melatonin, by changing the redox state of the cell, might influence receptor-mediated functions. Indeed, the function of several GPCRs has been shown to be sensitive to the cellular redox state. Whether this is also the case for MT receptors has to be addressed in future studies. Although endogenous melatonin levels are typically considered to range from low pM to low nM concentrations, much higher concentrations may be reached locally in the brain (Legros et al., 2014) and in activated immune cells (Conti et al., 2000). In addition, melatonin can be actively taken up through the GLUT1 glucose transporter (Hevia et al., 2015). In conclusion, the role of the MT3 binding site is still not fully understood and warrants further attention. Concerning the free radical scavenging properties of melatonin, it is surprising that opposing opinions are still in the literature. Overall, the antioxidant effects of melatonin appear to be complex, relying on a mixture of MT receptor-dependent and -independent processes.