Pharmacodynamics of the orexin type 1 (OX₁) receptor in colon cancer cell models: A two-sided nature of antagonistic ligands resulting from partial dissociation of Gq

Abstract

Background and Purpose

Orexins have important biological effects on the central and peripheral nervous systems. Their primary ability is to regulate the sleep–wake cycle. Orexins and their antagonists, via OX₁ receptor have been shown to have proapoptotic and antitumor effects on various digestive cancers cell models. We investigated, (1) the ability of orexin-A and its antagonists to regulate OX₁ receptor expression at the cell surface and (2), how OX₁ antagonists induced proapoptotic effect in cancer cells models.

Experimental Approach

The OX₁ receptor internalisation is determined by imaging flow cytometry in colon cancer cell models and the OX₁ receptor coupling to G proteins via bioluminescence resonance energy transfer and molecular dynamic simulation.

Key Results

Orexin-A induced rapid receptor internalisation within 15 min via β-arrestin 2 recruitment, whereas antagonists had no effect. Furthermore, Gq is critical for receptor internalisation and signalling pathways, and no other G proteins appear to be recruited. Surprisingly, antagonists induced recruitment and conformational changes in Gq protein. Simulated molecular dynamics of agonists/orexin receptor/Gq complexes show that antagonists exhibits a similar binding mode, stable at the binding site and show conformational changes of ECL2, similar to that of the agonists.

Conclusion and Implications

OX₁ receptor activation induced orexin/β-arrestin-dependent internalisation, which was independent of the apoptotic pathway induced by orexins and antagonists. In addition, antagonists activate the Gq protein, suggesting its putative partial dissociation. These results suggest that the development of OX₁ receptor targeting molecules, including orexin antagonists with antitumor properties, may pave the way for innovative cancer therapies.

Graphical Abstract

Abbreviations

CHARMM-GUI

Chemistry at Harvard macromolecular mechanics/Graphical User Interface

ECL2

extra cellular loop 2

EEA1

early endosome antigen 1

Gq-CASE

G protein tri-cistronic activity sensors

LAMP1

lysosomal-associated membrane protein 1

MD

molecular dynamic

MEF

mouse embryonic fibroblast

OxA

orexin A

OxB

orexin B

STX6

syntaxin-6

TM

transmembrane domain

YFP

yellow fluorescent protein

What is already known?

• On many digestive cancer cell models, orexins and their antagonists induce apoptosis and antitumor effects.

What does this study add?

• Orexin-A triggered OX₁ receptor β-arrestin 2/Gq-dependent internalisation
• Orexin antagonists mediated apoptosis by Gq conformational changes.

What is the clinical significance?

• Orexin agonists and/or antagonists represent innovative molecules as potential therapeutic agents in cancer treatment.

1 INTRODUCTION

Orexins, also known as hypocretins, include orexin-A (OxA) and orexin-B (OxB), which are two neuropeptides secreted primarily in the central nervous system (CNS), specifically in the hypothalamus (De Lecea et al., 1998; Sakurai et al., 1998). OxA and OxB bind selectively to two orexin receptors known as orexin/hypocretin type 1 (OX₁) receptor and orexin/hypocretin receptor type 2 (OX₂) receptor, which belong to the G protein-coupled receptor (GPCR) superfamily (Soya & Sakurai, 2020; Gotter et al., 2012). The interaction between orexins and their receptors activates the Gq protein, which consists of three subunits (αq and β/γ), leading to the activation of phospholipase C (PLC) by the αq subunit (Couvineau et al., 2019). PLC, among other molecules, catalyses the production of inositol-1,4,5-trisphosphate (IP₃), which triggers the transient release of Ca²⁺ into the cytoplasm (Couvineau et al., 2019). Orexins have significant biological effects that are primarily observed in the CNS, where they promote wakefulness (Jacobson et al., 2022; Wong et al., 2011). They are also involved in drug addiction, food intake, energy homeostasis, motivation and reward-seeking (Couvineau et al., 2018; Tyree et al., 2018). These effects in the CNS are mediated mainly by the Ca²⁺ signalling pathway (Kukkonen & Turunen, 2021). Notably, orexins can activate other signalling pathways, such as cAMP, JNK (c-JUN N-terminal kinase), PI3K (phosphoinositide 3-kinase)-Akt and MAPK (mitogen-activating protein kinase)-ERK1/2 (Couvineau et al., 2022; Kukkonen & Turunen, 2021). Furthermore, orexins regulate various biological effects in the peripheral nervous system, including reproductive functions, metabolism, gastrointestinal motility, neuroendocrine functions, blood pressure and energy balance (Couvineau et al., 2021). Several years ago, our group demonstrated that OX₁ but not OX₂ receptors ectopically expressed in human digestive cancers (colon, liver and pancreas cancers) and nondigestive cancers (prostate cancer) (Alexandre et al., 2014; Couvineau et al., 2019; Dayot et al., 2018; Voisin et al., 2011). More recently, OX₁ receptor expression was also detected in glioblastoma (Yang et al., 2023). The activation of OX₁ receptor by OxA in these cancer cells has been shown to have antitumor effects characterised by mitochondrial apoptosis both in vitro and in vivo (preclinical models) (Couvineau et al., 2019). Our group has also identified a newly discovered signalling pathway induced by OxA and OxB (Couvineau et al., 2019). We found that the OX₁ and OX₂ receptor sequences contain two immunoreceptor tyrosine-based inhibitory motifs (ITIMs) that are phosphorylated by Src kinases in the presence of orexins (El Firar et al., 2009). This phosphorylation leads to the recruitment of tyrosine-protein phosphatase non-receptor type 11 (SHP2) and the activation of the mitogen-stress protein kinase p38 via the RAS/MAPK signalling pathway, resulting in the activation of caspase 3 and 7, ultimately leading to the death of cancer cells. Several OX₁ and OX₂ antagonists, including suvorexant, lemborexant and daridorexant, have been developed for the treatment of insomnia (Couvineau et al., 2022). Surprisingly, suvorexant and almorexant (another antagonist but not used in insomnia treatment) have shown antitumoral effects on pancreatic cancer (Dayot et al., 2018). The following question arises, how can antagonists block the action of orexins in the CNS and induce proapoptotic effects in cancer cells? To answer this question, we analysed the ability of OxA and antagonists (almorexant, lemborexant and daridorexant) to discriminate between these two effects mediated by different signalling pathways. We explored this through their ability to induce receptor internalisation. In fact, the activity of G-protein coupled receptors (GPCRs) at the cell surface is regulated by two main mechanisms, desensitisation and internalisation (Slosky et al., 2021; Wingler & Lefkowitz, 2020). GPCR desensitisation, which results in a loss of affinity for ligands and receptor efficacy or signalling outcome, is induced by the phosphorylation of serine and threonine residues found in GPCR intracellular sequences by G protein kinases (GRKs), protein kinase A (PKA) and/or protein kinase C (PKC). After repeated stimulation by the ligand, the β-arrestin 1 and 2 were recruited. This recruitment leads to the internalisation and/or signalling of the receptors (Slosky et al., 2021; Wingler & Lefkowitz, 2020). During the internalisation process, GPCRs are transported from the cell surface to the intracellular endosomal compartments, where they undergo degradation and/or recycling (Martínez-Morales et al., 2022). In this context, the objective of our study was to determine the impact of OxA and OX₁ antagonists on, (1) the internalisation of OX₁ receptor using an imaging flow cytometry approach in colon cancer cell models and (2), the coupling of OX₁ receptor to G proteins (Gq, Gs and Gi) via bioluminescence resonance energy transfer (BRET) and molecular dynamic simulation. In our study, OxA but not the antagonists induced the internalisation of OX₁ receptor via β-arrestin 2. A surprising finding was that OxA and antagonists, which are prescribed to help sleep (Couvineau et al., 2022) by blocking Ca²⁺ signalling pathway, have antitumor properties (Dayot et al., 2018). In this report, we demonstrate that antagonists are able to discriminate these two signalling pathways involving the Gq protein.

2 METHODS

2.1 Cell line culture, cell growth test, apoptosis analysis and in vitro calcium mobilisation

The HEK293T cell line (ATCC, cat# CRL-3216, RRID: CVCL_0063) and two colon cancer cell lines HT29 (ATCC, Cat# HTB-38, RRID: CVCL_0320) and HCT116 (ATCC, Cat# CCL-247, RRID: CVCL_0291) were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). The mouse embryonic fibroblast (MEF) Q11 cell line invalidated for Gq protein was a gift from Pr. Wilkie (Heidelberg University, Heidelberg, Germany). The MEF β-arr1, β-arr2 and β-arr1–2 cell lines invalidated for β-arrestin 1, β-arrestin 2 or β-arrestin 1 and 2, respectively, were gifts from Pr. R. Lefkowitz (Duke University, Durham, NC, USA). Every recombinant cell line that expressed the wild-type (wt) human OX₁ receptor fused with yellow fluorescent protein (YFP) or with Discosoma red fluorescent protein (DsRed) was grown according to ATCC recommendations as previously described (Dayot et al., 2018; Gratio et al., 2024; Voisin et al., 2011). The 5.10⁴ cells per well in 24-well plates (Corning Costar, New York, NY, USA) were seeded for 24 h before being treated with or without OxA almorexant or lemborexant at 10⁻⁵ M and 10⁻⁶ M as previously described (Dayot et al., 2018). After 48 h of treatment, the cells were dissociated and counted as previously described (Dayot et al., 2018). Results are expressed as the percentage of untreated cells (Dayot et al., 2018). The cells were also immunostained with annexin V-AF647 (Thermo Fisher Scientific, Cat# MA5–36957, RRID: AB_2341149) in binding buffer (Thermo Fisher, Newton Drive, Carlsbad, USA) and the fluorescence was read on a Guava Easycyte Cytometer (Millipore, Darmstadt, Germany) in the green channel for YFP and in the ‘far red’ channel for Annexin V. The 2.10⁵ cells per well were cultured for 24 h and then stained with the FluoForte® Kit according to the manufacturer instructions (Enzo Life Science, Lausen Switzerland) as previously described (Dayot et al., 2018; Gratio et al., 2024) for the intracellular calcium mobilisation test.

2.2 Directed mutagenesis and cell transfection

Directed mutagenesis was performed using the QuickChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA), following the manufacturer instructions to generate the S262A mutation of OX₁ receptor on a 1.3-kb EcoRI/EcoRI cDNA fragment of the human wild type (WT) OX₁ receptor previously cloned into the expression vector pEYFP (Clontech Takara, Kyoto, Japan) in fusion with the yellow fluorescent protein (YFP) (pOX₁ receptor-YFP) as previously described (Dayot et al., 2018). Mutated plasmid DNA purification was performed with a Macherey–Nagel kit (Düren, Germany) according to the manufacturer instructions and the DNA mutation was verified by sequencing (Eurofins, Luxembourg). The HEK293T cell line, MEF cell lines (WT and KO for the Gq/11 protein or for the β-arrestins) and two colon cancer cell lines HT29 and HCT116 were transfected with the OX₁ receptor WT encoding plasmid or with the S262A-OX₁ receptor mutated encoding plasmid using the OptimusJet (Polyplus, Illkirch, France) transfection reagent following the manufacturer's instructions. Transfected cells were selected in the presence of geneticin (0.5 mg ml⁻¹; G418; Thermo Fisher, Newton Drive, CA, USA). After 1 week, the transfected cells were sorted via FACSMelody (Becton Dickinson, Franklin Lakes, NJ, USA) to obtain approximately 80–90% fluorescent cells. For the BRET experiments, 75,000 - HEK-OX₁ receptor DsRed cells were transiently transfected with the different G-case plasmids under the same conditions.

2.3 OX₁ receptor-YFP trafficking analysis, data acquisition and data analysis

The 3.10⁶ cells per well were plated in a six-well plate 24 or 48 h prior to OxA treatment and then incubated with 10⁻⁶ M or 10⁻⁵ M for 2 or 10 min at room temperature (RT). For inhibition assays, cells were treated with 10⁻⁶ M almorexant 1 h before OxA treatment or with 10⁻⁶ M YM-254890 24 h before OxA treatment. For PLC and SHP2 inhibition, cells were treated with 10⁻⁵ M U73122 for 1 h or 10⁻⁵ M RMC-4550 for 24 h prior to OxA treatment. The cells were incubated for 6 h with 10⁻⁵ M gallein before OxA treatment. The cells were incubated in fresh medium in the presence or absence of OxA for 0 min, 15 min, 30 min, 1 h and 2 h, dissociated with 1 ml of TrypLE, washed in phosphate buffered saline (PBS) and fixed with 300 μl of 2% (w/v) paraformaldehyde (PFA) for 15 min at 37°C as previously described. The cell pellet was resuspended in 300 μl of PBS. Then, the cells were immunostained as previously described (Gratio et al., 2024) with 10 ng ml⁻¹ early endosome antigen 1 (EEA1; Santa Cruz Biotechnology, Cat# 365652, RRID: AB_10850311), Leucyl-cysteinyl aminopeptidase (IRAP; Cell Signalling Technology, Cat# 6918, RRID: AB_10860248), STX6 (Proteintech Rosemont, Cat# 10841-1-AP, RRID: AB_2196506), CD71 (Cedarlane, Burlington, Cat# MO-L40042) or 2 ng ml⁻¹ Lamp1 (Merck, Cat# L1418, RRID: AB_477157) antibodies for 30 min at RT. The cells were washed and then incubated with the appropriate secondary antibody anti-IgG (10 ng ml⁻¹; anti-mouse-IgG/Alexa Fluor 405 [Thermo Fisher Scientific, Cat# A31553, RRID: AB-221604] and anti-rabbit-IgG/Alexa Fluor 647 [Thermo Fisher Scientific, Cat# A31573, RRID: AB_2536183]) for 30 min at RT. The cells were resuspended in 40 μl of PBS supplemented with 2% BSA and 2 mM ethylenediaminetetraacetic acid (EDTA).

Data acquisition and data analysis were performed using an ImageStream X MKII (ISX software, Amnis, Seattle, WA, USA) and its dedicated analysis IDEAS® software (version 6.1 Amnis, Seattle, USA) as described previously (Gratio et al., 2024). Briefly, all samples were acquired at 60X magnification. Around 10,000 cells were dissociated into single-cell events and each sample was collected. Data from single stained cells were also acquired to calculate the compensation matrix. All the images are displayed in pseudocolour (AF405 = purple, YFP = green and AF647 = red). The template used in this study was the software-defined ‘colocalisation’ template. To analyse the colocalisation of two fluorescent markers inside the cell, a mask removing the fluorescence of the OX₁ receptor-YFP membrane was generated by developing a morphology (M02) and an erode (M02, 6) function of the YFP channel. These two functions were subsequently combined using Boolean logic (AND) to create a specific mask that quantifies only cytoplasmic colocalisation.

2.4 BRET experiences

Two different luciferase plasmids with Renilla luciferase (Rluc) or NanoLuc (nLuc) were used. The β-arrestin 2 construct was a gift from X. Iturrioz (CEA, Saclay, France). The level of BRET was detected with a luminometer (Tecan; Männedorf, Switzerland). After adding 25 μl of coelenterazine-h (1 mg ml⁻¹), the mixture was recorded for 2 min before the addition of different compounds including OxA, almorexant or lemborexant at different concentrations ranging from 10⁻⁷ M to 2.10⁻⁵ M and then recorded for 3 min. The results are expressed as the ratio between the fluorescence intensities of YFP and coelenterazine-h before ligand addition. G protein tri-cistronic activity sensors (Gq-CASE), Gi1-CASE, Gi2-CASE, Gi3-CASE and Gs (short)-CASE were gifts from Gunnar Schulte (Addgene plasmid # 168125, 168120; 168121, 168122 and 168124).

2.5 Molecular dynamics (MD) simulations of OX₂ receptor with mini-G_sqi and compound 1 or lemborexant

The OX₂ receptor complex structure composed of mini-G_sqi, Gβ₁γ₂ and the small molecule agonist compound 1 was based on the cryo-EM active-state structure 7L1V from Protein Data Bank (PDB; scFv16 and Sb51 were removed – see Figure S1) (Hong et al., 2021). For preparation of the second complex of OX₂ receptor, Maestro (Schrödinger LLC, New York, NY, USA) was used to replace compound 1 in 7L1V by lemborexant oriented in the same way as in the inactive-state structure 7XRR (Asada et al., 2022) and to remove steric clashes and perform the short minimisation of atom positions. PrankWeb (Jakubec et al., 2022) was used to search for other druggable orthosteric or allosteric binding sites in the OX₂ receptor but none were found (Figure S2). Inputs for the MD simulations were prepared using Membrane Builder in CHARMM-GUI (Jo et al., 2008). PPM 3.0 (Lomize et al., 2022). The OX₂ receptor complexes were placed in a lipid bilayer consisting of a 3:1 ratio of POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and cholesterol. A periodic rectangular water box (TIP3P) was used and each system was neutralised by adding Na⁺ and Cl⁻ ions at a concentration of 0.15 M. The number of atoms in the simulations was approximately 200,000 atoms and the Charmm36 force field was used. The equilibration step included 10,000 steps of steepest descent minimisation and 25,000 steps of conjugated gradient minimisation. The equilibration simulation was performed in substance, volume and temperature using Langevin dynamics (303.15 K). The time integration step in the equilibration and production runs was set to 2 fs. The production run in the substance pressure temperature (NPT) was performed using the Langevin piston Nose–Hoover method (1 bar, 303.15 K) and lasted for 2 μs for the 7L1V-based systems including mini-G_sqi. The GPU-accelerated version of Nanoscale Molecular Dynamics (NAMD; Phillips et al., 2020) was used for all MD simulations. Every tenth frame of the simulation trajectories was combined using CatDCD, wrapped in PBCTools to account for the periodic box and analysed via VMD (Humphrey et al., 1996) and PyMOL (Schrödinger LLC, New York, NY, USA). To confirm, the results obtained for OX₂ receptor MD simulations were additionally performed for OX₁ receptor complexes with lemborexant in three replicas.

2.6 MD simulations of OX₁ and OX₂ receptor with orexin-A and -B and mini-G_sqi

The same procedure as above (Modeller, CHARMM-GUI and NAMD) was applied to generate eight simulation systems—four replicas for OX₁ receptor and four replicas for OX₂ receptor. The 7L1U PDB structure was used as a template for OxB-OX2R, with the missing region of OxB (UNIPROT id: O43612) taken from 1CQ0. The Rosetta loop modelling (the CCD algorithm) was used to reconstruct the missing N-terminus of OX₂ receptor (UNIPROT id: O43614) that is known to be essential for peptide binding. Two kinds of models were generated, one with 50 residues-long N-terminus and the other, with a 20 residues-long N-terminus. These Rosetta models were used as templates in building the whole OX₂ receptor–OxB–mini-Gsqi complex by using Modeller (1000 models per complex, discrete optimised protein energy (DOPE) scoring function, top 50 models per complex for verification). In total, four models were selected for MD simulations: two with a full-length N-terminus and two with a truncated N-terminus. These models were used as templates to build OX₁ receptor–OxA–mini-Gsqi complexes with Modeller. Here, OxA was based on the 1R02 PDB template. NAMD simulations lasted 1.5 μs for each eight replicas and trajectories were analysed according to the protocol described earlier.

2.7 Experimental design and data analyses

For each in vitro experiment, the randomisation of cell treatment was carried out. Bar charts did not reveal anything unusual or interesting about the data that was not obvious from the before–after charts. All the statistical analyses were performed using GraphPad Prism 8.0.2 software. All experiments were repeated at least three times and if a large variation between experiments was observed, additional experiments were included. The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2022). Sample sizes of at least n = 5 independent values were subjected to statistical analysis. P < 0.05 is set as the threshold for statistical significance and post hoc tests were conducted if F in the one-way analysis of variance achieved P < 0.05.

2.8 Materials

Almorexant, lemborexant, daridorexant, YM-254890, U73122 and PP2 were purchased from MedChemExpress (Monmouth Junction, NJ 08852, USA), while of OxA was purchased from GL Biochem, (Shanghai, China). Gallein was purchased from BioTechne (Bristol, UK), Details of other materials and suppliers were provided in the specific sections.

2.9 Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in the IUPHAR/BPS Guide to PHARMACOLOGY http://www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2023/24 (Alexander, Fabbro, Davenport, et al., 2023; Alexander, Fabbro, Kelly, et al., 2023).

3 RESULTS

3.1 OxA antagonists (almorexant, lemborexant and daridorexant) inhibited Ca²⁺ signalling pathways in HEK-OX₁ receptor-YFP cells while inducing apoptosis in HEK-OX₁ receptor-YFP cells and colon carcinoma cell lines (HT-116-OX₁ receptor-YFP and HT-29)

We evaluated the inhibitory effects of almorexant, lemborexant and daridorexant, which are part of the DORA (dual orexin receptor antagonist) class, on the inhibition of intracellular Ca²⁺ release induced by OxA and on the growth of colonic cancer cells (HT-29 cell line) and recombinant cells (HCT116-OX₁ receptor-RYFP and HEK-OX₁ receptor-YFP cell lines). In accordance with expectations, 1 μM almorexant (almo), lemborexant (lembo) or daridorexant (darido) inhibited Ca²⁺ release induced by 1 μM OxA, thus showing their ability to inhibit Ca²⁺ signalling pathway (Figure 1a) (Beuckmann et al., 2017; Callander et al., 2013; Treiber et al., 2017). In contrast, 48 h of treatment with 1 μM of these three antagonists inhibited the growth of HT-29, HCT116-OX₁ receptor-YFP and HEK-OX₁ receptor-YFP cells similar to OxA (Figure 1b). Furthermore, incubation of the three antagonists and OxA with HEK-OX₂ receptor-YFP cells yielded similar results to those observed in HEK-OX₁ receptor-YFP cells, indicating that these antagonists also inhibit cell viability in cells expressing OX₂ receptor (Figure 1b). In addition, OxA, almorexant, lemborexant and daridorexant induced apoptosis in HEK-OX₁ receptor-YFP cells, as determined by annexin V fluorescence (Figure 1c). As previously described, Src is involved in the phosphorylation of OX₁ receptor at immunoreceptor tyrosine-based inhibitory motif sites, which leads to OxA-induced apoptosis (Couvineau et al., 2019). Notably, phosphorylated Src and OX₁ receptor colocalised after both OxA or almorexant activation of OX₁ receptor. This colocalisation was completely abolished by the PP2 Src inhibitor (Figure S3), in support of OxA- and almorexant-induced Src recruitment by OX₁ receptor. Taking these findings into account, it can be concluded that orexin antagonists inhibit calcium release signalling while activating apoptosis signalling.

FIGURE 1

Effect of orexin-A (OxA) and antagonists on intracellular Ca²⁺ release, cell viability and apoptosis. (a) Inhibition by 1 μM almorexant (Almo), 1 μM lemborexant (Lembo) and 1 μM daridorexant (Darido) of intracellular Ca²⁺ release induced by 1 μM OxA. Cells were pretreated with1 μM antagonists and then incubated with 1 μM OxA before intracellular Ca²⁺ detection; (b) impact of OxA and antagonists on cell viability determined by cell counting after 48 h of treatment; (c) Proapoptotic effect of OxA and antagonists, determined by annexin V-AF647 immunostaining. Data are shown as means ± SEM, n = 6. * indicates P < 0.05. Abbreviations: OX1RYFP, OX₁ receptor-yellow fluorescent protein; OX2RYFP, OX₂ receptor-yellow fluorescent protein.

3.2 Impact of OxA, almorexant and lemborexant on the OX₁ receptor internalisation process

To investigate OX₁ receptor internalisation in the presence of OxA or antagonists (almorexant or lemborexant), the colonic cancer cell lines HT-29 and HCT116, as well as HEK cells, were transfected with cDNA encoding OX₁ receptor fused to YFP.

After incubating HEK-OX₁ receptor-YFP cells with 1 μM OxA for 0, 30, 60 and 90 min, we observed the presence of green fluorescence corresponding to OX₁ receptor-YFP at the cell surface for 0 min (control) (Figure 2a). In contrast, at 30, 60 and 90 min, OX₁ receptor-YFP was partially located in the cell cytoplasm (Figure 2a). Notably, purple fluorescence is associated with intracellular early endosomes (EEA1), which contain essentially internalised material transferred from endocytic vesicles (Hanyaloglu & von Zastrow, 2008; Jovic et al., 2010; Kamentseva et al., 2020). The combination with these two channels (Figure 2a) revealed that OX₁ receptor-YFP and EEA1 were colocalised. Quantification of the images showed that 30.1%±2.9% of the cells were colocalised with OX₁ receptor-YFP and EEA1 after 30 min of incubation (Figure 1b). This colocalisation lasted 60 to 90 min, during which the percentage of cells decreased to 18.6% ± 2.1% and 16.8% ± 1.9%, respectively (Figure 2a). In contrast, following a 30-min incubation with almorexant or lemborexant, OX₁ receptor-YFP localised to the cell surface (Figure 2a). Quantitative analysis of the images revealed a lack of colocalisation between OX₁ receptor-YFP and EEA1, with percentages similar to those of the control (Figure 2a). Similar results were observed with the colonic cancer cell lines HT29-OX₁ receptor-YFP and HCT116-OX₁ receptor-YFP (Figure 2b and c). Specifically, incubating HT29-OX₁ receptor-YFP or HCT116-OX₁ receptor-YFP cells with OxA for 30 min revealed a colocalisation of OX₁ receptor-YFP with EEA1, measuring at 45.9% ± 2.0% and 33.0% ± 2.2%, respectively (Figure 2b and c). However, this colocalisation decreased after 60 min to 17.9% ± 1.8% for HT29-OX₁ receptor-YFP cells and 24.6% ± 3.4% for HCT116-OX₁ receptor-YFP cells. As observed earlier, in the presence of almorexant, no colocalisation between OX₁ receptor-YFP and EAA1 was detected (Figure 1c and d) in these two cancer cell lines. Importantly, pretreatment with 1 μM almorexant prior to incubation with 1 μM OxA completely eliminated the colocalisation of OX₁ receptor-YFP with EEA1, indicating that almorexant inhibited the internalisation of OX₁ receptor induced by OxA (Figure S4).

FIGURE 2

Effect of orexin-A (OxA) and antagonists on OX₁ receptor (OX1R) trafficking. (a) Colocalisation of OX₁ receptor and early endosomes in the presence of OxA, almorexant and lemborexant in HEK-OX₁ receptor-yellow fluorescent protein (YFP) cells; (b and c) colocalisation of OX₁ receptor and early endosomes in the presence of OxA and almorexant in HT29-OX₁ receptor-YFP and HCT116-OX₁ receptor-YFP cells, respectively. BF, brightfield. Data are shown as means ± SEM, n = 5. * indicates P < 0.05.

3.3 OX₁ receptor was partially recycled to the plasma membrane and partially degraded

To investigate the fate of OX₁ receptors inside the cells following internalisation, we used various biomarkers (Evnouchidou et al., 2020) associated with different cell compartments. These biomarkers included IRAP (storage endosomes), STX6 (trans-Golgi-derived vesicles), CD71 (transferrin receptor with constitutive recycling) and LAMP1 (lysosomes). As shown in Figures 3 and S5, for the colonic cancer cell line, HCT116-OX₁ receptor-YFP was incubated with 1 μM OXA for different durations and analysed by IFC, and colocalisation of OX₁ receptor-YFP with all the other markers was observed from 15 to 120 min. The highest level of colocalisation was observed at 30 min for IRAP, STX6, CD71 and LAMP1 (Figure 3). However, the colocalisation of OX₁ receptor with the indicated markers decreased from 60 to 120 min (Figure 3). In contrast, when HCT116-OX₁ receptor-YFP cells were incubated with 1 μM almorexant for 15 to 120 min, no colocalisation of OX1 receptor with the various markers was detected (Figure 3a–d). These results indicated that OX₁ receptor was detected in various intracellular compartments. Additionally, the same experiments were conducted on HT29-OX₁ receptor-YFP and HEK-OX₁ receptor-YFP cells, yielding similar results, suggesting that approximately half of the OX₁ receptor is recycled back to the cell surface (Figures S6 and S7).

FIGURE 3

OX₁ receptor trafficking in the presence of orexin-A (OxA) and almorexant in HCT116-OX₁ receptor-YFP cells. Colocalisation between OX₁ receptor (OX1R)-yellow fluorescent protein (YFP) and (a) Leucyl-cysteinyl aminopeptidase (IRAP) vesicles; (b) STX6 for trans-Golgi network; (c) transferrin receptor (CD71); and (d) with LAMP1 for lysosomes. Data are shown as means ± SEM, n = 5. * indicates P < 0.05.

3.4 The Gq protein: a key player in OX₁ receptor internalisation and OxA-induced cell death

To explore the involvement of the Gq protein in the process of OX₁ receptor internalisation and OxA-induced cell death, two G protein inhibitors, YM-254890, which blocks Gαq GTP exchange from GDP (Uemura et al., 2006) and gallein, which inhibits β/γ-mediated signalling, were tested (Lehmann et al., 2008). In addition, we used mouse embryonic fibroblast (MEF) cells invalidated for Gq protein and transfected with the OX₁ receptor-YFP construct. HCT116-OX₁ receptor-YFP cells were pretreated with 1 μM YM-254890 and then incubated with 1 μM OxA, which did not result in intracellular Ca²⁺ release, compared to that in the control condition. (Figure 4a). Treatment of HCT116-OX₁ receptor-YFP cells with OxA or almorexant for 48 h inhibited cell growth. In contrast, OxA and almorexant did not exhibit any inhibition of cell growth in the presence of YM-254890 (Figure 4a). Furthermore, treating of HCT116-OX₁ receptor-YFP cells with YM-254890 significantly reduced the colocalisation of OX₁ receptor-YFP with EEA1 induced by OxA by approximately 60% (Figure 4a). These findings indicate that in colonic cancer cells, inhibition of Gαq activity not only suppressed OxA-induced cell death but also impeded OxA antagonist-triggered cell death. In contrast, it only affected the colocalisation of OX₁ receptor and EEA1 induced by OxA. Pretreatment of HCT116-OX₁ receptor-YFP cells with 10 μM gallein, followed by exposure to 1 μM OxA or 1 μM almorexant, demonstrated that gallein inhibited both OxA- and almorexant-induced apoptosis (Figure 4b). As shown in Figure 4b, gallein itself induced 19.54% cell apoptosis in HCT116-OX₁ receptor-YFP cells compared to control cells (8.12%). In contrast, pretreatment of HCT116-OX₁ receptor-YFP cells with gallein prior to OxA incubation did not inhibit OX₁ receptor-YFP and EEA1 colocalisation at 15, 30, 60 and 120 min (Figure 4b). A colocalisation maximum of 24.2% ± 0.7 for OxA and 19.9 ± 3.4 for OxA plus gallein was reached at 30 min. According to the study of MEF Q11 OX₁ receptor-YFP cells (Figure 4c) lacking Gq protein expression, (1) no Ca²⁺ release was observed in the presence of OxA compared to that in the control (WT MEF-OX₁ receptor-YFP cells expressing Gq); (2) in contrast to the control, neither OxA nor almorexant induced apoptosis; and (3) compared to the control, no colocalisation between OX₁ receptor and EAA1 was observed. These experiments indicated that the Gq protein was crucial for the proapoptotic impact of agonist/antagonist compounds and for OX₁ receptor internalisation induced by OxA. Notably, while the U 73122 inhibitor significantly abolished OxA-induced OX₁ receptor internalisation by inhibiting PLC, the RMC 4550 inhibitor had no significant effect on OX₁ receptor internalisation by inhibiting SHP2 (Figure S8).

FIGURE 4

Role of Gq protein in calcium mobilisation, inhibition of cell viability and OX₁ receptor (OX1R) trafficking in the presence of orexin-A (OxA) and almorexant (Almo) (a, left) Effect of YM 254890 on calcium mobilisation induced by OxA; (a, middle) impact of YM 254890 on cell viability and (a, right) OX₁ receptor trafficking in the presence of OxA and almorexant. (b) Impact of gallein (Gal) on cell apoptosis (left) and OX₁ receptor trafficking (right) in the presence of OxA and almorexant. (c) OxA-induced calcium mobilisation in MEF WT-OX₁ receptor-yellow fluorescent protein (YFP) and MEF Q11-OX₁ receptor-YFP cells; effect of OxA and almorexant on cell apoptosis (middle) and OX₁ receptor trafficking (right) in MEF WT-OX₁ receptor-YFP and MEF Q11-OX₁ receptor-YFP cells. Data are shown as means ± SEM, n = 5. * indicates P < 0.05. Abbreviations: EEA1; early endosome antigen 1; NT, no treatment

3.5 The S262A mutation in OX₁ receptor abolished OX₁ receptor trafficking but not apoptosis

A previous report indicated that the OX₁ receptor-S262A mutant prevents the protein from interacting with GRK2 (Cai et al., 2020). Determination of the apoptotic effects of 1 μM OxA and almorexant after 48 h of exposure to colonic cancer HCT116 cells and MEFs WT transfected with either OX₁ receptor-YFP or OX₁ receptor-S262AYFP revealed that the S262A mutation had no impact on the cell apoptosis induced by these two compounds (Figure 5a). Conversely, the S262A mutation strongly inhibited OX₁ receptor/EAA1 colocalisation compared to that in control cells transfected with the native receptor (Figure 5b). Based on these results, it appears that inhibiting the interaction between GRK2 and OX₁ receptor blocked the internalisation of OX₁ receptor but had no effect on the proapoptotic effects of OxA or almorexant.

FIGURE 5

(a) Impact of OX₁ receptor (OX1R)-S262A mutation on cell apoptosis and (b) receptor trafficking in the presence of Orexin-A (OxA) and almorexant in MEF WT-OX₁ receptor-yellow fluorescent protein (YFP) and MEF WT OX₁ receptor-YFP-S262A cells (left) and HCT116-OX₁ receptor-YFP and HCT116-OX₁ receptor-YFP-S262A cells (right). Data are shown as means ± SEM, n = 5. * indicates P < 0.05.

3.6 β-Arrestin 2 plays a major role in OX₁ receptor trafficking

Four arrestin isoforms were identified, including visual arrestins and nonvisual β-arrestin-1 and β-arrestin-2 (Jiang et al., 2022). To investigate the role of β-arrestin-1 and β-arrestin-2 in the endocytosis of OX₁ receptor, WT MEF-OX₁ receptor-YFP cells or MEF-OX₁ receptor-YFP cells lacking β-arrestin-1, β-arrestin-2, or both β-arrestins (Lefkowitz & Whalen, 2004) were used. As shown in Figure 6a, the absence of β-arrestin-1 or β-arrestin-2 had no impact on the ability of OxA or almorexant to induce cell apoptosis, confirming that β-arrestins were not involved in this effect. In contrast, the use of arrestins invalidated cells revealed a strong modification of the OX₁ receptor trafficking (Figure 6b). Moreover, colocalisation of OX₁ receptor with EEA1 was not observed in for β-arrestin-2- and β-arrestin-1/β-arrestin-2-invalidated MEF-OX₁ receptor-YFP cells (Figure 6b). Meanwhile, OX₁ receptor and EEA1 colocalised in β-arrestin-1-invalidated cells compared to control cells (Figure 6b). These results indicated that β-arrestin-2 was crucial for OX₁ receptor trafficking. To examine the interaction between OX₁ receptor and β-arrestin-2, we conducted BRET experiments using OX₁ receptor-YFP and β-arrestin-2-luciferase. HEK293T cells expressing OX₁ receptor-YFP and β-arrestin-2-luciferase were incubated with OxA. This led to the detection of a BRET signal, indicating a physical interaction between OX₁ receptor and β-arrestin-2 (Figure 6c). Conversely, in the presence of almorexant or lemborexant, no BRET signal was observed, suggesting that these OxA antagonists were unable to recruit β-arrestin-2 (Figure 6c).

FIGURE 6

β-arrestin 2 impact on OxA or almorexant-induced cell apoptosis, OX₁ receptor trafficking and recruitment by OX₁ receptor (OX1R). (a) Determination of cell apoptosis induced by OxA and almorexant (almo) in MEF WT OX₁ receptor-yellow fluorescent protein (YFP) and MEF-OX₁ receptor-YFP cells invalidated for β-arrestin 1 (βarr 1^−/−), β-arrestin 2 (βarr 2^−/−) and β-arrestins 1 and 2 (βarr 1&2^−/−); (b) OX₁ receptor trafficking in MEF WT-OX₁ receptor-YFP and MEF-OX₁ receptor-YFP cells in the presence of OxA; (c) BRET determination between β-arrestin 2 coupled to luciferase and OX₁ receptor-YFP in HEK OX₁ receptor-YFP cells in the presence of OxA (left panel) and almorexant or lemborexant (right panel). Data are shown as means ± SEM, n = 5. * indicates P < 0.05.

3.7 OxA and antagonists induced a conformational change of Gq protein

To assess the conformational changes in Gq, we used a BRET-based activity sensor known as Gq-CASE comprising an αq subunit fused to nanoluciferase, along with β/γ subunits fused to the fluorescent protein Venus. As shown in Figure 7a, a dose-dependent inhibition of BRET signal was observed when recombinant HEK293T cells expressing OX₁ receptor and Gq-CASE were incubated with OxA. Surprisingly, a comparable reduction of BRET signal but with different kinetic, which could suggest a partial dissociation, was obtained when cells were incubated with increasing doses of almorexant (Figure 7a). In contrast, this BRET signal was reduced in the presence of the YM-254890 inhibitor when HEK-Gq-CASE cells were incubated in the presence of OxA (Figure 7b). Conversely, the inhibitor YM-254890 did not affect the BRET signal induced by almorexant or lemborexant in HEK293T-Gq-CASE cells (Figure 7b). Notably, no reduction of BRET signal was detected in HEK cells expressing recombinant Gs-CASE, Gi1-CASE, Gi2-CASE or Gi3-CASE in the presence of OxA, almorexant or lemborexant (Figure S9). It should be noted that the transfection of these different G protein constructs had no impact on the ability of OxA or almorexant to inhibit cell growth (Figure S10). Based on these observations, it appears that only Gq physically and functionally interacts with OX₁ receptor. Furthermore, antagonists such as almorexant and lemborexant were able to recruit Gq and modify its structure.

FIGURE 7

BRET signal determination of G protein tri-cistronic activity sensors (Gq-CASE) in HEK-OX₁ receptor (OX1R)-DsRed cells in the presence of OxA and antagonists. (a) Impact of OxA on Gq-CASE BRET signal (left panel) and impact of almorexant (Almo) on Gq-CASE BRET signal (right panel) in HEK-OX₁ receptor-DsRed cells; (b) role of YM 254890 (YM) on BRET signal of Gq-CASE in the presence of OxA (left panel) and almorexant or lemborexant (Lembo), right panel, in HEK-OX₁ receptor DsRed cells. Data are shown as means ± SEM, n = 5.

3.8 MD simulations of the OX₂ receptor–mini-G_sqi protein complexes

The impact of lemborexant on structural changes in the OX₂ receptor–Gq complex during activation was analysed in comparison with compound 1, a potent OX₂ receptor-selective agonist. The three antagonists inhibited the viability of cancer cells and cells expressing either the OX₁ or OX₂ receptor (Figure 1b). These two receptor subtypes can both induce apoptosis (Couvineau et al., 2022), proving the relevance of using the 7L1V structure of OX₂ receptor to compare the effects of lemborexant and compound 1 on Gq. Moreover, our previous 3D model of the OX₁ receptor–OxB complex and the recent cryo-EM structure of OX₂ receptor–OxB were very similar (Couvineau et al., 2022). The sequence identity of the OX₁ and OX₂ receptors is also very high (67%), and lemborexant interacts with both receptor subtypes. This is why, because of the lack of active-state OX₁ receptor structures in PDB, we first chose OX₂ receptor (7L1V) to observe its interactions with lemborexant, only later performing simulations for OX₁ receptor systems. MD simulations were preceded by an analysis of the binding sites by PrankWeb (Jakubec et al., 2022), showing only two already known binding pockets in OX2R (the orthosteric ligand binding site and the intracellular G protein binding site – Figure S2) and no alternative allosteric sites that could explain the unexpected effect of lemborexant. At first, variance in the lemborexant/compound 1 positions in the orthosteric OX₂ receptor binding site was observed (Figure 8a). As an antagonist, lemborexant should be unstable in the active-state structure of the OX₂ receptor, additionally stabilised by mini-Gα (Hong et al., 2021). Surprisingly, no significant changes in the position of either ligand were observed and heavy-atom root mean square deviation (RMSD) values fluctuations were rather insignificant. Similar results confirming stable interactions with ligands were obtained for the lemborexant–OX₁ receptor–G_q simulations (Figure S19). In contrast to the preserved ligand position, ECL2 demonstrated high RMSD fluctuations regardless of the ligand type and the receptor subtype (Figures S20 and S21), leading to its closing above the binding site after over 2 μs. This was preceded by the formation of hydrogen bonds between ECL2 residues and TM4, TM6 and extra cellular loop 1 (OX₂ receptor, 2 μs, Figure 8b) or TM7 (OX₁ receptor, 1 μs, Figure S22), and by the switching of the Y^7.32 rotamer (Figure S23). A detailed comparison of the ECL2 conformations observed by the end of 2 μs MD simulations and those in the active-state (7L1V) and inactive-state (5WQC) OX₂ receptor structures was shown in Figure 8c. The closed position of ECL2 more closely resembled the active-state structure (7L1V). Additionally, the conformational change of the Y^7.32 rotamer during this ECL2 movement resembled the active-like structure of OX2R bound to orexin-B (7L1U). In both the OX₁ and OX₂ receptor simulation systems, the changing of the conformation of the microswitch-like residue Y^7.32 towards the exterior of the receptor prior to the closing of ECL2 was observed during the first microsecond for four out of five simulation replicas (Figure S23). Notably, the subsequent closing of ECL2 took place not earlier than in the second μs, that is in line with the MD simulation standard in terms of timescale (first movements of side chains and then larger loops). During the simulations, we observed that at first Y^7.32 was responsible for stabilising the terminal N,N-dimethylbenzamide of the ligand and thus could be essential for initial ligand binding, but later, the ECL2 closing was enough to keep the ligand in place and no interactions between Y^7.32 and the ligands were observed. The hydrogen bonding network between ECL2 and the rest of the receptor was similar for the OX₁ and OX₂ receptors except for the R/K and E/G substitutions in beta-turn of ECL2, respectively, leading to even better coverage in case of the OX₁ receptor (ECL2 closer to TM7).

FIGURE 8

A comparison of the compound 1 and lemborexant binding modes and their impact on the OX₂ receptor activation. (a) A comparison of the binding modes of compound 1 (left panel) and lemborexant (right panel). Receptor conformations from the first (grey receptor, orange ligand) and last (green receptor, red ligand) frame of the mini-Gq MD simulations were superimposed. Residues interacting with compound 1 that were described in Hong et al. (2021)) were marked. Polar interactions detected in PyMOL were shown as yellow dashed lines. (b) A comparison of the closing movement of extra cellular loop 2 (ECL2) for compound 1 (left panel) and lemborexant (right panel). Receptor conformations from the first (grey), middle (1 μs, blue) and last frame (2 μs, green receptor, red ligand) of the MD simulations were superimposed. The closing movements of the ECL2 loop were indicated by red arrows and refer to the largest peaks in the RMSD plots. (Figure S20). The hydrogen bonding network connecting ECL2 with TM4, TM6 and extra cellular loop 1 (ECL1) together with the Y^7.53 rotamer switching facilitates the ECL2 closing. (c) A comparison of ECL2 conformations in the inactive-state OX₂ receptor structure 5WQC (Suno et al., 2018) (grey), the active-state OX₂ receptor structure 7L1V (Hong et al., 2021) (blue) and the complex from the final frame of the compound 1 including simulation (green, left) or the lemborexant including simulation (green, right). (d) The same changes of the ionic lock involving the ‘DRW’ motif and the tyrosine Y^7.53 toggle switches during the OX₂ receptor activation induced by compound 1 and lemborexant. Here, the inactive-state OX₂ receptor structure 7XRR was shown in grey (with polar contacts shown as yellow dashed lines), whereas the active-state OX₂ receptor conformations stabilised by compound 1 and lemborexant were shown in green (with polar contacts shown as red dashed lines).

Both lemborexant and compound 1 in the OX₂ receptor systems were bound near Q^3.32, T^2.61, H^7.39 and Y^7.43 (Figure 8a), all of which were determined to be important for agonist binding (Hong et al., 2021). Interactions between the ligands and the equivalent of the first three residues (Q^3.32, S^2.61, H^7.39) were additionally detected in the lemborexant–OX₁ receptor–G_q systems (Figures S24 and S25). The formation of hydrogen bonds between the ligands and Q^3.32, H^7.39 and T^2.61 was shown in Figure S11. H^7.39 was described to stabilise compound 1 (Hong et al., 2021) and also contributed to the binding of lemborexant through π–π stacking with its pyrimidine ring, which similarly interacts with the pyridine ring, which stacks further with P^3.29 (Asada et al., 2022). OxB was shown to form hydrogen bonds with Q^3.32, N^6.55, H^7.39 and Y^7.43 in 7L1U, while interactions between Q^3.32 and H^7.39 and compound 1 were shown in 7L1V (Hong et al., 2022). H^7.39 directly interacts with agonists but in the case of antagonists, this interaction is water mediated (Hong et al., 2021). The outwards-facing Q^3.32 rotamer was suggested to be strictly associated with ligand type (agonist) based on 1-μs simulations by Hong et al. (2021), but it has not been confirmed either in 3-μs simulations by Karhu et al. (Karhu et al., 2019) or in any of our microsecond simulations of OX₁ and OX₂ receptors. Asada et al., describes Q^3.32 as important for lemborexant binding to the inactive-state OX₂ receptor (PDB ID 7XRR) (Asada et al., 2022) but not for suvorexant (Hong et al., 2021). The T/A^3.33, but not the T/A^2.61, mutation has also been reported as important for the lemborexant binding to the inactive-state OX₂ receptor. As shown in Figures 7a and S11, the lemborexant binding mode after 2 μs resembled the agonist-like binding modes, for example, by interactions with T^2.61, H^7.39 and Q^3.32.

The final Q^3.32 rotamer observed in 2-μs simulations of the lemborexant-including systems was in its inwards-facing state rather in than the outwards-facing state as described by Hong et al. (Hong et al., 2021). Moreover, the same inwards-facing Q^3.32 conformation was observed for the compound 1 simulations and Karhu et al. (2019) simulations of OxA-OX₂ receptor. Noteworthily, this inwards-facing rotamer of Q^3.32 was also observed in our simulations of OxA-OX₁ receptor, OxB-OX₂ receptor, lemborexant-OX₁/OX₂ receptors and compound 1-OX₂ receptor-Gq (Figures S12–S14 and S25). In addition, receptors were in their active-state throughout the simulations (lemborexant/compound 1-OX₂ receptor-G_q, OxA-OX₁ receptor-G_q, OxB-OX₂ receptor-G_q; lemborexant/-OX₁ receptor-Gq (Figures S15, S16 and S26). Namely, residues of the ‘DRW’ motif of OX_1/2 receptors (corresponding to the ‘DRY’ motif in other GPCRs) remained in their active-like conformation throughout the 2-μs simulations regardless the ligand type as compared with the inactive-state 7XRR structure (Figure 8d). In 7XRR, R^3.50 participates in a network of polar contacts involving T^2.39, D^3.49 and S^4.37, whereas in our case, it interacts with Y^5.58 and Y^7.53. The observed hydrogen bond between R^3.50 and Y^5.58 in the compound 1 and lemborexant-including systems (Figure S17) is expected to occur in an active-state conformation (Hong et al., 2021). Interactions between R^3.50 and T^2.39, typical for the inactive receptor state (Asada et al., 2022), were non-existent either in the lemborexant or compound 1-including simulation systems, and they were also not observed for the OxA- and OxB-including systems. Moreover, the Gα residue F341^G.H5.08, described by Ham et al. (Ham et al., 2021) as a microswitch that indicates the activation of Gα, was in its active ‘on’ state throughout the simulations, regardless of whether the ligand was compound 1, lemborexant, orexin A or orexin B, demonstrating that lemborexant was able to activate or more likely partially activate Gα as a result of its partial dissociation (Figure S18).

4 DISCUSSION AND CONCLUSION

Previous studies have shown that OX₁ receptor is expressed in cancers of the digestive tract such as colon, pancreatic and liver cancer (Couvineau et al., 2019), and in cancers of the nondigestive tract such as prostate cancer and glioblastoma (Alexandre et al., 2014; Yang et al., 2023). However, OX₁ receptor was not expressed in corresponding healthy tissues (Voisin et al., 2011). Activation of OX₁ receptor by OxA in cancer cells results in the induction of two independent but Gq protein-dependent signalling pathways. (1) the canonical signalling pathway leads to the production of intracellular calcium by activating PLC through the αq G-protein subunit and (2), the proapoptotic signalling pathway involves the phosphorylation of immunoreceptor tyrosine-based inhibitory motif (ITIM) sites, present in the OX₁ receptor sequence, by Src kinase activated by β/γ subunits, which induces the recruitment of the tyrosine phosphatase SHP2 leading to the activation of caspase 3 and 7 (Couvineau et al., 2019). Triggering the proapoptotic action of OxA on OX₁ receptor was shown to be responsible for the antitumor effect of orexins in preclinical models generated from cancer cell lines or patient-derived xenograft (PDX) models (Dayot et al., 2018; Voisin et al., 2011). Surprisingly, DORA molecules, such as almorexant, suvorexant, lemborexant and more recently daridorexant, completely abolished the canonical OX₁ receptor pathway but can activate the receptor proapoptotic pathway (Dayot et al., 2018). These observations suggested that OxA and/or DORA may be potential targets for cancer therapy. However, for the use of such molecules in cancer therapy, the availability of OX₁ receptor at the cancer cell surface is crucial.

The regulation of GPCRs in terms of activity and expression on the cell surface is modulated by various cellular mechanisms (Martínez-Morales et al., 2022). The expression of GPCRs at the cell surface is highly regulated by two concomitant mechanisms, desensitisation and internalisation (Martínez-Morales et al., 2022). Agonist-ligand interactions with GPCRs involve conformational changes that trigger G-protein-mediated signalling through G-protein heterotrimer dissociation and GDP–GTP exchange (Wingler & Lefkowitz, 2020). One of the consequences of these conformational changes is the unmasking of serine and/or threonine residues present in intracellular loops and C-terminal GPCR domains that are phosphorylated by GPCR kinases (Martínez-Morales et al., 2022) or other kinases (Casas-González & García-Sáinz, 2006). The phosphorylation of GPCRs has several advantages in terms of receptor regulation, including a loss of GPCR affinity for ligands, the inhibition of the agonistic response called desensitisation and the initiation of the internalisation process by the recruitment of β-arrestins (Martínez-Morales et al., 2022). GPCR endocytosis is essential for its dephosphorylation and recycling to the plasma membrane, providing a new pool of fully functional receptors (Bahouth & Nooh, 2017). It should be noted that during the internalisation process, different GPCRs in endosomes may have signalling and functional effects (Bahouth & Nooh, 2017). The internalisation of orexin receptors has been little studied, with only a few reports showing that OX₁ receptor is internalised in response to agonists (Cai et al., 2020; Ward et al., 2011). Additionally, only one report indicated that OX₂ receptor was internalised (Lesiak et al., 2020). The use of IFC has demonstrated that OX₁ receptor undergoes rapid internalisation in 15 min, primarily within early endosomes. Additionally, after 30 min, OX₁ receptor can also be found in other compartments, including storage endosomes, the trans-Golgi network and lysosomes. Our results showed that after internalisation, OX₁ receptor was transported through the trans-Golgi network and storage vesicles to the plasma membrane for recycling. In parallel, a fraction of OX₁ receptor was present in lysosomes, an indication of its degradation. We estimated that approximately 30% of the OX₁ receptor was internalised in each cycle, 50% of the OX₁ receptor was recycled and 50% was degraded. This process was dependent on the αq subunit of the Gq protein, but not on the β/γ subunit and on β-arrestin 2, but not on β-arrestin 1. It is clearly accepted that the desensitisation/internalisation of GPCRs induces decoupling between the receptor and the G protein (Martínez-Morales et al., 2022), leading to a minimal role of the G protein in the internalisation process. However, it has been shown that some GPCRs such as thromboxane A₂ (TP) receptor (Rochdi & Parent., 2003), vasopressin type 2 (V₂) receptor (Daly et al., 2023) and calcium-sensing (CaS) receptor (Mos et al., 2019) are dependent on the G protein for internalisation. However, it has also been shown that intracellular calcium produced in the cell by GPCRs could play a role in the internalisation process (Horinouchi et al., 2019). In this way, OX₁ receptor activation by OxA which produces intracellular calcium via the Gq protein and PLC could explain the role of the αq subunit in OX₁ receptor internalisation. In addition, our observation indicated that inhibition of PLC partially reduced OxA-induced OX₁ receptor internalisation. In contrast, the inhibition of SHP2 had no effect on OX₁ receptor internalisation but inhibited the apoptosis induced by OxA or almorexant, demonstrating the independence of OX₁ receptor internalisation from the proapoptotic signalling pathway. Note that neither β-arrestin 1 nor β-arrestin 2 have any effect on OxA- or almorexant-induced apoptosis, clearly demonstrating that β-arrestins have no role in this pathway. In contrast, antagonist ligands such as almorexant or lemborexant were unable to induce OX₁ receptor internalisation, although they were able to induce cell apoptosis (see Figure 1). This observation was not trivial. Several reports have indicated that antagonists, at least when given as such, are able to induce internalisation of GPCRs such as cholecystokinin (CCK) receptor (Roettger et al., 1997), V₂ receptor (Pfeiffer et al., 1998), glucagon receptor (Sachdev et al., 2009), CCR4 receptor (Sato et al., 2013) and NPY (Y₁) receptor (Pheng et al., 2003). Moreover, our results showed that the proapoptotic effects of OxA and its antagonists such as almorexant were dependent on the αq subunit and the ability to induce GDP–GTP exchange by the use of the YM-254890 inhibitor. In addition, the proapoptotic effects of OxA and almorexant were inhibited by the use of gallein, an inhibitor of β/γ subunit signalling. These critical observations strongly indicate that both the agonist OxA and the antagonist almorexant rely on Gq protein activation to induce apoptosis in cancer cells. Furthermore, these findings suggest that the underlying molecular mechanisms involved are similar. In the past, several publications have reported that orexin receptors can be coupled to other G proteins such as Gs or Gi (Couvineau et al., 2019; Kukkonen & Turunen., 2021). This putative coupling of OX₁ receptor to other G-proteins was investigated in a recombinant system using the BRET approach with G-protein constructs called G-case (Schihada et al., 2021), in which the α-subunit was coupled to luciferase and the γ-subunit was tagged with Venus. Our results confirmed that OX₁ receptor was functionally coupled to the Gq protein in the recombinant system but was not coupled to Gs or Gi (Gi1, Gi2 or Gi3), showing that only Gq was able to transduce the ligand signal into the cell. As mentioned earlier the GDP–GTP exchange inhibitor, YM-254890 inhibited the proapoptotic effect of OxA and almorexant. In contrast YM-254890 partially inhibited the functional interaction between OxA/OX₁ receptor and the Gq protein but not between almorexant/OX₁ receptor and the Gq protein. However, between the OxA and antagonist treatments, the dissociation kinetic curves of the Gq protein differed (see Figure 6). In the presence of OxA, the Gq protein dissociated rapidly between 0 and 13 s, whereas in the presence of almorexant or lemborexant, Gq dissociated after 15 s. These differences in kinetics could explain the difference in the effect of YM-254890 on the GDP–GTP exchange of the Gq protein between the two types of ligands. Furthermore, an analysis of the interactions between lemborexant and OX₂ receptor (or OX₁ receptor) observed in MD simulations confirmed that this compound interacts with the same residues known agonists, for example, T^2.61 (Asada et al., 2022), when bound to the active-state receptor. It should be noted that ligands form similar interactions with OX₁ receptor which is always in the active state. The ligand stabilising effect of the closing of ECL2 was similar for the OX₂ receptor selective agonist compound 1 and lemborexant, suggesting further similarities in their mode of action. Furthermore, the analysis of the DRW motif and the tyrosine toggle switch confirmed that lemborexant behaved as an agonist at the end of the 2-μs simulations. The canonical and dogmatic scheme of activated G protein is based on the full dissociation model of the α and β/γ subunits (Chung & Wong, 2021). Nevertheless, some evidence indicates the existence of a partial dissociation model between these subunits (Chung & Wong, 2021). Using BRET or FRET technologies (fluorescence resonance energy transfer), it has been demonstrated that activation of G protein may result in partial dissociation (Chung & Wong, 2021). This partial dissociation model may involve sliding of the β/γ subunits along the α subunit, or a clamshell model in which the G protein opens allowing access to the β/γ dimer to its effector (Galés et al., 2006). MD simulations and BRET experiments revealed that lemborexant activates the Gq protein and could induce its partial dissociation, which in turn could explain its ability to induce the apoptosis signalling pathway while inhibiting the Ca²⁺ signalling pathway. This aspect suggests that molecules that are able to activate partially Gq protein through OX₁ receptor leading to the discrimination of two signalling pathways may provide a basis for developing new clinical therapies.

In conclusion, OX₁ receptor activation induced orexin/β-arrestin-dependent internalisation that was independent of the apoptotic pathway induced by orexins and antagonists. Furthermore, antagonists such as almorexant and lemborexant were able to activate the Gq protein suggesting its partial dissociation, inducing apoptosis but not intracellular calcium production. There is a strong evidence that these antagonists prescribed for insomnia are not complete antagonists but partial agonists that are capable of discriminating between two signalling pathways. In this context, the development of innovative OX₁ receptor -targeting molecules specific for the apoptosis pathway paves the way for a new generation of cancer therapies having clinical interest.

AUTHOR CONTRIBUTIONS

All authors contributed to the study conception and design. Material preparation and analysis were performed by Valérie Gratio, Paulina Dragan, Laurine Garcia, Pascal Nicole, Thierry Voisin and Dorota Latek. The first draft of the manuscript was written by Valérie Gratio and Alain Couvineau. All authors read and approved the final manuscript.

ACKNOWLEDGEMENTS

We are grateful to Pr. R. Lefkowitz (Duke University, Durham, NC, USA) for providing the β-arrestin KO MEF cell lines, Pr. Wilkie and Pr. Offermanns (Institute of Pharmacology, University of Heidelberg) for providing the Gq/11 KO MEF cell line and Dr. X. Iturrioz (SIMOS, CEA, Saclay, France). We also thank M. Dumont for her technical assistance. Computational resources were provided by PLGrid (PLG/2023/016255) and LUMI (PLL/2023/04/016464).

CONFLICT OF INTEREST STATEMENT

The authors do not have any conflicts of interest.

DECLARATION OF TRANSPARENCY AND SCIENTIFIC RIGOUR

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Design and Analysis and as recommended by funding agencies, publishers and other organisations engaged with supporting research.

Open Research

DATA AVAILABILITY STATEMENT

The data are available upon request from the authors as [email protected] for cellular experiments. Molecular dynamic simulation data are available on: https://zenodo.org/records/13963128 Zenodo, DOI: 10.5281/zenodo.13963128, https://zenodo.org/records/13963636 and DOI: 10.5281/zenodo.13963635