International Union of Basic and Clinical Pharmacology CXIV: Orexin Receptor Function, Nomenclature and Pharmacology

A. Primary Signal Transducers of Orexin Receptors

B. Downstream Signal Transducers of Orexin Receptors

1. Phospholipase C

PLC practically refers to the phosphoinositide-specific phospholipase C, and is used in this meaning here as well. It comes in many isoforms stimulated by different factors (Bunney and Katan, 2011; Kadamur and Ross, 2013; Katan and Cockcroft, 2020). PLCβ is classically associated with GPCR signaling along the G_q subfamily pathway (Fig. 3, A and D), but GPCRs can, more or less directly, regulate any other PLC isoform as well (Bunney and Katan, 2011; Kadamur and Ross, 2013; Katan and Cockcroft, 2020). Upon hydrolysis of phosphatidylinositol-4,5-bisphosphate (PIP₂) by PLC, the product is inositol-1,4,5-trisphosphate (IP₃) − PIP₂ + H₂O → IP₃ + diacylglycerol (DAG) (Fig. 3, A and D) − but lower phosphorylated species inositol monophosphate and bisphosphate, of unknown physiological role, are released when the substrate is phosphatidylinositol or phosphatidylinositol monophosphate, respectively (Fig. 3, A and D) (Katan and Cockcroft, 2020). Even though PIP₂ is often referred to as the preferred PLC substrate (Katan and Cockcroft, 2020), no evidence is presented; in contrast, there are definite findings showing significant PLC activity on phosphatidylinositol and phosphatidylinositol monophosphate as well as high cellular levels of especially phosphatidylinositol (Harden et al., 1987; Homma et al., 1988; Gandhi et al., 1990; Zhou et al., 1993; Hwang et al., 1996; Zhang et al., 2013; Johansson et al., 2008). Endo-/sarcoplasmic reticulum IP₃ receptor channels open in response to IP₃ and thus give rise to intracellular Ca²⁺ elevation, which is usually followed by so-called capacitative Ca²⁺ influx, but Ca²⁺ elevation upon GPCR stimulation can also be achieved by other means (II.B.2.).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3

Mammalian orexin receptor signaling. The figure aims to summarize the more immediate actions of orexin receptor stimulation combining the knowledge from both neuronal and non-neuronal systems. Both well- and less-proven signaling is shown. (A) The potential phospholipase activations as mapped for human orexin receptors in CHO-K1 cells and partially verified in rodents ex vivo. Please note that the studies have been performed in detail with OX₁ and verified to a limited degree for OX₂. The faded parts connecting the different phospholipase ”trees" are hypothesized based on known metabolism of lipid-derived compounds. All free fatty acids (FFA) are not equal: For instance, cPLA₂ is more likely to release a polyunsaturated fatty acid with higher signaling potential. Even other components are not necessarily equal: For instance, LPA produced via PLA₁ and PLA₂ are sn2-LPA and sn1-LPA, respectively. The precise pathways to the activation of phospholipases are not known although G_q is implicated based on the effect of the inhibitor UBO-QIC/FR900359. Abbreviations: DAGK, DAG kinase; IP₁ and IP₂, inositol mono- and bisphosphates, respectively; PAP, phosphatidic acid phosphohydrolase; otherwise, they are the same as in the text. Star-shapes indicate molecules potentially active as messengers. (B) The potential Ca²⁺-elevating reverse mode action of NCX. Upon opening of NSCCs with low Ca²⁺-permeability, Na⁺ flows into the cells. Depolarization and the local elevation of [Na⁺] change the energetics of NCX to favor reverse mode action. (C) It is also possible that NSCCs even allow Ca²⁺ influx which activates NCX in normal mode, and thus both NSCCs and NCX contribute to the depolarization. It is also possible the actions of NSCC and NCX are not connected. (D) A map of potential orexin receptor responses. Star-shapes indicate molecules/ions active as messengers. Green and red arrows indicate stimulation and inhibition, respectively, and solid lines known/assumed cascades and dashed lines unclear intermediates or completely unknown signals. Activation of G_q leads to stimulation of a) cPLA₂, which produces at least AA as a messenger, b) PLD1, which produces PA, and c) PLC, which produces DAG and IP₃. IP₃ activates endoplasmic reticulum IP₃ receptors (IP₃R) leading to Ca²⁺ release, which then gives additional store-operated Ca²⁺ influx (not shown). DAG can act as a substrate for DAGL, which releases 2-AG. DAG also activates, among others, PKC, which here stimulates AC and PLD and activates or inhibits ion channels [NSCCs as well as VGCCs and VGKCs (voltage-gated Ca²⁺and K⁺ channels, respectively)]. In addition to PKC, AC is stimulated by G_s and inhibited by G_i; G_q also seems to somehow act together with G_i in AC inhibition. cAMP elevation activates PKA, which inhibits VGCCs and, among NSCCs, hyperpolarization-activated cyclic nucleotide-gated channels, while cyclic nucleotides themselves stimulate the latter (not shown). However, the NSCCs usually regulated by orexin receptors are of unknown type as also are the signal cascades involved. NCX may be regulated “passively” as in (B) and (C) or via an unknown signal. Several different types of VGCCs are regulated leading to different types of effects, and there appear to be other messages than just PKC phosphorylation. Cellular Ca²⁺ dynamics can be regulated by orexin receptors in multiple ways: In addition to the Ca²⁺ release by IP₃, Ca²⁺ influx can come from store-operated Ca²⁺ channels (not shown), NSCCs, NCX (reverse mode) and VGCCs, potentially passively regulated by the membrane potential but also via orexin receptor signals.

DAG, the other hydrolysis product of phosphoinositides, has plenty of targets (Carrasco and Merida, 2007) and it can also act as a source of other second messengers, i.e., phosphatidic acid (PA; II.B.5.), and a free fatty acid and 2-monoacylglycerol (MAG; see below) upon hydrolysis by diacylglycerol lipase (DAGL) (Fig. 3, A and D) (Merida et al., 2008; Fowler et al., 2017).

Orexin receptor signaling has been strongly associated with the G_q subfamily → Gα_q → PLCβ → IP₃ → Ca²⁺ release. This is deduced from the strong coupling of both human orexin receptor subtypes to Ca²⁺ elevation and PLC activation upon recombinant expression (II.A.1.), but has seldom been assessed in native cells. Since there are other ways for GPCRs to elevate intracellular Ca²⁺ concentration in addition to this pathway, Ca²⁺ elevation per se does not allow extrapolation of possible upstream pathways. The poor efficacy and selectivity and well-known toxicity of the small molecule PLC inhibitors (for instance, the most commonly used one, U73122; Taylor and Broad, 1998) preclude conclusions based on their effect on orexin signaling. In recombinant human OX₁- and OX₂-expressing CHO-K1 cells, the G_q/11/14 inhibitor UBO-QIC/FR900359 fully blocks orexin receptor simulation-induced PLC activity (Kukkonen, 2016a). PLC activity upon orexin receptor activation is also supported by the effect of protein kinase C (PKC) inhibitors on some responses in native neurons (see Leonard and Kukkonen, 2014, especially Supporting Information Table S1), though a) their selectivity is limited (Davies et al., 2000; Wu-Zhang and Newton, 2013), b) there are other sources for DAG (see below), and c) PKC can also be activated by other means (Zeng et al., 2012).

Bacteria express phosphatidylcholine-specific PLC and a similar activity has been suggested to be present in mammals, however without success in identifying the enzyme. A synthetic compound, D609 [O-(octahydro-4,7-methano-1H-inden-5-yl) carbonopotassium dithioate], has been suggested to inhibit this PLC, and has been often used to implicate involvement of the enzyme. D609 has also been used in orexin research (reviewed in Kukkonen, 2013). Sphingomyelin synthase classically catalyzes the reaction ceramide + phosphatidylcholine ↔ sphingomyelin + DAG. Sphingomyelin synthase is inhibited by D609, which has given one reasonable explanation to the reported mammalian “phosphatidylcholine-specific PLC activity” (Luberto and Hannun, 1998). Furthermore, both known mammalian sphingomyelin synthase enzymes (SMS1 and -2) were recently shown to exhibit phosphatidylcholine-specific PLC activity, i.e., phosphatidylcholine + H₂O → phosphocholine + DAG (Chiang et al., 2021b), which is not surprising since mammalian phosphatidylethanolamine-specific PLC is also related to sphingomyelin synthase (Chiang et al., 2021a). This offers a satisfactory explanation to the issue for the time being. GPCRs, including orexin receptors, have also been suggested to signal via phosphatidylcholine-specific PLC based on the inhibitory effect of D609 (reviewed in Kukkonen, 2013). However, the regulation of sphingomyelin synthase remains undetermined, and no known domains for regulation are found in UniProt. Thus, the molecular mechanism of potential regulation of sphingomyelin synthase by orexin receptors (or any GPCR) remains unresolved. Sphingomyelin synthase 2 would be the major candidate for GPCR regulation as it is mainly found in the plasma membrane (Chen and Cao, 2017).

Both human orexin receptor subtypes couple to PLC activation in recombinant cell lines, as demonstrated by robust elevation of total inositol phosphates (and IP₃ and DAG, as measured occasionally) (Lund et al., 2000; Holmqvist et al., 2002; Johansson et al., 2007; Johansson et al., 2008; Tang et al., 2008; Putula and Kukkonen, 2012). The exact type of PLC is not clear; a study with human OX₁ receptors in CHO-K1 cells implicates two different PLC activities with specificities for different phosphoinositides (Johansson et al., 2008). DAG generated upon the PLC activity in these cells further leads to 2-arachidonoyl glycerol (2-AG) production – upon apparent DAGL activity – as well as arachidonic acid (AA) release from 2-AG hydrolysis (Turunen et al., 2012). 2-AG exits CHO-K1 cells and acts as an endocannabinoid (Jäntti et al., 2013). PLC activity upon orexin receptor stimulation has been directly measured in a few native cells and cell lines (Mazzocchi et al., 2001a; Randeva et al., 2001; Karteris et al., 2004; Karteris et al., 2005; Wenzel et al., 2009) and 2-AG production upon orexin receptor activation has been implicated in several brain regions (see II.B.6.).

C. Regulation of Electrical Activity

Orexin receptor activation was found early on to be excitatory to native CNS neurons (de Lecea et al., 1998; reviewed in Kukkonen and Leonard, 2014; Leonard and Kukkonen, 2014). This is usually accomplished by postsynaptic depolarization mediated by K⁺ channel closure (Horvath et al., 1999b; Ivanov and Aston-Jones, 2000; Hwang et al., 2001; Bayer et al., 2002; reviewed in Kukkonen and Leonard, 2014; Leonard and Kukkonen, 2014) or activation of NSCCs or NCX in reverse mode (Eriksson et al., 2001; Brown et al., 2002; Burlet et al., 2002; Liu et al., 2002; Wu et al., 2002; Yang and Ferguson, 2002; Burdakov et al., 2003; Yang and Ferguson, 2003; reviewed in Kukkonen and Leonard, 2014; Leonard and Kukkonen, 2014). It is rarely possible to distinguish between NSCC and NCX by the methods used (II.B.2.). It should be noted that the membrane conductance-decreasing (the first) and -increasing (the two latter) actions produce different effects on the overall excitability of the neurons (Schöne and Burdakov, 2017). Tandem-pore K⁺ family channels (e.g., TASK1/3; Doroshenko and Renaud, 2009) or inward rectifier channels (K_IR or more specifically here GPCR-regulated K_IR, GIRK) are likely candidates for the PKC-mediated inhibition and the latter are also inhibited by reduced PIP₂; however, the channel types have usually not been directly identified. The precise types of NSCCs and NCXs are usually also not known, nor are the signals that regulate them. One study suggests TRPC4/5 channels as the NSCCs involved (Kolaj et al., 2014), but no other study has attempted to identify them. During recent years, more selective TRP channel inhibitors have become available (Bon et al., 2021; Koivisto et al., 2021).

In recombinant CHO-K1, HEK293 and IMR-32 cells, involvement of TRPC channels (Larsson et al., 2005; Näsman et al., 2006; Peltonen et al., 2009; Louhivuori et al., 2010) as well as NCX (Louhivuori et al., 2010) in OX₁ responses has been suggested, but the regulatory mechanisms may vary in different cell types (II.B.2.). Lipid signaling cascades are one potential regulatory mechanism for NSCCs (Kukkonen, 2011; Kukkonen, 2014), as also discussed and examined in a few original studies; the engagement of orexin receptors in the regulation of these cascades (PLC, PLD, PLA₂, DAGL) has been discussed in the context of the experimental findings above. The regulation of NCX is largely unknown. As indicated (II.B.2.), the forward mode action (Na⁺ in, Ca²⁺ out) that is depolarizing can be precipitated by Ca²⁺ influx by any means (Fig. 3C), while the reverse mode action (Na⁺ out, Ca²⁺ in) could come from Na⁺ influx (Fig. 3B), but the other forces to consider are the Na⁺ and Ca²⁺ gradients and the membrane potential; the direction of NCX action is principally in energetic downstream direction along the net force determined by the relevant two ionic gradients and the membrane potential (Harper and Sage, 2016; Khananshvili, 2022). The number of Na⁺ ions transported can vary, and therefore this factor also affects the energetics of the transport (Khananshvili, 2022). The cell surface levels of the transporters and the velocity of the reaction and even its directionality may be additionally regulated by Na⁺ and Ca²⁺ themselves, phosphorylation, proteolysis and interactions with other proteins or other molecules or ions etc. (Harper and Sage, 2016; Khananshvili, 2022). Of the known orexin receptor-regulated signaling components, PKA, PKC and PIP₂ have been clearly identified. It should be noted that, in addition to NCX, other members of the Ca²⁺/cation antiporter family can also take part in the cation transport (Khananshvili, 2022). Finally, even voltage-gated Ca²⁺ channels have been implicated in the depolarizing action of orexins (Chrobok et al., 2017).

Neuronal effects of orexin receptor signaling are not limited to simple excitation. Orexin receptor stimulation can affect (reduce or enhance) afterhyperpolarization mediated by different types of K⁺ channels [putatively K_Ca2/K_Ca3.1 (SK), K_Na1 (Slick/Slack), K_v4.3 (A) and unidentified types; faster or slower activation of the channel, altered open time] which may affect action potentials and other signals in multiple ways, e.g., by increasing the refractory time, changing the pattern of impulse firing, enhancing Ca²⁺ influx etc. (Zhang et al., 2010; Kolaj et al., 2014; Ishibashi et al., 2016; reviewed in Leonard and Kukkonen, 2014). The mechanisms used by orexin receptors to regulate these channels are putatively Ca²⁺ as well as PKA and PKC activities, but in most cases they are not known. PIP₂ levels, affected directly by PLC and indirectly by PLD, may be one relevant channel regulator (Hille et al., 2015). Recently, orexin receptors were also shown to enhance the persistent Na⁺ current, I_NaP (Tanaka et al., 2021), which is assumed to be carried by the voltage-gated Na⁺ channels (Na_v) and to act as a central regulator of cellular excitability (Lin and Baines, 2015; Stafstrom, 2007).

Recombinant CHO-K1 cells are depolarized by orexin receptor stimulation upon activation of an NSCC suggested to be a TRPC channel (Larsson et al., 2005). In human OX₁-expressing HEK293 cells, TRPC3/6 channels are suggested to contribute to the orexin-A induced Ca²⁺ oscillations (Peltonen et al., 2009) and in OX₁-expressing IMR-32 cells to reverse mode action of NCX, as in Fig. 3B (Louhivuori et al., 2010). Both human orexin receptor subtypes inhibit GPCR-regulated inward rectifier K⁺ (GIRK1 and -2) channels in recombinant HEK293 cells and depolarize the cells (Hoang et al., 2003).

Activation of human orexin receptors is capable of causing production of the endocannabinoid 2-AG (Fig. 3, A and D) as first demonstrated in recombinant CHO-K1 cells and then in native CNS neurons; significant additional information comes from the innovative utilization of inhibitors of relevant enzymes in several studies (II.B.6.). 2-AG is likely to act as a central mediator of orexin responses in the CNS. This implies that orexin receptors engage the PLC pathway also in the CNS, including its other characteristic messages such as PIP₂ decrease, DAG and IP₃ increase and intracellular Ca²⁺ release (II.B.1., II.B.2.). Interestingly, 2-AG may also act on ion channels and on the postsynaptic side (Oz, 2006; Gantz and Bean, 2017; Soderstrom et al., 2017), which may cloud the distinction between “direct” and “indirect” effects of orexin receptor signaling.

Presynaptic orexin actions have also been reported (van den Pol et al., 1998; Li et al., 2002; Smith et al., 2002; Davis et al., 2003; Lambe and Aghajanian, 2003; Borgland et al., 2008). Ca²⁺ elevation has been seen in a few studies (reviewed in Kukkonen and Leonard, 2014; Leonard and Kukkonen, 2014). Upon longer exposure to exogenous orexins, plastic effects on neuronal signaling, for instance long-term potentiation, take place (Selbach et al., 2004; Borgland et al., 2006; Chen et al., 2008; Selbach et al., 2010; Lu et al., 2016).

D. Cell Plasticity and Death in Non-Neuronal Cells

Long-term activation (24–96 hours of continuous exposure) of both orexin receptor subtypes has been shown to induce programmed cell death in recombinant cells, cancer cell lines and native cancer cells in culture. The findings and the mechanisms have been reviewed in detail in Kukkonen, 2013. In short, two different mechanisms, though potentially not mutually exclusive, have been reported in two different CHO cell clones. One relies on orexin receptor activation of a G_q subfamily G protein → an unidentified Src family kinase → orexin receptor phosphorylation → engagement of the protein phosphatase SHP-2 (Voisin et al., 2008; El Firar et al., 2009). The other is mediated via the p38 MAPK pathway (Ammoun et al., 2006b). These mechanisms have not been investigated further.

Regardless of the signaling cascades, it is very interesting that some carcinoma cells express orexin receptors (primary cells: colorectal carcinoma; cell lines: colorectal carcinoma, glioma, pancreatic carcinoma, prostate carcinoma, and some others suggested but not verified) which couple to (putative) programmed cell death (Rouet-Benzineb et al., 2004; Voisin et al., 2006; Voisin et al., 2011; Bieganska et al., 2012; Alexandre et al., 2014; Dayot et al., 2018; Huan et al., 2023). A peculiar finding reports cell death in the AsPC-1 pancreatic carcinoma cell line upon exposure to either orexin-A or the DORA almorexant (Dayot et al., 2018). As such it is not impossible in the world of biased agonism that a compound is an antagonist for one pathway and an agonist for another. However, the cell death upon orexin receptor activation has also been reported to rely on G_q (Voisin et al., 2008), but as yet, there is no evidence that almorexant acts as a G_q (or other) pathway agonist in any direct orexin receptor assay. Thus, the mechanism of the effect of almorexant may be distinct from action on orexin receptors.

Several orexin receptor agonists are in various clinical development stages for narcolepsy etc. (V.B.2.), raising the question of whether such ligands could also activate the cell death response. It is currently unknown whether native neurons also express the cell death response following long-term orexin receptor activation. Orexin receptor agonists for the indications would only be administered diurnally to consolidate daytime waking, and thus the receptor stimulation would not be continuous. Also, the current data from limited animal or human experiments with small molecule orexin receptor agonists does not indicate any demise of orexin receptor-expressing neurons, as concluded from the retained response (II.G.). However, a potential sustained nonmedical use (e.g., misuse) of orexin receptor agonists might carry risk.

In contrast to the cell death-stimulating effects of orexins, other studies associate them with promotion of cell survival and plasticity (reviewed in Kukkonen, 2013). Some studies suggest that orexin receptor stimulation increases cancer cell proliferation (see, e.g., Suo et al., 2018). Part of the explanation may arise from the fact that cancers are heterogeneous, and even the established, defined cancer cell lines tend to be subject to geno- and phenotypic drift. More recent studies suggest a protective role of orexins in, e.g., drug-induced kidney injury (Jo et al., 2022), though most such studies have focused on neuronal cells (IV.F.). No long-reaching conclusions about the actual role of orexins or molecular mechanisms can be drawn from such limited studies.

E. Other Orexin Receptor Signaling

Orexin receptor signaling has been suggested to activate a number of protein kinases, e.g., PKC, extracellular signal-regulated kinase (ERK), p38 and p90^RSK, in studies using either direct or indirect methods (Milasta et al., 2005; Ammoun et al., 2006a,b; Johansson et al., 2008; Wang et al., 2014). The primary signaling cascades may be extrapolated from the various kinase activities, but the approach is not generally valid.

AMP-activated protein kinase (AMPK) is a central regulator of cellular energy metabolism and cell growth. Orexin-A activates AMPK in cultured rat arcuate nucleus neurons via L-type voltage-gated Ca²⁺ channel-mediated Ca²⁺ influx (Wu et al., 2013); the simplest explanation is that orexin-A induces depolarization via some of the mechanisms discussed above (II.C.). Inhibition of AMPK abrogates orexin-mediated increase in neuropeptide Y production and food intake; however, the AMPK inhibitor was introduced intracerebroventricularly (icv) (Wu et al., 2013). Phosphoinositide-3-kinase (PI3K) → phosphoinositide-dependent kinase 1 (PDK1) → protein kinase B (PKB/Akt) cascade has been suggested for orexin receptor signaling in both direct and indirect studies (Mazzocchi et al., 2001a; Ammoun et al., 2006a; Göncz et al., 2008), although the activation mechanisms of this cascade are unclear. In the recombinant hypothalamic neuronal cell lines N41/OX₁ and N41/OX₂ (and likewise recombinant OX₁- or OX₂-expressing HEK293 cells), orexins activate mTORC1 (mammalian target of rapamycin complex 1) in a PI3K- and PKB- (and ERK-) independent manner. Instead, the signal is suggested to be relayed via Ca²⁺ influx and the lysosomal pathway. However, this is not specific to orexin receptors since any Ca²⁺ elevation mimics it (Wang et al., 2014). mTOR activation has also been associated with corticosteroid synthesis in the cell lines H295R and Y-1 (Zhang et al., 2024a). mTOR and AMPK are often put in opposition, but orexin receptors are known to activate even otherwise apparently counteracting pathways such as G_i and G_s or ERK and p38 (see above). PKB is also shown to be activated in rat embryonic cortical neuronal cultures (Sokolowska et al., 2014); it is unclear, how PKB phosphorylation could be so strong if only a small subset of cortical neurons expresses orexin receptors, as is suggested (IV.A.).

Normal brown adipose tissue (BAT) development in mice has been suggested to require OX₁ receptor signaling via p38 and BMP-7 (bone morphogenic protein-7) (Sellayah et al., 2011; Sellayah and Sikder, 2012). Interaction of orexin and bone morphogenic protein signaling pathways has also been reported in other studies (Fujita et al., 2018; Fujisawa et al., 2019; Fujisawa et al., 2021).

F. Biased Signaling

Where there is a receptor with several primordial signaling pathways, there is potential for biased signaling. For orexin receptors, the question has been raised in several reviews (e.g., Kukkonen, 2012; Leonard and Kukkonen, 2014; Kukkonen, 2019; Kukkonen and Turunen, 2021; Dale et al., 2022). Unfortunately, primary studies investigating the issue, even with regard to the physiological transmitters orexin-A and orexin-B, are rare. Since the original study (Sakurai et al., 1998), orexin-A is believed to be a more potent agonist on OX₁ than orexin-B while the ligands would be equipotent on OX₂. Information about the relative potencies of these two ligands for different responses in the same CHO-K1 cell clones has been collected in Table 1 of (Kukkonen, 2013). The analysis indicates that the EC₅₀(orexin-B)/EC₅₀(orexin-A) for OX₁ varies between 1.6–18-fold – depending on the response measured – while in the literature, regarding different cell lines and clones, this has been suggested to be as much as 83-fold (Sakurai et al., 1998). The potencies for orexin-A, orexin-B and the putatively OX₂-selective synthetic agonist, Ala¹¹, d-Leu¹⁵-orexin-B, at OX₁ and OX₂ receptors in CHO-K1 and HEK293 cell lines have also been compared with respect to Ca²⁺ elevation. In general, the preference of orexin-B and Ala¹¹, d-Leu¹⁵-orexin-B for OX₂ was confirmed although the results in the two cell lines were significantly different (Putula et al., 2011). Along the same lines, the apparent OX₂ receptor-selectivity of Ala¹¹-d-Leu¹⁵-orexin-B is highly variable in different cellular and physiological studies (Asahi et al., 2003; Putula et al., 2011; Yamamoto et al., 2022). Note, that the potential off-target effects of Ala¹¹-d-Leu¹⁵-orexin-B have not been assessed.

The small molecule agonist Nag 26 was compared for OX₁ and OX₂ receptor-mediated responses; some indication of biased signaling was found (Rinne et al., 2019). Firazorexton/TAK-994 has been compared with native orexin peptides for different responses mediated by the OX₂ receptor and shown to be equally potent and efficacious as orexin-B with respect to PLC activation, but clearly less potent (or efficacious) than orexin-B with respect to β-arrestin recruitment and ERK and CREB (cAMP response element-binding protein) phosphorylation (Ishikawa et al., 2023); however, this may alternatively be explained by lower intrinsic efficacy of firazorexton/TAK-994.

Thus, there is limited evidence of biased agonism for the few orexin receptor agonists that have been studied, but the primary studies are limited and more extensive work is required, especially with a greater number of agonists. The novel methods to directly measure G protein and other primary pathways (II.A.1.) would be preferred although it should be recognized that these studies are not always in perfect agreement either. Independently of the results of such studies, the use of agonists for orexin receptor classification in experimental work is not recommended; antagonist studies are more conclusive, especially when determining the rank orders of antagonistic potency.

G. Orexin Receptor Desensitization and Trafficking

GPCR stimulation is classically associated with receptor desensitization, internalization and downregulation, though these processes affect different GPCRs to different degrees; in certain cases, such phenomena may actually result in a redirection of the receptor signaling, most notably upon the action of β-arrestin as an adaptor protein for both internalization and signaling (Shenoy and Lefkowitz, 2011). Kinases such as GRKs (GPCR kinases) and adaptor proteins β-arrestin-1 and -2 are often but not always associated with these processes (Shenoy and Lefkowitz, 2011; Magalhaes et al., 2012; Moo et al., 2021). In recombinant HEK293 cells, both orexin receptor subtypes interact with both isoforms of recombinant β-arrestin upon orexin-A challenge; the interaction of OX₂ with both β-arrestin isoforms is proposed to be more sustained than that of OX₁ (Dalrymple et al., 2011). After a prolonged agonist challenge, the receptors and β-arrestin internalize together (Evans et al., 2001; Dalrymple et al., 2011). The C-termini of both orexin receptors are at least partially involved in this interaction (Milasta et al., 2005; Dalrymple et al., 2011; Jaeger et al., 2014); whether interaction with β-arrestin is required for internalization remains unclear due to contradictory data (Milasta et al., 2005). Blocking the interaction with β-arrestin-2 using receptor mutagenesis shortens the time of sustained ERK activation by OX₁ receptors, whereas the Ca²⁺ response is apparently not affected (Milasta et al., 2005). When quantified, the decrease in cell surface expression after orexin stimulation was quite substantial for both OX₁ and OX₂ receptors in HEK293 cells (>50%), but much more limited for OX₂ receptors in CHO-K1 cells (∼25%) (Dalrymple et al., 2011; Ward et al., 2011; Kumagai et al., 2015). However, there was no attempt to quantify the potential functional effect of this on receptor signaling. CHO-K1 cells expressing either orexin receptor subtype can be stimulated repeatedly with orexin-A with some minutes of washout between stimulations without any apparent decrease in the Ca²⁺ response (Johansson et al., 2007; Kukkonen et al., unpublished). OX₁ receptor-expressing CHO-K1 cells can be stimulated with 100 nM orexin-A continuously for 60 minutes with little apparent decrease in the Ca²⁺ signal (Ammoun et al., 2006a). OX₁ receptors give a much more sustained ERK activation in CHO-K1 cells (Ammoun et al., 2006a) than in HEK293 cells (Milasta et al., 2005; Dalrymple et al., 2011). It is difficult to draw firm conclusions, but the combined evidence from the studies thus appears to indicate that orexin receptor internalization is stronger in HEK293 than in CHO-K1 cells.

The results from cell lines tell us little about the therapeutically central question, whether a drug will retain its clinical efficacy and potency upon long-term use. The apparent sensitization or desensitization on a systemic level cannot be traced back to a specific mechanism but is a sum of effects on different levels. Orexin peptides and small molecular agonists ameliorate narcolepsy symptoms in mouse models and human patients (V.B.2.). Repeated orexin receptor stimulation with small molecule orexin receptor agonists has been tested on the systemic level in a few studies. No reduction in the effect was observed in narcoleptic PPO-KO mice upon YNT-185 administration during the active periods of three consecutive days (Irukayama-Tomobe et al., 2017). After a two-week-treatment of narcoleptic PPO-ataxin-3 mice (in which the orexinergic neurons die postnatally) during the daily active periods, the responses to TAK-925 and TAK-994 were either not at all or only mildly reduced (Evans et al., 2022; Ishikawa et al., 2022; Ishikawa et al., 2023). In NT1 patients (IV.C.), a similar protocol has not been applied but the effect of TAK-994 was sustained for an 8-week-treatment period (Dauvilliers et al., 2023). Along the same lines, unpublished clinical data presented at the recent 2023 World Sleep Congress with various new OX₂ agonists demonstrate sustained effects on various daily sleep parameters in narcolepsy patient populations, implying that receptor desensitization is not an issue in patient populations with narcolepsy type 1 or 2 (NT1 and NT2, respectively) or other excessive daytime hypersomnias (see also Jacobson et al., 2022). This is reminiscent of that observed with chronic somatostatin receptor agonists such as octreotide, used to treat acromegaly and other endocrine disorders, given over several decades (Weckbecker et al., 2003; Hoyer and Bartfai, 2012). It should also be noted that ectopic overexpression of PPO gave an amelioration of narcoleptic symptoms in PPO-ataxin-3 mice, though the persistence of the effect was not defined in the study (Mieda et al., 2004). Nevertheless, the findings on the cellular and systemic level are quite similar.

Compensatory effects may also be seen upon long-term antagonist treatment. However, orexin receptor antagonists in clinical use or testing retain their efficacy without the need to increase the dose during long-term treatment: Suvorexant, lemborexant and daridorexant were the longest ever tested hypnotics in phase 3 studies, some of which lasted for up to 14 months (Hoyer et al., 2020; Jacobson et al., 2022).

A. Receptor 3D Structures

Human orexin receptor 3D structures were initially constructed by homology modeling based on known GPCR crystal structures and used for modeling of the potential binding modes for receptor ligands and molecular dynamics simulations (Malherbe et al., 2010; Tran et al., 2011; Heifetz et al., 2012; Karhu et al., 2015; Turku et al., 2016; Turku et al., 2019). The first (inactive) orexin receptor 3D structure (OX₂–suvorexant) was published in 2015, followed by OX₁ structures in 2016 (OX₁–SB-674942 and OX₁–suvorexant) and then other bound ligands (Yin et al., 2015; Yin et al., 2016; Suno et al., 2018; Hellmann et al., 2020; Rappas et al., 2020; Hong et al., 2021; Asada et al., 2022; Yin et al., 2022) [Protein Data Bank (PDB) codes for OX₁: 4ZJ8, 4ZJC, 6V9S, 6TQ9, 6TQ7, 6TQ6, 6TQ4, 6TP6, 6TP4, 6TP3, 6TOT, 6TOS, 6TOD, 6TO7; PDB codes for OX₂: 4S0V, 5WS3, 5WQC, 7L1U, 7L1V, 6TPG, 6TPJ, 6TPN, 7SR8, 7SQO, 7XRR]. The studies have been mostly performed using X-ray crystallography except for two, where cryo-electron microscopy (cryo-EM) was used. The latter investigated the putatively active receptor conformations: one with the OX₂ receptor in complex with orexin-B and the small molecule agonist compound 1 (PDB: 7L1U and 7L1V, respectively) and the other with OX₂ receptor in complex with a G_q-mimetic (mini-Gα_s/q/iN/Gβ/Gγ) or G_i1 and a small molecule agonist danavorexton/TAK-925 (PDB: 7SR8 and 7SQO, respectively) (Hong et al., 2021; Yin et al., 2022).

When bound to OX₂, only the C-terminal part of orexin-B (N20–M28) had high enough density and was included in the final structure (Hong et al., 2021). This visible peptide region adopts extended conformation and does not retain the α-helix that is seen in the nuclear magnetic resonance (NMR) solvent structures of orexin-A and -B (see V.A.), which have been used in molecular modeling studies (Heifetz et al., 2012; Karhu et al., 2015; Karhu et al., 2019). In contrast, the sodium dodecyl sulfate (SDS) micelle-bound NMR structure of orexin-A C-terminus is less well defined (leaving a possibility of conformation freedom) and orexin-B is unstructured in its extreme C-terminus; thus the micelle-bound peptide structures (Miskolzie and Kotovych, 2003; Miskolzie et al., 2003) are somewhat more similar to the secondary structure of the receptor-bound orexin-B than are those in aqueous solvent (V.A.). The peptide bound to OX₂ is in contact with TM2–TM7 as well as the extracellular loops 2 and 3 (ECL2 and -3) and the N-terminus (Hong et al., 2021). The peptide residues G24–M28 bind deep in the binding pocket complementing the space. I25 is sandwiched, M28 forms lipophilic contacts with OX₂, and there are hydrogen bonds between T27 and Asn324^6.55 and Gln134^3.32 as well as L26 and His350^7.39, likely through a water molecule. Note that the orexin peptide amino acids are here given with single letter codes and the receptor amino acids with three letter codes. Because of the high similarity of the C-terminal region of orexin-A and orexin-B, it is likely that they form similar interactions in this region. The region N20–A23 has several polar interactions with the widened upper part of the binding pocket (Hong et al., 2021). The N-terminal regions of orexin-B and the extracellular regions of OX₂ receptor were of low resolution (low density) in cryo-EM. The model was complemented by molecular dynamics simulations, where the N-terminal part of the peptide adopts an α-helical conformation, and forms contacts with the α-helical N-terminus of OX₂ as well as the ECL2, indicating a role of extracellular receptor parts in orexin peptide binding (Hong et al., 2021), although the N-terminal part of neither orexin peptide is highly significant for receptor binding (V.A.).

The main differences between the putative active and inactive conformations of OX₂ receptors are in TM5–TM7, in a manner that has been described as the conserved mechanism of GPCR activation (Manglik and Kruse, 2017; Weis and Kobilka, 2018), in addition to several microswitches such as Asn-Pro-x-x-Tyr (a.k.a. NPxxY) motif and the agonist connector motif in the base of the orthosteric binding pocket (Pro-Val-Phe in orexin receptors) (Hong et al., 2021).

Both putative active structures with small molecule agonists (danavorexton/TAK-925 and compound 1; Fig. 4B2) are well aligned – despite the use of the nanobody Sb51 to obtain the compound 1-bound structure – showing a root mean square deviation of atomic positions (RMSD) of 0.6 Å when overall structures are compared (Hong et al., 2021; Yin et al., 2022). The level of difference between the peptide-bound structure and compound 1-bound structure is also similar, having an RMSD of 0.54 Å, when the overall structures (distinguishable α-carbons in the receptors and G proteins) are compared; the main difference is seen in the ECL sites in contact with Sb51 (Hong et al., 2021). Inactive OX₁ (SB-674042-bound, 4ZJC) and OX₂ structures (suvorexant-bound, 4S0V) also align well, having an RMSD of 0.4 Å over the distinguishable α-carbons of the receptors, mostly differing in the ECL2 and N-terminus (Yin et al., 2016).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4

Selected synthetic small molecule orexin receptor ligands. (A) antagonists, (B) agonists. Please note that for some compounds, stereochemical representation in different sources may differ. The molecular structures were drawn using ChemDraw 22.2.0.3300 (PerkinElmer Informatics, Inc.). For antagonists, the selectivity for the receptor subtypes (the ratio of K_d, K_b or K_i values) is given. If no value is given, there are no relevant published data, and if no range is given, there is just one published value. Please observe that the values determined by equilibrium binding and functional assays can differ, for instance due to kinetics of the ligands. The affinity data and the specific references can be found at the IUPHAR/BPS Guide to PHARMACOLOGY.

These studies show that the small molecule binding sites in OX₂ and OX₁ are highly similar, differing only in two residues: Thr111^2.61 versus Ser103^2.61 and Thr135^3.33 versus Ala127^3.33, which have been suggested to be important for subtype selectivity, possibly through water network interactions (Malherbe et al., 2010; Tran et al., 2011; Yin et al., 2015; Yin et al., 2016; Rappas et al., 2020; Hong et al., 2021; Yin et al., 2022). These positions contribute to the selectivity of danavorexton/TAK-925 (>10 000-fold OX₂ receptor-selective) as determined by reciprocal mutations: The potency at OX₁^{S103T, A127T} was more than 300-fold higher as compared with wildtype OX₁ while the potency at OX₂^{T111S, T135A} was 500-fold lower as compared with wildtype OX₂ (Yin et al., 2022). Interestingly, mutating Thr111^2.61 of OX₂ to Ser or Ala did not have any effect on the binding of the DORA lemborexant (Fig. 4A3), while Thr135^3.33Ala mutation led to a 10-fold decreased affinity (Asada et al., 2022). Among wildtype receptors, lemborexant prefers OX₂ only 2.3-fold (Beuckmann et al., 2017; Asada et al., 2022). Being a smaller residue, Ala127^3.33 allows two configurations for lemborexant as compared with one for threonine, which may explain the difference in the affinity in favor of threonine (Asada et al., 2022). The Ala127^3.33Thr mutation in the OX₁ receptor reduces the binding affinity of the OX₁-selective SB-674042 20-fold (Malherbe et al., 2010). The position 3.33 may also explain the selectivity of SB-674042 for OX₁: The larger residue threonine (as compared with alanine) in OX₂ creates a steric clash with SB-674042, and reciprocal mutations increase the affinity of OX₂ while decreasing the affinity of OX₁, whereas the binding of DORA-12 (Fig. 4A3) is not affected (Malherbe et al., 2010; Tran et al., 2011; Yin et al., 2016). A recent study presented an NMR analysis of a mutated OX₂ receptor labeled with ¹³C-labeled methionine upon binding of the OX₂ orexin receptor-selective antagonist (2-SORA) EMPA and the DORA suvorexant (Imamura et al., 2024). The analysis revealed conformational heterogeneity of the suvorexant-bound complex, in which suvorexant adopted multiple poses and the whole complex acquired multiple states. In contrast, EMPA was more rigidly fitted into OX₂ binding pocket, showing strong hydrogen bonds to T135^3.33 and P131^3.29. In molecular dynamics simulations, residues responsible of subtype-specificity, T111^2.61 and T135^3.33, were affected differently: Suvorexant showed higher affinity toward T111^2.61, while EMPA toward T135^3.33, which was in line with the conclusion of heterogeneity caused by T135^3.33 (Imamura et al., 2024).

Asn324^6.55 of OX₂ plays an important role for the orexin-B binding mode (hydrogen bonding) and receptor activation, while its role in danavorexton/TAK-925 binding is less prominent (Hong et al., 2021; Yin et al., 2022). The Asn324^6.55Ala mutation is better tolerated by danavorexton/TAK-925 than orexin-B with respect to PLC activation (Yin et al., 2022). Orexin-B, compound 1 and danavorexton/TAK-925 share an important contact with Gln134^3.32, which is critical for many GPCRs (Venkatakrishnan et al., 2013): Alanine mutation in this position markedly reduces responses to orexin-A, orexin-B and danavorexton/TAK-925 (Ca²⁺ or PLC activity) (Malherbe et al., 2010; Hong et al., 2021; Yin et al., 2022). Residues at the base of the binding pocket (Gln134^3.32 as well as Tyr317^6.48) show a clear effect in the β-arrestin recruitment assay for danavorexton/TAK-925. Both residues undergo large conformational changes between inactive and active receptor forms: For instance, Gln134^3.32 projects upwards in the active and downward in the inactive structures (Yin et al., 2015; Yin et al., 2016; Rappas et al., 2020; Hong et al., 2021; Asada et al., 2022; Yin et al., 2022). Interestingly, Gln126/134^3.32 (OX₁ and OX₂ position numbers, respectively) also adopts different conformations between OX₁ and OX₂ upon the DORA lemborexant binding: In both cases, lemborexant forms a hydrogen bond to Gln126/134^3.32, but the bond is to the sidechain nitrogen in OX₁ and to the sidechain oxygen in OX₂. Gln126^3.32Ala reduces the affinity of OX₁ for SB-674042 50-fold (Malherbe et al., 2010); Gln126 has the largest surface contact with SB-674042, when compared with any other residue (Yin et al., 2016). Mutation of Gln134^3.32Ala in OX₂ reduces the affinity for the DORA lemborexant 50-fold, but only fourfold for the DORA suvorexant whereas the binding of the OX₂-selective antagonist EMPA is not affected (Asada et al., 2022).

Lemborexant was also uniquely sandwiched by Pro131^3.29 and His350^7.39 in OX₂ while other small molecule antagonists (EMPA: 5WQC, suvorexant: 4S0V, HTL6641: 6TPN) formed a water-mediated hydrogen bond with His350^7.39 (Yin et al., 2015; Suno et al., 2018; Rappas et al., 2020). For compound 1 binding, the contact is not water-mediated (Hong et al., 2021). Mutation of His350^7.39 to alanine was deleterious for lemborexant and suvorexant – but not EMPA – binding to OX₂ (Asada et al., 2022). His344^7.39 in OX₁ has also been described to be important for SB-674042 binding by forming an aromatic contact with the ligand (Malherbe et al., 2010; Yin et al., 2016).

B. Receptor Dimers or Oligomers

GPCRs have been proposed to form di- or oligomeric homo- or heteromeric complexes (reviewed in Milligan, 2009; Farran, 2017). The current view is that all GPCRs of the (originally defined) class A and class C make homo- or heteromeric complexes, and possibly all GPCRs do. Assay methods do not reveal the exact number of protomers and thus the complexes are often referred to as dimers. Dimerization is an obligatory feature for (possibly all) class C receptors (Jones et al., 1998; Kaupmann et al., 1998; White et al., 1998; Kuner et al., 1999; Wu et al., 2014; Mao et al., 2020; Papasergi-Scott et al., 2020; Gao et al., 2021), while the case for class A receptors is unclear. For class A GPCRs, heterodimerization has been demonstrated to affect properties such as trafficking, ligand affinity and potency and receptor signaling pathways, but in the vast majority of cases the significance of dimerization is unknown (Maggio et al., 2007; Milligan, 2009; Gomes et al., 2016; Gaitonde and Gonzalez-Maeso, 2017). Both human orexin receptor subtypes have been demonstrated to form homo- and heteromeric complexes with each other and with every GPCR tested in recombinant setting. Complex partners determined to date for the OX₁ receptor include apelin, CB₁, CCK₁, CRF₁, CRF₂, ghrelin, GPR103 and κ opioid receptors, and for the OX₂ receptor, 5-HT_1A, CB₁ and GPR103 receptors (reviewed in Kukkonen, 2017; Kukkonen and Turunen, 2021). The non-GPCR σ₁ and σ₂ receptors have also been suggested to complex with orexin receptors (Navarro et al., 2015; Navarro et al., 2019). Heterodimerization of OX₁ and κ opioid receptors has been proposed to shift the signaling from G_q and G_i, respectively, of the homomeric receptors to G_s (Chen et al., 2015). However, another study demonstrated a functional interaction between OX₁ and κ opioid receptors (Robinson and McDonald, 2015), providing an alternative explanation; this is similar to the question regarding dimerization versus functional interaction of OX₁ and CB₁ receptors (Jäntti et al., 2013). So far, the only reported complex with a likely physiological significance is that between OX₁, CRF₁ and σ₁ receptors in the rat VTA (Navarro et al., 2015). However, there are also issues with this study regarding the in vitro and in vivo methodology for testing of the inhibition of dimerization.

A fully reliable demonstration of dimerization/oligomerization is not simple. Colocalization demonstrated with high-resolution microscopy does not prove anything else but colocalization, though comobility may be indicative for complexes. Bioluminescence and fluorescence resonance energy transfer (BRET and FRET, respectively) may be slightly more specific, but not all receptor complexes work well in these assays. Bimolecular fluorescence complementation (BiFC) has the built-in problems that the two pieces of the fluorescent protein have affinity for each other and thus promote dimerization, and that the complex formation is irreversible (Kerppola, 2006). Importantly, these methods are difficult to apply to native receptors. Antibody-based methods such as the proximity ligation assay also only prove colocalization and are totally dependent on the antibodies' specificities (generally low for orexin receptors). Native gel electrophoresis may miss dynamic interactions and is not able to determine the molecular weights of the complexes. Physiological rather than molecular interactions need to be excluded (Turunen et al., 2012; Jäntti et al., 2013; Robinson and McDonald, 2015).

As concerns orexin receptor complexes, there is thus little evidence of any physiological significance, although some of the findings may be interesting as such. In most cases, the receptors reported to form complexes with orexin receptors upon recombinant overexpression have not been demonstrated to be expressed in the same native cells. The specificity of cell-permeating peptides used to inhibit dimerization is by no means clear among GPCRs or other proteins (see, e.g., Ekokoski et al., 2010) and there is no unequivocal theoretical framework to support specificity. A single study has tackled the question of whether the orexin receptor (homomeric) complexes are stable or transient (Xu et al., 2011). The issues with these methods are additionally discussed in Kukkonen, 2017, 2019.

A. Orexin Neurons and Their Projections

Orexin neurons are found in the perifornical, dorsal and lateral hypothalamic areas, as determined in several mammalian species using antibodies against PPO, detection of PPO mRNA, labeling of the orexin neurons with the EGFP gene under the Hcrt promoter (Li et al., 2002) and tracing of the orexin neurons from target areas (more recent studies: Ciriello et al., 2013; Ciriello and Caverson, 2014; Garcia-Garcia et al., 2013; Yokota et al., 2016; reviewed in Sakurai, 2007; Kukkonen, 2013; Sakurai et al., 2021). Isolated orexin neurons have been reported in other nearby regions by peptide detection (reviewed in Kukkonen et al., 2002) and recently, based on PPO mRNA, even on clearly separate CNS regions (see below). In humans, the number of orexin neurons is estimated at 80 000 at most and just a few thousand or hundred in rats and mice (Peyron et al., 1998; Harrison et al., 1999; Fronczek et al., 2005; Thannickal et al., 2000; Fronczek et al., 2007; Thannickal et al., 2009; Obukuro et al., 2010; Black et al., 2018; Gonzalez and Prehn, 2018; Berteotti et al., 2021) (see also IV.C.1. and IV.D. for the observed changes in the neuron count). In the lateral hypothalamus, orexin neurons are interspersed among other neuron types, e.g., melanin-concentrating hormone (MCH)-ergic neurons (Elias et al., 1998; reviewed in Kukkonen, 2013). There has been considerable interest in finding molecularly or functionally distinct subpopulations of orexin neurons but with few notable results so far. It has been suggested that rats have anatomically distinct neuronal populations (medial versus lateral), based on the projection sites, whereas in mice, the neurons with different projection sites appear intermixed (reviewed in Sagi et al., 2021). In addition to tracing studies, functional and expression analyses in mice identify potential subsets of neurons, though these also appear intermingled (Schöne et al., 2011; Dalal et al., 2013; Mickelsen et al., 2017). Lateral hypothalamic neurons – between populations or within them – have been suggested to be minimally connected with each other (Burdakov and Karnani, 2020), although this may not align with previous anatomical and functional studies (reviewed in Kukkonen, 2013). If true (and this would not be unusual in the brain in, e.g., the hypothalamus or dorsal raphe nucleus), it would allow subpopulations of orexin neurons (or even individual neurons) to act independently in different functions.

No other unique marker for orexin neurons, other than PPO, has thus far been unequivocally reported (Dalal et al., 2013; Seifinejad et al., 2019; reviewed in Kukkonen, 2013). Orexin neurons express, in addition to orexins, other neuropeptides/neurotransmitters, e.g., dynorphin (and other prodynorphin products in different degrees), neurotensin, glutamate and amylin (these neuromodulators are present in most, if not all neurons) as well as galanin and GABA (present in a fraction of neurons) (Ciriello et al., 2013; Furutani et al., 2013; Muschamp et al., 2014; Apergis-Schoute et al., 2015; Li et al., 2015; Dergacheva et al., 2016a; Mickelsen et al., 2017; reviewed in Kukkonen, 2013); these cotransmitters may support the actions of each other (Schöne et al., 2014; Concetti and Burdakov, 2021). The neurons also express neuronal activity-regulated pentraxin II (NARP/NPTX2) (Reti et al., 2002), an early intermediate gene that plays a critical role in guiding synaptic plasticity.

Orexin neurons are regulated by many transmitter systems, and the neuronal pathways for some of these have been identified by retrograde or anterograde tracing while others remain speculative. Orexin neurons express many different types of receptors based either on functional studies (reviewed in Kukkonen, 2013) or mRNA expression studies (Harthoorn et al., 2005; Dalal et al., 2013). Figure 5 summarizes the transmitters regulating orexin neurons. Note that the studies cannot identify which regulatory systems apply to all orexin neurons versus only subpopulations of neurons, as would seem sensible from the functional perspective. Some studies have suggested that orexin neurons also contain orexin receptors (see Kukkonen, 2013) while others dispute this view (Vassalli et al., 2015). A recent study demonstrates that about 5% of the orexin-A-positive cells express either OX₁ or OX₂ receptor mRNA (Tsuneoka and Funato, 2024).

Dietary cues represent another central regulator of the activity of orexin neurons. Fasting increases PPO mRNA expression (Sakurai et al., 1998). Glucose directly hyperpolarizes the neurons via unidentified leak-type K⁺ channels and hypoglycemia depolarizes them (reviewed in Kukkonen, 2013). This has been suggested to give rise to the compensatory adrenal catecholamine secretion via stimulation of OX₂ receptors in the rostral ventrolateral medulla (Korim et al., 2016) and (partially) to pancreatic glucagon secretion via OX₁ receptors in the dorsal vagal motor nucleus (Paranjape et al., 2007), though these suggestions are only based on the effects of orexin receptor antagonists. Other studies have demonstrated that the glucose-mediated hyperpolarization of orexin neurons is only transient and thus may represent a sensor for the change in glucose levels rather than for absolute glucose levels (Williams et al., 2008). In contrast, lactate and nonessential amino acids depolarize orexin neurons (Parsons and Hirasawa, 2010; Karnani et al., 2011a; Viskaitis et al., 2022). Orexin neurons are also pH-sensitive, responding to acid with depolarization (IV.I.).

An early study found that the proximal promoter region (spanning 3.2 kB upstream of the Hcrt gene) was sufficient to direct the expression to orexin neurons (I.C.) (Sakurai et al., 1999; Moriguchi et al., 2002). Several transcription factors and other proteins are suggested either to repress (interferon-α and IGFBP3; Waleh et al., 2001; Honda et al., 2009) or to promote PPO expression (FOXA2, NR6A1, EBF2 and PLAGL1; Silva et al., 2009; Tanaka et al., 2010; Sanchez-Garcia et al., 2018; Tanaka et al., 2019). On the network level, orexin neurons have been suggested to be activated by, for instance, hunger (IV.B.), wakefulness (IV.C.), anticipation (at least in motor activity context; IV.D. and IV.J.), pain (IV.G.), stress and anxiety (Cohen et al., 2020; Yamashita et al., 2021), reward (IV.D.) (Hassani et al., 2016) and imminent cataplexy (in PPO-KO mice; Zhou et al., 2022) while inflammation may silence orexin signaling of these neurons (Grossberg et al., 2011; Seifinejad et al., 2023).

The activity of orexin neurons and the orexin production/release have been suggested to show diurnal rhythmicity in animal models (Taheri et al., 2000; Yoshida et al., 2001; Martinez et al., 2002). The original orexin measurements were performed using antibodies and are thus somewhat uncertain, given that these antibodies may possibly also detect the parent peptide PPO, orexin-A fragments or altogether different molecules (VI.), but measurements with mass spectrometry-based methods or biosensors, which are not subject to these liabilities, have also recently confirmed the diurnal rhythmicity of orexin release (Hopkins et al., 2021; Duffet et al., 2022a; Narita et al., 2023).

Potential targets for orexin neurons have been mapped by the methods described above as well as by detection of synapse formation, orexin receptor mRNA expression, and functional testing. These aspects have been thoroughly reviewed previously (see, e.g., Sakurai, 2007; Kukkonen, 2013; Sakurai et al., 2021) and thus are not specifically dealt with here. In general, orexinergic projections from hypothalamus span all major parts of the brain, though only specific sites. Early studies reported very sparse innervation of the cerebellum (Kukkonen et al., 2002) and thus the role of orexins there was questioned (Sakurai, 2007). More recent studies have identified some potential orexin actions there as well (Yu et al., 2010; Nisimaru et al., 2013; Ciriello and Caverson, 2014; Zhang et al., 2024b), but another recent study finds no orexin receptor mRNA in the cerebellum (Tsuneoka and Funato, 2024).

Studies suggest that orexinergic projections to the neocortex are scarce though significant. Originally, only layer 6b was pointed out (Bayer et al., 2004) and thus most studies have focused on it (Wenger Combremont et al., 2016; Hay et al., 2015; Zolnik et al., 2024), yet a recent study demonstrates orexin receptor mRNA expression in several cortical layers (Tsuneoka and Funato, 2024). ³H-EMPA autoradiography was used to map OX₂ receptor expression in the rat brain and peripheral tissues (Mitsukawa and Kimura, 2022); the results are in general agreement with findings of receptor mRNA in situ hybridization (Marcus et al., 2001; Trivedi et al., 1998), although higher receptor densities are detected by autoradiography, in particular in some cortical areas. By contrast, some ex vivo studies using isolated rodent cortical neurons suggest that a large proportion of total cortical neurons respond to orexins (Urbanska et al., 2012; Sokolowska et al., 2014; Palomba et al., 2020). It is possible that the culture of the embryonic or neonatal neurons prior to experiments changes the properties of the neurons or selects particular neuron subpopulations (Urbanska et al., 2012; Sokolowska et al., 2014; Palomba et al., 2020). Nevertheless, there may be more orexin activity in neocortex than previously assumed as revealed in the recent mapping of the peptidergic neuronal networks in the human prefrontal cortex by mRNA analysis (Zhong et al., 2022), demonstrating wide-spread expression of OX₂ mRNA and limited expression of OX₁ mRNA in these regions. The authors also suggested that PPO mRNA is expressed in pyramidal glutamatergic projection neurons and in GABAergic interneurons, in addition to the putative “classical” projection neurons from the hypothalamus; it is unclear whether the Hcrt promoter acts in the same way in these neurons as in the hypothalamus. Thus far there is, however, no other evidence to support these findings (i.e., orexin peptide or receptor expression, orexin receptor responses). In transgenic rodents, ectopic PPO expression may occur extrahypothalamically due to other processes (VI.). Altogether, only very few cells and regions seem to express mRNA for both orexin receptor subtypes (Tsuneoka and Funato, 2024).

There is certainly stronger evidence for some connections than for others. Additional confusion is generated by some methodological issues. Simple stimulation of neurons by exogenous ligands does not prove that there is endogenous signaling via the receptors, and many studies have not properly distinguished between pre- and postsynaptic effects. In essence, tetrodotoxin is not sufficient to eliminate all potential effects on presynaptic terminals, though it blocks action potentials and thus input to transmitter release. We have here altogether excluded the information based on the use of orexin receptor antibodies since they are likely to produce misleading results due the generally poor immunogenicity of GPCR epitopes (see, e.g., Dahl et al., 2023); specific evidence as to this has been obtained with orexin receptors (Kukkonen, 2012; Kukkonen, 2013). Ideally, validation of such antibodies should be performed in cells, tissues and/or animals from which the target receptor has been knocked out/down, but such controls are still not commonly performed.

B. Orexins in Feeding and Metabolism

Originally, orexin receptors were associated with appetite stimulation and arousal and maintenance of wakefulness. Appetite regulation was first reported by Sakurai and coworkers, since fasting increased brain PPO mRNA expression and icv injection of orexin-A or -B increased short-term feeding in fed rats (Sakurai et al., 1998). A major reason for testing appetite regulation originated from the hypothalamic expression of orexins. Orexin-A induced feeding more potently than orexin-B and thus the receptor of interest was thought to be OX₁ (Sakurai et al., 1998). Several studies replicated such findings in both rats and mice (reviewed in Kukkonen et al., 2002). As described above, orexin neurons are regulated by glucose, lactate and nonessential amino acids. A more detailed mapping of the orexin projections and orexin receptor mRNA distribution identified orexinergic innervation of feeding-associated nuclei such as the arcuate nucleus (Broberger et al., 1998; Elias et al., 1998; Peyron et al., 1998). Furthermore, injection of orexin-A into specific brain sites, such as the arcuate nucleus, the perifornical hypothalamic area, the rostral portion of the lateral hypothalamus, the paraventricular nucleus of the hypothalamus (PVN), and the nucleus accumbens, but not the substantia nigra, increased food intake (Dube et al., 1999; Kotz et al., 2002; Thorpe and Kotz, 2005; Kotz et al., 2008; Yang et al., 2018). Orexin neurons are also regulated by feeding-associated neuronal and endocrine input (IV.A.) (Figs. 7 and 9 in Kukkonen et al., 2002; Fig. 8 in Kukkonen, 2013).

Orexins were soon found to increase metabolic rate in rodents (Lubkin and Stricker-Krongrad, 1998), largely through activation of BAT via second order neurons (Oldfield et al., 2002; Berthoud et al., 2005; Zheng et al., 2005; Tupone et al., 2011; Martins et al., 2016). Orexins also increase locomotor activity (Hagan et al., 1999; Yamanaka et al., 2003a; Garau et al., 2020; Karnani et al., 2020). Thus, at least in rodents, orexins could also lead to increased energy expenditure. This was further validated by ectopic overexpression of PPO which protected mice from diet-induced obesity by burning excess calories in BAT (Funato et al., 2009). Interestingly, orexins produced by the placenta have been proposed to be required for prenatal BAT development in mice (Sellayah et al., 2011), although this view has been challenged (Kakizaki et al., 2019). Overweight and obesity are more common in human NT1 patients resulting in an average 10–20% higher body mass index (BMI) than in healthy controls (Kok et al., 2003; Scammell, 2015; Bassetti et al., 2019) although their basal metabolic rate is similar to BMI-matched controls (Fronczek et al., 2008; Dahmen et al., 2009) – assuming the comparison with BMI-matched controls is relevant. It should, however, be kept in mind that the loss of orexin neurons would not only affect orexins, but also other neurotransmitter systems involved in feeding and weight control, including the transmitters coexpressed in orexin neurons. Of note, the effects on weight gain in transgenic rodent models of deficient orexin signaling under normal and high-calorie food intake are not fully consistent with the human situation and may be mouse strain- and sex-dependent (Hara et al., 2005; Fujiki et al., 2006; Kakizaki et al., 2019).

Cold is a central trigger for BAT metabolic activity. PPO-ataxin-3 rats showed reduced thermogenesis upon cold exposure which may be a result of reduction in both BAT activation and locomotor activity as compared with normal controls (Mohammed et al., 2016). By contrast, in a study of PPO-ataxin-3 and PPO-KO mice, locomotor activity of both strains was normal and PPO-KO mice showed normal thermogenesis (Takahashi et al., 2013). Along these lines, human NT1 patients show normal BAT activation upon cold exposure (Enevoldsen et al., 2018). In addition to methodological differences between these studies, it remains possible that the role of orexins in BAT regulation in rodents and humans is different, since the structure and function of BAT are also distinct in adult rodents and humans (Wu et al., 2012; de Jong et al., 2019). Furthermore, different mouse models of orexin- or orexin neuron-deficiency have produced inconsistent results (Kukkonen, 2013; Takahashi et al., 2013; Inutsuka et al., 2014; Mohammed et al., 2016).

It has been suggested that central orexins regulate peripheral energy metabolism not only via BAT and locomotor activity, but may also involve hormonal regulators such as glucagon, insulin and adrenocorticotropic hormone, although the physiological significance remains unclear (Kukkonen, 2013). More recently, it was shown that a high fat diet induces steatohepatitis, hepatic inflammation and hepatocellular carcinoma more strongly in PPO-KO mice than in wildtype mice (Tsuneki et al., 2022). The occurrence of hepatic inflammation was largely abolished by icv orexin supplementation whereas obesity and fatty liver were not affected (Tsuneki et al., 2022). On the other hand, increased orexin signaling by optogenetic stimulation of orexin neurons reduced glucose tolerance in obese mice on a high fat diet (Xiao et al., 2021). Importantly, in the latter study, the expression of channelrhodopsin-2 in orexin neurons was (partially) validated, as the specificity of expression for transgenes under the proximal orexin promotor can strongly vary (see section VI. for discussion). Indeed, some ectopic expression was detected and not all expression was quantitated, introducing an aspect with which to question these results. However, the questionability of the results with orexin–cre lines in general is much higher when there is no attempt to assess the efficiency and specificity of the expression, as is the case in most studies (VI.). Finally, peripheral tumor-induced metabolic (higher plasma glucose, lower glucose tolerance) and sleep disturbances co-occur with abnormal activity of orexin neurons and are inhibited by blocking orexin receptors with the DORA almorexant (Fig. 4A3) (Borniger et al., 2018).

Overall, based on transgenic animal models as well as animal and human studies with orexin receptor antagonists (V.B.1.), it seems appetite regulation is not a major physiological role of orexins, or at least not one that can be easily manipulated, and thus appetite suppression appears unlikely to be a major therapeutic indication for orexin receptor antagonists (V.B.1.). This may be partially explained by the complicated relationship between appetite, metabolism and reward in orexin signaling (see below), or altogether lower significance of orexins in the regulation of appetite. While orexin neurons respond to energy shortage and orexins obviously increase feeding behaviors in the short term, the role of orexins in this may be rather in activating vigilance and seeking of essential nutrients (Yamanaka et al., 2003a; Linehan and Hirasawa, 2022; Viskaitis et al., 2022; reviewed in Peleg-Raibstein et al., 2023). Nevertheless, there are ongoing clinical trials with OX₁ receptor-selective antagonists (1-SORAs), such as JNJ-61393215, CVN45502 or CVN7666 (Fig. 4A1) in feeding-related conditions.

Studies with orexin receptor antagonists with regard to appetite and feeding are described under V.B.1. For the role of orexins in the reward and motivated behavior associated with food ingestion, please see IV.D.

C. Orexins in Wakefulness and Sleep

The central role of orexins in the regulation of wakefulness and sleep emerged soon after the discovery of orexins. Central application of orexin-A icv or locally at structures involved in sleep–wake control increased wakefulness in rats (Hagan et al., 1999; Bourgin et al., 2000; Methippara et al., 2000; Piper et al., 2000). Soon after the discovery of the orexin system, hereditary canine narcolepsy was found to result from three distinct OX₂-inactivating Hcrtr2 mutations in three breeds of dogs (Lin et al., 1999; Hungs et al., 2001). Along the same lines, PPO-KO mice did not lose weight as expected based on the hypothesis of reduced appetite, but instead suffered from narcolepsy including cataplectic attacks (Chemelli et al., 1999), i.e., rapid, transient loss of skeletal muscle tone (except for respiratory function and eye movement). Cataplexy is typically induced by strong positive or negative emotions (laughter, fear) in humans suffering from narcolepsy with cataplexy, a.k.a. NT1. The cataplectic features of narcoleptic dogs and PPO-KO mice expanded to the investigation of the role of the orexin system in human NT1 subjects. Unlike in dogs, there is no evidence for OX₂ mutations in human NT1 but, instead, the patients have low orexin-A levels in their CSF and lack of orexin expression in the lateral hypothalamus (IV.C.1.). Further studies in rodents demonstrated that narcolepsy could also be induced by postnatal genetic destruction of the orexin neurons in mice and rats via the toxin ataxin-3 expressed under the Hcrt promoter (Hara et al., 2001; Beuckmann et al., 2004; Mieda et al., 2004), postnatal toxic destruction of the orexin target neurons by orexin-B conjugated with the toxin saporin (Gerashchenko et al., 2001) or by deletion of both Hcrtr1 and Hcrtr2 genes (double-KO) in mice (Willie et al., 2003). Interestingly, OX₁-KO mice show only minimal sleep–wake dysfunction, whereas OX₂-KO mice show a modest sleep phenotype with very rare cataplectic attacks (Willie et al., 2003), which contrasts with the receptor double-KO where cataplexy is relatively easily triggered (Kisanuki et al., 2001; Willie et al., 2003; Sakurai, 2007; Kalogiannis et al., 2011; Mang et al., 2012), like in the PPO-KO mice (Chemelli et al., 1999; Espana et al., 2007; Clark et al., 2009; Morawska et al., 2011; Oishi et al., 2013). Of note, selective knockout of OX₂ in dopaminergic neurons gives paradoxical hyperarousal (Bandarabadi et al., 2023).

While the symptoms of NT1 patients and the narcoleptic phenotype in animals were major indicators of a central role of orexins in the regulation of sleep and wakefulness, converging evidence was also obtained from healthy animals, i.e., centrally applied orexin-A increased wakefulness and reduced sleep (Hagan et al., 1999; Bourgin et al., 2000; Methippara et al., 2000; Piper et al., 2000). This was also supported by mapping of the orexinergic projections to wake-promoting nuclei (Trivedi et al., 1998; Date et al., 1999; Horvath et al., 1999b; Mondal et al., 1999; Marcus et al., 2001). Further studies have refined and expanded these findings. Optogenetic activation of orexin neurons in mice stimulates wakefulness while their silencing induces sleep (Adamantidis et al., 2007; Tsunematsu et al., 2011; Carter et al., 2012; Tsunematsu et al., 2012; Tsunematsu et al., 2013). Similarly, optogenetic stimulation of vasopressinergic neurons of the hypothalamic paraventricular nucleus increases wakefulness via activation of orexin neurons (Islam et al., 2022). However, it should be noted that optogenetic constructs needed for the light-mediated control of the neurons may have additional effects: High expression of archeorhodopsin 3 has been reported to silence orexin neurons even in the absence of illumination (Williams et al., 2019).

Exogenous expression of PPO results in a partial recovery from narcoleptic-like symptoms in PPO-ataxin-3 mice (Mieda et al., 2004; Liu et al., 2011b) and exogenous stimulation with synthetic OX₂ receptor agonists improves the symptoms in these and PPO-KO mice (Evans et al., 2022; Yamamoto et al., 2022) as well as in NT1 and NT2 patients (Evans et al., 2022). On the other hand, orexin receptor antagonists, both DORAs and 2-SORAs, are effective in inducing and maintaining sleep (V.B.1.). Evidence for the role of orexins in sleep-wakefulness circuitry has been reviewed previously (Sakurai, 2007; Scammell et al., 2017; Sakurai et al., 2021; Seifinejad et al., 2021; Jacobson et al., 2022). In short, orexin neurons are most active during wakefulness, somewhat active during nonrapid eye movement (NREM) sleep and cease to fire upon transition to rapid eye movement (REM) sleep – except for a subpopulation that is weakly active during REM sleep (Ito et al., 2023). Orexin neurons stimulate the wake-active aminergic neurons of the brain stem. They are themselves inhibited by these neurons and the preoptic area (POA) GABAergic NREM sleep-active neurons as sleep pressure develops. In addition, there is reciprocal inhibition between the aminergic neurons and the GABAergic neurons. During wakefulness, orexinergic drive activates aminergic neurons, which increases the inhibition of POA and allows cortical activation. Orexin neurons also regulate the wake-active/rapid eye movement (REM) sleep-active cholinergic neurons of the brain stem and the basal forebrain. The neuronal circuitry involved in sleep regulation, including the effects of orexins, has been mathematically modeled (Sorooshyari et al., 2015). With increased sleep drive, the activity of the POA increases, suppressing the activity of orexin neurons and the aminergic neurons. When orexin is lost in NT1, the reciprocal inhibition between the aminergic neurons and POA is disrupted and becomes a chaotic oscillator. The regulation of REM sleep is less clear (Scammell et al., 2017; Sakurai et al., 2021; Seifinejad et al., 2021), and the role of orexin receptor subtypes in REM versus NREM sleep (or inhibition of these) is different (Mieda et al., 2011; Clark et al., 2020; Sun et al., 2021; Ito et al., 2023). There is also a potential link between orexins and melatonin: a) orexinergic neurons project to the pineal gland, b) there are orexin receptors in the pineal gland and c) exogenous orexins inhibit or stimulate melatonin expression, depending on other stimuli (Mikkelsen et al., 2001; Fabris et al., 2004; Zhang et al., 2005; Appelbaum et al., 2009; Zieba et al., 2011; Kirsz et al., 2020). It has also been suggested that there are inhibitory melatonin receptors on orexinergic neurons, but the evidence is only based on receptor antibodies and no direct functional effects have yet been demonstrated (Sharma et al., 2018).

The phenotype of narcoleptic canines and orexin receptor KO-mice puts OX₂ in the focus for regulation of wakefulness and sleep. It is, yet, possible that, at least in rodents, both orexin receptors affect sleep, as seen in the orexin receptor knockout mice and by comparing the effects of OX₁ or OX₂ antagonism alone and in combination on sleep. However, the effect of OX₁ elimination/inhibition on sleep is only seen when OX₂ is inhibited at the same time (Willie et al., 2003; Dugovic et al., 2009; Mang et al., 2012; Betschart et al., 2013; Hoyer et al., 2013; Dugovic et al., 2014; Gotter et al., 2016; Brooks et al., 2023). Various pharmaceutical companies have thus been developing either 2-SORAs (Fig. 4A2) or DORAs (Fig. 4A3). There are currently three DORAs on the market as hypnotics, namely suvorexant, lemborexant and daridorexant (Fig. 4A3), while further DORAs as well as a few 2-SORAs are under development. The evidence and the ideas regarding antagonism of orexin receptor subtypes versus REM and NREM sleep are presented under V.B.1.b. The ligand discovery has significantly contributed to the understanding of the respective role of the two orexin receptor subtypes in the regulation of sleep and wakefulness.

Insomnia and reduced sleep quality become more frequent in humans and animals upon aging. A recent mouse study suggests this is caused by hyperactivity of the orexin neurons (Li et al., 2022c), despite the concurrent loss of orexin neurons. One of the major functions of sleep, in addition to memory consolidation, is suggested to be “clearing” of toxic waste products from the brain by the glymphatic system (reviewed in Rasmussen et al., 2018). Note that a recent study (Miao et al., 2024) disputes the view of the positive impact of sleep on brain clearance. The substances cleared from the brain include (phospho-)tau and amyloid β, linking sleep and AD (Rasmussen et al., 2018), and consequently the orexin system via regulation of sleep and wakefulness. Wakefulness, and especially sleep deprivation, increases soluble amyloid β in the brain interstitial fluid and CSF, whereas the DORA almorexant reduces this burden (Kang et al., 2009). Limited human studies with respect to NT1 do not allow clear conclusions to be drawn about the relationship between orexin signaling and AD (Gabelle et al., 2019; Dauvilliers, 2021; Lucey et al., 2023), and the orexin receptor antagonists have not yet been on the market long enough to test this concept. The potential interplay of orexins, neuronal degeneration and sleep disturbances is discussed under IV.F.

D. Orexins in Motivation and Addiction

The orexin system appears to play a major role in motivation and addiction. Activation of orexin receptors is associated with the motivation to work for different kinds of palatable foods (Thorpe et al., 2005; Borgland et al., 2009; Choi et al., 2010; Kay et al., 2014; Castro et al., 2016; Terrill et al., 2016). Injection of orexin-A into the nucleus accumbens shell increases motivation for food in a progressive ratio task, i.e., in which the effort (i.e., number of lever presses) required to receive a food reward, such as a sucrose pellet, progressively increases (Thorpe and Kotz, 2005; Choi et al., 2010). The 1-SORA SB-334867 has been reported to reduce motivation for (Thorpe and Kotz, 2005; Nair et al., 2008; Borgland et al., 2009) and consumption of (Borgland et al., 2009; Choi et al., 2010) palatable food rewards in a progressive ratio task but a recent study utilizing three different 1-SORAs (including SB-334867) questions this with respect to sucrose (Bergamini et al., 2024). Orexin neurons are activated in drug- or food-paired contexts (Harris et al., 2005) or in response to stimuli that predict the availability of food (Campbell et al., 2017). Furthermore, optogenetic stimulation of orexin terminals in the VTA, a region that encodes the significance of motivationally relevant stimuli, increases dopamine release in the nucleus accumbens as well as preference for contexts associated with that optical stimulation (Thomas et al., 2022). Orexin neuronal activity also correlates with reward anticipation in a Go/No-Go task (Tyree et al., 2023). In this study, optogenetic stimulation of orexin neurons prior to the cue increased premature responses in the Go/NoGo task, and the effect was blocked by an undisclosed 1-SORA suggesting a role of orexin in impulse control. In several studies, human NT1 patients were shown to have a reduced ability in specific decision-making tasks as compared with NT2 patients or controls, but whether this is associated with impulsivity is debated (Bayard et al., 2011; Dimitrova et al., 2011; Bayard et al., 2013). On the other hand, orexin signaling has been associated with risk avoidance via the nucleus accumbens (Blomeley et al., 2018).

Given the role of orexin signaling in the motivation for food, orexin receptor antagonists have been proposed as a pharmacotherapy for binge eating disorder. In animal models systemic administration of 1-SORAs and DORAs, but not 2-SORAs, reduced binge-like consumption of palatable food at doses that do not affect homeostatic eating (Piccoli et al., 2012; Alcaraz-Iborra et al., 2014; Rorabaugh et al., 2014; Olney et al., 2015; Vickers et al., 2015). These studies suggested that OX₁ antagonism may be effective at reducing binge eating in humans, but the results of a phase 2 clinical trial (NCT04753164; see https://clinicaltrials.gov/) with the 1-SORA, ACT-539313 (Fig. 4A1), have been disappointing (V.B.1.).

A role for orexin signaling in addiction was first suggested from behavioral studies showing that orexin signaling can reinstate extinguished place preference paired with morphine and that inhibition of OX₁ can block reinstatement of place preference or drug self-administration in rats (Boutrel et al., 2005; Harris et al., 2005). Orexin receptor activation contributes to the reinforcement, spatial and contextual memory and stress-associated drug relapse (reviewed in Baimel and Borgland, 2017; James et al., 2017; Fragale et al., 2021b). Many preclinical studies claim a major involvement of OX₁ receptors in drug seeking, although some studies implicate OX₂ over OX₁ for alcohol self-administration, conditioned place preference, and reinstatement (Shoblock et al., 2011). Addiction is associated with neural circuit dysfunction characterized by serial changes in synaptic transmission in the VTA. Morphine or cocaine-induced neuroplasticity within the VTA requires OX₁ signaling at dopamine neurons (Baimel and Borgland, 2017). Interestingly, elimination of OX₂ receptors in dopamine neurons led to faster acquisition of an operant task but more premature and perseverative responses, suggesting impaired inhibitory control, while motivation remained unchanged (Bandarabadi et al., 2023). NT1 patients had normal reinforcement learning when tested in a task known to be dopamine-dependent. However, they had decreased learning to avoid losses, suggesting that individuals with NT1 had reduced sensitivity to punishment, an effect likely linked to altered vigilance, as this was also seen in other central disorders of hypersomnolesence (Strauss et al., 2024) and even reported in controls deprived of sleep (Gerhardsson et al., 2021). Furthermore, exposure to cocaine, morphine or stress increases the activity of orexin neurons which may, in turn, further drive orexin inputs to the VTA, modulating motivated behavior while simultaneously promoting arousal states leading to insomnia, a common comorbidity of substance use (Fragale et al., 2021b). Indeed, an OX₁ receptor-preferring agonist, YNT-3708 (Fig. 4B1), increases conditioned place preference for cocaine (Iio et al., 2023). Thus, concern has been raised regarding the safety of using orexin receptor activators (e.g., in narcolepsy and other sleep disorders), whereas orexin receptor antagonists (especially 1-SORAs) may be protective (Fragale et al., 2021b; Zlebnik et al., 2021). However, current agonists in development for NT1 and other sleep disorders are targeted specifically to the OX₂ and there is little reported research activity on OX₁ agonists to date (V.B.2.).

DORAs have proved effective in animal models in decreasing cocaine-related impulsivity (Gentile et al., 2018b), cocaine motivation (Gentile et al., 2018a) and opioid seeking (Illenberger et al., 2023). Furthermore, Illenberger et al. (2023) recently showed that suvorexant blocked oxycodone seeking in male Wistar rats, while reducing it in females which tended to self-administer oxycodone twice as much as males. This idea has initiated several clinical trials for off-label use of the DORAs suvorexant and lemborexant (Fig. 4A3) for the treatment of opioid, cocaine, alcohol, or nicotine use disorders (e.g., NCT04229095, NCT03937986, NCT03999099, NCT04818086, NCT05145764) and substance use disorders comorbid with sleep disorders (NCT03897062, NCT05458609, NCT04287062). While most trials have not been completed yet, early results using suvorexant have been somewhat surprising. In a small study of non-treatment-seeking cocaine users (n = 7), suvorexant enhanced the reinforcing effects of low-dose intravenous cocaine and increased the motivation to choose cocaine over monetary rewards (Stoops et al., 2022). At clinical doses, the abuse potential of suvorexant and lemborexant appears low, although there may be small differences in abuse potential at higher doses of these drugs (Schoedel et al., 2016; Born et al., 2017; Moline et al., 2023). In a study examining the efficacy of suvorexant on reducing opioid withdrawal-related insomnia, there was an increase in sleep duration and no abuse potential of suvorexant (Huhn et al., 2022). Some studies have shown an increase in number of orexin-expressing neurons in opioid users and animal models of opioid or cocaine use disorder (Thannickal et al., 2018; James et al., 2019; Fragale et al., 2021a). This effect is likely due to increased orexin peptide expression levels in individual neurons allowing these neurons to show above the detection threshold. In addition, decrease in orexin cell body size has been observed in opioid users (Thannickal et al., 2018). Interestingly, even the number of orexin-expressing fibers within the VTA is increased after morphine treatment and repeated suvorexant treatment during morphine administration prevents this increase (McGregor et al., 2024). Therefore, it will be interesting to evaluate the effects of DORAs in pending clinical trials for substance use disorders as they are completed and published in peer reviewed journals. The 1-SORA JNJ-61393215 (Fig. 4A1) is being developed, for instance, for addiction. No data on its impact in addiction setting have been published yet, but a recent phase 1 study reported no sleepiness or other significant CNS side effects (Brooks et al., 2023).

E. Orexins and Other Psychiatric Disorders

A potential role for orexins in anxiety and depression is not immediately apparent, yet it seems reasonable to associate orexins with anxiety-like behavior because they increase vigilance, activity and stress response (reviewed in Kukkonen, 2013) and may reduce risk-taking (Blomeley et al., 2018) in rodent models. However, they also increase motivated behaviors (IV.D.) that may counteract stress (reviewed in Peleg-Raibstein and Burdakov, 2021). Whether the orexin system is involved in innate anxiety is not clear, and animal models of anxiety are not without their pitfalls (reviewed in Jacobson and Cryan, 2010; Peleg-Raibstein and Burdakov, 2021). Nevertheless, central administration of orexins increases anxiety-like behavior (reviewed in Kukkonen, 2013), whereas OX₁ antagonism reduces anxiety in a variety of anxiety and stress paradigms in both rodents and humans (Flores et al., 2015; Bonaventure et al., 2017; Salvadore et al., 2020; Yaeger et al., 2022; Ten-Blanco et al., 2023). On the other hand, there are also instances where orexin shows anxiolytic effects via OX₂ activation (Zhang et al., 2024b; reviewed in Kukkonen, 2013). In a posttraumatic stress model in rats, orexin-A, applied before the stressful stimulus, reduced the development of stress-related behavior whereas almorexant did the opposite (Cohen et al., 2020). Hypothalamic deep brain stimulation was identified as a potential anxiolytic intervention in a mouse model, and it was suggested that the action requires orexinergic neuronal activity (Li et al., 2022a). Insulin-like growth factor I (IGF-1) activates orexin neurons and loss of this signaling by IGF-1 receptor-KO in these neurons exposes animals to stronger anxiety response (Zegarra-Valdivia et al., 2020; Fernandez de Sevilla et al., 2022). Escitalopram, used commonly in both major depressive disorder (MDD) and in some anxiety disorders, was recently shown to upregulate orexin-A expression (Golyszny et al., 2022). These discrepancies may relate to a) different etiology of anxiety forms, b) potentially different roles of orexin signaling in the different stages (e.g., exposure to the anxiogenic stimulus versus the later symptomatic phase) and c) issues with the animal models. A few clinical studies have been performed or initiated to investigate the effect of orexin receptor antagonism in anxiety and stress, sometimes in connection with sleep. (However, note that when searching at https://clinicaltrials.gov/, the majority of clinical studies with orexin receptor antagonists do not contain orexin or hypocretin as a searchable term.) Interestingly, suvorexant was shown to reduce anticipatory anxiety in healthy humans (Gorka et al., 2022), raising the question whether this links to the role of orexins in anticipation (Donegan et al., 2022b) or anxiety (or both). The 1-SORA compound 56 (Fig. 4A1) has been reported to reduce stress-induced hyperarousal in rats without any hypnotic effects (Bonaventure et al., 2015) and another 1-SORA, JNJ-54717793 (Fig. 4A1), reduces panic attacks in rats (Bonaventure et al., 2017). Currently, the 1-SORA JNJ-612393215 (Fig. 4A1) is undergoing clinical trials for anxiety disorders (Brooks et al., 2023).

Animal models of depression suffer from similar problems as the models of anxiety (Cryan and Holmes, 2005; Jacobson and Cryan, 2007), and the results for orexins in this domain are not clear. In general, DORAs and to some extent 2-SORAs, but not 1-SORAs, are antidepressive, but the pharmacological investigations with SORAs and DORAs aiming at amelioration of the symptoms of are not unequivocal (V.B.1.c.). By contrast, it has been proposed that depression in humans correlates with lower CSF orexin levels (reviewed in Peleg-Raibstein and Burdakov, 2021) – based on the (unreliable) antibody-based orexin measurements in the brain or in the CSF (VI.). Some rodent studies suggest that orexin signaling itself has an antidepressant effect (Lutter et al., 2008; Scott et al., 2011). Insomnia is a common symptom in MDD and it is well-known that good sleep improves mood (reviewed in Riemann et al., 2020). Hence the question: Could improved sleep, by contributing to mood elevation, hinder the development of MDD (reviewed in Riemann et al., 2020)? In agreement, insomnia and/or hyperarousal are thought to increase the risk of depression (and metabolic disorders), providing potential links between depression and a hyperactive orexin system (reviewed in Khan and Aouad, 2017; Riemann et al., 2020). Some experimental evidence supports the involvement of orexins in this interplay (Nakamura and Nagamine, 2017), and there could be bidirectional interactions, with depression favoring insomnia and vice versa, resulting in a vicious cycle. With respect to psychosis, increased dopaminergic activity and putative behavioral correlates have been reported to be reduced by orexin receptor antagonists (Elam et al., 2021; Perez and Lodge, 2021). Cerevance has a 1-SORA, CVN766, under development for a variety of psychiatric conditions (V.B.1.c.).

Finally, both suvorexant and lemborexant have been found to effectively prevent delirium – which is considered a somatic rather than psychiatric condition – in small groups of hospitalized patients (Adams et al., 2020; Matsuoka et al., 2022, 2023).

F. Orexins in Inflammation and Neuronal Cell Plasticity and Death

The role of orexins in neuroinflammation and neurodegeneration including Parkinson's disease (PD) and AD is actively investigated. The questions are interesting and a potential role of orexins in such processes would not be surprising based on their other reported effects on cell plasticity, death and survival (see, e.g., Rouet-Benzineb et al., 2004; Ammoun et al., 2006a; Ammoun et al., 2006b; Sellayah and Sikder, 2012) (II.C–D.) However, many studies suffer from potential methodological issues, as discussed in Kukkonen and Turunen, 2021, and although exogenous orexin stimulation is generally suggested to be protective, conclusions are often difficult to draw. For instance, orexin-A is reported to alleviate neuronal cell death in experimental models of intracerebral hemorrhage (Kitamura et al., 2010). Different mechanisms such as induction of hypoxia-induced factor-1α (Yuan et al., 2011), reduced inflammation (Xiong et al., 2013), inhibition of endoplasmic reticulum stress (Xu et al., 2021) and inhibition of autophagy (Zhang et al., 2022) have been proposed, but the causalities are difficult to demonstrate. A chapter of its own are the cellular model systems. A very popular model for PD research with focus on orexins, SH-SY5Y neuroblastoma cells, are sympathetic ganglion-derived (hence peripheral, not central) noradrenergic (not dopaminergic) neurons, in which no expression of orexin receptors has been demonstrated to date. However, there also are potentially more interesting neuroprotective findings for orexin peptides in PD (Guerreiro et al., 2015; Liu et al., 2018b), though there are still confounding factors in these. One such study reports a functional motor improvement by exogenous orexins in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced PD mouse model (Wang et al., 2019b).

As concerns AD, there is more direct evidence regarding orexins. Brain interstitial fluid amyloid β and tau levels are aligned with the sleep–wake cycle, increasing during wakefulness and decreasing during sleep in mice (IV.C.) (Kang et al., 2009; Holth et al., 2019). Brain interstitial fluid amyloid β levels and accumulation in plaques in the brain are increased by sleep deprivation and icv orexin-A whereas they are decreased by the DORA almorexant (Fig. 4A3) below the background levels, as investigated in the Swedish amyloid precursor protein (APP) mutation and APP+presenilin-1 mutation mouse models (Tg2576 and APPswe/PS1dE9 mice, respectively) (Kang et al., 2009). Intranasal orexin-A worsens disease symptoms and increases amyloid β plaque accumulation in the 3 × Tg mouse model of amyloidogenesis and tauopathy (harboring the previous two mutations + the tau P301L mutation), potentially via increased β-secretase 1 (BACE1) and decreased neprilysin expression (Li et al., 2023). Both tau and amyloid β rodent models show hyperarousal (reviewed in Kron et al., 2024). Orexin neuron count and activation are elevated in the APPswe/PSldE9 mice (Zhao et al., 2022), while the tau P301L mutation mice show a selective reduction of OX₁ mRNA in the locus coeruleus (Keenan et al., 2021). These two datasets appear conflicting, but might be explained by OX₁ downregulation counteracting orexin system hyperactivity.

Some studies find elevated CSF orexin-A levels in AD patients, while others report decreased levels: A meta-analysis failed to detect a significant change in orexin levels in AD (Treu and Plante, 2021). The numbers of orexin-producing neurons and their projections are reduced in AD (Fronczek et al., 2012; Kasanuki et al., 2014; Oh et al., 2019). The apparently conflicting results may be explained by different disease stages, i.e., potentially an increase in early stage and reduction later upon the demise of orexin neurons. However, further consideration is required, given that the immunological methods regularly used to detect CSF orexin levels may lack specificity (VI.). The impact of orexins versus other transmitter systems in the sleep–wakefulness cycle is also difficult to pinpoint in the presence of the generalized neurodegeneration in advanced AD. In some studies, correlations of CSF orexin-A with CSF amyloid precursor protein fragments, tau and phospho-tau levels have been determined (Liguori et al., 2014). Currently, reduction in CSF amyloid β42 (or the ratio amyloid β42/β40) and increase in CSF tau and phospho-T¹⁸¹-tau are associated with AD in clinical diagnostics, though the causality behind this is not fully clear (Olsson et al., 2016; Delaby et al., 2022; Gorelick, 2023).

In summary, orexins stimulate wakefulness which may reduce the glymphatic clearance of amyloid β fragments and (phospho-)tau, which may increase plaque and neurofibrillary tangle formation. One hypothesis is that part of the pathology in AD and frontotemporal dementia may result from distortion of the sleep–wakefulness cycle upon which amyloid β fragments and/or (phospho-)tau, among others, are no longer effectively cleared from the brain by the glymphatic system, but this has not been validated. Orexin neuron overactivity may be seen in AD but it is unclear whether this is (part of) the cause or a result of the AD pathogenesis – or both (reviewed in Wang and Holtzman, 2020). Aging-enhanced orexinergic drive (Li et al., 2022c) should thus be considered with regard to the glymphatic system, should the relationship between sleep and glymphatic clearance hold true (Miao et al., 2024). Elderly patients with NT1 have been reported to have less amyloid plaque formation in the brain than age-matched controls (Gabelle et al., 2019) and the DORA suvorexant (Fig. 4A3) has been reported to have some positive effect in a mouse model of AD and in human AD patients (Herring et al., 2020; Zhou et al., 2020; Lucey et al., 2023). The I308V variant of OX₂ has been reported to confer a slightly increased risk for AD (Gallone et al., 2014). However, the studies are thus far too preliminary and limited to allow conclusions, and the orexin receptor antagonists have not been on the market long enough to allow assessment in a cohort context. Further research in this area is thus clearly required. It should be noted that SH-SY5Y cells have been used as a cellular model in AD research as well, which is equally questionable as in PD.

Several studies address the role of orexins in immune regulation in the CNS or in the periphery. For instance, orexin-A reduces the expression and exosomal delivery of PD-L1 in colon cancer cell lines, which in turn promotes T cell activity against the cancer cells (Wen et al., 2022). Intranasally administered orexin-A reduces mRNA for inflammatory markers in a rat model of cardiac arrest-induced neuroinflammation (Modi et al., 2017). Intrathecally administered orexin-A reduces diabetic neuropathy in streptozocin-induced diabetes model in rats (Niknia et al., 2019). Icv orexin-A ameliorates experimental autoimmune encephalopathy in mice (Becquet et al., 2019). Palmitic acid activates the BV2 microglial cell line whereas orexin-A reduces this activity and the expression of inflammatory markers (Duffy et al., 2015). However, these are isolated findings and thus do not allow definitive conclusions about the physiological role of orexins in these processes or the molecular mechanisms involved.

G. Orexins and Antinociception

Orexin receptor activation, likely via OX₁ receptors, is antinociceptive in animal models (Inutsuka et al., 2016; Iio et al., 2023; Nakamoto et al., 2023; reviewed in Kukkonen, 2013), possibly via endocannabinoid-mediated retrograde presynaptic inhibition (see II.B.6) especially in the periaqueductal gray (Ho et al., 2011; Lee et al., 2016; Chen et al., 2018; Lee et al., 2021). In contrast, orexin-induced analgesia is independent of opioids and opioid receptors, as assessed using naloxone (Bingham et al., 2001; Cheng et al., 2003; Kajiyama et al., 2005; Mobarakeh et al., 2005; Lee et al., 2016). Orexin neurons can be activated upon painful stimuli and their activity may be a normal physiological response to counteract pain (Zhu et al., 2002; Watanabe et al., 2005; Inutsuka et al., 2016). The novel OX₁-selective agonist (R)-YNT-3708 (Fig. 4B1) was reported to be antinociceptive in the mouse tail-flick test (Iio et al., 2023). On the other hand, activation of orexin neurons has been suggested to be involved in itch response, since they are activated by intradermal chloroquine, a histamine-independent pruritogen, and ablation or inhibition of orexin neurons reduces acute and chronic itch; the periaqueductal gray has been associated with this as well (Kaneko et al., 2022; Kaneko et al., 2024). Interestingly, one major group of analgesics, opioids, can induce itch via μ receptor-dependent central mechanisms, i.e., independent of peripheral mechanisms such as histamine (Ikoma et al., 2006; Buddenkotte and Steinhoff, 2010). Potential linking of orexin and opioid systems in itch has not been assessed, but orexins are linked to opioids on the level of reward and addiction (IV.D.).

H. Orexins and Epilepsy

High doses of icv orexin-A can induce behavioral seizures, although this has not been scrutinized with EEG recordings (Ida et al., 1999). REM sleep is antiepileptic due to its attenuation of cortical synchronization, and the kindling model of epilepsy is enhanced by REM sleep deprivation (Frauscher et al., 2016; Ng, 2017). Orexins contribute to increased excitability in the brain (II.C.), hence a potential, though weak, link has been suggested between orexin activity and epilepsy. Blocking of either orexin receptor subtype reduces pentylenetetrazol-induced seizures in rats (Ni et al., 2014; Asadi et al., 2018). Hippocampal optogenetic stimulation-induced seizure intensity in mice is reduced by high-dose (likely nonselective between orexin receptors) SB-334867 (Fig. 4A1). Furthermore, it was found that orexin neuron preseizure activity correlated positively with the seizure intensity and that inhibition of this activity optogenetically or by deep brain stimulation reduced seizure intensity and/or probability (Li et al., 2024). Though sleep deprivation is known to promote epileptic seizures (Roliz and Kothare, 2023), these data indicate that the role of orexins in seizure activity is distinct from their role in sleep and wakefulness. Cannabidiol, used to treat a small subset of epilepsy patients (Franco et al., 2021), has been suggested to act as an orexin receptor antagonist (Vitale et al., 2021), although its affinity for orexin receptors appears too low to explain its physiological effects.

I. Sympathetic, Cardiovascular and Respiratory Regulation by Orexins

Orexin neurons project to brain stem sites regulating sympathetic outflow in rodents (reviewed in Kukkonen et al., 2002; Kukkonen, 2013). This leads to a) increased metabolic activity (at least partly by activation of BAT; IV.B.), and b) cardiovascular effects, e.g., increased mean arterial pressure and heart rate (Narai et al., 2024; reviewed in Kukkonen et al., 2002; Kukkonen, 2013). Both orexin receptor subtypes have been implicated (Shahid et al., 2012; Beig et al., 2015). Recently, intranasal orexin-A administration in humans was shown to increase sympathetic vasoconstrictor action, suggested to take place via changed baroreflex sensitivity, though without concomitant changes in blood pressure or heart rate (Meusel et al., 2022). NT1 patients may have higher heart rate than controls during sleep (Berteotti and Silvani, 2018) and cardiovascular disorders (and risk factors for such) are altogether more frequent in NT1 than in control populations (Jennum et al., 2021). However, other somatic and psychiatric comorbidities are also more frequent in NT1 and thus the causalities are not evident (Jennum et al., 2021). For findings concerning the metabolic rate in NT1 patients and the role of orexins in the heart, see IV.B. and IV.M., respectively.

Several studies point to a role for orexins in respiratory regulation. Orexin neurons project to the pons, medulla and spinal sites responsible for regulation of breathing in rats, exogenous orexins stimulate breathing via these sites, even via direct effect on motoneurons, and PPO-KO mice show impaired central hypercapnic response, which causes central sleep apnea (Young et al., 2005; Dutschmann et al., 2007; Nakamura et al., 2007; Kuwaki, 2008; Yokota et al., 2016). Both orexin receptors have been associated with respiratory regulation (Corcoran et al., 2010; Beig et al., 2015). However, injection of orexin-B or Nag 26 (Fig. 4B2) into the Kölliker–Fuse nucleus of anesthetized mice slowed down the respiratory rate (Varga et al., 2022), while orexin-B within spinal cord attenuated the respiratory depression (suppression of cervical C4 motor root activity) caused by sevoflurane, propofol and remifentanil in a rat medulla–spinal cord preparation (Umezawa et al., 2015).

Within the CNS, respiration is thought to be regulated by the sensing of carbon dioxide upon its acidifying effect (Sharabi et al., 2009; Nattie and Li, 2012). Orexin neurons are sensitive to extracellular pH (Burdakov et al., 2006; Williams et al., 2007). They express the acid-sensitive ion channels ASIC1a (Song et al., 2012) and TASK1 and -3 (Burdakov et al., 2006; Gonzalez et al., 2009b; Wang et al., 2021), although the specific role of each channel type is unclear and some results appear contradictory: ASIC inhibition fully blocks the acid-induced increase in respiratory drive (Song et al., 2012), TASK1 and -2 inhibition has a partial effect (Wang et al., 2021) while TASK1/3-KO does not affect the acid response on respiration (Gonzalez et al., 2009b). Since anandamide, bupivacaine and ruthenium red, used in (Wang et al., 2021), are not selective TASK inhibitors, one conclusion may be that the acid sensitivity of orexin neurons is ASIC1a-dependent (Song et al., 2012), but this is contrasted by other findings suggesting inhibition of a leak-like K⁺ conductance by lowered pH (Williams et al., 2007; Gonzalez et al., 2009b). Part of the response to acidification may come via orexin-mediated enhancement of acid responses of other nuclei (Lazarenko et al., 2011) or vice versa. Intrathecally administered orexin-A sensitizes the respiratory response to hypoxia in rats, which has been interpreted to originate from sensitized peripheral chemoreceptor output (Kim et al., 2016). It has also been suggested that the hypoxia-stimulated C1 cell group could activate orexin neurons (Kim et al., 2016). In contrast, hypoxia alone (weakly) or combined hypoxia and hypercapnia (more strongly) have been reported to inhibit orexin neurons in rat hypothalamic slice preparations (Dergacheva and Mendelowitz, 2017).

In summary, orexin neurons display acid sensitivity themselves and may also have contacts with other neurons with acid-, CO₂- or O₂-sensitivity. Whether orexins then increase or decrease respiration seems to depend on the preparation and conditions in each experimental setting, and thus the physiological role of orexins in the regulation of respiration is unclear. It should be noted that NT1 patients do not have a reduced hypercapnic response (Han et al., 2010) and the DORAs suvorexant and daridorexant (Fig. 4A3) in clinical use do not affect breathing in chronic obstructive pulmonary disease (with reduced hypercapnic response) or obstructive sleep apnea (Sun et al., 2015; Sun et al., 2016; Boof et al., 2021a; Boof et al., 2021b), contrasting with the phenotype of PPO-KO mice and the effect observed in neonatal rats (Nakamura et al., 2007; Kuwaki, 2008; Spinieli et al., 2021).

J. Orexins in Premotor and Motor Control

A few studies have identified orexinergic innervation of lower motoneurons to skeletal muscles and potential excitation of these synapses (Zhang and Luo, 2002; Xi et al., 2003) (see also V.I.). Excitation of brain stem motor nuclei has also been demonstrated (Zhang et al., 2011; Liu et al., 2018a). Lack of orexinergic tone contributes to the lack of muscle tonus during REM sleep and in cataplexy (Xi et al., 2002; Burgess and Peever, 2013). Orexins have been recently linked to anticipation and initiation of voluntary movements (Concetti and Burdakov, 2021; Donegan et al., 2022b) and even motor adaptation (Donegan et al., 2022a), though this may be difficult to distinguish from motor control (Tyree et al., 2023). On the other hand, orexins have been suggested to be centrally involved in the cardiovascular adaptation for locomotion (Narai et al., 2024).

K. Other Putative Orexin Functions in the CNS

Retinal signaling involving at least bipolar, amacrine and ganglion cells (including the melanopsin-expressing ones) is modified by orexins, as indicated by the endogenous expression of the peptides and the receptors, the responses to exogenous orexin-A and the impact of the DORA TCS1102 (Fig. 4A3) on the endogenous signaling (Liu et al., 2011a; Zheng et al., 2015; Ruan et al., 2021; Zhou et al., 2021). Interestingly, orexin signaling has been suggested to modulate processing of visual information in the brain (Chrobok et al., 2017; Chrobok et al., 2021b) and potentially the pupillary light reflex (Chrobok et al., 2021a; Grujic et al., 2023). Disturbances in the visual system are not reported as side-effects of the three clinically available DORAs, suggesting that the visual system is not a major target of orexins or that the DORAs in use do not enter the eye in a sufficient degree. However, the development of the orexin receptor agonist JZP-441 was recently halted partially due to visual disturbances (V.B.2.).

As described under IV.D., orexin signaling contributes to contextual memory with respect to addictive drug use. In addition, it has been reported that orexin signaling is central for “normal” memory functions, as especially investigated with respect to avoidance and spatial memory (see, for instance Jaeger et al., 2002; Telegdy and Adamik, 2002; Aitta-Aho et al., 2016; Akbari et al., 2007; Liao et al., 2024). Thus, increased orexinergic activity during wakefulness plays a role in memory acquisition whereas decreased orexinergic activity during sleep allows for memory consolidation (IV.C.). The memory functions promoted by orexins include social memory (Yang et al., 2013), and social interaction may be promoted by orexins (Abbas et al., 2015; Faesel et al., 2021; Dawson et al., 2023). This links to the psychiatric conditions, especially anxiety and depression, and the actual functions are in many cases difficult to distinguish in animal models (IV.D.).

L. Pituitary Regulation by Orexins

Isolated pituitary cells and cell lines express orexin receptor mRNA and respond to orexins; both increased and decreased release of hormones has been observed (reviewed in Kukkonen, 2013). However, a central question is how the orexins could reach pituitary endocrine cells. As pointed out in Kukkonen, 2013 potential physiological orexin actions could take place via regulation of the release of the known hypothalamic releasing factors and factors alike, but these do not explain the reported direct effects of exogenous orexins on the pituitary gland. There is no evidence based on extensive phase 3 clinical trials or postmarket assessments that DORAs affect pituitary functions.

M. Other Potential Orexin Functions in the Periphery

A few peripheral tissues/organs have been reported to express orexin receptor or PPO protein or mRNA or respond to ex vivo orexin stimulation. While the receptor protein expression data based on receptor antibodies are uncertain (IV.A.), other data may be more convincing. ³H-EMPA autoradiography suggested the presence of very low levels of OX₂ in the rat adrenal cortex (Mitsukawa and Kimura, 2022), whereas OX₂ mRNA levels and ex vivo functional assays suggested significant receptor levels (Jöhren et al., 2001; Karteris et al., 2005). Another study reported expression of OX₂ mRNA and potential functional responses in the rat testis (Karteris et al., 2004), where, however, no ³H-EMPA binding was detected (Mitsukawa and Kimura, 2022). It should be noted that mismatches between mRNA and protein levels are not uncommon.

OX₁ receptors were found in rat osteoblast precursor cells (Ziolkowska et al., 2007). PPO-overexpressing mice and OX₁-KO mice were shown to have a higher bone mass, whereas PPO-KO mice and OX₂-KO mice appeared to have lower bone mass (Wei et al., 2014). OX₁ was suggested to promote the adipocyte differentiation instead of osteoblast differentiation by suppressing local ghrelin production. The local PPO production in bone was proposed to come from mesenchymal stem cells, osteoblasts and bone marrow adipocytes, since these expressed PPO mRNA. The OX₂-dependent effect was instead suggested to require central OX₂ signaling-mediated suppression of leptin expression. However, the causality of ghrelin and leptin in these processes was not verified (Wei et al., 2014). In contrast, OX₁ receptor activation was reported to increase osteoblastic differentiation of the mouse cell line MC3T3-E1 via p38 MAPK activity and hypoxia-induced factor-1α induction and to inhibit hypoxia-induced reduction in osteogenic factors (Han et al., 2018; Gu et al., 2022). Subcutaneous orexin-A injections reversed hypoxia-induced bone mass loss in mice (Gu et al., 2022). Thus, there are only a few studies of the role of orexins in bone remodeling and the studies are in apparent contradiction as concerns the role of OX₁.

Placentally produced orexins have been suggested to contribute to prenatal development of BAT in mice, though these data have been questioned (IV.B.). An SNP upstream of human HCRTR2 gene conferred a poorer prognosis in systolic heart dysfunction and after myocardial infarction (Perez et al., 2015; Wohlfahrt et al., 2023) (III.). While the reported lower protein levels for this polymorphism are not reliable (see above), OX₂-KO mice were more prone to diastolic dysfunction. Wildtype mice were also partially protected by orexin-A infusion in the angiotensin II and isoproterenol chemical stress model of heart failure (Perez et al., 2015). Rat cardiomyocytes were reported to express OX₂ receptor mRNA and isolated cells contracted upon orexin-B or Ala¹¹, d-Leu¹⁵-orexin-B challenge (Patel et al., 2018). Orexin-B and/or Ala¹¹, d-Leu¹⁵-orexin-B gave partial protection against the ischemic injury in the isolated rat heart and in intact rats and mice (Patel et al., 2018; Jiao et al., 2024). Even human heart muscle samples were shown to respond to orexin-B. It was recently suggested that the effect of exogenous orexin stimulation (here Ala¹¹-d-Leu¹⁵-orexin-B) is inhibition of sensory neurons which further inhibit excessive sympathetic activation (Jiao et al., 2024). However, none of the studies has identified any endogenous orexin source. PPO-KO mice have higher numbers of circulating monocytes and neutrophiles (McAlpine et al., 2019); the authors suggest that the link is via hypothalamically produced orexins that are released in the blood stream where they act upon bone marrow to suppress production of colony-stimulating factor 1.

Any putative peripheral orexin signaling has the problem of PPO production. Are orexins coming from a peripheral source and used a) locally or b) systemically via circulation, or would central orexin neurites c) target peripheral tissues directly or d) release orexins in the circulation? There is little evidence for c except for the pineal gland (IV.C.), and thus this possibility might be excluded. As concerns circulating orexin (alternatives b and d), the release of orexins would need to be quite high. Using radio- and enzyme immunoassay kits, highly variable plasma levels of orexin-A have been measured in both human and rodents. For instance, a recent very high impact paper reports about 20 pg/ml in mice (McAlpine et al., 2019), which corresponds to 5.6 pmol/l and total of 6.2 fmol (the plasma volume = 25 g × 80 μl blood/g × 0.55 = 1.1 ml). The reported numbers of orexin neurons in the mouse brain are quite variable, e.g., ranging between 440 and 2300 (Obukuro et al., 2010; Black et al., 2018; Gonzalez and Prehn, 2018; Berteotti et al., 2021). Assuming there were 540 neurons (the median of these studies), every neuron would have to produce 0.011 fmol of PPO to reach this a plasma concentration. ¹²⁵I-orexin-A and ¹²⁵I-orexin-B have half-lives of 30.7 minutes and 4.4 minutes, respectively, in the mouse blood (Kastin and Akerstrom, 1999). Keeping a steady-state concentration of 5.6 pmol/l of orexin-A would thus require an additional constant production, processing, transport and release of 0.14 fmol/min orexin-A, which is equal to 155,000 molecules/min/neuron. In some human studies plasma levels below 0.3 pmol/l are reported while other studies report levels as high as 700 pmol/l (summarized in Mäkelä et al., 2018). The metabolism of orexin-A in the human plasma has not been investigated but its half-life in rats has been suggested to be similar to mice (Ehrström et al., 2004) and thus we may assume it to be similar even in humans. The number of orexin neurons in the human brain can be approximated with 80 000 (Kukkonen et al., 2002). Thus the corresponding values for 5.6 pmol/l in human plasma would be 0.19 fmol/neuron + 350 fmol/min/neuron (2 600 000 molecules/min/neuron). We can compare this to human pancreatic β-cells. The reports are not fully consistent but, for instance, the normal fasting plasma insulin level in the clinical tests (Helsinki University Hospital Laboratory; Helsinki, Finland) is 12–120 pmol/l. Assuming a half-life of 5 minutes, this would correspond to 2800–28,000 molecules/min/cell (in all 10⁹ β-cells), which is similar to published studies (Polonsky et al., 1988; Ramchandani et al., 2010; Knopp et al., 2019). It, thus, appears that either the central orexin neurons or the peripheral tissues (PPO expression has been reported in gastrointestinal tract, adrenal gland, epididymis, testis, pancreas, kidney and placenta; Kukkonen, 2013) would have to be quite efficient (neuro)endocrine organs, which seems unlikely. Surprisingly, more than hundredfold higher circulating orexin-A levels than 5.6 pmol/l have been reported (e.g., 790 pmol/l; Goldstein et al., 2021), making the comparison even more absurd at the current level of understanding. The most likely explanation lies in the failure of the antibody-based methods in orexin detection (VI.), but even other factors need to be considered, for instance the much higher number of orexin neurons in relation to the blood volume in rodents (which have been the subject for most studies), potentially different pharmacokinetics of orexin-A in different species and maybe pulsatile secretion, but still there are central questions to be answered before the alternatives b and c would make any sense. Thus, a paracrine signaling (alternative a) would be most logical, though there is a mismatch of the receptor and peptide expression (above).

The final significant issue of peripheral orexin receptors concerns their expression in cancer cells (II.D.), which currently rather has bearing on pharmacology than physiology.

In summary, there are potentially very interesting findings of the peripheral role of orexins and orexin receptors. However, the findings are isolated and, as viewed in the light of our current understanding, both inexplainable and contradictory. There is also evidence for a mismatch between orexin mRNA expression and functional responses to exogenous orexins on the one hand and peripheral PPO peptide or mRNA expression on the other (Kukkonen, 2013); in most cases, there is no clear source for orexin peptides and thus the receptor stimulation is uncertain. Thus, alternative explanations for the orexin receptor expression should be considered, such as roles for orexins under certain conditions (induced expression?), other ligands for the receptors, constitutively active receptors, other roles for the receptors (and the peptides), or an altogether superfluous role of the peripheral orexin receptors (could these be evolutionary relics?). Clearly, more research is required. One source of information will be the potential phase 4 trials and other postmarketing data on DORAs.

A. Orexin Receptor Agonist Peptides

The original study that reported both human orexin receptors and native peptides (Sakurai et al., 1998) demonstrated that both OX₁ and OX₂ were activated by orexin-A and -B when recombinantly expressed in CHO-K1 cells. While the peptides had equal affinity and potency for OX₂, orexin-A showed an approximately 100-fold higher potency for OX₁ compared with orexin-B, i.e., orexin-B was deemed OX₂-selective whereas orexin-A did not distinguish between OX₂ and OX₁ receptors. However, later studies have demonstrated that the ratio of the peptide potencies on OX₂ depends on the response measured, pointing at biased signaling (II.F.).

The C-termini of orexin-A and -B are highly conserved (Fig. 1), and studies involving peptide truncation and amino acid exchange have confirmed its central role in orexin receptor binding and activation (Darker et al., 2001; Okumura et al., 2001; Ammoun et al., 2003; Lang et al., 2004; Lang et al., 2006) (see below). In 2021, a study presented an orexin-B-bound human OX₂ cryo-EM structure, revealing for the first time the structure of a potentially active orexin receptor state (Hong et al., 2021). As expected, the C-terminal region of orexin-B bound deep in the binding pocket, occupying the earlier described small molecule binding site (see III.A.), in a complementary shape. The structure of the peptide was extended, meaning that the α-helical solution structure (see below) of orexin-B's C-terminus is not preserved in the binding mode. Because of the low resolution of the structure on the extracellular site, the N-terminal part of orexin-B was excluded from the final structure.

Successive truncation of the orexin peptides from the N-terminal end is initially well tolerated, but the potency starts to gradually decrease when reaching beyond the first 12 or so amino acids (Darker et al., 2001; Okumura et al., 2001; Ammoun et al., 2003; Lang et al., 2004; Lang et al., 2006). The shortest peptides acting in the nanomolar range have been described to be composed of the 15–19 C-terminal amino acids (Darker et al., 2001; Ammoun et al., 2003; Lang et al., 2004; German et al., 2013), while an orexin-A variant as short as 7 amino acids has been demonstrated to have moderate activity on OX₂ at 10 μM concentration (Ammoun et al., 2003); the potency is naturally also dependent on the receptor expression levels and the cellular background. By contrast, C-terminal manipulation of the peptides is not tolerated (Darker et al., 2001; Lang et al., 2004).

In most studies, reduction of the disulfide bridges or substitution of the cysteines with alanines leads to 10-fold or less decrease in the potency for OX₁ and OX₂ (Okumura et al., 2001; Lang et al., 2004; unpublished studies by Kukkonen, Rinne et al.). However, potential spontaneous (re)formation of the disulfide bridges has not been investigated in any study. Orexin-A modifications are generally better tolerated by OX₂ than OX₁ (Darker et al., 2001; Okumura et al., 2001; Holmqvist et al., 2002; Ammoun et al., 2003; Zhu et al., 2003; Lang et al., 2004; Lang et al., 2006; Johansson et al., 2007; Putula et al., 2011). Substitution with alanine, d-amino acid, etc. in orexin-A and -B has also offered little conclusive information (Darker et al., 2001; Ammoun et al., 2003; Lang et al., 2004; Lang et al., 2006). The putative OX₂-selective agonist peptide Ala¹¹, d-Leu¹⁵-orexin-B may represent an exception (Asahi et al., 2003), though its actual selectivity for OX₂ is still debated and may be signaling pathway-dependent (Asahi et al., 2003; Putula et al., 2011; Yamamoto et al., 2022). Nevertheless, Ala¹¹, d-Leu¹⁵-orexin-B is frequently used to support OX₂-mediated effects. The use of agonists of limited selectivity is generally not recommended in pharmacological studies as there is always potential for biased signaling, hence the use of combinations of an agonist and at least two antagonists (one 1-SORA and one 2-SORA) should be the norm (Kukkonen, 2019).

NMR structures of orexin-A and orexin-B both in aqueous solution (Lee et al., 1999; Kim et al., 2004; Takai et al., 2006) and bound to sodium dodecyl sulfate (SDS) micelles have been published (Miskolzie and Kotovych, 2003; Miskolzie et al., 2003). The set of NMR structures of orexin-A show bent or straight conformations, while orexin-B always adopts a bent conformation. In solution, the orexin-B structure is composed of two α-helical regions (L7–G19 and A23–M28; 1CQ0), while orexin-A is composed of three (C6–G9, L16–A23 and N25–T32; 1WSO) or two α-helical regions (C14–H21 and N25–T32; 1R02). In micelles, orexin-A is composed of the α-helices D5–Q9 and L16–G22 and orexin-B of α-helices L7–S18 and A22–L26. While the secondary structure of the C-terminus of orexin-A is unclear (the last 11 residues), the end of the C-terminus in orexin-B (T27 and M28) is clearly nonhelical (Miskolzie and Kotovych, 2003; Miskolzie et al., 2003). The micelle-bound structures, with unstructured C-termini, were assumed to mimic the membrane-bound structures of the peptides; thus, the conformational freedom was suggested to potentially have a role in receptor binding via the membrane interface (Miskolzie and Kotovych, 2003; Miskolzie et al., 2003) (see below). In the NMR structures, the α-helical regions are connected by a kink or a linker. It was hypothesized that the truncation of peptides destabilizes their α-helical structures, and thus leads to a decrease in the binding affinity (Karhu et al., 2018). Chemically stabilized ("stapled") α-helical parts of C-terminal orexin-A fragments were thus examined for orexin receptor activation, but all stapled peptides were found to be less potent than the nonstapled peptide of the same length, likely due to a reduced binding affinity (Karhu et al., 2018). The reason for the lower potency remained unclear at the time, but it was speculated that the staple itself or a too rigid peptide conformation hindered the receptor binding or activation. Later the 3D-structure of orexin-B-bound OX₂ revealed that the C-terminal region of orexin-B does not retain α-helical conformation in the binding mode (Hong et al., 2021), offering a likely explanation to the behavior of the stapled peptides (see III.A.).

Orexins are amphipathic peptides, possessing hydrophilicity and hydrophobicity on the opposite sides of the α-helix 1 (Lee et al., 1999; Miskolzie and Kotovych, 2003; Miskolzie et al., 2003; Kim et al., 2004; Takai et al., 2006). A recent study investigating the binding of orexin-A to synthetic lipid bilayer vesicles (Ball et al., 2023) demonstrated that the α-helix content of orexin-A increases when bound to the membrane and that the C-terminal helix aligns in parallel to the surface of the membrane. The significance of this observation is unclear, but the authors speculated that this would “prepare” the peptide for receptor binding (Ball et al., 2023), as was also proposed for the NMR structures of the micelle-bound orexin peptides (see above). The corresponding L20A and L15A mutations in N-terminally truncated orexin-A and orexin-B variants, respectively, reduce the potencies by more than 10-fold (Darker et al., 2001; Ammoun et al., 2003; Lang et al., 2004; German et al., 2013) which has been suggested to be caused by reduced membrane lipid binding of the lipophilic helix side (Miskolzie and Kotovych, 2003; Miskolzie et al., 2003), though it is unclear why the –CH₃ sidechain would be so much less lipophilic than –CH(CH₃)₂. However, the C-terminus of orexin-B loses its α-helical conformation when bound to the reported active receptor conformation (Hong et al., 2021) and the same is seen with micelle-bound orexin-A (see III.A.). In essence, structural information gained from studies with unbound peptides is of limited utility, since the same peptide can likely adopt multiple conformations depending on the target receptor, the complexed accessory proteins involved in the signaling apparatus, etc.

An apparent requirement of extracellular Ca²⁺ for the binding of ¹²⁵I-orexin-A to OX₁ has been reported, which suggests a similar effect on native orexin-A (Putula et al., 2014). It is unclear whether the total binding capacity or the affinity is affected (Lund et al., 2000; Johansson et al., 2007; Putula et al., 2014). It also remains unclear whether the property is related to the peptide or the receptor, but Ca²⁺ did not have any effect on the circular dichroism (CD) spectrum of orexin-A or its stability at different temperatures (Putula et al., 2014). If Ca²⁺ affects the receptor, it is difficult to determine whether the effect is intracellular or extracellular, since manipulation of the extracellular Ca²⁺ levels affects the intracellular levels, especially close to the plasma membrane. The fast Ca²⁺ buffer BAPTA (10–20 mM) has been introduced inside the cells in patch-clamp studies of orexin receptor signaling (Liu et al., 2002; Burdakov et al., 2003; Acuna-Goycolea and van den Pol, 2009; Kargar et al., 2015; Ishibashi et al., 2016; Usui et al., 2019; Jin et al., 2021). Its impact ranges from no effect to almost total inhibition of the response. All studies have used high concentrations of orexins (0.1–1 μM); thus, it is not possible to judge whether there is any effect of BAPTA on the orexin concentration–response relationship in those studies in which BAPTA was not effective.

B. Small Molecule Orexin Receptor Ligands

Most drug discovery efforts in both industry and academia have focused on orexin receptor antagonists. The original indications for therapeutic intervention using orexin receptor antagonists were obesity and insomnia, illnesses that affect a large proportion of the total population and are in urgent need of novel therapies. Conversely, the obvious indication for orexin receptor agonists, narcolepsy with cataplexy, is rare – although many of the recent new drug approvals are for rare or very rare diseases. Small molecule nonpeptide agonists are generally much more difficult to discover/develop than antagonists, especially in the absence of a good pharmacophore model, as is often the case with neuropeptides and their receptors (Hoyer and Bartfai, 2012). Thus, there is a great number of orexin receptor antagonist structures published, with many antagonists undergoing clinical studies and three DORAs on the market. Unfortunately, only a few antagonist structures have been thoroughly characterized. Much fewer agonist structures have been published, and two are in clinical trials. These advances are summarized below. Further details on orexin receptor drug discovery and development can be found in recent reviews (see, e.g., Perrey and Zhang, 2020; Jacobson et al., 2022; Bonifazi et al., 2023).

C. Labeled and Other Types of Special Ligands

¹²⁵I-orexin-A has been the “traditional” radioligand as it was the only one commercially available for years, and ¹²⁵I-orexin-B is also available. Agonists are theoretically less suitable as radioligands, and orexins (and their labeled variants) have additional biophysical and chemical properties that make them even less optimal (Kukkonen, 2012). Some ³H-labeled antagonists have become available more recently; the current commercial selection contains at least ³H-SB-674042, ³H-almorexant, ³H-EMPA and ³H-TCS1102/-BBAC, which had originally been custom-labeled by pharmaceutical companies for internal purposes. Radiolabeled orexin receptor antagonists have only been characterized to a limited extent, with some exceptions (Langmead et al., 2004; Mould et al., 2014). ¹²⁵I-orexin-A is also used in the standard radioimmunoassay for measurement of orexin peptides in the CSF for research purposes and in narcolepsy diagnostics. Radiolabeling (³H) of the agonists danavorexton/TAK-925 and firazorexton/TAK-994 has also been reported (Ishikawa et al., 2022; Ishikawa et al., 2023).

There have been a few attempts to generate orexin receptor-specific small molecule positron emission tomography (PET) tracers. Such tracers might be useful in detecting either orexin receptor levels (when able to effectively compete away endogenous orexins) or orexinergic neuronal activity (when competed away by endogenous orexin peptides). There may be additional advantages and disadvantages of antagonist and agonist probes (Colom et al., 2019). The currently published orexin receptor positron emission tomography probes, when taken up in the brain, show rather high nonspecific binding as identified by, for instance, labeling of areas known not to express significant amounts of orexin receptors. Thus, positron emission tomographic imaging cannot currently be reliably applied in orexin research.

Orexin peptides have been fluorescently custom-labeled (Darker et al., 2001; Evans et al., 2001; Karteris et al., 2001), and both fluorescent and biotin-labeled orexin-A and -B are also commercially available, but have not been widely used. A time-resolved fluorescence resonance energy transfer-based binding assay utilizing labeled orexin-A and orexin receptors is also commercially available.

Nonradioactive heavy labels (¹³C, ¹⁵N) have been used in custom-made internal orexin-A and -B standards for mass spectrometry-based detection assays (VI.) (Hirtz et al., 2016; Bårdsen et al., 2019; Lindström et al., 2021; Hopkins et al., 2021; Ball et al., 2023; Narita et al., 2023) and in NMR (III.A., V.A.). A caged version of orexin-B – that can be uncaged upon a UV or near-UV light pulse – has recently been developed (Duffet et al., 2022b). This can be used in discrete temporal or regional settings which may allow for increased probing of the functions of orexin-B in the modulation of synaptic transmission and other physiological processes. A photoactivatable/-inactivatable orexin-B_6–28 version (photorexin) has also been reported; however, the achieved degree of photoisomerization does not appear to be able to fully turn the compound off (Prischich et al., 2024).