Evolution and divergence of teleost adrenergic receptors: why sometimes ‘the drugs don't work’ in fish Free

In response to altered metabolic demands and stress, adrenaline and noradrenaline, released as hormones and/or neurotransmitters, co-ordinate a multitude of interacting physiological responses. In teleost fishes, the functions under adrenergic control include cardiorespiratory physiology (Altimiras et al., 1997; Hanson et al., 2006; Pan et al., 2023; Wood, 1976), tissue metabolism (Fabbri and Moon, 2016; Milligan, 1996), osmoregulation (Kumai et al., 2012; McCormick, 2001), splenic contraction (Holmgren and Nilsson, 1975; Joyce and Axelsson, 2021), swim bladder inflation (Dumbarton et al., 2010), pigmentation of the skin (Burton, 2002; Wang et al., 2009; Xu and Xie, 2011) and even behaviour (Marrone et al., 1966).

The actions of adrenaline and noradrenaline depend upon binding of these ligands to specific G-protein coupled receptors (GPCRs), adrenergic receptors (adrenoreceptors; ARs) (Bylund et al., 1994). The capacity for tissue-specific responses to these ligands hinges on the expression of different types of ARs in different organs, as well as compartmentalization of subsequent intracellular signalling cascades. Some organs such as the brain express many different ARs, presumably with regional and cellular heterogeneity, whereas others express only a limited complement of one or two ARs (Chen et al., 2007; Ruuskanen et al., 2005a; Wang et al., 2009) (Fig. 1A). Broadly, there are three types of AR: α1-, α2- and β-ARs, which were defined in mammals through a combination of pharmacological and molecular approaches (Box 1). In mammalian species, there are three α1-ARs (α1A, α1B, α1D), three α2-ARs (α2A, α2B, α2C) and three β-ARs (β1, β2, β3) (Strosberg, 1993). α1-ARs are typically involved in contraction of smooth muscle through modulation of Ca²⁺ signalling (Fig. 1B; Chen and Minneman, 2005; Docherty, 2019), α2-ARs mediate neuronal negative feedback mechanisms at the pre-synaptic terminal (Fig. 1C; Aantaa et al., 1995) and β-ARs classically activate stimulatory G-proteins (G_s) and are famous for their regulation of cardiac contractility and heart rate, as well as other diverse functions (Fig. 1D; Vasudevan et al., 2011).

A large and ever-growing body of work is devoted to the study of adrenergic signalling in teleosts. These studies include experiments aimed at understanding the basic physiology of fish (Abramochkin et al., 2022; Dumbarton et al., 2010; Joyce et al., 2022; Steele et al., 2011), as well as applied work employing zebrafish as models for human diseases or for drug discovery (Giacomotto and Ségalat, 2010; Li et al., 2021; Maciag et al., 2022, 2023; MacRae and Peterson, 2015; McMacken et al., 2018; O'Daniel and Petrunich-Rutherford, 2020; Sysoev et al., 2019). The effect of some adrenergic drugs, particularly β-blockers (i.e. β-AR antagonists), on fish is also under intensive study in an ecotoxicological context, as they are known to pollute aquatic environments (Fabbri and Moon, 2016; Giltrow et al., 2011; Owen et al., 2007).

Many of the studies employing adrenergic pharmacological agents in fish presume they have equivalent actions to those in mammals (Owen et al., 2007), where their specificity has been defined. In many cases, this may hold true (Ruuskanen et al., 2005b). However, easily overlooked are the abundant historical reports of atypical, ambiguous or otherwise unusual pharmacological properties of fish ARs published over the past decades (Brighenti et al., 1987; Fabbri et al., 1998, 1999a; Johansson, 1984; Jozefowski and Plytycz, 1998; Lortie and Moon, 2003; Marshall et al., 1993; Steele et al., 2011; Stene-Larsen, 1981; Wood, 1976). In recent years there have been major advances in our understanding of the evolution of ARs in vertebrates (Céspedes et al., 2017; Joyce, 2023; Spielman et al., 2015; Zavala et al., 2017). These include prominent examples of genome duplication and pseudogenization (see Glossary) endowing different vertebrate groups with distinct genetic repertoires of ARs. The purpose of this Review is to present some of the challenges associated with studying fish ARs with pharmacological agents that have been defined by studies in mammals. Our aim is not to discredit or discourage pharmacological studies in fish, but to provide information on how they should be designed and interpreted appropriately in the context of the evolutionary history of ARs. Building on earlier arguments (see Fabbri and Moon, 2016), we discuss how comparative pharmacology can be powerfully combined with molecular biology, including the study of gene expression and generation of knockouts. We also draw attention to fertile areas for future research that are emerging from contemporary advances in ‘omics’ technologies (e.g. comparative genomics and transcriptomics; see Glossary).

The genetic basis of the pharmacological classification of ARs was established in the late 1980s and early 1990s during an intense period of gene cloning and sequencing in which the nucleotide and inferred protein sequences were determined for a range of mammalian β-ARs (Dixon et al., 1986; Emorine et al., 1989; Frielle et al., 1987), α1-ARs (Cotecchia et al., 1988; Lomasney et al., 1991; Perez et al., 1991; Ramarao et al., 1992; Schwinn et al., 1990; Voigt et al., 1990) and α2-ARs (Kobilka et al., 1987; Lomasney et al., 1990; Regan et al., 1988; Weinshank et al., 1990; Zeng et al., 1990). These studies proved instrumental in consolidating our understanding of previously ambiguous classifications (Box 1; Bylund et al., 1994; Hieble et al., 1995). Gene cloning was later expanded to sequence ARs in other species of vertebrates including teleosts (Chen et al., 2007; Giltrow et al., 2011; Nickerson et al., 2001, 2003; Ruuskanen et al., 2004; Svensson et al., 1993) and amphibians (Devic et al., 1997; Mori et al., 2013). The advent of whole-genome sequencing and advances in automated genome annotation have enabled the retrieval of accurate predicted gene sequences from a vast array of species representing all of the major vertebrate lineages (Aris-Brosou et al., 2009; Céspedes et al., 2017; Joyce, 2023; Zavala et al., 2017). This lends itself to phylogenetic and other comparative genomic analyses (e.g. conserved synteny, see Glossary) to elucidate the evolutionary history and interrelationships of the receptor subtypes.

Within each major type of AR there are multiple (three to four in most vertebrates) subtypes, each encoded by a different gene. This diversification presents a prime example of the ‘two rounds’ (2R) whole-genome duplication (WGD) hypothesis (Aris-Brosou et al., 2009; Céspedes et al., 2017; Ruuskanen et al., 2004; Wang et al., 2023; Zavala et al., 2017). The 2R hypothesis, pioneered by Susumu Ohno (1970), describes how after diverging from chordate invertebrates, the common ancestor of jawed vertebrates underwent two rounds of WGD, generating up to four copies (paralogues; see Glossary) of each given gene (Dehal and Boore, 2005). Cyclostomes (jawless vertebrates) likely diverged from the jawed vertebrate lineage after the first (1R) WGD (Joyce et al., 2023; Nakatani et al., 2021; Yu et al., 2023 preprint). Candidates for gene families that expanded in the 2R WGDs are thus expected to accord to the so-called 1:4 rule, wherein four paralogues of a gene can be traced to the common ancestor of jawed vertebrates, where only a single gene is known in invertebrates (Meyer and Schartl, 1999), or traced to a potentially monophyletic group owing to lineage-specific gene or genome duplications. In reality, the 1:4 rule is not strictly adhered to because psuedogenization, i.e. non-functionalization, is a common fate of duplicated genes with redundant functions (Prince and Pickett, 2002). Nevertheless, it is well recognized that the 2R WGDs generated protein diversity responsible for many core physiological functions in vertebrates, including oxygen-transport globins (Hoffmann et al., 2012; Opazo et al., 2015), muscle sarcomeric proteins (Joyce et al., 2023), ion channels (Mackrill and Shiels, 2020), hypoxia-inducible factors (Mandic et al., 2021; Townley et al., 2022) and various neurohormonal receptors (Brunet et al., 2016; Hwang et al., 2013; Ocampo Daza et al., 2022; Pedersen et al., 2018; Ravhe et al., 2021), including ARs (Aris-Brosou et al., 2009; Céspedes et al., 2017; Wang et al., 2023; Zavala et al., 2017). Importantly, there is also evidence for a third (3R) WGD in the common ancestor of teleosts, generating an additional array of genes (Jaillon et al., 2004; Near et al., 2012; Santini et al., 2009; Taylor et al., 2001, 2003). Some teleost lineages, such as salmonids (Berthelot et al., 2014; Lien et al., 2016; Macqueen and Johnston, 2014), have undergone a fourth (4R) WGD which doubled the genetic complement yet again.

Numerous studies have approached the evolution of β-ARs (where adrb1, adrb2 and adrb3 encode β1-, β2- and β3-ARs, respectively) with phylogenetic methods (Aris-Brosou et al., 2009; Giltrow et al., 2011; Imbrogno et al., 2022; Joyce, 2023; Nickerson et al., 2003; Wang et al., 2009; Zavala et al., 2017). Although some preliminary work presumed (Ruuskanen et al., 2004) or found (Giltrow et al., 2011) the β1- and β2-ARs to be most closely related to each other, subsequent studies using more comprehensive and robust approaches (incorporating more outgroups) typically recovered β1 and β3 as sister groups (Aris-Brosou et al., 2009; Imbrogno et al., 2022; Joyce, 2023; Zavala et al., 2017). This indicates that the 1R WGD generated both the common ancestor of adrb1 and adrb3 and the common ancestor of adrb2, as well as a now extinct gene, adrb4 (Zavala et al., 2017). The 2R WGD resulted in the separation of adrb1 and adrb3, as well as adrb2 and adrb4, although the latter quickly underwent pseudogenization prior to the divergence of extant jawed vertebrates (Zavala et al., 2017).

The 3R event in teleosts generated duplicated copies of genes for β1-, β2- and β3-ARs. Most extant teleosts have only a single β1-AR gene (adrb1), indicating the redundant copy was most likely lost soon after its origin (Zavala et al., 2017). Nevertheless, many lineages retain two β2-AR genes (adrb2a and adrb2b) and some retain two paralogues of β3-AR (adrb3a and adrb3b). Zebrafish (Danio rerio), for example, have one gene for β1-AR and two for each of β2-AR and β3-AR (Wang et al., 2009). Comparative phylogenetics (see Glossary) have also helped clarify some pharmacological anomalies amongst β-ARs. For example, in the turkey, a ‘novel’ avian β-AR, termed β4C-AR, was identified (Chen et al., 1994) and received notable attention (Baker, 2010). However, phylogenetic studies confidently place it as a β3-AR homologue (Aris-Brosou et al., 2009; Nickerson et al., 2003). An unfortunate legacy of this misnomer is that many adrb3 genes, including those of teleosts (Giltrow et al., 2011), have been automatically annotated as adrb4C orthologs in databases such as NCBI GenBank and Ensembl. This exemplifies the utility and necessity of evolutionary studies in establishing a universal classification system.

In light of the vast body of literature on adrenergic regulation in rainbow trout (Oncorhynchus mykiss) over the past 50 years (e.g. Abramochkin et al., 2022; Farrell et al., 1986; Gamperl et al., 1994; Nickerson et al., 2003; Wood and Shelton, 1975), the effect of the salmonid 4R WGD on the β-AR repertoire was recently studied (Joyce, 2023). Salmonids typically have seven adrb genes, including duplicates of adrb2a, adrb2b and adrb3a in addition to a single adrb3b. Surprisingly, salmonids lack adrb1, and are the first vertebrate lineage known to lack this gene. It may be little coincidence that salmonids have a boosted repertoire of adrb2 and adrb3 genes, presumably to ensure functional redundancy to compensate for the lack of adrb1 (Joyce, 2023). Further work is required to understand the physiological consequences of the loss of adrb1 in salmonids.

Compared with β-ARs, scant attention has been devoted to the evolution of α1-ARs in vertebrates (Aris-Brosou et al., 2009; Chen et al., 2007) but they appear to have followed a similar trajectory. Paralogous genes (adra1a, adra1b and adra1d) of each of the three α1 subtypes (α1A, α1B- and α1D-ARs, respectively), with little deviation, have been identified in each jawed fish lineage, forming monophyletic clusters (see Glossary; Aris-Brosou et al., 2009; Chen et al., 2007). α1B and α1D are sister proteins so appear to have shared a common ancestor after 1R and diverged in the 2R WGD, whilst the sister gene to α1A became extinct. Teleosts possess duplicated copies of genes for α1-AR (adra1aa and adra1ab) and α1B-AR (adra1ba and adra1bb) but generally lack a second copy for α1D-AR.

The evolution of α2-ARs is somewhat more complex because a fourth gene family (i.e. beyond the α2A, α2B and α2C triad defined in mammals), encoding the α2D receptor, is found in teleosts and other non-mammalian lineages (Céspedes et al., 2017; Ruuskanen et al., 2004). Given its presence in all lineages of jawed vertebrates except crocodilians and mammals, it is most parsimonious to assume that all four α2-AR genes were present in the common ancestor of jawed vertebrates, and likely diversified in the 2R WGDs. However, this is somewhat ambiguous, as both Aris-Brosou et al. (2009) and Céspedes et al. (2017) reported α2D-AR to be sister to a monophyletic group composed of the other three α2-ARs (α2A, α2B and α2C). If the α2-AR family diversified in the 2R WGD, each subtype would be predicted to be sister to one other family of paralogues (in two sets of two); this ambiguity awaits resolution in future work. α2D-AR was later convergently lost in the mammalian and crocodilian lineages (Céspedes et al., 2017). Teleost-specific duplications, attributable to 3R, can be identified for α2A-, α2C- and α2D-ARs, but the duplicated α2B was lost near the base of teleost phylogeny and is not present in any extant representatives (Céspedes et al., 2017).

Although these phylogenetic studies strongly support that the major AR subtypes (α1 and α2 and β) are monophyletic, the wider relationships between the types are more contentious. Some earlier studies suggested that the different AR types are closely related to each other, so could potentially have arisen from duplications of a common ancestral AR gene (Aris-Brosou et al., 2009; Owen et al., 2007; Ruuskanen et al., 2004; Yang-Feng et al., 1990). However, broader comparisons with other monoamine receptors, such as serotonin and dopamine receptors, indicate α1-, α2- and β-ARs are not particularly closely related to one another amongst other monoamine receptors, suggesting that their specialized affinity for catecholamines could have evolved separately at least twice or all three times (Fig. 2; Callier et al., 2003; Candiani et al., 2005; Céspedes et al., 2017; Le Crom et al., 2003; Ravhe et al., 2021; Spielman et al., 2015; Yamamoto et al., 2013; Zavala et al., 2017). Together with structural modelling studies, this shows how minor sequence changes can shift ligand affinity with relative lability (Xhaard et al., 2006). Given the apparent phylogenetic disparity between α1- and α2-ARs (Spielman et al., 2015), it is remarkable that they have converged upon shared specificity for some general α-AR pharmacological agents (such as phentolamine; Table 1). In other words, in our current classification system, the α1 and α2 receptors are grouped together rather artificially on the sole basis of limited pharmacological similarities and not evolutionary ancestry.

Evolutionary studies have also been useful in determining how natural selection has acted upon the ARs. For example, there is evidence for a burst of positive selection on α1-ARs in the period shortly after their diversification (Chen et al., 2007). Other studies focused on mammals have shown how α2B-AR (Madsen et al., 2002), β2-ARs and β3-ARs (Cagliani et al., 2009) have undergone rapid evolutionary change attributable to selection. This emphasizes how the pharmacological and physiological differences between receptor subtypes must have emerged immediately following the 2R WGDs, but could later have been subject to lineage-specific changes.

Taken together, this evolutionary insight emphasizes the vulnerability of transferring mammalian classification of receptor subtypes, founded upon pharmacological studies, to other vertebrates including fishes. In particular, the actions of genome duplication (including the teleost-specific 3R WGD) and independent cases of gene loss in some lineages (such as the loss of ADRA2D in mammals) gives rise to distinct AR repertoires in different vertebrate groups. In the next section, we will present some of the most notable pharmacological differences previously identified between mammalian and fish ARs.

Earlier authors have noted the difficulties in using specific adrenergic agonists and antagonists, defined as such in mammalian work, in fish (Fabbri et al., 1998; Johansson, 1984). Numerous studies have indicated receptor pharmacological properties (e.g. order of agonist potency) that differ between fish and mammals. For example, binding studies of radiolabelled ligands (such as [3H]CGP-12177A) in rainbow trout skeletal muscle demonstrated inconsistent binding of classical antagonists and agonists (Lortie and Moon, 2003). The skeletal muscle ARs were refractory to the β1-AR-selective antagonist atenolol, but sensitive to a β2-AR-selective antagonist ICI-118,551 (Table 1). Interestingly, the receptors in red skeletal muscle were more sensitive to noradrenaline than to adrenaline, which is not typical of β2-ARs (Box 1). Moreover, white skeletal muscle showed a relatively low affinity for the selective β2-AR agonist procaterol (Lortie and Moon, 2003). Similar ligand binding studies on goldfish (Carassius auratus) head kidney preparations identified β-ARs with a low affinity for the β2-AR-specific agonist fenoterol and the β2-AR antagonist butoxamine, indicating that the β-ARs were unlikely to be of the β2 subtype (Jozefowski and Plytycz, 1998). However, the receptors also displayed little affinity for the β1-AR antagonist atenolol, leading the authors to conclude ‘ligand potency does not allow their being classified into any known mammalian subtype’ (Jozefowski and Plytycz, 1998).

In a study on black bullhead catfish (Ictalurus melas) hepatocytes (Fabbri et al., 1999b), it was demonstrated that yohimbine, in mammals defined as an α2-AR antagonist, binds to both α1 and α2 receptors and, remarkably, at high experimental concentrations enhances β-AR signalling (Fabbri et al., 1999b). Also in catfish hepatocytes, it was shown that the actions of isoproterenol, a classical β-AR agonist, were little affected by propranolol, a general β-blocker, suggesting isoproterenol is not a ‘pure’ β-AR agonist in catfish liver (Brighenti et al., 1987). Furthermore, there are reports that some AR ligands have opposing effects in mammals and fish. Clonidine, oxymetazoline and xylometazoline are defined as α-AR agonists in mammals (Table 1). However, in Atlantic cod (Gadus morhua) arterial preparations, these agents bind to α-ARs but do not elicit a contractive response, rather acting as competitive antagonists of adrenaline (Johansson, 1979).

Perhaps the most compelling and powerful experiments are those that have directly compared mammalian and fish ARs using heterologous expression in cultured cell lines (Ruuskanen et al., 2005b; Steele et al., 2011). In a systematic comparison of ligand binding to zebrafish and human α2-ARs expressed in a Chinese hamster ovary (CHO) cell line, Ruuskanen et al. (2005a,b) reported very similar responses between fish and mammal receptors of a given α2-AR subtype. However, some notable differences were reported. For instance, oxymetazoline exhibited no activation of the zebrafish α2B-AR or α2C-AR, which could help explain Johansson's (1979) surprising observation of its lack of agonist activity. It is important to note that Ruuskanen and colleagues’ (2005a,b) study focused on α2-ARs, so does not discount the differences that pharmacological agents can have across major AR types. For example, although yohimbine has similar effects on zebrafish and human α2-ARs (Ruuskanen et al., 2005b), this does not negate the surprising side effects it has as a fish α1-AR antagonist and β-AR signal enhancer, as described above (Fabbri et al., 1999b).

There are particularly disturbing reports of the non-specific effects of phenylephrine, as the drug remains especially widely used as an α-AR-specific agonist in fish studies (Abramochkin et al., 2022; Bonvissuto et al., 2022; Filogonio et al., 2017; Sandblom et al., 2009), despite it being well documented that phenylephrine also activates β-ARs in fish (Ask et al., 1980; Brighenti et al., 1987; Fabbri et al., 1992; Moon and Mommsen, 1990). This anomalous binding of phenylephrine to β-ARs was confirmed in human embryonic kidney 293 (HEK293) cells expressing transfected zebrafish or human β-ARs, wherein phenylephrine bound zebrafish β2-ARs with a high affinity (Steele et al., 2011). This study also showed that procaterol, considered to be a β2-AR agonist, stimulated intracellular cyclic adenosine monophosphate (cAMP) levels equally in cells expressing zebrafish β1-ARs, which could help explain previously contradictory results employing this drug in fish (Lortie and Moon, 2003).

The growing availability of teleost AR protein sequences and recent improvements in predictive physical modelling of GPCRs allow insights into the basis of differences in ligand affinity between fish and mammalian receptors. As a specific example, here we compare β2-receptor structure between teleosts and mammals (Fig. 3). Although the structures of the ligand binding domains are largely similar, we identified the formation of a salt-bridge between an aspartic acid (or physico-chemically similar glutamic acid) in extracellular loop 2 (ECL2) that is highly conserved across vertebrates and an arginine residue in transmembrane domain 7 (TM7), a mutation only identified in bony fish β2 receptors (including teleost β2a and β2b) and teleost-specific β3a receptors (Fig. 3B–D). The salt-bridge identified for zebrafish β2-ARs is formed by amino acids at equivalent positions, so that the small differences in sidechain locations (arginine in particular) between zebrafish β2a and β2b receptors (Fig. 3C,D) are likely to be indicative of a range of allowed movement. In contrast, K305 of the human β2 receptor (Fig. 3B) is at a different location to the arginine site of zebrafish β2-ARs, closer towards the pocket opening and more distant to the acidic sidechain (D192 in Fig. 3B). It is possible that D192 could form a charge interaction with H93 (which is closer than K305) in human β2-AR. If so, that would be on the side of the binding pocket, rather than blocking it. The salt-bridge of fish β2-ARs, found at a pocket of ligand entry, is likely to have limited effects on the receptor affinity for small agonists (including adrenaline and noradrenaline), but may impede the binding of larger antagonists, such as the β2-specific blocker ICI-118,551, which has inconsistent efficacy in teleosts (Abramochkin et al., 2022; Joyce et al., 2022; Lortie and Moon, 2003; Nickerson et al., 2003). This may explain why in adrb1 knockout zebrafish, the persistent positive chronotropic responses of adrenergic stimulation were surprisingly not attenuated by ICI-118,551 (Joyce et al., 2022), even though adrb2b receptor expression was increased compared with that in wild-type fish, suggesting a likely role for β2-ARs.

Together, comparative genomics, pharmacological studies and structural modelling reveal that not only do teleost fish have different AR repertoires to mammals (even when they share orthologues; see Glossary) but also their ligand-binding properties may be distinct. Accordingly, pharmacology alone cannot be used to reliably infer AR function and expression in fish. Nevertheless, pharmacological approaches can be effectively combined with molecular genetic techniques that are able to more precisely target teleost receptors, as discussed in the next section.

The discovery and characterization of a cAMP-dependent Na⁺/H⁺ exchanger (NHE) in teleost red blood cells (RBCs) was a seminal event in comparative physiology, which demonstrated a strategy for optimizing blood oxygen transport that is unique among vertebrates (Baroin et al., 1984; Cossins and Richardson, 1985; Mahé et al., 1985; Nikinmaa and Huestis, 1984). During periods of severely impaired blood oxygen transport, as might occur during acute hypoxia or exhaustive exercise, catecholamines are released into the circulation (Perry and Bernier, 1999; Reid et al., 1998), yielding adaptive adrenergic responses in a variety of organs and tissues including the RBCs (Motais et al., 1992; Nikinmaa, 1992; Randall and Perry, 1992; Thomas and Perry, 1992). Here, catecholamines bind to β-ARs, leading to cAMP accumulation and βNHE activation by protein kinase A (PKA) (Borgese et al., 1992). Unlike other vertebrate NHEs, the C-terminal region of the teleost βNHE contains two consensus PKA binding sites that are required for its activation by catecholamines (Borgese et al., 1992). Early studies on the mechanisms and functional consequences of βNHE activation paid little attention to the identity of the teleost RBC β-AR. Typically, βNHE activation was achieved using high doses of adrenaline (Cossins and Richardson, 1985; Nikinmaa, 1992) or isoproterenol (Baroin et al., 1984) with or without the non-specific antagonist propranolol. The first pharmacological assessment of the teleost RBC β-AR was conducted in 1988 using trout blood (Tetens et al., 1988). Based on a ligand potency order of isoproterenol>noradrenaline⪢adrenaline, it was concluded that the trout RBC β-AR is likely to be of the β1 subtype, which was referred to by the authors as a ‘noradrenaline receptor’ (Tetens et al., 1988). At the time of this study, the existence of a β3-AR subtype, which exhibits a similar ligand potency order to the β1-AR (Emorine et al., 1989), had not yet been confirmed, and the discovery that salmonids actually lack β1-ARs would wait another 35 years (Joyce, 2023). Based on a 60-fold greater affinity for noradrenaline than adrenaline, it was suggested that the elevation of circulating noradrenaline levels during acute stress in trout could account for 100% of the RBC adrenergic response in vivo. In contrast, a subsequent study using blood of Atlantic cod (Berenbrink and Bridges, 1994) demonstrated a potency order of adrenaline>noradrenaline, which is a characteristic of the mammalian β2-AR (Strosberg, 1993). Upon reviewing the literature on circulating catecholamine levels in Atlantic cod, it was concluded that activation of the RBC βNHE during periods of acute stress in cod was likely attributable to elevated adrenaline (rather than noradrenaline) levels. Thus, separate studies on two teleost species reached significantly different conclusions with respect to the identity of the RBC β-AR subtype and the relative contributions of adrenaline and noradrenaline to the activation of βNHE in vivo.

The ambiguity surrounding the identity of the trout β-AR was ultimately resolved with the cloning and characterization of two trout β3-ARs, β3a and β3b (Nickerson et al., 2003). The trout RBC, while expressing high levels of β3b, lacked β3a expression. Despite the successful cloning of these two trout β3-ARs, the authors were unable to clone the β1-AR, a result that was unexplained at the time. In hindsight, this finding, when coupled with the absence of the β1-AR in the trout genome (Joyce, 2023), reveals that the previously reported ligand potency order for activation of βNHE inadvertently led to the misidentification of the RBC β-AR as β1 rather than β3b (Tetens et al., 1988). Because the rainbow trout RBC contains a single β-AR subtype, blood can be used effectively as a model system to characterize fish β3b-ARs instead of the more cumbersome cell transfection route. Using the former approach, it was demonstrated (Nickerson et al., 2003) that the trout RBC β3b-AR exhibits ligand binding and functional properties (i.e. responses to agonists and antagonists) that are unique among vertebrate β3-ARs.

The fish liver stores significant quantities of glucose in the form of glycogen. Under conditions where additional circulating glucose is required (e.g. food deprivation, strenuous activity or hypoxia), glucose can be derived rapidly from hepatic glycogen via glycogenolysis and released into the bloodstream (Polakof et al., 2012). Glycogenolysis is regulated hormonally by catecholamines, with adrenaline and noradrenaline believed to play a dominant role (Fabbri and Moon, 2016; Fabbri et al., 1998). In brief, the binding of catecholamines to hepatic ARs initiates signal transduction pathways culminating in the activation of glycogen phosphorylase.

Identifying and establishing the roles of the AR subtypes in the fish liver in activating hepatic glycogenolysis has been described as a ‘challenging path’ (Fabbri and Moon, 2016), owing in part to methodological limitations and a reliance on pharmacological approaches designed and validated for mammalian research. The current consensus is that catecholamine-mediated hepatic glycogenolysis is linked to the activation of β2- and α1-ARs leading to classical cAMP and inositol trisphosphate (IP₃)/Ca²⁺ signalling, respectively (Fabbri and Moon, 2016), which mirrors the situation in mammals (see Fig. 1B,D). For many years, however, it was thought that β-adrenergic signalling was the predominant (if not exclusive) pathway for activating glycogenolysis in fish (Fabbri et al., 1998). This view was based on the inability of typical α-AR agonists and antagonists to stimulate glycogenolysis or inhibit its adrenergic activation, respectively (Birnbaum et al., 1976; Janssens and Lowrey, 1987; Ottolenghi et al., 1984), and the abolishment of glycogenolytic activity with propranolol (Janssens and Lowrey, 1987) (for a review, see Fabbri and Moon, 2016).

Based on the results of molecular cloning and receptor ligand binding assays, it is well established that fish hepatocytes express β2-ARs, which exhibit similar pharmacology to those of mammals (Dugan and Moon, 1998; Dugan et al., 2008; Fabbri et al., 1992, 1995; Nickerson et al., 2001; Reid et al., 1992). Despite being largely dismissed in terms of playing a significant role in hepatic glycogenolysis, there was indirect evidence for the involvement of α-ARs, beginning with the study of Birnbaum et al. (1976), which showed that ‘low’ doses (0.1 µmol l⁻¹) of adrenaline resulted in cAMP-independent activation of glycogenolysis in goldfish liver. However, in the absence of any effects of the α-AR agonist phenylephrine in stimulating glycogenolysis or the antagonist phentolamine in inhibiting adrenergic activation of glycogenolysis, a role for α-ARs was deemed unlikely (Birnbaum et al., 1976). A dissociation between activation of glycogenolysis and cAMP accumulation also was reported for rockfish hepatocytes exposed to catecholamines (Danulat and Mommsen, 1990), although this study also reported a reduction of adrenaline-mediated glycogenolysis by the α-AR antagonist prazosin. Further indirect evidence of cAMP-independent (Ca²⁺-dependent) pathways regulating hepatocyte glycogenolysis included the induction of glucose release from trout hepatocytes by the Ca²⁺ ionophore A23187 and the blockade of adrenaline-mediated glucose release by the Ca²⁺ channel blocker verapamil (Michelsen and Sheridan, 1990). Finally, Moon and Mommsen (1990) demonstrated increased glycogenolytic activity in several species in response to the mammalian α-AR agonist phenylephrine. Although the results appeared to indicate the presence of hepatic α-ARs, meaningful interpretation was confounded because phenylephrine caused an increase in cAMP levels and the effects of phenylephrine on glycogenolysis were either reduced or abolished by α- and β-AR antagonists, respectively. Further clouding the interpretation was the report of phenylephrine apparently binding to β-ARs in catfish liver (Fabbri et al., 1992). Thus, entering the 1990s, there was considerable indirect, albeit ambiguous evidence for hepatic α-ARs in fish.

Compelling evidence for the existence of α-ARs linked to Ca²⁺ signalling in eel and catfish liver was provided by Zhang et al. (1992a,b), who used fluorescence imaging to measure intracellular Ca²⁺ levels ([Ca²⁺]_i) in isolated hepatocytes. With this, it was shown that [Ca²⁺]_i was increased by adrenaline in eel and catfish hepatocytes (Zhang et al., 1992a). Later, an analogous study revealed the same mechanism in goldfish (Krumschnabel et al., 2001). The increase in [Ca²⁺]_i was blocked by the α-AR antagonist phentolamine and was unaffected by the β-AR antagonist propranolol (Zhang et al., 1992a). The potency order of the ligands tested was consistent with the presence of an α1-AR subtype, a conclusion that was supported, in part, by a more detailed pharmacological analysis of catfish liver (Moon et al., 1993). Interestingly, the effects of adrenaline on raising [Ca²⁺]_i in that study were blocked equally by the α1-AR antagonist phentolamine and the α2-AR antagonist yohimbine, leading to further confusion. It was not until years later that yohimbine was identified as a dual α1/α2-AR antagonist in catfish liver (Fabbri et al., 1999b). Moreover, the results of that same study revealed that yohimbine exhibited properties of a β-AR agonist and could increase cAMP levels, and it was promptly concluded that yohimbine is not suitable for use with fish (Fabbri et al., 1999a,b). With the finding of these atypical antagonist properties of yohimbine, the puzzling results of Moon et al. (1993) of equal antagonism of adrenaline-induced increases in [Ca²⁺]_i in catfish hepatocytes is readily explained by the binding of both antagonists to α1-ARs. Surprisingly, however, these increases in [Ca²⁺]_i were not associated with any increases in glycogen phosphorylase activity (Moon et al., 1993).

Clarity was finally achieved by using a perifusion hepatocyte incubation system exhibiting increased hormone responsiveness (Ottolenghi et al., 1994). At physiological levels (10 nmol l⁻¹) of adrenaline, glucose release from catfish hepatocytes was caused by interactions with both β- and α-ARs (Fabbri et al., 1999a), whereas at higher concentrations of adrenaline, similar to those used in previous studies, the adrenergic response was mediated exclusively by β-ARs. In isolated hepatocytes and hepatic membranes, it was also demonstrated that the α-AR-dependent signalling pathway (acting through the messenger IP₃; see Fig. 1B) is activated at lower adrenaline concentrations than the β-AR cAMP transduction cascade (Fabbri et al., 1995). This study also used radioligand binding experiments to confirm the presence of functional α-ARs (Fabbri et al., 1995). Thus, one can speculate that the extent of participation of α- and β-ARs in vivo will depend on the levels of circulating catecholamines, which, in turn, will reflect the severity of the stressor (Caselli et al., 2002; Randall and Perry, 1992). Hence, during mild stress, glycogenolysis may be activated by adrenergic activation of both α- and β-ARs, whereas during more severe stress, the β-ARs may be exclusively involved.

Establishing the AR profile of the teleost heart has been, and continues to be, a significant challenge fraught with complexities. In mammals, inotropic and chronotropic responses to adrenergic stimulation are largely attributable to stimulation of β1-ARs, although a smaller proportion of β2-ARs may also be co-expressed (Brodde et al., 1982; Costin et al., 1983; Heitz et al., 1983; Khamssi and Brodde, 1990; Myagmar et al., 2017). The predominance of β1-ARs is strongly supported by the finding that deletion of the adrb1 gene encoding β1-AR renders the mouse heart insensitive to β-AR stimulation (Rohrer et al., 1996). A small population of cardiomyocytes express β3-ARs (Myagmar et al., 2017), which likely play a minor, potentially inhibitory role, in the regulation of function (Cannavo and Koch, 2017; Dessy and Balligand, 2010; Imbrogno et al., 2015).

Initial studies in teleosts identified that ventricular contractility was more sensitive to adrenaline than to noradrenaline in rainbow trout (Gannon, 1971), cod (Holmgren, 1977) and plaice (Pleuronectes platessa) (Falck et al., 1966), providing the first hint that β2-ARs may be the predominant cardiac subtype. This notion was supported by later experiments showing positive inotropic and chronotropic responses in spontaneously beating sino-atrial preparations from trout to the β2-selective agonist salbutamol but not the β1-AR selective agonist prenalterol (Ask et al., 1980). In carp (Cyprinus carpio) ventricle, it was shown that the contractile responses to isoproterenol were insensitive to β1-AR antagonists atenolol and practolol (Temma et al., 1986a) and that contractility was more sensitive to adrenaline than to noradrenaline (Temma et al., 1986b). In a radioligand binding study, Gamperl et al. (1994) reaffirmed the greater sensitivity of trout cardiac ARs to adrenaline than to noradrenaline and reported insensitivity to atenolol, thereby concluding that ‘trout ventricular β-adrenoreceptors are exclusively of the β2 type’ (Gamperl et al., 1994). These convincing results contributed to the dogma that fish hearts are most commonly dependent on β2-AR stimulation (Olsson et al., 2000; Owen et al., 2007; Shiels, 2017; Stene-Larsen, 1981). The predominance of β2-AR-mediated adrenergic regulation is certainly true in salmonids where the adrb1 gene encoding the β1-AR was evolutionarily lost (Joyce, 2023). However, for other species, the picture remains more complicated.

As we have sought to emphasize in this Review, purely pharmacological approaches have proven unreliable. The use of atenolol in in vitro studies (Gamperl et al., 1994; Mendonça and Gamperl, 2009; Temma et al., 1986a) yields particularly ambiguous data. Atenolol has amongst the lowest binding affinities of β-AR antagonists (Baker, 2005) and, in mammals, its use in vitro may elicit only transient binding (Nakasone et al., 1988), yielding results that may be inconsistent with in vivo effects (Vandeplassche et al., 1991). Other radioligand binding studies, including those in bird (Lindgren and Altimiras, 2013) and reptile (Abraham et al., 2019) hearts, have employed the β1-AR antagonist CGP-20712A, which has a greater binding affinity than atenolol. It would be worthwhile to study its effects on adrenergic responses of myocardium from different teleost species. It may also be problematic to draw conclusions based on sensitivity to adrenaline and noradrenaline; in HEK293 cells transfected with human β1-ARs, radioligand binding, as expected, appeared to be more sensitive to noradrenaline than to adrenaline (i.e. exhibiting a lower dissociation constant; Steele et al., 2011). However, HEK293 cells transfected with zebrafish β1-ARs exhibited radioligand binding that instead appeared to be more sensitive to adrenaline than to noradrenaline (Steele et al., 2011), which could account for the fish heart being more sensitive to adrenaline (Falck et al., 1966; Gannon, 1971; Holmgren, 1977) even if they express β1-ARs.

Recent studies in zebrafish have provided direct evidence for β1-AR responses in the fish heart. Atenolol, when used in vivo, reduced heart rate and contractility in adult zebrafish studied with echocardiography (Hein et al., 2015). In larval zebrafish, atenolol and morpholino knockdown of β1-ARs reduced positive chronotropic responses to hypercapnia (Miller et al., 2014). Also in larval zebrafish, treatment with the β1-specific agonist dobutamine incurs an even greater increase in contractility than does treatment with adrenaline (Shin et al., 2010). Recently, we developed a β1-AR knockout zebrafish line (Joyce et al., 2022), which displays chronically reduced heart rate during larval development, while retaining a positive chronotropic response to adrenergic agonists in larvae and isolated hearts from adults. Together, these results indicate an important, but not exclusive, functional role for the β1-AR in zebrafish. Although we acknowledge that mRNA levels do not necessarily correlate with protein expression (Buccitelli and Selbach, 2020; Nickerson et al., 2002), adrb1 also appears to be highly expressed in the hearts of other bony fishes, including spotted gar (Lepisosteus oculatus) (Zavala et al., 2017), fathead minnow (Pimephales promelas) (Giltrow et al., 2011), pike (Esox lucius) and cod (Joyce, 2023).

Additionally, there is a growing appreciation of the possible roles of β3-ARs in fish hearts (Imbrogno et al., 2015; Petersen et al., 2013; Shiels, 2017). As β3-ARs can sometimes be coupled to inhibitory G-proteins, this may include a negative inotropic component, as shown in eel hearts (Imbrogno et al., 2006). However, this line of research is still in its infancy (Imbrogno et al., 2015) and it remains to be clarified how functionally relevant β3-ARs are in other species.

In mammals, α1-AR stimulation may fine-tune the cardiac response to adrenergic stimulation by lengthening action potential duration (APD) of cardiomyocytes, prolonging the time for Ca²⁺ influx (Endoh et al., 1991; Jensen et al., 2011; Joyce et al., 2021). Earlier studies in fish found little evidence to support a relevant contribution of α-ARs in the teleost heart (Ask et al., 1980; Farrell et al., 1986). However, recently, it was suggested that α-ARs could mediate a prolonged APD in trout cardiomyocytes via inhibition of the delayed rectifier K⁺ current (I_Kr) (Abramochkin et al., 2022). This suggestion was founded on pharmacological evidence that the suppressed I_Kr induced by adrenaline treatment was not inhibited by combined treatment with β1-AR and β2-AR antagonists (atenolol and ICI-118,551, respectively) and could be mimicked with phenylephrine (α1-AR agonist) but not isoproterenol (β-AR agonist) (Abramochkin et al., 2022). However, as we have argued, conclusions on fish AR function based on pharmacological data alone should be treated with caution, particularly with regard to the use of phenylephrine (which is capable of non-specific actions on β-ARs), atenolol (which displays transient binding in vitro) and ICI-118,551 (which is potentially blocked by a salt bridge in teleost β2-ARs; Fig. 3).

Taken together, it is emerging that adrenergic regulation in the fish heart is more complex than has long been believed and likely involves species-specific expression of various AR subtypes. Regrettably, different techniques have often been employed in different species, and this lack of consistency makes it difficult to synthesize across studies. It is retrospectively frustrating that so much work has been dedicated to the rainbow trout heart, which because of the loss of adrb1 in salmonids, is poorly representative of other teleosts (Joyce, 2023). Nevertheless, there are encouraging signs that by combining pharmacological and molecular genetic approaches, we are on the road to disentangling the contributions of different ARs in the hearts of different fish species.

An inherent assumption when employing adrenergic drugs with type and subtype specificity in fish is that the receptors are well conserved with those of mammals. Although this is often the case, research on fish ARs historically has been mired with pharmacological anomalies. In this Review, we have developed the argument that an understanding of the evolutionary history of fish ARs is pivotal in the application of pharmacological agents in comparative contexts. Evolutionary studies have indicated that the three main types of AR perhaps attained sensitivity to adrenaline and noradrenaline independently (Spielman et al., 2015). Given that any conserved differences between the subtypes must have been established before the separation of teleosts and lobe-finned fishes >350 million years ago, it is remarkable that many pharmacological characteristics have persisted (Ruuskanen et al., 2005b), especially considering that subtle structural changes can substantially alter ligand binding affinity. Some of the divergence between fish and mammalian ARs that previously clouded a clear understanding of AR expression in fish is beginning to be elucidated by complementing pharmacology with molecular genetics. The ‘omics’ era presents an opportunity for an even deeper understanding of the evolution and diversity of adrenergic regulation in fish, and some of the most pressing outstanding questions are summarized below.

The duplication of ARs as a result of the teleost-specific 3R WGD provides a valuable opportunity to gain insight into the potential fates and physiological importance of duplicated genes. Recent phylogenetic work has clarified that only some lineages of fish have retained the duplicated copies of β3-AR (adrb3a and adrb3b genes), while, by contrast, most teleosts retain both β2-AR duplicates (Zavala et al., 2017). It remains to be established why adrb3b is disposable in only some lineages, and it would be worthwhile investigating whether its retention correlates with any particular physiological advantages or whether the lineages maintaining it show any common traits. Presuming adrb genes become sub-functionalized in the lineages with duplicated copies, it would also be interesting to assess whether adrb3a is able to re-expand its ‘physiological niche’ in the specific lineages that subsequently lost adrb3b. Such an investigation could be reinforced experimentally by evaluating the effects of adrb3 paralogue-specific gene knockouts. Together, these experiments could help clarify some of the elusive roles of the β3-ARs (Imbrogno et al., 2015).

One of the most important recent advances is the appreciation of receptor multiplicity in the fish heart, particularly in non-salmonids, because it opens the possibility of different receptors being preferentially expressed under different environmental conditions. In adrb1 knockout zebrafish, there is greater expression of adrb2b and adrb3b in larval hearts, so certainly there is the capacity for fish to significantly alter cardiac adrb expression (Joyce et al., 2022). It is possible that some receptors are preferentially expressed following acclimation to different temperatures, which is known to affect adrenergic sensitivity of the fish heart (Keen et al., 1993, 2017). Such studies may additionally help explain the emerging interspecific differences. The study of single-cell gene expression has been useful on this front in mammals by establishing that only a subpopulation of cells express certain ARs (Myagmar et al., 2017); similar investigations could be tremendously informative in fishes.