The hypothalamic-pituitary-thyroid (HPT) axis is the neuroendocrine thyrotropic axis (Figure 2). After indirect evidence of the existence of a thyrotropin-releasing factor in dogs (Shibusawa et al., 1955), a fraction, purified from bovine or ovine hypothalamic extracts, was found to stimulate the release of TSH (Schreiber et al., 1961; Guillemin et al., 1962, 1963). The TRH gene encodes multiple identical repeats of the three amino-acid (pyroglutamic acid-histidine-proline) TRH sequence (Galas et al., 2009). This is an example of intragenic duplication which allows an amplification of the synthesis and production of this neurohormone.

TRH-R are members of membrane class A G-protein coupled receptors (GPCR). Two subtypes, TRH-R1 and TRH-R2, have been characterized in mammals. A third one, TRH-R3, has been identified in birds (Li et al., 2020), amphibians (Bidaud et al., 2002), reptiles (Li et al., 2020), and teleosts (Mekuchi et al., 2011; Saito et al., 2011; Li et al., 2020). In birds, only TRH-R1 and TRH-R3 have been identified so far (Li et al., 2020). Four subtypes have been cloned in sockeye salmon Oncorhynchus nerka (TRH-R1, TRH-R2a, TRH-R2b, and TRH-R3; Saito et al., 2011), and in medaka Oryzias latipes (TRH-R1a, TRH-R1b, TRH-R2, and TRH-R3; Mekuchi et al., 2011). As in medaka (Mekuchi et al., 2011), duplicated TRH-R1a and b are found in many teleosts including zebrafish Danio rerio, and might be the result of 3R, even if only one TRH-R1 is identified in sockeye salmon (Saito et al., 2011). The duplicated TRH-R2s present in sockeye salmon may be due to the 4R; TRH-R2b is truncated (Saito et al., 2011). According to the recent hypothesis of Li et al. (2020) on the evolutionary history of TRH-Rs, multiple gene loss events might have occurred during evolution, including TRH-R2 loss in birds, and many mammals, as well as TRH-R3 loss in mammals. Interestingly, pituitary TRH-R3 expression was recently shown to elevate as metamorphosis progressed in bullfrog tadpole (Nakano et al., 2018). It would be interesting to investigate whether the possible involvement of TRH-R3 in amphibian metamorphosis is also observed in other major life history transitions, such as teleost metamorphoses and sauropsid hatching.

TSH is a glycoprotein hormone composed of two subunits, the alpha subunit (Gpα) common to gonadotropins, and the hormone-specific TSHβ subunit, which is paralogous to the luteinising hormone (LH) and follicle-stimulating hormone (FSH) β subunits. The genes encoding the β subunits of the two gonadotropins LH, FSH and of TSH are paralogous genes that likely arose from the two successive 1R/2R in ancestral vertebrates. A single TSHβ is present in mammals and other tetrapods, as well as in actinopterygians, and is considered as the “classical” TSH. Its β subunit is now named TSHβ1 since a second TSHβ, called TSHβ2, has been identified in some representative species of early vertebrates such as in chondrichthyans and basal sarcopterygians (Maugars et al., 2014). These data suggest that the gene coding for tshβ2 was lost twice independently during osteichthyan evolution, in a tetrapod ancestor and in an actinopterygian ancestor. Indeed tshβ2 could not be retrieved in the genome of an holostean (spotted gar Lepisosteus oculatus) nor any teleosts, supporting an early loss of tshβ2 in the actinopterygian lineage, leading to inheritance of only tshβ1 by the teleost lineage (Maugars et al., 2014). Two tshβ paralogs were found in teleosts, resulting from the 3R duplication of tshβ1 (Maugars et al., 2014; Fleming et al., 2019), and named tshβ1a and tshβ1b (Fleming et al., 2019). Recently, up to three tshβ1 paralogs were identified in Oncorhynchus species, due to the conservation of the duplicated 4R-paralogs of tshβ1a (tshβ1aα and tshβ1aβ) while only the two tshβ1 paralogs issued from 3R (tshβ1a and tshβ1b) are present in Atlantic salmon Salmo salar as in other teleosts (Fleming et al., 2019). Interestingly, the two tshβ paralogs identified in Atlantic salmon are expressed in different pituitary cells (Fleming et al., 2019): tshβ1a in the anterior adenohypophysis, and tshβ1b in cells near to the pituitary stalk, cells comparable to the pars tuberalis TSH cells involved in seasonal physiology and behavior in birds (Yoshimura, 2013) and mammals (Dardente et al., 2019).

TSH-R is present at the surface of the thyroid follicle cells in the thyroid gland. It belongs to the class A GPCR receptors and is paralogous to the receptors for gonadotropins. In teleosts, due to the 3R, duplicated receptors tshra and tshrb have been characterized (Maugars et al., 2014). TSH-R are not only expressed in the thyroid gland, but also in a variety of other tissues (Williams, 2011). Of particular interest is the presence of tshr in the brain and the gonads, where they mediate TSH retrograde action on brain deiodinase and direct regulatory effect on reproduction, in birds and mammals (Nakane and Yoshimura, 2019).

Two main forms of TH are found in vertebrates, T₄ with four iodinated tyrosine residues (3,3′,5,5′), the hormonal precursor produced by the thyroid gland, and T₃ with three iodinated tyrosine residues (3,3′,5), the active hormone which binds to TR with a 10-fold higher affinity compared to T₄ (Holzer et al., 2017). T₄ is produced, by thyrocytes under TSH control, from thyroglobulin, a large dimeric protein containing many tyrosine residues.

TR belong to the nuclear receptor superfamily. Two homologous receptors, TRα and TRβ, are present in tetrapods (Sap et al., 1986; Weinberger et al., 1986; Thompson et al., 1987; Brooks et al., 1989; Helbing et al., 2006; Kanaho et al., 2006). The 3R led to two duplicated trα paralogs and two duplicated trβ paralogs in basal teleosts, such as conger and eel. So far, four tr cDNAs have been cloned in Japanese flounder: two α types and two β types (Yamano et al., 1994; Yamano and Inui, 1995). In some other teleosts, such as zebrafish, two trα genes but a single trβ gene have been retrieved, suggesting that one of the trβ duplicated paralog would have been lost in the course of teleost evolution (Lazcano and Orozco, 2018). Xenopus laevis possesses two distinctive genes for each tr likely due to its genome polyploidisation (Yaoita et al., 1990). In tetrapods, alternative splicing of TR leads to various protein isoforms with different binding activities and biological functions. In mammals, two TRα variants (α1 and α2) are thus generated from a single gene (Izumo and Mahdavi, 1988). Similarly, in addition to the 3R-duplicated genes, alternative splicing may lead to a large variety of TR isoforms in teleosts, with species-specific variations (Marchand et al., 2001).

Deiodination of TH corresponds to the removal of iodine from their outer or inner rings, and is performed by iodothyronine deiodinase enzymes in the thyroid, peripheral or central target tissues. By regulating the bioavailability of T₄ and T₃, they thus control their action (Bianco et al., 2002; Orozco and Valverde-R, 2005; Darras and Van Herck, 2012; Orozco et al., 2012; Steegborn and Schweizer, 2020). The first deiodinase to be cloned was dio1 from rat (Berry et al., 1991), then dio3 was cloned from Xenopus laevis (St. Germain et al., 1994), and dio2 from Rana catesbeiana (Davey et al., 1995). So far, these three types of iodothyronine deiodinases, DIO1, DIO2, and DIO3, have been identified in most vertebrates (Bianco et al., 2002; Darras and Van Herck, 2012; Orozco et al., 2012). DIO1 is a low selectivity enzyme with both outer and inner ring deiodinase activity. DIO2 is an outer ring deiodinase, which removes an iodine residue from T₄ and gives active T₃. DIO3 is an inner ring deiodinase, which inactivates T₄ and T₃ in reverse T₃ (rT₃ or 3,3′,5′-T₃), and diiodothyronine (T₂ or 3,3′-T₂), respectively. Orozco et al. (2012) reviewed the evolutionary history of deiodinase family and point out that dio1 is the oldest vertebrate deiodinase gene and dio2 the most recent one. Some teleosts possess 3R-duplicated dio2 and dio3 genes (Orozco et al., 2012; Alves et al., 2017). The widespread and differential tissue distribution for the three types of deiodinase in vertebrates has been reviewed previously (Darras and Van Herck, 2012; Orozco et al., 2012). In the Atlantic salmon, up to six deiodinase paralogs have been characterized, one dio1, two dio2 (dio2a and dio2b) and three dio3 (dio3a1, dio3a2, and dio3b) (Lorgen et al., 2015). A single dio2 gene is present in Northern pike Esox lucius (Rondeau et al., 2014), which suggests that the two dio2 paralogs of the Atlantic salmon arose recently from the salmonid-specific WGD, 4R (Lorgen et al., 2015). Dio3a1/3a2 likely also resulted from 4R (Lorgen et al., 2015). Functional divergence of the two salmon dio2 paralogs is reported: while gill dio2a expression is induced by seawater exposure (Lorgen et al., 2015), dio2b expression in the brain is sensitive to photoperiod during smoltification (Lorgen et al., 2015; Irachi et al., 2021).

The hypothalamic-pituitary-adrenal axis (HPA), in mammals and sauropsids, and hypothalamic-pituitary-interrenal (HPI), in amphibians and teleosts, is the neuroendocrine corticotropic axis (Figure 3). In 1955, a substance, present in extracts of mammalian hypothalamus and able to stimulate ACTH secretion in vitro, was named corticotropin-releasing factor (CRF) (Guillemin and Rosenberg, 1955; Saffran et al., 1955). CRF (or CRH) was first isolated from sheep hypothalamus (Vale et al., 1981) and identified in all vertebrates thereafter (Lovejoy et al., 2014; On et al., 2019). Together with urotensin I (UI) in teleosts, sauvagine (SVG) in amphibians, and urocortins (Ucn) in mammals, it forms a large family of peptides, the CRH/urocortin family. As CRH, urotensin I, sauvagine and urocortins have roles in the regulation of ACTH and MSH, as well as energy metabolism and reproduction (Lovejoy and Balment, 1999). It was first suggested that two ancestral crh/ucn1 and ucn2/ucn3 genes likely arose by specific gene duplication before vertebrate WGD events (Hwang et al., 2013), as a single crh-like gene is identified in amphioxus and tunicates (On et al., 2019). Both ancestral genes were duplicated twice in ancestral vertebrates via 1R and 2R, followed by some paralog losses, leading to up to 5 genes (crh1, crh2, ucn1 coming from ancestral crh/ucn1; ucn2, ucn3 issued from ancestral ucn2/ucn3) in extant representative species of some vertebrate lineages such as chondrichthyans, holosteans and actinistians (Cardoso et al., 2016; On et al., 2019). Teleost specific 3R resulted in the duplication of crh1 into two paralogs crh1a and crh1b conserved in many species (Grone and Maruska, 2015; Cardoso et al., 2016). Crh2 may have been lost in recent teleosts (Cardoso et al., 2016), while one 3R-crh2 paralog has been conserved in basal groups of teleosts such as European eel (Maugars et al., 2016).

In tetrapods, CRH has been shown to act via two GPCRs, corticotropin-releasing hormone receptors (CRHR), CRHR1, and CRHR2, which belong to the class 2 subfamily B1 of secretin-like GPCR superfamily (Lovejoy et al., 2014). Crhr1 was duplicated by 3R into two paralogs (crhr1a and crhr1b) which were conserved in many extant teleosts, while one of 3R-duplicated crhr2 paralogs would have been lost (Cardoso et al., 2016). A third CRHR, CRHR3, identified in the catfish, Ameiurus nebulosus (Arai et al., 2001), highly similar to catfish CRHR1, likely results from gene-specific duplication, that may also occurred in various teleosts (Lovejoy et al., 2014; Cardoso et al., 2016).

The main pituitary hormone involved in the HPA/HPI axis is ACTH. ACTH is derived from tissue-specific post-translational processing of its precursor, the proopiomelanocortin (POMC); POMC also gives rise to melanocyte stimulating hormone (MSH), and β-endorphin (β-END). POMC forms the opioid/orphanin gene family together with proenkephalin, prodynorphin and proorphanin (Dores et al., 2002; Sundström et al., 2010). ACTH and MSH, called melanocortins (MC), act via MC receptors (MCR), while β-END acts via opioid receptor. Teleost 3R gave rise to pomc gene duplicates: pomc-α (a or A) and pomc-β (b or B), with pomc-β having lost a functional β-endorphin (De Souza et al., 2005). Further independent gene-specific duplications during teleost evolution resulted in duplicates of pomc-α, such as in sea bream Sparus aurata (Cardoso et al., 2011) and Astatotilapia burtoni (Harris et al., 2014).

Melanocortin receptors (MCR) belong to the rhodopsin class A13 family of GPCRs. In tetrapods, five MCR genes have been identified: mc1r to mc5r (Cortés et al., 2014; Dores et al., 2014). In teleost fish, probably due to 3R, the number of receptors increases up to six in zebrafish, which has two MC5R paralogs (Västermark and Schiöth, 2011), while pufferfish Fugu rubripes has only four, with no mc3r and only one copy of mc5r (Logan et al., 2003). Concerning the ligand selectivity of MCRs, all of the paralogous MCRs can be activated by both ACTH and MSH in extant cartilaginous fishes, while in extant teleosts and tetrapods, only ACTH can activate MC2R (Cortés et al., 2014; Dores et al., 2014). In mammals, the MCRs have distinct tissue expression (Cone, 2006; Dores et al., 2014): mc1r is mainly expressed in melanocytes; mc2r in adrenal cortex; mc3r and mc4r in the brain; and mc5r in a variety of exocrine glands, such as sebaceous, lacrimal and preputial glands. In concordance with their expression sites, each MCR has its proper function: MC1R is involved in skin and hair pigmentation; MC2R in adrenal steroidogenesis and stress response; MC3R and MC4R in energy homeostasis; and MC5R in exocrine gland secretion. These features can also be observed in non-mammalian vertebrates, with some differences. Of particular interest, mc5r is co-expressed with mc2r in the interrenal of Xenopus tropicalis (Dores and Garcia, 2015) and several teleosts [rainbow trout Oncorhynchus mykiss (Haitina et al., 2004; Aluru and Vijayan, 2008); common carp Cyprinus carpio (Metz et al., 2005); barfin flounder Verasper moseri (Kobayashi et al., 2011)], as well as in the chicken adrenal (Takeuchi and Takahashi, 1998), suggesting a possible role of MC5R, in addition to MC2R, in the regulation of HPI axis in these non-mammalian vertebrates.

Corticosteroids (CS) are steroid hormones, divided into glucocorticoids (GC), and mineralocorticoids (MC). In mammals and sauropsids, CS are synthetized by the adrenal gland. In amphibians and teleosts, CS are synthetized by the interrenal gland, a tissue embedded inside the anterior part of the kidney (head kidney) and homologous to the adrenal cortex of the adrenal gland in mammals (Chester-Jones, 1987). Cortisol is the primary GC in most mammals and teleosts, while it is corticosterone in birds, reptiles, amphibians and many rodents (Mommsen et al., 1999; Aerts, 2018). Both GC can co-exist in most vertebrates. Aldosterone, which is the principal MC in mammals (Gilmour, 2005), is lacking in all teleosts studied so far. In teleosts, it is generally accepted that cortisol plays both GC and MC roles (McCormick, 2001; McCormick et al., 2008).

GC and MC receptors (respectively, GR and MR) belong to the nuclear receptor superfamily. 2R resulted in the emergence of GR and MR, present in all extant vertebrates. In teleosts, 3R gave rise to duplicated gr and mr (Bury, 2017). Zebrafish is the only known teleost to have conserved only one of the two gr paralogs (Schaaf et al., 2008). The two mr paralogs are present in a basal teleost, the European eel Anguilla anguilla (Lafont et al., 2014), while only one mr paralog has been conserved in the other extant teleosts studied so far (Baker and Katsu, 2019).

After hatching and a period of growth, the tadpole undergoes a rapid and irreversible physiological, anatomical and environmental transition marking the transition from the larval state to the adult state, the metamorphosis. Gudernatch observed that the administration of extracts of the thyroid gland (and only this organ) via the water induced tadpole premature metamorphosis (Gudernatch, 1912). Following TH measurements (Leloup and Buscaglia, 1977), the elucidation of the roles and mechanisms of action of TH in vertebrates has thus benefited enormously from amphibian model (Grimaldi et al., 2013; Buchholz, 2017; Sachs and Buchholz, 2017). The activity of TH is regulated locally by deiodinases (DIO) and cytosolic proteins (Shi et al., 1994). In particular, the expression of DIO2 (enzyme activating TH) is high in tissues undergoing metamorphosis, while the expression of inactivating deiodinase (DIO3) is high in tissues before and after metamorphosis (Leloup et al., 1981; Galton, 1989).

Another level of TH signaling control is linked to the expression profiles of thyroid hormone receptors (TR) (Yaoita and Brown, 1990). A dual role of thyroid signaling was proposed (Sachs et al., 2000; Grimaldi et al., 2013; Buchholz and Shi, 2018). Thus, before metamorphosis, the relative absence of TH and the expression of TRα contribute to repressing target genes allowing tadpole growth. During the climax of metamorphosis, the high concentration of TH concomitant with the expression of TRα and TRβ contributes to the activation of the transcriptional program leading to tadpole transformation. The two TR isoforms have also different roles during metamorphosis. TRα was shown to be more involved in cell proliferation while TRβ in cell differentiation and apoptosis (Furlow and Neff, 2006; Denver et al., 2009). TRα is required for TH-dependent neurogenesis (Wen et al., 2019). The autoinduction of TRβ is one of the earliest responses to the thyroid signal (Yaoita and Brown, 1990) because trβ is a direct TH-target gene (Ranjan et al., 1994; Bilesimo et al., 2011).

Transgenic studies in Xenopus laevis revealed that TR are necessary and sufficient for mediating the effects of T₃ during metamorphosis (Buchholz et al., 2003, 2004). However, recent gene knockout studies in Xenopus tropicalis showed that TR are not required for most metamorphic transformations, although tadpoles lacking TR die in the middle of metamorphosis (Shi, 2021). Removal of TR enables premature metamorphosis of several adult tissues (Buchholz and Shi, 2018), likely due to the absence of T₃-inducible gene repression. This result supports the role of unliganded TR to avoid precocious metamorphosis for proper tadpole growth. Finally, removal of TR prevents the disappearance of tadpole-specific tissues (Shibata et al., 2020, 2021).

Glucocorticoids (GC) are also essential during anuran metamorphosis. In Xenopus laevis tadpoles, corticosterone concentration is high during pre-metamorphosis, decreases during pro-metamorphosis and increases again during metamorphic climax (Jaudet and Hatey, 1984; Jolivet-Jaudet and Leloup-Hâtey, 1986; Kloas et al., 1997; Glennemeier and Denver, 2002). However, the actions of GC on anuran development are difficult to dissect because of their involvement in many biological processes. Thus, GC inhibit growth, both in pre- and pro-metamorphic tadpoles and the administration of exogenous GC to pre-metamorphic tadpoles inhibits the emergence of lower limbs (Kobayashi, 1958). However, the use of drugs which antagonize the molecular action of GC or prevent the production of GC by the adrenals, inhibits metamorphosis (Kikuyama et al., 1982).

During metamorphosis, GC act through glucocorticoid receptors (GR) that have dynamic and tissue-specific expression profiles (Krain and Denver, 2004). In the tail, GR expression follows the concentration profile of GC, but in the gut and brain the number of transcripts were constant during pro-metamorphosis, and slightly decreased or increased, respectively, during climax. GR knockout in Xenopus leads to complete abrogation of the corticosterone-responsive gene induction by exogenous hormone (Sterner et al., 2020). In addition, tadpoles lacking GR developed faster than wild-type sibling until forelimb emergence. Then, they developed more slowly and died at the climax of metamorphosis, indicating that GR is required for metamorphosis and control the developmental rate. The essential requirement of the HPA axis was confirmed with the gene-editing disruption of the pomc gene (Sterner et al., 2020). Mutant tadpoles had a reduced level of plasma corticosterone at metamorphosis, as well as lower expression of the corticosterone-responsive genes. Last but not least, these tadpoles had reduced rates of growth and development that finally led to death, late during metamorphosis at the time of tail resorption.

Although TH are the trigger for metamorphosis and are required throughout this process (Sachs and Buchholz, 2017), their actions strongly intersect with those of CS (Sachs and Buchholz, 2019). Stressors such as predation or pond desiccation can be experienced by tadpoles and cause activation of the HPI axis (Glennemeier and Denver, 2002). In the tadpole, the synthesis and release of TSH by the anterior pituitary is under the control of CRH (Denver, 1993, 2021), as in other non-mammalian vertebrates (De Groef et al., 2006). When injected to prometamorphic Bufo arenarum larvae, CRH is able to accelerate metamorphosis, likely by direct action on both TSH and ACTH pituitary cells (Miranda et al., 2000). This dual role of CRH allows the tadpole to modulate the rate of metamorphosis in response to environmental stimuli (Denver, 1997). CS are closely linked to metamorphosis by delaying or accelerating its rate of progression. Elevation in circulating CS during the pre-metamorphic stage slows tadpole growth (Denver, 2017). However, during pro-metamorphosis, tadpoles increase the production of TH and CS in response to environmental stress, and are therefore able to accelerate metamorphosis, which can allow the animal to escape a deteriorating larval habitat by transitioning to the next life history stage (Denver, 1997, 2009; Denver et al., 2009; Kulkarni and Buchholz, 2014). Remarkably, CS are capable of triggering metamorphosis in Buffo boreas (Hayes et al., 1993). This action is probably due to the endogenous levels of TH with which GC are able to act synergistically, but insufficient, on their own, to induce the metamorphosis.

How does the crosstalk between TH and CS occur? CS are known to promote the expression of activating deiodinases and repress the expression of inactivating deiodinases, thus contributing to the accumulation of active T₃ in target tissues. This model has long been the basis for explaining the potentiating effect of GC on metamorphosis (Galton, 1990). Furthermore, GC can also act in synergy with TH via cross-regulation at the level of their respective nuclear receptors. GC are able to increase the binding capacity of T₃ in cell nucleus (Suzuki and Kikuyama, 1983) and the amount of trβ transcripts in the tail, brain and intestine (Bonett et al., 2010). Conversely, the expression of GR seems to be partly controlled by TH depending on the tissues observed: while in the brain and intestine, treatment with T₃ induces a decrease in the expression of GR, the opposite occurs in the tail (Krain and Denver, 2004). These results suggest that CS are able to increase the sensitivity to TH in some tissues, thus accelerating metamorphosis. These actions are synergistic with low concentrations of TH (Bonett et al., 2010). Brown et al. (2014) reported similar hormonal synergy during larval development in fish.

Other genes expressed during metamorphosis are synergistically regulated by TH and CS (Kulkarni and Buchholz, 2012; Bagamasbad et al., 2015), some through direct transcriptional regulation by liganded TR and GR. One of the best studied examples of this type of regulation is the transcription factor krüppel-like factor 9 (Klf9), which is directly and synergistically regulated by TH and CS via an ultra-conserved super enhancer located upstream of the transcription start site (Bonett et al., 2009; Bagamasbad et al., 2015). Klf9 strongly enhances TRβ autoinduction in tadpole tissues, among other roles modulating gene transcription and hormone action during metamorphosis (Hu et al., 2016). Other genes show synergistic regulation by TH and CS in tadpole tissues, but whether their protein products function to modulate the rate of metamorphosis requires further study (Kulkarni and Buchholz, 2012, 2014).

Finally, the recent results with tadpoles lacking pomc reinforce the inseparable link between TH and GC at metamorphosis (Sterner et al., 2020). Tadpoles lacking pomc had reduced expression levels of TH-responsive genes such as klf9 and trβ. Even more significant, mutant death at metamorphosis is rescued by exogenous TH, suggesting a strong entanglement of the signaling by both hormones.

Beside the striking larval metamorphosis in eels and flatfishes, studies also report developmental changes controlled by TH during larval to juvenile transition in various teleosts such as grouper Epinephelus coioides (de Jesus et al., 1998), seabream (Campinho et al., 2010), gobiid Sicyopterus lagocephalus (Taillebois et al., 2011), or clownfish Amphiprion percula (Salis et al., 2021). Our review will focus on the well-known pleuronectiform/flatfish metamorphosis. Larval metamorphosis with the migration of one eye to the opposite side of the head is unparalleled in vertebrate development (Inui and Miwa, 2012). Spectacular morphological changes are accompanied by behavioral (pelagic to benthic life and locomotion) and physiological (such as development of gastrointestinal tract to adapt to the novel food resources in a new habitat) changes.

Smoltification is the transition of sedentary juvenile parr into downstream migratory smolt, which will pursue its growth phase in the ocean. Smoltification gathers many changes, morphological ones, such as body silvering and fin darkening, behavioral ones, with swimming activity in open space, formation of schools and downstream migration, and physiological ones related to adaptation to seawater and imprinting (Hoar, 1976, 1988; Boeuf, 1993; McCormick, 2012; Rousseau et al., 2012).

During most of the gestation in human, T₄ is converted in rT₃, while toward term, developmental changes in tissue deiodinase activity occur, leading to preferential deiodination of T4 to T3 (instead of rT₃) and rise in plasma T₃ concentration near term (Forhead and Fowden, 2014). In contrast, in the rat, circulating T₃ concentrations do not increase before birth (Dubois and Dussault, 1977; Harris et al., 1978; Lamers et al., 1986). Comparison of hormonal profiles during the perinatal period in two closely related murine species with distinct modes of development (altricial vs. precocial), the rat (altricial), and the spiny mouse Acomys cahirinus (precocial) allowed to show a correlation between perinatal increase in T₃ levels and precocial timing of birth (Lamers et al., 1986).

TH have a major regulatory role on fetal growth (Forhead and Fowden, 2014), and maturation of the central nervous system (Patel et al., 2011; Stenzel and Huttner, 2013). Furthermore, maternal hypothyroidism is associated with increased rates of respiratory distress syndrome in new-born (Casey et al., 2005; Männistö et al., 2013) and cardiorespiratory disorder (Rousseau et al., 2019). In various mammalian species, TH change the synthesis of the components of surfactant (Ballard et al., 1984; Das et al., 1984; Warburton et al., 1988; Gilbert et al., 2001; Van Tuyl et al., 2004). TH are also involved in the intestinal structural development (Trahair and Sangild, 1997; Sirakov and Plateroti, 2011; Sirakov et al., 2014). Specific TRα1 loss-of-function leads to low development and impaired activity of murine intestinal stem cells in culture (Godart et al., 2021). In addition, in vivo treatment confirms the positive action of T₃ on intestinal crypt cell proliferation and demonstrates its key action in modulating the number of stem cells, the expression of their specific markers and the commitment of progenitors into lineage-specific differentiation (Godart et al., 2021).

An increase of plasma fetal GC is detected near parturition in many mammals (Alexander et al., 1968; Mulay et al., 1973; Lamers et al., 1986; Silver and Fowden, 1989; Yoon et al., 1998). The involvement of HPA axis in the timing of birth is first evidenced in ovines (Van Rensburg, 1967; Barnes et al., 1977). Furthermore, premature delivery is induced by injection of ACTH or corticosteroids into the ovine fetus (Van Rensburg, 1967; Halliday and Buttle, 1968; Liggins, 1968, 1969) as well as in fetal piglet (Bosc, 1973). More recently, the use of antalarmin, a CRH-R type I antagonist, can delay the onset of parturition in sheep, which suggests that CRH is involved in the induction of parturition in this species (Chan et al., 1998).

The many roles of GC on organ maturation have been the subjects of recent reviews (Fowden and Forhead, 2015; Fowden et al., 2016; Jellyman et al., 2020). As for TH, we will concentrate on GC involvement in lung and gastro-intestinal tract (GIT) maturations, which are crucial to cope with the change of environment (from aquatic to terrestrial) and of nutrition (from parenteral to enteral), respectively. In lambs born prematurely by infusion of dexamethasone, partial aeration of the lungs is noted, suggesting an accelerated appearance of surfactant activity (Liggins, 1969). Infusion of ACTH into one of twin pairs of lamb fetuses accelerates the morphological changes, that would normally occur in the lung, before birth (Sundell et al., 1979). GC treatment changes the synthesis of the phospholipid and protein components of surfactant (Warburton et al., 1988; Gilbert et al., 2001). At birth, there is a shift from mainly parenteral nutrition in the fetus (via the placenta) to enteral nutrition in the neonate, and the GIT has to prepare during gestation for coping with this change. This prenatal development of GIT is regulated by GC (Trahair and Sangild, 1997; Sangild et al., 2000). After bilateral adrenalectomy of fetal sheep, growth of mucosal structures and villus height are reduced in the small intestine (Trahair et al., 1987a). After infusion of cortisol, no change in enterocyte morphology is detected, but the proportion of crypt cells and the migration of enterocytes are increased (Trahair et al., 1987b). In addition, cortisol influences the prenatal development of gastric acid and gastrin secretion, and of GIT hydrolase activities in both the fetal pig and sheep (Trahair and Sangild, 1997; Sangild et al., 2000).

Before birth, TH and GC synergize for the maturation of various organs, especially the lung, the liver and the brain. This synergism is partly linked to the fact that GC induce local deiodinase expression and activities resulting in the increase of circulating T₃ and thus T₃ bioavailability (Forhead et al., 2006; Fowden and Forhead, 2009). A synergism of TRH/T₃ and cortisol has been reported on lung maturation in fetal sheep, as infusion of cortisol and TRH (Liggins et al., 1988) or T₃ (Schellenberg et al., 1988; Warburton et al., 1988) increase the distensibility and the stability of the lung. In explant culture of human fetal lung, a supra-additive response in the synthesis of surfactant is observed in the presence of both dexamethasone and T₃ (Gonzalez et al., 1986).

T₄ injection to rats induces a precocious appearance of pepsinogen in the oxyntic gland mucosa and increases acid secretion, while in propylthiouracil-induced hypothyroid pups, pepsinogen content and basal acid secretion are low (Tseng and Johnson, 1986). However, T₄ has no such effects in adrenalectomized rats, while corticosterone is able to increase pepsinogen content and basal acid secretion in the absence of normal levels of TH (Tseng and Johnson, 1986).

Teleost fish as ectothermic vertebrates show complex responses to temperature variation (Pinsky et al., 2019). Temperature has in teleost fishes a direct effect on oxygen delivery, cardiovascular function, muscle function, food conversion, mitochondria efficiency, and biochemical reaction rates (Little et al., 2020). Fish physiology is thus likely to be affected by increases in average temperature and temperature variability leading to energy metabolism changes with impact on growth and locomotion. All these physiological processes have been associated with thyroid function.

T₃ regulates thermal acclimation in zebrafish (Little et al., 2013) with the decrease of metabolism, skeletal muscle function and swimming performance in hypothyroid cold-acclimated (18°C) but not hypothyroid warm-acclimated (28°C) animals. The interactions between temperature levels, thyroid conditions and swimming performance are also linked to performance of the heart and oxygen transport (Little and Seebacher, 2014). In the context of global warming, the low sensitivity to TH at warm temperature is a concern because it may reduce the capacity of zebrafish to match this environmental change. Furthermore, in medaka, exposure until hatching to warm temperatures (32°C) increases the activation of TH biosynthesis, via TSH (Castañeda-Cortés et al., 2020), as well as the activation of the stress axis, via CRH (Castañeda-Cortés et al., 2019).

Rainbow trout is a typical cold-water fish species for which seasonally warmer water correlates with a decrease in fry survival and an increase of T₃ and T₄ concentrations (Giroux and Schlenk, 2021). Transcriptome analysis of liver tissues from adult rainbow trout under heat stress (24°C) and control conditions (18°C) identifies the differential expression of 428 long non coding RNA some of which can be involved in maintenance of homeostasis or adaptation to stress (Quan et al., 2020). In Atlantic salmon, elevated temperature increases T₄ nocturnal levels (Nisenbaum et al., 2020). Thus, future climate changes can induce lower salinity tolerance and accordingly result in poor survival in seawater after smoltification.

To show the wide diversity of the links between global warming and TH, we will give three more examples. First, in the mosquitofish Gambusia holbrooki the thermal acclimation is mediated by TH and leads to decreased mitochondrial efficiency, metabolic rates, and swimming performance (Le Roy and Seebacher, 2020). Second, in the desert fish Amargosa pupfish Cyprinodon nevadensis amargosae, temperature contributes to altered TH levels and morphological development (Lema et al., 2016). We can thus expect that many desert fishes will be vulnerable to rising temperatures because they occupy habitats with already high temperatures. Finally, increased temperature induces a decrease of TH levels in convict surgeonfish Acanthurus triostegus (Besson et al., 2020). The effect on TH levels correlates with affected sensory development leading to higher predation.

Chronic temperature increases (persisting for a long time) are also known to induce acute stress responses in fishes (Alfonso et al., 2020). Global warming can cause water freshening from increased freshwater inputs. In Antarctic spiny plunderfish Harpagifer antarcticus, salinities decrease leads to high plasma cortisol levels compare to normal salinity (Vargas-Chacoff et al., 2021). This has implications for fish species that have evolved in stable environmental conditions like in the Antarctic. In a totally different environment, climate-induced bleaching of coral reefs alters anemonefish hormonal stress, resulting in decreased reproductive hormones and severely impacted reproduction (Beldade et al., 2017). Finally, in juvenile brown trout, Salmo trutta, water temperature affects cortisol and GR mRNA levels (Filipsson et al., 2020).

Amphibians are another ectodermic taxon with complex interactions between costs and benefits of life in the aquatic and terrestrial environments. They are undergoing a precipitous decline linked with environmental pressures including climate change. Many amphibian species exploit temporary or even ephemeral aquatic habitats for reproduction. Survival from desiccation of pond drying is reached by maximizing larval growth and acceleration of development, including precocious metamorphosis.

The western spadefoot toad tadpoles, Pelobates cultripes, usually have long larval period, but it can shorten it in response to pond drying (Gomez-Mestre et al., 2013). Developmental acceleration correlates with increased corticosterone and TH levels as well as increased TRβ levels to increase metamorphic process. Anuran tadpole shows thus phenotypic plasticity in age and size at metamorphosis as a response to temperature variation (Ruthsatz et al., 2020). In Arizona tiger salamander Ambystoma tigrinum, the small increase in temperature tends also to decrease the time to metamorphosis and results in a worse body condition (Park et al., 2016). Plasticity in rates of growth and development is beneficial to allow a more rapid transition into the juvenile stage where rates of mortality are lower. However, early metamorphosis leads to juveniles with smaller size correlated with low survival and locomotor performance (Székely et al., 2020). Pond drying may also lead to the loss of the capacity to respond to desiccation following decreased food availability (Enriquez-Urzelai et al., 2013), liver damage with upregulation of DIO2 and TRα transcript levels while there is a decrease of TRβ (Chen et al., 2021), or increased oxidative damage during metamorphosis (Petrović et al., 2021).

TH and GC level variations underlie differences in the timing of metamorphosis. In response to pond drying, western spadefoot toad and New Mexico spadefoot toad Spea multiplicata that have long larval periods and large size at metamorphosis, accelerate metamorphosis and elevate whole-body content of TH and corticosterone (Kulkarni et al., 2017). In contrast, in Couch’s spadefoot toad Scaphiopus couchii that has a short larval period, whole-body TH and corticosterone content are high during metamorphosis and weakly affected by pond drying. Thus, species exhibiting less plasticity will be more sensitive to environmental changes. The TH and CS level variations to accelerate metamorphosis during habitat desiccation may originate from elevated CRH (Denver, 1997). CRH can be a transducer of environmental stimuli to modulate the rate of metamorphosis via the endocrine response. Interestingly, CRH has also been shown to control the term of pregnancy in mammals. Maternal plasma CRH concentrations are elevated early in pregnancy in those patients destined to deliver preterm, and are lower in patients destined to deliver post-dates (Challis and Hooper, 1989; McLean et al., 1995).

Temperature increase alters avian phenotypes such as advanced reproduction, migration schedules and individual appearance. Bird ability to maintain a stable body temperature in a wide range of thermal environments is regulated via endocrine pathways including TH and with a lesser extent CS (Ruuskanen et al., 2021). The knowledge on endocrine regulation of thermogenesis concerns mainly poultry, but poorly describe for environmental temperature variation. Offspring development are dependent of females via hormones deposited in eggs. Ambient temperature changes affect the hormone levels in the yolk including TH and CS. More specifically, T₄ levels were negatively correlated with ambient temperature in great tits Parus major (Ruuskanen et al., 2016). However, in wild pied flycatchers Ficedula hypoleuca, there is no evidence for context-dependent effects of prenatal TH related to postnatal temperature on growth, survival, and plasma TH levels (Hsu et al., 2020), suggesting species differences or unknown confounding effects.

In reptiles and turtles, temperature effect on TH and GC plasma levels is controversial being either absent, positive or negative. Highlighting only a few examples, in the alligator lizards Elgaria coerulea and Elgaria multicarinata, corticosterone levels increase with temperature (Telemeco and Addis, 2014) and in the South-American tegu lizard Salvator merianae, T₃, T₄, and corticosterone levels show a positive relationship with body temperature (Zena et al., 2020). Similar correlation between body temperature and plasma corticosterone is observed in the eastern fence lizards Sceloporus undulatus, under laboratory conditions, but not in the field (Racic et al., 2020). In contrast, in the common lizard Zootoca vivipara, baseline corticosterone levels decrease with increasing thermal conditions (Dupoué et al., 2018).

In mammals, we will only highlight two of the numerous examples. First, the transition between late gestation and early lactation is a developmental window sensitive to warming. Dairy cows experience stress due to the high energy and nutrient requirements of the fetus and the mammary gland. In early lactation, the decline of TH levels is more pronounced in cow maintained at 28°C compared to the one maintained at 15°C (Weitzel et al., 2017). The second example is an Artic mammal, which has highly evolved to these extreme environments and its capacity to adapt to this change may be limited. The polar bear is emblematic to this environment. It was shown that CS binding capacity of plasma CS binding globulin increases in polar bears as a consequence of climate warming (Boonstra et al., 2020). Evidence is still lacking to have a complete view of the interactions between GC and TH in response to global warming.

As previously mentioned, TH orchestrates metamorphosis, brain development, and metabolism representing a potential target for endocrine disruptor chemicals/compounds (EDC). Although the molecular mechanisms for TH disruption is still unknown for most of the EDC, several affect TH synthesis, transport or metabolism and downstream effects (Thambirajah et al., 2019). One of the main concerns is disruption during early neurogenesis, which may affect several TH actions such as proliferation, differentiation, migration, synaptogenesis and myelination in the developing nervous system (Préau et al., 2015).

Because anuran metamorphosis is strictly dependent on TH, amphibians represent sensitive models for the detection and mechanistic elucidation of TH disrupting activities (thyroid histology, metamorphosis progression). EDC impact wild anurans and contribute to population declines. Some pesticides and biocides may interfere with TH signaling during nervous system development (Leemans et al., 2019; Trudeau et al., 2020) or metamorphosis timing (Orton and Tyler, 2015). Last but not least, exposition of Xenopus tadpoles to the benzo[a]pyrene leads to delayed metamorphosis and sexual maturity with transgenerational disruption of metabolism and population decline (Usal et al., 2021).

In zebrafish, exposure during the 7 first days of development to 25 known TH disrupting compounds leads to morphological defects and variations of transcripts involved in the HPT axis (Spaan et al., 2019). In another teleost fish, the metamorphosing convict surgeonfish Acanthurus triostegus, high doses of the pesticide chlorpyrifos induce defects by decreasing TH levels impacting olfactory, visual and mechano-sensory structure development (Besson et al., 2020). Interestingly, similar phenotypes were observed following temperature increase, highlighting the profound threat anthropogenic stressors pose to fish communities. To highlight the EDC mode of action, liver transcriptomic responses in yellow perch Perca flavescens population show variation of genes transcripts related to reproduction, retinol, iron, TH, oxidative stress, lipid metabolism, and immune functions between strongly impacted population vs. less impacted groups (Defo et al., 2018).

In birds as in teleost fishes and amphibians, because GC regulate metabolism, ability to modulate corticosterone in response to stressor is essential to face a wide array of environmental challenges. EDC contamination was shown to impair this GC role. Methylmercury exposure reduces stress-induced GC response in zebra finches, Taeniogypia guttata (Moore et al., 2014). TH function is also EDC target. The rural nestling peregrine falcon Falco peregrinus has significantly lower circulating concentrations of TH compared to urban nestlings where flame retardants are environmental contaminants that accumulate in predatory birds (Fernie et al., 2017).

EDC also affect TH signaling in mammals. Considering the large amount of recent data available, only a few examples will be given here. In mice, exposure to the flame-retardant, tetrabromo bisphenol A, modulates hypothalamic set-points controlling metabolic responses by targeting TR regulation of trh and mc4r (Decherf et al., 2010). In the polar bear Ursus maritimus, an heavily polluted organism for which some EDC are banned for decades (Routti et al., 2019), some EDC, such as organochlorine and perfluoroalkyl, are negatively correlated with plasma TH levels (Bourgeon et al., 2017). In wild chimpanzees, contamination by polluted drinking waters containing pesticides (Krief et al., 2017) and bisphenols (Krief et al., 2020) has been reported, and the water samples exhibit TH disrupting activities (Spirhanzlova et al., 2019). Experimental exposure of Xenopus laevis tadpoles to a mixture of 15 chemicals at concentrations reported in human amniotic fluid induces variation of behavior, of TH-responsive gene expression and of nervous system development (Fini et al., 2017). Considering the conservation of TH function across vertebrates, one may speculate on such impacts of amniotic fluid chemicals on human fetal brain development.

Environmental and health concerns remain particularly challenging when addressing thyroid hormone axis disruption. Only few chemicals have been tested, the number of test methods is limited and the thyroid system is complex. The development of specific chemical safety testing is required (Browne et al., 2020) with special attention to neurological development (Gilbert et al., 2020; Kortenkamp et al., 2020; O’Shaughnessy and Gilbert, 2020). Attention is also paid to break down the wall between mammalian and non-mammalian vertebrate regulatory testing (Couderq et al., 2020; Holbech et al., 2020). In addition to phenotypic or histological end-points, the use of molecular assays (transcripts, proteins or metabolites) will be more sensitive and most of all will allow detection before adverse effects occur (Fini et al., 2007; Kulkarni and Buchholz, 2013). Development of non-destructive (without euthanasia) biomarkers that provide causal link between the presence of a chemical and an ecological effect are also needed to allow the study of vulnerable populations.

The adrenal has been neglected in endocrine disruption regulatory testing strategy while the negative consequences of adrenocortical dysfunction during development have been recognized (Hinson and Raven, 2006; Harvey, 2016). Upstream of the biological effect, an activity mimicking that of GC was measured in surface waters (Schriks et al., 2013). Hydrocortisone was detected but could not explain a significant fraction of the observed GR activity.

GC synthesis pathway is also a target. Different environmental toxicants, including phthalate esters together with bisphenols and their analogs pose deleterious effects on the biosynthesis of GC in vitro (Ahmad et al., 2017; Verma et al., 2018). These observations were confirmed in vivo using nitrate, a major anthropogenic contaminant in the FW environment. In three-spined stickleback Gasterosteus aculeatus L., dissolved inorganic N directly exerts a disruptive influence on the function of the stress axis and GC levels, supporting concerns that nitrate is an EDC (Pottinger, 2017). GC disruption was linked to adverse effect in zebrafish exposed to methylparaben, widely used as antimicrobial preservative. In addition to increase in cortisol levels, animals show alteration of heart rate and hatching percentage as well as signs of anxiety-like behavior (Luzeena Raja et al., 2019).

Finally, because persistent organic pollutants can interact with GR, EDC mixtures were screened for GR translocation and GR transactivation in an in vitro assay (Wilson et al., 2016). EDC mixtures did not induce GR translocation in nucleus nor produce an agonist response in the GR transcriptional assay. However, in the presence of cortisol, some chemicals were found to decrease GR transcriptional activity.