Teleost fishes offer an alternative model in which there is a clear neuroanatomical link between estrogen production in the brain and neurogenesis. Indeed, in teleost fishes, radial glial cells strongly express aromatase and the questions are to understand when and why this feature emerged during evolution. Whether this particular situation is found only in fishes or is also observed in basal tetrapods is unclear. Recently released information on frogs will also be examined in this review for evolutive and comparative purposes.

2. Estrogen receptors, aromatase and radial glia in the brain of fish

Teleost fish (Teleostei) represent one of three infraclasses of Actinopterygians, corresponding to the so-called ray-finned fishes. This is a highly persified group that arose 280–250 millions years ago and comprises around 26,000 species in about 40 orders and that include most of the living fishes, including the economically-relevant species that are the topic of intense research. A major characteristic of these animals is that, compared to tetrapods, they experienced an additional whole genome duplication that took place early in the emergence of the group. As a result of this duplication, known as the 3R [34], teleost fish have potentially twice as much genes than other vertebrates (2:4:8 rule). Although many of these duplicated genes have been lost during evolution, some of them are conserved and developed new or exaggerated functions.

2.1. Estrogen receptors in teleost fishes

The first estrogen receptor cloned in fish was that of the rainbow trout [35,36], soon after the cloning of the human ERα (esr1) [37]. Compared with the mammalian ERα, this rainbow trout (rt ERα) of 65 kDa, named ERα short, exhibited at the N-terminus a deletion of 45 amino acid residues corresponding to the A domain. However, subsequently a longer ERα form of 71 kDa was retrieved from an ovarian library [38,39]. By S1 nuclease protection assays, it was shown that these two isoforms derived from two classes of mRNA generated by an alternative usage of two promoters. Consequently, these mRNA species differ in their 5′-untranslated region and the presence of an ATG in exon 2a permits adding 45 residues at the N-terminus of rtERα-long. Analysis of the transcriptional activities of these isoforms in a yeast cell system demonstrated that, in contrast with rtERα-long, rtERα-short exhibits a ligand-independent transactivation capacity representing 15–25% of the full-length receptor activity. Structural analysis of the AF1 function showed that as it is the case of the mammalian ERα, in the absence of ligand, the A domain of the rtERα-long interacts with the C-terminal region in the absence of ligand, causing inhibition of the AF1 activity located in the B domain [38,39]. Studies in rainbow trout showed that the full-length ERα is expressed in liver, brain, pituitary, and ovary, whereas expression of the ERα short is restricted to the liver, demonstrat-ing a tissue-specific expression of these two ERα isoforms [40].

Following these pioneer studies and the discovery of two estrogen receptors in mammals [41], it rapidly appeared that teleosts have not only one ERα, but also two ERβ resulting from an ancient duplication [42,43]. Zebrafish, probably the best-documented fish, has three ER, two ERβ, ERβ1 (esr2b), and ERβ2 (esr2a) and one ERα (esr1). These three estrogen receptors bind estradiol with a high affinity within the 0.5–0.7 nM range, similar to what has been observed in mammals [43]. They also have transactivation capacity on reporter genes bearing estrogen responsive element, with ERβ2 exhibiting a little more efficien-cy than the other two receptors, starting at 10−12–10−11 M. All three ERs are expressed in a wide variety of tissues including the brain [43].

One of the characteristics of ERα, at least in the liver, is its spectacu-lar up regulation by its own ligand. This phenomenon has been first deciphered at the molecular level in the rainbow trout and is the basis for the strong induction of vitellogenin expression in the liver upon es-trogenic modulation [35]. Such an up regulation of ERα is also observed in the brain, but the induction is much lower than what is observed in the liver [44].