We present here strong evidence confirming the presence of two AQP isoforms in vertebrates that are structurally and functionally related to mammalian AQP1. These isoforms, Aqp1a and Aqp1b, seem to coexist exclusively in teleost fish since Aqp1b was not found in mammals, amphibians or birds. The main difference between Aqp1a and Aqp1b is in the C-terminal tail, which contains specific residues for regulation of intracellular trafficking in Aqp1b.
Previous sequence analyses of MIPs suggest that substrate selective modes (AQPs and aquaglyceroporins) were acquired early in the history of the family by gene duplication and functional shift, with the highest level of diversification occurring in vertebrates and higher plants . Analysis of Aqp1a and Aqp1b distribution by searching currently available genome sequence information and by cDNA cloning suggest that both isoforms are present exclusively in the teleost genome. Phylogenetic reconstruction of vertebrate AQP1-like proteins indicates that Aqp1a and Aqp1b share a common origin and are likely to have evolved from duplication of a common ancestor. Because both isoforms are present in fish species belonging to distant taxonomic groups, from basal (e.g., Anguilliformes) to more modern (e.g., Gasterosteiformes, Perciformes, Pleuronectiformes and Tetraodontiformes) groups [26, 27], this duplication must be ancient and is likely to have affected most teleosts. As suggested for many duplicated genes in teleosts, the origin of Aqp1b might be the whole-genome duplication (WGD) event that occurred specifically in the ray-finned (Actinopterygian) lineage after splitting from the tetrapod lineage about 350 million years ago . However, as it has not been possible to identify Aqp1b in most basal actinopterygians (e.g., paddlefish and sturgeon) with the genomic information available, it is not known whether the AQP1 duplication also affected these groups. In all teleosts examined, the aqp1a and aqp1b loci were found to be closely linked, indicating that Aqp1b possibly arose by a gene duplication event at a local level rather than at the chromosome or genome level. Local gene duplication has also been proposed, for instance, to explain the repertoire of teleost opsins [29, 30] and the generation of the Xiphophorus Xmrk oncogene .
The radiation of teleosts in the ocean most likely required the evolution of new osmoregulatory mechanisms in eggs and early embryos to alleviate the passive water loss imposed by the hyper-osmotic environment . In this scenario, it is plausible to hypothesize that duplication of aqp1 genes in teleosts allowed for one duplicate to encode a product with a new function through innovating mutations in regulatory and/or structural sequences ('neofunctionalization'). This seems to be the case for sea bream Aqp1b which plays a specialized physiological role in the oocyte mediating water uptake during meiotic maturation [10, 11]. In other marine and catadromous teleosts, such as sole and eel, respectively, that like sea bream also spawn highly hydrated eggs, we show here that aqp1b also encodes a functional water channel whose RNA is predominantly accumulated in the ovary, suggesting a similar role of Aqp1b during oocyte hydration. In other pelagophil teleosts [32, 33], as well as in some species of catfish, in which oocyte hydration may also occur despite having a freshwater life cycle (e.g., ), Aqp1b-encoding ESTs have also been found in the ovary. In contrast, in zebrafish, a freshwater species where almost no oocyte hydration is observed , we found a completely different aqp1b expression pattern, together with a higher mutation rate in the amino acid sequence of the encoded protein. Based on these data, we argue that, in marine teleosts producing highly hydrated eggs, Aqp1b possibly represents a neofunctionalized water channel adapted to oocytes to facilitate water transport. Finn and Kristoffersen  recently proposed that neofunctionalization of duplicated Vtg genes, which allowed one paralog to be proteolyzed into FAAs driving hydration of the maturing oocytes, was a key event in the evolution and success of the teleosts in the oceanic environment. The duplication and neofunctionalization of aqp1b may have occurred in parallel to this mechanism to facilitate oocyte water uptake in marine teleosts.
The Aqp1b isoform found in freshwater teleosts that spawn non-hydrated eggs, such as zebrafish, might be inactivated by mutations, or be eventually lost in the genome. This hypothesis is supported by the absence of aqp1b in advanced freshwater species, such as medaka, while the synteny between the aqp1a chromosome loci and downstream genes (e.g., thoc1) is conserved. However, in other extant freshwater teleosts that arose later in evolution, such as Tetraodon, aqp1b is retained in the genome. The retention of aqp1b in freshwater pufferfishes is intriguing, given that there appears to have been a massive elimination of DNA after WGD in most modern teleost genomes, resulting in the retention of only a subset of the duplicates [36, 37]. It is possible to speculate, however, that the recent evolution of Tetraodontiformes did not last long enough to allow specific divergence of the genome and hence of aqp1b. The relatively high amino acid sequence identity (77%) between fugu, a marine pufferfish which produces hydrated eggs, and Tetraodon Aqp1b supports this conjecture. In any event, additional studies aiming at the characterization of Aqp1b in more freshwater fish species, as well as the determination of sites of gene expression and protein accumulation, are required to better understand the driving force behind aqp1b isoform evolution.
In addition to the ovary, aqp1b mRNA has been detected in the posterior intestine, kidney, gills and esophagus of marine fish (this work, and [10, 14, 38]). The intestine of marine teleosts has an important osmoregulatory role, as hypo-osmoregulating fish have long been known to drink seawater to replace water lost by diffusion to their environment (see  for review). Accordingly, Aqp1a and Aqp1b proteins have been reported to be localized in the intestinal epithelia of teleosts [12–15]. However, in the sea bream gastrointestinal tract Aqp1a and Aqp1b have a different distribution pattern. Whereas Aqp1a is localized in the apical and basolateral membrane of enterocytes in duodenum and hindgut, Aqp1b is exclusively localized in the apical microvilli of rectal epithelial cells . Moreover, freshwater acclimation reduces the synthesis of Aqp1a in all intestinal segments, and of Aqp1b in rectum. Conversely, seawater acclimation of eels increases Aqp1a expression and protein synthesis in the intestine [13, 19]. Therefore, although the specific physiological functions of Aqp1a and Aqp1b in the teleost gastrointestinal tract remain unknown, these data may point to an additional role of Aqp1b in water movement across the intestinal epithelia.
The primary structure of AQP1-like proteins corresponding to the TM2 and TM5 domains and connecting loops B and E, which are involved in the formation of the water-selective pore, is highly conserved between teleost and mammals. However, Aqp1a and Aqp1b show different permeability efficiencies when expressed in X. laevis oocytes, and teleost Aqp1b isoforms also show a marked structural divergence at the C-terminal cytoplasmic tail with respect to Aqp1a and mammalian AQP1. Functional experiments, using artificial expression in oocytes of sea bream wild-type Aqp1a and Aqp1b, and chimeric proteins in which the C-terminus of Aqp1a was totally exchanged for that of Aqp1b, or the reverse, revealed that the Aqp1a tail drives constitutive targeting to the plasma membrane, unlike that of Aqp1b which produces partial retention of the expressed proteins in intracelluar vesicles. These data strongly suggest that Aqp1b independently acquired specific regulatory domains in the C-terminal region for the control of Aqp1b intracellular trafficking.
To investigate the nature of putative regulatory sites in the Aqp1b C-terminus, we analyzed its amino acid sequence in different teleosts. Based on this analysis, selected residues of sea bream Aqp1b were mutated into Ala or Asp and the resulting proteins were expressed in oocytes to determine their intracellular localization and permeability properties. In the Aqp1b C-terminus, we found typical sorting and internalisation signals that are common in many mammalian transmembrane proteins for targeting from the trans-Golgi network to the lysosomal-endosomal compartment . These motifs, however, were also detected in the Aqp1a C-terminal tail, although their sequence appeared to be different between the Aqp1a and Aqp1b isoforms. Thus, in all six teleost Aqp1b sequences analyzed, but not in Aqp1a, a di-Leu or Leu-Ile signals appear to be conserved. Mutation of sea bream Aqp1b Leu234Leu235 motif into Ala234Ala235 produced the retention of the protein in intracellular compartments and apparently increased its degradation, thereby reducing water permeability of these oocytes. These results are similar to those observed with the mammalian AQP2 mutant which has an altered and extended C-terminal tail, retained in late endosomes/lysosomes triggering degradation . Similarly, mutation of the C-terminal Leu345Leu346 motif in the human vasopressin V3 receptor produces mis-folding of the protein and abolishes receptor export . It is possible therefore, that the di-Leu motif in sea bream Aqp1b also plays a role in conformation, ensuring correct routing to the plasma membrane. However, di-Leu motifs are also involved in basolateral membrane targeting and microvilli anchoring of mammalian cell adhesion proteins, ion channels and receptors [42–45], including AQP4 , in polarized epithelial cells. Further studies are needed to establish whether the di-Leu motif has an additional function in the control of Aqp1b expression on the cell surface.
Most notably, functional analyses revealed that two residues in the sea bream Aqp1b C-terminal sequence, Thr229 and Ser254, were responsible for sea bream Aqp1b translocation from intracellular vesicles to the oocyte plasma membrane. The probability of the Thr229 residue being phosphorylated was low, and accordingly the Aqp1b-T229A mutant did not affect the phosphorylation state of Aqp1b, although it did inhibit Aqp1b cell surface expression and oocyte water permeability. Since Thr229 did not match any kinase phosphorylation consensus site other than protein kinase C (which apparently is not relevant here), the specific function of this residue is unknown and awaits further experimentation. Nevertheless, it was observed that the Aqp1b-S254A mutant prevented phosphorylation and increased Aqp1b translocation into the plasma membrane and subsequent water permeability, whereas the Aqp1b-S254D mutant, which mimicked the constitutively phosphorylated state of Aqp1b, was predominantly located in intracellular vesicles. These results suggest that dephosphorylation of Ser254 triggers Aqp1b shuttling to the cell surface, while its phosphorylation may retain the protein in intracellular vesicles. This mechanism is thus apparently the opposite to that described so far for mammalian and amphibian AQPs (i.e., AQP2 and AQP-h2), where channel insertion in the plasma membrane of collecting duct cells or granular cells of the anuran urinary bladder is triggered by protein kinase A-mediated phosphorylation of specific Ser residues in the C-terminal tail [21, 47]. Interestingly, the Ser254 in sea bream Aqp1b, a consensus site for a Pro-directed kinase, seems to be conserved in modern marine teleosts which produce hydrated eggs (Ser244 in fugu Aqp1b and Ser244 in sole Aqp1b). Pro-directed kinases are a large family of mitogen-activated protein kinases (MAPK) and cyclin-dependent kinase-like kinases, some of which (e.g., p38 MAPK) are involved in transduction pathways leading to the activation of the maturation promoting factor (MPF) during oocyte meiotic maturation [48–50]. In sea bream, Aqp1b translocation into the oocyte plasma membrane is a tightly regulated process thought to occur transiently downstream of MPF activation during meiotic maturation, just before complete hydrolysis of yolk proteins and maximum K+ accumulation is reached in the oocyte . Therefore, it will be of interest to investigate the potential role of cell-cycle related kinases, or other kinases activated during oocyte maturation, in Ser254 phosphorylation and regulation of Aqp1b trafficking.