Isolation of a novel aquaglyceroporin from a marine teleost (Sparus auratus): function and tissue distribution

The aquaporin family [formerly called the major intrinsic protein (MIP)family] consists of transmembrane channel proteins (AQP), which function as channels for non-ionic compounds (Heymann and Engel, 1999). In vertebrates two principal groups have been identified, the CHIP (CHannel forming Integral Protein) and GLP (GLycerol intrinsic Protein) proteins (Froger et al., 1998; Heymann and Engel,1999; Park and Saier,1996; Reizer et al.,1993) and phylogenetic studies suggest that the groups arose from two divergent bacterial paralogues. Members of the CHIP cluster function exclusively as water channels (aquaporins), while members of the GLP group have a much broader function and, although considered to be primarily glycerol transport proteins (aquaglyceroporins; Park and Saier, 1996), they are also permeable to water and other small solutes (for reviews, see Ishibashi et al., 1998a,b; Verkman et al., 1996; Wintour, 1997).

Several CHIP homologues have been identified in mammals: AQP0, AQP1 and AQP2 (Lanahan et al., 1992; Preston and Agre, 1991; Yang et al., 1999), AQP4, AQP5 and AQP6 (Krane et al., 1999; Raina et al., 1995; Sobue et al., 1999; Turtzo et al., 1997; Yasui et al., 1999) and AQP8(Ishibashi et al., 1997a; Koyama et al., 1998). Most of them have been shown to stimulate osmotic water permeability of plasma membrane when expressed in Xenopus oocytes. Four aquaporin homologues belonging to the GLP group have been identified in mammals: AQP3(Echevarria et al., 1994; Ishibashi et al., 1994; Ma et al., 1994), AQP7(Ishibashi et al., 1997b),AQP9 (Kuriyama et al., 1997; Tsukaguchi et al., 1998) and AQP10 (Hatakeyama et al.,2001; Ishibashi et al.,2002). Members of the GLP group have a widespread distribution in tissues associated with water transport. For example, AQP3 is abundantly expressed in the principal cells of the mammalian kidney(Ecelbarger et al., 1995; Echevarria et al., 1994), in the large airways, urinary bladder, conjunctiva, epidermis(King et al., 1997; Nielsen et al., 1997) and gastrointestinal tract (Koyama et al.,1999; Ramirez-Lorca et al.,1999). AQP7 is abundantly expressed in the rat seminiferous tubules and plasma membrane of late spermatids and is also present in rat heart, skeletal muscle and kidney(Ishibashi et al., 1997a). AQP9 is expressed at high levels in liver and testis and at lower levels in brain (Tsukaguchi et al.,1998) and AQP10 is expressed in the small intestine(Ishibashi et al., 2002).

Relatively few reports about AQP proteins in non-mammalian vertebrates exist. In amphibians AQP1–AQP4(Abrami et al., 1994; Ma et al., 1996; Schreiber et al., 2000) have been identified. In fish, a CHIP protein (AQP0) has been identified in Fundulus heteroclitus (Virkki et al., 2001) and recently an AQP3 homologue was reported in eel Anguilla anguilla (Cutler and Cramb, 2002). Fish are a useful group in which to study aquaporin evolution as they inhabit diverse aquatic environments and face a constant osmotic and ionic challenge, which varies according to the waters they inhabit(freshwater <0.1 mOsmol kg^–1; seawater approximately 1000 mOsmol kg^–1). The importance of water and ion transport by renal and extrarenal epithelial tissues for maintenance of osmotic homeostasis in fish is well documented (Karnaky,1998). However, the importance of the aquaporins in water movements and associated osmotic homeostasis in fish has received surprisingly little attention. The present paper reports the cloning and functional characterisation of an aquaglyceroporin cDNA from a marine teleost, the gilthead sea bream (Sparus auratus L.). In addition, northern blotting and in situ hybridisation were used to characterise its tissue and cellular distribution.

Adult gilthead sea bream Sparus auratus L. (approximately 350 g)were maintained in through-flow seawater tanks at 17±2°C under natural photoperiod for winter in the Algarve, Portugal. To obtain tissues for mRNA extraction fish were killed by stunning and decapitation, and liver,kidney, gastrointestinal tract (GIT), skin, gills and skeletal muscle were frozen in liquid nitrogen or fixed overnight at 4°C in 4%paraformaldaheyde (PFA). Frozen material was stored at –80°C until utilised for RNA extraction and PFA-fixed material was washed with PTw(phosphate-buffered saline+0.1% Tween-20) and immediately processed for preparation of sections for in situ hybridisation.

Total RNA was extracted from the collected tissue using `TRI reagent'(Sigma-Aldrich, St Louis, MO, USA). The poly(A)⁺ RNA fraction was obtained from total RNA by chromatography on columns of oligo-dT cellulose(Amersham Biosciences, Lisbon, Portugal).

A cDNA library was constructed from 5 μg of gilthead sea bream kidney poly(A)⁺ RNA using the UNI-ZAP XR cDNA cloning kit according to the manufacturer's instructions (Stratagene, La Jolla, CA, USA). The double-stranded cDNA was ligated into the corresponding vector and packaged into Gigapack Gold III packaging extracts (Stratagene). The sea bream kidney library was screened with a 700 bp [α-³²P]dCTP-labelled sea breamAQP cDNA probe (Rediprime, random labelling kit, Amersham Biosciences)generated by RT-PCR. Filters were hybridised overnight at 65°C in Church–Gilbert buffer (1 mmol l^–1 EDTA, 0.25 mol l^–1 NaHPO₄, 7% SDS) and washed 2× 10 min at 65°C in 0.1× SSC/0.1%SDS (1/200 dilution of a solution of 3 mol l^–1 NaCl, 0.3 mol l^–1 sodium citrate, pH 7.0– 20× SSC). One positive plaque was isolated out of 400 000 recombinants, automatically excised into pBluescript (Stratagene) and sequenced to give threefold coverage. The sequence has been deposited in GeneBank (accession number AY363261).

Capped RNA (cRNA) transcripts corresponding to S. auratus AQP were synthesized in vitro with T3 RNA polymerase (Promega, Madison, WI,USA) from XhoI-digested pBluescript vector containing the sbAQP cDNA. The isolation and microinjection of oocytes have been described previously(Deen et al., 1994). In brief,pieces of the ovary from Xenopus laevis were treated with collagenase A (2 mg ml^–1; Roche, Holland) in modified Barth's culture medium (MBS), 0.33 mmol l^–1 Ca(NO₃)₂,0.4 mmol l^–1 CaCl₂, 88 mmol l^–1NaCl, 1 mmol l^–1 KCl, 2.4 mmol l^–1NaHCO₃, 10 mmol l^–1 Hepes, pH 7.5, 0.82 mmol l^–1 MgSO₄, for 2 h at room temperature. After repeated washing with fresh MBS, ovarian follicles at stages V–VI were selected and equilibrated in MBS at 18°C for 24 h. The resulting intact follicles and partially denuded oocytes were then injected with either 50 nl of distilled water (negative control), 50 nl of distilled water containing 10 ng cRNA of sbAQP, or with 1 ng cRNA of human AQP1 as a positive control (data not shown; Denker et al.,1988), and incubated in MBS at 18°C for another 24 h. Subsequently oocytes were completely defolliculated with watchmaker forceps,equilibrated again for 24 h at 18°C, and then used in water and solute transport assays.

The osmotic water permeability (P_f) was measured from the time course of osmotic oocyte swelling in a standard assay. Oocytes were transferred from 200 mOsm MBS to 20 mOsm MBS medium at room temperature, and the swelling of the oocytes was followed by video microscopy using serial images at 2 s intervals during the first 20 s period. The P_f values were calculated taking into account the time-course changes in relative oocyte volume[d(V/V_o)/dt], the oocyte surface area(S), and the molar volume of water (V_w=18 cm³ ml^–1), using the formula V_o[d(V/V_o)/dt]/[S×V_w(Osm_in–Osm_out)]. To determine glycerol permeability, the oocytes were transferred to an isotonic solution containing 160 mmol l^–1 glycerol and complemented with 40 mOsm MBS to adjust the solution to 200 mOsm. The apparent glycerol permeability coefficient(⁠

To examine the effect of mercury on the P_f, the oocytes were incubated in MBS containing 1 mmol l^–1 HgCl₂for 15 min before the swelling assay, which was also performed in the presence of HgCl₂. To determine if the mercurial effect was reversible, the same oocytes were rinsed 3 times in MBS, incubated with 5 mmol l^–1 mercaptoethanol for 15 min, and subjected to the swelling assays 2 h later.

Urea transport activity mediated by sbAQP was measured by uptake of[¹⁴C]urea (Amersham Biosciences). Injected oocytes were transferred to 1 mmol l^–1 urea containing 2 μCi ml^–1(74 kBq ml^–1) in MBS for 20 min, rapidly rinsed 3 times with ice-cold MBS, and lysed in 10% SDS at room temperature followed by liquid scintillation counting.

Between 2 and 5 μg of mRNA from sea bream liver, kidney,gastro-intestinal tract (duodenum, midgut, hind-gut and rectum), skin, gill and skeletal muscle were fractionated on a 5.5% formaldehyde/1.5% agarose gel in MOPS, transferred to a nylon filter (Hybond-N, Amersham) with 10× SSC and cross-linked at 80°C for 2 h. Prior to hybridisation, the filter was washed at 60°C for 20 min in 1× SSC, 0.1%SDS and pre-hybridized in Church–Gilbert buffer for 2 h at 58°C. Hybridisation was allowed to proceed overnight at 58°C in fresh pre-hybridization solution containing the full-length sea bream AQP as the probe, isolated from the kidney library labelled with [α-³²P]dCTP (Rediprime, random labelling kit,Amersham Biosciences). Stringency washes were carried out for 2× 15 min at 60°C in 1× SSC, 0.1% SDS and then at 65°C for 10 min in 0.1× SSC, 0.1% SDS. The blot was exposed to a Phosphoimager (Biorad,Hercules, CA, USA) for 7 days. In order to evaluate the relative amounts of mRNA loaded onto the filter for each tissue, the blot was also hybridized with a sea bream β-actin probe (Santos et al., 1997) following the same protocol as before but reducing exposure time to 2 h. Exposure data was analyzed using the Multi-Analyst/PC software package (Biorad, Lisbon, Portugal).

The distribution of sbAQP mRNA along the gastrointestinal tract, in the gill and kidneys was further investigated by in situ hybridisation. The full-length sbAQP cDNA cloned in pBluescript was digested with BamHI (Promega, Madison, WI, USA) at 37°C for 1.5 h in order to linearise the DNA. The linearised vector was purified and in vitrotranscription was carried out using 20 units of T7 RNA polymerase in transcription buffer (Amersham Biosciences) with 2 μl of digoxigenin-RNA labeling mix (Roche, Lisbon, Portugal), for 2 h at 37°C. The riboprobe synthesis was stopped with 2 μl of 0.2 mol l^–1 EDTA,precipitated with sterile sodium acetate (0.1 volume, 3 mol l^–1; pH 5.2) and 75% ethanol (2.5 volumes) and resuspended in 50 μl of water. Riboprobe purity and concentration were determined by electrophoresis in a 1% agarose gel containing ethidium bromide.

Tissue sections were dewaxed, rehydrated and then pre-hybridised at 55°C for 4 h in hybridisation solution (50% formamide, 4× SSC, 1 mg ml^–1 torula RNA, 0.1 mg ml^–1 heparin,1× Denhart's, 0.1% Tween 20, 0.04% CHAPS). Tissues were then hybridised overnight in a humidified box at 55°C in 40 μl per section of hybridisation solution containing 15 μl ml^–1 of riboprobe. Controls were pretreated with RNase prior to hybridization with riboprobe, or the riboprobe was excluded from the hybridization. High stringency washes were carried out for 3× 5 min at 55°C with 2× SSC. Tissue sections were then washed 2× 5 min with 2× SSC:0.12% CHAPS at 22°C,followed by a wash for 5 min in 2× SSC:PTw (1:1, v/v) and finally 5 min in PTw. Detection of hybridised probe was carried out using anti-digoxigenin-alkaline phosphatase (AP) Fab fragments (1/100) (Roche). The chromagens for colour detection were NBT (4-nitroblue tetrazolium chloride)and BCIP (5-bromo-4-chloro 3-indolylphosphate) and colour development was carried out over 2–17 h at30°C. Stained sections were rinsed in PBS,fixed for 15 min in 4% paraformaldehyde at room temperature, rinsed in PBS and mounted in glycerol gelatine. Sections were analysed using a microscope(Olympus BH2) coupled to a video attachment (Sony 3CCD DXP-930P) linked to a computer for digital image analysis.

A frequent problem in phylogenetic analysis is the limited representation of fish sequences. For a more robust phylogenetic analysis, the genome of the model organism Fugu rubripes(http://fugu.hgmp.mrc.ac.uk)was searched for GLP members using the sbAQP. Fugu scaffolds containing GLP protein amino acid signatures were then analysed using NIX, a Web tool to view the results of a suite of DNA analysis programs(http://menu.hgmp.mrc.ac.uk). The predicted coding regions were extracted and further confirmed using BLAST(Altschul et al., 1990) against SWISSPROT and non-redundant TREMBL. The human genome was also analysed for the presence of GLP homologues. The deduced sbAQP amino acid sequence was aligned with other available aquaporin amino acid sequences in public databases (see Table 1) using CLUSTAL X Version 1.64b (Thompson et al.,1997) and the percent identity between different AQP forms was determined (Table 2). Phylogenetic trees were generated in PAUP* Version 4.0b(Swofford, 1998) to identify maximum parsimony trees for different hypothesis of relationships using Allium cepa AQP1 (AF255795) as an outgroup in this analysis. The sbAQP deduced protein structure was analysed using HMMTOP V.2.0 software(Tusnady and Simon, 2001) and a hydropathy profile obtained using the Kite and Doolittle method(Kyte and Doolittle,1982).

The clone isolated from the newly prepared kidney cDNA library(Fig. 1) spans 1047 bp and encodes a protein of 298 amino acids (aa). The 5′untranslated region(UTR) comprises 84 bp and the 3′UTR comprises 69 bp ending in poly(A)⁺. However, as no polyadenylation signal could be identified this suggests that the 3′UTR of the clone was truncated. Sequence comparison revealed that the encoded protein was most similar to members of the GLP group and shared 48% identity with the recently identified human AQP10, 47% identity with human AQP3, 43% identity with mouse AQP7 and 42%identity with mouse AQP9 (Table 2). The sequence identity between the cloned sbAQP and eel AQP3(46%) was lower than that with mammalian and amphibian AQP3 (50%) and this may be partly explained by the phylogenetic distance between these fish. The sbAQP also exhibited low sequence conservation with CHIPs (<20%, Table 2). None of the previously isolated AQP showed a higher sequence similarity with the sbAQP(Table 2). This is in contrast to eel AQP, which shared 67% identity with human AQP3, and makes it difficult to unambiguously identify sbAQP on the basis of sequence similarity,suggesting that it may be a novel form of AQP.

In common with other aquaporins, six transmembrane domains were predicted in sbAQP from the hydropathy profile and are underlined in the deduced aa sequence (Fig. 1). The MIP family signature, Asn-Pro-Ala (NPA), was present at aa 76–78 and at aa 208–210 of sbAQP, which are located, respectively, in the second (B) and fifth (E) connecting loop. Both the N terminus and C terminus were predicted to be in the cytosol, as for other AQPs. Amino acid residues 8, 18, 33, 84,167, 244, 272 and 291 were cysteine. No potential N-linked glycosylation site[NX(S/T)] was found in sbAQP. The deduced molecular mass of the protein is 31.6 kDa and the theoretical pI 6.5.

Xenopus laevis oocytes injected with cRNA encoding sbAQP had a sevenfold larger P_f than that of oocytes injected with water alone (control, Fig. 2A). Moreover, oocytes expressing sbAQP showed increased P_glyand P_urea (measured, respectively, using volumetric and radiotracer methods) relative to the values found with water-injected oocytes(Fig. 2B,C). Comparison of the relative permeability of sbAQP to water, glycerol and urea revealed that its functional characteristics are similar to mammalian AQP3 and AQP10: it is most permeable to water and its permeability to glycerol and urea is several-fold lower (Echevarria et al.,1996). The water channel function of sbAQP was sensitive to mercuric chloride (HgCl₂). In the presence of 1 mmol l^–1 HgCl₂ there was a ∼70% decrease in osmotic water permeability (P_f). However, it was not possible to reverse this effect by mercaptoethanol treatment. This is in contrast to the observations of human AQP3 and AQP10, where at 1 mmol l^–1HgCl₂ there is significant inhibition of water transport, which is partially reversed in the presence of mercaptoethanol(Ishibashi et al., 2002; Kuwahara et al., 1997b).

Two transcripts of 2 kb and 1.6 kb were identified by northern blotting in the gastrointestinal tract (mid-gut, hind-gut and rectum) and in all these tissues the relative abundance of both transcripts was similar(Fig. 3A). In kidney only the large transcript of 2 kb and in gill only the 1.6 kb transcript were detected. No signal was detected in duodenum or in any of the other tissue analysed. Comparison of the expression of sbAQP with that of β-actin(Fig. 3B) indicated that hind-gut and kidney are the tissues with the highest expression of sbAQP followed by the rectum, mid-gut and gill.

In situ hybridisation was carried out to characterise the cellular expression of sbAQP in the gastrointestinal tract, gills and kidney(Fig. 4). An intense sbAQP signal was found in the epithelial cell lining some of the kidney tubules(Fig. 4A,C). In gill sbAQP expression was most intense in the primary filaments, where the chloride cells are normally found, and expression of sbAQP was also observed at the proximal ends of some of the secondary lamellae(Fig. 4D,F). The expression of sbAQP in the hind-gut was principally in cells scattered in the lamina propria(Fig. 4G), although the identity of these cells remains to be established. Further expression of sbAQP was also found in cells localised at the interface of the circular and longitudinal muscle layer in the hind-gut(Fig. 4I). sbAQP was not found to be expressed in either the columnar enterocytes of the villi or in the numerous goblet cells of the hind-gut. A similar pattern of sbAQP distribution was found in the mid-gut, although the signal was far weaker. Despite being unable to detect sbAQP transcripts in the duodenum by northern blot a very weak signal was found in cells of the mucosa by in situ hybridisation(data not shown). No sbAQP signal was detected in the liver(Fig. 4J) and the controls utilised to verify probe specificity were also negative(Fig. 4B,E,H).

Analysis of the Fugu genome led to the identification in silico of six scaffolds containing putative GLP genes. The extracted Fugu genes were introduced in a multisequence alignment(Fig. 5) used in the phylogenetic analysis. Two Fugu genes (from scaffolds M000233 and M004004) were excluded because they were extremely truncated. The first phylogenetic analysis done showed two principal clades that corresponded,respectively, to GLP and CHIP proteins. To better exemplify the relationship between the GLP family members, phylogenetic analysis was performed with only members of this group and Allium cepa AQP1 was the outgroup(Fig. 6). The GLP clade was composed of separate groups that corresponded to AQP3, AQP7, AQP9 and AQP10. The Fugu gene in scaffold M001883 grouped with AQP3 and was most similar to the available fish sequences. The genes in scaffolds M000386 and M001042 grouped, respectively, with mammalian AQP7 and AQP10. AQP9 was the only GLP in which there was no match with a predicted Fugu GLP gene. The sbAQP formed an independent group clustering with the Fugu gene from scaffold M000648. Extensive searches of the human genome failed to identify a sbAQP paralogue.

We report the cloning of an aquaglyceroporin from a sea bream kidney cDNA library. Its amino acid composition and certain structural features, namely,the presence of two MIP family signatures (NPA) and longer amino acid sequences in the second and third extra-cellular loops(Ishibashi et al., 1998a),support its classification as a GLP protein.

Sequence similarity and phylogenetic analysis incorporating extracted Fugu genes, and the identification of a Fugu gene that clustered in an independent group with sbAQP, strongly suggest that sbAQP is a novel GLP member. Extensive searches of the human genome failed to identify a paralogue, suggesting that the gene found in sea bream and Fugu may have evolved only in the fishes after divergence from the tetrapod lineage.

Characterisation of sbAQP function using the Xenopus oocyte water and solute transport assay indicates that it is a functional fish aquaglyceroporin and transports water, glycerol and urea. Moreover, the osmotic water permeability is mercury (HgCl₂) sensitive. However,treatment with mercaptoethanol fails to reverse the latter effect and it is unclear if this is an artefact caused by non-specific effects, such as the toxicity of the high-dose of HgCl₂ (1 mmol l^–1),on Xenopus oocyte viability or reflects a characteristic of the sbAQP protein (Kuwahara et al.,1997b). Mercurial reagents are thought to block the aqueous pore of AQPs by binding specifically to cysteine residues, and Cys-11 in mammalian AQP3 and Cys-212 in mammalian AQP2 have been shown to confer this characteristic (Kuwahara et al.,1997a). The absence of Cys-11 but presence of Cys-8 in sbAQP may indicate that the latter is responsible for the observed mercury sensitivity,although candidate cysteine residues are present nearby at positions 18, 22 and 33.

Analysis of sbAQP tissue distribution in adult sea bream, demonstrated that it is widely distributed in water-transporting epithelia. The gastrointestinal tract, but more specifically the hind-gut and the kidney, express the highest levels of sbAQP mRNA. The origin of the two transcripts identified in the sea bream gastrointestinal tract has not been determined, although the relatively limited conservation of sbAQP with other members of the GLP group isolated so far in vertebrates (Table 2),and the high stringency conditions utilised for the northern blot, appear to rule out probe cross-hybridisation with these forms. However, identification in silico in the Fugu genome of at least nine AQP raises the possibility that the probe used in the present study could have cross-hybridised with an as-yet-unknown isoform and further work will be required to resolve this question. In human, two forms of AQP10 have been identified, a ∼2 kb transcript and a splice variant of 2.3 kb that lacks the sixth transmembrane domain (Hatakeyama et al., 2001; Ishibashi et al., 2002). Multiple transcripts of another GLP family member,AQP7, have also been reported in rat with two transcripts found in kidney,heart and skeletal muscle (Ishibashi et al., 1997a). It will be of interest to determine the nature of the two transcripts of AQP identified in sea bream and to characterise their function in vivo. For example, in man the 2.3 kb splice variant of AQP10 lacking the sixth transmembrane domain is permeable to water but not to glycerol or urea, unlike the full-length 2 kb transcript.

In man the duodenum and jejunum are the major sites of expression of AQP10,and these are also the sites for entrance and secretion of large volumes of water (Hatakeyama et al.,2001). A similar observation is made with AQP3, which is expressed in the basolateral membrane in the collecting duct of kidney(Ecelbarger et al., 1995; Echevarria et al., 1994; Ma et al., 2000) and in the rectum and hind-gut (Ramirez-Lorca et al.,1999). The general presence of AQPs in membranes important for water movement has led to the suggestion that transcellular as well as paracellular pathways of water movement exist in this tissue in man. The behaviour of aquaporins reported in mammals and their response to movements of water driven by osmotic gradients, fits in well with experimental observations of water movement in the fish intestine, where it has been linked to ion movements similar to those reported in ouabain-sensitive water absorption in Atlantic salmon (Usher et al.,1991) and Fundulus heteroclitus (Marshall et al., 2002a,b). In the latter situation it has been demonstrated that the intestine not only absorbs water but also secretes fluid by mechanisms linked, at least in part,to cystic fibrosis transmembrane conductance regulator-like (CFTR-like) ion channel. More studies will be required to determine if sbAQP is directly involved in water absorption in the intestine in order to establish whether it provides a mechanism by which drinking and water absorption can be independently regulated in the fish intestine(Fuentes and Eddy, 1997).

In teleost fish the only members of the MIP family identified are an AQP0 homologue in the killifish Fundulus heteroclitus (Accession number,AF191906), AQP3 homologues in the European eel Anguilla anguilla(Accession number, AJ319533) and zebra fish Danio rerio (Accession number, BC044188), and an AQP in Tribolodon hakonenis (Accession number, AB055465) (Cutler and Cramb,2002; Virkki et al.,2001). In killifish AQP0 is restricted to the lens and is present as two transcripts of 2.8 and 1.8 kb, which are suggested to correspond to alternative splice variants. AQP3 in the European eel has a wider tissue distribution, being present in low abundance in the eye, oesophagus and intestine as a single transcript of 2.4 kb and in high abundance in the gills,where an additional transcript of 7 kb is also identified. Moreover, in the eel the level of expression of AQP3 appears to be regulated by the composition of the bathing medium, so that the expression of AQP3 in the gills of freshwater-adapted eel is several-fold higher than in those adapted to saltwater. Comparison of the eel AQP3 and sbAQP tissue distribution reveals an overlap, although the relative abundance of transcripts seems to differ. The gill in freshwater fish plays an important role in total body water influx and is suggested to account for up to 90% of water uptake(Haywood et al., 1977). However, this is not true in marine fish where the presence of leaky epithelia ensures high water loss through the gills, in particular the chloride cell complex. The distribution of sbAQP mRNA in the gills of the sea bream coincided with that described by immunohistochemistry for AQP3 in seawater-adapted European eel (Lignot et al., 2002). The mRNA of AQP in seabream and protein of AQP3 in European eel is principally localised in the chloride cells within the primary filaments of the gills and at the proximal ends of the secondary lamellae. The physiological role of sbAQP in the gill remains to be established, whether it is directly involved in water movement and/or also associated with elimination of nitrogenous waste products (Isaia,1984).

The tissue distribution of sbAQP coincides with that of the mammalian GLP proteins, AQP3, AQP7 and AQP10. Surprisingly the expression of eel AQP3 in the kidney is very low and does not coincide with the abundant expression in the basolateral membrane of the collecting duct(Ecelbarger et al., 1995; Echevarria et al., 1994; Ma et al., 2000) in mammals. In seawater fish such as the sea bream, the kidney is unable to produce highly concentrated urine and instead produces low quantities of urine containing abundant quantities of divalent ions such as Mg²⁺, Ca²⁺or even SO ^2– 4. The relatively high expression of sbAQP in the kidney may be related to functional adaptations in the kidney of seawater fish, where water movements are thought to be linked to monovalent ions such as Na⁺, Cl^– resorption and secretion in the tubules (Beyenbach and Liu,1996).

In conclusion, a sbAQP cDNA was isolated from a kidney cDNA library and its primary structure was found to be homologous to other GLP proteins, with the same functional characteristics. Phylogenetic analysis of sbAQP, including putative GLP genes from Fugu, demonstrated that it did not group with any of the previously isolated GLP and instead formed a separate group,suggesting that it may be a novel GLP member.

This work was supported by project PRAXIS XXI/2/2.1/BIA/211/94 from the Portuguese National Science and Technology Foundation (FCT), co-financed by EU structural funds, DG-Fisheries Project Q5RS-2002-00784 (CRYOCYTE) and an EU Biotech grant (QLRT2000-00778). C.R.A.S., J.C.R.C. and J.F. were in receipt of FCT fellowships PRAXIS XXI/BPD/22040/99, PRAXIS XXI/BD/19925/99BPD/22033/99,respectively.