Endothelin and endothelin converting enzyme-1 in the fish gill:evolutionary and physiological perspectives

Endothelin (EDN) is a family of three autocrine/paracrine peptides (EDN1,EDN2 and EDN3) that function in a variety of physiological processes such as the regulation of vascular tone(Yanagisawa et al., 1988) and natruresis in the kidney (Zeidel et al.,1989). Endothelins are translated as ∼200 amino acid (aa)preproendothelins (preproEDN) that are initially cleaved by a furin-like enzyme (Yanagisawa et al.,1988) to form the relatively inactive 38 aa proendothelin (proEDN,also known as Big-EDN) (Kimura et al.,1989). Proendothelin is further cleaved to form the active 21 aa EDN by the endothelin converting enzyme (ECE-1 and/or ECE-2)(Shimada et al., 1994; Xu et al., 1994). In mammals,actions of EDNs are mediated via two G-protein coupled receptors:endothelin A receptor (EDNRA), which preferentially binds EDN1(Arai et al., 1990), and endothelin B1 receptor (EDNRB1), which binds all three EDNs with equal affinity (Sakurai et al.,1990). In non-mammalian vertebrates, EDNs also equally bind to a third G-protein coupled receptor, endothelin B2 receptor (EDNRB2)(Lecoin et al., 1998), and in amphibians EDN3 binds to an amphibian-specific receptor termed endothelin C receptor (EDNRC) (Karne et al.,1993).

Preproendothelin genes have been found in all major gnthastome clades(Fig. 1), and there is evidence for EDN1 regulation of vascular tone in fishes(Olson et al., 1991; Evans et al., 1996; Evans, 2001; Evans and Harrie, 2001; Wang et al., 2001). In addition, EDN1 inhibition of transport by the multidrug resistance-association protein was demonstrated in shark (Squalus acanthias) rectal tubules(Miller et al., 2002) and killifish (Fundulus heteroclitus) renal tubules(Masereeuw et al., 2000). Recently, Evans et al. (Evans et al.,2004) determined that exogenous (mammalian) EDN1 inhibited net chloride transport in the killifish opercular epithelium, a tissue used as a model for the SW teleost gill (Karnaky et al., 1977). In teleosts, the gill is the main site for ion balance, nitrogen excretion, acid–base regulation and gas exchange(Evans et al., 2005). Estuarine euryhaline fishes like the killifish (Fundulus heteroclitus) encounter varying environmental salinities throughout the day (Marshall, 2003),resulting in a net gain or loss of ions depending on the water salinity; thus the regulation of gill ion transport is an important mechanism to maintain ionic homeostasis. Evans et al. hypothesized that EDN1 signaling cascades in the gill may be a local regulator of ion balance in fishes(Evans et al., 2004). Thus,the purpose of this study was to determine if EDN1 and ECE1 are produced in the killifish, and secondarily to determine if environmental salinity regulates gill EDN1 and/or ECE mRNA expression. We were also interested in determining the phylogenetic/evolutionary relationships of the EDNs and ECE family of proteins.

Killifish Fundulus heteroclitus L. were trapped in North East Creek, Mount Desert Island, ME, USA and maintained in free flowing, 32 p.p.t. seawater (SW) tanks at the Mount Desert Island Biological Laboratory, under a natural summer photoperiod, before being shipped to the University of Florida. There they were maintained in 32 p.p.t., 23°C SW, under a 12 h:12 h light:dark photoperiod. Tank pH was maintained between 7.8 and 8.0, and ammonia, nitrate and nitrite levels were below 1 p.p.m. The fish were fed commercial fish pellets to satiation every other day.

All protocols and procedures were approved by the Institutional Animal Care and Use Committee at the University of Florida. Previously published molecular protocols were used (Hyndman et al.,2006). Killifish were decapitated, and the gills of the right side were removed and snap frozen in liquid nitrogen. Total RNA was then isolated with TRI-reagent (Sigma, St Louis, MO, USA), and 5′ and 3′ RACE cDNA was synthesized from 4 μg of total RNA using a GeneRacer™ Kit(Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocols. Published degenerate reverse EDN1 primers(Wang et al., 2006) were used in our initial 5′ Touchdown RACE PCR following Invitrogen's protocols. The polymerase used was 0.625 Units of Ex Taq, hot start, DNA polymerase(Takara Bio, Madison, WI, USA) and the reactions were run in an Express thermocycler (ThermoHybaid, Franklin, MA, USA). The PCR parameters were:94°C for 2 min, 5 cycles of 94°C for 30 s, 72°C for 30 s, 5 cycles of 94°C for 30 s, 70°C for 30 s, 30 cycles of 94°C for 30 s,45°C for 30 s, 72°C 30 s, and a final 72°C for 5 min. PCR products were visualized by ethidium bromide staining in 1.5% agarose gels, ligated into pCR®4-TOPO vectors, and transformed into TOP10 chemically competent cells using a TOPO TA Cloning® Kit for sequencing (Invitrogen). Plasmid DNA was then sequenced in both directions at the Marine DNA Sequencing Facility at the Mount Desert Island Biological Laboratory (Salisbury Cove, ME,USA). Once we had the 5′ end, specific killifish EDN1 primers were designed (Table 1) and 3′ Touchdown RACE PCR was performed to complete the cDNA sequences. The PCR parameters were: 94°C for 2 min, 5 cycles of 94°C 30 s, 72°C for 30 s, 5 cycles of 94°C for 30 s, 70°C for 1 min, 30 cycles of 94°C for 30 s, 50°C for 30 s, 72°C for 1 min, and a final 72°C for 10 min. PCR products were cloned and sequenced as above.

To obtain ECE cDNA, 4 μg of total gill RNA (extracted as described above) was reverse transcribed using the First-strand cDNA Superscript™ III reverse transcriptase kit (Invitrogen) with oligo-dT primers. Initial degenerate primers used to generate ECE were designed using CODEHOP (Rose et al., 2003)and are recorded in Table 1. PCRs were run on 0.5 μl of the oligo-dT cDNA, with 0.625 units of Ex Taq,hot start (Takara) and standard cycling parameters. PCR products were cloned and sequenced as described above.

Sequence results for each transcript were assembled with GeneTools software(BioTools Inc., Edmonton, Alberta, Canada) and killifish EDN1 and ECE1 nucleotide sequences were searched for open reading frames(ORFs). The resulting amino acid translations were analyzed with the basic local alignment search tool (Blast) on the National Center for Biotechnology Information (NCBI) website. The predicted amino acid sequences were aligned with other full-length vertebrate EDN or ECE proteins using Clustal X(Chenna et al., 2003). All sequences were taken from GenBank or the Genome projects in Ensembl (e:!44 April 2007). Preproendothelin-1 sequences from each major vertebrate clade(mammals to teleosts) were separately aligned, and similarities among the sequences highlighted with GeneDoc (available at http://www.psc.edu/biomed/genedoc),including the expected cleavage sites for furin and ECE(Yanagisawa et al., 1988; Opgenorth et al., 1992). To determine the relationship among our sequences and those from other organisms,EDN and ECE alignments were exported to PHYML(Guindon et al., 2005) and a Fast Maximum-Likelihood test was performed, following the WAG model of amino acid substitutions (Whelan and Goldman,2001) and a calculated gamma of 1.023 and 1.03, respectively. Branches were then tested for statistical significance by bootstrapping with 500 replicates.

To determine the distribution of EDN1A, EDN1B and ECE1mRNA among tissues, relative duplexing semi-quantitative PCR was performed on total RNA from gill, opercular membrane, brain, heart, stomach, intestine and kidney tissue, as described previously(Choe et al., 2004; Choe et al., 2005). Briefly,cDNA was produced from the tissues of a SW killifish as described above, but random hexamer primers (not oligo-dT primers) were used so that ribosomal and messenger RNA would be reverse transcribed. Non-degenerate primer pairs were designed to amplify a product with high efficiency (e.g. high melting temperature), and to minimize the chance of amplifying contaminating genomic DNA, the primer pair was designed to include at least one intron–exon boundary when possible (Table 1). A QuantumRNA™ 18S internal standard primer kit (Ambion,Woodward Austin, TX, USA) was used to control for variability in cDNA quality and quantity between the different tissues tested. Duplexing PCR with primers for 18S and either EDN1A, EDN1B or ECE1 were then optimized to ensure that the reactions were terminated during the exponential phase. Lastly, the products were visualized by ethidium bromide staining in 1.5%agarose gels and digitized using the Biorad Gel Doc™ XR System.

Killifish were acclimated to SW (approximate concentrations in mmol l^–1: Na⁺ 517, Ca²⁺ 9, K⁺ 12,Cl^– 486) (Choe and Evans,2003) or freshwater (FW; Gainesville dechlorinated tapwater,approximate concentrations in mmol l^–1: Na⁺ 4,Ca²⁺ 1, K⁺ 0.03, Cl^– 0.40)(Choe and Evans, 2003) for 2 weeks, at which point the SW killifish were transferred into FW (SW→FW)and the FW killifish were transferred into SW (FW→SW). An additional set of killifish were removed and replaced into SW or FW as sham controls(SW→SW and FW→FW, respectively). Immediately after transfer, 5 or 6 killifish from each treatment were sacrificed, the gills excised and snap frozen for RNA extraction and cDNA synthesis. Killifish (N=5 or 6/treatment) were further sacrificed at 3, 8 and 24 h post transfer (acute acclimations), as well as 30 days post transfer (chronic acclimation). RNA was extracted from all of the samples and oligo-dT cDNA synthesized as described above.

To determine the effects of environmental salinity on killifish gill EDN1A, EDN1B and ECE1 mRNA levels, quantitative real-time PCR (qRT-PCR) was performed. Nondegenerate primers were designed to amplify a product between 50–100 bp across a predicted intron–exon boundary(Table 1). L8 was used as an internal control gene, as previously described(Choe et al., 2005; Choe et al., 2006). Each sample was run in triplicate using 2 μl of 1/10 diluted original cDNA, 7.4 pmol of primers and SYBR® Green Master Mix (Applied Biosystems, Foster City, CA, USA) in a total volume of 25 μl. The cycling parameters used were: an initial denaturing step of 95°C for 10 min, 40 cycles of 95°C for 35 s, 60°C for 30 s and 72°C for 30 s, followed by a melting curve analysis to ensure only one product was amplified. Random samples were also sequenced following qRT-PCR, confirming amplification of the target of interest. To determine the degree of possible genomic contamination, qRT-PCR was run using RNA samples that were not reverse transcribed, and we determined that there was no genomic contamination. All qRT-PCRs were run on a MyiQ quantitative thermocycler (Biorad, Hercules, CA, USA).

Each primer pairs' efficiency was determined by performing a tenfold dilution curve using plasmid cDNA. Efficiency (E) for each primer pair was calculated using the equation: E=–1+10^(–1/slope), where`slope' was the slope of the dilution curve. Each cycle threshold (CT) value was subtracted from a randomly chosen control sample resulting in a ΔCT,and were analyzed using the Pfaffl equation(Pfaffl, 2001): ratio=E^ΔCTtarget/E^ΔCTL8. Each Pfaffl ratio was then standardized to the average chronic seawater Pfaffl ratio.

Values are expressed as means ± s.e.m. (standard error). For qRT-PCR data, a two-factor ANOVA was performed to determine whether effect of environmental salinity over time differed between SW→FW transfers and SW→SW shams or FW→SW transfers and FW→FW shams. If statistical significance was found a one-factor ANOVA was run to determine the effect of time over a treatment group. Finally, all time points were compared to sham time points with unpaired t-tests to determine if salinity transfers altered mRNA expression. All values that did not meet homogeneity or equal variance tests were log transformed to meet the assumptions of the ANOVA; P=0.05.

Killifish gills were fixed in 4% paraformaldehyde in 10 mmol l^–1 phosphate buffered saline (PBS) pH 7.3, for 24 h,dehydrated in an increasing concentration of ethanol, cleared in Citrisolv(Fisher Scientific, Pittsburgh, PA, USA), and embedded in paraffin wax. The tissue blocks were cut at 7 μm, placed on Superfrost Plus slides (Fisher Scientific), and heated at 37°C overnight.

mRNA for EDN1A, EDN1B and NKA(Na⁺,K⁺-ATPase) mRNA were visualized using in situ hybridization. An ECE1 mRNA probe was not made because our partial sequence of that transcript was from the middle of the sequence, a region that in other fishes is >65% identical to ECE2, and we were afraid of the potential cross-reactivity of this probe. Specific digoxigenin(DIG)-RNA probes (sense and antisense) were made against the 3′ end of the transcripts [including untranslated regions (UTRs) for the EDNs;these regions were <60% identical]. For EDN1A the probe was made from position 520 to the end of the transcript, including the polyA tail (420 bp long). The EDN1B probe was made from position 515 up to and including the polyA tail (419 bp). Both of these transcripts were cloned as described above. A killifish NKA mRNA probe was also made based upon the complete killifish NKA sequence (AY057072). The probe was made from base pairs 915–3115. This transcript was also cloned and sequenced to ensure it was indeed NKA. All of the transcripts were linearized by T3/T7 PCR amplification from the plasmids containing the sequences of interest. DIG-RNA probes were generated by incubating 100–200 ng of the linearized transcripts with the DIG RNA Labeling mix (Roche Applied Science,Indianapolis, IN, USA) following the manufacturer's protocols, at 37°C for 16 h followed by treatment with DNAse for 1 h at 37°C. The DIG-RNA probes were purified using mini Quick Spin RNA columns (Roche) following the manufacturer's instructions, eluted in 80 μl of diethyl pyrocarbonate(DEPC) treated water and stored at –80°C until use.

To determine which gill cells expressed EDN1A or EDN1BmRNA, gill tissue slides were rehydrated in two changes of Citrisolv, followed by incubation in a series of decreasing concentration of ethanol washes. The slides were placed in sterile 10 mmol l^–1 PBS and post-fixed in 4% PFA for 10 min at room temperature (25°C). Following this, the slides were rinsed in sterile 10 mmol l^–1 PBS and incubated in proteinase K (5 mg ml^–1) at room temperature for 5 min. Again, they were washed in 10 mmol l^–1 PBS and post fixed in 4% PFA for 10 min to inactivate the proteinase K. After a final PBS wash, the slides were incubated in prehybridization solution (50% formamide, 10% dextran sulphate, 2% blocking reagent, 0.1% CHAPS, 1% Tween 20, 5 mmol l^–1 EDTA, pH 8.0, 5× SSC, 50 μg ml^–1 heparin, 1 mg ml^–1 tRNA, in DEPC-water)for 2 h at room temperature. Next 200–500 ng of DIG-RNA probes were added to fresh prehybridization solution and the slides were left to incubate at 60°C for 18–24 h. Following this, the tissues were washed for 30 min in 2× SSC at room temperature, 2× SSC at 60°C, two 0.2× SSC 60°C washes, one 0.2× SSC at room temperature and KTB(50 mmol l^–1 Tris pH 7.5, 100 mmol l^–1 NaCl and 10 mmol l^–1 KCl) at room temperature. The tissues were then blocked in 20% normal goat serum diluted in KTB for 1 h at room temperature and incubated in 7.5 U ml^–1 of sheep anti-DIG-AP,Fab fragments (Roche) diluted in normal goat serum (NGS), overnight at 4°C. The slides were then washed in three changes of KTB and incubated in alkaline phosphatase buffer (100 mmol l^–1 Tris, pH 9.5, 100 mmol l^–1 NaCl, 50 mmol l^–1 MgCl₂)for 30 min at room temperature. Visualization of the probes was achieved by incubating the tissues in BCIP/NBT Substrate Kit, 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium (Vector Labs, Burlingame, CA, USA) with levamisole, following manufacturer's instructions, at room temperature until the signal developed (2–6 h). Images were captured using an Olympus BX60 light microscope with a Hitachi KP-D50 digital camera. Image contrast and brightness were adjusted with Photoshop CS (Adobe, San Jose, CA, USA).

Slides were analyzed following published methods(Piermarini et al., 2002; Hyndman et al., 2006). Slides with chronic SW and FW acclimated killifish gill tissue were incubated in primary antibodies: polyclonal, anti-human-proEDN1 (1/1000 dilution) (Phoenix Pharmaceutical, Burlingame, CA, USA) made against the complete 38 aa of human proEDN1, which is 74% identical to both killifish EDN1s. Monoclonal, anti-NKA(α5, 1/1000) was developed by Dr Douglas Fambrough, and was obtained from the Developmental Studies Hybridoma Bank, which was developed under the auspices of the National Institute of Child Health and Human Development of the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA.

From the killifish gill we have sequenced two EDN1 transcripts and designated them EDN1A (accession no. EU009474) and EDN1B(accession no. EU009475). EDN1A is 917 bp with an ambiguous base at position 762 (C or T), and a predicted ORF of 483 bp that translates into a preproendothelin-1A (preproEDN1A) of 189 aa. EDN1B is 912 bp with a predicted ORF of 429 bp that translates into a preproendothelin-1B(preproEDN1B) of 143 aa. The predicted cleavage sites in the preproEDN1s for furin and ECE are depicted in Fig. 2A. As found in other vertebrate preproEDN1 peptides, preproEDN1A and preproEDN1B contain the dibasic cleavage sites for furin and the conserved Trp²¹–Val²² cleavage site for ECE(Fig. 2A). The two active (21 aa) killifish EDN1s are 80% identical to human EDN1 and are 100% identical to each other (Fig. 2B). As seen in Fig. 1, our killifish preproEDN1s group with the other fish preproEDN1 sequences and group outside of the fish preproEDN2 and preproEDN3 sequences. We were unable to find a preproEDN orthologue in the Ciona intestinalis or Branchiostoma genomes, suggesting that the EDNs are found only in vertebrates.

We have sequenced 1696 bp from the middle of the killifish ECE1cDNA (accession no. EU009476). This translates into 565 aa from the killifish ECE1. Endothelin converting enzymes are found in all organism including Bacteria and Archaea (Fig. 3). As seen in Fig. 3, the non-vertebrate ECEs are not well resolved (i.e. lancelet ECE groups with locust and sea urchin, while sea squirt groups with hydra), and although there is no clear explanation for this, Fig. 3 shows three distinct ECE clades: ECE1, ECE2 and non-vertebrate ECE. Our partial sequence of ECE1 from the killifish groups with the other fish ECE1 sequences confirming it is ECE1.

Using duplexing relative semi-quantitative PCR we found EDN1A mRNA in the gill, opercular epithelium, brain, heart, stomach, intestine and kidney of the killifish (Fig. 4). Relatively high expression was found in the gill, brain and kidney. EDN1B mRNA was not found in the opercular epithelium, and had very little expression in the gill, but was highly expressed in the brain, kidney and intestine. Finally, ECE1 mRNA was found in all of the tissues tested, with highest expression in the stomach, intestine and gill(Fig. 4).

EDN1A mRNA was localized to gill epithelial cells in the interlamellar region (Fig. 5A). These cells were large, round and were dispersed along the entire length of the filament (not shown). NKA mRNA was also found in epithelial cells in the interlamellar region (Fig. 5C) but was concentrated only near the afferent filamental artery(trailing edge of the filament, not shown). The morphology of these NKA positive cells is consistent with mitochondrion-rich cells (MRCs)of the killifish gill (Katoh et al.,2001; Marshall,2003; Choe et al.,2006; Hyndman et al.,2006). In contrast, EDN1B mRNA expression was rare (as was seen in the EDN1B tissue distribution described above); however,it was found in some lamellar pillar cells(Fig. 5D,F) and in epithelial cells adjacent to the environment (Fig. 5D). Incubation of slides with sense probes produced a little non-specific staining (Fig. 5B,E).

Proendothelin-1, the precursor to active EDN1, was immunolocalized to epithelial cells in the interlamellar region of the killifish gill, and on the lamellar pillar cells (Fig. 6A,C). Proendothelin-1 immunoreactivity was seen in epithelial cells adjacent to cells immunoreactive for NKA(Fig. 6A,C), and these cells share similar morphology to neuroendocrine cells (NECs) described in fishes(Zaccone et al., 1992; Goniakowska-Witalinska et al.,1995; Zaccone et al.,1996; Mauceri et al.,1999). This immunolocalization matches the killifish EDN1mRNA localization shown in Fig. 5. Slides incubated in NGS and double labeled with NKA were done as negative controls, and showed no non-specific staining of the proEDN1 secondary antibody and chromagen (Fig. 6B,D).

When killifish were transferred from FW to SW there were no statistically significant changes in EDN1A or EDN1B mRNA levels over 24 h(acute acclimation) compared to sham (FW→FW) transfers(Fig. 7A,C). ECE1 mRNA increased four- and sixfold compared with sham ECE1 mRNA levels at 8 and 24 h post a FW→SW transfer (Fig. 7E). EDN1A mRNA levels did not change significantly with acute acclimation from SW→FW compared to sham (SW→SW) treatments(Fig. 7B), but 24 h EDN1B mRNA levels were almost threefold higher compared to sham EDN1B mRNA levels at 24 h (Fig. 7D). ECE1 mRNA levels were twofold higher after 3 and 24 h of acclimation to FW but did not differ from sham values at 8 h post transfer (Fig. 7F). With chronic acclimation (30 days), there were no statistical differences between the SW- and FW-acclimated killifish for EDN1A, EDN1B or ECE1mRNA levels (Fig. 8). In addition, preproEDN1 (protein) immunolocalization did not differ between fish chronically acclimated to FW (Fig. 6A,B) or SW (Fig. 6C,D).

We have sequenced two cDNAs for EDN1 from the killifish gill. These two transcripts have identical 5′ ends until base pair 510, where there is an AC insertion resulting in a frame shift mutation in EDN1Athat changes the stop codon TGA to ACT GAG. Conversely, EDN1B may have lost an AC at position 510 resulting in a stop codon and a truncated preproEDN. Past this insertion/deletion (indel), the two transcripts differ along the rest of the sequence, but are only 5 bp different in total length. This suggests that the two transcripts are from duplicated EDN1 genes and are not transcript variants. We did not find duplicate EDN1 genes in the Takifugu, Tetraodon or Danio genomes, suggesting the EDN1 duplication was a Fundulus-specific event (see note added in proof). As a consequence of the indel at position 510, the predicted preproEDN-1a is 46 aa longer than preproEDN-1b. Even though there are differences between the killifish preproEDN-1s, the proEDN-1s and EDN1s are 100% identical. Why would a tissue (cell) produce two different mRNAs if the end protein translated is 100% identical? We hypothesize that these two EDN1s have remained in the killifish because they have different regulatory pathways, and/or are stimulated by different signals. This hypothesis might suggest that the EDN1s would have different tissue/cellular distributions. Supporting this hypothesis, from our tissue distribution analysis (Fig. 4), we found EDN1a was ubiquitously expressed in all the tissues tested, with relatively high expression in the gill, brain and kidney. Conversely, EDN1b was found in very low levels in the gill and opercular epithelium, but relatively high in the brain. Nevertheless, until complete killifish EDN1A and EDN1B genomic sequences are determined,it is unclear how and what factors may regulate these genes, and thus to suggest why these two have been retained in the killifish.

In the killifish gill we found EDN1A mRNA expression in epithelial cells of the interlamellar region, and EDN1B mRNA expression was found in pillar cells, and in cells adjacent to the environment, in the interlamellar region. Not only are these two transcripts expressed in different levels within a tissue (Fig. 4), they are also expressed in different cells within the gill. From our immunohistochemical experiments, we found proEDN1 immunoreactivity in epithelial cells adjacent to the NKA immunoreactive cells. NKA is commonly used as a marker for the ion transporting, MRCs of the fish gill(Katoh et al., 2001; Marshall, 2003; Evans et al., 2005). Proendothelin immunoreactivity was also found on pillar cells, which is in agreement with our in situ hybridization findings. ProEDN was immunolocalized to gill neuroendocrine cells (NECs) of the eel (Conger congo), catfish (Heteropneustes fossilis) and dogfish(Scyliorhfnus canicula), and the morphology of these cells matches that of the proEDN immunoreactive epithelial cells in the killifish gill(Zaccone et al., 1996).

Endothelin production in the pillar cells was suspected by earlier workers(Sundin and Nilsson, 1998; Stenslokken et al., 1999), who showed that infusion of mammalian EDN1 into the lamellae of the rainbow trout resulted in a `constriction of the vascular sheet' (of the lamellae)and that this was likely due to constriction of the pillar cells. The authors hypothesized that hormonal control of pillar cell tone may be one mechanism to match respiratory needs of a fish while minimizing ion fluxes. There is no evidence that pillar cells are innervated(Bettex-Galland and Hughes,1972; Bettex-Galland and Hughes, 1973); thus endocrine/paracrine/autocrine signaling molecules may be the regulators of pillar cell tone. Pillar cells contain contracting filamentous material(Bettex-Galland and Hughes,1972; Bettex-Galland and Hughes, 1973) and recently an actin-binding protein was described,FHL5, that is highly expressed in these cells, suggesting that they are capable of contraction (Mistry et al.,2004). Video microscopy was used to demonstrate in vivothat pillar cells do contract with EDN1 infusion(Stenslokken et al., 1999). Recently, EDN receptors EDNRA and EDNRB were immunolocalized in the fish gill,and EDNRB was found throughout the gill vasculature, NECs and pillar cells of the cod Gadus morhua (Stenslokken et al., 2006). EDNRA was described in nerve fibers running along the length of the filament and innervating the gill vasculature(Stenslokken et al., 2006). In the fugu Takifugu rubripes EDNRA was found on the pillar cells and in erythrocytes (Sultana et al.,2007). Studies from our lab in the killifish have found EDNRB receptors throughout the gill vasculature and pillar cells, and EDNRA receptors on the mitochondrion-rich cell (K.A.H. and D.H.E., unpublished observations). In the long horn sculpin Myoxocephalus octodecimspinosus, EDNRA receptors were found on the pillar cells, while EDNRB was found throughout the gill vasculature (K.A.H. and D.H.E.,unpublished observations). Evidently, there is species-specific EDN receptor distribution in the gill of fishes.

Killifish usually live in estuaries where there are rapid changes in environmental conditions such as salinity and temperature(Marshall, 2003). We tested the effects of rapid changes of environmental salinity on mRNA expression of EDN1A, EDN1B and ECE1. EDN1 transcript levels did not change with chronic acclimation to FW or SW (Fig. 8). In addition, proEDN1 immunoreactivity in the gill did not differ between SW and FW acclimated killifish(Fig. 6). However, EDN1B and ECE1 mRNA levels increase with acute FW acclimation, suggesting that more active EDN1 protein is produced. 10^–8 mol l^–1 mammalian EDN1 can inhibit net chloride transport in the killifish opercular epithelium, and this is predominately due to stimulation of cyclo-oxygenase (COX) and subsequent prostaglandin production (Evans et al.,2004). These findings suggest that during transfer to a hypo-osmotic environment, EDN1B and ECE1 protein levels increase, resulting in an increase in active EDN1 that could potentially inhibit net chloride transport, helping the fish retain ions. However, we cannot rule out that EDN1 signaling in the gill is different than what was described in the killifish operculum (Evans et al., 2004)(see below). In addition, it is undetermined how volume stress, such as occurs during a rapid change to a hypo-osmotic environment, effects blood flow through the gill. EDN1B was found on gill pillar cells, and may play a role in regulating blood flow during blood volume increases; however, this is an unexplored area of fish gill physiology.

Although we found an increase in EDN1 during acclimation to FW, we unexpectedly found a sixfold increase in ECE1 mRNA levels with acute SW acclimation, suggesting that there is an increase in ECE1 production during this period. This in turn would result in more EDN1 production because the proteolytic cleavage of proEDN1 to EDN1 by ECE1 is a rate-limiting step(D'Orleans-Juste et al.,2003). Our attempts to measure EDN1 production in the fish gill by enzyme immunoassay and Tris-Tricine western blotting were unsuccessful, but measurements of EDN1 levels are necessary to fully understand the role of EDN1 cell signaling in the fish gill. Recently, a 3.4-fold increase was shown in COX-2 mRNA levels in the killifish gill(Choe et al., 2006), 3 h post a FW→SW and a 2.6-fold increase 3 h post a SW→FW transfer, and the authors hypothesized that the increase in COX-2 is an important mechanism for gill cell survival during osmotic stress. A similar result has been demonstrated in mammalian medullary interstitial cells (that experience large changes in osmotic stress), which require functional COX-2 to survive(Hao et al., 1999; Hao et al., 2000). Medullary interstitial cells also contain EDNRs but do not produce EDN1(Dean et al., 1996). Endothelin has been shown to stimulate COX-2 in a variety of mammalian tissues(Hughes et al., 1995; Chen et al., 2003), and EDN1 signaling via endothelial ENDRB results in the production of prostacyclin (Warner et al.,1989; Hirata et al.,1993). Taken together, our findings suggest that during rapid changes in environmental salinity, gill cell survival during this osmotic stress may be accomplished by increased EDN1 production and subsequent stimulation of COX production of prostaglandins. To the best of our knowledge,it is unclear what aids cell survival during salt or water load in fishes, and it is plausible, since EDN1 and ECE1 are ubiquitously expressed, that this may be a more global change in their signaling patterns and is not a gill-specific phenomenon; however, this is yet to be determined and experiments testing these hypotheses are needed. In addition, studies blocking aspects of EDN1 signaling in the gill and subjecting these fish to salinity challenges are vital in understanding EDN1 function during osmotic stress. This technique has been successful in mice models, where kidney collecting duct EDN1 (or EDNRB1)knockout mice who are fed a high salt diet are unable to excrete the excess Na⁺ accumulated and are severely hypertensive(Ahn et al., 2004; Ge et al., 2006), suggesting that EDN1 is necessary for salt excretion in mammals. Applications of these types of techniques to fish models are necessary to fully understand the in vivo role of EDN signaling in the fish gill.

While searching the completed genome projects, we were unable to find EDN orthologues in any organism basal to the teleost fishes. From the maximum likelihood analyses presented in Fig. 1, there are three distinct groups, representing preproEDN1,preproEDN2 and preproEDN3 in vertebrates. Aortic vascular smooth muscle rings from the sea lamprey and Atlantic hagfish constrict in response to mammalian EDN1, suggesting that the receptors and EDN1 are expressed endogenously in these basal, Agnathan vertebrates (Evans and Harrie, 2001). Physiological responses to EDN1 have also been demonstrated in the spiny dogfish shark(Evans et al., 1996; Evans and Gunderson, 1999),again suggesting that EDN1 is produced endogenously. Currently, the exact evolutionary history of this family of peptides is not clear. EDN sequences from hagfish, lamprey and sharks are necessary to determine when these peptides arose, and to determine if it was due to gene/genome duplications or other evolutionary events over vertebrate evolution. Interestingly, mutations in EDN1, endothelin receptors (EDNRs) or ECE result in severe craniofacial developmental abnormalities, and these phenotypes are often lethal(Kurihara et al., 1994; Brand et al., 1998; Clouthier and Schilling, 2004; Nair et al., 2007), suggesting the EDN signaling is necessary for development in vertebrates, and it may be a key innovation in the radiation of vertebrates.

Unlike EDNs, which are only found in vertebrates, ECE is found in all organisms, including Bacteria and Archaea. In Fig. 3, our maximum likelihood analyses reveal three distinct groups of ECEs: prokaryote, fungal and invertebrate ECE, vertebrate ECE1 and vertebrate ECE2. This suggests a gene duplication event sometime after the chordate–vertebrate split, but before the teleost radiation. Since there is no molecular evidence for EDNs or EDNRs in animals basal to the vertebrates why would they have an ECE?Endothelin converting enzymes are zinc-dependant metalloendoproteases and part of the Neprelysin and Kell family (Shimada et al., 1994; Xu et al.,1994). In vertebrates, ECE can function as a monomer or dimer;however, for effective proteolytic cleavage of proEDN1 to EDN1, dimerization at Cys⁴¹² is preferential(Shimada et al., 1996). In contrast, hydra ECE (Zhang et al.,2001) and the other invertebrate, fungal and prokaryote ECEs are missing Cys⁴¹² and are believed to function as monomers(Zhang et al., 2001). Vertebrate ECE has been shown to cleave peptides other than proEDN, including bradykinin, angiotensin I and substance P(Hoang and Turner, 1997; Johnson et al., 1999),suggesting that ECE may be a generalist protease. Although the native substrates cleaved by ECE in non-vertebrate organisms are undetermined, it is plausible that ECE originally cleaved substrates found in all organisms, and during vertebrate evolution started functioning as a dimer and preferentially cleaving proEDN.

To summarize our findings, we propose the following model(Fig. 9) of paracrine and autocrine EDN1 signaling in the fish gill. The diagram shows a lamellar cross-section of the gill (same orientation as the gills in Figs 5, 6, 7), with pillar cells (PCs)highlighted in grey and adjacent pavement cells (PVCs) in white. In the intralemallar region there are two MRCs and an NEC above the gill vasculature. Cyclo-oxygenase-2 (COX-2) and neuronal nitric oxide synthase (nNOS) were previously immunolocalized in the killifish gill, to MRCs(Choe et al., 2006) and NECs,nerve fibers and lamellar arterioles (LA)(Hyndman et al., 2006),respectively. NKA was immunolocalized to the basolateral membrane of the MRC(see Katoh et al., 2001; Choe et al., 2006; Hyndman et al., 2006) and the chloride channel, cystic fibrosis transmembrane conductance regulator (CFTR),to the apical membrane of the MRC (Katoh et al., 2001). From our studies and others, EDNRB were found throughout the gill vasculature (K.A.H. and D.H.E., unpublished)(Stenslokken et al., 2006),and EDNRB and EDNRA were on the pillar cells, depending on the species (K.A.H. and D.H.E., unpublished observations)(Stenslokken et al., 2006; Sultana et al., 2007). EDNRA were found on MRCs in the killifish gill (K.A.H. and D.H.E., unpublished observations). Here we present EDN1 expression in cells adjacent to the MRC(likely NECs) and pillar cells. This suggests a paracrine role of EDN1 signaling, given that it is produced in the NEC and can bind to receptors on the adjacent MRCs (Fig. 9,pathway 1) where it potentially stimulates COX-2 activity, resulting in cell survival during osmotic stress and/or alter ion transport by the MRC as previously hypothesized (Evans et al.,2004). EDN1 can also potentially act as a paracrine binding to EDNRB receptors on the gill vasculature and lamellar arterioles, suggesting it can regulate perfusion of the lamellae(Fig. 9, pathway 1). It also can act as an autocrine on the pillar cells, further supporting the role of regulation of local perfusion across a lamella to meet the respiratory needs of the fish (Fig. 9, pathway 2)(Sundin and Nilsson, 1998; Stenslokken et al., 1999). It may also help maintain lamella integrity during rapid increases in plasma volume during exposure to a hypo-osmotic environment. This is the first model to depict EDN1 signaling in the fish gill, and in the future, studies determining the specific function of EDN1 in the gill and whole fish, are necessary to understand its role in normal fish physiology.

List of abbreviations and symbols

 
• CFTR

  cystic fibrosis transmembrane conductance regulator

 
• COX

  cyclo-oxygenase

 
• CT

  cycle threshold

 
• DIG

  digoxigenin

 
• E

  efficiency

 
• ECE

  endothelin converting enzyme

 
• EDN

  endothelin

 
• EDNRA

  endothelin receptor A

 
• EDNRB1

  endothelin receptor B1

 
• EDNRB2

  endothelin receptor B2

 
• EDNRC

  endothelin C receptor

 
• indel

  insertion/deletion

 
• LA

  lamellar arteriole

 
• MRC

  mitochondrion-rich cell

 
• NEC

  neuroendocrine cell

 
• NKA

  sodium, potassium ATPase

 
• nNOS

  neuronal nitric oxide synthase

 
• ORF

  open reading frame

 
• PC

  pillar cell

 
• preproEDN

  preproendothelin

 
• proEDN

  proendothelin

 
• PVC

  pavement cell

 
• qRT-PCR

  quantitative real-time polymerase chain reaction

 
• UTR

  untranslated region

Since this paper was submitted, we have found EDN1 duplicate genes in the zebrafish, stickleback, Takifugu and Medaka genome projects(Ensembl e!47), suggesting that this is not necessarily a killifish-specific duplicate as we initially thought. This may be the result of a teleost-specific genome duplication event that occurred after the split from the ancestor leading to the tetrapods.

Ensembl numbers:

 
• zebrafish

  ENSDARP00000093511

 
• stickleback

  ENSGACP00000017902

 
• Takifugu

  SINFRUP00000152472

 
• Medaka

  ENSORLP00000011633.

We thank Dr Michael Miyamoto for guidance with our phylogenetic analyses,and Dr Keith Choe for his help with molecular biology. We also thank Drs Michael McCoy and Michelle Monette for their statistical analysis advice. Funding was provided by NSF IOB-0519579 to D.H.E. and Simga Xi GIAR to K.A.H.