Molecular mechanisms for intestinal HCO₃⁻ secretion and its regulation by guanylin in seawater-acclimated eels

Marine teleosts drink seawater (SW) copiously and absorb significant amounts of imbibed SW across the intestine to compensate for osmotic water loss in the hypertonic environment (Grosell, 2011). The ingested SW is diluted to isotonicity by coordinated actions of NaCl removal by the esophagus (Hirano and Meyer-Gostan, 1976; Parmelee and Renfro, 1983; Takei et al., 2017) and Ca²⁺/Mg²⁺ removal by precipitation as carbonate in the intestine (Grosell et al., 2009b; Wilson et al., 2009; Faggio et al., 2011). Water is then absorbed by the intestine in parallel with NaCl absorption from diluted SW (Grosell, 2011). The molecular mechanisms for NaCl absorption across the intestinal epithelia have been well characterized in a few teleost species; Na⁺-K⁺-2Cl⁻ cotransporter 2 (NKCC2 or Slc12a1) is the major player for NaCl uptake on the apical membrane of enterocytes, and Na⁺/K⁺-ATPase (NKA) and an unidentified Cl⁻ channel on the basolateral membrane transport NaCl into the extracellular fluid (Cutler and Cramb, 2008; Gregŏrio et al., 2013; Ando et al., 2014; Esbaugh and Cutler, 2016). The molecular mechanism for HCO₃⁻ secretion has also been investigated in the intestine of marine teleosts (Grosell, 2011). The major transporter for HCO₃⁻ secretion on the apical membrane was identified as Pat-1 (Putative anion transporter-1)/Slc26a6, and that for HCO₃⁻ uptake on the basolateral side was suggested as electrogenic Na⁺-HCO₃⁻ cotransporter (NBCe1/Slc4a4) because of upregulation of these genes after high salinity acclimation in the pufferfish (Kurita et al., 2008), toadfish (Grosell et al., 2009b; Taylor et al., 2010) and seabream (Gregŏrio et al., 2013). The combination of these transporters accomplishes transcellular HCO₃⁻ secretion.

Intestinal ion and water transport is regulated by various hormones (Takei and Loretz, 2011), of which intestinal guanylin (GN) family peptides are promising candidates for the major hormone involved in SW adaptation (Kalujnaia et al., 2009; Takei and Yuge, 2007). We have identified GN, uroguanylin (UGN) and renoguanylin (RGN), and two guanylyl cyclase-C receptors (GC-C1 and GC-C2) in the intestine of Japanese eel (Yuge et al., 2003, 2006). These hormone and receptor genes are expressed abundantly in the different segments of digestive tract, and importantly, their expression is consistently upregulated after transfer of fish from freshwater (FW) to SW. Among the guanylin family, only GN is intestine-specific and secreted from goblet cells into the lumen with mucus (Yuge et al., 2003). Because homologous GN from the eel is available, this euryhaline teleost serves as a good model with which to analyze the mechanisms of GN action on the HCO₃⁻ secretion. It was shown that GN decreased NaCl absorption through inhibition of NKCC2b (Ando et al., 2014) and decreased Cl⁻ secretion into the lumen through stimulation of diphenylamine-2-carboxylic acid (DPC)-sensitive anion channels in the intestine of SW-acclimated eels (Ando and Takei, 2015). In the marine toadfish, eel RGN was shown to decrease HCO₃⁻ secretion through inhibition of Slc26a6 in the posterior intestine, whereas RGN increased HCO₃⁻ secretion in the fish acclimated to concentrated (60 ppt) SW (Ruhr et al., 2015). In view of the highly diversified osmoregulatory mechanisms and their hormonal regulation in fishes (Takei et al., 2014), it is desirable to examine how homologous GN affects the suite of transporters for HCO₃⁻ secretion in the eel.

In this study, we examined the effect of GN on HCO₃⁻ secretion in the intestine of SW-acclimated eels using the pH-stat method in an Ussing chamber. As GN increased HCO₃⁻ secretion, we examined the effect of various inhibitors for transporters on the GN-induced HCO₃⁻ secretion to probe the mechanism. To identify target transporters, we then performed RNA-seq analyses and listed all candidate genes expressed in the eel intestine for HCO₃⁻ secretion. Molecular species of the transporters were further narrowed down by quantitative real-time PCR using paralog-specific probes based on the upregulation after transfer of eels from FW to SW. Among upregulated genes, we finally selected candidate transporters involved in GN-induced transcellular HCO₃⁻ secretion based on sensitivity to the inhibitors.

Cultured eels (Anguilla japonica Temminck and Schlegel 1846) weighing around 200 g were purchased from a pond culture in Yamakine, Chiba, and kept in freshwater aquaria at 18°C for more than 1 week. Some eels were then transferred to SW tanks and kept there for more than 1 week before experimentation (termed SW eels). Eels were not fed after purchase and used up in 1 month after SW acclimation. The loss of body mass was less than 10% after 1 month in this condition. All experiments in this study were approved by the Animal Experiment Committee of The University of Tokyo and were performed according to the guidelines prepared by the Committee.

Eel GN (Peptide Institute, Osaka, Japan), bumetanide (Sankyo Co., Tokyo), 4,4′-dinitrostilbene-2,2′-disulfonic acid (DNDS) and 4,4′-diisothiocyano-2,2′-disulfonic acid (DIDS) (Tokyo Chemical Industry Co., Ltd, Tokyo), and DPC (Wako Pure Chemical Industries Ltd, Tokyo) were purchased from commercial sources. GN was dissolved in water at 10⁻⁴ mol l⁻¹, aliquoted by 100 µl, kept frozen at −20°C, and diluted by the appropriate Ringer’s solution at the time of experimentation. Bumetanide and DPC were dissolved in ethanol at 10⁻² mol l⁻¹, and the final concentration of ethanol was less than 1% after dilution with HCO₃⁻-free Ringer’s. One-percent ethanol alone had no effects on the electrical parameters of tissues in the Ussing chamber. DNDS and DIDS were dissolved in HCO₃⁻-free Ringer’s (for mucosal side addition) or in normal Ringer’s solution (for serosal side) at 10⁻² mol l⁻¹, and diluted to 5×10⁻⁴ mol l⁻¹ before administration.

After decapitation, the intestine of SW eels (n=31 in total) was removed and stripped of the serosal muscle layers to increase the responsiveness (Ando and Kobayashi, 1978). The stripped intestine was opened and mounted as a flat sheet in an Ussing chamber with an exposed area of 0.785 cm². The serosal side of the intestine was bathed with Ringer’s solution (2.3 ml), containing (in mmol l⁻¹) 118.5 NaCl, 4.7 KCl, 3.0 CaCl₂, 1.2 MgSO₄, 1.2 KH₂PO₄, 24.9 NaHCO₃, 5.0 glucose and 5.0 alanine, and was bubbled with a 95% O₂–5% CO₂ gas mixture (pH 7.4). Although eel plasma pH is ca. 7.8, all experiments were performed at pH 7.4 because electrical parameters were more stable (Ando et al., 2014) and the effects of GN and bumetanide were more reproducible at this pH (Ando and Takei, 2015). The mucosal side of the intestine was bathed with HCO₃⁻-deficient Ringer’s solution (2.3 ml), where NaHCO₃ was replaced by NaCl and gassed with 100% O₂ passed through 20 mmol l⁻¹ NaOH to remove CO₂, which would acidify the mucosal saline. The mucosal pH was measured by a pH electrode (MI-410, Microelectrodes Inc., Bedford, NH, USA) and a pH meter (HM-5B, Toa Electronics, Tokyo, Japan), and clamped at pH 7.4 by pH-stat (HSM-10A, Toa Electronics).

The rate of HCO₃⁻ secretion was estimated by the rate of mucosal alkalinization, which was titrated with 0.05 mol l⁻¹ HCl to the clamped pH (7.4). The volume of 0.05 mol l⁻¹ HCl for titration (µl×0.05) was used as the amount of secreted HCO₃⁻ (μmol l⁻¹). The potential difference (PD) was measured at the serosa relative to the mucosa through a pair of electrodes (EK1, WPI, Sarasota, FL, USA) and a voltage/current clamp system (CEZ-9100, Nihon Kohden, Tokyo, Japan). To determine the tissue resistance (R_t), rectangular pulses (20–30 µA, 500 ms) were applied across the intestinal epithelium every 5 min. These data were recorded automatically on a chart recorder (EPR-121A, Toa Electronics). From the deflection of the PD (ΔPD), total resistance was calculated and the R_t was obtained by subtracting fluid resistance (120 Ω cm²) from the total resistance. The short-circuit current (I_sc) was calculated as ΔPD/R_t.

To assess the ion transporter target(s) of GN that lead to increased HCO₃⁻ secretion, various inhibitors were applied on either the mucosal or serosal side of the intestinal epithelia in the Ussing chamber before or after GN administration. The inhibitors applied on the mucosal side were bumetanide, DPC and DNDS/DIDS, because GN inhibited NKCC2 and stimulated Cl⁻ channels in the eel in our previous study (Ando et al., 2014; Ando and Takei, 2015), and because UGN inhibited Pat-1 in the toadfish (Ruhr et al., 2016). DNDS/DIDS was also applied to the serosal side to evaluate the role of basolateral AE and NBC, which transport HCO₃⁻ into the cell.

In order to identify all candidate transporters that are expressed in the eel intestine and could be involved in HCO₃⁻ secretion, we performed RNA-seq analyses using the methods that we reported previously (Wong et al., 2014). For this purpose, eels were transferred directly from FW to SW and the intestine was collected before and 7 days after transfer (n=5 at each time point). After total RNA extraction using Isogen (Takara Bio Inc., Shiga, Japan), the RNA quality was monitored using an Agilent 2100 Bioanalyzer system (Agilent Technologies, Santa Clara, CA, USA) and only RNA with an integrity number >7.0 were used for subsequent procedures and sequencing. The cDNA libraries were prepared from the RNA samples using TruSeq RNA Sample Preparation v2 and sequenced by an Illumina HiSeq 2500 (Illumina, San Diego, CA, USA). The detailed protocol for bioinformatic analysis was described in Wong et al. (2014).

In this experiment, the anterior, middle and posterior parts of the intestine were separately collected at 0, 3, 12 and 24 h and 3 and 7 days after FW–FW and FW–SW transfer (n=6 at each time point). The tissues were immediately frozen, and total RNA was extracted and treated with DNase I (Life Technologies, Carlsbad, CA, USA) to remove genomic DNA contamination. Then, 1 μg of RNA was reverse-transcribed with the Iscript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's protocols. Real-time PCR was performed using Kapa SYBR 2X PCR mix (Kapa Biosystems, Wilmington, MA, USA) and the ABI 7900HT Fast Real-Time PCR System (Life Technologies). The amplification of a single amplicon was confirmed by analyzing the melting curve after the real-time cycling. The melt curve was analyzed by an inbuilt function of real-time PCR package SDS 2.4 provided by Thermo Fisher Scientific. Each reaction contained cDNA transcribed from 2.5 ng total RNA. The PCR reaction efficiency of each assay was monitored by serial diluted cDNA samples, which were pooled from the aliquots of all tested samples. Each sample was assayed in duplicates. All sample runs were successful. Eel elongation factor 1α (eef1a) was used as an internal control to normalize the gene expression among different samples. We also measured the expression of other transporter genes, such as nkcc2b, known to be upregulated in SW eel intestine to confirm the reliability of the assay. We aligned the nucleotide sequences of isoforms to find the mismatched regions to design primers that are specific to one isoform only. Most primers were designed with the assistance of the PrimerQuest program provided by Integrated DNA Technologies. Primer sets that form dimers or multiple peaks in melt curve analysis were discarded. Primer sequences are listed in Table S1.

Statistical analyses of the data were performed using the Wilcoxon signed rank test or the Mann–Whitney U-test, programmed by KyPlot (version 5.0, Kyens Lab Inc., Tokyo, Japan). Results are given as means±s.e.m. and considered significant at P<0.05.

The baseline values before treatments were −25.1±3.2 mV for PD, −501.8±63.1 µA for I_sc and 21.9±1.4 Ω cm² for R_t at the middle intestine (n=20). After application of GN to the mucosal side, serosa-negative PD and I_sc decreased within 5 min, but HCO₃⁻ secretion increased slowly with a latency of ca. 20 min (Fig. 1A). The GN effect on HCO₃⁻ secretion was segment dependent, with the greatest effect on the early segment of middle intestine (Fig. 1B). The GN-induced HCO₃⁻ secretion was accompanied by a remarkable inhibition of the I_sc. However, no correlation was observed between GN-induced HCO₃⁻ secretion and I_sc (Fig. 2).

As GN profoundly inhibited NKCC2 in SW eel intestine (Ando et al., 2014), we compared the effects of GN and bumetanide on HCO₃⁻ secretion in the early segment of the middle intestine, where the GN effect was greatest. When applied to the mucosal side, both GN and bumetanide decreased PD and increased HCO₃⁻ secretion to the same degree, and these effects were not additive after consecutive administration (Fig. 3). Pretreatment with bumetanide abolished the GN effect, and vice versa (Fig. 3C), indicating a common target.

Similar time courses were observed for GN and bumetanide actions, supporting the common target of these drugs (Table 1). Latent periods for the GN effect on PD were longer than those for bumetanide. However, the time from the PD effect to the HCO₃⁻ secretion were similar between GN and bumetanide.

As the effect of DNDS was more stable and consistent than DIDS in the eel intestine, we report here only the results of DNDS treatment. Mucosal application of DNDS failed to change HCO₃⁻ secretion, and subsequent GN application consistently increased it (Fig. 4A). The mucosal DNDS also failed to inhibit the GN-induced HCO₃⁻ secretion (Fig. 4B). Another stilbene inhibitor, DIDS, showed similar effects (data not shown). The inhibition of apical anion channels by DPC did not inhibit the GN action when both were given to the mucosal side (Fig. 4C). However, DNDS applied to the serosal side decreased the GN-induced HCO₃⁻ secretion gradually to a significant level (P<0.001; Fig. 4D).

RNA-seq analyses showed that the intestine expressed some genes of the Slc4 and Slc26 families of AE that are potentially transporters of HCO₃⁻ from serosa to mucosa (Table S2). Some of them are upregulated after acclimation in SW for 7 days. However, because the known paralogous isoforms (e.g. pat1a, b and c) reported in eels could not be distinguished in transcriptome analysis, we used real-time PCR to quantify the expression of each isoform to assess their potential involvement in HCO₃⁻ secretion.

Gene-specific primers for each paralog were designed for qPCR based on the mRNA sequence deduced from the genome database of Japanese eel (Henkel et al., 2012) (Table S1). For instance, two alternative transcripts were identified for Slc26a3 (DRA), and draa expression was more abundant in the posterior region, whereas drab expression was more abundant in the anterior region (Fig. 5). Moreover, draa was upregulated in the posterior intestine, whereas drab was upregulated during the course of SW acclimation in all intestinal regions, particularly in the anterior intestine. Three paralogous pat1 genes were expressed in the eel intestine, and pat1a and pat1b were major transporters as suggested by their expression levels (Fig. 5). The expression of pat1a was greatest in the posterior intestine, whereas pat1b was expressed most abundantly in the anterior intestine. Further, pat1a expression increased gradually and reached maximum after 3 days in SW, whereas the expression decreased for pat1b after SW transfer. The expression of pat1c increased after 3 days in SW, although the expression level was low (Fig. 5).

Concerning basolateral HCO₃⁻ transporters, the expression of sat1 was greater posteriorly along the intestine, and its expression level increased 3 days after SW transfer in all segments (Fig. 6). The expression of slc4a2a and b (ae2a, b) increased at some time points after SW transfer, but their expression levels were very low in all segments (Fig. 6). The expression of slc4a1 (ae1) was even lower than that of ae2s. Three nbc genes were expressed in the intestine, of which nbce1a was the most abundant, followed by nbcn1 and nbcn2a (Fig. 7). The expression of nbce1a decreased along the anterior–posterior axis, whereas that of nbcn1 and nbcn2a increased posteriorly. The expression of the nbc genes increased during the course of SW after acclimation in all intestinal segments (Fig. 7).

The intestine is an important osmoregulatory organ for adaptation to the hyperosmotic SW environment, where water is absorbed from imbibed SW to compensate for osmotic water loss (Grosell, 2011). To achieve efficient water absorption, the teleost intestine secretes HCO₃⁻ into the lumen to remove concentrated divalent ions (Mg²⁺ and Ca²⁺) in SW as carbonate precipitates and to decrease luminal fluid osmolality (Grosell et al., 2005; Wilson et al., 2002, 2009). Hormones such as growth hormone/IGF-I and cortisol, which re-organize osmoregulatory organs to a SW-type, have long been implicated in SW adaptation (McCormick, 2001; Takei and McCormick, 2013), but investigations of the short-acting, oligopeptide hormones are still new. One of the promising candidates for SW-adapting hormone is the GN family of peptides as they are the only oligopeptide hormones thus far identified that are upregulated at both ligand and receptor levels after transfer of eels to SW (Comrie et al., 2001a,b; Yuge et al., 2003, 2006). Thus, it is quite intriguing to examine how GN regulates HCO₃⁻ secretion through its action on the transport proteins. In this study, we showed that eel GN enhances HCO₃⁻ secretion in SW-acclimated eel intestine; however, the effect does not appear to be a direct action on Pat-1 (Slc26A6) on the apical membrane as suggested in other species but mediated by its inhibition of apical NKCC2b. The resulting decrease in cytosolic Na⁺ and Cl⁻ stimulates apical DRAs and/or Pat-1s and basolateral NBCs for HCO₃⁻ secretion.

The transporters involved in HCO₃⁻ secretion have been pursued in several teleost species. On the apical membrane of intestinal epithelial cells, Pat-1 may play an important role in HCO₃⁻ secretion as its stoichiometry is nHCO₃⁻/1Cl⁻ as shown in the euryhaline pufferfish Takifugu obscurus (Kato et al., 2009), although there are conflicting reports on the stoichiometry of Pat-1 in mammals (see Alper and Sharma, 2013). In fact, pat1 is profoundly upregulated after transfer of this euryhaline pufferfish from FW to SW (Kurita et al., 2008) and after transfer of toadfish (Ruhr et al., 2016) and sea bream (Gregŏrio et al., 2013) from SW to concentrated SW. Consistently, mucosal DIDS, but not serosal DIDS, inhibited HCO₃⁻ secretion in the sanddab, Citharichthys sordidus (Grosell et al., 2001). In eels and other teleosts, however, serosal application of DIDS/DNDS, but not its mucosal application, was effective for inhibition (Dixon and Loretz, 1986; Ando and Subramanyam, 1990; Faggio et al., 2011), as confirmed in the present study. Further, HCO₃⁻ concentration of the serosal fluid greatly influences HCO₃⁻ secretion in these teleosts. The responsible transporter for HCO₃⁻ uptake on the serosal side has been suggested as NBCe1 in a few species of teleost because of upregulation in hyperosmotic environments (Kurita et al., 2008; Taylor et al., 2010; Grosell, 2011; Gregŏrio et al., 2013). Thus, transcellular HCO₃⁻ transport could be conducted by apical Pat-1 and basolateral NBCe1 in some teleost species. The cytosolic HCO₃⁻ may also be provided by hydration of CO₂ catalyzed by cytosolic carbonic anhydrase (Grosell et al., 2009a; Gregŏrio et al., 2013).

By RNA-seq and subsequent qPCR, we found that three pat1 paralogs were expressed in the eel intestine, and pat1a and pat1c were upregulated after SW transfer in different parts of the intestine, suggesting their potential roles in HCO₃⁻ secretion. However, the expression of two dra paralogs was higher than that of pat1, although the reverse is true for the sea bream (Gregŏrio et al., 2013). In particular, the expression of drab was greatest in the anterior intestine and most profoundly upregulated after SW transfer, which coincides with the observation that white precipitates occur in the lumen of anterior intestine within a day after transfer of eels to SW (authors' unpublished observation). In the posterior intestine, draa may take over the role of drab for HCO₃⁻ secretion and Cl⁻ absorption as suggested by the profound upregulation of the gene at this segment. Thus not only Pat-1, but also DRA is likely involved in the HCO₃⁻ secretion in different segments of SW eel intestine. However, mucosal DNDS/DIDS failed to decrease HCO₃⁻ secretion in this study, which suggests that eel Pat-1 and DRA are not involved in HCO₃⁻ secretion in SW eel intestine, or these AEs of eels are insensitive to these stilbene inhibitors. It is important to note that DNDS/DIDS effectively blocked Pat-1 in mammals (Stewart et al., 2009) and teleosts (Boyle et al., 2015), but DIDS was ineffective to block DRA in the rat (Barmeyer et al., 2007) and mouse (Whittamore and Hatch, 2017). There are no studies on DIDS/DNDS sensitivity to teleost DRA, but it is possible that they are ineffective in blocking eel DRAs.

We identified three NBC genes in the eel intestine as basolateral transporters, nbce1a, nbcn1 and nbcn2a, and they are upregulated after SW transfer. Thus, HCO₃⁻ may be taken up from the serosal fluid by electrogenic NBCela and electroneutral NBCn1 and NBCn2a. Interestingly, the stoichiometry of the human NBCe1 is 1Na⁺/2HCO₃⁻; however, external Na⁺ concentration affects the stoichiometry in the NBCe1 of euryhaline pufferfish and the ratio reached 1Na⁺/4HCO₃⁻ at 120 mmol l⁻¹ Na⁺ (Chang et al., 2012). Another candidate for HCO₃⁻ uptake is Sat-1, and this gene was upregulated after SW transfer in the eel intestine. We also identified AE1 and AE2, which are usually located on the basolateral membrane of transport epithelia (Romero et al., 2013), but their expression levels were very low in the SW eel intestine.

As the blood of teleost fishes has higher pH (7.6–8.0) compared with mammals, more NaHCO₃ (ca. 10 mmol l⁻¹) was added to teleost Ringer’s and lower CO₂ gas (0.3–1%) was used for bubbling for teleost in vitro experiments (Table 2). However, the current experiments were performed using the Ringer’s that contains 24.9 mmol l⁻¹ NaHCO₃ and the gas that contains 5% CO₂ under a pH of 7.4. We used this condition because we could not detect any difference in the effects of GN on electric parameters (PD, I_sc and R_t) under the two different conditions, i.e. the current condition (24.9 mmol l⁻¹ HCO₃⁻ Ringer’s under 5% CO₂ bubbling, pH 7.4) and the condition using 10 mmol l⁻¹ HCO₃⁻ Ringer’s under 1% CO₂ bubbling, pH 7.8 as shown in a previous study (Ando and Takei, 2015). However, this unphysiological condition for teleosts should have profoundly affected the rate of HCO₃⁻ secretion in the present study. In fact, HCO₃⁻ secretion increased 4.25-fold when pH 7.4 Ringer’s was used compared with pH 7.8 Ringer’s in the intestine of European flounder (Wilson and Grosell, 2003), and it increased 2.64-fold when 20 mmol l⁻¹ HCO₃⁻ Ringer’s was used compared with 5 mmol l⁻¹ HCO₃⁻ Ringer’s (Grosell and Genz, 2006; Taylor et al., 2010). Thus, it must be noted that HCO₃⁻ secretion is highly enhanced in the current unusual condition, which could allow the GN response to be more easily studied.

The present study showed that GN activates HCO₃⁻ secretion after a latent period of approximately 20 min. The slow enhancement indicates that GN may not act directly on the HCO₃⁻ transporters on the apical membrane of epithelial cells but through its action on NKCC2 inhibition (Ando et al., 2014). Consistently, the effects of GN and bumetanide were similar in both magnitude and time course, and they were not additive. We suggest that the NKCC2 inhibition decreases intracellular Cl⁻ concentration, which then activates DRA/Pat-1 on the apical membrane to take up Cl⁻ and excrete HCO₃⁻. In this way, intracellular Cl⁻ appears to govern the whole process of HCO₃⁻ secretion. The upregulation of two dra and three pat1 genes supports this hypothesis. If eel Pat-1s are electrogenic as discussed above, the serosa-negative PD should decrease after GN action. However, there was no correlation between changes in HCO₃⁻ secretion and PD after GN, suggesting that the induced HCO₃⁻ secretion is electroneutral. As DRA is electroneutral in mammals (see Alper and Sharma, 2013), it is possible that electroneutral and DNDS-insensitive DRAs are involved in GN-induced HCO₃⁻ secretion in the eel (Fig. 8).

As a second candidate for the route of HCO₃⁻ secretion, we considered Cl⁻ channels such as cystic fibrosis transmembrane conductance regulator (CFTR), because GN stimulates apical Cl⁻ channel in the SW eel intestine (Ando and Takei, 2015) and HCO₃⁻ is known to pass through the Cl⁻ channels (Rubenstein, 2018). Two CFTR genes exist in the eel, but the cftr expression was low and decreased after SW transfer in the eel intestine (Ando et al., 2014; Wong et al., 2016). Considerable expression of cftr was reported in some teleost species (Marshall and Singer, 2002; Gregŏrio et al., 2013), and it was recently reported that eel RGN induced CFTR insertion into the apical membrane of toadfish enterocytes (Ruhr et al., 2018). However, mucosal DPC, which effectively abolished GN-induced Cl⁻ secretion in the eel intestine (Ando and Takei, 2015), failed to inhibit GN-induced HCO₃⁻ secretion in the present study. Thus, it is unlikely that GN directly stimulates the Cl⁻ channel including CFTR for HCO₃⁻ secretion in the SW eel intestine. The slow electroneutral effect of GN on HCO₃⁻ secretion also excludes the direct involvement of the Cl⁻ channel in the GN effect. The remaining possibility is that GN further decreased cytosolic Cl⁻ through the Cl⁻ channel, and the decreased cytosolic Cl⁻ activates DNDS-insensitive DRA (Fig. 8). The apical Cl⁻ channel may be active in SW eel intestine in vivo, because luminal Cl⁻ concentration is very low and the function of the DPC-sensitive Cl⁻ channel becomes evident when mucosal Cl⁻ concentration is very low in the Ussing chamber in vitro (Ando and Takei, 2015).

In the intestine of mammals, GN enhances HCO₃⁻ secretion through its action on CFTR, but the effect is much faster than that observed in the eel intestine (Guba et al., 1996; Sellers et al., 2008). The GN effect was diminished in CFTR-deficient mice or after blocking the apical Cl⁻ channels by inhibitors (Seidler et al., 1997; Zuang et al., 2007). The plausible model for GN action is that cGMP produced after binding to GC-C inhibits phosphodiesterase, resulting in cAMP accumulation and activation of CFTR for HCO₃⁻ secretion (Sindic and Schlatter, 2006). However, as GN enhances intestinal HCO₃⁻ secretion in the CFTR-deficient mice (Sellers et al., 2008), GN also acts on HCO₃⁻ transport systems in addition to CFTR (Ko et al., 2002).

In contrast to the mucosal application, serosal DNDS inhibited GN-stimulated HCO₃⁻ secretion. This indicates that HCO₃⁻ uptake across the basolateral membrane is a limiting step for the transcellular HCO₃⁻ flux from serosa to mucosa. Of the HCO₃⁻ secreted into the lumen, 80% was suggested to be taken up from the serosal fluid and the rest was produced within the cell from CO₂ by carbonic anhydrase in the enterocyte of sea bass (Faggio et al., 2011). NBCe1 has been suggested to be responsible for HCO₃⁻ uptake in the pufferfish (Kurita et al., 2008), toadfish (Taylor et al., 2010) and sea bream (Gregŏrio et al., 2013). Six possible HCO₃⁻ transporters, NBCela, NBCn1, NBCn1a, Sat-1, AE1 and AE2, appear to exist on the basolateral membrane of eel enterocytes and three nbc genes and sat1 were upregulated after SW transfer. Thus, it is possible that GN stimulates these basolateral transporters (Fig. 8). The uptake of HCO₃⁻ across the basolateral membrane may explain the slow process of GN-induced HCO₃⁻ secretion. However, it is also likely that the decrease in intracellular Na⁺ after NKCC2 inhibition stimulates Na⁺ and HCO₃⁻ cotransport by the NBCs, which then promote apical DRA for HCO₃⁻ secretion (Fig. 8). The intestinal HCO₃⁻ secretion is inhibited by prolactin in the seabream, which was mediated through the downregulation of nbce1 (Ferlazzo et al., 2012), supporting the rate-limiting role of basolateral NBCe1 in HCO₃⁻ secretion in this species.

In the toadfish intestine, RGN contrarily inhibited HCO₃⁻ secretion, but it increased the secretion slightly in fish acclimated in concentrated SW (Ruhr et al., 2015). The weak effect of RGN may be due to the low HCO₃⁻ concentration of serosal fluid in that study, and the greater dependence on the cytosolic HCO₃⁻ produced by carbonic anhydrase in this species (Taylor et al., 2010). In the eel, GN-induced HCO₃⁻ secretion depends deeply on the HCO₃⁻ uptake from the serosal fluid. Judging from the greater dependence of the RGN effect on Pat-1 and CFTR in the toadfish, the molecular mechanisms for HCO₃⁻ secretion in the intestine and its regulation by hormones are highly diverse among teleost species, as observed in other osmoregulatory organs (Takei et al., 2014). Concerning other hormones, stanniocalcin stimulated HCO₃⁻ secretion in the sea bream intestine and parathyroid hormone-related peptide inhibited it (Fuentes et al., 2010, Table 2).

The authors thank Drs Haruka Ozaki, Wataru Iwasaki and Yuzuru Suzuki of the University of Tokyo for transcriptomic analysis, and Dr Christopher A. Loretz of State University of New York at Buffalo for his valuable comments on the manuscript and polishing of the English.