The mineralocorticoid receptor contributes to barrier function of a model fish gill epithelium Free

FW fishes are aquatic vertebrates that exist in an environment where ions are scarce and water is abundant compared with the inside of the fish (Bentley, 2002). FW fishes are, therefore, tasked with limiting the paracellular loss of ions via exposed epithelia (e.g. gill and skin) down the concentration gradient, i.e. from the extracellular fluid into the surrounding FW. In this regard, osmoregulation of FW fishes offers excellent insight into the regulation of paracellular permeability of vertebrate epithelia. The teleost fish gill is a primary osmoregulatory organ that constitutes approximately 50% of the surface area of the fish interface with the surrounding environment – it, therefore, plays an important role in the ability of the fish to maintain salt and water balance (Evans et al., 2005). Permeability of the gill must be tightly regulated in order to facilitate effective and directional ion transport that counteracts obligatory diffusional ion movement across the exposed area.

Physical properties of the fish gill epithelium tight junction (TJ) as a structure have been implicated in the regulation of ion traffic across the paracellular pathway of the gill for quite some time (Sardet et al., 1979; Perry, 1997; Evans et al., 2005). Paracellular permeability in the fish gill epithelium is regulated by the molecular machinery of the TJ complex (Chasiotis et al., 2012; Kolosov et al., 2014). However, the cellular heterogeneity and structural complexity of the teleost gill epithelium hinder mechanistic investigation of the roles individual TJ proteins may play in the maintenance of gill epithelial permeability. Cultured gill epithelial models, however, partially circumvent these problems and allow the study of gill epithelium function in vitro (Schnell et al., 2016).

Cortisol is a multifunctional steroid hormone in vertebrates involved in processes like growth, osmoregulation and the stress response (Mommsen et al., 1999). In the gill of FW fishes specifically, cortisol is known to: (i) stimulate active ion uptake and (ii) reduce paracellular loss of ions into FW (Mommsen et al., 1999; Wood et al., 2002; Kelly and Wood, 2001a,b, 2002, 2008; Chasiotis and Kelly, 2011). In mammals, cortisol binds to two receptors: glucocorticoid receptor (GR) and MR (Baker, 2003). Mammals, however, possess a separate mineralocorticoid ligand, aldosterone, that is also able to bind to the MR, but with much higher affinity and at lower concentrations (Baker, 2003). In order to prevent overstimulation of the MR with cortisol (which is present in the plasma at 100-fold greater concentration than aldosterone), mammalian tissues co-express the cortisol degradation enzyme 11β-hydroxysteroid dehydrogenase (11β-HSD) with the MR. 11β-HSD converts cortisol into its inactive form cortisone (Baker, 2003; Chapman et al., 2013). All of these elements of the mineralocorticoid pathway (i.e. cortisol, MR and 11β-HSD) have been described in teleosts, and in salmonids specifically. What is missing from the fish systems is an MR-specific mineralocorticoid or an ability to synthesize one (Baker, 2003). The search for a mineralocorticoid ligand in salmonids returned empty-handed when the response of the gill to 11-deoxycorticosterone and aldosterone was used to consider this possibility in a study dedicated to ion uptake and salt secretion, suggesting that neither of these factors relays corticosteroid function in the gill of these fishes at least (McCormick et al., 2008). Further support for this idea comes from the observations of Kelly and Chasiotis (2011), who reported that the mineralocorticoid ligands aldosterone and deoxycorticosterone did not alter paracellular permeability of a primary cultured FW rainbow trout gill epithelium, but MR blockade, cortisol and the glucocorticoid agonist dexamethasone did (Kelly and Chasiotis, 2011).

Cortisol reduces the paracellular movement of inert permeability markers as well as solutes across cultured gill epithelia in a dose-dependent manner under symmetrical (Leibovitz's L15 medium on basolateral and apical sides) and asymmetrical (basolateral L15, apical FW; where basolateral to apical ion movement is equivalent to ion loss) conditions (Kelly and Wood, 2001a,b, 2002; Zhou et al., 2003). Studies utilizing MR and GR receptor blockers have helped to elucidate that this physiological response to cortisol in cultured trout gill epithelia was relayed through both the MR and GR (Kelly and Chasiotis, 2011). Furthermore, cortisol was reported to increase TJ depth in a cultured trout gill epithelium (Sandbichler et al., 2011a) and to increase protein and/or transcript abundance of TJ proteins such as occludin (ocln/Ocln) (Chasiotis et al., 2010), tricellulin (tric/Tric) (Kolosov and Kelly, 2013) and claudin (cldn/Cldn)-3a, -6, -7, -8b, -8d, -10e, -12, -20a -28b, -30 and -31 (Chasiotis and Kelly, 2011; Kolosov et al., 2017; Kolosov and Kelly, 2017). Interestingly, the mRNA abundance of select TJ proteins did not appear to be responsive to dexamethasone, a GR-specific ligand (Chasiotis and Kelly, 2011). Therefore, it appears that cortisol relays its ‘tightening’ effect on cultured trout gill epithelia in part by increasing the mRNA/protein abundance of cldns/Cldns-28b and -30, and acting exclusively via the MR. Notably, while many cldns seemed to respond to cortisol stimulation through both GRs and MR, cldn-28b and -30 appeared to be responsive through the MR only (Kelly and Chasiotis, 2011).

The GR and MR are a part of the equation that regulates the production, clearance and tissue availability of cortisol (Mommsen et al., 1999; Baker, 2003). Levels of transcript encoding MR in the gill of salmonids are known to alter in response to changes in environmental salinity and cortisol supplementation (Kiilerich et al., 2007, 2011; Chen et al., 2016). Transcript encoding MR has been detected in cultured trout gill epithelia (Kelly and Chasiotis, 2011). However, the specific role of trout MR in a cortisol-mediated reduction in paracellular permeability has only been studied using corticosteroid receptor (CR) blockers, whose specificity in teleost systems can sometimes be contentious (Kelly and Chasiotis, 2011). Nevertheless, select MR blockers have been shown to significantly reduce the cortisol-induced tightening response of a primary cultured trout gill epithelium, so it can be hypothesized that the MR plays a role in mediating the tightening effect of cortisol. To consider this further, the objective of the current study was to further clarify the contribution of the MR to a cortisol-mediated reduction in paracellular permeability of the FW fish gill epithelium by using a primary cultured gill epithelium derived from FW rainbow trout to investigate the response of specific TJ proteins following transcriptional knockdown of mr using double-stranded RNA.

Adult rainbow trout of both sexes, Oncorhynchus mykiss (Walbaum 1792), purchased from Humber Springs Trout Hatchery (Orangeville, ON, Canada), were held under a constant 12 h light:12 h dark photoperiod in 600 l opaque polyethylene tanks continuously supplied with flow-through dechlorinated FW (composition in μmol l⁻¹: 590 Na⁺, 920 Cl⁻, 760 Ca²⁺, 43 K⁺, pH 7.35). Fish were fed once daily (ad libitum) with commercial trout pellets (Martin Profishent, Elmira, ON, Canada). All experimental procedures and animal care were conducted in accordance with an approved York University Animal Care Committee protocol (which conformed to the guidelines of the Canadian Council on Animal Care).

A primary cultured gill epithelium model was prepared using methodology developed by Wood and Pärt (1997) and described in detail by Kelly et al. (2000). This model is composed exclusively of gill pavement cells (PVCs). The procedure begins with an ∼6 day period of flask culture, after which cells are harvested and seeded onto semi-permeable polyethylene terephthalate filter inserts (0.9 cm² area, 0.4 μm pore, 1.6×10⁶ pores cm⁻²; BD Falcon^TM, BD Biosciences, Mississauga, ON, Canada). Companion cell culture plates (Falcon BD) were used to keep filter inserts immersed on both sides under symmetrical conditions, where both apical and basolateral compartments of the culture system contained L15 medium supplemented with 6% fetal bovine serum (FBS).

Where cultured preparations were treated with cortisol (hydrocortisone 21-hemisuccinate, Sigma-Aldrich), a single dose of 500 ng ml⁻¹ was used. Cortisol was added to the basolateral side of the experimental set up only. Cortisol was replenished with each change of medium and the dose used was selected as effective and physiologically relevant based on previous dose–response studies using this model (Kelly and Wood, 2001a) and observations of elevated cortisol levels in rainbow trout (Barton and Iwama, 1991).

Chopstick electrodes (STX-2) attached to a custom-modified EVOM epithelial voltohmmeter (World Precision Instruments, Sarasota, FL, USA) were used to monitor TER in insert-cultured epithelia. Measured TER values were background corrected by subtracting TER of a blank insert containing only media of appropriate composition. All TER measurements are expressed in Ω cm². Flux of the paracellular tracer, [³H]polyethylene glycol (MW 400 kDa, PEG-400; American Radiolabeled Chemicals, Inc., St Louis, MO, USA) was used as a measure of paracellular permeability prior to harvesting cells. Basolateral medium was loaded with PEG-400 and flux was determined according to calculations reported by Wood et al. (1998).

Western blot analysis of MR was conducted using methods previously outlined in Chasiotis et al. (2010). Briefly, insert cultured cells were rinsed with ice-cold phosphate-buffered saline (PBS) at 4°C and incubated in a lysis buffer (10 mmol l⁻¹ Tris-HCl, pH 7.5, 1 mmol l⁻¹ EDTA, 0.1 mmol l⁻¹ NaCl, 1 mmol l⁻¹ PMSF) with 1:200 protease inhibitor cocktail (Sigma-Aldrich Canada Ltd). Epithelial tissue was collected, homogenized by repeatedly passing cells in solution through a 26G syringe needle, and centrifuged at 10,000 g for 10 min at 4°C to obtain the supernatant. Protein concentration of the supernatant was determined using a Bradford assay (Sigma-Aldrich Canada Ltd). A total of 5 μg protein was then used for western blotting using a 12% SDS-PAGE gel. Wet transfer to a polyvinylidene difluoride (PVDF) membrane was performed at 100 V for 1 h. Following transfer, the membrane was incubated in 5% skimmed milk Tris-buffered saline with Tween (TBS-T: 10 mmol l⁻¹ Tris, 150 mmol l⁻¹ NaCl, 0.05% Tween-20, pH 7.4) solution for 1 h at room temperature (RT). After this, the membrane was incubated with a rabbit anti-MR custom-synthesized antibody overnight at RT with constant agitation. The custom-synthesized MR polyclonal antibody was raised in rabbit against a synthetic peptide (Ac-CKSRPGMSQGPRGEG-amide) corresponding to a 618–631 amino acid region of rainbow trout MR (GenScript, Piscataway, NJ, USA) (1:1000 dilution in TBS-T). Signal detection was performed by incubating the membrane with a horseradish peroxidase (HRP)-conjugated goat anti-rabbit secondary antibody (Bio-Rad Laboratories Canada Ltd) for 1 h at RT. Antigen reactivity was examined by incubating the membrane with Clarity™ Western ECL Blotting Substrate (Bio-Rad Laboratories Canada Ltd) for 5 min at RT. Imaging of immunoreactivity was conducted using a Chemi-Doc™ MP System (Bio-Rad Laboratories Canada Ltd). Specificity of the custom-made primary antibody was confirmed using a peptide pre-absorption technique described in detail by Bui and Kelly (2014).

Procedures used for the immunolocalization of the MR in cultured gill epithelia are described in detail in Chasiotis et al. (2010). Briefly, insert-cultured epithelia were fixed with 3% paraformaldehyde, permeabilized with ice-cold methanol for 5 min at −20°C, rinsed with Triton X-100 in PBS and incubated with antibody-dilution buffer (ADB; 10% goat serum, 3% bovine serum albumin and 0.05% Triton X-100 in PBS) for 1 h at RT. Preparations were then incubated with a custom-made rabbit anti-MR antibody (as detailed above) overnight at RT (1:100 in ADB). After washing in PBS, epithelia were then incubated with a TRITC-labelled goat anti-rabbit antibody (1:500 in ADB; Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA) for 1 h at RT. After incubation with antibodies, preparations were mounted on slides using mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI) for nuclear visualization. Images of MR immunoreactivity and DAPI fluorescence were captured using differential interference contrast (DIC) and laser-scanning confocal microscopy with an Olympus BX51 microscope in conjunction with a Fluoview unit and Melles Griot green and red argon lasers (Olympus Canada). As a negative control, epithelia were prepared as previously described in conjunction with the peptide pre-absorption procedure outlined in Bui and Kelly (2014).

For quantification of MR protein abundance, western blot analysis was used to image MR immunoreactivity as described above, after which membranes were incubated in stripping buffer, blocked with 5% skimmed milk solution as outlined above and incubated with mouse monoclonal anti-actin JLA20 antibody (Developmental Studies Hybridoma Bank, Iowa City, IA, USA). Actin immunoreactivity was imaged and used for normalization of MR abundance after validating, by statistical comparison, that actin protein abundance between experimental groups did not significantly change in response to experimental treatment. MR and actin protein abundance were quantified using ImageJ software (National Institutes of Health, Java 1.6.0_45, 64-bit).

Custom-made double-stranded RNA (dsRNA) spanning nucleotides 590 and 1598 of the mr transcript (GenBank accession no. NM_001124483.1) was prepared using a T7 RiboMAX™ Express RNAi System (Promega, Madison, WI, USA) following the manufacturer’s instructions. dsRNA concentration was determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Nepean, ON, Canada) and 15 pmol was used for transcriptional knockdown of MR in cultured epithelia combined with Lipofectamine 3000^® transfection reagent (Invitrogen Canada Inc.). Previous studies have reported a dsRNA dose of 15 pmol is effective for knockdown in epithelia transfected at 24 h post-seeding (hps) (Kolosov et al., 2017; Kolosov and Kelly, 2017). Therefore, control and cortisol-stimulated epithelia were treated with dsRNA at 24 hps, i.e. mr-KD followed 24 h exposure to cortisol in the cultured epithelia that experienced both treatments.

For analysis of the presence of mr transcript in transfected tissues, total RNA was isolated and cDNA synthesized as previously described (Kolosov and Kelly, 2018). The presence of mr transcript was then determined by reverse transcription PCR (RT-PCR). In order to detect transcriptional knockdown of mr, a separate set of primers for rainbow trout mr was designed using the coding sequence (forward: 5′-ATTGCTGCTCATCTGGAACC-3′ and reverse: 5′-CGTTGTTGTTGTTCTCTTGG-3′, predicted amplicon size ∼1100 bp). PCR conditions were as follows: denaturation at 95°C, 4 min, followed by 40 cycles of: (1) denaturation (95°C, 30 s), (2) annealing (56°C, 30 s) and (3) extension (72°C, 30 s). The RT-PCR product was examined using agarose gel electrophoresis and sequenced to confirm the identity of the amplicon.

The transcript abundance of gr1 (GenBank accession no. NM_001124730), gr2 (NM_001124482) and 11β-hsd (NM_001124218.1) was examined in the cultured gill epithelia following RNA extraction and cDNA synthesis using techniques outlined in previous studies. The transcript abundance of cortisol-responsive cldn isoforms cldn-3a (BK007964), cldn-6 (KF445436), cldn-7 (BK007965), cldn-8d (BK007966.1), cldn-12 (BK007967), cldn-20a (BK009404), cldn-23a (BK008775), cldn-28b (EU921670), cldn-30 (BK007968) and cldn-31 (BK007969) was examined as outlined above. Primer sets have been previously published by Kelly and Chasiotis (2011) (gr1, gr2, cldn-3a, cldn-7, cldn-12, cldn-30 and cldn-31), Chen et al. (2015) (11β-hsd), Kolosov et al. (2014) (cldn-6 and cldn-23a), and Kolosov and Kelly (2017) (cldn-20a). The selection of cldn isoforms as cortisol responsive was based on observations of either Kelly and Chasiotis (2011) or Gauberg et al. (2017). Transcript abundance of ocln (GQ476574) and tric (KC603902) was also examined, using primers previously published by Kelly and Chasiotis (2011) and Kolosov and Kelly (2013), respectively. Both ocln and tric were first shown to be cortisol responsive in the fish gill epithelium in the aforementioned studies. Rainbow trout elongation factor 1-alpha (ef-1a; GenBank accession no. AF498320.1; primers in Kolosov et al., 2017) was used as a reference gene. The use of ef-1a as a reference for gene of interest normalization was validated by statistically comparing ef-1a threshold cycles between treatments to confirm that experimental conditions had not significantly altered the values.

Using western blotting, an examination of Cldn-8d, Cldn-28b and Ocln protein abundance was performed (as described above). A custom-synthesized rabbit anti-Ocln polyclonal antibody was used for western blot in accordance with methods previously outlined by Chasiotis et al. (2010). A custom-synthesized rabbit anti-Cldn-8d polyclonal antibody was used for western blot according to Kolosov and Kelly (2017). A custom-synthesized Cldn-28b polyclonal antibody was raised in rabbit against a synthetic peptide (Ac-CPPKDENHNVKYSK-amide) corresponding to the 185–198 amino acid region of rainbow trout Cldn-28b (EU921670) (GenScript, Piscataway, NJ, USA).

Statistical analysis was conducted using Student's t-test, or a two-way ANOVA protocol coupled with a multiple comparison test where appropriate, in SigmaPlot (v11) statistical software. SigmaPlot tested for normality and equal variance before proceeding with ANOVA analysis. Statistical significance was based on the observation of a fiduciary limit of P<0.05. All P-values generated by statistical analysis are presented in Table S1.

A custom-synthesized rainbow trout MR antibody detected a protein band corresponding to the theoretical molecular weight of trout MR, ∼110 kDa, and when a peptide pre-absorption procedure was followed, immunoreactivity was not detected (Fig. 1A). MR immunoreactivity was localized to the cytosol of cells in the cultured trout gill epithelium, and following peptide pre-absorption, MR immunoreactivity could not be observed (Fig. 1B).

Transcriptional knockdown of mr in a cultured trout gill epithelium eliminated detectable transcript at 24, 48 and 72 h post-transfection (Fig. 2A) and reduced relative protein abundance of MR to 29±2% (Fig. 2B; t-test P<0.0001).

When the control epithelium was cultured in the presence of 500 ng ml⁻¹ cortisol, an increase in TER and a decrease in paracellular PEG-400 flux was observed (Fig. 3A,B; two-way ANOVA, P<0.001 control versus cortisol). Concurrently, MR protein abundance decreased in cortisol-treated preparations (Fig. 3C, two-way ANOVA, P=0.017 control versus cortisol). In contrast, mr-KD epithelia exhibited a decrease in TER and an increase in PEG-400 permeability versus control epithelia (i.e. in the absence of cortisol) (two-way ANOVA, P<0.001 control versus cortisol). mr-KD resulted in a significant elevation in TER and decrease in PEG-400 permeability upon cortisol stimulation (two-way ANOVA interaction term, P=0.022 for TER and P<0.001 for PEG-400; see Table S1). A statistically significant interaction between cortisol and mr-KD was found with respect to MR protein abundance (see Table S1; two-way ANOVA interaction term, P=0.028) and MR protein abundance remained low in cortisol-treated mr-KD preparations.

Transcript abundance of gr1 remained unaffected by cortisol, mr-KD or a combination of the two treatments (Fig. 4A). However, mr-KD significantly altered the transcript abundance of gr2 (Fig. 4B; P=0.017) and 11β-hsd (Fig. 4C; P<0.001), where an elevation in both could be observed. Although cortisol treatment alone, and the interactive effect of cortisol and mr-KD did not significantly alter gr2 transcript abundance (Fig. 4B), transcript abundance of 11β-hsd was further elevated by cortisol treatment (Fig. 4C; P<0.001) and statistical analysis indicated a significant interaction between the two experimental variables (see Table S1; two-way ANOVA interaction term, P=0.002).

Transcript abundance of cldn-3a, -6, -7, -8d, -12, -20a, -23a, -28b, -30 and -31, as well as ocln significantly increased in response to cortisol treatment, while tric mRNA abundance decreased (Fig. 5; for P-values, see Table S1). mr-KD influenced the mRNA and protein abundance of TJ proteins in a cultured gill epithelium in several ways. First, transcript abundance of cldn-6, -8d, -23a and -28b decreased, and cldn-20a mRNA abundance increased in response to mr-KD alone (i.e. in the absence of cortisol) (Fig. 5; for P-values, see Table S1). Second, when mr-KD epithelia were stimulated with cortisol, transcript abundance of ocln, cldn-8d and cldn-23a was elevated to a lesser degree (Fig. 5; two-way ANOVA interaction term, P<0.001 for ocln, P=0.004 for cldn-8d and P<0.001 for cldn-23a). Finally, the transcriptional responses of cldn-20a, -28b and -30 were found to exhibit statistically significant interactive effects of the two experimental variables (for P-values, see Table S1).

Protein abundance of Cldn-8d (Fig. 6A; P=0.006), Cldn-28b (Fig. 6B; P=0.015) and Ocln (Fig. 6C; P=0.008) increased in cortisol-treated epithelia, which mirrored changes in mRNA abundance and reflected their cortisol sensitivity. Additionally, mr-KD was found to reduce Cldn-8d protein abundance (Fig. 6A; P=0.030), but an interactive effect of cortisol and mr-KD on Cldn-8d was not found to be significant (see Table S1; P=0.447). Cldn-28b protein abundance was not significantly altered by mr-KD itself (Fig. 6B; P=0.317), but a significant interaction between cortisol and mr-KD was observed (see Table S1; P=0.016), and cortisol treatment failed to elevate Cldn-28b protein levels in mr-KD epithelia (Fig. 6B). In contrast, Ocln protein abundance increased following mr-KD (Fig. 6C, P=0.010 control versus mr-KD) and although no further alterations in Ocln abundance were observed in cortisol-treated mr-KD preparations, statistical analysis indicated a significant interaction between the two factors (see Table S1; P=0.050).

Cortisol, a GR ligand and an identified MR ligand in teleost fish circulation, has been broadly documented to enhance the barrier properties of externally exposed teleost fish epithelia, such as the gill epithelium and epidermis (Kelly and Wood, 2001a,b, 2002; Zhou et al., 2003; Kelly and Wood, 2008; Sandbichler et al., 2011a,b; Chasiotis and Kelly, 2012; Kwong and Perry, 2013; Gauberg et al., 2017). This takes place, in part, through a reduction in the permeability of the paracellular pathway and in a natural setting would help to mitigate passive solute movement and assist in the maintenance of homeostasis. The epithelial tightening brought about by cortisol in fish epithelia has more recently been linked to alterations in molecular components of the epithelial TJ complex (Bui et al., 2010; Chasiotis et al., 2010; Chasiotis and Kelly, 2011; Kelly and Chasiotis, 2011; Sandbichler et al., 2011a; Kolosov and Kelly, 2013; Kwong and Perry, 2013; Gauberg et al., 2017; Kolosov et al., 2017; Kolosov and Kelly, 2017, 2018), and at least one study using a model gill epithelium preparation has implicated both the GR and MR pathways in the modulation of gill epithelium TJ complex physiology (Kelly and Chasiotis, 2011). In this regard, it has been suggested that select TJ proteins may be responsive to cortisol exclusively via the MR and because the MR appears to have evolved in the teleost lineage prior to the evolution of a separate and dedicated MR ligand as well as aldosterone synthase (Arterbery et al., 2011), it is of special interest to examine what role the MR plays in relaying the physiological response of these organisms to cortisol. Using a model FW rainbow trout gill preparation with depleted MR levels, the current study supports the idea that the MR is important in cortisol-induced epithelial tightening by revealing that in the absence of added cortisol, the MR is needed for the establishment of full electroresistive properties in this model (i.e. mr-KD compromised the integrity of gill epithelium barrier properties). Furthermore, mr-KD alone significantly reduced the transcript and/or protein abundance of several Cldn TJ proteins. In addition to this, the exposure of mr-KD preparations to cortisol enhanced the barrier properties of the model preparation, which underscores the idea that the MR is required for normal barrier properties to develop in this model. Finally, mr-KD revealed that several cortisol-sensitive TJ proteins do not appear to respond to cortisol via the MR but that cortisol-induced alterations in mRNA abundance of cldn-8d, cldn-23a, cldn-28b, ocln and tric were perturbed by mr-KD. Of particular interest, and in support of the idea that select TJ proteins may be responsive to cortisol via the MR exclusively, was the observation that mr-KD diminished the ability of cortisol to increase Cldn-28b protein abundance.

An important observation of the current study is that in this model, cortisol-mediated changes in permeability rely on regulatory networks of TJ proteins that are at least in part attuned to the abundance of the MR. Part of the regulatory network of the molecular machinery of the fish gill TJ complex probably utilizes the MR in the presence of basal levels of cortisol to develop resistive properties. Therefore, it seems that the presence of both GR and MR in the teleost gill epithelium has functional significance even in the absence of a dedicated mineralocorticoid ligand.

The finding that the MR is needed for the establishment of full-fledged electroresistive properties of the cultured trout gill epithelia is interesting in the light of previous findings regarding the effects of mineralocorticoid ligands on fish CRs. A previous study reported that cortisol (but not aldosterone or deoxycorticosterone) was the only MR ligand capable of producing an increase in the barrier properties of the cultured trout gill epithelium (Kelly and Chasiotis, 2011). In contrast, Arterbery et al. (2011) have shown that aldosterone and deoxycorticosterone can bind MR and GR constructs of plainfin midshipman and daffodil cichlid in HeLa cells with various affinities. There appear to be species-specific differences in the response of CRs to mineralocorticoid ligands in teleosts. As: (i) neither aldosterone, nor deoxycorticosterone is found to circulate at a significant level in adult teleosts, and (ii) neither of these ligands has any effect on the barrier properties of the rainbow trout gill epithelium, we conclude that barrier properties in the rainbow trout gill are regulated by cortisol and the MR is an important part of the response to cortisol.

A cultured FW trout gill epithelium composed exclusively of PVCs develops its resistive properties in media supplemented with 6% FBS over ∼14 days, and the mRNA abundance of many TJ proteins, including those sensitive to cortisol, increases (Kolosov et al., 2014). FBS contains basal levels of cortisol (Milo et al., 1976): 6% supplementation results in estimated final concentration in media of ∼0.1–1 ng ml⁻¹. In addition to cortisol, mechanisms coordinating TJ assembly during epithelial development have been described to rely on a multitude of cytokines, transcription factors and post-translational regulation (Khan and Asif, 2015; Boivin and Schmidt-Ott, 2017; Na et al., 2017). The basal level of cortisol in conjunction with the presence of cytokines and transcription factors in serum-supplemented media probably plays a role in up-regulating transcription of TJ proteins. Interestingly, dimerization after nuclear shuttling and in vitro binding to DNA by the MR has been shown to not depend on the presence of the ligand (Grossmann et al., 2012). Specifically, the binding of the MR depends on the presence of HSP90 and was inhibited by the HSP90 inhibitor geldanamycin (Grossmann et al., 2012). Additionally, the GR and MR are reported to heterodimerize in some systems, which elicits a composite transcriptional response distinct from that of GR and MR homodimers (Trapp et al., 1994; Kiilerich et al., 2015). In the gill epithelium model, mr-KD may have interfered with the steady-state equilibrium of MR monomers, MR homodimers, MR/GR heterodimers, and DNA-bound cortisol-stimulated MR, changing the transcription of cortisol-responsive genes in developing epithelia supplemented with FBS. This probably resulted in a decrease in the constitutive transcription of MR-sensitive TJ proteins.

Concurrently with the above-described changes, the molecular composition of the TJ complex in mr-KD epithelia was altered. More specifically, a reduction in the mRNA abundance of cldn-6, -23a and -28b, as well as a reduction in the mRNA and protein abundance of Cldn-8d was observed. In addition, cldn-20a mRNA abundance and Ocln protein abundance significantly increased. All of these TJ proteins have been demonstrated to be, or characterized as, putative barrier-forming TJ proteins (Chasiotis et al., 2012; Kolosov et al., 2014; Kolosov and Kelly, 2017). Therefore, reduced abundance of putative and known barrier-forming gill cldns/Cldns in mr-KD preparations would be consistent with a modest but significant compromise of gill epithelium barrier properties in the current study. Furthermore, the transcript abundance of cldn-20a has previously been shown to increase in response to a knockdown-mediated reduction in Cldn-8d levels in the gill epithelium (Kolosov and Kelly, 2017). The protein (but not mRNA) abundance of Ocln-a has been shown to increase in the skin of larval zebrafish specifically in response to cortisol-GR stimulation (Kwong and Perry, 2013). Therefore, an upregulation in cldn-20a and Ocln may be a compensatory response triggered by increased paracellular permeability of mr-KD epithelia experiencing reduced levels of mr, and cldn-6, -8d, -23a and -28b, resulting in modest rather than dramatic alterations in epithelial integrity.

In a situation where one ligand mediates both a mineralocorticoid and a glucocorticoid response within a tissue, much of the effect of this ligand is determined by the mosaic of receptors and ligand-metabolizing enzymes at the tissue level (Savory et al., 2001). Therefore, the availability and abundance of a certain receptor can shape the response of the containing tissue to the ligand in question (e.g. Kiilerich et al., 2015). For example, it's been shown that the GR, but not the MR, mediates cortisol effects on the development of epidermal ionocyte in larval zebrafish (Cruz et al., 2013). Similarly, regional differences in the TJ response to cortisol in rainbow trout skin were suggested to be mediated by differences in receptor and enzyme abundance following cortisol loading (Gauberg et al., 2017).

GR protein levels are known to autoregulate gr mRNA, increasing transcript abundance when GR protein is occupied, and available receptor levels decrease in fish tissues (Sathiyaa and Vijayan, 2003; Vijayan et al., 2003). Additionally, mr-knockout (KO) in medaka has been reported to increase GR levels to compensate for the absence of the MR (Sakamoto et al., 2017). Similarly, it seems from the current study that if the levels of MR are altered in the absence of added cortisol, the model gill epithelium of rainbow trout can increase the abundance of gr2, possibly to compensate by enabling a full-fledged response to cortisol once the ligand is present. The model gill epithelium may also have employed a cortisol clearing mechanism involving 11β-HSD to prevent overstimulation of the remaining MR. 11β-HSD has been proposed as a mechanism for protection of the MR from cortisol activation in the gills of smolting Atlantic salmon (Kiilerich et al., 2007). Since then, an upregulation of 11β-hsd has similarly been reported in the gill and skin of cortisol-loaded rainbow trout (Chen et al., 2015, 2016; Gauberg et al., 2017). Interestingly, our findings seem consistent with the role of 11β-hsd in mammalian CR signalling, where it protects the MR from overstimulation with cortisol (Ferrari et al., 2000), allowing aldosterone to act on the MR exclusively (Edwards et al., 1988; Funder et al., 1988). However, the role of 11β-hsd in teleost CR signalling is probably confined to ensuring MR specificity of the CR response, rather than MR protection as seen in mammals.

Previous studies have suggested that endogenous mechanisms may prevent certain steroids from binding to the MR in teleosts in a manner similar to MR protection in tetrapods (Edwards et al., 1988; Funder et al., 1988; Arterbery et al., 2011). In the cultured gill epithelium, alterations in the mRNA abundance of 11β-hsd and, to a lesser extent, gr2 may be a part of this MR-specificity response. This could mean that the role of 11β-HSD in facilitating the specificity of the MR response evolved prior to the evolution of the second ligand and was important for the dichotomy of the MR–GR signalling pathways in response to one ligand, cortisol. These mechanisms may be important in organs that constantly have to differentiate between stress and osmoregulatory cortisol and may be a fine-tuning mechanism for the corticosteroid regulation of TJs in fish epithelia.

Despite the fact that mr-KD model gill epithelia exhibited compromised barrier properties, when stimulated with cortisol, mr-KD epithelia mounted a response that exceeded that of cortisol-stimulated epithelia with unaltered CR levels. This is evident from the TER and flux rates of PEG-400, which were increased and decreased, respectively. An upregulation of gr2 coupled with an increase in 11β-hsd abundance in mr-KD epithelia may have resulted in an enhanced stimulation of the GR pathway in mr-KD epithelia treated with cortisol. This observation is in line with previously reported higher resistive properties of cultured gill epithelia stimulated with a GR-exclusive agonist such as dexamethasone (Kelly and Wood, 2002; Chen et al., 2015). Interestingly, a mechanism that could explain the increased activity of the GR in mr-KD epithelia was presented in a recent study by Kiilerich et al. (2015), who demonstrated that transcriptional activity of CRs decreases when both MR and GR are transfected into the same COS-7 cell (Kiilerich et al., 2015). This suggests that the interaction between MRs and GRs regulates a response to cortisol by cells that express both receptors. As MR levels were reduced and GR levels increased in mr-KD epithelia in the current study, it is plausible that GR activity stimulated by cortisol was enhanced compared with the control condition where MR levels were higher, which in turn contributed to increased transcriptional activation of barrier-building machinery with cortisol stimulation.

Alterations in mRNA abundance, evident in the cortisol-stimulated epithelia in the current study, have been reported previously for every TJ protein examined in this suite of experiments: cldn-3a, -7, -8d, -12, -28b, -30 and -31 (Kelly and Chasiotis, 2011; Sandbichler et al., 2011b); cldn-6 (Bui et al., 2010); cldn-20a and -23a (Kolosov and Kelly, 2017); ocln (Chasiotis et al., 2010); and tric (Kolosov and Kelly, 2013). However, there were several notable differences in the response of mr-KD epithelia to cortisol stimulation when compared with that of epithelia with unperturbed CR levels. The first of these was that cortisol-treated mr-KD epithelia exhibited modest alterations in cldn-8d/Cldn-8d, ocln and tric abundance versus cortisol-treated control preparations. This suggests that transcriptional regulation of these TJ proteins is connected to the MR pathway, which is supported further by the fact that the abundance of cldn-8d also decreased in response to mr-KD itself. These findings corroborate previous studies that have described cldn-8d and ocln to be sensitive to cortisol stimulation via both GR and MR pathways (Kelly and Chasiotis, 2011). However, observations of cldn-23a and tric are novel and significant. With respect to cldn-23a, no previous study has considered which CR modulates its transcript abundance following exposure to cortisol. So, this study provides the first evidence that the MR is a significant contributor to adjustments in cldn-23a abundance following cortisol exposure. Following this, the novelty of tric observation in mr-KD preparations relates to the fact that the molecular components of the tricellular tight junction (tTJ) complex in teleost fishes have only recently been credited with relaying part of the cortisol-mediated epithelial tightening response (Kolosov and Kelly, 2013, 2018) and, as for cldn-23a, nothing is known regarding which receptor is used to relay the response to cortisol.

A second notable difference in the response of mr-KD epithelia to cortisol stimulation versus control cortisol-treated epithelia was that no significant changes in the abundance of cldn-20a and Ocln were observed. However, the abundance of these molecular components had already increased in response to mr-KD alone, probably in response to depleted Cldn-8d levels and restructuring of the CR pathways (see above). This suggests that mechanisms involved in the putative compensatory response of cldn-20a and Ocln to mr-KD can override the cortisol sensitivity of these TJ components.

Lastly, and of particular importance, was that the protein abundance of Cldn-28b was not elevated in cortisol-treated mr-KD epithelia, indicating that, as previously suggested based on CR agonist/antagonist studies, the abundance of this TJ protein is probably controlled via the MR pathway (Kelly and Chasiotis, 2011). However, cldn-28b mRNA abundance was significantly elevated in cortisol-stimulated mr-KD epithelia (perhaps via remaining MRs in mr-KD epithelia). However, we suggest that this may have happened as a compensatory response to depleted Cldn-28b protein levels via a separate regulatory pathway independent of the MR, as it did not result in increased protein abundance.

Interestingly, previous studies have reported dose-dependent quantitative differences as a result of which pathway (GR or MR) is stimulated by cortisol. For example, epithelia stimulated with the GR agonist dexamethasone typically exhibited higher mRNA levels of cldns compared with epithelia stimulated with an agonist that was proposed to relay through the MR pathway (Kelly and Chasiotis, 2011). Additionally, the effects of cortisol on the transcription of some cldns is known to be dose dependent (Bui et al., 2010; Chasiotis and Kelly, 2011). Thus, if 11β-HSD (which exhibited further elevated mRNA abundance in mr-KD epithelia treated with cortisol) contributed to cortisol clearance, it is possible that it contributed to the reduced ability of cortisol to increase the mRNA abundance of select cldns.

Previous studies have reported that cortisol-induced tightening of the gill epithelium is relayed though both the GR and MR, but these studies were conducted using pharmacological agents that arguably lack specificity or may have off-target effects in teleost systems (Kelly and Chasiotis, 2011). The current study took a more direct approach to examining the MR contribution to the tightening effect via knockdown. A compensatory response to mr-KD appeared to take place in the absence of added ligand. This may indicate the presence of a regulatory network responsible for CR abundance in this peripheral tissue that ensures the sufficiency of the response to the ligand once it is present. This may be an example of an early-divergent receptor pair function that adapted to a single ligand with multiple effectors in peripheral tissue. However, previous studies, employing heterologous MR-specific ligands (e.g. aldosterone) and loss-of-function, have suggested that the MR is not likely to play a role in relaying cortisol-mediated changes in the fish skin and gills (Cruz et al., 2013; Kwong and Perry, 2013; Sakamoto et al., 2016). Sakamoto et al. (2016) have demonstrated, specifically, that osmoregulation was normal in mr-KO medaka, suggesting that peripheral osmoregulatory tissues like the gill epithelium are able to respond to cortisol in a manner that enables normal gill function (Sakamoto et al., 2016). Our study corroborates this finding to some extent by showing that mr-KD model gill epithelium preparations can exhibit an enhanced cortisol-mediated reduction in paracellular permeability. However, the current study also suggests that the MR may be necessary for the development of normal barrier function of the rainbow trout gill epithelium. This is because in control preparations (i.e. mr-KD without added cortisol), mr-KD unequivocally compromises resistance and paracellular marker flux in the model trout gill and when mr-KD preparations are treated with cortisol, the barrier properties of the model are enhanced. Moreover, the MR pathway seems to be responsible for mediating some of the cortisol-induced epithelial tightening to specific TJ proteins (in the case of Cldn-28b, seemingly, exclusively so). The MR has been reported to be present in many epithelial/endothelial tissues of teleosts, including rainbow trout (Greenwood et al., 2003; Sturm et al., 2005). These tissues would be responsible for maintaining blood–brain, blood–urine and blood–water barriers in fishes to name just some. Although at least one teleost species appears to survive in the absence of an MR (Sakamoto et al., 2016), the MR-KD condition in rodents (that have a separate MR ligand, aldosterone) is lethal (Gass et al., 2001). In the current study, mr-KD epithelia were able to mount an enhanced cortisol-induced tightening response, possibly by mounting an MR specificity response and overstimulating the GR pathway in light of reduced MR levels (see Kiilerich et al., 2015). Similarly, upregulation of GR mRNA abundance has been described to take place in mr-KO medaka organs/tissues (Sakamoto et al., 2017). Interestingly, Sakamoto et al. (2016) have also reported a pathological behavioural phenotype and connected the mr-KO phenotype with perturbed brain function. TJ protein expression profiles in teleosts (and other vertebrates) exhibit organ/tissue and isoform specificity, making TJ assembly heterogeneous in different barriers (reviewed in Kolosov et al., 2013). This may preclude epithelia (or endothelia), other than the gill epithelium examined in our study, from successfully rearranging TJ proteins in response to cortisol if the MR pathway is compromised. It may be possible that the blood–retina barrier of the choroid plexus epithelium adjacent to the eye of medaka (where most of the MR immunoreactivity is localized in healthy medaka) was compromised in mr-KO (Sakamoto et al., 2016). Therefore, the importance (or redundancy) of the MR pathway for cortisol-induced epithelial ‘tightening’ could be region specific. For example, a recent study has found that effects of increased circulating levels of cortisol are relayed in the skin TJ complex of rainbow trout in a region-dependent manner (Gauberg et al., 2017). Therefore, further studies are warranted that aim to investigate how TJ proteins of different epithelia/endothelia and even different regions of the same epithelium respond to cortisol in teleosts.

The authors would like to thank Drs Barry Loughton, Andrew Donini and Jean-Paul Paluzzi (York University) for insightful feedback on this study.