Defensive slime formation in Pacific hagfish requires Ca²⁺- and aquaporin-mediated swelling of released mucin vesicles

Hagfishes (Myxinidae) are benthic, marine craniates that are notable for their unique defense mechanism of rapidly releasing large amounts of slime exudate when threatened (Janvier, 1996; Spitzer and Koch, 1998). The exudate combines with seawater to form slime, an effective defense mechanism because of its ability to lodge on gills and impair respiration (Lim et al., 2006; Zintzen et al., 2011). The exudate consists of two principal components, long (~up to 15 cm), thin (~1–3 μm wide) protein threads and mucin-like glycoproteins (Luchtel et al., 1991; Salo et al., 1983). Recent findings have demonstrated that during mixing of slime exudate with seawater, the mucins elongate into strands and assist in the unravelling of condensed slime threads (Winegard and Fudge, 2010). Mucins also likely contribute to the slime's high hydrodynamic resistance (Winegard and Fudge, 2010).

The slime glands are located in pairs along the entire ventro-lateral length of the animal (Fig. 1) and contain both gland thread cells (GTCs) and gland mucous cells (GMCs) (Spitzer and Koch, 1998). Holocrine secretion of mucins occurs when the gland is stimulated to expel its contents, and both mature GTCs and GMCs are forced through the narrow gland duct, causing the plasma membrane to be stripped from each cell that passes through it (Spitzer and Koch, 1998). In contrast to exocytotic release, in which mucin vesicles fuse with the apical plasma membrane of a mucous cell, holocrine release from hagfish slime glands results in the rapid and mass release of numerous mucin vesicles from many GMCs (Deyrup-Olsen and Luchtel, 1998). When the vesicles come into contact with seawater, they rapidly burst and the mucins interact with seawater and threads to produce slime (Luchtel et al., 1991).

Hagfish mucin vesicles are disc-shaped structures that are ~7 μm along the major axis, and are encased by a lipid bilayer (Luchtel et al., 1991). Vesicles can be stabilized in concentrated (1 mol l⁻¹ or higher) solutions containing polyvalent anions such as citrate, sulfate and phosphate, but appear to rupture in solutions containing monovalent anions, regardless of the valency of the associated cation (Luchtel et al., 1991; Salo et al., 1983). Based on these observations, Luchtel et al. (Luchtel et al., 1991) hypothesized that ions from seawater enter the vesicle down their concentration gradients, resulting in an influx of water and the swelling and rupture of the vesicle.

In contrast to mucin granules released by exocytosis (Verdugo, 1991), hagfish slime mucins are released with intact membranes. Given the rapidity with which the slime can be deployed (~100 ms) (Lim et al., 2006), we hypothesized that the movement of ions and water molecules across the vesicle membrane may be accelerated by the presence of ion transporters and aquaporins (AQPs), the latter of which can increase the rate of water permeation by 10- to 100-fold (Agre et al., 2002).

The present study aimed to elucidate the mechanisms involved in Pacific hagfish (Eptatretus stoutii) slime mucin vesicle rupture in seawater. Our data demonstrate that the vesicle membrane plays an important role in regulating vesicle swelling, and suggest that the movement of ions and water molecules is facilitated by at least two kinds of membrane transporters, at least of one of which is Ca²⁺ activated.

In order to determine whether mucin vesicle swelling is dependent on the vesicle membrane, we examined the effects of the membrane-disrupting detergent Triton X-100. Vesicles exposed to the simplified seawater (SSW; 10 mmol l⁻¹ CaCl₂ + 545 mmol l⁻¹ NaCl) control exhibited substantial variability in both the rate and onset of swelling, with an apparent bimodal distribution of some vesicles swelling relatively slowly and early, and others swelling rapidly after an initial delay (Fig. 2A,C). Exposure to SSW containing 0.1%

Triton-X 100 caused a striking reduction in the variability of swelling behaviour (Fig. 2B,D; supplementary material Movie 1). The mean swelling rate for vesicles in SSW was 2.0±0.4% increase in area per millisecond, and for detergent-treated vesicles it was substantially slower (0.20±0.03% area ms⁻¹).

We exposed vesicles to an array of solutions that differed in the presence or absence of three abundant ions in natural seawater: Na⁺, Cl⁻ or Ca²⁺. Exposure to artificial seawater [ASW; following a recipe in Bidwell and Spotte (Bidwell and Spotte, 1985)] resulted in the rupture of 99.1±0.5% of vesicles (Fig. 3A) within 120 s (supplementary material Movie 2). All solutions containing 10 mmol l⁻¹ Ca²⁺ or higher (SSW, 10 mmol l⁻¹ Ca²⁺ gluconate + 980 mmol l⁻¹ sucrose, and 333 mmol l⁻¹ CaCl₂) also caused rupture in >98% of vesicles. This was not significantly different from ASW, indicating that the effect of Ca²⁺ was not dependent on the presence of either Na⁺ or Cl⁻. In the Ca²⁺-free solution containing only 545 mmol l⁻¹ NaCl, significantly fewer vesicles ruptured (39.7±2.2%) than in Ca²⁺-containing controls. Other Ca²⁺-free solutions also resulted in less rupture compared with the ASW control. However, 545 mmol l⁻¹ NMDG-Cl (N-methyl-d-glucamine chloride) and ASW + 15 mmol l⁻¹ EGTA [ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid] both caused a higher percentage of vesicles to rupture than 545 mmol l⁻¹ NaCl, but these were still both significantly lower than the ASW control.

To test the effects of mono- and divalent cations and anions (Na⁺, Ca²⁺, Cl⁻ and SO₄²⁻), mucin vesicles were exposed to a salt-free solution of 1 mol l⁻¹ betaine as an osmotic control, and four betaine solutions with 10 mmol l⁻¹ CaCl₂, CaSO₄, NaCl or Na₂SO₄. The two solutions containing Ca²⁺ caused >98% of vesicles to rupture, regardless of the valency of the anion present in solution (Fig. 3B), while the Ca²⁺-free solutions caused no more rupture than the betaine control.

We exposed vesicles to a series of concentrations of CaCl₂ in NaCl to determine the threshold at which Ca²⁺ induces rupture. We also tested whether another divalent cation (Mg²⁺) could substitute for Ca²⁺. Concentrations of Ca²⁺ >3 mmol l⁻¹ resulted in significantly more rupture than NaCl only (Fig. 3C). Mg²⁺, the other prevalent divalent cation in seawater, only increased rupture above NaCl levels at concentrations higher than 100 mmol l⁻¹ Mg²⁺. Even 415 mmol l⁻¹ MgCl₂ only caused 69±5% rupture, which was significantly less than the effect of Ca²⁺ concentrations two orders of magnitude lower.

Hagfish slime mucin vesicles were treated with three ion channel inhibitors to test whether Ca²⁺-induced rupture could be inhibited. Vesicles were treated with 500 μmol l⁻¹, 0.1% DMSO (dimethyl sulfoxide) or no vehicle prior to exposure to ASW, SSW, 545 mmol l⁻¹ NaCl, or 10 mmol l⁻¹ Ca²⁺ gluconate + 980 mmol l⁻¹ sucrose. Only 37±10% of vesicles ruptured when exposed to Ca²⁺-free NaCl as previously demonstrated. In addition, there was no significant effect of either the DMSO vehicle control or DIDS (4,4′-diisothiocyanatostilbene-2,2′-disulfonic acid disodium salt hydrate) on vesicle rupture in the absence of Ca²⁺ (Fig. 4A). However, the three Ca²⁺-containing solutions (ASW, SSW and Ca²⁺ gluconate + sucrose) showed >97% rupture, which was reduced to <27% by the presence of DIDS.

The inhibitory effect of DIDS did not depend on the presence of Cl⁻, as vesicles exposed to 1 mol l⁻¹ betaine or 10 mmol l⁻¹ CaSO₄ + 980 mmol l⁻¹ betaine had a similar response, with DIDS exposure resulting in significantly less rupture in the presence of Ca²⁺ than the DMSO control. However, DIDS also caused a small but significant (P=0.032) reduction in rupture in betaine alone (Fig. 4B). Similarly, treatment of vesicles with 20 μmol l⁻¹ tannic acid (Fig. 4C) or 20 μmol l⁻¹ PGG (penta-O-galloyl-β-d-glucose hydrate; Fig. 4D) resulted in significantly less rupture in solutions containing Ca²⁺ compared with controls. Tannic acid also caused a decrease in rupture in 545 mmol l⁻¹ NaCl but PGG did not.

We tested for the presence of AQP by exposing vesicles to four solutions (ASW, SSW, 545 mmol l⁻¹ NaCl and 5 mmol l⁻¹ Tris) with and without pre-treatment with 1 mmol l⁻¹ HgCl_2, a known AQP inhibitor. HgCl₂ pre-treatment and solution type both had significant effects on swelling rate, and the interactive effect was also significant (Fig. 5A). Under control conditions, swelling was significantly faster in solutions containing Ca²⁺ (ASW 1.95±0.22% area ms⁻¹ and SSW 2.03±0.40% area ms⁻¹) compared with Ca²⁺-free 545 mmol l⁻¹ NaCl (0.947±0.193% area ms⁻¹). Vesicles treated with ASW under control conditions exhibited both fast and slow swelling (Fig. 5B), but very few vesicles treated with HgCl₂ demonstrated fast swelling (Fig. 5C; supplementary material Movie 3). With HgCl₂, there was no significant effect of solution on swelling rate. The overall effect of HgCl₂ treatment was an approximate 10-fold reduction in swelling rate (mean <0.21% area ms⁻¹).

In order to determine whether AQPs facilitate water movement across the vesicle membrane during vesicle deployment in seawater, we searched for evidence of AQP homologues in the hagfish slime gland. A partial and a full-length sequence for two putative hagfish AQP homologues were identified following searches of two hagfish Illumina transcriptomes obtained from a combined gill/slime gland or kidney preparation. A full-length sequence for the first homologue was obtained by homologous cloning using the Eptatretus burgeri AQP4 (AB258403.1) as a template. Maximum likelihood phylogenetic analysis demonstrated that one sequence was a member of the aquaglyceroporin subfamily that shares a common ancestor with mammalian AQP3 (Fig. 6A). The other identified sequence grouped most closely with E. burgeri AQP4 and shares a common ancestor with other members identified as AQP4 (Fig. 6B). Full-length sequences can be found in the NCBI database for EsAQP3 (KF895385) and EsAQP4 (KF895386).

RT-PCR amplification of both EsAQP3 and EsAQP4 from slime gland tissue of two hagfish revealed that these two aquaporin homologues are both expressed in the slime gland (see supplementary material Fig. S1). The PCR products were sequenced and confirmed to be segments of the EsAQP3 and EsAQP4 genes, respectively.

We found that Pacific hagfish vesicles exposed to SSW exhibited two distinct categories of swelling behaviour, a pattern also found in Atlantic hagfish Myxine glutinosa (Herr et al., 2010). When the vesicle membrane was disrupted by Triton X-100, swelling behaviour was homogenized. This suggests that the two kinds of swelling can be traced to differences in the vesicle membranes. Luchtel et al. (Luchtel et al., 1991) also proposed that vesicle rupture was mediated by the vesicle membrane, and suggested that it is permeable to monovalent anions and cations of all valencies, and that the movement of these ions into the vesicle causes osmotic influx of water. Given the high concentrations of Na⁺ and Cl⁻ in seawater, it is indeed likely that their concentration gradients would favour net influx. Our results do not support this notion, however, as neither Na⁺ nor Cl⁻ was required for vesicle rupture. Ca²⁺ is required for rupture in approximately 60% of vesicles. The remaining 40% of vesicles could only be stabilized in concentrated solutions of salts with polyvalent anions, and ruptured in nearly all aqueous solutions, including hyperosmotic sugar solutions. These data raise the possibility that the two classes of vesicle swelling kinetics correspond with differences in Ca²⁺ dependence. Luchtel et al. (Luchtel et al., 1991) failed to detect Ca²⁺ sensitivity in their study, but this is not surprising given that they assessed vesicle rupture by whole slime formation and were not able to examine the behaviour of individual vesicles. Ca²⁺-independent vesicles ruptured in the absence of Ca²⁺, Na⁺ and Cl⁻, and even in high concentrations of sucrose or betaine. This non-specific rupture may be a result of damaged membranes, which would let these typically impermeant molecules diffuse into the mucin network and cause swelling. Alternatively, the membranes of these vesicles may contain non-selective transporters that allow the passage of both ions and molecules as large as sucrose.

Mg²⁺, the most abundant divalent cation in seawater, was not able to substitute for Ca²⁺ in its ability to trigger rupture, which suggests a Ca²⁺-specific mechanism. Furthermore, Ca²⁺-dependent rupture occurred in the absence of any monovalent cations or anions. Interestingly, Ca²⁺-dependent rupture was inhibited by the anion channel inhibitors DIDS, tannic acid and PGG. DIDS is an effective Cl⁻ channel inhibitor (Jentsch et al., 2002; Schultz et al., 1999), but also has effects on a variety of integral membrane proteins, including Ca²⁺-activated Cl⁻ channels (CaCCs) (Duran et al., 2010; Ferrera et al., 2010; Huang et al., 2012), bicarbonate transporters (Bevensee et al., 2000; Romero et al., 2000) and pannexins (Ma et al., 2009). Tannic acid and PGG target CaCCs but not other common Cl⁻ channels (Namkung et al., 2010); however, extensive pharmacological profiles of these compounds are lacking. Despite these effects, it seems unlikely that the mechanism of rupture relies on a chloride channel, as rupture occurred in the presence of impermeant monovalent anions such as gluconate, and even in the absence of any monovalent anion.

The ability of vesicles to rupture in most test solutions suggests that the vesicle membrane contains transporters, and that these transporters are Ca²⁺-activated in 60% of vesicles, and constitutively open in the other 40%. While we do not know the identity of these putative transporters, one possibility is that they are similar to those that form gap junctions or gap junction hemi-channels. Connexins and innexins form gap junctions in vertebrates and invertebrates, respectively, while vertebrate pannexins, which share homology with innexins, are more often found as hemi-channels without connections to other cells (reviewed in Scemes et al., 2009) and allow the passage of ions and metabolites up to 1.5 kDa (Wang et al., 2007). It is possible that a pannexin-like channel, or a similar non-specific transporter, is responsible for the movement of materials across the mucin vesicle membrane. Furthermore, some pannexins are activated by Ca²⁺ (Locovei et al., 2006). A pannexin-like transporter in the mucin vesicle membrane would allow the rapid influx of ions in seawater such as Na⁺ and Cl⁻ down their concentration gradients. Even if the vesicle is iso-osmotic to seawater, it is likely that the most osmotically active molecules are the mucins, which are too large to exit the intact vesicle. An influx of ions would thus tip the osmotic balance in favour of net water influx. The only particles that are likely to leave the vesicle are the cationic counterions, which have not yet been identified, but must be present for electrostatic neutrality. Positive counterions also likely play a charge-shielding role and allow charged mucins to remain condensed within developing mucus cells. If these putative positive counterions are more effective at charge shielding than the cations that enter the vesicle during vesicle deployment (such as Na⁺), efflux of counterions will further drive swelling via increased electrostatic repulsion among mucins.

We identified two AQP-like genes expressed in hagfish slime gland tissue that our phylogenetic analysis suggests are closely related to AQP3 and AQP4. Furthermore, when vesicles were treated with the AQP inhibitor Hg²⁺ (Agre et al., 2002; Cerdà and Finn, 2010), the mean swelling rates decreased almost 10-fold. The effect of Hg²⁺ on swelling rate was consistent regardless of whether Ca²⁺ was present, and was even observed in the ion-poor 5 mmol l⁻¹ Tris treatment. These data support the hypothesis that AQPs in the vesicle membrane facilitate the rapid influx of water by greatly increasing the permeability of the vesicle membrane to water. However, it should be noted that Hg²⁺ can have inhibitory effects on a variety of other transporters including K⁺ channels (Gallagher et al., 1995; Leonhardt et al., 1996; Liang et al., 2003), Na⁺ channels (Kurata et al., 1998; Leonhardt et al., 1996), Ca²⁺ channels (Leonhardt et al., 1996; Pekel et al., 1993) and Na-K-2Cl cotransporters (Kinne-Saffran and Kinne, 2001). However, given that Hg²⁺ caused a reduction in swelling rate, and not the inhibition of rupture, as was seen with other chemical inhibitors (DIDS, tannic acid and PGG), ion flux was likely not affected. Indeed, the application of Hg²⁺ to vesicles resulted in a small increase in the proportion of vesicles that ruptured in the absence of Ca²⁺ (supplementary material Fig. S2), which is the opposite of what one would expect if Hg²⁺ inhibited ion channels crucial to vesicle rupture. Mercurial inhibition of AQPs is known to be reversible by reducing agents such as β-mercaptoethanol (reviewed in Cerdà and Finn, 2010); however, such agents also cleave the disulfide bonds that link mucin molecules, and have strong effects on the cohesive properties of hagfish slime (Fudge et al., 2005; Koch et al., 1991; Winegard and Fudge, 2010). For these reasons, we did not test the reversibility of Hg²⁺ inhibition, as it would be impossible to determine whether an increase in swelling rate was due to the reversal of AQP inhibition, or to changes to the mucin network (Verdugo, 1990).

AQPs have been found in secretory vesicles, and are thought to assist vesicle swelling and product release in many cases such as secretory vesicles in human small intestine (Li et al., 2005), rat parotid gland secretory vesicles (Matsuki et al., 2005) and zymogen granules of pancreatic acinar cells (Cho et al., 2002; Itoh et al., 2005). While several AQPs have been localized to intracellular vesicles, including AQP1 (Cho et al., 2002), AQP5 (Matsuki et al., 2005), AQP6 (Yasui et al., 1999), AQP10 (Li et al., 2005) and AQP12 (Itoh et al., 2005), there is no evidence of either AQP3 or AQP4 in vesicle membranes. The detection of these two AQPs in the slime gland may indicate their role in vesicle rupture, but we acknowledge that the slime gland contains several cell types, including an epithelial cell layer. We thus cannot yet conclude that these AQPs are expressed in mucous cells, and further work is required to determine whether one or both of these AQPs are found specifically in the vesicle membranes.

AQP4 is considered to be a mercury-insensitive AQP (Agre et al., 2002; Cerdà and Finn, 2010; Hasegawa et al., 1994; Yang et al., 1997), and this appears to hold true for E. burgeri AQP4 (Nishimoto et al., 2007). However, a more recent study of AQP4 suggests that the Hg²⁺-sensitive domain of AQP4 is located intracellularly and is inaccessible to impermeable Hg²⁺ (Yukutake et al., 2008). While the amino and carboxy termini of AQPs typically face the cytoplasm (Verkman and Mitra, 2000), the orientation of these domains in intracellular vesicle membranes can be intravesicular (Cho et al., 2002) or extravesicular (Matsuki et al., 2005). Thus it is possible that AQP4 is Hg²⁺-sensitive in hagfish slime mucin vesicles, as we do not know the orientation of AQPs in hagfish mucin vesicle membranes. Likewise, although western blot analysis of E. burgeri tissues with a EbAQP4-specific antibody does not reveal EbAQP4 expression in the slime gland (Nishimoto et al., 2007), our evidence of RNA expression of EsAQP4 in the slime gland suggests that AQPs may be more widespread in hagfish tissues than was previously assumed.

Our experiments and analysis lead us to the following model of hagfish slime mucin vesicle rupture. Ca²⁺-sensitive vesicles are released from the slime gland into seawater, exposing them to Ca²⁺ ions that bind to transporters in the membrane, prompting their opening. A net influx of Na⁺, Cl⁻ and other ions ensues, and water follows via osmosis, with influx facilitated by AQPs, leading to swelling and rupture of the vesicle (Fig. 7). In the case of Ca²⁺-insensitive vesicles, the simplest model is the same without the requirement of Ca²⁺ to open the transporter.

The hagfish slime gland is unique from other mucus delivery systems in that it allows the rapid and directed release of mucus from many mucous cells. The evolution of GMCs and GTCs most likely occurred via the invagination and specialization of epidermal cells, which include epidermal thread cells (ETCs), small mucous cells (SMCs) and large mucous cells (LMCs) (Blackstad, 1968; Spitzer and Koch, 1998). SMCs are the most abundant cells in the epidermis (~92%) and are involved in merocrine secretion of epidermal mucus (Spitzer and Koch, 1998). In contrast, LMCs are fewer in number, and little is known about their function; however, Blackstad (Blackstad, 1968) indicated that mature LMCs, as well as ETCs, migrate to the surface of the epidermis and are expelled. We propose that selection for rapid and copious mucus secretion resulted in the development of larger and larger mucous cells whose contents could be ejected via the holocrine mode.

Evolutionarily speaking, holocrine secretion was surely more effective than the much slower process of exocytotic release, which requires fusion of intracellular vesicles with the apical membrane. Explosive holocrine release created its own challenges, however, as mucus granules were expelled into seawater surrounded by a lipid bilayer, which hindered the movement of ions and water into the mucin gel. The addition of transporters and AQP into the vesicle membranes helped alleviate this problem. Making the opening of the transporters dependent on Ca²⁺, an ion that is abundant in seawater, but scarce within cells, may have reduced the risk of premature swelling of the vesicles within the gland.

The evolution of defensive sliming in hagfishes involved a switch from exocytotic to holocrine release of mucin vesicles, which increased the speed at which mucus could be expelled, but limited the rate at which it could be hydrated because the granules were enclosed in a lipid bilayer membrane. Our data suggest that this problem was overcome by incorporating proteins in the vesicle membrane, which increased its permeability to ions and water molecules.

Pacific hagfish [Eptatretus stoutii (Lockington 1878)] were obtained from the Bamfield Marine Sciences Centre, Bamfield, BC, Canada, and were maintained at the University of Guelph Aqualab according to a protocol approved by the University of Guelph Animal Care Committee (protocol 09R128).

Slime exudate samples were collected from anaesthetized hagfish as described in Herr et al. (Herr et al., 2010), and were stored in a stabilization buffer {SB; 0.9 mol l⁻¹ sodium citrate + 0.1 mol l⁻¹ PIPES buffer [piperazine-N,N′-bis(ethanesulfonic acid)], pH 6.7} (Downing et al., 1981; Fudge et al., 2003). Stabilized slime was filtered to separate mucin vesicles from thread skeins, and samples of isolated vesicles were quantified and standardized using the linear relationship between vesicle number and absorbance (350 nm) (Downing et al., 1981).

Slime gland tissue samples were collected following euthanasia by decapitation. Slime glands were frozen in liquid nitrogen for storage at −80°C until they were used for RNA extraction.

Mucin vesicle rupture assays followed the methods described in Herr et al. (Herr et al., 2010). Drug/solution effects on mucin vesicles were quantified by calculating the percentage of mucin vesicles that ruptured when exposed to each solution for 120 s. Vesicle swelling rates were measured by recording videos of vesicle rupture at ~14 frames s⁻¹, and analyzing the rate at which vesicle area increased, frame-by-frame as a percentage of the initial area. The swelling rate for each vesicle was calculated as the slope of a line of best fit through the swelling data.

To test the effect of membrane disruption, mucin vesicles from six hagfish were washed with SB containing 0.1% Triton X-100 prior to exposure to a simplified seawater (SSW) solution (10 mmol l⁻¹ CaCl₂ + 535 mmol l⁻¹ NaCl). To test the effect of HgCl₂ on swelling rate, vesicles were exposed to artificial seawater (ASW; 545 mmol l⁻¹ NaCl), SSW and distilled water buffered with 5.0 mmol l⁻¹ Tris under control conditions and after treatment with 1 mmol l⁻¹ HgCl₂ (MacIver et al., 2009).

To test whether mucin vesicle rupture is dependent on specific ions in seawater, vesicles were exposed to solutions that mimicked the osmotic strength and pH of seawater, but excluded one or more of the most abundant ions found in natural seawater (Na⁺, Cl⁻ and Ca²⁺). These solutions included: ASW, SSW (includes Na⁺, Cl⁻ and Ca²⁺), 545 mmol l⁻¹ NaCl (Ca²⁺-free), 545 mmol l⁻¹ NMDG-Cl⁻ (Na⁺- and Ca²⁺-free), 545 mmol l⁻¹ Na⁺ gluconate (Ca²⁺- and Cl⁻-free), 10 mmol l⁻¹ Ca²⁺ gluconate + 980 mmol l⁻¹ sucrose (Na⁺- and Cl⁻-free), 1 mol l⁻¹ sucrose (osmotic comparison, ion-free), ASW + 15 mmol l⁻¹ EGTA (Ca²⁺-free seawater), and 333 mmol l⁻¹ CaCl₂ (high Ca²⁺, Na⁺-free). To test the effectiveness of a divalent anion versus a monovalent anion in the presence or absence of Ca²⁺, vesicles were exposed to a control solution of 1 mol l⁻¹ betaine, and four betaine solutions containing 10 mmol l⁻¹ of CaCl₂, CaSO₄, NaCl or Na₂SO₄. The concentration of Ca²⁺ required to cause 100% rupture was measured by exposing vesicles to a series of CaCl₂ concentrations, and was compared with a range of concentrations of MgCl₂ to determine the interchangeability of divalent cations. The osmolarities of the varying concentrations of CaCl₂ or MgCl₂ were maintained by adjusting the NaCl concentration for a total osmolarity of 1 osmol l⁻¹. Solution osmolarity was measured with a Vapro Vapor Pressure Osmometer (model 5520, Wescor, Inc., Logan, UT, USA).

Vesicles were exposed to the channel inhibitor DIDS at a concentration of 500 μmol l⁻¹ (0.1% DMSO) in SB prior to exposure to the following three solutions: SSW, 545 mmol l⁻¹ NaCl and 10 mmol l⁻¹ Ca²⁺ gluconate + 980 mmol l⁻¹ sucrose. Rupture of DIDS-treated vesicles was compared with control solutions with 0.1% DMSO (vehicle control), as well as solutions without a vehicle control.

Tannic acid effects were tested by exposing vesicles to the following solutions with and without 20 μmol l⁻¹ tannic acid: ASW, SSW, 545 mmol l⁻¹ NaCl and 10 mmol l⁻¹ Ca²⁺ gluconate + 980 mmol l⁻¹ sucrose. Similarly, mucin vesicles were exposed to 20 μmol l⁻¹ PGG in 0.1% DMSO in SB prior to exposure to the four solutions above as well as a DMSO control.

Protein sequences of known AQP3 genes from a variety of species were obtained from NCBI using the standard BLASTp algorithms (Altschul et al., 1990) using Xenopus laevis aquaporin 3 (NP_001081876) as a query sequence. Eighteen AQP homologues were obtained and aligned with MUSCLE (www.ebi.ac.uk/Tools/msa/muscle/), and using HMMER3 (v3.0; Janelia Farm; hmmer.janelia.org), a hidden Markov model (HMM) profile was calculated from the resulting alignment. We conducted two separate HMMER searches through both a translated hagfish gill/slime gland transcriptome database and a translated hagfish kidney transcriptome (BGI, Beijing, China) using the compiled AQP HMM profile as a query. The sequences that were returned by the HMMER search were BLAST analyzed on NCBI to verify that they belonged to the AQP family of genes. From these searches, we were able to obtain a full-length sequence of EsAQP3 and a partial sequence of EsAQP4. In order to obtain a full-length coding sequence for EsAQP4, interspecies primers (see supplementary material Table S1) were developed using published AQP4 from E. burgeri (NCBI accession BAE93686).

Total RNA was obtained from the hagfish gill (~100 mg) using a TRIzol extraction. The RNA sample was then cleaned of genomic contents using DNase I (Ambion/Life Technologies, Carlsbad, CA, USA) and first-strand gill cDNA was synthesized using RevertAid H-minus M-MuLV reverse transcriptase (Fermentas/Thermo Scientific, Pittsburgh, PA, USA). PCR reactions were conducted using Phusion DNA polymerase (Thermo Scientific) and interspecies or species-specific primers using the manufacturer's specifications for the reaction for 30 cycles. DNA products were sequenced using BigDYE Terminator 3.1 according to the manufacturer's specifications.

Identified EsAQP3 and EsAQP4 protein sequences were aligned separately against other known AQPs belonging to the AQP and aquaglyceroporin subfamilies, respectively (Tingaud-Sequeira et al., 2010), using MUSCLE (Edgar, 2004) in Seaview (Galtier et al., 1996; Gouy et al., 2010) for MacOS. The two aligned sequence files were then refined using GBlocks (Castresana, 2000) to subtract gaps and residues of low/noisy homology with parameters selected to allow for more relaxed stringency (Talavera and Castresana, 2007). Phylogenetic analysis was conducted using RAxML-HPC BlackBox v7.6.3 (Stamatakis, 2006; Stamatakis et al., 2008) hosted on the CIPRES science gateway (Miller et al., 2010) using the LG evolutionary model (Le and Gascuel, 2008). For phylogenetic analysis, base frequencies were determined by the model and the proportion of invariable sites was determined using the GTRGAMMA model. Bootstrap analysis was conducted with 1000 replications. Nineteen (EsAQP3) and 21 (EsAQP4) sequences from a range of aquaglyceroporin and aquaglyceroporin subfamilies, respectively, were used in the analysis (Fig. 6), with GlfP (AP003882) and AQPZ (AAC43518) from Escherichia coli selected as an outgroup for the aquaglyceroporin and AQP trees, respectively.

RT-PCR was used to test whether the identified AQP sequences were expressed in slime gland tissue. Total RNA was extracted from the gland tissue from two hagfish using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA), and reverse-transcribed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA, USA), following treatment with Amplification Grade DNase 1 (Sigma-Aldrich, St Louis, MO, USA). Detection of EsAQP3 and EsAQP4 sequences in the sample via PCR was through a TopTaq DNA Polymerase Kit (Qiagen Sciences, MD, USA) using species-specific primers (see supplementary material Table S2). Sequencing was performed at the University of Guelph Advanced Analysis Centre Genomics Facility to confirm identity. All kits were used following the manufacturer's protocol.

Statistical analyses on Triton X-100 and HgCl₂ experiments were conducted using SPSS Statistics v19 (IBM, Armonk, NY, USA). Vesicle swelling rates were calculated as the slope of the linear equation function, and were log₁₀ transformed. The mean of the log-transformed swelling rates was calculated for each trial to obtain a single value for each true replicate (hagfish) under each treatment and solution condition. Because the hagfish used for the control solution experiments were different animals from the six used for the HgCl₂ treatments, the comparison of control vesicles to HgCl₂-treated vesicles was analyzed using a split plot design. Both percent rupture data and mean swelling rate data were analyzed with a two-way ANOVA for solution and HgCl₂ treatment effects, with HgCl₂ treatment nested within hagfish identity. ANOVA results that indicated significant differences were followed by Tukey HSD post hoc pairwise comparisons based on both split plot and whole plot error terms.

Statistical analysis of ion dependence and chemical inhibition experiment data was conducted using SigmaStat for Windows (v3.5). All percent rupture data were transformed by taking the arcsine of the square root of each value. The ion sensitivity experiments, as well as concentration threshold experiments for Ca²⁺ and Mg²⁺, were analyzed by repeated-measures one-way ANOVA. A t-test was used to compare the highest concentrations of Ca²⁺ and Mg²⁺ to which the vesicles were exposed. Repeated measures two-way ANOVAs were used to test for the effects and interactions of solution and drug treatments for the experiments involving DIDS, tannic acid and PGG. All ANOVA results that indicated significant differences (P<0.05) were followed by Holm–Sidak post hoc analysis.

The authors would like to thank Kyle Crans, Shannon Ferraro, Michael Lawrence and Jean-Luc Stiles for assistance with video analysis, and Dr Oualid Haddad, Dr Atsuko Negishi and Tim Winegard for suggestions and advice. We would also like to thank Dr Todd Gillis, Courtney Deck, Jordan Klaiman, Dr Brian Allen (for statistics), Bob Frank and Matt Cornish from the University of Guelph's Hagen Aqualab, and staff at Bamfield Marine Sciences Centre.