Breathing versus feeding in the Pacific hagfish

Hagfish (Phylum Agnatha) may be the oldest extant representatives of the most ancient vertebrate lineage, though this remains controversial (Bardack, 1998; Oisi et al., 2013; Miyashita, 2020). Certainly, they are very different from modern fishes, and of key importance in phylogenetic debates with respect to anatomical, physiological and molecular aspects of early vertebrate evolution. Sadly, they are also endangered by a targeted, rapidly expanding and unsustainable world-wide fishery; 10% of known species are now red-listed by the International Union for the Conservation of Nature, and another 40% are classified as data deficient (Ellis et al., 2015). Hagfish are difficult to study in captivity, for multiple reasons elaborated by Glover et al. (2019), and many aspects of their biology remain unknown. There is a critical need to learn more about the basic physiology of this evolutionarily significant group.

What little is known about their respiratory physiology has been summarized by Johansen and Strahan (1963), Bartels (1988) and Malte and Lomholt (1988). Their breathing mechanism is unique amongst fish. Water is inspired exclusively via a single central nostril; there appears to be no water entry through the mouth. In the Pacific hagfish Eptatretus stoutii, Eom and Wood (2019) described a two-phase pumping system in which ventilatory flow is powered by rhythmic contraction of a single velum chamber that creates suction for inhalation via the nostril, then forces the water down a very long pharynx leading to 24 separate gill pouches (12 per side) as well as into the common pharyngo-cutaneous duct (PCD). The PCD is an apparent bypass shunting system with a separate exit on the left side. The gill pouches and PCD contract at a much lower rate than the velum chamber, and these contractions help drive exhalation through the 24 separate efferent branchial ducts and the PCD aperture. There are gill lamellae in each pouch (but not in the PCD) and the direction of water flow is thought to be countercurrent to the direction of blood flow (Mallatt and Paulsen, 1986). There are no data on ventilation or O₂ extraction at the gills in fed hagfish. However, in fasting hagfish, resting oxygen consumption rate (Ṁ_O2) and O₂ utilization at the gills are low relative to those of other fish (Eom and Wood, 2021). There is also a gradient in O₂ extraction, with the highest expired O₂ partial pressure (P_O2) values at the exits of the anterior gill pouches (i.e. least O₂ utilization), and lowest expired P_O2 values at the exits of the most posterior gill pouches and PCD (i.e. greatest O₂ utilization) (Eom and Wood, 2021). When ventilatory flow spontaneously increases above resting levels, Ṁ_O2 increases in parallel. Increases in ventilatory flow can be achieved by increases in either ventilatory stroke volume or velar contraction frequency (hereafter velar frequency), or both (Eom and Wood, 2019; Eom et al., 2019), but O₂ extraction efficiency (utilization) remains fairly constant (Eom and Wood, 2021).

There is only sparse information on the feeding physiology of hagfish, and its metabolic consequences (reviewed by Martini, 1998; Glover and Bucking, 2015; Glover and Weinrauch, 2019). Indeed, in a recent Commentary in Journal of Experimental Biology, Glover and Weinrauch (2019) highlighted the need for more studies on this aspect of their physiology. Hagfish are traditionally known as scavengers on dead and dying vertebrates, burrowing into the carcasses (Shelton, 1978; Martini, 1998), and some nutrients may be directly absorbed through the skin (Glover et al., 2011). Less widely appreciated is the fact that hagfish are also active predators. Zintzen et al. (2011) documented targeted predation on fish buried in sand at depth in the wild. Hagfish are virtually blind, but find their prey by olfactory cues (Tamburri and Barry, 1999; Glover et al., 2019), probably detected by densely packed chemosensory cells (Schreiner organs; Braun and Northcutt, 1998) in the anterior body surface together with olfactory rosettes (Theisen, 1973). Although lacking jaws, hagfish are very capable of grasping prey with their dental plates and teeth, which can be advanced on either side of the mouth (Clark and Summers, 2007; Clark et al., 2010), followed by ingestion through the mouth apparatus – feeding by ‘engulfment’ (Eom and Wood, 2019). Various cartilaginous structures and complex deep protractor, clavatus, perpendicularis and tubulatus muscles are involved, enabling the dental plates and keratinous teeth to move forwards and backwards for grasping and pulling the prey objects. Several contraction cycles may be needed to force the food down the tract (Clark and Summers, 2007). Subsequent digestion and absorption appear to be slow processes, with morphological changes in the gut and long-lasting but variable post-prandial elevations (8–72 h) in ammonia excretion and O₂ consumption (Ṁ_O2) (Wilkie et al., 2017; Weinrauch et al., 2018; Glover et al., 2019). However, these two studies did not examine feeding by engulfment, but rather used either force feeding of low ration (5% of body mass) by gavage (Glover et al., 2019) or voluntary feeding of unknown ration by allowing the animals to bury into the carcass of a dead hake (Weinrauch et al., 2018).

In the present study, we employed behavioural, physiological and morphological approaches to investigate the interface of feeding versus breathing in Eptatretus stoutii. Our particular focus was on voluntary feeding by engulfment, which ensures consumption of a high ration (∼20% of body mass), on the structures and pathways involved, and on the respiratory consequences. Clearly, there would be a potential for large chunks of ingested prey to interfere with the rhythmic pumping of the velum chamber. We employed both classical morphological approaches (dissection, silicone injection, video recording) and state of the art imaging techniques: diffusible iodine-based contrast-enhanced micro-computed tomography (DICE micro-CT) and high field micro-magnetic resonance imaging (micro-MRI). The latter was used to construct a visual tour through the breathing and feeding passages of the hagfish head. Previous anatomical uncertainties were resolved. We videoed the natural engulfment feeding behaviour in the laboratory, recorded the swallowing action, and then followed the hagfish to see how long it took to complete digestion. As the two previous reports of Ṁ_O2 elevation after feeding differ in both methodology and observed responses (Weinrauch et al., 2018; Glover et al., 2019), we characterized patterns for 24 h after feeding by engulfment. In addition to Ṁ_O2, we also examined whether ventilatory flow (the product of velar frequency and stroke volume), branchial O₂ utilization or both increased, and whether the anterior-to-posterior gradient in O₂ utilization persisted. Eom and Wood (2021) suggested several possible explanations for this gradient, including rebreathing of water from one pouch to the next as it passed posteriorly prior to expulsion. Our objective was to see whether there was any morphological basis for such re-breathing.

Based on the prior information summarized above, our specific hypotheses were: (i) that there would be anatomical separation of feeding versus breathing passages, such that rhythmic velum chamber contraction would not be blocked by engulfed prey objects; (ii) that there would be a morphological basis for the gradient in O₂ utilization from the anterior to posterior gill pouches seen in resting fasted hagfish; (iii) that the expected Ṁ_O2 elevation after feeding would be accompanied by increases in ventilatory flow as a result of elevations in both stroke volume and velar frequency, but little change in O₂ utilization; and (iv) that the high ration associated with engulfment feeding would result in more marked elevation of Ṁ_O2 than reported in previous studies (Weinrauch et al., 2018; Glover et al., 2019).

Pacific hagfish, Eptatretus stoutii (Lockington 1878) (average body mass 99.1 g, range 71.8–158.3 g) were collected in Trevor channel (48° 50.844′ N, 125° 08.321′ W, depth 100 m) on the west coast of Vancouver Island, BC, Canada under permits (XR 194 2017, XR 204 2018, XR 212 2019, XR 136 2021) from the Department of Fisheries and Oceans Canada, using bottom-dwelling traps baited with strips of Pacific hake (Merluccius productus). The captured hagfish were transferred to Bamfield Marine Sciences Centre (BMSC) and placed in fibreglass tanks served with flowing seawater (temperature 11–13°C, salinity 30–31 ppt) and several PVC pipes for hiding shelters. The hagfish were fasted over 2–3 weeks and fed only during experiments. Prior to experiments, the hagfish were weighed. Most experiments were performed during the hours of darkness, starting in the early evening, and the animals were shielded from direct light. After experiments, the hagfish were euthanized using an overdose of tricaine methane sulfonate (MS-222, 5 g l⁻¹ neutralized to pH 7.8 with 5 mol l⁻¹ NaOH; Syndel laboratories, Parksville, BC, Canada) and then eviscerated to ensure death. The experiments reported here were performed under animal usage permits (AUP) approved by the University of British Columbia (UBC, A14-0251, A18-0271) and BMSC Animal Care Committees (AUP RS-17-20, RS-18-20, RS-19-20-21-15), and followed the guidelines of the Canadian Council of Animal Care.

The hagfish generates unidirectional ventilation, inhaling water through a single central nostril duct and exhaling it to the ambient environment through ∼12 pairs of gill pouches and the PCD. The flow is generated by high-frequency velum movement, with low-frequency gill pouch and PCD contraction aiding exhalation (Eom and Wood, 2019; Eom and Wood, 2021). Direct measurement of total ventilatory flow is only possible by placing a flow probe on a tube implanted in the nostril (Perry et al., 2009; Eom and Wood, 2019). However, hagfish cannot feed when the nostril tube is in place. Therefore, in order to monitor ventilation immediately after feeding, hagfish were allowed to feed normally as described below, and then, within 1 min, an operation was initiated to install the nostril tube. The hagfish were quickly anaesthetized in MS-222 (0.6 g l⁻¹, neutralized to pH 7.8 with 5 mol l⁻¹ NaOH) for 1–2 min. Hagfish are known to be hypoxia tolerant, so the anaesthetized hagfish were placed on an operating table without gill irrigation, and their bodies were moistened with wet tissue. A 3 cm length of transparent silicone tubing (6.35 mm o.d. and 4.32 mm i.d.) was inserted into the nostril duct and sutured by two stitches (26 mm 1/2C taper, Perma-Hand Silk, Ethicon, Somerville, NJ, USA) to the skin around the nostril opening area. After the operation, each hagfish was placed in a plastic container (20 cm×20 cm×15 cm) served with flowing seawater. The hagfish resumed their normal coiled posture within 5 min.

An ultrasonic microcirculation blood flow probe (V-series, Transonic Systems Inc., Ithaca, NY, USA) connected to a flow meter (T206, Transonic Systems Inc.) was gently connected to the silicone nostril tube of the recovered hagfish with minimal disturbance, and the total ventilatory flow rate was measured. The measured ventilation signal was converted into digital signals in a PowerLab data integrity system (ADInstruments, Colorado Springs, CO, USA), and visualized in LabChart software v. 7.0 (ADInstruments). Using analysis functions in LabChart, the ventilation flow (V̇_w, ml min⁻¹) and frequency (f_R, min⁻¹) were analysed and averaged per 30 s intervals. The averaged ventilation flow data were divided by fish body mass in order to yield mass-specific V̇_w (ml kg⁻¹ min⁻¹). Then, the standardized V̇_w was divided by the measured frequency (f_R) so the mean stroke volume (SV_w, ml kg⁻¹) was calculated. During ventilation measurement, artifacts such as electrical and physical non-specific noise signals were also detected, especially between 1 and 10 Hz, so they were automatically removed by a low-pass filter in LabChart. It was essential to set the correct orientation of the flow probe as it detects the direction of flow during experiments. Two-point system calibration was checked in the flow meter (Transonic Systems Inc.), and also extrinsic calibration was performed by flowing BMSC seawater (12°C, 31 ppt) through the flow probe at known rates using an aqua lifter vacuum pump (Cheng Gao Plastic and Hardware Electricity, Dongguan, Guangdong, China).

Prior to the experiment, frozen anchovies (Engraulis mordax) were completely thawed in seawater. Two or more hagfish were placed in a round glass bowl filled with 5 l of seawater. Within 5 min, the hagfish exhibited a coiled posture and stayed at the bottom of the glass bowl. Four or five thawed chunks of anchovies (∼10 g per chunk) were then placed into the glass bowl. In order to stimulate feeding behaviour, 3–5 ml of anchovy juice extracted from decayed anchovy chunks incubated in seawater at room temperature was also added. In order to record the overall feeding event, the glass bowl was elevated to a height of 60 cm, so as to facilitate video recording by a portable industrial camera (USB 2 µEye ME, IDS Imaging Development Systems GmbH, Obersulm, Germany) mounted underneath the glass bowl. Review of the video recordings confirmed the amount consumed by each hagfish. In some trials, each anchovy chunk contained a small magnet (<1 g), so the engulfed anchovy chunk could be detected externally by a magnetic detector (AMY6, General Tools Secaucus, NJ, USA). After completion of feeding behaviour, the hagfish resumed their coiled posture and did not show further attraction to the remaining anchovy chunks. In some trials, the fed hagfish were moved to individual plastic containers served with flowing seawater and monitored at least 10 times per day for up to 5 days. In some of these fed hagfish, their responsiveness was tested by gentle pinching or transfer to another container, procedures which invariably elicit activity in fasted hagfish. In other trials, the fed hagfish were then immediately used for whole-animal respirometry (series II), nostril operation (series III), and/or P_O2 recording (series IV). In all tests, in order to minimize stress, which could result in slime generation, we moved the hagfish using a fine net instead of bare hands.

Experiments were started in the evening (19:00–21:00 h). Hagfish (N=8) were allowed to feed voluntarily on pre-weighed chunks of anchovy as described above, and the amount of food ingested was recorded. We watched the individual hagfish, identified by slightly different sizes, throughout the feeding period, so we could quantify how much food in total was eaten by each one. The animals were then placed individually in 1 l darkened glass jars served with flowing seawater, and standing in a running seawater bath to maintain 12°C. Periodically, flow was stopped, the jars were sealed, and the decline in P_O2 over approximately 30 min (exact time noted) was recorded using a galvanic oxygen electrode and meter (DO 6+, Oakton Instruments, Vernon Hills, IL, USA), so as to measure Ṁ_O2 at 0.5, 1, 2, 4, 8, 12, 18 and 24 h after feeding. A parallel control group of hagfish were put through an identical procedure, including exposure to anchovy juice to initiate excitement, but were not allowed to feed, followed by Ṁ_O2 measurements at the same times. Blank measurements demonstrated no significant decline in P_O2 over 30 min.

After feeding (again at 19:00–21:00 h) and nostril operation as described above, ventilatory flow (V̇_w) recordings were started by attaching the flow probe to the nostril tube as soon as the hagfish resumed its resting coiled posture (∼5 min). The flow probe was left in place for periods of up to 60 min for recordings at 1, 4 and 24 h after feeding (N=6–7 at each time). We found that if the flow probe was left attached in the intervening periods, the hagfish became restless. Data for fasted controls were compiled from measurements in non-fed control animals (N=40). In these control experiments, the hagfish were completely resting and had not been subjected to exposure to anchovy juice to initiate excitement. The measured V̇_w, and inspired and expired P_O2 (Pi_O2 and Pe_O2) were used for calculating the Ṁ_O2 (Eqn 3), as in series IV.

Individual hagfish were placed in a 2 l plastic container served with flowing seawater, and P_O2 recordings were made within 0–1 h of feeding, and again at 4 and 24 h after feeding (N=6–8 at each time). As in series III, data for fasted controls were compiled from measurements in non-fed resting animals (N=16) that had not been subjected to exposure to anchovy juice to initiate excitement. To make measurements, we gently flipped over the coiled hagfish to expose the openings of the 12 paired gill pouches and PCD. Although inverted, hagfish still maintained the coiled, resting posture. An O₂ micro-optode probe which was connected to a micro-fibre optic oxygen meter (PreSens, Microx TX3, Precision Sensing GmbH, Regensburg, Germany) was applied within 5 mm of each respective gill opening on one side so as to measure expired water Pe_O2. The micro-optode was similarly applied to measure Pe_O2 at the outflow of the PCD. A three-axis micro-manipulator (Narishige, Tokyo, Japan), mounted by magnet on an iron plate placed underneath the plastic container, was used to manoeuvre the micro-optode probe so that the probe position could be finely adjusted during experiments. The order of measurement of the different openings was randomized. Similar to our previous report (Eom and Wood, 2021), the measured Pe_O2 showed very small rhythmic fluctuations over time. Their rhythmic frequency was much slower than velar movements, and co-ordinated with faint gill pouch contractions. The inhaled water oxygen partial pressure (Pi_O2) was also measured at the entrance of the silicone tubing in the fish nostril.

To image the internal anatomy of the Pacific hagfish, four specimens were used in total for high-field micro-MRI (n=1) and DICE micro-CT (n=4; one of these was the same specimen used for micro-MRI). The hagfish were euthanized as detailed above. Before micro-MRI, the 5 cm cranial portion of the specimen was excised from the remaining body and placed in a 50 ml Falcon™ tube and immersed in Fomblin Y, a H⁺ devoid medium providing a completely dark background in H⁺-MRI, under a low vacuum to extract air bubbles. MRI was performed on an Agilent 9.4T system equipped with a Rapid-40-H⁺ body coil and using a 3D gradient echo sequence with the following parameters: repetition time 19 ms, echo time 3.848 ms, flip angle 30 deg, field of view 50×50×50 mm³, spatial resolution 0.0977 mm isotropic, number of averages 44, total acquisition time 45 h. Before DICE micro-CT, the samples were stained in Lugol's solution (8.3 g l⁻¹ I₂ and 16.6 g l⁻¹ KI) for 1–2 weeks (depending on specimen size) with rocking. The cranial portion of the same specimen used for micro-MRI was also used for DICE micro-CT along with the cranial portion of another specimen. Two additional specimens were used to scan the gill pouch portion. Micro-CT was performed using a Scanco Medical XtremeCT system with the following parameters: x-ray tube voltage 59.4 kVp, x-ray tube current 119 µA, 1000 projections per 180 deg, integration time 132 ms, field of view 70×70×70 mm³, spatial resolution 0.041 mm.

We also performed dissection of freshly euthanized hagfish for which injection of coloured silicone into the nostril duct (red) and the mouth cavity (green or yellow) had been performed. The coloured silicone was prepared by adding red or green commercial food dye (Club House, London, ON, Canada) into ∼20 ml of commercial silicone (Advanced Silicone 2, GE). The silicone was well stirred with a wooden stick for 30 s, placed in a 20 ml disposable syringe, and injected into the nostril duct (red) and mouth cavity (green), respectively. After that, the hagfish was decapitated and the silicone-injected head region was fixed in 4% formalin solution for a week, and then used for dissection.

All the measured inspired water O₂ (Pi_O2), expired water O₂ (Pe_O2), ventilation flow rate (V̇_w), frequency of ventilation (f_R), calculated stroke volume (SV_w), O₂ utilization and oxygen consumption rate (Ṁ_O2) parameters are reported as means±s.e.m. with the number of animals (N). Data were transformed when necessary so as to ensure normality and homogeneity of variance. Two-way repeated measures ANOVA with Tukey's post hoc test was used to examine overall effects of feeding, time and their interaction in series II. One-way ANOVA followed by Dunnett's post hoc test was performed to evaluate whether the changed physiological variables in fed hagfish were significantly different relative to fasted controls in series III and IV. In series IV, in order to assess potential differences in O₂ utilization among the various gill pouches and PCD, two different analyses were performed. In the first, linear regressions were performed of O₂ utilization against gill pouch order, and the significance of the correlation coefficient was assessed. In the second, one-way ANOVA followed by Tukey's post hoc tests were carried out to identify specific differences among pouches. All the statistical tests were conducted in GraphPad Prism software v. 6.0 (La Jolla, CA, USA), and the threshold for statistical significance was P<0.05.

Movie 1 is a video of a bout of active feeding by two hagfish. Hagfish seemed more willing to feed when more than one animal was present. Feeding behaviour appeared to be triggered by olfactory cues, because the freshly thawed anchovy chunks were ignored by the hagfish until the extract of rotten anchovy was added. Shortly thereafter, the hagfish became active, exhibiting orientation towards the chunks, and exposing the dental plates. The following is based on detailed observations on N=8 hagfish. Four stages of feeding behaviours were identified: (1) orientation to chemical cues of anchovy chunks (Fig. 1A), (2) searching for and approaching the chunks (Fig. 1B), (3) exposing the dental plates, biting the chunks, and engulfing the bitten-off parts sequentially (Fig. 1C), and (4) vigorously shaking the anterior body (Fig. 1D). After the hagfish exposed the dental plates, the ventral muscle contraction was observed and it lasted until the engulfed chunks were transferred to the intestine, where they were identified by the magnet detector (Fig. 1E,F,K). The swallowing process took only about 5 s. Maximum engulfing capacity appeared to be about 20% of body mass; each hagfish (∼100 g) usually engulfed two chunks (∼10 g per chunk). The fed hagfish did not show any further attraction to the remaining anchovy chunks (Fig. 1E) and rested motionless in a typical coiled posture. At 1.7 days after feeding, the hagfish appeared to have digested the soft tissues, and defaecation of anchovy bony structures and scales through the anus was observed (Fig. 1G). At 5 days, the fish excreted intestinal membranes (Fig. 1H,I,J; N=3) together with bile and appeared to have finished the digestive and absorptive processes.

The fed animals voluntarily ingested 19.74±0.67% (N=8) of their body mass in anchovy chunks. The two-way repeated measures ANOVA revealed significant effects of feeding (P=0.0076) and time (P<0.0001) but their interaction was not significant (P=0.4234) over the 24 h experimental period. Ṁ_O2 approximately doubled by 0.5 h after feeding, becoming significantly different from that of the fasted controls at 0.5–4 h and also at 12 h (Fig. 2). Note that in this series, the controls had been exposed to anchovy juice to initiate excitement, so the significantly higher Ṁ_O2 values in the experimental animals were solely due to actual feeding.

In comparison to completely resting non-fed control animals, fed hagfish significantly increased all of their ventilatory parameters, O₂ utilization and Ṁ_O2, but with differential contributions at different times (Fig. 3). Thus, increases in velar frequency (f_R; Fig. 3A, one-way ANOVA, P=0.0022) and total ventilatory flow rate (V̇_w; Fig. 3B, P=0.0255) were significant at 24 h, whereas increases in ventilatory stroke volume (SV_w; Fig. 3C, P=0.0013) and mean O₂ utilization (Fig. 3D, P=0.0028) were significant at 1 h, while increases in Ṁ_O2 were significant at both 1 h and 24 h post-feeding (Fig. 3E, P=0.0001). Overall, the increases in Ṁ_O2 relative to controls measured here by the Fick principle were about 2.5-fold, only slightly greater than those measured directly by respirometry in series II (Fig. 2), perhaps reflecting differences in the treatment of the non-fed controls.

Measurements of Pe_O2 during the first hour after feeding demonstrated that the fed hagfish showed almost 2-fold higher O₂ utilization (P=0.0010), averaging ∼66% versus ∼35% in fasting hagfish (Fig. 4). The characteristic trend for progressively increasing O₂ utilization from anterior to posterior pouches and PCD, as previously reported by Eom and Wood (2021), was also seen in the current dataset for fasting fish (Fig. 4A). There was a significant linear correlation (r=0.849, P=0.002) between O₂ utilization and pouch order. One-way ANOVA revealed significant overall variation in utilization among pouches and PCD (P=0.029), though none of the individual differences were significant among the 13 means. However, when the data were combined into three groups (pouches 1–5, pouches 6–12 and PCD) based on the anatomical findings of series V (see below), the one-way ANOVA result was highly significant (P=0.002), with the post hoc tests revealing significant differences of pouches 1–5 (lower) versus pouches 6–12 and PCD (higher); pouches 6–12 and PCD were not significantly different. This pattern disappeared in the fed hagfish at 1 h, with no significant differences among the different gill pouch openings, as evaluated by the same analyses (Fig. 4B). As mean utilization gradually decreased at 4 and 24 h after feeding (Fig. 3D), the pattern tended to re-appear but was not significant (data not shown).

The muscle and cartilage connections were observed in the anterior head region by DICE micro-CT imaging (Fig. 5). The dorsal longitudinal bar is connected to the paired tentacles and ring-shaped cartilage in the nostril duct, while the dental plates are connected to the medio-rostral part of basal plates, in accord with the observations of Oisi et al. (2013).

During hyperventilation, the hagfish vigorously moved specific areas (white dotted areas in Fig. 6A) located 1–2 cm posterior from the eye spot, indicating the position of the velum chamber, which is the fast-suction pump used for inhalation in the hagfish (Eom and Wood, 2019). This movement can be clearly seen in Movie 2. The location of the velum chamber was matched to other anatomical images between lines 4 and 5 in Fig. 6B–F. Fig. 6B is an MRI image showing a mid-sagittal view of the hagfish head. Note that the oronasal septum divides the nostril duct from the mouth cavity, and ends at line 4. Fig. 5C is a lateral view of the hagfish head in micro-MRI; internal bony and cartilaginous structures were sagittally (Fig. 6D) and dorsally (Fig. 6E) visualized in micro-MRI images. Note the presence of four pairs of cartilaginous or bony structures (between lines 4 and 5) that likely form parts of the velum apparatus. Note also the paired dental plates attached to the protractor muscle (between lines 1 and 3). Fig. 6F,G shows photographs of mid-sagittal views of the dissected hagfish head, which had been injected through the nasal duct with red silicone and through the mouth cavity with yellow (Fig. 6F) or green (Fig. 6G) silicone. Note that the yellow silicone injected through the mouth cavity started to merge with the red silicone in a particular area (Fig. 6F); this area was visualized after removing the silicone and matched to the position of the velum chamber (Fig. 6G). A schematic diagram of the mid-sagittal view of the hagfish head was prepared using the information from the above panels (Fig. 6H). The dental plates and the various bony and cartilaginous structures of the velum chamber were outlined from Fig. 6D and recreated in Fig. 6H.

Movie 3 is a video of a micro-MRI tour through the breathing and feeding passages of the anterior part of the hagfish head. We enter anteriorly through the nostril, move under the olfactory rosettes to the potential velum chamber area, and then turn around 180 deg, re-enter, and exit to the environment through the mouth cavity. At the start, we are viewing the hagfish's head from the anterior; note the nostril surrounded by paired tentacles with the mouth opening visible in the midline below the nostril. The dental plates are retracted and are not visible. At 14 s, we enter the nostril and reach the start of the olfactory rosettes, which can be seen on the dorsal surface at 25 s. At 40 s, we reach a small wrinkled chamber which ventrally lies over and is linked to the oronasal septum, leading to a large chamber-like area, likely the start of the velum chamber. At 55 s, we enter into the potential velum chamber and at 60 s we turn direction by 180 deg. After entering the underside of the wrinkled small chamber which is matched to the posterior end of the oronasal septum at 80 s, we reach the mouth cavity and exit to the environment at 110 s.

Connections from the gill pouches to respiratory and vascular tracts were observed by means of DICE micro-CT (Fig. 7A–C) and dissections (Fig. 7D–F). Each gill pouch possessed three visualized connections: pharynx to gill pouch representing connection 1 (C1) and gill pouch to gill opening representing C2 in the respiratory tract, and ventral aorta to gill pouch representing C3 in the vascular tract. Based on the DICE micro-CT imaging, we initially (mis)interpreted the structures marked ‘C3’ in Fig. 7B,C as functional connections for water flow between adjacent gill pouches. However, careful dissection of preserved material could not confirm their presence. We then realized that these were the afferent branches from the ventral aorta to the gill pouches passing into the plane of sectioning of the DICE micro-CT. In the vascular tract, C3 in the anterior gill pouches between gill pouch 1 and gill pouch 5 goes to the separate branches of the bifurcated ventral aorta, but the remainder of the C3 connections in the posterior gill pouches are to the single median ventral aorta, which in turn arises from the heart. As a result of C1 connections, all gill pouches including the PCD are indirectly connected via the pharynx to the intestine.

In the Pacific hagfish, a representative of the oldest extant connection to the ancestral vertebrates, the pharynx is a part of the respiratory tract and connects to the intestine. It is used for both breathing and feeding. As the pharynx serves dual purposes, the rhythmic movement of breathing could be impeded by engulfment of large prey chunks during feeding or vice versa. In this study, therefore, we documented that feeding by engulfment occurs in E. stoutii and tried to understand how the hagfish coordinated breathing with the engulfment and swallowing processes, using visual observation together with physiological measurements and anatomical studies of the anterior hagfish body including the respiratory tract. Our study also looked at the behavioural and physiological aspects of feeding and respiration.

Our first hypothesis, that there would be anatomical separation of feeding versus breathing passages, was not supported. To provide some background, most published models (e.g. Strahan, 1958; Johansen and Strahan, 1963; Li and Wang, 2021) are based on the Atlantic hagfish (Myxine glutinosa) and portray the feeding pathway (mouth) and breathing pathway (nostril duct) uniting anterior to the velum chamber. This would not be a problem when the ingested food is a slurry obtained by rasping. However, in the Pacific hagfish (E. stoutii), which can feed by engulfment of large chunks of prey, Eom and Wood (2019) earlier proposed that the point of union would be posterior to the velum chamber such that feeding and breathing would not conflict. However, Holmes et al. (2011) published an image of a species which they tentatively identified as E. stoutii that appears to contradict this conclusion.

Using CT (Fig. 6B), MRI (Fig. 6C–E) and dissection (Fig. 6F,G) images of the anterior hagfish body, and visual observation of velar movement (Movie 2), we rejected the more anterior location of the velum chamber suggested earlier (Eom and Wood, 2019), and accepted the more posterior location of the velum chamber labelled by Holmes et al. (2011) in their published magnetic resonance image. This means that the oronasal septum, which separates the water flow pathway from the food pathway, terminates at the anterior end, rather than the posterior end of the velum chamber (Fig. 6F–H). Our earlier conclusion had been supported by measurements of reciprocal pressure changes in the nostril and mouth cavity of non-feeding animals, a point to be discussed below.

With the new interpretation as to the location of the velum chamber, prey objects hooked into the mouth cavity by the horny dental teeth and the contraction of the medio-rostral part of the basal plates (Fig. 5) would be transferred into the velum chamber, rather than behind it. When the hagfish is not feeding, the cyclic velum chamber contractions generate suction for continuous unidirectional inflow of water through the nostril, and positive pressure forcing water down the long pharynx (Eom and Wood, 2019). Considering the size of the prey chunks (up to 10% of body mass), the prey objects would block this major respiratory pump such that the nostril ventilation and water flow down the pharynx would be transiently disabled.

However, the hagfish can still exhale water stored in the gill pouches by gill pouch contraction, creating a vacuum in the respiratory tract, because the decreased pressure from the gill pouches cannot be relieved by water inhalation through the nostril duct at this time. This vacuum would help to quickly move the food item past the velum chamber. Additionally, the hagfish frequently showed ventral muscle contraction (Fig. 1D; Movie 1) when they extruded their dental plates (Fig. 1A–C). After engulfment, the hagfish vigorously shook their anterior body in an undulatory, peristaltic fashion in association with very clear ventral muscle contraction. The prey objects were quickly transferred to the intestine (within 5 s) across the ∼10 cm length of pharynx (Movie 1). This repeated peristaltic contraction moved from the anterior to the posterior gill pouches, and lasted until the hagfish successfully transferred the engulfed prey objects into the intestine (Fig. 1D). The ventral muscle contraction stopped as soon as the hagfish resumed a coiled posture. We conclude that hagfish use the nostril and mouth alternately, and that the interruption of breathing is short lasting. In this regard, it is relevant that non-feeding hagfish routinely exhibit periods of complete apnoea (Eom and Wood, 2019), and can stop breathing for prolonged periods when exposed to noxious ammonia (Eom et al., 2019). In future, measurements of pressure in various parts of the system and of nostril flow during the actual feeding event will be needed to substantiate these ideas, and unfortunately will be very difficult to obtain.

Another point needing further exploration is that our earlier measurements in non-feeding hagfish clearly showed pressure peaks in the mouth coinciding with negative pressure troughs and flow peaks in the nostril (Eom and Wood, 2019). Thus, even though the two pathways are not separate but rather meet at the entrance of the velum chamber, they appear to be kept functionally separate. Perhaps the posterior end of the oronasal septum acts as a uvula-type flap, alternately closing off one or the other pathway, an idea to be tested in future experiments.

Our second hypothesis was that there would be a morphological basis for the increasing gradient in O₂ utilization from the anterior to posterior gill pouches seen in resting fasted hagfish (Fig. 4A). Earlier, we had suggested that one possibility would be sequential direct passage of some respired water from anterior to posterior gill pouches (Eom and Wood, 2021). However, based on the anatomical results shown in Fig. 7, we are now confident that there are no direct water connections from one gill pouch to the next, so another explanation is needed. Partial rebreathing of water by backflow from anterior pouches into the pharynx remains a possibility, as does O₂ uptake by the pharyngeal wall itself (Eom and Wood, 2021). However, the present findings on vascular anatomy provide another possible explanation – the different anatomy of the ventral aorta serving anterior versus posterior gill pouches. The anterior ventral aorta is bifurcated into parallel, progressively narrowing branches supplying blood to the 5th to 1st gill pouches (Fig. 7D), but the posterior ventral aorta is a single larger vessel directly connecting the heart to the 12th to 6th gill pouches (Fig. 7E). As a result of bifurcation, the cross-sectional area of the anterior ventral aorta is much smaller than that of the posterior ventral aorta. Thus, we propose that as we move anteriorly, the gill pouches receive progressively less blood flow, and therefore are less effective in O₂ extraction from the water (Fig. 4A). Further studies are required to understand the benefit of ventral aorta bifurcation and lower O₂ utilization in the anterior gill pouches in resting, non-fed hagfish, and how this is overcome after feeding in hagfish (Fig. 4B).

Our third hypothesis was that increased Ṁ_O2 after feeding would be accompanied by little change in O₂ utilization, but rather by increases in V̇_w due to elevations in both SV_w and f_R. In the first hour after feeding, when Ṁ_O2 was already significantly elevated by 2.5-fold (Fig. 3E), this hypothesis was clearly not supported because V̇_w did not change significantly (Fig. 3B) despite an increase in SV_w (Fig. 3A), whereas O₂ utilization almost doubled (Figs 3D, 4). However, at 24 h after feeding, Ṁ_O2 was still elevated, but utilization had returned to control values (Fig. 3D), and now increased SV_w (Fig. 3B), driven by increased f_R (Fig. 3A), was the major contributor. Clearly, the situation is more complex than seen previously with spontaneous hyperventilation (Eom and Wood, 2021), on which the hypothesis was based, and the hagfish has several options at its disposal for increasing Ṁ_O2. In the first hour, the strategy of high O₂ extraction efficiency, and unchanged flow, as a result of increased SV_w but reduced f_R, may represent a more efficient mechanism. This increased O₂ utilization (66%) after feeding brings the low utilization (35%) of the resting hagfish into the range (55–80%) reported for resting teleosts with similar benthic lifestyles (Eddy, 1974; Lomholt and Johansen, 1979; Wood et al., 1979). Possible explanations include increased blood flow to the gill pouches, increased diffusive conductance, and/or decreased venous O₂ levels. However, there are no measurements of these parameters in post-feeding hagfish, or to our knowledge even in post-feeding teleosts, so this is an important topic for future investigation. At later times, elevations in Ṁ_O2 became more dependent upon elevation of flow due to increased f_R.

One important caveat with respect to the results of Fig. 3 is that there was only a single pre-feeding control point, and no simultaneous control group measurements at the same times (1, 4, 24 h) as for the fed hagfish. It is possible that some of the same changes in respiratory parameters could have occurred in non-fed animals at these times. However, this concern is somewhat alleviated by the observation that there were no significant variations in these parameters in the non-fed control hagfish at these same times in series II (Fig. 2).

Our results are supportive but not unequivocal with respect to our fourth hypothesis. We had predicted that the large ration (∼ 20% of body mass) associated with engulfment feeding would result in more marked elevation of Ṁ_O2 than reported in two previous studies. These investigations, one on E. stoutii (Weinrauch et al., 2018) using natural feeding by carcass burrowing (ration unknown), the other on Eptatretus cirrhatus (Glover et al., 2019) using gavage feeding (5% ration), reported comparable relative increases in Ṁ_O2 after feeding, peaking at 1.9- to 3-fold, but with different time courses, with maxima at 8 h in the former and 24 h in the latter. Comparison of our results with those of Weinrauch et al. (2018) is particularly relevant, as the two studies were performed at the same temperature, on the same species collected from the same location, but with different feeding protocols. Our series II experiment (Fig. 2) provides the fairer comparison, as the fish were not instrumented and the measurements were made by closed system respirometry in both investigations. In our experiments, the elevation in Ṁ_O2 was very rapid (within 1 h) and remained significant to 12 h (Fig. 2), perhaps reflecting the very large meals ingested. Weinrauch et al. (2018) made their first measurement at 4 h, where there was no increase, and detected a significant increase only at 8 h; subsequent Ṁ_O2 measurements were slightly elevated at 12–36 h, but none of these values were significantly different from their pre-feeding control value. Another difference was in absolute Ṁ_O2 values; our measurements (Fig. 2) under both control and post-feeding conditions were about 50% lower than those of Weinrauch et al. (2018). However, in our series III, where the fish were instrumented and therefore disturbance was greater, both control and post-feeding peak Ṁ_O2 values were similar to those reported by Weinrauch et al. (2018), but our Ṁ_O2 elevation remained significant at 24 h (Fig. 3E). Overall, these results suggest that had we extended our measurements to a longer time course, even larger cumulative increases in Ṁ_O2 might have been recorded, especially in light of the slow time course of digestion.

The post-prandial elevations in Ṁ_O2 reflect specific dynamic action (SDA), the metabolic cost of processing the meal – digestion, absorption, metabolic interconversions and protein synthesis (Secor, 2009). Note that in series II, typical ‘pre-feeding activity’ was induced in the non-fed controls by exposure to anchovy juice, yet immediate post-feeding elevations in Ṁ_O2 were still significantly greater in the experimental animals that were also exposed to the juice and allowed to ingest the anchovy chunks (Fig. 2). This eliminates excitement as a major cause of the increased metabolic rate. There was however a tendency (non-significant) for Ṁ_O2 to rise over time in the controls, peaking at the same time (12 h) as the highest values in the animals that fed (Fig. 2). Possible explanations include a diurnal variation in metabolic rate and/or initial absorption of some nutrients from the anchovy juice through the body surface (Glover et al., 2011), eliciting a minor SDA response in the controls.

Based on the appearance of faecal material, the hagfish completely digested the soft tissue of engulfed prey objects equivalent to ∼20% of body mass within 2 days and excreted the used gut membrane in 5 days. During this whole period, the fed hagfish maintained a coiled posture and up to 48 h barely responded to any physical contact such as pinching by forceps or transferring from tank to tank. Known defence behaviours, such as knotting behaviour and slime generation, were also suppressed. However, overall responsiveness to disturbance was dramatically increased after the hagfish excreted bony structures and scales within 2 days post-feeding, thereafter showing similar responsiveness to non-fed hagfish between 2 and 5 days.

After finishing the digestion process, the hagfish excreted the used gut membrane (Fig. 1H–J), known as the peritrophic membrane (Adam, 1966). Comparable specialized chitin-based mesh membrane structures are commonly observed in the guts of invertebrates, but occur in few vertebrates (Engel and Moran, 2013; Nakashima et al., 2018). The major functions of the peritrophic membrane in invertebrates have been summarized by Glover and Weinrauch (2019), and include promoting digestion, serving as a structural barrier to protect the gut epithelium and segregating the gut microbiome from the epithelium. Further studies are required to understand the functional roles of the peritrophic membrane in the hagfish digestive process, and especially the nature of the hagfish intestinal microbiome.

In a recent review of the feeding physiology of hagfish, Glover and Weinrauch (2019) highlighted the need for far more work in this field on an important group close to the base of vertebrate evolution. However, they noted that: ‘Many of the key tools of the experimental biologist are compromised by a species that does not readily feed in captivity, is difficult to instrument, and which produces copious quantities of slime’. In the present study, we have been able to overcome some of these impediments by using simple approaches such as very gentle handling, careful instrumentation, and experimentation during the hours of darkness in diffuse light, thereby producing valuable new knowledge. This knowledge has been supplemented by respirometry and both classical and state-of-the art morphological techniques. Future studies should employ comparable methodologies to examine the feeding versus breathing interface over a longer post-prandial period, with different meal sizes, under different O₂ regimes and on hagfish species with different trophic habits. Ideally, the present methodology should be ultimately improved such that physiological changes during the actual feeding event can be recorded.

We thank Drs Tony Farrell, Colin Brauner and Bill Milsom for the loan of equipment, BMSC Research Coordinators Dr Eric Clelland and Tao Eastham for invaluable support, and two anonymous reviewers whose constructive comments improved the manuscript.