Morphology and kinematics of feeding in hagfish: possible functional advantages of jaws

From the Lower Silurian to the Upper Devonian, approximately 80 million years, a diversity of agnathans (jawless fishes) coexisted with the earliest gnathostomes (jawed vertebrates) (Carroll,1988). Agnathans have been reduced to two successful groups, the hagfishes and lampreys, which have persisted since the Paleozoic and collectively comprise just 0.2% of extant craniates. Despite their minimal current diversity, agnathans were the most abundant craniates for over 140 million years (Purnell, 2002). Fossil agnathans were diverse, and often possessed a protective dermal layer of bone and dentine. Jawed craniates became the dominant form following the extinction of most agnathans in the late Devonian. The selective pressures on craniates in the late Devonian are not known and how jaws were initially used is unclear (Mallatt, 1996). Virtually all extant craniates have jaws that are used for various behaviors including breathing, communication, and combat. Jaws are especially important for prey capture and feeding and allow gnathostomes to exploit a variety of food types.

Hagfish are basal craniates and the sister taxon to the vertebrates(Liem et al., 2001). Unlike the lampreys, hagfish lack any traces of vertebrae and are exclusively marine(Liem et al., 2001; Martini, 1998). There are approximately 60 identified species, which are primarily recognized as demersal scavengers feeding on dead or dying marine invertebrates and vertebrates (Fernholm, 1998; Martini, 1998). Though hagfish morphology is unusual, it is highly conserved through time. The earliest known hagfish specimen (Myxinikela siroka), from the late Paleozoic of Illinois (approximately 300 million years ago), is morphologically similar to extant genera (Bardack, 1991). Although the phylogeny of extant hagfish is not well resolved, two subfamilies are recognized: the Myxininae and Eptatretinae(Fernholm, 1998).

The jawless feeding apparatus of hagfish is complex and functions in a very different fashion than vertebrate jaws, though it does have the same basic components: a supporting skeleton, muscles to power movement, and a dental battery to capture and process prey. Hagfish capture and transport food with two rows of non-serrated, grasping keratinous `teeth', which are anchored to dental plates, a bilaterally folding, paired series of cartilages(Fig. 1A)(Cole, 1905; Cole, 1907; Dawson, 1963). Dental plates are supported by anteroventrally situated basal plate cartilages(Fig. 1B) and, during feeding,are pulled in and out of the mouth via retractor and protractor muscle groups (Fig. 2). When feeding, hagfish occasionally tie their bodies into knots to forcefully remove chunks of flesh from large carcasses(Martini, 1998). A single posteriorly curved tooth situated in the palate augments knot-tying behaviors by allowing a hagfish to anchor itself to the prey(Fig. 1C)(Cole, 1905; Dawson, 1963).

Though we cannot directly apply what we learn from the extant jawless fishes to their armored, and in some cases exceedingly large, ancestors,studying jawless feeding systems is a useful window into the functional advantages provided by jaws. In this study, we aim to (1) compare the morphology of the feeding apparatus of two hagfish species, Eptatretus stoutii and Myxine glutinosa, (2) compare the feeding kinematics in E. stoutii and M. glutinosa, (3) calculate forces generated by the musculature during feeding, (4) propose a physical model of the hagfish feeding mechanism, and (5) compare hagfish feeding performance to that of gnathostomes to evaluate the functional constraints of jawlessness. We employed dissection, morphometrics and video techniques to address the following hypotheses: (1) there is little or no interspecific variation in feeding morphology and mechanics between E. stoutii and M. glutinosa, as dietary diversity is minimal in hagfishes and (2) the absence of jaws and a rigid skeleton limits force generation in hagfish feeding muscles and dental plate protraction–retraction speed.

Care and handling of all living specimens were performed ethically, though as invertebrate chordates hagfish are not required to be covered by an animal care protocol. Live Pacific hagfish, Eptatretus stoutii (Lockington)were obtained from Scripps Institute of Oceanography (La Jolla, CA, USA) and live Atlantic hagfish, Myxine glutinosa Linnaeus, were obtained from Mount Desert Island Biological Laboratories (Salisbury Cove, ME, USA). Both species were housed together in aquaria with recirculating artificial saltwater maintained at approximately 10°C and 34 p.p.t.(Gustafson, 1935). Most animals were housed in an aquarium (58.42 cm×33.02 cm×33.02 cm)with shades and a cover to block ambient light and to prevent escape. Experimental specimens were filmed feeding in a smaller aquarium (50.80 cm×25.4 cm×31.75 cm) under moderate lighting from two 15 W fluorescent lamps. Individual animals in the smaller aquarium were corralled in a filming area (17.78 cm×30.48 cm×33.20 cm) bordered with perforated white plastic sheets. The smaller aquarium remained unlit when individuals were not being filmed. For maintenance, animals were fed small pieces of squid once every 2 weeks.

Hagfish were selected based on their willingness to feed. Once shifted to the filming tank, an animal was offered rectangular portions (1.0 cm×2.0 cm×0.25 cm) of squid. Each squid was loosely secured to a plastic tie and then positioned directly in front of, and sometimes touching, the animal's mouth. Feeding behaviors were recorded with a JVC digital camcorder at 30 frames s^–1, an appropriate frame rate for the feeding performance of these animals (see Movie 1 in supplementary material).

We collected dental plate kinematic data from five specimens of M. glutinosa (TL=32.0–42.0 cm) and three specimens of E. stoutii (TL=28.0–38.0 cm). We analyzed 9–31 feeding bouts from each M. glutinosa and 5–18 from each E. stoutii, with 5–10 min intervals between feeding bouts. Dental plate kinematic variables included gape cycle time, time to maximum gape,dental plate retraction time, and maximum protraction angle. Gape cycle time(GCT) was defined as the time required for a hagfish to fully protract and retract its dental plate (Fig. 4A). Time to maximum gape (TMG) was the elapsed time for maximal dental plate protraction to be attained from a retracted position(Fig. 4A). Dental plate retraction time (DPRT) was the elapsed time for dental plate retraction from a maximally protracted position (Fig. 4A). From both hagfish species, we also compared mean GCT and mean TMG from capture and transport phases(Table 1). We defined maximum protraction angle (MPA) as the angle between the anterior tip of the maximally protracted dental plate, the mouth opening, and the anterior tip of the snout(Fig. 4B).

Kinematic data determined from cranial movements other than dental plate movements were obtained from 11 feeding bouts in two specimens of E. stoutii (TL=29.0 cm and 38.0 cm) and 25 feeding bouts in four specimens of M. glutinosa (TL=32.0–37.5 cm). Cranial kinematic variables included head depression angle(Fig. 4B), head depression time(Fig. 4A) and head elevation time (Fig. 4A). Head depression angle (HDA) was the angle between the anterior tip of the snout just before the onset of head depression, the pivoting point of the head, and the anterior tip of the snout once the head was maximally depressed(Fig. 4B). Head depression time(HDT) was the elapsed time from the onset to the completion of head depression(Fig. 4A). Head elevation time(HET) was the elapsed time from the onset to completion of head elevation from a maximally depressed state (Fig. 4A). Cranial kinematic variables were only recorded from food transport phases. MPA and HDA were measured with Image J software (NIH). We only used lateral views of hagfish feeding events for determining MPA and HDA. Time variables (GCT, TMG, DPRT, HDT and HET) were only recorded from video clips that clearly showed dental plate and cranial movements.

Because of a limited sample size and size-range of E. stoutii and M. glutinosa, we could not accurately scale hagfish GCT with length. However for comparative purposes, we plotted the relationship between body length and GCT from both hagfish species in this study and from most of the gnathostome species listed in the Appendix. From these data, we graphed the relationship between body length and GCT in aquatic and terrestrial feeding craniates.

We used a one-way analysis of variance (ANOVA) in SPSS 12.0 to compare the means of dental plate dimensions, basal plate dimensions, HFA length, GCT,TMG, DPRT, MPA, HDA, HDT and HET in M. glutinosa and E. stoutii. Each species represented an independent group and our test variables included anatomical measurements and kinematic variables. A one-way ANOVA was also used for comparing mean GCT and TMG from food capture and transport phases, in which the feeding phases (food capture or transport) were test variables and kinematic variables were independent groups. The significance level of P=0.05 was used in all analyses.

Least-squares (LS) regressions of GCT and body lengths were plotted in JMPIN (SAS Institute, Cary, NC, USA). Independent (body length) and dependent(GCT) axes were log-transformed in each regression. LS regressions were appropriate for this analysis, as the probability of measurement error is greater in the dependent data (GCT) than in the independent data (body length)(Sokal and Rohlf, 1995). Plots included fitted mean lines (slope of zero) and fitted log-transformed lines with 95% confidence limits. We tested each LS regression for statistical significance (P=0.05) by comparing the slopes of the fitted log-transformed line and fitted mean line using ANOVA.

Morphology of the hagfish feeding apparatus is illustrated in Figs 1 and 2. For additional reference,excellent descriptions of the morphology of the HFA can be reviewed in the literature (Cole, 1905; Cole, 1907; Dawson, 1963).

When normalized to total length (TL), mean HFA length, basal plate dimensions (length and width) and dental plate dimensions (length and width)were significantly greater in Eptatretus stoutii(Fig. 5). Normalized feeding apparatus length in E. stoutii was 22% longer than in M. glutinosa, basal plates were 15% longer and 16% wider, and dental plates were 20% longer and 16% wider.

Neither absolute nor TL-scaled deep protractor muscle PCSA and P_o were significantly different in E. stoutii and M. glutinosa. Deep protractor muscles theoretically exert 2.76–2.99 N on the dental plate during feeding. When scaled to length,mean force production of E. stoutii clavatus muscle was significantly larger (10.23 N) than that of M. glutinosa (6.73 N)(Fig. 5G,I).

Feeding in both E. stoutii and M. glutinosa can be divided into four general stages: identification, positioning, food ingestion and intraoral transport. Smell and touch appear to be the means by which hagfish identify food. Identification involves independent movement of the tentacles as they contact the food. Simultaneously with or immediately following identification, the mouth is positioned onto or next to the food. Once mouth positioning is established, the food capture (ingestion) stage begins; the dental plates are repeatedly protracted and retracted until the food is engulfed. During protraction, dental plates laterally unfold as they are pulled out of the mouth. Protraction is coupled with simultaneous unveiling of the oral mucosa from the dorsal (toothed) surface of dental plates, which exposes teeth. The toothed surface of the dental plate often slides against the food during protraction. During retraction, the dental plates are pulled back into the mouth. Food becomes hooked on the teeth once retraction begins and becomes even more secure as the dental plates begin to fold medially. Upon entering the mouth, oral mucosa begins to envelope the dental plate, which unhooks the food from the teeth and pushes the food into the esophagus. Intra-oral transport, which begins once food is ingested,involves repeated dental plate protraction and retraction events coupled with repeated head depression and head elevation.

Animals that responded well to filming tank settings typically remained motionless at the bottom of the tank within 30 min of introduction. M. glutinosa specimens would lie sprawled and E. stoutii specimens would lie coiled. We did not observe any hagfish actively pursuing food. Food was usually positioned just in front of, and sometimes touching the animal to elicit a feeding bout. Approximately 2 out of 10 attempts to feed animals in the filming tank were successful. On average, both hagfish species required 3 dental plate protraction–retraction events to ingest a squid rectangle. Because squid rectangles were loosely secured to their plastic ties,knot-tying behaviors were not observed in any feeding event. Hagfish would engage in knot-tying if food was firmly tied; however, these data were excluded from this study.

Mean GCT, TMG, and DPRT, graphed in Fig. 6 and listed in Table 2, are pooled from both food capture and transport events. Although GCT, TMG, and DPRT varied considerably in both hagfish species, with more variation in M. glutinosa, mean values did not differ significantly (Table 2, Fig. 6A–C). GCT averaged about 1 s, one third of which was protraction and two thirds of which were retraction. Mean MPA in E. stoutii and M. glutinosa were 175° and 178°, respectively, and did not differ significantly(Table 2, Fig. 6F). Mean GCT and mean TMG from capture and transport events were similar in M. glutinosa(Table 1). Mean capture and transport TMG were similar in E. stoutii; however, mean GCT in E. stoutii were significantly longer in food transport events than in food capture events (Table 1).

Cranial movements were similar in E. stoutii and M. glutinosa during food transport events. We observed no significant differences in head depression times, head elevation times or head depression angles (Table 2, Fig. 6D,E,G). The onset of head depression occurred between the onset of dental plate protraction and the time of maximum gape (Fig. 7). The head would usually remain depressed for 477 ms prior to the onset of head elevation. Head depression time overlap with dental plate retraction time was only apparent in E. stoutii; however, head elevation in both species occurred in the final moments of dental plate retraction and was completed by the end of the gape cycle (Fig. 7).

The long GCT in E. stoutii and M. glutinosa are noticeable outliers of the LS regression of GCT on body lengths for aquatic feeding craniates. The hagfish data lie outside of the 95% confidence limits and are an order of magnitude above the regression line at their respective body length (Fig. 8A). The slopes of the graphs for both aquatic and terrestrial species are significantly different from zero, indicating a significant relationship between body length and GCT (R²=0.4197; P<0.0001 and R²=0.3727; P<0.005,respectively) (Fig. 8). GCT increases hypoallometrically in aquatic and terrestrial feeding craniates(b=0.5499 and b=0.4792, respectively).

In some respects, the capabilities of the jawless hagfish feeding apparatus are comparable to those of vertebrate jaws. For example, gape size, a combination of gape angle and jaw length, is an important parameter in determining the feeding niche of gnathostomes because it limits the size of prey that can be consumed whole(Hambright, 1991; Luczkovich et al., 1995; Nilsson and Bronmark, 2000). Despite lacking jaws, hagfish can evert their dental plates to an angle of approximately 180°. Among jawed vertebrates, this extreme is achieved during feeding only by snakes (Cundall and Greene, 2000) and algal scrapers such as Loricariid and Mochikid catfishes (Geerinckx et al.,2007), anuran tadpoles(Wassersug and Yamashita,2001) and the prowfish (Zaprora silenus)(Carollo and Rankin, 1998; Clemens and Wilby, 1961). Large gape angles in male Hippopotamus (110°) and male Babirussa (∼180°) are produced for threat displays and not during feeding (Herring, 1972; Herring and Herring, 1974). Because the protraction angle attained in hagfish is so wide, food size is limited only by the perimeter of the soft tissue around the mouth. Though it might appear that a 180° gape is unimportant to a scavenging carnivore,wide protraction angles enable hagfish to remove relatively large pieces of flesh. Returning the dental plate from a long, 180° excursion prior to ingestion imposes strain on prey tissue, which may facilitate dismembering. Removing large portions of food requires fewer feeding bouts, with the potential advantage of less competition with conspecifics and a shorter window of vulnerability in and on a carcass.

Another surprising area where hagfish function as well as gnathostomes is in generating force to accomplish the prey capture cycle. Lacking jaws and a rigid skeleton does not limit the forces that can be generated in hagfish feeding musculature. In lateral view, the HFA can be modeled as a fixed pulley system, in which the input force or speed is equivalent to output force or speed (Fig. 9A). Assuming the effects of friction, inertia and angular changes of the dental plate during protraction and retraction are negligible, the input forces from the feeding muscles are reasonable approximations of the forces exerted by the dental plate. However, in jaws, input forces from muscles are not directly translated to the bite forces (Fig. 9B). We calculated that the mean forces generated by the clavatus and deep protractor muscles of both hagfish species are equal to or exceed the bite forces in four wrasse species (Clifton and Motta, 1998), six turtle species(Herrel et al., 2002), and several (27) finch species (van der Meij and Bout, 2004). Bite force in jaws and the rasping force in dental plates are important parameters for feeding performance because the amount of force can determine the hardness of food that can be processed and the amount of food that can be held in the mouth. Bite force measurements serve as an index for diet and demonstrate how cranial morphology influences the ecology of a species (Hernandez and Motta, 1997; Herrel et al.,2001; Huber et al.,2005; Wainwright,1987). Considering the absence of jaws and forelimbs in hagfish,the forces exerted on dental plates are necessary for grasping and processing chunks of flesh.

Estimated force production in the deep protractor muscles (DPM) and clavatus muscles (CM) and their roles in feeding demonstrate a clear trade-off in force and speed. DPM are longitudinal muscles designed for relatively rapid, yet less forceful protraction of the dental plate while the clavatus muscle is a bipennate muscle that retracts the dental plate more slowly but with greater force. Clavatus muscle force, which is almost an order of magnitude greater than that of the deep protractor muscle, is comparable to some gnathostome jaw closing muscles. The force generated by the clavatus muscle is transmitted to the dental plates via the clavatus tendon,which has been shown to be as strong (47.8±3.5 MPa) and stiff(290±29 MPa) as some gnathostome tendons(Summers and Koob, 2002). Considering the high forces generated in retracting musculature, it is not surprising that the clavatus tendon has a similar response to load as in gnathostome tendons.

Because hagfish feeding forces and gapes are comparable to, or in some cases exceed, the forces and gapes in gnathostomes, we can assume that vertebrate jaws are not prerequisites for producing considerable muscle forces and wide gapes. Challenges other than muscle force and wide gape may have been associated with the selective pressures placed on the gnathostome common ancestor. Extant phylogenetic bracketing (EPB) can infer obscure traits, such as feeding mechanics in extinct agnathans and the common ancestor to gnathostomes. Causal biological relationships between preservable traits (e.g. hard tissue) and unpreservable traits (e.g. soft tissue and behavior)determine the degree of likelihood (level of inference) of obscure traits occurring in extinct taxa (Witmer,1995). EPB is a well-supported method for reconstructing unclear scenarios in a phylogenetic context(Barrett and Rayfield, 2006; Carrano, 2000; Erickson et al., 2002; Jasinoski et al., 2006; Perry and Sander, 2004). Our results demonstrate a probable causal relationship between preservable cranial endoskeletons and unpreservable traits such as feeding muscle force and wide gapes. Because this relationship is maintained in hagfishes and gnathostomes,we can assume with a minimal level of speculation [level 1 inference(sensuWitmer, 1995)]that muscle force production and wide gapes are characteristic of all agnathan taxa and therefore are not functional innovations of craniate jaws.

In general all gnathostome species bite faster than hagfish(Appendix, Fig. 8). Rapid jaw movements are important in the capture of elusive prey. Furthermore, short GCT and TMG are required to generate the low buccal pressures of suction feeding, a common means of food capture in the aquatic medium(Grubich and Wainwright, 1997; Lauder, 1980; Liem, 1990; Svanback et al., 2002). Dental plate GCT and TMG are long in hagfish, making it difficult to feed on active prey. Although gentle ciliary-generated suction needed for filter feeding is the primitive mode in chordates (occurring in both urochordates and cephalochordates), it is not the immediate precursor to the strong suction seen in most aquatic vertebrates. Instead, a grasping-tearing mode induced by repeated movements of muscularly suspended cartilaginous dental plates appears to be the intermediate.

The significant correlation between GCT and body size across various species of aquatic and terrestrial feeding craniates shows both that hagfish are slow for their size and that the scaling relationship for terrestrial and aquatic feeders is similar (Fig. 8). The GCT of both aquatic and terrestrial feeders scales approximately with L^0.5. That is, feeding in water appears to require similar speed per length as feeding on land. This could indicate that the scaling of GCT with body size is independent of prey capture mode. For example, suction feeding, the dominant prey capture mode in the aquatic medium, does not impose shorter GCT per unit body length in aquatic feeders than in terrestrial feeders.

The functional flexibility of the gnathostome feeding apparatus, at the individual level, and in an evolutionary context, is driven by the system of levers and kinetic chains that determines kinematic parameters(Westneat, 1990; Westneat, 2004). In general,jaws are third order levers or 4-bar linkages geared to increase closing velocity. In hagfish, the muscular attachments on the dental plate, and movement of the dental plate about the basal plate, resemble a fixed pulley,which neither offers speed nor force amplification as in levers or linkages(Fig. 9). The key innovation of jaws may be that they allow a lever system or linkage system to increase kinematic transmission efficiency (KT) of the jaw muscles at the expense of high force transmission (Fig. 9B).

The differences in feeding apparatus morphology between these two hagfish species beg further studies of morphology and diet. Although the diet of M. glutinosa is more known than that of E. stoutii, marine invertebrates appear to be the primary diet in both species(Martini, 1998). The more robust apparatus of E. stoutii relative to M. glutinosa is supported by larger dental plates that have more area for dentition. Presumably the larger number of teeth in E. stoutii allows them to grasp prey more firmly. Greater inertial resistance of the dental plates might explain the greater force we calculated in E. stoutii clavatus muscles. However, these morphological disparities do not have significant effects on the kinematics of feeding, except that E. stoutii has a longer gape cycle time during the transport phase and head depression generally overlaps the onset of dental plate retraction. We suppose that the morphological differences between E. stoutii and M. glutinosa are barely manifested in feeding kinematics because (1)morphological disparity is not always translated to feeding behavior and mechanics (Hulsey and Wainwright,2002) or (2) the particular kinematics we examined are unaffected by the morphological parameters we measured.

In summary, bite speed appears to be a major functional innovation allowed by jaws. Even the most basal gnathostomes, the arthrodire placoderms, managed to overcome the constraint of heavy dermal cranial armor with 4-bar linkages of considerable KT (Anderson and Westneat,2006). The configuration of the jawless hagfish feeding apparatus as a fixed pulley provides both an advantage, minimal reduction in force transmitted from the feeding muscles to dental plates, and a disadvantage,impaired rapid dental plate movements. Lever and linkage systems in vertebrate jaws permit functional flexibility, enabling gnathostomes to occupy a diversity of dietary niches. Nevertheless, hagfish protract their dental plates to extremely wide angles rarely attained by jawed vertebrates and posses feeding muscles that can generate forces comparable to that in some gnathostomes. Our data, coupled with EPB principles, suggest that generating considerable muscular forces and attaining wide gape angles were present in the common ancestor to the craniates, not the common ancestor to gnathostomes. Long gape cycle time, wide protraction angle, and considerable muscle force in hagfish are suitable for a diet consisting of dead or dying prey. The feeding mechanism in hagfish, which also occurs in lampreys and possibly some extinct agnathan lineages (Janvier,1993; Yalden,1985), appears to be an intermediate form between the cephalochordates and gnathostomes. Morphological variation exists in the feeding apparatuses of E. stoutii and M. glutinosa,nevertheless the feeding mechanism is conserved in both species.

List of abbreviations

 
• ρ

  muscle density

 
• θ

  muscle pennation angle

 
• ANOVA

  analysis of variance

 
• CM

  clavatus muscle

 
• DPM

  deep protractor muscle

 
• DPRT

  dental plate retraction time

 
• EPB

  extant phylogenetic bracketing

 
• GCT

  gape cycle time

 
• HDA

  head depression angle

 
• HDT

  head depression time

 
• HET

  head elevation time

 
• HFA

  hagfish feeding apparatus

 
• K

  specific tension

 
• KT

  kinematic transmission efficiency

 
• L_M

  muscle length

 
• LS

  least-squares

 
• M

  muscle mass

 
• MPA

  maximum protraction angle

 
• PCSA

  physiological cross-sectional area

 
• P_o

  maximum force production

 
• TL

  total length

 
• TMG

  time to maximum gape

Eddie Kisfaludy and Tom Koob generously provided Pacific and Atlantic specimens, respectively. Sabreena Kasbati, Ahmed Ibrahim, and Peggy Kozick assisted with aquarium maintenance. Drafts of this manuscript were improved by critical, intellectual input from two anonymous reviewers, Kate Loudon, Doug Fudge, the Comparative Physiology Group and Biomechanics Labs at UCI. We thank the National Science Foundation (IOB-0616322 awarded to A.P.S.) and the American Physiological Society for a Porter Fellowship awarded to A.J.C.