Function of the hypobranchial muscles and hyoidiomandibular ligament during suction capture and bite processing in white-spotted bamboo sharks, Chiloscyllium plagiosum

Suction feeding performance in fishes is influenced by many factors, including but not limited to, the magnitude of orobranchial pressure generated during a suction feeding strike. Furthermore, orobranchial pressure generation is dependent on factors such as the change in orobranchial volume and the power output of musculature involved in orobranchial expansion (Muller et al., 1982; Norton and Brainerd, 1993; Sanford and Wainwright, 2002; Svanback et al., 2002; Carroll and Wainwright, 2006; Wilga and Sanford, 2008; Holzman et al., 2012). To generate high negative pressures in the orobranchial cavity, the muscles that actuate orobranchial expansion have to contract rapidly while generating forces exceeding that produced by the orobranchial pressure. In other words, they need to contract at high velocity (v) while producing large forces (F), thus suction feeding is a power (P=Fv)-dependent behavior.

Power production by skeletal muscle is governed by the force–velocity relationship (Hill, 1949), which results in a trade-off between high force production and high velocity of contraction. Many suction feeding teleosts compensate for the trade-off by using combinations of orobranchial depressors (hypobranchials) and the much larger trunk muscles (epaxials and hypaxials) to aid in expansion of the orobranchial cavity (Tchernavin, 1953; Lauder, 1982; Carroll, 2004; Carroll and Wainwright, 2006; Coughlin and Carroll, 2006; Van Wassenbergh, 2007; Camp and Brainerd, 2014). Other species also use catch mechanisms that allow for storage and release elastic strain energy in elastic elements, such as tendons, to amplify power production during suction feeding (de Lussanet and Muller, 2007; Van Wassenbergh et al., 2008; Roos et al., 2009; Van Wassenbergh et al., 2014).

Power amplification mechanisms in the feeding apparatus are not limited to teleosts, they are also found in other vertebrates such as toads (Lappin et al., 2006), salamanders (Deban et al., 2007) and chameleons (Herrel et al., 2000; de Groot and van Leeuwen, 2004; Anderson and Deban, 2012). Toads increase power production associated with jaw depression and ballistic tongue projection by storing elastic strain energy in the muscles associated with jaw depression (Lappin et al., 2006). Although storage and recovery of elastic strain energy in muscle is not common in vertebrates, the muscles of the feeding apparatus, such as the hypobranchials, are prime candidates for such mechanisms as they typically lack long tendons (Alexander, 1988; Lappin et al., 2006). Thus under circumstances where a hypobranchial muscle (whole muscle) is active, and being lengthened, the muscle tissue is more likely to be the main contributor to elastic energy storage.

Elasmobranch fishes that are obligate suction feeders, such as Ginglymostoma cirratum (nurse shark) and Chiloscyllium plagiosum (white-spotted bamboo shark), are capable of generating high levels of negative orobranchial pressure (−100 kPa) during feeding compared with that of teleost fishes (Sanford and Wainwright, 2002; Carroll et al., 2004; Wilga et al., 2007; Motta et al., 2008). Hypobranchial muscles, which actuate orobranchial expansion, of elasmobranch fishes are either parallel or in-series with each other. The coracomandibularis (CM) muscle, which depresses the lower jaw, originates from the pectoral girdle, extends rostrally and inserts onto the left and right halves of the lower jaw just lateral to the mandibular symphysis. The coracoarcualis (CA) and coracohyoideus (CH) muscles depress the hyoid, are in-series with each other, parallel to the CM and extend from the pectoral girdle to the basihyal of the hyoid (Wilga et al., 2000) (Fig. 1A). The muscles also have very short tendons and aponeuroses at the origins and insertions (Motta and Wilga, 1995, 1999; Ramsay, 2012). The CM, CH and CA of G. cirratum and C. plagiosum are also hypertrophied relative to other elasmobranchs that use combinations of suction and bite feeding such as Squalus acanthias (spiny dogfish), which does not generate such high negative orobranchial pressures (Motta and Wilga, 1999; Wilga et al., 2007; Motta et al., 2008; Ramsay, 2012).

The high negative orobranchial pressures and rapid jaw and hyoid movements generated by C. plagiosum (Wilga et al., 2007) suggests that this species may amplify hypobranchial power production during prey capture, while the hypobranchial anatomy (Wilga et al., 2007; Ramsay, 2012) suggests that a muscle may be the vessel for elastic energy storage rather than a tendon. For example, the CH and CA may initially contract isometrically prior to the onset of hyoid depression and then are suddenly allowed to shorten, or the CH may be actively lengthened by the in-series CA and then suddenly allowed to shorten. Both scenarios would result in high-powered contractions, yet would require a catch mechanism. However, such a catch mechanism has not been identified in suction feeding elasmobranchs such as C. plagiosum and the kinematics of suction feeding in C. plagiosum lacks a fundamental feature of the high-powered suction feeding mechanisms of bony fishes: cranial elevation does not occur (Wilga et al., 2007; Wilga and Sanford, 2008; Nauwelaerts et al., 2007, 2008).

The same muscles (CM, CH and CA) are also used to open the jaws and depress the hyoid during prey processing in elasmobranchs, which typically does not involve high negative pressures, but could result in positive hydrodynamic pressures as prey is manipulated in the orobranchial cavity and compressed between the jaws (Motta et al., 1997; Wilga and Motta, 1998a,b, 2000). Jaw and hyoid depression during prey processing are typically reduced relative to that exhibited during prey capture to avoid releasing captured prey (Motta et al., 1997; Wilga and Motta, 1998a,b, 2000). Therefore, strain and activation in the CM, CH and CA of a strong suction feeder such as C. plagiosum may be altered from that used during suction feeding to promote steady controlled movements of the jaw and hyoid. The shortening velocities of the CM, CH and CA may be reduced so the muscles are slower but more forceful (Hill, 1949). Furthermore, the CH and CA may not necessarily shorten together or as much to actuate hyoid depression; shortening of just the CH or CA while the other contracts isometrically may be sufficient and more economical.

In this study, we investigate the anatomy of feeding apparatus of white-spotted bamboo sharks (C. plagiosum) and experimentally determine how C. plagiosum coordinates activation and strain in the parallel and in-series hypobranchial musculature during suction capture and prey processing, in order to address the following questions: (1) does the feeding apparatus of C. plagiosum exhibit any morphological similarities with suction-feeding bony fish that may be beneficial to muscle power amplification during suction feeding?; (2) are patterns of activation and strain in the hypobranchial musculature of C. plagiosum coordinated in a manner that facilitates elastic energy storage in the hypobranchials, thus increasing potential power output of the muscles during suction feeding?; (3) are shortening velocities in the hypobranchials related to pressure generation?; and (4) are skeletal kinematics and muscle activation patterns conserved across suction capture and prey processing behaviors?

Ten Chiloscyllium plagiosum (Bennett 1830) (55–72 cm total length, TL; four preserved specimens and six fresh frozen) individuals of both sexes were obtained from fish wholesalers and used for morphological analysis. Four C. plagiosum of both sexes (70–72 cm total length) were used for the in vivo part of this study. Sharks were obtained from our breeding population at the University of Rhode Island. Sharks were housed in a 1900 liter circular tank containing filtered seawater of 35 p.p.t. salinity at a temperature of 24±2°C, kept on a 12 h:12 h light:dark cycle and maintained on a diet of squid (Illex sp.) and silversides (Menidia menidia). Maintenance diet was discontinued and sharks were placed in a 757 liter circular experimental tank to acclimate 5–7 days prior to surgery and data collection. Individuals were anesthetized for surgery by immersion in a seawater bath containing 0.175–2 g l⁻¹ of MS-222 (tricane methane sulfonate), then moved to a surgical tray with a reservoir of aerated seawater containing 0.06–0.07 g l⁻¹ MS-222 and intubated for instrumentation. After surgery, sharks were returned to the experimental tank and allowed to recover for 1–3 h. In most instances sharks were swimming and ventilating normally within 45 min. Sharks were fed sections of squid mantle scaled to 1.5 mouth width (mw) during data collection until satiated. Prey scaled in that manner elicited processing bites after suction capture. All housing and experimental procedures were in accordance with the University of Rhode Island Institutional Animal Care and Use Committee (IACUC) protocol (AN05-07-001).

The anatomy of the feeding apparatus was investigated by gross dissection of the fresh and preserved specimens. This part of the study focused primarily on the CM, CH and CA muscles and the cartilages and ligaments of the mandibular and hyoid arches. Anatomical structures were recorded with photographs taken using a JVC GR-DV2000U mini DV digital camcorder mounted on a Bogen Max Repro copy stand. If ligament fiber arrangement was not easily distinguishable, individual ligament fascicles were removed using stainless-steel microforceps to reveal fiber direction and photographed. Figures illustrating the gross anatomy of the feeding apparatus were drawn from the photographs. Anatomical terminology used to describe the morphology of C. plagiosum followed that of Motta and Wilga (1995, 1999) and Wilga and Motta (1998a).

Manual manipulation of fresh and preserved heads were used to visualize ligament and joint interactions at the chondrocraniohyomandibular articulation (ACHM), hyomandibuloceratohyal articulation (AHMC), hyomandibulomandibular articulation (AHMM) and interaction of CM, CH and CA with the hyoid. To simulate jaw and hyoid depression, 0.4 cm diameter cotton–polyester cord was stitched into the tendinous insertions of the cranial musculature associated with jaw depression (CM) and hyoid depression (CH and CA). Cords were attached using a 5 cm quilter's curved sewing needle and threaded down the line of action of the muscles using a 6.4 cm straight sewing needle. Loops were tied at the end of the exposed sections of cord, and cords were pulled along the line of action of the muscles to simulate skeletal kinematics exhibited in studies of muscle activation and kinematics in live sharks (Wu, 1994; Motta et al., 1997; Wilga and Motta, 1998a,b; Motta and Wilga, 2001; Matott et al., 2005).

To record muscle strain, two 1 mm piezoelectric crystals (Sonometrics Corporation, London, ON, Canada) were surgically implanted into the CM, CH and CA through 1–2 mm incisions in the skin (Fig. 1A). Prior to crystal introduction a small puncture for each crystal was made in the muscle through the incisions along fascicle lines using a 12-gauge stainless-steel hypodermic needle to create an introduction path for each crystal and accompanying plastic introducer. Crystals were implanted 7–11 mm apart. To implant the CH, which is deep to the CM and thus not visible through the incisions in the skin, depth was measured from the center of the CH muscle belly to the ventral body surface (12.41±0.24 mm) of six fresh-frozen C. plagiosum specimens (70–72 cm TL) obtained from fish wholesalers for a previous study (Ramsay, 2012). The depth measurements were used to mark the CH muscle belly depth on the 12-gauge introduction needle. To reduce the chances of crystals dislodging from the CH due to action of the muscles superior to it, crystals were introduced just lateral to the lateral border of the CM and two 1 mm hooks of stainless-steel 0.002 mm insulated bifilar wire (California Fine Wire Co.) were glued to the base of the lead wire leaving the 1 mm crystals with a tiny dab of epoxy. The CA consists of many small myomeres. Crystals implanted into the CA crossed more than one myomere. All 1 mm crystals were tested in seawater and at set distances before and after hooks were added to ensure that addition of the hooks did not affect signal transduction.

Skeletal kinematics were measured using five 2 mm piezoelectric crystals with suture loops that were divided into a group of two (crystals 4 and 5) and three (crystals 1 to 3), passed into the buccal cavity through the fifth gill slits and sutured to the skin at the anterior portions of the jaws just posterior to the tooth rows to measure gape distance (crystals 4 and 5), the basihyal of the hyoid arch (crystal 3) and two positions on the cranium within the buccal cavity, on the sagittal plane of the shark (crystals 1 and 2) (Fig. 1B). Crystals 1–3 were used for triangulation of vertical displacement of the hyoid and jaws using the law of cosines (Sanford and Wainwright, 2002; Wilga and Sanford, 2008). Crystals 4 and 5 had to be placed lingual to the exposed tooth rows to protect them from damage during the feeding event, thus linear distances are presented as percentage changes from crystal distances measured at onset of feeding. Sonomicrometry (SONO) uses ultrasound transmitted and received between all pairs of crystals to quantify distance based on the time sound takes to travel through a medium (e.g. muscle, water). The speed of sound through muscle tissue (1560 m s⁻¹) (Mol and Breddels, 1982) was used for all 1 mm crystals, and through seawater at a temperature and salinity of 24°C and 35 p.p.t. (1531 m s⁻¹) was used for all 2 mm crystals (Chen and Millero, 1977). Buccal pressure was also recorded using a 2 mm Millar SPR-799 microcatheter side-tipped pressure transducer threaded into the buccal cavity though the fifth gill slit on the left or right side of the shark (whichever one contained the fewest crystal leads) and sutured to the ventral surface of the cranium just lateral to the two reference crystals 1 and 2 (Fig. 1B), as in Wilga and Sanford (2008). Pressure transducers were calibrated using a similar method as in Higham et al., (2006), which tested over a range of −60 to 0 kPa and found a linear relationship between transducer output and pressure. Transducers used in this study were sent to Millar Instruments, Inc. (Houston, TX, USA) where they were placed in a vacuum to test for linearity in transducer output over the range −101 to 0 kPa. The output voltage of the transducer was found to be a linear function of pressure down to −101 kPa (r²=0.99, see Table S1 for calibration and output curves). Van Leeuwen and Muller (1983) investigated the accuracy of pressure transducer recording during suction feeding in relation to recording frequency and found that most reliable recordings occur at 3 kHz and above. The transducer used in this study samples at a rate of 10 kHz, thus any time delay in pressure change due to the rapid nature of the suction feeding event was considered to be negligible.

Muscle activation was recorded using electromyography (EMG) by implantation of stainless-steel 0.002 mm insulted bifilar wire loaded into 25-gauge hypodermic needles. One millimeter of insulation was stripped from one end and the two exposed wires were bent orthogonally 180 deg, forming a double hook. The hooked ends were inserted into muscle tissue unilateral to the implanted crystals (Gillis and Biewener, 2000) (Fig. 1A).

Wire leads (SONO and EMG) were secured by suturing them to the outer fascia of the muscle they were implanted in and/or suturing skin from the incision tightly around them (Gillis and Biewener, 2000; Carroll, 2004). EMG electrode wires were attached to a multi-pin connector attached to 3 m low-noise cables attached to 16-channel AM Systems model 1700 AC differential amplifiers at a gain of 10 k, bandpass 1.0–10 kHz with 60 Hz notch filter. All muscles were recorded simultaneously using an AD Instruments Powerlab 16SP analog-to-digital converter and software. All SONO crystals were connected to a Sonometrics 16-channel TRX series 16 digital ultrasound measurement system and recorded using Sonoview software. The pressure probe was connected to a Millar instruments pressure control unit (model PCU-2000) and recorded simultaneously with crystal distances in Sonoview [this followed the detailed methods described by Wilga and Sanford (2008), except the inhibit delay was set at 3.25 mm]. To synchronize the EMG and SONO/pressure data, the pressure output line was split and connected to a separate channel on the EMG Powerlab (see above). Peak pressure from the dual recordings was used to match EMG traces with SONO/pressure traces. At the termination of each experiment, sharks were euthanized by MS-222 overdose according to University of Rhode Island Institutional Animal Care and Use Committee guidelines and dissected in order to verify crystal and electrode position.

All sonometric and pressure data were smoothed by removal of obvious outlying points followed by a rolling 3-point average using SonoSOFT software (version 3.4.55, Sonometrics Corporation). To determine the onset and offset of muscle activity, a rectified, integrated EMG signal was used to determine when the signal increased to 2.5 times the signal baseline using Chart software (version 5.4.2, AD Instruments, Colorado Springs, CO, USA). Sonometric, pressure and EMG data were converted to ASCII files in SonoSOFT and Chart, and opened in Microsoft Excel spreadsheets for further analysis. The time of peak gape was set as time zero (t₀) to reference all other temporal variables. The timing of changes between slow and fast linear kinematic and muscle shortening velocities were identified as the times where velocities increased by 1.5 mm s⁻¹ or fiber lengths (FL) s⁻¹, respectively (Carroll, 2004). Chiloscyllium plagiosum is a suction ventilator; therefore, feeding events typically interrupt a ventilation cycle, which involves small increases and decreases in gape, hyoid depression and buccal pressure. Onset of gape increase and hyoid depression were defined as the time when gape and hyoid depression increased to 10% of peak gape and hyoid depression (Wilga and Sanford, 2008). The degree of gape and hyoid motion were calculated as percentages of the resting distances to account for variation resulting from slight differences in implant location.

To compare feeding behaviors, a two-way mixed model ANOVA on repeated measures (RM ANOVA) was performed, with individual as the random effect and behavior as the fixed effect. To compare variables within a behavior, several paired t-tests were performed. Multiple linear regression was used to evaluate the relationship between muscle and kinematic and pressure variables. Separate models were run with gape velocity (mm s⁻¹), gape opening (%), hyoid depression velocity (mm s⁻¹), peak hyoid depression (%) and peak pressure (kPa) as dependent variables and CM, CH and CA velocities (FL s⁻¹) and strain (%) as independent variables. To evaluate the relationship between jaw depression and hyoid depression, possibly due to the LHML linkage, separate models were run with gape velocity (mm s⁻¹) and gape opening (%) as dependent variables and hyoid depression velocity (mm s⁻¹) and peak hyoid depression (%) as independent variables; N=6, five replicates per individual. To remove effects of multicollinearity, any independent variables with variance inflation factors (VIF) of 4.0 or higher were removed from models and the analysis was re-run (Zar, 2009). In cases where only one independent variable remained, a linear regression was run. Statistical tests were run using SPSS (version 12, IBM Corporation) or Systat (version 11, Systat Software Inc.).

As in other elasmobranchs, the feeding apparatus of C. plagiosum is formed from six small cartilages, the paired dorsal, medial and ventral labial cartilages (DLC, MLC and VLC) and 10 large cartilages: a chondrocranium (CR), a mandibular arch consisting of the left and right halves of the palatoquadrate (PQ, upper jaw) and Meckel's cartilage (MC, lower jaw), and a hyoid arch consisting of the paired hyomandibulae (HMD) and ceratohyals (CER) and a singular basihyal (BH) (Fig. 2). The proximal ends of the HMD articulate with the lateral surfaces of the CR at the chondrocraniohyomandibular articulations (ACHM), then extend laterally such that the distal ends articulate Meckel's cartilage at the Meckel's knob (MK) and the proximal end of the CER, forming the hyomandibulomandibular articulation (AHMM) and hyomandibuloceratohyal articulation (AHMC), respectively. The CER extend anteromedially and articulate with the lateral surfaces of the BH, forming the ceratohyobasihyal articulations (ACBH). When in the resting position, the BH of C. plagiosum rests against the ventral surface of the CR, moving the line of action of the coracohyoideus and coracoarcualis (CH and CA) to an orientation that is in-line with the jaw joint and the articulation between the CER and HMD (Fig. 2B). The cartilages of the feeding apparatus have ligaments present in other species that help to guide skeletal motion (Ramsay, 2012); however, one ligament in C. plagiosum differs from that of non-suction feeding elasmobranchs. The hyoidiomandibular ligament extends from the ventral surface of the cranium, medial to the articulation between the CR and HMD. The ligament extends laterally at a slight posterior angle down the length of the HMD and attaches to the posterior surface of the proximal end of the CER, forming the medial hyoidiomandibular ligament (LHMM). In C. plagiosum the fibers of the LHMM extend around the lateral surface of the proximal CER, attach to the anterior surface of the proximal CER and extend anteriorly to attach to the posterior surface of Meckel's cartilage, forming the lateral hyoidiomandibular ligament (LHML). The insertion of the LHML on Meckel's cartilage is in a retroarticular position, ventral to the jaw joint (Fig. 2).

Manual depression of Meckel's cartilage by pulling down the line of action of the CM results in posteroventral rotation of the jaw joint (JJ), while the hyoid remains still and the LHML becomes visibly slacked. Manual manipulation of the hyoid results in hyoid and Meckel's cartilage depression when the BH is pulled posteriorly in the direction of the CH and CA line of action. As the CER rotate posteroventrally at the articulation between the distal hyomandibulae and proximal ceratohyals (AHMC), the LHMLs become visibly taut and pull on Meckel's cartilage, generating posteroventral rotation at the articulation between Meckel's knob and the proximal ceratohyal (AHMM) and the JJ. The orientation of the LHMLs moves from horizontal to posterodorsally angled relative to the rostrocaudal axis of the shark as the hyoid and Meckel's cartilage depress (Fig. 2C,D).

Suction capture events are much shorter in duration, but exhibit more distinct phases than bite processing events. Relative to peak gape, all comparable timings of kinematic, activity and strain onset, offset and peak phases are shorter (P<0.001; Figs 3 and 4, Table 1). The duration of CM and CH activity is shorter during suction capture and the CA is not active during bite processing. Active shortening strain in the CM is greater during bite processing, but CM shortening velocity is the same in both feeding behaviors (P<0.001, P=0.110; Table 2). Active shortening strain and velocity in the CH are higher during suction capture. Gape opening, hyoid elevation and depression percentages and velocities are all greater during suction capture (P<0.001; Table 2). No variables exhibited individual effects (P>0.05).

Prior to the onset of gape opening, the CA begins to actively shorten, the inactive CH is lengthened, the hyoid is elevated and the jaws close (Figs 3A, 4A, Table 1). The CH becomes active while passively lengthening, but continues to lengthen, reaching peak length just after the hyoid reaches peak elevation. Activity in the CM occurs after the onset of CH activity; however, the CM does not begin to shorten until after the CH reaches peak length. Active CM shortening precedes the onset of gape increase. The increase in gape progresses through two phases: a slow phase prior to the onset of hyoid depression and a fast phase where opening velocity increases rapidly (P<0.001; Tables 1 and 2). Shortening of the CM occurs with the slow phase of gape opening and continues until the CH begins to slowly shorten. The onset of the fast phase occurs simultaneously with the onset of hyoid depression, pressure decrease, CM active lengthening, and fast shortening of the CH. Peak gape is reached prior to peak hyoid depression and is simultaneous with peak CH shortening. Peak hyoid depression occurs simultaneously with peak pressure decrease and CM lengthening and the offset of CH activity (Figs 3A, 4A, Table 1).

The hyoid is elevated and the jaws are closed just past the resting distances prior to the onset of gape opening. Onset of CM activity occurs just prior to the onset of hyoid elevation and continues until peak CM shortening and gape closure. The gape increases through distinct slow and fast phases (P<0.001) that occur during hyoid elevation (Figs 3B, 4B, Tables 1 and 2). The CM begins to shorten just prior to the slow phase of gape opening and continues to peak gape. Onset of CH active shortening occurs just prior to peak hyoid elevation and the onset of pressure increase, and active shortening continues past peak gape and the onset of hyoid depression. The CA shows negligible activity (Figs 3B, 4B, Table 1).

Linear regression plots show a strong relationship between the percentage of peak gape opening and peak hyoid depression (r²=0.921, P<0.001), and gape (slow) opening velocity and CM shortening velocity during suction capture (r²=0.947, P<0.001; Fig. 5B,C). Multiple linear regression plots show strong interactions between peak pressure and two muscle velocity variables (CH total shortening velocity and CA total shortening velocity) during suction feeding (r²=0.936; Fig. 5A). Multiple linear regression plots from bite processing data show interactions between total gape opening velocity and two CM variables (total shortening velocity and peak shortening strain) and between hyoid depression velocity and two CH variables (total shortening velocity and total shortening strain) (r²=0.902, P<0.001; r²=0.957, P<0.001; Fig. 5D,E).

Morphological investigation and manual manipulation of the feeding apparatus of C. plagiosum suggests the presence of a catch mechanism involving the line of action of the hyoid depressing muscles (CH and CA) and a linkage involving the distal HMD, proximal CER, Meckel's cartilage, and LHML that couples jaw and hyoid depression during suction capture. The coupling mechanism appears to be a four-bar linkage system, which will be referred to as the ceratomandibular coupling (Fig. 6). Four-bar linkages are a common feature in the jaw depression mechanisms of bony fishes (Anker, 1974; Westneat, 1990, 1991, 2002, 2004), but have not yet been described in elasmobranchs. Four-bar linkages consist of a stable ground link and three mobile links: an input link where force is applied to the system, an output link where force and motion is transferred and a coupler link that connects the input link and output link and transfers the force and motion of the input link to the output link (Westneat, 1990, 1991, 2002, 2004). In C. plagiosum the ground link (Fig. 6C, a) is the distal end of the hyomandibula between the AHMM and AHMC. The input link (b) is the proximal end of the CER from the AHMC to the connection of the LHML on the CER. The coupler link (c) is the LHML from the attachment on the proximal CER to the attachment on Meckel's cartilage. Finally, the output link (d) is the proximal end of Meckel's cartilage from the LHML attachment on Meckel's cartilage to the AHMM (Fig. 6C). As we will explain below, the proposed catch mechanism and ceratomandibular coupling are supported by evidence from in vivo measurement of skeletal kinematics, muscle strain and activation during suction capture. Furthermore, the timing of peak orobranchial pressure, peak hyoid depression and rapid shorting of the CH suggest that the more caudal hypaxial muscles may be working with the hypobranchials to achieve peak suction power. However, during bite processing C. plagiosum alters muscle strain patterns and frees the feeding apparatus from constraints resulting from the ceratomandibular coupling.

The CM is the main jaw depression muscle in most elasmobranchs, extending from the pectoral girdle to the lower jaw, with activity coinciding with jaw depression (Marion, 1905; Motta and Wilga, 1995; Motta et al., 1997; Wilga and Motta, 1998a,b; Motta and Wilga, 1999; Wilga and Motta, 2000). The CM of C. plagiosum has a similar arrangement and activity pattern relative to jaw depression. When biting onto prey, the CM shortens during the entirety of jaw depression; however, during suction capture, active shortening only occurs during the slow jaw depression phase which ends simultaneously with the onset of hyoid depression, fast jaw depression and suction pressure generation (Figs 3A and 4A). The major increase in jaw depression occurs while the CM is actively lengthening. The strong interaction of CM shortening velocity and strain with gape opening velocities in the linear and multiple regression models support the CM as an actuator of jaw depression during the slow phase of jaw depression in suction capture and all phases of depression of the jaw in bite processing (Fig. 5B,D).

In addition, the CM may be functioning as hyoid support. The majority of hyoid displacement during suction feeding in C. plagiosum is posteroventrally directed (Wilga and Sanford, 2008). The CM is ventral to the hyoid arch, therefore the hyoid will be pressed into the dorsal surface of the CM during the expansive phase of suction feeding (Figs 1 and 2). Active lengthening of the CM may aid in decelerating or controlling the rapidly depressing hyoid, possibly reducing the chances of over-extension. Another possibility is that the hyoid pushing into the CM aids in jaw depression by pushing the middle of the CM ventrally, which would pull the jaw ventrally as well.

The CH and CA of elasmobranchs are in-series muscles that extend from the pectoral girdle to the BH of the hyoid arch (Marion, 1905; Motta and Wilga, 1995, 1999) (Fig. 1A). Activity of the CH and CA in elasmobranchs shows that the muscles are activated together to depress the hyoid during prey capture (Motta et al., 1997; Wilga and Motta, 1998a,b, 2000; Matott et al., 2005; Motta et al., 2008). In C. plagiosum, the CH is active during the expansive phase of suction capture and bite processing; however, the CA is only active during the expansive phase of prey capture. Small increases in baseline activity are present in the CA during the bite processing, but they are less than 2.5 times the baseline (Fig. 3).

The strain, activation and kinematic profiles exhibited during suction capture suggest that the CH is functioning to store elastic strain energy from the contraction of the larger CA, thus amplifying power output during suction capture. This idea is supported by the long periods of CA activation (192 ms) prior to any motion of the hyoid, active lengthening of the CH during those periods, followed by rapid shortening of the CH and CA (Table 1, Fig. 3). The pre-activation period of the CA is similar to that observed in feeding muscles of vertebrates that use power-amplified feeding mechanisms such as pipefish (de Lussanet and Muller, 2007; Van Wassenbergh et al., 2008), seahorse (Roos et al., 2009; Van Wassenbergh et al., 2014), toads (Lappin et al., 2006) and chameleons (Herrel et al., 2000; de Groot and van Leeuwen, 2004; Anderson and Deban, 2012). The interaction of the CH and CA results in a total mean shortening velocity of −22.58±0.94 FL s⁻¹ for the CH (Table 2). The velocity of the CH is comparable to some of the highest velocities reported for vertebrate skeletal muscle, such as that of the pectoralis muscle in birds like quail (Coturnix chinensis, 26 FL s⁻¹) and starling (Sturnus vulgaris, 22.3 FL s⁻¹) (Medler, 2002).

Furthermore, C. plagiosum generates some of the highest levels of peak negative orobranchial pressure recorded in a suction feeding vertebrate: −72 kPa (present study) to −100 kPa (Wilga et al., 2007; Wilga and Sanford, 2008). The level of sub-ambient pressure inside the orobranchial cavity during the expansive phase of suction feeding is a major contributor to muscular loading and power production centrarchid fishes and clariid catfishes (Carroll, 2004; Carroll et al., 2004; Van Wassenbergh et al., 2005a; Carroll and Wainwright, 2006). Linear regressions of CH–CA total shortening velocities indicates that hypobranchial contraction velocity may be a major contributor to pressure generation during suction feeding in C. plagiosum and provides further evidence that suction feeding is power dependent (Fig. 5A). However, the timing of peak pressure generation is somewhat peculiar. Peak suction power typically occurs with peak negative pressure generation in fish (Camp et al., 2015; Van Wassenbergh, 2015; Van Wassenbergh et al., 2015). In C. plagiosum the rapid recoil of the CH ends just before peak buccal pressure and hyoid depression are achieved (Fig. 7A, Table 1). The offset in timing suggests that other muscles are working with the CH and CA to achieve peak depression of the hyoid and maximum negative pressure generation. These muscles may be the large segmented hypaxials (HPX), which are in-series with the CH and CA (Fig. 1A). The in-series connection between the two muscle groups is in the form of the highly mobile coracoid bar (described below), which serves as the insertion of the HPX and origin of the CA (Figs 1A and 8B). Caudoventral depression of the coracoid bar by contraction of the HPX would be transferred to the hyoid through the CH–CA and result in further hyoid depression if the contractile phase of the HPX extends past that of the CH and CA. Thus the large HPX may serve as another source of power generation for suction feeding as documented in bony fish (Camp and Brainerd, 2014; Camp et al., 2015).

The catch mechanism for the CH–CA appears similar to that hypothesized in bony fish (Aerts et al., 1987; Muller, 1987), in that alignment of CH–CA line of action with the articulations of the hyoid and jaws results in a small moment arm that allows the CA to actively lengthen the CH without generating any skeletal motion (Figs 2B and 7B). The release of the catch mechanism is not as straightforward, but may involve the medial and lateral portions of the hyoidiomandibular ligament (LHMM, LHML) and the CM, and would progress through five distinct stages during suction capture (Fig. 7). Stage one is initiated when the CA begins to actively shorten. The CH is passively lengthening, due to elevation of the hyoid (preparatory phase; Wilga and Sanford, 2008) and by the shortening CA, which pulls the CH origin caudally. The CH lengthens by 12.83% after becoming active, is now loaded, and ready to be released (Fig. 7A). During stage one the LHMM and LHML form a sling that stabilizes the loose articulation between the HMD and CER, preventing any gliding or rotation of the CER about the HMD (Fig. 7B). Stage two begins once the CH reaches peak active lengthening. The CH remains isometric at 12.83% strain until after the onset of slow jaw depression generated by active shortening of the CM. The CA continues to actively shorten during the isometric phase of the CH. Release of the catch occurs when the CM depresses the jaw during the slow phase of jaw depression (Fig. 7B). The caudoventral rotation of the lower jaw at the JJ may slacken the LHML, leading to stage three where the proximal ends of the CER glide caudodorsally as the still tense LHMM rotates at the attachment to the ventral cranium (Fig. 7B). The motion of the CER with the LHMM moves the distal ends of the CER and attached BH caudoventrally as tension is re-established in the LHML, resulting in slight depression of the hyoid at the end of stage three (Fig. 7B). The slight depression of the hyoid will result in a change in the alignment of the CH–CA line of action, forming a moment arm with the hyomandibuloceratohyal joint that releases the catch (Fig. 7B). Stage four is initiated when the CH–CA begin to rapidly shorten together at the beginning of fast jaw and hyoid depression and continues to shorten until peak hyoid depression and pressure generation (Figs 3A and 4A). The release of the hyoid then engages the ceratomandibular coupling as the CER rotate at the hyomandibuloceratohyal articulations and pull the LHML posterodorsally to synchronize jaw and hyoid depression until peak jaw and hyoid depression is reached (stage five) (Fig. 7).

Another potential release mechanism for the catch may involve the plate-like extension of the caudal-facing rim of Meckel's cartilage (MC) at the MC symphysis, which will be referred to as the MCS, and the resting position of the BH and CER (Fig. 8A). Prior to the expansive phase, the BH sits dorsal to the MCS such that the MCS would block any caudoventral depression of the hyoid if the jaws were held closed by the jaw adductors during the activation phases of the CH and CA (Fig. 8Ai). Release of the catch would be initiated by relaxation of the jaw adductors and depression of the MC by contraction of the CM, thus moving the MCS caudoventrally and allowing depression of the hyoid to occur (Fig. 8Aii).

Another possible method of catch release may involve movement of the CH–CA origin (Fig. 8B). The CH–CA complex has a skeletal origin on the rostroventral portion of the coracoid bar of the pectoral girdle (Figs 1A and 8B). The pectoral girdle of C. plagiosum is a U-shaped element consisting of a single fused coracoid bar and two scapular processes. The girdle attaches to the trunk via the cucullaris muscle rostrodorsally, epaxials caudodorsally, and the hypaxials caudally and ventrally (Wilga and Lauder, 2001). The lack of a skeletal articulation between the pectoral girdle and the trunk or cranial skeleton, along with the muscular support, results in a highly mobile pectoral girdle capable of caudoventral rotation of the girdle during hypaxial contraction (Wilga and Lauder, 2001) (Fig. 8B). Contraction of the hypaxial musculature during the contraction phase of the CH–CA would drop the skeletal origin of the CH–CA ventrally, thus moving the line of action of the muscles ventrally from its alignment with the hyoid articulation, increasing the moment arm of the CH–CA and releasing the catch (Fig. 8Bii). Involvement of the hypaxials to augment power production during suction feeding has been reported in largemouth bass, Micropterus salmoides (Camp and Brainerd, 2014; Camp et al., 2015). Hypaxial contribution to suction feeding in C. plagiosum is supported by the continued hyoid depression and time of peak pressure after peak CH contraction during the expansive phase (Figs 3A, 4A and 7A). However, a very recent study examining the motion of the pectoral girdle during suction feeding in C. plagiosum using x-ray reconstruction of moving morphology (XROMM) has revealed that the girdle does rotate caudoventrally during suction capture, but the motion occurs after peak gape (Camp et al., 2017). This suggests that girdle motion via HPX contraction is more closely associated with transport of prey to the esophagus after initial capture rather than suction capture (Camp et al., 2017).

The results of Camp et al. (2017) provide further support to the CH-CA being the main contributers to power production via catch mechanisms hypotheses 1 or 2 (Fig. 7 and 8A, respectively) during suction capture. Regardless of which method is used to release the catch, the strong interaction between per cent jaw opening and the per cent hyoid depression illustrated in the linear regression (Fig. 5B) supports the LHML as a functional component of the ceratomandibular coupling connecting jaw and hyoid depression.

The CH and CA in C. plagiosum exhibit different strain profiles during bite processing. The CH begins to actively shorten during hyoid elevation as positive pressure increases in the orobranchial cavity. Depression of the hyoid by shortening of the CH decreases the positive orobranchial pressure (Fig. 3B). Kinematic differences include jaw closure prior to peak hyoid depression and reduced hyoid depression when biting (Table 2). Hyoid elevation during bite processing may function to pin large prey against the roof of the mouth as the jaws open and prepare to deliver a bite. This may account for the increase in positive orobranchial pressure (Table 2). The skin on the roof of the mouth and the dorsal surface of the BH contain small caudally facing denticles that the prey may catch on, thus reducing the chance of prey escaping the jaws. The CA is not active during bite processing; however, shortening of the in-series CH does not appear to passively lengthen the CA. The CA is a segmented muscle, composed of myomeres separated by tendinous myosepta (Fig. 1A). It may be that the CA exhibits compartmentalization and only a few myomeres are active to stabilize the origin when the CH is active. Another possibility may involve contraction velocity of the CH and the viscoelastic properties of the CA. Muscle and tendon are viscoelastic tissues and are stiffer when loaded rapidly (Vogel, 2003). The CH contracts more slowly during bite processing than suction captures, but the contraction velocities (−2 to −6 FL s⁻¹) are still high compared with many other vertebrate skeletal muscles (Medler, 2002). Rapid loading of the CA by the CH may result in a stiffened CA that the CH can contract against. Regardless of the method, the CH appears to be the main contributor to hyoid depression. The role of the CH as the main hyoid depressor during bite processing in C. plagiosum is supported by evidence from multiple linear regression models of hyoid depression velocity, CH shortening velocity and CH shortening strain (Fig. 5E).

The hypobranchial muscles of C. plagiosum are used to increase power production during suction feeding by active lengthening of the CH by the CA. Thus the CH function stores strain energy to be released during hyoid depression. Hypobranchials interact without generating skeletal motion due to a catch mechanism similar to that originally proposed by Muller (1987) and Aerts et al. (1987) for bony fishes. The CM, relaxing jaw adductors, and/or the hypaxials may be the key(s) that unlocks the catch, resulting in high-powered suction feeding. Chiloscyllium plagiosum also employs a four-bar linkage system, the ceratomandibular coupling, containing a ligament link, the LHML, linking hyoid depression and jaw depression. During suction capture, LHML function is analogous to that proposed for the mandibulohyoid ligament of bony fish by Alexander (1970), Lauder and Shaffer (1985), Bemis and Lauder (1986) and Westneat (2004). Alternative functions of the mandibulohyoid ligament have been illustrated by Van Wassenbergh et al. (2005b, 2008), which suggest that the ligament works as a hyoid levator linked with lower jaw elevation. If the ceratomandibular coupling of C. plagiosum does function to elevate the hyoid during the compressive phase it may be counterproductive to the suction capture mechanism. Elevation of the hyoid prior to jaw closure may compress the buccal cavity and force water and prey out of the still-open mouth. Function of the ceratomandibular coupling during the compressive phase of suction feeding needs further investigation, but see Ramsay (2012). The linkage appears to disengage during bite processing with only the CM depressing the lower jaw. The ability to disengage the ligament linkage may be useful in bite processing, by allowing the jaws to move independently from the hyoid, while the hyoid may be used to secure prey against the roof of the mouth as the jaws are opened and repositioned. The anatomy of the hypobranchial muscles is conserved among shark species (Marion, 1905; Motta and Wilga, 1995, 1999; Wilga, 2005). Similarly, basic suction feeding kinematic and muscle activity are conserved among elasmobranchs, bony fish and aquatic feeding salamanders (Wainwright et al., 1989; Wilga and Motta, 1998a; Motta and Wilga, 2001). Yet strain profiles during suction feeding and bite processing in C. plagiosum suggest that selection can generate divergent patterns of strain that can result in new behaviors without major changes in morphology, skeletal kinematics and muscle activity.

We gratefully thank Jocelyne Dolce, Shannon Gerry, Anabela Maia and Christopher Sanford for their assistance. Special thanks are due to Elizabeth Brainerd, Thomas Roberts and one anonymous reviewer for providing insightful comments that greatly improved the manuscript. Karel Liem, Andrew Carroll, Sandra Nauwelaerts, Christopher Sanford and Christine Newton Ramsay provided great scientific discussions related to the work. We also thank SeaWorld of San Diego for donating sharks; and Lazaro Garcia, Dawn Simmons, Andrea Scott, Dillon Reilly, Aaron Hawkins and Ed Baker for overseeing animal husbandry. Food for sharks was donated by Quaker Lane Bait and Tackle and Sea Freeze Inc.