Aquatic feeding in pipid frogs: the use of suction for prey capture

Inertial suction feeding is ancestral to vertebrates and remains the most common mode of prey capture among aquatic vertebrates, including most teleost fish (Alexander, 1970; Ferry-Graham and Lauder, 2001; van Leeuwen and Müller, 1984), some turtles (Lemell et al., 2000; Lemell et al., 2002; Summers et al., 1998; van Damme and Aerts, 1997) and salamanders (Deban and Marks, 2002; Deban and O'Reilly, 2005; O'Reilly et al., 2002). Pipid frogs are unusual among anurans in that they are strictly aquatic. Unlike typical anurans, pipid frogs feed in their aquatic environment (Duellman and Trueb, 1986). Pipids are highly specialized for life in water and exhibit a suite of derived characters, including the retention of a lateral line system, which can be used for detecting movement of prey, and the complete loss of the tongue. Previously, modes of feeding in frogs have been defined by the pattern of tongue protraction (Nishikawa, 2000). Because pipid frogs have no tongue, they cannot feed like other frogs and thus must employ alternative means of prey capture.

Feeding in water poses substantially different challenges than capturing prey in air because water has very different physical properties [~800× more dense and ~50× more viscous (Vogel, 1988)]. Movements that are effective for a terrestrial animal may not function well in an aquatic environment. Rather than trying to overcome the resistance of the water, aquatic organisms often use hydrodynamic properties to their advantage. A prey item can be entrained in a bolus of water and moved by the flow of water into the predator's mouth. This brief flow of water is generated by an explosive expansion of the buccopharyngeal cavity, creating a transient drop in pressure inside the buccopharyngeal cavity (Lauder, 1980). Water then moves toward the area of low pressure, carrying the prey item along with it (Müller et al., 1982). When a flow toward the buccopharyngeal cavity is generated (relative to an earth-bound frame of reference), this type of prey capture is referred to as inertial suction feeding (van Damme and Aerts, 1997). Alternatively, the prey could simply be engulfed by the predator so long as there is no resulting forward movement of the water, which would only push the prey further away (Müller and Osse, 1984). In fishes, the opercular opening permits unidirectional flow thereby preventing a fluid build-up and resulting bow wave. In animals without opercula, the expanding buccopharyngeal cavity can act as a reservoir for the engulfed water, compensating for the forward movement of the predator (i.e. compensatory suction) (van Damme and Aerts, 1997). Inertial suction is often identified by the movement of the prey toward the predator. Inertial and compensatory suction can each be used alone but are often used in combination to varying degrees (Ferry-Graham and Wainwright, 2003).

Although inertial suction feeding is the most common method of aquatic prey capture, not all aquatic-feeding vertebrates use inertial suction feeding. For example, when the eastern box turtle, Terrapene carolina, feeds in water, it overtakes its prey by rapid forward extension of its neck and large expansion of the buccopharyngeal cavity (Summers et al., 1998). The prey are not sucked toward the turtle (i.e. compensatory suction is used). The aquatic-feeding garter snakes Thamnophis couchii and Thamnophis rufipunctatus capture prey using fast forward strikes and exhibit no expansion of the buccopharyngeal cavity but do not appear to create a bow wave (Alfaro, 2002). These snakes may benefit from head morphology that creates little drag and no bow wave, as well as drag-reducing kinematics during prey capture.

There are few published studies of the feeding mechanism in pipid frogs, and these have produced conflicting conclusions. Feeding in Hymenochirus boettgeri was first described by Sokol (Sokol, 1969). These small pipids were said to draw food into their mouths using a water current produced by hyoid depression and retraction. The same mechanism of prey capture was described for the genera Xenopus and Pipa (Sokol, 1969), and later ethological studies of X. laevis and P. pipa supported his observation that these frogs created a ‘partial vacuum’ that could draw prey into the frog's mouth (Avila and Frye, 1977; Avila and Frye, 1978). However, more recent studies of pipid frogs concluded that only the genus Hymenochirus used inertial suction to feed underwater, making this species ‘unique among anurans’ (Dean, 2003). Xenopus and Pipa were said to rely on forearm scooping for prey capture (O'Reilly, 1998; O'Reilly et al., 2002). However, given that the natural position of the forelimbs obscures the view of the mouth, it is likely that the lack of consensus regarding the mode of prey capture in pipid frogs has been a result of the difficulty in making behavioral observations and the lack of a direct measurement for inertial suction feeding.

Various methods have been employed in the study of inertial suction feeding. Researchers have used ratios of distances (RSI) moved by predator and prey (Norton and Brainerd, 1993), measurements of pressure changes (Nemeth, 1997; Svänback et al., 2002), flow visualization at the mouth (Holzman et al., 2008a; van Leeuwen, 1984) and force measurements on the prey themselves (Holzman et al., 2008b; Holzman et al., 2007). There has been much discussion of the best performance measure of an inertial suction feeder but without general consensus (Carroll et al., 2004; Ferry-Graham and Wainwright, 2003; Svänback et al., 2002; van Wassenbergh et al., 2006; Wainwright et al., 2007). The complications arise from the current acceptance that inertial suction does not adhere to steady-state hydrodynamics, is influenced by both behavior and morphology, and is therefore complex and unpredictable (Wainwright et al., 2007). We have used multiple indicators of inertial suction feeding that give us a direct indication of presence and type of suction feeding used in pipid frogs.

Our determination of inertial suction uses real-time measurement of buccopharyngeal pressure in combination with multi-view behavioral kinematics. Pressure measurements provide evidence of the minimum requirement for inertial suction feeding, and observations of prey movement in the direction of the predator is the most direct measure of inertial suction feeding. It has been shown that there is a direct link between the pressure generated within the buccopharyngeal cavity and fluid flow at the mouth opening (Higham et al., 2006). Without a minimum amount of pressure, the bow wave generated by forward movement of the predator would push the prey away. Flow towards the predator can be generated when the subambient pressure is greater than the minimum needed to overcome the bow wave. Therefore, although subambient pressure does not necessarily result in inertial flow, it is certainly a minimum requirement (van Wassenbergh et al., 2006). Observations of prey movement provide the most direct measure of inertial suction feeding because the combined effects of pressure, flow, behavior and morphology are reflected on the prey itself (Wainwright et al., 2001). When pressure measurements are coupled with observations of prey movement towards the mouth and the lack of the use of hands to move the prey into the mouth, it can then be concluded that prey capture was the result of flow created by inertial suction. We used this combination of techniques to determine the presence and degree to which pipid frogs use inertial suction during prey capture.

Four genera of pipids were used in this study. Xenopus laevis Daudin (African clawed frog), Pipa pipa Linné (Surinam toad) and Hymenochirus boettgeri Noble (dwarf African clawed frog) were obtained from commercial suppliers (Xenopus Express, Brooksville, FL, USA). Pseudhymenochirus merlini Chabanaud (Merlin's frog) were obtained from T. Papenfuss (UC Berkeley, CA, USA). All frogs were adults. The size ranges (snout-to-vent length, SVL) for each species were as follows: H. boettgeri 30.5–32.1 mm (N=3); P. merlini 36.5–43.7 mm (N=3); X. laevis 79.3–85.5 mm (N=3); P. pipa 94.0–143.4 mm (N=4). Frogs were housed in aquaria with water temperature maintained at ~23°C using aquarium heaters, and under a natural light cycle. Frogs were maintained on commercial pipid food (Xenopus Express or Purina, Flagstaff, AZ, USA), earthworms (Eisenia fetida) and bloodworms (Chironomus sp.). All methods were approved by the NAU IACUC Protocol #03-085.

Feeding behavior was recorded at 250 frames s⁻¹ using a digital imaging system (Redlake™ Motionscope PCI, San Diego, CA, USA). Illumination was provided by synchronized infrared LED strobes and/or multiple fiber optic lights. Feeding sequences were imaged in various combinations of two views (lateral, plus frontal or ventral) using a mirror placed at a 45 deg. angle under the filming tank (ventral view) or at one end (frontal view). Grids were placed in each field of view to ascertain scaling and aspect ratio. To minimize parallax, the distance between the frog and the background grid was kept to a minimum, and the distance of the camera to the frog was many times greater than the length of the feeding field. Because of the large differences in size between species, two types of prey were used. The large frogs, X. laevis and P. pipa, were fed earthworms and the small frogs, H. boettgeri and P. merlini, were fed bloodworms. Because both prey items were non-elusive and similar in shape and movement, they were expected to elicit similar feeding responses. Feeding events were analyzed using Didge 2.02 image analysis software (written by A. Cullum, 1999).

Buccopharyngeal pressure was recorded during feeding using a pressure transducer catheter (0.46 mm diameter, Millar™ Mikro-Tip SP671, Houston, TX, USA) inserted into a guide cannula. A trigger signal synchronized the imaging and pressure systems. A Millar™ TCB-600 pressure control unit was used to interface the pressure transducer with a SonoMetrics™ DC amplifier (London, ON, Canada), and SonoView software was used to capture unfiltered signals at 1000 Hz.

For implantation of guide cannulae, frogs were anaesthetized by immersion in a buffered solution of 0.1% tricaine methanesulphonate (MS-222, Finquel^®, Argent Chemical Laboratories, Redmond, WA, USA). A 17-gauge hypodermic needle was used to pierce the posterior end of the buccopharyngeal cavity and lateral body wall musculature, just posterior to the scapula. A polyethylene cannula (PE50, 0.58 mm inner diameter/0.965 mm outer diameter) was then threaded through the needle and the needle was removed. A plastic gasket was fitted on the internal end of the cannula and a terminal flange was formed by heating and flattening the cannula. The cannula was pulled such that the gasket and flange were flush with the internal surface of the buccopharyngeal cavity. An external rubber gasket flush with the frog's skin prevented the cannula from being pulled inside the buccopharyngeal cavity. Animals were given at least 24 h to recover from surgery before recordings were attempted. Immediately prior to feeding trials, the pressure transducer was threaded through the guide cannula so that the tip of the pressure sensor was flush with the wall of the buccopharyngeal cavity. A plug of window caulk sealed the external end of the cannula and held the catheter in place.

The following variables were calculated: (1) peak subambient pressure – the maximum change between buccopharyngeal and ambient pressure, (2) rate of pressure drop – the difference in pressure between the buccopharyngeal cavity and ambient, measured for each data point at 1000 Hz, averaged from when buccopharyngeal pressure began to fall below ambient until maximum subambient pressure was reached, (3) prey distance at mouth opening – the distance between the closest point on the prey item and the tip of the frog's upper jaw measured at the beginning of mouth opening, (4) distance at first prey movement – the distance between the closest point on the prey item and the tip of the frog's upper jaw measured in the frame prior to the first prey movement, and (5) distance of prey movement toward frog – the distance traveled by the prey item toward the frog during the time when the frog begins to open its mouth until the prey is within the mouth. Peak subambient pressure and the rate of pressure drop were calculated for two categories of prey capture. If the prey was captured without contacting the hands, the behavior was categorized as suction (inertial or compensatory). If any prey–hand contact was observed, the behavior was considered to be sweeping. The three distance measurements were calculated for suction feeding trials only. All distances were measured with respect to an earth-bound frame of reference. Means, standard error of means (s.e.m.) and sample sizes for each variable are listed in Table 1.

Unpaired t-tests were performed on values of the peak subambient pressure to determine if the experimental values differed from the expected values. The mean of all trials for each individual was used. The null hypothesis was that the peak subambient pressure produced by pipid frogs during prey capture was no different than ambient (ambient pressure=0). Kruskal–Wallis tests were used to assess potential differences among species. Because of small sample sizes (N<5), critical values (α=0.05) were obtained from the exact distribution (Kruskal and Wallis, 1952). Multiple comparisons were then made with Tukey–Kramer HSD post-hoc tests to determine how species differed from each other. Analyses were performed using JMP 8.0 statistical software (SAS Institute, Inc., Cary, NC, USA).

Feeding began with the prey within one body length (BL) in front of the frog. A lunge was often initiated before mouth opening, powered by extension of the hind limbs. Mouth opening began when the frog was less than 20% of its BL away from the prey item (P. pipa 16% BL, X. laevis 9% BL, H. boettgeri 14% BL, P. merlini 6% BL). The differences in absolute distances (Table 1) were found to be significant (P=0.0167) with P. pipa initiating mouth opening at a greater distance from the prey than H. boettgeri and P. merlini. Gape was increased by depressing and bending the mandible. Movement of the prey toward the frog (in an earth-bound frame of reference) without the use of the hands (inertial suction feeding) was observed in each species of pipid frog. Initial movement of the prey toward the frog was first observed during the frog's approach to the prey, at a distance as far as 11% of the frog's BL (P. pipa 11% BL, X. laevis 2% BL, H. boettgeri 6% BL, P. merlini 6% BL). The first movement of the prey occurred at a significantly greater absolute distance in P. pipa than in H. boettgeri and P. merlini (P=0.1151; Table 1). The prey item was sucked a significantly greater absolute distance in P. pipa than in the other three species (P=0.0359; P. pipa 10% BL, X. laevis 3% BL, H. boettgeri 5% BL, P. merlini 4% BL; Table 1).

The most notable difference among the species of pipids was the use of forelimbs during prey capture. Two of the species, H. boettgeri and P. merlini, were never observed using forelimbs for prey capture. During the lunge, H. boettgeri and P. merlini adducted their forelimbs posteriorly, giving the frog a fusiform shape and probably contributing to the power of the lunge (Figs 1 and 2). Once the frog was close to the prey, the prey was sucked into the frog's mouth by a current of water apparently generated by rapid buccopharyngeal expansion. Immediately following engulfment of the prey, the fore- and hindlimbs were rotated anteriorly, halting the forward movement of the frog.

Pipa pipa and X. laevis exhibited a different pattern of movement during prey capture (Figs 3 and 4). The forelimbs of P. pipa and X. laevis were often used to grasp and shove the prey into their mouths. Before prey capture, the forelimbs were held in a forward flexed position, so that each manus (hand) was positioned in front of the head. During prey capture, the forelimbs were extended and drawn toward the mouth. If the prey item was encountered during this sweeping movement, it was grabbed or pushed into the frog's mouth. If the prey item was already within the area described by the forelimbs at rest, it was either sucked into the mouth without any use of the limbs or was simply grasped between the frog's jaws. Pipa pipa exhibited a high degree of manual dexterity compared with other pipids. If the prey item was first encountered with the manus, it was usually grasped between the digits before being moved into the mouth.

All species of pipids showed a marked drop in pressure inside the buccopharyngeal cavity during prey capture. The buccopharyngeal pressure began to drop as the mouth began to open and reached its peak about midway in the gape cycle (Fig. 6). The prey began to move toward the frog before the minimum subambient was reached. Buccopharyngeal pressure was still below that of ambient at the time of mouth closing but continued to increase rapidly until equilibrium was reached. The magnitude of the subambient pressure was significantly different from that of ambient for Hymenochirus (P=0.0349) and for Pipa during all feeding events combined (P=0.0348) and for suction feeding alone (P=0.0442; Table 1). There were no significant differences among species in the magnitude of subambient pressure nor in the rate of change in pressure (Table 1).

Surprisingly little is known about the feeding behavior of the aquatic pipid frogs, even though X. laevis is a model organism for biological research and is widely used (Cannatella and de Sa, 1993). The exception is the smallest pipid H. boettgeri for which it has been established that inertial suction is the sole method of prey capture (Dean, 2003; Sokol, 1969). It has been suggested that X. laevis is also capable of producing inertial suction and uses this method of prey capture a significant proportion of the time it feeds (Avila and Frye, 1977). Perhaps because these data were purely observational, this conclusion has not been readily accepted (Cannatella and Trueb, 1988; O'Reilly et al., 2002). The reluctance to accept inertial suction feeding in X. laevis may be the result of the ease of observation of the more obvious sweeping behavior. Feeding in P. pipa has been reported anecdotally but not studied (Sokol, 1969). This is the first report on the feeding behavior of P. merlini because of its rarity outside its limited natural range.

All four species of pipids exhibited a sharp drop in buccopharyngeal pressure that coincided with the start of mouth opening (Fig. 6). The creation of subambient pressure is a minimum physical requirement for inertial suction, but is not necessarily an indication of inertial suction feeding because it does not demonstrate an effect on the prey. To feed using inertial suction, the magnitude of the pressure drop must be sufficiently large to create a flow that will entrain the prey outside of the mouth. Most vertebrates that employ inertial suction for prey capture create greater drops in buccopharyngeal pressure than those found in pipid frogs (Fig. 5). However, the magnitude of the subambient pressure in pipids is similar to that found in aquatic salamanders (Lauder and Shaffer, 1986) (Fig. 5). Pipid frogs also tend to feed on prey that are generally not elusive (primarily invertebrates), unlike many teleost fish that often feed on elusive prey (De Bruyn et al., 1996; Kazadi et al., 1986; Measey, 1998; Measey and Royero, 2005; Simmonds, 1985). Because these invertebrate prey are not elusive and are close to neutrally buoyant, they are likely to be entrained in a relatively weak flow. However, even with the small subambient buccopharyngeal pressures that are generated by pipid frogs, enough flow is created to capture such elusive prey as small fish. Xenopus laevis has been found to consume fish in the wild, although fish are not a major portion of the diet of pipid frogs (McCoid and Fritts, 1980). Analysis of gut contents from exotic X. laevis living in the Santa Clara River estuary in California found fish in the guts of three out of six frogs examined (Lafferty and Page, 1997). One of these frogs contained a fish that was 54% of its SVL (three other fish had also been consumed by this individual). Small fish are easily captured by X. laevis (Measey, 1998) as well as P. pipa (C.A.C., personal observation). Thus, the generation of small subambient buccopharyngeal pressures is more than sufficient for the capture of the typical prey of pipid frogs using inertial suction feeding.

In the pipids, there was visible movement of the prey item toward the frog even before the greatest subambient pressure was reached. The drop in buccopharyngeal pressure occurred during every feeding event, regardless of whether the forelimbs were used. After prey capture there were subsequent drops in pressure during prey manipulation; these were usually smaller in magnitude than the initial peak during prey capture. Like many other aquatic vertebrates (Gillis and Lauder, 1994), pipid frogs use suction for intraoral prey transport.

Hymenochirus boettgeri and P. merlini demonstrated similar feeding kinematics. The feeding bout began with a lunge, during which both the fore- and hindlimbs were used to power the movement forward. The lunge is a behavior that is basal to anurans (Deban et al., 2001). Although the distance lunged varied, the synchronized movement of all four limbs always occurred. This behavior was first described by Sokol (Sokol, 1969) for H. boettgeri and has since been analyzed quantitatively (Dean, 2003). When the frog had almost made contact with the prey, the mouth began to open followed by a depression of the hyoid and the prey was then sucked into the frog's mouth. During aquatic feeding, the lunge may focus the flow in front of the frog and give a more elongated shape to the bolus of water being moved (Higham et al., 2005). The increase in gape was accomplished by both a lowering of the mandibles as well as mandibular bending (Figs 2 and 3). Mandibular bending may contribute to occlusion of the lateral gape creating a more rounded gape (Deban et al., 2001). A circular gape will orient the flow more in front of the mouth and affect prey movement from a greater distance (Lauder, 1985).

Unlike H. boettgeri and P. merlini, feeding in X. laevis was often characterized by forelimb movement. During prey capture, X. laevis used their forelimbs to ‘sweep’ a small area anterior to their mouth. If a potential prey item was encountered, it was pushed with the hands into the frog's mouth. If the prey item was already within the area described by the forelimbs, it would be sucked into the frog's mouth without actually contacting the limbs. Usually, prey capture resulted from a combination of these behaviors. Xenopus laevis would sweep with its forelimbs, bringing the prey into close proximity to its mouth where it would then be captured by suction. After the first successful prey capture, active searching for additional prey often began and the sweeping behavior was used repeatedly. The switch to an active foraging mode was also noted by McCallum (McCallum, 1997). The combination of behaviors makes sense in light of both predicted and observed hydrodynamic properties. Inertial suction results in fluid movement in front of the mouth. However, fluid speed decreases exponentially over a very short distance (Day et al., 2005; Ferry-Graham and Wainwright, 2003; Müller et al., 1982). The sweeping behavior in X. laevis expands the distance over which prey capture can occur, which may be especially useful when capturing larger prey.

Although feeding behavior in X. laevis had been previously described qualitatively, there was no consensus on its mode of feeding. Some authors reported the exclusive use of forelimbs to ‘shovel’ food into its mouth (McCallum, 1997; O'Reilly et al., 2002), while others reported a ‘water current’ that drew prey into the mouth (Avila and Frye, 1977; Sokol, 1969). These conflicting results were likely to be a consequence of two factors. First, the forelimbs of X. laevis are held in front of the head and the elbows are flexed toward the mouth when searching for prey and during feeding. Because this is a frequent and easily observed behavior, it was simple to conclude that it was the only feeding behavior. The effects of inertial suction are more subtle. Also, the position of the forelimbs obscures the lateral view of the mouth, the most common angle for observation. These two factors resulted in conflicting conclusions about the feeding mode in X. laevis. Our combined approach of direct measurement of buccopharyngeal pressure during prey capture with multiple views for kinematic analyses, demonstrated that, like H. boettgeri, X. laevis is also capable of inertial suction feeding. Feeding in P. pipa was most similar to X. laevis. These frogs

were found to use a combination of sweeping forelimb movements and inertial suction during prey capture. If a prey item was within the sweep area of the forelimbs, it was transported to the mouth by inertial suction. If a prey item was farther away or was large, the forelimbs were used to grasp and bring the prey into the frog's mouth. Perhaps the most notable characteristics of feeding in P. pipa were the dexterity and control of its movements. Unlike H. boettgeri and P. merlini, which have a stereotyped pattern of movement during prey capture, feeding in P. pipa appears to be highly modulated. There is a high level of dexterity in the digits, which are able to grasp prey. One frog was able to grab a worm off its back and pull it into its mouth. Grasping is a behavior thought to be limited to arboreal anuran groups that use it for climbing (Gray et al., 1997). It now appears that grasping has also evolved independently in the Pipidae, as seen in X. laevis and especially in P. pipa. Movement of the mandibles of P. pipa was also modulated during capture and manipulation of prey. Pipa pipa was able to move each mandible independently, often opening only one lateral half. In doing so, the frogs appeared to be directing the flow into the side of the mouth. During manipulation of prey, asymmetrical movements appear to limit the gape size and flow out of the mouth. This high degree of control was not observed in the other pipid frogs.

Through the use of kinematic analyses and direct measurements of pressure in the buccopharyngeal cavity, the use of inertial suction was found to be an integral part of prey capture in all four species of pipid frogs examined in this study (X. laevis, P. pipa, H. boettgeri, P. merlini). There is variation among species in the extent to which inertial suction feeding is used to capture prey. Hymenochirus boettgeri and P. merlini feed using inertial suction exclusively; X. laevis and P. pipa can use inertial suction feeding alone, forelimb scooping alone, or a combination of inertial suction and forelimb scooping for prey capture.

We would like to thank A. P. Summers for assistance in developing the technique for measuring pressure in frogs, T. Papenfuss for sharing his frogs, and A. C. Gibb for many discussions about suction feeding. The paper was greatly improved by comments from members of the Gibb and Nishikawa labs, M. L. Crump and J. O'Reilly. P. C. Wainwright and an anonymous reviewer provided valuable and much appreciated feedback on the manuscript. This research was supported by the NIH IMSD Program (R25-GM56931), an NSF grant awarded to K.C.N. (IBN-0240349), and a Sigma Xi Grant to C.A.C. Deposited in PMC for release after 12 months.