Aquatic–terrestrial transitions of feeding systems in vertebrates: a mechanical perspective

It is generally appreciated that characteristics of the environment impose important selective pressure on an animal's build or ‘habitus’ (see Glossary) (Losos, 1990). From a mechanical point of view, interaction forces with, for instance, the substrate, food items and items used in a behavioral context, adaptively shape the movement apparatus (e.g. Drucker and Lauder, 2002). This holds true also for the interactions with the medium – aquatic or aerial – that animals live in. Mechanical properties (density and viscosity) differ considerably between both environments: water is approximately 800 times more dense and 50 times more viscous than air (e.g. Hughes and Brighton, 1967). To the benefit of most aquatic animals, this high density results in considerable hydrostatic lift forces contributing largely to the support of the body and food items against gravity (e.g. Losilevskii and Papastamatiou, 2016). However, when movement is involved, the flipside of the same coin shows up. Animals or objects moving through the water experience considerably higher drag forces, not only because of the higher viscosity, but (as soon as size and/or speed becomes significant) because of the high pressure differential over the body, the magnitude of which is largely determined by the fluid density (Fish, 2006).

These differences in density and viscosity also constrain aquatic and terrestrial feeding in different ways (Herrel et al., 2012). Whereas grasping and transporting food are mostly challenged by (but can also benefit from) gravitational and inertial effects on the food item in the terrestrial environment, it is largely the hydrodynamic forces that determine the feeding strategies and performance in the aquatic environment. Reaching or heading towards a food item on land normally does not affect the item's position and a food particle can ‘easily’ be grasped by hand, tongue or jaws, after which the particle's inertia may be exploited by the animal during further processing and intra-oral transport. An aquatic animal moving towards a food item will inevitably push water forward in front of it (creating a ‘bow wave’), which will in turn exert a hydrodynamic force on the food item or prey (Van Wassenbergh et al., 2009b). The precise effect on the food particle depends on the interplay with all other forces (e.g. gravity, substrate friction, active escape forces, etc.), but the chances are rather high that, without remedy, food particles are simply carried with the flow induced by the approaching animal and might never be caught.

Therefore, ways to avoid this ‘bow wave’ effect are needed, and primary aquatic feeders, such as fishes, can do this by ensuring a backwards flow relative to the approaching body and, ultimately, relative to the buccal cavity. Options for this are essentially threefold. The first option is to enable free-flow through the gaping mouth and the buccal cavity while moving forwards. Food particles in this water volume remain unaffected and can thus be overtaken by the forward-moving buccal cavity (this is known as ‘pure ram’ feeding). The second option represents the other extreme: a backwards inertial flow is actively generated in front of the ‘stationary’ mouth aperture and through the buccal cavity. Particles in this flow, even if neutrally buoyant and passive (Van Wassenbergh and Aerts, 2009), will experience hydrodynamic forces that can transport them towards and through the buccal cavity (i.e. pure inertial suction). As for the ‘bow wave’, the precise effect depends on the interplay with all other forces acting on the food particles (Wainwright and Day, 2007). The third option is any active or passive combination of these first two options. However, regardless of which option is used, sufficiently wide gaping and expansion of the buccopharyngeal and opercular cavities (see Glossary) are required to capture and transport food. In the case of inertial suction, this expansion must be powerful and properly tuned (see Box 1) and the generation of the flow to capture and transport food will thus require considerable muscular effort. Clearly, such functional requirements represent a strong adaptive pressure and head morphology, muscle properties and motor control can be expected to be shaped accordingly (see Box 1).

Aquatic feeding is the ancestral feeding mode in vertebrate groups that gave rise to amphibious and terrestrial forms, but making use of aquatic feeding strategies out of water seems problematic. A ‘ram strategy’ might be useful to some extent to grasp prey (yet hampered by the fact that items are most often physically positioned on a solid substrate instead of ‘floating’ in the air), but intra-oral transport would be more problematic without an accompanying adjustment of the motor program and participation of functional units other than the head. Applying the ‘inertial suction strategy’ for food capture and transport on land seems virtually impossible for an aquatic-optimized feeder: because of the low viscosity and density of the air, fluid forces (friction and pressure differentials) are far too low to carry the prey with the flow, unless the generated flow rates could be increased massively. However, given a fixed capacity for volume expansion (determined by head morphology and dimensions), this would require much faster buccal expansion rates, and muscles adapted for aquatic feeding would need to function far away from their optima for power and efficiency (Hill, 1950).

Nevertheless, about 400 million years ago, some aquatic vertebrates transitioned to the land, and this raises the important question as to how the aquatic feeding morphology and motor strategies of the ‘colonists’ could be used successfully for their initial terrestrial feeding and how they were further modified and optimized to give the sophisticated feeding strategies used by later tetrapods. Mechanical analyses of terrestrial feeding in amphibious fishes (i.e. fishes that regularly leave water), salamanders and turtles may be very revealing in this respect. This Review focuses on aquatic–terrestrial transitions in vertebrates from a mechanical perspective in an evolutionary context. We first give an overview on suction feeding, the mode used by most aquatic vertebrates, and explain the challenges a primarily aquatic vertebrate is faced with when aiming for terrestrial feeding. We introduce fishes that do leave the water to feed on land as well as salamanders and turtles that can switch between aquatic and terrestrial feeding environments. We discuss how experimental approaches using extant organisms can help us understand how early tetrapods have left water, and conclude that the muscular tongue was one of the central evolutionary innovations for successful land invasion of tetrapods.

An animal with the generalized anatomy of a suction-feeding fish (Box 1, Fig. 1) will face several mechanical problems with capturing food on land, as illustrated schematically in Fig. 2. The first problem (problem 1 in Fig. 2) relates to posture and stability on land: because of their laterally flattened body shape, most fish will fall to their side when placed on land. The animal should therefore be able to stabilize at least its head against uncontrolled rolling. Second (problem 2 in Fig. 2), the orientation of the line between the upper and lower jaw tips of a suction feeder will generally be near-vertical. This gape orientation is unfavorable for grabbing small food items lying on the ground. Third (problem 3 in Fig. 2), strong depression of the hyoid (see Glossary), a typical movement by fish that assists in opening the mouth and generating suction (Box 1, Fig. 1), may result in contact between the hyoid and the ground surface. Such a hyoid–ground collision can push the head up, and move the mouth away from food on the ground. Fourth (problem 4 in Fig. 2), the retraction of the pectoral girdle (see Glossary) by the hypaxial muscles (see Glossary) during suction feeding (Box 1, Fig. 1) can interfere with support (and stability) of the body by the pectoral fins or limbs. Sudden backward pulling of the pectoral girdle can cause a sudden, forward ground reaction force on the pectoral girdle. Fifth (problem 5 in Fig. 2), as mentioned above, sucking prey from a distance into the mouth using fluid-dynamic forces will no longer work in air. The animals will be forced to develop an alternative strategy for the initial capture of the food. Finally (problem 6 in Fig. 2), intra-oral fluid-dynamic transport will no longer work without water inside the buccopharyngeal cavity. To overcome each of these six problems, several species of amphibious fishes have developed solutions to enable them to exploit food on land.

Probably the most limited form of a transition from aquatic to terrestrial feeding by a fish is observed in the European catfish (Silurus glanis) (Table 1). These fishes have been reported to swim alongside the gravel beaches in shallow waters where pigeons gather together for drinking and cleaning. They manage to capture the pigeons by ‘beaching’ often more than half of their body outside of the water (Cucherousset et al., 2012). During successful attacks, pigeons that are grabbed between the upper and lower jaws are drawn backwards into the water. Similar beaching behavior is also known for aquatic mammals such as orca (Orcinus orca; Guinet and Bouvier, 1995) and bottlenose dolphins (Tursiops sp.; Sargeant et al., 2005). The head and anterior body of the European catfish is dorso-ventrally flattened, which appears to be sufficient to solve problem 1 (Fig. 2) and maintain a posture that is suitable for biting at birds, and quickly turn and crawl back into the water. However, because problems 2 and 3 (Fig. 2) persist, the catfish's mouth will be relatively high above the ground (hyoid bumping of the ground might occur, although there are no images showing this) with a horizontally pointing surface normal of the mouth opening. This will inevitably limit the range of potential terrestrial prey to those that are relatively large, presumably even to those taller than the height of the head of the beaching catfish.

Some amphibious fish that have a long, eel-like body exhibit a terrestrial feeding behavior in which the posture of the head is controlled in such a way that problems 2, 3 and 4 are also circumvented. They use the ground support and flexibility of their eel-like body to perform an elevation of the trunk and a dorso-ventral flexion of the anterior trunk region. This brings their head into a nose-down posture, which allows the mouth to grab relatively small food items from the ground (Fig. 3, Table 1). This behavior has been described for the eel-catfish Channallabes apus (Van Wassenbergh et al., 2006; Van Wassenbergh, 2013) and for the reedfish Erpethoichthys calabaricus (Van Wassenbergh et al., 2017). This feeding behavior is more versatile than that of the European catfish, as the ground surface can be scanned for smaller food by left and right yawing of the head (with help of the chemotactile barbels of C. apus) (Van Wassenbergh et al., 2006).

A re-orientation of the gape is also observed in another amphibious fish, the largescale foureyes (Anableps anableps). However, it uses a markedly different mechanism. Instead of lifting the trunk up and rotating the head down as in the catfish and reedfish (Fig. 4A–C, Table 1), the largescale foureyes entirely relies on the kinematics of its jaws to pick up food from the ground (Michel et al., 2015a). The tooth-bearing part of the upper jaw, the premaxilla (see Glossary), first performs mainly a forward translation while the lower jaw is depressed (resembling typical upper jaw protrusion) (Fig. 4A,B). Afterwards, the premaxilla can be rotated tip-down towards the food, driven by the retraction of the maxilla (see Glossary) by the A3-jaw adductor muscle division (Fig. 4C). The latter action can be performed seemingly independently from lower jaw adduction by the other jaw adductor muscle divisions, allowing more precise ‘picking’ of food from the bottom compared with other (i.e. non-cyprinodontiform) fishes (Hernandez et al., 2008, 2009; Michel et al., 2015a). Michel et al. (2015a) proposed that the increased amplitude of ventral protrusion (i.e. the continued downward movement of the premaxilla during lower jaw closing; Fig. 4B,C) shown by A. anableps compared with that of non-terrestrial feeding cyprinodontiform fishes could have been an important factor to allow terrestrial feeding in this species. Hyoid bumping (problem 3 in Fig. 2) is not observed as hyoid depression is limited (Fig. 4C) and the hyoid is located high enough above the ground owing to the pointed shape of the head. Whether pectoral girdle support interferes with the terrestrial feeding mechanism of A. anableps (problem 4, Fig. 2) is unknown owing to the lack of kinematic data regarding the pectoral girdle and/or fins. Problem 6 persists as A. anableps must return into the water to swallow the food grabbed between the jaws on land.

Mudskippers (Oxudercinae), which feed on the muddy shores of mangroves (Stebbins and Kalk, 1961; Gordon et al., 1969), also make use of their protrusible upper jaw to capture prey on land. However, they do so in a way clearly different from that of the largescale foureyes (Table 1). Reorientation of the mouth opening (problem 2; Fig. 2) is much less the result of ventral protrusion of the premaxilla, but results from the nose-down rotation of the head (Fig. 4D,E). Although functionally similar to the downward-pointing head with elevated trunk region of the eel-catfish (Fig. 3), the pectoral fins are now used to lift the head and anterior trunk (Sponder and Lauder, 1981; Kawano and Blob, 2013; Michel et al., 2014) (Fig. 4, Table 1). X-ray video images have indicated not only that the lower jaw rotates over a relatively large angle but also that the shape of the lower jaw suggests that the anterior bony part (dentary) is rotated relative to the posterior bony part (angulo-articular) (Michel et al., 2014). This flexible lower jaw is placed against (or close to) the ground in front of the prey, while the maxilla is placed just behind the prey. Thereafter, a quick closing action of both the maxilla and lower jaw ‘scoops’ the food into the mouth (Michel et al., 2014) (Fig. 4E,F).

It was recently discovered that the mudskipper Periophthalmus barbarus use a ‘hydrodynamic tongue’ to capture and transport food on land (Michel et al., 2015b) (Table 1). From the moment the mouth opens, a convex meniscus of water becomes visible at the mouth aperture (Fig. 4D). This water further protrudes out of the mouth (Fig. 4E), and, just before the jaws are placed around the food, the water comes into contact with the food and spreads along the surface surrounding the food (Fig. 4F). While the jaws are closing and the food is engulfed, part of the expelled water is sucked back into the buccopharyngeal cavity. Essential in this process is the pronounced lifting of the hyoid while the mouth is opening – this compresses the mouth cavity and pushes the intra-oral water forward (Fig. 4D,E). Next, a depression of the hyoid (as typical for aquatic suction feeding; Fig. 1) expands the buccopharyngeal cavity and thereby causes water (plus food, and presumably some air) to be sucked back into the mouth and towards the back of the buccopharyngeal cavity. The term ‘hydrodynamic tongue’ has been used previously to describe the different, but more common, intra-oral, flow-driven transport of prey in aquatic fishes (Liem, 1991). In addition to this function, the hydrodynamic tongue of mudskippers also includes an extra-oral component during the initial capture of prey (Michel et al., 2015b).

The intra-oral water used as a hydrodynamic tongue enables the mudskipper to capture, transport and swallow food on land. X-ray videos have revealed that food is transported towards the back of the mouth in a single gape opening–closing cycle, and that more than one food item can be captured, transported and swallowed on land before the fish needs to return to the water (Michel et al., 2015b). Mudskippers are currently the only amphibious fishes known to deal with the transport and swallowing problem on land (problem 6; Fig. 2). To avoid losing water on land, their gill slits are reduced, leaving only a small opening at the dorsolateral sides of the opercular bones (Michel et al., 2016). The typical rotation of the pectoral girdle relative to the rest of the skull during suction feeding (Fig. 1), which would cause the head to be rotated nose-up instead of nose-down if the pectoral girdle is anchored to the ground, is hardly noticeable during terrestrial feeding in the mudskipper (Michel et al., 2015b). Consequently, mudskippers show effective solutions for all of the mechanical problems (Fig. 2) associated with performing terrestrial feeding with an aquatic suction feeding apparatus. Their capacity to transport and swallow food on land implies that mudskippers reflect a higher level of terrestrialization compared with the other amphibious fishes mentioned above.

Solutions to the mechanical problems faced by fish when feeding on land can evolve before an amphibious lifestyle is adopted. Mechanical demands of certain modes of aquatic feeding can result in evolution of a trait that later proves beneficial for the terrestrial capture of prey. Such exaptations (see Glossary) for terrestrial feeding are likely to occur in groups that specialized in benthic feeding. Aquatic benthic feeders of the Gobiidae, the family to which also the mudskippers belong, have a sub-terminal mouth, and show a depression of the hyoid which does not hinder food capture even when the head and body remain close to the substrate (Maie et al., 2009). The problems of reorientation of the mouth opening (problem 2; Fig. 2) and ground-impact of the depressing hyoid (problem 3; Fig. 2) are thus similar for adopting an aquatic, benthic feeding lifestyle. However, most fish that feed on food lying on the substrate in the aquatic environment will tend to incline the entire body to steeper angles, e.g. cichlids (Rupp and Hulsey, 2014), eels (Mehta and Wainwright, 2007a,b), catfish (Van Wassenbergh et al., 2009a) and reedfish (Van Wassenbergh et al., 2017). As these postures are impossible outside water, not all aquatic solutions to benthic feeding facilitate the transition to terrestrial feeding.

In conclusion, amphibious fishes show a multitude of solutions to the mechanical problems they face when feeding out of water. Terrestrial food capture can be successful with or without support from the pectoral fins, and with or without a complex protrusible jaw system. Transport and swallowing, however, remain impossible without water, and only the mudskipper has evolved a solution to perform this action on land. Yet, the terrestrial feeding kinematics of several groups of amphibious fishes (overviewed in Wright and Turko, 2016) remains unstudied. For example, some blennies (Blennoidea) seldom submerge themselves under water, and are therefore described as terrestrial animals (Hsieh, 2010). Trichomycterid catfishes occupy a range of trophic niches, including small, eel-like fishes that live hidden among the leaf litter and mud (Glanapteryginae; de Pinna, 1998). Consequently, further studies will help us to clarify the morphological shifts that allow different fish to forage on land (Ashley-Ross et al., 2013). A better understanding how extant fish manage to feed on land with a feeding system primarily adapted for an aquatic environment will in turn shed light on the evolution of terrestrial feeding mechanisms in early tetrapods.

Today there is little doubt that tetrapods evolved in an aquatic environment (Coates and Clack, 1991; Clack, 2012; Schoch, 2014) and that the first tetrapods lived and foraged in water. For example, the well-known stem tetrapod Acanthostega from the Late Devonian (ca. 360 Mya; Fig. 5) was aquatic throughout its life (Coates and Clack, 1991; Clack and Coates, 1995; Schoch and Witzmann, 2011; Schoch, 2014). Similarly, the stem tetrapod Ichthyostega (Fig. 5), which lived approximately at the same time as Acanthostega, was basically an aquatic animal with internal gills, a lateral line system (see Glossary) and a typical fish-eater dentition (Schoch, 2014), but its axial and limb skeleton might have allowed short bouts of locomotion out of water (Pierce et al., 2012). However, Ichthyostega was not amongst the first tetrapods that stepped out of water. Trace fossils from the early Middle Devonian period of Poland (ca. 400 Mya) that pre-date any tetrapod body fossil by 18 Mya tell us that other as-yet-unknown four-legged vertebrates ventured for short excursions along the shore (Niedźwiedzki et al., 2010). In other words, the first tetrapods were still heavily bound to aquatic realms, and their limbs evolved as an alternative type of strong fins that were exapted for terrestrial locomotion (and solving the problem of stability shown in Fig. 2) in a second step (Clack, 2012; Pierce et al., 2012; Schoch, 2014).

But why did stem tetrapods with a body primarily adapted for aquatic life leave water at all? Many theories were introduced in the last few decades trying to tackle that question (for an overview, see Romer, 1958; McNamara and Selden, 1993; Carroll, 2009; Niedźwiedzki et al., 2010; Clack, 2012; Schoch, 2014). Here, we will focus on the theory most appealing to us: that leaving water might have opened new possibilities to exploit food sources. As suggested by Niedźwiedzki et al. (2010) and Clack (2012), stem tetrapods might have exploited the niche of predating on stranded and trapped fish or invertebrates during low tide in intertidal zones. The trackways from Poland were found along intertidal and lagoonal zones and therefore can well support the hypothesis that the track-makers were heading along the shore in search for stranded and trapped marine animals (Niedźwiedzki et al., 2010; Clack, 2012).

How could prey be captured out of water with a feeding system primarily adapted for aquatic strikes (see mechanical problems shown in Fig. 2)? Similar to terrestrial actinopterygians (see Glossary) (Fig. 1), stem tetrapods still had a fish-like hyobranchial apparatus bearing internal gills (Clack, 2012) and lacked a movable tongue but were able to grasp prey out of water with their jaws. Acanthostega was probably capable of using suction feeding (Neenan et al., 2014; Porro et al., 2015), but, by making deductions from its dentition and skull structure, it also used direct bites, probably to capture larger prey (Markey and Marshall, 2007). Accordingly, it seems likely that, analogous to the fin-to-limb transition, the shift from a high-performance suction feeding mechanism to a firm biting mechanism (jaw prehension), as an exaptation for a terrestrial feeding mode, had already happened in water, and use of this mechanism on land to grasp prey was probably the next functional step. However, biting alone is useless if the gape cannot be directed to the food item. Although mudskippers are an example of how weight-bearing support from the pectoral fins can be used to bring their mouth towards terrestrial food, it is questionable whether the whole-body pivoting observed in mudskippers would have been possible in the considerably larger adult specimens of Late Devonian tetrapodomorphs. Mudskippers seldom surpass 0.2 kg (Khaironizam and Norma-Rashid, 2002), but fossils such as Tiktaalik, Acanthostega and Ichthyostega (all surpassing 0.5 m in length; Fig. 5) probably weighed 10 times more. However, around the time of the origin of the weight-supporting pectoral girdle and appendages, the first necks appeared in tetrapodomorphs (Fig. 5) – the pectoral girdle that previously attached directly to the back of the skull (e.g. red elements in Fig. 1) became disconnected from the skull to form a pectoral joint more posterior to the vertebral column (Daeschler et al., 2006; Shubin et al., 2015). This allowed the head to be moved while the pectoral girdle and appendage were statically supporting the head.

Upon the discovery of the first tetrapodomorph that had a neck, Tiktaalik roseae (Fig. 5), it was first proposed that the neck would allow dorsoventral flexion that could be used during feeding (Daeschler et al., 2006), perhaps in a way analogous to amphibious fishes (Van Wassenbergh et al., 2006). Later, it was proposed that T. roseae would use lateral instead of dorsoventral movements of the head to capture food on land and underwater (Hohn-Schulte et al., 2013). The investigators argued that (1) the head shape of T. roseae is somewhat reminiscent of crocodiles, which use lateral strikes of the head, and (2) the relatively long and narrow T. roseae jaws make it biomechanically more advantageous (less inertial cost and hydrodynamic resistance) to use side-snapping to capture prey in water (Hohn-Schulte et al., 2013). Additionally, it can be argued that a relatively long lower jaw, like those of the Late Devonian Osteolepididae (Fig. 5), might be too large to allow the mouth to be placed over the food, as observed in amphibious fishes (Figs 3 and 4). The smaller jaws of amphibious fishes require less height above the ground to be rotated over a large angle, allowing food to be grabbed from above. Rolling of the head and grabbing food on land with one side of the jaw could overcome this problem.

The ability to adjust the movement pattern of the head in order to effect the function of food prehension on land, such as sideways rolling of the head, could have made it unnecessary to evolve the morphology of the jaws during the transition(s) from water to land in Late Devonian tetrapodomorphs. This idea seems supported by analyses of the morphology and mechanics of the lower jaws in the transition from Devonian tetrapods to aquatic/semi-aquatic Carboniferous species – the changes are negligible (Neenan et al., 2014). The lower jaws of tetrapods only start showing considerable changes much later (Pennsylvanian age), when herbivory evolved within amniotes (see Glossary) (Anderson et al., 2011, 2013). These studies indicate that no different mechanical constraints are imposed on the lower jaw to function effectively during terrestrial feeding compared with aquatic feeding.

Once a food item is grasped by the jaws, it must be transported to the back of the oral cavity for swallowing. Though variable and overlapping, roughly three phases have been suggested both for aquatic and terrestrial gnathostomes (Lauder, 1983; Hiiemae and Crompton, 1985): food first needs to be transported from the front to the back of the oral cavity (oral phase) and to the back of the pharyngeal cavity (pharyngeal phase) for the final transition into the esophagus (swallowing phase). The oral and pharyngeal transport phases are not strictly separated events but rather a continuum, and in the following we will use ‘transport’ or ‘intraoral transport’ when referring to food transport through the oropharyngeal cavity. In water, transport phases can be accomplished by the use of water currents, but how can transport be accomplished on land without the aid of water flow (see mechanical problem 6 in Fig. 2)? Stem tetrapods and actinopterygians feeding on land are faced with the same problem for transport, and, by making deductions from extant analogues (see Glossary), three plausible solutions have been proposed: (i) carrying a gulp of water out to land and using water in the mouth as a hydrodynamic tongue, (ii) dragging the prey back into the water to use suction for intraoral transport and (iii) inertial transport by quick lateral or dorso-ventral head rotations while temporarily releasing the tight grip on the prey.

Option i, carrying a gulp of water to be used as hydrodynamic tongue, is a mechanism found in mudskippers (Fig. 4D–F) (Michel et al., 2015b), and it might well be that some as-yet-unknown stem tetrapods used a convergent mechanism (Michel et al., 2016). Option ii, dragging prey back to the water, might be the easiest way to circumvent the problem of transport as well as swallowing prey captured on land, and is used in a variety of aquatic and amphibious vertebrates that occasionally strike prey on land (Lopez and Lopez, 1985; Miller and Larsen, 1990; Peddemors and Thompson, 1994; Bels et al., 1997; Summers et al., 1998; Werth, 2000; Stayton, 2011; Heiss et al., 2013a, 2015; Michel et al., 2015a; Natchev et al., 2015; Van Wassenbergh et al., 2017). Option iii is found in crocodilians (Cleuren and de Vree, 1992, 2000) and varanid lizards (Smith, 1982, 1986), where food is literally thrown through the mouth (Smith, 1982) by fast dorsal head rotation in concert with coordinated gape opening to accelerate prey backwards (Smith, 1982, 1986; Cleuren and de Vree, 1992, 2000). This form of inertial transport requires a movable neck that allows fast head movements relative to the trunk. In fact, as the shoulder girdle lost its tight connection to the skull during early tetrapod evolution (Fig. 5), the resulting movable neck can be considered one of the key innovations of tetrapods. Accordingly, aside from bringing a gulp of water on land, dragging captured prey back to the water for use in aquatic intra-oral transport mechanisms, it is likely that some stem tetrapods used a simple form of inertial oral transport on their occasional terrestrial excursions in search for food. Note that one group of ray-finned fishes, moray eels, has evolved the exceptional capacity to directly transport food from the jaws to the esophagus by extreme protrusion of the pharyngeal jaws (Mehta and Wainwright, 2007a,b). However, forward displacements of the pharyngeal jaws in fishes are generally confined to the posterior region of the pharynx (Liem and Greenwood, 1981; Sponder and Lauder, 1981; Wainwright, 1989). Yet, pharyngeal jaw transport has been shown to be involved in the mudskipper Periophthalmus when feeding on land (Sponder and Lauder, 1981). Given that denticles were also present on the gill arches in temnospondylians (e.g. Schoch, 2009), it cannot be entirely excluded that pharyngeal teeth movements played a role in the final phases of intra-oral transport in some tetrapods too. Nevertheless, permanent colonization of land was only possible after efficient feeding mechanisms independent of water had evolved.

A fish-like hyobranchial apparatus bearing gills might allow short excursions and occasional feeds on land but prevents any permanent adoption of terrestrial domains. Permanent ‘terrestrialization’ was only possible after reduction of the gills and development of a tool that allowed efficient food capture and intra-oral transport independent of water flow – the tongue. The tongue recruits a large part of the hyobranchial musculoskeletal system and therefore replaces, to a high degree, the suction-feeding apparatus otherwise used for food ingestion in water and the suspension of the gills. The functional consequences of this include a reduced capability to suction feed and the need for another respiratory system. Gill respiration was effectively replaced by lung and cutaneous respiration, and giving up high-performance suction feeding paved the way for further degrees of freedom to remodel the hyobranchial musculo-skeletal system into a lingual system (Heiss et al., 2013b).

However, owing to the poor preservation of soft tissue and cartilage in general and non-mineralized tissue of the hyobranchial system in early tetrapods in particular (Schoch, 2014), it is hard to define when and where the first tongue appeared in the evolution of tetrapods. Inferring from fossil records, we know that stem tetrapods were basically aquatic forms (Clack, 2012), but that stem amphibians and stem amniotes (see Glossary) evolved amphibious lifestyles and gradual transitions from an aquatic, gill-bearing larva to a more-or-less terrestrial adult with reduced gills (Schoch, 2014). While in the dissorophoid (see Glossary) zatracheids (stem amphibians) we have the first evidence for a remodeling of a larval, gill-bearing hyobranchial apparatus to an adult, tongue-supporting hyolingual skeleton, semouriamorphs (stem amniotes) evolved terrestrial forms along a separate line (Witzmann, 2013; Schoch, 2014). It is not clear whether stem amphibians and stem amniotes have evolved a tongue completely independently (Witzmann, 2013), but, if we consider lingual systems in extant tetrapods, it seems likely that the history of the tongue was subject of many parallelisms.

The tetrapod tongue is mainly used as a prehensile organ for the initial uptake of food and to manipulate or transport food within the mouth (Schwenk, 2000a). The tongue can be a highly movable organ moved out of the mouth to grasp food items that are then brought into the mouth by tongue retraction. Such initial food uptake by the tongue requires an adhesive potential of the tongue to overcome the inertia and gravitational forces acting on the food (Bramble and Wake, 1985). Specialized tongue surface microstructures such as papillae (see Glossary), in conjunction with mucus secretions, can enhance the adhesive potential (Bramble and Wake, 1985). The horned frog Ceratophrys has mastered tongue adhesion so comprehensively that, in theory, it is capable of lifting its own body weight by its tongue alone (Kleinteich and Gorb, 2014).

Once food is brought into the mouth, it is directly transported to the esophagus for swallowing, or to the posterior teeth for mechanical reduction, if chewing is applied beforehand (Schwenk, 2000a; Schwenk and Rubega, 2005; Herrel et al., 2012). Swallowing can be a complex affair and mechanisms differ substantially amongst tetrapods, but in most cases tightly coordinated contraction of pharyngeal and lingual muscles widen the esophageal sphincter and push the food bolus into the esophagus (e.g. Smith, 1984, 1992; Bramble and Wake, 1985; Schwenk, 2000a). However, next to the use of the tongue in concert with pharyngeal musculature, some tetrapods have evolved alternative mechanisms to carry food to the esophagus, as summarized in Table 1. Nonetheless, the tongue remains one of the central evolutionary innovations in tetrapods enabling all feeding stages independent of water, but tongue functions can be very different and tongue morphology can vary substantially amongst groups.

In general, a tetrapod tongue is supported by skeletal elements – the hyolingual skeleton – and integrates mucosal, muscle and connective tissue elements. In lissamphibians (see Glossary), most birds and crocodilians, the tongue is a relatively simple mechanical system and consists of a mucosal sheath covering a simple (extrinsic) muscular system that rises from the supporting hyolingual skeleton, jaw or shoulder girdle (Özeti and Wake, 1969; Regal and Gans, 1976; Schwenk, 2000a; Tomlinson, 2000; Wake and Deban, 2000; Putterill and Soley, 2004; Erdoğan and Iwasaki, 2014). Other tetrapods, such as tortoises, squamates and mammals, have large bulged tongues and a muscular system with a complex arrangement of highly integrated internal (intrinsic) fibers oriented longitudinally, transversally, vertically, diagonally or circularly.

The muscular complexity mentioned above accounts for the unusual biomechanical properties of such tongues being able to function as muscular ‘hydrostats’ (see Glossary). In muscular hydrostats, the uncompressible volume of the muscle bodies has the effect that any decrease in one dimension causes a compensatory increase in at least one other dimension (Kier and Smith, 1985; Kier, 2012). For example, if we assume a cylindrical shape of such a system, elongation is achieved by diameter reduction caused by contraction of transverse, vertical or circular running fibers, whereas shortening is effected by an increase in diameter, caused by contraction of longitudinally oriented fibers. More-complex fiber constellation and activation patterns can cause coiling, bending or twisting movements. Accordingly, tongues exhibit a great range of movements that becomes especially evident in mammal and sauropsid tongues, which show motion ranges comparable with those of elephant trunks or squid tentacles (Kier and Smith, 1985; van Leeuwen et al., 2000; Kier, 2012). Muscular hydrostats are highly efficient lingual systems, but relatively simple tongues with marginal or no intrinsic musculature at all can be high-performing tools too, as exemplified by some frogs and salamanders that use elastic energy recoil mechanisms to catapult their sticky tongue pad towards distant prey that has virtually no chance to escape (Regal and Gans, 1976; Deban et al., 1997; Nishikawa, 2000; Wake and Deban, 2000; Deban and Richardson, 2011).

To understand the evolution of a functional tongue, salamanders are of special importance within extant tetrapods as they offer the unique opportunity to track the remodeling from a larval fish-like hyobranchial system bearing gills to a postmetamorphic tongue-bearing hyolingual system (see Fig. 6). Similar to bony fishes, larval salamanders are suction feeders with open gill slits (Lauder and Shaffer, 1985, 1988; Lauder and Reilly, 1988; Reilly and Lauder, 1992). As salamanders metamorphose, the head region undergoes massive changes, and the secondary tongue pad forms just anterior to the primary larval tongue (Opolka et al., 2001; Greven et al., 2013), while the hyobranchial musculoskeletal system gets reduced and rearranged in order to allow movements between what were previously fixed elements of hyoid and branchial arches (Reilly, 1987). In a simplified model, protraction and retraction of the tongue in the generalized postmetamorphic salamander is mainly achieved as the tongue-pad-bearing skeletal element – the basibranchial (in conjunction with hypo- and ceratobranchials) – is protracted and retracted relative to the ceratohyal (see Glossary) by action of the antagonistic subarcualis rectus and rectus cervicis muscles (Deban, 2003) (Fig. 6).

The cellular mechanisms controlling hyobranchial remodeling during metamorphosis in lissamphibians are still poorly understood. Most likely, thyroid-hormone-affected gene regulation accounts for the changes associated with hyobranchial remodeling during metamorphosis (Brown and Cai, 2007; Faunes et al., 2017; Laudet, 2011) and convergently triggered changes in gene regulation might also be the cause of evolutionary hyobranchial remodeling processes. So, similar to the fin-to-limb transition, regulatory changes in specific patterning genes (Shubin et al., 1997) might have led to the hyobranchial–hyolingual transition in tetrapod evolution. Currently, we do not have any knowledge on the form and function of the tongue in the first terrestrial tetrapods, but salamanders teach us that relatively small anatomical rearrangements in the hyobranchial system can lead to significant mechanical changes leading to a functional shift. Accordingly, it is plausible that similar modifications in the hyobranchial system in early crown tetrapods led to the evolutionary innovation of the tongue, which in turn paved the way for a successful invasion of land and the exploitation of terrestrial food sources.

Demands between feeding in water and feeding on land are diverse and, to a large extent, conflicting (Bramble and Wake, 1985; Deban, 2003). Because of the different dynamic behavior of food particles when in water or air, substantially different morphological designs of the feeding apparatus are required (Bramble and Wake, 1985). For example, robust hyobranchial skeletons and associated muscles allow high-performance suction feeding in water, but such a design is useless on land, whereas a slender hyolingual skeleton supporting a muscular tongue is of great use on land but suboptimal for aquatic feeding (Bramble and Wake, 1985; Deban, 2003; Stinson and Deban, 2017). As shown in the previous sections, a feeding system primarily adapted for aquatic (suction) feeding can also be used to catch prey on land. Similarly, remodeling of a gill-bearing hyobranchial to a hyolingual tongue-bearing system might, in many cases, reduce suction feeding performance but does not prevent suction feeding per se. Indeed, we find many examples within salamanders, turtles and actinopterygian fishes that regularly switch between aquatic and terrestrial environments and have found different mechanical solutions to circumvent the trade-offs between efficient food uptake on land and in water.

Merely aquatic salamandrid newts use suction in water and a slightly modified suction strike on land and grasp prey using jaws (Miller and Larsen, 1989, 1990; Heiss et al., 2013a, 2015). A similar, but convergently evolved, strategy is used by semi-aquatic turtles, where suction feeding is performed in water, but their jaws seize prey on land (Fig. 7) (Natchev et al., 2009, 2015; Heiss et al., 2010; Stayton, 2011). Once food is held between the jaws, it is dragged back to the water, or in some cases swallowed on land by backward transport and transition into the esophagus by the tongue (Natchev et al., 2009, 2015). Some salamanders go further, and, while they still use suction feeding in water, they change to lingual prehension on land and catch prey by their protractible tongue (Fig. 7) (Reilly and Lauder, 1989; Lauder and Gillis, 1997; Deban and Wake, 2000; Wake and Deban, 2000; Heiss and De Vylder, 2016). In contrast, turtles are less flexible and only the fully terrestrial tortoises make use of their tongue for initial food uptake on land, but unlike their turtle relatives, extant tortoises have lost any capacity to feed under water (Natchev et al., 2015).

Flexible feeding systems allowing aquatic and terrestrial uptake of food demand a high degree of behavioral plasticity and flexibility, but structurally represent compromises and perform suboptimally in at least one medium (Bramble and Wake, 1985). The only solution to increase performance in both media would be to change the morphology of the feeding apparatus when changing from one to the other feeding medium. In fact, some salamandrid newts are such masters of change – they use suction feeding in water and catch prey by the tongue on land and reversibly change the anatomy of their feeding apparatus to meet the demands of the respective feeding circumstances (i.e. aquatic versus terrestrial feeding) (Heiss et al., 2013a, 2015, 2016, 2017; Van Wassenbergh and Heiss, 2016). Specifically, some salamandrid newts change seasonally between an aquatic and a terrestrial life and as they switch habitat, they also undergo modifications of their feeding apparatus. In the aquatic phase, they grow labial lobes (Matthes, 1934), which are skin folds that restrict the gape opening anteriorly and increase suction feeding performance measured by a suction flow velocity that is increased by up to 30% (Van Wassenbergh and Heiss, 2016). Labial lobes then disappear when the newts leave the water and change to the terrestrial lifestyle. At this point, the formerly smooth tongue grows microscopic slender papillae and the mucus-producing goblet cells increase in number to supply the tongue with more mucus (Heiss et al., 2017). The integration of microscopic slender papillae and mucus has been shown to increase adhesive potential of the amphibian tongue (Kleinteich and Gorb, 2015) and accordingly, the newts increase their potential prey capture success by gaining a sticky tongue pad as they change to their terrestrial lifestyle. As newts seek their home-waters again in the spring, the lingual papillae disappear and the goblet cell number decreases, whereas labial lobes recommence growing again (Matthes, 1934; Heiss et al., 2017).

Even if a feeding apparatus undergoes some changes to account for different demands, it will always represent some kind of compromise, and amphibious species that switch between aquatic and terrestrial feeding usually perform at a lower efficiency than those specialized for one medium. By contrast, salamanders, turtles and actinopterygians have convergently found ways to exploit food sources from two very different environments where prey abundance can vary in time. Considering such analogues in conjunction with fossil records (fossil tracks) it makes it more plausible that the first tetrapods that left the water some 400 Mya were indeed driven by the search for new food sources in intertidal zones.

Transitions from aquatic to terrestrial habitats are well known as pivotal events in vertebrate evolution, and this article synthesizes the mechanical demands of capturing, transport and swallowing of food on land for ancestrally aquatic as well as amphibious species. We have illustrated how several amphibious fishes and tetrapods deal with these different mechanical challenges to feed on land. Such experimental studies on extant organisms allow useful insights and deductions regarding how the first terrestrial tetrapods managed the physical challenge to feed on land on their way towards land invasion. In fact, biomechanical studies on fishes, amphibians and other vertebrates that feed in both environments (e.g. Fig. 7) will remain of central importance to improving our understanding of the functional–morphological trade-offs involved in aquatic–terrestrial transitions. Promising approaches for future work include, amongst others, studying the effect of phenotypic plasticity and flexibility on aquatic–terrestrial feeding transitions. For example, it was shown that the basal extant actinopterygian Polypterus can be raised out of water and that such individuals adopt terrestrial-style kinematics and anatomy of their locomotor system (Standen et al., 2014). Analogous to this, behavior and anatomy of its feeding apparatus might change accordingly when raised out of water. Similarly, representatives of the tetrapod sister group, i.e. extant lungfishes, might be raised on land or at least trained to feed on land, and it will be interesting to know how they change their feeding system to capture, transport and swallow food out of water. Furthermore, paedomorphic or larval salamanders are a promising model to study how a basically aquatic tetrapod with limbs and a hyobranchial system bearing gills manages to feed out of water: a very similar condition first tetrapods were faced with. Using such model organisms in integrative experimental, analytical and computational approaches such as biplanar X-ray recordings, XROMM (X-ray reconstruction of moving morphology), finite element analysis (FEA) or computational fluid dynamics (CFD) offers the opportunity to obtain deeper insights into the mechanics of feeding transitions in extant vertebrates, which in turn helps to explain one of the most crucial steps in our own history: the evolution of terrestriality and successful invasion of land.

We thank A. Herrel, D. Adriaens, K. Michel, B. Vervust, J. Brecko for their help in our research on the biomechanics of feeding in amphibious fishes, N. Natchev, P. Lemell, C. Beisser and J. Weisgram for their contributions on turtle feeding research and David Hattan along with two anonymous reviewers for their valuable comments on the manuscript.