Launches, squiggles and pounces, oh my! The water–land transition in mangrove rivulus (Kryptolebias marmoratus)

The capacity to transition from aquatic to terrestrial environments is most easily recognized when accompanied by morphological specializations associated with weight bearing in a gravity-dominated environment; e.g. fins that adopt a limb-like structure. However, such anatomical changes are not always required for transient exploitation of the terrestrial habitat. The mangrove rivulus Kryptolebias marmoratus [formerly Rivulus marmoratus (Poey 1880)] is an amphibious fish native to mangrove swamps ranging from central Florida to northern Brazil (Davis et al., 1990). These small, fusiform teleosts of the order Cyprinodontiformes are notable for their ability to fully emerse from the water when conditions are poor (Abel et al., 1987; Taylor, 2000). Tidal fluctuations, elevated hydrogen sulfide levels and anoxic water conditions are common in upland mangrove environments (Harrington, 1961). Mangrove rivulus respond to high sulfide levels and low oxygen levels with a threshold-mediated emersion behavior (Abel et al., 1987; Regan et al., 2011). Taylor (Taylor, 2000) has also tied emersion events to intraspecific competition. Kryptolebias marmoratus have the ability to respire cutaneously, a trait well suited for their habitat and emersion behaviors (Grizzle and Thiyagarajah, 1987). Abel et al. (Abel et al., 1987) showed that K. marmoratus can survive out of the water for more than 30 days in moist leaf litter, and Taylor (Taylor, 1990) witnessed terrestrial survivability of 66 days in an artificial muddy environment. Kryptolebias marmoratus have recently been found concentrated inside of logs far removed from the water, with anywhere from a dozen to 100 individuals in a single log (Taylor et al., 2008). The discovery of K. marmoratus in this surprising environment points to another adaptive trait: the ability to locomote on land.

Cursory descriptions of terrestrial motion such as ‘burrowing’ in leaf litter, ‘slithering’ on land and using an ‘S-shaped posture’ when jumping were described by Huehner et al. (Huehner et al., 1985). Taylor (Taylor, 1992) also employed adjectives with snake-like connotations in his descriptions, using the term ‘serpentine’ to characterize the motions of K. marmoratus. In a laboratory study, Huehner et al. (Huehner et al., 1985) used the terms ‘slithering’ and ‘jumping’ to describe the capture of termites on land and the fish's subsequent return to the water. A natural history study (Taylor, 1992) found terrestrial food items in the stomachs of fish that were sampled out of the water, providing further evidence that K. marmoratus use directed terrestrial locomotion to exploit various land resources in the wild.

How, then, are these fish able to move from water to land? Amphibiousness is not a new evolutionary trait in modern fishes. Proto-tetrapods such as Acanthostega most likely moved terrestrially by dragging their bodies across the substrate with modified pectoral fins in the late Devonian period (Carroll and Holmes, 2008). Anatomical modifications to facilitate terrestrial movement are common in amphibious fish. Extant teleosts such as the mudskipper (Periopthalmus koelreuteri), the clingfish (Sicyases sanguineus) and the climbing perch (Anabas spp.) actively direct their terrestrial movement, aided by morphological adaptations to their gill covers or paired fins (Harris, 1960; Ebeling et al., 1970; Davenport and Abdul Martin, 1990). In contrast, quasi-amphibious fish such as leaping blennies (Hsieh, 2010) and K. marmoratus have no obvious phenotypic modifications for locomotion on land. Here we describe the kinematic and behavioral characteristics that allow K. marmoratus to transition successfully from water to land in pursuit of terrestrial prey.

Twelve specimens of K. marmoratus were used, representing a mix of wild-caught (Florida Keys, FL, USA, and Belize) and laboratory-reared individuals. Selected K. marmoratus had a mean (±s.e.m.) total length (TL; tip of the snout to end of the tail fin) of 4.144±0.438 cm. Five of these specimens were housed separately in 10 liter tanks with brackish water (25 ppt salinity), maintained at 24±1°C with a 12 h:12 h light:dark photoperiod, and fed ‘pinhead’ crickets once a day. These fish were conditioned to traverse a barrier made of Permoplast modeling clay (AMACO, Indianapolis, IN, USA), flush with the water surface, in the middle of the tank before feeding on the opposite side of the tank. Conditioning lasted 3 to 4 weeks, and involved gradually increasing the barrier length until the fish had to cross a terrestrial environment before feeding. The ultimate barrier was sloped at 15 deg and plateaued above the water line (Fig. 1). The slope angle at the water–land interface of red mangrove swamps in the Florida Keys, FL, USA, and Lighthouse Reef Atoll, Belize, was measured to determine an ecologically relevant slope (range: 8–16 deg; B.M.P., unpublished); the slope used experimentally was within the range of natural slopes.

Seven of the 12 specimens were used in trials to quantify the squiggle behavior (defined below). Fish were housed in 1 liter plastic containers under the same conditions as above, and fed 1.5 ml of Artemia nauplii solution per day. Squiggle trials were captured on a flat, rectangular surface (22.5×34.5 cm) made of the same Permoplast clay, slightly moistened with water to simulate a transitional microenvironment from water to land. One specimen was filmed on wetted bench-liner paper at a different facility. Similar trends were observed for the squiggle behavior on both flat surfaces and the sloped barrier. All procedures were approved by the Wake Forest University IACUC (protocol no. A11-134).

All transitional behaviors observed on the sloped barrier were filmed in the same tank in which the fish were housed. Room temperature remained at 24±1°C. Data collection took place between 10:00 and 16:00 h, during the fish's daily feeding schedule. Two Fastec high-speed video cameras (Fastec Imaging, San Diego, CA, USA) were used to record all events; first at 250 frames s⁻¹ illuminated by two CN-126 LED lights (CowboyStudio, Allen, TX, USA), and later at 125 frames s⁻¹ under ambient lighting. No differences in behavior were noted between the two light conditions. One camera was suspended above the tank at a 15 deg angle from the horizontal to capture the dorsal aspect of motion (this camera was parallel to the slope of the clay barrier). The second camera was oriented laterally to the tank (Fig. 1). Immobilized crickets were placed in the water on the right side of the tank for transitional trials and placed on the barrier within 4 cm of the water's edge for feeding trials. Events were considered successful if the fish emersed at least half of its body from the water. After traversing the barrier, fish were returned to the left side of the tank using a dip net. All recorded behaviors occurred spontaneously.

Additional squiggle trials were recorded by taking fish out of their 1 liter containers and placing them on the wetted flat clay (or wetted bench liner). A Fastec camera was suspended parallel to the clay's surface, and recorded motion from a dorsal view at 250 frames s⁻¹. No extra artificial lighting was necessary. Only squiggle behaviors of at least one wavelength were analyzed.

High-speed videos were split into sequential BMP images and imported into the program Didge (A. Cullum, Creighton University, Omaha, NE, USA). Images were calibrated for distance using an object of known size placed in the video field. Two-dimensional midline coordinates (x,y) of the squiggle transitional behavior were digitized frame-by-frame using the ‘segment’ function of Didge (Pace and Gibb, 2011) with 20 segments per total length of each fish, starting at the tip of the snout and ending at the tip of the tail. These coordinates were then imported into a spreadsheet program. Trials that exhibited body contact along the wall of the tanks during the transitional period were excluded from the analysis. Sigma Plot (SYSTAT Software, Chicago, IL, USA) was used to create both a midline trace and a body percentage trace of a representative squiggle. Adobe Illustrator (Adobe Systems, San Jose, CA, USA) was used to create a full-body trace of the squiggle movement. Wave duration and amplitude of the squiggle motion were analyzed from the coordinates. Wave amplitude is defined as the distance that the body of the fish is displaced from a direct line of motion, and is analyzed by drawing a line between the start and end locations of a body point in a single wave and then determining how far away the body point deviates from that line in each frame. Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) was used to create an aggregate illustration of the amplitude of each body point at any given time percentage through a wave. All amplitude data were standardized to body length. The distance ratio of the squiggle behavior was also tabulated (Pace and Gibb, 2011). The distance ratio is the linear distance a body point travels from the beginning to the end of a wave divided by the total, non-linear distance that body point actually travels from frame to frame through the same wave. This ratio hints at the mechanical efficiency of a motion (Pace and Gibb, 2011): larger ratios have less lateral movement per forward distance traveled. Distance ratio and maximum amplitude were evaluated for significant differences between body points in SPSS (IBM Corporation, Armonk, NY, USA) using a one-way ANOVA and Tukey's post hoc test (α=0.05). Homogeneity of equal variances was found in the distance ratio data (P=0.964) and maximum amplitude data (P=0.295) using Levene's test. Videos were analyzed in ImageJ (NIH, http://rsb.info.nih.gov/ij) to determine the take-off velocity, take-off angle and height of the launching mode of locomotion, as well as the distance and velocity of the pounce mode of locomotion.

Kryptolebias marmoratus were observed leaving the water using three distinct modes of locomotion. ‘Launching’ makes up 54% of the transitional behaviors, ‘squiggling’ accounts for 26% and ‘pouncing’ accounts for the remaining 20%. The first mode, launching, involved the fish leaping out of the water (N=5, 1–4 trials per specimen; Fig. 2, see supplementary material Movie 1). Fish oriented themselves immediately below the water's surface, facing their eventual direction of motion, and used their tail fin for rapid propulsion, starting from rest. Prior to launch, the fish curled into an S-bend and rapidly accelerated out of the water (Fig. 3, see supplementary material Movie 2). Its tail continued to oscillate even when completely airborne. Take-off velocity was calculated at 27.5±3.69 (mean ± s.e.m.) body lengths (BL) per second (N=5 fish; 16 total trials), ~1 m s⁻¹. Take-off angle from the water was consistent, averaging 47.6±8.19 deg (range: 32.6–64.8 deg), allowing the fish to reach a mean height of 3.2±0.60 cm (~0.77 BL) above the water. If the fish did not clear the barrier, other behavioral modes were used to reach the water on the other side, such as the squiggle.

The second-most observed transitional behavior was the squiggle, which is characterized by a bending of the head and tail toward one another in an oscillatory motion (N=9, 1–2 trials per fish, 1–3 waves per trial; Fig. 4, see supplementary material Movie 3). In a squiggle, the fish curls into a C shape, splaying the pectoral fin located on the outside of the body curvature over the substrate. The fish then yaws slightly, using the pectoral fin as a pivot point to push and lift its body while it sweeps its tail and head through an S-shaped bend until it reaches the opposite curvature. In the process, the non-splayed pectoral fin arches off the substrate, pivots around the splayed pectoral fin, and plants in a splayed configuration in a near mirror image of the original C shape. This motion, representing a half wave, is then repeated. Multiple waves may occur throughout a squiggle. The distinctive bends of the squiggle behavior are best illustrated by tracing the midline of the fish (Fig. 5). Each tracked point along the body (20, 40, 60, 80 and 100%) behaved in a repeated, traceable spatiotemporal pattern (Fig. 6). The broad ‘figure eight’ motion of the tail and the tight S trace of 20% TL were consistent among specimens.

Maximum amplitude of the squiggle wave-form motion differed significantly between some, but not all, points along the body (F=29.393, P<0.001; Fig. 7). The head and tail had the largest maximum amplitudes during any single oscillation, while the mid-body points had smaller amplitudes. The point at 20% of the total length had the lowest maximum amplitude, which corresponded with the location of pectoral fin insertion; the fish pivot at this point during the squiggle motion. Fig. 8 illustrates the average amplitude of each point along the body at any given time percentage through a wave. At approximately 45% through a wave, body points at 80 and 100% TL had a small, sudden drop in amplitude, which differed from the overall trends of the other body points. At this point in time, the vertebral column bent at a point immediately anterior to 80% TL, reversing the curvature of the segment between 80% TL and the tail. The same overall path of motion was maintained as the tail continued toward the head region. The head, 80% TL and tail body points of the fish had the smallest distance ratio (linear displacement divided by total distance travelled in a single wave) during the squiggle, while the point at 20% TL had the highest distance ratio (F=36.851, P<0.001; Fig. 9). The squiggle movement was rapid, with a wave duration of 0.34±0.17 s.

The third transitional behavior was a prey-capture technique, designated the pounce (N=2, 1–6 trials; Fig. 10, see supplementary material Movie 4). Upon locating a cricket placed on the clay barrier, the fish oriented its body with its head region flush with the water–land interface, resting on its ventral side, with both pectoral fins splayed. The fish then curled its tail toward its body while slightly moving its head in the opposite direction to make an S configuration (Fig. 10). This wind-up motion was followed by the fish's immediate acceleration toward the cricket, sliding across the substrate in order to grasp the prey in its jaws. If successful, the fish returned to the water, reversing its S-shaped behavior and curling its head toward its tail to complete a 180 deg turn. Once immersed, the fish shook its head multiple times and then consumed the prey item. If unsuccessful at capturing the prey item on land, the fish either returned to the water with a similar behavior, reversing the S shape, or continued to the other side of the tank via the squiggle motion. Only a single attempt at prey capture was made while the fish was on land, regardless of making a successful or unsuccessful pounce. Pounce behaviors were only observed for crickets placed no more than 2 cm from the water's edge, with a mean distance of 1.21±0.62 cm. Pounce velocity was relatively slow when compared with launching behavior, with a mean of 5.38±2.42 BL s⁻¹, or 20.9±8.00 cm s⁻¹.

Kryptolebias marmoratus were observed to use three very distinct, consistent behaviors when transitioning from water to land: launches, squiggles and pounces. Launches and squiggles were directed locomotory behaviors onto land, while pounces were used exclusively for terrestrial prey capture. All of these behaviors could be independently elicited without a stimulus, although some were generated with specific conditional variables such as the distance of cricket placement from the water–land interface on the sloped barrier.

Ballistic launches from the water by aquatic vertebrates have been described in numerous species [blacktip sharks (Brunnschweiler, 2005); flying fish (Davenport, 1990); silver arowana (Lowry et al., 2005); sockeye salmon (Lauritzen et al., 2005); Trinidadian guppies (Soares and Bierman, 2013)]. The larger species, including sharks and cetaceans, typically launch into the air after a period of burst swimming (Brunnschweiler, 2005; Davenport, 1990; Hester et al., 1963; Hui, 1989; Lauritzen et al., 2005), while smaller fish accelerate out of the water after at most a few tail strokes from an S-start windup (present study; Lowry et al., 2005; Soares and Bierman, 2013). The launch behavior of K. marmoratus seems suited to rapidly propel the fish across a maximal distance. The mean angle of the launch, 47.6±8.19 deg from the horizontal plane, is close to the angle of optimum projection (45 deg) in a ballistic motion consisting of equal vertical and horizontal forces (Price and Romano, 1998). However, there is substantial deviation in this launch angle both within and among specimens, likely because of slight differences in the initial position of the fish. Drag from air resistance should be minimal in this motion, as the fish is in the air for a very short period of time. Fish sometimes landed in the water on the other side of the tank, but often hit the barrier and slid across the clay; the slope of the ramp might inhibit the ability of the fish to visually predict where it will land before launch.

The squiggle of K. marmoratus is a novel mode of transitional locomotion, yet employs muscle and bone structure common to all teleosts. The movement, involving the splaying of the pectoral fin and yawing motions of the body axis, does not fit into any of the classic lateral undulation, sidewinding or concertina movements of snakes, opposing any description of the squiggle as a ‘serpentine movement’ (Jayne, 1986; Pace and Gibb, 2011; Huehner et al., 1985; Taylor, 1992). During the squiggle, K. marmoratus relies largely on its tail for propulsion while the planted pectoral fin acts as a pivot. Therefore, traction against the substrate plays an important role in movement efficiency. Slipping was periodically observed in the area surrounding the water–land interface of the sloped clay barrier, where the fish could not make enough purchase on the partially submerged substrate to move forward. Many squiggle trials were observed and recorded, but some included wall effects and could only add to a qualitative description of the movement. However, this suggests that the fish actively seeks complex, three-dimensional surfaces to aid its terrestrial locomotion, a possible adaptation to the dense prop roots of its mangrove environment. Distance ratios for each body point support the tail-based propulsion method described above (Fig. 9). The pectoral fin insertion point at 20% TL has the least amount of lateral movement during the forward motion, while ratios for the head, tail and 80% TL of the fish were the lowest, and were not significantly different (Fig. 6). However, the head's lateral displacement appears to be due to both the tail-driven rotation at the pectoral fins and the active pushing of the pectoral fins off the substrate. Comparison to distance ratios for terrestrial movement in the limbless ropefish shows a higher ratio than K. marmoratus in every region but 20%, where distance ratios are similar. These differences indicate that ropefish need a smaller amplitude of oscillation for forward motion. Ropefish lack clear, significant differences in the distance ratios between each body percentage point, which is most likely due to the equality of each body point's contribution to oscillatory motion (Pace and Gibb, 2011). This serves as a contrast to K. marmoratus, where a large disparity is found in distance ratios between the extremes of the fish and the pectoral fin insertion point at 20% TL. Distance ratio is a proxy for the mechanical ‘efficiency’ of a movement, as a lower distance ratio means that less time is spent moving forward (Pace and Gibb, 2011), and more time spent moving laterally.

A single trial was recorded in which the fish squiggled through three microenvironments – aquatic (0–24.9% emersion), transitional (25–74.9% emersion) and terrestrial (75–100% emersion) – in three uninterrupted, sequential waves. Each wave had a trend of increasing oscillatory amplitude as the fish moved from water to land (supplementary material Fig. S1). The squiggle motion of the fish while still submerged (0–25% emersion), included the addition of aquatic tail-based propulsion, and a smaller lateral displacement was observed. Distance ratios were higher in the aquatic zone than the terrestrial microenvironment; however, the transitional zone tended to have the highest distance ratios for each body point (supplementary material Fig. S2). Further kinematic analyses and a larger sample size are needed for each zone to elucidate solid conclusions.

The pounce behavior is also seemingly novel in its execution. The splaying of pectoral fins for stability on the substrate and the ‘line-up’ behavior while still immersed are unique. Mudskippers (Periopthalmus argentilineatus) are well known for their use of pectoral fins during terrestrial locomotion, and the strategy of fin splaying seems to be similar; fins are oriented to maximize contact with the ground (Pace and Gibb, 2009). However, a kinematic mode similar to the pounce has not been described in mudskippers (Pace and Gibb, 2009). Kryptolebias marmoratus often swam to and from the water–land edge for an indefinite amount of time, stopping to line up with their prey and then swimming away multiple times before attempting a pounce. Although the S-bend wind-up behavior for the pounce is similar to the launch wind-up, the fish's body orientation is much more horizontal with respect to the water's surface and the velocity of the motion is much slower. This suggests that the fish are able to modulate the force of their rapid accelerations. Similar S-bend ambush mechanisms have been recorded aquatically for elongate fishes (Porter and Motta, 2004). The narrow range of predator–prey distances observed for the pounce suggests that this mode of behavior is very opportunistic. Kryptolebias marmoratus only attempted one pounce behavior per line-up. If unsuccessful, the fish remained stationary, and eventually either continued to locomote across the clay or returned to the water. There was no second consecutive attempt at prey capture. Upon successful prey capture, the fish always returned to the water, shaking its prey vigorously before attempting to consume it. No incidence of complete terrestrial food consumption was observed. Thus, K. marmoratus may go through a period of starvation when on land, which corresponds with Brockmann's (Brockmann, 1975) observation of emaciation in K. marmoratus found far from sources of water.

The recorded data for the slope trials show that when given an environment conducive to all three modes of locomotion, the launch behavior represents 54% of the fish's transition kinematics, while the pounce and the squiggle made up 20 and 26% of the fish's movements, respectively. These numbers are not entirely accurate descriptors of the observed frequency of the locomotory modes. The frequency of the launching behavior when compared with the other behaviors may be the product of the laboratory environment, and may not necessarily reflect the incidence of behaviors in the natural habitat of K. marmoratus. The squiggle behavior was used somewhat more frequently than the analyzed data reflect; many of these instances involved wall-effects, non-transitioning behavior, or could not be properly filmed and identified and were thus discarded. Because the fish had to travel a set distance to reach their food and encountered no predators during this experiment, it can be postulated that the launch transitional behavior was the most energetically efficient when compared with the squiggle because it covered a longer distance with a single effort. The squiggle seems to be a more cautious approach to locomotion: in their natural habitat, fish would be able to alter the direction of movement in order to find shelter or escape potential predators. The blind landing nature of the launch and inability of the fish to alter its direction once in the air suggests that the launch may be a more dangerous method of transitioning from water to land in a mangrove swamp. We noted that fish were more likely to squiggle on the wetted, flat clay surface in places where water had pooled or beaded up. When on completely dry clay, the fish implemented another, strictly terrestrial mode of jumping, a ‘tail-flip’, in which the fish lifts and curls its head over the center of mass and uses its tail to propel itself away from a stimulus (Gibb et al., 2011; Gibb et al., 2013). As the tail-flip is not a transitional behavior, we did not describe it in this study. It is possible that the squiggle behavior is only used in a narrow range of environments, where there is not enough water to execute an aquatic launch and not enough traction to execute a tail-flip. Squiggle transitional behaviors have been observed in the wild from fish that were captured and returned to their habitat (B.M.P., unpublished).

Though this study is the first published description of the transitional behavior of K. marmoratus, limited evidence suggests that some other fishes may move terrestrially using a similar squiggle method, such as three walking catfish species [Clarias batrachus, Clarias gariepinus and Ictalurus punctatus (Pace et al., 2010)]. Furthermore, Cucherousset et al. (Cucherousset et al., 2012) have reported instances of a pounce-like behavior in the wels catfish (Silurus glanis) when ambushing birds found along the shore. The kinematics of terrestrial motion in such organisms have not yet been quantified; however, a comparative study between K. marmoratus and similarly behaving teleosts would be a useful complement to this study. We did not quantify fully terrestrial behavior in this study. However, K. marmoratus jump using a directed, repeatable motion on land that requires further, in-depth analysis. Huehner et al. (Huehner et al., 1985) described K. marmoratus as active foragers, citing personal observations of fish leaping out of the water to catch moving prey. This was not observed in any of the feeding trials, as almost all pounces were directed at stationary crickets. Further study of prey-capture behavior is needed to determine whether K. marmoratus will actively hunt terrestrial food or is strictly an opportunistic terrestrial feeder.

The ability of K. marmoratus to have multiple directed modes of transitional kinematics is evolutionarily significant. Although K. marmoratus is not a basal teleost, having possibly diverged from its South American congeners as recently as 2 million years ago (Tatarenkov et al., 2009), its amphibious nature and ability to exploit land resources with behavioral modification may serve as a model for earlier fishes. Currently, an increasing pressure to exploit land resources is considered one of the main ‘pushes’ for tetrapod evolution in amphibious fish (Long and Gordon, 2004). Taxa such as K. marmoratus suggest that pressure and eventual ability to venture onto land does not necessarily correlate with limb development. Kryptolebias marmoratus has no obvious morphological adaptations for transition to a terrestrial environment, but is still able to successfully exploit land habitats. Amorphologically similar ancient amphibious fish, though not necessarily an evolutionarily close relative, could also have exploited terrestrial resources with modified behavior, which would not be readily apparent in the fossil record.

We thank Dr Ryan Earley for contributing a portion of the specimens used in this study. We also thank Dr Susan Fahrbach, Dr Miles Silman, members of the Ashley-Ross lab and two anonymous reviewers for their helpful feedback on the manuscript.