Swimming of larval zebrafish: fin–axis coordination and implications for function and neural control

Locomotion by fishes has traditionally been subdivided into swimming with paired or median fins, or with the body axis and caudal fin (e.g. Webb, 1994). Often, slow swimming is powered by the pectoral fins actuated in phase (synchronously)with one another (Blake,1983). At higher relative speeds, pectoral fin locomotion may be supplemented by intermittent caudal fin movement(Drucker and Jensen, 1996a). Frequently, at high swimming speeds, the pectoral fins are tucked along the sides of the body and axial movement alone generates propulsive thrust(Webb, 1994).

Many species, including adult zebrafish(Thorsen et al., 2004), use axial body bending at all speeds, and the fins to maneuver and stabilize(Webb, 1994). When used for maneuvering, pectoral fins have been shown to alternate out of phase temporally, such that abduction of one fin coincides with adduction of the contralateral fin (Drucker and Lauder,2003). Other species employ their pectoral fins in synchrony as their primary mode of locomotion during steady swimming across a wide range of speeds (Webb, 1973, 1993; Gibb et al., 1994; Drucker and Jensen, 1996a,b; Walker and Westneat, 1997). The morphologically unusual burrfish (Chilomycterus schoepfi)alternates the pectoral fins during swimming, which are active simultaneously with the caudal fin (Arreola and Westneat,1996).

Research in locomotion of larval fishes has focused on axial movements during steady swimming, turning, prey capture(Budick and O'Malley, 2000; Borla et al., 2002), startle behaviors (Batty, 1981; Hale, 1996, 1999; Budick and O'Malley, 2000; Müller and van Leeuwen,2004) and swimming performance(Fisher et al., 2000; Bellwood and Fisher, 2001; Fisher and Bellwood, 2003). Work by Batty (1981), and Müller and van Leeuwen(2004), demonstrated that plaice larvae (Pleuronectes platessa) and zebrafish larvae,respectively, can swim with simultaneous axial and pectoral fin movements. However, the detailed kinematics and role of coordinated pectoral fin and body movements have gone unstudied primarily due to the technical difficulty of visualizing pectoral fins of larvae(Budick and O'Malley,2000).

The combined movement of the limbs and axis during locomotion has been studied in depth in tetrapods. Axial bending is often coordinated with limb rhythms so that a flexion–extension limb cycle corresponds to one cycle of axial bending (Ritter,1992; Ashley-Ross,1994). The limbs within a fore limb or hind limb pair alternate with each other so that, for most of the stride cycle, one side is in its swing phase while the other is in its support phase(Biewener, 2003). There is a short period of overlap when both limbs are on the ground with one limb at the beginning of the support phase and the other at the end during walking. The production of axial movements via standing or traveling waves of bending (Williams et al.,1989; Frolich and Biewener,1992; Ritter,1992; Reilly and Delancey,1997) varies among species, developmental stage and gait. The basic temporal pattern of this locomotor activity involves the integrated activity of central pattern generators (CPGs) in the spinal cord (for reviews,see: Stein, 1978; Grillner, 1981; McClellan, 1996).

Pectoral fin muscles of adult fishes have been studied in many species and include an array of muscles that control fin adduction and abduction during different locomotor modes (e.g. Winterbottom, 1974; Geerlink, 1979, 1983, 1989; Westneat, 1996). Several muscles, including their subdivisions and individual bundles, perform various roles in actuating the fin during locomotion(Thorsen and Westneat, in press). Despite widespread interest in limb development (e.g. Sordino et al., 1995; Ahn et al., 2002), the muscle morphology and function of early developing fish fins remains to be explored.

To investigate the use of pectoral fins in larval zebrafish locomotion, we examined axial bending and fin movement during routine swimming, and compared it with swimming following the startle response, a behavior thought to be produced at near-peak velocity that does not involve fin movement. Both our preliminary observations and reports in the literature(Batty, 1981; Borla et al., 2002; Müller and van Leeuwen,2004) found that the fins and axis were active simultaneously during routine larval fish swimming. Previous work on tetrapod locomotion demonstrating that alternation of the limbs and lateral bending of the axis are coordinated tightly during locomotion(Ashley-Ross, 1994; Bennett et al., 2001) drove our hypothesis that relative movements of limbs and axis of larval fish would be similarly patterned. Through the comparison of slow and fast swimming we suggest that the use of fins may be associated with the hydrodynamics experienced by the fish at different swimming speeds. In addition, we describe the pectoral fin musculature and discuss its functions in fin movement.

This work complements the previous work of Budick and O'Malley(2000), and Müller and van Leeuwen (2004), and focuses on the coordination of fin movements during slow swimming and the neural implications of kinematic patterns. Based on our data in larval zebrafish, and similar data in plaice larvae(Batty, 1981), we suggest that fishes and tetrapods may use similar neural coordination of axial and appendicular structures, and that the mechanisms for that coordination may have been conserved from an ancestral condition.

Eggs of wild-type zebrafish Danio rerio Hamilton 1822 andα-actin GFP transgenic zebrafish(Higashijima et al., 1997)were obtained from a breeding laboratory population maintained at 28.2°C. Embryos and larvae were raised at 28.2°C on a 14 h:10 h light:dark cycle until 5 days post-fertilization (dpf). Ten wild-type individuals were used at 5 dpf for behavioral experiments [3.94±0.17 mm total length (TL), mean± s.d.]. Muscle morphology was examined at 5 dpf in fiveα-actin transgenic individuals (3.97±0.07 mm TL). We found no difference (Student's t-test, P=0.6948) in size between wild type and α-actin populations.

For behavioral imaging, larvae were transferred to 10% Hanks solution and placed into Petri dishes (3.5 cm indiameter). Behavioral observations were made after acclimation to room temperature (25°C) for 15 min. Fish were placed under a Leica MZ 6 microscope (Wetzlar, Germany) with an attached high-speed Redlake MotionScope PCI 2000S video camera (San Diego, CA, USA). Black and white video at 1000 frames s^–1 and 240×210 pixel resolution was saved directly to a PC utilizing the Redlake Imaging MotionScope 2.21.1 software. Only spontaneous swimming events were collected for slow swimming trials. A glass micropipette was directed at the caudal region of the fish to elicit fast swimming responses.

Behavioral trials for slow swimming (30 total; three trials per individual for ten fish) and post-startle swimming [15 total; three trials per individual for five fish (a subset of the individuals used in slow swimming trials)] were analyzed with a customized program for digitizing the axial midline using LabView 5.0.1 software (National Instruments, Austin, Texas, USA; with virtual instruments designed by J. R. Fetcho, Cornell University, NY, USA). In addition, the timing and coordination of fin movements and parameters used to calculate Reynolds number (Re=VLρ/μ,where V and L are the velocity and length of the fish, andρ and μ are the density and viscosity of water) were determined by viewing trials frame-by-frame in NIH Image 1.62 (NIH, Bethesda, MD, USA). For each trial, we quantified kinematic data during the middle of a straight swimming bout for one tail-beat cycle and the three fin strokes that overlapped it (two on one side of the body, one on the other).

We defined each fin cycle (locomotor cycle) using three events: the frame just prior to start of fin abduction, the frame of maximum lateral abduction,and the first frame post adduction. The refractory period between fin cycles was defined by indeterminate fin activity, which results from a fin positioned adjacent to the body. Points of maximal medial axial curvature correspond to when the tip of the fish's tail changes direction(Budick and O'Malley, 2000). Only swimming after the first two tail strokes was examined, both for fast and for slow locomotion, to avoid the asymmetric initial bends and acceleration associated with the initiation of movement. Asymmetrical bends begin when maximal convexity is achieved for the first time in the same direction as the initial turn (which is at the end of the second beat), and ends when the axis cycles back to this configuration. All statistical tests were performed with JMP 3.1.6 (SAS Institute, Cary, NC, USA).

A subset of the α-actin transgenic zebrafish were stained with Calcein green (Molecular Probes, Eugene, OR, USA) to visualize the cleithrum and endoskeletal components of the pectoral fin. Fish were immersed in a 0.2%Calcein green solution following Du et al.(2001) for 15 min and allowed to swim freely. Fish were then rinsed in 10% Hanks and anesthetized with MS222 and embedded in agar for confocal imaging. α-actin GFP fish and Calcein-stained fish (Calcein green + α-actin GFP) were positioned with their left side down in 1.2% agar on a glass coverslip floor of a small Petri dish. The agar was covered with a 50% mixture of 10% Hanks solution and MS222 to prevent desiccation and fish movement while imaging. The pectoral girdle musculature was imaged under a Zeiss LSM 510 laser-scanning confocal imaging system (Thornwood, NY, USA).

Single optical sections and image stacks (40 × objective, 1028×1028 pixel resolution, 100 slices, 0.8 μm interval for three-dimensional reconstruction) of the pectoral girdle musculature and fin membrane were saved to a PC. Three-dimensional reconstructions were produced using Zeiss LSM 510 software. Fin surface area was calculated in ImageJ 1.30(NIH, USA) using three-dimensional lateral view projections of the fin(musculature and membrane). The number of muscle fibers constituting the pectoral musculature were counted using three-dimensional projections and Z stacks to aid in the visualization of the fin.

We found that slow and fast swimming in larval zebrafish represent distinct swimming gaits. Larval zebrafish synchronize their pectoral fin movements with the body axis (18–28 Hz) during slow swimming and tuck their fins against their body during fast swimming. During slow swimming, the axial muscle bends the body with the same frequency as the fins such that one left–right axial cycle corresponds to one abduction–adduction cycle of the fins (Fig. 1). When the right fin is initially at maximum abduction and ready to initiate adduction toward the body, the left fin is in its adducted position against the body. As the right fin is adducted, the tail flips toward it while the left fin is abducted. The right fin becomes fully adducted while the left fin is fully abducted. This cycle is then repeated with the other fin and subsequent tail-flip towards it.

Slow swimming with the fins and axis was significantly slower,approximately an order of magnitude lower, than fast swimming(Fig. 2A). Fast swimming was more effective at propelling the larval fish forward, covering more than four times the distance traveled during slow swimming(Fig. 2B). The average duration of locomotor cycles was significantly shorter during fast swimming events(Fig. 2C). Recalculated for the slow swimming condition averaged 43±3(Fig. 2D). During fast swimming, in which the fins are tucked and the axis alone propels the fish, Re numbers were significantly higher (427±31) than those of slow swimming, corresponding to a change in velocity (P<0.0001, Fig. 2A). Axial movement of zebrafish swimming possesses attributes of traveling and standing waves with a loose node present slightly posterior to the pectoral girdle (in agreement with Müller and van Leeuwen,2004).

We further investigated the fin movements and coordination of the fins and axis in the slow swimming gait (Table 1). The duration of a complete fin abduction–adduction cycle, including the refractory period(Drucker and Jensen, 1996a)when a fin is positioned against the body, averaged 41.23±0.94 ms and is not significantly different from the duration of a tail-beat cycle(40.23±0.94 ms, P=0.4602, Fig. 3A). The mean duration of the refractory period was 5.27±0.62 ms. The mean durations of the abduction and adduction phases across three fin cycles were not significantly different (17.66±0.44 ms vs 18.30±0.54 ms, P=0.3680; Fig. 3B). The mean time points of maximum fin abduction during slow swimming events(–0.10, 19.80 and 40.07 ms) coincided with, and were not significantly different (P>0.05) than, maximum axial bending (1.00, 21.17 and 40.30 ms), indicating that the fins and axis are highly coordinated(Table 1).

Pectoral fins in the larval stage are composed of a flexible endoskeletal disk (Grandel and Schulte-Merker,1998), fin membrane with actinotrichia, and muscles that actuate the fin membrane. The fin musculature, composed of two relatively simple muscles, is separated along the sagittal plane by an endoskeletal disk. Confocal microscopy sections (Fig. 4A–C) through these muscles in a transgenic fish that expresses green fluorescent protein in muscle fibers(Higashijima et al., 1997)indicate the position of the abductor/adductor musculature along the fin. Planar views of the abductor and adductor muscles(Fig. 4A,B,E,F), illustrate that muscle fibers run in a sheet on the fin extending upwards from its base. The abductor muscle is located on the rostral side of the fin and pulls the fin forward when it contracts. The adductor muscle is located on the caudal side of the fin and pulls the fin back against the body when it contracts. The abductor and adductor originate along the anterolateral and anteromedial surface of the cleithrum, respectively, and insert onto the fin membrane(Fig. 4F).

The fin musculature represents a functional fin blade surface area of 25590±993 μm² (82055±2535 μm² total fin blade area) in the lateral plane of the abductor(Table 2). The majority of the fin musculature is only one muscle fiber thick. The abductor and adductor muscles are composed of essentially the same number of muscle fibers(54.4±1.4 and 51.8±1.9, respectively; P=0.1902; Table 2). A number of fibers converge at the origination of the fin musculature, along the midline of the fin, which is about two muscle fibers thick. Muscle fibers along the midline run parallel from origination to insertion. The musculature servicing the leading edge and trailing-edge of the fin travel at opposite angles of curvature with respect to the midline fibers, with trailing-edge fibers having the largest relative curvature (Fig. 4A,B,E).

Many fishes can use axial and fin movements during swimming. However, in adult fishes, these two modes of locomotion tend to function independently of one another (e.g. Webb, 1994). By contrast, slow swimming of larval zebrafish is characterized by the coordinated movement of the fins and body axis. The pattern of coordination identified for slow swimming is similar to that observed in basal tetrapod groups, and may be controlled by an evolutionarily conserved neural circuit. Here we use locomotor behavior and morphology to develop hypotheses of locomotor function.

Axial locomotion in larval zebrafish has been well described in several recent studies (Budick and O'Malley,2000; Müller and van Leeuwen, 2004), the first at 6 and 9 dpf, the second at 2–5,7 and 14 dpf. We chose to focus on 5 dpf larvae because previous work on the neural control of swimming (e.g. Liu and Fetcho, 1999; Hale et al.,2001) and previous kinematic studies (e.g. Budick and O'Malley, 2000; Borla et al., 2002) have been done at that age. In the present study we further examine the pectoral fins,focusing on the steady swimming component of the slow start and compare it with straight forward swimming following the, previously described, fast start(Budick and O'Malley, 2000; Müller and van Leeuwen,2004).

To examine the relationship between fin and axial movement during steady swimming, we examined only the component of slow swimming in which the fish is swimming straight with equivalent left and right angular head movement, and with little change in angle between tail strokes (limiting initiation and end movement bias). We restricted trials because of our primary interest in the coordination of steady forward locomotion. For these components of the swimming trials, we found that pectoral fin beats matched both the frequency and phase of tail beats. We conclude that the fins and axis are highly coordinated by showing that the number of fin movements matched the number of points of maximal axial curvature (Fig. 1) and that there are no significant differences between the timing of fin maximum lateral abduction and maximum axial curvature(Table 1). This differs slightly from the findings of Müller and van Leeuwen(2004) that the pectoral fins are active during slow starts at the same frequency (typically below 30 Hz)but not necessarily the same phase as axial movements during what they call`slow-start swimming'. The pectoral fins and tail were found to be in phase in most sequences of slow start swimming in Müller and van Leeuwen(2004), although during burst swimming the pectoral fins were occasionally found to be out of phase with the tail (U. Müller, personal communication). We attribute our findings of tight fin–axial coordination to the extent of the slow swimming events examined (i.e. steady swimming – no burst of acceleration or deceleration).

The pattern, and relative timing, of fin and axial movement observed during slow swimming (Fig. 5A) bears a striking resemblance to the fore limb and axial coordination of some amphibians and reptiles, and the walking and running gait of many tetrapods(Fig. 5C,D; Daan and Belterman,1968; Ritter,1992). In most cases, the limbs are coordinated so that one cycle of axial bending corresponds to one limb cycle. Periods of maximal axial curvature generally coincide with maximal extension of limbs. In tetrapods,coordination of these behaviors involves the integrated activity of central pattern generators controlling the abduction–adduction rhythms of the limbs and lateral bending of the body(Devolvé et al., 1997; Bem et al., 2003).

The timing of limb activity, specifically the duration of abduction–adduction phases, varies among larval and adult fishes as well as tetrapods. Unlike in larvae, the abduction–adduction phases of pectoral swimming of adult fishes are not equal in duration. Studies by Gibb et al. (1994) on bluegill sunfish, Walker and Westneat(1997) on a wrasse, and Drucker and Lauder (1997) on a surfperch, demonstrate that fin adduction is shorter in duration than abduction. A comparison of timing of larval fin cycles to the swing/stance cycle in tetrapods reveals that cycle duration varies depending on locomotor speed (Biewener, 2003). Work on the salamander, Dicamptodon(Ashley-Ross, 1994), has shown that the stance and swing phase durations during the step cycle are nearly equal (Fig. 5C). However,kinematics of lizard limbs have shown that stance is longer in duration than the swing phase during running (Reilly and Delancey, 1997). Differences in timing of limb movements in vertebrates may reflect specializations based on morphological, behavioral and function requirements.

Fricke and Hissman (1992) have shown that the coelacanth (Latimeria chalumnae) can coordinate its pectoral fins with the caudal fin in a similar fashion to tetrapods and larval fishes. Fins were coordinated with a phase difference of 180° (abduction of one fin and adduction of the other). Pectoral fins were employed in an alternating fashion during accelerated forward movement, and have the ability to synchronize after a sudden start and during the following behaviors: curve swimming, accelerated movement to gliding, and upside down swimming (Fricke and Hissman, 1992). The pattern observed in the coelacanth provides additional behavioral evidence that a fin–axis motor pattern may be primitive among Osteichthyes(Sarcopterygii and Actinopterygii).

We suggest that the neural control of fin–axis coordination observed in tetrapods and larval fishes evolved prior to the split of sarcopterygian(lobe-finned) and actinopterygian (ray-finned) fishes and, although not common in the swimming modes of adult fishes, may have been conserved in the larvae of some species. Work by Grillner and Wallen(1985) suggests that the neural circuits controlling rhythmic axial oscillation in lamprey, one of the most basal vertebrate lineages, could be employed with limb CPGs to generate the pattern of axial muscle activity observed in tetrapods. Our data and other larval data (Batty, 1981)support this hypothesis by demonstrating that an axial traveling wave of bending, typical of fishes, can occur with rhythmic limb movements and may represent an intermediate condition of circuit coordination in which the axial bending and fore limb CPGs are integrated temporally but without substantial modification to the axial movement pattern. Furthermore, axial kinematics of adult eels (Gillis, 1996), in which axial movement alone generates propulsion, demonstrate a similar pattern to slow swimming in zebrafish, suggesting that use of pectoral fins does not necessarily alter axial patterns.

The diversity of vertebrates provides many opportunities to examine the diversity and evolution of coordination of limbs and axis. For example, Azizi and Horton (2004) recently found that the elongate salamander (Siren lacertina), which lacks hind limbs, is able to decouple appendicular movements and tail movements during aquatic walking, which the authors suggest may be related to elongation. This example highlights one of several possible evolutionary modifications of a primitive limb–axial circuit.

The presence of coordinated fin activity during slow swimming does not necessarily mean that the fins are participating in generating propulsive force. Fin movement may contribute to respiration(Osse and van den Boogaart,1999) or may be used to stabilize the body during swimming. Equal abduction and adduction phases of pectoral fin movements are highly coordinated with axial movement. As suggested by Batty(1981) for pectoral and axial movements in plaice larvae, the synchronization of these pectoral fin movements with axial movements may serve to offset head yaw by counteracting the recoil effect produced by the tail movement. Larval zebrafish pectoral fin strokes are timed precisely to do this, improving efficiency by reducing drag induced by axial swimming movements. The functions of the fins in respiration,stability and propulsion remain to be tested. Clarifying the roles of fin and axial coordinated movement patterns may provide important insight into the evolution and diversification of vertebrate locomotion.

The difference in Re number between slow and fast swimming suggests that hydrodynamic forces may be related to fin use during steady swimming. For larval fishes, the pattern of fin–axis locomotor coordination (Batty, 1981; Müller and van Leeuwen,2004; this paper) seems to be associated with swimming in low Re conditions. There was a tenfold difference in Re values between slow swimming with pectoral fins and fast swimming with axial movement alone (43±3 and 427±31, respectively). Re values and movement pattern reported here are similar to those described by Batty(1981), and Müller and van Leeuwen (2004), in larval plaice and zebrafish, respectively. Zebrafish maintain pectoral fin and axial coordination in a significantly decreased Re environment (Reranging from 3–11), which was achieved by increasing the viscosity of water using polyvinyl pyrrolidone (Sigma-Aldrich, Saint Louis, MO, USA; D.H.T. and M.E.H., unpublished). This finding suggests that coordinated alternating fin movements with the axis can occur through a wide range of low Renumbers.

The 5 day time period of the zebrafish studied here represents the first phase of pectoral fin development (Grandel and Schulte-Merker, 1998). Despite their early development, larval zebrafish pectoral fins are fully functional and perform normal locomotor behaviors. Based on kinematic and morphological data(Grandel and Schulte-Merker,1998; Thorsen et al.,2004; this paper), we believe that larval zebrafish musculature moves along with the fin membrane and is a functional component of the fin blade. We predict the abductor/adductor muscles are able to bend with the fin through its full range of motion. The only stationary structure of the pectoral girdle appears to be the cleithrum, which anchors both abductor and adductor muscles.

Muscle fibers are relatively evenly distributed along the fin membrane,although the muscle fibers inserting at the midline of the fin are longer than those of the leading or trailing-edge fibers. A distributed network of muscle fibers along the abductor and adductor muscles suggests an even force distribution along the fin. Neural innervation patterns(Thorsen et al., 2004) suggest independent control of the leading, middle and trailing-edge components of the fin musculature. We predict then, that in the larval condition, the pectoral musculature has variable control of the fin due to innervation patterns and muscle curvature enabling asymmetries in fin movement. High-resolution,high-speed video technology could be used to test these predictions.

The patterns of movement described here suggest a similarity in the neural control of limbs and the body axis. We suggest that the same basic limb–axis motor control circuit has been conserved evolutionarily and is present in fishes and salamanders; however, in fishes it is only used during early development when animals experience low Re conditions, whereas tetrapods have retained and modified it for function in adults. We believe that a number of factors, including Re, stability, fin musculature and a primitive neural circuit, contribute to produce the behavior of the zebrafish during slow swimming. Many questions remain regarding the function of fins throughout development, how fins are controlled through sensory–motor mechanisms, neural circuitry for generating fin abduction–adduction rhythms and fine control of motion. The simplicity of the pectoral fin musculature composed of one muscle at one limb joint makes the larval zebrafish an excellent model to address many of these questions.

For their insight and comments on the manuscript, we thank H. Bierman, A. Rice, J. Socha, S. Thorsen, J. VanTassell and M. Westneat. The manuscript was improved by the comments of two anonymous reviewers. We thank J. R. Fetcho for a digitizing program. Thanks to Ryan Day and Ru Yi Teow for laboratory assistance. This work was supported by a Brain Research Foundation Grant, NSF Grant No. IBN0238464, NIH Grant No. NS043977 to M.E.H. and NSF graduate research fellowship to D.H.T.