Locomotor corollary activation of trigeminal motoneurons: coupling of discrete motor behaviors

During rhythmic locomotion, a number of otherwise independent movements are influenced by the rhythm of the motor commands. In vertebrates, such coupling is observed for respiration (Bramble and Carrier, 1983), eye movements (Lambert et al., 2012), tail motion (Wada et al., 1993) or phase-locked arm and trunk motor adjustments (Earhart, 2013). While some coupled motor behaviors such as tail motion, e.g. in dogs, improve locomotor performance (Wada et al., 1993), others improve sensory acquisition and processing as observed for retinal image stabilization (Chagnaud et al., 2012a; Lambert et al., 2012). One possibility for coupling otherwise unrelated motor behaviors to locomotion is via corollary discharge that derives from spinal or supraspinal locomotor centers. These signals allow fast and reliable phase locking of the respective motor behavior to the locomotor rhythm (Chagnaud et al., 2012a). Corollary discharges are independent of locomotor style and occur during rhythmic limb-based locomotion in terrestrial vertebrates as well as during body/tail-based swimming in aquatic vertebrates (Chagnaud et al., 2012a). Traditionally, motor corollary discharges have been described as mechanisms to differentiate between environmental and self-generated sensory inputs (Crapse and Sommer, 2008; Cullen, 2004, 2011; von Holst and Mittelstaedt, 1950; Poulet and Hedwig, 2007; Sommer and Wurtz, 2008). In contrast, the impact of motor corollary discharges on motor systems and the indirect effects on the sensory encoding have been less well investigated. Some studies have shown that corollary discharge can cause reflex inhibition during swimming in Xenopus embryos (Sillar and Roberts, 1988), suppress withdrawal responses in gastropods during feeding behavior (Davis et al., 1974; Kovac and Davis, 1980) and drive compensatory eye movements during swimming in larval Xenopus (Lambert et al., 2012). However, attempts to decipher the mechanisms underlying the effects of corollary discharge on motor behaviors are often constrained by the complexity of the central nervous system and the limb-based locomotion. In contrast, tail-based swimming in amphibians (Wassersug and Hoff, 1985) and fishes represents a simple, stereotyped locomotor pattern with correspondingly simpler spatio-temporal profiles of corollary discharges.

A number of aquatic anamniotes possess various numbers of mobile appendages on the head that contain several types of sensors (see Fox, 1999). One example is the pair of rostrally protruding tentacles in Xenopus laevis Daudin tadpoles that are equipped with Merkel cells (Ovalle, 1979). The mechanosensory nature of these cells (Maricich et al., 2009) suggests an important role for touch discrimination and surface structure recognition. Because of the location of the tentacles on the head and the tadpole's undulatory swimming style with prominent horizontal head oscillations, the tactile function of these appendages might be impaired during locomotor behavior. Given the previously reported spinal efference copy-driven compensatory eye motion in Xenopus tadpoles during swimming (Combes et al., 2008; Lambert et al., 2012), a protection of the sensor and a potential improvement of propulsive efficacy could be achieved by a similar spinal locomotor corollary discharge that causes a tentacle retraction.

Here, we provide direct evidence that locomotor corollary discharges during rhythmic swimming in X. laevis tadpoles initiate a bilateral tentacle retraction. Fluorescent tract tracing, Ca²⁺ imaging, electrophysiological recordings and video analysis of tentacle motor behavior during fictive locomotion in semi-intact in vitro preparations outline the underlying trigeminal motoneuronal populations, their firing pattern and their link to ongoing locomotor commands. The observed coupling of tentacle retraction to swimming is an excellent example of intrinsic control of a particular motor behavior that is otherwise unrelated to locomotion.

Xenopus laevis tadpoles are equipped with a bilateral, mobile pair of tentacles (Cannone and Kelly, 1977; Ovalle, 1979) that are attached to the upper jaw at the lateral aspect of the mouth. At rest, these appendages extend rostro-laterally (Fig. 1Ai,B) at an angle of 19±10 deg (mean±s.d., 61 tentacles) relative to the longitudinal body axis. During fictive locomotion, indicated by rhythmic bursts in spinal ventral roots (VRs) in isolated tadpole preparations (Combes et al., 2004; see below for details), video recordings revealed that the tentacles were concurrently retracted (Fig. 1Aii). The maximal retraction angle during swimming was 115±24 deg (61 tentacles, n=41 swimming episodes in N=7 preparations; Fig. 1B). During locomotor activity, the tentacles essentially remained in a lateral position interspersed by variably timed and sized oscillations (Fig. 1C). As the locomotor activity ceased, the tentacles protracted towards the initial resting position (Fig. 1Aiii, Aiv, Ci).

Simultaneous video recordings of the tentacles on both sides (n=20 episodes in N=4 preparations) yielded very similar motion dynamics and trajectories of both tentacles during a given locomotor episode (see red and blue traces in Fig. 1Ci). The close coupling of onset, duration and motion pattern of the left and right tentacle was verified by correlating the respective motion trajectories (Fig. 1Cii). The distribution of the correlation coefficients (median: 0.81; Fig. 1D) was significantly different from zero (P<0.0001; Wilcoxon signed-rank test) indicating that both tentacles move symmetrically. This close match of the bilateral motor activity was exploited in subsequent experiments by combined recordings of the motion pattern of one tentacle and the neuronal commands of the trigeminal motor nerve that innervates the tentacle on the other side. Moreover, the presence of tentacle movements during locomotion in completely isolated and immobile Xenopus tadpole preparations and thus in the absence of tail motion-related sensory signals suggests that the motor commands derive from a locomotor corollary discharge that activates the tentacle motor system.

Tentacle movements in Xenopus tadpoles are produced by a single muscle (m. levator mandibulae pars lateralis) that retracts the tentacle (Nieuwkoop and Faber, 1994); in contrast, protraction of the tentacle occurs in the absence of a respective muscle but likely involves the mechanics of a cartilaginous spring (Ovalle et al., 1998). The tentacle muscle, responsible for the retraction, has a joint proximal origin with the jaw closing muscles (m. levator mandibulae pars intermedius and medialis) but remains otherwise separate (dashed line in Fig. 2B), and attaches distally at the cartilaginous base of the tentacle (Fig. 2A,B). The exclusive control of tentacle movements by the lateral portion of the m. levator mandibulae was demonstrated by a set of differential muscle lesions. A specific transection (three preparations) of the m. levator mandibulae pars lateralis, which left the other two portions of the m. levator mandibulae intact, abolished all tentacle movements. In contrast, a reciprocal surgical intervention (three preparations), that preserved the lateral portion, left tentacle movements unaffected. The sensory-motor innervation of the tentacle and its muscle is formed by the trigeminal ophthalmic (rOP) and mandibular branches (rMA) as illustrated by the pattern of labeled peripheral nerves following application of biocytin to the trigeminal nerve root close to the hindbrain (Fig. 2A). The motor innervation of the tentacle muscle, however, originates exclusively from the smaller branch (tentacle nerve, TN) after ramification of the rMA into two branches at the level of the m. levator mandibulae (red asterisk in Fig. 2B).

Apart from innervating the tentacle muscle, the TN branch also innervates the jaw closing muscle (Fig. 2A,B). This branch exclusively contains motor axons as indicated by the absence of sensory hindbrain projections following tracer application to this particular nerve (left side in Fig. 2C). This differs from the observed pattern after labeling of the second rMA branch (blue arrowhead in Fig. 2B) or the trigeminal nerve root, where additional dense afferent terminations were observed in the dorsal hindbrain throughout r1–r7 (arrowheads on the right side in Fig. 2C,E), indicating the presence of sensory fibers in the respective nerves.

The hindbrain location of motoneurons innervating the tentacle muscle was compared with the entire population of trigeminal motoneurons following application of biocytin to the TN at its entrance into the muscle or to the trigeminal nerve close to the hindbrain (cell groups on the left and right side, respectively, in Fig. 2C). Tracer application to the TN consistently labeled a group of motoneurons in hindbrain segments r2 and r3 (left side in Fig. 2C–E). This bi-segmental location coincides with that of the entire population of trigeminal motoneurons, labeled from the trigeminal nerve root (right side in Fig. 2C). The number of retrogradely labeled tentacle motoneurons was variable and ranged from 9 to 87 neurons in different experiments (median: 27 cells, N=10 preparations). These motoneurons only innervated the m. levator mandibulae because labeling of the TN and the second rMA branch (blue arrowhead in Fig. 2B) in different colors (Alexa Fluor dextran 488 and 546; N=4 preparations) did not yield double-labeled neurons (see absence of yellow cells in r2 and r3 in Fig. 2E). However, tentacle motoneurons were intermingled with other rMA motoneurons (green and red cells in r2 and r3 in Fig. 2E) within the trigeminal nucleus.

Tentacle motoneurons form a heterogeneous population with respect to soma size and shape (inset in Fig. 2C,D). A histogram of soma diameters (120 cells in N=6 preparations) revealed that cell size extends over a large range that suggests the presence of up to three groups of neurons with a tendency for a size-related dorso-ventral separation (see Fig. 2C,D inset for z-axis color-coded confocal reconstruction). Neurons were categorized as either small (<12 µm) and mostly round (group 1: S) or medium sized (≥12 and <30 µm) with oval somata (group 2: M) with the consistent exception of one to two very large (≥30 µm) neurons with elongated cell bodies (group 3: L), located at the most dorso-medial aspect of the labeled cell group (L in Fig. 2C,D inset). Medium-sized neurons were located at more medial and dorsal positions relative to the smaller cells (Fig. 2C,D inset). The mean size of the neurons was 8.3±1.9 µm (56 cells) for small neurons, 18.1±3.8 µm (58 cells) for medium neurons and 34.6±4.4 µm (six cells) for large neurons. Accordingly, tentacle motoneurons form a morphologically diverse and dispersed cell group within the trigeminal motor nucleus in r2 and r3 with a size-related dorso-ventral arrangement.

The morphological diversity of tentacle motoneurons prompted us to test whether all or only a particular subpopulation of motoneurons become active during swimming and cause the observed tentacle retraction. Neuronal activity was measured as Ca²⁺ transients in the entire tentacle motor nucleus or in individual motoneurons, following unilateral or bilateral application of Calcium Green-1 dextran to the peripheral motor target(s) and retrograde transport to the cell bodies (Fig. 3Ai inset, Bii, Cii). During episodes of fictive locomotor activity, visible as rhythmic bursting in the spinal VR (green trace in Fig. 3Ai), the population of tentacle motoneurons showed fluorescence changes that were timed to the VR bursting (Fig. 3Ai). Simultaneous Ca²⁺ imaging of the bilateral tentacle motor nuclei (red and blue outline in Fig. 3Ai inset) yielded almost identical dynamics of the transients (red and blue traces in Fig. 3Ai), matching the symmetric motion profiles of the two appendages (Fig. 1Ci,D). The duration of these population responses (n=15 swimming episodes in N=8 preparations), measured as half-width of the Ca²⁺ transients (Fig. 3Ai), was highly correlated with the duration of the fictive swimming episodes (Fig. 3Aii). Thus, the duration of Ca²⁺ responses in the tentacle motor nuclei closely matches the duration of a swimming episode.

In a separate set of experiments, we determined the fraction and morphology of the activated tentacle motoneurons by recording Ca²⁺ responses of individual motoneurons (34 cells; n=9 swimming episodes in N=3 preparations; Fig. 3Bii,Cii). Ca²⁺ transients were encountered in 14 out of 34 retrogradely identified motoneurons during fictive swimming episodes (Fig. 3Bi,Ci). Responses were observed only in the population of medium-sized motoneurons with oval cell bodies (16.1±1.6 μm, 14 cells), whereas neither the subgroup of small motoneurons nor the one to two very large motoneurons displayed any locomotion-related responses (Fig. 3B,C). Based on the dynamics of the Ca²⁺ responses, the 14 responsive neurons could be separated into a majority of cells (11 cells; color-coded traces in Fig. 3Bi and blue trace in Fig. 3Ci) with a half-width of the transients that significantly correlated with the duration of the corresponding swimming episode (blue dots in Fig. 3Biii). In contrast, a few motoneurons (two swimming episodes in three neurons) exhibited Ca²⁺ responses that were highly transient and only present at the beginning of a swimming episode (red trace in Fig. 3Ci; red dots in Fig. 3Biii). However, the morphology of these neurons was indistinguishable and the size not significantly different from the size of those with the predominant response pattern. Thus, motoneurons responsible for tentacle motion during tadpole swimming appear to form a morphologically homogeneous group with Ca²⁺ response profiles that suggest the presence of either transient or sustained activity during locomotor episodes.

To reveal the discharge profile of the motor commands that provoke tentacle movements during locomotor CPG activity, we made simultaneous recordings of a VR on one or both sides of the spinal cord, the TN on one side and the motion of the tentacle on the other side (Fig. 4A). During fictive swimming, indicated by an episode of burst activity in VRs (green and gray traces in Fig. 4B), the otherwise silent TN became active (black trace and firing rate in Fig. 4B). In addition, the intact tentacle contralateral to the recorded TN was simultaneously retracted (blue trace in Fig. 4B). The similarity of tentacle motion and tentacle motor profiles (blue and black traces in Fig. 4B) was quantified by correlating the normalized position of the tentacle with the normalized firing rate of the contralateral TN (Fig. 4C). Normalized tentacle motion and contralateral TN firing rates were correlated with a coefficient ranging between 0.50 and 0.93 (median: 0.81; n=18 swim episodes in N=4 preparations; Fig. 4D). The distribution of correlation coefficients was significantly different from zero (P<0.001; Wilcoxon signed-rank test). Despite the variability in the movements of both tentacles (Fig. 1Ci), the distribution of correlation coefficients indicates that the firing pattern of the TN on one side faithfully predicts the motion profile of the tentacle on the other side.

Locomotor influence on tentacle motion was studied by analyzing the coupling pattern between spinal and tentacle motor activity. Tadpole swimming is a fluctuating motor behavior in terms of duration and strength/speed, which is reflected during fictive locomotion by variable episode durations, burst frequencies and amplitudes (Combes et al., 2004). By comparing spinal VR and TN activity, we identified those parameters of the locomotor commands that are transmitted to the tentacle motor system. The most obvious of these parameters was the duration of the fictive swimming episode: the discharge in spinal VRs and in the TN closely matched in time (compare black with green traces in Fig. 5A), as indicated by the significant linear regression with a slope close to unity (Fig. 5B; r²=0.94, P<0.0001, Wilcoxon signed-rank test, n=57 swim episodes in N=10 preparations). This agrees with the behavioral observation that the tentacles are retracted during the entire swimming event.

Alterations in swimming strength in vivo essentially derive from changes in the magnitude and frequency of tail excursions that in turn are represented in vitro by variations in the cycle frequency of the VR bursts and in the intra-burst firing rate (Combes et al., 2004). A spontaneous increase in spinal VR intra-burst discharge, represented by a larger integral of the firing rate (green traces in Fig. 5C) was accompanied by a similar increase in retraction angle of the tentacle (blue trace in Fig. 5C). With increasing locomotor strength from low to medium to high swimming amplitudes (Fig. 5D), the tentacle reached progressively more eccentric positions. In fact, the tentacle positions for the three levels of swimming strength were significantly different from each other (P<0.0001, Kruskal–Wallis test and post hoc comparisons; n=4 swimming episodes in N=3 preparations). The variation of position with swimming strength prompted us to compare the swimming strength (green traces in Fig. 5E) during a given locomotor episode with the rate of TN firing (black traces in Fig. 5E). Spontaneous alterations of VR burst amplitude caused a comparable modulation of the TN firing rates, with the latter discharge ceasing as soon as the VR bursting decreased below a particular level (red arrowhead in Fig. 5E). Pooling the data from four different swimming episodes in three preparations where the VR burst discharge frequency displayed a spontaneous modulation showed that the TN activities at the three levels of swimming strength were also significantly different from each other (Fig. 5F; P<0.0001, Kruskal–Wallis test with post hoc comparisons), with a more pronounced TN activity during stronger swimming.

While the previously described corollary discharge in extraocular motoneurons during rhythmic locomotion in Xenopus tadpoles is strictly phase coupled to the VR bursting in a 1:1 fashion (Lambert et al., 2012), the dynamics of tentacle motor commands appears to be more variable and temporally more complex (Fig. 6). To assess the coupling of TN spiking during episodes of fictive swimming as a function of the timing of VR bursts, we generated VR burst-triggered cumulative spike time histograms (see Materials and methods). Based on entropy statistics (Kajikawa and Hackett, 2005), recordings from 18 out of 31 swimming episodes in N=7 preparations (58%) showed significant phase coupling between spinal VR and TN discharge (Fig. 6A,B,D), while the TN discharge during 13 swim cycles displayed no coupling to the VR burst rhythm (Fig. 6C). In the group of phase-coupled recordings, a mono-phasic coupling (four out of 18; Fig. 6A) could be clearly distinguished from a bi-phasic pattern (five out of 18; arrowhead in Fig. 6Bii inset), while the remainder (nine out of 18) exhibited a phase coupling that could not be unambiguously distinguished as one or the other. The different coupling patterns during individual swimming bouts were summarized as circular plots, emphasizing the variability among the different coupling categories (Fig. 6Aiii–Ciii). Functionally, the mono- or bi-phasic timing of TN bursts (Fig. 6Aiii, Biii) suggests that the inputs to trigeminal motoneurons originate from one or both sides of the spinal cord during a particular recording. This might also explain the difficulty of assigning a particular phase relation to the coupling pattern in most cases (mono- versus bi-phasic), assuming a continuum between unilateral and bilateral locomotor corollary drive to tentacle motoneurons.

During fictive swimming in Xenopus tadpoles, a corollary discharge informs the tentacle motor system about ongoing locomotor activity. This intrinsic locomotor signal activates a discrete set of trigeminal motoneurons on each side of the rostral hindbrain that causes a symmetrical retraction of the appendages. The motor command encodes both duration and magnitude of the locomotor activity, thereby ensuring a reliable coupling between propulsive locomotion and tentacle motion. Given the phase coupling between trigeminal nerve discharge and spinal motor rhythm, the locomotor corollary discharge derives at least in part from the spinal central pattern generator (CPG) circuitry.

Xenopus laevis tadpoles possess a pair of mobile appendages between developmental stage 47 and 61 (Nieuwkoop and Faber, 1994), which contain Merkel cells and thus were suggested to play a role in touch reception (Ovalle, 1979; Ovalle et al., 1998). These tentacles are usually protruded forward but can be actively retracted into a lateral position (Fig. 1A). The use of appendages for mechanoreceptive exploration of the environment in these animals is useful for near-field orientation and navigation in the mostly murky aquatic environment in which these animals naturally live (Nieuwkoop and Faber, 1994). A similar mechanoreceptive role in mammals is performed by mechanoreceptive facial hairs – whiskers. Rodents use whisker movements for tactile exploration of the environment, for instance while walking along the wall of a cage (Hartmann, 2011). Similarly, tadpoles drift slowly with their tentacles touching the walls and floor of the tank (S.H., R.B., H.S. and B.P.C., unpublished observation), potentially collecting tactile information. However, several differences between the two structures exist: tentacles are formed by a single protrusion located on each side of the head, while whiskers are usually arranged in groups (Berg and Kleinfeld, 2003). Moreover, whisker movements in rodents are controlled by multiple muscles (Berg and Kleinfeld, 2003), while tentacle motion in Xenopus tadpoles is exerted by a single muscle and a spring-like antagonistic mechanism (Ovalle et al., 1998). For tactile exploration, whiskers are actively moved (‘whisking’), while tentacles are kept in an extended position. However, despite these morphological differences, amphibian tentacles appear to be similar, yet simpler versions of touch-receptive mammalian appendages.

The presence of Merkel cells at a relatively high density in Xenopus tentacles (Ovalle, 1979) indicates a particular sensitivity of these structures to touch. Because of the hydrodynamic drag between the appendages and water, permanently extended tentacles could rhythmically stimulate these touch receptors during swimming-related head undulations. A tonic retraction of the tentacles during locomotor episodes might considerably reduce excessive stimulation of the sensory cells, in particular during strong and fast swimming. Under this assumption, locomotor corollary discharge-mediated tentacle retraction would minimize self-generated stimulation of the Merkel cells during swimming by reducing the water flow/pressure impinging on the tentacles. In a number of other sensory systems, corollary discharge similarly reduces or even suppresses self-generated sensory inputs (for review, see Crapse and Sommer, 2008); this also complies with the classical role of motor efference copies or corollary activity (von Holst and Mittelstaedt, 1950; Sperry, 1950). As a side effect, tentacle retraction during locomotion reduces the likelihood of mechanical damage to the tentacles that can reach a maximal length of 1 cm (∼20% of larval body length).

In addition to this putative protective function, tentacle retraction might also improve the hydrodynamics of the tadpole. Appendage retraction streamlines the body shape, thereby facilitating energetically more efficient swimming (Crespi et al., 2013; Liu et al., 1997). This is particularly important during strong swimming, when protruded tentacles would cause a considerable drag and impair the sinusoidal motion of the head during swimming, specifically at those larval stages where these appendages are at their maximal length. In fact, a similar streamlining of the body shape is observed during tail-based swimming in other amphibians such as salamanders, where both forelimbs and hindlimbs are aligned to the body (Delvolvé et al., 1997) or in axolotl, where the external gills are retracted (D'Août and Aerts, 1997). Thus, locomotor-coupled head appendage or limb retractions in amphibians represent distinct behavioral reactions that improve the hydrodynamic signature and, in the case of retracted limbs during tail-based swimming, considerably enhance locomotor performance (Crespi et al., 2013). Such a strategy is not restricted to amphibians but occurs also during swimming in alligators (Manter, 1940; Fish, 1984) and ‘terrestrial swimming’ in sandfish lizards (Maladen et al., 2009) where the tail-based propulsion is accompanied by similar limb adductions as in salamanders (Delvolvé et al., 1997).

The absence of propulsive movements and movement-related sensory feedback in semi-intact Xenopus preparations (Straka and Simmers, 2012) demonstrates that the coupling between locomotion and tentacle motion relies exclusively on intrinsic signals, i.e. corollary discharge from locomotor areas. Theoretically, these signals could derive from midbrain locomotor centers (Cabelguen et al., 2003; Saitoh et al., 2007) or from CPGs in the rostral spinal cord or from a combination of the two.

A spinal contribution of the corollary activity in tentacle motoneurons is supported by the presence of strict phase coupling of the TN burst discharge with the burst rhythm of the spinal VR on one (mono-phasic) or on both sides (bi-phasic; Fig. 6A,B) along with the highly correlated amplitudes of spinal VR and TN firing or tentacle motion. The absence of a clear VR–TN burst coupling in a number of experiments (Fig. 6C) might reflect the consequence of signal integration in a particular motoneuronal population that only becomes apparent in some experiments depending on the number and variety of recorded TN axons. Alternatively, or in addition, the ascending spinal locomotor corollary discharge in tentacle motoneurons, unlike in extraocular motoneurons (Lambert et al., 2012) might be supplemented by tonic signals from locomotor centers in the midbrain and/or hindbrain. In fact, midbrain neurons in the salamander Notophthalmus viridescens, for instance, display tonic activity during locomotion (Cabelguen et al., 2003), similar to that seen in the TN nerve during locomotor episodes. Even though the relative contributions of midbrain locomotor regions and spinal CPGs to the tentacle motor corollary discharge remain unclear, we hypothesize that both neuronal structures contribute to the trigeminal activation during locomotor activity.

While locomotion-related tentacle retraction potentially serves several distinct, not mutually exclusive purposes (see above), the coupling could be a vestige of the rostro-caudally distributed activity in the spinal cord-like nervous system of vertebrate ancestors during undulatory swimming (Fetcho, 1992; Wada, 1998). However, given the rather specialized structure and function of these rostral appendages, this notion appears too simplistic. The inherent linkage between the two motor systems rather suggests a particular functional role of tentacle retraction during swimming, which is not necessarily exclusive or in opposition with a pre-existing vestigial signaling pathway. Accordingly, ascending locomotor corollary discharges from the spinal CPG circuitry might represent a widely distributed signaling component within the brain during rhythmic locomotion and might influence numerous sensory and/or motor systems. Among the hindbrain motor systems currently known to receive spinal inputs are the extraocular system (Combes et al., 2008; Lambert et al., 2012; von Uckermann et al., 2013) and the trigeminal motor component (this study). Evidence for an even more widespread function arises from studies on spiny dogfish (Russell and Roberts, 1974) and larval Xenopus (Chagnaud et al., 2012b), in which connections from the spinal cord locomotor CPG to mechanosensory efferent neurons were demonstrated. Further hindbrain targets might include hindbrain nuclei in axolotl that cause a retraction of the gills during swimming (D'Août and Aerts, 1997).

Based on the various examples, we hypothesize that appendage retraction (e.g. limbs, tentacles, gills) during undulatory swimming is generally caused by locomotor corollary activity, which allows a fast and reliable coupling of independent motor behaviors with locomotion. Such a functional role expands the traditional concept of motor efference copies/corollary discharge in simply differentiating between external and self-generated sensory signals (Cullen, 2004, 2011). Accordingly, corollary discharge during locomotor activity appears to influence a large number of both sensory and motor systems throughout the brain, including the trigeminal system, potentially optimizing locomotor performance and sensory perception (Chagnaud et al., 2012a). Depending on the type of motor behavior (rhythmic, task oriented), the effect of corollary activity might vary considerably (Crapse and Sommer, 2008; Cullen, 2014; King, 2013). Nonetheless, intrinsic coupling offers a convenient substrate for a context-dependent, direct linkage between otherwise discrete motor behaviors.

Experiments were performed on semi-intact in vitro preparations of 33 larval Xenopus at stages 51–55 (Nieuwkoop and Faber, 1994) in compliance with the ‘Principles of Animal Care’ publication by the National Institutes of Health and the German law for animal protection (Tierschutzgesetz). Permission for the experiments was granted by the Regierung von Oberbayern (55.2-1-54-2531.3-18-10). All animals were obtained from the in-house breeding facility at the Biocenter-Martinsried of the LMU Munich.

In all experiments, animals were first anesthetized in 0.02% 3-aminobenzoic acid ethyl ester (MS-222; Sigma-Aldrich Pharma Ltd, UK) in ice-cold frog Ringer solution (composition in mmol l⁻¹: NaCl, 75; KCl, 2; CaCl₂, 2; MgCl₂, 0.5; NaHCO₃, 25; glucose, 11: pH 7.4). The ventral part of the head/body, including the lower jaw and visceral organs, was carefully removed, leaving the tail attached to the head and the sensory-motor innervation of the tentacles intact. The skin covering the dorsal portion of the head was removed, the soft skull tissue opened and the forebrain disconnected. Preparations were transferred to a Petri dish (volume 5 ml) and fixed dorsal-side up to the Sylgard floor with insect pins. In some cases the spinal cord was exposed up to segment 20 and the VRs were transected bilaterally. The remaining, caudal part of the tail was firmly secured on both sides to the Sylgard floor with insect pins at the level of segments 21–25, with the remainder of the tail left free to perform undulatory swimming movements. After repeated rinsing in fresh Ringer solution, preparations were continuously superfused with oxygenated Ringer solution at a rate of 1.3–2.1 ml min⁻¹. The temperature of the bath solution was maintained at 17±0.5°C. Recordings lasted up to 5 h.

Tentacle motor behavior during spinal VR activity (i.e. fictive swimming, hereafter referred to as swimming) in these isolated preparations was quantified by video analyses of tentacle motion and/or recording the respective motor nerve discharge. For monitoring tentacle movements during fictive swimming, one or both tentacle(s) remained attached and innervated by the respective trigeminal nerve branch. Motor commands for tentacle movements were captured after further unilateral or bilateral isolation of the trigeminal motor nerve branch that innervates the tentacle muscle.

Movements of one or both tentacles during spinal locomotor activity (see below) were video-captured from the top in isolated preparations (n=8) with a CCD camera (Axiocam, Zeiss, Germany) with a resolution of 272×208 pixels at a rate of 20–50 frames s⁻¹ either separately or in combination with the spike activity of the contralateral tentacle motor nerve. The motion trajectory of the tentacle was analyzed offline by manually tracking the distal yet still straight portion of the tentacle with respect to the base using the Vidana software (courtesy of Dr M. Hofmann). From the position coordinates, the angle of the tentacle relative to the longitudinal body axis was calculated in Matlab (MathWorks, Natick, MA, USA) and plotted as motion trajectory over time.

Spontaneous or evoked locomotion in isolated tadpole preparations was recorded as bilaterally alternating rhythmic burst discharge in spinal VRs, termed fictive swimming (e.g. Combes et al., 2004). Multi-unit activity from one or two spinal VRs of segments 10–15 along with the discharge of one or both tentacle motor nerves (N=12) were recorded with individually adjusted glass suction electrodes, fabricated with a horizontal puller (P-97 Brown/Flaming, Sutter Instruments, Novato, CA, USA). The recorded neuronal activity was amplified (EXT 10-2F; npi electronics, Tamm, Germany), digitized at 10 kHz (CED 1401, Cambridge Electronic Design, Cambridge, UK), processed with commercial software (Spike 2, Cambridge Electronic Design), stored on a PC and analyzed off-line with custom-written scripts using Igor Pro (Wavemetrics, Tigard, OR, USA) or Matlab.

To compare the tentacle retraction on one side with the TN firing on the other side, we calculated the correlation coefficient after normalizing the angular tentacle position and the TN firing rate to their maximum for each swimming bout. The temporal relationship between rhythmic spinal VR and trigeminal motor activity was evaluated by comparing the duration of the concurrent discharge episodes, obtained from multi-unit recordings of at least one VR and one TN. These data were extracted from n=57 individual swimming episodes (these included only rhythmic alternating activity and no escape) in N=10 preparations and compared using a scatter plot and linear regression.

The degree and variation of swimming strength during fictive locomotion were estimated from integrals of the VR bursting in recordings of those locomotor episodes that showed a modulation of the swimming strength. In these traces, the swimming amplitude was calculated from integrals of VR bursts with a bin width of 30 ms (N=2 preparations) or with a bin width that matched the frame rate of simultaneously video-recorded tentacle motion (21 ms, N=3 preparations; 71 ms, N=1 preparation). Integrated and binned VR burst amplitudes were normalized to the maximal magnitude within a given swimming episode and were grouped according to amplitude into three categories (low, medium, high) indicating weak, medium and strong swimming. These three different levels were used to plot the normalized firing rate of the simultaneously recorded TN or the video-recorded tentacle position. Significance between the three levels was tested using a Kruskal–Wallis test, followed by post hoc multiple sample comparisons (Wilcoxon signed-rank tests).

To determine potential phase correlations between swimming (i.e. VR activity) and TN activity, the instantaneous firing rate of the VR bursts was calculated from recordings of 31 locomotor episodes in N=7 preparations. The phase relationship between the TN discharge and VR bursts was calculated and plotted as a histogram of spikes with a bin size of 5 deg (one burst cycle=360 deg). The histogram was normalized by dividing the spike count by the number of swimming cycles. Histogram data were displayed as circular plots, with their maxima aligned at 0 deg. To assess whether the distribution of spikes in the TN was phase coupled to unilateral or bilateral VR activity, a previously described procedure was employed (Kajikawa and Hackett, 2005). Briefly, the entropy (D score) of the plotted spike time histogram was calculated as the difference between unity and a normalized entropy measure (D=1−E/E_max). D may vary from 0 to 1, with 0 being no phase coupling (probability of firing is equally distributed over the entire swimming cycle) and 1 being maximal phase coupling, independent of the underlying spike time distribution. In order to determine the significance of D, the inter-spike intervals were reshuffled 10,000 times for each swimming episode and a D-value was calculated for each repetition (Kajikawa and Hackett, 2005). The D-value from the initial analysis was then compared with this distribution of reshuffled D-values by calculating the proportion of those that were equally large or larger than the original value. Phase coupling was considered significant if this proportion was smaller than 0.05 (i.e. one-tailed probability with α=0.05).

To assess the location and spatial arrangement of tentacle motoneurons in the hindbrain, fluorescent tracers (Alexa Fluor dextran 488, 546; Life Technologies, Carlsbad, CA, USA) and biocytin (Sigma-Aldrich) were applied to different portions of the trigeminal nerve in isolated in vitro preparations (Straka et al., 2001). Crystals of the respective tracer, melted to the tip of an injection needle, were inserted into various trigeminal nerve branches or the main trigeminal nerve root. After overnight incubation in oxygenated Ringer solution at 14°C, preparations were fixed in 4% paraformaldehyde in 0.1 mol l⁻¹ phosphate buffer (PB, pH 7.4) at 10°C for 5–12 h and rinsed (3×10 min) in cold 0.1 mol l⁻¹ phosphate-buffered saline (PBS, pH 7.4). In experiments where biocytin was applied, preparations were fixed and rinsed as described above and kept in Dent's solution [80% methanol, 20% dimethylsulfoxide (DMSO)] at room temperature for 12–14 h, transferred into 100% methanol and stored at −80°C for at least 1 h. Then, preparations were rehydrated at room temperature in 70%, 50% and 35% methanol solutions in distilled water for 1 h each, followed by 0.1 mol l⁻¹ PBS (3×30 min), incubated in a 1:200 solution of streptavidin-Cy2 (Dianova, Germany) in 0.1 mol l⁻¹ PBS for 2 h and rinsed in PBS (3×30 min). The brainstem of all preparations was removed, cleaned of surrounding tissue, mounted on slides and coverslipped with Vectashield (Vector Laboratories, Burlingame, CA, USA). Retrogradely labeled neurons and afferent nerve terminations were reconstructed from stacks of optical sections (1.5–3 µm) obtained from a confocal microscope (SP5, Leica, Germany). In order to map the position of retrogradely labeled motoneurons onto the rhombomeric scaffold, preparations were additionally scanned with an illumination wavelength of 633 nm, outlining the spatial arrangement of the rhombomeres (r). Z-axis projections, image processing and quantification of neuronal numbers were carried out using the Fiji software package (http://fiji.sc/wiki/index.php/Fiji).

Tentacle motoneurons were retrogradely labeled 1–2 days prior to the experiment with Calcium Green-1 dextran (Life Technologies) by applying crystals of this Ca²⁺ sensor to the peripheral part of the tentacle motor nerve. Imaging of Ca²⁺ transients was performed with a confocal scanning microscope (LSM 700, Carl Zeiss, Germany) in the presence and absence of fictive locomotor activity. To prevent potential movement artifacts during imaging, all residual muscular elements of the isolated preparation were removed, except for the most caudal portion of the tail, which was left free to perform swimming movements (see above). Images were acquired at a rate of 5–10 frames s⁻¹ (ZEN black, Carl Zeiss, Germany), stored and analyzed post hoc. Image analysis was performed off-line using the Fiji software package and custom-written scripts in Igor Pro. Background fluorescence was subtracted and bleaching was corrected using a linear regression algorithm. All data are presented as relative changes in fluorescence (ΔF/F). The duration of Ca²⁺ transients were determined as the time at half-maximal amplitude of the fluorescence change (half-width) during a given swimming episode.

The authors thank Dr M. Hofmann for the Vidana software and D. Wickmaier for contributions to initial experiments.