Biol Cybern (2012) 106:669679 DOI 10.

1007/s00422-012-0528-0

PROSPECTS

Predictability of visual perturbation during locomotion: implications for corrective efference copy signaling
Boris P. Chagnaud John Simmers Hans Straka

Received: 23 July 2012 / Accepted: 9 October 2012 / Published online: 20 November 2012 Springer-Verlag Berlin Heidelberg 2012

Abstract In guiding adaptive behavior, efference copy signals or corollary discharge are traditionally considered to serve as predictors of self-generated sensory inputs and by interfering with their central processing are able to counter unwanted consequences of an animals own actions. Here, in a speculative reection on this issue, we consider a different functional role for such intrinsic predictive signaling, namely in stabilizing gaze during locomotion where resultant changes in head orientation in space require online compensatory eye movements in order to prevent retinal image slip. The direct activation of extraocular motoneurons by locomotor-related efference copies offers a prospective substrate for assisting self-motion derived sensory feedback, rather than being subtracted from the sensory signal to eliminate unwanted reafferent information. However, implementing such a feed-forward mechanism would be critically dependent on an appropriate phase coupling between rhythmic propulsive movement and resultant head/visual image displacement. We used video analyzes of actual locomotor behavior and basic theoretical modeling to evaluate head motion during stable locomotion in animals as diverse as Xenopus laevis tadpoles, teleost sh and horses in order to assess the potential suitability of spinal efference copies to
This article forms part of a special issue of Biological Cybernetics entitled Multimodal and Sensorimotor Bionics. B. P. Chagnaud (B) H. Straka Department Biology II, Ludwig-Maximilians-Universitt Mnchen, Grohaderner Strae 2, 82152 Planegg-Martinsried, Munich, Germany e-mail: b.chagnaud@gmail.com J. Simmers Institut de Neurosciences Cognitives et Intgrative dAquitaine, CNRS UMR 5287, Universit Bordeaux, 146 rue Lo Saignat, 33076 Bordeaux, France e-mail: john.simmers@u-bordeaux2.fr

the stabilization of gaze during locomotion. In all three species, and therefore regardless of aquatic or terrestrial environment, the head displacements that accompanied locomotor action displayed a strong correlative spatio-temporal relationship in correspondence with a potential predictive value for compensatory eye adjustments. Although spinal central pattern generator-derived efference copies offer appropriately timed commands for extraocular motor control during self-generated motion, it is likely that precise image stabilization requires the additional contributions of sensory feedback signals. Nonetheless, the predictability of the visual consequences of stereotyped locomotion renders intrinsic efference copy signaling an appealing mechanism for offsetting these disturbances, thus questioning the exclusive role traditionally ascribed to sensory-motor transformations in stabilizing gaze during vertebrate locomotion. Keywords Corollary discharge Spinal cord Gaze stabilization Vestibular system Vertebrates

1 Introduction The use of predictive feed-forward signals in the central nervous system (CNS) plays an important role in an animals sensory perception of its environment during behavior. So-called efference copies (von Holst and Mittelstaedt 1950) or corollary discharge (Sperry 1950) during the execution of a particular motor task are essential for the event-correlated detection and modulation of sensory signals during active probing of the environment by echo- (Suga and Jen 1975) or electrolocation (Bell 1981), for example. In such goal-directed foraging behaviors a temporally limited shunt in sensory sensitivity is required to allow the animal to distinguish between self-generated (reafferent) signals

123

670

Biol Cybern (2012) 106:669679

and the detection of information arising from the external world. In addition, an efference copy of motor commands can be employed as a predictive reference for calibrating and comparing unwanted self-induced sensory signals with sensory inputs from external sources. Such forward models of motor discharge are now recognized as indispensible for the precision of sensory perception and the production of robust and adaptive behavior in both invertebrates (Webb 2004) and vertebrates (Cullen 2004) alike (for recent review, see Crapse and Sommer 2008). Efference copy signaling plays an important role in the social interactions between animals in the acoustic and electroreceptive domains. The necessity to suppress vocal reafference in order to maintain receptivity to social partners is a requisite for most vocalizing animals. One of the best understood examples of acoustic reafferent suppression is the stridulating behavior of crickets (Poulet and Hedwig 2007) where, during vocalization, the stridulation central pattern generator (CPG) drives a separate set of interneurons that directly shuts-off central auditory neurons, without affecting peripheral auditory sensing. This inhibitory corollary discharge is rhythmically phase-locked to stridulating motor activity and serves to protect the crickets auditory sensory system from desensitization, allowing it to remain receptive to the acoustic environment. In sound-producing teleost sh, vocal motor system corollary discharge to the inner ear and lateral line organs is likely to decrease the sensitivity of endorgan hair cells (Weeg et al. 2005), thereby attenuating or shunting reafferent signals during auditory processing. In contrast, in echolocating bats, a corollary discharge during sonar calls is sent from the laryngeal motor cortex to the inferior colliculus to prevent the over-stimulation of central auditory circuits (Sinha and Moss 2007). Despite these diverse applications of corollary discharge signals in terms of both source and level of impact on sensory processing pathways, a unifying feature is the predictive sensory consequence of a selfgenerated action, thereby providing a reference framework for the actual sensory response. A more complex situation arises during locomotion or any other form of motor activity that causes relative body motion, such as rhythmic breathing. Locomotion is usually unrelated to sensory probing of the surrounding environment and generally deteriorates the perception of the external world in the absence of any compensatory mechanisms for the consequences of self-motion. All motor activity induces changes in sensory perception due to either a displacement of sensory structures (such as the eyes and ears) relative to an external target or stimulus source, or to the perception of the motor action by the sensory organs themselves (e.g., lateral line system in sh). Both conditions lead to an unwanted stimulation of the sensors in question, causing ambiguity in whether their input signals arise from the external environment (exafference), are self-generated by the animals own

movements (reafference), or are due to variable combinations of the two. This is particularly relevant in the case of sense organs that rely on a high spatio-temporal precision, such as the visual and auditory systems. Similarly, correct discrimination between the two signal sources is fundamental for animals to execute meaningful behavioral repertoires, like escape responses or prey capture. Based on the pioneering work of von Holst and Mittelstaedt (1950) and Sperry (1950), the theoretical framework for the employment of efference copies and/or corollary discharges has been established. Independent of the different terminology, the functional consequence of these signals is achieved by a subtraction of the negative blueprint of the impending motor activity from the expected sensory inputs (von Holst and Mittelstaedt 1950; Bell 1981). Moreover, the corollary discharge can occur at any given level of motor activity and is not necessarily a precise mirror image of reafferent neuronal activity, but rather, represents particular spatio-temporal aspects of the impending action and potential sensory stimuli (for review, see Crapse and Sommer 2008). With regard to the functional role of intrinsic efference copies during motor activities that impair sensory perception, two major consequences can be distinguished: 1. Temporal disruption of sensory encoding and 2. Maintenance of sensory encoding. A temporally-specic disruption of sensory encoding occurs during, e.g., vocalization, visual saccades, active electrolocation, or electroreception. During these behavioral repertoires, a corollary discharge causes a temporally limited reduction or shunt of the sensation and/or perception process in order to prevent a potentially over-stimulation of the sensory pathways by the induced motor activity and/or its sensory reafference signal. Consequently, perturbations of sensory encoding systems during navigation and orientation within the environment become minimized. For example, this has been demonstrated in weakly electric sh, which generate a local short-term electric eld, and bats, which emit ultrasound signals (Bell 1981; Suga and Jen 1975). In both cases, the activity-related sensory signals are peripherally or centrally suppressed in order to mask the emitted signal and to detect and enhance the exafferent component as the sensory image of the explored environment. In contrast, the maintenance of sensory encoding is attempted during motor activities that, as a secondary consequence, alter and often disturb the sensory encoding process. This is the case for lateral line receptors during swimming (Roberts and Russell 1972), sh electroreceptors during breathing (Montgomery and Bodznick 1994) or involuntary changes of eye position during locomotor-related head perturbations (Angelaki and Cullen 2008). In these cases, various mechanisms are implemented that allow counteracting the undesired disturbing consequences of sensory perception by subtracting the motor activity-related signals from those detected by the

123

Biol Cybern (2012) 106:669679

671

sensory organs or by stabilizing the position of the sensory organ (e.g., eyes and ears) relative to the environment. Confronting the disruptive consequences of behavioral actions for sustaining accurate perception of the visual world is common to all vertebrates, whether running, swimming, or ying. In the absence of any compensatory mechanism, self-generated movement causes an undesired continuous retinal image motion that deteriorates visual perception (Cullen 2004; Angelaki and Hess 2005). This is due to the fact that the eyes of vertebrates are tightly coupled to the head, which in turn is connected to the rest of the body and thus are constantly in motion during animal displacement. The degradation of visual signal processing results from an omnidirectional motion of the photoreceptors with respect to the environment. At higher motion frequencies and amplitudes, the visual scene becomes blurred, thus rendering precise visual perception impossible. However, the detrimental consequences for visual signal processing are counteracted by eye movements that minimize the retinal image slip. This is classically considered to be achieved by the transformation of bodymotion derived visuo-vestibular sensory feedback signals. As the major component, sensory signals from the balance organs are transformed into adequate motor commands for maintaining positional stability of the eyes in space. Such a positional stabilizing mechanism is provided by the vestibulo-ocular reex (VOR) that assists in counteracting the disturbing effects of locomotion on visual acuity (Angelaki and Cullen 2008). This motor reex employs sensory feedback signals during locomotor activity that are detected by several sets of vestibular endorgans and are subjected to appropriate sensory-motor transformation within the central vestibulo-extraocular circuitry (for review, see Straka and Dieringer 2004). However, this mechanism, which relies on reactionary sensory feedback signaling, is relatively slow and spatio-temporally constrained due to the ltering characteristics of the sensory periphery. At least theoretically, efference copy signaling provides an alternative mechanism for eliciting image-stabilizing eye movements during locomotion. In principle, intrinsic efference copies of the motor commands that generate locomotor propulsion offer a faster and more reliable substrate for counteracting the disruptive visual consequences of self-motion, especially since this CPG-driven behavior is rhythmic, relatively stereotyped and therefore produces predictable body and head motion. Indeed, earlier behavioral observations (Easter and Johns 1974; Harris 1965; Lyon 1900) as well as more recent electrophysiological studies (Combes et al. 2008; Lambert et al. 2012; Stehouwer 1987) have indicated that the spinal rhythmogenic circuitry might provide direct feed-forward signals to the extraocular motor nuclei which are appropriate for stabilizing ocular orientation during loco-

motion. To date, the evidence for such an intrinsic control mechanism has derived from animal models such as sh and larval amphibians that exhibit spatio-dynamically relatively simple locomotor behaviors. Undulatory swimming movements in these animals consist of a leftright bending of the trunk and tail along the longitudinal axis that causes head displacements in phase-opposition to tail excursions in the horizontal plane (Figs. 1, 2). Thus, the minimization of retinal image disturbances requires eye movements occurring essentially in the same plane and in counter-direction to the alternating head displacements. Direct evidence for the contribution of intrinsic efference copy signaling to gaze stabilization has been recently obtained in the isolated CNS of Xenopus tadpoles (Combes et al. 2008; Lambert et al. 2012). Such reduced in vitro preparations offer the possibility to record the motor output patterns that normally drive actual locomotor movements in the intact animal, consisting of bilaterally alternating rhythmic burst discharge in spinal ventral roots. During such so-called ctive locomotion in the absence of any actual body motion, the horizontal extraocular motor nerves were found to be cyclically active in time with ventral root bursts. Importantly, moreover, the latter occurred in a spatio-temporal relationship appropriate for producing conjugate eye movements that would counteract oppositely directed head displacements in the intact swimming animal (Beyeler et al. 2008). The absence of actual movement and thus sensory reafferent inputs during such spinal CPG-driven activity in these in vitro conditions conrmed the existence of a feedforward coupling mechanism between the spinal locomotor and the extraocular motor systems that is inherent to the CNS, and which would be functionally appropriate for stabilizing gaze during real locomotion. Such a process thus ts with the general forward-model concept (Webb 2004) of employing intrinsic corollary discharge to predict the consequences (here visual image disturbance) of self-motion in order to stabilize sensory (visual) perception. In the case of the tadpole, the predictability of visual degradation during steady swimming is facilitated by the relatively stereotyped, alternating left/right movements of the body and head, as well as by the mechanical contiguity between these two anatomical regions. Consequently, the expected retinal image disturbances resulting from the cyclic spatio-temporal alterations in head/body position during locomotion should be able to be determined with reasonable accuracy. One of the rst researchers to suggest a potential role for corollary discharge in stabilizing gaze was Sperry (1950). In an elegant set of experiments, in which the eyes of a teleost sh were experimentally rotated by 180 , he concluded that the visual system must receive a non-sensory, intrinsic anticipatory signal that encodes for both speed and direction, thereby promoting gaze stabilization. This conclusion was drawn from the nding of an accentuation of the

123

672 Fig. 1 Correlation of head and tail movements during swimming of a trout. a Superimposed color-coded outlines of the swimming animal obtained from three consecutive frames of a video sequence. (The video was provided by Dr. J. Liao, Whitney Labs, Miami). Crosspoints indicate the position of markers used in b. b Analysis of body/tail (black/green, left vertical axis) and head deections (red, right vertical axis) over several swimming cycles. Deections were normalized to percent changes in body length. c Color-coded cross-correlograms of body/tail and head deections. The high correlation values indicate a strong coupling of rhythmic body/tail and head motion. (Color gure online)

Biol Cybern (2012) 106:669679

Fig. 2 Correlation of head and tail movements during swimming in larval Xenopus. a Superimposed color-coded outlines of a swimming tadpole obtained from three consecutive frames of a video sequence. (The video was provided by Dr. M. Beraneck, University Paris Descartes). Crosspoints indicate position of markers used in b. b Analysis of the body/tail (black/green, left axis) and head deections (red, right axis) over several swim cycles. c Cross-correlogram of body/tail and head deections indicating motion coupling. (Color gure online)

movement of the rotated eye in the gaze (headeye) rather than in the opposite direction, as would be expected from the known functional organization of optokinetic reexes. These results were also in line with the earlier observations of von Helmholtz (1866), who proposed that during active head motion, the visual eld remains stable due to a counterbalancing mechanism that appears to be independent of sensory signals. Evidence for this conclusion was supported

by the blurring of the visual world during passive motion of the eyes with dynamics similar to those occurring during active head movements (von Helmholtz 1866). The increasing number of reports speculating on, or describing specic aspects of intrinsic sensory-independent mechanisms for gaze control prompted us to explore the basic requirements and limiting factors for the possibility that efference copies of locomotor commands maybe

123

Biol Cybern (2012) 106:669679

673

more generally involved in driving compensatory eye movements during vertebrate self-motion. However, the suitability of such an internal mechanism is critically dependent on the spatio-temporal relationship between actual locomotor movements and the resulting image motion and disturbance. Particularly crucial here is the reliability with which the rhythmic motor patterns for locomotion are translated into phase-coupled perturbations of the head/eyes and thus of image motion on the retina. In a rst step, therefore, we examined the relationship between locomotor movement dynamics and visual disturbance by employing video recordings of swimming in two aquatic animals (Xenopus tadpole and the trout) and two different (walking and running) quadrupedal movement patterns of a mammal (horse) on a treadmill for behavioral and motion-tracking analyses. These data were then combined with simple numerical modeling to assess the predictive value of intrinsic locomotor signals for gaze stabilization during different types of aquatic and terrestrial locomotor strategies. Based on our ndings, we conclude that locomotor-related corollary discharge in general can indeed provide a predictive extraocular motor command suitable for minimizing retinal image slip during locomotion, although the precision of compensatory eye motion requires presumably additional sensory feedback signals. Nevertheless, a combination of efference copy and sensory signaling related to head/body motion in space, rather than exclusively to sensory-motor transformations, offers a prospective bimodal substrate for retinal image stabilization during vertebrate locomotion.

rotations that require compensatory eye movements to maintain image stability on the retina. Accordingly, the degree of head displacement, the phase relation with propulsive activity during diverse locomotor behaviors and the extent to which a particular movement pattern provides a correlative reference for underlying spinal CPG activity was assessed. Video tracking of different body regions allowed quantifying the relative head motion during tadpole/trout swimming in a ow tank and horse walking and galloping on a treadmill.

2.1 Fish ow-tank swimming The undulatory body/tail excursions that drive forward propulsion in aquatic vertebrates such as sh and larval amphibians are generated by bilaterally alternating contractions of axial musculature, which propagate rostro-caudally along the trunk (Combes et al. 2004). The anatomically tight connection between body and head in most aquatic animals results in a strict phase coupling between the movements of these two structural components. As a result, the dynamics and magnitude of the leftright head excursions during swimming are intimately linked to the body undulations and therefore to the spinal CPG activity that drives the axial musculature. In theory, therefore, a likely consequence of this strict coupling between body/tail and head movement during stable locomotion is a highly predictable modulation of retinal image displacement. This hypothesis was rst tested by making a video analysis of stationary swimming of a trout in an opposing water ow of 45 cm/s (the video was kindly provided by Dr. J. Liao, Whitney Labs, Miami). A semi-manual, frame-by-frame detection of the spatial relationship between several visual markers placed along the longitudinal axis of an animals body was conducted with custom made software (courtesy of M. Hofmann). This enabled monitoring the trajectories of the trouts body/tail and head movements during swimming activity (Fig. 1a) and determining their temporal features and relationships (Fig. 1b). Both trajectories corresponded to regular leftright sinusoidal oscillations, although the amplitude of tail deection was 23 times larger (20 % of body length) than the associated head displacement (Fig. 1b) (8 % of body length). Importantly, however, cross-correlation analysis conrmed that the two signals occurred with a strict in-phase relationship that was indicative of a tight temporal coupling between body/tail and head movement (Fig. 1c). As a consequence, the retinal image displacements resulting from the head undulations would also occur in a highly predictable manner, in phase-locked coordination with propulsive body/tail motion. On this basis, therefore, intrinsic copies of spinal locomotor signals conveyed to appropriate sets of eye muscles would be suitably timed for counteracting the visual disturbance after adequate scaling.

2 Quantication of head motion during locomotion All vertebrates locomote by using n, limb, or trunk muscles to produce displacement of the body relative to the environment, such as during swimming, walking, or ying. The motor commands that drive the coordinated contractions of the various agonistic and antagonistic muscle sets responsible for a particular locomotor act are produced by so-called central pattern generators located primarily within the spinal cord (Orlovsky et al. 1999; Grillner 1981, 2006; Rossignol 1996). Rhythmogenic CPG networks are composed of bilaterally distributed premotor interneurons and motoneurons that generate precisely timed patterns of output for producing appropriate sequences of muscle contractions. Although triggered by startstop signals from supraspinal centers and modulated by sensory feedback pathways (Grillner 2006), spinal CPG networks are relatively autonomous in producing the basic motor output patterns that underlie largely stereotyped body motion in space, ranging from the undulatory swimming of aquatic vertebrates to running in terrestrial tetrapods including humans (Dietz 1992). Direct consequences of such regular propulsive movements are associated rhythmic head

123

674

Biol Cybern (2012) 106:669679

2.2 Larval Xenopus swimming Like most sh species, locomotion in amphibian larvae is also achieved by lateral undulations along the longitudinal axis (Azizi et al. 2007; Fig. 2a) produced by alternating left right contractions of epi-axial musculature (Beyeler et al. 2008). However, a major biomechanical difference between amphibian tadpoles and sh is the formers weaker anatomical linkage between the head/body and tail (Azizi et al. 2007). This lack of inter-regional rigidity presumably diminishes the strength of longitudinal coupling and thus, theoretically at least, reduces the predictability of retinal image displacements during locomotion. To quantify the coupling of body/tail during locomotion, a video analysis of swimming Xenopus tadpoles in a ow tank was also made as for the trout (the video was kindly provided by Dr. M. Beraneck, University Paris Descartes). Although larval frogs exhibit a positive rheotactic behavior (Simmons et al. 2004), in contrast to trout, these animals fail to perform steady stationary locomotion, but instead, swim in episodic bouts in order to maintain overall position in the water column. As a result, an associated drift in the tail and body movement trajectories occurs (Fig. 2b), thereby inuencing the subsequent correlation analysis as seen in the cross-correlogram of Fig. 2c. Nonetheless, the undulatory motion of the tadpoles tail during locomotor bouts produced noticeable head displacements (Fig. 2a, b). The more pronounced deection of the caudal tail compared to the trout was presumably due to a greater exibility of the tadpoles thinner tail and the weaker mechanical coupling with the rest of the body (note that for this reason, locomotor-related tail oscillations were measured at two points: one relatively rostral, at the level where the hindlimbs develop during subsequent metamorphosis [hereafter referred to as body] and the other at the tip of the tail). As in the trout, body and head motion were in a stable out-ofphase relationship, with the amplitude of tail motion being 3 times larger than the corresponding head displacements. Therefore, as further conrmed by the cross-correlation analysis (Fig. 2c), the swimming tadpole also displays robust spatio-temporal coupling between its tail and head/body during swimming behavior. 2.3 Horse treadmill stepping In contrast to the relatively simple propulsive strategies of adult sh and larval amphibian, locomotion in terrestrial vertebrates, in particular those with well developed and exible necks, generates head movements that are more complex, potentially less stereotyped and more weakly phase-coupled to locomotor movement. Thus, the resultant retinal image displacements could be similarly complicated and therefore potentially less predictable, especially in the case of quadrupeds where individual limb movements occur at multiple

phases of the locomotor cycle. Moreover, an increase in quadrupedal locomotor velocity is often achieved by a significant change in stepping pattern (or gait), rather than by a simple increase in rhythm frequency or amplitude of the same basic pattern as in the case of undulatory swimming in anamniotes. However, terrestrial vertebrates that increase speed through alterations in gait (e.g., walking to trotting to galloping in horse) might still produce head movements and associated retinal image displacements that remain spatio-temporally correlated with the rhythmic motor drive originating from the spinal CPG. The potential consequences for retinal image displacement during quadrupedal locomotion were therefore quantied by analyzing limbbodyhead movements of a horse on a treadmill from several publicly available sets of video records (Fig. 3). This analysis showed that galloping on the treadmill was accompanied by a rhythmic sinusoidal vertical modulation of the head in phase-coordination with foreleg stepping (Fig. 3a, b). Although, the magnitude of head motion varied considerably with the changing dynamics of the locomotor patterns, i.e., from walking to galloping (see below), head movements were clearly rhythmic at all times, and remained phase-coupled to specic phases of the foreleg motion cycle (Fig. 3c). Thus, despite the larger degrees of freedom in head movement compared to limbless aquatic vertebrates with their more compact biomechanical properties, quadrupedal vertebrates such as horses also display considerable stability of head motion and consequently rhythmic retinal image displacements that theoretically can be predicted by the locomotor CPG pattern. Similar results were obtained when the locomotor pattern of the horse changed from galloping to walking. During the latter, the right and left front legs alternate and both correlate with the head motion (Fig. 4a). However, this results in a temporal relationship in which head motion occurs in a movement trajectory that is twice the frequency of an individual limb. A better correlation was thus achieved by summing the left and right leg signals and correlating the resulting sum with the head displacement signal (Fig. 4b, bottom panel). In summary, the three examples presented heretadpole, trout and horseindicate that spinal CPG-driven locomotor rhythmicity produces visual disturbances that are phase-coupled with, and proportional to, ongoing locomotor performance and independent of the locomotor mode and the animals body plan. Accordingly, resulting sensory perturbations will be directly linked to the temporal patterning of the locomotor command itself. Thus, processes of gaze stabilization could in theory employ preemptive copies of locomotor output to directly elicit spatio-temporally appropriate ocular counter-rotations without the need for reactionary sensory feedback signals arising from locomotor-related head/body motion.

123

Biol Cybern (2012) 106:669679

675

Fig. 3 Correlation of horse head and leg movements during galloping. a Three different superimposed color-coded images from a movie of a horse running on a treadmill showing changes in head position during one step cycle. Crosspoint indicates position of marker used to

track head motion in b. b Trajectories of the front left and right legs (green/black, left axis) and head motion (red, right axis) during a gallop sequence. c Cross-correlogram indicating a strong coupling of rhythmic head and leg movements. (Color gure online)

Fig. 4 Correlation of horse head and leg movements during walking. a Trajectories of front left/right legs (green/black, left axis) and head motion (red, right axis) during a walking sequence. Bottom panel shows the sum (blue) of the left and right leg signals. b Cross-correlograms indicating strong coupling of rhythmic head and leg movements for both the left and right legs (color-coded; top panel) and their summed trajectories (bottom panel). (Color gure online)

3 Theoretical considerations of eye motion relative to visual distance In order to optimize and sustain visual acuity during locomotion, the magnitude of compensatory eye motion must vary with the distance of a given visual target in space: more distant targets will require smaller eye movements to keep retinal images stable, while accurate perception of a closer visual target necessitates a larger relative motion

of the retina and thus eye movements with correspondingly larger amplitudes, despite locomotor-derived head displacements remaining the same. To illustrate this relationship, we calculated and quantied the rhythmic vertical head/retinal image displacements that accompany human bipedal walking (Fig. 5). Specifically, a set of trigonometric functions was used (Fig. 5a) to assess the consequences of head and compensatory eye motion amplitudes at different focal distances during walking. This allowed calculating the

123

676

Biol Cybern (2012) 106:669679

Fig. 5 Computation of the variability in human head elevation with respect to visual target distance. a Schematic of the variables used in the calculations. Arrows indicate two variables that need to be compensated for during locomotion in order to maintain visual eld stability.

b Relationship between distance and angle from the eyes to xation points for different head elevations (head position color-coded from 1[red] to 10 [black]). Note that changes in head position or distance to target causes changes in the xation angle. (Color gure online)

Fig. 6 Summary of the most likely anatomical pathways mediating spinal CPG-derived efference copies for gaze stabilization during locomotion

tance and just sufcient eye motion amplitude must necessarily derive from sensory feedback signals from the visual system (target distance), or from the sensing of the relative position of the head on the neck (neck muscle proprioception) or relative to the earth gravitational vector (utricle). In the absence of such afferent information, locomotor efference copies would only be able to predict the direction of compensatory eye movements but would not compute the actual movement amplitude required for precise retinal image stabilization. Nevertheless, given that animals have relatively constant and species-specic ranges of viewing distance, it is likely that intrinsic locomotor-CPG signals would offer an overall gain that is reasonably adapted to a given animals preferential target distance.

relationship between the angle and the distance to xation points for different body sizes (110 a.u.). Regardless of body size, the resulting function was exponential, indicating that with increasing distance of the focal plane, the magnitude of eye motion required to stabilize gaze decreases (Fig. 5b). However, as also revealed by the plots, there was always a predictable relationship between distance and angle that is dependent on the height of the head above the ground. The above correlations have important implications for any compensatory eye motion that is attributable to spinal CPG efference copy signaling. To optimally employ this intrinsic mechanism for gaze stabilization, knowledge of the effective viewing distance to the visual target is required in order to ensure the appropriate gain of counteractive eye movements. However, the information required for computing the exponential relationship between viewing dis-

4 Central integration sites for spino-extraocular motor coupling The gaze-stabilizing coupling between spinal locomotor commands and the extraocular motor system could theoretically occur at several CNS levels that are not necessarily mutually exclusive. 4.1 Oculomotor nuclei The most direct interconnecting pathway could conceivably employ projections from spinal locomotor circuitry to appropriate bilateral populations of extraocular motoneurons in the mid- and hindbrain (Fig. 6, left). Depending on the spatial direction of locomotion-derived head perturbations that vary according to species, distinct sets of horizontal, vertical, and oblique extraocular motoneurons would have to be activated

123

Biol Cybern (2012) 106:669679

677

at different phases of each locomotor cycle. A direct spinoextraocular motor coupling has the advantage of fast signal transmission and thus would ensure tight phase coordination between the two motor systems. In sh and amphibian tadpoles, the integrative requirements for coupling appear to be relatively straightforward. The spatially restricted horizontal leftright head movements during their undulatory swimming essentially require oppositely directed conjugate eye movements produced by motoneurons innervating the medial and lateral rectus muscles (Combes et al. 2008). Consistent with this requirement, a direct connectivity has been recently established for the spinal CPG-extraocular motor coupling in Xenopus tadpoles (Lambert et al. 2012). In quadrupedal vertebrates where locomotor activity is often accompanied by complex 3D head displacements, appropriate activation of compensatory eye movements would require similarly complex, multiple connections with the different pools of extraocular motoneurons and a locomotor cycle phase-specic activation by efference copy signals. However, although the degrees of freedom are larger in these cases, a direct coupling mechanism is still conceivable. 4.2 Vestibular nuclei An alternative coupling pathway could involve the locomotor-CPG efference copies being rst integrated with sensory feedback signals in order to adjust for adaptive modications related to viewing distance. A potential site for such convergence is the hindbrain vestibular nucleus (Fig. 6, right), which also serves as the major integrating center for the transformation of passive head acceleration signals into extraocular motor commands for compensatory eye movements (Angelaki and Hess 2005). Although the necessarily additional synaptic connectivity would slow transmission compared to direct spino-extraocular coupling, conveying spinal efference copy signals via this nucleus offers the functional advantage of the entire VOR circuitry being available for a spatially-specic distribution of combined intrinsic feedforward and sensory feedback signals to the oculomotor nuclei. Indeed, some evidence for such an indirect vestibular pathway trajectory has been obtained in decerebrated guinea pig (Marlinsky 1992), where an increased ring rate occurs in vestibular neurons during locomotion. In Xenopus tadpoles, however, the focal blockade of synaptic transmission in the vestibular nuclei fails to disrupt spino-oculomotor coupling during ctive swimming, thereby indicating that here at least, vestibular nuclear circuitry is not interposed in the efference copy pathway (Lambert et al. 2012). 4.3 Cerebellum Given the important role of the cerebellum in gaze stabilization (Kheradmand and Zee 2011), this structure is a poten-

tial, additional site for conveying spinal efference copies to the extraocular motor system. The activity of vestibulo-spinal neurons is modulated prior to, as well as during, bouts of locomotor activity in cats (Orlovsky 1972; Udo et al. 1982), although significantly, this modulation is only observed in animals with an intact cerebellum. Given the prominent role of the cerebellum in adaptive plasticity of vestibular reexes (Gittis and du Lac 2006; Angelaki and Cullen 2008), this brain structure might serve as a convergence locus for self-motion-derived multisensory inputs such as visual, vestibular, and proprioceptive signals, as well as for intrinsic locomotor efference copies. After adequately integrating and combining the various spatio-temporally-specic signals, the premotor commands would then be distributed via well-established cerebellar pathways and VOR circuitry to extraocular motoneurons. 4.4 Neck motoneurons In contrast to most sh and amphibian larvae, vertebrates with a well developed and exible neck could also use this multi-segmental structure to facilitate gaze stabilization. In this way, locomotion-derived head movements would not simply involve passive pendulum-like displacements with respect to the body, but at least in theory, could be offset by counteractive contractions of neck muscles, thereby stabilizing head position and thereby gaze. However, such a compensatory mechanism could possibly also disturb or diminish locomotor performance since rhythmic head/neck motion often contributes itself to the effectiveness of an animals forward propulsion. Thus, compensatory eye movements rather than head motion would most likely remain the predominant target for intrinsic locomotor-related signals underlying gaze stabilization.

5 Summary Spinal CPG-driven locomotor activity is generally highly stereotyped and generates similarly rhythmic head perturbations and a deterioration of visual information processing in the absence of a compensatory mechanism. The inevitably tight coupling between locomotor rhythmicity and corresponding head displacements allows predicting the visual consequences during self-motion, independently of speciesspecic differences in locomotor patterns. The stereotypy of locomotion and the predictability of the resultant visual disturbance thus offer the necessary correlative substrate for relaying efference copies of spinal locomotor commands to the brainstem nuclei involved in ocular motor control. Even though such intrinsic predictive signals may be unable to ensure a spatio-temporally perfect stabilization of retinal image position during propulsive movement, they are

123

678

Biol Cybern (2012) 106:669679 Dietz V (1992) Human neuronal control of automatic functional movements: interaction between central programs and afferent input. Physiol Rev 72:3369 Easter SS, Johns PR (1974) Horizontal compensatory eye movements in goldsh (Carassius auratus). J Comp Physiol A 92(1):3757. doi:10.1007/bf00696525 Gittis AH, du Lac S (2006) Intrinsic and synaptic plasticity in the vestibular system. Curr Opin Neurobiol 16(4):385390. doi:10.1016/ j.conb.2006.06.012 Grillner S (1981) Control of locomotion in bipeds, tetrapods, and sh. In: Brookhart JM, Mountcastle VB (eds) Handbook of physiologythe nervous system II. American Physiological Society, Bethesda, pp 11791236. Grillner S (2006) Biological pattern generation: the cellular and computational logic of networks in motion. Neuron 52(5):751766. doi:10.1016/j.neuron.2006.11.008 Harris LR (1965) Visual motion caused by movements of the eye, head and body. In: Smith AT, Snowman RJ (eds) Visual detection of motion. Academic Press, London, pp 397435 Kheradmand A, Zee DS (2011) Cerebellum and ocular motor control. Frontiers Neurol 2. doi:10.3389/fneur.2011.00053 Lambert FM, Combes D, Simmers J, Straka H (2012) Gaze stabilization by efference copy signaling without sensory feedback during vertebrate locomotion. Curr Biol 22:16491658 Lyon EP (1900) Compensatory motion in shes. Am J Physiol 4(2): 7782 Marlinsky VV (1992) Activity of lateral vestibular nucleus neurons during locomotion in the decerebrate Guinea pig. Exp Brain Res 90(3):583588. doi:10.1007/bf00230942 Montgomery JC, Bodznick D (1994) An adaptive lter that cancels self-induced noise in the electrosensory and lateral line mechanosensory systems of sh. Neurosci Lett 174(2):145148 Orlovsky GN (1972) Activity of vestibulospinal neurons during locomotion. Brain Res 46(0):8598. doi:10.1016/ 0006-8993(72)90007-8 Orlovsky GN, Deliagina TG, Grillner S (1999) Neuronal control of locomotion: from mollusc to man. Oxford University Press, New York Poulet JFA, Hedwig B (2007) New insights into corollary discharges mediated by identied neural pathways. Trends Neurosci 30(1):1421 Roberts BL, Russell IJ (1972) The activity of lateral-line efferent neurones in stationary and swimming dogsh. J Exp Biol 57(2):435448 Rossignol S (1996) Visuomotor regulation of locomotion. Can J Physiol Pharmacol 74(4):418425 Simmons AM, Costa LM, Gerstein HB (2004) Lateral line-mediated rheotactic behavior in tadpoles of the African clawed frog (Xenopus laevis). J Comp Physiol A 190(9):747758 Sinha SR, Moss CF (2007) Vocal premotor activity in the superior colliculus. J Neurosci 27(1):98110 Sperry R (1950) Neural basis of the spontaneous optokinetic response produced by visual inversion. J Comp Physiol Psychol 43: 482489 Stehouwer DJ (1987) Compensatory eye movements produced during ctive swimming of a deafferented, reduced preparation in vitro. Brain Res 410(2):264268. doi:10.1016/0006-8993(87)90323-4 Straka H, Dieringer N (2004) Basic organization principles of the VOR: lessons from frogs. Prog Neurobiol 73:259309 Suga N, Jen PH (1975) Peripheral control of acoustic signals in the auditory system of echolocating bats. J Exp Biol 62(2):277311 Udo M, Kamei H, Matsukawa K, Tanaka K (1982) Interlimb coordination in cat locomotion investigated with perturbation. Exp Brain Res 46(3):438447. doi:10.1007/bf00238638 von Helmholtz H (1866) Handbuch der Physiologischen Optik. English translation by J. P. C. Southall (1924) for the Optical Society of America.

excellently suited to phase-couple extraocular motor output with locomotor activity. In this context also, it is conceivable that the fast feed-forward intrinsic commands are complemented by reactionary sensory feedback signals that yield the necessary information for ensuring the complete precision of the dynamic stabilizing process. However, recent evidence from tadpoles has indicated that during regular swimming and in direct contrast to the spinal efference copy, the horizontal semicircular canal-derived VOR is actively suppressed and plays no role in the gaze-stabilizing process (Lambert et al. 2012). Finally, from an evolutionary perspective, since locomotion through undulatory body movements was already part of the behavioral repertoire of pre-vertebrate ancestors, it is possible that the locomotor efference copy signaling served as the initial mechanism for maintaining the rst movable eyes xed in space during self-motion, before more complex and specic body-motion sensing vestibular organs had evolved.
Acknowledgments The authors thank J. Leo van Hemmen for inviting us to participate in this special issue. In addition we thank Drs. M. Beraneck and J. Liao for permission to use their videos of swimming trout and Xenopus laevis, respectively, and Dr. M. Hofmann for use of video-tracking software. This work was supported by the German Science Foundation (SFB 870, B12) and by funds from the German Federal Ministry of Education and Research under Grant code 01 EO 0901. The collaboration between the two laboratories was nancially supported by the Bayerisch-Franzsisches Hochschulzentrum (BFHZ).

References
Angelaki DE, Cullen KE (2008) Vestibular system: the many facets of a multimodal sense. Annu Rev Neurosci 31:125150 Angelaki DE, Hess BJ (2005) Selfmotion-induced eye movements: effects on visual acuity and navigation. Nat Rev Neurosci 6:966 976 Azizi E, Landberg T, Wassersug RJ (2007) Vertebral function during tadpole locomotion. Zoology 110(4):290297. doi:10.1016/j.zool. 2007.02.002 Bell C (1981) An efference copy which is modied by reafferent input. Science 214:450453 Beyeler A, Mtais C, Combes D, Simmers J, Le Ray D (2008) Metamorphosis-induced changes in the coupling of spinal thoraco-lumbar motor outputs during swimming in Xenopus laevis. J Neurophys 100(3):13721383 Combes D, Merrywest SD, Simmers J, Sillar KT (2004) Developmental segregation of spinal networks driving axial- and hindlimbbased locomotion in metamorphosing Xenopus laevis. J Physiol 559(1):1724 Combes D, Le Ray D, Lambert FM, Simmers J, Straka H (2008) An intrinsic feed-forward mechanism for vertebrate gaze stabilization. Curr Biol 18:R241243 Crapse TB, Sommer MA (2008) Corollary discharge across the animal kingdom. Nat Rev Neurosci 9(8):587600 Cullen KE (2004) Sensory signals during active versus passive movement. Curr Opin Neurobiol 14:698706

123

Biol Cybern (2012) 106:669679 von Holst E, Mittelstaedt H (1950) Das Reafferenzprinzip. Naturwissenschaften 37:464476 Webb B (2004) Neural mechanisms for prediction: do insects have forward models? Trends Neurosci 27:278282

679 Weeg MS, Land BR, Bass AH (2005) Vocal pathways modulate efferent neurons to the inner ear and lateral line. J Neurosci 25(25):5967 5974

123