A SOMEWHAT NEGLECTED ASPECT OF INSECT NEUROPHYSIOLOGY

Tom Matheson discusses Keir Pearson's 1972 paper entitled ‘Central programming and reflex control of walking in the cockroach’.

A copy of the paper can be obtained from http://jeb.biologists.org/cgi/content/abstract/56/1/173

Keir Pearson began his influential 1972 paper on cockroach walking with the statement that “a somewhat neglected aspect of insect neurophysiology is the nervous control of walking”. He then proceeded to outline a series of experiments that, together with his earlier work (Pearson and Iles, 1970), helped to shape a very large body of subsequent research that has shown how the central nervous system and sensory signals act together to control limb movements. Almost 40 years later, his 1972 paper continues to be cited four or five times a year. The list of citing authors reads like a who's who of motor control and neuroethology, and the 240 or so citations span topics as diverse as vertebrate spinal cord pattern generation (Grillner, 1974), motor deficits in patients with large-fibre sensory neuropathy (Sanes et al., 1984), artificial neural nets for controlling legged robots (Cruse et al., 1995), and motor memories in foraging wood ants (Lent et al., 2009). This diversity is testament to the fundamental nature of the questions addressed by Pearson's experiments and to his clarity of expression, which makes his paper very accessible.

As Keir himself pointed out, by the early 1970s there was a growing body of work describing both central and peripheral neuronal contributions to the motor control of repetitive behaviours such as stridulation, breathing and flight. Extracellular nerve and muscle recordings had been used since the 1940s – a great deal of the resulting work being published in The Journal of Experimental Biology by leaders in the field such as Pringle (e.g. Pringle, 1939) – and intracellular studies were beginning to appear (Hoyle, 1970).

For walking, however, there was considerable uncertainty about the relative roles of central and peripheral signals in generating the motor pattern. Charles Sherrington had shown that stepping reflexes could be elicited in cats whose spinal cords were severed (Sherrington, 1906); and, in locusts, the removal of leg sensory receptors was known to markedly disrupt leg movements (Usherwood et al., 1968; Wendler, 1966). Such experiments had given rise to a strong notion that walking rhythms were largely generated by sensory reflexes – although this view was beginning to lose favour (Wilson, 1966).

Pearson set out to resolve this question. To do so, he focussed on the hind legs of the cockroach Periplaneta americana and, in particular, on a subset of the muscles that raise (levate) and lower (depress) the limbs during the swing and stance movements of a step cycle, respectively. In 1970, he and John Iles had used restrained, headless and completely deafferented animals (i.e. lacking a brain or any sensory feedback from the leg) to demonstrate for the first time that alternating patterns of levator and depressor motor activity similar to those seen in walking were generated centrally in the absence of leg sensory feedback. Put simply, the basic motor pattern was generated within the central nervous system (Fig. 1). This decisive finding was a turning point because it laid to rest the idea that the motor pattern might be generated by a chain-reflex mechanism that was dependent upon the activation of sensory receptors in the limbs. Pearson's subsequent 1972 paper provided important additional experiments along with a lucid discussion that synthesised the two studies.

In the first part of the 1972 paper, from work carried out in restrained cockroaches, Pearson confirmed the 1970 observation that deafferented animals could produce centrally generated reciprocal patterns of activity in the coxal motor neurones. But, this new work also addressed a problem of the earlier study using headless animals: that it had not been possible to fully disentangle the effects of descending brain signals from those of sensory signals from the legs – a problem that had left some ambiguity in the interpretation of the earlier results. The 1972 experiments also extended the earlier result by showing that removal of leg sensory signals led to changes in the motor pattern; most notably an increased activation of one of the motor neurones, the slow depressor coxae D_s. This second important result provided strong evidence that sensory feedback acted to modify the centrally generated pattern of activity.

To determine whether his results were meaningful in freely walking cockroaches, Pearson recorded the motor neurone signals using flexible electromyogram wires inserted directly through the cuticle into the muscles of interest. The cockroaches were able to walk or run freely in an arena, and could be given weights to carry or pull in different ways. Pearson's recordings provided beautifully clear examples of alternating activity in the antagonistic levator and depressor coxae motor neurones during real walking, and formed the basis for quantitative analyses of specific relationships between the bursts of motor activity.

These analyses gave important insights into the detailed action of individually identified motor neurones during walking. The electromyogram recordings showed that changes in walking and running speed were associated with the recruitment of different motor neurones. The first of several further key observations was that as walking speed changed, the durations of the bursts of activity in the depressor motor neurone D_s changed much more than did those of the levators. This correlated well with contemporary behavioural data for walking (Delcomyn, 1971), which showed that changes in walking speed affected the stance phase (to which the depressor contributes) much more markedly than the swing phase. Pearson was also able to describe how the motor activity was patterned within a burst, and to show how this changed with walking speed or experimental manipulation. Crucially, loading the animals led to alterations of motor activity, so sensory feedback here – as in the dissected preparations – clearly influenced the overall pattern. Moreover, Pearson deduced that the sensory feedback was much less important at the fastest walking speeds. One of the nicest lines of argument employed in this part of the paper was that the stance phase coxal depressor motor activity was intensified at times when there was self-induced loading of the leg as the other ipsilateral legs entered the swing phase and thus no longer contributed to the support of the body mass (Fig. 2). Now, almost 40 years later, it is becoming increasingly clear that passive biomechanical properties of the limbs provide additional mechanisms for stabilising limb movements alongside sensory feedback (Zakotnik et al., 2006; Proctor and Holmes, 2010). Such mechanisms might dominate at high stepping frequencies when there is insufficient time for sensory signals to reach the central nervous system and exert their influence.

To identify the sensory signals that might contribute to the load-induced changes in motor activity, Pearson carefully denervated different regions of the leg before observing whether or not an added load could still influence the motor pattern. These experiments ruled out major contributions from some of the coxal receptors, and led Pearson to suggest that the trochanteral campaniform sensilla might be responsible. These small dome-like sensory receptors on the leg were already known to excite the depressor motor neurone D_s in response to strain in the leg cuticle (Pringle, 1940), but nobody knew whether they actually responded to limb forces during normal walking. More recent work on these and other leg campaniform sensilla has provided strong indications that they do respond during walking (e.g. Hofmann and Bässler, 1982; Delcomyn, 1991; Noah et al., 2001), and their artificial stimulation has now been shown to affect ongoing walking motor patterns (Schmitz, 1993; Akay et al., 2007).

Pearson's 1972 paper was presented in an exceptionally clear and lucid style of writing, which remained tightly focussed on the questions at hand. Its accessibility has undoubtedly contributed to the paper's long-lasting influence on the field. It touched on several key questions that have exercised sensory motor neurophysiologists of both vertebrate and invertebrate persuasions over the subsequent 40 years: load compensation, gating of reflexes during active behaviour, and the site and nature of feedback influences on central networks. Pearson provided a concise summary of the rapidly changing state of knowledge at the time. For instance, he proposed a model for the observed patterns of coordination of the coxal depressor and levator motor neurones in the cockroach, and used this to explain possible mechanisms by which limb proprioceptors of locusts might have caused the profound effects of sensory manipulation seen by Usherwood et al. (Usherwood et al., 1968). In the light of subsequent intracellular studies and an increased understanding of the patterns of connectivity between the various motor neurones and interneurones involved in leg movements, some details of Pearson's basic model seem now to be incorrect – and others remain untested (e.g. Burrows, 1996) – but it nevertheless provided a significant stimulus and focus for much subsequent work.

An important area of modern biomedical research is the design of effective limb prostheses controlled by brain signals (‘neural prostheses’ or ‘brain-machine interfaces, BMIs’). Despite remarkable progress in this area (e.g. Velliste et al., 2008), one fundamental remaining problem is how best to incorporate sensory signals into control systems driven by brain activity: a question that was already at the forefront of Pearson's mind some 40 years ago – albeit in a rather different context.