Review
Proprioceptive control of posture: a review of new concepts
J.H.J. Allum
a,*, B.R. Bloem
b, M.G. Carpenter
c, M. Hulliger
d, M. Hadders-Algra
eaDepartment of ORL, Uni6ersity Hospital, Basel, Switzerland
bDepartment of Neurology, Leiden Uni6ersity Medical Center, Leiden, The Netherlands cDepartment of Kinesiology, Uni6ersity of Waterloo, Waterloo, Ont., Canada dDepartment of Clinical Neurosciences, Uni6ersity of Calgary, Calgery, Alta., Canada eDepartment Medical Physiology, Uni6ersity of Groningen, Groningen, The Netherlands
Received 26 February 1998; received in revised form 27 May 1998; accepted 8 June 1998
Abstract
The assumption that proprioceptive inputs from the lower legs are used to trigger balance and gait movements is questioned in this review (an outgrowth of discussions initiated during the Neural Control of Movement Satellite meeting held in Cozumel, Mexico, April 1997). Recent findings presented here suggest that trunk or hip inputs may be more important in triggering human balance corrections and that proprioceptive input from the lower legs mainly helps with the final shaping and intermuscular coordination of postural and gait movements. Three major questions were considered. First, what role, if any, do lower-leg proprioceptive inputs play in the triggering of normal balance corrections? If this role is negligible, which alternative proprioceptive inputs then trigger balance corrections? Second, what is the effect of proprioceptive loss on the triggering of postural and gait movements? Third, how does proprioceptive loss affect the output of central pattern generators in providing the final shaping of postural movements? The authors conclude that postural and gait movements are centrally organized at two levels. The first level involves the generation of the basic directionally-specific response pattern based primarily on hip or trunk proprioceptive input and secondarily on vestibular inputs. This pattern specifies the spatial characteristics of muscle activation, that is which muscles are primarily activated, as well as intermuscular timing, or the sequence in which muscles are activated. The second level is involved in the shaping of centrally set activation patterns on the basis of multi-sensorial afferent input (including proprioceptive input from all body segments and vestibular sensors) in order that movements can adapt to different task conditions. © 1998 Elsevier Science B.V. All rights reserved.
Keywords: Balance corrections; Infant development; Leg movements; Locomotion; Postural control; Proprioceptive feedback; Sensory neuropathy
Introduction
Our present concepts regarding the influence of nor-mal proprioceptive function (propriocepsis) on human postural responses stem, first of all, from clinical obser-vations. Epidemiological surveys have established that a reduction of leg propriocepsis is a risk factor for falls in the elderly [82,102,115]. The importance of intact
propriocepsis for maintaining upright stance is further underscored by the sometimes devastating clinical con-sequences induced by a loss of propriocepsis, for exam-ple in patients with peripheral neuropathy. The resultant balance disorder can be very impressive, and accounts of such patients continue to reach not only the scientific literature [25,30,47], but also the popular press [38,107]. These clinical observations raise a number of scientific questions: If our footing slips, we trip on some unforeseen obstacle, or if we miss a stairway step, is the ensuing proprioceptive information from the ankle joints the most important factor shaping the balance
* Corresponding author. Present address: University HNO-Klinik, Petersgraben 4, CH-4031 Basel, Switzerland. Tel.: + 41 61 2652040; fax: + 41 61 2652750; e-mail: allum@ubaclu.unibas.ch
0966-6362/98/$ - see front matter © 1998 Elsevier Science B.V. All rights reserved.
correction or are other sensory inputs, more directly related to the primary task of posture and gait, to maintain a stable torso of greater importance? If pro-prioceptive signals from other joints such as the knee, hip and neck together with vestibular signals are more important in triggering and modulating balance correc-tions, how does ankle proprioceptive input contribute to balance corrections and modify gait patterns? Given that the central nervous system (CNS) must integrate these sensory signals together to yield a single motor output how is the process acquired during development or reacquired following sensory loss? These and related questions are explored in five original contributions which examined the role of proprioceptive inputs in the control of posture and gait. These contributions examined:
1. The interactions between proprioceptive and vestibular inputs in generating human balance corrections;
2. how these trunk proprioceptive and vestibular inter-actions change balance corrections in different directions;
3. the changes in balance corrections caused by loss of proprioceptive input from the lower legs;
4. the changes in gait coordination in cats caused by global proprioceptive loss;
5. how the developing human infant gradually devel-ops a fixed pattern of balance corrections appropri-ate for the direction of perturbed trunk motion. The quintessence of these contributions is that ankle proprioceptive inputs are not essential for triggering balance corrections and gait movements but may play some role in shaping the final form of the muscle synergy underlying both types of movements.
The new synthesis presented in these contributions highlights the control by CNS in the selection of appro-priate balance correcting and gait timing and amplitude patterns. This central processing implies a central syn-thesis of relevant afferent information which could be simplified by relegating one or more parts of this task to central pattern generators (CPG). One of these task simplifications would be for the CPG to select the basic timing of postural responses of balance corrections [42] prior to selection of the response scaling at different muscles. Allum et al. [7] have argued that in man this timing pattern selection is set on the basis of a number of links that the body is forced to move with at the onset of a balance disturbance. For example, if the disturbance forces the knees into a locked position so that the legs move as one rather than two elements, it is to be expected that trunk, knee and ankle muscles will provide different proprioceptive trigger signals than if the knees are not locked but flexing. Another way the CNS may simplify the task of computing the correct amplitude response shaping is to pre-weight groups of muscles to receive more proprioceptive or more
vestibu-lar weighting. Presumably the selection of which mus-cles are more shaped by vestibular inputs and which have emphasized proprioceptive weighting during the control of trunk forward pitching preventing total body rearward falling, is based on a learning process in early infancy in which the growing infant employs these muscles to prevent a backwards fall as Hadders and Forssberg [58 – 60] point out in their contribution. The fact that subjects who suffer vestibular loss or proprio-ceptive loss as adults cannot switch this weighting between muscles to prevent a fall even after years of experience controlling falls [12], implies that this basic level of pre-processing of balance corrections cannot be modified and is pre-set. Thus, the final central compu-tation of the metrics of the balance-correcting synergy would be the selection of a base-level amplitude of muscle activation about which proprioceptive and vestibular inputs would interact by increasing or de-creasing this base-level of activation. We hypothesize that this basic activation level is rapidly selected using prediction of the amplitudes with which the trunk and legs will move as the balance correction occurs. With this organization sequence the CNS would have the advantage of a rapid reaction to any postural distur-bance and have computational freedom to make minor postural adjustments during the following stabilizing period when the body is repositioned in a new upright position with low levels of muscle activity.
Originally, it was proposed that for balance correc-tions ankle inputs trigger responses in stretched lower-leg muscles and that this trigger signal is then transmitted in a distal-to-proximal fashion upwards to elicit a balance-correcting muscle response synergy with onsets of 100 – 120 ms across a number of links [36,68,94,95]. The movement of the body which resulted was attributed to resemble the action of an inverted pendulum-the so-called ‘ankle strategy’ [67]. The first feature of this concept that was called into question concerns the distal-to-proximal triggered activation of automatic balance-correcting muscle responses. Allum et al., and Keshner et al. [7,75] used similar support-surface induced balance perturbations noted triggered responses in the trunk and neck muscles that occurred at the same time as the proposed triggering response at 100 ms in gastrocnemius [67,94]. Moreover, the pro-posed triggering response at 100 ms in gastrocnemius is delayed compared with the earlier 50 ms latency of soleus muscles to stretch of the ankle muscles and the next response observed at 80 ms in stretched quadriceps muscles [7]. Thus, it appears that proprioceptive reflex systems in other than the ankle muscles could well trigger postural responses with onsets of 100 – 120 ms. Following this line of reasoning, a number of authors have suggested that rotation of the trunk [9,42] rotation of the knee [7] or more distally, stretch of the intrinsic foot muscles [113] could trigger postural responses. It
was therefore, the role of proprioceptive signals other than those from the ankle joint which we wished to explore in this set of contributions.
Patterns of proprioceptive and vestibular interactions underlying human balance corrections
J.H.J. Allum, F. Honegger
Department of ORL, Uni6ersity Hospital, Basel, Switzerland
Balance corrections must be triggered from sensory sources which early and reliably detect the disturbance to the upright stance. Ideally, this sensory information would also be employed to establish the timing and metrics of balance corrections. Instability in human balance can be registered very early by proprioceptive sensory systems responding to motion and muscle stretch at the ankle, knee and trunk joints, as well as by head linear and angular accelerations [7]. Once balance corrections are triggered it is generally accepted that a confluence of proprioceptive and vestibular modulation to the basic centrally initiated template of activity es-tablishes the amplitude pattern of the muscle response synergy [9,35,42,70]. Visual inputs appear to primarily influence later stabilizing reactions to the initial balance corrections [5,7,74]. It is generally assumed that the contributions of one sensory input can be assessed by testing subjects suffering from a deficit of the to-be-tested sensory input [9,42,71], or by perturbing one sensory input alone while holding the others constant [12,36,41] or by perturbing one sensory input a constant amount and allowing other inputs to vary [6,7]. Most of these techniques have inherent disadvantages as we will document. The deficit approach assumes, for example, that absence of vestibular input does not affect proprio-ceptive stretch reflexes prior to balance corrections and therefore the body movements and proprioceptive mod-ulation during balance corrections. Furthermore, when perturbing one sensory input alone, the metrics of the perturbation are seldom exactly those encountered dur-ing a stance disturbance with multiple inputs. Lastly multiple sensory inputs, for example, trunk and vestibu-lar inputs, may be required for a third input at the knee to have a fully developed modulating effect. By specifi-cally controlling for variations at the ankle and knee joints, and comparing responses of normal and vestibu-lar-loss subjects to stance perturbations we hoped to overcome some of these disadvantages and shed some light onto the complexity of proprioceptive and vestibu-lar interactions during balance corrections.
1. Methods
The techniques used to probe balance control mecha-nisms in human subjects have been described in Refs. [5,7]. Subjects stood on a support-surface which could either rotate about the ankle joints or translate horizon-tally. Stimuli consisted of one of three types of balance perturbations which were presented in random order 10 times for a total of 30 in a series. The first series was presented under eyes-open conditions and, after a 5 – 10 min pause, the second identical series was presented with eyes closed. The three types of stimuli were 4° dorsi-flexion rotation (‘normal’ ankle – angle protocol), or a combined 4 cm rearward translation and 4° plan-tar-flexion rotation to yield approximately 0° of ankle dorsi-flexion (‘nulled’ ankle – angle protocol), or a com-bined 4 cm rearward translation and 4° dorsi-flexion to yield approximately 6° of ankle dorsi-flexion (‘en-hanced’ ankle – angle protocol). Stimulus durations were 150 ms. Over the first 200 ms of the ‘nulled’ and ‘enhanced’ ankle angle protocols, a separate micropro-cessor provided extra servo-rotation of the support-sur-face. The purpose of this rotation was to keep ankle angle as measured between the lower-leg and the sup-port-surface zero for the ‘nulled’ protocol and equal to the average measured profile of normals for the ‘en-hanced’ protocol. The subjects (15 normal subjects and five otherwise healthy subjects who had a bilateral peripheral vestibular deficit) were asked to return to upright as quickly as possible in response to each stimulus.
Several measurements were recorded as in previous studies in the pitch plane: trunk angular velocity (Wat-son Industries 9200°/s range, 50 Hz bandwidth); up-per leg angular velocity (Watson Industries 9100°/s range); lower leg angle with respect to vertical (using a potentiometer system); and ankle torque about the left and right ankle joints (from the outputs of strain gauges imbedded in the support-surface).
Surface electromyographic (EMG) recordings were obtained from the left and right tibialis anterior (TA) and soleus muscles, and from the right medial gastroc-nemius, right quadriceps, paraspinals (PARAS) and upper trapezius (TRAP) muscles, using pairs of surface electrodes placed 3 cm apart along the muscle belly. For data analysis purposes, the first three responses were ignored to avoid adaptation effects entering popu-lation responses [74]. The remaining nine responses from each subject to the same rotation – translation protocol were averaged together once a zero latency for the responses had been defined. Zero latency (stimulus onset) was defined using the computed velocity of ankle dorsi-flexion. The first inflexion of this velocity trace was used as zero latency (the vertical line at 0 ms in Fig. 1). Average responses from all subjects in the same population tested under the same condition (eyes open
Fig. 1. Differences between the responses of normal and vestibular-loss subjects to a simultaneous rearward translation and dorsi-flexion of the support-surface yielding a 6° ankle dorsi-flexion. The balance correcting responses in soleus and gastrocnemius are not altered by vestibular loss but those of paraspinals are profoundly changed. Muscle activation patterns of five subjects with bilateral peripheral vestibular-loss (thick traces) are compared to the average responses of 15 normal subjects (thin traces). The traces are aligned in time with the first deflection of ankle angular velocity (similar to the first deflection in ankle torque). Rearward rotations of angles and angular velocities are plotted as positive (upward) deflection of traces as is decreased ankle torque on the platform. The insert shows the time course of head linear and angular accelerations over the first 150 ms. Upward linear accelerations are plotted as a positive trace deflections as is head backwards pitching angular acceleration. Other methodological details may be found in Ref. [12].
or closed) were used to create the population averages shown in Fig. 1.
2. Results
This report concentrates on the responses to com-bined rearward translation and dorsi-flexion rotation of the support-surface (‘enhanced’ ankle – angle protocol). As Fig. 1 shows this produced identical profiles of a rapid 6° ankle flexion, identical traces of first a knee flexion followed by a rapid knee extension as the upper leg rotated forward, and similar ankle torque record-ings over the first 200 ms for the two populations and conditions. Despite these identical lower and upper leg movement profiles significant differences were observed in the way muscle activity was modulated following vestibular loss. Furthermore, trunk angular velocity traces had dissimilar profiles for normal and vestibular-loss subjects and were accompanied by paraspinal re-sponses different in latency and amplitude modulation. The differences (or lack thereof) between the muscle activity in normal and vestibular-loss subjects could be classified into three types of modulation. Triceps surae muscles were not influenced by vestibular loss. Fig. 1 shows that the soleus and gastrocnemius activity were practically overlapping in both normal and vestibular-loss subjects for eyes open and closed conditions. The activity consisted of two bursts, the first at 56 ms was larger in amplitude in soleus, and the second at 106 ms was larger in amplitude in gastrocnemius. These onset latencies are for normals with eyes closed. The onset latencies were on average 2 ms, respectively, 5 ms shorter for vestibular-loss subjects. The earlier onset in vestibular-loss subjects was probably due to the slightly forward leaning posture adopted by these subjects as can be observed from the increased baseline activity in soleus over the 100 ms prior to the stimulus onset. It is worth noting that the difference in latency onsets of the first and second burst of triceps surae activity is similar to the latency of onset of upper leg rotation suggesting that the second burst of activity may be a stretch reflex response to knee reextension.
The second type of vestibular modulation pattern appeared in quadriceps. It consisted of a triphasic change in the difference between normal and vestibular-loss responses. The most noticeable aspect of the re-sponse difference was the change in the amplitude of the stretch reflex response in quadriceps induced by knee flexion. The amplitude of this activity commencing at 80 ms was reduced in patients compared to controls during the time that the head was accelerated down-wards (see insert at the top right of Fig. 1).
Responses in TA, PARAS and TRAP muscles pre-sented a third type of modulation. The pattern was characterised by an early reduction in activity after 120
ms in the muscle responses of vestibular-loss subjects, followed by a major increase in activity. Although there was no significant shift in latency for TA and TRAP muscles, the overall impression was a shift in the onset of the balance correcting responses in vestibular-loss subjects. This overall impression could be confirmed in the case of PARAS as latencies of the balance correct-ing responses were significantly later, some 22 ms on average. This change was accompanied by a reduction in the early extension velocity of the trunk and by an elongation of the period of early decreased PARAS activity present for vestibular-loss subjects. That is, the early decreased activity commencing at ca. 80 ms is an unloading reflex caused by early trunk extension. The main effect of the balance correcting activity in PARAS is presumably to break the forward flexion of the trunk. As this is some 30% faster at 250 ms in vestibular-loss subjects, it is not surprising that PARAS activity is also increased at this time. In addition, the increased trunk flexion would require the head to be flexed back more in order to maintain gaze stable in space, thus extra neck extensor activity in TRAP muscles would be nec-essary in vestibular-loss subjects as we indeed observed.
3. Discussion
The current set of results emphasise features of bal-ance correcting responses that have been known for some time. Vestibular and proprioceptive inputs both contribute to the modulation of the muscle response synergy [5,8,68,74,95]. The special feature focussed on this report is the differential weighting of vestibular and proprioceptive inputs to the activity of different groups of muscles. We wish to emphasize that an interaction of proprioceptive and vestibulo-spinal modulation is nor-mally required for balance corrections. Some muscles, though, such as triceps surae, appeared to receive negli-gible vestibular modulation for the perturbation to stance we applied, whereas other muscles such as quadriceps and tibialis anterior were influenced by vestibular inputs as early as 80 – 120ms, respectively. These changes in the vestibular modulation observed following vestibular loss were not due to different profi-les of perturbation parameters such as ankle and knee flexion, and therefore altered proprioceptive reflex am-plitudes, because we carefully controlled these, as Fig. 1 confirms, to be identical for all subjects. Rather, this reduction of activity during both the stretch reflex in quadriceps and the balance correcting response in tib-ialis anterior appears to be due the failing influence of head vertical accelerations on these responses. This illustrates well the advantages of a confluence of sen-sory inputs on muscle responses, when the head drops due to a sudden flexing of the knees. Both the proprio-ceptive stretch and vestibulo-spinal reflexes act together
to extend the knees and right the body again. In the absence of these mutually reinforcing effects, the trunk motion is more unstable. More specifically, the trunk initially extends backwards less rapidly than normal and for a longer duration in vestibular-loss subjects. Consequently, the unloading response in paraspinal muscles is longer.
Based on our observations questions that can be asked are threefold. Is the proprioceptive unloading responses in paraspinals the primary trigger signal for balance corrections? Is the shift in balance-correcting response latencies in paraspinals with vestibular loss the result of absent vestibular modulation, that is, is the effect of vestibular loss similar to that observed for quadriceps where the amplitude of the response is dependent on the presence of both sensory inputs? Are the changes in early trunk velocities and thus paraspinal balance correcting responses the result of decreased responses in quadriceps muscles with vestibu-lar loss? The essential point raised by these questions is that when two reflexes interact exact information on the interdependency for the interaction can only be ob-tained by holding one input identical for both test populations as was done for the ankle and knee inputs in this report. Furthermore what appears to be proprio-ceptive reflex as in the case of quadriceps may in fact functionally be primarily a vestibulo-spinal reflex. Fol-lowing this line of reasoning, it appears that early changes in paraspinal muscle responses result from a combination of biomechanical effects ensueing when quadriceps and tibialis anterior responses are lacking vestibular modulation, plus an additional direct lack of input from the vestibular system in paraspinal muscles. If balance correcting responses are centrally organised as many authors have assumed [12,42,67,94] then the most parsimonious explanation for the shift in balance-correcting paraspinal responses in vestibular-loss sub-jects is that the shift occurs because of the lack of direct vestibular input to centres generating this triggered response. The changes in paraspinal responses after 200 ms with respect to those of normal subjects are pre-sumably caused by two factors: the delay in onset; and secondly the excessive trunk velocity in vestibular-loss subjects stretching paraspinal muscles more than in normal subjects.
The shift in the latencies of paraspinal responses without a corresponding shift in the latencies of triceps surae muscles is evidence indicating that balance cor-recting responses are not organised as an ascending
synergy triggered and starting at the ankle joint as other authors have maintained [67,70,71,94,96]. If in-deed the ascending synergy hypothesis was valid, then the latencies should not have been altered in paraspinals in vestibular-loss subjects. In fact we have previously demonstrated that ankle inputs are not re-quired to trigger balance corrections [9]. Based on our current and previous results [9] it seems more likely that balance correcting responses are triggered by trunk inputs as Forssberg and Hirschfeld [42] have suggested. Our report adds details to this general hypothesis by suggesting that a confluence of knee, trunk and vestibulo-spinal inputs triggers human balance correc-tions depending on the mode of movement the body is forced into by the perturbation and on the differential weighting of proprioceptive and vestibulo-spinal inputs in the triggered muscle’s balance correcting response. According to this hypothesis when the body is forced into knee flexion then extension, with trunk extension and then flexion commencing shortly after the knee flexion, the centrally triggered response in triceps surae muscles at 100 ms is triggered by trunk motion unload-ing paraspinals and knee flexion stretchunload-ing quadriceps and is therefore not altered by vestibular loss. Paraspinal responses which are delayed to 135 ms to ensure an optimal movement strategy is further delayed by vestibular loss indicating that this sensory input plays a role in the central triggering process for these muscles. The normal confluence of sensory inputs pre-sumably permits the CNS to cope with a variety of situations in which proprioceptive and vestibular inputs can reinforce each others righting effects and thereby prevent a fall. The disadvantage of this arrangement is that vestibular rather than lower-leg proprioceptive loss has major effect on balance instability and some sup-posedly ‘purely’ upper-leg and trunk proprioceptive-based responses are altered more than would be expected from changes in movement profiles. Thereby, the study of proprioceptive contributing inputs to bal-ance corrections has an added level of complexity which needs to be unravelled.
Acknowledgements
Support for this work was provided by the Swiss National Research Foundation grant 32-41957.94 to J.H.J. Allum. Typographical assistance was provided by Ms P. Mu¨ller and E. Clarke.
Effects of proprioceptive loss on postural control
B.R. Bloema, J.H.J. Allumb, M.G. Carpenterc,
F. Honeggerb
aDepartment of Neurology, Leiden Uni6ersity Medical Center, Leiden, The Netherlands
bDepartment of ORL, Uni6ersity Hospital, Basel, Switzerland
cDepartment of Kinesiology, Uni6ersity of Waterloo, Waterloo, Ont., Canada
According to classic neurological teaching, normal functions of the nervous system can be estimated from studies on patients with selective lesions in areas that normally supply a particular function. Following this principle, the contribution of lower leg proprioceptive inputs to normal postural control can be explored in patients with distal peripheral neuropathy. However, this approach is hampered by the fact that patients with polyneuropathy often have a lesion in sensory and motor postural control systems because they typically suffer from loss of both proprioception and muscle strength. Although subjectively sensory complaints usually pre-dominate, almost all patients have both sensory and motor impairment upon clinical or neurophysiological examination [26,30]. To avoid this shortcoming, we have studied postural control in five carefully selected patients with subtle sensory polyneuropathy, but a fully preserved muscle strength.
4. Methods
4.1. Subjects
Five patients with diabetic polyneuropathy and 15 healthy controls participated. All subjects gave informed consent to a protocol approved by the Institutional Review Board of the University Hospitals in Basel and Leiden. All patients had bilaterally absent Achilles ten-don reflexes and weak patella tenten-don reflexes, but a fully preserved muscle strength during careful clinical exami-nation. Extensive neurological investigation, including assessment of vision, vestibular function, lower leg sen-sory function and balance, revealed no other abnormal-ities. Subjects with other neurological disorders (includ-ing muscle atrophy), vestibular disease, orthopaedic pro-blems (including diabetic arthropathy) and diabetic retin-opathy or other visual disturbances were excluded. Pa-tients taking medication that might affect postural con-trol (e.g. psychotropic medication) were also excluded. 4.2. Experimental design
The experimental design has been described in more
detail by Allum et al. [9,12]. Subjects were standing on a movable support-surface. The stimuli were identical to those described in the previous contribution. The first response for each perturbation type was ignored for data analysis purposes. The experiment was performed both with eyes open and closed. In this paper, only the results obtained with pure ‘toe-up’ rotational perturbations (eyes open) will be reported. A full account will appear elsewhere [24].
Subjects received no prior information regarding per-turbation type or magnitude, but were merely instructed to return to the upright as quickly as possible. We recorded surface electromyography (EMG) from leg and trunk muscles as well as several biomechanical variables (see prior contribution).
Data analysis was described in the prior contribution from Allum and Honegger [12].
All responses for an individual subject were combined, then the data were averaged across subjects in the same population. Differences between population means were tested with Student’s t-test, using a cut-off of PB0.05.
5. Results
Averaged postural responses of patients and controls to a dorsi-flexion rotation of the support surface are shown in Fig. 2. The pattern of postural activity in normal subjects consisted of a short latency stretch reflex in soleus at ca. 50 ms, followed by posturally stabilizing responses at 134 ms (S.D. 9 ms) in tibialis anterior and a coactivation of soleus. The coactivation response in soleus has been termed a medium latency stretch reflex by some authors [22,33,113]. The earliest activity in paraspinal muscles occurred around 80 ms, followed by a larger response at 118 ms (S.D. 9 ms) well before onset of muscle activity in the more distal tibialis anterior muscle. Early paraspinal muscle activity conceivably represents a stretch response, triggered by forward flex-ion of the trunk which commenced some 60 ms after the postural perturbation. The ankle torque trace showed first an increase in dorsi-flexion torque on the support surface due to the body inertia and intrinsic ankle stiffness. The ankle torque trace then rapidly moved in a plantar-flexion direction, mainly due to stabilizing muscle activity in tibialis anterior [23,74]. Monosynaptic short latency stretch reflexes in soleus were absent in patients. Instead, a soleus response occurred earlier in patients than in controls. Possibly this response is a medium latency stretch reflex, because it commenced when short latency reflex responses would have termi-nated in controls. Most markedly, the initial part of stabilizing responses in tibialis anterior were delayed and reduced in amplitude in patients. In contrast, the onset latency of paraspinal muscle activity was normal in patients (115 ms, S.D. 5 ms). The amplitude of paraspinal
muscle activity was, however, considerably reduced. These abnormalities of postural responses were reflec-ted by a delayed onset of plantar-flexion (stabiliz-ing) ankle torque, suggesting inadequate stabilizing
muscle activity across several leg muscles. Furthermore, the weak paraspinal activity was parallelled by an increased forward flexion of the trunk compared to controls.
Fig. 2. Postural responses in patients with diabetic sensory polyneuropathy (thick lines) and healthy subjects (thin lines) following sudden toe-up rotational perturbations. All responses represent population averages, calculated from the average responses to nine trials of all subjects. The traces were aligned in time with respect to zero latency. (Reproduced from Ref. [24]).
6. Discussion
The results of this study indicate that a selective reduction of proprioceptive feedback from the lower legs affected some, but not all balance correcting responses evoked by support surface rotations. Specifically, the onset of balance correcting responses in ankle muscles (soleus and tibialis anterior) were altered in patients, while the onset of muscle activity in paraspinal muscles was normal. These results confirm that lower leg propri-ocepsis (conceivably mainly from the ankle joint) is used to trigger and modulate coupled ankle reflexes [2,35,37,40,81,112,114]. On the other hand, it seems that lower leg propriocepsis does not trigger balance correct-ing responses in more proximal body segments, such as the trunk. The onset of paraspinal muscle activity pre-ceded the onset of muscle activity in the more distal tibialis anterior muscle and was not affected by changes in stretch reflexes in soleus muscles, observations which are at odds with the concept of a distally triggered ‘ankle strategy’ [93]. Keshner et al. have also noted contradic-tions to this concept in normal subjects [75]. A more proximally located trigger is apparently used to trigger postural activity in trunk muscles.
The exact nature of this ‘proximal’ trigger cannot be inferred with certainty from our study, but may be located in the trunk which has functionally effective proprioceptive sensation [118]. Forward flexion of the trunk preceded onset of muscle activity in paraspinal muscles, which seems compatible with a role for trunk movement as a trigger for postural activity in paraspinal muscles. Interestingly, trunk flexion was increased in our patients, but paraspinal muscle activity was reduced in amplitude. If we assume that anterior trunk movements induced by tibialis anterior activity add to those caused passively by support surface dorsi-flexion and that this trunk movement is corrected by paraspinal activity, then increased trunk flexion in patients may represent an ‘active’ compensation centred at the hips to counterbal-ance their insufficient ankle responses. This could be achieved by a reduced amount of paraspinal muscle activity, which indeed occurred in our patients.
The concept of a ‘proximal trigger’ for at least some postural responses is supported by several lines of evidence. First, although backward platform translations and toe-up platform rotations cause a similar ankle dorsi-flexion, these two perturbation types evoke differ-ent postural strategies and synergies [6]. Second, recdiffer-ent evidence suggests that, at least for lateral sway during quiet stance, equilibrium is mainly organized around the hip joints and not (as predicted by the ‘inverted pendu-lum’ model of postural control) around the ankle joints [31]. In fact, it appears that quiet stance is more unstable when movement is restricted to the ankles only [41]. Third, clinical experience indicates that postural instabil-ity is less impressive in patients with impaired
proprio-ception in the lower legs due to length-dependent axonal degeneration, as compared to patients with an equally severe proprioceptive loss in more proximal body seg-ments due to length-independent ganglionopathies [30,47]. Fourth, analyses of ontogenic development of postural control suggest that infants initially use a proximal trigger to stabilize their trunk while sitting. It is possible that older children never fully abandon this postural strategy when they learn to walk and balance their trunk in an upright position [42].
Our observations differ from those of Inglis et al. [71] who noted that postural responses in legs and trunk were equally delayed in patients with a peripheral neuropathy. This discrepancy is perhaps related to differences in experimental design because Inglis and colleagues used horizontal platform translations, as opposed to the rotational perturbations in the present study. However, it is more likely that differences in patient selection can explain the contrasting observations. All our patients were carefully screened to exclude muscle weakness, whereas all patients in the study of Inglis et al. [71] had a reduced distal muscle strength that likely affected postural control [123].
As expected, short latency stretch responses in soleus were absent in our patients with bilaterally absent Achilles tendon reflexes. Interestingly, the onset latency of the subsequent medium latency response in soleus was reduced in patients, suggesting that the absence of early stretch reflex activity facilitated the appearance of later occurring stretch responses. The precise neural circuitry underlying this medium latency stretch reflex activity remains unknown [61]. Our observation is perhaps best reconciled with the theory that firing of separate subpop-ulations of motorneurons [14], possibly due to grouped afferent discharges from stretched muscles, gives rise to the segmented pattern of stretch reflex activity [17]. According to this latter theory, the delay of medium latency responses relative to short latency responses is caused by the fact that spinal motorneurons are tempo-rarily refractory to successive Ia afferent bursts. Such mechanisms appear to be operative in human triceps surae [17].
We conclude that intact propriocepsis from the lower legs is not required to trigger all balance corrections. Lower leg propriocepsis serves to trigger postural re-sponses in ankle muscles and helps to shape other responses within a given postural strategy, once as we hypothesize, these have been triggered by trunk movement.
Acknowledgements
G. van de Giessen and M. Cramer are gratefully acknowledged for their assistance. We thank Dr H.P.P.H. Lemkes for the kind referral of his patients. We also thank Prof R.A.C. Roos and Dr J.G. van Dijk for
their critical comments. This study was supported by Swiss National Research Foundation Grant 32-41957.94 to J.H.J. Allum and a grant by Eli Lilly Netherlands to B.R. Bloem.
Contributions of proprioceptive and vestibular inputs to postural control in the roll and pitch planes
M.G. Carpentera, J.H.J. Allumb, F. Honeggerb aDepartment of Kinesiology, Uni6ersity of Waterloo, Waterloo, Ont., Canada
bDepartment of ORL, Uni6ersity Hospital, Basel, Switzerland
Previous investigations of balance responses to rota-tions of the support surface have been limited to pertur-bations directed in the anterior – posterior or pitch plane [7,11,15,77,119]. When human balance is observed out-side of the laboratory, however, perturbations to equi-librium can be experienced from multiple directions within both the pitch and roll planes. It may well be that initial head and limb movements and therefore the resulting muscle synergies and movement strategies in non-pitch planes are very different from these previously observed in pitch planes.
Therefore, the purpose of this investigation was to examine both the muscular and biomechanical responses of normal healthy adults to platform rotations in the pitch and roll planes. From the results of this experiment we hoped to gain insights into two previously unan-swered questions. First, to determine the directional sensitivity of postural muscles involved in balance correc-tions and the consequent biomechanical results of such activity. Second, to discover the relative contributions from the hip, knee, and ankle proprioceptive inputs in triggering automatic postural responses to unexpected perturbations.
In contrast to surface translations, rotational stimuli in the pitch plane elicit stretch reflexes in ankle muscles which increase instability and are antagonistic to those recruited for balance correcting responses [32,93] whereas the opposite situation has been documented for paraspinal muscles [7,9]. Given these contrasting func-tional response characteristics for ankle and hip muscles and the possibility that these differences may vary as roll stimuli are additionally present we considered it neces-sary to examine stretch and balance correcting responses separately as they may be differentially affected by the direction of the perturbation.
7. Methods
Fourteen healthy normal adults volunteered for this
study. Our sample consisted of seven male and seven female participants with ages between 18 and 28 years. Subjects were free from any neurological or orthopedic disorders as verified by self report.
Subjects were required to stand on a movable force platform capable of generating support surface rotations in multiple directions of pitch and roll combinations. Prior to each perturbation, subjects used on-line visual (eyes open condition) and auditory (eyes closed condi-tion) feedback from anterior – posterior ankle torque to maintain a locked knee standing position within91 Nm of their preferred stance position set prior to the exper-iment. Subjects were presented with platform rotations in 16 different directions all separated by 22.5°. Platform rotations had a constant amplitude of 7.5° and rotated at a constant velocity of 50°/s. Platform rotations were presented in two series performed on separate days. One series contained eight directions (45° separation), with each direction randomly presented five or six times for a total of 44 trials per series. The other series contained the remaining eight directions. The order of series presentation was counterbalanced among subjects to minimize any ordering effects. Perturbations were pre-ceded by a random time interval between 5 and 20 s to minimize any effects due to anticipation.
EMG data was collected starting 100 ms prior to the onset of platform rotation, for a duration of 1 s. Surface electrodes were applied bilaterally to tibialis anterior, soleus, vastus lateralis (quadriceps) and paraspinal mus-cles. Raw EMG data was sampled at 1000 Hz, after band-pass filtering, full wave rectification, and smooth-ing. Background activity during the 100 ms prior to perturbation onset was subtracted from each signal and EMG areas were calculated within predetermined time intervals associated with previously identified stretch (40 – 100 ms); (80 – 120 ms); balance correcting (120 – 220 ms); (240 – 340 ms) secondary balance-correcting re-sponses and stabilizing reactions (350 – 700 ms) [7,10]. Angular velocity of the trunk and upper leg were measured using angular velocity transducers oriented in both the pitch and roll planes. Vertical, anterior – poste-rior and lateral linear accelerations were recorded at the level of the temples from accelerometers mounted to a tight fitting headband. EMG and biomechanical signals were averaged with respect to direction within each subject and population averages were compiled from the 14 subjects.
8. Results
As illustrated in Fig. 3, the overall body response to multiple direction rotations of the support surface were stereotypical and dependent upon the direction of the perturbation. As reported by Keshner and Allum [76], response to pure toe-upward tilt of the platform (defined
Fig. 3. Stick figure illustration of stimulus induced link movements in response to different rotational perturbations in the pitch and roll planes. The curved arrows represent angular velocities of the legs and trunk or angular accelerations of the head, and the straight arrows the linear accelerations of the head.
ms at the right and left sides of the head consistent with an early roll angular acceleration at the head. A flexion of the ‘uphill’ knee was observed later (ca. 120 ms) to help maintain equilibrium under the new requirements of the platform position. The magnitude of the biome-chanical responses in both the pitch and roll plane were changed in amplitude with the direction of the pertur-bation.
The area of EMG responses for soleus and paraspinal muscles are illustrated in Fig. 4. Mean areas and positive standard errors of the means have been plotted with respect to the direction of surface rotation for both left and right muscles. Earliest onset of the stretch reflex in soleus muscles was observed at ca. 50 ms. During the period between 40 and 100 ms increased muscle activity occurred in backward directions be-tween 113 and 248° with maximum activity oriented approximately at the 180° (toe up) direction. Balance correcting responses in soleus muscle were sensitive to directions very different from those of its stretch re-sponse. For balance correcting responses (120 – 220 ms) the greatest activity was observed in the forward direc-tions with maximum activity tuned at ca. 30° lateral of 0°. Activity in the backward directions during this time period can be attributed to co-contraction with activity in tibialis anterior [74].
Stretch reflexes in paraspinal muscles can be ob-served as early as 45 ms. As shown in Fig. 4, maximum activity during the period between 40 and 100 ms was directed at ca. 135 and 225° for right and left muscles, respectively. An earlier unloading response (associated with negative EMG area) was observed at ca. 35 ms during rotations of the support surface forward and to one side with maximum unloading responses observed at 315° for the right paraspinal and 45° rotations for the left.
Maximum balance correcting responses were oriented 90° from both the stretch and unloading reflexes. Max-imum activity during this period tuned at ca. 225 and 135 for right and left paraspinals, respectively, with most activity observed in contralateral backward directions.
9. Discussion
Previous studies utilizing multi-directional transla-tions of the support surface have also reported direc-tionally sensitive activity of postural muscles and biomechanical variables in humans [90] and cats [83 – 85,106]. However, these studies did not separate stretch and later balance correcting responses to determine how each were influenced by the direction of the pertur-bation. The results of the present experiment clearly demonstrate that stretch and balance correcting re-sponses are sensitive to the direction of the surface here as the 180° direction) can be described as a
‘stiffen-ing strategy’ characterized by backward rotation of the lower limbs coupled with a forward rotation of the trunk segment with an onset of trunk velocity of ca. 60 ms. Linear accelerations of the head were observed as early as 15 ms with vertically upward directed accelera-tions greater in magnitude than in the horizontal A – P or lateral directions. When the platform was tilted only in the toe-down direction (direction 0°) a multi-segmen-tal response characterized by flexion of the knee joint and backward rotation of the trunk was observed at approximately 60 ms. This is accompanied by a down-ward acceleration of the head at ca. 15 ms onset and again smaller accelerations in the horizontal directions. As a roll component was added to the platform pertur-bation, the trunk consistently rotated in the opposite direction to that of the platform rotation. Trunk roll angular velocity was observed with an onset latency of 20 ms in all perturbations containing a roll component ca. 40 ms prior to any movement of the trunk in the pitch direction. For pure roll directions, vertical linear accelerations of different directions were observed at 50
rotation. In addition, stretch reflexes are sensitive to directions much different than those for balance cor-recting responses within the same muscle. In the soleus muscles, stretch reflexes are tuned along the pitch plane and maximum balance correcting responses are ori-ented ca. 180° to maximum stretch reflex. However, in the paraspinal muscles, the maximum balance correct-ing response is tuned 90° from both the stretch reflex and unloading reflex in the same muscle. These differ-ences in directional sensitivity strongly suggest that a complex, presumably centrally computated pattern un-derlies balance corrections. Further, that the disparity between proprioceptive reflex vectors and those of bal-ance corrections indicates that other inputs such as vestibular inputs must contribute to balance corrections.
Observed stretch reflexes in the paraspinals demon-strated onset latencies equal to those of the soleus muscles. Earlier unloading paraspinal reflexes, charac-terized by a decrease in EMG activity below back-ground levels, were observed as early as 35 ms. Similar unloading reflexes have been reported in response to sudden unloading of the trunk during backward trans-lations [9] and in trunk and neck muscles in subjects abruptly tilted from an upright sitting position [64,73]. Early onset of trunk angular velocity in the roll direc-tion prior to any movement in the pitch plane is evidence that directionally-sensitive trigger information from proprioceptive receptors at the level of the hip may be available to generate balance corrections in addition to less directionally-sensitive somatosensory information received from stretch receptors at the ankle
Fig. 4. Polar plots of mean EMG activity (dark shade) and positive standard error (grey shade) for the soleus and paraspinal muscles of 14 subjects tested under eyes-open conditions. The response amplitudes were measured for rotational perturbations in 16 directions during stretch reflex (40 – 100 ms) and balance correcting (120 – 220 ms) time intervals. The arrows indicate the direction of the calculated maximum activity vector for each muscle and response interval.
joint. Early rotation of the pelvis has also been ob-served in sitting adults and infants during platform rotations with reported latencies of 10 ms [42,64]. These experiments also identified early vertically directed head accelerations with similar onset latencies to those found in this present study. The early (15 ms onset) vertically directed linear accelerations of the head may provide early complementary trigger information in the pitch direction prior to the time pitch information is fully elicited at the trunk. Therefore, it appears that both muscular and earlier biomechanical responses at the trunk and head are more sensitive to the direction of the platform perturbation than those at the legs.
We wish to emphasize that stretch reflex and balance correcting responses are sensitive to very different direc-tions within the same muscle. In the soleus muscle, the balance correcting response is oriented at ca. 180° to the stretch reflex, however in paraspinal muscles, the balance correcting response is tuned 90° to both the stretch and unloading reflexes. The orientation of balance correcting responses in paraspinal muscles may reflect the instability of the trunk in the lateral directions and the differences to reflex responses in soleus muscles may reflect the necessity to coordin-ate leg activity with that of the trunk. The func-tional characteristic of this interlimb coordination is the presence of early roll velocities of the trunk prior to velocities in the pitch direction. Early movement of the trunk in the roll direction, as well as stretch and unloading reflexes in paraspinal muscles prior to so-matosensory information received from lower leg mus-cles, also provides evidence that earlier and directionally sensitive trigger information is available at the level of the hip. Unloading reflexes in paraspinals with earlier onsets than stretch reflexes may suggest that Ib force-related afferent fibres may contribute to triggering balance correcting responses even prior to Ia stretch-related fibres. In addition, early vertically-di-rected linear-acceleration responsive vestibular informa-tion appears to be available prior to angular acceleration information and may be used to vectorially modify postural responses. Thus it is to be expected that vestibular-loss subjects would suffer gross asym-metries in muscle responses when tilted in pitch and roll.
Acknowledgements
Support for this work was provided by the Swiss National Research Foundation grant 32.41957.94 to J.H.J. Allum and a Canadian National Scientific Re-search Council grant to M.G. Carpenter.
Large fibre sensory contributions to the control of locomotion studied by non-invasive deafferentation
M. Hulligera, Greg M. Bishop, M. Djupsjo¨backab,
A. Zanussi
aDepartment of Clinical Neurosciences Uni6ersity of Calgary, Calgery, Alta., Canada
bDepartment of Work Physiology and Technology, National Institute for Working Life, S-907 13
Umea, Sweden
For more than a century experimental deafferenta-tion has been used to study the role of peripheral feedback systems in the control of movement [43 – 45,49,62,80,91,92,120]. The results were controversial, mainly due to methodological limitations [117]. For instance, the classical method of limb deafferentation by dorsal rhizotomy is non-selective, in that it abolishes sensory feedback of all modalities. Further, it is re-stricted to a few spinal segments, and complicated by surgical trauma and extended periods of recovery, al-lowing for motor deficits to be masked by compensa-tion. The bulk of the earlier observations seemed to indicate that peripheral deafferentation often produced remarkably insignificant motor deficits, since basic forms of voluntary movement were either preserved or observed to recover rapidly. However, clinicians contin-ued to be impressed by often striking motor dysfunc-tion in patients with apparently selective large-fibre sensory neuropathies [104,110,111]. The etiology of these neuropathies was mostly unknown, and the asso-ciated motor deficits spanned a wide range. For in-stance, some patients were unable to compensate for unforeseen disturbances, while posture and gait were little affected [104,110]. In one case an even more remarkable inability to learn new motor skills was described by Rothwell et al. [104]. On the other hand, more generalized motor deficits (including general ataxia and abnormal gait) have also been described [46,108,109]. However, the interpretation of these find-ings was again controversial, since clinical observations rarely permit unambiguous identification of neu-ropathies as purely sensory, and since in practice it is often impossible to rule out additional primary deficits in the CNS.
Overdoses of pyridoxine (vitamin B6) cause a large
fibre peripheral sensory neuropathy, which is associated with major motor impairment [45,111,122]. The pyri-doxine neuropathy was first observed in dog [13,65,66], subsequently also studied in rat [78,79,122] and eventu-ally described in man as a megavitamine syndrome [111]. Further, in laboratory animals pyridoxine causes selective degeneration of large sensory fibres, while sparing motor axons [66,122]. Clinically it is associated
with ataxia without loss of strength [79,124]. The selec-tivity of the experimental B6 neuropathy was further confirmed in histopathological studies. In dogs and rats, neurone degeneration in the CNS was shown to be restricted to projection regions of large-diameter first-order afferent fibres (spinal cord and dorsal column nuclei, e.g. [66,79]).
However, the nature and extent of the motor impair-ment induced by such chemical deafferentation have never been examined in any detail. In addition, being non-invasive and rapidly induced, this form of deaf-ferentation avoids the complications from surgical trauma associated with dorsal rhizotomy. Since large sensory fibres appear to be affected preferentially, it seems possible in principle to produce modality-specific deafferentation of the largest sensory afferents, i.e. the group I axons, by controlled and carefully adjusted administration of pyridoxine. Therefore, experimental pyridoxine intoxication might provide a new tool for behaviourial and electrophysiological studies of the function of exclusively proprioceptive group I feedback systems in the control of involuntary and voluntary movement.
We set out to analyze the motor deficits elicited by pyridoxine-mediated deafferentation in the cat, using kinematic recordings during treadmill locomotion, vestibular and proprioceptive reflex testing in chronic animals, as well as single unit and compound potential recordings in terminal acute experiments. This report focuses on preliminary observations on the impairment of walking and of the responses to perturbations during acute pyridoxine neuropathy.
10. Methods
10.1.Animals
Chronic recordings were made in four awake male cats (3.6 – 4.5 kg) studied before and after administra-tion of pyridoxine. The treatment and recording proto-col for the chronic observations did not require surgical interventions. Upon completion of the chronic record-ings, the sensory neuropathy induced by pyridoxine was investigated in a terminal acute electrophysiological experiment. The entire protocol was approved by the University Animal Care committee.
The animals were trained to walk at different speeds on an enclosed treadmill for food reward and verbal encouragement. Training sessions lasted 30 min and took place every day over a period of 2 – 3 weeks. Before recording sessions began, the animals were shaved under short lasting Halothane anaesthesia. The dominant axes of the main joints of the right forelimb and hindlimb were determined by palpating bone mo-tion during imposed movement. Joint axes as well as
prominent bony locations were marked by lightly tat-tooing overlying skin.
10.2. Treatment and animal care
The treatment and recording protocol lasted 2 weeks and encompassed a pre-treatment recording period (days − 3 to 0; control measurements), a treatment period (days 1 to 3 or 4), a post-treatment recovery period (days 4 or 5 to 10 or 11), and a terminal acute experiment (day 11 or 12). The animals received three or four daily intraperitoneal injections of pyridoxine hydrochloride (350 mg/kg, Sigma; dissolved in isotonic saline or Ringer solution) on consecutive days. Within 3 – 4 days of the first injection, the animals developed a motor syndrome with gait impairment and general ataxia (see Results), which was so severe that it com-promised adequate food and liquid intake over a period of about 5 days. During this time, the animals were fed by one of the experimenters. If necessary, they also received injections of Dextrose saline solution (300 ml, i.p.). With this care all animals tolerated the treatment and maintained satisfactory health, although most of them lost some weight (on average about 0.25 kg). On day 11 or 12, typically one week after full manifestation of the clinical syndrome the electrophysiology of the peripheral nerves was investigated in a terminal acute experiment that was conducted under pentobarbi-tone anaesthesia. At the end of this experiment the animals were perfused with formalin for histological examination of tissue samples from the peripheral and central nervous system. The results of the terminal experiments and the histological analysis are reported elsewhere [21].
10.3. Recording
Kinematic recordings were made every day before, during and after treatment. At the height of the motor syndrome, the animals had to be supported to enable them to walk. This was done with a customized trunk harness, which was suspended from the ceiling of the treadmill’s enclosure and adjusted to provide between 25 and 50% body weight support.
At the beginning of each session self-adhesive reflec-tive markers were attached to a thoracic vertebra (Th4), shoulder, elbow and wrist of the right forelimb, and over the 7th lumbar vertebra, hip, knee, ankle and toe of the right hindlimb. The tattooed skin marks ensured reproducible marker topography between sessions. The positions of these markers were digitized in three di-mensions at 200 Hz, using a Kintrac motion analysis system. Recordings were made during walking, both with and without weight support, at three treadmill speeds (0.2, 0.4, 0.8 m/s), with 10 – 15 trials at each speed. Individual trials lasted 2 – 3 s. When locomotor
ability was reduced (i.e. during and after treatment), recordings were typically limited to the lowest speed. However, at a given speed, two sets of recordings, i.e. with and without harness support, were performed, provided the animal’s locomotor capability and/or mo-tivation allowed to do so. In addition, when animals landed from short falls (15 cm), kinematic recordings were taken for a 2 s episode centred around the mo-ment of foot contact. All kinematic recordings were made with the animals situated in the enclosed tread-mill box. Further, the kinematic recordings were always supplemented by video recordings of the entire acquisi-tion session.
Tendon reflexes were elicited by light finger taps on a transistor switch that was placed on the skin over the Achilles tendon. Surface EMG electrodes were attached longitodinally over the lateral gastrocnemius muscle. The EMG signal was suitably amplified (typically with a gain of 3,000) band-pass filtered (10 Hz – 1kHz) and sampled with a digital oscilloscope (80 kHz) during a 50 ms period after the trigger signal that was generated by the mechanical switch. Tendon tap responses were recorded first when the animals were maximally re-laxed, resting on the lap of one of the experimenters. The recordings were repeated during standing, when the animals supported most of their body weight by their hindlimbs, while being restrained by an experi-menter holding their trunk.
Vestibular function was evaluated by clinical assess-ment of classical vestibular reflex responses: righting responses were elicited by slowly tilting the animal (head up, head down, to the left or to the right) and by observing compensatory/corrective trunk and limb movements; rapid roll responses were elicited by rapidly rotating the animal’s trunk around the longitudinal trunk axis and observing compensatory head and eye movements.
10.4.Analysis
Kinematic data were pre-processed with commercial software (Kintrac) to generate files of 3-dimensional marker coordinates and to calculate basic parameters of marker position records (maxima, minima, excursion etc.). In particular, successive minima of the vertical coordinate of the wrist and hind paw toe marker were used to demarcate individual step cycles and measure cycle duration [20]. Further processing, to calculate joint angle profiles and derived variables and to gener-ate angle – angle phase plots, was carried out with cus-tomized software [20].
Digitized EMG records of responses to tendon taps were transferred to the laboratory computer network using a serial interface. The EMG data was further analyzed with customized software, to generate displays of superimposed raw EMG profiles and to measure
response magnitude. This was calculated as the average of rectified smoothed EMG envelopes during time win-dows that were defined interactively by the user. All graphic processing and statistical analysis was carried out with public domain software (ACE/gr).
11. Results
11.1. General features of the pyridoxine syndrome
Following pyridoxine treatment on 3 – 4 consecutive days, all animals developed an incapacitating motor syndrome, which precluded spontaneous walking and treadmill locomotion without weight support (see Methods). Typically, the syndrome developed rapidly, on day 3 or 4 of the treatment period, and it persisted with only minimal signs of recovery during the subse-quent 7 days.
Qualitative observations from withdrawal responses and measurements of tetanic force in terminal acute experiments demonstrated that the motor deficit was not due to loss of strength arising from muscle or motoneurone dysfunction. Qualitatively, withdrawal re-sponses elicited by moderate pinching of the animals’ paws were as vigorous as before treatment; EMG activ-ity during assisted standing appeared to be normal and tetanic force in hindlimb muscles as well as antidromic compound potentials recorded in the ventral roots were within the range of normal values.
Clinically, vestibular righting and roll responses were not affected. At the height of the motor syndrome, when the animals were tilted slowly in the sagittal plane, they responded appropriately with extension of the forelimbs, and when tilted in the frontal plane, with extension of the ipsilateral limbs, in either case so as to counteract the tilting motion. During rapid roll rota-tions along the longitudinal body axis, the head was maintained in horizontal position in the frontal plane even for the largest trunk rotations (about 60°). 11.2. Kinematic analysis of treadmill locomotion
Motor deficits spanned a wide range. The syndrome encompassed severe ataxia, unsteady gait, limited pos-tural stability, abnormal limb excursions and erratic interlimb coordination during locomotion, and a strik-ing inability to compensate for external disturbances, e.g. when landing from short falls.
Preliminary multivariate kinematic analysis indicated that during walking the basic cyclic pattern of individ-ual limb movements was little affected ([20] and unpub-lished), while cadence was reduced and stride length and angular excursions were often increased. Before treatment all animals comfortably walked on the tread-mill at 0.8 m/s. At the height of the syndrome, during
Fig. 5. Interlimb coordination during harness supported slow treadmill walking. Row 1, abnormal locomotion and disruption of interlimb coordination 10 days after first pyridoxine injection. Row 2, control recordings from the same animal, before pyridoxine treatment. A – C, illustration of three separate trials for each condition. B.2, illustration of the measurement of the duration of a step cycle (horizontal bar), defined as the interval between successive instances of foot contact (arrows) at the beginning of the stance phase. B.1, C.1, illustration of the variability of step cycle duration (horizontal bars) and of the inconsistent relation of forelimb cycle demarcation (arrows) relative to the hindlimb cycle (bars) during walking without sensory feedback. See also text.
the first week after the manifestation of the motor impairment, none of the animals were capable of walking without weight support, not even for brief episodes. Three of the four animals were able to walk with harness support towards the end of the post-treatment period (on days 6 – 10). However, the maximum speed they attained was no higher than 0.2 m/s.
While the cyclic motion patterns of individual limbs appeared little affected, direct observation of the animals during walking trials indicated that interlimb coordina-tion was severely compromised. This was confirmed by careful inspection of video recordings and by the analysis of the temporal profiles of the sagittal motion of forelimb and hindlimb markers.
Fig. 5 shows three episodes of walking at low speed (0.2 m/s) that were recorded from the same animal during the control period, before commencement of pyridoxine treatment (Fig. 5, bottom row), and during full manifestation of the ataxic syndrome, 10 days after the first injection (Fig. 5, upper row). In either case, the animal walked on the treadmill with partial weight support provided by a harness (see Methods). Movement is illustrated by sagittal marker motion. Forelimb wrist motion is shown by the thick lines, hindlimb toe motion by the hatched lines. Treadmill motion was backward and is shown as downward displacement. During stance,
the paws rested on the treadmill and travelled backwards (downward sloping segments of the two limb traces), while during swing they are brought forward, ahead of treadmill motion (upward sloping segments). In these displays touchdown occurs at the maxima, takeoff at the minima, of the marker motion profiles (Fig. 5A.2).
During normal locomotion (Fig. 5), fore and hindlimb step cycles showed a consistent phase relationship, in that forelimb touchdown occurred consistently at about 30% of the hindlimb step cycle (horizontal bar and arrows in Fig. 5B.2). In contrast, during ataxic walking, forefoot contact (arrows in Fig. 5B.1, C.1) occurred at practically any phase of the hindlimb step cycle. In addition, occasionally the forelimb performed two com-plete step cycles during a single hindlimb cycle or vice versa (asterisks in [5]B.1, C.1). Moreover, compared with the normal gait pattern (Fig. 5.2), both forelimb and hindlimb step cycle durations revealed appreciable vari-ability (horizontal bars in Fig. 5B.1, C.1).
In Fig. 5, weight-supported, rather than unsupported walking is illustrated, simply because during the full manifestation of the motor syndrome the animals were not able to walk without harness assistance. The obser-vations of Fig. 5 were confirmed for two other animals. For the fourth animal, the interlimb coordination during pyridoxine impairment could not be assessed, since it was
not able to walk, even with partial weight support. The phase relationship between forelimb and hind-limb step cycles was analyzed by systematic measure-ments of the forelimb cycle demarcation relative to the concomitant hindlimb cycle and by expressing it as a percentage of the normalized hindlimb step cycle. Fig. 6 shows representative cross cycle-phase histograms for all recordings from days 0 to 10 of the same animal as in Fig. 5. Fig. 6 confirms that, in normal animals, walking with partial weight support was associated with a relatively tight coupling of forelimb and hindlimb step cycles, with the forelimb lagging behind the hindlimb by 20 – 50%. This coupling is reflected by the relatively narrow peak in the cross-cycle phase histogram. In the absence of weight support, cycle phase coupling was somewhat tighter (not illustrated). In contrast, in the ataxic condition after pyridoxine treatment, the cou-pling between forelimb and hindlimb step cycles was appreciably reduced, as is evident from the much wider profile of the cross-cycle phase histogram in Fig. 6. Analysis of the video recordings indicated that the decoupling was not a constant feature. Instead, episodes of several cycles with consistent phase locking between forelimb and hindlimb (Fig. 5B.1.) tended to be followed by episodes of several successive cycles with erratic phase coordination (Fig. 5C.1.).
11.3. Landing from falls
During short falls, animals typically prepare for land-ing by fully extendland-ing all four limbs. Upon foot con-tact, they yield and absorb the inertial impact by partial flexion in the principal joints of each limb. This yield phase lasts 100 – 150 ms and is sufficiently short to enable the animals to land and stand in a slightly crouched position. With the kinematic recording ar-rangements of this study, this sequence of events was conveniently monitored by displaying the temporal profiles of the vertical displacement of the L7 vertebral marker. In Fig. 7, the upper family of traces (thin lines) illustrates the yield upon landing and subsequent stabi-lization at an intermediate height in control recordings immediately before pyridoxine treatment was initiated. The displacement of around 25 cm that was maintained after landing corresponded to the normal standing height of this animal. The lower family of traces (thick lines) illustrates landing sequences at the height of the motor syndrome. It can be seen that both the duration and amplitude of the yield phase were significantly increased. In fact, the animal was unable to absorb the inertial impact and landed on its trunk, unable to stand. The residual displacement of 10 – 15 cm after the falls reflects the height of the animals trunk, with the variation between trials being attributable to different degrees of body rotation while lying. This observation was confirmed in all four animals of this report. 11.4. Response to tendon tap
The kinematic recordings of landing sequences in Fig. 7 provide no clues on the nature of the neural mechanisms underlying the failure to absorb the inertial impact during landing. On kinematic evidence alone, failure of anticipatory reactions cannot be ruled out, nor can failure of short-latency segmental reflex re-sponses be automatically inferred. However, the latter was to be expected, if pyridoxine intoxication indeed caused a selective large fibre sensory neuropathy that would compromise feedback from muscle spindle affer-ents and hence the segmental stretch reflex.
During control recordings, crisp tendon tap re-sponses were readily elicited, even in relaxed muscles. Minimum latency was 8 ms, strongly suggesting that the early components were mediated by Ia afferents and the monosynaptic pathway. During the early stages of the pyridoxine syndrome (around day 3), reflex re-sponses in relaxed muscles were diminished, occurring inconsistently. However, once the motor syndrome was fully manifest, tendon tap responses could no longer be elicited.
Abolition of reflex responses might simply reflect changes in central, especially motoneurone, excitability. This possibility was evaluated in reflex recordings
dur-Fig. 6. Cross cycle-phase histograms to illustrate the absence of tight cycle coordination between forelimb and hindlimb steps during har-ness supported slow treadmill locomotion at 0.2 m/s. 1, pyridoxine syndrome, 10 days after commencement of treatment. 2, control data from the same animal before treatment. See also text.
