Development of a universal measure of quadrupedal forelimb-hindlimb coordination using digital motion capture and computerised analysis

During the past 10–15 years, neuroscientists have provided statistical evidence for efficacy of many novel interventions in ameliorating the loss of function that accompanies spinal cord injury (SCI) in proof-of-concept experiments in rodent models (see reviews [1, 2]). These data prompt asking whether any of these interventions can be successfully translated into therapeutically valuable treatments for clinical SCI patients [3, 4]. One response has been to initiate immediate investigations in human patients and several groups throughout the world have already commenced clinical trials on various cell transplant strategies [5–7]. Indeed, progress towards application in human patients has been sufficiently rapid that a series of guidance articles has recently been published to ensure results of clinical trials are readily interpretable [8–10].

However, there remains a recognised need for further investigations in non-human species, especially with regard to cell transplantation repair therapies [11], such as bone marrow stromal cells [12–14], Schwann cells [15, 16] and olfactory ensheathing cells [17, 18]. Such extended testing might also be of benefit for screening many potential pharmacological interventions such as chondroitinase ABC [19], or combinations of therapies [20, 21] prior to application in humans. These animal studies can address the potential obstacles in translating success in treatment of small, uniform lesions in experimental rodents into similar results in large, heterogenous lesions in human clinical patients. The inevitable heterogeneity of clinical cases provides the most difficult potential obstacle to detection of clinical efficacy, and thus implementation in humans patients, yet constitutes the most probable reason for failure of therapies to translate from lab to clinic [22].

Spinal cord injury (SCI) is common in domestic dogs and provides a useful large cohort of clinical patients, with a comparable degree of lesion heterogeneity to their human counterparts, in which therapies can be screened prior to introduction into human patients. Preliminary studies have confirmed the usefulness of this approach and the safety of olfactory ensheathing cell (OEC) transplantation in the dog [23]; robust, quantitative methods of functional outcome analysis for clinical SCI are now required for progression to Phase II trials in this species.

There are many methods for analysing the functional outcome after SCI in experimental rodents, including both categorical and numerical scores based on various types of observation (reviewed by [24]), most notably the widely-used BBB score [25]. Of specific relevance are reported numerical methods by which coordination between fore and hind limbs can be accurately quantified and correlated with the severity of experimental rodent SCI [26, 27]. Whilst potentially applicable to dogs, these methods are time-consuming and potentially liable to failure when applied to analysis of locomotion of dogs of variable size and conformation.

The widespread availability of highly sophisticated motion capture and analysis equipment, such as high speed video images acquired from infrared cameras, has revolutionised the ability to very precisely analyse kinetic and kinematic disturbances in the dog (reviewed by [28]). Moreover, domestic dogs are excellent candidates for motion analysis since they are highly amenable to training and are relatively large, which facilitates limb labelling. However, a drawback is their great variability in size and conformation meaning that gait patterns (e.g. between the walk, trot, pace etc) may vary between different uninjured patients, thus hindering easy comparison of outcome between groups of heterogenous individuals. Additionally, domestic dogs suffer a wide range of orthopaedic conditions, any of which may impact on the kinematics of their gait; and, since the presence or extent of such conditions are highly unlikely to be known at the time of injury, may subsequently confound attribution of gait abnormalities to SCI.

Our aim in this study was to use digital video motion capture equipment and modern computerised analysis to devise a simple numerical parameter to describe forelimb-hindlimb coordination that would be independent of size, conformation, speed of movement or pattern of gait and therefore could be applied to all quadrupeds.

Devising a method to quantify the effect of spinal cord injury on coordination between fore and hind limbs in dogs is a challenging task, since there are a great many variables that must be taken into account. Several previous studies investigating the effects of SCI in cats and rodents have concentrated on measures ranging from the recovered ability to take weight-bearing steps [29], to parameters relating to an individual limb, namely joint angles and electromyograph recordings [30, 31], whilst others have focussed on forelimb-hindlimb coordination [27, 32, 33]. We show here that the mean diagonal coupling interval is unaffected by speed of walking, gait pattern, size of animal or orthopaedic disease. Therefore, it represents an extremely robust, yet simple, measure of coordination between fore and hind limbs. This will be of great value for future analysis of outcome of SCI treatments in dogs and other quadrupeds since we can now define in precise numerical terms the gait disturbance associated with thoracolumbar SCI thereby providing a replacement for methods of visual clinical scoring [25].

Previous studies that aimed to define fore-hind limb co-ordination in normal quadrupeds have also examined the correlation between forelimb and hindlimb step cycles [26, 32–34]. However, their parameters were often derived by complicated and time consuming methods, and, as we show here, simple relationships between timing of fore and hind limb paw placement at each step are not sufficiently robust to summarise locomotion as 'normal' within a varied population of normal individual dogs. In our current study, by use of a treadmill to maintain a specified speed of locomotion and through computer-aided analysis, we have extended previous investigations to define mean diagonal pairing interval as a highly robust parameter with applicability beyond one species to include those within and outside a laboratory setting. Mean diagonal pairing interval also has the advantage of being a summary parameter, being the end result of analysis of a large number of step cycles and critically dependent on the rigidly time-locked sequence of forelimb and hindlimb step cycles. It would be expected that this parameter will be unaffected by alterations in gait during locomotion, although this remains to be tested.

Any method of gait analysis that relies on an animal walking on a treadmill could be criticised because of the possibility that apparent coordination is caused by the common substrate along which the fore and hind limbs are walking, causing the limbs of the two girdles to fall into an apparently entrained pattern simply due to the physical linkage. Similarly, the physical linkage between the fore and hind limb girdles or the spine could be a cause for apparent, but not 'real', coordination between fore and hind limbs. However, both these possible caveats are not compatible with the loss of coordination between fore and hind limbs that we have defined and quantified here in SCI dogs, and further supported by a previous report showing that mice with complete SCI can coordinate between the hindlimbs while walking on a treadmill, but not between the forelimbs and hindlimbs [27]. In addition, a previous study comparing overground with treadmill kinematics in the intact rat found only small differences in step cycle timing and joint angles, although coordination was not specifically examined, suggesting that treadmill walking is representative of normal locomotion [35]. Therefore we believe that coordinated walking on a treadmill in normal dogs represents a true observation and is not an epiphenomenon.

Provision of abdominal support is essential for evaluating the gait of animals with severe thoracolumbar SCI, since many are unable to stand or walk unaided, despite the ability to generate weak hindlimb movements. Therefore the question arises as to whether this weight support might confound interpretation of our data. Previous work in neonatal rats [36] and human patients [37, 38] suggests that increasing the weight load on the pelvic limbs can increase their step frequency. However, these studies detected such effects when testing a range of weight support varying from zero to 75% of body weight, which is quite different from our study. Also, this previous work examined pelvic limb movement only, rather than its coordination with the thoracic limbs, for which there is no published information on the effects of partial weight support; there is the additional caveat that animals injured as neonates exhibit different recovery from those injured as adults [39]. In dogs of variable conformation of leg and spinal dimensions, such as the population we examine here, there is good reason to suppose that there will be some variability in the proportion of body weight 'normally' carried by the hindlimbs. In addition, the proportion of body weight carried by the hindlimbs will vary depending on the precise position on the treadmill and instant to instant acceleration or deceleration. Therefore we consider that our use of a support band to maintain a visually normal posture with adjustments to optimise stepping was the most reasonable solution, and, indeed, was shown not to affect forelimb-hindlimb coordination (measured by the mean diagonal coupling interval) in normal dogs.

In SCI dogs the mean diagonal coupling interval became a negative figure, implying that the hindlimbs were walking at a slower step cycle rate than the forelimbs. Such independent movement of the hindlimbs caudal to a thoracolumbar lesion is likely to be due to inherent rhythmogenicity of the lumbar spinal cord pattern generators, or to subsequent plasticity during recovery [40–44]. Our data would further suggest that although the hindlimbs are able to generate a stepping pattern independently of the forelimbs, this inherent pace is slower than that of the forelimbs. A similar result has been found in extensively lesioned cats [29, 31, 32]. The implication therefore is that, in normal animals, a more cranial instructive signal is required to accelerate the hindlimb movements, in contrast to the suggestions made from in vitro studies on neonatal rodent spinal cord that the hindlimbs are capable of regulating speed of activity in the forelimbs [45, 46].

The neural substrate of coordination between fore and hindlimbs has been investigated in many previous studies and, theoretically at least, could be mediated by either propriospinal or supraspinal pathways. In rats and cats, it has been shown that destruction of the ventral and ventrolateral regions of the thoracolumbar spinal cord is sufficient to interrupt normal coordination between fore and hind limbs [32, 47], correlating with in vitro data from neonatal rats that shows normal rhythmic activity in the isolated ventral third of the spinal cord [38]. Evidence also exists to suggest that the dorsolateral funiculus may be involved in fore-hindlimb coordination in the cat [31], and the corticospinal tract in monkeys [48]. In the SCI dogs examined here, the majority of lesions were ventral because of the nature of the inciting cause, as is also the case in most human SCI. Interestingly, it appears that long propriospinal axons may be less susceptible to contusive injury than descending tracts such as the corticospinal tract [49], thus providing a mechanism by which forelimb and hindlimb coordination could be preserved following even severe SCI in dogs.

Although forelimb-hindlimb coordination represents an important aspect of recovery after SCI in quadrupeds it is incomplete as a description of functional recovery, especially with regard to implications for therapy in humans since it is not a measure of goal-directed (implying supraspinally-driven) behaviour. For that reason, it may appear that measures of coordination between fore and hind limbs have limited impact as a means to define the efficacy of an intervention in human SCI patients. However, there is considerable evidence that the human nervous system retains much of the coordination between 'fore' and 'hind' limb coordination observed in quadrupeds [50, 51], suggesting the possibility that physical tasks in which human subjects were required to coordinate effort between their unaffected arms and their paretic legs might be devised. Nevertheless, the major importance of devising a measure of fore-hind coordination is to provide a robustly quantifiable parameter for determining the effect of an intervention in pre-human trials. Therefore, the ability to analyse the gait of dogs in great detail using this technique will have important implications for determining the rate of progress towards human clinical trials.