It is well established that the spinal cord contains intricate computing units capable of performing rapid ongoing motor processing of complex proprioceptive and cutaneous input during coordinated motor behaviors such as standing and stepping. Neural networks in the lumbosacral spinal cord (i.e., central pattern generators (CPG)) can function autonomously (without any brain control) to produce the characteristic alternating motor patterns of gait and to compensate for errors and obstacles[2, 3] using only sensory information from the limbs[4–7]. More recently it has become recognized that these networks have the ability to process complex sensory ensembles that can serve as the controller of posture and locomotion[6, 8, 9].
The rat or cat spinal cord isolated from supraspinal control via a complete low- to mid-thoracic spinal cord transection produces locomotor-like patterns in the hindlimbs when facilitated pharmacologically and/or by epidural spinal cord stimulation[10, 11]. Thus, locomotor-like patterns can be modulated by stimulation of the networks intrinsic to the spinal cord without the contribution of descending signals. To take advantage of these properties, a more thorough knowledge of the mechanisms of spinal cord stimulation, along with a more detailed understanding about specific sites and parameters of stimulation and their corresponding motor output is needed.
Ichiyama et al. reported that epidural electrical stimulation of the spinal cord can induce rhythmic, alternating hindlimb locomotor activity in chronic spinal rats. Stimulation at the L2 spinal segment at frequencies between 30 and 50 Hz consistently produced successful bilateral stepping. Similar epidural stimulation at other spinal segments were less effective, e.g., epidural stimulation at the T13 or L1 evoked rhythmic activity in only one leg and stimulation at the L3, L4, or L5 produced mainly flexion movements.
More recently, completely paralyzed (motor complete, sensory incomplete) human subjects were implanted with a commercially available spinal cord electrode array and stimulation package originally designed for pain suppression, unpublished observations. Stimulation of specific spinal segments (caudal electrodes, ~ S1 spinal level) in combination with the sensory information from the lower limbs and weeks of stand training was sufficient to generate full weight-bearing standing. These subjects also recovered some voluntary control of movements of the toe, ankle, and the entire lower limb, but only when epidural stimulation was present. Thus it appears that the epidural stimulation provided excitation of lumbosacral interneurons and motoneurons that, when combined with the weak excitatory activity of descending axons that were not otherwise detectable, achieved a level of excitation that was sufficient to activate the spinal motor circuits. These results demonstrate that some patients clinically diagnosed as having complete paralysis can use proprioceptive input combined with some synaptic input from descending motor signals, perhaps residual but functionally silent without epidural stimulation to the spinal motor circuits, to generate and control a range of motor functions during epidural stimulation.
These studies suggest that the intrinsic circuits of the spinal cord, if intact, are desirable targets for stimulus-based therapies and strategies. Secondly, the specific stimulation parameters are highly critical to the pattern and quality of functional motor output. The technological hurdles to reach these targets are non-trivial. We have designed an electrode array capable of selectively stimulating specific segments of the rat spinal cord to generate discrete motor responses using a high-density grid of epidural electrodes embedded within a thin-film flexible substrate[13, 14]. Although stimulation occurs at the surface level, miniaturization of the electrode contacts and the use of materials specific to our design restrict the effective field of stimulation to a smaller area as compared with conventional wire surface electrodes.
The specificity and high-density features of the electrode array enable us to capitalize on two key features of the spinal cord circuitries that are believed to be essential for rehabilitating posture and locomotion after spinal cord injury (SCI). Firstly, the spinal circuitry can be neuromodulated and the stimulation can be carefully delimited to affect only relevant areas of the spinal cord, thus optimizing the motor outcome. Secondly, as locomotor circuitries are highly plastic and adapt when provided with sensory cues during motor training, the density and versatility of the multi-electrode array allows for rapid adjustments of stimulation protocols and adaptations to physiological changes that may occur in the spinal cord over time after injury.
Several design features were taken into account including the flexibility of the array, biocompatibility of the base, and stability of the electrodes for a chronic implant. Parylene C has emerged as an ideal electrode array substrate due to its biocompatibility, insulative properties, flexibility, and tear resistance. The tear resistance of parylene C is large, making the arrays robust to surgical manipulation, as well as to stresses produced in a moving animal. The techniques needed to manufacture these multi-electrode devices are not unprecedented. This is the first time, however, that this technology has been adapted for the express purpose of controlling stimulation at specific sites of the spinal cord in a chronic preparation. Given these basic principles and the results observed in the animal models with conventional wire electrodes[10, 17] and from the human subjects with commercially available electrode arrays, it seems likely that use of a high-density electrode array could greatly improve the quality of standing and stepping after paralysis.
Rather than attempting to impose exogenous motor commands, this strategy will capitalize on the intrinsic neural control mechanisms of the spinal cord that remain functional post-SCI, enabling the spinal circuits to process sensory input and to serve as the primary source of control. Using this technology, we can selectively and differentially activate distinct neuronal groups distributed throughout the spinal cord, allowing stimulation of specific electrodes on the array to modulate the physiological state of the spinal circuitry so that sensory input can control various hindlimb motor outputs. To examine the potential capabilities of this stimulation system, we used this novel, flexible, high-density stimulating electrode array during the recovery of standing and stepping in adult rats after a complete mid-thoracic spinal cord transection.