Brain-controlled modulation of spinal circuits improves recovery from spinal cord injury - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Aug 1;9(1):3015.
doi: 10.1038/s41467-018-05282-6.

Brain-controlled modulation of spinal circuits improves recovery from spinal cord injury

Marco Bonizzato  1 , Galyna Pidpruzhnykova  2 , Jack DiGiovanna  1 , Polina Shkorbatova  2   3 , Natalia Pavlova  2   3 , Silvestro Micera  1   4 , Grégoire Courtine  5   6
Affiliations

Brain-controlled modulation of spinal circuits improves recovery from spinal cord injury

Marco Bonizzato et al. Nat Commun. .

Abstract

The delivery of brain-controlled neuromodulation therapies during motor rehabilitation may augment recovery from neurological disorders. To test this hypothesis, we conceived a brain-controlled neuromodulation therapy that combines the technical and practical features necessary to be deployed daily during gait rehabilitation. Rats received a severe spinal cord contusion that led to leg paralysis. We engineered a proportional brain-spine interface whereby cortical ensemble activity constantly determines the amplitude of spinal cord stimulation protocols promoting leg flexion during swing. After minimal calibration time and without prior training, this neural bypass enables paralyzed rats to walk overground and adjust foot clearance in order to climb a staircase. Compared to continuous spinal cord stimulation, brain-controlled stimulation accelerates and enhances the long-term recovery of locomotion. These results demonstrate the relevance of brain-controlled neuromodulation therapies to augment recovery from motor disorders, establishing important proofs-of-concept that warrant clinical studies.

PubMed Disclaimer

Conflict of interest statement

S.M. and G.C. are founders and shareholders of GTX Medical, a company developing therapies in partial relationships with the topic of the submitted manuscript. M.B., S.M., and G.C. hold a patent on spatiotemporal neuromodulation algorithms (WO2015/063127). The remaining authors declare no competing interests.

Figures

Fig. 1
Fig. 1
Conceptual and technological design of the BSI. The rats were implanted with a microwire array (32 wires) into the leg area of the right motor cortex. The raster plot shows neural recordings over three successive gait cycles. Each line represents spiking events identified from one electrode, while the horizontal axis indicates time. Stance and swing are colored in black and blue, respectively. Two types of BSI were tested. First, a decoder anticipated the onset of the swing phase, which triggered the delivery of stimulation protocols applied to the lumbar spinal cord, wherein motoneurons innervating flexor muscles reside. Second, the cumulative firing calculated from multiunit activity was directly linked to the intensity of stimulation protocols delivered to the same location. Shaded region (bottom left): electromyographic activity of a flexor muscle (tibialis anterior) together with a stick diagram decomposition of leg movements during the stance (dark gray) and swing (blue and light gray) phases of gait. The occurrence of the stimulation is highlighted in blue. During testing, the rats walked in a gravity-assist that personalized the amount of upward force for each rat. Copyright Jemère Ruby (2017)
Fig. 2
Fig. 2
Development and validation of a binary BSI after contusion. a Recording performed on a treadmill 10 days after the severe contusion. From left to right: sub-threshold stimulation of S1 and L2 segments, stimulation of S1 and L2 segments, stimulation of S1 segment plus brain-controlled of L2 segment (flexion stimulation). From top to bottom: color-coded leg kinematics, neuronal signal from a representative channel, multiunit activity, normalized cumulative firing, electromyographic activity of the tibialis anterior muscle, and vertical displacement of the foot. The gait phases are color coded. The blue dots indicate foot-off events decoded from cortical population ensemble. The region colored in blue highlight the occurrence of stimulation over L2. b Confusion matrix of Foot-off decoding calculated across the five rats. c Bar plots reporting mean values and individual mean values of parameters modulated during continuous stimulation versus brain-controlled stimulation. The relative activation of the tibialis anterior was calculated as a percent of the maximum activity recorded during locomotion. *, P < 0.05
Fig. 3
Fig. 3
Design of the BSI decoder. a Successive steps involved in the elaboration of the decoders. Step 1: Neural signals were synchronized with kinematic and muscle activity recordings during locomotion. Each of the 32 channels from the microwire array implanted into the leg area of the motor cortex was filtered, and then transformed in spiking events. The spiking events were calculated from multiunit activity, when passing a threshold that was set manually for each channel. Step 2: The six channels that displayed the largest correlation with the muscle activity measured from the tibialis anterior were isolated. Step 3: A linear decoder linking multiunit activity from the six isolated channels with the control variable (step height) was calibrated for each rat. The linear decoder was then implemented in the online processing pipeline. Step 4: The neural recordings were processed online to obtain spike-rate estimates before passing the resulting cumulative firing through the decoder that tracked neural correlates of foot-off events. When the cumulative firing crossed a threshold corresponding to 100 ms before the predicted occurrence of foot-off events, the pulse generator delivered a 200 ms burst of stimulation over the L2 segment. b Online instantaneous firing-rate, averaged across gait cycles, as applied on Step 4 to control the brain–spine interface (n = 5 rats on treadmill). c Receiver operating characteristic (ROC) illustrating the accuracy of foot-off event detections, which lied well above chance level for all the rats (n = 5 rats). Bar plot reporting the variability in the timing of actual and decoded foot-off events across a period of 2 min of continuous locomotion. The intrinsic variability of foot-off events was larger than the average error in foot-off event detections
Fig. 4
Fig. 4
Rehabilitation leads to encoding of leg flexion in the motor cortex. a Bipedal locomotion recorded on treadmill during continuous stimulation after 5 weeks of gravity-assisted gait rehabilitation. A reward was presented in front of the rat to encourage the variations of foot trajectories. Conventions are the same as in Fig. 2. b Correlation between cumulative firing at foot-off and the subsequent step height obtained at 10 days post-injury and after gait rehabilitation. Data are from the rat shown in a
Fig. 5
Fig. 5
Modulation of reflex responses with increasing stimulation amplitude. a EMG responses recorded from the tibialis anterior muscles following single pulses of stimulation delivered at L2 with increasing amplitudes of stimulation. The value 0% correspond to the smaller amplitude that was functional to facilitate locomotion. Each response is an average of 10 repetitions. The shaded areas distinguish direct responses (direct stimulation of the motor nerve) from post-synaptic responses (reflex responses), which are elicited from the recruitment of proprioceptive feedback circuits. These temporal windows are defined from the expected latencies and durations of these responses. The blue region highlights the range of amplitudes over which the reflex responses remained functional, i.e. the stimulation evoked a functional increase in leg flexion components during locomotion without causing co-contraction with other muscles. b Bar plot reporting the mean amplitude of reflex responses over the entire range of tested amplitudes (mean ± SEM, n = 5 rats). The amplitude of these responses was calculated as the integral of the averaged and rectified signals over the temporal window compatible with trans-synaptic responses (n = 10 repetitions per amplitude). *, P < 0.05
Fig. 6
Fig. 6
Validation of the proportional BSI. a Design of the BSI decoder, following the conventions of Fig. 3. Steps 1–3 remain unchanged. Step 4B: The neural recordings were processed online to obtain spike–rate estimates before passing the resulting cumulative firing through the proportional decoder. Every pulse of stimulation (40 Hz) delivered over the L2 segment was instantly shaped to hold an amplitude linearly proportional to the current cumulative firing, as displayed in the inset. b Online instantaneous firing-rate, averaged across gait cycles, as applied in Step 4, to control the brain–spine interface (n = 7 rats overground). c Bipedal locomotion recorded on a treadmill during proportional stimulation 10 days after injury. A reward was presented in front of the rat to encourage the variations of foot trajectories. Conventions are the same as in Fig. 2. First inset: close up of a single gait cycle, illustrating co-variation between ankle flexor EMG magnitude and cumulative firing. Second inset: further zoom-in, showing that the cumulative firing rate (top) set the stimulation amplitude (middle), which induces motor responses in the ankle flexor that is proportional to this amplitude (bottom). d–e Peak ankle flexor activity during swing co-variates with cumulative firing at foot-off when the proportional brain–spine interface is active (n = 5). f Actual step heights (black) and predicted step heights (blue) during a continuous sequence of steps with continuous stimulation and the proportional BSI. The same data are shown in the correlation plots. g Bar plots reporting the percent of explained variance in step height from the cumulative firing of cortical ensemble population. The same rats (n = 5) were tested with the three conditions of stimulation at 10 days after injury. The same analysis was performed in a group of eight rats that underwent gait rehabilitation with robotic assistance and continuous stimulation for 5 weeks. *, P < 0.05
Fig. 7
Fig. 7
The proportional BSI improves overground walking. a Stick diagram decomposition of leg movements during a trial with continuous stimulation, binary stimulation and proportional stimulation. Data recorded four weeks after injury. b Bar plot reporting the mean values of distances between intact rats (n = 5 rats) and the different groups of injured rats in the PC space. This value decreases with improved locomotor performance. c Bar plots reporting the mean duration of paw dragging (d) and distance covered per unit of time during locomotion. e Bar plots reporting the mean values of intra-limb coordination for each joint of the leg, calculated from the cross-correlation values between the oscillations of adjacent segments with respect to the direction of gravity. The abrupt burst of stimulation delivered with the binary control partly disrupted intra-limb coordination. *, P < 0.05. **, P < 0.01. ***, P < 0.001
Fig. 8
Fig. 8
The proportional BSI improves stair climbing. a Scheme of the staircase climbing task. b Stick diagram decomposition of leg movements and foot trajectory during a trial with continuous stimulation. The instantaneous stimulation amplitude is indicated at the bottom. c Bar diagram representing the average step height in trials on a flat surface or when a staircase is positioned in front of the animal, with continuous stimulation and proportional control (n = 5). Right: Stick diagram decomposition of leg movements and foot trajectory during a trial with proportional brain–spine interface. The instantaneous stimulation amplitude is indicated at the bottom. d Circular plots reporting the relative percent of trials with a successful step onto the elevated platform (pass), a tumble (hitting the foot against the staircase) and a fall when climbing the staircase with continuous stimulation or proportional brain–spine interface (n = 5 rats). Quantification was performed blindly by two independent experts and averaged. Right: Mean trajectory of the foot (n = 10 trials) when passing the first staircase with continuous and with proportional brain–spine interface (n = 5 rats). e Continuous cumulative firing when progressing on the flat surface and during the first two steps on the staircase. Bar plots reporting the mean change in cumulative firing when progressing along a flat surface and up the staircase (n = 5 rats). *, P < 0.05. **, P < 0.01
Fig. 9
Fig. 9
Proportional BSI encodes information not found in muscle activity. a Binary control: confusion matrix (n = 5) of Foot-off decoding calculated online across the rats using cumulative firing as in Fig. 2 (bottom) and ankle flexor EMG envelope (top). b Proportional control: cumulative firing, ankle flexor envelope and foot trajectory during three consecutive steps. c Bar diagram representing the anticipation of cortical activity to leg trajectory is superior to the length of the flexion phase of swing, while the EMG correlates at shorter latencies (n = 5 rats). d Ankle flexor muscle envelope evaluated at foot-off fails to explain the final position of the foot during swing 10 days after injury or 5 weeks after injury. Conversely, cortical ensemble activity explains more than 50% of this variance when the stimulation is controlled by the proportional brain–spine interface 10 days after injury, or when tested with continuous stimulation after several weeks of rehabilitation. *, P < 0.05
Fig. 10
Fig. 10
Rehabilitation enabled by the proportional BSI improved recovery. a Individual gait cycles recorded during overground locomotion with continuous stimulation every week, from week 2 to week 5, are displayed in the PC space for two representative rats. The PC analysis was applied on all the gait cycles from all rats at all the time-points (w, week). b Bar plot reporting the mean values of distances between trained rats (n = 6 and 7 rats for the trained group) and intact rats (n = 6) in the PC space over the course of the gait rehabilitation program. This value decreases with improved locomotor performance. c Bar plots reporting the mean values of body weight support capacities and maximum foot speed during swing. *, P < 0.05
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Bouton CE, et al. Restoring cortical control of functional movement in a human with quadriplegia. Nature. 2016;533:247–250. doi: 10.1038/nature17435. - DOI - PubMed
    1. Ajiboye AB, et al. Restoration of reaching and grasping movements through brain-controlled muscle stimulation in a person with tetraplegia: a proof-of-concept demonstration. Lancet. 2017;389:1821–1830. doi: 10.1016/S0140-6736(17)30601-3. - DOI - PMC - PubMed
    1. Capogrosso M, et al. A brain-spine interface alleviating gait deficits after spinal cord injury in primates. Nature. 2016;539:284–288. doi: 10.1038/nature20118. - DOI - PMC - PubMed
    1. Ethier C, Oby ER, Bauman MJ, Miller LE. Restoration of grasp following paralysis through brain-controlled stimulation of muscles. Nature. 2012;485:368–371. doi: 10.1038/nature10987. - DOI - PMC - PubMed
    1. Moritz CT, Lucas TH, Perlmutter SI, Fetz EE. Forelimb movements and muscle responses evoked by microstimulation of cervical spinal cord in sedated monkeys. J. Neurophysiol. 2007;97:110–120. doi: 10.1152/jn.00414.2006. - DOI - PubMed

Publication types