ARTICLE IN PRESS
G Model NSR 3614 1–9
Neuroscience Research xxx (2013) xxx–xxx
Contents lists available at ScienceDirect
Neuroscience Research journal homepage: www.elsevier.com/locate/neures
Review article
1
Corticospinal neuroprostheses to restore locomotion after spinal cord injury
2
3
4 5 6
Q1
David Borton a , Marco Bonizzato b , Janine Beauparlant a , Jack DiGiovanna b , Eduardo M. Moraud b,c , Nikolaus Wenger a , Pavel Musienko a , Ivan R. Minev d , Stéphanie P. Lacour d , José del R. Millán e , Silvestro Micera b,f , Grégoire Courtine a,∗ a
Center for Neuroprosthetics and Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland Translational Neural Engineering Laboratory, Center for Neuroprosthetics and Institute of Bioengineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland 9 c Automatic Control Laboratory, Swiss Federal Institute of Technology (ETH), Zurich, Switzerland 10 11 Q2 d Laboratory for Soft Bioelectronic Interfaces, Center for Neuroprosthetics, IMT/IBI, EPFL, Switzerland e Laboratory for Non-Invasive Brain-Machine Interface, Center for Neuroprosthetics and Institute of Bioengineering, Ecole Polytechnique Fédérale de 12 Lausanne (EPFL), Lausanne, Switzerland 13 f The BioRobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy 14 7 8
b
15
16
a r t i c l e
i n f o
a b s t r a c t
17 18 19 20
Article history: Received 1 October 2013 Accepted 3 October 2013 Available online xxx
21
26
Keywords: Neuroprosthetics Brain–machine interface Spinal interface Neuromotor rehabilitation
27
Contents
22 23 24 25
28 29 30 31 32 33 34 35 36 37 38
1. 2. 3. 4. 5. 6. 7. 8.
In this conceptual review, we highlight our strategy for, and progress in the development of corticospinal neuroprostheses for restoring locomotor functions and promoting neural repair after thoracic spinal cord injury in experimental animal models. We specifically focus on recent developments in recording and stimulating neural interfaces, decoding algorithms, extraction of real-time feedback information, and closed-loop control systems. Each of these complex neurotechnologies plays a significant role for the design of corticospinal neuroprostheses. Even more challenging is the coordinated integration of such multifaceted technologies into effective and practical neuroprosthetic systems to improve movement execution, and augment neural plasticity after injury. In this review we address our progress in rodent animal models to explore the viability of a technology-intensive strategy for recovery and repair of the damaged nervous system. The technical, practical, and regulatory hurdles that lie ahead along the path toward clinical applications are enormous – and their resolution is uncertain at this stage. However, it is imperative that the discoveries and technological developments being made across the field of neuroprosthetics do not stay in the lab, but instead reach clinical fruition at the fastest pace possible. © 2013 Published by Elsevier Ireland Ltd and the Japan Neuroscience Society.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spinal neuroprostheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flexible multi-electrode arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Closed-loop control algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Locomotor related information in the motor cortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Corticospinal neuroprosthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Corticospinal neuroprosthesis to enhance plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conflict of interest statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
00 00 00 00 00 00 00 00 00 00 00
39
∗ Corresponding author at: International Paraplegic Foundation Chair in Spinal Cord Repair, EPFL SV BMI – Station 19, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland. Tel.: +41 21 693 8343; fax: +41 21 693 07 40. E-mail address: gregoire.courtine@epfl.ch (G. Courtine). 0168-0102/$ – see front matter © 2013 Published by Elsevier Ireland Ltd and the Japan Neuroscience Society. http://dx.doi.org/10.1016/j.neures.2013.10.001
Please cite this article in press as: Borton, D., et al., Corticospinal neuroprostheses to restore locomotion after spinal cord injury. Neurosci. Res. (2013), http://dx.doi.org/10.1016/j.neures.2013.10.001
G Model NSR 3614 1–9
41
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
ARTICLE IN PRESS D. Borton et al. / Neuroscience Research xxx (2013) xxx–xxx
2
1. Introduction Various neurological diseases and traumatic injuries significantly impair motor function, often leading to a severe paralysis dramatically impacting quality of life. In the majority of neuromotor disorders, however, the spinal neuronal circuitry producing the motor patterns for standing and walking remains intact and sufficiently functional. Even after a complete spinal cord injury (SCI) abruptly disconnecting supraspinal and spinal circuits, the neuronal networks embedded in lumbosacral segments retain the capacity to generate complex motor behaviors in rats (Courtine et al., 2009a; Dominici et al., 2012), cats (Mushahwar et al., 2002, 2006; Rossignol et al., 2001) and humans (Harkema et al., 2011). Likewise, the cortical circuits largely retain their ability to control upper limb prostheses (Hochberg et al., 2012). In the spinal cord, these intrinsic motor control capabilities only emerge when delivering electrical and/or chemical excitation to the region, which transforms neuronal circuits from non-functional to highly functional networks. While these findings have been primarily obtained in rodent and cat models of SCI, preliminary results in paraplegic individuals revealed that the spinal cord of humans responds in a strikingly similar manner (Harkema et al., 2011). These encouraging results in experimental animal models and in humans have triggered a surge of interest for the development of neuroprosthetic technologies capable of taking full advantage of the intrinsic circuit properties of the spinal cord in order to restore motor control after SCI (Edgerton and Harkema, 2011). The first objective of this conceptual paper is to summarize the key concepts and complex challenges underlying the design of such spinal neuroprosthetic systems. In parallel to the search for neural repair strategies after SCI, nearly three decades of developments in neural engineering have led to the design of alternative solutions to replace lost sensorimotor functions. The proliferation of high-density neuronal recording devices (Borton et al., 2013; Miranda et al., 2010; Normann; Rizk et al., 2009) together with high-performance computing capabilities has allowed the identification of core principles through which cortical areas may contribute to coordinating movement (Schwartz, 2007). Accordingly, motor intention and execution has been extracted from neuronal ensemble modulation in the primary motor cortex (MI) (Amirikian et al., 2000; Collinger et al., 2013; Donoghue et al., 1998; Georgopoulos et al., 1992, 1982; Georgopoulos and Stefanis, 2007; Paninski et al., 2004a; VargasIrwin et al., 2010), premotor cortex (PMv and PMd) (Bansal et al., 2011; Kang et al.), parietal (PPC and PRR) (Buneo and Andersen, 2012; Hauschild et al., 2012; Lindner et al., 2010), and frontal areas (Fried et al., 2011) of able-bodied non-human primates and disabled humans. Through probabilistic and statistical algorithms, the decoded signals have been mapped onto prosthetic actions and command signals to control computers (Birbaumer et al., 1999; Hochberg et al., 2006), prosthetic limbs (Collinger et al., 2013; Hochberg et al., 2012; Velliste et al., 2008), movement of wheelchairs (Carlson and Millán, 2013) and even functional electrical stimulations of muscles (Biasiucci et al., 2013; Moritz et al., 2008; Muller-Putz et al., 2005) and spinal circuits (Nishimura et al., 2013). These advances have established the conceptual and technological bases for the design of corticospinal neuroprostheses whereby cortical modulations could directly adjust electrochemical stimulations of spinal circuits in order to restore neural communication between the brain and spinal cord after a severe SCI. The second objective of this brief communication is to summarize the concerted efforts of the Center for Neuroprosthetics (CNP) at the Swiss federal institute of technology in Lausanne (EPFL) and Zurich (ETHZ), in collaboration with a large European consortium (http://www.neuwalk.eu), to pioneer experimental corticospinal neuroprostheses with two aims: (i) restore cortical control over
adaptive locomotor movements after a complete SCI; (ii) facilitate motor execution during robot-assisted training to enhance use-dependent remodeling of spared neuronal circuits and their connections after a severe, but incomplete SCI.
2. Spinal neuroprostheses The spinal neuronal circuits associated with locomotion are generally referred to as central pattern generators (CPGs) (Grillner, 1981). Locomotor CPGs are operatively defined as neural networks that can endogenously produce rhythmic patterns of motor output, i.e. in the absence of phasic input from descending or peripheral projections. However, there is overwhelming evidence that the properties of spinal neuronal circuits expand well beyond those of a basic machine producing stereotypical patterns of motor activity. Instead, the spinal brain acts as a smart information-processing interface that actively processes multifaceted sensory information and, on this basis, takes decision on how to adjust lower limb kinematics in order to meet environmental constraints while maintaining stability. In the complete absence of supraspinal influences, the lumbosacral spinal cord of rats, cats, and humans remains capable of producing organized motor patterns for standing (Harkema et al., 2011), walking at various speeds and in many directions (Beres-Jones and Harkema, 2004; Courtine et al., 2009a; Harkema et al., 2011, 1997; Musienko et al., 2007), and even climbing a staircase (Dominici et al., 2012). However, after a severe SCI the interruption of descending pathways deprives spinal circuitry of the essential source of modulation and excitation to enable permissive states of sensorimotor circuits. Consequently, the motor control capacities of spinal neuronal network fail to manifest (Edgerton et al., 2008), which results in complete and permanent paralysis. Over the past decade, we have methodically developed an electrochemical spinal neuroprosthesis with the aim to transform non-functional lumbosacral circuits to highly functional networks (Fig. 1A) (Courtine et al., 2009a; Gerasimenko et al., 2007; Ichiyama et al., 2008; Lavrov et al., 2008; Musienko et al., 2011). This electrochemical neuroprosthesis combines a cocktail of monoaminergic agents, and epidural electrical stimulation (EES) applied over the dorsal aspect of lumbar and sacral segments of the spinal cord (Fig. 1B). These inputs replace the missing source of modulation (chemical stimulation) and excitation (electrical stimulation) after the interruption of descending pathways. Together, electrochemical stimulations enable well-coordinated hindlimb movements during overground locomotion in otherwise paralyzed rats (Courtine et al., 2009a; Dominici et al., 2012). Preliminary testing showed that electrical spinal cord stimulation similarly enabled standing and manually assisted stepping in a chronically paralyzed individual (Harkema et al., 2011). The locations, features, and functional effects of stimulation were identical to those used in rats. Most studies in experimental animals have delivered continuous (tonic) electrical stimulation through one or two electrodes implanted along the midline of the spinal cord. To date, empirical observations have guided the positioning of electrodes. Typically, electrodes are located at lumbar segment L2 and sacral segment S1 in order to promote flexion and extension of the lower limbs, respectively (Fig. 1B). Adjustment of the stimulation parameters such as amplitude and frequency led to predictive modulation of lower limb kinematics during stepping. These combined results suggest that the lumbosacral spinal cord is composed of widely distributed, yet highly synergistic, combination of neural circuits that can be stimulated in a task specific manner to generate a wide variety of movements (Courtine et al., 2009a; Musienko et al., 2009). These findings also emphasize that the current empirical approach is sub-optimal and impractical for conceiving efficient
Please cite this article in press as: Borton, D., et al., Corticospinal neuroprostheses to restore locomotion after spinal cord injury. Neurosci. Res. (2013), http://dx.doi.org/10.1016/j.neures.2013.10.001
106 107 108 109
110
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
G Model NSR 3614 1–9
ARTICLE IN PRESS D. Borton et al. / Neuroscience Research xxx (2013) xxx–xxx
3
Fig. 1. Site specific modulation of standing and stepping through epidural spinal cord stimulation (EES) (adapted from Courtine et al., 2009a,b). Rats with complete midthoracic transection trained on a treadmill with robotic support system (A). Transformation of non-functional spinal sensorimotor circuits into highly functional states 7 days post injury (B). Site specific modulation of EES during standing. Stimulation of L2 electrodes resulted in whole limb flexion, stimulation of S1 electrodes in whole limb extension.
169 170 171 172 173 174
175
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
multisite stimulation protocols to restore motor control after SCI. Optimization of electrical spinal cord stimulation paradigms will critically rely on (i) a computational model to predict optimal electrode positioning, (ii) novel technology yielding flexible multielectrode arrays, and (iii) closed-loop control systems to achieve the online adjustment of multisite stimulations.
interface large areas of the spinal cord with high-density electrode configurations capable of delivering both electrical, and potentially chemical, stimulations for extended durations in experimental animal models.
3. Flexible multi-electrode arrays
New implantable technologies provide the basis for the development of stimulation strategies that take full advantage of the distributed neural circuitry in the mammalian spinal cord. By finetuning the parameters of stimulation adaptively, specific aspects of locomotion may be reinforced to ameliorate individual deficits in the gait of the animal. This type of stimulation is likely to improve the outcome of rehabilitative training when employed during treadmill training (Gerasimenko et al., 2007; Ichiyama et al., 2008; van den Brand et al., 2012). However, the tuning of multiple
The efficiency of a spinal neural interface rests on its ability to endure the extreme mechanical strain inherent in such an active biological substrate. Innovative use of materials is paramount to achieving electrode-to-neuron interfaces that establish a bidirectional link between the stimulating electrodes and the neural circuitry. Until recently, implantable neural interfaces were manufactured with stiff metallic foils embedded with conductors and coated in elastomeric packaging for flexibility and ionic protection – processes and technologies borrowed from the microelectronic industry (Cheung, 2007). The structural and mechanical biocompatibility of the neural interface has only recently been a concern in the design of neuroprostheses (Bellamkonda et al., 2012). Most neural electrodes available today are stiff and brittle, while neural tissues conversely are soft, fluid-bathed, and viscoelastic materials. The mechanical mismatch at this interface, combined with local micromotions, may induce an inflammatory reaction of immune cells, the generation of fibrotic tissue, withdrawal or death of the nearby neurons, progressive loss of electrode contact, and ultimately implant failure. Soft microelectrode arrays (SMEAs) are electrode interfaces embedded in a thin (
G Model NSR 3614 1–9
Neuroscience Research xxx (2013) xxx–xxx
Contents lists available at ScienceDirect
Neuroscience Research journal homepage: www.elsevier.com/locate/neures
Review article
1
Corticospinal neuroprostheses to restore locomotion after spinal cord injury
2
3
4 5 6
Q1
David Borton a , Marco Bonizzato b , Janine Beauparlant a , Jack DiGiovanna b , Eduardo M. Moraud b,c , Nikolaus Wenger a , Pavel Musienko a , Ivan R. Minev d , Stéphanie P. Lacour d , José del R. Millán e , Silvestro Micera b,f , Grégoire Courtine a,∗ a
Center for Neuroprosthetics and Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland Translational Neural Engineering Laboratory, Center for Neuroprosthetics and Institute of Bioengineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland 9 c Automatic Control Laboratory, Swiss Federal Institute of Technology (ETH), Zurich, Switzerland 10 11 Q2 d Laboratory for Soft Bioelectronic Interfaces, Center for Neuroprosthetics, IMT/IBI, EPFL, Switzerland e Laboratory for Non-Invasive Brain-Machine Interface, Center for Neuroprosthetics and Institute of Bioengineering, Ecole Polytechnique Fédérale de 12 Lausanne (EPFL), Lausanne, Switzerland 13 f The BioRobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy 14 7 8
b
15
16
a r t i c l e
i n f o
a b s t r a c t
17 18 19 20
Article history: Received 1 October 2013 Accepted 3 October 2013 Available online xxx
21
26
Keywords: Neuroprosthetics Brain–machine interface Spinal interface Neuromotor rehabilitation
27
Contents
22 23 24 25
28 29 30 31 32 33 34 35 36 37 38
1. 2. 3. 4. 5. 6. 7. 8.
In this conceptual review, we highlight our strategy for, and progress in the development of corticospinal neuroprostheses for restoring locomotor functions and promoting neural repair after thoracic spinal cord injury in experimental animal models. We specifically focus on recent developments in recording and stimulating neural interfaces, decoding algorithms, extraction of real-time feedback information, and closed-loop control systems. Each of these complex neurotechnologies plays a significant role for the design of corticospinal neuroprostheses. Even more challenging is the coordinated integration of such multifaceted technologies into effective and practical neuroprosthetic systems to improve movement execution, and augment neural plasticity after injury. In this review we address our progress in rodent animal models to explore the viability of a technology-intensive strategy for recovery and repair of the damaged nervous system. The technical, practical, and regulatory hurdles that lie ahead along the path toward clinical applications are enormous – and their resolution is uncertain at this stage. However, it is imperative that the discoveries and technological developments being made across the field of neuroprosthetics do not stay in the lab, but instead reach clinical fruition at the fastest pace possible. © 2013 Published by Elsevier Ireland Ltd and the Japan Neuroscience Society.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spinal neuroprostheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flexible multi-electrode arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Closed-loop control algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Locomotor related information in the motor cortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Corticospinal neuroprosthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Corticospinal neuroprosthesis to enhance plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conflict of interest statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
00 00 00 00 00 00 00 00 00 00 00
39
∗ Corresponding author at: International Paraplegic Foundation Chair in Spinal Cord Repair, EPFL SV BMI – Station 19, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland. Tel.: +41 21 693 8343; fax: +41 21 693 07 40. E-mail address: gregoire.courtine@epfl.ch (G. Courtine). 0168-0102/$ – see front matter © 2013 Published by Elsevier Ireland Ltd and the Japan Neuroscience Society. http://dx.doi.org/10.1016/j.neures.2013.10.001
Please cite this article in press as: Borton, D., et al., Corticospinal neuroprostheses to restore locomotion after spinal cord injury. Neurosci. Res. (2013), http://dx.doi.org/10.1016/j.neures.2013.10.001
G Model NSR 3614 1–9
41
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
ARTICLE IN PRESS D. Borton et al. / Neuroscience Research xxx (2013) xxx–xxx
2
1. Introduction Various neurological diseases and traumatic injuries significantly impair motor function, often leading to a severe paralysis dramatically impacting quality of life. In the majority of neuromotor disorders, however, the spinal neuronal circuitry producing the motor patterns for standing and walking remains intact and sufficiently functional. Even after a complete spinal cord injury (SCI) abruptly disconnecting supraspinal and spinal circuits, the neuronal networks embedded in lumbosacral segments retain the capacity to generate complex motor behaviors in rats (Courtine et al., 2009a; Dominici et al., 2012), cats (Mushahwar et al., 2002, 2006; Rossignol et al., 2001) and humans (Harkema et al., 2011). Likewise, the cortical circuits largely retain their ability to control upper limb prostheses (Hochberg et al., 2012). In the spinal cord, these intrinsic motor control capabilities only emerge when delivering electrical and/or chemical excitation to the region, which transforms neuronal circuits from non-functional to highly functional networks. While these findings have been primarily obtained in rodent and cat models of SCI, preliminary results in paraplegic individuals revealed that the spinal cord of humans responds in a strikingly similar manner (Harkema et al., 2011). These encouraging results in experimental animal models and in humans have triggered a surge of interest for the development of neuroprosthetic technologies capable of taking full advantage of the intrinsic circuit properties of the spinal cord in order to restore motor control after SCI (Edgerton and Harkema, 2011). The first objective of this conceptual paper is to summarize the key concepts and complex challenges underlying the design of such spinal neuroprosthetic systems. In parallel to the search for neural repair strategies after SCI, nearly three decades of developments in neural engineering have led to the design of alternative solutions to replace lost sensorimotor functions. The proliferation of high-density neuronal recording devices (Borton et al., 2013; Miranda et al., 2010; Normann; Rizk et al., 2009) together with high-performance computing capabilities has allowed the identification of core principles through which cortical areas may contribute to coordinating movement (Schwartz, 2007). Accordingly, motor intention and execution has been extracted from neuronal ensemble modulation in the primary motor cortex (MI) (Amirikian et al., 2000; Collinger et al., 2013; Donoghue et al., 1998; Georgopoulos et al., 1992, 1982; Georgopoulos and Stefanis, 2007; Paninski et al., 2004a; VargasIrwin et al., 2010), premotor cortex (PMv and PMd) (Bansal et al., 2011; Kang et al.), parietal (PPC and PRR) (Buneo and Andersen, 2012; Hauschild et al., 2012; Lindner et al., 2010), and frontal areas (Fried et al., 2011) of able-bodied non-human primates and disabled humans. Through probabilistic and statistical algorithms, the decoded signals have been mapped onto prosthetic actions and command signals to control computers (Birbaumer et al., 1999; Hochberg et al., 2006), prosthetic limbs (Collinger et al., 2013; Hochberg et al., 2012; Velliste et al., 2008), movement of wheelchairs (Carlson and Millán, 2013) and even functional electrical stimulations of muscles (Biasiucci et al., 2013; Moritz et al., 2008; Muller-Putz et al., 2005) and spinal circuits (Nishimura et al., 2013). These advances have established the conceptual and technological bases for the design of corticospinal neuroprostheses whereby cortical modulations could directly adjust electrochemical stimulations of spinal circuits in order to restore neural communication between the brain and spinal cord after a severe SCI. The second objective of this brief communication is to summarize the concerted efforts of the Center for Neuroprosthetics (CNP) at the Swiss federal institute of technology in Lausanne (EPFL) and Zurich (ETHZ), in collaboration with a large European consortium (http://www.neuwalk.eu), to pioneer experimental corticospinal neuroprostheses with two aims: (i) restore cortical control over
adaptive locomotor movements after a complete SCI; (ii) facilitate motor execution during robot-assisted training to enhance use-dependent remodeling of spared neuronal circuits and their connections after a severe, but incomplete SCI.
2. Spinal neuroprostheses The spinal neuronal circuits associated with locomotion are generally referred to as central pattern generators (CPGs) (Grillner, 1981). Locomotor CPGs are operatively defined as neural networks that can endogenously produce rhythmic patterns of motor output, i.e. in the absence of phasic input from descending or peripheral projections. However, there is overwhelming evidence that the properties of spinal neuronal circuits expand well beyond those of a basic machine producing stereotypical patterns of motor activity. Instead, the spinal brain acts as a smart information-processing interface that actively processes multifaceted sensory information and, on this basis, takes decision on how to adjust lower limb kinematics in order to meet environmental constraints while maintaining stability. In the complete absence of supraspinal influences, the lumbosacral spinal cord of rats, cats, and humans remains capable of producing organized motor patterns for standing (Harkema et al., 2011), walking at various speeds and in many directions (Beres-Jones and Harkema, 2004; Courtine et al., 2009a; Harkema et al., 2011, 1997; Musienko et al., 2007), and even climbing a staircase (Dominici et al., 2012). However, after a severe SCI the interruption of descending pathways deprives spinal circuitry of the essential source of modulation and excitation to enable permissive states of sensorimotor circuits. Consequently, the motor control capacities of spinal neuronal network fail to manifest (Edgerton et al., 2008), which results in complete and permanent paralysis. Over the past decade, we have methodically developed an electrochemical spinal neuroprosthesis with the aim to transform non-functional lumbosacral circuits to highly functional networks (Fig. 1A) (Courtine et al., 2009a; Gerasimenko et al., 2007; Ichiyama et al., 2008; Lavrov et al., 2008; Musienko et al., 2011). This electrochemical neuroprosthesis combines a cocktail of monoaminergic agents, and epidural electrical stimulation (EES) applied over the dorsal aspect of lumbar and sacral segments of the spinal cord (Fig. 1B). These inputs replace the missing source of modulation (chemical stimulation) and excitation (electrical stimulation) after the interruption of descending pathways. Together, electrochemical stimulations enable well-coordinated hindlimb movements during overground locomotion in otherwise paralyzed rats (Courtine et al., 2009a; Dominici et al., 2012). Preliminary testing showed that electrical spinal cord stimulation similarly enabled standing and manually assisted stepping in a chronically paralyzed individual (Harkema et al., 2011). The locations, features, and functional effects of stimulation were identical to those used in rats. Most studies in experimental animals have delivered continuous (tonic) electrical stimulation through one or two electrodes implanted along the midline of the spinal cord. To date, empirical observations have guided the positioning of electrodes. Typically, electrodes are located at lumbar segment L2 and sacral segment S1 in order to promote flexion and extension of the lower limbs, respectively (Fig. 1B). Adjustment of the stimulation parameters such as amplitude and frequency led to predictive modulation of lower limb kinematics during stepping. These combined results suggest that the lumbosacral spinal cord is composed of widely distributed, yet highly synergistic, combination of neural circuits that can be stimulated in a task specific manner to generate a wide variety of movements (Courtine et al., 2009a; Musienko et al., 2009). These findings also emphasize that the current empirical approach is sub-optimal and impractical for conceiving efficient
Please cite this article in press as: Borton, D., et al., Corticospinal neuroprostheses to restore locomotion after spinal cord injury. Neurosci. Res. (2013), http://dx.doi.org/10.1016/j.neures.2013.10.001
106 107 108 109
110
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
G Model NSR 3614 1–9
ARTICLE IN PRESS D. Borton et al. / Neuroscience Research xxx (2013) xxx–xxx
3
Fig. 1. Site specific modulation of standing and stepping through epidural spinal cord stimulation (EES) (adapted from Courtine et al., 2009a,b). Rats with complete midthoracic transection trained on a treadmill with robotic support system (A). Transformation of non-functional spinal sensorimotor circuits into highly functional states 7 days post injury (B). Site specific modulation of EES during standing. Stimulation of L2 electrodes resulted in whole limb flexion, stimulation of S1 electrodes in whole limb extension.
169 170 171 172 173 174
175
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
multisite stimulation protocols to restore motor control after SCI. Optimization of electrical spinal cord stimulation paradigms will critically rely on (i) a computational model to predict optimal electrode positioning, (ii) novel technology yielding flexible multielectrode arrays, and (iii) closed-loop control systems to achieve the online adjustment of multisite stimulations.
interface large areas of the spinal cord with high-density electrode configurations capable of delivering both electrical, and potentially chemical, stimulations for extended durations in experimental animal models.
3. Flexible multi-electrode arrays
New implantable technologies provide the basis for the development of stimulation strategies that take full advantage of the distributed neural circuitry in the mammalian spinal cord. By finetuning the parameters of stimulation adaptively, specific aspects of locomotion may be reinforced to ameliorate individual deficits in the gait of the animal. This type of stimulation is likely to improve the outcome of rehabilitative training when employed during treadmill training (Gerasimenko et al., 2007; Ichiyama et al., 2008; van den Brand et al., 2012). However, the tuning of multiple
The efficiency of a spinal neural interface rests on its ability to endure the extreme mechanical strain inherent in such an active biological substrate. Innovative use of materials is paramount to achieving electrode-to-neuron interfaces that establish a bidirectional link between the stimulating electrodes and the neural circuitry. Until recently, implantable neural interfaces were manufactured with stiff metallic foils embedded with conductors and coated in elastomeric packaging for flexibility and ionic protection – processes and technologies borrowed from the microelectronic industry (Cheung, 2007). The structural and mechanical biocompatibility of the neural interface has only recently been a concern in the design of neuroprostheses (Bellamkonda et al., 2012). Most neural electrodes available today are stiff and brittle, while neural tissues conversely are soft, fluid-bathed, and viscoelastic materials. The mechanical mismatch at this interface, combined with local micromotions, may induce an inflammatory reaction of immune cells, the generation of fibrotic tissue, withdrawal or death of the nearby neurons, progressive loss of electrode contact, and ultimately implant failure. Soft microelectrode arrays (SMEAs) are electrode interfaces embedded in a thin (
