Spinal cord injury (SCI) induces haemodynamic instability that threatens survival1,2,3, impairs neurological recovery4,5, increases the risk of cardiovascular disease6,7, and reduces quality of life8,9. Haemodynamic instability in this context is due to the interruption of supraspinal efferent commands to sympathetic circuits located in the spinal cord10, which prevents the natural baroreflex from controlling these circuits to adjust peripheral vascular resistance. Epidural electrical stimulation (EES) of the spinal cord has been shown to compensate for interrupted supraspinal commands to motor circuits below the injury11, and restored walking after paralysis12. Here, we leveraged these concepts to develop EES protocols that restored haemodynamic stability after SCI. We established a preclinical model that enabled us to dissect the topology and dynamics of the sympathetic circuits, and to understand how EES can engage these circuits. We incorporated these spatial and temporal features into stimulation protocols to conceive a clinical-grade biomimetic haemodynamic regulator that operates in a closed loop. This ‘neuroprosthetic baroreflex’ controlled haemodynamics for extended periods of time in rodents, non-human primates and humans, after both acute and chronic SCI. We will now conduct clinical trials to turn the neuroprosthetic baroreflex into a commonly available therapy for people with SCI.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Data that support the findings and software routines developed for the data analysis will be made available upon reasonable request to the corresponding authors. Source data are provided with this paper.
Squair, J. W., Phillips, A. A., Harmon, M. & Krassioukov, A. V. Emergency management of autonomic dysreflexia with neurologic complications. Can. Med. Assoc. J. 188, 1100–1103 (2016).
Readdy, W. J. et al. Complications and outcomes of vasopressor usage in acute traumatic central cord syndrome. J. Neurosurg. Spine 23, 574–580 (2015).
Inoue, T., Manley, G. T., Patel, N. & Whetstone, W. D. Medical and surgical management after spinal cord injury: vasopressor usage, early surgerys, and complications. J. Neurotrauma 31, 284–291 (2014).
Squair, J. W. et al. Spinal cord perfusion pressure predicts neurologic recovery in acute spinal cord injury. Neurology 89, 1660–1667 (2017).
Squair, J. W. et al. Empirical targets for acute haemodynamic management of individuals with spinal cord injury. Neurology 93, e1205–e1211 (2019).
Cragg, J. J., Noonan, V. K., Krassioukov, A. & Borisoff, J. Cardiovascular disease and spinal cord injury: results from a national population health survey. Neurology 81, 723–728 (2013).
Wu, J.-C. et al. Increased risk of stroke after spinal cord injury: a nationwide 4-year follow-up cohort study. Neurology 78, 1051–1057 (2012).
Illman, A., Stiller, K. & Williams, M. The prevalence of orthostatic hypotension during physiotherapy treatment in patients with an acute spinal cord injury. Spinal Cord 38, 741–747 (2000).
Carlozzi, N. E. et al. Impact of blood pressure dysregulation on health-related quality of life in persons with spinal cord injury: development of a conceptual model. Arch. Phys. Med. Rehabil. 94, 1721–1730 (2013).
Furlan, J. C., Fehlings, M. G., Shannon, P., Norenberg, M. D. & Krassioukov, A. V. Descending vasomotor pathways in humans: correlation between axonal preservation and cardiovascular dysfunction after spinal cord injury. J. Neurotrauma 20, 1351–1363 (2003).
Capogrosso, M. et al. A brain-spine interface alleviating gait deficits after spinal cord injury in primates. Nature 539, 284–288 (2016).
Wagner, F. B. et al. Targeted neurotechnology restores walking in humans with spinal cord injury. Nature 563, 65–71 (2018).
Saadoun, S., Chen, S. & Papadopoulos, M. C. Intraspinal pressure and spinal cord perfusion pressure predict neurological outcome after traumatic spinal cord injury. J. Neurol. Neurosurg. Psychiatry 88, 452–453 (2016).
Vale, F. L., Burns, J., Jackson, A. B. & Hadley, M. N. Combined medical and surgical treatment after acute spinal cord injury: results of a prospective pilot study to assess the merits of aggressive medical resuscitation and blood pressure management. J. Neurosurg. 87, 239–246 (1997).
Rawlings, A. M. et al. Association of orthostatic hypotension with incident dementia, stroke, and cognitive decline. Neurology 91, e759–e768 (2018).
Phillips, A. A., Krassioukov, A. V., Ainslie, P. N. & Warburton, D. E. R. Baroreflex function after spinal cord injury. J. Neurotrauma 29, 2431–2445 (2012).
Phillips, A. A., Krassioukov, A. V., Ainslie, P. N. & Warburton, D. E. R. Perturbed and spontaneous regional cerebral blood flow responses to changes in blood pressure after high-level spinal cord injury: the effect of midodrine. J. Appl. Physiol. 116, 645–653 (2014).
Courtine, G. et al. Transformation of nonfunctional spinal circuits into functional states after the loss of brain input. Nat. Neurosci. 12, 1333–1342 (2009).
Angeli, C. A. et al. Recovery of over-ground walking after chronic motor complete spinal cord injury. N. Engl. J. Med. 379, 1244–1250 (2018).
Gill, M. L. et al. Neuromodulation of lumbosacral spinal networks enables independent stepping after complete paraplegia. Nat. Med. 24, 1677–1682 (2018).
Wenger, N. et al. Spatiotemporal neuromodulation therapies engaging muscle synergies improve motor control after spinal cord injury. Nat. Med. 22, 138–145 (2016).
West, C. R. et al. Association of epidural stimulation with cardiovascular function in an individual with spinal cord injury. JAMA Neurol. 75, 630–632 (2018).
Harkema, S. J. et al. Epidural spinal cord stimulation training and sustained recovery of cardiovascular function in individuals with chronic cervical spinal cord injury. JAMA Neurol. 75, 1569–1571 (2018).
Harkema, S. J. et al. Normalization of blood pressure with spinal cord epidural stimulation after severe spinal cord injury. Front. Hum. Neurosci. 12, 83 (2018).
Darrow, D. et al. Epidural spinal cord stimulation facilitates immediate restoration of dormant motor and autonomic supraspinal pathways after chronic neurologically complete spinal cord injury. J. Neurotrauma 36, 2325–2336 (2019).
Grimm, D. et al. In vitro and in vivo gene therapy vector evolution via multispecies interbreeding and retargeting of adeno-associated viruses. J. Virol. 82, 5887–5911 (2008).
Witten, I. B. et al. Recombinase-driver rat lines: tools, techniques, and optogenetic application to dopamine-mediated reinforcement. Neuron 72, 721–733 (2011).
Tomer, R., Ye, L., Hsueh, B. & Deisseroth, K. Advanced CLARITY for rapid and high-resolution imaging of intact tissues. Nat. Protocols 9, 1682–1697 (2014).
Strack, A. M., Sawyer, W. B., Marubio, L. M. & Loewy, A. D. Spinal origin of sympathetic preganglionic neurons in the rat. Brain Res. 455, 187–191 (1988).
Capogrosso, M. et al. A computational model for epidural electrical stimulation of spinal sensorimotor circuits. J. Neurosci. 33, 19326–19340 (2013).
Gradinaru, V., Thompson, K. R. & Deisseroth, K. eNpHR: a Natronomonas halorhodopsin enhanced for optogenetic applications. Brain Cell Biol. 36, 129–139 (2008).
Phillips, A. A., Elliott, S. L., Zheng, M. M. Z. & Krassioukov, A. V. Selective alpha adrenergic antagonist reduces severity of transient hypertension during sexual stimulation after spinal cord injury. J. Neurotrauma 32, 392–396 (2015).
Beauparlant, J. et al. Undirected compensatory plasticity contributes to neuronal dysfunction after severe spinal cord injury. Brain 136, 3347–3361 (2013).
Minev, I. R. et al. Biomaterials. Electronic dura mater for long-term multimodal neural interfaces. Science 347, 159–163 (2015).
Formento, E. et al. Electrical spinal cord stimulation must preserve proprioception to enable locomotion in humans with spinal cord injury. Nat. Neurosci. 21, 1728–1741 (2018).
Cornwell, W. K. III et al. Restoration of pulsatile flow reduces sympathetic nerve activity among individuals with continuous-flow left ventricular assist devices. Circulation 132, 2316–2322 (2015).
Purohit, S. N., Cornwell, W. K. III, Pal, J. D., Lindenfeld, J. & Ambardekar, A. V. Living without a pulse: the vascular implications of continuous-flow left ventricular assist devices. Circ. Heart Fail. 11, e004670 (2018).
Cheng, A., Williamitis, C. A. & Slaughter, M. S. Comparison of continuous-flow and pulsatile-flow left ventricular assist devices: is there an advantage to pulsatility? Ann. Cardiothorac. Surg. 3, 573–581 (2014).
Saleem, S. et al. Wavelet decomposition analysis is a clinically relevant strategy to evaluate cerebrovascular buffering of blood pressure after spinal cord injury. Am. J. Physiol. Heart Circ. Physiol. 314, H1108–H1114 (2018).
Phillips, A. A., Warburton, D. E. R., Ainslie, P. N. & Krassioukov, A. V. Regional neurovascular coupling and cognitive performance in those with low blood pressure secondary to high-level spinal cord injury: improved by alpha-1 agonist midodrine hydrochloride. J. Cereb. Blood Flow Metab. 34, 794–801 (2014).
Courtine, G. & Bloch, J. Defining ecological strategies in neuroprosthetics. Neuron 86, 29–33 (2015).
Phillips, A. A. & Krassioukov, A. V. Contemporary cardiovascular concerns after spinal cord injury: mechanisms, maladaptations, and management. J. Neurotrauma 32, 1927–1942 (2015).
Phillips, A. A. & Krassioukov, A. V. in Neurological Aspects of Spinal Cord Injury (eds Weidner, N. et al.) 325–361 (Springer International Publishing, 2017).
Richardson, R. R., Cerullo, L. J. & Meyer, P. R. Autonomic hyper-reflexia modulated by percutaneous epidural neurostimulation: a preliminary report. Neurosurgery 4, 517–520 (1979).
Ramsey, J. B. G. et al. Care of rats with complete high-thoracic spinal cord injury. J. Neurotrauma 27, 1709–1722 (2010).
Anderson, M. A. et al. Required growth facilitators propel axon regeneration across complete spinal cord injury. Nature 561, 396–400 (2018).
Squair, J. W. et al. High thoracic contusion model for the investigation of cardiovascular function after spinal cord injury. J. Neurotrauma 34, 671–684 (2016).
Asboth, L. et al. Cortico-reticulo-spinal circuit reorganization enables functional recovery after severe spinal cord contusion. Nat. Neurosci. 21, 576–588 (2018).
Scheff, S. W., Rabchevsky, A. G., Fugaccia, I., Main, J. A. & Lumpp, J. E., Jr. Experimental modeling of spinal cord injury: characterization of a force-defined injury device. J. Neurotrauma 20, 179–193 (2003).
van den Brand, R. et al. Restoring voluntary control of locomotion after paralyzing spinal cord injury. Science 336, 1182–1185 (2012).
Krassioukov, A. V. & Weaver, L. C. Connections between the pontine reticular formation and rostral ventrolateral medulla. Am. J. Physiol. 265, H1386–H1392 (1993).
Ueno, M., Ueno-Nakamura, Y., Niehaus, J., Popovich, P. G. & Yoshida, Y. Silencing spinal interneurons inhibits immune suppressive autonomic reflexes caused by spinal cord injury. Nat. Neurosci. 19, 784–787 (2016).
Sundt, D., Gamper, N. & Jaffe, D. B. Spike propagation through the dorsal root ganglia in an unmyelinated sensory neuron: a modeling study. J. Neurophysiol. 114, 3140–3153 (2015).
McIntyre, C. C. & Grill, W. M. Extracellular stimulation of central neurons: influence of stimulus waveform and frequency on neuronal output. J. Neurophysiol. 88, 1592–1604 (2002).
Pan, C. et al. Shrinkage-mediated imaging of entire organs and organisms using uDISCO. Nat. Methods 13, 859–867 (2016).
Lee, E. et al. ACT-PRESTO: Rapid and consistent tissue clearing and labeling method for 3-dimensional (3D) imaging. Sci. Rep. 6, 18631 (2016).
Bria, A. & Iannello, G. TeraStitcher - a tool for fast automatic 3D-stitching of teravoxel-sized microscopy images. BMC Bioinformatics 13, 316 (2012).
Rizk, A. et al. Segmentation and quantification of subcellular structures in fluorescence microscopy images using Squassh. Nat. Protocols 9, 586–596 (2014).
Kirshblum, S. C. et al. International standards for neurological classification of spinal cord injury (revised 2011). J. Spinal Cord Med. 34, 535–546 (2011).
Bogert, L. W. J. & van Lieshout, J. J. Non-invasive pulsatile arterial pressure and stroke volume changes from the human finger. Exp. Physiol. 90, 437–446 (2005).
Jansen, J. R. et al. A comparison of cardiac output derived from the arterial pressure wave against thermodilution in cardiac surgery patients. Br. J. Anaesth. 87, 212–222 (2001).
Westerhof, B. E., Gisolf, J., Stok, W. J., Wesseling, K. H. & Karemaker, J. M. Time-domain cross-correlation baroreflex sensitivity: performance on the EUROBAVAR data set. J. Hypertens. 22, 1371–1380 (2004).
Wieling, W., Ganzeboom, K. S. & Saul, J. P. Reflex syncope in children and adolescents. Heart 90, 1094–1100 (2004).
Whinnett, Z. I. et al. Multicenter randomized controlled crossover trial comparing haemodynamic optimization against echocardiographic optimization of av and VV delay of cardiac resynchronization therapy: the BRAVO trial. JACC Cardiovasc. Imaging 12, 1407–1416 (2019).
Notay, K. et al. Validity and reliability of measuring resting muscle sympathetic nerve activity using short sampling durations in healthy humans. J. Appl. Physiol. 121, 1065–1073 (2016).
Incognito, A. V. et al. Evidence for differential control of muscle sympathetic single units during mild sympathoexcitation in young, healthy humans. Am. J. Physiol. Heart Circ. Physiol. 316, H13–H23 (2019).
Wallin, B. G. et al. Sympathetic single axonal discharge after spinal cord injury in humans: activity at rest and after bladder stimulation. Spinal Cord 52, 434–438 (2014).
Incognito, A. V. et al. Pharmacological assessment of the arterial baroreflex in a young healthy obese male with extremely low baseline muscle sympathetic nerve activity. Clin. Auton. Res. 28, 593–595 (2018).
See Supplementary Notes. Support: Compute Canada, the Natural Sciences and Engineering Research Council, the Canadian Institutes of Health Research, Banting Fellowship, Alberta Innovates Health Solutions, Campus Alberta Neuroscience, the Libin Cardiovascular Institute, the Hotchkiss Brain Institute, PRAXIS, International Foundation for Research in Paraplegia, McCaig Institute for Bone and Joint Health, European Union’s Horizon 2020 (no. 785907 Human Brain Project SGA2, no. 842578 and no. 665667), RESTORE: Eurostars E10889, OPTISTIM: Eurostars E!12743, the Swiss National Science Foundation (NCCR Robotics, 310030_185214, CRSII3_160696), European Research Council (ERC-2015-CoG HOW2WALKAGAIN 682999), and Bertarelli Foundation. M.A.A. is supported by a SNF Ambizione fellowship (PZ00P3_185728).
G.C., J.B., J.W.S., A.A.P. and S.P.L hold various patents in relation with the present work. R.B. and N.B. are Medtronic employees. Y.C. and Q.L. are Motac neuroscience employees. E.B. is consultant and shareholder of Motac neuroscience. G.C., J.B., S.P.L. and A.A.P. are shareholders of Onward, a company with potential relationships with the invention.
Peer review information Nature thanks Patrice Guyenet, Ryan Solinsky, Veronica Tom and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Step 1 (a): We first tested the capacity for an orthostatic challenge to reduce blood pressure in rats 30 days after a T3 spinal cord injury (n = 4). Tilting rats 90° upright did not lead to any reduction in systolic blood pressure (one-way repeated measures ANOVA; F3 = 0.612; P = 0.62), diastolic blood pressure (one-way repeated measures ANOVA; F3 = 1.105; P = 0.40), or mean arterial pressure (one-way repeated measures ANOVA; F3 = 0.915; P = 0.47). Data are mean ± s.e.m. Step 2(b): To confirm that our contusion model disrupted descending control of haemodynamics we used AAV-DJ-hSyn-flex-mGFP-2A-synaptophysin-mRuby injected into the RVLM of TH-Cre rats (n = 4) one month after T3 spinal cord injury. We found a near complete disruption of descending THON sympatho-excitatory axons (paired one-tailed t-test; t = 5.08; P = 0.007) and synapses (paired one-tailed t-test; t = 4.64; P = 0.009) when comparing counts above and below the injury. Bar charts represent the mean with raw data overlaid. Step 3(c): Overview of the time-course used to examine the natural history of haemodynamic instability in uninjured (n = 6) and spinal cord injured animals (n = 6). Confirmation that the lesion site spared minimal white matter (identified using GFAP; mean = 2%). Step 4(d): We implanted a wireless recording system to monitor haemodynamics and sympathetic nerve activity. A blood pressure cannula was inserted into the abdominal aorta and microelectrodes sutured to the sympathetic renal nerve. Step 5(e): Data were recorded 24/7 and automatically uploaded to a server where automated analyses were triggered to quantify blood pressure and sympathetic nerve activity throughout the day and night. Step 6(f): We established the natural history of haemodynamic instability by recording the response to spinal cord injury and automatically detecting outliers (see Methods, ‘Haemodynamic and sympathetic nerve activity monitoring’) for blood pressure and sympathetic nerve activity data for a total of 7 weeks. Step 7(g): Because rats do not exhibit haemodynamic instability in response to an orthostatic challenge (see ‘Step 1(a)’), we developed a servo-controlled negative-pressure approach whereby animals are placed in a chamber and the pressure is dropped and monitored in closed loop (see Methods, ‘Implementation of a simulated orthostatic challenge in rodents and non-human primates’). *P < 0.05; **P < 0.01; ***P < 0.001. Source data
Extended Data Fig. 2 Rodent model develops haemodynamic instability across the natural history of spinal cord injury.
a, We recorded haemodynamics and sympathetic nerve activity during and after the spinal cord contusion (n = 6). b, We observed an immediate increase in blood pressure and sympathetic nerve activity following the onset of the contusion. c, Quantifications revealed a significant increase in systolic blood pressure (paired one-tailed t-test; 91 mmHg vs 128 mmHg; t = 5.40; P = 0.001) and mean arterial pressure (paired one-tailed t-test; 63 mmHg vs 96 mmHg; t = 4.50; P = 0.003), a decrease in heart rate (paired one-tailed t-test; 309 bpm vs 100 bpm; t = −10.26; P = 7.56 × 10−5), and an increase in sympathetic nerve activity (paired one-tailed t-test; t = 2.26; P = 0.037) during contusion. These were followed by complete reversal after the contusion, where systolic blood pressure (paired one-tailed t-test; 116 mmHg vs 79 mmHg; t = −5.97; P = 0.0009), mean arterial pressure (paired one-tailed t-test; 85 mmHg vs 53 mmHg; t = −5.14; P = 0.002), sympathetic nerve activity (paired one-tailed t-test; t = −3.29; P = 0.011), and heart rate (paired one-tailed t-test; 350 bpm vs 313 bpm; t = −2.91; P = 0.017) decreased compared to pre-injury. d, We next established the natural history of haemodynamics using 24/7 recordings. We found that throughout the recording period animals with spinal cord injury deviated outside key thresholds representing ‘normal’ values (points scaled by size and transparency based on their deviation outside our set thresholds (dotted lines)). e, We found a left-shift in the distribution of haemodynamic values, and a right-ward shift in heart rate values, indicating generally lower blood pressure, higher heart rate, and more aberrant sympathetic nerve activity (Kolmogorov–Smirnov test; all P < 2.2 × 10−16). f, Quantifications revealed an increase in the number of deviations for systolic blood pressure (independent samples one-tailed t-test; t = 5.92; P = 0.0005), diastolic blood pressure (independent samples one-tailed t-test; t = 3.68; P = 0.007), mean arterial pressure (independent samples one-tailed t-test; t = 3.23; P = 0.011), heart rate (independent samples one-tailed t-test; t = 2.0; P = 0.0499), and sympathetic nerve activity (independent samples one-tailed t-test; t = 3.20; P = 0.006). We also found an increase in the variance of systolic blood pressure (independent samples one-tailed t-test; t = 2.70; P = 0.011), diastolic blood pressure (independent samples one-tailed t-test; t = 2.01; P = 0.036), mean arterial pressure (independent samples one-tailed t-test; t = 2.85; P = 0.009), and sympathetic nerve activity (independent samples one-tailed t-test; t = 3.20; P = 0.006), fitting the criteria for haemodynamic instability after SCI. g, Formal baseline recordings revealed that baseline systolic blood pressure (two-way repeated measures ANOVA; interaction effect F6,1 = 7.05; P = 1.10 × 10−5; all Tukey post hoc P < 0.001) and mean arterial pressure (two-way repeated measures ANOVA; interaction effect F6,1 = 4.93; P = 0.0004; all Tukey post hoc P < 0.001) were reduced, in agreement with spontaneous 24/7 data recordings. Data are mean ± s.e.m. h, We devised a closed-loop negative-pressure system to mimic an orthostatic challenge in rats. Animals with spinal cord injury could not respond to decreasing pressures, whereas uninjured animals responded and slowly recovered. This response was consistent across all six weeks post injury and is contrasted against week 0 (no injury for both groups). i, Quantification of this response revealed increased negative deltas for the spinal cord injury group for systolic blood pressure (two-way repeated measures ANOVA; interaction effect F6,1 = 3.71; P = 0.003; all Tukey post hoc P < 0.05) and mean arterial pressure (two-way repeated measures ANOVA; interaction effect F6,1 = 4.20; P = 0.001; all Tukey post hoc P < 0.05). j, After SCI we observed a time-dependent increase in the linear relationship between chamber pressure and blood pressure (likelihood ratio test of nested models; P < 0.001), indicating that haemodynamics cannot be stabilized during orthostatic challenge without a functioning baroreflex. In all panels, percentage change is presented for clarity as needed, while all statistics are calculated from raw values. Bar charts represent the mean with raw data overlaid. *P < 0.05; **P < 0.01; ***P < 0.001. HR, heart rate; iSNA, integrated sympathetic nerve activity; MAP, mean arterial pressure. Source data
Extended Data Fig. 3 Functional and anatomical mapping reveal haemodynamic hotspots preferentially enriched to respond to epidural electrical stimulation.
Step 1(a): We confirmed the role of the splanchnic ganglia in activating pressor responses. Electrical stimulation of splanchnic ganglia led to pressure responses that linearly increased with the stimulation amplitude (n = 4, linear mixed modelling; all R2 > 0.71; all P < 6.72 × 10−12). These relationships were robust across systolic blood pressure, diastolic blood pressure, and mean arterial pressure. Step 2(b): Retrograde tracing of splanchnic ganglia using fluorogold labelled sympathetic pre-ganglionic neurons, which we confirmed using ChAT staining. Successful injection was confirmed in splanchnic ganglia by colocalization of fluorogold with TH. We found a peak density in the number of sympathetic pre-ganglionic neurons projecting to splanchnic ganglia in the lower thoracic segments. Step 3(c): We next completed functional mapping of the spinal cord by iterating through each segment and stimulating epidurally to activate pressor responses. We completed these experiments in animals one hour after injury (acute, n = 3), one-week after injury (sub-acute, n = 3), two weeks after injury (intermediate, n = 5), and four weeks after injury (chronic, n = 5). We recorded blood pressure using a terminal carotid catheterization preparation, and calculated the peak change in blood pressure during the stimulation. Step 4(d): We found that the response to targeted epidural electrical stimulation (EES) increased with time after injury, and that there was a haemodynamic hotspot in the lower thoracic spinal cord. Step 5(e): We found a linear relationship between the functional and anatomical mapping results, providing a clear rationale for the existence of the observed haemodynamic hotspot (linear modelling; all R2 > 0.72; all P < 0.003). Step 6(f): We used computational modelling (Extended Data Fig. 4) and found that lateral edge (near root entrance/exit) placement of electrodes resulted in preferential recruitment of T12, even with EES placed at L2. This was in contrast to more midline stimulation where we saw preferential recruitment off L2. Indeed, with electrodes placed with a subtle ‘lateral shift’ the shift in recruitment to T12 already begins. We reasoned this may be the mechanism by which serendipitous clinical observations were made. We confirmed this hypothesis using a rhizotomy experiment, in which we cut the T12 root after stimulating with EES at L2 (n = 5). We found a significant blunting of the EES response in systolic blood pressure (paired one-tailed t-test; 10 mmHg vs 6.9 mmHg; t = −3.95; P = 0.008), mean arterial pressure (paired one-tailed t-test; 9.9 mmHg; vs 5.6 mmHg; t = −6.03; P = 0.002), and diastolic blood pressure (paired one-tailed t-test; 9.8 mmHg; vs 5.3 mmHg; t = −5.90; P = 0.002). Bar charts represent the mean with raw data overlaid. Source data
Extended Data Fig. 4 Epidural electrical stimulation induces pressor responses through the recruitment of posterior afferents and excitatory interneurons.
Step 1(a): We developed a hybrid computational model based on real anatomical structures generated through high-resolution computed tomography and MRI scans. This model combines a geometrically realistic 3D finite element model of the spinal cord with realistic compartmental cable models of all afferent neurons, efferent neurons and some interneurons. We established a computational pipeline to obtain anisotropic tissue property maps, discretize the model, perform simulations using an electro-quasi-static solver and couple these simulations with NEURON-based electrophysiology models (Sim4Life by ZMT, www.zurichmedtech.com). We investigated the recruitment patterns of various afferent and efferent fibres within the spinal cord structure. We found that stimulation over the dorsal aspect of the spinal cord led to high levels of recruitment of major afferents, before any recruitment of efferent neurons directly from the stimulation. This suggested that epidural electrical stimulation activates pressor responses by recruiting afferents. Step 2(b): Next, we experimentally tested the hypothesis that pressor responses induced by EES were dependent on afferent activation. We completed successive dorsal rhizotomies at T11, T12 and T13 and found a graded reduction in the response to stimulation (one-way ANOVA; all P < 0.001; post hoc results indicated), with the largest decrease when removing T12, consistent with our functional and anatomical mapping results. Grey box indicates stimulation. Bar charts represent the mean with raw data overlaid. Step 3(c): Next, we developed a NEURON-based spiking neural network model composed of integrate-and-fire neurons to predict the presence of direct, indirect excitatory, and indirect inhibitory connections. Indirect inhibitory connections resulted in poor sympathetic pre-ganglionic neuron recruitment (left) and in the minimization of membrane potentials in response to increasing stimulation amplitude (right; various stimulation amplitudes indicated by alpha; action potential threshold indicated by horizontal dotted line; stimulation onset indicated by vertical dotted line). This suggested that pressor responses to EES likely are mediated by either direct, monosynaptic connections between afferents and sympathetic pre-ganglionic neurons or by indirect circuits including excitatory interneurons. Step 4(d): We completed anterograde tracing of the dorsal root ganglia. Using dynamic image registration we generated a digital dorsal horn whereby we could select a region of interest (ROI; grey box) and determine the mean intensity (‘Observed ROI’) of either axons (orange) or synapses (red). Using 1000 bootstraps of random ROIs as a null distribution we found a depletion of axons (empirical P = 0.019) and synapses (empirical P = 0.001) in the intermediolateral column. We confirmed this result by counting neurons with appositional synapses on ChATON neurons in the ventral horn versus the lateral horn and found a similar statistical depletion (n = 10 images, 294 neurons; Fisher’s exact test; odds ratio (OR): 0.082; P < 2.2 × 10−16). This suggested that the most likely circuit mediating these responses instead included an excitatory interneuron. We therefore completed retrograde trans-synaptic tracing and found interneurons trans-synaptically connected to splanchnic ganglia that were SLC17A6 positive, and had VGLUT1 synaptic puncta in their immediate vicinity (see Fig. 2), suggesting direct connections with large diameter afferents. *P < 0.05; **P < 0.01; ***P < 0.001. Rz, rhizotomy. Source data
Extended Data Fig. 5 Epidural electrical stimulation induces pressor responses through the recruitment of splanchnic ganglia.
Step 1(a): To determine whether epidural electrical stimulation activated splanchnic ganglia we stimulated rats with T3 spinal cord injury for 30 min (n = 5), or did not stimulate them (n = 5). We collected the splanchnic ganglia and subjected them to immunohistochemisty. We stained each section for TH, FOS (immediate early gene), DAPI, and NISSL to confirm neuronal phenotypes. We then used an automatic cell segmentation algorithm to identify cells that coexpressed TH, NISSL and/or FOS. Step 2(b): Quantifications revealed that animals stimulated with 30 min of EES had more FOS-positive neurons than unstimulated rats (independent samples one-tailed t-test; t = 6.56; P = 0.001), and that these neurons colocalized with TH and NISSL stains (both P < 0.001). Step 3(c): We next completed two loss of function studies to examine the causal role of the splanchnic ganglia in the pressor responses to EES. We conducted an optogenetics silencing experiment (n = 4) and an axotomy experiment whereby we surgically severed the connection between the sympathetic pre-ganglionic neurons in the spinal cord and the splanchnic ganglia (n = 4). We completed these experiments on animals 60 days after T3 spinal cord injury. For optogenetics experiments viral injections occurred at 30 days after injury. Activating the inhibitory opsin eNpHr3.0 with yellow light suppressed the pressor response to EES, and therefore increased the difference between the peak response to EES and the minimum response during light (or control) for systolic blood pressure (independent samples one-tailed t-test; 7 mmHg vs 12 mmHg; t = 2.90; P = 0.031) and mean arterial pressure (independent samples one-tailed t-test; 7 mmHg vs 13 mmHg; t = 2.60; P = 0.040). Step 4(d): Axotomy of the connection between sympathetic pre-ganglionic neurons and splanchnic ganglia likewise blunted the pressor response for systolic blood pressure (n = 4; independent samples one-tailed t-test; 20 mmHg vs 9.2 mmHg; t = −4.54; P = 0.001), diastolic blood pressure (independent samples one-tailed t-test; 19.6 mmHg vs 7 mmHg; t = −2.40; P = 0.048), and mean arterial pressure (independent samples one-tailed t-test; 19 mmHg vs 7 mmHg; t = −3.08; P = 0.027). These experiments indicate the splanchnic ganglia are necessary to induce a pressor response with EES. Step 5(e): We next tested whether blocking α1 receptors on systemic blood vessels would blunt the response to EES. We administered prazosin intravenously and found an immediate loss of pressor responses to EES (n = 7). The response returned after drug washout (one-way repeated measures ANOVA; 18 mmHg vs 5.7 mmHg vs 13.1 mmHg; F2 = 15.63; P = 0.0001; Tukey HSD). *P < 0.05; **P < 0.01; ***P < 0.001. Source data
Extended Data Fig. 6 Mechanistic insight enabled the development of a targeted electronic dura mater and biomimetic stimulation strategy to recapitulate the natural dynamics of the sympathetic nervous system.
Step 1(a): To develop the spatial features of an electrode array specifically targeting haemodynamic hotspots we first quantified all the features of the low thoracic spinal column. We measured the precise length of each spinal segment and vertebrae using a combination of gross anatomical dissections, high resolution computed tomography scans, and custom MRI sequences. Step 2(b): We used these anatomical features, driven by the identification of haemodynamic hotspots to develop an electronic dura mater specifically targeting T11, T12 and T13 spinal segments. Detailed dimensions of the active stimulation sites (500 μm diameter, 0.002 cm2 geometric surface area). Step 3(c): When placed on the spinal cord, the array increases blood pressure and all electrodes are functional. Top, electrochemical impedance spectrum (modulus, left and phase, right) of an electrode array acquired in vitro post-fabrication. Step 4(d): To recapitulate the natural dynamics of the sympathetic nervous system we first recorded neural activity from the renal sympathetic nerve and blood pressure from the descending aorta. We measured these signals in response to a hypotensive stimulus (sodium nitroprusside (SNP)) in both injured and uninjured animals (n = 5; uninjured example shown). We found that there was an impaired response in the sympathetic nerve activity after SCI. To quantify the changes in these dynamics we trained a feed-forward neural network to predict a continuous output from a given input. For example, predicting systolic blood pressure from sympathetic nerve activity (iSNA). We found that in uninjured animals, there were strong correlations whereby the model could predict one from the other. In injured animals this correlation was absent. Here, we show the ability of the model to predict SBP from iSNA in response to this stimulus (strength of correlation (Pearson correlation) presented as −log10(P) for each group). Responses are presented on a normalized delta scale to account for absolute differences between animals. Step 5(e): To understand the timing delay of RVLM activation to sympathetic outflow from the spinal cord we stimulated the RVLM electrically, and measured the efferent volley over T11, T12 and T13 (n = 5). Stimulation of the RVLM dramatically increased blood pressure, confirming localization of the stimulation. We then measured the delay between action potentials in response to 100 Hz 10 s pulse trains of RVLM stimulation and found a 2.5 ± 0.4 ms delay between segments. Representative traces across segments are shown for one animal. We therefore integrated this delay into the stimulation design between segments. Step 6 (f): Finally, to understand the precise role of frequency dynamics in blood pressure control we stimulated the RVLM using optogenetics in TH-Cre rats. We found that stimulation with blue light led to a robust increase in wavelet spectrogram within the 0.4–1.0 Hz band (paired samples one-tailed t-test; t = 2.67; P = 0.028). This was in contrast to activation of an inhibitory opsin using a yellow laser (which would, in this case, inhibit the RVLM due to the presence of an inhibitory opsin), which showed significantly less activation compared to blue light (paired samples one-tailed t-test; t = 2.44; P = 0.035). In response to an orthostatic challenge, the wavelet power response in uninjured rats was less pronounced in the presence of inhibitory (yellow) light (bottom; independent samples one-tailed t-test; t = 4.04; P = 0.008). *P < 0.05; **P < 0.01; ***P < 0.001. Source data
Step 1(a): We tested whether we could stabilize haemodynamics using the neuroprosthetic baroreflex, operating in a closed loop, in animals with acute (n = 7; 12 h after injury) and chronic (n = 6; one month after injury) T3 spinal cord injury. We implemented the neuroprosthetic baroreflex within research-grade technology to achieve precise control over stimulation parameters. Step 2(b): Schematic of electronic dura mater electrode arrays targeting haemodynamic hotspots. Bar chart shows relative pressor responses (systolic blood pressure; n = 5, independent samples one-tailed t-test; t = 3.90; P = 0.006). Step 3(c): We found a linear relationship between stimulation amplitude and the pressor response to stimulation in both animals with acute (mixed model linear regression; R2 = 0.84, P < 2.2 × 10−16) and chronic (mixed model linear regression; R2 = 0.81, P < 1.0 × 10−15) spinal cord injury. Step 4(d): We completed a series of trials to test the ability of the neuroprosthetic baroreflex to stabilize haemodynamics in rats with acute spinal cord injury. The neuroprosthetic baroreflex was activated in a closed loop before the activation of the lower-body negative-pressure chamber. In trials where the stimulation was ON, we found a reduction in the target (baseline) error (paired samples one-tailed t-test; systolic blood pressure (SBP): t = −6.12, P = 5.50 × 10−4; mean arterial pressure (MAP); t = −6.08, P = 4.48 × 10−4; diastolic blood pressure (DBP); t = −5.85, P = 4.34 × 10−4) reduced time outside key thresholds (−10 mmHg; paired samples one-tailed t-test; SBP: t = −12.52, P = 9.94 × 10−6; MAP; t = −12.29, P = 8.83 × 10−6; DBP; t = −11.73, P = 1.15 × 10−5), a restoration of the nonlinear relationship between blood pressure and chamber pressure, and a concomitant reduction in the linear model coefficient (likelihood ratio test of nested models; all P < 0.001). These quantifications held for systolic blood pressure (top), diastolic blood pressure (middle), and mean arterial pressure (bottom). Step 5(e): We completed the exact same experiments on animals with chronic spinal cord injury and found similar results to those of the acutely injured rats. Specifically, in trials where the stimulation was ON, we found a reduction in the target (baseline) error (paired samples one-tailed t-test; SBP: t = −3.84, P = 0.006; MAP; t = −3.83, P = 0.006; DBP; t = −3.83, P = 0.006), reduced time outside key thresholds (−10 mmHg; paired samples one-tailed t-test; SBP: t = −4.37, P = 0.004; MAP; t = −4.43, P = 0.003; DBP; t = −4.21, P = 0.004), a restoration of the nonlinear relationship between blood pressure and chamber pressure, and a concomitant reduction in the linear model coefficient (likelihood ratio test of nested models; all P < 0.001). These quantifications held for systolic blood pressure (top), diastolic blood pressure (middle), and mean arterial pressure (bottom). Step 6(f): We found that in response to stimulation blood pressure rapidly reached the set-point, with convergence times of 0.76 s in the example case presented in Fig. 3, and 1.15 s (95% confidence interval: 0.36–2.5 s) across n = 13 animals in response to the negative-pressure chamber. In this case convergence was defined as stable within 2.5 mmHg. Step 7(g): The neuroprosthetic baroreflex, acting in closed loop, re-established natural frequency dynamics (increased wavelet power in the 0.4–1.0 Hz spectrogram) in both animals with acute (paired samples one-tailed t-test; t = 4.46; P = 0.002) and chronic SCI (paired samples one-tailed t-test; t = 3.37; P = 0.014). *P < 0.05; **P < 0.01; ***P < 0.001. Source data
Step 1(a): To further establish the efficacy of the neuroprosthetic baroreflex, we performed experiments in three non-human primates. First, we measured arterial blood pressure using an invasive catheter in the subclavian artery. Next, we performed a T3 spinal cord injury to mimic the experimental conditions of our rodent experiments. We next mapped the pressor responses to epidural stimulation from T7 to L3. We combined these results with ex vivo dissections of the anatomical dimensions of the lower thoracic spinal cord in rhesus macaques (n = 3) to design an electronic dura mater. Finally, we implemented all the features of the neuroprosthetic baroreflex. Step 2(b): We tested the efficacy of the neuroprosthetic baroreflex specifically within the context of acute traumatic SCI. We emulated all the features of standard neurointensive care including arterial blood pressure measurements, clinical grade anaesthesia (intravenous propofol), as well as temperature and respiration control. We integrated our stimulation approach into clinical-grade technologies using an implantable pulse generator and a spatially selective spinal implant. Step 3(c): All the features of the neuroprosthetic baroreflex were injected into our previously used clinical-grade stimulation approach. In brief, the neuroprosthetic baroreflex received beat-by-beat continuous blood pressure to provide closed loop control. Stimulation output control was sent to the neural research programmer interface, which communicates with the implantable pulse generator through a series of Bluetooth and infrared links. These commands were then sent directly to the customized spinal implant. Step 4(d): Similar to rodent experiments, we found that T3 spinal cord injury induced a significant surge in systolic blood pressure (paired one-tailed t-test; 127 mmHg vs 213 mmHg; t = 4.15; P = 0.027), mean arterial pressure (paired one-tailed t-test; 110 mmHg vs 172 mmHg; t = 3.96; P = 0.029), diastolic blood pressure (paired one-tailed t-test; 102 mmHg; vs 151 mmHg; t = 3.80; P = 0.031), and an accompanying decrease in heart rate (paired one-tailed t-test; 111 bpm vs 76 bpm; t = −4.05; P = 0.028). By one-hour after injury, we observed clinically relevant neurogenic shock, characterized by decreased systolic blood pressure (paired one-tailed t-test; 127 mmHg vs 110 mmHg; t = −3.20; P = 0.043), mean arterial pressure (paired one-tailed t-test; 110 mmHg vs 95 mmHg; t = −5.23; P = 0.017), and diastolic blood pressure (paired one-tailed t-test; 96 mmHg vs 102 mmHg; t = −6.24; P = 0.012). Step 5(e): Despite the fact that we observed an immediate decrease in resting blood pressure, epidural electrical stimulation was able to cause an immediate and transient pressor response in all three animals. In all cases, where percentage change is presented statistics were completed on raw values. *P < 0.05; **P < 0.01; ***P < 0.001. Source data
Extended Data Fig. 9 Neuroprosthetic baroreflex implemented in non-human primates with spinal cord injury.
Step 1(a): We recorded blood pressure using an axillary artery catheterization preparation, and calculated the peak change in blood pressure during stimulation. We found that there was a haemodynamic hotspot in the lower thoracic spinal cord, specifically over segments T10, T11 and T12. Step 2(b): Accounting for the offset in segments between species, we found a linear relationship between the functional mapping results in rats and the functional mapping results in non-human primates, further confirming the localization of a haemodynamic hotspot in the last three thoracic segments (linear modelling; R2 = 0.65; P = 0.018). Step 3(c): To develop the spatial features of an electrode array specifically targeting haemodynamic hotspots in the non-human primate we quantified all the features of the low thoracic spinal column. We measured the precise length of each spinal segment and vertebra using a combination of gross anatomical dissections, high resolution computed tomography scans, and custom MRI sequences. Step 4(d): We used these anatomical features, driven by the identification of haemodynamic hotspots to develop an electronic dura mater specifically targeting T10, T11 and T12 spinal segments. Detailed dimensions of the active stimulation sites (0.7 mm diameter, 0.014 cm2 geometric surface area). Step 5(e): Electrochemical impedance spectrum (modulus, left and phase, right) of an electrode array acquired in vitro post-fabrication indicates functional, low-impedance electrodes. Step 6(f): We fully implemented a negative-pressure chamber designed to fit a non-human primate to induce a stimulated orthostatic challenge. We found that stimulation using the chamber led to an immediate and consistent decrease in blood pressure. Step 7(g): We next implemented our biomimetic stimulation protocols. We found a linear relationship between stimulation amplitude and the pressor response (n = 3 monkeys; linear modelling, R2 = 0.85; P = 9.33 × 10−5). Compared to stimulating only one hotspot, stimulating with all three sets of electrodes, and therefore targeting each haemodynamic hotspot led to a more robust increase in blood pressure (n = 2 monkeys). Step 8(h): Implementation of the neuroprosthetic baroreflex, acting in closed loop, led to sustained increases in blood pressure that did not fatigue (n = 1 monkey). In comparison, continuous open-loop stimulation, using the same stimulation parameters, elicited an increase in blood pressure that was immediately followed by rapid fatigue. For continuous stimulation, the amplitude was set to the maximum observed stimulation value in the trials with the closed-loop controller. Only biomimetic stimulation re-established the natural frequency dynamics, revealed using wavelet decomposition (n = 2 monkeys). Step 9(i): Using this closed loop approach, we found a reduction in the target (baseline) error, reduced time outside key thresholds (−5 mmHg; −2 mmHg for diastolic blood pressure), a restoration of the nonlinear relationship between blood pressure and chamber pressure, and a concomitant reduction in the linear model coefficient. These quantifications held for systolic blood pressure (top), diastolic blood pressure (middle), and mean arterial pressure (bottom). Source data
Extended Data Fig. 10 Targeted epidural electrical stimulation enables control of haemodynamics in a human with spinal cord injury.
Step 1(a): A 5-6-5 Medtronic paddle array was implanted below the T10 and T11 vertebral bodies in a 38-year-old patient with clinically-complete cervical SCI, presenting with medically-refractory orthostatic hypotension. We confirmed the paddle array location with computerized tomography and magnetic resonance imaging. Step 2(b): We recorded haemodynamics, MSNA and catecholamine levels in the patient with and without stimulation. We used personalized computational modelling and a genetic algorithm to predict the ideal spatial configurations to activate the lower thoracic spinal segments. We then used the illustrated decision-making process to determine whether to move forward with the use of a given configuration. Step 3(c): We identified an optimal configuration, which recruited the lower thoracic spinal segments, increased blood pressure, normalized plasma noradrenaline levels, and increased muscle sympathetic nerve activity. We tested other configurations that are not optimally targeted to these roots and found the blood pressure responses were not robust. Step 4(d): Activation of the posterior columns using medial electrodes did not lead to a pressor response. Step 5(e): Using the optimal configuration, we found a stepwise increase in blood pressure as we increased the amplitude of stimulation (R2 = 0.86; P = 0.02; linear regression). Data are mean ± s.e.m. Step 6(f): Blood pressure and MSNA recordings during EES. The top left panel shows the blood pressure recording and the bottom left panel shows the rectified and integrated (mean voltage) neurogram demonstrating multi-unit MSNA. Triangles identify the accepted multi-unit MSNA bursts (broad base width with a peak surpassing three times the noise width). The mean voltage neurogram is derived from the raw MSNA neurogram in the top right panel, which is the amplified and bandpass filtered neurogram. The bottom right tracing shows the MSNA action potential spike events over time that were derived from the raw MSNA neurogram above. Step 7(g): To test whether the concept of the neuroprosthetic baroreflex could be implemented in humans, we verticalized the patient to induce an orthostatic challenge while recording blood pressure. We performed feature extraction to obtain a rolling mean blood pressure, from which we calculated a continuous error (Δblood pressure). We then adjusted the weighted proportion to generate a calculated stimulation output, which was implemented using a clinician controller to modulate blood pressure in a closed loop. Step 8(h): When our closed-loop system was activated, blood pressure was tightly regulated evidenced by stabilized blood pressure and mitigated target error. Step 9(i): This patient now uses the stimulation daily and has been able to cease other treatments for orthostatic hypotension. Furthermore, the daily use of this therapy has reduced self-reported burden of orthostatic hypotension for this patient, as assessed using clinical questionnaires. AIS, American Spinal Injury Association Impairment Scale. Source data
This file contains Supplementary Methods and associated references, and extended Acknowledgements and Author Contributions.
The neuroprosthetic baroreflex implemented using clinical grade software and hardware in a non-human primate model of spinal cord injury (SCI). We emulated neuro-intensive care in an acute model of complete SCI in nonhuman primates. We connected e-dura implants to a clinical-grade implantable pulse generator with wireless communication modules and custom-made software interfaces that enable real–time control of stimulation parameters. This video shows the continuous streaming of the mean arterial pressure during simulated orthostatic challenges in a negative pressure chamber, and how the neuroprosthetic baroreflex instantly normalized blood pressure despite these pronounced orthostatic challenges.
Clinical implementation, an interview with the participant An interview with the participant that uses this targeted EES approach to stabilize blood pressure.
About this article
Cite this article
Squair, J.W., Gautier, M., Mahe, L. et al. Neuroprosthetic baroreflex controls haemodynamics after spinal cord injury. Nature 590, 308–314 (2021). https://doi.org/10.1038/s41586-020-03180-w
Nature Reviews Neuroscience (2021)