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.
From: Neuroprosthetic baroreflex controls haemodynamics after spinal cord injury
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.
