Implantable bioelectronics for gut electrophysiology

Abstract

A major regulator of gastrointestinal physiology is the enteric nervous system. This division of the autonomic nervous system is unique in its extensiveness, with neurons distributed from the esophagus to the rectum, and its capability for local information processing. However, the constant motion of the gut, arising from its relative movements in the peritoneal cavity and the peristaltic movements associated with gut motility, as well as the sparse distribution of the neurons constituting the enteric nervous system, has made access and analysis exceedingly challenging. Here, we present the construction and validation of a bioelectronic implant for accessing neural information from the distal colon. Our bioelectronic monitoring system demonstrates real-time electrophysiological recording in response to chemical and mechanical distension under anesthesia and to feeding and stress in freely-moving animals. Direct access to the communication pathways of the enteric nervous system paves the way for neuromodulation strategies targeting the gut-brain axis.

Fig. 1. Conformable devices can be surgically implanted on top of the submucosal plexus of the colon.

a Brightfield image taken in transmission showing the device layout for accessing the ENS. Devices consist of gold tracks insulated with parylene-C with poly(ethylene dioxythiophene):poly(styrene sulfonate) (PEDOT:PSS)-coated gold electrodes, containing a central array of 28 recording electrodes, each 20 µm in diameter, flanked on either side by a pair of large electrodes, 200 µm in size, which are wired but were used as surgical markers. Given that the device substrate (parylene-C) is translucent, the outline of the device has been highlighted in orange. b Photograph of device with bonded flat flexible cable (FFC) for connection to Intan recording system. Device is situated on a glass slide in this image for transport, but is removed for implantation. Black box indicates the region of device shown in a. Loops designed into the device for suturing are visible on the left, outside the black box. c Schematic detailing recording electrode layout, where electrodes are arranged as a linear array of 7 tetrodes, with each electrode separated by 50 µm from its tetrode neighbors. d–h, Steps for implanting the device into the colonic wall, images taken via stereoscope during surgery. Top shows surgical images of the steps, and the bottom shows a representative schematic for each image. Dotted line in d indicates the periphery of the section of colon under analysis. Brackets in e indicate the section of the needle that is situated in the wall of the colon. Brackets in (f, g) indicate the section of forceps that are within the wall. Dotted line in (g), highlights the outer edges of the implant. Brackets in (h, i) show the section of the implant that is in the colonic wall. Inset from (h) in the blue box is shown in (i) with the device situated beneath the muscularis externa. Dotted lines indicate where the device enters (bottom dotted line) and exits (top dotted line) the tissue. j Hematoxylin and eosin histological stain of needle placement in colonic wall, which is used to create the tunnel for the implant, as shown in (e). This image shows implantation directly below the muscularis externa. Asterisk indicates the tunnel in the tissue where the needle was threaded through the colonic wall. The lumen of the gut, which contains fecal matter, is shaded in gray.

Fig. 2. Conformable electronics enable in vivo acquisition of gut electrophysiological responses to in vivo mechanical stimuli.

a Ligated portion of the colon with a device implanted into the colonic wall prior to distension at t₀. b Distended colon during saline injection. For a and b, the colon is outlined with a dashed blue line. The implant is outlined with a dashed black line, and a bracket is used to designate the portion of the implant that is within the colon wall. c High-frequency bandpass (300–2000 Hz) and low pass (<300 Hz) representative full voltage traces. Segments highlighted in color represent the distension and the following 10 s intervals after distension. Isoflurane was initially at 1.3% (blue) and was increased to 5% (red) at the time represented by the dashed line. d Zoom in on the high-frequency bandpass traces corresponding to the distension segments. e Zoom in on the low-pass traces corresponding to the distension segments. f Normalized Area Under the Curve (nAUC) for all the low-frequency distension traces (10 s window) before and after the rise in isoflurane concentration (n = 4 rats, 2 values as low isoflurane and 2 values at high isoflurane per rat, see Supplemental Fig. 4a–f for traces from other rats). The nAUC standardizes the AUC obtained from each 10 s window by the range value within each experiment, ensuring a fair assessment of the responses across different rats. *: Statistically significant difference with p = 0.003 (t-statistic: 3.54, degrees of freedom: 14, Cohen’s d: 1.77, 95% Confidence Interval of mean difference: [0.409, 0.763]) using two-sided unpaired t-test after criteria for normality and variance homogeneity assumptions were met. Statistics for boxplots: [blue (1.3 % isoflurane): minima (0.53), maxima (0.89), median (0.69), Q1 (0.54), Q3 (0.85), whisker low bound (0.09), whisker high bound (1.31), percentiles (0.53, 0.55, 0.69, 0.85, 0.88)], [red (5% isoflurane): minima (0.005), maxima (0.34), median (0.06), Q1 (0.01), Q3 (0.17), whisker low bound (−0.22), whisker high bound (0.41), percentiles (0.006, 0.01, 0.06, 0.17, 0.31)]. The quantification of the nAUC for the high-frequency segments was not deemed meaningful due to the short duration of the response (<20 ms).

Fig. 3. Conformable electronics enable in vivo acquisition of distinct gut electrophysiological responses to various in vivo pharmacological stimuli.

Representative responses for (a) topical administration of 1 µM bradykinin, b topical administration of 500 nM capsaicin, and c intraluminal administration of 500 nM capsaicin (administered per the overall regimen shown in Supplemental Fig. 5), showing (i) spectrogram (0−2000 Hz) of power spectral density (PSD), (ii) raw voltage signal. Drug administration occurred at approximately t₀ for (a–c). Asterisk in c indicates initial high frequency response as discussed in the text. Power spectra corresponding to each data point, t_b, t₀, t₁, t₂, t₃ in the spectrograms for each drug are shown in Supplemental Fig. 6. d Normalized response per animal to different drug additions (n = 6 rats, independent experiments). The response activity elicited by each drug was quantified with the maximum [dB/(Hz*s)], mean [dB/(Hz*s)], and variance {[dB/(Hz*s)]²} of the temporal power spectral density (PSD), computed as the area under the curve (AUC) of the spectrogram power data using non-overlapping 1 s rolling windows. The responses were normalized [0,1] to the maximum and minimum values obtained among all additions (including saline) to allow for meaningful comparisons. Statistical summary: Mann-Whitney U Test: ^aU = 0.0, p = 0.0034, ^bU = 3.0, p = 0.015, ^cU = 5.0, p = 0.04, ^dU = 0.0, p = 0.0034, ^eU = 3.0, p = 0.016; Bar plots overview: Max - saline (0.05 ± 0.12), bradykinin(0.71 ± 0.33), capsaicin topical (0.26 ± 0.41), capsaicin intraluminal (0.54 ± 0.51); µ - saline (0.12 ± 0.21), bradykinin(0.66 ± 0.41), capsaicin topical (0.25 ± 0.39), capsaicin intraluminal (0.55 ± 0.50); Var - saline (0.01 ± 0.03), bradykinin(0.60 ± 0.44), capsaicin topical (0.20 ± 0.40), capsaicin intraluminal (0.51 ± 0.54); e, e, f, Administration of bradykinin for (e) and (f) was performed as follows: 1st addition, 2nd addition, saline wash, 3rd addition, 4th addition, saline wash, 5th addition, saline wash, 6th addition. e Mean of the temporal PSD (300–2000 Hz) for an initial and secondary administration of bradykinin (repeated twice, separated 2 to 3 min in between—1st and 3rd additions compared to the 2nd and 4th) with no intermediate saline wash (n = 4 rats, independent experiments). Error bars: first (0.43), second (0.35). f Mean of the temporal PSD (300–2000 Hz) for 4 additions of bradykinin with at least one saline wash step in between (n = 4 rats, independent experiments). 1st addition compared to the 3^rd, 5^th, and 6^th (Mann-Whitney U-test: *1^st vs. 5^th U = 16.0, p = 0.0265; **1^st vs. 6^th, U = 16.0, p = 0.0265; Error bars: first (0.09), second (0.37), third (0.19), fourth (0.06). Isoflurane was kept at 1.3% for the whole duration of the recordings.

Fig. 4. Surgical placement and electrophysiological recording demonstration in mouse, rat, and pig.

Image showing surgical placement of implantable device in a mouse colon, b rat colon, and c pig colon. For each image, the device is outlined with black dotted lines, and the colon is outlined with blue dotted lines. The black bracket indicates the portion of the device that resides within the colonic wall. Representative voltage traces, bandpass filtered between 0 and 2000 Hz, for d mouse, e rat, and f pig. We see similar signal-to-noise ratios for each system. g Distension trace for mouse showing time-synced (top) low-pass (0–300 Hz) and (bottom) high-band pass (300–2000 Hz). h Spectrogram showing response to capsaicin administration in the pig colon. For this experiment, ~10 mL of capsaicin was poured topically onto the region containing the implant over the course of 1-min. This spectrogram begins (time at 0 s) at the culmination of the capsaicin dosing. The maximal visible response is highlighted using an asterisk. These pilot experiments were performed in a similar fashion to the rat data, although further refinement may be necessary if conducting an entire study in a different species. However, given the general conservation in neuronal density across vertebrate species^,, the data we collect should be of similar origin to those data from the rat.

Fig. 5. High spatiotemporal resolution electrophysiologic acquisition permits identification of putative single neurons in the enteric nervous system.

a Custom implantable wire was designed to facilitate chronic device implantation. The wire has two symmetric plugs / connectors, used to insert into a zero-insertion force (ZIF)-type plug (right), which, via a custom-printed circuit board (PCB), can be interfaced with an Intan headstage for recording. The other end of the wire has an identical geometry but is bonded onto the microfabricated implant (left), which has the same replicated connector on the device itself. b Inset showing plug with intermediate wiring visible on the left of the image. The intermediate wiring is double-sided to reduce footprint and bifurcated to allow for the placement of sutures around the wire. Small tabs connect the bifurcated wires across the length of the wire, enabling the wire to be anchored at the point where it crosses the peritoneal wall for access to the colon. c Schematic detailing the placement of wiring and device within the rat. For the surgery, the wiring was routed subcutaneously between the percutaneous port down to the ventrolateral flank of the rat, where it crossed the peritoneal wall, allowing for implantation of the microfabricated portion of the device into the colon. d Illustration of the tetrode arrangement of four electrodes in one device in a chronically implanted rat. e Sample high-pass filtered traces obtained during a recording session conducted 12 days post-surgery from three implanted electrodes (color refers to matching electrode in (e) during (right; scale bar, 100 ms and 30 µV). f Averaged extracellular spike waveforms for three sample putative single neurons recorded from the implanted electrodes presenting acceptable (<500 kΩ) electrical impedance, which also presented good signal quality confirmed through offline analysis in the same chronic recording session demonstrating distinct and anatomically consistent localization across electrodes (n = 4718 spikes/neuron 1, 4748 spikes/neuron 2, 2497 spikes/neuron 3; trace, 3.5 ms, right; scale bar 3.5 ms and 10 µV). g Autocorrelations of spike occurrence for putative single neurons shown in b demonstrate physiologic refractory periods. h Cross-correlations of spike occurrence between putative single neurons reveal different co-activation patterns. i Spatiotemporal features of putative single neuron activity form distinct clusters in principal component space. j Spike occurrence attributable to putative single neurons is stable across a 30-min recording session as visualized using example channel-projected principal components.

Fig. 6. Colonic activity is modulated by food ingestion and stress.

a Normalized instantaneous firing rate of all putative single neurons aligned to the onset of feeding and sorted by strength of pre-meal firing rate (warmer colors indicated higher z-scored firing rate modulation; n = 89 neurons, from 4 rats across 3 days of recording). b Average normalized instantaneous firing rate of putative single neurons during pre-meal (magenta), meal (cyan), and post-meal (green) epochs (shaded error bars represent standard error of mean; n = 89 neurons, from 4 rats across 3 days of recording). c Comparison of putative single neuron average normalized firing rate during pre-meal, meal, and post-meal epochs (Kruskal-Wallis ANOVA, p = 4.25 × 10⁻⁷; Wilcoxon signed-rank, pre-meal vs post-meal p = 4.58 × 10⁻⁶, meal vs post-meal p = 1.90 × 10⁻⁸; 89 neurons, from 4 rats across 3 days of recording; box center = medians; box edges = upper and lower quartiles; whiskers = non-outlier minimum and maximum; notch = 5% significance level). d Image of rat in novel environment on Day 1. e Traces showing the temporal evolution of the normalized power per frequency band (μV²/H) per minute for the full duration of the chronic recording for rat 1 on day 1. The frequency bands used were: “0 - 0.2 Hz” to capture slow wave activity associated with interstitial cells of Cajal (ICCs), originated locally in the submucosal plexus or potential volume-conducted components from deeper layers such as the myenteric plexus; “0.2–1 Hz” for slow rhythms such as those in circular smooth muscle^,; “1–5 Hz” for faster rhythmic activities in smooth muscle^,; “5–300 Hz” for primary skeletal muscle activity related to tissue and body movements (EMG related); and “300–2000 Hz” for high-frequency neural components^,–. f Violin plots showing the distribution of the normalized power per frequency range (μV²/Hz) within the first 15 min (data points) for all rats for each frequency band and day to visualize the decrease in the overall response along days.