Miniature endoscope for high resolution electrophysiological recordings from the colon of live mice

Abstract

A major weakness in the field of neurogastroenterology research has been a lack of technology to determine the spatial and temporal coordination of electrical activity along the gastrointestinal (GI) tract in-vivo, without requiring a surgical procedure. To overcome this weakness, we developed a miniaturized endoscope consisting of 128 iridium oxide recording sensors that allowed us to make high resolution intraluminal electrophysiological recordings in-vivo from the mucosal surface of the terminal large intestine of anesthetized mice. Recordings revealed discharges of smooth muscle action potentials organized into complex spatiotemporal patterns. The patterns were modified by pharmacological agents donepezil and atropine that stimulated or suppressed cholinergic neurotransmission, respectively. The patterns were also ablated by benzalkonium chloride, known to disrupt the function of the enteric nervous system. The endoscope was further validated under ex-vivo recording conditions, where blocking enteric neural activity with tetrodotoxin (TTX) again altered spontaneously occurring action potential patterns. This approach offers a unique opportunity to easily characterize normal and dysfunctional patterns of GI electrical activity in genetically modified and/or diseased mouse models, including drug discovery and high-throughput studies.

Preclinical neurogastroenterology research needs tools to characterize signals underlying gut functions. Here, the authors describe a mini-endoscope for high-resolution intraluminal electrophysiological recordings from colons of live mice in wild type and disease model animals.

Introduction

Preclinical research in small animal models is advanced mainly by increasing the number of tools for detailed characterization of neurobiology spanning from immunohistochemistry to advanced high-resolution microscopy. Ex-vivo studies have been crucial in dissecting the fundamental mechanisms that govern colonic function in the absence of systemic influence, such as hormonal or extrinsic innervation^1,2. Whole colon preparations in an organ bath also facilitate targeted pharmacological manipulations without systemic effects and allow video monitoring of (micro)motion of the entire specimen^3–5. However, functional in-vivo assessment of the colon, including the enteric nervous system (ENS), is still limited to very simplified approaches like total transit time^6,7 or bead expulsion assay^8,9. Hence, the underlying, complex electrophysiological patterns of gastrointestinal (GI) motor activity that drives propulsion of content along the GI tract, and the modulations of those patterns by functional GI disease models or preclinical pharmacological interventions, remain hidden from researchers.

Most electrical recordings from the GI tract of live animals have been made after surgical implantation of recording electrodes in the outer muscle layers (the circular and longitudinal coats), known as the muscularis externa, in a variety of species, including dogs^10–12, rats¹², rabbits and pigs¹³, sheep and cows¹⁴, and humans¹⁵. Notable studies have also correlated the electrical activity with motor activity in the upper GI tract of anesthetized animals¹⁶, for review see ref. ¹⁷.

In previous studies, intraluminal suction electrodes have been used to record the myoelectric activity from the GI tract of a variety of animal models and in human large intestine^18–20. These studies used low-resolution recordings, such that only single or small numbers of recording sites were obtained. A downfall of using low-resolution data acquisition from organs like the GI tract is that the directionality and degree of spatial coordination of unique patterns of electrical activity is difficult to discriminate. This is partly because smooth muscle action potentials propagate over short distances^21,22 due to the low spatial constant of the syncytial properties of smooth muscle²³. It would be a major step forward to be able to record in high resolution the electrical activity along the smooth muscle of the GI tract in live animals without requiring invasive surgical procedures, e.g., laparotomy. Improved resolution recordings have been made, on the other hand, of the mechanical activity of the intestine using fiber Bragg grating technology²⁴; however, these recordings were based on slow mechanical distortion of the gut wall, rather than the underlying electrophysiological function driving it.

Two recent studies made intraluminal recordings from the intestine of rabbits²⁵, and another in pigs and mice²⁶. Both approaches share similar, very interesting design concepts focused on recording from larger luminal diameters, clearly prioritizing future translatability to human medical devices. For this reason, however, they necessarily compromise on the resolution, ease of manufacture, and ease of use of a device specifically needed and designed for high-resolution, high-throughput preclinical (mouse) medical research using modern technology.

Thus, to advance the state-of-the-art, we have developed a miniaturized endoscope patterned with a matrix of 128 electrodes, each capable of resolving single smooth muscle action potentials, and together offering exhaustive visualization of complex patterns of activity those action potentials form. Despite the high resolution, the endoscope remains small enough to be introduced per rectum into the colon of anesthetized mice with minimal workload burden on the experimenter and high repeatability, thus facilitating high-throughput investigations into functional GI changes in the vast array of disease models, including genetically modified mice.

Results

Contact impedance characterization

The endoscope comprises a semi-rigid ⌀2 mm and 30 mm-long tube patterned with a matrix of 128 electrode contacts (Fig. 1). Impedance spectroscopy data of the endoscope’s contacts are presented in Supplementary Fig. 5. Bench tests of the endoscope in a saline bath demonstrated median functional electrode impedances of 23.65 kΩ (9.90 kΩ interquartile range Fig. 1F). Notably, one major concern we had, pertaining to the entire concept of intraluminal recordings, was whether such values would carry over to in-situ (in-vivo), i.e., whether the endoscope electrodes would establish sufficient electrical contact with the tissue when simply inserted into the colon. Median in-situ impedances were 32.75 kΩ (12.20 kΩ interquartile range; Fig. 1F), significantly higher than the preceding saline bath measurement (p < 0.001, paired two-sided Wilcoxon signed rank test), but still within range for electrophysiological recordings for such small footprint contacts. Individual electrode impedances correlated strongly between saline and in-situ conditions (Pearson’s r = 0.76; Fig. 1F), suggesting that it is the manufacturing variability that explains much of the in-situ variance, ruling out a concern of some uncontrolled confounding factors (e.g., anatomy-related) systemically compromising intraluminal recordings.

Fig. 1. Mini-endoscope design, manufacture, and interaction with colonic mucosa.

A The polyimide electrode matrix with: 1—tissue contacts, 2—conductive paths, and 3—printed circuit board (PCB) connection pads. B Exploded view of constituent parts of the endoscope: 1–PCB) routing signal paths to connectors, 2—nylon tube, the substate of the endoscope, 3–5—housing components, 6—polyimide film bearing electrode contacts, connection pads and conducting paths, which is wrapped around the nylon tube, 7,8—reference and ground wires terminating in gold-plated subdermal needle electrodes. C Rendering of the endoscope and its positioning when inserted into the animal (mouse cartoon created in BioRender. Planchette, A. (2025) https://BioRender.com/s4zvt3q). D Microscope image of the endoscope shaft wrapped in the electrode matrix. Several tissue contacts have been marked with arrows. E Schematic cross-section of the endoscope and its relation to the colon anatomy. F Box plot of impedances of the endoscope’s electrodes in-situ, as well as in a saline bath immediately preceding and following the in-situ recording in one animal. Boxes denote the median, the lower and upper quartiles; the whiskers denote the minimum and maximum values that are not outliers. P-values denote significance in median differences in a paired two-sided Wilcoxon signed rank test; r-values denote Pearson’s coefficients of correlation across the electrodes between measurements. The few (n = 6) channels broken during device manufacturing were excluded from this data.

In-vivo recordings of intraluminal electrophysiology with high spatial resolution

We present recordings with the final iteration of the device, which were performed in six animals (three healthy and three with colonic tissue lesions described below). A hallmark feature of our intraluminal colonic recordings was the detection of prominent action potentials (“spikes”, Fig. 2A, B). These spikes had a mean half-width of 9.5 ms (no variance across three healthy mice; Fig. 2C) and mean peak-to-peak amplitude of 285, 294, and 315 µV in the three animals; Fig. 2C). A critical point regarding intraluminal recordings was whether the electrodes could maintain spatial selectivity or whether physiological fluid between them would create a common electrical environment, blurring spatial resolution. However, isolated action potentials were recorded in neighboring contacts of the endoscope with less than half the amplitude (Fig. 2D). Thus, with such steep spatial decay, we can be certain of the intraluminal recording’s ability to spatially capture signals of very local origin. Simultaneously, this also verifies optimal contact pitch of the endoscope: since spatially the sources of electrophysiological activity are somewhat oversampled, this ensures no blind spots. On average, the isolated detected action potentials exhibited not only amplitude attenuation but—along the circumferential axis—also phase reversals (Fig. 2D), suggesting some possible non-trivial interplay of electrophysiological sources and sinks across the colonic wall.

Fig. 2. Characterization of intraluminal electrophysiology.

A Example of a raw signal from one electrode of the endoscope showing the spikes, which constitute the predominant feature of the recordings. A single spike is magnified on the inset. B The blue trace is the signal from (A) high-pass filtered above 10 Hz, the blue histograms are an RMS quantification (“envelope”) of the spiking activity in that example in 0.1 and 1 s sliding windows. Since RMS is a metric of the spiking activity we rely on frequently herein, we include this illustration of how it is derived. C Mean spike traces from six mice (n = 39881, 28502, 20041, 14951, 27490, 22841), demonstrating the time course of waveform of an average isolated action potential. The spike pool used for the average was obtained by finding all negative peaks in data with a defined minimal prominence (maximum value of −150 µV and no neighbors within 50 ms) and extracting and averaging epochs centered on those peaks. D Spatial spread of an average (n = 35299) isolated spike from an example mouse (iv13wt3). Median, lower and upper quartile traces are drawn. On the inset below, the same spread is presented but in both spatial dimensions on one plot and the temporal dimension across consecutive plots (timestamps are relative to the timing of main negative peak of the spike). Diagram was created using Blender (2D).

The recorded spikes formed readily discernible and consistent spatiotemporal patterns (Fig. 3). In ketamine/medetomidine anesthetized animals, the most prominent pattern consisted of ~10 s-long bursts of spiking activity observed primarily in the oral channels ~2 times per minute, covering the distal ~10 mm of the endoscope (black stars in Fig. 3A). Each recorded burst typically had an internal pattern of activity, composed of periodic (~1 per second) burstlets (black diamonds in Fig. 3B), frequently forming propagating anterograde waves. From the aboral sections of the colon, we recorded different patterns of colonic spikes, frequently with a predominant retrograde propagation of action potentials (black triangles in Fig. 3A).

Fig. 3. Multi-channel electrophysiological patterns at rest and in response to pharmacological modulation.

A Example 3 min of signals recorded from all electrodes of the endoscope in a ketamine/medetomidine anesthetized wild type mouse (animal ID: iv13wt3) covering 24.8 mm of the distal colon. The low frequency component, visible in Fig. 2A, has been removed with 10 Hz high-pass filter, a common-average spatial filter has been applied to reduce any remaining power line noise, and a few (n = 4) broken channels have been interpolated. B Example 15 s of signals from (A) at a higher temporal magnification. C RMS heatmap representation of the spiking activity of the signal in (A). D Spectra of RMS of all the channels of 30 min data from (E), showing various rhythmic spiking patterns discernible in that data. The power spectral density has been log-transformed to better visualize power in higher frequencies. Note that the frequencies are expressed in cycles-per-minute, rather than typical Hz, as the spectra are of RMS envelopes of spike-band signals, not the signals themselves, and cycles-per-minute better cover the range of temporal dynamics of these envelopes. E Full 30-min recording from same mouse (iv13wt3). Cyan and orange outlines denote significant (bank of 1920 one-sided Wilcoxon rank-sum tests, p < 0.0000052, i.e., Bonferroni-corrected p < 0.01/1920, see Supplementary Information for details) increases and drops of spiking activity due to donepezil and atropine application (respectively) compared to the baseline before the donepezil application. The histogram above the signals shows summed RMS of spiking activity in from the oral (distal) half of the channels; the blue curve overlaid on the histogram traces a 200 s moving mean trend, with the dashed line indicating the mean pre-donepezil baseline. The blue rectangle outlines the time span zoomed in on in (A).

To quantify these periodicities in spiking activity, as well as capture any less conspicuous rhythms, we performed Fourier analysis of the RMS (calculated with a 100 ms window) of 30-min signals (Fig. 3D). This spectral analysis showed that the orally located anterograde bursts occur at 2.2 c.p.m. (black star in Fig. 3D), whereas anterograde burstlets occur indeed at 60 c.p.m. (black diamond in Fig. 3D). These findings were consistent across wild type mice (Fig. 4). Spectral representation of the signals also uncovers other rhythmicity hotspots (e.g., oral 18 c.p.m. and aboral 30 c.p.m. in Figs. 3D and 4A), that are not easily distinguished with the naked eye in the raw electrophysiology signal.

Fig. 4. Examples of electrophysiological patterns in gut-lesioned animal model.

A 3-min examples of signals from all channels from three wild type mice and log-transformed power spectra of the RMS envelopes of full 30-min recordings. Note that the frequencies are expressed in cycles-per-minute, rather than typical Hz, as the spectra are of RMS envelopes of spike-band signals, not the signals themselves, and cycles-per-minute better cover the range of temporal dynamics of these envelopes. Narrow harmonic bands seen in high frequencies in oral channels of mouse iv14wt1 are artifacts. B 3-min examples of signals from all channels from three BAC-lesioned mice and log-transformed power spectra of the RMS envelopes of full 30-min recordings. Narrow harmonic bands seen in high frequencies in oral channels of mouse iv14bac3-1 are breathing artifacts. C Aggregate spiking activity (summed RMS envelopes) from oral (distal) half of the channels in the three wild-type and three BAC-lesioned mice over 30 min (left), and their spectra (right). D An example of aberrant activity patterns in a BAC-lesioned mouse, consisting of trains of spikes with steady linear decrease in rate, likely representing spasms.

Raw recordings also consistently contain low-amplitude electrocardiographic (ECG) signals, which are straightforwardly suppressed during signal conditioning (see Supplementary Fig. 2A), and in fact afford the opportunity of very precise cardiac monitoring (see Supplementary Fig. 2C). In some mice, we also see electromyographic (EMG) activity related to breathing on the distal channels of the endoscope. Breathing EMG forms very distinct patterns both in temporal (signal waveform) and spectral domains (see Fig. 4 and Supplementary Fig. 3) and thus can be ignored during data interpretation. Nevertheless, it also can obscure some colonic signals and further work is required to develop an algorithm to remove it from the data.

In-vivo pharmacological modulation of colonic electrophysiological activity

To demonstrate in real-time the sensitivity of the device to pharmacological interventions, we modulated the activity by intraperitoneal injections of well-known pharmacological agents (Fig. 3E). Donepezil, a cholinesterase inhibitor used to enhance cholinergic tone, rapidly and significantly upregulated spiking activity (Fig. 3E, t = 7 min), while atropine, a cholinergic antagonist, had a predictable reversing effect and produced a rapid, short-lived significant reduction of spiking activity (Fig. 3E, t = 17 min). Areas of statistically significant change in activity are outlined in cyan (upregulation) and orange (downregulation). The rapid onset and decay of pharmacological effects were most evident in the oral regions that primarily exhibit ~2.2 c.p.m. bursts of activity (black stars in Fig. 3A), with the exception of a prolonged upregulation of activity at the aboral end of the colon (black triangles in Fig. 3A).

In-vivo aberrant electrophysiological patterns in gut-lesioned animal model

To verify the capability of the device to detect aberrant patterns between healthy and disrupted organs, we implemented a benzalkonium chloride (BAC) model known to disrupt colonic motility through tissue damage in three mice. None of the treated mice exhibited outward signs of distress or other spontaneous behavioral alterations. However, in all of them endoscopic intraluminal recording of the colon was altered from that described above in untreated wild type mice. We still recorded action potentials in all animals after BAC treatment (although their morphology appeared altered: Fig. 2C), but the typical spatiotemporal patterns they formed in untreated animals were abolished (Fig. 4B, C). Instead, they formed different configurations of activity in each BAC-lesioned animal. Spiking frequency spectral analysis revealed the loss of the frequency bands seen commonly in healthy mice (18-30 c.p.m. and 60 c.p.m. bands), as well as broadening of the ~2 c.p.m. band (indicating “arrythmia”) as seen in Fig. 4B. Various different patterns of electrical activity were recorded, for example, trains of spikes with steady linear decrease in firing rate (Fig. 4D, bottom).

At the oral end of the endoscope, where the propagating waves of spike bursts occur at ~2 c.p.m. in healthy controls, summed spiking activity plots reveal the global loss of rhythmicity and amplitude of spike bursts in BAC-lesioned animals (Fig. 4C, orange plots) compared to unlesioned controls (Fig. 4C, blue plots). In BAC animals, a larger spread of spiking frequency with no prominent peaks at ~2 c.p.m. was seen in Fourier frequency plots comparing the two groups (Fig. 4C). Each BAC animal exhibits different profiles of dysfunction, most easily observed in Fig. 4B; this is likely due to the variation in the effects that BAC exposure can have on tissue.

Ex-vivo colonic electrophysiological patterns

To validate our endoscope, we used an established ex-vivo setup to characterize the intrinsic neural and muscular properties of the colon in isolation^27–29. Resected intact sections of the colon were threaded over the endoscope shaft to mimic intraluminal recording. The ex-vivo set-up was slightly modified to allow the endoscope and the colon to be submerged in the bath at an angle to protect the electronics. Additionally, a benchmark suction pipette electrode was attached to the colon from the serosal side (see photo in Fig. 5A).

Fig. 5. Validation of mini-endoscope in established organ bath setup.

A Example 50-min of signals recorded from all electrodes in an ex-vivo preparation of 40 mm of the distal colon, showing spontaneous activity, as well as effects of application of tetrodotoxin (TTX, neuronal voltage-gated sodium channel blocker) and subsequent application of BAYK8644 (L-type voltage-gated calcium channel agonist). The blue histogram above the signals shows summed spiking activity (RMS envelope) across all channels. B–D Example 2-min signals zoomed in on blue boxes from (A), showing magnification of spontaneous activity and activity after TTX and BAYK8644 application, as well as concurrent instantaneous variations of colon diameter extracted from video, corresponding in time and location to the signals. See Supplementary Video 1 for actual colon in motion alongside concurrent electrophysiological signals. E 15-min, 100-s, and 5-s examples of signals from the benchmark state-of-the-art external suction pipette electrode and one channel of the endoscope nearest to it. F 40-min example of RMS envelope of spiking activity of signals recorded from a different ex-vivo sample, showing the effects of a calcium channel blocker (nicardipine); signals themselves are shown in Supplementary Fig. 4. G Example 20-min signals from another ex-vivo preparation, where mucosa was removed from distal third of the colon (see inset photo). Zoomed in 20-s sections of the signals are also shown separately from the intact and mucosa-stripped segments.

In the ex-vivo setup, we readily recorded spikes. Moreover, the spikes were confirmed to be directly related to motor activity of the colon. Instantaneous variations of colon diameter along the entire length of the preparation extracted from the video demonstrate a precise relationship between patterns of observed spikes and colonic (micro)motion (Fig. 5B–D and Supplementary Video 1). Spike bursts were markedly different than those in anesthetized animals, as they appeared only once every ~5 min (compared to twice per minute in-vivo), lasted much longer (up to 90 s compared to ~10 s in-vivo) and travelled the extent of the entire colon preparation of 40 mm (Fig. 5A).

To confirm the role of enteric neurons in regulating colonic muscle activity²⁷ using our device, the sodium channel blocker tetrodotoxin (TTX), was applied to the preparation (Fig. 5A) and repeated in four biological samples. TTX did not abolish action potentials, however the patterns of electrical activity were modified such that complex spike bursts were replaced with single spike waves that discharged rhythmically at an increased frequency (4.6 times per minute in Fig. 5A, C, and D), specifically in the aboral region of the colon. To characterize the effects of increased action potential discharge following abolition of enteric nerve conduction, we applied the Ca²⁺ channel opener, BAYK8644. This agonist did not restore the propagating waves of activity that occurred prior to TTX application. The global effects of these pharmacological interventions are evident in the RMS plots (Fig. 5A). Finally, we confirm that signals recorded with the endoscope are L-type Ca²⁺ channel-dependent action potentials by applying nicardipine to the colon from another animal, resulting in the complete removal of action potentials (Fig. 5F and Supplementary Fig. 4).

In another experiment, we compared the action potentials recorded with the suction pipette electrode, which has been established as a benchmark approach in numerous foundational studies^30–33, to the signals recorded from the contact of our endoscope nearest to it. The resemblance between the two signals was very close (Fig. 5E), with Pearson’s coefficients of correlation of their RMS envelopes at 0.87 and 0.91 for 0.1 and 1 s RMS windows (precision levels), respectively (correlations were significant at p = 0 to the level of machine precision).

Lastly, to confirm the spatial origin of the signals, the colon was opened longitudinally, and the mucosa was peeled off from the distal one third of the preparation. The tissue was mounted lumen-side down over the endoscope shaft (see photo in Fig. 5G). Electrical activity was recorded in all regions (Fig. 5G) and the propagating waves of action potentials were observed in all preparations. Partial lengthwise surgical removal of the mucosa enabled selective recording with the tissue in direct contact with the circular muscle layer, while at the same time recording from intact mucosa confirming that the endoscope records electrophysiological signals originating below the mucosa that propagate through the full thickness of the gut wall.

Discussion

We have developed a miniature, easy-to-use endoscope capable of recording in high resolution the propagation of action potentials over tens of millimeters simultaneously along the rostro-caudal and circumferential axis of colonic smooth muscle in live mice. We believe that this device can uncover anomalies and offer new insight into mechanistic understanding of gut function using disease models and genetically modified mice. Because recordings can be made minimally invasively, we can use this technique to provide repeated and rapid high-throughput experimentation without surgical intervention.

We have validated the sensitivity of our device in-vivo using pharmacological drugs (donepezil and atropine), known to modulate the activity of the ENS and/or the intestinal smooth muscle. We also performed a validation in a simple disease model using benzalkonium chloride (BAC), known to disrupt colonic motility through tissue damage, in particular to the ENS. The device consistently uncovered aberrant spiking patterns in BAC-treated mice. The data from BAC-treated mice also hints at the device’s ability to detect a disease-altered morphology of the action potential itself, however more work would be required to verify this initial observation. However, future experiments should aim to evaluate the effectiveness of the device in identifying aganglionic patterns/absences of electrical activity in other relevant rodent models of Hirschsprung disease, e.g., endothelin receptor B (Ednrb)-deficient mice^34,35. We also plan to test the utility of our device in identifying electrophysiological correlates of anal sphincter tone, with the aim of comparing them with manometric changes in animal models of anal incontinence for evaluation of potential experimental treatments.

Previous electrophysiological recordings from GI-smooth muscles in live animals have mostly—with the exception of the approaches discussed below—involved surgically invasive techniques from the outmost gut layers, often with low spatial resolution. Commonly such electrodes also used special attachment to tissue in order to ensure sufficient electrical contact and create an isolated electric compartment.

Thus, one justifiable concern with intraluminal recordings using an endoscope was tissue contact, particularly with no features for enhancing tissue adhesion through, e.g., suction, distension²⁵, or adhesion²⁶. We found, however, that a well-chosen endoscope diameter and specialized low-impedance coating on the sensors was sufficient for high-signal-to-noise ratio thus enabling consistent capture of single action potentials. However, it is possible that our approach works well in mice because their colons have no haustration and may not work as well in other species. Nevertheless, such an approach is far less cumbersome and possible to use at scale, as required, e.g., for drug discovery in preclinical research.

It is possible the distension provided by the endoscope induces patterns of myoelectric activity recorded in live animals. However, the diameter of the endoscope was made at 2 mm, which is similar to that of a murine fecal pellet and consistent with a view that the patterns recorded are physiologically relevant.

Another concern of intraluminal recordings with individual sensors bearing no isolating attachment features could be the spatial selectivity of the electrodes. Immersion of all contacts within a shared conductive fluid environment could, in principle, lead to signal spread and reduce the ability to resolve localized activity. However, our results demonstrate that this is not the case when the electrode design is properly optimized. In developing our device, we took inspiration from recent advances in micro-electrocorticography (µECoG³⁶). µECoG electrode arrays achieve high spatial resolution recording from the cerebral cortex - even sufficient for brain-computer interfacing—despite no compartmentalization, relying only on passive tissue adhesion. Similarly to µECoG, we made the footprint of the electrodes very small, but coated them with a low-impedance iridium oxide layer. As a result, we found that the amplitude of action potentials declined significantly across neighboring contacts (Fig. 2D), indicating that the signals originate from highly localized sources. Moreover, the precise electrode geometry and slight spatial oversampling of our device open the door to applying advanced computational methods. These methods will allow advances toward localization and extraction of underlying signal sources.

Electrophysiological recordings from contracting tissue are susceptible to motion artifacts. These artifacts may introduce spurious waveforms into the data, with time courses that reflect the physical displacement of tissue rather than genuine neural or myogenic activity. Ensuring that the contact is well attached to tissue is not enough to counter this: in any case, the downstream signal path (whether realized as a cable or otherwise) must change its geometry (move, bend) and generate a triboelectric artifact³⁷. To mitigate this, we restricted our analysis to high-frequency events, specifically action potentials, whose spectral signature is mostly above 10 Hz, much faster than the motion of the gut and filtered off slower activity (<10 Hz) that can be seen in raw data (Fig. 2A). To err on the side of caution, we limit our claims in the present publication to the spike components—those portions of the data that can be confidently attributed to electrophysiological activity. This decision comes at the cost of discarding potentially valuable insights contained in the slow-wave signals (see Supplementary Fig. 1), which are more susceptible to confounding factors. The confound could be resolved in the future by a multimodal approach, e.g., concurrent imaging of motility from within the endoscope thanks to its transparency or measurement of pressure if used with a compressible substrate, which would allow controlling for potential motion artifact contamination.

The in-vivo spike data do, however, still contain artifacts of physiological origin, namely ECG signals and—in some animals—EMG activity related to breathing. We found the ECG signals easily distinguishable and readily dampened to a negligible level by simple signal processing methods (see Supplementary Fig. 2A, B). Additionally, they can be easily extracted to serve as a precise heartrate readout (Supplementary Fig. 2C). Breathing EMG forms also very distinct patterns (see Fig. 4 and Supplementary Fig. 3) and, while it thus can be ignored when interpreting the data, further work is required to develop a more sophisticated suppression algorithm. Given the well-defined geometry of the endoscope, a promising approach involves volumetric modelling of electric fields coupled with independent component analysis^38,39, which would allow limiting the signal content to sources within a certain distance from the endoscope.

A recent study²⁵ used a scalable balloon catheter which enabled electrical recordings from the colon of anesthetized rabbits. That study also recorded action potentials through the mucosa in the form of periodic spike bursts. That device had longitudinal resolution limited to two locations. This additionally prompted us to develop the current device, which provides high spatial resolution of the spatial spread and coordination of action potentials, and direct assessment of the complex patterns they form.

A second recent publication proposed a device tested in pigs, as well as a version used in mice²⁶. The design of this innovative device, named Luminal Electrophysiological Neuroprofiling System (LENS), was optimized to improve and stabilize the contact between the sensors and the GI wall by combining mechanical and chemical innovation. Specifically, the developers used inflatable balloons further enhanced by an adhesive hydrogel, which is activated by an amide reaction upon contact with the mucosa milieu. The design of the device seems to reflect the authors’ ambition to progress to a human medical device, and the work appropriately focuses on the porcine tests. Consequently, not much information was provided on the mouse device (which would be relevant to contrast here), nor data recorded with it, thus it is difficult to draw comparisons. The dimensions of the device are comparable (20-by-2 mm), though the channel count was limited to16 electrodes. As a result, compared to our device, the murine LENS allowed recordings over a somewhat shorter length (2 cm vs 2.5 cm) and at a globally lower resolution. The lower specifications of the LENS mouse device are, however, understandable seeing that the authors’ ambitions lay in large animal tests and subsequent clinical translation, rather that development of an optimal small animal preclinical research tool.

The validation of our device presented here also included recordings of intraluminal electrophysiological activity in excised mice colon in an organ bath combined with simultaneous video imaging of colonic wall movements using spatiotemporal mapping. This approach confirmed that the patterns of action potentials were directly related to motor activity of the colon This allows us to infer a similar relationship occurs in-vivo, where large coverage recording of colonic motility and other micromotion at high temporal- and spatial-resolution would otherwise not be possible.

In conclusion, we believe that our easy-to-use device, with its ability to record colonic electrical activity in live mice at high resolution—down to singular action potentials—and high spatial coverage offers a unique opportunity to study GI dysfunction in disease models for drug discovery, as well as fundamental research.

Methods

Animals

In-vivo tests were performed in wild-type adult C57BL6/J mice, ranging from 25 to 35 weeks of age, raised under specific pathogen free conditions and handled in compliance with Swiss Veterinary Law guidelines. Food and water were provided ad libitum with a light/dark cycle of 7AM–7PM. All procedures were approved by the Veterinary Office of the Canton of Geneva (ethics approval GE241D).

Ex-vivo tests were performed in excised colon samples from experimentation approved by the Animal Welfare Committee of Flinders University (ethics approvals #4004 and #3999), and all protocols carried out in accordance with the National health and Medical Research Council (NHMRC) Australian code for the care and use of animals for scientific purposes (8th edition, 2013) and recommendations from the NHMRC Guidelines to promote the wellbeing of animals used for scientific purposes (2008).

Mini-endoscope for high resolution electrophysiology recording

The developed device consists of a semi-rigid ⌀2 mm and 30 mm-long cylindrical endoscope (Fig. 1). The substrate of the endoscope is a transparent nylon tube with a hemispherical distal end cap. The tube is wrapped in an electrode matrix that is a 10 μm-thick polyimide film bearing 128 tissue contacts connected to external readout pads through isolated conducting paths (Fig. 1A). The electrode matrix is designed in Python using the KLayout integrated circuit layout library (www.klayout.de) and custom Python modules, and manufactured in an ISO-7 level cleanroom on 10 inch wafers. The tissue contacts, readout pads, and conducting paths are made of platinum. The tissue contacts are circular with a 200 μm diameter. They are additionally coated with iridium oxide for improved impedance and charge injection capacity. The tissue contacts are laid out on a 32-by-4 grid with a 0.8 mm pitch along the endoscope shaft (thus longitudinally covering 24.8 mm of the colon) and 1.57 mm pitch (= 2 mm × π/4) on the circumference of the shaft. The manufactured polyimide electrode matrix is L-shaped (Fig. 1A), so that after wrapping the longer arm of the L longitudinally around the nylon tube, the shorter arm of the L forms a freely floating tab bearing readout pads for connection to a custom printed circuit board (PCB). The tab has alignment holes for precise positioning over the connection pads of the PCB, which has matching alignment holes. Connection pads of the electrode matrix are bonded to the connection pads of the custom PCB with conductive silver epoxy applied through a matching stencil. The custom PCB pins out the 128 channels to four 36 Position Dual Row Male Nano-M Connector (Omnetics Connector Corporation, USA) for connection to downstream recording/stimulating equipment (here we used the Grapevine Nomad, Ripple, LLC, USA). The extreme pins on each side of each Omnetics connector are fused and connected via the PCB to two 30 AWG (0.254 mm) ~10 cm-long insulated silver-plated copper wires terminated with ⌀0.22 mm and 13 mm-long gold-plated subdermal needles forming ground and reference electrodes. The PCB is housed in a custom 3D-printed enclosure, which also clamps the substrate nylon tube. The nylon tube contains a central channel for introduction of optional optical devices, e.g., for imaging of tissue motion. The electrode matrix wrapped around the tube is transparent enough to allow this. The design was optimized in repeated bench recordings in a saline bath, as well as in-vivo recordings in C57BL/6J mice (n = 20 in total), until the final design was converged upon as presented in Fig. 1.

In-vivo experimental recordings in anesthetized mice

During the recordings, the animals were anesthetized with a mixture of ketamine (50 mg/kg dose, QN01AX03 KETANARKON 100 ad us. Vet. Streuli Pharma) and medetomidine (5 mg/kg dose, QNC05CM91 DORBENE® ad us. vet., Dr. E. Graeub AG) by intraperitoneal injection. Prior to anesthesia induction, animals were handled for ~5 min to naturally expel any feces. Once anesthesia was established, the distal colon was additionally flushed with 2 ml of warm saline using a slightly lubricated ⌀2 mm plastic tubing attached to a syringe, inserted to a depth of 30 mm, and vitamin A ointment (1667290 VITAMINE A Blache ong opht, Bausch & Lomb Swiss AG) was applied to the eyes for hydration. The endoscope was inserted per rectum into the distal colon of the mouse up to the depth of 35 mm. Subdermal ground and reference needle electrodes were inserted under the skin of each of the hind legs of the mouse (Fig. 1). Endoscope electrode signals were monitored in real-time and recorded for up to 90 min with a Grapevine Nomad neural signals processor (Ripple, LLC, USA) at 2 kHz sampling rate with a 0.1 Hz high-pass-filter and an antialiasing low-pass filter.

In-vivo pharmacological manipulations

To ascertain the physiological origin of the observed signals in-vivo, we tested their response to intraperitoneal injections of several compounds known to modulate the activity of the ENS and/or the intestinal smooth muscle, specifically: donepezil (3 mg/kg dose, donepezil hydrochloride D6821 Sigma-Aldrich) and atropine (2 mg/kg dose, atropine sulphate A03BA01 Amino AG).

In-vivo recordings in mice with colonic tissue lesions

To assess the sensitivity of the observed signals to local lesions of the tissue, we used a benzalkonium chloride (BAC) model known to disrupt colonic motility⁴⁰. Briefly, the treatment consisted of inserting a swab soaked in 0.2% BAC to the desired depth up to 3 cm of the colon, holding it in place for 15 s before removing it for 5 min and repeating the process for a total of 8 cycles. Swabbing was chosen to minimize the invasiveness of the approach, contrary to previously published methods relying on laparotomy to treat the exterior lining of the gut wall^41,42 or local intrarectal injections^40,43.

Ex-vivo organ bath recordings of mouse colon

To validate the physiological origin and significance of activity observed in-vivo in a more controlled setting, we used excised mouse colon preparations in an organ bath. Ex-vivo experiments also provided an opportunity for measurements over a slightly larger distance, since in-vivo recordings were limited to 3.5 cm insertion to avoid tissue damage. As a result, we also created a modified version of the device with an increased longitudinal pitch of the electrodes to 1.29 mm to cover 40 mm of the excised colon.

C57BL/6 J mice (n = 11 in total) were euthanized by inhalation overdose of isoflurane. The procedures were approved by Animal Welfare Committee of Flinders University (approval No. 4004). The entire colon was removed via midline laparotomy and placed in a Petri dish containing Krebs solution (containing in mM: NaCl 118; KCl 4.7; NaH₂PO₄ 1; NaHCO₃ 25; MgCl₂ 1.2; D-Glucose 11; CaCl₂ 2.5, gassed with 95% O₂ and 5% CO₂, 36.5 °C; Chem-Supply Pty Ltd., Australia, and Vetec Fine Chemicals Ltda, Brazil). Intraluminal content was flushed with Krebs solution via syringe and the mesentery was removed with spring scissors. Preparations were then transferred to a 100 ml organ bath where the full length of the endoscope was inserted within the colonic lumen. The colon was fixed in position by ⌀100 μm stainless steel entomology pins at the distal and proximal ends to prevent longitudinal displacement during recordings. Preparations were continuously superfused with Krebs solution (36.5 °C) at ~ 3.5 mL/min.

Gut movements were recorded by video camera fixed above the organ bath (1280 × 960 pixels, 9.15 f.p.s.; Dino-Lite AM7515MZT, AnMo Electronics Corporation, Taiwan). Video was transformed into maps of circumferential gut diameter (diameter maps) with the spatiotemporal mapping technique described by ref. ⁴⁴ using custom-made software in Matlab (MathWorks, Inc., USA). Regions of minimal diameter (i.e., contraction) are represented as lightest pixels and maximal diameter (i.e., dilatation) is represented by darkest pixels.

Ex-vivo pharmacological manipulations

During ex-vivo validation, we investigated effects of application on the tissue preparation of tetrodotoxin (TTX, neuronal voltage-gated sodium channel blocker), BAYK8644 (L-type voltage-gated calcium channel agonist), and nicardipine (calcium channel blocker). Drugs were dissolved as stock solutions in water or dimethylsufoxide before dilution to the final concentrations in organ baths: tetrodotoxin (1 μM; T-550, Alomone Laboratories, Israel), nicardipine (3 μM, N7510), and BAYK8644 (0.1 μM, B112, Sigma Chemical Co., USA).

Statistical and signal processing procedures

One of the most conspicuous features of the recordings from our endoscope are smooth muscle action potentials. To visualize and quantify the intensity of this spiking activity we use a RMS (root mean square) “envelope” of the signals, after first high-pass filtering them above 10 Hz and applying a common-average reference. We use either 0.1 or 1 s-long windows for RMS calculation, depending on the time frame of interest (minutes or tens of minutes, respectively). The derivation of this metric envelope timeseries is illustrated in Fig. 2B. To capture and quantify any rhythmicity in the spiking activity, we calculate spectra of the RMS envelope using the short-time Fourier transform in 1- or 2-min Hamming windows. As RMS are not zero-mean signals, we detrend them prior to spectrum calculation.

The above signal processing pipeline and the one more involved statistical procedure used for Fig. 3E are described in detail in the Supplementary Information.

For any other results presented in this article, the more straightforward data analysis methods used are identified where applicable. All data analysis has been performed in Matlab (MathWorks, Inc., USA, version R2024b).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

A.S., A.P., and M.J.G. formulated the initial idea and hypothesis for the study. A.S., K.W., and M.J.G. developed and manufactured the device. A.S., A.P, Y.C., T.J.H., L.T., N.J.S., and M.J.G. designed and performed experiments. A.S. developed the software, processed, and analyzed the collected data. A.S., A.P., Y.C., T.J.H., M.D., N.J.S., and M.J.G interpreted the collected data. A.S. drafted the initial version of the paper. A.S., A.P., K.W., Y.C., T.J.H., L.T., M.D., N.J.S., and M.J.G. critically reviewed and revised the manuscript for intellectual content.

Peer review

Peer review information

Nature Communications thanks Michael Camilleri, Haitao Liu, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the article and its supplementary files. Any additional requests for information can be directed to and will be fulfilled by the corresponding author. Source data are provided with this paper and in the Zenodo repository, accessible via the following link: 10.5281/zenodo.17912944. Source data are provided with this paper.

Code availability

The code for the analyses in this study is available in the Zenodo repository under the following link: https://zenodo.org/records/16964635.

Competing interests

A.S., A.P., M.J.G., K.W., and Y.C. are inventors on a patent application related to the findings reported in this manuscript. The patent application (Application No. PCT/IB2025/056531) is currently pending and was filed by the Wyss Center for Bio and Neuro Engineering. The application covers the method for evaluation of electrophysiology of tubular organs via intraluminal route. These authors have no other competing interests. T.J.H., L.T., M.D., and N.J.S. have no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69144-2.