Brain-stomach interactions revealed via simultaneous Electrogastrography and fMRI

by Dr. Ignacio Rebollo
German Institute of Human Nutrition Potsdam-Rehbrücke

This user research article summarizes our publication “The sensory and motor components of the cortical hierarchy are coupled to the rhythm of the stomach during rest” Rebollo, I., & Tallon-Baudry, C. . Journal of Neuroscience, 42(11), 2205-2220.(2022) https://doi.org/10.1523/JNEUROSCI.1285-21.2021

By combining electrogastrography, a tool to measure the electrical activity of the stomach non-invasively, resting state fMRI, and phase synchronization methods, we developed a novel framework to examine brain-stomach interactions. We found that during rest, the stomach is coupled to a wide-spread cortical and subcortical network, with large effect sizes in sensory-motor regions.

Introduction

Spontaneous brain activity is organized into resting state networks (RSNs), consisting of regions displaying correlated activity and serving different sensory and cognitive functions (Yeo et al., 2011). While the spatial layout of RSNs can be described along a series of sensory-motor and transmodal cortical gradients (Huntenburg et al., 2018), the classical division of the brain into sensory-motor and transmodal regions has limited our understanding of the processing of internal bodily information. Indeed, given that the brain has to continuously monitor bodily inputs to keep the organism alive, the question of how spontaneous activity, RSNs and cortical gradients are related to bodily functions has received scarce attention. In this work, we present a framework for examining how spontaneous fMRI activity is related to the gastric rhythm, a slow (0.05 Hz) electrical oscillation intrinsically generated in the stomach wall by specialized pacemaker cells (Sanders et al., 2014). The gastric rhythm is produced at all times, setting the pace for the contractions necessary for digestion, and can be measured non-invasively with cutaneous abdominal electrodes. Known as electrogastrography (EGG), this tool can be safely implemented in the MRI environment using MRI-compatible amplifiers (Rebollo et al., 2018; Wolpert et al., 2020). Here, we combined EGG and resting state fMRI to establish how the brain is phase coupled with the gastric rhythm across RSNs, cortical regions and gradients.

Methods

Participants:

A total of 63 participants (31 females) were included in the analysis described below. The study was approved by the ethics committee Comité de Protection des Personnes Ile de France III. All participants gave written consent.

MRI acquisition and preprocessing:

MRI acquisition and preprocessing steps are described in detail in Rebollo and Tallon-Baudry (2022). Briefly,  MRI data was acquired at 3 Tesla using a Siemens MAGNETOM Verio scanner (Siemens) with a 32-channel phased-array head coil. Resting-state functional MRI time series were acquired with an echoplanar imaging (EPI) sequence with 450 volumes, a TR of 2000 ms and  3 mm3 isometric voxels. High-resolution T1-weighted structural MRI scans of the brain were acquired for anatomic reference after the functional sequence using a 3D gradient-echo sequence. fMRI preprocessing steps included slice timing and motion correction, normalization to Montreal Neurologic Institute (MNI) space, spatial smoothing, removal of linear and quadratic trends, bandpass filtering between 0.01 and 0.1 Hz and correction for cerebrospinal fluid motion.

EGG acquisition and preprocessing:

The EGG was simultaneously recorded during functional MRI acquisition with bipolar EMG electrodes (20 kΩ, 120 cm long cables) using a BrainAmp EXG amplifier (Brain Products), placed between the legs of participants. Before the recordings, the skin of participants was rubbed and cleaned with alcohol to remove dead skin, and electrolyte gel was applied. The EGG was recorded via four bipolar electrodes placed in three rows over the abdomen, with 4 cm distance between the negative and positive derivation (Figure 1, see Rebollo et al 2022 for a more detailed explanation of the electrode layout).  EGG was acquired at a sampling rate of 5000 Hz and a resolution of 0.5 μV/bit with a low-pass filter of 1000 Hz and no high-pass filter (DC recordings). Data were recorded 30 s before and after the resting state acquisition to avoid the spreading of the ringing artifact caused by the start of the MRI acquisition. The amplifier received a trigger signaling the beginning of each MRI volume.

Brain-stomach interactions revealed via simultaneous Electrogastrography and fMRI

Figure 1: The electrogastrogram (EGG).
The EGG was recorded via four bipolar electrodes placed in three rows over the abdomen.

EGG preprocessing:

Data analysis was performed using the FieldTrip toolbox (Oostenveld et al., 2010). Data were low-pass filtered below 5 Hz and down-sampled from 5000 to 10 Hz. In order to identify the EGG peak frequency  between the normal range of the stomach (0.033–0.066 Hz) of each participant, we computed the spectral density estimate at each EGG channel using Welch’s method on 200-s time windows with 150-s overlap. Data from the EGG channel displaying the largest peak power were then bandpass filtered to isolate the signal related to gastric rhythm (linear phase finite impulse response filter, FIR, designed with MATLAB function FIR2, centered at EGG peaking frequency, filter width ±0.015 Hz, filter order of 5). Data were filtered in the forward and backward directions to avoid phase distortions and down-sampled to the sampling rate of the BOLD acquisition (0.5 Hz). Filtered data included 30 s before and after the beginning and end of MRI data acquisition to minimize ringing effects. Due to the slow frequency of the gastric rhythm, no MRI gradient removal technique was necessary.

Quantification of gastric-BOLD phase coupling and effect sizes:

We used the procedure described in Rebollo et al. (2018), to quantify Gastric-BOLD coupling. Briefly, the BOLD signal of each voxel was bandpass filtered at the gastric frequency of each participant. The first and last 15 volumes (30 s) were discarded from both the BOLD and EGG time series as it is distorted by the ringing artifact produced by the start and end of  fMRI data acquisition. The Hilbert transform was then applied to the BOLD and EGG time series and the phase-locking value (PLV, Lachaux et al., 1999) was computed between gastric and each voxel phase time series. We then used a two-step statistical procedure to identify which brain regions were significantly coupled to the stomach at the group level (Rebollo et al., 2018; Richter et al., 2017). First, we calculated a chance level coupling by computing PLV between each participant’s brain data to the stomach data of the other 62 participants and taking the median value across voxels. This allows us to account for any biases in coupling that arise from using a signal with similar spectral content but not tailored to each participant’s gastric rhythm. In a second step, we used a cluster-based permutation approach that intrinsically corrects for multiple comparisons to determine significantly coupled regions at the group level (Maris & Oostenveld, 2007). Finally, for each significant gastric network voxel, we computed the effect sizes (Cohen’s d) between empirical and chance PLV and projected them into the brain surface.

Results

The gastric network (Figure 2A) comprises bilateral regions along the central, cingulate, and lateral sulci, as well as occipito-parietal-temporal regions and portions of the left striatum (Figure 2B), bilateral thalamus (Figure 2C), and cerebellum (Figure 2D).  With respect to RSNs (Yeo et al., 2011), most of the gastric network (67%) is included in the somato-motor (38%) network, which also includes the auditory cortices, and in the visual (29%) network. Gastric coupling in the somato-motor-auditory network spans somatosensory, motor and auditory regions surrounding the central, lateral and cingulate sulci. Gastric coupling in the visual network spans striate and extrastriate visual regions, and is particularly pronounced in medial and ventral occipital regions. Effect sizes were similar across visual and somato-motor networks (Figure 2E, right). Effects sizes in the few coupled regions of the default, saliency, control and attention networks were only slightly smaller than those in sensory and motor cortices. The overlap with the default network (9.5%) occurs mostly in one medial node of the default network, the retrosplenial cortex as well as in the lateral node in the superior temporal sulcus, and a small rostral prefrontal region (Figure 2E, left). Only 8.1% of the gastric network is found in the saliency network, sparing core regions such as the anterior insula and the fundus of the dorsal anterior cingulate sulcus. Only very few regions of the gastric network belonged to the control network (4.6%), dorsal attention network (3.6%), or limbic network (0.4%). Subcortical regions represented 6.8% of the gastric network. Finally, 11.9% of the gastric network is found in the cerebellum.

Brain-stomach interactions revealed via simultaneous Electrogastrography and fMRI

Figure 2: The gastric network and RSNs.
A, Effect sizes (Cohen’s d) of gastric-BOLD coupling are plotted in orange in regions significantly phase synchronized to the gastric rhythm (n = 63, voxel-level threshold = 0.05 one-sided and cluster significance <0.025 one-sided, intrinsically corrected for multiple comparisons), overlaid on top of the cortical parcellation in seven RSNs proposed by Yeo et al., 2011, color codes as in E. The gastric network also comprises left striatum (B), bilateral thalamus (C), and cerebellum (D). E, Percentage of the gastric network in each of the brain’s RSNs (left), and average effect size across significant voxels within each network (right). F, Twenty-five regions from Glasser et al. (2016) parcellation showing the largest effect sizes averaged across hemispheres. Arrowheads depict effect sizes for left (<) and right (>) hemispheres. Somatomot, somato-motor-auditory. Modified from Rebollo et al. (2022).

We then sought to characterize effect sizes and overlap across a more fine grade parcellation of the cortex (Glasser et al., 2016) consisting of 180 areas per hemisphere. For each region of the parcellation overlapping with the gastric network, we computed the percentage of the area in the gastric network and the average effect sizes across significant gastric network vertices. The 25 regions with the largest effect sizes across both hemispheres (Figure 2F), consisted mainly of primary and non-primary sensory and motor regions. Notably, primary motor and somatosensory cortices (regions 4, 3b, 1, and 3a), premotor (6d, supplementary motor and cingulate eye fields (SCEF), 6 mp, cingulate motor regions (24dd), auditory (STSda, A5, MBelt, primary auditory cortex, PBelt, RI, STGa), the granular insular, opercular area 43, early visual regions (V1, V2, V3 and VVC). Only few transmodal regions displayed large effect sizes, such as FEF from the saliency/attentional networks, the retrosplenial complex (RSC), area STSda and area 10d, three regions typically associated with the default network.

Gastric network and cortical gradients of functional connectivity

We found gastric coupling to be stronger in sensory and motor regions, which are located  opposite to transmodal regions in the gradients of functional connectivity described by (Margulies et al., 2016). In this approach, each cortical voxel can be defined by its location along two different axes, one that goes from unimodal to transmodal regions, and another one that goes from visual to transmodal to somato-motor-auditory regions. Figure 3A reproduces the findings of Margulies et al. (2016). The projection of the probability density on the first gradient shows two prominent peaks at the extremities, corresponding to transmodal and unimodal regions (Figure 3A, red curve). When considering only the gastric network, the probability density shows a markedly different profile, with an increase in the unimodal extreme only (Figure 3B, red curve). With respect to the second gradient, the projection of the whole-brain probability density on the second gradient shows a prominent peak in transmodal regions (Figure 3A, blue curve). However, gastric network voxels are more densely represented in the visual and somato-motor-auditory extremes of the gradient (Figure 3B, blue curve), indicating that coupling with the gastric rhythm is more likely to be present in unimodal than in transmodal brain regions. We then tested whether the distribution of gastric network voxels along the first and second gradients could be due to chance. For this, we computed the percentage of brain voxels that belong to the gastric network for each of the two gradients across one hundred equidistant bins (Figure 3C, orange lines). By permuting the position of the gastric network across the cortex one thousand times, we verified that such spatial biases in the distribution of the gastric network across bins could not occur randomly. Indeed, for both of the gradients, the percentage of brain voxels overlapping with the gastric network across bins is systematically larger than chance in the unimodal bins of the gradient (Figure 3C).

Brain-stomach interactions revealed via simultaneous Electrogastrography and fMRI

Figure 3: The gastric network and cortical gradients of functional connectivity. A, Density plot depicting the distribution of all cortical voxels along the first two gradients of functional connectivity described previously (Margulies et al., 2016). The first gradient (y-axis), runs from unimodal (negative values, sensory or motor regions) to transmodal regions (positive values). The second gradient (x-axis) runs from visual (positive values) to somato-motor and auditory regions (negative values). The color scale depicts the logarithm of the number of voxels. The projection of the probability density on the first and second gradients are shown in red and blue, respectively. B, Density plot of gastric network voxels on the first two gradients of functional connectivity. Gastric network voxels are located in the unimodal extremes of the two gradients. The projection of the probability density on the first and second gradients are shown in red and blue, respectively. C, Percentage of all brain voxels that belong to the gastric network (orange) for each of the two gradients, computed on 100 equidistant bins. The black circles depict the distribution of chance level overlap obtained by reallocating randomly the spatial position of gastric network voxels in the cortex. Reproduced from Rebollo et al. (2022).

Conclusions

We present here an approach combining electrophysiology and resting state fMRI with phase synchronization methods to examine brain-stomach interactions non-invasively in humans. We describe in detail the  gastric network, a set of brain regions phase coupled to the rhythm of the stomach during rest, and analyzed the spatial layout of the gastric network at the level of its constituent regions (Glasser et al., 2016), resting-state networks (Yeo et al., 2011), and its extent and position along the first two gradients of cortical connectivity that underlie the topological structure of the cortex (Margulies et al., 2016). We found that the gastric network is overrepresented in unimodal sensory-motor regions, and underrepresented in transmodal regions.

References

Glasser, M. F., Coalson, T. S., Robinson, E. C., Hacker, C. D., Harwell, J., Yacoub, E., Ugurbil, K., Andersson, J., Beckmann, C. F., Jenkinson, M., Smith, S. M., & Essen, D. C. V. (2016). A multi-modal parcellation of human cerebral cortex. Nature, 536(7615), 171. https://doi.org/10.1038/nature18933

Huntenburg, J. M., Bazin, P.-L., & Margulies, D. S. (2018). Large-Scale Gradients in Human Cortical Organization. Trends in Cognitive Sciences, 22(1), 21–31. https://doi.org/10.1016/j.tics.2017.11.002

Lachaux, J.-P., Rodriguez, E., Martinerie, J., Varela, F. J., & others. (1999). Measuring phase synchrony in brain signals. Human Brain Mapping, 8(4), 194–208.

Maris, E., & Oostenveld, R. (2007). Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods, 164(1), 177–190. https://doi.org/10.1016/j.jneumeth.2007.03.024

Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2010). FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data. Computational Intelligence and Neuroscience, 2011. https://doi.org/10.1155/2011/156869

Rebollo, I., Devauchelle, A.-D., Béranger, B., & Tallon-Baudry, C. (2018). Stomach-brain synchrony reveals a novel, delayed-connectivity resting-state network in humans. ELife, 7, e33321. https://doi.org/10.7554/eLife.33321

Rebollo, I., & Tallon-Baudry, C. (2022). The sensory and motor components of the cortical hierarchy are coupled to the rhythm of the stomach during rest. Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.1285-21.2021

Richter, C. G., Babo-Rebelo, M., Schwartz, D., & Tallon-Baudry, C. (2017). Phase-amplitude coupling at the organism level: The amplitude of spontaneous alpha rhythm fluctuations varies with the phase of the infra-slow gastric basal rhythm. NeuroImage, 146, 951–958. https://doi.org/10.1016/j.neuroimage.2016.08.043

Sanders, K. M., Ward, S. M., & Koh, S. D. (2014). Interstitial cells: Regulators of smooth muscle function. Physiological Reviews, 94(3), 859–907. https://doi.org/10.1152/physrev.00037.2013

Wolpert, N., Rebollo, I., & Tallon‐Baudry, C. (2020). Electrogastrography for psychophysiological research: Practical considerations, analysis pipeline, and normative data in a large sample. Psychophysiology, e13599. https://doi.org/10.1111/psyp.13599

Yeo, B. T. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead, M., Roffman, J. L., Smoller, J. W., Zöllei, L., Polimeni, J. R., Fischl, B., Liu, H., & Buckner, R. L. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology, 106(3), 1125–1165. https://doi.org/10.1152/jn.00338.2011