Author response: Spontaneous neuronal oscillations in the human insula are hierarchically organized traveling waves
Year: 2022
Type: peer-review
Abstract: Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract The insula plays a fundamental role in a wide range of adaptive human behaviors, but its electrophysiological dynamics are poorly understood. Here, we used human intracranial electroencephalographic recordings to investigate the electrophysiological properties and hierarchical organization of spontaneous neuronal oscillations within the insula. We analyzed the neuronal oscillations of the insula directly and found that rhythms in the theta and beta frequency oscillations are widespread and spontaneously present. These oscillations are largely organized along the anterior–posterior (AP) axis of the insula. Both the left and right insula showed anterior-to-posterior decreasing gradients for the power of oscillations in the beta frequency band. The left insula also showed a posterior-to-anterior decreasing frequency gradient and an anterior-to-posterior decreasing power gradient in the theta frequency band. In addition to measuring the power of these oscillations, we also examined the phase of these signals across simultaneous recording channels and found that the insula oscillations in the theta and beta bands are traveling waves. The strength of the traveling waves in each frequency was positively correlated with the amplitude of each oscillation. However, the theta and beta traveling waves were uncoupled to each other in terms of phase and amplitude, which suggested that insular traveling waves in the theta and beta bands operate independently. Our findings provide new insights into the spatiotemporal dynamics and hierarchical organization of neuronal oscillations within the insula, which, given its rich connectivity with widespread cortical regions, indicates that oscillations and traveling waves have an important role in intrainsular and interinsular communications. Editor's evaluation This manuscript is of interest to neuroscientists willing to deepen their knowledge related to the role of the insula and to any scientist interested in oscillatory activity and traveling waves in the brain. The substantial dataset and the novel methodological approach provide interesting insights on the functional organization of this brain region. https://doi.org/10.7554/eLife.76702.sa0 Decision letter Reviews on Sciety eLife's review process Introduction The insula plays a critical role in cognitive control (Menon and Uddin, 2010), awareness (Craig, 2009; Critchley et al., 2004), and adaptive human behaviors (Menon and Uddin, 2010; Singer et al., 2009). Dysfunction of the human insula is prominent in many psychiatric and neurological disorders (Jilka et al., 2014; King-Casas et al., 2008; Sha et al., 2019). Human functional magnetic resonance imaging (fMRI) studies have demonstrated insula involvement in detection and attentional capture of goal-relevant stimuli across a wide range of cognitive tasks (Cai et al., 2014; Dosenbach et al., 2008; Dosenbach et al., 2006; Menon and Uddin, 2010; Sridharan et al., 2008). Despite its critical role in cognitive function and dysfunction (Menon, 2011; Uddin, 2015), the electrophysiological foundations of the insula, its hierarchical organization, and sub-second spatiotemporal dynamics are poorly understood, in part due to the low temporal resolution of fMRI data. Neuronal oscillations are a fundamental feature of cortical networks and are critical for facilitating communication among brain areas (Berger, 1929). Understanding the fundamental features of insula oscillations is therefore important, given its rich connectivity with widespread cortical regions (Menon and Uddin, 2010). Here, we address these challenges and investigate, for the first time, the hierarchical organization and spatiotemporal gradients in the insula using a unique human intracranial electroencephalography (iEEG) dataset. Human fMRI studies have shown that the insula plays a foundational role in human cognitive processes by interacting with other large-scale brain networks such as the medial default mode network (DMN) and lateral frontal–parietal network (FPN), which are two other large-scale brain networks that play an important role in external stimulus-related cognition and monitoring internal mental processes (Menon, 2011; Power et al., 2011). Crucially, the intrinsic connectivity of the insula displays close correspondence with task-related coactivation patterns (Sridharan et al., 2008; Supekar and Menon, 2012), and this correspondence has allowed intrinsic and task-related connectivity associated with the insula to be demarcated and studied under a common, triple-network model framework (Menon, 2011). Cytoarchitectonics studies in non-human primates have shown functional subdivisions within the insula. These studies revealed that the non-human primate insula can be divided into an anterior-basal agranular part and a posterior granular part (Augustine, 1996; Mesulam and Mufson, 1985). These studies, as well as direct tract-tracing studies, have also shown that the anterior insula (AI) is structurally connected to orbitofrontal, entorhinal, amygdala, periamygdaloid, piriform, olfactory cortices, and the temporal pole whereas the posterior insula (PI) is connected to primary and secondary sensory and motor cortices (Fudge et al., 2005; Höistad and Barbas, 2008; Mesulam and Mufson, 1985; Stefanacci and Amaral, 2002). Similar to the findings in non-human primates, MRI studies in humans have also shown subdivisions in the insula (Ghaziri et al., 2017; Morel et al., 2013). Previous fMRI studies have shown functional subdivisions within the insula, as the AI was involved in task-related engagement and disengagement with the DMN and FPN across a wide range of cognitive and affective tasks (Cai et al., 2016; Cai et al., 2021; Chen et al., 2016; Lamm and Singer, 2010; Sridharan et al., 2008). The AI, along with the anterior cingulate cortex, is known as the salience network, which is activated simultaneously across most cognitive task domains, thus forming a core task-set system (Dosenbach et al., 2008; Dosenbach et al., 2006; Ouyang et al., 2008). Together, these findings suggest a common hierarchical organization within the insula across species. Neuronal oscillations are a fundamental feature of cortical networks and are critical for facilitating communication among a wide range of brain areas across species (Berger, 1929). Understanding the fundamental features of insula oscillations is therefore important, given its rich connectivity with widespread cortical regions (Menon and Uddin, 2010). However, the electrophysiological basis of the functional subdivisions in the insula and its spatiotemporal gradients and hierarchical organization of spontaneous neuronal oscillations are poorly understood. iEEG recordings with adequate sampling of electrodes from both the anterior and posterior subdivisions of the insula provide one way to address this challenge. We examined a resting-state iEEG dataset of 10 participants who underwent treatment for intractable epilepsy and had a total of 239 electrodes implanted across both the anterior and posterior subdivisions of the insula. The first goal of our study was to determine the hierarchical spatial organization of the electrophysiological patterns across the insula and, in particular, explaining how they relate to the frequency and power of neuronal oscillations. Global brain intrinsic gradients have been known to be closely associated with the structural connectivity of the brain areas and the degree of task-related activations (Huntenburg et al., 2018; Paquola et al., 2019; Zhang et al., 2019). Analyses of intrinsic oscillatory gradients in the insula are, therefore, crucial to understand the structure–function relationship of areas within the insula and between the insula and other regions of the cortex (Huntenburg et al., 2018; Vázquez-Rodríguez et al., 2019) and provide a more mechanistic understanding of oscillations as a means of intrainsular and interinsular communications in the human brain. Based on recent findings that showed a brain-wide posterior-to-anterior frequency gradient and an anterior-to-posterior power gradient using resting-state magnetoencephalography (MEG) recordings from humans (Mahjoory et al., 2020), as well as similar findings from direct recordings at the cortical surface (Zhang et al., 2018), we hypothesized that the insula would have a posterior-to-anterior frequency gradient and an anterior-to-posterior power gradient as well. Recent studies in humans suggested that the hierarchical organization of neuronal activity across the cortex is not a stationary phenomenon, rather that it has a dynamic phenomenon, which propagates across the cortex in space and time (Alamia and VanRullen, 2019; Halgren et al., 2019; Hangya et al., 2011; Kleen et al., 2021; Lozano-Soldevilla and VanRullen, 2019; Massimini et al., 2004; Muller et al., 2016; Stolk et al., 2019; Zhang and Jacobs, 2015; Zhang et al., 2018). This phenomenon, also known as ‘traveling waves,’ consists of oscillations that propagate progressively across the cortex. Traveling waves have been shown to be relevant for visual perception (Besserve et al., 2015; Davis et al., 2020; Gabriel and Eckhorn, 2003; Muller et al., 2014; Nauhaus et al., 2009; Townsend et al., 2015; Vinck et al., 2010; Zanos et al., 2015) and movement initiation (Balasubramanian et al., 2020; Denker et al., 2018; Rubino et al., 2006; Rule et al., 2018; Takahashi et al., 2015) in non-human primates and visual processing (Liang et al., 2021; Stroh et al., 2013; Xu et al., 2007) and spatial navigation (Agarwal et al., 2014; Patel et al., 2012; Patel et al., 2013) in rodents. In humans, traveling waves have been observed during sleep (Dickey et al., 2021; Hangya et al., 2011; Massimini et al., 2004; Muller et al., 2016) as well as during behavior (Alamia and VanRullen, 2019; Kleen et al., 2021; Stolk et al., 2019; Zhang and Jacobs, 2015; Zhang et al., 2018). However, it is unclear whether these traveling waves, which are mostly observed on the surface of the cortex, are also relevant for deep neocortical structures such as the insula. In light of this emerging literature, the second goal of our study was to detect traveling waves within the human insula. We adopted a novel method using circular statistics to identify traveling waves amidst the neuronal oscillations that we measured in each participant individually. In addition to detecting traveling waves, we also analyzed differences in the characteristics of the observed traveling waves across participants, and we measured traveling waves at a range of frequencies. Based on findings from human electrocorticography studies that showed low-frequency alpha and high-frequency beta traveling waves were weakly or uncorrelated with each other in the sensorimotor cortex (Stolk et al., 2019), we tested whether there was a link between the properties of traveling waves in the human insula between different frequencies. Our findings provide the first electrophysiological evidence for the presence of spontaneous oscillatory gradients and traveling waves in the human insula. Results Theta frequency and power gradients in the insula Our first goal was to investigate the spatial properties of the frequency and power gradients in the neuronal oscillations in the insula. These analyses were inspired by recent findings showing a posterior-to-anterior fast-to-slow frequency gradient and an anterior-to-posterior power gradient in oscillations measured from resting-state MEG recordings from humans (Mahjoory et al., 2020) and a hippocampal posterior-to-anterior frequency gradient using iEEG recordings in human spatial navigation (Goyal et al., 2020). We examined a dataset of subjects with iEEG recordings conducted during rest and used spectral analyses to measure the peak frequency and spectral power of the oscillations from each electrode in the insula (Materials and methods). Individual subjects in this dataset had multiple electrodes in their insula, including contacts in both hemispheres (Figure 1). To characterize the properties of the insula oscillations in these data, for each electrode we first detected oscillations by measuring the frequency bands where power exceeded one standard deviation above the 1/f spectrum using the multiple oscillations detection algorithm (MODAL) (Watrous et al., 2018). Theta (~6–9 Hz) and beta (~15–32 Hz) were the most prominent oscillations in the insula (Figure 1). However, there were no 9–15 Hz oscillations in the insula (Figure 1). In the left hemisphere,~90% of electrodes (n=107) showed a peak frequency in the theta range and 99% showed a peak in the beta frequency band. In the right hemisphere,~80% of electrodes (n=95) had theta-band peaks and ~81% had beta-band peaks. We next conducted tests of these theta and beta oscillations to evaluate their spatial and hierarchical organization. Figure 1 with 2 supplements see all Download asset Open asset Theta-band and beta-band frequency oscillations in the human insula. (a) Lateral and frontal views of the brain showing electrodes in the left insula (n=125) and right insula (n=114) across all participants. (b) An oscillation-detection algorithm known as multiple oscillations detection algorithm (MODAL) detected theta (~6–9 Hz) and beta (~15–32 Hz) oscillations across all electrodes in the insula (Materials and methods). The black line depicts the power spectrum, the red dotted line depicts the 1/f regression line, and the red shaded region around the spectrum represents the standard error of the mean power for each frequency. Gray shaded regions indicate the detected frequency bands at which the power spectrum was higher than the 1/f background signal. (c) The distribution of the peak frequencies detected across all insula electrodes. In the left insula, 90% of electrodes had theta oscillations and 99% had beta oscillations. In the right insula, 80% of electrodes had theta oscillations and 81% had beta oscillations. Peak frequencies of theta and beta oscillations appear to have relatively normal distributions across the insula. To localize the electrode positions within the insula while accounting for inter-subject anatomical variability (Kurth et al., 2010; Naidich et al., 2004), we used a novel approach of modeling the shape of each subject’s insula and then transforming each subject’s electrode position to a standard Montreal Neurological Institute (MNI) coordinate space using an iterative, semiautomated coregistration approach (Materials and methods). Using this pooled dataset, we then evaluated how the spectral dynamics (frequency and power) of the measured neuronal oscillations correlated with the coordinates of the electrodes within the insula (y-axis and z-axis) across all participants, using linear mixed-effect (LME) models (see Materials and methods). The y-coordinate of each electrode represents its position along the AP axis and the z-coordinate corresponds to the superior–inferior (SI) axis. We separately measured oscillations within the left and right hemispheres because interhemispheric differences in structure and function could affect oscillatory organization (Jakab et al., 2012; Sridharan et al., 2008). The AP axis coordinates (y-coordinates) of electrodes within the left insula predicted the peak frequency and power of the measured oscillations in the theta band (Figure 2). In the left insula, the frequency of theta was slower in the anterior insula (t[74] = –4.09, p<0.001, LME model) (Figure 2a); the amplitude of theta oscillations was higher in the left anterior insula (t[84] = 2.52, p<0.05) (Figure 2c). There was no indication of theta frequency (t[74] = –0.56, p>0.05) or theta power (t[74] = –0.19, p>0.05) gradients along the SI axis of the left insula. Figure 2 Download asset Open asset Frequency and power gradients of theta oscillations in the insula. (a) Peak frequencies within the theta band (~6–9 Hz) showed posterior-to-anterior gradient in the left insula. Coordinates of electrodes along the AP axis predicted peak theta frequency, where slower theta frequencies were more prominent in the anterior insula and faster theta frequencies were prominent posteriorly (t[74]=–4.09, p<0.001). (b) Theta frequencies in the right insula were not hierarchically organized along the AP axis (t[57]=–0.14, p>0.05), or the superior–inferior (SI) axis (t[57]=0.07, p>0.05). (c) Theta power showed anterior-to-posterior gradient in the left insula with theta power being higher in the anterior insula (t[74]=2.52, p<0.05). (d) Theta power showed a superior-to-inferior gradient in the right insula where higher amplitude theta oscillations were prominent in the superior portion of the insula (t[57]=2.92, p<0.01). A marginally significant AP gradient suggests that theta power was slightly more concentrated in the right anterior insula, similar to the power gradient in the left hemisphere (t[57]=1.98, p=0.053). No interaction effect was observed between AP and SI axes (t[57]=–1.14, p>0.05). Results were computed using linear regression where the y-axis (AP) and z-axis (SI) electrode coordinates were used as continuous independent variables to predict spectral power and frequency dynamics. Variability across participants has been accounted for in all t-statistics. Images were generated via automatic selection of insular vertices from anatomical parcellations using the Desikan–Killiany (DK) atlas. In the right insula, the frequencies of theta oscillations were not hierarchically organized along the AP axis (t[57]=–0.14, p>0.05) or the SI axis (t[57]=0.07, p>0.05) (Figure 2b). However, theta power trended toward an AP gradient, where theta power was greater in the right anterior insula (t[57]=1.98, p=0.053) (Figure 2d), consistent with the left hemisphere. Theta power also showed a significant SI gradient with higher amplitudes in the superior portion of the right insula (t[57]=2.92, p<0.01) (Figure 2d). Beta frequency and power gradients in the insula In addition to theta oscillations, we also observed widespread insula oscillations in the beta band. In the left insula, the power of beta oscillations showed an anterior-to-posterior gradient, with higher power anteriorly (t[84]=3.07, p<0.01) (Figure 3c). However, the frequency of beta rhythms was not hierarchically organized along the AP axis (t[84]=–0.95, p>0.05) or the SI axis (t[84]=–0.59, p>0.05) in the left insula (Figure 3a). In the right insula, beta power also showed a significant anterior-to-posterior gradient (t[58]=2.97, p<0.01) (Figure 3d), with higher power oscillations in the anterior insula. Thus, beta amplitude was greater in the anterior insula across both hemispheres. However, in the right hemisphere, the frequency of insula beta rhythms showed no gradients, in neither the AP (t[58]=–0.63, p>0.05) nor SI directions (t[58]=–1.10, p>0.05) (Figure 3b). Figure 3 Download asset Open asset Frequency and power gradients of beta oscillations in the insula. (a) Peak frequencies within the beta band (~15–32 Hz) showed no spatial organization in the left insula (t[84]=–0.95, p>0.05). (b) Beta frequencies in the right insula did not show AP (t[58]=–0.63, p>0.05), or SI gradients (t[58]=–1.10, p>0.05). (c) Electrode coordinates along the AP axis of the left insula predicted beta power, revealing that higher power oscillations were prominent in the anterior insula (t[84]=3.07, p<0.01). (d) Beta power in the right insula also showed an AP gradient where higher amplitude beta oscillations were prominent in the right anterior portion (t[58]=2.97, p<0.01). Results were computed using linear regression where the y-axis (AP) and z-axis (SI) electrode coordinates were used as continuous independent variables to predict spectral power and frequency dynamics. Variability across participants has been accounted for in all t-statistics. Images were generated via automatic selection of insular vertices from anatomical parcellations using the Desikan–Killiany (DK) atlas. These findings reveal that spontaneous theta and beta oscillations are spatially organized across the insula, especially along the AP axis. In both hemispheres, higher amplitude beta oscillations were concentrated anteriorly, whereas theta frequency and power showed opposing AP gradients in the left insula. More importantly, we used LME models to analyze these patterns, which included a categorical factor for subject identity and thus ensured that our results were not caused by inter-subject differences. Therefore, all reported t-statistics above specifically reflect the overall strength (effect size) and spatial direction of spectral gradients across electrodes, while accounting for differences between individual subjects. The number of electrodes contributed by each subject was also similar (mean=23.9 contacts, standard error=±2.43 contacts). Notably, the single-subject gradients also directionally align with the group level results (Figure 1—figure supplement 1). Oscillation clusters are traveling waves in the human insula Previous work on the computational modeling of weakly coupled oscillators has shown that traveling waves can naturally emerge from spatially varying gradients of oscillations across different frequencies (Ermentrout and Kopell, 1984). Therefore, given that we observed spatial gradients in the amplitude and frequencies of insula oscillations, we next examined the phase dynamics of these signals to test whether the frequency gradients contributed to traveling waves. A prerequisite for identifying a neural traveling wave is that a spatially contiguous region of cortex must show an oscillation at a single frequency. Thus, as a first step to identify traveling waves in the human insula, we sought to find oscillation clusters in each participant. We defined an oscillation cluster as a spatially contiguous group of electrodes having narrowband oscillations at a given frequency band (see Materials and methods). To identify electrode clusters, as in earlier work (Zhang et al., 2018), we first computed the mean power spectrum of each electrode and then removed the 1/f background signal to find narrowband oscillation peaks at a particular frequency. We then used a spatial clustering algorithm to identify all the groups of contiguous electrodes that showed narrowband oscillations at the same frequency. We defined these groups of electrodes, adjacent in both frequency and space, as an oscillation cluster. Figure 4 shows examples of oscillation clusters in the theta and beta frequency bands for the insula electrodes in participant #3. We found 28 oscillation clusters (15 in the left insula and 13 in the right insula) across 10 participants (see Figure 4—figure supplements 1–9 for oscillation clusters from other participants). Oscillation clusters were present at both theta and beta frequencies, with 14 oscillation clusters in the theta band (8 in the left insula and 6 in the right insula) and 14 oscillation clusters in the beta band (7 in the left insula and 7 in the right insula), in 9 out of 10 participants (participant #4 did not have any oscillation clusters). Overall, 79% of all electrodes (186 out of 239 electrodes) participated in at least one oscillation cluster. Figure 4 with 9 supplements see all Download asset Open asset Example of spatiotemporal clustering in the human insula. A cluster was identified as a spatially contiguous groups of electrodes that had narrowband oscillations at either theta or beta frequency band (Materials and methods). (a) Insula electrodes visualized in a brain surface plot in participant #3. (b) Insula electrodes projected on to a flat brain which was subsequently used for visualization of oscillation clusters and traveling waves. (c) Theta frequency oscillation cluster in participant #3 (electrodes with black circles denote electrodes without any narrowband peaks in the power spectrum). (d) Beta frequency oscillation cluster in the same participant. (e) Mean power spectrum (blue) in log–log coordinates from an example electrode in participant #3. Black line denotes fitted 1/f line to the mean power spectrum in log–log coordinates using robust linear regression (see Materials and methods). (f) Normalized power spectrum (red) after removal of the 1/f background signal. This approach emphasizes narrowband oscillations as positive deflections. Stars (cyan) denote the narrowband peaks in the normalized power spectrum, identified as any local maximum greater than one standard deviation above the mean (see Materials and methods). Our next goal was to test whether the signals on these oscillation clusters behaved as traveling waves (Materials and methods). We first informally inspected the signals from the oscillation clusters and found that the of their oscillations with electrode the presence of traveling waves. Figure shows an example of the of the across a of electrodes for both low-frequency theta-band and beta-band oscillation clusters in participant #3. this example shows that a single group of electrodes could traveling waves at multiple frequencies Figure Download asset Open asset of traveling waves in the human insula. We of electrodes in the oscillation clusters, and they with the of the electrodes then they a traveling We used a regression approach to detect patterns of traveling waves in addition to linear traveling waves, in an oscillation cluster. This regression approach the direction of traveling waves for each electrode (Materials and methods). (a) from electrodes as in (b) for participant #3. (c) signals in the theta frequency band. (d) signals in the beta frequency band. (e) and (f) are of signals in (c) and to show as red a spatially of the across the electrodes (electrodes for theta frequency and electrodes for beta of regression approach to detect traveling waves. coordinates of electrodes were to coordinates using 1 and 2 denote the first and second and of the electrodes were fitted with these coordinates in the regression circles denote the of the electrodes with the the between the and the predicted (Materials and methods). plot showing direction of the theta traveling wave to right plot showing direction of the beta traveling wave to the same as in the left 1 shows of the directions across time for theta traveling showing of directions of the theta traveling waves to and 1 shows of the directions across time for the beta traveling showing of directions of beta traveling waves to and We used circular statistics to identify traveling waves at each and electrode in a given oscillation cluster by measuring the phase of the oscillations on each electrode and for between electrodes. To this for each electrode in a given oscillation we first identified a of electrodes, as within a of for each we used a regression to characterize the spatial between the of the oscillations and the positions of the electrodes (Figure Thus, this models each phase distribution as a by the spatial gradient that the to the of across the electrodes. The fitted phase gradient a of the and direction of traveling wave at each (Figure this we found that the features of traveling waves across time We computed the of the traveling waves as the mean of phase by the models across the cluster (Materials and methods). Figure 6 Download asset Open asset of traveling waves in the human insula. Traveling waves their directions with to Insula electrodes visualized in a brain surface plot in (a) participant and (b) participant (c) and (d) are of the directions across time for theta Hz) and beta Hz) traveling waves in participant (e) and (f) are the same as in (c) and for participant show
Source: eLife
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $2,500
OpenAlex ID: W4282923337