Cardiac and Gastric Interoceptive Awareness Have Distinct Neural Substrates

Abstract Interoceptive awareness, an awareness of the internal body state, guides adaptive behavior by providing ongoing information on body signals, such as heart rate and energy status. However, it is still unclear how interoceptive awareness of different body organs are represented in the human brain. Hence, we directly compared the neural activations accompanying attention to cardiac (related to heartbeat) and gastric (related to stomach) sensations, which generate cardiac and gastric interoceptive awareness, in the same population (healthy humans, N = 31). Participants were asked to direct their attention toward heart and stomach sensations and become aware of them in a magnetic resonance imaging (MRI) scanner. The results indicated that the neural activations underlying gastric attention encompassed larger brain regions, including the occipitotemporal visual cortices, bilateral primary motor cortices, primary somatosensory cortex, left orbitofrontal cortex, and hippocampal regions. Cardiac attention, however, selectively activated the right anterior insula extending to the frontal operculum compared with gastric attention. Moreover, our detailed analyses focusing on the insula, the most relevant region for interoceptive awareness, revealed that the left dorsal middle insula encoded cardiac and gastric attention via different activation patterns, but the posterior insula did not. Our results demonstrate that cardiac and gastric attention evoke different brain activation patterns; in particular, the selective activation may reflect differences in the functional roles of cardiac and gastric interoceptive awareness.


Introduction
Subjective experiences of internal bodily states are referred to as interoceptive awareness (Khalsa et al., 2018), and such senses indicate ongoing physiological changes and guide adaptive behavior (Paulus et al., 2019;Quigley et al., 2021). Although interoceptive awareness serves multiple functions, depending on its source, e.g., thirst, hunger, heartbeat perception, it remains unclear how subjective differences in interoceptive awareness are represented in the human brain (Azzalini et al., 2019). We used functional magnetic resonance imaging (fMRI) to focus on the brain activations accompanying attention to cardiac (related to the heartbeat) and gastric (related to the stomach) sensations that elicit interoceptive awareness of different body organs, which have been suggested to serve different functional roles.
Considering the types of situations where people become aware of their heartbeat, the heartbeat signals and cardiac interoceptive awareness inform changes in physiological arousal (Paulus et al., 2019). In experimental settings, accelerated false cardiac feedback, i.e., exaggerated cardiac interoceptive awareness, has been found to alter the emotional salience of neutral faces (Gray et al., 2007) and perceived physical effort (Iodice et al., 2019) by enhancing physiological arousal. Similarly, gastric interoceptive awareness, such as fullness or hunger, modulates foraging and feeding behavior. In fact, people with eating disorders show altered gastric function and interoceptive awareness (Walsh et al., 2003;van Dyck et al., 2021), which may cause their abnormal eating behavior. Despite considerable differences in their functional roles, extant research has failed to establish whether cardiac and gastric interoceptive awareness are differently represented in the brain.
The invasive procedures needed to manipulate bodily signals have limited detailed functional brain mapping of interoceptive awareness. The brain regions responding to changes in visceral signals and interoceptive awareness have been studied using physical stimulation of viscera (Lu et al., 2004;Jarrahi et al., 2015) or pharmacological disturbance (Hassanpour et al., 2016(Hassanpour et al., , 2018 during fMRI scanning. However, such paradigms are difficult to combine in a single study because of invasiveness and the requirement of specific instruments, e.g., a barostat or an intravesical infuser. In addition, these brain activations would reflect the participant's discomfort and anxiety, induced by the abnormal sensations, and thus may not be appropriate for mapping interoceptive awareness of different organs. Another way to map interoceptive awareness is to use an interoceptive attention paradigm, asking participants to direct their attention toward sensations originating from the body. Importantly, although exteroceptive information, such as cutaneous sensations, may contribute to interoceptive awareness (Khalsa et al., 2009), interoceptive attention has been found to elicit interoceptive awareness and robust activations of brain regions, including the insula, middle cingulate cortex, and supplementary motor area (Caseras et al., 2013;Tan et al., 2018;Haruki and Ogawa, 2021). In particular, researchers have consistently associated activation of the right anterior insula with interoceptive attention (Zaki et al., 2012;Simmons et al., 2013) and objective accuracy of heartbeat perception (Critchley et al., 2004;Pollatos et al., 2007;Caseras et al., 2013). These results appear to support an influential neuroanatomical model of interoceptive awareness, where the posterior insula receives the initial cortical input of visceral signals while awareness of internal bodily states is represented in the right anterior insula (Craig, 2009(Craig, , 2011Evrard, 2019).
Previous studies using the interoceptive attention paradigm suffer from the limitation of having extensively used heartbeat attention, eliciting cardiac interoceptive awareness, over other bodily sensations. Even when attentional focus on gastric sensations was deployed, direct comparison within the modality of interoceptive attention has not been discussed (Simmons et al., 2013;Kerr et al., 2016;DeVille et al., 2020). Therefore, it is still unclear whether the neural encodings of interoceptive awareness of different bodily signals differ, limiting theoretical advances in this field. To address this issue, we directly compared the neural activations for cardiac and gastric interoceptive awareness using the interoceptive attention paradigm in a healthy population. We hypothesized that attention to cardiac and gastric sensations, and thus their interoceptive awareness, would elicit distinct neural activations, which would be modulated according to their functional roles. That is, cardiac attention might activate the regions underlying physiological arousal, while brain areas that modulate feeding and foraging behavior might show enhanced activation in gastric attention. Moreover, we considered that the insula, the most relevant region for interoceptive awareness (Craig, 2009;Critchley and Harrison, 2013), would show a subregionspecific representation of cardiac and gastric interoceptive awareness. To test this idea, we performed a region of interest (ROI) analysis, focusing on the insula, by combining multivoxel pattern analysis (MVPA) with a basic comparison of neural activation.

Participants
A total of 35 right-handed people participated in the experiment (15 women). One participant (a woman) requested an interruption of the experiment, yielding incomplete data. Of the remaining participants, the data of three (one man and two women) were excluded from the analysis because their maximum head movement was .3 mm during the experiment. Thus, the final analysis included 31 participants (12 women) who were 21.61 6 2.45 years of age on an average (range: 20-31). Their handedness was assessed by a modified version of the Edinburgh Handedness Inventory for Japanese participants (Hatta and Nakatsuka, 1975). The sample size deemed sufficient for determining brain activation was based on previous studies that used a similar task for interoceptive attention (Wiebking et al., 2015;Haruki and Ogawa, 2021). Written informed consent was obtained from all participants. This study was conducted in accordance with the Declaration of Helsinki and all its amendments. The Ethics Committee of Hokkaido University approved the experimental protocol.

Task procedures
To quantify brain activation for cardiac and gastric interoceptive awareness noninvasively, we asked participants to perform an established interoceptive attention task in an MRI scanner. We used a task procedure that elicited robust brain activation for interoceptive awareness (Simmons et al., 2013;Kerr et al., 2016;DeVille et al., 2020), which was modified to suit our purpose, i.e., to perform MVPA, and to increase statistical power. The present task had a simple block design that had three conditions with resting periods: heart attention (for cardiac interoceptive awareness), stomach attention (for gastric interoceptive awareness), and visual attention (for control). Participants were asked to focus on sensations from their heartbeat or stomach, or visual stimuli, and to be aware of the slight sensations for each condition (Fig. 1). In the heart and stomach attention trials, the words "HEART" and "STOMACH" were presented for 10 s each on the screen to allow participants to realize which sensation they were focusing on. In the visual attention trial, the word "TARGET" was presented on the screen for 10 s, with the color of the word gradually and slightly fading from black to gray. The color changed every 1.5 s for a total of five times. Therefore, in each task trial, participants directed their attention to a particular, vague sensation without a salient stimulus. During the rest period, participants watched a fixation crossbar for 12 s with their eyes open. Preparing for the next task trial was prohibited. Each condition of the task trial was presented five times per run: a single run included a total of 15 task trials and 15 rest periods. The order of the task trials was pseudorandomized, but the same trial was not presented three times in succession. At the end of each run, participants rated the subjective intensity of the sensations (heart, stomach, and visual) throughout the run using a Likert scale ranging from 1 (not intense at all) to 7 (extremely intense). An ;5.5-min run was repeated five times; thus, we obtained 25 data points (125 volumes) for each condition per participant. This procedure was designed to include a larger number of trials than those of previous studies (Kerr et al., 2016;DeVille et al., 2020) because we needed a sufficient number of task trials to perform MVPA. Before the first run began, all participants underwent a resting-state scan that lasted for 5 min, which was not analyzed in the present study. Stimuli were presented on a liquid crystal display and projected onto a custom-made viewing screen. Participants took a supine position in the scanner and viewed the screen via a mirror. Participants experienced a practice trial and filled its intensity report (for each condition), to learn how to perform the task and report the subjective intensity, before entering the MRI scanner. Through the practice trials, we verbally confirmed that the participants successfully directed their attention to each sensation depending on the condition.

Preprocessing of fMRI data
All image preprocessing was performed using SPM12 software (Wellcome Department of Cognitive Neurology; http://www.fil.ion.ucl.ac.uk/spm). All the functional images were initially realigned to adjust for motion-related artifacts. Volume-based realignment was performed by co-registering images using rigid-body transformation to minimize the squared differences between volumes. The realigned images were then spatially normalized with the Montreal Neurologic Institute template based on the affine and nonlinear registration of co-registered T1-weighted anatomic images. They were resampled into 3 mm-cube voxels with sinc interpolation. The images were spatially Figure 1. Schematic presentation of the task procedure. An example of the time course of the functional magnetic resonance imaging (fMRI) task is depicted. Our block design fMRI task included three types of task trials (heart attention, stomach attention, and control visual attention) that lasted 10 s each, followed by rest (12 s). Participants were asked to focus on sensations arising from the heart, stomach, and color changes in the word, according to the type of task. At the end of each run, participants rated the subjective intensity of the sensations for each condition using a Likert scale ranging from 1 (not intense at all) to 7 (extremely intense). smoothed using a Gaussian kernel of 6 Â 6Â6 mm full width at half-maximum. The images used for MVPA were not smoothed to avoid blurring the information in the multivoxel activity pattern.

Statistical analysis Subjective intensity of the sensations
We used R (version 4.0.3) for all our statistical inferences, except with functional imaging data. First, we tested whether the subjective intensity of the sensations differed between the modalities (cardiac, gastric, and visual). A linear mixed-effects (LME) model analysis, implemented in the lme4 package (Bates et al., 2015), was performed. With all the data obtained (N = 465, 3 trial types for 5 runs for 31 participants) as the dependent variable, we modeled the type of sensation as the fixed effect. Random intercepts and slopes for the effects of participants, and random intercepts for the runs, were modeled as random effects, ensuring the maximal random structure for our models (Barr et al., 2013). The LME allowed us to avoid averaging the values across five runs compared with the traditional ANOVA (Baayen et al., 2008). The parameters were estimated using the restricted maximum likelihood method, with the degrees of freedom estimated using the Satterthwaite method.

Whole-brain activation
We first evaluated the brain regions activated under each condition (heart attention, stomach attention, and visual attention) using a generalized linear model (GLM). An individual-level GLM included three regressors of interest for each condition as a separate box-car function that was convolved with the canonical hemodynamic response function. The rest period was used as a baseline. To reduce motion-related artifacts, six motion parameters were included as nuisance regressors. By combining the three conditions, we obtained six contrast images (heart attention compared with stomach attention, heart compared with visual, stomach compared with visual, and their opposite contrasts) for each participant. We then performed a group-level random effects analysis for these images using a one-sample t test. Through these statistical inferences, we directly compared the activation for each condition across the whole brain.
Moreover, we performed a group-level analysis of covariance (ANCOVA) to exclude the effect of the subjective ratings of stimulus intensity on brain activation. This was because we considered that the differences in subjective intensity could affect the brain activation pattern, independently of the object of attentional focus. Using the image of interoceptive attention, contrasted to the visual control, we modeled individual subjective ratings for each condition, averaged for five runs, as covariates. Then ANCOVA, excluding the effects of subjective rating as nuisance, was performed. By doing so, we assessed the brain activations of cardiac attention, contrasted with visual attention, without the effect of individual differences in subjective heartbeat intensity; and gastric attention, contrasted with visual, without the subjective intensity of stomach sensation. Furthermore, we explored variations in brain activation as a function of subjective intensity; a regression analysis with subjective intensity as a covariate of interest was performed using the same model used in the ANCOVA. For all the analyses, the voxel-level threshold was set to p , 0.001 (uncorrected), and the cluster-level threshold was set to p , 0.05, corrected for family-wise error (FWE).

ROI analysis of the subregions of the insula
We tested whether cardiac and gastric interoceptive awareness at the voxel level were differently represented in the subdivisions of the insula. In particular, we first performed MVPA, which allowed us to investigate more sophisticated neural representation than did conventional analysis (Norman et al., 2006), suspecting that conventional analysis might fail to detect activation differences in the insula. The ROIs were defined as the anatomic subdivisions of the insula in the Hammersmith brain atlases (Brain Development; www.brain-development.org). These images were constructed as a 3D probabilistic atlas using in vivo T1 MR images, including anatomic structures commonly seen in the human insula: the anterior short gyrus (ASG; the most dorsal anterior portion of the insula), middle short gyrus (MSG; the dorsal mid-anterior), posterior short gyrus (PSG; the dorsal mid-posterior), anterior inferior cortex (AIC; the ventral anterior), anterior long gyrus (ALG; the dorsal posterior), and posterior long gyrus (PLG; the ventral posterior; Faillenot et al., 2017). Importantly, these ROIs have been confirmed to show a subregion-specific activation pattern for heart attention (and, therefore, cardiac interoceptive awareness; Haruki and Ogawa, 2021).
The MVPA for heart and stomach interoceptive attention was performed with a two-class classifier based on a linear support vector machine (SVM) implemented in LIBSVM (http://www.csie.ntu.edu.tw/;cjlin/libsvm/). We first created another individual-level GLM, apart from the whole-brain analyses that included 15 task trials as independent regressors per run with six motion parameters as nuisance regressors. By doing so, we obtained parameter estimates of all the voxels in each ROI, for a total of 75 trials per participant, labeling them as heart, stomach, or visual, depending on the condition of each trial. We then trained SVM to classify the identity of the brain activation pattern of heart and stomach interoceptive attention using these parameter estimates as inputs to the SVM. Individual-level classification accuracy was estimated with a 5-fold "leave-one-out" cross-validation to avoid overfitting. This procedure uses inputs in four runs as training data, and inputs in one remaining run as test data, which was repeated five times for all possible combinations. The averaged classification accuracy across five repetitions of the tests was computed for each ROI for each participant, independently. We used a default hyperparameter (a fixed regularization parameter C = 1). The parameter estimates of the trial were not used as inputs to the classifier. One-sample t tests were performed to test whether the activation patterns for heart and stomach interoceptive attention in each subregion were classifiable using SVM. That is, we tested whether the computed classification accuracy was higher than the chance level (50%) at the group level, separately for each ROI. Because we used 12 ROIs (six anatomic regions for both hemispheres), the calculated p values were corrected for Benjamini and Hochberg's false discovery rate (FDR; Benjamini and Hochberg, 1995).
We also compared the differences in the averaged activations in the ROIs among each condition because we suspected that the classification accuracy calculated by MVPA could merely reflect the mean signal change in the ROIs. First, the mean signal changes in each ROI were extracted for each participant, separately for the heart and stomach attention conditions. Then, we performed a repeated-measures ANOVA on these values with the conditions (heart and stomach) and anatomic locations (the ASG, MSG, PSG, AIC, ALG, and PLG) as within-factors, independently for each hemisphere. Multiple comparison correction for post hoc analyses was performed using Shaffer's modified sequentially rejective Bonferroni procedure implemented in R. The combined use of MVPA and direct comparisons of the averaged signal change allowed detailed investigation of how interoceptive awareness specific to the cardiac and gastric domains are represented in the human insula.
Furthermore, we tested whether the averaged activation for interoceptive and exteroceptive attention differed in the insula, using a similar procedure for cardiac and gastric interoceptive attention. We extracted mean signal changes for interoceptive attention, the averaged activation across heart and stomach attention, and exteroceptive attention (visual attention) in each ROI. Then repeatedmeasures ANOVA, with the condition (interoceptive and exteroceptive) and the anatomic location as within-factors, was performed separately for each hemisphere.

Data availability
The .nii format figures of brain activation that support our findings are available at https://neurovault.org/collections/ 13262/. The .mat files containing the group-level brain activation, the results of our ROI analyses, and the behavioral data are also available at https://osf.io/8pg37/. The computer codes are available from the corresponding author upon reasonable request.

Subjective intensity of the sensations
We performed an LME model analysis for the subjective ratings of the intensity of each sensation. The LME included the type of stimuli (heart, stomach, and visual) as the main factors; random intercept and slope for the effects on the participants and random slope for the sequence of runs, were modeled as random effects. The subjective intensity of visual stimuli was rated the highest (marginal mean contrasted to heart = 0.90, t (30.00) = 4.51, p , 0.001) and that of the stomach the lowest (marginal mean = À0.41, t (30.00) = À2.22, p = 0.03) among the three stimuli (Table 1).

Whole-brain activation for cardiac and gastric attention
We directly compared whole-brain activation across the three conditions (heart attention, stomach attention, and visual attention as the control). We found that stomach attention that elicited gastric interoceptive awareness activated larger brain areas, including the occipitotemporal visual cortices, bilateral primary motor, primary somatosensory, left orbitofrontal, and bilateral posterior hippocampus. In contrast, the right dorsal anterior insula extending to the frontal operculum only showed higher activation in heart attention, i.e., selectively for cardiac interoceptive awareness ( Fig. 2A; Table 2). In particular, the medial visual area that was enhanced during gastric interoceptive attention largely overlapped with the "gastric network," that showed a temporal coupling with the gastric basal rhythm (Fig. 2B; Rebollo and Tallon-Baudry, 2022). Moreover, by comparing the heart and stomach conditions to the visual control, we found that cardiac and gastric interoceptive attention activated similar brain regions, including the insula, frontal operculum, parietal operculum, middle cingulate, and supplementary motor area; the results were highly comparable to previous studies (Caseras et al., 2013;Tan et al., 2018;Haruki and Ogawa, 2021). A notable exception was that the hippocampus and medial visual areas were activated only in stomach attention (Fig. 3A). The visual control condition, contrasted with the heart and stomach conditions, activated brain regions critical for visual attention, such as the middle frontal gyrus, superior parietal lobule, posterior thalamus (geniculate nucleus), and lateral visual association area (Dosenbach et al., 2007;Mayer et al., 2007), supporting the validity of our experimental design (Fig.  3A). Furthermore, we confirmed that the brain activation elicited by cardiac and gastric interoceptive attention was almost unaffected by differences in subjective stimulus intensity by performing an ANCOVA excluding the effect of the subjective ratings of stimulus intensity. The results of the ANCOVA were fairly comparable to the brain activation obtained with cardiac and gastric interoceptive attention contrasted with that in visual control (Fig. 3B, 3C). There was no brain activation covarying with the subjective The linear mixed-effects model included the condition as the fixed effect. The random structure was maximal for appropriate inference: random intercept and slope for the effect of the participants, and random intercept for the effect of the run sequence, were modeled as random effects. SE, standard error for marginal means; CI, confidence interval; e.d.f., estimated degrees of freedom.
intensity rating even at a more moderate threshold (cluster-level p , 0.05, uncorrected). All results are reported with a voxel-level threshold of p , 0.001 (uncorrected) with cluster size correction for p values , 0.05 (FWE).

Multivoxel pattern classification for cardiac and gastric attention in the insula
We performed MVPA using an SVM classifier that distinguished the activation patterns in the insula in heart and stomach conditions. Activation patterns in the anatomic subregions of the insula (Faillenot et al., 2017;Fig. 4A) were used as inputs to the SVM. We found that classification accuracy in the left PSG, which corresponds to the dorsal middle insula, was significantly higher than the chance level (50%; mean classification accuracy = 56.32, t (30) = 3.39, d = 0.61, corrected p-value for FDR = 0.024), suggesting that the region had distinct representation for cardiac and gastric interoceptive awareness

Comparison of the signal change for cardiac and gastric attention in the insula
We directly compared the signal change in the insular subregions for cardiac and gastric interoceptive attention. Averaged parameter estimates were extracted from the insular ROIs independently for heart and stomach conditions, i.e., 12 values for each condition. We then conducted a repeated-measures ANOVA, with the anatomic location and condition as the within-factors, independently for each hemisphere. The results indicated that, in both hemispheres, the main effects of anatomic location and the interaction between location and condition were significant (left, F  Fig. 4C), replicating the results of a previous study (Haruki and Ogawa, 2021). Post hoc analysis of the left ROIs revealed that the ALG (corresponding to the dorsal posterior insula) showed a marginally significant activation that was higher for gastric attention than for cardiac attention (F (1,30) = 3.83, corrected p = 0.059). Post hoc analysis of the right ROIs also revealed that the right ASG and MSG exhibited higher activation for cardiac attention Figure 2. Whole-brain activation for cardiac and gastric attention. A, Significant activation in cardiac attention contrasted with gastric attention (warm color) and vice versa (cold color) are depicted with a height threshold of p , 0.001 (uncorrected) and an extent threshold of p , 0.05 (corrected for family-wise error, corresponding to .42 voxels). Sagittal views show that attention to cardiac and gastric sensations activated the brain region relevant to each interoceptive awareness. The sensorimotor regions (S1, primary sensory; M1, primary motor; V3, visual area 3; FG, fusiform gyrus), orbitofrontal cortex (OFC), and hippocampus (HP) showed enhanced activation in gastric attention, while the right anterior insula (AI) was activated in cardiac attention. B, The internally extended visual area that was activated in gastric attention largely overlapped with the brain regions that have been shown to couple with the gastric basal rhythm (Rebollo and Tallon-Baudry, 2022; https://neurovault. org/collections/9985/). L, left hemisphere; R, right hemisphere. All figures are shown in axial slices with z and sagittal slices with x denoting locations in the Montreal Neurologic Institute coordinates. than for gastric attention (F (1,30) = 5.93, corrected p = 0.021; F (1,30) = 6.57, corrected p = 0.016, respectively) while a marginally significant activation, higher for gastric attention, was found in the ALG (F (1,30) = 4.13, corrected p = 0.051).

Discussion
Our results empirically showed that cardiac and gastric interoceptive awareness had similar but distinct neural substrates. Direct comparisons of brain activations revealed that interoceptive attention to heart and stomach sensations activated similar brain regions, including the insula, frontal operculum, parietal operculum, middle cingulate, and supplementary motor area, in contrast to the control visual attention. However, compared with gastric attention, cardiac attention activated the right anterior insula extending to the frontal operculum more; gastric attention enhanced activations in the occipitotemporal visual cortices, bilateral primary motor, primary somatosensory, left orbitofrontal, and bilateral posterior hippocampus. Moreover, our MVPA-based ROI analyses revealed that the left middle-anterior insula had distinct neural representations for cardiac and gastric interoceptive attention.
Previously, heartbeat perception has been studied as a representative of interoceptive awareness in general, However, activations in the medial visual cortex and hippocampus were found only in gastric attention, compared with visual attention. The height threshold was set to p , 0.001 (uncorrected) with an extent threshold of p , 0.05 (corrected for family-wise error), corresponding to .52 voxels. B, The distribution and mean scores of the subjective reports of stimulus intensity are plotted separately for each condition. Participants reported the intensity at the end of the scanning runs for each condition, yielding 155 datasets (31 participants for 5 runs) for each condition. Our maximal linear mixed-effects model analysis revealed that the visual stimulus (slight change in word color) was rated the most intense (regression coefficient compared with heart = 0.90, t (30.00) = 4.51, p , 0.001) and the stomach sensation was rated the least intense (regression coefficient = À0.41, t (30.00) = À2.22, p = 0.034). Error bars show the 95% confidence intervals, while the half-violin plot represents the kernel density estimation. C, The brain activation underlying cardiac and gastric interoceptive awareness did not vary as a function of the subjective report of the stimulus intensity. We performed the analysis of covariance that excluded the effect of the subjective reports from brain activation, revealing almost the same activation patterns compared with the activations reported in A. The height threshold was set to p , 0.001 (uncorrected) with an extent threshold of p , 0.05 (corrected for family-wise error), corresponding to .52 voxels. R, right hemisphere. All figures are shown in axial slices with z denoting locations in the Montreal Neurologic Institute coordinates.
suggesting the right anterior insula as the most relevant region for the generation of subjective experiences of internal bodily states (Craig, 2009;Critchley and Garfinkel, 2017). We, however, found that cardiac attention activated the right anterior insula more than gastric attention did, suggesting that the right anterior insula may predominantly code cardiac interoceptive awareness, rather than interoceptive awareness arising from other sources, e.g., gastric. In line with this idea, the most consistent activation in the right insula has been found by a meta-analysis focusing on cardiac interoceptive attention (Schulz, 2016). Furthermore, previous studies have found a robust correlation between individual accuracy of heartbeat perception and heightened activation of the right anterior insula (Critchley et al., 2004;Pollatos et al., 2007;Caseras et al., 2013;Haruki and Ogawa, 2021), but this was not the case for awareness of breathing (Wang et al., 2019) or skin conductance (Baltazar et al., 2021). The activation dominance of the right anterior insula may be explained by the functional roles of cardiac interoceptive awareness: people can notice increased arousal with heartbeat perception (Paulus et al., 2019). The right insula appears essential to perceive arousal as resection of the region causally diminishes physiological and emotional arousal (Terasawa et al., 2021;Holtmann et al., 2022). Moreover, presenting accelerated cardiac feedback increases perceived physiological arousal (Story and Craske, 2008) and, more importantly, activates the right anterior insula (Gray et al., 2007;Kleint et al., 2015). These findings, in conjunction with the present results, imply that cardiac interoceptive awareness closely involves physiological arousal and thus activates the right anterior insula by merely focusing on cardiac sensation. To sum up, the relationship between the right anterior insula and interoceptive awareness appears stronger in cardiac interoceptive awareness than in other modalities; such dominance of the cardiac domain may reflect the functional role of cardiac interoceptive awareness, which informs physiological arousal.
We also found that the brain regions underlying gastric interoceptive awareness encompassed the visual cortex. The involvement of the visual cortices in attention to stomach sensations appears puzzling, but converging evidence suggests a strong connection between the visual cortex and stomach function. Recently, Rebollo and colleagues found temporally delayed connectivity between brain activity and the intrinsic electrical rhythm generated by the stomach, which included the occipitotemporal visual cortices in addition to the somatosensory and motor areas The left MSG exhibited significantly higher classification accuracy above the 50% chance level. A marginally significant classification accuracy, i.e., p , 0.1, was observed in the right ASG, left ASG, and left MSG. The point plot represents the mean classification accuracy with a 95% confidence interval, while the half-violin plot represents the kernel density estimation. C, The results of direct comparison of the signal strength between cardiac and gastric attention are plotted. Post hoc analysis of repeated-measures ANOVA revealed that the right ASG and MSG showed significantly higher activation in cardiac than gastric attention. There was a marginally significant effect that implies stronger activation in gastric than cardiac attention in the left ALG and right ALG. The point plots represent the mean signal change, with 95% confidence interval, while the half-violin plot represents the kernel density estimation (green for cardiac, orange for gastric attention). *p , 0.05; †p , 0.10 (corrected for false discovery rate). As Extended Data, we performed a similar analysis focusing on the subdivisions of the insula for averaged brain activation in interoceptive and exteroceptive attention (Extended Data Fig. 4-1). (Rebollo et al., 2018;Rebollo and Tallon-Baudry, 2022). Crucially, the visual areas activated in the current experiment largely overlapped with the clusters included in the gastric network. Other past research showed that electrical or vibration stimuli on the stomach evoked neural activation in the occipital area in rats (Cao et al., 2019), cats (Pigarev et al., 2013), and even humans (Mayeli et al., 2021). Considering all these reports, it can be suggested that gastric functions are strongly tied to the visual cortex; here, we demonstrated that interoceptive attention to the stomach, which elicits gastric interoceptive awareness, activates the visual cortex even without direct stimulation of the stomach.
The reason for the involvement of the visual area may be that gastric signaling, and its subjective awareness, are closely related to foraging and feeding behavior that requires the integration of visuospatial information with energy status and motor coordination (Kanoski and Grill, 2017;Rebollo et al., 2021). For example, hungry people would change their sensory sampling and behavior in the external world to maximize the chance of food consumption. The current results support this idea; we observed activation related to gastric attention in the left orbitofrontal cortex, bilateral hippocampus, primary motor cortex, and the visual areas, which are all associated with food intake. For example, the orbitofrontal cortex has been suggested to modulate eating behavior by encoding the nutritional and reward values for food or food cues (Seabrook and Borgland, 2020). A meta-analysis of fMRI studies, including viewing food pictures compared with nonfood pictures, indicated that the left orbitofrontal cortex was most consistently activated for visual food stimuli (van der Laan et al., 2011). Furthermore, the role of the hippocampus in controlling food intake and regulating energy status has received considerable attention in the past few years (Suarez et al., 2019;Quigley et al., 2021); recent evidence indicates that the human hippocampus encodes ongoing nutritional states in response to food cues (Jones et al., 2021). Based on these findings, we consider that neural responses to gastric interoceptive awareness could be encoded in relation to food intake, which was activated by focusing on stomach sensations without any stimulation or food-related cues in the present study.
The present ROI analyses revealed that cardiac and gastric interoceptive attention had distinct representations in the insula. Only the right ASG and MSG, corresponding to the dorsal anterior insula, showed higher activation in cardiac than gastric attention; in the right mid-posterior and left insula, there was no activation difference. Nevertheless, we identified the left PSG, corresponding to the dorsal middle insula, in different activation patterns for cardiac and gastric attention. These results elaborate on how interoceptive awareness is represented in the human insula; in particular, we suggest, for the first time, that the left middle insula codes viscera-specific interoceptive awareness. Interestingly, the posterior insula did not show separable activation for cardiac and gastric attention, nor higher activation than baseline. Researchers have considered the posterior insula as the "primary" interoceptive cortex as it receives an initial cortical input of visceral signals (Craig, 2002;Evrard, 2019). The current inactivity of the posterior insula may support its role in encoding ongoing changes in bodily signals (Craig et al., 2000;Meier et al., 2018), rather than a subjective awareness of bodily states. This is because, in the interoceptive attention task, participants must have interoceptive awareness without homeostatic perturbation. Together, our ROI analyses would support the gradual processing of bodily signals in the insula along the posterioranterior axis (Craig, 2009): the posterior codes physical changes in the signal while interoceptive awareness would be represented in the middle-anterior insula.
Unfortunately, we did not record any physiological data during the fMRI scanning, potentially limiting the current study. Simultaneous recording of electrocardiograms (ECGs) and electrogastrograms would address the interaction between brain activation and physiological changes evoked by interoceptive attention. However, we believe the lack of physiological recording does not challenge the validity of the present study because past research indicated that cardiac parameters (heart rate and ECG amplitude) did not differ by directing attention between interoceptive and exteroceptive signals (Petzschner et al., 2019). On a related note, previous studies suggested that external processes, such as cutaneous sensations, could affect interoceptive awareness (particularly in the cardiac domain; Khalsa et al., 2009). Although we asked participants to focus on their internal sensations, we could not completely rule out the possibility that participants felt the heartbeat sensation from their skins. However, in the first place, interoceptive awareness may not arise solely from a single visceral sensation, such as when heartbeat perception is felt with the contraction of blood vessels and beating of the heart or hunger is mediated by both the physical contents of the stomach and a chemical signal. If so, researchers may benefit from isolating the neural substrates of cardiac and gastric interoceptive awareness localized in the present study from visceral topography to reveal the process of (visceral) multisensory integration. Another possible limitation is that the difference in task difficulty in each condition could affect the results. Although we tried to make the difficulty comparable for each condition, the task condition had a significant effect on the subjective intensity of stimulus; the subjective intensity of the visual stimulus was rated the highest while that of the stomach sensation was the lowest. We therefore performed another analysis that excluded the effect of the subjective rating of stimulus intensity, finding that brain activation was almost unaffected by individual differences in the subjective ratings. The results support our original idea that the present brain activation was elicited by the subjects (the heartbeat and stomach sensations) of attentional focus not by the difference in task difficulty. Recording a trial-by-trial fluctuation of the subjective intensity of stimuli would strengthen the present findings that the subjective differences in interoceptive awareness are encoded differently in the brain.
All the participants of this study were healthy adults aged between 20 and 31 whose native language was Japanese. However, we consider the present results to be replicable in other countries in healthy young adults because the brain activations induced by cardiac attention were comparable to that of previous studies conducted in Europe and the US (Caseras et al., 2013;DeVille et al., 2020). The stimulus used in the present experiment was a single word in English; therefore, we expect the results to be generalizable to healthy young adults as long as they understand the meaning of the word. The participants performed the task while they were lying in a supine position in an MRI scanner. Thus, it could be argued that the differences in brain activation of cardiac and gastric attention may be specific to interoceptive awareness in a supine position. We have no reason to believe that the results depend on other characteristics of the participants, materials, or context.