If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Anorexia nervosa (AN) is characterized by low body weight, disturbed eating, body image disturbance, anxiety, and interoceptive dysfunction. However, the neural processes underlying these dysfunctions in AN are unclear. This investigation combined an interoceptive pharmacological probe, the peripheral beta-adrenergic agonist isoproterenol, with resting state functional magnetic resonance imaging (rsfMRI) to examine whether individuals with AN relative to healthy comparisons (HC) show dysregulated neural coupling in central autonomic network brain regions.
Method
23 weight-restored AN female and 23 age, and BMI-matched HCs completed rsfMRI scanning before and after receiving isoproterenol infusions. We examined whole-brain functional connectivity (FC) changes using central autonomic network seeds in the amygdala, anterior insular cortex, posterior cingulate cortex, and ventromedial prefrontal cortex after performing physiological noise correction procedures.
Results
Relative to HCs, adrenergic stimulation caused widespread FC reductions in the AN group between central autonomic network regions and motor, premotor, frontal, parietal, and visual brain regions. Across both groups, these FC changes were inversely associated with trait anxiety (STAI-trait), trait depression (PHQ-9), and negative body image perception (Body Shape Questionnaire) measures but not with changes in resting heart rate. These results were not accounted for by baseline group FC differences.
Conclusions
Weight-restored AN females show a widespread state-dependent disruption of signaling between central autonomic, frontoparietal, and sensorimotor brain networks that facilitate interoceptive representation and visceromotor regulation. Additionally, trait associations between central autonomic network regions and these other brain networks suggest that dysfunctional processing of interoceptive signaling may contribute to affective and body image disturbance in AN.
Anorexia nervosa (AN) is an eating disorder defined clinically by the subjective disturbance of one’s own body image, and preoccupation with reducing one’s own body size, despite a low objective body mass (
American Psychiatric Association (2022): Diagnostic and statistical manual of mental disorders (5th ed., text rev.). https://doi.org/10.1176/appi.books.9780890425787
). Beyond these consensus-based diagnostic features, it is characterized by the abnormal perception of homeostatic processes within the body including decreased hunger ratings or tolerance of sustained hunger sensations (
), the primary brain network implicated in the processing of peripheral organ system (i.e., interoceptive) signals.
Resting state functional magnetic resonance imaging (rsfMRI) studies have found that individuals with AN show evidence of disrupted functional connectivity (FC) in several brain networks, including the default mode and salience networks (
A systematic review of resting-state functional-MRI studies in anorexia nervosa: Evidence for functional connectivity impairment in cognitive control and visuospatial and body-signal integration.
A systematic review of resting-state functional-MRI studies in anorexia nervosa: Evidence for functional connectivity impairment in cognitive control and visuospatial and body-signal integration.
). From a neuroanatomical perspective the CAN overlaps with both the default mode (i.e. ventromedial prefrontal cortex (vmPFC) and posterior cingulate cortex (PCC)) and salience networks (i.e. amygdala and insula) (
). However, its primary function is the hierarchical interoceptive processing of neurovisceral signals and the associated visceromotor and regulatory activity (
). Accumulating task-based fMRI studies of AN also identify abnormal neural activity related to interoceptive processing. The insular and cingulate cortices have been particularly implicated in AN responses to, or anticipation of, aversive perturbations of interoceptive signals including nociceptive stimulation (
). Abnormal activation of these regions in AN has also been observed relative to healthy comparisons (HC) during an interoceptive attention task performed under physiological resting conditions (
). Collectively, these studies suggest that dysregulated neural monitoring of interoception in AN occurs across the continuum of homeostatic states.
From a dynamic perspective, the CAN is proposed to exert “tonic, reflex, and integrative control of autonomic function [that] depends on the physiologic and behavioral state of the person” (
). This implies the brain’s direct responsibility for monitoring, integrating, and ultimately, regulating interoceptive signaling. Such activities are likely facilitated by the coordinated synthesis of afferent autonomic information with not only homeostatic parameters, but also contextual information from somatic and behavioral/cognitive processes that are represented outside of CAN regions. Recently, FC within the CAN was found to be greater overall and more strongly correlated with heart rate in HCs compared to those with acute AN (
Cruz Fdl, Schumann A, Suttkus S, Helbing N, Bar KJ (2022): Dynamic Changes in the Central Autonomic Network of Patients with Anorexia Nervosa. preprint.
). There is also some evidence of abnormal interactions among psychological traits and various cognitive functions in AN, including deficits in cognitive control, habit learning, value processing (e.g., reward learning), and heightened negative affect (e.g., trait anxiety, fear learning/generalization) (
Friederich H-C, Herzog W (2010): Cognitive-Behavioral Flexibility in Anorexia Nervosa. Behavioral Neurobiology of Eating Disorders: Springer Berlin Heidelberg, pp 111-123.
). However, the links between these psychological traits and interoceptive processing are unclear.
We addressed this gap in the current study by examining changes in CAN dynamics in AN following the pharmacologic perturbation of cardiorespiratory interoception. Specifically, we used rsfMRI to compare network-level changes in the blood oxygen-level-dependent (BOLD) signal induced by intravenous infusions of isoproterenol, an adrenaline analog with a short half-life that acts preferentially on peripheral beta-adrenergic receptors (
). We have consistently found isoproterenol to elicit focal changes in insular cortex activity using multiple fMRI sequence modalities in healthy participants (
Association of Generalized Anxiety Disorder With Autonomic Hypersensitivity and Blunted Ventromedial Prefrontal Cortex Activity During Peripheral Adrenergic Stimulation.
). In behavioral studies, we have shown that AN individuals report heightened cardiorespiratory sensations during stimulation and during saline infusion (
). Thus, isoproterenol is a promising tool for illuminating the brain networks contributing to distorted interoceptive processing in AN.
We evaluated CAN function in weight-restored AN females relative to female HCs matched on age and body mass index via rsfMRI scanning both before and following peripheral cardiorespiratory stimulation with isoproterenol. We provided participants a meal immediately following the experimental paradigm to enhance the relevance of affective and interoceptive processing during the scanning period, as prior studies have demonstrated that meal anticipation represents a clinically relevant context in which individuals with AN exhibit heightened meal-related anxiety (
The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
). Based on the hypothesis of abnormal CAN functioning within AN, we predicted group differences in pre- to post-infusion changes in whole-brain, voxelwise FC across several a priori CAN seed regions of interest (ROI). These ROIs consisted of the left and right anterior insula, the left and right amygdala, and the bilateral ventromedial prefrontal cortex (vmPFC) and posterior cingulate cortex (PCC). We also examined the dimensional relationship that traits relevant to AN, anxiety, depression, and body shape perception may have on changes in CAN FC following peripheral adrenergic stimulation.
Methods
Participants
AN participants were required to meet lifetime DSM 5 criteria for a diagnosis of AN, and current diagnoses were verified using the clinician-administered Mini International Neuropsychiatric Interview (MINI version 6.0)(
The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
). However, all participants were required to be weight-restored as defined by a minimum body mass index (BMI) of 18.5 or greater for at least 1 month to control for the well-described state effects of acute starvation (
). Self-reported race/ethnicity were also recorded.
No participant had any lifelong history of neurological, diabetic, cardiovascular, or respiratory disorders, including asthma. Lifetime history of schizophrenia, bipolar disorder, or antisocial personality disorder were also exclusionary. However, to maintain a sample representative of community treatment samples, we allowed for comorbid mood and anxiety disorders except for panic disorder to reduce potential complications due to isoproterenol-induced panic anxiety. We also allowed participants taking selected psychotropic medications into the study provided the dose was stable during a 2-week period prior to scanning (see Supplemental Table S1). See supplement for pre-scan instructions regarding medications.
This study was approved by the Western IRB. All participants provided written informed consent and were compensated for their participation. It was part of a larger project investigating interoceptive dysfunction AN and generalized anxiety disorder (
Association of Generalized Anxiety Disorder With Autonomic Hypersensitivity and Blunted Ventromedial Prefrontal Cortex Activity During Peripheral Adrenergic Stimulation.
)(ClinicalTrials.gov Identifier: NCT02615119). The project included not only the isoproterenol infusion perturbation task but also an interoceptive attention task, both performed between collection of the two rsfMRI scans reported here. On arrival, participants consumed a 300 Calorie snack. Immediately after the scan session all participants consumed a 1000 Calorie meal after choosing from one of three isocaloric options prior to the scanning session (as in (
The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
Data were collected on a 3 Tesla General Electric (GE) MR750 MRI scanner with a GE 8-channel receive-only head coil. 32-channel EEG signals were recorded simultaneously but not analyzed as part of this study. First, high-resolution anatomical T1-weighted volume scans (MP-RAGE) with SENSE were obtained in sagittal orientation (TR=1900 ms, TE=2.26 ms, matrix 256×256, flip angle=9°, 186 slices, and an isotropic resolution of 1×1×1 mm3). Next, T2*-weighted images were obtained using a gradient-echo EPI sequence with Sensitivity Encoding (TR=2000 ms, TE=27 ms, acquisition matrix=128×128 mm2, flip angle=78°, 39 interleaved transverse slices of 2.9 mm thickness, and an in-plane resolution of 1.9×1.9 mm2). A total of 240 whole-brain volume sets were acquired in each 8-minute resting state session.
Experimental design
Two eight-minute eyes open resting state fMRI scans were collected, one before and the other following the completion an interoceptive perturbation and then an interoceptive attention task. During the interoceptive perturbation task, participants received six intravenous bolus infusions in a double-blinded, randomized manner: isoproterenol (0.5 micrograms (μg) and 2 μg) and saline, each administered twice. Isoproterenol was obtained from Valeant Pharmaceuticals. The isoproterenol task duration was 30 minutes. During the interoceptive attention task (adapted from (
)) participants focused their attention on cardiac, respiratory, or stomach sensations, or on a visual target. No infusions were delivered during this 18-minute task. Thus, the resting state scans were collected approximately 50 minutes apart. During the resting state scans, cardiac and respiratory signals were recorded at 40 Hertz via MRI-scanner-equipped pulse oximeter attached to the nondominant index finger and a respiratory transducer belt placed around the torso. To assess key illness-related traits during the baseline visit, we administered several self-report scales including the Patient Health Questionnaire (PHQ-9; (
). Vital signs (heart rate, systolic and diastolic blood pressure) were measured during the baseline visit and after meal completion at the end of the fMRI scanning visit.
). After discarding the first five images to ensure a steady-state tissue magnetization, we identified volumes with excess motion (Euclidean norm of the motion derivatives>0.3 mm or fraction of voxel outliers>10%) and removed spikes in the signal intensity time courses. Slow BOLD fluctuations and time-locked cardiac and respiratory artifacts were removed using respiration volume per time (RVT) regressors (
). The RVT and RETRICOR regressors consisted of the RVT function and four delayed terms at 5, 10, 15, and 20 seconds and eight RETROICOR regressors generated on a slice-wise basis by AFNI's “RetroTS.m” script (
). See Supplement for further preprocessing steps.
Regions-of-interest and FC analysis
To investigate functional abnormalities of autonomic brain regions, we defined six CAN regions of interest (ROI) implicated in cardiovascular regulation (
) to increase anatomical accuracy given the variable morphology and size of these structures. To define the anterior insula, we divided the whole insula ROI into two subregions using a perpendicular plane that cuts the posterior-anterior axis at y=6 and used the anterior section (
Figure 1Location of central autonomic network (CAN) seed regions of interest: ventromedial prefrontal cortex (orange), posterior cingulate cortex (green), amygdala (left and right, red), and anterior insular cortex (left and right, blue).
We obtained resting-state FC maps for each of the pre-infusion and post-infusion rsfMRI scans by extracting the time series from each seed region and correlated them against all brain voxels. The resulting correlation coefficients were then converted to Fisher z-statistics to produce a more normally distributed variable (
). Next, we tested for group differences in the effects of isoproterenol on FC in healthy and AN individuals over time. All resulting clusters of this group x time interaction analysis were further corrected for multiple comparisons. For this, we used the 3dClustSim function in AFNI to run 10,000 Monte Carlo simulations of whole‐brain fMRI data and determine the cluster size at which the false positive probability was below α<0.05 with an uncorrected voxelwise threshold of p<0.001. For these thresholds, the simulation yielded a cluster size threshold of 130 contiguous voxels. Finally, we conducted an exploratory correlational analysis evaluating associations between FC changes and trait measures of anxiety (STAI-trait), depression (PHQ-9) and body image (BSQ), using the Bonferroni method for multiple comparison correction. This dimensional analysis was conducted across both groups as these traits are expected to vary across the spectrum of healthy and clinical conditions.
Results
Of the 70 individuals allocated to the intervention, 58 completed it with good quality data for both pre and post-isoproterenol rsfMRI scans (Figure S1). Six HCs were excluded for not completing all self-report measures. Another 6 HCs were removed by a propensity matching procedure (MatchIt in R (
)), to obtain an equal analysis sample of 23 ANs and 23 HCs matched on age and body mass index (Table 1). The groups did not differ on blood pressure (BP) at baseline or resting heart rate (HR) during the pre-isoproterenol scan. However, the HC group showed evidence of increased systolic and diastolic BP relative to the AN group, whose values were unchanged. The AN group also showed higher HR during the baseline visit, the post-isoproterenol scan, and the end of the fMRI visit (Table 1).
Table 1Descriptive and Inferential Statistics for Demographic and Diagnostic Variables
HC
AN
t-Stat (df)
p-Value
Mean (SD)
Mean (SD)
Gender
2F
23F
Education (years)
14.3(1.7)a
13.7(1.3)
t(38.93)=-1.21
p=0.234
Age (years)
22.3(3.9)
21.2(3.7)
t(43.90)=-0.97
p=0.338
BMI
22.9(2.8)
22.4(3.2)
t(43.23)=-0.61
p=0.544
Anxiety (STAI-Trait)
28.6(7.4)
53.1(10.1)
t(40.24)=9.42
p<0.001
Depression (PHQ-9)
0.8(1.2)
7.6(5.4)
t(24.12)=5.82
p<0.001
Body image (BSQ)
46.7(10.9)
127.0(36.6)
t(25.89)=10.07
p<0.001
Physiological measures - Baseline
Systolic BP (mmHg)
113.3(10.0)
110.2(10.3)
t(43.98)=-1.03
p=0.307
Diastolic BP (mmHg)
72.2(9.1)
72.2(9.0)
t(43.99)=-0.02
p=0.987
HR (bpm)
69.2(9.1)
76.9(12.5)
t(40.20)=2.40
p=0.021
Physiological measures - During fMRI
HR - pre ISO (bpm)
65.5(11.4)
69.8(7.4)
t(37.67)=1.53
p=0.135
HR - post ISO (bpm)
65.6(8.4)
71.3(9.4)
t(43.5)=2.34
p=0.034
Physiological measures - Post fMRI
Systolic BP (mmHg)
118.3(14.4)
110.9(9.2)
t(37.27)=-2.08
p=0.045
Diastolic BP (mmHg)
78.0(8.2)
73.0(7.3)
t(43.30)=-2.19
p=0.034
HR (bpm)
75.8(10.7)
83.1(10.6)
t(43.99)=2.34
p=0.024
a One healthy comparison participant did not provide their years of education.
Abbreviations: AN, anorexia nervosa; BMI, Body Mass Index; BP, blood pressure; bpm, beats per minute; BSQ, Body Shape Questionnaire; ISO; isoproterenol; PHQ-9, Patient Health Questionnaire; F, Female; HC, healthy comparison; M, Male; STAI, State Trait Anxiety Index.
Across all group x time interactions for each seed, the analysis of rsfMRI data revealed FC increases for the HC group and decreases for the AN group for a total of 19 significant multiple comparison corrected seed-to-target associations (Table 2). Specifically, FC change within CAN regions was observed between the vmPFC and both anterior insula regions, and the left anterior insula was coupled with the PCC. The right amygdala showed no significant FC change. Also prominently correlated with the CAN seeds were the sensorimotor and lateral prefrontal cortices. Figure 2 displays group x time (pre-post change) FC interaction maps for the left anterior insula, left amygdala, vmPFC, and PCC ROIs (see Supplement for group and condition effects).
Table 2Brain regions showing Group x Time FC interaction.
Seed and Target Regions
L/R
BA
Voxel Count
Peak (MNI)
T-
Anxiety STAI-Trait
Body Image BSQ
Depression PHQ-9
X
Y
Z
stat
r
p
r
p
r
p
stat
stat
stat
stat
stat
stat
Seed: Right Anterior Insula
Target Medial PFC
L/R
10
410
2.6
46.8
8.4
-4.77
-0.63
p<0.001
-0.58
p<0.001
-0.53
p<0.001
Primary Motor
L
4
248
-13.1
-33.8
64.4
-5.61
-0.67
p<0.001
-0.58
p<0.001
-0.54
p<0.001
Seed: Left Anterior Insula
Target Medial PFC
L/R
10
440
-0.9
67.8
-3.9
-4.54
-0.56
p<0.001
-0.53
p<0.001
-0.48
p<0.001
Primary Sensory/Motor
L
1/4
293
-42.9
-29
66.1
-5.45
-0.55
p<0.001
-0.54
p<0.001
-0.39
p=0.007
PCC/Precuneus
L
23/30/31
260
-14.9
-56.5
20.6
-4.26
-0.54
p<0.001
-0.57
p<0.001
-0.41
p=0.004
Primary Motor
L
4
204
-14.9
-30.2
64.4
-5.15
-0.65
p<0.001
-0.54
p<0.001
-0.51
p<0.001
Lingual/Mid-Occipital Gyrus
R
18
160
27.1
-84.5
-3.9
-4.46
-0.60
p<0.001
-0.56
p<0.001
-0.43
p=0.003
Seed: Left Amygdala
Target Posterior Insula/STG
R
13/22
190
44.6
-14.5
6.6
-5.20
-0.62
p<0.001
-0.64
p<0.001
-0.52
p<0.001
Seed: Posterior Cingulate Cortex
Target Parahippocampal/PCC/Cuneus
L
17/18/27/30
462
-20.1
-75.8
8.4
-4.85
-0.53
p<0.001
-0.61
p<0.001
-0.44
p=0.002
MFG
L
10/46
199
-37.6
38
3.1
-4.36
-0.57
p<0.001
-0.57
p<0.001
-0.48
p<0.001
IFG
L
11/47
158
-23.6
34.5
-12.6
-4.42
-0.59
p<0.001
-0.60
p<0.001
-0.43
p=0.003
Parahippocampal
R
27/30
151
11.4
-42.5
1.4
-4.61
-0.56
p<0.001
-0.64
p<0.001
-0.43
p=0.003
Seed: Ventromedial Prefrontal Cortex
Target Superior Parietal/Supramarginal
L
7/40
603
-41.1
-40.8
45.1
-4.85
-0.53
p<0.001
-0.50
p<0.001
-0.36
p=0.015
IFG
R
44/45
228
53.4
6.5
24.1
-4.62
-0.55
p<0.001
-0.57
p<0.001
-0.51
p<0.001
OFC
L
11/47
187
-25.4
32.8
-10.9
-4.86
-0.65
p<0.001
-0.62
p<0.001
-0.47
p<0.001
IFG/Premotor
L
9/44
174
-53.4
1.2
22.4
-4.39
-0.51
p<0.001
-0.52
p<0.001
-0.33
p=0.024
OFC
R
11/47
158
27.1
32.8
-9.1
-4.70
-0.58
p<0.001
-0.60
p<0.001
-0.44
p=0.002
LPFC
L
45/46
158
-41.1
-34.5
10.1
-4.60
-0.57
p<0.001
-0.61
p<0.001
-0.21
p=0.003
Premotor
L
6
151
-27.1
-5.8
64.4
-4.36
-0.57
p<0.001
-0.53
p<0.001
-0.43
p=0.002
The criterion for detecting significant clusters was p<0.05 as estimated by AFNI’s 3dClustSim (cluster size > 130 voxels, voxel level thresholded at p<0.001). Significant correlations with the trait facet of the State Trait Anxiety Inventory (STAI), Body Shape Questionnaire (BSQ), and Depression Patient Health Questionnaire (PHQ-9) were Bonferroni-corrected for multiple comparisons. Abbreviations: dlPFC, dorsolateral prefrontal cortex; FC, functional connectivity; IFG, inferior frontal cortex; PCC, posterior cingulate cortex.
Figure 2Top: T-maps of group x time interaction for the left anterior insula (blue), left amygdala (red), vmPFC (yellow) and PCC (green). Bottom: Interaction plots for selected clusters with connectivity values before (T1) and after isoproterenol infusion (T2), specifically for (A) the left anterior insula to mPFC, (B) the left amygdala to right insula, (C) the vmPFC to left parietal cortex. Abbreviations: AN, anorexia nervosa; HC, healthy comparison; mPFC, medial prefrontal cortex; PCC, posterior cingulate cortex; vmPFC, ventromedial prefrontal cortex.
To evaluate associations between FC and traits relevant to AN, the change in FC for each of the 19 significant seed-to-target associations was then correlated with anxiety (indexed via STAI trait score), body shape perception (indexed via BSQ total score), and depression (indexed via PHQ-9 total score). We computed these associations across groups to explore whether they might suggest a transdiagnostic dimensional relationship. After applying Bonferroni correction for multiple comparisons (corresponding to a threshold of p<0.0026), all region pairs showed at least a moderate inverse correlation with the STAI and BSQ and similar but smaller magnitude correlation with the PHQ-9 (Table 2 and Figure 3). To ensure that the observed changes in FC following adrenergic stimulation were not simply reflecting pre-existing widespread group differences, we assessed for FC group differences in the baseline rsfMRI scans collected prior to the infusions. The only FC difference observed was a decrease for AN relative to HC between the PCC seed and the anterior PFC (MNI x=11, y=61, z=15) (Figure 4).
Figure 3Regional associations between changes in FC with trait measures of anxiety (STAI-Trait) and body image (BSQ) across AN and HC groups. For illustrative purposes, selected relationships are displayed: each scatterplot shown reflects the FC-trait association indicated by the black arrow in the adjacent glass brain. The black line in each plot indicates the correlation slope. Blue diamonds represent healthy individuals, and red circles represent AN individuals. All reported p-values are Bonferroni-corrected and two-tailed. Abbreviations: AIns, anterior insula; BSQ, Body Shape Questionnaire; IFG, inferior frontal gyrus; IP, inferior parietal; M1, primary motor cortex; PCC, posterior cingulate cortex; OFG, orbitofrontal gyrus; S1, primary somatosensory cortex; STAI: State Trait Anxiety Inventory; V1, primary visual cortex; vmPFC, ventromedial prefrontal cortex.
Figure 4FC differences between AN and HC for the PCC seed region at baseline (i.e., before receiving the isoproterenol infusion). A cluster of reduced functional connectivity was observed between the PCC seed and an anterior medial PFC cluster in AN. The cluster is significant at α < 0.05 after cluster-based correction (> 130 contiguous voxels as determined by AFNI' 3dClustSim) with voxel-wise p < 0.001. For visualization purposes, the violin plot indicates the FC values of each group in the cluster. Abbreviations: FC, functional connectivity.
Finally, to ensure that the observed changes in FC following adrenergic stimulation were not simply reflecting the observed post-scan group differences in BP and HR (Table 1), we evaluated and found no associations between CAN seed-to-target regions and the post-scan systolic and diastolic BP as well as HR (Supplement). We also found no significant association between changes in post-scan HR or BP and CAN seed regions for any cluster (Figure S3 and Supplement).
Discussion
In the current study, we observed strong evidence of changes in central autonomic network FC induced by a peripheral beta-adrenergic challenge in weight-restored individuals with AN relative to HCs. Specifically, FC was diminished between CAN seed regions and a wide range of brain areas in the AN group, as compared to HCs, who consistently showed opposing FC increases in those regions. The reductions in FC were observed both within the CAN and between the CAN and frontal, parietal, somatosensory, motor, and visual brain regions. Each correlated ROI pair also showed an inverse relationship with trait anxiety, trait depression, and a perceptual measure of body image disturbance, but showed no relationship with changes in resting heart rate or blood pressure, suggesting that interoceptive signaling between CAN regions and other brain networks may facilitate the representation and regulation of these affective and body image concerns.
The widespread FC group differences occurred between regions of the CAN and portions of networks that primarily represent default mode/self-processing, cognitive control, motor, somatosensory, and visual processing. While we saw only one group difference in resting state FC prior to adrenergic stimulation (reduced PCC-to-mPFC connectivity in AN), our observed pre-post infusion FC changes align with frontoparietal and CAN FC differences seen previously in the cross-sectional rsfMRI AN literature (
A systematic review of resting-state functional-MRI studies in anorexia nervosa: Evidence for functional connectivity impairment in cognitive control and visuospatial and body-signal integration.
). We observed several FC changes in the amygdala, but that was the only subcortical region for which we observed FC with another CAN region (i.e., the posterior insula), either as a seed or target. This is noteworthy as several CAN regions coordinate with subcortical regions in other circuits to integrate viscerosensory input and affective information to initiate visceromotor output (
LeDoux JE (1992): Emotion and the amygdala. The amygdala: Neurobiological aspects of emotion, memory, and mental dysfunction. New York, NY, US: Wiley-Liss, pp 339-351.
). Prior cross-sectional rsfMRI studies of AN have primarily observed altered FC within or between CAN regions and subcortical regions only in studies of underweight AN compared to HC (
The reduced FC observed in AN following a cardiorespiratory perturbation could be interpreted as a stable consequence of the interoceptive dysfunctions commonly ascribed to this disorder (
). However, we believe this to be unlikely for several reasons. First, the described interactions represent state-dependent changes occurring after resolution of an acute perturbation of interoception and were not reflected by substantial group differences in CAN FC at baseline. Second, associations between reduced resting state FC and abnormal interoception and self-representation have been reported previously, though primarily based on cross-sectional data. For example, acutely ill and recovered women with AN have shown decreased activity in the DMN (precuneus), as well as attenuated salience (ACC) and motor network activity relative to HCs (
). Although those decreases were attributed to state-dependent abnormalities of interoception or body image disturbance, such constructs were not manipulated acutely in a state-dependent manner. The current findings therefore represent the first demonstration of a state-dependent alteration of FC showing associations with anxiety, depression, and body image disturbance that is in keeping with, but builds upon, the cross-sectional evidence. Moreover, the lack of BP changes combined with HR increases for AN versus HC in the post-infusion setting may reflect additional evidence of reduced vascular reactivity in AN (
), even though we saw no meaningful associations between these physiological indices and the group differences in FC. Finally, based on the observed associations between FC and BSQ scores, abnormalities of body image processing may relate to abnormal interoceptive processing and body image perception in those with AN, and may be particularly influenced by maladaptive frontoparietal-CAN communication. This is an important observation, as interoceptive processing has been hypothesized to be an ingredient involved in the formation of body image perception (
). Body image disturbances in weight-restored AN extend beyond self-evaluation to body ownership on experimental tasks such as the rubber hand illusion (
Guterstam A, Björnsdotter M, Gentile G, Ehrsson H, H. (2015): Posterior Cingulate Cortex Integrates the Senses of Self-Location and Body Ownership. Current Biology. 25:1416-1425.
). The breadth of the FC changes and their association with body image and affective traits may thus reflect the broader role of interoception in multisensory integration (
). Our results may also support recent arguments on the neural basis of embodied self-processing networks, which posit that effective coordination of neural activity across self-processing, cognitive control, and salience networks (including the insula and amygdala regions of the CAN) is required for generation of an embodied self-representation (
). This raises the possibility that increased top-down regulation in AN may reflect efforts to self-regulate one’s physiological condition to limit exposure to environmental factors with aversive associations, including food, through disciplined behavior (i.e., food avoidance). We speculate that such a process would likely leverage coordinated communication across numerous networks such as the frontoparietal and CAN networks, but also motor and visual networks, as observed in the present study, and could reflect the identification of large-scale deficits in neuronal synchrony and brain connectivity (
) in AN. Pursuant to this view, interoceptively triggered changes in body state in AN individuals would have widespread and disparate effects on emergent properties (
) of central autonomic but also cognitive, executive, salience, visual, and even motor networks.
We observed inverse relationships between CAN FC and self-reported trait measures of body shape representation, depression, and anxiety across groups. These results are consistent with prior reports that activity in CAN regions, such as the amygdala (
), correlates with anxiety traits regardless of group status. Our findings cannot be attributed to effects of acute starvation in AN, given the recruitment of a weight-restored sample, and may thus reflect individually meaningful trait differences across the dimensions of affective and body image concerns. Further insight into the predisposing contributions of these signals is needed, particularly in premorbid samples at risk for development of an eating disorder. Children with avoidant restrictive food intake disorder also show increased risk for other eating disorder diagnoses in adulthood (
), and it remains unclear to what extent interoceptive dysfunction plays a role in the disorder. Our results add to the cross-sectional literature showing relatively decreased cortical FC in both underweight and weight-restored AN (
), highlighting the need for investigation into whether it may be a predictive factor for AN in those at risk for developing severe eating disorders. Further, our acute repeated-measures results suggest that an instability of brain-body interactions after adrenergic perturbation may be a part of the noradrenergic pathophysiology of AN (
). Potential translational and/or clinical implications of this work could include efforts to develop new laboratory tools or clinical tests capable of delineating interoceptive network dysfunction at the individual patient level. Alternatively, targeting abnormal CAN FC metrics might be incorporated into the development of new treatments focused on the principles of interoceptive exposure, particularly from the standpoint of gastrointestinal processing (
). Studies investigating the effects of enhancing inter-network communication between the CAN and other self-processing networks may also help to refine the search for new treatment targets capable of addressing anxious and maladaptive body image symptomatology, especially in underweight AN.
This study has several limitations. The low sample size is balanced by a repeated measures experimental medicine design which allowed us to examine the influence of an interoceptive perturbation on resting brain network activity, and by the paucity of FC studies of AN, most of which have used similar or smaller sample sizes. Another limitation is the subset of CAN regions used for seed-based correlation analysis. We focused on six classically defined cortico-limbic CAN regions linked to sympathovagal control (
) that could play a role in the regulation of cardiovascular oscillations during exertion of sympatho-vagal control. Inclusion of other seeds in future studies (e.g., posterior/mid insula) could help to identify additional networks not reported here. We included AN individuals with certain comorbid conditions who were taking selected psychiatric medication but did not think this had a meaningful impact on our findings for several reasons. First, they were stably medicated and screened for recent use of medications (e.g., sedatives) known to acutely modulate fMRI signals. Second, the repeated measures design allowed us to detecting rapid shifts in brain activity that would be unaffected by chronic medication use. Third, most AN individuals have comorbid psychiatric conditions (
), thus findings from our sample have ecological validity for patients receiving treatment in real-world settings and can be supplemented by future studies controlling more tightly for medications and comorbid conditions. Fourth, the strong correlations between trait measures of psychopathology and changes in FC across the whole sample could have been influenced by oversampling both ends of the transdiagnostic spectrum. Balanced sampling will be important for future studies, particularly those seeking to identify group differences in associations between these measures. However, our comparison between groups did not reveal compelling evidence suggestive of group differences in association (see figure S2), and we speculate that the observed relationships reflect neural trait associations along a linear spectrum that serve as a basis for further investigation. Finally, our study sample has two demographic limitations. We recruited an all-female AN sample, which we emphasized to address the far higher (10:1) disorder prevalence rate in females than males. The other is the lack of racial diversity in the AN sample. Since eating disorders are expressed across the spectrum of race and ethnicity (
), increasing the diversity of ethnic and racial sampling in future studies will be important in establishing the representativeness of the current findings.
In conclusion, we found that peripheral adrenergic stimulation induced widespread central autonomic FC reductions in weight-restored individuals with AN compared to HCs. Associations of CAN-FC changes in frontal, parietal, somatosensory, motor, and visual brain regions with trait measures of anxiety, depression, and body image disturbance provide evidence supporting the argument that interoceptive signaling between CAN regions and other brain networks may facilitate the representation and regulation of affective and body image concerns.
Acknowledgements
Megan Sinik, BS, Dhvanit Raval, BS, and Chloe Sigman, BS, provided physiological data processing assistance; and Megan Cole, RN, Jeanne Echols, RN, Lisa Augustine, RN, Susan Maxey, RN, and Lindsey Bailey, NT, provided isoproterenol preparation assistance. The authors would also like to acknowledge the key contributions of the late Jerzy Bodurka to the acquisition of the functional neuroimaging data.
American Psychiatric Association (2022): Diagnostic and statistical manual of mental disorders (5th ed., text rev.). https://doi.org/10.1176/appi.books.9780890425787
A systematic review of resting-state functional-MRI studies in anorexia nervosa: Evidence for functional connectivity impairment in cognitive control and visuospatial and body-signal integration.
Cruz Fdl, Schumann A, Suttkus S, Helbing N, Bar KJ (2022): Dynamic Changes in the Central Autonomic Network of Patients with Anorexia Nervosa. preprint.
Friederich H-C, Herzog W (2010): Cognitive-Behavioral Flexibility in Anorexia Nervosa. Behavioral Neurobiology of Eating Disorders: Springer Berlin Heidelberg, pp 111-123.
Association of Generalized Anxiety Disorder With Autonomic Hypersensitivity and Blunted Ventromedial Prefrontal Cortex Activity During Peripheral Adrenergic Stimulation.
The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
LeDoux JE (1992): Emotion and the amygdala. The amygdala: Neurobiological aspects of emotion, memory, and mental dysfunction. New York, NY, US: Wiley-Liss, pp 339-351.
Guterstam A, Björnsdotter M, Gentile G, Ehrsson H, H. (2015): Posterior Cingulate Cortex Integrates the Senses of Self-Location and Body Ownership. Current Biology. 25:1416-1425.
The authors are supported by the National Institute of Mental Health (R01MH127225, K23MH112949 to SSK), National Institute of General Medical Sciences Center Grant Award Number (1P20GM121312 to MPP, SSK), and The William K. Warren Foundation. The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Author contributions
FDC, ART and SSK had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the analysis. SSK developed the study design. FDC, ART, and SSK conducted the data analysis. ART, RCL, VU, and SSK acquired the data. SSK and MPP obtained the funding for the study. FDC, ART and SSK wrote the manuscript. All authors provided critical revisions of the manuscript for important intellectual content.
Disclosures
Dr. Paulus is an advisor to Spring Care, Inc., a behavioral health startup, and he has received royalties for an article about methamphetamine in UpToDate, all unrelated to the current work. All other authors report no biomedical financial interests or potential conflicts of interest.
00344-5&id=gr1.jpg){kind=link}
00344-5&id=gr2.jpg){kind=link}
00344-5&id=gr3.jpg){kind=link}
00344-5&id=gr4.jpg){kind=link}