Brain activation and heart rate variability as markers of autonomic function under stress

Abstract

Efficient brain-heart interactions, mediated by the central autonomic network (CAN), are crucial in regulating physiological and psychological stress. The ability of the autonomic nervous system to adapt to stress predicts resilience to cardiovascular, anxiety, and mood disorders. Since the neural dynamics underlying brain-heart interactions remain poorly understood, this study investigated brain activation and heart rate variability (HRV) during stress and relaxation. Functional magnetic resonance imaging (fMRI) and peripheral heart rate assessment were used to assess brain-heart coupling during breathing-induced relaxation, psychosocial stress and stress recovery in 32 healthy participants. We assessed the relation between perceived stress and brain activation, and employed non-linear generalized additive models to forecast changes in HR based on brain activation in the CAN. Both breathing-induced relaxation and stress induction significantly affected HR variation and triggered brain activation in CAN-related regions. HR variation was related to CAN activity during stress induction, and that chronic stress was linked to reduced brain activation during stress recovery. Finally, we demonstrated that brain activation within the CAN predicts changes in HRV. Our results offer novel insights into dynamic brain-heart interactions during stress-related autonomic regulation and emphasize the brain-heart axis's potential as a target for therapeutic interventions aimed at enhancing stress resilience.

Keywords: Central autonomic network; Functional MRI; Generalized additive modelling; Heart rate variability; Stress.

Fig. 1

Graphical depiction of the Deep-Breathing stress-induction fMRI task. (A) The experimental task starts with 4 min of unpaced (free) breathing, followed by 4 min of paced breathing with an interval of 10 breaths per minute and 4 min of slow breathing (deep-paced breathing, 5.5 breaths per minute). Following this, participants had to complete a modified version of the Montreal Imaging Stress Task (Dedovic et al. 2005), where simple arithmetic problems have to be solved within a time-frame individually calculated to be too short to solve the task, and feedback regarding their own performance relative to the average performance of other participants was provided on a mock performance bar, manipulated to exaggerate the average performance. The task finished with a “recovery-from-stress” block of 4 min of slow breathing.

Fig. 2

Changes in Heart Rate Variation during the experimental task. (A) Relation between the experimental task and peak-to-peak interval of photoplethysmography signal (inter-beat intervals, IBI, P = 0.0584, η² = 0.08). UB = unpaced breathing, NB = normal-paced breathing, DB = deep (slow)-paced breathing, SI = Stress induction, RDB = Recovery deep(slow)-paced breathing. Sig. differences were observed between DB and NB (P < 0.001), SI and DB (P = 0.001), RDB and SI (P < 0.001) and RDB and NB (P < 0.001). (B) Relation between the experimental task and the root mean square of the successive difference in adjacent IBIs (RMSSD, non-sig.). (C) Negative correlation between mean scores of the “Perceived Stress Scale” and differences in Interbeat-Intervals when comparing recovery deep(slow)-paced breathing (RDB) with deep(slow)-paced breathing (DB, r = − 0.383, P = 0.0304). Importantly, this finding did not survive Bonferroni correction for multiple comparisons (20 comparisons per HRV metric: α = 0.0025). Pearson product-moment correlation coefficient (2-tailed). Regression line is depicted with 95% confidence bands.

Fig. 3

Whole-brain fMRI activation during paced breathing and stress induction. (A) Significant brain activation during the “Relaxation” contrast (deep-paced breathing vs. normal-paced breathing), shown here are activations in the thalamus and parietal operculum (left) and bilateral postcentral/somatosensory gyrus (right), (B) Significant brain activation during the “Stress Reactivity” contrast (stress induction vs. deep-paced breathing), activations in the periaqueductal gray, bilateral middle temporal gyrus (left) and anterior cingulate cortex (right) are shown. Results are exclusively masked with the contrasts “stress induction” vs. “unpaced breathing” and “stress induction” vs. “normal-paced breathing”, shown at the right. (C) “Stress Recovery Specific Relaxation” (recovery deep-paced breathing vs. deep-paced breathing), significant activation in the middle temporal gyrus is shown. (D) “Post Stress Relaxation” (recovery deep-paced breathing vs. normal-paced breathing), significant activations in the thalamus and medial cingulate gyrus are shown. (E) Influence of inter-beat intervals values on brain activation during the contrast “Stress Reactivity” (stress induction vs. deep-paced breathing), inclusion of IBI values (re-sampled at the fMRI TR) in individual first-level GLMs. Shown here are significant activations in the thalamus, putamen, anterior insula and medial cingulate. (A, C, D) All results significant at a cluster-level FWE-corrected significance threshold of P < 0.05, with a cluster defining threshold of FWE P < 0.05 and minimal cluster size k > 10. (B) Results significant at a cluster-level FWE-corrected significance threshold of P < 0.05, with a cluster-defining voxel-wise threshold of FWE P < 0.0001 and minimal cluster size k > 10. (E) All results significant at a cluster-level FWE-corrected significance threshold of P < 0.05, with a cluster-defining voxel-wise threshold of P < 0.001 uncorrected and minimal cluster size.

Fig. 4

Whole brain regression with chronic stress. Significant negative association between brain activation during the contrast “Stress Recovery Specific Relaxation” (recovery deep-paced breathing vs. deep-paced breathing) and scores of the TICS-SCSS scale (chronic stress), significant activation in the anterior cingulate cortex/medial prefrontal cortex is shown. Results significant at a cluster-level FWE-corrected significance threshold of P < 0.05, with a cluster-defining voxel-wise threshold of P < 0.001 uncorrected and minimal cluster size k > 10.

Fig. 5

GAM results for the predictive relationship between brain activation and HRV. (A) Smooth function for thalamus time-series activation significantly predicts RMSSD (F = 4.155, P < 0.001, edf = 8.655). (B) Smooth function for anterior insula time-series activation significantly predicts RMSSD anterior insula (F = 3.121, P = 0.0024, edf = 7.991). (C) Smooth function for vmPFC time-series activation significantly predicts IBI (F = 8.770, P < 0.001, edf = 8.480), (D) Smooth function for medial cingulate time-series activation significantly predicts IBI (F = 5.755, P < 0.001, edf = 7.986). All models included subject and condition as fixed effects, and smooth terms were estimated using REML. The shaded band represents the 95% confidence interval around the smooth fit. The dashed lines outline the upper and lower bounds of the confidence interval. The gray dots represent the model residuals for each observation. Masks used to extract brain activation time series are shown at the lower right corner in each plot.