Association between heart rate variability and fluctuations in resting-state functional connectivity

Abstract

Functional connectivity has been observed to fluctuate across the course of a resting state scan, though the origins and functional relevance of this phenomenon remain to be shown. The present study explores the link between endogenous dynamics of functional connectivity and autonomic state in an eyes-closed resting condition. Using a sliding window analysis on resting state fMRI data from 35 young, healthy male subjects, we examined how heart rate variability (HRV) covaries with temporal changes in whole-brain functional connectivity with seed regions previously described to mediate effects of vigilance and arousal (amygdala and dorsal anterior cingulate cortex; dACC). We identified a set of regions, including brainstem, thalamus, putamen, and dorsolateral prefrontal cortex, that became more strongly coupled with the dACC and amygdala seeds during states of elevated HRV. Effects differed between high and low frequency components of HRV, suggesting specific contributions of parasympathetic and sympathetic tone on individual connections. Furthermore, dynamics of functional connectivity could be separated from those primarily related to BOLD signal fluctuations. The present results contribute novel information about the neural basis of transient changes of autonomic nervous system states, and suggest physiological and psychological components of the recently observed non-stationarity in resting state functional connectivity.

Fig. 1

Locations of dACC (top row; blue) and amygdala (bottom row; red) seed regions, shown in neurological convention.

Fig. 2

Schematic of approach for calculating the temporal relationship between functional connectivity and heart rate variability. A sliding-window analysis of heart rate variability is performed on the heart rate time series (upper left), producing a sequence of HRV values across the scan (upper right). Similarly, a sliding-window analysis of correlation between the BOLD signal of the seed time series and that of each voxel in the brain is performed (illustrated for one voxel; lower left), producing a sequence of functional connectivity values (lower right). Finally, the HRV and functional connectivity sequences are compared using linear regression.

Fig. 3

Regions for which temporal fluctuation in functional connectivity with the dACC (top row) and amygdala (bottom row) during resting state had significant positive correlation with fluctuations in HRV. Clusters are significant at p<0.05 corrected.

Fig. 4

Regions for which temporal fluctuation in functional connectivity with the dACC (top row) and amygdala (middle row) during resting state had significant positive correlation with fluctuations in Fourier-based LF (blue) and HF (red) measures of HRV. Bottom row: regions for which BOLD activity was correlated with the time series of LF- and HF-HRV. All maps are cluster thresholded at p<0.05 (corrected), with a height threshold of p<0.0001 uncorrected (t>4.17) for the connectivity maps and p<0.001 (t>3.34) for the activation maps. Peak coordinates are listed in Tables 3A and 3B.

Fig. 5

Significant effects of HF-HRV beyond LF-HRV were found in the indicated brainstem cluster (left); values across subjects (mean±2 SE) are illustrated on the right (*p<0.05, **p<0.01), indicating that the specificity of the effect to HF is due to insignificant LF covariation with connectivity between dACC and brainstem.

Fig. 6

Positive conjoint effect of the LF-HRV component on FC with dACC and amygdala (p<0.05 corected), masked by the respective conjoint HF-HRV effect on both ROIs at p<0.05 uncorrected to obtain a conservative estimate of LF-specific effects. Peak coordinates are listed in Table 3B.

Fig. 7

Relationship between HF-HRV and dACC functional connectivity, overlapping with (red) and distinct from (blue) effects of BOLD signal variation.

Fig. 8

Areas for which the inter-subject analysis of mean functional connectivity and mean HRV survived a threshold of p<0.05 uncorrected, masked with clusters significant at p<0.05 (corrected) in the analysis of the temporal relationship between functional connectivity and HRV.