Heart-brain synchronization breakdown in Parkinson's disease

Abstract

Heart rate variability (HRV) abnormalities are potential early biomarkers in Parkinson's disease (PD) but their relationship with central autonomic network (CAN) activity is not fully understood. We analyzed the synchronization between HRV and brain activity in 31 PD patients and 21 age-matched healthy controls using blood-oxygen-level-dependent (BOLD) signals from resting-state functional brain MRI and HRV metrics from finger plethysmography recorded for 7.40 min. We additionally quantified autonomic symptoms (SCOPA-AUT) and objective autonomic cardiovascular parameters (blood pressure and heart rate) during deep breathing, Valsalva, and head-up tilt, which were used to classify the clinical severity of dysautonomia. We evaluated HRV and BOLD signals synchronization (HRV-BOLD-sync) with Pearson lagged cross-correlations and Fisher's statistics for combining window-length-dependent HRV-BOLD-Sync Maps and assessed their association with clinical dysautonomia. HRV-BOLD-sync was lower significantly in PD than in controls in various brain regions within CAN or in networks involved in autonomic modulation. Moreover, heart-brain synchronization index (HBSI), which quantifies heart-brain synchronization at a single-subject level, showed an inverse exposure-response relationship with dysautonomia severity, finding the lowest HBSI in patients with severe dysautonomia, followed by moderate, mild, and, lastly, controls. Importantly, HBSI was associated in PD, but not in controls, with Valsalva pressure recovery time (sympathetic), deep breathing E/I ratio (cardiovagal), and SCOPA-AUT. Our findings support the existence of heart-brain de-synchronization in PD with an impact on clinically relevant autonomic outcomes.

Fig. 1. Methodological sketch.

1 Simultaneous and synchronous data acquisition for brain resting-state functional MRI (rs-fMRI) and finger photoplethysmography (PPG); 2A Heart Rate Variability (HRV) metrics computation using the PhysioNet Cardiovascular Signal Toolbox and definition of K overlapping sliding time windows with integer multiples of the Repetition Time (TR) parameter of fMRI acquisition; 2B Preprocessing of blood-oxygen-level-dependent (BOLD) signal from rs-fMRI encompassed slice-time correction, removal of movement and physiological noise, band-pass filtering, registration to the Montreal Neurological Institute 152 (MNI152) template and spatial smoothing with a 6 full width at half maximum (FWHM) kernel; 3 Maximum Lagged cross-correlations were assessed for building HRV-BOLD Sync Maps for each participant. One map was obtained per window length; 4 Group differences were assessed for the HRV-BOLD Sync Maps, separately using the maps obtained at different window lengths; 5 Group comparison results across different windows were combined using the Fisher’s statistic and Bonferroni-correction.

Fig. 2. Central Autonomic Network and its intersection with HRV-BOLD de-synchronization maps.

a Anatomical representation of the Central Autonomic Network (CAN) created by pooling several regions extracted from different public MRI atlases (see Methods section in main text). b Statistically significant brain regions after group comparisons with the contrast patient <control in NNiqr-BOLD synchronization maps and surviving multiple comparisons. CAN network is colored in gray. The map of significant results is shown in a color scale that represents the value of the X² statistic of the Fisher’s test. Significant brain regions overlay with those belonging to CAN, including cerebellum, brainstem, lateral parietal-temporal cortex, medial prefrontal cortex, insula, hypothalamus, and anterior cingulate cortex (see Results section in main text for further details). HYP hypothalamus, INS insular cortex, ACC anterior cingulate cortex, AMG amygdala, mPFC medial prefrontal cortex, PAG periaqueductal gray matter, MED medulla, PBC parabrachial complex, BNST bed nucleus of the stria terminalis.

Fig. 3. Relationship between heart-brain synchronization index and clinical autonomic outcomes.

Panel a shows in its left side that the method to classify the severity of cardiovascular dysautonomia was based on a cardiovascular severity composite including E/I ratio, Valsalva PRT and ΔSBP phase IV (see Methods). PD patients with a composite value belonging to the 1st-quartile of the distribution of values were classified as PD sev-dys (n = 8), those belonging to the 4th-quartile as PD mild-dys (n = 8), and the rest of patients were considered as PD mod-dys (n = 14). On the right side, panel a shows a boxplot with differences in the distribution of heart-brain synchronization index (HBSI) (see Methods) for the dysautonomia severity PD subgroups and the controls (n = 21). All group comparisons were statistically significant (*p < 0.05). Panels b, c show scatter plots that relate the HBSI with two cardiovascular autonomic parameters, deep breathing E/I ratio (left) and Valsalva PRT (right), (b), and with autonomic manifestations (total SCOPA-AUT) (c). The statistical significance and values of r (Rs) of the correlations calculated independently for patients and controls are shown in each graph, as well as the corresponding adjusted regression lines. All represented variables are dimensionless except for Valsalva PRT, measured in seconds. PD Parkinson’s disease, PD sev-dys PD patients with severe dysautonomia, PD mod-dys PD patients with moderate dysautonomia, PD mild-dys PD patients with mild dysautonomia, Valsalva PRT Valsalva pressure recovery time, Deep breathing E/I ratio expiratory-to-inspiratory ratio for heart rate variability.*p < 0.05; **p < 0.01; ***p < 0.005.