Heart-brain mapping: Cardiac atrial function is associated with distinct cerebral regions with high free water in older adults

Abstract

Left atrial (LA) dysfunction has been linked to cognitive impairment and cerebrovascular dysfunction. Higher brain free-water (FW) derived from diffusion-MRI was associated with early and subtle cerebrovascular dysfunction and more severe cognitive impairment. We hypothesized that LA dysfunction would correlate with higher brain free-water (FW) among healthy older adults. 56 community older adults (73.13 ± 3.56 years; 24 female) with normal cognition and without known cardiovascular disease who had undergone cardiac-MRI, brain-MRI, and neuropsychological assessments were included. Whole-brain voxel-level general linear models were constructed to correlate brain FW measures with LA indices. We found lower scores in LA function measures were related to higher grey matter (GM) FW in regions including orbital frontal and right temporal regions (p < 0.01, family-wise error corrected). In parallel, LA dysfunction was associated with higher FW in white matter (WM) fibres including superior longitudinal fasciculus, internal capsule, and superior corona radiata. However, LA dysfunction was not related to WM tissue reduction and GM cortical thinning. Moreover, these cardiac-related higher brain FW were associated with lower executive function and higher serum B-type natriuretic peptide (p < 0.05, Holm–Bonferroni corrected). These findings may have implications for anti-ageing preventive strategies targeting cardiac and cerebral vascular functions to improve heart and brain outcomes.

Introduction

Dementia, a major contributor to disability and dependency among older adults, poses a significant public health challenge. ¹ With the expectation that dementia cases will rise to affect one in six individuals in the aging population by 2030, it is crucial to understand the cardiac contributions to cognitive decline for the development of early intervention and prevention strategies. ² Studies have shown that cardiac dysfunction may cause blood vessel alterations and neurovascular damage via both vascular and Alzheimer’s disease (AD) pathophysiological pathways in the brain, which eventually lead to neuronal injury and cognitive impairment. ³ Previous study have demonstrated the association between heart failure and the increased risk of Alzheimer’s disease, suggesting a possible shared pathophysiology between cardiac dysfunction and cerebral neurodegeneration. ⁴ Similarly, another study found that atrial fibrillation is associated with different forms of dementia, including AD. ⁵ Although atrial fibrillation is a common cause of cardiac-related stroke, ⁶ the paradigm of searching for atrial fibrillation only when cardiac-related stroke occurs is likely inadequate for characterizing pre-existing relationships between the left atrium (LA) and cognition. ⁷ Investigations between the LA and cerebral function performed upstream prior to clinical disease may depict pre-existing correlations that can help map cardiac and brain functions associated with cognition, serving as useful blueprints for dementia prevention in cardiac disease.

The LA is a highly dynamic chamber whose function is conventionally understood in three phases (illustrated in Supplementary Figure 1): reservoir, conduit, and booster. In the reservoir phase, the LA expands during left ventricular (LV) contraction and isovolumetric relaxation to receive venous return from the pulmonary circulation; in the conduit phase, the LA drains blood passively into the LV; in the booster phase, the LA contracts upon stimulation by a sinus node depolarization, which contribute to 15–30% of LV stroke volume. ⁸ Atrial cardiopathy refers to the pathophysiological alterations of the LA including atrial fibrosis, deleterious electrical and autonomic remodelling, eventually manifesting as atrial fibrillation and thromboembolic events.^6,9 In non-disease cohorts, LA volumes and reservoir function have been associated with subclinical cerebrovascular disease. For example, among stroke-free community participants, greater LA volumes and reduced LA reservoir function were associated with silent brain infarcts and white matter hyperintensity volumes. ¹⁰ Additionally, the presence of LA enlargement has been associated with elevated brain amyloid by positron emission tomography, among non-demented community dwelling older individuals. ¹¹

Diffusion MRI (dMRI) has emerged as an important method for studying cerebrovascular dysfunction related disorders (e.g., cerebrovascular disease (CeVD), AD and vascular dementia) in ageing adults.^2,12 Free-water (FW) measures, derived from dMRI using a bi-tensor model, reflect the contribution of freely diffusing extracellular water molecules, which are not confined by their local microenvironment. ¹³ Patients with CeVD and AD have higher brain FW levels compared to healthy individuals, and this is linked to declines in cognitive function.^{12,14 –16} Moreover, recent studies also report that the FW compartment can be used to detect early and subtle cerebrovascular-related abnormalities in normal-appearing white matter (WM) without WM hyperintensities.^{16 –18} On the other hand, the FW-corrected tissue content abnormalities represent neurodegenerative microstructural tissue changes such as axonal damage and myelin sheath alterations. ¹³ However, it is unclear how LA function is related to these brain microstructural alterations and their possible contribution to cognitive performance in community older adults.

Here, we analysed the cross-sectional associations between LA phasic functions derived from cardiac magnetic resonance feature tracking (CMR-FT) and brain FW together with neurodegenerative imaging markers (GM cortical thickness (CTh) and WM tissue compartment fractional anisotropy (FAt)) in cognitively normal older adults without known cardiovascular disease. We hypothesized that perturbations in specific phases of LA function would correlate with higher brain FW, which is an early imaging marker for cerebrovascular dysfunction. We sought to test whether these brain measures related to LA function would be associated with cognitive performance and cardiac blood biomarkers in cognitively normal older adults.

Material and methods

Participants

The participants were from the Cardiac Aging Study (CAS), ¹⁹ a prospective study initiated in 2014, which focused on exploring the characteristics and determinants of cardiovascular aging. The primary objective of this study is to investigate the various factors that contribute to cardiovascular health in the aging population and their associations with brain health. By examining these elements, the study aims to develop strategies for preventing cardiovascular diseases in older adults. The CAS participants were recruited from the local community and from the Singapore Chinese Health Study (SCHS), a population-based cohort study established in 1993 focusing on diet and health patterns among the Chinese population in Singapore. ²⁰ The SCHS study involves over 63,000 Chinese men and women aged 45–74 years recruited from community in Singapore, with the primary aim of investigating diet and other lifestyle factors in relation to age related chronic diseases. The comprehensive data collected from this cohort provides valuable insights into the health trends and risk factors prevalent in the aging population in Singapore. The current study cohort consisted of men and women who participated in the baseline CAS 2014 examination and who had no self-reported history of physician-diagnosed cardiovascular disease (such as coronary heart disease, stroke) or cancer. Ethics approval was obtained from National Healthcare Group Domain-Specific Review Board, Singapore (2014/628/C). Participants gave informed consent according to the Declaration of Helsinki.

All participants were examined and interviewed by trained study coordinators. Participants completed a standardized questionnaire that included medical history and cardiovascular risk factors. Cardiovascular risk factors including hypertension, diabetes mellitus, dyslipidaemia, smoking history, and body mass index. Hypertension was defined by current use of antihypertensive drugs or physician-diagnosed hypertension. Diabetes mellitus was defined by current use of anti-diabetic agents or physician-diagnosed diabetes mellitus. Dyslipidaemia was defined by current use of lipid-lowering agents or physician-diagnosed dyslipidaemia. Smoking history was defined as ever smokers or never smokers. Body mass index was calculated as weight in kilograms divided by the square of height in meters. Moreover, an office-based 10-year general cardiovascular disease risk score derived from the Framingham Heart Study ²¹ was calculated for overall cardiovascular risks (Table 1). Sinus rhythm status was ascertained by resting electrocardiogram. Clinical data were obtained on the same day as the assessment of MRI and serum collection. Participants did not have atrial fibrillation or significant valvular heart disease on cardiac imaging. Participants with physician-diagnosed neuropsychiatric conditions, neoplasms, heart, kidney or liver failure, Mini-Mental State Examination (MMSE) score of less than 24, incidental findings on structural MRI, abnormal electrocardiograms (ECGs) or symptoms of heart disease were excluded.

Table 1.

Clinical characteristics of study participants.

Demographic (n = 56)
Age, mean (SD), years	73.2 (3.6)
Gender, male %	32/24 (57%)
Education, mean (SD), years	8.4 (3.8)
MMSE (Max = 30), mean (SD)	27.4 (1.6)
Cardiovascular risk factors
Body mass index, mean (SD), kg/m2	23.9 (3.5)
Systolic blood pressure, mean (SD), mmHg	151.5 (17.2)
Hypertension, %	34 (61)
Hyperlipidaemia, %	31 (55)
Diabetes mellitus, %	11 (20)
Ever-smoker, %	2 (3.6)
FHS-CVD, %	39.4 (18.8)
Left atrial function, mean (SD)
Reservoir strain during systole (εs), %	30.3 (7.5)
Conduit strain during early diastole (εe), %	12.6 (3.7)
Reservoir strain rate (SRs), %, s−1	1.6 (0.5)
Cognitive domain
Attention, T-score, mean (SD)	51.0 (5.6)
Processing speed, T-score, mean (SD)	51.7 (7.6)
Executive, T-score, mean (SD)	51.1 (6.1)
Serum biomarkers (n = 51)
B-type natriuretic peptide (BNP),  median (IQR)	35.0 (37.0)

MMSE: Mini-mental state examination; SD: standard deviation; FHS-CVD: office-based non-laboratory Framingham 10-year cardiovascular disease risk score; IQR: interquartile range.

The sample consisted of 56 participants aged from 68 to 79 years old, with a mean age of 73.2 ± 3.6 years. It included 24 women and 32 men, reflecting a gender ratio of approximately 1:1.33. All participants had good-quality neuroimaging and cardiac data, and completed neuropsychological assessments. Most participants had cardiovascular risk factors of hypertension (60.7%), hyperlipidaemia (55.4%) and diabetes mellitus (19.6%), with 10-year general cardiovascular disease (CVD) risk of 39.4 ± 18.8% (Table 1).

Neuropsychological testing

All participants underwent a comprehensive neuropsychological assessment conducted by trained researchers. These assessments covered three cognitive domains relevant to cognitive ageing and cardiac/cerebrovascular disease: executive function, attention, and processing speed. Executive functioning was assessed using a combination of the Categorical Verbal Fluency Test, the Design Fluency Test in the Delis-Kaplan Executive Function System, ²² and the Trail Making Test B. ²³ Attention was assessed through the Digit Span Test and the Spatial Span Test in Wechsler Memory Scale III. ²⁴ Processing speed was evaluated using the Symbol Digit Modalities Test ²⁵ and the Trail Making Test A. ²⁶ To ensure uniformity in interpretation across different tests, the scores from each test were standardized to T scores (mean = 50, SD = 10). For domains evaluated with multiple tests, domain-average composite scores per participant were computed by averaging the summated T scores from the relevant tests.

Cardiac MRI acquisition and processing

Cine cardiac magnetic resonance (CMR) scans were performed using balanced steady state free precession sequence. All participants were imaged on a 3T magnetic resonance imaging system (Ingenia, Philips Healthcare, The Netherlands) with a dStream Torso coil (maximal number of channels 32). BFFE end-expiratory breath hold cine images were acquired in multi-planar long-axis views (2-, 3-, and 4-chamber views) and a stack of parallel short-axis views to cover the left ventricle (LV) from base to apex. Typical parameters were as follows: TR/TE 3/1 ms; flip angle, 45°; in-plane spatial resolution, 1.0 mm × 1.0 mm to 1.5 mm × 1.5 mm; slice thickness, 8 mm; pixel bandwidth, 1797 Hz; field of view, 300 mm; frame rate, 30 or 40 per cardiac cycle. We developed an in-house semi-automatic algorithm to track the distance (L) between the left atrioventricular junction and a user-defined point at the mid posterior LA wall on standard CMR 2- and 4-chamber views.^27,28 Both 2- and 4-chamber views were used to generate the average strain and strain rate results. Longitudinal strain ( $\epsilon$ ) at any time point ( $t$ ) in the cardiac cycle from end-diastole (time 0) was calculated as: $\epsilon \left(t\right)=\left(L\right(t)-L_0)/L_0$ . LA reservoir strain ( ${\epsilon }_s$ ), conduit strain ( ${\epsilon }_e$ ) and booster strain ( ${\epsilon }_a$ ) were calculated at $t$ equals left ventricular end-systole, diastasis and pre-LA systole, respectively. To derive the peak strain rate (SR) indices, peak values of the first time derivative of the strain-time curve at systole, diastasis and LA contraction were measured. Strain and SR parameters from both 2- and 4-chamber views were averaged to obtain mean results for analysis (Supplemental Figure 2). Other cardiac structural and functional metrics obtained from MRI and echocardiography are detailed in Supplemental Table 1.

Cardiac-related serum biomarker testing

Antecubital venous blood samples (20–30 ml) were taken from consenting participants in the morning. After collection, the blood samples were immediately placed on ice for transportation and were processed within 6 h to obtain serum samples, which were subsequently stored at −80°C. Plasma levels of B-type natriuretic peptide (BNP) were measured by chemiluminescent microparticle immunoassay (ARCHITECT BNP produced by Fujirebio Diagnostics Inc for Abbott Laboratories) using the Abbott ARCHITECT i2000SR analyser. ²⁹

Brain MRI acquisition

Each participant underwent brain MRI scanning using a 3-T MAGNETOM Trio™ from Siemens, Germany. High-resolution T1-weighted structural MRI were conducted using magnetization-prepared rapid gradient echo sequence (MPRAGE; 192 continuous sagittal slices, TR/TE/TI = 2300/2.98/900 ms, flip angle = 9°, FOV = 256 × 240 mm², matrix = 256 × 240, isotropic voxel size = 1.0 × 1.0 × 1.0 mm³, bandwidth = 240 Hz/pixel). Diffusion MRI scans were acquired using echo-planar imaging (EPI) sequence (30 non-collinear diffusion gradient directions at b = 1000 s/mm², six volumes of b = 0 s/mm², TR/TE = 9600/107 ms, FOV = 256 × 256 mm², matrix = 128 × 128, 54 contiguous slices, and voxel size = 2.0 × 2.0 × 2.0 mm³).

Cortical thickness processing

We derived cortical thickness (CTh) from T1-weighted structural MRI images using Freesurfer V6.0 (http://surfer.nmr.mgh.harvard.edu) following previous work.^12,30,31 It included motion correction, removal of the non-brain mass, automated Talairach transformation, intensity correction, volumetric segmentation, and cortical surface parcellation and reconstruction. A Gaussian kernel of 15 mm full width at half maximum (FWHM) was applied to the cortical thickness map of each participant before further analyses.^12,32

Diffusion MRI data processing and free water derivation

Data processing steps are illustrated in Supplementary Figure 3.

Diffusion imaging pre-processing

The diffusion MRI data were pre-processed using FSL V6.0 (http://www.fmrib.ox.ac.uk/fsl). ¹⁶ Eddy current distortions and head movements were corrected to the first b = 0 volume via affine registration of the diffusion MRI. Data were discarded if the maximum displacement relative to the initial b0 volume was more than 3 mm. The diffusion gradients were rotated to improve consistency with the motion parameters. Geometric distortion caused by magnetic field inhomogeneity was corrected using gradient echo sequences (GRE) field maps. Individual maps were visually checked for artefacts, signal dropout, and extra motion. Individual DTI indices maps were created by fitting the DTI model to the pre-processed diffusion data at each voxel.

Free-water imaging method

We utilized the free-water imaging technique on the processed diffusion MRI data to calculate both the fractional volume of extracellular water molecules that move freely (FW) and the fractional anisotropy of water molecules near tissue structures (FAt). ¹³ In essence, the FW compartment accounts for water molecules that are not hindered or confined during diffusion, with a set diffusivity of 3 × 10⁻³ mm²/s (the diffusion coefficient for free water at 37 °C). A map of FW is generated by determining the fractional volume of this specific compartment in each voxel. The FAt map, on the other hand, is derived from a diffusion tensor that models water molecules close to cellular membranes, thereby correcting for the influence of freely diffusing extracellular water. As a result, the FAt compartment is anticipated to offer greater specificity and sensitivity to changes in brain tissue compared to metrics obtained from a single tensor model. Voxel-wise maps for both FW and FAt were generated for each study participant. ³³

White matter and cortical grey matter indices derivation

For WM analysis, we utilized tract-based spatial statistics (TBSS) ³⁴ to carry out a voxel-wise evaluation of fractional anisotropy (FA) across major WM tracts in the brain, in line with our earlier methodology. ¹⁶ Subsequently, the aligned FW and FAt maps for each participant were mapped onto a standardized FA skeleton, yielding individual-level skeletonized images.

For GM analysis, we applied a surface-based method to derive cortical FW using Freesurfer V6.0 following our previous work. ³⁵ Briefly, 1) T1-weighted MRIs underwent cortical reconstruction.^12,30 2) An affine registration matrix was computed between the skull-stripped b0 diffusion image and the segmented T1-weighted image, using a boundary-based algorithm. This method leverages the detailed segmentation of the WM and pial surfaces, ensuring accurate registration between the b0 and T1-weighted images. At this point, we excluded those individuals with coregistration errors through visual inspection. 3) FW volumes were then mapped onto each participant’s individual surface space, generated from the structural cortical segmentation. Cortical FW was sampled at the midpoint along the normal vector between the GM and WM surfaces at each vertex. ³⁶ 4) A partial volume correction for the FW surface maps was applied using Freesurfer following the previous study. ¹² 5) The spherical registration achieved during the cortical thickness processing was administered to normalize each FW surface map to the standard surface. 6) A Gaussian kernel of 15 mm FWHM was applied to individual-level FW surface maps for subsequent analyses. ¹²

Statistical analyses

We generated the mean and standard deviation of the demographic, cardiac markers, and cognitive measures across all participants using Statistical Package Social Sciences (SPSS v.27.0) software (Table 1). Normality of these clinical data, which were utilized in subsequent analyses, was confirmed through the Kolmogorov-Smirnov test. For data that were not normally distributed, such as blood biomarker levels, we applied a logarithmic transformation to achieve normality.

Associations between left atrial cardiac imaging indices and brain measures

Informed by the capacity of voxel/vertex-wise General Linear Models (GLMs) to efficiently handle complex imaging data, we employed these models to explore region-specific brain FW associated with cardiac imaging indices (Figure 1, Step-1). This approach facilitates a detailed examination of region-specific brain-cardiac associations using whole-brain search with multiple comparison correction. The analysis involved vertex-wise surfaced grey matter (GM) FW/Cortical Thickness (CTh) or voxel-wise skeletonized white matter (WM) FW/Fractional Anisotropy (FAt) images as dependent variables, with each left atrial (LA) strain or strain rate parameter as the independent variable. Control variables included age, gender, education, ethnicity, and total brain volume, which are standard in brain MRI research focusing on clinical and aging populations. Additionally, the Framingham 10-year cardiovascular disease risk score was incorporated to adjust for cardiovascular risk factors.

Figure 1.

Study design schematic. 56 community older adults with normal cognition and without known cardiovascular disease were studied. General linear models were performed to identify region-specific associations between brain measures (grey matter and white matter free-water) and left atrial function indices (Step 1). Mean values of brain measures were extracted from significant left atrium-related regions. Partial correlation were calculated between the cognition and left atrium-related brain free-water (Step 2).

In addition, for GM measures, regions were tested for significance using a Monte Carlo simulation with 10,000 repeats (Freesurfer, Glmfit). All GLM results were reported at p < 0.01, family-wise error (FWE) corrected. For WM measures, WM skeletons were examined for statistical significance using threshold-free cluster enhancement (TFCE) and permutation-based non-parametric testing (FSL Randomise).

Associations of brain free water with cardiovascular biomarker levels and neuropsychological assessments

We extracted the mean GM and WM FW from the overlapping regions showing associations with all three cardiac markers. Partial correlation was calculated between the mean GM (or WM) FW and neuropsychological assessments (executive function, attention, and processing speed) across all participants, controlling for age, gender, years of education and total GM (or WM) volume (Figure 1, Step-2). We applied the Holm–Bonferroni multiple comparison correction method, setting the significance level at 0.05 divided by the number of cognitive domains and the two tissue types tested (GM/WM) (0.05/3/2), p < 0.0083). This rigorous approach ensures the robustness of our results. ³⁷ We validated our results by adding office-based non-laboratory Framingham 10-year cardiovascular disease risk score as a nuisance covariate.

Associations of cardiovascular biomarker levels with brain free water or left atrial index

Partial correlation was calculated between the mean GM (or WM) FW and logarithmically transformed BNP levels across all participants, controlling for age, gender, years of education and GM (or WM) volume. Holm–Bonferroni multiple comparison correction for GM/WM (0.05/2, p < 0.025). We validated these findings by adding office-based non-laboratory Framingham 10-year cardiovascular disease risk score as nuisance covariates.

Besides, we also calculated the association of logarithmically transformed BNP levels with the left atrial index, controlling for age, gender, and years of education. Holm–Bonferroni multiple comparison correction for number of LA indices (0.05/3, p < 0.0133).

Results

Associations between left atrial phasic functions and brain grey matter measures

Overall, we found that higher brain GM FW was related to lower LA function (p < 0.01, FWE corrected) in these cognitively normal older adults (Figure 2, Supplemental Table 2). In contrast, no association was found between cardiac imaging indices and cortical thickness.

Figure 2.

Higher grey matter free-water (FW) correlated with lower left atrial function. The whole-brain vortex-wise linear regression indicated that higher FW values in the grey matter regions (highlighted in blue-cyan) were associated with reduced reservoir strain (εs), conduit strain (εe) and reservoir strain rate (SRs). The overlapping brain regions (in red) showing associations with all three cardiac markers were in right orbitofrontal cortex and fusiform gyrus. Boundary of Yeo’s functional intrinsic networks parcellation ⁵³ is shown. Results are reported at p < 0.01, FWE corrected.

The overlapping brain regions showing associations with all three cardiac markers were in the right orbitofrontal cortex and a small part of the fusiform gyrus (Figure 2, bottom right). Specifically, reduced LA reservoir strain (εs) was associated with greater FW in bilateral middle/orbital frontal cortices and right temporal regions, mainly within the default mode network (DMN), executive control networks (ECN), limbic and parts of dorsal/ventral attention and visual networks. Similarly, reduced LA conduit strain (εe) was associated with higher FW in bilateral frontal-temporal regions, within the DMN, ECN, limbic, and parts of attention and visual networks. Reduced LA reservoir SRs was associated with higher FW in the right orbitofrontal and occipital cortices, which contains ECN, limbic and visual networks (Figure 2, Supplemental Table 2). There was no association between GM FW with booster strain (εa), conduit strain rate (SRe) and booster strain rate (SRa).

Similar findings were obtained by 1) adding office-based non-laboratory Framingham 10-year cardiovascular disease risk score as covariates (Supplemental Figures 4A); 2) controlling for the regional cortical thickness (Supplemental Figures 4B).

Associations between left atrial phasic functions and brain white matter indices

Higher WM FW was associated with lower LA functions (p < 0.01, FWE corrected) in cognitively normal individuals (Figure 3, Supplemental Table 3). In contrast, no association was found between cardiac imaging indices and tissue compartment FAt.

Figure 3.

Higher white matter free-water (FW) correlated with lower left atrial function. The whole-brain voxel-wise linear regression indicated that higher FW values in the white matter skeleton (highlighted in blue) were associated with reduced reservoir strain (εs), conduit strain (εe) and reservoir strain rate (SRs). The overlapping brain regions (in red) showing associations with all three cardiac markers were mainly in right superior longitudinal fasciculus, internal capsule, and superior corona radiata. Results are TFCE enhanced, reported at p < 0.01, FWE corrected. The white matter skeleton is shown in green.

The overlapping brain regions showing associations with all three cardiac markers were in right projection and association fibres including right superior longitudinal fasciculus, internal capsule, and superior corona radiata (Figure 3, bottom right). Specifically, reduced reservoir strain (εs) was associated with greater WM FW in bilateral parietal-frontal projection fibres, right temporal association fibres and the corpus callosum. Reduced conduit strain (εe) was associated with higher WM FW in widespread parietal and frontal regions, and right temporal association fibres. Reduced reservoir strain rate (SRs) was associated with enhanced WM FW in multiple right hemispheric WM tracts including superior longitudinal fasciculus, internal capsule, corona radiata and a small region in the genu of the corpus callosum (Figure 3, Supplemental Table 3). There was no association between WM FW with booster strain (εa), conduit strain rate (SRe) and booster strain rate (SRa).

The associations remained significant after 1) further controlling for office-based non-laboratory Framingham 10-year cardiovascular disease risk score (Supplemental Figures 5A); 2) controlling for regional WM volume (Supplemental Figures 5B).

Associations of left atrium-related brain free water with executive function

There was no association between LA phasic function measures and cognitive function. In contrast, we found that higher mean FW in the overlapping left atrium-related brain regions, i.e., those showing associations with all three cardiac markers (Figures 2 and 3, bottom right panel), was associated with lower executive function scores (Figure 4) in old adults. This was true for both GM regions (Figure 4 left, r = −0.37, CI = [−0.62, −0.11], p = 0.0053, survived multiple comparison correction) and WM regions (Figure 4 right, r = −0.36, CI = [−0.61, −0.11], p = 0.0060, survived multiple comparison correction). However, both GM and WM FW were not associated with processing speed and attention. The associations remained significant by further controlling for office-based non-laboratory Framingham 10-year cardiovascular disease risk score (GM: r = −0.36, CI = [−0.62, −0.10], p = 0.0068; WM: r = −0.35, CI = [−0.61, −0.10], p = 0.0073; both are survived multiple comparison correction).

Figure 4.

Higher brain free-water correlated with lower executive function. Mean FW is extracted from the overlapping brain regions showing associations with all three cardiac markers. GM: grey matter; WM: white matter; FW: free-water.

Association between left atrium-related brain free water and serum BNP

In the subset of 51 participants with serum cardiac biomarkers, we found that higher FW in those LA-relevant regions was associated with higher log-transformed serum BNP level (GM: r = 0.40, CI = [0.16, 0.65], p = 0.002; WM: r = 0.38 CI = [0.11, 0.64], p = 0.006; both are survived multiple comparison correction). The associations remained significant by further adjusting the office-based non-laboratory Framingham 10-year cardiovascular disease risk score (GM: r = 0.42, CI = [0.19, 0.64], p = 0.0025; WM: r = 0.36, CI = [0.10, 0.63], p = 0.0085; both are survived multiple comparison correction).

Higher serum BNP level was also correlated with lower LA MRI indices (εs: r = −0.38, CI = [−0.66, −0.11], p = 0.0066; εe: r = −0.34, CI = [−0.60, −0.08], p = 0.0131; SRs, r = −0.37, CI = [−0.65, −0.10], p = 0.0076; all are survived multiple comparison correction).

Discussion

To our knowledge, this is the first study examining the associations between brain FW and LA function measured by phasic LA strain and SR in cognitively normal older adults without heart disease. We found lower LA reservoir and conduit measures were associated with higher extracellular FW in bilateral middle/orbital frontal and right temporal GM regions, and bilateral parietal-frontal projection WM fibres, right temporal association fibres and the corpus callosum. In contrast, none of these cardiac imaging markers was associated with WM tissue FAt and GM cortical thickness. Moreover, these LA-related higher brain FW was associated with worse executive function. Our findings suggest that in the context of older adults with normal cognitive functions, disturbances in LA function are linked to alterations in brain FW, supporting a novel relationship between the vascular nervous system and the cardiac atrial function, although causal relationships are undefined.

Mechanism of the association between left atrial function and brain free-water

Our study demonstrates that FW measures in the brain are associated with LA reservoir and conduit in cognitively normal older adults without cardiac dysfunction. Recently, brain FW alterations have gained increasing attention because of their capability of detecting early brain abnormalities in cerebrovascular disease^14,38 and subtle vascular-related abnormalities in normal-appearing WM.^16–18 Previous studies suggest that increases in FW are most likely beginning from brain extracellular water features. ¹³ However, the exact reasons leading to the higher FW and hence increased extracellular water signal in cardiac-related conditions are not yet well-defined. The observation of a positive association between serum BNP and brain FW may point to a systemic body water connection between cardiac chamber stretch (represented by BNP ³⁹ ) and extracellular water signals represented by brain FW. It is also plausible that mechanistically, a subtle form of cerebrovascular-related insult^15,16,18 is occurring in the circulatory system that explains our observations. In more severe clinical scenarios, cardiac dysfunction leads to reductions in blood supply to the brain or cardioembolic strokes.^40,41 These processes, in turn, lead to early cerebrovascular inflammation, and endothelial and blood brain barrier (BBB) dysfunction ³ which manifests as increases in brain FW.

Reservoir and conduit but not booster strains associate with brain free-water

Reservoir and conduit strains in the LA were distinctively associated with brain FW. Based on left atrial mechanics, reservoir and conduit strains are predominantly ‘passive’ (for the LA) phenomena that occur over cardiac circulatory cycle, processes that are associated with stretch and compliance state of the cardiac chambers. ⁴² The LA is intimately linked to the LV as it is directly exposed to the LV cavity during diastole. LV mechanical events occur during the LA atrial reservoir period, from isovolumic contraction, ejection to isovolumic relaxation. The LA conduit phase covers early LV filling and diastasis. ⁴³ In disease states where there are increased LV filling pressures such as heart failure, LA dilatation is associated with poor clinical outcomes. ⁴⁴ Overall, LA structural and functional remodelling has been described as a barometer of diastolic burden that is related to diastolic dysfunction severity and cardiovascular death. ⁴⁵ In the absence of clinical disease and normal LV function, our results expand on the potential for LA mechanical function as a corollary marker of cardiovascular ageing that is related to other vascular beds such as the brain. We had observed that both LA reservoir strain and reservoir strain rate were associated with brain FW. In the atrial fibrillation literature, reservoir strain and reservoir strain rates are impaired in patients with atrial fibrillation, additionally lower in patients with recurrent atrial fibrillation. ⁴⁶ LA reservoir strain is also an independent predictor of stroke, even in lower-risk patients, adjusted for abnormalities in LV function and mass. ⁴⁷ Notably, our participants were in normal sinus rhythm and were stroke-free community-living older adults. Further, our observations between LA reservoir strain and brain markers occurred among participants with normal LA sizes, suggesting earlier upstream connections between the heart and brain, in contrast to other studies that included participants with subclinical but abnormal LA volumes.^10,11 This may suggest that targeting asymptomatic participants only when LA volumes are abnormal may be insufficient (too late) as a preventative strategy.

On the other hand, LA booster strain represents left atrial contractile phase performance that depends on multiple phenomena such as preload, afterload, intrinsic contractility as well as electromechanical coupling. ⁴³ The lack of association between booster strain and brain markers in our participants is notable and warrants broader validation. However, they are not surprising and may point against electromechanical activation as a mechanism that links the heart and brain functions represented by this study. Finally, LA booster functions are seen to deteriorate in clinical disease states such as heart failure, ⁴⁸ which would not be expected in otherwise well-community participants.

Left atrium related free-water alteration in heart function control centres and executive control network

Interestingly, we observed that lower LA markers were related to higher FW in the regions including the insula, orbitofrontal cortex and cingulate, which are recognised as heart function control centres.^49,50 Moreover, we also found higher FW in the WM association fibres linking cortical neural control of the cardiovascular system, and WM project fibres of afferent and efferent pathways collecting cortical region to heart controlling centres in subcortical forebrain structures and brain stem. ⁵¹ Therefore, we hypothesize a potential vicious cycle in the progression of cerebrovascular disease. Subtle cardiac-derived hemodynamic dysfunction, possibly indicated by an increase in brain FW within brain regions governing cardiac functions and their associated white matter fibres, might further exacerbate the disorganization of cardiac function. ⁵⁰ This interpretation aligns with findings from both animal and human studies that demonstrate early hemodynamic changes can precede significant neuronal damage. ⁵² Future studies could prioritize the evaluation of sympathetic activity to shed light on the neural mechanisms driving the intricate heart-brain interactions. ⁵⁰

The implied GM regions overlapped with the executive control networks (ECN) in the mid-frontal, inferior-parietal and insula. ⁵³ This is aligned with the previous work demonstrating that ECN was related to cerebrovascular disease. ³⁵ In line with the finding, we also demonstrated that cardiac function-related higher FW was associated with worse cognitive test of executive function. Intriguingly, we found that brain FW, but not neurodegenerative alterations (i.e. WM tissue FAt and cortical thickness) was associated with left atrial function. These higher FW in the cognitive normal might be susceptible to early and mild cerebrovascular dysfunction, which could appear early than the cortical neuronal loss and WM tissue damage.^54,55 Conceptually, future preventative strategies that target vascular risk factors may meaningfully achieve protective effects on neurological tissues and cardiac tissues.

Our study highlights a critical connection between left atrial function and brain free-water levels in older adults, emphasizing the interplay between cardiac function and cerebrovascular health. This heart-brain relationship suggests potential strategies for the prevention and management of cognitive decline in older adults. Non-invasive imaging measures, like brain FW derived from diffusion-MRI, may serve as early biomarkers for cerebrovascular dysfunction.^14,38 This early detection, even among older adults with normal cognition and no known cardiovascular disease, provides opportunities for timely preventive interventions. Our work provides insights for future research and clinical trials including pharmaceutical therapies that target this heart-brain connection, for preventive/early treatment of cognitive and cardiac declines with ageing.

Limitations

We had focused our current analyses on LA function as this was a sensitive and novel cardiac marker observed in association with early cardiac ageing processes. ²⁸ In contrast to disease cohorts where gross abnormalities in larger cardiac ventricular chambers have been observed in cerebrovascular diseases, the aim of the current work was to identify early asymptomatic connections between the heart and the brain relevant for future preventive strategies. In this sample, LV and structures were within clinically normal limits. We focussed on LA strain and SR parameters (and not other LV functions) as MRI-derived LA strain has been shown to be an early upstream perturbation of cardiac ageing in this cohort, for which gross LV functions were clinically normal. ²⁸ Our study is observational, hence no causal inferences should be made. The sample size is small (with limited power for sub-analyses) and future larger studies are warranted to verify the associations, despite statistically significant results. Sinus rhythm was ascertained at single time point based on electrocardiogram while sufficient for community-based studies would not definitively exclude paroxysmal or undiagnosed atrial fibrillation. LA strain reflects global left atrial function in the longitudinal direction only, hence segmental deformation circumferential and radial strains are not considered. However, left atrial strains are generally longitudinal due to thin atrial walls and fibre orientations. ⁵⁶ The use of MRI scanners provided state-of-the-art parameters of interest, but may limit research to settings that may not have similar resources, processing software and knowledge. For the brain MRI processing, we have conducted a boundary-based registration step and a visual quality control step to minimize dMRI-T1 miss registration. However, because of the relatively low resolution of dMRI space compared to the high-resolution T1 maps, partial volume effects cannot be entirely excluded, in which case brain atrophy may also influence the FW values. To address these influences, we have carried out partial volume.

Conclusions

In conclusion, as proof of concept, we have defined new relationships between phasic LA function and brain FW increase in community older adults with normal cognition. The advent of advanced imaging methodologies in both the nervous and cardiovascular systems has led to a suite of experimental investigations for the determination of neuro-cardiologic interactions and the prevention of cognitive and cardiac dysfunctions.

Data availability statement

Data are available upon reasonable request.

Supplementary material

Supplemental material for this article is available online.