The effects of cerebral white matter changes on cardiovascular responses to cognitive and physical activity in a stroke population

Abstract

Autonomic nervous system (ANS) control may be disrupted by cerebrovascular disease. We investigated the relationship between alterations in white matter integrity and regulation of the ANS in 23 participants who sustained a stroke within five years. These participants underwent diffusion tensor imaging and fractional anisotropy values were calculated (DTI-FA) for each hemisphere and lobe. Cognitive and physical exertion tasks were performed while recording an electrocardiogram (ECG). Respiratory sinus arrhythmia (RSA) decreased more during a verbal fluency task with lower left hemisphere DTI-FA. Further, the physical stressor yielded decreases in RSA with lower frontal DTI-FA, and higher temporal lobe DTI-FA, p<.05 (perhaps a release effect on the central autonomic network). Decrements in ANS regulation may have functional consequences that alter behavior, as well as potentially increasing the risk for further vascular disease.

Introduction

Cerebrovascular disease is one of the major causes of neurobehavioral disorders, including dementia in the United States and Europe (Gorelick, Román, & Mangone, 1994) and hypertension is a major risk factor for the development of cerebrovascular disease. In addition to injuring the networks that directly mediate voluntary activities such as cognition and purposeful motor behaviors, the cerebral cortex controls the autonomic nervous system (ANS). Autonomic nervous system functions are important in almost every task we perform and impairments of the functional systems involved in control of the ANS may have significant impact on behaviors such as cognition, emotion and physical activity, as well as altering cardiovascular factors that can influence the risk for future strokes, heart and kidney diseases. Though little research exists on the impact of cerebrovascular disease on the cerebral top down control of autonomic functions, we can gain insight by studying autonomic function in participants with cerebrovascular disease.

Patients who suffer with cerebrovascular disease often demonstrate a deterioration of frontal-executive functions (Desmond, 2004; Kramer et al., 2002; Boyle et al., 2003; Román and Royall, 1999; Pohjasvaara et al., 2002). The reason for this relationship between cerebrovascular disease and frontal-executive dysfunction may be related to two factors. Whereas major strokes often do cause cortical injury, cardiovascular disease is also associated with the development of white matter disease (Ylikoski et al., 2000). Moser et al. (2001) notes that vascular injury to the brain most often induces an impairment in the frontal-subcortical networks. These vascular disorders that injure the subcortical networks include the Wallerian degeneration related to cortical neuronal injury, ischemic demyelization and lacunes. These disorders of the white matter are likely to cause frontal lobe–executive dysfunction (Pantoni & Garcia, 1997; Annoni, et al. 2003). Frontal-executive dysfunction with subcortical white matter injury is related to the extensive interconnectivity of the frontal lobes with other brain areas.

While the frontal lobes do receive input from the olfactory system, (Nauta, 1973), unlike the posterior neocortex the frontal lobes receive no direct visual, tactile or auditory sensory projections, and have no primary sensory association areas for these modalities. The sensory information received by the frontal lobes, therefore comes by means of white matter pathways from these posterior sensory association areas, as well as from the polymodal inferior parietal lobe (Nauta, 1971; Petrides and Pandya 1984). Based on this pattern of connectivity, Nauta (1971) suggested that the entire prefrontal cortex should be considered association cortex. Whereas the inferior parietal lobe association cortex receives projections from the pulvinar and the posterior lateral nuclei of the thalamus, the frontal lobes, in addition to receiving extensive connections from the thalamic dorsomedial nucleus also receive connections from the pulvinar nucleus, the thalamic reticular nucleus, and intralaminar nuclei. The frontal lobe is unique in that it has the greatest variety of connections to the limbic system, including the cingulate cortex, the amygdala and the entorhinal-hippocampal system. Another critical subcortical connection is with septal–hypothalamic-mesencephalic network (Nauta, 1971). In addition to receiving dopaminergic input from the ventral tegmental area of the mesencephalon via the mesocortical pathway, the frontal lobes have strong connections with the ventral and dorsal striatum and receive important input from the cerebellum. Thus, disorders of the cerebral white matter are more likely to cause frontal-executive dysfunction, than temporo-parietal cognitive dysfunction.

Also, a neuroimaging study of white matter disease in patients with vascular dementia demonstrated preferential damage of frontal white matter (Gootjes et al., 2004), illustrating a more direct explanation of disease-behavior interaction. Further, another study, using volumetric measures of MRI and regional PET glucose metabolism in cerebral cortex, showed hypometabolism in the frontal cortex regardless of region of detected white matter hyperintensities and also demarcated significantly higher frequency of white matter hyperintensities in the prefrontal region compared with other brain regions. No relationship was found between white matter hyperintensities in any region of the brain and altered metabolism in parietal or occiptotemporal regions (Tullberg et al., 2004). Overall, based on the studies reviewed above there is substantial evidence to suggests that prefrontal/subcortical networks are disrupted in individuals who have a diminished integrity of their white matter. These frontal-subcortical networks are important in the mediation and control of voluntary goal oriented behaviors; however, impairments of cerebral subcortical white matter may also influence the functions of the involuntary ANS.

Cerebral Interactions in Autonomic Nervous System Control

Classic research on the (ANS) recognized the importance of the cerebrum in modulating peripheral physiological responses to stressors (e.g., Cannon, 1914; Pavlov, 1927) and frontal networks have especially been implicated in cerebral control of the ANS (Loewy, 1991). In comparative neuroscience research, consistent results demonstrate a role of the frontal cortex in inhibiting sympathetic nervous system activation, possibly by modulating parasympathetic action (Owens & Verberne, 2001; Owens et al., 1999; Hardy & Holmes, 1988) including baroreceptor reflexes (Resstel et al., 2004). In humans, cardiovascular arousal (increased heart rate and mean arterial pressure) has been associated with decreased rCBF (PET) in prefrontal cortex (Critchley et al., 2000). In a review of the literature on cortical control of the cardiovascular system, Verberne and Owens (1998) cite evidence from anatomical, lesion, and electrical stimulation studies that the medial prefrontal cortex is preferentially involved in modulating inhibitory responses of the sympathetic division of the ANS. They conclude that medial prefrontal cortex has consistently been demonstrated to be a depressor area with respect to autonomic function (e.g., activity is associated with decreased blood pressure). Gianaros et al. (2011) study of healthy adults found that during a stressful cognitive task, fMRI activity in the dorsal and posterior anterior cingulate, as well as the posterior cingulate cortex correlated with reductions in baroreceptor suppression (increases in BP and heart rate).

The full extent of anatomical bridges between the frontal cortex and nuclei involved in autonomic control is not fully known. There are, however, cortical projections to the nucleus of the solitary tract, which also receives cardiopulmonary afferents. These interconnections are important and are thought to be involved in blood pressure, vasomotor, and heart rate regulation through baroreceptor suppression (an autonomic control system that modulates and coordinates cardiovascular responses to changes in blood pressure).

Although there have been some mixed results, both animal and human studies suggest that there is lateralized hemispheric differentiation of the systems that control the ANS (Oppenheimer et al., 1992; Oppenheimer et al., 1996; Yokoyama et al., 1987; Yoon et al., 1994; Porges at al., 1994). For example, electrical stimulation of left insular cortex yields bradycardia whereas stimulation of the right insular cortex yields tachycardia (Oppenheimer et al., 1992). Further, right hemispheric damage in the rat is associated with elevated catecholamines with corresponding increases in heart rate and blood pressure (Hachinski et al., 1992) and in the human with tachyarrhythmia (Lane et al., 1992). Thus it appears that the right hemisphere more strongly controls sympathetic functions and the left hemisphere, parasympathetic functions. The hemispheric brain systems that are involved in the regulation of autonomic activity are integrated with the systems involved in the regulation of behavioral processes including executive functions (Gianoros et al., 2004), affect (Damasio et al., 1990; Berntson, Sarter, & Cacioppo, 1998), and state regulation (Porges, 1995, 2007).

One way to study cortical/subcortical contributions to the control of the ANS has been to examine stressor responses including responses to physical and cognitive activities as well as social tasks. Translating this particular experimental stressor method of research to participants with cerebrovascular disease may allow an understanding of the role of decrements of white matter integrity induced by ischemic demyelization (white matter hyperintensities or leukoaraiosis) and regional lacunes in autonomic control in the context of related factors such as executive dysfunction and alteration of emotions.

Cerebrovascular Disease and the Autonomic Nervous System

There is evidence that cerebrovascular disease alters autonomic functions, though much of this evidence comes from comparative neuroscience studies and studies of patients with acute strokes (Cechetto, 1993; Vingerhoets, 1993). However, autonomic dysregulation associated with cerebrovascular disease has been understudied, with the exception of some recent research demonstrating linkages between cardiovascular reactivity and hypertension as risk factors for the development of stroke and vascular cognitive impairment. Greater cardiovascular reactivity/variability to orthostatic manipulation, mental stress (e.g. mental arithmetic), and during sleep have all been associated with increased risk of stroke and cerebrovascular disease (Kario et al., 2002; Kario et al., 1997; Everson et al., 2001). Further, higher levels of resting blood pressure are associated with increased prevalence of white matter disease (Vermeer et al., 2002).

Stritmatter et al. (2003) examined the location dependent differences in autonomic function in patients with ischemic stroke. They studied patients with right and left hemisphere strokes, as well as patients with brainstem and cerebellar strokes to learn if there were differences in norepinephrine and epinephrine levels as well as blood pressure, heart rate, and cardiac output during the first five days following stroke. They concluded that sympathetic functioning was most affected by right hemisphere stroke (increased nor-epinephrine levels and increased overall blood pressure). Waldstein et al., (2004), using anger recall, a social speech task, and mental arithmetic as stressors, found greater stress-induced blood pressure reactivity was associated with progression of cerebrovascular disease. The design was cross-sectional and they used visually identified white matter hyperintensities as their imaging marker (Fazekas criteria). They claim that their study was the first to examine blood pressure reactivity to indices of cerebrovascular disease using MRI.

The aim of the present study is to evaluate the impact of regional white matter injury on heart rate variability in a population of participants who suffered with an ischemic stroke within five years. The severity of white matter injury was evaluated using diffusion tensor imaging – fractional anisotropy (DTI-FA). Our major hypothesis was that impairment of white matter connections within the functional systems that regulate the ANS would result in altered ANS function and that this relationship might be regional. The frontal lobes perform many inhibitory functions and therefore, damage to prefrontal cortex often results in release phenomena (e.g., the return of rooting, grasp and sucking reflexes, as well as emotional disinhibition). Though it is unknown if the frontal lobes also have a similar control of the ANS there is some evidence to support this role (as reviewed). Thus, we posited that frontal lobe white matter damage might induce dyscontrol of ANS activation with exaggerated cardiovascular responses to stressors. In contrast, temporal lobe white matter damage might induce a failure of activation with a lack or reduction of cardiovascular response to stressors. In addition, we anticipated that there may be a laterality effect that would depend on the systems recruited by the type of stressor used (e.g, verbal fluency, left hemisphere; nonverbal fluency, right hemisphere).

Methods

Participants

Twenty-three subjects (14 men, 9 woman; average age = 69+-8 years) were selected from the Risk Markers for Dementia after Stroke (RMDAS) sample. RMDAS is a five-year longitudinal study of 108 ischemic stroke patients with baseline and annual neuropsychological and imaging examinations (Williamson et al., 2008; Williamson et al., 2010). RMDAS exclusion criteria included severe aphasia (≤50% on the Boston Diagnostic Aphasia Examination Commands subtest) or severe dementia (MMSE score < 10); prior dementing or degenerative neurological disease other than stroke, current psychiatric illness, or an inability to complete brain MRI. Further exclusion criteria for this study included multi-factorial cardiovascular disease/systemic issues (e.g., heart attack + atrial fibrillation + diabetes + congestive heart failure). Originally, the intention was to exclude subjects taking beta-blockers, angiotensin converting enzyme (ACE) inhibitors, calcium channel blockers, and other medications that may influence the autonomic nervous system. However, exclusion of subjects taking these medications was not feasible within the available subject pool, since the majority of participants in this study are being treated with these medications. Instead, these variables are statistically examined. Unfortunately, the grip-strength task introduced substantial artifacts in the ECG data in several subjects. Therefore, only 18 subjects with both imaging and quantifiable ECG data were used for this condition and 23 for the other two conditions. The Institutional Review Board (IRB) at the University of Illinois approved all study procedures, and written informed consent was obtained from all patients.

Neuroradiological evaluation

All subjects underwent magnetic resonance imaging (MRI) examination according to the following protocol:

Image acquisition

Axial scout images using a short TR, short TE (T1-weighted) gradient-echo pulse sequence; T1 weighted 3D SPGR (124 slices, coronal acquisition, in-plane acquisition matrix=256×192, slice thickness=1.6 mm, gap=0, FOV=22×22, TR/TE=34/7, flip angle=35 degrees, 1 NEX); and T2 weighted 3D SPGR (coronal acquisition, in-plane acquisition matrix=256×144, slice thickness=1.5 mm, gap=0, FOV=26×19.5, TR/TE= 2625/110, ETL:32, bandwidth 62.5 kHz, flip angle=35 degrees, 2 NEX) were acquired on a GE 1.5 Tesla imaging system (Signa, General Electric Medical Systems, Milwaukee, WI) with XL high-speed gradients (Rev 10) in a single session.

Diffusion-tensor weighted imaging scans included an axial diffusion weighted single shot spin echo, echo planar scan (FOV = 24 cm, 128 × 128, zero filled to 256 × 256, 19 slices, 6 mm thick, 0 gap) acquired at the same locations as the in-plane structural images. Two b-values were used, b = 0 s/mm² and a high b-value (b = 800 s/mm²). The high b-value was obtained by applying gradients along two axes simultaneously according to the following pattern that yield measurements along six non-collinear directions: (x, y, z) = [(1, 1, 0), (0, 1, 1), (1, 0, 1), (−1, 1, 0), (0, −1, 1), (1, 0, −1)]. This was repeated four times for each slice, with the sign of all of the gradients inverted for two of the repetitions. The magnitude images were averaged prior to calculation of the apparent diffusion coefficients (ADCs); this approach eliminates cross-terms with imaging gradients. An additional set of inversion recovery images with cerebrospinal fluid nulling (TI ~ 2100 ms) was acquired for each slice with b = 0 s/mm². These images were used to unwrap the eddy current effect of the diffusion gradients in the diffusion-weighted images prior to ADC calculations. From the averaged and unwarped diffusion weighted images, six apparent diffusion coefficients (ADC) were calculated, from which the six independent elements of the diffusion tensor were determined for each voxel. As an estimate of white matter integrity, a fractional anisotropy (FA) value was calculated as follows: From the diffusion tensor in each voxel, three eigenvectors were derived, defining the direction of the diffusion system, with the corresponding eigenvalues. Based on the three eigenvalues and the mean eigenvalue, the anisotropy was measured as FA, yielding values between 0 and 1.

Region of interest derivation

Region of interest (ROI) masks were generated for the frontal, parietal, temporal, and occipital lobes, and also the left and right hemispheres, using the Wake Forest University Pick Atlas plugin for SPM2 (Statistical Parametric Mapping). The Wake Forest Pick Atlas (Maldjian et al., 2003) is a tool for creating ROI masks using normalized brain template regions. The atlas has been successfully used in a diverse range of region of interest studies (e.g., Woolley et al., 2007; Newman et al., 2007; Shirley et al., 2007). These masks were then applied to each image using an automated MatLab (Mathworks, 2009) script to generate the regional FA values. Each mask was visually inspected to ensure accurate regional representation.

Lesion identification

Ischemic strokes were identified by the study neuroradiologist and a neuroradiological fellow working under her supervision, consistent with the Cardiovascular Health Study (Manolio et al., 1994) criteria for image analysis. The fellow was trained to a standard reliability (i.e., >.8 convergence) target on a practice dataset prior to evaluation of the study data to match the neuroradiologist. Images consisted of T1 and T2- weighted images reformatted to 6 mm slices reconstructed in the axial plane and standardized to the AC-PC axis. Strokes were defined as lesions greater than 3 mm in maximum antero-posterior or lateral dimension. Clear boundaries as indicated by hyperintensities on T2- weighted images when involving the cortex or deep nuclear regions (basal ganglia and thalamus), and hypo-intensities on T1 weighted images when involving the white matter, and with no mass effect while also presenting in a definable vascular distribution were required. Further, lesions were coded for location (cortical, subcortical white matter, deep gray structures, infratentorial), side (right, left, bilateral, midline), and size.

Stroke volumes

Ischemic strokes were outlined based on involvement in T2-weighted images (anatomical determination referenced to T1-weighted images). Areas were calculated, multiplied by slice thickness, and summed over consecutive involved slices, to yield a volume measurement for each identified lesion. Lesion volumes were then summed to arrive at a total stroke volume for each patient. The Analyze software package was used to do tracings and generate volumes.

Autonomic Nervous System Monitoring

Heart rate data were collected using a three-lead ECG configuration. The ECG signal was sampled at 1 kHz. Offline software detected the R-wave in sequential heart beats from the ECG to generate a series of heart periods defined by R-R intervals in msec. The heart period data were visually inspected and edited to identify missed and faulty R-wave detections and to remove the effect of ventricular arrhythmias that might confound the quantification of periodic components due to atrial rhythms in the beat-to-beat heart rate pattern (e.g., RSA). Editing consisted of integer arithmetic (e.g., dividing intervals when detections were missed or adding intervals when spuriously invalid detections occurred). These data were used to calculate the heart rate variability statistic, respiratory sinus arrhythmia (RSA). Editing and heart rate variability quantification were performed with Cardioedit and Cardiobatch programs (Brain-Body Center, University of Illinois at Chicago). The algorithm included in CardioBatch for the calculation of RSA is based on the methods developed to quantify the amplitude of heart rate variability (see Porges, 1985; Porges & Bohrer, 1990; Denver, Reed, & Porges, 2007). The quantitative procedures include: 1. R-R intervals derived from the ECG are timed to the nearest millisecond to produce a time series of sequential heart periods; 2.

Sequential heart periods are re-sampled into 500 ms intervals to produce equal interval time-based data; 3.The time-based series is detrended by a 21-pont cubic (to quantify RSA) moving polynomial filter (Porges & Bohrer, 1990). The polynomial is stepped through the data to create a smoothed template that is subtracted from the original time-based series to generate a detrended residual series free of baseline drift; 4. The detrended time series is band-passed to extract the variance in the heart rate pattern associated with spontaneous breathing for RSA (i.e., 0.12-0.40 Hz). The natural logarithm of the variance of the band-passed time series is calculated as the measure of the amplitude of RSA (Riniolo & Porges, 1997). The result is mathematically equivalent during steady state conditions to spectral analysis (see Denver, Reed, & Porges, 2007). This method is not moderated by respiration and conforms to the assumptions necessary for parametric statistics (Lewis et al., 2011).

Stimuli

Verbal fluency

The Controlled Oral Word Association Test (COWAT) assesses the oral or written production of words beginning with a designated letter (Benton & Hamsher, 1976). It consists of three trials in which, in this instance, participants are instructed to write as many words as possible in one minute beginning with a specified letter. Proper names, numbers, and the same word with different suffixes are not permitted. The letters F, S, and T were used. These letters were chosen based on the tendency of normals to produce an equal number of responses for each letter (approximately 11-12 words per minute). The COWAT has been used as a physiological stressor in previous research (Everhart & Harrison, 2002; Williamson & Harrison, 2003).

Nonverbal fluency

For nonverbal fluency the Ruff Figural Fluency Test (RFFT) (Ruff, 1988) was used. It consists of five parts, each containing different stimulus presentations. Each part has 35 dot matrices arranged in a 5 × 7 array. Participants were instructed to connect the dots in as many unique ways that they can conceive within a one-minute period. The first three sheets of the test protocol, instead of the standard five, were used in this experiment in order to maintain consistency with the COWAT. The RFFT has been used as a physiological stressor in previous research (Everhart & Harrison, 2002; Williamson & Harrison, 2003).

Grip Strength

A hand dynamometer [WW-1596-NP, Laffayette Instrument] was used as a physical stressor. There were three maximum strength trials at the right hand. Each trial required a sustained grip of thirty seconds. Order of administration was counterbalanced. Maximum grip strength was recorded at the start of each trial and sustained grip was recorded within 2 seconds prior of instructed release.

Experimental Procedure Condition 1 – Cognitive

Condition 1 began after a five minute rest period for acclimination to the testing environment. The participant was seated. A baseline physiological measurement of three minutes was taken. Immediately following the recording of the physiological data, the participant was asked to complete the RFFT or the COWAT using the appropriate instruction set for the test. Order of the fluency tests was administered in a counterbalanced fashion. Upon completion of each fluency test, physiological measurement was continued for three minutes. Physiological data was aggregated across the three RFFT and the COWAT conditions in order to provide three minutes of active condition data from which to derive heart-rate variability statistics.

Experimental Procedure Condition 2 – Grip Strength

Subsequent to condition 1, participants were asked to stand. Physiological data acquisition was continued. Participants stood for a three minute standing baseline. The grip strength protocol was implemented (see above). After each thirty-second grip period, participants stood for two minutes at rest while physiological data was collected (to measure recovery). The thirty-second grip strength trials were aggregated in order to have enough data from which to derive heart-rate variability statistics. This was done due to experimental constraints (not reasonable to expect sustained grip beyond 30 seconds).

Statistical Methods

For the cognitive variables, bivariate correlations were calculated to measure the relationship between whole brain white matter and the regions of interest to autonomic baseline, task-dependent (cognition and exertion), and recovery stages. Difference scores¹ were calculated for the task dependent (baseline subtraction) and recovery (task-dependent subtraction) stages using average change data over all trials (3 trials verbal fluency, 3 trials nonverbal fluency) for the cognitive variables. Multivariate, hierarchical stepwise linear regressions accounting for stroke variables (forced into the model: stroke volume, total number of strokes) were calculated for white matter regions with significant correlations at each stage for the two cognitive tasks. For the grip-strength task, because of the 30 second trials, a repeated measures analysis was calculated using 3 levels of 30 second epochs (before, during, and after) and co-varying the white matter variables of interest.

Results

Patient Characteristics\demographics

Sample characteristics including age, sex and education, measures of functional impairment including the Stroke Severity Score and MMSE, medication status, and stroke characteristics including total number of strokes, stroke volumes, and stroke locations are reported in table 1.

Table 1.

Sample characteristic

	Seated Baseline 1	Seated Baseline 2	Standing Baseline
HR (X, SD)	66.5 ± 11.2	65.73 ± 11.9	72.6 ± 12.4
RSA (X, SD)	4.9 ± 2.1	4.5 ± 2.0	3.7 ± 1.5

Note scores of “0” for total # of strokes are an artifact of the diagnostic criteria (stroke databank criteria, [13] vs Cardiovascular Health Study criteria, [18]) due to the required lesion size

Baseline Conditions

Across all baseline conditions, one before each experimental condition, left hemisphere fractional anisotropy was positively related to heart rate (baseline 1 - seated, r = .485, p < .05, baseline 2 - seated, r = .433, p= <.05 baseline 3 – standing - , r = .470, p <.05) (see table 2 for baseline HR and RSA averages). No relationships to other regions were noted. There were no significant correlations between fractional anisotropy and RSA at baseline. Demographic and health status factors were not related to the autonomic variables (i.e., sex, age, medication status).

Table 2.

Baseline HR and ROI correlations

ROI Baseline HR	Whole Brain	Frontal	Temporal	Parietal	Left	Right
Standing	.293	.328	.375	.427	.453*	.444
Seated 1	.362	.334	.348	.321	.467*	.339
Seated 2	.360	.286	.387	.327	.415*	.381

^*p <.05

Multivariate hierarchical linear regressions accounting for stroke variables in the first step (stroke volume and total number of strokes) and left hemisphere FA in the second, revealed that left hemisphere FA contributed independent variance to both standing and seated baseline heart rates (Standing baseline heart rate, B = 1.027, p < .05, OR = .100 – 2.053; combined seated baseline heart rate, B = .207, p = <.05, OR = .145 – 2.297). Higher left hemisphere FA is related to a faster baseline heart rate. Stroke volume and total number of strokes were not related to autonomic behavior.

Further, to directly compare the impacts of left hemisphere and right hemisphere FA on baseline heart rate, two hierarchical linear regression sets were calculated. In the first, left hemisphere FA was entered in the first step and right hemisphere FA in the second step, with the three baselines as dependent variables in separate analyses. In all three, the left hemisphere model was a significant predictor of baseline heart rate whereas adding the right hemisphere diminished the predictive power of the model (no longer significantly predicted heart rate even with the left hemisphere FA in the model) (seated baseline 1 model 1, F (1, 21) = 5.883, p < .05, model 2, F (2, 21) = 2.881, p = >.05; seated baseline 2 model 1, F(1, 21) = 5.581, p <.05, model 2, F(2, 21) = 3.070, p > .05; standing baseline model 1, F(1, 21) = 4.648, p < .05, model 2, F(2, 21) = 2.273, p = .13). In the second set of hierarchical regressions, right hemisphere FA was entered in the first step and left hemisphere FA in the second step. In all three baselines, neither step was a significant model (seated baseline 1 model 1, F (1, 21) = 2.600, p = .123, model 2, F (2, 21) = 3.070, p = .07; seated baseline 2 model 1, F(1, 21) = 3.558, p = .07, model 2, F(2, 21) = 2.099, p = .149; standing baseline model 1, F(1, 19) = 4.411, p = .05, model 2, F(2, 19) = 2.273, p = .13.

Task-dependent Conditions

Nonverbal Fluency

Bivariate correlations revealed no significant relationships between DTI-FA whole brain and other regions and autonomic response to the nonverbal fluency (Ruff) task. Medication status, other stroke variables, and demographic factors were not related to white matter integrity or ANS factors (see Table 3).

Table 3.

Fluency RSA and ROI correlations

ROI Fluency RSA	Whole Brain	Frontal	Temporal	Parietal	Left	Right
Nonverbal	−.209	−.077	−.052	−.048	−.130	−.110
Verbal	.117	.408	.434*	.580*	.486*	.408

^*p <.05

Verbal Fluency

Bivariate correlations revealed significant relations between several white matter regions and RSA change in response to verbal fluency. Fractional anisotropy of the left hemisphere's (r = .486, p<.05), parietal (r = .547, p<.05), and temporal (r = .434, p <.05) regions all related to RSA change (see table 2). The parietal region demonstrated the strongest relationship (r = .580, p <.05). (see Table 3). Trends, but no significant relationships, among the FA variables and HR were also observed. Demographic factors, medication status, and other stroke variables did not correlate with autonomic changes, though it should be noted that power is a limiting factor in evaluating these relationships.

Multivariate hierarchical linear regressions accounting for stroke variables in the first step (stroke volume and total number of strokes) and parietal FA in the second step revealed that parietal FA contributed independent variance to RSA response during verbal fluency (B = 22.055, p < 0.05). (See Figure 1). Further, the same analysis with left hemisphere FA (in lieu of parietal FA), revealed a laterality effect (B = 14.563, p < 0.05) (no relationship to right hemisphere FA). To clarify the laterality effect, similar to the baseline heart rate models, two additional hierarchical linear regression sets were calculated. In the first, left hemisphere FA was entered in the first step and right hemisphere FA in the second step. In the second, right hemisphere FA was entered in the first step and left hemisphere in the second. RSA response during verbal fluency was the dependent variable. In the first analysis, the left hemisphere model is a significant predictor of RSA response, F(1, 20) = 5.883, p < 0.05. Including the right hemisphere FA in the 2^nd stop decreases the predictive power of the model, F (2, 20) = 2.881, p > 0.05. Entering the right hemisphere FA first yields two non-predictive models, with model 1 yielding, F (1, 20) = 3.784, p > 0.05.

Figure 1.

Scatter plot of RSA change as a function of parietal lobe DTI-FA during the verbal fluency task.

Grip Strength

Within-subject, repeated measures ANOVA of autonomic performance in light of frontal lobe white matter integrity revealed significant differences in RSA from rest, during grip, and to recovery, F(2, 13) = 4.243, p < 0.05, and in heart rate, F(2,13) = 4.154, p < 0.05 such that with lower DTI-FA in the frontal lobes, RSA decreased more during this task and also showed a greater decrease in heart rate during task. Higher-DTI-FA in the temporal lobes yielded a greater suppression of RSA during grip (decrease in RSA), F(2, 13) = 3.946, P <0.05. We see a mirror of the frontal lobe effects in the parietal lobes with greater RSA suppression with lower DTI-FA, F(2, 18) = 3.528, p < 0.05. RSA does not violate Mauchly's test of sphericity, thus sphericity was assumed in determining probability values.

Recovery

No significant relationships between apparent white matter integrity and autonomic response to recovery (after task return to baseline) minus (baseline) were observed.

Discussion

The primary findings of this study are that a loss of white matter integrity is associated with a disordered task-dependent autonomic response, and that effect is regional and specific to vagal pathways dynamically regulating heart rate, as assessed by quantifying RSA. Consistent with literature suggesting that the left-hemisphere has preferential control of the parasympathetic nervous system (e.g., Wittling et al., 1998; Oppenheimer et al., 1992), deterioration of the left-hemisphere's white matter integrity (as measured by DTI-FA) is associated with RSA changes on cognitive tasks that recruit left-hemisphere resources, such as letter-phoneme fluency but not on cognitive tasks that, arguably, recruit right hemisphere resources such as design fluency (e.g., Williamson and Harrison, 2003). Further, parietal lobe DTI-FA was the strongest predictor of RSA response to verbal fluency. This is consistent with previous work from our lab demonstrating parietal FA sensitivity to general cognitive performance in a similar sample (Williamson et al., 2010). ). As we did not directly monitor the activity of the sympathetic nervous system or measure the changes in the systems innervated by this system we cannot draw any conclusions about how white matter integrity influences sympathetic activity. .

Baseline heart rate, but not RSA, was related to left hemisphere white matter integrity. Better white matter integrity was associated with a faster baseline heart rate. Though we would normally consider a slower heart rate as indicative of greater parasympathetic influence, perhaps this relationship reflects a difference in the calibration of “healthy” in this population. Faster heart rate may reflect a greater flow of oxygenated blood to the brain that would functionally enable greater mobilization and health.

With the motor stressor, grip-strength, a commonly used autonomic stress test, we saw an ANS pattern that is consistent with our understanding of the action of underlying regional systems of the central autonomic network (Loewy, 1991). Participants with deteriorated frontal-lobe white matter integrity demonstrated a disinhibited profile with a larger drop of RSA during grip (recall, normally, activity in the prefrontal cortex is correlated with decreased sympathetic nervous system response). In contrast, with deteriorated temporal lobe FA there is no drop in RSA during grip, and with higher DTI-FA in the temporal region, we observed a healthy appearing decrease in RSA during grip.

The functional neuroanatomy of cortical and subcortical inputs interacting with the vagal regulation of the heart has not been entirely elucidated. Research on telencephalic contributions to control of the ANS suggests hemispheric asymmetries. However, the literature on how exactly the systems are lateralized is mixed. Several researchers posit that ANS control is primarily concentrated in the right hemisphere (e.g., Heller, 1993, Porges, 1994). Perhaps, this is because much of the interest in brain correlates of autonomic change has theoretically emerged from the pioneering works of Charles Darwin and William James, who postulated a linkage between emotion and physiological arousal. Theorists have argued for right hemisphere dominance in the processing of emotion (e.g., Code, 1986; Heilman & Gilmore, 1998) and damage to the right hemisphere is associated with clinical syndromes that affect emotional processing, including anosognosia, anosodiaphoria, and aprosodia. Porges (1994) suggests that since the primary vagal access to the sino-atrial node (i.e., the pacemaker of the heart) is mediated by the right vagus nerve, the neuroanatomy of cortical, subcortical, and brain stem nuclei as they pertain to the primary vagal access suggest a right hemisphere dominance, regardless of emotional involvement. Although the vagus, the primary modulator of parasympathetic changes in cardiovascular function, is bilateral, as are the output nuclei (nucleus ambiguus and the dorsal motor nucleus) to the vagus, the right vagus, due to relative influence on the sino-atrial node, modulates the rhythmic beat-to-beat heart rate changes. Further, there are ipsilateral connections directly from the amygdala to the output nuclei, and from the cortex to the amygdala, suggesting a clear anatomical pathway for right hemisphere parasympathetic influence.

In contrast, functional data, namely changes in autonomic patterns in laterality paradigms, suggest left hemisphere involvement in parasympathetic control of the ANS. The anatomic area where contralateral fibers from the left-hemisphere cross to influence right vagal output has not been fully defined, but several stimulation studies have suggested (e.g., Oppenheimer et al., 1992) that the left-hemisphere can alter vagal activity. For instance, electrical stimulation of the right insular cortex in humans results in tachycardia while stimulation of the left insular cortex results in bradycardia (Oppenheimer et al. 1992). In a study examining unilateral migraine, significant bradycardia and vasodilation were demonstrated with left sided migrainers versus right (Avnon et al., 2004).

Control of the sympathetic nervous system activity appears to be primarily mediated by the right hemisphere. ty (Wittling & Genzel, 1995; Heilman et al., 1978; Critchley et al. 2000). Perhaps, like spatial attention, both hemispheres are involved in an asymmetrical fashion in the processing of autonomic information. Further, there may be significant cross-talk as a function of task integration (e.g., listening to music involves substantial bilateral functional networks and has complex autonomic and cognitive effects). Alternatively, since fMRI cannot distinguish activation of inhibitory circuits from activation of excitatory circuits, the data may not be easily interpreted in building models of linear connectivity. In addition, neural feedback circuits may exist that are not necessarily the product of interhemispheric cross-communication. In the current study, it appears, when tasks that require left hemisphere resources are performed regardless of task-type, left hemisphere white matter integrity predicts the parasympathetic response. Though we did not measure sympathetic nervous system response, a future study analyzing mobilization of sympathetic nervous system resources in this population may be illuminating.

Cerebrovascular disease has a complex relationship with autonomic function. Hypertension is the strongest one risk factor for stroke and white matter disease. We know that autonomic reactivity to stress is also associated with later development of cardiovascular disease as well as with a variety of emotional states/traits including anxiety and hostility (e.g., Williamson and Harrison, 2003; Manuck & Krantz, 1986). Recent research has suggested a top-down element to the effects of cerebrovascular disease on autonomic control (e.g., white matter hyperintensities correlate with blood pressure variability in response to cognitive tasks, Waldstein et al., 2004). Consistent with our findings, there is a large literature reporting depressed RSA in individuals with hypertension (Masi et al., 2007).

In the current study, we observed task-dependent autonomic changes as a function of a loss in white matter integrity in a sample of patients with ischemic stroke. These results are suggestive that these effects are due to changes in top down control of the ANS. The changes in task-dependent RSA as a function of regional white matter integrity suggest that there may be a negative synergistic impact of the disease process with behavioral activation. In other words, white matter disease impacts cortical and subcortical networks essential for proper integration of the ANS with cognitive and physical resources. In our sample, poorer parasympathetic performance is associated with weaker grip strength output and thus this deterioration of autonomic control may not only lead to an acceleration of the disease process, but also induce alterations in behavior including alterations of emotions and mood, cognitive activities and motor performance. .

This study has several limitations. Because we used a cross-sectional design, we cannot make definitive conclusions about the progression of white matter disease, specific neural network involvements, and autonomic changes. Further, our sample size imposes some statistical limitations in that we are not able to compare specific lesion hypotheses, or run more complicated hierarchical regressions accounting for specific lesions in our evaluation of white matter effects due to the potential for type II error. Also, our physical stressor procedures were brief, introducing some additional technical limitations in our approach to the data. We kept them brief to reduce patient burden in our elderly sample, but perhaps a better approach to this would be to use a different physical stressors such as a treadmill or stationary bike in which we could get a longer autonomic in-task assessment.

In addition, the use of a 1.5T magnet and nine diffusion directions makes it impossible to analyze the impact of damage to the specific white matter tracts we suspect are involved in autonomic/behavioral integration, limiting us to a more global analysis. Further, by nature of the subject population of interest (stroke and cerebrovascular disease), this is not a healthy population. Medications, other diseases, and general frailty offer significant noise in which to evaluate the impacts of white matter changes on our variables of interest. Most subjects were prescribed at least one form of anti-hypertensive medication. Though we statistically evaluated whether or not medication status was related to white matter integrity and autonomic response, an ideal situation would be to evaluate patients with no cardio-active medications. Although it is known that certain antihypertensive medications, such as beta blockers, can influence heart rate, it is not entirely clear what the effects are of the various classes of blood pressure medications on heart rate variability components. In our clinical environment, a medication holiday would have been potentially dangerous as an experimental control. While we did not find any correlations with blood pressure medications and task-dependent measures of heart rate or heart-rate variability, pharmaceutical manipulations of blood pressure do influence autonomic regulation of the heart. However, due to the size and clinical heterogeneity of our sample and the variations in medication and dose, the study design precludes the opportunity to evaluate the specific impact of the blood pressure medications prescribed to our participants on our measures of heart rate and heart rate variability. Further, while some blood pressure reducing medications may influence baseline heart rate variability, there is no evidence that top-down mobilization of ANS control is affected by these medications. If there were substantial effects of blood pressure medications, we might expect them to wash out any results, rather than amplify them.

Future research examining the effects of white matter disease on specific tracts and how these tracts relate to ANS and behavioral changes may be valuable. Tracts of interest may include the uncinate fasciculus, the stria terminalis, and cortico-spinal tracts. The results of this research suggest that there is an asymmetry of cortical control of the ANS such that, with a loss of the integrity of the white matter in the left hemisphere, there is a deterioration of vagal response to cognitive stressors that recruit those resources. Subjects had both left and right hemisphere white matter integrity decrements, yet only the left hemisphere white matter integrity loss was associated with alterations in autonomic performance during a left hemisphere cognitive task. This supports the postulate that the left hemisphere is perhaps more critical with respect to aspects of parasympathetic nervous system control of the heart. Analysis of our grip strength data revealed some crude indicators of differential impact of white matter damage of regional functional systems on autonomic control. For example, fronto-parietal lobe white matter decrements are associated with a greater decrease in RSA during task performance and temporal lobe white matter decrements yield a smaller RSA decrease during task performance. While these results should be considered preliminary (small sample size, fragile population, etc. . .), the data suggest that neural regulation of the ANS is impacted by white matter damage. Alteration of white matter in a stroke population significantly attenuates task dependent vagal control, likely reflecting changes in neural networks necessary for either efferent or afferent integration of resources for cognitive and motor performance.

Figure 2.

Scatter plot of RSA change as a function of left hemisphere DTI-FA during the verbal fluency task.

Figure 3.

Regional DTI-FA relationships (low vs. high) to RSA progression through baseline, during grip, and recovery (each point represents a 30-s block).