Long-Term Blood-Brain Barrier Permeability Changes in Binswanger’s Disease

Abstract

Background and Purpose

The blood brain-barrier (BBB) is disrupted in small vessel disease (SVD) patients with lacunes and white matter hyperintensities (WMHs). The relationship of WMHs and regional BBB permeability changes has not been studied. We hypothesized that BBB disruption occurs in normal appearing WM (NAWM) and regions near the WMHs. To test the hypothesis, we repeated BBB permeability measurements in patients with extensive WMHs related to Binswanger’s disease (BD).

Methods

We selected a subset of 22 BD subjects from a well-characterized larger prospective vascular cognitive impairment cohort. We used 16 age-matched controls for comparison. The abnormal WM permeability (WMP) was measured twice over several years using dynamic contrast-enhanced MRI (DCEMRI). WMP maps were constructed from voxels above a predetermined threshold. Scans from first and second visits were co-registered. WM was divided into 3 regions: NAWM, WMH ring and WMH core. The ring was defined as 2mm on each side of the WMH border. WMP was calculated in each of the three specific regions. We used paired t-test, ANOVA and Fisher’s exact test to compare individual changes.

Results

WMP was significantly higher in subjects than controls (p<0.001). There was no correlation between WMH load and WMP. High permeability regions had minimal overlap between first and second scans. Nine percent of WMP was within the WMHs, 49% within the NAWM, and 52% within the WMH ring (p<0.001; ANOVA).

Conclusions

Increased BBB permeability in NAWM and close to the WMH borders supports a relationship between BBB disruption and development of WMHs.

INTRODUCTION

Cerebral small vessel disease (SVD) is the main cause of vascular cognitive impairment (VCI) and a major cause of ischemic stroke.¹ Increased permeability of the blood brain barrier (BBB) has been observed in mixed groups of SVD patients.² BBB disruption might play an important causal role in SVD, possibly through toxic effects of leaked fluid and blood-derived proteins within the white matter (WM).³ Plasma proteins, such as IgG, complement and fibrinogen, have been identified in the WM of patients with SVD.⁴ Also, there is an increase of the CSF/serum albumin ratio in patients with vascular dementia, which correlates with the degree of white matter hyperintensities (WMHs), suggesting that the abnormal WM might be the site of BBB leakage.⁵ More recently, magnetic resonance imaging (MRI) studies in patients, using dynamic contrast-enhanced MRI (DCEMRI), have generated permeability maps showing regions of increased WM permeability, and an association of WM BBB permeability with the load of WMHs.⁶ However, long-term BBB changes in WM and the association of WMHs with regions of altered BBB have not been studied longitudinally.

Binswanger’s disease (BD) is a progressive form of SVD in patients with chronic hypertension and/or other associated vascular risk factors that have MRIs characterized by large WMHs on fluid attenuated inversion recovery (FLAIR) and T2-weighted imaging sequences, cognitive impairment, mainly in executive function, gait ataxia, and focal neurological signs.⁷ These patients usually progress to dementia of the vascular type. We hypothesized that in patients with BD, the WM permeability relates to the WMHs load and that the BBB permeability increases as the disease progresses, preceding the growth of new WMHs. We postulate that BBB disruption occurs in different regions over time with increased permeability in normal appearing WM (NAWM) in regions near WMHs. To test these hypotheses, we used DCEMRI to generate longitudinal permeability maps to measure long-term changes of the WM permeability. We further compared regions of abnormally high permeability from the two scans and analyzed the distribution of these regions within the WMHs, normal appearing WM (NAWM) and newly formed WMHs.

METHODS

Subjects

From a prospective longitudinal VCI cohort of 95 subjects, we retrospectively selected a subset of 22 SVD subjects thought to have Binswanger-like features. VCI subjects had a clinical evaluation by a board certified neurologist, neuropsychological studies, brain MRI studies and CSF for demyelinating profile, CSF/albumin ratio, and matrix metalloproteinases (MMPs). The inclusion criteria for the SVD Binswanger-like subjects were: Two completed brain MRI with contrast more than 10 months apart, large WMHs (Fazekas >2), more than 2 vascular risk factors, cognitive complaints and focal neurological symptoms or gait disturbances. We excluded patients with cortical strokes, SVD due to genetic/toxic/metabolic causes and patients with suboptimal DCEMRI (motion artifact). We selected sixteen asymptomatic age-matched controls for comparison. In agreement with recently published literature, we retrospectively applied the Binswanger score to our patient selection.⁸ All patient selection was blinded to the permeability data. The University of New Mexico Human Research Review committee approved all aspects of the study.

Neuropsychological testing

Standardized measures of cognitive function were given to all patients in the study. All tests were administered and scored according to standard procedures for that test and were administered by a trained psychologist. Standardized (T) scores were calculated for each test using published norms for each test. Averaged composite T scores were calculated for each of the three cognitive areas of interest (memory, executive functioning, and processing speed) as well as an overall composite of cognitive functioning. Tests for each composite included: memory (Hopkins Verbal Learning Test-Delay, Rey Complex Figure Test-Long Delay), executive (Digit Span Backwards, Trail Making Test B, Wisconsin Card Sorting-Total errors or Stroop Color and Word Test-interference score), processing speed (two subtests from the WAIS-III-- Digit Symbol and Symbol Search). The overall composite included these three composites, plus attention (Digit Span Forward, and Trial Making Test A) and language (Boston Naming 60 item test, Controlled Oral Word Association (FAS)), averaged.

MRI acquisition

The study started on the Siemens 1.5T Sonata MRI scanner but is currently being done on the Siemens 3T Trio scanner. Fourteen BD patients were scanned twice on 1.5T MRI and 8 patients on 3.0T; all repeat studies were done on the same instrument used in the first scan. The measurements consist of a structural T1 scan, a FLAIR scan for characterizing WMHs, and a TAPIR scan for BBB permeability calculations.⁹ TAPIR uses sequential, rapidly acquired, axial T1 measurements with the first image taken before the Gadolinium diethylenetriaminepentacetate (Gd-DTPA; Magnevist, Bayer Schering Pharma) injection and multiple T1 measurements after injection. A quarter of the standard dose of Gd-DTPA is injected with an automatic power injector (Medrad Spectris SolarisVR MR injection system; Siemens). The scan parameters were similar at the two scanner field strengths with slightly higher resolution images at 3T and TAPIR images also having better temporal resolution. The scan parameters are summarized in the supplemental table.

Structural image analysis

The FLAIR images were segmented for WMHs, using a semiautomated software package (JIM V.6.0, Xinapse Systems Ltd, Northants, UK, http://www.xinapse.com). A single experienced physician blinded to the clinical information, used the semi-quantitative Fazekas scale for rating WMHs.¹⁰ The same rater used T1, T2 and FLAIR for grading perivascular spaces (PVS) and counting lacunar and cortical strokes for each MRI scan.¹¹

Permeability analysis

The voxel-by-voxel permeability maps were calculated from TAPIR measurements based on the Patlak model of Gd-contrast agent leaking through the BBB as described earlier.¹² The T1 images, the FLAIR image and the permeability maps from each visit were spatially registered to the FLAIR image of the first visit for evaluating longitudinal changes. Permeability was calculated only within WMH and the normally appearing white matter (NAWM) regions. The WM was defined by segmenting the T1 image by fast/FSL, followed by eroding the WM segment by 7 mm to avoid proximity with gray matter and the CSF. This gave a more consistent permeability calculation in the WM. The regions of active BBB leakage were defined by exceeding a threshold permeability of 0.003 min⁻¹. The permeability threshold was varied from 0.001 to 0.005 min⁻¹, and 0.003 min⁻¹ gave the maximum accuracy of correctly predicting controls and patients. The threshold was obtained from the data itself, because of the small sample size. WM permeability (WMP) is the sum of all permeability voxels over the active area of BBB leakage. WMP depends on the value of the cut-off threshold.

To understand the relationship of WMP and development of new WMHs we calculated the voxel overlap between the initial visit WMP map and the newly formed WMHs from visit 2 (WMH visit2 – WMH visit1). The overlap was reported as percentage of newly WMHs voxels with prior abnormal permeability.

We combined data from the first and second visits to study the spatial distribution of the WMP in relation to the WMHs. This was done because visual examination of the permeability maps indicated that regions of high permeability clustered at the boundary of the WMHs. In this analysis, we divided the WM into three ROIs: 1) NAWM, 2) 4 mm ring bordering the WMHs constructed by dilating and eroding the WMH mask by 2mm on each side of its border, and 3) the WMHs within the ring. We calculated the WMP for each of each of these regions.

Statistics

For group comparison of normally distributed data, we used Satterthwaite student’s t-test and paired t-test for long-term changes of individual variables. For group comparison binomial data, we applied Fisher test and ANOVA for repeated covariate analysis. Statistical analyses were conducted with SAS v12.1 and SPSS v20.0.

RESULTS

Twenty-two subjects with clinical features of BD and sixteen-age matched who meet inclusion and exclusion criteria completed brain scans and neuropsychological evaluation at both visits. Table 1 shows the baseline characteristics of patients and controls. Increased total WMP was seen in all BD subjects in both visits compared with controls (ANOVA; p = 0.01). The variance of WMP was significantly higher in patients than controls, p=0.002 (FIGURE 1). The median time between scans was 16.5 months (IQR: 12.5, 22.6). Repeated measures using months between scans as covariate did not alter the WMP (ANOVA p=0.74). WMP between visit 1 and visit 2 had individual fluctuations but did change in the group (paired t-test; p = 0.22). There was a significant increase of WMHs load (paired t-test, p<0.01), Fazekas score (paired t-test, p < 0.01) and number of PVS (paired t-test, p < 0.01) between the first and second brain MRI. We found no association between the WMP and WMHs load within the subjects (ANOVA p= 0.22). (FIGURE 2) There was a no significant increase of asymptomatic lacunes (paired t-test, p=0.07) between visit one and two. Neither WMP nor neuropsychological testing worsened over time (data not shown). In both visits, BD patients consistently displayed abnormal high permeability regions within the entire WM.

TABLE 1.

Comparison of clinical, MRI, cognitive, and CSF factors between BD subjects and controls.

Features	BD=22	Controls=12	p
Age	67±10	61±9.5	ns
Gender (f)	41%	44%	ns
Hypertension	86%	19%	<0.001
Diabetes	27%	0%	0.03
Focal neurological signs	68%	13%	0.001
Previous Stroke	50%	0%	0.001
Ataxia/imbalance	73%	6%	0.001
MRI Features:
Lacunes	82%	0%	<0.001
WMHs*	39472±2474	702±449	<0.001
Fazekas score	5.2±0.88	1.46±0.77	<0.001
Perivascular spaces	5.3±0.95	3.28±1.13	<0.001
CSF-albumin ratio†	7.16±2.6	5.6±2.8	0.04
Cognitive Evaluation:
Global Score	46.68±8.2	52.73±8.9	0.02
EXE	43.86±8.2	50.6±5.9	0.01
MEM	46.5±11.6	53.1±8.9	0.07
SPEED	48.29±10.4	55.6±9.3	0.03
BS‡	5.8±1.7	1.2±1.8	<0.001

^*Supratentorial white matter hyperintensities (WMHs) volume in voxels

^†Cerebrospinal fluid (CSF) from different age-matched controls receiving spinal surgery.

^‡Binswanger Score (BS) of >5 have a 78% likelihood of BD.⁸

FIGURE 1. WM permeability (WMP).

WM permeability from visit 1, visit 2 and controls. There are significant differences in WMP between controls versus BD subjects’ 1^st and 2^nd visits (ANOVA p<0.001), but no difference between patient visits. Box plots are shown with outliers as circles. WMP variances were higher in BD subjects than controls (ANOVA p=0.01)

FIGURE 2. WMHs versus WM permeability.

In subjects with BD, there was no association between WMHs load and WMP (ANOVA, p=0.22).

Control subjects showed very little permeability and when it was present, the pattern showed the permeability was scattered randomly around the WM. Figure 3 shows representative WMP maps in the age-match controls. Occasionally a small patch of periventricular WMH was seen in the controls.

FIGURE 3. Permeability maps in four representative control patients.

The normal controls had some areas of WMHs mainly in the periventricular regions (arrows). Green delineates the WMHs, The red areas of scattered increased permeability were lower than found in the BD patients. None of the controls had regions of high permeabillty (yellow). The color code used showed increasing permeability from red to yellow.

The overlap of newly formed WMHs with the WMP map from visit 1 showed 11% of new WMHs voxels had prior abnormal permeability. Similarly, WMHs showed 14% of WMHs voxels with abnormal permeability. WMP map in visit 1 was compared with visit 2. Most of the regions showing permeability at visit 1 were gone in visit 2, and only 5% ±2 of the voxels with increased WMP overlapped between the two visits. The regions with increased permeability in visit 1 are shown in one color and those in the second visit in another color with the overlapped areas in a third color (FIGURE 4).

FIGURE 4. Permeability maps from visit 1 and 2.

Examples of 5 BD patients (rows) measured twice over a one-year period showing the different permeability areas from scan in visit one to scan in visit two. WMHs are shown in green, high permeability areas in red (visit 1) and blue (visit 2) Column A: FLAIR, Column B: WMHs and permeability at visit 1, Column C: WMHs and permeability at visit 2, Column D: Combined high permeability maps from visit 1 (red) and 2 (blue). Average overlapped areas represented only 5% on the permeability maps (in yellow).

After visual inspection of the WMP regions in scans from visit 1 and 2, we observed that the increased permeability seemed to cluster around the edges of the WMHs. To explain the location of the permeability in relation to the WMHs, we drew a series of new ROIs that consisted of a 4mm ring, which was traced along all WMH borders. An illustrative BD patient is shown with a rim around the WMHs and the permeability superimposed on the FLAIR image (FIGURE 5A). The 4mm rim is drawn in the enlargement of one of the WMHs with the permeability shown in red (FIGURE 5B). Plots of individual subjects are shown to indicate the rationale for the selection of the rim size. The exponential slope shows the grouping of high permeability voxels within the 4mm ring. We observed a clustering of the high permeability voxels close to the border of WMHs (FIGURE 5C). Using the permeability maps from the 3.0 Tesla patients only, we found that 51% of total permeability voxels were located inside the 4mm ring, 9% were located within the core of the WMHs, and 49% were located within the NAWM (ANOVA, p<0.001).

FIGURE 5. WMH border and ring.

A: FLAIR MRI shows the WMHs outlined with a green band to show the construction of the rim around the WMHs. B: Higher magnification of one of the traced WMHs shows the border of the WMH (black line) and the 2mm on both sides of the WMH border (green rim). C: Fraction of high permeability voxels from WMH border. Individual lines represent individual subjects. Arrow indicates border of WMHs. Rim extends 2mm on both sides. Exponential slope is observed close to WMH borders due to grouping of high permeability voxels on the inner part of the ring. The average value of all patients shown as a red line.

DISCUSSION

We found that BD patients have increased BBB permeability, which remains elevated after one to two years. Sites of abnormal BBB permeability on the first scan tended to resolve with new areas arising in the second scan. The volume of WMHs failed to correlate with the WMP. Comparing voxel permeability maps between the two scans showed very little overlap, suggesting a continuous, but fluctuating, pattern of BBB disruption. The small variability in WMP from controls compared with the high variability in BD patients argues against these changes being spurious and suggests that WMP fluctuations arise from biological processes

A unique aspect of this study is the ability to follow the regions of high permeability with voxel mapping. The initial WMP scan did not predict future WMH formation mostly due to the small overlap between initial permeability and the new WMHs. We also observed that regions of permeability in the two scan rarely overlapped. Original high permeability regions resolved over time with new regions of leakage appearing in other locations. Such variability in permeability might account for the modest overlap found between the first WMP map and second scan.

Since these observed spatial fluctuations in permeability could be attributed in part to co-registration errors, poor spatial resolution or reproducibility of the imaging technique, we further studied the spatial relationship between WMHs and the BBB permeability maps from each individual scan. By tracing a ring around the WMHs borders, we demonstrated that most high permeability voxels were located either in NAWM or the surrounding edges of the WMHs with only a few voxel seen inside the core of the WMHs. Our results strongly suggest that the WMHs borders obtained on FLAIR do not dictate absolute permeability changes in the WM. Instead, it seems that there is a gradual transition of high WM permeability from outside of the WMHs core to NAWM regions. These findings are of special interest for understanding the disease progression. Studied using fractional anisotropy (FA), an MRI method that measures the WM microstructure, has shown that a penumbral region of reduced FA surrounds WMHs.¹³ Long term follow up showed that baseline abnormalities in FA predicted future WMHs formation.^{14, 15} These FA changes are larger within regions of enlarging WMHs than in other NAWM regions.¹⁶ Our findings that WM BBB disruption appears to cluster around WMHs might represent another indicator of disease activity associated with local WMHs growth. Defective WM BBB might be an earlier phenomenon that indicates endothelial dysfunction followed by WM tract abnormalities. This initial BBB disruption in the WM has been observed in the hypertensive stroke prone rat animal model before the appearance of subcortical ischemic changes.¹⁷

Lacunar infarct is another form of SVD responsible for 20% of all ischemic strokes and it is strongly associated with the presence of WMHs.¹⁸ There is mounting evidence indicating that the NAWM surrounding the visible WMHs on FLAIR correspond to regions of future WMHs growth and new lacunar stroke occurrence. This has been observed also in patients with cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), a genetic form of SVD.¹⁹ New small DWI lesions occurring at the borders of the WMHs have also been postulated to be related to WMHs growth.²⁰ BBB disruption has been observed in lacunar stroke patients but it is unknown if it precedes the occurrence of new lacunar infarcts.² Patients with BD typically have lacunar infarcts mainly discovered incidentally after brain imaging. However, lacunar strokes can also occur in absence of WMHs, perhaps due to other pathophysiological mechanisms such as intracranial atheromatous plaque or microemboli.²¹ In these instances the earlier BBB disruption might not play an important role. Therefore it is possible that lacunar strokes have different pathophysiologic mechanisms. Future studies using DCEMRI and permeability maps could differentiate the pathophysiologic mechanisms of lacunar strokes.

Our patients have larger WMHs and it might be that these lesions, which represent gliosis, no longer have inflammation related to changes in permeability. Since both DCEMRI and FA demonstrate abnormalities in the NAWM in patients with large WMHs, future use of the combination of these two methods needs to be assessed as possible imaging biomarkers for interventional trials.

We also observed an increase of WMP in NAWM distant from the WMHs. Using a different BBB MRI technique; another group has shown in a mixed SVD group of patients an increase of BBB permeability in the NAWM that was associated with the WMH load.²² Our results are consistent with these observations, but we were unable to show an association between the load of WMHs and WMP. Our finding supports the notion that WMP is a migrating process while WMH load is an accumulating process.

Based on the alterations of BBB in the NAWM, and the greater density of high permeability voxels in the rim surrounding the WMHs, we postulate that areas of NAWM with increased permeability are regions vulnerable to growth of WMHs. Furthermore, we found pathological BBB changes occurring in the white matter of BD patients in both the NAWM and around the edges of the WMHs, suggesting a diffuse pathological process. Strengths of our study include the first demonstration of regional visualization of the WM permeability maps. There are several caveats of our study, including small sample size, retrospective design, different scanner strength, suboptimal spatial resolution and lack of autopsy diagnostic verification.

In summary, our study reveals that patients with BD have a persistent disruption of the BBB that fluctuates between WM regions over time. The increased number of high permeability voxels found in regions around WMHs strengthens the association of BBB disruption with the development of WMHs, but further studies will be needed for confirmation.