Heterogeneous blood‐brain barrier dysfunction in cerebral small vessel diseases

Abstract

INTRODUCTION

We explored how blood‐brain barrier (BBB) leakage rate of gadolinium chelates (K^trans) and BBB water exchange rate (k_w) varied in cerebral small vessel disease (cSVD) subtypes.

METHODS

Thirty sporadic cSVD, 40 cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), and 13 high‐temperature requirement factor A serine peptidase 1 (HTRA) ‐related cSVD subjects were investigated parallel to 40 healthy individuals. Subjects underwent clinical, cognitive, and MRI assessment.

RESULTS

In CADASIL, no difference in K^trans, but lower k_w was observed in multiple brain regions. In sporadic cSVD, no difference in k_w, but higher K^trans was found in the whole brain and normal‐appearing white matter. In HTRA1‐related cSVD, both higher K^trans in the whole brain and lower k_w in multiple brain regions were observed. In each patient group, the altered BBB measures were correlated with lesion burden or clinical severity.

DISCUSSION

In cSVD subtypes, distinct alterations of k_w and K^trans were observed. The combination of K^trans and k_w can depict the heterogeneous BBB dysfunction.

Highlights

• We measured BBB leakage to gadolinium‐based contrast agent (K^trans) and water exchange rate (k_w) across BBB in three subtypes of cSVD.

• CADASIL is characterized by lower k_w, HTRA1‐related cSVD exhibits both higher K^trans and lower k_w, while sporadic cSVD is distinguished by higher K^trans.

• There are distinct alterations in k_w and K^trans among subtypes of cSVD, indicating the heterogeneous nature of BBB dysfunction.

1. BACKGROUND

Cerebral small vessel diseases (cSVD) are highly prevalent and represent a common cause of ischemic and hemorrhagic strokes in the elderly, and vascular dementia. ¹ The cSVD represent a heterogeneous group of diseases of various etiologies and underlying mechanisms. The majority of cSVD cases are sporadic and primarily associated with age and hypertension. ² In addition, a minority of cSVD patients have a monogenic origin. The two most common hereditary subtypes are cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) and the high‐temperature requirement factor A serine peptidase 1 (HTRA1) ‐related cSVD ²

The pathophysiology of cSVD remains poorly understood. A prevailing view is that blood‐brain barrier (BBB) dysfunction could play a vital role in the pathogenesis of cSVD. ¹ , ³ , ⁴ BBB tightly regulates molecular transport between the bloodstream and the cerebral tissue to maintain a stable microenvironment. Severe BBB alterations during the course of cSVD presumably lead to leakage of plasma or cellular components from blood vessels, causing cerebral microvascular damage, brain tissue edema, and neuroinflammation, that finally can disrupt the balance of the microenvironment in brain. ¹ The compromised BBB also affects cerebral blood flow (CBF) regulation and hinders the removal of metabolic byproducts. ⁵ These pathophysiological mechanisms may contribute to the onset and progression of cSVD‐related tissue lesions.

Accumulating evidence suggests a correlation between cSVD severity and alterations in BBB integrity, as observed using dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI), a widely used technique for assessing the BBB leakage rate of gadolinium chelates from blood to brain (K^trans) in vivo under the condition of BBB structural disruption. Recently, a noninvasive diffusion‐prepared pseudocontinuous arterial spin labeling (ASL) imaging (DP‐pCASL) has been validated for measuring water exchange rate across the BBB (k_w), which reflects the combination of alterations of active transport and structural disruption. Recent data, partly from our group, suggest that k_w would be lower in CADASIL patients as well as in symptomatic HTRA1‐mutation carriers. ⁶ , ⁷ Walsh et al. recently reported a higher BBB influx rate (K_i) of gadolinium‐based contrast agents (GBCAs) in sporadic cSVD, which was not found in CADASIL. ⁸ Measures of both k_w and K^trans have been previously investigated only in a group of elderly individuals (n = 16). ⁹

Considering the diverse etiologies and vascular pathologies observed in subtypes of cSVD, alongside the intricate structural complexity of the BBB, it is plausible to anticipate the significant heterogeneity in the manifestation of BBB dysfunction. We hypothesized that the metrics of DCE‐MRI and DP‐pCASL could vary not only among patients with cSVD but also across different cSVD subtypes. For this purpose, we measured both K^trans and k_w data to explore BBB dysfunction in healthy individuals and in patients with sporadic cSVD, CADASIL, and HTRA1‐related cSVD.

2. METHODS

2.1. Participants

Four groups of subjects were included in the present study: (1) patients with CADASIL, (2) patients with HTRA1‐related cSVD, (3) patients with sporadic cSVD, and (4) healthy individuals (control group). In the present study's participants, 23 of the 40 CADASIL, 8 of the 13 HTRA1‐related cSVD, and 23 of the 40 healthy controls were also participants in our previous study. ⁶

CADASIL patients were included based on a confirmed genetic diagnosis of a typical cysteine mutation in the NOTCH3 gene and age ≥ 18 years. HTRA1‐related cSVD patients were included after a genetic test showing a characteristic heterozygous or homozygous mutation of HTRA1. Sporadic cSVD patients were individuals who had a recent lacunar stroke confirmed by an acute small subcortical ischemic lesion on diffusion‐weighted imaging or presented with vascular cognitive impairment (VCI) presumably related to cSVD lesions. More specifically, VCI was defined as a cognitive impairment in at least one cognitive domain at neuropsychological testing and cSVD lesions should be moderate to severe white matter hyperintensities (WMH; Fazekas score deep ≥ 2 or periventricular = 3) or mild WMH (Fazekas score deep = 1 or periventricular = 2) combined with lacunes or microbleeds ¹⁰ ). Any individual presenting with a suggestive feature of monogenic cSVD (WMH in temporal lobes) or with a family history of stroke, dementia or motor disability was excluded. When they were aged below 60 years, ¹¹ whole exome sequencing (WES) was conducted to exclude any already identified cause of hereditary cSVD. The healthy control group comprised subjects with no history of stroke or other major neurological disorders. They were recruited from both the family of hereditary cSVD patients having a negative genetic test or within the community. All individuals who were presented with a contraindication to MRI or an eGFR (estimated glomerular filtration rate) ≤ 59 ml/min/1.73m² were excluded from the study.

The following demographic and clinical data were collected in each subject: age, sex, years of education, vascular risk factors (hypertension, diabetes, hyperlipidemia, and smoking history), and medical history (history of acute ischemic stroke and intracerebral hemorrhage). Disability was assessed using the modified Rankin Scale (mRS) and significant disability was considered when the mRS score ≥ 2. ¹² Global cognition was assessed using the Montreal Cognitive Assessment (MoCA).

This study was approved by the Institutional Review Board and Ethics Committee at Huashan Hospital (KY2021‐561) and Beijing Chaoyang Hospital (2021‐52). Written informed consent was obtained from each participant.

RESEARCH IN CONTEXT

1. Systematic review: We reviewed the literature using traditional resources (e.g., PubMed). Dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI) and the technique based on arterial spin labelling (ASL) are the two main MRI techniques for blood‐brain barrier (BBB) assessment and could measure complementary characteristics of the BBB. Few studies have directly compared the two techniques among cerebral small vessel disease (cSVD) subtypes.

2. Interpretation: Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) is characterized by lower k_w, sporadic cSVD exhibits higher K^trans, while high‐temperature requirement factor a serine peptidase 1 (HTRA1) ‐related cSVD is distinguished by lower k_w and higher K^trans. Our results indicate the heterogeneous nature of BBB dysfunction in cSVD.

3. Future directions: Further animal models and pathological studies are needed to investigate the potential mechanisms underlying the alterations of the two BBB measurements. Longitudinal studies are also required to explore how k_w and K^trans vary across different stages of diseases, as well as determine whether k_w and K^trans at baseline could predict disease progression.

2.2. MRI protocol

Participants underwent brain MR scan on a Siemens 3T Prisma MRI system (Erlangen, Germany) with a 64‐channel head coli. Structural brain MRI sequences included: (i) T1‐weighted magnetization‐prepared rapid acquisition gradient echo (MPRAGE) images, resolution = 1 × 1 × 1 mm³, inversion time = 900 ms, echo time (TE) = 2.98 ms, repetition time (TR) = 2300 ms, (ii) T2‐weighted images, TE = 94 ms, TR = 4000 ms, flip angle (FA) = 150°, field of view (FOV) = 230 mm × 230 mm, matrix size = 320 × 230, (iii) T2‐weighted fluid attenuated inversion recovery (FLAIR) images, resolution = 1 × 1 × 1 mm³, inversion time = 1800 ms, TE = 388 ms, TR = 5000 ms, and (iv) susceptibility‐weighted images (SWI), TE = 20 ms, TR = 28 ms, FA = 15°, FOV = 220 mm × 220 mm, matrix size = 256 × 220.

Water exchange rate across BBB was measured by the DP‐pCASL sequence with background suppressed 3D gradient and spin‐echo (GRASE) readout, acquisition voxel size = 3.5 × 3.5 × 8 mm³, TE = 36.5 ms, TR = 4000 ms, FA = 120°, FOV = 224 mm × 224 mm, matrix size = 64 × 64, 12 slices (10% oversampling), label pulse duration = 1500 ms; centric ordering and background suppression timing were optimized for suppressing gray and WM signals to ∼10% of M0. A two‐stage approach was used to measure arterial transit time (ATT) and k_w: fifteen repetitions were acquired during the flow encoding arterial spin tagging (FEAST) scan at postlabeling delay (PLD) = 900 ms and diffusion weighting (b‐value) of 0 and 14 s/mm² with a total acquisition time of 4 minutes for estimating ATT. ¹³ The k_w metric was calculated from scans acquired at PLD = 1800 ms, when the labeled blood reached the microvascular compartment, with b = 0 and 50 s/mm², respectively. Twenty repetitions were acquired for each b‐value of the k_w scan, and the total acquisition time was 6 minutes.

Last, DCE‐MRI was acquired to assess the BBB leakage rate of GBCAs from blood to brain. DCE‐MRI consisted of a precontrast T1‐mapping protocol and a dynamic T1‐weighted gradient‐echo acquisition. T1 map was estimated from 3D fast low angle shot (FLASH) images with five different FAs (2°, 5°, 10°, 12°, and 15°). Imaging parameters for T1 mapping were: FOV = 175 × 175 mm², 14 oblique coronal slices through the  hippocampus and basal ganglia as well as carotid arteries for individual arterial input function (AIF), resolution = 1.1 × 1.1 × 5 mm³, TE = 2.18 ms, TR = 5.14 ms, matrix size = 320 × 320. Imaging parameters for dynamic T1w acquisition were: 3D FLASH with gradient and RF spoiling, FOV = 175 × 175 mm², 14 slices, resolution = 1.1 × 1.1 × 5 mm³, TE = 3 ms, TR = 8 ms, matrix size = 320 × 320, temporal resolution = 15 s, 64 frames were acquired with a total acquisition time of 16 minutes. It was acquired in a coronal direction to minimize the inflow effects of arterial input function. Contrast agent (Dotarem, Gadoterate meglumine, 0.5 mmol/mL) was injected after 30 s of image acquisition with an average injection rate of 3 mL/s followed by a 25 mL saline flush at a dose of 0.1 mmol/kg body weight. We described the MRI methods of DP‐pCASL and DCE‐MRI according to the DCE and ASL lexicons from Open Science Initiative for Perfusion Imaging (OSIPI). ¹⁴ , ¹⁵

FIGURE 1.

The flow chart of imaging processing. BBB water exchange rate (k_w) and BBB leakage rate of gadolinium‐based contrast agent (K^trans) were quantified for ROI analysis. BBB, blood‐brain barrier; DGM, deep gray matter; NAWM, normal‐appearing white matter; ROI, region of interest; WMH, white matter hyperintensity.

2.3. MRI analysis and ROI definition

Postprocessing of DP‐pCASL data was performed off‐line using LOFT BBB Toolbox (http://www.loft‐lab.org/index‐5.html). The details of processing procedure were described previously. ⁶ DCE‐MRI dynamic series were corrected for rigid head motion off‐line using SPM12 (Well‐come Trust Centre for Neuroimaging, UCL). The volume transfer constant, K^trans, measures the efflux rate of GBCAs extracted across the BBB. ¹⁶ Therefore, K^trans depends on BBB permeability and capillary surface area, while k_w is independent of surface area. We calculated the fractional plasma volume (Vp) to investigate whether it was not different between groups. We also generated the exchange rate of GBCA (k_Gad) as the ratio between K^trans and Vp: k_Gad = K^trans/Vp to further validate the findings of K^trans. The results of k_Gad are shown in Table S1. The Patlak model was fit to DCE‐MRI data to provide estimates of K^trans and Vp after noise filtering. ¹⁷ We used all the data during the Patlak model fit. The parameter bounds for K^trans were 1*10^‐7–2 and 0.001–1 for Vp. AIF was determined from each individual's internal carotid artery (ICA). We showed an example to demonstrate that the ICA was free from inflow effects in Figures S1–S3.

Average values of k_w, Vp and K^trans were measured in four regions of interest (ROIs) corresponding to the whole brain, deep gray matter (DGM), WMH, and normal‐appearing WM (NAWM), as shown in Figure 1. The whole brain and DGM (including basal ganglion and thalamus) ROIs were delineated using the Anatomical Labeling (AAL) Template in SPM12, and T1w images, as previously described. ⁷ WM was segmented on T1w images using SPM12 and eroded (2 mm) to avoid the same potential partial volume effects (PVE) from GM or CSF. WMH mask was automatically segmented on FLAIR images using the lesion segmentation tool toolbox. ¹⁸ WMH was extracted from the WM mask to obtain the NAWM mask. As WMH was scarce in healthy controls, WMH ROI was not defined in the control group. In this group, we obtained k_w/K^trans values in the NAWM by using the WMH mask of each patient group for the comparison of k_w/K^trans in WMH between patients and healthy controls. 

TABLE 1.

Demographics, vascular risk factors, medical history, and conventional MRI markers of cSVD in the three groups.

	Group				p‐Value
Parameter	CADASIL	HTRA1‐related cSVD	Sporadic cSVD	HC	CADASIL versus HC	HTRA1‐related cSVD versus HC	Sporadic cSVD versus HC
Number	40	13	30	40
Age, years	45.70 ± 12.65	40.62 ± 13.73	65.33 ± 7.53	42.70 ± 11.80	0.232	0.597	< 0.001
Sex, male	16/40	7/13	20/30	8/40	0.051	<0.001	< 0.001
Education, years	12 (9–16)	15 (6–15)	9 (6–12)	11 (9–18)	0.332	0.317	0.007
Vascular risk factors
Hypertension	18/40	2/13	21/30	1/40	<0.001	0.145	< 0.001
Diabetes	2/40	0/13	4/30	0/40	0.247	/	0.030
Hyperlipidemia	0/40	0/13	4/30	6/40	0.013	0.167	0.563
Smoking history	9/40	4/13	18/30	5/40	0.239	0.137	< 0.001
Medical history
AIS	15/40	6/13	18/30	0/40	<0.001	<0.001	< 0.001
ICH	2/40	0/13	1/30	0/40	0.247	/	0.429
MOCA	23 (12)	26 (8)	21.5 (5)	25.5 (7)	0.075	0.936	<0.001
Imaging markers
Lacune counts, n	2 (0–20.5)	9 (0–30)	1 (0.75–9.75)	0 (0–0)	<0.001	<0.001	< 0.001
WMH volume, mL	22.20 (5.66–34.28)	13.46 (2.44–23.38)	6.06 (2.12–19.56)	0.16 (0.03–0.38)	<0.001	<0.001	< 0.001
Relative WMH volume, %	1.4 (0.59–2.43)	1.31 (0.23–2.04)	0.51 (0.21–1.74)	0.02 (0.01–0.04)	<0.001	<0.001	<0.001
Microbleeds counts, n	1 (0–41.5)	1 (0–3)	1.5 (0–6)	0 (0–0)	<0.001	<0.001	< 0.001
ePVS	2 (2–2)	2 (2–3)	2 (2–2)	2 (1–2)	0.170	0.014	< 0.001

Note: Values are presented as number (%) for categorical variables, mean ± SD for normally distributed continuous variables or median (interquartile range) for non‐normally distributed continuous variables.

Abbreviations: AIS, acute ischemic stroke; CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; cSVD, cerebral small vessel disease; ePVS, enlarged perivascular space; HC, healthy control; HTRA1, high‐temperature requirement factor A serine peptidase 1; ICH, intracerebral hemorrhage; SD, standard deviation; WMH, white matter hyperintensity.

^*Bold p‐values indicate statistical significance (p < 0.05).

2.4. Quantification of MRI lesion burden

MRI markers of cSVD were defined according to the STandards for ReportIng Vascular changes on nEuroimaging (STRIVE). ¹⁹ The number of lacunes and cerebral microbleeds (CMBs) was obtained in each individual after visual assessment of 3D T1, 3D FLAIR and SWI images, respectively. The severity of enlarged perivascular space (ePVS) was semiquantitatively scored based on the slice and hemisphere with the largest number of ePVS in basal ganglia (BG) and centrum semiovale (CSO), respectively (0 = no ePVS, 1 = 1–10 ePVS, 2 = 11–20 ePVS, 3 = 21–40 ePVS, 4 = over 40 ePVS). The ePVS score corresponded to the sum of BG‐ePVS score and CSO‐ePVS score for each individual. The volume of WMH was calculated from WMH masks and as the product of the number of voxels and the volume per voxel. To account for variations in brain volume, we normalized the volume of WMH to the intracranial volume.

2.5. Statistical analysis

The normality of data was assessed by the Shapiro–Wilk test. Normally and non‐normally distributed continuous data were described as mean ± standard deviation (SD) and median and interquartile range, respectively. Categorical data were reported as numbers of cases and percentages (n, %). For continuous data, t‐tests or Mann–Whitney rank‐sum tests were applied accordingly, while for categorical data, chi‐squared tests or Fisher exact tests were used. 

FIGURE 2.

K_w and K^trans comparisons between the healthy control (orange) and CADASIL (red), HTRA1‐related cSVD (blue), and sporadic cSVD (purple) in the four ROIs (columns). (A) BBB water exchange rate (k_w). K_w values were lower in CADASIL and HTRA1‐related cSVD compared with healthy controls for whole brain, DGM and NAWM, while values were not significantly different between the sporadic cSVD and healthy controls. (B) BBB leakage rate of gadolinium‐based contrast agent (K^trans). K^trans values were higher in HTRA1‐related cSVD for the whole brain, and higher in sporadic cSVD for the whole brain and NAWM compared with healthy controls, but not significantly different between CADASIL and healthy controls. * p < 0.05; *** p < 0.001. BBB, blood‐brain barrier; CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; cSVD, cerebral small vessel disease; DGM, deep gray matter; HC, healthy control; HTRA1, high‐temperature requirement factor A serine peptidase 1; NAWM, normal‐appearing white matter; ROI, region of interest; WMH, white matter hyperintensity.

K_w, Vp, and K^trans were compared between the three cSVD groups and healthy controls using general linear regression analysis with age, sex, and group (CADASIL vs. controls, HTRA1‐related cSVD vs. controls, and sporadic cSVD vs. controls) as independent variables, and k_w, Vp, or K^trans as the dependent variable respectively. Correlations of k_w and K^trans in ROIs were evaluated by Spearman correlation analysis.

General regression model was performed to determine the associations between K^trans/k_w and cSVD imaging markers and mRS (adjusted for age and sex), and global cognition (adjusted for age, sex, and education level). Statistical analyses were performed using IBM SPSS 26.0. In this exploratory study, all statistical data analyses were two‐sided at a 0.05 level of significance as the primary analysis. For more strictly statistical testing, we further did Bonferroni correction to adjust for multiple comparisons, the results of which were shown in Tables S1—S6.

3. RESULTS

3.1. Subject characteristics

In total, we recruited 123 subjects. DP‐pCASL data were available from 40 CADASIL, 13 HTRA1‐related cSVD, 30 sporadic cSVD, and 40 healthy controls, and DCE‐MRI data were acquired in 33 CADASIL, 10 HTRA1‐related cSVD, 25 sporadic cSVD and 25 control participants (Figure S4). Example images of k_w and K^trans map of the representative patient in each cSVD group were shown in Figure S5. Demographics, vascular risk factors, and medical history were shown in Table 1. Compared with controls, age was not significantly different in the two monogenic cSVD groups, but as expected, it was significantly higher in the sporadic cSVD group. Besides, males were more common in HTRA1‐related cSVD and sporadic cSVD. Compared with the healthy control group, sporadic cSVD subjects had poorer cognition.

3.2. Comparison of k_w, Vp, and K^trans between the three groups of cSVD and healthy controls

Compared with healthy ,   controls, the CADASIL group and HTRA1‐related cSVD group showed significantly lower k_w in the whole brain (p < 0.001), DGM (p = 0.006 and p = 0.001, respectively), and NAWM (p < 0.001) as shown in Figure 2A and Table S2. No significant difference in k_w was observed between the sporadic cSVD group and healthy controls.

FIGURE 3.

Correlations between K^trans and k_w in each group. (A) In the whole brain, significantly negative correlation was found in the HTRA1‐related cSVD group (r = −0.636, P = 0.048). (B) In the DGM, significantly negative correlation was found in the CADASIL group (r = −0.381, p = −0.029). There were no significant correlations between K^trans and k_w in WMH (C) or NAWM (D) in any group. The r and p values were calculated using the Spearman rank correlation. CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; cSVD, cerebral small vessel disease; DGM, deep gray matter; HC, healthy control; HTRA1, high‐temperature requirement factors A serine peptidase 1; NAWM, normal‐appearing white matter; WMH, white matter hyperintensity.

FIGURE 4.

Associations between cSVD imaging markers and k_w and K^trans. Forest plot summarized the associations between cSVD imaging markers (lacune count (triangle), relative WMH volume (rhombus), CMB count (circle) and PVS score (square) and k_w/K^trans in general regression models after adjusting for age and sex. The x‐axis displays the standardized β and 95% CI. In the CADASIL group, k_w in the whole brain, WMH, and NAWM were negatively associated with lacune count. In the HTRA1‐related cSVD group, reduced k_w in the whole brain, DGM, and NAWM were related to larger relative WMH volume, elevated K^trans in the NAWM was associated with more CMB, and increased K^trans in the DGM was related to higher PVS score. In the sporadic cSVD group, no significant associations between imaging markers and k_w or K^trans were found. The orange lines indicate statistically significant associations. * p < 0.05; *** p < 0.001. CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; CMB, cerebral microbleed; cSVD, cerebral small vessel disease; CI, confidence interval; DGM, deep gray matter; HTRA1, high‐temperature requirement factors A serine peptidase 1; NAWM, normal‐appearing white matter; PVS, perivascular space; WMH, white matter hyperintensity.

For Vp, we found no significant difference between any patient and control groups (Table S3). Therefore, K^trans in our study primarily reflected BBB permeability across cSVD and control groups. HTRA1‐related cSVD exhibited significantly higher K^trans values in the whole brain (p = 0.023). Similarly, the sporadic cSVD group demonstrated significantly higher K^trans in both the whole brain (p = 0.040) and NAWM (p = 0.017). There were no significant differences in K^trans values between CADASIL and the control group (Figure 2B and Table S4).

To compare k_w/K^trans in WMH between patient groups and healthy controls, we obtained k_w/K^trans values of healthy controls by using the WMH mask of each patient group. The WMH k_w of both the CADASIL group (p = 0.004) and HTRA1‐related cSVD group (p = 0.005) was significantly lower compared to healthy controls, while WMH K^trans of sporadic cSVD was significantly higher compared to healthy controls (p = 0.006)

3.3. Correlations between k_w and K^trans in four ROIs

The scatter plots depicting the relationships between k_w and K^trans in the four ROIs were shown in Figure 3. Significantly negative correlations were found only in the DGM of CADASIL (r = −0.381, p = 0.029) and in the whole brain of HTRA1‐related cSVD (r = −0.636, p = 0.048). Conversely, no significant correlations were identified in the remaining two groups.

3.4. Associations between cSVD imaging markers and k_w and K^trans in patient groups

Next, we examined the associations of the two BBB parameters with conventional cSVD imaging markers, respectively. In the CADASIL group, there was a significantly negative association between lacune count and k_w in the whole brain (β = −0.410, p = 0.015), WMH (β = −0.329, p = 0.044), and NAWM (β = −0.380, p = 0.023) after adjusting for age and sex. In HTRA1‐related cSVD, we found that k_w in the whole brain (β = −0.283, p = 0.038), DGM (β = −0.272, p = 0.043), and NAWM (β = −0.270, p = 0.048) was negatively correlated with WMH volume, K^trans in the NAWM was positively related to CMB count (β = 0.595, p = 0.014), and K^trans in the DGM was positively associated with ePVS score (β = 0.483, p < 0.001) after adjustment for age and sex. In sporadic cSVD, no significant relationship of k_w or K^trans with any imaging marker was found (Figure 4 and Table S5).

3.5. Associations between clinical severity and k_w and K^trans in patient groups

Further, we detected whether the two BBB parameters could reflect the clinical severity, including functional dependence and cognition. After adjusting for age and sex, in the CADASIL group, lower k_w in the whole brain (odds ratio [OR] = 0.260, p = 0.024), WMH (OR = 0.298, p = 0.032), and NAWM (OR = 0.283, p = 0.028) were associated with increased risk of the presence of mRS ≥ 2 (Table 2).

TABLE 2.

Associations between the presence of abnormal mRS and k_w/K^trans in three patient groups.

		kw		Ktrans
Parameter	Region	OR (95% CI)	p‐Value	OR (95% CI)	p‐Value
CADASIL	Whole brain	0.260 (0.081, 0.839)	0.024	2.162 (0.790, 5.918)	0.134
	DGM	0.449 (0.177, 1.137)	0.091	1.478 (0.636, 3.442)	0.364
	WMH	0.298 (0.099, 0.900)	0.032	0.987 (0.374, 2.609)	0.980
	NAWM	0.283 (0.091, 0.872)	0.028	1.962 (0.764, 5.043)	0.161
HTRA1‐related cSVD	Whole Brain	0.879 (0.211, 3.655)	0.859	6.773 (0.283, 162.065)	0.237
	DGM	0.788 (0.184, 3.374)	0.748	17.850 (0.392, 812.406)	0.139
	WMH	0.712 (0.169, 2.998)	0.644	1.025 (0.192, 5.479)	0.977
	NAWM	0.847 (0.205, 3.494)	0.818	7.338 (0.465, 115.816)	0.157
Sporadic cSVD	Whole Brain	1.946 (0.675, 5.613)	0.218	0.897 (0.330, 2.440)	0.832
	DGM	2.532 (0.830, 7.721)	0.102	0.313 (0.052, 1.891)	0.206
	WMH	2.117 (0.712, 6.297)	0.177	0.194 (0.017, 2.252)	0.190
	NAWM	1.944 (0.691, 5.468)	0.208	0.934 (0.360, 2.351)	0.889

Note: Adjusted for age and sex.

Abbreviations: CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; CI, confidence interval; cSVD, small vessel disease; DGM, deep gray matter; HTRA1, high‐temperature requirement factor A serine peptidase 1; NAWM, normal‐appearing white matter; OR, odds ratio; WMH, white matter hyperintensity.

Bold p‐values indicate statistical significance (p < 0.05).

The results of cognition were shown in Figure 5 and Table S6. In the CADASIL group, after adjusting for age, sex, and education level, lower k_w in the whole brain (β = 0.824, p < 0.001), DGM (β = 0.571, p = 0.006), WMH (β = 0.672, p = 0.001), and NAWM (β = 0.789, p < 0.001) was associated with poorer global cognition. In the patients with HTRA1‐related cSVD, higher K^trans in the DGM was correlated with poorer global cognition (β = −0.559, p = 0.042). In the sporadic cSVD, higher K^trans in the whole brain was found to be associated with poorer global cognition (β = −0.333, p = 0.036).

FIGURE 5.

Relationships of k_w and K^trans with cognition in the 3 patient groups. The bubble matrix shows the associations between k_w/K^trans and MOCA in general linear regression models after adjusting for age, sex, and education levels. In the CADASIL group, decreased k_w in the whole brain, DGM, WMH, and NAWM were associated with poorer cognition. In the HTRA1‐related cSVD group and sporadic cSVD, increased K^trans in the DGM and the whole brain was linked with poorer cognition, respectively. Color depicts standardized β; circle size depicts p value. The direction of the association is indicated by plus (positive association) and minus (negative association) signs. CADASIL, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; cSVD, cerebral small vessel disease; DGM, deep gray matter; HTRA1, high‐temperature requirement factor A serine peptidase 1; MoCA, Montreal Cognitive Assessment Test; NAWM, normal‐appearing white matter; WMH, white matter hyperintensity.

4. DISCUSSION

In this study, we measured both the BBB water exchange rate and BBB leakage rate of GBCA in patients with sporadic cSVD and with two different types of monogenic cSVD. We found that (1) in the CADASIL group, patients had normal levels of K^trans, but significantly lower k_w in multiple brain regions, and the lower k_w was associated with more lacunes and clinical severity; (2) in the HTRA1‐related cSVD group, patients had both higher K^trans in the whole brain and lower k_w in multiple brain regions. The lower k_w was related to larger WMH volume, and the higher K^trans was associated with more CMB, higher PVS score, and poorer cognition; and (3) in the sporadic cSVD group, patients had no significant change in k_w, but higher K^trans in the whole brain and NAWM, which was associated with poorer cognition. These results suggest that the BBB dysfunction may largely vary among the different subtypes of cSVD and that k_w and K^trans can help depict distinct and heterogeneous changes of the BBB across cSVDs of different origins.

The hypothesized mechanisms of BBB dysfunction and the modification on k_w and K^trans in cSVD subtypes are visualized in Figure 6. As GBCAs passively traverse BBB mainly via the paracellular pathway, an increase in K^trans (when Vp does not change) indicates damage of endothelial cells (ECs), tight junctions (TJs), or pericytes. ²⁰ , ²¹ In contrast, water passes across BBB through various pathways, including passive transport, aquaporin‐4 (AQP4) assisted exchange, and some cotransport proteins. ²² , ²³ , ²⁴ , ²⁵ Therefore, k_w may be altered in a range of BBB pathologies. For instance, Ohene et al. used AQP4 knockout mice and found a longer pre‐exchange lifetime of water, indicating a slower water exchange rate. ¹⁸ It can be speculated that down‐regulated or depolarized AQP4 channels on astrocyte end‐feet could also decrease k_w. In the case of PVS dysfunction, interstitial fluid (ISF) can become trapped in the enlarged PVS due to insufficient fluid movement, ²⁶ resulting in heightened osmotic pressure and preventing water from passing BBB. The impact of BBB structural disruption on k_w might be complex. In the Alzheimer's disease (AD) rat model, elevated PSw (k_w multiplied by Vc) was observed, which was inversely correlated with occludin expression, ²⁷ while reduced BBB water exchange rate in the model of middle cerebral artery occlusion (MCAO) ²⁸ and patients with glioblastoma ²⁹ was reported. The different degrees of structural disruption may have opposite effects on k_w. When structural disruptions are subtle, like in AD and cSVD, there is no significant increase in osmotic pressure in the PVS and brain tissue, making passive water transport easier. However, severe structural disruption in the MCAO model and brain tumor can lead to excessive water accumulation within brain tissue, resulting in hyperosmotic pressure that hinders water passage across BBB. The studies investigating changes in k_w in different animal models were summarized in Table S7.

FIGURE 6.

Hypothesized mechanisms of BBB dysfunction in subtypes of cSVD. (A) Schematic representations of BBB structure and transport mechanisms in healthy (left) and cSVD (right). In health, intact BBB is composed of ECs and their TJs, the basement membrane, pericytes and astrocyte end‐feet with polarized AQP4 channels. In cSVD, there are depolarized or less expression of AQP4 (reduced AQP4 activity), enlarged PVS, dysfunction of ECs and pericytes, and TJ loss. The reduced AQP4 activity results in less efficient water transport across BBB, which would decrease k_w. The enlarged PVS indicates dysfunction of fluid clearance and interstitial edema, which impedes water passing across BBB and leads to reduced k_w. The dysfunction of ECs/TJs/pericytes facilitates passive transport of GBCAs, but the k_w might be decreased or increased depending on the severity of BBB leakage. B‐D shows the structure of the arteriole (upper panel) and capillary (lower panel) of the three subtypes of cSVD. (B) In CADASIL, the arteriopathy is characterized by the presence of GOM within the media and prominent morphological alterations of SMCs, while maintaining a predominantly normal endothelium. The reduced AQP4 activity and enlarged PVS/interstitial edema lead to reduced k_w. (C) In HTRA1‐related cSVD, the autopsy studies demonstrate fibrous intimal proliferation and loss of vascular SMCs. The EC/TJ disruption results in increased K^trans, while k_w can increase or decrease at different stages of EC/TJ disruption. For the HTRA1‐related cSVD subjects enrolled in our study, the EC/TJ disruption might be subtle that increased k_w. In addition, there may be reduced AQP4 activity and enlarged PVS, leading to decreased k_w, which was more pronounced than the effect of subtle EC/TJ disruption on k_w. (D) In sporadic cSVD, vascular pathologies are characterized by EC dysfunction, TJ disruption and SMC degeneration. The EC/TJ disruption results in elevated K^trans. For the sporadic cSVD subjects in our study, the EC/TJs disruption might also be subtle which leads to elevated k_w. Meanwhile, the depolarized AQP4 and enlarged PVS lead to reduced k_w, which might be overruled by the impact of EC/TJ disruption on k_w. By Figdraw. AQP4, aquaporin‐4; BBB, blood‐brain barrier; cSVD, cerebral small vessel disease; EC, endothelial cell; GOM, granular osmiophilic material; HTRA1, high‐temperature requirement factor A serine peptidase 1; PVS, perivascular space; SMCs, smooth‐muscle cells; TJ, tight junction.

Our results in the CADASIL group are consistent with recent findings of k_w ⁷ or K^trans. ⁸ The lack of significant difference in K^trans aligns with findings from studies on post mortem human brain samples and the mouse model, which showed no fibrinogen leakage in the pure abnormal WM and their presence only in few areas with enlarged PVS or lacunes. ³⁰ Altogether, our findings further support that BBB leakage may not play a central role in the pathogenesis of WM lesions in CADASIL. Conversely, the existing evidences of k_w support that the water exchange rate would be significantly reduced, of which the exact mechanisms remain unknown. Some pathological studies already showed the displacement of AQP4 ³¹ and enlarged PVS in CADASIL, ³² , ³³ which could be hypothesized as a source of the k_w decrease. Additional studies are obviously needed to understand whether the accumulation of NOTCH3‐ECD (extracellular domain) around smooth muscle cells (SMCs) or pericytes could reduce k_w without altering its physical integrity.

In patients with HTRA1‐related cSVD, we observed both significantly higher K^trans and lower k_w. Histopathologically, the vascular pathologies of HTRA1‐related cSVD include fibrous intimal proliferation, thickening, and splitting of the internal elastic membrane, hyaline degeneration of the media, and loss of SMCs, ³⁴ , ³⁵ similar to sporadic cSVD. ³⁶ Studies on BBB dysfunction in HTRA1‐related cSVD are scarce. The higher K^trans in the present study suggests a possible disruption of the ECs, TJs or pericytes. The reduction of k_w is of high interest since subtle alterations of the BBB physical integrity are expected to increase dramatically the exchange rate of water in general. ²⁷ , ³⁷ This discrepancy might be attributed to the extensive modifications in water transportation, such as depolarized AQP4 and enlarged PVS, which surpass the effects of EC/TJ disruption. There are no available pathological studies linking these results to the potential AQP4 alteration or enlarged PVS in this disorder.

Patients with sporadic cSVD showed higher K^trans in the whole brain and NAWM, in agreement with previous results obtained in patients with lacunar infacts ³⁸ or VCI. ³⁹ In sporadic cSVD, pathological studies in animal models ⁴⁰ , ⁴¹ and patients ³⁶ already revealed EC dysfunction and TJ disruption. Such structural modifications could explain the higher K^trans that we observed, while k_w could increase in such a condition. Depolarization of AQP4 and enlarged PVS have also been reported post mortem in patients of sporadic cSVD, ⁴² , ⁴³ , ⁴⁴ which could decrease k_w. However, we did not observe a significant difference in k_w between sporadic cSVD and controls, which requires cautious interpretation. Sporadic cSVD exhibits greater heterogeneity of pathophysiology than monogenic cSVDs. The impact of depolarized AQP4 or enlarged PVS on k_w (leading to lower k_w) might be greater, the same, or less than the BBB structural disruption (leading to higher k_w) in different disease stages. Thus, exploring the change of k_w can be complex, and the sample size in the present study may be insufficient. It may be beneficial to include patients across different stages and track changes in BBB measurements over time.

In each patient group, the significantly altered BBB measures were correlated with the clinical severity. Particularly in the CADASIL group, the reduced k_w in multiple regions was associated with more lacunes, poorer cognition, and higher mRS. BBB measurements might serve as indicators for assessing the severity of vascular wall alterations, ultimately leading to lesion accumulation. Additional investigations are needed to determine how these parameters are actually involved independently of the accumulation of WMH, microbleeds, and lacunes and according to their measures in different cerebral locations.

We only found significantly negative correlations between k_w and K^trans in the DGM of CADASIL and in the whole brain of HTRA1‐related cSVD. We speculated that the two BBB measurements were associated with vascular wall damage or disease severity, thereby leading to the indirectly negative correlations. In other brain regions, the correlation was missing, possibly due to the different transport mechanisms represented by k_w and K^trans. In addition, water diffusion between compartments can modify the MR signal enhancement in the tissue and the measured K^trans, as GBCA concentration is quantified indirectly through water relaxation. ²² Larsson et al. found that water exchange rate can significantly affect perfusion estimation in the brain when using Gd‐DTPA as an intravascular contrast agent. ⁴⁵

This study has several limitations. First, due to the rarity of HTRA1‐related cSVD, the sample size of this group is relatively small. Second, DCE data were not available in several cases due to GBCA injection failure, excess motion, and participants’ worries about the side effects of GBCA, which might reduce the statistical power to detect associations. Third, we enrolled a single control group younger than the sporadic cSVD, although we adjusted for age in all analyses. Fourth, when analyzing the associations between clinical severity and BBB measurements, we did not adjust for the imaging markers due to the limited sample size. Fifth, we hypothesized the BBB dysfunction properties of each cSVD subtype based on the available cross‐sectional data. Indeed, the pathophysiology of cSVD is a dynamic process and BBB dysfunction may vary across disease stages. Further studies are needed to follow up the changes in the two BBB parameters. Sixth, the spatial resolution and signal‐to‐noise ratio of k_w maps are relatively low compared to standard ASL images due to diffusion preparation and total‐generalized‐variation (TGV) regularized modeling. Ventricular CSF could cause PVE on k_w in WM with low spatial resolution. However, our DP‐pCASL method utilized a single TE and moderate labeling pulse duration and PLD, thus it is not feasible to reliably measure the signal in the CSF compartment. Finally, we lack direct evidence linking BBB dysfunction mechanisms and alterations of k_w. The precise contribution of each distinct transport pathway to the pathological functioning of the brain and the changes in BBB measurements remains poorly understood.

To conclude, our findings demonstrate distinct alterations in k_w and K^trans among subtypes of cSVD, indicating the heterogeneous nature of BBB dysfunction. Identifying the potential mechanisms underlying the changes in BBB measurements needs further investigation in animal models or patients with pathological samples.

CONFLICT OF INTEREST STATEMENT

The authors report no competing interests. Author disclosures are available in the supporting information.

CONSENT STATEMENT

Informed consent was obtained by all human participants.

Supporting information

Supporting Information

Supporting Information

Ying Y, Li Y, Yao T, et al. Heterogeneous blood‐brain barrier dysfunction in cerebral small vessel diseases. Alzheimer's Dement. 2024;20:4527–4539. 10.1002/alz.13874