Continuous Blood-Brain Barrier Breakdown In Acute Ischemic Stroke Patients

Neurological diseases such as ischemic stroke and dementia are associated with compromised blood-brain barrier (BBB) permeability. Knowledge about the time course of BBB leakage may have impact on therapeutic interventions and diagnostic measures such as testing for blood biomarkers. However, reports on the timeline and pattern of this leakage are contradictory. Therefore, we aimed to assess the time course of BBB permeability in ischemic stroke patients during the rst 24 hours after symptom onset using dynamic contrast enhanced (DCE) MRI at 3 Tesla. We categorized time from stroke symptom onset to imaging into the following groups 1) 0-6 hours (n=10), 2) 6-16 hours (n=14) and 3) 16-24 hours (n=29). BBB permeability differed signicantly between stroke lesions and the contralesional tissue for groups 2 and 3 (p=0.006, p<0.001, Wilcoxon-signed rank test). Using univariate or multivariate linear regression analyses we found no association between BBB leakage and age, sex, hyperintense reperfusion marker (another marker of BBB permeability) hemorrhagic transformation, white matter lesion load, symptom severity, functional disability and cerebrovascular risk factors. The results of our study therefore suggest continuous BBB leakage in the rst 24 hours after stroke.


Introduction
The blood-brain barrier (BBB) acts as an interface between the vasculature and the brain tissue regulating the bi-directional passage of substances as well as protecting the central nervous system. The endothelial cells are the main physical barrier with the junctional complex of tight junctions, adherens junctions and GAP junctions. These are surrounded by pericytes, astrocytes and the basal membrane 1 .
Neurological diseases such as ischemic stroke compromise the BBB function with increased permeability and potential consequences like brain edema formation, hemorrhagic transformation and polymorphonuclear neutrophil in ltration 2 .
However, there are contradicting reports on the time course of the blood-brain barrier leakage after ischemic stroke and its association with reperfusion. Data from experimental animal models have suggested a biphasic 3,4 or even triphasic pattern 5 of BBB leakage. Huang et al. 3 reported a very early peak between 1.5-2 minutes and a late BBB opening occurring between 4 and 22 hours after ischemic injury. Klohs et al. 4 using near infrared spectroscopy and a radiolabeled albumin compound found an initial BBB opening at 4-8 hours and a second one at 12-18 hours after reperfusion. Preston et al. 5 using [ 3 H]sucrose also found the rst opening 10 minutes after ischemia induction, followed by a second one after 6 hours in certain brain structures, such as striatum and hippocampus and a third opening between 6 -24 hours in the neocortex. These regional differences suggest that different brain areas may have different sensitivity to ischemia and reperfusion. Applying various models, tracers or contrast agents injected impede comparison between studies.
On the other hand, there are also reports of a more continuous opening of the BBB after ischemic stroke in animal models as well as in humans 2,6,7 . The earliest time point of BBB leakage is reported to occur within minutes after stroke onset and the leakage lasts for about 5 weeks in animal models 2 ; in humans, however, the peak occurs between 6 to 48 hours after stroke onset 7 .
Other factors besides ischemic stroke may in uence blood-brain barrier permeability as well, for instance hyperglycemia 8 though, even here with contradicting reports 9 . Furthermore, when using MRI, factors such as main magnetic eld strength, coil setups, sequence types, injection protocols as well as choice of arterial/venous input function can in uence the assessment of BBB leakage as well as the quality of the generated parametric maps 10 .
The aim of our study was to determine whether a time dependency of increased BBB permeability can be observed within the rst 24 hours after stroke onset using dynamic contrast enhanced (DCE) MRI at 3 Tesla (T). We also assessed potential factors in uencing BBB leakage.

Patients
Fifty-three patients of the ongoing prospective observational LOBI-BBB study were enrolled (clinicaltrails.gov NCT02077582). Inclusion criteria was time between stroke onset and imaging ≤ 24 hours. Exclusion criteria were age < 18 years, contraindications to MRI examinations such as pacemaker, strokes of unknown onset as well as renal dysfunction, since contrast agent was applied.
Written informed consent was obtained from all patients. The study design was approved by the local ethics committee of the Charité-Universitätsmedizin Berlin, Germany (EA4/056/08).
All methods were performed in accordance with relevant guidelines and regulations such as the Declaration of Helsinki and the "Strengthening the Reporting of Observational Studies in Epidemiology" (STROBE) guidelines for the reporting of observational studies.
Imaging MR examinations were performed on a 3 T MRI scanner (Prisma_ t, Siemens Healthineers Erlangen, Germany) using a standard stroke protocol 22 as well as a T1 dynamic protocol. For T1 mapping T1 measurements with 4 different ip angles (2°, 10°, 20°, 35°) were acquired. Subsequently, a continuous serial acquisition of 60 volumes of T1-weighted images after administration of 10 mL gadobutrol (Gadovist®, 1 M, Bayer Schering Pharma AG, Berlin, Germany) at a ow rate of 1 mL/s and a delay of 1 minute, followed by a 20-mL saline ush covering a period of 6 minutes was performed. The imaging parameters for the T1 measurements were as follows: echo time (TE) 2.5 ms, repetition time (TR) 55 ms, 7 slices, 5 mm slice thickness, gap 0.5 mm, matrix size 128 x 102. TR for the different ip angle measurements was 60 ms.
Postprocessing of the data included motion correction 23 (FLIRT=FMRIB's linear image registration tool, FSL, FMRIB, Oxford, UK, (https://fsl.fmrib.ox.ac.uk/fsl/)) and coregistration of diffusion-weighted (DWI) and T1 dynamic images using SPM12 (Statistical Parametric Mapping, Wellcome Trust Centre for Human Neuroimaging, UCL, London, https://www. l.ion.ucl.ac.uk/spm/). Regions of interests (ROI) of stroke lesions were created on coregistered DWI images and mirrored to the contralesional side (https://www.nitrc.org/projects/mricron/). Hemorrhagic transformed parts of ischemic stroke lesions are easily identi ed on DWI images and were excluded from the lesion ROI. Another ROI was placed in a major venous sinus and used as venous input function since the ΔR1 measured in the sagittal sinus agrees well with the time course of ΔR1 in arterial blood 24 .

Statistics
Statistical analysis was performed using R (https://www.R-project.org/) 25 . To compare K trans values between stroke and mirror regions the Wilcoxon signed-rank test was employed. Univariate analysis was rst used to assess the association between log-transformed K trans values and parameters like TSI, HARM 12 sign (the presence of contrast agent in the cerebrospinal uid detected on FLAIR), hemorrhagic transformation, and cerebrovascular risk factors . The log transformation of the K trans values was performed to deal with some violations of the assumptions of univariate and multivariate linear regression. A p-value of < 0.05 was considered as signi cant.
Merali et al. 7 also found a continuous increase in BBB permeability, again most pronounced between 6 to 48 hours after stroke onset, but also in the hyperacute phase, that is within the rst 6 hours after ischemia. The authors though, used the parameter permeability-surface area product (PS) instead of K trans , which only corresponds to K trans under very speci c conditions. Only if cerebral blood ow considerably exceeds PS does K trans approximately equal PS 15 . Compared to Merali et al. 7 with 20 patients in the hyperacute phase we only had 10 patients and used a different software package for the Patlak analysis, which might contribute to the difference in the results. Furthermore, BBB leakage is not only restricted to the stroke region but was also found in contralesional brain tissue 14 in the acute phase of stroke which we found in 6/10 patients in the rst group. This might also contribute to the lack of signi cant difference in BBB permeability in the rst 6 hours after symptom onset. When contralesional BBB leakage recovers, as has been reported 14 , differences in BBB leakage between stroke and contralesional tissue become more pronounced.
As already pointed out, differences in equipment such as MR scanner eld strength, coil con guration, duration of ischemia, tracers, experimental design, software packages and kinetic models impact the quanti cation of BBB leakage 2,10 , in particular with respect to a probable second BBB opening.
When applying Patlak analysis, which is used in most of the studies, Manning et al. 16 found factors yielding erroneous results when assessing BBB permeability. These include the rate of injection of contrast agent (CA). This can result in fast BBB water exchange and thereby in a substantial bias of the permeability surface product and occurs during the early part of the time course of the contrast agent application. A slow injection is less sensitive to plasma ow, since the slower exchange in Gadolinium concentration results in similar arterial, capillary and venous concentrations. Excluding early data points resolves the problem of the fast BBB water exchange, reduces blood ow effects and the sensitivity to plasma ow. Furthermore, a small delay for the bolus injections has signi cant impact on the estimated parameters and signi cant errors remain even when early data points are excluded, while slow injection of Gadolinium virtually eliminates the problem. T1 saturation and transverse dephasing caused by the high concentration of a bolus injection can be avoided as well 16 . Since we employed a slow injection protocol and started the analysis using the very rst time points of signal intensity increase we could address these factors mentioned above to minimize bias in our analysis.
Complicating the reported results is the relative lack of longitudinal studies of BBB leakage. One and evidence of deposition in certain brain regions 18, 19 .
As previously reported 14 , we found no association between K trans values and HARM sign, another reported marker of BBB disruption, which was observed in 21 out of 53 of patients in this current study. This can be explained by a different pathophysiological mechanism resulting in HARM sign in contrast to an increased K trans . Additional factors in uencing HARM sign are infarct size (median volume size 0.73, IQR 0.35-1.89 mL), which showed a large variability in our study cohort, as well as risk factors such as hypertension and diabetes 20 . Both were reported to have an impact on HARM sign and showed no association with K trans values in our study using DCE-MRI, again probably due to the two different techniques applied. We also did not nd an association between K trans and hemorrhagic transformation, but the number of patients was too small (n=3) for any satisfactory analysis and conclusion.
Our study has some limitations, foremost the small number of patients assigned to each time between symptom onset and imaging category. This was largely due to the fact that many patients who were eligible for mechanical thrombectomy did not receive a DCE-MRI scan, as this would have led to unjusti ed delays in treatment. This also might have introduced a bias with respect to symptom severity in our cohort, which was clearly skewed towards minor strokes as is re ected in the NIHSS score.
Secondly, the patient sample was very heterogeneous regarding stroke location with infra-as well supratentorial strokes as well as pronounced differences in lesion volumes. The latter might also be contributing to the missing evidence of detectable BBB leakage in the hyperacute stage.
Thirdly, the impact of various risk factors on BBB leakage must be interpreted with caution due to the small number of patients. This is despite the fact that the distribution of the number of risk factors was comparable across the different time groups.
A different and novel promising technique is currently emerging for assessing BBB leakage noninvasively: DP-pCASL (diffusion prepared pseudo-continuous arterial spin labeling) using the water exchange rate across the BBB as a means to describe BBB leakage. Shao et al. 21 found a good agreement analyzing BBB leakage using DCE-MRI and DP-pCASL as well as high test-retest reproducibility measuring water exchange rate immediately after and about 6 weeks later in elderly participants with cerebral small vessel disease. Using this technique with an endogeneous contrast agent that is water, more insight into the time course of BBB permeability under various conditions may be obtained.

Conclusion
Our results support the view of a continuous BBB leakage over the rst 24 hours after stroke onset, in line with previously published experimental and human data. We found no association between cerebrovascular risk factors and BBB leakage.