MRI of Blood–Brain Barrier Permeability in Cerebral Ischemia

Abstract

Quantitative measurement of blood–brain barrier (BBB) permeability using MRI and its application to cerebral ischemia are reviewed. Measurement of BBB permeability using MRI has been employed to evaluate ischemic damage during acute and subacute phases of stroke and to predict hemorrhagic transformation. There is also an emerging interest on the development and use of MRI to monitor vascular structural changes and angiogenesis during stroke recovery. In this review, we describe MRI BBB permeability and susceptibility-weighted MRI measurements and its applications to evaluate ischemic damage during the acute and subacute phases of stroke and vascular remodeling during stroke recovery.

Introduction

The capillary endothelium plus the basal lamina, the pericytes enclosed within the basal lamina, and the surrounding cuff of astrocytic foot processes form the blood–brain barrier (BBB) complex. This barrier reduces the passive movement of water and restricts water-soluble materials from crossing the BBB, thereby protecting brain cells from exposure to neurotoxins and other unwanted neuroactive blood-borne agents.

BBB leakage after cerebral ischemic damage is present during the acute and subacute phases of ischemic damage [1–8] and during the early stage of angiogenesis associated with stroke recovery [9, 10]. BBB disruption during the acute and subacute stages post-stroke contributes to tissue damage and drives cerebral edema and intraparenchymal hemorrhage [1–8]. The BBB leakage at the early stage of angiogenesis during stroke recovery as an expression of vascular remodeling is a relatively new area of interest, requiring further investigation [9, 10].

Magnetic resonance imaging (MRI) plays a major role in the management of the patient after stroke [11]; among many other uses, MRI has been employed to evaluate BBB integrity via a measurement of the leakage rate of an intravascular contrast agent. BBB leakage can be quantitated by measurement of the blood-to-brain transfer constant (K^trans) as,

$$K^{\text{trans}}=E\times F$$ (1)

where F is the plasma flow and E is the extraction fraction (Fig. 1). E, in turn, is related to the permeability-surface area (PS) product of the local capillary network to the indicator via the expression,

$$E=1-e^{-\frac{\mathrm{P}\mathrm{S}}{F}}$$ (2)

where F is the plasma flow. Using classical tracer kinetic theory, dynamic contrast-enhanced (DCE) MRI permits quantitative evaluation of the leakage of contrast agents across the BBB [12–14], most frequently via the blood-to-brain transfer constant, K^trans, of gadolinium diethylenetriaminepentaacetic acid (Gd-DTPA).

Fig. 1.

Vascular parameters of tracer transport and exchange in the damaged tissue. Flow plasma flow, Fp; shunt flow where there is no exchange of waste or nutrients with the tissue, perfusion nutritive flow, PS permeability-surface area product, E extracted fraction, VD* volume of distribution—not all of the tissue volume is “available” to the extracted tracer (in the case of Gd-DTPA, this represents the interstitial volume, Ve). Modified from ref. [133]

Recently, high-resolution susceptibility-weighted MRI (SWI) has been used to provide complementary information to K^trans for detecting hemorrhage and angiogenesis after stroke. We will also review the application of SWI in staging hemorrhage and angiogenesis after stroke. In this review, we focus on MRI methodologies of BBB measurements and its applications during the acute and subacute phases of ischemic damage, and at the early stage of angiogenesis during stroke recovery.

MRI methodologies employed to evaluate BBB leakage

MRI with gadolinium-based contrast media has been employed to characterize pathophysiological changes during stroke due to its ability to accentuate the underlying flow variations and to aid in visualization by vascular and/or parenchymal contrast enhancement due to BBB leakage of the contrast agent [15–17]. Quantitative evaluation of BBB leakage is based on one of two tracer kinetic models, either the single-capillary model first proposed by Renkin [13] or the Johnson–Wilson tissue homogeneity model [14, 18]. In MRI, because the tissue homogeneity model requires a higher degree of signal-to-noise and higher sampling rates than the single-capillary model, the single-capillary model is almost universally employed, but perhaps with modifications for water exchange across compartmental boundaries, as in the shutter-speed model. The single-capillary kinetic modeling of contrast agent distribution has a basis in the simple rate equation describing diffusive flux across a permeable membrane [13]. This is determined by the differences in concentration between two compartments that are separated by a membrane and by the membrane’s permeability. For the case of transport across the capillary wall, the flux is equal to PS (C_p−C_e), where C_p and C_e are the concentrations of contrast agent in the plasma and interstitial spaces, respectively.

It is easily understood that the plasma concentration (and therefore capillary–tissue exchange) is dependent upon delivery of the indicator to the microvascular bed, i.e., upon flow. As indicated in Eq. 2, the extraction fraction, E, is the relative difference between arterial and venous concentrations of contrast agent and is defined as the fractional reduction of contrast agent in the plasma during its passage through the tissue [13]; as indicated in Eq. 1, E is the blood-to-brain transfer constant, K^trans divided flow [19]. Unless flow is independently determined, MRI-DCE using T1-weighted dynamic contrast cannot be performed quickly enough to estimate flow, and K^trans becomes the major summary parameter of vascular permeability.

In order to produce an absolute measure of K^trans, an estimate of its concentration-time curve in tissue, C(t), and arterial input function (AIF) must be generated. Although the relative change in signal intensity can sometimes be used to approximate these curves [20], in general, an estimate of the change in R₁ (R₁=1/T₁) is required. This requirement in turn ordains that the dynamic imaging be accompanied by an estimate of baseline T₁. The tracer kinetic model used in most MRI studies to date is the modified single-capillary model described by Larsson and Tofts [21, 22]. A major limitation of such one-compartment models is the difficulty in interpreting the parameter K^trans. If the single compartment reflects the interstitial volume, v_e, alone (i.e., the vascular signal is negligible) and plasma flow, F»PS, the so-called permeability limited regime, then K^trans reflects PS. Conversely, if PS»F (the capillary wall presents almost no barrier to the contrast agent), then K^trans reflects F, and the distribution volume is the sum of v_e and vascular volume, v_p. If the contribution to the signal from contrast agent in the v_p of the tissue is significant, then a multi-compartment model is needed. This has long been recognized in PET studies [23] and explicitly modeled in the popular Patlak approach to data analysis [24]. Also, a correction for efflux from the “irreversible” tissue compartment back into the blood is considered in a later modification by Patlak and Blasberg [25]. The Patlak model and graphical methods [24, 25] along with simplified model assumptions have been used successful to analyze dynamic contrast-enhanced (DCE) MRI data and to evaluate K^trans with and without considering efflux [26, 27]. However, quantitative measurement of K^trans using DCE remains a challenge due the possible confound of the “shutter-speed effect” [28–30]. We have demonstrated [31–33] that multipoint estimates of R₁ using a Look–Locker sequence [34] do not demonstrate the serious nonlinearities that are attributed to the faster DCE sequences employing a short TR with highly saturated signals. Our data using the Look–Locker R₁ data acquisition [34] and Patlak plots to analyze the MRI data were confirmed by comparison of K^trans estimates obtained with quantitative autoradiography in the same animals [26, 31, 32, 35]. The Look–Locker R₁ data acquisition with Patlak data analysis provides an accurate estimate of K^trans in most cases of stroke [36–40]. However, these studies showed only moderate rates of leakage. The techniques employed could become biased in the case of a very severe BBB leakage due to the relatively slow rate of data acquisition.

The future direction in the quantitative measurements of BBB leakage will focus on minimizing errors from current methodologies, including more accurate estimate of the AIF, the influence of blood hematocrit, variation of the density of the tissues, water exchange between the various tissue compartments, and the tissue-specific relaxivity of an MRI contrast agent, as well as new methodologies to estimate flow and PS product simultaneously.

SWI uses fully velocity-compensated, radio frequency-spoiled, high-resolution 3D gradient echo scans and incorporates phase information to provide high sensitivity in detecting deoxygenated blood [39, 41–43]. While both phase and amplitude information are present in every MR image acquisition, phase information is generally discarded. It should be noted, however, that the phase information is strongly dependent on local variations in susceptibility, such as caused by deoxygenated blood. Thus, in the MRI experiment, small changes in the concentration of deoxyhemoglobin will be sensitively reflected in the phase of the local magnetization vector as it evolves. This sensitivity is utilized in SWI to form images of very small venous vascular features or hemorrhage [39, 42, 44–46].

Application to acute ischemic damage

A number of studies with various animal models of cerebral ischemia indicate that the BBB opens several hours after occluding a supplying artery. After 4 h of permanent occlusion, the influx of α-aminoisobutyric acid (AIB) and urea (both low molecular weight markers of BBB opening) increases, and leakage of albumin and horseradish peroxidase begins [1–5]. After 1–3 h of transient MCA occlusion and 2–3 h of reperfusion, the BBB opens to AIB and sucrose [47–49], but this opening may reverse temporarily several hours later [49]. Following similar occlusion times, BBB penetration of Evans blue-tagged albumin (EBA) was observed after 2–6 h of reperfusion and appeared to be dependent on rate of tissue perfusion and brain region [50, 51]; BBB opening to EBA may also be biphasic [50]. In these studies, vasogenic edema appeared to accompany the leakage of EBA. As for briefer periods of ischemia, 10–30 min of arterial occlusion plus reperfusion led to BBB opening to ³H inulin and ¹⁴C-sucrose after 3–6 h of reperfusion [52, 53] and was, thus, somewhat later than with 1–3 h of occlusion. Blood–brain barrier opening and vasogenic edema exacerbate ischemic injury and may contribute to reperfusion injury [49, 54, 55].

BBB disruption may also lead to hemorrhagic transformation (HT) in many affected subjects. HT of an ischemic brain is common in stroke patients, occurring in up to 70% of cases [56]. However, it has only been with the advent of sensitive imaging techniques, performed at prescribed times after stroke onset, that the true incidence of HT is becoming clear. Interest in HT after ischemic stroke has been heightened by the introduction of reperfusion strategies designed to remove the occluding clot and reestablish cerebral blood flow. All of the reperfusion strategies proven to be effective in improving functional outcome after ischemic stroke have been associated with an increased risk of HT, which is sometimes fatal [57–59]. Currently, diagnosis of hemorrhage is the domain of computed tomography (CT) rather than MRI, especially in acute stroke [60–62]. However, diagnosis of symptomatic hemorrhage using CT is only 57% efficient [63], and while CT can diagnose hemorrhage once it has occurred, it cannot predict hemorrhage unless high-dose contrast-enhanced CT is used [64, 65]. Magnetic resonance imaging, however, has demonstrated excellent predictive ability for hemorrhage [6, 8, 37]. The ability of MRI to sensitively detect and predict hemorrhage using a contrast agent, and magnetization transfer MRI may be related to early endothelial damage leading to increased permeability to small molecules (i.e., Gd-DTPA contrast agent or small proteins) across the blood–brain barrier (BBB) [6, 8, 37]. These early studies using a contrast agent provide only indirect, technique-dependent information of permeability increases associated with BBB damage [6, 8]. Later investigation using quantitative MRI BBB measurement has demonstrated that BBB transfer constant, K^trans, of the MRI contrast agent is the most sensitive early (2 to 3 h) predictor of the cerebral tissue that will progress to fibrin leakage [37].

MRI can not only predict HT based on endothelial cell damage and the ensuing severe disruption of the BBB but can also identify hemorrhage after it forms, based on changes in the signal induced by magnetic susceptibility [66, 67]. During hemorrhage, hemoglobin gradually changes from the oxy-, deoxy- (ferrous), and met- (ferric) heme proteins found in erythrocytes to the iron (ferric) storage forms of ferritin and hemosiderin contained within phagocytes, depending on the stage of hemorrhage [66]. Traditional MRI images of T₁, T₂, proton density, and gradient echo sequences exhibit different patterns of hemorrhage at different stages of development [68–71]. However, traditional MRI based on signal intensity is not as sensitive, especially during the hyperacute stage of hemorrhage [61, 62, 66]. The deoxy- and met- heme proteins in erythrocytes and the iron storage forms of ferritin and hemosiderin in phagocytes are paramagnetic molecules that cause signal loss (darkening) in images due to magnetic susceptibility, best seen on susceptibility-weighted T₂^* images [66, 72]. Susceptibility-weighted T₂^* MRI can sensitively detect hemorrhage even in the hyperacute stage [72, 73]. Thus, SWI exhibits high sensitivity in detecting hemorrhage, even to microscopic bleeds [39, 42]. Signals from substances with different susceptibilities (such as hemorrhage) will cause phase differences with adjacent tissues [39, 42]. Hemorrhage can be identified on the processed images, which combine magnitude and phase information to maximize the negative signal intensities of the regions containing degraded hemoglobin [39, 42]. By combining MRI SWI and BBB permeability measurements, we can obtain information about the stage and extent of hemorrhage transformation.

Quantitative determination of BBB damage and early prediction of HT during the hyperacute stage of ischemic damage are important to the acute management of stroke. Advances in MRI quantitative BBB measurements have improved its reliability for early prediction and identification of HT and determination of acute ischemic damage. Implementing these approaches in patients will provide new opportunities for better management of acute treatment of stroke.

Application to stroke recovery

The traditional application and development of MRI in stroke have been primarily focused on its acute management, i.e., focused on staging of ischemic damaged tissue, especially identification of reversibly damaged ischemic tissue such as penumbra, with the goal of extending the treatment window, and early detection and prediction of ischemic damage [74–79]. However, only a small percentage of stroke patients can be accessed and treated within the thrombolytic treatment window of 4.5 h [80], even with assistance from advanced neuroimaging tools. It is important to have alternate treatment strategies, particularly restorative therapies, with a less restrictive therapeutic window that can be applied to a large population of stroke patients, and concomitantly to develop advanced neuroimaging methodologies to monitor stroke recovery.

Studies in laboratory animals suggest that cell-based or pharmacological-based neurorestorative treatments can enhance brain reorganization and substantially improve functional recovery, even if treatment is initiated up to weeks after stroke [81–87]. Neurorestorative treatments amplify endogenous processes of brain plasticity, including angiogenesis and neuronal remodeling through neurogenesis and axonal reorganization, which likely contribute to improvement in neurologic function after stroke [81, 88–90]. Current understanding of angiogenesis, neuronal remodeling, and the interaction between angiogenesis and neuronal remodeling after stroke, however, has been derived mainly from regional measurements of stained cerebral tissue sections using histological and immunohistological methods [91–102]. These techniques do not allow dynamic assessment of tissue remodeling and permit only one measurement per experimental animal. MRI has been used as a modality complementary to postmortem studies to noninvasively monitor functional recovery and tissue remodeling after stroke [9, 89, 103]. More recent developments, especially neurorestorative treatment of preclinical models of stroke, have demonstrated that neurological outcome is correlated to vascular and neuronal remodeling, processes that can be monitored by MRI [9, 89]. Because the MRI measurement is noninvasive, MRI indices of vascular and neuronal remodeling related to neurological outcome demonstrated in preclinical stroke study can be translated to the clinic.

It is likely that vascular remodeling contributes to neurological recovery post-stroke [9, 10, 83, 90, 104–106]. Higher cerebral blood vessel density predicts improved outcome and survival in stroke patients [104, 107]. Neurological improvement is stimulated by neurorestorative treatments after stroke induces angiogenesis [81]. Treatment of stroke in rats with neurorestorative therapy increases levels of rat vascular endothelial growth factor (VEGF) [83, 108–111], which consequently enhances angiogenesis and reduces functional deficits [81, 90, 112]. The expression of VEGF promotes vascular remodeling and induces angiogenesis and arteriogenesis, primarily in the boundary zone of the ischemic lesion. These newly formed vessels are important for tissue perfusion, but most likely their benefit derives from the factors expressed by angiogenic and newly formed vessels, such as BDNF, VEGF, VEGFR2, and matrix metalloproteinases (MMP), such as MMP 2 and MMP 9 [113, 114]. Based on histopathological investigation of angiogenesis after stroke, MRI methodologies have been developed and implemented to monitor the spatial and temporal evolution of vascular remodeling [9, 10, 39].

Angiogenesis and vasculogenesis are complex processes by which new capillaries form by sprouting from preexisting vessels or de novo, respectively [115]. These newly formed cerebral vessels are inherently leaky; it can take several weeks to form a functional BBB [116]. An early approach in detecting angiogenesis by MRI was to monitor changes in blood volume which, with time, may reflect the growth of new blood vessels [117–120]. A significant correlation was found between dynamic contrast-enhanced MRI blood volume measurements and histological determination of microvessel density in angiogenic hot spots [117–120]. Elevated cerebral blood volume (CBV) correlates with increased vascular density in patients with glioma [121]. However, an inherent flaw in the correlation of CBV with vascular remodeling may arise from increased vessel diameter, resulting in an increased blood volume fraction that does not necessarily reflect increased number of vessels as determined by mean vascular density [121].

A link between angiogenesis and vascular permeability has been established through the work of Dvorak on VEGF [122]. Upon activation of VEGF receptors by VEGF, microvessels become hyperpermeable to plasma proteins and other circulating macromolecules. Such hyperpermeability was found to accompany angiogenesis in tumors, healing wounds, retinopathies, inflammatory conditions, and physiological ovarian angiogenesis [121]. Quantitative evaluation of the leakage of contrast across the BBB has been applied to reproducibly measure vascular permeability in cancer patients [123]. MRI CBV, cerebral blood flow (CBF), and K^trans have also been employed to measure angiogenesis after stroke [9, 39]. MRI in combination with 3D laser scanning confocal microscopy (LSCM) images of neural progenitor cell therapy of stroke in rat shows enhanced angiogenesis in ischemic boundary regions after cell-based treatment, confirmed by an increase in vascular density and the appearance of large thin-walled mother vessels in LSCM [9]. The enhanced angiogenesis was coincident with increases of CBF and CBV at 6 weeks after treatment, and also coincident with transient increases of blood-to-brain transfer constant (K^trans) of Gd-DTPA with a peak at 1 to 3 weeks after cell therapy [9, 124]. Compared to early BBB leakage in the angiogenic areas after neurorestorative treatment, the non-treated stroked animals exhibited a delayed BBB leakage (2 to 5 weeks) in the angiogenic areas (Fig. 2) [39]. Previous investigations also reported that new microvessel formation started from 12 h up to 21 days or longer in a peri-infarcted area [125–128]. The K^trans and other MRI measurements, along with advanced analysis using iterative self-organizing data, identify the location and area of vascular remodeling and angiogenesis [9, 129]. K^trans is a sensitive parameter to detect early stage of angiogenesis, based on similar mechanisms of angiogenesis in tumor. However, a tumor exhibits a constant increase in K^trans due to the continuous growth of new vessels, while a transient increase in K^trans, with a short time window, is demonstrated after restorative treatment of stroke [9]. Thus, the sensitivity of K^trans to detect angiogenesis is time dependent during stroke recovery.

Fig. 2.

Angiogenesis appearances in MRI K^trans and SWI after stroke with and without treatment: K^trans maps (a, b) and SWI (c, d) images exhibited early increases in K^trans (b, red arrow) and the dark lines in SWI (d, white arrowhead) in the angiogenesis related areas for the treated compared to the control rat (a, c)

In addition to K^trans, SWI also provides high sensitivity in detecting angiogenesis [39, 41]. Because angiogenesis typically occurs in regions of high oxygen extraction, SWI will generate early images of small draining veins in peri-infarct regions that are likely to promote angiogenesis. This deoxyhemoglobin can also be detected by T2* maps [39]. A combination of SWI, T2*, and K^trans may therefore provide information about the stage of angiogenesis [39, 41].

Although K^trans measurement is applied to clinical tumor patients [12, 130], there are few applications of K^trans measurement in stroke patients [131, 132]. This may be attributed to an under appreciation of the importance of the K^trans measurement and the need for an additional injection of contrast agent other than that for perfusion. Perfusion is an important measurement in clinical stroke patient study. Therefore, K^trans measurement can be performed by either eliminating perfusion or using a dose of contrast agent in two separate injections. However, arterial spin labeling CBF measurement becomes available in more clinical MRI scanner and could be an ideal combination with K^trans measurement for monitoring vascular remodeling after stroke.

We have reviewed MRI methodologies used in the evaluation of BBB leakage and its applications in both the acute and recovery phase after cerebral ischemic damage. Although the available MRI measurements of BBB leakage have demonstrated a great potential in the management of patient treatment, these techniques need to be further developed to increase the accuracy of the measurements and to provide information from both flow and permeability with a single MRI contrast agent injection. The application of MRI measurements of BBB leakage during both the acute phase of stroke and stroke recovery may also need to be combined with other measurements to provide staging information of ischemic and hemorrhagic damage as well as stroke recovery. Since the noninvasive nature of MRI permits translation of MRI methods from animals to patients, validation of MRI techniques for studying prediction of ischemic damage and hemorrhage as well as therapy-induced stroke recovery could lead to optimization of treatment protocols and improved management of stroke.