The role of oxidative stress in blood-brain barrier disruption during ischemic stroke: antioxidants in clinical trials

Abstract

Ischemic stroke is one of the leading causes of death and disability. Occlusion and reperfusion of cerebral blood vessels (i.e., ischemia/reperfusion (I/R) injury) generates reactive oxygen species (ROS) that contribute to brain cell death and dysfunction of the blood-brain barrier (BBB) via oxidative stress. BBB disruption influences the pathogenesis of ischemic stroke by contributing to cerebral edema, hemorrhagic transformation, and extravasation of circulating neurotoxic proteins. An improved understanding of mechanisms for ROS-associated alterations in BBB function during ischemia/reperfusion (I/R) injury can lead to improved treatment paradigms for ischemic stroke. Unfortunately, progress in developing ROS targeted therapeutics that are effective for stroke treatment has been slow. Here, we review how ROS are produced in response to I/R injury, their effects on BBB integrity (i.e., tight junction protein complexes, transporters), and the utilization of antioxidant treatments in ischemic stroke clinical trials. Overall, knowledge in this area provides a strong translational framework for discovery of novel drugs for stroke and/or improved strategies to mitigate I/R injury in stroke patients.

Graphical Abstract

1. Introduction

Globally, stroke is the second-leading cause of death and the third-leading cause of combined morbidity and mortality (1). In the United States, stroke is the fifth leading cause of death and approximately 87% of strokes are ischemic in nature (2). The only FDA approved therapeutic to treat ischemic stroke is recombinant tissue plasminogen activator (r-tPA), which must be administered within 4.5 h of the onset of stroke and carries with it an increased risk of intracranial bleeding. Additionally, surgical interventions (i.e., endovascular thrombectomy (EVT)) where an occluding thrombus is mechanically removed in an effort to restore cerebral perfusion have become routine in recent years(3). Both of these treatment strategies for stroke involve reperfusion, which can accelerate neuronal injury and contribute to post-stroke functional neurological deficits. As such, an improved understanding of pathological processes associated with ischemic stroke is critically needed to facilitate development of more effective and safer treatment paradigms.

Ischemic stroke occurs when a vessel supplying blood to the brain becomes obstructed and the brain is deprived of oxygen, glucose, and other essential nutrients. Although the human brain accounts for only 2% of the body weight, it uses 20% of the body’s energy (4). This high metabolic demand causes the brain to be especially susceptible to ischemia/reperfusion (I/R) injury. The pathogenesis of stroke is complex, but oxidative stress caused by the overproduction of reactive oxygen species (ROS) and/or reactive nitrogen species (RNS) has been shown to be centrally involved (5–7). ROS are formed by the partial reduction of oxygen and, under physiological conditions, directly interact with critical signaling molecules involved in a wide range of cellular processes including proliferation and survival, cerebral vascular tone, transcriptional regulation, apoptosis, aging, iron homeostasis, DNA damage response, and inflammation (8–11). When the physiological balance between ROS generation and ROS clearance exceeds antioxidant capacity, oxidative stress occurs and is followed by cellular injury. ROS can damage cells during I/R by reacting directly with macromolecules (e.g. proteins, nucleic acids, lipids), thereby increasing the permeability of the mitochondrial transition pore, reacting with free iron (Fenton reaction), activating matrix metalloproteinases (MMPs), and upregulating chemokines and adhesion molecules (12). The brain is especially vulnerable to oxidative stress due to its high energy consumption, lipid rich environment, and weak antioxidant capacity (13).

The blood-brain barrier (BBB) protects brain parenchyma from potentially neurotoxic substances in the circulation by regulating the permeability of substances across the cerebral endothelium. Permeability across the BBB is very low compared to the peripheral microvasculature. Demonstrating a transendothelial electrical resistance of 1500–2000 Ω•cm², the BBB is generally only passively permeable to small, lipophilic molecules with few hydrogen bond donors and/or acceptors (14–16). Passive diffusion through the paracellular route is limited by the tight junction (TJ) and adherens junction protein complexes linking endothelial cells at the BBB (17, 18). Transcellular diffusion is regulated by low rates of pinocytosis and efflux transporters (e.g. P-glycoprotein (P-gp), Breast Cancer Resistance Protein (BCRP)) that pump lipophilic molecules from the endothelial plasma membrane back into the bloodstream (18–20). Important nutrients such as glucose, amino acids, nucleosides, and certain neurotransmitters are able to gain entry into the CNS through carrier-mediated transport at the BBB (21). Alterations in the BBB’s ability to regulate entry of peripheral substances into the CNS can lead to cellular damage and neurotoxicity. Important TJ and transporter proteins that are expressed at the BBB and are important in the context of ischemic stroke are shown in Figure 1.

Figure 1: Localization of tight junction proteins, ABC, and SLC transporters in brain microvascular endothelial cells that have been studied in the context of I/R injury.

(A) Key tight junction proteins regulating paracellular permeability at the BBB include occludin, claudins-3 and -5, and ZO-1, which anchors occludin and the claudins to the actin cytoskeleton. (B) Transporters that have been shown to be upregulated in experimental stroke models and contribute to development of edema and neurological injury through dysregulation of ion gradients (i.e., SGLT1, NKCC1, NHE1/2) are shown. Additionally, MRP isoforms that include the critical endogenous antioxidant glutathione within their substrate profile are depicted at the luminal and abluminal membrane of the brain microvascular endothelial cells. Finally, ABC and SLC transporters that are involved in disposition of drugs relevant to treatment of I/R injury are shown. These include uptake transporters for statins (i.e., Oatp1a4) and memantine (Oct1/2) as well as efflux transporters that are essential determinants of CNS penetration for many currently marketed and experimental therapeutics (i.e., P-gp, Bcrp). As the evaluation of transporters in the setting of I/R injury advances, this figure will evolve to include multiple other BBB transporters.

Cerebral I/R leads to disruption of the BBB, a process that results in deleterious consequences for the CNS such as cerebral edema. Cerebral edema is a serious complication arising from ischemic stroke that can lead to death. Cytotoxic edema, characterized by increased accumulation of fluid into brain cells, and vasogenic edema are both important components of cerebral edema caused by ischemic stroke. Vasogenic edema is caused by a disruption of the TJs at the BBB and leads to extracellular accumulation of fluid and extravasation of potentially neurotoxic serum proteins. BBB disruption during reperfusion can lead to hemorrhagic transformation that may be a natural consequence of ischemic stroke or may be enhanced with r-tPA therapy (22). Later stages of TJ and BBB permeability changes following I/R are thought to be beneficial as they are associated with vascular recovery, repair, and angiogenesis. I/R induced changes to BBB permeability must also be considered in the context of drug delivery. Alterations in TJ proteins or drug transporter functional activity at the BBB can affect the ability to deliver therapeutics to the CNS at pharmacologically relevant concentrations. Due to the ability of oxidative stress to impact BBB integrity, targeting the production and/or downstream effects of ROS generation is an opportunity to develop effective therapeutic strategies to treat ischemic stroke (5). The remainder of this review will focus on how oxidative stress disrupts the BBB and the ongoing results of antioxidant therapy in human clinical trials to treat ischemic stroke.

2. Generation of ROS during ischemia/reperfusion injury

ROS include oxygen ions such as the superoxide anion (O₂•⁻), free radicals such as the hydroxyl radical (OH•), the hypochlorite ion (OCl⁻), peroxyl radicals (ROO), and peroxides such as hydrogen peroxide (H₂O₂), which are unstable and capable of reacting with a wide range of molecules. O₂•⁻ is highly reactive and can oxidize macromolecules or form H₂O₂ when protonated. Additionally, O₂•⁻ readily reacts with nitric oxide (NO) to form peroxynitrite (ONOO⁻), a highly toxic RNS that contributes to I/R brain injury primarily by reacting with tyrosine residues on proteins to form 3-nitrotyrosine adducts (23, 24). H₂O₂ is less reactive than superoxide but can diffuse a longer distance across membranes because it is non-polar. OH•, formed by the reduction of peroxides by reduced iron or copper ions, is highly reactive and capable of stripping electrons from nearby molecules, which further generates and propagates the production of more ROS (25). Mitochondria have been identified as a major source of ROS production during I/R (26, 27). Cellular deprivation of oxygen and glucose induces a decoupling of the mitochondrial respiratory chain that causes a build-up of reduced intermediates which can then react with oxygen to form O₂•⁻ (28). Interestingly, restoration of blood flow to the ischemic tissue leads to a burst-like production of ROS which can cause reperfusion injury (29–32). Thus, both ischemia and reperfusion are involved in generating ROS during stroke.

In addition to mitochondria, several different sources of ROS generation have been identified including endogenous enzymes such as the NADPH oxidase (NOX) family, xanthine oxidase (XO), nitric oxide synthase (NOS), myeloperoxidase (MPO), lipoxygenase family (LOX), and cyclooxygenase (COX) (33–36). Seven isoforms of the catalytic subunit of NOX have been identified, five of which are expressed in the CNS (NOX1-5). Every cell type of the neurovascular unit expresses at least one NOX enzyme complex (37). NOX enzymes produce O₂•⁻ under physiological and pathological conditions except for NOX4 that primarily produces H₂O₂ (12). Cerebral NOX expression and superoxide generation is increased in animal stroke models and post-mortem brain tissue of human stroke patients (38–40).

NOS enzymes synthesize NO from L-arginine. Under physiological conditions, NO can act as a signaling molecule and its production is involved in synaptic plasticity in the CNS, central regulation of blood pressure, smooth muscle relaxation, and vasodilatation (41). Under pathological conditions, excess generation of NO leads to formation of ONOO⁻ and oxidative stress may result. The four isoforms of NOS are endothelial NOS (eNOS), neuronal NOS (nNOS), inducible NOS (iNOS), and mitochondrial NOS (mtNOS) (42). In the context of stroke, eNOS and nNOS are generally thought to be neuroprotective while expression of iNOS is associated with oxidative stress and I/R injury (43). The role of mtNOS in ischemic stroke is not well characterized.

Other ROS generating enzymes, such as XO, may also be involved in the pathogenesis of stroke. XO catalyzes the formation of uric acid and the oxidants O₂•⁻ and H₂O₂ in the presence of purine substrate and molecular O₂•⁻. The role of XO in I/R injury, however, is quite controversial with different studies showing either beneficial or deleterious effects of XO activity (44–47).

Another significant source of oxidative stress during cerebral I/R is MPO, which is expressed on neutrophils, microglia, and monocytes/macrophages (36). MPO catalyzes the reaction of H₂O₂ with Cl⁻ to form chlorous acid/hypochlorite ion (HOCl/OCl⁻), which can directly participate in redox reactions with macromolecules and trigger cell death (36, 48). An increase in neutrophils and MPO activity has been reported in both experimental stroke and ischemic stroke patients (49–51) while MPO polymorphisms are associated with infarct size and functional outcome (52).

To combat oxidative stress, cells express several classes of antioxidant enzymes which convert ROS to less reactive species (Table 1). These enzymes generally metabolize either O₂•⁻ or peroxides and include superoxide dismutases (SODs), catalase, thiol peroxidases such as glutathione and thioredoxin peroxidases, and peroxiredoxins (25). Glutathione, a three-amino acid reducing substrate utilized by glutathione peroxidases and peroxiredoxin 6 to remove hydroperoxides, is also an important endogenous antioxidant (53). Additionally, dietary antioxidants such as vitamins, carotenoids, polyphenols, and flavonoids obtained from plant sources can be protective against oxidative stress generated during ischemic stroke (54).

Table 1 –

Enzymes involved in the production of ROS/RNS, their role in BBB disruption, and their inhibitors in pre-clinical I/R models.

ROS/RNS generating enzyme	Species generated	Effects on BBB during I/R	Pharmacological inhibitors
NADPH Oxidase (NOX family)	O2−, H2O2 (directly by NOX4 only, indirectly by dismutation of O2−)	↑Brain edema, ↑Evans blue and sodium fluorescein extravasation, ↑hemorrhagic transformation; ↓claudin-5, ↓occludin, ↓collagen IV expression (38,40, 82–85)	Apocyanin (non-specific), VAS2870 (NOX-2/4)
Xanthine oxidase (XO)	O2−, H2O2 (indirectly by dismutation of O2−)	↑Brain edema, ↑Sodium fluorescein extravasation (87)	Allopurinol
Nitric oxide synthase (NOS family)	NO, ONOO-(indirectly through reaction of O2− with NO)	↑Brain edema, ↑Evans blue extravasation, ↑vasodilation (104)	Nitro-L-arginine methyl ester
Lipoxygenase (LOX family)	ROO−	↑Brain edema, ↑Evans blue extravasation, ↑IgG extravasation, ↑hemorrhagic transformation; ↓claudin-5, ↓ZO-1 expression (83, 84)	ML351 (12/15 LOX)
Cyclooxygenase (COX family)	O2−	↑Brain edema, ↑Evans blue extravasation, ↑IgG extravasation, ↑hemorrhagic transformation; ↓occludin, ↓ZO-1 expression (86)	Nimesulide (COX-2), CAY10404 (COX-2)
Myeloperoxidase (MPO)	(HOCl/OCl−)	↑Brain edema,↑IgG extravasation (92–94)	N-acetyl lysyltyrosylcysteine amide, 4-aminobenzoic acid hydrazide

A key signaling pathway inducing expression of antioxidant enzymes is the nuclear factor erythroid 2-related factor 2 (Nrf2)–antioxidant response element (ARE) pathway. Under physiological conditions, Nrf2 is inactivated by binding to the Kelch-like ECH-related protein 1 (Keap1) at the actin cytoskeleton. Keap1 senses electrophilic stress via redox-sensitive Cys residues and these events disrupt the Keap1-Nrf2 interaction to suppress proteasomal degradation. Under conditions of oxidative stress, ROS react with Nrf2 and cause it to dissociate from Keap1 and translocate to the nucleus where it binds to AREs resulting in increased expression of antioxidant enzymes (55). The ability of Nrf2 signaling to regulate cellular responses to ROS requires involvement of other intracellular signaling pathways (56). This is supported by the observation that pharmacological inhibitors of p53 (i.e., pifithrin), p38 mitogen-activated protein kinase (MAPK) (i.e., SB203580), and nuclear factor (NF)-κB (i.e., N4-[2-(4-phenoxyphenyl)ethyl]-4,6-quinazolinediamine, SN50) blocked effects of the Nrf2 activator sulforaphane on BBB target proteins such as P-gp (56).

3. Effects of oxidative stress on BBB Integrity during I/R

Neuronal homeostasis depends on proper functioning of the BBB. Cerebral I/R both generates ROS and alters expression, activity, or localization of antioxidant enzymes resulting in oxidative stress. The increased production of ROS combined with reduced cellular antioxidant capacity causes several pathophysiological changes at the BBB that contribute to brain injury and recovery following ischemic stroke. BBB disruption caused by I/R leads to leakage of blood constituents into the brain parenchyma, vasogenic edema, hemorrhagic transformation, and leukocyte infiltration. Leukocytes are also a source of ROS generation and their entry into the CNS further exacerbates oxidative stress in the brain (57).

Many previous studies have shown that oxidative stress is associated with cerebral I/R injury. Middle cerebral artery occlusion (MCAO) in rats has shown a steady increase in ROS production during ischemia with a burst of ROS production upon reperfusion (30). MCAO in mice induces formation of O₂•⁻ and ONOO^– in areas of microvessels colocalized with markers of BBB disruption, such as Evans blue leakage (58). An increase in lipid peroxidation and 3-nitrotyrosine formation has also been found in rat brains following I/R (59–61). Levels of the antioxidant enzyme SOD and the endogenous antioxidant glutathione are also altered by I/R in rats (62). Higher SOD indicates a cellular response to ROS generation while depletion of GSH suggests ROS formation due to I/R. In human stroke patients, plasma levels of 4-hydroxynonenal (4-HNE) and malondialdehyde, markers of lipid peroxidation, are significantly higher than in control subjects (63, 64). Plasma levels of the oxidative biomarker F2-isoprostane, a prostaglandin-like compound formed by free radical-initiated peroxidation of arachidonic acid, are elevated in the hyperacute phase of ischemic stroke and are predictive of infarct growth (65). Nitrates and nitrites, markers of NO production, are also elevated in the CSF of stroke patients in a manner which correlates with infarct size (66).

The BBB shows an increase in permeability during I/R in both animal models and humans (67). The degree of BBB opening in experimental stroke is either continuous and monophasic in nature or biphasic depending on the animal model, species, severity of I/R injury, method of detection, and time point examined (67–72). Bernardo-Castro et al. have proposed that the hyperacute phase (< 6 h) of BBB opening is caused by hypoxia while the acute phase (6–72 h) is mediated by inflammation and reperfusion injury. The subacute phase (72 h-6 wks) is associated with angiogenic and recovery processes, and the chronic phase (> 6 wks) coincides with neovascularization (73). Thus, early phases of BBB disruption likely contribute to cell injury and death while the later phases are associated with vascular recovery and repair. ROS production is able to alter BBB function by modifying proteins (e.g. TJ complexes, basement membrane, adherens junctions), lipids, DNA, and intracellular signaling pathways (74).

Several NOX isoforms have shown involvement in BBB disruption in animal models of stroke. In mice lacking gp91^phox, a subunit of NOX, mice displayed reduced infarct volume and brain edema 1 h into reperfusion (75). Apocyanin, a nonspecific NOX inhibitor, reduced Evans blue leakage after MCAO in a dose-dependent manner (75). NOX2 and NOX4, but not NOX1, expression is increased in the brains of mice during reperfusion after MCAO (38). Administration of the NOX2/4 inhibitor VAS2870 30 min before reperfusion reduced infarct size, Evans blue extravasation, and hemorrhagic transformation 24 h after ischemia. VAS2870 also prevented reperfusion-induced alterations to the TJ proteins claudin-5 and occludin as well as the basement membrane protein collagen IV (38). NOX4 expression is upregulated in neurons and endothelial cells in mice after MCAO and in human stroke patients after routine autopsy (40). Kleinschnitz et al. found that MCAO in mice induces a decreased infarct volume in NOX4-null mice compared to wild-type mice while mice deficient in NOX1 or NOX2 show no difference in infarct volume (40). In contrast, De Silva et al. found decreased in 3-nitrotyrosine formation in the MCA and reduced infarct volume in NOX2^−/− mice after MCAO (76). NOX4 knockout mice also show reduced 3-nitrotyrosine formation, reduced Evans blue extravasation, and less edema after MCAO suggesting ROS formation is involved in the development of cerebral edema after I/R (40, 77). However, one study in NOX1 knockout mice showed no differences in brain edema following MCAO while another study showed a decrease in brain edema and BBB permeability to Evans blue and sodium fluorescein (78, 79). In a humanized NOX5 knock-in mouse model, mice displayed an increase in ROS formation, BBB leakage to Evans blue, and infarct size after MCAO compared to wild type mice (80). In short, NOX 5 is a developing story in I/R.

An important role for the lipoxygenase family in promoting BBB damage following I/R has been demonstrated in several studies. LOX generates ROO and can directly oxidize lipid membranes containing polyunsaturated fatty acids (81). 12/15 LOX levels are increased in neurons and endothelial cells following transient focal ischemia and LOX metabolites are increased in gerbils following bilateral common carotid occlusion and reperfusion (82, 83). LOX gene knockout or inhibition with baicalein reduced brain edema, extravasation of IgG, and claudin-5 down-regulation after MCAO (83). In an embolic MCAO model, the 12/15 LOX inhibitor ML351 significantly reduced tPA-related hemorrhage, IgG extravasation, and loss of the TJ proteins claudin-5 and ZO-1 (84). Given LOX’s roles in producing pro-inflammatory mediators and cell signaling, it should be noted that beneficial effects of LOX inhibition in cerebral I/R models may be independent of its role in generating oxidative stress.

COX enzymes are also capable of generating free radicals and impacting BBB integrity during cerebral I/R injury (85). Cortical, but not subcortical, Evans blue leakage was reduced in an MCAO model when the COX-2 inhibitor nimesulide was administered either immediately or 6 h after the onset of ischemia (86). Following MCAO, reduced extravasation of IgG, significantly lower brain hemoglobin levels, and higher expression of the TJ proteins occludin and ZO-1 was observed in COX-2 knockout mice or mice that had been administered the COX-2 inhibitor CAY10404 compared to controls (86). Similar to LOX, COX enzymes are also prominently involved in producing inflammatory mediators that may affect BBB permeability independent of ROS generation.

XO activity is also involved in BBB disruption during cerebral I/R injury. One study showed the XO inhibitor allopurinol reduced brain edema and BBB permeability to the small molecule tracer sodium fluorescein (87). Interestingly, XO is involved in the formation of uric acid which itself possesses antioxidant properties and shows therapeutic potential for the treatment of ischemic stroke (88–90). In addition to uric acid’s antioxidant properties, it also activates several intracellular signaling pathways, suggesting its therapeutic potential may not be due solely to ROS scavenging (91).

Following cerebral I/R injury, circulating neutrophils enter the brain and contribute to oxidative stress and BBB disruption (57). MPO is abundantly expressed in neutrophils and contributes to the formation of 3-nitrotyrosine adducts after MCAO (23). Matsuo et al. showed a significant decrease in brain edema following MCAO by depleting circulating neutrophils with an antibody (92). Pharmacological inhibition of MPO with N-acetyl lysyltyrosylcysteine amide in an MCAO mouse model reduced infarct volume, IgG extravasation and the oxidative stress markers chlorotyrosine, 3-nitrotyrosine, and 4-HNE (93). In mouse brain endothelial cells, addition of the MPO/H₂O₂/Cl⁻ system resulted in compromised barrier function. This affect was attenuated by the MPO inhibitor 4-aminobenzoic acid hydrazide (94).

It has also been demonstrated that activation of the Nrf2 pathway with dimethyl fumarate (DMF), a drug already in clinical use, protects BBB integrity in ischemic stroke models. DMF causes a dissociation of keap-1 from Nrf2 which results in nuclear translocation of Nrf2 and transcription of antioxidant genes (95). In a mouse model of hypoxia-ischemia, pre-treatment with DMF up-regulated Nrf2 mRNA, enhanced expression levels of Nrf2 target antioxidant proteins, and reduced brain edema and infarct size (96). In an MCAO mouse model, pre-treatment with DMF prevented BBB tight junction alterations and brain edema (97).

NOS isoforms generate NO, which is important in the pathogenesis of stroke due to its reactive nature as a free radical, ability to increase vasodilation, role in cellular signaling, and ability to generate ONOO⁻ when reacting with O₂•⁻. Cerebral I/R has been shown to increase the expression of eNOS, iNOS, and nNOS (98, 99). NO can have protective or detrimental effects during I/R depending on the isozyme that is involved. For example, mice deficient in nNOS have smaller infarcts after MCAO while mice deficient in eNOS have larger infarcts after MCAO, suggesting that eNOS is protective and nNOS is detrimental in the context of cerebral I/R injury (100, 101). Genetic deletion of the iNOS gene and pharmacological inhibition of iNOS both reduced infarct size following I/R in mice (102, 103). Administration of the non-specific NOS inhibitor nitro-L-arginine methyl ester (L-NAME) to rats undergoing MCAO resulted in reduced brain edema and Evans blue extravasation (104). The role of different NOS isoforms in mediating BBB disruption during I/R injury needs further investigation, but the increased production of NO or ONOO⁻ associated with activation of NOS is likely involved due to the ability of ROS to activate matrix metalloproteinases (MMPs) by oxidation or nitrosylation (105–107) leading to leak.

MMPs are a family of zinc-containing endopeptidases that are activated during cerebral I/R and are capable of degrading TJ proteins and components of the extracelluar matrix. In addition to direct activation by ROS, MMP expression can be increased through actions on redox-sensitive elements of transcription factors (such as NF-κB and AP-1) that are known to be binding sites for MMP transcription (108). MMP-9 knockout mice subjected to MCAO display reduced Evans blue leakage and degradation of ZO-1 compared to wild type (109). In Mn-SOD knockout mice undergoing transient focal ischemia, there is increased MMP-9 expression which is associated with TJ alterations and Evans blue extravasation (110). Administration of the broad-spectrum MMP inhibitor BB-1101 reduced BBB leakage to [¹⁴C]-sucrose and degradation of the TJ proteins occludin and claudin-5 following MCAO in spontaneously hypertensive rats (111). Treatment with ONOO⁻ decomposition catalysts prevented MMP-2 and MMP-9 activation and reduced brain edema volume in rats undergoing MCAO (24, 106). These observations were associated with a preservation of collagen IV and brain endothelium (106). L-NAME reduced Evans blue leakage, ZO-1 degradation, and MMP-2 and 9 activity in rats when administered 15 min before MCAO (112). Overexpression of Cu/Zn-SOD has been shown to reduce hyperglycemia-enhanced Evans blue leakage and MMP-9 activation after MCAO in rats (113). MMPs also play a prominent role in the development of hemorrhagic transformation. MMP-3^−/− mice, but not MMP-9^−/− mice, treated with r-tPA after MCAO showed an increase in intracranial bleeding compared to wild-type mice (114). The free radical scavenger edaravone and the spin-trapping agent α-phenyl tert butyl nitrone also prevented r-tPA-induced intracerebral bleeding after MCAO in rats (115, 116). Taken together, these observations suggest ROS activate MMPs which subsequently disrupt BBB integrity during I/R leading to leak and edema.

It is plausible that ROS-induced alterations of TJ proteins contribute to increased BBB permeability observed in I/R injury. Not only do ROS activate the TJ-degrading MMPs, but ROS can also induce TJ changes by structural modifications (i.e. – protein nitrosylation, lipid peroxidation, etc.) or inducing translocation away from the TJ protein complex (117). As little as 1 h of hypoxia followed by 20 min of reoxygenation in rats increased BBB permeability to [¹⁴C]-sucrose and induced changes in structure and localization of the TJ protein occludin and these alterations can be prevented by pre-administration with the antioxidant TEMPOL (118). In bovine brain endothelial cells, hypoxia and reoxygenation increased permeability of [¹⁴C]-sucrose and iNOS expression (119). Although many studies have shown alterations in TJ protein expression or localization is associated with an increase in BBB permeability during experimental stroke (72, 109, 111, 112, 120–125), other mechanisms of BBB disruption during I/R have been proposed. Some studies performed in experimental stroke models have shown an increase in BBB permeability without apparent disruption of the TJ. An increase in brain endothelial transcytosis through vesicles or non-selective channels and endothelial degeneration have also been suggested as mechanisms responsible for increasing BBB permeability in response to cerebral I/R (126–129). Differences in ischemic pre-clinical models, time points examined, and/or detection methods may influence whether TJ alterations are observed during cerebral I/R. These different mechanisms are not mutually exclusive, and all may play a role in BBB opening at different time points following ischemic stroke. Although oxidative stress is well known to cause endothelial cell death (130–133), the effects of oxidative stress on transcytosis at the BBB is an area that warrants further investigation.

Studies in rodents with genetically altered levels of endogenous antioxidant enzymes also indicate an important role for ROS production on BBB disruption and infarct volumes. Mice subjected to MCAO and deficient in Cu/Zn-SOD (SOD1) or Mn-SOD (SOD2) exhibit larger infarct volumes, Evans blue extravasation, and brain swelling, while rats overexpressing Cu/Zn SOD have reduced infarct volumes and edema formation after MCAO (110, 134, 135). Overexpression of human glutathione peroxidase also reduces brain edema and preserves vascular integrity in mouse I/R models (136, 137).

4. Effects of oxidative stress on BBB Transport Systems during I/R

It should also be noted that cerebral I/R can alter functional expression of putative membrane transport systems at the level of brain microvascular endothelial cells. In fact, many members of solute carrier (SLC) family of transport proteins are susceptible to modulation in the setting of pathological states with an oxidative stress component. For example, Yeh and colleagues observed upregulation of glucose transporter 1 (Glut1) in rat brain endothelial cells exposed to conditioned medium derived from rat C6 glioma cells exposed to hypoxic conditions (138). Similar results were obtained in an in vivo model of experimental ischemic stroke where male CD-1 mice subjected to 1 h occlusion of the middle cerebral artery followed by 24 h reperfusion exhibited increased Glut1 expression and transport activity at the BBB(139). I/R injury was also shown to induce functional expression of the sodium-glucose cotransporter 1 (Sglt1) in brain capillaries isolated from male Wistar rats(140). Sglt1 was previously shown to be a critical determinant of blood-to-brain glucose delivery and development of cerebral edema in the setting of I/R(141). Of particular significance, pharmacological inhibition of Sglt1 in male CD-1 mice subjected to MCAO reduced both infarction volumes and edema ratios, which suggests that this transporter may be a viable target for stroke treatment(141). More recently, Stanton and colleagues showed a non-significant increase in organic cation transporter 1 (Oct1) in cortical microvessels isolated from male Sprague-Dawley rats subjected to 90 min MCAO followed by 2 h reperfusion(142). Additionally, Williams et al. showed decreased expression of organic anion transporting polypeptide 1a4 (Oatp1a4) at the BBB in ipsilateral cerebral cortex tissue from male Sprague-Dawley rats under the same MCAO conditions used in the study by Stanton(143). Despite the need to interrogate the temporal relationship between reperfusion and Oct1/Oatp1a4 changes in brain microvascular endothelial cells, these results can have profound implications for CNS delivery of drugs capable of improving post-stroke outcomes such as memantine (i.e., an Oct transport substrate) and atorvastatin (i.e., an Oatp transport substrate)(142–144).

Transport mechanisms for ions are also modulated in response to ischemic conditions. In fact, progression of edema following I/R can be exacerbated by increased brain uptake of sodium ions. Increased flux of Na⁺ and Cl⁻ and, subsequently, movement of osmotically obliged water across the BBB results from enhanced functional expression of Na-K-Cl cotransporter NKCC(145) and Na-proton exchangers NHE1 and NHE2(146). Indeed, NHE1 activity is critical to the regulation of edema volume following I/R injury(147). Additionally, cerebral sodium uptake and cytotoxic edema were reduced in the presence of TRAM-34, a small molecule inhibitor of the BBB calcium-activated potassium channel KCa3.1, suggesting a paramount role for this ion exchanger in the onset of edema following MCAO(148).

ATP-binding cassette (ABC) efflux transporters are established determinants of drug penetration into brain parenchyma. As such, members of this transporter superfamily have been evaluated in the context of I/R injury. Functional expression of P-glycoprotein (P-gp), a 170 kDa integral membrane protein encoded by the multidrug resistance (MDR) gene, has been shown to be altered at the BBB following an ischemic insult. Specifically, Spudich and colleagues showed increased P-gp protein expression in CD31-positive brain microvacular endothelial cells at 3 h and 24 h post-MCAO(149). Additionally, P-gp expression has been shown to be enhanced in cerebral microvessels following 90 min MCAO as well as in cultured rat brain endothelial cells subjected to hypoxia/aglycemia(150). Regulation of P-gp expression and/or activity under oxidative stress conditions is likely to be complex and involve multiple molecular pathways in brain endothelial cells. For example, P-gp transport activity was shown to be increased in rat brain endothelial cells in vitro and in rat hippocampal tissue in vivo following activation of NF-κB signaling in response to hyperammonemia, a pathological condition associated with oxidative stress(151). In contrast, hydrogen peroxide-induced oxidative stress caused a rapid decrease in P-gp functional expression in a human brain microvessel endothelial cell line (hCMEC/d3)(152). This effect was triggered by activation of Abl and Src kinases, which phosphorylated caveolin-1 and promoted cellular internalization of P-gp(152). These two studies points towards complex regulation of P-gp under oxidative stress conditions and suggest that rigorous molecular studies are needed to fully understand the pathways involved. Indeed, altered transport activity of P-gp can have profound implications for CNS delivery of therapeutics relevant to treatment of ischemic stroke. Indeed, several drugs with neuroprotective properties have been shown to be P-gp transport substrates. For example, P-gp has been shown to be involved in CNS disposition of tanshinone IIA, a component of Salvia miltiorrhiza bunge that has been studied for effectiveness as a stroke drug (153). More recently, studies in cultured human endothelial cells (154) and in vivo in female Sprague-Dawley rats (155) have identified the commonly prescribed statin drug atorvastatin as a P-gp transport substrates. Indeed, preclinical (143) and clinical studies (156) have shown that atorvastatin can provide therapeutic benefits in the setting of ischemic stroke. These findings can be interpreted to imply that inhibition of P-gp transport activity can be an effective strategy to optimize drug delivery to ischemic brain tissue. Using the established P-gp inhibitor tariquidar, Bauer and colleagues have reported proof-of-concept data that near-complete inhibition of P-gp mediated transport is achievable at the human BBB (157); however, the ubiquitous expression profile of P-gp must be considered before incorporating this strategy as a stroke treatment paradigm. P-gp is expressed in multiple barrier epithelial cells including colon, small intestine, renal proximal tubules, and bile canaliculi(3). As such, inhibition of P-gp at the BBB using pharmacological antagonists will also block P-gp function in other tissues, an effect that can increase disposition of drugs throughout the body and enhance the probability of off-target effects and/or dose-limiting toxicities (158, 159).

P-gp is known to function in synergy with breast cancer resistance protein (BCRP), a 72-kDa protein that is encoded by the ABCG2 gene (160). Recently, quantitative targeted proteomics studies have shown that the relative abundance of BCRP is higher at the human BBB as compared to P-gp (161). Given the considerable overlap in substrate profile with P-gp (162), BCRP may provide a greater contribution than P-gp to overall BBB efflux transport in health and disease. Functional expression of BCRP has been shown to be altered in response to pathophysiological conditions relevant to ischemic stroke. For example, experimental work conducted in a co-culture model comprised of bovine brain endothelial cells and rat astrocytes demonstrated that Abcg2 mRNA was decreased in response to oxygen/glucose deprivation conditions (163). Of particular significance, Abcg2 transcript levels returned to baseline expression after reoxygenation for 24 h or 48 h (163). Interestingly, it has been demonstrated that phenethyl isothiocyanate (PEITC), a known antioxidant, is a Bcrp transport substrate (164). Indeed, changes in Bcrp expression and transport activity in diseases with an oxidative stress component may affect the CNS disposition and potential therapeutic efficacy of such compounds. As such, it is critical that these findings on Bcrp be validated using in vivo models of experimental stroke, particularly due to the fact that this efflux transporter is known to play a critical role in CNS disposition of stroke drugs such as statins (154, 155).

Additionally, ABC transporters have potential to play a critical role in oxidative stress responses in cellular compartments of the NVU. Indeed, members of the multidrug resistance protein (MRP) family are known to actively efflux the endogenous antioxidant glutathione (GSH), which can have significant implications for pathologies with an oxidative stress component such as I/R. GSH is responsible for maintenance of cellular redox balance and antioxidant defense in the brain. It has previously been reported that various Mrp isoforms are upregulated in response to oxidative stress conditions. For example, increased expression of mRNA transcripts for Mrp1, Mrp2, and Mrp4 were observed in cerebral microvessels isolated from female Sprague-Dawley rats subjected to hypoxia/reoxygenation stress (165). Additionally, acetaminophen intoxicated male C57BL/6J mice, an in vivo condition where oxidative stress was demonstrated by induction of Nad(p)h:quinone oxidoreductase 1 (Nqo1) and hemooxygenase 1 (Ho-1) mRNA and protein, showed increased brain expression of Mrp2 and Mrp4 (166). In both studies, induction of Mrp expression resulted from activation of Nrf2 signaling (165, 166). In primary cultures of rat astrocytes exposed to the human deficiency virus (HIV-1) envelope protein gp120, increased functional expression of Mrp1 resulted in enhanced cellular efflux of GSH (167). In this in vitro model of HIV-associated glial cell injury, oxidative stress was demonstrated by increased expression of heat-shock protein 70 and iNOS as well as increased intracellular concentrations of oxidized glutathione (GSSG) and 2’,7’-dichlorofluorescein fluorescence (167). Molecular mechanisms involved in regulation of Mrp1 expression in cultured rat astrocytes following treatment with gp120 include NF-κB signaling and the c-Jun N-terminal kinase (JNK) pathway (168) Enhanced expression and activity of Mrp isoforms at the BBB/NVU can cause reduced brain and/or endothelial cell concentrations of GSH, an alteration in cellular redox status, and increased potential for cellular injury. Additionally, studies in rat cerebellar granule neurons showed that hydrogen-peroxide induced oxidative stress induced Mrp1-mediated efflux transport of GSH (169). Therefore, the implications of changes in Mrp activity in cellular components of the NVU following exposure to ROS and subsequent oxidative stress require more rigorous evaluation.

5. Clinical use of antioxidants to treat ischemic stroke

Data generated in animal models of stroke indicates ROS production significantly contributes to BBB disruption and detrimental outcomes during cerebral I/R. The use of antioxidants capable of inhibiting oxidation or ROS production to treat ischemic stroke is a strategy that has been attempted over the past several decades. For example, TEMPOL was shown to reduce cerebral glutamate concentrations, decrease infarction volume, and improve neurobehavioral outcomes in male Sprague-Dawley rats subjected to 2 h MCAO (170) Although TEMPOL, as well as other antioxidant drugs, have not received FDA approval as stroke therapeutics, one such compound, edaravone, has been approved to treat ischemic stroke in Japan since 2001.

In a mouse model of prolonged cerebral hypoperfusion, edaravone reduced both MMP-9 expression and BBB permeability to immunoglobulin G while a rat model of transient focal ischemia showed edaravone significantly lowered brain edema (171, 172). In rats subjected to MCAO, edaravone inhibited MMP-9 activity and intracranial bleeding after r-tPA administration (115). Nearly two dozen clinical trials have been performed utilizing edaravone to treat stroke. These trials were typically small, conducted in Asian patients, and yielded mixed results (173–181). A systematic review with meta-analysis of randomized controlled trials (RCTs) and observational studies administering edaravone for treatment of ischemic stroke showed edaravone use was associated with improved chances of a good or excellent outcome after 90 days and lower mortality but no differences in intracranial bleeding (182). Another systematic review with meta-analysis suggested edaravone treatment for acute ischemic stroke showed improved function in daily activities, improved neurological deficit on short-term follow-up, and improvements in re-canalization, but no significant effect on death or disability (183). Limitations of these meta-analyses include only a small number of RCTs, heterogeneity in the included trials, few studies reporting long term results, and many studies being performed in Asian populations.

Ebselen is an organoselenium compound with glutathione peroxidase-like activity. In pre-clinical I/R models, ebselen reduced markers of oxidative stress and significantly lowered brain edema (184–186). In a placebo-controlled, double-blind clinical trial to treat acute MCAO, oral ebselen was well-tolerated and significantly reduced infarction volume and improved neurological outcomes in patients who started treatment within 6 h of onset, which suggests that ebselen may exhibit neuroprotective properties relevant to ischemic stroke treatment (187). In a different trial, patients treated with ebselen had a significantly better outcome at 1 month, but not at 3 months, and the outcome of patients receiving ebselen early (≤ 24 h) was superior to that of patients who received later (> 24 h) treatment (188).

Tirilazad is a synthetic aminosteroid capable of scavenging hydroxyl and lipid peroxyl radicals and stabilizing membranes (189). Although tirilazad displayed neuroprotection in a number of experimental stroke models, a meta-analysis of six double-blind, placebo-controlled clinical trials showed tirilazad increased death or disability by about one-fifth (190). A systemic review of tirilazad treatment in experimental stroke models suggested that the broad range of doses tested, long time interval (~ 5 h) between stroke onset and initiation of treatment, and lack of pre-clinical efficacy in gyrencephalic species all may have influenced the discrepancy between animal and human studies (191).

Disodium 2,4-disulphophenyl-N-tert-butylnitrone (NXY-059), a free-radical spin trapping agent, is neuroprotective in pre-clinical I/R models and has been used in clinical trials to treat ischemic stroke. A meta-analysis of experimental stroke in rodents, rabbits, or primates in 15 different studies showed NXY-059 to be neuroprotective when administered up to 4 h after occlusion (192). The beneficial effects observed in pre-clinical stroke models led to the use of NXY-059 to treat ischemic stroke in human trials. Although the first study showed NXY-059 administered intravenously within 6 h of stroke onset to be significantly better than placebo in improving outcome, the second study with a larger sample size showed NXY-059 was neutral for the primary and all secondary outcomes. A pooled analysis of these two trials confirmed neutral results in the overall population as well as those treated early or with alteplase (193). It has been suggested the discordance between animal and human studies is due to the pre-clinical studies not being relevant to the human condition, the short delay (typically 5–30 min) between onset of ischemia and administration of NXY-059 to animals, study quality biases, and low BBB permeability of NXY-059 (192, 194, 195).

6. Conclusions

The past several decades of pre-clinical experiments in cell culture and animal models of I/R and clinical observations have suggested oxidative stress is prominently involved in the pathogenesis of ischemic stroke. ROS production during cerebral I/R damages BBB integrity by modulating tight junction protein complexes and altering homeostatic transport mechanisms in brain microvascular endothelial cells. These molecular changes contribute to brain edema, hemorrhagic transformation, extravasation of serum proteins, and altered drug delivery to the CNS (Figure 2). A rigorous understanding of these mechanisms and time-course of ROS production and toxicity in experimental stroke models will provide critical information that will lead to new drugs, repurposing of currently marketed therapeutics, and/or strategies to target oxidative stress following ischemic stroke. Many antioxidant drugs have shown the capability to provide neuroprotection in experimental stroke models; however, results have been somewhat discordant in human trials. Adherence to Stroke Therapy Academic Industry Roundtable (STAIR) criteria for preclinical studies (i.e., in vivo dose-response experiments with appropriate allometric scaling, establishment of a time window for effective therapy, blinded study design in reproducible and physiologically controlled animal models, assessment of both histological and functional outcomes at acute and chronic poststroke time points, and use of rodents for initial proof-of-concept studies before consideration of higher order species) provides important guidance to improve the probability of successfully identifying novel stroke drugs(144). This certainly applies to development of novel antioxidant compounds. Other considerations for improved outcomes of antioxidant therapy in human ischemic stroke trials can occur if these drugs are administered earlier, in conjunction with r-tPA treatment, or co-administered along with therapeutic targeting of different ROS pathways. Overall, rationale design and development of novel antioxidant drugs offers an opportunity to advance stroke pharmacotherapy; however, careful experimental design and a more detailed understanding of oxidative stress mechanisms in the brain is required to move these compounds from the bench to the bedside.

Figure 2 – Effects of ROS production on BBB integrity during ischemic stroke.

In the healthy brain (A), the BBB is intact and acts as a barrier allowing for selective entry of substances to the brain from the blood. During cerebral I/R injury (B), the BBB is compromised and substances in the blood may enter the brain through alterations in paracellular or transcellular permeability. Potential consequences of BBB damage during ischemic stroke include leak (vasogenic edema), extravasation of serum proteins, hemorrhagic transformation, and altered drug delivery. Therapeutics may exhibit higher or lower BBB permeability during ischemic stroke depending on whether they are substrates of transporter proteins which have been affected by I/R stress (Image created using BioRender.com).