Neuroinflammation, Stroke, Blood-Brain Barrier Dysfunction, and Imaging Modalities

Abstract

Maintaining blood-brain barrier (BBB) integrity is crucial for the homeostasis of the central nervous system (CNS). Structurally comprising the BBB, brain endothelial cells interact with pericytes, astrocytes, neurons, microglia, and perivascular macrophages in the neurovascular unit (NVU). Brain ischemia unleashes a profound neuroinflammatory response to remove the damaged tissue and prepare the brain for repair. However, the intense neuroinflammation occurring during the acute phase of stroke is associated with BBB breakdown, neuronal injury, and worse neurological outcomes. Here, we critically discuss the role of neuroinflammation in ischemic stroke pathology, focusing on the BBB and the interactions between CNS and peripheral immune responses. We highlight inflammation-driven injury mechanisms in stroke, including oxidative stress, increased matrix metalloproteinase production, microglial activation, and infiltration of peripheral immune cells into the ischemic tissue. We provide an updated overview of imaging techniques for in vivo detection of BBB permeability, leukocyte infiltration, microglial activation, and upregulation of cell adhesion molecules following ischemic brain injury. We discuss the possibility of clinical implementation of imaging modalities to assess stroke-associated neuroinflammation with the potential to provide image-guided diagnosis and treatment. We summarize the results from several clinical studies evaluating the efficacy of anti-inflammatory interventions in stroke. Although convincing preclinical evidence suggests that neuroinflammation is a promising target for ischemic stroke, thus far, translating these results into the clinical setting has proved difficult. Due to the dual role of inflammation in the progression of ischemic damage, more research is needed to mechanistically understand when the neuroinflammatory response begins the transition from injury to repair. This could have important implications for ischemic stroke treatment by informing time- and context-specific therapeutic interventions.

Introduction

Ischemic stroke is characterized by a highly interconnected and multiphasic neuropathological cascade of events in which an intense and protracted inflammatory response plays a crucial role in worsening brain injury. A large body of evidence indicates that post-ischemic inflammation is associated with acute blood-brain barrier (BBB) disruption, vasogenic edema, hemorrhagic transformation, and worse neurological outcomes in animals and humans. While neuroinflammation contributes to brain damage in the early phase of ischemic stroke, the inflammatory response could promote recovery at late stages by facilitating neurogenesis, angiogenesis, and neuronal plasticity. A better mechanistic understanding of how the neuroinflammatory response transitions from injury to repair would help identify novel strategies to intervene therapeutically in a time- and context-specific manner.

The BBB is a highly selective structural and functional barrier between the blood and the CNS, creating a unique microenvironment required for normal brain function and homeostasis. Comprising the BBB, brain endothelial cells interact with astrocytes, pericytes, nerve terminals, microglia, and perivascular macrophages in the neurovascular unit (NVU). The NVU concept was put forward to highlight the complex interactions among all the cells associated with the BBB (Figure 1)^1–3. A more detailed description of the structural and functional aspects of the BBB is provided in the Supplemental Material.

Figure 1. Schematic of the neurovascular unit (NVU).

Anatomically comprising the blood-brain barrier (BBB), the endothelial cells are surrounded by pericytes, astrocytic end-feet processes, neuronal terminals, and the basement membrane, comprising extracellular matrix (ECM) proteins that provide mechanical support and facilitate cell-ECM and cell-cell interactions at the NVU. Perivascular microglial cells contact the brain vasculature and modulate BBB permeability. Perivascular macrophages are key mediators in immune surveillance and are located in penetrating arterioles and post-capillary venules. Situated on the luminal side of the endothelial cells, the glycocalyx is composed of proteoglycans and their linked glycosaminoglycan chains, including heparan sulfate and hyaluronan. Glycocalyx degradation and shedding dramatically increase BBB permeability. An intricate complex of tight junction and adherens junction proteins “zip” together adjacent endothelial cells, conferring the low permeability of the BBB under physiological conditions. Claudins and occludin form the seal between endothelial cells, and they associate with the actin cytoskeleton via accessory proteins, such as ZO-1. Junctional adhesion molecules (JAMs) form part of the tight junctions and facilitate the attachment of endothelial cell membranes. The brain vasculature expresses high levels of platelet endothelial cell adhesion molecule-1 (PECAM-1), also known as CD31, which contributes to BBB function. Vascular endothelial cadherin (VE-cadherin) and catenins are the main structural proteins comprising the adherens junctions. They are necessary to form tight junctions and provide physical integrity to the BBB.

This review provides a critical evaluation of the role of neuroinflammation in ischemic stroke, with a particular emphasis on the BBB and peripheral immune responses. We provide an overview of the functional and structural elements determining the BBB phenotype and how this is disrupted after stroke by inflammation-driven mechanisms, including oxidative stress, matrix metalloproteinases, microglial activation, and peripheral immune cell invasion into the ischemic tissue. We discuss the role of peripheral immune cells and their cytokines in the resolution and recovery following ischemic stroke. We next provide an overview of methodologies and applications of medical imaging modalities to assess neuroinflammation after ischemic stroke and their potential implementation in clinical practice for image-guided diagnosis and treatment. We finally discuss the results of several clinical interventional trials targeting inflammatory pathways to reduce brain injury and improve outcomes.

Mechanisms of inflammation-driven BBB damage in ischemic stroke

Oxidative stress

As an essential contributor to cell death and BBB injury, excessive reactive oxygen species (ROS) production occurs shortly after vessel occlusion and is closely associated with the stroke-induced inflammatory response. Stroke also impairs endogenous antioxidant mechanisms, which leads to oxidative stress. ROS directly damage endothelial cells of the BBB and contribute to vasogenic edema. The superoxide radical (O₂^−.) is a primary ROS involved in increased vascular permeability after cerebral ischemia⁴ and O₂^−. radical scavenging reduces stroke-induced BBB permeability and vasogenic edema⁴.

In a transient mouse MCAO model, transgenic overexpression of the intracellular form of glutathione peroxidase (GPx1) reduces infarct size, decreases MMP-9, and prevents BBB disruption⁵. Conversely, GPx1-null mice display larger infarcts and a dramatic exacerbation of BBB leakage compared with wild-type controls⁶.

The Nox2/gp91^phox containing NADPH oxidase in microglia and infiltrating immune cells is a significant source of ROS after transient cerebral ischemia. Compared to wild-type controls, gp91^phox knockout mice display smaller lesions and reduced BBB breakdown after ischemic stroke⁷, suggesting that NADPH oxidase-driven oxidative stress plays a critical role in BBB injury in stroke.

Depending on the cell type in which it is generated, nitric oxide (NO) could have beneficial or deleterious effects after cerebral ischemia⁸. Increased NO formation by neuronal nitric oxide synthase (nNOS) is detrimental in the context of ischemic stroke⁹. Endothelial-derived NO is neuroprotective in models of ischemic brain injury through mechanisms involving increased CBF and attenuation of peripheral leukocyte invasion^8,10. However, excessive eNOS-derived NO generation under oxidative stress conditions could be detrimental via the formation of peroxynitrite, the product of the reaction between NO and O₂^−. radical⁸. After stroke, peroxynitrite formation is dramatically increased in microvessels and astrocytic end-foot processes, which is associated with BBB disruption and active MMP-9 labeling, suggesting an association between peroxynitrite and vascular damage⁸.

Matrix Metalloproteinases

As critical mediators of cerebrovascular damage after stroke, matrix metalloproteinases (MMPs) are a family of zinc-dependent proteases that are rapidly increased in activated microglia, astrocytes, and extravasated leukocytes, particularly neutrophils. Among the MMPs, MMP-2 (gelatinase A), MMP-9 (gelatinase B), MMP-3 (stromelysin-1), and MMP-12 have been extensively documented to contribute to BBB damage in stroke¹¹. MMP-dependent BBB disruption results from the degradation of the basal lamina and TJPs by active MMPs. The MMPs exist as zymogens, and several mechanisms activate these proteases during neuroinflammatory and hypoxic conditions. MMP-14 (also known as MT1-MMP) converts pro-MMP-2 into active MMP-2, and proMMP-9 can be activated by MMP-3 or oxidative modification¹¹.

Poor neurological function and hemorrhagic transformation in stroke patients correlate with elevated plasma MMP-9 levels following rtPA thrombolysis¹². BBB breakdown after stroke in humans correlates with increased MMP-9¹³, and postmortem studies have found MMP-9 mainly localized in infiltrating neutrophils, microglia, and brain endothelium¹⁴. As the injury progresses, neutrophils are the primary source of active MMP-9 in stroke¹⁵, and release of MMP-9 from human neutrophils is significantly increased by tPA. This mechanism contributes to hemorrhagic complications following rtPA thrombolytic therapy¹⁶.

There is a sustained increase in MMP-2 and MMP-9 after transient MCAO, and post-ischemic administration of MMP inhibitors reduces BBB opening and stroke damage¹¹. MMP-9 genetic deletion results in better functional outcomes and smaller infarcts in models of ischemic stroke¹⁷. Genetic or pharmacological inactivation of MMP-9 leads to less degradation of tight junction proteins and neurovascular protection¹⁷.

Most studies have focused on gelatinases (MMP-2 and MMP-9) in stroke pathophysiology. However, MMP-3 and MMP-12 are also implicated in neuroinflammation-mediated BBB damage after stroke. Compared to wild-type mice, MMP-3 knockout mice have reduced BBB leakage following delayed rtPA thrombolysis treatment¹⁸. In a model of transient cerebral ischemia in rats, MMP-12 levels are dramatically increased, and systemic administration of a plasmid expressing MMP-12 shRNA reduces stroke volume and BBB permeability¹⁹.

Targeting MMPs to reduce stroke injury has been proven difficult due to the lack of selective inhibitors with low toxicity. In addition to mediating a detrimental role in the acute stroke phase, MMPs are beneficial in the recovery phase by enhancing neuronal plasticity and vascular remodeling^20,21.

Role of inflammatory mediators released by microglia and peripheral immune cells in acute ischemic brain injury

Dying cells in the ischemic brain release many molecules that activate the surrounding viable tissue or infiltrating cells. These molecules include, among others: ATP, high-mobility group box 1 protein (HMGB1), uridine triphosphate, and heat shock proteins²². These danger signals or danger-associated molecular patterns (DAMPs) activate immune cells through the engagement of pattern recognition receptors (PRRs) such as Toll-like receptor family (TLR1 to 13) and suppression of tumorigenicity 2 (ST2). As a second signal, they can activate purinergic receptors and intracellularly the inflammasome family, which induces secretion of proinflammatory cytokines by innate immune cells²². One example of such a pathway is the ATP/P2X7 axis, which can activate the NLRP3 inflammasome and secretion of IL-1β. Overall, these events trigger proinflammatory intracellular signaling cascades and transcription factors, for example, NF-κB, ROS, MMPs, and the release of proinflammatory cytokines, especially IL-1β, −6, −17, −18, and tumor necrosis factor (TNF)-α, to initiate a local sterile immune response, which is associated with BBB dysfunction after stroke²².

The first cells to respond are microglia, the brain’s resident immune cells. Activated microglia migrate to the injured area and release proinflammatory cytokines, NO, ROS, prostaglandins, and chemokines, resulting in the additional chemoattraction of circulating leukocytes. Treatment with minocycline to block microglial activation markedly reduces infarct volume and BBB disruption, and improves neurological function²³, suggesting that inhibition of microglial activation during the acute phase of stroke is neuroprotective. However, emerging data indicate that microglia actively participates in neurorepair after stroke by releasing growth factors and anti-inflammatory cytokines such as TGF-β1²⁴.

Like other paradigms of sterile tissue injury, the innate immune cells comprise the predominant cellular composition of the brain infiltrates in the early acute post-stroke phase followed by lymphocytes. The dynamics of leukocyte invasion into the ischemic brain differ substantially among different cell types²⁵. After stroke, first microglia and then invading macrophages appear within minutes to hours after ischemia, followed by invading neutrophils peaking at three days post infarct²⁵. Their presence persists until day seven and declines afterwards²⁵. Atypical innate-like T cells such as γδ T cells and CD8+ cytotoxic T cells also arrive early²⁵, while conventional CD4 T cells arrive in a time window similar to neutrophils²⁵. T cells migrate preferentially to the lesion borders and can be detected throughout at least 30 days post-infarct in the brain parenchyma. Treg cells arrive several days after brain ischemia but are still present more than 30 days post lesion²⁵.

T cells play a significant role in secondary neuroinflammation and stroke outcome. Lymphocyte deficient transgenic mice have smaller infarcts after transient or permanent middle cerebral artery occlusion (MCAO) than immunocompetent control animals^25–31. Antibody-mediated depletion of single T cell subsets, namely CD8+, CD4+ and γδ T cells or their secreted cytokines, has also been shown to reduce secondary lesion progression^27,28,32. T cell-directed therapies in the acute phase after stroke are antigen-independent³⁰. Therefore, it is likely that innate-like lymphocytes or antigen-independent mechanisms of T cell activation are most important in acute ischemic tissue damage.

One of these innate-like T lymphocytes that have been implicated in tissue damage after stroke are the γδ T cells²⁷. They arrive early (within 6 h) and are marked by the expression of Vγ6 segment of the γδ T cells receptor and the chemokine receptor CCR6³³. Their signature cytokine in stroke is IL-17 which has been shown to harm the peri-lesional tissue. Correspondingly, γδ T cell-deficient mice have reduced infarct volumes^27,28, and neutralizing IL-17-specific antibodies can significantly improve stroke outcome²⁸.

Another critical cytokine is IL-1β. It is upregulated rapidly in experimental stroke and worsens ischemic injury in experimental models³⁴. Its release is usually triggered by TLRs’ engagement and purinergic receptors such as P2X7. The endogenous IL-1 receptor antagonist (IL-1Ra) reduces lesion size in experimental cerebral ischemia models, even in aged and comorbid animals and with delayed administration³⁴. A meta-analysis of preclinical stroke studies demonstrated that IL-1Ra reduced infarct volume by 36% in pooled data from 1283 animals³⁵. A recent human trial could verify a reduction in peripheral IL-6 levels after IL-1Ra treatment³⁶.

The ligand-receptor pair TNF/TNFR1 was suggested as relevant for stroke pathology by genome-wide association studies in the form of a polymorphism in the TNF gene that increases the susceptibility for stroke³⁷. After experimental stroke, TNF is mainly secreted by microglia, astrocytes, and invading macrophages and peaks at 12–24 h but remains elevated for days to weeks^25,38. Interestingly, TNF-knockout mice³⁸ and conditional TNF-KO mice with ablation of TNF in myeloid cells, including microglia³⁹, develop larger infarcts and worse behavioral deficits. Therefore, microglial-derived TNF mediated via TNFR1 is thought to be protective in stroke, while mice with a loss of TACE-mediated cleavage of TNF and, therefore, loss of soluble TNF develop smaller infarcts⁴⁰. Targeting soluble TNF experimentally, for example, with etanercept resulted in some studies in better behavioral outcomes but had no effect on infarct volume⁴¹.

Taken together, we can conclude that a lot of experimental data exists showing that acutely after an ischemic stroke, a robust sterile inflammation is triggered. This inflammation is induced by several different cell types and humoral factors. Inhibition of T cells, IL-17, IL-1β, or soluble TNF can experimentally result in better outcomes.

Role of peripheral immune cells and their cytokines in the resolution and recovery of ischemic brain injury

The peripheral immune cell most involved in the resolution of ischemic injury is the T regulatory (Treg) cell. Several reports have shown a protective role of Tregs in post-stroke neuroinflammation (reviewed by Liesz et al.⁴²). IL-10 is the primary mediator of Treg-induced effects. IL-10 is an anti-inflammatory cytokine that inhibits IL1-β, TNF-α, and potentially even IL-17 secretion⁴³. In experimental stroke, IL-10 mRNA and protein and IL-10R mRNA increase, and transgenic mice overexpressing IL-10 show reduced infarct volumes^44,45. Administration of IL-10 is protective in experimental stroke and limits post-stroke inflammation^43,46. Other sources of IL-10 are infiltrating immune cells such as B cells⁴⁷ and resident cells, for example, microglia or astrocytes^48,49. In addition, Treg might also affect BBB integrity during acute stroke. Besides effects on cytokines, adoptive Treg transfer also inhibits MMP-9 activity and protects the BBB integrity⁵⁰. Tregs, microglia, and cytokines like IL-10 limit the acute inflammatory response after ischemic brain injury. However, there is convincing data of a prolonged presence of inflammatory cells and cytokines in the brain after stroke, which might be necessary for repair efforts.

In part, the recovery after stroke resembles the processes occurring in the early stages of nervous system development⁵¹. T cells and locally proliferating T cells have recently been documented for the chronic phase⁵². Recovery after stroke can be triggered partly by the release of growth factors and cytokines secreted by both microglia and invading T cells⁵³. One exemplary mechanism for the pro-regenerative effect of T cells might be the production of BDNF⁵⁴. Also, the production of the Th2-signature cytokine IL-4 and the regulatory T cell-derived cytokine IL-10 has been associated with neuroprotective functions of T cells after brain injury⁵⁵. Moreover, T cell-derived IFN-γ can modulate social behavior by affecting GABAergic projections⁵⁶. T cells are also potent modulators of the local inflammatory micromilieu aside from direct neuroprotective or modulatory effects. The reaction to acute brain injury and repair is a multicellular process involving multiple non-neuronal cell populations that can potentially be affected by T cells patrolling the injured brain⁵⁷.

Similar observations are also valid for different proinflammatory cytokines. For example, the second peak in IL-17 expression occurs around day 28 after stroke⁵⁸. Recent studies show that IL-17A+ γδ T cells populate the meninges from perinatal stages and control neuronal signaling, behavior⁵⁹, and memory⁶⁰. IL-17 is likely also involved in the repair mechanisms after stroke. The fact that IL-17A can have these dual roles can be explained by the dependence of the IL-17A receptor signaling on synergistic stimulation by other cytokines, such as IL-1β and TNF-α⁶¹, which are also differentially upregulated in ischemic brain tissue⁶².

Under physiological conditions, TNF has an important role in modulating glutamatergic synaptic transmission and plasticity⁶³. In addition, the absence of TNF has been reported to impact cognition and behavior⁶⁴, making general strategy to block TNF in stroke difficult. The different receptors add to the difficulties. TNFR1 plays a role in sustaining chronic inflammation in the experimental allergic encephalomyelitis model, while TNFR2 is important for remyelination⁶⁵. Intracortically infused soluble TNFR1 for one week after photothrombotic stroke preserved axonal plasticity in the brain by competing for soluble TNF with TNFR1 receptors⁶⁶.

Taken together, most if not all parts of an inflammatory reaction have at least a dual role or, more likely multiple roles. While certain aspects in the acute stage might be harmful, they are later essential for repair and recovery. These events are context-specific and dependent on the receiving receptors and cells.

Imaging markers of neuroinflammation

Medical imaging modalities like CT, MRI, and PET have accelerated stroke diagnostics. These modalities are available in clinical and preclinical settings, providing ideal conditions for translational research. Over the years, numerous studies have demonstrated how imaging techniques can be applied for in vivo detection of BBB permeability, leukocyte infiltration, microglial activation, and upregulation of cell adhesion molecules. Here, we provide a brief overview of methodologies and recent applications of medical imaging modalities to assess neuroinflammation after stroke, including implications for clinical implementation.

BBB permeability imaging

BBB disruption can be measured from the extravasation of contrast agent or radioactive tracer, which results in image contrast enhancement. Quantification of BBB permeability is feasible with dynamic contrast-enhanced (DCE) or perfusion imaging protocols on MRI or CT scanners, for which paramagnetic or iodine-based contrast agents, respectively, are intravenously injected^67,68. The efflux rate of the contrast agent from plasma into the tissue can be calculated through pharmacokinetic modeling of DCE T₁-weighted MRI signal intensities, from which different permeability measures, such as the extraction fraction, blood-to-brain transfer constant, and permeability-surface area product, can be quantified⁶⁹. These measures can also be calculated from perfusion imaging with dynamic susceptibility contrast-enhanced (DSC) T₂^(*)-weighted MRI or dynamic CT⁷⁰, which enables fast acquisition at the expense of spatial resolution.

Most BBB permeability imaging studies use clinically available contrast agents like gadolinium chelates and iodinated contrast media. Alternative contrast agents with varying dimensions and pharmacodynamic properties allow assessment of different degrees of BBB opening and dynamics of BBB breakdown, but this has been limited to preclinical studies in animals⁶⁷. Recent developments with arterial spin labeling techniques, typically used for perfusion MRI, have shown promise for the non-invasive measurement of BBB permeability⁷¹. This approach utilizes water as a tracer rather than an injected exogenous contrast agent.

Increased BBB opening is associated with a reduced likelihood of a favorable neurological outcome at 90 days in stroke patients^72–74. Various MRI and CT studies have shown that BBB permeability imaging can aid in predicting hemorrhagic transformation after acute ischemic stroke (see Bivard et al. for a recent multicenter study⁷⁵ and Arba et al. for a recent systematic review and meta-analysis⁷⁶) (Figure 2, top panel). The association between BBB permeability and hemorrhagic transformation is particularly apparent in hypoperfused areas after thrombolytic treatment^77,78. A prospective study has recently demonstrated a significant correlation between pre-treatment BBB permeability and post-treatment hemorrhagic transformation in acute ischemic stroke patients who underwent perfusion CT before intravenous thrombolysis or endovascular thrombectomy⁷⁹. On the other hand, the absence of an association between (reversible) BBB permeability before thrombolysis and unfavorable clinical outcome after 90 days has also been reported⁸⁰.

Figure 2. Imaging neuroinflammation after stroke.

Multimodal brain images of different aspects of neuroinflammation (left images) and corresponding MR images of tissue injury (right images) after clinical or experimental stroke. Top: BBB permeability map, calculated from perfusion CT <4.5 h after ischemic stroke (before intravenous thrombolysis), and follow-up FLAIR MRI displaying hemorrhagic transformation after 24 h (after intravenous thrombolysis) (modified from Bivard et al.⁷⁵, licensed under CC BY-NC-ND 4.0). Middle: Map of [18F]DPA-714 uptake, a marker of activated microglia, calculated from PET, overlaid on a T2-weighted MR image of the ischemic tissue lesion, at 7 days after transient MCAO in mice (from Zinnhardt et al.¹¹⁸, licensed under CC BY-NC-ND). Bottom: T₂*-weighted MRI after injection of anti-VCAM-1 antibody functionalized MPIO, revealing upregulation of VCAM-1 in and around the lesion, shown on a T₂-weighted MR image, at 2 days after transient MCAO in mouse (from Gauberti et al.¹¹², reproduced with permission, Copyright Clearance Center, license# 5233371245585). Arrows or crosshair point at elevated BBB permeability preceding hemorrhagic transformation (top), maximal microglial activation in the lesion core (middle), and high VCAM-1 expression in the lesion borderzone (bottom).

Post-procedural hemorrhage and worse functional outcomes have also been associated with late contrast enhancement of cerebrospinal fluid (CSF) space on follow-up T₂-weighted fluid-attenuated inversion recovery (FLAIR) MR images of acute ischemic stroke patients⁸¹. This phenomenon was termed “hyperintense acute reperfusion marker” (HARM)⁸². CSF enhancement in patients may start focally in the acute phase after stroke and become widespread and diffuse throughout the subarachnoid space at later stages⁸³. Delayed CSF enhancement on post-contrast FLAIR images has been recognized as a marker of BBB leakage in different neurological disorders⁸⁴. Moreover, the retention of the contrast agent in the subarachnoid space may also inform on (deficiencies in) CSF dynamics and (‘glymphatic’) clearance mechanisms, as demonstrated with MRI in rodents after intrathecal or intracisternal infusion of a gadolinium chelate^85,86.

Immune cell imaging

It is possible to identify immune cells after labeling with contrast agents or tracer-binding to cell-specific markers. PET-based detection of the 18-kDa translocator protein (TSPO), an outer mitochondrial membrane protein, can be considered the gold standard for imaging of activated (residential) microglia or (infiltrated) macrophages (although TSPO may also be expressed by endothelial cells and astrocytes)⁸⁷. Over the years, different TSPO radiotracers have been synthesized and tested, also for other nuclear imaging methods, like SPECT, with varying levels of sensitivity and specificity for microglia/macrophage detection⁸⁷. This is partly due to human TSPO polymorphisms causing differential affinity for TSPO ligands, although this does not appear to be an issue in preclinical animal studies^87,88.

PET studies in animal models and patients have shown that TSPO expression is low in the first days after stroke, increases in the following weeks (Figure 2, middle panel), and normalizes again after several months^88,89. Nevertheless, activation of microglia, co-localized with secondary tissue degeneration, has been detected in non-infarcted areas of chronic patients beyond 6–12 months after stroke^90,91. Relatively little is known about the role of microglia activation in the development of post-stroke injury. Alternative PET tracers have been used to learn more about the profile and impact of microglial activation. A possible neuroprotective effect was suggested in a study in which early detection of post-stroke microglial activation, i.e., 24 h after experimental stroke in rats, was demonstrated with PET of the endocannabinoid receptor type 2 (CB2)⁹². Another study reported detection of gliogenesis, which may be part of restorative processes after stroke, with a PET marker for cell proliferation at day 7 after transient MCAO in rats⁹³. Uptake of the PET tracer [¹⁸F]-2-fluoro-2-deoxy-D-glucose (a marker for glucose consumption) in infarcted brain tissue has also been associated with activated microglia or macrophages⁹⁴. However, this may be obscured by other metabolically active cells in or around the lesion. Additional markers showing promise for PET-based assessment of post-stroke inflammatory responses include α7 nicotinic acetylcholine receptors⁹⁵ and adenosine A1 receptors⁹⁶.

Activated tissue-resident microglia and blood-borne macrophages may also be detected with MRI, enabling imaging at higher resolutions. The most frequently applied strategy for MRI of immune cells involves intravenous injection of ultrasmall particles of iron oxide (USPIOs), which may be internalized by circulating monocytes or phagocytosed by macrophages or microglia after extravasation^97,98. Superparamagnetic iron oxide particles are available in different sizes and configurations and induce hypointense or hyperintense signals on T₂^(*)- or T₁-weighted MR images, respectively. However, image contrast enhancement in (pathologic) brain areas after systemic injection of iron oxide particles may not always have a cellular origin, as these particles can get trapped intravascularly or accumulate interstitially without cellular uptake, possibly leading to false-positive assessments⁹⁹. Moreover, some types of iron oxide particles were shown not to accumulate in post-stroke mouse brains after intravenous injection¹⁰⁰. As an alternative strategy, cells can be labeled in vitro and subsequently injected to spatiotemporally track their distribution more specifically, as demonstrated for monocytes¹⁰¹ and T lymphocytes¹⁰² in rodents with an experimental stroke lesion. Nanoparticles specifically designed for phagocytic internalization may also improve cell-specific uptake. This was recently shown for a nanoprobe consisting of a gadolinium fluoride core coated with polyethylene glycol and functionalized with a fluorophore¹⁰³. The multimodal properties, i.e., being paramagnetic and fluorescent, enabled combined in vivo MRI and intravital (two-photon) microscopy, which confirmed uptake by phagocytic immune cells in a mouse stroke model.

In addition to multimodal probes, the development of multifunctional (theranostic) probes may allow combined image-guided diagnosis and treatment. For example, platelet-mimetic nanoparticles, coloaded with iron oxide and an anti-inflammatory agent, are internalized by activated neutrophils in a mouse stroke model, where they can exert therapeutic effects and be detected with MRI¹⁰⁴.

Only a few studies have evaluated the potential of immune cell imaging with MRI in humans. Pilot studies in stroke patients reported heterogeneous patterns of contrast enhancement after injection of USPIOs^105–108, which could be due to different forms of accumulation (intra- and extracellular) and variation in timings of contrast agent injection and post-contrast MRI acquisition. At present, the clinical potential of MRI-based detection of immune cells with superparamagnetic iron oxide particles as contrast agents remains unclear and requires further investigations. Alternatively, labeling autologous leukocytes with radioisotopes enables detection with nuclear imaging techniques, like SPECT and PET, but this remains an invasive procedure that is further hampered by the release of radiolabel from cells after injection¹⁰⁹.

Imaging molecular markers of inflammation

Molecular markers of neuroinflammation can be detected using contrast agents functionalized with ligands (e.g., peptides, proteins, or antibodies) that enable specific targeting and binding. This approach has been successfully applied to identify endothelial activation after experimental stroke with MRI^89,97,110. Different contrast agents have been used, but micron-sized particles of iron oxide (MPIOs) have become the most popular for molecular imaging purposes^110,111. Because of their size, these iron oxide particles generate large contrast-enhancing effects on MR images. They usually do not extravasate and remain inside the vasculature, allowing binding to molecular entities expressed on the luminal side of the cerebral endothelium. The relatively short plasma half-life (order of minutes) of MPIOs (as compared to many USPIO types (order of hours)) limits contrast effects arising from the circulating free agent.

Functionalization of MPIOs by conjugation with specific monoclonal antibodies has enabled in vivo MRI of the upregulation of P-selectin, E-selectin, VCAM-1, and ICAM-1 in rodent stroke models^89,97,110. These cell adhesion molecules have been detected in and outside the lesion territory. For example, Gauberti et al. observed strong contrast-enhancing effects around the ischemic lesion on T₂^(*)-weighted MR images after intravenous injection of anti-VCAM-1 antibody functionalized MPIO at 24 h after permanent MCAO in mice¹¹² (Figure 2, bottom panel). This inflammatory response preceded secondary tissue injury, and therefore the authors coined the term “inflammatory penumbra”. Strong expression of VCAM-1 in a transient MCAO model in rats has also been revealed with MRI in the first hours after recanalization, which could affect reperfusion dynamics and contribute to expansion of the cerebral infarct¹¹³.

At subacute to chronic stages after stroke, cerebrovascular inflammation can be detected with anti-ICAM-1 antibody functionalized MPIO¹¹⁴. Colocalization with ICAM-1-positive leukocytes inside the vasculature, days to weeks after transient MCAO in mice, suggested that targeting MR contrast agents to cell adhesion molecules may also be employed to identify immune cells. This has indeed been demonstrated for microglia or macrophages using Iba-1 antibody-conjugated superparamagnetic nanoparticles, measured in rats up to four weeks after transient MCAO¹¹⁵. Conversely, iron oxide particles may be coated with a natural leukocyte membrane to target activated endothelium, demonstrated using neutrophil-mimetic magnetic nanoprobes in post-stroke mice ¹¹⁶.

Another approach for in vivo imaging of neuroinflammation involves using responsive contrast agents. In a study by Breckwoldt et al., mice were injected with a gadolinium-based paramagnetic contrast agent specifically activated by MPO¹¹⁷. MPO is secreted by activated neutrophils and microglia/macrophages, which peaks in the first days after experimental stroke in rodents. Correspondingly, these authors found clear signal enhancement on T₁-weighted MRI, caused by the enzyme-activatable contrast agent, reflecting MPO activity that peaked on day 3 after transient MCAO.

Activation of MMPs, another family of enzymes involved in the neuroinflammatory cascade after stroke, can also be identified with medical imaging modalities. Zinnhardt et al. conducted an original dual-tracer PET study by injecting a radiofluorinated MMP inhibitor and a radiofluorinated TSPO ligand at different time-points after transient MCAO in mice¹¹⁸. Their study revealed differential time courses for MMP (maximal expression after seven days) and microglial activation (maximal expression after fourteen days). The presence of MMPs has also been visualized with ¹⁹F MR spectroscopic imaging, following injection of a fluorinated molecular ligand with a high affinity for MMPs¹¹⁹. The signal from this ligand, injected around 24 h after transient MCAO in mice, was significantly elevated inside the lesion area, which was further increased after tPA treatment, correlating with total MMP-2 and MMP-9 activity.

Future implications for clinical and preclinical imaging of neuroinflammation

Although many preclinical studies have demonstrated the ability of medical imaging modalities to assess neuroinflammation, most methods are not routinely utilized for stroke diagnosis in clinical practice. Some procedures, like BBB permeability imaging, have shown considerable promise in patient studies and could be easily implemented. BBB permeability imaging could aid in the identification of patients who can be safely treated with intravenous thrombolysis or endovascular thrombectomy, even beyond the currently recommended time windows for these treatments. However, standardized protocols for image acquisition and data analysis would need to be developed.

Imaging cellular or molecular markers of neuroinflammation may provide complementary diagnostic information after stroke. Yet, several methods await further preclinical evaluation of sensitivity, specificity, and safety. Developing biocompatible and biodegradable contrast agents with good labeling efficiency for human cells or strong binding affinity to human-specific epitopes will be critical for effective clinical implementation. Translational studies should continue to explore the potential of imaging methods to elucidate the role of different inflammatory processes, identify treatment targets and monitor immunomodulating interventions after stroke. Multimodal and multifunctional (theranostic) (nano)probes that can be detected with different imaging modalities and deliver therapeutic agents could be particularly promising for image-guided diagnosis and treatment.

Translation to Humans and Concluding Remarks

While the data from animal studies support a prominent role of inflammatory cells in the pathogenesis of stroke, human data, let alone positive human studies, are scarce. Some histopathological data documents inflammatory cells’ presence in acute and chronic stages after ischemic stroke^120,121. However, the overall involvement of T cells seems to be less in human lesions and more prone to MHC class I-restricted CD8+ T-cells compared to the broader spectrum of lymphocytes in rodents¹²¹. Microglia and macrophages seem to follow similar patterns in human and rodent stroke, but in aged human stroke lesions, these cells often possess a pre-activated phenotype not seen in young rodents.

So far, our attempts to translate the experimental findings to a clinical application have failed to a large extent. Early studies blocking adhesion molecules (ICAM-1, MAC-1) or the recombinant neutrophil inhibitory factor have been ineffective in clinical trials¹²². For the ICAM-1 study, the negative outcome was attributed to deleterious immunoactivation resulting from the administration of a mouse antibody to humans. Other factors that likely contributed are the difference in human and rodent immune systems and different age structures in humans and rodents; blocking one inflammatory signal might not be enough and a failure to select the right patient cohort. This selection problem is well documented by the only preclinical randomized controlled trial (pRCT) followed by a human trial. In their pRCT, Llovera and colleagues investigated the effects of CD49d-specific antibodies¹²³. A significant reduction in leukocyte invasion and infarct volume was only documented for permanent MCAO but not for temporary MCAO. Subsequent human trials (ACTION 1 and 2) using CD49d-specific antibodies failed to find any effect on stroke progression^124,125. However, the variance in stroke volume with 1–212 ml was comparable to the non-significant tMCAO (0.001–0.095 ml) but not to the significant pMCAO results (0.003–0.016 ml)¹²³. A much better selected patient cohort might have yielded a different result. However, there is another explanation for this discrepancy. A follow-up study showed that a single application of CD49d-specific antibodies only temporarily blocks T cell invasion and T cells accumulate in the brain afterwards⁵², making it difficult to compare the results of the acute rodent studies to the three-month follow-up time point in humans.

The use of large gyrencephalic animal models, including sheep, dogs, swine, and nonhuman primates, has been recommended by the Stroke Therapy Academic Industry Roundtable to complement the preclinical testing in rodents and enhance the clinical translation of promising cerebroprotective strategies. Large animal stroke models could aid in the selection of the most viable therapies to test in clinical trials^126. A recent study in swine subjected to stroke via an endovascular approach showed that while gadolinium enhancement on T1-weighted MR images was marginal at 24h, likely the result of restricted access of the contrast agent due to limited perfusion of the ischemic territory, immunoglobulin extravasation into the ischemic territory was elevated both (sub)acutely (24h) and chronically (up to 3 months)¹²⁷, which is similar to findings from clinical studies showing persisting BBB dysfunction in ischemic stroke^128,129.

Despite the mostly neutral or negative outcomes of clinical trials, there are some positive preliminary findings in ischemic and hemorrhagic stroke for anti-inflammatory treatments, such as with lymphocyte-trapping agent fingolimod^130–132, approved to treat relapsing forms of multiple sclerosis. While the data presented is compelling, these were non-blinded, single-center trials that need to be repeated in larger cohorts. Also, intravenous IL-1Ra in patients with acute stroke significantly reduces plasma inflammatory markers¹³³. In a recently published study, IL-1Ra significantly reduced IL-6 and C-reactive protein plasma levels. IL-1Ra was not associated with a favorable outcome, but the study was not powered to provide significant outcome data.

Despite reports of improved neurological outcomes in patients with stroke or traumatic brain injury treated with perispinal etanercept^134,135, none of the currently used anti-TNF therapeutics have so far been approved as a neuroprotective strategy in combination with tissue plasminogen activator treatment.

According to the ClinicalTrials.gov registry of clinical intervention trials, more than 40 are exploring anti-inflammatory strategies with pending outcomes in ischemic stroke. They include a broad spectrum of potential therapies, including cell therapies, modulating lymphocyte migration, or repurposing drugs approved for multiple sclerosis.

Thus far, the lack of positive clinical data using anti-inflammatory therapies suggests that our understanding of the role of neuroinflammation in stroke is insufficient. It is essential to recognize and mechanistically determine the multiphasic nature of inflammation at different stages of the ischemic injury to better develop and implement anti-inflammatory strategies as stroke treatments. Many factors could explain the failure to translate animal studies into the clinic¹³⁶. Before conducting clinical trials, potential anti-inflammatory therapies should be evaluated in clinically relevant experimental models that incorporate aging, comorbidities, and gender to better mimic the human stroke population. Moreover, to better assess responses to treatment, direct imaging of drug delivery could be valuable to confirm arrival of the drug in the ischemic territory.

Non-standard Abbreviations and Acronyms

BBB

Blood-brain barrier

CBF

Cerebral blood flow

CNS

Central nervous system

ECM

Extracellular Matrix

ICAM-1

Intercellular adhesion molecule-1

IL

Interleukin

JAM

Junctional adhesion molecule

MCAO

Middle cerebral artery occlusion

MMP

Matrix metalloproteinase

NO

Nitric oxide

NVU

Neurovascular unit

ROS

Reactive oxygen species

TNF-α

Tumor necrosis factor α

TJP

tight junction protein

Treg

T regulatory cell

VCAM-1

Vascular cell adhesion molecule-1

ZO

zona occludens