First described by Hippocrates nearly 2,400 years ago, stroke, or “apoplexy,” is the second leading cause of death globally and accounts for approximately 1 out of every 19 deaths in the United States (1). Stroke is also a leading cause of long-term disability and places a high economic burden on global healthcare systems. Despite significant advances in primary and secondary stroke prevention, the annual number of strokes and stroke-related deaths have persistently increased over the past two decades (2). An estimated 87% of strokes are ischemic (3), in which a sudden interruption in cerebral blood flow results in rapid cell death within the ischemic core. Specifically in the neuropathological progression of stroke, neuronal necrosis and apoptosis result in profound neuroinflammation and secondary tissue injury, which can be counterproductive to both short and long-term recovery (4, 5).

Current treatment strategies for acute thrombotic or embolic stroke are focused on early reperfusion with intravenous thrombolytics or mechanical thrombectomy, supplemented by supportive care and acute complication management. At present, alteplase is the only Food and Drug Administration (FDA)-approved medical therapy in the United States for the treatment of acute ischemic stroke (6). Mechanical thrombectomy with or without intravenous thrombolysis has revolutionized the treatment of stroke. Similar progress in treating secondary inflammatory injury is necessary to optimize patient outcomes.

At the onset of ischemic stroke, activated, amoeboid microglia rapidly migrate to the site of ischemic injury. The reduction in cerebral blood flow to the area of injury initiates a sequence of events often referred to as the ischemic pathway, which is notable for energy depletion and glutamate excitotoxicity leading to cell death (76). The ischemic core, defined as the region in which irreversible cell death has occurred, is characterized by elevated ion concentrations and glutamate levels as well as tissue acidosis (76). Elevated glutamate levels drive the release of microglial chemoattractants, such as ATP, CCL21, CXCL10, and IL-1β, leading to microglial recruitment (77). Imaging studies of ischemic stroke mouse models have found that within the first 24 hours of ischemia, and as early as thirty minutes after stroke onset (78) activated microglia (as characterized by an amoeboid morphology) populate the area of injury (19, 79). Adenosine triphosphate (ATP) release by dying neurons serves as a chemoattractant which activates purinergic receptors, such as P2Y12, on microglia and guides their directional migration to the site of injury (80). Downstream signaling pathways then enable chemotaxis via adhesion disassembly (81). Once microglia reach the site of injury, they are activated by damage-associated molecular patterns (DAMPs). High mobility group box 1 (HMGB1) is one such DAMP secreted by dying neurons following ischemic stroke. HMGB1 binds to TLR4 on microglia and induces production of pro-inflammatory cytokines, including interleukin (IL)-6 and tumor necrosis factor-alpha (TNF-alpha). Heat shock proteins, such as Hsp70, also activate microglia via nuclear factor-kappa B (NF-kB) activity (21). Microglia in turn secrete matrix metalloproteinase (MMP)-9, inducible nitric oxide synthase (iNOS), and reactive oxygen species (ROS) (81). These activities are temporally coordinated as some studies suggest that during the earlier stages of injury, microglia primarily promote tissue repair and reconstruction of the extracellular matrix (82, 83). In the later stages of injury, microglia shift towards a pro-inflammatory phenotype in which they release inflammatory cytokines and generate ROS (82, 83). This transition is not absolute, however, as other authors have shown that activated microglia can also secrete anti-inflammatory cytokines, such as transforming growth factor-beta (TGF-β) and IL-10 (84) at later timepoints ( Figure 2 ).

In a transient ischemic rat model, microglia activation was evident 3.5 hours after the onset of ischemia (79). In a permanent ischemic mouse model, microglia featuring hypertrophic cell bodies and shortened processes were observed as early as the 30-minute time point (79, 85). Activated microglia are also recruited to the penumbra, the potentially salvageable region of brain tissue surrounding the ischemic core. In the penumbra, activated microglia are notably associated with blood vessels. Some studies have suggested that perivascular activated microglia play an important role in blood vessel repair as well as promoting the integrity of the BBB, which is compromised during ischemic stroke (86, 87). However, the role of perivascular activated microglia in mitigating tissue injury remains controversial, with other studies suggesting that they promote blood vessel disintegration (88).

Gender and aging also influence the microglial response to stroke. Mouse model studies have demonstrated that resting state male microglia exhibit higher expression of inflammatory genes regulated by the NF-kb transcription factor compared to female microglia, and male mice subjected to cerebral ischemia developed larger infarcts than female mice (89, 90) The difference in infarct size may be due to increased expression of genes significant for cellular plasticity in female microglia compared to male microglia (91). Additionally, aged mice have been found to experience more severe neurologic deficits following MCAO as well as increased serum levels of inflammatory cytokines compared to young mice (92). One hypothesis for this finding is the upregulation of IRF5 signaling which occurs with aging (92). Microglia in aged mice have been found to resist phenotypic transformation to an M2 anti-inflammatory state in response to IL-4 (93). Therefore, male sex and increased age may both impair anti-inflammatory aspects of the microglia response to ischemic injury.

Mobilization of monocytes from the bone marrow to the blood and ultimately into the infarcted tissue is dependent on the CCL2/CCR2 axis (59, 94, 95). CCL2 is produced by astrocytes and microglia in states of hypoxia, as it is under direct transcriptional control of Hypoxia-inducible factor 1-alpha (HIF-1alpha) (58). Because CCL2 exclusively binds to CCR2, which is only expressed on the classically activated, pro-inflammatory monocyte subset, this is the subset recruited to the ischemic site, a finding supported in experimental models of brain ischemia and hemorrhage (10, 72, 96). Garcia-Bonilla et al. utilized ischemic stroke models with CX3CR1^GFP/+CCR2^RFP/+ bone marrow (BM) chimeric mice to study the effects of CCR2 and CX3CR1 on monocyte/macrophage recruitment following stroke and showed that non-classical, alternatively activated CX3CR1⁺ monocytes/macrophages were absent in the brain of CCR2 null mice. Furthermore, they found that while circulating hematogenously-derived alternatively-activated monocytes were absent in NR4A1-deficient mice (a nuclear receptor responsible for the differentiation and survival of non-classical, alternatively activated monocytes), these mice still had increased alternatively-activated monocytes in the brain 14 days after stroke occurred (51, 75). This led the authors to conclude that the alternatively-activated monocytes were, in fact, derived from the classically-activated monocytes once they had arrived at the site of ischemia, and were not produced de novo in the bone marrow.

Breakdown of the BBB facilitates the transmigration of cells including monocytes from the periphery into the site of injury. Nadareishvili et al. quantitatively measured gadolinium leakage through serial MRIs to assess the degree of BBB disruption after thrombolytic therapy for acute stroke and found that increased BBB permeability was associated with worse outcomes in patients independent of the severity and size of stroke. In fact, for every 1% increase in BBB permeability there was a 75% decrease in the chance of a good long-term functional outcome (97). BBB permeability has been shown to occur in two phases: reversible disruption occurs early after ischemic onset and is driven by the release of MMP2, later BBB disruption is mediated by MMP-3, MMP-9, and cyclooxygenases days after the ischemic insult (98). This second wave of BBB disruption allows for the infiltration of systemic immune cells, including neutrophils, dendritic cells, T-cells, NK cells and monocytes. Endothelial activation also results in chemokine release and upregulation of adhesion molecules, both of which facilitate recruitment and transmigration of circulating monocytes. Endothelial and tissue resident macrophages regulate the infiltration of monocytes and neutrophils through the production of cytokines such as granulocyte macrophage-colony stimulating factor (GM-CSF) (99). Ischemia also triggers expression of adhesion molecules such as selectins, integrins, and intercellular and vascular adhesion molecules along the microvasculature, which facilities rolling and high affinity interactions for firm adhesion of circulating leukocytes (100).

Anatomically, the choroid plexus may be a particularly important route of monocyte recruitment and ingress into the infarcted brain region. Through gene expression studies and the use of chimeric mouse models, Ge et al. found that the choroid plexus responded to stroke by upregulation of several key mediators of MoDM trafficking (such as Vcam1, Madcam1, Cxcl1, CCL2, NT5E, and IFNy) which resulted in increased trafficking of MoDMs into the choroid plexus and CSF. If primed for the M2-phenotype in vitro using treatment with macrophage-colony stimulating factor (M-CSF), IL4, and IL13 prior to administration into the lateral ventricle ipsilateral to the ischemic lesion, MoDMs homed to the area of ischemia and promoted post-stroke recovery and improved cognition (101).

Once monocytes have infiltrated the area of ischemia, the ischemic environment promotes their differentiation into macrophages. While CCR2⁺ monocytes retain a round, amoeboid phenotype and are limited to the ischemic core, CX3CR1⁺ monocyte/macrophages can adopt three different phenotypes in the ischemic brain based on where they are localized relative to the infarct core: in the penumbra, they generally take on a ramified phenotype similar to homeostatic microglia, while in the infarct region, they take on an amoeboid, phagocytic macrophage phenotype, and when associated with blood vessels, resemble perivascular macrophages (75). Using the C-X-C Chemokine Receptor Type 4 (CXCR4) signature to trace cells of hematopoietic stem cell origin, it was found that monocyte infiltration occurs in both the peri-infarct and infarct areas after transient MCAO (102). In a photothrombosis infarct model, infiltration primarily occurred in the peri-infarct region. Interestingly, in CXCR4 knockout mice that underwent photothrombosis, monocyte infiltration and microglial proliferation were both reduced, suggesting that MoDMs are responsible for microglial repopulation of the infarct core. Furthermore, the authors also demonstrated that MoDMs were the main source of microglia-activating mediators following photothrombosis and maintained microglial activation in the peri-infarct region until they were cleared (102).

Stroke is clinically staged into the acute, subacute, and chronic periods ( Figure 3 ). The acute period is generally defined as the period of minutes to days following the ischemic insult, while the subacute period refers to the time from days to weeks following stroke, and the chronic period refers to the time period from weeks to months and beyond. The chronic stage can last for years and continue for the remainder of a patient’s life (8). Depending on the stage of stroke, microglia and MoDMs are preferentially polarized towards different activation states in order to carry out specialized functions. Mismatch between these activation states and the timing of recovery, or inflammation that becomes chronic can both be detrimental to long-term outcomes.

Clinical therapies for stroke are currently focused on salvaging the penumbra in the acute phase via a combination of reperfusion techniques including intravenous thrombolytics and mechanical thrombectomy. Treatment strategies for the ensuing brain edema are mainly supportive, involving interventions such as hyperosmolar therapy, corticosteroids, hyperventilation and CSF diversion to reduce intracranial pressure during the acute swelling period. By the subacute and chronic stages, the treatment focus shifts to recovering function through physical and neuropsychological rehabilitation (137). At present there are no approved therapies that target the inflammatory pathways to limit secondary injury. Several pharmacological agents have been explored for their ability to promote neuroplasticity and neurogenesis during the later stages of stroke, including antidepressants, amphetamines, and neurotrophins (138). However, randomized controlled trials for these agents have failed to meet their efficacy endpoints (139).

The induction or promotion of the anti-inflammatory M2 microglia phenotype is one treatment strategy that has been employed in both preclinical and clinical studies to decrease inflammation in the acute and subacute phases and promote brain plasticity through in the chronic phase (113). Promotion of the M2 phenotype can be accomplished by administering molecular compounds which activate specific cell signaling pathways, such as the STAT3, AMPK, and PPARγ pathways (140–142). In preclinical studies, intravenous injection of compounds, such as xuesaitong (a Chinese patent medicine) and Ac2-26 (an annexin/lipocortin 1-mimetic peptide), as well as intraperitoneal injection of compounds such as recombinant human fibroblast growth factor 21 (rhFGF21 and melatonin, have been reported to promote the M2 microglia phenotype by modulating cell signaling (143–146) ( Table 1 ). HP-1c, an activator of the AMPK and Nrf2 pathways, as well as CDDO-EA, an activator of the Nrf2 pathway, also promote the M2 microglia phenotype (29, 30). Inhibition of polarization to the M1 phenotype via inhibition of the PDE5 pathway is another treatment which has shown preclinical success. Administration of sildenafil, a PDE5 inhibitor, after MCAO has been found to decrease the number of M1 microglia during the later stages of stroke as well as reduce the extent of the ischemic lesion (149). Nitric oxide (NO) and hydrogen sulfide (H2S)-releasing hybrid) (NOSH-NBP) has been similarly found to promote the M2 microglia phenotype in murine models of cerebral ischemia (152). Furthermore, clinical trials evaluating the safety and efficacy of NBP in mild to moderate ischemic stroke patients are ongoing with results still pending ( Table 2 ). Minocycline is an antibiotic found to promote the M2 microglia activation state in preclinical studies and has demonstrated some preliminary success for improving neurofunctional recovery following acute ischemic stroke in early clinical trials (161, 162). The MINOS (minocycline to improve neurological outcome in stroke) study was a phase 1 open-label, dose-finding study which found that minocycline could be safely tolerated in acute stroke patients at intravenous doses of up to 10 mg/kg (163). In a multicenter prospective randomized open-label pilot study of intravenous minocycline in a small sample of acute stroke patients, Kohler et al. reported that minocycline was safe but not efficacious. However, this study was not powered to reliably identify modest differences in clinical outcomes (164). In a small, open-label evaluator-blinded trial, Amiri-Nikpour et al. reported a gender-dependent effect on minocycline neuroprotection in ischemic stroke, noting improved clinical outcomes (lower National Institutes of Health Stroke Scale (NIHSS) scores) in minocycline-treated male patients, but no significant difference in minocycline-treated female patients (165). Of note, in these clinical trials, minocycline was administered as an oral or intravenous formulation to both ischemic and hemorrhagic acute stroke patients exclusively during the acute phase (161, 164–166). Additional clinical trials which test the efficacy of such pharmacological compounds are needed to determine whether modulating microglia phenotype can lead to improvement in neurologic outcome post ischemic stroke.

Several other drugs with known immunomodulatory properties have been repurposed in an effort to downregulate stroke-induced neuroinflammation. Montelukast, an anti-asthmatic drug, is a compound which has been found to promote microglia polarization to the M2 phenotype in mice (155). Montelukast acts as an antagonist of the CysLT-1 receptor and has been found to increase the number of M2 phenotype microglia during the acute phase of stroke (155). Edaravone is a free radical scavenger used in the treatment of amyotrophic lateral sclerosis (ALS), which has been found to promote the M2 microglia activation in preclinical studies (160, 167). In a retrospective study by Enomoto et al, clinical outcomes of patients who underwent endovascular reperfusion therapy and edavarone therapy within two days of admission were compared with patients who underwent endovascular reperfusion alone. In the group that received edavarone, the authors reported significantly lower in-hospital mortality and greater functional independence at discharge (168). Additional clinical trials designed to evaluate the efficacy of Edaravone are in preparation or currently in progress ( Table 2 ). Several clinical trials are investigating the efficacy of combining Edaravone with dexborneol, a food additive which has demonstrated anti-inflammatory effects in preclinical stroke models (169). Interestingly, the combined treatment of Edaravone with dexborneol has been found to have a greater benefit for female patients compared to male patients, suggesting that sex may significantly influence the efficacy of such treatments (169). Fingolimod, a sphingosine l-phosphate receptor modulator that is FDA-approved for relapsing multiple sclerosis, is another drug that was shown in preclinical studies to skew microglia towards the M2 phenotype following chronic cerebral hypoperfusion and is now being studied in several clinical trials. In one trial of 25 patients with acute hemispheric ischemic stroke, combined therapy of fingolimod and alteplase was associated with fewer circulating leukocytes, smaller infarct volumes, attenuated reperfusion injury and improved functional outcomes compared to alteplase monotherapy (170).

Exosome therapy has also been used to increase the overall number of M2 microglia in the post-stroke environment. This strategy uses extracellular vesicles (EVs) derived from multipotent mesenchymal stromal cells to deliver proteins, lipids, and/or nucleic acids. In the context of stroke, EVs have been shown to facilitate transfer of miRNAs between cells, which can influence post-transcriptional gene regulation in microglia to increase the expression of M2 markers (171). In traumatic brain injury mouse models, delivering miR-124-3p via EVs derived from microglia cells has been shown to promote the anti-inflammatory phenotype (172). Furthermore, infusion of EVs derived from microglia treated with IL-4 into mice following MCAO has been found to effectively promote microglia polarization to the M2 phenotype during the chronic stage of stroke. IL-4 treatment promotes functional recovery by enhancing the neuronal myelination capacity of GPR17-expressing oligodendrocyte precursor cells (111). Therefore, exosome therapy is another promising, microglia-focused therapy for ischemic stroke.

Compared to microglia-focused therapies, treatments targeting MoDMs are more limited. Preclinical studies have focused on manipulating the pro- and anti-inflammatory monocyte subtypes, as well as administering anti-inflammatory MoDMs as a form of therapy. A study by Schmidt et al. using clodronate liposomes to deplete peripheral macrophages showed no beneficial therapeutic effect after ischemic stroke in mice (153). However, when monocyte-derived macrophages were skewed to an M2 phenotype in vitro prior to administration into the CSF of mice after MCAO, improved cognitive and motor function was observed although there was no difference in infarct volume (101). In another preclinical study, Yang et al. demonstrated that shifting blood monocytes toward a CCR2+ proinflammatory state using remote ischemic limb conditioning (RLC) prior to stroke onset reduced brain injury and improved recovery, and that adoptive transfer of CCR2 deficient monocytes abrogated the proinflammatory shift and resulted in worse functional outcomes (157). Another study utilized hypoxic preconditioning prior to MCAO to upregulate CCL2, the receptor for CCR2, and found that it resulted in a neuroprotective phenotype with reduced infarct volume, blood-brain barrier disruption, and leukocyte migration during MCAO (58). These findings emphasize not only the importance of pro-inflammatory monocytes in stroke recovery, but also the benefit of manipulating peripheral immune cells or their chemokine signals before infiltration into the brain. There may be potential for future preventative strategies to shift monocytes to a CCR2+ subset in the acute phase of stroke or prime the microenvironment for CCR2+ cell migration earlier in the post-stroke process, such as through remote ischemic limb conditioning or hypoxic preconditioning. Furthermore, adaptive cell therapy or autologous transplantation of M2-like MoDMs into the CSF could be a promising avenue for treatment, but this will require additional studies to optimize timing of administration, dosage, and efficacy in humans.

Approaches to skewing peripheral monocytes into the M2-like phenotype have targeted factors such as PPARγ, NR4A1, and micro-RNAs (21 and 146-a) (173). Using a model of stroke-prone spontaneously hypertensive rats, Nakamura et al. demonstrated that pioglitazone, a PPARγ agonist, was protective against hypertension-induced stroke by inhibiting macrophage infiltration and suppressing the expression of inflammatory cytokines CCL2 and TNF-α (174). NR4A1, a pro-oncogenic nuclear receptor, is integral to the differentiation of classical monocytes into the M2 anti-inflammatory phenotype and may be a potential therapeutic target (175). MicroRNAs (miRNAs) have been shown to play an integral role in regulating monocyte development and function. MiRNA 146-a has been the most extensively studied and shows the largest difference in expression between classical and non-classical monocytes, with non-classical monocytes featuring higher expression. Depletion of miRNA 146-a augments the pro-inflammatory response of classical monocytes (176).

Timing of microglia and monocyte-derived macrophage migration and activity is a key consideration in developing effective therapeutic strategies. The current model is that the anti-inflammatory functions of activated microglia in the acute phase wane in the subacute phase. During this period, MoDMs fulfill an anti-inflammatory role, and the activated microglia shift toward a pro-inflammatory state. This delayed infiltration of MoDMs makes them potential candidates for immunomodulation in the subacute phase. However, our knowledge of immune cells present at each stage of ischemic stroke is currently limited to specific “snapshots” of cell populations, measured primarily through techniques such as flow cytometry and immunohistochemistry. A more fluid understanding will be necessary to target the correct cell type at the correct time (4). Future studies must also consider potential downstream effects of eliminating conventionally proinflammatory cells during each phase of stroke. As discussed, pro-inflammatory cells may have the capability to differentiate into anti-inflammatory subtypes, and therefore may be sensitive to the dynamic changes in the tissue microenvironment. Thus, nonspecific deactivation of MoDMs may decrease local tissue damage, but may also disable subsequent debris clearance and other repair mechanisms. Further elucidation of the specific contexts in which these activities occur will be critical in developing targeted immunotherapeutics.

Immunotherapies that target monocytes and MoDMs have already shown promise in neurological disorders such as encephalitis, multiple sclerosis (MS), and aneurysmal subarachnoid hemorrhage (SAH) (177–181). In a broader sense, inflammation has been shown to contribute significantly to the pathogenesis of many peripheral and central nervous system diseases, including but not limited to fibromyalgia, neuropathic pain, Alzheimer’s disease, Parkinson’s disease, and traumatic brain injury (182). A core pattern of activation by resident microglia followed by systemic myeloid cells has been shown in the aforementioned neurological diseases, with cells displaying an initial state of proinflammatory activation and a later anti-inflammatory phenotype (183). Though the inflammatory response is similar between ischemic stroke and these other neuropathologies in the sequential activation and deactivation of the innate and adaptive immune response, ischemic stroke is unique among neuropathologies in that it is caused by an acute insult that, left untreated, rapidly results in oxidative stress, apoptosis, and inflammation, leading to a massive immune response. Unlike neurodegenerative diseases, which have a more chronic, persistent immune response characterized by continuously active microglia and infiltrating leukocytes, ischemic stroke has different phases of inflammation, ranging from the acute to chronic stage. The complex orchestration of the immune response in ischemic stroke is highly dependent on the switch in activation states of microglia and MoDMs at specific times following the onset of the insult. In sum, the main differences between the way inflammation contributes to the progression of individual neurological diseases arise in a disease-specific and lesion stage-specific manner with regard to the contribution of resident versus recruited myeloid cells and their activation profiles during each stage. Ultimately, targeting the progression of neuroinflammation may be of translational benefit for a wide variety of neurological diseases.

In the setting of ischemic stroke, microglia and MoDMs phagocytose cellular debris and mediate the inflammatory response by adopting pro- and anti-inflammatory activation states. These activation states are dependent upon environmental and temporal factors, and available studies suggest that inducing a phenotypic switch in microglia and MoDMs may promote stroke recovery. There are important limitations to translating this work. Markers of differentiation between microglia and monocyte-derived macrophages have historically been lacking; however, RNA sequencing data have elucidated more specific markers such as Tmem119, paving the way for targeted studies of the spatiotemporal dynamics of microglia and MoDMs in the setting of ischemic stroke. Persistent inflammation during the chronic stage of stroke is associated with impaired neurofunctional recovery, and there are no current treatments for stroke in the chronic stage beyond rehabilitation. So far, clinical trials have identified compounds which can induce anti-inflammatory microglia and MoDM activation states; however, the clinical efficacy of these compounds has yet to be confirmed. Furthermore, clinical trials have largely focused on the acute phase of stroke. To optimize neurofunctional outcomes of ischemic stroke patients, it may be necessary to apply specific immunotherapies across the spectrum of acute, subacute, and chronic inflammatory events following stroke. Future research will determine precisely which pathways should be targeted and when.

EW and KR wrote and edited the manuscript. JK was responsible for the primary literature review and edited and revised the manuscript. RX and RL edited the manuscript. CJ conceived of the manuscript and oversaw the literature review, organization and writing, and edited the final version of the manuscript. All authors contributed to the article and approved the submitted version.

CJ is a scientific co-founder of Egret Therapeutics with equity interests in the company and inventor on a patent filed by Johns Hopkins for using PD-1 agonists to treat cerebral vasospasm and ischemia.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.