The Role of Microglia in Ischemic Preconditioning

Abstract

Ischemic preconditioning (IPC) is an experimental phenomenon in which a brief ischemic stimulus confers protection against a subsequent prolonged ischemic event. Initially thought to be due to mechanistic changes in neurons, our understanding of IPC has evolved to encompass a global reprogramming of the CNS after transient ischemia/reperfusion that requires innate immune signaling pathways including Toll-like receptors (TLRs) and type I interferons. Microglia are the CNS resident neuroimmune cells that express these key innate immune receptors. Studies suggest that microglia are required for IPC-mediated neuronal and axonal protection. Multiple paradigms targeting TLRs have converged on a distinctive type I interferon response in microglia that is critical for preconditioning-mediated protection against ischemia. These pathways can be targeted through administration of TLR agonists, cytokines including interferon-β, and pharmaceutical agents that induce preconditioning through cross tolerance mechanisms. Transcriptomic analyses and single cell RNA studies point to specific gene expression signatures in microglia that functionally shift these mutable cells to an immunomodulatory or protective phenotype. Although there are technological challenges and gaps in knowledge to overcome, the targeting of specific molecular signaling pathways in microglia is a promising direction for development of novel and effective pharmacotherapies for stroke. Studies on preconditioning in animal models, including non-human primates, show promise as prophylactic preconditioning treatments for selected at risk patient populations. In addition, our growing understanding of the mechanisms of IPC mediated protection is identifying novel cellular and molecular targets for therapeutic interventions that could apply broadly to both acute stroke and chronic vascular cognitive impairment patients.

Graphical Abstract

Introduction

The brain’s resistance to ischemic injury, or ischemic tolerance, can be transiently augmented by prior exposure to a non-injurious preconditioning stimulus (Gidday 2006). The first in vivo report of cerebral preconditioning was published more than fifty years ago (Dahl and Balfour 1964) with demonstration that a pre-exposure to a brief period of anoxia resulted in increased anaerobic glycolysis with no corresponding decrease in the rate of ATP utilization in the rat brain, which led to prolonged survival following a subsequent extended anoxic exposure. In 1986, Schurr et al., provided the first electrophysiologic demonstration of cerebral ischemic tolerance by showing that the amplitude of an electrically-evoked population spike in hippocampal CA1 neurons was maintained during anoxia if the slice was exposed earlier to brief anoxia (Schurr et al. 1986). Kitagawa et al. subsequently published two landmark papers documenting the ability of brief bilateral carotid occlusion to protect gerbil hippocampal CA1 pyramidal neurons (Kitagawa et al. 1990), as well as the cerebral cortex, putamen, and thalamus (Kitagawa et al. 1991) from subsequent prolonged global ischemic exposure, thereby firmly establishing the concept of ischemic tolerance in the brain. A substantial number of studies confirmed the phenomenon of ischemic tolerance in forebrain ischemia models and then in the penumbral region in focal ischemia models (reviewed in (Kirino 2002)). Subsequently, multiple robust and reproducible experimental models of cerebral ischemic tolerance have been described involving a variety of metabolic and/or physical stressors including hypoxia, hyperthermia, hypothermia, global hypoperfusion, oxidative stress, cortical spreading depression, seizures, inflammatory cytokines and lipopolysaccharide (LPS) as the preconditioning stimuli (reviewed in (Gidday 2006), (Kirino 2002), and (Kariko et al. 2004)). Although rodent models have served as the foundation for the field, ischemic tolerance is an evolutionarily conserved form of cerebral plasticity that has been demonstrated in invertebrates (reviewed in (Gidday 2006)) and multiple vertebrate models (reviewed in (Gidday 2006), (Kirino 2002), and (Kariko et al. 2004)) including non-human primates (Bahjat et al. 2017; Bahjat et al. 2011). Preconditioning triggers adaptive responses characterized by at least two distinct time frames: (i) a short-lasting protective phenotype that can be induced within minutes and results in changes in ion channel permeability, protein phosphorylation, and other post-translational modifications (‘rapid preconditioning’) (Perez-Pinzon et al. 1997; Stagliano et al. 1999); and (ii) a more extended ischemic tolerance that evolves over hours to days and requires new gene expression and de novo protein synthesis (‘classical preconditioning’ or ‘delayed tolerance’) (reviewed in (Gidday 2006), (Kirino 2002), and (Kariko et al. 2004)). Rapid preconditioning has been best described in response to anesthetics such as isoflurane (de Klaver et al. 2002) as well as solvents such as dimethyl sulfoxide (DMSO) (Nakamuta et al. 2001) and ethanol (reviewed in (Collins et al. 2009)). In contrast, classical preconditioning, the primary focus of the current review, is best described as a response to injury-related conditions including global hypoperfusion, focal ischemia, seizure, or hypoxia. Protection induced by classical preconditioning can last for days to several weeks and is more potent in the CNS than rapid preconditioning (Gidday 2006; Kariko et al. 2004; Kirino 2002). The term ‘ischemic preconditioning’ (IPC) represents a subset of classical preconditioning in which both the preconditioning stimulus and the subsequent injury-inducing exposure are ischemic in nature. The term ‘cross tolerance’ is used when the preconditioning exposure differs from the subsequent injury (Kariko et al. 2004; Kirino 2002).

Ischemic Preconditioning

As preconditioning became increasingly recognized as a potent form of endogenous neuroprotection, a number of investigators began to look for clinical correlations (Dirnagl et al. 2003). Results from two retrospective analyses of prospectively collected clinical data (Moncayo et al. 2000; Sitzer et al. 2004) suggested that stroke patients who suffered a recent prior transient ischemic attack (TIA) had better neurological outcomes than those who did not, implying that a human correlate to experimental IPC may exist. Measurement of cerebral blood flow in animal models of IPC demonstrated that ischemic tolerance was not accompanied by an improvement of regional tissue perfusion during or after ischemia (Barone et al. 1998; Chen et al. 1996; Matsushima and Hakim 1995). These findings agree well with the results of a clinical neuroimaging based retrospective study that reported that although final infarct volumes were significantly smaller in patients that had suffered a prodromal TIA than in those with first ever brain ischemia, patients in both groups had similar cerebral blood flow and perfusion parameters (Wegener et al. 2004). Based on these results, IPC-mediated protection was hypothesized to be due to changes induced in the neuropil itself.

Most of the early theories relating to the mechanism of IPC-mediated protection focused on effects specifically on and within neurons (Dirnagl et al. 2003; Kirino 2002). Proposed mechanisms included neuronal membrane stabilization, inhibition of neuronal excitability/apoptosis, and induction of neuronal stress responses (Dirnagl et al. 2003; Kirino 2002). Although the early neuron-centric studies did reveal important clues to the mechanisms of ischemic tolerance, including a central role for increased expression of neuronal chaperones such as heat shock proteins (HSPs), they could not readily explain the growing body of evidence implicating innate immune signaling pathways in both IPC- and cross tolerance-mediated protection (Kariko et al. 2004). In particular, multiple studies demonstrated that pretreatment with LPS (endotoxin) induced robust protection against subsequent cerebral ischemia (Bordet et al. 2000; Rosenzweig et al. 2004; Tasaki et al. 1997). Importantly, the tolerance inducing effects of LPS have the same temporal pattern and involve the same pro-inflammatory cytokines characteristic of other preconditioning treatments (reviewed in (Kariko et al. 2004)). LPS preconditioning and IPC can both be abolished when protein synthesis inhibitors are co-administered (Barone et al. 1998; Bordet et al. 2000) or when the function of inflammatory cytokines are blocked (Rosenzweig et al. 2007; Tasaki et al. 1997). LPS activates Toll-like receptors (TLRs) including TLR4 (Hoshino et al. 1999; Poltorak et al. 1998) and TLR2 (Yang et al. 1998). Microglia are known to express a large repertoire of TLRs (Olson and Miller 2004) and are competent to respond potently to a variety of TLR agonists including LPS (reviewed in (Hanisch and Kettenmann 2007)). In 2004 the concept of damage associated molecular patterns (DAMPs), endogenous TLR agonists released in brain following CNS injury, was beginning to take hold and a unifying theory of IPC and cross tolerance involving inhibition and down-regulation of TLRs and cytokine signaling was proposed (Kariko et al. 2004). Microglia were thus posited as critical cellular mediators of both IPC- and cross tolerance-mediated neuroprotection (Kariko et al. 2004).

In 2003 a landmark study (Stenzel-Poore et al. 2003) was published characterizing the genomic (transcriptomic) response in the mouse cortex to three related but distinct transient middle cerebral artery occlusion (tMCAO)-induced conditions: (i) prolonged injurious cerebral ischemia alone, (ii) IPC alone, and (iii) IPC followed 72 hours later by prolonged injurious cerebral ischemia. Considering that all of the conditions involved ischemia/reperfusion of different lengths or sequence, perhaps the most striking finding of this study was how different the transcriptomic profiles from each experimental paradigm were - less than 4% of regulated genes were shared among any two or more conditions (Stenzel-Poore et al. 2003). A key difference among the gene profiles was in the third group (IPC + prolonged ischemia) where the overall transcriptional response to injury was downregulated and most of the genes that were suppressed involved pathways that regulate metabolism, molecular transport, or cell cycle control (Stenzel-Poore et al. 2003). Thus, the response to injury in unsorted cortical tissue in the context of preconditioning was disposed towards dampened cellular activity in contrast with heightened expression of immune defense and repair genes in the first group (prolonged ischemia alone) (Stenzel-Poore et al. 2003). The transcriptional response to preconditioning alone (group two) or IPC followed by prolonged ischemia (group three) also had a profile that suggested suppression of cellular energy use and attenuation of ion-channel activity similar to evolutionarily conserved mechanisms observed in hibernation states (Stenzel-Poore et al. 2004). A major component of preconditioning-mediated protection was thus proposed to be secondary to reprogramming the brain’s genomic responses to transient ischemia/reperfusion (Stenzel-Poore et al. 2007).

In 2008 a methodology focused study was published characterizing the optimal experimental stimulus and response time for IPC in mouse. The authors examined the preconditioning effects of either a single ischemic exposure (15 min tMCAO) or multiple interrupted short bursts of ischemia (three bouts of 5 min tMCAO 45 min apart) and different intervals between IPC and stroke (1–4 days) (Zhang et al. 2008). Using this paradigm, the optimal time to establish preconditioning was determined to be three days (72 hours), although this was also dependent on the preconditioning stimulus itself, with milder inductions (multiple sessions) requiring longer periods of establishment (Zhang et al. 2008). Although the optimal preconditioning stimulus should not, in and of itself, result in any permanent brain injury, data suggests that 15 min tMCAO in some rodent models can induce delayed selective neuronal loss with topographically associated activation of microglia even in the absence of overt infarction (Ejaz et al. 2015a; Ejaz et al. 2015b; Pedrono et al. 2010). It seems unlikely however that selective neuronal loss is required for IPC-mediated protection (Zhang et al. 2008).

Innate Immune Signaling Pathways in Ischemic Preconditioning and Cross Tolerance

Building upon this work were further investigations into molecular pathways implicated in preconditioning with a focus on LPS preconditioning. Microarray analysis of cortical tissue from LPS-induced preconditioned mice revealed a distinctive gene expression profile with high expression of innate immune response genes, of which type I interferon stimulated genes (ISGs) were overwhelmingly represented (Marsh et al. 2009a). Furthermore, preconditioning against prolonged ischemia could be induced with intracerebroventricular injection of interferon-β (IFNβ) when given prior to prolonged MCAO in wild-type mice. LPS preconditioning was attenuated in animals with genetic deficiency in downstream type 1 IFN signaling: interferon regulatory factor 3 knockout (Irf3^−/−) mice (Marsh et al. 2009a). Other contemporaneous studies highlighted the importance of TLR4 signaling in both ischemia-related brain injury and preconditioning-mediated protection. Tlr4^−/− mice exhibited reduced infarct volumes after prolonged MCAO (Cao et al. 2007; Caso et al. 2007), but attenuated IPC-mediated reductions in post-stroke infarct volume (Pradillo et al. 2009). Genetic data on stroke patients correlating specific Tlr4 single nucleotide polymorphisms (SNPs) with clinical outcomes also hinted at a crucial role for TLRs in the human neuroinflammatory response to cerebral ischemia (Weinstein et al. 2014). TLR4 activation leads to several downstream events important for establishment of IPC and for the microglial response to damage. One critical downstream effector induced by TLR4 activation is TNFα, which is required for IPC (Marsh and Stenzel-Poore 2008; Stevens et al. 2008) and reduced in Tlr4^−/− animals after IPC (Pradillo et al. 2009). Transcriptomic analysis of cultured primary microglia exposed sequentially to hypoxia/hypoglycemia followed by normoxia/normoglycemia (H/H-N/N), ‘ischemia/reperfusion’-like conditions, demonstrated that wild-type and Tlr4^−/− microglia showed markedly disparate responses to H/H-N/N (Weinstein et al. 2010). This suggested that direct exposure of microglia to ‘ischemia/reperfusion’-like conditions induced unique mechanisms that specifically differentiate microglia with intact or deficient TLR4 signaling. Other in vitro studies involving microglial-neuronal co-cultures (Kaushal and Schlichter 2008; Lehnardt et al. 2008) or transfer of conditioned medium from hypoxia-exposed microglia to cultured neurons (Lai and Todd 2006) also identified TLR4 activation and/or pro-inflammatory cytokine release as critical mediators of microglial-neuronal cross-talk in the setting of ischemia-like conditions.

Activation of other TLRs also induces cross tolerance against cerebral ischemia. Preconditioning against stroke can be achieved by the activation of TLR9 through exposure to cytosine-guanine (CpG) oligodideoxynucleotide motifs (Marsh et al. 2009b). When mice were administered CpGs 72 hours before prolonged MCAO, the transcriptomic response after preconditioning and stroke shifted to predominantly type I IFN-associated transcripts (Marsh et al. 2009b) comparable to LPS preconditioning. When nonhuman primates (rhesus macaques) were administered CpGs 72 hours prior to a focal ischemic injury of the cortex the cortical damage was reduced and neurological recovery was improved (Bahjat et al. 2017; Bahjat et al. 2011). TLR3 recognizes polyinosinic polycytidylic acid (poly: IC), a synthetic analog of double stranded RNA associated with viral infection, and is also capable of protecting against prolonged ischemia and inducing a type I IFN gene expression profile (Gesuete et al. 2012). Similar effects were achieved using the TLR7 agonist Gardiquimod (GDQ) (Leung et al. 2012). GDQ induced a significant increase in plasma IFNα levels and protection against ischemia was not seen in mice with genetic deficiencies in type I IFN signaling (i.e Irf7^−/− and Ifnar1^−/−) (Leung et al. 2012). TLR4-mediated preconditioning with LPS is also well characterized (Marsh et al. 2009a; Rosenzweig et al. 2004; Rosenzweig et al. 2007) but absent in Irf3^−/− mice (Marsh et al. 2009a). Based on these studies it was apparent that TLRs and type I IFNs are closely entwined and important in multiple forms of preconditioning (Stevens et al. 2011). Activation of TLR signaling from multiple TLRs converges on type I IFN signaling (Figure 1). Thus, many forms of cross tolerance share common downstream molecular signaling mechanisms. Downstream components include the signaling adaptor proteins MyD88 and TRIF; the latter molecule being essential for TRIF-mediated IFN expression after TLR4 activation (reviewed in (Kumar et al. 2011)) (Figure 1). Vartanian et al. found that TRIF, but not MyD88, was required for LPS preconditioning mediated neuroprotection in the mouse tMCAO stroke model (Vartanian et al. 2011) and the TRIF/IRF3 pathway has been implicated in both LPS preconditioning (Marsh et al. 2009a) and IPC (Stevens et al. 2011).

Figure 1.

Multiple innate immune signaling pathways involved in preconditioning converge on type I interferon stimulated gene (ISG) expression. TLR9 and TLR7 can be activated through endogenous ligands such as CpG-rich DNA and single-stranded RNA, respectively. Additionally, the pharmacological agent Gardiquimod (GDQ) can activate TLR7 to induce preconditioning against cerebral ischemia. Both TLR9 and TLR7 upregulate ISGs through MyD88-dependent mechanisms. TLR4 responds to the exogenous ligand LPS, which can induce preconditioning, or to various endogenous DAMPs including high mobility group box 1 (HMGB1), peroxiredoxins (PRXs), and heat shock proteins (HSPs). Binding of TLR4 can result in signaling through adaptor protein MyD88, which leads downstream to NFκB activation and TNFα expression, and through a different adaptor protein TRIF to induce ISG expression via IRF3. The co-receptor CD14 skews TLR4 activation to the TRIF/IRF3 pathway. TLR3 can be activated by double-stranded RNA or poly(I:C), which induces preconditioning against cerebral ischemia, and induces an ISG response via TRIF. Additionally, preconditioning can be induced through type I IFN signaling (IFNα or IFNβ ligands binding to IFNAR1/IFNAR2), which transduces the activation signal via STAT1 and IRF9. Intracellularly, RIG-I and MDA-5 recognize short and long double-stranded RNA DAMPs, respectively, and induce ISGs via MAVS. cGAS recognizes double-stranded DNA resulting in STING-mediated ISG expression. Additionally, NOD-like receptor (NLR) activation leads to ISG expression, though the specific DAMPs inducing NLRs are less well defined. MicroRNAs are capable of acting on multiple components of these signaling pathways to modulate the inflammatory response. MiR-155 suppresses expression of several mRNA targets with gene products that are key components in the TLR/MyD88 pathways while promoting type I IFN signaling via inhibition of proteins such as suppressor of cytokine signaling 1 (SOCS1) that ordinarily restrain this pathway, thus shifting the milieu from TLR-based signaling to enhancement of type I IFN signaling. After the early pro-inflammatory response driven by miR-155, miR-145 and miR-146a are expressed, which target IFNβ and components of type I IFN signaling system to restrain this pathway and resolve inflammation.

Cell surface co-receptors may also play a role in channeling TLR-activation toward the type I IFN pathway. CD14 is the prototypical co-receptor for TLR4 (Jiang et al. 2005). CD14 facilitates TLR4 responses to LPS and it is required for LPS-induced TLR4 internalization and engagement of downstream signaling components (Zanoni et al. 2011) (Figure 1). Janova et al defined a critical role for CD14 in microglia, specifically related to regulating the sensitivity of microglia to pathogens and DAMPs, and in restraining microglial activation in response to a significant immune challenge – thereby preventing hyperactivation (Janova et al. 2015). The study also found that genetic deficiency in CD14 causes impairment in IFNβ production and signaling in microglia and that CD14 was mandatory for microglial and CNS responses to TLR4-agonist DAMPs (Janova et al. 2015). The authors proposed that CD14 signaling control is distinct for microglia and is critical in differentiating responses to varying TLR4 endogenous agonists. Notably, CD14 deficiency also resulted in increased infarct volume in a 60 min tMCAO model of stroke (Janova et al. 2015). Similar to CD14, the CD24/Siglec G pathway has also been proposed as an important signaling pathway allowing immune cells to distinguish and differentiate responses to endogenous (DAMP) and exogenous (PAMP) TLR4 agonists (Chen et al. 2009b). Studies exploring the impact of these pathways on microglial signaling, stroke, and IPC are all ongoing.

Cellular Mediators of Preconditioning

Multiple cells in the CNS and peripheral immune system respond to IPC (reviewed in (McDonough and Weinstein 2016)). Microglia are the likely first responders to ischemia (Umekawa et al. 2015), undergo a metabolic shift as a consequence of TLR4 activation (Nair et al. 2019), and are capable of recruiting peripheral immune cells and activating astrocytes. Astrocytes demonstrate in vitro capabilities of responding to IPC on their own with alterations in glucose transport and metabolic pathways (Pang et al. 2015; Yu et al. 2008) affecting their ability to survive subsequent prolonged ischemia. Neurons also shift their metabolic profile in response to IPC (Venna et al. 2012; Yu et al. 2008). This metabolic adjustment primes cells for surviving prolonged ischemia (Stenzel-Poore et al. 2007; Tanaka et al. 2015). Peripheral immune cells infiltrate the CNS within 72 hours of IPC (McDonough and Weinstein 2016). The contribution of peripheral immune macrophages and monocytes to IPC is unknown, however peripheral macrophages and microglia share numerous innate immune functions and pathways. Recently a study was published demonstrating that monocytes isolated and adoptively transferred from LPS-preconditioned mice were able to confer protection against a prolonged subsequent MCAO (Garcia-Bonilla et al. 2018). However, for the purposes of this review we will focus on the role of microglia, the CNS resident tissue macrophage, in IPC.

Microglia

Our emerging understanding of the molecular cues that control microglial lineage, viability, proliferation, differentiation and gene expression has direct relevance to understanding the IPC-related neuroimmune response. Microglia originate from c-kit⁺ hematopoietic progenitors in the yolk sac around embryonic day (E)7.0 (in the mouse), with an initial early wave of microglia colonizing the developing CNS around E8.0 (Ginhoux et al. 2010). Colony stimulating factor 1 receptor (CSF1R) expression on microglia is required for development of microglia (Ginhoux et al. 2010). Recent studies have demonstrated CSF1R is required for maintenance of adult populations of microglia, with inhibition of this pathway resulting in depletion of microglia in the adult CNS (Elmore et al. 2014). These inhibitors, including PLX3397 and PLX5622, form the basis for investigations into identifying putative adult microglial progenitors (Elmore et al. 2014; Huang et al. 2018), functional consequences of microglial depletion and replenishment (Elmore et al. 2015; Rice et al. 2017), and biological roles of microglia in injury and disease models (Hilla et al. 2017; Rice et al. 2015). Other critical regulators of microglial maturity include transcription factors Pu.1 and Irf8 (Kierdorf et al. 2013), as well as cytokine transforming growth factor-β (TGF-β) (Butovsky et al. 2014). A transcriptomic and proteomic study of microglia identified numerous proteins and genes differentially expressed in acutely isolated adult microglia and revealed a TGF-β-dependent microglial signature (Butovsky et al. 2014). Neither microglia cell lines (N9, BV2), cultured primary microglia, nor embryonic stem cell microglia express this adult ex vivo microglial signature, but the transcriptomic profile of cultured microglia can be adjusted to a state more closely resembling acutely isolated adult microglia by culturing cells in the presence of TGF-β (Butovsky et al. 2014). Furthermore, as the microglial transcriptome is rapidly altered under cell culture conditions (Bohlen et al. 2017), it is important to consider these limitations and differences between in vitro and in vivo microglial responses in designing experiments to test microglial responses to immune stimuli. Despite these experimental considerations, transcriptomic data on acutely isolated microglia has translated well from mouse to humans (Gosselin et al. 2017). In addition, microglia-specific genes demonstrate a high degree of overlap with genes linked to neurodegenerative conditions such as Alzheimer’s disease (AD) (Gosselin et al. 2017). These findings support the clinical relevance of investigation into specific disease-associated microglial gene expression patterns using animal models.

Under homeostatic conditions in the healthy brain, microglia exhibit ramified and mobile processes that surveil surrounding brain tissue (Nimmerjahn et al. 2005) for the purposes of detecting and responding to perturbations. Facilitating this biological function, microglia possess a ‘sensome’ of innate immune receptors (Hickman et al. 2013), including TLRs, that respond to pathogen-associated molecular patterns (PAMPs) from exogenous agonists derived from viral and bacterial structures as well as to DAMPS, endogenous biological markers of damage and cell stress such as heat shock proteins (Benarroch 2013; Lehnardt et al. 2008), cytosolic DNA (Cox et al. 2015; Marsh et al. 2009b) or peroxiredoxins (Shichita et al. 2012). Microglia also respond to complement cascade proteins (Hong et al. 2016; Lian et al. 2016). Under homeostatic conditions, microglia regulate the number of functional synapses (Ji et al. 2013; Paolicelli et al. 2011) through complement, which also serves important synaptic pruning functions in development (Stevens et al. 2007). However, aberrant responses to complement, or excessive accumulations of complement proteins, may be contributing factors to neurological diseases such as AD (Hong et al. 2016) and multiple sclerosis (MS) (Michailidou et al. 2016). Other cues engage the phagocytic functions of microglia, for example TREM-2 (Takahashi et al. 2005) and CD68 (Simpson et al. 2007) promote microglial phagocytosis.

Recent studies into microglial biology have utilized the rapidly developing single-cell RNA sequencing technologies to gain insights into the microglial transcriptome over the lifespan of an organism and microglial responses to CNS injury. During development – both prenatal and early postnatal – microglia are more transcriptionally diverse than in aged animals (Hammond et al. 2018). The transcriptomic profile of aged microglia is more similar to the activated rather than surveying microglia observed in young adult animals (Hammond et al. 2018; Holtman et al. 2015). The microglia of aging mice also downregulate transcripts for endogenous ligands, while those for microbe recognition and host defense are upregulated (Hickman et al. 2013), indicating evolving transcriptomic and sensomic profiles for microglia over the lifespan of an organism. Single-cell RNA sequencing studies demonstrate that microglia downregulate expression of several canonical microglial differentiation markers (P2ry12, Cx3cr1, Tmem119) after injury (Hammond et al. 2018; Holtman et al. 2015), while some upregulate proliferation markers and ISGs such as Cxcl10 and Ccl4 (Hammond et al. 2018). Hammond et al. also identified upregulated ISGs in aging mice, including genes Ifitm3 and Oasl2, and characterized the microglial phenotype as having a more immunogenic profile in the aging mouse brain compared to young adult mice.

Microglia in Ischemic Preconditioning

Few groups research microglia specifically in the context of IPC. Longer monophasic periods of cerebral ischemia are more commonly reported using tMCAO time intervals that result in stroke-like infarction and neuroinflammatory responses. From these we can infer microglial responses and immune pathways that might be targetable for stroke therapies or which might be altered after IPC to confer tolerance against prolonged ischemia. One study in mice revealed proliferation of microglia after moderate ischemia (30 min tMCAO) that was reduced in more severe stroke (60 min tMCAO) (Denes et al. 2007). Published the same year, another study using a transgenic system with thymidine kinase gene expression driven by the CD11b promoter combined with pharmacologic administration of ganciclovir found that ablation of proliferating microglial cells prior to and after a 60 min tMCAO resulted in larger infarct volumes, more apoptotic neurons, and altered expression of proinflammatory cytokines (Lalancette-Hebert et al. 2007). Using PLX3397, a CSF1R antagonist that depletes microglia, a recent study demonstrated increased infarct volumes after prolonged tMCAO (60 min) following depletion of microglia (Szalay et al. 2016). Based on these studies there are a few suggestions of mechanisms by which preconditioned microglia can provide protection against prolonged ischemia. However, as discussed above, studies into the transcriptomic profile of preconditioned and stroked brains demonstrate little overlap in gene expression (Stenzel-Poore et al. 2007; Stenzel-Poore et al. 2004; Stenzel-Poore et al. 2003). There is also evidence that pharmacologic preconditioning in nonhuman primates generates a distinctive proteomic signature similar to transcriptomic data in preconditioned rather than stroked mice (Stevens et al. 2018). These studies urge caution in over-interpreting microglial responses to prolonged ischemia as they may diverge from protective preconditioning responses. However, studies examining the responses to prolonged ischemia are important for developing novel therapeutic targets to modulate the neuroimmune response of stroke patients.

Due to the status of microglia as the resident immune cell of the CNS we have focused specifically on microglial responses to IPC. We developed a combined in vivo/ex vivo experimental model to study IPC in white matter, specifically in the mouse optic nerve (MON). This white matter-centric approach is important clinically as one of the critical neuroanatomical differences between rodent and human brains is that the latter possess a much greater volume percentage of white matter (Hamner et al. 2011; Zhang and Sejnowski 2000). In addition, both ischemic stroke and vascular cognitive impairment in humans heavily impacts white matter (O’Brien and Thomas 2015; Shi and Wardlaw 2016; Wardlaw et al. 2013). Scientifically, the focus on white matter in IPC is also important because ischemia-associated effects in white matter and gray matter are known to advance through distinct molecular mechanisms (Matute et al. 2013; Pantoni et al. 1996; Stys et al. 1992). In our model a unilateral common carotid artery occlusion (CCAO) was used to make one MON transiently ischemic (preconditioned). Reperfusion was then established and the animals were sacrificed 72 hours later. Both MONs were isolated and exposed ex vivo to oxygen glucose deprivation (OGD) with compound action potentials (CAPs) recorded throughout OGD exposure and after restoration to normoxic/normoglycemic conditions. We found that preconditioned MON CAP recovery was markedly improved relative to control (contralateral or sham-operated) MONs (Hamner et al. 2015). Not only was CAP recovery improved in preconditioned MONs, but axonal integrity, as determined by immunohistochemistry for phosphorylated neurofilament, was improved relative to control MONs and oligodendrocyte loss was attenuated (Hamner et al. 2015). Genetic deficiency of two key innate immune signaling receptors, TLR4 or IFNAR1, abolished preconditioning effects, as did the myeloid cell-type specific knockdown of IFNAR1 in microglia (Hamner et al. 2015). Overall, this study was the first demonstration of IPC as an intrinsic capability of white matter and the first in vivo demonstration that cell-type specific expression of an individual gene plays an indispensable role in IPC-mediated protection.

A significant body of work suggests that type I IFN signaling in microglia plays a critical role in multiple forms of preconditioning-mediated protection. It is worth noting that the effects of type I IFNs on outcome following prolonged ischemia (stroke) has been investigated using several different animal model systems and temporal administration paradigms. Veldhuis et al. demonstrated that subcutaneous injections of recombinant rat IFNβ once daily starting as late as six hours following permanent MCAO reduced infarct volume and edema as determined by MRI diffusion and T2 sequences, respectively (Veldhuis et al. 2003a). The protection was associated with reduced infiltration of inflammatory cells from the periphery and attenuation of blood-brain barrier (BBB) disruption, but did not affect cerebral blood flow (Veldhuis et al. 2003a; Veldhuis et al. 2003b). In contrast, Maier et al. found that treatment of rats with varying dosages of daily intravenous rat IFNβ or monopegylated IFNβ following tMCAO failed to confer neuroprotection as determined by histology and neurobehavioral testing one week following stroke (Maier et al. 2006). Marsh et al. reported that intracerebroventricular injection of IFNβ prior to 40 min tMCAO reduced infarct volume (Marsh et al. 2009a). In a mouse tMCAO model, Kuo et al. showed that intravenous administration of recombinant mouse IFNβ three hours before and after MCAO significantly reduced infarct volume in wild-type, but not IFNAR1^−/−, mice (Kuo et al. 2016). IFNβ also suppressed stroke-induced infiltration of peripheral monocytes and neutrophils and reduced the number and size of microglia/macrophages in the brain parenchyma (Kuo et al. 2016). In contrast to preconditioning effects of IFNβ, several acute stroke studies in the mouse demonstrated that neither genetic deficiency of IFNβ (Inacio et al. 2015; Marsh et al. 2009a) nor IFNAR1 (Inacio et al. 2015) had an impact on stroke outcome. One recent study reported a significant reduction in infarct volume in IFNAR1^−/−, but not IFNAR2^−/−, mice (Zhang et al. 2017), which suggests a role specifically for IFNβ, but not IFNαs, in responding to ischemia. IFNβ is capable of initiating signaling by binding the IFNAR1 subunit independently of IFNAR2, which results in Ifnar1^−/−, but not Ifnar2^−/−, mice being protected against LPS-induced lethality (de Weerd et al. 2013). Based on these studies it appears that type I IFNs can have both beneficial and deleterious effects in the setting of cerebral ischemia/reperfusion depending on the context, timing, and duration of signaling activation as well as which type I IFNs and IFNAR subunits are present. Given the diversity of models and genetic paradigms, there are likely multiple mechanistic explanations for the differing degrees of impact of type 1 IFN signaling in stroke. Nevertheless, one recurring theme is that early temporal activation of the type 1 IFN signaling system yields resistance to subsequent prolonged ischemic exposure. This temporal pattern fits well with the concept of microglial priming (Garcia-Bonilla et al. 2014) and allows for the possibility that downstream microglial pathways, including ones involving ISG products, are critical effectors in mediating protection and/or enhanced recovery.

Given the importance of microglial type 1 IFN signaling in IPC, we decided to look specifically at transcriptomic responses of microglia both in vitro and in vivo to ischemia-reperfusion conditions (McDonough et al. 2017). We re-analyzed our transcriptomic data on cultured primary microglia exposed sequentially to hypoxia/hypoglycemia followed by normoxia/normoglycemia (H/H-N/N) at higher stringency levels and found that H/H-N/N induced a TLR4-dependent ISG response to ‘ischemia/reperfusion’-like conditions in vitro. Strikingly, 50% of all differentially regulated genes were ISGs and bioinformatic analyses identified interferon regulatory factors (ISGF3, IRF-2, IRF-1, and IRF-7) as the most active transcription factors under these experimental conditions. The ISG response was furthermore dependent on IFNAR1 and characterized by phosphorylation of signal transducer and activator of transcription 1 (STAT1) and down-regulation of IFNAR1 surface expression (McDonough et al. 2017). Using the classic grey matter predominant model of preconditioning after a 15 min tMCAO in mice, we also found robust expression of ISGs in acutely isolated preconditioned cortical microglia (McDonough et al. 2017). Similar to our in vitro findings, this microglial ISG response was dependent on IFNAR1, but interestingly was independent of TLR4. Nevertheless, the convergence of both the in vitro and in vivo transcriptomic responses to ischemia/reperfusion on IFNAR1-dependent ISG expression in microglia suggested that type 1 IFN signaling in microglia is critical in this physiologic context. The findings are of particular interest in light of other recent data describing microglial ISG responses in aging (Grabert et al. 2016; Holtman et al. 2015) and neurodegenerative disease (Holtman et al. 2015). They were also of interest given our recent data on microglial type 1 IFN signaling being necessary for IPC-mediated axonal protection in white matter (Hamner et al. 2015). Finally, these data provided cell-type specific context to earlier studies implicating type 1 IFN signaling in unsorted preconditioned cortical tissue (Marsh et al. 2009a; Stenzel-Poore et al. 2003).

We reported previously that a brief IPC pulse (15 min tMCAO) induced a robust increase in the number of cortical microglia and macrophage subsets as determined by ex vivo flow cytometry (McDonough and Weinstein 2016). Follow up work in our laboratory has characterized the IPC-induced microglial proliferative response in cortex at both the microglial transcriptomal and in situ histological levels. Our findings indicate that the microglial transcriptomic response to IPC is characterized by massive up-regulation of cell cycle and cellular proliferation genes even in the absence of cortical infarction (A.M. and J.W., unpublished observations). The molecular signaling pathways spurring this IPC-induced microglial proliferative response are under investigation. The IPC-induced microglial transcriptomic response is interesting in light of recent single-cell RNA sequencing studies focused on microglia that have identified both cell cycle/proliferation and IFN signaling as critical microglia gene expression clusters conserved among different injury and disease paradigms (Hammond et al. 2018). We also observe downregulation of multiple microglial differentiation genes in the setting of IPC (A.M. and J.W., unpublished observations). Suppression of canonical microglial differentiation markers has been reported in several neurodegenerative conditions (reviewed in (Butovsky and Weiner 2018)).

There are commonalities in microglial responses to different injuries, for example engagement of type I IFN signaling, downregulation of mature microglial markers (dedifferentiation), and an increase in cell proliferation, that may be key for mounting a protective response. However, the identity of the important protective pathways and critical regulators requires more investigation. Based on work from our lab and others, we hypothesize that microglia are primary responders to signals of cellular damage (i.e. DAMPs) released from other cells in the CNS including astrocytes, neurons and oligodendrocytes (Figure 2). These intracellular signals clearly engage microglia through TLRs (Figure 1), but there are also other molecular mediators that facilitate a protective and/or immunomodulatory rather than a cytotoxic inflammatory response. Other signaling system candidates for mediating the microglial response to DAMPs include: (i) nucleotide-binding domain leucine-rich repeat NOD-like receptors (NLRs) (Chen et al. 2009a; Mason et al. 2012), (ii) the retinoic acid gene I-like helicase/mitochondrial antiviral signaling protein (RIG-I/MAVS) complex (Seth et al. 2005), and (iii) the cytosolic DNA sensing cyclic GMP-AMP synthase/adaptor protein stimulator of type 1 IFN genes (cGAS/STING) pathway (Chen et al. 2016). All three of these systems function as intracellular pattern-recognition receptors active in microglia (Cheng et al. 2017; Reinert et al. 2016; Shiau et al. 2013) that potently regulate innate immunity and lead to induction of type 1 IFN signaling (Buskiewicz et al. 2016; Coutermarsh-Ott et al. 2016; Reinert et al. 2016) (Figure 1). Recently the RIG-I pathway was implicated along with type 1 IFN signaling in cerebral preconditioning-mediated protection against stroke-induced brain injury (Gesuete et al. 2016).

Figure 2.

In response to an ischemic preconditioning stimulus, multiple cells in the CNS release DAMPs and cytokines that signal to the microglia, predominantly through TLRs and IFNAR1. In response to these signals, microglia upregulate ISGs and undergo proliferation. Consequentially the altered microglial phenotype is protective and it exerts numerous modulatory and supportive capabilities on astrocytes, neurons, and oligodendrocytes. Microglia modulate astrocyte activation, which in turn results in decreased astrogliosis and maintenance of the blood-brain barrier (BBB). Furthermore, astrocytes and microglia both provide trophic support by regulating glutamate and Ca²⁺ levels, which leads to increased survival of neurons and oligodendrocytes and increased axonal integrity after prolonged cerebral ischemia. Microglial ISG chemokines can also regulate and modulate the peripheral immune response. IPC exposed but long-lived microglia may contribute to innate immune memory (IMM).

Microglia as effectors of IPC mediated protection

Studies into downstream effects of preconditioning stimuli suggest a myriad of microglia-mediated outcomes and cell-cell interactions are involved in modulating the CNS response to subsequent prolonged ischemia. Polyinosinic polycytidylic acid (poly-ICLC; a stabilized variant of poly-IC)-mediated preconditioning triggers a type I IFN response in microglia and IFNβ production by microglia and astrocytes, which attenuates BBB dysfunction after ischemic injury by modulating tight junctions in endothelial cells through type I IFN signaling pathways (Gesuete et al. 2012). LPS preconditioning reduces neutrophil infiltration into the brain after prolonged tMCAO in mice (Rosenzweig et al. 2004), potentially through similar type I IFN-mediated pathways and effects on endothelial cells. Type I IFNs also act directly on microglia to modulate phagocytosis, including the clearance of degenerating axons, degraded myelin, and apoptotic cells (Hosmane et al. 2012; Rajbhandari et al. 2014), functions necessary for restoration of tissue homeostasis. In addition to both producing and responding to type I IFNs, microglia are capable of secreting and/or responding to interleukins (IL-1b, IL-6, IL-10, IL-23), TNFα, TGFβ, and both brain- and glial-derived neurotrophic factors (BDNF and GDNF) (Benarroch 2013). Thus, depending on the regional milieu, microglia may promote: (i) a localized pro- or anti-inflammatory state, or (ii) toxic or protective effects on surrounding CNS cells, axons and synapses. Through these signaling molecules, microglia are capable of responding to and modulating astrocytes and peripheral immune cells (PICs) – both by recruiting PICs into the CNS and modulating present peripheral infiltrates to resolve an inflammatory episode (Figure 2) (reviewed in (McDonough and Weinstein 2016)). In addition to recognizing molecular markers of cellular injury and stress, microglia have receptors that receive constitutive quiescent signals from neurons, among such signals are CD200 (Hoek et al. 2000) and CX3CL1 (fractalkine) (Paolicelli et al. 2011). Microglia express receptors CD200-R and CX3CR1 for these ligands, which are engaged by neurons in a healthy state to promote the surveying microglial phenotype. Loss of either CD200 or fractalkine is an activation cue for microglia (Hoek et al. 2000; Paolicelli et al. 2011). Fractalkine/CX3CR1 signaling has been studied extensively in stroke models. Mice genetically deficient for fractalkine are less susceptible to cerebral ischemia/reperfusion injury (Soriano et al. 2002) and Cx3cr1^−/− mice show reduced infarct volumes and improved neurobehavioral outcomes following MCAO (Denes et al. 2008; Tang et al. 2014). Furthermore, in wild-type mice, siRNA silencing of cx3cr1 attenuated white matter injury following bilateral carotid artery occlusion (Liu et al. 2015). Notably, both the genetic- (Denes et al. 2008; Tang et al. 2014) and siRNA- (Liu et al. 2015) mediated reductions in CX3CR1 also attenuated ischemia-induced increases in microglia activation markers, proliferation and cytokine release. Understanding which molecular signaling pathways are engaged in shaping the immunomodulatory/protective microglial phenotype in IPC is critical for future investigations into mechanisms.

miRNA regulators of innate immune signaling in microglia

Micro RNAs (miRNA) are a class of small non-coding RNAs that regulate gene expression post-transcriptionally. A growing body of research has developed to focus on mechanisms of miRNA regulation and influence on cellular function and responses. General biological functions of miRNAs include epigenetic regulation of gene expression, post-transcriptional gene silencing/regulation, and promotion of cellular differentiation (Su et al. 2016). Within the context of neuroinflammation, miR-21, miR-124, miR-146a, miR-155, and miR-689 have been implicated as key regulatory miRNAs (reviewed in (Slota and Booth 2019; Su et al. 2016)). Of particular interest to mechanisms of preconditioning, miR-155 is upregulated in response to LPS (via NF-κB), multiple TLR agonists, IFNβ (Wang et al. 2010), and IPC (Su et al. 2014). The IPC-induced miR-155 response is driven through activation of p53 with a functional consequence being suppression of c-Maf, a transcription factor known to promote differentiation and anti-inflammatory responses in myeloid cells, and subsequent upregulation of pro-inflammatory cytokines (Su et al. 2016; Su et al. 2014). Within the context of ischemia, miR-155 has been observed to be downregulated after stroke (Liu et al. 2010) and studies focused on inhibiting miR-155 have demonstrated improved post-stroke outcomes following distal MCAO (Caballero-Garrido et al. 2015; Pena-Philippides et al. 2016). These outcomes included improved blood flow and BBB integrity, reduced endothelial damage, reduced cerebral infarction and neuronal death, and improved sensorimotor behaviors (Caballero-Garrido et al. 2015). Additionally, miR-155 inhibition resulted in smaller, less phagocytic microglia/macrophages after distal MCAO and reduced expression of CD68 by microglia (Pena-Philippides et al. 2016).

The miRNA response to preconditioning, both IPC and TLR-mediated cross tolerance, features both early-acting and late-induced miRNA components that modulate the microglial activation state dynamically. MiR-155 is expressed hyperacutely after an immune stimulus has been detected and inhibits components downstream of TLR/MyD88 pathways (Forster et al. 2015) while promoting type I IFN signaling via inhibition of proteins that ordinarily restrain this pathway, most notably by inhibiting suppressor of cytokine signaling 1 (SOCS1) (Wang et al. 2010), thus shifting the milieu from TLR-based signaling to enhancement of type I IFN signaling. MiR-155 is strongly induced as early as three days after an IPC stimulus (Su et al. 2014) and maintained for at least seven days (Su et al. 2016). The seven day post-IPC miRNA profile showed induction of miR-34a and miR-145 through the same p53-dependent pathway (Su et al. 2016; Su et al. 2014). Of these late-expressed molecules, miR-145 directly targets IFNβ in the CNS (Figure 1) (Witwer et al. 2010). Similarly, late-expressed miR-146a downregulates pro-inflammatory cytokines and gene expression in microglia (reviewed in (Slota and Booth 2019; Su et al. 2016)), is capable of targeting ISGs including STAT1 and IRF5 in macrophages (He et al. 2016; Tang et al. 2009), and may induce similar responses in microglia after IPC (Figure 1). MiR-146a also inhibits multiple downstream components of TLR-signaling such as interleukin receptor-associated kinases IRAK-1 and IRAK-2 (reviewed in (Forster et al. 2015; Su et al. 2016)), thus counteracting pathways implicated in acute neuroinflammation and also temporally limiting multiple preconditioning pathways including TLRs and type I IFNs. Intriguingly, miR-155 and miR-146a are also induced by these same pathways suggesting they are part of a positive feedback process (miR-155 promotes type I IFN signaling) and in other contexts a negative feedback process (miR-146a inhibits TLR and type I IFN signaling, miR-155 attenuates TLR signaling) (Forster et al. 2015; Slota and Booth 2019; Wang et al. 2015).

The time course of miR-155/miR-146a expression dynamics after an immune challenge suggests that early microglial responses skew pro-inflammatory (miR-155) and may be later resolved by miR-146a (Su et al. 2016). This appears to also have direct consequences on type I IFNs, engaging an early IFN response via miR-155 and restraining that same signaling pathway at later time points through miR-145 and miR-146a. Based on these functions, one possible mechanism of IPC is that the mild, non-injurious IPC stimulus induces a mild inflammatory episode that is subsequently resolved within days by late-expressed microRNAs including miR-145 and miR-146a at or prior to the time of the second (prolonged) ischemic insult, promoting tolerance against this second insult. This occurs in contrast to a singular prolonged ischemic event (stroke) that engages inflammation lasting several days and which is driven by cellular death and the resultant prolonged exposure to DAMPs from injured and dying cells. Thus, the kinetics of miRNA regulation in microglia may explain the temporal dynamics and potency of delayed preconditioning in the brain. Furthermore, there are some studies demonstrating that type I IFN signaling regulates components of the miRNA machinery including Argonaute family members and Dicer expression levels (reviewed in (Wiesen and Tomasi 2009)), suggesting long-term impacts on the microglial response to subsequent immune challenges.

Additional miRNA regulators of innate immune signaling include miR-21, which is induced by TLR signaling via MyD88 and functions as a negative regulator of these pathways, and the Let-7 miRNA family which directly targets TLR4 (reviewed in (Slota and Booth 2019)). Dysregulation of these immune modulating miRNAs has been reported in several neuroinflammatory and neurodegenerative disorders including MS, AD, Parkinson’s disease, and encephalitis (reviewed in (Slota and Booth 2019; Su et al. 2016)). The role of these, and other miRNAs, in modulating the microglial response to ischemia warrants further study.

Chronic type I interferon signaling and microgliopathies

As discussed above, type I IFN signaling specifically in microglia is required for IPC-induced axonal protection (Hamner et al. 2015). Downstream ISG expression is a core transcriptional feature of microglia after IPC (McDonough et al. 2017) and in the context of aging and neurodegeneration (Grabert et al. 2016; Hammond et al. 2018; Holtman et al. 2015). While acute exposure to type I IFNs is protective in both IPC and cross tolerance, chronic and/or unrestrained type I IFN signaling is associated with microglial dysfunction – or microgliopathies - in the developing and aging nervous system. Congenital infections during pregnancy can result in TORCH syndrome and effects on the CNS such as microcephaly, leukoencephalopathy, cerebral atrophy, and calcifications. Genetic dysregulation of type I IFN signaling results in a clinical phenotype identical to TORCH syndrome in the absence of infectious cause, classified as pseudo-TORCH syndrome (PTS). In humans, genetic loss-of-function mutations in ubiquitin-specific peptidase 18 (USP18), a key negative regulator of type I IFN signaling, results in unrestrained type I IFN signaling in microglia and is associated with multiple and severe developmental neurological abnormalities (intracerebral hemorrhage, ventriculomegaly) along with microgliopathy (Meuwissen et al. 2016). Genetically targeted deletions of Usp18 specifically in microglia in mice result in a destructive interferonopathy with distinctive white matter microgliopathy (Goldmann et al. 2015), while mice with systemic knockout of Usp18 exhibit severe hydrocephalus, enlarged ventricles, and high mortality (Knobeloch et al. 2005; Ritchie et al. 2002). These data suggest that exposure to chronically amplified type I IFN signaling, of either infectious or genetic etiology, has significant developmental consequences and points to microglia as key mediators of CNS pathology in TORCH and PTS (Blank and Prinz 2017). USP18, and its downstream protein substrate ISG15 (Malakhov et al. 2002; Ritchie et al. 2002), the genes for which are both upregulated in cortical microglia by IPC (McDonough et al. 2017), represent potential targets in the type I IFN signaling pathway that could be modulated by conventional pharmacologic approaches (Owens et al. 2014) or through novel microglia directed therapies (Nance et al. 2017; Nance et al. 2015; Wieghofer et al. 2015).

In the adult mouse, type I IFN signaling appears to be protective in experimental autoimmune encephalomyelitis (EAE), a model of MS and CNS autoimmunity. When IFNAR was deleted specifically in myeloid cells using a LysM^Cre transgenic line crossed with Ifnar^fl/fl mice, the bigenic mice had increased mortality and a more severe disease phenotype compared with Ifnar^fl/fl controls (Prinz et al. 2008). While IFNβ treatment can ameliorate many symptoms of MS in human patients, some develop undesirable side effects after chronic treatment (Neilley et al. 1996; Waubant et al. 2003). There are hints as to the detrimental effects of long-term type I IFN exposure in other neuroimmunological studies. Elevated type I IFNs are a clinical hallmark of systemic lupus erythematosus (SLE) patients, and increased type I IFN signaling has been noted in the brains of these patients (Bialas et al. 2017). In transgenic lupus-prone mice with IFNAR-dependent autoimmunity, microglia were observed to have increased uptake of neuronal material and decreased synaptic density compared to controls, which was attenuated after anti-IFNAR antibody treatment (Bialas et al. 2017). During natural aging in mice, type I IFN signaling components and downstream ISGs are upregulated in microglia (Deczkowska et al. 2017; Grabert et al. 2016; Hammond et al. 2018; Holtman et al. 2015). In one study 63% of genes upregulated in aged microglia were related to type I IFN signaling in a bioinformatics analysis (Deczkowska et al. 2017). To examine the effects of chronic type I IFN exposure, Deczkowska et al. created a transgenic mouse that overexpressed IFNβ in the choroid plexus and examined behavioral phenotypes and presynaptic puncta in the hippocampus. These mice had cognitive impairments and reduced numbers of VgluT2⁺ and synaptophysin⁺ puncta in the hippocampus, which were reversed when the IFNβ-overexpressing mice were crossed with mice with microglia deficient in type I IFN signaling (Deczkowska et al. 2017). These data suggest that the negative effects of chronic type I IFN exposure are due entirely to effects on microglia, and they also agree with the previous work that suggests chronic exposure to type I IFNs can result in aberrant synaptic engulfment. Increased complement factor 4b (C4b) expression in the aged and IFNβ−overstimulated microglia (Deczkowska et al. 2017) may also be linked to increases in synaptic pruning observed in aging-related pathologies like AD and glaucoma (Hong et al. 2016; Stevens et al. 2007). Additionally, Minter et al demonstrated that in a mouse model of AD, type I IFNs contributed to cognitive decline and altered the microglial response to neuroinflammation (Minter et al. 2016).

Though many of these studies demonstrate negative consequences of chronic type I IFN exposure (or over-exposure), they have also identified potential pharmacological targets such as USP18 (Goldmann et al. 2015), IFNAR1 (Bialas et al. 2017), and MEF2C (Deczkowska et al. 2017) that can be manipulated to restrain or antagonize type I IFN signaling. Interestingly, in our gene array data comparing microglia isolated from IPC-exposed wild-type mice, we observed not only a significant ISG response (McDonough et al. 2017) but also downregulation of Mef2c (Table 1). Deczkowska et al. reported that when Mef2C was downregulated inflammatory genes including Ccl2, Ccl5, Il1b, and Tnf were upregulated along with c4b (Deczkowska et al. 2017). These are effects we also see in acutely isolated preconditioned cortical microglia (Table 1). Differentiating between protective and toxic effects of type I IFN signaling in microglia therefore merits additional study. The differences may be related in part to timing and duration of exposure (acute vs. chronic) with the latter inducing additional detrimental pathways, or other co-factors necessary for determining the physiologic outcome in microglia. Context of the overall neurologic state (healthy vs. diseased) and status of surrounding neurologic tissue, particularly levels of released DAMPs and inflammatory effectors are likely also critical factors. It is important to note that microglia are long-lived (Fuger et al. 2017) CNS resident immune cells that have a unique ability to assimilate multiple distinct immune priming responses over time (Neher and Cunningham 2019) and have been recently hypothesized to play a key role in the development of innate immune memory (IIM) (Wendeln et al. 2018). Diverse innate immune activating (or ‘priming’) stimulations followed by secondary (sometimes repetitive) stimulations can skew the microglial phenotype away from an exaggerated inflammatory (‘trained’) response toward an immunomodulatory or suppressed (‘tolerized’) response. As such, it may be useful to consider IPC as a functionally and clinically important experimental model for IIM with the microglial response to IPC being the critical cellular/molecular manifestation of this phenomenon (Figure 2).

Table 1.

Expression of Mef2c and Mef2c-regulated genes in ischemic preconditioned cortical microglia.

gene name	description	log2 fold change	linear fold change	p-value
Mef2c	myocyte enhancer factor 2C	−0.27	0.83	0.04093777
Ccl2	chemokine (C-C motif) ligand 2	1.53	2.89	0.00012651
Ccl5	chemokine (C-C motif) ligand 5	2.60	6.06	0.0002623
Il1b	interleukin 1 beta	1.59	3.01	0.00210394
Tnf	tumor necrosis factor	1.39	2.62	0.00048202

Mef2c is downregulated, and its target genes are upregulated, in acutely isolated cortical microglia 72 hours after an ischemic preconditioning stimulus. Mef2c restrains type I IFN signaling and downregulates target inflammatory genes Ccl2, Ccl5, Il1b, and Tnf; the dysregulation of this pathway contributes to chronic type I IFN mediated cognitive decline in mice (Deczkowska et al. 2017). Methods: transient middle cerebral artery occlusion (tMCAO) or sham surgeries were performed as described (McDonough et al. 2017) for 15 minutes to induce preconditioning. Cerebral blood flow was monitored via laser doppler flowmetry to confirm occlusion (McDonough et al. 2017; McDonough and Weinstein 2016). Mice were allowed to recover for 72 hours before they were anesthetized, perfused, and cortices were removed for digestion and ex vivo flow cytometry following methods published in (McDonough et al. 2017; McDonough and Weinstein 2016; Su et al. 2014). Cells were lysed and RNA was collected for cDNA synthesis and cDNAs were hybridized to a Mouse Gene 1.0 ST Array (Affymetrix). Our data was processed and analyzed as described in (McDonough et al. 2017) with assistance and support from the University of Washington (UW) Functional Genomics and Proteomics Core. N = 12 – 14 mice per group (sham vs. tMCAO IPC) with samples collected by ex vivo flow cytometry sorting in four separate experiments per group. The transcriptomic data is available at the NCBI Gene Expression Omnibus at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE107983.

Translational Studies and Clinical Importance of IPC

IPC is observed under experimental conditions in mice, rats, and nonhuman primates, and immune pathways and genes appear to be conserved in the microglia of rodents and humans (Gosselin et al. 2017). This evolutionary conservation along with the potency of IPC and the high degree of applicability to human patients point to IPC as an experimental phenomenon with strong translational potential. Investigations into mechanisms of IPC have already identified several potential therapeutic targets in stroke. As noted above, retrospective studies suggest that IPC-like protection may occur following TIA in human patients (Moncayo et al. 2000; Wegener et al. 2004). However, the application of direct transient ischemic pulses to the brain in patient populations at high immediate stroke risk (i.e. cardiac or carotid surgery, sub-arachnoid hemorrhage, or TIA patients) is impractical and potentially dangerous. The length of time between a TIA and a stroke can vary greatly between patients and experimental IPC has a narrow temporal window of protection that dwindles, so efficacy in human patients may decline as a function of time. Furthermore, TIAs may cause damage and undesirable outcomes (Gad et al. 2019; Yu and Coutts 2018), and TIAs correlate with higher incidence of stroke (Coutts 2017; Graham et al. 2019). Differentiating between an IPC-like ‘beneficial’ TIA and injury-inducing TIA may be difficult.

Remote ischemic preconditioning (RIPC) is a related and less invasive prophylactic treatment option that relies on remote limb ischemia to induce neuroprotection. The seminal papers characterizing RIPC suggest that neuroprotection is rapid and equally potent as IPC (Dave et al. 2006; Meller and Simon 2015; Ren et al. 2008). However, there are still many unknown variables and mechanistic questions in RIPC, a relatively new field of study, in particular how global preconditioning can be achieved remains to be determined. One possible mechanism may be via extracellular vesicles (EVs) containing miRNAs, which were found in the serum of human patients that received RIPC (Frey et al. 2019). Among the upregulated miRNAs in these EVs were miR-29, miR-146a, and miR-19 (Frey et al. 2019), transcripts that target TLRs and TLR-signaling and generally promote an anti-inflammatory state (Forster et al. 2015; Slota and Booth 2019). Another microRNA upregulated in the EVs, miR-24 (Frey et al. 2019), targets stimulator of IFN genes (STING) for degradation (Huang et al. 2012). Although these findings are intriguing, the focus of the Frey et al study was on cardiac protection and it remains to be seen if these EVs reach the brain and how CNS cells respond to these epigenetic signals.

Despite the present lack of mechanistic understanding about RIPC, clinical trials into the effectiveness of RIPC have nonetheless been performed. Two large clinical trials published their findings comparing patients treated with RIPC to those with a sham treatment prior to cardiac surgery. Both trials were randomized, controlled, double-blinded, and appropriately powered with patient cohorts from multiple treatment centers (Hausenloy et al. 2016; Meybohm et al. 2016). However, the chosen end-point parameters evaluated in the studies included patient mortality, myocardial infarction, stroke occurrence, and acute renal failure. Considering the biology of IPC, the negative findings of these studies are unsurprising, as none of the measured outcomes are anticipated biological benefits of preconditioning. There is no evidence that preconditioning through any mechanism reduces stroke occurrence or impacts the rate of additional vascular events. In contrast, one might anticipate improved long-term neurological outcomes in the preconditioned patients. Although designed with rigorous controls and selection, the chosen end-point parameters of these studies do not reflect the anticipated biological effects of RIPC, thus leaving the clinical efficacy of RIPC undetermined. A much smaller (26 patients) single-center, randomized, outcome-blinded and placebo-controlled Phase II trial in the UK examined the efficacy of administering remote ischemic post-conditioning to patients who had previously had a stroke within the past 24 hours using the modified Rankin Scale and NIHSS scores as outcomes (England et al. 2017). The authors noted a slight trend towards improvement in neurological outcomes after post-conditioning, but conceded that their study was not powered to detect clinical changes, and instead designed to evaluate whether or not post-conditioning was feasible and could be tolerated by stroke patients (England et al. 2017). Nonetheless, the results are intriguing and additional trials are underway. The authors identified an increase in serum levels of HSP27, which has neuroprotective effects in experimental stroke (England et al. 2017) and which may be a possible molecular signaling mechanism shared by remote ischemic preconditioning. Modulating the post-stroke inflammatory response via the same preconditioning stimulus (blood pressure cuff cycles to induce remote limb ischemia) suggests there may be certain IPC-activated pathways that can be targeted in stroke patients within an expanded treatment delivery window.

More accessible and promising methods of preconditioning include treatment with pharmacological agents that target, perhaps in combination, key neuroimmune pathways such as TLRs and type I IFN signaling. Encouragingly, non-human primate preclinical studies suggest efficacy for a pharmacological agent targeting TLR9 in stroke (Bahjat et al. 2017; Stevens et al. 2018). The repurposing of IFNβ (clinicaltrial.gov study NCT00097318), already approved by the FDA as a treatment for MS, is another promising biopharmaceutical approach for stroke inspired by mechanistic studies on preconditioning. Non-TLR pattern recognition receptors (NLRs, RIG-I-MAVS, cGAS/STING), TLR-modulating co-receptors (CD14, CD24/Siglec), and intracellular signaling molecules downstream of TLRs (TRIF, MyD88) or type I IFNs (STAT1, IRF3, IRF9) represent a rich source of additional preconditioning-inspired pharmacologic targets. As noted above, recent work has validated the concept of an adoptive transfer cellular based preconditioning approach for stroke (Garcia-Bonilla et al. 2018). Other cellular based therapies, including transfer of cultured microglia (Imai et al. 2007; Kitamura et al. 2004), have shown promise in rodent stroke models. Targeted modulation specifically of microglial proliferation (Lalancette-Hebert et al. 2007), senescence (Bussian et al. 2018) and/or viability/replacement (Elmore et al. 2018; Jin et al. 2017; Szalay et al. 2016) may also enhance or attenuate (depending on the context) preconditioning-mediated protection. Novel nanoparticle technologies (Nance et al. 2017) and new molecular methodologies (Wieghofer et al. 2015) to specifically target microglia have increased our ability to modulate specific signaling pathways in a targeted cell-type specific manner. Given the extended temporal window for modulation of the neuroimmune response in stroke (Dirnagl 2012), all of these innate immune cell- and signaling pathway-focused approaches may well be applicable clinically beyond the immediate high stroke risk populations typically targeted in preconditioning studies. In addition, the recent findings on preconditioning in white matter (Hamner et al. 2015) suggest that manipulation of microglial function and innate immune signaling represents a viable clinical target for patients with vascular cognitive impairment/dementia, a disorder characterized neuropathologically by diffuse leukoencephalopathy.

Summary

Microglia play a critical role in IPC-mediated neuronal and axonal protection against injurious ischemic events. Implicated in the establishment of IPC are innate immune pathways that include TLRs and type I IFN signaling. Multiple lines of evidence converge on microglial ISG expression as a hallmark feature of preconditioning- (both IPC and cross tolerance) induced protection. Curiously, some of the pathways upregulated in preconditioned microglia appear to have commonalities with other injury or aging models, suggesting that there may be a common component in the microglial response to disparate neuropathologies. However, there also exist key differences in the microglia response to both distinct and similar pathologies that modify the cellular phenotype from cytotoxic to protective. The temporal sequence and duration of exposure are critical factors here – brief exposure prior to a serious injury induces preconditioning-mediated protection, whereas extended exposure in the setting of chronic injury exacerbates damage. Discerning how the microglial molecular damage sensors, intracellular signaling pathways, and downstream effector cascades differ in these scenarios is necessary for developing effective modulatory therapies. Recent advances in single cell transcriptomics offer a technological tool for obtaining disease specific signatures of microglia and may identify microglial subsets and unique responses to IPC - thus driving innovation. Furthermore, nonhuman primate preclinical studies suggest that pathways identified in mice may translate well to humans, and preventative therapies for at-risk patients may be on the horizon. Preconditioning-inspired microglia- and/or innate immune signaling-targeted therapeutic approaches may also translate well to the broader acute stroke and vascular cognitive impairment patient populations.

Main points:

• Ischemic preconditioning confers protection against neuronal and axonal loss induced by cerebral ischemia.

• Microglia and innate immune signaling pathways, especially TLRs and type I IFNs, are critical for establishing ischemic preconditioning.