Microglial depletion abolishes ischemic preconditioning in white matter

Abstract

Ischemic preconditioning (IPC) is a phenomenon whereby a brief, non-injurious ischemic exposure enhances tolerance to a subsequent ischemic challenge. The mechanism of IPC has mainly been studied in rodent stroke models where gray matter (GM) constitutes about 85% of the cerebrum. In humans, white matter (WM) is 50% of cerebral volume and is a critical component of stroke damage. We developed a novel CNS WM IPC model using the mouse optic nerve (MON) and identified the involved immune signaling pathways. Here we tested the hypothesis that microglia are necessary for WM IPC. Microglia were depleted by treatment with the colony stimulating factor receptor-1 (CSFR1) inhibitor PLX5622. MONs were exposed to transient ischemia in vivo, acutely isolated 72 hours later, and subjected to oxygen-glucose deprivation (OGD) to simulate a severe ischemic injury (i.e. stroke). Functional and structural axonal recovery was assessed by recording compound action potentials (CAPs) and by microscopy using quantitative stereology. Microglial depletion eliminated IPC-mediated protection. In control mice, CAP recovery was improved in preconditioned MONs compared with non-preconditioned MONs, however, in PLX5622-treated mice, we observed no difference in CAP recovery between preconditioned and non-preconditioned MONs. Microglial depletion also abolished IPC protective effects on axonal integrity and survival of mature (APC⁺) oligodendrocytes after OGD. IPC-mediated protection was independent of retinal injury suggesting it results from mechanistic processes intrinsic to ischemia-exposed WM. We conclude that preconditioned microglia are critical for IPC in WM. The ‘preconditioned microglia’ phenotype might protect against other CNS pathologies and is a neurotherapeutic horizon worth exploring.

Graphical Abstract

Introduction

Ischemic injury to white matter (WM) is a critical component of acute stroke pathology and clinical outcomes. However, traditional models of ischemic injury have neglected the unique pathophysiology of WM in stroke injury (Baltan et al., 2008). In the human brain, WM makes up approximately half of the total brain volume but the WM volume in rodents is only ~13% (Dirnagl, 2012; Zhang & Sejnowski, 2000). The failure of rodent models to fully recapitulate the pathophysiology of ischemic injury in human stroke is in part due to the limited representation of WM, with its unique cytoarchitecture and cellular makeup, in the rodent brain (Hamner et al., 2015; Ransom, Acharya, & Goldberg, 2004; Sozmen, Hinman, & Carmichael, 2012). Moreover, the pathophysiology of ischemic injury differs markedly in WM compared to gray matter (GM); glutamate excitotoxicity mediated by N-methyl-D-aspartate (NMDA) glutamate receptors is a predominate mechanism of ischemic injury in GM, while this is not the case in WM (Ransom et al., 2004; Tekkok, Brown, Westenbroek, Pellerin, & Ransom, 2005). Experimental pharmacotherapies for stroke based on traditional rodent ischemic injury models are thus not optimally targeting WM ischemic injury. In contrast, newer experimental models, whether in rodents or higher order animals, that specifically focus on ischemic injury to WM, may offer novel and translationally relevant insights that could lead to identification of new and more promising cellular and molecular therapeutic targets.

Ischemic preconditioning (IPC) is an endogenous protective phenomenon in which a short ischemic event induces protection against subsequent ischemic injury (Gidday, 2006). IPC-mediated protection is robust and present in multiple invertebrate and vertebrate models (Gidday, 2006), including non-human primates (Bahjat et al., 2011). Retrospective clinical analyses of human stroke patients suggest that exposure to brief ischemia during a transient ischemic attack (TIA) confers protection against subsequent ischemic injury and reduces stroke severity (Moncayo, de Freitas, Bogousslavsky, Altieri, & van Melle, 2000; Wegener et al., 2004). Defining the cellular and molecular mechanisms of IPC is vital for identifying and advancing neurotherapeutics of stroke (McDonough & Weinstein, 2020).

To study IPC and ischemic injury in WM, we developed a novel sequential in vivo/ex vivo model of IPC and acute ischemia with functional measurements of axonal protection and recovery in the mouse optic nerve (MON) (Hamner et al., 2015). The model consists of an in vivo IPC stimulus, an intervening period of recovery, and then removal of the MONs for exposure to an ex vivo ischemic injury in the form of oxygen-glucose deprivation (OGD). The MON is a purely myelinated CNS WM tract and is one of only a few models that have been used extensively to study ischemic injury in WM (Baltan et al., 2008; Hamner et al., 2015; Hilla, Diekmann, & Fischer, 2017; Matute, Domercq, Perez-Samartin, & Ransom, 2013; Ransom et al., 2004; Stys, Waxman, & Ransom, 1992). It offers several advantages including capacity to quantitatively and dynamically assess ischemic injury using both physiological and anatomical methods (Baltan et al., 2008; Hamner et al., 2015). The optic nerve is also ideal for studying axonal energy metabolism dynamics ex vivo (Looser, Barrett, Hirrlinger, Weber, & Saab, 2018; Tekkok et al., 2005) due to its location and ease of ex vivo study. As such the MON has become a standard model for characterizing fundamental WM-specific molecular mechanisms of ischemic injury (Baltan et al., 2008; Hilla et al., 2017; Matute et al., 2013; Ransom et al., 2004; Stys et al., 1992).

Induction of IPC causes marked genomic reprogramming in whole tissue sections from mouse cortex (Stenzel-Poore, Stevens, King, & Simon, 2007; Stenzel-Poore, Stevens, & Simon, 2004). We attributed many of these changes to cortical microglia (McDonough et al., 2017; McDonough et al., 2020), the specialized tissue macrophages of the CNS. These findings were intriguing as microglia play a significant role in the neuroinflammatory response of many neurological diseases and injuries including stroke (Garden & Moller, 2006; Perry & Holmes, 2014; Weinstein, Koerner, & Moller, 2010). Although activation of microglia was once considered to be an undesirable and destructive pro-inflammatory process, we and others are building a body of literature that suggest microglia can enhance and accelerate recovery and regeneration after neurologic injury (Hanisch & Kettenmann, 2007; Kettenmann, Hanisch, Noda, & Verkhratsky, 2011; Michell-Robinson et al., 2015) particularly in the context of IPC-mediated protection (Hamner et al., 2015; McDonough et al., 2017; McDonough et al., 2020; McDonough & Weinstein, 2020). Using our MON model of IPC, we demonstrated that IPC-mediated protection is present in WM (Hamner et al., 2015). Furthermore, we showed that this protection is dependent on innate immune signaling pathways (Hamner et al., 2015). However, the specific cell type(s) involved in establishing IPC-mediated axonal protection in the WM are unknown. Due to the high expression of these innate immune signaling pathways in microglia (Hamner et al., 2015; McDonough et al., 2017; McDonough et al., 2020), we hypothesize that microglia are critical mediators of IPC-induced axonal protection in WM.

Colony stimulating factor 1 receptor (CSF1R) is selectively expressed on microglia and is required for the development of microglia and maintenance of adult populations (M. R. Elmore et al., 2014). Several pharmacological agents including PLX5622 have been repurposed for their ability to selectively deplete microglia via antagonism of CSF1R (Hilla et al., 2017; Huang et al., 2018; Rice et al., 2017; Rice et al., 2015). Data from some microglial depletion studies suggests that there may be benefits to depletion/replenishment in certain CNS pathologies (M. R. P. Elmore et al., 2018; Rice et al., 2017), however several recent stroke studies report that CSF1R inhibitor-induced depletion of microglia worsens outcomes (Jin et al., 2017; Szalay et al., 2016). The latter findings support our hypothesis that microglia are important for endogenous cytoprotection and recovery processes after ischemia. Establishing a key role for microglia in WM IPC is important. We and others have previously reported findings that implicate microglia (Weinstein et al., 2010) in the physiology of ischemic injury (Nedergaard & Dirnagl, 2005) and recovery (McDonough & Weinstein, 2020) in rodent models of stroke that primarily affect GM. However, there are significant differences between WM and GM microglia at the levels of morphology, cell density, and in subpopulation dynamics, transcriptomics, and surface protein expression (Bottcher et al., 2020; Lawson, Perry, Dri, & Gordon, 1990; Sankowski et al., 2019).

Using our novel IPC in MON model, we examined the effects of pharmacologic depletion of microglia on IPC-mediated axonal protection in WM to demonstrate preservation of WM structure and axonal function. The model of ex vivo OGD allows us to record compound action potentials (CAPs) and assess axonal function in real-time. To assess the structural integrity of both axons and their supporting glia post-ischemia, we used immunohistochemical (IHC) methods to determine the role that microglia play in the preservation of structural components in the MON after IPC.

Methods

Treatment with PLX5622

Adult C57BL/6 (WT) mice (Jackson laboratory) 12–16 weeks of age were fed AIN-76A rodent diet (control) or AIN-76A formulated with 1200 ppm colony stimulating factor-1 receptor (CSF1R) antagonist PLX5622 (Plexxikon) for 21 days; this dose and duration have been previously shown to deplete microglia in the brain and optic nerve (Hilla et al., 2017; Rice et al., 2015). After this period of treatment, animals were sacrificed and MONs were collected for analysis of microglial depletion (Figure 1A), electrophysiological recordings with OGD only (Figure 2A), or received IPC and subsequent electrophysiological recordings with OGD (Figure 2D). All experiments were done in accordance with National Institute of Health (NIH) guidelines and approval from our Institutional Animal Care and Use Committee (IACUC). No adverse events or untimely mortality were present in any treatment group or study cohort.

Figure 1.

Microglial depletion after PLX5622-treatment. A. Diagram showing experimental paradigm for depleting microglia. Animals were fed either AIN-76A control rodent chow or AIN-76A formulated with 1200 ppm PLX5622 for 21 days prior to collection of MONs for IHC analysis on effects of microglia depletion in the MON. Confirmation of microglial depletion is reported here, effects on axonal function and structure are reported in Figures 2 and 3, respectively. Effects on mature oligodendrocytes are reported in Figure 4. B. Quantification of microglial numbers (Iba1⁺/Tmem119⁺ cells) in control and PLX5622-treated MONs demonstrates significant decrease in microglial number after PLX5622 treatment ***p-value = 0.0006; unpaired t-test. C. Representative images from animals receiving control diet (top row) or PLX5622 (bottom row). Sections are stained for DAPI (blue), Iba1 (green), and Tmem119 (red); scale bar = 50 μm.

Figure 2.

Recovery of supramaximal compound action potentials (CAP) of MONs after ex vivo oxygen-glucose deprivation (OGD). A. Experimental scheme for panels B and C: mice were fed control or PLX5622-infused chow for 21 days prior to removal of MONs for ex vivo OGD. B. CAP area in MONs exposed to 45 min OGD. Percent CAP recovery at 60 min after restoration of normoxic/normoglycemic conditions is represented in b’; PLX5622 treatment had no effect on axonal function under these conditions: p-value = 0.8787 (unpaired t-test), N=6. C. CAP area in MONs exposed to 60 min OGD. c’. Percent recovery comparison shows no effect of PLX5622-treatment on axonal function and recovery in response to a more severe injury: p-value = 0.4204 (unpaired t-test), N=7. D. Experimental scheme for panels E and F: mice were fed control or PLX5622-infused chow for 21 days, IPC (15 min CCAO) was induced, and MONs were collected for ex vivo studies 3 days later. E. CAP area in MONs from control chow fed animals receiving IPC. Preconditioned optic nerves are labeled as ipsilateral (ipsi) to the IPC stimulus. e’. Percent recovery 60 min after restoration of normoxic/normoglycemic conditions demonstrates IPC-mediated protection in ipsi MONs compared to contralateral (contra) MONs that did not receive preconditioning: *p-value = 0.0270, N=5–6. F. CAP area in MONs from PLX5622-treated animals receiving IPC. Percent recovery is depicted in f’ and indicates loss of IPC-mediated protection: p-value = 0.3212, N=6.

Common carotid artery occlusion surgery

To induce IPC, unilateral common carotid artery occlusion (CCAO) surgery was performed as previously published (Hamner et al., 2015). In brief, animals were anesthetized and the common carotid artery was exposed. The common carotid artery was ligated for 15 min and then released to permit reperfusion. Animals were exposed to 4% isoflurane by inhalation for induction of surgical plane and tapered to 2.5% for maintenance of surgical plane during surgery. Temperature was monitored pre-, peri-, and post-operatively until animals were ambulatory with body temperature maintained at 37^oC through heat pad support. 50 mg/kg body weight of buprenorphine was administered prior to surgery for pain management. Only one MON is exposed to ischemia which allows us to use the second MON as an internal control. This eliminates a key source of biological variability since ipsilateral and contralateral MONs are subjected to the same conditions (i.e. diet, surgery), obtained from the same animals, and are isolated in parallel.

Ex vivo oxygen-glucose deprivation (OGD)

For ex vivo studies, MONs were acutely excised from animals sacrificed following approved IACUC methods. CAPs were recorded as previously described (Hamner et al., 2015; Rich, Patrick, Hamner, Ransom, & Brown, 2020). In brief, ipsilateral and contralateral MONs were placed in an interface perfusion chamber (Harvard Apparatus) superfused with artificial cerebral spinal fluid (CSF) containing the following (in mmol/L): 125 NaCl, 3.0 KCl, 2.0 CaCl2, 2.0 MgSO4–7H2O, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose, pH 7.45. MONs were maintained at 37^oC for the duration of ex vivo experiments. Suction electrodes were placed on either end of the MON. Supramaximal stimuli were applied orthodromic to the MON to elicit maximal CAPs. To induce OGD in MONs the superfusion solution was rapidly switched to an oxygen/glucose-free solution (containing an osmotically similar concentration of sucrose; oxygen was depleted by continuously bubbling with 95%N₂/5%CO₂). MONs were allowed to equilibrate for 30 – 60 min to establish stable baseline CAP measurements before exposure to OGD followed by a recovery period. Data were recorded as previously described (Hamner et al., 2015) using proprietary software (Clampex). MON function was quantified by integrating the area under the CAP (Hamner et al., 2015). The normalized CAP area at any time is proportional to the relative number of functioning axons (Stys et al., 1992).

Immunohistochemistry

Animals were sacrificed following approved IACUC methods. For treatment (PLX5622 vs control diet) controls and determining effect of CCAO on the retina, the MONs or retinas were removed acutely and post-fixed for 24 hours in 4% paraformaldehyde (PFA). For IHC analysis of MONs that were exposed to OGD (with or without IPC), MONs were removed from the perfusion chamber at the conclusion of recordings and drop-fixed for 24 hours in 4% PFA. After fixation, MONs were rinsed in phosphate buffered saline (PBS), cryo-protected in 10%, 20%, and 30% sucrose, embedded in optimal cutting temperature (OCT) compound (Fisher), and sectioned at 12 μm on a cryostat. Slides were blocked in 10% donkey serum (Jackson ImmunoResearch) and 0.1% Triton X-100 (Sigma) for 30 min. Primary antibodies used: goat anti-Iba1 (Abcam, 1:500) or rabbit anti-Iba1 (Wako, 1:250), rabbit anti-TMEM119 (Abcam, 1:1000), sheep anti-Chx10/Vsx2 (Exalpha, 1:300), mouse anti-HuC/D (Invitrogen, 1:500), mouse anti-SMI31 (BioLegend, 1:500), and mouse anti-APC (Millipore, 1:200). Slides were incubated with appropriate secondary antibodies conjugated to Alexa-488 or Alexa-568 (Abcam; 1:500) and stained with 4’,6-diamidino-2-phenylindole (DAPI) (Sigma, 1:1000). Slides using Iba1, Tmem119, and/or APC antibodies were boiled for 3 minutes in citrate buffer for antigen retrieval prior to the block step.

Imaging and analysis

Imaging was done with a Marianas imaging system using a Zeiss Axiovert 200M microscope and Slidebook software. We used standard stereological methods (Howard & Reed, 2010) and the Slidebook stereology module for selecting random sample sites across MONs. For quantification of cells within these sample sites, we used the optical dissector method to quantify the number of cells within a 50 μm x 50 μm field of view following standard counting rules (Howard & Reed, 2010). To assess the extent of SMI31 staining in MONs, the ImageJ particle analysis plugin was used following standard published protocols (Bennett & Brody, 2015; Kneynsberg, Collier, Manfredsson, & Kanaan, 2016). To briefly summarize: images were converted to 8-bit and channels were split to select the SMI31 immunopositive channel for analysis. Where applicable, regions of interest (ROIs) were chosen to omit edge or blur effects. The MaxEntropy thresholding algorithm was applied and then the “analyze particle” function was used with following parameters (Bennett & Brody, 2015; Kneynsberg et al., 2016): size (micron²) 0–∞ and circularity 0–1. Data output included object count, total area of detected objects, average object size, and percent area.

Experimental design and statistics

Animals were assigned to treatment groups by block randomization and investigators were blinded to treatments for electrophysiological and IHC outcomes; animals were assigned non-descriptive codes. Animal numbers are indicated in text, figures, and/or figure legends, and were calculated in order to obtain statistically significant results based on power analysis (p<0.05, 90% power, variance <20% and anticipated effect size >40% (Hamner et al., 2015)). Data available from the authors on request.

Statistical evaluation was performed using PRISM 8 software (GraphPad). Comparisons between multiple experimental groups were made using one or two-way ANOVA with appropriate post-tests; these are indicated in the Results section and/or figure legends for each experiment. For comparisons between single experimental and control groups we used student’s t-test; p-values <0.05 are considered significant. Presented data are mean ± SEM.

Results

Treatment with PLX5622 depletes microglia in the mouse optic nerve

Mice were fed either control or PLX5622-infused chow for 21 days to deplete microglia from the MON (Hilla et al., 2017) (Figure 1A). Myeloid cells in the MON were quantified based on expression of Iba1 and/or Tmem119. Iba1 is a pan-myeloid marker that labels both microglia and peripheral monocytes/macrophages, whereas Tmem119 is a microglia specific marker (Satoh et al., 2016). In control MONs we counted 0.3846 ± 0.03734 (N=6) Iba1⁺/Tmem119⁺ cells per stereological position. In MONs from PLX5622 treated mice we quantified 0.01256 ± 0.005398 (N=6) Iba1⁺/Tmem119⁺ cells per position (Figure 1B) indicating a 96.4% reduction. The overwhelming majority of cells were positive for both Iba1 and Tmem119 (Figure 1C). In control MONs we counted 166 cells across 440 positions in six animals. Of these, 163 were Iba1⁺/Tmem119⁺ (98.1% ± 1.4% of all cells), two were Iba1⁻/Tmem119⁺, and one was Iba1⁺/Tmem119⁻. In PLX5622-treated MONs we counted six cells across 469 positions in six animals; all were Iba1⁺/Tmem119⁺.

Microglial depletion does not affect axonal recovery after exposure to acute ischemia

To test if microglial depletion alters the functional response to acute ischemic injury alone, we recorded supramaximal CAPs from control and PLX5622-treated MONs exposed to acute injury of varying duration (Figure 2A): 45 minutes (Figure 2B) or 60 minutes of OGD (Figure 2C). In MONs from both control and PLX5622-treated mice, longer duration of OGD reduced the recovery of CAPs after restoration of normoxic/normoglycemic conditions (comparing 2c’ to 2b’; p-value = 0.0003, two-way ANOVA, N=6). These findings are consistent with our prior reports (Baltan et al., 2008; Hamner et al., 2015) and indicate that our methods of measuring axonal function and recovery are sensitive and reproducible. Notably, CAP recovery in PLX5622-treated MONs was not significantly different from control MONs after acute injury (Figure 2B–C; p-value = 0.8316, two-way ANOVA, N=6). These findings are important because they allow us to ascertain baseline values of expected recovery after exposure to acute ischemic injury alone and establish that PLX5622 treatment does not affect this recovery.

Microglial depletion abolishes IPC-mediated protection of axonal function

Our prior work in MON demonstrated that the protective effect of IPC on axonal function after OGD is dependent on innate immune signaling pathways such as Toll-like receptors (TLRs) and type 1 interferons (IFNs) (Hamner et al., 2015). These pathways are highly expressed in microglia (Hamner et al., 2015; Janova et al., 2015; McDonough et al., 2017; Weinstein et al., 2010), indicating that microglia may play a critical role in IPC. To test this hypothesis, animals treated with control or PLX5622-infused chow received a 15 min CCAO to precondition one MON (ipsilateral). After 72 hours of reperfusion, MONs were collected for ex vivo recording and exposure to 45 min OGD (Figure 2D). In animals receiving control chow, CAP recovery after OGD was significantly greater in ipsilateral MONs (38.72% ± 1.923%) compared to contralateral MONs (24.19% ± 3.169%) (Figure 2E, p-value = 0.0270, FDR posthoc test, N=5–6). This finding confirms our previous work reporting that IPC-mediated axonal protection is present in this model of WM IPC in the MON (Hamner et al., 2015). In contrast, in PLX5622-treated animals we did not observe differences in CAP recovery after OGD between ipsilateral (25.98% ± 3.876%) and contralateral (31.87% ± 5.997%) MONs (Figure 2F, p-value = 0.3212, FDR posthoc test, N=6). Based on these data, we conclude that IPC-mediated axonal protection is absent in nerves treated with PLX5622 (and with corresponding depletion of microglia) (p-value = 0.0248, two-way ANOVA, N=5–6). Furthermore, based on our previous experiments (Figure 2B–C), we are able to rule out the possibility that PLX5622 treatment results in altered CAP recovery after OGD alone.

IPC-mediated protection of axonal structure is blocked by CSF1R antagonism

In our previous report we observed abnormal neurofilament staining and reduced axonal structural integrity in MONs exposed to OGD, which was partially attenuated by IPC (Hamner et al., 2015). Here we examine effects of microglial depletion via CSF1R antagonism on axonal structure after IPC and OGD (Figure 3) using anti-phosphorylated neurofilament antibody (SMI31). SMI31 staining decreases after injury due to loss of neurofilament phosphorylation specifically (Kashiwagi et al., 2003; McLean, Popescu, Gordon, Zochodne, & Verge, 2014; Schultz et al., 2017) and/or axonal degradation generally (Bennett & Brody, 2015; Kneynsberg et al., 2016). We used ImageJ to quantify the brightness and size of SMI31⁺ fragments to evaluate phosphorylated neurofilament staining in the MON as an indicator of neurofilament health (Schultz et al., 2017) and preservation of axonal structure (Bennett & Brody, 2015; Kneynsberg et al., 2016). The first outcome we examined was the percent area of immunostaining of phosphorylated neurofilaments in each field of view we imaged within the MON (Figure 3B–C). We compared SMI31 staining in MONs from animals that received control or PLX5622-infused chow (see Figure 1A for experimental paradigm). We observed bright and extensive SMI31 staining in these treatment groups (Figure 3A). The average percent area of immunopositivity was similar in both control-treated (7.418 ± 1.508%) and PLX5622-treated MONs (4.700 ± 0.6534%) (Figure 3B; p-value = 0.1492, unpaired t-test, N=4) indicating that PLX5622 did not alter the extent of neurofilament phosphorylation or axonal integrity. When MONs were exposed to 45 min ex vivo OGD (see Figure 2D for experimental paradigm), the amount and brightness of SMI31 staining decreases in all experimental groups (Figure 3A), which matches our previous report (Hamner et al., 2015). Qualitatively, we observed that in control diet animals the OGD-induced loss of SMI31 signal in ipsilateral (preconditioned) MONs is attenuated compared to contralateral MONs, which matches our previously reported data (Hamner et al., 2015). However, IPC-mediated attenuation of OGD-induced loss of SMI31 signal appears absent in MONs from PLX5622-treated mice (Figure 3A). Our quantitative analysis (Figure 3C) substantiates the qualitative observations. PLX5622-treatment eliminates the IPC-mediated protection seen in the MON from control diet fed mice (p-value = 0.0224, two-way ANOVA, N=6). In control-treated animals there is a statistically significant increase in SMI31⁺ area in ipsilateral MONs relative to the contralateral (p-value = 0.0170, Sidak’s multiple comparisons test, N=6). Furthermore, our quantitative analysis reveals no significant difference between contralateral and ipsilateral MONs from PLX5622-treated animals (p-value = 0.8114, Sidak’s, N=6).

Figure 3.

SMI31 (phosphorylated neurofilament) staining is used to assess axonal structural integrity in MONs. A. Representative staining of SMI31 (green) in multiple treatment groups; scale bar = 50 μm. MONs from control- or PLX5622-treated animals were preconditioned (ipsi) or not (contra) 72 h prior to 45 min ex vivo OGD. To assess the extent of SMI31 staining in MONs, the ImageJ particle analysis plugin was used to quantify the percent area (% area) that was immunopositive (B-C) and determine object size after thresholding (D-E). B. The percent area immunopositive for SMI31 does not vary significantly between MONs treated with control- or PLX5622- chow alone (p-value = 0.1492, N=4, t-test); experimental paradigm illustrated in Figure 1A. C. After exposure to OGD, the percent area immunopositive for SMI31 decreases relative to non-OGD exposed MONs (note the differing axis scales between B and C); experimental paradigm illustrated in Figure 2D. Microglial depletion attenuates IPC-mediated effects (*p-value = 0.0224, two-way ANOVA, df = 23, N=6). In control-treated animals, there is a significant increase in ipsi MONs relative to contra MONs (#p-value = 0.0170, Sidak’s multiple comparisons test, N=6). There is no significant difference between contra and ipsi PLX5622-treated MONs (p-value = 0.8114, Sidak’s, N=6). D. The average size of SMI31⁺ objects between control and PLX5622-treated MONs is non-significant (p-value = 0.2157, N=5, t-test). E. After exposure to OGD, the average size of SMI31⁺ objects is decreased relative to non OGD controls (note differing axis scales between D and E). The effect of microglial depletion on object size after exposure to OGD is non-significant but trending (p-value = 0.0511, two-way ANOVA, df = 20, N=6). However, in control-treated MONs, there is a statistically significant increase in average size in ipsi MONs compared to contra (*p-value = 0.0400, N=6, Sidak’s). Contra and ipsi values are similar in PLX5622-treated MONs (p-value = 0.9007, N=6, Sidak’s).

Next, we examined the size of the SMI31⁺ fragments in each field of view using the particle analysis tool in Image J. Axonal degradation should result in both reduced and more fragmented SMI31 staining (Kashiwagi et al., 2003; Kneynsberg et al., 2016; Schultz et al., 2017) (as observed in OGD-exposed MONs in Figure 3A) and thus a decrease in the average object size. We quantified object (fragment) size in animals that received only PLX5622 treatment (or control diet) and the means were similar (p-value = 0.2157, t-test, N=5) (Figure 3D). This is consistent with the percent area outcome (Figure 3B) and our qualitative observations (Figure 3A). Object size decreases in all OGD-exposed MONs (compare Figure 3E to 3D, and note scale of the Y-axes in both). Our two-way ANOVA p-value of this dataset is 0.0511 (N=6), which just misses statistical significance but suggests a trending effect of microglial depletion on reducing IPC-mediated protection for this parameter as well. There is a significant increase in the average size of SMI31⁺ fragments in ipsilateral control MONs compared to contralateral control MONs (p-value = 0.0400, Sidak’s, N=6). In PLX5622-treated animals, the mean object size is similar between contralateral and ipsilateral MONs (p-value = 0.9007, Sidak’s, N=6). These observations are consistent with our percent area findings (Figure 3C), qualitative observations (Figure 3A), and our CAP recovery (Figure 2) data.

IPC-mediated protection of APC⁺ oligodendrocytes is eliminated by CSF1R antagonism

As we previously published, WM IPC protects APC⁺ oligodendrocytes in the preconditioned (ipsilateral) MON (Hamner et al., 2015). We sought to confirm these prior findings and determine the impact of CSF1R antagonism via PLX5622 on IPC-mediated protection of oligodendrocytes (Figure 4). To control for possible off-target effects of PLX5622 on oligodendrocytes we quantified the number of APC⁺ cells in MONs that received either control diet or PLX5622-infused diet for 21 days (Figure 4B). In MONs fed a control diet, we found a similar number of APC⁺ cells (2.524 ± 0.4551 per stereological section) compared to those fed the PLX5622 diet (2.625 ± 0.7283). Thus, we found that 21 days of treatment with PLX5622 does not impact total APC⁺ oligodendrocytes in the MON (p-value = 0.7994, unpaired t-test, N=5). When MONs were exposed to 45 min ex vivo OGD, the average number of APC⁺ oligodendrocytes decreased across all experimental conditions (Figure 4A, 4C). Furthermore, control-treated MONs exhibited the anticipated preconditioning effect based on our prior publication (Hamner et al., 2015): the number of APC⁺ cells per position was decreased in the contralateral MON relative to the ipsilateral MON (p<0.0001, Sidak’s multiple comparisons test, N=6). In PLX5622-treated MONs, there is no significant difference between APC counts in the contralateral and ipsilateral MONs (p-value = 0.9180, Sidak’s, N=6). Thus, we conclude that the IPC-mediated preconditioning (protective) effect on mature oligodendrocytes is lost in MONs derived from mice treated with PLX5622.

Figure 4.

APC⁺ oligodendrocytes in control-diet and PLX5622-diet MONs decrease when exposed to OGD. MONs from control- or PLX5622-treated animals were preconditioned (ipsi) or not (contra). A. Representative staining of APC⁺ oligodendrocytes (green) in multiple treatment groups; scale bar = 50 μm. B. Control fed and PLX5622 fed animals that did not receive OGD (experimental paradigm in Figure 1A) have abundant APC⁺ oligodendrocytes and the number of APC⁺ cells is not affected by microglial depletion (p-value = 0.7994, N=5, t-test). C. Microglial depletion significantly alters the IPC-mediated protection of mature (APC⁺) oligodendrocytes (***p-value = 0.0002, two-way ANOVA, N=6). In control-treated MONs, there is a significant increase in APC⁺ oligodendrocyte survival in ipsi MONs relative to contra MONs after exposure to OGD (experimental paradigm in Figure 2D) (####p<0.0001, N=6, Sidak’s multiple comparisons test). In contrast, oligodendrocyte survival in PLX5622-treated MONs does not differ between contra and ipsi MONs (p-value = 0.9180, N=6, Sidak’s).

Transient ischemia does not result in retinal injury

The 15 min CCAO that induces IPC-mediated protection in control MONs also exposes the ipsilateral retina to ischemia (May & Lutjen-Drecoll, 2002). One possible mechanism of IPC could be that retinal neurons become injured during this brief ischemic event and subsequently produce anterograde signals. In turn, this could result in microglial activation in the MON and initiate transcriptomic programs that result in cytoprotection in the optic nerve. To rule out this possibility, we examined the effect of transient CCAO on retinal tissue (Figure 5A). There was no difference in the numbers of Vsx2⁺ bipolar neurons (Figure 5B) or HuC/D⁺ ganglion and amacrine cells (Figure 5C) between ipsilateral and contralateral retinas 72 hours after IPC. We did not observe morphological changes in either microglia (Figure 5D) or astrocytes (Figure 5E), or microglial proliferation (Figure 5D), indicative of glial activation in preconditioned retinas. We conclude that our preconditioning stimulus (15 min CCAO) does not result in retinal injury or glial activation, thus attributing the protective responses we observe in IPC to intrinsic cellular processes in the MON rather than external effects of retinal neurons or glia.

Figure 5.

IPC does not cause neuronal loss or glial activation in the retina. A. Experimental scheme for retinal experiments: IPC (15 min unilateral CCAO) was performed and three days later the retinas were collected for analysis. B. Vsx2⁺ neurons were quantified in control (contralateral; contra) and preconditioned (ipsilateral; ipsi) retinas. No neuronal loss due to this transient ischemic pulse was observed: p-value = 0.7518, N=6. C. HuC/D⁺ ganglion (p-value = 0.8690, N = 6) and amacrine cells (p-value = 0.6714, N = 4–5) did not differ in morphology or number between contralateral and ipsilateral retinas. D. Immunostaining for Iba1 did not reveal changes in microglial morphology or number after IPC (p-value = 0.7676, N = 3). E. GFAP immunostaining also did not demonstrate changes in astrocyte morphology. Scale bars: 50 μm. Abbreviations: ONL - Outer nuclear layer, INL - inner nuclear layer, GCL - Ganglion cell layer.

Discussion

Protection from ischemic preconditioning in axons is mediated by microglia

Here we demonstrate that CSF1R antagonism profoundly depletes microglia in MONs and abolishes IPC-induced protection for axons. This is the first direct demonstration that IPC-mediated protection in the CNS is dependent on a particular cell type, namely microglia. These findings fit well with our prior work (Hamner et al., 2015), which demonstrated a critical role for innate immune signaling in IPC-mediated axonal protection. Our electrophysiology results oriented around MON function are supported by our structural/histological investigations of axonal integrity and oligodendrocyte survival. Interestingly, microglial depletion affected axonal recovery from ischemic injury only in experimental paradigms where the microglia were primed by an IPC stimulus (CCAO) prior to prolonged ischemic exposure (OGD). This finding is consistent with the recent demonstration that adoptive transfer of lipopolysaccharide (LPS) preconditioned, but not naïve, monocytes can induce protection against subsequent cerebral ischemic injury (Garcia-Bonilla et al., 2018). Our finding that microglial depletion abolishes IPC mediated protection is also consistent with findings from prior studies demonstrating that: (i) cellular proliferation is required for induction of ischemic tolerance (Maysami, Lan, Minami, & Simon, 2008) and (ii) the vast majority of proliferating cells in brain following IPC are microglia (McDonough et al., 2020). Our WM results also suggest that the microglial response to IPC and subsequent prolonged ischemia represent a case study of innate immune memory – a functionally altered immune response, specifically from a myeloid cell population, after a secondary stimulus is delivered (Neher & Cunningham, 2019).

Microglial depletion does not alter axonal function at baseline or recovery from acute ischemic injury alone

Notably, CSF1R inhibition had no effect on either baseline axonal function or axonal response/recovery to prolonged acute ischemic injury alone. These data suggest that although microglia are required for IPC-mediated axonal protection in WM, they do not significantly affect either axonal functional homeostasis or responses to acute ischemia. The latter finding is surprising considering that prior reports indicate that stroke outcomes in mice are exacerbated with CSF1R antagonist treatment (Szalay et al., 2016). There are several possible reasons for this discrepancy including differences in timing of ischemic exposure and experimental model choice. The study by Szalay et al used a GM-predominant model, middle cerebral artery occlusion (MCAO) in mice, and assessed outcomes of infarct volume and neuronal function 24 hours after ischemic injury (Szalay et al., 2016). Here we report outcomes of axonal structure and function in a CNS WM tract during a period within hours of ischemic injury. However, our results are in line with another acute WM injury model (optic nerve crush) where microglial depletion also had no effect on outcome (Hilla et al., 2017). The lack of effect of CSF1R antagonist treatment at baseline is also consistent with several prior studies demonstrating that administration of CSF1R antagonists in developmentally mature (but not aged) mice does not influence synaptic or neurobehavioral function in the absence of other neuropathology (M. R. Elmore, Lee, West, & Green, 2015; M. R. Elmore et al., 2014).

Effects of microglial depletion on preconditioning mediated protection of axonal integrity

Measurements of axonal integrity (Figure 3) suggest that microglial depletion eliminates IPC mediated protection of axons against subsequent prolonged ischemic injury. Analyses comparing multiple groups within this data demonstrate that only the ipsilateral and contralateral values in the control chow fed group are significantly different when compared to each other. This IPC mediated protection is abolished in the MONs from PLX5622 treated mice. Similarly, qualitative observations in Figure 3A are supported by our electrophysiological data in Figure 2 showing abrogation of IPC-mediated protection of axonal function. Our axonal integrity data correlates well with our electrophysiology data and this is hardly surprising given that structural integrity of axons goes hand in hand with improved conductivity and function (Hamner et al., 2015; Stys et al., 1992; Tekkok & Goldberg, 2001). Furthermore, this indicates that the IPC-induced attenuation of OGD-induced loss of axonal function has a structural correlate within the MON WM itself.

Effects of microglial depletion on preconditioning mediated protection of oligodendrocytes

Our finding that CSF1R antagonist treatment eliminates IPC-mediated effects on viability of mature APC⁺ oligodendrocytes in the MON suggests a direct connection between microglial presence and oligodendrocyte resilience to ischemic WM injury. The results in control fed animals confirm our prior results (Hamner et al., 2015) but the absence of preconditioning effect in MONs from the PLX5622 fed mice is a new finding and implicates microglia specifically. Interestingly, the average counts for PLX5622-treated MONs are close to the ipsilateral value of MONs from control diet fed animals (Figure 4C). Sidak’s multiple comparisons tests performed after the two-way ANOVA support a nonsignificant difference between control-treated ipsilateral MONs and PLX5622-treated ipsilateral (p-value = 0.8609) and contralateral (p-value = 0.6331) MONs. Although the difference in the number of APC⁺ cells in the ipsilateral and contralateral MONs from the control diet fed mice is clear (and statistically significant), we might have anticipated that the PLX5622-treated MON groups would have lower APC⁺ cell numbers closer to those of the contralateral MON group from the control fed mice. Although treatment with PLX5622 did not affect the number of APC⁺ cells in otherwise naïve mice under homeostatic conditions (Figure 4B), it did eliminate IPC mediated protection against the deleterious effects of prolonged OGD on APC⁺ cell survival. This suggests that preconditioning may have skewed the microglial phenotypic response to OGD such that it was ‘less toxic’. In other words, in response to acute ischemic injury alone, microglia play a role in decreasing oligodendrocyte survival. However, if microglia were preconditioned (or primed) prior to prolonged ischemia they shift toward being protective (or at least benign) to oligodendrocytes. Identifying the molecular cues responsible for this shift in microglial behavior after ischemic injury is critically important, as microglial depletion seems unlikely to be a viable therapy for acute stroke (Szalay et al., 2016). Our prior work suggests that type 1 IFN signaling pathways in microglia/macrophages are activated following preconditioning (McDonough et al., 2017) and may contribute to this phenotypic adjustment (Hamner et al., 2015). These findings further suggest that modulating microglial function in the context of acute ischemic stroke could preserve brain connectivity and improve clinical outcomes.

There is a subtle difference between our axonal integrity and oligodendrocyte viability results following microglial cell depletion (Figures 3, 4). Both sets of data demonstrate: (i) no effect of microglial depletion on baseline MON status and (ii) microglial depletion eliminates IPC mediated effects. In the SMI31 data set (Figure 3), the IPC mediated effect on axonal integrity appears to be protective against ischemic injury whereas in the APC data set (Figure 4), the IPC mediated effect appears to be an elimination of cytotoxicity after ischemic injury. With regards to axonal structural integrity, preconditioning presumptively shifts the microglial phenotype toward an actively protective state such that OGD-induced damage to axon structure is partially attenuated; a protection that is not seen in microglia-depleted MONs. In contrast, with regards to oligodendrocyte survival in MONs from control diet fed mice, we observe that preconditioning alters the response to OGD such that the outcome is similar to microglia-depleted MONs (regardless of preconditioning status). With respect to this parameter only, preconditioning skews the microglial phenotype functionally toward a neutral state, equivalent to the response in the absence of microglia entirely. These differential effects of microglia on oligodendrocytes and axons are intriguing but not without precedent. A recently published study deleted the Csf1r enhancer in mice as a method of genetically ablating microglia (Rojo et al., 2019). In these mice, the lack of microglia did not affect oligodendrogliosis or myelination during development, but displayed disorganization of myelin sheath structures (Rojo et al., 2019). Furthermore, chronic lack of microglia did not affect oligodendrocytes, but did affect formation of new myelin in response to training/learning paradigms and was also sufficient to induce spontaneous demyelination in 6-month old mice (Rojo et al., 2019). Similarly, a study of ischemic injury in neonatal rats also suggests that axons are more susceptible to injury than oligodendrocytes (Suryana & Jones, 2014). These studies and our findings here suggest that myelin/axon structure may be more sensitive to changes in microglia than oligodendrocytes. Characterizing the specific molecular signals in microglia that produce these related but distinct parametric responses remains an important future research direction.

IPC-induced axonal protection is independent of neuronal injury

In our model of IPC the retina is also exposed to transient ischemia (May & Lutjen-Drecoll, 2002), which may be injurious. However, our IHC analysis demonstrates no loss of neuronal viability or glial activation in the retina after IPC (Figure 5) that would indicate injury to these cells (Todd, Finkbeiner, Wong, Hooper, & Reh, 2020). This matches our previous findings, that a 15 minute ischemic event is non-injurious to neurons in the mouse cortex (McDonough et al., 2020). However, it is interesting that we do not see microglial proliferation in the retina in response to brief ischemia, as we observe this in the mouse brain even in the absence of signs of neuronal injury or death (McDonough et al., 2020). Thus, IPC-induced axonal protection in MON is independent of retinal injury or response to brief ischemia and is caused by factors intrinsic to the WM. This is particularly important to note since both IPC (Roth et al., 1998; Zhu, Ohlemiller, McMahan, & Gidday, 2002) and other forms of cross tolerance (Halder et al., 2013) have been reported in the retina. Other reports note that energy status in axons is independent of their neural soma and that preconditioning should not be presumed to originate in soma and spread to the axons (Nguyen, Zerimech, & Baltan, 2021). Localization of the microglia-dependent protective effect to WM has important implications for both mechanism and development of future therapeutics.

Limitations

It is important to acknowledge some limitations of our study. The first is that non-microglial myeloid cells play a role in IPC-mediated protection in GM (Garcia-Bonilla et al., 2018) and may be affected by PLX5622 (Lei et al., 2020). However, our cellular co-localization studies indicate that the infiltrating macrophage population in MON is minimal. Newly developed transgenic mouse lines with high levels of microglial-specificity in their cellular Cre recombinase activity have been described (Kim et al., 2021; Masuda et al., 2020; Rojo et al., 2019) and, when combined with Csf1r-flox mice (Masuda et al., 2020), could be used to try to confirm the microglial specificity of our results. Though powerful, the genetic approach to microglia depletion can result in marked phenotypic changes in the surviving cells (Masuda et al., 2020) and generation of transgenic mice with microglial specific Cre expression can also lead to functional changes in microglia (Cardona et al., 2006; McDonough et al., 2020). Specifically, we have seen that Cx3cr1-haploinsufficiency in transgenic mice alters microglial responses to IPC (McDonough et al., 2020). A second limitation is that CSF1R antagonism could be exerting effects that are independent of microglial viability (Nissen, Thompson, West, & Tsirka, 2018). However, given the context of our current understanding of the mechanism(s) in IPC (McDonough & Weinstein, 2020) we think that PLX5622-induced microglial depletion is a more likely explanation for the elimination of the marked IPC-specific protective effects we see here. A third limitation is that due to the acute time frame in which we can study the MON ex vivo, we cannot identify longer term consequences as a result of improved oligodendrocyte survival in either preconditioned or PLX5622-treated MONs to better elucidate the relationship between oligodendrocyte survival, axonal integrity, and axonal function. Future IPC studies, perhaps involving high resolution in vivo imaging of WM tracts, in which both the primary and secondary ischemic exposures are in vivo, would be better suited for looking at these longer-term outcomes. Despite these limitations, the findings here support our view that microglia are critical cellular contributors to IPC mediated protection in WM and that microglia could be therapeutically targeted to provide robust endogenous protection against ischemic injury in WM.

Clinical relevance of the WM IPC model

The 2019 Stroke Treatment Academic Industry Roundtable (STAIR) committee identified a critical need to study WM protection in rodents to improve translational outcomes of basic science research in the stroke field (Savitz, Baron, Fisher, & Consortium, 2019). However, the rodent brain consists of 13% WM by volume compared to human brains, which are 53% WM (Dirnagl, 2012; Zhang & Sejnowski, 2000). Clinical stroke in humans heavily impacts WM, whereas only approximately 10% of rodent MCAO infarcts affect WM regions of the brain; as such many experimental rodent models of stroke are sub-optimal for evaluating cellular or pharmaceutical interventions that may preserve WM connectivity and function. Mechanisms of ischemic injury differ between WM and GM (Matute et al., 2013; Pantoni, Garcia, & Gutierrez, 1996; Stys et al., 1992) which further limits our ability to elucidate WM-specific mechanisms of injury and protection in rodent whole-brain models. Our MON model directly addresses these limitations and the call to action for improved methods of translating research from rodent to human by focusing on WM-centric outcomes. Furthermore, modeling of WM injury can identify WM-specific effects relevant to multiple forms of human cerebrovascular pathology including acute ischemic stroke and chronic ischemia resulting from cerebral small vessel disease. As noted, our WM IPC model does have limitations, however, the primary alternative - studies using animal models with gyrencephalic brains and a larger volume of WM (ex: ferret, macaque) - are expensive and time intensive. Our model lends itself well to proof of principle evaluations of cytological and axonal protection that can form scientific rationale and justification for subsequent studies in higher order animals.

Conclusions

Microglial depletion, achieved by pharmacological CSF1R antagonism, results in complete loss of IPC-mediated axonal protection in the MON. This has important implications for the role of microglia in protecting axonal structure and function in all CNS WM tracts. Although we previously demonstrated IPC-mediated axonal protection in WM with reliance on innate immune signaling (Hamner et al., 2015), the addition of microglial depletion to the experimental paradigm allows us to evaluate how microglial reprogramming due to IPC affects surrounding glial cells and axons. From this we determine that microglia are critical cyto- and axonal- protective agents in the WM. Longer term studies with in vivo metrics of WM function and integrity are necessary to more fully evaluate how microglia support chronic phases of stroke recovery, however our studies here provide justification for targeting a physiologically relevant cell type in the WM for clinical therapeutics.

Main points:

1. Microglia are required for IPC mediated axonal protection in WM.

2. In the absence of IPC, microglial depletion does not alter axonal function or recovery from injury.

3. The cellular mechanisms of IPC in WM are anatomically intrinsic to WM.