Imaging a concussion and the ensuing immune response at the blood–brain barrier

Significance

Repetitive concussion results in activation of microglia, which work to seal areas of the damaged blood–brain barrier (BBB). This results in progressive phagocytosis of astrocytes, which may ultimately contribute to the long-term cognitive dysfunction that is observed in patients with concussion. By clearing astrocytes at the damaged vasculature, microglia may alter BBB dynamics, cerebrovascular autoregulation, and cellular signaling in the brain, features that are common to brain injury and neurodegenerative diseases.

Abstract

Concussions can cause debilitating symptoms despite no evidence of structural changes on diagnostic imaging. The cellular events occurring in the brain parenchyma following concussion, especially repetitive concussion, are not well elucidated. We developed a concussion model to induce a confined area of injury without causing frank hemorrhage. Using intravital microscopy, we observe activation of the vasculature that supported neutrophil rolling and platelet adhesion but no overt cellular recruitment from blood into brain parenchyma. Activated resident, not monocyte-derived, macrophages relocated to the injury site via Cx3cr1 and phagocytosed dysfunctional/detached astrocytes via scavenger receptors and TLR4, particularly after repetitive concussion. Additionally, microglia sealed areas of blood–brain barrier (BBB) disruption via purinergic pathways. Using a splitCre approach to dissect microglia and perivascular macrophages, we show that microglial invasion into the injury site is key to reducing BBB disruption. Our data suggest that microglia repair the BBB following concussion, but in doing so significantly alter the cellular ultrastructure of the brain milieu.

Traumatic brain injury (TBI) is the leading cause of death and disability in adults (1, 2). TBI is divided into primary and secondary insult. The primary insult is the impact of the trauma itself and is addressed with preventative measures. Secondary brain injury occurs in the hours to days following the initial impact as a result of direct, rotational, and/or shear forces of the trauma. These debilitating mechanisms include axonal injury (3), glutamate excitotoxicity (4), cerebral metabolic derangements (5), ischemia (6), and neuroinflammation (7). Termed a “silent epidemic,” concussion is the most common and mild type of TBI, often overlooked as cases do not present for medical attention and are not represented in epidemiological statistics. The pathogenesis underlying how concussions can cause significant impairment and dysfunction is incompletely understood (8).

Current TBI models cause significant trauma and cerebral contusion (9), involve a craniotomy prior to injury (10) or skull fracture (11). These profound structural changes and the neuroinflammatory response that ensues are not representative of what occurs following a concussion injury. Another study of a mild closed-head TBI involved thinning of the skull bone and induction of compression injury by downward pressure, causing capillary damage (frank hemorrhage) and significant indiscriminate cell death (12). This led to recruitment of peripheral immune cells from the bloodstream including swarming of neutrophils to the meninges (12). Furthermore, monocyte-derived macrophages were recruited and localized to the sites of injury and participated in neoangiogenesis (13). However, this protocol results still in more profound form of brain injury with more structural damage than concussions in humans. For example, in the patients that lost consciousness and/or had amnesia postconcussion and were hospitalized and imaged, only 12% had detectable meningeal hemorrhage and less than 10% had subarachnoid or subdural bleeding (12).

In the current study, we sought to develop a concussion model in mice where no frank hemorrhage was evident. We then examined the neuroinflammatory events occurring in both the meningeal compartment and, specifically, in the brain parenchyma below the dura. This model also allowed us to examine repeated concussions known to lead to greater cognitive and functional decline (14, 15). Our data show a profound loss of astrocytes and a replacement by microglia that help seal the leaky brain vasculature particularly in repeated concussion. The latter occurred at the expense of altering the cellular constituents of the blood–brain barrier (BBB), which may explain the long-term cognitive impairments seen especially in children, athletes, and military personnel subjected to repeated head injuries.

Results

Closed Skull Concussion Model Devoid of Peripheral Immune Cells.

We developed a closed-skull concussion model using a cortical-controlled impactor with a modified rubber tip to prevent skull fracture (termed CON 1X). To examine the effect of repeated concussion (CON 2X), mice were injured for a second time, 24 h following the primary insult. In our closed-skull concussion model, the brain appears normal with no areas of injury visible to the naked eye. Behavioral studies were conducted, revealing minimal differences between the sham and brain-injured mice (SI Appendix, Fig. S1), consistent with the unappreciable neurological deficits observed in humans with concussion. By contrast, a positive-control open-skull injury shows obvious contusion as well as ischemia underlying the injury site (Fig. 1 A and B), in addition to massive peripheral immune cell infiltration of neutrophils and monocytes in whole brain slices, as identified in LysM-GFP animals (Fig. 1C). These dramatic changes were absent in our concussion model (Fig. 1 A–C). We also used intravital imaging to examine in real-time the involvement of peripheral immune cells in the vasculature in the parenchymal compartment below the dura, directly below the injury following concussion and repeated concussion. There was an increase in neutrophil-endothelial cell interactions at the site of concussion suggesting activation of endothelium, but not sufficient activation to induce firm adhesion or extravasation of neutrophils into the parenchyma (Fig. 1 D and E and Movies S1 and S2). Despite the absence of frank hemorrhage, some platelets transiently adhered to the endothelium, but we never saw any inflammatory monocytes adhere to the vasculature (Movies S3 and S4).

Fig. 1.

A closed skull concussion model does not result in invasion of peripheral immune cells. (A) C57 whole mouse brains demonstrate no signs of macroscopic injury in a closed skull TBI model compared to visible contusion in an open skull injury. (B) Brain slices incubated with 2,3,5-triphenyltetrazolium chloride (TTC) show no signs of ischemic injury in a closed skull model versus areas of nonviable tissue (yellow squares) in an open-skull injury model (Left). Quantification of percent of nonviable brain from TTC slice image analysis (Right). Data are expressed as mean ± SEM; individual points represent individual animals. One-way ANOVA compared to sham with Dunnett’s multiple comparisons test, ****P < 0.0001. (C) Whole brain stitched fluorescent confocal microscopy images in the LysM-gfp mouse show infiltration of neutrophils and/or monocytes (green, gfp) in open versus closed skull injury. (Scale bar, 200 μm.) Quantification of neutrophils/monocytes in whole brain slice imaging, ***P = 0.0005, unpaired two-tailed t test. (D) Intravital confocal imaging of mouse brain following closed skull injury shows neutrophil (red, PE) trafficking to the vasculature increases following CON 1X and CON 2X, with resolution beginning at 24 h. n = 3 to 5 per time point, per injury group. (Scale bar, 40 μm.) (E) Intravital confocal microscopy of C57 mouse shows a neutrophil (red, PE) traveling through the vasculature but no neutrophil extravasation following CON 1X or CON 2X, n = 3 per time point, per injury group. (Scale bar, 20 μm.)

Regardless of the lack of peripheral immune cell recruitment, there was significant cell injury. Using propidium iodide applied transcranially, we observed significant microscopic cell death in concussions, which was further increased following multiple challenges (Fig. 2A). While there was no peripheral recruitment of immune cells from the vasculature, microglia responded impressively to injury (Movie S5 and Fig. 2B). Following concussion, there was a dramatic increase in the total GFP signal at the site of injury in Cx3cr1^gfp/+ animals, reflective of microglial activation (via increased soma size), relocation of microglial cell bodies to respond to injury, and thickening of dendrites. These alterations were further augmented in repeated concussion where the activated microglia were noted to preferentially localize around the vasculature (Fig. 2C). To ensure that the relocated microglia were truly resident and not recruited via inflammatory monocytes, we performed repeated concussions in Ms4a3^Cre:LSL-TdTom mice, in which all monocytes and hence monocyte-derived macrophages, but not resident microglia, are permanently red fluorescent. Fig. 2D clearly demonstrates tdTomato-positive cells in the vasculature, but a negligible number of positive inflammatory macrophages among the mass of microglia recruited to the injured parenchyma, even after repeated concussion.

Fig. 2.

Microglia become activated in response to BBB damage. (A) Intravital imaging of C57BL/6 mice following injury. Brain surface stained with propidium iodide. One-way ANOVA, *P = 0.03, **P = 0.002. (Scale bar, 20 μm.) (B) Intravital multiphoton microscopy images of CX3CR1-gfp/wt mice. Left, xyz 3D view stack images. (Scale bar, each square = 30 μm.) Middle, extended focus z-stack images. (Scale bar, 50 μm.) Right, quantification. Microglia (green, gfp) become activated and migrate to the superficial cortical site of injury following TBI. One-way ANOVA, **P = 0.004, ***P = 0.0008. (C) Images (Left) and quantification (Right) of the number of microglia (individually colored) intimately associated with vasculature (red) following brain injury. One-way ANOVA, *P = 0.01, ***P = 0.005. (Scale bar, 50 μm.) (D) Intravital imaging (Left) and quantification (Right) demonstrating Ms4a3-positive macrophages are only found in the intravascular compartment following CON2X, unpaired two-tailed t test, *P = 0.01. (E) Intravital imaging (Left) of Aldh1L1-gfp mice IV injected with 70,000 MW dextran rhodamine show BBB permeability and astrocyte end-feet (ALDH1L1-gfp, green) retract from vasculature following injury (white arrow). (F) BBB permeability following single injury over multiple time points (hours). 300 ng Pertussis toxin (PTX) was injected IV over 2 d as a positive control for BBB disruption. Significance markers represent comparison to Sham. Statistical analysis with one-way ANOVA and multiple comparisons post hoc testing. *P = 0.03, ***P = 0.001, ****P < 0.0001. (G) BBB permeability in single and repetitive concussion. One-way ANOVA, **P = 0.001. (H) Quantification of astrocyte distance from vasculature, one-way ANOVA, ***P = 0.003, ****P < 0.0001 Each dot represents one mouse. Data represent three independent experiments with 4 to 6 mice per group.

We next explored whether concussion causes microvascular damage. We injected 70,000 MW fluorescent dextran intravenously, which remains almost exclusively within the vasculature under sham conditions (Fig. 2E; red in vessel). Following concussion, the dye leaked into the brain parenchyma (Fig. 2 E, Center: red surrounding yellow vessel) peaking at 12 h and declining at 24 and 48 h (Fig. 2F). The vascular leakage occurred in the concussion area and not in unaffected areas. Interestingly, when a second concussion was performed 24 h later and imaged at 12 h post–repeated injury, there was almost no increase in vascular permeability (Fig. 2 E, Right panel and quantification Fig. 2G intravital imaging). As astrocytic end-feet contact the vasculature and play a critical role in maintaining the BBB, these experiments were completed in Aldh1l1^gfp mice to visualize the astrocytes at the BBB. Following concussion, astrocytes detached from the vasculature appearing to be displaced by the leaked fluid (Fig. 2 E, Middle panel). Following repeated concussions, where there was almost no increase in vascular permeability, most astrocytes that normally juxta-position next to the vasculature, had disappeared with only a few cells remaining in the field of view (Fig. 2 E, Right panel), but at significant distances from the vessels (Fig. 2H).

Activated Microglia Attenuate Vascular Permeability Following TBI by Restoring the BBB.

To better understand why the permeability decreased in repeated concussion, we imaged microglia in this low vascular permeability setting in vivo. Fig. 3A shows a time progression of the interaction of microglia with blood vessels. We observed microglia extending their processes around the vasculature and over 1 h completely encasing a significant part of the blood vessel in the injured area (Fig. 3A and Movie S6). To understand the molecular mechanisms underlying this process, we systematically examined a number of pathways that might contribute to activation and recruitment of the microglia to the repeated concussion area. First, because chemokine receptor Cx₃cr1 found on microglia has been shown to play a key role in microglial activation (16) we examined microglial migration after repeated concussion in Cx3cr1^−/− mice. At baseline conditions and following a single injury, there was no significant difference in the number of microglia at the site of impact. The recruitment of microglia was significantly reduced at the site of injury (Fig. 3 B and V). Concomitantly, permeability remained elevated in the repeated concussion model consistent with our observation that the recruited microglia sealed vessels to decrease permeability (Fig. 3D).

Fig. 3.

Activated microglia protect the BBB in repetitive mild TBI. (A) Intravital multiphoton microscopy video sequences in CX3CR1^gfp/+ mice. Microglia (green, gfp) extend their processes to encase injured blood vessels (red) following a TBI. (Scale bar, 10 μm, time stamp in minutes.) (B) Intravital 2P images and (C) quantification of recruitment of microglial cells to the site of injury in CX3CR1^+/− and CX3CR1^−/− mice. (D) Permeability results following CON2X in CX3CR1^−/− mice, unpaired t test, **P = 0.002. Each dot represents one mouse. Data represent three independent experiments with 3 to 5 mice per group. (E) Number of microglia at the site of injury in normal and apyrase-treated conditions. (F) Apyrase-treated microglia can still be activated as measured by soma size (one-way ANOVA, ****P < 0.0001, ***P = 0.0007. (G) xy scatter plots of individual microglia show that apyrase alters microglia morphology by longest-principal axis. Microglia colored based on total area size. (H) Schematic of longest principal axis measurement. The yellow arrowed line shows the length measured. (I) Representative individual microglia in untreated and apyrase-treated conditions highlights shorter principal axis in apyrase-treated microglia. (Scale bar, 20 μm.) (J) Quantification of changes to microglia morphology/longest principal axis following treatment with apyrase. One-way ANOVA, *P = 0.03, ***P = 0.0004. (K) Intravital 2P images of BBB permeability (Left) and quantification (Right) following apyrase treatment, unpaired t test *P = 0.01, **P = 0.006. (L) BBB permeability results following repetitive mild TBI in P2RX7KO mice. Unpaired t test, ***P = 0.0001.

ATP has been shown to function as a danger-associated molecular pattern and to be critical to direct microglial processes toward dying neurons (17). To test for ATP involvement in our model, we hydrolyzed extracellular ATP released by dying cells via transcranial administration of apyrase immediately following repeated concussion. Importantly, following apyrase treatment, microglia still arrived at the repeated concussion site at levels seen in untreated mice (Fig. 3E). This is consistent with previous reports that ATP is unlikely to function as a chemoattractant (18). Microglia at the site of injury following apyrase treatment also still had enlarged soma sizes suggesting that despite ATP inhibition microglia were still activated (Fig. 3F). However, detailed image analysis revealed that the morphology and dendritic pattern of the microglia were markedly altered after apyrase inhibition of ATP. When we characterized the morphology of hundreds of individual microglial cells, those of mice treated with apyrase had a shorter dendrite-to-dendrite length span compared to those of animals that were not treated, indicated by dense grouping in the left lower quadrant (blue) in scatter plots of individual microglia (Fig. 3 G–J). Apyrase-treated microglia were unable to stretch their dendrites to sense and respond to the damaged BBB. Correspondingly, we observed a significant increase in BBB permeability in repetitive concussion when ATP was hydrolyzed with apyrase and microglia could not interact with the vasculature (Fig. 3K). Interestingly, in the single concussion model, apyrase increased the already elevated vascular permeability (Fig. 3K) suggesting that even following one challenge, microglia limited vascular dysfunction. To exclude untoward effects of apyrase we also genetically targeted P2rx7, the purinergic receptor found on microglia that is activated by high concentrations of ATP (19). Repetitive injury significantly increased BBB permeability in P2rx7^−/− mice compared to wild-type controls (Fig. 3L). In total, these experiments suggest that the recruitment, but also the physical interaction with blood vessels is required for microglia to protect the BBB following repeated concussions. The Cx₃cr1 pathway recruits the microglia to the concussion site consistent with previous evidence that Cx₃CR1 is necessary for microglial migration (16), while the ATP-P2RX7 pathway induces the dendrite wrapping around the vasculature in two separate, but complementary effector functions.

Progressive Phagocytosis of Astrocytes By Microglia in Repetitive Injury.

Key roles of microglia are to prune neurons, remove debris under homeostatic conditions, and to clear dead cells during neuroinflammation. Herein using imaging we noted very significant numbers of microglia interacting with detached astrocytes near vascular junctions, but the astrocytes were not dead, as assessed by propidium iodide (PI) (Fig. 4A). We further investigated the specific cell death program that was occurring in astrocytes following TBI via flow cytometric assessment with YO-PRO-1 fluorescent dye, which is permeable to apoptotic cells and PI, which is not. Less than 5% of astrocytes in the injured hemisphere were undergoing cell death (Fig. 4B). The majority of dead or dying astrocytes were apoptotic versus necrotic. In the repeated concussion model, the astrocyte-specific dye SR101 was found extensively in the microglia that were localized at the site of detached astrocytes (Fig. 4C), but only in the injury area. This suggested that either the dye was now being taken up by activated microglia (never reported before) or alternatively, the microglia were engulfing detached astrocytes. To test whether indeed bona fide microglial phagocytosis of live detached astrocytes occurs in the concussed brain, we used PrismPlus mice, which express distinct fluorophores for astrocytes and microglia (20). Following a single concussion, we observed occasional microglial uptake of astrocytes as shown by internalization of Aldh1l1⁺ material in CX₃CR1⁺ microglial cells (Fig. 4D), whereas after repeated concussion, there was much greater microglial phagocytosis of astrocytes (Movies S7–S9). To further verify that microglial phagocytosis of astrocytes occurred after concussion, we performed flow cytometry with Aldh1l1^gfp mice and screened for green fluorescence within microglial cells. Although microglia were isolated from the whole hemisphere and not only the injury site, there was clear phagocytosis of astrocytes in the single concussion model, but not in sham controls, and the uptake was dramatically increased in the multiple concussion model (Fig. 4E). These data indicate that microglia phagocytose astrocytes particularly after a second injury. Additional flow cytometric analysis of neuronal death and contribution of astrocytic phagocytosis was performed (SI Appendix, Fig. S2). This demonstrates that <10% of neurons were apoptotic or necrotic in repetitive concussion with an increase in apoptotic neurons in repetitive injury. The contribution of phagocytosis via astrocytes in repetitive concussion is negligible.

Fig. 4.

Microglia progressively phagocytose astrocytes in repeated concussion. (A) Quantification of the percentage of propidium iodide cells that had colocalization signal with astrocyte-gfp fluorescence, one-way ANOVA, n.s. (B) Representative flow cytometry analysis (Right) for apoptotic versus necrotic astrocytes. Cells were gated on size, ACSA⁺, PI⁺, YO-PRO-1⁺. Apoptotic cells are identified as those that have positive signal for YO-PRO-1, quantification (Left). 2-way ANOVA, n.s. (C) Intravital imaging of Sham (Left) and CON 2X (Right) CX3CR1-gfp/wt mice injected IV with SR101 demonstrates sulforhodamine 101 stained astrocytes and gfp-microglia are distinct from one another in noninjured conditions but show colocalization of microglia signal (green, gfp) with astrocytes (red, SR101) in CON 2X. (Scale bar, 40 μm.) (D) Images (Left) and quantification (Right) of intravital imaging with PrismPlus mice. Microglia (green, gfp), phagocytosing astrocytes (gray, ALDH1L1, DsRedMax, pseudocolored), astrocytes that are not phagocytosed by microglia have been pseudocolored purple (ALDH1L1, DsRedMax). (Scale bar, 20 μm.) Enlarged view shows a single microglia phagocytosing an astrocyte with the microglia partially clipped away in the first panel to demonstrate that the astrocyte is within the microglia. (Scale bar, 5 μm.) One-way ANOVA, **P = 0.003. (E) Flow cytometry analysis for gfp-positive astrocytes in microglia following TBI. Cells were gated on size, viability, CD45^lo, CX3CR1⁺, CD11b⁺. Plots show quantification of percent of gfp-positive microglia, indicating phagocytosis of astrocytes. One-way ANOVA, **P = 0.002, ***P = 0.0002. (F) Flow cytometry analysis for ACSA-2 astrocytes within CD45⁺, CX3CR1⁺, CD11b⁺ microglia in C57 wild-type mice. One-way ANOVA **P = 0.001, ***P < 0.0001. (G) Flow cytometry quantification of ACSA-2 -positive microglia in TLR4^−/− mice, indicating microglial phagocytosis. One-way ANOVA, *P = 0.02. (H) Flow cytometry quantification of ACSA-2⁺ microglia in CD36^−/− mice. Each dot represents one mouse. Data represent three independent experiments (A, B, and D–H) with 3 to 6 mice per group or two independent experiments.

The fact that the majority of astrocytes did not take up propidium iodide suggests that this was not simple clearance of dead cells but perhaps additional phagocytic pathways were at play (21). Analysis of datasets of phagocytic inflammatory microglia suggested specific detection receptor and scavenger receptor pathways were enriched including toll-like receptor-4 (TLR4) and the CD36 scavenger receptor pathways, observations confirmed by other groups immediately after injury in more profound TBI models (22–24). In fact, CD36 also acts as a coreceptor to TLR4 to promote signaling (25). We tested the role of TLR4 and CD36 in our concussion model to elucidate the mechanism underlying microglial phagocytosis of astrocytes. We used astrocyte cell surface antigen-2 (ACSA-2) as a marker for astrocytes in TLR4^−/− and CD36^−/− mice. Flow cytometric analysis verified that wild-type C57BL/6 mice with ACSA-2 antibody showed a similar proportion of astrocyte phagocytosis to that of Aldh1l1^gfp mice (Fig. 4F). In TLR4^−/− mice, microglia were no longer able to increase astrocyte phagocytosis to the same extent after repetitive injury, although phagocytosis still occurred following a single injury (Fig. 4G). By contrast, phagocytosis of astrocytes was completely abrogated in CD36^−/− mice, both after a single and repeated concussion (Fig. 4H) suggesting that in concussion, CD36 is the dominant receptor for microglial clearance of astrocytes.

In addition to microglia, there is a population of Cx3cr1-positive perivascular macrophages located at the brain borders (26). To further explore the phenotype of the Cx3cr1⁺ cells that were localizing to the sites of injury, we caused concussion in binary transgenic mice, in which a combination of Sall1^ncre and CX3CR1^ccre transgenes specifically target microglia, while Lyve1^ncre and Cx3cr1^ccre coexpression marks vascular-associated macrophages (27). In steady state, the majority of cells interacting with blood vessels are perivascular macrophages despite the fact that they make up only 5% of the total brain macrophage pool (Fig. 5A) (27). The percentage of the two types of macrophages interacting with blood vessels did not change significantly after a single concussion, whereas in repeated concussion, there was a clear increase in the number of microglia that surrounded the microvasculature (Fig. 5 B and C). The number of Lyve1+ cells which already associate with the vasculature was not changed (Fig. 5 A and C) and therefore could not explain the large increase in macrophage numbers seen in the concussion area (Fig. 2B). In sum, these data suggest that following repetitive concussion, microglia specifically are rapidly recruited to the site of injury and have a role in sealing the BBB and in clearing astrocytes.

Fig. 5.

Perivascular macrophages and microglia have distinct responses following repetitive concussion. (A) Intravital 2P imaging of Lyve-1^ncre:CX3CR1^ccre:R26 tdTomato mice under CON1X and CON2X TBI demonstrating perivascular location of Lyve1-positive (tdTomato) cells. (Scale bar, 40 μm.) (B) Intravital 2P imaging of Sall1^ncre:CX3CR1^ccre:R26 tdTomato mice under CON1X and CON2X TBI demonstrating progressive localization of CX3CR1+ microglial cells (tdTomato) with repetitive injury. (C) Quantification of vascular associated cells in Cx3cr1ccre: Lyve1ncre and Cx3cr1ccre: Sall1ncre. Two-way ANOVA *P = 0.01, **P = 0.002. Each dot represents one mouse.

Discussion

Concussions are a significant public health problem as up to 15% of patients have persistent neurocognitive dysfunction (28). Furthermore, repetitive concussion can lead to progressive neurodegeneration in a clinical syndrome known as chronic traumatic encephalopathy (CTE) (29). Despite the considerable functional impact of concussions, the underlying molecular events remain unelucidated. In this study, we report that even a single mild blow causes permeability increases in the BBB within hours. However, while we anticipated that repeated concussion would further increase BBB dysfunction, we were surprised to find that the increased BBB permeability after a first hit had completely disappeared after a second hit. In addition, in our single concussion model astrocytes were now at a distance from the vasculature with detached end-feet, whereas repeated concussion resulted in complete loss of astrocytes due to microglial uptake of these altered but not dead cells. Clearly, while the single concussion model appeared to be more dysfunctional than the repeated concussion model based on permeability, there was a very significant change in the cellular architecture after two hits which may result in long-term altered biology. Preventing microglial recruitment to the concussed area resulted in increased BBB dysfunction in the single concussion model consistent with the microglia helping to reduce microvascular permeability. In the repeated concussion model, we showed that microglia physically surrounded the disrupted vasculature while other microglia removed detached astrocytes. The mere presence of the microglia was not sufficient to reduce the vascular leak. Indeed, when microglia entered the concussed area but were inhibited from wrapping around the vasculature for example with an ATP inhibitor, vascular permeability remained elevated. While we cannot exclude the possibility that microglia can release factors that might help close the BBB, their physical wrapping around the injured vessel was necessary.

Our data show that there may be at least two different functions performed by microglia suggesting the need for division of labor. The first function we identified was a limitation of BBB dysfunction and the second was to remove astrocytes that detached from the vasculature. Whether this was performed by two different subpopulations of microglia or one type of microglia that is able to adapt according to environmental changes (ie., increased permeability versus detached astrocytes) is unclear. In addition to microglia, the brain is also seeded with perivascular macrophages. Using a split-Cre system to distinguish the microglia from the perivascular macrophages, it appears that the majority of the macrophage recruitment to the BBB was from the microglial population whereas the perivascular macrophage numbers did not change. Although their participation in the sealing of the vasculature is likely due to their perivascular location, the microglia had to be recruited to completely seal the BBB. In fact, since these perivascular macrophages were previously shown to induce immune cell recruitment in addition to protecting the BBB (30, 31), the lack of recruitment of any blood-borne immune cells including neutrophils or monocytes in our model suggests limited activation of the perivascular macrophages.

In models of TBI ranging from moderate to severe, the data suggest that activated microglia secrete chemokines and cytokines that promote BBB disruption (32, 33). In addition, there is now a vast amount of literature suggesting that these cells are also key to the massive recruitment of neutrophils and monocytes, the latter turning into inflammatory macrophages and together causing significant inflammation and damage. In our concussion model, the immune response was more clandestine and reparative in nature with no recruitment of blood-borne immune cells. Moreover, the increase in permeability was likely a result of astrocytes in part detaching from the vascular wall (34) and not overt microglial injury. Despite cell death, increased vascular dysfunction, and large relocalization of microglia, this did not induce recruitment of peripheral immune cells as no neutrophils or monocytes were seen to infiltrate the brain parenchyma. In fact, in this model, the local resident immune cells entirely deal with cellular injury and death. This differs from not only more severe brain injury but also from subtle injuries in other organs where even the death of a few cells or multiple injuries leads to significant neutrophil and monocyte recruitment (18, 35). Clearly, in the brain, the microglia cloak minor injuries to prevent excessive inflammation and peripheral immune cell recruitment, induce repair independent of peripheral immunity, and thereby limit subsequent tissue damage.

Our concussion model looks specifically at the parenchymal compartment and intracerebral vessels, which are protected by astrocytic end-feet whereas meningeal vessels are not. We demonstrate that when the vasculature of the parenchymal compartment is injured, astrocytic end-feet are displaced from the vasculature and activated microglia work quickly to contain the damage and protect the BBB, a process that does not involve participation of peripheral immune cells. It is possible that with a single concussion, the astrocytes may recover contact with the vessel wall, but in multiple concussions, the astrocytes were phagocytosed. The consequences of these actions in the long term are yet to be determined. While our work implicates a key role for the scavenger receptor CD36 as well as contribution from the pattern recognition pathway TLR4, other molecules may also be involved including, for example, Mer tyrosine kinase (MerTK) a receptor tyrosine kinase that is highly expressed on microglial cells and is required for phagocytosis of apoptotic cells (36). In disease processes, MerTK has been demonstrated to be upregulated in phagocytic microglia following TBI (37). In vivo knockdown of Mer worsened functional outcomes in the acute phase of injury and increased cerebral edema, consistent with our findings that microglial activation was important in regulating BBB dysfunction. The role of this receptor versus CD36 and perhaps other molecules is worth exploring further. TBI is an evolving neurological event with acute and chronic consequences. We demonstrate that activated microglia work quickly to contain the acute injury and clear cellular debris, including dysfunctional astrocytes. Astrocytes play an important role in maintenance of the BBB, neurovascular coupling, and also protect neurons from apoptosis following brain injury (38, 39). In our concussion model, repeated injury results in progressive phagocytosis of astrocytes, which may ultimately contribute to the long-term cognitive dysfunction that is observed in patients with concussion. The potential for astrocyte regeneration in concussion is unclear particularly after multiple concussions, but by clearing astrocytes at the damaged vasculature, microglia may alter BBB dynamics, cerebrovascular autoregulation, and cellular signaling in the brain, features that are common to brain injury and neurodegenerative diseases (40, 41).

Materials and Methods

Mice.

C57BL/6, Aldh1L1-egfp-DTA, Cx3cr1^GFP/+, LysM ^gfp/gfp, Cx3cr1^gfp/gfp, PrismPlus, Tlr4^−/−, and CD36^−/− mice were obtained from Jackson Laboratory. P2RX7^−/− mice were a gift from Dr. Tanya Mayadas. Ms4a3^Cre:LSL-TdTom mice were provided by Dr. Florent Ginhoux, now publicly available. Cx₃ cr1^ccre:Sall1^ncre and Cx₃cr1^ccre:Sall1^ncre mice were provided by Dr. Steffen Jung. All animals were 6 to 8 wk of age. Animals were maintained in a specific pathogen-free facility at the University of Calgary Animal Resource Center. All experiments were approved by the University of Calgary Animal Care Committee and in compliance with guidelines established by the Canadian Council for Animal Care.

Mild TBI.

Mice are anesthetized with 200 mg/kg ketamine (Bayer Animal Health) and 10 mg/kg xylazine (Bimeda-MTC). Injury is induced by an electromagnetically controlled impact device (ImpactOneTM Stereotaxic Impactor, Leica Microsystems, St. Louis, MO). To induce a mild, closed-skull injury, mice are placed in a stereotaxic frame and an incision made in the scalp to expose the right parietal bone. A rubber-covered impactor tip is attached to the impactor. The impactor is rigidly mounted at 20° from the vertical plane and aimed at the right parietotemporal region, midway between the lambda and bregma. The tip is then impacted onto the skull 1 mm past the zero point at a velocity of 4.0 m/s, duration of 0.5 s, to cause a nonpenetrating blow. For a repetitive injury, the skin incision is sutured and mice are returned to their cages and recovered for 24 h at which point the incision is reopened and a second impact is delivered. Imaging is performed within 30 min following impact.

Brain Tissue Viability Staining.

Following a mild TBI while under anesthetic, mice were killed with the craniocervical dislocation technique. The brains of killed mice were removed, immediately placed in ice-cold PBS, and sectioned coronally into 2 mm slices by a brain matrix (Zivic Instruments, Pittsburgh, PA). Brain slices were incubated in 2% 2,3,5-triphenyltetrazolium chloride monohydrate (Sigma) at 37 °C for 30 min, followed by 4% paraformaldehyde fixation overnight. As a comparison to mTBI, a severe injury was also performed and tissue evaluated for viability. In a severe injury, mice were anesthetized and positioned under the impactor as above described. A craniotomy was performed at the right parietal bone. A bare metal impactor tip was attached to the impactor. The tip was impacted onto the brain 1 mm past the zero point at a velocity of 4.0 m/s, duration of 0.1 s, to cause an open-skull, penetrating blow.

Microscopy of Brain Slices.

Following mild or severe injury, Cx3cr1^GFP/+ mice were injected with 10 μL Ly6G-1A8 PE (BioLegend) and 10 μL CD31-AF647 (eBioscience, conjugated with Invitrogen kit) antibody. LysM^egfp/egfp mice did not receive additional antibody. Mice were then killed, perfused, and brains harvested and sliced as above described. The third slice from the frontal pole at the level of the third and lateral ventricle was then mounted onto a coverslip and placed under a microscope. Stitched imaging of the entire brain slice is performed using Volocity (Quorum).

Intravital Multiphoton Microscopy.

To prepare mice for imaging, a craniotomy is performed over the area of impact using a high-speed drill. The underlying brain is kept moist with saline and a coverslip placed on top. The tail vein is cannulated to administer additional anesthetic and fluorescent dyes or antibodies. Image acquisition is performed using an upright multiphoton microscope (Leica SP8, Leica Microsystems) equipped with a 25 × 0.95 NA water objective lens, InSight DeepSee pulsed infrared lasers with dual output (fixed 1,040 nm and tunable 680 to 1,300 nm; SpectraPhysics). Emitted fluorescence was detected by nondescanned HyD detectors with 650 to 700, 565 to 620, and 500 to 500 nm. Image stacks were acquired using a 1.0 μm z-step to a maximum depth of 300 μm. Some images were also acquired with an Olympus BX61W1 FV300 using a ×20/1.0 XLUMPlanFl water immersion objective equipped with a modified FV300 unit and Ti:sapphire laser. The laser emitted red and near-infrared light in the range of 680 to 1,020 nm, which was tuned to an excitation wavelength of 940 nm for all experiments except for those in which SR101 was used, in which it was tuned to 900 nm. A dichroic filter cube placed before the photomultiplier tube detectors allowed for simultaneous two-color imaging. The filter cube was assembled from a beamsplitter and two-photon detector emission filters (RXD2 [GFP] and RXD4 [RFP], Olympus). Olympus Fluoview 5.0 software was used to drive the microscope. For time-lapse imaging (xyt) of neutrophil trafficking, a confocal spinning disk microscope was used. Images were acquired with an upright microscope (BX51; Olympus) using 20X/0.75 NA XLUM Plan F1 objective (Olympus). The microscope was equipped with a confocal light path (WaveFx, Quorum) based on a modified Yokogawa CSU-10 head (Yokogawa Electric). A 512 × 512 pixel back-thinned electron-multiplying charge-coupled device camera (C9100-13, Hamamatsu) was used for fluorescence detection. Simultaneous behavior of multiple cell types in the brain was assessed using three (488, 561, and 635-nm) laser excitation wavelengths (Cobalt, Stockholm, Sweden) in rapid succession and visualized with the appropriate long-pass filters (Semrock, Rochester, NY).

Fluorescent Dyes.

To visualize brain vasculature, mice are injected i.v. prior to imaging with 300 μL of 1 mg/mL 70,000 MW dextran-rhodamine (Lifetech/Invitrogen). Fluorescent staining of astrocytes was performed by intravenous injection of sulforhodamine 101 (SR101) (Sigma). 500 μM solution of SR101 was injected at a dose of 20 mg/kg body weight 2 h prior to imaging. Cell death was visualized by incubating the exposed meninges and brain with propidium iodide (1.5 mM) in saline for 10 min, followed by a saline wash.

BBB Permeability Assay.

To assess BBB permeability, 70,000 MW dextran-rhodamine was injected i.v following brain injury. Mice were imaged within 1 h following injection. During image acquisition, 3D stacks of four separate fields of view over the area of injury were obtained. To analyze the degree of BBB permeability, individual blood vessels in the field of view are traced using ImageJ, and a 10 μm area around each vessel is selected as a region of interest. The average fluorescent intensity of this region is measured and subtracted from the background fluorescent intensity. The mean fluorescent intensity around three or more blood vessels is calculated and taken as the percent permeability for the field of view. For single-injury experiments using apyrase, immediately following injury, a craniotomy was performed and 200 μL apyrase (300 U/mL, Sigma) was transcranially applied and incubated for 1 h, followed by a saline wash. Imaging was performed immediately afterward. For repetitive injury, apyrase was transcranially applied at the time of the second injury.

Image Analysis.

Quantitative image analysis was performed using FIJI and Imaris (Bitplane, v9.5). To quantify cell death, propidium iodide–positive cells were identified using the Analyze Particles tool. The number of cells was divided by the area analyzed and represented as cells/mm². To measure astrocyte distance from blood vessels, individual astrocytes were counted and the distance from an astrocyte to the fluorescently labeled blood vessel was measured. The mean distance per astrocyte was measured for each acquisition image. To visually assess microglial activation, we used cell soma size as a marker of activation as previously described (42, 43). Microglial cell perimeters were traced in the 3D z-stack images to threshold and eliminate background fluorescence and the cell body area was measured. Microglia were considered activated if the soma were greater than 50 μm² (42). The number of microglia with soma greater than 50 μm² was divided by the total number of microglia in the field of view and multiplied by 100 to give a percentage of activated microglia. To characterize individual microglia morphology, the Surfaces module in Imaris v 9.5 for semiautomated rendering of individual microglia. Accuracy of the software reconstruction was independently verified by two different investigators for each acquisition image. Similarly, surface reconstruction of cranial vessels and astrocytes was also performed using the Imaris Surface module. To characterize the degree to which microglia can stretch its processes, we used the Imaris Bounding Box OO Length C function, where individual reconstructed microglia are enclosed in a box with axes oriented to the principal axes of the object. BoundingBoxOO Length C measures the length of the longest principal axis inside of the object. To quantify the number of microglia that are intimately associated with the vasculature, Imaris Shortest Distance to Surfaces statistical function was used, with Surfaces = Vasculature. All microglia with a surface distance to vasculature equal to zero were considered to be a vasculature-associated microglia. The number of vasculature-associated microglia was divided by the total number of microglia in the field of view to determine the percentage of vasculature-associated microglia. To assess the percentage of astrocytes that were phagocytosed by microglia on imaging, Imaris Shortest Distance to Surfaces statistical function was used, with Surfaces = Microglia. All astrocytes with a surface distance to microglia less than zero were considered to be engulfed/within the microglia. This number was divided by the total number of astrocytes to obtain the percentage of astrocytes phagocytosed by microglia. For all imaging analysis, three to four fields of view were acquired per mouse. Each field of view was analyzed, and the mean of all fields of view was used as the average number per mouse.

Cell Isolation and Flow Cytometry.

For single-injury experiments, a head injury was administered as previously described and brains were harvested 24 h following injury. For repetitive injury experiments, brains were harvested 24 h following the second injury. To harvest mice brains, mice were anesthetized with CO₂ and perfused with ice-cold HBSS. Brains were extracted and single-cell suspension generated by both mechanical disruption and enzymatic digestion as previously described (44, 45). Myelin and other cellular debris were removed using myelin removal magnetic beads (Miltenyi Biotec). Cells were pelleted and stained with CX3CR1 (BV421or PE, BioLegend), CD45 (BV510, BioLegend), CD11b (PerCPCy5.5, Pharmingen), ACSA-2 (APC, Miltenyi) at 1:100 concentration. In instances where flow-cytometry was performed using AldhL1^GFP/+ mice the gfp fluorescence was used to identify astrocytes. Otherwise, astrocytes were identified as ACSA-2⁺ cells. Microglia were identified as CD45^loCX3CR1⁺CD11b⁺ cells. Dead cells were excluded using ghost dye^TM red 710 (TONBO Biosciences). Samples were run using the BD FACS Canto flow cytometer and analyzed using FlowJo software (Tree Star).

Statistics.

Data are expressed as mean ± SEM and were analyzed using Student’s t test or one-way ANOVA with Sidak post hoc multiple comparisons test, as indicated in figure legends. All statistical tests were two sided, and a P-value < 0.05 was considered significant. Data were analyzed using GraphPad Prism v 8.4.3.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Intravital spinning disk microscopy time-lapse imaging in non-injured control C57BL/6 mouse injected with q-tracker 655 vascular labeland Ly6G-1A8 (neutrophils, PE). There is no evidence of neutrophil recruitment. Scale bar 30 um, time-stamp in minutes.

Movie S2.

Intravital spinning disk microscopy time-lapse imaging in C57BL/6 mouse injected with CD31 (endothelial cells, AF647) and Ly6G-1A8 (neutrophils, PE). Neutrophils traffick through the vasculature but are not recruited to the site of injury and do not extravasate. Scale bar 30 um, time-stamp in minutes.

Movie S3.

Intravital spinning disk microscopy time-lapse imaging in CX3CR1-gfp/wt mouse injected with CD49b (platelets, PE). Scale bar 40 um, time-stamp in minutes.

Movie S4.

Intravital spinning disk microscopy time-lapse imaging in CX3CR1-gfp/wt-CCR2rfp/rfp mouse with mTBI, injected with Q-tracker 655 vascular label. Two monocytes travel through the vasculature but do not extravasate. Peripheral immune cells are not highly recruited following closed-skull injury. Scale bar 40 um, time-stamp in minutes.

Movie S5.

Intravital multiphoton microscopy time-lapse imaging in CX3CR1-gfp/wt mouse under sham and TBI conditions, injected with 70,000 MW dextran rhodamine. Microglia have small cell bodies and fine thin dendrites and processes that constantly survey the brain environment under Sham conditions. Following injury, microglia increase their soma size and dendrites become thickened. Scale bar as indicated, time-stamp in hh:mm:ss.

Movie S6.

Intravital multiphoton microscopy time-lapse imaging in CX3CR1-gfp/wt mouse with repeated TBI, injected with 70,000 MW dextran rhodamine. Microglia extend their processes and wrap around the damaged BBB following rmTBI. Scale bar 10 um, timestamp in hh:mm:ss.

Movie S7.

Intravital multiphoton microscopy time-lapse imaging in PRISMPlus mouse with repeated TBI. Microglia (green, gfp) use their processes to probe and phagocytose fragments of astrocytes (red, DSRedMax). Scale bar 20 um, time stamp in hh:mm:ss.

Movie S8.

Intravital multiphoton microscopy time-lapse imaging in PRISMPlus mouse with repeated TBI. Microglia (green, gfp) use their processes to probe and phagocytose fragments of astrocytes (red, DSRedMax) that are detached from blood vessels (blue, q-tracker 655 vascular label). Scale bar 10 um, time stamp in hh:mm:ss.

Movie S9.

Animated 3D reconstruction of microglia phagocytosis in PrismPlus mice. Astrocytes that have not been phagocytosed by microglia (green) are coloured purple, astrocytes that are completely within microglia are coloured gray. Animation shows clipping planes where astrocyte fragments are shown to be completely phagocytosed by microglia.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.