Blast injury leads to chronic shifts in protein expression of distinct astrocyte populations

Abstract

Blast traumatic brain injury results in chronic pathology, especially for those receiving repetitive injuries. To evaluate cellular changes induced by these pressure waves, we studied post-mortem prefrontal cortex of military personnel with a history of multiple blast exposures and military controls with no battlefield experience. Chronic increases in IBA1 (microglia) and GFAP (glial fibrillary acidic protein; astrocyte) immunoreactivity occurred in injured brains and also confirmed that GFAP-expressing astrocytes altered predominantly at interface regions of the brain: around blood vessels, the grey-white matter interface and in layer 1, consistent with the pattern of damage seen with blast exposure. We focused on pathologic implications of the astrocyte derived proteins GFAP, aquaporin-4, and connexin-43. Astrocyte morphology in injured samples altered significantly, revealing a disintegrated, beaded shape, with a loss of fine processes. We also observed a shift in astrocyte immunoreactivity, where control brains showed two dominant populations, labelling as either GFAP+ or aquaporin-4+ only, with a smaller portion of co-labelled cells. Samples from injured brains revealed the emergence of a third dominant population of cells with abnormal morphology co-labelled with GFAP and aquaporin-4; significant increases in astrocytes with abnormal morphology also occurred, including those both GFAP+ and aquaporin-4+. Connexin-43, which helps maintain neural homeostasis, significantly co-labelled with aquaporin-4, and not GFAP, in both control and injured brains, suggesting the aquaporin-4 subtype to be homeostatic. Interlaminar astrocytes consistently showed abnormal morphology in injured brains, featuring extensive GFAP+ beaded processes. The GFAP+ beaded processes showed additional characteristics in the injured brains, being surrounded in a ring-like fashion by aquaporin-4 immunoreactivity as well as co-labelling with phosphorylated connexin-43, indicating an inflammatory phenotype. To investigate secondary pathology that might relate to immune dysfunction, immunoreactivity with IgG revealed the presence of autoantibody in injury samples, which primarily labelled neurons in layer 2–3 that also co-immunoreacted with complement C3. Half of the interlaminar astrocytes in the injured brains also showed immunoreactivity with C3. We conclude that control human cerebral cortex contains at least two distinct populations that express either GFAP or aquaporin-4, but not both. After military related blast traumatic brain injury, a third population of astrocytes emerges expressing both GFAP and aquaporin-4. Connexin-43 continues to be co-expressed with aquaporin-4, but shifts toward an abnormal morphology. We also find overall chronic alterations in expression of astrocytic proteins that coincide with induction of autoantibodies directed towards neurons and recruit complement.

Unique astrocytes occur in blast exposed military personnel. The authors report that neocortical astrocytes express either glial fibrillary acidic acid (GFAP) or Aquaporin4, but after blast show both markers. They additionally demonstrate that Connexin43 (a homeostatic gap junction protein) is more likely co-expresses with the Aquaporin 4+ astrocytes, indicating greater stability than the GFAP+ cells.

Graphical Abstract

Graphical Abstract.

Introduction

Traumatic brain injuries (TBIs) are among the most common physical damages incurred by military personnel.¹ TBIs are varied, ranging from mild to moderate to severe, but share common features such as acute and chronic inflammation, cellular oedema, glial cell activation, cell damage, impaired neural transmission, blood brain barrier disruption, meningeal injury, and more.² The damage associates with short term symptoms such as headache and sleep disruption, or long-term effects including fatigue, chronic pain, depression, mood changes, cognition, and other psychological consequences.^3-5 While various therapies and treatments attempt to mitigate symptoms,⁶ very few therapeutic approaches aid individual suffering from continued complications, perhaps due to their dynamic nature and diffuse pathology.⁷

Because of the enduring underlying behavioural problems that accompany brain injury, we evaluated blast exposure in our study of human pathology. We focused on post-mortem samples from the prefrontal cortex (PFC) of military personnel exposed to blast TBIs. Blast injury is a common occurrence in most of the recent conflicts involving the US military.⁸ We assessed levels of inflammation via changes in microglia using ionized calcium binding adaptor molecule 1 (IBA1) and astrocytes using glial fibrillary acidic protein (GFAP). Since astrocytes contribute significantly to maintaining homeostasis, we concentrated on these glial cells.^9,10 We also studied aquaporin-4 (AQP4) and connexin-43 (Conn43) as both proteins are integral to astrocytic function.^11-13 AQP4 is a water channel protein considered vital to the integrity of the blood brain barrier (BBB) and the glymphatic system.¹⁴ Physiologically, AQP4 is polarized (localized) mostly to astrocyte endfeet, where it integrates with blood vessels, aiding in the removal of interstitially derived waste products, a function attributed to the glymphatic system.^15,16 Conn43 is a gap junction protein expressed by astrocytes that contributes to the formation of the glial syncytium—a connectome of astrocytes across broad anatomical regions—allowing communication to rapidly flow from one astrocyte to the next.¹⁷ Conn43 also aids in spatial buffering, a process that allows astrocytes to act as sponges, soaking up extracellular ions when excessive concentrations arise as a result of hyperexcitability, thus protecting the health of neighbouring neurons.^18,19

Due to the nature of chronic and prolonged symptoms of military personnel who experience TBIs, we expect chronic and prolonged astrogliosis in post-mortem samples, as previously reported.^20-22 After blast injury, astrocytic changes concentrate in a distinct pattern involving areas of interface, including the subpial layer,²³ the glial limitans surrounding vessels,²⁴ and the white-grey matter border.²² We also find that interlaminar astrocytes show abnormal morphology and reactivity in these samples. Interlaminar astrocytes are a sub-type likely originating from radial glial cells²⁵; They reside in layer 1, making connections with the pia via their soma or processes, while also extending deep into the cortical layers.^25,26 Little is known of their function, however they may participate in BBB functions and tripartite synaptic communication, as they make connections with neurons and capillaries.²⁷ Our study found extensive disruption in interlaminar astrocytes, indicating they may be specifically susceptible to blast injury.

Autoimmune related threats may arise from this type of injury and are also of particular concern.²⁸ A significant correlation exists between individuals who experience a TBI and the incidence of multiple sclerosis, suggesting that TBI may trigger an autoimmune response.²⁹ Furthermore, BBB integrity may be impacted, permitting peripheral toxins and non-central nervous system (CNS) resident immune cells to flow into the parenchyma, increasing the inflammatory environment and contributing to oedema.³⁰ Optimal conditions that contribute to the onset of autoimmune disorders include physical damage to cells, releasing components such as DAMPs (Damage Associated Molecular Patterns),³¹ an inflammatory environment, and the invasion of peripheral immune cells into the parenchyma with a disrupted BBB.^28,32,33 All of these conditions have an increased risk after military TBIs.^32,34 Of note, the post-mortem samples analyzed here consist of individuals with repeated TBI exposure. We therefore searched for the presence of autoantibody in the brain tissue and its potential colocalization with astrocytes. We also explored further pathology by evaluating for the presence of complement in the injured and control brains.

Methods

Histology and imaging

Tissue samples were acquired from the Department of Defense/Uniformed Services Brain Repository and immediately chilled at 4°C in a 4% paraformaldehyde solution. See Supplementary Table 1 to view characteristics of the subjects. To process the tissue, we positioned the block in a plastic mold and surrounded it with prepared agarose. We cut the tissue block using a vibratome (Leica VT1000; Leica) at 50 µm thickness, and stored the sections in a solution of 0.04% sodium azide in phosphate buffered saline (PBS), at 4°C. For immunohistochemistry, we used antigen retrieval by floating sections in 1× citrate epitope retrieval buffer (Thermo Scientific, Waltham, MA) in an 80°C water bath for 20 min. This was followed by a series of washes with 1×, 7.4 pH PBS. Next, we performed a 3-h blocking step with 3% Normal Goat Serum (NGS) in PBS-Triton. The primary antibodies used for this study included: chicken anti-glial fibrillary acidic protein (GFAP) (1:500; Abcam Cat# ab4674), rabbit anti-aquaporin-4 (AQP4) (1:1000; Abcam Cat# ab128906), rabbit anti-ionized calcium binding adaptor molecule 1 (IBA1) (1:1000; Wako Cat# 019-19740), mouse connexin-43 (Conn43) (1:500; ThermoFisher Cat# CX-1B1), rabbit phosphorylated-connexin-43 (p-conn43) (1:500; ThermoFisher Cat# PA5-104820), and mouse immunoglobulin-G sub-class 1 (IgG1) (1:500; Cell Signalling Cat# RWP49). Incubation with primary antibodies lasted overnight at 4°C on a slow-rocker. A series of washes followed with 1× PBS; next, secondary antibodies were diluted to 1:500. The secondary antibodies included Alexa Fluor 555- or 488-conjugates; nuclei were stained with 40,6-diamidino-2-phenylimdole (DAPI) (Invitrogen, Waltham, MA: D21490, 1:2000). Incubation lasted for 90-minutes at room temperature on a shaker, followed by a series of washes with 1× PBS. In some cases, use of 1× autofluorescent quencher in 70% ethanol TrueBlack® Lipofuscin (Cat#23007 Biotum, Inc. Fremont, CA) was followed by secondary antibody incubations and washes. Finally, we mounted the tissue on Azer Scientific Unifrost slides. For long-term storage, slides are kept at 4°C.

To image sections we used a Zeiss Axio Observer.Z1 microscope with an Apotome using Zen 2012 software [Blue edition version 1.1.2.0 (Carl Zeiss Microscopy)] to capture 10× and 20× optical image fluorescent stacks. When quantifying the fluorescence intensity and percent area, we first optimized the exposure limits for each wavelength channel (e.g. Red: 400 ms, Green: 60 ms, and Blue: 15 ms) to avoid fully saturated pixels that might skew results. To quantify intensity fluorescence and percent area fluorescence we adapted procedures in FIJI (https://imagej.net/software/fiji/) and from https://www.youtube.com/watch?v=nLfVSWcxMKw&list=TLPQMDUwNjIwMjAhHiT0TuG0XA&index=4.

Morphology with Fiji

To determine the changes in astrocyte morphology in control versus injury samples in the PFC, we captured 20× magnifications with the Apotome and compressed Z-stacks of fluorescently labelled GFAP in the regions of interest (around grey matter blood vessels). Brightness was adjusted by applying the auto option. The backgrounds were subtracted on a rolling ball radius of 50.0 pixels. To help remove excess background fluorescence we applied the smooth application, made the photo binary, and filled holes. The remaining particles were analyzed with a threshold size of 0.001-infinity, allowing removal of background immunofluorescence and the results summarized (see Supplementary Fig. 1). We then captured shape descriptor measurements that included percent area of particles, total area occupied, and count. This process was also described previously in Ryu et al.³⁵ There were 4 human subjects per group with at least 2–3 sections per person, 2–4 images per section, and 1–3 cells analyzed per image for a total of 35 control cells and 31 injury cells (see Supplementary Tables 1 and 2). Results were analyzed using unpaired T-tests.

Sub-population and statistical analysis

To determine sub-populations of astrocytes we double-labelled tissue sections with GFAP (green) + AQP4 (red), GFAP (green) + Conn43 (red/mauve), or AQP4 (green) + Conn43 (red/mauve). This allowed us to determine the astrocytes that had a more reactive profile or a homeostatic profile. The images for this analysis were taken from the PFC at 20× magnification using the Apotome and z-stacks, targeting astrocytes at blood vessels. We used 4 control subjects and 4 blast injured subjects; control and injured samples included 3 sections each with 2–3 images per section. Controls had 7–16 cells per image for a total of 286 cells, while the injury samples had 3 sections each with 2–3 images per section, 2–28 cells per image, for a total of 216 cells. In the case of Conn-43 (red/mauve) + AQP4 (green), we used the same number of samples for control and injured subjects. Control samples had 2–3 sections each with 2–3 images per section and 1–17 cells per image for a total of 178 cells. Injury samples had 3 sections each with 2 images per section and 1–16 cells per image for a total of 111 cells. For labelling with Conn-43 (red/mauve) + GFAP (green), we included the numbers of subjects. Control samples had 2–3 sections each with 2–3 images per section and 2–18 cells per image for a total of 215 cells. Injury samples had 2–3 sections per sample with 2–3 images per section and 2–21 cells per image for a total of 194 cells. See Supplementary Table 2.

As considerable astrocytic morphological changes occurred when comparing control and injury samples, we took shape into account when quantifying sub-populations. If the astrocytes retained fine branches, they were counted as normal, but if the astrocytes appeared fragmented or beaded, they were described as abnormal. The individual P values comparing the proportion of cells of each type between the control and injured groups are based on chi-square tests with a Bonferroni adjustment for multiple comparisons.

The initial introduction to the all of the samples was blinded by those doing the analysis and imaging. After successful identification of the tissue belonging to either the injured or control group, the subjects were identified.

Results

Microglia increase their intensity in the prefrontal cortex after injury

We used IBA1 to label microglia and measured the percent area immunoreactive and the intensity levels of the fluorescent marker (see Methods). Figure 1A and C shows representative images of the measured areas; each sample included an area of about 400× 400 µm taken through layers 2–5. The morphology of the microglia after injury appeared distinctly altered. The control cells showed fine, distinct processes (Fig. 1A and B), while the microglia in the injured brains presented many fewer fine processes and displayed a fractured and broken appearance (Fig. 1C and D). Quantification of the area labelled and intensity indicates a significant increase in both measurements for the injured samples (Fig. 1E and F). Significance was determined using an unpaired T-test.

Figure 1.

Immunoreactivity for microglia in the PFC of control and injured brains. Images show immunoreactivity for microglia (IBA1) in coronal sections of the post-mortem human brain. Control samples (A) are shown at higher power (B), as are the injured samples (C), and magnified in (D). An unpaired T-test reveals that the injured samples are significantly increased in (E) intensity (*P = 0.0047) and percent area labelled (unpaired T-test, *P = 0.0053) (F) of IBA1 fluorescence. Four subjects were included in the analysis with 3 tissue sections taken from each subject. Each tissue section included 2 or 3 images. The mean value of each subject was used for the T-test analysis and each point on the graphs represents a single person. Scale bars are the same for A and C and B and D.

Astrocyte Reactivity

Through layers of the cerebral cortex and interlaminar astrocytes

Figure 2A and B shows examples of GFAP reactivity in interlaminar astrocytes of the PFC. These astrocytes in control brains showed smooth processes that extended from layer 1 into deeper layers (Fig. 2A). Those in the injured tissue displayed beaded and often discontinuous processes (Fig. 2B). The percent area labelled and the intensity of immunoreactivity in a region of astrocyte reactivity through the neocortical layers revealed both measures significantly increased in the injured brains (Fig. 2E and F). We determined significance using unpaired T-tests.

Figure 2.

Astrocyte GFAP immunofluorescence in the PFC of control and injured brains. Images of GFAP+ (glial fibrillary acidic protein) astrocytes in coronal sections of the post-mortem human brain: Control samples of interlaminar astrocytes (A) show smooth processes that extend into deeper layers, injury samples are seen in (B). Representative images of individual astrocytes at blood vessels are shown in control (C) and injured (D). Quantifications of PFC samples from Layer-1 through Layer-6 of the neocortex show significant increases comparing control to injury intensity (*P = 0.0158) (E) and percent area labelled (*P = 0.017) (F) of GFAP fluorescence using unpaired T-tests. Four subjects were included in the intensity and percent area analysis with 3 tissue sections taken from each subject. Each tissue section included 2 or 3 images. The mean of each subject was used for the T-test analysis and each point on the graphs represents the mean value of one subject. Randomly selected astrocytes were subjected to ‘analyze particles’ using Fiji to display the impact of injury cell shape. A particle was defined as any connected structure. Injured astrocytes are more discontinuous than controls, which retain a connected soma structure. A sample control cell mask processed in Fiji is seen in (G), while an injury sample cell mask is seen in (H). An unpaired T-test reveals that injury samples are significantly increased in particle count (J; P = 0.0004) and significantly decreased in (I; P = 0.0262) average particle size when compared with controls. Four subjects were included in each group, with 8–9 randomly selected astrocytes per subject. Each point on the graphs represents the particle size or count for a given tissue section. Histogram lines represent the sample means and lines the standard error of the mean (SEM).

Astrocytes surrounding blood vessels show abnormal morphology

We also sampled astrocytes whose soma was within 200 µm from histologically identifiable blood vessels in the grey matter; the blood vessels were identified by their distinct morphology. The vessel structure represents another example of an interface between tissue types. In control tissue, the astrocytes near blood vessels show normal morphology with numerous fine processes (Fig. 2C). The GFAP reactive astrocytes in injured samples display abnormal morphology surrounding the vasculature with beaded and interrupted processes extending from the cell soma (Fig. 2D). We quantified the morphology of these astrocytes using a particle analysis in Fiji, which counted the number of discretely labelled and discontinuous components in individual cells. Figure 2G provides an example of a thresholded image of an astrocyte from control tissue revealing relatively continuous processes, whereas Fig. 2H provides an image from an injured brain with many discontinuous elements. Figure 2I and J reports quantification of the control and injured morphology using particle analysis. The particle count is significantly increased in injured conditions compared with control tissue samples, whereas the average particle size significantly increases in the controls (Fig. 2I and J). Our finding supports the observation that astrocytes in the injured brains show more fractured and disrupted distributions compared with the relatively smooth patterns seen in control astrocytes. The astrocytes with fractured morphology were considered to be ‘abnormal’ whereas those with continuous, smooth processes were counted as having ‘normal’ morphology. The significance was determined by calculating an unpaired T-test (P < 0.0001 for both measurements).

Aquaporin-4 immunoreactivity labels a large population of astrocytes

Astrocytic endfeet often ensheath vessels, creating a layer of the BBB known as the glial limitans; these endfeet express high levels of a water channel protein called AQP4.¹⁶ Fig. 3A and B provides representative images of tissue immunoreacted against AQP4 in the PFC. Quantification of the intensity and percent area fluorescence demonstrate no significant changes with either measurement, determined with an unpaired T-test (Fig. 3E and F). In addition, AQP4 labels a large population of single astrocytes, as observed in Figs 3C and D and 4A. Many of these surrounded blood vessels, but the soma of AQP4+ astrocytes not directly adjacent to vessels were also observed. Figure 3C and D shows individual astrocytes taken from layer 3 in control and injured tissue. Although the astrocytic processes revealed with AQP4 are not as smooth and unbroken as seen with GFAP, those in control brains are more continuous than in the injured brains. The AQP4 processes in the immunoreactive pattern of the injured tissues appear as small open circles in this (coronal) plane of section (Fig. 3D). Figure 4A portrays single AQP4+ astrocytes in layer 3 surrounded by AQP4 and GFAP immunoreactivity in the interlaminar astrocytes of an injured brain. Of note, the AQP4 label in the interlaminar astrocytes is also fractured, appears circular in this plane of section and, in many instances, surrounds the GFAP positive immunoreactivity (seen in Fig. 4C and E). Figure 4B and D provide an image taken from a control brain, which shows relatively smooth processes extending from interlaminar astrocytes. In the control brain, AQP4 immunoreactivity surrounds blood vessels but does not interact with the interlaminar astrocytes. We also tested the possibility that that the beaded processes might represent colocalization with synapses. Although we believe this was unlikely, since the GFAP+ beads are larger than synapses and GFAP does not typically colocalize with synaptic markers, we evaluated potential colocalization of GFAP and synaptophysin and did not find consistent overlap between the 2 markers (Supplementary Fig. 2).

Figure 3.

AQP-4 immunofluorescence in the PFC of control and injured groups. Images captured of aquaporin-4 water channel protein—(AQP-4) in coronal sections of post-mortem human brain samples. A control image is seen in (A), and an injury in (B). A higher power image of individual astrocytes labelled with AQP4 can be seen for the control sample (C) and an injured sample (D). An unpaired T-test reveals no significant changes in (E) intensity and percent area (F) of AQP-4 fluorescence between control and injury groups, although there is a trend toward increases after injury. Four subjects were included in the analysis with 3 tissue sections taken from each subject. Each tissue section included 2 or 3 images. The mean of each subject was used for the T-test analysis and each point on the graphs represents the mean value of one subject. Histogram lines represent the sample means and lines the standard error of the mean (SEM). PFC, Prefrontal Cortex.

Figure 4.

Examples of aquaporin 4 (AQP-4) and GFAP immunoreactivity in neocortical and interlaminar astrocytes. (A) Shows individual AQP4 immunoreacted astrocytes (red) surrounded by interlaminar astrocytes labelled with GFAP and AQP4. (B–E) shows higher power views of interlaminar astrocytes for control (B and D) and injured (C and E) brains. For the injured brains, the AQP4 surrounds the beaded GFAP (green) immunoreactivity in this plane of section. (B, D) Shows GFAP (green) immunofluorescence in control interlaminar astrocytes, which do not display interrupted and beaded morphology. Merged areas can be seen as yellow.

Astrocyte sub-populations

GFAP and AQP4 alter colocalization after injury

Interestingly, during our analysis of GFAP and AQP4 immunofluorescence, most astrocytes labelled as GFAP+ or AQP4+ only, and to a lesser degree, co-labelled with both markers. Since the expression of AQP4 is previously reported to localize to astrocytic endfeet, it was surprising to find a large number of astrocytes labelling entirely with AQP4, covering the endfeet, processes, and soma of the cells, with little to no GFAP immunoreactivity in the same cell. To evaluate these distinct sub-populations, we quantified the patterns of expression for GFAP and AQP4 in astrocytes residing at or within 200 µm from the lumen of a blood vessel within the neocortex. The AQP4 immunoreactivity of astrocytes occurred throughout the neocortex, but for quantification, we focused on those in the close vicinity of blood vessels. Figure 5 provides representative images of astrocytes labelled with AQP4+ only (red), GFAP+ only (green), and co-labelled (yellow) astrocytes. Examples of astrocytes with normal morphology, abnormal morphology, or co-labelled are shown at increased magnification (Fig. 5A–C). Figure 5D and E shows lower power views of AQP4+ and GFAP+ astrocytes that surround blood vessels. Normal morphology and reactivity are seen in Fig. 5D, while abnormal morphology and increased reactivity can be seen in the image from an injured brain (Fig. 5E). We indicate cells considered to be single or double labelled with differently coloured arrows in Fig. 5D and E. Counts of astrocyte sub-populations along with their morphology were statistically assessed using chi-square tests with a Bonferroni adjustment for multiple comparisons to reveal significant shifts in sup-populations comparing control to injury (Fig. 5F and G). The largest population of astrocytes in controls was from the ‘AQP4+ only—Normal’ (normal morphology) group, with the second most prominent group being ‘GFAP+ only—Normal’. Making up less than a quarter of the quantified cells were the AQP4+ and GFAP+ co-labelled astrocytes with normal morphology. Very few cells showed an abnormal morphology in the control tissue (Fig. 5F). After blast injury, samples revealed a sub-population shift, where the most numerous sub-population was ‘GFAP+ only—Abnormal’ (with abnormal morphology), ‘GFAP+/AQP4+ co-labelled—Abnormal’ cells also showed a large increase. A significant increase in ‘AQP4+ only—Abnormal’ cells also occurred in the injury group (Fig. 5G). Note that all groups showed significant differences in expression patterns and morphology in the blast injured samples compared with control samples.

Figure 5.

Distinct subpopulations of GFAP+ and AQP4+ astrocytes experience a shift after injury. The dominant population of astrocytes were AQP4+ only or GFAP+ only, with fewer astrocytes co-labelling with both markers. We aimed to define these populations and observe how they change after injury. (A–C) represents examples of astrocyte morphologies as either Normal (A), abnormal (B) or colabelled (C). (D, E) show representative images of astrocytes double labelled with GFAP and AQP4 before and after injury and closely adjacent to blood vessels. Also indicated are astrocytes classified as colabelled (as shown in C) and designated with yellow arrows, while examples of astrocytes classified as single labelled are indicated with white arrows (D, E). The quantification shown in (F, G) indicates that the proportion of astrocytes shifts from being dominantly AQP4+ with normal morphology to predominantly GFAP+ with abnormal morphology. The individual P values comparing the proportion of cells of each type between the control and injured groups are based on chi-square tests with a Bonferroni adjustment for multiple comparisons. Each subpopulation experienced a significant shift after injury. The sample numbers for this analysis are shown Supplementary Table 2. AQP4 abnormal **P = 0.0016, AQP4 normal ***P < 0.0000, GFAP abnormal ***P < 0.0000, GFAP normal ***P < 0.0000, Co-label abnormal P < 0.0000, Co-label normal *P = 0.0063.

GFAP/AQP4 co-labelling with connexin-43 alters after injury

Since astrocytes react to injury or disease, we sought to understand if the sub-populations inclined toward a healthy homeostatic profile, or a non-homeostatic profile. While the increased expression of GFAP can be viewed as a mechanism allowing astrocytes to respond quickly to pathologic conditions in an effort to allow for optimal healing processes, the chronic upregulation of GFAP also associates with negative consequences.³⁶ The injury samples used here, which show increased GFAP reactivity compared with controls, are years removed from their repeated exposures to military blast TBIs, indicating chronically reactive astrocytic changes. We aimed to understand if the AQP4+ and/or GFAP+ sub-populations were adaptive in nature, supporting regulation of CNS homeostasis, or non-homeostatic and potentially contributing to an ongoing inflammatory environment. To evaluate the functional profiles of these astrocytes, we studied the immunoreactivity of Conn43, a gap junction protein involved in the regulation of CNS homeostasis via spatial buffering and the creation of glial syncytia. We tried to determine if Conn43 co-labels with either GFAP+, AQP4+ astrocytes or both, assisting our understanding in the role of these sub-populations. We focused on Conn43, as a potential indicator of an adaptive profile (positive for Conn43), or maladaptive profile (negative for Conn43).

Figure 6A–C and E–G provide representative images of human tissue from control and blast injured groups co-stained with antibodies against AQP4 and Conn43. We counted the number of astrocytes labelled with Conn43, AQP4 or both, and included whether those cells had normal or abnormal morphologies. Over 75% of cells in control sections were co-labelled with both Conn43 (mauve) and AQP4 (green) and normal in appearance (note, AQP4 was previously labelled in red). In the injured brains, the most dominant population remains co-immunoreactive for Conn43 and AQP4 with over 50% retaining a normal morphology, although significantly diminished from control (Fig. 6D and H). The second largest population of astrocytes were also co-immunoreactive for Conn43 and AQP4, but with an abnormal morphology (Fig. 6H). The individual P values comparing the proportion of cells of each type between the control and injured groups are based on chi-square tests with a Bonferroni adjustment for multiple comparisons. Significant changes in the populations of cell types were seen after injury with increases in the AQP4+ abnormal morphology, and Conn43+/AQP4+ co-labelled of abnormal morphology, and decreases in co-labelled astrocytes with normal morphology (Fig. 6D and H).

Figure 6.

AQP4+ or GFAP+ subpopulations of astrocytes differentially co-label with connexin-43. (A–C and E–G) Representative cells in control and injured PFC labelled with markers against AQP4 (green) and connexin-43 (Conn43—mauve. When comparing controls (A–C) to injury (E–G), a noticeable change in astrocyte morphology appeared where processes were slightly more beaded and/or discontinuous. The individual P values comparing the proportion of cells of each type between the control and injured groups are based on chi-square tests with a Bonferroni adjustment for multiple comparisons (D, H). The number of samples included in this analysis can be seen in Supplementary Table 2. A significant increase occurred in abnormal AQP4+ cells after injury ***P < 0.0000 and in cells co-labelled with abnormal morphology (***P < 0.0000) and a significant decrease appeared in co-labelled cells with normal morphology (***P < 0.0000). (I–K and M–O) Antibody markers against GFAP (green) and Connexin-43 (Conn43—mauve) in the PFC of control and injured brain sections. (L, P) Quantification of the different populations. A significant decrease in Conn43 normal cells (**P = 0.0011), GFAP normal cells (P < 0.0000—not shown because there are so few), and co-labelled normal cells (*P = 0.0174) occurred after injury when compared with controls; a significant increase in GFAP abnormal cells and co-labelled abnormal cells appeared after injury (++P = 0.0158). (L, P) The P values comparing the proportion of cells of each type between the control and injured groups are based on chi-square tests with a Bonferroni adjustment for multiple comparisons. The number of samples used in this analysis can be seen in Supplementary Table 2. (I) We compared cells that labelled with antibodies against Conn43 and co-labelled with either AQP4 or GFAP. The graph shown here indicates that AQP4+ cells are significantly (****P < 0.0001) more likely to co-label with Conn43 compared with the GFAP+ cells, determined using a chi-square contingency.

Next, we assessed the percent of GFAP+ cells co-labelled with Conn43 (Fig. 6I–P). Representative images display astrocytes co-labelled with GFAP and Conn43 in the PFC of control and injury groups (Fig. 6I–K and M–O) taken from layers 2–3 in the PFC. The images in Fig. 6 reveal that GFAP+ cells (green) and Conn43+ cells (red) remain largely distinct and separate. Quantification shows that the most dominant population of astrocytes in the control brains are those expressing Conn43 with normal morphology and the second most dominant population are those expressing GFAP with normal morphology (Fig. 6L). For the injured group, the GFAP+ only astrocytes with abnormal morphology are the most dominant, while Conn43 cells with normal morphology make up less than half of the total population (Fig. 6P). The total number of astrocytes co-labelled with GFAP and AQP4 are similar in the in the control (9% of the total population) and injured groups (6% of the total population). Overall, we observed that AQP4+ astrocytes frequently co-label with Conn43, while GFAP+ astrocytes rarely co-label with Conn43. Taken together, this information suggests the AQP4+ astrocyte sub-population possesses an adaptive (homeostatic) profile, while GFAP+ astrocytes may not. Because of the increase in abnormal morphology of these cells, they may be negatively impacted by an ongoing inflammatory environment, partially losing their homeostatic functions.

When phosphorylated, connexin-43 (p-Conn43) can become maladaptive and contribute to inflammation.^37,38 To evaluate if p-Conn43 changes after injury, we assessed p-Conn43 immunoreactivity in control and injured samples. Figure 7A–I shows images taken from interlaminar astrocytes; p-Conn43 is not expressed in control tissue, while the abnormal appearing and beaded interlaminar astrocytes in all of the injured brains express p-Conn43 that colocalizes with GFAP (Fig 7J–L).

Figure 7.

Interlaminar GFAP+ astrocytes express phosphorylated-connexin-43. (A–I) Co-labelling of fluorescent antibody markers against GFAP (green) (D–F) and phosphorylated connexin-43 (p-Conn43—red) (A–C) to PFC coronal brain sections can be seen in the abnormal, beaded distribution seen in the injured brain samples (images from 2 different samples are shown) (B–L). Very little fluorescence of p-Conn43 can be seen in the control samples (A–C, J). The merged images show a distinct overlap between GFAP and p-Conn43 in the interlaminar astrocytes (H, I, K, L). Higher power images can be seen in (J–L).

Immunoglobulin mediated complement targets neurons

TBIs have the capacity to create an inflammatory environment, release DAMPs, and allow the peripheral adaptive immune system to gain access to a semi-immuno-privileged space (the CNS). As described earlier, the BBB protects the CNS from peripherally circulating fluids, including immune cells. After TBI, the BBB often becomes compromised, permitting innate and adaptive immune cells to enter and react. Individuals exposed to repeat TBIs face increased risk of the peripheral immune system entering the CNS, contributing to chronic inflammation in the brain. The entry of peripheral immune cells, a chronically inflamed environment, and DAMPs brings enhanced danger for the development of autoimmune conditions, where the immune system recognizes, tags, and attacks autoantigens.

To determine if autoantibodies exist in the injured brain samples and the potential impact of their presence reduced homeostasis, we immunoreacted the control and injured tissue samples for the presence of immunoglobulin class G (IgG) and complement component C3. We hypothesized that if an autoimmune reaction occurred, it would target astrocytes, as they exist at the centre of TBI pathology and could potentially aid in understanding the observed abnormal morphologies. We selected IgG1 as the most common class and isoform of autoantibody present in another autoimmune condition (neuromyelitis optica spectrum disorder), where immune cells mark and destroy astrocyte derived AQP4.¹⁴ After the target of autoantibodies opsonizes with IgG, several possible mechanisms may occur to destroy the marked target. One possibility is that complement can recognize the Fc portion of the autoantibody and bind, initiating complement cascade activation. For this reason, we considered the presence of C3 in injured brains as an effect of any antibody presence. Figure 8 reveals that IgG was detected in each of the injured brain samples, while not present in control samples (not shown). Surprisingly, IgG labelled neurons residing in and around layer 2 and 3 of the PFC, rather than astrocytes. When the samples were labelled with C3, we discovered IgG and C3 to co-label the same neurons (Fig. 8A–F). We observed this pattern of co-labelling in every injured brain sample. To a lesser degree, appearing in only 2 of the 4 injury samples, we detected the abnormal interlaminar astrocytes labelled with both IgG and C3 (Fig. 8G and H). Taken together, these observations reveal that an autoimmune reaction may occur in the brains of military blast TBI wounded resulting in a complement response. They also reveal interlaminar astrocytes to be included, albeit to a lesser degree, as the antigenic target of autoantibodies.

Figure 8.

Layer-2–3 neurons and layer 1 interlaminar astrocytes label with autoantibody and complement. Immunoreactivity against immunoglobulin G subclass 1 (IgG1-red) and complement component 3 (C3—green) in layers 2–3 of the PFC revealed co-labelling of neurons (A–F) and (G, H) interlaminar astrocytes from injury samples, while controls showed no label (controls not shown).

Discussion

Summary

This project studied the pathological response to military blast TBIs by focusing on astrocytes, which play an important role in the homeostatic regulation of the brain. We also found augmented inflammation by revealing increased immunoreactivity of microglia in the PFC. We concentrated on assessing the expression patterns of astrocyte-derived proteins involved in regulation of homeostasis, including AQP4 and Conn43. We re-confirm that military blast related TBIs induce chronic astrogliosis and neuroinflammation in interface regions of the neocortex²² and show that in this reactive environment, astrocytes display abnormal morphology, specifically in interlaminar astrocytes. We also note astrocytic changes near the vasculature, solidifying the idea of interface concentrated pathology. We observed separate and distinct groups of GFAP expressing and AQP4 expressing astrocytes, identifying a novel population. Blast injuries shift astrocytes towards a reactive profile, with increases in abnormal morphology and changes in expression of multiple proteins including GFAP, AQP4, and Conn43. We conclude that post blast injury, GFAP+ astrocytes particularly shift from normal morphology, compared with those that primarily express AQP4. We suggest that blast TBI creates chronic pathology leading injured individuals toward secondary neuronal injuries involving immunoglobulin mediated complement pathology.

Types of astrocytes

From this and previous work we confirmed that the PFC of individuals exposed to blast related TBIs in the military show chronic repercussions in their cellular and molecular environment, where microglia and GFAP+ astrocytes reveal increased immunoreactivity.²² This type of injury significantly impacts the morphology of astrocytes; they lose fine branches and instead have a discontinuous, beaded shape, particularly prominent in areas of interface tissue (sub-pial, perivascular, grey-white border). Various types of astrocytes (e.g. protoplasmic, fibrous, varicose) reside in specific brain regions.³⁹ Protoplasmic astrocytes are located predominantly in the grey matter contributing to maintenance, care, and protection of CNS components.⁴⁰ Fibrous astrocytes typically reside within the white matter and provide support to neurons.⁴⁰ Varicose astrocytes also have a beaded appearance, similar to the description outlined in our findings and reside mostly in neocortical layers 5–6 of hominoids.^39,41 While the function of the varicose astrocyte is not clear, the beaded shape may assist in harbouring ions such as calcium, which allow for modulation of neural activity/synaptic activity, a phenomenon that aids in rapid intercellular communication and information processing as well as having involvement in the regulation of vasculature.^42-44 Others suggest that varicosities only emerge on astrocytes because of certain conditions, suggesting pathogenic development, while also lending to ambiguity and lack of understanding in the field at this point in time.⁴⁵ Varicose astrocytes exist in specific neocortical layers (5–6), however, and are likely not the reactive astrocytes we observed. The varicose astrocytes, in addition to residing in the deeper layers also present with a tangential extension of their processes, compared with the predominant version of the varicosity presenting astrocytes observed here.

The interlaminar astrocytes appear specifically involved in pathology after blast injury in this blast injured population. We observed dramatic alterations of their normal morphology, as the relatively smooth processes became interrupted and beaded (Figs 2, 5, 7, and 8). The beaded structures were also surrounded by AQP4 immunoreactivity, colocalized with p-Conn43 and C3 complement after blast injury, but not in control tissue, suggesting that these specific cells are strongly impacted by the pathologic conditions. Their role is not clear, but because they extend into, and presumably contact, deeper layers, it is likely this communication is impaired after TBI.^25,27 Falcone et al.⁴⁵ also report that various morphologic alterations occur in interlaminar astrocytes with no clear explanation of the underlying cause. They also find that varicosities observed in interlaminar astrocytes appear in conjunction with similar changes in varicose astrocytes, suggesting the long processes of both cells may be subject to similar pathologies.⁴⁵

Clasmatodendrosis

A strong beaded morphology of astrocytes occurs throughout the entire neocortex after blast TBI, suggesting they belong to the pathologic condition known as clasmatodendrosis. Clasmatodendrocytic astrocytes are irreversibly injured cells presenting with a loss of fine continuous processes, replaced by disintegrated processes with the development of vacuolizations or beads⁴⁶.⁴⁷ The onset usually follows trauma, neurodegenerative diseases, or old age because of hypoxic conditions, vessel leakage, and/or faulty clearance⁴⁶.⁴⁷ A study investigating post stroke survivors with and without dementia concluded that the presence of clasmatodendrocytic astrocytes is a pathological substrate contributing to dementia.⁴⁸ The team also reported that the presence of clasmatodendrocytic astrocytes in Alzheimer’s disease (AD) with cerebrovascular disease (CVD), but not in AD without CVD, suggesting a vascular event contributed to the abnormality.⁴⁸ This idea aligns with our findings here, i.e. a concentration of injury occurs at interface regions, including between astrocytes and blood vessels. The same group explored the effects of hypoperfusion injury on GFAP and AQP4 astrocytes in a baboon animal model (non-hominoid) and found a redistribution of cells transformed into clasmatodendrocytes, further suggesting hypoxic conditions impact the importance of gliovascular networks.⁴⁸ It is important to note that others report the beaded abnormal morphologies described here and elsewhere are not a result of post-mortem autolysis⁴⁹; several investigations provide evidence that the abnormality emerges pre-mortem and therefore a result of ensuing pathologies.^47,48,50 Here, we quantify the abnormal morphologies and show significant differences between military blast TBI and control brain tissues collected and processed under similar circumstances, strongly reinforcing the idea that this phenomenon is pathologic.

A study researching the presence of clasmatodendrocytic astrocytes in influenza associated encephalopathy concluded the abnormality of cell shape links with synapses on dendritic spines but does not represent autophagic astrocyte cell death. The authors came to this conclusion using LC3B and p62, common markers for autophagy and protein degradation, which revealed no association between the markers and the astrocytes.⁵¹ In contrast to those findings, Sakai et al.⁵⁰ found that for individuals who suffered various head traumas (contusions, haematomas, haemorrhage), clasmatodendrocytic astrocytes associate with accumulations of p62 and K48 in the beaded regions, suggesting dysfunctional protein degradation. A more recent study investigating the involvement of age-dependent cognitive decline in a mouse model identified a group of astrocytes that shared a similar morphological appearance, however, they also concluded the astrocyte beads associated with dysfunctional autophagy, decreasing the ability to regulate homeostatic CNS environments.⁵² They showed that inhibiting mTOR or the proteasome induced the abnormal phenotype, which they called autophagy dysregulated astrocytes, or APDAs. This group concluded that APDAs distinctly differ from the clasmatodendrocytic astrocytes described above. Taken together, these findings may indicate that APDAs, while sharing a similar morphology as clasmatodendrocytes, are a newly identified class of pathologic astrocyte. Future work will need to be conducted regarding military blast-related TBIs to determine if protein degradation machinery is compromised within the astrocytes reported here.

Astrocyte populations shift after blast related TBI

Blast injury leads to a shift in astrocyte sub-populations, moving from groups either AQP4+ or GFAP+ in roughly equal numbers, with normal morphology, to a significant increase in astrocytes that co-label with GFAP+ and AQP4+ and have an abnormal cell shape. Astrocytes possess both beneficial and detrimental characteristics in various circumstances and phases of recovery.⁵³ To assist in determining the functional relevance of the astrocytes in our samples, we investigated Conn43 as another marker to determine if either of our identified GFAP+ or AQP4+ astrocytes are more likely to contribute to the restoration and homeostasis of the injured brain. As previously mentioned, Conn43 contributes to homeostasis through spatial buffering and creation of glial connectomes.^12,13,17 Conn43 significantly co-labelled with AQP4+ astrocytes compared with GFAP+ astrocytes, suggesting that AQP4+ astrocytes are more beneficial in nature, working toward a healthy environment, whereas the GFAP+ sub-population may not be as accommodating. To support this idea, we investigated the presence of phosphorylated Connexin-43, a modified version of the protein connected to the promotion of inflammation. p-Conn43 did not appear in any of the control brains in our samples. In the injured samples, we observed p-Conn43 co-labelled with the beaded interlaminar GFAP+ astrocytes and not the AQP4+ astrocytes, suggesting that GFAP is more likely to coincide with pathologic markers and not contribute to homeostasis.

A recent study by Braun et al.⁵⁴ also studied AQP4 immunoreactivity in human PFC that received similar blast injuries. Our results resemble several features of their findings, although we did not measure the laminar parameters identical to those in their report. We agree that AQP4 reactivity increases around blood vessels after blast injury and that upper layers of neocortex show increased fluorescent density of AQP4. We also agree that increased AQP4 immunoreactivity occurs in the grey-white matter junction, although we did not precisely measure this feature, which may account for any subtle distinctions our observations.

Astrocytes + autoimmune involvement

Finally, we aimed to understand if threats from maladaptive secondary injuries contribute to the dramatic changes seen in the abnormal morphology observed in astrocytes after blast injury. Given the context that blast exposure is repetitive during the individual’s military tenure and the chronic reactivity seen here, we evaluated for the presence of IgG autoantibodies. We found immunoreactivity for this autoantibody predominantly in neurons residing in layers 2–3, and in some cases, the intralaminar astrocytes originating in layer 1 of the PFC. In other prominent CNS autoimmune diseases targeting astrocytes, such as neuromyelitis optic spectrum disorder, IgG is the primary isoform of autoantibody.⁵⁵ As indicated earlier, TBIs result in the increased presence of DAMPs from damaged cells. Most DAMPs from TBIs are high mobility group box 1, or HMGB1, a protein not often exposed extracellularly, and can linger in the brain, persisting well after an initial trauma.⁵⁶ In the samples analyzed here, the individuals had a history of repeated exposure to blast-related TBI, likely increasing the concentration of DAMPs, such as HMGB1. Interestingly, HMGB1 often results from hypoxia and blast injuries also result in increased incidences of hypoxia in the brain.^57,58

We also found that the presence of autoantibodies in the injured tissue co-labelled with complement protein C3. Complement in the inflamed brain is primarily derived from reactive astrocytes^59,60 and inflicts toxicity towards various CNS cells including neurons, microglia, and endothelial cells.⁶¹ Often the complement system is assumed to be a part of the innate arm of the immune system, however, it can be involved in the adaptive immune response as well, by interacting with immunoglobulin. The Fc portion of autoantibodies can bind to complement C3, initiating the subsequent cascade that culminates in the formation of the membrane attack complex.^62,63 This suggests blast-related military TBIs, resulting in chronic reactive CNS environments, are subject to the development of autoimmune conditions arising from autoantibodies with the capacity to attract complement activation, directed toward neurons and astrocytes.

Limitations

The limitations to this study include a relatively small sample size for both the non-injured and blast-exposed cases. The results in some instances indicate a trend toward significance and it would be useful to increase the number of samples to elaborate on our findings. Likewise, other areas of the brain, such as the anterior cingulate cortex and the hippocampus may be useful to describe the unique pathophysiology originating from military related TBIs. In addition, a genetic analysis of the astrocytes would help to confirm the sub-populations of astrocytes and how their phenotypic profiles can change over time. Considering that these samples are human creates a challenge as human samples have a high degree of genetic and environmental variability. To elevate our understanding of autoimmunity, we would also like to assess blood samples; potential blood related biomarkers may contribute to diagnosis and subsequent treatment options.

Conclusions

Our findings suggest that military blast-related TBIs have the capacity to offset CNS homeostasis through disruption of astrocyte sub-type populations. We report distinct populations of neurons that preferentially express AQP4 or GFAP; the GFAP expressing astrocytes rarely co-label with Conn43, while the majority of AQP4+ cells are Conn43+. Individuals who suffered from blast injuries display chronic inflammatory markers, which correlate with the incidence of autoantibodies, primarily directed towards neurons, with the ability to attract complement proteins.

Supplementary material

Supplementary material is available at Brain Communications online.

Funding

This work was supported by USU-PAT-74-10982 (Defense Health Agency) and USU-PAT-74-12492 (Jean Perkins Foundation).

Competing interests

The authors have no conflicts of interest to report.

Data availability

All data are available for inspection upon request.

Disclaimer

The views presented here are those of the authors and do not represent views of the Uniformed Services University, the Department of War, or the US Government.