Introduction

Concussion is now recognized as a major health issue, with many millions of cases reported worldwide each year [11, 18, 61]. Indeed, although alternatively referred to as mild traumatic brain injury (mTBI), for many there is nothing ‘mild’ about concussion, with approximately 15% of patients suffering persisting neurocognitive dysfunction [68, 70]. Moreover, an increasing number of reports link exposure to repetitive concussions with progressive neurodegeneration, including ‘chronic traumatic encephalopathy’ (CTE) [33, 57, 83].

Despite growing public alarm regarding the aftermath of concussion, there remains no consensus on the underlying pathophysiology of the injury. In part, this is due to the limited availability of human post-mortem brain tissue for examination from this typically non-lethal injury. As such, concussion remains in part defined by an absence of the overt brain pathologies characteristic of higher levels of TBI, including, hematoma, contusion or marked brain swelling [51]. However, emerging experimental and clinical data indicates that concussion may share one of the most consistently observed pathologies of moderate and severe TBI, diffuse axonal injury (DAI).

Across all injury severities, DAI is thought to stem from the viscoelastic nature of the brain under dynamic mechanical loading during head rotational acceleration [3, 26, 36, 37, 43, 58, 84, 86, 87]. Potentially due to their highly structured organization in tracts and unique anatomic architecture, axons appear especially vulnerable to rapid brain tissue deformation during TBI. This has been demonstrated in a swine model of concussion that employs head rotational acceleration parameters scaled to human concussion [17, 44]. Neuropathological analysis of the injured swine brains identified selective, multifocal regions of axonal pathology throughout the white matter very similar in appearance to characteristic DAI observed at autopsy after severe TBI in humans [44]. In addition, a growing number of reports using advanced neuroimaging and fluid biomarker analyses lend further support for DAI as a key pathological substrate of concussion in both humans and animal models [12, 55, 59, 76, 81, 94, 95].

Intriguingly, emerging evidence suggests that other brain structures also suffer mechanical injury during concussion, which like DAI typically are invisible to conventional, non-invasive examination. In particular, while by definition there is no focal hemorrhage in concussion, techniques such as dynamic contrast enhanced (DCE) MRI have demonstrated contrast accumulation in the brains of football players [93], suggesting there is subtle disruption of the cerebral vasculature affecting the integrity of the blood–brain barrier (BBB). In addition, in previous preclinical studies, BBB dysfunction has been observed in the absence of other significant neuropathologies following an open-skull, direct impact model of TBI in cats [66, 67]. In these studies, BBB permeability was demonstrated to occur via multiple routes including increased transcytosis as well as hypertension-induced endothelial breakdown. Furthermore, autopsy studies in humans have demonstrated acute and long-lasting disruption of BBB integrity as a common neuropathological finding after just a single moderate or severe TBI [34]. However, the specific contribution of isolated mechanical forces on the BBB, including rotational forces common to concussion, has yet to be explored directly.

Here, we evaluated potential BBB disruption following injury in the swine head rotational acceleration model of concussion. We found acute and marked BBB leakage at levels of injury insufficient to cause hemorrhage. Detailed neuropathological examination revealed that BBB permeability occurred in a pattern and distribution consistent with a biomechanically induced pathology and was associated with an acute astroglial response. Parallel studies of human cases of moderate and severe TBI confirmed neuropathological features similar to those found in the swine concussion model, though with evidence of more extensive BBB disruption, consistent with previous observations. Together, these data indicate that, in addition to DAI, BBB dysfunction may also be an important pathology of concussion.

Materials and methods

Experimental design

We hypothesized that dynamic forces caused by rapid head rotational acceleration–deceleration in adult swine will induce acute extravasation of blood-borne proteins into the brain parenchyma that are detectable via immunohistochemistry (IHC). In a controlled laboratory experiment, 6-month-old (adult) Hanford miniature swine underwent the rotational acceleration model of TBI under general anesthesia and survived for 6 h (n = 3), 48 h (n = 2) and 72 h (n = 4), and were compared with shams (n = 3). Based on previous data [44] the proposed number of animals was expected to provide significant power, specifically where the neuropathological outcome was anticipated to be absent in sham animals. All histological experiments, analyses and quantification were performed blind to the injury status of the animal.

Rotational acceleration model of concussion

All animal experiments were conducted in accordance with protocols approved by The University of Pennsylvania Institutional Animal Care and Use Committee and in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

Six-month-old female Hanford miniature swine were subjected to an established concussion model that induces injury via forces caused by rotational acceleration/deceleration due to pure impulsive centroidal head rotation [17, 44, 58, 71, 82, 85]. For TBI in humans, the size of the brain is critical as significant mass effects between regions of tissue can create high strains during dynamic brain deformation during rapid accelerations [36, 37, 54]. Accordingly, injury parameters were scaled up relative to brain mass to create the mechanical loading of brain tissue relevant to that which occurs in human TBI [36, 37, 54, 58, 89]. Notably, in this model, there is an absence of any head impact or forces arising from linear acceleration/deceleration.

Following the induction of anesthesia using 0.4 mg/kg midazolam IM and 5% inhaled isoflurane, animals were intubated and anesthesia maintained via 2.5% inhaled isoflurane. The HYGE pneumatic actuator device was used to induce rapid head rotation. Specifically, the HYGE actuator generates linear motion via the triggered release of pressurized nitrogen. This linear motion is then converted to angular motion via custom-designed linkage assemblies to induce rotation over 20 ms. Rotational kinematics were recorded using angular velocity transducers (Applied Technology Associates) mounted to the linkage sidearm coupled to a National Instruments data acquisition system running custom LabView software (10 kHz sampling rate). In this fashion, we produced pure impulsive centroidal head rotation of up to 110° in the coronal plane with peak angular velocity of 225–250 rad/s. Animals were recovered from anesthesia and returned to the housing facility. While the procedure is non-surgical, preemptive analgesia was provided post-injury in the form of 0.1–3 mg of Buprenex (slow release preparation) SQ and acetaminophen 50 mg/kg PR. Sham animals were also euthanized 72 h after being subjected to identical procedures absent head rotation.

At the study endpoint, all animals were deeply anesthetized and transcardially perfused using chilled heparinized saline (2 L) followed by 10% neutral buffered formalin (NBF) (8 L). Perfusions were performed at standardized pressures calibrated for the provision of adequate fixation and the absence of any structural damage to the vasculature that would permit artefactual extravasation of serum proteins. This was verified histologically (described below) and fixation was sufficient to permit quality tissue morphology and immunoreactivity. Brains were subsequently post-fixed for 7 days in 10% NBF, sectioned into 5 mm blocks in the coronal plane and processed to paraffin using standard techniques.

Human post-mortem acute severe TBI cohort: demographic and clinical data

Since uncomplicated concussion is typically non-lethal, scarce cases come to autopsy examination. Therefore, post-mortem brain tissue from patients who died following single severe TBI was examined to determine if there were shared features in the distribution of serum protein extravasation and the cellular response to BBB breakdown between human autopsy material and our swine injury model across injury severities. All tissue was obtained from the Glasgow TBI Archive, Department of Neuropathology, Queen Elizabeth University Hospital, Glasgow, UK. Tissue was acquired at routine diagnostic autopsy and approval for use was granted by the NHS Greater Glasgow and Clyde Bio-repository Governance Committee.

TBI cases were selected to include patients with survival times ranging from 6 to 72 h following acute severe TBI (n = 12). Detailed reports from the diagnostic autopsy and/or forensic reports were available for all cases and indicated a history of a single severe TBI, supported by autopsy findings. TBI cases were compared to material from age/sex-matched controls (n = 7) acquired at routine diagnostic post-mortem at the same institution. Controls had no documented history of TBI. Full clinical and demographic information, including age, sex and cause of death, is provided for all groups in Table 1.

Table 1 Demographic and clinical data
Full size table

Human brain tissue preparation

For all examinations, the intact brain was immersed in 10% formol saline at autopsy and fixed for at least 3-weeks prior to dissection. Sampling using a standardized protocol and paraffin embedding was performed as described previously [31]. Analyses were performed using hemi-coronal sections of the parasagittal cortex at the level of the mid-thalamus to include the cingulate gyrus, corpus callosum and superior frontal gyrus. This region was selected given its midline location and known susceptibility to injury [1, 2, 42, 43, 49].

Single immunohistochemical labeling

Human and swine tissue was examined for extravasation of the serum protein fibrinogen (FBG) and axonal pathology using IHC. Following deparaffinization and rehydration, human and swine tissue sections were immersed in aqueous hydrogen peroxide (15 min) to quench endogenous peroxidase activity. Antigen retrieval was performed in a microwave pressure-cooker with immersion in preheated Tris EDTA buffer. Subsequent blocking was achieved using 1% normal horse serum (Vector Labs, Burlingame, CA) in Optimax buffer (BioGenex, San Ramon, CA) for 30 min. Incubation with the primary antibodies was performed overnight at 4 °C. Specifically, to identify axonal pathology in human and swine tissue, sections were labeled with an antibody reactive for the N-terminal amino acids 66–81 of the amyloid precursor protein (APP) (Millipore, Billerica, MA) at 1:80 k permitting visualization of APP accumulating within transport-interrupted axonal bulbs or swollen and tortuous varicosities along the length of damaged axons. In addition, adjacent sections were incubated with antibodies specific for the serum protein FBG including rabbit polyclonal antibodies targeting the full-length swine FBG at 1:5 k (Abcam, Cambridge, MA) and human FBG at 1:17.5 K (Dako, Carpinteria, CA). FBG does not normally accumulate in the brain parenchyma under physiologic conditions. After rinsing, sections were incubated with the appropriate biotinylated secondary antibody for 30 min (Vector Labs, Burlingame, CA), followed by avidin–biotin complex (Vectastain Universal Elite kit, Vector Labs, Burlingame, CA). Finally, visualization was achieved using 3,3′-diaminobenzidine (DAB) (Vector Labs, Burlingame, CA) and counterstaining with haematoxylin performed.

Positive control tissue for APP IHC included sections of swine tissue with previously established DAI. Positive control tissue for FBG IHC included a section of swine brain tissue with contusional injury, and thus overt BBB disruption. Omission of primary antibodies was performed on the same material to control for non-specific binding.

Multiple antigen labeling

Immunoenzymatic double labeling

Sections were examined to directly compare the pattern and distribution of FBG extravasation with that of axonal pathology in swine. Specifically, APP staining was first performed using DAB visualization as above. The same sections were subsequently incubated with rabbit anti-FBG antibody (1:3.5 K) for 20 h at 4 °C. Detection was achieved via the ImmPRESS™-AP anti-rabbit IgG (alkaline phosphatase) polymer detection kit (Vector Labs, Burlingame, CA) followed by the Vector blue alkaline phosphatase substrate kit (Vector Labs, Burlingame, CA). After rinsing, slides were coverslipped using aqueous mounting medium (Dako, Carpinteria, CA).

Immunofluorescent (IF) labeling

Double and triple IF labeling was performed in swine tissue to (1) compare the extravasation of FBG with that of another blood-borne protein IgG in a section of tissue at the level of the basal ganglia at the level of the head of caudate nucleus in all cases and; (2) determine the cell types associated with the uptake of FBG/IgG in all sections where observed on single labeling IHC.

Following deparaffinization and rehydration, antigen retrieval and blocking was performed as above. Primary antibodies were applied serially for 20 h (4 °C) and were specific for FBG (Abcam, Cambridge, MA, 1:3 K), glial fibrillary acidic protein (GFAP) (Millipore, Billerica, MA; 1:8 K), ionized calcium-binding adaptor molecule 1 (IBA-1), and microtubule-associated protein 2 (MAP2) (Abcam, Cambridge, MA; 1:2 K). After rising, the corresponding Alexa Fluor (Invitrogen, Carlsbad, CA) secondary antibody was applied at 1:500 in a 2% species-specific blocking solution for 2 h at room temperature. For the identification of IgG, a goat anti-swine IgG conjugated with Alexa 488 (Jackson ImmunoResearch, PA; 1:20) was applied. Serial sections of positive control tissue (contused tissue) were subjected to the entire procedure with the omission of subsets of primary antibodies to control for non-specific immunofluorescence. Negative controls included individual omission of primary antibodies to control for non-specific binding. Following rinsing, all double fluorescent-immunolabeled sections were coverslipped using an aqueous mounting medium (Dako, Carpinteria, CA) and visualized using a Nikon A1RSi Laser Scanning Confocal (Nikon, Tokyo, Japan). High resolution confocal scanning was performed at 0.5 μm increments through the full depth of the tissue sections and image reconstruction achieved using Nikon Elements AR.

Analysis and quantification of immunohistochemical findings

All neuropathological assessment and quantification was performed blinded to the injury status of animals by 2 independent examiners and inter-rater reliability was determined. Routine H&E staining was performed on serial sections for all brain regions examined (see below).

Extent and distribution of FBG extravasation

All tissue examinations were performed on 8 μm whole-brain coronal paraffin sections at 5 brain levels including: (1) the frontal cortex at 5 mm from the frontal pole, including prefrontal cortex; (2) the basal ganglia at the level of the head of caudate nucleus; (3) the anterior hippocampus; (4) the posterior hippocampus at the level of the posterior commissure; and (5) the medulla. Levels were selected to incorporate a wide sampling spanning the rostrocaudal extent of the entire brain.

For each animal, FBG labelled whole-brain sections from each of these 5 levels (see Fig. 2a for examples of brain levels examined) were subjected to high resolution scanning at 20× magnification using the Aperio ScanScope and viewed using associated Aperio ImageScope software (Leica Biosystems, Wetzlar, Germany). These digital images were reviewed by 2 independent observers (VJ and MW) each of whom manually outlined all regions of extravasated FBG to generate detailed maps of BBB disruption (Fig. 2). This was achieved using the annotation tool on the Aperio ImageScope software. These subsequent outlines then permitted the quantification of the percentage area of FBG immunoreactivity for each hemisphere (across all 5 brain levels) per animal.

Quantifying regional overlap between FBG extravasation and axonal pathology in swine

The pattern and distribution of axonal pathology as identified by APP immunohistochemistry has previously been characterized in the non-impact rotational acceleration injury swine model of mild TBI used in this study [44].

To determine the potential relationship between axonal pathology and BBB disruption, tissue dual-labelled for APP (axonal pathology) and FBG were examined. Animals with the survival of 48–72 h post-TBI were selected to encompass the maximal extent of axonal pathology, which is identifiable due to the abnormal accumulation of APP secondary to a failure of axonal transport.

Dual-labelled coronal sections for all cases at the level of the basal ganglia at the level of the head of caudate nucleus (Fig. 3a, b) were selected and, as above, underwent high resolution digital scanning at 20× magnification using the Aperio ScanScope and viewed using associated Aperio ImageScope software (Leica Biosystems, Wetzlar, Germany).

The entire region of white matter above the level of the corpus callosum for each animal/section in both hemispheres was selected for examination and has previously been shown to be a region where axonal pathology is stereotypically observed [44]. Within this region, injured axons were identified and defined as those that were APP positive axonal profiles with an injured morphology including (1) terminal axonal swellings or axonal bulbs, formerly known as retraction balls, and (2) axons with a beaded or fusiform morphology representing multiple points of apparent transport interruption as has been historically described [27, 28, 42, 43, 77, 88]. Upon identification, injured axons were tagged using the annotation tool in the Aperio ImageScope software. To determine the percentage of axons regionally associated with BBB disruptions, all axons were reviewed and designated as either (1) existing in a region of white matter with associated FBG extravasation or (2) existing in a regions where FBG immunoreactivity was absent.

Human tissue neuropathological analyses

For human tissue studies, the extent of FBG extravasation was classified using standard semi-quantitative assessments and designated in a blinded fashion as: 0 = absent, 1 = minimal, 2 = moderate, or 3 = extensive pathology based on the frequency and intensity of immunoreactivity as previously described [34]. Score 3 (extensive) represents sections with > 50% of the section being FBG immunoreactive. Representative examples are shown in Supplementary Fig. 2. This scoring was representative of the entire hemi-coronal section of the parasagittal cortex at the level of the mid-thalamus to include the cingulate gyrus and corpus callosum. The presence or absence of axonal pathology was noted.

Statistical analyses

Statistical analyses for IHC quantification were performed using GraphPad Prism statistical software (GraphPad Software Inc. La Jolla, CA). Differences in the mean of the percentage area of FBG immunoreactivity were assessed using the two-sample t test for all comparisons. Differences were considered significant if p < 0.05.

The Intraclass Correlation Coefficient (ICC) was used to determine inter-rater reliability of quantitative measurements of IHC findings using IBM SPSS Statistics (IBM Corporation, Armonk, NY).

Results

Isolated head rotational acceleration results in a clinical presentation consistent with concussion

Following the rotational acceleration model of concussion and subsequent withdrawal of isoflurane, all animals mobilized rapidly, were fully conscious and alert, mental status was unaltered and normal feeding and drinking behavior resumed within 1–2 h. Consistent with the clinical presentation of mild TBI, no animals displayed evidence of focal neurological deficit. In all animals, neurological examination was normal and animals displayed normal posture, tone, gait, power, sensation, proprioception and cranial nerve examination where possible to examine.

Neuropathological characterization of trauma-associated pathologies was performed following survival of 6 h (n = 3), 48 h (n = 2) and 72 h (n = 4) post-injury and compared with shams (n = 3). As previously reported [44], brains were normal on gross examination and indistinguishable from shams at all time points post-injury. Consistent with clinical concussion, there was no evidence of any focal pathology, including hemorrhage. The brain hemispheres were symmetrical with no evidence of swelling or raised intracranial pressure (ICP). H&E staining confirmed an absence of any focal hemorrhagic or ischemic foci in any animals.

IHC specific for APP revealed swollen and morphologically altered axons consistent with transport-interruption secondary to axonal cytoskeletal damage and indistinguishable from that observed in human DAI [1,2,3, 27, 28, 42, 77]. DAI was observed in injured animals in all brain levels examined in a stereotyped multifocal distribution consistent with biomechanical forces as previously described in detail [44] and discussed below.

Experimental concussion results in acute BBB disruption

Multifocal FBG extravasation into the brain parenchyma was observed at 6–72 h following a single, rapid rotation in swine as perivascular FBG immunoreactivity in both white and gray matter in injured animals, but was observed minimally in shams (Fig. 1a–i). Further, detailed mapping of FBG extravasation in multiple whole-brain coronal sections revealed BBB leakage in a stereotyped, multifocal distribution consistent with a biomechanical etiology. Specifically, structural interfaces within the brain were preferentially affected, with the evidence of BBB disruption typically observed at gray–white interfaces, in periventricular regions (Figs. 1d–f, 2c, e) and at subpial zones within the cortex, often towards the dorsal surface anteriorly (Figs. 1h, 2c). Occasionally, inflection points in the brain architecture were also affected, including cortical sulci and the lateral aspect of the parahippocampal cortex where it meets the inferior temporal gyrus (Fig. 1f), further supporting a biomechanical etiology. In all cases, the hippocampus was affected bilaterally, with prominent FBG extravasation at the interface between the hilus/molecular layer and the dentate gyrus and the vessels of the hippocampal fissure (Fig. 1g). Finally, FBG extravasation was observed in the thalamus at all time points (Fig. 2d).

Fig. 1
figure 1

Experimental concussion induces multifocal BBB leakage. a An absence of FBG immunoreactivity in the subcortical white matter in a sham animal. Immunoreactivity specific for FBG; b surrounding a vessel in the parenchyma of the thalamus 6 h following a single experimental concussion; c in the subcortical white matter at the level of the head of the caudate nucleus 48 h following experimental concussion; d at the tip of the lateral ventricle preferentially in left hemisphere (leading hemisphere in rotation) 6 h following experimental concussion; e at the ventricular interface of the corpus callosum 72 h following experimental concussion; f in the periventricular white matter and depth of the sulci at the lateral aspect of the parahippocampal cortex where it meets the inferior temporal gyrus, 6 h post-experimental concussion; g adjacent to the dentate gyrus and vessels of the hippocampal fissure 6 h following experimental concussion; and h at the cortical surface in the frontal lobe 72 h post-experimental concussion. i Graph showing the percentage area of FBG immunoreactivity for the whole coronal brain hemispheres for each animal in all 5 levels combined. Scale bars a, c 200 μm, b, e 300 μm, d 3 mm, f 2 mm, g 400 μm, h 500 μm

Full size image
Fig. 2
figure 2

BBB leakage occurs in a biomechanical distribution after experimental concussion. a MRI image with vertical dotted lines indicating the levels of brain examined in the coronal plane. b–f Whole brain coronal maps showing FBG extravasation annotated manually following high resolution digital scanning in each of the 5 levels denoted in a. Scale bars 5 mm

Full size image

Quantification of the mean percentage area of FBG immunoreactivity in each hemisphere across 5 coronal sections was minimal in sham animals [left hemisphere: 0.18 ± 0.01%; right hemisphere: 0.29 ± 0.05%]. In comparison with sham, the mean percentage area of FBG immunoreactivity was increased in both hemispheres at 6 h [left hemisphere: 4.39 ± 2.30% (p = 0.03), right hemisphere 3.87 ± 1.88% (p = 0.03)], and at 48 h post-concussion [left hemisphere: 1.16 ± 0.46% (p = 0.02), right hemisphere 1.31 ± 0.05% (p = 0.0002)]. By 72 h post-concussion, FBG immunoreactivity remained increased versus shams in the left hemisphere [2.30 ± 1.32% (p = 0.04)], with a trend to increased staining in the right hemisphere [2.50 ± 2.27% (p = 0.16)] (Fig. 1i). No difference in the extent of FBG immunoreactivity was observed between TBI groups at 6, 48 and 72 h survival (p > 0.05). Inter-rater reliability between observers was excellent with an ICC of 0.98.

Notably, while the extent of FBG immunoreactivity did not differ between hemispheres (p > 0.05), a stereotyped hemispheric asymmetry was observed at all three time points and was highly consistent among animals. Specifically, in rostral brain regions (Fig. 2a, c), the left hemisphere (leading hemisphere during coronal rotation) displayed an increased density of extravasated FBG closer to the midline, including extensive leakage at the dorsolateral aspect of the lateral ventricle. This was in contrast to the right (non-leading) hemisphere, where more extensive extravasation was observed closer to the brain surface, including the digitate white matter and gray matter of the insular cortex and primary somatosensory cortex (Fig. 2c).

Immunoglobulin (IgG) IHC revealed perivascular extravasation consistent in both extent and distribution with that of FBG in swine (Fig. S1). Additional double immunofluorescent labeling in a subset of sections confirmed co-localization of FBG and IgG, consistent with evidence of BBB disruption leading to global leakage of large, blood-borne proteins, versus being a FBG specific phenomenon (Fig. S1).

Regional overlap of BBB disruption and axonal pathology following experimental concussion

The spatial relationship between axonal pathology and BBB disruption in swine was revealed by examining whole-brain sections at 48–72 h post-concussion at the level of the head of caudate nucleus (Level b; Fig. 3) labeled with both APP and FBG (multiple immunoenzymatic labeling) and compared to sham animals.

Fig. 3
figure 3

BBB leakage has limited regional overlap with axonal pathology after experimental concussion. a Whole section map of annotated FBG immunoreactivity at 48 h post-experimental concussion. b Adjacent section to a showing APP positive axonal pathology in a distribution that overlaps with FBG leakage in the white matter. c Graph showing the percentage of APP positive profiles identified in regions where coincident FBG extravasation is observed via multiple chromagen staining at 48 and 72 h post-experimental concussion. d FBG (blue) confined to the vascular compartment in a sham with no extravasation into the brain parenchyma. No axonal pathology is observed (brown). e, f, h Regions of extensive FBG extravasation (blue) and axonal pathology (brown) in the e deep cortical white matter 48 h following experimental concussion, f gray–white interface of the frontal cortex at 72 h post-experimental concussion and h the periventricular region of the leading hemisphere at 48 h post-experimental concussion. i Scattered axonal pathology both within and beyond regions of extravasated FBG in the white matter at 48 h post-experimental concussion. j APP immunoreactive axonal pathology in the perivascular region without FBG extravasation. g, k Overt FBG extravasation without axonal pathology in the gray–white interface (g) and deep white matter (k) at 48 h post-experimental concussion. Scale bars a, b 6 mm, d–k 100 μm

Full size image

Distinct axonal profiles with an injured morphology, including terminal swellings, beading and intact fusiform profiles were present in a stereotypical, multifocal distribution affecting the subcortical and periventricular white matter and midline structures of all injured animals, consistent with previous reports [44] (Fig. 3b). Notably, mapping of APP positive injured axons revealed overlap between regions of high-density axonal pathology and extensive FBG extravasation in white matter (Fig. 3a, b). However, although frequent, this overlap between APP positive axonal profiles and regions of FBG extravasation in white matter was not absolute (Fig. 3c, e–k). Specifically, though 62.8% (SD 1.8%) and 60.6% (SD 9.3%) of APP positive axons at 48 and 72 h, respectively, fell within a region of co-incident FBG extravasation, the remaining axons did not (Fig. 3c). There was no difference between time points (p = 0.8). FBG extravasation was also frequently observed in the absence of any axonal pathology in both white and gray matter (Fig. 3g, k).

Cellular uptake of FBG occurs acutely following experimental concussion

Single labeling IHC revealed that, following concussion in swine, in addition to parenchymal FBG immunoreactivity, staining was also present in cellular structures within regions of high extravasation (Fig. 4). Morphologically, these cells appeared consistent with astrocytes and neurons, confirmed in subsequent double and triple immunohistochemical labeling with GFAP (Fig. 4f–i) and MAP2 (Fig. 4j, k). In contrast, only very occasional microglia (IBA-1 positive) were observed co-labeled with FBG, typically closely adjacent to the vessel (Fig. 4h, i). High resolution confocal microscopy demonstrated this FBG immunoreactivity was within neurons and astrocytes, rather than confined to the cell surface, indicating internalization of the protein. All studies were repeated with IgG, which showed a similar pattern of neuronal and astrocytic uptake (Fig. S1).

Fig. 4
figure 4

Astrocytes and neurons internalize blood-borne proteins after experimental concussion. a FBG immunoreactive cells with a glial morphology at the depth of the sulcus within the inferior temporal gyrus 48 h post-experimental concussion. b Higher magnification of box in a. c Cells with neuronal morphology and immunoreactive for FBG in the caudate nucleus 72 h post-experimental concussion. d Normal vessels without FBG (red) extravasation. FBG is confined to the vessel lumen without co-localization with astrocytes (GFAP; green) or microglia (IBA-1; purple). e Higher magnification of box in d. f Vessels in premotor cortex showing marked FBG (red) extravasation 72 h post-experimental concussion. Co-localization with astrocytes (GFAP; green) is observed. Only minimal co-localization with microglia (IBA-1; purple) was observed in cells immediately adjacent to the vessel. g Higher magnification of region in f. h Penetrating surface vessels of parietal cortex showing marked FBG (red) extravasation 72 h post-experimental concussion with co-localization with astrocytes (GFAP; green). i Higher magnification of box in h. j Cells in the frontal cortex with neuronal morphology that are negative for IBA-1 (purple) and GFAP (green), yet co-localize with FBG (red). k Neuronal cell-type was confirmed with evidence of co-localization of MAP-2 (green) cells and FBG (red). Scale bars a, c 100 μm, b, e 25 μm, d, f–k 50 μm

Full size image

APP and FBG IHC following human severe TBI: comparisons with experimental concussion

Following severe human TBI, APP IHC revealed axonal pathology in all TBI cases in a pattern and distribution consistent with DAI, with axonal profiles observed individually scattered or in small clusters, indistinguishable to that observed following concussion in swine [42, 43]. However, in contrast with the concussion model, other cases displayed additional superimposed, widespread waves of axonal APP immunoreactivity in keeping with axonal pathology as a result of the vascular complications of raised intracranial pressure [23, 25, 32, 35, 42, 69]. This compared to human controls where just 1 of 7 cases displayed axonal pathology in a pattern and distribution in keeping with hypoxic/ischemic injury, likely indicative of an agonal event near the time of death [23, 25, 32, 35, 42, 69].

Moderate or extensive FBG immunostaining was observed in all but one TBI case, consistent with previous reports (Fig. 5) [34, 52]. This compared to controls in which there was less extensive FBG immunoreactivity, with just 1 case showing extensive, 1 moderate and 5 minimal extravasation (Fig. 5a), again in keeping with previous descriptions [34, 52]. Despite the more widespread and extensive FBG immunoreactivity in human severe TBI, consistent with observations in swine studies, there was increased FBG immunoreactivity adjacent to subpial vessels in the superficial layers of the cortex (Fig. 5c, d), as well as at the interface between cortical gray and underlying white matter (Fig. 5e).

Fig. 5
figure 5

FBG extravasation after acute severe TBI in humans. a An absence of FBG immunoreactivity out with the vascular compartment in a 20-year-old male control that died following a sudden cardiac event secondary to hypertrophic cardiomyopathy. b Marked FBG extravasation in a perivascular distribution at the gray–white interface in the superior frontal gyri in a 20-year-old male 48 h following TBI caused by assault. c FBG extravasation and associated glial uptake at the depth of sulci in a 58-year-old male that died 6 h following TBI caused by assault. d A penetrating vessel at the cortical surface of the cingulate gyrus surrounded by extravasated FBG (same case as b). e Extensive neuronal uptake of FBG in the superior frontal gyrus in a 58-year-old male 6 h after TBI caused by a fall. f High magnification of a neuron immunoreactive for FBG in an 18-year-old male 10 h after a fall. g An astrocyte immunoreactive for FBG with process extending towards a vessel at the surface of the cingulate cortex in a 56-year-old female 24 h following TBI caused by a road traffic accident. Scale bars a–e, g 100 μm, f 50 μm

Full size image

Furthermore, both neuronal and glial uptake of FBG was also observed in TBI and control cases (Fig. 5e–g), typically in regions with high parenchymal FBG extravasation. The extent of this cellular uptake was frequently extensive and widespread in human TBI cases, with immunoreactive astrocytes often present at the cortical surface (Fig. 5g).

Discussion

Using a swine model of concussion delivering pure head rotational acceleration and based on the kinematics of human concussion, we found evidence of multifocal BBB disruption shortly after injury. Extravasation of serum proteins was identified in a distribution that coincided with brain regions previously predicted to undergo greatest tissue deformation during injury. While this included overlap with areas of axonal pathology in the white matter, BBB disruption was also found independent of axonal injury throughout the gray matter. Notably, extravasated serum proteins were observed within astrocytes and neurons in proximity to regions of BBB disruption shortly after injury, suggesting rapid uptake by these cells. These data indicate that acute dysfunction of the neurovascular unit may be a common pathological feature of concussion.

While a wide range of pathologies have been characterized in human tissue studies from patients exposed to moderate or severe TBI [30], the rarity of post-mortem examination following concussion has meant that current understanding of the pathophysiology of the injury is based largely on indirect and non-invasive studies in vivo. Moreover, although there are multiple rodent models of ‘mild’ TBI, there is debate regarding their fidelity to replicate the biomechanics and pathophysiology of human concussion. In particular, the relatively large and gyrencephalic human brain endures substantial mass effects during head rotation associated with concussion, reflected in high strains as it is dynamically deformed. Therefore, through extensive characterization, we established a scaling approach to create similar mechanical loading of brain tissue between human TBI and the swine model. The parameters are based on Holbourn’s Scaling Equation [63], where the interaction of brain mass with the level of head rotational acceleration is taken into account [19, 39]. Specifically, for a swine brain of 90–100 g, the rotational acceleration forces must be increased eightfold or more to create the same tissue deformation that a human brain undergoes during concussion [41, 43]. This model recapitulates many aspects of the acute clinical presentation of mTBI, including loss of consciousness dependent on the angle of head rotation, as well as DAI in the absence of overt neuropathologies. However, there are anatomic differences between humans and swine, including skull architecture, that may affect the distribution of subtle neuropathologies and potentially outcome. In addition, the strain of swine used does not have nearly as prominent a falx cerebri as do humans. Since this structure may serve as a physical barrier during brain deformation, axonal pathology is much more prominent in midline structures in human TBI compared with the swine model.

Consistent with previous studies [44], here we identified DAI following a single rotational event in swine in a pattern and distribution consistent with mechanical deformation, and in the absence of any hemorrhage, brain swelling or ischemia. These findings support emerging clinical, neuroimaging and blood biomarker data indicating that DAI is a key pathological substrate of concussion in humans due to mechanical vulnerability of white matter axons [12,13,14, 21, 43, 55, 59, 62, 80, 94, 95]. Nonetheless, these findings do not preclude the possibility that other brain structures also suffer mechanical damage within and beyond white matter.

Using this swine model, our current data also demonstrate that the brain vasculature is vulnerable to damage as a consequence of the mechanical forces of concussion. While the specific ultrastructural nature of this mechanical damage to vessels remains to be determined, the resulting abrupt disruption of BBB integrity is clearly evidenced by multifocal pathological extravasation of large serum proteins, FBG and IgG. Of note, FBG immunoreactivity was observed at all time points post-injury and was most marked with a survival duration of 6 h. Interestingly, a biphasic course of BBB dysfunction has previously been described in the acute phase following models of severe TBI in rodents [9, 10]. While our data trends towards a relative decrease in FBG immunoreactivity at 48 versus 72 h, a greater number of animals would be required to determine the time course of acute BBB opening. In addition, as the presence of serum proteins in brain parenchyma is determined both by their rate of extravasation and clearance, this is unlikely to be a precise measure of BBB opening at any given time-point. Future studies utilizing the terminal intravascular administration of exogenous labels, e.g. Evans blue, will be important to accurately determine the acute temporal time course of BBB leakage in this model.

The duration of BBB permeability in the swine model beyond the acute phase will also be important to explore. Indeed, considering that FBG extravasation has been observed in post-mortem human brain months and even years after a single moderate-severe TBI [34], a milder form of persisting BBB may also occur in some cases of concussion. While the resulting downstream pathological effects of BBB dysfunction remain to be determined in single concussion, BBB permeability along with perivascular neuropathologies including neurofibrillary and glial tauopathy have been reported in individuals with CTE following exposure to repetitive concussion [24, 56, 83].

In the present study, FBG and IgG immunoreactivity was observed in a stereotypical, multifocal distribution, typically surrounding vessels of variable size. Extravasated serum proteins were frequently present surrounding vessels at anatomical boundaries with juxtaposing material properties, including the gray–white interface, subpial cortex, and adjacent to the dentate gyrus and lateral ventricles. Moreover, while there was no difference in the overall extent of BBB leakage between hemispheres, stereotyped hemispheric differences in the distribution of extravasated serum proteins suggests BBB disruption occurs in a distribution consistent with a mechanical etiology.

To determine the clinical relevance of the findings in the swine concussion model, parallel examinations were performed on post-mortem tissue following human severe TBI. As expected, due to the differences in injury severity, far more extensive and widespread FBG extravasation was found in the human TBI cases than was observed in swine concussion. Nonetheless, there were notably similarities in the pattern of BBB disruption between the two species. As found in swine concussion, a multifocal distribution of FBG extravasation was apparent, with subpial vessels frequently affected and occasional focal disruptions at the gray–white boundary were also observed in the human cases. As with the findings in swine, the location of these disruptions suggests a mechanical origin with a predominance at anatomic boundaries. Unlike the swine concussion model, however, there were additional changes typical of severe TBI in humans, including evolving hemorrhagic lesions, ischemic change and brain swelling. Moreover, human control brains also displayed some degree of serum protein extravasation, likely induced by protracted agonal or prolonged ischemic events around the time of death.

Following concussion in swine, within the white matter, there was considerable gross regional overlap of abnormal FBG immunoreactivity and axonal profiles accumulating APP. Specifically, approximately 60% of axons with an injured morphology were observed in regions with coincident FBG extravasation. This suggests that the overlapping pattern of both pathologies in white matter might reflect regions of higher mechanical injury during rotational acceleration. Since physical and computational models of swine TBI have shown both gray and white matter undergo a heterogeneous pattern of high stresses and strains, future studies could examine whether BBB disruption following in vivo injury could be used as a pathological marker for loading distribution throughout the entire brain.

The overlap of different neuropathologies agrees with biomechanics literature showing the mechanical threshold for compromise of the BBB overlaps with thresholds for axonal injury [8, 79]. In general, the direction of impact, acceleration profile, and relative size and shape of the brain all contribute to areas of maximum deformation within the brain during ‘real world’ impacts that cause concussion [22]. Thus, identification of BBB disruption may serve as a type of “injury strain gauge”, showing regions that endured the highest deformations in both the gray and white matter.

With regards to white matter, at the local scale, tissue deformations can parallel the trajectory of axons within a region, maximally stretching these axons. In this scenario, axonal injury may appear in the absence of BBB compromise. In comparison, if tissue stretch appears perpendicular to primary axonal pathways, axons are not stretched significantly and BBB compromise can occur prior to any evidence of axonal damage. Finally, when the local tissue deformation is enough to exceed both the local tolerance of the microvasculature and the tolerance of axons aligned with the deformation field, one would see the co-localization of BBB and axonal injury. We observed each of these three injury phenotypes in our swine concussion model. The process of accounting for the microanatomy of the brain in predicting injury risk is now appearing in computational models of the brain to predict concussion injury risk [29, 40, 64], and our observations reinforce the need to accurately align the biomechanical features of injury with the anatomy. In the future, we expect these computational efforts will evolve to include estimates of damage to the brain network structure [47], which would significantly enhance our ability to connect the initial injury to the consequences of the injury over time.

Intriguingly, in addition to leaking into the brain parenchyma, serum proteins were also found to be taken up by local cells. Perivascular clusters of both neurons and astrocytes immunoreactive for FBG and IgG were observed following both concussion in swine and severe TBI in humans. Similar cellular uptake of serum proteins, including FBG, has been described in other disorders where BBB dysfunction features, including cerebrovascular disease [15, 90], epilepsy [53], multiple sclerosis [7, 45, 48], cerebral malaria [16], acquired immunodeficiency syndrome [65], HIV encephalitis [20] and Alzheimer’s disease [72, 90, 92]. Surprisingly, there was minimal uptake of serum proteins by microglia when compared with astrocytes and neurons, which may reflect cell-specific uptake for the clearance of serum proteins after TBI. Notably, astrocytes have previously been demonstrated to selectively clear FBG in vitro [38]. Since neurons staining for FBG following concussion appeared morphologically normal at the time points examined in the present study, further temporal studies would be required to examine if this uptake has pathological consequences.

The specific mechanisms by which the extravasation of serum proteins in TBI are potentially neurotoxic remain to be explored. However, in addition to its role in hemostasis, FBG is increasingly recognized for its additional and diverse contribution to CNS pathology via mediation of inflammatory processes [4, 5], astrocyte activation [75] autoimmunity and demyelination [6, 73]. Moreover, FBG has been demonstrated to inhibit CNS neurite outgrowth in vitro and in vivo [74], and thus may be a potential barrier to axonal recovery in concussion, where white matter injury is a primary pathology [12,13,14, 21, 55, 59, 62, 80, 94, 95]. Interestingly, exposure of the brain to serum proteins has been speculated as a potential mechanism driving age-related tau astrogliopathy (ARTAG), a pathological accumulation of abnormally phosphorylated tau in astrocytes that occurs in a subpial, subependymal and perivascular distribution [46]. However, the presence of ARTAG following TBI has yet to be characterized.

In addition to the exploration of potential downstream neuropathological consequences of serum protein extravasation, it will be critical to examine the pathophysiological mechanisms that drive BBB dysfunction in concussion. This may include the immediate mechanical failure of the barrier, potentially caused by direct disruption of endothelial tight-junctions. In contrast, downstream pathophysiological alterations to the BBB, including altered transcellular transport, will also be important to examine since it has been demonstrated following an impact model of TBI in cats [67]. Interestingly, the matrix metalloproteases (MMPs), including MMP-9 in particular, are increasingly recognized as mediators of BBB permeability following TBI and their role in concussion should also be determined [50, 60, 78, 96].

Overall, we demonstrate that acute BBB disruption is a primary consequence of head rotational acceleration in a swine model of concussion. This observation could help direct non-invasive diagnosis of concussion, such as advanced neuroimaging techniques to identify the extent and distribution of BBB leakage [91]. In addition, characterizing the nature and time course of BBB permeability may have direct relevance for the development of serum biomarkers for concussion [44, 81], and for the timing of delivery of potential pharmacological interventions.