In traumatic brain injury mechanical forces applied to the cranium and brain cause irreversible primary neuronal and astroglial damage associated with terminal dendritic beading and spine loss representing acute damage to synaptic circuitry. Oedema develops quickly after trauma, raising intracranial pressure that results in a decrease of blood flow and consequently in cerebral ischaemia, which can cause secondary injury in the peri-contusional cortex. Spreading depolarizations have also been shown to occur after traumatic brain injury in humans and in animal models and are thought to accelerate and exacerbate secondary tissue injury in at-risk cortical territory. Yet, the mechanisms of acute secondary injury to fine synaptic circuitry within the peri-contusional cortex after mild traumatic brain injury remain unknown. A mild focal cortical contusion model in adult mouse sensory-motor cortex was implemented by the controlled cortical impact injury device. In vivo two-photon microscopy in the peri-contusional cortex was used to monitor via optical window yellow fluorescent protein expressing neurons, enhanced green fluorescent protein expressing astrocytes and capillary blood flow. Dendritic beading in the peri-contusional cortex developed slowly and the loss of capillary blood flow preceded terminal dendritic injury. Astrocytes were swollen indicating oedema and remained swollen during the next 24 h throughout the imaging session. There were no recurrent spontaneous spreading depolarizations in this mild traumatic brain injury model; however, when spreading depolarizations were repeatedly induced outside the peri-contusional cortex by pressure-injecting KCl, dendrites undergo rapid beading and recovery coinciding with passage of spreading depolarizations, as was confirmed with electrophysiological recordings in the vicinity of imaged dendrites. Yet, accumulating metabolic stress resulting from as few as four rounds of spreading depolarization significantly added to the fraction of beaded dendrites that were incapable to recover during repolarization, thus facilitating terminal injury. In contrast, similarly induced four rounds of spreading depolarization in another set of control healthy mice caused no accumulating dendritic injury as dendrites fully recovered from beading during repolarization. Taken together, our data suggest that in the mild traumatic brain injury the acute dendritic injury in the peri-contusional cortex is gated by the decline in the local blood flow, most probably as a result of developing oedema. Furthermore, spreading depolarization is a specific mechanism that could accelerate injury to synaptic circuitry in the metabolically compromised peri-contusional cortex, worsening secondary damage following traumatic brain injury.
The end-stage damage in traumatic brain injury (TBI) is attributed to both primary and secondary brain injury. Primary injury results from mechanical forces applied to the cranium and brain. The main insult activates multiple factors that further increase damage in the peri-contusional cortex (Graham et al., 2000). These factors include increased intracranial pressure, cytotoxic/vasogenic oedema, spreading depolarizations, elevated extracellular potassium, calcium dysregulation, excitotoxicity, free-radical production and inflammatory processes, to name a few of the most often studied mechanisms (Greve and Zink, 2009). Cerebral ischaemia has been long thought of as one of the secondary injury mechanisms of TBI (Bouma et al., 1991), because an increase in intracranial pressure from tissue oedema results in a loss of blood flow (Nortje and Menon, 2004). Much like the focal stroke, damage after TBI appears to be exacerbated by recurrent propagating waves of spreading depolarization in ∼55% of human patients (Hartings et al., 2009; Dreier, 2011; Hartings et al., 2011b; Lauritzen et al., 2011). Moreover, spreading depolarization-induced intracellular calcium increase and glutamate release indicate that excitotoxicity during TBI may accomplish detrimental effects through the spreading depolarization route (Obrenovitch and Urenjak, 1997).
Since most excitatory synapses occur on the dendritic arbor, dendrites were predicted to be the initial site of excitotoxic injury leading to neuronal damage and death (Bindokas and Miller, 1995). Indeed, sustained high calcium levels that were shown to develop in distal dendrites upon excitotoxic insult are capable of slowly spreading to the soma, ultimately resulting in acute neuronal death (Shuttleworth and Connor, 2001; Vander Jagt et al., 2008). Spreading depolarization recorded in vivo under conditions of severe metabolic compromise, such as during global ischaemia, rapidly damages fine synaptic circuitry (Murphy et al., 2008) and in the absence of reperfusion, as seen in the ischaemic core, dendrites remain terminally beaded and spines are lost (Zhang and Murphy, 2007; Risher et al., 2010). Concurrently, terminal dendritic beading is indicative of an irreversible acute neuronal injury (Hori and Carpenter, 1994; Risher et al., 2010), and it is an early sign of cell death pathway activation (Enright et al., 2007). Spreading depolarizations facilitate injury to fine synaptic circuitry within the ischaemic penumbra (Risher et al., 2010), but the real-time evolution of the spatiotemporal damage to synaptic circuitry in the peri-contusional cortex is unknown and the underlying mechanisms are not clear. Here, we hypothesized that the expansion of secondary injury to synaptic circuitry in the peri-contusional cortex after mild TBI depends on the availability of local blood flow (i.e. the degree of ischaemia) with spreading depolarizations exacerbating damage.
Controlled focal non-penetrating deformation of the cortex has been widely used to induce TBI in rodents (Dixon and Kline, 2009). However, for in vivo imaging, one of the biggest obstacles to monitoring the development of secondary damage at the cellular level is bleeding on the brain surface after TBI. Hence, we have adapted and modified this highly reproducible controlled cortical impact injury model to induce a mild trauma with reduced subdural brain haemorrhage that made it suitable for high-resolution 2-photon laser scanning microscopy (2PLSM) imaging. Our data suggest local cerebral ischaemia as one of the chief mechanisms of the secondary injury to fine synaptic circuitry in the peri-contusional cortex with spreading depolarizations greatly accelerating acute damage.
Materials and methods
All procedures followed National Institutes of Health guidelines for the humane care and use of laboratory animals and underwent annual review by the Animal Care and Use Committee at the Georgia Health Sciences University. Founders of the B6.Cg-Tg(Thy1-YFPH)2Jrs/J colony (YFP-H) were purchased from Jackson Laboratories. YFP-H mice display bright fluorescence in a fraction of pyramidal neurons of the neocortex, thus facilitating in vivo 2PLSM imaging. The wild-type littermates of YFP-H mice were also used in experiments for behavioural testing, laser Doppler blood flowmetry, in some electrophysiological experiments and for histology. Founders of the FVB/N-Tg(GFAP-EGFP)GFEA-FKi colony (GFAP-EGFP) were kindly provided by Dr. H. Kettenmann (Max Delbruck Centre for Molecular Medicine, Berlin, Germany). Mice of this strain display bright fluorescence in astrocytes from multiple areas of the CNS. In total, 40 YFP-H, eight GFAP-EGFP and 12 wild-type adult male and female mice of average age ∼7 months were used in this study.
Preparation of mice for in vivo imaging
Surgical procedures for cranial window followed a standard protocol used previously (Risher et al., 2010). For acute imaging, anaesthesia was induced with an intraperitoneal injection of urethane (1.5 mg/g body weight), or intraperitoneal injection of Avertin (0.025 mg/g body weight) for multi-day imaging sessions or behavioural testing. Body temperature was maintained at 37°C with a heating pad (Sunbeam). A short ∼1 cm L-shaped glass capillary (1.2 mm diameter) was inserted into the trachea and secured with two sutures in mice used for acute imaging. The skull was exposed after a midline incision. The dental drill (Midwest Stylus Mini 540S) with one-quarter bit was used to thin the circumference of a 4 mm diameter circular region of the skull centred at stereotaxic coordinates −1.5 mm from bregma and 2.0 mm lateral over sensory-motor cortex. The thinned bone was lifted up with forceps and an optical chamber was constructed by covering the intact dura with a thin layer of 1.5% agarose prepared in a cortex buffer containing (in mM) 135 NaCl, 5.4 KCl, 1 MgCl2, 1.8 CaCl2, 5 HEPES, pH 7.3. The optical chamber was left open for later TBI induction. A small aluminium bar with two tapped screw holes was glued to the skull with dental acrylic cement (Co-Oral-Ite Dental) and then the mouse head was stabilized by screws tightened to a custom-made L-shaped adjustable metal arm affixed to the baseplate mounted on the Luigs and Neumann microscope stage for imaging. Rectal temperature was monitored continuously and maintained at 37°C with a heating blanket (Harvard Apparatus). Blood oxygen saturation level was assessed with MouseOx® pulse oximeter (STARR Life Sciences) mounted on the left thigh to ensure O2 saturation remained >90% for the duration of imaging, indicating that mice were respiring properly. Depth of anaesthesia was gauged by the lack of a toe-pinch reflex and heart rate (450–650 beats/min) monitoring with MouseOx® pulse oximeter and maintained with only minimal supplementation (<10% of the initial urethane dose) for up to 8 h. Hydration was maintained by intraperitoneal injection of 0.9% NaCl (200–300 µl) with 20 mM glucose at 1–2 h intervals. A 0.1 ml bolus of 5% (w/v) Texas Red® Dextran (70 kDa) (Invitrogen) in 0.9% NaCl was injected into the tail vein for blood flow visualization. All chemicals were from Sigma Chemical unless indicated otherwise.
Traumatic brain injury model
Mild TBI was induced by the controlled cortical impact injury device (Pittsburgh Precision Instruments) following a protocol adapted from Dixon and Kline (2009). The baseplate containing head holder and mouse was moved from the microscope stage to the controlled cortical impact injury device. Agarose covering the cortex was removed and the pneumatic piston containing a 2.0 mm diameter flat edge impactor tip was adjusted to allow for an impact normal to the cortical surface. To reduce a chance of inaccuracy while zeroing the impactor tip on the dura, the tip was positioned to contact the dura under the high magnification of the Zeiss Stemi SV6 stereo zoom microscope using the impact depth adjustment vernier with the margin of error equal to 5 μm. The cortex was struck at 5.0 m/s to a depth of 0.1 mm with impact duration of 100 ms and the dura was then carefully removed. The cortex was covered with 1.5% agarose prepared in a cortex buffer and sealed by a circular glass coverslip (Bellco, #1943-00005) using the dental cement to prevent any shifting of the cover-slip due to brain swelling.
Two-photon laser scanning microscopy
Images were collected with infrared optimized water-immersion objective ×40/0.80 NA (Carl Zeiss), using the Zeiss LSM 510 NLO META multiphoton system mounted on the motorized upright Axioscope 2FS microscope. The scan module was directly coupled with the Spectra-Physics Ti:sapphire broadband mode-locked laser (Mai-Tai) tuned to 910 nm for two-photon excitation. Emitted light was detected by internal photomultiplier tubes of the scan module with the pinhole entirely opened. The imaged dendrites and astrocytes were typically within 100 μm of the pial surface and therefore in layer I. To monitor structural changes, 3D time-lapse images were taken at 1.0-μm increments, using ×3 optical zoom, resulting in a nominal spatial resolution of 13.65 pixels/μm (12 bits/pixel, 1.26 μs pixel time) across a 75 × 75 µm imaging field or at 2.0-μm increments with ×0.7 optical zoom, resulting in a nominal spatial resolution of 3.19 pixels/μm (8 bits/pixel, 2.51 μs pixel time) across a 318 × 318 µm imaging field. We were able to successfully image the same individual astrocytes before and after TBI although the entire peri-contusional area experienced dramatic swelling. This was achievable due to the relatively large size of astrocytes and sparse EGFP expression labelling only a small fraction of astroglia. Individual image stacks of astrocytes were taken before and 0.5 h after TBI and time lapse imaging continued in 0.5 h intervals. It was more technically challenging to find the same dendrites before and after injury in the neuropil with high density of yellow fluorescent protein (YFP) labelled dendrites that were shifting out of register in 2D space due to tissue swelling. Hence, first we verified that in general, dendrites were uninjured and appeared healthy before TBI and then we obtained the image stacks of particular dendrites at 0.5 h after TBI and time lapse imaging was performed in 1 h intervals starting 1 h after TBI. When shifting of the focal plane occurred, the field of focus was adjusted and re-centred before acquiring the next image stacks (Risher et al., 2010). The blood flow was imaged in a repetitive line scan mode (1000 lines per scan) along the central axis of a capillary (1.15 ms/line, 0.2 µm/pixel, 8 bits/pixel, 4.57 μs pixel time). Data acquisition was controlled by Zeiss LSM 510 software.
An LSM 510 Image Examiner (Zeiss) was used together with NIH ImageJ for 2PLSM image analyses and processing. A median filter (radius = 2) was applied to images in Figs 3 and 4 to reduce photon and photomultiplier tube noise. The Scientific Volume Imaging Huygens Professional image deconvolution software was used to process images of dendrites and astrocytes in Figs 5–8. Dendritic beading was identified as the appearance of rounded regions extending beyond the diameter of the parent dendrite separated by ‘interbead’ segments. Dendritic recovery was defined as the disappearance of rounded ‘beaded’ regions. Progression of dendritic injury was assessed by quantifying the amount of dendritic beading in an imaging field following published protocols (Murphy et al., 2008). Briefly, image stacks were converted to maximum intensity projections and then the percentage of beaded dendrites was scored in 20 µm squares that were superimposed over the images using ImageJ. Each box was scored as beaded or not. Only boxes with dendrites in them were counted.
Because axial resolution of 2PLSM is relatively poor (∼2 µm), small astroglial processes were not readily measurable and therefore, volume changes of fine astroglial processes were not quantified. To simplify the interpretation of morphometric data imposed by the lower axial resolution of 2PLSM as compared with the lateral resolution (∼0.4 µm), we used 2D maximum intensity projections of image stacks (∼20-µm thick) to assess relative changes in the cross-sectional soma area of individual astroglial cells as was described previously (Andrew et al., 2007; Risher et al., 2009, 2012). This analysis of maximum intensity projection images assumed that astroglial soma volume is changing uniformly in all directions, as based on viewing astrocytes along the z-axis in control and at each experimental time point. Such morphometric analysis of changes in the lateral dimensions was adequate to determine relative volume changes, which underestimated the actual volume changes assuming they are approximately isotropic. Therefore, as in our previous studies, three techniques were used for analyses of astrocytes: (i) maximum intensity projection images were digitally traced by hand to measure the area of astroglial soma profiles in control and for each time point. Alternatively, maximum intensity projection images were filtered with a median filter (radius = 1), background subtracted, and thresholded using ImageJ. The thresholded image with the astrocytic somata was then outlined and automatically measured with the ‘analyze particles’ function of ImageJ. The relative change in soma profiles traced either manually or automatically yielded similar results; (ii) control and experimental maximum intensity projection images were pseudocoloured red and green, aligned and overlaid to reveal different areas that remained green in contrast to yellow overlapping areas; and (iii) control outlines were created and filled to create a mask which showed areas of swelling when overlaid on post-impact images. The velocity of red blood cells was measured from a space-time image formed by a stack of line scans along the central axis of a capillary where moving red blood cells are seen as dark bands (Kleinfeld et al., 1998). The sign and the inverse of the slope of these bands reflect the direction of flow and speed. The velocity of red blood cells is calculated as (Δx/Δt), where Δx is the average distance between bands at a fixed time and Δt is the average time between bands at a fixed position.
Cerebral blood flow measurements
Two-dimensional maps of cerebral blood flow with high spatiotemporal resolution were acquired by laser speckle imaging as described elsewhere (Dunn et al., 2001; Risher et al., 2010). Briefly, the cortical surface was illuminated through Edmund Optics anamorphic beam expander by a 785 nm StockerYale laser at an angle of ∼30° and imaged with a ×4/0.075 NA objective (Zeiss). The Zeiss AxioCam MRm CCD camera controlled by AxioVision software (Zeiss) was used to capture 150 images at 13 Hz with 20 ms of exposure time. Individual images of variance were created from raw speckle images by using 2D variance filtering (3 × 3 pixel kernel size, 3.23 µm/pixel) function of ImageJ. Individual 32-bit images of the standard deviation were calculated by taking the square root of the variance images. An image of laser speckle contrast (k) was obtained by dividing the standard deviation image by the mean of each raw image and then averaging the stack to obtain a single image.
Speckle correlation time values (τc) were used to quantify relative changes in cerebral blood flow velocity as described elsewhere (Dunn et al., 2001; Ayata et al., 2004; Risher et al., 2010). Briefly, speckle contrast (k) images were converted to speckle correlation time images (τc) using the equation τc ≈ 2Tk2, where T is the exposure duration of the camera (Tom et al., 2008). The percentage of baseline cerebral blood flow was computed by calculating the ratio of a baseline image of τc values before TBI with images of τc values at different time points after TBI (Dunn et al., 2001; Ayata et al., 2004).
Conventional laser Doppler recordings of cerebral blood flow were conducted in wild-type mice using the PeriFlux 5000 system (Perimed PF5010 laser Doppler perfusion monitoring unit) equipped with small (1 mm diameter) Probe 407-1 (Perimed) and PeriSoft analysis software. The probe was positioned at stereotaxic coordinates −1.0 mm from bregma and 0.7 mm lateral to contact a cover-slip sealing the optical chamber and control regional cerebral blood flow was measured to obtain baseline values. After controlled cortical impact, craniotomy was sealed again and cerebral blood flow was monitored continuously for 6 h. Cerebral blood flow at 30 min time intervals were calculated from average values in 1 min.
Electrophysiological recordings and spreading depolarization induction
The cortical slow direct current potential was recorded with a glass microelectrode (filled with 0.9% NaCl, 1–2 MΩ) inserted through a small opening in a cover-slip into the peri-contusional area to the site of imaged dendrites within layer I of cortex. The Ag/AgCl pellet reference electrode (A-M Systems) was installed under the skin above the nasal bone. Spreading depolarization was induced by pressure injecting ∼50 nl of 1 M KCl with Picospritzer (Parker Instruments) through a micropipette (1 MΩ) inserted to the depth of 200–300 μm within ∼2 mm away from the site of imaging. Signals were recorded with MultiClamp 700B amplifier, filtered at 1 kHz, digitized at 10 kHz with Digidata 1322A interface board, and analysed with pClamp 9 software (Molecular Devices).
To assess vestibulomotor function beam walk tests were performed at four different time points; before TBI, and 1, 3 and 7 days after TBI. The procedures for beam walk test followed the protocol adapted from Feeney et al. (1982). Mice were trained to escape a bright light and loud white noise by traversing a narrow wooden beam (50 cm length, 6 mm wide) to enter a darkened glove box at the opposite end of the beam. Performance was rated on a 7-point scale as described previously (Masuda et al., 2010): 1, unable to traverse the beam and unable to place the left hindlimb on the horizontal surface of the beam; 2, unable to traverse the beam, but places the left hindlimb on the horizontal surface of the beam; 3, traverses the beam while dragging the left hindlimb; 4, traverses the beam and at least once places the left hindlimb on the horizontal surface of the beam; 5, traverses the beam and uses the left hindlimb in less than half of its steps along the beam; 6, traverses the beam and uses the left hindlimb in more than half of its steps along the beam; 7, traverses the beam with no more than two foot slips. Training trials were continued until the mouse traversed the beam with a score of 7 on three consecutive trials, after which sham operation or controlled cortical impact injury was initiated a day later. The data for each daily session consisted of the best of three trials.
After TBI at 8, 24, 48 and 72 h a subset of mice were anaesthetized with urethane and perfused intracardially with 25 ml cold PBS, followed by 25 ml of 4% paraformaldehyde (Electron Microscopy Sciences) in phosphate buffer. Brains were removed and post-fixed in the same fixative for 36 h. Coronal sections were cut at a thickness of 100 μm with a vibrating blade microtome (VT1000S, Leica Instruments), mounted with distilled water onto gelatin-coated slides, dried overnight and then hydrated through graded alcohols to distilled water. For haematoxylin and eosin staining, sections were stained with haematoxylin (Fisher Scientific) for 15 min and then rinsed in distilled water for 20 min. Sections were then stained with eosin-phloxine solution (Thermo Fisher Scientific) for 2 min, dehydrated, cleared with xylene (Fisher Scientific) and mounted on the microscope slides for further examination using the Aperio ScanScope CS digital slide scanner with the Olympus ×20/0.75 NA Plan Apo objective. For Fluoro-Jade staining (wild-type mice), sections were placed into a solution of 0.06% KMnO4 for 15 min, rinsed in distilled water, and then stained with 0.001% Fluoro-Jade solution (Histo-Chem) for 30 min. Sections were then rinsed in distilled water, dried, cleared with xylene and mounted on the microscope slides with DPX mounting media (Aldrich) for further examination using the Zeiss LSM 510 system with the Zeiss 10×/0.30 NA objective.
SigmaStat (Systat) was used for statistical analyses. A paired two-tailed t-test and one-way repeated measures ANOVA followed by the Tukey or the Student Neuman–Keuls post hoc tests were used to compare group means for parametric data. Kruskal–Wallis ANOVA or Friedman one-way repeated measures ANOVA on Ranks followed by the Tukey or the Student Neuman–Keuls post hoc tests were used to compare the median values for non-parametric data. The linear regression analysis was applied to quantify the strength of the relationship between two variables. The significance criterion was set at P < 0.05. Data are presented as mean ± standard deviation.
Mild tramatiuc brain injury induced by a non-penetrating localized deformation of the cortex creates an area suitable for real-time in vivo two-photon imaging
While previous histological studies have investigated post-traumatic injury after controlled cortical impact contusion (Kochanek et al., 1995; Smith et al., 1995; Hall et al., 2005, 2008), individual dendrites and astrocytes have not been monitored sequentially in the peri-contusional cortex to reveal the development of damage to synaptic circuitry over time. Using 2PLSM imaging at the cellular level of resolution we investigated dynamic recruitment of the peri-contusional cortex into the injury territory. Because bleeding on the brain surface was too great to allow imaging after 750–1000 µm tissue deformation depth used in murine models of severe controlled cortical impact injury (Smith et al., 1995; Hall et al., 2005; Deng et al., 2007; Hall et al., 2008; von Baumgarten et al., 2008), we decreased the cortex compression depth to 100 µm, which significantly reduced bleeding and enabled imaging in the peri-contusional cortex adjacent to the site of the primary injury (Fig. 1A). Imaging and electrophysiological recordings were started 0.5 h after the controlled cortical impact. This delay was necessary to ensure high quality assembly of the optical chamber after cerebral contusion. As assessed with electrophysiological recordings in 18 animals, there were no recurrent spontaneous spreading depolarizations occurring in the peri-contusional cortex between 0.5 and 6 h after the primary impact (Fig. 1B). Histological assessment with haematoxylin and eosin and Fluoro-Jade staining showed that the tissue damage was localized in the cortex (Fig. 1C and D), suggesting that this TBI model is milder than other murine controlled cortical impact injury models (Smith et al., 1995; Hall et al., 2005; Deng et al., 2007; Hall et al., 2008). Yet, the scores of the beam walk test revealed a significant motor deficit at 7 days post-TBI (Fig. 1E), indicating that this model can be used to investigate the development of secondary damage in mild TBI.
To quantify changes in regional cerebral blood flow, we conducted laser Doppler measurements with a small probe (1 mm diameter) positioned over the peri-contusional cortex. Regional cerebral blood flow dropped to 74.3 ± 8.2% of the control value at 0.5 h post-TBI (Fig. 2A, P < 0.05) and remained significantly decreased during the next 6 h. Fixed coronal sections collected from animals perfused at 8 h after TBI also revealed that Texas Red dextran, which was injected to label blood plasma before injury, was trapped in the stalled vessels surrounding the site of the primary impact (Fig. 2B). These data indicated the existence of an ischaemic zone in the peri-contusional cortex. Two-dimensional maps of cerebral blood flow with high spatiotemporal resolution acquired by laser speckle imaging validated blood flow loss at the site of the primary injury as well as confirmed progressive hypoperfusion in the peri-contusional cortex extending 0.5–1 mm outside the primary impact area (Fig. 2C–F). We therefore used laser speckle imaging to delineate the peri-contusional cortex and to guide 2PLSM imaging in all experiments. We conclude that in this mild TBI mouse model, the peri-contusional cortex can be precisely anatomically localized and compared between animals to study the spatiotemporal development of secondary injury to synaptic circuitry.
Dendritic injury in peri-contusional cortex evolves relatively slowly
With the stalled vessels in the peri-contusional cortex, it seemed plausible that acute terminal ischaemic damage to dendrites would develop in the region over time. We imaged the peri-contusional cortex to track changes in dendritic structure for up to 8 h post-impact. Surface cortical tissue in the peri-contusional area retained a relatively normal appearance immediately after impact with a small amount of bleeding in the cortex (Fig. 3A). Two-photon laser scanning microscopy imaging with low magnification over a large field surrounding the transition from damaged to relatively intact tissue showed that the peri-contusional cortex contained many dendrites with a normal appearance at 0.5 h after TBI (Fig. 3B). However, most of the dendrites had beaded by 4 h after impact (Fig. 3B). Time lapse higher magnification 2PLSM imaging over the border of transition between normal and damaged dendrites revealed relatively slow progressive dendritic beading during the next 6 h after TBI with the characteristic ‘beads-on-a-string’ appearance of damaged dendrites (Fig. 3C). Quantification of 2PLSM image series in five animals confirmed that not all dendrites in the peri-contusional cortex are immediately damaged but are susceptible to terminal injury developing over time (Fig. 3D).
Blood flow loss precedes dendritic beading in peri-contusional cortex
It has been demonstrated that without spreading depolarizations the development of dendritic beading in the metabolically compromised ischaemic penumbra is gradual and depends on the availability of local blood flow (Zhang and Murphy, 2007; Risher et al., 2010). Hence, it is conceivable that after mild controlled cortical impact contusion, the post-traumatic dendritic injury in the metabolically compromised peri-contusional cortex is similar to the relatively slow developing dendritic injury shown in the ischaemic penumbra in the absence of spreading depolarizations. Furthermore, it has been shown that individual flowing blood vessels can support intact dendritic structure within ∼80 μm into the ischaemic territory (Zhang and Murphy, 2007). Thus, we determined how the presence of a nearby flowing or non-flowing vessel will affect synaptic circuitry outcome in the peri-contusional cortex over time. In the example shown in Fig. 4A, dendrites were normal and the blood vessel was flowing at 1 h after TBI, as measured with a line scan along the vessel (Fig. 4B and C). Complete loss of blood flow at 2 h after TBI (Fig. 4B and C) was paralleled by the beginning of dendritic beading that developed gradually during the next 2 h (Fig. 4A). Another image in Fig. 4D taken over the large field shows the transition from injured dendrites at the site of the primary impact to mostly normal appearing dendrites in the peri-contusional cortex but all the blood vessels were already stalled in the entire imaging field. Under this condition of severe hypo-perfusion dendritic beading developed within the next 45 min (Fig. 4D). These observations quantified in Fig. 4E confirmed near complete beading of all dendrites in the imaging field within the next 1 h after loss of blood flow.
Spreading depolarizations accelerate acute dendritic injury in the metabolically compromised peri-contusional cortex
Stroke-induced spreading depolarizations result in dendritic beading in metabolically compromised tissue that is reversible if neurons are not severely energetically compromised (Murphy et al., 2008; Risher et al., 2010). Yet, recurrent spreading depolarizations increase metabolic stress in the ischaemic penumbra greatly accelerating damage to synaptic circuitry even in the presence of nearby flowing blood vessels (Risher et al., 2010). Here, we determined whether spreading depolarizations could augment and facilitate acute dendritic injury in a metabolically compromised peri-contusional cortex. As there were no spontaneous spreading depolarizations in this mild TBI model we have elicited a series of spreading depolarizations by pressure injecting ∼50 nl of 1 M KCl far away from the site of imaging. Rapid dendritic beading was observed when the first spreading depolarization invaded the peri-contusional cortex (Fig. 5A). The beaded dendrites recovered during repolarization, but the subsequent three rounds of spreading depolarization resulted in the terminal injury to synaptic circuitry despite blood continuing to flow within nearby vessels (Fig. 5A). Quantification of 2PLSM image series in five animals confirmed that the fraction of beaded dendrites increased from 23.5 ± 15.6% before to 96.2 ± 5.7% during the first induced spreading depolarization (Fig. 5B). However, dendrites swiftly recovered after passage of the first spreading depolarization with a fraction of beaded dendrites in the imaging field returning to 23.8 ± 13.0% at 75 ± 17 s of repolarization. Likewise, the fraction of beaded dendrites increased to 99.2 ± 1.8%, 100.0 ± 0.0% and 99.4 ± 1.4% during the passage of the second, third and fourth spreading depolarization, respectively. Concurrently, these subsequent rounds of spreading depolarizations increased the fraction of terminally beaded dendrites to 61.2 ± 13.8%, 73.7 ± 18.7% and 84.8 ± 22.9% as measured at 101 ± 17 s, 74 ± 33 s and 115 ± 95 s during repolarization, respectively. These findings strongly indicate that accumulated stress caused by spreading depolarizations accelerated acute dendritic injury in the peri-contusional cortex.
Spreading depolarizations do not cause accumulating dendritic injury in healthy neocortex
Spreading depolarization-induced dendritic morphological alterations in the normoxic cortex are equivocal. Indeed, it was reported that dendrites can undergo different degrees of structural change during the passage of normoxic spreading depolarization, ranging from no changes to various degrees of dendritic beading (Takano et al., 2007). Here, we examined whether normoxic KCl-induced spreading depolarizations could result in dendritic disruption with accumulating dendritic damage in normoxic healthy neocortex compared with metabolically compromised peri-contusional cortex. Similarly to experiments with the mild TBI model, we elicited four rounds of spreading depolarizations in five healthy mice by pressure-injecting ∼50 nl of 1 M KCl far away from the site of imaging. Rounds of rapid beading and recovery were observed when subsequent spreading depolarizations invaded the site of imaging in four of five animals, but there was no accumulating dendritic injury as all dendrites quickly recovered during repolarization (Fig. 6A). Furthermore, dendrites were remarkably stable during the passage of spreading depolarization in one mouse (images not shown). Quantification of 2PLSM image series in five animals confirmed that during passage of four subsequent spreading depolarizations the fraction of beaded dendrites reached 74.0 ± 41.9%, 70.0 ± 43.7%, 60.8 ± 41.6% and 61.7 ± 46.3%, respectively (Fig. 6B). Importantly, dendrites recovered to pre-spreading depolarization level as measured at 87 ± 24 s, 92 ± 58 s, 52 ± 40 s and 58 ± 42 s during repolarization, respectively. Thus, based on quantifications presented here we conclude that spreading depolarizations do not carry risk of accumulating dendritic injury in normoxic neocortex (Fig. 6B) whereas dramatically accelerating terminal dendritic injury in metabolically challenged peri-contusional cortex (Fig. 5B).
Persistent astroglial swelling in the peri-contusional cortex
Astroglial swelling has previously been proposed as a possible cause of injury and target for therapy after TBI (Kimelberg, 1992; Mongin and Kimelberg, 2005). Recently we have shown that in the early minutes to hours following stroke, astrocytes swell and remain swollen in the ischaemic neocortex with moderate to severe energy deficits (Risher et al., 2012). To reveal dynamic reorganization of astrocytes in the peri-contusional cortex we imaged seven astrocytes alongside blood vessels in five animals. There was a significant swelling of astroglial soma in the peri-contusional cortex at 0.5 h after the impact along with a swelling of fine astroglial processes (Fig. 7A), but the latter observation was not quantified because small processes were not readily measurable, especially along the z-axis. Furthermore, an astrocyte was swollen even in the presence of the nearby flowing vessel (Fig. 7A). The astroglial cross-section soma area increased by 30.4 ± 24.0% above the control at 30 min and by 42.3 ± 23.6% at 60 min after TBI (Fig. 7B). The blood flow in vessels nearby these astrocytes had declined by 54.3 ± 41.5% (P < 0.05) at 60 min after TBI (Fig. 7C), but the magnitude of astroglial swelling was not correlated (P = 0.6) with the degree of blood flow loss. Yet, astroglial swelling was apparent, indicating oedema. Indeed, during imaging the entire peri-contusional area experienced dramatic swelling. Additionally, in another cohort of experiments we conducted longitudinal imaging of 10 astrocytes in three animals at 1 h and 24 h after TBI. Astroglial swelling persisted for 24 h (Fig. 8A) with an average increase in the soma size by 18.5 ± 13.3% above the control at 1 h and by 26.5 ± 8.0% at 24 h (Fig. 8B), reflecting enduring oedema. Concurrently, the blood flow had declined by 69.2 ± 12.6% (P < 0.02) alongside eight astrocytes as measured at 24 h after TBI in two animals. There was no correlation (P = 0.9) between the magnitude of astroglial swelling and the degree of blood flow loss.
We applied in vivo real-time 2PLSM to the mild focal cortical contusion model to monitor the development of secondary injury in the peri-contusional cortex at the cellular level. To the best of our knowledge, these are the first experiments to examine real-time dynamics of progressive secondary injury at the level of synaptic networks after mild TBI. In the absence of spreading depolarizations, 2PLSM imaging showed slowly progressing dendritic beading in the peri-contusional cortex during the acute phase (8 h) after cortical impact, while astrocytes were persistently swollen. Simultaneous observation of blood vessels and dendrites revealed that after complete blood flow loss all dendrites in the imaging field became beaded within 1 h. When successive rounds of spreading depolarizations were induced to invade the imaging site in uninjured cortex of control mice, dendritic beading coincided with passage of depolarizations in most animals with complete recovery during repolarization. In contrast, when spreading depolarizations were induced to invade the peri-contusional cortex, rounds of dendritic beading consistently coincided with passage of spreading depolarizations but with only partial recovery during repolarization indicating accumulating terminal dendritic damage, which signified expansion of the injury into the peri-contusional cortex.
We have used a controlled cortical impact brain trauma model that is widely used to mimic a range of human contusion injuries (Dixon et al., 1991; Morales et al., 2005). A particular advantage of this model for the current study was that deformation parameters such as time, velocity and depth of the impact could be easily controlled to create a peri-contusional area suitable for 2PLSM imaging. In the preliminary set of experiments we used deformation parameters to produce severe injury (Hall et al., 2005; Deng et al., 2007), but bleeding on the brain surface was too substantial for the successful imaging. Therefore, we reduced the depth of the impact from ∼1000 μm to ∼100 μm and the impactor diameter from 3 to 2 mm to produce milder injury. Importantly, the craniotomy was sealed with a cover glass to preserve the effect of the post-traumatic increase in the intracranial pressure (Morales et al., 2005). The controlled cortical impact produced a small focal contusion while still causing measurable injury, as was validated with histological analyses and behavioural testing. Furthermore, laser speckle imaging confirmed a loss of blood flow at the site of the primary impact and revealed the presence of the hypo-perfused peri-contusional cortex adjusted to the primary injury site. Therefore, throughout the study, laser speckle imaging was used to guide 2PLSM to the peri-contusional cortex that could be precisely localized, thus facilitating high resolution in vivo imaging.
In mild and moderate models of TBI regional blood flow frequently remains above ischaemic levels (Morales et al., 2005). In our study laser Doppler flowmetry in a 1-mm diameter region adjusted to the primary impact site also showed a moderate (∼30%) decrease of regional blood flow. However, Texas Red® trapped in the vessels surrounding the site of primary injury revealed the presence of ischaemic vessels. This finding is in agreement with early studies reporting significant local reduction of blood flow at the site next to the primary impact even at mild injury levels (Yamakami and McIntosh, 1991; Dietrich et al., 1996). Two-photon laser scanning microscopy imaging of individual blood vessels further confirmed the loss of blood flow in the peri-contusional cortex, showing that our model can be used for studying ischaemia as the mechanism of secondary injury to the cellular components of synaptic circuitry in mild TBI.
At this time, the exact mechanism causing perfusion deficits in our model is unknown. Micro-thrombosis resulting from abnormal platelet activation in the early stage after TBI (Dietrich et al., 1994, 1996; Maeda et al., 1997) may lead to perfusion loss in the peri-contusional cortex (Bramlett and Dietrich, 2004). Oedema quickly develops during brain trauma, raising the intracranial pressure within a rigid cavity of the skull (Greve and Zink, 2009). Hence, an increase of intracranial pressure from tissue oedema has been thought to be one of the major mechanisms of blood flow loss (Nortje and Menon, 2004). As expected, the oedema was developing in our TBI model as evident from persistent astroglial swelling. Furthermore, during the imaging, the focal plane was slowly shifting over several hours providing an internal indicator of continuous tissue swelling. However, the intracranial pressure was not monitored in our experiments and future studies are required to establish whether increased intracranial pressure contributed to loss in blood flow. In the current study, we focused on high-resolution real-time imaging of dendrites, astrocytes and nearby blood vessels after mild TBI.
Gradual degeneration of dendritic structure was observed over a wide area 8 h after TBI, reflecting secondary injury in the peri-contusional cortex. Previously, dendritic beading has been reported in the ischaemic condition (Hori and Carpenter, 1994; Zhang et al., 2005) and without recovery of sufficient blood flow it is a reliable indicator of terminal injury to fine synaptic circuitry (Zhang and Murphy, 2007; Li and Murphy, 2008; Risher et al., 2010). Indeed, recent studies using in vivo 2PLSM imaging showed that without recurrent spreading depolarizations injury to dendritic structure is gated by the degree of ischaemia after local blood flow loss in a photothrombotic model of focal stroke (Zhang et al., 2005; Risher et al., 2010). During severe ischaemia (∼90% reduction of blood flow) dendritic structure was lost within 10–40 min, but during moderate ischaemia (∼50% reduction of blood flow) dendritic beading developed within a few hours (Zhang et al., 2005). In the photothrombotic model, dendritic structure could be maintained in the ischaemic tissue in the presence of a flowing capillary ∼80 μm away (Zhang and Murphy, 2007), and arterioles can supply oxygen even over longer distances, well above 100 μm (Kasischke et al., 2011). Photothrombotic occlusion results in the abrupt, severe loss of blood flow within the irradiated area, which is separated by a narrow hypo-perfused penumbral zone (<380 μm) from the area with normal flow (Zhang and Murphy, 2007). In our study, the hypoperfused peri-contusional area is wider (up to 1 mm). Hence, it is conceivable that a milder gradient of metabolic stress over a large area would result in a slower (over several hours) loss of dendritic structure. Yet, when blood flow was completely stalled, dendritic beading developed within the first hour, closely approximating a time course of dendritic injury during severe ischaemia (Zhang et al., 2005; Zhang and Murphy, 2007). These data taken together suggest that decreased local blood flow is responsible for the dendritic beading in the peri-contusional area after mild TBI.
Diffusion of potentially damaging substances released from the site of the primary impact and during extravasation might contribute to the acute dendritic injury in our TBI model. Indeed, leaking plasma contains 30–80 μM of glutamate (Meldrum, 2000). Surprisingly, extravasation was shown to cause little or no additional acute damage to dendrites in the photothrombotic model of stroke (Zhang and Murphy, 2007). Additionally, rupturing a single penetrating arteriole did not cause dendritic injury as demonstrated in another recent 2PLSM study (Rosidi et al., 2011). Yet, it is conceivable that large leakage of diffusible serum-derived factors (including excitotoxic glutamate) can contribute to acute dendritic injury. However, systematic measurements between cortical areas with and without extravasation were not possible in our study. Future 2PLSM experiments will be necessary to address whether extravasation contributes to acute dendritic injury in the peri-contusional cortex.
The effects of spreading depolarizations in animal models of TBI are equivocal. This could be partially explained by the complexity of primary and secondary injury of TBI in humans that animal models rarely reproduce (Lauritzen et al., 2011). In rodent models the primary insult usually results in spreading depolarization, but spontaneous spreading depolarizations are fewer with a shorter duration or do not occur (Sunami et al., 1989; Ozawa et al., 1991; Nilsson et al., 1993; Williams et al., 2005; von Baumgarten et al., 2008). Moreover, spreading depolarizations duration and their frequency vary with severity of trauma e.g. a mild injury results in short spreading depolarization similar to the migraine event, whereas a severe injury elicits longer spreading depolarization (Rogatsky et al., 1996, 2003). In humans, prolonged spreading depolarizations are associated with worse clinical outcome in brain trauma (Hartings et al., 2011b) with two delayed peaks of depolarizations at ∼24 h and on Days 6–7 (Hartings et al., 2009). In contrast, such prolonged spreading depolarizations were not reported in animal models of TBI (Lauritzen et al., 2011). Yet, the clinical data show the whole spectrum of waves and even short spreading depolarizations will impose a significant metabolic challenge (Hashemi et al., 2009; Feuerstein et al., 2010), especially when they will invade the hypoperfused peri-contusional cortex. The lack of spontaneous spreading depolarizations in our model (0.5–6 h after trauma) could be due to the mild injury that was implemented to enable imaging but longitudinal recordings were not conducted. Another limitation is that we did not record during the controlled cortical impact and in the first 30 min after trauma.
An in vivo study by von Baumgarten et al. (2008) failed to reveal the effect of spreading depolarizations on contusion volume increase assessed with cresyl violet histology at 24 h. However, single time point histological studies cannot reveal dynamic cellular reorganization over time. In contrast, using intrinsic optical signal imaging to track spreading depolarization in time and space in neocortical slice model of trauma, Church and Andrew (2005) have shown that a single spreading depolarization expands traumatic injury. In our experiments we have used KCl application at the remote site to elicit rounds of spreading depolarization to invade the peri-contusional cortex as was previously used in TBI (von Baumgarten et al., 2008) and stroke models (Back et al., 1996; Busch et al., 1996). It should be noted that there is a depolarization continuum between terminal long-lasting and short-lasting spreading depolarization and biophysical features of depolarization remain preserved as it spreads in tissue along a gradient of metabolic stress (Dreier, 2011; Dreier et al., 2013). This provides additional rationale for experiments with artificially induced spreading depolarizations as they represent phenomena of the same nature as recurrent spontaneous spreading depolarizations that could be harmful when invading the hypoperfused peri-contusional cortex.
In the experiments reported here, spreading depolarizations resulted in rapid dendritic beading within a few seconds. As in the energy-deprived ischaemic penumbra with recurrent spontaneous spreading depolarizations (Risher et al., 2010, 2011), the spread of artificially-induced depolarization waves through metabolically challenged peri-contusional cortex has greatly augmented terminal dendritic injury. These results indicate that under conditions when recurrent depolarizations occur spontaneously, their spread through energy-deprived peri-contusional cortex could be detrimental, resulting in fast terminal injury to fine synaptic circuitry.
An in vivo study by Takano et al. (2007) has shown that even under normal conditions, spreading depolarization can be accompanied by a complex pattern of dendritic beading or no beading. Therefore, we have evoked multiple KCl-induced spreading depolarizations to test if equal to the TBI model number of artificially-induced spreading depolarizations could possibly cause terminal dendritic injury in the healthy normal neocortex. The results reveal that unlike the metabolically compromised peri-contusional cortex, spreading depolarizations caused no aggregating dendritic injury in the normal neocortex. Interestingly, rapid dendritic beading and recovery coincided with rounds of spreading depolarization in four mice, but in one animal the imaged dendrites were remarkably stable during the passage of depolarization. In normal cortex spreading depolarization is the pathophysiological correlate of the migraine aura (Lauritzen, 1994), but even here it can cause a shortage of energy supply (Hashemi et al., 2009) and can be accompanied by a short period of tissue hypoxia with a complex pattern of distribution across the capillary bed (Takano et al., 2007; Yuzawa et al., 2012). This could possibly translate into a complex pattern of spreading depolarization-induced dendritic beading thus reflecting patterns of tissue hypoxia across the capillary bed with perhaps no beading occurring in areas of luxury oxygen supply in the vicinity of penetrating arterioles (Kasischke et al., 2011). Future 2PLSM experiments will be necessary to test this notion.
Recently, using high-resolution in vivo 2PLSM imaging, we have shown that astroglial soma swell by ∼21% following hypo-osmotic stress induced by intraperitoneal distilled water injection and by ∼33% following complete global ischaemia accomplished by cardiac arrest (Risher et al., 2009). Likewise in another study (Risher et al., 2012), we observed persistent astroglial swelling by ∼28% between 30–60 min of severe episode of global ischaemia induced by a bilateral common carotid artery occlusion. Furthermore, astrocytes in the metabolically compromised ischaemic penumbra showed a persistent swelling of ∼31% between 30–60 min following photothrombotic occlusion (Risher et al., 2012). This study has also revealed an early and long-lasting astroglial swelling response in the peri-contusional cortex of similar magnitude to the swelling resulting from other pathological conditions such as ischaemia and osmotic stress. Likewise with hypoperfused ischaemic penumbra, astrocytes were swollen despite the presence of a nearby flowing vessel with reduced blood flow and remained swollen at 24 h, while blood vessels continued to flow at the decreased rate. These data suggest that a dissipation of the brain oedema after TBI (i.e. astroglial swelling) could be considerably affected by the reduced blood flow in the peri-contusional cortex. Further experiments are necessary to determine how a time-dependent resolution of astroglial swelling will be affected by the degree of ischaemia defined by the reduction in the blood flow.
Swelling may be reflective of neuroprotective processes occurring in astrocytes such as extracellular glutamate and potassium uptake, increased glycogen metabolism and the release of glutathione to defend neurons from oxidative injury (Dienel and Hertz, 2005; Mongin and Kimelberg, 2005). Conversely, water influx through aquaporin 4 during swelling could elicit activation of astrocytic signalling events that may exacerbate the pathological outcome (Thrane et al., 2011). Therefore, astroglial swelling can be harmful and affect their ability to provide neuronal support (Kimelberg, 1992, 2005). Astroglial swelling could result in the release of neuroactive and excitotoxic substances such as glutamate, aspartate, ATP and d-serine (Kimelberg et al., 1990; Schell et al., 1995; Mongin and Kimelberg, 2002) leading to neuronal death (Orellana et al., 2011). Future experiments will be necessary to determine the fate of the swollen astrocytes in the peri-contusional cortex and their ability to protect neighbouring neurons.
A brief outlook: implications of the results regarding potential target for treatment in patients with traumatic brain injury
Among key factors that require immediate attention following brain trauma is the control of the spread of secondary injury from hypoxia and ischaemia (Bullock, 2007). Indeed, a third of patients with TBI and 90% of patients who die from TBI show ischaemic injury (Winchell and Hoyt, 1997; Rudehill et al., 2002; Greve and Zink, 2009). Using a preclinical model of TBI we have shown the presence of ischaemic blood vessels in the peri-contusional cortex that could lead to a dangerous ischaemic condition and therefore result in delayed ischaemic damage to synaptic circuitry. Cytotoxic oedema is one of the major sources of brain swelling after TBI (Unterberg et al., 2004) and astroglial swelling, which occurs within the first hours after head trauma in humans (Bullock et al., 1991), has been proposed as a possible cause of injury and target for therapy after TBI (Kimelberg, 1992; Mongin and Kimelberg, 2005). Here, we have observed persistent astroglial swelling after TBI indicating enduring oedema and possible disrupted neurovascular coupling as well. Hence, development of treatment strategies aimed to improve blood flow in the peri-contusional zone should target not only microthrombosis or oedema but also disordered neurovascular coupling. Among other important factors exacerbating secondary injury from TBI are spreading depolarizations, which are found to be common in patients with head trauma (Lauritzen et al., 2011). The full spectrum from short- to very long-lasting spreading depolarization waves has been recorded in the evolution of TBI in the human brain (Hartings et al., 2011b). Furthermore, patients with multiple or prolonged spreading depolarizations have very poor prognoses for recovery (Hartings et al., 2011a, b). Here, we provided direct evidence that just a few rounds of spreading depolarizations have dramatically accelerated acute injury to synaptic circuitry in the metabolically challenged peri-contusional cortex thus, pointing to spreading depolarization as a useful pharmacological target in TBI. Yet, short-lasting spreading depolarizations that invade normoxic surrounding tissue could have indirect beneficial effects as recently reviewed (Dreier, 2011). Future combined efforts between basic and clinical scientists and physicians are required to address when and under what circumstances spreading depolarizations could switch from harmful to protective.
This work was supported by National Institutes of Health Grants NS057113 (S.A.K.).
The authors thank Dr. Jianghe Yuan for his excellent technical assistance.
two-photon laser scanning microscopy
enhanced green fluorescent protein
traumatic brain injury
yellow fluorescent protein