Attenuation of tonic inhibition prevents chronic neurovascular impairments in a Thy1-ChR2 mouse model of repeated, mild traumatic brain injury

Rationale: Mild traumatic brain injury (mTBI), the most common type of brain trauma, frequently leads to chronic cognitive and neurobehavioral deficits. Intervening effectively is impeded by our poor understanding of its pathophysiological sequelae. Methods: To elucidate the long-term neurovascular sequelae of mTBI, we combined optogenetics, two-photon fluorescence microscopy, and intracortical electrophysiological recordings in mice to selectively stimulate peri-contusional neurons weeks following repeated closed-head injury and probe individual vessel's function and local neuronal reactivity. Results: Compared to sham-operated animals, mTBI mice showed doubled cortical venular speeds (115 ± 25%) and strongly elevated cortical venular reactivity (53 ± 17%). Concomitantly, the pericontusional neurons exhibited attenuated spontaneous activity (-57 ± 79%) and decreased reactivity (-47 ± 28%). Post-mortem immunofluorescence revealed signs of peri-contusional senescence and DNA damage, in the absence of neuronal loss or gliosis. Alteration of neuronal and vascular functioning was largely prevented by chronic, low dose, systemic administration of a GABA-A receptor inverse agonist (L-655,708), commencing 3 days following the third impact. Conclusions: Our findings indicate that repeated mTBI leads to dramatic changes in the neurovascular unit function and that attenuation of tonic inhibition can prevent these alterations. The sustained disruption of the neurovascular function may underlie the concussed brain's long-term susceptibility to injury, and calls for development of better functional assays as well as of neurovascularly targeted interventions.


Introduction
Mild traumatic brain injury (mTBI) is the most common type of traumatic brain injury globally, with an incidence of 369 in 100,000 [1]. Ninety percent of all TBI are mild [2,3] and many cases of mTBI are thought to go undiagnosed [3]. In addition to short-term cognitive dysfunction such as fatigue, mental clouding, and confusion, mTBI can lead to long-term neuropsychiatric and neurological deficits [4][5][6][7][8][9], with mTBI raising the probability of dementia up to 6-fold [6]. Adverse effects are compounded when mTBI occurs repeatedly [10,11]. Many organizations have consequently implemented a Ivyspring International Publisher three-strike rule in which individuals suffering mTBI for the third time are isolated from situations that previously posed risk of injury irrespectively of any symptoms [12]. Clinical assessment of mTBI is confounded by the absence of contrast in the peri-contusional (proximal to injury focus) region on structural neuroimaging [13][14][15]. Hitherto, the majority of the preclinical mTBI research effort has focused on acute neurological changes and diagnosis, leaving major gaps in our understanding of the lasting deficits after repeated mTBI. The absence of detectable neuroimaging contrast and the gaps in the understanding of chronic pathology have encumbered development of effective interventions: currently, pharmacological treatments are prescribed in the subacute and chronic stage of TBI to treat symptoms and not the root causes of dysfunction [16].
Particularly significant to understanding the lasting deficits post TBI is evaluation of the neurovascular unit, given its importance for normal brain functioning [17]; and its high dynamicity in the subacute stage post injury [18]. To date, studies of the neurovascular unit at the individual vessel level post mTBI have been limited to recordings at rest or ex vivo [19][20][21][22], leaving injury-induced changes to functional hyperemia at the single vessel level unclear. At the same time, dysregulation of functional hyperemia has been suggested as a promising therapeutic target post TBI [23]. In parallel, neuronal evaluations post-mTBI have shown reduced peri-contusional firing and attenuated evoked potentials [8,9,[24][25][26][27][28][29], and neuronal loss [20,25,27,30,31]. Attenuated neuronal activity may result from differential susceptibility of excitatory vs. inhibitory neurons to injury [32], and in turn lead to further neurovascular unit disruptions [33]. Addressing post-injury excitatory/inhibitory imbalances [8,9,27] has recently been proposed a means of correcting both neuronal and cerebrovascular disruptions [9]. Two recent studies accordingly investigated GABA-A receptor (GABAAR) inverse agonism [26,27]. In ischemia and chronic stress, beneficial effects on functional recovery have been shown [34][35][36] by GABAAR antagonism via sustained, low dose administration of L-655,708, a GABAAR inverse agonist with dose-dependent specificity [37]. Its applications in mTBI models have resulted in partial behavioural recovery and alleviation of post-injury increases in GABAergic tone in slice preparations [26,27]; however, the effects of GABAA receptor inverse agonism on in situ neurophysiological state are not known. These gaps are particularly significant given the critical role of the re-instatement of neurovascular coupling for functional recovery [17]. Finally, quantitative functional assays are key to enable reliable cross-study comparisons and synthesis of results [13,38].
Here we use simultaneous two-photon fluorescence microscopy and intracerebral electrophysiology all recordings in an optogenetic murine model of repeated mTBI to quantify the sustained effects on the neurovascular unit with low invasiveness and high spatio-temporal precision, and assess the potential for neurovascular function normalization by a disinhibitory intervention. For this study, mTBI was modeled by three closed head injuries, an intervention that has previously been shown to produce no contrast on conventional neuroimaging [39]. Neurovascular readouts were quantified in the impacted tissue region (henceforth peri-contusional tissue). Focused photostimulation of channelrhodopsin-2 was used to probe the neurovascular unit function in light of our earlier work showing this paradigm provides more robust neuronal and vascular responses than do physiological stimuli and offers greater spatiotemporal control over the stimulation, while still providing a correlate of responses to physiological stimuli [40]. We found sustained neurovascular function aberrations that parallel earlier mesoscopic scale observations in clinical populations but that have not been thus far observed in preclinical models of mTBI. The observed mTBI-induced neurovascular deficits were largely prevented via a disinhibitory treatment with L-655,708, an inverse GABAAR agonist.

Animals
All experimental procedures in this study followed the ARRIVE guidelines and were approved by the Animal Care Committee of the Sunnybrook Research Institute, which adheres to the Policies and Guidelines of the Canadian Council on Animal Care and meets all the requirements of the Provincial Statute of Ontario, Animals for Research Act as well as those of the Canadian Federal Health of Animals Act. Ninety-four adult (51 male and 43 female) Thy1-ChR2-YFP mice (wild-type Channelrhodopsin-2, ChR2, in pyramidal neurons [41], Jackson labs 007612, line 18), 3-6 months of age, were used in the study. Cohort sizes and sex ratios are summarized in Table 1.

Closed head injury surgery
The induction of closed head injury was performed by adapting a previously published protocol [25]. Mice were anesthetized in an induction chamber with 5% isoflurane in oxygen enriched medical air (30% O 2 , balance N 2 ) and transferred to a stereotaxic frame, where they were maintained under 1.5-2% isoflurane via a nose cone. Mice were subcutaneously administered Baytril (0.02 mg/mL), Marcaine (0.1 mL), and Ringer's lactate solution (0.5-1 mL of solution with 130 mM Na, 4 mM K, 1.5 mM Ca, 109 mM Cl, 28 mM Lactate, Hospira, Canada). The scalp was shaved and following sterile surgical procedures, an incision was made in the scalp to expose the skull surrounding the impact site at +1mm ML, +0.5 mm AP. Connective tissue was removed and an impact delivered at the aforementioned location via a stereotaxic impactor (Leica Cortical Impactor, Leica, US) using the following parameters: 1.5 mm tip diameter, 2 m/s impact speed, 1 mm impact depth and 200 ms impact dwell time (corresponding to 359 kPa). Following the impact, the scalp was sutured, Ringer's lactate solution re-administered, and Betadine reapplied over the incision site. Mice were then monitored and allowed to recover in a recovery cage on a warming pad. After approximately 30 min, fully awake and alert animals were returned to their home cage in the colony. Each mouse was subjected to three cortical impacts, delivered with an inter-impact interval of 72 h. A 9% mortality rate was observed in the 47 mice that underwent impact procedures and in vivo experiments, due to skulls cracking in three mice as well as one incident of poor recovery from surgery. Sham animals underwent the same surgical procedures but with no impact delivery.

Osmotic pump implant surgery
Osmotic pump implant surgeries were performed 3 days following the third closed head injury surgery. The 3-day delay to start of L-655,708 administration was chosen based on previous reports of the timing of hyper-/hypoexcitability shifts occurring in the days following CNS damage [8,9,34,35] Osmotic pumps (Alzet micro-osmotic pump Model 1002, Durect, USA) were prepared one day prior to implantation following established procedures [35]. Pumps were loaded with 5 mM L-655,708 (Bio-techne, USA) in 50% DMSO/50% saline to be delivered at a dose of 200 μg/kg/day over 12 days, previously described as a dose level for GABAAR ɑ5-subunit specificity with significant effects counteracting increased tonic inhibition [35,37,42] in the days following injury [8,9]. The dosing was set to both significantly affect the ɑ5-subunit containing GABAARs and avoid convulsive behaviour [35,37]. Vehicle pumps were loaded with 50% DMSO/50% saline. Loaded pumps were incubated at 37°C overnight prior to surgery. On the day of the pump implantation surgery, mice were anesthetized in an induction chamber under 5% isoflurane and transferred to a stereotaxic frame with auxiliary ear bars for head fixation where they were maintained under 1.5-2% isoflurane delivered via the nose cone. Mice were given subcutaneous Baytril (0.05 mg/mL), Marcaine (0.1 mL), and Ringer's lactate solution (1 mL). The skin behind the right shoulder was shaved and cleaned with alcohol wipes and Betadine and ocular gel applied to the eyes. Following sterile surgical procedures, an incision was made behind the right shoulder and subcutaneous space was gently cleared through the incision with blunt surgical clamps. Pumps were then placed subcutaneously through the incision such that they were fully enclosed under the skin. The incision was then sutured, Betadine reapplied and another 1 mL of Ringer's lactate solution administered. Mice were allowed to recover on a warming pad in a recovery cage. When mice were awake and alert, they were returned to their home cage. Over the subsequent 48 h, the mice were monitored closely for any signs of complications due to the implant surgery.

Surgical procedures in preparation for imaging
Mice were anesthetized with 5% isoflurane and maintained at 1.5-2-2.5% isoflurane while on a feedback-controlled temperature pad (CWE Inc., Ardmore, PA) set to 37°C. A tracheostomy was performed to enable mechanical ventilation (SAR 830/P, CWE Inc.) and a tail vein catheter placed to allow for fluorophore and anesthetic infusion. Breath rate, heart rate, arterial oxygen saturation, and pulse/breath distention were measured throughout experiments via a pulse oximeter (MouseOx, STARR Life Sciences) and recorded via an MP150 acquisition system (Biopac Systems, Canada). Representative physiological data acquisition is shown in Figure S2 and the recorded values are summarized in Table 2.
Cranial window implantation was performed following established protocols [43]. Isoflurane was lowered to 1.25-1.5% and mice were head-fixed on a stereotactic frame (Narishige, Japan). Subcutaneous Ringer's lactate solution and Xylocaine (10 mg/mL, 50-100 μL volume, AstraZeneca Canada) were administered for hydration and local anesthesia, respectively. The scalp was removed and a dental drill used to create a 3-4mm craniotomy centered over the impact location, at +1 mm ML, +0.5 mm AP. The skull cap was removed; the dura was left intact unless it remained attached to the removed skull cap, and 1% agarose in phosphate-buffered saline (PBS) (Sigma-Aldrich, Canada) applied to the brain surface. A 5 mm glass coverslip for imaging only experiments or a small section of ~1 mm thick silicone-based polydimethylsiloxane (PDMS) [44] for combined imaging and electrophysiology experiments was placed over the agarose and secured with cyanoacrylate glue. Dental cement (Land Dental, USA) was then used to create an immersion well surrounding the window. Prior to imaging, a 70 kDa Texas Red dextran (25 mg/kg body weight, Invitrogen, USA) dissolved in PBS (8.33 mg/mL) was injected via a tail vein catheter.

Two-photon fluorescence microscopy (2PFM) and focused optogenetic stimulation
Two-photon fluorescence microscopy was performed following established methods [40]. 2PFM was performed on an FVMPE-RS multiphoton microscope (Olympus, Japan) with 25x/1.05 NA or 10x/0.6 NA objective lenses (Olympus, Japan) for imaging and imaging/electrophysiology experiments, respectively. The mouse was positioned under the microscope and an Insight Ti:Sapphire laser (SpectraPhysics, USA) was used to excite YFP-labelled ChR2-expressing neurons and Texas Red-labelled vasculature at 900 nm. Emitted signals were acquired by photomultiplier tubes (PMTs) preceded by 485-540 nm or 575-630 nm barrier filters, separated by a 570 nm dichroic mirror (Chroma Technology, USA). Images down to cortical depths of 500-600 μm were captured prior to functional imaging ( Figure 2). Thereafter, isofluorane was discontinued and anesthesia switched to alpha-chloralose (75 mg/kg induction, 40 mg/kg/hr maintenance) via an infusion pump (Harvard Apparatus, USA). The objective was positioned over the injury focus, and cortical penetrating vessels within that 2PFM imaging FOV (508 x 508 μm for the 25x objective, 1272 x 1272 μm for the 10x objective) were interrogated. The red blood cell (RBC) speed in these vessels was recorded through galvanometer-driven line scans (1.1-1.3 ms/line, 2 μs/pixel) along the longitudinal axes of these vessels, at the level of pre-diving point segments parallel to the cortical surface. The 458 nm (0.5 mW/mm 2 ) photostimulation was raster scanned with a separate galvanometer over a 120 μm-diameter circular ROI centered over each penetrating vessel with 144-502 ms repetition period, 4 μs/pixel, 2-2.1 s duration, so as to stimulate neuronal soma and processes proximal to the imaged vessel while maintaining >70% of the photostimulated area extravascular. Photostimulation power was measured via a power meter (Model 842-PE, Newport, USA). Power density conversion was performed using the area of the stimulation ROI (120 μm diameter circular ROI).

Local field potential recordings and data analysis
In a subset of experiments, local field potentials (LFPs) were recorded simultaneously with RBC speed. Mice were prepared as described above, but the cranial window was covered using PDMS instead of glass so as to allow for tungsten electrodes to be positioned in the cortical tissue below the cranial window [44], at a cortical depth of ca. 50 μm. LFP activity was captured via tungsten electrode mounted on a multi-manipulator (Scientifica, UK) and recorded through a patch-clamp amplifier operating in current clamp mode (MultiClamp 700B, Molecular Devices, USA) with an analog 20 Hz low-pass filter used to decrease background noise and enable real-time response visualization; and subsequently digitized for computer acquisition (Digidata 1440A, Molecular Devices, USA). Recordings were low pass-filtered post-hoc and LFP response parameters were extracted aparametrically. Extracted response parameters were peak LFP magnitude (ΔLFP) and area under the curve (AUC) during stimulus; as well as peak LFP magnitude and AUC following stimulus offset. To characterize neuronal activity at rest, power spectra of 20 s pre-stimulus baseline segments were acquired via FFT and divided into frequency bands as previously described [45]. ΔLFP increased with decreasing distance between the region of the photostimulation and the location of the recording electrode tip. This effect lead us to scale ΔLFP, on a subject-wise basis, to the level observed at the mean electrode-tophotostimulation ROI distance of all mice ( Figure 2B): all LFP readouts depicted were linearly corrected in this way, as shown in Figure S1.

RBC speed data analysis
RBC speed was estimated from line scans as described earlier [46], allowing for robust speed estimation of fast-moving RBCs commonly seen in cortical penetrating arterioles. After image normalization and DC offset correction, sequential line scans were Fourier transformed, one FFT was conjugated and multiplied by the FFT of the other, and the inverse FFT was performed on the result to compute their cross-correlation. Gaussian peak-fitting was performed to calculate the spatial shift between adjacent line scans. RBC speed was calculated from this spatial shift while taking into account the line scan acquisition rate. RBC speed traces were then low-pass filtered and a 10-point moving average was applied to minimize noise and effects of respiratory/cardiac fluctuations. RBC traces following each stimulus presentation were fitted to a gamma function to estimate response magnitudes (ΔvRBC), time to peak (TTP), full-width-at-halfmaximum (FWHM), and area under the curve (AUC).

Paired neurovascular measurements and neurovascular coupling ratio
In mice that underwent combined electrophysiology and 2PFM experiments, simultaneous neuronal and vascular recordings were made at rest and in response to photostimulation. At rest, Θ band power, acquired from spectral characterization of LFP baseline, and baseline RBC speed (v base ) were linearly regressed. In the evoked state, the neurovascular coupling (NVC) ratio was calculated as the quotient of venular Δv RBC AUC to ΔLFP AUC.

Histological characterization
At experimental endpoint peripheral blood was flushed with phosphate buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 ; pH 7.4) containing 0.1% heparin (1 i.u./mL; LEO Pharma Inc, Thornhill, ON) followed by 4% paraformaldehyde (PFA) (w/v) in PBS. Brains were post-fixed in 4% PFA overnight followed by cryoprotection in 30% sucrose (w/v) in PBS at 4°C. Brains were sectioned at a thickness of 40 µm via sliding freezing microtome (Leica Biosystems, Wetzlar, Germany). Brain slices were imaged on a Zeiss Observer Z1 microscope and images were analyzed via custom analysis in ImageJ. LaminB1 fluorescence was normalized on a slicewise basis to DAPI co-stain fluorescence to account for intersubject variance in staining. ( Figure 7C). ɣH2AX expression was calculated from the fluorescence within neuronal nuclei, as determined by co-expression of DAPI and NeuN, and expressed as a percentage of area of those nuclei ( Figure 7F). GFAP fluorescence was summed across images and normalized to the number of slices per image.

Statistical analysis
The effects of injury and treatment were assessed by generalised linear modelling using the lme4, lmerTest, and nlme packages in R, with subjects treated as a random effect. This modelling accounts for variation within and between subjects to yield robust estimates in the presence of unbalanced group sizes. Fixed effects in linear mixed effect models were condition (injury/sham), treatment (L-655,708/ vehicle), and, for histological data, hemisphere (contra-/ipsilateral to injury). Interaction terms between fixed effects were included in the modelling. The contrast across condition and treatment groups of the different readouts was thus modeled as: where multiplication in this notation denotes both the main effects and interaction between the two terms. Since the LFP magnitude estimates were not normal, they were log transformed prior to linear modelling. Sex was included as a fixed effect in preliminary analysis but was not found to exert a significant effect on any readout and was thus not made a part of the model in the final analysis.

Resting venular RBC speed and neuronal Θ power are altered in mTBI mice
We first examined the resting hemodynamics by comparing pre-stimulus v RBC (v base ) across groups. At rest, arterioles did not exhibit any baseline RBC speed contrast. In contrast, RBC speed in penetrating venules of mTBI mice was twice that of sham animals (Δv base injury = 115 ± 25%, p injury = 0.000005, p treatment = 0.78). Such increased venular baseline RBC speeds were not present in mTBI mice treated with L-655,708 (Δv base = -88 ± 35%, p injury x treatment = 0.01) as displayed in Figure 3A. Resting vessel diameters were not found to be significantly different across groups aside from a trend toward venular dilation due to mTBI (Δd venule = 14.5 ± 9.1%, p venule injury = 0.1). The standard deviation of venular baseline v RBC (σ base ) was increased in mTBI (Δσ base = 60 ± 39%, p injury = 0.0002, p treatment = 0.54, p injury x treatment = 0.07, Figure S3D). There was, however, no difference in the corresponding coefficient of variation (COV = σ base /v base ), indicating that the contrast in standard deviation resulted from the proportionate increase in v base ( Figure 3B). Analysis of power spectra of the baseline LFP recordings indicated that the theta band (4-8 Hz [45]) neuronal power was reduced in mTBI mice ( Figure 3B, ΔΘ injury = -57 ± 79%, p injury = 0.01). We also observed a trend toward a reduced effect of injury on neuronal theta band power in L-655,708 treated mTBI mice (ΔΘ injury x treatment = 27 ± 59%, p injury x treatment = 0.08).

Reduced neuronal reactivity in mTBI mice
To investigate changes in local neuronal activity and neurovascular unit function post-injury, 2PFM measurements of vascular reactivity were paired with local field potential recordings via a tungsten microelectrode [44]. ΔLFP was attenuated in the mTBI group (ΔLFP = -47 ± 28%, p = 0.018), as shown in Figure 4. A simplified linear model further showed that there was no statistically significant difference between sham and mTBI L-655,708-treated mice (p = 0.59). Of note, as shown in Figure 2B, the Euclidean distance between photostimulation ROI/vessel diving point and the injury focus was recorded: however, no significant correlation between this distance and any neuronal measures were observed within the 1272 x 1272 μm imaging FOV used here. These findings show that mTBI may elicit sustained reduction in neuronal reactivity and that low-dose L-655,708 intervention can prevent this effect.

Augmented venular reactivity in mTBI mice
Vascular responsivity was evaluated in the chronic stage of repeated mTBI by measuring single cortical penetrating vessel responses to focused photostimulation ( Figure 5). Area under the curve (AUC) of arteriolar responses to focused photostimulation was found to be indistinguishable between cohorts ( Figure 5B), whereas venular reactivity was elevated in mTBI mice ( Figure 5C, ΔAUC = 53 ± 17%, p = 0.001). This venular hyperreactivity was prevented by L-655,708 treatment (ΔAUC = -66 ± 24%, p = 0.005). Our findings here indicated sustained venular hyperreactivity post mTBI and a potential for its prevention by an early, low dose L-655,708 intervention. Similarly to neuronal recordings, no significant correlations between distance from injury focus and any vascular measures were observed within our 508 x 508 μm imaging FOV.

Neurovascular coupling is altered in mTBI mice
Lastly, we examined the relationship between neuronal and vascular reactivity to test for changes to neurovascular coupling. Paired neuronal and venular function measurements at rest were linearly regressed to assess the effects of mTBI on resting state ( Figure 6). We observed a significant dependence (p = 0.009, slope = 9 ± 4) of the baseline venular vRBC on neuronal theta power at baseline ( Figure 6A), with no effect of either injury or treatment. For responses to photostimulation, the neurovascular coupling (NVC) ratio was calculated as the ratio of vascular to neuronal AUC following photostimulation. We observed a significant increase in NVC ratio due to mTBI as shown in Figure 6B (ΔNVC = 674 ± 304%, p = 0.039) and further that mTBI mice treated with L-655,708 were indistinguishable from their uninjured counterparts (p = 0.88), implying an increased venular NVC component in the injured state that may be prevented through GABAAR inverse agonism.

Reduced LaminB1 expression and increased ɣH2AX expression in mTBI mice
To independently assess the mTBI sequelae, we evaluated several histological markers in the pericontusional tissue. Neuronal survival (on NeuN) and glial activation (on GFAP) were indistinguishable between mTBI and sham cohorts. We next evaluated laminB1 and ɣH2AX as prior clinical and preclinical studies reported on TBI-elicited senescence, manifested as nuclear membrane loss (and thus reduced cerebral laminB1 expression) [47,48], and DNA damage (and hence increased ɣH2AX expression in neuronal nuclei) [47]. LaminB1 expression was visibly reduced in mTBI vs. sham mice as shown in Figure 7C (ΔLaminB1/DAPI = -17 ± 8%, p injury = 0.047). ɣH2AX expression was increased in mTBI cohort, as demonstrated in Figure 7F (ΔɣH2AX = 43 ± 13%, p injury = 0.011). YFP expression in brain slices was also quantified to test whether ChR2 expression were altered following mTBI; however, no such changes were observed ( Figure S4). No significant differences were observed between ipsi-and contracontusional hemispheres in mTBI mice in any histological readout (pLaminB1 hemisphere = 0.25, p ɣH2AX hemisphere = 0.84).   LaminB1 (red) staining of a brain slice maximum intensity projection along the A-P dimension in the ipsi-contusional hemisphere of a vehicle-treated sham mouse at the sham impact coordinates. B. DAPI/LaminB1 staining of a brain slice maximum intensity projection along the A-P dimension in the right hemisphere of an mTBI mouse at the impact coordinates. Scale bar = 50 μm. C. Total LaminB1 fluorescence normalized by total DAPI fluorescence in each image slice across injury/treatment groups. LaminB1 fluorescence is significantly reduced in mTBI mice (ΔLaminB1/DAPI = -17 ± 8%, pinjury = 0.047). Nmice = 23 (6 sham/vehicle, 5 sham/L-655,708, 6 TBI/vehicle, 6 TBI/L-655,708), Nimages = 184 (4 slices/subject, 2 contra/ipsi images/slice). Representation of imaging locations on the brain slices. Image from biorender.com. D. NeuN (green) /ɣH2AX (magenta) staining of a brain slice maximum intensity projection along the A-P dimension in the right hemisphere of a vehicle-treated sham mouse at the sham impact coordinates. E. NeuN/ɣH2AX staining of a brain slice maximum intensity projection along the A-P dimension in the right hemisphere of an mTBI mouse at the impact coordinates. Scale bar = 50 μm. F. Fractional area of ɣH2AX expression within neuronal nuclei in each image slice across injury/treatment groups. ɣH2AX expression is significantly increased in mTBI mice (ΔɣH2AX = 43 ± 13%, pinjury = 0.011). Nmice = 20 (4 sham/vehicle, 4 sham/L-655,708, 5 TBI/vehicle, 7 TBI/L-655,708), Nimages = 160 (4 slices/subject, 2 contra/ipsi images/slice).

Discussion
The cellular level changes elicited by mild traumatic brain injury are not well understood [49][50][51]. Previous preclinical mTBI studies have largely focused on structural changes and/or mesoscopic scale functional measures [21,24,25,31,[52][53][54][55][56]. Here, for the first time, we paired single vessel measurements with intracortical recordings and applied focal photostimulation to elucidate neurovascular unit dysfunction post mTBI. We demonstrated beneficial effects of delayed, chronic, low dose disinhibition, via GABAAR inverse agonism, on the neurovascular function. Given the importance of neurovascular physiology for brain health [17], our findings motivate the development of post-concussion interventions that target the neurovascular unit to promote long term resilience of the concussed brain. Successful application of such interventions in heterogeneous clinical populations will be predicated on the development of translational biomarkers of neurovascular dysfunction post mTBI, thereby allowing optimization of treatment to individual patients' needs.
The impact parameters in this model were set within previously-described guidelines for inducing mild injury with the controlled cortical impact [57,58]. The resulting injury was designated as 'mild' given lack of contrast on structural MRI, the criterion commonly applied in clinical practice [13][14][15]. That the injury produced was mild was further supported by the lack of group contrast on both NeuN and GFAP staining. While microhemorrhages, dural damage, and vessel rupture are common occurrences in the acute stage of mTBI [59][60][61][62], they are expected to resolve rapidly and were indeed not presently observed at two weeks post final insult. Prior preclinical studies have characterized TBI pathophysiology via morphological, functional, and behavioural assays [7-9, 19, 20, 22, 24-26, 30, 31, 50-52, 54-56, 63-71]. The presence and extent of these changes has varied widely across studies depending on the age of subject [30,65,72], time post-injury [19,31,54,63,64,70], number of injuries [30,31], and injury model [65,67]. In the case of mild TBI, it has been shown that long-term changes may arise in both patients and in rodent models from either single or repeated injuries [11,49]. Consistent with this, we observed neurovascular impairments two weeks post third impact. The current study used laminB1 and ɣH2AX expression as markers of senescence in the absence of cohort-wise differences with respect to either NeuN or GFAP expression. Dramatic laminB1 expression reductions, which are associated with nuclear membrane ruptures and cell death [73], have been observed in TBI and in preclinical models of TBI [47,72]. In the present mild TBI model, a limited decrease of laminB1 expression was observed as well. Furthermore, heightened ɣH2AX expression, indicative of DNA damage, has been reported post mortem in TBI brains [47], a feature recapitulated by our model. Of note, these effects were not specific to the impacted hemisphere, indicating that closed head impacts elicit bilateral damage and/or that the injury is relayed by the two-week time point. Acute bilateral damage is likely a result of contrecoup injury [61], with the impactor causing deformation of the impacted cortex and contusion of the contralateral cortex. This is in line with cortical impact models being described as "mainly focal" [57]. Neuronal survival and astrogliosis, as measured by NeuN and GFAP expression, were indistinguishable among groups. In a similar model, we have previously reported no NeuN contrast alongside ipsicontusional GFAP contrast [25], as with earlier work [31,74]; other studies, in contrast, did find reductions in neuronal survival [75][76][77]. Despite the lack of group contrast, gliosis was visible in some of our mTBI mice.
The cerebrovascular changes observed in mTBI patients in the chronic stage post trauma exhibit high variance with respect to presence, directionality, and magnitude of cerebral blood flow (CBF) signal changes [7, 13, 20, 25, 50-54, 68, 78]. The majority of studies have reported a reduction in baseline arterial blood flow [13,52,78] and an attenuation of stimulation-elicited cerebrovascular responses measured 6+ months post mTBI [13,52]. Several studies have reported decreases [79][80][81][82] in CBF in the weeks to months following mTBI via ASL fMRI, whereas others have shown increases [78,[83][84][85], no change [86][87][88][89], or a combination of spatiallydependent changes [90][91][92][93] in this timeframe. In preclinical models of mTBI, hypoperfusion has been observed at 1-4 weeks post TBI [25,70] and within 1 hour to 1 day of injury [19,22,24]. Similar to the clinical data, findings in preclinical models are highly heterogeneous but widely report CBF and cerebrovascular reactivity reductions post-injury [38,94], whereas we presently report venular hyperreactivity, analogous to previous clinical findings [13,78] and the first preclinical study to report this to our knowledge. While conventional readouts may be valuable for establishing the presence and severity of injury, quantitative functional measurements at the single-unit level are necessary for an in-depth understanding of the neurovascular sequelae of TBI pathophysiology [8,9,49,50,56] at a single-unit level [49,50], and for establishing the absolute magnitude of these changes [13,95]. Here, we observed doubling of venular baseline RBC velocities, a trend toward venular dilatation of about 15%, as well as a ~50% elevation in venular reactivity. Analogously, earlier work in mTBI patients reported increased CBF using BOLD and ASL fMRI 5-6 weeks [83] and 6+ months [52,56,68,78,87] post injury. Evaluation of BOLD fMRI responses to working memory task in mTBI patients 4+ months post-injury has revealed attenuated responses of peri-contusional vasculature [87], whereas other studies using hypercapnic challenge with BOLD fMRI have shown increased reactivity 2+ weeks following injury [89] and with ASL 1 week following injury [88]. The presently-observed venular specificity of the treatment effect on the brain vasculature may be due to venular measures reflecting an integration of all upstream changes in the microvascular network [96] capillaries are likely affected by trauma as well [21,97], acquiring a representative sample size of these numerous microvessels would have taken prohibitively long. Additionally, the intrinsically slow nature of laser scanning (vs. relatively high speed of RBCs in cortical penetrating vessels) prevented us from making robust measurements of photostimulation-induced changes in the vessel diameter concomitant to the vessel's vRBC estimation.
Neurovascular coupling was interrogated in this study through ratios of paired neuronal and venular measures. While resting coupling between neuronal and vascular function remain unaltered, the NVC ratio was increased by mTBI. This finding suggests an alteration of neurovascular signaling that may be an adverse sequela of mTBI or a compensatory effect to potentiate the support of the injured tissue. At rest, venular baseline vRBC was linearly dependent on baseline neuronal theta power; this dependence was not affected by either injury or treatment. While gamma band neuronal power would have been interesting to evaluate due to its role in neurovascular coupling [98], such analysis was presently precluded by the 20 Hz low-pass analog filtering used to mitigate the noise in the electrophysiological recordings introduced by the concurrent microscopy and by the use of a tungsten electrode (to accommodate a bulky objective of concurrent 2PFM and electrophysiology recordings). On the whole, our observations support the notion of evoked response assays being more sensitive measures of brain (dys)function than the baseline state recordings [13].
Interneuron-mediated tonic inhibition has been previously implicated in preclinical mTBI pathophysiology [8,9,26,27]. L-655,708 distribution in the brain following i.p. administration of 3 mg/kg (vs. 2.4 mg delivered subcutaneously, over 12 days, in the current study) has previously been shown to result, within 1-2 h following injection, in widespread distribution across the neocortical tissue [99]. In contrast to the neocortex, little L-655,708 was observed in the cerebellum and the subcortical regions. Excitatory/inhibitory signaling balance is disrupted in mTBI as well as in severe injuries [8,9] and this imbalance is thought to cause hyperreactivity in the immediate aftermath of injury followed by hyporeactivity in the subacute phase of injury [9]. A delayed interventional window may thus be present in the hyporeactive phase post-injury. In months following mTBI, it has been shown that functional changes mediated by GABAA receptor currents occur in absence of an increase in receptor count [27]. Additionally, one week following mTBI, impaired behaviour and synaptic plasticity were previously prevented via administration of L-655,708 intraperitoneally [26]. Here we demonstrate the neurophysiological effects of such treatment at the single-unit level in vivo: specifically, a prevention of venular deficits and a normalization of neuronal responses in the L-655,708 treated vs. vehicle administered mTBI animals. The venular normalization observed in mTBI/L-655,708 mice may have been secondary to the normalization of neuronal tone and/or induced by the decreased availability of GABA, which is itself vasoactive [100,101]. This indicates that a well-timed (subacute phase), sustained, low-dose GABAA receptor inverse agonism may have preventative effects on the neurovascular aberrations elicited by mTBI and identifies the excitation/inhibition signaling imbalance as an important mechanism of injury and hence treatment target. Furthermore, this study was performed with functional recordings in vivo, providing a greater translational potential than the previously reported ex vivo measures [8,9,26,27]. Of note, L-655,708 administration in this study was systemic (via a subcutaneous osmotic pump) and is thus easily translatable. The resulting widespread effect on the neocortex [99] may have been beneficial considering the aforementioned spatial propagation of injury.
In summary, the present study describes sustained neurovascular impairments in a preclinical model of repeated closed head injury. We showed that in addition to neuronal hyporeactivity (-47 ± 28%) and reduced activity at rest (-57 ± 79%), the pericontusional venular function is abnormally elevated at rest (115 ± 25%) and upon stimulation (53 ± 17%). Increased resting cerebral blood flow has been observed at a regional level via MRI in mTBI patients [13,56,68,78] but has not been shown at the level of individual vessels. Elevated venular responses to neuronal activation is a novel finding in preclinical mTBI, with analogous pathology described clinically [13,88,89]. Delayed, low dose GABAA receptor inverse agonism via L-655,708 [26,27,34,35] to alleviate hypothesized tonic inhibition post TBI [8,9,26,27,34,35] prevented the mTBI-induced venular impairment and normalized the neuronal reactivity. Our study provides new insight into the neurophysiological sequelae of mTBI, constitutes a first step in the development of translational markers of neurovascular pathology in mild TBI that is silent on neuroimaging, and suggests benefits of attenuating tonic inhibition in the subacute to chronic stage of injury.