Persistent dysmorphology and dysfunction of the brain microvasculature following repeated trauma

Abstract

Mild traumatic brain injury typically produces no abnormalities on neuroimaging yet elicits symptoms that, in an increasing fraction of survivors, linger for years, particularly with recurring injuries. To elucidate the underlying biological substrates, we leveraged two-photon fluorescence microscopy resonant scanning and our recently developed deep-learning based pipeline to evaluate the cerebrovascular network in the subacute stage following three intact-skull cortical impacts in adult mice under low isoflurane. Microvascular volume density increased by 19.1 ± 18.6% after moderate impacts, while mean capillary diameters increased by 5.6 ± 3.2% for mild and by 1.1 ± 2.6% for moderate impact series; consistent with network remodeling and tone dysregulation, which limits cerebrovascular reserve. Mild impacts cohort showed a paradoxical hypercapnia response with decreases in red blood cell velocities (vRBC) (−20 ± 13%) in the majority (62%) of cortical penetrating arteries and a net ipsicontusional hypercapnia-induced arteriolar vRBC decrease, in stark contrast to the physiologically normal increase seen in sham mice (+27 ± 15%). Downstream, the mild impacts resulted in attenuation of hypercapnia-induced cortical capillaries’ flow increase while the moderate impacts cohort showed no ipsicontusional capillary flow response. Sustained aberrations in brain microvasculature post-trauma may underlie the persistence of symptoms and mediate susceptibility to further injury.

Introduction

Traumatic brain injury (TBI) is a major cause of disability worldwide, with an estimated 69 million TBIs a year ¹ and a heavy social and economic burden.^2,3 The bulk of the TBIs are of mild-type or “concussions,”^3–5 clinically defined as lack of skull fracture, less than 10 min of unconsciousness, and at least one of: post injury amnesia, altered mental state at the time of accident, and a transient or sustained neurological deficit.^6,7 While symptoms often resolve within 3 months, they may endure for decades.^8–13 Given the increasing population of patients who experience long-term symptoms (up to ~50%),^14–16 a better understanding of the sustained TBI-elicited neuropathologies is of wide importance. However, the heterogeneity of the injury progression and the lack of conventional neuroimaging contrast make it difficult to monitor recovery and predict outcomes in the clinic.^17,18 Repeated injury is of particular interest given its high frequency in multiple settings and its pernicious long-term deficits. Although most mild TBIs occur in isolation, repetitive mild TBIs, often seen in contact sports, military service, and domestic violence^19–25; are especially insidious as recurrence increases the brain’s vulnerability to damage, particularly if encountered before symptoms resolve.^26–30 Mounting evidence suggests that a history of TBIs increases susceptibility to depression, dementia, memory deficits, neurodegenerative diseases and stroke.^31–41 An understanding of the sustained microscopic damage post-TBI could help address persistent symptoms by identifying potential targets for disease-modifying treatments.

Stable cerebral blood flow (CBF) is critical to brain function, yet cerebrovascular damage is common in TBI cases. ⁴² To maintain blood flow within the physiological range, the diameters of the brain’s resistance vessels are continuously adjusted in response to variations in blood pressure (cerebral autoregulation), levels of neuronal activity (neurovascular coupling), and in response to vasodilators or vasoconstrictors such as acid/base balance, nitric oxide amounts, and CO₂ concentration (cerebrovascular reactivity).^43–46 CBF impairments are frequently observed in repetitive TBI, manifesting as acute decrease in autoregulation, ⁴⁷ chronic reductions in cerebral autoregulation and cerebrovascular reactivity, ⁴⁸ and a range of other CBF alterations following mild TBI.^49–51 Impaired autoregulation may result in a non-physiological level of blood flow, increasing the susceptibility of tissue to damage from subsequent hits, potentially to catastrophic ends.^19,52 Cerebrovascular reactivity (CVR), in particular, can be a useful indicator of the brain’s vascular reserve capacity and vascular health, by quantifying the ability of blood vessels to dilate in response to a vasodilatory challenge. The global cerebrovascular dilation in response to increased CO₂, or hypercapnia, is a common CVR challenge that can be easily implemented through inhalation of a gas mixture with 5%–10% CO₂ tension.^53,54 The CVR reduction in mild TBI patients is often co-existent with normal resting cerebral blood flow, indicating that CVR is a more sensitive measure of brain vascular damage than baseline CBF.^49,55–57

Preclinical studies are particularly useful for elucidating subtle TBI-induced microvascular changes that are not detectable on conventional neuroimaging, although grading for TBI in experimental models has been rather inconsistent.^58,59 Although rodent TBI models do not recapitulate all aspects of human injury, ⁶⁰ they afford a finely-controlled setting in which to examine the trauma sequelae at the cellular level. Controlled focal brain deformations using the closed head controlled cortical impact (CCI) model ⁶¹ in particular provide control over the location, severity and potential spread of the injury via adjustments in the impact velocity, the depth of impact, the duration of contact and the tip size; all the while leaving the skull intact for multiple impacts, as a craniotomy procedure will cause inflammation separate from that produced by the impact. ⁶² At the subacute and chronic timepoints following “mild” TBI cortical impacts, neurovascular impairments reported by our group and others to date include reduced CVR to hypercapnia and reduced baseline CBF in conjunction with blood-brain barrier breakdown, gliosis, and pericyte degeneration without vessel loss.^63–66 On the whole, however, the subacute stage of injury has been understudied in experimental models of TBI, and most hitherto work has been largely qualitative in nature. The subacute injury stage is arguably the most likely interventional window as patients with mild TBI often delay treatment, ⁶⁷ but the impacted tissue still shows a high level of dynamicity.^68,69

To address these gaps, we here set out to quantify persistent morphological and functional alterations induced by graded (either mild or moderate) repetitive impacts at a microscopic level in the sub-acute stage of injury. We contrasted the ipsi- to contra-contusional cortex to increase the sensitivity of our metrics. By employing two photon fluorescence microscopy (2PFM) in low isoflurane anesthetized mice, we examined the CVR to hypercapnia via measurements of red blood cell velocity (vRBC) in cortical penetrating supply and draining vessels—that are key for normal cortical perfusion^70,71—as well as via tracking the transit of a bolus of fluorescent contrast agent through the cortical microvascular bed at varying cortical depths. We selected CO₂ as a vasodilatory stimulus because of its use in human CVR studies, its robustness, and easy repeatability of the challenge. We performed morphometric analysis of the microvasculature using our computational pipeline for cerebrovascular segmentation and graph analyses from in situ 2PFM data. We revealed a series of persistent deficits in microvascular structure and function, including tone loss and attenuation of microvascular reactivity.

Materials and methods

Animals

Experimental procedures followed the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines ⁷² and were approved by the Animal Care Committee of the Sunnybrook Research Institute, which adheres to Policies and Guidelines of the Canadian Council on Animal Care and meets all requirements of the Provincial Statute of Ontario, Animals for Research Act as well as those of the Canadian Federal Health of Animals Act. All mice were kept under a 12 h light/dark cycle, with access to food and water ad-libitum. A total of 61 mice were used for this study: microscopy imaging data sets were acquired in 34 mice (age range 2.8–10.9 months old with an average age of 5.8 ± 2.6 months, 22 males, 14 females, mean weight 26.6 ± 5.4 g, see Table S1 for sex breakout and Figure S1 for group breakout). Thirteen mice were excluded from imaging (Figure S1) four were excluded due to TBI skull cracks during; five due to post-surgical health decline; and four due to failures during surgical preparation for imaging (one from probable air introduction during IV injection, three from excessive cranial window bleeds obscuring optical imaging clarity). Following the 3Rs ⁷³ in order to reduce the number of total animals, we used mice strains from related study projects so that the obtained data here might be usable in the related projects as well. Strains used were on a C57BL/6 mouse strain (Charles River Canada) or mutant mouse strains on a C57BL/6 genetic background: Cre XTdTomato (B6.Cg-Tg(Tek-cre)1Ywa/JxB6; 129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J, Toronto Centre for Phenogenomics), and Thy1-ChR2-YFP (B6.Cg-Tg(Thy1-COP4/EYFP)18Gfng/J, Jackson Laboratories 007612, line 18). Cre XTdTomato mice express a red fluorescent protein in endothelial cells under the Tek (Tie2) promoter^74,75 and thus enabled examination of endothelium; while Thy1-ChR2-YFP line 18 mice express YFP and channelrhodopsin in pyramidal neurons, required in a related TBI project.

Closed-head controlled cortical impact

Mice started the protocol for repeated non-penetrative traumatic brain injuries of graded magnitude from age 2.5 to 10.6 months of age (thus spanning young adult to middle-aged animals), to be imaged two weeks following the third and final hit (Figure 1(a) and (b)). Each mouse received three closed head cortical injuries spaced 72 h apart using a version of the mild protocol described earlier, ⁶³ falling within the temporal window of vulnerability to repeated injury.^23,76 Skulls were examined via 25× surgical stereoscope microscopy to ensure they remained intact. Though “mild trauma” parameters can vary from study to study, in general its depth of impact is below 1 mm and the impact velocity lower than 4 m/s.^59,77 Mild TBI in our study referred to the following impact parameters as used in our previous studies^63,78: 1.5 mm tip diameter, 2 m/s impact velocity, 200 ms dwell time, with an impact depth of 1 mm at (+0.5 mm AP and 1.75 mm ML) in the right hemisphere, as shown in Figure 1. Moderate TBI was induced with a 3 mm tip diameter, 4 m/s impact velocity, 200 ms dwell time, and an impact depth of 1 mm. Sham animals received the same anesthesia, stereotaxic positioning and surgical incision, over the same duration of time, but with no impact. The acute and subacute stage neurological deficits assessments were performed as described in the Supplemental Methods. To ensure the moderate impact showed little or no contrast on MRI, T2 weighted structural MRIs were acquired using a 7T animal system (Bruker, BioSpec, Ettlingen, Germany), as described in the Supplemental Methods.

Figure 1.

Experimental design and methodological overview. (a) Experimental timeline showing mice receiving non-penetrative impacts of either mild or moderate grade, spaced 3 days apart. (b) Impactor was targeted to +0.5 mm AP, and 1.75 mm ML on the skull over the right hemisphere. (c) Structural MRI of three separate representative mice, showing a coronal slice at a location of the center of the impact region, imaged during the 24 h following the first moderate impact. (d) Structural MRIs of three different representative mice showing coronal slices at a location of the center of the impact region, imaged during the sub-acute stage following the three moderate impacts. (e) An XYZ stack maximum intensity projection showing the vascular lumen (MIP) with overlaid bolus injection time series (XYT scan) (blue) placed at three different cortical depths used for bolus scan acquisition: one bolus XYT scan was collected at the top of layer II/III, one toward the bottom of layer II/III and one in layer V. Each location was collected first under normocapnia, followed by hypercapnia (10% FiCO₂) condition. (f) Signal intensity transit traces over time for the same vessel under normocapnic (grey trace) and under hypercapnic state (purple trace) overlaid on the same time scale. (g) Maximum intensity projection as viewed from the surface with the pial vessel’s diving point that was line scanned indicated with turquoise line. Inset shows a sample XYT line scan data collected in an ascending venule. (h) Example of a venule line scan under normocapnic conditions, with the red blood cells appearing as black lines against a dextran labeled plasma. Flow velocity is estimated via the slope of the red blood cell black lines. (i) The same vessel under hypercapnic (10% CO₂) conditions showing, through the decreased slope, red blood cell velocity increase relative to that during normocapnia. (j) Example of a vascular graph extracted from the 2PFM volumetric data from a sham mouse, used for morphometric analysis. (k) Vascular graph colorized according to vessel radius.

Two-photon laser scanning microscope

Following preparatory surgeries (see Supplemental Methods for detailed description), isoflurane was reduced to low levels for imaging (~0.75%–1%). Animal physiology was monitored throughout imaging sessions (see Supplemental Methods for detailed description) (Table 1). Blood pressure (systolic, diastolic, and mean) was also measured in a small cohort of 6-month-old mice (n = 6; 3 male, 3 female, evenly distributed across groups) using the CODA VPR tail-cuff system (Kent Scientific) (Figure S7). Two-photon fluorescence microscopy (2PFM) was performed using the Olympus XLPLN25XWMP2 multiphoton water immersion objective (25×, 1.05 NA, 2 mm WD) mounted on the Olympus FVMPE-RS microscope (Evident Scientific, Waltham, MA) coupled to an InSight DeepSee laser (Spectra-Physics, Santa Clara, CA) with a tuning range of 680–1300 nm, <100 fs pulse width, 80 MHz pulse repetition rate and a Mai-Tai HP DS laser (Spectra-Physics, Santa Clara, CA), tuning range of 690–1010 nm, ~100 fs pulse width, 80 MHz pulse repetition rate. Excitation and emission barrier filters (Chroma Technology, USA) for fluorophores used are detailed in Table S6. The order of imaging the injured and non-injured hemispheres was alternated to diminish confounding effects of worsening animal physiology over time. The emitted signals were separated by a 570 nm dichroic mirror (Chroma Technology, USA). High speed resonant scanning was used to acquire fast scanning in the X direction for bolus scans and XYZ stacks. Galvanometer scanning was used for the vRBC line scans. The number of animals by scan type is summarized in Table S2.

Table 1.

Physiological monitoring data.

Physiological Measurement	Sham	Mild TBI	Moderate TBI
paCO2 estimate based on microcapnography (mmHg)	35.7 ± 12.1	36.9 ± 8.2	33.8 ± 3.5
Heart rate at normocapnia (bpm)	433.3 ± 55.3	427.2 ± 71.9	458.5 ± 67.4
O2 saturation at normocapnia (%)	97.1 ± 1.0	95.1 ± 2.7	96.7 ± 3.0

Physiological monitoring metrics across groups. Mean values ± standard deviations of physiological readouts of sham and TBI cohorts under normocapnic conditions recorded during imaging. There was no statistically significant contrast between cohorts on any of these metrics.

Bolus scan acquisition and analysis

To assess cerebrovascular reactivity through the microvascular response to a hypercapnia, bolus passage was tracked via XYT scans over the same plane, first during normocapnia (20%–30% O₂ with balance nitrogen) and next during hypercapnia (20%–30% O₂, 10% CO₂, and balance nitrogen), for a bolus pair. The XYT scans of 60 s duration were 509 µm × 509 µm, 2 µm × 2 µm nominal in-plane resolution, 2 μs/pixel, frame averaged 4 times, generating 3.8 frames per second. Scan acquisition began 60 s after the peak ETCO₂ (76.0 mmHg) on the microcapnograph. The gas composition was returned to normocapnic mixture immediately following completion of the hypercapnia scan. The animal was allowed to return to the baseline (normocapnic) ETCO₂ level for at least 5 min before starting a subsequent normocapnia/hypercapnia pair. Bolus tracking acquisition plane pairs were collected at 150 µm cortical depth and the next pair at 300 µm (within layer II/III). When image quality allowed, a third pair was acquired at 700 µm up to 1100 µm (within layer V/Layer VI) (Figure 1(e)). The bolus was injected at the 30 s timepoint of the scan, the first 30 s being used for baseline fluorescence. Each bolus (of 5.5–8.25 mg/ml with total end volume of 33 mg/kg) was injected at a rate of 45 µl/s via a catheter infusion line connected to an infusion pump (Harvard Apparatus, USA). Signal intensity increases with bolus passage (Figure 1(f)), where arteries tend to exhibit the signal increase first, and veins last. Images were rendered with ImageJ Fiji ⁷⁹ and Bitplane Imaris (Belfast, UK). Analyses were performed in Matlab (The MathWorks inc, USA). Transit time estimates were made by evaluating the period of time it takes for a bolus of fluorescent dye to pass through the vascular tree normalized by the arrival time, that is, the onset time of the fluorescence increase in the imaged plane. ⁸⁰ The average signal intensity for each vessel segment (vessel flanked by two branching points and/or edge of the FOV) was computed and normalized to the peak signal intensity of that vessel. Transit times were separated according to vessel type (i.e., arterioles, capillaries, venules). Using volumetric images, penetrating vessels were designated as arteries or venules based on their morphological features and branching patterns.^81–83 Vessels less than 10 µm in diameter were classified as capillaries. The bolus passage, during normocapnia and hypercapnia, was modeled using the gamma variate function^84,85 as described previously.^80,86

Line scans acquisition and analysis

Red blood cell (RBC) velocity measurements were made via galvanometer driven line scanning along the long axis of the penetrating vessels at the pre-diving horizontal section (Figure 1(g)). Line scans were acquired over 12 s at 1.2 ms/line, 0.5 µm in plane resolution, 2 µs/pixel. Images were rendered in ImageJ Fiji and Imaris. The displacement of the red blood cell between sequential line scans (Figure 1(h) and (i)) was used to calculate the speed, using custom Matlab scripts we have described previously, ⁸⁷ leveraging the code provided in Kim et al. ⁸⁸

Volumetric imaging

For volumetric images, a stack of slices parallel to the cortical surface was acquired from the pial surface down to at most 1100 µm, every 1.5 µm, with a nominal in-plane resolution of 1 µm × 1 µm (512 pixels × 512 pixels). Excitation powers were increased with focal depth to collect enough signal without either heating or saturation, keeping illumination powers at less than 50% of the damaging levels. ⁸⁹ Fast resonant scanning was used (0.067 μs/pixel) with each 3-frame averaging to achieve image quality similar to that of slower high resolution galvanometer driven scanning (2 μs/pixel), reducing imaging time. ⁹⁰ Despite averaging, the time spent acquiring a volumetric stack was reduced by a factor of five compared to galvanometer scanner, thereby limiting total acquisition time and improving the temporal sampling.

Vascular segmentation for morphometrics

Vascular segmentations of 2PFM volumetric scans are challenging due to limited signal-to-noise ratio and variable image quality across animals (due to variability in the cranial window fidelity, the preponderance of large pial vessels in the FOV, and cross talk between fluorescent reporters). Ground truth images were created by a rater manually annotating blood vessels in 42 image stacks from 25 mice, labeling microvasculature, pial and penetrating vasculature, and non-vascular signal (e.g., background and non-vascular cellular signal); these images were used to train a deep learning model to segment the vasculature.^91,92 Pericytes and endothelial cells were segmented via thresholding and connected component analysis to ensure that they met a minimum size threshold, 500 pixels. Following segmentation, vascular segmentation masks were eroded down to centerlines, which were then converted to graphs (Figure 1(j)). Boundary detection of intensity gradients of the fluorescent signal radiating out from the centerlines was then used to estimate the diameter of each blood vessel. ⁹² Distances from pericytes to vessels were defined as the shortest distance, estimated from a distance transform on the vessel centrelines and adjusting for vessel radius.

For vascular network morphometrics, the vasculature was cropped to a maximal depth of 800 μm for mild TBI vascular networks and reference SHAM group comparisons. Moderate TBI and SHAM comparisons were uncropped. This was done to account for differences in tissue clarity in mouse strains used for the different groups. The total volume of the vasculature was estimated by summing the total volume for each vessel, which was estimated via the mean diameter and length, assuming each vessel segment could be modeled as a cylinder. The vascular branch point density was estimated using the number of nodes with an order greater than 2, and the vessel segment density from the number of edges in the extracted graph. Vessel length density was estimated by adding the total length of the vascular network and dividing by the volume of the stack. Incidences of pericytes along each vessel were counted by looking for vessel centerline segments with pericytes within 10 μm of the vascular wall. These were then used to classify vessels as expressing observable pericytes or not and counting the number of pericytes on each vessel. Unique incidents of pericytes were counted using connected component analysis, and ellipsoids were fitted to each endothelial cell to estimate the length of each cellular axis. The largest two axes were used to estimate the eccentricity of each endothelial cell.

Statistics

The animal numbers by scan type are listed in Table 2. The change in red blood cell velocity (ΔvRBC) between normocapnia and hypercapnia was calculated as:

Table 2.

RBC velocity change induced by hypercapnia.

Vessel type	Sham (%)	Mild TBI (%)	p-Value: mild TBI vs sham	Moderate TBI (%)	p-Value: moderate TBI vs sham
Mean change in ipsicontusional vRBC
Arterioles	27 ± 15	−20 ± 13	0.0309*	3 ± 13	0.197
Venules	43 ± 16	24 ± 16	0.4925	47 ± 22	0.7805
Mean change in contacontusional vRBC
Arterioles	17 ± 24	−18 ± 23	0.1482	9 ± 14	0.9447
Venules	82 ± 43	17 ± 18	0.1048	11 ± 14	0.0749

Percent change in arteriolar and venular RBC velocity in response to hypercapnia, relative to normocapnia. *Indicates p < 0.05

$$\mathrm{\Delta }\mathbf{v}\mathbf{R}\mathbf{B}\mathbf{C}[\%]=\left(\mathbf{v}\mathbf{R}\mathbf{B}\mathbf{C}\_\mathbf{H}\mathbf{C}/\mathbf{v}\mathbf{R}\mathbf{B}\mathbf{C}\_\mathbf{N}\mathbf{C}-1\right)\times 100$$

Data were then stratified by vessel class (arteries vs veins). To examine regional effects, we separated contracontusional from ipsicontusional data. Linear mixed-effects modeling was performed using the lme4 ⁹³ and lmerTest packages in R (version 4.2.2). The models included treatment (moderate TBI, mild TBI, or sham) as a fixed effect and subject ID as a random effect, enabling robust evaluation of treatment-specific effects:

$$\mathrm{\Delta }\mathrm{v}\mathrm{R}\mathrm{B}\mathrm{C}~\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}+\left(1|\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{j}\mathrm{e}\mathrm{c}\mathrm{t}\_\mathrm{i}\mathrm{d}\right)$$

Similarly, for microvascular morphometry analyses, statistical significance was assessed using lme4 and lmerTest packages, with vessel class as the fixed effect and subject as a random effect. Post hoc comparisons were done using Mann-Whitney U tests implemented in rstatix, ⁹⁴ and the significance threshold was set at p = 0.05. The following linear mixed effects models used to look for contrast between treatment groups:

$$\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{a}\mathrm{m}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{r}~\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}+\left(1|\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{j}\mathrm{e}\mathrm{c}\mathrm{t}\_\mathrm{i}\mathrm{d}\right)$$

Microvessels above 10 μm in diameter were analyzed separately from capillaries due to the known structural differences in their vessel walls. All values are quoted as mean ± standard deviation unless otherwise noted.

Results

Our results elucidated that morphological alterations and pronounced vascular function deficits persist into the subacute stage of injury and that the level of brain microvascular dysfunction may not be proportional to impact severity. We analyzed 34,886 vessel segments from 11 SHAM mice; 25,197 vessels from 12 mild TBI mice; and 47,568 vessels 10 moderate TBI mice were extracted from the imaged volumes and used for the morphological analysis.

Mild to moderate impacts do not lead to robust behavioral changes

Neurological deficit tests revealed no differences between groups (see Tables S3–S5), ruling out severe injuries. ⁵⁹ There was no influence of injury on sensorimotor function at any time point. Activity level (via the novel cage test) showed increased grooming activity in the mild group (p = 0.009). Nestlet shredding for nest building, a metric of general wellbeing, showed the injured groups had incomplete nests at the acute and subacute injury stages. Preliminary behavior analyses thus suggest the presence of mild, but not severe deficits.

Remodeling of microvascular morphology post moderate TBI

Intravascularly injected dextran dye labeled the blood plasma, and volumetric structural stacks of the cortical tissue were acquired as deep as 1100 µm in the case of Alexa 680 dextran injected animals in sham, mild and moderate injury cohorts on either hemisphere (Figure 2). The vascular segmentation model, when compared to manual rating achieved, a mean Dice score of 0.76 ± 0.13, a mean precision of 0.75 ± 0.11, mean recall of 0.77 ± 0.16, mean Hausdorff 95% distance of 18.68 ± 14.36 μm, and mean surface distance of 3.39 ± 3.01 μm. Following vascular segmentation, a vascular graph was extracted and microvascular morphometric analysis performed using our deep-learning pipeline. ⁹² We first examined the morphological patterning for angiogenesis signatures, as angiogenesis and rarefaction have been reported following experimental TBI^65,95–98 and angiogenesis has been seen by 14–21 days in chronic hypoxia mouse model, ⁹⁹ as hypoxia likely happens to some degree in TBI. ¹⁰⁰ Vascular volume density was significantly larger ipsicontusional to moderate impacts when compared to that of shams (20 ± 17%, p = 0.02 Wilcoxon); however, there was no significant difference in vascular volume density for the mild impact cohort versus sham (Figure 3(a) and (b)). Previous morphological examinations by our group had also reported a significantly increased vascular volume within the TBI-injured cortex relative to that of shams. ⁶³ Since vascular number density, vascular branch point density, and endothelial cell density for were not significantly different in moderate relative to sham when examining the acquired cortex as a whole (Figure S2), we ascribed the increased vascular volume density in moderate impacts to vasodilatation, that is, tone dysregulation. To investigate this possibility, we compared the vessel diameters of capillary (vessels <10 μm in diameter) and larger vessels (>10 μm in diameter) of TBI cohorts versus sham animals. This cutoff was supported by prior work^101,102 and the current vessel size distribution (Figure S6) indicating that only a small proportion of vessels were in the 7–10 μm range. We found a significant increase in mean capillary diameter for both ipsicontusional mild TBI (0.25 ± 0.15 μm, 5.6 ± 3.2%, p < 0.0001, Wilcoxon) and ipsicontusional moderate TBI (0.03 ± 0.12 μm, 1.1 ± 2.6%, p = 0.043, Wilcoxon) compared to shams (Figure 3(c) and (d)). Furthermore, the contracontusional hemisphere’s capillaries in the moderate TBI cohort also showed a significant difference from shams, as a decrease in diameter (−0.09 ± 0.13 μm, p = 0.003, Wilcoxon) (Figure 3(d)). Interestingly, the capillary diameter in the moderate TBI cohort showed elevated intravessel heterogeneity, manifested as a significant difference in intravascular diameter standard deviation (vessel’s diameter variation of 0.006 ± 0.049 μm) in ipsicontusional hemispheres relative to sham hemispheres (p < 0.0001, Wilcoxon) (Figure 3(e)), which was not observed in the mild TBI group (Figure S2). Larger vessels (>10 μm in diameter) in the ipsicontusional moderate TBI also showed a significant increase, of 1.12 ± 0.71 μm, in mean diameters relative to those in sham hemispheres (p < 0.001, Wilcoxon) (Figure 3(f)).

Figure 2.

Network-level visualization of the cortical microvasculature. Maximum intensity projections (MIP) under normocapnia showing representative two-photon fluorescence microscopy volumetric (XYZ) stacks from ipsicontusional (a–c) and contracontusional (d–f) hemispheres. Sham injury: (a, d); Mild repetitive traumatic brain injury: (b, e); Moderate repetitive injury: (c, f). Scale bars: (a) = 150 μm, (b) = 100 μm, (c) = 100 μm, (d) = 150 μm, (e) = 150 μm, (f) = 150 μm.

Figure 3.

Changes consistent with microvascular network remodeling evident with TBI. Microvascular volume density obtained from volumetric stacks for (a) mild and (b) moderate TBI: the microvascular volume was increased ipsicontusional to moderate TBI when compared to either sham (25 ± 15%, p = 0.016 Wilcoxon), or contracontusional to moderate TBI hemisphere (20 ± 17%, p = 0.02 Wilcoxon). Capillary (d < 10 μm) mean diameter for (c) mild and (d) moderate TBI: the capillary diameter was increased by 0.25 ± 0.15 μm ipsicontusional to mild TBI versus sham (p < 0.0001, Wilcoxon), and by 0.35 ± 0.04 μm contracontusional to mild TBI versus ipsicontusional to mild TBI (p < 1e-16, Wilcoxon). Moderate TBI led to a 0.05 ± 0.12 μm increase in ipsicontusional capillary diameter versus sham (p = 0.043, Wilcoxon) and a 0.12 ± 0.18 μm increase versus contracontusional to moderate TBI cortex (p < 0.0001, Wilcoxon). (e) Moderate capillaries showed elevated intravessel diameter heterogeneity with a vessel diameter variation of 0.006 ± 0.049 μm in ipsicontusional hemisphere relative to shams (p < 0.0001, Wilcoxon). (f) Large vessels’ (d > 10 μm) mean diameter for moderate TBI: ipsicontusional to moderate TBI larger vessels’ diameters increased by 1.12 ± 0.71 μm when compared to that of larger vessels in sham hemispheres (p < 0.0001, Wilcoxon), and by 0.93 ± 0.24 μm when compared to larger vessels contracontusional to moderate TBI (p = 0.0002, Wilcoxon). (g) Density of vessels with pericytes increased by 1907 ± 626 pericyte expressing vessels/mm³ (p = 0.045, Wilcoxon). Angiogenic metrics were re-examined within the surface layers only (i.e., down to the bottom of layer II/III at 304 µm). (h) Maximum intensity projection of vasculature from a sham mouse showing the pericyte cells (green) on the endothelial wall (red). Scale bar = 100 µm. (i) The significant increase of vessels within pericytes within the moderate cohort ipsicontusional compared to shams for microvascular volume density persisted (19 ± 19%, p = 0.032 Wilcoxon). Considering layers I–III only revealed an increased branch point density (j) (29 ± 12%, p = 0.020 Wilcoxon), increased vessel number density (k) (42 ± 11%, p = 0.001 Wilcoxon), and increased vessel length density (l) (20 ± 7%, p = 0.013 Wilcoxon). Regarding hemispheric differences, we observed changes ipsicontusional to moderate TBI when compared to the contracontusional moderate TBI: in vessel volume density (i) (17 ± 17%, p = 0.049 Wilcoxon), in vascular branch point density (j) (29 ± 14%, p = 0.009 Wilcoxon), and in vessel number density (k) (22 ± 10%, p = 0.034 Wilcoxon). (m) Mean extravascular distance in cortical layer I ipsicontusional cortex of moderate TBI cohort was lower than that of sham by 2.49 ± 2.43 μm (16.0 ± 15.7%, p = 0.008). (n) Mean extravascular distance in cortical layer II/III ipsicontusional cortex of moderate TBI cohort was decreased, relative to sham’s, by 1.67 ± 0.88 μm (14.3 ± 7.6%, p = 0.0007). The following number of subjects, volumes and vessels were used to generate the figure: Sham N = 11 (19 volumes, 31,283 vessels), Mild Contracontusional N = 5 (5 volumes, 3436 vessels), Mild Ipsicontusional N = 11 (14 volumes, 19,054 vessels), Moderate Contracontusional N = 9 (9 volumes, 16,253 vessels), Moderate Ipsicontusional N = 10 (12 volumes, 26,362 vessels).

We investigated capillary pericyte distribution, as pericytes and their adequate coverage play a key role in capillary tone, blood brain barrier permeability, vascular resistance, retrograde signaling, and blood flow distribution.^103–108 Here we observed a significant increase in density of vessels containing pericytes ipsicontusional to the moderately impacted tissue (p = 0.045, Wilcoxon) (Figure 3(g)), though the average number of pericytes per vessel was not significantly different, suggesting that the aforementioned difference was not due to an excess proliferation of pericytes (Figure S2(h)).

As an increased number of pericyte containing vessels was not due to excess pericytes, and pericyte labeling via superfusion was only visible in the superficial layers, we re-examined the previously mentioned angiogenic metrics within the surface layers only, clipped down to the bottom of layer II/III (here estimated at 304 µm). ¹⁰⁹ Within this region, we continued to see a significant increase in microvascular volume density post ipsicontusional to moderate TBI, when compared to sham (19 ± 19%, p = 0.032 Wilcoxon) (Figure 3(i)), as well as increased branch point density (29 ± 12%, p = 0.020 Wilcoxon) (Figure 3(j)), increased vascular number density (42 ± 11%, p = 0.001 Wilcoxon) (Figure 3(k)), and vascular length density (20 ± 7%, p = 0.013 Wilcoxon) Figure 3(l)). Additionally, within the moderate TBI cohort, compared to contracontusional cortex, the ipsicontusional cortex showed increased vascular volume density (17 ± 17%, p = 0.049 Wilcoxon) (Figure 3(i)), vascular branch point density (29 ± 14%, p = 0.009 Wilcoxon) (Figure 3(j)), and vascular number density (22 ± 10%, p = 0.034 Wilcoxon) (Figure 3(k)). Finally, the extravascular distance in the ipsicontusional cortex of the moderate TBI cohort was decreased relative to sham, by 2.49 ± 2.43 μm (16.0 ± 15.7%, p = 0.008) in cortical layer I (down to 69 μm, ¹⁰⁹ Figure 3(m)); and by 1.67 ± 0.88 μm (14.3 ± 7.6%, p = 0.0007) in cortical layer II/III (Figure 3(n)).

In the Cre XTdTomato mice containing endothelial cell labeling, we looked at the eccentricity of the endothelial cells. The endothelial cell shape and cell orientation can indicate chronic flow changes, as chronically reduced flows may produce rounder cells while higher flows may produce elongated cells.^110,111 We did not see any significant cell shape changes between cohorts.

As an aside, we did notice off-target tdTomato labeling contained within the parenchyma in our moderate TBI cohort, which was not present in shams (Figure S2(i) and (j)). It has been reported that a small amount of circulating cells can be positive in Tg(Tek-cre)1Ywa model ⁷⁵ and therefore could be circulating immune cells recruited to the damaged area, or macrophages with a shared progenitor cell as endothelial cells ¹¹² also capable of expressing tdTomato through off target expression. This could also be indicative of endothelial cell damage, with neuroimmune-responding cells such as microglia, macrophages, or blood-derived monocytes clearing tdTomato-containing debris, ¹¹³ or inflamed cell signaling via extracellular vesicles of endothelial origin spreading to parenchymal-residing cells. ¹¹⁴ We did not observe significant differences in the density of endothelial cells, suggesting that neither rarefaction nor angiogenesis was taking place at this stage (Figure S2).

TBI impaired arterial reactivity to hypercapnia bilaterally

Penetrating vessels play a critical role in cortical perfusion and are compromised in neurological diseases such as CADASIL, Alzheimer’s Disease, as well as aging.^115–118 Damage to just one penetrating vessel may affect cortical perfusion.^70,119,120 We thus evaluated the changes induced by hypercapnia on the penetrating vessels’ vRBC. In sham treated animals, we observed the expected increase in arteriolar RBC velocity (27 ± 15%), whereas mild TBI animals showed a decrease in RBC velocity (−20 ± 13%, p = 0.03 vs sham), while moderate TBI showed no change ipsicontusionally (3 ± 13%, p = 0.197 vs sham) (Figure 4(a) and Table 2; Table S9). With respect to the directionality of the ipsicontusional arteriolar responses to hypercapnia, hypercapnia-elicited decreases in vRBC were observed in 31% of arterioles in sham, 62% of arterioles in the mild TBI cohort, and 50% of arterioles in the moderate TBI cohort (Figure 4(b)). Similarly, contracontusionally, sham animals showed an overall increase in arteriolar vRBC with CO₂ inhalation (17 ± 24%), whereas mild injury animals showed a decreased vRBC to 10% CO₂ inhalation (−18 ± 23%; p = 0.15) and the moderate injury cohort showed a small increase (9 ± 14%; p = 0.94) (Figure 4(c) and Table 2; Table S9). A similar percentage of vessels exhibited decreased vRBC during hypercapnia in the contracontusional hemisphere as was seen ipsicontusionally: 33% of sham arteriolar, compared to 67% post-mild TBI and 54% post-moderate TBI (Figure 4(d)). In contrast to arterioles, we saw no cohort-wise differences in venular vRBCs responses to hypercapnia for either hemisphere (Table 2; Table S9 and Figure S3). Under normocapnic conditions, TBI cohort arterial RBC velocities were not significantly different from sham ipsicontusionally (MILD p = 0.59; MOD p = 0.59) or contracontusionally (MILD p = 0.97; MOD p = 0.23).

Figure 4.

Microvascular reactivity to hypercapnia is reduced in mild TBI. (a) and (b) Hypercapnic challenge elicited changes in arteriolar RBC velocities within the ipsicontusional hemispheres. (a) Whereas sham cohort showed the expected increase in RBC velocity to hypercapnia, the vRBC in vessels ipsicontusional to mild injuries decreased during hypercapnia, while vessels ipsicontusional to moderate TBIs showed no changes (Sham +27 ± 15% vs mild impact −20 ± 13% (p = 0.0309); sham vs moderate impact 3 ± 13% (p = 0.197)) (Sham N = 9 (16 samples), Mild N = 12 (17 samples), Moderate N = 10 (19 samples)), (b) Hypercapnic challenge led the majority of vessels in the sham treatment group to increase in RBC velocity, with only 31% of vessels decreasing their RBC velocity. In the mild TBI cohort, the majority (62%) of arterioles decreased their RBC velocity with hypercapnic challenge. In the moderate TBI cohort, 50% of vessels showed decreases in their RBC velocity with hypercapnia (Sham N = 9 (32 samples), Mild N = 12 (34 samples), Moderate N = 10 (38 samples)). (c) and (d) Hypercapnic challenge elicited changes in the arteriolar RBC velocity within the contracontusional hemisphere. (c) Hypercapnic challenge had a similar effect on the contracontusional as on the ipsicontusional hemispheres: Mild TBI cohort showed hypercapnia-elicited a net decrease in the RBC velocity, while sham cohort vessels increased their vRBCs and moderate TBI cohort vessels’ vRBC did not change (Sham +17 ± 24% vs. mild impact −37 ± 14% (p = 0.148); sham vs moderate impact 9 ± 14% (p = 0.945)) (Sham N = 4 (10 samples), Mild N = 5 (9 samples), Moderate N = 9 (13 samples)). (d) Contracontusionally, the majority of vessels in the sham treatment group increased their RBC velocity, while the mild and moderate impact cohorts had many vessels show decreased RBC velocity in response to hypercapnia (Sham N = 4 (20 samples), Mild N = 5 (18 samples), Moderate N = 9 (26 samples)). Linear regression (lm(TT_capillary_HC~TT_capillary_ NC) was applied to centered mean transit times of capillaries (aggregating data on both hemispheres) in hypercapnic (HC) state versus normocapnic (NC) state; with MILD and SHAM shown in (e) and MODERATE and SHAM shown in (f). In all cohorts, capillaries showed significant hypercapnia-induced flow increases (lm slope_SHAM = 0.22 ± 0.04, lm slope_{MILD TBI} = 0.33 ± 0.03, lm slope_{MODERATE TBI} = 0.29 ± 0.02), with significant differences in slope between MILD and SHAM: p < 0.0001, and between MODERATE and SHAM: p = 0.005, where the difference in slopes was assessed from the interaction term in the linear mixed effects model: lme(TT_capillary ~ state * treatment, random = ~1 | subject_id). The shaded region indicates the 95% confidence interval.

Capillary response to hypercapnia is reduced in mild TBI

Under physiological conditions, excess CO₂ in the blood during hypercapnic breathing results in vasodilation and increased brain blood flow, ⁴⁴ increased vRBC and decreased capillary transit time heterogeneity.^121–123 The mesoscopic vasodilatory response to hypercapnia is blunted in severe TBI clinically. ¹²⁴ Here, all mice showed the expected overall increase in capillary flow during hypercapnia versus normocapnia, and a smaller dispersion of capillary transit times during hypercapnia versus during normocapnia (Figure 4(e) and (f)) (slope_SHAM = 0.22 ± 0.04, slope_Mild TBI = 0.33 ± 0.03, slope_Mod TBI = 0.29 ± 0.02). However, the injury resulted in decreased capillary reactivity to hypercapnia when compared to the response in shams (mild TBI vs SHAM: p < 1e-5) (Figure 4(e)); and moderate TBI versus sham: p = 0.005 (Figure 4(f)). Capillary transit time heterogeneity was investigated, as excessively high and low transit times would produce poor oxygen tissue delivery and could be indicative of a microvascular disease state, during which an increase in cerebral blood flow would reduce delivery further. ¹²⁵ Under normocapnia, mean capillary transit times did not differ significantly from sham in either the ipsicontusional hemisphere (MILD p = 0.96; MOD p = 0.67) or the contracontusional hemisphere (MILD p = 0.76; MOD p = 0.74). During normocapnic state, Levene’s test for homogeneity of variance revealed an increased capillary transit time heterogeneity in the injury cohorts versus sham cohort (p < 0.001) (Figure S4). When examining the hemispheric differences in the cerebrovascular response to hypercapnia, the moderate TBI cohort revealed lateralization, whereby the ipsicontusional capillaries showed less transit time shortening than contracontusional capillaries (Table S11).

Discussion

Repetitive mild traumatic brain injury can result in a progressive accumulation of damage which typically goes undetected by conventional structural neuroimaging. Patients with repetitive TBI often have lingering symptoms and are at an increased risk for developing neurodegenerative diseases such as Alzheimer’s, Parkinson’s and chronic traumatic encephalopathy. Cerebrovascular injury is thought to underlie a large amount of the related long-term disability following TBI. ¹⁰⁰ In the present work, we used a 10% inspired CO₂ tension to evaluate the microscopic scale cerebrovascular reserve. Although baseline CBF was not directly assessed in this study, our prior ASL MRI work in the same TBI model and time point demonstrated ~50% hypoperfusion in peri-contusional cortex. ⁶³ In the present study, mean normocapnic capillary transit times and arterial RBC velocities were not affected by treatment either ipsi- and contracontusionally, indicating preserved baseline perfusion at the microvascular level. However, injured mice exhibited increased capillary transit time heterogeneity and reduced capillary reactivity to hypercapnia, revealing subtle microvascular impairments and functional disruptions in baseline blood flow distribution. These findings emphasize the importance of dynamic functional assessments, such as cerebrovascular reactivity challenges, to uncover impairments that may be missed by mean CBF measurements, particularly in response to mild or moderate TBI. Under physiological conditions, brain blood flow will increase during hypercapnia relative to normocapnia due to relaxation of smooth muscles cells in resistance vessels, i.e. arteries and arterioles, as a result of decreased pH, ¹²⁶ corresponding to a reduction in the transit time of a bolus of fluorescently labeled plasma passing through the vascular bed. ¹²⁷ At the same time, the heterogeneity of the microvessels’ transit times is expected to decrease during hypercapnia. ¹²⁸ In some preclinical models of neurological diseases for which TBI survivors may be at an increased risk of developing, microvasculature shows a reduced vascular reactivity and reduced transit time homogenization to hypercapnia.^129,130 Our examination of the cerebrovascular consequences of TBI at the microvascular level revealed cerebrovascular reactivity differences in the repeated injury mouse brain when compared to the uninjured sham mouse. Counterintuitively, the reactivity differences were more pronounced in the repeated mild versus moderate TBI, with only subtle differences in the repeated moderate injury. In turn, penetrating arteries in the mild TBI cohort exhibited an overall decrease in RBC velocities during hypercapnia (whereas an overall increase in RBC velocities is expected during a vasodilatory challenge, as was seen in the sham cohort). Such decrease was also observed contracontusionally in the mild TBI cohort. We believe the observed changes in mild TBI may indicate the vasculature is functioning near the limit of the reserve capacity for vasodilation given the lack of structural remodeling. The observed decreases in vRBC in penetrating arterioles could indicate the presence of vascular intracerebral steal phenomenon, or negative CVR, where blood is redistributed from regions with inadequate vasodilatory capacity toward regions with normally functioning vasculature.^131,132 In contrast, moderate TBI did elicit structural remodeling and baseline vessel calibers were maintained during hypercapnia (relative to normocapnia), with no apparent “steal” phenomenon. In short, we posit that the endogenous recovery processes are triggered only at moderate levels of injury: the ensuing structural remodeling mitigates post trauma vascular dysfunction. The absence of a dose-response relationship between mild and moderate TBI in our study may reflect recruitment of distinct pathophysiological processes,^133,134 whereby additional responses are elicited under greater severity of injury, influencing both functional and structural changes.

Mesoscopic scale manifestations of the changes in the microvasculature may be detected on neuroimaging. Indeed, reduced CVR post TBI has been demonstrated in the clinical populations via non-invasive imaging,^{48,49,55,135,136} with multiple reports of a CVR dysfunction in TBI patients despite maintained baseline cerebral blood flow.^49,55 Our previous experimental study revealed, using arterial spin labeling MRI, a 70% reduction in regional CVR in repetitive mild TBI cohort compared to sham. ⁶³ Other preclinical studies have also seen alterations in CVR in TBI models. Lynch et al observed, using laser speckle imaging, a 30%–46% decreased CVR to 5% CO₂ in repetitive TBI mouse model at the chronic stage post injury both proximal and distal to injury. Accompanying was a reduction in markers of cells regulating vascular tone and resistance—such as smooth muscle cells and pericytes—indicating alterations to the vascular wall components but preserved vascular density at 3 months post injury, with a trend toward progressive recovery. ⁶⁴ Wei et al reported a 30% reduction in pial vascular reactivity to 5% CO₂ in the semi-acute to chronic stage post single impact in rats, in addition to reduced vasodilation in response to vasodilators targeting the smooth muscle cells or the endothelial cells. ¹³⁷ Our results add to these studies’ observations of microvascular dysfunction following TBI, by assessing individual vessels’ structure and function. We here observed a blunted change in the capillary transit during hypercapnia as well as aberrant penetrating arterial flow whereby the majority of vessels exhibited flow decrements during hypercapnia.

As TBI involves a multifaceted secondary injury cascade, there are a multitude of processes involved in cerebrovascular regulation to hypercapnia which could be affected by trauma, including damage to endothelial cells, pericytes and smooth muscle cells, altered arterial electrical communication and altered levels of vasoactive signaling (e.g., nitric oxide) and/or diminished reactivity to dilatory stimuli, increased basal tissue acidosis producing baseline dilation, damage to the extrinsic sympathetic pial vessel innervation, impaired dilatory signal propagation through the vessel bed, or dysfunctional autoregulation.^64,138–147

In addition to functional deficits, we also observed morphological changes consistent with microvascular network remodeling. When compared to the sham treatment mice, the moderate injury cohort showed a significant increase in vessel volume density (25% larger) ipsicontusional to the injury. Changes in cerebrovascular density have been observed within experimental TBI models, with a transient decrease in vessel density acutely post TBI,^96,97 and an increase in vessel density at later time points post injury.^65,95,97 Our vessel volume density increase could result from angiogenesis and/or dilations. Indeed, the larger vessels (>10 µm diameter) as well as capillaries (<10 µm diameter) were significantly larger ipsicontusional to moderate TBIs, when compared to shams, suggesting sustained dilations. In contrast, capillaries within the contracontusional hemisphere in the moderate injury group showed a significant decrease in diameter compared to shams. Mild injury, in turn, did not result in such capillary changes. In the case of the moderately injured hemisphere, the capillaries showed significant increase in capillary diameter heterogeneity. Increased capillary diameters have been noted previously following experimental TBI up to 4 weeks post injury.^65,148

We next tested whether this heterogeneity and the observed basal capillary dilatation were due to loss of pericyte coverage. Acute pericyte loss (under 24 h) and later proliferation (at day 5) has been reported in experimental mouse controlled cortical impact TBI. ¹⁴⁹ However, the average number of pericytes per vessel was conserved here. Given the loss and rebound of pericyte coverage in Zehendner et al., ¹⁴⁹ the pericytes may have been in a state of recovery at our sub-acute time-point, resulting in dilatory dysfunction. Our results did show, however, an increased number of vessels containing pericytes ipsicontusional to moderate TBIs (the pericytes’ dye ¹⁵⁰ did not penetrate to depths beyond layer II/III, so our observations were restricted to the top three cortical laminae). The increased number of pericyte-covered vessels, combined with the preserved number of pericytes per vessel, suggests potential angiogenesis, so we re-focused the analyses to the superficial layers that are experiencing the most impact force in the current TBI model.^98,151,152 Restricting the analyses to the superficial layers did reveal significant changes in angiogenesis metrics, with increased vascular volume density, increased branch point density, increased vascular number and increased length densities; as well as decreased mean extravascular distance ipsicontusional to the moderate injury. Notwithstanding, we did not observe any differences in endothelial cell density although label within the endothelial cell affords sensitivity to new yet unperfused vessels as well as collapsed vessels. In a chronic hypoxia mouse model, new sprouts formed patent connections by 14–21 days. ⁹⁹ On the whole, our findings suggest that the observed vascular density changes reflect active remodeling rather than atrophy. ¹⁵³ The microvascular changes seen here may set the stage for maladaptive remodeling, impaired hemodynamics, and cortical atrophy that typically emerge at chronic time points following TBI.^154,155

Limitations

It is important to note several limitations related to this study. For one, our classification of capillaries by diameter (<10 µm) may lead to some misclassification of vessels and would best be replaced via use of specific markers of blood vessel wall components. Our open-source analysis pipeline (https://github.com/AICONSlab/novas3d/tree/main) does allow users to adjust cut-off diameter specifying capillaries. Future work using molecular markers or high-resolution structural mapping will be important to refine vessel categorization. In addition, our animals were anesthetized with isoflurane for the injury induction and the imaging sessions, which affects cerebral hemodynamics. Isoflurane can elevate baseline CBF, potentially reducing apparent vasodilatory reserve to CO₂; however, we kept the mice ventilated with low levels of Isoflurane (<1.25%), which has been shown to preserve hemodynamic responsivity).^156–158 It was not possible to perform the imaging session under awake conditions, as exposure to 10% CO₂ in awake animals often elicits a hyperpnea response, whereas anesthesia combined with mechanical ventilation allows precise control over the degree of induced CO₂ tension elevation and enables lengthy imaging data acquisition and thus acquisition of robust imaging data sets. Isoflurane, nonetheless, is not only a vasodilator, ¹⁵⁹ but potentially also neuroprotective. ¹⁶⁰ The vasodilatory effect of inhalation anesthetics such as isoflurane is dose dependent, yet cerebrovascular dilation to CO₂ inhalation is preserved ¹⁶¹ at low dose isoflurane (0.75%–1.25%). A recent study by Evans et al. showed preserved hemodynamic activity under 1.5% isoflurane with no noticeable physiological effects from repeated usage. ¹⁶² Notwithstanding, we do expect some blunting of the cerebrovascular reactivity due to isoflurane-induced dilation. Further, the closed cortical impact does not recapitulate all aspects of clinical TBIs: notably, we used stereotaxic head immobilization, curbing extracranial damage and head rotational movements characteristic of clinical impacts. ⁶² However the use of the head immobilized closed impact model allows for repeated injury at a precise location, permitting imaging at a later time point within that exact injury location. A further limitation is that vessel diameter changes in vivo reflect the combined influence of vascular tone and structural remodeling. Disentangling these requires direct assessment of passive vessel diameter, typically achieved by pressure-diameter measurements conducted in Ca²⁺-free buffer. ¹⁶³ Of note, Mughal et al. reported functional dysregulation in TBI mice at an earlier time point (4–7 days) demonstrated by impaired K⁺-evoked responses in vivo and in ex vivo pressurized vessels under Ca²⁺ free conditions, alongside Kir2.1 channel dysfunction, despite preserved capillary architecture. ¹⁶⁴ Further studies incorporating passive-condition measurements at subacute time points will be essential to determine whether the vessel changes we observe arise predominantly from tone loss, structural remodeling, or both.

Conclusion

The current study provided evidence of reduced cerebrovascular reserve following mild impacts, as evidenced by the paradoxical blood flow decreases upon CO₂ inhalation. The three moderate impacts abrogated the response to CO₂ in the “moderate” TBI cohort that exhibited significant angioarchitectural differences including baseline diameter increases, as well as angiogenesis within the superficial cortex. The moderate impacts were severe enough to trigger vascular adaptations that were absent following mild TBIs, where the microvessel function post injury was at or near the reserve capacity. A lack of a strict dose-response between injury severity and outcome in our data highlights that TBI pathophysiology is not solely determined by injury magnitude. Instead, distinct endogenous repair mechanisms and injury cascades may be recruited at different severity thresholds, resulting in nonlinear and sometimes divergent functional and structural outcomes. The present findings support the importance of developing treatments to target the brain microvasculature post repetitive TBIs, especially given the role neurovascular dysfunction plays in TBI as well as in diseases for which TBI patients are at an increased risk.