Cerebral microcirculatory failure after subarachnoid hemorrhage is reversed by hyaluronidase

Abstract

Aneurysmal subarachnoid hemorrhage remains one of the more devastating forms of stroke due in large part to delayed cerebral ischemia that appears days to weeks following the initial hemorrhage. Therapies exclusively targeting large caliber arterial vasospasm have fallen short, and thus we asked whether capillary dysfunction contributes to delayed cerebral ischemia after subarachnoid hemorrhage. Using a mouse model of subarachnoid hemorrhage and two-photon microscopy we showed capillary dysfunction unrelated to upstream arterial constriction. Subarachnoid hemorrhage decreased RBC velocity by 30%, decreased capillary pulsatility by 50%, and increased length of non-perfusing capillaries by 15%. This was accompanied by severe brain hypoxia and neuronal loss. Hyaluronidase, an enzyme that alters capillary blood flow by removing the luminal glycocalyx, returned RBC velocity and pulsatility to normal. Hyaluronidase also reversed brain hypoxia and prevented neuron loss typically seen after subarachnoid hemorrhage. Thus, subarachnoid hemorrhage causes specific changes in capillary RBC flow independent of arterial spasm, and hyaluronidase treatment that normalizes capillary blood flow can prevent brain hypoxia and injury after subarachnoid hemorrhage. Prevention or treatment of capillary dysfunction after subarachnoid hemorrhage may reduce the incidence or severity of subarachnoid hemorrhage-induced delayed cerebral ischemia.

Introduction

Despite numerous attempted interventions aneurysmal subarachnoid hemorrhage (SAH) remains one of the most deadly neurovascular disorders.¹ The neurologic deficits that develop in the days to weeks following SAH are particularly pernicious and continue to pose an intractable challenge for patients and healthcare providers. The failure of recent clinical studies (CONSCIOUS-2 and -3) targeting vasospasm to show benefit has redirected the field away from angiographic vasospasm as the dominant mechanism of delayed injury after SAH. Nimodipine, a dihydropyridine-type calcium channel blocker is currently the sole FDA-approved treatment available for patients with SAH for which there is conclusive evidence, and the benefits are marginal at best.² A systematic review found no difference in arterial diameters with CCBs indicating that the therapeutic mechanism of CCBs is independent of changes in lumen sizes of angiographically visible vessels.³ Thus, to provide improved therapies for SAH patients, we must look beyond arterial vasospasm.

As was first proposed by Herz et al.⁴ four decades earlier we hypothesized that capillary blood flow is acutely reduced after SAH independent of blood flow changes caused by vasospasm, leading to brain hypoxia and neuronal loss. Additionally we hypothesized that shedding of the endothelial glycocalyx, a glycoprotein layer (300–500 nm thick) secreted by endothelial cells on their luminal surface⁵ would improve microvascular circulation and reverse brain hypoxia. Prior ex vivo histological studies have suggested that SAH causes changes in capillary structure or function, but the degree to which these changes alter blood flow in vivo remains relatively unexplored.⁶ In addition, how upstream arterial vasospasm changes parameters of capillary blood flow after SAH is undefined, and the impact of altered capillary blood flow on oxygen delivery and neuronal loss remains unknown.

Materials and methods

Ethics statement

All mouse experiments were written and performed in accordance with and approved by the Institutional Animal Care and Use Committee of the University of Rochester and in accordance with the ARRIVE guidelines

Mouse preparation, SAH injections

Mice were randomly grouped by coin flip or die roll (depending on number of groups). Mice were only included if they survived to the end of each experiment. FVB/NJ mice (Jackson Laboratory), 8–12 weeks old, male and female, were fasted overnight and anesthetized with ketamine (0.12 g/kg body weight) and xylazine (0.01 g/kg) intraperitoneally. The femoral artery and vein were cannulated with a polyethylene catheter (PE-10); the arterial catheter was then attached to a microsyringe preloaded with a small volume (10 µL) of heparinized saline. The technique for injecting into the cisterna magna is as previously described.⁷ Briefly, the head was fixed in a stereotactic apparatus and the skull was exposed at the occipitocervical junction; 60–80 µL of autologous arterial blood was injected in the cisterna magna over 1 min. Similar procedures were used for sham-treated mice, except artificial CSF was used for injections (126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 2 mM MgCl₂, 2 mM CaCl₂, 10 mM glucose, and 26 mM NaHCO₃ at pH 7.4). After injection, if mice were to be used for 2-photon laser scanning microscopy (2-PLSM), the skin incision was closed with 4–0 monofilament nylon suture and mice were recovered. For SAH or saline injections with simultaneous intracranial pressure (ICP) and laser Doppler recording, the injections were done as described above, but the scalp incision was extended to allow for exposure of the calvarium overlying the cerebral hemispheres. Through a small burr hole over the left somatosensory cortex, an ICP monitor (Millar) was inserted and the signal was converted by a DigiData 1332 A interface (Axon Instruments) with an interval of 200 s and recorded with the pCLAMPEX 9.2 program (Axon Instruments). Through the same incision over the right somatosensory cortex, relative changes in local perfusion was detected by laser Doppler flowmetry (PF5010 Laser Doppler Perfusion Module with a microtip, PR 418-1, Perimed), and the signals were converted as described for the ICP monitor.

Mouse preparation, in vivo imaging

Techniques used for in vivo imaging are as previously described.⁸ Briefly, 48 h after induction of SAH or sham, mice were anesthetized with ketamine and xylazine as above, and the femoral artery was catheterized along with the femoral vein. The mice were intubated and artificially ventilated with a small-animal ventilator (SAAR-830, CWE). An open cranial window was then prepared as described previously.⁸ A craniotomy (3 mm diameter) was made over the primary somatosensory cortex, centered 1–2 mm posterior to the bregma and 2–3 mm lateral to midline. The dura was removed, and a custom-made metal plate was glued to the skull with dental acrylic cement. Sulforhodamine 101 (to fluorescently label astrocyte cell bodies and endfeet) was loaded as described previously,⁹ dissolved in aCSF, and the cranial window was then covered with NaCl 0.9% solution containing 1% agarose (37℃) with a glass coverslip then positioned to provide stability against brain pulsations. FITC-dextran (Sigma, 2000 kDa) was injected intravenously as previously described to outline the microcirculation.⁸ Body temperature was maintained at 37℃ by a heating blanket (BS4, Harvard Apparatus). Blood gasses were analyzed with a Rapidlab 248 (Bayer, sample size 40 µL). Periodic (every hour) fluid resuscitation was provided by injection of lactated Ringers solution (100 µL) i.p., and we completed experiments only if physiological variable remained within normal limits. We used a custom-built microscope attached to a Ti:sapphire laser (Tsunami, Spectra-Physics) and a scanning box (FV300, Olympus) using Fluoview software and a 20 × water immersion objective (0.95 numerical aperture, Olympus) for two-photon imaging. FITC-dextran and sulforhodamine were exited at 825 nm, and we detected emitted light with 525/50 and 607/45 filters, respectively. Laser power was on average 5–10 mW. Surface pial vessels were visualized under brightfield optics to confirm identity as arteries or veins prior to quantification with LSM. For measurements of pial vessels using LSM, the cranial window was made by thinning the bone but the inner table of the skull, as well as the dura, remained intact (and sulforhodamine application was not performed), except as described below. Naïve-control mice were used in place of sham-control for certain experiments enumerated below, as we found no difference between the two groups when we measured arteriolar diameter (p = 0.399, one-way Student t-test) and capillary RBC velocity ((Mann Whitney p < 0.001 [p = 0.0125, Bonferroni correction]).

Endothelin-1 superfusion

Naïve mice were prepared for in vivo imaging of capillary flow, and a coverslip was mounted over the cranial window and the space over the brain surface was filled with aCSF. Polyethylene tubing was then mounted at one side of the cranial window, and an outlet was fashioned at the other side. During imaging, baseline capillary flow data was collected from 5 to 10 capillaries per mouse and baseline surface vessel diameters were collected, and then 50–60 µL of endothelin-1 (ET-1) dissolved in aCSF was superfused from a Hamilton syringe connected to the PE tubing over 5 min. Data collection started again after superfusion was complete, and typically took 15–30 min to complete. The same capillaries and arteries measured before ET-1 superfusion were measured after ET-1 perfusion. Different mice were used for each of the ET-1 concentrations examined (1, 5, and 10 µM), and a separate set of mice were tested with superfusion of aCSF alone.

Hyaluronidase treatments

Type IV-S hyaluronidase from bovine testes (Sigma) was given over 15 s in a bolus dose of 100 µL working concentration via the femoral venous catheter to effect an end plasma concentration of 150 U/mL plasma assuming an average of 1.5 mL blood volume in a 25-g mouse. Stock hyaluronidase (10,000 units/mL) was made by dissolving lypholized powder in PBS pH 7.4 with 2% glycerol (for cryoprotection). Bolus dose was made by diluting the stock solution in PBS pH 7.4. This effectively resulted in an end plasma concentration of 0.03% glycerol. Sham-injected control mice were given bolus infusions of a similar volume of PBS with glycerol (0.03%). Mean arterial pressure (MAP), ICP, and HR were recorded during each experiment, and blood gasses were measured before and after hyaluronidase treatment. Another set of SAH mice were treated with i.v. injections of hyaluronidase or vehicle at 12 and 24 h after SAH, and then allowed to survive for 7 days, at which point they were sacrificed and their brains processed for stereologic counting of neurons (see below).

Fluorescent microsphere injections

To determine whether non-perfusing capillaries might still have movement of blood plasma, in a subset of naïve control and SAH mice the carotid artery ipsilateral to the cranial window was catheterized prior to imaging. At the time of LSM, red fluorescent latex microspheres (0.5 µm, Invitrogen) were hand-injected through the catheter (20–30 µL).

Hypoxia probe immunohistochemistry

After 48 h of induction of SAH or sham-controls, mice were infused with hypoxyprobe (HP1-100Kit, 60 mg/kg body weight, NPI) and subsequently sacrificed 1 h after infusion. Brains were removed and post-fixed in PBS containing 4% paraformaldehyde overnight at 4℃. Coronal 100-µm-thick vibratome sections of brain slices were prepared. Sections were blocked with 5% normal donkey serum (Jackson ImmunoResearch Laboratories) and 0.5% Triton X-100 (Sigma-Aldrich) for 30 min at room temperature. Sections were then incubated with a monoclonal antibody to Hypoxyprobe (HP1-100Kit, 1:50) and a polyclonal antibody to MAP2 (AB5622 1:1000, Millipore) at room temperature for 2 h. Sections were then washed and incubated in secondary antibodies (Jackson ImmunoResearch Laboratories) and counterstained with 4′,6′-diamidino-2-phenylidole (DAPI, D-21490, 1:5000, Invitrogen). Images were collected with a confocal laser-scanning microscope (FV500, Olympus).

Examination of effects of SAH on pericytes, astrocytes, endothelial cells, neuron morphology, and number

To visualize pericyte morphology in vivo, we used transgenic adult mice expressing DsRed in pericytes under the NG2 promoter (Jackson Laboratory). To visualize endothelial cell morphology, we used transgenic mice expressing eGFP in endothelial cells under the Tie-2 promoter (Jackson Laboratory). To visualize neuronal morphology, we used transgenic mice expressing YFP in excitatory neurons under the Thy1 promoter (Jackson Laboratory). Mice underwent SAH injections as described above, and after 36–48 h survival, mice were prepared for in vivo imaging as described above. Injection of a fluorescent dextran allowed visualization of vessel lumens during in vivo imaging. For examination of neuronal morphology and endothelial cell volume, mice underwent SAH injections and after 36- to 48-h survival were sacrificed and perfused with 4% paraformaldehyde in PBS. Sham mice were used as controls for both in vivo imaging of pericytes and endothelial cells as well as ex vivo imaging of neurons and endothelial cells. For examination of neuronal morphology, coronal 50 µm vibratome sections were prepared and imaged using a confocal microscope (FV500, Olympus). YFP was excited at 455 nm and emission collected with a 515 nm emission filter with 50 nm bandwidth. To evaluate the effects of SAH on neuron number, FVB/NJ mice were subjected to SAH and perfused 7 days after SAH. Coronal 50-µm vibratome sections were prepared, and cresyl violet staining was performed. Sections were visualized with brightfield optics, and optical sections were then analyzed for neuron number in the cerebral cortex and hippocampus using StereoInvestigator software (optical fractionator method). Astrocyte cell number and endothelial cell volume were calculated using custom-made software (Matlab Inc.) as described previously.¹⁰

Measurement of brain tissue oxygen tension

We measured tissue pO₂ using a modified Clark-type polarographic oxygen microelectrode with a guard cathode for tissue (3–4 µm tip; Unisense) as previously described.¹¹ We calibrated the electrode before and after each experiment using distilled H₂O equilibrated with 8% O₂ balanced with N₂; air; 100% N₂; and a reductant solution of 0.1 M ascorbate and 0.1 M NaOH.

Analysis of LSM data

Velocity of RBCs and fluorescent beads was measured using line-scanning with 2–3 ms temporal resolution as described previously,¹² using 1- to 5-s long line scans analyzed for average velocity at every 30 ms using a modification of an automated routine in Matlab (Mathworks) kindly provided by Dr Chris Schaffer (Cornell University).¹³ Pulsatility was defined as the standard error mean (S.E.M.) divided by the mean velocity of RBCs. Capillary diameter, astrocyte endfeet thickness, and arteriole endfeet thickness were measured by taking multiple XYZ imaging stacks of capillary profiles orthogonal to the plane of imaging and quantified using ImageJ software (NIH).

Measurement of non-perfusing capillary length was performed using the “spaceballs” stereology technique,¹⁴ a method derived from Buffon’s principle in which the probe is a sphere virtually embedded in the tissue of interest and the parameter counted is the number of profiles that transect the sphere’s edge. Briefly, XYZ stacks of capillary images (250 µm × 250 µm × 100 µm collected every 4 µm) were collected in vivo using Fluoview software and analyzed using StereoInvestigator. Capillaries were identified by their diameter (<10 µm) and presence of a single column of RBCs when the capillary was perfusing. Perfusing capillaries were defined by the presence of moving RBCs within the capillary at the time of imaging (typical image acquisition time frame 30–60 s). Because of an artifact caused by fluorescence bleed-through above and below the plane, intersections of a sphere and capillary were only counted if it occurred at the plane of the clearest image. In all, 36–42 sites were visited in a systematically random boustrophedon-like pattern per each XYZ stack, and the radius of each spaceball was set at 20 µm. Sources of XYZ-stacks and line-scans were blinded during rating and analysis.

Evans Blue quantification of blood–brain barrier disruption

Mice were injected with 4 mL/kg of 4% albumin in PBS that had been pre-tagged with 2% Evans blue, as described elsewhere.¹⁵ For positive controls, mice received intravenous injection of 500 uL of 20% mannitol 5 min after injection of Evans blue. After 1-h circulation of Evans blue, mice were transcardially perfused with cold PBS (50 mL), and the brains were extracted. Tissue was homogenized in 500 µL TCA (7%), optical density was measured using spectrophotometry at 620 nm, with standardization performed using known Evans blue concentrations, and results were expressed as (grams Evans blue)/(gram of tissue).

Edema quantification

Mice were anesthetized with ketamine and xylazine as above, and were then decapitated. Brains were removed, the cerebellum and olfactory bulbs were discarded for standardization, and the remaining brain tissue was immediately weighed (wet weight). The brains were then dried for 72 h at 70℃, and reweighed (dry weight), and results were expressed as (wet weight-dry weight-)/wet weight. For positive controls, mice were anesthetized and then injected with distilled water (i.p.; 20% of body weight) and DDAVP 0.4 µg/kg (i.p.); 30 min later mice were decapitated and processed for edema determination.

Results

Cisterna magna injection of blood in mice closely mimics acute human SAH physiology

We first tested the validity of autologous blood injection into the cisterna magna as a mouse model of acute brain changes caused by SAH in humans. When an aneurysm ruptures, it causes a sudden increase in ICP, which rapidly equals or exceeds MAP, thereby causing cerebral blood flow (CBF) to transiently cease. This, in turn, leads to an increase in MAP and a decrease in heart rate (the Cushing reflex).¹⁶ Based on ICP and laser Doppler flowmetry measures of CBF (LDF, expressed as % of baseline), autologous blood injection into the cisterna magna caused a sudden increase in ICP above MAP, coupled with a sudden decrease in LDF down to zero, followed by an increase in MAP and decrease in heart rate that closely matched human conditions (Figure 1a). These observations show that our mouse model closely approximates the acute consequences of aneurysm rupture in humans confirming prior findings in other small animal models of SAH.¹⁷

Figure. 1.

The physiological features of SAH are replicated in an experimental model of SAH in mice. (a) Representative traces of ICP (red), MAP, and laser Doppler flow (LDF, % of baseline) recorded for 30 min. (b) Brightfield images of surface vessels in somatosensory cortex in a sham-control mouse 48 h after injection of saline. Red arrows point to pial arteries and blue arrows point to pial veins (Scale bar, 500 µm). (c) Surface vessels of a representative mouse 48 h after injection of autologous blood (SAH). (Difference between sham control arterial diameter and SAH arterial diameter p = 0.006, t-test, see Figure 2.) (d) BBB breakdown, as detected by extravasation of Evans blue into brain tissue samples from control mice and mice 48 h after SAH (p = 0.325, N = 4–6 mice). Intravenous mannitol (0.5 mL of 20% mannitol in saline) induced a significant increase in Evans blue extravasation (positive control; differences in Evans Blue p = 0.0024, ANOVA). (e) Percent brain water content ([wet weight-dry weight]/[wet weight]) at different time points after SAH or water intoxication (positive control – i.p. injection of 10% body weight distilled water with DDAVP [0.4 µg/kg]). Each box represents its respective group average (p < 0.001, N = 5–7 mice). (f) Mouse feeding behaviors as measured by body weight were impaired after SAH (N = 6–8 mice, differences in body weight p < 0.001, ANOVA).

SAH caused surface artery spasm when viewed through a cranial window 48 h later (Figure 1b and c), but sham-controls had no visible effects on surface artery diameter. Using Evans blue spectrophotometry as an assay of blood–brain barrier (BBB) breakdown,¹⁸ we saw that our SAH model did not result in opening of the BBB (Figure 1d). Using brain water weight measurements, we found that our SAH model also did not develop brain edema (Figure 1e). Although our SAH model did not cause BBB breakdown or brain edema, it was clear that SAH did affect animal behavior; mice were much less active in terms of feeding and grooming, as demonstrated by 16% decrease in body weight by 48 h after SAH (Figure 1f), compared to a 3.4% decrease in body weight in controls.

SAH acutely disrupted capillary RBC movement unrelated to arterial vasospasm

We next quantified the extent of vasospasm found after SAH in our model, and tested the effect of arterial constriction on capillary blood flow compared to the effect of SAH. We used 2-PLSM to measure surface arteries through a cranial window that was thinned but not completely removed to avoid herniation-induced changes in surface vessels. Measurements of surface arteries in control and SAH mice showed that injection of blood into the subarachnoid space caused a 12% reduction in the diameter of surface arteries by 48 h after injection (Figure 2a and e). To assess SAH-induced changes in capillary blood flow in vivo, we used 2-PLSM through an open cranial window in sham-control (un-injected) and SAH mice to visualize blood flow through individual capillaries. In control brains 93–97% of capillaries are perfused with RBCs that transit through the capillaries in an orderly fashion (Figure 2b left), in concordance with classic studies in rats.¹⁹ In contrast, blood flow through capillaries was dramatically altered after SAH (Figure 2b right). Line scans taken from a random selection of capillaries within control and SAH mice were then assessed for differences in RBC velocity and RBC pulsatility. This showed that SAH causes a dramatic reduction in capillary RBC velocity and pulsatility compared to control mice (Figure 2f–g).

Figure. 2.

SAH causes a decrease in capillary RBC velocity that is not replicated by arterial constriction with ET-1 in naïve-control mice. (a) Two-photon images of pial arteries (gray) in a control mouse (left) and pial arteries constricted in an SAH mouse (right). Red arrows point to pial arteries (Scale bar, 50 µm). (b) 2-photon image of brain capillaries in control and SAH mice. White arrows point to capillaries with representative line scans in a control mouse (left) and an SAH mouse (right) (Scale bar, 20 µm). (c) Two-photon images of pial arteries at baseline and after superfusion of 5 nM ET-1 (Scale bar, 50 µm). (d) Capillaries before and after ET-1 superfusion with line scans of capillaries at the white arrows (Scale bar, 20 µm). (e) Quantification of changes in pial arterial diameter. Arterial constriction after SAH and after 5 nM ET-1 superfusion are statistically equivalent (ANOVA with Tukey test p = 0.05, Holm-Sidak method unadjusted p = 0.321), 10 nM ET-1 caused a more than twice the degree of arterial spasm (p < 0.05). (f) Changes in RBC velocity after superfusion of ET-1. The percent decrease in RBC velocity after SAH is significantly larger than the percent decrease in RBC velocity in the ET-1 mice (Mann-Whitney p < 0.001 [p = 0.0125, Bonferroni correction]) even at the higher concentration of ET-1. (g) RBC pulsatility after superfusion of ET-1. ET-1 at all concentrations failed to alter capillary pulsatility, but SAH caused a very significant decrease (Mann Whitney p < 0.0001 [p = 0.0125, Bonferroni correction], n = 40–60 capillaries, N = 4–6 mice in each group).

To test the effect of arteriolar vasoconstriction on downstream capillary flow we superfused the brain surface with endothelin-1 (ET-1) and measured surface artery diameter, capillary RBC velocity, and pulsatility before and after. ET-1 is a potent vasoconstrictor that is increased in the brain after SAH and has been implicated as a cause of vasospasm and DIND; inhibition of ET-1 results in reduced arterial vasospasm in animal models and in humans after SAH.²⁰ Superfusion of ET-1 caused dose-dependent arterial constriction, and at 5 nM was statistically indistinguishable from the arterial spasm caused by SAH in our mouse model (Figure 2c–e). However, line scans from capillaries before and after ET-1 superfusion showed no significant change in RBC velocity or pulsatility at 5 nM ET-1 (Figure 2d, f–g). Lower dose ET-1 (1 nM) had no effect on surface arteries, capillary RBC velocity or capillary RBC pulsatility, and higher dose ET-1 (10 nM) caused more than double the reduction in arterial diameter seen after SAH. Only at this higher degree of surface artery constriction (10 nM ET-1 caused a 28% decrease in arterial diameter, compared to 12% decrease after SAH) did ET-1 cause a change in RBC velocity similar to SAH, but there was still no change in RBC pulsatility compared to the decrease in RBC pulsatility seen after SAH (Figure 2g).

SAH causes capillary narrowing that prevents RBC flow without capillary occlusion

In addition to causing reduced RBC velocity and pulsatility, SAH caused many capillaries to stop perfusing with RBCs (Figure 3a). We applied unique stereological techniques for in vivo data acquired from 2-PLSM, while also discriminating between capillaries that maintained RBC perfusion and those that were devoid of moving RBCs during the time of image acquisition (typically 30–60 s). These studies showed that the percent of total capillary length devoid of RBC perfusion (non-perfusing capillaries) in SAH mice is six times the value in control mice (Figure 3b). Quantification of total capillary length within cortical tissue in control and SAH mice showed no difference in the total capillary length between groups (control, 538.32 ± 24.64 mm/mm³; SAH, 548.86 ± 24.61 mm/mm³), suggesting that our fluorescent labeling and stereologic quantification techniques were not biased by capillary thrombosis or other phenomena that could obscure adequate intravascular labeling of all brain capillaries.

Figure. 3.

SAH causes capillaries to narrow, blocking RBC movement. (a) Left, collapsed optical section (20 µm) showing multiple non-perfusing capillaries lacking RBC movement (blue arrows) 48 h after SAH. Yellow arrows indicate perfusing capillaries containing RBC movement in the same field. High magnification outlined in blue from left (middle). High magnification outlined in yellow from left (right). Stereological analysis of optical sections shows that non-perfusing capillaries are much more common after SAH (mean ± SEM of non-perfusing capillaries in control and SAH mice, N = 12–13 mice, p < 0.001, t-test) (Scale bars, 50 µm and 10 µm in left and middle frames, respectively). (b) Histogram of percent non-perfusing capillaries. (c) Capillaries without RBCs permit passage of nanospheres. A solution containing nanospheres with a diameter of 0.4 µm was repeatedly delivered by a catheter inserted in the internal carotid artery ipsilateral to the cranial window. A representative field shows a perfusing (yellow bar) and a non-perfusing capillary (blue bar). Subsequent line scanning of the two capillaries shows that the nanospheres can travel in both capillaries, but with a slower velocity in the capillary with RBC (blue bar) (Scale bar, 10 µm). (d) Capillary diameters of perfusing capillaries are significantly larger than the average diameter of capillaries that are without red blood cells (N = 8 mice, n = 1900–2900 capillaries, p < 0.001, Mann Whitney test).

This, however, led us to question whether capillaries devoid of RBC movement are completely occluded; presumably, if fluorescent tracer gets into a capillary but RBCs do not, then the capillary must still be open to the movement of blood plasma. To definitely prove that capillaries without RBC movement were not occluded, SAH mice (n = 7) were prepared for 2-PLSM and the carotid artery ipsilateral to the cranial window was cannulated for injection of fluorescently-labeled latex beads (0.5 µm diameter) during imaging of capillaries with and without RBC perfusion. Numerous examples were found in SAH mice of absent RBC flow through capillaries with preserved transit of fluorescent beads, and presumably blood plasma (Figure 3c), and line scans from capillaries with and without RBC perfusion showed bead transit confirming that capillaries without RBC perfusion are not completely occluded. We measured the diameters of capillaries in naïve-control brains and in SAH brains; capillaries with RBC perfusion in control and SAH brains had similar diameters, but the capillaries without RBC perfusion were 16% smaller (Figure 3d).

Capillary constriction is not caused by pericyte constriction or astrocyte swelling

To test whether pericytes contribute to capillary flow restriction in either naïve-control or SAH mice, we used transgenic mice expressing DsRed under the NG2 promoter, which results in selective DsRed fluorescence in pericytes. In adult mice used for this study, greater than 99% of DsRed-positive cells were located along the vasculature (Figure 4a–c) in the expected position of pericytes. Transgenic mice were subjected to SAH or control (no cisternal injection), and then prepared for in vivo imaging of the fluorescent pericytes as well as an intravascular fluorescent tracer. Capillary lumen diameter in control and SAH mice was measured at the location of pericytes as well as 10–20 µm away along the same capillary (Figure 4d). Our measurements failed to demonstrate any evidence of focal capillary narrowing co-localized with pericytes in non-perfusing capillaries in either control or SAH mice. Conceivably, capillaries that are collapsed due to pericyte constriction would inherently fall in a smaller size range of capillaries (e.g. less than 3 µm in diameter), but our measurements in this subset of non-perfusing capillaries also failed to demonstrate any evidence of focal capillary narrowing colocalized to the presence of pericytes. We next measured astrocyte endfeet thickness in vivo in control and SAH mice, which revealed no difference between groups (Figure 4f–g).

Figure. 4.

Capillary flow abnormalities after SAH are not caused by pericyte constriction or swelling of astrocyte endfeet. (a) Images captured with confocal microscopy from cortical slices in transgenic NG2::DS-red mice stained astrocytes with monoclonal GFAP, endothelial cells with CD31 and DAPI nuclei stain. White arrows point to pericyte cell bodies and processes arrayed along the endothelial cells of the capillary beds (Scale bar, 50 µm). (b) White arrows point to pericyte cell bodies and green arrows point to GFAP-positive astrocytes hugging the capillary wall (Scale bar, 15 µm). (c) White arrows correspond with pericyte cell bodies at bifurcations in capillary network (Scale bar, 10 µm). (d) In vivo optical slice of pericytes along the vasculature. The image includes both the DS-red and intravascular FITC. (Scale bar, 10 µm). (e) Measurements of capillary lumen diameter showed no significant difference between the non-perfusing capillary diameters at the pericyte or 10–20 µm away from the pericyte (N = 4 mice, n = 109 capillaries, p = 0.193, Wilcoxon signed rank test). We also failed to find a difference in capillary diameters at the pericyte or 10–20 µm away from the pericyte in perfusing capillaries (N = 4 mice, n = 904 capillaries, p = 0.938, Wilcoxon signed rank test). (f) In vivo optical slice of sulforhodamine-labeled astrocytes (orange) and intravenous FITC (green). The green arrows point to astrocyte cell bodies and their corresponding endfeet surrounding capillaries. (g) Histogram showing the difference in astrocyte endfeet diameter compared in SAH and control mice is not statistically significantly different (n = 19–22 capillaries, N = 3–4 mice, p = 0.311, t-test).

Together, these data along with our data showing lack of brain edema in SAH (Figure 1e) rule out any changes in “extraluminal” structures that could cause capillary flow restriction, and instead suggests that capillary blood flow is reduced by a phenomenon that occurs within the capillary lumen itself.

Defects in capillary RBC perfusion are reversed by hyaluronidase

All endothelial cells produce 300- to 500-nm-thick glycocalyx on their luminal surface, which contributes 0.6–1.0 µm to the total cross-sectional diameter of a blood vessel.⁵ The glycocalyx makes up a significant percentage of total luminal cross-section diameter in capillaries (average diameter 4 µm), but is a minor component of cross-section diameter in pial arteries (average diameter 35–45 µm) or intraparenchymal arterioles (average diameter 15–20 µm). For this reason, we hypothesized that treatment with hyaluronidase would have a relatively specific effect on capillary luminal diameter and blood flow compared to upstream arteries or arterioles, a concept that has been confirmed in studies of cardiac capillary physiology.²¹

When administered i.v. to mice after SAH (n = 8), hyaluronidase treatment caused an expected increase in free blood plasma hyaluronan (Figure 5e) but did not alter blood pressure (MAP 77 ± 7.9 mm Hg before hyaluronidase, 82 ± 5.9 mm Hg after hyaluronidase, p > .05, n = 5), and in vivo LSM showed that hyaluronidase did not increase parenchymal arteriolar diameter (18.4 ± 1.8 µm before hyaluronidase, 17.7 ± 1.7 µm after hyaluronidase) nor did it affect pial artery diameter (average diameter 34.7 ± 1.6 µm before hyaluronidase, 35.2 ± 1.4 µm after hyaluronidase).

Figure. 5.

Treatment with hyaluronidase reverses the capillary disturbances caused by SAH. (a) Before hyaluronidase treatment, two-photon LSM shows many poorly perfusing capillaries in the cortex, with slow moving or static RBCs (white arrows) (Scale bar, 50 µm). (b) Ten minutes after hyaluronidase treatment, many of the poorly perfusing capillaries now show improved perfusion with rapid transit of RBCs. (c and d) Higher magnification views of the area outlined by the white box in (a) showing the differences in capillary perfusion before (c) and after (d) hyaluronidase treatment (Scale bar, 10 µm). (e) Left, Schematic of hyaluronidase treatment and data acquisition for capillary RBC flow data; Right, serum hyaluronan measurement confirms in vivo hyaluronidase activity. (f) Summary histograms showing (from left to right) % change in capillary diameter in control and SAH mice after hyaluronidase treatment (N = 4–5 mice unpaired t-test, p = 0.04), % non-perfusing capillary length in both control and SAH mice before and after injection of hyaluronidase (Control N = 4 mice, paired t-test, p = 0.01817; SAH N = 4 mice, paired t-test, p = 0.031), RBC velocity in both naïve-control and SAH mice before and after hyaluronidase treatment (Control N = 3 mice, 35 capillaries, paired t-test, p = 0.1767; SAH N = 4 mice, 44 capillaries, paired t-test, p < 0.0001), and pulsatility of RBC velocity in control and SAH mice before and after treatment of Hyaluronidase (control N = 3 mice, 35 capillaries, paired t-test, p = 0.2435; SAH N = 4 mice, 44 capillaries, paired t-test, p < 0.0001). Results are mean ± S.E.M.

In spite of the negligible effect on proximal arteries, hyaluronidase treatment reversed the changes in capillary RBC flow caused by SAH (Figure 5a–d). Hyaluronidase made capillaries wider in both sham-control and SAH mice, and after hyaluronidase the percentage of non-perfusing capillaries caused by SAH returned to the level found in control mice, RBC velocity returned to or even surpassed control levels of velocity, and pulsatility (Figure 5e). Interestingly, hyaluronidase also reduced the percentage of non-perfusing capillaries present even in control mice (typically 3–10% before hyaluronidase, <1% after, Figure 5f) without any significant effect on RBC velocity or pulsatility.

Brain hypoxia and neuron loss after SAH can be reversed with hyaluronidase

Next, we performed experiments to test whether changes in capillary RBC flow seen after SAH are associated with brain hypoxia. We injected sham-control and SAH mice with pimonidazole hydrochloride (Hypoxyprobe), a bioreductive agent that binds to protein sulfur groups when local tissue oxygen tension (p_O2) <10 mm Hg, which has been used as an immunohistochemical marker of tissue hypoxia.²² Comparison of control and SAH mice showed that Hypoxyprobe binding was essentially non-existent in control brain, but was widespread after SAH in ex vivo tissue specimens (Figure 6a). Immunohistochemical double-labeling with MAP-2 confirmed that many of the cells that exhibited robust binding of Hypoxyprobe after SAH were neurons (Figure 6b); immunohistochemistry for GFAP also showed occasional double-labeled cell bodies indicating hypoxic astrocytes, but these were not as numerous.

Figure. 6.

SAH causes brain hypoxia and neuron death that can be reversed with hyaluronidase treatment. (a) Immunohistochemistry for hypoxyprobe binding in representative sections in sham-control and SAH mice (left two panels) (Scale bar, 10 µm). (b) MAP-2 immunohistochemistry (right two panels) is used to identify neurons in the cortex. (c) Summary histogram showing tissue pO₂ in sham-control and SAH mice recorded using an oxygen sensing microelectrode (N = 5 sham-control, 11 SAH mice, t-test, p < 0.001). (d) Schematic showing timing of hyaluronidase injection and data acquisition for neuron morphology and neuron number. (e) Confocal photomicrographs of apical dendrites from sham-control and SAH transgenic mice expressing YFP under the Thy-1 promoter (labeling excitatory neurons) 48 h after SAH. At this time, the overall number of neurons does not appear significantly reduced (insets) (Scale bar, 10 µm). (f) Stereologic counting of cresyl violet-stained sections from sham-control and SAH mice shows significant reduction in neuron number; hyaluronidase treatment reverses neuron loss at 7 days after SAH (N = 7, 1-way ANOVA, p = 0.019, Tukey test p < 0.001).

Additionally, we measured in vivo cerebral p_O2 using microelectrodes. Comparison of control and SAH mice showed that control mice had brain parenchymal p_O2 of 20–35 mmHg, while SAH mice had significantly reduced p_O2 (Figure 6c). When we measured p_O2 in vivo in SAH mice before and after hyaluronidase treatment, oxygen levels improved significantly back to normoxic levels (>20 mmHg), confirming that hyaluronidase-induced changes in capillary RBC flow could improve blood flow and oxygen delivery to the brain after SAH.

After observing that hyaluronidase reverses the changes in capillary RBC flow and hypoxia caused by SAH, our next question was whether SAH causes neuron loss and can this also be reversed by hyaluronidase delivered at 12 and 24 h post-injection of autologous blood (Figure 6d). Although there are data documenting infarcts in animal models and patients after SAH,^23,24 there are few studies examining diffuse, widespread loss of neurons after SAH as would be expected with widespread reduction in capillary RBC flow and these studies do not adequately quantify the extent of neuronal loss because they depend on secondary indicators of neuronal death (e.g. immunohistochemistry of apoptotic markers which provide a “snap shot” of dying cells) or employ non-stereologic techniques.

To examine the acute effects of SAH on neuronal morphology we used ex vivo confocal imaging of brain slices of dendrites and cell bodies in transgenic mice expressing yellow fluorescent protein (YFP) under the Thy1 promoter 48 h after SAH, neurons demonstrated blebbing and irregularities along their dendrites, with reduced density of dendrites in layers II–III compared to that of control mice (Figure 6e). In order to obtain a comprehensive counting of neuron number, we used standard stereologic techniques combined with Nissl stain (cresyl violet) to obtain an unbiased sample of neuron number in both cortex and hippocampus at 7 days after SAH. Stereologic counting of neurons 7 days after SAH showed that neuron population is reduced by one-third after SAH (Figure 6f). The degree of neuron loss after SAH corresponds to the severe hypoxia observed acutely after SAH, and shows that the extent of neuronal injury in our model correlates well with neuron loss documented in human autopsy studies after SAH.²⁵

To test whether changes in capillary RBC flow contributed to neuronal death, and to show that rescue of neurons could be accomplished by reversing capillary flow defects seen after SAH, we performed hyaluronidase treatment and examined its effects on neuron number in SAH mice. Mice received intravenous injections of hyaluronidase 12 h and 24 h after SAH (N = 7). Supporting our hypothesis, treatment with hyaluronidase prevented SAH-induced neuron loss at 7 days (Figure 6d); the number of neurons counted stereologically in mice treated with hyaluronidase after SAH was indistinguishable from that of control mice. This shows that a treatment selectively targeting capillary blood flow changes after SAH can rescue neurons that would ordinarily have died.

We next attempted to confirm our earlier results at the ultra-structural level using electron microscopy. SAH and sham-control mice were prepared 48 h prior to perfusion fixation. Mice were randomly assigned to one of three groups, Sham-control, SAH + vehicle, or SAH + hyaluronidase. Vehicle and hyaluronidase were administered i.v. at 12 and 24 h post-injection of autologous blood and all mice were perfusion-fixed 24 h later (48 h total). We measured capillary luminal cross-section and astrocyte endfeet area as well as number of endothelial protrusion. Consistent with our earlier experiments cross-sectional area of capillary lumen was significantly decreased in the SAH group that received vehicle as well as the SAH group that received hyaluronidase but was more pronounced in the former (Figure 7a–d). We did not find any evidence of astrocytic endfeet swelling, as each group was statistically indistinguishable (Figure 7e). Surprisingly we did find an increased number of endothelial luminal protrusions in the SAH + vehicle cohort compared to the SAH + hyaluronidase as well as sham control mice (Figure 7f).

Figure. 7.

SAH causes narrowing of capillary lumen and increases small luminal protrusions that can be reversed by hyaluornidase treatment. (a–c) Electron micrographs of representative capillaries from a sham-control, SAH + vehicle and SAH + hyaluronidase mice (scale bar, 10µm). (d) Summary histogram showing luminal cross-sectional area (µm²) (n = 25 capillaries, N = 3–4 mice per group, 1-way ANOVA, p = 0.017, Tukey test p < 0.05). (e) Summary histogram showing astrocyte endfeet cross-sectional area (µm²) (n = 25 capillaries, N = 3–4 mice per group, 1-way ANOVA, p = 0.322). (f) Summary histogram showing number of small luminal protrusions (n = 25 capillaries, N = 3–4 mice per group, 1-way ANOVA, p = 0.01, Tukey test p < 0.05).

Discussion

The aim of the present study was to demonstrate that SAH causes a profound decrease in capillary blood flow leading to brain hypoxia and neuron loss, and these sequelae of SAH can be prevented by specifically targeting the endothelial glycocalyx, a physical resistor to RBC transit. These changes are acute and unrelated to arterial vasospasm, which is the “culprit” typically implicated in poor patient outcome after SAH. Capillaries without RBC flow continue to have flow of blood plasma, thus flow is not occluded but restricted to the point that RBCs cannot pass through. This flow restriction is not caused by brain edema, astrocyte endfeet swelling, endothelial swelling, or pericyte constriction, suggesting by process of elimination that a luminal component restricts RBC movement. Therapies targeting capillary RBC flow represent a much needed new avenue for improving patient outcome after SAH, and our work shows that measured hyaluronidase treatment after SAH is able to restore capillary RBC flow, reverse brain hypoxia, and rescue neurons that die after SAH.

We used hyaluronidase to specifically manipulate capillary luminal diameter and RBC flow after SAH. This is a well-documented means of manipulating the glycocalyx in other studies of capillary blood flow in brain²⁶ and other organ systems,²¹ and we have shown that selectively improving capillary perfusion can improve brain oxygenation after SAH and prevent SAH-induced neuronal death without altering systemic blood pressure or local upstream artery/arteriole diameter. Arterial vasomotor tone is an important regulator of blood flow and delayed arterial spasm could worsen the capillary RBC flow deficits we have described, or cause different problems with cerebral perfusion (e.g. focal infarction as opposed to global ischemia). These data show that changes in capillary blood flow are also an important component of cerebral blood flow derangement after SAH, and this derangement can be selectively treated. Most studies of DIND after SAH have focused on intracranial large vessel vasospasm because it is logical to assume that arterial vasospasm is the culprit of DIND and poor patient outcome. Unfortunately, therapies that reduce vasospasm have failed to realize significant clinical benefit.²⁷ Therapy targeted at improving blood flow through capillaries, combined with anti-vasospasm treatments, may prove more effective at improving patient outcome.

The microcirculatory changes we have demonstrated are distinct from angiographic vasospasm, the most commonly implicated cause for poor outcome in patients after aneurysmal SAH prior to the CONSCIOUS-1 trial.²⁷ As noted by others, the timing of vasospasm in mice (24–48 h after SAH) is different from vasospasm in humans (4–14 days).¹⁷ Unlike SAH, which dramatically reduces both RBC velocity and pulsatility, ET-1-induced vasospasm of similar severity had no effect on RBC velocity or pulsatility. The higher dose of ET-1-induced spasm eventually led to reduced RBC velocity, but still had no effect on pulsatility. Together, these data show that SAH has effects on capillary RBC perfusion that are independent of effects caused by proximal artery vasospasm, and this argues for specific SAH-induced changes in capillaries. Additional in vivo imaging studies may clarify the interplay between upstream arterial and capillary flow disturbances in SAH as well as a variety of other cerebrovascular disease states where the impact of arterial narrowing on capillary blood flow (and the delivery of oxygen) remains incompletely understood.

Many studies suggest that aneurysm rupture is associated with a profound decrease in oxygen consumption for reasons that are not fully understood.²⁸ Conceivably, the reduction in RBC velocity seen after SAH in our model could be due to decreased metabolic demand and compensatory decrease in CBF due to autoregulation, but our hyaluronidase studies suggest otherwise. If RBC flow was decreased because metabolic demand was decreased, then “force feeding” the brain by increasing blood flow with hyaluronidase treatment should not have resulted in improved neuronal survival. Instead, our brain oxygenation and neuronal survival data before and after hyaluronidase treatment strongly suggests that improving capillary RBC flow improved oxygen delivery, and this prevented neuronal loss after SAH.

By combining in vivo 2-PLSM and stereologic techniques, we are able to measure the length of perfusing and non-perfusing capillaries, unlike other reports where conclusions were drawn from ex vivo samples.²⁹ There are no direct means for making similar in vivo observations in patients with SAH, but established as well as evolving technologies allow for assessment of microcirculatory function in patients with sepsis.³⁰ Positron emission tomography (PET) and computed tomographic perfusion (CTP) provide indirect quantification of cerebral blood flow,³¹ and new invasive brain monitors that can measure brain tissue oxygen levels³² may allow for extension of our work in mice models to human clinical trials. Both PET and CTP demonstrate CBF is reduced acutely after SAH, with CTP showing increased mean transit time (MTT) that correlates with our finding of acute reduction in RBC velocity after SAH.³³ PET and CTP also demonstrate an increase in cerebral blood volume (CBV) after SAH, which may seem at odds with our finding that capillary lumen is effectively decreased after SAH, but in vivo imaging does not allow for a global assessment of CBV, and if other parts of the circulatory system dilate in response to hypoxia, then our findings are not necessarily contradicted by PET and CTP.

The physiological deformation of erythrocytes as they travel through capillaries has been studied since the 1960s^3,4 and is a phenomenon conserved across many mammalian species, including humans.³⁵ RBC shape is thought to influence oxygen transport by erythrocytes in capillaries and alterations in RBC deformability can contribute to the pathology of a variety of human diseases. The relative conservation of erythrocyte size (6.8 µm in mice, 7.5 µm in rats, 8.0 µm in humans) and flow speeds through capillaries between different species suggest that the physical properties of RBC movement through capillaries is important to the ability of RBCs to carry and release oxygen to surrounding tissue.³⁵ Additionally, we found that SAH tended to increase the number of small luminal protrusions observed within capillaries, confirming what has previously been reported.³⁶ While the significance of this finding has yet to be elucidated, it would not be surprising if these small luminal protrusions have a direct effect on RBC transit.

Hyaluronidase degradation of the glycocalyx may dislodge microthrombi or adherent inflammatory cells that may narrow or occlude capillaries after SAH.³⁷ Whether capillary narrowing and RBC obstruction after SAH is caused by changes in the glycocalyx, microthrombi, inflammatory cell adhesion, or a combination of these intraluminal factors as has previously been reported remains to be elucidated, but what is clear from our data is that therapies focused on any or all of these factors could have significant impact on patient outcome after SAH. Therapy specifically targeted at any one of these possibilities may be more effective and have fewer side effects than hyaluronidase treatment which we used as a “proof of principle” that specific manipulation of capillary blood flow could alter the course of brain injury after SAH. Future experiments will include behavioral testing, but a new animal model may be necessary as previous studies involving mouse models of SAH have not produced conclusive results.³⁸ Further work with hyaluronidase in large animal models of SAH, or other agents targeting capillary blood flow, may lead to the development of therapies that can be translated into human treatment. Additionally, the work we presented here is also consistent with the SAH Neurovascular Inversion hypothesis put forward by the Wellman Lab in that the peak severity of neurovascular inversion (∼48 h) fits with the time course of microvascular disruption we observed as well as the lack of structural changes (endothelial and astrocytic swelling, perecytic constriction).³⁹

Our work provides further evidence that SAH causes specific disturbances of capillary blood flow that affect oxygen delivery and neuronal survival. Future work can focus on better defining the mechanical, cellular, and molecular events that conspire to cause capillary narrowing after SAH and the effects that decreased capillary blood flow has on brain function. Conceivably, similar changes in capillary function may also be confirmed in other forms of brain injury. With a better understanding of the elements that change in capillaries after brain injury, therapies can be developed to restore the microcirculation and preserve neurologic function.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:

This work was supported in part by NS057522 (GEV), NS 30007 and NS37073 (MN) from the US National Institutes of Health and the National Institute of Neurologic Disease and Stroke, by the Neurosurgery Research and Education Foundation of the American Association of Neurological Surgeons, by the Anspach Effort, and by the Brain Aneurysm Foundation.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

Evan D McConnell contributed to the design of the study, acquired, analyzed and interpreted data as well as drafting, revising the manuscript, and final approval of the version to be published. Helen S Wei contributed to the design of the study, analysis and interpretation of data in addition to drafting, revising the manuscript, and final approval of the version to be published. Katherine M Reitz contributed to the study by acquiring and analyzing data as well as drafting the manuscript and final approval of the version to be published. Hongyi Kang contributed to the acquisition and analysis of data, drafting the manuscript, and final approval of the version to be published. Takahiro Takano contributed to study design and critically important intellectual content and final approval of the version to be published. G Edward Vates contributed to the design of the study, analysis and interpretation of data, drafting, revising manuscript, and final approval of the version to be published. Maiken Nedergaard contributed to the study design, drafting, revising manuscript, and final approval of the version to be published.