Mapping the dynamics of brain perfusion using functional ultrasound in a rat model of transient middle cerebral artery occlusion

Abstract

Following middle cerebral artery occlusion, tissue outcome ranges from normal to infarcted depending on depth and duration of hypoperfusion as well as occurrence and efficiency of reperfusion. However, the precise time course of these changes in relation to tissue and behavioral outcome remains unsettled. To address these issues, a three-dimensional wide field-of-view and real-time quantitative functional imaging technique able to map perfusion in the rodent brain would be desirable. Here, we applied functional ultrasound imaging, a novel approach to map relative cerebral blood volume without contrast agent, in a rat model of brief proximal transient middle cerebral artery occlusion to assess perfusion in penetrating arterioles and venules acutely and over six days thanks to a thinned-skull preparation. Functional ultrasound imaging efficiently mapped the acute changes in relative cerebral blood volume during occlusion and following reperfusion with high spatial resolution (100 µm), notably documenting marked focal decreases during occlusion, and was able to chart the fine dynamics of tissue reperfusion (rate: one frame/5 s) in the individual rat. No behavioral and only mild post-mortem immunofluorescence changes were observed. Our study suggests functional ultrasound is a particularly well-adapted imaging technique to study cerebral perfusion in acute experimental stroke longitudinally from the hyper-acute up to the chronic stage in the same subject.

Introduction

The consequences of acute focal cerebral ischemia on brain tissue range from none to massive, i.e. complete infarction with loss of all tissue elements. It is well established from animal models using middle cerebral artery occlusion (MCAo) in various species that this wide range of consequences depend on three key factors, namely the depth of hypoperfusion achieved following MCAo, the timing of recanalization and the efficiency of tissue reperfusion.^1–4 These concepts have been translated to the clinical arena, and imaging-based clinical studies have documented that similar factors also determine tissue fate and hence neurological outcome after stroke. ⁵ Nevertheless, numerous issues remain incompletely resolved regarding the precise time course of hypoperfusion and reperfusion in relation to tissue outcome, for instance how stable is perfusion during MCAo until recanalization occurs, how deleterious is rebound hyperperfusion, is reperfusion slope important for tissue outcome? To address such issues, a three-dimensional quantitative in vivo imaging technique able to map perfusion in the rodent brain at high spatiotemporal resolution would be desirable. Currently available methods such as laser Doppler, speckle laser or near infrared spectroscopy ⁶ do not have all of these features, allowing only surface imaging, limited field-of-view (FOV) and/or limited spatial resolution. More cumbersome techniques such as PET, ⁷ SPECT, ⁸ CT perfusion ⁹ and perfusion MRI using dynamic susceptibility contrast or ASL ¹⁰ have inadequate temporal and/or spatial resolution. Finally, MR-based cerebral blood volume (CBV) mapping ¹¹ has good spatial resolution but limited temporal resolution, and raises practical issues in the laboratory environment, particularly in the context of MCAo.

In this respect, a novel and promising approach called functional ultrasound (fUS) imaging has been developed to study brain hemodynamic changes associated with various somatosensory stimulation in head-fixed rats. ¹² This method has been recently extended to brain imaging in freely moving rats during active tasks. ¹³ fUS allows assessment of the changes in CBV at high spatiotemporal resolution (100 μm and 400 ms, respectively) without contrast agent and in both physiological ¹² and pathological conditions. ¹⁴ As will be seen here, and relative to other imaging techniques, fUS allows a precise mapping of ischemia and reperfusion following MCAo.

The present pilot study was therefore designed to evaluate the ability of fUS to longitudinally assess tissue perfusion over several days in a rat tMCAo model, as well as the fine dynamics of early reperfusion following release of occlusion. To ensure reperfusion of viable tissue over several days, we avoided long MCAo durations where reperfusion can be modest or even absent as a result of the no-reflow phenomenon. ¹⁵ We therefore opted for an MCAo duration and rat strain expected to result in limited infarction or even pure neuronal loss, namely 45 min in Sprague-Dawley rats. ¹⁶ In addition, our focus was on the dorsal somatosensory cortex and motor cortex because (i) being located at the center and border of the MCA territory, respectively, they allow the assessment of different depths of ischemia and (ii) they are readily assessable with fUS using the chronic thinned-skull surgery procedure developed in our laboratory. ¹²

Materials and methods

Animals

This investigation was performed in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals. The protocol was approved by the Local Animal Ethics Committee of Paris 5 (CEEA 34) and conducted in accordance with Directive 2010/63/EU of the European Parliament. Adult male Sprague-Dawley rats (n = 18; Janvier Labs, France) weighting 250 g were used for this study and kept in a 12-h reverse dark/light cycle environment at a temperature of 22℃ with unlimited water and controlled access to food (around 20 g per day per animal), with a follow-up of weight (twice a week) allowing to maintain their initial weight during all the experiments. Rats were housed four per cage. Subjects were randomized to tMCAo or sham according to a 2:1 ratio, for a total of 12 MCAo and 6 sham rats. All subjects will be reported below. This manuscript was written up according to the ARRIVE guidelines for reporting animal experiments.

Anesthesia

Briefly, initial surgery was performed under 2.5% isoflurane in 100% O₂, 0.5 L/min (Vetflurane, Virbac, France), delivered continuously through a nose mask. Isoflurane was systematically reduced at 1.25% for all imaging sessions. The depth of anesthesia was assessed regularly by pinching the interdigital region of the foot using forceps, as well as controlling the correct respiration rate and pattern, eye blinking and whiskers movements. Body temperature was monitored by a rectal probe and maintained between 36 and 37℃ throughout the experiment using a digitally controlled heating blanket (ATC1000, WPI, USA). Regarding recovery, rats were placed in a warm cage and monitored periodically until they wake up. For analgesia, intra-peritoneal injections of buprenorphine (0.03 μg g, Buprecare) were performed immediately after surgery and again 12 h later.

Thinned-skull surgery

To reduce fUS signal attenuation from bone and to perform fUS in the chronic post-MCAo stage, we used a previously described thinned-skull surgery protocol. ¹² As compared to craniotomy, this procedure prevents problems such as damage to parenchyma and inflammation, which may affect cerebral perfusion. ¹⁷ Briefly, the scalp was shaved and cleaned with betadine. Prior to skin incision, lidocaine was injected subcutaneously (1 ml/kg). Then, the scalp was incised over the sagittal suture and the periosteum removed with a scalpel blade. The skin was trimmed laterally and the temporal muscle gently detached from the bone on both sides of the skull. The frontoparietal bone was thinned on a 0.7 cm² area (from Bregma + 3.50 mm to −4.00 mm, and laterally by ∼5.00 mm from the sagittal suture) over the somatosensory cortex at low speed with a dental drill using a 1.4 mm burr. To avoid overheating the brain, saline was added repeatedly between drilling sessions until the skull began to be flexible, and the pial vessels are visible through wet bone. This thickness of around 50 µm is required for ultrasound imaging of the brain in chronic conditions. ¹² An echogenic 250 µm glass bead was fixed at Bregma 0.00 with cyanoacrylate glue to provide a landmark to calibrate the position of the ultrasound probe. The complete procedure lasted ∼1 h and was immediately followed by fUS imaging to assess the quality of the thinned-skull procedure. After each imaging session, the bone was protected with a 5 mm thick well of dental cement (Dentalon plus, Heraeus-Kuzler, Germany) cast around the thinned window and filled with low-melting agarose (1%) containing ampicillin at a concentration of 100 µg/ml. The agarose ensures proper acoustic coupling required for fUS imaging, and ampicillin was used to inhibit bacterial growth.

tMCAo surgery

Forty-five minutes transient focal cerebral ischemia was achieved by AM, using the intraluminal MCAo model as previously reported.^8,18 Assignment to the sham or MCAo group was by draw just prior to surgery. Briefly, in MCAo rats a 4-0 nylon suture microfilament of 2.6 cm length was introduced into the right external carotid artery up to the level where the MCA branches out. The ipsilateral common carotid artery (CCA) was also ligated to reduce collateral flow. After 45 min, the filament and then the CCA clip were removed to allow reperfusion. Sham rats received a cervical incision with dissection of the right internal carotid artery without intraluminal introduction. For both groups the skin was then sutured. After MCAo, subjects were caged individually and carefully monitored. Rats were monitored to check they regained consciousness and showed no signs of discomfort, severe motor deficit or seizures.

fUS imaging

General characteristics

According to the characteristics of the ultrasound probe, the fUS image had an FOV of 12.8 mm width × 8 mm depth. The probe was oriented in the coronal plane above the brain and aligned to the region of interest using a piezoelectric linear stage (M403.4DG, Physik Instrumente, Germany) to allow motion of the animal from anterior to posterior. The acoustic coupling was ensured by isotonic gel (Uni'Gel US, Aspet'Inmed, USA) applied for each imaging session between the agarose and the probe. The ultrasound scanner is composed of an electronic module for ultrasound emission and reception (V1, Verasonics, USA), and a linear array transducer with a central frequency of 15 MHz (L15-128, 128 elements, 0.10 mm pitch, Vermon, France). A bi-CPU computer workstation on which custom MATLAB R2012 scripts (MathWorksInc, USA) is used to generate the pulse wave and to reconstruct all images (including beamforming and filtering steps). Acquisition parameters are fully described in Urban et al. ¹²

fUS protocol

fUS imaging was performed under 1.25% isoflurane at day 0 before, during and after tMCAo, and repeated three and six days after surgery (Figure 1). At each time-point, fUS images were acquired at six coronal planes (Bregma + 2.50, + 1.00, 0.00, −1.00, −2.00, −3.00 mm) and separately for each hemisphere; in addition, a 60 min fUS movie was acquired directly after thread removal over one coronal plane of the affected hemisphere (see below for details).

Figure 1.

Schematic representation of the experimental timeline (top panel). After an initial behavioral training of two weeks, thinned-skull surgery was performed from Bregma (ß) + 2.50 to −3.00 mm. The motorized ultrasonic probe was used over the thinned skull window. fUS imaging was acquired before, during and early after reperfusion, and at days 3 and 6 following 45 min tMCAo (blue box). From days 2 to 20, Garcia's Neuroscore, modified sticky label test and beam-walking test were carried out to evaluate behavioral changes (red box). Finally, after sacrifice (at day 21), immunofluorescence assays were used to assess ischemic neuronal loss, astrocytosis and microglial activation using NeuN, GFAP and Iba1, respectively (green box).

For fUS data acquisition, the animal was head fixed in the stereotaxic frame, the ultrasound probe was positioned and fUS data were acquired on an anti-vibration table to minimize external sources of vibration. No contrast agent was used. The fUS images acquired during occlusion were started ∼8 min after MCAo. After 45 min, the animal was removed from the fUS apparatus to carefully withdraw the filament. The animal was then repositioned under the fUS apparatus and fUS data acquisition restarted (one fUS image, then movie acquisition). The duration of the thread removal and repositioning procedure was 6 to 10 min. Following each scanning session, the animal was placed in a warm cage and monitored periodically until it had recovered.

fUS image acquisition

Each session consisted in automatic acquisition at six coronal planes on each hemisphere, each plane data consisting of three consecutive fUS acquisitions, subsequently averaged to minimize noise. This was repeated for the opposite hemisphere, for a total acquisition time ∼3 min. This process therefore generated 12 images, 6 on each side. In each image, pixel size was 100 × 100 µm, with a slice thickness of 500 µm and FOV of 12.8 mm wide × 9 mm deep, covering approximately three-quarter of the adult rat brain in the coronal direction. In order to obtain coronal images covering the whole brain, the corresponding images on each side were subsequently stitched together using a Matlab script, resulting in six coronal images in total (see Figure 2(a) for illustration). One such set of six coronal images was obtained for each scanning session, namely control, occlusion, early reperfusion, and three days and six days after MCAo. Further details on fUS data acquisition can be found in Urban et al. ¹²

Figure 2.

(a) Representative coronal fUS images at Bregma −2.00 mm before, during and after MCA occlusion, highlighting severe focal ischemia affecting the primary somatosensory (S1) and less severely so the motor cortices (M1-M2) followed by tissue reperfusion. The dotted line represents the depth of fUS signal quantitative analysis, which was restricted to the dorsal cortex due to attenuation from skull regrowth after day 3. This resulted in excluding the hippocampus (HPC), corpus callosum (CC) and caudate-putamen (CP). (b) fUS imaging across six coronal sections, regularly spaced between Bregmas +2.50 and −3.00 mm and covering most of the MCA territory, shown here for the imaging five time-points (top to bottom). The cartoon shown above the top sections illustrates the motor (blue) and somatosensory ROIs (red) used for analysis. For illustration, the red contours over the right hemisphere depict the brain area showing markedly reduced fUS signal during MCAo (see Methods). Following thread withdrawal, clear-cut reperfusion of the previously ischemic area is apparent already in the immediate post-occlusion scan and maintained at both later time-points. The high signal at the top of some images at days 3 and 6 represents skull thickening and granulation tissue causing artefactual signal. The grey scale on the right represents the normalized signal intensity. Scale bar = 2 mm.

The fUS signal can be separated in a positive and a negative Doppler signal, proportional to the number of red blood cells (RBCs) flowing towards and away from the transducer, respectively. This allows the detection of individual arterioles and venules lying in the axial direction of the ultrasound beam, i.e. oriented perpendicularly to the pial surface (see Figure 2(a) for illustration). Importantly, the noise resulting from tissue motion (i.e. pulse pressure and respirations) that generates low Doppler frequencies was cancelled by applying a 20 Hz cutoff high-pass filter. A consequence of this filtering is that RBC motion speed lower than 2 mm/s was cancelled as well, meaning that slow blood flow from capillaries and other small vessels does not contribute to the finally acquired fUS signal.

fUS movie acquisition

As mentioned above, the early post-reperfusion fUS movie was acquired at a single coronal plane (Bregma −2.00 mm) and over the affected hemisphere only, in order to image the center of the occluded MCA cortical territory. It was started 10–15 min post-thread withdrawal and was acquired at 0.2 Hz (i.e. one image every 5 s) for 60 min.

fUS data analysis

Because bone regrowth caused a marked reduction of fUS signal in structures distant from the surface, ¹² the analysis of the longitudinal data through the chronic stage was restricted to dorsal cortical areas only. This was based on the fUS signal, as confirmed by the low level of attenuation not exceeding 15 dB throughout the six-day acquisition period (data not shown).

All fUS images were first adjusted to the size of the corresponding masks extracted from the anatomical atlas of the rat brain ¹⁹ with the help of a custom GUI interface on Matlab allowing pixel per pixel adjustment. In order to ensure automatic data extraction according to a fixed region of interest (ROI) template, we adapted the method previously described by Hughes et al. ²⁰ Accordingly, the fUS data were analyzed within a set of 20 ROIs per hemisphere spanning the motor cortex and the dorsal portion of the somatosensory cortex across the six coronal sections, as follows: primary somatosensory cortex (S1, including paw, forelimb, hindlimb, trunk region, shoulder) and primary and secondary motor cortex (M1 and M2, respectively). This template was then superimposed on the co-registered fUS maps (Figure 2(b)).

A dedicated MATLAB script then allowed automatic processing for the quantification of the hemodynamic signal in each ROI. First, in order to mitigate the influence of aberrant pixels with extremely high signal, we used the median value across all voxels within the ROI for further analyses. Second, we excluded all ROIs from the contralateral (non-occluded) hemisphere with a median value <0.2, as these low signal values are related to attenuation of the ultrasound signal and do not allow proper rCBV quantification. In this event, the pair of ROIs was excluded from further analysis. Third, we computed for each ROI the ratio of the median fUS signal of the affected-side ROI to that of the homologous unaffected-side ROI, generating relative CBV (rCBV) values. In order to reduce the effects of global variations in signal across the brain from subject to subject, we then normalized this ratio to the same ROI's pre-occlusion (i.e. baseline) ratio. Finally, the weighted median values (i.e. weighted by the size of each ROI) for the motor and somatosensory areas (n = 12 and 8 original ROIs, respectively) were computed for each rat.

An automated segmentation of the ischemic area during occlusion was performed using a dedicated MATLAB script based on a pixel-by-pixel analysis of the intensity of the CBV compared with the average CBV signal on the entire image. All pixels having a value below the average CBV intensity signal + 3 SD were grouped together and a Gaussian filter of 5 × 5 pixels was applied to define the ischemic area.

Regarding the fUS movie, the same set of template ROIs was applied onto the fUS data following size adjustment as described above. In order to illustrate the time course and slope of reperfusion at high temporal resolution, the fUS data for the somatosensory and motor areas were then normalized to the first image (to be referred to as %CBV in what follows).

Statistical analysis

All statistical analyses were performed using GraphPad Prism (Version 6.0, GraphPad Software Inc.). Statistical significance was set at two-tailed p < 0.05. Time-series were subjected to one-way repeated measures (RM)-ANOVAs followed by Dunnett's post hoc test if a significant time effect was present.

Behavioral assessment

The behavioral effects of the tMCAO were assessed by carrying out three tests repeatedly over a period of 21 days after tMCAo. The Garcia Neuroscore (NS), the modified Sticky Label Test (mSLT) and the Beam Walking Test (BWT) were performed at days 1, 2, 5, 8, 14 and 20 (Figure 1). All behavioral tests were performed in a dark environment. To reduce investigator bias, all behavioral experiments were carried out by the same operator over the whole period of five weeks, including two weeks of habituation and three weeks of behavioral tests. The investigator was blinded to group identity during behavioral testing.

The NS is an evaluation of the neurological deficit scored from 3 to 18 (the lower the score, the worse the deficit). Evaluation of the deficit due to tMCAo is made from motor, sensory, reflex and observational tests as described previously. ²¹ The mSLT allows assessment of subtle sensorimotor dysfunction induced by the occlusion. A small patch of paper tape wrapped around the contralateral wrist to the ischemic hemisphere was used to assess time attended. ²² Each behavioural test is composed of five experimental sessions, each trial lasting up to 30 s. After each trial, the tape is removed and rats receive a resting time lasting at least 3 min. The mSLT performance is calculated by dividing the time attending to the stimulus by 30 s, which expresses the fraction of the observation period that the animal spends attending to the tape. The best results obtained each day are then averaged to obtain the final score.

For the BWT, rats were trained to perform a task in a 1.5 cm wide, 120 cm long rectangular wooden bar placed 50 cm above the floor between the start point and the home area. The time to reach the home area was measured and the best trial out of three was selected. ²³

Within each group (sham or tMCAo), the mSLT and BWT data were assessed for a main effect of time (within-subject factor) using one-way repeated-measures ANOVA. A group effect was sought only if the within-group ANOVA showed a significant time effect.

Immunohistological staining

Experimental procedure

At day 21, rats were deeply anesthetized with sodium pentobarbital (100 mg/kg i.p.) and transcardially perfused with 50 ml saline followed by 150 ml of 4% paraformaldehyde in 0.1 M phosphate buffer saline (PBS), pH 7.4, using a peristaltic pump and flow rate of 10 to 25 mL/min. Brain was removed and placed overnight in 4% paraformaldehyde fixative in PBS, then 40 µm thick sections across the MCA territory were prepared using a vibratome (Leica VT1000S, Leica Microsystems, Germany). Six coronal slices located at Bregma +2.50, + 1.00, −000, −1.00 −2.00 and −3.00 mm respectively, were immunolabeled overnight with antibodies against the neuronal marker NeuN (1:1000; Millipore, USA), the astroglial marker GFAP (glial fibrillary acidic protein) (1:500; Invitrogen, USA) and the microglial cell marker ionized calcium-binding adaptor molecule 1 (Iba1) (1:500; Wako, Japan). Then slices were washed 3 × 10 min at room temperature followed by incubation with species-appropriate secondary antibodies conjugated to Alexa Fluor 488 nm (Molecular Probes, Life Technologies, France, 1:1000) in PBS, washed again (3 × 10 min) in PBS and mounted with a DAPI Prolong Antifade kit (Molecular Probes, Life Technologies, France). Standardized images acquisition was performed with a digital slide scanner (Hamamatsu NanoZoomer, USA) at 20 × magnification. Visual analysis classified the ischemic lesions present 21 days post-MCAo as either pan-necrosis, defined by complete loss of neurons, microglia and astrocytes and destroyed extracellular matrix leading to cavitation and marked volume loss, or selective neuronal loss (SNL), characterized by incomplete neuronal loss, presence of activated microglia and astrocytes within the abnormal area, and preserved extracellular matrix without any cavitations or marked tissue loss. ²⁴

Lesion segmentation

The digital images were post-processed using ImageJ (National Institutes of Health, Bethesda, USA). Segmentation of the region showing lack of or reduced NeuN staining (neuronal loss) or increase of GFAP or Iba1 staining (inflammation) was performed manually by two senior investigators (JCB and AU) presented slice by slice the data from all three immunostains simultaneously and blinded to rat group. Affected regions were independently identified and contoured for the sham and tMCAo rats. Any disagreement between the two readers was resolved post-hoc by consensus. For NeuN, the cross-sectional surface area encompassed within the contoured ROIs was then measured within each section using ImageJ. Whole lesion volume across adjacent sections was then estimated using the planimetric Cavalieri method. ²⁵ For each rat, the total volume across all sections, to be referred to as Vol_total below, was then calculated according to three groupings: cortical, striatal and both together.

Statistical analysis

Cohen's kappa on NeuN ROIs was first carried out to determine inter-rater reliability of manual lesion contouring. Agreement between the two raters was only moderate for the entire dataset (kappa = 0.379), but was good for the striatum area (kappa = 0.533) and poor for cortical areas (kappa = 0.209). Differences in occurrence of damage between the two groups were assessed using Fisher's exact test. Differences in Vol_total between tMCAo and sham rats were assessed using Mann-Whitney test.

Results

Animals

Seven rats (five MCAo and two sham) were excluded from the analysis. Four died during the procedures: three from surgery-related massive local hemorrhage (one during thinned-skull surgery and three during tMCAo) and one found dead in its cage without warning signs nine days after MCAo (autopsy not carried out). The remaining three rats were excluded either a priori because of failure to induce sufficient hypoperfusion during MCAo based on fUS data (two rats) or because of occurrence of symptomatic inflammation, i.e. granulation tissue, over the thinned skull (one sham rat). Thus, seven tMCAo rats and four sham rats were available for the final analysis.

2D mapping of cerebral perfusion using fUS imaging

Pre-occlusion data

Thanks to the high spatial resolution of fUS, individual cortical vessels were readily identifiable in the control fUS images before occlusion (Figure 2(a), leftward image). The cortical vasculature was homogeneously distributed over the imaging plane and between planes. Similar images were obtained in sham rats throughout the protocol.

fUS during and following MCAo

The time course of fUS images before during and after MCAo in a typical rat is illustrated for one selected whole coronal section in Figure 2(a), and for the dorsal cortical areas at six different levels and all five time points in Figure 2(b). Note the striking decrease in fUS signal in the ischemic MCA territory during occlusion, which was maximal for the somatosensory area at Bregma +1.00 mm (mean% value relative to mirror region: 9.1%, range: 3.8–16.6%).

The mean rCBV data for the somatosensory and motor ROIs are shown in Figure 3 for the MCAo (left) and sham rats (right). In the somatosensory region, the rCBV declined to 19.7 ± 16% of baseline during MCAo, returned progressively to control values from early reperfusion (63 ± 17.9%) to slightly above control values (121.2 ± 65.5% and 115.7 ± 51.3%) at days 3 and 6. There was a significant effect of time with respect to baseline (p = 0.0015, RM-ANOVA) with on post-hoc tests a highly significant decrease vs. baseline during occlusion (p <  0.00001, Dunnett's test), still significant immediately after release of occlusion (p < 0.001), and not significant anymore at subsequent time points. In the motor region, the rCBV was reduced to 71.4 ± 31.1% of baseline during occlusion. It was still reduced immediately after release of the occlusion (78.1 ± 8.6%), and subsequently was slightly higher than baseline (119.5 ± 24.6% and 142 ± 9.7% at days 3 and 6, respectively). However, there was no significant effect of Time using RM-ANOVA. There were no significant rCBV changes over time in the sham group.

Figure 3.

Left: Time course of the mean (and 1 SD) affected-to-unaffected CBV ratio (normalized to the baseline value) before (Ctl), during (Occ) and at reperfusion (Rpf) and at post-MCAo days 3 and 6, for the somatosensory (top) and motor (bottom) ROIs in the MCAo group, (n = 7 rats). These data illustrates the reduction in CBV in these cortical areas during occlusion, followed by partial reperfusion immediately after thread withdrawal, becoming complete and stable thereafter. These effects were more conspicuous for the somatosensory cortex. The repeated-measures ANOVA showed a significant time effect for the somatosensory cortex (p = 0.0015). Stars show statistically significant differences of post-MCAo values relative to baseline (**P < 0.01, ****P < 0.0001). Note that although the ANOVA for the motor region was not significant, there was a trend for significant rCBV reduction at reperfusion time point. Right: corresponding data for the Sham animals (n = 4) at day 0 before (control: Ctl; immediately after sham surgery: ‘Sham’), and at days 3 and 6 after sham surgery, showing essentially stable CBV over time with no statistically significant changes.

Table 1 shows the individual rCBV data measured during occlusion for the two cortical regions, illustrating the depth of rCBV reduction in individual subjects across the group.

Table 1.

Individual rCBV values in somatosensory and motor regions during occlusion.

Rat	Primary somatosensory cortex	Motor cortex
1	0.16	0.51
2	0.17	0.23
3	0.04	1.18
4	0.17	0.61
5	0.51	0.67
6	0.05	0.38
7	0.27	1.11

Individual occluded/control hemisphere ratio of weighted median fUS signal, normalized to baseline. See Methods for details.

Figure 4 shows the data extracted from the 60 min movie. Figure 4(a) shows the mean %CBV (n = 7) for the occluded-side somatosensory and motor ROIs. There was a continuous rise of %CBV over the entire duration of the fUS movie. Figure 4(b) shows the individual data for the seven rats and the two brain regions. This illustrates the remarkable temporal resolution and good signal-to-noise ratio in individual animals, showing also clear inter-subject differences in reperfusion slope. The individual slope values are shown underneath each time–signal curve, together with r ² . There was no significant relationship between the reperfusion slopes and the rCBV values measured during occlusion or immediately after recanalization (i.e. just before the start of the movie).

Figure 4.

(a) One-hour fUS movie of %CBV recorded over Bregma −2.00 mm within the somatosensory and motor ROIs following reperfusion, averaged across the seven MCAo rats (see Materials and methods section for details). CBV% represents the difference in fUS signal (i.e. not normalized to the non-occluded hemisphere) between the first 5 s frame of the fUS movie (t = 0) and subsequent time points (up to 60 min after thread withdrawal). (b) Corresponding individual CBV time-signal curve for each rat, illustrating marked inter-subject variability. Note that both areas showed similar mean reperfusion slopes. Next to each slope are shown the r2 Pearson linear coefficient and the slope (α/min).

Behavior

Garcia NS

No neurological deficit was present at any time point in any rat, except one MCAo rat that had a score of 17/18 on day 20 (decreased response to left whisker stimulation).

mSLT

Based on the one-way RM-ANOVAs, no significant change in mSLT performance was found in either the MCAo or the sham group (Supplementary Figure S1A).

BWT

The one-way RM-ANOVA for the tMCAo group revealed a significant time effect (p = 0.005), but post-hoc tests showed no significant difference from baseline for any time point. As shown in Supplementary Figure S1B, apparently random changes occurred over time without any clear physiological pattern. Regarding the sham group, the one-way RM-ANOVA showed no significant time effect. The two-way RM-ANOVA comparing directly the two groups showed no significant interaction (p = 0.21).

Immunofluorescence

No instance of pan-necrosis (i.e. infarction) was observed in any rat, but SNL was present in at least one area in each MCAo rat, as compared to one in four sham rats (p = 0.024). Figure 5 illustrates typical findings for the striatum and cortex. Marked striatal SNL, associated with moderate-to-marked microglial activation and marked astrocytosis, was present in five MCAo rats as compared to 0/4 sham rats (p = 0.06; Fisher test). Ischemic changes were milder in the cortical areas as compared to striatum and were present in 4/7 MCAo rats vs. 1/4 sham (p = 0.55). Importantly, the sham rat had the lowest volume of delineated cortical SNL of all subjects (0.043 mm³). The mean (± 1 SD) volumes for SNL (i.e. Vol_Total) are shown in Table 2. There was a significant difference between the MCAo and sham groups for total SNL, i.e. both striatum and cortex together (p = 0.0061, Mann-Whitney test) with a strong trend for striatal SNL (p = 0.06) but not for cortical SNL (p = 0.23). There was no significant correlation between mSLT or BWT data and Vol_Total data across the seven tMCAo rats for cortical, striatal or total SNL nor between the latter and the reperfusion slopes.

Figure 5.

Selective neuronal loss (SNL) revealed 21 days after tMCAo by NeuN immunofluorescence on coronal sections through the cortex (a) and striatum (b) in a representative rat. For the cortex, a low magnification image is shown on the top, with higher magnification images of the colored boxes (blue: occluded side; red: contralateral side) below. In addition, high-magnification glial fibrillary acidic protein (GFAP) and ionized calcium-binding adaptor molecule 1 (Iba1)-stained sections (representing astrocytosis and microglial activation, respectively) obtained at the same anatomical level in the same rat, are shown underneath the NeuN sections. These images illustrate clear-cut patchy SNL affecting the cortex (red arrows), and irregularly shaped SNL in the striatum of this rat. See Materials and methods for the definition of SNL. Note the incomplete loss of neurons within the abnormal areas, and the close topographical congruence between NeuN staining loss and increased Iba1 and/or GFAP staining, indicating a close association between SNL, microglial activation, and astrocytosis. The colored contours illustrate for the striatum the consensus ROIs for each immunostain (see Materials and methods section).

Table 2.

Selective neuronal loss volume assessed by immunofluorescence.

	Cortical	Striatal	Total
tMCAo	0.08 ± 0.13	0.22 ± 0.19	0.29 ± 0.14
Sham	0.01 ± 0.02	0	0.01 ± 0.02

tMCAo: Transient middle cerebral artery occlusion;

Mean (and one SD) total volume (mm³) of selective neuronal loss (SNL) as assessed by means of NeuN immunofluorescence, for the cortical and striatal areas separately and both together, across the seven tMCAo rats. For each rat, a weighted mean volume was obtained across all brain sections with ROIs delineating SNL, so called ‘Vol-Total’ (see Materials and methods section).

Discussion

This pilot study is the first to document that fUS imaging is able to map the changes in CBV occurring during and after MCA occlusion. Thus, we were able to obtain images of CBV at high spatial resolution (∼100 µm, Figure 2) and to chart the fine dynamics of tissue reperfusion in the individual rat at high temporal resolution (Figure 4). In addition, the thinned-skull preparation allowed us to longitudinally acquire chronic rCBV maps over six days following tMCAo without affecting brain physiology (Figure 3). The fUS therefore appears particularly well adapted to image cerebral perfusion in experimental stroke at high spatiotemporal resolution not only in the hyper-acute stage but also chronically in the same subject.

fUS imaging

We showed that fUS is able to provide quantitative mapping of perfusion following experimental stroke. As expected for brief tMCAo inducing limited tissue damage, the time course of rCBV involved first a marked decrease during occlusion by more than 50% relative to baseline (with greater involvement of the somatosensory relative to motor areas, which are located more peripherally in the MCA territory), ²⁰ followed by a return towards normal values immediately after reperfusion, and a steady increase thereafter leading to normal values by day 3 that remaining stable at day 6. Previous perfusion imaging studies using PET,^26–28 MRI^29–31 or CT ⁹ in animal stroke models have consistently reported CBV decreases in the ischemic area relative to contralateral hemisphere, though with variable degrees of reduction probably reflecting differences in vascular compartments assessed by and spatial resolution of the imaging technique used, MCAo model and timing of measurement as well as species used. Interestingly, a recently published rat study using photoacoustic imaging ³² reported CBV reductions around 70% following branch MCA occlusion, similar to our figure of ∼80%.

Specifically regarding the data extracted from the 1 h movie (Figure 4), we observed a clear and continuous reperfusion across rats in both the motor and somatosensory areas, not reaching a clear plateau by the end of recording. The motor and sensory areas exhibited approximately the same mean slope during reperfusion (15% and 12% per hour relative to first frame, respectively), while at the individual level, the slope ranged from −4.0% to + 27.8%/h (Figure 4(b)). Since the fUS data used to determine these slopes were expressed relative to that in the first 5 s frame, the latter could be affected by the actual signal present during MCAo or just before start of the movie. However, no clear relationship was present between the slopes and the rCBV value either during MCAo or immediately after it. We are not aware of any previous study that has assessed the relationship between the individual or group-based perfusion slope following recanalization and the severity of previous ischemia during MCAo. More generally, to our knowledge no data of the kind reported here are available in the literature. This serves to highlight the power of fUS to record reperfusion dynamics in real-time, in turn opening new avenues in understanding the extent and dynamics of post-ischemic reperfusion, which are poorly understood at present. Further developments of fUS should allow us to follow in real-time the effects of occlusion and reperfusion, for instance using five frames/s as compared to 0.2 here.

Another advantage of fUS relative to currently available optical methods such as intrinsic optical imaging (IOI), voltage sensitive dye imaging (VSDi) or 2-photon microscopy (2 PM), is that it is not limited to the cortical surface or a small tissue volume but provides CBV measurements in the depth of the cortex and with a wide FOV (see Introduction section).

The currently available fUS imaging system does have some limitations. Although it enables a macroscopic view of the vascular network including deeper structures, the latter are affected by low signal-to-noise ratio due to bone regrowth and were therefore excluded from further analysis in this longitudinal study. This is a concern given these areas are the most severely involved after proximal MCAo as carried out here, and as also documented by our immunofluorescence findings. This limitation is currently being addressed in the laboratory. Second, we applied here a 20 Hz high-pass filter to discard blood compartments that would otherwise overwhelm the signal. However, by doing this, small penetrating arterioles and ascending venules with a mean RBCs velocity <2 mm/s as well as capillaries were also discarded. Thus, only a fraction of total CBV could be measured. Third, for the analyses of the fUS images, we normalized the fUS signal to the contralateral hemisphere to generate rCBV data. There are several reasons for doing this: i) the measured fUS signal is dimensionless, and fUS imaging is currently capable to measure relative CBV changes only, not absolute CBV; ii) as just stated, in its current state of development the fUS signal from slow-moving vessels such as capillaries cannot be reliably measured and is therefore excluded at acquisition using low-pass filter; thus, the rCBV obtained is biased towards larger blood vessels; iii) even using the same experimental set up and fUS parameters, the absolute fUS signal may differ from one rat to the next due to e.g. minute differences in (thinned) skull thickness and other local factors; and iv) as can be seen in the fUS images presented in figure 2A, the brain vasculature is asymmetrical between the left and right hemispheres due to physiological variability in vascular anatomy. Thus, in order to reduce this intrinsic variability and meaningfully compare rCBV in each ROI over time and in subject groups, normalization was carried out first between hemispheres and then to the baseline ratio. Normalizing to the non-occluded hemisphere is widely used to reduce the effects of inter-subject global variations in perfusion measurements, and is underpinned by the consistently reported lack of acute hemodynamic changes in the contralesional hemisphere in various species including rats,^26–29 while normalizing to baseline is common in the stroke literature when assessing temporal changes relative to pre-intervention baseline. Nevertheless, we also analyzed post-hoc the non-baseline normalized occluded/non-occluded ratios, see Supplemental Figure 2. Despite the expected greater variability, the pattern of changes post-MCAo was the same as with the baseline-normalized ratios, and the statistical significance was similar. Finally, longitudinal fUS imaging could not be carried out later than six days after thinned-skull procedure because of bone regrowth and thickening possibly exacerbated by local inflammation. Research on stabilization of the cranial window is underway in our lab to adapt the protocol for longer time periods.

Additional limitations were that, because of procedural requirements of the thread MCAo model requiring the animal to be laid on its back at the time point of thread withdrawal, we were unable to record fUS data at the time reperfusion started. We could therefore have missed transients such as rebound hyperperfusion, although the latter phenomenon would be expected to last longer than 15 min.^33–35 Likewise, we did not acquire a fUS movie at the time MCA occlusion was initiated and during it which would be of interest to assess stability of CBV. Again, animal placement for safe thread pushing up impeded concomitant imaging, although a model variant allowing to push the thread without moving over the animal has been recently reported. ³⁶ Note that these limitations do not necessarily apply to MCAo models where access to the MCA is lateral or from above the skull, such as with distal occlusion. Assessing the perfusion changes immediately after occlusion and following them in real-time over the duration of the latter and after reperfusion will likely afford new insights into the mechanisms of ischemic brain damage.

Another strong interest of fUS is that it holds potential for brain imaging in the freely moving condition. Indeed, the ultrasonic probe can be miniaturized and implanted on the rodent's head. ¹³ The chronic fUS recordings presented in this paper are an important first step toward this with the possibility to induce ischemia in awake rodents. Finally, it might be possible in the future to image the full 3D brain in a single acquisition by using 2D matrix transducers, and to image both CBV and CBF.

Immunofluorescence

Some SNL was present in all MCAo rats versus one-fourth sham rats, a significantly different distribution, and striatal SNL in 5/7 tMCAo rats as compared to 0/4 sham (p = 0.06). The SNL engulfed most of the striatum and was associated with marked astrocytosis and variably severe microglial activation. Regarding the cortex, SNL was present in 4/7 tMCAo rats only, was less marked than in the striatum and was associated with clear astrocytosis but less conspicuous microglial activation (Figure 5). Accordingly, the inter-rater reproducibility for detecting SNL was good for the striatal area but poor for the cortex. Surprisingly, even after consensus, a sham rat had one (very small) delineated area of apparent SNL in the cortex. Because this observation did not seem biologically plausible, we performed post-hoc direct cell counting for NeuN, GFAP and Iba1 staining in all rats with delineated SNL areas. These measurements were carried out within a box positioned in the centre of the ROI, where the percentage of remaining NeuN-labelled cells was quantified as compared to the mirror area, by means of the Particle Analysis command in ImageJ. For Iba1 and GFAP, pixel intensity was quantified by means of the Integrated Density measuring function also in ImageJ, again relative to the mirror area. This revealed that the apparent SNL area in the sham rat had in fact the least reduced NeuN affected/contralateral ratio (0.67, as compared to 0.40 ± 0.17 (range 0.20–0.59) for cortical SNL ROIs in MCAo rats) with no increase for either inflammatory stain (Iba1: 0.94 vs. 1.37 ± 0.40; GFAP: 0.081 vs. 1.23 ± 0.09; comparable values for the striatal ROIs in the MCAo rats: 0.22 ± 0.14 (range: 0.065–0.44), 1.21 ± 0.21 and 1.34 ± 0.25, respectively), suggesting it probably represented a staining artifact.

The lack of infarction in any rat in the present study fits our aims (see Introduction section). It is also consistent with the complete and permanent tissue reperfusion with return to baseline CBV seen with fUS from day 3 onward. ²⁰ Also, late-appearing (≥ 12 h) post-ischemic hyperperfusion, which was not present in our MCAo group, has been regularly associated with the presence of tissue infarction in rodents.^34,35 The lack of early hyperperfusion in our study (Figure 3) is also consistent with the absence of infarction. Thus, in their classic 2 h MCA clip occlusion study, Tamura et al. ⁴ noted that in cats with moderately reduced CBF during occlusion, gradual reperfusion without hyperperfusion occurred over 1–2 h following clip release, and was associated with good tissue outcome, whereas in rats with severe occlusion-phase ischemia, release of the clip resulted in highly variable patterns of reperfusion, the frequent occurrence of early hyperperfusion being associated with tissue infarction. Classic PET studies in cats subjected to 30–120 min MCAo also showed that acute post-clip release hyperperfusion was associated with worse tissue outcome. ³⁷ Another frequently observed post-MCAo phenomenon, also not observed here, is delayed hypoperfusion, i.e. a secondary drop in perfusion following early complete reperfusion. ³⁸ This is again consistent with the mild ischemic damage present in our subjects, given that this phenomenon is seen following prolonged severe ischemia and is a marker of proceeding infarction involving secondary no-reflow from endothelial damage, inflammation and pericyte constriction.^4,36

The presence of SNL rather than infarction with 45 min tMCAo in Sprague-Dawley rats also agrees with current literature that almost consistently reports infarcts with 60 min occlusion and pure SNL with 30 min occlusion. ¹⁶ Although given the notorious difficulties with the thread occlusion model ³⁹ incomplete MCAo could be raised as a possible explanation to our finding, adequate occlusion of the origin of the MCA was likely achieved firstly because the post-mortem data showed striatal damage, and secondly because MCAo induced extensive fUS signal reductions (Figure 2(a)) that were marked in each rat (Table 1). Thus, rCBV during occlusion was on average 20% of baseline in the whole somatosensory area (Figure 3), and reached even lower values in the somatosensory ROI at the most severely affected coronal plane (see Results section). The finding of more consistent and conspicuous SNL and inflammation in the striatum as compared to the cortex was expected with brief proximal MCAo, where cortical SNL is generally mild or even absent. ¹⁶ In agreement with previous reports, ¹⁶ although striatal SNL was conspicuous, it varied markedly from subject to subject, being absent in two rats and of variable intensity in the remaining five. Intriguingly, in some rats striatal SNL was severe and associated with marked microglial activation and astrocytosis (Figure 5), reminiscent of previous reports of striatal SNL gradually progressing to near complete neuronal loss over several weeks after proximal MCAo.^40–43 Finally, there was no significant correlation between the fUS-based reperfusion slope and the severity of SNL. This does not, however, rule out the possibility that reperfusion slope relates to more severe degrees tissue damage. To our knowledge, no previous study has assessed the relationship between reperfusion slope and tissue damage after MCAo.

Behavior

In this study, we used both the simple Neuroscore and two sensitive sensorimotor tests to assess the behavioral effects of the ischemic injury over three weeks. There were no effects whatsoever on the Neuroscore and the mSLT. The unchanged Neuroscore is consistent with the lack of actual infarction. ¹⁶ Regarding the BWT, significant time effects were present with the ANOVA but without any clear and biologically plausible pattern; in addition, there was no significant difference between tMCAo and sham groups, altogether indicating the lack of meaningful changes. Although the histopathological changes were mild, more conspicuous behavioral effects might have been expected based on the single previous report on brief proximal MCAo in Sprague-Dawley rats. ⁴⁴ However, the site of SNL is known to dictate the behavioral effects of stroke, with striatal SNL seemingly having less clear sensorimotor effects than cortical SNL.^16,45 Severe striatal SNL has, however, been reported to impair performance on sophisticated cognitive tests,^46,47 but these were not implemented in the present study.

Conclusion

Although further validation of fUS imaging against gold-standard methodologies, and further documentation of its utility in experimental stroke using different occlusion durations and MCAo models are required, the present pilot study outlines how recent advances in fUS imaging enable high-resolution mapping of mesoscale-level structures, namely arterioles and venules within cortical blocks; the characterization of the ischemic area with respect to regional differences in blood volume during transient occlusion and subsequent reperfusion; and the assessment of the dynamics of reperfusion in the individual rat with an unprecedented time resolution. Importantly, the thinned-skull preparation allowed longitudinal fUS recording for up to six days post-surgery. We anticipate that with the use of existing surgical techniques for distal MCAo and further miniaturization of the ultrasound equipment, high-resolution measurements will be feasible in anesthetized or awake animals to continuously chart the effects of MCAo and reperfusion in the whole brain, including deeply situated structures, both in the resting state and after sensory stimulation. Finally, fUS could be important for pharmacological studies of potential protective agents in rodent stroke models.