All animal preparations and procedures were performed in accordance with the Institutional Animal Care and Use Committee (IACUC) of Sungkyunkwan University. Male, 9-week-old, Sprague-Dawley rats (280~320g) were used (OrientBio, Seongnam, South Korea). All rats were housed in humidity- and temperature-controlled facilities with 50~60% humidity at 23℃ under a 12-h light/dark cycle.

For the MCAO procedure, the rats were anesthetized and maintained with 2% isoflurane and oxygen gas (VetEquip, CA, USA). A heating pad was used to maintain the animal's body temperature throughout the procedure (FHC, ME, USA). We carefully made an incision in the neck skin and exposed the right common carotid artery (CCA). A 4-0 monofilament nylon suture coated with silicon was introduced from the CCA into the middle cerebral artery (MCA). The insertion length of the suture (18~20 mm) and thickness of the silicon head (330~380 µm) varied according to the body weight of the animal [17]. After MCA occlusion, the incision was temporarily closed and the animal was returned to home cage for either 60 min (N=3) or 90 min (N=12). Then, the animal was anesthetized again with 2% isoflurane, the silicon-coated suture was removed for reperfusion, and the incision was closed.

According to prior studies, the infarct volume of the MCAO model reached a maximum after 24 h of reperfusion [18]. Therefore, we conducted the mNSS behavioral testing after 24 to 72 h of reperfusion. We focused on assessing primary sensory cortical functions to compare the hemodynamic changes in the primary sensory cortex and behavioral consequences in the MCAO model. The mNSS is composed of three different behavioral tests: motor, sensory, and beam balance tests (Fig. 1A). Detailed mNSS procedures were described by Chen et al. (2001) [19].

We performed the ORIS imaging right after completion of the behavioral test. We carefully removed the skull of both hemispheres under 2% isoflurane anesthesia. The animal's health was carefully monitored throughout the experiment. Two tungsten electrodes (Plastics One, Roanoke, VA, USA) were placed onto the intact dura mater of each hemisphere (Fig. 1B), which were connected to a pulse generator (Master-8, A.M.P.I., Jerusalem, Israel). Functional hemodynamic imaging was performed using an optical imaging system (Imager 3001, Optical Imaging, Rehovot, Israel) with an external light source (Leica Microsystems, Manheim, Germany). Reflected light was collected through two 50 mm tandem lenses (Nikon, Tokyo, Japan) and filtered using a 570±10 nm band-pass filter. The 570 nm is the isosbestic wavelength of total hemoglobin, which is equivalent to CBV changes. The images were acquired for 300 s at 3.33 Hz. The images recorded during the initial 30 s before applying 10 s stimulation (1 mA, 3.33 Hz, 10 ms pulse width) to the cortex were averaged and designated as the baseline image. Both the ipsilateral hemisphere (I.H), which had the infarction, and the opposite contralateral hemisphere (C.H) were alternatively imaged. In order to ensure hemodynamic recovery to baseline levels, the inter-trial interval was set at 900 s (Fig. 1C).

After ORIS imaging, the entire brain of the rat was extracted and sectioned into six coronal slices, 2 mm thick, using a rat brain matrix (World Precision Instruments, Sarasota, FL, USA). All coronal slices were stained with 2% 2,3,5-triphenyltetrazolium chloride (TTC) solution (Sigma-Aldrich, St. Louis, MO, USA) for 20 min and fixed with 4% paraformaldehyde at 4℃. After 24 h of initial staining, the six slices were photographed with a graduated ruler using a digital camera (Canon, Tokyo, Japan). The infarct volume of each slice was quantified using Fiji software (NI.H, MD, USA). We classified the MCAO models into two groups according to their infarct volume; the mild MCAO group had less than 90 mm³ of infarct area that was primarily located in the striatum, while the severe MCAO group had greater than 90 mm³ of infarct area that included areas in the cortex as well as the striatum.

The activation maps shown in Fig. 3A, C, and E were calculated by dividing every frame by the mean of the baseline frame. We measured the time course of CBV change in the region of interest (ROI, 7×7 pixels) of the cortical tissue, quantifying both the maximum value of CBV change and the time to reach that maximum value. The ROI that was selected included the tissue area that showed the strongest CBV changes but excluded the large vessels. The ORIS data were analyzed using custom-written MATLAB software (MathWorks, Natick, MA, USA).

To compare hemodynamic differences in the arteries between the MCAO and normal groups, we defined the reactive artery from images and quantified them as shown in Fig. 4A. First, the raw images were cropped (3.7×5.9 mm²) and segmented into cerebral blood vessels and surrounding tissues using the trainable Waikato Environment for Knowledge Analysis (WEKA) segmentation plugin, which combines a collection of machine learning algorithms in Fiji [20]. Based on the segmentation results, the probability of each pixel in the raw image belonging to the segmented vessel was verified by the WEKA tool. The verified pixel probabilities, from 0% to 100%, were mapped in an intensity form. For accurate extraction of the blood vessel without intervention, such as a cerebral hemorrhage and unintended light reflection from the dura mater, we set the threshold value at 60% of the probability for vessel classification (Fig. 4A). This threshold value was a balance between the detection ratio and a false-positive ratio and was determined through an empirical analysis. In a sequence, the reactive region was defined by mapping the standard deviations (SD) of each pixel's value over time. All maps were filtered by a 3×3 Gaussian kernel to prevent the signals from interfering with the outliers. To exclude a low level of variation, which was defined as no response over time, a total of 1,500 pixels were gathered from the highest value in the calculated SD map and were binary coded. A binary coded image was multiplied to all frames of the vessel probability maps in order to extract the reactive vessels, which were defined as arteries, from the vessel probability map (Fig. 4A). The final step of quantification was to measure the time course of CBV change in the extracted artery in each frame. The measured intensity of artery was calculated to extract hemodynamic characteristics, such as maximum ΔCBV and time to peak to compare the differences between the groups using statistical analyses with Kruskal-Wallis nonparametric tests. SD mapping, binary coded images, and statistical analyses were all performed using the custom-written MATLAB software.

The MCAO rats were separated into two groups, a mild MCAO group (N=5) and a severe MCAO group (N=10), according to the volume of the infarct (Fig. 2B-D). The majority of the animals that received 90 min of occlusion had severe ischemia, while all of the animals that received 60 min of occlusion and 2 animals that received 90 min of occlusion had mild ischemia. The infarct area in the mild MCAO model was primarily in the striatum (Fig. 2C), while the severe MCAO group exhibited infarcts in both the cortex and subcortical areas (Fig. 2D). Fig. 2A describes the behavioral and physical discrimination between the mild and severe MCAO groups. The average infarct volume of the severe MCAO model (275.44±32.97 mm³) was significantly greater (p value<0.001) than that of the mild MCAO model (15.69±5.88 mm³). In the severe MCAO groups, animals had a higher mNSS score (mild MCAO, 6.0±1.05 points; severe MCAO, 9.9±0.35 points; p value<0.001) and greater weight loss than the mild MCAO group (normalized to weight before MCAO; average±SEM; mild MCAO, 95.02±2.50%; severe MCAO, 83.88±1.80%; p value=0.003; independent t-test).

In the normal group, CBV around the stimulating electrode increased immediately after the electrical stimulation and peaked in both hemispheres at about 105 s, as shown in two-dimensional images of the whole hemisphere (Fig. 3A, B). In the mild MCAO group, the CBV increase following electrical stimulation was also observed in both hemispheres (Fig. 3C, D). In the severe MCAO group, the CBV of the I.H decreased slightly in the region around the stimulating electrode, while the CBV of the C.H increased after electrical stimulation (Fig. 3E, F).

The maximum CBV change in both hemispheres increased by 30.13±1.16% at 104±4.16 s after stimulation in the normal group. As expected, CBV change in the I.H of the severe MCAO group (-8.98±0.72%) was significantly lower than CBV change in the corresponding C.H (30.83±0.94%) (average±SEM; p value<0.001, paired t-test) (Fig. 3G). The I.H of the severe MCAO group showed significant differences in maximum CBV (p value<0.001) and time to reach to maximum peak (p value<0.001) compared to the normal group, while both the maximum CBV change (C.H, 26.45±3.92%; I.H, 27.19±1.98%) and the time to reach the maximum CBV change (C.H, 91.65±12.87 s; I.H, 89.80±3.69 s) in the mild MCAO group showed non-significant differences compared to the normal group (max CBV, p value=0.912; time to max ΔCBV, p value=0.923) (Fig. 3G-H). In the severe MCAO animal, direct electrical stimulation onto the somatosensory cortex created an altered pattern of CBV change in the ipsilateral hemisphere compared with standard CBV changes in that region. Thus, normal autoregulation of neurovascular coupling may be significantly disrupted on the ipsilateral side of severe MCAO animals.

To examine arterial responsiveness as the increase in CBV spreads throughout the infarcted brain, we segmented arteries using the WEKA methods (Fig. 4A). Representative images of the reactive artery (red color) are shown in Fig. 4B. However, segmented arteries could not be obtained from the I.H of the severe MCAO model because electrical stimulation to the I.H of the severe MCAO group did not cause a significant change in CBV. The averaged maximum change in CBV in arteries increased up to 39.91±1.57% at 93.85±4.87 s in the normal group (Fig. 4C). The CBV change in arteries of the mild MCAO group was faster (C.H, 78.45±7.19 s, p value=0.090; I.H, 53.70±14.10 s, p value=0.005) and smaller (C.H, 31.99±2.60%, p value=0.275; I.H, 26.65±7.38%, p value=0.379) than normal group in both hemispheres during electrical stimulation (Fig. 4D), which was distinct from the hemodynamic reactivity in tissue (Fig. 3G-H). In the C.H of the severe MCAO group, the reaction of the arteries was the slowest (109.45±3.58 s) and the smallest (27.75±2.34%) compared with the C.H of the other groups (Fig. 4E). The differences of the time to reach maximum CBV change of the artery were statistically significant in the C.H of the severe MCAO group compared with the normal (p value=0.023) and C.H in mild MCAO (p value=0.014) groups. However, the maximum CBV change of the tissue area in the C.H of the severe MCAO group was not significantly different among the three groups (Fig. 3G). These results suggest that in the severe MCAO group, the autoregulatory function of arteries in the contralateral hemisphere was disrupted.

We next investigated whether hemodynamic reactivity was correlated with the severity of ischemia in Fig. 5. We correlated pathophysical indexes (mNSS, weight loss, and infarct size) with the hemodynamic indexes (maximum change in pixel intensity and time to maximum pixel intensity) of segmented artery and tissue ROI. The time to peak ΔCBV in artery has the highest correlation score with infarct size among all correlations (r=0.738, p value=0.004) (Fig. 5A). Fig. 5B summarizes correlation coefficient values from correlation analysis of quantified hemodynamic values of the three groups along with pathophysical information.