Delayed perfusion phenomenon in a rat stroke model at 1.5T MR: An imaging sign parallel to spontaneous reperfusion and ischemic penumbra?

2.1. Animal models of stroke 

This experiment was performed in compliance with the guidelines of the International Committee on Thrombosis and Hemostasis [10] and the current institutional regulations for use and care of laboratory animals. Eight male Sprague–Dawley rats weighting approximately 350g were included in this study. Anesthesia was performed with initial inhalation of 4% isoflurane for 3min and maintained with 2% isoflurane in a mixture of 20% oxygen and 80% room air. Body temperature was kept at 37.5±0.5°C during the surgical procedure using a heating pad. MCA was occluded by the PIT approach, as described in detail elsewhere [5], [6], [7].

The presence and degree of stroke-induced brain damage were assessed using a method of neurological deficit score described by Bederson et al., which scores systematically neurological and behavioral aspects of the rat [11].

2.2. MR imaging 

MR imaging was conducted with a 1.5T Siemens Sonata scanner (Erlangen, Germany) using a commercially available 4-channel phased array wrist coil (MRI Devices Corporation, Waukesha, WI, USA). Rats were placed supinely into a plastic holder and connected to the same breathing anesthesia system as for the surgery. The penile vein of rats was cannulated with a G25 infusion needle for contrast agent injection. For each imaging sequence 12 coronal slices were acquired with a slice thickness of 2mm gapped at 0.2mm.

T1-weighted spin echo MRI (T1WI, TR/TE=500ms/15ms) was acquired with a field of view (FOV) of 55mm×34.4mm and an imaging matrix of 256×120, resulting in an acquisition time of 3min and 21s. T2-weighted turbo-spin echo (T2WI, TR/TE=5860ms/100ms) was acquired with a FOV of 55mm×31mm and an imaging matrix of 256×146, resulting in an acquisition time of 3min and 9s.

Diffusion-weighted images (DWI, TR/TE=3000ms/80ms) were obtained using a 2D spin echo echo-planar imaging sequence (SE-EPI) with a FOV of 140mm×70mm and an imaging matrix of 192×96, resulting in an in-plane resolution of 0.7mm×0.7mm and an acquisition time of 4min and 20s. In order to reduce susceptibility artifacts and scanning time, a parallel imaging technique namely generalized autocalibrating partially parallel acquisition (GRAPPA) (Siemens, Germany) with an acceleration factor of 2 was applied. For DWI, three directions (x, y, and z) were measured and averaged for calculation of the isotropic apparent diffusion coefficient (ADC) value. ADC maps were calculated based on three different b-values (0s/mm2, 500s/mm2, 1000s/mm2).

DSC-PWI was acquired using a T2* weighted (TR/TE=2000ms/40ms) EPI sequence in combination with the GRAPPA technique to increase spatial homogeneity in the images. A dynamic series of 100 measurements (12 images per measurement and 1200 images in total) resulted in a scan time of 3min and 34s, with a FOV of 156mm×43.5mm and an image acquisition matrix of 128×64, leading to an in-plane resolution of 0.7mm×0.7mm. During the dynamic series, bolus injection of Omniscan (Amersham S.A., Norway) at a triple dose (0.3mmol/kg) was started after the 30th imaging acquisition to ensure a sufficient precontrast baseline. The bolus injection was administered manually in less than 1s without saline flush.

The entire MRI protocol approximately lasted for 25min for each animal at each time point.

To observe the serial changes of cerebral ischemia, all rats were scanned with MR imaging at 1h, 2h, 3h, 6h, 12h, 24h, and 72h after MCA occlusion under the s above mentioned anesthetic regime.

2.3. Postmortem procedures 

For the verification of the stroke model, the documentation of the patency of MCA and the quantification of arterial collaterals across the longitudinal fissure, we conducted the following postmortem macro- and microscopic procedures at 72h after MCA occlusion. After euthanasia of the animals with an intravenous overdose of pentobarbital, the chest was opened, the thoracic aorta was clamped at diaphragm level, and the left ventricle of the heart was infused with 1ml barium suspension (Micropaque®, Guerbet, Cedex, France). Subsequently, the brain was excised and radiographed with a mammographic soft X-ray using a microfocus radiographic technique (25–26kV, 2.8–5.0mAs).

Brains were then sliced into six preselected coronal sections (2mm thick slices from anterior 3.5mm to 13.5mm) using a brain matrix (Agar Scientific, England). Each section was incubated at 37°C in 2% triphenyl tetrazolium chloride (TTC) solution for 15min and normal brain tissue stained brick red while cerebral infarct remained pale. For microscopy, the brain sections were then fixed with 10% formalin and processed with cresyl violet staining, which is widely used for studying neuronal tissue. As a basic stain, it readily interacts with the acidic components of the neuronal cytoplasm, such as the RNA-rich ribosomes, as well as nuclei and nucleoli of the cells.

2.4. Image analysis 

The image analysis was performed off-line on a LINUX workstation using dedicated software (Biomap, Novartis, Basel, Switzerland).

For the quantification of relative signal intensity (rSI) and relative lesion volume (rLV) on MR images of relevant sequences, the areas of the lesion and contralateral region and/or the entire brain were delineated with consensus by three experienced authors (F.C., N.N., Y.N.) using an operator-defined region of interest (ROI) on each of the lesion-containing slices. The software automatically generated the SI, LV, and brain volume for each animal, from which rSI and rLV were calculated, i.e., rSI=lesion SI/contralateral SI×100%; rLV=lesion volume/brain volume×100%.

At the contralateral region covering the origin of the right MCA, a 25mm2 ROI, which was further divided into 49 pixels of 0.72mm×0.72mm, was placed to measure the arterial input function, from which we selected those pixels on the panel representing the right MCA branch to automatically generate PWI-derived rCBF map with the built-in software.

For the PWI and the PWI-derived parameter map, only one central lesion-containing slice (the sixth slice from the tentorium of cerebellum) involving both cortex and striatum was chosen for lesion delineation. The hyperintense lesion on the PWI source images, which was more convincing than that on the PWI-derived map, was first contoured with consensus, and subsequently transferred onto the rCBF map. The automatic readout of abnormal rCBF from the region of perfusion deficiency as a percentage of that measured in the contralateral brain was generated using the software.

Similarly, TTCrLV was obtained with commercial image analysis software (Adobe Photoshop 7.0) as a postmortem standard of infarct size.

The penumbra or brain tissue at risk of infarction [4] at each time point was operationally defined as the mismatch found from the PWI source image and the DWI, i.e., PenumbrarLV=PWIrLV−DWIrLV, where PenumbrarLV>0 denotes the presence of a penumbra [12], [13], [14].

A DP sign at the PWI source image was defined as the delayed occurrence of a hypointense area at the left MCA territory due to a sluggish contrast arrival compared to the corresponding contralateral brain [2] (Fig. 1). It appeared initially as a part of the entire hyperintense zone (perfusion deficit) on the first image at the contrast arrival, became hypointense on one of the following images and returned to isointense within a few seconds afterwards. To quantify the rLV and rCBF changes in the DP region, the hyperintense zone on the first image (Fig. 1B Zone1) and the second image (Fig. 1C Zone2) were delineated; the contour of Zone1 on the first image was transferred to the second image and subsequently the zone of DP was derived, i.e., Zone3 (DP)=Zone1−Zone2. The rLV of the DP and rCBF for Zone1, Zone2, and Zone3 were generated. The duration of the DP sign was recorded for each animal at each time point based on the mean SI-time curve (intervals of A–C at yellow curve and A–D at red curve in Fig. 1E). A DP sign was positive if a secondary nadir occurred from Zone3 (C at the red curve in Fig. 1E), whereas the DP sign was negative if the nadir was overlapping with that in the contralateral normal brain cortex (B at the yellow curve in Fig. 1E).

View Large Image Download to PowerPoint

Fig. 1. Determination of delayed perfusion phenomenon on PWI in rats after photothromotic MCA occlusion. (A–D) Serial source images of PWI at the same slice at four different time points. (A) Baseline before Gd-DTPA arrival. (B) Showing hyperintense region, or perfusion deficit (Z1=Zone1) 2s after Gd-DTPA arrival with hypointense normal brain noted. (C) 2s after (B), a delayed perfusion of hypointense area within the Z1 was noted resulting in a reduced perfusion deficit region (Z2=Zone2). (D) 2s after (C), the signal of delayed perfusion region (Z3=Zone3) recovered to normal intense suggesting a delayed arrival of Gd-DTPA. (a–d) Schematic representation of serial source images of PWI at four different time points matched with (A–D). (E) Mean signal intensity-time curves from delayed perfusion region (Z3, red line curve corresponding to the region of interest in red in (C)) and contralateral cortex (yellow line curve corresponding to the region of interest in yellow in C). The lines of (A–D) across the curves correspond to the time points to acquire images of (A–D).

From the cerebral surface of the brain specimens, the barium-filled arterial collateral branches in the area of 2mm laterally alongside the longitudinal fissure were counted and compared among the animals.

2.5. Statistical analysis 

Statistical analysis was carried out with the SPSS for windows software package (release 11.5, SPSS Inc., Chicago, Illinois). The numerical data were averaged over all animals grouped together and reported as mean±standard deviation (S.D.). Paired or unpaired Student's t-tests were applied for parameter comparisons between different time points and groups. The volume and occurrence of penumbra were correlated with those of DP using linear regression analysis. A significant difference was considered if the p-value was less than 0.05.

3.1. General conditions 

All animals survived the surgical intervention and subsequent imaging protocol. The neurological symptoms for brain damage were assessed using Bederson's score [11] with a median of 3.8 (range: 2–6, n=8) before sacrifice at 72h (Table 1).

Table 1. Comparisons between the rats with and without cortex damage

	Brain damage	Neuro scorea	Reperfusion time pointb (h)	DP durationc (second)	MCAd	Collateral arteriolese	TTC positive
	Striatum only (n=4)f	5	6	6.7	Occluded	20	Striatum
		4	12	5.7	Occluded	17	Striatum
		2	6	6.3	Patent, narrow	25	Striatum
		3	6	6.0	Patent, narrow	19	Striatum
	Mean±S.D.	3.5±1.3	6.0±3.0	6.2±0.4		20.3±2.9
	Cortex and striatum (n=4)	4	6	8.7	Occluded	9	Cortex and striatum
		4	12	7.4	Occluded	8	Cortex and striatum
		2	12	6.7	Patent, narrow	13	Cortex and striatum
		6	12	6.0	Occluded	12	Cortex and striatum
	Mean±S.D.	4.0±1.6	10.5±3.0	7.2±1.2		10.5±2.1
	pg	0.65	0.21	0.012		0.004

a Neurological deficits assessment using Bederson scoring method. b Spontaneous reperfusion start time based on the rCBF of the ischemic region. c Delayed perfusion (DP) duration based on PWI source images. d Qualitative assessment of the MCA patency based on barium microangiography. e Collateral arterioles counted on brain specimen of barium microangiography within the region of 2mm to the midline of rat brain. f Four rows on the left represent data from four individual rats. g Comparisons of means between the rats with and without cortex involvement.

3.2. Signal intensity changes of T2WI, DWI, and ADC 

After MCA occlusion, the rSI of ischemic lesions on T2WI and DWI increased gradually from 1h, peaked at 12h, dropped slightly at 24h and then declined considerably at 72h. The changes of T2WI and DWI paralleled, but the rSI value was higher on DWI (T2 shine-through) than on T2WI. Whereas, the time course of the ADC of the ischemic area appeared reverse in comparison with that on DWI. The rSI value of the ADC increased to a maximum at 72h (Fig. 2A).

View Large Image Download to PowerPoint

Fig. 2. Longitudinal changes of relative signal intensity (SI) (A) and relative volume (B) in ischemic brain lesion after MCA occlusion.

3.3. Changes of relative lesion volume 

At 1h after MCA occlusion, the DSC-PWI source image revealed a larger cerebral ischemic lesion (PWIrLV=31.5±6.6%) compared to the DWI (DWIrLV=21.6±4%) and the T2WI (T2WIrLV=8.3±1.2%) images. Over time the lesion volume expended from the striatum to the left temporal cortex, and reached a maximum at 12h on PWI (PWIrLV=33.5±5.9%) or at 24h on DWI (DWIrLV=34.6±4%) and T2WI (T2WIrLV=33±4.1%) followed by a remarkable reduction at 72h (Fig. 2B). During the time course, the PWIrLV was constantly higher than DWIrLV until 12h. Comparing to TTCrLV (22±5.7%) at 72h, being the definite infarct size, PWIrLV was significantly higher (p<0.01) at 1h after MCA occlusion, while T2WIrLV and DWIrLV were significantly lower (p<0.01). However, at 72h both DWIrLV and T2WIrLV became significantly higher (p<0.01), whereas PWIrLV was comparable (p=0.13) (Fig. 2B).

3.4. Delayed perfusion and penumbra 

3.4.1. Imaging findings 

The DP phenomenon could be readily discerned on serial source images of PWI in all rats as exemplified in Fig. 3. Twelve hours after MCA occlusion, at baseline before contrast bolus, the SI appeared homogeneous over the entire brain, except for the local hyperintensity at the site of surgical and photochemical intervention to induce MCA occlusion, due to the T2-weighting nature of the applied DSC-PWI sequence (Fig. 3A1). Two seconds later when the contrast bolus arrived, the contralateral hemisphere turned suddenly into hypointense as a result of normal blood perfusion (Fig. 3A2). After a delay of 2 more seconds, the SI dropped at the left cortex region, suggesting the delayed contrast arrival or delayed local perfusion (Fig. 3A3). The SI of both hemispheres tended to recover after another 2s (Fig. 3A4). The durations of the DP phenomenon in each rat as obtained from the perfusion curves (Fig. 3B) are listed in Table 1.

View Large Image Download to PowerPoint

Fig. 3. Delayed perfusion phenomenon on PWI in a rat 12h after photothrombotic MCA occlusion with corresponding perfusion curves, images of T2WI and DWI, microangiography and histomorphology. (A1–A4) Serial source images of PWI at four time points. (A1) Baseline, 62s after starting the PWI sequence. (A2–A4) 64s, 66s, and 68s after starting the PWI sequence, a delayed perfusion (delayed contrast arrival) zone in the left cortex was visible. (B) The mean intensity–time curves generated from 100 images of PWI at the same slice indicated a delayed perfusion of contrast in the left cortex (red line) compared with the contralateral cortex (yellow line). The SIs for the curves was derived from the corresponding ROI shown on the inner picture. (C1–C2) T2WI and DWI displayed the infracted lesion. (D1–D2) Photographs of the dorsal (D1) and ventral (D2) view of the brain showed, respectively, the collateral anastomosis (within dotted rectangle) and disrupted left MCA (black arrow). (D3) Corresponding microangiography showed MCA occlusion (white arrows). (E1) TTC staining at 72h proved the presence of infarct. The rectangular frame denotes where the microscopy is focused. (E2) Cresyl violet staining clearly demonstrated the infarction in the striatum and viable tissue in the cortex (I: infarction, V: viable brain tissue).

The images of T2WI (Fig. 3C1) and DWI (Fig. 3C2) showed that the infarcted lesion was limited to the striatum but not the cortex, suggesting the presence of a penumbra at the cortex if cross-referenced with the PWI findings (Fig. 3A1–A4 and B).

3.4.2. Occurrence rates 

Similar to the clinical report on stroke patients [2], DSC-PWI displayed a DP phenomenon at the cortex region of the left MCA supplied territory in all rats (n=8) with the incidence of: 100% at 1–3h, 63% at 6–12h, 38% at 24h, and 25% at 72h (Fig. 4A). The incidence of penumbra showed a parallel trend among the same animals and both the penumbra and the DP sign showed an occurrence rate above 60% till 12h after MCA occlusion (Fig. 4A). The linear regression of the occurrence rates between these two features was significant (p<0.0001) with an r2 value of 0.96 (Fig. 4B).

View Large Image Download to PowerPoint

Fig. 4. Relationship between the occurrence of ischemic penumbra (PWI–DWI mismatch) and delayed perfusion (DP) region. (A) The occurrence of delayed perfusion phenomenon was similar with that of penumbra between 1h and 72h after MCA occlusion. (B) The linear regression between the occurrence of delayed perfusion and penumbra was significant (p<0.0001, r2=0.96).

3.4.3. Volume changes 

All the DP (Zone3) appeared about 2s lagging behind the maximum T2* effect at the contralateral normal brain and the hypointense area gradually returned to normal (Fig. 3A1–A4 and B).

The volume change of the DP region shared similar trends with the penumbra. Its volume reduced from 10.9±2.6% (of the brain volume) at 1h to 3.0±5.2% at 72h (Fig. 5A). Anatomically, the penumbra and DP area involved mainly the same cortex region (Fig. 3A1–A4, C). The linear regression between the volumes of DP region and penumbra were significant (p<0.01), with r2=0.85 (Fig. 5B).

View Large Image Download to PowerPoint

Fig. 5. Relationship of relative volume between the ischemic penumbra (PWI–DWI mismatch) and delayed perfusion region. (A) The volume change of the delayed perfusion region shared similar trends with that of the penumbra. The penumbra remained positive until 12h after MCA occlusion. (B) The linear regression between the volume of delayed perfusion and penumbra was significant (p<0.01, r2=0.85).

The penumbrarLV decreased from 10±5.4% (of the brain volume) at 1h to −3.3±4.5% at 72h, indicating that the ischemic lesion on PWI became smaller than that on DWI at late time points (Fig. 5A). However, the penumbrarLV remained positive during the time between 1h and 12h and unchanged during the time between 6h and 12h.

3.5. Changes of rCBF 

In the striatum (Figs. 1 Zone2 and 6), the rCBF remained almost unchanged during the whole ischemic period but slightly increased up to 72h. In the cortex or DP region (Figs. 1 Zone3 and 6), the rCBF decreased from 88.9% at 1h to the lowest level of 64.9% at 6h. Subsequently, the rCBF increased significantly at 12h (p<0.05) and this increase continued over the level of the contralateral area up to 72h. If the global perfusion deficiency (including both the DP area and the striatum) is to be considered (Zone1 in Fig. 1), the rCBF demonstrated a similar pattern with that in the striatum, except for a significant drop of rCBF at 6h (p<0.05) (Fig. 6). The reperfusion time points at which rCBF started to increase in each individual rat are given in Table 1.

View Large Image Download to PowerPoint

Fig. 6. rCBF changes over 72h. The rCBF of ischemic lesion in the cortex, striatum and total (cortex plus striatum) region decreased from 1h after MCA occlusion to a minimum at 6h after MCA occlusion. The rCBF then increased starting at 6h up to 72h indicating that a reperfusion phenomenon occurred (*p<0.05 compared to prior time points).

3.6. Microangiography, TTC staining, and histology 

Based on the extension of cerebral involvement on TTC staining, retrogradely the eight rats tended to form two groups of four rats each with different severities of ischemia as reflected on neurological scores, MR images, MCA patency, number of collaterals and tissue viability (Table 1; Fig. 3).

With cresyl violet staining, no histological abnormalities were microscopically observed in the contralateral cerebral regions of the ischemic rats. Necrotic neurons appeared with karyolysis and pyknosis of nuclei as well as vacuolation of cytoplasm, resulting in a regional loss of pigmentation and tissue structure (the region denoted by I in Fig. 3E2). In four rats widespread neuronal necrosis was found in the striatum and the cortex of the left MCA supplied region. However, in the other four rats limited necrosis was observed only at the striatum while the cortex was almost intact. In-between a transition zone existed, consisting of residual viable neurons mixed with infiltrating inflammatory cells, proliferated gliocytes and capillaries as a sign of early scar formation 3 days after cerebral infarction (Fig. 3E1 and E2).

4.1. Spontaneous reperfusion 

Although the PIT model of proximal MCA occlusion in rats has been applied for studying stroke with MRI [5], [7], [8] and testing new anti-stroke agent [9], spontaneous recanalization as a common feature of clinical thromboembolic stroke [15] has been very sparsely studied experimentally using this model. Using in vivo non-invasive MR imaging, we have been able to confirm, for the first time to our knowledge, the previously found reperfusion phenomenon based on postmortem autoradiography in the same rat model [6].

The evolution of rCBF in both the striatum and the cortex of the ischemic lesion obtained with DSC-MRI simulates the rCBF evolution observed with autoradiography during the insult process [6], i.e., the rCBF increased significantly starting at 6h after MCA occlusion, especially in the involved cortex region. This provides an in vivo evidence for the spontaneous reperfusion occurring in this particular model. In the previous experiments with the animal model identical to that used in the present study, the spontaneous reperfusion occurred between 3h and 8h [6], [16]. Our result (6h) falls well into this time range and was consistent with observations in animal and in clinical patients [17]. The exact reason for the variation was unknown. However, it may be related to the different techniques used, i.e., autoradiography versus MRI, and to the complex factors involved in the mechanism of the MCA occlusion in the model. A photochemically induced thrombus may first obstruct the proximal MCA initially. Over time the extent of the occlusion expanded toward the distal part of MCA, and fibrin-rich microthrombi formation occurred on the superficial capillaries and subsequently extended to the cortex and the striatum [18]. The rCBF decreased over time and reached a minimum at 6h after MCA occlusion. Next the diffused thrombus was dissolved probably due to an endogenous thrombolytic mechanism through, e.g. leptomeningeal collateral circulation [19] and/or leukocytes infiltration [18] starting at 6h following ischemia, resulting in a significant increase of rCBF.

4.2. Penumbra in PIT stroke model 

Saving the penumbra is the main purpose of the therapy for hyperacute and acute stroke and is the theoretical basis behind the reperfusion concept [13], [14], [20]. Previous study implied the presence of the penumbra in the PIT model [7] and the present study evaluated the detailed penumbra evolution.

Though decreasing in a time-dependent manner due to the expansion of the infarct core, the penumbrarLV (PWIrLV−DWIrLV) remained positive from 1h up to 12h after the ischemic insult and almost unchanged between 6h and 12h, which is consistent with the clinical results [14]. Since the existence of a penumbra or rescuable region is ideal for the evaluation of anti-stroke agents, a relatively long time window of 12h for the presence of penumbra in this study would potentially benefit therapeutic strategy for ischemic stroke, which is complementary to the previously documented results at a shorter time window in the same stroke model [9]. In experimental focal ischemia, prompt and effective reperfusion has been reported to salvage the penumbra and therefore to prevent infarct growth [20], [21]. In the present study, half of the rats (4/8) showed a cortex free of infarct on TTC staining at 72h, which was partially attributed to the reperfusion occurring at 6h following ischemia.

After 12h, however, the penumbrarLV decreased further, which can be explained by the relatively long duration of prior ischemia (6h), the limited efficiency of reperfusion, and subsequent reperfusion injury [20], [21].

4.3. Delayed perfusion phenomenon 

The cerebral collateral circulation, which plays a pivotal role in the hemodynamics of cerebral ischemia, remains largely unappreciated, especially for the secondary collateral routes via the leptomeningeal vessel [3]. Since conventional angiography is still the gold standard for direct visualization of collaterals, non-invasive imaging techniques were normally precluded for evaluation of the leptomeningeal and other secondary collateral pathways due to their limited resolution [3]. Recently, however, it is reported that the cortex delayed arterial transit on arterial spin labeling MR perfusion imaging [22] and a late-appearing hypointensity on serial source images of DSC-MRI (the “delayed perfusion” or DP sign) [2] in acute ischemic stroke patients might be markers of collateral leptomeningeal blood flow that may be present in MCA occlusion [17], [23]. Experimentally, this clinically identified DP sign was further confirmed in the present study on source images of DSC-MRI in a rat photothrombotic stroke model. Similar to the clinical report on patients with hyperacute stroke [2], DSC-PWI in this longitudinal rat experiment displayed the DP phenomenon at the cortex region of the left MCA supplied territory exclusively in all rats. Furthermore, we documented the incidence of DP declining gradually within 72h, a message complementary to the first clinical study where only one DSC-PWI was performed at 6h after the onset of stroke [2].

4.4. Potential link between delayed perfusion, penumbra and spontaneous reperfusion 

One of the major findings in this study was the temporal evolution and the presence of the DP sign being parallel and well correlated with those of the penumbra.

The DP sign, an indirect evidence of secondary collateral blood flow, may be indicative of impaired hemodynamic status [3]. According to the observations in this study, the infarct tissue gradually expanded over the entire region of impaired hemodynamics (Fig. 2B), leading to an enlarged volume of Zone2 (Fig. 1) most likely due to the insufficient collateral blood flow and reperfusion. Therefore, the DP region reduced with time as seen with penumbra, despite a spontaneous reperfusion. Conceptually, the penumbra has been defined as ischemic tissue but potentially salvageable if local perfusion is effectively restored [13], [14], [24], whereas DP area is also indicative of impaired hemodynamic status. Therefore, they both may overlap anatomically as proved in this study with MCA occlusion induced stroke.

The hemodynamic effects of the collateral circulation may be critical for maintaining perfusion to penumbral regions and facilitating clearance of fragmented thrombus from more proximal locations [19], [25]. However, the presence of collateral blood flow does not necessarily reduce the extension of infarction in all cases. Besides, spontaneous reperfusion of an occluded MCA is another important factor to maintain certain blood supply to the MCA territory. Therefore, neither factor could prove sufficient by itself [17]. In this study, the early developed leptomeningeal blood flow or DP sign (Fig. 4, Fig. 5, Fig. 6) in conjunction with the spontaneous reperfusion of the occluded MCA observed at 6h may render favorable impacts on the following events. They might improve the rCBF in the ischemic region and especially in the cortex, resulting in a further increase of rCBF over the level of the contralateral area starting at 12h up to 3 days. In this way, they help to maintain the presence of a penumbra up to 12h and allow thrombolytic access to proximal and distal aspects of the thrombi [19]. They also might have prevented larger MCA territory infarctions by sparing only the cortex but not the striatum as seen among the four rats in this study, since deep cerebral collaterals within the subcortical region were less effective, causing an incomplete clearance of thrombus [3], [9].

However, both the striatum and the cortex were infracted in the remaining four rats in this study. This could be partly explained by the fact that they had a relatively longer DP duration of 7.2s and a delayed reperfusion time at 10.5h compared to 6.2s (p<0.05) and 6h (p>0.05), respectively, in the other four rats with striatum involved only, although the difference for delayed reperfusion time was not significant. It seems that only the early development of an effective collateral blood flow and rapid reperfusion of the occluded MCA within the first 8h may prevent infarct expansion [17].

4.5. Study limitations 

Some limitations exist in this study. Firstly, due to the limited resolution of the clinical MR machine for the imaging of small animals, our in vivo methodology did not allow direct visualization of MCA occlusion and collateral vessels in this stroke model. However, the DP sign observed on source images of DSC-PWI offers a non-invasive approach to visually and semi-quantitatively assess collateral circulations with added values for both anatomical and functional evaluation of stroke. Secondly, evidence obtained from serial MRI suggests that the current concept based on PWI–DWI mismatch for ischemic penumbra as adopted in this study appears non-optimal for defining the penumbra [14], [26]. As demonstrated in this study, lesions on DWI were even larger than those on TTC staining at the end of the experiment, suggesting that both the irreversibly infracted tissue (infarct core) and part of the viable penumbral tissue were included in DWI-defined lesion [14]. However, the current PWI–DWI definition derived from MRI provides an approximation of the penumbra, is proved practical and is therefore widely used in clinic.

In summary, there are three major findings in the present study. (1) A spontaneous reperfusion approximately 6h after MCA occlusion has been confirmed for the first time with MR imaging in the rat model of photothrombotic stroke; (2) a penumbra has been identified for up to 12h after ischemic stroke onset in this model; (3) the DP sign as a marker of collateral leptomeningeal blood flow has been found to be parallel and well correlated with the penumbra. This study also suggested that the collateral blood flow shown as a DP sign and the early reperfusion seem to be two critical factors in maintaining perfusion to penumbral regions and preventing infarction growth. Therefore, the prolonged image-based thrombolytic window may benefit the therapeutic strategy for acute stroke [27]. Further clarification of the relationship between collaterals, ischemic penumbra and spontaneous reperfusion may offer a better insight into the mechanisms of acute stroke.