|Home | About | Journals | Submit | Contact Us | Français|
Occlusions of penetrating arterioles, which plunge into cortex and feed capillary beds, cause severe decreases in blood flow and are potential causes of ischemic microlesions. However, surrounding arterioles and capillary beds remain flowing and might provide collateral flow around the occlusion. We used femtosecond laser ablation to trigger clotting in single penetrating arterioles in rat cortex and two-photon microscopy to measure changes in microvessel diameter and red blood cell speed after the clot. We found that after occlusion of a single penetrating arteriole, nearby penetrating and surface arterioles did not dilate, suggesting that alternate blood flow routes are not actively recruited. In contrast, capillaries showed two types of reactions. Capillaries directly downstream from the occluded arteriole dilated after the clot, but other capillaries in the same vicinity did not dilate. This heterogeneity in capillary response suggests that signals for vasodilation are vascular rather than parenchymal in origin. Although both neighboring arterioles and capillaries dilated in response to topically applied acetylcholine after the occlusion, the flow in the territory of the occluded arteriole did not improve. Collateral flow from neighboring penetrating arterioles is neither actively recruited nor effective in improving blood flow after the occlusion of a single penetrating arteriole.
In the healthy brain, the cerebral vasculature is highly regulated and demonstrates an impressive capacity to control the distribution of blood flow with spatial and temporal precision. For example, the brain vasculature can reroute flow toward specific groups of active neurons as small as a single cortical column (Attwell and Iadecola, 2002). This neurovascular coupling has important scientific and clinical implications because it provides the basis for localizing normal and abnormal brain function using methods such as functional magnetic resonance imaging (Ogawa et al, 1990). The neocortex in both rodents and humans has an interconnected network of surface arterioles that consists of many loops (Blinder et al, 2010; Schaffer et al, 2006; Vander Eecken and Adams, 1953). To feed the capillaries within the cortex, penetrating arterioles branch off surface arterioles and plunge perpendicularly into the brain. It has been suggested that this combination of a highly redundant network of surface arterioles and nonredundant penetrating arterioles might serve to facilitate rerouting of flow from one region to another (Nishimura et al, 2007). Such a network might provide a mechanism for flow redistribution not only during normal neurovascular coupling but also in response to vascular occlusions. In experimental models of occlusion of large cerebral arteries, such as the middle cerebral artery (MCA), acute (Belayev et al, 2002; Shih et al, 2009; Tomita et al, 2005; Wei et al, 1998) and chronic (Coyle and Heistad, 1987) dilation of cortical arteries and arterioles has been observed, suggesting that a mechanism for compensatory collateral flow is activated. However, it is not known whether such compensatory routing mechanisms are activated after occlusions of smaller vessels such as penetrating arterioles.
Cortical microinfarcts, 400 to 500μm in size, presumably resulting from occlusion of single cortical arterioles, have been recognized as a prevalent cause of focal neurologic deficits as well as cognitive impairment (Kovari et al, 2004; Vermeer et al, 2003). However, the hemodynamic mechanisms underlying such lesions are poorly understood. Recent developments in optical techniques enable the occlusion of only a single segment of an arteriole or capillary (Nishimura et al, 2006, 2007; Schaffer et al, 2006). These methods, in combination with two-photon excited fluorescence (2PEF) microscopy, provide the tools to study the changes in blood flow and role of active vascular regulation after an occlusion at the level of single microvessels. Experiments with clots in surface arterioles established that there is great redundancy in the surface arteriole network, which provides alternate routes for blood flow around the occluded vessel (Schaffer et al, 2006). In contrast, occlusions in penetrating arterioles, which have little redundancy, result in severe decreases in capillary blood flow in a region ~500μm in diameter (Nishimura et al, 2007), indicating that existing collateral flow is limited. However, it is not known whether the capacity for compensatory flow has been maximized or if additional compensation by vessel dilation might be possible.
In this study, we sought to determine whether the vasculature surrounding an occluded penetrating arteriole is recruited to provide collateral flow to alleviate blood flow decreases. First, we examined whether occlusion of a penetrating arteriole causes active vasodilation in the neighboring penetrating and surface arterioles similar to the dilation observed after occlusion of larger cerebral vessels. Surprisingly, we found no such response. We then examined the effect of the occlusion on the diameter and flow in capillaries around the occluded vessel and found that the only capillaries to dilate were those directly downstream from the occluded vessel; these capillaries also slowed drastically as a result of the clot. Finally, we examined whether the capillary blood flow in the territory of the occluded arteriole could be improved by forcing the neighboring arterioles to dilate, but found that although applying a vasodilator leads to overall increases in flow, it does not resolve the severe blood flow decreases in capillaries downstream from the clot.
All animal procedures were approved by Cornell University Institutional Animal Care and Use Committee. Fifty-nine male Sprague-Dawley rats (Harlan, Inc, South Easton, MA, USA), ranging from 200 to 400g in weight were used in the experiments. Glycopyrrolate (50μg/100g rat) was injected intramuscularly to facilitate respiration. Rats were anesthetized by 5% isoflurane and maintained at 1.5% to 2%. Body temperature was kept constant at 37.5°C, with a heating blanket controlled by rectal thermometer (50 to 7053P; Harvard Apparatus, Holliston, PA, USA). A pulse oximeter (MouseOx; Starr Life Sciences Corp., Oakmont, PA, USA) clipped to the hind paw of the rat was used to monitor blood oxygen saturation and heart rate. A local anesthetic, bupivacaine (0.1mL, 0.125%), was administered at each incision site. The femoral artery was cannulated to monitor blood pressure (BP-1; World Precision Instruments, Sarasota, FL, USA). A tracheotomy was performed to allow intubation for artificial ventilation (SAR-830/P; CWE Inc, Ardmore, PA, USA) and to monitor exhaled carbon dioxide levels (Capstar-100; CWE Inc). Ventilation rates and breath volumes were adjusted to maintain stable end-tidal carbon dioxide. Rats were ventilated with a mix of medical air and the minimum amount of oxygen needed to maintain a minimum arterial blood oxygen saturation of ~95%. Physiological parameters were recorded throughout the experiments to ensure that physiological variables stayed the same during the multiple stages of each imaging session (Supplementary Table S1). Rats were also given 5% (wt/vol) glucose in physiological saline (0.1mL/100g rat) subcutaneously every hour to maintain hydration. An ~3 × 6-mm2 craniotomy was performed over the parietal cortex and the dura was removed. Tubing (inner diameter: 0.25mm, outer diameter: 0.76mm; Microbore, Tygon, Shanghai, China) was glued at the edge of the craniotomy to allow perfusion of artificial cerebrospinal fluid (ACSF) (Kleinfeld and Delaney, 1996) and acetylcholine (ACh). An 8-mm, No. 1.5 glass cover slip (50201; World Precision Instruments) was then glued over the exposed brain using cyanoacrylate and dental cement (Lang Dental Mfg Co, Wheeling, IL, USA and Co-Oral-Ite Dental Mfg Co., Diamond Springs, CA, USA). The lateral edge of the craniotomy was left without glue, but covered in 1.5% agarose (A9793; Sigma, St Louis, MO, USA) in ACSF to allow for drainage of ACSF and the ACh solution while preventing exposure of the cortex to outside air. In most experiments (n=47), rats were gradually transitioned off of isoflurane and onto urethane (1.5g/kg (U2500; Sigma)) administered by intraperitoneal injection. Imaging was started 30minutes after urethane administration. In some experiments (n=12), rats were kept on isoflurane for the entire duration of the experiment. The vasculature was visualized by intravenously injecting 0.3mL of 5% (wt/vol) solution of 2-MDa fluorescein-conjugated dextran (FD2000S; Sigma) in physiological saline.
In vivo vascular measurements were made with a custom-built 2PEF microscope using 800-nm, 87-MHz, 100-femtosecond pulses from a Ti:sapphire laser oscillator (MIRA HP; Coherent, Santa Clara, CA, USA), pumped by a continuous wave laser (Verdi-V18; Coherent, Santa Clara, CA, USA). Laser pulses were prechirped to compensate for dispersion in the microscope with a prism-based compressor (Muller et al, 1998). Laser scanning and data acquisition was controlled by MPScope software (Nguyen et al, 2006). Images spanning the entire cranial window were taken using a 0.28 numerical aperture × 4 air objective (Olympus, Center Valley, PA, USA). For high-resolution imaging, red blood cell (RBC) speed measurements, vessel diameter measurements, and vascular lesioning, we used a 0.95-numerical aperture, × 20, water immersion objective (Olympus).
To map and categorize the arterioles, capillaries, and venules in the area around the target vessel, stacks of images spaced 1μm axially were obtained (Supplementary Text). To measure vessel diameter, we recorded images of individual vessels stepping from above to below the vessel, accumulating at least 20 image frames (~6seconds). For surface and penetrating arterioles, these frames were axially projected, guaranteeing a measurement across the thickest (middle) portion of the vessel. Diameters were calculated by manual selection of a region of interest that included the vessel segment to be measured. The area above threshold (20% of maximum intensity in the projection) was divided by the length of the measured area to yield an average measure of diameter across a 10 to 50-μm segment. For capillaries, because the motion of the brain made averaging frames impractical, diameters were measured from ~5 to 10 individual frames in which the motion appeared to be minimal and then averaged.
To quantify centerline RBC velocity in individual vessels, linescan measurements were obtained by scanning single vessels at a line rate of 1.7kHz for 40seconds and extracting the average speed over this period (Kleinfeld et al, 1998; Schaffer et al, 2006). In these measurements, the scan is aligned along the vessel axis, which produces diagonal dark streaks in the resulting space-time image due to moving RBCs (which exclude the intravenously injected dye), with the slope of the streaks inversely proportional to the RBC velocity. Slope is calculated with an automated image-processing algorithm (Kleinfeld et al, 1998; Schaffer et al, 2006). In arterioles, RBC flow (volume of RBCs/time) was taken to be proportional to the product of the centerline RBC velocity and the cross-sectional vessel area, so we calculated the ratio of flow in a vessel before and after an occlusion as
where v is the centerline RBC speed and R is the vessel radius. The proportionality constant depends on the exact shape of the dependence of RBC speed and hematocrit on radial position in the vessel. We assume this proportionality constant does not change between baseline and later measurements in the same vessel, and thus cancels out in the ratios reported here (Supplementary Text). In penetrating arterioles, RBC velocity and vessel diameter were measured in the portion of the vessel that ran parallel to the cortical surface proximal to where the arteriole dove into the brain. In capillaries, RBCs must travel nearly single file so we report RBC speed and RBC flux (number of RBCs/time) (Supplementary Text). Images and linescans were taken in the same vessels about 1hour before and after penetrating arteriole occlusion. Approximately 10 to 25 vessels were measured in each animal. Following postocclusion measurements, ACh (A9101; Sigma; 10μmol/L in ACSF) (Park et al, 2008) was topically applied to the neocortex in some animals. Vessels were remeasured starting 10minutes after topical application of ACh.
Occlusions of penetrating arterioles were produced by damaging the endothelium of targeted vessels using tightly focused femtosecond laser pulses, leading to localized clotting of the vessel (Nishimura et al, 2006). Vessels were irradiated with 50femtosecond, ~0.1 to 2-μJ pulses from a 1-kHz pulse train produced by a Ti:sapphire regenerative amplifier (Legend 1k USP; Coherent) pumped by a Q-switched laser (Evolution 15; Coherent) and seeded by a Ti:sapphire oscillator (Chinhook Ti:sapphire laser; Kapteyn-Murnane Laboratories Inc, Boulder, CO, USA; pumped by Verdi-V6; Coherent, Inc) (Nishimura et al, 2006). The imaging and ablation beams were combined using a polarizing beam splitter located just after the scan mirrors of the 2PEF microscope and were focused in the same plane. The energy was varied with neutral density filters and the number of pulses deposited on the targeted vessel was controlled by a mechanical shutter with 2milliseconds minimum opening time (VMM-D4; Uniblitz, Rochester, NY, USA). During imaging with 2PEF microscopy, the amplified beam was focused in the lumen of the targeted penetrating arteriole in the segment below the surface but proximal to the first capillary branch off the arteriole. To minimize possible collateral damage, we began irradiation with one pulse using an energy that is below the expected damage threshold (~100nJ at about 100-μm depth). Next, the number of pulses was increased by factors of 10 up to 1,000 pulses with the same energy, often trying each pulse number a few times while watching for signs of vessel damage. The pulse energy was gradually increased by ~25%, and the sequence of increasing pulse number was repeated until some extravasation of fluorescently labeled blood plasma outside the vessel lumen was observed. Once extravasation occurred, multiple nearby areas along the inner wall of the arteriole segment were irradiated with this laser energy and pulse number. Irradiation was continued until RBC motion as visualized in the segment of the penetrating arteriole at the surface of the cortex was stalled. If the occlusion recanalized, irradiation was repeated until the target vessel stopped flowing. Only one occlusion was produced per animal. As an alternative method for producing occlusions, we use photothrombotic clotting (4/59 rats) (Supplementary Text) (Nishimura et al, 2007; Schaffer et al, 2006; Sigler et al, 2008; Watson et al, 1985).
Distributions were nonnormal, so nonparameteric statistical tests were necessary. To compare the effects of different anesthetics and other parameters such as occlusion method or vessel type (Figures 1F–1H, 2A–2C, 2G–2I) two-way analysis of variance (ANOVA) on ranks was used (JMP Statistical Software, SAS Institute Inc, Cary, NC, USA). When this yielded significant differences (Figure 1F) or when multiple comparisons analysis was not necessary (Figures 4A and 4B), we compared pairs of groups with the Wilcoxon–Mann–Whitney ranks sum test. Whenever measurements were made in two instances in the same vessels, the more powerful Wilcoxon rank sign test for paired data was used (Figures 6A and 6C). The significance of trends dependent on topological separation from the targeted arteriole was tested with Cuzick's trend test (Cuzick, 1985) (Stats Direct, Altrincham, Cheshire, UK), which is a variant of the Wilcoxon tests (Figures 2A–2C, ,4A,4A, 4B, 6B, and 6C). A P value of <0.05 was considered statistically significant for all tests. Matlab was used to generate box plots and trend lines as a function of spatial distance (details in Supplementary Text). We used G*Power (Faul et al, 2007) for post hoc power analysis using the Wilcoxon–Mann–Whitney test to compare two groups and the Wilcoxon signed-rank test in cases where we had matched pairs (Supplementary Text). Sensitivity was also calculated with G*Power. In all cases, we used α=0.05. A summary of all raw data is given in the Supplementary Information.
The RBC speed and vessel diameter changes caused by penetrating arteriole occlusions were measured in neighboring penetrating and surface arterioles as well as in nearby capillaries in anesthetized, adult, male, Sprague-Dawley rats. We used 2PEF microscopy to image cortical vasculature labeled with intravenously injected fluorescein dextran through closed cortical windows (Figure 1A). Occlusions were produced by focusing femtosecond-duration laser pulses (0.1 to 2-μJ energy) onto the descending segment of a penetrating arteriole just above the first downstream branch (Figures 1B and 1C). Nonlinear absorption of laser energy injures the vessel wall and triggers clotting (Nishimura et al, 2006). Before and after occlusion of a single penetrating arteriole, lumen diameter, and RBC speed were measured with 2PEF microscopy in neighboring arterioles (Figures 1D and 1E) and capillaries (Figure 3). The targeted and neighboring penetrating arterioles had similar sizes and speeds (Supplementary Figure S1). Separate animals in which no occlusions were induced with only a matched time delay between vessel measurements were used as controls. We found that the active dilatory responses of neighboring arterioles and nearby capillaries after occlusion of a single penetrating arteriole were strikingly different.
We found no evidence for active vasodilation in neighboring arterioles in response to a penetrating arteriole occlusion (Figures 1D and 1F). As vasoactivity can be altered by anesthesia, we performed the measurements using either urethane (19 occlusions and 12 controls) or isoflurane (7 occlusions and 3 controls) anesthesia during vessel measurements and clot formation (Figures 1F–1H). Rather than dilating, neighboring penetrating arterioles constricted slightly after occlusion (Figure 1F; two-way ANOVA on ranks, P=0.041). Post hoc sensitivity analysis showed that we should have been able to resolve a 5% increase in diameter under urethane and a 12% increase under isoflurane with a statistical power of 0.8 (Supplementary Text), suggesting that any vasodilation after a penetrating arteriole occlusion is less than these values. We observed a slight, but not statistically significant, drop in both RBC speed and in RBC flow in neighboring penetrating arterioles after the occlusion, indicating that blood flow to the area surrounding the occlusion was mildly decreased (Figure 1G and 1H; two-way ANOVA on ranks P=0.26 and P=0.16).
To rule out the possibility that nearby or closely connected penetrating arterioles responded differently than distant arterioles, we categorized neighboring penetrating arterioles by the topological separation (i.e., number of vessel branches) and spatial distance from the occluded arteriole. We found no dependence of changes in vessel diameter, RBC speed or RBC flow on topological separation (Figures 2A–2C; Cuzick's trend test, P>0.14), indicating that vessels closely or distantly connected to the occluded arteriole are equally unaffected by the occlusion. Similarly, changes in vessel diameter, RBC speed or RBC flow did not depend on the spatial distance between the vessel and the occluded arteriole (Figures 2D–2F). In addition to no dilation in the penetrating arterioles that dive into the brain tissue to feed capillaries, we found that communicating arterioles that stay on the cortical surface and serve as conduits to more distal regions of brain also did not dilate after a penetrating arteriole occlusion (Figure 2G, two-way ANOVA on ranks, P=0.4). Communicating arterioles decreased slightly in both velocity and RBC flow after the clot, but did not differ significantly from penetrating arterioles (Figures 2H and 2I; two-way ANOVA on ranks, P=0.5 for both). Finally, we tried an alternate clotting method based on photochemical thrombosis with rose bengal, and again found no dilation in neighboring penetrating arterioles (Supplementary Text; Supplementary Figure S2). In all these experiments, urethane and isoflurane anesthetics generated similar results.
A majority of the oxygen and nutrient exchange occurs in the subsurface vessels, so we measured vessel diameter and RBC speed in parenchymal capillaries before and after occlusion of a single penetrating arteriole (Figure 3). We used only urethane for these measurements because the two anesthetics showed similar behavior in arterioles. We found a highly heterogeneous mix of diameter and speed changes in nearby capillaries after a penetrating arteriole occlusion. In the example of Figure 3, the immediate vicinity of the occluded penetrating arteriole included capillaries 10 branches from the target (red in Figure 3B) and capillaries that we were unable to trace back to the target arteriole or were traced back to other arterioles (orange in Figure 3B). One capillary (marked with * in Figures 3D and 3E) was five branches away from both the target vessel and a neighboring penetrating arteriole. Note that the blood flow in this capillary was originally from the target vessel, but after the occlusion (Figure 3C), this vessel reversed direction and supplied flow from the neighboring unoccluded penetrating arteriole. Despite this source of collateral flow, there were large decreases in RBC speed in the majority of the vessels 10 branches away from the occluded arteriole (Figure 3D). Interestingly, nearly all of these vessels dilated (Figure 3E).
Changes in capillary diameter and RBC speed after a penetrating arteriole clot depended strongly on the topological separation between the capillary and the occluded vessel. We categorized each measured capillary by the number of branches that separate it from the trunk of the target vessel (Figure 4A, inset), or the nearest penetrating arteriole for controls. After occlusion of a penetrating arteriole, median RBC speed slowed to 1% (4%) of the baseline speed for capillaries one to two (three to four) branches downstream from the clot (Figure 4A). The amount of slowing decreased with increased topological separation from the clotted vessel (Cuzick's trend test, P<0.0001), with no statistically significant slowing, relative to controls, in vessels more than four branches downstream. Capillaries one to two (three to four) branches downstream from the clot dilated to a median of 113% (110%) of baseline diameter after a penetrating arteriole clot (Figure 4B). The amount of dilation decreased further downstream from the clotted vessel (Cuzick's, P<0.0001), with no statistically significant dilation, relative to controls, more than four branches downstream (Figure 4B). Tube hematocrit remained unchanged after the occlusion for capillaries at all topological separations from the clotted arteriole (Supplementary Text; Supplementary Figure S3).
For capillaries with the same topological separation from the clotted vessel, the dilation and RBC speed decrease did not depend on the distance of the capillary from the occluded vessel. Vessels categorized as closely connected to the occluded penetrating arteriole (four or fewer branches downstream) slowed to a median speed of 2% of baseline. The speed was relatively constant for these closely connected capillaries in a 250-μm radius region around the occluded penetrating arteriole (red in Figure 4C). Distantly connected vessels (more than four branches downstream), however, remained at baseline speeds regardless of their distance from the clotted arteriole. If topological separation is ignored, the median capillary blood flow speed was severely decreased near the occluded vessel and returned to baseline over 300μm (black line in Figure 4C), confirming previous experiments that used a different clotting method based on photochemical interactions to form occlusions (Nishimura et al, 2007). This gradual increase in postclot blood flow speed with distance is consistent with the decrease in the number of closely connected vessels with distance (Supplementary Text; Supplementary Figure S4). Diameter changes also showed little dependence on lateral distance from the occluded vessel, when capillaries are segregated into closely and distantly connected groups (Figure 4D). Uniformly over a 250-μm radius region around the occluded vessel, closely connected vessels dilated to a median diameter of 111% of baseline, whereas distantly connected vessels did not dilate.
We asked whether the amount of dilation after a penetrating arteriole occlusion depended on the amount of reduction in RBC speed. We plotted the postclot diameter versus postclot speed, each as a percentage of baseline, for all penetrating arterioles (Figure 5A) and capillaries (Figure 5B), including both clot and control measurements. The penetrating arteriole measurements, both clot and control, grouped into a single distribution that showed no change, on average, in either speed or diameter (Figure 5A). In the capillaries, categorization into two groups by k-means clustering (Matlab) revealed one cluster that slowed drastically and dilated (red in Figure 5B). In the second cluster (blue in Figure 5B), average speed and diameter did not change, similar to the behavior of the arterioles. This second cluster contained 91% of the control capillaries. The capillaries from clot experiments in these two clusters differed in their topological separation from the occluded penetrating arteriole (Kolmogorov–Smirnov test, P<0.0001), with the slowed and dilated group including more capillaries that were fewer branches from the occlusion (Figure 5B, inset). For example, 88% of capillaries four or fewer branches downstream from an occluded penetrating arteriole were in the slowed and dilated cluster. The difference in the cumulative probability of being in the slowed and dilated cluster vs. the unchanged cluster was largest at four branches downstream from the occlusion (Figure 5B, inset), indicating that vessels one to four branches downstream responded, on average, differently from those more than four branches downstream.
The slowed, dilated capillaries appeared to be spatially intermingled with unaffected capillaries. The clustering analysis suggested that closely connected capillaries (four or fewer branches downstream from an occluded penetrating arteriole) tended to belong to the slowed and dilated group, whereas distantly connected capillaries (more than four branches downstream) belonged to the unchanged group (Figure 5B). Therefore, we analyzed the three-dimensional distribution of closely and distantly connected capillaries in three animals by categorizing every capillary within an image stack by its connectivity to the occluded arteriole and by the three-dimensional position of the midpoint of each capillary. In the example of Figures 5C and 5D, the flow speed and diameter changes in 19 capillaries were measured after the target penetrating arteriole was occluded, and in all but three cases, the topological classification (closely or distantly connected) correctly predicted the measured physiological response (either slowed and dilated or unchanged). This consistency supports the observed relationship between the topological classification and the vessel's response (Figure 5B, inset). In this example, many distantly connected capillaries were adjacent to closely connected vessels (Figure 5D). Across three animals, we find that at a distance of ~120μm from the targeted penetrating arteriole, about 16% (95% confidence interval, 10% to 24%, 1,101 vessels) of distantly connected capillaries had a capillary that was closely connected to the target penetrating arteriole as their nearest spatial neighbor (Supplementary Text; Supplementary Figure S4). The average distance between a closely connected capillary and the nearest distantly connected capillary was 46±12μm (mean±s.d.; Supplementary Figure S5A). For the subset of distantly connected capillaries that had a closely connected vessel as a nearest neighbor, the separation between the two vessels was only 35±15μm (Supplementary Text; Supplementary Figure S5B).
After the measurements during occlusion, ACh (10μmol/L in ACSF) was perfused into the cranial window and each vessel was remeasured in a subset of animals. Penetrating arterioles neighboring the occlusion dilated with ACh relative to the postclot diameter (Figure 6A; Wilcoxon paired rank sign test, P=0.0002). The ACh also caused the RBC flow in these vessels to increase relative to the RBC flow during the occlusion (Figure 6A; Wilcoxon paired rank sign test, P<0.0001). Aggregated across all measurements capillaries dilated relative to clot diameters with ACh (Figure 6A; Wilcoxon paired rank sign test, P=0.00025). Capillaries with different topological separation from the occluded penetrating arteriole dilated similarly in response to ACh application (Figure 6B; Cuzick's tend test, P=0.85).
With ACh application after occlusion, the median RBC speeds one to two (three to four) branches downstream from the occluded penetrating arteriole remained at <1% (13%) of baseline (Figure 6C). Ten or more branches away, the ACh causes an increase in median RBC speed to 105% of baseline values. Despite the dilation of both neighboring penetrating arterioles and capillaries, RBC speed in the downstream capillaries most affected by a penetrating arteriole occlusion was not substantially improved by ACh, whereas distantly connected vessels became slightly hyperemic. Similarly, RBC flux, the number of RBCs flowing through the capillary segment per unit time, decreased severely in the capillaries closest to the occlusion, and was not improved with ACh application (Supplementary Text; Supplementary Figure S6).
Femtosecond laser ablation can trigger clotting in penetrating arterioles while preserving structure and flow in neighboring vessels, enabling the study of physiological processes, such as vasoreactivity, in the nearby microvessels. This laser ablation technique is a variation on the method previously developed to clot single cortical capillaries (Nishimura et al, 2006). As the interaction of the laser light and the tissue depends on nonlinear absorption, only the targeted portion of the vessel is injured by the laser and the subsequent clotting and occlusion are limited to a single vascular segment. Previous work on occlusions of penetrating arterioles used photochemical thrombosis with rose bengal (Nishimura et al, 2007). As the rose bengal method uses linear optical excitation to produce the singlet oxygen that damages the vessel wall and initiates clotting, it is not possible to confine the effect to a single vessel segment and some level of clotting in surrounding capillaries is very difficult to avoid. To rule out the possible effects of our clotting technique on vasoreactivity, we tested both the femtosecond laser ablation and photochemical methods for clotting penetrating arterioles and saw no dilation in neighboring penetrating arterioles for both methods (Supplementary Figure S2). In addition, previous measurements of flow changes in capillaries after a penetrating arteriole occlusion produced using the rose bengal method (Nishimura et al, 2007) are similar to those reported here (Figures 4A and 4C).
We used 2PEF microscopy to investigate whether active vascular regulation in surrounding penetrating arterioles mitigates the decreased flow in capillaries downstream from an occluded cortical penetrating arteriole. We first investigated whether the neighboring penetrating arterioles that feed territories adjacent to the occluded vessel dilate in response to the occlusion (Figure 1). Contrary to the expectation that the occlusion could trigger a compensatory vasoactive response in the surrounding area, the neighboring penetrating arterioles did not dilate (Figure 1F). Even the closest arterioles, both in topological and spatial distance, did not dilate after occlusion (Figure 2). Our methods had sufficient statistical power (Supplementary Text) to resolve changes in arteriole diameter and RBC speed of a magnitude comparable to or smaller than those observed in neurovascular coupling. Taken together, this data suggest that it is unlikely there is any physiologically relevant compensatory vasoregulatory mechanism that is recruited to increase flow in neighboring penetrating arterioles to increase perfusion in the tissue adjacent to the occluded arteriole.
This absence of dilation is in marked contrast to the coordinated dilation of local and upstream arterioles that occurs in response to increases in neural activity (Devor et al, 2008; Iadecola et al, 1997). Upstream dilation of arterioles in neurovascular coupling (and other systems such as muscle) is likely to be mediated by signal conduction in the endothelium (Dietrich et al, 1996; Welsh and Segal, 1998) or astrocytes (Xu et al, 2008). Our observation of dilation in the neighboring penetrating arterioles after ACh application (Figure 6A), an endothelium-dependent dilator (Faraci and Heistad, 1998; Rosenblum and Nelson, 1988), indicates that both endothelium and smooth muscle were functional despite experimental manipulations. This suggests that the lack of dilation after penetrating arteriole occlusion comes from a lack of signaling. Although we used two different mechanisms for clotting that show similar results, we cannot eliminate the possibility that the lack of dilation of neighboring arterioles after the clot is an artifact of our technique. Injured endothelial cells or astrocytes at the clot location might not be capable of conducting the vasodilatory signals that might otherwise be present (Emerson and Segal, 2000; Xu et al, 2008). However, because these cells are quite susceptible to ischemia or hypoxia (Fisher, 2008), it is likely that a naturally forming thrombus or an embolus would also result in a disruption of endothelial or astrocyte function that would interfere with any possible vasodilatory signal (Emerson and Segal, 2000).
Although no dilation was observed in arterioles, a subset of capillaries do dilate after the occlusion (Figures 4B and 4D). Cluster analysis of downstream capillaries suggests that there is a trend for lower flow to be associated with a larger dilation (Figure 5B). No such trend is observed in the neighboring penetrating arterioles (Figure 5A), but this may be explained by the fact that after a single penetrating arteriole occlusion, flow in other arterioles does not drop to the extremely low levels observed in the downstream capillaries. After larger strokes, such as MCA occlusions, other investigators do observe acute vasodilation in the surface and penetrating arterioles that sit downstream from the occlusion (Belayev et al, 2002; Shih et al, 2009; Tomita et al, 2005; Wei et al, 1998), consistent with the idea that drastically reduced flow or intraluminal pressure is necessary to trigger dilation. In the work by Shih et al (2009), mean velocities in penetrating arterioles dropped to ~30% of baseline and vessel diameter increased to ~120% of baseline after MCA occlusion. Wei et al (1998) made ministrokes by ligating several surface arterioles and noted a marked dilation and also described considerable slowing in nearby surface arterioles. Taken together, these data suggest that the vasodilation signal does not activate in a vessel until hemodynamics have substantially changed in that vessel.
The spatial distribution of dilated and undilated capillaries rules out a parenchymal source for the vasodilation signal. There are frequent occurrences in which an extremely slowed and dilated capillary is adjacent to a rapidly flowing and undilated capillary. This is reflected in the aggregate data as a strong dependence of the speed and diameter changes on the topological separation (Figures 4A and 4B), but not the spatial distance (Figures 4C and 4D) between a capillary and the occluded penetrating arteriole. Analysis of the spatial distribution of vessels with different topological separation (Supplementary Figures S4 and S5) shows that capillaries only a few branches downstream from the target, which slow and dilate after the occlusion, are spatially interspersed with more distantly connected capillaries, which are largely unaffected by the clot (Figures 5C and 5D). The close proximity of dilated and unaffected capillaries suggests that there is no signal for vasodilation that originates from the neurons or astrocytes or other parenchymal cells. If such a diffusible signal from the parenchymal tissue existed, one would expect that capillaries spatially near but many branches away from the occluded arteriole would dilate along with nearby, but more closely connected vessels.
Past work using a pimonadozole probe that precipitates in severe hypoxia suggests that tissue within 150μm of a penetrating arteriole occlusion is relatively uniformly hypoxic (Nishimura et al, 2007). Our observations of undilated vessels in this region suggest vessels that maintain high levels of flow do not dilate even if they run through regions of likely hypoxia near the occluded vessel. This argues that tissue hypoxia is not sufficient to drive vasodilation. Current ideas that neurovascular coupling in the normal brain is linked to synaptic release and byproducts of neuronal activity rather than to metabolism or oxygen usage are consistent with our findings (Attwell and Iadecola, 2002; Devor et al, 2008; Lindauer et al, 2009; Sukhotinsky et al, 2010). For example, recent work on neurovascular coupling shows blood flow still increases in regions of neuronal activity even under hyperbaric hyperoxygenation, which demonstrates another case in which dexoygenation is decoupled from vasodilation (Lindauer et al, 2009). We find that the alterations in oxygenation and energy delivery caused by blood flow decreases due to a penetrating arteriole occlusion do not generate upstream or diffusible vasodilatory signals, even though the oxygenation and metabolite deficits due to occlusions are more severe than those caused by normal neuronal activity.
The dilation in downstream capillaries after penetrating arteriole occlusion (Figure 4B) suggests that there is a perfusion-dependent signal for dilation. Possible candidates for the vasodilatory signal include flow-dependent forces such as changes in shear stress or transmural pressure. Arterioles do respond to changes in shear stress, but arterioles tend to dilate in response to an increase in shear stress (Ngai and Winn, 1995). This response is the opposite of what is observed in the capillaries in our experiments and in surface and penetrating arterioles after MCA occlusion (Belayev et al, 2002; Shih et al, 2009; Tomita et al, 2005; Wei et al, 1998) in which vessels dilate in response to decreased flow. Hematocrit in the capillaries did not change significantly after occlusion (Supplementary Text; Supplementary Figure S3), but the RBC flux did drop in proportion to the speed change, suggesting that dilated capillaries may respond to some change in the rate of RBCs that pass through (Supplementary Text; Supplementary Figure S6). A more likely candidate is change in transmural pressure. Brain arterioles constrict with increased pressure, suggesting that decreased transmural pressure could trigger dilation (Ngai and Winn, 1995; Schmid-Schonbein, 1999). The vessels immediately downstream from a penetrating arteriole are likely at reduced luminal pressure after the occlusion relative to before the occlusion because these vessels are now effectively farther downstream from a high-pressure arteriole. Similarly, surface and penetrating arterioles would be at reduced luminal pressure after MCA occlusion, suggesting that a myogenic mechanism for vasodilation could be also responsible for the acute dilation observed after large stroke models (Belayev et al, 2002; Shih et al, 2009; Tomita et al, 2005; Wei et al, 1998). Thus, our work suggests that acute postclot vascular reactivity could use the same mechanisms involved in autoregulation, which keeps brain perfusion constant under changing systemic blood pressure.
We applied a vasodilator, ACh, to investigate whether increasing flow in unoccluded arterioles can substantially improve flow in capillaries downstream from an occluded penetrating arteriole (Figure 6). Despite dilation in both neighboring penetrating arterioles and capillaries, ACh did not substantially improve perfusion in the territory of decreased flow caused by the occlusion. The ACh leads to only a very small increase in the magnitude of RBC speed (Figure 6C). The first few branches downstream from the occlusion are most severely affected by the clot and are also the least improved by vasodilator application. This result suggests that although connections between penetrating arteriole territories through the capillary bed exist (Moody et al, 1990), they are too infrequent to provide flow between territories. Clinical trials for vasodilators such as xanthine derivatives for treatment of vascular dementia (Kittner et al, 1997; Pantoni, 2004) have not shown significant efficacy and have also been disappointing for the treatment of stroke (Bath and Bath-Hextall, 2004; Bereczki and Fekete, 2008). If vascular dementia is driven by the effects of many small vessel occlusions, our finding that ACh does not improve flow downstream from the occlusion might be a contributing factor to the drug trials' somewhat disappointing results. However, we cannot rule out that other patterns of dilation driven by mechanisms other than ACh might be more effectual or that vasodilators might have significant impact in the case of partial occlusions. Alternative strategies for increasing flow include reducing blood viscosity (Nishimura et al, 2006) or leukocyte adhesion (Belayev et al, 2002), each of which have been found to increase flow downstream from local occlusions. However, it is not clear that these flow increases are sufficient to be therapeutic.
The authors declare no conflict of interest.
Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)