Penetrating arterioles are a bottleneck
in the perfusion of neocortex
Nozomi Nishimura*†‡, Chris B. Schaffer*†‡, Beth Friedman§, Patrick D. Lyden§¶, and David Kleinfeld*†¶?
Departments of *Physics and§Neurosciences,¶Graduate Program in Neurosciences, and†Center for Theoretical Biological Physics, University of California
at San Diego, La Jolla, CA 92093
Communicated by Harry Suhl, University of California at San Diego, La Jolla, CA, November 5, 2006 (received for review June 8, 2006)
Penetrating arterioles bridge the mesh of communicating arte-
rioles on the surface of cortex with the subsurface microvascular
bed that feeds the underlying neural tissue. We tested the con-
jecture that penetrating arterioles, which are positioned to regu-
late the delivery of blood, are loci of severe ischemia in the event
of occlusion. Focal photothrombosis was used to occlude single
penetrating arterioles in rat parietal cortex, and the resultant
laser-scanning microscopy in individual subsurface microvessels
that surround the occlusion. We observed that the average flow of
red blood cells nearly stalls adjacent to the occlusion and remains
downstream from the occlusion. Preservation of average flow
emerges 350 ?m away; this length scale is consistent with the
spatial distribution of penetrating arterioles. We conclude that
penetrating arterioles are a bottleneck in the supply of blood to
neocortex, at least to superficial layers.
blood flow ? hemodynamics ? ischemia ? stroke ? two photon
blood. The large cerebral arteries supply blood to the cortex
grid may be considered as a robust compliant source of blood (1,
2). In contrast to the 2D architecture of this surface network, the
subsurface microvascular bed consists of a tortuous plexus of
small vessels that course in three dimensions, in which much of
the exchange of gas, metabolites, and heat occurs. The surface
and subsurface networks are bridged by penetrating arterioles,
which branch from the surface arterioles and dive radially into
the brain tissue (3). However, it is not known whether the
penetrating arterioles associate with spatially segregated terri-
tories of microvasculature, as opposed to territories that are
interdigitated (4, 5).
Past work has shown that the location of a clot within the
anatomical subregions of the microvascular network strongly
determines the extent of the resulting ischemia. For example, a
clot in a surface arteriole, which constitutes one branch of the
surface network with its extensive anastomoses, leads to a
relatively weak reduction in flow in downstream surface vessels
(2). A clot in deep-lying microvessels, where the vasculature is
relatively less interconnected (6, 7), results in a significant
decrement in blood flow (8). In contrast to the substantial
interconnections within the surface and subsurface networks,
the penetrating arterioles that bridge these networks appear to
be largely devoid of anastomoses (4, 9). This architecture casts
of blood flow to columnar-sized regions of cortex (10–12). Yet
this very same feature is likely to cause pronounced ischemia if
the penetrating arterioles are occluded. In support of this
hypothesis, the microstrokes observed in human patients are
often centered around penetrating arteries and arterioles with
obstructed lumens (13, 14). Further, the occlusion of penetrating
arterioles in animal models by the intraarterial injection of
he mammalian brain has developed specialized vascular
architecture to maintain and regulate the distribution of
occluding particles, such as microemboli or microbeads, can
result in infarcts (15–18).
In this work, we examine the hypothesis that penetrating
ask: (i) What is the magnitude and extent, both in terms of
topology and space, of the reduction of blood flow after the
occlusion of a single penetrating arteriole? (ii) Is there a
relationship between the size of the territory with flow reduc-
tions and the spacing between an occluded arteriole and neigh-
boring patent penetrating arterioles?
Large-scale maps of the brain vasculature of rat in the field of
a craniotomy, obtained with in vivo two-photon laser-scanning
microscopy (TPLSM) (19) and labeling of the blood plasma with
fluorescein/dextran, reveal branches of the surface communi-
cating arteriole network as well as penetrating arterioles. Can-
didate vessels for photothrombotic clotting are identified from
the maps and traced back to a readily identifiable artery or vein
on the pial surface; we targeted only arterioles that had branches
in the upper 400 ?m of cortex (Fig. 1A). High-resolution planar
images were used to measure the diameter of the penetrating
arteriole and to map vessels that lie up- and downstream from
the target. Line-scans along the axis of a vessel were used to
measure red blood cell (RBC) velocity (2, 20–24). After baseline
velocities were determined, the photosensitizer rose bengal was
injected, and green laser light was focused at high-numerical
aperture into the target penetrating arteriole to form a clot (2)
Our photothrombotic clots occluded the most proximal por-
in the section from the last surface branch to at least the first
vessel that branches laterally below the pial surface. Only one
penetrating arteriole per animal was occluded. Before occlusion,
per s (mean ? SD; n ? 16), and the average diameter was 10.5 ?
3.4 ?m. Blood flow in the surrounded vasculature was quantified
before and after occlusion of a penetrating arteriole.
Topological Dependence of Impacted Flow.Wecategorizedtheflow
in microvessels by their connectivity to the target penetrating
arteriole (Fig. 2B Inset). The measured vessels included those
that lie both upstream and parallel to the target vessel, and
Author contributions: N.N., P.D.L., and D.K. designed research; N.N., C.B.S., and B.F.
performed research; N.N., P.D.L., and D.K. contributed new reagents/analytic tools; N.N.,
B.F., and D.K. analyzed data; and N.N., B.F., and D.K. wrote the paper.
The authors declare no conflict of interest.
Abbreviations: TPLSM, two-photon laser-scanning microscopy; D, downstream; U, up-
stream; P, parallel.
Ithaca, NY 14853.
?To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2006 by The National Academy of Sciences of the USA
www.pnas.org?cgi?doi?10.1073?pnas.0609551104 PNAS ?
January 2, 2007 ?
vol. 104 ?
no. 1 ?
that branched laterally off the penetrating arteriole were de-
noted as downstream (D) and were categorized by the number
of branches between the measured vessel and the penetrating
vessel. Thus, D1 vessels were directly connected to the pene-
trating arteriole, whereas D2 vessels were two branch points
away, etc. In a few cases, the penetrating vessel did not maintain
a trunk that stayed larger than capillary diameter, i.e., ?5 ?m,
and in these cases, the penetrating vessel was categorized as a D1
vessel after its size diminished to capillary diameter. Vessel
segments from surface arterioles that fed the penetrating vessel
were categorized as upstream (U), and arterioles that shared the
same source as the penetrating vessels were categorized as
parallel vessels (P).
An example data set (Fig. 2A) demonstrates both increases
and decreases in blood speed after the occlusion, as well as
reversal in the direction of flow (8), and is suggestive of a high
variability in the observed changes in the speed of RBCs. As
indicated by the values in each branch order (105 vessels total
across 16 animals), the speed of RBCs in vessels downstream
from the occluded penetrating vessel was severely reduced and
did not recover to baseline values even as far as 10 branches
distal to the target vessel (Fig. 2 B and C). In the first branch
downstream from the penetrating arteriole (D1), RBCs are
almost stalled as median speeds dropped to 0.03-times their
baseline values after penetrating arteriole occlusion (D1; Fig.
2B). Branches further downstream showed smaller decreases in
speed, but median speeds only approached 0.1-times their
baseline speed even as far as 10 branches downstream (D3–D10;
Fig. 2B). If we consider the mean values of speed, which
highlights trends in the data, we see that the speed tends toward
its baseline level from a value of 0.03 ? 0.01 (mean ? SEM)
?times baseline for D1 vessels to 0.33 ? 0.13 ?times baseline for
D6–D10 vessels (Fig. 2C).
Spatial Dependence of Flow Change. The decrement in RBC flow
in response to occlusion of a penetrating arteriole extended far
downstream, so much so that our ability to maintain accurate
count of the branch order was limited by the penetration depth
of TPLSM imaging. We thus analyzed changes in RBC speed in
deep microvessels with respect to the lateral distance from the
occluded arteriole and without respect to connectivity. As in the
analysis according to branch order (Fig. 2), we found a broad
range of changes in blood flow after occlusion of the penetrating
include both increases and decreases in RBC speed in vessels
that lie hundreds of micrometers away from the target vessel. A
compendium of all data, pooled across the 50- to 400-?m depth
of the measurements, shows a trend in the mean RBC speed
relative to baseline that increases from near zero at the target
penetrating arteriole and returns to baseline by 350 ?m away
(Fig. 3B; the branch order of the measured vessels is color-
coded), with half of the baseline speed obtained by 165 ? 25 ?m
(red line is running mean of 50 data points ? SEM). Last, the
pia (data not shown) within the range of our measurements.
The mean RBC speed remained at baseline values for dis-
tances ?350–900 ?m from the target arteriole, yet these postclot
speeds appeared to show increased variability (Fig. 3B). This
spread in values is most apparent from a graph of the postclot
velocity, denoted vpostclot, vs. the baseline velocity, denoted
vbaseline (Fig. 3C). We quantified the spread about a linear
dependence between vpostclotand vbaselinein terms of the ratio
(vpostclot ? vbaseline)/vbaseline (Fig. 3D), which has a root-mean-
Schematic of experimental set-up. Green laser light is focused into a pene-
trating arteriole by a microscope objective concurrent with TPLSM of the
directions of image stacks through a penetrating vessel clot; schematics
indicate the directions and volumes of projection. The clot is visualized as a
dark mass in the target vessel that is often surrounded by bright areas of
stalled, labeled plasma. The images are inverted for clarity. B Inset shows
tangential view before irradiation.
Induction of clots in penetrating arterioles by photothrombosis. (A)
image stacks in the vicinity of the occluded arteriole and schematics of RBC
velocity before (Upper) and after (Lower) clot generation in a penetrating
arteriole. The target vessel is marked by a yellow X and spirals into the
parenchyma. Negative numbers indicate flow direction has reversed in direc-
tion after clot relative to baseline. (B) The fraction of baseline speed after the
clot across all measurements. Each box contains the middle two quartiles of
range of data, and dots indicate outliers that are outside the range of the
middle two quartiles ? 1.5 times the interquartile distance. Downstream
branches are grouped by number of branches from target penetrating vessel
(D1–D10), as indicated in schematic on the left. Upstream (U), parallel (P), and
vessels that had no connection to the target vessel visible in TPLSM stacks are
grouped separately. (C) Fractional change from baseline speed vs. connectiv-
the trend in the data.
Blood-flow changes categorized by topology. (A) Projections of
www.pnas.org?cgi?doi?10.1073?pnas.0609551104Nishimura et al.
square (rms) width of ? ? 0.45. To establish that the spread was
dominated by variability in the value of RBC velocities that is
induced by the clot, as opposed to intrinsic variability in RBC
speed (21) and/or biases in our measurement process, we per-
formed two control experiments. In the first set, we sought to
estimate the inherent variability in RBC flow through microves-
sels. We used animals with no clots in penetrating arterioles and
measured the baseline speed of RBCs at two different times,
30–60 min apart, in deep microvessels (Fig. 3E). The data set
consisted of vessels of similar depth and caliber as those mea-
sured in experiments with occluded vessels (Fig. 3 A and B), and
the spread in measurements was quantified in terms of the
analogous ratio (vbaseline2 ? vbaseline1)/vbaseline1 (Fig. 3F). The
distribution for this control data had a rms width of ? ? 0.30, less
than that for the postclot data. In a second set of control
experiments, we sought to establish whether the flow in vessels
after an occlusion evolves over the timescale of our measure-
ments. A penetrating arteriole was blocked and RBC velocities
were measured at two different times, again 30–60 min apart,
after the formation of the clot (Fig. 3G). The mean speeds for
the two sampling times were statistically indistinguishable, i.e.,
vpostclot1? 0.85 ? 0.28 mm/s (mean ? SEM) vs. vpostclot2? 0.90 ?
0.35 mm/s, whereas the spread in value between the two mea-
sures, quantified in terms of the ratio (vpostclot2 ? vpostclot1)/
vbaseline(Fig. 3H), also had a rms width of ? ? 0.30. There is no
statistical difference between the distributions of velocities for
the two control groups (P ? 0.75 as compared with the Fisher–
Snedecor f distribution). Critically, the normalized distribution
of the change in velocities among distant vessels after vs. before
a clot was statistically broader than the distributions for the
the variability in blood flow through microvessels is increased
after occlusion of even a distant penetrating arteriole.
Validation of a Localized Occlusion. Several aspects of the blood
flow measurements made within ?200 ?m of the targeted
penetrating arteriole indicated that the observed ischemia does
not reflect nonlocalized effects of photothrombosis. First, we
monitored blood flow in upstream and parallel vessels on the
surface. Flow in upstream vessels, at one branch point away from
occluded vessel, showed a decrease to about half the baseline
distal to the upstream source (Fig. 2 B and C). Flow in parallel
vessels showed almost no change on average (Fig. 2 B and C).
These results for the surface vessels suggest that the effects of
photothrombosis are localized only to the targeted penetrating
arteriole. Second, we measured RBC flow in deep vessels that
were present in the same field of view as the vessels downstream
of the target but had no obvious connection to the occluded
penetrating arteriole. In the case illustrated in Fig. 2A, examples
of unconnected vessels, indicated in blue, were connected to a
radially draining venule and showed little effect from the laser.
On average, blood flow in nonconnected vessels within 200 ?m
of the target vessel slowed to only 0.88 ? 0.15 (mean ? SEM)
of the initial RBC speed after penetrating vessel occlusion,
whereas flow in vessels within the same spatial range that were
observed to be connected to the target slowed to 0.16 ? 0.03 of
the initial speed (Fig. 2 B and C). In vivo image stacks also
showed that the majority of vessels in the vicinity of the target
vessels appeared normal. In the case illustrated by Fig. 4 A and
B, we consider two microvessels that lie close to the clot. Flow
in one vessel is essentially stopped after the clot (vessel 2),
whereas flow in the other continues largely unchanged (vessel 1).
Two deeper vessels are observed to have substantially reduced
flow (vessels 3 and 4); additional details are in supporting
information (SI) Fig. 8. In general, RBC flow in vessels in the
vicinity of the occluded arteriole indicates that the majority of
structurally affected by the laser light.
Histological effects of a penetrating arteriole occlusion were
assessed in qualitative assays of brain pathology. We identified
the territory in which the clotted vessel resided by the fluores-
cence of groups of vessels that retained fluorescein/dextran dye.
This effect is likely to be a consequence of laser irradiation but
is independent of measured blood flow changes (2) (Fig. 5A2?).
Maximal projection of vessels across the cortex, with the fraction of baseline
RBC speed after photothrombotic clot indicated. The clotted penetrating
arteriole is indicated by the red circle. (B) Fraction of baseline RBC speed after
the occlusion vs. distance from the occluded arteriole. The connectivity of the
measured vessels is color-coded, as indicated. The thick red line through the
data is the smoothed response averaged over a window that included 50
points, with ?1 SEM limits indicated by the thin red lines, and the blue line is
the prediction of the change in RBC speed from a model with an exponential
spatial distribution for the flow away from a blocked penetrating arteriole
from the measured vessels. Two outliers, with relative speeds of 2.6 and ?2.8,
were excluded from analysis in D. (E and F) The inherent variability in two
measurements separated by 30–60 min, with no clot present. (G and H) The
inherent variability in two measurements separated by 30–60 min, with both
and second measurements after induction of the clot are shown in blue and
red, respectively, with black lines linking measurements on the same vessel.
Insets show the full range of the data.
Nishimura et al.PNAS ?
January 2, 2007 ?
vol. 104 ?
no. 1 ?
These fluorescent regions were found to be selectively marked
by tissue hypoxia, as inferred from immunocytochemical iden-
tification of cellular uptake of pimonidazole hydrochloride
(HypoxyProbe, Chemicon, Temecula, CA) (Fig. 5 A2 and A2
Inset) and from the absence of similar concentrations at levels
0.5 mm both medial and lateral (Fig. 5 A1 and A3). The clot
territory was also marked by vascular immunoreactivity to IgG,
a sign of abnormal vessel leakiness (25) (Fig. 5 B2 and B1?–B3?).
Immunostaining for IgG also extended more weakly to tissue
(Fig. 5 B1 and B3). This spatially extensive vascular immunore-
activity for IgG is consistent with the observation of increased
variability in blood flow, in which some vessels show substantial
decreases after clot formation in a distant penetrating arteriole
(Fig. 3 C and D).
Spatial Distribution of Penetrating Arterioles. We measured the
distribution of penetrating arterioles to analyze how their sta-
tistics might affect the area impacted by occlusion of a single
penetrating arteriole. The locations of all arterioles were iden-
tified from large-area images of the cortical vasculature (Fig.
6A). The mean distance between nearest-neighbor pairs of
penetrating arterioles was 130 ? 60 ?m (mean ? SD) with an
irregular distribution (n ? 125 arterioles across four animals). A
measure relevant to the flow of blood is the radial distance from
every pixel across the cortical surface to the nearest arteriole,
denoted rbaseline (see SI Text and SI Fig. 9). We numerically
calculated the distribution of rbaseline(n ? 71 after arterioles at
the edge of the imaging field are excluded), and observed a mean
distance of ?rbaseline? ? 145 ?m (blue histogram; Fig. 6B). The
projection in the coronal (x-z) plane of image stacks around a penetrating
and spatial profile was measured before and after the occlusion. High-
magnification images and line-scan data for indicated vessels are shown before
as dark streaks on a bright background; the sign and magnitude of the slope of
the streaks reflects the direction and speed, respectively, of RBC motion.
Blood flow in vessels in the immediate neighborhood of the clot. Side
section series taken across a 2.2-mm span centered on the fluorescent vessels
that demarcate the core territory of the clotted arteriole (A2?, inverted
in Fig. 3B. (A) Sections stained to localize tissue hypoxia demonstrate a
confined high-density of staining at the level of the center of this series. The
location of the clot is confirmed by trapped fluorescein/dextran (A2?). (B)
Sections stained for immunoreactivity to IgG illustrate that vessel leakiness
(B1?, B2?, and B3?) is concentrated at the center of this series (B2) yet extends
up to 1.1 mm in medial and lateral directions.
Tissue reactions in region of occluded penetrating arteriole. Sagittal
vasculature with penetrating arterioles marked with red dots; largest for the
nearest penetrating arteriole (rbaseline, blue) and the second-nearest arteriole
(rbaseline, red). (C) The difference in radial distance from the first- and second-
nearest penetrating arteriole, ?r, as a function of distance from the nearest
arteriole, rbaseline(red curve). The calculation was performed for penetrating
arteriole locations that were randomly assigned while maintaining a mean
density of arterioles equal to the measured value (black curve) and for
arterioles arranged in regular patterns, i.e., hexagonal (blue curve), square
(gray curve), and triangular (green curve) lattices; see SI Text. (D–F) Model of
blood flow from penetrating arterioles, in which the contribution of flow in
the tissue was described by an exponential decay, i.e., e?r/??, where r is the
radial distance from the arteriole, and ?? ? 40 ?m. The calculated sum of
the flow from all penetrating arterioles, located as in A, is shown in D. The
of the arteriole marked in red in A, is shown in E. Black lines in D and E
demarcate the pixels that are equidistant from more than one arteriole. The
calculated ratio of blood flow that remains after occlusion of a penetrating
arteriole (A) is shown in F.
Spatial analysis of vascular territories. (A) Image of the surface
www.pnas.org?cgi?doi?10.1073?pnas.0609551104Nishimura et al.
effect of an occlusion of a penetrating arteriole was estimated by
deleting an arteriole and calculating the radial distance from
pixels within the territory of the deleted vessel to the nearest
remaining arteriole, denoted rpostclot. The distribution of rpostclot
was calculated across all arterioles, and the mean value was
?rpostclot? ? 205 ?m (red histogram; Fig. 6B). Finally, the effect
of the clot on flow at a given location across cortex may be
related to the increased distance to unclotted arterioles relative
to the distance to the original but now clotted arteriole. We
denote this change in distance by ?r ' rpostclot? rbaseline. Near
the clotted arteriole, ?r ? 130 ?m, which equals the mean
distance between nearest neighboring arterioles, as expected.
The value of ?r decreases to an asymptotic value ?r 3 0 far from
the vessel, where the average distance to the clotted penetrating
arteriole equals to the distance to an unperturbed arteriole (red
trace; Fig. 6C). The distance to achieve half of this decrement is
135 ?m, which is also close to the value of the mean spacing
the spatial arrangement of the penetrating arterioles, we numeri-
cally calculated the form of ?r for different spatial arrangement of
arterioles. We used a distribution of randomly located penetrating
vessels that maintained the same average density of penetrating
arterioles as the experimental measurement, i.e., 13 ? 3 vessels/
mm2(mean ? SEM; four animals) and found that the dependence
and red traces in Fig. 6C). For comparison, we calculated ?r for
simulated distributions of arterioles arranged in triangular, square
and hexagon lattices (SI Text and SI Fig. 10). The edge length of
each lattice was set equal to the measured mean distance between
pairs of nearest neighbor arterioles to force ?r for the simulations
to equal the experimental value at rbaseline? 0. These simulations
with regular lattices show a much more rapid decrease in ?r with
distance from the occluded arteriole than the actual data. In an
alternative calculation, the density of penetrating arterioles in each
lattice was set equal to the experimentally observed value, with a
concomitant increase in nearest neighbor distance from the mea-
sured mean value of 130 ?m, to values of 277 ?m for square, 290
?m for triangular, and 243 ?m for hexagonal lattices. In all three
regular lattices, ?r decreases to zero at rbaseline ? 300 ?m. We
penetrating arteriole after a clot is consistent with a random spatial
distribution of arterioles (Fig. 6C).
A Model for Blood-Flow Reduction. The parsimonious interpreta-
tion of our data is that collateral flow from other penetrating
arterioles makes a limited contribution to the territories about
each diving arteriole. We attempted to quantify this notion by
relating the dependence of the decrement in RBC speed on
distance (Fig. 3B) to the spatial distribution of penetrating
arterioles (Fig. 6 D–F). A 2D continuum model of the contri-
bution of one penetrating arteriole to the perfusion of local
tissue was constructed by using the measured locations of
penetrating arterioles, as above. We represented the parenchy-
mal flow from each arteriole by a monotonically decreasing
function with cylindrical symmetry about the arteriole and
assumed that the blood flow at any point in the tissue was the
sum of the contributions from every penetrating arteriole (SI
Text and Fig. 11). We then calculated the expected decrement in
?m, which corresponds to the nominal distance between branch
points of microvessels (26), together with the observed random
distribution of arteriole spacing (Fig. 6C), yields a good fit to the
in Fig. 3B). The predicted flow within the territory that sur-
rounds each penetrating arteriole is relatively uniform except
near some boundaries between neighboring territories (Fig. 6D).
The occlusion of a single penetrating arteriole yields a well
localized region of decremented flow (Fig. 6 E and F). Thus our
model accounts for the major effect of an occlusion.
Our measurements of the changes in blood flow induced by an
occlusion suggest that a single penetrating arteriole contributes
?m in radius (Figs. 3 and 6). Detailed flow measurements on a
to the occluded arteriole (Figs. 2 and 4). These measurements
support the idea that the microvascular bed contains distinct
territories that are supplied by a single penetrating arteriole.
However, there is also significant variability in the speed of RBC
interpret this variability as evidence that different territories are
also interconnected at the microvascular level, albeit with insuffi-
cient capacity to maintain normal levels of flow.
The measured distribution of distances between penetrating
vessels is consistent with random spacing (Fig. 6C). Further, the
density is sparse compared with lattices of penetrating arterioles
whose spacing is set equal to the mean distance between nearest
neighboring arterioles. In lattices, the territory of each arteriole
is smaller than the experimentally measured spatial configura-
tions (Fig. 6C). A model of blood flow that incorporates the
measured distribution suggests that removal of a penetrating
distribution (Fig. 6E), as observed (compare data and model in
Figs. 3B and 6C). The result of this analysis suggests that the
changes observed in blood flow after penetrating vessel occlu-
sion are passive properties of the vascular geometry. Thus,
variation in nearest-neighbor arteriole distances may play a role
in modulating the severity and area of ischemia after occlusion.
Flow Within the Cerebrovascular Hierarchy. A comparison of the
sensitivity to changes in blood flow after a clot among three
different levels in the cerebral vascular hierarchy suggests that
ischemia is most severe when a clot is located in a penetrating
vessel (Fig. 7). Previous work has shown that a clot in the surface
network of arterioles generates only a mild decrease in blood
flow, even as close as the first branch downstream (2). This is
consistent with the richly interconnected 2D nature of this
network. In contrast, a clot in a deep microvessel leads to a
severe drop in the speed of RBCs in the first downstream branch
(8), similar to the drop observed after the occlusion of a
D1 D2 D3 & D4
o i s
u l c
o r e t f a d
n i l e
b f o n
o i t c
(B) Bars show the mean and SEM of fraction of baseline RBC speed after
surface communicating arterioles (2) for vessels at different downstream
branches relative to the occluded vessel.
Summary of flow changes from occlusions at different levels of
Nishimura et al.PNAS ?
January 2, 2007 ?
vol. 104 ?
no. 1 ?
penetrating arteriole. However, the flow returns to 0.5-times Download full-text
baseline values within three branches from the blockage, con-
sistent with collateral flow through microvessels that originates
from the same penetrating arteriole. In contrast to the case for
surface arterioles and deep microvessels, occlusion of a pene-
trating arteriole has a devastating effect on the flow through
downstream microvessels, as collateral flow from neighboring
bottlenecks in the link between surface arterioles and the
tortuous network of microvessels that supplies blood throughout
the depth of cortex (Fig. 7).
after the photothrombotic occlusion of penetrating arterioles is
consistent with observations of pathology in human patients. Our
results support the clinically based hypothesis that penetrating
(13, 14). For example, subcortical regions are particularity vulner-
able to ischemia that results in lacunar lesions, because the blood
arteries are an important mechanism in cerebrovascular disease in
humans. Recent work indicates that microstrokes that are com-
pletely or largely confined to the gray matter (28–30), along with
small pial infarcts after certain strokes (31), have clinical conse-
benefit from animal models of small stroke that target penetrating
Our subjects were 16 male Sprague–Dawley rats, 100–350 g in
mass, that were anesthetized by interperitoneal injection of
urethane (150 mg per 100 g of rat) and maintained and surgically
prepared for in vivo TPLSM imaging of parietal cortex as
described (2). Image stacks of surface vasculature that were
taken with a 0.28-N.A., ?4 air objective (Olympus, Melville,
NY) and a 0.30-N.A. ?10 dipping objective. High-resolution
imaging, line-scan measurements, and photoexcitation of the
photosensitizer dye made use of a 0.8-N.A. ?40 dipping objec-
tive. The care and experimental manipulation of our animals
have been reviewed and approved by the Institutional Animal
Care and Use Committee at University of California at San
Occlusion of Penetrating Arterioles. Penetrating arterioles were
defined as vessels that branched off of a surface arteriole,
penetrated into the brain parenchyma, and served as the sources
for capillaries. All occluded penetrating arterioles selected for
study had at least a short segment in which the flow was confined
to the brain surface. This segment was used to measure the RBC
velocity before the occlusion.
Photothrombotic occlusions made use of injections of rose
bengal, as described (2). Green laser light, 0.1–5 mW, was focused
in a diffraction-limited spot coplanar with the imaging beam to
permit near real-time monitoring of clot progression. The spot was
aimed into the lumen of the target vessel in locations that ranged
from just distal from its upstream source on the brain surface to
above the first branch to the capillary bed, but no deeper than 50
?m below the pia. The vessel was irradiated in bouts that lasted 2–5
s with the spot scanned to ensure a complete clot.
Immunohistochemistry. In seven cases, animals were injected with
animals were perfused transcardially with 100 ml of phosphate-
buffered saline (PBS), followed by 100 ml of 4% (wt/vol) parafor-
maldehyde in PBS. Fiducial marks were made at known locations
relative to the target vessel by passing ?20 ?A through a single
tungsten electrode that was translated at 2 mm/s through the tissue.
The brain was removed and cryoprotected with 50% (wt/vol)
on a freezing-sliding microtome. Sections near the location of the
clots were selected based on the location of the targeted vessels
relative to the fiducial marks.
Reduced pimonidazole was visualized in tissue sections by
incubation for 48 h in primary antibody (anti-HypoxyProbe)
(90204; Chemicon), diluted 1:1,000 in PBS with 10% (vol/vol)
goat serum and 2% (vol/vol) Triton X-100, followed by biotin-
ylated rat-adsorbed anti-mouse IgG antibody (BA2001, Vector
Laboratories, Burlingame, CA), avidin-biotinylated peroxidase
complex (PK4000; Vector Laboratories), and finally a diamino-
benzidine reporter (SK4100; Vector Laboratories). Immunore-
activity to IgG was performed by overnight incubation of
sections in biotinylated antiuniversal IgG antibody (PK6200;
Vector Laboratories) at 1:1,000 dilution in the PBS-based di-
luent followed by the previous sequential incubation steps.
We thank Scott Lee, Harry Suhl, Philbert Tsai, and Thomas Woolsey for
useful discussions and Earl Dolnick, Rodolfo Figueroa, and Naomi Kort
for technical assistance. This work was funded by the National Institutes
of Health Grants EB/003832 (to D.K.), NS/043300 (to P.D.L.), and
RR/021907 (to D.K.) and by National Science Foundation Grant DBI/
0455027 (to D.K.).
1. Brozici M, van der Zwain A, Hillen B (2003) Stroke 34:2750–2762.
2. Schaffer CB, Friedman B, Nishimura N, Schroeder LF, Tsai PS, Ebner FF,
Lyden PD, Kleinfeld D (2006) PLoS Biol 4:e22.
3. Scremin OU (1995) in The Rat Nervous System, ed Paxinos G (Academic, San
Diego) pp 3–35.
4. Moody DM, Bell MA, Challa VR (1990) Am J Neuroradiol 11:431–439.
5. Woolsey TA, Rovainen CM, Cox SB, Henger MH, Liange GE, Liu D,
Moskalenko YE, Sui J, Wei L (1996) Cereb Cortex 6:647–660.
6. Harrison RV, Harel N, Panesar J, Mount RJ (2002) Cereb Cortex 12:225–233.
7. Motti ED, Imhof, H-G, Yasargil MG (1986) J Neurosurg 65:834–846.
8. Nishimura N, Schaffer CB, Friedman B, Tsai PS, Lyden PD, Kleinfeld D (2006)
Nat Methods 3:99–108.
9. Rosenblum WI, Zweifach BW (1963) Arch Neurol 9:414–423.
10. Cox SB, Woolsey TA, Rovainen CM (1993) J Cereb Blood Flow Metab
11. Rovainen CM, Woolsey TA, Blocher NC, Wang D-B, Robinson OF (1993)
J Cereb Blood Flow Metab 13:359–371.
12. Iadecola C (2004) Nat Rev Neurosci 5:347–360.
13. Fisher CM (1969) Acta Neuropathol 12:1–15.
14. Gan R, Sacco RL, Kargman DE, Roberts JK, Boden-Albala B, Gu Q (1997)
15. Millikan C, Futrell N (1990) Stroke 21:1251–1257.
16. Roos MW, Sperber GO, Johansson A, Bill A (1996) Exp Neurol 137:73–80.
17. Miyake K, Takeo S, Kaijihara H (1993) Stroke 24:415–420.
18. Macdonald RL, Kowalczuk A, Johns L (1995) Stroke 26:1247–1250.
19. Svoboda K, Denk W, Kleinfeld D, Tank DW (1997) Nature 385:161–165.
20. Zhang S, Boyd J, Delaney KR, Murphy TH (2005) J Neurosci 25:5333–5228.
21. Kleinfeld D, Mitra PP, Helmchen F, Denk W (1998) Proc Natl Acad Sci USA
22. Chaigneau E, Oheim M, Audinat E, Charpak S (2003) Proc Natl Acad Sci USA
23. Dirnagl U, Villringer A, Einhaupl KM (1992) J Microsc 165:147–157.
24. Hutchinson EB, Stefanovic B, Koretsky AP, Silva AC (2006) NeuroImage
25. McDonald DM, Choyke PL (2003) Nat Med 9:713–725.
26. Tata DA, Anderson BJ (2002) J Neurosci Methods 113:199–206.
27. O’Brien JT, Erkinjuntti T, Reisberg B, Roman G, Sawada T, Pantoni L, Bowler
JV, Ballard C, DeCarli C, Gorelick PB, et al. (2003) Lancet Neurol 2:89–98.
28. Vinters HV, Ellis WG, Zarow C, Zaias BW, Jagust WJ, Mack WJ, Chui HC
(2000) J Neuropathol Exp Neurol 59:931–945.
29. Mezzapesa DM, Rocca MA, Pagani E, Comi G, Filippi M (2003) Arch Neurol
30. Kovari E, Gold G, Herrmann FR, Canuto A, Hof PR, Michel JP, Bouras C,
Giannakopoulos P (2004) Stroke 35:410–414.
31. Lee DK, Kim JS, Kwon SU, Yoo S-H, Kang D-W (2005) Stroke 36:2583–2588.
www.pnas.org?cgi?doi?10.1073?pnas.0609551104Nishimura et al.