Cortical spreading depression causes and coincides with tissue

© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
ARTICLES
Cortical spreading depression causes and coincides
with tissue hypoxia
Takahiro Takano1, Guo-Feng Tian1, Weiguo Peng1, Nanhong Lou1, Ditte Lovatt1, Anker J Hansen2,
Karl A Kasischke1 & Maiken Nedergaard1
Cortical spreading depression (CSD) is a self-propagating wave of cellular depolarization that has been implicated in migraine
and in progressive neuronal injury after stroke and head trauma. Using two-photon microscopic NADH imaging and oxygen sensor
microelectrodes in live mouse cortex, we find that CSD is linked to severe hypoxia and marked neuronal swelling that can last
up to several minutes. Changes in dendritic structures and loss of spines during CSD are comparable to those during anoxic
depolarization. Increasing O2 availability shortens the duration of CSD and improves local redox state. Our results indicate
that tissue hypoxia associated with CSD is caused by a transient increase in O2 demand exceeding vascular O2 supply.
CSD is a slowly moving wave of synaptic activity cessation and tissue
depolarization first described in 1944 (ref. 1). The mechanism of CSD
wave propagation remains poorly understood2. Here we have observed
the cellular events during the propagation of CSD using two-photon
microscopy of exposed cortex of live mice3,4. Two-photon imaging and
spectroscopy applied to NAD+ and its reduced form, NADH, provides
three-dimensionally resolved, quantitative metabolic imaging in brain
tissue with cellular resolution5. NADH is the principal electron carrier
in glycolysis, the Krebs cycle and the mitochondrial respiratory chain.
Only the reduced form, not NAD+, is fluorescent. Therefore intrinsic
NADH tissue fluorescence serves as a sensitive indicator of the cellular
redox state in vivo6. The technical advantages of two-photon microscopy enabled us to visually monitor tissue redox state, capillary
perfusion and cellular morphology while a CSD wave passed across
the field of view. Combining optical observation with electrical
recording of tissue oxygenation, we found that CSD was associated
with depletion of tissue O2 and marked neuronal swelling. Our study
puts forward a new concept: that a hypoxic state can occur in the
absence of reduction in cerebral blood flow under conditions of high
energy demand. This observation may have direct clinical implications,
as several lines of work support the notion that CSD constitutes the
neurological basis of migraine with aura, and spontaneous waves
of CSD may contribute to secondary injury in stroke and traumatic
brain injury7,8.
RESULTS
Mixed pattern of changes in NADH during CSD
Two-photon imaging of the exposed cortex of adult mice in vivo
showed that intrinsic NADH fluorescence under resting conditions
was evenly distributed and individual cells could not be outlined
(Fig. 1a). Upon induction of CSD, massive and complex
spatiotemporal changes in the NADH fluorescence developed. As the
wave front of CSD advanced through the field of view, NADH
fluorescence dropped –15.6 ± 1.4% (n ¼ 5) across the field of view.
This initial decrease in NADH fluorescence was transient and followed
by an additional decrease in NADH in small elongated areas with radii
of 10–30 mm. The areas with reduced NADH fluorescence (dip) were
surrounded by irregularly shaped zones of NADH fluorescence
increases (overshoot) (Fig. 1a and Supplementary Video 1 online).
These changes in the tissue redox state occurred along with a deflection
of the local direct-current (DC) potential characteristic of spreading
depression (Fig. 1b). Both the dip and the overshoot of NADH
fluorescence were of large amplitudes, peaking at –19.1 ± 1.1% and
+26.0 ± 3.5% of control values (n ¼ 5), respectively (Fig. 1c,d). The
amplitudes of these CSD-associated NADH fluctuations were one
order of magnitude larger than the modest 2–4% changes in NADH
fluorescence evoked by physiological activity5, clearly indicating pathophysiological metabolic transitions during CSD. The transition
between areas of NADH dip versus overshoot was abrupt within a
temporal resolution of 2 s, giving rise to a characteristic pattern of dark
areas set on a bright background of increased NADH. This heterogeneous pattern of NADH fluorescence change was followed by a
uniform increase in NADH fluorescence across the field of view that
lasted several minutes (Fig. 1a,c). Integrating NADH fluorescence
signal across the whole imaging field (500–700 mm) revealed that the
overall NADH signal first showed an initial dip peaking –5.5 ± 1.5%
(n ¼ 5) followed by a prolonged overshoot (Fig. 1c,d), in accordance
with earlier reports9,10.
Because NADH fluorescence is an indirect measurement of tissue O2
(ref. 6), our next approach was to administer dextran-conjugated
fluorescein-5-isothiocyanate (FITC-dextran) intravenously to outline
the vasculature. Combined imaging of NADH and intravascular
1Center
for Aging and Developmental Biology, Department of Neurosurgery, University of Rochester Medical Center, 601 Elmwood Avenue, Box 645, Rochester, New York
14642, USA. 2Discovery, Novo Nordisk A/S, DK-2760, Maaloev, Denmark. Correspondence should be addressed to T.T. ([email protected]).
Received 8 March; accepted 29 March; published online 29 April 2007; doi:10.1038/nn1902
754
VOLUME 10
[
NUMBER 6
[
JUNE 2007 NATURE NEUROSCIENCE
ARTICLES
57 s
60 s
63 s
66 s
69 s
72 s
60
Figure 1 CSD is associated with marked
changes in NADH signal in cortex of adult mice.
(a) Pseudocolored images of NADH fluorescence
changes during the passage of a wave of CSD
50 mm below the cortical surface. NADH
decreases (dips) are displayed in green-blue and
NADH increases (overshoot) in red-yellow. Dashed
lines indicate the wave front moving across the
field. Arrowheads indicate areas with NADH
overshoot. Scale bar, 100 mm. (b) LFP recording
of the CSD wave in a. (c) Time-course relative
changes of NADH fluorescence in dip (green),
overshoot (orange) and integration of the whole
field (black) of the CSD wave shown in a. Means
and s.d. were calculated from eight areas of
dip and overshoot in the field. (d) Summary
histogram of initial peak NADH fluorescence
changes (mean ± s.e.m.).
b
40
20
0
–20
5 mV
–40
135 s
30 s
–60
Overshoot
Whole field
20%
Dip
30 s
[
NUMBER 6
[
30
20
10
0
–10
–20
–30
–40
FITC-dextran showed that each region with decreased NADH signal
contained a capillary in its center (Fig. 2a). This heterogeneous pattern
of NADH fluorescence changes lasted longer than local DC depolarization, with an average of 84.3 ± 1.9 s (n ¼ 15 events in seven mice)
(Fig. 2b,c). Measurements of NADH fluorescence changes close to the
electrode tip showed that tissue depolarization occurred 3.6 ± 0.7 s
(n ¼ 15 in seven mice) before the changes in the NADH fluorescence
signal (Fig. 2b,c). The characteristic pattern of NADH fluorescence
dips and overshoots moving horizontally along the CNS surface is
apparently CSD specific and does not result from reduced O2 delivery,
because it could not be replicated by lowering O2 supply. Cardiac arrest
resulted in a uniform NADH fluorescence increase, and anoxic depolarization was not associated with a mixed pattern of changes in NADH
signal (Fig. 2d).
Taken together, our results showed that the observed changes in local
redox state upon CSD were a function of the distance to the vasculature: tissues located closer than 8–10 mm from a capillary vessel wall
showed NADH dips, whereas those located further away showed
NADH overshoots. The steep phase of tissue depolarization in the
forefront of CSD waves preceded the characteristic pattern of changes
in NADH signal, implying that the sudden demand for O2 reflected the
intense increase in workload of membrane ion pumps responsible for
normalization of ion homeostasis. A simple yet logical explanation for
the observed pattern of changes in NADH fluorescence is that perivascular tissue with prime access to O2 increased the rate of oxidative
metabolism and microwatershed areas located further away from the
vasculature became hypoxic as a result of excessive consumption of O2
in perivascular tissue (Fig. 2a). In support of this idea, topical
NATURE NEUROSCIENCE VOLUME 10
40
JUNE 2007
Whole field
d
Dip
c
Overshoot
105 s
Fluorescence change (%)
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
a
application of high K+ (100 mM) induced DC
depolarization and NADH fluorescence
changes similar to those observed during
CSD (Fig. 2d). Prior studies of CSD-evoked
changes in NADH fluorescence, based on
fluorimetry or conventional imaging techniques, lacked cellular resolution to detect the
local dynamics. In these studies, NADH fluorescence was reported to show a biphasic
response consisting of an initial decrease
followed by an increase during CSD, as in
our observations9–11.
Dynamics of local perfusion during CSD
One hypothetical explanation for the increase
in NADH fluorescence in microwatershed
areas is that local perfusion might transiently decrease as a result of
swelling of astroglial endfeet. The rationale for this model is that
astrocytes respond to high K+ with a rapid increase in cell volume, and
swelling of pericapillary endfeet might compress capillary flow, resulting in local oligemia12. To test this idea, we quantified the local rate of
perfusion by measuring arterial cross-sectional area during the passage
of CSD4. Contrary to the expectations of this hypothesis, the arterial
diameter increased after the onset of CSD, and this was followed by
prolonged vessel constriction13 (Fig. 3a,b and Supplementary Videos
2 and 3 online). The vessel dilation lasted 101.3 ± 6.8 s (n ¼ 9 in five
mice), outlasting the DC depolarization by approximately 30 s
(Fig. 3b,c). Penetrating arteries showed a +50.7 ± 8.2% (n ¼ 12 in
six mice) increase in their cross-sectional area, followed by a –16.9 ±
7.3% (n ¼ 10 in five mice) constriction compared with baseline
(Fig. 3d). Veins also dilated during CSD and then underwent a
moderate constriction, but the amplitudes of changes in veins were
modest compared with those in arteries (+11.0 ± 4.4% and –7.3 ± 1.8%
respectively, n ¼ 9 in five mice) (Fig. 3d). Line scanning with a
temporal resolution of 2–3 ms showed that basal capillary flow rate
followed a similar pattern to the dynamic changes in arterial diameter.
A prolonged period of increased capillary flow with a duration of up to
several minutes was followed by a relative decrease in capillary flow
(Fig. 3e). Thus, continuous imaging of arterial cross-sectional
areas and capillary flow rates contradict the idea that the NADH
fluorescence increases resulted from transient oligemia. In accordance
with previous studies14,15, our analysis indicates that CSD is associated
with a transient increase in local perfusion that is followed by a
prolonged oligemia.
755
ARTICLES
a
NADH
Blood vessels
Merge
255
0
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
b
d
60
CSD
40
20
DC
deflection
0
–20
–40
–60
5 mV
20 s
Cardiac arrest
before depolarization
100
50
Lag
0
–50
–100
20%
Cardiac arrest
after depolarization
100
50
20 s
NADH
dip and overshoot
0
Figure 2 The microvasculature determines the
pattern of NADH fluorescence changes during
CSD. (a) NADH fluorescence changes during peak
depolarization (left panel). Intravascular labeling
with FITC-dextran (middle panel) combined with
NADH fluorescence imaging (right panel) shows
that NADH dips strictly colocalized with
capillaries. NADH overshoots occur in
microwatersheds at greater distances from the
capillaries. Scale bar, 100 mm. (b) Comparison
of tissue depolarization and NADH fluorescence
changes in the areas adjacent (o50 mm) to
the recording electrode. (c) Comparison of the
duration of the CSD-evoked DC depolarization, the
duration of NADH dip and overshoot, and the lag
between the onset of NADH fluorescence changes
and the onset of tissue depolarization (mean ±
s.e.m.). (d) The complex pattern of changes in
NADH (dip – overshoot) is triggered by high K+
and not by reduced blood flow. The pattern of
NADH fluorescence changes during CSD (first
panel) is distinctly different from the uniform
increase in NADH fluorescence changes evoked
by cardiac arrest before (second panel) or after
(third panel) the anoxic depolarization. Topical
application of 100 mM KCl evokes a mixed
pattern of NADH fluorescence dips and overshoots
(fourth panel) similar to that seen in CSD.
Scale bar, 100 mm.
–50
c
–100
DC
deflection
KCI–induced
depolarization
NADH
dip and overshoot
Lag
0
20
40
60
Time (s)
80
100
Because CSD has been reported to evoke an initial vasoconstriction
in mice13, we next used laser Doppler flowmetry as an alternative
approach to analyze changes in cortical blood flow during the passage
of a CSD wave. We confirmed that an initial decrease in cerebral blood
flow preceded the hyperemic phase of CSD (Supplementary Fig. 1
online). The early phase of CSD-induced oligemia was noted in mice
anesthetized with isoflurane, with pentobarbital or with urethane plus
a-chloralose (Supplementary Fig. 1), but was of smaller amplitude
(5–10% of baseline) than previously reported13. The smaller amplitude
of oligemia observed in our study may reflect strain differences. We
used FVB mice, whereas previous studies employed SV129 or C57BL
mice13. The hyperemic response during CSD averaged B110% of
baseline in mice anesthetized with isoflurane and with urethane plus
a-chloralose, but was smaller (B50%, P o 0.05) in those treated with
pentobarbital (Supplementary Fig. 1).
CSD is associated with severe tissue hypoxia
An alternative explanation for localized increases in NADH (overshoot)
is that the increase in blood flow is not sufficient to provide the
additional O2 required for the restoration of normal transmembrane
ion gradients in the wake of CSD. To address this question, we
combined recordings of the DC potential with measurements of local
tissue O2 tension (pO2) using microelectrodes with tip diameters of
3–4 mm (Fig. 4a)16. CSD caused a substantial drop (420 mmHg) in
756
tissue pO2 that lasted for a period of 125 ± 15 s,
or more than 2 min (n ¼ 7 in six mice). The
40
hypoxic period was 56 s longer than the DC20
potential deflection, but recovered 92 s faster
0
than the electroencephalogram (EEG) activity
–20
(Fig. 4a,b). The decrease of pO2 occurred
–40
shortly (1.6 ± 1.4 s, n ¼ 7 in six mice) after
–60
the onset of the DC-potential deflection
(Fig. 4a–d). Most remarkably, CSD substantially decreased tissue pO2 to 3.7 ± 0.8 mmHg,
from a baseline of 26.7 ± 4.1 mmHg (n ¼ 8 in six mice) (Fig. 4a,e). The
decline of O2 tension during the passage of CSD has been reported both
in early studies of CSD17 and in more recent studies9,11. Notably, the O2
electrode used in this study integrates pO2 in a volume with a diameter
of B8–10 mm, and we were unable to compare pO2 in areas of NADH
dips versus overshoots.
To compare the O2 changes induced by CSD with those induced by
anoxia, we next induced cardiac arrest by intravenous administration
of a bolus of saturated KCl. Cardiac arrest induced changes in the DC
potential with a similar amplitude to those induced by CSD (Fig. 4c).
As expected, O2 depletion preceded anoxic depolarization by 116.1 ±
12.1 s (n ¼ 4), consistent with the concept that terminal depolarization
is caused by ATP depletion18 (Fig. 4c,d). Upon cardiac arrest, pO2
decreased to 0.1 ± 0.3 mmHg (n ¼ 4), or to a state of complete O2
depletion (Fig. 4c,e).
These studies indicate a fundamental distinction between CSD and
anoxic depolarization: in CSD, tissue depolarization precedes the pO2
decline, whereas the order is reversed after cardiac arrest; in the latter
case, a pO2 decline precedes tissue depolarization by approximately
2 min (Fig. 4d). Another difference is the characteristic mixed pattern
of high-low NADH fluorescence triggered by CSD, but not by
anoxia (Fig. 2d).
To further investigate the relation between tissue pO2 and NADH
changes during CSD, tissue pO2 was artificially increased by superfusing
60
VOLUME 10
[
NUMBER 6
[
JUNE 2007 NATURE NEUROSCIENCE
ARTICLES
3
Figure 3 CSD is associated with a transient increase in local perfusion
followed by oligemia. (a) Time-course images of cross-sectional view of a
penetrating artery during CSD. Scale bar, 10 mm. (b) Comparison of DCpotential changes and vessel cross-sectional area during spreading
depression. Gray lines 1, 2 and 3 indicate the times at which images
in a were captured. (c) A summary histogram of the duration of tissue
depolarization, duration of arterial dilation, and latency of onset of
vasodilation and depolarization (mean ± s.e.m.). (d) Maximal vasodilation
during the first 2 min of passage of CSD and maximal constriction afterwards
in arteries (diameters 15–20 mm) and veins (diameters 15–30 mm) (mean ±
s.e.m.). (e) Line scan images of capillary perfusion during the passage of
CSD. The velocity with which erythrocytes pass through capillaries mirrors
changes in arterial vessel diameters.
DC
Dilation
Lag
0
25
50
75
100
Time (s)
d
DC
5 mV
20 s
1
50 µm
20 s
Arterial
crosssectional
area
3
Dilation
Constriction
2
e
70
60
50
40
30
20
10
0
–10
–20
–30
Artery
Vein
Max
Before During
After
0.33
0.42
0.24
mm s–1 mm s–1 mm s–1
Min
when superfusion with aCSF was discontinued. Combined, these
observations support the notion that CSD-induced increases of
NADH result from a rapid increase in O2 consumption, which
transiently depletes O2 supply in the microwatershed areas.
the exposed cortex with 100% oxygenated artificial cerebrospinal fluid
(aCSF). This artificial increase in tissue pO2 converted NADH fluorescence during CSD from the mixed pattern of high-low changes to a
relatively uniform dip in NADH fluorescence. Thus increased tissue pO2
converted the overshoot in NADH fluorescence in a microwatershed
area into a dip of NADH (Fig. 4f,g). Conversely, perfusion with
oxygen-depleted aCSF (equilibrated with N2) increased the basal
NADH fluorescence and resulted in an additional increase in the areas that showed an
NADH overshoot (Fig. 4f,g). The effect of
a CSD
manipulating pO2 was fully reversible: the
mixed pattern of high-low NADH reappeared
Neuronal swelling and distortion
It has been reported that during hypoxia dendrites experience swelling
and loss of spines within minutes19. We reasoned that if CSD were
associated with an episode of severe hypoxia lasting B2 min, CSD
might trigger structural changes of dendritic processes. To address this
b
EEG
DC
0.5 mV
EEG
20 s
Figure 4 Severe tissue hypoxia during CSD.
(a) Combined measurements of EEG, DC potential
and pO2 during CSD. (b) A summary histogram of
the durations of EEG arrest, DC depolarization
and drop of tissue pO2 (mean ± s.e.m.).
(c) Changes in EEG, DC potential and tissue
pO2 evoked by cardiac arrest. (d) A histogram
comparing lag between onset of tissue
depolarization and drop of tissue pO2 in CSD and
in cardiac arrest (mean ± s.e.m.). (e) Comparison
of changes in tissue pO2 before and during CSD,
and after cardiac arrest (mean ± s.e.m.,
*P o 0.02, t-test). (f) NADH fluorescence
changes during CSD are modulated by tissue pO2.
Control CSD (left panel) gives rise to the distinct
pattern of NADH dips and overshoots. Superfusion of the exposed cortex with oxygenated
aCSF (middle panel) results in a relatively uniform
dip in NADH fluorescence in the same imaging
field. Superfusion with deoxygenated aCSF (right
panel, equilibrated with N2) gives rise to a
homogeneous increase in NADH fluorescence
during the passage of a CSD wave. Scale bar,
100 mm. (g) NADH fluorescence in arbitrary
fluorescence unit (AFU) before and during CSD
in perivascular and distant area shown in f. In
controls, NADH fluorescence decreases in
perivascular areas (green lines) and increases in
distant areas (orange lines). When the tissue is
oxygenated, both perivascular and distant area
show decreases in NADH fluorescence.
Superfusion with deoxygenated aCSF increases
the baseline NADH fluorescence, and CSD results
in further increases of NADH in both the
perivascular and distant areas.
NATURE NEUROSCIENCE VOLUME 10
[
O2
0
5 mV
DC
20 s
d
100
Duration (s)
e
20
CSD
O2
c
20 s
Cardiac arrest
0.5 mV
EEG
20 s
–60
–80
–100
20 s
20
15
5
0
Base
3,000
20 s
f
NUMBER 6
Control
O2
Peak
g
10 mmHg
O2
25
10
–120
5 mV
Cardiac
arrest
30
–40
–140
DC
35
–20
10 mmHg
200
40
0
Lag DC – O2 (s)
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
2
Cross-sectional area (%)
b
Tissue pO2 (mmHg)
2
N2
100
50
0
NADH (AFU)
c
1
CSD
Cardiac arrest
a
2,500
2,000
1,500
1,000
500
Perivascular
Control
O2
N2
Distant
Control
O2
N2
–50
0
–100
[
JUNE 2007
Base
Peak CSD
757
a
Before CSD
b
Before CSD
12 s
42 s
Figure 5 CSD is associated with marked neuronal
swelling and distortion of dendritic spines.
(a) Dendrites of excitatory neurons expressing
YFP (Thy1 promoter) showing blebbing and loss
of dendritic spines during CSD. Arrowheads
indicate the spines. (b) The blebbing and the
loss of spines are unaltered by ventilating 100%
O2 and are reproduced by topical application
of 100 mM KCl. (c) Anoxic depolarization also
causes a similar dendritic distortion. (d) Marked
increase in cell body volume of YFP-expressing
neurons during CSD. (e) A summary histogram of
maximal area changes in cell bodies of neurons
and astrocytes (mean ± s.e.m., *P ¼ 0.014,
t-test). (f) Astrocytes loaded with the astrocyte
specific indicator sulforhodamine 101 do not
experience detectable increases in cell volume
during the passage of a wave of spreading
depression. A general decrease in fluorescence
emission occurs concurrently with the tissue
depolarization, possibly reflecting tissue swelling.
Scale bars, 10 mm.
245 s
KCI
c
During CSD
Before anoxia
Anoxia
100% O2
Before CSD
15 s
18 s
66 s
42 s
132 s
120 s
question, we imaged dendrites in cortical layer 2 during the passage of a
wave of CSD in transgenic mice expressing fluorescent protein (yellow
fluorescent protein (YFP) or enhanced green fluorescent protein
(eGFP)) under either the Thy1 or the Gad1 promotor20,21. Notably,
dendrites showed various degrees of morphological change within
seconds of CSD depolarization, ranging from no apparent changes to
severe blebbing and loss of spines (Fig. 5a,b). Out of ten mice we
studied, four showed severe—an average of 77.9 ± 7.0%—loss of
dendritic spines during CSD. Additional analysis suggested that the
blebbing and the loss of spines resulted from K+-induced depolarization rather than from hypoxia. Changes in dendritic structure evoked
by cardiac arrest were intimately linked to anoxic depolarization
(Fig. 5c). Dendritic structures remained normal for minutes after
cardiac arrest, but showed marked changes concomitant with anoxic
depolarization. Loss of spines after anoxic depolarization averaged
78.5 ± 2.8% (n ¼ 4). Exposing cortex to an aCSF containing 100 mM
K+ (K+ replacing Na+) was similarly associated with the tissue depolarization and the loss of dendritic spines, averaging 78.7 ± 2.9% (n ¼ 3)
(Fig. 5b). Increasing O2 in the breathing mixture from 21% to 100%
during repeated passages of CSD in the same mouse did not reduce the
loss of dendritic spines (Fig. 5b). Combined, this analysis indicates that
loss of dendritic spines is caused by neuronal depolarization rather than
reduction in tissue pO2 and may be mediated by the concomitant influx
of Ca2+ (ref. 19). The structural changes in dendritic spines may
contribute to the transient cessation of neuronal activity during CSD,
which persists after the normalization of ion homeostasis (Fig. 4a).
We also quantified volume changes in neuronal somas; the crosssectional area of neuronal somas increased by 23.1 ± 4.8% (n ¼ 8 in
three mice), corresponding to a 37.1 ± 0.1% increase in the cell body
volume during CSD (Fig. 5d,e and Supplementary Video 4 online).
758
e
*
30
20
The morphology of most dendrites returned
to normal after CSD, and neuronal cell bodies
assumed their pre-CSD volumes within
0
8–10 min. In contrast to the marked swelling
of neurons, the structural appearance of astro–10
cyte processes and somas remained essentially
unaffected by CSD, with an insignificant
–20
change in peak soma area of –2.0 ± 6.1%
(n ¼ 16 in three mice), corresponding to a
0.9 ± 9.0% volume change (Fig. 5e,f). Thus,
whereas neurons experienced extensive volume shift and structural
changes during CSD, astrocytes remained relatively unaffected. These
results indicate that tissue swelling during CSD is likely to be mainly
neuronal. The rapid activation of volume-sensitive anion channels that
are linked to efflux of amino acid osmolytes may limit astrocytic
volume changes during the passage of CSD22.
10
Neurons
f
Before CSD
Astrocytes
d
Cell body area (% change)
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
ARTICLES
Manipulation of tissue hypoxia by O2 inhalation
Because hypoxia-induced neuronal injury is a function of the severity
and duration of O2 depletion, we asked whether it was possible to
reduce the hypoxia associated with CSD. In this regard, we note that
hyperbaric O2 is used in the treatment of migraine with aura and more
commonly in cluster headache23. We observed here that the duration of
CSD was a direct function of O2 concentration in breathing air.
Increasing the percentage of O2 from room air concentration (21%)
to 100% shortened the duration of CSD by –24.0 ± 7.3% (n ¼ 5),
whereas decreasing O2 to 15% or 10% prolonged the duration by
+25.0 ± 8.5% (n ¼ 4) and +43.5 ± 13.1% (n ¼ 5), respectively
(P o 0.01, analysis of variance; Fig. 6a,b). The amplitudes of DC
depolarization and velocity of CSD were not significantly altered by
manipulating the percentage of O2 in air (P 4 0.05, analysis of
variance; Fig. 6a,b). Thus, the CSD waves were shortened in mice
ventilated with 100% O2 and were prolonged by reducing O2 to 15% or
10%. In fact, hypoxia lowered the threshold for elicitation of CSD, and
spontaneous waves of CSD were observed in many mice when
O2 content fell below 8% (data not shown). Moreover, the NADH
signal during CSD was sensitive to changes in the percentage of inspired
O2 (Fig. 6c,d). With normal air (21% O2), an average of 24.5 ± 1.1% of
the total imaging field underwent an NADH fluorescence dip, whereas
NADH increased (overshot) in an area of 18.0 ± 0.8% (n ¼ 6) of the
VOLUME 10
[
NUMBER 6
[
JUNE 2007 NATURE NEUROSCIENCE
ARTICLES
a
b
60
100% O2
100%
15%
10%
5 mV
40
Percent change
20 s
20
0
–20
15% O2
d
100% O2
15% O2
Air
e
f
100% O2
10 mmHg
20 s
Baseline
shift
300
250
*
200
150
100
*
50
Percent area change
c
Velocity Amplitude Duration
80
60
40
20
0
–20
–40
–60
Amplitude (% change)
–40
Baseline pO2 (% change)
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
Air
100%
15%
*
*
*
*
Overshoot Dip
area
area
0
100% Air 15%
50
*
100
150
Figure 6 The severity of CSD-induced tissue
hypoxia is reduced by increasing O2 in breathing
mix. (a) Traces depicting three waves of CSD in a
mouse ventilated with 100% O2, room air (21%
O2) or 15% O2. The duration of CSD, but not the
velocity or amplitude, are inversely related to the
percentage of O2 in air. (b) Summary of velocity,
amplitude and duration of DC depolarization
normalized to room air in mice ventilated with
100%, 15% or 10% O2 (mean ± s.e.m.).
(c) NADH fluorescence dip (green) and overshoot
(red) areas during CSD wave passage in a mouse
ventilated with 100% O2, room air or 15% O2.
Left panel, NADH signal in grayscale and
vasculature in red in the same field during the
peak of CSD with room air. The mixed pattern of
high-low NADH signal is evident in the raw image.
The recording electrode is located in the lower
left corner. Scale bar, 100 mm. (d) Summary
histogram of the relative size of NADH dip and
overshoot area in mice ventilated with 100% and
15% O2 compared with dip and overshoot in room
air (mean ± s.e.m., *P o 0.01, Dunnett’s test
compared with room air). (e) Tissue pO2 in a
mouse ventilated with 100% O2, room air or 15%
O2. (f) Summaries of pO2 during baseline (resting)
conditions, amplitude of maximal pO2 dip during
CSD, and duration of CSD-induced hypoxia in
mice ventilated with 100% O2, room air or 15%
O2. All values were normalized to room air in the
same mouse (mean ± s.e.m., *P o 0.05,
Dunnett’s test compared with room air).
200
250
*
with 15% O2 and shortened by increasing the
percentage of O2 to 100% (Fig. 6e,f).
Air
100%
In summary, these observations indicate
pO2
*
that duration of CSD-induced hypoxia is an
dip
Air
inverse function of the availability of O2 and
15%
*
15% O2
that the tissue is capable of extracting and
0
50
100
150
200
consuming additional O2: the NADH fluoresDuration (% change)
cence signal during CSD shifted toward a
decrease, indicative of increased oxidative
metabolism6, when the percentage of O2 was
same field. When the percentage of O2 in air was increased from 21% to increased from 21% to 100%. The increased consumption of
100%, the NADH dip area expanded and the overshoot area contracted O2 resulted in a shortening of hypoxia and of tissue depolarization
(n ¼ 6). Conversely, switching to 15% O2 diminished the relative area (Fig. 6a,f). Conversely, decreasing the percentage of O2 to 15% was
of tissue undergoing an NADH dip and increased the overshoot area associated with a shift of NADH fluorescence changes toward an
(n ¼ 5). Because the NADH fluorescence signal indirectly detects increase, reflecting a decline in oxidative metabolism and a prolongacellular O2 consumption, these observations indicate that use of O2 is tion of tissue hypoxia and depolarization (Fig. 6a,f).
limited by vascular supply of O2.
We also studied the effect of O2 on tissue pO2 in cortex during CSD. DISCUSSION
When mice were ventilated with room air (21% O2), pO2 dropped The present study shows that CSD is associated with a more than 2-min
from 26.7 mmHg to 3.7 mmHg during the passage of CSD, period of tissue hypoxia, massive neuronal swelling and temporary loss
and hypoxia lasted an average of 125 s ± 15 s (n ¼ 7 in six mice) of dendritic spines. The tissue hypoxia was not a consequence of
(Fig. 4b,e). As expected, the baseline pO2 shifted when the percentage hypoperfusion, because local blood flow actually increased during CSD
of O2 in air was altered. When the mice were ventilated with 100% or (Fig. 3 and Supplementary Fig. 1). We propose that the drop in pO2
15% O2, tissue pO2 respectively increased to 237.0 ± 29.6% or reflects a period during which O2 consumption transiently exceeded
decreased to 55.6 ± 6.7% (n ¼ 9 and 8, respectively) (Fig. 6e,f). vascular delivery of O2. During the peak of CSD, we have previously
Because the resting pO2 was higher in 100% O2, the amplitude of the shown that extracellular K+ increases from a resting concentration of
pO2 dip during CSD became larger (226.1 ± 22.1% that in room air). B3–4 mM to 60–80 mM (ref. 18). As with anoxic depolarization, the
Conversely, the dip amplitude was smaller in 15% O2 (54.4 ± 7.1% loss of transmembrane ion gradients occurs within seconds.
that in room air) (Fig. 6e,f). The duration of the CSD-induced pO2 Subsequent normalization of ion concentrations requires extensive
decline followed a similar pattern to that of the direct-current (Na+ + K+)ATPase activity. Several lines of evidences presented here
depolarization: the hypoxic period was prolonged in mice ventilated support the notion that O2 supply during CSD is insufficient and that
0
100% Air 15%
NATURE NEUROSCIENCE VOLUME 10
[
NUMBER 6
[
JUNE 2007
300
759
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
ARTICLES
tissue adjacent to arteries and capillaries consumes O2 at the expense of
more distant tissue. The perivascular tissue with prime access to O2
showed a decrease in NADH fluorescence, indicating a shift of the
NADH/NAD+ ratio toward NAD+, which can only occur with adequate
O2 supply6. In contrast, NADH fluorescence increased in microwatershed areas, consistent with a shift in the NADH/NAD+ ratio
toward NADH because of inadequate oxygenation to support increased
aerobic metabolism. In parallel, tissue pO2 dropped from B40 mmHg
to B4 mmHg. The depletion of tissue O2 was not a result of reduced
supply, because capillary perfusion rate remained constant or increased
during the passage of CSD (Fig. 3). In fact, the elimination of O2 supply
by cardiac arrest was associated with a uniform increase in NADH
regardless of the distance to the vessels (Fig. 2). The two-dimensional
extension of the NADH transitions followed a wavefront and clearly did
not reflect a cellular pattern. This further supports our interpretation
that the underlying cause of the NADH transitions is predominately
metabolic and not cell-specific pathophysiological responses.
The results presented in this study further emphasize the applicability and suitability of intrinsic signals such as NADH for highresolution imaging of nervous tissue. Two-photon NADH imaging has
previously been shown to be capable of spatiotemporally resolving
activity-dependent metabolic transitions between neuronal and astrocytic processes within the neuropil of the hippocampal CA1 region5.
The present study applies spreading depression as a pathophysiological paradigm to evoke extreme metabolic changes, which now
reveals a clear spatiotemporal relationship between cerebral energy
metabolism and the cerebral microvasculature (Fig. 2). Such phenomena can, of course, only be observed in the living, intact animal, where
blood circulation and the spatial relationship between the vascular
tissue and the neuropil is preserved. This further emphasizes the
necessity of conducting metabolic imaging studies in vivo, if feasible.
Aforementioned earlier studies9,10,24 using NADH surface fluorimetry have reported an early NADH decrease (‘dip’), which occurred
during the peak of DC depolarization, followed by a longer-lasting
NADH overshoot. Surface fluorimetric imaging has lower spatial
resolution than our technique. Indeed, we found that by integrating
and averaging our NADH fluctuations across the whole imaging field
the response indicated an initial NADH dip followed by an NADH
overshoot during CSD. The previous studies also showed that anesthesia can convert the initial NADH dip to an NADH overshoot in adult
gerbils, but not in adult rats. Switching the anesthetic to pentobarbital
did not alter the NADH pattern formation (Supplementary Video 5
online). Several anesthetics dampen the amplitude of local hyperemia,
and can restrict O2 delivery and possibly revert the initial oxidative
cycle to a reductive cycle25.
Our observations indicated that small pockets of tissue located
between capillaries were exposed to hypoxia despite a robust increase
in cerebral perfusion during CSD. This local tissue hypoxia had a
functional consequence, as the inverse correlation between O2 supply
and the duration of CSD (Fig. 6) implied that O2 supply was the ratelimiting factor for normalization of extracellular K+. This finding raises
the question of whether other CNS pathologies are associated with local
hypoxia despite the absence of a blood flow reduction. Several other
pathological conditions, including seizure, are associated with a sharp
increase in energy demand and a reduction in tissue pO2 (ref. 26).
More subtle mismatches in O2 delivery and consumption may over
longer periods of time contribute to neuronal demise in neurodegenerative diseases27.
Another new observation is that CSD is associated with severe
neuronal swelling and the transient breakdown of dendritic structures,
with an apparent loss of many dendritic spines. The marked changes in
760
dendritic structure observed may result from the neuronal depolarization in response to high extracellular K+ and are likely to contribute to
the depression of EEG activity that accompanies CSD. Increasing O2 in
air to 100% did not cause a notable reduction of the morphological
alterations of dendritic structures during CSD (Fig. 5). Our observations confirm the original report that dendrites are most sensitive to
hypoxic depolarization and show structural alteration within minutes
of hypoxia28. The remarkable stability of astrocytic morphology during
CSD is probably a consequence of the fast volume recovery in these
cells. Astrocytes express a large number of volume-sensitive channels
that effectively limit transient changes in cell volume by rapid efflux of
small osmolytes, including taurine, aspartate and glutamate22.
Several compounds can block the initiation of CSD, including
NMDA receptor antagonists and gap junction inhibitors2, but attempts
to define a chemical mediator of CSD have failed. Once a wave of CSD
is evoked, its propagation is insensitive to a number of NMDA receptor
antagonists and channel inhibitors29 (unpublished observations). It is
therefore unlikely that a simple transmitter mechanism accounts for
the steady propagation of CSD. The slow velocity of CSD implies that a
diffusible factor may mediate its migration across cortex30. The wave of
CSD advances as K+ in its forefront reaches the threshold of 10–12 mM,
resulting in an abrupt decrease in membrane resistance and massive
efflux of cytosolic K+ (refs. 18,31). Consequently, in an attempt to clear
the high K+, the (Na+ + K+)ATPase exhausts the local O2 supply. Our
data support the idea that the limited availability of O2 slows the
clearance of K+. The inverse relationship between duration of CSD and
the percentage of O2 in air indicated that O2 availability was the
limiting factor for repolarization8. We propose that K+ diffusion32 in
combination with tissue hypoxia is an essential component of CSD. As
a result of hypoxia, extracellular K+ remains high for a prolonged
(B70 s) period. We propose that the long-lasting increase in extracellular K+ is required for the continued propagation of CSD, because it
allows K+ to diffuse forward along its concentration gradient, and CSD
thereby to continue its propagation across cortex. Further studies will
be required to establish the necessity of hypoxia for the propagation of
CSD, and a key question for future studies is whether further increases
of pO2 can abrogate CSD. Nevertheless, the inability to identify a
transmitter system responsible for propagation of CSD does indirectly
support the notion that K+ and hypoxia drive the forward movement of
CSD. The ability of two-photon imaging to analyze dynamic changes in
cellular redox state, cytosolic Ca2+, pH, cell volume and capillary
perfusion on a microscopic level in intact adult mice is likely to
prove useful in future examination of the neurovascular regulation.
The observations reported here have implications for a number of
neurological diseases. CSD-like waves are spontaneously elicited in
stroke and head trauma33,34. Such depolarization events may have an
essential role in secondary injury by recruiting additional tissue into the
expanding ischemic or traumatic lesion35–37. In vivo experiments
indicate that CSD-likes waves may be evoked at the infarct rim,
probably as a consequence of the high concentration of K+ and
glutamate within the ischemic core38. The propagation of CSD-like
waves in the penumbral regions adds metabolic stress to the already
hypoperfused penumbral tissue8. Therefore, tissue located adjacent to
the ischemic lesion is unable to restore normal extracellular K+ after the
passage of a CSD wave. The tissue remains depolarized and is
eventually incorporated in the infarct. Moreover, recent observations
have shown that CSD waves are commonly encountered in people
suffering from acute head trauma: a total of 73 spontaneous CSD-like
waves were monitored in six of twelve people studied37. These CSD-like
events may contribute to directly to peritraumatic hypoxia, vasoconstriction and post-traumatic confusion.
VOLUME 10
[
NUMBER 6
[
JUNE 2007 NATURE NEUROSCIENCE
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
ARTICLES
Several lines of evidence support the concept that the appearance
and evolution of scotoma in classic migraine involves CSD39,40. Highfield functional MRI mapping of blood oxygenation level–dependent
(BOLD) events during the auras preceding a migrainous attack show a
close spatial and temporal relationship between visual scotomas and a
self-propagating wave whose velocity and BOLD characteristics closely
resemble CSD7. An initial focal increase in BOLD signals progresses
contiguously and slowly across the occipital cortices of humans as they
experienced retinotopic progression. Because progression of scotomas
is temporally linked to an increase in the BOLD signal, the visual
disturbances do not result from ischemia7. Based on the observations
reported here, we suggest that migraine aura, which includes visual
defects or black spots, is the clinical manifestation of transient focal
tissue hypoxia resulting from insufficient supply of O2 during CSD. If
so, the severity of metabolic stress associated with migraine aura is
comparable to that associated with transient ischemic attacks, and the
striking distortion of dendritic structures may be linked to long-term
changes of synapses and cortical networks in those suffering from
repeated migraine attacks.
Migraine is considered a largely benign disorder, despite the frequent
impairment of cognitive performance during and after headaches and,
less commonly, of sensorimotor function in hemiplegic migraine41,42.
Subtle but significant deficits in memory, reaction time and attention
in migraineurs have been noted in some, but not in other, studies43,44.
Moreover, long-lasting dysfunction of contrast discrimination, color
perception, prolonged visual evoked potentials and decreased visual
field sensitivity have been noted in people with migraine45–48; all of
these are compatible with the common involvement of the occipital
cortex in visual aura and possible sequels of hypoxia during CSD waves.
Given the probable dependence of aura on CSD, it is anticipated that
prophylaxis, rather than pain relief, will benefit people with migraine in
the long term49.
METHODS
Mouse preparation. FVB mice, 8–12 weeks old, of either sex (Taconic), fasted
overnight, were anesthetized with ketamine (0.12 g per kilogram body weight)
and xylazine (0.01 g kg–1) intraperitoneally. We made a craniotomy, removed
the dura mater and glued a metal plate to the skull with dental acrylic cement.
The mice were intubated and artificially ventilated with a small-animal
ventilator (SAAR-830, CWE). We poured NaCl 0.9% solution containing 1%
agarose (37 1C) on the cranial window glued a glass coverslip to the metal plate.
Body temperature, mean arterial blood pressure and blood gas were monitored
and maintained as previously described3,4. In experiments involving manipulation of pO2 tension, we left the dura intact and superfused the exposed cortex
with artificial cerebrospinal fluid (aCSF). FITC-dextran (Sigma), 2,000 kDa,
was intravenously administered as previously described to outline the vasculature. We chose for study arteries and veins with diameters of 15–20 mm and
15–30 mm, respectively4. Cortical spreading depression was evoked by pressureinjecting 1 M KCl through a micropipette (tip diameter 2–3 mm) into cortex at
100–300 mm depth and 2–3 mm away from the imaging field (Picospritzer,
Parker). All experiments were approved by the Institution Animal Care and Use
Committee of University of Rochester.
In vivo two-photon imaging. After surgery, the mice were reanesthetized with
urethane and a-chloralose (1 g kg–1 and 50 mg kg–1, intraperitoneally). Body
temperature was maintained at 37 1C by a heating blanket (BS4, Harvard
Apparatus). Blood gasses were analyzed with a Rapidlab 248 (Bayer, sample size
40 ml). We completed experiments only if physiological variables remained
within normal limits. We used a custom-built microscope attached to a
Ti:sapphire laser (Tsunami, Spectra-Physics) and a scanning box (FV300,
Olympus) using Fluoview software and a 20 objective (0.95 numerical
aperture, Olympus) for two-photon imaging. NADH was detected with
740 nm excitation and an emission filter of either 460 nm emission filter with
50 nm bandwidth, or BGG22 colored glass. FITC-dextran was simultaneously
NATURE NEUROSCIENCE VOLUME 10
[
NUMBER 6
[
JUNE 2007
detected with a 515 nm emission filter with 50 nm bandwidth. We quantified
the acquired images with ImageJ (NIH) software and analyzed pseudocolor
images of NADH fluorescence changes using custom programs in Matlab
(MathWorks, Inc.)5. For noise suppression, all images were filtered using a
rotationally symmetric gaussian filter with a size of 5 pixels and an s.d. of 1. The
baseline image was numerically constructed as the average image of the last six
images before induction of CSD. For each pixel, the pixel intensity values of the
NADH fluorescence changes after induction of CSD were subtracted from the
pixel intensity values of the baseline image. The resulting differential images
were pseudocolored using colormaps ranging from –60 to 60 or from
–100 to 100 pixels.
Electrophysiological recording. DC potential was recorded by a patch pipette
(TW150F-4, WPI; outer diameter, 1.5 mm; inner diameter, 1.12 mm; tip
diameter, 2–3 mm), containing 1% FITC-dextran (20 kDa, Sigma) in 0.9%
NaCl. LFP signals were amplified at 0–1,000 Hz and digitized at 10 kHz. For
EEG, signals were further bandpass-filtered at 1–100 Hz. We measured tissue
pO2 using a modified Clark-type polarographic oxygen microelectrode with a
guard cathode for ptissue (3–4 mm tip; Unisense). We calibrated the electrode
before and after each experiment using distilled H2O equilibrated with 8% O2
balanced with N2; air; 100% N2; and using a reductant solution of 0.1 M
ascorbate and NaOH.
Quantification of volume changes in neurons and astrocytes. To visualize
neurons, we used transgenic mice expressing YFP in excitatory neurons under
the Thy1 promoter20 or expressing eGFP in interneurons under the Gad1
promoter21. YFP and eGFP were both excited at 910 nm and emission collected
with a 515 nm emission filter with 50 nm bandwidth. The data from Thy1-YFP
and Gad1-eGFP mice were pooled, as no differences were noted. Astrocytes
were visualized by sulforhodamine 101 loading50. The volume of the cell bodies
was calculated as area3/2.
Laser Doppler flowmetry. Relative change in local perfusion was detected by
laser Doppler flowmetry (PF5010 Laser Doppler Perfusion Module with a
microtip, PR 418-1, Perimed). The DC recording electrode was placed adjacent
to the probe. Signals were converted by a DigiData 1332A interface (Axon
Instruments) with an interval of 200 ms and recorded with the pCLAMPEX 9.2
program (Axon Instruments).
Note: Supplementary information is available on the Nature Neuroscience website.
ACKNOWLEDGMENTS
This work was supported in part by grants NS30007 and NS38073 from the US
National Institutes of Health and the National Institute of Neurological Disorders
and Stroke (to M.N.), the New York State Spinal Cord Injury Program, The
Dana Foundation and the Philip-Morris Organization.
COMPETING INTERESTS STATEMENT
The authors declare no competing financial interests.
Published online at http://www.nature.com/natureneuroscience
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Leao, A. Spreading depression of activity in cerebral cortex. J. Neurophysiol. 7,
359–390 (1944).
2. Martins-Ferreira, H., Nedergaard, M. & Nicholson, C. Perspectives on spreading
depression. Brain Res. Brain Res. Rev. 32, 215–234 (2000).
3. Tian, G.F. et al. An astrocytic basis of epilepsy. Nat. Med. 11, 973–981 (2005).
4. Takano, T. et al. Astrocyte-mediated control of cerebral blood flow. Nat. Neurosci. 9,
260–267 (2006).
5. Kasischke, K.A., Vishwasrao, H.D., Fisher, P.J., Zipfel, W.R. & Webb, W.W. Neural
activity triggers neuronal oxidative metabolism followed by astrocytic glycolysis. Science
305, 99–103 (2004).
6. Chance, B., Cohen, F., Jobsis, F. & Schoener, B. Intracellular oxidation-reduction states
in vivo. Science 137, 499–508 (1962).
7. Hadjikhani, N. et al. Mechanisms of migraine aura revealed by functional MRI in human
visual cortex. Proc. Natl. Acad. Sci. USA 98, 4687–4692 (2001).
8. Nedergaard, M. Mechanisms of brain damage in focal cerebral ischemia. Acta Neurol.
Scand. 77, 81–101 (1988).
761
© 2007 Nature Publishing Group http://www.nature.com/natureneuroscience
ARTICLES
9. Mayevsky, A., Zarchin, N. & Friedli, C.M. Factors affecting the oxygen balance in the
awake cerebral cortex exposed to spreading depression. Brain Res. 236, 93–105
(1982).
10. Mayevsky, A. & Weiss, H.R. Cerebral blood flow and oxygen consumption in cortical
spreading depression. J. Cereb. Blood Flow Metab. 11, 829–836 (1991).
11. Mayevsky, A., Lebourdais, S. & Chance, B. The interrelation between brain PO2 and
NADH oxidation-reduction state in the gerbil. J. Neurosci. Res. 5, 173–182 (1980).
12. Tomita, M. et al. Initial oligemia with capillary flow stop followed by hyperemia during
K+-induced cortical spreading depression in rats. J. Cereb. Blood Flow Metab. 25,
742–747 (2005).
13. Ayata, C. et al. Pronounced hypoperfusion during spreading depression in mouse cortex.
J. Cereb. Blood Flow Metab. 24, 1172–1182 (2004).
14. Hansen, A.J., Quistorff, B. & Gjedde, A. Relationship between local changes in cortical
blood flow and extracellular K+ during spreading depression. Acta Physiol. Scand. 109,
1–6 (1980).
15. Fabricius, M. & Lauritzen, M. Transient hyperemia succeeds oligemia in the wake of
cortical spreading depression. Brain Res. 602, 350–353 (1993).
16. Revsbech, N. An oxygen microsensor with a guard cathode. Limnol. Oceanogr. 34,
474–478 (1989).
17. Van Harreveld, A. & Fifkova, E. Glutamate release from the retina during spreading
depression. J. Neurobiol. 2, 13–29 (1970).
18. Hansen, A.J. Effect of anoxia on ion distribution in the brain. Physiol. Rev. 65, 101–148
(1985).
19. Hasbani, M.J., Schlief, M.L., Fisher, D.A. & Goldberg, M.P. Dendritic spines lost during
glutamate receptor activation reemerge at original sites of synaptic contact. J. Neurosci.
21, 2393–2403 (2001).
20. Feng, G. et al. Imaging neuronal subsets in transgenic mice expressing multiple spectral
variants of GFP. Neuron 28, 41–51 (2000).
21. Oliva, A.A., Jr, Jiang, M., Lam, T., Smith, K.L. & Swann, J.W. Novel hippocampal
interneuronal subtypes identified using transgenic mice that express green fluorescent
protein in GABAergic interneurons. J. Neurosci. 20, 3354–3368 (2000).
22. Takano, T. et al. Receptor-mediated glutamate release from volume sensitive channels in
astrocytes. Proc. Natl. Acad. Sci. USA 102, 16466–16471 (2005).
23. Capobianco, D.J. & Dodick, D.W. Diagnosis and treatment of cluster headache. Semin.
Neurol. 26, 242–259 (2006).
24. Mayevsky, A. & Sclarksy, D.L. Correlation of brain NADH redox state, K+, PO2 and
electrical activity during hypoxia, ischemia and spreading depression. Adv. Exp. Med.
Biol. 159, 129–141 (1983).
25. Schulte, M.L. & Hudetz, A.G. Functional hyperemic response in the rat visual cortex
under halothane anesthesia. Neurosci. Lett. 394, 63–68 (2006).
26. Hansen, A.J. & Nedergaard, M. Brain ion homeostasis in cerebral ischemia. Neurochem.
Pathol. 9, 195–209 (1988).
27. Iadecola, C. Neurovascular regulation in the normal brain and in Alzheimer’s disease.
Nat. Rev. Neurosci. 5, 347–360 (2004).
28. Van Harreveld, A. Changes in volume of cortical neuronal elements during asphyxiation.
Am. J. Physiol. 101, 233–242 (1957).
29. Somjen, G.G. Mechanisms of spreading depression and hypoxic spreading depressionlike depolarization. Physiol. Rev. 81, 1065–1096 (2001).
762
30. Nicholson, C., Kraig, R.P., ten Bruggencate, G., Stockle, H. & Steinberg, R. Potassium,
calcium, chloride and sodium changes in extracellular space during spreading depression in cerebellum [proceedings]. Arzneimittelforschung 28, 874–875 (1978).
31. Nedergaard, M. & Hansen, A.J. Characterization of cortical depolarizations evoked in
focal cerebral ischemia. J. Cereb. Blood Flow Metab. 13, 568–574 (1993).
32. Grafstein, B. Mechanism of spreading cortical depression. J. Neurophysiol. 19,
154–171 (1956).
33. Nedergaard, M. & Astrup, J. Infarct rim: effect of hyperglycemia on direct current
potential and [14C]2-deoxyglucose phosphorylation. J. Cereb. Blood Flow Metab. 6,
607–615 (1986).
34. Umegaki, M. et al. Peri-infarct depolarizations reveal penumbra-like conditions in
striatum. J. Neurosci. 25, 1387–1394 (2005).
35. Mayevsky, A. et al. Cortical spreading depression recorded from the human brain using a
multiparametric monitoring system. Brain Res. 740, 268–274 (1996).
36. Mayevsky, A., Manor, T., Meilin, S., Doron, A. & Ouaknine, G.E. Real-time multiparametric monitoring of the injured human cerebral cortex–a new approach. Acta
Neurochir. Suppl. 71, 78–81 (1998).
37. Fabricius, M. et al. Cortical spreading depression and peri-infarct depolarization in
acutely injured human cerebral cortex. Brain 129, 778–790 (2006).
38. Strong, A.J. & Dardis, R. Depolarisation phenomena in traumatic and ischaemic brain
injury. Adv. Tech. Stand. Neurosurg. 30, 3–49 (2005).
39. Gorji, A. Spreading depression: a review of the clinical relevance. Brain Res. Brain Res.
Rev. 38, 33–60 (2001).
40. Bolay, H. et al. Intrinsic brain activity triggers trigeminal meningeal afferents in a
migraine model. Nat. Med. 8, 136–142 (2002).
41. Meyer, J.S., Thornby, J., Crawford, K. & Rauch, G.M. Reversible cognitive decline
accompanies migraine and cluster headaches. Headache 40, 638–646 (2000).
42. Calandre, E.P., Bembibre, J., Arnedo, M.L. & Becerra, D. Cognitive disturbances and
regional cerebral blood flow abnormalities in migraine patients: their relationship with
the clinical manifestations of the illness. Cephalalgia 22, 291–302 (2002).
43. Waldie, K.E., Hausmann, M., Milne, B.J. & Poulton, R. Migraine and cognitive function:
a life-course study. Neurology 59, 904–908 (2002).
44. Riva, D. et al. Cognitive and behavioural effects of migraine in childhood and adolescence. Cephalalgia 26, 596–603 (2006).
45. McKendrick, A.M. & Badcock, D.R. Decreased visual field sensitivity measured
1 day, then 1 week, after migraine. Invest. Ophthalmol. Vis. Sci. 45, 1061–1070
(2004).
46. Shepherd, A.J. Colour vision in migraine: selective deficits for S-cone discriminations.
Cephalalgia 25, 412–423 (2005).
47. Akin, R., Unay, B., Sarici, S.U., Ulas, U. & Gokcay, E. Evaluation of visual evoked
potentials in children with headache. Turk. J. Pediatr. 47, 150–152 (2005).
48. Yenice, O. et al. Assessment of spatial-contrast function and short-wavelength sensitivity
deficits in patients with migraine. Eye 21, 218–223 (2007).
49. Ayata, C., Jin, H., Kudo, C., Dalkara, T. & Moskowitz, M.A. Suppression of
cortical spreading depression in migraine prophylaxis. Ann. Neurol. 59, 652–661
(2006).
50. Nimmerjahn, A., Kirchhoff, F., Kerr, J. & Helmchen, F. Sulforhodamine 101 as a specific
marker of astroglia in the neocortex in vivo. Nat. Methods 1, 31–37 (2004).
VOLUME 10
[
NUMBER 6
[
JUNE 2007 NATURE NEUROSCIENCE