Critical role of aquaporin-4 (AQP4) in astrocytic Ca2+
signaling events elicited by cerebral edema
Alexander S. Thranea,b,c,1, Phillip M. Rappolda,1, Takumi Fujitaa,1, Arnulfo Torresa, Lane K. Bekara, Takahiro Takanoa,
Weiguo Penga, Fushun Wanga, Vinita Rangroo Thranea,b,c, Rune Engerb,c, Nadia N. Haj-Yaseinb,c, Øivind Skared,e,
Torgeir Holenf, Arne Klunglandg, Ole P. Ottersenb,2, Maiken Nedergaarda,2, and Erlend A. Nagelhusa,b,c,h,2
aDivision of Glial Disease and Therapeutics, Center for Translational Neuromedicine, Department of Neurosurgery, University of Rochester Medical Center,
Rochester, NY 14642;bCentre for Molecular Biology and Neuroscience, Letten Centre, Institute of Basic Medical Sciences, University of Oslo, 0317 Oslo, Norway;
cCentre for Molecular Medicine Norway, Nordic European Molecular Biology Laboratory Partnership, University of Oslo, 0318 Oslo, Norway;dDivision of
Norway;fDepartment of Anatomy, Institute of Basic Medical Sciences, University of Oslo, 0317 Oslo, Norway;gCentre for Molecular Biology and Neuroscience,
Institute of Medical Microbiology, Oslo University Hospital, 0027 Oslo, Norway; andhDepartment of Neurology, Oslo University Hospital, 0027 Oslo, Norway
Edited* by Peter Agre, The Johns Hopkins Malaria Research Institute, Baltimore, MD, and approved December 6, 2010 (received for review October 14, 2010)
Aquaporin-4 (AQP4)is aprimaryinflux routeforwaterduringbrain
edema formation. Here, we provide evidence that brain swelling
triggers Ca2+signaling in astrocytes and that deletion of the Aqp4
gene markedly interferes with these events. Using in vivo two-
photon imaging, we show that hypoosmotic stress (20% reduction
in osmolarity) initiates astrocytic Ca2+spikes and that deletion of
Aqp4 reduces these signals. The Ca2+signals are partly dependent
on activation of P2 purinergic receptors, which was judged from
the effects of appropriate antagonists applied to cortical slices.
Supporting the involvement of purinergic signaling, osmotic stress
was found to induce ATP release from cultured astrocytes in an
AQP4-dependent manner. Our results suggest that AQP4 not only
serves as an influx route for water but also is critical for initiating
downstream signaling events that may affect and potentially ex-
acerbate the pathological outcome in clinical conditions associated
with brain edema.
rapidly swell in the initial phase of edema development (1). These
data are consistent with a number of studies suggesting that
aquaporin-4 (AQP4) serves as a primary influx route for water
from blood to brain (2, 3). AQP4 is strongly expressed in astro-
cytic endfeet (4), which form a continuous pericapillary sheath
that is interrupted only by a narrow extracellular space (5).
In vitro studies clearly show that swelling of astrocytes leads to
the activation of a number of signaling cascades (6, 7). Because
astrocytes are prone to swell in experimental conditions associ-
ated with edema formation (8), this raises the question of
whether the same signaling cascades are activated in early edema
formation in vivo and whether they affect the clinical outcome.
Hypoosmotic stress induces brain edema with early accumu-
lation of water in astrocytes (1, 9, 10). Thus, hypoosmotic stress
provides a suitable experimental model to explore potential
signaling mechanisms initiated by astrocytic swelling in vivo.
Here, we use optical imaging to show that hypoosmotic stress
induces Ca2+spikes in astrocytes in vivo and that these spikes
are potentiated in the presence of AQP4. We also provide in
vitro data indicating that the AQP4-dependent Ca2+signals are
mediated in part by autocrine purinergic signaling. Our findings
show that brain edema formation should not be seen merely as
a process of passive water accumulation in brain but as a condi-
tion that sets in motion specific signaling processes that may
significantly affect disease progression and morbidity.
sing in vivo two-photon imaging of mice subjected to hypo-
osmotic stress, we have shown previously that astrocytes
Aqp4 Deletion Reduces Swelling of Cortical Astrocytes Exposed to
Mild Hypoosmotic Stress. Immunofluorescence and Western
blots confirmed the efficacy of the Aqp4−/−KO strategy (Fig. 1 A
and B). Light microscopy revealed normal cytoarchitecture of
cortex in Aqp4−/−mice. Specifically, astrocytes, visualized by
GFAP immunolabeling, displayed normal morphology and intact
endfeet (Fig. 1B).
To further validate the Aqp4 deletion, we performed volu-
metric analysis of astrocytic somata in acute cortical slices ex-
posed to solution of reduced osmolarity (Fig. 1C). Astrocytes
were readily detected by two-photon imaging after the slice had
been incubated with the fluorescent dye Texas red hydrazide, an
approach similar to that described for sulphorhodamine 101 (11,
12). Dye loading of slices obtained from transgenic mice that
express EGFP under control of the Glt-1 promoter (Glt-1–EGFP
BAC transgenic mice) (13) confirmed that Texas red hydrazide
was selectively taken up by GFP-expressing astrocytes (Fig. 1D).
When exposed to 20% reduction in osmolarity, WT astrocytes
which wassignificantlyhigher than the 3 ± 0.8% increaseobserved
in Aqp4−/−astrocytes (P < 0.001, two-tailed Student t test) (Fig.
linear volume reduction, reflecting regulatory volume decrease.
To explore whether volume recovery was dependent on the
magnitude of osmotic stress, we exposed WT and Aqp4−/−
astrocytes to artificial cerebrospinal fluid (aCSF) with 30% re-
duction in osmolarity. Under this condition, astrocytes from both
genotypes showed continuous increase in soma volume, and
magnitude of swelling was similar (Fig. 1E).
Hypoosmotic Stress Enhances Ca2+Signals in Cortical Layer 1
Astrocytes in Vivo. In vivo two-photon imaging of Glt-1–EGFP
BAC transgenic mice confirmed that the Ca2+indicator Rhod2
AM was taken up by astrocytes (Fig. 2 A and B). Increase in
Rhod2 signal intensity was not associated with altered GFP
signal (Fig. 2B). The ratio between Rhod2 and GFP signal in-
tensities provided a more reliable measure of astrocytic Ca2+
signals than the Rhod2 signal (Fig. 2C), reflecting that the for-
mer measure is less sensitive to inadvertent small shifts in focal
plane. In neither WT nor Aqp4−/−mice did frequency of astro-
cytic Ca2+spikes change over time in the control state (0.092 ±
0.038 vs. 0.197 ± 0.065 for the first and last 15 min in WT, n = 76
Author contributions: A.S.T., P.M.R., T.F., A.T., L.K.B., T.T., M.N., and E.A.N. designed re-
search; A.S.T., P.M.R., T.F., A.T., L.K.B., W.P., F.W., R.E., N.N.H.-Y., and E.A.N. performed
research; L.K.B., T.T., T.H., A.K., O.P.O., M.N., and E.A.N. contributed new reagents/
analytic tools; A.S.T., P.M.R., T.F., V.R.T., Ø.S., and E.A.N. analyzed data; and A.S.T.,
P.M.R., O.P.O., M.N., and E.A.N. wrote the paper.
The authors declare no conflict of interest.
*This Direct Submission article had a prearranged editor.
Freely available online through the PNAS open access option.
1A.S.T., P.R., and T.F. contributed equally to this work.
2To whom correspondence may be addressed. E-mail: firstname.lastname@example.org, nedergaard@
urmc.rochester.edu, or email@example.com.
| January 11, 2011
| vol. 108
| no. 2www.pnas.org/cgi/doi/10.1073/pnas.1015217108
cells, P = 0.15; 0.179 ± 0.111 vs. 0.232 ± 0.076 in Aqp4−/−, n =
56, P = 0.55; mixed model analyses) (Methods).
Brain edema induced by i.p. water injection was associated
with enhanced Ca2+signals in WT mice (Fig. 2 D–G). Frequency
of astrocytic Ca2+spikes and percentage of astrocytes with Ca2+
spikes (active cells) increased already in the first 15-min period
post water injection and became even higher as brain swelling
progressed (Fig. 2G). Spike amplitude and proportion of long-
lasting (≥30 s) spikes also became higher in the late phase of
osmotic brain edema (Fig. 2 E and F).
Osmotically Induced Astrocytic Ca2+Spikes in Vivo Are Dependent on
AQP4.In the control state, frequency of astrocytic Ca2+spikes did
not differ between WT and Aqp4−/−mice (P = 0.69) (Fig. 2G).
In contrast to WT, Aqp4−/−mice did not respond to water in-
jection with altered astrocytic spike frequency (Fig. 2G). Only
the proportion of active astrocytes increased somewhat in the
late phase of osmotic brain swelling (Fig. 2G). At this stage,
proportion of active astrocytes (P = 0.0042) and spike frequency
(P = 0.0038) differed between WT and Aqp4−/−mice. In the first
15 min after water injection, these values did not differ signifi-
cantly between genotypes (P = 0.098 and P = 0.159, respectively;
mixed model analyses) (Methods), probably reflecting the time it
takes for brain swelling to develop after i.p. water injection.
Mild Hypoosmotic Stress Induces AQP4-Dependent Astrocytic Ca2+
Responses in Acute Cortical Slices. Supporting our in vivo find-
ings, we found that exposing acute cortical slices to mild hypo-
osmotic aCSF (20% reduction in osmolarity) robustly triggered
astrocytic Ca2+spikes in WT mice (Fig. 3 A and B). The pro-
portion of astrocytes that responded with Ca2+spikes was much
lower in Aqp4−/−than in WT mice (Fig. 3 A and B). Moreover,
Ca2+spikes in Aqp4−/−mice had a lower amplitude (Fig. 3B)
and delayed onset (272 ± 9 s in Aqp4−/−vs. 162 ± 4 s in WT, P <
0.001, two-tailed Student t test). More severe osmotic stress
(30% reduction in osmolarity) diminished the difference in re-
sponder rate between WT and Aqp4−/−mice (Fig. 3B), possibly
reflecting robust astrocyte swelling in both genotypes during this
condition (compare with Fig. 1E). However, at 30% reduction in
osmolarity, the spike amplitude was still lower in Aqp4−/−than in
WT mice (Fig. 3B).
sites upstream and downstream of exons 1–3. Floxed mice were bred with Cre-expressing mice to produce mice with the Aqp4 KO allele. Western blot
confirmed the absence of AQP4 in Aqp4−/−mice. (B) Immunofluorescence micrographs of mouse cortex probed with primary antibodies against AQP4 (green)
and GFAP (red) with DAPI-labeled nuclei (blue) for orientation. The AQP4 immunofluorescence signal is absent in Aqp4−/−mice. Insets display perivascular
AQP4 and GFAP labeling at higher magnification. (Scale bar: 25 μm; Inset, 5 μm.) (C) Experimental design for validating the effect of Aqp4 deletion on os-
motically induced astrocyte swelling. Acute brain slices were prepared from WT and Aqp4−/−mouse pups. Slices were loaded with Texas red hydrazide and
perfused with aCSF with normal (isotonic) or reduced osmolarity (−20% or −30% Osm). Sectional images were acquired for 3D volume analysis. (D) Two-
photon imaging of Texas red hydrazide-loaded slices obtained from mice expressing GFP in astrocytes (Glt-1–EGFP BAC transgenic mice) confirmed that the
dye was selectively taken up by astrocytes. (Scale bar: 5 μm.) (E) Exposure of acute cortical slices to 20% reduction in osmolarity (−20% Osm) induced more
prominent swelling of astrocytic somata in WT (n = 37) than in Aqp4−/−mice (n = 26; P < 0.001 at 5 min, two-tailed Student t test). The initial swelling was
followed by shrinkage reflecting regulatory volume decrease. More severe osmotic stress (−30% Osm) induced continuous swelling in both genotypes (n = 30
and 31). Error bars represent SEM. Lower shows representative images of astrocytes exposed to −20% Osm. The red ring marks the astrocyte soma cir-
cumference at baseline. (Scale bar: 5 μm.)
Aqp4 KO strategy and validation. (A) The targeted allele contained a flippase recognition target (FRT)-neomycin-FRT cassette after exon 3 and LoxP
Thrane et al.PNAS
| January 11, 2011
| vol. 108
| no. 2
AQP4-Dependent Ca2+Signals Are Mediated in Part by Autocrine
Purinergic Signaling. Incubating WT slices with the nonselective
P2 antagonists suramin and pyridoxal-phosphate-6-azophenyl-
2′,4′-disulfonate (PPADS) delayed onset of Ca2+responses to
mild hypoosmotic stress (327 ± 8 s vs. 241 ± 9 s without antag-
onists, P < 0.001, two-tailed Student t test) and reduced per-
centage of responding astrocytes (Fig. 3B). Microinjection of
ATP in cortical slices induced astrocytic Ca2+responses with
similar kinetics in WT and Aqp4−/−mice (Fig. 3B Lower) (mean
amplitude = 68 ± 2% vs. 69 ± 3%; time to response = 11 ± 1.2 s
vs. 12 ± 1.2 s, respectively), suggesting that Aqp4 deletion did not
interfere with signaling mechanisms downstream of purinergic
Aqp4 Deletion Abrogates Osmotically Induced ATP Release from
Cultured Astrocytes. Cultured WT astrocytes exposed to hypo-
osmotic medium (−20% Osm) for 15 min released more ATP
than those kept in isotonic solution (Fig. 3C). In contrast, cul-
tured Aqp4−/−astrocytes subjected to similar stress showed no
significant change in ATP release (Fig. 3C). Taken together, our
data are compatible with a role for AQP4 in amplifying signaling
events triggered by cell swelling (Fig. 3D).
Brain edema formation is commonly regarded as a passive pro-
cess by which water accumulates in the brain because of changes
in osmotic driving forces or perturbations of the blood–brain
barrier. Hence, the symptomatology and treatment of this seri-
ous condition is usually discussed in the context of the accom-
panying intracranial pressure changes that may eventually cause
herniation and compromise blood flow to the brain. The possi-
bility that brain edema formation sets in motion pathophysio-
overlooked. This is remarkable given the fact that cell swelling—
the hallmark of cytotoxic edema—is known to activate a number
of signaling cascades that may have profound effects on cell
function (14). Specifically, in vitro studies have shown that
swelling causes astrocytes to release neuroactive substances such
as glutamate and ATP (6, 15).
This study shows that incipient edema is associated with
astrocytic Ca2+signals in vivo. These signals are causally linked
to water influx and cell swelling, because they were significantly
reduced in animals deficient in AQP4. Previously, deletion of
Aqp4 has been shown to abrogate osmotically induced astrocytic
swelling and counteract build-up of brain edema (8, 16).
processes haslargely been
detected in anesthetized Glt-1–EGFP BAC transgenic mice after loading of Rhod2 AM onto the cortical surface. (B) Images show Rhod2 fluorescence in a GFP-
expressing astrocyte. Because increase in Rhod2 signals were not associated with changes in GFP fluorescence, the ratio between the two signals provided
a measure of the astrocytic Ca2+signal. (C) During brain swelling, the ratio between Rhod2 and GFP fluorescent signal intensities (lower trace) was less
sensitive to inadvertent small shifts in focal plane than Rhod2 signals (upper trace). Thus, Rhod2/GFP signals were used for reliably defining Ca2+spikes [i.e.,
transients exceeding 20% (dashed line) of baseline]. (D) Pooled Ca2+traces (measured as relative changes in Rhod2/GFP) for all astrocytes (n = 24) within an
image field in a WT mouse subjected to i.p. water injection (indicated by arrow; 200 mL/kg) to induce osmotic brain swelling. Note increase in spike frequency
and amplitude as brain edema develops. (E) In WT mice, relative spike amplitude and proportion of spikes lasting ≥30 s were higher in the last 15 min than in
the first 15 min after water injection. (F) Traces of relative Ca2+changes in WT astrocytes 30 min after i.p. water injection. The position of the respective
astrocytes (confluent Rhod2 and GFP signals in yellow) is indicated in Left. FITC-dextran (green) was injected i.v. at the beginning of the experiment to outline
the vasculature, and it confirmed vascular perfusion. Time-lapse sequence of the Ca2+responses is shown in Right. Note the intense and long-lasting Ca2+
surge in the astrocytic soma and endfoot surrounding a vein. (Scale bar: 25 μm.) (G) In WT mice, the frequency of astrocytic Ca2+spikes was higher already in
the first 15 min after water injection compared with the control state, and it was even higher in the last 15 min of observation. In Aqp4−/−mice, water
injection did not increase spike frequency. The percentage of astrocytes with more than or equal to one Ca2+spike(s) per 15-min observation (active
astrocytes) increased profoundly in WT mice after water injection. In Aqp4−/−mice, the number of active astrocytes increased only at the late phase of osmotic
brain swelling. Error bars represent SEM. Mixed model analyses were performed using a binomial (Bernoulli) model with logit link for binary observations
(passive or active cells), a Poisson model for count data (spike frequency), and a linear model for spike amplitudes (24).
In vivo two-photon imaging of astroglial Ca2+signals during hypoosmotic stress. (A) Diagram of experimental setup. Astrocytic Ca2+transients were
| www.pnas.org/cgi/doi/10.1073/pnas.1015217108 Thrane et al.
The in vivo analyses were complemented with monitoring of
blood flow in the microvascular bed. Despite absence of overt
changes, it is difficult to rule out small alterations in cerebral
perfusion caused by the incipient brain edema. Thus, it was
deemed necessary to include complementary in vitro studies in
slices. Such studies also allowed us to dissect the mechanisms
underlying the AQP4-sensitive Ca2+responses. Analyses in
acute cortical slices supported the data obtained in vivo. Notably,
slices exposed to hypoosmotic media displayed Ca2+signals in
astrocytes reminiscent of those seen in vivo. These signals were
attenuated after Aqp4 deletion. The attenuation was particularly
pronounced at 20% decrease in osmolarity.
Previous in vitro studies have shown that activation of puri-
nergic receptors triggers astrocytic Ca2+transients (17–19). We
hypothesized that Ca2+signals elicited during edema formation
depend—at least in part—on ATP release from swollen astro-
cytes. Application of P2 antagonists to acute cortical slices sup-
ported this view. The quantitative analysis indicated that, in
∼25% of WT astrocytes, the Ca2+response was contingent on
ATP signaling. Obviously, additional mechanisms are at play and
contribute to the observed Ca2+signals. Stretch-sensitive recep-
tors are likely to be among these mechanisms.
Next, we set out to resolve whether astrocytes could serve as
a source of ATP. In cultured astrocytes, osmotic stress induced
ATP release, and this release was abolished after Aqp4 deletion.
Taken together, the data suggest that AQP4 not only mediates
water influx but also is essential for initiating signaling events
associated with edema formation. This may explain the rather
pronounced protective effect of Aqp4 KO or AQP4 mislocaliza-
tion in stroke models (8, 20). It is well-known that water passes
through the lipid bilayer of the plasma membrane (although to
a limited extent compared with the water flux through aquaporin
channels) and that diffusion also occurs through the thin slits that
separate the astrocyte endfeet. In AQP4-deficient mice, a dis-
crepancy between the extent of water transport restriction on the
onehandandthe protective effectin stroke on theother handcan
easily be explained if loss of AQP4 also interferes with signaling
mechanisms that exacerbate the pathological outcome.
Edema formation and cytotoxicity likely engage in a vicious
cycle, where cell swelling causes release of cytotoxic compounds
that, in turn, lead to tissue damage and more swelling. ATP is
known to act as a cytotoxic compound in stroke, as judged by
a number of in vitro and in vivo studies (21). Thus, an early-stage
intervention with AQP4 inhibitors would interfere with this vi-
cious cycle by counteracting not only the swelling per se but also
deleterious secondary events like ATP release. Further studies
are required to resolve whether AQP4 is also involved in swelling-
activated glutamate efflux through volume-sensitive channels (6).
An obvious question is whether the effect of Aqp4 deletion
solely depends on the change in swelling response or whether
AQP4 (alone or in combination with other molecules) serves as
an osmosensor upstream of the above signaling events. To dis-
tinguish between these possibilities, we exposed acute slices to an
osmotic stress (30% reduction in osmolarity) severe enough to
override the mechanisms that normally limit transmembrane
water transport, as evidenced by the reduced sensitivity to Aqp4
deletion. With an osmotic stress at this scale, the percentage of
astrocytes that responded with Ca2+spikes was nearly as high in
the Aqp4−/−animals as in WT. This observation is consistent with
the idea that the Ca2+signals are elicited by the AQP4-induced
swelling response rather than through an osmoreceptor response.
Our study has revealed that induction of brain edema sets in
motion specific signaling events in brain cells. Notably, we have
shown by in vivo two-photon imaging that osmotic stress and
edema formation are associated with brisk Ca2+signals in cor-
tical astrocytes. This observation prompted us to resolve whether
these signals are dependent on AQP4, which is assumed to
constitute the main influx route for water at the brain–blood
interface. Using a Aqp4−/−line, we show that deletion of Aqp4
interferes with the frequency and amplitude as well as the du-
ration of the Ca2+signals observed. Taken together with com-
plementary analyses in reduced experimental models, our data
are consistent with the idea that AQP4-mediated cell swelling is
inextricably coupled with activation of signaling pathways that
from WT (Upper) and Aqp4−/−(Lower) mice. Traces shown are from astrocytes marked in Left. Exposure of slices from WT mice to aCSF with 20% reduction in
osmolarity (−20% Osm) induced brisk astrocytic Ca2+spikes. In contrast, this osmotic stress failed to elicit Ca2+spikes in most AQP4-deficient astrocytes. (B)
Quantitative analysis of astrocytic Ca2+responses to osmotic stress. Deletion of Aqp4 or blocking P2 purinergic receptors with PPADS/suramin significantly
reduced the number of astrocytes that responded with Ca2+spikes during exposure to −20% Osm. When more severe hypoosmotic stress (−30% Osm) was
applied, a larger fraction of the Aqp4−/−astrocytes responded. The amplitude of the Ca2+spikes differed between the genotypes for both types of stress.
Image shows intense Ca2+signals in Rhod2-loaded astrocytes after microinjection of ATP and FITC-dextran (green; to verify injection) into the slice. Repre-
sentative traces from WT and Aqp4−/−mice are shown. (C) Cultured astrocytes exposed to hypoosmotic media (−20% Osm, 15 min) released significantly more
ATP than those kept in isotonic media. Astrocytes from Aqp4−/−mice did not show osmotically induced ATP release. P values were obtained by two-tailed
Student t test. Error bars represent SEM. (Scale bar: 25 μm.) (D) Diagram showing proposed involvement of AQP4 in astrocyte signaling cascades during
hypoosmotic stress. AQP4-mediated water influx triggers Ca2+transients, partly by promoting release of ATP and activation of P2 purinergic receptors.
Osmotically induced astrocytic Ca2+responses and ATP release in vitro. (A) Two-photon images of Rhod2 AM-loaded acute cortical slices obtained
Thrane et al. PNAS
| January 11, 2011
| vol. 108
| no. 2
may profoundly affect the pathological and pathophysiological
outcome in clinical conditions associated with brain edema.
Mice. Aqp4−/−mice were generated by GenOway by cloning and sequencing
of a targeted region of the murine Aqp4 gene in a 129/Sv genetic back-
ground. The strategy was to design a targeted locus allowing us to delete
exons 1–3 to avoid any expression of putative splice variants. Hence, a flip-
pase recognition target (FRT)-neomycin-FRT-LoxP–validated cassette was
inserted downstream of exon 3, and a LoxP site was inserted upstream of
exon 1 as depicted in Fig. 1A. After homologous recombination in ES cells,
ES-cell injection into blastocytes, and generation of chimeras, heterozygous
floxed mice were obtained by breeding chimeras with C57BL/6J females.
Heterozygous floxed mice were bred with C57BL/6J Cre expressing mice to
generate mice heterozygous for the KO allele, Aqp4+/−. The Aqp4+/−mice
were then backcrossed with C57BL/6J mice for five generations before
intercrossing to yield Aqp4−/−and Aqp4+/+(WTs). For acute cortical slice
experiments, we also used C57BL/6J pups from Jackson Laboratory as WT
controls. For in vivo experiments, we used Aqp4−/−and WT mice expressing
EGFP in astrocytes. These mice were generated by breeding Aqp4−/−and WT
mice with BAC promoter reporter transgenic mice that express EGFP under
the control of the natural Glt-1 promoter (13). The latter mice were provided
by J. D. Rothstein (Johns Hopkins University, Baltimore, MD).
Western Blot and Immunohistochemistry. After homogenization and solubi-
lization, extracts of Aqp4−/−and WT brains were loaded onto a 10% SDS/
PAGE gel and subsequently transferred onto 0.2-μm poly(vinylidene difluor-
ide) membrane (Bio-Rad). The membrane was probed with 0.02 μg/mL goat
anti-AQP4 (Cat# sc-9888; Santa Cruz Biotechnology), developed using alka-
line phosphatase substrate (ECF Western blotting reagents; Amersham
Pharmacia), and visualized with a Typhoon Variable Mode Imager (Amer-
sham Pharmacia). Fixation of mice, preparation of tissue slices, and immu-
nohistochemistry were performed as described previously (22). We used
a monoclonal antibody against GFAP (G3893, 1:100; Sigma) and a polyclonal
antibody against AQP4 (AB3068, 1:100; Chemicon International).
Animal Preparation for in Vivo Imaging. Eight- to twelve-wk-old WT and
Aqp4−/−mice expressing EGFP in astrocytes were anesthetized with ure-
thane and α-chloralose (1 g/kg and 50 mg/kg i.p., respectively), intubated,
and artificially ventilated with room air using a small animal ventilator
(SAR830; CWE) set to ∼100 breaths/min with a tidal volume of 0.3–0.4 mL.
Body temperature was kept at 37 °C by a temperature-controlled heating
blanket. A craniotomy (3 mm in diameter) was made over the cortex 1 mm
lateral and 0.5 mm posterior to the bregma, and the dura was removed. The
Ca2+indicator Rhod2 AM (2 mM; Invitrogen) was loaded to the exposed
cortex for ∼50 min. After washing for 10 min with aCSF, the craniotomy was
covered with 1% agarose in aCSF and sealed by a coverslip. The femoral
artery was cannulated for continuous monitoring of mean arterial blood
pressure and analysis of blood gases. Only mice with blood gases within the
physiological range (pO2= 80–150 mmHg, pCO2= 30–45 mmHg, pH 7.25–7.5)
were included. To outline the vasculature, we administered FITC-dextran
(2,000 kDa, ∼0.4 mL, 2.5% in saline; Sigma) intravenously. WT and Aqp4−/−
mice (n = 6 for each genotype) were injected with distilled water (200 mL/kg)
i.p. immediately before imaging. Mice not receiving water injections (n = 3
for each genotype) served as controls.
In Vivo Two-Photon Imaging. A Mai Tai laser (SpectraPhysics) attached to
a confocal scanning system (Fluoview 300; Olympus) and an upright micro-
scope (IX51W; Olympus) was used for in vivo imaging, as previously described
(23). A 20× (0.9 NA) water immersion lens with 3× additional zoom was used
to image astrocytes in cortical layer 1. Excitation wavelength was 860 nm,
and emission was collected at 575–645 nm. Dual-channel (Rhod2 and GFP)
images with 512 × 512-pixel frames were acquired in a region with intact
capillary perfusion every 5 s for 60 min or until capillary perfusion stopped.
The low sampling rate was used to avoid photo damage. The two-photon
laser power was adjusted to a power that was less than 40 mW at the
sample. Two of six WT and one of six Aqp4−/−mice subjected to water in-
jection died before 60 min (at 30, 48, and 55 min, respectively). During brain
swelling, the focus was adjusted for persistent imaging of the same area. A
Ca2+spike was defined when the relative ratio between the Rhod2 and GFP
signal intensities exceeded 20% of baseline over a 10-min recording period.
Frequency of spikes was calculated per astrocyte per 15 min. Because of the
low frequency of Ca2+spikes in the control state, we averaged data for two
consecutive 15-min periods. Active astrocytes were defined as cells that had
more than or equal to one Ca2+spike(s) per 15 min. Values are expressed as
mean ± SEM. Mixed model analyses were performed using a binomial
(Bernoulli) model with logit link for binary observations (passive or active
cells), a Poisson model for count data (spike frequency), and a linear model
for spike amplitudes (24). A hierarchical structure was assumed for all these
models, with random effects representing variation between animals and
cells inside animals. All experiments were approved by the Institution Ani-
mal Care and Use Committee of University of Rochester.
Preparation of Acute Cortical Slices and Dye Loading. Coronal cortical slices
were prepared from 10- to 20-d-old mice of either sex as described previously
(22, 25). In brief, the brains were submerged in gassed (95% O2and 5% CO2)
ice-cold cutting solution, and coronal slices (400 μm) were cut on a Vibra-
tome (TPI). Slices were incubated for 20 min in aCSF and then loaded with
either Rhod2 AM (2 mM) or Texas red hydrazide (1.5 μM) in aCSF at 35 °C for
50 min. The aCSF, gassed as described above, contained (in mM) 126 NaCl,
2.5 KCl, 1.25 NaH2PO4, 2 MgCl2, 2 CaCl2, 10 glucose, and 26 NaHCO3(pH 7.4).
Two-Photon Imaging in Acute Cortical Slices. Dye-loaded slices were trans-
ferred to a recording chamber (1.5 mL), held in place by a nylon-threaded
used for imaging (excitation wavelength = 820–860 nm; emission collected at
575–645 nm). A 60× (0.9 NA) water immersion lens with 6–10× additional
zoom was used for the volume assessment, whereas a 20× (0.9 NA) water
immersion lens was used for calcium imaging. All experiments were per-
formed at room temperature. After imaging in normal aCSF, we switched
the perfusion solution to hypoosmotic aCSF. This solution differed from the
control aCSF only with respect to the NaCl concentration, which was either
reduced by 20% or 30% (NaCl = 100.8 or 88.2 mM, respectively). Osmolari-
ties were verified by freezing-point depression.
Volumetry. Astrocytes with endfeet visibly extending to vessels at a tissue
depth >40 μm were selected, and 3D image stacks were collected; frame sizes
of 256 × 256 at intervals of 1.5–3.0 μm in the z direction were collected with
an acquisition time of <20 s using minimal laser power (<40 mW). Images
were acquired for a 60-min period and then analyzed for changes in soma
volume using custom-made software (Matlab Inc.). A median filter with
a radius of 5 pixels was used to reduce background noise. A region of interest
was defined around the soma in maximum intensity projections using fixed
landmarks for all time points within an experiment. Pixels over a certain
threshold within this region were counted for each xy frame in the z stack.
Laser power, photomultiplier tube sensitivity, and thresholds were kept
constant for each image sequence. Automated thresholding was performed
by normalizing pixel intensity to a decay constant extracted from an intensity
histogram of each image. The sum of pixels for each xyz stack was then
compared with the sum at baseline. Values are expressed as mean ± SEM.
Intragroup and intergroup analyses were performed using a two-tailed Stu-
dent t test.
Calcium imaging. Time-lapse images of astrocytic Ca2+signals were recorded in
the slices every 5 s. Images were acquired 1 min before the solution change as
a baseline and 10 min after; 512 × 512 frames were acquired every 5 s using
minimal laser power (<40 mW). Changes in Rhod2 intensity (ΔF/F0) were an-
alyzed using custom-made software (Matlab Inc.) within a manually defined
region of interest (10-μm-diameter circle for soma). Relative Rhod2 increases
>20% (or >2 SDs) above baseline value were defined as a Ca2+spike. After
administration of ATP, the Ca2+response was imaged for 168 s, one frame per
and intergroup analyses were performed using a two-tailed Student t test.
Drug delivery. For experiments with P2 purinergic receptor antagonists, the
slices were exposed to suramin (100 μM; Tocris Bioscience) and PPADS (30 μM;
Tocris Bioscience) >30 min before imaging and hypoosmotic challenge. Slices
exposed to microinjected ATP were perfused with control aCSF. A fine
electrode filled with aCSF containing 500 μm ATP was inserted 40–80 μm into
the slice. After a 33.6-s baseline recording, ATP was puffed through the
electrode using a Picospritzer (10 psi, 100 ms; Parker Instrumentation); 1%
FITC was used to visualize the puff.
ATP Release from Cultured Astrocytes. Serum-free media (control) or solution
with 20% reduction in osmolarity (by removing NaCl) was added (400 μL; 2×)
to astrocyte cultures derived from WT or Aqp4−/−mice. After 15-min expo-
sure, 350-μL samples were collected (450 μL remained in the wells). ATP
content in samples collected from cultures grown in 24-well tissue culture
plates was measured by using a bioluminescent ATP assay kit (Sigma) and
a Victor2 plate reader (Wallac) and was normalized to the cell number (26).
A total of eight plates were examined for each genotype and their re-
| www.pnas.org/cgi/doi/10.1073/pnas.1015217108Thrane et al.
spective control. Values are expressed as mean ± SEM. Intergroup analyses
were done using a two-tailed Student t test.
ACKNOWLEDGMENTS. We thank Mr. Justin Chang (University of Rochester
Medical Center, Rochester, New York) for making the software used for
analyzing astrocytic soma volumes and Ca2+transients, Professor Petter
Laake (Institute of Basic Medical Sciences, University of Oslo, Oslo, Norway)
for advice on the statistical analysis, and Professor Jeffrey D. Rothstein
(Johns Hopkins University, Baltimore) for providing Glt-1–EGFP BAC trans-
genic mice. This work was supported by the US National Institutes of
Health Grants P01NS050315 and R01NS056188 (to M.N.), the Research
Council of Norway, and the Letten Foundation.
1. Nase G, Helm PJ, Enger R, Ottersen OP (2008) Water entry into astrocytes during brain
edema formation. Glia 56:895–902.
2. Amiry-Moghaddam M, Ottersen OP (2003) The molecular basis of water transport in
the brain. Nat Rev Neurosci 4:991–1001.
3. Verkman AS (2009) Knock-out models reveal new aquaporin functions. Handb Exp
4. Nielsen S, et al. (1997) Specialized membrane domains for water transport in glial cells:
High-resolution immunogold cytochemistry of aquaporin-4 in rat brain. J Neurosci 17:
5. Mathiisen TM, Lehre KP, Danbolt NC, Ottersen OP (2010) The perivascular astroglial
sheath provides a complete covering of the brain microvessels: An electron
microscopic 3D reconstruction. Glia 58:1094–1103.
6. Mulligan SJ, MacVicar BA (2006) VRACs CARVe a path for novel mechanisms of
communication in the CNS. Sci STKE 2006:pe42.
7. Kimelberg HK (2005) Astrocytic swelling in cerebral ischemia as a possible cause of
injury and target for therapy. Glia 50:389–397.
8. Manley GT, et al. (2000) Aquaporin-4 deletion in mice reduces brain edema after
acute water intoxication and ischemic stroke. Nat Med 6:159–163.
9. Nagelhus EA, Lehmann A, Ottersen OP (1993) Neuronal-glial exchange of taurine
during hypo-osmotic stress: A combined immunocytochemical and biochemical
analysis in rat cerebellar cortex. Neuroscience 54:615–631.
10. Risher WC, Andrew RD, Kirov SA (2009) Real-time passive volume responses of
astrocytes to acute osmotic and ischemic stress in cortical slices and in vivo revealed by
two-photon microscopy. Glia 57:207–221.
11. Nimmerjahn A, Kirchhoff F, Kerr JN, Helmchen F (2004) Sulforhodamine 101 as
a specific marker of astroglia in the neocortex in vivo. Nat Methods 1:31–37.
12. Kang J, et al. (2010) Sulforhodamine 101 induces long-term potentiation of intrinsic
excitability andsynaptic efficacyin
13. Regan MR, et al. (2007) Variations in promoter activity reveal a differential expression
and physiology of glutamate transporters by glia in the developing and mature CNS.
J Neurosci 27:6607–6619.
14. Hoffmann EK, Lambert IH, Pedersen SF (2009) Physiology of cell volume regulation in
vertebrates. Physiol Rev 89:193–277.
15. Kimelberg HK, Macvicar BA, Sontheimer H (2006) Anion channels in astrocytes:
Biophysics, pharmacology, and function. Glia 54:747–757.
16. Papadopoulos MC, Verkman AS (2005) Aquaporin-4 gene disruption in mice reduces
brain swelling and mortality in pneumococcal meningitis. J Biol Chem 280:13906–13912.
17. Salter MW, Hicks JL (1994) ATP-evoked increases in intracellular calcium in neurons
and glia from the dorsal spinal cord. J Neurosci 14:1563–1575.
18. Shao Y, McCarthy KD (1995) Receptor-mediated calcium signals in astroglia: multi-
ple receptors, common stores and all-or-nothing responses. Cell Calcium 17:187–
19. Wang X, Ge S, Crooks GM (2009) Fluorescent immunohistochemistry and in situ
hybridization analysis of pancreas. Methods Mol Biol 560:191–201.
20. Amiry-Moghaddam M, et al. (2003) Alpha-syntrophin deletion removes the
perivascular but not endothelial pool of aquaporin-4 at the blood-brain barrier and
delays the development of brain edema in an experimental model of acute
hyponatremia. Proc Natl Acad Sci USA 100:2106–2111.
21. Rossi DJ, Brady JD, Mohr C (2007) Astrocyte metabolism and signaling during brain
ischemia. Nat Neurosci 10:1377–1386.
22. Simard M, Arcuino G, Takano T, Liu QS, Nedergaard M (2003) Signaling at the
gliovascular interface. J Neurosci 23:9254–9262.
23. Wang X, et al. (2006) Astrocytic Ca2+ signaling evoked by sensory stimulation in vivo.
Nat Neurosci 9:816–823.
24. Cnaan A, Laird NM, Slasor P (1997) Using the general linear mixed model to analyse
unbalanced repeated measures and longitudinal data. Stat Med 16:2349–2380.
25. Kang J, Nedergaard M (2000) Imaging astrocytes in acute brain slices. Imaging
Neurons: A Laboratory Manual, eds Yuste R, Lanni F, Konnerth A (Cold Spring Harbor
Laboratory Press, Plainview, NY), pp 1–42.
26. Cotrina ML, et al. (1998) Connexins regulate calcium signaling by controlling ATP
release. Proc Natl Acad Sci USA 95:15735–15740.
Thrane et al.PNAS
| January 11, 2011
| vol. 108
| no. 2