Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844

& 2008 ISCBFM All rights reserved 0271-678X/08 $30.00

www.jcbfm.com

Metabolic challenge to glia activates an

adenosine-mediated safety mechanism that
promotes neuronal survival by delaying the
onset of spreading depression waves
Santiago Canals1,2, Belén Larrosa3, Jesús Pintor4, Marı́a A Mena5 and Oscar Herreras3
1
Instituto de Investigaciones Biomédicas ‘Alberto Sols’, CSIC-UAM, Madrid, Spain; 2Max Planck Institute
for Biological Cybernetics, Tübingen, Germany; 3Cajal Institute, CSIC, Madrid, Spain; 4School of Optics,
University Complutense, Madrid, Spain; 5Department of Investigación, Hospital ‘Ramón y Cajal’,
Madrid, Spain

In a model of glial-specific chemical anoxia, we have examined how astrocytes influence both
synaptic transmission and the viability of hippocampal pyramidal neurons. This relationship was
assessed using electrophysiological, pharmacological, and biochemical techniques in rat slices and
cell cultures, and oxidative metabolism was selectively impaired in glial cells by exposure to the
mitochondrial gliotoxin, fluoroacetate. We found that synaptic transmission was blocked shortly
after inducing glial metabolic stress and peri-infarct-like spreading depression (SD) waves
developed within 1 to 2 h of treatment. Neuronal electrogenesis was not affected until SD waves
developed, thereafter decaying irreversibly. The blockage of synaptic transmission was totally
reversed by A1 adenosine receptor antagonists, unlike the development of SD waves, which
appeared earlier under these conditions. Such blockage led to a marked reduction in the electrical
viability of pyramidal neurons 1 h after gliotoxin treatment. Cell culture experiments confirmed that
astrocytes indeed release adenosine. We interpret this early glial response as a novel safety
mechanism that allocates metabolic resources to vital processes when the glia itself sense an
energy shortage, thereby delaying or preventing entry into massive lethal ischemic-like depolariza-
tion. The implication of these results on the functional recovery of the penumbra regions after
ischemic insults is discussed.
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844; doi:10.1038/jcbfm.2008.71; published online 9 July 2008

Keywords: adenosine; glia; ischemia; penumbra; spreading depression

Introduction scenario in stroke, anoxia, hypoglycemia, and con-

Neuronal activity and viability rely on the contin-
uous supply of metabolic substrates by the vascular
system. It is now well known that an imbalance
between the rate of energy production and utiliza-
tion causes the collapse of ionic gradients and leads
to massive cell depolarization. This is a common
Correspondence: Dr S Canals, Instituto de Investigaciones
Biomedicas ‘Alberto Sols’, CSIC-UAM, Arturo Duperier 4, Madrid
28029, Spain.
E-mail: scanals@iib.uam.es
Funding: Spanish Ministry of Education and Science (BFU 2005/
8917, BFU2006-28426-E/BFI); Comunidad Autónoma de Madrid
(S-SEM-0255-2006); Consejo Superior de Investigaciones Cientı́-
ficas (I3/200620I186); SC is supported by the Human Frontier
Science Program.
Received 15 April 2008; revised and accepted 16 June 2008;
published online 9 July 2008
cussion, which initiate the signaling cascades lead-
ing to loss of function and cell death (Somjen et al,
1990). Also well known is the tight metabolic
coupling between neurons and glial cells (reviewed
in Araque et al, 2001), which, in the context of
energy shortage, is indicated by the disruption of
tissue homeostasis in parallel with the gradual
decline of the electrical activity and irreversible
injury to neurons on selective blockage of the glial
Krebs cycle (Largo et al, 1996). In this process, a
rapid loss of excitatory synaptic transmission is
followed by the spontaneous development of peri-
infarct-like waves of spreading depression (SD)
(Hossman, 1996; Canals et al, 2005b). Although SD
has little impact in normal tissue (Herreras and
Somjen, 1993), it kills neurons in brain regions
deprived of glial activity according to a spatiotem-
poral pattern, which closely reproduces the
 Glial adenosine and energy shortage
S Canals et al
1836
sequence of events in the penumbra of an ischemic
focus (Fabricius et al, 2006). These results under-
score the fundamental role of astrocytes in main-
taining normal physiologic activity. However, less is
known about their active participation in the
protective mechanisms activated during energy
imbalance (for a review see Ames, 2000).
One such protective mechanism involves the
accumulation of adenosine in the extracellular
space (Dunwiddie and Masino, 2001). Several
studies have shown that during periods of energy
shortage, such as ischemia, hypoxia, or hypoglyce-
mia, the concentration of interstitial adenosine
increases several folds and it reduces neuronal
activity (Pull and McIlwain, 1972; Gribkoff et al,
1990; Frenguelli et al, 2007). Adenosine may be an
ideal molecule to fight against the functional
collapse of neurons as it can block synaptic
transmission by decreasing neurotransmitter release
and hyperpolarizing the postsynaptic membrane
(Greene and Haas, 1991), while increasing the local
energy supply by its potent vasodilatory activity
(Phillis, 2004). Therefore, not surprisingly, blockage
of the adenosine A1 receptor during an ischemic
episode restores synaptic transmission, precipitates
the occurrence of anoxic depolarization, and
increases the infarct size (Fowler, 1989; Rudolphi
et al, 1992). Adenosine-mediated neuroprotection
also occurs in experimental models in which
cellular respiration has been pharmacologically
depressed (chemical anoxia; Ilie et al, 2006).
In this study, we have examined the relationship
between neurons and astrocytes and the effects of
adenosine released under conditions of metabolic
stress in rat slices and cell cultures. This issue was
investigated using electrophysiological and pharma-
cological techniques in a model of glial-specific
chemical anoxia in which fluoroacetate (FAc), a
mitochondrial gliotoxin, selectively impaired the
oxidative metabolism of glial cells. Fluoroacetate
and its active metabolite fluorocitrate inhibit the
mitochondrial enzyme aconitase, which catalyzes
the conversion of citrate into isocitrate (Clarke,
1991). The specificity of these agents is because of
their selective uptake by the acetate transporter
present only in glial cells (Clarke et al, 1970;
Waniewski and Martin, 1998). Indeed, they have
been used in numerous experimental models to
study the role of astrocytes or neuro-glial coupling
on synaptic transmission and neuron survival (e.g.,
Largo et al, 1996; Hülsmann et al, 2000; Willoughby
et al, 2005; Zielke et al, 2007).
Our results indicate that synaptic transmission is
blocked shortly after selective glial metabolic stress
because of the rapid extracellular accumulation of
adenosine as an active response of astrocytes. Thus,
while adenosine A1 receptor antagonists preserve
synaptic transmission, they also promote the devel-
opment of SD waves and accelerate the loss of
neuronal viability. Importantly, cell culture experi-
ments show that adenosine is directly produced by
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
astrocytes. We conclude that in periods of energy
shortage, glia participates in the overall adjustment
of neuronal activity via the adenosinergic blockage
of synaptic transmission as a general mechanism to
prevent or delay massive depolarization and death.
We believe our data unveil a glial-controlled and
adenosine-mediated safety mechanism, which
could serve as a new therapeutic target in the
context of brain ischemia.
Materials and methods
All experimental procedures conformed to ECC regula-
tions for animal care and every effort was made to
minimize the suffering and the number of animals used.
Slice Preparation and Treatments
Hippocampal slices (450 mm thick) were obtained from
female Sprague–Dawley rats (120 to 150 g) and placed in a
holding chamber (351C) at the interface between humidi-
fied air (95% O2/5% CO2) and artificial cerebrospinal fluid
(ACSF). Slices were incubated for at least 90 mins and
then transferred to a recording chamber (‘Oslo’ type)
continuously superfused by ACSF at 2 mL/min.
To block Krebs cycle in glial cells, we used 10 and
20 mmol/L FAc (Sigma-Riedel de Häen, Seelze, Germany)
in ACSF. The effect of increased osmolarity was compen-
sated by adding mannitol to the ACSF for 30 mins before
FAc (final osmolarity 315 and 335, respectively).
The general adenosine receptor antagonist, 8-(p-sulfo-
phenyl)-theophylline (8-SPT; Sigma, St Louis, MO, USA),
and the A1-selective antagonist, 8-cyclopentyl-1,3-dipro-
pylxanthine (DPCPX; Sigma), were added to the ACSF
before or after exposure to FAc.
Cell Culture
Rat glial cultures were prepared as described previously
(Canals et al, 2001). Briefly, the hippocampus from 1-day-
old neonatal rats was dissected, and the meninges were
carefully removed and digested with 0.25% trypsin for
15 mins at 371C. Trypsinization was stopped by adding an
equal volume of culture medium (Dulbecco’s modified
Eagle’s medium with 15% fetal bovine serum) to which
0.02% DNase I was added. The solution was pelleted,
resuspended in culture medium, and a single-cell suspen-
sion was prepared by repeated pipetting. Cells were
seeded at a density of 35,000 cells/cm2 in P-24 multiwell
dishes and cultured at 371C in humidified 5% CO2. The
medium was replaced every 6 to 8 days and the cultures
reached confluency after 7 to 10 days in vitro. After 10 to
15 days in culture, the number of astrocytes (GFAP + cells)
was about 80% to 90% of the total population of cells.
Hippocampal glial cultures were treated with 20 mmol/L
FAc or solvent for 1, 3, 6, and 24 h at 11 DIV. At the end of
the incubation, the culture medium was collected in two
aliquots, one was used for high-performance liquid
chromatography (HPLC) determination of the adenosine

concentration and the other aliquot was used to evaluate
the lactate dehydrogenase (LDH) activity. The cells were
homogenized and used for protein determination.
Adenosine Determination
The samples were pretreated before the injection in the
HPLC. The proteins were eliminated by acid treat-
ment (perchloric acid) followed by centrifugation at
13,000 r.p.m. for 10 mins and the supernatants neutralized
with KOH. The supernatants were submitted to chromato-
graphy through Accell QMA cartridges (from Waters,
Milford, MA, USA). These single-use columns retain all
the negatively charged nucleotides, permitting the elution
of noncharged molecules (nucleosides). Eluates were
processed by HPLC, as described in the following
paragraph.
Aliquots (25 mL) were analyzed by HPLC on a system
made up of a Waters 1515 isocratic HPLC pump, a 2487
dual-wavelength absorbance detector, and a 717 autosam-
pler, which were all controlled by the Breeze software
(Waters). The column used was a mbondaPak C18 (22 cm
length, 0.4 cm diameter) and the mobile phase was
0.1 mol/L KH2PO4, 5% methanol (pH 6.0). The flow rate
was 1.5 mL/min and the column was maintained at room
temperature (221C±11C). Detection was monitored at a
wavelength of 254 nm and the system was equilibrated
overnight before the injection of the corresponding
external standards. External standards were prepared with
adenosine purchased from Sigma, prepared in ultrapure
water. The quantification of adenosine in the samples was
performed by comparing the areas of the peaks obtained in
the samples with the corresponding areas of the external
standards (Breeze software).
To confirm that the peak that elutes with the same
retention time as the standard was actually adenosine, a
24 h sample was enriched with 1 mmol/L commercial
adenosine and submitted to chromatography. The analysis
of the chromatographic profile showed a single peak with
increased size, suggesting the substance contained in the
sample was adenosine (Figure 3A, middle panel). To
further verify the identity of the measured compound, a
sample containing the putative adenosine peak was
incubated in the presence of 50 adenylic acid deaminase
from Aspergillus sp, which transforms adenosine into
inosine. The enzyme (0.12 U) was incubated with the 24 h
sample in a final volume of 1 mL of 50 mmol/L HEPES (pH
6.5). Incubation was carried out at 371C for 1.5 h. The
HPLC analysis clearly showed that the compound was
adenosine as there was a reduction in the putative
adenosine peak with a concomitant increase in the inosine
peak (Figure 3A, lower panel).
Electrophysiology
Monopolar stimulating electrodes (40 mm tungsten wires)
were placed at the alvear region and the lower third of the
stratum radiatum to elicit anti- and orthodromic (synaptic)
responses, respectively. Two recording pipettes filled with
150 mmol/L NaCl (3 to 6 MO) were placed along the main
Glial adenosine and energy shortage
S Canals et al
1837
CA1 pyramidal cell axis to record from the same
population. One was located at the stratum pyramidale
to record the population spike (PS) and the other at the
stratum radiatum to record the field excitatory postsynap-
tic potential (fEPSP; Canals et al, 2005a). A third pipette
was located at the stratum pyramidale of the CA3 region to
record the antidromic PS from this region after Schaffer
stimulation in the stratum radiatum. The collision test was
used to assess whether the ortho- and antidromic stimulat-
ing electrodes were essentially activating the same popula-
tion in the CA1 (López-Aguado et al, 2002). Suitable
stimulating/recording arrangements were chosen so that
the antidromic PS totally collided with a previous
orthodromic PS. After filtering (DC: 5 kHz) and amplifying,
signals were recorded on VCR, acquired by a computer
(20 to 40 kHz acquisition rate, Digidata 1200; Axon
Instruments, Burlingame, CA, USA) and processed using
the Axoscope and Clampfit software (Axon Instruments).
In a group of experiments, intracellular recordings were
also made using micropipettes (1.5 mm o.d.) backfilled
with 4 mol/L potassium acetate (80 to 100 MO). The
signals obtained were amplified using a bridge circuit
amplifier (Axoclamp 2A; Axon Instruments), filtered at
10 kHz, and stored in a VCR for offline analysis. Impale-
ments were made from the somata of electrophysiologi-
cally identified pyramidal cells at least 80 mm below the
slice surface.
Measurement and Statistics
The antidromic PS in the stratum pyramidale of both the
CA1 and CA3 regions was measured as the amplitude of
the peak of its negative limb with respect to the preceding
baseline. When the artifact partially overlapped the
evoked potential, the most positive value between them
was used instead. The orthodromic PS in the CA1 was
measured as the amplitude from the preceding positive
crest and negative peak. The fEPSP was measured as the
slope of its negative phase. Data were quantified as
the mean±s.e.m. Statistical comparisons were made using
the Student’s t-test. Multiple comparisons were made with
one-way analysis of variance followed by the Newman–
Keuls test as a post hoc evaluation. Differences were
considered statistically significant when P < 0.05.
Results
The Impairment of Glial Oxidative Metabolism Blocks
Synaptic Transmission in the CA1 and CA3 Regions
of the Hippocampus
We first characterized the short-term effects of FAc
treatment on the electrophysiological properties of
pyramidal neurons in the CA1 and CA3 regions of
the hippocampus using three recording (R1–3) and
two stimulating (S1,2) electrodes in the hippocampus
(Figure 1A). The ortho- and antidromic field
responses were recorded before and 30 mins after
20 mmol/L FAc treatment (black versus gray tra-
cings,respectively, from a representative experiment;
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
 Glial adenosine and energy shortage
S Canals et al
1838

Figure 1 Inhibition of oxidative metabolism in glia blocks synaptic transmission without producing changes in the electrogenic
properties of neurons. (A) Schematic illustration of the position of the recording (R1–3) and stimulating (S1,2) electrodes in this and the
following experiments. (B) Representative tracings corresponding to the sample field potentials in control conditions (black) and after
30 mins of FAc perfusion (20 mmol/L, gray). Orthodromic PS (S1–R1), fEPSP (S1–R2), and synaptic recurrence in CA3 (S1–R3,
arrowhead) are blocked, whereas antidromic potentials (arrows) and the fiver volley (asterisk in S1–R2) remain unaffected. Same
calibration bars apply to recordings R1 and R2. (C) Evolution over time of the ortho- and antidromic potentials during exposure to FAc
(20 mmol/L, mean±s.e.m. from n = 19 slices and 9 animals). (D) Upper panel: intracellular recordings from the soma of a
pyramidal cell showing the ortho- and antidromic responses in control conditions (black), and after FAc perfusion for 30 mins
(20 mmol/L, dashed) and 90 mins (gray). Lower panel: the same cell shows normal responses to a depolarizing current pulse
( + 0.2 nA, 1 sec) after a 90 min perfusion of FAc.

Figure 1B) and the time course of the fEPSP, ortho-
and antidromic PS in the CA1, and the antidromic
PS in CA3 was established from the pooled data for
20 mmol/L FAc (n = 19 slices, 9 animals; Figure 1C).
Synaptic transmission in the hippocampus was
strongly depressed after a 30 mins perfusion of
FAc. Indeed, the orthodromic PS recorded in the
CA1 soma layer and the fEPSP in the stratum
radiatum of CA1 decreased to 98.6%±0.5% and
81.4%±6.7% of their control values, respectively
(Figure 1C). Orthodromic transmission in the CA3
region also declined, as noted by the selective
blockage of the second PS because of recurrent
excitation (Figure 1B, arrowhead). This effect com-
menced 10 to 15 mins after FAc treatment, and once
it began, the depression of the orthodromic PS was
complete within 10 mins. In sharp contrast, the
antidromic field responses in CA1 and CA3 were
largely unaffected, as was the fiver volley recorded
in the stratum radiatum (arrows in R1–S2 and R3–S1
and asterisk in R2–S1, respectively). Indeed, only a
small increase in their amplitude was noted from
the onset of the decline in orthodromic transmission
(Figures 1B and 1C). Similar blockage of synaptic
transmission, and intact action potential initiation
and conduction, was observed in intracellular
recordings from the CA1 pyramidal cells (Figure
1D). In addition, the intracellular injection of
depolarizing current pulses ( + 0.2 nA, 1 sec) elicited
very similar responses both before and up to 90 mins
after FAc perfusion (in the neuron shown in Figure
1D). A progressive decline of antidromic transmis-
sion was only observed after 40 to 90 mins
(59±24 mins, mean±s.d.) of FAc treatment, conco-
mitant with the appearance of the first SD waves (see
below). These results are in agreement with previous
studies using FAc, both in preparations in vitro and
in vivo (Stone et al, 1990; Keyser and Pellmar, 1994,
Largo et al, 1996; Larrosa et al, 2006), and indicate
that the selective impairment of glial oxidative
metabolism inhibits synaptic transmission without
interfering with the electrogenic membrane function.
Blockage of Synaptic Transmission is because of the
Activation of A1 Adenosine Receptors
Adenosine accumulates in the extracellular
space because of the inhibition of mitochondrial

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844

 Glial adenosine and energy shortage
S Canals et al
1839
respiration during hypoxic and chemical hypoxia
(Pull and McIlwain, 1972; Gribkoff et al, 1990;
Frenguelli et al, 2007). Hence, adenosine accumula-
tion because of the selective glial intoxication may
be responsible for the fast interruption of synaptic
transmission in the present model. We tested the
effects of general and A1-selective antagonists of
adenosine receptors on the blockage of orthodromic
potentials in the hippocampus during glial cell
intoxication. Synaptic transmission was recovered
by adenosine antagonists in both the CA1 and CA3
regions of the hippocampus. This recovery was
evident through the temporal evolution of the field
potentials, and their quantification on exposure to
FAc treatment, both before and after selective
inhibition of A1 receptors with DPCPX, is shown
in Figures 2A and 2B. A similar effect could be seen
in the pooled data for DPCPX (eight slices in five
animals) and the general adenosine receptor antago-
nist 8-SPT (eight slices in five animals; Figure 2C).
Likewise, recovery of synaptic transmission was
obtained when the slices were exposed to 10 mmol/L
FAc in the presence of adenosine antagonists. Indeed,
the blockage of synaptic transmission was also Figure 2 Adenosine A1 receptor antagonism restores synaptic
prevented when the slices were pretreated with transmission in the hippocampus. Representative tracings (A)
DPCPX or 8-SPT before FAc perfusion (data not and the evolution of field potentials over time (B) in the same
experiment illustrating the return of synaptic transmission when
shown).
0.1 mmol/L DPCPX is introduced into the perfusion medium.
These results show that glial respiration regulates The numbers in brackets designate the time at which the
synaptic transmission through adenosine binding tracings were obtained. Same calibration bars apply to
to A1 receptors. This mechanism affects different recordings R1 and R2. (C) Quantitative analysis of orthodromic
synapses within the hippocampus, as Schaffer potentials before and after perfusion of the adenosine antago-
collateral-evoked fEPSPs in the stratum radiatum nists during FAc treatment (10 mmol/L 8-SPT or 0.1 mmol/L
of CA1 and synaptic recurrence in CA3 are regulated DPCPX).
in the same way.

space from the first evaluated time point (300%

Glial Cells and Extracellular Adenosine in Response
to Mitochondrial Inhibition
Having shown the mechanistic role of adenosine in
synaptic blockade during glial intoxication, we next
studied the ability of glial cells in releasing
adenosine in response to mitochondrial inhibition.
To address this question we cultured glial cells from
rat hippocampus and incubated them with
20 mmol/L FAc for different periods of time. We
chose the cell culture model to unequivocally show
the capability of glial cells from the hippocampus to
accumulate adenosine in the extracellular medium
in response to a metabolic challenge, independent
of the neuron-glia interplay. In addition, the cell
culture model allowed us to assay adenosine
concentrations in a very robust and reproducible
way. The cultures were first characterized immuno-
cytochemically to show that they consisted mainly
of astrocytes (80% to 90% GFAP-positive cells) and
almost no neurons were present (less than 1% to 2%
b-tubuline-positive cells). Moreover, cell membrane
integrity (LDH release) was assessed in parallel with
adenosine determinations. On exposure to FAc,
adenosine clearly accumulates in the extracellular
increase) and rises in a time-dependent manner
(Figures 3B and 3C). This accumulation was not
accompanied by LDH release into the culture
medium and thus could not be attributed to the
breakdown of the plasma membrane (Figure 3D).
Although the time course of adenosine accumula-
tion in a closed system (cell culture) cannot be
directly compared with that of an open one (super-
fused brain slices), these data clearly show that glial
cells accumulate adenosine in the extracellular
space when their oxidative metabolism is impaired.
Our results are in good agreement with previously
published data on glial cell cultures showing
adenosine accumulation under metabolic stress
(Parkinson and Xiong, 2004).
The Glial-Mediated Accumulation of Extracellular
Adenosine Delays or Prevents the Occurrence
of Lethal SD Waves
Spreading depression waves develop spontaneously
when the glial oxidative metabolism is inhibited for
long periods leading to neuronal cell death in
the hippocampus (Nedergaard and Astrup, 1986;

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844

 Glial adenosine and energy shortage
S Canals et al
1840

Figure 3 Glial cell-mediated extracellular accumulation of adenosine in response to mitochondrial inhibition. Hippocampal glial cells
in culture were exposed to FAc (20 mmol/L) for different periods of time. (A) Identification of adenosine in biological samples. A
sample containing the putative adenosine peak and the same sample enriched with 1 mmol/L commercial adenosine (the two upper
panels). The treatment with adenylic acid deaminase transforms the putative adenosine into inosine (third panel from the top). The
lower chromatogram is a standard of 1 mmol/L commercial adenosine (Ado), guanosine (Gno), inosine (Ino), and hypoxanthine (Hyp).
(B) Representative HPLC chromatograms of samples collected at the indicated times. Lower chromatogram same as in (A).
(C) Adenosine concentration in the culture medium. (D) The lactate dehydrogenase (LDH) activity measured in the culture medium
remained constant throughout the experiment, ruling out the possible contribution of cell lysis to the extracellular adenosine
measured. AUFS, absorbance units full scale.

Hossman, 1996; Largo et al, 1996, 1997; Lian and
Stringer, 2004). If the blockage of synaptic transmis-
sion represents a safety mechanism controlled by
glial cells to allocate energy resources to vital
processes, then antagonism of A1 receptors should
precipitate such processes. We tested this prediction
by increasing the time of FAc perfusion (up to 2 h)
and evaluating the electrical viability of neurons
(n = 6 slices from three rats). Longer exposure to FAc
in our model led to the appearance of SD waves
(Figure 4) and a decrease of the antidromic PS
amplitude (75.6%±11.3% of the control 1 h after
exposure to FAc). Stepwise reductions in the
antidromic PS were observed after each SD episode,
and the amplitude of the antidromic PS was
inversely correlated with the number of SD events
(data not shown). This effect can be appreciated
through the evolution of the antidromic PS on
exposure to FAc (Figure 4A), where the experiments
in which SD waves developed or not can be
compared. In the presence of A1 blockers, the
restoration of synaptic transmission was accompa-
nied by a hastening of the SD waves (Figure 4B), an
increase in its frequency (Figure 4C), and a marked
reduction in the electrical viability of pyramidal
neurons (Figure 4D). Thus, the antagonic effect of
adenosine on SD occurrence during glial intoxica-
tion represents a fundamental protective mechan-
ism for neurons.
This phenomenon is also illustrated in Figure 5,
where the effects of A1 antagonism were investi-
gated well after the inhibition of glial respiration. In
this experiment, FAc treatment produced the de-
scribed effects on synaptic and antidromic transmis-
sion with a stepwise reduction in the a-PS (black
arrow) only after the occurrence of a SD episode
(black asterisk). The subsequent administration of
DPCPX, more than 90 mins after the FAc treatment
started, was still very efficient in recovering the
synaptic transmission (gray arrow); however, the
recovery process was curtailed by the immediate
precipitation of a second SD episode (gray asterisk),
which was now lethal for the tissue (note the
complete abolition of the antidromic response).
Discussion
The main finding of this study is that inhibition of
oxidative metabolism in glial cells produces an
adenosine A1-receptor-mediated blockage of synap-
tic transmission in the hippocampus, which delays

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844


Figure 4 The glial-dependent accumulation of adenosine is a
safety mechanism that preserves neuronal viability by opposing
SD waves. Cell viability in CA1 is quantified by the amplitude of
maximal antidromic PS during the perfusion of FAc (20 mmol/L)
(A). The experiments were grouped depending on whether SD
waves did (filled circles) or did not develop (open circles). The
occurrence of SD waves (normally starting by the end of the first
hour of perfusion) caused an almost total and irreversible
blockage of the PS in individual experiments. In another group
of experiments, the slices were incubated with FAc (20 mmol/L)
for 2 h in the presence or absence of the general antagonist of
adenosine receptors, 8-SPT, or the A1-specific antagonist,
DPCPX. The time of onset (B) and frequency (C) of SD waves,
as well as the remaining antidromic PS at the end of the
experiment (D), were analyzed. In the presence of adenosine
antagonists, SD events appear earlier and with higher
frequency. A concomitant higher decrease in the electrical
viability of neurons is observed, when compared with only FAc
treatment. Data represent mean±s.e.m. from n = 6 slices and
3 rats; *P < 0.05, **P < 0.01.
the appearance of SD waves and preserves neuronal
viability. We interpret these results as representing a
glial-operated safety mechanism that allocates meta-
bolic resources to vital processes when the energetic
capacity of the tissue (or the capacity sensed by glial
cells) is surpassed. In turn, this delays the onset of
SD waves that are lethal to neurons in regions
of glial malfunction, increasing the time window for
survival.
Specificity of Glial Toxins
Numerous reports indicate that FAc and fluoroci-
trate are selective inhibitors of glial metabolism. Not
only are they selectively taken up by astrocytes
(Clarke et al, 1970; Waniewski and Martin, 1998),
their initial effects are also unambiguously related
to the impairment of functions commonly attributed
to glial cells. These include the regulation of pHo,
spatial potassium buffering, and the metabolic
Glu–glutamine cycle (Berg-Johnsen et al, 1993;
Largo et al, 1996; Stringer and Aribi, 2003). Also,
Glial adenosine and energy shortage
S Canals et al
1841
Figure 5 Adenosine released in response to glial intoxication
preserves neuronal viability. The time course of the evoked field
potentials (anti- and orthodromic PS, black and white circles,
respectively) and DC potential (black line) in CA1 during FAc
perfusion are shown. At 95 mins of FAc treatment, 0.1 mmol/L
DPCPX was introduced in the perfusion medium (gray bar).
Note the occurrence of two SD waves, one in a period of only
FAc treatment (black asterisk), inducing a transient inactivation
of the antidromic transmission (during the time length of the SD
episode; black arrow) and followed by the partial recovery of the
a-PS amplitude; and a second one after DPCPX inclusion,
producing the permanent abolition of electrical activity in the
slice. The synaptic transmission was also recovered by the A1
antagonist, shortly before the ignition of the second and lethal
SD event.
their exogenous application causes the gradual
selective depolarization of astrocytes (Largo et al,
1997; Hülsmann et al, 2000). In contrast, these
toxins have no effect on the antidromic PS (Stone
et al, 1990; Keyser and Pellmar, 1994; Largo et al,
1996, 1997; Larrosa et al, 2006) or on the membrane
properties of pyramidal neurons (Larrosa et al, 2006;
Hülsmann et al, 2000). The neuronal demise
observed after long-term exposure to these inhibitors
(several hours) is therefore secondary to the disrup-
tion of glial homeostatic activity, the unavoidable
perturbation of the reciprocal neuro-glia metabolic
and signaling pathways, together with the occur-
rence of SD waves (Largo et al, 1996; Lian and
Stringer, 2004).
Mechanisms of Synaptic Blockage
It is generally agreed that the loss of synaptic
transmission on FAc treatment is produced by the
exhaustion of the neurotransmitter pool of glutamate
because of the inhibition of the glutamine/glutamate
cycle in glial cells (Quastel, 1978). This view is
supported by the experimental observation that the
glutamine precursor is depleted after FC treatment
(Largo et al, 1996), and that exogenous glutamine
maintains synaptic potentials in brain slices (Keyser
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
 Glial adenosine and energy shortage
S Canals et al
1842
and Pellmar, 1994). However, this interpretation is
weakened by the fact that glial-derived glutamine is
not the only source of the neurotransmitter gluta-
mate, as it can also be synthesized from glucose in
neurons (Ward et al, 1983). Furthermore, the highly
variable rate and degree of synaptic depression
observed in different studies (from 30 mins to 5 h
and from total suppression to only 85%; Berg-
Johnsen et al, 1993; Keyser and Pellmar, 1994; Largo
et al, 1996; Martin et al, 2007) raise some doubt
whether the lack of transmitter is the main cause of
synaptic failure during glial poisoning. One may
argue that because of the compartmentalization of
metabolites, only the neurotransmitter glutamate
pool is affected by glial intoxication and that this
represents a small fraction of the total glutamate
pool (Hamberger et al, 1979). Even if this was the
case, it is also known that glutamine can feed
the tricarboxylic acid cycle, bypassing the FAc
inhibition (Hassel et al, 1994) and, thus, restoring
the energetic status of the glial cells. Indeed,
glutamine supplementation prevents the decrease
in ATP levels induced in glial cell cultures by FAc
(Hassel et al, 1994). Based on the above evidence, it
does not appear that neurotransmitter depletion is
responsible for the blockage of synaptic transmis-
sion during glial poisoning or vice versa. Our results
clearly showed that the glutamate pool is available
for synaptic transmission during FAc treatment
(at least for the length of our experiments, some up
to 3 h), as synaptic transmission is restored shortly
after it has been fully blocked, and that the
mechanism for synaptic blockade is, indeed,
the accumulation of extracellular adenosine and
subsequent activation of A1 receptors in the
pyramidal neuron.
A variety of manipulations that alter energy
metabolism, such as oxygen and/or glucose depriva-
tion and chemical anoxia, have been shown to
produce adenosine accumulation in the interstitial
space as well as adenosine-mediated blockage of
synaptic transmission (Latini and Pedata, 2001).
Adenosine could be released from both pre- and
postsynaptic neurons, as well as from nonneuronal
cells. Normally, adenosine accumulation is thought
to arise from neural elements by either direct release
or through the extracellular metabolism of released
nucleotides (Cunha, 2001; Latini and Pedata, 2001).
However, our results, together with the data
reported by others (Parkinson and Xiong, 2004;
Martin et al, 2007), indicate that astrocytes are a
source of adenosine. The adenosinergic blockage of
synaptic transmission reported here is by far the
earliest noticeable effect of FAc challenge (B5 to
10 mins) well before any indication of altered ion
homeostasis, which occurs in vivo after at least 1 to
2 h. Thus, it is likely that adenosine is released
directly from astrocytes as a direct response to the
blockage of the tricarboxylic acid cycle. This inter-
pretation is also supported by the experiments in
glial cultures, which show that in absence of
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
neurons, glial cells produce considerable amounts
of adenosine when challenged with FAc.
Glial-Mediated Accumulation of Extracellular
Adenosine Provides Neuroprotection by Delaying
the Occurrence of SD Waves: Implications for
Ischemic Brain Injury
For the first time, we have shown that glial-
mediated interruption of synaptic transmission is a
reversible phenomenon that delays the appearance
of SD waves, decreases their frequency, and pre-
serves the electrical viability of pyramidal neurons.
It is well known that neuronal damage at the core
of a brain infarct cannot be reversed. Massive
depolarization is accompanied by the breakdown
of membrane integrity and the release of a number of
neuroactive agents. However, the surrounding re-
gion, the so-called ischemic penumbra, is initially
viable and only moderate signs of neuron distress
can be detected (Hossman, 1996). A causal relation-
ship between the SD waves arising from the borders
of the ischemic nucleus and the neuronal death in
the surrounding penumbra has already been estab-
lished (Iijima et al, 1992; Fabricius et al, 2006). It
was previously shown that energy shortage in glial
cells is a necessary condition for SD waves to kill the
neurons. Accordingly, it was proposed that SD
waves and glial dysfunction together are responsible
for cell death in the ischemic penumbra (Largo et al,
1996). Here, we show that the specific metabolic
stress of glia is by itself a reliable trigger of SD
waves, and that their occurrence, as well as the
subsequent loss of neuronal function, can be
delayed by the activation of A1 receptors. In the
context of brain ischemia, we consider that the
release of adenosine by glia could function as a
protective response against cell death, diverting
energy from processes that are not essential for cell
survival to vital processes, and delaying the occur-
rence of lethal SD waves in the ischemic penumbra.
Such a mechanism could increase the survival time
window of any brain region confronted by an
ischemic episode.
It is also well established that glial cells partici-
pate in the regulation of cerebral blood flow (Harder
et al, 1998) and that adenosine fulfils a key role in
the so-called neurovascular coupling (Phillis, 2004).
Together, these observations suggest that by con-
comitantly decreasing synaptic transmission and
increasing local cerebral blood flow, adenosine from
glial origin could restore the ratio of energy produc-
tion and consumption during periods of energy
imbalance in vivo. It is also plausible that in the
ischemic penumbra, adenosine release alleviates
hypoxic episodes (Lauritzen et al, 1982; Shin et al,
2006; Strong et al, 2007) and the excessive workload
associated to the passage of peri-infarct depolariza-
tions. Hence, the neuroprotective mechanism de-
scribed here may be envisaged as a safeguard against

the otherwise lethal outcome of breaking the normal
interplay between neurons and glia when these are
unable to maintain the normal flow of energy
metabolites. In vivo studies have already shown
interferences between endogenous adenosine and
the generation of SD waves (Kaku et al, 1994).
Further studies will be required to understand the
mechanism of such an interaction, which is of
particular clinical relevance as a potential new
target for the treatment of ischemic and traumatic
brain injury.
Acknowledgements
We thank Rosa Solano for help with cell cultures,
Eduardo Martin for critical reading of the paper,
Alfonso Araque for helpful comments, and Mark
Sefton (BiomedRed SL) for editorial assistance.
Disclosure/conflict of interest
The authors declare no conflict of interest.
References
Ames A (2000) CNS energy metabolism as related to
function. Brain Res Brain Res Rev 34:42–68
Araque A, Carmignoto G, Haydon PG (2001) Dynamic
signaling between neurons and astrocytes. Annu Rev
Physiol 63:795–813
Berg-Johnsen J, Paulsen RE, Fonnum F, Langmoen IA
(1993) Changes in evoked potentials and amino acid
content during fluorocitrate action studied in rat
hippocampal slices. Exp Brain Res 96:241–6
Canals S, Casarejos MJ, Rodrı́guez-Martı́n E, De Bernardo
S, Mena MA (2001) Neurotrophic and neurotoxic
effects of nitric oxide on fetal midbrain cultures.
J Neurochem 76:56–68
Canals S, López-Aguado L, Herreras O (2005a) Synapti-
cally recruited apical currents are required to initiate
axonal and apical spikes in hippocampal pyramidal
cells: modulation by inhibition. J Neurophysiol 93:
909–918
Canals S, Makarova J, López-Aguado L, Largo C, Ibarz JM,
Herreras O (2005b) Longitudinal depolarization gradi-
ents along the somatodendritic axis of CA1 pyramidal
cells: a novel feature of spreading depression.
J Neurophysiol 94:943–51
Clarke DD (1991) Fluoroacetate and fluorocitrate: mechan-
ism of action. Neurochem Res 16:1055–8
Clarke DD, Nicklas WJ, Berl S (1970) Tricarboxylic acid-
cycle metabolism in brain. Effect of fluoroacetate and
fluorocitrate on the labelling of glutamate, aspartate,
glutamine and g-aminobutyrate. Biochem J 120:345–51
Cunha RA (2001) Adenosine as a neuromodulator and as a
homeostatic regulator in the nervous system: different
roles, different sources and different receptors. Neuro-
chem Int 38:107–25
Dunwiddie TV, Masino SA (2001) The role and regulation
of adenosine in the central nervous system. Annu Rev
Neurosci 24:31–55
Glial adenosine and energy shortage
S Canals et al
1843
Fabricius M, Fuhr S, Bhatia R, Boutelle M, Hashemi P,
Strong AJ, Lauritzen M (2006) Cortical spreading
depression and peri-infarct depolarization in acutely
injured human cerebral cortex. Brain 129:778–90
Fowler JC (1989) Adenosine antagonists delay hypoxia-
induced depression of neuronal activity in hippocam-
pal brain slice. Brain Res 490:378–84
Frenguelli BG, Wigmore G, Llaudet E, Dale N (2007)
Temporal and mechanistic dissociation of ATP and
adenosine release during ischaemia in the mammalian
hippocampus. J Neurochem 101:1400–13
Greene RW, Haas HL (1991) The electrophysiology of
adenosine in the mammalian central nervous system.
Prog Neurobiol 36:329–41
Gribkoff VK, Bauman LA, VanderMaelen CP (1990)
The adenosine antagonist 8-cyclopentyltheophylline
reduces the depression of hippocampal neuronal
responses during hypoxia. Brain Res 512:353–7
Hamberger A, Chiang GH, Nylén ES, Scheff SW, Cotman
CW (1979) Glutamate as a CNS transmitter. I. Evalua-
tion of glucose and glutamine as precursors for the
synthesis of preferentially released glutamate. Brain
Res 168:513–30
Harder DR, Alkayed NJ, Lange AR, Gebremedhin D,
Roman RJ (1998) Functional hyperemia in the brain
hypothesis for astrocyte-derived vasodilator metabo-
lites. Stroke 28:229–34
Hassel B, Sonnewald U, Unsgard G, Fonnum F (1994)
NMR spectroscopy of cultured astrocytes: effects of
glutamine and the gliotoxin fluorocitrate. J Neurochem
62:2187–94
Herreras O, Somjen GG (1993) Effects of prolonged
elevation of potassium on hippocampus of anesthe-
tized rats. Brain Res 617:194–204
Hossman KA (1996) Periinfarct depolarizations. Cerebro-
vasc Brain Metab Rev 8:195–208
Hülsmann S, Oku Y, Zhang W, Richter DW (2000)
Metabotropic glutamate receptors and blockade of glial
Krebs cycle depress glycinergic synaptic currents of
hypoglossal motoneurons. Eur J Neurosci 12:239–46
Iijima T, Mies G, Hossmann KA (1992) Repeated negative DC
deflections in rat cortex following middle cerebral artery
occlusion are abolished by MK-801. Effect on volume of
ischemic injury. J Cereb Blood Flow Metab 12:727–33
Ilie A, Ciocan D, Zagrean AM, Nita DA, Zagrean L,
Moldovan M (2006) Endogenous activation of adeno-
sine A(1) receptors accelerates ischemic suppression of
spontaneous electrocortical activity. J Neurophysiol
96:2809–14
Kaku T, Hada J, Hayashi Y (1994) Endogenous adenosine
exerts inhibitory effects upon the development of
spreading depression and glutamate release induced
by microdialysis with high K+ in rat hippocampus.
Brain Res 658:39–48
Keyser DO, Pellmar TC (1994) Synaptic transmission in
the hippocampus: critical role for glial cells. Glia
10:237–43
Largo C, Cuevas P, Somjen GG, Martı́n del Rı́o R, Herreras O
(1996) The effect of depressing glial function on rat
brain in situ on ion homeostasis, synaptic transmission
and neuronal survival. J Neurosci 16:1219–29
Largo C, Ibarz JM, Herreras O (1997) Effects of the
gliotoxin fluorocitrate on spreading depression
and glial membrane potential in rat brain in situ.
J Neurophysiol 78:295–307
Larrosa B, Pastor J, López-Aguado L, Herreras O (2006) A
role for glutamate and glia in the fast network
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
 Glial adenosine and energy shortage
S Canals et al
1844
oscillations preceding spreading depression. Neuro-
science 141:1057–68
Latini S, Pedata F (2001) Adenosine in the central nervous
system: release mechanisms and extracellular concen-
trations. J Neurochem 79:463–84
Lauritzen M, Jorgensen MB, Diemer NH, Gjedde A,
Hansen AJ (1982) Persistent oligemia of rat cerebral
cortex in the wake of spreading depression. Ann Neurol
12:469–74
Lian XY, Stringer J (2004) Energy failure in astrocytes
increases the vulnerability of neurons to spreading
depression. Eur J Neurosci 19:2446–54
López-Aguado L, Ibarz JM, Varona P, Herreras O (2002)
Structural inhomogeneities differentially modulate ac-
tion currents and population spikes initiated in the
axon or dendrites. J Neurophysiol 88:2809–20
Martin ED, Fernandez M, Perea G, Pascual O, Haydon PG,
Araque A, Ceña V (2007) Adenosine released by
astrocytes contributes to hypoxia-induced modulation
of synaptic transmission. Glia 55:36–45
Nedergaard M, Astrup J (1986) Infarct rim: effect
of hyperglycemia on direct current potential and
[14C]2-deoxyglucose phosphorylation. J Cereb Blood
Flow Metab 6:607–15
Parkinson FE, Xiong W (2004) Stimulus- and cell-
type-specific release of purines in cultured rat
forebrain astrocytes and neurons. J Neurochem 88:
1305–12
Phillis JW (2004) Adenosine and adenine nucleotides as
regulators of cerebral blood flow: roles of acidosis, cell
swelling, and KATP channels. Crit Rev Neurobiol
16:237–70
Pull I, McIlwain H (1972) Metabolism of [14C]adenine and
derivatives by cerebral tissues, superfused and elec-
trically stimulated. Biochem J 126:965–73
Quastel JH (1978) Cerebral glutamate–glutamine interrela-
tions in vivo and in vitro. In: Dynamic properties of glia
Journal of Cerebral Blood Flow & Metabolism (2008) 28, 1835–1844
cells (Schoffeniels E, Frank G, Tower DB, Hertz L, eds),
Oxford: Pergamon, 153–62
Rudolphi KA, Schubert P, Parkinson FE, Fredholm BB
(1992) Neuroprotective role of adenosine in cerebral
ischaemia. Trends Pharmacol Sci 13:439–45
Shin HK, Dunn AK, Jones PB, Boas DA, Moskowitz MA,
Ayata C (2006) Vasoconstrictive neurovascular cou-
pling during focal ischemic depolarizations. J Cereb
Blood Flow Metab 26:1018–30
Somjen GG, Aitken PG, Balestrino M, Herreras O,
Kawasaki K (1990) Spreading depression-like depolar-
ization and selective vulnerability of neurons. A brief
review. Stroke 21(11 Suppl):III179–83
Stone EA, Sessler FM, Weimin L (1990) Glial localization
of adenylate-cyclase-coupled b-adrenoceptors in rat
forebrain slices. Brain Res 530:295–300
Stringer JL, Aribi AM (2003) Effects of glial toxins on
extracellular acidification in the hippocampal CA1
region in vivo. Epilepsy Res 54:163–70
Strong AJ, Anderson PJ, Watts HR, Virley DJ, Lloyd A,
Irving EA, Nagafuji T, Ninomiya M, Nakamura H, Dunn
AK, Graf R (2007) Peri-infarct depolarizations lead to
loss of perfusion in ischaemic gyrencephalic cerebral
cortex. Brain 130:995–1008
Waniewski RA, Martin DL (1998) Preferential utilization
of acetate by astrocytes is attributable to transport.
J Neurosci 18:5225–33
Ward HK, Thanki CM, Bradford HF (1983) Glutamine and
glucose as precursors of transmitter amino acids: ex
vivo studies. J Neurochem 40:855–60
Willoughby JO, Mackenzie L, Pope KJ, Broberg M, Nilsson M
(2005) Localised astroglial dysfunction disrupts
high-frequency EEG rhythms. J Neural Transm
112:205–13
Zielke HR, Zielke CL, Baab PJ, Tildon JT (2007) Effect of
fluorocitrate on cerebral oxidation of lactate and
glucose in freely moving rats. J Neurochem 101:9–16