Neuron, Vol. 43, 729–743, September 2, 2004, Copyright 2004 by Cell Press
Neuronal Synchrony Mediated by
Astrocytic Glutamate through Activation
of Extrasynaptic NMDA Receptors
characterized by rapid offset kinetics, dominates at the
synapse, while the NR1/NR2B complex, characterized
by slow kinetics, is mainly in extrasynaptic membrane
(Rumbaugh and Vicini, 1999; Stocca and Vicini, 1998;
Tovar and Westbrook, 1999). The activation of extrasyn-
aptic NMDARsby glutamateescaping fromthe synaptic
cleft during episodes of highsynaptic activity (Conti and
Weinberg, 1999; Kullmann, 1999) suggests the hypothe-
sis that they have a distinct role (Hardingham et al.,
2002; Scimemi et al., 2004; Tovar and Westbrook, 2002).
Extrasynaptic NMDARs might also represent prefer-
ential targets of glutamate released from a nonsynaptic
source, such as astrocytes. These glial cells release
glutamate through a Ca2?-dependent mechanism (Ar-
aque et al., 2000; Bezzi et al., 1998; Parpura et al., 1994;
cal interactions with neurons (Carmignoto, 2000; Hay-
don, 2001). In the hippocampus, astrocytic glutamate
modulates inhibitory transmission (Kang et al., 1998;
Liu et al., 2004) and increases the frequency of sponta-
pionic acid receptor (AMPAR)-mediated events in pyra-
midal neurons (Fiacco and McCarthy, 2004), probably
through activation of presynaptic metabotropic gluta-
mate receptors (mGluRs). While these studies demon-
on specific events in neuronal transmission, the general
role of this process as a widespread phenomenon in
the brain remains to be discovered.
Here we show that activation of [Ca2?]ielevations in
astrocytes by various stimuli elicits in CA1 pyramidal
neurons repetitive, NMDAR-mediated responses mainly
due to the NR1/NR2B complex. A striking feature of
this response is that it occurs with a high degree of
synchrony in multiple neurons.
Tommaso Fellin,1,3Olivier Pascual,2,3
Sara Gobbo,1Tullio Pozzan,1
Philip G. Haydon,2and Giorgio Carmignoto1,*
1Istituto CNR di Neuroscienze and
Dipartimento di Scienze Biomediche Sperimentali
Universita ` di Padova
viale G. Colombo 3
2Department of Neuroscience
University of Pennsylvania
School of Medicine
215 Stemmler Hall
3610 Hamilton Walk
Philadelphia, Pennsylvania 19104
vation of synaptic ionotropic glutamate receptors. In
hippocampal slices, we report that stimulation of
Schaffer collaterals evokes in CA1 neurons delayed
inward currents with slow kinetics, in addition to fast
excitatory postsynaptic currents. Similar slow events
also occur spontaneously, can still be observed when
neuronal activity and synaptic glutamate release are
blocked, and are found to be mediated by glutamate
released from astrocytes acting preferentially on ex-
trasynaptic NMDA receptors. The slow currents can
be triggered by stimuli that evoke Ca2?oscillations in
astrocytes. As revealed by paired recording and Ca2?
imaging, a striking feature of this NMDA receptor re-
sponse is that it occurs synchronously in multiple CA1
neurons. Our results reveal a distinct mechanism for
neuronal excitation and synchrony and highlight a
functional link between astrocytic glutamate and ex-
trasynaptic NMDA receptors.
Schaffer Collateral Stimulation Evokes
Slow Inward Currents Mediated
by NMDA Receptors
Intense stimulation of neuronal afferents is commonly
used to study LTP, a long-lasting increase in the re-
sponse of the postsynaptic neuron that is believed to
level (Bliss and Collingridge, 1993). In hippocampal
slices, in 6 of 22 CA1 pyramidal neurons tested, high-
frequency stimulation of Schaffercollaterals (SCs) in the
absence of extracellular Mg2?triggered delayed, slow
inward currents (SICs; Figure 1A). In three of these neu-
rons, SICs were repetitively evoked by successive stim-
ulations. The SICs occurred at low frequency (Figure
1B), and 12 of 14 events were recorded within the first
minute after turning off the stimulation (mean delay ?
SEM, 40.1 ? 8.2 s; n ? 14). Compared to the excitatory
postsynaptic current (EPSC), SICs displayed a notably
slower rise time (92.3 ? 29.0 ms, n ? 14 versus 6.4 ?
0.6 ms, n ? 44), a decay fit by a single exponential
function with a mean ?decay? 568.5 ? 176.4 ms (EPSC
The N-methyl-D-aspartate receptors (NMDARs) (Din-
gledine et al., 1999) play key roles in physiopathological
phenomena, such as synaptic transmission, long-term
dependent refinement of synaptic connections (Bourne
toxic neuronal damage (Choi and Rothman, 1990), and
epilepsy (Dingledine et al., 1990). NMDARs are hetero-
meric complexes assembled from NR1 and different
cological and kinetic properties (Dingledine et al., 1999).
Depending on the subunit composition, the subcellular
the peak of synaptogenesis, the NR1/NR2A complex,
3These authors contributed equally to this work
sion with t-ACPD (an agonist of mGluRs) triggered slow
NMDAR-mediated inward currents (Parri et al., 2001)
campal neurons, slice perfusion with t-ACPD resulted
in [Ca2?]ielevations mediatedby ionotropic GluR activa-
tion (Pasti et al., 1997). We thus asked whether mGluR
stimulation with the group I mGluR agonist (S)-3,5-dihy-
droxyphenylglycine (DHPG; 10–30 ?M) can trigger SICs
in hippocampal neurons.
Patch-clamp recordings show that in a subpopulation
of CA1 pyramidal neurons (27 of 98, 28%) DHPG trig-
gered D-AP5-sensitive SICs that were indistinguishable
fromthe SICsevokedby SCstimulation(Figure 1E;Sup-
full/43/5/729/DC1]). In the presence of (5S,10R)-(?)-
5,10-imine maleate (MK-801, 20 ?M), an open channel
NMDAR blocker, a few SICs could be observed upon
the first but not the second DHPG stimulation (n ? 4).
SICs reverse polarity at positive potentials (Supplemen-
tal Figures S1A and S1B) and are superimposed on a
steady-state inward current mediated by activation of
neuronal mGluR1 receptors (Congar et al., 1997; Cre ´pel
et al., 1994) (Supplemental Figure S2). SICs were ob-
served with unchanged amplitude and kinetics in 6-nitro-
7-sulphamoylbenzo[f]quinoxaline-2,3-dione (NBQX) (30
?M), a specific AMPAR antagonist (Supplemental Fig-
ures S1C and S1D). Desensitization of the AMPAR ac-
counts for the absence of an AMPA component in SICs
because, in the presence of D-AP5 and TTX, cyclothia-
zide, which reduces AMPAR desensitization (Yamada
and Tang, 1993), unmasked DHPG-evoked inward cur-
rents that were abolished by NBQX (30 ?M; Supplemen-
tal Figure S3).
The percentage of neurons displaying SICs upon
DHPG stimulation and the frequency of SICs do not
change significantly during the second (30%, n? 66;
SICs/min, 1.2 ? 0.2; n ? 20) or the third (24%, n ? 29,
SICs/min, 1.8 ? 0.7; n ? 7) postnatal week. SICs were
detected from all the three cells recorded in slices ob-
tained from rats older than 21 days (data not shown).
Because experiments were performed in 1 ?M tetro-
dotoxin (TTX) to block action potential generation, we
suggest that SICs arise from a nonneuronal origin. In
support of this notion, DHPG-evoked D-AP5-sensitive
SICs were still present after slice incubation (2 hr) with
2 ?M tetanus neurotoxin (TeNT), which blocks the syn-
aptic release of neurotransmitters (Schiavo et al., 1992)
applied to SCs at least 10-fold in TeNT-incubated slices
cally impaired. In controls, each stimulus evoked an
EPSC, while in TeNT the stimulus often failed to evoke
an EPSC (Figure 1G, left), and the amplitude of the few
responses that were observed was drastically reduced
(Figure 1G, right and inset). Thus, SICs are due to NMDA
receptor activation by glutamate of nonsynaptic origin.
Figure 1. Slow Inward Currents Mediated Exclusively by NMDA Re-
ceptors Can Be Recorded from CA1 Pyramidal Neurons
(A) Representative whole-cell patch-clamp recording from a neuron
after episodes of high-frequency SC stimulation. Three successive
episodes of stimulation of 20 s duration were applied. While the first
trial failed to trigger SICs, the second (shown in the figure) and the
third trial evoked four and two SICs, respectively (some of these
currents are shown at an expanded time scale). Scale bars, 100 pA
and 10 s (top); 50 pA and 200 ms (bottom).
(B) Average number of SICs/min in six responsive neurons before
and after SC stimulation. *p ? 0.05.
(C) Examples of spontaneous SICs from a different neuron. Scale
bars, 100 pA and 10 s; 50 pA and 200 ms.
tions, in the presence of D-AP5 and after D-AP5 washout (n ? 2).
(E) Recordings from a pyramidal neuron showing SICs triggered by
3 min application of 15 ?M DHPG (thick line) in 1 ?M TTX. Scale
bars, 100 pA and 50 s; 50 pA and 200 ms. The EPSCs evoked by
SC stimulation to the same neuron which displays DHPG-induced
SICs have the typical fast activation of synaptic currents (rise time,
4.4 ? 0.2 ms, n ? 51).
(F) In a slice incubated in 2 ?M TeNT and perfused with TTX, 10
?M DHPG triggers SICs. Scale bars, 100 pA and 50 s; 50 pA and
(G) Average percentage of failures and EPSC amplitude in controls
and in slices preincubated with TeNT. (Inset) Representative exam-
ples of four EPSCs recorded from neurons in a control slice (left)
and in a TeNT-incubated slice (right). Scale bars, 100 pA and 100
ms. In this as well as the other figures, *p ? 0.05, **p ? 0.001.
?1? 27.6 ? 3.0 ms, ?2? 146.1 ? 12.0 ms, n ? 44), and
a mean amplitude of ?95.0 ? 36.7 pA (Supplemental
Table S1 [http://www.neuron.org/cgi/content/full/43/5/
729/DC1]). A fast AMPA component was always absent
from the SIC.
SICs also occurred spontaneously (Figures 1B and
1C) at a very low frequency (number of SICs/min, 0.16 ?
0.04, n ? 65; Supplemental Table S1 [http://www.neuron.
org/cgi/content/full/43/5/729/DC1]), with the exception
of three neurons (number of SICs/min, 3.1 ? 0.6). SICs
onist (Figure 1D).
In neurons of the ventrobasal thalamus, slice perfu-
Nonsynaptic Origin of Glutamate that Evokes SICs
Since [Ca2?]ielevations trigger glutamate release from
astrocytes (Bezzi et al., 1998; Parpura et al., 1994; Pasti
et al., 2001) and SC stimulation (Pasti et al., 1997; Porter
and McCarthy, 1996) evokes [Ca2?]ielevations in astro-
Astrocyte-Mediated Neuronal Synchrony
Figure 2. DHPG Triggers [Ca2?]iOscillations
in Astrocytes and NMDAR-Mediated Ca2?
Responses in CA1 Pyramidal Neurons
(A) Sequence of pseudocolor images show-
ing the [Ca2?]i changes of two astrocytes
(spots 1 and 2) and two pyramidal neurons
(spots 3 and 4) after stimulation with 10 ?M
DHPG. In each frame, the timing after the on-
set of DHPG application is indicated. Sam-
pling rate, 4 s. Scale bar, 20 ?m. The ratio (R)
of the intensity of the light emitted by Indo-1
at thetwo wavelengths (405/485)is displayed
as a pseudocolor scale. Arrows indicate the
[Ca2?]iincreases in astrocyte 2 and neuron 3.
(B) Time course of the 405/485 changes from
ter distinguish the response from each cell,
in this as well as the other figures, some of
485 values in astrocytes and neurons were
similar and ranged from 0.63 to 0.71. Aster-
isks mark the timing for the images shown
(C) Average percentage of DHPG-responsive
neurons before D-AP5 application (n ? 7), in
the presence of D-AP5 (n ? 7) and after its
washout (n ? 3).
cytes and nonsynaptic, glutamate-mediated SICs in py-
ramidal neurons, astrocytes are good candidates for a
nonsynaptic source of glutamate.
To study the response of astrocytes to DHPG, we
performed Ca2?imaging experiments. In 1 ?M TTX,
DHPG triggered [Ca2?]i elevations in the majority of
astrocytes (mean ? SEM, 71.1% ? 9.2%, n ? 14) and
in a subpopulation of CA1 neurons (38.4% ? 5.5%, n ?
14, Figures 2A and 2B; Supplemental Movie 1A [http://
Neuronal but not astrocytic [Ca2?]iresponses were re-
versibly blocked when DHPG was applied in 200 ?M
D-AP5 (Supplemental Movie 1B; Figure 2C). NBQX
(10–50 ?M; n ? 2) and incubation with 2 ?M TeNT did
not block the action of DHPG (data not shown).
The results indicate that both SICs and transient
[Ca2?]ichanges in pyramidal neurons reflect the same
event, i.e., NMDAR activation by DHPG-induced gluta-
mate release from a nonsynaptic origin, presumably
Other stimuli that are effective in evoking [Ca2?]ioscil-
lations in astrocytes, such as purinergic receptor ago-
nists (Arcuino et al., 2002) and low Ca2?(Parri et al.,
2001; Zanotti and Charles, 1997), also evoke SICs (Sup-
plemental Figure S4 [http://www.neuron.org/cgi/content/
full/43/5/729/DC1]). The amplitude and kinetics of SICs
evoked by the various stimuli are similar (Supplemental
Table S1 and Supplemental Figure S4).
2004). UV photolytic elevation of astrocytic Ca2?evoked
a SIC in the associated pyramidal neuron in 36% of the
examples tested (n ? 98; Figure 3). Photo release of
Ca2?in the astrocytic cell body locally elevates [Ca2?]i,
which slowly propagates throughout the processes (Ca2?
wavefront velocity 1.05 ? 0.17 ?m/s; n ? 13). In each
of four preparations where the dye-loaded dendrite and
the astrocyte cell body were in the same focal plane,
the Ca2?wavefront reached the dendrite at the same
time that the SIC was detected in the pyramidal neuron.
Photolysis-induced SICs were of similar amplitude and
kinetics (Supplemental Table S1 [http://www.neuron.
org/cgi/content/full/43/5/729/DC1]; latency 12.2 ? 1.25
s) to those evoked by other stimuli which trigger [Ca2?]i
elevations in astrocytes. UV photolysis neither evoked
a calcium elevation in the stimulated astrocyte nor a
SIC in the associated pyramidal neuron (n ? 8) when
NP-EGTA was omitted.
To control for inadvertent stimulation of neuronal pro-
cesses, we repositioned our photolysis beam to directly
stimulate the neuronal dendrite. In this configuration,
photolysis fails to evoke a neuronal SIC (n ? 8), demon-
strating that the calcium wavefront within the astrocyte
is necessary for the generation of the neuronal SIC.
Using pairs of stimuli, we compared the relative SIC
amplitudes in the presence and absence of NMDAR
antagonists. In control conditions (0 Mg2?saline), the
SIC evoked by the second stimulus (P2) was similar in
amplitude to the response evoked by the first (P1). In
contrast, the addition of D-AP5 (50 ?M) or 1 mM Mg2?
significantly depressed the SIC amplitude with no effect
on the astrocytic [Ca2?]ielevation (Figures 3B and 3C).
Ifthe SICis dueto theastrocyte, repetitivestimulation
of the same cell should evoke neuronal responses with
a similar latency. Figure 3D shows that the stimulus-
response latencyfor pairs of photolysispulses was sep-
arated by 2 min; the first photolysis stimulus (P1) does
predict the SIC latency following the second photostim-
Stimulation of Individual Astrocytes Evokes SICs
in Adjacent Neurons
To test the causal link between [Ca2?]i increases in
astrocytes and the NMDAR-mediated response in neu-
astrocytes while measuring SICs in a CA1 pyramidal
neuron. Slices were loaded with the Ca2?indicator fluo-
4 AMand theCa2?cageNP-EGTA AM,which selectively
loads into astrocytes but not into neurons (Sul et al.,
Figure 3. Photolytic Elevation of [Ca2?]iin a Single Astrocyte Elicits an NMDA Receptor-Dependent SIC in an Adjacent Pyramidal Neuron
(A) Left image shows a fluo-4 and NP-EGTA-loaded astrocyte (green) adjacent to dendrites of an Alexa 568-filled CA1 pyramidal neuron (red).
Subsequent images display percent changes in fluo-4 intensity in pseudocolor with an overlay image (white) showing the location of the
pyramidal neuron dendrite. Images were taken at the times indicated by the dashed lines in the simultaneous current recording from the
neuron (lower trace). Delivery of a UV pulse, between image 1 and 2, elevates [Ca2?]iin the astrocyte cell body which slowly spreads through
the astrocyte processes. By the seventh image in the sequence, changes in astrocyte fluorescence were detected in the same pixels as
occupied by the neuronal dendrite (red pixels, arrows) that coincided with the onset of the SIC recorded in the neuron (lower trace).
(B) Pairs of photolysis pulses (P1 and P2) evoke similar amplitude SICs (top trace; 0 mM Mg2?). Application of D-AP5 (50 ?M; p ? 0.01) or 1
mM Mg2?(p ? 0.01) 5 min before P2 significantly reduced the SIC amplitude.
(C) Histograms showing the average (?SEM) ratio of SIC amplitude (P2/P1) from experiments discussed in (B).
(D) The latencies of SICs following P1 predict the SIC latency in response to P2 (r ? 0.92 p ? 0.01). The point labeled in red represents the
example shown in (A).
ulus (P2), supporting our contention that the SIC is
evoked by the calcium elevation in the astrocyte.
Together with the results obtained in response to the
other stimuli which mobilize astrocytic Ca2?and evoke
SICs, these data demonstrate a causal link between
astrocytic [Ca2?]i elevations and pyramidal neuronal
SICs and show that this current is mediated largely by
not significantly changed (Figures 4C and 4D). The
slower decay time is consistent with the delayed clear-
ance of glutamate in the extracellular space.
Extrasynaptic NMDARs Mediate SICs
The slow decay time of ?90% of SICs (Supplemental
Figure S4E [http://www.neuron.org/cgi/content/full/43/
5/729/DC1]) suggests that the NMDAR mediating SICs
is composed mainly of the NR1/NR2B complex, which
is characterized by slow offset kinetics (Dingledine et
al., 1999). To test the hypothesis that astrocytic gluta-
mate acts preferentially on the NR1/NR2B complex, we
used the selective NR2B antagonist ifenprodil (10 ?M),
which blocks 80% of the current mediated by the NR1/
NR2B complex (Williams, 1993). In the presence of 5–10
?M ifenprodil, the SIC amplitude was drastically and
reversibly reduced (Figure 4E), while the NMDA-medi-
ated EPSCs evoked by SC stimulation were affected
only slightly (Figures 4F and 4G).
Reverse Operation of Glutamate Transporters
Does Not Mediate SICs
To determine whether glutamate is released through a
reversal operation of glutamate transporters, experi-
ments were performed in the presence of the glutamate
transporter inhibitor DL-threo-?-benzyloxyaspartate
(TBOA) (Shimamoto et al., 1998). The increase in the
extracellular concentration of glutamate that follows
TBOA application generated a D-AP5-sensitive inward
current (?109.1 ? 20.3 pA; n ? 15; Figure 4A). Under
these conditions, astrocyte stimulation with DHPG still
evoked SICs in 4 of 14 pyramidal neurons (29%) that
were reversibly blocked by D-AP5 (Figures 4A and 4B).
The mean decay time of SICs was 3-fold slower than in
controls, while both mean rise time and amplitude were
Glutamate Released from Astrocytes Triggers
Synchronized Responses in Multiple Neurons
We frequently observed simultaneous transient [Ca2?]i
increases in several CA1 neurons following astrocytic
Astrocyte-Mediated Neuronal Synchrony
Figure 4. SICs Are Not Inhibited by the Glutamate Transporter Antagonist TBOA and Are Mediated Mainly by the Activation of NR1/NR2B
(A) In the presence of 100 ?M TBOA, 10 ?M DHPG still triggers SICs (left). In the same neuron, in the presence of D-AP5, a successive
stimulation with DHPG fails to evoke SICs (right). In the presence of TBOA, DHPG evoked SICs in 4 of 10 (40%) cells, with a mean frequency
of 1.3 ? 0.3 SICs/min. Scale bars, 50 pA and 5 s.
(B) Total number of SICs in the four responsive neurons in the presence of TBOA (50–100 ?M), TBOA plus D-AP5, and after TBOA washout.
(C) Average amplitude (n ? 17) and rise (n ? 15) and decay (n ? 17) time of SICs in the presence of TBOA. As a control, we reported the
average value of amplitude (n ? 259) and rise and decay time (n ? 202) of SICs recorded in in the absence of TBOA.
(D) Examples of normalized SICs under control conditions and in the presence of 100 ?M TBOA. Scale bar, 200 ms.
(E) Mean amplitude of DHPG- and low Ca2?-induced SICs in control condition (n ? 21, 2 cells), in the presence of 5–10 ?M ifenprodil (n ?
57; 5 cells) and after its washout (n ? 45, 5 cells). The inset shows representative examples of SICs under the different experimental conditions.
Scale bars, 50 pA and 200 ms.
(F) In four cells, including one of the cells included in (E), 10 ?M ifenprodil slightly but not significantly reduces the peak amplitude of NMDA
EPSCs. All EPSCs are recorded in the presence of 30 ?M NBQX.
(G) Representative experiment showing the time course of the NMDA EPSC amplitude at basal conditions, during 10 ?M ifenprodil application
and after its washout. The inset shows averages of 20 EPSCs in the different experimental conditions. Scale bars, 100 pA and 200 ms.
Movie 2 [http://www.neuron.org/cgi/content/full/43/5/
729/DC1]), the [Ca2?]ielevation in a stratum radiatum
astrocyteis followedbyan [Ca2?]iincreasein threeadja-
cent pyramidal neurons. This increase is short lasting
(2 s) and occurs synchronously in the three neurons
(Figure 5B). Paired recordings made from pyramidal
neurons while stimulating glutamate release from astro-
that SICs can occur in the two neurons with a high
degree of temporal correlation. Figure 5C shows that
only a few milliseconds separate the onset of synchro-
nous SICs that occur in two neurons spontaneously as
well as following DHPG stimulation. Note that the two
subsequent SICs that occur in the first neuron are ac-
companied by only one synchronized event in the sec-
ond neuron. Synchronized SICs are always character-
ized by the typical slow kinetics of SICs. Depolarizing
voltage pulses applied to each neuron of the pair re-
vealed no evidence of electrical coupling (Figure 5D;
n ? 8). At least two synchronized events were detected
in 10 of 25 pair recordings. The interevent time interval
histogram of SICs from a total of ten pairs demonstrates
that 23% of SICs (44 of 195) were synchronized within
a time window of 100 ms (Figure 5E). Such a level of
coincidence did not arise randomly among events from
paired recordings. Indeed, predictions obtained by a
Monte Carlo simulation or by an analysis of the Poisson
distribution demonstrate that in a 100 ms time window
the probability of detecting a pair of coincident events
is less than 0.001 and that to attain by chance the level
of SICs coincidence that we observed in our paired
recordings would require a time window of 30 s (Fig-
Figure 5. SICs from Distinct Neurons Can Occur with a High Degree of Synchronization
(A) [Ca2?]ichanges occurring in one astrocyte (1) and three neurons (2, 3, and 4) upon low Ca2?stimulation. Sampling rate, 2 s. Scale bar, 20 ?m.
(B) Time course of the response from the same cells indicated in (A). Scale bars, 0.1 (R), 60 s. Note that the Ca2?elevation in astrocyte 1
precedes the synchronous response in neurons 2, 3, and 4 (right). Scale bars, 0.1 (R), 8 s.
(C) Patch-clamp, paired recordings from two CA1 pyramidal neurons showing synchronized SICs after 20 ?M DHPG stimulation in 1 ?M TTX.
Scale bars, 200 pA and 50 s; inset 200 pA and 400 ms. The somata of the two neurons are 30 ?m apart.
(D) Same pair of neurons as in (C), showing that voltage steps (from ?80 to ?10 mV) applied to each cell do not propagate to the other cell
of the pair. Traces are not leak subtracted. Scale bars, 250 pA and 40 ms.
(E) Interevent time interval histogram of SICs occurring in the two neurons from ten pairs. A total number of 39 time intervals were counted
in a window of ?6 s. The majority of these (31/39) are restricted to a time window of ?1.5 s (shown in the figure). Note that 44 SICs, i.e., 22
of these time intervals, are restricted within a time window of 100 ms. Bin, 25 ms. The mean frequency of SICs in neurons from pair recordings
is 0.45 ? 0.08 events/min.
(F) Probability of observing coincident events according to a Monte Carlo simulation or analysis of the Poisson distribution. With a time
window of 100 ms and a frequency of events of 0.45 events/min, the probability of detecting a pair of coincident events is less than 0.001.
Monte Carlo simulation (?); probability of 1 or more coincidences (?); probability of exactly 1 coincidence (?).
(G) Paired recording showing a large-amplitude SIC in neuron 2 and a synchronized, small-amplitude SIC in neuron 1 triggered by 10 ?M
DHPG stimulation. Scale bars, 200 pA and 10 s; 200 pA and 2 s.
(H) Paired recording from two neurons showing spontaneous, synchronous SICs. Scale bars, 200 pA and 10 s; 200 pA and 400 ms. The
distance between the somata of the neurons from the pairs shown in (G) and (H) is 100 ?m.
The synchronized SICs could have very different am-
plitudes (Figures 5G and 5H), and large-amplitude events
detected in one neuron were not always accompanied
by a synchronized event from the other neuron of the
pair (Figure 5H, right).
Neuronal somata separated by up to 100 ?m could
release from astrocytes acts simultaneously on more
than two neurons. We thus performed additional confo-
cal imaging studies with relatively high temporal resolu-
tion (1 or 2 s). Slices were perfused with 1 ?M TTX and
astrocytes were stimulated with low Ca2?, DHPG, or
(RS)-2-chloro-5-hydroxyphenylglycine (CHPG), a spe-
cific agonist of the mGluR5 subtype. This compound
efficiently evoked [Ca2?]i oscillations in astrocytes as
well as [Ca2?]ielevations (Figure 6) and SICs in neurons
(see Supplemental Figure S2D [http://www.neuron.org/
cgi/content/full/43/5/729/DC1]). In Figures 6A and 6A1
(see also Supplemental Movie 3), following low Ca2?
stimulation, groups of neurons displayed three succes-
sive, short-lasting, synchronous [Ca2?]iresponses. Sev-
eral neurons participated in the three responses, and in
a synchronous [Ca2?]ielevation of one frame duration,
mental Movie 3). Repetitive responses from the same
neurons were frequently observed. In the example illus-
trated in Figures 6B and 6C, astrocyte stimulation with
1 mM CHPG first triggered in four neurons a synchro-
nous response (Figures 6B and 6B1) and then a short-
Astrocyte-Mediated Neuronal Synchrony
lasting [Ca2?]ielevation restricted to a dendrite together
with a synchronous [Ca2?]ielevation at the somata of
two other neurons (Figures 6C and 6C1; Supplemental
Movie 4). In all neurons from all experiments, synchro-
nous responses were abolished by D-AP5. The inter-
event time interval histogram of the neuronal response
from the experiment illustrated in Figures 6B and 6C
reveals that the large majority of [Ca2?]ielevations from
each neuron occurred in the same time frame with at
least one [Ca2?]ielevation from another neuron in the
field (Figure 6D). Only 22% (25 of 113) of the responses
occurred in solitary neurons. The majority of the re-
sponses in the astrocytes (65 of 72, 90%) were coinci-
dent with or preceded those in neurons (Figure 6E).
Results from a total of 22 experiments reveal that the
majorityof domainsarecomposed oftwoto fourcontig-
uous neurons, although domains comprising a higher
number of neurons are also present (Figure 6F). Figure
6G reports the maximal spatial extent of the domain
expressed as a function of the number of neurons in
the domain. Besides the expected positive correlation
between these two values, the graph reveals that two
of approximately 100 ?m can display synchronized re-
In one additional experiment in which a time acquisi-
ulation. From a total of four experiments with a time
resolution of 1 or 2 s, 70% ? 6% of [Ca2?]ielevations
of 2 s with at least one other neuron in the field of view.
another neuron in the field (Supplemental Figure S5A
DC1]). The majority of the responses in the astrocytes
(31 of 39, 75.5%) were coincident with or preceded that
in neurons (Supplemental Figure S5B). As expected, the
domain response was reversibly inhibited by D-AP5
of domains are composed of two to four neurons (mean,
2.56 ? 0.02), although domains comprising a higher
number of neurons are also present (Supplemental Fig-
ure S5E). As a whole, we observed delayed responses
in 47 of 133 neurons (35% ? 6%; n ? 4), and 59% ?
in synchrony with at least one other neuron in the field
of view. The mean amplitude and frequency of [Ca2?]i
elevations in these responsive neurons was 0.16 ? 0.02
(R405/485 change) and 1.3 ? 0.47 events/min, respec-
tively. SCstimulation triggered long-lasting[Ca2?]ioscil-
lations in 21 of 46 astrocytes (48% ? 12%, n ? 4).
Astrocyte-Mediated Neuronal Synchrony
under Physiological Conditions
We next investigated whether the domain response
cal conditions at 35?C in the presence of 1 mM extracel-
lular Mg2?, in the absence of picrotoxin, and in 1 ?M
TTX. Activation of astrocytes by either DHPG or PGE2
evoked SICs (Figures 7A and 7B, respectively), although
in a significantly lower percentage of neurons (10.7%,
n ? 56 versus 28%, n ? 98, at 1 mM and 0 Mg2?,
respectively, Fisher’s exact test, p ? 0.05) and lower
frequency (Figure 7C). Under these physiological condi-
tions, the mean rise and decay times of SICs were un-
changed, although their mean amplitude was signifi-
cantly reduced(Figure 7D).SC stimulationin 1mM Mg2?
also evoked SICs in 6 of 11 neurons (Figures 7E and
7F; Supplemental Table S2 [http://www.neuron.org/cgi/
Similar to results obtained in the absence of Mg2?, in
1 mM Mg2?SICs can occur in pairs of neurons with a
high degreeof temporalcorrelation (Figures7G and7H).
Depolarizing voltage pulses applied to each neuron of
the pair revealed no evidence of electrical coupling (Fig-
ure 7G1; n ? 3). In 7 of 13 pairs that displayed SICs, we
detected at least two synchronized events. In these 7
pairs of neurons, we observed a total of 67 SICs, and
30% of these events were synchronized within a time
synchronous SICs, we confirmed that astrocyte-medi-
larly detected under these conditions. Stimulation of
SCs in the presence of Mg2?evoked long-lasting [Ca2?]i
oscillations in 23 of 35 astrocytes (62% ? 6%, n ? 4)
in a similar manner to when stimulation was performed
in 0 Mg2?. The sequence of images in Figure 8A shows
the resulting [Ca2?]ielevations that were evoked in CA1
pyramidal neurons by SC stimulation (Figure 8A2) and a
delayed, synchronous [Ca2?]ielevation in a domain that
tal Movie 5 [http://www.neuron.org/cgi/content/full/43/
5/729/DC1]). The large majority of [Ca2?]ielevations (37
of 43, 86%) from each neuron occurred in the same time
Synaptic Activation of Astrocytes Evokes
Feedback Neuronal Synchronization
Since synaptic release of glutamate can trigger [Ca2?]i
oscillations in astrocytes (Pasti et al., 1997; Porter and
McCarthy, 1996), we asked whether activated astro-
cytes can signal back to neurons and trigger NMDAR-
dependent synchronized responses in neuronal do-
mains. This series of experiments (n ? 4) was performed
at physiological temperature (35?C) and in the absence
of Mg2?and picrotoxin. Application of short stimulus
trains (see Experimental Procedures) to SCs evoked in
CA1 pyramidal neurons [Ca2?]ielevations in response
to each applied stimulus (Figures 6H2 and 6I) and a
delayed [Ca2?]ielevation in astrocytes (cells 5, 6, and 7;
Figures 6H3and 6I). Turning off the stimulation while
applying TTX caused the neuron responses to cease
immediately. In contrast, [Ca2?]ioscillations in the astro-
cytes continued (Figure 6I). Our prediction was that
these [Ca2?]ioscillations could mediate the release of
glutamate from these cells and activate synchronized
responses in pyramidal neurons. In the example re-
ported, a synchronous [Ca2?]ielevation in a neuronal
6H4and 6I). One of these neurons displayed a second
response in synchrony with a contiguous neuron that
was not involved in the first response. The interevent
the majority of [Ca2?]ielevations (22 of 30, 73%) from
each neuron occurred in the same time frame (time ac-
quisition, 1 s) with at least one [Ca2?]ielevation from
Figure 6. Astrocytic Glutamate Triggers Synchronized [Ca2?]iElevations in Neuronal Domains
(A) Low Ca2?stimulation triggers synchronous [Ca2?]iincreases in nine pyramidal neurons in TTX. Sampling rate, 2 s. Scale bar, 20 ?m.
(A1) Time course of 405/485 changes for the neurons shown in (A). The inset marks the responsive neurons in the field (red) displaying
(B) CHPG-induced [Ca2?]ioscillations in one astrocyte (1) is followed by a synchronous response in four nearby neurons. Two of these are
indicated (2 and 3). Sampling rate, 2 s. Scale bar, 20 ?m. The time course of the 405/485 change from the astrocyte and two of the responsive
neurons is reported in (B1).
(C) Same field as in (B), illustrating the synchronous response from the soma of two neurons (4 and 5) and the dendrite of a third neuron (6)
to CHPG stimulation. Sampling rate, 2 s. Scale bar, 20 ?m. The time course of the 405/485 change for the dendrite and two other neurons is
shown in (C1).
Astrocyte-Mediated Neuronal Synchrony
frame (time acquisition, 1 s) with at least one [Ca2?]i
elevation from another neuron in the field (Figure 8C).
In comparison to conditions that favor the activation of
the NMDA receptor, these events were observed in a
reduced percentage of neurons (19 of 156, 12% ? 5%;
n ? 5; p ? 0.05) and at a lower frequency (0.48 ? 0.08
events/min; p ? 0.05), while their mean amplitude (405/
408 change, 0.19 ? 0.04) was not significantly changed.
The distribution of the number of neurons composing
the domain (Figure 8D) as well as the mean number of
neurons in the domains (3.7 ? 0.1) are also similar. From
four experiments with a time resolution of 1 s, 72% ?
10% (n ? 176) of [Ca2?]ielevations from an individual
neuron occurred within a time window of 1 s with at
least one other neuron. Thus, the presence of SICs and
the synchronous activation of groups of neurons can
be observed in more physiological conditions when
GABAergic synaptic transmission is intact and Mg2?is
present in the external saline.
ated response in neurons.
The slow rise and decay times that characterize SICs
may be due to a slow glutamate release mechanism,
the distance between the site of release and the target
membrane receptors, as well as the coefficient of diffu-
sion. We favor the possibility that these kinetics arise
from a slow increase in glutamate concentration in the
vicinity of the extrasynaptic neuronal receptors. Such
a slow increase in glutamate concentration can also
account for the absence of an AMPA component in
SICs. Indeed, when AMPA receptor desensitization was
reduced by addition of cyclothiazide, astrocyte activa-
tion evoked AMPA-mediated events with a slow rise
time. This observation also indicates that glutamate and
not aspartate is the principal mediator of SICs, since
the latter does not activate the AMPA receptor (Patneau
and Mayer, 1990).
The drastic reduction in SIC amplitude by the antago-
nist of the NR1/NR2B complex, ifenprodil, indicates that
SICs are mediated mainly by the NR1/NR2B receptor.
subunit dominates at the synapse, the NR2B subunit is
confined mainly to the extrasynaptic membrane (Rum-
baugh and Vicini, 1999; Tovar and Westbrook, 1999).
Our results thus suggest that the release of glutamate
leads to a slow yet selective activation of extrasynap-
The reverse operation of glutamate transporters (Att-
well et al., 1993) is not involved in the generation of
SICs, since they were still evoked in the presence of the
transporter inhibitor TBOA. While astrocytes are known
to release glutamate through various mechanisms (Hay-
don, 2001; Fellin and Carmignoto, 2004), a Ca2?-depen-
dent mechanism is involved in SIC generation. Indeed,
all the stimuli that activate [Ca2?]ielevations in astro-
cytes also trigger SICs and, most important, photolysis
neurons. While further studies will be necessary for a
full clarification of the cellular mechanism underlying
glutamate release that generates SICs, our results are
compatible with the recent demonstration that astro-
cytes possess a mechanism for a regulated, vesicular
release of this transmitter (Bezzi et al., 2004).
Astrocytic glutamate-mediated responses were ob-
served only in a subpopulation of neurons. While this
may be due to the fact that the stimuli that we provided
only activated a subpopulation of astrocytes, it is also
important to appreciate that astrocytic processes are
not uniformly distributed around the synapses (Ventura
Glutamate Released from Astrocytes
Triggers Slow NMDAR Responses
in CA1 Pyramidal Neurons
The inward currents generated by the activation of
NMDARs described in our study are different from syn-
aptic EPSCs because they have one order of magnitude
slower rise times, never have an AMPA component, and
are mediated mainly by the NR1/NR2B complex. Block-
ade of neuronal activity and synaptic transmission with
TTX and TeNT allows us to rule out synaptic release
from axon terminals as a source of the neurotransmitter,
including asynchronous release that, in principle, may
be responsible for the slow kinetics of SICs (Diamond
and Jahr, 1995). Accordingly, the most likely source of
this glutamate is from elements that are in proximity to
the neuronal membrane, such as the astrocytic pro-
cesses. This conclusion is supported by the observa-
tions that (1) stimuli that evoke [Ca2?]ioscillations and
glutamate release from astrocytes trigger NMDAR-
mediated SICs and [Ca2?]ielevations in CA1 pyramidal
neurons; (2) stimulation of SCs, which evokes [Ca2?]i
SICs and [Ca2?]i elevations in neuronal domains; (3)
[Ca2?]ielevations in activated astrocytes precede or are
coincident with [Ca2?]ielevations in pyramidal neurons;
and (4) selective stimulation of [Ca2?]ielevations in indi-
triggers SICs in nearby neurons. This last observation
provides conclusive evidence for a causal link between
(D) Interevent time interval histogram for the [Ca2?]ielevations in neurons (bin, 2 s) in the experiment shown in (B)–(C).
(E) Interevent time interval histogram reporting the timing for the [Ca2?]ielevation in each astrocytes with respect to the timing for the [Ca2?]i
elevation in neurons (bin, 2 s) in the same experiment shown in (B)–(C). The average time interval between the astrocytic and the neuronal
response is 2.7 ? 0.5 s.
(F) Histogram of the number of responsive neurons in the domain after stimulation with either DHPG, CHPG, or low Ca2?from 22 experiments.
(G) Maximal spatial extension of domains as a function of the number of neurons in the domain which display synchronous response to
DHPG/CHPG or low Ca2?stimulation.
(H) Pseudocolor images showing [Ca2?]ielevations in neurons due to SC stimulation (H2), delayed [Ca2?]ielevations in astrocytes (H3), and a
synchronous response in three pyramidal neurons (H4). Sampling rate, 1 s. Scale bar, 20 ?m. These experiments were performed at 35?C in
1 ?M TTX and in the absence of picrotoxin.
(I) Time course of 405/485 changes from the four neurons and the three astrocytes indicated in (H1).
Figure 7. Astrocyte Activation in 1 mM Mg2?Triggers Solitary as well as Synchronized SICs
(A) DHPG-evoked SIC at 35?C, 1 mM extracellular Mg2?, in TTX and in the absence of picrotoxin. DHPG triggers SICs in 8 of 74 CA1 neurons.
Scale bars, 50 pA and 5 s; 50 pA and 400 ms.
(B) SICs evoked by stimulation with 5 ?M PGE2, under the same experimental conditions as in (A). PGE2triggers SICs in 3 of 19 CA1 neurons.
Scale bars, 50 pA and 5 s; 50 pA and 400 ms.
(C) Average number of SICs/min following DHPG stimulation in either 0 mM Mg2?(n ? 27) or in 1 mM Mg2?(n ? 8). The two values are not
significantly different (p ? 0.06).
(D) Average amplitude and rise and decay time for spontaneous and evoked SICs in 1 mM Mg2?(n ? 83). As a control, we reported the
average value of amplitude (n ? 259) and rise and decay time (n ? 202) of spontaneous and evoked SICs recorded in 0 extracellular Mg2?.
(E) Stimulation of SCs evokes SICs in 1 mM Mg2?(bottom, individual SICs at expanded time scale; top, EPSCs triggered by a stimulus train).
(F) Average frequency of SICs before and following SC stimulation (n ? 6).
(G) Synchronous (left) and solitary (right) SICs evoked by astrocyte stimulation with DHPG in 1 ?M TTX. Scale bars, 50 pA and 5 s; 50 pA and
400 ms. The somata of the two neurons were 70 ?m apart.
(G1) Same pair of neurons as in (G), showing the trace following voltage steps (from ?80 to 0 mV) applied to each cell of the pair. Scale bars,
250 pA and 40 ms.
SICs. The somata of the two neurons were 10 ?m apart. Scale bars, 50 pA and 5 s; 50 pA and 200 ms.
(I) Interevent time interval histogram of SICs occurring in the two neurons from seven distinct dual recordings. A total of 19 time intervals
were counted in a window of ?6 s. The majority of these (18/19) are restricted to a time window of ?1.5 s (shown in the figure). Note that 20
SICs, i.e., 10 of these time intervals, are restricted within a time window of 100 ms. Bin, 50 ms. The mean frequency of SICs in neurons from
pair recordings is 0.22 ? 0.07 events/min.
Astrocyte-Mediated Neuronal Synchrony
Figure 8. Neuronal Activity-Dependent Domain Responses under Physiological Conditions
(A) Images showing synchronous [Ca2?]ielevations in neurons following stimulation of the SCs (A2) and a synchronous delayed response from
eight contiguous pyramidal neurons (A4) under physiological conditions; scale bar, 20 ?m. Note that no astrocytes could be visualized in this
field at the same focal plane of responsive neurons.
(B) Time course of 405/485 changes from five of the neurons composing the domain.
(C) Interevent time interval histogram for the [Ca2?]ielevations in neurons (bin, 1 s) from the same experiment.
(D) Histogram of the number of responsive neurons in the domain from five experiments.
and Harris, 1999) and that hippocampal astrocytes rep-
ent functional properties (Matthias et al., 2003).
glutamate release from two distinct sites, of the same
or different astrocytes impinging on the two neurons.
These two release episodes can be triggered by an
[Ca2?]ielevation occurring simultaneously in two astro-
cytic processes from the same or from two different
The presence of glutamate release sites in astrocytes
at different distances from the recorded neuron could
account for the variability in the rise and decay times
of SICs. Additionally, the observation that two or more
SICs recorded in the same neuron can have strikingly
different kinetics suggests the presence of multiple re-
lease sites, from either one or many astrocytes imping-
ing onto an individual CA1 neuron.
The complexity of astrocyte-to-neuron communica-
tion that emerges from these observations is further
emphasized by the results obtained in confocal micros-
copy experiments in which [Ca2?]ielevations triggered
neously in two to twelve CA1 neurons. An [Ca2?]ieleva-
tion occurring in a solitary neuron was rare. This sug-
gests that activation of multiple neurons by astrocytic
glutamate isa common featureof this formof astrocyte-
neuron crosstalk. With respect to the percentage of
synchronized [Ca2?]ielevations in neurons (75%), the
percentage of synchronized SICs is apparently lower
(23%). This discrepancy is most likely due to the fact
the Ca2?response of tens of neurons simultaneously
and thus have, with respect to paired recording experi-
ments, a higher probability of observing a synchronized
response in any two neurons in the field.
Glutamate Released from Astrocytes
Synchronizes Activity in Neuronal Domains
The synchronized activation of groups of neurons medi-
ated by extrasynaptic NMDARs represents one of the
main findings of this study. In paired recordings, we
observed that SICs could occur in two neurons with a
high degree of temporal correlation. Such a high syn-
chronization cannot derive from spreading of the cur-
rent through gap junctions, since we never detected
signs of electrotonic coupling between neurons which
display synchronized responses. Additionally, synchro-
nized events cannot be due to activation of NMDARs
affected by TTX, which blocks action potential dis-
and because TeNT, which blocks synaptic release of
glutamate, does not impair SICs. It is unlikely that the
ates neuronal synchrony because this wave propagates
at a speed of approximately 10 ?m/s (Sul et al., 2004).
The most likely explanation for synchronized events is
that they are derived from a single episode of astrocytic
glutamate releasethat activatesextrasynaptic NMDARs
from the adjacent dendrites of neurons that are close
the extracellular volume. An alternative, although not
mutually exclusive, hypothesis is that a pair of synchro-
nized SICs derives from two synchronous episodes of
Importantly, SICs as well as synchronized [Ca2?]iele-
vations in neuronal domains were also observed when
[Ca2?]ioscillations in astrocytes were triggered by in-
when neuronal activity is high, activated astrocytes can
signal back to neurons and synchronize activity in dis-
tinct neuronal domains. Such a response may represent
a physiological signaling mode of astrocyte-to-neuron
communication in the brain. The recent evidence for
the presence of [Ca2?]ioscillations in astrocytes in vivo
astrocyte interactions in brain function. While episodes
of high neuronal activity similar to those that in our
experiments activate the astrocyte response are com-
monly used to study the plasticity of synaptic transmis-
sion in brain slices (Bliss and Collingridge, 1993; Hu-
meau et al., 2003; Malinow et al., 2000; Wang et al.,
2003), it is conceivable that they might not be compara-
ble to those that occur under physiological conditions
in vivo. Future experiments should elucidate this issue.
Activation of astrocytes by neuronal activity triggered
domain responses in neurons also in the presence of
1 mM Mg2?, although with a reduced probability com-
pared to those from experiments that were carried out
in the absence of Mg2?. Furthermore, activation of
astrocytes in 1 mM Mg2?by DHPG, PGE2, or neuronal
activity still activates SICs. These observations further
emphasize the physiological relevance of this form of
This ability of astrocytic glutamate to activate NMDARs
at a physiological concentration of Mg2?was similarly
observed in the ventrobasal thalamus (Parri et al., 2001).
While the mechanism leading tothe removal of the Mg2?
block is unclear, several possibilities can be proposed.
For example, in addition to glutamate, astrocytes re-
lease D-serine (Schell et al., 1995), an effective ligand
for the glycine binding site of the NMDA receptor, which
may enhance the opening of the channel. Other factors
released from astrocytes, such as ATP (Guthrie et al.,
1999), may contribute to the removal of the Mg2?block
by depolarizing the neuronal membrane. Furthermore,
larization, an action of astrocytic glutamate on this neu-
ronal receptor may also contribute. After the discovery
that astrocytes can be activated by synaptic release of
various neurotransmitters and havethe ability to release
neuroactive molecules, such as glutamate (Bezzi et al.,
1998; Haydon, 2001; Parpura et al., 1994; Pasti et al.,
1997) and ATP (Guthrie et al., 1999), several studies
revealed at least some of the functional roles which may
be ascribed to the reciprocal communication between
these glial cells and neurons, including the control of
the neurovascular coupling (Zonta et al., 2003) and the
modulation of inhibitory transmission in the hippocam-
pus (Kang et al., 1998; Liu et al., 2004). In the same brain
region, astrocytic glutamate mediates an increase in
the probability of spontaneous glutamate release from
glutamatergic axon terminals, by acting on extrasynap-
tic mGluR receptors (Fiacco and McCarthy, 2004). This
tial action of astrocytic glutamate on extrasynaptic re-
ceptors. The results we report reveal a hitherto un-
recognized role of astrocytes in promoting coordinated
activity of distinct subsets of CA1 pyramidal neurons.
Synchronization of neuronal activity is hypothesized
to be of fundamental relevance to information pro-
cessing in the brain (Singer, 1999). Although no consen-
sus has yet emerged regarding its cellular mechanism,
synchrony is believed to arise from the convergence
of excitatory synaptic inputs and inhibitory interactions
within the neuronal network (Harris et al., 2003). We
show here that NMDAR activation by astrocytic gluta-
mate represents an additional mechanism for neuronal
synchrony. This mechanism can be operative under se-
lected conditions, when more intense synaptic activity
can increasingly activate the release of glutamate from
astrocytes (Pasti et al., 1997). By cooperating with the
excitatory synaptic inputs to recruit specific subsets
of neurons in the neuronal network, the activation of
extrasynaptic NMDA receptors by astrocytic glutamate
may represent a flexible, additional mechanism that fa-
vors the formation of dynamically associated assem-
blies of neurons.
While this NMDA receptor-mediated signaling be-
tween astrocytes and neurons may contribute to the
overall dynamics of neuronal synchrony, its very pres-
ence raises a series of questions on its possible role
in pathological changes in the hippocampus, such as
excitotoxicneuronal damage(Choi,1988)or thegenera-
tion of epileptiform activity (Dingledine et al., 1990).
In conclusion, we reveal that glutamate released from
astrocytes acts on extrasynaptic NMDA receptors to
promote synchronized activity in distinct neuronal do-
mains in the CA1 hippocampal region. These results
raise the possibility that astrocytes contribute to the
formation of the basal functional module in brain infor-
Transverse hippocampal slices (300–400 ?m) were prepared from
Wistar rats at postnatal days 10–22 as described (Edwards et al.,
1989; Pasti et al., 1997). Slices were cut with a Leica VT1000S vibra-
tome and incubated at 37?C for a recovery period of at least 1 hr.
All experiments were performed within 3 hr after the recovery. The
physiological saline for slice cutting and incubation was NaCl, 120
mM; KCl, 3.2 mM; NaH2PO4, 1 mM; NaHCO3, 26 mM; MgCl2, 2 mM;
CaCl2, 1 mM; glucose, 2.8 mM; Na-pyruvate, 2 mM; and ascorbic
acid, 0.6 mM at pH 7.4 with O295%, CO25%. In the experiments
that used picrotoxin to block GABAergic inhibition (see below), the
connection between the CA3 and CA1 regions was cut to overcome
the spreading of epileptic-like activity. Slices for confocal micros-
copy were loaded with the Ca2?indicator indo-1/AM (Molecular
Probes, Eugene, OR) and 0.02% pluronic for 50 min under mild
stirring at 37?C. Slice incubation with TeNT (2 ?M, for 2 hr) was
Patch-Clamp Recordings and Analysis
Sliceswereput intherecordingchamber andcontinuouslyperfused
with NaCl, 120 mM; KCl, 3.2 mM; NaH2PO4, 1 mM; NaHCO3, 26 mM;
CaCl2, 2 mM; glucose, 2.8 mM; glycine, 1 mM; at pH 7.4 with O2
95%,CO25%. TheexperimentsinFigures 7and8were performedin
thepresenceof 1mMMg2?intherecording saline.Unlessotherwise
specified, picrotoxin (100 ?M, Sigma) was added to the saline to
block GABAa-mediated inhibition. Low Ca2?solution was obtained
by replacing CaCl2with EGTA (0.25 mM). Typical pipette resistance
was 3–4 M?. Intracellular pipette solution was K-Gluconate, 145
mM; MgCl2, 2 mM; EGTA, 5 mM; Na2ATP, 2 mM; NaGTP, 0.2 mM;
HEPES, 10 mM; to pH 7.2 with KOH. Patch-clamp recordings were
performed using standard procedures and one or two Axopatch-
200B amplifiers (Axon Instruments, Union City, CA). Data were fil-
Astrocyte-Mediated Neuronal Synchrony
tered at 1 KHz and sampled at 5 KHz with a Digidata 1200 interface
and pClamp software (Axon instruments). Neurons were voltage
clamped at ?60 mV, unless otherwise stated. Evoked postsynaptic
currents were triggered using a bipolar tungsten electrode (FHC,
Bowdoinham, ME) positioned at the stratum radiatum to stimulate
the SC pathway, 100–200 ?m from the cell of interest. Single pulses
(100 ?s duration, 0.2–1 mA) were applied at 0.2 Hz. To study evoked
synaptic tranmission in slices incubated with TeNT, the intensity of
the stimulus applied to neuronal afferents was increased to 5–10
mA. To trigger SICs, stimuli were delivered to SCs at 25–30 Hz (100
or 200 ms duration trains repeated at 0.3 or 1 Hz for 10–30 s, 0.2–0.5
mA). Experiments were performed either at room temperature or at
35?C. Data analysis and fitting were performed with Clampfit 8.2
(Axon instrument) and Origin 6.0 (Microcal Software, Northampton,
at the peak; rise time was calculated with the 20%–80% criterion,
and the decay time as the time constant of a single or double
exponential fit. Inward currents with rise time slower than 10 ms
and amplitude greater than ?20 pA were classified as SICs. SICs
with an amplitude smaller than ?20 pA or rise time faster than 10 ms
were analyzed only when they occurred synchronously in neurons
from the pair recording experiments. Due to the difficulty in applying
a reliable kinetic analysis, some of these events were only consid-
ered for the calculation of the mean amplitude. Cells displaying an
increased number of SICs during the period of stimulation with
respect to an equal time period before stimulation were considered
responsive. The frequency of spontaneous SICs was measured in
these responsive neurons. The interevent time interval between two
SICs was calculated as the time interval between the onset of the
current in cell 1 and the onset of the current in cell 2. Data are
expressed as mean ? SEM. Predictions obtained by a Monte Carlo
simulation or by analysis of the Poisson distribution were used to
estimate the probability of randomly occurring coincident SICs in
our paired recordings. If the events are noncorrelated and occur
randomly, the probability of each event is independent of the state
of the other, and the probability density D of SICs in each neuron
is estimated by N/T (N ? total number of events recorded in the
p(n) ? exp (?D ? ?) ? (D ? ?)n/n!and is plotted in the figure for both
arrival of events on two channels with probability density of 0.0075
event/s by means of a Monte Carlo method. The probability of the
simultaneous occurrence of two events as a function of the ampli-
tude of the coincidence window is in agreement with the Poisson
model and it is also plotted in the figure.
the astrocytic and the neuronal response is obtained by measuring
the timing for the [Ca2?]ielevation in each astrocyte with respect to
that for the [Ca2?]ielevation in each neuron of the field.
In some experiments, photolysis was performed in conjunction with
wide-field fluorescence imaging and electrophysiology. Similar ap-
proaches to those discussed above were used but employed 8–13
day Swiss Webster mice. Since SICs that were similar in properties
to those detected in rat slices were seen in recordings from these
murine slices in response to activation of metabotropic receptors
(data not shown), we pooled data from both of these rodents in this
study. Hippocampal slices were prepared as detailed previously
(Sul et al., 2004), and astrocytes were bulk loaded for 2 hr at room
temperature in ACSF containing fluo-4 AM (12.5 ?g/ml), NP-EGTA
(25 ?g/ml), DMSO (0.1%), and pluronic (0.05%) saturated with O2
95%, CO25%. Since greater than 98% of the calcium indicator-
loaded cells were astrocytes, we used the presence of the indicator
loading to determine cell identity (Sul et al., 2004). Parallel whole-
were recorded with the patch-clamp technique using an Axoclamp
amplifier 1C. Pipettes had a resistance of 4–5 M? when loaded with
a solution containing K-Gluconate, 130 mM; CaCl2, 1 mM; MgCl2,
2 mM; EGTA, 11 mM; MgATP, 1.5 mM; NaGTP, 0.3 mM; HEPES, 10
mM; to pH 7.2 with KOH. The fluorescent dye Alexa 568 (0.1 mM)
(Molecular Probes) was added to this solution to visualize the den-
drites of the recorded neuron. Photorelease of Ca2?was performed
by a 3 ?m diameter UV pulse (351 and 364 nm) generated by an
argon ion laser (Coherent Enterprise II; duration 100 ms, power
200–250 ?W) connected by an optical fiber to an Uncager system
(Prairie Technologies, Inc., Middleton, WI). Images were acquired
using a Q imaging cooled CCD camera and Image-Pro software.
The UV pulse, camera acquisitions, and electrophysiology were all
controlled by pClamp software (version 9.0; Axon Instruments) con-
nected to a Digidata 1332A. Analysis of the images was performed
using Metamorph software (Universal Imaging Corp., Downing-
DHPG, CHPG,D-AP5, MK-801, NBQX, cyclothiazide,DL-TBOA, and
LY 367385 were obtained from Tocris Cookson (Buckhurst Hill, UK);
Sigma (Milan, Italy); PGE2was from Biomol (Plymouth Meeting, PA).
Purified TeNT was kindly provided by Dr. O. Rossetto, Department
of Experimental Biomedical Sciences, University of Padova, Pa-
A confocal microscope (Nikon RCM8000) was used for monitoring
the [Ca2?]ichange at the single-cell level as previously described
(Pasti et al., 1997). Slices were continuously perfused with the same
extracellular solution that was used in electrophysiological re-
cording with sulfinpyrazone (0.2 mM). The sampling rate was 1–4 s
and 16 to 32 images were averaged for each frame. To activate
[Ca2?]ielevationsin astrocytes,weusedthe samestimulationproto-
col that in patch-clamp experiments triggered SICs. Experiments
were performed at either room temperature or 35?C. Cells in the
focal plane 10–30 ?m beneath the surface of the slice were moni-
tored. Cells located at different depths displayed a similar value of
R405/485, indicating that the neurons and astrocytes under study
were not damaged by slicing procedures. Neurons and astrocytes
were distinguished on the basis of the distinct kinetics of their re-
sponse to high K?stimulation, which, as previously reported (Pasti
et al., 1997), was always performed at the end of the experiment.
The maximal extension of the neuronal domain that displayed a
synchronized response was estimated as the distance between the
centers of the somata of the two neurons positioned at the borders
of the responsive domain. The interevent time interval of [Ca2?]i
elevations in neurons was estimated by comparing the timing for
the [Ca2?]i response from each responsive neuron within a time
window of 12 s. In the experiments that used an acquisition time
interval of 1 or 2 s, responses were classified as synchronous when
they occurred at the same time frame. The time interval between
We thank Micaela Zonta for helpful discussion and for the prepara-
experiments in an initial stage of our study; Gian Michele Ratto for
simulationanalysis;and YolandeHaydonforediting ourmanuscript.
This work was supported by grants from the Armenise-Harvard Uni-
RBNE01RHZM_003) and ST/Murst: “Neuroscienze” to G.C., the Ital-
ian Association for Cancer Research (AIRC), and European Commu-
nity (QLG3-CT-2000-00934) to T.P., and the NIH (RO1 NS43142; R37
NS37585; P20-MH-071705) (to P.G.H.). The authors of this paper
have declared a conflict of interest. For details, go to http://www.
Received: February 24, 2004
Revised: July 26, 2004
Accepted: August 6, 2004
Published: September 1, 2004
Araque, A., Li, N., Doyle, R.T., and Haydon, P.G. (2000). SNARE
protein-dependent glutamate release from astrocytes. J. Neurosci.
Arcuino, G., Lin, J.H.-C., Takano, T., Liu, C., Jiang, L., Gao, Q.,
Kang, J., and Nedergaard, M. (2002). Intercellular calcium signaling
mediated by point-source burst release of ATP. Proc. Natl. Acad.
Sci. USA 99, 9840–9845.
Haydon, P.G. (2001). GLIA: listening and talking to the synapse. Nat.
Rev. Neurosci. 2, 185–193.
Hirase, H., Qian, L., Bartho ´, P., and Buzsa ´ki, G. (2004). Calcium
dynamics of cortical astrocytic networks in vivo. PLoS Biol. 2(4):
e96 DOI: 10.1371/journal.pbio.0020096.
Humeau, Y., Shaban, H., Bissie `re, S., and Lu ¨thi, A. (2003). Presynap-
tic induction of heterosynaptic associative plasticity in the mamma-
lian brain. Nature 426, 841–845.
Kang, J., Jiang, L., Goldman, S.A., and Nedergaard, M. (1998).
Nat. Neurosci. 1, 683–692.
Kullmann, D.M. (1999). Synaptic and extrasynaptic roles of gluta-
mate in the mammalian hippocampus. Acta Physiol. Scand. 166,
Liu, Q.-S., Xu, Q., Arcuino, G., Kang, J., and Nedergaard, M. (2004).
Astrocyte-mediated activation of neuronal kainate receptors. Proc.
Natl. Acad. Sci. USA 101, 3172–3177.
Malinow, R., Mainen, Z.F., and Hayashi, Y. (2000). LTP mechanisms:
from silence to four-lane traffic. Curr. Opin. Neurobiol. 10, 352–357.
Matthias, K., Kirchhoff, F., Seifert, G., Hu ¨ttmann, K., Matyash, M.,
Kettenmann, H., and Steinha ¨user, C. (2003). Segregated expression
of AMPA-type glutamate receptors and glutamate transporters de-
fines distinct astrocyte populations in the mouse hippocampus. J.
Neurosci. 23, 1750–1758.
Parpura, V., Basarsky, T.A., Liu, F., Jeftinija, K., Jeftinija, S., and
Haydon, P.G. (1994). Glutamate-mediated astrocyte-neuron signal-
ling. Nature 369, 744–747.
Parri, H.R., Gould, T.M., and Crunelli, V. (2001). Spontaneous
astrocytic Ca2?oscillations in situ drive NMDAR-mediated neuronal
excitation. Nat. Neurosci. 4, 803–812.
lular calcium oscillations in astrocytes: a highly plastic, bidirectional
form of communication between neurons and astrocytes in situ. J.
Neurosci. 17, 7817–7830.
Pasti, L., Zonta, M., Pozzan, T., Vicini, S., and Carmignoto, G. (2001).
Cytosolic calcium oscillations in astrocytes may regulate exocytotic
release of glutamate. J. Neurosci. 21, 477–484.
Patneau, D.K., and Mayer, M.L. (1990). Structure-activity relation-
ships for amino acid transmitter candidates acting at N-methyl-
D-aspartate and quisqualate receptors. J. Neurosci. 10, 2385–2399.
Porter, J.T., and McCarthy, K.D. (1996). Hippocampal astrocytes
in situ respond to glutamate released from synaptic terminals. J.
Neurosci. 16, 5073–5081.
Rumbaugh, G., and Vicini, S. (1999). Distinct synaptic and extrasyn-
aptic NMDA receptors in developing cerebellar granule neurons. J.
Neurosci. 19, 10603–10610.
Schell, M.J., Molliver, M.E., and Snyder, S.H. (1995). D-serine, an
endogenous synaptic modulator: localization to astrocytes and glu-
tamate-stimulated release. Proc. Natl. Acad. Sci. USA 92, 3948–
Schiavo, G., Benfenati, F., Poulain, B., Rossetto, O., Polverino de
Laureto, P., DasGupta, B.R., and Montecucco, C. (1992). Tetanus
teolytic cleavage of synaptobrevin. Nature 359, 832–835.
Scimemi, A., Fine, A., Kullmann, D.M., and Rusakov, D.A. (2004).
NR2B-containing receptors mediate cross talk among hippocampal
synapses. J. Neurosci. 24, 4767–4777.
Shimamoto, K., Lebrun, B., Yasuda-Kamatani, Y., Sakaitani, M.,
Shigeri, Y., Yumoto, N., and Nakajima, T. (1998). DL-threo-beta-
porters. Mol. Pharmacol. 53, 195–201.
tion of relations? Neuron 24, 49–65.
Stocca, G., and Vicini, S. (1998). Increased contribution of NR2A
subunit to synaptic NMDA receptors in developing rat cortical neu-
rons. J. Physiol. 507, 13–24.
Sul, J.-Y., Orosz, G., Givens, R.S., and Haydon, P.G. (2004).
Astrocytic connectivity in the hippocampus. Neuron Glia Biol. 1,
Attwell, D., Barbour, B., and Szatkowski, M. (1993). Nonvesicular
release of neurotransmitter. Neuron 11, 401–407.
Bezzi, P., Carmignoto, G., Pasti, L., Vesce, S., Rossi, D., Rizzini,
B.L., Pozzan, T., and Volterra, A. (1998). Prostaglandins stimulate
calcium-dependent glutamate release in astrocytes. Nature 391,
Bezzi, P., Gundersen, V., Galbete, J.L., Seifert, G., Steinha ¨user, C.,
Pilati, E., and Volterra, A. (2004). Astrocytes contain a vesicular
compartment that is competent for regulated exocytosis of gluta-
mate. Nat. Neurosci. 7, 613–620.
Bliss, T.V.P., and Collingridge, G.L. (1993). A synaptic model of
memory: long-term potentiation in the hippocampus. Nature 361,
Bourne, H.R., and Nicoll, R. (1993). Molecular machines integrate
coincident synaptic signals. Cell Suppl. 72, 65–75.
astrocytes and neurones. Prog. Neurobiol. 62, 561–581.
Choi, D.W. (1988). Glutamate neurotoxicity and diseases of the ner-
vous system. Neuron 1, 623–634.
Choi, D.W., and Rothman, S.M. (1990). The role of glutamate neuro-
toxicity in hypoxic-ischemic neuronal death. Annu. Rev. Neurosci.
Congar, P., Leinekugel, X., Ben-Ari, Y., and Cre ´pel, V. (1997). A long-
lasting calcium-activated nonselective cationic current is generated
by synaptic stimulation or exogenous activation of group I metabo-
tropic glutamate receptors in CA1 pyramidal neurons. J. Neurosci.
Constantine-Paton, M., Cline, H.T., and Debski, E. (1990). Patterned
activity, synaptic convergence, and the NMDA receptor in devel-
oping visual pathways. Annu. Rev. Neurosci. 13, 129–154.
Conti, F., and Weinberg, R.J. (1999). Shaping excitation at gluta-
matergic synapses. Trends Neurosci. 22, 451–458.
Cre ´pel, V., Aniksztejn, L., Ben-Ari, Y., and Hammond, C. (1994).
Glutamate metabotropic receptors increase a Ca2?-activated non-
specific cationic current in CA1 hippocampal neurons. J. Neuro-
physiol. 72, 1561–1569.
Diamond, J.S., and Jahr, C.E. (1995). Asynchronous release of syn-
aptic vesicles determines the time course of the AMPA receptor-
mediated EPSC. Neuron 15, 1097–1107.
Dingledine, R., McBain, C.J., and McNamara, J.O. (1990). Excitatory
amino acid receptors in epilepsy. Trends Pharmacol. Sci. 11,
Dingledine, R., Borges, K., Bowie, D., and Traynelis, S.F. (1999). The
glutamate receptor ion channels. Pharmacol. Rev. 51, 7–61.
Edwards, F.A., Konnerth, A., Sakmann, B., and Takahashi, T. (1989).
A thin slice preparation for patch clamp recordings from neurones
of the mammalian central nervous system. Pflugers Arch. 414,
Fellin, T., and Carmignoto, G. (2004). Neuron-to-astrocyte signaling
in the brain represents a distinct multifunctional unit. J. Physiol.
Fiacco, T.A., and McCarthy, K.D. (2004). Intracellular astrocyte cal-
cium waves in situ increase the frequency of spontaneous AMPA
receptor currents in CA1 pyramidal neurons. J. Neurosci. 24,
Guthrie, P.B., Knappenberger, J., Segal, M., Bennett, M.V., Charles,
A.C., and Kater, S.B. (1999). ATP released from astrocytes mediates
glial calcium waves. J. Neurosci. 19, 520–528.
Hardingham, G.E., Fukunaga, Y., and Bading, H. (2002). Extrasynap-
tic NMDARs oppose synaptic NMDARs by triggering CREB shut-
off and cell death pathways. Nat. Neurosci. 5, 405–414.
Harris, K.D., Csicsvari, J., Hirase, H., Dragoi, G., and Buzsa ´ki, G.
(2003). Organization of cell assemblies in the hippocampus. Nature
Astrocyte-Mediated Neuronal Synchrony Download full-text
Tovar, K.R., and Westbrook, G.L. (1999). The incorporation of NMDA
receptors with a distinct subunit composition at nascent hippocam-
pal synapses in vitro. J. Neurosci. 19, 4180–4188.
Tovar, K.R., and Westbrook, G.L. (2002). Mobile NMDA receptors
at hippocampal synapses. Neuron 34, 255–264.
between hippocampal synapses and astrocytes. J. Neurosci. 19,
Wang, Z., Xu, N.-L., Wu, C.-P., Duan, S., and Poo, M.-M. (2003).
Bidirectional changes in spatial dendritic integration accompanying
long-term synaptic modifications. Neuron 37, 463–472.
Williams, K. (1993). Ifenprodil discriminates subtypes of the
N-methyl-D-aspartate receptor: selectivity and mechanisms at re-
combinant heteromeric receptors. Mol. Pharmacol. 44, 851–859.
Yamada, K.A., and Tang, C.-M. (1993). Benzothiadiazides inhibit
rapid glutamate receptor desensitization and enhance gluta-
matergic synaptic currents. J. Neurosci. 13, 3904–3915.
Zanotti, S., and Charles, A. (1997). Extracellular calcium sensing by
glial cells: low extracellular calcium induces intracellular calcium
release and intercellular signaling. J. Neurochem. 69, 594–602.
Zonta, M., Angulo, M.C., Gobbo, S., Rosengarten, B., Hossmann,
K.-A., Pozzan, T., and Carmignoto, G. (2003). Neuron-to-astrocyte
signaling is central to the dynamic control of brain microcirculation.
Nat. Neurosci. 6, 43–50.