silver 2003.pdf - Alice

R EPORTS
High-Probability Uniquantal
Transmission at Excitatory
Synapses in Barrel Cortex
R. Angus Silver, 1 *† Joachim Lübke, 2 Bert Sakmann, 3
Dirk Feldmeyer 3 *
The number of vesicles released at excitatory synapses and the number of
release sites per synaptic connection are key determinants of information
processing in the cortex, yet they remain uncertain. Here we show that the
number of functional release sites and the number of anatomically identified
synaptic contacts are equal at connections between spiny stellate and
pyramidal cells in rat barrel cortex. Moreover, our results indicate that the
amount of transmitter released per synaptic contact is independent of
release probability and the intrinsic release probability is high. These properties
suggest that connections between layer 4and layer 2/3 are tuned for
reliable transmission of spatially distributed, timing-based signals.
Synaptic weight is thought to be a primary
determinant of neural computation. According
to the quantal hypothesis (1, 2) the efficacy
of a connection is determined by the
product of the probability of release, the number
of release sites, and the size of the
postsynaptic response to a quantum of transmitter.
Quantal parameters determine how
reliably information is transmitted (3, 4),
with higher rates of information transfer possible
at connections with larger numbers of
release sites and higher (sustained) release
probabilities (3). However, basic synaptic
properties, such as number of functional and
anatomical release sites per connection and
the release probability per site, remain uncertain
at most cortical synapses. Moreover, the
number of vesicles released at individual synaptic
contacts is controversial. Correlation of
functional release sites and anatomically
1
Department of Physiology, University College London,
Gower Street, London WC1E 6BT, U.K. 2 Forschungszentrum
Jülich GmbH, Institut für Medizin,
D-52425 Jülich, Germany. 3 Max-Planck Institut für
Medizinische Forschung, Abteilung Zellphysiologie,
Jahnstrasse 29, D-69120 Heidelberg, Germany.
*These authors contributed equally to the work.
†To whom correspondence should be addressed. E-
mail: a.silver@ucl.ac.uk
Fig. 1. EPSP recordings from an L4 to L2/3 cell pair in rat barrel cortex under different release
probability conditions at 36°C. (A) Current protocols for AP stimulation in the presynaptic
spiny stellate cell (top trace) and for testing input resistance stability in postsynaptic pyramidal
cell (second trace). Third trace shows the AP in the spiny stellate cell. The bottom trace shows
voltage responses to current injection followed by individual EPSPs (gray lines) and the mean
EPSP (black line) in the postsynaptic pyramidal cell. (B) Mean EPSPs recorded in different
extracellular Ca 2 and Mg 2 concentrations at a potential of –73 mV. (C) Input resistance
during different extracellular Ca 2 and Mg 2 concentrations.
Fig. 2. Quantal parameters estimated
from evoked EPSPs under low probability
conditions. (A) The top trace
shows the mean action potential
evoked by a brief depolarizing current
pulse at –72 mV; the bottom trace
shows 14 individual EPSPs (gray
lines), the mean of the 44 EPSP successes
(black solid line), and the mean
of the 279 failures (dashed line) recorded
at –72 mV in 1 mM [Ca 2 ]
and 5 mM [Mg 2 ]. (B) The amplitude
histogram of EPSPs (bins) and scaled
baseline noise (dashed line). The coefficient
of variation of the quantal
EPSP amplitude (CV Q
)was calculated from the background-subtracted
variance. (C) Relation between the variance of the EPSP peak
amplitude (corrected for background variance) and mean peak EPSP
amplitude for an individual L4 to L2/3 cell pair. Each data point shows
a different probability condition. Error bars indicate the theoretical
standard error in the estimate of the variance. Solid line shows the fit
to a multinomial model giving Q 0.09 mV, N F
5.25 and
19800.
www.sciencemag.org SCIENCE VOL 302 12 DECEMBER 2003 1981

R EPORTS
identified synaptic contacts in inhibitory neurons
of goldfish lead to the hypothesis that a
maximum of one vesicle is released per active
zone (“the single vesicle hypothesis”)
(5). This hypothesis was subsequently supported
at connections between hippocampal
pyramidal cells and interneurons (6). However,
recent studies indicate that the concentration
of glutamate in the synaptic cleft changes
with release probability, which implies that
multiple vesicles are released at each synaptic
contact per action potential (AP) (7, 8).
We investigated the number of functional
and anatomical release sites, the release probability,
and the mode of release at synaptic
connections between pairs of layer 4 (L4) and
layer 2/3 (L2/3) excitatory neurons in rat
barrel cortex. Paired whole-cell recordings
were made from acute slices from 17- to
23-day-old rats at near-physiological temperature
(9, 10). In the presence of 2 mM [Ca 2 ]
and N-methyl-D-aspartate (NMDA) receptor
antagonists, single APs in L4 spiny neurons
evoked unitary excitatory postsynaptic potentials
(EPSPs) in L2/3 pyramidal cells with a
mean peak amplitude of 0.54 0.06 mV at
–70 mV (n 32 cells; Fig. 1A). When
stimulated at low frequency (0.1 to 0.033
Hz), EPSP amplitudes remained stable for
long periods (2 to 3 hours). The amplitude
of unitary EPSPs could be increased (in eight
of nine cells) or reduced by altering the
[Ca 2 ] and [Mg 2 ], which suggests that
transmitter release probability was not maximal
under our control conditions (Fig. 1B).
Such alterations in the [Ca 2 ]/[Mg 2 ] ratio
had little effect on the input resistance, and
thus on the voltage response to a given synaptic
conductance (Fig. 1C).
We first determined the mean postsynaptic
response to a single quantum of
transmitter (Q) by evoking release under conditions
of low release probability. In the presence
of 1 mM extracellular Ca 2 , the proportion
of presynaptic APs that failed to release
transmitter was high (0.82 0.02; 0.8 to 1.0
mM Ca 2 ; n 5), and thus, the vast majority
of the EPSPs were single quantal events (Fig.
2A) (10). Quantal EPSPs were identified on
the basis of their characteristic shape, after
each trace was digitally filtered. We checked
that small-amplitude EPSPs were not missed
by averaging the failures (Fig. 2A). On average,
the mean amplitude of quantal EPSPs
was 0.15 0.02 mV at –70 mV (n 5
cells). The mean number of quanta contributing
to the evoked EPSP (quantal content)
under control conditions (2 mM Ca 2 ) was
thus 3.6.
The number of quantal events generated
per synaptic contact can be determined by
comparing the number of sites that can generate
a quantal event (functional release sites,
N F
) and the number of anatomically identified
synaptic contacts (N A
). Because the
quantal content does not provide information
about N F
or about the average release probability
across these sites (P R
), we estimated
these quantal parameters by applying multiple-probability
fluctuation analysis (MPFA)
(11, 12), also known as variance-mean analysis
(13). We first estimated the total nonuniformity
in Q from amplitudes of quantal
EPSPs under low P R
conditions, using the
same peak and baseline measurement windows
used for MPFA (Fig. 2B) (10). This
gave a coefficient of variation (standard deviation/mean;
CV Q
0.43 0.06; n 5
cells) similar to that reported for other excitatory
synapses (11, 14). The variability in
quantal size at an individual release site (intrasite
variability) has been measured directly
at single site cerebellar synapses at physiological
temperature (CV QI
0.21) (15). If we
assume the same value for intrasite quantal
variability at L4 to L2/3 synapses (see verification
below), the variation arising from
differences in the mean quantal amplitude
across sites (intersite variability) can be calculated
(CV QII
0.37) from the total quantal
variability (10). These estimates of quantal
Fig. 3. Relation between the number of anatomically
identified synaptic contacts and functional
release sites. (A) Camera lucida reconstruction of
a cell pair, L4 stellate cell with L2/3 pyramidal
cell. The soma and dendrites of the presynaptic L4
cell are given in red, the axonal arbor in blue. The
somatodendritic configuration of the postsynaptic
L2/3 pyramidal cell is given in black, the axonal
arbor in green. Lower inset shows the distribution
of putative, light microscopically identified synaptic contacts. (B) Variance-mean plot and
multinomial fit for the cell pair shown in (A), where Q and N F are the estimated quantal size
and number of functional release sites, respectively. (C) Electron micrographs (EMs) of the
synaptic contacts identified by light microscopy in (A), where the letters next to the blue dots
in (A) indicate the position of each EM image shown in (C). Labeling in (C): b, presynaptic
bouton with clearly visible synaptic vesicles; d, postsynaptic dendrites.
1982
12 DECEMBER 2003 VOL 302 SCIENCE www.sciencemag.org

R EPORTS
variability were used to constrain a nonuniform
quantal model, which allowed more
accurate estimates of Q, N F
, and P R
.
Time-stable epochs of EPSP amplitudes
were identified for each release probability
condition with a Spearman rank-order analysis
(15). The mean and variance of the peak
amplitude were calculated and corrected for
deviations in driving force from –70 mV,
with an assumed reversal potential for AMPA
receptors of 0 mV (9). EPSP variance-mean
plots were constructed after subtracting background
variance and were fit with a multinomial
model with fixed values of quantal variance
(10); goodness of fit was assessed with
the chi-square test. A variance-mean plot and
multinomial fit are shown in Fig. 2C that
gave a Q of 0.09 mV and an estimate of 5.3
functional release sites for this cell pair. The
mean Q varied little across cell pairs (0.14
0.02, n 9) and was not significantly different
from that measured directly from quantal
EPSPs under low P R
conditions (P 0.55,
unpaired t test, unequal variance). The N F
across cells was 5.1 0.9 (n 9) and varied
between 2.3 and 6.5 except for one outlying
cell in which N F
11.1 (the mean value
excluding this cell was 4.4 0.5, n 8).
Simulations suggest that part of the variance
in N F
for the eight clustered cells arose from
uncertainty in estimating the variance from a
limited number of EPSPs (10). Under control
conditions, the P R
at each site was high
(0.79 0.04 in 2 mM Ca 2 , n 9), consistent
with a previous study (16). An upper
limit for the quantal variability at an individual
site was calculated by dividing the variance
remaining when P R
was maximal
(1.02 0.05, n 9) by N F
. This gave a value
of CV QI
close to that assumed from measurements
at single-site synapses (0.26 0.02
versus 0.21) (15). Estimation of the coefficient
of variation of release probability across
sites gave a value (0.28 0.07 in 2 mM
Ca 2 , n 9), similar to that obtained in
hippocampal synapses in culture (17).
Anatomical identification of axondendrite
intersections with high-resolution
light microscopy of biocytin-filled cell pairs
(10) indicated between four and six putative
synaptic contacts per connection located on
the proximal dendritic segments of the
postsynaptic neuron (blue dots, Fig. 3A, inset).
Across cells, an average of 4.6 0.2
(n 14) putative synaptic connections were
found, a value not significantly different from
Fig. 4. Fractional block of the EPSP by low-affinity glutamate antagonists is independent of release
probability. (A) The upper trace shows the presynaptic APs in a L4 spiny neuron; middle traces show
mean EPSPs recorded in 2 mM Ca 2 ,1mMMg 2 under control conditions and in 1 mM -DGG (red
trace). The bottom trace shows mean EPSPs recorded in 1.25 mM Ca 2 under control conditions
(blue) and in 1 mM -DGG (green). (B) Plot of EPSP amplitudes as a function of time through the
experiment. (C) EPSPs in (A) normalized to the peak of the first control EPSPs. (D) Relation
between percentage block of the EPSP and extracellular [Ca 2 ] for 0.5 mM kynurenic acid and
1mM-DGG.
the N F
(P 0.58 unpaired t test, unequal
variance). In four cell pairs, we were able to
recover the anatomy after performing quantal
analysis. A camera lucida reconstruction of a
cell pair is shown in Fig. 3, for which we
estimated an N F
of 5.5 and an N A
of 5 (Fig. 3,
A and B). For cell pairs in which both quantal
analysis and anatomical reconstruction were
performed, N F
and N A
were not significantly
different (P 0.21, paired t test, n 4; table
S1). We further examined whether these
close appositions were indeed synaptic contacts
by making serial ultrathin sections
through the entire dendritic domain of the
postsynaptic neuron and examining them at
the electron microscopic (EM) level. Synaptic
contacts were identified on the basis of a
clear presynaptic bouton-like structure containing
closely packed synaptic vesicles together
with closely apposing (20 nm) preand
postsynaptic membranes (Fig. 3C). The
presence of electron-dense biocytin reaction
product both pre- and postsynaptically and
the long duration of the recordings precluded
the use of other anatomical identifiers such as
the pre- and postsynaptic densities. Full EM
analysis of two quantal analysis pairs of neurons
and one additional cell confirmed identification
of 13 out of 14 putative synaptic
contacts, which gave a detection rate close to
that observed at other cortical synapses (18,
19). Synaptic contacts were located directly
on dendritic shafts (Fig. 3C) and occasionally
on spines. Applying the observed optical detection
rate for synaptic contacts (93%)
across the four pairs of neurons, we estimated
N F
/N A
to be 0.90 0.16, which suggests a
one-to-one relation between these synaptic
properties (table S1).
The all-or-none behavior of excitatory synapses
could arise either presynaptically, from the
release of a single quantum per AP, or postsynaptically,
from saturation of postsynaptic receptors
(20, 21) following multivesicular release
(22, 23). We therefore examined whether the
cleft glutamate concentration changes with P R
by assaying the level of block of the EPSPs by
low-affinity glutamate receptor antagonists (7,
12, 24). The effect of 1 mM -d-glutamylglycine
(-DGG) on the mean EPSPs recorded from a
cell pair in normal Ca 2 (2 mM) and in the
presence of a low Ca 2 (1.25 mM) concentration
is shown in Fig. 4, A and B. Normalization to the
first EPSP amplitude illustrates that the fractional
block of the EPSP by -DGG was similar
under these two conditions (Fig. 4C). On average,
the fractional block of the EPSP by -DGG
was not significantly different in 2 mM Ca 2
and 1.25 mM Ca 2 (66 6% and 60 5%,
respectively; P 0.43; n 5). Moreover, the
paired-pulse ratio in 2 mM Ca 2 was unchanged
in the presence of -DGG (0.59 to 0.52, P
0.36; n 10), which indicates a similar fractional
block on each pulse. We tested these results
further by carrying out a series of experiments
www.sciencemag.org SCIENCE VOL 302 12 DECEMBER 2003 1983

R EPORTS
using kynurenic acid, a different rapidly equilibrating
antagonist. As for -DGG, the fractional
block of the EPSP by kynurenic acid was not
significantly different in 2 mM Ca 2 and 1.25
mM Ca 2 (55 3% and 53 6%, respectively;
P 0.61; n 4; Fig. 4D). Our results contrast
with the substantial increase in the level of block
of excitatory postsynaptic currents (EPSCs) by
-DGG and kynurenic acid at the climbing fiber,
as the release probability was lowered (7) from a
similar initial value (11). The fact that rapidly
equilibrating competitive antagonists are equally
effective at blocking the EPSP when P R
was
reduced by a factor of 4 suggests that the concentration
of transmitter in the synaptic cleft,
following release, is independent of P R
at L4 to
L2/3 synapses. These results indicate that each
excitatory synaptic contact releases a single
quantum of glutamate (probably one vesicle but
could be a group of vesicles whose number is
independent of P R
) and operates independently,
in an all-or-none manner.
The high P R
at L4 to L2/3 synapses is well
suited to promote reliable transmission of the
one (76%) or two spikes that whisker stimulation
usually evokes in L4 cells in vivo (25),
because short-term depression becomes significant
during longer bursts of L4 activity [Fig.
4A and (9)]. The depression of high P R
synapses,
together with the precise synaptic latency
between L4 and L2/3 (1 to 3 ms) and rapid
EPSP kinetics (9), may also be important for
maintaining and strengthening L4 to L2/3 synapses,
because this connection exhibits a spiketiming–based
plasticity rule with a narrow coincidence
window (26). Curiously, the small
numbers of release sites, all-or-none uniquantal
signaling and small quantal size are likely to
endow an individual L4 to L2/3 connection
with a relatively low capacity for transmitting
information. These properties contrast with
synapses in the retina, where large numbers of
release sites (200) and graded transmission
allow high information transmission rates (27).
Moreover, transmission is finely tuned because
only 10% of the 300 to 400 spiny stellate cells
that innervate an individual L2/3 pyramidal cell
fire synchronously, within 15 ms of single
whisker stimulation (25), close to the minimum
number required for a pyramidal cell to reach
voltage threshold (9). Our results suggest that
the functional and anatomical properties of L4
to L2/3 synaptic connections are well adapted
for efficient, timing-based distributed signaling.
References and Notes
1. B. Katz, The Release of Neural Transmitter Substances
(Liverpool Univ. Press, Liverpool, 1969).
2. D. Vere-Jones, Aust. J. Stat. 8, 53 (1966).
3. A. Zador, J. Neurophysiol. 79, 1219 (1998).
4. M. London, A. Schreibman, M. Häusser, M. E. Larkum,
I. Segev, Nature Neurosci. 5, 332 (2002).
5. H. Korn, A. Triller, A. Mallet, D. S. Faber, Science 213,
898 (1981).
6. A. I. Gulyas et al., Nature 366, 683 (1993).
7. J. I. Wadiche, C. E. Jahr, Neuron 32, 301 (2001).
8. T. G. Oertner, B. L. Sabatini, E. A. Nimchinsky, K.
Svoboda, Nature Neurosci 5, 657 (2002).
9. D. Feldmeyer, J. Lübke, R. A. Silver, B. Sakmann,
J. Physiol. (London) 538, 803 (2002).
10. Methods, table S1, and references are available as
supporting online material on Science Online.
11. R. A. Silver, A. Momiyama, S. G. Cull-Candy, J. Physiol.
(London) 510, 881 (1998).
12. J. D. Clements, R. A. Silver, Trends Neurosci 23, 105 (2000).
13. C. A. Reid, J. D. Clements, J. Physiol. (London) 518,
121 (1999).
14. J. M. Bekkers, J. D. Clements, J. Physiol. (London) 516,
227 (1999).
15. R. A. Silver, S. G. Cull-Candy, T. Takahashi, J. Physiol.
(London) 494, 231 (1996).
16. M. A. Castro-Alamancos, B. W. Connors, Proc. Natl.
Acad. Sci. U.S.A. 94, 4161 (1997).
17. V. N. Murthy, T. J. Sejnowski, C. F. Stevens, Neuron
18, 599 (1997).
18. H. Markram, J. Lübke, M. Frotscher, A. Roth, B. Sakmann,
J. Physiol. (London) 500, 409 (1997).
19. J. Lübke, A. Roth, D. Feldmeyer, B. Sakmann, Cereb.
Cortex 13, 1051 (2003).
20. J. J. Jack, S. J. Redman, K. Wong, J. Physiol. (London)
321, 65 (1981).
21. P. Jonas, G. Major, B. Sakmann, J. Physiol. (London)
472, 615 (1993).
22. R. A. Silver, in Central Synapses: Quantal Mechanisms
and Plasticity, D. S. Faber, H. Korn, S. J. Redman, S. M.
Thompson, and J. S. Altman, Eds., 4th Human Frontier
Science Workshop, Strasbourg, France, 22 to 24 April
1997 (Human Frontier Science Program, Strasbourg,
France, 1998), pp. 130–139.
23. K. A. Foster, A. C. Kreitzer, W. G. Regehr, Neuron 35,
1115 (2002).
24. J. S. Diamond, C. E. Jahr, J. Neurosci. 17, 4672 (1997).
25. M. Brecht, B. Sakmann, J. Physiol. (London) 543, 49
(2002).
26. D. E. Feldman, Neuron 27, 45 (2000).
27. S. B. Laughlin, R. R. de Ruyter van Steveninck, J. C.
Anderson, Nature Neurosci. 1, 36 (1998).
28. Supported by the Max-Planck Society and the Wellcome
Trust. R.A.S. is in receipt of a Wellcome Trust
Senior Fellowship. We thank A. Roth and C. Saviane
for help with simulations; D. Attwell, D. DiGregorio,
M. Farrant, S. Mitchell, T. Nielsen, and I. Vida for
comments on the manuscript; and N. Nestel and B.
Joch for technical assistance.
Supporting Online Material
www.sciencemag.org/cgi/content/full/302/5652/1981/DC1
Materials and Methods
Table S1
References
23 May 2003; accepted 15 October 2003
Anterior-Posterior Guidance of
Commissural Axons by
Wnt-Frizzled Signaling
Anna I. Lyuksyutova, 2 Chin-Chun Lu, 1 Nancy Milanesio, 1
Leslie A. King, 2 Nini Guo, 4 Yanshu Wang, 4 Jeremy Nathans, 4
Marc Tessier-Lavigne, 5 * Yimin Zou 1,2,3 †
Commissural neurons in the mammalian dorsal spinal cord send axons ventrally
toward the floor plate, where they cross the midline and turn anteriorly toward the
brain; a gradient of chemoattractant(s) inside the spinal cord controls this turning.
In rodents, several Wnt proteins stimulate the extension of commissural axons after
midline crossing (postcrossing). We found that Wnt4 messenger RNA is expressed
in a decreasing anterior-to-posterior gradient in the floor plate, and that a directed
source of Wnt4protein attracted postcrossing commissural axons. Commissural
axons in mice lacking the Wnt receptor Frizzled3 displayed anterior-posterior
guidance defects after midline crossing. Thus, Wnt-Frizzled signaling guides commissural
axons along the anterior-posterior axis of the spinal cord.
Axonal connections are patterned along the
anterior-posterior (A-P) and dorsal-ventral
(D-V ) neuraxes. Guidance molecules that
play essential roles in the D-V guidance of
axons have been identified, whereas the nature
of the A-P guidance cues has remained
an enigma (1, 2). The dorsal spinal cord
commissural neurons form several ascending
somatosensory pathways. During embryonic
development, they project axons to the ventral
midline (Fig. 1A). At the floor plate,
commissural axons cross the midline, enter
the contralateral side of the spinal cord, and
make a sharp anterior turn toward the brain
(Fig. 1B) (3). The initial ventral growth of the
commissural axons is directed by a collaboration
of two chemoattractants, netrin-1 (4–
6) and Sonic hedgehog (Shh) (7), and chemorepellents
of the bone morphogenetic protein
(BMP) family (8). As the axons cross the
midline, they lose responsiveness to these
chemoattractants (9) but gain responsiveness
to several chemorepellents, including Slit and
semaphorin proteins (10), which guide axons
from the D-V axis into the A-P axis (10).
To determine why axons turn in an anterior
direction, we studied the turning of commissural
axons after midline crossing in “open-book” spi-
1
Department of Neurobiology, Pharmacology and Physiology,
2 Committee on Developmental Biology, 3 Committee
on Neurobiology, University of Chicago, Chicago, IL 60637,
USA. 4 Departments of Molecular Biology and Ophthalmology,
Howard Hughes Medical Institute, Johns Hopkins
Medical School, Baltimore, MD 21205, USA. 5 Department
of Biological Sciences, Howard Hughes Medical Institute,
Stanford University, Stanford, CA 94305, USA.
*Present address: Genentech Inc., South San Francisco,
CA 94080, USA.
†To whom correspondence should be addressed. E-
mail: yzou@bsd.uchicago.edu
1984
12 DECEMBER 2003 VOL 302 SCIENCE www.sciencemag.org