Department of Neurosciences, University of New Mexico, School of Medicine, Albuquerque, New Mexico 87131
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Cormier, R. J., A. C. Greenwood, and J. A. Connor. Bidirectional Synaptic Plasticity Correlated With the Magnitude of Dendritic Calcium Transients Above a Threshold. J. Neurophysiol. 85: 399-406, 2001. The magnitude of postsynaptic Ca2+ transients is thought to affect activity-dependent synaptic plasticity associated with learning and memory. Large Ca2+ transients have been implicated in the induction of long-term potentiation (LTP), while smaller Ca2+ transients have been associated with long-term depression (LTD). However, a direct relationship has not been demonstrated between Ca2+ measurements and direction of synaptic plasticity in the same cells, using one induction protocol. Here, we used glutamate iontophoresis to induce Ca2+ transients in hippocampal CA1 neurons injected with the Ca2+-indicator fura-2. Test stimulation of one or two synaptic pathways before and after iontophoresis showed that the direction of synaptic plasticity correlated with glutamate-induced Ca2+ levels above a threshold, below which no plasticity occurred (~180 nM). Relatively low Ca2+ levels (180-500 nM) typically led to LTD of synaptic transmission and higher levels (>500 nM) often led to LTP. Failure to show plasticity correlated with Ca2+ levels in two distinct ranges: <180 nM and ~450-600 nM, while only LTD occurred between these ranges. Our data support a class of models in which failure of Ca2+ transients to affect transmission may arise either from insufficient Ca2+ to affect Ca2+-sensitive proteins regulating synaptic strength through opposing activities or from higher Ca2+ levels that reset activities of such proteins without affecting the net balance of activities. Our estimates of the threshold Ca2+ level for LTD (~180 nM) and for the transition from LTD to LTP (~540 nM) may assist in constraining the molecular details of such models.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Learning and memory are widely
associated with enduring forms of synaptic plasticity such as long-term
potentiation (LTP) (Bliss and Gardner-Medwin 1973) and
long-term depression (LTD) (Fujii et al. 1991
), which
respectively enhance and diminish synaptic strength. Despite their
opposing effects, LTP and LTD share some common biochemical mechanisms,
including intracellular Ca2+ transients. Here, we
focus on the relation of transient Ca2+ levels to
LTP and LTD in postsynaptic CA1 pyramidal neurons.
A wealth of data demonstrates the requirement for
Ca2+ transients in synaptic plasticity. Direct
measurements of Ca2+ using ratio-fluorescence
microscopy showed micromolar levels in postsynaptic neurons in response
to tetanic stimulation (Müller and Connor 1991;
Petrozzino et al. 1995
; Yuste and Denk
1995
). In addition, raising postsynaptic
Ca2+ directly by photolysis of
caged-Ca2+ compounds loaded into postsynaptic
neurons induced LTP as well as LTD (Malenka et al. 1988
;
Neveu and Zucker 1996
). On the other hand, inhibiting
Ca2+ transients by loading postsynaptic neurons
with calcium chelators, such as EGTA (Lynch et al. 1983
)
or caged-chelators (Malenka et al. 1992
), inhibited
induction of LTP and LTD. Furthermore induction protocols associated
with higher Ca2+ levels induced LTP, while
protocols associated with more moderate Ca2+
levels induced LTD (Hansel et al. 1996
, 1997
). Despite
these compelling studies relating Ca2+ to
synaptic plasticity, a quantitative relationship between measured Ca2+ levels and synaptic plasticity in the same
neurons has not been directly demonstrated, prompting the present study.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We prepared coronal hemispheric brain slices (350- to 400-µm
thick) from adult male Sprague-Dawley rats (Harlan; 5-10 wk old) by
standard humane methods involving quick decapitation of anesthetized animals (Connor and Cormier 2000). The slices were then
maintained at 28°C in artificial cerebrospinal fluid (ACSF;
containing, in mM: 126 NaCl, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 1 MgCl2, 2 CaCl2, and 10 dextrose; gassed with 95%
O2-5% CO2) until use. All
chemicals were from Sigma.
For experiments, individual slices were placed in a submersion chamber
and perfused at 1-2 ml/min with ACSF at 31 ± 0.5°C. Intracellular recordings were made from CA1 neurons ~75 µm below the slice surface, using glass micropipettes with tips filled with 12 mM fura-2 and shanks filled with a solution of 3 M KCl and 1 M
K-acetate. The electrode resistance was initially ~200 M with
fura-2 present but dropped to ~120 M
in ~20 min before data
collection. An Axoclamp 2A amplifier in bridge mode was used to record
membrane potential. To evoke excitatory postsynaptic potentials
(EPSPs), one or two monopolar stimulating electrodes were placed in
stratum radiatum with the tip (20 µm diameter) ~100 µm below the
slice surface. Stimulating electrodes were placed in the proximal third
of s. radiatum to activate afferents on proximal dendrites or in the
distal third to activate afferents on distal dendrites. Pairs of EPSPs
separated by 50 ms were evoked by constant-current pulses (50-100
µA) every 15 s.
After stopping the afferent stimulation used to evoke baseline EPSPs,
an iontophoresis pipette (1 M glutamate, pH 7.0, ~10 M) was
positioned 20-50 µm from the primary apical dendrite of the
fura-filled neuron. Glutamate was ejected by five iontophoretic pulses
(duration: 10 s, amplitude: 4 µA, interval: 60 s), and the
pipette was then withdrawn from the slice
5 min before afferent stimulation was resumed. In control experiments, 1 M NaCl was used in
place of glutamate.
Ca2+ levels were measured by successively
illuminating fura-filled neurons with 350 and 380 nm wavelength light
for 200 or 400 ms and imaging the resulting fluorescence with a
cooled-CCD camera on an upright microscope (Zeiss Axioskop) with a
water-immersion, ×40 objective lens. Ca2+ levels
were calculated by standard ratiometric methods after fluorescent-background correction, assuming a
Ca2+-affinity KD
of 225 nM for fura-2 (Grynkiewicz et al. 1985). Fura levels were estimated as follows. Cells were injected with 0.7 nA for
20 min. Conservatively ignoring dilution by the backfilling solution
and assuming a cell volume of 3 nl, the final intracellular [fura]
would be ~60 µM, by the following equation: [Fura-2] = (n · i · t)/(z · F · v). Here n is the transport number (0.1), i is the iontophoresis current, t is the time of
iontophoresis, z is the net charge, F is
Faraday's constant, and v is the cell volume. A published
comparison of similar sharp-electrode methods to whole cell
fura-injection suggests that our methods achieved intracellular fura
levels of 20-30 µM (Petrozzino et al. 1995
), unlikely
to slow the already slow Ca2+ kinetics associated
with glutamate iontophoresis.
Statistical comparisons were made using Tukey's two-tailed
t-test, corrected for multiple comparisons by Bonferroni's
method (Snedecor and Cochran 1980). All error bars show
the SE. Error propagation to determine uncertainty in plasticity factor
followed standard methods (Bevington and Robinson 1992
).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We measured intracellular Ca2+ levels during
iontophoresis of glutamate onto the apical dendrites of CA1 pyramidal
cells in coronal brain slices from rats (n = 67). The
bulk of these experiments (n = 53) served to
characterize the ensuing Ca2+ transients, guiding
a study of the relation between these transients and effects on
transmission in 20 synaptic inputs to 14 cells with one
(n = 8) or two (n = 6) stimulated
synaptic pathways. Cells that were included in this study had stable
resting potentials more negative than 65 mV and stable resting
Ca2+ levels averaging 71 ± 11 (SE) nM.
Glutamate iontophoresis (10 s, 4 µA) typically depolarized the
membrane to values between 50 and
10 mV (data described in this
paragraph not shown) (see Cormier and Kelly 1996
;
Cormier et al. 1993
). After a glutamate pulse, cells
repolarized and sometimes also hyperpolarized, perhaps from activation
of Ca2+-gated potassium channels (Hotson
and Prince 1980
). Action potentials were often observed during
the first two glutamate pulses in a train (at 1 pulse/min) but rarely
during later pulses, though these later pulses were associated with
more rapid depolarization than were the first two pulses. Control
experiments with equimolar substitution of sodium chloride for
glutamate demonstrated that iontophoresis current (up to 10 times
that used here) had no detectable effect on membrane potential,
intracellular Ca2+, or synaptic transmission
(n = 3) (see Cormier and Kelly 1996
; Cormier et al. 1993
).
Focusing first on the temporal information in our imaging data, we
measured Ca2+ levels in the region of dendrite to
which glutamate was applied in most experiments, ~70 µm from the
soma (marked in Fig. 1D, ).
In this region, Ca2+ levels typically rose during
6-10 s of the standard 10-s pulse, reached a peak level between 100 and 600 nM, and returned to basal levels by 5-120 s after the pulse
(Fig. 1A: data from typical cell). Increasing the
iontophoretic current produced higher Ca2+ levels
(Fig. 1B: data from same cell) and very prolonged
Ca2+ transients in some cases, as reported
elsewhere (Connor and Cormier 2000
). In our study
relating Ca2+ transients to plasticity, we set
the iontophoretic current at an intermediate level (4 µA) that
produced Ca2+ levels <1 µM to reduce the risk
of excitotoxic effects.
|
A series of five successive pulses of glutamate at one-minute intervals
led to multiphasic Ca2+ transients (Fig.
1C). In some experiments, the basal
Ca2+ level was higher for successive pulses
(n = 20 of 67; typical data shown in Fig.
1C1). Also the peak Ca2+ level often
increased with successive pulses (n = 38 of 67; Fig. 1C1). Saturation of glutamate uptake (Asztely et al.
1997; Mennerick and Zorumski 1995
) or of
Ca2+ sequestration or extrusion, among other
mechanisms, may have contributed to these phenomena. On the other hand,
in many experiments, the peak Ca2+ level was very
similar from pulse to pulse (n = 29 of 67; Fig. 1C2 shows typical data).
The spatial distribution of glutamate-induced
Ca2+ transients within cells was inhomogeneous,
suggesting that synapses at widely separated locations might show
synaptic plasticity of differing magnitude and direction. In our
initial experiments, we placed the iontophoresis pipet at ~70 µm
along the proximal apical dendrites, producing
Ca2+ transients centered near the point of
application. Typically, the initial peak was closer to the soma than
the pipette was (Fig. 1D, 4 s), possibly the result of
activating dendritic voltage-gated Ca2+ channels
(Miyakawa et al. 1992; Regehr and Tank
1992
). However, by 8 s, the Ca2+
levels most proximal to the soma had declined, leaving an increasingly well-defined peak centered near the glutamate source. In some later
experiments (n = 6), the iontophoresis pipette was
placed in distal stratum radiatum, ~250 µm from the base of the
apical shaft. Distal glutamate led to Ca2+
transients resembling those that followed proximal glutamate application, except that the location of maximal change tended to be
shifted distally. An example is shown in Fig. 1E.
Having examined the effects of glutamate application on dendritic
Ca2+, and as a prelude to correlating
glutamate-induced Ca2+ levels with synaptic
plasticity, we next asked what does glutamate iontophoresis do to
synaptic transmission near and away from the application site in
fura-filled cells? To answer this question, we used the fura-filled
micropipette to record evoked intracellular EPSPs before and 15 min
after glutamate application. In eight experiments, a single stimulating
electrode was placed close to the cell-body layer to activate synapses
on the proximal dendrites, the site of glutamate application (Fig.
2A1). These experiments were
classified into three groups based on their plasticity factor, defined
as mean slope of 20 consecutive EPSPs arbitrarily selected from 40 to
60 min after the end of iontophoresis, normalized with respect to a
similarly calculated mean baseline slope: LTP 115% (A); LTD
85% (B); 85%
no
plasticity
115% (C). These specific group
boundaries were chosen because the 95% confidence interval for the
baseline mean was 100 ± ~15% in every experiment. However, varying the group boundaries to 100 ± 10 or 100 ± 30% had
no effect on group membership. Furthermore due to the stability of the
recorded EPSPs, the 95% confidence interval for a plasticity factor
was never more than 7.1% of the plasticity factor for any pathway in
this study, and the mean normalized confidence interval for all
plasticity factors was only 2.9%. Thus no 95% confidence interval for
a pathway's plasticity factor ever crossed a group boundary as
initially defined. Applying this classification scheme to EPSPs that
were evoked in eight single-pathway experiments, we observed LTD of
synaptic transmission in five cells (to 56 ± 10% of baseline) and LTP in three cells (to 166 ± 11%), as summarized in Fig.
2B (single pathway). These results showed that glutamate
applied to the proximal dendrites affected synaptic transmission in
that dendritic region, prompting a series of two-pathway experiments with distal glutamate application.
|
In these two-pathway experiments (n = 6), the first
stimulating electrode in proximal s. radiatum was complemented with a second one in distal s. radiatum, as diagrammed in Fig. 2A2.
Glutamate was applied to the dendrites in the distal third of s.
radiatum (as in Fig. 1E) to increase the likelihood that
EPSPs from the distal synapses would undergo LTP (as in the
representative experimental time course in Fig. 2C). We
expected that we might also observe plasticity in the proximal pathway
due to the activation of voltage-gated Ca2+
channels on the primary apical dendrites (Miyakawa et al.
1992; Regehr and Tank 1992
). In fact, when we
applied our plasticity-factor classification scheme to the proximal
pathway, we observed no plasticity (98 ± 3%) in four cells, LTD
(72%) in one cell, and LTP (302%) in one
cell1 at 40-60 min
after the glutamate treatment. Meanwhile, the distal pathway showed no
plasticity (106 ± 1%) in two cells, LTD (32%) in one cell, and
LTP (231 ± 51%) in three cells. These data are plotted on the
right in Fig. 2B, with proximal and distal plasticity factors from individual neurons connected by lines. Thus both sides of
Fig. 2B show that the range of observed plasticity factors was well suited to answer the question: do the
Ca2+ levels reached during glutamate
iontophoresis and subsequent effects on synaptic transmission correlate
as they would in a simple model in which Ca2+
caused these effects?
To answer this question, we first pooled the electrophysiological data from the 8 single-pathway experiments and the six two-pathway experiments to obtain a total of 20 experimental time courses of EPSP slope, divided as before into three groups according to plasticity factor. The average time courses for the LTP, LTD, and no-plasticity groups of pooled data are plotted in Fig. 2, D-F, respectively (n = 7, 7, and 6). Next, we sorted the glutamate-induced Ca2+ transients into the same three groups. In these data, the distal or proximal location of stimulus-activated synapses was inferred from the position of the stimulating electrode. To quantify the Ca2+ levels that may have affected synaptic transmission, we analyzed the iontophoresis imaging data with measurement boxes placed on the dendrite near these putative synaptic locations. Figure 3A shows such box placement on a typical cell in a two-pathway experiment, while Fig. 3B shows the average multiple-pulse Ca2+-level time course obtained from all such boxes for synaptic pathways that exhibited LTP. For comparison between pathways or groups of pathways, multiple-pulse time courses were collapsed to a single-pulse time course by averaging across the five glutamate pulses, as illustrated for the LTP group in Fig. 3C. Then, the three points at 6, 8 and 10 s during the single-pulse time courses were averaged to obtain a single measure of peak-Ca2+ level for each pathway or for each group in the case of the averaged data. Fura saturation had minimal effect on the average peak-Ca2+ level, as at most one of five peaks in the average approached 1 µM (e.g., 850 nM in Fig. 1C1).
|
Looking first at the results for each pathway separately, Fig.
3D shows a scatter plot of percentage plasticity factor
versus the calculated peak Ca2+ level. Different
symbols indicate data collected from single (proximal) pathways and
distal and proximal pathways in two-pathway experiments (see legend).
Despite a statistically adequate fit (r = 0.60, P < 0.005), a linear model was rejected on
physiological grounds because it predicted the ongoing induction of
profound LTD at resting Ca2+ levels, which of
course was not observed. A linear model constrained to start from no
plasticity at Ca2+ baseline was rejected
(r = 0.12, P < 0.59, not plotted). A
fourth-order polynomial fit is also plotted in Fig. 3D to
help visualize the multiphasic dependence of plasticity on
Ca2+ that a satisfactory model would need to
explain. One satisfactory class of models would involve a balance of
effectors with opposing effects on synaptic strength, with the activity
of at least one effector being reset by
Ca2+ levels above a threshold
(Grzywacz and Burgi 1998; Lisman
1989
). In such a model, the no-plasticity group would be
predicted to consist of two subgroups, one that was exposed to
Ca2+ levels below that required to reset the
enzyme activities and a second that was exposed to
Ca2+ levels at which the activities of individual
enzyme molecules were reset without affecting the balance of activities.
Turning next to the summary data shown in Fig. 3E, pair-wise comparisons among all three plasticity groups showed that the peak Ca2+ levels in the LTP group were significantly larger than in the LTD group (P < 0.05, n = 7 and 7). Thus our data support the conventional view that low Ca2+ levels lead to LTD, while higher levels lead to LTP. In contrast, the mean peak Ca2+ level of the no-plasticity group barely differed from that of the LTD group (not significant, P > 0.9, n = 6 and 7). Meanwhile, the standard deviation (SD = 220) of the peak Ca2+ levels in the no-plasticity group was so large that its mean was not significantly different from that of the LTP group (P > 0.2, n = 6 and 7). However, as suggested by Fig. 3D, this extreme variability within the no-plasticity group could be accounted for by dividing it into two subgroups ("low" and "high" Ca2+) with markedly different average peak Ca2+ levels with small standard deviations [147 ± 20 and 541 ± 63 (SD) nM, n = 3 and 3].
Of course, the significance of this difference in
Ca2+ levels could not be tested because
membership in the subgroups depended on Ca2+,
violating the standard assumption of independence. However, the
apparent bimodality of the distribution of peak
Ca2+ levels within the no-plasticity group was
supported by the failure of a normal model to fit this distribution
(P 0.001,
2 = 35). Also,
despite the marked difference between the Ca2+
levels associated with the high-Ca2+ and
low-Ca2+ no-plasticity subgroups, it followed
from the no-plasticity group's definition that the mean plasticity
factors associated with these subgroups were statistically
indistinguishable (Fig. 3F, P = 0.27, n = 3 and 3). Finally, although the mean peak
Ca2+ level associated with LTP was not much
higher than that of the high-Ca2+ no-plasticity
subgroup (Fig. 3E, no statement of significance possible),
this result is not surprising because the iontophoretic current (4 nA)
was selected to achieve Ca2+ levels below 1 µM,
typically ~500 nM. Previous work showed that larger iontophoretic
currents led to LTP (Cormier et al. 1993
) and larger
Ca2+ transients (Connor and Cormier
2000
).
Returning to Fig. 3D, the fourth-order polynomial fit provided initial estimates for the Ca2+ levels associated with the low-Ca2+ transition between no plasticity and LTD (165 nM) and the 100% cross-over point between the LTD and LTP groups (485 nM). To address the arbitrariness of the fourth-order polynomial, we also estimated each of these transition points by a second method. For a second estimate of the low-Ca2+ transition, linear interpolation between the data points straddling the 85% plasticity line yielded an 85% intercept of 180 nM. For a second estimate of the 100% cross-over point, we took the mean peak Ca2+ level in the high-Ca2+ no-plasticity subgroup (541 nM). As they include no arbitrary assumptions, we accepted these second estimates in place of the estimates from the fourth-order polynomial, with which they are in reasonable agreement.
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
This paper reports that glutamate iontophoresis elevated
intracellular Ca2+ in the apical dendrites of CA1
pyramidal cells and induced synaptic plasticity in the same cells.
Although glutamate iontophoresis was previously shown to induce
synaptic plasticity in hippocampal cultures (Malgaroli and Tsien
1992) and slices (Cormier and Kelly 1996
;
Cormier et al. 1993
), the present study included
Ca2+ imaging, allowing us to relate
glutamate-induced Ca2+ levels to synaptic
plasticity quantitatively.
Glutamate iontophoresis, Ca2+, and cell viability
The iontophoretic protocol produced Ca2+
transients that returned to baseline within 120 s after a series
of five 10-s pulses at 1-min intervals. Earlier work suggested that
exposure to appropriate Ca2+ levels for 5 min was
sufficient to induce LTD (Mulkey and Malenka 1992),
while only 2.5 s of appropriate Ca2+ levels
was reported to be required for LTP induction (Malenka et al.
1992
). With these putative temporal requirements in mind, our
protocol was designed to reduce the importance of
Ca2+-transient kinetics as a factor affecting
synaptic plasticity and to result in LTD or LTP, depending on the peak
Ca2+ levels achieved. The between-cell
variability that we observed in peak Ca2+ levels
is consistent with the many sources of variability inherent in the
iontophoretic application of glutamate in the slice, including cell
depth, variations in local perfusate flow, and distance from the
dendrite. This variability is also consistent with an earlier study of
glutamate-induced Ca2+ transients in dentate
granule cells, not involving synaptic plasticity (Kudo et al.
1987
).
In contrast to our protocol, which required five sustaining 10-s pulses
of glutamate to achieve ~5 min of elevated
Ca2+, a neurotoxic exposure of acutely
dissociated hippocampal cells to glutamate or
N-methyl-D-aspartate (NMDA) can induce
Ca2+ transients that last for many minutes
after iontophoresis (Connor et al. 1988;
Wadman and Connor 1992
). Although we found that a similarly prolonged recovery could be achieved using glutamate iontophoresis in hippocampal slices, in our hands it required twice as
much iontophoresis current as was used in the present study
(Connor and Cormier 2000
). Thus our five-pulse protocol was designed to achieve relatively prolonged Ca2+
transients, while avoiding the prolonged single-pulse recovery that we
feared would be associated with neurotoxic effects. As the higher
Ca2+ levels were associated with LTP, whereas
sick cells might be expected to show LTD, we infer that we successfully
avoided complicating neurotoxic effects. This inference is further
supported by our observation of stable resting potential and input
resistance throughout the experiments.
Glutamate iontophoresis and synaptic plasticity
The clear correlation between Ca2+ levels
and plasticity that we observed suggests that
Ca2+ is a key mediator of plasticity induced by
glutamate iontophoresis. However, our experiments do not rule out
effects of glutamate on synaptic plasticity that are not entirely
mediated by postsynaptic Ca2+. For example, in
addition to their role in mobilizing intracellular Ca2+ (Jaffe and Brown 1994;
Linden et al. 1994
; Llano et al. 1991
; Murphy and Miller 1988
), metabotropic glutamate
receptors can affect synaptic transmission in a variety of other ways
(Conn and Pin 1997
). Also, glutamate iontophoresis may
release neuromodulators, such as nitric oxide or arachidonic acid
(Medina and Izquierdo 1995
). Addressing the possibility
that these substances or glutamate contributed to presynaptic
plasticity in our experiments, we note that we did not stimulate
afferent fibers during glutamate exposure and that previous work
established that presynaptic action potentials and transmitter release
are not required for the induction of plasticity by glutamate
iontophoresis (Cormier et al. 1993
).
In contrast to the absence of synaptic stimulation in our induction
protocol, an earlier study found that sustained dendritic Ca2+ transients elicited by action potentials
evoked with current pulses at 3 Hz for 5 min were insufficient to
elicit LTD without paired synaptic stimulation (Christie et al.
1996). This study also differed from the present study in that
F/F measurements of fura-2 fluorescence were
reported instead of ratiometrically determined
Ca2+ levels, preventing a direct comparison of
results. However, in harmony with the present study, these authors
interpreted their results to suggest that Ca2+
played a critical role in the efficacy of their LTD-induction protocol,
noting that LTD induction was blocked by nimodipine, Ni2+, or APV (independently) and suggesting that
the essential role of paired synaptic stimulation may have been to
enhance Ca2+ influx specifically into spines (see
also Yuste and Denk 1995
). These localized synaptic
Ca2+ transients apparently did not add
significantly to the dendritic
F/F observed
with action potentials alone. In contrast, our admittedly less
"physiological" induction protocol may have permitted a more direct
assessment of the Ca2+ levels affecting
transmission, as glutamate-induced Ca2+
transients are relatively homogeneous spatially, especially in fura-filled cells (see following section).
Ca2+ measurements with fura-2
Two concerns naturally arise regarding the use of fura-2. First,
its high affinity for Ca2+ (~225 nM) could lead
to saturation. However, saturation occurs at ~1 µM
Ca2+, above the levels that we obtained during
plasticity induction. Second, it has been reported that exogenous
Ca2+ buffers, like fura-2, can interfere with the
induction of LTP by diminishing the peak amplitude of brief
Ca2+ transients while incidentally prolonging the
recovery (Hansel et al. 1996, 1997
; Kimura et al.
1990
; Lynch et al. 1983
; Malenka et al.
1988
). However, the use of fura-2 in our experiments is unlikely to have interfered with the relation between our
Ca2+ measurements and plasticity for the
following reasons. Similar methods were reported to lead to
intracellular fura levels of 20-30 µM (Petrozzino et al.
1995
), while our conservatively high calculated estimate was 60 µM. Also, as the postsynaptic Ca2+ transients
were induced by iontophoresis of glutamate at some distance from the
putative synaptic location, the transients were expected to be
inherently slow, reducing concerns that fura-2 would further retard
their kinetics or diminish their magnitude. Granted, it is likely that
mobile fura-2 molecules tended to keep the Ca2+
level at the synapse and in the dendrite more homogeneous than if local
Ca2+ level depended only on local
Ca2+ channels and other physiological sources
(Carnevale and Rosenthal 1992
). However, this enhanced
homogeneity coupled with our slow Ca2+-transient
kinetics would have the advantage of reducing extra variance in the
relationship between measurements of plasticity and dendritic
Ca2+. In summary, the main goal of these
experiments was to study effects of
Ca2+-transient magnitude in relative isolation
from duration effects, irrespective of "normal"
Ca2+-level kinetics and spatial inhomogeneity.
Synapse localization
Nonetheless, imprecision in synapse localization must have contributed to variance in the relationship between the Ca2+ measurements and synaptic plasticity, tending to obscure our results. To limit this source of variance, we placed a fine stimulating electrode (20 µm diameter tip) as close as 0.5 mm from the dendrite and measured "plasticity-related" Ca2+ along a ~40 µm length of dendrite. Also relevant to the probable magnitude of this source of variance, the Ca2+ gradient along the apical dendrite was typically ~1 nM/µm within 50 µm of the presumed synaptic locations at the time of peak Ca2+. In spite of this variance, however, some significant and intriguing results emerged from the data.
Distinct Ca2+ thresholds for LTD and LTP
These data provided indirect support for a class of models in
which Ca2+ levels above a threshold reset the
activity-level balance between opposing molecular effectors of synaptic
strength (e.g., Lisman 1989). Specifically, we observed
LTD at lower peak Ca2+ levels and LTP at higher
levels (means: 335 ± 46 and 574 ± 27 nM, P < 0.05). Our data also provide estimates of the transition Ca2+ levels that separate the LTD plasticity
group from the no-plasticity subgroup on the
low-Ca2+ side (180 nM) and from the LTP group on
the high-Ca2+ side (541 nM). The latter
transition level is the mean peak Ca2+ level of
the no-plasticity subgroup intermediate between LTD and LTP, with peak
Ca2+ levels ranging from 450 to 600 nM. Ours is
the first study to document the existence of this no-plasticity
subgroup, an essential prediction of a Ca2+-based
version of the BCM learning rule (Bienenstock et al.
1982
).
In this study, we focused on peak Ca2+ levels
because prior evidence suggested that peak Ca2+
levels were related to synaptic plasticity (Hansel et al. 1996, 1997
;
Malenka et al. 1988
; Müller and Connor 1991
; Neveu and Zucker
1996
; Petrozzino et al. 1995
; Yuste and Denk 1995
). Of course, it is
likely that Ca2+ levels during the decay phase
also affect synaptic plasticity. However, because decay is fast when
Ca2+ is high, most of the recovery phase (the
long tail) is similar for a transient peaking at 600 nM and one peaking
at 300 nM, for example. To a rough approximation, the main difference
between such transients is the peak at 600 nM and swift decay to 300 nM. This difference in peak Ca2+ may contribute a
difference in net plasticity, while the remaining decay phase in common
may make similar contributions to net plasticity. Our use of five
relatively closely spaced transients, interrupting the first four decay
phases, may accentuate the effects on plasticity of differences in peak
Ca2+.
Relating our findings to earlier work, this laboratory previously
showed that increasing Ca2+ levels to 20 µM by
tetanic stimulation (Petrozzino et al. 1995) or bath
application of tetraethylammonium (Petrozzino and Connor 1994
) reliably induced robust LTP. These early studies differ from the present study in that a low-affinity indicator (mag-fura-5) was used to focus exclusively on LTP induced by large
Ca2+ transients in spines and fine dendrites.
Also an early study where Ca2+ transients were
produced by flash-photolysis of the caged-Ca2+
compound nitr-5 that was injected into the postsynaptic cell showed
that 2-4 µM Ca2+ induced only LTP
(Malenka et al. 1988
). A more recent nitr-5 study
suggested that Ca2+ levels of 300-500 nM induced
LTD and LTP in separate cells (Neveu and Zucker 1996
).
Unlike the present study, this study did not measure the
Ca2+ levels resulting from the
plasticity-induction protocol (flash photolysis) and, instead,
estimated Ca2+ from the duration of the flash,
cell depth, and model assumptions. Further support for distinct
Ca2+ thresholds came from imaging studies where
different stimulation protocols induced LTD or LTP and caused
relatively small and large Ca2+ transients,
respectively (Hansel et al. 1996
, 1997
). However, the
synaptic plasticity and imaging measurements in these studies were
conducted in separate experiments and calibrated estimates of
Ca2+ levels were not provided (Hansel et
al. 1996
, 1997
).
Another study used two tetanic protocols in the presence or absence of
picrotoxin to evoke Ca2+ transients that were
related to LTP and LTD (Otani and Connor 1998).
Several factors motivated the execution of the present study with a
single protocol of glutamate iontophoresis, achieving results that
extend the results of these earlier experiments without contradiction.
Compared to glutamate iontophoresis, tetani are likely to activate
different mechanisms of Ca2+ entry and additional
plasticity factors. As a result of recruitment of additional pathways,
the spatial domain of the Ca2+ transients
associated with tetani may not accurately reflect the synaptic
locations under test conditions. Also, Ca2+ flows
directly into the postsynaptic structure during tetani, reaching
transient local peaks that are likely to affect plasticity but are
difficult to measure accurately. In contrast, glutamate iontophoresis
generated slow and relatively homogeneous Ca2+
increases that could be measured accurately. Despite these
methodological differences, however, no contradiction exists between
the results of these two studies. The mean peak
Ca2+ associated with LTD in the earlier tetanic
study was ~460 nM, consistent with the range of peak levels
determined for LTD in the present study (180-500 nM). The tetanic
value could be a little high within the present range because
Ca2+ levels were near their peak for only ~10 s
with the tetanic protocol or because baseline
Ca2+ was relatively high in the earlier study.
Conditions of insufficient Ca2+ increase for
plasticity were also consistent between studies. In the earlier study,
no plasticity was observed under tetanic conditions leading to a peak
Ca2+ increase of ~30 nM from baseline, safely
below our new threshold for the induction of LTD (an increase of 120 nM
from baseline). Finally, the mean Ca2+ increase
associated with LTP in the earlier tetanic study was >1 µM,
well above our minimum threshold for LTP induction.
In summary, our new results are consistent with earlier data while
going beyond them to show, with one induction protocol and one
Ca2+ indicator, that low
Ca2+ levels induce LTD, high
Ca2+ levels induce LTP, and intermediate
Ca2+ levels are associated with a no-plasticity
domain. In general, of course, Ca2+ thresholds
for the induction of synaptic plasticity are likely to be affected by
Ca2+-signal duration and various pre- and
postsynaptic factors, perhaps including prior neuronal activity
(Abraham and Tate 1997; Bienenstock et al.
1982
). During our iontophoretic protocol, however, prolonged postsynaptic Ca2+ transients are uncoupled from
presynaptic activity. With this protocol, it may be that a steady-state
balance of Ca2+-dependent enzymes plays a
relatively large role in determining the ensuing synaptic plasticity.
Thus our estimates of the threshold Ca2+ level
for LTD (~180 nM) and for the transition from LTD to LTP (~540 nM)
in this simplified context may contribute to the ongoing effort to
construct a molecular model of Ca2+-dependent
synaptic plasticity (Grzywacz and Burgi 1998
;
Lisman 1989
; Malenka and Nicoll 1999
;
Soderling and Derkach 2000
).
![]() |
FOOTNOTES |
---|
Present address and address for reprint requests: R. J. Cormier, Dept. of Psychiatry, Washington University School of Medicine, St. Louis, MO 63110 (E-mail: cormierb{at}psychiatry.wustl.edu).
1 In one cell, both pathways were unusually responsive to distal glutamate application. However, no other basis existed for excluding these data and their relation to corresponding Ca2+ levels provided nonessential support for this study's conclusions.
Received 1 March 2000; accepted in final form 21 September 2000.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|