HZPPOCAMPUS, VOL. I , NO. 1, PAGES 79-91, JANUARY 1991
Long-term Enhancement of CAI Synaptic Transmission is Due to Increased Quantal Size, Not Quantal Content T. C. Foster and B. L. McNaughton Department of P s y c h o l o g y , University of Colorado, Boulder, CO 80309-0345 U.S.A.
ABSTRACT Quantal components of Schaffer collateral synaptic transmission recorded intracellularly from CAI pyramidal cells were examined using 2 methods: simultaneous recordings of CA3-CAI cellpairs, and minimal electrical stimulation in stratum radiatum. Quantal parameters estimated by the method of failures and by a computer algorithm that optimized parameter estimates using deconvolution of background noise were highly correlated. EPSP-amplitude histograms of CA3CAI cell pairs (N = 10) and minimal electrical stimulation (N = 33) could be adequately described either by Poisson o r binomial statisticc, or by both. and exhibited similar estimates of unit quantal size ( 4 ) and mean quantal content ( m ) .Paired-pulse stimulation with SO msec between stimuli resulted in an expected facilitation in the EPSP amplitude and increase in In during the second response, as estimated by noise deconvolution, by the decrease in apparent failures, and by a decrease in the coefficient of variation of the EPSP. Tetanization of the Schaffer collaterals that induced long-term enhancement (LTEILTP) of the population response was associated with an average increase in y for minimal-stimulation responses, with no significant change in any estimate of r n . Taken together, these data indicate that, under the present experimental conditions. LTE is expressed as an increase in quantal size, rather than an increase in the number of quanta released per presynaptic impulse. Although this is t 7 o r definitive evidence for a postsynaptic mechanism, these findings d o further restrict the classes of possible presynaptic mechanisms that may be proposed to account for LTE expression. Key words: long-term potentiation, quanta1 a n a l p i s , hippocampus, synaptic transmission
Although there has been substantial progress in recent years in understanding the physiological mechanisms underlying the factors that control the indwtion of long-term enhancement (LTEILTP) of hippocampal synaptic transmission (e.g., see reviews by Bliss and M. A. Lynch, 1988; G. Lynch et al., 1988), the cellular changes through which this increased synaptic efficacy is c~xpwssedremain a matter of open controversy. In particular, although there is little dispute that the induction process involves integration of afferent input (McNaughton et al., 1978) by the postsynaptic neurons, postsynaptic depolarization (Gustafson et al., 1987). and activation of postsynaptic NMDA receptors (Collingridge et al.. 1983; Harris et al., 1984), a substantial case has been developed suggesting that the expression of L T E involves an increase in evoked transmitter release from presynaptic terminals (Bliss and M. A. Lynch, 1988). As suggested by M. A. Lynch (1989), such a presynaptic change could, in principle, be brought about through chemical feedback from postsynaptic sites. On the other hand, there is also substantial
_ _ _ _ _ _ ~ __ _ _ _
~-
Correspondence and reprint requests to T. C Fostei, Ari7ona Research Laboratories, Division of Neural Systems, Memory and Aging, 384 Life Sciences North Building, Tucson, AZ U S A 85724
evidence that would favor a postsynaptic locus of expression as well (e.g.. Kauer et al.. 1988; Mueller and G . Lynch, 1988; Davies et al., 1989). Thus, the matter remains a major open question in the field. Since this manuscript was accepted for publication. a report has appeared in which evidence is presented for an increase in mean quantal content of the EPSP (Malinow and Tsien, 1990). Because the latter finding is at variance with our own conclusions, consideration of the possible sources of discrepancy is deferred until the discussion section. Unlike the relatively short-term phenomenon of potentiation seen at both neuromuscular (Magleby and Zengel, 1975) and perforant path (McNaughton, 1982) synapses, L T E is not due to an increase in the quantal release probability ( p )at the presynaptic terminal’. However. the possibility was left open for other presynaptic mechanisms of expression, such as an increased number of release sites ( n ) ,o r an increase in the amount of transmitter released per quantum. The experimental results described in this report represent a further restriction of these presynaptic possibilities by demonstrating
’
T h k is one of the major reamns for our persistence in the use of the term “enhancement” in referring to this phenomenon.
79
80
HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991
that. under our experimental conditions, LTE in Schaffer collateral synapses in CAI is due to an increased quantal size ( 4 ) .We leave open the question of whether this increase is pre- or postsynaptically generated. Quantal analysis of synaptic transmission, first proposed by del Castillo and Katz (1954) as a statistical description of factors contributing to variance in synaptic potentials. represents a powerful physiological method for clarifying the basic nature and location of changes in synaptic efficacy. Although applications of the technique to hippocampal circuits have been hampered by the problem of resolution of the small synaptic signals arising from individual afferent fibers (Voronin. 1988: Hess et al., 1987: Sayer ct al.. 1989, 1990: Fricdlander et al., 19901, a substantial improvement has recently become available through the application of the method of noise deconvolution (Edwards et al.. 1976: Korn et al.. 1982: Wojtowicz and Atwood, 1986: Sayer et al., 1989, 1990). This is a statistical procedure that takes account of variation in the distribution of quantal events caused by background recording noise, in order to estimate the parameters of the probability distribution underlying the fluctuation of evoked EPSPs. In its least constrained form (Edwards et al.. 1976. Sayer et al.. 1989. 1990) no assumptions are made about the underlying statistical distribution other than that the distribution is discrete, and that the variation in response size is independent of the background noise. Because of this lack of constraint on the statistical model, relatively high signalto-noise ratios and/or large sample sizes are required in order to obtain statistically reliable parameter estimates. However, in the few cases in which reliable estimates have been obtained for Schaffer collateral synapses ( e g . Sayer et al., 1989, 1990). these have not differed substantially from the predictions of the binomial model. This provides N prior-ijustification for making the binomial assumptions that all release sites from a single axon have uniform probabilities of release, and generate equivalent-sized quantal events in which the intrinsic variance of quantal sizes is negligible. The advantage of assuming a binomial process (or its Poisson approximation) is that reasonably reliable parameter estimates can be obtained with smaller samples and/or a lower signal-lo-noise ratio. Moreover, even if the estimates for individual cases are not highly reliable, the means from a number of experiments should nevertheless provide a valid estimate of the population mean. which can be used to assess how these parameters change on average under different experimental conditions. In the present report, we compare how the estimated binomial (or Poisson) parameters change during L T E and during the short-term phenomenon of paired-pulse facilitation. for which there is good evidence for an increased quantal release at hippocampal and most other synapses. The general conclusions derived from the binomial deconvolution method are verified by 2 other methods: the method of failures, and the coefficient of variation method. Although both of the latter methods rely upon the Poisson approximation to the binomial distribution for the accuracy of their parameter estimates, a significant change in EPSP size, without significant changes in m estimated by these methods, would constitute relatively assumption-free evidence for increased quantal size. In principle, the most convincing studies of synaptic change at individual connections should involve simultane-
ous recording from pairs of connected neurons (Kuno, 1971), for example, one pyramidal cell in CA3 and another in C A I . Indeed. although the success rate for obtaining synaptically connected dual CA3-CAI impalements is very low (Friedlander et al.. 1990: Sayer et al., 1989. 1990). quant'd 1 events at this synapse have been characterized in this manner. However, the low probability of such successful dual impalements, in addition to an unexpectedly low success rate of inducing LTE (Friedlander et al., 1990) at these connections (a problem we shall discuss further below). has made the use of this method for analyzing the mechanism of LTE expreysion impractical. On the other hand. previous studies indicate that minimal stimulation can be used successfully to activate unitary synaptic responses. This method has been used to examine differences in quantal parameters due to the location of synaptic contacts, as well as short-term changes in quantal parameters during repetitive stirnulation (Hess et al.. 1987; Sayer et al., 1989). For example. McNaughton et al.. (19x1) showed that synaptic depression in perforant path synapses during repetitive activity was accounted for by a decrease in tn with little o r no decline in 4 . They also concluded from the step-like increase in EPSP amplitude near threshold, that excitation failure in the presynaptic axon contributed relatively little to response variation. Increased confidence in the method of minimal stimulation would, however, be provided by evidence that this method yields estimates of unitary response size and general quantal parameters that agree closely with the dual impalement method. This issue is taken up in the present studies. The minimal-stimulation method has the advantage that the timing of presynaptic activity can be better controlled. We make use of this advantage to obtain additional verification of the statistical procedures, by showing that they yield the expected increase in the /zp product during paired-pulse facilitation, a process already known to involvc an increase in transmitter release at most synapses (del Castillo and Katz, 1954: Martin and Pilar. 1964: Wernig, 1972: Zucker, 1973; McLachlan, 1975: Hirst et al.. 1981: Smith, 1983; Hess et al.. 1987).
MATERIALS A N D METHODS Brains were removed from adult rats (300-400 g) under ether anesthesia, and the hippocampi were dissected out. Slices (450-500 p,m) were cut parallel to the alvear fibers using a tissue chopper, and transferred to a standard recording chamber (Haas, et al., 1979). During the experiment. slices were perfused ( 1 mlimin) at 30-32°C with oxygenated artificial cerebrospinal fluid containing (in mM): NaCl 124. KCI 2. K H 2 P 0 4 1.25, MgS04 2. CaCll 2. NaHCO, 26, glucose 10. Humidified air (95% Oz. 5% COz) was continuously blown over the slices. Intracellular recordings were obtained from CA 1 pyramidal cells located in the CAI stratum pyramidale (approximately 600 p m from the CA3ICAI border) with glass micropipettes (20-90 M a ) pulled from 1.0 mm O.D., thin-walled tubing, and filled with 3 M K' acetate. Resting membrane potential and input resistance were monitored and recorded throughout the experiment. C A I cells were considered suitable if they exhibited a resting membrane potential of - 60 mV o r below (mean, -73.9 mV ? 10.2 SD). a spike of at least 50 mV from onset to peak (excluding the EPSP) (mean, 65.0 mV +- 1 I .4
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster and McNaughton
INTRACELLULAR
-ij
CAI INTRACELLULAR
\I R
STIM
CA3 INTRA OR XTRACELLULAR
CA3-CA1
Minimal Stimulation
=-
J Fig. 1 . Top: Stimulating and recording electrode configurations for CA3-CA 1 spike-triggered EPSPs (Left) and Schaffer collateral minimal-stimulation EPSPs (Right). Bottom: Examples of intracellularly recorded evoked EPSPs (a and b), apparent failures (c) and the EPSP averages (d) for single CA3 spikes (Left) and for minimal-stimulation (Right). The average of the CA3 unit spike, which precedes the apparent onset of the EPSP by about 3 msec, is shown in (el. Calibrations. 2 msec and 250 kV.
SD), and an input resistance greater than 16 MR (mean, 26.8 MR & 8.2 SD). Additional constraints for quantal analysis included the establishment of steady state release conditions prior to tetanic stimulation, such that unitary EPSPs averaged over 50 trial blocks for 150 or more trials did not show a systematic change in response tendency or fluctuations in mean amplitude of more than about 20%. Fig. 1 illustrates the 2 methods used to obtain intracellular CAI EPSPs from stimulation of presumed single CA3 afferents. For studies involving simultaneous recording from CA3 and CAI cell-pairs, single CA3 cells were recorded intracellularly as described above, or extracellularly with glass pipettes (2-10 MR) filled with perfusion medium and glutamate (10 mM). In these CA3 cells, action potentials were elicited by injection of depolarizing current o r iontophoretic ejection of glutamate, respectively. In both cases the currents were adjusted in order to elicit less than one spike per second. Online spike-triggered averages of CAI responses were used for initial identification of CA3-CAI connections. For minimalstimulation experiments, diphasic constant current stimulation (100 psec, 5-50 FA) was delivered via twisted-pair, plat-
81
inum-iridium wire stimulating electrodes located in stratum radiatum. Stimulus intensity was adjusted with a high resolution potentiometer. Minimal-stimulation consisted of paired pulses of identical intensity, separated by 50 msec, with one such pulse pair delivered every 5 seconds. Once a CAI cell met the criteria for recording, stimulation intensity was adjusted to evoke the smallest consistent averaged response for conditioning and test stimulation (McNaughton et al., 1981). At these stimulus levels, failures of transmission could be frequently observed in response to the first stimulus at to, and less frequently in the second response at tSo. stimulation was continued for 10I5 minutes in order to insure stabilization of the responses. One hundred fifty to 250 response pairs were then recorded and averaged in real time over 50 trial blocks to insure response stability. If the average amplitude tended to increase or decrease, another 150 trials were collected until stability was achieved. After collection of control data, the stimulus intensity was adjusted to elicit a population response of 2-8 mV, and 5-10 records were taken. Stimulus intensity was then raised further, to above the threshold to elicit an action potential, and high-frequency stimulation (4 bursts of 8 pulses at 400 Hz) was delivered in an attempt to induce L T E by the cooperativity method (McNaughton et al., 1978). Stimulus parameters were then returned to pre-tetanus values and minimal-stimulation responses were collected during the next 2060 minutes. Enhancement of the population EPSP was examined after tetanus and again at 20 and 40 minutes later, using the same stimulus level used to collect the population response baseline. Stimulus timing and data collection were under computer control. Electrophysiological signals were stored on FM tape for off-line analysis or stored directly on computer as 50 msec records of evoked responses o r 50 msec records of background noise only (i.e., baseline fluctuations in membrane potential and electronic noise in the absence of a CA3 spike or electrical stimulation). Individual records for each evoked response consisted of a 2 msec baseline period and the first 48 msec after stimulation. Noise records were collected just prior to each response trial for minimal-stimulation and between CA3 discharges for spike-triggered responses. Two time-windows ( I -2 ms), corresponding to the baseline period before the stimulus artifact and the EPSP peak, were defined for the average response. These window regions were then used to measure unitary EPSP amplitudes for individual records by subtracting the mean voltage within the baseline window from the mean within the peak window. This process was also applied to the noise-only records, using the same window parameters, to obtain a measure of the expected noise contribution to the EPSP measures. In order to analyze fluctuations in evoked responses, separate EPSPand noise-amplitude histograms were constructed. These were divided into 30 bins over the range of observed responses (bin width for EPSP histograms range, 19-85 kV). A chi-square test was used to insure that the noise histogram conformed to a Gaussian distribution at P < .05. Two independent methods were used t o extract the quantal parameters from the EPSP histograms. The parameters for a Poisson distribution ( m , 4 ) o r binomial distribution ( n . p , q ) were determined by a computer algorithm that found the pa-
82
HIPP0CAMPU)S VOL. 1, NO. 1 , JANUARY 1991
rameters that best fit the observed response distribution. The expected proportion of responses for each of the 30 bins was calculated by the binomial or Poisson formula, using sums of Gaussian curves whose variances were obtained from the noise distribution, whose means represented multiples of the single quantal size, and whose integrals were equal to the expected proportion of multiple release events. The expected distribution was then compared to the observed distribution using a chi-square test. Parameter values were altered according to a nonlinear optimization routine (Nclder and Mead. 1965; Press et al., 1988) until the lowest chi-square value. and thus best estimate for each parameter, was obtained. The algorithm was constrained to positive values for all variables. Mean quantal content was also calculated by the method of failures (del Castillo and Katz. 1954) according to the equation:
a
I
4
1000
5
d
L
800
600 400 200 0
2
4
6
8
b where N is the total number of samples and tio i s the number of times an evoked response failed to appear. Because the noise was found to be normally distributed. the number of failures was estimated as two times the number of responses less than Lero plus those that were equal to zero (Nicholls and Wallace. 1978). Quanta1 size was then calculated as the mean EPSP amplitude divided by m,-. The failures method was not used if the number of failures was less than 6% or greater than 95% of the total number of samples (Martin. 1977). This method was used exclusively tor one CA3-CAI cell-pair in which only 80 responces were obtained. Because of the closc correspondence of quantal parameters estimated by the 2 methods (see results) and the inability to make quantal estimates for some cells resulting from a ION failure rate, all statistical analyses were performed on estimates calculated by the optimization procedure except where specified. Statistical significance was assessed using Student {-tests with a = 0.05. Finally, for comparison with the recent results of Malinow and Tcien (1990). possible changes in EPSP mean quantal content were assesscd using the coefficient of variation method according to 1.he lormula w,,
=
M'lcr'.
where M is the mean EPSP amplitude and v2 is the variance of the EPSP amplitude after subtraction of the background noise variance.
RESULTS Characterization of CA3-CA1 spike-triggered synaptic responses CA3 spike-triggered responses were observed in 10 out of 176 cell-pairs tested. The mean time from CA3 spike peak to apparent EPSP onset wits 5.19 +- 1.95 (SD) msec. Synaptic responses to single CA3 spikes consisted of clear small EPSPs interspersed with apparent failures (Fig. 2). The mean baseline amplitude for all cell pairs varied around zero (mean. - 8.5 & 9.1 SD FV), with an overall mean for the noise stan-
10
12 14
16
18
TIME (ms)
1
30 1
rz
25
0
15
3
u
!
q SIZE
= 233 U V
m = 1.2 BINSlZE=47uV
2o
10
5 0
0 mV
0.5
1.o
Fig. 2. ( a ) Illustration of firilui-cs and mono5ynaptic EPSPs I-ecordcd intracellularly in one CAI cell i n response to 50 single CA3 spikcs whose time of occurrence is indicated by the arrow. The quantal amplitude estimated by the method of failures was 350 p V . (b) Example of an EPSP amplitude frequency distribution (BARS) for 395 responses from ;inother CA3-CAI cell-pair and the best-fitting Poisson distribution ( L I N E ) estimatcd by the optimization algorithm. Arrows indicate multiples of estimatcd quantal amplitude.
dard deviation of 155.6 2 92.4 (SD) k V . Estimated nz values calculated by the method of failures and by the coefficient of variation method were highly correlated with the results fi-om the optimization procedure foi- 9 cells with at least I50 responses (mf: r2 = 0.85, P < .OOO5: 1 7 1 ~ r'~ : = 0.88, P < .OOO2). Descriptive statistics indicated that response histograms for 7 of the 9 cell-pairs could be adequately described by either Poisson o r binomial statistics at the 0.01 level of significance. The contribution of quantal parameters to synaptic strength for the population of cell-pairs was tested by plotting 172 and q against mean EPSP amplitude. A simple regression indicated that both q and m were significantly correlated with EPSP amplitude. Examination of several factors that might have contributed to the variability of q and rn revealed no correlations with input resistance. resting membrane potential, onset latency, 10-90% rise-time o r EPSP half-width. Also, 4 and 171 were not correlated. For 4 cell-pairs the CA3 cell was activated by antidromic stimulation of the Schaffer collaterals. However, relatively high stimulation currents, which produced large EPSPs and
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster and McNaughton
Table 1. Quantal Parameter Estimates for 3 CA3-CA1 Connections for which Sufficient Data Were Obtained to Estimate Effects of LTE Inducing Stimulation ___ EPSP (pV)
qq,t (I*V)
83
~~
m,,,,, ~~~
Cell
Pre
Post
Pre
Post
Pre
I 2 3 ~~
Post ~
~~~
-
72 2 126 S 389 9
69 1 181 1 130 8 __
141 210 204
IS9
252 225
051 060 I91
0 42 075 0.67
*
~
spikes in the CAI cell, were necessary for antidromic activation, indicating that placement of the stimulating electrode was relatively distant from the fiber of interest. In one case, the stimulating electrode was moved to 2 other locations along the vertical axis of the stratum radiatum with little success in reducing the current necessary for antidromic driving. Once a cell could be activated by antidromic stimulation. the intensity was further increased to elicit a CAI cell spike, and tetanizing stimulation was delivered. Tetanization of the Schaffer collaterals resulted in enhancement of the population test response measured 15 minutes later. Only 3 cell-pairs were maintained such that at least 150 CA3 spikes occurred during the 5-30 minutes following tetanization, and hence could be subjected to quantal analysis. Based on the considerable fluctuations that occur in rn in the minimal-stimulation experiments (see below), and on the uncertain reliability of the parameter estimates from individual experiments, we d o not consider 3 cells a sufficient population from which to draw any conclusions concerning the question of how quantal parameters change following LTE induction. We believe that this question can be answered only from the mean of a reasonably large number of experiments. Nevertheless, the data from these 3 cells are presented in Table 1, for the sake of completeness. In some cases, facilitation and summation of CA3 spikeelicited EPSPs were observed during a burst of CA3 cell spikes. Figure 3 shows examples of bursts in which failures were observed for the initial spike, and facilitation and/or summation was observed for later spikes.
*
n , b
C
Quantal analysis using minimal-stimulation: Similarity to CA3-CAI spike-triggered responses Paired-pulse stimulation of the Schaffer collaterals was used to elicit minimally evoked conditioning (to) and test (tso) EPSPs separated by 50 msec. Background recording noise in these experiments was comparable to the dual cell impalement experiments (mean, 172 p V ? 59 SD). Sufficient failures were observed for application of the method of failures in all but 1 of 33 cells at to and all but 5 cells at t5(). Estimates of rn, calculated by the method of failures and by the optimization procedure for to and tS0 responses, were highly consistent (rnf vs. rnopt: r2 = 0.74, P < ,0001, slope = I .04). As expected (see discussion) rncv was less well correlated with mopt, and increasingly overestimated rn at higher values (mCv vs. mopt:r2 = 0.3 P < .0001, slope = 2.08). These relationships are illustrated in Figure 4. One response histogram for to and 1 for tSocould not be described by Poisson or binomial statistics at the 0.01 level of significance. For to the observed
Fig. 3 . Postsynaptic CAI responses to bursts of CA3 unit activity. (a) and (b) illustrate pairs of spikes in a single CA3 cell in which failure of transmitter release at the CAI synapse can be seen (*). Another CA3-CAI cell-pair in which failure of release was observed for the first spike in a burst is shown in (c). The CAI responses in (b) and (c) are composed of multiple quanta. Calibration, 4 msec, CA3: 200 p V , CAI: 500 PV. response distributions of 20 of the 33 cells were best fit by a Poisson function. On the other hand, there was a tendency for t5() amplitude histograms to be better fit by a binomial distribution (26 out of 33) and to exhibit fewer failures. This is to be expected when p is increased, as is known to occur during synaptic facilitation. The mean and range of to minimal-stimulation quantal pa-
84 HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991 12 -
9-
m, 6-
0
2
1
3
opt
l2
9
6 mcv
1
1
0 0 0 0
1
nificant increase in the estimated rn for to and t5()with no significant change in q or p. However, only one Vu and two V5(,response histograms could adequately be fit by a binomial distribution, most likely because of differences in q for different afferent fibers. In addition, in 2 cases, increasing the EPSP above 300% resulted in a small population EPSP recorded extracellularly upon withdrawal from the cell, which may also have contaminated these results. Previous studies have indicated that EPSP rise-time and q are influenced by synaptic location (Andersen et al., 1980; ’Turner, 1988; Sayer et al., 1989). The mean overall 10-90% rise-time was 4.46 ? 1.61 (SD) msec (N = 33). The rise-time was reduced for responses elicited by stimulating electrodes localized to the inner one-third of the stratum radiatum (3.56 & 1.16 msec, N = 8), and increased when the stimulating 4 electrodes were localized to the outer one-third of the stratum radiatum (5.62 f 1.23 msec, N = 6). While mean 4 was slightly larger for the proximal vs. distal electrode locations, statistical analysis did not indicate a significant difference. Neither mean EPSP, quantal size, nor quantal content were correlated with input resistance o r rise-time. EPSP amplitudes were somewhat correlated with resting potential (r’ = 0.214, P < .046). However, the correlation of resting potential with quantal size was not significant. The main result of this part of the experiment was the rather close correspondence of quantal parameters obtained with the dual impalement and minimal-stimulation techniques as summarized in Table 2 . columns I and 2. Quantal analysis of facilitation at the CA3-CA1 synapse
Y
0
1
2
3
4
opt Fig. 4. ‘lop: Correlation between quantal contents for pairedpulse minimal-stimulation of the Schaffer collateral.; calculated by the method of failures (mr) and by the binomial deconvolution method using nonlinear parameter optimization (mupr).Data are plotted for the to and tso responses obtained prior to LTE induction. The correlation (r2 = 0.74) was significant at P < .0001. There were no significant differences between methods in the estimates of either rn or q. Bottom: Same as top but data. for the coefficient of variation method (mcv). This method significantly overestimated m,,, and underestimated q. Moreover, the correlation with mOprwas poorer (r’ = 0.3). The increased variance at higher WI is expected theoretically because of the nonlinear effect of increased p (due partly to facilitation at t5(J on mC.. Solid lines in both figures indicate one to one correspondence, not regression.
rameter values were very similar to those observed for the CA3-CA I cell-pairs (Table 2). Like CA3-CA I spike-triggered responses, minimal-stimulation EPSP means were correlated with estimates of both y (r’ = 0.19, P < .Of I ) and i n (r’ = 0.3 I , P < .0007). Increasing the mean EPSP amplitude (150-4S0%5) by increasing the stimulation intensity (N = 7) resulted in a sig-
The mean overall V,,, EPSP amplitude was increased 67 f 34.8 (SD)%’relative to VO(Fig. 5 ) . Table 2, columns 2 and 3, illustrate the quantal parameter means and ranges for conditioning and test stimuli. During paired-pulse facilitation, all 3 estimates of M were significantly increased. Quantal size estimated by optimimtion and by the failures method was not changed. A significant decrerise occurred in the quantal size estimated by the coefficient of variation method. This is to be expected because of the strong nonlinear tendency of this method to overestimate m as p increases. The magnitude of facilitation was negatively correlated with mo (r’ = 0.32, P < ,0006). suggesting that facilitation was inversely dependent on the initial release level. For binomial analysis, there was a significant tendency for n to increase in the tso response, but this increase was not correlated with facilitation of the EPSP (r2 = 0.036. P < .29). On the other hand. the expected significant increase in p was observed and was correlated with facilitation (r’ = 0.35: P < .0003). Quantal analysis of LTE for minimal-stimulation of CA3 fibers
For all 33 cells included in the present analysis, tetanic stimulation elicited enhancement of the population response. Data were rejected if this failed to occur. In order to avoid biasing the results, however, data were nor rejected if the population response was enhanced but the minimal-stimulation response failed to show an increase.
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster a n d M c N a u g h t o n Table 2a. S u m m-.a r.y of Q u a n t a 1 Analysis Results for D u a l Recording a n d Minimal-stimulation Experiments* .~ ~~~
EPSP (pV) (FV)
qupl
mopt
nopt Pop1 qfad (CLV) mk,,l
(jcv
(PV)
in,,
..
1
2
3
4
5
6
CA3-CA1
pairs
Minimal Stim Baseline to
Minimal Stim Baseline tso .~
Minimal Stim LTE 20 min to
Minimal Stim LTE 20 min tso _.
Minimal Stim LTE 40 min to
264 (60) 72-674 220 (18) I 4 1-350 1.18 (0.17) 0.51-2.04 3.8 (0.29) 2-5 0.29 (0.03) 0.17-0.46 227 (33) 136-485 I.09 (0.16) 0.52-2.08 183 (19) 89-296 1.38 (0.22) 0.49-2.58
290 (32) 101-75 1 245 (15) 88-398 1.30 (0.09) 0.43-2.70 5.28 (0.3) 2-9 0.24 (0.01) 0.09-0.43 231 (27) 7 1-460 1.38 (0.11) 0.45-2.57 1.61 (16) 8-390 2.83 (0.51) 0.56-16.31
474 (38) 198-1 181 258 (15) 77-407 1.96 (0.12) 0.68-3.9 6.03 (0.39) 3-1 I 0.34 (0.02) 0. 15-0.61 216 (17) 7 1-460 2.04 (0.12) 0.69-2.92 126 (14) 26-380 4.66 (0.39) 0.73-10.25
371 (33) 8 1- I004 301 (20) 98-553 I.36 (0.09) 0.36-2.70 5.47 (0.46) 2-12 0.26 (0.02) 0. I 1-0.60 265 (17) 99-544 1.40 (0.11) 0.29-2.93 230 (38) 27- 126 I 2.62 (0.45) 0.13-14.65
580 (62) 136- I833 298 (18) 122-512 1.99 (0.14) 0.75-4.65 4.49 (0.22) 2-8 0.42 (0.02) 0.23-0.6 I 2.48 (18) I 15-447 2.01 (0.13) 0.75-2.94 208 (53) 27-179 I 5.07 (0.66) 0.10-17.27 .
386 (37) 67-760 323 (23) 98-560 1.25 (0.10) 0.32-2.92 5.28 (0.47) 2-1 I 0.26 (0.03) 0.09-0.66 301 (27) 88-583 I .37 (0.15) 0.25-2.94 255 (44) 45-1 I12 2.31 (0.33) 0.38-6.25..
-..
~
~~
~
~~
~~~
~
85
7 Minimal Stim LTE 40 min tsi, -~
586 (67) 135-1529 326 (23) 120-580 1.86 (0.14) 0.68-3.68 4.62 (0.21) 2-7 0.38 (0.10) 0.2 1-0.60 307 (37) 125-565 1.79 (0.15) 0.96-2.96 211 (35) 11-813 3.95 (0.60) 1.14- 13.10
* Using the binomial nonlinear optimization technique (opt). the failures method (fail) and the coefficient of variation method (cv). The following columnwise statistical comparisons were performed using t-tests with 01 set at 0.05: 1-2, 2-3, 2-4. 2-6. 3-5. 3-7, 4-5, 4-6. 5-7. 6-7. These tests addressed the general questions of how quantal parameters compared between C A 3 - C A I cell-pair recording and minimal stimulation; how quantal parameters changed at 2 different times after LTE induction in the population response: how quantal parameters changed during paired pulse facilitation at a 50 nisec interstimulus interval; and whether quantal parameters changed over time between the 2 post LTE epochs. The statistical summary is given in Table 2b.
Table 2b. C o l u m n Comparison ~~
1 vs L ~~~
~~~
ycvIkV)
0.80 (43) 0.92 142) 0.19 (42) 0.50 (43)
nicv
0.13 143)
q,.,,,(+v) mt.,,)
~.
...
~~~
~
n.0001 133) 0.09 133) 0.0001 133) 0.0006 (331 0.0001 (33) 0.45 12x) 0.0001 I281 0.007 f33) 0.003 133) ...
~~
2 vs. 4
~~~~~
EPSP I ~ V ) 0.62 (43) q,,,, ( + V ) 0.40 (43) 0.26 143) n7p, nt,pt 0.24 143) POP1
~
2v5 3
~
~
0.0015 133) 0.0001 (33) 0.40 133) 1.00 133) 0.36 ( 3 3 ) 0.0031 132) 0.76 132) 0.042 (331 0.93 ( 3 3 )
L vs. 6
3 vs.
~
5
~
0.025 1261 0.0001 126) 0.62 126) 0.33 (26) 0.71 126) 0.001 124)
0.76 (24) 0.074 (261 0.94 (26)
0.0047 133) 0.1)001 ( 3 3 ) 0.63 133) 0.23 0.15 0.029 0.065 0.085 0.54
.~
(3.11 1331 122) (22) (33) 133)
..
~~~~~~
~
3 vs. 7
0.093 0.0002 0.47 0.40 0.33 o.ooah 0.048 0.044 0.20
_
126)
126) 126) (26) 126) 118) 118)
(26) (26)
4
VS.
5
4
v5.
6
_ ~ . o.oo01(331 0.46 126) 0.75 133) 0.0001 133) 0.24 133) 0.0001 133) 0.54 (24) 0.0001 124) 0.29 133) 0.0001 1331 .
0.61 126) 0.16 (26) 0.53 126) 0.54 126) 0.25 124) 0.30 (24) 0.99 (26) 0.22 (26)
. .
-.
5 vs. 7
6 vs. 7
0.21 126) 0.27 (26) 0.20 126)
0.0001 1261
.
0.81 I261 0.0001 (26) 0.013 (26)
0.86 126) 0.082 126)
0.001 126)
0.26 117) 0.022 117) 0.82 (26) 0.039 126)
0.92 (23) 0.0002 123) 0.021 (261 0.0001 126) ~
Summary of statistical results for columnwise comparisons on data from Table 2, showing (Y levels for student's I-tests. Repeated measures comparisons were employed for all tests except 1 vs. 2. in which the samples were independent. Results that were statistically significant at 01 = 0.05 are underlined. Except for the italicized comparisons, the results support the following conclusions: ( 1 ) minimal-stimulation samples single fibers from the same population a s the dual i C A 3 - C A I ) recording experiments: (2) paired-pulse facilitation leads to an increase in m with no change in q: (3) L T E expression involves an increase in q with no change in m . The discrepant dec,r-euAein q,, during pairedpulse facilitation is expected from the nonlinear relationship between tn,, and true m under changes in p.
The population EPSP amplitude increased 40 k 34 (SD)% immediately following tetanus (i.e., during post-tetanic potentiation), and remained elevated 30 k 43% (N = 33) at 20 minutes post-tetanus and 21 2 47% (N = 26) by 40 minute post-tetanus. The mean EPSP amplitude for minimal-stimulation during the 5-20 minute period was increased 35 t- 57% and 27 2 40% for Vo and Vs0, respectively. During the 2040 minute period Vo and Vso were increased by 27 t- 36% and 21 k 55%, respectively. It is important to note that the overall mean fractional increase in EPSP amplitude for minimal-stimulation elicited EPSPs was similar to that observed for the population response. Figure 6 shows the time course
of these increases relative to the corresponding mean pretetanus amplitudes. While on average, the minimal-stimulation mean EPSP amplitude was increased, not all cells exhibited such an increase. The mean EPSP amplitude decreased for 9 of 33 cells at t,, and 10 of 33 cells at tS0during the 5-20 minute period after tetanization. During the 20-40 minute period, 10 of 26 cells at to and 9 of 26 cells at t50exhibited a decrease in mean EPSP amplitude relative to pre-tetanus. As with the pre-tetanus data, most post-tetanus amplitude histograms at to were best fit by a Poisson function (20 of 33). and only 1 histogram did not conform to a Poisson distribution
86
HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991
-0.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
mV
-0.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
mV
Fig. 5. Synaptic facilitation evoked by paired-pulse stimulation of the Schaffer collaterals for 1 cell. Top: Mean (solid line) and standard deviation (dotted line) of CAI EPSPs (N = 150). Calibrations, 4 msec, 100 k V . Bottom: Best fitting binomial distributions (line) of the observed response amplitude distributions (bars) for conditioning (Left) and test (Right) stimuli. Bars = 50 bV.
at P < .01. Histograms for t5,, were best fit by a binomial function (21 of 33), with 2 unable to conform to a binomial function and 4 unable to conform to a Poisson function at P < .01. The means and ranges of the alterations in quanta1 parameters resulting from tetanization are summarized in Table 2. columns 2-7. and Figure 7 . The major observation was that, whereas there was a significant increase in 4 at both time points after tetanization, there was no significant change in either m , n , o r p . Moreover, the average fractional increase in 4 was sufficient to account for the mean increases in both the minimal-stimulation EPSP and the population EPSP. Although there was no average increase in rn, nevertheless, there was often substantial variation i n this parameter between the baseline and post-tetanization periods. Indeed, the standard deviation for fractional change in tn was substantially greater than the corresponding measure for change in q (to: 0.51 vs. 0.28: t50:0.51 vs. 0.27). As a result, sometimes a decrease in m was large enough to mask an increase in Vo in spite of an increase in 4 . In other cases, increases in m produced an increase in Vo beyond that caused by the increase in 4 . There was a significant negative correlation (I' = 0.31, P < .001) between the change in m following teta-
nization and the change in paired-pulse facilitation. suggesting that much of the variance in m was due to variance in p (McNaughton, 1980. 1982; Zalutsky and Nicoll, 1990). This was corroborated by a significant positive relation between the change in m and the change in p estimated from binomial optimization (6 = 0. I I . P < .04). and by the lack of significant rclation to the change in binomial n. A possible bias may result from selection of data for inclusion in the analysic on the basis of whether the single fiber EPSP is increased, rather than the population EPSP (Voronin, 1988; Malinow and Tsien, 1990). 'To illustrate this possibility, we present the following reanalysis of our data, based on only those fibers that exhibited a V,, increase greater than 25% during the 5-20 minutes after tetanus. Mean EPSP values increased approximately 2 times that observed for the population of afferents (83 2 1 I%, N = IS), and the number of included single-fiber experiments was reduced by approximately one-half. For these experiments, both 4 and m exhibited significant increases "4": P < ,0006;mo:P
Long-term Enhancement of CAI Synaptic Transmission is Due to Increased Quantal Size, Not Quantal Content T. C. Foster and B. L. McNaughton Department of P s y c h o l o g y , University of Colorado, Boulder, CO 80309-0345 U.S.A.
ABSTRACT Quantal components of Schaffer collateral synaptic transmission recorded intracellularly from CAI pyramidal cells were examined using 2 methods: simultaneous recordings of CA3-CAI cellpairs, and minimal electrical stimulation in stratum radiatum. Quantal parameters estimated by the method of failures and by a computer algorithm that optimized parameter estimates using deconvolution of background noise were highly correlated. EPSP-amplitude histograms of CA3CAI cell pairs (N = 10) and minimal electrical stimulation (N = 33) could be adequately described either by Poisson o r binomial statisticc, or by both. and exhibited similar estimates of unit quantal size ( 4 ) and mean quantal content ( m ) .Paired-pulse stimulation with SO msec between stimuli resulted in an expected facilitation in the EPSP amplitude and increase in In during the second response, as estimated by noise deconvolution, by the decrease in apparent failures, and by a decrease in the coefficient of variation of the EPSP. Tetanization of the Schaffer collaterals that induced long-term enhancement (LTEILTP) of the population response was associated with an average increase in y for minimal-stimulation responses, with no significant change in any estimate of r n . Taken together, these data indicate that, under the present experimental conditions. LTE is expressed as an increase in quantal size, rather than an increase in the number of quanta released per presynaptic impulse. Although this is t 7 o r definitive evidence for a postsynaptic mechanism, these findings d o further restrict the classes of possible presynaptic mechanisms that may be proposed to account for LTE expression. Key words: long-term potentiation, quanta1 a n a l p i s , hippocampus, synaptic transmission
Although there has been substantial progress in recent years in understanding the physiological mechanisms underlying the factors that control the indwtion of long-term enhancement (LTEILTP) of hippocampal synaptic transmission (e.g., see reviews by Bliss and M. A. Lynch, 1988; G. Lynch et al., 1988), the cellular changes through which this increased synaptic efficacy is c~xpwssedremain a matter of open controversy. In particular, although there is little dispute that the induction process involves integration of afferent input (McNaughton et al., 1978) by the postsynaptic neurons, postsynaptic depolarization (Gustafson et al., 1987). and activation of postsynaptic NMDA receptors (Collingridge et al.. 1983; Harris et al., 1984), a substantial case has been developed suggesting that the expression of L T E involves an increase in evoked transmitter release from presynaptic terminals (Bliss and M. A. Lynch, 1988). As suggested by M. A. Lynch (1989), such a presynaptic change could, in principle, be brought about through chemical feedback from postsynaptic sites. On the other hand, there is also substantial
_ _ _ _ _ _ ~ __ _ _ _
~-
Correspondence and reprint requests to T. C Fostei, Ari7ona Research Laboratories, Division of Neural Systems, Memory and Aging, 384 Life Sciences North Building, Tucson, AZ U S A 85724
evidence that would favor a postsynaptic locus of expression as well (e.g.. Kauer et al.. 1988; Mueller and G . Lynch, 1988; Davies et al., 1989). Thus, the matter remains a major open question in the field. Since this manuscript was accepted for publication. a report has appeared in which evidence is presented for an increase in mean quantal content of the EPSP (Malinow and Tsien, 1990). Because the latter finding is at variance with our own conclusions, consideration of the possible sources of discrepancy is deferred until the discussion section. Unlike the relatively short-term phenomenon of potentiation seen at both neuromuscular (Magleby and Zengel, 1975) and perforant path (McNaughton, 1982) synapses, L T E is not due to an increase in the quantal release probability ( p )at the presynaptic terminal’. However. the possibility was left open for other presynaptic mechanisms of expression, such as an increased number of release sites ( n ) ,o r an increase in the amount of transmitter released per quantum. The experimental results described in this report represent a further restriction of these presynaptic possibilities by demonstrating
’
T h k is one of the major reamns for our persistence in the use of the term “enhancement” in referring to this phenomenon.
79
80
HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991
that. under our experimental conditions, LTE in Schaffer collateral synapses in CAI is due to an increased quantal size ( 4 ) .We leave open the question of whether this increase is pre- or postsynaptically generated. Quantal analysis of synaptic transmission, first proposed by del Castillo and Katz (1954) as a statistical description of factors contributing to variance in synaptic potentials. represents a powerful physiological method for clarifying the basic nature and location of changes in synaptic efficacy. Although applications of the technique to hippocampal circuits have been hampered by the problem of resolution of the small synaptic signals arising from individual afferent fibers (Voronin. 1988: Hess et al., 1987: Sayer ct al.. 1989, 1990: Fricdlander et al., 19901, a substantial improvement has recently become available through the application of the method of noise deconvolution (Edwards et al.. 1976: Korn et al.. 1982: Wojtowicz and Atwood, 1986: Sayer et al., 1989, 1990). This is a statistical procedure that takes account of variation in the distribution of quantal events caused by background recording noise, in order to estimate the parameters of the probability distribution underlying the fluctuation of evoked EPSPs. In its least constrained form (Edwards et al.. 1976. Sayer et al.. 1989. 1990) no assumptions are made about the underlying statistical distribution other than that the distribution is discrete, and that the variation in response size is independent of the background noise. Because of this lack of constraint on the statistical model, relatively high signalto-noise ratios and/or large sample sizes are required in order to obtain statistically reliable parameter estimates. However, in the few cases in which reliable estimates have been obtained for Schaffer collateral synapses ( e g . Sayer et al., 1989, 1990). these have not differed substantially from the predictions of the binomial model. This provides N prior-ijustification for making the binomial assumptions that all release sites from a single axon have uniform probabilities of release, and generate equivalent-sized quantal events in which the intrinsic variance of quantal sizes is negligible. The advantage of assuming a binomial process (or its Poisson approximation) is that reasonably reliable parameter estimates can be obtained with smaller samples and/or a lower signal-lo-noise ratio. Moreover, even if the estimates for individual cases are not highly reliable, the means from a number of experiments should nevertheless provide a valid estimate of the population mean. which can be used to assess how these parameters change on average under different experimental conditions. In the present report, we compare how the estimated binomial (or Poisson) parameters change during L T E and during the short-term phenomenon of paired-pulse facilitation. for which there is good evidence for an increased quantal release at hippocampal and most other synapses. The general conclusions derived from the binomial deconvolution method are verified by 2 other methods: the method of failures, and the coefficient of variation method. Although both of the latter methods rely upon the Poisson approximation to the binomial distribution for the accuracy of their parameter estimates, a significant change in EPSP size, without significant changes in m estimated by these methods, would constitute relatively assumption-free evidence for increased quantal size. In principle, the most convincing studies of synaptic change at individual connections should involve simultane-
ous recording from pairs of connected neurons (Kuno, 1971), for example, one pyramidal cell in CA3 and another in C A I . Indeed. although the success rate for obtaining synaptically connected dual CA3-CAI impalements is very low (Friedlander et al.. 1990: Sayer et al., 1989. 1990). quant'd 1 events at this synapse have been characterized in this manner. However, the low probability of such successful dual impalements, in addition to an unexpectedly low success rate of inducing LTE (Friedlander et al., 1990) at these connections (a problem we shall discuss further below). has made the use of this method for analyzing the mechanism of LTE expreysion impractical. On the other hand. previous studies indicate that minimal stimulation can be used successfully to activate unitary synaptic responses. This method has been used to examine differences in quantal parameters due to the location of synaptic contacts, as well as short-term changes in quantal parameters during repetitive stirnulation (Hess et al.. 1987; Sayer et al., 1989). For example. McNaughton et al.. (19x1) showed that synaptic depression in perforant path synapses during repetitive activity was accounted for by a decrease in tn with little o r no decline in 4 . They also concluded from the step-like increase in EPSP amplitude near threshold, that excitation failure in the presynaptic axon contributed relatively little to response variation. Increased confidence in the method of minimal stimulation would, however, be provided by evidence that this method yields estimates of unitary response size and general quantal parameters that agree closely with the dual impalement method. This issue is taken up in the present studies. The minimal-stimulation method has the advantage that the timing of presynaptic activity can be better controlled. We make use of this advantage to obtain additional verification of the statistical procedures, by showing that they yield the expected increase in the /zp product during paired-pulse facilitation, a process already known to involvc an increase in transmitter release at most synapses (del Castillo and Katz, 1954: Martin and Pilar. 1964: Wernig, 1972: Zucker, 1973; McLachlan, 1975: Hirst et al.. 1981: Smith, 1983; Hess et al.. 1987).
MATERIALS A N D METHODS Brains were removed from adult rats (300-400 g) under ether anesthesia, and the hippocampi were dissected out. Slices (450-500 p,m) were cut parallel to the alvear fibers using a tissue chopper, and transferred to a standard recording chamber (Haas, et al., 1979). During the experiment. slices were perfused ( 1 mlimin) at 30-32°C with oxygenated artificial cerebrospinal fluid containing (in mM): NaCl 124. KCI 2. K H 2 P 0 4 1.25, MgS04 2. CaCll 2. NaHCO, 26, glucose 10. Humidified air (95% Oz. 5% COz) was continuously blown over the slices. Intracellular recordings were obtained from CA 1 pyramidal cells located in the CAI stratum pyramidale (approximately 600 p m from the CA3ICAI border) with glass micropipettes (20-90 M a ) pulled from 1.0 mm O.D., thin-walled tubing, and filled with 3 M K' acetate. Resting membrane potential and input resistance were monitored and recorded throughout the experiment. C A I cells were considered suitable if they exhibited a resting membrane potential of - 60 mV o r below (mean, -73.9 mV ? 10.2 SD). a spike of at least 50 mV from onset to peak (excluding the EPSP) (mean, 65.0 mV +- 1 I .4
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster and McNaughton
INTRACELLULAR
-ij
CAI INTRACELLULAR
\I R
STIM
CA3 INTRA OR XTRACELLULAR
CA3-CA1
Minimal Stimulation
=-
J Fig. 1 . Top: Stimulating and recording electrode configurations for CA3-CA 1 spike-triggered EPSPs (Left) and Schaffer collateral minimal-stimulation EPSPs (Right). Bottom: Examples of intracellularly recorded evoked EPSPs (a and b), apparent failures (c) and the EPSP averages (d) for single CA3 spikes (Left) and for minimal-stimulation (Right). The average of the CA3 unit spike, which precedes the apparent onset of the EPSP by about 3 msec, is shown in (el. Calibrations. 2 msec and 250 kV.
SD), and an input resistance greater than 16 MR (mean, 26.8 MR & 8.2 SD). Additional constraints for quantal analysis included the establishment of steady state release conditions prior to tetanic stimulation, such that unitary EPSPs averaged over 50 trial blocks for 150 or more trials did not show a systematic change in response tendency or fluctuations in mean amplitude of more than about 20%. Fig. 1 illustrates the 2 methods used to obtain intracellular CAI EPSPs from stimulation of presumed single CA3 afferents. For studies involving simultaneous recording from CA3 and CAI cell-pairs, single CA3 cells were recorded intracellularly as described above, or extracellularly with glass pipettes (2-10 MR) filled with perfusion medium and glutamate (10 mM). In these CA3 cells, action potentials were elicited by injection of depolarizing current o r iontophoretic ejection of glutamate, respectively. In both cases the currents were adjusted in order to elicit less than one spike per second. Online spike-triggered averages of CAI responses were used for initial identification of CA3-CAI connections. For minimalstimulation experiments, diphasic constant current stimulation (100 psec, 5-50 FA) was delivered via twisted-pair, plat-
81
inum-iridium wire stimulating electrodes located in stratum radiatum. Stimulus intensity was adjusted with a high resolution potentiometer. Minimal-stimulation consisted of paired pulses of identical intensity, separated by 50 msec, with one such pulse pair delivered every 5 seconds. Once a CAI cell met the criteria for recording, stimulation intensity was adjusted to evoke the smallest consistent averaged response for conditioning and test stimulation (McNaughton et al., 1981). At these stimulus levels, failures of transmission could be frequently observed in response to the first stimulus at to, and less frequently in the second response at tSo. stimulation was continued for 10I5 minutes in order to insure stabilization of the responses. One hundred fifty to 250 response pairs were then recorded and averaged in real time over 50 trial blocks to insure response stability. If the average amplitude tended to increase or decrease, another 150 trials were collected until stability was achieved. After collection of control data, the stimulus intensity was adjusted to elicit a population response of 2-8 mV, and 5-10 records were taken. Stimulus intensity was then raised further, to above the threshold to elicit an action potential, and high-frequency stimulation (4 bursts of 8 pulses at 400 Hz) was delivered in an attempt to induce L T E by the cooperativity method (McNaughton et al., 1978). Stimulus parameters were then returned to pre-tetanus values and minimal-stimulation responses were collected during the next 2060 minutes. Enhancement of the population EPSP was examined after tetanus and again at 20 and 40 minutes later, using the same stimulus level used to collect the population response baseline. Stimulus timing and data collection were under computer control. Electrophysiological signals were stored on FM tape for off-line analysis or stored directly on computer as 50 msec records of evoked responses o r 50 msec records of background noise only (i.e., baseline fluctuations in membrane potential and electronic noise in the absence of a CA3 spike or electrical stimulation). Individual records for each evoked response consisted of a 2 msec baseline period and the first 48 msec after stimulation. Noise records were collected just prior to each response trial for minimal-stimulation and between CA3 discharges for spike-triggered responses. Two time-windows ( I -2 ms), corresponding to the baseline period before the stimulus artifact and the EPSP peak, were defined for the average response. These window regions were then used to measure unitary EPSP amplitudes for individual records by subtracting the mean voltage within the baseline window from the mean within the peak window. This process was also applied to the noise-only records, using the same window parameters, to obtain a measure of the expected noise contribution to the EPSP measures. In order to analyze fluctuations in evoked responses, separate EPSPand noise-amplitude histograms were constructed. These were divided into 30 bins over the range of observed responses (bin width for EPSP histograms range, 19-85 kV). A chi-square test was used to insure that the noise histogram conformed to a Gaussian distribution at P < .05. Two independent methods were used t o extract the quantal parameters from the EPSP histograms. The parameters for a Poisson distribution ( m , 4 ) o r binomial distribution ( n . p , q ) were determined by a computer algorithm that found the pa-
82
HIPP0CAMPU)S VOL. 1, NO. 1 , JANUARY 1991
rameters that best fit the observed response distribution. The expected proportion of responses for each of the 30 bins was calculated by the binomial or Poisson formula, using sums of Gaussian curves whose variances were obtained from the noise distribution, whose means represented multiples of the single quantal size, and whose integrals were equal to the expected proportion of multiple release events. The expected distribution was then compared to the observed distribution using a chi-square test. Parameter values were altered according to a nonlinear optimization routine (Nclder and Mead. 1965; Press et al., 1988) until the lowest chi-square value. and thus best estimate for each parameter, was obtained. The algorithm was constrained to positive values for all variables. Mean quantal content was also calculated by the method of failures (del Castillo and Katz. 1954) according to the equation:
a
I
4
1000
5
d
L
800
600 400 200 0
2
4
6
8
b where N is the total number of samples and tio i s the number of times an evoked response failed to appear. Because the noise was found to be normally distributed. the number of failures was estimated as two times the number of responses less than Lero plus those that were equal to zero (Nicholls and Wallace. 1978). Quanta1 size was then calculated as the mean EPSP amplitude divided by m,-. The failures method was not used if the number of failures was less than 6% or greater than 95% of the total number of samples (Martin. 1977). This method was used exclusively tor one CA3-CAI cell-pair in which only 80 responces were obtained. Because of the closc correspondence of quantal parameters estimated by the 2 methods (see results) and the inability to make quantal estimates for some cells resulting from a ION failure rate, all statistical analyses were performed on estimates calculated by the optimization procedure except where specified. Statistical significance was assessed using Student {-tests with a = 0.05. Finally, for comparison with the recent results of Malinow and Tcien (1990). possible changes in EPSP mean quantal content were assesscd using the coefficient of variation method according to 1.he lormula w,,
=
M'lcr'.
where M is the mean EPSP amplitude and v2 is the variance of the EPSP amplitude after subtraction of the background noise variance.
RESULTS Characterization of CA3-CA1 spike-triggered synaptic responses CA3 spike-triggered responses were observed in 10 out of 176 cell-pairs tested. The mean time from CA3 spike peak to apparent EPSP onset wits 5.19 +- 1.95 (SD) msec. Synaptic responses to single CA3 spikes consisted of clear small EPSPs interspersed with apparent failures (Fig. 2). The mean baseline amplitude for all cell pairs varied around zero (mean. - 8.5 & 9.1 SD FV), with an overall mean for the noise stan-
10
12 14
16
18
TIME (ms)
1
30 1
rz
25
0
15
3
u
!
q SIZE
= 233 U V
m = 1.2 BINSlZE=47uV
2o
10
5 0
0 mV
0.5
1.o
Fig. 2. ( a ) Illustration of firilui-cs and mono5ynaptic EPSPs I-ecordcd intracellularly in one CAI cell i n response to 50 single CA3 spikcs whose time of occurrence is indicated by the arrow. The quantal amplitude estimated by the method of failures was 350 p V . (b) Example of an EPSP amplitude frequency distribution (BARS) for 395 responses from ;inother CA3-CAI cell-pair and the best-fitting Poisson distribution ( L I N E ) estimatcd by the optimization algorithm. Arrows indicate multiples of estimatcd quantal amplitude.
dard deviation of 155.6 2 92.4 (SD) k V . Estimated nz values calculated by the method of failures and by the coefficient of variation method were highly correlated with the results fi-om the optimization procedure foi- 9 cells with at least I50 responses (mf: r2 = 0.85, P < .OOO5: 1 7 1 ~ r'~ : = 0.88, P < .OOO2). Descriptive statistics indicated that response histograms for 7 of the 9 cell-pairs could be adequately described by either Poisson o r binomial statistics at the 0.01 level of significance. The contribution of quantal parameters to synaptic strength for the population of cell-pairs was tested by plotting 172 and q against mean EPSP amplitude. A simple regression indicated that both q and m were significantly correlated with EPSP amplitude. Examination of several factors that might have contributed to the variability of q and rn revealed no correlations with input resistance. resting membrane potential, onset latency, 10-90% rise-time o r EPSP half-width. Also, 4 and 171 were not correlated. For 4 cell-pairs the CA3 cell was activated by antidromic stimulation of the Schaffer collaterals. However, relatively high stimulation currents, which produced large EPSPs and
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster and McNaughton
Table 1. Quantal Parameter Estimates for 3 CA3-CA1 Connections for which Sufficient Data Were Obtained to Estimate Effects of LTE Inducing Stimulation ___ EPSP (pV)
qq,t (I*V)
83
~~
m,,,,, ~~~
Cell
Pre
Post
Pre
Post
Pre
I 2 3 ~~
Post ~
~~~
-
72 2 126 S 389 9
69 1 181 1 130 8 __
141 210 204
IS9
252 225
051 060 I91
0 42 075 0.67
*
~
spikes in the CAI cell, were necessary for antidromic activation, indicating that placement of the stimulating electrode was relatively distant from the fiber of interest. In one case, the stimulating electrode was moved to 2 other locations along the vertical axis of the stratum radiatum with little success in reducing the current necessary for antidromic driving. Once a cell could be activated by antidromic stimulation. the intensity was further increased to elicit a CAI cell spike, and tetanizing stimulation was delivered. Tetanization of the Schaffer collaterals resulted in enhancement of the population test response measured 15 minutes later. Only 3 cell-pairs were maintained such that at least 150 CA3 spikes occurred during the 5-30 minutes following tetanization, and hence could be subjected to quantal analysis. Based on the considerable fluctuations that occur in rn in the minimal-stimulation experiments (see below), and on the uncertain reliability of the parameter estimates from individual experiments, we d o not consider 3 cells a sufficient population from which to draw any conclusions concerning the question of how quantal parameters change following LTE induction. We believe that this question can be answered only from the mean of a reasonably large number of experiments. Nevertheless, the data from these 3 cells are presented in Table 1, for the sake of completeness. In some cases, facilitation and summation of CA3 spikeelicited EPSPs were observed during a burst of CA3 cell spikes. Figure 3 shows examples of bursts in which failures were observed for the initial spike, and facilitation and/or summation was observed for later spikes.
*
n , b
C
Quantal analysis using minimal-stimulation: Similarity to CA3-CAI spike-triggered responses Paired-pulse stimulation of the Schaffer collaterals was used to elicit minimally evoked conditioning (to) and test (tso) EPSPs separated by 50 msec. Background recording noise in these experiments was comparable to the dual cell impalement experiments (mean, 172 p V ? 59 SD). Sufficient failures were observed for application of the method of failures in all but 1 of 33 cells at to and all but 5 cells at t5(). Estimates of rn, calculated by the method of failures and by the optimization procedure for to and tS0 responses, were highly consistent (rnf vs. rnopt: r2 = 0.74, P < ,0001, slope = I .04). As expected (see discussion) rncv was less well correlated with mopt, and increasingly overestimated rn at higher values (mCv vs. mopt:r2 = 0.3 P < .0001, slope = 2.08). These relationships are illustrated in Figure 4. One response histogram for to and 1 for tSocould not be described by Poisson or binomial statistics at the 0.01 level of significance. For to the observed
Fig. 3 . Postsynaptic CAI responses to bursts of CA3 unit activity. (a) and (b) illustrate pairs of spikes in a single CA3 cell in which failure of transmitter release at the CAI synapse can be seen (*). Another CA3-CAI cell-pair in which failure of release was observed for the first spike in a burst is shown in (c). The CAI responses in (b) and (c) are composed of multiple quanta. Calibration, 4 msec, CA3: 200 p V , CAI: 500 PV. response distributions of 20 of the 33 cells were best fit by a Poisson function. On the other hand, there was a tendency for t5() amplitude histograms to be better fit by a binomial distribution (26 out of 33) and to exhibit fewer failures. This is to be expected when p is increased, as is known to occur during synaptic facilitation. The mean and range of to minimal-stimulation quantal pa-
84 HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991 12 -
9-
m, 6-
0
2
1
3
opt
l2
9
6 mcv
1
1
0 0 0 0
1
nificant increase in the estimated rn for to and t5()with no significant change in q or p. However, only one Vu and two V5(,response histograms could adequately be fit by a binomial distribution, most likely because of differences in q for different afferent fibers. In addition, in 2 cases, increasing the EPSP above 300% resulted in a small population EPSP recorded extracellularly upon withdrawal from the cell, which may also have contaminated these results. Previous studies have indicated that EPSP rise-time and q are influenced by synaptic location (Andersen et al., 1980; ’Turner, 1988; Sayer et al., 1989). The mean overall 10-90% rise-time was 4.46 ? 1.61 (SD) msec (N = 33). The rise-time was reduced for responses elicited by stimulating electrodes localized to the inner one-third of the stratum radiatum (3.56 & 1.16 msec, N = 8), and increased when the stimulating 4 electrodes were localized to the outer one-third of the stratum radiatum (5.62 f 1.23 msec, N = 6). While mean 4 was slightly larger for the proximal vs. distal electrode locations, statistical analysis did not indicate a significant difference. Neither mean EPSP, quantal size, nor quantal content were correlated with input resistance o r rise-time. EPSP amplitudes were somewhat correlated with resting potential (r’ = 0.214, P < .046). However, the correlation of resting potential with quantal size was not significant. The main result of this part of the experiment was the rather close correspondence of quantal parameters obtained with the dual impalement and minimal-stimulation techniques as summarized in Table 2 . columns I and 2. Quantal analysis of facilitation at the CA3-CA1 synapse
Y
0
1
2
3
4
opt Fig. 4. ‘lop: Correlation between quantal contents for pairedpulse minimal-stimulation of the Schaffer collateral.; calculated by the method of failures (mr) and by the binomial deconvolution method using nonlinear parameter optimization (mupr).Data are plotted for the to and tso responses obtained prior to LTE induction. The correlation (r2 = 0.74) was significant at P < .0001. There were no significant differences between methods in the estimates of either rn or q. Bottom: Same as top but data. for the coefficient of variation method (mcv). This method significantly overestimated m,,, and underestimated q. Moreover, the correlation with mOprwas poorer (r’ = 0.3). The increased variance at higher WI is expected theoretically because of the nonlinear effect of increased p (due partly to facilitation at t5(J on mC.. Solid lines in both figures indicate one to one correspondence, not regression.
rameter values were very similar to those observed for the CA3-CA I cell-pairs (Table 2). Like CA3-CA I spike-triggered responses, minimal-stimulation EPSP means were correlated with estimates of both y (r’ = 0.19, P < .Of I ) and i n (r’ = 0.3 I , P < .0007). Increasing the mean EPSP amplitude (150-4S0%5) by increasing the stimulation intensity (N = 7) resulted in a sig-
The mean overall V,,, EPSP amplitude was increased 67 f 34.8 (SD)%’relative to VO(Fig. 5 ) . Table 2, columns 2 and 3, illustrate the quantal parameter means and ranges for conditioning and test stimuli. During paired-pulse facilitation, all 3 estimates of M were significantly increased. Quantal size estimated by optimimtion and by the failures method was not changed. A significant decrerise occurred in the quantal size estimated by the coefficient of variation method. This is to be expected because of the strong nonlinear tendency of this method to overestimate m as p increases. The magnitude of facilitation was negatively correlated with mo (r’ = 0.32, P < ,0006). suggesting that facilitation was inversely dependent on the initial release level. For binomial analysis, there was a significant tendency for n to increase in the tso response, but this increase was not correlated with facilitation of the EPSP (r2 = 0.036. P < .29). On the other hand. the expected significant increase in p was observed and was correlated with facilitation (r’ = 0.35: P < .0003). Quantal analysis of LTE for minimal-stimulation of CA3 fibers
For all 33 cells included in the present analysis, tetanic stimulation elicited enhancement of the population response. Data were rejected if this failed to occur. In order to avoid biasing the results, however, data were nor rejected if the population response was enhanced but the minimal-stimulation response failed to show an increase.
LONG-TERM ENHANCEMENT OF CAI SYNAPTIC TRANSMISSION / Foster a n d M c N a u g h t o n Table 2a. S u m m-.a r.y of Q u a n t a 1 Analysis Results for D u a l Recording a n d Minimal-stimulation Experiments* .~ ~~~
EPSP (pV) (FV)
qupl
mopt
nopt Pop1 qfad (CLV) mk,,l
(jcv
(PV)
in,,
..
1
2
3
4
5
6
CA3-CA1
pairs
Minimal Stim Baseline to
Minimal Stim Baseline tso .~
Minimal Stim LTE 20 min to
Minimal Stim LTE 20 min tso _.
Minimal Stim LTE 40 min to
264 (60) 72-674 220 (18) I 4 1-350 1.18 (0.17) 0.51-2.04 3.8 (0.29) 2-5 0.29 (0.03) 0.17-0.46 227 (33) 136-485 I.09 (0.16) 0.52-2.08 183 (19) 89-296 1.38 (0.22) 0.49-2.58
290 (32) 101-75 1 245 (15) 88-398 1.30 (0.09) 0.43-2.70 5.28 (0.3) 2-9 0.24 (0.01) 0.09-0.43 231 (27) 7 1-460 1.38 (0.11) 0.45-2.57 1.61 (16) 8-390 2.83 (0.51) 0.56-16.31
474 (38) 198-1 181 258 (15) 77-407 1.96 (0.12) 0.68-3.9 6.03 (0.39) 3-1 I 0.34 (0.02) 0. 15-0.61 216 (17) 7 1-460 2.04 (0.12) 0.69-2.92 126 (14) 26-380 4.66 (0.39) 0.73-10.25
371 (33) 8 1- I004 301 (20) 98-553 I.36 (0.09) 0.36-2.70 5.47 (0.46) 2-12 0.26 (0.02) 0. I 1-0.60 265 (17) 99-544 1.40 (0.11) 0.29-2.93 230 (38) 27- 126 I 2.62 (0.45) 0.13-14.65
580 (62) 136- I833 298 (18) 122-512 1.99 (0.14) 0.75-4.65 4.49 (0.22) 2-8 0.42 (0.02) 0.23-0.6 I 2.48 (18) I 15-447 2.01 (0.13) 0.75-2.94 208 (53) 27-179 I 5.07 (0.66) 0.10-17.27 .
386 (37) 67-760 323 (23) 98-560 1.25 (0.10) 0.32-2.92 5.28 (0.47) 2-1 I 0.26 (0.03) 0.09-0.66 301 (27) 88-583 I .37 (0.15) 0.25-2.94 255 (44) 45-1 I12 2.31 (0.33) 0.38-6.25..
-..
~
~~
~
~~
~~~
~
85
7 Minimal Stim LTE 40 min tsi, -~
586 (67) 135-1529 326 (23) 120-580 1.86 (0.14) 0.68-3.68 4.62 (0.21) 2-7 0.38 (0.10) 0.2 1-0.60 307 (37) 125-565 1.79 (0.15) 0.96-2.96 211 (35) 11-813 3.95 (0.60) 1.14- 13.10
* Using the binomial nonlinear optimization technique (opt). the failures method (fail) and the coefficient of variation method (cv). The following columnwise statistical comparisons were performed using t-tests with 01 set at 0.05: 1-2, 2-3, 2-4. 2-6. 3-5. 3-7, 4-5, 4-6. 5-7. 6-7. These tests addressed the general questions of how quantal parameters compared between C A 3 - C A I cell-pair recording and minimal stimulation; how quantal parameters changed at 2 different times after LTE induction in the population response: how quantal parameters changed during paired pulse facilitation at a 50 nisec interstimulus interval; and whether quantal parameters changed over time between the 2 post LTE epochs. The statistical summary is given in Table 2b.
Table 2b. C o l u m n Comparison ~~
1 vs L ~~~
~~~
ycvIkV)
0.80 (43) 0.92 142) 0.19 (42) 0.50 (43)
nicv
0.13 143)
q,.,,,(+v) mt.,,)
~.
...
~~~
~
n.0001 133) 0.09 133) 0.0001 133) 0.0006 (331 0.0001 (33) 0.45 12x) 0.0001 I281 0.007 f33) 0.003 133) ...
~~
2 vs. 4
~~~~~
EPSP I ~ V ) 0.62 (43) q,,,, ( + V ) 0.40 (43) 0.26 143) n7p, nt,pt 0.24 143) POP1
~
2v5 3
~
~
0.0015 133) 0.0001 (33) 0.40 133) 1.00 133) 0.36 ( 3 3 ) 0.0031 132) 0.76 132) 0.042 (331 0.93 ( 3 3 )
L vs. 6
3 vs.
~
5
~
0.025 1261 0.0001 126) 0.62 126) 0.33 (26) 0.71 126) 0.001 124)
0.76 (24) 0.074 (261 0.94 (26)
0.0047 133) 0.1)001 ( 3 3 ) 0.63 133) 0.23 0.15 0.029 0.065 0.085 0.54
.~
(3.11 1331 122) (22) (33) 133)
..
~~~~~~
~
3 vs. 7
0.093 0.0002 0.47 0.40 0.33 o.ooah 0.048 0.044 0.20
_
126)
126) 126) (26) 126) 118) 118)
(26) (26)
4
VS.
5
4
v5.
6
_ ~ . o.oo01(331 0.46 126) 0.75 133) 0.0001 133) 0.24 133) 0.0001 133) 0.54 (24) 0.0001 124) 0.29 133) 0.0001 1331 .
0.61 126) 0.16 (26) 0.53 126) 0.54 126) 0.25 124) 0.30 (24) 0.99 (26) 0.22 (26)
. .
-.
5 vs. 7
6 vs. 7
0.21 126) 0.27 (26) 0.20 126)
0.0001 1261
.
0.81 I261 0.0001 (26) 0.013 (26)
0.86 126) 0.082 126)
0.001 126)
0.26 117) 0.022 117) 0.82 (26) 0.039 126)
0.92 (23) 0.0002 123) 0.021 (261 0.0001 126) ~
Summary of statistical results for columnwise comparisons on data from Table 2, showing (Y levels for student's I-tests. Repeated measures comparisons were employed for all tests except 1 vs. 2. in which the samples were independent. Results that were statistically significant at 01 = 0.05 are underlined. Except for the italicized comparisons, the results support the following conclusions: ( 1 ) minimal-stimulation samples single fibers from the same population a s the dual i C A 3 - C A I ) recording experiments: (2) paired-pulse facilitation leads to an increase in m with no change in q: (3) L T E expression involves an increase in q with no change in m . The discrepant dec,r-euAein q,, during pairedpulse facilitation is expected from the nonlinear relationship between tn,, and true m under changes in p.
The population EPSP amplitude increased 40 k 34 (SD)% immediately following tetanus (i.e., during post-tetanic potentiation), and remained elevated 30 k 43% (N = 33) at 20 minutes post-tetanus and 21 2 47% (N = 26) by 40 minute post-tetanus. The mean EPSP amplitude for minimal-stimulation during the 5-20 minute period was increased 35 t- 57% and 27 2 40% for Vo and Vs0, respectively. During the 2040 minute period Vo and Vso were increased by 27 t- 36% and 21 k 55%, respectively. It is important to note that the overall mean fractional increase in EPSP amplitude for minimal-stimulation elicited EPSPs was similar to that observed for the population response. Figure 6 shows the time course
of these increases relative to the corresponding mean pretetanus amplitudes. While on average, the minimal-stimulation mean EPSP amplitude was increased, not all cells exhibited such an increase. The mean EPSP amplitude decreased for 9 of 33 cells at t,, and 10 of 33 cells at tS0during the 5-20 minute period after tetanization. During the 20-40 minute period, 10 of 26 cells at to and 9 of 26 cells at t50exhibited a decrease in mean EPSP amplitude relative to pre-tetanus. As with the pre-tetanus data, most post-tetanus amplitude histograms at to were best fit by a Poisson function (20 of 33). and only 1 histogram did not conform to a Poisson distribution
86
HIPPOCAMPUS VOL. 1, NO. I , JANUARY 1991
-0.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
mV
-0.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
mV
Fig. 5. Synaptic facilitation evoked by paired-pulse stimulation of the Schaffer collaterals for 1 cell. Top: Mean (solid line) and standard deviation (dotted line) of CAI EPSPs (N = 150). Calibrations, 4 msec, 100 k V . Bottom: Best fitting binomial distributions (line) of the observed response amplitude distributions (bars) for conditioning (Left) and test (Right) stimuli. Bars = 50 bV.
at P < .01. Histograms for t5,, were best fit by a binomial function (21 of 33), with 2 unable to conform to a binomial function and 4 unable to conform to a Poisson function at P < .01. The means and ranges of the alterations in quanta1 parameters resulting from tetanization are summarized in Table 2. columns 2-7. and Figure 7 . The major observation was that, whereas there was a significant increase in 4 at both time points after tetanization, there was no significant change in either m , n , o r p . Moreover, the average fractional increase in 4 was sufficient to account for the mean increases in both the minimal-stimulation EPSP and the population EPSP. Although there was no average increase in rn, nevertheless, there was often substantial variation i n this parameter between the baseline and post-tetanization periods. Indeed, the standard deviation for fractional change in tn was substantially greater than the corresponding measure for change in q (to: 0.51 vs. 0.28: t50:0.51 vs. 0.27). As a result, sometimes a decrease in m was large enough to mask an increase in Vo in spite of an increase in 4 . In other cases, increases in m produced an increase in Vo beyond that caused by the increase in 4 . There was a significant negative correlation (I' = 0.31, P < .001) between the change in m following teta-
nization and the change in paired-pulse facilitation. suggesting that much of the variance in m was due to variance in p (McNaughton, 1980. 1982; Zalutsky and Nicoll, 1990). This was corroborated by a significant positive relation between the change in m and the change in p estimated from binomial optimization (6 = 0. I I . P < .04). and by the lack of significant rclation to the change in binomial n. A possible bias may result from selection of data for inclusion in the analysic on the basis of whether the single fiber EPSP is increased, rather than the population EPSP (Voronin, 1988; Malinow and Tsien, 1990). 'To illustrate this possibility, we present the following reanalysis of our data, based on only those fibers that exhibited a V,, increase greater than 25% during the 5-20 minutes after tetanus. Mean EPSP values increased approximately 2 times that observed for the population of afferents (83 2 1 I%, N = IS), and the number of included single-fiber experiments was reduced by approximately one-half. For these experiments, both 4 and m exhibited significant increases "4": P < ,0006;mo:P
