Exposure to novel environments or behavioral training is associated with increased strength at hippocampal synapses. The present study employed quantal analysis techniques to examine the mechanism supporting changes in synaptic transmission that occur following differential behavioral experience. Measures of CA1 synaptic strength were obtained from hippocampal slices of rats exposed to novel environments or maintained in individual cages. The input/output (I/O) curve of extracellularly recorded population excitatory postsynaptic potentials (EPSPs) increased for animals exposed to enrichment. The amplitude of the synaptic response of the field potential was related to the fiber potential amplitude and the paired-pulse ratio, however, these measures were not altered by differential experience. Estimates of biophysical parameters of transmission were determined for intracellularly recorded unitary responses of CA1 pyramidal cells. Enrichment was associated with an increase in the mean unitary synaptic response, an increase in quantal size, and a trend for decreased input resistance and reduction in the stimulation threshold to elicit a unitary response. Paired-pulse facilitation, the percent of response failures, coefficient of variance, and estimates of quantal content were not altered by experience but correlated well with the mean unitary response amplitude. The results suggest that baseline synaptic strength is determined, to a large extent, by presynaptic release mechanisms. However, increased synaptic transmission following environmental enrichment is likely due to an increase in the number or efficacy of receptors at some synapses and the emergence of functional synaptic contacts between previously unconnected CA3 and CA1 cells.
It is believed that cognitive ability is a function of the dynamics of synaptic connectivity between neurons, and memory operations may depend on adjustments in synaptic transmission properties due to the history of synaptic activity (Foster 1999). Notable advances have been made in identifying and characterizing two forms of activity-dependent synaptic plasticity, long-term potentiation and long-term depression (LTP and LTD), which are manifest following patterned activation of neural circuits (Bliss and Collingridge 1993). However, information on endogenous mechanisms for naturally occurring changes in synaptic strength with experience is still lacking.
Differential experience (e.g., exposure to enriched or novel environments) influences hippocampal anatomy, physiology, and biochemistry. Relative to “normal” laboratory animals, an increase in hippocampal synaptic efficacy is observed in vitro following exposure to novel environments or behavioral conditioning (Foster et al. 1996; Green and Greenough 1986; Power et al. 1997). The observation of increased synaptic transmission recorded in vitro indicates that synaptic changes are long-lasting and expression is not due to extrahippocamal systems involved in arousal. Thus experience-dependent growth in synaptic strength is preserved in the hippocampal slice providing an opportunity to examine the mechanisms for expression of a naturally occurring change in synaptic function.
Statistical analysis of the fluctuations in unitary responses can be used to provide estimates of biophysical parameters of transmission. These techniques have been used to explore the mechanisms for endogenous changes in CA3-CA1 synaptic function across the life span (Barnes et al. 1992; Dumas and Foster 1995; Hsia et al. 1998). Therefore as a starting point for examination of potential mechanisms for expression of the experience-dependent increase in synaptic transmission, we examined unitary responses recorded from pyramidal cells in region CA1 of control animals and animals exposed to environmental enrichment.
Sprague-Dawley rats were bred and housed in our vivarium and maintained on a reversed 12:12 h light/dark cycle. On postnatal day 1 (P1), litters were restricted to a maximum of 10 pups with no less than 2 female pups per litter. Males were separated from females and weaned on P21. At P42, rats were randomly assigned to individual cages (IC) or enriched conditions (EC). Animals in the IC group were individually housed in wire cages (25 × 18 × 21 cm) and only handled for routine upkeep. Each day, EC rats (2–3 per cage) were taken from the home cage (25 × 18 × 66 cm) and placed for 1–6 h in a novel environment (e.g., empty water maze, large wooden box, or large wire cage) that contained three-dimensional objects (e.g., coffee cans, children's plastic toys, cardboard boxes, large plastic drain pipes), rat chow, and a water bottle. The exposure to each environment and three-dimensional objects was nonsystematically randomized across days. Following 25–32 days of differential experience (i.e., P67–74), animals were deeply anesthetized, their hippocampi were harvested, and hippocampal slices (450–500 μm) were cut parallel to the alvear fibers. Slices were then transferred to a standard recording chamber, and perfused at 32°C with oxygenated artificial cerebrospinal fluid (ACSF) containing the following (in mM): 124 NaCl, 2 KCl, 1.25 KH2PO4, 2 MgSO4, 2 CaCl2, 26 NaHCO3, and 10 glucose. Humidified air (95% O2-5% CO2) was continuously blown over the slices.
Extracellular field potentials were recorded with ACSF-filled glass micropipettes (4–6 MΩ). The recording and stimulating electrodes were separated by ∼1 mm and localized to the middle of the s. radiatum. The signals were amplified, filtered (1 Hz to 1 kHz), and stored on computer disk for off-line analysis. The viability of the slices was examined, and only those slices exhibiting a field potential synaptic response of at least 1.5 mV for 200 μA stimulation were employed for further analysis. Biphasic constant current stimuli (100 μs) were delivered such that paired stimulation pulses (50-ms interstimulus interval) occurred once every 30 s. The paired-pulse ratio of the field potential synaptic responses was calculated as the quotient of the response to the test pulse divided by the response to the conditioning pulse. Input/output (I/O) curves of the field potentials were constructed using the means of five conditioning stimulus pulses at three stimulation intensities (100, 200, and 300 μA). In addition, I/O curves were constructed across seven levels based on the fiber potential amplitude. The seven levels were chosen to ensure that ∼15 of the total responses were included at each level. For induction of LTP, the stimulation intensity was set to elicit excitatory postsynaptic potential (EPSP) of ∼1 mV. Following 20–30 min of stable baseline recording (0.033 Hz), LTP was induced (2 1-s bursts of 100 Hz, each burst separated by 10 s), and recording continued for another 30 min.
Intracellular recording and quantal analysis
Methods for collection of unitary responses and quantal analysis have previously been published in detail (Dumas and Foster 1995; Foster and McNaughton 1991). Briefly, intracellular responses were obtained using glass micropipettes (50–90 MΩ) filled with 3 M K+-acetate. Only cells that exhibited a resting membrane potential of −60 mV or below, a spike of at least 60 mV greater than the excitatory synaptic potential, and an input resistance >20 MΩ were considered suitable. The membrane potential was held at or near −72 mV (using less than ±0.2 nA). Synaptic responses were elicited by biphasic minimal-stimulation (100 μs, 5–80 μA) and consisted of paired pulses of identical intensity, separated by 50 ms, with one such pulse pair delivered every 5 s. Once a cell had met the criteria for recording, the stimulation intensity was lowered such that the smallest consistent averaged response was obtained. At this intensity, transmission failures were apparent and interspersed among responses that appeared to fluctuate in discreet steps (see Fig. 5). Responses to the conditioning stimulation were averaged on-line to ensure response stability. If the average response amplitude tended to increase or decrease over blocks of 50 trials, recording was continued until a stable response was obtained for at least 150 trials (294 ± 12 mean number of trials, mean ± SE). Stimulus timing and data collection were computer controlled, and data were stored on computer as 50-ms records (2 ms of baseline and 48 ms after stimulation). Background system noise records were collected just prior to conditioning stimulation for each trial. Response amplitudes were calculated as the difference between the membrane potential averaged with in a 1- to 2-ms window just prior to stimulus onset and a similar window centered over the peak of the evoked response. This procedure of obtaining the average potential difference between two discreet time windows was then performed on the noise-only records to estimate the noise contribution to the EPSP measurements.
Several independent methods were employed to examine the biophysical mechanisms of synaptic transmission. First, the number of transmission failures (n 0) was estimated as two times the number of responses less than zero plus those that were equal to zero. The number of failures was used to estimate the percent of transmission failures (n 0/total number of responses). Six cells were not included in the analysis due to transmission failures beyond the range of >5 or <0.95% of the total number of responses for the conditioning stimulation. Our previous experience indicates that estimates of biophysical parameters become unreliable for responses beyond this range (Foster and McNaughton 1991). Second, the coefficient of variation (CV) was calculated according to the equation: CV = ς/M, whereM is the mean EPSP amplitude, and ς is the standard deviation of the response minus the standard deviation of the noise-only records. Third, to estimate quantal parameters, noise deconvolution and parameter optimization analyses were applied under assumptions of Poisson (m, q) release (Dumas and Foster 1995; Foster and McNaughton 1991). Response amplitudes were divided into 30 bins, and the parameters of best fit were determined by comparing the observed distribution with a distribution generated for Poisson release parameters and convolved with the noise distribution. The parameter values were altered according to a nonlinear optimization routine. The algorithm was constrained to positive values for all variables and the best fit determined by a χ2 test. As a final method, the data were subjected to a computer optimization algorithm that compares the entropy (i.e., smoothness or flatness) and detection of peaks in the response distribution (i.e., maximum entropy noise deconvolution, MEND) (Kullmann 1992). The MEND analysis enables the detection of peaks and makes no assumption concerning the underlying transmission mechanisms. In this case, when peaks are detected, the distance between peaks provides an estimate of quantal size.
To establish that transmission at CA3-CA1 synapses was altered by the environmental enrichment procedure, I/O curves were constructed using three stimulus levels. In cases where recordings were taken from more than one slice from the same animal, the responses within each level were averaged. Figure 1 illustrates that the fiber potential amplitude (Fig. 1 A) and paired-pulse ratio (Fig. 1 B) were not different between EC (n = 15) and IC animals (n = 15) across the three stimulus intensities of the I/O curve. In contrast the field potential synaptic response was increased across the I/O curve [F(1, 56) = 4.2, P < 0.05]. Figure1 D shows the EPSP slope plotted against fiber potential amplitude across the 50 individual slices (EC, n = 23; IC, n = 27). In this case, the responses at each level of the I/O curve were assigned to one of seven groups according to the fiber potential amplitude (Fig. 1 C). An ANOVA on the EPSP slope across the seven fiber potential levels confirmed a significant increase in synaptic strength for responses elicited from slices of EC animals [F(1, 137) = 27.73, P < 0.0001].
While difference in synaptic efficacy due to experience was not related to changes in the fiber potential amplitude and paired-pulse ratio, the measures correlated with the overall level of synaptic strength and much of the variability in the response could be accounted for by these factors. The slope of the population synaptic response of the field potential was positively correlated with the fiber potential at each level of the I/O curve (300 μA: R 2 = 0.29, P < 0.005; 200 μA:R 2 = 0.43, P < 0.0001; 100 μA: R 2 = 0.52,P < 0.0001). Further, the paired-pulse ratio was inversely correlated with the field potential synaptic response at 200 μA (R 2 = 0.36, P < 0.0005) and 100 μA (R 2 = 0.27,P < 0.005). In contrast, the treatment condition accounted for <15% of the variability in the population synaptic response at each stimulus level (300 μA:R 2 = 0.12, P = 0.06; 200 μA: R 2 = 0.11, P= 0.08; 100 μA: R 2 = 0.14,P < 0.05).
For some slices, the stimulation intensity was set to elicit a 1-mV field potential synaptic response, and, after collection of the baseline responses, LTP was induced (Fig.2). An examination of the slope of the field potential synaptic response 30 min after LTP induction indicated a significant increase in the synaptic response for slices from EC (n = 12; 164 ± 16%) and IC rats (n = 13; 149 ± 8%) with no difference between groups.
Intracellular recording of unitary responses
Acceptable intracellular recordings were obtained from 32 and 31 cells from EC and IC animals, respectively. Table1 combines the data from the two groups and summarizes the results comparing transmission characteristics for the conditioning and test pulses of paired-pulse stimulation. Paired-pulse stimulation resulted in an increase in the test response [F(1, 62) = 44.62, P < 0.0001]. Facilitation is thought to result from an increase in the probability of transmitter release such that the percent of transmission failures is reduced for the test pulse compared with the conditioning pulse (Fig. 3). As predicted, the percent of transmission failures decreased for the test pulse [F(1, 62) = 41.60, P < 0.0001], consistent with the notion that the increase in synaptic strength is due to an increase in transmitter release. The coefficient of variance was observed to decrease [F(1, 62) = 35.06, P < 0.0001], which again is consistent with an increase in presynaptic function (Korn and Faber 1991). Finally, nonlinear parameter optimization using noise deconvolution indicated an increase in quantal content [F(1, 62) = 29.49,P < 0.0001] for the test response, in the absence of a change in quantal size (Table 1). Together, the results indicate that the techniques can detect a change in synaptic strength due to an increase in transmitter release.
Table 2 shows the means for intracellular recording parameters and waveform characteristics of the averaged unitary responses for the two treatment groups. Comparison of the mean EPSP amplitude for unitary responses across the two experience conditions indicated a significant effect of experience [F(1, 61) = 3.98, P < 0.05], with the EC group exhibiting an increase in the mean unitary response (Fig.4 A). Due to the fact that the EPSP amplitude was increased for the EC group, comparison of biophysical parameters was performed using one-tailed Student'st-tests. No difference was observed for any of the waveform characteristics of the averaged unitary synaptic responses (10–90% rise time, half-width, time-to-peak), suggesting that synaptic responses were generated at similar locations along the dendrite. In addition, no difference was observed for the resting membrane potential and action potential amplitude. The paired-pulse ratio was not different between groups, suggesting that the increase in synaptic strength for the EC group was not due to a change in the probability of transmitter release. There was a tendency (P = 0.07) for the input resistance to decrease for cells from animals exposed to enrichment and for greater stimulation intensity to elicit unitary responses from cells of the IC group (P = 0.06; Fig.4 B).
The increase in the EPSP for the EC group was not associated with a change in the percent of transmission failures (Table3). Furthermore, the coefficient of variance was not different between groups. A change in synaptic strength, in the absence of a change in the coefficient of variance, suggests that quantal size is altered (Korn and Faber 1991). Results of the optimization procedure indicated a significant increase in quantal size [t(61) = 1.79,P < 0.05] and an increase in the signal-to-noise ratio [t(61) = 1.80, P < 0.05] (Table 3).
The noise deconvolution procedure makes assumptions about the distribution of responses (i.e., Poisson release), which may influence the results. Therefore the data were fit (P < 0.05) using a procedure that makes no assumption about the release process or the distance between the response amplitudes (Kullmann 1992). However, this technique requires a larger number of samples. Therefore the MEND procedure was employed only in cases in which at least 450 responses were recorded from the cell (EC:n = 9; IC: n = 10). For individual cells, the number of peaks that had a probability of occurrence of >5%, and were one noise standard deviation above 0 mV, ranged from 1 to 7 (3.53 ± 0.38, mean ± SE) and did not differ between the two groups (Table 3). Across all cells, variability in the distance between the peaks was ∼16% of the mean distance between the peaks. Finally, the mean distance between the peaks was greater for the EC group [t(18) 2.48, P < 0.05], consistent with the idea that the increase in the EPSP amplitude for EC animals is associated with an increase in quantal size. Figure5 provides an example of a cell from an EC animal, with a large signal-to-noise ratio and apparent quantal transmission.
Regression analyses indicated that mean unitary responses were not correlated with intracellular recording parameters, waveform characteristics, or stimulation intensity. The experience condition accounted for only 10% of the variability in averaged unitary response amplitude (R 2 = 0.10,P < 0.01) and, for 19 cells in which estimates were obtained by MEND analysis, quantal size accounted for <30% of the EPSP variability (R 2 = 0.28,P < 0.05). In contrast, approximately one-half of the variability in the mean synaptic responses could be explained by variability in release parameters including the percent of transmission failures (R 2 = 0.49, P< 0.0001), coefficient of variance (R 2 = 0.48, P < 0.0001), and mean quantal content estimated by nonlinear parameter optimization involving noise deconvolution (R 2 = 0.40, P < 0.0001; Fig. 6). In addition, the mean EPSP amplitude was negatively correlated with the paired-pulse ratio across all cells (R 2 = 0.29,P < 0.0001).
Effects of environmental enrichment on synaptic function
This study demonstrates that, following environmental enrichment, an increase in CA3-CA1 synaptic strength can be observed in vitro. A naturally occurring increase in synaptic strength associated with environmental enrichment has previously been described for perforant path synapses of the dentate gyrus (Foster et al. 1996,2000; Green and Greenough 1986). The enrichment-dependent increase in perforant path synaptic strength is mediated by increased α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor function and interacts with mechanisms for LTP. Previous research indicates that, like perforant path LTP, CA3-CA1 LTP induction is impaired shortly after an initial exposure to a novel environment or stressful situation (Garcia et al. 1997;Waters et al. 1997; Xu et al. 1997). Interestingly, the current study found that, in contrast to perforant path synapses, LTP could still be induced at CA3-CA1 synapses 24 h after the final exposure to novel environments. One possible explanation for regional and temporal differences is that the effects of enrichment may be more persistent in the dentate gyrus. For example, increased AMPA receptor binding is more enduring in the dentate gyrus (Cammarota et al. 1996). Alternatively, dentate granule cells and CA1 pyramidal cells may exhibit differential growth of new spines due to enrichment (Juraska et al. 1985;Moser et al. 1994). The emergence of new spines in CA1 might reflect new synapses that retain the potential to undergo subsequent LTP (Engert and Bonhoeffer 1999). Regardless, the results indicate that the mechanisms for LTP induction are differentially regulated by experience in the two regions.
Presynaptic and postsynaptic contributions
to synaptic strength
Recent controversies concerning presynaptic and postsynaptic contribution to synaptic plasticity have raised questions concerning the limitations of quantal analysis methods and validity of the parameter estimates. In many cases, the source of the differences can be traced to differences in experimental techniques (e.g., developmental state, temperature), the method for data selection, and inappropriate assumptions concerning the quantal analysis techniques (Korn and Faber 1991; McNaughton and Foster 1990; Redman 1990). The frequency of synaptic transmission failures and consequent estimates of quantal content provide classic measures of presynaptic function (del Castillo and Katz 1954). However, for central synapses, it is unclear whether a change in failure rate is due to altered probability of release or the number of transmission sites. Furthermore, the coefficient of variation is also influenced by variance in quantal size (Faber and Korn 1991). As Korn and Faber (1991) have pointed out, in the case of altered synaptic transmission, the only unequivocal conclusion is that quantal size is altered when the coefficient is unchanged as was observed in the current study. To provide valid estimates of quantal size, various deconvolution procedures have been devised. While noise deconvolution under constraints of Poisson release appear to adequately describe release at CA3-CA1 synapses of adults (Foster and McNaughton 1991; Sayer et al. 1990) and hippocampal cultures (Bekkers and Stevens 1995), constrained parameter optimization may over fit the data. This has led to the development of deconvolution procedures that do not make any assumption about the release model (Edwards et al. 1976;Kullmann 1992). However, estimates of quantal parameters according to unconstrained deconvolution are still limited by the sample size and background noise. Thus the limitations and assumptions of any single parameter extraction technique will determine the utility of that technique for quantal analysis.
The current study utilized a number of parallel techniques to reduce the reliance on assumptions underlying any single technique. The strongest support for the validity of the techniques is that the expected results were obtained for paired-pulse facilitation, for which an extensive literature indicates an increase in presynaptic function (Debanne et al. 1996; del Castillo and Katz 1954; Foster and McNaughton 1991; Hess et al. 1987). Quantal content was similar to that previously observed for synaptically connected cell pairs in adult rats (Foster and McNaughton 1991; Sayer et al. 1990), indicating that one or very few fibers were activated by the minimal stimulation. The mean unitary response amplitude for the conditioning pulse was inversely related to the paired-pulse ratio, the level of transmission failures, and coefficient of variance and was positively correlated with mean quantal content estimated by the nonlinear parameter optimization using noise deconvolution. The finding that presynaptic function accounts for most of the variability in synaptic strength is a consistent finding and is likely due to the large variability in mean quantal content compared with quantal size (Dumas and Foster 1995; Foster and McNaughton 1991; Markram et al. 1997). The strong relationship between mean quantal content and the synaptic response emphasizes the importance of this biophysical parameter in determining baseline synaptic strength.
Although presynaptic function is the major determinant of synaptic strength, changes in transmitter release are not responsible for the increase in synaptic strength due to experience. Despite the increase in synaptic strength for the EC animals, all measures of presynaptic function including the failure rate, coefficient of variance, quantal content, and the paired-pulse ratio were not altered by differential experience. The increase in synaptic strength, in the absence of a change in the coefficient of variation, provides strong evidence of an increase in quantal size (Korn and Faber 1991). This conclusion is supported by the results from the two noise deconvolution procedures that indicate that the increase in synaptic strength associated with differential experience is due to an increase in quantal size. Thus while quantal size contributed less to overall variability in synaptic strength, increased quantal size was observed in animals exposed to enrichment conditions.
It is unclear what mechanism underlies the increase in quantal size; however, experience-dependent enhancement of synaptic strength is associated with an increase in AMPA receptor binding, suggesting increased affinity or number of receptors at preexisting synapses (Foster et al. 1996; Gagne et al. 1998) resulting in increased postsynaptic responsiveness to transmitter. In addition, increased AMPA binding may represent new transmission sites that arise through the unmasking of postsynaptic receptors at previously “silent” synapses or the growth of new synaptic contacts. The absence of a change in quantal content or the MEND analysis estimate of the number of transmission sites (i.e.,n) suggests that the growth in synaptic strength was not due to the addition of more transmission sites between connected CA3-CA1 cell pairs. It is possible that differential experience was associated with divergence of synaptic contacts and the appearance of new transmission sites between cells that previously lacked functional connections. The tendency for reduced input resistance in the EC group is consistent with reports of an increase in spine density (Moser et al. 1994). The reduction in input resistance is not likely due to an increase in the spontaneous release of transmitter (e.g., GABA) since no difference was observed in the standard deviation of the noise-only records. Furthermore, a decrease in the stimulation intensity needed to activate a unitary response is consistent with an increased likelihood of a synaptic contact between individual CA3 fibers and the population of CA1 cells.
The results do not limit other mechanisms for naturally occurring changes in synaptic strength due to experience. The diversity of conclusions for quantal analysis of LTP indicate that the manifestation of distinct pre- or postsynaptic mechanisms depends on a number of relevant variables including developmental state and initial transmission parameters (Durand et al. 1996;Larkman et al. 1992; McNaughton and Foster 1990; McNaughton et al. 1994; Williams et al. 1993). For example, several laboratories have recently provided evidence that the increase in quantal content observed following LTP in the developing neonatal hippocampus may be due to postsynaptic mechanisms involving the unmasking of silent synapses (Durand et al. 1996; Isaac et al. 1995;Liao et al. 1995). It will be important for future studies to determine whether the new transmission sites represent an increase in the control of previously connected and communicating CA3-CA1 neurons or an increase the divergence of CA3 synaptic contacts with the population CA1 cells. Finally, it may be significant that LTP in CA1 of hippocampal slices from adults is associated with an increase in quantal size (Foster and McNaughton 1991). This initial discovery for a major role of quantal size in expression of LTP has been confirmed by a number of different investigators (Cormier and Kelly 1996; Isaac et al. 1998; Kullmann and Nicoll 1992; Liao et al. 1992; Manabe et al. 1992; Stricker et al. 1999). These results suggest thatN-methyl-d-aspartate (NMDA) receptor–dependent mechanisms may underlie the increase in synaptic strength associated with differential experience (Foster et al. 2000), while regulation of presynaptic function may depend on other physiological or behavioral processes (McNaughton et al. 1994;Stricker et al. 1999). Alternatively, activity-dependent release of neurotrophic factors can increase quantal size and influence synaptic connectivity (Sherwood and Lo 1999;Thoenen 1995). Regardless of the mechanism, the results suggest that postsynaptic alterations are important for the maintenance of a stable increase in synaptic strength associated with differential experience.
Special thanks go to K. Sharrow and J. Masse for help in preparing the manuscript.
This research was supported in part by National Institutes of Health Grant AG/NS-14979 and National Science Foundation Grant IBN-97230055.
T. C. Foster (E-mail:).
- Copyright © 2001 The American Physiological Society