Developmental Changes in Release Properties of the CA3-CA1 Glutamate Synapse in Rat Hippocampus

P. Wasling, E. Hanse, B. Gustafsson


Developmental changes in release probability (Pr) and paired–pulse plasticity at CA3-CA1 glutamate synapses in hippocampal slices of neonatal rats were examined using field excitatory postsynaptic potential (EPSP) recordings. Paired-pulse facilitation (PPF) at these synapses was, on average, absent in the first postnatal week but emerged and became successively larger during the second postnatal week. This developmental increase in PPF was associated with a reduction in Pr, as indicated by the slower progressive block of the N-methyl-d-aspartate (NMDA) EPSP by the noncompetitive NMDA receptor antagonist MK-801. This developmental reduction in Pr was not homogenous among the synapses. As shown by the MK-801 analysis, the Pr heterogeneity observed among adult CA3-CA1 synapses is present already during the first postnatal week, and the developmental Pr reduction was found to be largely selective for synapses with higher Pr values, leaving Pr of the vast majority of the synapses essentially unaffected. A reduction in Pves, the release probability of the individual vesicle, possibly caused by reduction in Ca2+ influx, seems to explain the reduction in Pr. In vivo injection of tetanus toxin at the end of the first postnatal week did not prevent the increase in PPF, indicating that this developmental change in release is not critically dependent on normal neural activity during the second postnatal week.


Following their formation, synapses undergo various collective changes, commonly referred to as synaptic maturation. One salient example of such maturation is a developmental decrease in release probability to a single action potential (Pr). This decrease in Pr has been documented among glutamate synapses in several regions, including the developing neocortex (Kumar and Huguenard 2001; Reyes and Sakmann 1999), calyx of Held (Iwasaki and Takahashi 2001; Taschenberger et al. 2002), striatum (Choi and Lovinger 1997), and hippocampus (Bolshakov and Siegelbaum 1995; but see Hsia et al. 1998; Muller et al. 1989), suggesting that it represents an essential step in the maturation of the glutamate synapse. In the striatum, patterned neural activity supports the induction of a long-term depression (LTD) that is expressed as a decrease in Pr (Choi and Lovinger 1997), suggesting that neural activity may underlie this form of maturation. However, whether inhibition of neural activity or of LTD during a critical period interferes with the maturation of Pr has never been directly tested.

Studies of hippocampal CA3-CA1 glutamate synapses have shown a considerable Pr heterogeneity among the mature synapses (Hessler et al. 1993; Huang and Stevens 1997). During the first postnatal week, however, the synapse population has been described both to have a uniformly high Pr (Bolshakov and Siegelbaum 1995) and to display, as the mature population, a considerable Pr heterogeneity (Hanse and Gustafsson 2001b). Likewise, conflicting data exist regarding a developmental change in Pr in that both a developmental decrease in Pr (Bolshakov and Siegelbaum 1995; Muller et al. 1989), and the absence of such a decrease (Hsia et al. 1998), have been reported. Whether there is a developmental decrease in Pr among these hippocampal synapses is thus uncertain. Moreover, if there is one, do the synapses change from a uniformly high Pr state to a lower one, or is there a more general shift in Pr within a heterogeneous Pr population?

The aim of this study is to examine to what extent the CA3-CA1 synapse changes its release properties in the early postnatal period, and if so, if this change is homogenous or heterogeneous among the synapses. Moreover, if such a change in Pr occurs, is it explained by a decrease in the number of vesicles that are immediately available for release (Dobrunz and Stevens 1997) and/or by a change in Pves, the probability of release of the vesicles themselves (Hanse and Gustafsson 2001b). Finally, is such a developmental change in Pr affected by an altered neural activity during the critical period?


Slice preparation and solutions

Experiments were performed on hippocampal slices from 3- to 27-day-old Wistar rats (n = 142). The animals were killed in accordance with the guidelines of the local ethical committee for animal research. Rats older than 8 days were anesthetized with isoflurane (Abbott) prior to decapitation. The brain was removed and placed in an ice-cold solution containing (in mM) 124 NaCl, 3 KCl, 0.5 CaCl2, 6 MgCl2, 26 NaHCO3, 1.25 NaH2PO4, and 10 d-glucose. In the whole cell experiments, the solution was composed of (in mM) 140 cholineCl, 2.5 KCl, 0.5 CaCl2, 7 MgCl2, 25 NaHCO3, 1.25 NaH2PO4, 1.3 ascorbic acid, and 7 dextrose. Transverse hippocampal slices (300–400 μm thick) were cut with a vibratome (Campden Instruments; or Slicer HR 2, Sigmann Elektronik) in the same ice-cold solution, and they were subsequently stored in artificial cerebrospinal fluid (ACSF) containing (in mM) 124 NaCl, 3 KCl, 2 CaCl2, 2 MgCl2, 26 NaHCO3, 1.25 NaH2PO4, 0.5 ascorbic acid, 3 myo-inositol, 4 d,l-lactic acid, and 10 d-glucose at 25°C. After ≥30 min of storage, a single slice was transferred to a recording chamber where it was kept submerged in a constant flow (∼2 ml/min) at 30–32°C. The perfusion ACSF contained (in mM) 124 NaCl, 3 KCl, 4 CaCl2, 4 MgCl2, 26 NaHCO3, 1.25 NaH2PO4, and 10 d-glucose. Bicuculline methiodide (20 μM) was always present in the perfusion ACSF to block GABAA receptor–mediated activity. All solutions were continuously bubbled with 95% O2-5% CO2 (pH ∼ 7.4). A surgical cut between the CA1 and CA3 regions and the higher than normal Ca2+ and Mg2+ concentrations were used to block spontaneous network activity.

Recording and analysis

Electrical stimulation of Schaffer collateral/commissural afferents and recordings of synaptic responses were carried out in the CA1 hippocampal region. Stimuli consisted of 0.2-ms negative or biphasic constant current pulses (15–80 μA) delivered through bipolar tungsten wires (resistance ∼0.1 MΩ, custom-made stimulator) or a glass pipette (resistance ∼0.5 MΩ; STG 1002, Multi Channel Systems, Reutlingen, Germany), in field and whole cell recordings, respectively. Stimulation electrodes were positioned in the stratum radiatum. Inputs received a test stimulus every 10 (field recordings) or 5 s (whole cell recordings). The stimulation intensity was set not to evoke firing in the postsynaptic neurons as evidenced by the absence of a population spike distorting the field excitatory postsynaptic potential (EPSP).

Field EPSP recordings were made by means of a glass micropipette (filled with 1 M NaCl) in s. radiatum. Field EPSPs were amplified with an Axoclamp-2A (Axon Instruments) and filtered at 3 kHz. Data were digitized (10-kHz sampling rate) and collected using a PC computer. Whole cell patch-clamp recordings were performed on visually identified CA1 pyramidal cells, using infrared-differential interference contrast videomicroscopy (CV-M50 IR, JAI Corp.) mounted on a Nikon E600FN microscope (Nikon). The pipette solution contained (in mM) 130 Cs-methanesulfonate, 2 NaCl, 20 HEPES, 0.2 EGTA, 5 QX-314, 4 Mg-ATP, and 0.4 GTP (pH ∼ 7.2 and osmolality 290–300 mOsm). Patch pipette resistances were 3–5 MΩ. Excitatory postsynaptic currents (EPSCs) were recorded at a sampling frequency of 10 kHz and filtered at 1 kHz, using an EPC-9 amplifier (HEKA Elektronik, Lambrecht, Germany). Cells were held in voltage-clamp mode at –70 mV for AMPA EPSC recordings. Series resistance was monitored using a 5-ms 10-mV hyperpolarizing pulse. Series resistance was not allowed to change more than ∼10% during an experiment; otherwise the experiment was discarded. Evoked responses were analyzed off-line using custom-made IGOR Pro (WaveMetrics, Lake Oswego, OR) software. Field EPSP magnitude was estimated by linear regression over the first 0.8 ms of the initial slope. Field NMDA receptor–mediated responses were measured by the initial 40-ms area. The presynaptic volley was measured as the peak-to-peak amplitude of the initial positive-negative deflection, and it was not allowed to change by >5% during the experiment. AMPA EPSCs were measured as the difference between the baseline level immediately preceding the stimulation artifact and the mean amplitude during a 1-ms time window around the negative peak between 3 and 8 ms after the stimulation artifact. Slope measurements (least-square linear regression from 20 to 80% of the rising phase) of EPSCs confirmed the results obtained with amplitude measurements (data not shown). In Fig. 3, all experiments were performed in 50 μM d-AP5 and 100 μM Ni2+ to block induction of long-lasting changes of synaptic strength (Wasling et al. 2002). In experiments in which the sensitivity of the field EPSP and paired-pulse facilitation (PPF) to extracellular [Ca2+] was examined, the sum of [Ca2+] and [Mg2+] was kept at 10 mM to minimize changes in membrane excitability while altering Ca2+ ion concentration. We also did not observe any changes in the amplitude of the presynaptic volley while changing the divalent ion concentrations in this manner (data not shown). When using extracellular [Ca2+] above 4 mM, recordings always started at the higher [Ca2+] solution before changing to the lower [Ca2+] solution. When field EPSPs were compared when changing from 4/0 to 4/4 Ca2+/Mg2+, the change in volley was corrected for. In experiments estimating the releasable pool of vesicles, high-frequency (20 and 50 Hz) stimulation was used (Fig. 6). The axons appeared to follow this frequency without failure as judged from the stability of the amplitude of the fiber volley. Thus the ratio between the 10th and 1st volley in a 50-Hz train was 0.96 ± 0.05 in experiments (n = 4) in which the volley was obtained in isolation during blockade of AMPA receptors (10 μM NBQX). To calculate average release probability based on MK-801–dependent decrease in NMDA field EPSP magnitude, the following equation was used: Pr = [1 − exp(−1/τ)]/FB, where FB is the fraction of NMDA receptors blocked by MK-801 after transmitter release at a synapse (Hessler et al. 1993). Data are expressed as means ± SE. Statistical significance for paired and independent samples was evaluated using Student's t-test.

Tetanus toxin experiments

A previous study from our laboratory (Groc et al. 2003) has shown that injection of tetanus toxin into the hippocampus in vivo (at P1) leads to an ∼80% reduction of spontaneous transmitter release for several days; the reduction still was >50% at P6–P8. The same degree of reduction was found in slices taken close to the injection site as in those taken >600 μm away, indicating that the toxin had spread over a large part of the hippocampus. Nevertheless, in this study, only the two slices closest (on either side) to the injection site were taken from each tetanus toxin-injected hippocampus, to ascertain an as extensive blockade as possible.

To examine the possible effect of neural activity on the developmental change in Pr, tetanus toxin was injected into the hippocampus of rat pups at P1 or at P6–P7, by the same person and using the same procedures as previously (Groc et al. 2003). The rats were anesthetized by inhalation of isoflurane, and their heads were placed in a surgical mask to maintain the skull stable. A constant flux of a mixture isoflurane/air was applied inside the surgical mask. A heat sheet was used to maintain body temperature. A midline incision was made on the head and a hole was drilled in the skull (OD, 0.4 mm). The stereotactic coordinates for injection were (P1/P6) as follows: anterioposterior, –1.4/–2.0 mm; mediolateral, +2.0/+2.9 mm, coordinates relative to the bregma; dorsoventral, –2.0/–2.9 mm from the cortical surface. Tetanus toxin was dissolved in a phosphate-buffered saline (0.1 M, pH 7.4) to the final concentration of 10 μg/ml. Five nanograms of tetanus toxin was injected into the hippocampus at ∼125 nl/min using a fused silica needle (ID, 75 μm; OD, 150 μm; Skandinavia Genetec AB). After injection, the needle was left in situ for 5 min to reduce reflux. The incision was chemically sutured (Vet-Seal, B. Braun Med).


Chemicals were from Sigma-Aldrich, except for d-AP5, ω-conotoxin VIA, MK-801, NBQX, and QX-314 (Tocris Cookson, Bristol, UK). Cytochrome C (0.2 mg/ml; bovine heart) was included in all ω-conotoxin GVIA experiments to block nonspecific peptide binding sites in the perfusion system.


PPF increases during the second postnatal week

PPF (interstimulus interval, 50 ms), i.e., the synaptic response following the second stimulus in a stimulus pair is larger than that following the first one, is considered a sensitive measure of the release probability (Pr) of a synapse (Zucker and Regehr 2002). When examined for field EPSPs in slices from P3–P6 rats, PPF was, on average, found to be absent (Fig. 1A ). PPF then emerges and becomes successively larger during the second postnatal week, reaching a value of about 1.6 at P15 (Fig. 1A). This result would suggest that the second postnatal week constitutes a period of substantial decrease in Pr among these synapses.

FIG. 1.

Paired-pulse (PP) ratio increases in a largely activity-independent manner during the 2nd postnatal week. A: summary graph showing PP ratio as a function of postnatal day (n = 86). Individual experiments (○) and mean (▪) at indicated postnatal day. Insets: example PP field excitatory postsynaptic potential (EPSP) responses from P6 (left) and P13 (right) rat. Calibration bars: 100 μV/10 ms. B: average PP ratio in control and tetanus toxin-injected animals at P5–P6 (injected at P1) and P12 (injected at P6–P7), respectively. *P < 0.001. Inset: schematic view of in vivo injection site of tetanus toxin in the hippocampus. See methods for further details.

To examine whether the change in PPF during the second postnatal week is driven by neural activity, tetanus toxin was injected in vivo into the hippocampus of P6–P7 rats (n = 9; see methods). When examined in slices taken from rats at P12, PPF averaged 1.48 ± 0.03 (n = 31) compared with 1.55 ± 0.04 (n = 25) in slices from noninjected rats (Fig. 1B). It thus appears that the change in PPF proceeds unabated, despite the considerable activity reduction caused by such an injection (Groc et al. 2003). However, the tetanus toxin injection may, either by a differential action on synapses with different Pr or by the activity reduction itself, have caused an increase in PPF, thus masking an effect on the second week developmental change in PPF. The effect of tetanus toxin was thus examined prior to the second week when Pr appears to be rather stable (see also Hanse and Gustafsson 2001b). Injection of tetanus toxin and examination of PPF at P5–P6, did, however, not reveal significantly larger PPFs (1.15 ± 0.06, n = 9) than in age-matched controls (1.08 ± 0.04, n = 21; Fig. 1B).

Release probability changes during the second postnatal week in a nonuniform manner among the synapses

Since Pr is a major determinant of PPF, with a decrease in Pr leading to an increased PPF, the simplest explanation for the observed change in PPF is a second week decrease in Pr. The neonatal synapse was originally reported to have (uniformly) a very high Pr, approaching 1.0 (Bolshakov and Siegelbaum 1995), and a reduction of this Pr during the second postnatal week would then seem a likely explanation for the increase in PPF. However, more recent studies have shown that Pr in the neonatal synapse, as in the more adult one, is very heterogeneous, with values from close to one to close to zero (Hanse and Gustafsson 2001b). With such a heterogeneous population, it is less clear in what manner the synapses may have changed, uniformly or not, to explain the PPF change. To explore this matter, we examined the MK-801–induced decay of pharmacologically isolated NMDA receptor–mediated field EPSPs (10 μM NBQX). Figure 2A shows the average decay curves obtained from such experiments on P6 (○, n = 9) and P12 (•, n = 9) rats, respectively. In both cases, the NMDA EPSP decays with a rapid initial phase followed by a slower phase. The decay time course obtained in MK-801 experiments was originally fitted by a double exponential indicating two major categories of CA3-CA1 synapses, a high and a low Pr group (Hessler et al. 1993; Rosenmund et al. 1993). When examined for the individual experiments, the present decay curves were also better fitted by a double than by a single exponential [sum of squares for the single exponential fits were at P6 7.4 ± 2.2 (n = 9) and at P12 7.6 ± 2.35 (n = 9) times larger than for the double exponential fits]. This result is consistent with a similar kind of Pr heterogeneity among the synapses during the first as well as after the second postnatal week (Dobrunz and Stevens 1997; Hanse and Gustafsson 2001b). However, the decay is, on average, substantially slower for the P12 synapses (Fig. 2A, •) than for the P6 ones (Fig. 2A, ○), suggesting a developmental decrease in Pr. As judged from the parameters of the double exponential fits obtained from the individual experiments, the decrease in Pr was mainly restricted to the high Pr group of synapses. Thus, whereas the time constant of the fast component changed considerably from 3.7 ± 0.2 (n = 9) at P6 to 7.8 ± 0.7 (n = 9) at P12 (P < 0.001), there was no significant difference between the time constants of the slow component: 55.2 ± 1.8 (n = 9, P6) and 63.8 ± 3.6 (n = 9, P12).

FIG. 2.

Developmental difference in MK-801 induced decrease of N-methyl-d-aspartate (NMDA) field EPSP magnitude. A: summary graph of the decrease of NMDA field EPSPs in the presence of 20 μM MK-801. Field EPSP magnitudes are plotted against stimulus number and are normalized to the 1st response. The progressive block at P6 (○, n = 9 experiments) and at P12 (•, n = 9 experiments) was fitted with a double exponential equation: A1exp(–t1) + A2exp(–t2). The best fit (at P6; P12) was obtained with A1 = 62.9; 59.0, τ1 = 3.4; 7.2, A2 = 37.1; 41.0, and τ2 = 52.5; 59.3. Note that these values represent the best fit to the average decay and differ somewhat from the values in the text that are based on averaging the values obtained from best fits to the decay from each individual experiment. B: stimulus interruption (in the absence of MK-801) for 10 min caused an increase in NMDA field EPSP magnitude that subsequently decayed back to baseline values. Values for the subsequent decay normalized to the 1st field EPSP following the resumption of stimulation are shown for P6 (○) and P12 (•) experiments. C: averaged NMDA field EPSPs, taken from the experiments in A, recorded before the addition of MK-801 (P6, gray lines; P12, black lines). NMDA field EPSP from each experiment represents the average from 15 consecutive responses. Average NMDA field EPSPs from each experiment were normalized to the same peak amplitude.

Knowing the fraction of NMDA receptors that are blocked at each trial (when glutamate is released), the (average) Pr of the high and low Pr groups of synapses can be calculated (Hessler et al. 1993). Moreover, the proportion of synapses belonging to the high and low Pr groups, respectively, can be obtained from the relative area of the fast and slow exponential components (exponential amplitude × time constant) (Rosenmund et al. 1993). Assuming a value of 0.3 for the fraction of NMDA receptors blocked/trial (Hessler et al. 1993; Huang and Stevens 1997; Rosenmund et al. 1993), the synaptic population at P6 would consist of a high Pr group with an average Pr value of 0.88 ± 0.04 (n = 9) and of a low Pr group averaging 0.067 ± 0.003 (n = 9) that would constitute 10 and 90% of the population, respectively. At P12, the corresponding Pr values would be 0.52 ± 0.02 (n = 9) and 0.068 ± 0.004 (n = 9), respectively, for 13 and 87% of the population. Thus Pr changes during the second week, but this change is restricted to the high Pr synapses, constituting a minority of the synapses. Nonetheless, these synapses will produce more than one-half of the field EPSP, and the decrease in their Pr value would result in an overall reduction of Pr (at P12) to 85% of the P6 value (Pr = weighted average of the average Pr of the high and of the low Pr group).

The question arises whether such a nonuniform change in Pr can alter PPF in the manner observed above. The MK-801 experiments were performed in 4/0 extracellular Ca2+/Mg2+, implying different release conditions from those in which the values in Fig. 1A were obtained (4/4 Ca2+/Mg2+). When PP ratio for the (non-NMDA) field EPSP was examined in 4/0 Ca2+/Mg2+, it averaged 0.90 (n = 12) and 1.25 (n = 6) for P6 and P12 synapses, respectively. The PP ratio difference to be explained is thus one from 0.90 to 1.25. Using data from experiments using minimal stimulation to examine the relation between PP ratio and Pr for single CA3-CA1 synapses during the first postnatal week (Hanse and Gustafsson 2001a), the PP ratios for high Pr (0.88) and low Pr (0.07) synapses at P6 would be about 0.25 and 2.20, respectively, resulting in an overall PPF of 1.04. Corresponding values for the P12 synapses would be PP ratios of about 0.60 and 2.20, respectively, resulting in an overall PPF of 1.32. The observed change in PPF from P6 to P12 can thus be accounted for by the rather selective decrease in Pr of high Pr synapses indicated from the MK-801 experiments.

Possible complicating factors with the MK-801 experiments

The NMDA EPSP observed when stimulation commenced after a 10-min-long stimulus interruption was substantially larger than the baseline level present before stimulus interruption (Fig. 2B). The NMDA EPSP decay in MK-801 should then include a decay of this enhancement, which is unrelated to the MK-801 block of open channels. However, the decay of this enhancement is blocked by NMDA receptor antagonists (Niu et al. 1999), implying that it will only occur following NMDA receptor activation. Since NMDA receptor activation will lead to block by MK-801, it seems reasonable to believe that the decay of this enhancement will not occur in the presence of MK-801. Another possible complication is that the MK-801–induced decay is not only a function of the Pr of a synapse, but also of the open probability of the NMDA receptor channels. The NMDA EPSPs observed in the P12 synapses were found to decay faster than those in P6 synapses (P6, τ = 66 ± 5 ms, n = 9; P12 = 42 ± 4 ms, n = 9; P < 0.05; Fig. 2C), possibly reflecting a developmental shift in NMDA receptor subunit composition to an increasing proportion of NR2A subunits (Kirson et al. 1999). However, such a shift from NR2B to NR2A subunits should not be expected to affect the total open probability of the NMDA receptor channels during an EPSP, since the reduced duration is compensated for by a higher peak open channel probability (Chen et al. 1999).

It should also be noted that the good double exponential fit to the MK-801 data not necessarily implies the existence of two separate homogeneous groups of synapses, but may also be compatible with more continuous Pr distributions (Huang and Stevens 1997). The high and low Pr groups of synapses may thus represent the upper and lower parts of a continuous distribution rather than two distinct groups.

Does the sensitivity of Pr and PPF to Ca2+/Mg2+ change during the second postnatal week?

The change in Pr indicated by the MK-801 experiments can be explained by a change in Pves as well as in the number of vesicles that are immediately available for release. To explore whether a Pves change is involved, the effect of changes in the extracellular Ca2+/Mg2+ concentration on field EPSPs was examined at both P6 and P12. To minimize changes in axonal excitability, divalent ion concentration was kept constant at 10 mM, and Ca2+/Mg2+ concentrations used were from 1/9 ≤ 10/0. Field EPSP values obtained at these different concentrations, normalized to the value obtained at 4/6 Ca2+/Mg2+, show that P6 and P12 synapses respond differently to these changes in Ca2+ influx (Fig. 3A). Thus, while reductions in Ca2+ influx decrease the EPSPs of P6 and P12 synapses in a rather similar manner, an increase in Ca2+ influx causes a considerably larger increase of the EPSP at P12 than at P6. This result suggests that P6 synapses are closer to saturation of their release machinery to Ca2+ ions, indicative of either a larger influx of Ca2+ ions or of a shift in the Ca2+ sensitivity curve. Nonetheless, this result seems compatible with a reduction in Pves as underlying the change in Pr from P6 to P12. These synapses also responded differently with respect to PPF to changes in the Ca2+/Mg2+ relation (Fig. 3B). Thus, while a larger Ca2+ influx resulted in a substantial PPF decrease in P12 synapses, there was little change at P6. On the other hand, a decrease in Ca2+ influx caused a large PPF increase at P6 and a small one at P12. These results are also compatible with the release machinery being closer to saturation by Ca2+ ions at P6 than at P12.

FIG. 3.

Age difference in the sensitivity to changes in extracellular Ca2+ and Mg2+ concentrations. A and B: summary graphs showing changes in field EPSP magnitude and PP ratio as a function of [Ca2+]o and [Mg2+]o at P6 (open symbols) and at P12 (closed symbols). C: changes in field EPSP magnitude plotted against PP ratio. Arrows on the P6 curve indicate PPF values at P6 (□) and at P12 (▪) in 4:4 [Ca2+]/[Mg2+] solution. In A and C, field EPSP magnitude is normalized to that obtained at 4:6 [Ca2+]/[Mg2+]. Note that some error bars are within symbols. For the various [Ca2+]/[Mg2+] concentrations (1/9, 2/8, 3/7, 4/6, 7/3, and 10/0) the n values were 3, 5, 7, 12, 5, and 5 (P6) and 0, 6, 6, 10, 7, and 6 (P12).

Can a reduced Ca2+ influx quantitatively account for both the decrease in Pr and increase in PPF? The plot in Fig. 3C between (relative) field EPSP amplitude and PPF shows that the field EPSP (at P6) has to decrease to 40% of control for PPF to increase from its control value (1.11 ± 0.04, n = 12) to the value found at P12 (1.55 ± 0.04, n = 16). As indicated from the MK-801 experiments, overall Pr at P12 would be 85% of that at P6. However, these experiments were performed at 4/0 Ca2+/Mg2+. A change from 4/0 to 4/4 Ca2+/Mg2+ decreased the field EPSP on average to 67 (n = 12) and 52% (n = 6) (after volley correction, see methods) for P6 and P12 synapses, respectively. This would indicate that, at 4/4 Ca2+/Mg2+, Pr at P12 should be 66% [85% × (52/67)] of that at P6. This value (66%) differs substantially from the decrease to 40% expected from the data in Fig. 3C. Thus a reduction in Ca2+ influx (by altered Ca2+/Mg2+ ratio) does not replicate the developmental change in release properties. A close match may, however, not be expected since the developmental change in release properties only affects high Pr synapses. Moreover, the alteration of the Ca2+/Mg2+ ratio may have effects on the release machinery unrelated to the change in Ca2+ influx per se (such as changes in resting intracellular Ca2+ levels).

Is the expression of presynaptic voltage-gated Ca2+ channels altered?

In interpreting the above results as indicating a developmental decrease in Pves, possibly related to a decrease in Ca2+ influx, the question arises of what may underlie such a change. One possibility is a developmental change in the type of voltage-gated Ca2+ channels that is expressed in the presynaptic membrane and whose openings underlie the Ca2+ influx leading to release. In fact, at some synapses, including hippocampal glutamate synapses in culture, there is a developmental reduced contribution from N-type voltage-gated channels (see Iwasaki et al. 2000). We thus examined the effect of ω-conotoxin GVIA, a blocker of these channels, on field EPSPs from P6 and P12 synapses, respectively. As shown from one such experiment (P6 rat), this toxin had a substantial effect on release, reducing the field EPSP to less than one-half its amplitude (Fig. 4A ). On average, the ω-conotoxin–sensitive fraction of the EPSP was 55 ± 4% (n = 4) at P6. This fraction was not significantly different (51 ± 2%, n = 6) at P12, indicating that there is no major shift in N-type channel expression during this time period (Fig. 4B). This fraction is also similar to what has been reported in 3- to 4-wk-old animals (Wheeler et al. 1994). The reduction in the field EPSP by ω-conotoxin was associated with an increase in PPF [from 1.15 ± 0.04 to 1.72 ± 0.02 at P6 (n = 4) and from 1.62 ± 0.08 to 2.12 ± 0.10 at P12 (n = 6)]. Thus (by linear approximation) the EPSP, i.e., Pr, at P6 has to be reduced to 60% to produce a PPF increase corresponding to that at P12. Such a reduction is in line with that calculated above to occur between P6 and P12 (66%), suggesting that a reduction in Ca2+ influx, by blockade of N-type channels, can essentially replicate the developmental change in release properties.

FIG. 4.

Effect of Ca2+ channel blocker on synaptic transmission is not age dependent. A: example experiment (P6) of the reduction in field EPSP magnitude following application of 10 μM ω-conotoxin GVIA. Insets: average field EPSPs (n = 10) taken at indicated time-points. Calibration bars: 100 μV/10 ms. B: summary graph of the ω-conotoxin GVIA-sensitive fraction at P6 and P12.

Variation in action potential duration with age

During the second postnatal week, there is a substantial change in action potential characteristics, at least as observed from recordings from CA1 pyramidal cell somata and from Calyx of Held (Spigelman et al. 1992; Taschenberger and von Gersdorff 2000). The amplitude is increased, and the duration is substantially reduced, likely as a consequence of changes in the expression of voltage-gated sodium and potassium channels. If such a change also occurs for the action potential invading the presynaptic bouton, one may expect a developmental decrease in Ca2+ influx associated with the reduced duration of the spike. As previously described (Berg-Johnsen and Langmoen 1992; Laerum and Storm 1994), procedures that alter the action potential duration, such as TEA and temperature, have significant effects on the fiber volley duration. Following the total blockade of the NMDA receptor–mediated EPSP by MK-801, the fiber volley could be observed in isolation and thereby be accurately measured with respect to its duration. The volley observed at the end of the first postnatal week averaged 1.59 ± 0.03 (n = 24) ms compared with 1.32 ± 0.02 (n = 42) ms for that observed 1 wk later, i.e., there was a substantial shortening (Fig. 5). To examine the activity dependence of this shortening, the fiber volley was also measured in slices from P12 rats that had been injected with tetanus toxin at P6. The volley duration in these slices averaged 1.28 ± 0.03 (n = 13), indicating that the action potential shortening proceeds independent of neural activity.

FIG. 5.

The presynaptic volley duration shortens during development. A: summary graph of volley duration at different postnatal days and following tetanus toxin injection at P6. *P < 0.001. Inset: example volleys at P6 (thin line) and at P12 (thick line). The example volley at P6 is scaled to the same peak amplitude as the P12 volley (dotted line). Calibration: 0.2 mV, 0.5 ms. B: volley duration histogram for P4–P7 (open bars) and P10–P14 (filled bars) with Gaussian fits superimposed (dashed lines). In these experiments, both AMPA and NMDA receptors were blocked (10 μM NBQX and 20 μM MK-801).

Does the immediately releasable pool of vesicles change with age?

Pr is determined not only by Pves but also by the number of vesicles that are immediately available for release, i.e., the preprimed pool (Hanse and Gustafsson 2001b). On one hand, a developmental decrease in pool size could contribute to the developmental decrease in Pr in high Pr synapses. On the other hand, a developmental increase in pool size that parallels the developmental decrease in Pves may explain the nonuniform developmental change in Pr indicated by the MK-801 data. This is because an increased pool size will have a larger impact on the Pr of low than of high Pves synapses (Hanse and Gustafsson 2001b), and may thus counteract a developmental decrease in Pr in low but not in high Pr synapses.

An estimate of the preprimed pool can be obtained by evoking EPSCs by a brief high-frequency train and plotting the cumulative EPSC amplitude against train length (Schneggenburger et al. 1999). The rationale is that when the cumulative amplitude rises in a linear manner, release is no longer coming from vesicles in the preprimed pool but only from newly recruited ones (that are assumed to be recruited/released at the same rate). By assuming that this recruitment starts at zero time, the intersection of the linear relation with the y-axis will give an estimate of the preprimed pool. This estimate will not be in absolute terms, but will be the relation between the preprimed pool and the number of vesicles released by the first stimulus in the train, i.e., it will be a function of Pr. Such estimates of the preprimed pool at P6 and P12 were obtained by evoking synaptic activity by 10 impulse trains at 20 and 50 Hz. Since the estimate should be independent of train frequency, the use of both frequencies can be seen as an internal control. In these experiments, synaptic activity was recorded as EPSCs using whole cell recording (rather than field recording) primarily to avoid buildup of depolarization (and subsequent spike activation) during the train. An example of such cumulative EPSC curves, normalized to the peak amplitude of the first EPSC, given by 20- and 50-Hz trains, respectively, are shown in Fig. 6A. The latter half of both curves is well described by a linear relation; these linear relations intersect with the y-axis at the same value, about twice that of the first EPSC. Averaged data from such experiments showed similar pool size estimates using 20- and 50-Hz trains (1.64 and 1.66 at P6; 2.06 and 2.11 at P12), as should be expected. However, compared for P6 and P12 synapses, there was a significant age group difference, the values being 1.64 ± 0.11 (n = 15) and 2.07 ± 0.09 (n = 13) for P6 and P12 synapses, respectively (Fig. 6B). Since these estimated pool values are a function of Pr, they will, however, only be useful if one knows the relation between the Pr values of the P6 and P12 synapses when these pool estimates were performed. It should first be noted that PPF of the synapses involved averaged 1.07 (P6) and 1.65 (P12) (estimated from the 1st of the 30 trains given for each synapse; see following text), indicating that these synapses were representative of P6 and P12 synapses, respectively. As calculated above, at 4/4 Ca2+/Mg2+, Pr at P12 should be 66% of that at P6. However, during the initial stages of the train activation (5–10 1st of the 30 trains given at 0.2 Hz) release conditions were altered in that the EPSC amplitude and/or PPF changed. Thus the pool estimations were performed after a steady state had been reached, during which the EPSC (following the 1st stimulus of the train) of P6 synapses had decreased (on average) to 0.68 of the initial value, whereas that of P12 synapses was less altered (0.93). Thus, under these conditions, Pr of the P12 synapses should be 90% [66% × (0.93/0.68)] of that of the P6 ones. The pool estimates of 1.64 ± 0.11 (n = 15) and 2.07 ± 0.09 (n = 13) given above should be translated into 1.64 ± 0.11 and 1.86 ± 0.08 (90% of 2.07) after correcting for this estimated Pr difference, indicating no significant change in the overall value of preprimed pool during the second postnatal week.

FIG. 6.

Estimation of vesicle pool size. A: example experiment showing cumulative excitatory postsynaptic current (EPSC) amplitudes (normalized with respect to the amplitude of the 1st response) obtained by 20- (○) and 50-Hz (▪) stimulation. Dashed lines are extrapolations from linear regression over the last 5 data points. Intersection between dashed lines and the y-axis represents the estimated pool size as a function of initial release probability (indicated by arrow). B: mean relative pool size at P4–P7, at P12–P15, and at P12–P15, with mean value corrected for lowered Pr.


These results show that the second postnatal week constitutes a period of substantial alterations in the release properties of CA3-CA1 hippocampal synapses manifested as a large increase in paired-pulse plasticity. Hence, these results do not substantiate a previous report that paired-pulse plasticity is unaltered throughout this time period (Hsia et al. 1998), but rather those reports indicating such a change (Bolshakov and Siegelbaum 1995; Muller et al. 1989). However, the results also suggest that the alteration takes place in only a subpopulation of the synapses, those with high Pr values, and leaves most of the population largely unaffected. This may explain the failure of Hsia et al. (1998) to observe the developmental change in paired-pulse plasticity, since fewer synapses will be sampled in whole cell than in field recordings. The alteration in paired-pulse plasticity was found to take place despite a profound blockade of synaptic transmission during this time period, by local in vivo tetanus toxin injection, suggesting that normal neural activity during the second postnatal week is not a prerequisite for the alteration in paired-pulse plasticity to occur. Our results suggest that the observed alteration in release properties is explained by a large reduction in the release probability of the individual vesicles (Pves) in the affected synapses, possibly mediated by a reduced Ca2+ influx. Since the observed modification in release largely occurs within a few days, it is not likely a consequence of synaptogenesis per se, which occurs continuously throughout the first 4 wk (Steward and Falk 1991).

The CA3-CA1 synapses in the first neonatal week have previously been described as uniformly characterized by a Pr close to one and by paired-pulse depression (Bolshakov and Siegelbaum 1995). This behavior was in contrast to that found in 2- to 3-wk-old animals, where Pr was substantially lower and there was PPF (Bolshakov and Siegelbaum 1995), indicating a switch from high to low Pr synapses in the intervening period. More recently, using minimal stimulation technique, a more heterogeneous first week Pr distribution was described (Hanse and Gustafsson 2001b). The present results support the existence of such first week heterogeneity. As in older animals (Hessler et al. 1993), the NMDA EPSP decayed, following MK-801 application, with an initial rapid and a later slower exponential phase that indicates the presence of both high and low Pr synapses. There is presently no clear understanding of the exact nature of the Pr distribution in older (>2 wk) animals. That is, MK-801 data taken to indicate the existence of two distinct groups of synapses (Hessler et al. 1993; Rosenmund et al. 1993) can equally well be accounted for by more continuous Pr distributions (Huang and Stevens 1997). Nevertheless, these results suggest that the neonatal population consists of synapses with higher and lower Pr, the low Pr synapses constituting the majority of the population. Assuming the fraction of NMDA receptors channels to be blocked following vesicular release to be 0.3–0.4 (Hessler et al. 1993; Huang and Stevens 1997; Rosenmund et al. 1993), the high Pr group would have an average Pr of 0.88–0.66 (at 4/0 mM Ca2+/Mg2+) and the low Pr group would average 0.07–0.05. These deduced Pr values are not incompatible with data from the minimal stimulation experiments (Hanse and Gustafsson 2001b). Although these latter data show a continuous distribution of Pr values, they indicate two peaks, at low (<0.1) Pr and high (0.4–0.6) Pr that suggests the presence of two broad categories of synapses. These peaks may represent the presently deduced high and low Pr groups, considering the uncertain value of the blocking fraction, and the fact that the minimal stimulation data were obtained at 4/4 Ca2+/Mg2+, and using train stimulation, both procedures that would decrease Pr.

A main finding in this study was that the second week synaptic maturation in Pr only affects the high Pr synapses, changing the estimated average Pr from 0.88 to 0.52 (blocking fraction = 0.3). Calculations using PP ratio–Pr data from minimal stimulation experiments (Hanse and Gustafsson 2001a) suggested (see results) that such a selective Pr change would produce an overall change in PPF consistent with that observed experimentally. However, whether this developmental Pr change is a distinct shift within a homogenous group of synapses with high Pr, or a variable shift in Pr among the synapses with the higher Pr values within a continuous distribution is, as indicated above, uncertain. Nevertheless, the maturation is not a switch of release properties from one group to another but an adjustment of the release probability within one group, or population, of synapses. This second week maturation thus differs from that observed in cultured hippocampal neurons where a proportion of the high Pr synapses converts into low Pr ones (Chavis and Westbrook 2001). On the other hand, it agrees with the developmental change within the Calyx of Held synapse (Iwasaki and Takahashi 2001). In this synapse, containing hundreds of release sites, a double-exponential fit of MK-801 data also suggested a shift within a high Pr group of release sites from much the same values and to much the same extent as shown presently for the hippocampal synapses, whereas little change was observed in the low Pr group. Thus, interestingly, the developmental change observed among a population of heterogeneous hippocampal synapses, containing a single release site, is mirrored by that occurring in a heterogeneous population of release sites within a single synapse.

What underlies the decrease in Pr in the high Pr group?

Controversy exists to what extent heterogeneity in Pr is determined solely by a variation in the immediately releasable vesicle pool (Dobrunz and Stevens 1997), or in addition, by variation in Pves among the synapses (Hanse and Gustafsson 2001b). According to the first viewpoint, the developmental change in Pr should be explained by a reduced vesicle pool in high Pr synapses. If Pr is decided by the pool alone, the high Pr group would produce the majority of the releasable vesicles, and a reduction of its pool size would be detectable. However, this study found no evidence for a developmental reduction in pool size, estimated using the cumulative EPSC amplitude arising from train stimulation (Schneggenburger et al. 1999). On the other hand, high Pr synapses that differ from the low Pr synapses mainly with respect to Pves will produce only a minority fraction of the releasable pool, and a reduction in this fraction may go undetected in our pool estimations. However, a pool decrease in high Pr synapses would not be expected to lead to a developmental increase in PPF (Hanse and Gustafsson 2001a).

Evidence for the involvement of a Pves change, rather than a pool change, in the developmental Pr reduction is the differential effect of changes in Ca2+/Mg2+ ratio on the EPSPs evoked from P6 and P12 synapses, respectively. Specifically, high Ca2+/Mg2+ ratios increased the EPSP of P6 synapses considerable less than that of P12 synapses, suggesting that the P6 synapses are closer to saturation with respect to Ca2+, i.e., operate higher up on the upper limb of the Pr–Ca2+ relation. This suggestion seems reasonable with respect to the high Pr group of synapses at P6. However, considering the large number of low Pr synapses in the P6 population, one may have expected greater latitude for increased EPSP amplitude with increased Ca2+/Mg2+ ratios. Previous MK-801 experiments (Hessler et al. 1993) have, however, indicated a smaller effect of Ca2+ influx changes (induced by the addition of Cd2+ to the perfusion solution) on the low than on the high Pr synapses. The low Pr group (at the high Ca2+/Mg2+ ratios used) may also operate on the upper limb of its Pr–Ca2+ relation. That is, the maximal Pr value of these synapses is substantially <1, as has been observed for cortico-geniculate synapses (Granseth and Lindstrom 2003, 2004). The effect of changes in Ca2+/Mg2+ ratio on PPF was also consistent with P6 synapses operating higher up on the upper limb of the Pr–Ca2+ relation. These data point toward a second week reduction in Pves that should be largely restricted to high Pr synapses, possibly explained by a reduction in the amount of Ca2+ reaching the Ca2+ acceptors (triggering release). In support of this explanation, a decrease in Ca2+ influx induced by a Ca2+ channel blocker could quantitatively replicate the combined Pr and PPF shift occurring from P6 to P12.

It should be noted, however, that a restriction of the developmental Pr reduction to the high Pr synapses not necessarily implies that the Pves reduction is so restricted. Since an increased pool size will have more impact on Pr in low than in high Pves synapses (Hanse and Gustafsson 2001b), a developmental increase in preprimed pool size for all synapses could counteract a decrease in Pves in low but not in high Pves synapses, specifically if, as noted above, Pves in low Pr synapses is less affected by changes in calcium influx than is Pves in high Pr synapses. However, this scenario is not supported by our pool estimations, which did not indicate any significant increase in preprimed pool size.

These results cannot explain why the amount of Ca2+ reaching the Ca2+ acceptors should decrease with age. A possible explanation is a developmental decrease in action potential duration that results in a reduced Ca2+ influx. The action potential in the CA1 neuron somata undergoes a second week maturation characterized by increased amplitude and shortened duration (Spigelman et al. 1992) that likely reflects an increased density of voltage-gated sodium and potassium channels. We also observed a substantial developmental shortening of the fiber volley duration that possibly reflects a shortening of the axonal (and bouton) action potential. In common with PPF, this shortening was not affected by tetanus toxin-induced activity blockade. However, whether such a shortening should lead to a decreased Ca2+ influx that selectively affects high Pr synapses is unknown. Another possibility is a developmental shift in the type of voltage-gated Ca2+ channel that is involved in the release. N-type voltage-gated Ca2+ channels differ from P/Q and R types in their coupling to release (Qian and Noebels 2001; Urbano et al. 2003), and developmental changes in N-type channels have been described for some synapses (Iwasaki et al. 2000). However, we found no evidence for such a developmental shift in that ω-conotoxin, a N-type channel blocker, did not differentially affect EPSPs at P6 and P12. This lack of developmental shift in N-type channel contribution agrees what has been reported for another cortical glutamate synapse, a thalamo-cortical one (Iwasaki et al. 2000).

Alternatively, a reduction in the amount of Ca2+ reaching the Ca2+ acceptors may relate to an increase in presynaptic buffering capacity of Ca2+ ions in the high Pr group (Blatow et al. 2003; Oleskevich and Walmsley 2002). However, whereas an increase of slow, as well as fast, buffering would lead to a decreased Pr, it would also lead to a decreased PPF (Rozov et al. 2001), and can thus not explain the developmental change in release. Moreover, a fast buffer, if present, is expected to lead to an increase in PP ratio when extracellular calcium levels are increased (Blatow et al. 2003). It was presently noted (Fig. 3B) that, at the lowest Ca2+ levels, P12 synapses did not increase their PPF when Ca2+ was lowered, indicative of a small concentration of fast buffers in these synapses. However, since, for P12 synapses, the overall dominating relation between Pr and PP plasticity is an inverse one, fast buffering is likely to have only a minor effect on Pr and PPF in these synapses.

An alternative explanation to a reduced amount of Ca2+ reaching the Ca2+ acceptors is a developmental change of the distance between Ca2+ channels and the Ca2+ acceptors or of the affinity of Ca2+ acceptors. Such changes would manifest themselves in the same manner as a reduced Ca2+ influx, and if occurring selectively in high Pr synapses, may potentially explain the developmental switch in release properties.

Activity dependence of PPF maturation?

We used in vivo injections of tetanus toxin into the hippocampal CA1 area to address the question of to what extent the second postnatal week synaptic maturation depends on neural activity. Such injections have been found to cause a profound decrease, for several days, of synaptic release over a large part of the hippocampus (Groc et al. 2003), and when applied at P1, to impair dendritic growth (Groc et al. 2002). However, tetanus injection at P6 did not affect the second week increase in PPF, indicating that this increase does not require the neural activity that normally exists during the second postnatal week. This result differs from work on cultured neurons in which NMDA receptor blockade prevents the developmental decrease in Pr (Chavis and Westbrook 2001). However, this latter Pr change differs from the present one, in that synapses changed from high Pr to low Pr, rather than shifting within the high Pr group. Moreover, there may be a difference when reducing activity in different manners (AP5 vs. tetanus toxin). At anteroventral cochlear nucleus (AVCN) glutamate synapses, Pr is, at P11–P16, lower in control than in deaf mice, possibly reflecting an activity dependent lowering of Pr (Oleskevich and Walmsley 2002). However, whether these synapses actually start out with a high Pr is not known. Moreover, the synapses differed only with respect to Pr and not with respect to PPF.

In the striatum, there is a temporal correlation between the developmental increase in PPF and a decreased susceptibility for the induction of an LTD based on Pr reduction (Choi and Lovinger 1997). Neural activity leading to induction of LTD may then cause the developmental increase in PPF. In the hippocampus, LTD may also show age dependence, being more easily induced early on, and be presynaptically expressed (Li et al. 2002; Wasling et al. 2002). Since these LTDs are evoked by rather weak neural activity, it cannot be excluded that tetanus toxin blockade does not interfere with their induction. However, these LTDs do not show the temporal profile of the second week PPF increase, in that they can be induced at a later date given proper induction conditions. They do thus not appear likely candidates for a second week activity-dependent change in release properties. This study also showed that tetanus injection at P1 did not affect the value of PPF at P6. The lack of effect of tetanus injection on the second week PPF increase was thus not masked by an effect of this activity blockade by itself on PPF (and Pr). Previous work on the effect of activity reduction on Pr has indicated that reduced activity, induced either by TTX application on cultured hippocampal neurons (Murthy et al. 2001), or by overexpression of K+ channels at the neuromuscular junction of Drosophila (Paradis et al. 2001), leads to an increase in Pr. This result can thus not be reproduced by tetanus toxin application, possibly either because of a difference in cells involved or of the difference in type of activity blockade.

In conclusion, these results show that the CA3-CA1 synapse undergoes a second postnatal week reduction in Pr, restricted, however, to the minority group of high Pr synapses. This developmental Pr reduction (and PPF increase) seems produced by a reduction in the vesicle release probability, possibly explained by a reduced Ca2+ influx. Such a reduction in Ca2+ influx may be secondary to maturation of the action potential leading to faster time characteristics. Interestingly, postnatal lack of thyroid hormone leads to lower PPF values (Vara et al. 2002) as well as lack of action potential maturation (Potthoff and Dietzel 1997), pointing toward a possible hormonal regulation of the synaptic Pr maturation.


This project was supported by the Swedish Medical Research Council (Project nos. 05180 and 12600) and Göteborgs Läkaresällskap.


We thank Dr. Laurent Groc for help with the tetanus toxin in vivo injections.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract