Differential long-term depression in CA3 but not in dentate gyrus following low-frequency stimulation of the medial perforant path


  • Thomas K. Fung,

    1. Department Physiology and Pharmacology, University of Western Ontario
    Search for more papers by this author
  • Pascal Peloquin,

    1. Department Physiology and Pharmacology, University of Western Ontario
    Search for more papers by this author
  • Kun Wu,

    1. Neuroscience Program, University of Western Ontario, London, Ontario, Canada N6A5C1
    Current affiliation:
    1. Department of Neurosurgery, Yale University School of Medicine, New Haven, CT 06520, USA
    Search for more papers by this author
  • L. Stan Leung

    Corresponding author
    1. Department Physiology and Pharmacology, University of Western Ontario
    2. Neuroscience Program, University of Western Ontario, London, Ontario, Canada N6A5C1
    • Department of Physiology and Pharmacology, The University of Western Ontario, London, ON N6A5C1, Canada
    Search for more papers by this author


Synaptic plasticity may depend not only on the afferent fibers but also on the recipient structure. The medial perforant path (MPP) from the entorhinalcortex projects to both the dentate gyrus (DG) and CA3, resulting in excitatory postsynaptic potentials (EPSPs) in both areas. In this study, we showed that long-term depression (LTD) following low-frequency stimulation of MPP was found only in CA3a, a CA3 subfield, but not in DG. Field potentials were recorded and current source density (CSD) analyzed in CA3a and DG following stimulation of MPP in urethane-anesthetized rats. MPP evoked a short-latency population spike (PS) and EPSP in CA3a, <2.5 ms delayed from the respective events in DG. A small electrolytic lesion of CA3a abolished the locally recorded PS in CA3a but did not affect the responses in the DG. Low-frequency stimulation of the MPP for 600 pulses at 5 Hz, but not at 1 Hz, resulted in LTD of up to 2 h in CA3a but not in DG. High-frequency stimulation (400 Hz bursts) of the MPP resulted in long-term potentiation (LTP) in both CA3a and DG. LTD at CA3a was blocked by a prior intracerebroventricular administration of an N-methyl-D-aspartate receptor (NMDAR) antagonist DL-2-amino-5-phosphonovaleric acid or a nonselective group I/II metabotropic glutamate receptor (mGluR) antagonist (RS)-α-methyl-4-carboxyphenylglycine. We conclude that an NMDAR and mGluR sensitive LTD is induced in CA3 but not in the DG following low-frequency MPP stimulation in vivo, and the bi-directional synaptic plasticity in CA3 may be responsible for its behavioral functions. Synapse, 2011. © 2011 Wiley-Liss, Inc.


A direct projection from the entorhinal cortex (EC), through the perforant path, to the Cornu Ammonis (CA) in addition to the dentate gyrus (DG) has been known for some time (Amaral and Witter,1989; Steward and Scoville,1976). A concept of parallel pathways from EC to DG, CA3 and CA1 (Amaral and Witter,1989; Yeckel and Berger, 1991) replaces that of a sequential, trisynaptic circuit (Andersen et al.,1971). Functional significance of the parallel circuits has been shown by the properties of place cells in the hippocampus. McNaughton et al. (1989) found that the place correlates of CA3 and CA1 cells were unchanged after lesioning more than 75% of the granule cells, and a direct EC to CA3 and CA1 connection was suggested to sustain the functional correlate of the CA cells. Dependence of CA1 place cells on a direct EC input has been demonstrated (Brun et al.,2008), and a direct EC to CA3 pathway may be involved in pattern completion (Nakazawa et al.,2002).

Synaptic plasticity, in particular long-term potentiation (LTP), has been proposed to be a cellular correlate of memory (Bliss et al.,2007). The EC to DG pathway was the first pathway studied for LTP induced by high-frequency stimulation (HFS) (Bliss et al.,2007). Subsequently, LTP of the direct EC to CA3 synapse (Breindl et al.,1994; Do et al.,2002; McMahon and Barrionuevo,2002) and EC to CA1 synapse (Colbert and Levy,1993; Leung et al.,1995; Remondes and Schuman,2002) have also been reported in vivo and in vitro.

Long-term depression (LTD) is a long-lasting decrease in synaptic transmission. It also serves an essential function in memory, since storage of information in a neural network must involve both decrease and increase in synaptic strengths (Brigman et al.,2010; Collingride et al.,2010; Ge et al.,2010; Malenka and Bear,2004). LTD in hippocampal CA1 is proposed to underlie behavioral flexibility in a spatial task (Duffy et al., 2007; Nicholls et al.,2008). LTD was effectively induced by prolonged low-frequency stimulation (LFS) of 1-5 Hz (Doyere et al.,1996; Dudek and Bear,1992) in CA1 or by repeated low-frequency paired-pulse stimulation in CA1 and DG (Thiels et al.,1996). Metabotropic glutamate receptors (mGluRs), in addition to N-methyl-D-aspartate receptors (NMDARs), have been suggested to play a role in LTD in hippocampal CA1 (Kleppisch et al.,2001; Manahan-Vaughan,1997; Neyman and Manahan-Vaughan,2008; Wang et al.,2008). To our knowledge, there has been no study on LTD of the direct EC to CA3 pathway.

In this study, synaptic plasticity of CA3a, a subfield of CA3, and DG was studied by recording field potential responses to medial perforant path (MPP) stimulation, using multichannel electrode arrays. CA3a subfield is selected because of its large population spike (PS) response following single-pulse stimulation of the MPP (Wu and Leung,1998). We reported that LFS of the MPP elicited LTD of the CA3a synaptic responses without significant long-term changes of the MPP to DG responses. We also showed that LTD of the MPP-CA3a responses was blocked by intracerebroventricular (icv) infusion of an NMDAR antagonist or a nonspecific mGluR antagonist.


Forty Long Evans rats (220–450 g) were used. Rats were anesthetized with urethane (1.2–1.5 g/kg i.p.). The recording electrode array was a 16-channel silicon probe angled at 5–10° from the vertical and entered the skull at P 3.2 and L3.5, with respect to bregma, to reach CA3a (Fig. 1A). A glass microelectrode was placed at P3.6, L2.5, at ∼3.8 mm ventral to the skull (V) to record from the hilus of the DG. In later experiments (MCPG and saline icv infusions), a second 16-channel silicon probe was placed in the DG. A stimulating electrode (125 μm Teflon-insulated wire except at the cut tip) was placed in MPP at P8, L4.4 and lowered to V 2.9–3.3 mm (Wu and Leung,1998). In some experiments, a Teflon-insulated wire electrode was used to record from CA3, before and after lesion; lesion was made by passing a DC current of 0.3–0.5 mA for 0.5–1 s. A jeweller's screw in the skull plate over the frontal cortex served as the anode, while another skull screw over the cerebellum served as a recording ground. Cathodal stimulus currents were delivered (with pulse duration of 0.2 ms) through a photo-isolated stimulus isolation unit (PSIU6, Astro-Med/Grass Instrument). Stimulation repetition rate was <0.1 Hz.

Figure 1.

Recording of average evoked potentials (AEP) and current source density (CSD) in CA3a and the dentate gyrus (DG) following medial perforant path (MPP) stimulation at 60 μA. A: Histology of track of the silicon probe through CA3a and schematic of microelectrode at the granule cell layer of the DG. B: AEPs (n = 4 sweeps) following 1st pulse MPP stimulation at electrode 4–12 of the silicon probe in CA3a and at the DG microelectrode. AEPs in CA3a show double population spikes (PSs) at Lines 1 and 2, corresponding to PS at the DG and CA3a, respectively. C: AEPs in CA3a showing paired-pulse facilitation [2nd pulse PS (P2) > 1st pulse PS (P1)] after MPP stimulation; electrode 7 was at the CA3a cell layer. D: Time course of the CSDs in CA3a plotted with the DG-AEP. First PS in the DG (onset 3.6 ms, peak 4.5 ms) occurred slightly earlier than P1 at electrode 7 of CA3a (onset 5.8 ms, peak 6.6 ms). E1 in CA3a had onset at 2.5 ms, DG-AEP onset 2.1 ms. Voltage calibration (1 mV and 10 ms) applies to both Panel C and DG-AEP (Panel D). Filled circle indicates shock artifact. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Silicon recording probes were provided by the NIH Center of Neural Communication Technology, University of Michigan, or purchased from NeuroNexus, Ann Arbor, MI. The probe used had 16 recording sites, each of 177 μm2 recording area, with the centers of the sites spaced 50-μm apart on a vertical shank. The signals were amplified 200–1000× by preamplifier and amplifier and acquired by custom made software using an AC instrument amplifier or Tucker Davis Technologies (Alachua, FL) real-time processor system RA-16 (Leung et al.,2008; Townsend et al.,2002). Single and average (n = 4) sweeps were stored and one-dimensional CSD(z,t) as a function of depth z and time t was calculated by a second-order differencing formula (Leung,1990; Nicholson and Freeman,1975):

equation image(1)

where Φ(z, t) is the potential at depth z and time t, Δz is the spacing (50 μm) between adjacent electrodes on the 16-channel probe. No spatial smoothing of the CSDs was necessary. The conductivity σ was assumed to be constant and the CSDs were reported in units of V/mm2.

CA3a subfield was selected because of its large PS response following stimulation of the MPP (Wu and Leung,1998). CA3a subfield also provided a relatively flat area for penetration of the multichannel silicon probe. The one-dimension CSD analysis above assumes spatially extensive planar layers of sources and sinks that are perpendicular to the mapping electrode (Leung,1990). CSD analysis removed the volume-conducted PS form the DG (Results) and revealed a local PS in CA3a with a soma source and dendritic sink profile (Fig. 1), suggesting that it was generated only by local (CA3a) action currents. CSD analysis also revealed CA3 distal dendritic sinks that corresponded to expected sites of synaptic excitation by the MPP (Wu and Leung,1998). However, in about 15% of the preparations, a distal dendritic sink was not clearly detected. The latter likely happened because MPP excited CA3 over a curved surface, and 1-dimensional CSD may not reliably detect the smaller and more spatially distributed excitatory apical dendritic sink.

Drug or vehicle solution was injected intracerebroventricularly (icv) through a cannula placed in the lateral ventricle (P0.8, L1.4) on the same side as the recording and stimulating electrode(s). NMDAR antagonist DL-2-amino-5-phosphonovaleric acid (APV) (100 nmoles) was dissolved and injected in a volume of 2-4 μl of saline; control experiments used either no injection or the same volume of saline. The dose of APV used blocked basal dendritic LTP in CA1 (Leung and Shen,1999). In another experiment, a nonselective group I/II metabotropic glutamate receptor (mGluR) antagonist (RS)-α-methyl-4-carboxyphenylglycine (MCPG) of 1 μmole was dissolved in 5 μl saline, with a small volume of 1 M NaOH added, and injected icv; control experiments used the same volume of saline of similar pH without MCPG. The icv dose of MCPG used did not induce a significant change in CA1 evoked potentials (Ma and Leung,2002) but blocked late LTD in CA1 (Manahan-Vaughan,1997).

After an experiment, the site of a stimulating electrode was lesioned by passing 0.3 mA current for 1-2 s duration. The rat brain was removed after intracardial perfusion with phosphate buffered saline and 4% formalin, and was later sliced into 40-μm-thick coronal sections. The stimulus and cannula sites and the recording track were identified in sections stained with thionin. The silicon probe track could be readily discerned in the brain sections, even without lesion through its electrode (Fig. 1A).

MPP-evoked excitatory sink (ES) in CA3a was measured as the maximal slope of the sink in a 1-ms interval (E1 after the first pulse in Fig. 1D), during the rising phase and before the onset of a population spike (PS). The PS in CA3a was measured at the site of the maximal PS sink amplitude, which was considered as the CA3a cell layer (Figs. 1C and 1D). The peak amplitude of the PS was measured by the maximal vertical distance of the CSD trace from a tangent line linking the two positive peaks surrounding the PS negative sink. In experiments in which the recording from the DG was derived from a single microelectrode near the granule cell layer, the ES was measured by the maximal (positive) slope of the field potential before the PS, and the PS was measured from the field potential using the tangent line method as described above. In experiments in which a 16-channel silicon probe was placed in the DG, the ES in the DG was measured as the maximal slope of rise (over 1 ms) of the sink at the middle molecular layer sink while the PS was measured from the CSD trace recorded at the granule cell layer of the dorsal blade of the DG. A test stimulus at the MPP was at an intensity that evoked a PS of 50-75% of the maximum at CA3a (typically 100-200 μA) and the DG. The high-frequency stimulation (HFS) or low-frequency stimulation (LFS) was delivered to the MPP at 100-400 μA, after a stable baseline of at least 30 min was established. A HFS consisted of 8 bursts of 12 pulses at 400 Hz, with the burst repeating at 10 s. A LFS consisted of 600 pulses of 1, 3, or 5 Hz pulses. Responses were recorded for at least 2 h following HFS or LFS. In most experiments, a single LFS train was delivered at 5 Hz, and occasionally at 1 and 3 Hz. Some rats received a HFS train at >2 h after LFS. Five rats were given the sequence of HFS followed by LFS, after a 2–3 h interval. Each measure (ES and PS) was normalized by the grand average of the measure before HFS/ LFS. Repeated measures analysis of variance (ANOVA) was used for statistical analysis of the normalized data at different times, and P < 0.05 was considered statistically significant.


Pattern of Excitation in CA3a by the Medial Perforant Path

Typical field potential recordings in CA3a following MPP stimulation at a moderate intensity (50–400 μA) showed two fast transients (line 1 and 2, Figs. 1B and 1C) corresponding PS in the DG (volume-conducted) and CA3a (local), respectively. After CSD analysis (Fig. 1D), the DG volume-conducted PS was removed and a single PS generated by local CA3a neurons remained. Stimulation of MPP also evoked an apical dendritic excitatory sink (E1 following the 1st pulse in Fig. 1D) at short latency (1.5–2.5 ms) in CA3a, preceding the onset of the PS; the onset of the ES in CA3a was ∼1 ms after the onset of population EPSP in the DG (Fig. 1D). The negative PS peak in CA3a was ∼ 2 ms after that in the DG, as shown by the interval between line 1 and line 2 in Figure 1B. In a group of 7 rats given MPP stimulation (214 ± 38 μA intensity), the peak latency of the PS in CA3a was 5.6 ± 0.3 ms (n = 7), which was 1.9 ± 0.2 ms later than that of the PS at the dorsal blade of the DG. The short latency difference in the excitatory sinks (∼1 ms) and PSs (∼2 ms) in DG and CA3a suggests that MPP directly activates CA3a rather than indirectly through the DG (see also Wu and Leung,1998).

To further confirm that the late PS peak (at 5–8 ms latency) recorded in CA3a was locally generated, a single wire electrode was lowered into CA3a for both recording and lesion. After passing a small DC current of 0.3 mA for 0.5 s through the CA3a recording electrode (Fig. 2B), the PS in CA3a was abolished, while the field response at the DG was suppressed transiently but recovered after 5 min (Fig. 2A). In a group of 5 rats subjected to CA3a lesion, the postlesion responses after 5 min showed a significant (P < 0.05, paired Wilcoxon test) reduction of the PS in CA3a to 5 ± 3% (n = 5) of the baseline, and a nonsignificant decrease of the DG peak response at 82 ± 10% of the baseline (n = 5).

Figure 2.

Representative lesion of CA3a abolished the local response. A: Average evoked potentials (n = 4 sweeps) at CA3a and dentate gyrus (DG) were recorded during baseline before (−5 min), 1 min and 5 min after lesion of CA3a. DG recording was made at the granule cell layer. Lesion was made by passing a DC current of 0.3 mA for 0.5 s through the CA3a electrode. B: Subsequent histological section stained with thionin shows a discrete lesion (arrow) at the pyramidal cell layer of CA3a. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

LTD in CA3a and the DG

In initial experiments, after MPP-evoked responses were stable for 30 min, long-term plasticity in CA3a and the DG were assessed following LFS (600 pulses) of 1, 3, and 5 Hz. No significant change in the PS (or ES) in the DG was found after LFS of any frequency, or in CA3a after 1-Hz LFS (Table I). LFS at 3 or 5 Hz resulted in depression at short-term (<60 min) and long-term (60–120 min), respectively. In the following, properties of the LTD in CA3s were studied using 5-Hz LFS.

Table I. Comparison of normalized amplitude of the population spike recorded at the CA3 and dentate gyrus, at different time periods following low-frequency stimulation (LFS) at 1, 3 or 5 Hz
Inducing stimulusRecording locationN10–30 min30–60 min60–90 min90–120 min
  1. Asterisk (*) indicates statistical significance (P < 0.05), post hoc Newman Keuls test, of at least one time point within time period. The average baseline population spike in each area is normalized to one.

NoneCA331.02 ± 0.041.00 ± 0.041.01 ± 0.061.03 ± 0.04
 DG41.09 ± 0.101.10 ± 0.111.10 ± 0.121.09 ± 0.12
1 HzCA341.16 ± 0.101.06 ± 0.130.98 ± 0.060.86 ± 0.05
 DG51.05 ± 0.121.02 ± 0.140.97 ± 0.100.89 ± 0.12
3 HzCA340.75 ± 0.16*0.83 ± 0.090.78 ± 0.070.60 ± 0.11*
 DG40.96 ± 0.090.94 ± 0.091.00 ± 0.061.04 ± 0.04
5 HzCA340.28 ± 0.08*0.45 ± 0.11*0.71 ± 0.12*1.09 ± 0.11
 DG41.00 ± 0.060.97 ± 0.041.03 ± 0.061.08 ± 0.05

In CA3a, LFS of 5 Hz resulted in an immediate decrease of ES and near complete abolition of PS for <10 min, followed by LTD of both ES and PS (Fig. 3A, 3C, and 3D). In a group of six rats, the PS recorded in CA3a was elicited by an average MPP stimulus intensity of 108 ± 30 μA, with the LTD of the PS and the ES for the 60–120 min duration averaging 0.71 ± 0.7 and 0.83 ± 0.06 (n = 6) times the baseline, respectively. The PS peak latency during baseline (before LFS) was 6.70 ± 0.19 ms (n = 6) and it increased by 0.27 ± 0.08 ms (n = 6) at 60–120 min after LFS. During baseline, paired-pulse facilitation (PPF) at an interpulse interval of 50 ms was observed for the ES in most rats and for the PS in all rats. Baseline E2/E1 was 1.17 ± 0.1 (n = 6) and P2/P1 was 1.52 ± 0.1 (n = 6). LFS did not significantly change the E2/E1 ratio, which measured 1.11 ± 0.1 after LFS. However, P2/P1 was significantly increased to 1.8 ± 0.15 (n = 6) after LFS, on account of a decreased P1 and similar P2 after LFS (Fig. 3B).

Figure 3.

Representative rat showing long-term depression (LTD) in CA3a induced by low frequency stimulation (LFS) of the medial perforant path (MPP). A: Average (n = 8 sweeps) of the evoked potential in dentate gyrus (DG) and of the current source density (CSD) trace recorded at the CA3a cell layer following 100 μA MPP stimulation. The recording at 5, 30, and 120 min after LFS was overlaid on top of the baseline recording (−5 min, dotted trace) before LFS. Filled circles indicate shock artifacts. B: Overlaid CSD trace before (−5 min) and 120 min after LFS, showing electrode 10 (presumed CA3a cell layer), 11 and 12 of the silicon probe, with excitatory sink E1 and E2 following 1st and 2nd pulse, and population spike P1 and P2 following 1st and 2nd pulse, respectively. C,D:Time plot of normalized measures showing LTD in CA3a induced by LFS, followed 170 min later by high frequency stimulation (HFS) that induced potentiation in CA3a and the DG. LFS was 5 Hz stimulation for 2 min, and HFS was eight bursts of 12 pulses at 400 Hz with interburst interval at 12 s; MPP stimulus intensity was 400 μA. Test MPP stimulus intensity was 100 μA. C: Time plots of rising slope of the excitatory sink in CA3a (E1 in Figure 3B) and the slope of the population excitatory postsynaptic potential (pEPSP) in DG show LFS-induced LTD in CA3a but not the DG. A subsequent HFS-induced potentiation in both CA3a and the DG. D: LTD was found for the population spike amplitude (P1 in Figure 3A) in CA3a but not in the DG. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

At the DG, LFS induced a short-term depression of the PS that recovered to near baseline values in <30 min (Figs. 3C and 3D and 4), but there was no statistically significant long-lasting change of ES or PS for the group at any given time after 30 min (one-factor ANOVA, n = 14 rats combining no-injection and icv saline infusion control rats). The example illustrated shows a late LTP (>90 min) in DG, as measured by either ES or PS (Figs. 3C and 3D).

Figure 4.

High-frequency stimulation (HFS) resulted in long-term potentiation in CA3a and dentate gyrus (DG) whereas low-frequency stimulation (LFS) resulted in long-term depression in CA3a only. Normalized population spike amplitude in CA3a (A) and DG (C) following HFS and then, 140 min later, LFS of the medial perforant path. HFS was eight bursts of 10 pulses at 400 Hz, and LFS was 2 min (600 pulses) at 5 Hz. HFS was given at time zero, while LFS was aligned to occur at 140 min (actual time, 148–178 min in five rats). n = 5 rats for CA3a, n = 4 after HFS in DG, and n = 3 after LFS in DG. B: Population spike in CA3a normalized for the last 30 min before LFS, in a group preceded by HFS and LTP (n = 5) and another group where LFS was given alone without preceding HFS (n = 7). D: Same as B except for population spike in DG; LFS only (n = 7), LFS after HFS (n = 3). There was no difference in the magnitude of the response at CA3a or DG after LFS whether it was preceded by HFS or not. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Responses in CA3a and DG were recorded during 5-Hz LFS. PSs in CA3a and DG both fluctuated in amplitude during LFS, above and below the baseline values, and the responses in CA3a was particularly labile. The PS in CA3a varied from no response, decreased or increased early-latency (<10 ms) PS, and occasional occurrence of a late-latency PS that suggests a disynaptic DG to CA3a response, with >5 ms delay from the PS in DG.

Bidirectional Synaptic Plasticity in CA3a

In some experiments, HFS was first given to the MPP to elicit LTP, followed 145–178 min later by LFS. HFS was found to elicit LTP that persisted for >140 min at both CA3a (Fig. 4A) and DG (Fig. 4C). The average LTP at 110–140 min was 1.6 ± 0.28 (n = 5) in CA3a and 1.2 ± 0.08 (n = 4) in the DG. LTP of the ES in CA3a and DG was also observed (data not shown). A subsequent 5-Hz LFS induced LTD in CA3a (Fig. 4B) and not in the DG (Fig. 4D). LTD and not depotentiation is suggested in CA3a since the response decreased below the baseline (Fig. 4A). LTD was also induced by the 5-Hz LFS alone, without a preceding LTP (Fig. 4B). The magnitude and duration of LTD were comparable with or without a preceding LTP, when normalized by the average response immediately (5–30 min) before LFS (Fig. 4B). Similarly, LTP induced by HFS was observed after LTD. HFS, delivered 140 min after LTD induced by LFS, elicited potentiation in both CA3a and DG (Figs. 3C and 3D), with an average potentiation magnitude similar to that elicited by HFS alone. Only 5-Hz LFS was delivered to the following experiments.

LTD in CA3a was Blocked by NMDAR Antagonist APV

Infusion of APV icv did not significantly change the evoked responses in CA3a or DG. However, APV infusion suppressed the effect of 5-Hz LFS on ES and PS in CA3a (Fig. 5). When compared to a no-infusion control group, two-factor repeated measures ANOVA revealed a significant main (drug) effect on the PS magnitude in CA3a [F(1,11) = 29.8, P < 0.0002, five APV, and seven control rats] and a significant drug × time interaction [F(19, 209) = 2.17, P < 0.005]. Posthoc comparisons indicated that differences between APV and control groups started at 15–20 min post-LFS and extended to 110 min post-LFS (Fig. 5A). In the presence of APV, PS peak latency was not significantly decreased after LFS, and the decrease averaged 0.2 ± 0.07 ms (n = 5) at 60–120 min after LFS. There was also a significant effect of APV vs. saline infusion on ES in CA3a after LFS [F (1,9) = 9.35, P < 0.02; two-factor repeated measures ANOVA, five APV and six control rats, rats without a clear ES were excluded]. Similar results were obtained when the APV icv group was compared to a saline icv group.

Figure 5.

Group data of the CA3a and DG population spike (PS) amplitudes after low-frequency stimulation (LFS) of the medial perforant path, in rats infused intracerebroventricularly with NMDA receptor antagonist D,L-APV (n = 5) or in no-injection control rats (n = 8). In each rat, PS amplitude was normalized by the average baseline (−30 min to 0 min). A: LFS resulted in an immediate depression of the CA3a PS in both control and APV-infused rats. However, APV-infused rats showed no LTD at >30 min, compared with control rats. B: The population spike in the DG showed similar response to LFS in both APV (n = 5) and control rats (n = 8). *P < 0.05, Newman-Keuls posthoc test difference between APV and control groups. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

APV had no significant effect on the DG responses after LFS. No significant drug effect was found for the PS in DG after LFS [Fig. 5B; repeated measures ANOVA F(1,10) = 0.13, P > 0.7, five APV and seven control rats]. Similarly, APV had no significant effect on ES in the DG after LFS [F(1,10) = 3.77, P < 0.08].

LTD in CA3a Was Blocked by mGluR Receptor Antagonist MCPG

Infusion of MCPG icv did not significantly change the evoked responses in CA3a or DG. LFS delivered to the MPP resulted in little LTD (at >30 min) in CA3a if MCPG icv was administered, as compared to pH-controlled saline icv (Figs. 6A and B). The short-term depression immediately after the LFS was not significantly different between saline and MCPG groups. Two-factor repeated measures of ANOVA showed a significant drug (MCPG vs. saline) effect on the PS magnitude in CA3a [F(1,12) = 6.06, P < 0.03, eight control and six MCPG rats] and a significant drug × time interaction effect [F(23, 276) = 2.25, P < 0.002; Fig. 6B]. In addition, MCPG, as compared with saline icv, diminished the LTD of the ES in CA3a in a time-dependent manner. Repeated measures ANOVA revealed no significant group effect on ES [F(1,12) = 3.06, P > 0.11, seven control and five MCPG rats] but a significant group × time interaction [F(17, 187) = 3.26, P < 0.0001] with post hoc differences only during the period of 45–50 min after LFS.

Figure 6.

Attenuated long-term depression (LTD) of the population spike (PS) in CA3a following low-frequency stimulation (LFS) of the medial perforant path in rats infused intracerebroventricularly (icv) with a nonspecific mGluR antagonist MCPG as compared with saline. A.: Representative current source density traces recorded at the CA3a cell layer after icv infusion of either MCPG or saline (SAL), with a trace at 30, 60, and 120 min after 5-Hz LFS of the medial perforant path overlaid with the baseline (−5 min) trace; each trace was an average of four sweeps. B: Group data of PS in CA3a in rats given saline (n = 8) or MCPG (n = 6) and then LFS. In each rat, PS was normalized by the average baseline response (−30 min to 0 min). LFS resulted in an immediate depression of the PS in both saline- and MCPG-infused rats, but MCPG-infused rats showed smaller LTD (at time >30 min) compared with saline infused rats. *P < 0.05, Newman-Keuls posthoc test difference between MCPG and saline groups. C: Group data for the PS in DG show statistically similar response after LFS in both MCPG-infused (n = 6) and saline-infused rats (n = 6). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

There was no MCPG vs. saline effect on the PS magnitude in the DG after LFS (Fig. 6C), as confirmed by repeated measures ANOVA [F(1,10) = 1.17, P > 0.3, six control and six MCPG rats]. There was also no MCPG vs. saline effect on the ES in DG after LFS [data not shown; repeated measures ANOVA F(1, 9) = 0.16, P > 0.6; six control and five MCPG rats].


This study reported original results that LTD of the direct EC to CA3a synapse, but not the EC to DG synapse, was induced by 5-Hz stimulation of the MPP. HFS of the MPP elicited LTP in both CA3a and DG. The LFS-induced LTD in CA3a was blocked by icv infusion of either an NMDAR antagonist APV or a nonspecific Group I/II mGluR antagonist MCPG.

The MPP fibers arise from stellate cells in Layer II of the EC and project to CA3 and DG of the hippocampus (Amaral and Witter,1989; Steward and Scoville,1976; Tamamaki and Nojyo,1993). We showed that MPP stimulation activated an EPSP in CA3 ∼1 ms after the EPSP in the DG (see also Wu and Leung,1998). Tamamaki and Nojyo (1993) showed that single Layer II neurons in the EC typically project through the perforant path first to DG before CA3 and CA2 (CA3a). Since the distance between DG and CA3a was 2–3 mm, a synaptic delay of 1 ms was consistent with a conduction velocity of ∼2 m/s for the MPP within the hippocampus. The ∼2 ms interval between a PS in CA3a and that in the DG is not sufficient for a direct excitation of CA3a by DG granule cells. Direct stimulation of the mossy fibers 1.5 mm away from a CA3a recording site resulted in a mossy-fiber evoked EPSP latency of >2.5 ms and PS latency of >5 ms (n = 5 rats, unpublished data), which is consistent with the data of Yeckel and Berger (1998) and a conduction velocity of the mossy fibers of 0.67 m/s estimated in vitro (Langdon et al.,1993).

LTP of the direct EC to CA3 synapse has been reported in anesthetized and behaving rats in vivo (Breindl et al.,1994; Do et al.,2002) and in hippocampal slices in vitro (McMahon and Barrionuevo,2002). The present study extended the previous studies, and reported a stable LTP of the EC to CA3 and the EC to DG pathway for at least 140 min after HFS (Fig. 4).

We showed that LTD of the ES and PS in CA3a was induced by 5-Hz LFS of the MPP. During LTD, ES and PS magnitudes were decreased and the PS onset and peak latencies were increased as compared to baseline (Fig. 3). The lack of a significant change of PPF of the ES (E2/E1 ratio) suggests a lack of presynaptic plasticity. The increase of PPF of the PS (P2/P1 ratio) is likely caused by a decrease in P1, i.e., a smaller P1 activated less feedback inhibition and allowed a larger P2 response (Leung et al.,2008).

LTD of the direct EC to CA3 synapses, in vivo or in vitro, has not been reported before. At the CA3 to CA1 apical dendritic synapse, 900 pulses (15 min) of 1-Hz stimulation induced LTD in CA1 in immature rats in vitro (Dudek and Bear,1992) and in behaving rats in vivo (Manahan-Vaughan,1997), although one study reported that LFS of 1-5 Hz was unable to induce LTD in anesthetized rats in vivo (Errington et al.,1995). Another procedure of low-frequency paired-pulse stimulation (25 ms interpulse interval) was found to more reliably induce LTD of the CA3 to CA1 synapse in anesthetized rats in vivo (Doyere et al.,1996; Thiels et al., 1994) or in hippocampal slices in vitro (Kleppisch et al.,2001). Dvorak-Carbone and Schuman (1999) reported that prolonged 1-Hz stimulation of the direct EC to CA1 synapses in hippocampal slices in vitro readily induced an NMDAR-dependent LTD, and a LTP induced by HFS if GABAA receptors were blocked (see also Colbert and Levy,1993). In the present study, the direct EC to CA3 synapse in vivo showed bi-directional plasticity without GABAA receptor blockade, but LTD could only be induced at LFS of 3 or 5 Hz.

LFS of the MPP at 1 Hz (900 pulses) was reported to induce LTD of the PS in the DG in some studies (e.g., Altinbilek and Manahan-Vaughan,2009) but not in others (e.g., Errington et al.,1995). The difference in LTD in the DG in different laboratories may partly depend on the rat's strain and other experimental conditions. The sensitivity of LTD in the DG in behaving rats was demonstrated by the facilitation of this LTD by large orientational or environmental cues (Kemp and Manahan-Vaughan,2007). In this study, LFS at 1-5 Hz did not induce a significant LTD in the DG, whether the ES or the PS was measured. HFS stimulation of the MPP induced LTP of the ES and PS in the DG, as was expected from previous studies (Bliss et al.,2007; Leung et al.,1995; Manahan-Vaughan,1997).

Infusion of an NMDAR antagonist APV icv blocked the LTD induced by LFS at the EC-CA3a synapse. An NMDAR antagonist also blocked the LTD induced by LFS at the CA3-CA1 synapse (Dudek and Bear,1992; Fox et al.,2006; Liu et al.,2004; Mulkey and Malenka,1992) and at the EC-CA1 synapse in vitro (Dvorak-Carbone and Schuman,1999). Ca2+ influx through activated NMDARs is necessary for LTD in CA1 neurons (Makenka and Bear, 2004; Mulkey and Malenka,1992), and presumably also for LTD at the EC-CA3 synapse.

The late LTD (>30 min) at the EC-CA3a synapse was also blocked by a nonspecific Group I/II mGluR antagonist MCPG. MCPG was shown to block the late LTD at the CA3-CA1 synapses induced by 1-Hz LFS or by low-frequency paired-pulse stimulation (Kleppisch et al.,2001; Manahan-Vaughan,1997). Activation of group I mGluRs, which are located postsynaptically and coupled positively to phospholipase C via Gq proteins, causes release of intracellular Ca2+ through stimulation of inositol triphosphate and diacylglyercol (Anwyl,1999; Conn and Pin,1997). Activation of Group II mGluRs, typically present at presynaptic sites, negatively affect adenylyl cyclase and presumably reduces presynaptic release via protein kinase A inhibition (Altinbilek and Manahan-Vaughan,2009; Santschi et al.,2006). Both Groups I and II mGluRs (Manahan-Vaughan,1997; Neyman and Manahan-Vaughan,2008; Santschi et al.,2006; Wang et al.,2008) have been implicated in LTD of the CA3-CA1 and MPP-DG synapses. The subtype(s) of mGluRs responsible for LTD at the MPP to CA3 synapse remains to be studied.

CA3 receives directly from the entorhinal cortex via the MPP and indirectly through MPP excitation of the DG, which projects to the proximal apical dendrites of CA3 via the mossy fibers. In agreement with Yeckel and Berger (1998), LFS of 5 Hz or less induced mainly a monosynaptic excitatory response in CA3a without large participation of indirect (polysynaptic) excitation of the CA3a through the DG. However, both direct and indirect excitation may participate in the activation of NMDARs and mGluRs during LTD induction by LFS. Since CA3 is excited physiologically by both direct and indirect pathways in vivo, the induction of synaptic plasticity with the participation of both pathways may be physiologically relevant as well.

LTD and Memory Acquisition

Hippocampal LTD has been associated with spatial memory (Bliss et al.,2007; Malenka and Bear,2004; Nakao et al.,2002; Wong et al.,2007), novelty and environmental cues (Kemp and Manahan-Vaughan,2004,2007). Altinbilek and Manahan-Vaughan (2009) showed that repeated treatment with a group II mGluR antagonist resulted in impairment of both spatial reference memory and LFS-induced LTD in the DG. LTD in CA1 is closely related to behavioral flexibility in a cognitive task (Brigman et al.,2010; Collingride et al.,2010; Duffy et al., 2007; He et al., 2010; Nicholls et al.,2008). Various manipulations of NMDAR function in CA3 disrupted behavioral performance. Selective knockout of NMDAR subunit 1 in CA3 of mice resulted in deficits of rapid formation of memory representations and pattern completion (Nakazawa et al.,2002,2003), and APV locally injected into CA3 disrupted novelty detection (Kesner,2007). These behavioral disruptions may involve the NMDAR-dependent LTD at the MPP to CA3 synapse.

As reported here, the synaptic plasticity of the same afferents (MPP) projecting to two different areas—CA3 and DG—is different. Differential LTP sensitivity at different targets of the same afferent system was reported for the collaterals of CA3 to the lateral septum and CA1 (McNaughton and Miller,1986). In this study, we also showed that CA3 was capable of bi-directional plasticity, i.e., both LTD and LTP, while the DG was only inclined towards LTP. The bidirectional synaptic plasticity would endow CA3 neurons with an increased sensitivity to environmental signals. Both theoretical and experimental evidence suggest that place fields of hippocampal including CA3 neurons depend on synaptic plasticity (Dragoi et al.,2003; Muller et al.,1996). While the current theory of place field formation centers on LTP, LTD should also be a powerful process to alter place fields and other types of adaptive networks. The 5-Hz frequency used for LFS is in the lower range of theta frequency of the rat during waking behaviors or rapid-eye-movement sleep (Bland,1986; Leung,1984), and whether a 5-Hz theta rhythm relates to the physiological induction of LTD remains to be studied.