Corresponding author S. Otani: Laboratoire de Neurobiologie et Neuropharmacologie du Développement, Institut des Neurosciences, Université de Paris VI, 7 Quai Saint Bernard, Paris 75005, France. Email: email@example.com
1During block of γ-aminobutyric acid-A-mediated inhibition, low-frequency stimulation (2 Hz, 900 pulses) to Schaffer collateral-CA1 neuron synapses of adult rat hippocampus induced an N-methyl-D-aspartate receptor-independent, postsynaptic Ca2+-dependent depression of synaptic strength (long-term depression; LTD).
2Ratio imaging with fura-2 revealed moderate dendritic [Ca2+] increases (≈500 nM) during only the initial ≈30 s of the 7.5 min stimulation period. Conditioning for 30 s was, however, insufficient to induce LTD.
3The [Ca2+] changes were insensitive to the metabotropic glutamate receptor (mGluR) antagonist (+)-α-methyl-4-carboxyphenylglycine (MCPG). MCPG, however, completely blocked LTD when present during conditioning.
4The [Ca2+] changes were abolished by postsynaptic hyperpolarization (-110 mV at the soma). Hyperpolarizing neurons to -110 mV during conditioning significantly attenuated LTD induction.
5LTD induction was also blocked by the postsynaptic presence of the protein kinase C inhibitor peptide PKC(19-36).
6These results suggest that LTD induction in adult hippocampus by prolonged low-frequency stimulation depends on both a rapid Ca2+ influx through voltage-sensitive channels and synaptic stimulation of mGluRs which may be coupled to phospholipase C.
We have recently reported that low-frequency stimulation applied asynchronously to two separate sets of synapses on the same CA1 pyramidal neuron induces LTD of these synapses in adult hippocampus (Otani & Connor, 1995). Induction of this LTD requires activation of both N-methyl-D-aspartate (NMDA) and metabotropic glutamate (mGlu) receptors, postsynaptic Ca2+ increases, and nitric oxide (NO) synthesis. Although more detailed mechanisms of this LTD have not been investigated, the asynchronous stimulation may generate a sufficient postsynaptic excitation without exceeding a LTD-inducible input frequency at one set of synapses (<10 Hz; Dudek & Bear, 1992). This notion predicts that in adult CA1, a pharmacological manipulation that enhances postsynaptic excitation lowers the threshold for LTD induction in single-pathway condition. In fact, 2 Hz single-pathway stimulation can induce LTD if it is combined with exposure to the mGluR agonist (1S, 3R/1S, 3S)-1-aminocyclopentane-1,3-dicarboxylic acid (ACPD) (Otani & Connor, 1995). Furthermore, 1 or 2 Hz single-pathway stimulation induces LTD in the condition where γ-aminobutyric acid-A (GABAA)-mediated synaptic inhibition is blocked (Kerr & Abraham, 1995; Wagner & Alger, 1995; Otani & Connor, 1996a). The second protocol points to the significant role of the balance between synaptic excitation and inhibition in the generation of LTD, as has been long proposed for long-term potentiation (LTP) induction (Wigström & Gustafsson, 1983; Artola et al. 1990).
In the present study, we have measured Ca2+ changes by fura-2 ratio imaging simultaneously combined with intracellular electrical recording to investigate the mechanisms underlying LTD induction during block of GABAA synaptic inhibition in the adult hippocampus. Combining these techniques with pharmacological manipulations, we demonstrate a unique co-operativity between moderate dendritic Ca2+ entry and mGluR activation in the induction of LTD. Some of the results have appeared in the form of a short communication (Otani & Connor, 1996b).
Male adult Sprague-Dawley rats (180-220 g, ∼50 days old) were anaesthetized by intramuscular injection of ketamine (40 mg kg−1) and xylazine (5 mg kg−1) and decapitated. The brain was rapidly removed from the skull, and the right hippocampus was dissected out in fully oxygenated (95 % O2-5 % CO2) artificial cerebrospinal fluid (ACSF) of the following composition (mM): 124 NaCl, 3 KCl, 26 NaHCO3, 2 CaCl2, 1 MgSO4, 10 D-glucose, and 1.25 NaH2PO4. Using a vibratome, transverse slices (400 μm) were prepared from the dorsal-mid portion of the hippocampus in chilled oxygenated ACSF. The slices were allowed to recover for at least 2 h at room temperature (∼20°C). A slice was then transferred to an interface recording chamber and continuously perfused with ACSF (1 ml min−1) at 32.5 ± 0.3°C. The CA1 region was separated from CA3 by a knife-cut to the Schaffer collaterals to reduce spontaneous presynaptic activity.
Stable intracellular recordings were made from over seventy CA1 pyramidal neurons using standard glass microelectrodes filled with 3 M potassium acetate (80-100 MΩ tip resistance). The mean resting membrane potential of the cells was -67 ± 0.8 mV, with an input resistance of 48 ± 1.0 MΩ. A spike amplitude of at least 85 mV was required for acceptable data. The membrane potential was held between -70 and -75 mV (-73 ± 0.5 mV on average) by negative current injection using a Neurodata amplifier (IR 283; New York) and monitored with a digital voltage meter (Precision 280). Monosynaptic excitatory postsynaptic potentials (EPSPs) of about 10 mV amplitude were evoked orthodromically by bipolar stimulating electrodes (25 μm twisted tungsten wires; monophasic square pulses of 150 μs duration) placed on the Schaffer collateral pathway. The responses were digitized at 4 kHz and stored on an IBM computer (model 5170; Armonk, NY, USA), using a TL-1 DMA interface (Axon Instruments) and the pCLAMP acquisition program (Axon Instruments).
During experiments, test EPSPs were evoked at 0.017 Hz. Standard LTD-inducing, low-frequency stimulation consisted of 900 pulses delivered at 2 Hz. Test stimulation was resumed 30 s after the end of 2 Hz stimulation. Only neurons in which spike height, input resistance and resting potential remained within 10 % of initial values were included in the analysis. Responses were averaged for each 5 min period, and changes were expressed as a percentage increase/decrease of the EPSP slope (mV per fixed time) from baseline level (a 5 min period just prior to conditioning). Student's two-tailed t test or analysis of variance (ANOVA) repeated measures was used for statistical analysis, with the level P < 0.05 considered as significant. All data are expressed as means ±s.e.m.
For the measurement of Ca2+ concentration ([Ca2+]), cells were impaled with microelectrodes initially containing 20 mM fura-2 (Molecular Probes) in the tip and 3 M potassium acetate in the barrel. After penetration, hyperpolarizing current (0.2-0.4 nA) was applied for 10–20 min to eject fura-2 which then diffused throughout the neuron. During this time, mixing of fura-2 and potassium acetate within the electrode also occurred, lowering the resistance from several hundred megaohms to ∼100 MΩ. Neurons near the top surface of the slice were examined for epifluorescence, using an upright microscope (Zeiss Axioskop, Thornburg, NY, USA) and a long-distance × 20 dry objective lens (Zeiss) in the interface configuration. A CCD camera system (Photometrics, Tucson, AZ, USA) was used in the frame transfer mode to acquire image pairs at 350 and 380 nm excitation wavelengths (100-150 ms exposure time). An image pair was recorded approximately every 60 s before and after 2 Hz conditioning stimuli and every 10–20 s during the 7.5 min conditioning protocol. Once the occurrence of consistent rises in [Ca2+] shortly after the onset of conditioning stimuli was established, the acquisition rate was accelerated up to 2–4 Hz during the first 20–30 s of conditioning. For imaging during 100 Hz stimuli, a logic signal from the camera controller served as a trigger to synchronize image acquisition and the delivery of the stimuli. [Ca2+] was determined from background corrected image pairs using the ratio method (Grynkiewicz et al. 1985).
All drugs were applied in the bathing medium, except BAPTA (Molecular Probes) and the protein kinase C inhibitor peptide PKC(19-36) (Gibco BRL), which were loaded into cells via the recording electrode (20 mM BAPTA and 250 μm PKC inhibitor in electrode). Other drugs used were dl-2-amino-5-phosphonovaleric acid (APV, Sigma), (+)-α-methyl-4-carboxyphenylglycine (MCPG, Tocris Cookson) and picrotoxin (Sigma).
LTD induction in adult hippocampus during block of GABAA-mediated inhibition
The upper left inset of Fig. 1 illustrates the experimental configuration employed for electrical measurements. In the presence of the GABAA antagonist picrotoxin (50 μm), which was included in the bathing medium throughout these experiments, prolonged 2 Hz stimulation (900 pulses, 7.5 min) induced LTD of the slope of the EPSP in the stimulated pathway (▪, -32 ± 12 %, n= 7 at 45 min, P < 0.03) (also see Otani & Connor, 1996a). After the first few pulses of the conditioning stimulus, the EPSPs typically elicited double action potentials (see Fig. 4A, left inset). This firing persisted for 20–30 s, after which the EPSPs again decreased to subthreshold levels in the majority of cases (7/10, including three experiments in which recordings were lost before 45 min post conditioning but the cells were expressing depression). In the remaining three cells, firing towards the end of the conditioning period was only sporadic and consisted of single action potentials. In the absence of picrotoxin, the same 2 Hz stimulation did not induce LTD (2.8 ± 4.7 % at 45 min, n= 6, Fig. 1, □). In this case, the conditioning stimulus still produced action potentials in four cells, but the firing period was shorter than in the presence of picrotoxin and/or the stimuli did not trigger multiple action potentials. Figure 1 also shows that LTD was still inducible in the presence of APV (100 μm, present throughout the experiments; •), suggesting that activation of NMDA receptors is not necessary in this induction protocol (-36 ± 6.3 % at 45 min, n= 5, P < 0.02). In the presence of APV, relatively more of the depression appeared to develop over a longer time course, during the post-conditioning 0.017 Hz test pulses.
mGluR activation and postsynaptic Ca2+ increases are critical steps in LTD induction in our protocol, consistent with several other reports (Mulkey & Malenka, 1992; Bashir et al. 1993; Bolshakov & Siegelbaum, 1994). First, the mGluR antagonist MCPG (0.3-0.5 mM), which was applied in the bath during conditioning, completely blocked LTD (14 ± 12 % at 45 min, n= 6, P < 0.03, Fig. 2, ▪). Second, the postsynaptic presence of the calcium chelator BAPTA also blocked LTD (7.7 ± 4.2 %, P < 0.05, n= 5; Fig. 2, □). Injection of BAPTA from the recording electrode was facilitated by applying a 0.2-0.4 nA hyperpolarizing current for approximately 15 min prior to experiments. These sets of data also illustrate that the LTD, when observed, is not due to a systematic change in the presynaptic pathways (cf. McNaughton et al. 1994).
[Ca2+] transients during conditioning stimuli
Figure 3 shows intracellular [Ca2+] increases measured by fura-2 ratio imaging for one representative neuron during 2 Hz stimuli in control medium and in the medium containing 50 μm picrotoxin and 100 μm APV, as in Fig. 1. [Ca2+] increases measured during 2 Hz stimuli in control medium (upper panels) were negligibly small even though the recording/injection electrode contained a relatively high Ca2+ level and could be expected to supply a steady Ca2+ input to the soma. In contrast, with picrotoxin present (lower panels), there were [Ca2+] increases in both the dendrites and the cell body during the initial period (< 30 s) of the conditioning stimulation, i.e. during the more or less regular action potential firing. During this low-frequncy firing in the presence of APV, Ca2+ increases were largest in the primary and the proximal secondary dendrites (see also Regehr et al. 1989).
Figure 4A summarizes the increases in [Ca2+] in the medial/distal primary or the proximal secondary dendrites during 2 Hz conditioning. Expressed as a percentage increase over resting levels, the picrotoxin-treated group (LTD condition, Fig. 1) showed a transient increase in [Ca2+] during initial periods (∼30 s) of conditioning stimulation (n= 7). The peak increase was 244 ± 84 % at 10 s after stimulus onset (P < 0.01 over control at the same time point). The left inset shows an example of the EPSP and action potential firing during the initial periods of conditioning in picrotoxin. Without picrotoxin, [Ca2+] increases were very small (35 ± 13 % at peak, n= 9). The presence of the NMDA antagonist APV (100 μm) had only a small effect on the [Ca2+] rise (198 ± 62 %, P < 0.03 over control and P > 0.1 over the ‘picrotoxin only’ group, n= 13), consistent with its lack of effect on LTD induction (Fig. 1). Simultaneous electrical recordings were made in twelve of these fura-injected neurons: nine neurons were treated with picrotoxin (5 were also treated with APV) and eight neurons were examined under control conditions (4 with APV). Conditioning stimuli induced LTD in six out of the nine picrotoxin-treated neurons (on average -39 ± 10 % at 20–30 min, n= 9). In contrast, in only one experiment out of eight was a reduction of synaptic strength in control medium demonstrated. Thus, the [Ca2+] increases observed in the presence of fura-2 are within the range sufficient for LTD induction. Figure 4A clearly shows that mean Ca2+ elevations lasted for only the initial 30 s or so of the induction stimulation. If the 2 Hz stimulation was stopped at 30 s, no LTD was induced (n= 5, right inset of Fig. 4A). This result suggests that other processes besides the Ca2+ increases are necessary for LTD (see below).
Figure 4B shows that MCPG (0.3-0.5 mM) which blocked LTD (Fig. 2) had no effect on the transient increase in [Ca2+] (n= 4). In contrast, hyperpolarization of the neuronal soma to -110 mV completely blocked [Ca2+] increases during conditioning stimulation (n= 8, Fig. 4B). Taken together, these two findings suggest that the observed [Ca2+] increases are the result of voltage-gated influx rather than intracellular Ca2+ release triggered by mGluR activation. The effect of postsynaptic hyperpolarization on LTD was also tested in another set of neurons without injected fura-2 (n= 5, Fig. 5). Although some late decreases in the EPSP were seen (-15 ± 6.3 % at 45 min), the overall change in the EPSP that occurred during 45 min post conditioning was significantly smaller than the LTD induced under the normal condition (P < 0.01, ANOVA). This result confirms that postsynaptic depolarization plays a critical role in LTD induction.
Figure 6 shows the absolute peak levels of [Ca2+] reached in primary and secondary dendrites during three different stimulation protocols; 2 Hz for 30 s with no picrotoxin, 2 Hz for 30 s with picrotoxin, and 100 Hz for 1 s with picrotoxin. The latter stimulus is often used to induce LTP at these synapses. It is clear that Ca2+ levels reached during the LTP protocol are much greater than those involved in LTD (see also Yasuda & Tsumoto, 1996; Hansel et al. 1997). In the LTD-inducing condition, i.e. 2 Hz in picrotoxin, mean peak dendritic [Ca2+] reached to 464 ± 66 nM during conditioning (P < 0.05 over ‘no picrotoxin’ condition, n= 25). In contrast, a brief 100 Hz tetanus elevated [Ca2+] at the same dendritic locations to 1.25 ± 0.36 μm (P < 0.005 over both control and LTD groups, n= 6).
Involvement of postsynaptic protein kinase C in LTD induction
Although MCPG did not affect [Ca2+] increases during 2 Hz stimulation in the presence of picrotoxin (Fig. 4B), it did block the induction of LTD (Fig. 2A). Some subtypes of the mGluR family, mGluR1 and mGluR5 (Schoepp & Conn, 1993; Nicoletti et al. 1996), are positively coupled to phospholipase C, whose activation results in the generation of diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3) due to hydrolysis of membrane-bound phosphoinositide (Schoepp & Conn, 1993; Nicoletti et al. 1996). DAG is a strong endogenous activator for protein kinase C (PKC) (Tanaka & Nishizuka, 1994), and PKC is a well-known messenger required for LTP induction (e.g. Lovinger et al. 1987; Malinow et al. 1989). We tested whether PKC activation was necessary for the LTD observed in the present study using the specific inhibitor peptide PKC(19-36). PKC(19-36) was injected into the postsynaptic cells (250 μm, loaded in recording electrodes). Diffusion of the drug to postsynaptic sites was allowed for at least 60 min before the delivery of 2 Hz stimulation (Malinow et al. 1989). As shown in Fig. 7, this treatment completely blocked LTD (4.5 ± 7.6 % at 45 min, n= 7, P < 0.04).
Both homosynaptic LTD and reversal of LTP (depotentiation) can be induced in the hippocampus by long (∼2-15 min), relatively low-frequncy (1-5 Hz) afferent stimulation. Our measurements showed that LTD-associated increases in postsynaptic [Ca2+] are transient. They occurred only during the first 20–30 s of 2 Hz conditioning, but conditioning periods this short were insufficient to induce LTD (Fig. 4A, inset). It has been shown that, for a significant depression, at least about 200 synaptic inputs have to be applied if the stimulus frequency is set at 2 Hz (1.7 min) (Fujii et al. 1991). On the other hand, we showed that exposure to MCPG during conditioning, which itself did not affect the postsynaptic [Ca2+] increases, blocked LTD. These results suggest the possibility that, for LTD induction, the rapid influx of Ca2+ is co-operative with a more prolonged episode of synaptic activation of mGluRs.
Postsynaptic hyperpolarization, but not the blockade of NMDA receptors, abolished the [Ca2+] rises and significantly reduced LTD, suggesting the involvement of voltage-sensitive Ca2+ channels in LTD induction. However, we did not attempt to block LTD by abolishing only the rapid [Ca2+] rises, i.e. the rises during the first 20–30 s of the conditioning. Therefore the possibility that undetectable [Ca2+] increases later in the conditioning period contribute to LTD induction cannot be ruled out. Also, we did not determine which types of Ca2+ channels are involved in LTD. In this respect, Oliet et al. (1997) recently showed involvement of T-type Ca2+ channels in mGluR-dependent induction of LTD in CA1 neurons prepared from young rats.
There are significant contrasts between the factors necessary for LTD induction shown in the present study and those found in other studies, as well as contrasts among these other studies. We suggest that the most important experimental variables contributing to these differences are conditioning stimulus frequency and amplitude, and age of the experimental animal. Using a 2 Hz protocol, we showed that a level of Ca2+ entry critical for LTD induction in the adult hippocampus can occur via the activation of voltage-gated dendritic channels rather than NMDA receptor-linked channels, so that LTD is still inducible in the absence of NMDA receptor activation. Bolshakov & Siegelbaum (1994), using neonatal rats and a 5 Hz protocol, also showed that LTD induction was not blocked by APV, but could be blocked by holding neurons at -70 mV during conditioning, and by bath-application of MCPG. Similarly, Oliet et al. (1997) showed, using neonatal and juvenile rats and the same 5 Hz stimuli, that MCPG-sensitive, APV-resistant LTD depends on Ca2+ influx via Ni2+-sensitive channels. In contrast, other studies that used 1 Hz protocols in either adult or immature hippocampus showed a requirement for NMDA receptors in LTD induction (Dudek & Bear, 1992; Mulkey & Malenka, 1992; Selig et al. 1995; Kerr & Abraham, 1995; Heynen et al. 1996; Staubli & Ji, 1996). We have indeed observed that, in the presence of APV, 1 Hz stimuli did not induce LTD while 2 or 5 Hz stimuli did, in the same slices prepared from immature hippocampus (S. Otani & J. A. Connor, unpublished observations). Thus, the higher frequency conditioning protocols appear to lower the dependency on NMDA receptor activation. These previous and present results suggest that methodological differences determine which induction factor (i.e. NMDA receptors or voltage-gated channels/mGluRs) is more dominantly involved in LTD. Alternatively, it may be that there are two separate mechanisms for LTD induction that coexist, one controlled by NMDA receptor activation and the other by mGluRs (Oliet et al. 1997). Variations in induction protocols may give one more importance than the other.
The stimulus strength used to induce LTD has also varied between different experiments. With the exception of our present study, which employed intracellular recording, all cases of synaptic LTD induction in adult preparations have been shown by measuring field responses. This complicates the comparison of results from different laboratories because of electrode placement and other factors. However, Kerr & Abraham (1995) have noted that LTD induction in adults fails to occur at a field EPSP amplitude of 0.5 mV, but is successfully induced at 1.0 mV (without GABAA block). Staubli & Ji (1996) showed that weak conditioning pulses, which produced no appreciable population spike, induced LTD in only about 20 % of preparations, while larger stimuli, which triggered population spikes, raised the success rate to 90–100 %. Stimulation in our experiments was subthreshold, or gave only sporadic firing, without GABAA block. Although their use of field recording makes it difficult to detect poorly synchronized firing, the stimulus employed by Wagner & Alger (1995) was probably also in the low range, and they too failed to observe LTD induction without GABAA block in slices from adults using a 1 Hz, 900 pulse protocol. Thus, we propose that the factors of primary importance in LTD induction in adult preparations are whether the cell fires in response to the conditioning stimulus and whether GABAA block increases the effectiveness of low amplitude stimuli to elicit firing.
In normal saline, CA1 neurons fire at most once in response to a moderate presynaptic stimulus. The conjunction of this depolarization with NMDA receptor activation should allow localized Ca2+ entry at the dendritic spines as well as less focal, voltage-gated entry. It has been directly shown in this laboratory that Ca2+ changes are most intense at the spines during tetanic stimulation and remain fairly localized there, even during tetani of several hundred milliseconds (Müller & Connor, 1991; Petrozzino et al. 1995). It is presumed, although not directly shown to our knowledge, that this pattern holds for single spikes. If a large part of the Ca2+ signalling in LTD derives from NMDA receptor influx, we would expect a homosynaptic induction as in the 1 Hz protocols, although, at high stimulus intensities, a wider region of Ca2+ spread, encompassing neighbouring synapses, might occur, supporting a heterosynaptic component (see Otani & Connor, 1996a; Staubli & Ji, 1996).
LTD induction in our study was dependent on postsynaptic PKC activation (see also Oliet et al. 1997), which has also been implicated in LTP induction (Lovinger et al. 1987; Malinow et al. 1989). We suggest that there may be a difference in the manner and location of PKC activation between two conditions. It is known that Ca2+-dependent activation of PKC subspecies expressed in CA1 pyramidal cell dendrites (α, β-II, γ) is facilitated by the presence of DAG (Nishizuka, 1992; Tanaka & Nishizuka, 1994): when arachidonic acid is available together with DAG, PKC may be activated even at basal [Ca2+] levels. Importantly, the action of these co-factors can sustain activation of PKC (Nishizuka, 1992). It is therefore possible that dendritic production of DAG and arachidonic acid (Dumuis et al. 1990) by repeated mGluR activation during prolonged low-frequncy orthodromic stimulation may help activate PKC persistently at low or basal [Ca2+] levels. In contrast, large [Ca2+] increases during LTP-inducing high-frequncy stimulation (see Fig. 6), especially the NMDA-dependent increases in dendritic spines (Petrozzino et al. 1995), may activate synaptic PKC with a limited access to DAG and arachidonic acid.
However, we should be still cautious in interpreting the effect of the mGluR antagonist MCPG on LTD induction. MCPG blocks both group I mGluRs (i.e. mGluR1 and 5), which are positively coupled to phospholipase C cascades, and group II mGluRs (i.e. mGluR2 and 3), which are known to be negatively coupled to cAMP pathways (Watkins & Collingridge, 1994; Nicoletti et al. 1996). Although CA1 pyramidal neurons express mGluR5 (Lujan et al. 1996), antagonism of MCPG to mGluR5 is somewhat controversial (Joly et al. 1995vs.Saugstad et al. 1995). It is thus possible that MCPG acted on group II mGluRs to block LTD induction in a manner unrelated to inhibition of the phospholipase C pathways. It has been shown in dentate gyrus that a group II mGluR antagonist blocks LTD induction by 1 Hz stimuli (Huang et al. 1997). In mossy fibre-pyramidal neuron synapses in area CA3, lack of functional mGluR2 resulted in the impairment of LTD with 1 Hz conditioning (Yokoi et al. 1996). It is interesting to note here that Petralia et al. (1996) recently showed the existence of group II mGluRs in CA1 postsynaptic neurons, as well as in presynaptic components.
We thank J. Petrozzino, R. Sun and L. M. Verselis for technical assistance and D. Linden and A. Greenwood for critical reading of the manuscript. We also thank F. Crépel, W. C. Abraham, and G. Golarai for their comments on the work.