Endogenous opioids regulate long-term potentiation of synaptic inhibition in the dentate gyrus of rat hippocampus

Long-term potentiation (LTP) of excitatory transmission in the hippocampus has been extensively studied as a synaptic model of learning and memory. Here we report a new form of LTP in which inhibitory synaptic signals are potentiated following tetanic stimulation of an opioid-containing excitatory pathway in the presence of opioid antagonists. The lateral perforant path (LPP) was stimulated at the dentate outer molecular layer of hippocampal slices. Evoked synaptic currents were recorded from dentate granule cells using whole- cell voltage-clamp techniques. A high-frequency stimulus train (100 Hz, 1 sec) delivered to the LPP in the presence of naloxone (1 microM) was found to induce a long-lasting potentiation (20 min to 2 hr) in the amplitude of gamma-aminobutyric acidA (GABAA) receptor-mediated inhibitory postsynaptic currents (IPSCs) of granule cells. Such a potentiation was not observed when tetanizing the LPP in control medium. Naloxone-revealed LTP of LPP-evoked IPSCs did not depend upon the presence of granule cell discharge, and was not accompanied by potentiation of mossy fiber-evoked IPSCs, indicating that feedforward, but not feedback, inhibitory circuits were involved. Induction of this LTP could be completely blocked by the N-methyl-D-aspartate (NMDA) receptor antagonist D-2-amino-5-phosphonopentanoic acid (D-APV). However, it was not significantly affected by hyperpolarization of granule cells. These results suggest that LTP may occur at the excitatory synapses between LPP terminals and GABAergic interneurons, rather than at the inhibitory synapses between interneurons and granule cells. Further examination using selective opioid antagonists demonstrated that blocking delta, but not mu and kappa, receptors is critical for inducing LTP of IPSCs in granule cells.

, hippocampal CA1 neurons (Morishita and Sastry, 199 I ;Xie and Sastry, 1991) and cerebellar Purkinje cells (Kano et al., 1992) following tetanic stimulation of excitatory inputs to the same cells. It is unclear yet whether the LTP of inhibitory circuits is a common phenomenon inducible in other brain areas as well, and very little is known concerning the factors that modulate LTP induction in the inhibitory circuitry. In the present study, we describe evidence that endogenous opioids may regulate the LTP of IPSCs of dentate granule cells following high-frequency stimulation of the LPP The LPP originates in the entorhinal cortex and projects to the outer molecular layer of the dentate gyrus, providing both glutamatergic (White et al., 1977) and opioid (Gall et al., I98 I) inputs to distal dendrites of granule cells. These LPP terminals also synapse on the interneuron dendrites in the molecular layer (Zipp et al., 1989). Immunocytochemical studies have demonstrated abundant GA-BAergic interneurons distributed throughout the dentate gyrus (Seress and Ribak, 1983;Babb et al., 1988;Han et al., 1993). The GABAergic axons from these interneurons synapse on granule cell dendrites, somata, and axon initial segments, constituting the anatomic substrate for the inhibitory synaptic signals recorded in granule cells during LPP stimulation.
While high-frequency stimulation of the LPP has been reported to induce LTP of excitatory transmission at the LPP-granule cell synapses (Bramham et al., 1988;Xie and Lewis, 1991;Hanse and Gustafsson, 1992;Colino and Malenka, 1993) it remains to be examined whether LPP tetanization could induce similar potentiation of GABAergic input to the granule cell. Furthermore, opioid peptides contained in the LPP projection, mainly enkephalins (Gall et al., 1981), have been found to facilitate LTP of excitatory transmission through p. (Xie and Lewis, 1990) and 6 (Bramham et al., 199la) receptors. Since activation of p, and 6 receptors reduces GABA-mediated IPSCs while showing no direct enhancing effects on excitatory postsynaptic currents (EPSCs) in granule cells (Xie et al., 1992;Piguet and North, 1994), we have previously proposed (Xie and Lewis, I99 1) that endogenous enkephalins may facilitate LTP of excitation at LPP synapses by suppressing GABAergic inputs to the granule cell (i.e., disinhibition).
Here we further explore if endogenous opioids, through their disinhibitory actions, could also regulate the plasticity of inhibitory circuits in the dentate gylus.
Synaptic stimulation and extracrllular recording. A sharpened monopolar tungsten electrode was placed in the outer one-third of the dentate molecular layer to stimulate the LPP terminals as previously described (Xie and Lewis, 1991). A second stimulating electrode could be positioned in stratum lucidum of area CA3b/c at least 100 pm away from the dentate hilus to stimulate mossy fibers, or in the middle molecular layer to activate the medial perforant path (MPP). Constant-current rectangular stimulus pulses (0.1 msec, 30-500 PA) were delivered through the electrodes by an isolated stimulator. Extracellular responses were routinely monitored through recording electrodes placed in the granule cell layer. In some experiments test pulses subthreshold for granule cell population spikes were employed, in which case more than one site in the granule cell layer near the whole-cell recording electrode was tested, to ensure that the stimuli were subthreshold for all granule cells in the area.
Whole-cull voltage-clamp recording. Evoked synaptic currents were recorded from dentate granule cells using electrodes (3-5 Ma) pulled from World Precision Instruments glass on a Sutter puller. Electrodes were filled with a solution containing (mM) N-methyl-D-glucamine, 90; methanesulfonate, 90; CsF, 40; HEPES, IO; MgCl?, 2; QX314, 3; and Mg-ATE 2 at pH 7.2. In some cases CsF was replaced with CsCl. After establishing the whole-cell configuration as previously described (Xie et al., 1992), the membrane potential usually fell to values close to 0 mV within l-2 min. The cell was then held at 0 mV and whole-cell currents were recorded under voltage-clamp conditions using a patchclamp amplifier (Axopatch-ID).
The holding potential could be transiently altered by voltage steps, which began 200 msec before the synaptic stimulus and continued up to 1 sec. The cell input resistance was monitored by checking the deflection of holding current during the voltage steps. It was in a range of 150-600 Ma in the cells examined. Measured series resistance was usually IO-20 MR and could be compensated up to 80%. Recorded currents were filtered at 2 kHz, displayed on a digital oscilloscope (Nicolet 3 IO), and stored on video tapes using a PCM data recorder (Vetter 200T).
Fluoride and the quaternary lidocaine derivative QX3 14 contained in the recording electrode suppressed voltage-dependent calcium (Kay et al., 1986) and sodium (Connors and Prince, 1982) currents, respectively. This treatment allowed us to hold granule cells at depolarized potentials without complications due to voltage-dependent currents. In addition, cesium (Johnston et al., 1980) and QX314 (Nathan et al., 1990) blocked potassium conductances, including GABA,, receptor-mediated late IPSCs. Thus, GABA,-mediated IPSCs could be recorded in isolation by stepping cells transiently (1 set) to the reversal potential for EPSCs during synaptic stimulation.
The reversal potential for EPSCs determined in the presence of GABA, antagonist bicuculline methiodide (20 FM) was 30 ? 3 mV (N = 7) under our experiment conditions. Experimental protocol. After formation of whole-cell patch, the cell was allowed to stabilize for 20 min. The baseline responses evoked by single test pulses to the LPP or mossy fibers were observed during this period of time. The test pulse was delivered once every minute at a holding potential of +30 mV to isolate 1PSCs as mentioned above. The stimulus intensity of test pulses was adjusted to evoke baseline IPSCs of 200-500 pA. When stable responses were obtained, a single highfrequency stimulus train (100 Hz, 1 set) was delivered to the LPP at a holding potential of 0 mV. The train stimuli used the same stimulus intensity as test pulses, or had a intensity of 100 FA when the test pulse was below 100 FA. Posttrain responses to test pulses were monitored for at least 30 min.
In the experiments where opioid antagonists were applied, drugs were usually added into the perfusion medium starting 10 min before LPP tetanus and continuously applied through the duration of the experiment. When using p-funaltrexamine (P-FNA) to irreversibly block p. receptors, slices were incubated with /3-FNA (IO FM) for 20 min and then washed with drug-free ACSF for 1 hr to remove unbound P-FNA before start of experiments.
Data analysis. Data were analyzed off line. The amplitude of IPSC was measured at its peak. Reversal potentials of synaptic currents were determined by the intercept of the plot of currents versus holding voltages (I-V curve). Data are expressed as means ? SE and subjected to a one-way analysis of variance and Student's t tests. Statistical significance was defined as P < 0.05.

Results
Data were obtained from 102 stable (40 min to 3 hr duration) whole-cell recordings from dentate granule cells with each granule cell recorded in a separate slice.

Naloxone-revealed LTP of IPSCs
Single test stimuli to the LPP terminals in the outer molecular layer evoked an outward current in granule cells held at +30 mV (Fig. IA). The GABA, antagonist bicuculline methiodide (20 J*M) completely blocked this outward current, indicating that it was a pure GABA, receptor-mediated current (Fig. 2). This current will be referred to simply as IPSC. Figure 1 illustrates the effect of LPP tetanization on IPSC amplitudes in the presence of the opioid antagonist naloxone. Application of 1 PM naloxone for 10 min before the train did not cause significant change in IPSC amplitudes.
However, the following stimulus train delivered in naloxone induced long-lasting increases in IPSC amplitudes.
The time course of IPSC increase appeared slower than what one saw in studies of LTP of excitatory synaptic signals (Colino and Malenka, 1993). The potentiation of IPSCs started within 1 min, gradually reached the peak lo-25 min after the train, and typically lasted for the duration of the stable recording conditions (20 min to 2 hr). In control experiments, the LPP was tetanized in ACSF without naloxone.
The IPSC amplitude showed small increases (8-15%) within 10 min after the train, but returned to the pretrain level thereafter (Fig. IB). To exclude nonspecific effects of naloxone, the MPP, a non-opioid containing pathway converging to granule cells, was tetanized in the presence of naloxone (1 FM). The amplitude of MPP-evoked IPSCs measured 20 min after the train was 89 + 4% (N = 8) of the pretrain value, indicating no induction of LTP To exclude any unknown artificial effects of intracellular fluoride on IPSCs due to its influence on calcium currents (Kay et al., 1986), CsCl was included in the electrode solution to replace CsF in another group of experiments. LPP tetanus in the presence of naloxone still induced LTP of IPSCs under such conditions. The average increase of IPSC amplitudes recorded with CsCl-containing electrodes was 55 ? 22% (N = 3) 25 min after the train, which was not significantly different from that observed with CsF-containing electrodes (67 + 22%, N = 7).

Involvement
of the feedforward, but not feedback, inhibitory circuitry Stimulation of the perforant path activates both feedforward and feedback inhibitory circuitry (Anderson et al., 1966;Buzsaki and Eidelberg, 198 1;Seress and Ribak, 1984;Scharfman, 199 1). Feedforward inhibition is activated by excitatory perforant path synapses on interneurons. Feedback inhibition is activated if the perforant path stimulus causes granule cell discharge, thus activating recurrent mossy fiber collaterals that in turn excite inhibitory interneurons. Two groups of experiments were conducted to investigate which set of the inhibitory circuitry was involved in LTP of granule cell IPSCs.
First, we used very weak LPP test stimuli (30-80 FA) to evoke IPSCs before and after LPP tetanus ( Fig. 2A,B). These stimuli were subthreshold for evoking extracellular population spikes in the granule cell layer, thus causing no granule cell discharge. IPSCs evoked by these stimuli were considered to result from activation of the feedforward inhibitory circuitry. The amplitude of IPSCs increased by 50 i 14% (N = 8) after LPP tetanus in these experiments, indicating potentiation of the feedforward inhibitory circuitry. Second, we attempted to measure LTP of the feedback inhibitory circuitry by monitoring mossy fiber-evoked IPSCs in granule cells during LPP tetanus. As previously reported (Mott et al., 1993), mossy fiber stimulation under our experimental conditions polysynaptically evoked pure GABAergic currents that could be completely abolished by GABA antagonists or by blockade of excitatory transmission.
Since GABA,-mediated late IPSCs had been blocked by intracellular cesium and QX314 in the present study, mossy fiber-evoked IPSCs here were entirely GABA,-mediated currents that were sensitive to bicuculline methiodide (Fig. 2C). In nine cells, both LPP-and mossy fiber-evoked IPSCs were monitored before and after the LPP train in the presence of naloxone (Fig. 2CJ).
The LPP-evoked IPSCs were increased by 40 + 8% 20 min after the train (P < 0.05), while the mossy fiber-evoked IPSCs were only increased by 4 ? 4% at the same time, not significantly different from the pretrain amplitude.
Thus  1993). To determine if this is also the case for LTP of IPSCs observed here, the NMDA antagonist D-APV (40 PM) was simultaneously applied with naloxone (1 FM) before and during LPP tetanus (Fig. 3). Posttrain changes of IPSC amplitudes were less than 10% of the pretrain values within 30 min (N = 6), indicating a complete blockade of IPSC potentiation in these experiments.
Thus, similar to LTP of excitatory synaptic signals, LTP of IPSCs also depended on activation of NMDA receptors.
Effect of the P-adrenergic receptor antagonist Perforant path stimulation in the presence of naloxone was previously reported to cause a short-term increase in the amplitude of mossy fiber-evoked IPSCs of CA3 pyramidal cells. The increase of IPSCs could be blocked by the p adrenergic antagonist propranolol, indicating involvement of endogenous norepinephrine system (Caudle et al., 1991). To examine if such a mechanism was also involved in LTP of IPSCs observed in dentate, slices were simultaneously perfused with propranolol(1 FM) and naloxone (1 pM) for 10 min before LPP tetanus. No significant change in the baseline IPSCs was found in the presence of propranolol and naloxone. The following LPP train induced a 54 -t 18% increase in IPSC amplitudes 20 min after the train (N = 4, P < 0.05). Thus, the P-adrenergic antagonist did not block naloxone-revealed LTP of IPSC in the dentate. A representative experiment is shown in Figure 3.  Effect of hyperpolarization of granule cell membrane potentials As demonstrated above, LTP of IPSCs involved LPP-activated, feedforward inhibitory circuitry (Fig. 2). Thus, the increase in IPSC amplitudes could result from LTP occurring at either of two synapses, the LPP-interneuron synapse or the interneurongranule cell synapse. The latter possibility was further examined here.
Since induction of LTP of IPSCs is an NMDA receptordependent process (Fig. 3) manipulations that attenuate NMDAmediated currents in the postsynaptic neuron would be expected to impair this process. If LTP does occur at the interneurongranule cell synapse, hyperpolarization of the granule cell, which facilitates magnesium blockade of NMDA channels (Nowak et al., 1984), should significantly suppress LTP induction, preventing the increase of IPSC amplitudes.
To test this possibility, LPP trains were delivered when holding granule cell at different membrane potentials. It was found that changing the holding potential from 0 mV to -60 or -90 mV did not affect the induction of LTP (Fig. 4). In a total of 11 granule cells, LPP trains delivered at a holding potential of -90 mV induced a 34 ? 18% increase in IPSC amplitudes 25 min after the train. This value was not significantly different from that observed in granule cells held at 0 mV during the train (49 & 15%, N = 6). Thus, hyperpolarizing granule cell did not block LTP induction, suggesting that the potentiation might not occur at the interneuron-granule cell synapse.

Selective opioid antagonist experiments
To determine the subtype of opioid receptors involved in naloxone-revealed LTP of IPSCs, selective opioid antagonists were applied to different groups of slices to replace naloxone during LPP trains (Fig. 5) The Journal of Neuroscience, May 1995, f5 (5) 3793 one. In another two groups of experiments, slices either were pretreated with IO IJ,M B-FNA, an irreversible p antagonist, or were perfused with the K antagonist nor-binaltorphimine (nor-BNI; 100 nM) during LPP tetanus. LTP of IPSCs was observed in neither group of granule cells, indicating that blocking 6, but not p, and K, receptors was required to reveal LTP of IPSCs.

Discussion
The present study demonstrates that LPP-activated inhibitory circuits in the dentate are capable of undergoing LTP following LPP tetanization. However, the induction of this form of LTP portant feature of LTP observed here is its NMDA receptor dependence, which differs from LTP of inhibitory synapses pre- The hyperpolarization experiments provide further evidence viously reported in mammalian brain. Two brief reports arguing against LTP of the interneuron-granule cell synapse. (Morishita and Sastry, 1991;Xie and Sastry, 1991) have described LTP of GABA,-and GABA,-mediated IPSPs in CA1 pyramidal cells. This LTP is enhanced, rather than reduced, by LTP of IPSCs was not prevented by hyperpolarizing granule cell chelation of intracellular calcium and blockade of protein kinase C, making it unlikely that it is NMDA receptor dependent. Anmembrane potentials to the level that has been shown to block other example is LTP of monosynaptic IPSPs of layer V cells NMDA receptor-dependent LTP in CA1 (Kato et al., 1993) and in rat visual cortex, which can be induced in the presence of NMDA receptor antagonists (Komatsu, 1994).
CA3 (Katsuki et al., 1991) pyramidal cells. Since the LTP is NMDA receptor dependent, this result indicates that the relevant NMDA receptors may not be on granule cells, and the LTP therefore does not occur at the interneuron-granule cell synapse. Although in our experiments the entire granule cell might not be hyperpolarized uniformly because of imperfect dendritic space clamp, the presence of intracellular cesium, which increases membrane resistance, and the short electrotonic length of the granule cells  should have facilitated the appears to be inhibited by endogenous opioids. Blocking the vivo. The latter two studies indicated a tetanus-induced, longlasting potentiation of synaptic inhibition in the absence of effect of endogenous opioids on 6 receptors is necessary for the opioid antagonists, which we have not seen in this study. Differences in methodology and means of assessing inhibition full expression of LTP in these inhibitory circuits. Several earlier might account for this discrepancy. Changes in synaptic inhibition were inferred indirectly in these two reports, whereas IPSCs of granule cells were measured directly in the present study. On in viva studies have attempted to examine plasticity of inhibition the other hand, our experiments were conducted with in vitro in the dentate in the absence of opioid antagonists, and the reslice preparation, where inhibitory circuits could be partially disrupted, possibly making LTP of inhibition more difficult to obtain. If activation of 6 receptors normally suppresses LTP of suits have been mixed. A long-lasting reduction of paired-pulse inhibition, blockade of these receptors may be necessary to in-inhibition following perforant path LTP was observed by Maru et al. (1989) in anesthetized rats. On the other hand, Buzsaki and Eidelberg (1982) observed increased probability of interneurons firing to single test stimuli following perforant path tetanization in extracellular unit recordings. Kairess et al. (1987) inferred increases in feedforward inhibition indirectly from changes in the EPSP slope versus population spike amplitude curve (E-S relationship) following perforant path tetanization in hyperpolarization to extend well into granule cell dendrites.
Overall, it appears more likely that the site of LTP is the LPPinterneuron synapses. A number of investigations have observed enhanced excitability of GABAergic interneurons following hippocampal formation LTP (Buzsaki and Eidelberg, 1982;Taube and Schwartzkroin, 1987;Stelzer et al., 1994). If LTP indeed occurs at the LPP-interneuron synapse, the polysynaptic nature of the potentiation may explain the time course difference between LTP of IPSCs observed here and LTP of EPSCs reported in other studies. By monitoring IPSCs at the granule cell, we were remotely measuring the effect of LTP of interneuron EPSCs. The time course of LTP of IPSC could therefore differ crease the robustness of the phenomenon for it to be seen in the from that of interneuron EPSCs because of the influence of other slice preparation. This may explain why only a small and short-factors, such as the postsynaptic excitability of interneurons that lasting potentiation of IPSCs was observed in control medium determines the threshold of interneuron firing, or any unknown in this study.
In the present study, granule cell IPSCs were evoked polysynaptically by LPP stimulation that activated both feedforward and feedback inhibitory circuits in the dentate (Anderson et al., 1966;Buzsaki and Eidelberg, 198 1;Seress and Ribak, 1984;Scharfman, 1991). Since mossy fiber-evoked IPSCs were not enhanced by LPP tetanization in the presence of naloxone, it was unlikely that the feedback circuitry was potentiated in our model. Furthermore, some of our experiments used test stimuli that were subthreshold for inducing granule cell discharge, thus only activating the feedforward inhibitory circuitry that has a lower threshold for synaptic activation (Buzsaki and Eidelberg, 1982;Scharfman, 1991). IPSC potentiation was still manifested in these experiments, indicating that LTP of IPSCs involves primarily the feedforward, but not feedback, inhibitory circuitry. LPP-activated feedforward inhibitory circuitry includes two synapses, the excitatory synapse driving GABAergic interneurons (LPP-interneuron synapses) and the GABAergic synapse on granule cells (interneuron-granule cell synapses). The potentiation of granule cell IPSCs could have resulted from LTP at either of these two synapses. Of these two sites, LTP at the GABAergic inhibitory synapse has been considered the less likely. One im-transient changes in the inhibitory input to the interneurons. Finally, hippocampal interneurons are hyperpolarized by enkephalins (Madison and Nicoll,l988), providing a possible mechanism whereby release of endogenous enkephalins from LPP terminals during tetanization might inhibit LTP at the LPP-interneuron synapse. It is conceivable that released enkephalins may act at 6 receptors on interneurons to hyperpolarize their dendrites, which reduces calcium influx through NMDA channels and hence inhibits induction of LTI? Different opioid receptor subtypes are known to play distinct roles in modulating LTP of excitatory synaptic signals in the hippocampus (Bramham et al., 199la;Xie and Lewis, 1991;Wagner et al., 1993;Weisskopf et al., 1993;Terman et al., 1994). Our data suggest that this is also the case for opioid regulation of LTP in inhibitory circuits. Blockade of 6, but not p and K, receptors was found responsible for naloxone-revealed LTP of IPSCs. It is particularly of interest that the p. antagonist B-FNA did not facilitate IPSC potentiation, though activation of p receptors has been known to strongly inhibit IPSCs of granule cells (Xie et al., 1992;Piguet and North, 1994 et al., 1979;Madison and Nicoll, 1988) and the other by direct inhibition of GABA release from interneuronal terminals (Lambert et al., 1991;Cohen et al., 1992;Capogna et al., 1993 _^_._._ -_-__ _^ ----, r--D-._..,_., _._ ____ _._ TT---~'-~-l \--------., Capogna et al., 1993). Such a presynm -I'-' 346-350. aprlc mecnamsm may cause strong and I.
. terfere with induc short-lasting suppresslon of IPSCs, but not intion of LTP at LPP-interneuron synapses. Fu-Buzsaki G, Eidelberg E (1982) Direct afferent excitation and long-term potentiation of hippocampal interneurons. J Neurophysiol 48:597-607. ture studies using interneuron recordings will be needed to provide direct information regarding k and 6 opioid effects on these cells.
Regardless of the site of potentiation responsible for LTP of IPSCs reported here, our data reveal a novel aspect of opioid modulation of synaptic plasticity, the regulation of LTP in inhibitory circuits. Previous studies demonstrated that endogenous opioids acting at k and 6 receptors facilitated LTP of excitatory LPP input to granule cells (Xie and Lewis, 1990;Bramham et al., 199 1 a). Now we have evidence that, simultaneously, opioids suppress LTP of inhibitory inputs to the granule cell through 6 receptors. The combination of these two aspects of opioid actions could strongly influence excitability of the LPP circuit. Since high-frequency stimuli are required for releasing endogenous opioids from hippocampal formation (Caudle et al., 1991), sustained high-frequency firing in the entorhinal cortex that releases opioids from the LPP would enhance LTP of EPSCs and suppress LTP of IPSCs in the LPP to granule cell circuit. Conversely, firing patterns that does not effectively release opioids, for example, brief episodes of low-frequency theta pattern firing may tend to favor LTP of IPSCs over LTP of EPSCs. Thus, opioid release may be envisioned as determining the direction of synaptic plasticity in this circuit toward inhibition versus excitation.
Morphological studies have demonstrated innervation of granule cell dendrites by GABAergic axons in conjunction with the perforant path terminals (Halasy and Somogyi, 1993;Han et al., 1993). The pairing of excitatory and inhibitory inputs on the same dendritic segments of granule cells may provide a mechanism for powerful and selective GABAergic control over the gain of excitatory inputs from the perforant path. Indeed, activation of GABA, receptors causes shunting inhibition of NMDA currents in granule cells , while reduction of GABA, inhibition facilitates induction of LTP of perforant path-evoked field EPSPs (Wigstrijm and Gustafsson, 1983;Mott and Lewis, 1991). By preventing LTP in LPP-activated inhibitory circuits, endogenous opioids would be expected to attenuate tonic inhibitory control over granule cell dendrites, enhancing their responses to further excitatory input. This mechanism could play an important role in maintaining enhanced excitability of granule cell during repeated episodes of afferent activation, such as seizure activity.