Corresponding author B. E. Alger: Department of Physiology, University of Maryland School of Medicine, 655 West Baltimore Street, Baltimore, MD 21201, USA. Email: firstname.lastname@example.org
1We investigated depolarization-induced suppression of inhibition (DSI) under whole-cell voltage clamp in CA1 pyramidal neurons of rat hippocampal slices. DSI, a transient reduction in monosynaptic evoked GABAAergic IPSCs lasting for ∼1 min, was induced by depolarizing the pyramidal cell to −10 or 0 mV for 1 or 2 s.
2Raising extracellular Ca2+ concentration increased DSI, and varying the DSI-inducing voltage step showed that the voltage dependence of DSI was like that of high-voltage-activated Ca2+ channels.
3The P- and Q-type Ca2+ channel blocker ω-agatoxin TK (200 nm and 1 μm) and the R- and T-type Ca2+ channel blocker Ni2+ (100 μm) reduced IPSCs without reducing DSI.
4The specific N-type Ca2+ channel antagonist ω-conotoxin GVIA (250 nm) reduced IPSC amplitudes and almost completely abolished DSI.
5Blocking L-type Ca2+ channels with nifedipine (10 μm) had no effect on IPSCs or DSI induced by our standard protocol, but reduced DSI induced by the unclamped Na+- and Ca2+-dependent spikes that occurred when 2(triethylamino)-N-(2,6-dimethylphenyl)acetamide (QX-314) was omitted from the recording pipette solution.
6Although intracellular Ca2+ stores were not measured, DSI was not affected by cyclopiazonic acid (CPA, 20–40 μm), a blocker of Ca2+ uptake into intracellular stores.
7We conclude that DSI is initiated by Ca2+ influx through N- and, under certain conditions, L-type Ca2+ channels.
Calcium-dependent modulation of synaptic transmission is ubiquitous in the nervous system, taking place mostly in presynaptic nerve terminals. However, excitatory transmission can also be affected by increases in postsynaptic Ca2+ concentration ([Ca2+]i). Long-term potentiation (LTP) (Malenka et al. 1988) and long-term depression (LTD) (Cummings et al. 1996) are absolutely dependent on increases in postsynaptic [Ca2+]i. Moreover, there is growing evidence of a Ca2+-dependent modulation of inhibitory synaptic transmission. Elevations in postsynaptic [Ca2+]i can reduce the efficacy of inhibitory transmission (Mouginot et al. 1991; Stelzer, 1992). In these studies, the increased [Ca2+]i was found to decrease the responsiveness of postsynaptic GABAA receptors to GABA.
Another form of Ca2+-dependent modulation of GABAergic transmission also occurs in neurons recorded in hippocampal (Pitler & Alger, 1992, 1994) or cerebellar (Llano et al. 1991; Vincent & Marty, 1993) slice preparations as well as cultured neurons (Ohno-Shosaku et al. 1998), and is termed depolarization-induced suppression of inhibition (DSI) (Alger & Pitler, 1995). DSI is a transient (∼1 min) reduction in monosynaptic GABAergic transmission following a brief depolarization of the postsynaptic neuron. DSI can be induced in principal cells by either a depolarizing voltage step or by a train of action potentials induced by current injection. Unlike phenomena that involve a postsynaptic decrease in the sensitivity of the GABAA receptor, the decrease in inhibition observed during DSI is mediated presynaptically by a reduction in GABA release. There is no decrease in postsynaptic GABAA receptor sensitivity in either the cerebellum or hippocampus. Because DSI is induced postsynaptically and expressed presynaptically, the involvement of a retrograde messenger can be inferred (Llano et al. 1991; Pitler & Alger, 1994; Alger et al. 1996). The findings that decreases in inhibitory transmission can promote certain types of epilepsy (McNamara, 1994), as well as the induction of long-term potentiation (Wigström & Gustafsson, 1983), underscore the physiological importance of this phenomenon.
The hypothesis that DSI is Ca2+ dependent is based on several observations: (1) DSI reportedly runs down concomitantly with the run-down of voltage-dependent Ca2+ tail currents (Llano et al. 1991); (2) DSI can be prevented by application of Ca2+-free solutions (Llano et al. 1991), or addition of Cd2+ to the bathing medium (Llano et al. 1991; Ohnu-Shoshaku et al. 1998); (3) DSI is induced by large voltage steps that induce Ca2+ influx (Llano et al. 1991; Pitler & Alger, 1992), even when intracellular 2(triethylamino)-N-(2,6-dimethylphenyl)acetamide (QX-314) is used to prevent Na+ influx (Pitler & Alger, 1994), and DSI can, in some cells, be enhanced by the L-type Ca2+ channel agonist Bay K 8644 (Pitler & Alger, 1992); finally, (4) DSI is not seen when high concentrations of the high-affinity Ca2+ buffer BAPTA are present in the recording pipette (Pitler & Alger, 1992, 1994). Thus the evidence supports the inference that DSI is mediated by the increased [Ca2+]i that probably takes place through activation of voltage-dependent Ca2+ channels. Nevertheless, apart from the Bay K 8644 data, there is no evidence as to which Ca2+ channels can mediate DSI.
Increases in [Ca2+]i can occur through activation of voltage-dependent Ca2+ channels (VDCCs), of NMDA receptors (Mayer & Westbrook, 1987), or following Ca2+ release from intracellular stores (Berridge & Irvine, 1984). Because NMDA receptors were blocked in our experiments on DSI of monosynaptic inhibitory transmission, we have focused on VDCCs and intracellular stores as possible contributors to the rise in [Ca2+]i. There is evidence for intracellular Ca2+ stores in CA1 neurons (Garaschuk et al. 1997). These cells also express a variety of VDCCs that participate in a wide range of cellular functions such as control of cell excitability, gene expression, neurotransmitter release, and information processing (Dunlap et al. 1995). Pharmacological studies have revealed the existence of at least six distinct subtypes of Ca2+ channel, L, N, T, P, Q, and R (Dunlap et al. 1995; Tsien et al. 1995). In order to understand the DSI mechanism fully it will be important to know whether the particular route of Ca2+ entry is significant or if any rise in bulk [Ca2+]i is sufficient to induce DSI. Furthermore, these channels are modulated by numerous neurotransmitter systems. If VDCCs were involved in the induction of DSI, then neurotransmitter systems could modulate cell excitability by altering DSI. It is, therefore, important to determine which Ca2+ channels are necessary for induction of DSI, and whether intracellular Ca2+ stores contribute to DSI.
We have undertaken the present experiments to address these issues. We conclude that DSI is induced mainly by Ca2+ influx through VDCCs of the N- and, in some circumstances, the L-types. The P-, Q-, R- and T-types do not appear to play a role. It appears that Ca2+ release from intracellular stores may not be necessary for the induction of DSI. A preliminary report of this work has appeared in abstract form (Lenz et al. 1997).
Preparation of slices
Adult male Sprague-Dawley rats (125–300 g, 30–60 days old) were deeply anaesthetized with halothane and decapitated. Both hippocampi were removed, mounted on agar blocks, and placed in a slicing chamber containing oxygenated, partially frozen bath solution (see below). Transverse slices (400 μm thick) were cut with a Vibratome (Technical Products International) and transferred to a holding chamber where they were maintained at the interface of physiological saline and humidified 95 % O2-5 % CO2 atmosphere at room temperature (20–24°C). Slices were allowed at least 1 h to recover before being transferred to a submerged, perfusion-type chamber (Nicoll & Alger, 1981) where they were perfused with bath solution (29–31°C) at 0.5–1 ml min−1.
The bath solution contained the following (mm): 120 NaCl, 25 NaHCO3, 3 KCl, 2.5 CaCl2, 2 MgSO4, 1 NaHPO4 and 10 glucose. 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 μm) and 2-amino-5-phosphonovaleric acid (APV, 50 μm) were present in the saline in all experiments to block ionotropic glutamate receptor-mediated responses. In experiments examining the dependence of DSI on extracellular Ca2+ concentration ([Ca2+]o), the same bath solution was used, except that the concentration of Ca2+ was raised to 5 mm and MgSO4 was omitted.
Whole-cell patch electrodes had resistances of 3–6 MΩ and were filled with one of two solutions (mm): (A) 100 CsCH3SO3, 50–60 CsCl, 10 Hepes, 2 BAPTA, 0.2 CaCl2, 2 MgATP, 1 MgCl2, 0.3 Tris-GTP, 5 QX-314, pH 7.25; (B) 150–160 CsCH3SO3, 10 Hepes, 2 BAPTA, 0.2 CaCl2, 2 MgATP, 1 KCl, 1 MgCl2, pH 7.25. QX-314 blocks Na+-dependent action potentials and certain K+ channels (Andrade, 1991). We found that solution A often produced more stable recordings; however, except for the experiments shown in Fig. 6, there was no difference in the results using the two solutions, and the data have been combined. The experiments of Fig. 6 are discussed separately.
CNQX and QX-314 were purchased from Research Biochemicals International, and BAPTA was purchased from Molecular Probes. ω-Agatoxin TK, a selective blocker of both P- and Q-type Ca2+ channels (Teramoto et al. 1995), was purchased from Alomone Labs (Jerusalem, Israel). All other drugs and chemicals were obtained from Sigma.
Whole-cell recordings and data analysis
CA1 pyramidal cell recordings were obtained using the ‘blind’ whole-cell patch-clamp recording technique. An Axoclamp-2 or an Axopatch 200A (Axon Instruments) was used for all experiments. Cells were voltage clamped to −70 mV soon after break-in. Acceptable cells had resting membrane potentials equal to or greater than −55 mV and input resistances greater than 35 MΩ. Series resistance was less than 12 MΩ at the beginning of an experiment and was compensated by 60–70 %. Cells were discarded if series resistance increased to greater than 30 MΩ during an experiment. Liquid junction potentials were small, and corrections for them were not made.
Bipolar concentric stimulating electrodes (Rhodes Electronics) were positioned in either stratum (s.) oriens or s. radiatum to allow orthodromic activation of CA1 pyramidal cells. Evoked inhibitory postsynaptic currents (eIPSCs) were elicited at 0.3 Hz and were filtered at 2 kHz with an eight-pole Bessel filter (Frequency Devices, Haverhill, MA, USA) and digitized at 5 kHz by a Digidata 1200 analog-to-digital converter (Axon Instruments). Data were also digitized at 22 kHz with a 14 bit PCM digitizer system (Neuro-corder DR-484, Neuro Data Instruments Corp.), stored on a VCR-based tape recorder and played into a computer for off-line analysis using pCLAMP 6.0 software (Axon Instruments). Statistical analysis of the data was performed using Student's paired t test (SigmaStat, Jandel Scientific, Corte Madera, CA, USA). Statistical significance of group data from preincubated slices was assessed using Student's unpaired t test. The significance level chosen was P < 0.05, and all data are reported as means ±s.e.m.
DSI was induced with a 1 or 2 s depolarizing voltage step to −10 or 0 mV every 90–120 s except where noted. To quantify DSI of eIPSCs we ordinarily compared the mean of the five IPSCs after the depolarizing step with the mean of the eight IPSCs immediately preceding the depolarizing step. Group data are often presented as ‘%DSI’ which is calculated from the formula:
where control IPSC (IPSCC) is the mean of eight eIPSCs preceding the voltage step and test IPSC (IPSCT) is the mean of five eIPSCs following the step. In some experiments, as noted, we also determined the ‘absolute IPSC reduction’ during DSI in various conditions. Absolute IPSC reduction is simply the difference (in pA) between IPSCC and IPSCT as defined above. For each cell, three to six DSI trials in control conditions and then in the presence of drug were used to obtain a mean %DSI for that cell. Calcium currents were elicited by a 500 ms voltage step to either −10 or 0 mV. The currents were either Cd2+- or leak subtracted, and the peak current was measured. To quantify drug effects on control IPSCs, we took the mean of at least twenty-five IPSCs both in control conditions and in the presence of drug.
Previous work from our laboratory (Pitler & Alger, 1992) showed that high intracellular concentrations of the Ca2+ chelator BAPTA prevented hippocampal DSI, thereby suggesting that a rise in intracellular calcium concentration ([Ca2+]i) is necessary for its induction. However, it remains unclear whether the source of this Ca2+ is the extracellular space, intracellular stores or both. By lowering [Ca2+]i, BAPTA probably interferes with many processes and, hence, its prevention of DSI does not unequivocally demonstrate that DSI is directly dependent on [Ca2+]i. An additional drawback of the BAPTA experiments is that the rapid diffusion of BAPTA into the postsynaptic cell prevented the observation of DSI. Therefore, these cells could not serve as their own controls.
DSI is dependent on [Ca2+]o
To begin, we confirmed that DSI is dependent on extracellular Ca2+ by increasing [Ca2+]o. In the presence of 2.5 mm Ca2+ and 2.0 mm Mg2+, a 1 s depolarizing step to −10 mV applied to the postsynaptic cell resulted in a reduction in the IPSC amplitude (DSI) for a transient period following the voltage step (Fig. 1A, left trace). Raising the extracellular Ca2+ concentration to 5 mm (high Ca2+), while lowering the Mg2+ concentration nominally to 0 mm, resulted in a dramatic increase in the magnitude and duration of DSI in the same cell (Fig. 1A, middle trace). The effect of high [Ca2+]o on IPSC amplitude and DSI was reversible upon wash (Fig. 1A, right trace). Increasing [Ca2+]o (Fig. 1B, middle trace) increased DSI irrespective of the amount of DSI in control Ca2+ conditions. The increase in DSI remained even when the stimulus intensity was lowered to elicit IPSCs in high [Ca2+]o (Fig. 1B, right trace) of similar amplitude to those recorded in low [Ca2+]o conditions (Fig. 1B, left trace). A similar increase in %DSI (see Methods) in the presence of 5 mm Ca2+ was seen in all five cells tested. DSI increased significantly from 23.0 ± 5.9 % in control conditions to 41.9 ± 7.8 % in the presence of high [Ca2+]o. These results demonstrate a dependence of DSI on extracellular Ca2+, and indicate that the degree of DSI induced by the 1 s voltage step protocol can be modulated. Thus this protocol is suitable for the proposed experiments.
DSI is dependent on Ca2+ influx through VDCCs
If Ca2+ entry from the extracellular space via VDCCs contributes to the rise in [Ca2+]i, then DSI should show a voltage dependence that closely parallels the voltage dependence of Ca2+ channel activation. To test this prediction, we depolarized the postsynaptic pyramidal cell to various voltages ranging from −40 to +50 mV and quantified the amount of DSI induced by each step. As illustrated in Fig. 2A, a depolarizing voltage step to −30 mV produced little or no DSI, whereas a step to −10 or 0 mV suppressed the IPSCs for several seconds following the step. Interestingly, a large depolarizing step to a voltage approaching the zero current potential for Ca2+ did not result in any DSI. The current-voltage relationship for calcium currents (open circles) and the voltage dependence of DSI (filled circles) from the same cell are superimposed in Fig. 2B. Similar results were observed in five other cells, and the normalized voltage dependence of DSI from six cells is shown in Fig. 2C. The similar voltage dependences of Ca2+ channel activation and DSI support previous evidence that Ca2+ entry into the postsynaptic cell through VDCCs is necessary for DSI.
ω-Agatoxin TK reduces IPSCs but not DSI
Several types of VDCCs exist on CA1 pyramidal cells (Christie et al. 1995; Kavalali et al. 1997). P-type Ca2+ channels are present on pyramidal neurons as well as on certain of the inhibitory synaptic inputs to these cells (Yamamoto et al. 1994). In the cell illustrated in Fig. 3A, depolarizing steps to 0 mV reliably resulted in DSI following each step. Application of the P-type channel blocker ω-agatoxin TK, at a concentration that completely blocks P-type channels (200 nm) (Mintz et al. 1992; Teramoto et al. 1995), reduced the IPSC amplitude, but did not reduce DSI. The time course of the experiment (Fig. 3C) reveals the irreversible effects of the toxin on IPSC amplitude and DSI. The filled circles represent the mean of eight IPSCs immediately preceding each DSI-inducing step, and the open squares are mean values of the five IPSCs following each step. Blocking the P-type Ca2+ channels actually increased the calculated mean %DSI from 33.2 ± 4.3 % in control conditions to 50.0 ± 5.0 % in the presence of ω-agatoxin TK (n= 8 cells, P= 0.001, Fig. 3D).
Upon inspection of the data (Fig. 3A) it appeared that the absolute reduction in IPSC amplitude following the DSI-inducing step was the same in control conditions and in the presence of 200 nmω-agatoxin TK. Thus the ω-agatoxin-induced increase in %DSI could result from the method used to calculate %DSI, rather than an actual increase in the DSI mechanism per se. If ω-agatoxin suppressed GABA release from a population of interneurons that were all equally susceptible to DSI, then ω-agatoxin should reduce IPSC amplitudes without affecting the %DSI. On the other hand, some interneurons might not be susceptible to DSI. If ω-agatoxin blocked only the output of cells not susceptible to DSI, then the evoked IPSC would decrease, %DSI would increase and the absolute reduction in IPSC amplitude during DSI would not change. To test this idea we determined the absolute DSI-induced reduction of IPSCs (in pA) in control conditions and again in the presence of agatoxin. As expected, the absolute reduction of IPSC amplitude caused by DSI did not differ significantly between control and ω-agatoxin conditions (Fig. 3E, n= 8). Therefore, the calculated increase in %DSI in ω-agatoxin probably results from a constant DSI-induced reduction in absolute IPSC amplitude, divided by smaller control IPSC amplitudes that are produced by agatoxin. We conclude that agatoxin did not affect the DSI mechanism per se.
Because 200 nmω-agatoxin TK blocks approximately 50 % of the Q-type channels that are present on these CA1 neurons (Sather et al. 1993), it seemed likely that Q-type VDCCs were not involved in DSI. However, we could not rule out a contribution of Q-type channels to DSI based on these data. Contributions of Q-type channels can be virtually eliminated by 1 μmω-agatoxin. Therefore, we incubated slices in 1 μmω-agatoxin TK for several hours to ensure complete block of both P- and Q-type channels (Sather et al. 1993; Wheeler et al. 1994). In three of three cells from three agatoxin-pretreated slices, prominent DSI was present (Fig. 3B) and was comparable in magnitude to that observed in cells to which 200 nmω-agatoxin TK was bath applied (60.5 ± 5.6 % in 1 μmω-agatoxin TK vs. 50.0 ± 5.0 % in 200 nmω-agatoxin TK, difference not significant).
Because slices were incubated in 1 μmω-agatoxin before recording began, we did not directly observe the onset of the toxin's effects on IPSC amplitude. Nevertheless, we are confident that ω-agatoxin was active for two reasons: (1) stimulation intensities that were effective in eliciting typical IPSCs in control slices (335 ± 7 pA, n= 6) elicited much smaller IPSCs in slices pretreated with 1 μmω-agatoxin (197 ± 31 pA, n= 3), and (2) although 200 nmω-agatoxin from the same stock solution always dramatically reduced IPSCs when we applied it during the course of a normal recording (n= 8, Fig. 3A and C), when 200 nm agatoxin was bath applied to slices pretreated with 1 μm agatoxin it had no further effect on IPSCs. Thus pretreatment with 1 μm agatoxin had already produced maximal IPSC suppression, and hence was active. These data strongly argue against the involvement of either P- or Q-type channels in DSI.
N-type channel activation is necessary for DSI
N-type Ca2+ channels exist on CA1 pyramidal cell bodies and dendrites (Christie et al. 1995; Kavalali et al. 1997), and are involved in the presynaptic control of GABA release in the hippocampus (Horne & Kemp, 1991; Ohno-Shosaku et al. 1994). We found that the specific N-type channel blocker, ω-conotoxin GVIA (250 nm), significantly reduced IPSC amplitude (Fig. 4A), and, unlike ω-agatoxin, greatly reduced the %DSI. The effects of conotoxin could not be reversed with up to 30 min wash. The group data from seven cells reveal that ω-conotoxin GVIA reduced the mean DSI significantly from 50.5 ± 8.7 to 9.0 ± 4.4 % (P= 0.005). This effect was seen whether %DSI or the absolute reduction of IPSCs (in pA) during DSI was measured (Fig. 4D and E). Increasing the stimulus intensity partially recovered the IPSC amplitude, but did not recover DSI. In the four cells tested in this way, the mean reduction in IPSC amplitudes caused by ω-conotoxin was 83.5 ± 5.1 % (i.e. to 16.5 % of control values), and increasing the stimulus intensity in ω-conotoxin restored the IPSCs to 58.5 ± 10.1 % of control, without restoring DSI. The time course of an experiment is displayed in Fig. 4C and demonstrates the rapid nature of ω-conotoxin action.
The near-complete block of DSI by ω-conotoxin GVIA suggests that postsynaptic N-type channel activation is necessary for DSI. From this experiment it is, however, difficult to rule out that ω-conotoxin's effect of blocking DSI was mediated presynaptically through its potent reduction of IPSCs. ω-Conotoxin could simply occlude DSI by preventing the output of GABA neurons that are susceptible to DSI. However, if ω-conotoxin's effects were strictly presynaptic, then IPSCs in the presence of ω-conotoxin should not be larger than the IPSCs occurring during maximal DSI. To calculate ‘maximal DSI’, we used only the two IPSCs occurring at the peak of the DSI effect (a span of 6 s), rather than the five IPSCs normally used, as some recovery from DSI was always evident with five responses, and thus maximal DSI would be underestimated. In three of seven cells (see Fig. 4A for example), we observed clearly larger IPSCs during ω-conotoxin application than those occurring during maximal DSI in the same cells. In these three cells the mean IPSC amplitude in ω-conotoxin (267 pA) was 44 % larger than the IPSCs measured during maximal DSI (186 pA). Further evidence against a strictly presynaptic site of action for ω-conotoxin was the occasional observation that ω-conotoxin could reduce DSI prior to reducing IPSCs. An example is seen in Fig. 4B in which ω-conotoxin progressively blocked DSI (from 38 % on the first depolarizing step, to 32 % on the second, to 26 % on the third) before it reduced the IPSC amplitudes (which, at the end of the third DSI trial in this example, were 102 % of the amplitudes at the beginning of the trace). In three cells like this, mean IPSCs in ω-conotoxin were 98 % of control amplitudes at a time when DSI was only 76 % of control values. Thus, it appears likely that ω-conotoxin reduces DSI in part by blocking Ca2+ influx postsynaptically, although more complex presynaptic mechanisms cannot be ruled out by these data.
Ni2+ reduces IPSCs, but not DSI
It has been suggested that both T- and R-type Ca2+ channels exist on hippocampal CA1 neurons (Christie et al. 1995; Kavalali et al. 1997). Low concentrations of NiCl2 (50–100 μm) block both T- and R-type channels while not greatly affecting other voltage-dependent Ca2+ channels (Takahashi & Akaike, 1991; Ellinor et al. 1993; Randall & Tsien, 1995). To determine whether these channels contribute to the postsynaptic rise in [Ca2+]i that initiates DSI, we tested 100 μm NiCl2, but found it had no effect on DSI (Fig. 5A and C). NiCl2 reduced the evoked IPSC amplitude on average to 70.8 ± 8.7 % of control (n= 15, P= 0.04), and this effect was completely reversible upon washout (Fig. 5A and B). The mean %DSI was not significantly altered by NiCl2 (control DSI: 43.9 ± 2.7 %; NiCl2 DSI: 41.9 ± 2.6 %; n= 6, P= 0.2), although the absolute DSI-induced IPSC reduction was decreased.
Nifedipine can affect DSI under certain conditions
L-type Ca2+ channels exist on CA1 pyramidal cells (Westenbroek et al. 1990) and contribute significantly to the total somatic Ca2+ current measured in these cells. The L-type Ca2+ channel antagonist nifedipine (10 μm) did not block DSI of evoked IPSCs (Fig. 6A), in agreement with the previous finding that nifedipine did not block DSI of spontaneous IPSCs (Pitler & Alger, 1992). In the eight cells tested, the mean %DSI in control conditions (41.0 ± 4.8 %) was not significantly different from the %DSI in the presence of nifedipine (43.4 ± 5.8 %, P= 0.2).
We were somewhat surprised that blocking L-type Ca2+ channels had no effect on DSI, especially in light of their high density on the soma and apical dendrites of these cells (Westenbroek et al. 1990) and their large contribution to whole-cell Ca2+ currents. To determine whether L-type VDCCs could play a role in DSI under any conditions, we altered the recording conditions to try to maximize their relative contribution to Ca2+ influx by reducing the ability of the voltage step to depolarize the dendritic region directly. We did this by omitting QX-314 from the recording electrode solution (in addition to blocking Na+ channels, QX-314 blocks certain K+ channels and promotes voltage spread to the dendrites) and using small voltage steps (to −25 or −20 mV) that produced an unclamped Na+- and Ca2+-dependent spike. Interestingly, nifedipine partially blocked DSI elicited by small voltage steps under these conditions (mean control DSI: 22.5 ± 2.0 %; mean nifedipine DSI: 10.9 ± 1.0 %; P= 0.03, n= 4) while not affecting DSI induced by large voltage steps (to −5 or 0 mV) in the same cells (mean control DSI: 31.3 ± 8.3 %; mean nifedipine DSI: 27.9 ± 2.6 %; P > 0.4, n= 4, Fig. 6D). These results suggest that, under certain voltage protocols, L-type Ca2+ channel current can contribute to DSI.
The same reasoning might imply that R- or T-type channel currents could play a role in DSI induced by submaximal voltage steps in the absence of QX-314. To test this possibility, we repeated the same experiments with NiCl2. DSI induced with small voltage steps in the presence of NiCl2 (24 ± 3 %) was not different from control DSI (26 ± 4 %, n= 4, P > 0.1). These data argue against a contributory role for T- or R-type Ca2+ channels in DSI irrespective of the voltage protocol.
Effects of VDCC antagonists on postsynaptic voltage-dependent Ca2+ current (ICa)
We measured the effect of each of the channel blockers on ICa elicited under the standard conditions, i.e. QX-314 present in the recording electrode. Although ICa cannot be ideally clamped in these cells with extended processes, some comparison of effectiveness of the action of the various antagonists could be made. The currents were elicited by a 500 ms voltage step to −10 or 0 mV in control bath solution and in the presence of each channel blocker. Sample currents are displayed in Fig. 7A, and the group data for the peak currents are shown in Fig. 7B. In general agreement with previous studies (Christie et al. 1995; Kavalali et al. 1997), we found that nifedipine produced the largest reduction in ICa, resulting in a peak current that was 59 ± 13 % of control. Blocking N-type channels reduced ICa to 76 ± 8 % of control, while 200 nmω-agatoxin TK reduced it to 73 ± 8 % of control. NiCl2 had a very small effect on ICa, reducing it to 93 ± 3 % of control.
Cyclopiazonic acid does not affect IPSCs or DSI
Taken together, our results confirm that DSI is a Ca2+-dependent mechanism, and indicate that Ca2+ entry through N-type channels, and L-type channels under certain conditions, was necessary for the induction of DSI. However, these results did not rule out a possible contributory role for Ca2+-induced Ca2+ release (CICR) from intracellular stores that are present in these neurons (Garaschuk et al. 1997). Therefore, we tested the effect of cyclopiazonic acid (CPA) on DSI at concentrations (20 and 40 μm) that blocked caffeine-induced Ca2+ transients in hippocampal slices (Garaschuk et al. 1997) by blocking ATP-dependent Ca2+ uptake into intracellular compartments. The mean DSI from eight cells exposed to CPA (47.5 ± 8.6 %) was not significantly different from the DSI obtained in control conditions (53.8 ± 9.0 %, P= 0.3). Similarly, we found that bath application of dantrolene sodium (30 μm), an inhibitor of Ca2+ release from stores, did not affect DSI (control DSI: 55.7 ± 6.3 %; dantrolene DSI: 53.4 ± 4.2 %; n= 2). Therefore, although we did not measure Ca2+ stores directly, it appears that CICR may not be involved in DSI and that Ca2+ entry from the extracellular milieu may be sufficient for DSI induction.
Previous work had suggested that DSI modulates inhibitory synaptic transmission in a Ca2+-dependent manner (Llano et al. 1991; Pitler & Alger, 1992, 1994; Ohnu-Shosaku et al. 1998). The present results strongly support a role for Ca2+ influx through VDCCs in DSI. Increasing [Ca2+]i greatly increased DSI, and the voltage dependence of DSI closely paralleled the voltage dependence of postsynaptic Ca2+ current activation. We have presented evidence for a role for N-type Ca2+ channel activation in the induction of DSI, and showed that L-type channels can contribute to DSI under certain conditions. Preliminary pharmacological tests suggest that intracellular Ca2+ stores are not required for DSI induction.
The finding that raising [Ca2+]o increased DSI supports the conclusion that DSI is dependent on [Ca2+]o. However, intracellular Ca2+ stores are quite sensitive to alterations in [Ca2+]o (Parekh & Penner, 1997). Alternatively, raising [Ca2+]o could affect the release properties of interneurons, which could, in turn, alter DSI. Changing the magnitude of the voltage step, which affects the amount of Ca2+ influx into the cell during the step, altered the %DSI, and, indeed, it was possible to prevent DSI by depolarizing the membrane to voltages approaching the zero current potential for Ca2+. Thus, it appears that the voltage dependence of DSI is very similar to that of high-voltage-activated VDCCs. Because the degree of DSI was sensitive to the magnitude voltage step delivered to the postsynaptic cell, we can exclude presynaptic or extracellular factors in the effect. We conclude that Ca2+ influx through VDCCs is the essential trigger for DSI, in agreement with previous suggestions.
Blocking N-type channels produced a near-complete block of DSI, suggesting that Ca2+ influx through postsynaptic N-type channels could be responsible for DSI. It is, however, difficult to rule out the possibility that ω-conotoxin's effect on DSI was mediated presynaptically. ω-Conotoxin could occlude DSI by preventing GABA release from interneurons that are susceptible to DSI. Alternatively, the retrograde DSI signal could cause DSI by blocking N-type Ca2+ channels located on presynaptic terminals, in which case ω-conotoxin would also occlude DSI presynaptically. However, these explanations seem unlikely to account for all of ω-conotoxin's action in light of the following observations. (1) In some cells the IPSC amplitudes in ω-conotoxin were larger than they were during maximal DSI. If the sole effect of ω-conotoxin on DSI were to occlude it by reducing IPSCs, the IPSCs during ω-conotoxin application should not be larger than those during maximal DSI. (2) We found ω-conotoxin could reduce DSI during a period when IPSCs were not affected. Both of these observations can be simply explained if, in addition to its presynaptic effects, ω-conotoxin also impedes the induction of DSI by preventing postsynaptic Ca2+ influx through N-type channels. More complex presynaptic explanations, perhaps involving interneuron-interneuron inhibition, cannot be ruled out by these experiments, however, and more work will be necessary to establish the role of N-type Ca2+ influx with certainty. It is worth noting that, in view of the involvement of VDCCs in DSI (e.g. Fig. 2), and the lack of effect of antagonists of L-, P-, Q-, R- and T-type VDCCs on DSI, it would appear that either N-type channels are involved, or a VDCC with different antagonist properties must be.
The finding that blocking P- and Q-type channels reduced IPSC amplitudes without blocking DSI shows that neither reducing IPSCs nor Ca2+ influx is sufficient to reduce DSI. In fact, blocking the P-type channels significantly increased %DSI, without affecting the absolute IPSC reduction during DSI. One explanation for these data is that P-type channels are present only on interneurons that are not susceptible to DSI. There is evidence that certain interneurons express either P- or N-type channels on their axon terminals, but not both (Poncer et al. 1997; however, see Ohno-Shosaku et al. 1994). If ω-agatoxin revents GABA release from interneurons that are not susceptible to DSI then the relative contribution of DSI-susceptible interneurons to the evoked IPSC would be greater in ω-agatoxin. This would increase the %DSI while not changing the absolute IPSC reduction during DSI. The lack of change in absolute DSI-induced IPSC reduction suggests that the DSI process is not altered by ω-agatoxin, and hence that postsynaptic Ca2+ influx through P- or Q-type channels is not involved in DSI.
In contrast to ω-agatoxin, 100 μm Ni2+, a blocker of R- and T-type channels, reduced IPSC amplitudes and the mean IPSC reduction during DSI without affecting %DSI. This would be expected if Ni2+ caused a uniform reduction in all GABA IPSCs, both those susceptible and those not susceptible to DSI, while having no effect on the postsynaptic Ca2+ influx necessary for DSI induction. Many studies have reported the relative specificity of low concentrations of Ni2+ for T- and R-type channels (Ellinor et al. 1993; Soong et al. 1993). Ni2+ blocked only a very small portion of the total Ca2+ current measured in the postsynaptic cells and hence was not a potent non-specific blocker of VDCCs in our cells. Ni2+ could reduce IPSCs by blocking a component of presynaptic Ca2+ influx, as Wu et al. (1998) have reported a Ni2+-sensitive component of Ca2+ current in the calyx of Held. This would be consistent with the finding that interneurons in stratum oriens, radiatum, and lacunosum moleculare are immunoreactive for the R-type Ca2+ channel α1 subunit (Yokoyama et al. 1995). On the other hand, Ni2+ also blocks GABA responses in turtle retina directly (Kaneko & Tachibana, 1986), and this could explain the IPSC suppression we observed. In any case, the data rule out postsynaptic Ca2+ influx through R- or T-type channels as a factor in DSI.
It appears that L-type Ca2+ channels can contribute to induction of DSI under certain experimental conditions. In earlier work we found that the L-type channel agonist, Bay K 8644, enhanced DSI, but in only 50 % of the cells (Pitler & Alger, 1992). We now report, that while nifedipine had no effect on DSI elicited by large depolarizing steps (to −5 or 0 mV), it reduced DSI elicited by small steps (to −25 or −20 mV) when QX-314 was omitted from the recording pipette solution. Some dihydropyridines can block a portion of an ω-conotoxin-sensitive component of Ca2+ current (Regan et al. 1991) as well as a sustained component of a low-voltage-activated calcium current (Avery & Johnston, 1996). However, because Ni2+ (100 μm), which blocks low-voltage-activated Ca2+ channels, did not block DSI induced by small voltage steps, we conclude that nifedipine's effect was not exerted through these channels. Moreover, it seems unlikely that nifedipine blocked a portion of the Ca2+ current through N-type channels, because it did not reduce IPSC amplitudes. Rather, it may be that certain voltage protocols elicit a postsynaptic response in which the role of L-type current is more important for DSI induction than it is under other protocols. With large voltage steps the role of L-type currents would be bypassed. One possible model would be that N-type current alone can trigger DSI, and that large voltage steps directly activate the N-type Ca2+ channels, which are mainly localized to the dendrites. In the absence of QX-314, small voltage steps applied at the soma would not depolarize the dendrites sufficiently to activate N-type channels directly, but would elicit unclamped Na+ and L-type Ca2+ current-dependent action potentials. These action potentials would boost the voltage delivered to the dendrites and thereby allow activation of the N-type channels, unless the L-type current component were depressed. Alternatively, DSI may normally be saturated by the Ca2+ that enters through N-type channels opened by large voltage steps, and the contribution of L-type Ca2+ influx only detectable at subsaturating levels.
These results, together with the distribution of N-type channels on dendrites and spines, are compatible with the idea that Ca2+ entry must occur at the dendrites to induce DSI. It is as yet unknown which VDCC, N or L type, plays the greater role under physiological conditions. Interestingly, Simmons et al. (1995) have reported that dynorphin release from dendrites of dentate gyrus granule cells is mediated by Ca2+ entry through both L- and N-type channels. Recently Le Beau & Alger (1998) have discovered a DSI-like suppression of IPSCs following low-Mg2+-induced epileptiform burst potentials, so the issue of L-type channel involvement in DSI can be addressed more directly.
These experiments suggested that Ca2+ entry through postsynaptic VDCCs is essential for induction of DSI, but did not rule out a role for Ca2+ release from intracellular stores. Because of the relatively slow onset of DSI (Pitler & Alger, 1992), and its rather long duration, we thought that Ca2+ release from intracellular stores might contribute to DSI. Neither CPA nor dantrolene sodium, both of which deplete intracellular Ca2+ stores, affected DSI. Nevertheless, intracellular Ca2+ imaging was beyond the scope of the present work; we do not have a positive control for the efficacy of these drugs in our experiments, and it remains conceivable that all or part of the long duration of DSI is mediated by Ca2+ released from intracellular stores. Alternatively, DSI could be triggered by a brief bolus influx of Ca2+ into the pyramidal cell, and its relatively slow kinetics the result of downstream effectors.
It is interesting that the ability of VDCC blockers to antagonize DSI was not obviously related to the degree to which they reduced ICa. ω-Conotoxin and ω-agatoxin were essentially equivalent in their ability to reduce ICa, but, whereas the former was highly effective in blocking DSI, the latter was completely ineffective. Nifedipine, which did not affect the DSI elicited by our standard voltage protocol, reduced ICa evoked by the same protocol very substantially. These data indicate that Ca2+ entry through various sources is not equipotent in eliciting DSI. N-type Ca2+ channels are localized predominantly on dendrites and dendritic spines (Westenbroek et al. 1992), and L-type channels are localized primarily in cell bodies and proximal dendrites. P-, Q- and R-type channels are also found on dendrites, however, and do not seem to play a role in DSI. It could be that Ca2+ influx must occur at precisely localized sites to trigger DSI.
In view of copious evidence that DSI involves a retrograde signalling process from the pyramidal cell to the interneuron, localized Ca2+ influx may trigger the release of a retrograde messenger. Recent work from both cerebellum (Glitsch et al. 1996) and hippocampus (Morishita et al. 1998) has revealed that a glutamate-like substance plays a key role in DSI induction. It is tempting to speculate that Ca2+ influx through VDCCs leads to the release of glutamate from sites in the dendritic regions of principal neurons, perhaps via the membrane-fusion-dependent process recently found to be critical in the establishment of LTP in hippocampal CA1 cells (Lledo et al. 1998).
We would like to thank F. E. N. Le Beau, S. E. Mason and W. Morishita for their comments on a draft of this manuscript. This work was supported by United States Public Health Service Grants NS30219 and NS22010 (B. E. A.). R. A. L. was supported by National Institutes of Health Neurosciences Training Grant NS07375. This manuscript will comprise part of a thesis submitted in partial fulfillment of the PhD degree requirements of R. A. L. We thank E. Elizabeth for expert word processing and editorial assistance.