An I_d-like current that is downregulated by Ca²⁺ modulates information coding at CA3–CA3 synapses in the rat hippocampus

Abstract

Voltage-gated K⁺ channels localised on presynaptic nerve terminals control information coding by modulating presynaptic firing and synaptic efficacy in target neurones. We found that at CA3–CA3 connections in hippocampal slice cultures, a fast-activating, slowly inactivating K⁺ conductance similar to the so-called delay current (I_D) is responsible for the delayed appearance of the first spike upon membrane depolarisation, for action potential repolarisation and for modulation of transmitter release. The I_d-like current was downregulated by intracellular Ca²⁺, as indicated by the increased delay in the appearance of the first action potential following either the block of Ca²⁺ flux through voltage-dependent Ca²⁺ channels with Cd²⁺ or replacement of the bathing solution with one devoid of Ca²⁺. In both cases, this effect was reversed by blocking this conductance with a low concentration of 4-aminopyridine (4-AP, 10-50 μM). Application of 4-AP shortened the delay to the first spike generation, prevented the effect of Cd²⁺ and increased the spike duration. The earlier appearance of the first action potential was also observed in the presence of dendrotoxin-1 (100 nM). In voltage-clamp experiments larger currents were recorded in the absence of extracellular Ca²⁺, thus confirming the downregulation of the I_d-like current by Ca²⁺ due to the positive shift of its inactivation. Spike broadening was associated with an enhancement of synaptic efficacy in target neurones, as assessed by the increase in EPSC amplitude and in the percentage of successes. Moreover, in the presence of 4-AP, EPSCs appeared with a longer latency and were more scattered. This conductance is therefore crucial for setting the timing and strength of synaptic transmission at CA3–CA3 connections. It is conceivable that switching off I_D by increasing intracellular Ca²⁺ following activity-dependent processes may facilitate network synchronisation and crosstalk between CA3 pyramidal cells, leading to seizure activity.

Presynaptic K⁺ channels are responsible for setting the resting potential, repolarising the membrane after action potentials and regulating the firing rate and information coding (Meir et al. 1999). By combining these different effects they play a crucial role in modulating transmitter release and in controlling synaptic efficacy in target neurones. In particular, by shaping presynaptic action potentials, K⁺ channels control the Ca⁺ signal necessary to trigger the fusion of synaptic vesicles with the surface membrane, exocytosis and transmitter release (Sabatini & Regher, 1997).

Much of our current knowledge on the functional correlation between changes in spike shaping and transmitter release comes from studies on invertebrates (Katz & Miledi, 1967a; Llinas et al. 1981; Augustine, 1990). A direct examination of mammalian central synapses has been precluded by the small size of presynaptic terminals, which are not easily accessible to patch pipettes. Only in a few cases it has been possible to directly measure K⁺ currents in nerve endings. Thus, at the calyx of Held synapse a fast-activating, delayed-rectifier K⁺ current has been found to be involved in the high-frequency firing of fast action potentials (Forsythe, 1994). Moreover, patch-clamp recording from mossy fibre boutons in rat hippocampal slices has allowed the characterisation of voltage-gated K⁺ channels that during activity accumulate in the inactivating state, leading to spike broadening and increases in Ca²⁺ fluxes and synaptic efficacy (Geiger & Jonas, 2000). At cerebellar basket cell terminals, dendrotoxin (DTX)-sensitive K⁺ channels have been identified (Southan & Robertson, 1998), blockage of which produces a dramatic increase in both the number and amplitude of spontaneous IPSCs.

Although the shape of the action potential may change between soma and terminals (Bourque, 1990; Geiger & Jonas, 2000), an indirect estimate of how spike repolarisation and presynaptic firing affect transmitter release and synaptic efficacy can be inferred using simultaneous recordings from two synaptically connected neurones. To this purpose, hippocampal slice cultures provide an ideal model. Hence, while maintaining morphological and functional features that are similar to those of native brain tissue, they express a higher connectivity, which increases the probability of finding monosynaptically coupled cells (Debanne et al. 1995; Pavlidis & Madison, 1999). In preparations such as this, a fast-activating and slowly inactivating voltage-dependent K⁺ current that is very sensitive to 4-aminopyridine (4-AP), named I_K(AT), has been described in CA3 pyramidal neurones (Bossu & Gähwiler, 1996; Bossu et al. 1996). In the present study we have characterised a current that appears to be similar to the ‘delay’ current (I_D) described in hippocampal neurones by Storm (1988), so called because of its involvement in the delayed appearance of the first spike during membrane depolarisation. We have found that changes in membrane potential and intracellular Ca²⁺ concentration up- or downregulate I_D in CA3 pyramidal cells. Moreover, blocking I_D by applying low concentrations of 4-AP resulted in increased cell firing, spike broadening and enhanced transmitter release, suggesting that this conductance is fundamental in determining the timing and strength of synaptic transmission at CA3-CA3 synapses. Small changes in membrane potential or intracellular Ca²⁺ concentration may turn this conductance on and off, leading to modifications of synaptic efficacy.

METHODS

Organotypic hippocampal slice cultures

Hippocampi were removed from 4- to 7-day-old rats after they had been killed by decapitation, and organotypic cultures were prepared following a method that is described in detail elsewhere (Gähwiler, 1981). The procedure was carried out in accordance with the regulations of the Italian Animal Welfare Act and was approved by the local authority veterinary service. Transverse 400 μm thick slices were cut with a tissue chopper and attached to coverslips in a film of reconstituted chicken plasma (Cocalico, Reamstown, PA, USA) clotted with thrombin (Sigma, Milan, Italy). The coverslips were transferred to plastic tubes containing 0.75 ml of medium. The tubes were placed in a roller drum (6 revs h⁻¹) inside an incubator at 36 °C. The medium contained: basal medium (BME, Eagle, with Hanks’ Salts, without L-glutamine; Gibco, 100 ml), Hanks’ balanced salt solution (Gibco, 50 ml), horse serum (Gibco, 50 ml), L-glutamine (Gibco, 200 mM, 1 ml) and 50 % D-glucose in sterile water for tissue culture (Gibco, 2 ml).

Electrophysiological recordings

After 10-14 days in an incubator, the slices, which had flattened to near-monolayer thickness, were transferred to a recording chamber that had been fixed to the stage of an upright microscope. The cultured slice in the recording chamber was superfused at room temperature (22-24 °C) with a bath solution containing (mM): NaCl 150, KCl 3, CaCl₂ 2, MgCl₂ 1, Hepes 10, glucose 10 (pH 7.3, adjusted with NaOH). Electrophysiological experiments were performed on CA3 pyramidal cells using the whole-cell configuration of the patch-clamp technique in current or voltage-clamp mode. CA3 neurones were identified both visually and on the basis of their firing properties (i.e. their ability to accommodate in response to long (800 ms) depolarising current pulses). The identity of the connected cells as pyramidal cells was confirmed in some experiments in which cells were labelled with neurobiotin.

Patch electrodes were pulled from borosilicate glass capillaries (Hingenberg, Malsfeld, Germany). They had a resistance of 3-6 MΩ when filled with an intracellular solution containing (mM): KMeSO₄ 135 (125 when adding 10 mM BAPTA), KCl 10, Hepes 10, MgCl₂ 1, Na₂ ATP 2, Na₂ GTP 0.4 (solution 1) for the presynaptic neurone or CsMeSO₄ 135, CsCl 10, Hepes 10, Qx314 (a quaternary amine derivative of lidocaine) 5, EGTA 0.5, MgATP 2, NaGTP 0.3 (solution 2) for the postsynaptic one; pH was adjusted to 7.3 with KOH and CsOH, respectively. The stability of the patch was checked by repetitively monitoring the input and series resistances during the experiment. Cells exhibiting 15-20 % changes were excluded from the analysis. Single or pairs (50 ms interval) of action potentials were evoked in current-clamp mode by short (5 ms) depolarising current pulses at 0.05 Hz in the presence of 10 nM tetrodotoxin (TTX, Affinity Research Products, Nottingham, UK) in order to reduce polysynaptic activity. EPSCs were recorded from the postsynaptic neurones in voltage-clamp mode at a holding potential of -60 mV. In some experiments, a Ca²⁺-free solution was used in which CaCl₂ was omitted and replaced by MgCl₂ (5 mM) and EGTA (1 mM). Voltage-clamp experiments were also performed on presynaptic neurones (filling solution 1) in the presence of extracellular CsCl (3 mM), carbachol (50 μM, Sigma, Milan, Italy) and TTX (0.5 μM), in order to block I_h, I_M and the Na⁺ channels, respectively. In some of these experiments, CdCl₂ (200 μM) was added in order to reduce the activation of Ca²⁺-dependent K⁺ channels. The membrane potential values were not corrected for a liquid junction potential of 9 mV. The following drugs were used: 4-AP (Sigma); DTX-I (Alomone Labs, Jerusalem, Israel); DL-2-amino-5-phosphonopentaoic acid (D-AP5), 6,7-dinitroquinoxaline-2,3-dione (DNQX), bicuculline and CGP 55845 (all purchased from Tocris Cookson, Bristol, UK).

Data acquisition and analysis

Data were stored on magnetic tape and transferred to a computer after digitisation with an A/D converter (Digidata 1200). Data acquisition was achieved using pClamp (Axon Instruments, Foster City, CA, USA). Data were sampled at 20-100 kHz and filtered with a cut-off frequency of 1 kHz. Series resistance compensation was used only for current-clamp recordings.

Voltage responses were analysed with Clampfit software (Axon Instruments). Action potentials were characterised by their firing threshold, their amplitude (from threshold to peak) and their width at the threshold level. The delay to the first spike generation was evaluated as the time interval between the spike peak and the beginning of the depolarising current step. Input resistance was calculated by measuring the amplitude of voltage responses to hyperpolarising current steps.

EPSCs were analysed with AxoGraph 4.6 Program (Axon Instruments), which uses a detection algorithm based on a sliding template. The onset of the EPSC was given by the intersection of a line through the 10 and 90 % of the EPSC rise time with the baseline. The same program was used to fit the decay phase of the EPSCs with a monoexponential function. Failures were identified visually. Onset, rise and decay times were calculated after averaging only the successes. EPSC latency was calculated as the time gap between the onset of the mean EPSC and the peak of the presynaptic spike. An estimate of the jitter of the responses was obtained by calculating the standard deviation of the latencies of all the responses. Mean EPSC amplitude was obtained by averaging successes and failures. The paired-pulse ratio (PPR) was calculated as the ratio between the mean amplitude of EPSC2 over EPSC1. The coefficient of variation (CV) of the response amplitude was determined as the ratio between standard deviation and mean, and its inverse squared value (CV⁻²) was calculated.

Activation of the ‘macroscopic’ outward K⁺ current was studied by applying a conditioning pre-pulse to -110 mV (500 ms duration) followed by a depolarising voltage step from -80 to +20 mV (800 ms duration, 10 mV increment); inactivation of this current was studied by applying conditioning pre-pulses from -100 to -20 mV (10 mV increment) followed by a depolarising step to +30 mV. The normalised current amplitude was fitted according to a Boltzmann equation of the form: I/I_max = 1/[1 + exp±(V - V_1/2)/k] for activation (-) and inactivation (+), respectively. In the equation, V_1/2 is the voltage at which the current is half of its maximum and k indicates the slope of the voltage dependence of the processes. The time course of the removal of inactivation was studied by a fixed depolarisation to +20 mV following a pre-pulse of variable duration (between 20 and 10 000 ms) to -110 mV from a holding potential of -60 mV. Current amplitudes were normalised relative to the amplitude of the 10 s conditioning pre-pulse and the average normalised currents were plotted as a function of the conditioning pre-pulse duration.

The 4-AP- or DTX-I-sensitive currents were obtained by subtracting the current recorded in the presence and absence of these drugs.

Values are given as mean ± S.E.M. The significance of differences was assessed by Student's t test or the Wilcoxon test. The differences were considered significant at P < 0.05.

RESULTS

A voltage-dependent K⁺ current similar to I_D is responsible for the delayed appearance of action potentials during membrane depolarisation

Patch-clamp recordings were made in whole-cell configuration and current-clamp mode from CA3 pyramidal neurones in organotypic hippocampal slice cultures. These neurones were identified as principal cells both visually (using infrared differential interference contrast video microscopy) and on the basis of their firing properties (i.e. their ability to accommodate in response to long depolarising current pulses). In some experiments (n = 16) cells were identified morphologically as pyramidal neurones by neurobiotin injection. In order to investigate their firing properties, current steps of different amplitudes (800 ms long) were injected. As shown in Fig. 1A, a long depolarising current step applied from the resting membrane potential induced the generation of a single spike followed by a slow afterhyperpolarisation (sAHP). By increasing the current intensity, the firing of the cell increased, showing a strong adaptation followed by a pronounced sAHP (Fig. 1B). Interestingly, a certain delay in the generation of the first spike was always observed. Out of 54 neurones with a mean resting potential of -56.4 ± 0.7 mV, the firing was generated by a minimal current step of 380 ± 30 pA (varying between 100 and 1100 pA) with a delay of 340 ± 30 ms (varying between 47 and 790 ms). The minimal intensity of current required from each cell to fire was correlated with its input resistance (r = -0.60, n = 42, P < 0.001), whereas the delay in spike appearance was correlated with the resting potential (r = 0.34, n = 54, P < 0.05). Thus, increasing the intensity of the current steps caused a reduction in the delay (Fig. 1A-B), and hyperpolarising the membrane below the resting level (≈ 10 mV) caused an increase in the delay (Fig. 1C). In 12 neurones, changing the membrane potential from -56 ± 1 to -66 ± 1 mV induced a significant increase in the delay of the first spike from 100 ± 20 to 540 ± 60 ms (P < 0.001; Fig. 1D), when comparing the same current steps. At the single spike level, no changes in spike amplitude were observed (67 ± 3 mV at both potentials; n = 8). However, action potentials evoked from the hyperpolarised level exhibited a longer time to threshold (from 2.1 ± 0.2 to 3.0 ± 0.3 ms, n = 8, P < 0.05) and a reduction in width (from 3.3 ± 0.1 to 2.59 ± 0.09 ms, n = 8, P < 0.001; see Bossu et al. 1996). These results suggest the involvement of a voltage-dependent conductance in these phenomena. One possible candidate is the fast-activating, slowly inactivating voltage-dependent K⁺ current known as the ‘delay current’, or I_D, for its peculiar ability of retarding the appearance of the first spike upon prolonged membrane depolarisation (Storm, 1988). This current is known to be very sensitive to micromolar concentrations of 4-AP (Rudy, 1988; Storm, 1988, 1990). To see whether I_D was responsible for the delayed appearance of action potentials, 4-AP (10-50 μM) was applied in the bath. 4-AP significantly reduced the delay of the appearance of the first spike from 310 ± 50 to 50 ± 10 ms (n = 14, P < 0.001; Fig. 1E-G). This effect was associated with an enhancement of spontaneously occurring synaptic events and a significant increase in input resistance (160 ± 20 MΩ and 180 ± 20 MΩ in the absence and presence of 4-AP, respectively; n = 13, P < 0.05). 4-AP significantly reduced the spike amplitude (from 67 ± 2 to 65 ± 2 mV, n = 19, P < 0.05) and increased spike duration (from 2.39 ± 0.07 ms to 3.2 ± 0.2 ms, n = 20, P < 0.001; see Fig. 4A and Fig. 10 in Bossu et al. 1996). Often, 4-AP produced small fluctuations of membrane potential due to an increase in network activity. In these cases, appropriate current injections were applied in order to maintain the membrane potential at the same resting level and to compare the firing properties in the two different conditions. The increase in spontaneous synaptic activity with 4-AP was not responsible for the observed effects on spike delay and duration since similar modifications were observed when 4-AP was applied in the presence of AMPA, NMDA, GABA_A and GABA_B receptor antagonists (DNQX 20 μM, D-AP5 50 μM, bicuculline 10 μM and CGP 55845 10 μM, respectively). In these conditions, 4-AP (25 μM) significantly reduced the delay of the appearance of the first spike from 500 ± 100 to 25 ± 7 ms and increased the spike duration by 30 ± 10 % (n = 4; data not shown; P < 0.05). As it has been shown that I_D is sensitive to DTX-I (Storm, 1990; Luthi et al. 1996), some experiments were performed in the absence or presence of this toxin. Similarly to 4-AP, DTX-I (100 nM) increased spontaneous synaptic activity and significantly reduced the time of appearance of the first spike elicited in current-clamp experiments by a depolarising current step (from 620 ± 40 to 60 ± 20 ms, n = 4; P < 0.05; Fig. 1H-J). These data indicate that indeed an I_D-like current, activated from the resting membrane potential, is responsible for regulating the appearance of the first spike and, as expected (Storm, 1988), contributes to spike repolarisation.

Figure 1. An I_D-like current activated from the resting membrane potential is blocked by low doses of 4-AP.

A, firing response upon injection of a steady inward current (0.1 nA for 800 ms) from a resting membrane potential of -61 mV. Notice the afterhyperpolarisation at the end of the pulse. B, increasing the stimulation intensity (0.2 nA) produced in the same neurone repetitive firing with strong adaptation followed by a prominent afterhyperpolarisation. C, upon membrane hyperpolarisation (-68 mV), the same neurone responded to a current step of 0.2 nA with a reduced number of spikes and with a larger delay in the appearance of the first action potential. Insets in A-C represent the same traces on an expanded time scale. D, mean delay to the appearance of the first spike from the resting membrane potential (control, ctl, -56 ± 1 mV; n = 12) and from a more hyperpolarised membrane potential (hyper, -66 ± 1 mV). E and F, voltage responses to a 0.2 nA current step in control conditions (E) and in the presence of 4-AP (F) for a different neurone. G, mean delay to the generation of the first spike in control conditions and in the presence of 4-AP (n = 14). H and I, firing responses to a 0.2 nA current step in control conditions (H) and in the presence of DTX-I (DTX, I; scale bar as in E and F). J, mean delay to the first spike appearance in control conditions and in the presence of DTX-I (n = 4). *P < 0.05, **P < 0.001.

Figure 4. The I_D-like current is involved in action potential repolarisation.

Action potentials were generated from rest by injection of short (5 ms duration) current steps, and spike width was calculated at the threshold level. Action potentials generated in different conditions are aligned at the threshold and are superimposed (left). The corresponding changes in width are shown on the right. A, application of 4-AP significantly broadened the spike (n = 20). Thick lines correspond to spike recorded in the presence of 4-AP. B and C, application of Cd²⁺ did not affect action potential width (n = 10, B), while further addition of 4-AP to Cd²⁺ increased the spike duration (n = 5; C). In B and C, thick lines correspond to action potentials recorded in the presence of Cd²⁺ alone. *P < 0.05; **P < 0.001.

Upregulation of an I_D-like K⁺ current following blockade of Ca²⁺ fluxes through voltage-activated Ca²⁺ channels

Bath perfusion with the non-selective voltage-dependent Ca²⁺ channel blocker Cd²⁺ (200 μM) blocked the sAHP and reduced spike adaptation (Fig. 2), as usually observed in pyramidal neurones (Madison & Nicoll, 1984; Storm, 1989). However, in contrast to previous observations (Madison & Nicoll, 1984; Storm, 1989), in the presence of the divalent cation, the number of spikes evoked by long (800 ms) depolarising current pulses was reduced (Fig. 2B and D). Surprisingly, application of Cd²⁺ significantly increased the delay of the appearance of the first spike from 170 ± 40 to 570 ± 40 ms, with a mean increase of 400 ± 40 ms (n = 10, P < 0.001; Fig. 2C). This effect was reversed by the addition of a low concentration (10-50 μM) of 4-AP (Fig. 2E-F). In the presence of Cd²⁺ and 4-AP, the cells were able to fire with much shorter delays (on average from 540 ± 60 to 12 ± 3 ms, n = 6, P < 0.001). These results suggest the possibility that the I_D-like current is downregulated by an increase in intracellular Ca²⁺ through voltage-dependent Ca²⁺ channels, as the blockade of these channels by Cd²⁺ increased the delay in spike generation. To see whether blocking Ca²⁺ flux directly affects I_D, occlusion experiments were performed with 4-AP and Cd²⁺. While addition of 4-AP to the control solution significantly reduced the delay of the first spike (from 290 ± 40 to 50 ± 20 ms, n = 6, P < 0.05), further application of Cd²⁺ did not retard spike generation (from 50 ± 20 to 40 ± 20 ms; in 4-AP or 4-AP and Cd²⁺, respectively; Fig. 3). Surprisingly, in spite of a clear effect on firing properties, Cd²⁺ did not affect spike width (2.6 ± 0.1 and 2.57 ± 0.09 ms, in control conditions and in Cd²⁺, respectively; n = 10; Fig. 4B). However, addition of 4-AP to Cd²⁺ produced a significant broadening of the action potential (from 2.41 ± 0.09 to 3.5 ± 0.1 ms, n = 5, P < 0.05; Fig. 4C).

Figure 2. Intracellular Ca²⁺ downregulates the I_D-like K⁺ current.

A and B, firing patterns upon injection of a 0.1 nA current step in control conditions (A) and in the presence of Cd²⁺ (B). Insets represent the same traces on an expanded time scale. C, mean delay to the first spike generation in control conditions and in the presence of Cd²⁺ (n = 10). D, two superimposed responses of another neurone to a 0.3 nA current step in control solution (thin line) and in the presence of Cd²⁺ (thick line). Note the lack of action potentials in Cd²⁺. E, same neurone as in D in the presence of Cd²⁺+ 4-AP. F, mean delay in the generation of the first spike in the presence of Cd²⁺ or Cd²⁺+ 4-AP for the same current injections (n = 6). **P < 0.001.

Figure 3. Blocking the I_D-like current with 4-AP prevents the effect of Cd²⁺.

A, firing response upon injection of a steady inward current (0.4 nA) from the resting membrane potential. B, 4-AP increased the firing rate and reduced the delay of the first spike. C, further application of Cd²⁺ did not produce any increase in spike delay, in line with the occlusion of the Cd²⁺ effect by 4-AP. D, mean delay to the appearance of the first spike in control conditions, in the presence of 4-AP or in the presence of 4-AP + Cd²⁺ (n = 6).

To further investigate the Ca²⁺ hypothesis, additional experiments (n = 7) were performed by replacing the control solution with one devoid of Ca²⁺. In the absence of extracellular Ca²⁺ the delay of the first spike increased significantly from 370 ± 100 to 540 ± 70 ms (P < 0.05; data not shown). Also in this case, further addition of 4-AP induced a reduction in the appearance of the first spike (from 470 ± 100 to 50 ± 10 ms; n = 5, P < 0.05). These results further suggest a downregulation of the I_D-like current by intracellular Ca²⁺. In line with this, we found that neurones loaded with the Ca²⁺ chelator BAPTA (10 mM) had, in comparison with controls, significantly more hyperpolarised resting potentials (-63 ± 2 mV, n = 10, P < 0.05; data not shown). This difference could be attributed to the upregulation of the I_D-like current following the reduced concentration of intracellular Ca²⁺ even if we cannot rule out the possibility that BAPTA affects a different outward K⁺ current (Lancaster & Batchelor, 2000). These neurones still exhibited a delay in spike generation (490 ± 90 ms) indicating that the I_D-like current could be activated from rest. The effectiveness of BAPTA in chelating intracellular Ca²⁺ was assessed by its capability to block the sAHP in all cells tested.

Voltage dependence of the macroscopic K⁺ current

An outward, fast-activating, slowly inactivating, voltage-dependent K⁺ current that is sensitive to low concentration of 4-AP has already been described in CA3 neurones from organotypic hippocampal slices using voltage-clamp recordings (Bossu & Gähwiler, 1996; Bossu et al. 1996). Therefore, a first set of experiments (n = 9) in voltage-clamp mode was performed in order to compare the I_D-like current with the previously studied one. ‘Macroscopic’ outward K⁺ currents were elicited by 10 mV depolarising voltage steps from a holding potential of -60 mV, in the presence of extracellular CsCl (3 mM), carbachol (50 μM), TTX (0.5 μM) and CdCl₂ (200 μM) in order to block I_h, I_M, Na⁺ channels, Ca²⁺ channels and to reduce the activation of Ca²⁺-activated K⁺ channels, respectively. As shown in the example of Fig. 5A, the 4-AP-sensitive component was obtained by subtracting the outward currents evoked in the presence of the drug from the control ones. Like I_D, the 4-AP-sensitive component showed fast-activating and slowly inactivating kinetics. As expected (Storm, 1990), similar results were obtained with DTX-I (n = 3, data not shown). Moreover, the activation and inactivation curves of the ‘macroscopic’ outward K⁺ currents gave potentials for half-inactivation and half-activation of -48.6 and 10.5 mV, respectively, while the slope factors were 9.4 and 12.7 mV for inactivation and activation, respectively (not shown). These values are very close to those obtained previously by Bossu et al. (1996) for the ‘ensemble’ current. In addition, in agreement with Bossu et al. (1996), the existence of a window current above the resting membrane potential was disclosed from activation and inactivation curves, suggesting the involvement of this macroscopic current in stabilisation of the resting membrane potential. Moreover, it appears that fluctuations around the resting membrane potential would strongly modify the inactivating state of this current.

Figure 5. Ca²⁺ sensitivity and inactivation properties of the I_D-like current.

A, outward currents elicited by depolarising voltage steps from a holding potential of -60 mV (800 ms duration, 10 mV increment, bottom left), in control conditions and in the presence of 4-AP. On the bottom right, 4-AP-sensitive currents were obtained by subtracting from the control traces those recorded in 4-AP. B, mean current amplitudes obtained by a depolarising step to +20 mV after conditioning pre-pulses of different durations to - 110 mV have been normalised to the current recorded after the 10 s pre-pulse and plotted as a function of pre-pulse duration (n = 6). Data points were fitted with a biexponential function, having time constants of 12 and 714 ms. C, activation (circles) and inactivation (squares) curves for the I_D-like current in the presence (filled symbols) or in the absence (open symbols) of extracellular Ca²⁺. Data points were obtained following normalisation to the maximal current. Curves represent Boltzmann equation fits, obtained as described in Methods. The half-activation voltages were -37.4 and -34.6 mV and the slopes were 8.5 and 10.4 mV, with and without extracellular Ca²⁺, respectively (n = 11). In the absence of Ca²⁺, the inactivation curve was shifted in a positive direction (half-activation voltages were -80.5 and -55.7 mV and slopes were 9.4 and 16.4 mV in the presence or absence of Ca²⁺, respectively; n = 3). D, 4-AP-sensitive current amplitudes obtained in the presence and in the absence of Ca²⁺ at different membrane potentials (n = 11). *P < 0.05.

To further characterise the properties of the isolated I_D-like current, we investigated the time course of the removal of inactivation (Fig. 5B). As shown in Fig. 5, the recovery from inactivation could be fitted with two exponential functions with time constants of 12 and 714 ms (n = 6). In this regard, this current behaves differently from both I_K(AT) and I_D, whose complete recovery from inactivation was observed after 150 ms and 5 s, respectively. The slower recovery observed by Storm (1988) may reflect a contribution of the slowly inactivating 4-AP-insensitive I_K(slow) described by Luthi et al. (1996) in organotypic hippocampal slices.

Ca²⁺ sensitivity of the I_D-like current

The increased delay in the appearance of the first spike in the presence of Cd²⁺ or in the absence of extracellular Ca²⁺ (when Ca²⁺ was substituted by Mg²⁺) suggests that the I_D-like current can be downregulated by changes in intracellular Ca²⁺ concentration (see also Wu & Barish, 1999, 2000). To see whether this was indeed the case, activation and inactivation of 4-AP-sensitive K⁺ currents were studied in two groups of 11 neurones: one exposed to a control medium, the other to a nominally Ca²⁺-free solution. In both cases TTX, CsCl and carbachol were present in the bathing solution. As shown in Fig. 5C (filled symbols), the activation properties of the isolated I_D-like current were similar in both experimental conditions (in the presence or absence of Ca²⁺, half-activation voltages were -37.4 mV and -34.6 mV and slopes were 8.5 mV and 10.4 mV). In contrast, in the absence of Ca²⁺the inactivation curve was shifted towards more positive membrane potentials (in the presence or absence of Ca²⁺, half-activation voltages were -80.5 mV and -55.7 mV, respectively, and slopes were 9.4 mV and 16.4 mV, respectively; Fig. 5C, open symbols). In line with this, the absolute currents recorded between -50 and +10 mV were significantly larger in the absence of Ca²⁺ (Fig. 5D; P < 0.05).

The I_D-like current modulates neurotransmitter release

The broadening of the action potential induced by low concentrations of 4-AP reveals a role for the I_D-like current in spike repolarisation and suggests strongly the involvement of this conductance in modulating transmitter release. In order to properly study this possibility, double patch-clamp recordings from interconnected CA3 pyramidal neurones were performed. Out of 15 double recordings analysed, the mean amplitude of successes was 50 ± 10 pA, with a median of 28 pA and a mean percentage of successes of 60 ± 10 % (see Pavlidis & Madison, 1999). According to Debanne et al. (1995), the criteria for monosynaptically connected pairs consist of short-latency responses (4.9 ± 0.9 ms, n = 15) with small fluctuations. Seven different pairs of neurones were studied in control conditions and in the presence of a low concentration of 4-AP (10 μM). In general, pairs of presynaptic action potentials (50 ms apart) delivered at a frequency of 0.05 Hz, evoked two sequential EPSCs that fluctuated in amplitude from trial to trial. As shown in the example in Fig. 6A and C, a transmitter failure to the first spike was almost always associated with a success to the second one. On the contrary, a success to the first spike was associated with a synaptic response to the second one, whose size was half of the first EPSC. Application of 4-AP (10 μM) produced a broadening of the action potentials and strongly increased the number of successes to the first spike (Fig. 6B and C). Moreover, in 4-AP some delayed responses appeared on the top of EPSCs (see arrows in Fig. 6B), suggesting either a ‘delayed’ release or the activation of previously silent connections. Overall, 4-AP induced in all cells tested, spike broadening (from 16 to 53 %, with a mean of 29 ± 6 %, n = 7, P < 0.05) and a significant increase in the percentage of successes to the first pulse (from 51 ± 13 to 70 ± 12 %, P < 0.05; Fig. 6D). In line with a presynaptic effect, 4-AP decreased the EPSC PPR from 2.5 ± 0.9 to 1.2 ± 0.4 (n = 5; Fig. 6E) and significantly increased CV⁻² by 250 ± 100 % (n = 7, P < 0.05).

Figure 6. The I_D-like current modulates neurotransmitter release at CA3-CA3 connections.

A and B, pairs of action potentials were generated (50 ms intervals, 0.05 Hz) in the presynaptic cell (upper traces) while EPSCs were recorded from the postsynaptic cell (lower traces) in control conditions (A) and in the presence of 4-AP (B). Eight traces are superimposed. C, time course of the peak amplitude of the first (open circles) and second (filled circles) EPSCs recorded from the cell shown in A and B. D, percentage of successes to the first (white column) and to the second (black column) spike in control conditions and in the presence of 4-AP (5-7 pairs). E, mean PPR calculated from five different pairs of neurones in the two experimental conditions. *P < 0.05.

Figure 7 shows an example in which no failures were detected in control conditions. In this case, 4-AP strongly increased EPSC amplitude (Fig. 7A-C), indicating an increase of synaptic efficacy by blocking the I_D-like current. Interestingly, in 4-AP, the EPSC latency increased significantly, as demonstrated by a latency distribution histogram (Fig. 7D). Overall, a change in the mean latency from 5 ± 1 to 6 ± 2 ms was observed (Fig. 7E). This was associated with a small increase in the scattering of response latencies (on average the standard deviation of EPSC latencies was 0.38 ± 0.8 and 0.5 ± 0.2 ms in control and in 4-AP, respectively; n = 4).

Figure 7. The I_D-like current increases the latency of postsynaptic responses.

A and B, eight superimposed action potentials generated (0.05 Hz) in the presynaptic cell (upper traces) while EPSCs are recorded from the postsynaptic cell (lower traces) in control conditions (A) and in the presence of 4-AP (B). C, time course of EPSC amplitude in control conditions and during the application of 4-AP for the pair shown in A and B. D, normalised latency distributions in control (white columns) and in 4-AP (black columns) for the same neurone shown in A-C (bin size = 0.08 ms). The latency was calculated as the time interval between the onset of the response and the peak of the presynaptic spike. In this case the average latency increased from 2.9 to 3.4 ms. E, mean latency of postsynaptic responses obtained from seven different pairs. For each experiment, the latency was calculated after averaging all of the successes. *P < 0.05.

According to the voltage-dependence of I_D and its role in the modulation of synaptic efficacy, additional double-patch recordings were performed in order to check whether small changes in the resting potential could modulate neurotransmitter release by removing the inactivation of I_D in the presynaptic neurone (n = 8). As already mentioned, 10 mV hyperpolarisation significantly reduced the action potential duration, with relative changes ranging from 13 to 28 % and a mean of 20 ± 2 % (n = 8, Fig. 8D). Surprisingly, only the cell in which hyperpolarisation caused a 28 % reduction in action potential width showed a simultaneous reduction in the percentage of successes (from 97 to 87 %, Fig. 8A-C), suggesting a role for I_D in this phenomenon. However, the remaining seven neurones showed a small increase in the percentage of successes, giving an overall change from 72 ± 9 to 78 ± 9 % (Fig. 8E). No significant changes were detected in the kinetics or in the amplitude of EPSCs (P > 0.05). In five double patches in which a paired-pulse protocol was applied, the increased percentage of successes to the first and second EPSC was accompanied by a slight, but significant decrease in the PPR (from 1.5 ± 0.2 to 1.2 ± 0.2, n = 5, P < 0.05). These results suggest that different processes involved in neurotransmitter release are modulated in opposite directions by changing the membrane potential in the presynaptic neurone.

Figure 8. Modulation of neurotransmitter release by changes in presynaptic membrane potential.

A and B, eight action potentials were generated from holding potentials of -50 (A) and -60 mV (B). Corresponding EPSCs were recorded from the postsynaptic cell (middle traces, A and B) and their means over the whole recording period are shown for each condition (lower traces, A, B). C, percentage of successes for the neurones showed in A and B at -50 and -60 mV, respectively. Note that the percentage of successes is reduced at the more hyperpolarised membrane potential. D and E, mean action potential width (D) and mean percentage of successes (E) calculated for eight pairs at resting membrane potential (ctl, -52 ± 1 mV) and at a more hyperpolarised level (hyper, -62 ± 1 mV). **P < 0.001.

DISCUSSION

In the study reported here we have described a voltage- and Ca²⁺-dependent K⁺ current with biophysical and pharmacological properties similar to I_D that plays a crucial role in regulating the firing properties of neurones, spike repolarisation and transmitter release at CA3-CA3 connections in the rat hippocampus. This I_D-like current contributes strongly to a macroscopic current previously described by Bossu et al. (1996) and shows fast activation and slow inactivation kinetics and a pharmacological profile that conforms to that of I_D, which has been described previously in both CA1 (Storm, 1988) and CA3 (Luthi et al. 1996; Mitterdorfer & Bean, 2002) hippocampal neurones.

Functional voltage-gated K⁺ channels are formed by homo- or heteromeric tetramers of principal α subunits, each of them consisting of six transmembrane helices that are often supplemented by accessory β subunits (Jan & Jan, 1997). Several genes, grouped into four main subfamilies, Kv1, Kv2, Kv3 and Kv4, have been cloned so far (Hille, 2001). Although the genes coding for the pore-forming α-subunit polypeptides mediating I_D are still unknown, the sensitivity of the I_D-like current to DTX-I strongly suggests a role for some non-inactivating Kv1 members (Kv1.1, Kv1.2 and Kv1.6; Harvey, 2001). Therefore, it is likely that the co-assembly of these subunits with other α or β subunits may account for the slowly inactivating I_D-like current we have described. Indeed, it has been shown that Kv1.1, Kv1.2 and Kv1.4 can form heteromeric channels in vivo (Sheng et al. 1993; Wang et al. 1994; Rhodes et al. 1995) and that the addition of β subunits confers inactivation properties to delayed-rectifier K⁺ channels (Robertson, 1997).

Downregulation of the I_D-like current by intracellular Ca²⁺

Several lines of evidence suggest that the I_D-like current is downregulated by intracellular Ca²⁺: (1) blocking Ca²⁺ flux through voltage-dependent Ca²⁺ channels with Cd²⁺ increases the delay of the first spike, an effect that is reversed by a low concentration of 4-AP; (2) blocking I_D with 4-AP prevents the effect of Cd²⁺ on spike delay; (3) in Ca²⁺-free solution, an increased delay in the appearance of the first spike is observed that is further reduced by the addition of 4-AP; (4) neurones loaded with BAPTA exhibit a more hyperpolarising resting potential, in line with an upregulation of the I_D-like current. These results have been confirmed by voltage-clamp experiments showing that a significantly larger 4-AP-sensitive K⁺ current could be activated in the absence of extracellular Ca²⁺. This can be explained by the shift of the inactivation curve towards more depolarised potentials.

In agreement with this result, suppression of a K⁺ current similar to I_A by intracellular Ca²⁺ has been described (Chen & Wong, 1991). Moreover, a Ca²⁺-induced downregulation of I_D has recently been inferred by the observation that activation of metabotropic glutamate receptors (types I and II), which are known to trigger Ca²⁺-dependent mechanisms, was able to inhibit it (Wu & Barish, 1999). In addition, Wu & Barish (2000) used Ca²⁺-sensitive dyes to show that I_D amplitude increases when intracellular Ca²⁺ is reduced. However, the mechanisms responsible for the Ca²⁺ modulation of the I_D-like current are as yet unknown. It has been shown that the channel proteins responsible for I_A bind proteins that act as Ca²⁺ sensors (An et al. 2000). By coupling I_A to activity-dependent fluctuations of intracellular Ca²⁺, these proteins may dynamically regulate cell excitability. A similar mechanism may account for our observations and for the proposed switch of the native I_K(AT) channel between two different gating modes expressing either sustained or transient inactivation kinetics (Bossu & Gähwiler, 1996). The switch would be controlled by intracellular factors such as Ca²⁺-dependent kinases or phosphatases. A similar bimodal behaviour has been described for the channel composed of Kv1.1 and Kv β 1.1 subunits expressed in Xenopus oocytes. In this case, the underlying current had a fast-activating component that could be reduced by dephosphorylation (Singer-Lahat et al. 1999).

While Cd²⁺ strongly delayed or abolished spikes evoked by sustained membrane depolarisation, it did not affect spike width. This apparent discrepancy can be explained by the fact that Ca²⁺ produces opposite effects on different K⁺ currents: it suppresses I_D and activates the Ca²⁺-dependent K⁺ currents (I_C) involved in spike repolarisation (Storm, 1989). Thus, the expected decrease in spike width due to the upregulation of I_D (following block of Ca²⁺ flux) would be counterbalanced by its increase due to the lack of activation of I_C. The relative contribution of these two opposing effects would depend on the relative affinity of different K⁺ channels for Ca²⁺, the time course of Ca²⁺ rise, the location of voltage-dependent Ca²⁺ channels relative to K⁺ channels and the spatial distribution of Ca²⁺ microdomains.

A presynaptic I_D-like current regulates synaptic efficacy in target neurones

Changes in spike repolarisation should affect transmitter release at nerve endings. Indeed, paired recordings from interconnected neurones have demonstrated clearly that an I_D-like conductance regulates synaptic efficacy at CA3-CA3 connections. The effect of 4-AP on EPSCs was presynaptic in origin, as shown by the decrease in transmitter failures and PPR and by the increase in CV⁻², all of which are considered traditional indices of presynaptic modifications (Katz, 1969; Zucker, 1989). In particular, in a paired-pulse protocol, the PPR is inversely related to the initial release probability (Dobrunz & Stevens, 1997). Thus, it is likely that the observed reduction in PPR reflects an increased probability of release during 4-AP application.

Distinct sets of K⁺ channel localised on synaptic terminals have been revealed with the aid of light and electron microscopy techniques (Roeper & Pongs, 1996). In particular, Kv1 channel complexes have been shown to be predominantly localised on CA3 axon terminals (Grissmer et al. 1994; Wang et al. 1994; Monaghan et al. 2001) where they regulate the excitability of the CA3 recurrent axon collateral system (Smart et al. 1998).

In the present experiments, very low concentrations of 4-AP were able to broaden the presynaptic spike at the somatic level and to increase synaptic efficacy. Even if the shape of the action potential changes between the soma and the terminals (Bourque, 1990; Geiger & Jonas, 2000), the concomitant increase in synaptic efficacy suggests strongly that similar changes in spike width occur at nerve terminals. Thus, broadening of presynaptic action potentials increases Ca²⁺ influx primarily by prolonging the duration of presynaptic Ca²⁺ currents (Sabatini & Regher, 1997). The 4-AP-induced increase in the amplitude of EPSCs may reflect, in analogy to the neuromuscular junction, an increase in the number of quanta delivered simultaneously by a single nerve pulse due to enhanced synaptic vesicle recycling (Heuser & Reese, 1981). In line with an increased probability of release following blockade of an I_D-like current, the appearance of delayed responses with multi-peaks could be due to the activation by 4-AP of previously presynaptically ‘silent’ connections (see Gasparini et al. 2000).

In the same preparation, an I_A-like K⁺ conductance has been reported to modulate synaptic transmission by gating action potential propagation at axonal branches (Debanne et al. 1997). In those experiments, propagation failures were prevented by very high concentrations of 4-AP (40 mM into the patch pipette) or when presynaptic action potentials were preceded by a brief or tonic hyperpolarisation. In our case, an increase in synaptic efficacy was produced by very low concentrations of 4-AP (10-50 μM). Moreover, tonic hyperpolarisation produced a reduction in spike width, as expected by removal of the inactivation of I_D, but this effect was not associated with changes in failure rate, making axon propagation failure unlikely. Eventually, a tendency towards an increase in the number of successes was commonly observed. This can be explained by the fact that an increased driving force for Ca²⁺ leading to an increased transmitter release would in these cases counterbalance the I_D-induced hastening of spike repolarisation and reduced synaptic efficacy (Katz & Miledi, 1967a). Only in one pair was an increase in failure rate detected upon membrane hyperpolarisation, and this pair also exhibited a strong reduction in spike width. However, this procedure was able to significantly delay the appearance of the first spike, suggesting that even small modifications in resting membrane potential affect the temporal coding of information.

Moreover, in the present experiments application of 4-AP resulted in a clear increase in the latency of synaptic responses. According to Katz & Miledi (1967a, b), a prolonged depolarisation would facilitate release by raising the permeability to Ca²⁺, but at the same time would reduce its driving force. This may be responsible for the delayed action of the presynaptic spike. While the broadening of action potentials may account for changes in the latency of synaptic currents, the lack of synchronicity in vesicle exocytosis could be responsible for the enhanced scattering of EPSCs observed in some experiments with 4-AP (Heuser & Reese, 1981).

Switching off the I_D-like current may facilitate seizure activity in the hippocampus

Low concentrations of 4-AP have been shown to induce spontaneous interictal discharges in the hippocampus (Rutecki et al. 1987; Perreault & Avoli, 1989). As demonstrated recently, these discharges originate in the CA3 hippocampal area and then propagate to the CA1 and entorhinal cortex (Luhmann et al. 2000). The CA3 hippocampal region may act as the pacemaker for the generation of synchronised activity (Miles & Wong, 1983; Ben-Ari & Cossart, 2000). This may depend largely on the dense network of collateral of axons interconnecting pyramidal neurones (Miles & Wong, 1986). It is therefore conceivable that an enhancement of intracellular Ca²⁺ following activity-dependent processes (Heinemann & Hamon, 1986) may switch off I_D at CA3-CA3 excitatory connections, generating a positive-feedback mechanism. Thus, the higher firing rate would enhance intracellular Ca²⁺, which in turn would inactivate more I_D leading to a further enhancement of cell excitability, Ca²⁺ influx and so on. This mechanism, by boosting neurotransmitter release, may facilitate the crosstalk between neurones, leading to network synchronisation and eventually seizure activity. This may be relevant for human temporal lobe epilepsy.