Muscarinic Modulation of Spike Backpropagation in the Apical Dendrites of Hippocampal CA1 Pyramidal Neurons

Modulation of dendritic [Ca²⁺]_i changes

Backpropagating Na⁺ spikes cause [Ca²⁺]_i changes in the dendrites of CA1 pyramidal neurons. In the distal dendrites these [Ca²⁺]_i increases are dominated by contributions from the first few spikes that propagate to that distance with the largest amplitude (Jaffe et al., 1992; Miyakawa et al., 1992;Callaway and Ross, 1995; Spruston et al., 1995a). Because muscarinic activation increases the amplitude of later spikes in a train, we expected that there might be an associated increase in the [Ca²⁺]_i change under these conditions. In fact, an increase by CCh of the stimulus-evoked [Ca²⁺]_i increase in the dendrites has been described (Muller and Connor, 1991), but the mechanism was not determined. To check this prediction, we made patch recordings in the dendrites with electrodes containing Ca²⁺ indicators (see Materials and Methods). Figure 3 presents results from a typical experiment. The image shows the electrode and the dendrites filled with Calcium Green-1. The recording location was ∼250 μm from the soma. In response to stimulation in the alveus we recorded a train of action potentials with decreasing amplitudes. At the same time we recorded the fluorescence increase from a rectangular region close to the electrode. As previously shown (see above), the increase peaked at the time of the first few larger amplitude action potentials. When 1 μm CCh was added to the bath, the amplitude of the later spikes increased. The amplitude of the fluorescence change also increased and peaked at the time of the last spike in the train. When 1 μm atropine was added to the superfusate, both the spike profile and the fluorescence change returned to close to the control responses. CCh did not cause a significant change in the resting fluorescence of the cell, suggesting that resting [Ca²⁺]_i was not affected (Muller and Connor, 1991). However, it is possible that there were localized increases that were missed by our measurements.

In these experiments the recovery times of the transients were slightly longer than the 100–150 msec recorded with minimal levels of indicator (Callaway and Ross, 1995; Helmchen et al., 1996). Therefore, the [Ca²⁺]_i transients may have been buffered, reducing their amplitude and slowing their time course. Nevertheless, the fluorescence records qualitatively indicate the increase in amplitude and the change in time course of the [Ca²⁺]_i changes in CCh.

CCh-induced increases in ΔF/F in the dendrites also were measured with bis-fura-2 and were measured when whole-cell recordings were made in the soma. There was some variability in the responses, particularly in the extent to which the CCh-induced fluorescence changes were reversible (possibly because of photodynamic damage). Nevertheless, an increase in amplitude and change in time course always were observed.

Perforated patch recordings

We were concerned that the modulation of activity-dependent propagation by CCh might be an artifact resulting from the dialysis of the cytoplasm by the pipette solution. One possibility was that a change in [Mg²⁺]_i affected Ca²⁺-activated K⁺ conductances (Lancaster et al., 1991). However, Figure4A shows that the amplitude of individual spikes was not affected by removing Mg²⁺from the pipette solution. We also assessed the effect of removing Mg²⁺ on the frequency-dependent amplitude reduction. We used the ratio of the amplitude of the last spike in a train of 10 action potentials to the amplitude of the first spike as a measure of this modulation. Figure 4B shows that eliminating internal Mg²⁺ had no consistent effect on this ratio when the spikes were activated at 50 msec intervals.

A more general control for the consequences of changing the intracellular solution is to make perforated patch recordings (Horn and Marty, 1988; Spruston and Johnston, 1992). When nystatin is used as the ionophore, cations and Cl⁻ are exchanged through the membrane, but no other cytoplasmic constituents are washed out. Using 200 μg/ml nystatin in the pipette solution (see Materials and Methods), we faithfully recorded action potentials from the dendrites at all distances; the access resistance was <50 MΩ (Fig.4A). Similarly, there was no systematic change in the frequency-dependent propagation at 20 Hz, although there was considerable scatter in the data (Fig. 4B). When 1 μm CCh was added to the bathing solution, this modulation was eliminated at all tested spike intervals (Fig. 5). Similar results were obtained in seven other cells. Therefore, we conclude that whole-cell recording with our standard pipette solution did not introduce artifacts in the CCh experiments.

Dependence on distance and frequency

Qualitatively, these results show that CCh changes the frequency-dependent backpropagation of action potentials into the dendrites. To assess these effects quantitatively, we made recordings at different distances from the soma and measured the effect of CCh on the amplitude of the first and last spike in a train. As shown previously (Turner et al., 1991; Spruston et al., 1995a), the amplitude of the first spike shows smooth decrements from ∼100 mV in the soma to ∼40 mV at 350 μm (Fig. 6A, left). The amplitude in the distal dendrites is ∼20 mV less than that measured by Spruston et al. (1995a). This difference occurred because our experiments were made at 30–32°C and theirs were made at room temperature (22–24°C), which we could demonstrate directly by turning off the heat on the experimental chamber (data not shown). This panel also shows that at all distances CCh had no significant effect on the amplitude of the first spike.

Figure 6A, right panel, shows that in the middle dendrites (150–275 μm from the soma) CCh increased the amplitude of the last spike in the train, often to more than twice the original height. However, in the distal dendrites (>300 μm from the soma) the effect was much smaller. Figure 6B shows these data plotted in a different way; the ratio of the amplitudes of the last and first spikes is shown as a function of the distance from the soma. In control conditions (filled circles) the ratio decreased with increasing distance from the soma (Spruston et al., 1995a). However, in CCh the ratio was above 80% up to 300 μm from the soma. A lower ratio beyond this distance was observed in every cell. Figure 6C, which compares the amplitudes in CCh and in control conditions, emphasizes that the effect was on the last spike and not the first. Most of the points far above the dotted diagonal line were from the middle dendritic region.

The amount of activity-dependent amplitude reduction varies as a function of the frequency of stimulation (Callaway and Ross, 1995). This result is shown systematically in Figure7B, which summarizes the results from 55 pyramidal neurons. Maximum peak amplitude reduction was measured when the action potentials were evoked at 50–100 msec intervals and increased with distance from the soma. In many cells some modulation was detectable even with 1–2 sec spike intervals, especially when measured in the distal dendrites (275–300 μm from the soma). The apparently weak modulation at short (10–20 msec) spike intervals may be related more to the way we measured spike amplitudes than to a reduction in the modulatory effect in this frequency range. We define spike amplitude as the difference between the peak potential and the resting potential. In many cells, when the spikes were activated at short intervals, a depolarizing afterpotential developed, increasing the peak potential of later spikes in the train (Fig. 7A). If the spike amplitudes are measured from the point of inflection on the rising phase, there is more modulation at shorter time intervals than the graphs in Figure 7B indicate.

The effect of CCh on frequency-dependent modulation was measured in a subset of these cells. Figure 7C summarizes these data. At all distances up to 300 μm from the soma CCh reduced or eliminated the amplitude reduction during a train.

Measurements in very distal dendrites

The apparent conclusion from these data is that CCh converts action potentials that fail to propagate to the distal dendrites into action potentials that reach the tips of the dendrites. However, recordings from the most distal dendrites suggest that the conversion may not be complete. Figure 8A shows data from a pyramidal neuron recorded 350 μm from the soma in whole-cell mode. In normal saline only the first action potential in a train fully propagated to the recording site when the interval was 50 msec between stimuli. Consistent with the data of Figure 6, the peak amplitude of this action potential was ∼40 mV. When the interval was shortened to 10 msec, the afterdepolarizations summated until a wider action potential was generated. Generation of these wider spikes was a consistent observation at this distance when strong stimuli were given. They were probably Ca²⁺-dependent action potentials, because they were associated with large transient increases in [Ca²⁺]_i (data not shown). When CCh was applied at a concentration of 10 or 50 μm, there was no clear effect on the activity-dependent amplitude reduction (Fig.8B). Figure 8C shows that the same result was obtained at all spike intervals tested. We can be confident that CCh was reaching this cell, because the resting potential depolarized when this agent was applied (Fig. 8B). A similar failure of CCh to reduce the spike amplitude modulation in the very distal dendrites was observed in seven other cells recorded in whole-cell mode. In two of these cells the amplitude of the second and third spikes was increased, but later spikes were unaffected. The enhancement of the earlier spikes may have been a consequence of the tonic depolarization in CCh (Tsubokawa and Ross, 1996a). We attempted to confirm this result by perforated patch and cell-attached techniques, but we were unable to make satisfactory recordings >300 μm from the soma.

Receptor subtypes

Classification of the muscarinic receptor subtypes that are involved may provide a clue to the mechanisms responsible for the frequency-dependent spike amplitude modulation. We tested the effects of pirenzepine and gallamine, putative M₁ and M₂ receptor antagonists (Hammer and Giachetti, 1984), which have been used previously in this kind of experiment (Dutar and Nicoll, 1988). In the experiment shown in Figure 9, the control recording in the dendrites demonstrated the usual activity-dependent amplitude reduction. When 5 μm CCh and 20 μm gallamine were added to the perfusate, the modulation was reduced, suggesting that gallamine did not antagonize the effects of CCh completely, but the gallamine clearly had some effect because washing it out (while leaving in the CCh) reduced the modulation further (n = 4). This effect was reversible on reperfusion with gallamine (fourth panel). Finally, removal of gallamine and the addition of 0.3 μmpirenzepine completely, blocked the demodulatory effect of CCh. In other experiments, when pirenzepine was added with CCh at the beginning of the experiment, there also was no reduction in the amplitude modulation (n = 6). From these results we conclude that 0.3 μm pirenzepine was fully effective in antagonizing the effect of CCh, whereas gallamine was partially effective. These results are consistent with the high levels of M₁ and M₃ mRNA expression and low levels of M₂ and M₄ mRNA expression in rat CA1 pyramidal neurons (Buckley et al., 1988). Dutar and Nicoll (1988) found that gallamine was 75% effective in suppressing the M-current. The weak suppression by gallamine of the effect of CCh on spike amplitudes suggests that blockade of the M-current plays only a small role in the amplitude modulation. The complete suppression of the CCh effect by pirenzepine suggests that M₁ receptors are important in transducing this muscarinic response. However, Dutar and Nicoll (1988) found that pirenzepine blocked all of the different muscarinic responses in CA1 pyramidal neurons, making it difficult to separate their relative importance. We also note that gallamine and pirenzepine are not completely selective between M₁ and M₂receptors, nor is their antagonism to CCh purely competitive (Hammer and Giachetti, 1984).

Calcium dependence

One important effect of acetylcholine is the suppression of the slow Ca²⁺-activated K⁺conductance that is the primary cause of spike accommodation during a sustained depolarizing stimulus (Bernardo and Prince, 1982; Cole and Nicoll, 1984). Because Ca²⁺ enters the dendrites after action potentials (Jaffe et al., 1992; Regehr and Tank, 1992), blockade of this conductance might explain the frequency dependence of spike backpropagation. To test this possibility, we made dendritic patch recordings with pipettes containing 10 mm BAPTA and 0 added Ca²⁺. Figure 10(Control) shows that BAPTA did not prevent the amplitude reduction (n = 4). A similar result follows from the data of Spruston et al. (1995a) because many of their recordings were made with pipettes containing 10 mm EGTA. Therefore, a Ca²⁺-activated K⁺conductance is unlikely to contribute to this effect. However, BAPTA did suppress the effect of CCh, even when CCh was used at a concentration of 10 μm. This result was true for all spike intervals tested (Fig. 10C). The figure also shows that CCh did not cause a tonic depolarization in the BAPTA-filled cells (compare with Fig. 2), consistent with previous observations (Fraser and MacVicar, 1996). Similar results were obtained when the pipette contained 10 mm EGTA and 0 added Ca²⁺(n = 4) or a mixture of 5 mm EGTA and 0.5 mm CaCl₂. The latter solution was selected to suppress [Ca²⁺]_i transients while maintaining resting [Ca²⁺]_i close to normal values. We also note that the buffering effect of 200 μm Ca²⁺ indicator (Fig. 3) was not sufficient to block the effect of CCh.

Effects of other pharmacological agents

The experiments described so far have concentrated on the muscarinic demodulation of spike propagation induced by CCh. This compound had the clearest effect on the activity-dependent propagation normally observed in CA1 pyramidal neurons. However, it is possible that other conductances, unaffected by muscarinic agonists, could play a role in the activity-dependent spike amplitude reduction. We examined several possible candidates.

We tested compounds that are known to block fast-activating K⁺ conductances in hippocampal pyramidal neurons (Storm, 1993). Apamin (1 μm, n = 3), charybdotoxin (100 nm, n = 3), 4-aminopyridine (100 μm, n = 3), and tetraethylammonium (5 mm, n = 6) had small effects on the action potential shape in dendritic recordings (Andreasen and Lambert, 1995). However, there was no significant change in the frequency-dependent amplitude reduction (data not shown). These results seem to rule out contributions from the delayed rectifier and the fast Ca²⁺-activated K⁺conductance (I_C). External CsCl (0.5–5 mm), which blocks the Q-current (Halliwell and Adams, 1982;Spruston and Johnston, 1992), had variable effects. Usually the first few spikes in a train were increased in amplitude to match the first spike, but the later spikes were affected weakly (data not shown;n = 12). CsCl consistently hyperpolarized the cell. In the middle dendrites, hyperpolarization by itself often increased the amplitude of action potentials (Fig. 8; Tsubokawa and Ross, 1996a), so it is possible that this weak effect was mediated partially by the hyperpolarization alone.

Norepinephrine (NE; 20 μm), which blocks the slow Ca²⁺-activated K⁺ conductance and reduces spike accommodation (Madison and Nicoll, 1982), had no effect when the spike interval was 50 msec (n = 9), whereas CCh had dramatic effects. However, at longer intervals (0.5–2 sec) the modulation was reduced and was reversible (n = 2; data not shown). This effect was not examined further in this series of experiments.

Mechanism of activity-dependent backpropagation

Jaffe et al. (1992) suggested that a Ca²⁺-activated K⁺ conductance could be responsible for the variable propagation in hippocampal cells. Our experiments showing no change in the propagation pattern with intracellular BAPTA or EGTA (Fig. 10) rule out this explanation.Migliore (1996) suggested that either slow inactivation of Na⁺ channels or a slow voltage-activated shunting conductance in the dendrites could explain activity-dependent propagation in his computational model. There is some experimental evidence supporting the role of slow Na⁺ channel inactivation (Colbert and Johnston, 1996b), but it is not known if this slow inactivation is the only mechanism involved. If slow Na⁺ channel inactivation is responsible, then removal of this inactivation is not likely to be the mechanism for the effect of CCh because CCh reduces peak Na⁺ current and slows inactivation in channels from CA1 pyramidal neurons (Cantrell et al., 1996). In principle, the M-current or the nonspecific cation conductance described by Benson et al. (1988) and others could qualify as the shunting conductance.

The blockade of the CCh effect by BAPTA does not seem to be consistent with a role for the M-current because Madison et al. (1987) found that the CCh-mediated suppression of this current persisted in Cd²⁺-containing saline. More directly, Dutar and Nicoll (1988) found that buffering [Ca²⁺]_i with BAPTA (at a concentration sufficient to block the slow AHP) had no effect on the suppression of the M-current by CCh. However, other experiments suggest that the suppression of the M-current (at least in sympathetic ganglia) is mediated by a rise in [Ca²⁺]_i released from internal stores (Kirkwood et al., 1991). Although we did not measure a rise in [Ca²⁺]_i via this pathway, it is possible that our apparatus was not sufficiently sensitive. If a Ca²⁺-mediated suppression of a K⁺ conductance is involved in the CCh response, then the specific target may be the recently discoveredether-à-go-go channel (Warmke et al., 1991; Stansfeld et al., 1996).

CCh reverses the activity-dependent spike amplitude reduction. However, it is not clear that it does so by blocking the conductance responsible for the reduction. Indeed, the inability of CCh to reverse the amplitude reduction in very distal dendrites (Fig. 8) suggests that different mechanisms are at work. Therefore, even if suppression of the M-current does not contribute to the effect of CCh on spike propagation, it still could have a role in the activity-dependent amplitude reduction observed under normal conditions.

Functional significance

Carbachol and other cholinergic agonists have many effects in the hippocampus, including the induction of bursts and oscillations (Bianchi and Wong, 1994) and modulation of synaptic plasticity (Blitzer et al., 1990; Huerta and Lisman, 1993). These effects may be mediated via changes in membrane conductances (Nicoll, 1985), induction of protein synthesis (Feig and Lipton, 1993), or other cellular mechanisms. It is not certain whether the enhancement of dendritic spike propagation relates to these more global effects.

There is evidence that backpropagating spikes are important for the induction of some forms of Hebbian synaptic plasticity. Recent experiments in the hippocampus (Magee and Johnston, 1997) and cortex (Markram et al., 1997) have shown that blocking these spikes can prevent LTP or LTD in some protocols. One possibility is that the spikes act via the [Ca²⁺]_i increases they evoke. In this case the increased amplitude, longer duration, and increased spatial extent of the [Ca²⁺]_i increases in CCh (Fig. 3) presumably would enhance these effects. A second possibility is that the voltage change of the backpropagating spikes directly affects some cellular function. For example, the conductance of NMDA receptor channels has an apparent voltage dependence because of the blocking effect of Mg²⁺. In the dendrites the action potential usually is initiated on the falling phase of the synaptic potential (Fig. 1) when only NMDA receptors are likely to be open (Spruston et al., 1995b). At this time they may be particularly sensitive to the changes in spike amplitude induced by CCh.

Backpropagating action potentials also can affect dendritic physiology in ways that are more straightforward. To illustrate, the [Ca²⁺]_i increases caused by the spikes will turn on Ca²⁺-activated K⁺conductances in the dendritic membrane. The resulting hyperpolarization will oppose the depolarization of the summating EPSPs, preventing the activation of voltage-dependent channels in the dendrites (Magee and Johnston, 1995a), which, in turn, should influence the way the EPSPs contribute to action potential initiation.