INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The amacrine cells of the retina are interneurons located in the inner nuclear layer, and they extend their dendrites to the inner plexiform layer. Typical amacrine cells do not possess a long axon, and this morphological feature is the origin of their naming. Electron-microscopic studies have revealed that the dendrites of amacrine cells contain both synaptic vesicles and postsynaptic membrane thickenings (Dowling and Boycott 1965; Vaughn et al. 1981), and these morphological features of amacrine cells suggest that their dendrites are presynaptic as well as postsynaptic sites (Wässle and Boycott 1991).

A large portion of the amacrine cell population is found to be GABAergic cells in the retina of various vertebrate species (Kolb 1997; Yazulla 1986). The targets of the GABAergic synapse are bipolar cells (feedback inhibition), other amacrine cells (mutual inhibition), and ganglion cells (feed-forward inhibition). The effect of feedback inhibition has been studied extensively (Dong and Werblin 1998; Hartveit 1999; Maple and Wu 1996; Tachibana and Kaneko 1987). The feed-forward inhibition from amacrine cell to ganglion cell is thought to be a factor forming the surround of the concentric receptive field of ganglion cells (Flores-Herr et al. 2001). Because mutual inhibition by GABAergic dendro-dendritic synapses has been reported (Watanabe et al. 2000), the mutual inhibition may induce local hyperpolarization in the dendrite, and local membrane potential changes in the dendrites are thought to modulate transmitter release from the presynaptic amacrine cell dendrites.

Intracellular recording from the soma of amacrine cells in cold-blooded animals has revealed that they generate action potentials superimposed on the light-evoked graded depolarization (Kaneko 1970; Werblin and Dowling 1969). In mammals, some types of amacrine cells are known to generate action potentials (Feigenspan et al. 1998; Koizumi et al. 2001; Taylor 1996). The action potentials of amacrine cells are blocked by TTX (Feigenspan et al. 1998; Koizumi et al. 2001; Miller and Dacheux 1976), suggesting that they are Na⁺ spikes, and TTX has been shown to reduce the inhibitory potency of the receptive surround of retinal ganglion cells (Bloomfield 1996; Cook and McReynolds 1998; Taylor 1999). Therefore it is highly likely that the amount of transmitter released from presynaptic sites is increased by spikes propagating into the dendrite (Watanabe et al. 2000).

One aim of the present study was to directly demonstrate that the action potential of amacrine cells can propagate regeneratively in the dendrite by applying dual patch-clamp recording and the action potential clamp technique to cultured amacrine cells. The other aim was to demonstrate that local application of GABA locally and independently suppresses propagation of the action potential in each dendrite of an amacrine cell.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Culture and identification of GABAergic amacrine cells

The experimental procedure conformed to the Guidelines for the Care and Use of Laboratory Animals, Keio University School of Medicine, and the university animal welfare committee approved our experiments. The culture method has been described previously (Koizumi et al. 2001). Briefly, after decapitating newborn Wistar rats (P0 and P1), their retinas were isolated and incubated at 37°C for 25 min in Ca²⁺-, Mg²⁺-free Hanks' balanced salt solution with HEPES (10 mM) supplemented with 1 mg/ml trypsin. After rinsing with Dulbecco's modified Eagle's medium (DMEM) supplemented with 5% heat-inactivated fetal bovine serum and triturating with a fire-polished glass pipette in 10 ml of culture medium, the dissociated cells were seeded on poly-L-ornithine-coated glass coverslips at a density of <1.5 × 10⁵ cells/ml and cultured for 10-14 days in DMEM supplemented with 14 mM NaHCO₃, 2 mM glutamine, 100 U/ml penicillin, 0.1 mg/ml streptomycin, and 5% heat-inactivated fetal bovine serum in a 5% CO₂ environment at 37°C. Immediately after dissociation, the cells appeared round, and no dendrites were seen. After 10 days in culture, only large cells (soma diameter of >10 µm) survived, and the dendrites from their soma extended over hundreds of micrometers. Experiments were performed using cells cultured for 10-14 days after dissociation. Cultured amacrine cells were identified by immunostaining with anti-HPC-1/Syntaxin (Sigma) and anti-GABA antibodies (Sigma), which have been described previously (Koizumi et al. 2001). Almost all (>90%) cultured cells with multiple long processes (the diameter of dendritic field >200 µm) were identified as GABAergic amacrine cells, as reported previously (Koizumi et al. 2001).

Dual patch-clamp recordings

A coverslip to which cultured cells had adhered was placed into a recording chamber, and the chamber was mounted on the stage of an inverted microscope equipped with Nomarski optics (IX-70, Olympus, Japan) and an ×60 objective lens. The chamber was continuously superfused with solutions that were gravity-fed at a rate of ~1 ml/min at room temperature (25°C). Membrane voltages and currents were recorded by the patch-clamp method in the whole cell configuration. The patch pipette was made by pulling Pyrex tubing on a micropipette puller (P-87, Sutter Instrument, Novato, CA). The recording pipette was connected to the input stage of a patch-clamp amplifier (Axoclamp 2B or Axopatch 200B, Axon Instruments, Foster City, CA), and an Ag-AgCl wire connected to the bath via a ceramic bridge served as an indifferent electrode. The pipette used to record from the soma had a resistance of 5-10 M when filled with pipette solution, whereas the resistance of the pipette used to record from the dendrite was 30-150 M. The somatic and dendritic pipettes were coated with dental wax (GC, Tokyo, Japan) to reduce the stray capacitance. Residual pipette capacitance and the access resistance were compensated as much as possible. Signals were low-pass filtered (Bessel filter, cutoff frequency: 5 kHz) and sampled at 10 or 20 kHz with DigiData 1200 interface and pCLAMP8 software (Axon Instruments). Recorded data were analyzed with Igor Pro software (WaveMetrics, Lake Oswego, OR). The standard external solution for the current-clamp experiments contained (in mM) 135 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, and 10 glucose (pH 7.4), and the standard pipette solution contained (in mM) 10 NaCl, 130 K gluconate, 1 CaCl₂, 1.1 EGTA, 10 HEPES, and 2 ATP-Na₂ (pH 7.2). TTX (Sankyo, Japan) was dissolved into the external solution and applied by a gravity feeding system. To block synaptic inputs, the extracellular solution contained bicuculline (GABA antagonist, Sigma, 100 µM), strychnine (glycine antagonist, Sigma, 2 µM), 2-amino-7-phosphonoheptanoic acid (AP7) [N-methyl-D-aspartate (NMDA) receptor antagonist, Sigma, 30 µM], and 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX, non-NMDA receptor antagonist, Sigma, 2 µM). In the GABA application experiments (Figs. 5 and 6), Ca²⁺ (2 mM) was replaced with equimolar Mg²⁺ to block spontaneous synaptic inputs. Lucifer yellow (0.2%) was dissolved in the intracellular solution to assess the spread of the dendritic field. No dye coupling via gap junctions with neighboring amacrine cells was observed.

Passive spread of hyperpolarizing voltage changes

Before investigating the spread of action potentials, we examined the spread of hyperpolarizing potentials evoked by negative current injection under the current-clamp mode with simultaneous whole cell patch-clamp recordings. Because amacrine cells in culture have no ionic conductance under hyperpolarized conditions (under 65 mV) (Koizumi et al. 2001), the hyperpolarizing voltage changes evoked in the dendrites represent only electrotonic properties. In Fig. 1A, we made simultaneous whole cell patch-clamp recordings of the soma and a dendrite of an amacrine cell (shown in Fig. 2A; 80 µm apart). The amplitude of the hyperpolarizing voltage changes recorded in the dendrite was >90% of that evoked in the soma (Fig. 1A). When we examined how far the hyperpolarized potential spread into the dendrite (Fig. 1B, n = 19), we found that ~90% of the somatic hyperpolarization spread 200 µm along the length of the dendrite.

View larger version (12K): [in this window] [in a new window]   Fig. 1. Passive spread of somatic hyperpolarizing voltage changes. A: somatic hyperpolarization was evoked by negative current injection under current-clamp conditions. Simultaneous whole cell patch-clamp recordings from the soma and the dendrite at sites 80 µm apart (shown in Fig. 2A). The hyperpolarizing voltage changes recorded in the dendrite were >90% of the changes evoked in the soma. The input resistance of the soma was 630 M, and the membrane capacitance was 50 pF. B: we examined how far the hyperpolarized potential spread along the dendrite (Fig. 1B, n = 19). Almost 90% of somatic hyperpolarization spread 200 µm into the dendrite.

View larger version (58K): [in this window] [in a new window]   Fig. 2. Propagation of action potentials in the dendrite of a cultured amacrine cell. A: Nomarski photomicrography of a cultured amacrine cell on which dual whole cell clamping was performed with two patch pipettes, one on the soma and the other on the dendrite. The pipettes were 80 µm apart. Calibration bar: 20 µm. B: injection of +20 pA into the soma through the soma pipette induced depolarization and successive action potentials in the soma (black trace). Similar voltage changes but of smaller amplitude were recorded from the dendrite (red trace). The 2 recording pipettes were applied in the current-clamp (CC) configuration. C: action potential clamp (APC). The soma was voltage clamped to the waveform of the action potential recorded from the soma as in B. D: voltage response of the dendrite with the soma was clamped to the waveform of the action potential. The control response was recorded without TTX (red trace), and the trace labeled as TTX (blue trace) was recorded in the presence of 1 µM TTX.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Propagation of action potentials into the dendrites of cultured amacrine cells

To demonstrate the propagation of action potentials into dendrites, simultaneous whole cell recordings were made from the soma and a dendrite of a cultured amacrine cell (Fig. 2A, distance between the two pipettes: 80 µm, resting membrane potential: 53 mV). Under current-clamp conditions injection of +20 pA into the soma depolarized the amacrine cell and triggered action potentials (Fig. 2B). The transient voltage response recorded from the dendrite had a 0.8-ms peak-to-peak delay from the soma action potential, and the amplitude was almost half that of the soma action potential.

Because there is a question as to whether the voltage changes recorded from the dendrite were action potentials propagated into the dendrite or merely represented electrotonic spread of the voltage change from the soma, we employed the action potential clamp technique, first used on cortical pyramidal neurons by Stuart and Sakmann (1994), to distinguish between these possibilities. The soma was voltage clamped by the waveform of the action potentials recorded in Fig. 2B (reproduced in Fig. 2C, "simulated action potential"), and the resulting voltage changes were recorded in the dendrite in the current-clamp configuration (Fig. 2D, control). The amplitude of the dendritic voltage change evoked by the somatic action potential clamp (Fig. 2D) was almost the same as the amplitude of the dendritic voltage change evoked by the somatic voltage change elicited by somatic current injection (Fig. 2B dendrite; also compare black squares with red circles of Fig. 3B). Thus the action potential clamp method proved capable of adequately controlling the somatic membrane potential. As shown clearly in the figure, in the presence of 1 µM TTX, the amplitude of the transient voltage change recorded from the dendrite was approximately one-third the amplitude of the voltage change recorded under control conditions. These results show that the action potential generated in the soma propagates into the dendrite and that TTX-sensitive Na⁺ current contributes to the propagation of action potentials into the dendrites of cultured amacrine cells.

View larger version (23K): [in this window] [in a new window]   Fig. 3. Relationship between distance from the soma and the action potential amplitude in the dendrite. The amplitude of dendritic action potential was normalized with that of the somatic action potential. A: action potentials were evoked by somatic current injection under dual CC conditions. The amplitude of the action potential recorded in the dendrite declined with distance from the soma, but beyond 80 µm, it leveled off at ~60% of the somatic action potential (n = 19). A fitting line was calculated as a single exponential curve. B: three types of dendritic action potential recordings were made from the same cell: a recording under CC conditions with somatic current injection, a recording under somatic APC in the control solution, and a recording under somatic APC in TTX solution. Data from 8 cells each at different distances are shown. The fitting line was imported from A. Under the action potential clamp on the soma, TTX suppressed the propagation of the action potential into dendrites (n = 8).

To examine how far the action potential spread into the dendrite, the relation of the amplitude of dendritic voltage change to distance from the soma was examined under the current-clamp mode by the simultaneous whole cell patch-clamp technique [distances between the soma and the recording point on the dendrite: 200 µm, resting membrane potential: 55 ± 7 (SD) mV, n = 19]. As shown in Fig. 3A, the amplitude of the dendritic action potential evoked by somatic current injection decreased with distance from the soma but leveled off at ~60% of the somatic action potential beyond 80 µm.

To elucidate the contribution of the TTX-sensitive Na⁺ current to dendritic propagation of the action potential, three types of data were recorded from the same cell: the amplitude of the dendritic action potential evoked by somatic current injection under current-clamp conditions, by somatic action potential clamp under control conditions, and by somatic action potential clamp under TTX conditions. The data from eight cells, each at a different distance, are shown (Fig. 3B). The amplitudes of voltage responses were recorded several times at each condition in each cell but found to be almost identical. The dendritic amplitudes of the action potentials evoked by somatic current injection and by somatic action potential clamp were almost the same. The amplitude measured under the control conditions leveled off at 60% of the amplitude of the somatic simulated action potential, while the amplitude measured in the presence of TTX declined to <40% of the amplitude of the somatic simulated action potential. These findings support the notion that the action potential propagates to the dendrite as a result of activation of TTX-sensitive Na⁺ currents in the dendrite.

Local changes in the dendritic membrane potential modulate the propagation of action potentials

The action potential in the dendrite showed all-or-none properties. In the experiment whose results are shown in Fig. 4, action potential clamping of the soma was performed, and the resulting voltage changes in the dendrite were recorded (Fig. 4A). With the soma action potential clamped, small bias currents of various amplitudes were simultaneously injected into the dendrite, and the voltage changes in the dendrite were recorded under the current-clamp mode (Fig. 4B). Regenerative action potentials were recorded from the dendrite under conditions in which positive bias currents were injected into the dendrite (+3-pA injection, top trace; no injection, middle trace of Fig. 4B), and TTX abolished these action potentials (Fig. 4B, blue traces). Injection of negative bias currents suppressed the propagation of the action potential in the dendrite (3-pA injection, bottom trace of Fig. 4B), and during injection of the negative bias currents, the waveform of the voltage changes recorded from the dendrite were identical under both the control conditions and in the presence of TTX. This observation indicates that injection of negative bias currents into the dendrites suppressed the regenerative propagation of action potentials into the dendrite.

View larger version (28K): [in this window] [in a new window]   Fig. 4. Effect of extrinsic bias current injection into a dendrite on action potential propagation. A: the command voltage applied to the soma in the action potential clamp experiment. B: voltage recordings of the dendrite under control conditions (red trace) and in the presence of 1 µM TTX (blue trace). A small bias current was injected into the dendrite from the recording pipette. The bias current was +3 pA (top), 0 pA (middle), or 3 pA (bottom). C: the relation between the amplitude of the transient voltage change recorded in the dendrite and the amount of bias currents injected into the dendrite. Filled circles (red): amplitude of the dendritic voltage change recorded under the control conditions. Filled squares (blue): amplitude of the dendritic voltage change recorded in the presence of 1 µM TTX. D: the same data as in C but replotted against the membrane voltage of the dendrite immediately before the spike-like voltage changes. Filled circles (red): amplitude of the dendritic voltage change recorded under the control condition. Filled squares (blue): amplitude of the dendritic voltage change recorded in the presence of 1 µM TTX.

Figure 4C illustrates the relationship between the amplitude of the voltage changes recorded in the dendrite and the amount of bias current injected into the dendrite at the recording site. The amplitude of the voltage changes increased abruptly between bias currents of 3 and 0 pA. In the presence of TTX, the amplitude of the voltage change recorded in the dendrite was identical at all bias currents tested (between 33 and +23 pA). When the same data were replotted against the dendritic membrane voltage immediately before the spike-like voltage changes, discontinuity was seen at a membrane voltage of 44.7 ± 2.5 mV (n = 8), the threshold voltage of the dendritic action potential. These results show that dendrites have a threshold at which the action potential is propagated in an all-or-none fashion.

GABA suppresses the propagation of action potentials into dendrites

The dendrites of amacrine cells function as both presynaptic and postsynaptic sites. Amacrine cells receive GABAergic input from neighboring amacrine cells via dendro-dendritic synapses (Marc and Liu 2000; Watanabe et al. 2000; Zhang et al. 1997). GABA activates GABA_A receptor on dendrites and hyperpolarizes amacrine cells by elevating Cl conductance (Watanabe et al. 2000). In the present study, to examine the effect of GABA on the propagation of action potentials into the dendrite under study, the action potential clamp technique was carried out with local application of GABA to the dendrite focusing on sites very near the dendritic pipette. GABA application inhibited action potential propagation into the dendrites of cultured amacrine cells (Fig. 5, A and B). The effect of GABA was overcome by injecting a positive bias current (more than +1 pA) into the dendrite, and the action potential in the dendrite reappeared. The inhibitory effect of GABA was achieved in two ways (Fig. 5C). First, the threshold bias current level was shifted to the positive direction, suggesting that a positive bias current was necessary to overcome the hyperpolarization induced by GABA. This interpretation is supported by the data in Fig. 5D in which the spike amplitude is plotted against the membrane voltage of the dendrite. The threshold voltage of the propagated action potential in the presence of GABA was found to be almost identical to the threshold voltage measured without GABA. Second, in the presence of GABA the amplitudes of all dendritic voltage changes were decreased (Fig. 5, C and D). It seemed highly likely that GABA induced membrane shunting and reduced the amplitude of the dendritic voltage changes. We obtained similar results in six other cells. The shunting effect of GABA was confirmed in a separate experiment. With a simultaneous current-clamp configuration on the soma and the dendrite, we examined the effect of local application of GABA (n = 7). The input resistance in the soma (807 ± 315 M) decreased to 51 ± 37% by GABA application, and the decrease in input resistance required injection of much more current to initiate an action potential at the soma.

View larger version (21K): [in this window] [in a new window]   Fig. 5. Effect of GABA on the propagation of action potential into dendrites. A: the command voltage applied to the soma in the action potential clamp experiment. B: voltage recordings of the dendrite under control conditions (red trace) and during puff application of 50 µM GABA to the dendrite (blue trace). C: relationship between the amplitude of the voltage recorded in the dendrite and the amount of bias current injected into the dendrite. Filled circles (red): amplitude of the change in dendritic voltage recorded under control conditions. Filled squares (blue): amplitude of the change in dendritic voltage recorded during puff application of GABA (50 µM). D: the same data as in C but replotted against the membrane voltage of the dendrite immediately before the spike-like voltage changes. Filled circles (red): amplitude of the dendritic voltage changes recorded under control conditions. Filled squares (blue): amplitude of the dendritic voltage changes recorded during puff application of GABA (50 µM).

Propagation of action potentials was independently suppressed in different dendrites of the same amacrine cell

If GABA input into a dendrite of an amacrine cell locally hyperpolarizes its membrane potential, the dendritic membrane potential could not be the same as that of other dendrites of the same amacrine cell, and as a consequence, propagation of the action potential should be suppressed independently in different dendrites of the same amacrine cell. To test this hypothesis, we recorded spontaneous action potentials by dual whole cell patch clamp on two different dendrites of an amacrine cell (Fig. 6A). In the control, the trains of spontaneous action potentials of the two different dendrites were synchronous and very similar to each other in amplitude and waveform (Fig. 6, B and C, asterisks). However, GABA application (50 µM, 100 ms) to one dendrite (Fig. 6A, dendrite 2) induced a decrease in the amplitude of the action potential in dendrite 2 to which GABA was applied (Fig. 6C, arrowhead), while the action potential of the other dendrite was unaffected. These results suggested that GABA suppressed propagation of the action potential independently in different dendrites.

View larger version (49K): [in this window] [in a new window]   Fig. 6. Spontaneous action potentials recorded from two different dendrites of an amacrine cell by dual whole cell patch clamp. A: Nomarski photomicrography of a cultured amacrine cell on which dual whole cell clamp was made with two patch pipettes on two different dendrites of an amacrine cell. GABA (50 µM, 100 ms) was applied by puffer pipette near the dendritic pipette (dendrite 2). Calibration bar; 20 µm. B and C: action potentials recorded from dendrites 1 (B) and 2 (C). In the control, the trains of spontaneous action potentials of the two different dendrites were synchronous and very similar to each other in amplitude and waveform (*). GABA application to dendrite 2 suppressed the propagation of the action potential in dendrite 2 alone (arrow).

Could action potentials be initiated in the dendrite?

Because amacrine cell usually receive excitatory inputs at the dendrite, we examined whether the action potential can be initiated in the dendrite. We employed the dual whole cell patch-clamp recordings from the soma and the dendrite and locally applied glutamate (50 µM) to the dendrite (6 cells). The glutamate application to the dendrite initiated trains of action potentials both in the soma and the dendrite (100 µm away from the soma, Fig. 7A). Although it was expected that the action potential in the dendrite should precede the action potential in the soma, we were unable to find a significant time delay of the rising phase or the peak time between somatic and dendritic spikes. However, initiation of action potentials in the dendrite was verified by other observations (Fig. 7, B and C). In the same amacrine cell used in the glutamate experiment, we injected positive current into the dendrite or into the soma. Under some conditions (+6 pA injection to the dendrite), only the dendritic action potential (the 1st dendritic spike of Fig. 7B, asterisk) was evoked and appeared to have no effect on the soma voltage. The second dendritic spike in Fig. 7B appeared to follow the somatic spike, and it is likely that the soma was slowly depolarized until an action potential was initiated which then propagated into the dendrite. Dendritic action potentials without action potentials in the soma were observed in 3 of 19 cells, and in the remaining 16 cells, action potentials in the dendrite were always synchronized with action potentials in the soma. In response to somatic current injection to the cell of Fig. 7, action potentials were evoked synchronously in the soma and the dendrite (Fig. 7C). In 19 cells examined, action potentials were evoked synchronously in the soma and the dendrite by somatic current injection. These results suggest that the action potential could be initiated in the dendrite.

View larger version (19K): [in this window] [in a new window]   Fig. 7. A: action potentials were evoked by local application of glutamate to the dendrite (50 µM). Although it was expected that the action potential in the dendrite would precede the action potential in the soma, we were unable to find a significant time delay of the rising phase or the peak time between somatic and dendritic spikes. B: current (+6 pA) was injected to the dendrite of the same cell used in A. Note that the first transient voltage change (*) was detected only in the dendrite, and not in the soma. C: current (+9 pA) injection to the soma initiated synchronous action potentials in the soma and the dendrite of the same cell used in A.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Physiological significance of action potentials in the dendrites of amacrine cells

Dendrites of retinal amacrine cells are synaptic input sites as well as output sites. This means that the action potentials of amacrine cell dendrites have important implications in regard to the regulation of neurotransmitter release. We have already reported that the spontaneous postsynaptic events recorded in amacrine cells consist of spike-driven large inhibitory postsynaptic potentials (IPSPs) and miniature IPSPs (Watanabe et al. 2000). The large IPSP was suppressed by TTX, whereas the small IPSP remained unaffected. This is a clear indication that the amount of transmitter release by amacrine cell dendrites is regulated by action potentials.

The dendritic action potentials of amacrine cells were labile and different from the robust action potentials of the axons of typical neurons such as retinal ganglion cells. Slight hyperpolarization suppressed the dendritic action potential. The attenuation of transient voltage changes may be attributable to the membrane capacitance and the A-type K⁺ conductance present in amacrine cells (unpublished data).

It is speculated that the amplitude of the voltage-activated inward current barely exceeded that of the outward current. Perhaps the relatively small inward current accounts for both the small peak amplitude of the dendritic spike as well as the peak voltage, which is far more negative than the Na⁺ equilibrium potential. It is important to measure the density of Na⁺ and K⁺ channels in the dendritic membrane of amacrine cells and the physical properties of the cytoplasm of the dendrite.

In the present study, we showed that GABA hyperpolarized dendrites and suppressed the propagation of action potentials in cultured amacrine cells. Because the inhibitory synaptic input sites were shown to be diffusely distributed in wide-field amacrine cells (Famiglietti and Vaughn 1981; Vaughn et al. 1981), GABAergic modification could be induced at any site on the dendrite. If the membrane of a dendrite was hyperpolarized in response to GABAergic input, its synaptic output would be suppressed and its lateral inhibitory output to neighboring cells would be diminished. By contrast, if the membrane of a dendrite was depolarized, its synaptic output would be increased and the lateral inhibitory output to neighboring cells would be enhanced accordingly. In addition, it is likely that the dendritic membrane potentials of an amacrine cell are not equipotential to each other, meaning that each dendrite can function independently when local synaptic inputs occur.

The limitation of our present study is that it was carried out on cultured amacrine cells. There may be the criticism that the distribution of voltage-gated channels on the dendrites of cultured amacrine cells are different from that of amacrine cells in vivo. Active properties of the dendrites are demonstrated in various kinds of neurons in the mammalian CNS (Johnston et al. 1996; Magee et al. 1998; Stuart and Sakmann 1994; Stuart et al. 1997) and retinal ganglion cells (Velte and Masland 1999). In amacrine cells, Miller and Dacheux (1976) speculated that amacrine cells generate action potentials in both their soma and dendrites. More recently, Cook and Werblin (1994) recorded Na⁺ currents from the dendrites of tiger salamander amacrine cells in a slice preparation by positioning an extracellular recording electrode close to the dendrite and based on their recordings they suggested the existence of a self-regenerative process in the dendrites. These previous works strongly support our idea that the dendrites of amacrine cells have active properties and that the propagation of action potentials are regulated by dendritic membrane potentials in all-or-none fashion.

Dendrites could generate an action potential locally

In the present study, we showed that somatic action potentials can propagate regeneratively into the dendrites of an amacrine cell. Our data strongly suggest that dendrites of amacrine cells have a Na⁺ current that boosts the spread of action potentials into their dendrites. If the density of the Na⁺ channels was sufficient to generate an action potential, the dendrite could generate a local action potential independent of the somatic action potential. Because the dendrites of amacrine cells function as both presynaptic and postsynaptic sites, excitatory synaptic input could locally trigger dendritic action potentials that modifying the neurotransmitter release by this particular dendrite. Miller and Dacheux (1976) have actually suggested that the dendrites of amacrine cells can generate action potentials. In the present study, we showed that some action potentials initiated in the dendrite had no effect on the soma voltage. Considering our results, a spike in only one dendrite is inadequate to produce a spike in the soma. The rule appears to be that action potential initiated in the soma can travel down the dendrites of an amacrine cell. Perhaps multiple dendrites must be activated so that their spikes can sum to produce a somatic spike. Thus it is likely that the action potential in the soma and the dendrite has a different effect in the extent of lateral inhibition and therefore in the information processing mechanism in the inner plexiform layer.