INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Retinal ganglion cells (RGCs) are the output neurons of vertebrate retinae that carry visual information to the brain. RGCs are classified into several functional subtypes that integrate the presynaptic inputs of different types of light responsiveness in different manners. One of the best studied functional classifications is based on the linearity of the spatial summation of presynaptic inputs. Since the original work in the cat (Enroth-Cugell and Robson 1966), vertebrate RGCs clearly showing linear and nonlinear summation have been termed X(-like) and Y(-like) RGCs, respectively. Vertebrate RGCs showing light responsiveness unlike that of X and Y RGCs have been termed W(-like) RGCs (see Stone et al. 1979 for review). Differences in spatial summation and other modes of input integration confer each RGC subtype its characteristic firing pattern, which encodes specific aspects of visual information. For example, there are RGCs that show sustained firing to a step change in illumination and others that show transient firing (sustained and transient RGCs, respectively) (see, e.g., Cleland et al. 1973; Fukuda et al. 1984; Rodieck and Stone 1965; Saito 1983; Stone and Fukuda 1974; Werblin and Dowling 1969). These responses may efficiently encode relatively constant and abruptly changing light stimuli, respectively (Dacey 1994; Merigan and Maunsell 1993). In mammals, most sustained RGCs consist of X-like RGCs and a certain subset of W-like RGCs; most transient RGCs consist of Y-like RGCs and another subset of W-like RGCs (see Stone et al. 1979 for review). In goldfish, RGCs are also classified into sustained and transient subtypes although this classification cuts across the X/Y/W classification (Bilotta and Abramov 1989; Levine and Shefner 1979) (see DISCUSSION).

It has been thought that the sustained and transient light responses of RGCs are shaped depending primarily on the time courses and spatiotemporal interactions of the synaptic inputs. This notion is supported by several studies using in-situ retinal preparations. The basic shapes of both light responses appear to reflect the time courses of synaptic inputs from selective bipolar cells (see Awatramani and Slaughter 2000). Transient light response may be further shaped by transient excitatory synaptic inputs from amacrine cells or by the truncation of sustained excitatory synaptic inputs by amacrine cells (see Nirenberg and Meister 1997).

According to this idea, the intrinsic firing property of an RGC is assumed to play only a minor role in shaping its light response. In some retinal preparations, firing properties do not significantly differ among RGCs and therefore are thought to contribute little to the cell-to-cell distinction of the firing pattern of the light responses. In the turtle, most RGCs show sustained firing to current step stimuli regardless of their light response shapes (Baylor and Fettiplace 1979). In the tiger salamander, most RGCs linearly convert the intensity of the synaptic input into firing frequency without major temporal transformation (Diamond and Copenhagen 1995). However, some studies indicate that RGCs may have heterogeneous intrinsic firing properties that could aid the establishment of the light response distinction. In the tiger salamander, the sustained or transient light response of an RGC could be mimicked by the response to a current step stimulus (Mobbs et al. 1992). In the cat, the sodium current's recovery from inactivation is slower in W-like RGCs than it is in X-like RGCs; this slow recovery may emphasize a sluggish light response peculiar to the former RGC subtype (Kaneda and Kaneko 1991).

To further explore the heterogeneity of the intrinsic firing properties of RGCs, we compared current-evoked responses in isolated goldfish RGC preparations that were free from synaptic inputs. Recordings were made in a perforated-patch, whole-cell configuration to minimize cytoplasmic disturbance, which might alter the intrinsic firing property. Under these conditions, we found a striking difference in firing accommodation among the goldfish RGCs. We identified a low-threshold, non-inactivating K⁺ current as a principal ionic mechanism underlying the firing accommodation. In addition, there was a tendency for firing accommodation to be more obvious in larger RGCs whose soma size distribution was similar to that of the RGC subtype showing transient light responses in situ. These findings suggest that RGCs have heterogeneous intrinsic firing properties that could aid synaptic inputs in shaping light responses.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cell preparation

The electrophysiological and morphological measurements described in RESULTS were performed in isolated RGC somata of adult common goldfish (Carassius auratus; body length 5-10 cm) prepared as described by Ishida and Cheng (1991). Briefly, retinae were dissected from one eye of goldfish anesthetized in an aqueous solution of tricaine methanesulfonate (0.3% wt/vol) and killed by decapitation. The retinae were then treated with Ca²⁺-free saline (pH 7.6) containing 1 mg/ml Protease Type XXIV (P-8038, Sigma) for 5-10 min at room temperature (23-25°C) and then triturated with a Pasteur pipette. The dissociated cells were plated to surface-modified polystyrene dishes (Falcon 3801, Becton Dickinson), maintained at room temperature with a 1:9 mixture of Leibovitz L-15 medium (Gibco 41300-039, Life Technologies, Grand Island, NY) and a Ca²⁺-containing physiological saline, and allowed to recover overnight. At the time of recording, RGCs were identified by morphological features described by Ishida and Cohen (1988). RGCs identified in this way showed the same classes of ligand- and voltage-gated ion currents as did those identified by retrograde labeling with dyes applied to the optic nerve (Ishida and Cohen 1988; Tabata and Ishida 1996).

Electrophysiological measurements

Whole-cell recordings were made from isolated RGCs in a tight-seal, perforated-patch configuration (Horn and Marty 1988) at room temperature (23-25°C). A recording pipette was pulled from a borosilicate glass capillary (BF150-86-10, Sutter, Novato, CA) to a tip resistance of 3-5 M. The pipette solution consisted of (in mM) 100 potassium D-gluconic acid (K-DGA), 6 NaOH, 10 1,2-bis(2-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid tetrapotassium salt (K₄-BAPTA), 4 CaCl₂, and 3.3 MgCl₂; pH and osmolality were adjusted to 7.5 with DGA and to 310 mOsmol/Kg with sucrose, respectively. Immediately before recording, a DMSO solution containing 4% wt/vol amphotericin B and 1% wt/vol Pluronic F-127 (Molecular Probes, Eugene, OR) was mixed with the pipette solution at a ratio of 1:200. During recording, the culture dish was perfused at a rate of 1-2 ml/min with a control bath solution. The standard control bath solution consisted of (in mM) 142 NaCl, 3 KCl, 2.5 CaCl₂, 10 HEPES, and 10 D-glucose; pH and osmolality were adjusted to 7.5 with NaOH and to 310 mOsmol/Kg with sucrose, respectively. Bath solutions containing the test agents were applied through a wide-tipped pipette located near the recorded RGC. When Ba²⁺ was used as a test agent, equimolar Mg²⁺ or Co²⁺ was added to the control bath solution to minimize change in the membrane surface charge. Other exceptions in the composition of bath solutions are stated in the figure legends. Stimulation and signal acquisition were performed with an Axopatch-1D amplifier (Axon, Foster City, CA) controlled by a PULSE system (version 8.10, HEKA, Lambrecht, Germany). Signals were low-pass filtered at 2-5 KHz and digitized at 10 KHz. The capacitance cancellation circuitry was adjusted under voltage-clamp conditions to minimize the slowest component of capacitive currents elicited by a 5-mV voltage step; whole-cell membrane capacitance (C_m) was read from the corresponding dial. After this adjustment, measurements were made either under voltage- or current-clamp conditions. The command and measured membrane potentials (E_m) described in RESULTS are corrected for a liquid junction potential between the pipette and the bath solutions. When the amplitude of a voltage-clamp current exceeded 500 pA, we employed electronic compensation (~70%) of series resistance (R_series).

In the current-clamp measurements, to avoid biased sampling from RGCs of a certain subtype, we attempted to record from every RGC that we encountered. The data were discarded if the RGC could not fire spikes in response to step current stimuli given without background current injection. Different RGCs were compared by macroscopic firing pattern but not by very fast voltage changes included in single spikes because the latter might be distorted in the voltage responses recorded with the voltage-clamp amplifier (Magistretti et al. 1996).

Morphological measurement

The cross-sectional soma areas of the RGCs used in the electrophysiological measurements were measured as follows. During or after a recording, photomicrographs of the RGC were taken on slide film (Ektachrome Dyna, ISO 100, Kodak). The photomicrographs were projected onto paper at a magnification of ×2,000 and the outline of the soma was traced manually. The outline was digitized on a flatbed scanner (CS-6151, Seiko, Chiba, Japan) at a resolution of 400 dpi and the area in the outline was measured with NIHimage software (version 1.61 for Macintosh, National Institutes of Health).

Statistical analyses

Statistical comparisons were performed with JMP software (version 3.1.6 for Macintosh, SAS, Cary, NC). When groups of data were judged to consist of normally distributed data (P < 0.05, Shapiro-Wilk W test), the data groups were compared by t-test and are presented as means ± SE. Otherwise, data groups were compared by the Wilcoxon/Kruskal-Wallis rank sum test and are presented as medians. Linear or nonlinear regression was used to fit functions to data with SigmaPlot (version 5.0.1 for Macintosh, SPSS, Chicago, IL) or Igor Pro software (version 2.04 for Macintosh, WaveMetrics, Lake Oswego, OR).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Intrinsic firing property may change in ruptured-patch configuration

Ruptured-patch configuration has been widely used to record the whole-cell signals of vertebrate RGCs because of its convenience and the relatively low R_series thereby achieved. However, the "wash-out" of cytoplasmic molecules through a ruptured cell membrane may alter the neuronal firing pattern (Cuevas et al. 1997). To test whether such alteration occurs in goldfish RGCs, we compared the voltage responses recorded consecutively in perforated-patch and ruptured-patch configurations (Fig. 1A). In this experiment, a perforated-patch configuration was first established with amphotericin B in the pipette solution and later was switched to a ruptured-patch configuration (the cell membrane with negative air pressure). In the perforated-patch configuration, electrical access to the intracellular side was obtained without wash-out because amphotericin B formed ionophores in the cell membrane that permeated small ions but not the large cytoplasmic molecules necessary to maintain the normal functions of some ion channels (Horn and Marty 1988). In the cases of RGCs displaying transient firing in the perforated-patch configuration, firing became more sustained after membrane rupture (n = 5; Fig. 1A). This alteration cannot be ascribed to cell membrane deterioration caused by amphotericin B entering the cytoplasm because it occurred immediately after membrane rupture (typically within 30 s).

View larger version (15K): [in this window] [in a new window]   Fig. 1. The intrinsic firing property of retinal ganglion cells (RGCs) may be affected by cytoplasmic perfusion. A and B: sample voltage responses to step current stimuli. Bottom schematic: time course of the stimulus. Dotted lines, initial resting membrane potential (Erest) (70 mV in A; 74 mV in B). Records were taken from 2 different RGCs. A: responses recorded consecutively in a perforated-patch configuration and in a ruptured-patch configuration (30 s after membrane rupture). Stimulus intensity was fixed at 90 pA. Only in this experiment was the pipette solution supplemented with 2 mM ATP. B: responses recorded at an interval of 10 min in the perforated-patch configuration. Stimulus intensity was fixed at 200 pA.

In contrast, if voltage responses were continuously recorded in the perforated-patch configuration, the firing pattern remained unchanged for 10-30 min (n = 5; Fig. 1B). To avoid the alterations shown in Fig. 1A, we decided to perform electrophysiological analyses in the perforated-patch configuration.

Tonic and phasic RGCs

We recorded the voltage responses to depolarizing current step stimuli in the perforated-patch configuration (104 RGCs; C_m = 15.0 ± 0.1 pF; R_series = 57.0 ± 0.3 M). The resting potential (E_rest) was not adjusted with background current unless otherwise stated. E_rest did not shift by more than a few mV over 30 min (data not illustrated), suggesting that concentrations of major permeant ions were similar in the pipette solution and the cytoplasm.

We found a striking difference in firing accommodations among the RGCs (Fig. 2). Some of the RGCs displayed sustained spike trains that lasted throughout the stimulus period in a certain range of current stimulus intensity (Fig. 2A, middle two traces), like most RGCs of many other vertebrates (Baylor and Fettiplace 1979; Belgum et al. 1983; Diamond and Copenhagen 1995; Fohlmeister and Miller 1997; Lukasiewicz and Werblin 1988). At stimulus intensities above this range, the spike trains of these RGCs were gradually shortened with stimulus intensity (Fig. 2A, bottom trace). Similar stimulus intensity-dependent spike train shortening has been reported in tiger salamander RGCs (firing truncation) (Lukasiewicz and Werblin 1988). In contrast, in most of the remaining RGCs firing was always accommodated within a few hundreds of milliseconds of stimulus onset. This firing accommodation is qualitatively distinct from firing truncation in that it occurred both at lower and higher stimulus intensities (Fig. 2B); RGCs with firing accommodation never displayed sustained firing in response to current step stimuli, even when stimulus intensity was varied in very small increments (5 pA) from the threshold level (Fig. 3).

View larger version (34K): [in this window] [in a new window]   Fig. 2. Tonic and phasic current-evoked responses of RGCs. Typical examples of tonic (A) and phasic (B) voltage responses to 1-s step current stimuli. Here and in Figs. 3-11, all recordings were made in the perforated-patch configuration. Stimulus intensity is indicated above each trace. Bottom schematic: time-course of the stimulus. Erest: 74 mV in A; 75 mV in B. Records were taken from 2 different RGCs with membrane capacitance (Cm) of 5.5 pF (A) and 12.5 pF (B).

We quantitatively compared the firing accommodations of goldfish RGCs, using the duration of the most sustained spike train displayed by each RGC in response to 1-s current step stimuli (D_max) (Fig. 3A). To find the most sustained spike train, stimulus intensity was varied in 5-pA increments. The majority (84.6%) of the tested RGCs (n = 84) either fell into a group with D_max < 200 ms or one with D_max > 800 ms (Fig. 3B and Table 1). We term these groups phasic and tonic RGCs, respectively, and further compared some of their electrophysiological and morphological properties. Mean E_rest was not different between tonic and phasic RGCs (Table 1). Thus the distinction in D_max values was not an artifact caused by the E_rest-dependent modulation of firing pattern that is seen in thalamic neurons (Llinás and Jahnsen 1982).

View larger version (21K): [in this window] [in a new window]   Fig. 3. Heterogeneity in firing accommodation among RGCs. A: maximal duration of spike train evoked by a 1-s step current stimulus (Dmax) was measured in each RGC. Dmax is represented by the time from stimulus onset to the peak of the last spike of the most sustained spike train. To find the most sustained spike train, stimulus intensity was changed in 5-pA increments. B: Dmax for 84 RGCs. RGCs with Dmax <200 and >800 ms were termed phasic and tonic RGCs, respectively, and are further analyzed in Table 1 and Figs. 4-7.

Input-output dynamics of tonic and phasic RGCs

We further compared the intrinsic firing properties of tonic and phasic RGCs, using the firing frequency-stimulus intensity (F-I) relation, which is widely employed to characterize the input-output dynamics of functional RGC subtypes (see, e.g., Mobbs et al. 1992; Thibos and Werblin 1978). We represent firing frequency with an inverse of the interspike interval of the first two spikes elicited by each current step (Fig. 4A) because this representation can efficiently describe transient responses consisting of only a few spikes (McCormick et al. 1985). When current stimulus intensity was normalized by C_m (current density), cell-to-cell data deviation was small within each RGC group (Fig. 4, B and C). Tonic RGCs fired at very low current densities (~2 pA/pF). In contrast, phasic RGCs fired only at higher current densities (6 pA/pF), as indicated by an "instep" formed at the leftmost part of the F-I plot (Fig. 4C). This may result in truncation of the lower part of the dynamic range (the range of input intensity that could be effectively encoded into firing rate) of the phasic RGCs. We calculated the input resistance (R_input) at E_rest from a steady-state shift in E_m caused by a 10-pA hyperpolarizing current step (Table 1). The mean R_input of the phasic RGCs was significantly lower than that of the tonic RGCs. One possibility is that phasic RGCs (but not tonic RGCs) are equipped with an ion current that activates around E_rest and effectively counteracts depolarizing stimuli (Figs. 8-11, see DISCUSSION).

View larger version (12K): [in this window] [in a new window]   Fig. 4. Firing frequency-stimulus intensity (F-I) relations of tonic and phasic RGCs. A: interval between the first two spikes evoked by a 1-s step current stimulus was measured at various stimulus intensities and was used to calculate firing frequency. F-I plots for 20 tonic (B) and 30 phasic (C) RGCs. Stimulus intensity is normalized by Cm (current density). For clarity, data are binned for each 2-pA/pF band. Large dots and error bars: means ± SE. Small dots in B: raw data points. In the case of tonic RGCs, it was difficult to obtain data at stimulus intensities higher than 30 pA/pF because the tonic RGCs were readily damaged by such a strong stimulus. Smooth lines: sigmoid functions fitted to the data.

We quantitatively compared the shape of the F-I plots of the tonic and phasic RGCs by empirically using sigmoid functions (four-parameter logistic functions) that are well fitted to these plots. The sigmoid function was defined as
where F, I, a, b, c, and d are firing frequency, current density, asymptotic minimum, slope parameter, inflection point, and asymptotic maximum, respectively (d > a; Fig. 4, B and C, smooth lines). The slope parameter characterizes the "sigmoidality" of the function. When the slope parameter is close to 1, the function lacks its instep. As the slope parameter increases (above 1), the function becomes more sigmoidal with a more acutely curved instep. The slope parameter of the fitted function was much larger for phasic RGCs than for tonic RGCs (Table 1), which confirms the difference in the F-I plot shape between these RGC groups.

Soma size distributions of tonic and phasic RGCs

We noticed that phasic RGCs tend to be larger than tonic RGCs. We quantitatively compared the soma sizes of the RGCs examined in the current-clamp study, using C_m and cross-sectional soma area (Fig. 5). In the scatter plot of these measures (Fig. 5A), the phasic RGCs (open symbols) are concentrated around the point of 30 ms, 13 pF, and 210 µm². The tonic RGCs (black symbols) are concentrated around the point of 990 ms, 10 pF, and 170 µm² and are loosely scattered in a zone of 800-1000 ms, 25-40 pF, and 350-700 µm². In contrast, the non-tonic, non-phasic RGCs (gray circles) are not concentrated at any point in the plot. The RGCs concentrated on the points in the C_m-D_max dimension largely overlap those in the soma area-D_max dimension (Fig. 5A), suggesting that both measurements describe the same morphological aspect of the RGCs.

View larger version (34K): [in this window] [in a new window]   Fig. 5. Soma size distribution of tonic and phasic RGCs. A: three-dimensional scatter plot of Cm vs. cross-sectional soma area vs. Dmax. Data were obtained from 71 RGCs. B: cross-sectional soma areas for 23 tonic and 44 phasic RGCs. Mean soma area of phasic RGCs is significantly larger than that of tonic RGCs (P < 0.05, Wilcoxon/Kruskal-Wallis rank sum test) (see Table 1).

To compare in detail the soma size distributions of the tonic and phasic RGCs, we re-plotted their soma areas into histograms (Fig. 5B). The majority of the phasic RGCs had larger values than did the tonic RGCs. Of the tonic RGCs, 60% were concentrated in a range of 100-250 µm² whereas 70% of the phasic RGCs were concentrated in a range of 200-500 µm² (Fig. 5B). Mean soma area as well as mean C_m were significantly different between the tonic and phasic RGCs (Table 1). In addition, there were a few tonic RGCs that had exceptionally large soma areas (>650 µm²; Fig. 5B), which corresponds to RGCs with C_m  34 pF. These values exceeded those of the largest phasic RGC. It is noteworthy that larger RGCs survived better than did smaller RGCs in vitro (data not illustrated). Thus we might have underestimated the fractional population of smaller RGCs.

Possible ion current underlying firing accommodation

Mathematical models of vertebrate RGCs incorporating previously identified ion currents display sustained firing, but not transient firing, to current step stimuli (see, e.g., Fohlmeister and Miller 1997). This suggests that phasic goldfish RGCs might express an unidentified current which produces firing accommodation. We searched for such an ion current with perforated-patch current-clamp and perforated-patch voltage-clamp techniques.

In central neurons, firing accommodation could be produced by several classes of K⁺ currents including 1) a low-threshold, noninactivating K⁺ current with or without coupling to muscarinic acetylcholine receptor (AChR) (we hereafter term this class of currents I_K,ln); 2) a Ca²⁺-activated K⁺ current (I_K,Ca); and 3) a 4-aminopyridine (4AP)-sensitive, slowly inactivating K⁺ current (I_K,4AP-s) (Brown et al. 1990; Del Negro and Chandler 1997; Storm 1990). We used antagonists preferring either of the K⁺ currents to test for the involvement of these currents in producing firing accommodations in phasic RGCs (Fig. 6).

View larger version (17K): [in this window] [in a new window]   Fig. 6. Effects of various ion channel blockers on firing accommodation. A-F: sample voltage responses to 1-s step current stimuli before and during extracellular application of indicated drugs. In A, response after drug wash-out (Recovery) is also shown. Dotted lines: Erest in control bath solution (58 to 79 mV). Stimulus intensity was fixed at a certain level (40-210 pA) throughout each set of experiments. Records were taken from 6 different RGCs. A: Ba2+ (1 mM) abolishes firing accommodation reversibly. Inset: close-up of the responses in the top panels. To compare the repolarizing phases, traces were justified at the peak of the second spike (*). B-D: neither Co2+ (2.5 mM) (B), apamin (1 µM) (C), iberiotoxin (IbTx; 50 nM (D), 4-aminopyridine (4AP; 1 mM) (E), or DIDS (1 mM) (F) abolished firing accommodation. Ca2+ in the control bath solution was totally replaced with Co2+ to minimize change in membrane surface charge.

In the control bath solution, the current-evoked responses of the phasic RGCs did not alter for 10-30 min (Fig. 1B). Ba²⁺, which is known to antagonize I_K,ln (Rudy 1988), rapidly abolished firing accommodation at a concentration of 1 mM (n = 9; Fig. 6A) but not at a lower concentration (0.1 mM, n = 3, data not illustrated). The R_input at E_rest significantly increased from 0.32 ± 0.12 to 0.87 ± 0.12 G after application of 1 mM Ba²⁺ (P < 0.05, Student's paired t-test, n = 4). Thus the effect of 1 mM Ba²⁺ could be ascribed to a conductance decrease presumably caused by I_K,ln blockade but not to a conductance increase caused by Ba²⁺ influx through Ca²⁺ channels. In contrast, total or partial replacement of Ca²⁺ in a bath solution with Co²⁺ (n = 3 for 2.5 mM, n = 7 for 2.4 mM), which has been reported to suppress depolarization-induced increase in cytoplasmic free calcium levels ([Ca²⁺]_I) (Ishida et al. 1991) and I_K,Ca (Ishida 1991), did not abolish firing accommodation (Fig. 6B). Also, apamin (1 µM) (Blatz and Magleby 1986) (Fig. 6C) and iberiotoxin (IbTx, 50 nM) (Galvez et al. 1990) (Fig. 6D), specific blockers against major small- and large-conductance Ca²⁺-activated K⁺ channels (SK and BK channels), respectively, did not abolish firing accommodation. To confirm the potency of the batches of apamin and IbTx used in the present study, we used cultured rat cerebellar Purkinje neurons that express both SK and BK channels (unpublished data). 4AP (40 µM-1 mM, n = 6, Fig. 6E) did not abolish firing accommodation. These results suggest that, among the three classes of K⁺ currents, I_K,ln is the most important for producing firing accommodation in RGCs. Moreover, a current underlying firing accommodation might not couple to muscarinic AChR because neither muscarine (30 µM, n = 5) or carbachol (an agonist for muscarinic and nicotinic AchR) (50-500 µM, n = 4) affected firing accommodation (data not illustrated).

In addition, we tested for the involvement of an outwardly rectifying Cl current (I_Cl) identified in goldfish RGCs (Tabata and Ishida 1999) because I_Cl shares noninactivating kinetics with I_K,ln. At a concentration of 1 mM, DIDS, a potent blocker against I_Cl (Tabata and Ishida 1999), did not abolish firing accommodation (n = 4) (Fig. 6F). Thus I_Cl might be less important for producing firing accommodation.

Contribution of Ba²⁺-sensitive current to intrinsic firing property heterogeneity

To examine the extent of the contribution of Ba²⁺-sensitive current to the differences in firing property between tonic and phasic RGCs, we measured the F-I relation of phasic RGCs in the presence of 1 mM Ba²⁺ (Fig. 7) (n = 5). Under this condition, the instep characteristic of the F-I plot for untreated phasic RGCs (Fig. 4C) completely disappeared. The sigmoid function originally generated for the F-I plot of untreated tonic RGCs (Fig. 4B) was well fitted to that of Ba²⁺-treated phasic RGCs (Fig. 7; the function was scaled along the y-axis but its parameters were not modified to preserve the shape of the original function). Thus Ba²⁺ treatment makes the input-output dynamics of phasic RGCs indistinguishable from those of tonic RGCs. This result suggests that the Ba²⁺-sensitive current expressed in phasic RGCs may largely explain the differences in firing properties between tonic and phasic RGCs.

View larger version (20K): [in this window] [in a new window]   Fig. 7. F-I relation of Ba2+-treated phasic RGCs. Plot of F-I relation measured in 5 phasic RGCs in the presence of 1 mM extracellular Ba2+. Firing frequency was measured as in Fig. 4. For clarity, data are binned for each 2-pA/pF band. Dots and error bars: means ± SE. Smooth line: sigmoid function originally fitted to the F-I plot for 20 tonic RGCs (Fig. 4B). The function is scaled along the y-axis at a factor of 1.2 but the values characterizing the parameters of the original function's shape are not modified.

Activation kinetics of Ba²⁺-sensitive current

We characterized the Ba²⁺-sensitive current under voltage-clamp conditions in the perforated-patch configuration. The results shown in Figs. 8-11 were obtained from large RGCs with C_m of 15.3 ± 1.1 pF (n = 31; R_series = 33.3 ± 5.4 M), most of which were expected to be phasic, based on the C_m distribution (Fig. 5A).

View larger version (23K): [in this window] [in a new window]   Fig. 8. Activation kinetics of barium-sensitive current (IBa-s). A: in this and in Figs. 9-11, IBa-s is extracted as the difference between whole-cell currents recorded under voltage clamp before (Control) and during (Ba2+) extracellular application of 1 mM Ba2+. The control and test bath solutions contained 2.4 mM Co2+, 1 µM TTX, and 1 mM 4AP to reduce voltage-gated Ca2+ and Na+ currents and voltage-gated K+ currents, which might be less important for firing accommodation (cf. Fig. 6). Note that IBa-s (slow current) is preceded by a fast current (arrow) when the holding potential is set as negative as 90 mV (see RESULTS for further explanation). Bottom schematics: time-courses of command potentials. For simplicity, only 4 traces are shown in "Difference." B: steady-state I-V plot of IBa-s. Steady-state current amplitude was measured as mean current amplitude during 0.9-1.0 s of a test potential step and normalized by Cm (current density). Dots and error bars: means ± SE. Data were taken from 6 RGCs. C: IBa-s elicited at indicated test potentials. Note that the fast current in A is absent when the holding potential is set at 65 mV. Gray lines: single exponential functions fitted to the IBa-s. Bottom schematics: time courses of command potentials. D: time constant of exponential functions fitted to IBa-s as in C plotted against test potential. Dots and error bars: means ± SE. Data were taken from 4 RGCs.

View larger version (13K): [in this window] [in a new window]   Fig. 9. Carrier ion species and deactivation kinetics of IBa-s. A-C: tail currents of IBa-s on repolarization from 0 mV (400 ms) to test potentials of 40 to 110 mV (increment, 10 mV; schematics in A). IBa-s was extracted by the subtraction method as in Fig. 8A. A: typical examples of tail currents recorded from one RGC (dotted lines). For simplicity, only the tail currents at 3 test potentials are shown. Smooth lines: single exponential functions fitted to the data. B: instantaneous I-V relation of tail currents measured at 3.5 ms of repolarization. Current amplitude is normalized by the value at a test potential of 80 mV. Dots and error bars: means ± SE. Data were taken from 4 RGCs. The Erev estimated by linear regression to the data (line) is 92 mV and close to the potassium equilibrium potential (EK) (98 mV). C: time constant of the fitted function plotted against test potential. Large dots and error bars: means ± SE. Small dot: raw data point. Smooth line: sigmoid function fitted by eye. Data were taken from 4 RGCs.

View larger version (19K): [in this window] [in a new window]   Fig. 10. Pharmacology of IBa-s. A-C: IBa-s was extracted as a difference between current recorded in the presence of the indicated toxin [A: 30 mM tetraethylammonium (TEA); B: 1 µM apamin; C: 50 nM IbTx] and that recorded during additional application of Ba2+ (1 mM; Plus Ba2+). B and C: total currents recorded before toxin treatment are also shown (Before). Bath solutions are the same as in Fig. 8 except that 30 mM Na+ was replaced with 30 mM TEA in A and Co2+ was not included (2.5 mM Ca2+) in B and C. Total currents in A are relatively small because TEA blocks the major voltage-gated K+ currents. Schematic in A: time course of command potential. Holding potential was set at 65 mV. Dotted lines: zero-current level.

View larger version (17K): [in this window] [in a new window]   Fig. 11. Relative contribution of IBa-s to total membrane conductance. A: for each 20-mV test potential band, membrane conductance was calculated from the steady-state I-V slope measured as in Fig. 8A (slope conductance). Values are normalized by Cm. Open symbols: total conductance measured in standard bath solution containing 2.4 mM Co2+. To minimize change in membrane surface charge, Co2+ was applied by the partial replacement of 2.5 mM Ca2+ in the control bath solution. Black symbols: Ba2+-sensitive conductance extracted as a difference between conductances before and during extracellular application of 1 mM Ba2+. Dots and error bars: means ± SE. Data were taken from 10 RGCs. B: fractional contribution of Ba2+-sensitive conductance to total conductance for each 20-mV test potential band. Dots and error bars: means ± SE. Data were taken from the same RGCs as in A.

In the control bath solution containing Co²⁺ (2.4 mM), 4AP (1 mM), and TTX (1 µM), the total whole-cell current activated during a depolarizing test potential step was seen as an outward current (Fig. 8A, control). At the concentration at which it abolished firing accommodation (1 mM), extracellular Ba²⁺ selectively blocked two components of the total current that are demonstrated as differences between currents recorded before and during Ba²⁺ application (Fig. 8A, difference). One component was a "fast" current that was activated within a few msec and was inactivated within 200 ms (Fig. 8A, arrow). The other component was a "slow" current that became obvious following the decay of the fast current and was sustained throughout a 2-s test potential step without inactivation. The slow current also was distinguished from the fast current in its lower activation threshold. As shown by its I-V plot (Fig. 8B), the slow current was activated at potentials of 70 mV and above. In contrast, the fast current was not activated even at a test potential as positive as 50 mV (Fig. 8A, difference). These differences between inactivation and activation kinetics indicate that these two currents are carried by different ion channels. In the following analyses, we focus on the slow current (hereafter termed I_Ba-s) because this current appears to be much more important for shaping firing patterns than does the fast current. The fast current was completely inactivated at potentials as negative as 65 mV (Fig. 8C; note that the fast current is not seen with a holding potential of 65 mV). This result suggests that the fast current is usually inactivated around the normal E_rest (approximately 65 mV) (Table 1) and thus is not very active in the early period of a repetitively firing response. In addition, higher concentrations of Ba²⁺ were not used for the extraction of I_Ba-s because 3-10 mM Ba²⁺ blocked not only I_Ba-s but also a slowly inactivating current (n = 5, data not illustrated).

We examined the voltage-dependence of I_Ba-s activation kinetics, fitting the single exponential function to the activation phase (rise) (Fig. 8C). The time constant of the fitted function decreased with more positive test potentials (Fig. 8D), suggesting that the activation of I_Ba-s is accelerated at more positive potentials.

Carrier ion species and deactivation kinetics of I_Ba-s

To examine the carrier ion species and deactivation kinetics of I_Ba-s, we measured the deactivation phase (tail current) of I_Ba-s, which was seen on repolarization, from a conditioning potential of 0 mV to more negative test potentials (Fig. 9). The tail current reversed direction between test potentials of 90 and 100 mV (n = 4) (Fig. 9, A and B). From a line fitted to the instantaneous I-V plot of the tail currents, the reversal potential (E_rev) of I_Ba-s was estimated to be 92 mV (Fig. 9B). This value is close to the K⁺ equilibrium potential (E_K) set by the pipette and bath solutions (98 mV). Even if it is assumed that the deviations of E_rev from E_K were entirely caused by an auxiliary flux of Na⁺ through the I_Ba-s channels, the relative permeability of Na⁺ over K⁺ (P_Na/P_K) would be as small as 0.005 (Goldman-Hodgkin-Katz equation with E_Na of 68 mV). Therefore, I_Ba-s is thought to be selectively carried by K⁺. The tail currents were well fitted by single exponential functions (Fig. 9A). The time constant of the fitted functions decreased with more hyperpolarized test potentials (Fig. 9C), suggesting that the deactivation of I_Ba-s is accelerated by hyperpolarization.

Pharmacology of I_Ba-s

We examined whether I_Ba-s is identical to any of the voltage-gated K⁺ currents with slow kinetics previously reported in vertebrate RGCs. Goldfish RGCs possess a tetraethylammonium (TEA)-resistant, voltage-gated K⁺ current with very slow inactivation kinetics (see Tabata and Ishida 1996, 1999). In goldfish RGCs preincubated with extracellular TEA (30 mM), additional application of 1 mM Ba²⁺ did not block a time-dependent current (Fig. 10A). Thus I_Ba-s is sensitive to TEA and is distinguished from the TEA-resistant, voltage-gated K⁺ current.

Vertebrate RGCs possess I_K,Ca (see, e.g., Lipton and Tauck 1987; Lukasiewicz and Werblin 1988; Rothe et al. 1999; Wang et al. 1998). SK channel, one of the two major channel types responsible for I_K,Ca, is slowly activated following a rise in the intracellular Ca²⁺ level (time constant 1-1.6 s) (see Brown et al. 1990). Therefore, SK channel may produce a slowly activating K⁺ current during depolarizing events accompanied by Ca²⁺ entry through voltage-gated Ca²⁺ channels. However, SK channel did not appear to mediate I_Ba-s. We could extract I_Ba-s in RGCs preincubated with the full-blocking concentration of apamin (1 µM) (Blatz and Magleby 1986) (Fig. 10B). Also, BK channel, the other major I_K,Ca-responsible channel, did not mediate I_Ba-s because I_Ba-s could be extracted in RGCs preincubated with the full-blocking concentration of IbTx (50 nM) (Galvez et al. 1990) (Fig. 10C). In addition, the steady-state densities of the apamin- and IbTx-sensitive components of the total current activated at a potential of 30 mV (mean across 0.9-1.0 s of the test potential step; 2.5 mM Ca²⁺, Co²⁺ was not included in the bath to allow Ca²⁺ influx through the voltage-gated Ca²⁺ channels) (Fig. 10, B and C) were 5.32 ± 4.29 pA/pF (n = 5 including 3 cases in which the corresponding component was undetectable) and 7.46 ± 2.87 pA/pF (n = 6), respectively.

Moreover, the total currents (including I_Ba-s) activated at test potentials from 110 to 30 mV were not reduced by muscarine (100 µM, n = 5, data not illustrated).

Contribution of I_Ba-s to total membrane conductance

To understand how I_Ba-s produces firing accommodation in phasic RGCs, we evaluated the relative contribution of I_Ba-s to total membrane conductance (Fig. 11). First, we measured the steady-state I-V relations of control current and I_Ba-s, as in Fig. 8. Next, we used these I-V slopes to calculate the conductance of the control component (g_control) and that of the Ba²⁺-sensitive component (g_Ba-s) (Fig. 11A). Finally, we plotted the relative contribution (g_Ba-s/g_control) as a function of test potential (Fig. 11B). In this experiment, we suppressed Ca²⁺ currents with extracellular Co²⁺ because repetitive depolarization might cause the cytoplasmic deposit of Ca²⁺ and the gradual facilitation of I_K,Ca.

At test potentials between 90 and 30 mV, g_Ba-s constitutes nearly half of g_control (Fig. 11B). At more positive potentials, in contrast, g_Ba-s constitutes only ~10% of g_control (Fig. 11B). The actual contribution of g_Ba-s to the total conductance in this range of E_m might be smaller than this value because I_Ca and I_K,Ca, which typically activate at potentials above 45 mV (Bindokas and Ishida 1996; Tabata et al. 1996), were not included in g_control. These results suggest that I_Ba-s may substantially affect subthreshold changes in the E_m but not the waveforms of individual spikes (see DISCUSSION).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Heterogeneity of intrinsic firing properties in goldfish RGCs

In the perforated-patch, whole-cell configuration, many goldfish RGCs displayed phasic current-evoked responses (Figs. 2 and 3). This result contrasts with previous observations that most RGCs in other vertebrates display only tonic current-evoked responses (Baylor and Fettiplace 1979; Belgum et al. 1983; Diamond and Copenhagen 1995; Fohlmeister and Miller 1997; Lukasiewicz and Werblin 1988), except for some RGCs in the tiger salamander (Mobbs et al. 1992) and the immature rat (Barres et al. 1988). This discrepancy may be caused by species variation and/or differences in recording configuration. Some of the previous observations were made in ruptured-patch configurations. Under these recording conditions, cytoplasmic disturbance might alter the intrinsic firing properties, including firing accommodation (Fig. 1).

The majority of the goldfish RGCs examined in the perforated-patch configuration were classified into tonic and phasic subtypes based on maximal firing duration (Fig. 3). The tonic RGCs made little adaptation in firing frequency throughout current stimuli (Fig. 2). Tiger salamander RGCs with similar intrinsic firing properties show light responses whose temporal patterns directly reflect the time-course of the synaptic inputs (Diamond and Copenhagen 1995). Thus tonic goldfish RGCs may exhibit various temporal patterns of light response in vivo, depending on the synaptic inputs. Some studies using isolated retinae (Cohen 1998; Mobbs et al. 1992) showed that, in the tiger salamander and the cat, most RGCs receive depolarizing synaptic inputs that slowly decay during excitatory light stimuli. If this is also true of goldfish, tonic RGCs will exhibit sustained light responses. However, if some of the tonic RGCs are predominantly governed by fast decaying excitatory synaptic drives generated with the aid of amacrine cells (see Nirenberg and Meister 1997), they will exhibit transient light responses. In contrast, in phasic goldfish RGCs, firing accommodation may emphasize the transience of the light response, reducing an RGC's excitability during prolonged excitatory synaptic inputs.

Do the differences in the intrinsic firing properties of tonic and phasic RGCs actually contribute to shaping the distinct light responses seen in vivo? Although there is no direct evidence of this, there are some observations that support this possibility. First, tonic and phasic goldfish RGCs had input-output dynamics similar to those of sustained and transient vertebrate RGCs, respectively. As shown by the instep of the F-I plots (Fig. 4), phasic goldfish RGCs had a higher threshold than did the tonic RGCs. A similar difference is seen in situ between the transient and sustained RGCs of the tiger salamander (Mobbs et al. 1992) and the mudpuppy (Thibos and Werblin 1978). Second, tonic and phasic goldfish RGCs had soma size distributions similar to those of sustained and transient goldfish RGCs, respectively. Goldfish RGCs are classified into ON and OFF subtypes, which display relatively sustained responses to stepped illumination of a particular wavelength (typically red), and the ON-OFF subtype, which displays a relatively transient response (see, e.g., Spekreijse et al. 1972). An intracellular recording/staining study (Vallerga and Djamgoz 1991) and an axonal conduction velocity measurement (Northmore and Oh 1998) suggested that OFF goldfish RGCs correspond to a morphological subtype with the largest somata (Cook et al. 1992; Hitchcock and Easter 1986). A few tonic RGCs with exceptionally large somata (Fig. 5) may correspond to OFF RGCs. These RGCs might also overlap Y-like RGCs to a large extent because OFF RGCs often are identified as Y-like RGCs (Bilotta and Abramov 1989). Moreover, the velocity measurement (Northmore and Oh 1998) suggests that the ON subtype contains more small RGCs than does the ON-OFF subtype. With respect to relative soma size (Fig. 5; Table 1), tonic and phasic RGCs might partly correspond to the ON and the ON-OFF RGCs, respectively. In addition, most ON and ON-OFF RGCs are identified as W- and Y-like RGCs (Bilotta and Abramov 1989). Assuming that the morphological correlation of X/Y/W RGCs in the cat (cat X, Y, and W RGCs have medium-sized, large, and small somata, respectively; see Stone et al. 1979 for review) is applicable to goldfish, the main components of tonic and phasic RGCs might be W- and Y-like RGCs, respectively.

Taken together, the results of the present current-clamp study demonstrate that goldfish RGCs have heterogeneous intrinsic firing properties that may be consonant with the temporal patterns of light responses.

Identification and possible function of I_Ba-s

Under voltage clamp conditions, we extracted a Ba²⁺-sensitive, voltage-dependent K⁺ current in isolated goldfish RGCs as a difference between currents recorded before and after external Ba²⁺ application (I_Ba-s) in isolated goldfish RGCs. I_Ba-s had a low activation threshold negative to 70 mV, was activated and deactivated slowly with time constants of 10-100 ms, and was not inactivated during depolarization for as long as 2 s (Figs. 8 and 9).

There are at least two general possibilities concerning the molecular nature of I_Ba-s. One is that I_Ba-s is a new current mediated by a K⁺ channel(s) highly sensitive to Ba²⁺. Another possibility is that I_Ba-s reflects a time-dependent reduction of a voltage-gated K⁺ current(s) caused by Ba²⁺-induced K⁺ channel modulation such as facilitation of inactivation. At present, a biological toxin that selectively blocks a specific channel(s) responsible for I_Ba-s has not been found. Positive identification of the molecular nature of I_Ba-s has yet to be made by single-channel recordings and molecular analyses of the putative channel protein. However, several lines of evidence obtained in this and previous studies support the former possibility. I_Ba-s was resistant to 4AP and Co²⁺ (Fig. 8) but was completely blocked by TEA (Fig. 10). Therefore, if the latter possibility were the case, I_Ba-s might be derived from a 4AP/Co²⁺-resistant, TEA-sensitive, voltage-gated K⁺ current(s), which mainly consists of a delayed rectifier K⁺ current(s) (I_K,V) in vertebrate RGCs (see Ishida 1995 for review). Contrary to this expectation, I_Ba-s has a much more negative activation threshold than do these I_K,V's (above 55 mV) (see, e.g., Lipton and Tauck 1987; Lukasiewicz and Werblin 1988; see Ishida 1995 for review). Moreover, I_K,V is generally known as a primary ionic mechanism that forms the repolarizing phase of a spike (Storm 1990). Thus Ba²⁺ would hamper the repolarizing phase of a spike, assuming the second possibility, whereas Ba²⁺ did not reduce or slow the repolarizing phase in goldfish RGCs (n = 5) (Fig. 6A, inset). In addition, closely related retinal cells (photoreceptors) possess a Ba²⁺-sensitive K⁺ current that may be mediated by a specific channel (see the following paragraphs). We compare in detail the basic properties of I_Ba-s with those of various K⁺ currents with slow kinetics.

Vertebrate RGCs possess I_K,Ca as well as I_K,V (see Ishida 1995 for review). I_K,V constitutes the major part of depolarization-activated K⁺ currents in vertebrate RGCs (see, e.g., Lukasiewicz and Werblin 1988; Sucher and Lipton 1992). Therefore, the kinetics of the Ba²⁺-resistant current measured in the present study (Fig. 8A) may reflect those of I_K,V. I_Ba-s appears to differ from I_K,V because I_Ba-s showed much more slow activation than did the Ba²⁺-resistant current. I_Ba-s shares voltage sensitivity with I_K,Ca mediated by BK channel (see Sah 1996 for review). However, I_Ba-s differs from BK channel-mediated current in its resistance to IbTx (Fig. 10). In addition, I_Ba-s differs from I_K,Ca mediated by apamin-sensitive SK channels (see Sah 1996 for review) in its resistance to apamin (Fig. 10). Recent studies (Hirschberg et al. 1998; Kohler et al. 1996) show that mammalian central neurons express an apamin-insensitive SK channel (SK1). Moreover, mammalian neurons possess an apamin-insensitive I_K,Ca that is probably mediated by another channel. However, I_Ba-s may also differ from these two apamin-insensitive currents because the activation of I_Ba-s is voltage-dependent (Fig. 8), unlike these currents (Hirschberg et al. 1998; Sah 1996). In the present study, we measured I_Ba-s using a bath solution containing 2.4 mM Co²⁺ and 0.1 mM Ca²⁺ (Figs. 8, 9, and 10A). Under this condition, the [Ca²⁺]_i of goldfish RGCs is fixed to the resting level, even when E_m is depolarized (~120 nM) (Ishida et al. 1991). Thus the voltage-dependence of I_Ba-s is not an artifact produced by an increase in [Ca²⁺]_i caused by Ca²⁺ influx through voltage-gated Ca²⁺ channels or by Ca²⁺-induced Ca²⁺ release from the intracellular store associated with such Ca²⁺ influx. Taken together, the data show that I_Ba-s may differ from major I_K,Ca's, although we do not exclude the possibility that I_Ba-s belongs to a previously unidentified class of I_K,Ca.

In other cell types, there are several classes of low-threshold, noninactivating K⁺ currents (I_K,ln's), including S current in Aplysia sensory neurons (I_S) (Siegelbaum et al. 1982), standing outward current in rat cerebellar granule cells [I_K(SO)] (Watkins and Mathie 1996), muscarine-sensitive current in vertebrate neurons (I_M) (Brown and Adams 1980), and I_K,ln's that kinetically resemble I_M but lack muscarine sensitivity. I_Ba-s differs from I_S and I_K(SO) either in pharmacological properties or kinetics. I_S is resistant to 10 mM Ba²⁺ (Shuster and Siegelbaum 1987). I_K(SO) is rapidly activated and deactivated with time constants of sub-milliseconds (Watkins and Mathie 1996). I_Ba-s more resembles I_M, sharing susceptibilities to 10³ M of Ba²⁺ and 10² M of TEA, resistance to 10³ M of 4AP, and slow activation and deactivation kinetics (time constants, 10-100 ms) (Adams et al. 1982a,b). However, I_Ba-s was not downregulated by muscarinic agonists (data not illustrated), unlike I_M (Brown and Adams 1980), although vertebrate RGCs express muscarinic AChRs (see Fischer et al. 1998). I_Ba-s also differs from I_M in its lower activation threshold (Adams et al. 1982a). With respect to activation threshold, I_Ba-s resembles a muscarine-insensitive I_K,ln found in smooth muscle cells (Evans et al. 1996). Recently, a muscarine-insensitive I_K,ln was found in closely related retinal cells (I_Kx in salamander rod photoreceptors) (Beech and Barnes 1989; Wollmuth 1994). Therefore, one possibility is that I_Ba-s forms a new class of I_K,ln with I_Kx. Further identification of I_Ba-s would be established by molecular comparisons between I_Ba-s channels and previously cloned I_K,ln channels such as ether à go-go (Hoshi et al. 1998; Warmke et al. 1991) and aK5.1 (Zhao et al. 1994).

At the concentration at which it selectively blocked I_Ba-s (1 mM), external Ba²⁺ abolished firing accommodation in phasic RGCs (Fig. 6). Thus I_Ba-s is an ionic mechanism sufficient to explain the firing accommodation. In an analogy of I_M, a K⁺ current kinetically resembling I_Ba-s (see the previous paragraph) (Brown et al. 1990; McCormick 1990), a possible action of I_Ba-s is depicted as follows. When a goldfish RGC rests at a potential of approximately 65 mV (Table 1), the membrane conductance largely consists of leak K⁺ and Na⁺ currents (Tabata and Ishida 1996, 1999) and I_Ba-s (Figs. 8 and 11). At the onset of a depolarizing current step stimulus, E_m is readily depolarized because an opposing electrical force against the depolarizing stimulus, caused by the resting membrane conductance, is relatively small. E_m rapidly reaches the spike threshold (approximately 45 mV) (Ishida 1991). Following the repolarizing phase of each spike, the current stimulus again depolarizes E_m toward the spike threshold. The activation level of I_Ba-s may be slightly increased by depolarization during each spike (Fig. 8). This additional activation can be temporarily summated because of I_Ba-s's slowly deactivating property (Fig. 9). This summation causes a gradual increase in I_Ba-s-mediated K⁺ conductance through repetitive firing. Therefore, at the late period of a current step stimulus, the electrical force opposing the depolarizing stimulus should be greatly increased. With this increased opposing force, E_m becomes more reluctant to depolarize and stays at subthreshold levels for a longer time. This prolonged subthreshold depolarization causes failure of spike firing by hampering the disinactivation of, and by facilitating the inactivation of, voltage-gated Na⁺ channels (McCormick 1990). The involvement of I_Ba-s in the prolongation of subthreshold depolarization is indicated by the accelerated subthreshold depolarization seen after Ba²⁺ application (n = 5) (Fig. 6A, inset; note a Ba²⁺-induced change in the time-course after the asterisked spikes).

Voltage-gated K⁺ currents other than I_Ba-s and major I_K,Ca's appeared to be less important for firing accommodation than did I_Ba-s (Fig. 6). Total conductance recorded in the absence of external Ba²⁺ shows a drastic increase at potentials more than approximately 50 mV (Fig. 11), indicating that the majority of Ba²⁺-resistant K⁺ currents have relatively high activation thresholds as compared with I_Ba-s. In addition, I_K,V, the primary component of voltage-gated K⁺ currents in vertebrate RGCs (see, e.g., Lukasiewicz and Werblin 1988; Sucher and Lipton 1992) is deactivated within a few msec of the cessation of depolarization. These kinetic properties may prevent these voltage-gated K⁺ currents from cooperating with I_Ba-s to form an electrical force opposing depolarizing current stimuli at subthreshold potentials. In some neurons, I_K,Ca's (particularly those mediated by SK channels) are gradually activated by Ca²⁺, which enters during repetitive firing, and modulate the firing pattern in a time-dependent manner (Brown et al. 1990; McCormick 1990). However, potent blockers of I_K,Ca's did not abolish firing accommodation in goldfish RGCs (Fig. 6). The mean densities of the apamin- and IbTx-sensitive currents (at test potentials of 30 mV, 5.32 ± 4.29 and 7.46 ± 2.87 pA/pF, respectively; see RESULTS) were less than half that of I_Ba-s (15.73 ± 4.25 pA/pF) (Fig. 8). Thus the functional contribution of I_K,Ca's may be relatively smaller than that of I_Ba-s in goldfish RGCs.

The F-I relation of tonic RGCs was indistinguishable from that of phasic RGCs whose I_Ba-s was blocked with Ba²⁺ (Fig. 7). This suggests that the functional contribution of I_Ba-s is negligible in tonic RGCs. Therefore, I_Ba-s may be the primary factor producing the heterogeneity in intrinsic firing property between tonic and phasic RGCs.