Electrical resonance and Ca2+ influx in the synaptic terminal of depolarizing bipolar cells from the Goldfish retina


To whom correspondence should be addressed.


  • 1Whole-cell recordings and fura-2 measurements of cytoplasmic [Ca2+] were made in depolarizing bipolar cells isolated from the retinae of goldfish. The aim was to study the voltage signal that regulates Ca2+ influx in the synaptic terminal.
  • 2The current-voltage relation was linear up to about −44 mV. At this threshold, the injection of 1 pA of current triggered a maintained ‘all-or-none’ depolarization to a plateau of −34 mV, associated with a decrease in input resistance and a damped voltage oscillation with a frequency of 50–70 Hz and initial amplitude of 4–10 mV. A second frequency component of 5–10 Hz was often observed. In a minority of cells the response to current injection was transient, recovering with an undershoot.
  • 3Unstimulated bipolar cells generated similar voltage signals, driven by current entering the cell through a non-specific cation conductance that continuously varied in amplitude.
  • 4The threshold for activation of the Ca2+ current was −43 mV and free [Ca2+]i in the synaptic terminal rose during a depolarizing response. Simultaneous measurements of the fluorescence associated with the membrane marker FM1–43 demonstrated that these Ca2+ signals stimulated exocytosis. Regenerative depolarizations and associated rises in [Ca2+]i were blocked by inhibiting l-type Ca2+ channels with 30 μm nifedipine.
  • 5Depolarization beyond −40 mV also elicited an outwardly rectifying K+ current. Blocking this current by replacing external Ca2+ with Ba2+ caused the voltage reached during a depolarizing response to increase to +10 mV.
  • 6The majority of the K+ current was blocked by 100 nm charybdotoxin, indicating that it was carried by large-conductance Ca2+-activated K+ channels. A transient voltage-gated K2+current remained, which began to activate at −40 mV. High-frequency voltage oscillations were blocked by 100 nm charybdotoxin, but low-frequency oscillations remained.
  • 7These results indicate that the voltage response of depolarizing bipolar cells is shaped by l-type Ca2+ channels, Ca+-activated K+ channels and voltage-dependent K+ channels. This combination of conductances regulates Ca2+ influx into the synaptic terminal and confers an electrical resonance on the bipolar cell.

Retinal bipolar cells receive synaptic inputs from photo-receptors and contribute to the transformation of the visual signal as it is transmitted to ganglion cells that send the information back to the brain (DeVries & Baylor, 1993). These neurons fall into two classes: ‘on’ bipolar cells depolarize in the light, while ‘off’ bipolar cells hyperpolarize. Bipolar cells cannot fire sodium-dependent action potentials, but it is not known how the membrane conductances in these cells act to shape the voltage response to light.

Bipolar cells can act as high-pass filters (Bialek & Owen, 1990). In the retinae of carp and dogfish, the signal in ‘on’ bipolar cells in response to a bright step of light usually consists of a single transient depolarization with undershoot, followed by a maintained depolarization of smaller amplitude (Saito, Kondo & Toyoda, 1979; Ashmore & Falk, 1980; Saito & Kujiraoka, 1982). This transformation of the visual signal is apparent in the retinal output, since many ganglion cells respond more strongly to rapid increases in light intensity than to steady illumination (Hubel, 1995).

Bipolar cells can also generate responses characteristic of bandpass filters that respond maximally at a particular frequency. In dogfish, the synaptic transfer function from rods to bipolar cells for dim lights has maximum gain at 3–5 Hz, and voltage oscillations at these frequencies have been observed in response to light (Ashmore & Falk, 1979, 1980). Tuning at higher frequencies has been observed in the retinae of carp and goldfish, where ‘on’ bipolar cells can generate a damped oscillation at 30–40 Hz superimposed on the maintained depolarization caused by a step of light (Kaneko & Hashimoto, 1969; Kaneko, 1970). A similar oscillatory response has been observed in ‘off’ bipolar cells when they depolarize after light is turned off (Kaneko & Hashimoto, 1969). Electrical oscillations at frequencies of 30–80 Hz have also been noted using extracellular electrodes in the retinae of frogs (Tomita & Funaishi, 1952; Brindley, 1956) and dogs and monkeys (Svaetichin, 1961), and in each case these were largest when the electrode was in the region of the bipolar cells. The role of high-frequency oscillations in the retina is not clear, but they have been observed in the retinal output generated by ganglion cells in a wide variety of animals (Adrian & Matthews, 1928; Granit, 1933; Doty & Kimura, 1963; Bishop, Levick & Williams, 1964; Neuenschwander & Singer, 1996).

It is not known how far the responses observed in bipolar cells are generated by intrinsic membrane conductances as opposed to other elements of the retinal circuit. We have therefore investigated the voltage response to current injection of ‘on’ bipolar cells isolated from the goldfish retina, with particular reference to the voltage range in which Ca2+ enters the synaptic terminal. Kaneko & Tachibana (1985) have shown that goldfish bipolar cells possess a variety of voltage-dependent conductances, including an L-type Ca2+ conductance that is localized to the synaptic terminal and underlies neurotransmitter release (Tachibana, Okada, Arimura, Kobayashi & Piccolino, 1993), and voltage- and Ca2+-dependent K+ currents. We find that the L-type Ca2+ conductance can support a regenerative depolarization, and that Ca2+-and voltage-dependent K+ conductances limit the amplitude of this response to less than 20 mV. The opposing effects of the Ca2+ and K+ currents contribute to an electrical resonance that can cause the membrane potential to oscillate at about 60 Hz, with a second modulation at about 5 Hz. The conductances underlying the electrical resonance in ‘on’ bipolar cells appear similar to those underlying the resonance of hair cells in the cochlea and sacculus (Crawford & Fettiplace, 1981; Hudspeth & Lewis, 1988). A preliminary account of some of these results have been presented to the Biophysical Society (Burrone & Lagnado, 1997).


Cell isolation

The Ringer solution normally contained the following (mm): 110 NaCl, 2.5 CaCl2, 2.5 KCl, 1 MgCl2, 10 glucose and 10 Hepes (pH 7.3). Bipolar cells were acutely dissociated from the retinae of goldfish by papain digestion, using methods described by Tachibana & Okada (1991). Goldfish were dark adapted overnight, then killed by decapitation followed by immediate pithing of the brain and spinal cord. The eyeballs were removed and placed in standard goldfish Ringer solution containing 0.2 mm Ca2+. Retinae were dissected free and incubated in hyaluronidase (4mg ml−1; Fluka) for 20 min at room temperature (21–23 °C). Retinae were then washed and cut into eight to sixteen pieces and incubated with papain (12–16 units ml−1; Worthington) for 30–60 min at room temperature in a Ringer solution containing 3 mm cysteine and no added Ca2+. The digestion times and papain concentrations varied depending on the lot number of the enzyme. The pieces of retina were washed several times with Ringer solution containing 0.5mgml−1 bovine serum albumin (BSA) and 0.2mm Ca2+ and stored at 4 °C. Bipolar cells were dissociated from the pieces of retina by trituration with fire-polished glass pipettes in the BSA-containing Ringer solution and plated out on coverslips coated with poly-l-lysine. Plated cells were kept in a moist environment and used within 2 h. Depolarizing (‘on’) bipolar cells were recognized by their morphology: flask-shaped cell body, thick main dendrites, long axon and large (10–12 μm diameter) synaptic terminal. On this basis, they are assumed to be rod-dominant ‘on’ bipolars classified as B1 by Ishida, Stell & Lightfoot (1980).

Fura-2 and FM1–43 measurements

Fluorescence measurements were made with a photomultiplier tube (Thorn EMI) attached to an inverted microscope (Zeiss Axiovert 10) using a Zeiss × 40 Fluar oil-objective lens (NA, 1.3) and a 75 W xenon lamp with light guide. Cells were loaded with fura-2 (Molecular Probes) by incubation in 1 μm of the AM ester for 10–40 min. Excitation wavelengths were alternated between 340 and 380 nm using niters of 10–12 nm bandwidth (Baling Electro-Optics) mounted in a filter wheel/shutter assembly (Ludl Electronic Products). Background fluorescence was minimized by stopping down the epifluorescence beam to a diameter of 20 μm, centred on the synaptic terminal. A pinhole in the image plane in front of the photomultiplier tube limited light collection to a 20 μm circle in the object plane. Fura-2 measurements were carried out at a rate of 5 ratios s−1.

Fura-2 signals were calibrated by permeabilizing cells with 5 μm ionomycin in the presence of different concentrations of Ca2+ and measuring the ratio (R) of the emission using the 340 and 380 nm beams. The free [Ca2+]i, was calculated according to the equation: [Ca2+]i, =K(RRmin)/(RmaxR). Rmin was measured in 0 Ca2+ and Rmax in 2.5 mm Ca2+. R was also measured in solutions containing 400 nm and 2000 nm [Ca2+], from which the constant K was estimated. These buffered Ca2+ solutions were made using EGTA by the ‘pHmetric’ method of Tsien & Pozzan (1989) and checked with a Ca2+-selective electrode (Orion) calibrated with Ca2+ standards (World Precision Instruments). The apparent dissociation constant of fura-2 measured in this way was 1000 nm.

The membrane marker FM1–43 (Molecular Probes) was excited at 490 nm (10 nm bandwidth), and emitted light was collected through a wideband filter (515–565 nm) using a 505DRLP dichroic mirror (Omega Optical). Simultaneous fura-2 measurements were made by alternating the excitation wavelengths between 340, 380 and 490 nm. Wavelengths shorter than 400 nm did not excite FM1–43, and wavelengths longer than 450 nm did not excite fura-2, allowing uncontaminated signals to be measured from each dye (Lagnado, Gomis & Job, 1996).

Solutions were applied to cells using four quartz pipes (100 μmi.d.; World Precision Instruments) glued into a square array and fed by gravity. Solutions were selected by means of solenoid valves (Lee Electric Valve Company). This system allowed us to make solution changes which were complete within a few seconds. Charybdotoxin and nifedipine were obtained from Sigma.


Cells were voltage clamped using the perforated patch technique (Korn, Marty, Connor & Horn, 1991). Recordings were made using an EPC-9 amplifier (HEKA) and analysed using Igor Pro (Wavemetrics). Electrode resistances were usually 2–4 MΩ. The solution in the patch pipette contained the following (mm): 110 potassium gluconate, 4 MgCl2, 3 Na2ATP, 1 Na2GTP, 0.4 BAPTA and 20 Hepes (pH 7.2); with 250 μg ml−1 nystatin or amphotericin. In some experiments, potassium gluconate was replaced by 110 mm caesium gluconate. Electrodes were usually applied to the synaptic terminal. The series resistance was typically 20–30 MΩ. Input resistances at −60 mV were usually 1–5 GΩ. Linear capacitative and leak currents were removed from records by subtracting the summed responses to four voltage steps (one-quarter the amplitude of the test step) applied from a holding potential of −80 mV. The automatic compensation features of the EPC-9 amplifier were used for both fast and slow capacitance compensation. The capacitance of cells varied from 10 to 16 pF. The series resistance was corrected up to 50% with a lag of 10 μs. Signals were filtered at 2.5 kHz and sampled at 5 kHz.

The membrane voltage (Vm) during our recordings will be given by Vm=Vp+VpfVj, where Vp, is the voltage applied to the patch pipette, Vj is the liquid junction potential and Vpf is a Donnan potential reflecting the impermeability of the ionophore to large anions (Barry & Lynch, 1991). Vpf was calculated from the difference in the activation of the Ca2+ current in perforated and conventional whole-cell recordings made with the same caesium gluconate pipette solution, and was found to be 11 mV. Vj was calculated from the composition of the internal and external solution to be 14.7 and 15.1 mV in Cs+ and K+ intracellular solutions, respectively, using Junction Potential Calculator software (Axon Instruments). However, experimentally measured values of Vj varied from 5 to 10 mV. Taking both the theoretical and experimental values of Vj into account, it seems likely that values of Vm were within a few millivolts of Vp, so no corrections were made. In this study we compared the threshold for activation of the Ca2+ and K+ currents with the threshold for generation of regenerative responses under identical recording conditions, so errors in Vm should apply equally to both estimates. The threshold for activation of L-type Ca2+ channels measured in this study was −43 mV, similar to values obtained by Tachibana et al. (1993).

Where appropriate, data are given as means ±s.e.m.


All-or-none voltage responses to current injection

Isolated ‘on’ bipolar cells exhibited regenerative depolarizations in response to current injections designed to mimic the effect of a step of light on a bipolar cell in the retina. The current-clamp recording in Fig. 1A shows that these all-or-none responses exhibited a sharp threshold. The current–voltage (I–V) relation measured in current clamp was linear up to about −47 mV (Fig. 1B), when the injection of just 1 pA more current triggered a regenerative depolarization about 15 mV in amplitude. This response was maintained throughout the 5 s of stimulation, and recovery occurred as soon as the depolarizing current was reduced by 1 pA. These sudden voltage jumps to a fixed plateau were termed ‘flips’. When the holding current was just beyond the threshold for a flip, the injection of further depolarizing current had little effect on the membrane potential (Fig. 1A and B), indicating that this response was associated with a large decrease in membrane resistance. The voltage threshold for initiation of a flip was −44 ± 1.5 mV (n= 11), and the plateau averaged −34 ± 1 mV.

Figure 1.

Regenerative depolarizations with a sharp threshold in ‘on’ bipolar cells

A, simultaneous measurements of synaptic [Ca2+]i (upper panel), membrane voltage (middle panel) and membrane current (lower panel). A 1 pA increase in the injected current was applied for 5 s at a series of holding currents. The third response was a flip to about −33 mV, which recovered as soon as the current was reduced by 1 pA. The arrow marks a spontaneous flip (see text). Note that synaptic [Ca2]i rose when a flip was initiated and fell on repolarization. B, the voltage response to 1 pA of injected current plotted as a function of the holding current for the cell in A. The threshold for a flip was about −46 mV. At more hyperpolarized potentials the input resistance was 3.3 GΩ. C, a similar experiment in a cell exhibiting transient voltage responses to current injection. The holding current was −16 pA, and responses are shown to 1 s steps to -11, -9, −7 and −1 pA. Just beyond threshold (thick trace), the response was spike-like with a prominent undershoot followed by a maintained component. Further depolarization did not generate a maintained response. D, the peak voltage is plotted as a function of the current for the cell in C. The input resistance was constant (2 GΩ) until −40 mV, when there was a sudden depolarization to about −25 mV and a large fall in resistance.

The large synaptic terminal of depolarizing bipolar cells from the goldfish retina allowed measurement of spatially averaged [Ca2+]i (Fig. 1A, top panel). The average value of resting [Ca2+]i was 60 ± 10 nm (40 cells), while [Ca2+]i during a flip could rise to levels up to 1–2 μm (average 400 ± 40 nm). Imaging of fura-2 fluorescence with a CCD camera demonstrated that these Ca2+ signals were localized to the synaptic terminal (not shown, but see Tachibana et al. 1993).

About 60% of cells generated all-or-none responses to current injection, but these responses were sometimes transient. Figure 1C and D shows an example in which the injection of 1 pA of current at the threshold of −42 mV generated a response similar to a flip, except that it only lasted 100 ms and the repolarization phase displayed a prominent undershoot before steadying (thick trace). An obvious feature of these regenerative responses was an increase in high-frequency noise at the plateau (Fig. 1A), associated with oscillatory voltage changes (Fig. 1C). These voltage oscillations are described in greater detail below. The remaining 40% of the cells showed no clear threshold for the regenerative voltage response, but reached a similar plateau which was also associated with oscillatory voltage changes (Fig. 9A and D).

Figure 9.

Ca2+-activated K+ channels were involved in generating high-frequency resonance

A, voltage responses to current injection. The holding current was −5 pA and responses are shown to steps to -1, 2, 10 and 25 pA. B, voltage responses to current injection after addition of 100 nm charybdotoxin (same cell as in A). Holding current was −5 pA and responses are shown to steps to -1, 4, 9 and 21 pA. Repetitive depolarizations at about 4 Hz could be elicited (thick trace). C, thick traces from A (control) and B (100 nm charybdotoxin) are compared on a faster time scale, showing the high-frequency resonance blocked by addition of charybdotoxin. D, the peak amplitude of the response in the first 200 ms of current injection plotted as a function of the injected current in control conditions (•) and after addition of 100 nm charybdotoxin (Δ). Charybdotoxin increased the amplitude of the response elicited by depolarization beyond −40 mV.

Spontaneous regenerative depolarizations

Voltage responses very similar to flips induced by direct current injection also occurred when the membrane potential spontaneously reached threshold, as shown in Fig. 2A. When the amplifier was switched from voltage clamp to current clamp the membrane potential repeatedly jumped to a fixed plateau of about −37 mV, and these depolarizations were correlated with rises in [Ca2+]i in the synaptic terminal up to 1μm. The plateau of a flip was associated with a voltage oscillation, as shown in Fig. 2C. Initially (a in Fig. 2), there was a damped oscillation with a peak-to-peak amplitude of about 10 mV and a frequency of 60 Hz. A second frequency component of about 10 Hz was also apparent. At later times (b in Fig. 2), the oscillations were less well defined and smaller in amplitude (note the different scale). The membrane potential of isolated cells was usually noisy (Fig. 2B), as was the holding current when a cell was held in voltage clamp (Fig. 2A). Given that a flip could be turned on and off by 1 pA changes in the holding current (Fig. 1), it seems likely that spontaneous flips were driven by this background conductance whenever it was large enough to depolarize the cell beyond threshold. It was not possible, therefore, to assign a definite resting membrane potential to isolated cells exhibiting spontaneous flips.

Figure 2.

Spontaneous rises in [Ca2+]i Were associated with membrane depolarization

A, simultaneous measurements of synaptic [Ca2+]i, (upper panel) and membrane current (middle panel) and membrane current (lower panel). Initially the cell was voltage clamped at a holding potential of −60 mV. Two depolarizations to −20 mV lasting 200 ms caused transient increases in [Ca2+]i. At the arrow the amplifier was switched to current clamp with a holding current of −7 pA. The membrane potential then ‘flipped’ repeatedly to a depolarized state associated with a rise in [Ca2+]i. B, the first four flips in A shown on an expanded time scale. Each rise in [Ca2+]i was preceded by depolarization and each fall in [Ca2+]i was preceded by hyperpolarization. C, the high-frequency noise at the plateau of the first flip in B is shown on an expanded time scale. Initially (a) membrane potential oscillations with an amplitude of 10 mV were observed at a frequency of 60 Hz. A second frequency component of about 10 Hz could also be observed. Later (the period marked b in B) the oscillations were less distinct and had a maximal amplitude of about 3 mV.

The voltage responses to current injection shown in Fig. 1 were similar in several respects to light responses of ‘on’ bipolar cells in situ. These do not exceed 20 mV in carp and are associated with a large fall in input resistance (Saito et al. 1979). Transient responses with an undershoot, similar to those in Fig. 1C, have been observed in carp (Saito et al. 1979; Saito & Kujiraoka, 1982) and dogfish (Ashmore & Falk, 1980), while maintained depolarizations with a damped oscillatory component of 30–40 Hz have been recorded in goldfish and carp (Kaneko & Hashimoto, 1969; Kaneko, 1970). Lower-frequency oscillations, at about 3–5 Hz, have been observed in dogfish (Ashmore & Falk, 1980). It therefore seems likely that the conductances shaping the voltage response of isolated bipolar cells also contribute to shaping the response of bipolar cells in situ. The resting membrane potential in ‘on’ bipolar cells in the dogfish retina averages −47 mV in the dark (Ashmore & Falk, 1980), which is close to the threshold for the activation of a regenerative response and Ca2+ influx that we measured in this study. It should be noted, however, that spontaneous depolarizations unrelated to inputs from photoreceptors have not been observed in bipolar cells in the retina. Below we describe some properties of the fluctuating conductance responsible for spontaneously depolarizing isolated bipolar cells beyond threshold.

Spontaneous Ca2+ rises drive vesicle cycling in the synaptic terminal

We tested whether spontaneous Ca2+ rises caused by flips could drive vesicle cycling by making simultaneous measurements of the fluorescence associated with fura-2 and the membrane marker FM1–43, as in Fig. 3. FM1-43 is an amphipathic dye that is much more fluorescent in lipid than water but cannot cross membranes (reviewed by Betz, Mao & Smith, 1996). In this cell, the spontaneous rises in [Ca2+]i were infrequent and long lasting. The initial large jump in fluorescence when 10 μm PM1-43 was added reflected the staining of the plasma membrane. There was no further increase in FM1–43 fluorescence while [Ca2+]i was below about 50 nm, but when [Ca2+]i spontaneously jumped to about 250 nm there was a continuous increase in fluorescence representing the staining of vesicle membrane having come into contact with the external medium after exocytosis. The rate of exocytosis during the rise in [Ca2+]i was estimated to be about 1% of the membrane surface area per second, which is equivalent to about 500 vesicles s−1 for a terminal 10 μm wide (see Lagnado et al. 1996). Exocytosis stopped when [Ca2+]i fell back below 50 mm. Similar observations were made in ten other cells, where spontaneous rises in [Ca2+]i drove exocytosis at rates equivalent to up to 2% of the terminal membrane surface area per second.

Figure 3.

Spontaneous rises in [Ca2+]i were associated with increased exocytosis

The thin trace (left-hand axis) shows [Ca2+]i in the synaptic terminal and the thick trace (right-hand axis) shows the fluorescence associated with the membrane marker FM1–43 in arbitrary units. Application of 10 μm FM1–43 is indicated by the filled bar. The initial rapid increase in fluorescence represents staining of the plasma membrane of the terminal, and the slower rise correlated with a spontaneous increase in [Ca2+]i represents an increase in membrane staining associated with exocytosis of synaptic vesicles. This increase in FM1–43 fluorescence only occurred during periods of raised [Ca2+]i. Terminal diameter, 10 μm.

Experiments using the caged Ca2+ compound NP-EGTA indicate that [Ca2+]i in the submicromolar range can drive a continuous cycle of exocytosis and endocytosis in the synaptic terminal of ‘on’ bipolar cells (Lagnado et al. 1996). It therefore seems likely that the prolonged exocytosis shown in Fig. 3 was stimulated by the rise in bulk cytoplasmic Ca2+ rather than ‘Ca2+ microdomains’ near open Ca2+ channels. The limited time resolution of the FM1–43 measurements in this study did not allow us to assess whether the rate of exocytosis was also modulated by high-frequency voltage oscillations, although capacitance measurements indicate that there is a rapid and transient component of exocytosis in the bipolar cell terminal (Mennerick & Matthews, 1996), which might be expected to respond to rapid changes in membrane potential (see Discussion).

A non-specific cation conductance

Some of the properties of the background current responsible for the triggering of spontaneous flips are shown in Fig. 4. Figure 4A shows the current noise in Fig. 2A in greater detail. Increases in the inward current at the holding potential of −60 mV were correlated with small rises in [Ca2+]i in the synaptic terminal (arrows), suggesting that this conductance had some permeability to Ca2+. Consistent with this, it could be reversibly blocked by 10μm La3+, as shown in Fig. 4B. This conductance was isolated by blocking K+ channels using a Cs+-filled pipette and blocking L-type Ca2+ channels by adding 200 μm nifedipine. To obtain the I- V relation, a series of voltage steps were applied when the inward current was relatively steady, and then when it was blocked with 10 μm La3+. The La3+-sensitive current was not dependent on time (Fig. 4C) or voltage (Fig. 4D), and it reversed at potentials between 0 and +10 mV (4 cells). These properties indicate that the background current was carried by a non-specific cation conductance.

Figure 4.

Some properties of a spontaneously varying inward current

A, simultaneous measurements of synaptic [Ca2+]i (upper panel) and membrane current (lower panel) at a holding potential of −60 mV. The traces are from the recording shown in Fig. 2A. The arrows mark three instances where a rapid increase in the inward current was correlated with a small rise in [Ca2+]i. B, the inward current was reversibly blocked by 10 μm La3+. The dotted line represents a period of 1 min when the current was stable. C, the response of the inward current to a series of voltage steps from −60 mV to +50 mV. The recording was made with a Cs+-filled patch pipette (to block K+ currents) in the presence of 200 μm nifedipine (to block the Ca2+ current). The current traces were obtained by subtracting responses obtained in 10 μm La3+ from those in its absence. D, the I-V relation of the La3+-sensitive current was linear and reversed at +10 mV. Results from C.

‘On’ bipolar cells have a non-specific cation conductance opened by cytoplasmic cGMP which can be suppressed by glutamate released by photoreceptors in the dark (Nawy & Jahr, 1990). However, the conductance that we observed was not inhibited by 200 μm glutamate, nor increased by adding 100 μm 8-bromo-cGMP to the perfusate. Spontaneous depolarizations do not occur in ‘on’ bipolar cells in the retina, so variations in the size of this inward current are probably an abnormal consequence of isolation. To understand the function of this conductance it will be necessary to understand how it is normally controlled. In the experiments we report here, its effect was to trigger regenerative depolarizations in isolated bipolar cells by bringing the membrane potential to threshold.

Ca2+ and K+ conductances activate at threshold

Electrical resonance qualitatively similar to that shown in Figs 1 and 2 has been extensively studied in hair cells of the cochlea (Crawford & Fettiplace, 1981; Art & Fettiplace, 1987) and sacculus (Hudspeth & Lewis, 1988), which are electrically tuned to particular frequencies. The resonance in hair cells is thought to involve an interaction between voltage-activated Ca2+ channels of the L-type and large-conductance, Ca2+-activated K+ channels. Results presented below indicate that a similar mechanism underlies the electrical resonance in ‘on’ bipolar cells.

In voltage-clamp recordings with pipette solutions containing 110 mm K+, depolarization beyond - 40 mV generated a maintained net outward current (Fig. 5A) that exhibited outward rectification (• in Fig. 5E). On an expanded time scale, this outward current was found to be preceded by an inward current that activated within milliseconds of depolarization to −40 mV (Fig. 5C). Depolarizing to −30 mV increased the speed with which the outward current developed (Fig. 5D). When external Ca2+ was replaced by 2.5 mm Ba2+, the outward current was blocked, revealing a maintained inward current (Fig. 5B) that peaked at about −10 mV (▪ in Fig. 5E). These results support observations made by Kaneko & Tachibana (1985), who identified the inward current as being carried by Ca2+ channels permeable to Ba2+, and the outward current by K+ channels blocked by Ba2+. Consistent with this identification, the outward current reversed at about −80 mV, close to the K+ equilibrium potential, and was also blocked by 25 mm external TEA (not shown).

Figure 5.

Membrane currents around the voltage threshold for regenerative responses

A, currents elicited by a series of depolarizations from −60 to +50 mV in 10 mV steps. The patch pipette contained 110 mm K+ and the external solution contained 2.5 mm Ca2+. The I–V relation is plotted as the circles in E. B, currents in the same cell after replacing external Ca2+ with 2.5 mm Ba2+. Note the different current scale. The I–V relation is plotted as the squares in E. C and D, currents recorded immediately after depolarization to −40 mV (C) and −30 mV (D). The thick traces are from A (2.5 mm Ca2+) and the thin traces from B (2.5 mm Ba2+). Note the rapid inward current preceding the outward current blocked by Ba2+. F, I–V relation of the Ca2+ conductance recorded with 110 mm Cs+ in the pipette to block K+ conductances and with 2.5 mm Ba2+ as the charge carrier. The voltage was varied as a ramp from −60 to +50 mV over 1s. The current began to activate at −43 mV.

The inward current was sensitive to dihydropyridines (Fig. 6A), confirming that it is carried by L-type Ca2+ channels (Tachibana et al. 1993). The I–V relation was recorded by blocking K+ conductances by replacing K+ the pipette with 110 mm Cs+, as shown in Fig. 5F. The threshold for activation of L-type Ca2+ channels was −43 ± 3 mV (n= 8), defined by measuring the voltage at which the inward current was more than two standard deviations greater than the baseline at hyperpolarized potentials. This value was not significantly different from the threshold for initiation of a regenerative response (Fig. 1).

Figure 6.

Blocking L-type Ca2+ channels with nifedipine blocked regenerative depolarizations and associated rises in [Ca2+]i

A, Ca2+ currents recorded in 20 μm nifedipine (thick trace) and control conditions before and after addition of nifedipine (thin traces); 110 mm Cs+ in the patch pipette and 2.5 mm external Ca2+. B, simultaneous voltage and [Ca2+]i, measurements. Holding current of −10 pA. Nifedipine at 30 μm reversibly blocked spontaneous voltage flips (lower panel) and rises in synaptic [Ca2+]i, (upper panel). Note that the voltage noise was still present in 30 μm nifedipine (see text). C, expansion of recording in lower panel of B, comparing voltage flips in control solution with the inhibitory effects of nifedipine. The dotted lines show the threshold and plateau voltages of the flips and emphasize that in the presence of nifedipine the voltage noise varied continuously, without any stereotyped transitions between these voltage levels.

Nifedipine reversibly blocked spontaneous voltage flips and associated rises in [Ca2+]i in the synaptic terminal, as shown in Fig. 6B. Similar results were observed in ten cells, using 10–30 μm nifedipine. The membrane voltage during periods with and without 30 μm nifedipine is shown in greater detail in Fig. 6c. After blocking L-type Ca2+ channels, regenerative depolarizations and maintained flips were not supported, even though the membrane potential was noisy and regularly crossed above threshold (dotted line in Fig. 6C). These random fluctuations are probably due to the nonspecific cation conductance, which was not blocked by nifedipine. The small rises in [Ca2+]i observed during the application of nifedipine probably reflect Ca2+ influx through this non-specific cation conductance (Fig. 4A), although some Ca2+ influx through incompletely blocked Ca2+ channels might also have occurred. The results in Figs 5 and 6 indicate that the initiation of a regenerative response required depolarization beyond the threshold for activation of L-type Ca2+ channels, which also provided the major route for Ca2+ entry into the synaptic terminal.

K+ currents limit the amplitude of regenerative depolarizations

Charybdotoxin, which blocks large-conductance, Ca2+-activated K+ channels in a variety of preparations (Dreyer, 1990), blocked most of the outward K+ current at a concentration of 100 nm (Fig. 7A and B). The remaining voltage-dependent current was also carried by K+, since it reversed at −80 mV (not shown), and could be blocked with 2.5 mm Ba2+ (Fig. 5). The transient nature of this current identifies it as an A-type K+ current (Connor & Stevens, 1971).

Figure 7.

Separation of Ca2+- and voltage-activated K+ currents with charybdotoxin

A, currents elicited by a series of depolarizations from a holding potential of −60 mV to +50 mV in 10 mV steps. The steps. The steady state I–V relation is plotted as the open circles in D. B, currents recorded from the same cell after addition of 100 nm charybdotoxin (CTX). Depolarization activated a transient outward current. The peak I–V relation is plotted as the open triangles in D. C, the charybdotoxin-sensitive current was separated by subtracting the currents in B from those in A. The I–V relation is plotted as the filled circles in D.

The I–V relation for the Ca2+-activated K+ conductance could be obtained by subtracting records obtained in the presence of 100 nm. charybdotoxin from those obtained in its absence (Fig. 7C). As in Fig. 5A and C, activation of this current was associated with a noticeable increase in noise, supporting the idea that it is carried by Ca2+-activated K+ channels of large conductance. Both the Ca2+-activated K+ conductance and voltage-dependent K+ conductance began to activate between −40 and −30 mV (Fig.7D).

The role of the combined K+ conductances in shaping the voltage response of ‘on’ bipolar cells was tested by blocking them in one of two ways: replacing internal K+ with Cs+ through the patch pipette (4 cells), or by replacing external Ca2+ with 2.5 mm Ba2+ (3 cells). Both manoeuvres caused the potential achieved by regenerative depolarization to exceed 10 mV, and Fig. 8 shows the effects of Ba2+. It therefore seems likely that the activation of these K+ conductances at potentials between −40 and −30 mV causes the large decrease in membrane resistance that occurs over this voltage range (Fig. 1B), which, together with the hyper-polarizing effect of the outward K+ current, limits the amplitude of the regenerative response.

Figure 8.

K+ conductances limited the degree of depolarization

Blocking K+ conductances by replacing external Ca2+ with 2.5 mm Ba2+ (bar) reversibly increased the amplitude of spontaneous voltage flips.

Large-conductance, Ca2+-activated K+ channels contribute to high-frequency resonance

The role of the Ca2+-activated K+ conductance was investigated by blocking it with 100 nm charybdotoxin. Figure 9A shows control responses to current injection from a cell exhibiting both low (5 Hz)- and high (70 Hz)-frequency resonance, and Fig. 9B shows responses from the same cell in the presence of 100 nm charybdotoxin. Blocking the Ca2+-activated K+ conductance blocked the high-frequency component of the response but increased the amplitude of the low-frequency component, which was manifest as regenerative ‘spikes’ with obvious undershoots (Fig. 9B and C). In charybdotoxin, the threshold for the generation of the initial spike was about −40 mV, and its amplitude was considerably greater than in control (Δ in Fig. 9D). Similar behaviour was observed in five cells tested in this way. These results indicate that the Ca2+-activated K+ conductance contributes to the high-frequency resonance, as well as setting the degree of depolarization that can be achieved. The voltage-dependent K+ conductance appears to contribute to the lower-frequency component of the voltage response, as well as the undershoot observed in transient responses.


Responses of bipolar cells in situ

The synaptic response of ‘on’ bipolar cells is generated by glutamate released from photoreceptors. In the dark, glutamate acts through a metabotropic receptor to activate a cGMP phosphodiesterase and close non-specific cation channels gated by cGMP in the cytoplasm (Nawy & Jahr, 1990). Light, by reducing glutamate release from photo-receptors, leads to the opening of these channels and depolarization. The aim of this study has been to investigate how the membrane conductances intrinsic to the bipolar cell act to shape the resulting voltage response and control Ca2+ influx into the synaptic terminal.

Two basic findings are that the L-type Ca2+ conductance can cause the response to become regenerative once the cell is depolarized beyond about – 43 mV, while Ca2+- and voltage-activated K+ conductances that begin to activate at slightly higher voltages tend to limit the depolarization to below – 30 mV. These observations may explain why light responses in ‘on’ bipolar cells rarely exceed 20 mV in amplitude. The triggering of regenerative responses may cause the ‘on’ bipolar cell to act something like a threshold detector. Consistent with this idea, rods signal over about three log units of intensity, but the response of the rod-dominant ‘on’ bipolar cells (as used in this study) only covers about one log unit of intensity (Saito et al. 1979; Saito & Kujiraoka, 1982). Regenerative depolarizations with a threshold, similar in waveform to the responses to current injection shown in Fig. 1C, have also been observed in depolarizing bipolar cells of the dogfish, where the response to dim steps of light is square and scales linearly with intensity for dim lights, until, above a threshold voltage, transient ‘spikes’ become apparent at light onset (Ashmore & Falk, 1980). We have shown that these transient responses occur over the voltage range that Ca2+ channels in the synaptic terminal begin to activate, and so it seems possible that they will cause a phasic increase in the rate of transmitter release (discussed below). Such a mechanism may contribute to the transient component of the response to a step of light seen in many ‘on’ ganglion cells (Hubel, 1995).

A third feature of the bipolar cell response was voltage oscillations, with both high (60 Hz)- and low (5 Hz)-frequency components. High-frequency oscillations (30–40 Hz) have been recorded in carp and goldfish (Kaneko & Hashimoto, 1969; Kaneko, 1970), while low-frequency oscillations (3–5 Hz) have been recorded in dogfish (Ashmore & Falk, 1980). Our results indicate that voltage-dependent K+ channels are important in generating low-frequency resonance, while Ca2+-activated K+ channels are involved in generating the high-frequency resonance. Hair cells in the ear also respond to current injection with voltage oscillations (Crawford & Fettiplace, 1981). These responses can be accounted for by a model in which depolarization causes the regenerative activation of an L-type Ca2+ current, and the resulting rise in [Ca2+] near the membrane causes the delayed activation of an outward K+ current, leading to hyperpolarization and reduced Ca2+ influx (Hudspeth & Lewis, 1988). Given that the same types of conductances are involved, it seems likely that the resonance in bipolar cells and hair cells are generated by similar mechanisms.

Hair cells of the cochlea are tuned to different frequencies over the audible range, so the electrical resonance has a clear function - the spectral decomposition of a sound into its constituent frequencies (Art & Fettiplace, 1987). The role of resonating responses in bipolar cells is less clear, although Doty & Kimura (1963) suggested that they may be the source of oscillatory changes in the frequency of action potentials generated by ganglion cells. Such oscillations in the retinal output have been observed in eels (Adrian & Matthews, 1928), frogs (Granit, 1933), monkeys (Doty & Kimura, 1963) and cats (Doty & Kimura, 1963; Bishop, Levick & Williams, 1964; Neuenschwander & Singer, 1996). High-frequency oscillations can also be observed in neurons of the lateral geniculate nucleus that receive their input from ganglion cells (Ariel, Daw & Rader, 1983; Neuenschwander & Singer, 1996), as well as neurons of the visual cortex (Konig, Engel & Singer, 1995). However, it should be noted that oscillations are not a common feature of responses observed in ganglion cells, so it is not clear what role they might have in the representation of visual information (Neuenschwander & Singer, 1996).

It may be that frequency tuning of responses in bipolar cells simply acts to improve time resolution of signals transmitted to ganglion cells by selecting higher-frequency components in the signal generated by photoreceptors, and that obviously oscillating responses only occur when damping of the bipolar cells is reduced. Indeed, Kaneko & Hashimoto (1969) found that responses recorded in a given bipolar cell only gradually became oscillatory. Changes in the properties of the Ca2+ and K+ currents, or simply changes in the membrane resistance, are all mechanisms which might modify the response of a bipolar cell in the retina. Such variations may also account for the different responses to current injection that we observed in isolated cells (Fig. 1). Preliminary results (J. Burrone & L. Lagnado, unpublished observations) indicate that high-frequency oscillations in isolated ‘on’ bipolar cells can be blocked by Substance P, a peptide which is released by amacrine cells and inhibits the L-type Ca2+ current in ‘on’ bipolar cells. The non-specific cation conductance that we observed, if it were under the control of an external transmitter, might also act to modulate responses, perhaps by setting the resting membrane potential relative to the threshold for the opening of Ca2+ channels. Another potential modulator is GABA, which amacrine cells release onto the synaptic terminal of bipolar cells in response to light. GABA can activate a Cl conductance, as well as inhibiting the L-type Ca2+ current through a G-protein (Matthews, Ayoub & Heidelberger, 1994). The decrease in input resistance caused by activation of the Cl conductance would, if large enough, be expected to shunt the electrical resonance.

The time course of neurotransmitter release

The output of the bipolar cell is the release of glutamate from small vesicles (Tachibana & Okada, 1991), so to understand the input-output relation of this neuron it will be necessary to understand the relation between the voltage signal and exocytosis. We have shown that the membrane behaves passively until L-type Ca2+ channels are activated, when regenerative responses stimulate prolonged Ca2+ influx into the synaptic terminal that result in continuous exocytosis. Voltage responses in depolarizing bipolar cells have transient and maintained components (Ashmore & Falk, 1980; Saito & Kujiraoka, 1982), and it appears that neurotransmitter release from the synaptic terminal also has at least two components. Tachibana & Okada (1991) monitored glutamate release by pushing the terminal of a goldfish bipolar cell against the membrane of a voltage-clamped horizontal cell in which the current generated by NMDA receptor channels was recorded. These experiments indicated that there was a rapid component of secretion that could be triggered by calcium tail currents lasting less than 1 ms, but that high rates of release could be maintained for up to 500 ms. Depolarizations lasting 2 s indicated that there was also a maintained component of release that continued for some seconds after closure of Ca2+ channels (see Fig. 7 of Tachibana & Okada, 1991). Lagnado et al. (1996) combined FM1–43 staining of synaptic vesicles with measurements of cytoplasmic Ca2+ using fura-2 and found that the slow, continuous, component of exocytosis required submicromolar levels of [Ca2+]i; indicating that it could be driven by Ca2+ rises in the bulk cytoplasm. Heidelberger, Heinemann, Neher & Matthews (1994) have shown that the rapid component of exocytosis requires Ca2+ levels of the order of 100 μm, which can only occur in the ‘Ca2+ microdomains’ existing close to open Ca2+ channels.

Responses of ‘on’ bipolar cells rarely exceed 10–15 mV in amplitude, so light will only modulate Ca2+ influx through L-type channels over a small voltage range at the foot of the activation curve. A similar situation arises in photoreceptors, which hyperpolarize from a dark potential of about −40 mV in response to light, and so utilize only a small range of the L-type Ca2+ conductance (Attwell, Borges, Wu & Wilson, 1987). The resonating responses that we have observed in ‘on’ bipolar cells are similar in amplitude to those observed in hair cells (5–10 mV). It seems likely, therefore, that the ribbon synapses present in these sensory neurons are specialized to respond to small changes in membrane potential. Transmitter release from hair cells can be modulated at high frequencies, since action potentials in fibres of the auditory nerve are phase-locked to tones up to 2 kHz, but it is not known if exocytosis from the bipolar cell will follow signals at their own resonant frequency of 60 Hz. One of the factors that allows transmitter release from hair cells to be modulated by high-frequency changes in membrane potential is the rapid opening and closing kinetics of the L-type Ca2+ channels. These Ca2+ channels co-localize with large-conductance, Ca2+-activated K+ channels at synaptic release sites (Roberts, Jacobs & Hudspeth, 1990), where electron micrographs show small vesicles tethered to an electron-dense ball by the plasma membrane. Bipolar cells have similar dense bodies associated with vesicles (Lagnado et al. 1996), but it is not known if Ca2+ and/or K+ channels are also localized at these sites. To understand how the voltage response of the bipolar cell controls transmitter release it will be necessary to characterize the opening and closing kinetics of the Ca2+ conductance and measure the time course of exocytosis in response to small voltage changes in the range where these channels begin to activate.


This work was supported by the Human Frontiers in Science Program and an MRC Studentship to J.B.