Introduction

Thirty years of comprehensive research has given a deep understanding of the light—activated enzymatic cascade which leads to the closure of the outer segment cGMP-gated channels of vertebrate photoreceptors, producing photoreceptor hyperpolarisation (reviewed, for example, in: McNaughton 1990; Rispoli 1998; Moriondo and Rispoli 2003). The time course of the light-induced changes of the current entering the outer segment does not drive the inner segment voltage changes with the same time course, which ultimately modulates the rate of transmitter released at the synaptic terminal. The difference between the two time courses is produced by at least five kinds of voltage-sensitive channels located in the inner segment (reviewed in Barnes 1994): a nonselective cation current (I h), a voltage-gated Ca2+ current (I Ca), two Ca-activated currents permeable to Cl− (I ClCa) and to K+ (I KCa), and a voltage-gated, Ca-insensitive K+ current (I Kx; Attwell and Wilson 1980; Bader et al. 1982; Barnes and Hille 1989; Beech and Barnes 1989; Cia et al. 2005; Corey et al. 1984; Han et al. 2000; Hestrin 1987). Although these conductances have been extensively studied in both rods and cones, less attention has been devoted on how they shape the generator potential in response to light decrements. This is however an obvious key feature of the visual system in an enormous number of situations, from the swift detection of a predator’s shadow to reading black letters on a white sheet of paper (Hurley 2002). We have previously described how I KCa is able to clamp the photovoltage to its dark value (Moriondo et al. 2001). However, it is not clear yet how I KCa and/or other voltage- and time- dependent cationic conductances affect the photoresponse kinetics and, in particular, the kinetics of the recovery of the generator potential to its dark level, once the light stimulus is terminated. This issue has been addressed in the following work by using rods mechanically isolated from Ambystoma mexicanum and recorded under whole-cell patch-clamp conditions.

Materials and methods

Animals and dissection

Ambystoma mexicanum larvae (20–25 cm length) were kept in deionised water supplemented with 1.55 g of marine salt per litre, in small tanks at room temperature (20–23°C), fed two-three times a week with earthworms and maintained in a natural light-dark cycle. Water change was performed once per week. The relative immobility of water and the constant ionic composition helped to preserve animals’ health. Adequate care was taken to minimise pain and discomfort to animals, procedures and protocols were approved by the Institutional Animal Care Committee at the University of Ferrara, Ferrara, Italy, in accordance with the Health Research Extension Act.

The dissociation procedure has been slightly modified from the one already published (Moriondo et al. 2001). Briefly, light-adapted and dark-adapted animals were anaesthetized with tricaine methanesulphonate 1 g/l, and subsequently decapitated and pithed. Both eyeballs were rapidly enucleated; the retina was dissected in standard external solution plus 1.5% foetal bovine serum (heat inactivated). The retina was triturated in 500 μl of standard external solution plus 1.5% foetal bovine serum with a P1000 pipette to obtain both intact photoreceptors and isolated rod outer segments (OS); the serum was found to improve the cell quality, leading to more stable whole-cell recordings. To record the photoresponses from OS, all the above manipulations were performed in the dark, using infrared illumination and an infrared viewer (Find-R-Scope, FJW Optical Systems, Palatine, IL, USA), on dark adapted animals (for ~4 h).

Fifty microliter of cell suspension were plated onto 35 mm Petri dishes and let settle on ice for ≈30 min before filling the dish with 1.5 ml of standard extracellular solution and starting the experiment; OS were instead used immediately after the retina trituration.

Patch-clamp recordings

The patch-clamp set-up was made of an inverted microscope (IMT-2, Olympus) fitted with an Eppendorf PatchMan micromanipulator (Eppendorf, Hamburg, Germany). Rapid changes of the external solution were performed by moving by hand a multibarreled perfusion pipette in front of the recorded cell. The external solutions flowing in the perfusion pipette were gravity-fed (flow rate of 0.5–0.8 ml/min). Rod outer segments were illuminated with an ultrabright infrared LED (900 nm) and viewed on a TV monitor connected to a contrast enhancement camera (Till Photonics, Planegg, Germany). Light stimuli were delivered on dark adapted OS using pClamp voltage protocols: one of the two analogical outputs of the Digidata 1322A was connected to a calibrated voltage-to-current converter driving an ultrabright LED (555 nm) coupled to the binocular port of the microscope. The light emerging from LED passed through a neutral density filter (logarithmic attenuation factor: 2.0), a diffusion filter, a dicroic mirror (550 nm) housed in the exciter filter insertion slot and finally light was focused on the OS through a 60x objective. The LED was positioned so that the spot of light was focused and centered on the cell when the cell was viewed in sharp focus on the TV monitor. This set-up was enclosed in a Faraday’s cage made with a copper’s mesh, that was covered with a thick fully opaque cloth to carry on the experiments in the dark on the photoresponses from OS. The experiments on intact photoreceptors were instead carried out under room lights.

Electrical recordings were carried out at room temperature (20–22°C), using the patch-clamp recording technique in the “whole-cell” configuration under visual control, employing an L/M-EPC7 patch-clamp amplifier (HEKA, Lambrecht, Germany). Current was filtered at 1.3 kHz by a low-pass Bessel filter (mod. LPBF-48DG, npi electronic GmbH, Tamm, Germany) and sampled at 10 KHz by a Digidata 1322A interface, connected to an AMD K6-2-400 mHz personal computer running Clampex 8.1 software (Axon Instruments, Foster City, CA, USA). Data were analysed offline by Clampfit 8.1; access resistance was compensated at 70%.

Solutions

Standard intracellular solution for recordings from intact cells contained (in mM): NaCl, 4; KCl, 105; EGTA free acid (ethylene glycol-bis-(β-aminoethyl ether)N,N,N′,N′,-tetraacetic acid), 5; MgCl2, 2.5; HEPES free acid (N-[2-hydroxyethyl]piperazine-N′-[2ethanesulfonic acid]), 10; the intracellular solution for the OS recordings had the same composition but contained KCl, 95, adenosine 5′-triphosphate, dipotassium salt, 5, and guanosine 5′-triphosphate, sodium salt, 1. The pH and the osmolality of both solutions were adjusted to 7.2 with KOH and to 238 mOsm/Kg with D-glucose (usually 5 mM final concentration), respectively.

Standard extracellular solution contained: NaCl, 110; KCl, 2.5; CaCl2, 1; MgCl2, 1.6; HEPES free acid, 10; the pH and the osmolality were adjusted to 7.6 with NaOH and to 242 mOsm/Kg with D-glucose (usually 10 mM final concentration), respectively. Ba2+ and Cs+ external solutions had the same formulation of the standard external solution, but 1 mM BaCl2 substituted for CaCl2 in the former, while the latter contained 100 mM NaCl and 10 mM CsCl. Extracellular solution for I h recording contained: NaCl, 85; KCl, 2.5; TEACl, 30; CaCl2, 1; MgCl2, 1.6; CdCl2, 0.2; HEPES, 10; pH 7.6, 242 mOsm/Kg.

Charybdotoxin (ChTx) was dissolved in standard external solution plus 0.01% bovine serum albumin to a final concentration of 100 nM. Tetrodotoxin was diluted to a final concentration of 2 μM from a 1 mM stock solution kept frozen. When used, CdCl2 was added to the external solution to a final concentration of 200 μM from a 100 mM stock solution.

All chemicals were purchased from Sigma Aldrich (St. Louis, MO, USA) but ChTx and tetrodotoxin, purchased from Alomone Labs (Jerusalem, Israel), and foetal bovine serum and bovine serum albumin, purchased from GIBCO Life Technologies (Gaithersburg, MD, USA).

Stimulation with real flash responses

Responses to 5 ms flashes of light of increasing, sub-saturating intensity were recorded from isolated, dark-adapted OS recorded with the whole-cell voltage-clamp technique. To compensate for the different dark current amplitude (i.e. the current suppressed by a saturating step of light delivered just at the end of the response family) recorded from the outer segments, all flash responses were normalised to the average dark current (50 pA). The response family was then inverted and fed to Clampex as a stimulus protocol file.

Results are given as mean ± sem where appropriate.

Theory

Pure passive responses have been computed as follows:

$$ \left\{{\begin{array}{ll} V(t)=V_{h} & \quad \hbox{for}\,0 \le t < t_{\text{s}}\\ V(t)=V_{\text{h}}+R_{\text{m}} I_{\text{stim}} \left({1-e^{{\frac{{-t+t_{\text{s}}}}{{R_{\text{m}}C_{\text{m}}}}}}}\right) &\quad \hbox{for}\,t_{\text{s}} \le t < t_{\text{e}}\\ V(t)=V_{\text{h}}+R_{\text{m}} I_{\text{stim}} e^{{\frac{{-t+t_{\text{e}}}}{{R_{\text{m}}C_{\text{m}}}}}} &\quad \hbox{for}\,t \ge t_{\text{e}}\\ \end{array}}\right. $$

where:

t :

Time (ms)

t s :

Time to turn on the voltage pulse (ms)

t e :

Time to turn off the voltage pulse (t e − t s  = 500 ms)

V h :

Holding potential (mV)

R m :

Rod membrane resistance (GΩ)

C m :

Rod membrane capacitance (pF)

I stim :

Stimulus current (pA).

Results

Voltage responses to current injections at V h = −40 mV and V h = −60 mV

To assess the role of the inner segment cationic conductances in shaping the recovery kinetics of the photoresponse, current-clamped rods were depolarised to various potentials using current injection steps of size similar to the light-regulated current (5–70 pA, lasting 500 ms; Fig. 1, inset). The steps were applied on top of a background current, whose amplitude was adjusted to have a steady membrane potential (V h) of −40 mV (simulating the dark conditions) or −60 mV (simulating the steady-state response to a background of light). This strategy was therefore aimed at simulating also the photovoltage changes elicited by the reopening of the cGMP-gated channels following progressive reductions of intensity of a background of light. However, R m (typically >1 GΩ) was higher compared to that of dark-adapted rods (tipically ~370 MΩ), because the outer segment cGMP-gated channels are all closed in the present experiments, that were carried out under room lights.

Fig. 1
figure 1

Current-clamp voltage responses of a light-adapted rod starting from two different holding potentials (V h). a Responses to a series of increasing depolarising currents in 5 pA increments lasting 500 ms (inset) starting from an holding current (−9.3 pA in this cell) giving V h  = −60 mV. b Same cell and stimulus protocol of a, but V h  = −40 mV (holding current: 35.8 pA). c Voltage responses to a current injection of 25 pA (upper panel) and 70 pA (lower panel) superimposed to the pure passive responses computed from the measured R m (1.2 GΩ) and C m (26.1 pF; see Methods); V h  = −60 mV. d Same cell, protocol and simulations as c, but the current injections were 5 pA (upper panel) and 25 pA (lower panel); V h  = −40 mV

Full size image

At V h = −60 mV, positive current injections generated an initial depolarisation which declined, often with damped oscillations, to a plateau level. At the stimulus offset, the voltage transiently overshot the pre-stimulus level of −60 mV. By increasing the size of the injected current, the size of the voltage decline became smaller and its kinetics slower, but the peak amplitude of the overshoot at the stimulus offset was progressively enlarged. Moreover, the largest current stimuli (>40 pA) did not generate time-dependent voltage changes: the potential remained instead substantially stable to a saturating level of −15.7 ± 2.7 mV (maximum amplitude for a 70 pA current injection, n = 6; Fig. 1a). If the holding current was adjusted to give V h  = −40 mV (corresponding to dark conditions), the voltage responses were dramatically different in respect to the ones obtained at V h  = −60 mV. Indeed, neither the voltage decline nor the overshoot were observed but the responses were just monotonic. However, as the injected current was made larger, the “voltage saturation effect” of Fig. 1a was still present (average amplitude for a 70 pA current injection: −16.2 ± 1.1 mV, n = 7; Fig. 1b). The above results show that some voltage- and/or time- dependent conductances are recruited during the current injections, and this would therefore strongly affect the recovery phase of the light response. The effectiveness of these conductances is shown in Fig. 1c, d, where the pure passive responses, computed from the measured R m and C m (see Methods), are compared to the actual voltage responses for two current injections delivered at V h  = −60 mV (Fig. 1c; 25 pA, upper panel; 70 pA, lower panel) and at V h  = −40 mV (Fig. 1d, 5 pA, upper panel; 25 pA, lower panel). It is clear that only the 5 pA current injection at V h  = −40 mV can be (barely) fitted. Therefore, at least one outward delayed current must be activated during the stimulus, resulting in voltage response compression. Voltage-clamp experiments indicated that this was the case: a delayed current was indeed progressively activated following a depolarisation from V h  = −60 mV (Fig. 2a); this current was therefore already active at V h  = −40 mV (Fig. 2b).

Fig. 2
figure 2

Voltage-clamp current responses of a rod starting from two different holding potentials. a Responses to a series of 500 ms voltage pulses from −80 to 70 mV in 10 mV increments starting from V h  = −60 mV; the responses to voltage pulses from −80 to −30 mV are enlarged in the inset. b Same as a, but V h = −40 mV. Dotted line indicate zero current level

Full size image

For both V h, another noisy outward component appeared to be abruptly activated for depolarisations more positive than −30 mV. Moreover, when hyperpolarising from −60 mV, a progressively activating inward current was recorded, which should necessarily comprise the now deactivating above mentioned current already active at −40 mV. It is clear that it is necessary to pharmacologically dissect the contribution of the diverse cationic components in order to understand the ionic bases of the above voltage responses.

IKCa, ICa, and Ih contribution to voltage responses

A possible candidate acting as an outward delayed current is I KCa (Moriondo et al. 2001): therefore, the experiments of Fig. 1 were repeated during extracellular perfusion with 200 μM Cd2+ (which blocks I Ca, and, indirectly, I KCa) or 100 nM ChTx (a scorpion venom known to selectively block I KCa). The voltage decline during the current injection and the overshoot at the stimulus offset observed at V h  = −60 mV were unaffected by Cd2+ application, indicating that neither I KCa nor I Ca produced these features in the voltage responses (Fig. 3).

Fig. 3
figure 3

Effect of I Ca and I KCa blockade on the current-clamp voltage responses at V h  = −60 mV. a Control recording. b Application of 200 μM Cd2+. c Current-to-voltage plots (I–V), computed by averaging the points just following the current injection onset (where dV(t)/dt ≈ 0; empty symbols) and by averaging the last 20 ms of the response (where again dV(t)/dt ≈ 0; filled symbols) in control (circles) and after the application of 200 μM Cd2+ (squares)

Full size image

Moreover, the voltage saturation effect shown in Fig. 1 was lost in the presence of Cd2+ or ChTx, irrespective of the initial V h (Figs. 3, 4), showing that this effect was produced by I KCa activation. Given the time-dependent waveform of the voltage response in current-clamp experiments, the current-to-voltage plots (I–V) were computed by averaging the points at about the voltage peak just following the current injection onset (where dV(t)/dt ≈ 0; Figs. 3c, 4d, e, empty symbols) and by averaging the last 20 ms of the response (where again dV(t)/dt ≈ 0; Figs. 3c, 4d, e, filled symbols). Clearly, the voltage increased upon increasing the size of the injected current in the presence of Cd2+ (Figs. 3c, 4d, squares) or ChTx (Fig. 4e, squares) instead of being clamped at about −16 mV as in control (70 pA current stimulus; Figs. 3c, 4d, e, circles; Figs. 3a, 4a). In detail, a 70 pA injection gave a maximal depolarisation of −0.8 ± 2.0 mV (Fig. 3b, c, n = 6) at V h  = −60 mV in the presence of Cd2+ and 31.4 ± 5.5 mV (in Cd2+; Fig. 4b, d, n = 7) and 24.5 ± 7.1 mV (in ChTx; Fig. 4c, e, n = 7) at V h  = −40 mV. This shows again the effectiveness of I KCa at clamping the photovoltage to values close to the dark potential (Moriondo et al. 2001).

Fig. 4
figure 4

Effect of selective block of I Ca and I KCa on the current-clamp voltage responses at V h  = −40 mV. a Control recording. b Application of 200 μM Cd2+. c Application of 100 nM ChTx. d Current-to-voltage plots (I–V), computed as in Fig. 3, in control (circles) and after application of 200 μM Cd2+ (squares). e Same as d, but the I–V indicated by the squares refers to 100 nM ChTx. Cd2+ and ChTx were applied on two distinct cell populations instead of on the same population in two following applications: this strategy excluded possible interactions between the effects of the two drugs, due to incomplete wash. The I–V resulting from the application of Cd2+ (d) and ChTx (e) are plotted against their own control (although both controls were nearly identical)

Full size image

In conclusion, neither I KCa nor I Ca generated the voltage decline and the overshoot of the current-clamp responses at V h  = −60 mV (Fig. 1a). I h did not either generate the latter response features, since it is not expected to be active for any potential more positive than −60 mV (Fig. 6g).

Indeed, I h blockade by the external application of 10 mM Cs+ (or 10 mM Cs+ + 200 μM Cd2+; Fig. 5) did not affect the waveform of the photovoltage in response to depolarising current injections.

Fig. 5
figure 5

Effect of I h, I Ca and I KCa blockade on the current-clamp voltage responses at V h  = −60 mV (a, b and c) and V h  = −40 mV (d, e and f). Stimulus protocol as in Fig. 1 (inset). a, d Control. b, e Application of 10 mM Cs+. c, f Application of 10 mM Cs+ +200 μM Cd2+

Full size image

IKx contribution to voltage responses

The last candidate explaining the peculiar features of the response at V h  = −60 mV is the non-inactivating, slow delayed-rectifier-like K+ current (I Kx; Attwell and Wilson 1980; Barnes and Hille 1989). The simplest approach to study I Kx is to subtract from the control traces the ones where I Kx is fully blocked. Unfortunately, no selective blockers of I Kx are yet available: a non-specific one is Ba2+ (Beech and Barnes 1989; Wollmuth 1994; Fig. 6), because concentrations as high as 1 mM of this ion (replacing 1 mM external Ca2+) were necessary to shift the channel voltage half-activation (estimated from the plot of the tail-current amplitude vs. test voltage; Fig. 6h) of about 17.2 mV (i.e. from −50.5 ± 0.5 mV in control to −33.3 ± 0.7 mV in 1 mM Ba2+, n = 4). To isolate the effect of Ba2+ on I Kx, the Ca2+ channels (and the Ca-activated conductances as well), were blocked by the external application of Cd2+. Moreover, Ba2+ ions screen the cell surface membrane charges less well than Ca2+: therefore, the voltage activation curve of any voltage-dependent channel would be subjected to a negative shift once the external Ca2+ is substituted for Ba2+ (for example, this shift is about −6 mV for Na+ channels; Hille et al. 1975). Therefore, the shift in the voltage half-activation of I Kx would result larger than 17 mV once corrected for the surface charge screening effect. In addition to the effect on I Kx, Ba2+ application seems to block the currents elicited by the hyperpolarisation to −70 and −80 mV (compare the insets of Fig. 6e, f). This effect could be explained by assuming that the latter traces are generated by I Kx deactivation (i.e. I Kx is thought to be still active at −60 mV), their size being reduced in Fig. 6f due to the Ba-induced shift of their voltage activation curve. However, this is not the case, because these two traces were still present when I Kx was fully blocked by 30 mM TEA (Fig. 6g), but they were cancelled by 10 mM Cs+ application (Fig. 8c, d). Therefore, they were generated by I h activation, and their apparent block in the presence of 1 mM external Ba2+ is very likely due to the shift of I h voltage activation curve towards negative values, caused by the aforementioned Ba2+ screening effect.

Fig. 6
figure 6

IKx and Ih waveforms recorded in current-clamp (a, b and c; stimulus protocol as in Fig. 1) and in voltage-clamp (d, e, fVh = −60 mV, stimulus protocol as in Fig. 2a, b; gVh = −30 mV, stimulus protocol consisting of 500 ms voltage pulses steps from −40 to −120 mV in 10 mV increments) and effect of Ba2+ on their voltage sensitivity. a, d Controls. b, c, e, f Application of 200 μM Cd2+. c, f 1 mM Ca2+ was substituted with 1 mM external Ba2+. Scale bar of d, e, f is indicated in between e and f, but vertical scale bar of d is indicated on its right. Responses to voltage pulses from −80 to −30 mV of e and f are enlarged in the corresponding inset on the right of each panel; the two insets have the same scale bar. g isolation of Ih with 30 mM TEA (see Methods); the fast transients at the onset and the offset of the voltage pulses are capacitative artefacts. h Plot of the normalised peak tail current amplitude from e (circles) and f (squares) vs. test potential (n = 4)

Full size image

IKx could be isolated by blocking all the other known current sources present in the rod inner segment by using 200 μM Cd2+ and 10 mM Cs+. Assuming that IKx is not sensitive to the latter ions (as also indicated in Frings et al. 1998), the panels c and f of Fig. 5 illustrate the voltage changes produced by the isolated IKx in response to current injection steps. The size of these voltage changes indicated that IKx is a robust current, that was therefore studied under voltage-clamp conditions while perfusing the rod with Cd2+ and Cs+ (Fig. 7).

Fig. 7
figure 7

Voltage-clamp recording of I Kx  + I h (a, b) and I Kx (c, d) at V h  = −60 mV (a, b) and V h  = −40 mV (b, d). a, b: Application of 200 μM Cd2+. c, d Application of 10 mM Cs+ 200 μM Cd2+. The responses to voltage pulses from −80 to −30 mV are enlarged in the four insets for each V h and ion application; dotted line indicates the zero current level. All responses are normalised to the same amplitude but the four insets, having all the same scale bar. Stimulus protocol of a and c (and relative insets) as in Fig. 2a, b; stimulus protocol of b and d (and relative insets) as in Fig. 2c

Full size image

As expected, the waveform of the currents recorded in response to depolarisations in Cd2+ (Fig. 7a, b) were almost identical to the ones recorded in Cd2+ + Cs+ (Fig. 7c, d), irrespective of the initial V h. I Kx gate is very slow for voltages between −60 and −30 mV (Fig. 7, insets), as indicated by the single-exponential fittings to its activation and deactivation waveforms. Indeed, a depolarisation to −30 mV, starting from a V h of −60 or −40 mV, gave an average activation and deactivation time constant of 280 ± 31 and 115 ± 14 ms (n = 8), respectively. These time constants were even larger for smaller depolarisations: a 10 mV step from V h of −60 or −50 mV gave an average activation and deactivation time constant of 471 ± 95 and 344 ± 31 ms (n = 8), respectively. This slow gating explains well the voltage decline of the responses during the current injection and the overshoot at the stimulus offset (Fig. 1a). Remarkably, for voltages more positive than −10 mV the current decayed with time, after an initial peak, in a voltage independent manner (Fig. 7a–d).

In conclusion, I Kx is the major current contribution shaping the overall photoresponse: since current injections employed so far are step-wise, it would be important to see how I Kx affects the response to a current stimulation with the same time-dependent waveform of the real photoresponse, delivered in the “dark” (i.e. at −40 mV) and on a “background of light” (i.e. at −60 mV). To this aim, current-clamped rods were challenged with stimulation protocols derived from real photocurrent families recorded from OSs of dark-adapted rods (see Methods for details, Fig. 8a), and extracellularly perfused with Cd2+-Cs+, to isolate I Kx as the only current not blocked. The activation of I Kx by setting V h to −40 mV, produced a transient overshoot of the pre-stimulus voltage level in the recovery phase of the flash response. This overshoot was absent when V h was set to −60 mV (below the activation threshold of I Kx).

Fig. 8
figure 8

Photoresponse families recorded in current clamp at V h = −40 and −60 mV in Cd2+-Cs+ extracellular solution, in response to current stimuli shown in the upper panel. The stimulus protocol consisted in whole-cell current responses to 5 ms flashes of intensities: 12.5, 25, 51.2 and 101.6 Rh*/flash recorded from a dark-adapted OS; dark current amplitude of all responses was normalised to 50 pA. Lower panel: superimposition of the photoresponses obtained with the largest current stimulus at V h −40 mV (thick trace) or −60 mV (thin trace). Dotted trace represents the scaled and inverted current stimulus

Full size image

ICa accelerates the photoresponse rising phase and generates spikes

In the presence of Cd2+, R m increases since two conductances are blocked: therefore, Cd2+ is expected to slow down the kinetics of the voltage change produced by a given current injection, while it is expected to boost the response amplitude. Indeed, the response rise time in Cd-treated cells became significantly larger in respect to the control (Fig. 9a) for current injections ≥10 pA (p < 0.01, ANOVA, n = 5; Fig. 9b). While increasing the current injection amplitude, the rise time steadily increased to a plateau level of 41.2 ± 2.9 ms in Cd2+ and of 13.1 ± 0.5 ms in control. However, the voltage amplitude in response to a given current stimulus in the presence of Cd2+ was smaller with respect to the control (Fig. 9a).

Fig. 9
figure 9

Effect of I Ca blockade on the rise time of the current-clamp voltage responses at V h  = −60 mV. a Voltage response to a 15 pA current injection in control (thin trace) and after application of 100 μM Cd2+ (thick trace); the rising phase of both responses is enlarged in the inset. b 10–90% rise time of the voltage response for current injections from 5 to 70 pA in 5 pA increments in control (circles) and after application of 100 μM Cd2+ (squares). The rise time of the responses to 60, 65 and 70 pA current injections were averaged in both conditions

Full size image

The full block of I Kx (and the block of all the K+ currents as well, but the I h) was obtained with external TEA application (30 mM) which, on the other hand, produced an auto-regenerative voltage response to depolarising current injections. Spikes were consistently recorded in all cells when R m  > 1 GΩ, and they were clearly evidenced at V h  = −60 mV (Fig. 10a), where progressively larger current injections produced progressively faster depolarisations to an average steady state amplitude of 17.7 ± 2.3 mV (n = 11; R m ≈ 1.3 ± 0.1 GΩ). Even if the injected current was turned off, this depolarised level was often maintained for a longer duration in tighter cells (rods reported in Fig. 10a, b are an example; mean duration: 12.7 ± 1.9 s, n = 6; range: 2.4–64 s). This indicates that the eventual recovery to V h, once the stimulating current was turned off, was basically controlled by the leakage. This recovery occurred at progressively earlier times when recording from the same cell for several minutes, suggesting the involvement of I Ca in spike generation. Consistently with this view, the spike threshold amplitude (−28.1 ± 0.6 mV, n = 5) was close to the I Ca one. The former was assessed as the voltage amplitude where the first derivative of the membrane potential departed from a constant value in response to current ramps of same maximal amplitude (100 pA) and of duration varying from 70 to 7,000 ms (data not shown; threshold was not affected by the ramp steepness). Moreover, spikes were unaffected by 2 μM tetrodotoxin, but they were abolished by 200 μM Cd2+ and the voltage responses were almost identical to the ones generated by a circuit formed by the parallel between R m and C m (Fig. 10b), as described in Theory.

Fig. 10
figure 10

Effect of I Kx, I KCa and I Ca blockade on the current-clamp voltage responses at V h  = −60 mV. a Voltage responses to current injections from 65 to 85 pA in 5 pA increments lasting 500 ms of a rod with R m  ≈ 600 MΩ perfused with 30 mM TEA. b Response to repetitive 40 pA current injections in a rod perfused with 30 mM TEA (thin trace) and with 30 mM TEA + 200 μM Cd2+ (thick trace); R m was ≈2.6 GΩ, ≈405 MΩ, ≈920 MΩ, and ≈1.1 GΩ just before, and after 1, 21, and 41 s from the first current injection, respectively. All recordings were obtained in 200 μM tetrodotoxin

Full size image

Discussion

Contribution of Na+, Ca2+ and K+ fluxes to the rod photoresponse

In the dark, there is a steady flow of Na+ into the photoreceptor carried by the current through the cGMP-gated channels (I cG), a steady influx of Ca2+ via the I cG and I Ca, and a steady outflow of K+ via the I Kx and the I KCa, while I h  ≈ 0. The resulting membrane permeability to Na+ (P Na), K+ (P K), and Ca2+ (P Ca) generates a membrane potential V m that is ≈−40 mV in the dark, that can be roughly accounted for by the equation:

$$ V_{\text{m}} = \frac{RT}{2F} \cdot \ln \left( {\frac{{P_{\text{Na}} [{\text{Na}}]_{\text{o}} + P_{\text{K}} [K]_{\text{o}} + 4P_{\text{Ca}} [{\text{Ca}}]_{\text{o}} }}{{P_{\text{Na}} [{\text{Na}}]_{i} + P_{\text{K}} [K]_{i} + 4P_{\text{Ca}} [{\text{Ca}}]_{i} }}} \right) $$

that has been computed using the Goldman-Hodgkin-Katz formalism in the reasonable assumption that:

$$ P_{\text{Na}} \approx P_{\text{K}} \quad [{\text{Na}}]_{i} \approx \left[ {\text{K}} \right]_{\text{o}} \quad [{\text{Na}}]_{\text{o}} \approx [{\text{K}}]_{i} $$

and where R, T and F have the usual meaning. During light stimulation, the inflow of Na+ and Ca2+ via the I cG is reduced, and the following hyperpolarisation decreases I Kx, I Ca, and, in turn, I KCa due to the resulting [Ca2+] i fall. Therefore, the rising phase of the light-induced hyperpolarisation is expected to be accelerated by the decreased inflow of Ca2+ (as shown in Fig. 9) and slowed-down by the decreased outflow of K+. On the other hand, I Kx (and, to a lesser extent, I KCa) has a very slow gating (Fig. 7) with respect to I cG and I Ca: as a result, the outflow of K+ is still sustained in the early phase of the photoresponse, actually leading to an acceleration of its rising phase (Barnes and Hille 1989; Fig. 7). Moreover, large hyperpolarisations (i.e. in the presence of strong light stimuli) activate I h and produce a net cation inward current that hinders the hyperpolarisation kinetics. Since I h, I Kx and I KCa gate with 100–200 ms relaxation times (at room temperature), they cause the voltage response to light to fade during the first 100–200 ms, producing therefore sensory adaptation. This adaptation sums up to the one occurring at the level of the enzymatic cascade that progressively re-opens a fraction of the cGMP-gated channels in the presence of prolonged illuminations, leading to a further amplitude compression of the voltage response. Once the light stimulus is terminated and the I cG is progressively restored to its dark level, the following depolarisation is accelerated by I h, which has a fast deactivation kinetics (Fig. 6g). I Ca activation also contributes to this acceleration, especially in the range of current injections above 10 pA, corresponding to the changes produced by the I cG reactivation when moderate to saturating light intensities are turned off (Fig. 9b). This is also evident in Fig. 9a, where the response amplitude to a given current injection (thin trace) is decreased when I Ca is blocked (thick trace). Given the voltage sensitivity of I h and I Ca, this acceleration is particularly large once intense light stimuli are turned off, providing a mechanism that quickly de-saturates the photoreceptor. I Ca activation and I h deactivation produce therefore high pass filtering of the light response. Both I Kx and I KCa avoid that cation influx, occurring at the light stimulus offset via the I cG and the I Ca, could erratically depolarise the rod past its normal resting value. The slow gating of I Kx allows that the above “dark voltage stabilisation” occurs without slowing down the early phase of the photoresponse recovery kinetics. The latter is also accelerated by I Kx when the light stimuli are delivered in the dark with respect to the ones delivered on a background of light, where the responses are anyway greatly accelerated as a results of the feedback regulation of the enzymatic cascade (reviewed in Moriondo and Rispoli 2003). This allows the photoreceptor to integrate light for relatively long time to maximise the sensitivity, without slowing down the recovery kinetics of the response. On the other hand, response amplitude is little affected by the voltage (Fig. 8) when I Kx is the solely active current: the latter does not therefore contribute to the photoreceptor desensitization (i.e. the reduction of response amplitude when it is superimposed on a background of light). I Kx affects rather the response kinetics, which nevertheless plays a key role in the rod performances, given that second-order retinal neurons primarily sense changes in the presynaptic response rather than its steady state level. In conclusion, I Kx and I KCa give a key contribution to the exquisite sensitivity of the rod visual system (which is able to signal the ~1.5 mV hyperpolarisation following the absorption of a single photon), by minimising also the low-frequency voltage fluctuations that may occur in the dark (Moriondo and Rispoli 2003).

The similarity between the responses of Fig. 5c and f with the ones of Fig. 1a and b, respectively (apart the voltage saturation effect due to I KCa activation), shows that the differences between the response families at V h  = −40 mV and at V h  = −60 mV are mainly due to I Kx gating. In particular, I Kx and I Ca are inactive at V h  = −60 mV: positive current injections (simulating the off of the light stimulus) rapidly activate I Ca, which accelerates the rising phase of the voltage response. Current injection also turns on I Kx (Fig. 8d), whose slow activation accounts for the voltage decline of the responses of Fig. 1a, especially for amplitudes between 5 and 30 pA, i.e. just in the range of the current changes produced by dim to sub-saturating light intensities. Larger current injections abruptly activate I KCa (with a fast kinetics; Moriondo et al. 2001) if the following depolarisation is larger than 30 mV: this activation produces the voltage saturation effect described above. At V h  = −40 mV, I Kx is already active: progressively increasing positive current injections activate I Kx and I Ca with progressively faster kinetics (Fig. 8b) leading to monotonic voltage responses, as shown in all responses of Fig. 1b and for current injections above 50 pA in Fig. 1a. Finally, further current injections abruptly activate I KCa, producing again the aforementioned voltage saturation effect.

Spikes

The lack of stability of the dark potential when I KCa and I Kx are blocked is indicated by the generation of spikes observed in rods perfused with Ba2+ or Sr2+ (Beech and Barnes 1989; Brown and Flaming 1978) and by the large depolarisations observed in dark-adapted rods perfused with ChTx (Moriondo et al. 2001) or when blocking all the K+ currents with TEA (Fig. 10). The action potentials recorded from the low vertebrate rods are not generated by voltage-gated Na+ channels, as in human photoreceptors (Kawai et al. 2001), but rather by I Ca activation, as previously found in toads (Fain et al. 1980), turtles (Gerschenfeld et al. 1980), lizards (Maricq and Korenbrot 1988; reviewed in Owen 1987), and primates (Schnapf et al. 1990). Curiously, no Na+ current was found in porcine photoreceptors: this difference may arise from the cell preparation, since human rods were recorded in retinal slices (Kawai et al. 2001) whereas porcine photoreceptors (Cia et al. 2005) and salamander rods (present study) were recorded from dissociated cells, that could have lost some cell parts where the voltage-gated Na+ channels could be entirely segregated. However, the porcine photoreceptor images reported in the Cia et al. (2005) study showed remarkably intact cells, and we also fail to record Na+ spikes even in the most intact rods, retaining the synaptic pedicle. Therefore, the presence of Na+ spikes in mammalian photoreceptors could be restrained to human retina only.

The eventual recovery to V h once the injected current is turned off (Fig. 10) is basically governed by the leak. This explains why this recovery took minutes to occur in the tighter cells (Fig. 10b). It can be therefore concluded that K+ conductances can stabilise the voltage to the dark level, avoiding spike priming which would therefore occur when these conductances are blocked (by TEA, ChTx, Ba2+, or Sr2+).

The spike generation is a Ca-dependent process, since it is blocked by the presence of extracellular Cd2+, but it can not be excluded that I ClCa channels (Maricq and Korenbrot 1988; Barnes and Deschênes 1992) could be implicated in this phenomenon. Indeed, the average sustained depolarisation voltage in the recordings of Fig. 10 is −1.82 ± 0.75 mV (n = 6), in good agreement with the Cl− equilibrium potential of −1.10 mV, calculated by the Nernst equation from the Cl− concentrations used in the intracellular and extracellular solutions, therefore it is possible that I ClCa could be involved in spike generation. Nevertheless, the peak membrane potential recorded at spike early times is about 18 mV (Fig. 10a, see “Results”), which is more positive than the Cl− equilibrium potential: therefore, I Ca plays a major role in shaping the spike rising phase.

Nature of I Kx

In all voltage-clamp recordings of rod I Kx presented here, there was a marked current decline for depolarisations larger than −10 mV (Figs. 6e, f, 8; Wollmuth 1994). No sign of such decline has been observed for voltages up to 50 mV in mammalian rods (Cia et al. 2005). I Kx was remarkably similar to the current recorded from two splice variants of an ether-à-go-go channel cloned from bovine retina (named bEAG1 and bEAG2; Frings et al. 1998), transiently expressed in HEK293 cells, as regard ion selectivity, blockage, voltage activation, Ba2+ inhibition and Cs+/Cd2+ sensitivity. About 80% of the voltage-clamp recordings obtained by Frings et al. (1998) from bEAG1 and bEAG2 (see their Fig. 3) did not show any current decline for any voltage, as found in mammalian rod I Kx; the remaining recordings (20%) did instead show a pronounced current decline for voltages larger than +30 mV, as found in salamander rod I Kx. This gives the attractive perspective that I Kx of vertebrate photoreceptors are coded by the same ether-à-go-go gene, and the differences in the electrophysiological features arise from some post-transcriptional modifications.