The Amplitude and Inactivation Properties of the Delayed Potassium Currents Are Regulated by Protein Kinase Activity in Hair Cells of the Frog Semicircular Canals

In hair cells dissected from the frog crista ampullaris, the combination of a calcium-dependent (IKCa) and a purely voltage-dependent component (IKV) gives rise to the delayed potassium current complex (IKD). These currents have been recently reported to display slow depolarization-induced inactivation and biphasic inactivation removal by hyperpolarization. The amplitude and inactivation kinetics of both IKCa and IKV are drastically modulated by a previously unrecognized mechanism of protein phosphorylation (sensitive to kinase inhibitors H89 and KT5823), which does not interfere with the transient potassium current (IA) or the calcium current (ICa). IKD amplitude was stable in cells patched with pipettes containing 8 mM ATP or under perforated-patch; under these conditions, a 10 min treatment with 10 µM H89 or 1–10 µM KT5823 reduced IKD amplitude by a mean of 67% at +40 mV. Similarly affected was the isolated IKV component (ICa blocked with Cd2+). Thus, a large potassium conductance can be activated by depolarization, but it is made available to the cell to a variable extent that depends on membrane potential and protein kinase activity. The total gKD ranged 4.6–44.0 nS in control cells, according to the level of steady-state inactivation, and was reduced to 1.4–2.7 nS after protein kinase inhibition. When sinusoidal membrane potential changes in the −70/−10 mV range were applied, to mimic receptor response to hair bundle deflection, IKD proved the main current dynamically activated and the only one regulated by PK: H89 decreased the total outward charge during each cycle by 60%. Phosphorylation appears to control both the amount of IKCa and IKV conductance activated by depolarization and the fraction thereof which can be rescued by removal of inactivation. The balance between the depolarizing transduction current and the repolarizing potassium current, and eventually the transmitter release at the cytoneural junction, are therefore modulated by a phosphorylation-mediated process.


Introduction
Inactivation is a typical property of ion channels and determines, in a voltage and time-dependent manner, the number of channels which are available for activation at any moment. The kinetics, voltage-and time-dependence of inactivation onset and removal are quite different, for the various conductances, so that the cell response is molded by the previous history of membrane potential. In the hair cells of the frog semicircular canal virtually all conductances are influenced by membrane potential, independent of biochemical and receptor transduction processes. For example, the IA activates in response to voltage shifts within a few milliseconds, and similarly fast is the onset of current inactivation [1]. On the other hand, inactivation is removed by a mirroringthough relatively slower -voltage-dependent process. An unexpected and more complex set of mechanisms of inactivation (onset and removal) have been recently described for the delayed potassium current complex, IKD, in frog hair cells [1]. The membrane potential level modifies by several-fold the amplitude of the current evoked by depolarization; a sojourn of a few tens of milliseconds at a given holding level is sufficient to markedly modify the amplitude of the response. At difference with IA, the inactivation process is quite slow and only partial for IKD. Thus, the intensity of the current upon activation is influenced by previous conditioning, but the changes are much more persistent over time.
Two relevant experimental manipulations strongly interfere with potassium currents: the ionic-metabolic modifications of the internal cellular milieu which occur in patch-clamp experiments and the concentration of ATP in the pipette. In hair cells, ICa is particularly sensitive to such manipulations, in that its amplitude exhibits relevant up-or down-regulation, according to the internal ATP levels artificially imposed [2], [3].
Less clear, at least in semicircular canal hair cells, is whether currents might also be modulated by intracellular mechanisms related to second messenger systems. Activation of protein kinases, linked to various signaling pathways, has been shown to dynamically modulate ionic channel properties in excitable cells; this may entail major modifications of the excitability machinery (see, for example, the reviews by Jonas and Kaczmareck [4] and Park et al. [5]). Among the targets of protein kinase modulation, potassium channels have deserved special attention. The phosphorylation state of the channel-protein has been shown to affect the open-probability of the channel (Po) (for Kv2.1: Park et al. [6]; for Kir6.2: Light et al. [7]), the inactivation mechanism (for Kv4.2: Jerng et al. [8]; for Kv3.4: Covarrubias et al. [9]; for Kv1.4: Roeper et al. [10]), the current activation curve and the calcium affinity (for calcium-dependent channels of BK type [11]), the total number and clustering of channels in the plasma membrane as a consequence of relocation of the channel protein [12], [13].
Given the sensitivity of ionic currents to intracellular ATP levels, in isolated frog hair cells, in this study we explore the effects of treatments potentially involved in phosphorylation of the most relevant channels. After considering the effects on individual isolated currents, the dynamic interplay of the single currents is analyzed, in generating the integrated electrical behavior of the cell. Finally, a minimal electrical model of the IA current in the isolated hair cell is proposed, which might help to understand the complex current interplay during sensory stimulation.
The mechanisms that modulate ion channel properties have diversified voltage dependence and operate on quite different time scales, thereby producing a wide activity-dependent regulation of hair cell excitability.

Preparation of Hair Cells
Frog labyrinth preparation, stimulation and recording procedures have been described in detail previously [1][2][3], [14]. All procedures for handling of animals and surgery were approved by the Animal Care and Use Committee of the University of Ferrara. The experiments were performed on wild frogs (Rana esculenta, 25-40 g body weight) that were mainly harvested in summer. The animals were anesthetized by immersion in tricaine methane sulphonate solution (1 g/l in water) and subsequently decapitated. Dissection of the two labyrinths was performed in a solution of the following composition (mM): 120 NaCl, 2.5 KCl, 0.5 EGTA, 5 4-2-hydroxyethyl-1-piperazineethanesulfonic acid (HEPES), 3 glucose and 20 sucrose. The final pH was 7.2 and the osmolality 260 mOsmol/kg. Six ampullae were treated for a period of 20-30 s with subtilisin A, type VIII (50 mg/ml, Sigma); the protease was thereafter blocked by trypsin inhibitor type II-S (Sigma) added to the dissection solution (final concentration, 0.7 mg/ml). Ampullae were then transferred to the experimental chamber (500-ml volume), submerged in the standard extracellular solution (mM: 120 NaCl, 2.5 KCl, 2 CaCl 2 , 1 MgCl 2 , 5 glucose, Tris buffer, pH 7. [3][4][5] in the presence of the trypsin inhibitor. Hair cells were mechanically dissociated from the ampullae by gently scraping the epithelium with a fine forceps. The protease inhibitor was eventually washed out before starting with electrophysiology. The glass bottom of the experimental chamber was coated with chloro-tri-n-butyl-silane to prevent cell sticking. Cells were viewed through a TV monitor connected to a contrast enhanced video camera (T.I.L.L. Photonics, Planegg, Germany) and continuously superfused with the extracellular solution. The camera was coupled to an inverted microscope (Olympus IMT-2, Tokyo, Japan) equipped with a 406 Hoffman modulation contrast system.
Cd 2+ (200 mM) was used as a blocker of voltage-dependent calcium channels. It was applied by rapidly changing (typically ,50 ms) the external solution by horizontally moving a multibarrelled perfusion pipette positioned in front of the recorded cell (with a computer-controlled stepping motor). The perforatedpatch configuration was obtained by adding amphotericin B (300 mg/ml) to the pipette solution. To facilitate the giga-seal formation, the amphotericin-containing solution was used to backfill the electrode, while the tip was filled with an amphotericin-free solution.
Series resistance ranged between 8 and 22 MV. The cell capacitance and series resistance were electronically compensated (50-75%) before each voltage-clamp protocol was run. Leak was measured near resting potential with 10 mV615 ms hyperpolarizing pulses; this leakage current was subsequently subtracted, assuming a linear behavior, in correcting current recordings offline. Currents were low-pass filtered at 5 kHz and acquired online at 10 kHz with pClamp hardware and software (pClamp 9.1 and Digidata 1322A interface; Axon Instruments, Union City, CA, USA). Data were analyzed offline by using pClamp 9.1 and Matlab 10a (The MathWorks, Natick, MA, USA) software. They were not corrected for the liquid junction potential, estimated to be about +4 mV under our standard recording conditions [1].
The IKD conductance of the cell, and the fraction thereof which is available at any moment, were estimated, assuming linearity of the instantaneous I-V curve, according to the equation: gKID = IKD/(V-E K ), with E K = 295 mV. The peak conductance was evaluated for currents evoked at V = +40 mV, where the N-shaped I-V curve of IKD exhibits its maximum, close to the experimental equilibrium potential of the calcium current.
Previous results indicated that the magnitude and kinetic properties of single ionic currents do not display any consistent differences and/or correlations among the different cell types [1]. Since the basic resting properties of the isolated hair cells that survived the dissection procedures were relatively homogeneous, no attention was paid to the specific cell type in analyzing and comparing different experimental conditions. No systematic differences were produced by the levels of ATP here tested (1 vs. 8 mM) or by external (or internal, not shown) H89 application, in the general electrical properties of hair cells (see Table 1). This rules out that any conductance became activated or inhibited by the treatments in the unstimulated cell.

The IA Model
The transient outward conductance, g A , in the frog hair cell was assumed to obey a Hodgkin-Huxley paradigm, controlled by a pair of voltage-dependent parameters, a and h a , respectively related to activation and inactivation. IA was modeled according to the kinetic scheme: IA(V,t) = g A ?a(V,t)?h a (V,t)?(V m -E K ), where g A is the maximal conductance (20.2 nS per cell), in the absence of inactivation, V m is the membrane potential, and E K is the potassium ion equilibrium potential (295 mV). The steady-state values and the time constants of the activation and inactivation parameters are presented as functions of voltage, and are valid within the voltage range of physiological interest (275/0 mV). The complete set of equations describing the voltage dependence of the activation and inactivation variables are derived from the experimental data presented in a previous report (Martini et al. [1]) and are as follows:

Statistical Analysis
All data are reported in the text and in the figures as means6SE. Comparisons among different conditions were performed by Student's t-test (paired t-test for paired observations) or by ANOVA. Values of P,0.05 were considered significant.

Results
Typical ionic currents recorded in hair cells acutely isolated from frog semicircular canals are illustrated in Fig. 1, together with the voltage protocols that explore their main features: voltagedependent activation, isolation of the single current components and inactivation mechanisms. The pipette K-ATP complement was 1 mM. The compound delayed K current, IKD, is displayed in panel Aa, as evoked from a holding level of 240 mV (to inactivate the transient IA potassium current). IKD is a compound current contributed to by the calcium-dependent potassium current, IKCa, the purely voltage-dependent potassium current, IKV, and a smaller calcium component, ICa. For clarity, only three tracings from the complete I-V curve are illustrated: the current is well developed during the step to 0 mV; it reaches its maximal amplitude at +40 mV (I-V curve peak); the current amplitude displays a region of decrement at more positive voltage steps, conferring the I-V curve a typical N-like shape (see Fig. 2C in Martini et al. [1]). Fig. 1Ab illustrates the protocol to dissect the IA current in isolation, from the difference between the first response (following a 1-s pre-pulse at 2100 mV to remove IA inactivation) and the second response (following a 300 ms sojourn at 240 mV to keep IA inactivated). This procedure also reveals the recently reported inactivation of IKD: in fact, IKD amplitude is smaller at 240 mV holding potential (panel Aa) than the steady response following the hyperpolarizing pre-pulse (arrow in panel Ab, note different amplification). Thus, a 1-s sojourn at 2100 not only removes IA inactivation, but is also able to remove a fraction of steady inactivation of IKD (details in Martini et al. [1]). The fold-increase in IKD amplitude, following hyperpolarization, will be referred to as inactivation removal coefficient (irc). At each value of membrane potential, the irc constitutes an indirect measure of the extent of steady-state inactivation of the channels contributing to IKD. Both protocols were repeated every 2 min during a period of 6-10 min. Fig. 1B illustrates the current modifications that occurred during the first 8 min observation period. They have been systematically observed in 6 cells and can be listed as follows: 1) the delayed potassium currents decreased in amplitude (rundown) and ended up displaying a different I-V shape: the N-type profile was no more present in panel Ba (larger response at +80 mV than at +40 mV); 2) the IA current was not affected by any rundown: its peak amplitude remained virtually constant (4.1 vs. 4.4 nA at +80 mV; 1.5 vs 1.9 nA at 0 mV); 3) the inactivation machinery of IKD was operating throughout the observation period (actually, in this example the irc increased from the initial value of 1.6 to 7.4 at 8 min); 4) the time-course of IKD response was very similar, independent of the recording time; in particular, activation was very fast when the current was evoked from 240 mV holding potential, but much smoother after inactivation removal, at all times. Since ICa often runs down during long-lasting recordings with 1 mM internal ATP [2], some of the effects might be related to a decrease in the IKCa component of IKD.

Stabilizing Effect of ATP
The amplitude of IKD current is liable to run-down, independent of inactivation removal. Since steady-state conditions of the IKD machinery are a prerequisite for a reliable evaluation of the treatments, it seemed important to establish recording conditions such that IKD properties remained stable over time. The stabilizing action of high intracellular ATP concentration on ionic current amplitude was verified first. Alternatively, the 'perforated-patch' configuration was used to minimize the perturbation of the internal milieu inherent with 'whole-cell' recording. In Fig. 2 the effects of 8 mM K-ATP in the pipette are shown, in experiments mirroring those of Fig. 1 (the first response, containing IA, has been omitted in both ''b'' protocols, for clarity). The delayed current complement clearly remains constant during a 10-min observation period: the I-V curve is virtually identical, its N-type shape is preserved throughout, and the inactivation removal mechanism is remarkably constant, with an irc value around 1.8. Similar results were observed when the perforated whole-cell configuration was tested (see below). High ATP in whole-cell was generally preferred here to the perforated-patch configuration, in order to clamp the internal ATP at a known concentration; moreover, high ATP was previously shown to prevent ICa amplitude rundown in hair cells and to preserve it from inactivation even during prolonged depolarization [3]. Despite the powerful stabilizing effect of ATP, the rundown of IKD was not reversed when pipettes back-filled with high ATP, but with 1 mM ATP at their very tip, were used. Cell loading with ATP was thus delayed by the time required for high ATP to equilibrate within the pipette: the usual current decay started to take place; in 3 out of 5 cells, the current decay broke off, but it did never reverse. Figure 3 summarizes the mean behavior of currents in the presence of 1 or 8 mM ATP in the patch pipette (6 and 9 cells, respectively). Current amplitudes evoked at +40 mV are presented; this single point is representative of the whole I-V curve, since the action of ATP was equally evident at any membrane potential. Panel A illustrates responses from 240 mV holding potential: the IKD current amplitude decays with time in low, but not high Figure 1. Decomposition of the effects of long-lasting recording with 1 mM ATP-patch clamp-pipettes. A. Initial tracings of the currents evoked in a hair cell by steps to 0/+80 mV (only three steps illustrated for clarity) from a holding potential of 240 mV (a) and by the same protocol after a 1 sec conditioning at 2100 mV (b). In (b) the current families are evoked twice: immediately following conditioning (IA inactivation removed) and after a sojourn at 240 mV sufficient to maintain full IA inactivation (but with little effect on the removal of IKD inactivation by the previous conditioning at 2100 mV). The difference-currents between the two episodes dissect the pure IA at each test potential (see text and Martini et al. [1], for details). Arrows indicate the IKD amplitude used to evaluate the effect of inactivation removal. B. Same protocols as in A, applied in the same cell 8 min later. Note the drastic decrease in IKD amplitude and the monotonic increase in the I-V curves (as opposed to panel A). IA peak amplitude, after appropriate dissection, remains constant at each test potential, the apparent decrease being exclusively due to the IKD component. doi: 10.1371/journal.pone.0067784.g001 ATP. Panel B shows responses after inactivation removal (1 s at 2100 mV) in the same group of cells. The numbers associated to each symbol indicate the corresponding values of irc. Both IKD amplitude and the irc value remained remarkably constant with high internal ATP. Conversely, the amplitude of IKD clearly declined with time in low ATP, even after inactivation removal, in spite of the parallel increase in irc (i.e. the rundown was not accounted for by rapidly reversible inactivation alone). Panel C shows that the peak amplitude of IA at +80 mV was totally insensitive to ATP concentration, even over long-lasting recording (n = 5, high ATP; n = 7, low ATP). Insensitivity of IA peak amplitude to H89 treatment was observed not only with 1 mM ATP+H89 in the pipette ( Fig. 3C; n = 6), but also when the perforated-patch configuration was used (not shown; n = 6).
The mix of 1 mM ATP plus 7 mM K-ADP in the pipette did not substitute for the 8 mM ATP solution in preventing current rundown.

Effect of H89, a Protein Kinase Inhibitor, on IKD
Two protein kinase inhibitors have been tested: 10 mM H89 and 1-10 mM KT5823, presumably specific for PKA or PKG, respectively. H89 was bath applied, under the conditions (8 mM ATP or perforated patch) which have been shown above to stabilize the delayed currents. Figure 4A shows representative IKD tracings, evoked in the same cell over the 230/+40 mV range from a 240 mV holding potential, before and 8 min after H89 treatment. Figure 4B displays the time course of the peak IKD amplitude, evoked at +40 mV from 240 mV holding potential, with 8 mM ATP in the pipette (triangles, same data as in Fig. 3A), and the same measurements in the presence of 10 mM H89, applied to the bath at the arrow (filled squares, n = 7), or preincubated for a 15-30 min period (open squares, n = 6). The PKA inhibitor, H89, clearly produced a substantial IKD decay, in cells otherwise stable. Despite the intrinsic variability among single hair cells, the mean initial IKD amplitude was significantly lower in cells pre-incubated in H89 (404.76134 pA, n = 6) vs. the gross mean of controls (983.16168pA, n = 16; P,0.05, Student's t-test), suggesting that the effects of the treatment were already present when the recording started.
The mean I-V curves in the same groups of cells, before and after 8 min exposure to H89 (8 mM ATP in the pipette) are shown in Fig. 5A. From these data, some voltage dependence in the effect of H89 is apparent. In particular, the drug appears to more strongly affect currents at potentials positive to +10 mV. The distinct components which comprise the IKD complex (a mix of calcium and potassium currents) intrinsically exhibit different I-V curves, so that the overall effect of H89 on the potassium component alone can be best evaluated close to the equilibrium potential for calcium ion (about +40 mV), where calcium current contribution is minimal.
As mentioned, and clearly evident in Figs. 1 and 2, inactivation of both IA and IKD is readily removed at 2100 mV. Actually, the delayed current, which was originally thought to be exempt of inactivation, in addition to fast recovery also displays a welldefined process of slow removal of inactivation. The two processes are readily distinguished: the fast one develops with a time constant of tens of milliseconds, while the slow one only ensues after a delay of about 300 ms and thereafter reaches its maximum with a time constant of seconds [1].
The voltage dependence of the fast, partial removal of IKD inactivation was investigated by conditioning for 1 sec at various potential values (240/2120 mV voltage range) hair cells either untreated or exposed for 8 min to H89 (8 mM ATP-pipette). In addition to reducing the absolute IKD current, H89 clearly impaired the inactivation process, as shown in Fig. 5B. A 1-s sojourn at different conditioning potentials produced a markedly voltage-dependent removal of inactivation in control cells, whereas in the presence of H89 the inactivation removal I-V relation was almost completely flattened. In particular, when the responses at +20 mV were compared in each cell before and after the sojourn at 2120 mV, a mean irc value of 5.561.7 was obtained in control cells (n = 5), versus a irc of 1.760.1 following H89 treatment (n = 8; P,0.001; Student's t-test). Since an irc value .1 was still observed, following H89 treatment, it was of interest to investigate whether not only the magnitude but possibly the voltage dependence as well of the inactivation removal might be affected by H89. Steadystate inactivation is usually expressed by a voltage-dependent ''h'' parameter, which may range from 0 (full inactivation) to 1 (inactivation completely removed). Since inactivation of IKD is only partial, the h KD parameter was estimated by measuring peak currents, isolating the fraction subject to inactivation (i.e. the current remaining after 240 mV pre-conditioning was subtracted) and normalizing the results to their largest value (that obtained after 2120 mV preconditioning). The voltage dependence of h KD was described by Boltzmann-type equations fitting the data, which yielded the following parameters: half-potential, V1/ 2 = 288.262.1 mV, and slope coefficient, b = 210.261.5 in control (n = 6) vs. V1/2 = 283.861.7 mV, b = 210.861.4 in H89 (n = 10). ANOVA analysis indicated that the slight shift to the right of the h KD curve in H89 was not statistically significant. Thus, the extent, but not the voltage dependence, of fast inactivation is affected by pharmacological manipulations of PKA.
The slow component of IKD inactivation also is a likely target of regulatory mechanisms; thus, it was similarly investigated next, in the presence of 8 mM ATP and H89. Slow removal of inactivation was evaluated by applying a 100-s pre-conditioning pulse at 2100 mV. This produced an additional consistent increase of the delayed current amplitude evoked by a step to +40 mV, by 62611% in controls (n = 5) and 1665% (n = 6) in H89-treated hair cells. The difference was statistically significant (P,0.01; Student's t-test), suggesting that both fast and slow components of the process of removal of IKD inactivation are modulated by H89 (i.e., they presumably involve a phosphorylation process).

Effect of H89 on IKV, IA and ICa
The pure IKV component is readily extracted from the IKD complex in [Ca 2+ ] o = 0 or by applying external 200 mM Cd 2+ to block ICa and, consequently, IKCa. Figure 6A illustrates the normalized peak amplitude of IKV over time in the three conditions previously tested for IKD: internal 1 or 8 mM ATP, and internal 8 mM ATP with bath applied 10 mM H89. The corresponding inactivation removal coefficient, irc, is reported close to each data point, as evaluated from the absolute current amplitudes. Results mirror the IKD data in indicating that also IKV suffers from a similar rundown (a mean decrease by 57% within 6 min) if not supported by high ATP (low ATP, open circles; n = 5 vs. high ATP, triangles; n = 7), and that it is strongly inhibited by H89 (a mean decrease by 73% after 6 min application, squares; n = 5). Consistent with the observations above for total IKD, the recovery in IKV amplitude due to inactivation removal is impaired by the protein kinase inhibitor, as indicated by constant values of irc in the strongly decayed currents. IKV accounts for about 30% of the whole IKD in normal hair cells. Therefore, inactivation of IKV alone does not account for the effects observed for the whole IKD.
The IA dynamics (I-V curve, peak current amplitude, inactivation mechanism) proved remarkably insensitive to the above treatments. Fig. 3C shows that the combined effect of two manipulations, namely 1 mM internal ATP and H89 exposure, which would otherwise strongly depress the IKD complex, does not affect the IA amplitude over time. Numbers close to symbols indicate the ratio of current amplitude after/before inactivation removal (irc). C. Mean peak amplitude at +80 mV of pure IA, after current dissection as illustrated in Fig. 1 and text, in cells recorded with 1 mM ATP-pipettes (filled circles; n = 7), 8 mM ATP-pipettes (triangles; n = 5) or 1 mM ATP pipettes with 10 mM H89 added to the bath (squares; n = 6). doi:10.1371/journal.pone.0067784.g003 The calcium current, ICa, is potentially affected by run-down or (less frequently) by run-up in frog hair cells, unless they are stabilized by high internal ATP [2]. In these conditions, ICa amplitude becomes constant and calcium-dependent inactivation is canceled, even during long lasting current inflow [3]. Figure 6B shows ICa peak amplitude in stabilized hair cells, evoked at 210 mV (where ICa exhibits its maximum) at increasing times in the presence or absence of H89. In the inset of the same figure, representative ICa tracings evoked at 210 mV illustrate that both amplitude and time course of the current are unaffected by time and PK inhibition. ICa appears as the only component of the IKD complex insensitive to protein kinase manipulation by H89.

Effect of KT5823
The results of H89 application proved highly reproducible in isolated hair cells, in terms of both the time course of the effect and the extent of current decay. KT5823 (1-10 mM, 8 mM ATP in the pipette) gave homogeneous effects on IKD amplitude, similar to those observed with H89. In a group of 5 cells exposed to 1 mM KT5823 in [Ca 2+ ] o = 2 mM, the mean amplitude of IKD decreased to 36.8% within 6 min (690.44695.7 vs. 254.486120.5 pA; P,0.01, paired Student's t-test). In two cells treated with 10 mM KT5823, the decrease was to 44.7%. When external calcium was removed (only IKV surviving), the final mean decrease was to 36.0% of the value before drug application (n = 3). Less predictable were the effects on the inactivation process. In three of these cells (1 mM KT5823) an additional, unexpected impairment of the IKD activation/inactivation mechanism ensued. The IKD complex appeared to lose any voltage dependence in a cell kept at 240 mV holding potential, as if all components with voltage-dependent activation had been canceled by inactivation. A 1-s pre-pulse at 2100 mV was sufficient to restore the voltage dependence (Fig. 6Cab). Conversely, in a different cell the inactivation removal was completely inefficient, over the 240/2120 mV range, and the test pulse evoked currents of constant amplitude, independent of previous conditioning (Fig. 6Cc).

Effect of Inhibitors or Activators
The observations here reported suggest that PK activity is necessary to control the IKD machinery. To test whether the intracellular signaling mechanism(s) might directly involve cAMP or cGMP, the following tools were employed: 300 mM IBMX, a non-specific inhibitor of cAMP and cGMP phosphodiesterases (applied in the bath or via the patch pipette; n = 9); 10 mm Rolipram, a selective cAMP-specific phosphodiesterase (PDE4) Figure 4. Effect of H89 on IKD amplitude under whole-cell or perforated-patch configuration. A. Representative IKD current families evoked in the same cell before and 8 min after H89 bath application. Holding potential was 240 mV, to inactivate IA. B. Mean IKD amplitude, evoked at +40 mV from a holding potential of 240 mV, at increasing time in control hair cells (triangles; same data as in Fig. 3A), in cells exposed to 10 mM H89, bath-applied at arrow (filled squares; n = 7), and in cells previously exposed to H89 for 15-30 min before starting whole-cell recording (open squares; n = 6). Patch pipette contained 8 mM ATP throughout. C. Time course of the normalized mean IKD amplitude, evoked at +40 mV from 240 mV, in control (triangles; n = 7) and cells exposed to 10 mM H89, applied at arrow (squares; n = 6). Perforated-patch recording throughout. doi:10.1371/journal.pone.0067784.g004 inhibitor (n = 3); 500 mM 8-Bromo-cAMP, a cell-permeable cAMP analog, activator of PKA (n = 6); 50-200 mM 8-pCPT-cGMP, a membrane-permeant activator of cGMP-dependent protein kinases and cGMP-gated ion channels (n = 7). Single hair cells were internally perfused via pipettes containing 0-1-8 mM ATP. None of the listed treatments was able to modify the previously described behavior: modifications in IKD current amplitude and inactivation mechanism in low ATP or steady-state conditions in high ATP.
These observations, combined with the data obtained with H89, indicate that adequate availability of ATP and functionality of kinases are both required to maintain the native amplitude of the IKD currents and to remove inactivation of the channels at negative potentials.

Modulation of Sinusoidal Voltage Commands
We verified the effects of H89 under conditions which might mimic, more closely than the simple I-V curve, the physiological dynamic behavior of the hair cell during activity. The membrane potential of the hair cell is expected to fluctuate physiologically from a 'resting' status close to 240 mV (cilia at an intermediate position, a fraction of the receptor transduction channels open), toward 270 mV (all channels closed by the inhibitory deflection of the cupula), or toward a more positive voltage level (activation of additional transduction channels during the excitatory movements of the cupula). The natural membrane potential range of the hair cell might thus span from 210 mV, close to the equilibrium potential of the transduction current, during maximal sensory stimulation, to 270 mV, close to the zero-current voltage, in the absence of any transduction current flow. We thus mimicked the receptor potential time course by driving the hair cell membrane potential with a sinusoidal waveform that fluctuated between 270 and 210 mV, to simulate the whole excitatory/inhibitory cycle of the hair bundle, or between 240 and 210 mV to simulate the excitatory effects only. Figure 7A illustrates the current tracings evoked in the same cell by sinusoidal voltage shifts at 1 Hz in the 270/210 voltage range in control saline over an 8 min period (8 mM ATP in the pipette). Similarly, Fig. 7B shows the effect of an 8 min exposure to H89 (panel b), compared with the initial tracing in normal saline (panel a). The charge displaced by the voltage shifts was comparable during the two hemicycles of each episode. The total outward charge evoked during a single complete excitatory/inhibitory cycle was used to evaluate the effects. The same experimental conditions as in Table 1 have been tested; the results are partially summarized in Table 2, to show the final result of each treatment.
The same conclusions as in I-V experiments can be readily drawn. Even under these dynamic conditions, the outward potassium current is maintained for several minutes only in the presence of high internal ATP. Low internal ATP or H89 treatment, even in the presence of high ATP, are unable to sustain a stationary potassium outflow, independent of the voltage range explored.

Current Dissection during Sinusoidal Cycles
As illustrated by Fig. 1, the different components of the potassium current are readily dissected during instantaneous voltage commands at various potentials. The pure calcium current Figure 5. H89 effect on I-V curves of IKD and on IKD inactivation removal. A. Mean I-V curves in the same group of hair cells (n = 5) evaluated in normal saline (triangles) and 8 min after application of 10 mM H89 to the bath (squares). B. IKD fast inactivation was progressively removed by a 1 s sojourn at increasing conditioning potentials in the range 240/2120 mV, and thereafter tested at +20 mV, in control cells (triangles; n = 6) or hair cells previously exposed for 8 min to 10 mM H89 (squares; n = 8). Each conditioning episode was followed by a 30 s period at 240 mV to allow inactivation to fully develop [1]. See text for the fit of data with Boltzmann-type equations. Pipette with 8 mM ATP throughout. doi:10.1371/journal.pone.0067784.g005 can also be isolated by appropriate ionic manipulations. During slow fluctuations of membrane potential, which are more pertinent to the physiological hair cell behavior, the contributions of the single current components is less directly predictable. Figure 8 illustrates the decomposition of the total outward current in an ideal 'mean' hair cell during repeated sinusoidal cycles in the 270/210 mV voltage range, in control conditions or in the presence of H89. The total current and the pure ICa are derived from direct measurements in patch-clamp (Fig. 8A,B). Since IA cannot be isolated (as was done in Fig. 1) under these conditions, and no selective pharmacological blockers are available, IA contribution was simulated according to the mathematical model described in Methods (Fig. 8C).
The changes in membrane potential here considered are very slow, compared to the time constants of IA activation and inactivation onset/removal, so that the IA conductance does not depart much from its steady-state values, at any potential. Since either the steady-state activation is quite low (below 250 mV) or the steady-state inactivation predominates (above 250 mV), the overall contribution of IA results quite small over the whole cycle (Fig. 8C). Its contribution might obviously be much more relevant for rotations much faster than 1 Hz. This point is illustrated in Fig. 8C where, during the second sinusoidal cycle, voltage is abruptly stepped to the final 210 mV level (dashed current/ voltage tracings). The fast current activation now generates a much larger IA component. Although the labyrinthine organs and hair cells are generally looked at as low-frequency mechanical sensors (as compared to their cochlear counterparts), rapid movements of the head might entail components of angular acceleration at quite high frequencies. Under such conditions the contribution of IA might become more relevant in molding the signal that reaches the CNS and the subsequent physiological responses.
Based on the I-V curves reported above, ICa and IA have been considered to be unaffected by H89 during sinusoidal activation as well. The summed IKCa-IKV complex was thus isolated by algebraically subtracting ICa and IA from the total current (Fig. 8D). The decrease of IKCa+IKV amplitude following H89 Figure 6. Effect of protein kinase inhibitors on IKV and ICa amplitude. A. Isolated IKV are dissected by omitting external calcium or in the presence of 200 mM Cd 2+ (results were similar and pooled). The mean IKV amplitude (normalized) evoked at +40 mV from a 240 mV holding potential is plotted against time in hair cells recorded with 1 mM ATP pipettes (circles; n = 5), 8 mM ATP pipettes (triangles; n = 7) or with 8 mM ATP pipettes and 10 mM H89 bath-applied at arrow (squares; n = 5). Numbers close to each symbol report the inactivation removal coefficient (irc) for IKV following a 1 s conditioning at 2100 mV. B. Normalized ICa peak amplitude plotted versus time in control cells (triangles; n = 4) or hair cells exposed to 10 mM H89 for 6 min (bath-applied at arrow) (squares; n = 4). Pipette with 8 mM ATP throughout. Representative ICa tracings evoked by 270/ 210 mV voltage steps, before and after 6 min in control saline or in the presence of H89, are illustrated in the inset. Tail currents are truncated. C. 1 mM KT5823 application for 6 min cancels the voltage dependence of IKD activation when steps start from a holding potential of 240 mV (a). Preconditioning the same cell at 2100 mV for 1 s (b) removes inactivation, and restores activation at the same membrane potentials tested in (a). In another cell, preconditioning over the 240/2120 mV range is completely inefficient in removing IKD inactivation (c). doi:10.1371/journal.pone.0067784.g006 treatment is evident, as was quantitatively reported in Table 2, and an asymmetry in charge movements between the excitatory and inhibitory hemicycles is revealed, mirroring the two bouts of ICa activation.

Discussion
The present experiments confirm the presence of a voltagedependent inactivation mechanism in the delayed potassium current machinery, IKD, in frog hair cells. We show here that the two major potassium currents which build up the IKD complex, IKCa and IKV, exhibit amplitude and steady-state inactivation properties which are drastically influenced by a previously unrecognized phosphorylation-dependent modulation. Such control is independent and specific, for each of the two currents; the transient potassium current, IA (though it exhibits a powerful native inactivation mechanism), and the calcium current, ICa, are completely insensitive to this regulation.

The KD-conductance Complement of the Hair Cell
The frog hair cell is endowed with a large potassium conductance which could be activated by depolarization, but is made available to the cell in a variable amount, efficiently controlled by the resting membrane potential level. Despite the large variability among cells, in the unstimulated cell maintained at the presumed resting membrane potential around 240 mV, about 4.6 nS of total delayed potassium conductance are available. This figure can be raised by a factor up to 5.5 by progressively removing the fast inactivation, corresponding to the 240 mV status, by a short-lived sojourn (1 s) at more negative potentials; the maximal effect is reached with a conditioning pulse at 2120 mV and develops within few milliseconds. Inactivation removal is enhanced by maintaining the cell at hyperpolarized levels for a longer time: the available potassium conductance further increases by an additional factor of 1.6 (+60%), though the process takes several seconds, providing the cell with a total potassium conductance of about 44 nS. The starting conductance level, and the subsequent adjustments due to membrane potential shifts, are all controlled by protein kinase activity. H89 treatment decreases the resting conductance to a mean of 1.4 nS; the available conductance is still sensitive to hyperpolarization, but with a limited ceiling: a 1.7 factor for fast inactivation removal, and an additional 1.16 factor if the slow process of inactivation removal is also allowed to fully develop. The maximal available potassium conductance is now reduced to 2.7 nS (about 6% of the total conductance in control conditions). Normal hair cell activity develops over a voltage range more restricted than that explored by biophysical experiments: it presumably never exceeds 275 mV, the zero-current level below which no cell conductances are active. Still, the reported effects of voltage and protein kinase control are quite relevant even within these functional limits: the KCa-KV conductance available would fluctuate in the 4.6-10.2 nS range under control conditions, and in the 1.4-2.1 nS range following H89 treatment.
For the sake of comparison, the irc values (ratio between current amplitude evoked from a conditioning potential that completely removes inactivation and that evoked from the native resting membrane potential), for fast currents that typically exhibit a steady-state inactivation mechanism, are of the same order of magnitude as those here reported for IKD: irc = 1.7 for I Na in squid axon, 1.6 for I Na and 4.9 for IA in rat sympathetic neurons [15].

Protein Kinase Effects
The present experiments aimed at exploring whether kinase activity was implied in the IKD inactivation machinery, but also what kinases and cyclic nucleotides might be involved. We have no definite answer for the latter questions.
As regards the kinase identity, H89 is considered to be specific for PKA, while KT5823 for PKG. The concentrations used here, however, may make the specificity ill-defined. Previous studies have shown that purified PKG can be inhibited by H89 in vitro, although the K i is 10 times higher than that for PKA [16][17]. Most data on activation of PKG suggest that its targets are different from those investigated here: nitric oxide inhibited the voltage-dependent ICa in rat vestibular hair cells by the activation of a cGMP-signaling pathway [18], and activation of PKG drastically decreased ICa in hair cells of the frog saccule, via the release of nitric oxide [19].
Few reports in the literature have concerned the mechanism of action of protein kinases on the electrical activity of semicircular canal hair cells. Jagger and Ashmore suggested a regulatory role for protein kinase A in guinea-pig cochlear hair cell [20][21]. Their main evidence was based on the fact that bath-applied 8-Bromo-cAMP, an activator of PKA, increased the amplitude of Ikf, a current similar to IKCa.
In the present experiments permeable analogs of cAMP or cGMP, as well as more or less specific phosphodiesterase inhibitors, were apparently unable to influence the final current behavior expected in the presence of different internal ATP levels. This finding could be explained by a significant level of constitutive activation of kinases which maintain the channel protein in a highly phosphorylated state, as occurs in other cell types ( [6]; some reviews of the extensive literature: Jonas and Kaczmarek [4]; Schubert and Nelson [22]; Hou et al. [23]).
In the native hair cell the cytosolic millimolar ATP concentration is likely close to that imposed artificially via high-ATP pipettes. Since the basal level of channel phosphorylation would be determined by the equilibrium between phosphorylation and dephosphorylation processes, the use of PK inhibitors would shift such equilibrium toward the dephosphorylated state and account for the decrease in the available potassium conductance and the steady-state inactivation impairment here observed.

Effect on Current Amplitude
Protein kinase inhibition by H89, and by KT5823, produces two distinct effects: it decreases the amplitude of both IKCa and IKV, and it interferes with their steady-state inactivation mechanism. Both current components are similarly affected, when considered in isolation. This might reflect similar structures of the channel proteins. The IKV fraction contributes to the total IKD by about 30%, so that its depression cannot account for the final cumulative effect on IKD. The involvement of IKCa is demonstrated by considering the difference current before/after Cd 2+ application: it appears to be intrinsically and genuinely sensitive to kinase inhibition, since no changes are observed in ICa amplitude, which is insensitive to H89 treatment.
Detailed analysis of the effects of H89 on the IKD activation curve (Fig. 5A) is hard, since at least three distinct currents, with different I-V relations, are involved. The steady-state inactivation curve, on the other hand, only reflects the behavior of the two potassium currents (ICa does not exhibit any steady-state inactivation property), which have rather similar molecular structure and display very similar sensitivity to the treatments. Incubation in H89 did not produce any significant shift in the inactivation curve, nor did it have any effect on the slope factor. This would suggest that H89 reduces the number of readily available channels, which maintain their normal responsiveness to depolarization.

Effect on Steady-state Inactivation Mechanism
A number of voltage-gated channels sustain transient currents. Typically, the sodium and IA currents, once evoked, exhibit a fast decay due to an intrinsic mechanism of fast inactivation (open state inactivation, OSI). The underlying molecular mechanisms are relatively well understood and there is evidence that they are regulated by protein kinase activity: for example, phosphorylation of potassium channels by serine/threonine kinases alters the rate of current decay due to fast inactivation ( [4][5][6][7][8][9] [8]). For these channel types, when the holding membrane potential is made progressively more negative, the current intensity in response to depolarizing steps to the same voltage levels progressively increases (closed state inactivation, CSI; for a review see Bä ring and Covarrubias [24]). The latter principle also applies to IKCa and IKV in frog hair cells; it is indicated here with the more traditional note 'steadystate inactivation'. Contrary to OSI, however, the molecular basis of CSI is not fully understood. A conceptual model, originally proposed for spHCN channels, assumes that CSI may result from a temporary separation of the voltage sensor and the activation gate [25]. Analysis of these structural channel modifications are beyond the purpose of this paper, which remains mainly phenomenological. The novel observation, however, is that phosphorylation-mediated effects are able not only to influence the open state or the number of responsive potassium channels, but also to control, in a voltage-dependent manner, the availability of the closed channels to subsequent depolarizations. This is particularly evident in the case of the KT5823 treatment, which in some cases has canceled any voltage-dependent inactivation removal, resulting in an irreversible channel blockade, both at membrane potential values at which they are physiologically closed and in response to depolarization, which should move them to the open state. An important point that emerges from these experiments is that the mechanisms that control the amplitude of the evoked currents or the removal of inactivation are independent processes, differently affected by protein kinase manipulation. Were this not the case, the irc values would always mirror the maximal amplitude of the evoked responses.

A Physiological Role for the IKD Inactivation Mechanism?
The kinetics of onset and removal of IKD inactivation are in the order of milliseconds, so that any shift of the hair cell momentary potential (action of the efferent systems, modifications of the tonic transduction current, fluctuations of the ionic gradients…) that occurred during the ongoing cell activity would produce a corresponding adjustment of the largely variable KD conductance.
Voltage steps and sinusoidal voltage shifts simulating the ciliary excitatory/inhibitory deflections of the hair cell clearly illustrate the role of the single current components evoked by the transduction current. The transient role of IA is expected to contribute only in the presence of strong, rapid potential fluctuations, while the ongoing hair cell activity is presumably controlled by the IKCa and IKV components, which counteract the sensory transduction current in setting the momentary membrane potential of the cell. The calcium current does not exhibit any intrinsic regulatory mechanism, so that its amplitude, and the related cytoneural synaptic activity, is exclusively governed via its voltage dependence properties. Conversely, IKCa and IKV channels are controlled in a more complex way in frog hair cells. Their variable phosphorylation level, inhibited by H89 or by KT5823, might thus operate to continuously adjust the balance between the depolarizing transducer current and the repolarizing potassium current, thus setting the calcium inflow, and indirectly the transmitter release rate at the cytoneural junction, which ultimately generates the sensory information to be conducted to the central nervous system.