Pflügers Archiv

, Volume 450, Issue 1, pp 34–44

Regulation of the voltage-gated potassium channel KCNQ4 in the auditory pathway


  • J.-M. Chambard
    • Department of Physiology and Centre for Auditory ResearchUniversity College London
    • Department of Physiology and Centre for Auditory ResearchUniversity College London
Cell and Molecular Physiology

DOI: 10.1007/s00424-004-1366-2

Cite this article as:
Chambard, J. & Ashmore, J.F. Pflugers Arch - Eur J Physiol (2005) 450: 34. doi:10.1007/s00424-004-1366-2


The potassium channel KCNQ4, expressed in the mammalian cochlea, has been associated tentatively with an outer hair cell (OHC) potassium current, IK,n, a current distinguished by an activation curve shifted to exceptionally negative potentials. Using CHO cells as a mammalian expression system, we have examined the properties of KCNQ4 channels under different phosphorylation conditions. The expressed current showed the typical KCNQ4 voltage-dependence, with a voltage for half-maximal activation (V1/2) of −25 mV, and was blocked almost completely by 200 µM linopirdine. Application of 8-bromo-cAMP or the catalytic sub-unit of PKA shifted V1/2 by approximately −10 and −20 mV, respectively. Co-expression of KCNQ4 and prestin, the OHC motor protein, altered the voltage activation by a further −15 mV. Currents recorded with less than 1 nM Ca2+ in the pipette ran down slowly (12% over 5 min). Buffering the pipette Ca2+ to 100 nM increased the run-down rate sevenfold. Exogenous PKA in the pipette prevented the effect of elevated [Ca2+]i on run-down. Inhibition of the calcium binding proteins calmodulin or calcineurin by W-7 or cyclosporin A, respectively, also prevented the calcium-dependent rapid run-down. We suggest that KCNQ4 phosphorylation via PKA and coupling to a complex that may include prestin can lead to the negative activation and the negative resting potential found in adult OHCs.


CochleaHair cellsPotassium channelPhosphorylationPrestinCalcium activated binding proteins


KCNQ1–5, the five members of the family of voltage-gated potassium channels, share approximately 40% homology at the amino acid level [6]. KCNQ1 channels are implicated in the repolarisation of cardiac myocytes and are also involved in secretion of endolymph in the inner ear [26]. KCNQ2 and KCNQ3 are expressed exclusively in brain and sympathetic ganglia [43] and form heteromeric channels underlying the M-current [39]. KCNQ5 is the most recently identified channel and is widely distributed in the central and peripheral nervous systems [22].

KCNQ4 is expressed abundantly in the cochlea as well as in brain, heart and skeletal muscle [19]. Mutations in the gene for KCNQ4 underlie a non-syndromic hereditary hearing loss, DFNA2 [20]. The channel is expressed in both inner hair cells (IHCs) [24, 27] and outer hair cells (OHCs) [23]. KCNQ4 has been identified tentatively as the molecular correlate of an OHC potassium current, termed IK,n [17]. IK,n is distinguished by an activation curve which contributes to the large negative resting potential of OHCs. This activation does not match that of KCNQ4 found in expression systems [20]. The potential eliciting half-maximal activation (V1/2) seen in activation curves in OHCs is variable but, in general, very negative, −80 mV in guinea-pig [16] and −66 mV in mouse at post-natal day (P)12 [23]. In contrast, V1/2 for KCNQ4 in expression systems ranges from −10 mV in oocytes [20] to −32 mV in HEK-293 cells [36]. To investigate this discrepancy we examined the role of phosphorylation of KCNQ4 by PKA and found that phosphorylation shifts the activation curve to more negative potentials. A further negative shift is produced by co-expression of the OHC motor protein, prestin. Together these observations could explain some, if not all, of KCNQ4’s characteristics in OHCs.

To parallel findings on IK,n in OHCs and, in particular, that IK,n is sensitive to elevated intracellular calcium [18], we also describe the effects of Ca2+-dependent modulation of KCNQ4 currents via calmodulin (CaM) and calcineurin (CaN). The universal sensor CaM is a small protein with four EF-hand-type Ca2+-binding sites [30], and has been detected in hair cells [10]. CaM has been shown to be tethered constitutively to several ion channels such as the small-conductance Ca2+-activated potassium channel SK [42], the intermediate-conductance K+ channel IK [9] and the voltage-gated human ether-a-go-go potassium channel, EAG [31]. CaM also interacts with the intracellular C-terminal region of KCNQ channels [12, 41, 44]. CaN, sometimes known as protein phosphatase PP-2B, is a ubiquitous serine/threonine phosphatase [29] that is also found in the inner ear [21]. Both Ca2+ and CaM regulate the activity of CaN which acts through the control of de-phosphorylation of ion channels [3] and transcription factors [14]. We describe here the effect of a rise in [Ca2+]i on KCNQ4 currents and show that KCNQ4 phosphorylation by PKA prevents inactivation through the action of endogenous CaM or CaN.

Materials and methods

Cell culture and transfection

Chinese Hamster Ovary (CHO)-K1 cells were grown at 37 °C and in 5% CO2 in DMEM supplemented with 10% foetal calf serum. Cells plated in 35-mm plastic dishes were transfected using Lipofectamine (Life Technologies, Gaithersburg, Md., USA) according to the manufacturer’s recommendations. CHO cells were transfected where appropriate with plasmids containing cDNAs coding for human KCNQ4, KCNQ2, KCNQ3, pendrin and rat prestin and KCNA1. All plasmids contained cDNAs driven by the cytomegalovirus promoter. Co-expression of green fluorescent protein (pEGFP-N1, Clontech) was used as a reporter for successful transfection, and only cells that showed green fluorescence were chosen for electrophysiology. All chemicals were supplied by Sigma unless otherwise stated.


Whole-cell, tight-seal recordings were made at room temperature (20–22 °C) 1–3 days after transfection. Cells were superfused with a bath solution containing (in mM): NaCl 143, KCl 4, CaCl2 1, MgCl2 1.5, HEPES 5 and d-glucose 10, pH 7.2, osmolarity 315 mosmol/kg. Pipettes (3–4 MΩ) were filled with the following intracellular solution (in mM): KCl 145, MgCl2 1, Na2-ATP 1, EGTA 3, HEPES 5 and d-glucose 5. For pooled experiments investigating the effect of Ca2+ on KCNQ4 current, EGTA was increased to 5 mM and 2 mM CaCl2 was added to give a free [Ca2+] of 100 nM (calculated using WinmaxC software). Experiments were also carried out with 5 mM BAPTA to achieve rapid buffering and showed similar results. Mg2+ was occasionally increased to 1.5 mM and ATP to 3 mM with no discernable difference to the data described below. The pH was adjusted to 7.2 and the osmolarity 310 mosmol/kg. Currents were recorded using an appropriate amplifier (Axopatch 200B, Axon Instruments, Union City, Calif., USA) with 60–90% series resistance compensation and then digitised at 5 kHz. No corrections were made for the small liquid junction potentials (less than 4 mV).


Drugs were pressure applied from a pipette placed 50–100 µm from the recorded cell. Linopirdine (RBI, Dupont Merck) was applied at 20, 100 or 200 µM. 8-Br-cAMP (Sigma), a membrane-permeable cAMP analogue which activates PKA, was prepared from its sodium salt. The catalytic sub-unit of adenosine 3′,5′-monophosphate-dependent protein kinase (PKAc) was added to the internal solution at 50 U/ml. H-89 (N-{2-[(p-bromocinnamyl) amino]ethyl}-5-isoquinolinesulphoneamide dihydrochloride; Calbiochem), a potent membrane-permeable inhibitor of PKA [37], was dissolved as a stock solution in dimethylsulphoxide (DMSO) and added so that the final concentration of DMSO did not exceed 0.1%. W-7 (Sigma), a Ca2+/CaM antagonist, was applied at 100 µM. The Ca2+/CaM/CaN inhibitors Cyclosporin A (100 nM, Sigma) and calcineurin autoinhibitory peptide (corresponding to residues 457–482 of PP-2B) (100 µM, Calbiochem) were both tested. These two inhibitors are thought to act by blocking the binding of Ca2+/CaM with CaN and hence block the activation of the phosphatase.

Flash photolysis of caged compound

The tetrasodium salt of the caged Ca2+ compound nitr5 (Molecular Probes, Eugene, Ore., USA) was dissolved in EGTA-free internal solution and complexed with Ca2+. Following membrane rupture in the whole-cell recording mode, the cell contents were allowed to dialyse and reached an equilibrium with the pipette solution. Single cells were exposed to a 1-ms 350–400 nm light flash (Cairn Research, UK). Light was focussed directly onto the cell by projection through a fibre optic. As previously described [18], with 2 mM nitr5 (50% complexed with Ca2+) as the main buffer in the cell, free [Ca2+]i was calculated to be 20 nM before photolysis. Single flashes raised [Ca2+]i by 90 nM [18].


Cell stimulation, data acquisition and analysis were performed using standard software (PClamp, Axon Instruments). The threshold of the activation is defined below as the potential at which the slope of the outwards current/voltage (I/V) relationship at 0 mV extrapolated linearly to zero current. The voltage/activation curve was derived from standard analysis of tail currents on stepping back to a fixed potential of −20 mV. Activation curves were fitted with a single, first-order Boltzmann function:
$$ \phi {\left( {\text{V}} \right)} = \frac{I} {{I_{{{\text{max}}}} }} = \frac{{\text{1}}} {{{\text{1}} + e^{{{\text{ }} - \frac{{V_{{\text{m}}} {\text{ - }}V_{{{\text{1}}/{\text{2}}}} }} {k}}} }} $$
where I is the tail current, Imax the maximum tail current, Vm the membrane potential of the preceding step and k (in mv) is the slope factor. Data are given as means±SD.


Characteristics of KCNQ4 channel expressed in CHO cells

Functional expression of KCNQ4 in CHO produced a slowly activating outwards current (Fig. 1A) that was blocked by linopirdine. Linopirdine at 200 µM blocked currents by 89±4% (n=12, Fig. 1B); 100 and 20 µM linopirdine reduced the current by 69±8% (n=3) and 48±4% (n=3), respectively (data not shown). The expressed current, although quite variable from cell to cell, was 20–70 times larger than that in non-transfected cells. Indeed, native CHO cells expressed endogenous K+ current with an amplitude of no more than 100 pA at 0 mV (Fig. 1C), equivalent to a leak conductance of 1 nS. Linopirdine (200 µM) had no effect on this endogenous current. In KCNQ4-transfected CHO cells, the outwards current activated at −53.5±4.0 mV (n=35, Fig. 1D). An analysis of tail currents using Eq. 1 showed that the mean V1/2 was −27.1±5.7 mV and k 11.6±1.6 mV (n=35, Fig. 1E)
Fig. 1A–E

Transient expression of the human voltage-gated K+ channel KCNQ4 in Chinese hamster ovary (CHO) cells. A Current traces recorded from a CHO cell transfected with the cDNA encoding KCNQ4. Holding potential −60 mV. Currents were elicited by 2-s command steps from −100 to +30 mV in 10-mV increments, followed by a 1-s step to −20 mV. The dotted line indicates zero current. All data were obtained with this protocol unless otherwise stated. B Currents recorded from the cell in A after addition of 200 µM linopirdine. C Endogenous voltage-activated currents from a non-transfected CHO cell. D Current/voltage (I/V) relationship of all traces presented in A (filled squares), B (filled circles) and C (filled triangles). E Activation curve from cell in A (fitted by Eq. 1). Parameters as in text

The kinetics of KCNQ4 current, IKCNQ4, were best fitted using the sum of two exponential components:
$$ I_{{{\text{KCNQ4}}}} {\left( V \right)} = A_{{\text{f}}} {\left( {{\text{1}} - e^{{ - \frac{t} {{\tau _{f} }}}} } \right)} + A_{{\text{s}}} {\left( {{\text{1}} - e^{{ - \frac{t} {{\tau _{s} }}}} } \right)} $$
where the two components, a “fast” and a “slow” component, have amplitudes Af(V) and As(V) and exponential time constants τf(V) and τs(V) respectively. The fit using Eq. 2 yielded a considerably smaller chi-square than fitting with a single exponential component. Both time constants were voltage dependent, as found in HEK-293 cells [32]. Values for τf and τs at 0 mV were 72±20 and 365±48 ms (n=35), respectively. As shown in Fig. 2A, the relative amplitudes of the components Af(V) and As(V) varied with voltage. As(V) varied from 6.1±3.1% (at 0 mV) to 22.6±10.1% (at +30 mV) of the total current (n=10). The two kinetic components of the current therefore suggest that at least three states, two closed (C1, C2) and one open state (O), in a scheme represented by:
$$ {\text{C}}_{{\text{1}}} \underset{{k_{{{\text{21}}}} }}{\overset{{k_{{{\text{12}}}} }}{\rightleftarrows}}{\text{C}}_{{\text{2}}} \underset{{k_{{{\text{32}}}} }}{\overset{{k_{{{\text{23}}}} }}{\rightleftarrows}}{\text{O}} $$
are required to model KCNQ4 channel activation. This scheme predicts that the time evolution of the open state, O, and hence the macroscopic current, will be defined by the time constants τf and τs, which will be complex functions of the rate constants kij [40].
Fig. 2A–D

Action of 8-Br-cAMP on KCNQ4 expressed in CHO cells. Current traces recorded from a KCNQ4 transfected CHO cell before (A) and after (B) 120 s exposure to 500 µM 8-Br-cAMP. I/V (C) and activation (D) curves before (filled squares) and after (open squares) 120 s exposure to 8-Br-cAMP. The I/V curve corresponding to the trace before 8-Br-cAMP application has been decomposed (dotted line) to show the relative amplitudes of the two fitted components Af and As according to Eq. 2. The I/V curve obtained after 8-Br-cAMP application corresponds to a fast component only. The activation curve before and after application of 8-Br-cAMP was fitted by Eq. 1 with voltages for half-maximal activation of −24.8 and −39.0 mV and slopes of 10.9 and 10.7 mV, respectively

Modulation of KCNQ4 current by phosphorylation: 8-Br-cAMP

The OHC current IK,n is affected by its phosphorylation state and can be enhanced considerably by cAMP [18]. To follow the parallel, 500 µM 8-Br-cAMP was added to the bath, at which the KCNQ4 currents at +30 mV in CHO cells increased by 35±18% (n=8, Fig. 2A–C). 8-Br-cAMP also shifted the activation curve to more negative potentials. After exposure, V1/2 was −37.3±6.8 mV (n=8), a negative shift of 10 mV. The activation slope, k, 10.2±1.1 mV, was not significantly altered (Fig. 2D). Following application of 8-Br-cAMP, the kinetics of activation were best fitted with a single exponential component with time constant τ. The rise of the current at 0 mV was fitted by τ=71±19 ms (n=8), nearly identical to τf at that potential. The data, therefore, indicate that the kinetic component associated with the slower time constant, τs, was reduced greatly by 8-Br-cAMP. Except where stated, all subsequent currents consisted of essentially only one kinetic component, indicating that only one closed state and one open state dominated the kinetics of the phosphorylated channel. A similar apparent two-state behaviour has been described for IK,n in OHCs [16]. In this interpretation, phosphorylation moves the equilibrium in the inactivated state from between C1 and C2 towards C2.

As seen also in OHC recordings [17, 18], 8-Br-cAMP changed the kinetics of the current at the most hyperpolarised potentials. Figure 3A (+500 µM 8-Br-cAMP) shows that short commands to −160 mV de-activated the current with a time constant τ of 20 ms (mean 22±5 ms, n=3). With less hyperpolarising commands, τ increased progressively to a maximum of 140 ms at −50 mV (mean 145±44 ms, n=8). Figure 3B shows that the full voltage dependence of τ was described by a bell-shaped curve. This curve can be generated by a first-order Hodgkin-Huxley kinetic scheme as has been reported for IK,n. The same curve also fitted the data for τf obtained from KCNQ4 currents before addition of 8-Br-cAMP.
Fig. 3A,B

Action of 8-Br-cAMP on kinetics of the KCNQ4 current. A Effect of 500 µM 8-Br-cAMP using 200-ms commands. Holding potential −60 mV. Protocols as shown. B Exponential time constant, τ(V), of the current activation/deactivation (ordinate) to a command potential, V (abscissa). Open circles and squares are current rise times obtained after 8-Br-cAMP application from A and Fig. 2A, respectively. The smooth curve is a least-squares fit to all data given by:\( \tau {\left( V \right)} = {\text{15}}{\text{.4 }}\phi {\left( V \right)}{\text{exp}}{\left( { - V/{\text{28}}{\text{.2}}} \right)} \) where ϕ(V) is given by Eq. 1. This form is generated by a simple two-state (closed/open) Hodgkin-Huxley scheme with first-order kinetics (see e.g. [17]). Filled symbols correspond to the two time constants, τs (triangles) and τf (squares), obtained by fitting the KCNQ4 current in Fig. 2A. In the presence of 8-Br-cAMP (Fig. 2B), no slow time constant was apparent

The effects of 8-Br-cAMP were only partially reversible, even after 10 min wash-out. The simplest explanation is that 8-Br-cAMP was resistant to hydrolysis by the endogenous phosphodiesterases in the cell. However, 8-Br-cAMP had an effect on only 50% of cells tested, which may indicate a variable, but generally low, level of endogenous PKA in CHO cells. Only those cells showing a measurable effect were included for statistical analysis in this study.

Modulation of KCNQ4 current by intracellular PKA

To investigate the mechanism by which intracellular cAMP influenced KCNQ4 currents, PKAc was added via the pipette solution. Cell recordings made in this manner showed an accelerated activation with a rise time constant τ of 69±14 ms at 0 mV (n=6) almost identical to that obtained in cells recorded in presence of 8-Br-cAMP (Fig. 4A). The activation curves were shifted to more negative potentials (threshold, −64±4 mV, n=6), with a V1/2 of −49.3±9.2 mV and a k of 10.7±1.6 mV. Thus PKA induced a negative shift of 20 mV in the KCNQ4 activation curve.
Fig. 4A–D

Action of exogenous protein kinase A (PKA) on KCNQ4 channel expressed in CHO cells. A Currents recorded from a CHO cell transfected with KCNQ4 and with 5 U/ml PKA catalytic sub-unit (PKAc) in the pipette solution. B Same cell, currents after 5 min exposure to 1 µM H-89. I/V (C) and activation (D) curves before (filled circles) and after (open circles) application of H-89

To determine whether the effect was due to solely to PKA, H-89, a selective, membrane-permeable inhibitor was employed [4]. Application of 1 µM H-89 to a cell with PKAc in the pipette decreased the amplitude of the outward current (Fig. 4B). The maximum effect of H-89 on the activation curves produced by PKA occurred only after at least 7 min. Recording from cells with H-89 and PKAc was equivalent to having no PKAc present at all (Fig. 4C,D). At 1 µM, H-89 did not affect the activation slope (10.7±0.9 mV, n=6) but shifted V1/2 in the positive direction to −33.2±2.7 mV, a value close to that obtained without exogenous PKA (−27.1±4.7 mV, Fig. 1E). H-89 had no effect when PKA was absent from the pipette (n=4, data not shown).

Co-expression of SLC26A5 (prestin) and KCNQ4

OHC membranes contain large quantities of a motor protein, SLC26A5 (prestin, [47]). Figure 5A shows recordings from a CHO cell co-transfected with both KCNQ4 and prestin. Prestin had no significant effect on the rise times of the outwards current. Thus, at 0 mV, τf and τs were 90±8 and 397±60 ms and relative amplitudes 77.1±5.2% and 22.9±5.2% (n=12), respectively. Prestin, however, shifted the activation curves in the negative direction by about 15 mV. In such co-transfected cells, V1/2 was −47.8±5.2 mV with no change in k (9.9±1.4 mV, n=12).
Fig. 5A–F

Co-expression of prestin with channels of the KCNQ family. A Currents recorded from a CHO cell transfected with KCNQ4 and prestin. B Cell with a similar expression pattern recorded with 50 U/ml PKAc in the pipette solution. C Activation curves for A (filled squares) and B (open circles). D Currents recorded from a CHO cell transfected with both KCNQ2 and KCNQ3. E Current traces recorded from CHO cell transfected with KCNQ2, KCNQ3 and prestin. F Activation curves of the traces recorded in D (open squares) and E (filled circles)

When cells were co-transfected with KCNQ4 and prestin and recordings made with PKAc in the pipette (Fig. 5B), the rise time of the current at 0 mV was characterised by a single τ of 68±9 ms (n=5) and was not significantly different from that determined with PKAc alone. The voltage dependence of this time constant remained unchanged and could be fitted in a manner similar to that seen in Fig. 4B. There was however a significant (P<0.01) negative shift in V1/2 to −57.4±2.7 mV (Fig. 5C) compared with control KCNQ4 values with no prestin but with PKAc in the pipette. There was no change in k (10.8±0.8 mV, n=5).

The effect on the activation curve was specific to prestin. Co-expression of KCNQ4 and another member of the SLC26A transporter family, pendrin (SLC26A4), had no effect. In eight cells co-transfected with both KCNQ4 and pendrin, activation curves were not significantly different from control cells expressing KCNQ4 alone (V1/2 −31.6±4.6 mV, k 10.2±1.4 mV).

To determine whether prestin only affected KCNQ4, its effect on other members of the KCNQ family was examined. Co-expression of KCNQ2 and KCNQ3 recapitulate heteromeric KCNQ2/3 channels with the properties of the M-current in sympathetic neurones [39] and thus cells were co-transfected using cDNAs coding for KCNQ2+KCNQ3, with and without prestin (Fig. 5D,E). Without prestin, V1/2 was −31.4±1.6 mV and k 14.4±1.2 mV (n=3), consistent with the known properties of the M-current in CHO cells [34]. Co-expression with prestin produced a negative shift of 15 mV (Fig. 5F, V1/2 −44.3±2.5 mV, n=4). As with KCNQ4, there was no effect of prestin on the activation time constant or on the slope of the activation curve.

To determine whether prestin affected only the KCNQ family, prestin was co-expressed with a potassium channel from the Shaker family, KCNA1. The outwards currents in transfected CHO cells were tetraethylammonium (TEA) sensitive, thus verifying that the channel was being correctly expressed in these experiments [15]. These KCNA1 currents were used to construct activation curves using standard protocols [38]. Co-expression of KCNA1 with prestin produced a small negative shift in the KCNA1 activation from a V1/2 of −9.4±1.0 mV in control to −14.5±3.4 mV (n=5) in cells co-transfected with prestin (data not shown). This negative shift in the activation curve was significantly less than found for the activation curves when KCNQ family channels were expressed.

Regulation of KCNQ4 by intracellular calcium

In addition to examining the control of KCNQ4 channels through phosphorylation, we investigated the effects of intracellular calcium on the expressed current. All the recordings described above were made using internal solutions with an estimated [Ca2+] of 1 nM, nominally zero Ca2+. To ensure that Ca2+ was fully buffered, experiments were carried out with both 5 mM BAPTA and 3 mM EGTA with no significant differences between results. To minimise the effects of localised Ca2+ influx, the experiments were also repeated in a nominally Ca2+-free external bath, a manipulation that produced no significant differences in the responses. Nevertheless, in all cells tested, currents recorded with zero Ca2+ in the pipette showed a small, slow decline (run-down) of 2.4±1% min−1 (n=26) over 5 min (Fig. 6A).
Fig. 6A–D

KCNQ4 channel and internal calcium. Currents recorded from a CHO cell transfected with KCNQ4 with intracellular [Ca2+] (A) 1 or (B) 100 nM. In each case, a second recording was made after 5 min. Linopirdine (200 µM) was then applied to block the residual KCNQ4 current. C Relative KCNQ4 current was measured at −20 mV with intracellular solution containing zero Ca2+ (open circles, n=26), 100 nM Ca2+ (filled circles, n=11) or 2 mM nitr5 (filled triangles, n=6). UV flash was applied 90 s after patch rupture. Release of Ca2+ decreased outwards current and produced a decline similar to that produced by 100 nM Ca2+ in the pipette. D Histogram showing the relative whole-cell tail current recorded at −20 mV with (white bars) or without (black bars) Ca2+ in the pipette and following application of different compounds as shown. +P<0.0001 vs. control with similar pipette solution; *P<0.0001 (PKAc in pipette solution) or P<0.001 (cyclosporin A in pipette solution) vs. control with similar pipette solution

Increasing the pipette Ca2+ to 100 nM, with or without ATP, accelerated the decline of current amplitude by a factor of 7–8: it declined by over 90% of initial control within a 5-min period (Fig. 6B–D). During this run-down period the kinetics of the KCNQ4 current did not change. They were still well fitted by a bi-exponential function. Measured after 4 min, the activation curves likewise were not altered significantly (V1/2=−29.4±6.7 mV; k=10.2±1.9 mV, n=11). The effect on run-down was Ca2+-dependent. It could not be reproduced with high concentrations of Ba2+ (2.5 mM, n=3), Mg2+ (3 mM, n=3) or Zn2+ (200 µM, n=2) in the pipette (Fig. 6D). In all recordings with high intracellular calcium in the pipette, 200 µM linopirdine still blocked over 80% of the total current (Fig. 6A,B,D).

The rate at which [Ca2+]i acted on KCNQ4 was investigated using the caged Ca2+ compound nitr5 as the Ca2+ source. To allow for time for cell loading, the UV flash was applied 90 s after patch rupture. In control cells loaded with nitr5 alone, currents were not affected (data not shown) and showed only slight run-down. Increasing [Ca2+]i by photolysis (estimated to result in a 100 nM free Ca2+ load) immediately started to decrease the outwards current. The subsequent run-down was similar to that produced by 100 nM Ca2+ in the pipette (Fig. 6C,D).

Including 50 U/ml exogenous PKA with 100 nM Ca2+ in the pipette surprisingly prevented rapid run-down (Fig. 6D). In the six cells studied, there was, however, a negative shift of 20±2.3 mV in V1/2, as well as a reduction in the slower time constant, τs. These responses were the same as those obtained by including PKAc only in the pipette. Thus current activation parameters and current amplitude can be controlled separately. The observations suggest that the phosphorylation site is distinct from the site(s) where Ca2+ exerts its action.

Modulation of KCNQ4 current by Ca2+ binding proteins

To examine the mechanisms by which [Ca2+]i could be having an effect, we studied the possible involvement of calmodulin, CaM, on the run-down by adding the CaM antagonist W-7 (100 µM) to the bath solution. With <1 nM Ca2+ in the pipette, W-7 increased KCNQ4 currents slightly (by less than 3% at +30 mV, n=3). W-7 produced a negligible negative shift in activation (−4.5 mV from control) and, over 5 min, the compound did not affect the (normal) slow run-down. Nevertheless, W7 substantially reduced the rapid current run-down normally observed in presence of high [Ca2+]i. With inclusion of 100 nM Ca2+ in the pipette, after 5 min, 47.8±21.9% (n=5) of the KCNQ4 current remained (Fig. 6D). The effects of W-7 showed varied greatly, demonstrating either the difficulty of access of the compound or differences in the CaM expression from cell to cell.

Current run-down was also affected by cyclosporin A, an antagonist of CaN. Cyclosporin A (100 nM) in the bath and 100 nM free Ca2+ in the pipette prevented current run-down. Over 5 min, the current fell by only 35±6% of control (n=6, Fig. 6D). When applied intracellularly, 100 µM calcineurin autoinhibitory peptide produced a similar moderate reduction of the current run-down (to 39±7% of control after 5 min, n=6; data not shown). No significant shift of the activation occurred during run-down. The results show that PP-2B is involved in the basal modulation of KCNQ4 and also indicate that, while inhibition of CaN reduced rapid run-down, it did not alter the current’s activation.


Modulation of KCNQ4 current by phosphorylation

The present results suggest that phosphorylation by PKA is a significant determinant of how KCNQ4 is modified to contribute to the native IK,n found in sensory hair cells. In the mammalian cochlea, there is electrophysiological evidence that IK,n is present in both classes of hair cell [16, 27], although both immunohistochemical and in situ hybridization studies show a considerably stronger signal in OHCs.

Other KCNQ channels are known to be modulated by phosphorylation. Both the KCNQ1/KCNE1 complex [25] and the M-current KCNQ2/3 [33] can be regulated by PKA, with src tyrosine kinase regulating most KCNQ channels [13]. The sequence for KCNQ4 suggests that these potential phosphorylation sites lie on the C-terminal region (as defined by the Prosite database). For the OHC it has been already been proposed that IK,n is itself under control of a background phosphorylation through PKA [18] and background de-phosphorylation by a Ca2+-dependent protein phosphatase, probably PP-2A (based on the sensitivity of the current to okadaic acid). Phosphorylation of an analogous current, termed IK,L in type-I vestibular cells, seems likely to be the target of similar modulation [5]. Although in OHCs the PKA inhibitor H-89 reduces OHC currents [18], the present lack of effect of H-89 on CHO cells transfected with KCNQ4 may indicate simply that there is only a low level of PKA activity in this particular cell line.

Result of co-expression of SLC26A5 (prestin) and KCNQ4 in an expression system

Phosphorylation can account for a significant fraction of the disparity between KCNQ4 and IK,n activation in OHCs but the presence of SLC26A5 (prestin) in mature OHCs may also contribute an additional negative shift (from a V1/2 of −50 mV for the fully phosphorylated KCNQ4 channel to −65 mV in adult OHCs). Indeed, the time course of prestin and KCNQ4 expression during hair cell development run in parallel [1, 23]. The first appearance of IK,n in the mouse OHC is at P8 and the development of prestin immunoreactivity in rat OHCs is most prominent at P6 and P9. Nevertheless, immunohistochemistry studies show that the major KCNQ4 and prestin signals do not co-localise on OHC membrane, nor is there even established evidence for significant levels of prestin in IHCs. We propose that the discrepancy arises because of the relatively slight quantities of prestin required. Prestin was discovered using a subtractive DNA library strategy [47] by taking advantage of the enhanced message in OHCs. This does not preclude much lower levels of expression in IHCs which may have functional consequences for the small number of KCNQ4 channels at most in such hair cells. Further, the exceptionally high density of prestin in the OHC lateral membrane may obscure signals elsewhere in the cell and electrophysiological measurement of OHC charge movements in patches are compatible with a low level of expression throughout the OHC basolateral membrane [11].

The simplest hypothesis is that prestin exerts its effects on KCNQ4 by changing the membrane surface charge. Clusters of positive charge have been described on the cytoplasmic domains of prestin [27] and could produce the shift of −15 mV in activation if they were close to the channel sensor microdomain. The much larger shift produced in KCNQ currents than in KCNA suggest that there may be protein-specific effects as well. Despite these simple proposals, there still remain differences in drug sensitivities between IK,n and KCNQ4. The linopirdine concentration eliciting half maximal inhibition (IC50) of IK,n in both IHCs and OHCs is in the micromolar range (IHC 1.1 µM [27]; OHC 0.7 µM [23]), whilst the IC50 for expressed KCNQ4 is an order of magnitude higher (in HEK-293 cells, 14 µM [36], and in CHO cells our data are compatible with an IC50 of 28 µM). This suggests that the linopirdine binding site may be modified by any additional interactions contributing to IK,n.

The evidence thus supports the idea that IK,n is a membrane complex with KCNQ4 as a subunit [23]. It is known that PKA catalytic and regulatory sub-units, PP1 and yotiao, are components of the KCNQ1/KCNE1 macromolecular complex [25]. It is therefore possible that KCNQ4 could be part of a similar complex in OHCs and IHCs involving kinase, phosphatase and a targeting protein, possibly prestin and/or a protein yet to be described. If the formation of such a complex were sensitive to surface charge, differences in the biophysical properties of KCNQ4 observed between the different expression systems may well be apparent.

Modulation of KCNQ4 current by calcium

The results show that intracellular Ca2+ controls the amplitude and stability of the KCNQ4 current. The results also explain the long-standing observation that elevated calcium in OHCs leads to the apparent loss of IK,n (although in hair cells the activation of a non-selective cation channel at the same time experimentally often masks the effect) [18]. Functionally, the effect of Ca2+ would be more prominent in hair cells from the basal (high-frequency) end of the cochlea, where functional expression of IK,n is greatest. We note that this calcium pathway may represent part of a calcium-controlled switch between hair cell membrane potential controlled by IK,n and control by distinct small K channels regulated by the cholinergic efferent system. Such a mechanism would enhance calcium entry induced by the action of the efferent fibres which terminate on OHCs.

The present data also demonstrate that raised cytoplasmic Ca2+ leads to a rapid run-down of expressed KCNQ4 currents. The most economical explanation is that the calcium binding proteins CaM and CaN, when activated by Ca2+, interact with KCNQ4 in the membrane and lead to channel inactivation. CaM is an ubiquitous Ca2+ binding protein that controls many cellular events including the activation of several proteins, enzymes and ion channels [45]. It is certainly known to be present in OHCs [10]. CaM interacts with members of the KCNQ family [12, 41, 44] binding to an IQ domain motif on the protein [28], either controlling the tetrameric assembly into the membrane [41,44] or by direct binding [12] and conferring Ca2+ sensitivity. It is unresolved whether the Ca2+/CaM complex or the Ca2+-free apoCaM form binds to this sequence [8, 41]. The simplest model here compatible with the data is that Ca2+/CaM both binds to a site on the channel and to a site on CaN to activate the phosphatase.

Interaction between PKA and calcium

The data also show that PKA can reduce the effects of elevated Ca2+. This result suggests that whereas phosphorylation/de-phosphorylation has an effect on the voltage dependence of the channel, the Ca2+ binding proteins are likely to act at a site distinct from the phosphorylation sites. Although direct binding of CaM is proposed here, there are two qualifications. First, application of a CaM antagonist should, in principle, reveal the protein’s role. Although the data are consistent with binding of the Ca2+/CaM complex promoting run-down, the CaM antagonist, W-7, also blocks the activity of the endogenous phosphodiesterases. Any inhibition of KCNQ4 by CaM would therefore be masked partially by the rise of cytoplasmic cyclic AMP, leading to an increase in KCNQ4 currents mediated by PKA. The data (Fig. 6) suggest that this effect is small. Second, calcium could act through a regulatory pathway which involves phosphoinostide bisphophate (PIP2). Although a minor component of the membrane phospholipids (less than 1%), PIP2 is now thought to play an important role in a wide variety of cellular processes [7]. PIP2 has been proposed as a direct activator of KCNQ channels [46]; chelation of PIP2, either by polylysine or PIP2 antibody, abolishes the currents. PIP2 is degraded when the level of [Ca2+]i rises. In the absence of ATP this proteolysis is non-reversible [2]. As an alternative explanation for the results it could be that PIP2 is degraded by elevated cytosolic Ca2+ and it is through this mechanism, rather than through direct gating of the expressed KCNQ4 channels, that Ca2+ also has its effect. In preliminary experiments PIP2 introduced through the pipette was not effective in reducing KCNQ4 currents (data not shown). This pathway is therefore likely not to be as significant as the Ca2+/CaM/CaN modulation pathway we propose here.

The sensitivity of current run-down to cyclosporin A (Fig. 6) does suggest that the Ca2+/CaM-activated phosphatase, CaN, is one of the components of a complex forming IK,n. CaN has been identified in both mammalian hair cells [21] and in CHO cells [35]. Our results therefore provide a simple explanation for the observation that IK,n in OHCs and KCNQ4 expressed in diverse cell systems have different biophysical properties [20, 24, 36]. They also suggest that the IK,n current is the target of multiple regulatory steps additional components of which still need to be identified.


We thank A. Tinker for the gift of the KCNQ4 clone and advice, DA. Brown for providing the KCNQ2 and KCNQ3 clones, B. Fakler for the rat prestin clone and M. Stocker for the KCNA1 clone. We thank H. Dorricott and R. Louvel for experimental assistance and J.E. Gale and D.J. Jagger for helpful discussion and critical review of the manuscript. This work was supported by the Wellcome Trust.

Copyright information

© Springer-Verlag  2005