CDK5 interacts with Slo and affects its surface expression and kinetics through direct phosphorylation

Abstract

Large-conductance calcium-activated potassium (BK) channels are ubiquitous and play an important role in a number of diseases. In hair cells of the ear, they play a critical role in electrical tuning, a mechanism of frequency discrimination. These channels show variable kinetics and expression along the tonotopic axis. Although the molecular underpinnings to its function in hair cells are poorly understood, it is established that BK channels consist of a pore-forming α-subunit (Slo) and a number of accessory subunits. Here we identify CDK5, a member of the cyclin-dependent kinase family, as an interacting partner of Slo. We show CDK5 to be present in hair cells and expressed in high concentrations in the cuticular plate and in the circumferential zone. In human embryonic kidney cells, we show that CDK5 inhibits surface expression of Slo by direct phosphorylation of Slo. Similarly, we note that CDK5 affects Slo voltage activation and deactivation kinetics, by a direct phosphorylation of T847. Taken together with its increasing expression along the tonotopic axis, these data suggest that CDK5 likely plays a critical role in electrical tuning and surface expression of Slo in hair cells.

large-conductance calcium-activated potassium (BK) channels are ubiquitous and important in human diseases, including deafness, hypertension, epilepsy, and movement disorders (13, 14, 16, 38, 45, 50, 52, 56). In chickens and other nonmammalian vertebrates, BK channels in hair cells play a critical role in electrical resonance, a mechanism by which frequency selectivity is established (17). Indeed, hair cells in these animals serve as a model system for their study, since these channels exhibit tonotopically determined changes in kinetics. BK channels are also colocalized with voltage-gated Ca²⁺ channels (VGCC) at the basolateral aspect of hair cells and show tonotopically determined changes in expression on the surface of these cells (17).

BK channels in hair cells of these nonmammalian vertebrates show variable kinetics that correlates with the tonotopic axis (15). Thus hair cells at the high-frequency end of the papilla have shorter deactivation times and mean open times (the latter observation was noted in the turtle) (3, 4). Current recordings and immunolocalization studies have shown an increasing number of Ca²⁺ channels and BK channels in higher frequency hair cells (2, 4, 57). The exact molecular mechanisms that underlie these changes in channel kinetics and increased channel numbers on the surface of hair cells along the tonotopic axis are as yet indeterminate.

BK channels contain a pore-forming α-subunit encoded by the Slo gene, which shows extensive alternative splicing (8). In addition, a number of interacting proteins, including the best characterized β-subunits 1–4, have been identified (11, 26, 27, 60, 62). While alternative splicing and association with the β-subunits have been implicated in both of these cellular processes, it is clear that these mechanisms alone cannot explain the changing kinetics of the channel or the increasing number of these channels along the tonotopic axis (17, 28, 29, 44, 53–55). Thus, while a wide range of alternatively spliced variants of Slo are tonotopically expressed, these variants are unable to replicate the range of kinetic properties in native hair cells (17, 28–30, 43, 54, 55).

The β₁- and β₄-subunits are expressed in increasing amounts in low-frequency hair cells (5, 39, 53). The β₄-subunit changes the kinetic properties of these channels, allowing them to respond to a rise in intracellular Ca²⁺ with minimal changes in the resting membrane potential (5). Additionally, the β₁-subunit reduces the wide variation in Ca²⁺ sensitivities of individual splice variants to a more uniform one (5, 39, 53), a feature of native hair cells. However, the deactivation times induced by these two β-subunits exceed that seen in native hair cells by orders of magnitude (5, 39, 53). The β₁- and β₄-subunits also prevent surface expression of Slo, including in hair cells (5, 39, 53). However, the extent to which these subunits affect the gradient in surface expression of Slo along the tonotopic axis remains unclear. We hypothesize that additional mechanisms, including association with other proteins and modulation by kinases, contribute to the gradients in surface expression and kinetic properties of these channels.

In an attempt to identify other protein subunits that interact with the Slo protein, we performed a yeast two-hybrid screen using the intracellular COOH-terminus of Slo as bait. We screened a chick cochlea cDNA library and identified the proline directed serine/threonine protein kinase cyclin-dependent kinase 5 (CDK5) as interacting with Slo. In this paper, we characterize the effects of CDK5 phosphorylation on Slo surface expression and kinetics.

MATERIALS AND METHODS

Yeast two-hybrid.

Yeast two-hybrid experiments were done as previously described (42). Briefly, the COOH-terminus of Slo (amino acids 491–744) was subcloned into pGBKT7 and used as bait to interrogate a chick cochlea cDNA library in pGAD T7 (Clontech). AH109 cells were serially transfected with both constructs and plated on drop-out medium lacking histidine, adenine, tryptophan, and leucine. Single colonies were isolated and serially replated in similar conditions on plates containing X-Gal. Cultures from single isolated colonies were used for yeast mini preps. Plasmid was electroporated into DH101 E. coli, grown in ampicillin and plasmid DNA isolated for sequencing.

Immunofluorescence detection of CDK5 and Slo.

Immunofluorescence detection, including quantitative immunofluorescence, was done as previously described (7), with modifications. We used an LSM 510 meta confocal microscope (Zeiss) to obtain images and used the attendant Zeiss LSM software to analyze and extract data. Chickens were euthanized by CO₂ asphyxiation, and their cochlea were dissected. For whole mount experiments, the tectorial membrane was removed (58) and fixed in 4% paraformaldehyde, PBS for 20 min. For experiments using cross sections, cochleas were embedded in paraffin, deparaffinated, and antigen retrieved by boiling for 10 min in 1 mm Tris·Cl 0.1 mM EDTA, pH 7.5, before antigen detection. The cochlea were washed in PBS (×3) and placed in blocking solution (PBS, 1% BSA, 5% horse serum, 0.1% Tween 20). Tissue was incubated in 1:1,000 anti-CDK-5 (C-8) antibody (Santa-Cruz) in blocking solution overnight at 4°C. After washing in PBS, 0.1% Tween 20 (×3), the tissue was incubated with Alexa 546 (or Alexa 488, where actin was detected with Phalloidin-Alexa 546) conjugated anti-rabbit antibody (1:1,000) for 1 h at room temperature (RT). The tissue was washed again in wash buffer (×3) and incubated with a 1:50 dilution of mouse anti-Slo antibody (BD/Transduction Laboratories) for 1 h at RT. Alexa 647 goat anti-mouse antibody (1:1,000) was added after the tissue was washed. The tissue was mounted in Vectashield (Vector) and viewed using a Zeiss 510 meta confocal microscope. Sixteen-bit images were acquired using a ×64 water immersion lens (numerical aperture 1.2), with fixed laser settings, a scan rate of 1.6 μs per pixel, a pinhole aperture of 1.0 Airy units, and fixed detector gain. Regions of interest (ROIs) of tall hair cells from fixed distances from the apical end of the cochlea were identified, and fluorescence data extracted. We established that the fluorescence intensity was within the linear range and used mean fluorescence density as a measure of protein concentration. Surrounding supporting cells, where there is minimal Slo and CDK-5 expression, were used to subtract background fluorescence. Cells from three individual cochlea were used for these analyses. The specificity of the antibodies was established by Western blots of cell lysates from cells with the respective constructs. Both antibodies identified bands of the expected size with minimal additional bands.

Coimmunoprecipitation.

Reciprocal immunoprecipitation was done with anti-Flag (FLAG-cSlo) antibody and anti-CDK5 (CDK5) antibody using immunoprecipitation kits purchased from Sigma (St. Louis, MO). In brief, membrane-enriched protein lysates were generated 48 h after transfection with CDK5 into a stable cell line expressing FLAG-cSlo-YFP (yellow fluorescent protein). Sixty-microliter anti-Flag M2-agarose or 60-μl anti-CDK5-agarose were added to the lysates. The mixtures were incubated with constant agitation for 2 h at 4°C. Following this binding step, the beads were washed three times with washing buffer. To elute the proteins of interest, 100 μl of loading buffer were added to the resin. The samples were incubated with constant agitation for 30 min at 4°C and then eluted by centrifugation. The eluted samples were boiled and then were separated on a precast 4 ∼ 15% Tris·Cl SDS-PAGE gel (Bio-Rad, Hercules, CA). Proteins were transferred by wet transfer to polyvinylidene fluoride membrane (Roche, Indianapolis, IN). Western blots were probed with anti-FLAG antibody (M2, 1:1,000), followed by goat anti-mouse antibody-peroxidase conjugate (1:5,000) or anti-CDK5 (1:1,000) followed by goat anti-rabbit antibody-peroxidase conjugate (1:2,500) (Sigma) for 1 h at RT. Immunoreactive proteins were detected using SuperSignal West Dura Extended Duration Substrate (Thermoscientific/Pierce, Rockford, IL).

Fluorescence resonance energy transfer.

Human embryonic kidney (HEK) cells were transfected with HSlo-YFP and CDK5-CFP [or cyan fluorescent protein (CFP)], and fluorescence resonance energy transfer (FRET) was evaluated 48 h later. For these experiments, cells were fixed in ice-cold acetone, rehydrated, and mounted in Vectashield. Sixteen-bit images were acquired with a Zeiss 510 confocal microscope using a ×64 water immersion lens (numerical aperture 1.2), with fixed laser settings, a scan rate of 1.6 μs per pixel, a pinhole aperture of 1.0 Airy units, and fixed detector gain. ROIs were demarcated, and emission spectra from each of these areas were determined using a Zeiss LSM 510 meta detector, while exciting at 458 nm and separately at 514 nm. Photobleaching was done at 514 nm with the excitation of the 40-W argon laser set at maximum until the YFP emission decreased 50% (∼20 s). To minimize photo damage and bleaching of adjacent sections in the z-axis, the aperture was kept at 1 Airy unit. Immediately after bleaching, the emission spectra from the ROIs were again recorded, while exciting at 458 nm. The amounts of emission in the CFP window (474–495) before and after photobleaching were quantified using the LSM 510 software and used to calculate FRET efficiency. The percentile efficiency of FRET was defined as:

$$\text{Fret efficiency}\left(\text{percentile}\right)=\left(\text{Ea}-\text{Eb}\right)/\text{Ea}\times 100$$

where Ea and Eb are the emissions of CFP (at 474–494 nm) after and before photobleaching, respectively (31).

Cell culture, transfection, and stable cell line generation.

HEK-293 cells (American Type Culture Collection ATCC, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium (DMEM, high glucose) containing 10% fetal bovine serum and 50 U/ml each of penicillin and streptomycin at 37°C with 5% CO₂. Constructs of HSlo or phosphorylation site mutations of HSlo were subcloned into pcDNA3 and transfected into these cells using lipofectamine (Invitrogen), according to the manufacturer's instructions. To produce stable cell lines, transfected cells were first selected in the presence of 1.2 mg/ml G418 (Clontech) in DMEM-based medium. Stable polyclonal cell lines were then cultured in DMEM-based medium containing 0.6 mg/ml G418.

Oocyte preparation and RNA injection.

The preparation of stage IV Xenopus laevis oocytes for RNA injection was done as previously described (5). All experimental procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee, Yale University School of Medicine.

For RNA injection, chicken Slo (cSlo) human CDK5 and P35 cDNAs were inserted into the pGH 19 vector. Plasmids were linearized with NotI and cRNA was transcribed using the T7 mMESSAGE mACHINE kit (Ambion, Austin, TX). RNA purification was done as previously described (5).

RNA in a volume of ∼46 nl was injected into each oocyte on the day after oocyte preparation. cSlo RNA at a final concentration of 0.3 mg/ml was injected with or without CDK5/CDK5 and P35 RNA. We generally injected cSlo and CDK5/CDK5 and P35 at 1 to 3 ratios by weight to maximize the effects of CDK5/p35. Recordings were performed 2–4 days after injection.

Electrophysiological recording.

Macroscopic currents were recorded in inside-out patches from injected oocytes or polyconal stable lines using the voltage-clamp technique (22) with an Axon 200B amplifier (Axon Instruments, Sunnyvale, CA). All experiments were performed at RT. Command delivery and data collections were carried out with a Windows-based whole cell voltage-clamp program, jClamp (Scisoft, Ridgefield, CT), using a Digidata 1322A interface (Axon Instruments). Currents were digitized at 100 kHz and filtered at 5–10 kHz. Data analysis was accomplished with jClamp or Clampfit (Molecular Devices, Union City, CA). Capacitance and leak currents were subtracted with a P/−5 leak subtraction protocol. Conductance-voltage (G-V) curves were obtained by measuring the amplitude of tail currents 100–200 μs after repolarization to −100 mV from the various test voltages. Each G-V curve was fitted with a Boltzmann function:

$$G\left(V\right)=\frac{G_{\max }}{1+\exp \frac{-zF\left(V-V_{\text{h}}\right)}{R\text{T}}}$$ 1

where G_max is the fitted value for maximal conductance, V_h is the voltage of half-maximal activation of conductance, z reflects the net charge moved across the membrane during the transition from the closed to the open state, F is the Faraday constant, R is the gas constant, and T is temperature.

To characterize relaxation kinetics, tail currents were collected at various step voltages after a 100-ms depolarization to 180 mV, to maximize the channel opening at specific Ca²⁺ concentrations. The time constants were obtained by fitting the tail currents to a single exponential function.

Recording pipettes were pulled from thin-walled borosilicate glass (TW 150, World Precision Instruments, Sarasota, FL) with impedance of ∼1–2 MΩ. The standard pipette/extracellular solution contained (in mM) 140 potassium-methanesulfonate, 20 KOH, 10 HEPES, and 2 MgCl₂, pH 7.2. The composition of the bath/intracellular solution was (in mM) 140 potassium-methanesulfonate, 20 KOH, 5 mM HEDTA, and 10 HEPES, pH 7.2, and calcium-methanesulfonate₂ was added to reach the appropriate free Ca²⁺ concentration. No Ca²⁺ chelator was used in the solution containing 100 μM free Ca²⁺. The amount of total calcium-methanesulfonate₂ needed to obtain the desired free Ca²⁺ concentration was calculated with Max Chelator (6), which was downloaded from http://www.stanford.edu/∼cpatton/webmaxc.htm. Final free Ca²⁺ concentration was measured with a Ca²⁺ electrode (Thermo Electron, Beverly, MA). The bath/intracellular solutions were delivered with the ALA QMM micromanifold perfusion system (ALA Scientific Instrument, Westbury, NY).

FACS analysis.

Polyclonal stable cell lines of HSlo and its phosphorylation site mutants were used for FACS analysis. Since we failed to obtain stable lines for S659D of HSlo, we used transient transfection of HSlo or S655A/D into HEK cells for FACS analysis (48 h).

We determined surface expression of control and mutants by using a live staining method. Briefly, live cells were harvested and incubated with anti-Myc conjugated with Alexa 647 (Cell Signaling) in phosphate-buffered saline, 1% bovine serum albumin, at RT for 30 min. The cells were then fixed in 1% paraformaldehyde before analysis on a Facscalibur machine (Becton-Dickinson).

To control for changes in intracellular expression of Slo channels affecting surface expression levels, we also determined total expression of Slo by intracellular staining for Slo in the same batch of cells. In brief, harvested cells were fixed and permeabilized by perm/fix buffer (BD Biosciences) at 4°C for 15 min. The cells were then incubated with mouse anti-Slo-antibody directed against its intracellular COOH-terminus (catalog no.: 611248 BD Biosciences) at a concentration of 1 μg/ml in phosphate-buffered saline, 1% bovine serum albumin, at 4°C. The primary antibody was detected in turn with a secondary anti-mouse antibody conjugated to Alexa 647. The cells were fixed briefly in 1% paraformaldehyde before analysis. FACS analysis was carried out with FlowJo software (Tree Star, Ashland, OR), as previously described (41). Since instrument settings could vary between experiments, we normalized all data to surface-labeled HSlo expression and separately intracellular HSlo expression. Relative surface expression was determined by dividing the mean fluorescence intensity of surface-labeled Slo by the mean fluorescence intensity of total labeled Slo.

Data analysis.

All results are given as means ± SE. Where appropriate, ANOVA was used to test for significance in differences.

RESULTS

Yeast two-hybrid identifies CDK5 as a Slo interacting protein.

To identify proteins interacting with Slo, we performed a yeast two-hybrid experiment using the COOH-terminus of cSlo as bait. For these experiments, we used a region of Slo encoding its COOH-terminus from amino acids 306–1114 (NP_989555.1), which were divided into three segments (306–540, 540–745, and 746–1114). cDNA encoding each of these amino acid fragments were subcloned into pGBKT7. We then probed a chick cochlea cDNA library and sought interacting partners (Fig. 1). Double-transformed AH109 cells were plated on dropout media lacking adenine, histidine, tryptophan, and leucine. Three out of the 20 clones that we identified as binding partners of amino acids 746–1114 contained a partial sequence of the chick homolog of CDK5. All of these clones expressed α-galactosidase in the assay conditions, confirming the strength of their interaction with Slo. CDK5 was not identified as a binding partner of Slo among the clones that interacted with the two other sequences of Slo used as bait in the yeast two-hybrid screen. The nucleic acid sequence of these clones encoded the very COOH-terminal amino acids 261–292 of chick CDK5 (NP_001129258.1). Moreover, they all contained the identical nucleic acid sequence, suggesting that they were derived from the same parent clone. This region of the CDK5 mostly has an α-helical structure, with the very COOH-terminus forming a random coil (59). This region of CDK5 has not been identified as important in mediating its kinetic activity (59).

Fig. 1.

Slo interacts with cyclin-dependent kinase 5 (CDK5) in yeast two-hybrid experiments. The schematic figure shows the COOH-terminal region of Slo (top) used for the yeast two-hybrid experiments. The region extends from amino acids (AA) 746–1114 and includes the RCK domains 1 and 2 immediately after the transmembrane (Tm) domain. Additionally, a schematic of the CDK5 protein (bottom), including the region (AA 261–292) identified in the yeast two-hybrid experiment, is shown. The region of CDK5 includes its very COOH-terminus that has been determined to be largely an α-helix. As indicated, the interaction of these regions occurred in highly stringent conditions [−His/−Ade/+alpha galactosidase (GAL)/−Trp/−Leu].

Interactions between Slo and CDK5 are confirmed by reciprocal immunoprecipitation and FRET.

Since yeast two-hybrid experiments are known to give false positive results, we sought to confirm an interaction between Slo and CDK5 by other means. In the first instance, we performed reciprocal immunoprecipitations using a stable HEK cell line expressing cSlo fused to a FLAG epitope at its NH₂-terminus and YFP to its COOH-terminus. As shown in Fig. 2, we were able to detect CDK5 in immunoprecipitates of Slo (and vice versa).

Fig. 2.

Reciprocal immunoprecipitation experiments confirm an interaction between Slo and CDK5. CDK5 was immunoprecipitated from lysates of a stable cell line expressing FLAG-cSlo-YFP (yellow fluorescent protein) and CDK5 using the anti-CDK5 C-8 antibody. Slo was immunoprecipitated using the M2 anti-FLAG antibody. Immunoprecipitates were separated on SDS-PAGE gels, and the reciprocal protein was detected by Western blotting. A: imunoprecipitates of CDK5 contain Slo (large arrow) detected using anti-FLAG antibody (lane 1). Immunoprecipitates using rabbit serum served as a negative control (lane 2). The molecular mass marker sizes are indicated (lane 3). B: immunoprecipitates of Slo contain CDK5 (arrow) detected using the CDK5 antibody (lane 1). Negative control (immunoprecipitates using mouse serum, lane 2), positive control (CDK5 immunoprecipitate, lane 3), and protein molecular mass markers are also shown. Note that the exposure times were five times longer in lanes 1 and 2 compared with lane 3.

We also sought to demonstrate protein-protein interactions between these two proteins in vivo using FRET. For these experiments, we used acceptor photobleaching to confirm FRET. We used, for these experiments, the aforementioned stable cell line expressing FLAG-Slo-YFP. We transfected CDK5 that was fused to CFP (at the COOH-terminus of CDK5) into these cells. As shown in Fig. 3, acceptor photobleaching resulted in a significant increase in CFP fluorescence in cells expressing CDK5-CFP, but not in those expressing CFP. These data suggest that Slo and CDK5 are in close approximation (<50 nm) in vivo, with likely intermolecular interactions.

Fig. 3.

Fluorescence resonance energy transfer (FRET) confirms in vitro cellular interactions between Slo and CDK5. A–D: shown are cyan fluorescent protein (CFP) and YFP fluorescence detected with a LSM510 confocal with a meta detector of a stable cell line expressing Slo-YFP transfected with CDK5-CFP. CFP (A and B) and YFP (C and D) fluorescence before (A and C) and after (B and D) YFP photobleaching are shown. As evident, there is an increase in CFP fluorescence that accompanies a decrease in YFP fluorescence after YFP photobleaching. E: shown are the FRET efficiencies of cells expressing Slo-YFP CDK5-CFP and separately Slo-YFP and CFP. An increase in FRET efficiency after photobleaching is observed in cells expressing Slo-YFP and CDK5-CFP (2.2 ± 0.3 SE). In contrast, FRET efficiency decreases after photobleaching in cells expressing Slo-YFP and CFP (−5.2 ± 1.3 SE). N = 12 in SloYFP/CDK5CFP and 10 in SloYFP/CFP.

CDK5 is expressed in hair cells and shows a tonotopic gradient in expression.

The presence of CDK5 in hair cells was suggested by the detection of this protein in our yeast two-hybrid screen using a cDNA library derived from chick cochlea. To more specifically ascertain the localization of CDK5, we used immunolabeling of this protein in whole mount cochlea preparations. As shown in Fig. 4, CDK5 is expressed almost exclusively in hair cells and excluded from the supporting cell layer. Within individual hair cells, CDK5 was expressed throughout the cytoplasm, but was concentrated at the cuticular plate and circumferential zone where cytoskeletal elements are particularly enriched (40, 61). These data suggest that CDK5 has multiple roles within hair cells, including modulating hair cell cytoskeletal elements. Immunostaining of paraffin-embedded cross sections revealed CDK5 to be expressed in higher levels in tall hair cells (Fig. 5). Moreover, we also noted an increase in the expression of this protein along the tonotopic axis (Fig. 6). Tall hair cells from a more basal region (2.5 mm from the apex) had an almost twofold increase in CDK5 compared with tall hair cells from a more apical location (1 mm from the apex). The increase in protein expression also parallels expression of mRNA detected by quantitative PCR (data not shown).

Fig. 4.

CDK5 is expressed in hair cells. A–C: shown are fluorescent confocal images of the basilar papilla viewed end-on and stained with phalloidin (A) and anti-CDK5 antibody (B). C: the pseudo-colored merged images (red actin, green CDK5). As evident, CDK5 is present in the upper one-half of the epithelium containing hair cells and largely absent from the lower one-half of the epithelium containing supporting cells. D–G: shown are fluorescent confocal images of hair cells stained with phalloidin (D), anti-CDK5 (E), anti-Slo (F), and the merged pseudo-colored image (G). The arrow shows a cluster of Slo channels (red) at the basolateral aspect of the cell. CDK5 (green) is expressed globally through the cytoplasm. Actin is pseudo-colored purple. Scale bar = 5 μm.

Fig. 5.

CDK5 in hair cells is enriched in areas rich in cytoskeletal elements. A–C: shown are fluorescent confocal images of hair cells in the basilar papilla viewed end-on and stained with phalloidin (A) and anti-CDK5 (B), with the merged images in C (red actin, green CDK5). CDK5 is expressed in high levels in stereocillia (large arrows) and the cuticular plate (small arrow). D–F: shown are fluorescent confocal images of hair cells viewed from above and stained with actin (D) and anti-CDK5 (E), with the merged pseudo-colored images in F (red actin, green CDK5). As evident, CDK5 is expressed in the cuticular plate (smaller arrow) and the circumferential zone (larger arrows), along with actin. Scale bar = 5 μm.

Fig. 6.

CDK5 expression in hair cells increases along the tonotopic axis. A and B: shown are fluorescent confocal images of cross sections (paraffin embedded) of the basilar papilla labeled with anti-CDK antibody. CDK5 is expressed in greater amounts in tall hair cells (THC) than in short hair cells (SHC). Similarly, high-frequency THC express more CDK5 (B) than low-frequency THC (A). C: a quantitative estimate of CDK5 expression in the THC from 1.0 and 2.5 mm from the apical end of the papilla confirm a 2.2-fold increase in CDK5 expression in the latter [low-frequency cells 228.7 ± 20.6 (SE); high-frequency cells 495.5 ± 54.1 (SE)]. For these experiments, mean fluorescence intensity was used as a measure of protein expression. Four THC from the neural edge from different sections were used for these comparisons (n = 12 cells each). Scale bar = 20 μm.

CDK5 affects Slo surface expression by direct phosphorylation.

We sought to determine the effects of CDK5 phosphorylation on Slo surface expression. CDK5 has been shown to affect the surface expression of a number of receptors, including the N-methyl-d-aspartate and the nicotinic acetylcholine receptor (36, 37, 64). We established a stable cell line that expressed Slo that was tagged with the Myc epitope at its NH₂-terminus. The latter feature allowed us to label Slo on the surface of the cell in lieu of the extracellular location of its NH₂-terminus. Cells were transfected with CDK5-CFP and the CDK5 activator p35 (or CFP alone in control cells). Cells were labeled with Alexa 647 conjugated anti-MYC antibody and separately a Slo antibody directed against the intracellular COOH-terminus of Slo after cell permeabilization. Cells expressing CDK5-CFP (or CFP alone in control cells) were sorted, and total and surface expression of Slo were determined in cells showing CFP fluorescence. As shown in Fig. 7, there was a relative decrease in Slo surface expression in cells expressing CDK5-CFP (and p35). We also treated HSlo expressing HEK cells with 1 μM roscovitine for 3 h and determined the surface expression of Slo. Here too we measured total expression of Slo within the cell to confirm that changes in surface expression were not secondary to changes in total expression of Slo. As is evident in Fig. 7, there is an increase in the relative amounts of Slo expressed on the surface of the cell in cells treated with the specific CDK5 inhibitor roscovitine (1 μM). Together, these data suggest that phosphorylation by CDK5 decreases surface expression of Slo.

Fig. 7.

Coexpression of CDK5 decreases, and inhibition of its kinase activity increases, relative surface expression of Slo. A: shown are histograms of fluorescence intensity of total HSlo (intra) and Slo on the surface of the cell (surf) in stable Myc-HSlo expressing human embryonic kidney (HEK) cells. Fluorescence intensity was detected using FACS. Cells were transfected with CDK5-CFP and p35 or CFP alone, and Slo expression in CFP-positive cells were quantified. CDK5 induces an increase in total Slo expression that is accompanied by a decrease in Slo surface expression. B: stable HEK cells expressing Myc-Slo were treated with 1 μM roscovitine (Rosc) and total Slo and Slo on the surface detected by fluorescent antibody labeling. As is evident, cells treated with roscovitine had unchanged levels of total Slo, but had an increase in surface-labeled Slo. C: comparison of Slo expressed on the surface of the cell to total amounts of Slo. CDK5 produces a decrease in relative surface expression of Slo, while 1 μM roscovitine produces an increase in relative expression of Slo on the cell surface [HSlo alone 0.38 ± 0.01 (SE), CDK5 0.31 ± 0.01 (SE), and roscovitine 0.46 ± 0.03 (SE)]. One-way ANOVA gave a P value of 0.004. A multiple-comparison test (Student-Newman-Keuls) showed significance between Slo alone and Slo + CDK5 (P < 0.05), and Slo alone and Slo + roscovitine (P < 0.05). N = 3 for each data set.

We then sought to determine whether the effects of CDK5 on Slo surface expression were mediated by direct phosphorylation. Recently, work from Trimmers group identified a number of phosphorylation sites in Slo using mass spectrometric detection of phospho peptides (63). Of over 25 unambiguous phosphorylation sites, we determined 7 residues as predicted targets of CDK5 phosphorylation using 2 different prediction algorithms (that is, 7 residues were predicted by both GPS 2.1 and Netphos1.0 algorithms; a further 2 were predicted targets of phosphorylation based on each of 1 of these prediction algorithms). Thus S642, S655, S659, T847, S855, S859, and T965 were identified as potential targets of CDK5 phosphorylation. We then attempted to make stable cell lines expressing these individual phosphonull/phosphomimetic mutations. We were able to establish polyclonal permanent cell lines in six of the seven phosphorylation sites with both phosphonull (S/T to A) and separately phosphomimetic (S/T to D) mutations (we were unable to make a stable cell line with S655D). In three of these phosphorylation sites (847, 859, and 965), phosphomimetic mutations caused a decrease in relative surface expression of Slo (that is, surface expression corrected for total expression within the cell; Fig. 8), with T847D producing the biggest effect. Phosphonull mutations at these sites caused minimal increase in relative surface expression. In almost all of these mutations, the total amount of Slo paralleled the surface expression of Slo, although the effects on relative surface expression were more pronounced than the total expression of Slo in the phosphomimetic mutations (Fig. 8). Phosphomimetic and phosphonull mutations of one site (S655) produced no changes in total as well as relative surface expression and were assayed after transient transfection.

Fig. 8.

CDK5 produces a decrease in Slo surface expression by direct phosphorylation. A–C: specific phosphomimetic mutations of Slo (T847D; S859D; T965D) decrease surface expression of Slo. Shown are fluorescent histograms of intracellular and surface-labeled Slo in several phosphomimetic and phosphonull mutations (A: T847D/A; B: S859D/A; C: T965D/A). D: shown are the relative expression of Slo on the surface of the cell (as a fraction of total Slo) in phosphonull and phosphomimetic mutations of CDK5 phosphorylation sites in Slo. One-way ANOVA showed a significant difference in relative surface expression between the different groups (P < 0.0001). Three phosphomimetic mutations show a statistically significant decrease in Slo surface expression compared with HSlo using the Student-Neuman-Keuls multiple-comparison test (T847D, P < 0.001; S859D, P < 0.05; T965D, P < 0.05). N = 3 for each data set.

While these data collectively suggest that CDK5 reduces Slo surface expression by direct phosphorylation of Slo, we sought to further confirm this possibility. We compared the changes in total and surface Slo expression in T847A when coexpressed with CDK5. T847D produced the largest change in surface expression, and we reasoned that, if the effects of CDK5 were due to direct phosphorylation, we would observe minimal changes in total and surface expressed Slo when we coexpressed CDK5 with the phosphonull mutation T847A. As shown in Fig. 9, coexpression of CDK5 results in no change in total expression of Slo and a slight reduction in surface expression.

Fig. 9.

CDK5 is unable to induce the changes in total expression and surface expression of Slo in the phosphonull mutation T847A. Shown are fluorescence histograms obtained by FACS analysis of HSlo and T847A expressed with CDK5CFP and p35 (B) and separately CFP (A). CDK5 and p35 do not change the total expression of Slo and produce minimal changes in surface expression of Slo in T847A. We have reproduced A from Fig. 7 for comparison. These data confirm that CDK5 induces changes in Slo surface expression by direct phosphorylation.

It should be noted that the experiments described above were performed with hSlo and its phosphodeletional and phosphonull mutations. While these data can likely be extrapolated to cSlo, which shares a 96% sequence homology including preservation of all seven phosphorylation sites, we felt it important that the experiments on phosphonull and phosphodeletional mutations be performed on mammalian Slo where these phosphorylation sites have been unambiguously established.

CDK5 has effects on the electrophysiological properties of Slo.

We also determined the effects of CDK5 on the electrophysiological properties of Slo. For these experiments, we injected oocytes with cRNA from cSlo. Additionally, some of these oocytes were also coinjected with cRNA encoding CDK5 or CDK5 and p35. p35 has been identified as a potent activator of CDK5. We recorded Slo currents using inside-out patch recordings from these oocytes. As noted in Figs. 10 and 11, Slo from excised patch recordings of oocytes injected with cSlo/CDK5 or cSlo/CDK5/p35 showed decreased voltage activation (depolarizing shift in the G-V relationship) compared with Slo alone. These changes were observed in all concentrations of “intracellular” Ca²⁺ tested. It should be noted that there was no difference in voltage activation when injecting CDK5 cRNA or CDK5 cRNA and p35 cRNA in combination. We interpret this to mean that oocytes contain sufficient amounts of endogenous activators of CDK5 (either p35 or p39).

Fig. 10.

CDK5 changes voltage activation of Slo. Representative current traces in inside-out macropatches from oocytes injected with cSlo (top), cSlo and CDK5 (middle), and cSlo, CDK5, and P35 (bottom) in 1 μM Ca²⁺. Patches were held at −100 mV and stepped to a range of voltages varying from −80 to +180 mV (20 mV per step) before stepping back to −100 mV. cSlo coinjected with CDK5 or CDK5 and P35 are less activated by voltage in a given Ca²⁺ concentration than cSlo alone. Moreover, tail currents of the coexpressed channels show slightly faster deactivation compared with cSlo alone.

Fig. 11.

Coexpression of cSlo with CDK5 or CDK5 and P35 causes shifts in its conductance-voltage (G-V) relationship. Shown are G-V relationships from excised inside-out patches of oocytes injected with cRNA containing cSlo alone, cSlo and CDK5, and cSlo, CDK5, and P35 in different Ca²⁺ concentrations (0, 1, 10, and 25 μM). CDK5 or CDK5 and P35 cause a decrease in voltage activation (depolarizing shift in Slo G-V relationships). cSlo, n = 12; cSlo + CDK5, n = 15; cSlo + CDK5 + P35, n = 16. G_max, maximal conductance.

We also determined the effects of coinjecting CDK5 or separately CDK5 and p35 on deactivation kinetics. We noted that, in nominally 0 μM Ca²⁺, coinjection of CDK5 or CDK5 and p35 both increased deactivation times compared with cSlo alone (Fig. 12). However, with increasing concentration of Ca²⁺ (1 and 10 μM), there was a reversal in this relationship, with CDK5 or CDK5 and p35 both producing a shortening in deactivation times compared with cSlo alone (Fig. 12). As with voltage activation, there was no difference between deactivation times in Slo currents in excised patches from oocytes injected with CDK5 or CDK5 and p35. In contrast to the effects on deactivation times, there were no differences in activation times.

Fig. 12.

CDK5 or CDK5 and P35 induce effects on deactivation time constants (τ_deactivation) that are dependent on Ca²⁺ concentration. A: shown are the τ_deactivation as a function of voltage in nominally 0 μM Ca²⁺. Coinjection of CDK5 or CDK5 and P35 prolonged deactivation of cSlo currents. However, both CDK5 and CDK5 with P35 shorten deactivation times with increased Ca²⁺ concentration, as shown in B (1 μM Ca²⁺) and C (10 μM Ca²⁺). This effect is particularly notable at holding potentials from a more depolarized range (−40 ∼ 0 mV). cSlo, n = 11; cSlo + CDK5, n = 15; cSlo + CDK5 + P35, n = 14.

CDK5 effects on the electrophysiological properties of Slo are likely due to constitutive and conditional phosphorylation of Slo.

We sought to determine whether the effects of CDK5 on Slo kinetics were due to direct phosphorylation. We analyzed Slo channels in which phosphomimetic (serine/threonine to aspartate) or phosphonull (serine/threonine to alanine) mutations were introduced as described above. We tested six pairs of phosphonull and phosphomimetic mutants. We determined that three mutations produced, to varying degrees, effects on voltage activation (G-V relationships) that would be consistent with a direct phosphorylation of Slo by CDK5 (Fig. 13). Thus mutations in T847, S859, and T965 all resulted in the phosphomimetic mutation, producing a depolarizing shift in G-V relationship compared with HSlo (with T847D producing the biggest shift). However, it should be noted that the phosphonull mutation and phosphomimetic mutation straddled HSlo G-V relationships in only two (S859 and T965) of these three sites (Fig. 13). In contrast, both the phosphonull (T847A) and phosphomimetic (T847D) mutations at T847 both had reduced voltage activation, although T847D has more markedly reduced voltage activation than T847A (Fig. 13).

Fig. 13.

Phosphomimetic mutations at T847, S859, and T965 mimic the effects of CDK5 phosphorylation on HSlo's voltage activation. A: both phosphomimetic and phosphonull mutations at T847 cause depolarizing shift of G-V relationship in 1, 10, and 25 μM Ca²⁺. T847D produces a more depolarized shift in G-V relationship than T847A. B: phosphomimetic and phosphonull mutations of S859 produce depolarizing and hyperpolarizing shifts in Slo G-V relationships, respectively. This observation was true in all but the highest concentrations of Ca²⁺ tested (25 μM Ca²⁺). C: T965D and T965A cause depolarizing and hyperpolarizing shifts in G-V relationships at all tested Ca²⁺ concentrations. HSlo, n = 8; T847A, n = 8; T847D, n = 7; S859A, n = 9; S859D, n = 8; T965A, n = 9; T965D, n = 8.

In contrast to these mutations that mimicked the effects of CDK5 on Slo voltage activation, we noted that other mutations produced opposing effects. Thus, phosphonull (S to A) mutations at S642 and S855 both produced decreased voltage activation compared with the corresponding phosphomimetic mutations. However, the phosphomimetic mutations at these sites (S642D and S855D) had voltage activation that was almost super-imposable on HSlo, suggesting that these two sites were constitutively phosphorylated in HSlo. Our observations with the phosphonull and phosphomimetic mutation at S855 are similar to the effects observed previously, although our data differ in one significant aspect. In the previous work, the phosphonull mutation had voltage sensitivities that closely resembled HSlo. In contrast, our data suggest the G-V relationship of the phosphomimetic mutation (S855D) to be super-imposable on HSlo.

Finally, we also noted that three of the mutants that mimicked the effects of CDK5 phosphorylation on voltage activation also had similar effects on deactivation times (Fig. 14). Thus, T847D, S859D and T965D all had more prolonged deactivation times than HSlo in nominally 0μM Ca²⁺ (Fig. 14). All of these mutants had shorter deactivation times compared with HSlo in 1 μM Ca²⁺, an effect that was similar to coinjection of CDK5.

Fig. 14.

Phosphomimetic mutations of T847 and S859 have similar effects to CDK5 on Slo τ_deactivation. Shown are the τ_deactivation of HSlo, T847D, and S859D at various voltages in nominally 0 Ca²⁺ (A) and 1 μM Ca²⁺ (B). Phosphomimetic mutations at both sites prolong τ_deactivation in nominally 0 Ca²⁺ and shorten τ_deactivation in 1 μM Ca²⁺, which are similar to the effects of CDK5 phosphorylation on Slo kinetics. HSlo, n = 7; T847D, n = 5; S859D, n = 6.

The data presented above showing broad similarity between the electrophysiological properties of phosphomimetic mutants of Slo and when CDK5 and Slo were coinjected strongly suggest that CDK5 brings about these changes by direct phosphorylation. We sought to further confirm this possibility by testing the effects of CDK5 on the phosphonull mutation T847A. The phosphomimetic mutation at T847 (T847D) had the largest change in voltage activation (best seen in graphs of its G-V relationship) of all the mutants that we tested and that also mimicked the effects of CDK5. We reasoned that a phosphonull mutation at T847 would show no change or a minimal decrease in voltage activation in response to coexpression with CDK5 and p35. Unexpectedly, we found that coexpression of CDK5/p35 with T847A resulted in channels that had increased voltage activation evidenced by the hyperpolarizing (left) shift in its G-V relationship (Fig. 15). These data strongly suggest that CDK5 causes decreased voltage activation [right (depolarizing) shift in G-V relationship] by phosphorylation of T847. Moreover, we conclude that the primary effect of CDK5 phosphorylation that results in a decreased voltage activation (depolarizing shift in Slo G-V relationship) is mediated by T847.

Fig. 15.

CDK5 likely produces the majority of its effects on the electrophysiological properties of Slo through direct phosphorylation of T847. A: coexpression of CDK5 (and p35) with T847A results in increased voltage activation compared with T847A alone. B: in contrast, coexpression of CDK5 (and p35) with HSlo results in a decreased voltage activation. Taken together, these data strongly suggest that the effects of CDK5 are mediated primarily through phosphorylation of T847.

DISCUSSION

In this paper, we identified the proline-directed serine/threonine kinase, CDK5, as a Slo interacting protein. In a yeast two-hybrid screen of chick cochlea cDNA using the COOH-terminus of Slo as bait, we identified CDK5 as a binding partner of Slo. We establish that CDK5 is expressed in hair cells, where it is most abundant in areas of the cell rich in cytoskeletal elements. We confirm that the interaction between Slo and CDK5 is likely real with reciprocal immunoprecipitation and FRET. We also show that this interaction likely has physiological consequences, with CDK5 affecting Slo kinetics and its expression on the surface of the cell.

CDK5 is a member of the cyclin-dependent protein kinase superfamily that is expressed in high levels in the central nervous system (12). Unlike other CDKs, however, CDK5 is not directly involved in cell cycle control, but rather has been implicated in a number of processes integral to the physiology and morphology of the cell (12). Thus CDK5 has been shown to be important in modulating neuronal migration and synaptic function (9, 10, 12, 18, 19, 20, 47). Moreover, recent data suggest that it plays a critical role in degenerative diseases of the central nervous system, including Alzheimer's disease (1, 21, 33, 35, 46).

Our findings that CDK5 is enriched in the cuticular plate and stereocilliary bundle are consistent with the known ability of CDK5 to phosphorylate and affect different cytoskeletal components (1, 21, 33, 35, 46). These findings may also have relevance to our inability to maintain hair cells in the presence of 1 μM roscovitine, which led to hair cell death (J. Bai and D. Navaratnam, unpublished observations). We speculate that CDK5 plays a critical role in maintaining the integrity of cytoskeletal elements in hair cells, in addition to affecting BK channels.

CDK5 affects the electrophysiological properties of Slo. Coexpression of CDK5 or CDK5 and p35 results in reduced voltage activation, reflected by the depolarizing shift in G-V relationship. Moreover, coexpression of CDK5 produces notable effects on deactivation times that were Ca²⁺ dependent. We reason that these effects are likely due to direct phosphorylation of Slo. First, specific phosphomimetic mutations mimic the effects of coexpressing CDK5. In particular, phosphomimetic mutations of T847 have the largest change in voltage activation and also mimic the effects of coexpressing CDK5. Furthermore, coexpressing CDK5 with the phosphonull mutation T847A produced an increase in voltage activation (hyperpolarizing shift in G-V relationship). This latter result also suggests that CDK5 phosphorylation of different residues can produce opposing effects on voltage activation, with phosphorylation of T847 overwhelming these effects. This assertion is supported by our finding that phosphonull mutations at two other CDK phosphorylation sites (642 and 855) produce decreased voltage activation (depolarizing shift in voltage activation).

We speculate that the effects of CDK5 phosphorylation on Slo kinetics are likely physiologically significant in hair cells. We were unable to directly test his hypothesis, since use of roscovitine resulted in hair cell death. The effects of CDK5 phosphorylation were to increase deactivation times in nominally 0 μM Ca²⁺ and shorten deactivation times at 1 μM and greater Ca²⁺ concentrations. Since native hair cells are thought to have resting Ca²⁺ of 5 μM (which increases further on hair cell depolarization), these data argue that the most likely effects of CDK5 in native hair cells is to shorten deactivation times. In this context, our findings that CDK5 protein expression increases in higher frequency hair cells are intriguing, since these hair cells have BK channels with shorter deactivation times. The concordance of increasing CDK5 levels in high-frequency hair cells, its effects on shortening deactivation times (in high-calcium concentrations), and the observation that BK channels in high-frequency hair cells have shorter deactivation times suggest that CDK5 may mediate BK channel activity in hair cells and contribute to electrical resonance. In this context, it is important to note that the primary mechanism of CDK5 activation is by interaction with p35 and p39. Both of these proteins are cleaved by calpains, a group of Ca²⁺-activated cytosolic proteases (23, 32, 34, 48, 49, 51). The two products of calpain-mediated cleavage, p29 and p25, both have increased affinity for CDK5 and produce more sustained CDK5 activity (23, 32, 34, 48, 49, 51). BK channels and Ca²⁺ channels are in close approximation in the basolateral pole of hair cells, where the entry of Ca²⁺ is thought to be the primary mechanism of activating BK channels. We speculate that the increased Ca²⁺ concentrations may have other secondary effects through activation of calpains and the subsequent cleavage of the CDK5 activators p35 and p39. Taken together these data raise the possibility that Ca²⁺ mediated activation of BK channels may be more complex and dynamic than previously envisaged with entry of Ca²⁺ not only directly activating BK channels, but also affecting its kinetics through calpains and CDK5. In contrast, there is scant data on how CDK5 could affect VGCCs, which would seem a likely target, given the sharp colocalization between BK channels and VGCCs in hair cells.

In previous work on these mutant channels, it was noted that S855A and S855D produced a decreased and increased voltage activation, respectively, compared with HSlo. While these data were broadly similar to our data, it differed in one significant aspect. Our data show the voltage activation of the phosphomimetic mutation (S855D) to be super-imposable on that of HSlo. In contrast, previous experiments showed voltage activation of S855A to be closer to HSlo than S855D. While wide variations in voltage activation have been noted previously in single-channel data (from single experimenters) for reasons that have been unclear (24, 25), it is possible that the differences in our data represent different states of basal CDK5 activity. Thus we speculate that the HEK cells used in our experiments contain a higher basal level of CDK5 activity compared with those used by Yan et al. (63). In addition, Yan et al. also did not observe a difference in voltage activation from phosphonull and phosphomimetic mutations at T847 or T965 and noted decreased voltage activation with both S859A and S859D. While we do not have an explanation for these discrepancies, we speculate that these discrepancies are due to changes in basal expression of other kinases and phosphatases that modify voltage activation. Phosphorylation of Slo by specific kinases can conditionally allow effects of subsequent phosphorylation by other kinases. For instance, phosphorylation of S695 by PKC renders the channel insensitive to subsequent activation by PKA or PKG (65). Similarly, phosphorylation of S1115 by PKC renders the channel insensitive to subsequent activation by PKA (65). The complex reciprocal interactions between kinases, including PKC and CDK5, will need to be studied further.

We show that expression of CDK5 decreases, and inhibition of CDK5 kinase activity increases, surface expression of Slo. We reason that CDK5 affects Slo surface expression by direct phosphorylation of the protein. Three lines of evidence support our assertion. First, phosphomimetic mutations have similar effects to coexpressing CDK5 and p35. Second, inhibition of CDK5 kinase activity produces effects that are similar to phosphonull mutations. Third, coexpression with CDK5 and p35 produces minimal changes in total Slo and Slo on the surface of the cell in phosphonull mutations of T847. Phosphomimetic mutations of T847 produce the largest changes in surface expression, suggesting that this site is critical for regulation of CDK5-induced effects on surface expression. In this context, we would like to note that T847 is a residue that also plays a critical role in the voltage activation of Slo (see above), making phosphorylation of this residue a key determinant of its properties. The exact mechanism of how CDK5 brings about decreased surface expression (increased endocytosis/decreased delivery to the surface, etc.) has to be determined.

CDK5 has been shown to have a similar effect on the expression of the NR2B subunit of the N-methyl-d-aspartate receptor. Here too, increased CDK5 activity results in decreased surface expression of the NR2B subunit. However, the effect is an indirect one with CDK5-mediated phosphorylation of PSD-95 decreasing its binding to Src, leading to reduced Src activity. Decreased Src activity, in turn, reduces phosphorylation of the tyrosine residue Y1472 in the internalization motif YEKL in the NR2B subunit (64). The reduced phosphorylation of this tyrosine residue increases endocytosis of NR2B. Thus increasing phosphorylation by CDK5 causes decreases surface expression of NR2B (64). Additionally, CDK5 binds to calpains and increases their activity, resulting in the proteolytic degradation of NR2B with a consequent decreased surface expression (23).

In conclusion, we show here that Slo interacts with CDK5, with the latter inducing effects on kinetics and the surface expression of Slo. These interactions are likely to affect BK channel behaviors in hair cells, where their effects have to be established.

GRANTS

This study was supported by National Institute of Deafness and Other Communications Disorders Grant R01-DC-007894.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: J.-P.B. and D.N. conception and design of research; J.-P.B., A.S., P.J., L.G., and D.N. performed experiments; J.-P.B., A.S., and D.N. analyzed data; J.-P.B. and D.N. interpreted results of experiments; J.-P.B., A.S., and D.N. prepared figures; J.-P.B. and D.N. drafted manuscript; J.-P.B. and D.N. edited and revised manuscript; J.-P.B., A.S., and D.N. approved final version of manuscript.