Ca2+/calmodulin dependent protein kinase (CaMKII) and protein phosphatase 2B (calcineurin) are key enzymes in the regulation of synaptic strength, controlling the phosphorylation status of pre- and postsynaptic target proteins. Here, we show that the inactivation gating of the Shaker-related fast-inactivating KV channel, Kv1.4 is controlled by CaMKII and the calcineurin/inhibitor-1 protein phosphatase cascade. CaMKII phosphorylation of an amino-terminal residue of KV1.4 leads to slowing of inactivation gating and accelerated recovery from N-type inactivated states. In contrast, dephosphorylation of this residue induces a fast inactivating mode of KV1.4 with time constants of inactivation 5 to 10 times faster compared with the CaMKII-phosphorylated form. Dephosphorylated KV1.4 channels also display slowed and partial recovery from inactivation with increased trapping of KV1.4 channels in long-absorbing C-type inactivated states. In consequence, dephosphorylated KV1.4 displays a markedly increased tendency to undergo cumulative inactivation during repetitive stimulation. The balance between phosphorylated and dephosphorylated KV1.4 channels is regulated by changes in intracellular Ca2+ concentration rendering KV1.4 inactivation gating Ca2+-sensitive. The reciprocal CaMKII and calcineurin regulation of cumulative inactivation of presynaptic KV1.4 may provide a novel mechanism to regulate the critical frequency for presynaptic spike broadening and induction of synaptic plasticity.
- voltage-activated K channels
- N-type inactivation
- C-type inactivation
- protein phosphatase
The strength of synaptic connections within neuronal circuits is flexible. This plasticity in neuronal excitability has been recognized as an important property underlying short- and long-term changes in the concerted activity of pre- and postsynaptic elements including ion channels (Kandel et al., 1991; Bliss and Collingridge, 1993; Zucker, 1993). It has been shown for a number of ligand- and voltage-gated ion channels that their activity can be modulated by the activation of protein kinases and phosphatases, which are regulated, in turn, by second messenger systems, e.g., Ca2+ and cAMP (Schulman, 1995). Potassium channels that constitute an extremely diverse superfamily involved in the control of pre- and postsynaptic excitability in this respect are particularly interesting. The regulation of potassium channel activity by protein phosphorylation may alter very distinctly neuronal excitability (Jonas and Kaszmarek, 1996).
Several recent in vitro studies have focused on theShaker superfamily of voltage-activated potassium (KV) channels (Chandy and Gutman, 1994). It was shown that phosphorylation by protein tyrosine kinase reduced the activity of KV1.2 and KV1.3 channels (Huang et al., 1993;Lev et al., 1995; Holmes et al., 1996). Also, cAMP-dependent protein kinase A (PKA) phosphorylation upregulates the activity levels of Kv1.2 (Huang et al., 1994) and KV2.1 (Wilson et al., 1994) and downregulates that of KV3.2 channels (Moreno et al., 1995). These KV channels belong to the class of delayed-rectifier channels mediating currents that do not inactivate rapidly. Other Shaker-related KVchannels, e.g., Shaker itself, KV1.4, and KV3.4, express outward currents that rapidly inactivate within milliseconds because of the presence of an amino-terminal inactivation domain (Hoshi et al., 1990). It was shown that phosphorylation of the amino-terminal inactivation domain of KV3.4 channels by protein kinase C (PKC) completely eliminated N-type inactivation (Covarrubias et al., 1994). Also, dephosphorylation of a modulatory C-terminal site of Drosophila Shaker channels considerably slowed the inactivation rate (Drain et al., 1994).
Ca2+/calmodulin-dependent kinases (CaMKIIs) are prominently expressed in mammalian brain, both pre- and postsynaptically. It is well accepted that CaMKII, which phosphorylates several pre- and postsynaptic proteins, has an important function in regulating neuronal excitability and synaptic strength (Braun and Schulman, 1995). It has been shown that members of theShaker KV channel family, e.g., KV1.1, KV 1.2, and KV 1.4, are localized to pre- and postsynaptic compartments (Sheng et al., 1992, 1993; Veh et al., 1995), making them possible targets for CaMKII phosphorylation. When we screened the cytoplasmic regions of KV1.4, a rapidly inactivating KV channel (Stühmer et al., 1989), for possible CaMKII phosphorylation sites, we detected three CaMKII consensus sequence motifs in the cytoplasmic amino-terminal sequence (RXXS/T at serine 101/102 and 123, and threonine 191). We show here that serine 123 of KV1.4 is a substrate for CaMKII phosphorylation and that CaMKII-phosphorylated KV 1.4 channels are dephosphorylated by the Ca2+-regulated calcineurin (protein phosphatase 2B)/inhibitor-1 protein phosphatase cascade. This Ca2+-sensitive phosphorylation/dephosphorylation of KV 1.4 has profound functional consequences for the inactivation properties of KV 1.4-mediated A-type potassium currents.
MATERIALS AND METHODS
Construction of HEK 293-KV1.4 cell line.Briefly, rat KV1.4 cDNA nucleotide (nt) 387–2658 (Stühmer et al., 1989) subcloned into pBluescript pKS+ was combined with a blunt-end Ear I fragment of the metallothionein IIA promotor (MT) [nt 30–830; (Karin et al., 1984)] cloned into the SalI polylinker restriction site of Bluescript pKS+. The resulting KV1.4 pKS+ clone was digested withEcoRI/BglII. The isolatedEcoRI/BglII MT-KV1.4 restriction fragment was ligated with EcoRI/BglII cut pmL2 eucariontic expression vector, containing polyadenylation and transcription termination signals from the SV40 late region. The KV1.4 pTMT-construct was checked by restriction analysis. HEK 293 cells were transfected with KV1.4 pTMT DNA after the protocol of Chen and Okayama (Chen et al., 1987). HEK 293 cells were grown in DMEM:F12 (Life Technologies, Gaithersburg, MD) supplemented with 10% FCS (Biother), 2 mm l-glutamine (Life Technologies) and penicillin-streptomycin (50 IU/ml-50 μg/ml; Life Technologies). Stably transfected cells were selected with 1 mgl G418 (Life Technologies). Selected HEK 293-KV1.4 cell clones were analyzed by Northern blot analysis for KV1.4 RNA expression and by Western blot analysis for KV1.4 protein expression.
In vitro Mutagenesis. Point mutations in the KV1.4 amino terminus were introduced by a PCR-based site-directed mutagenesis (Ho et al., 1989). For the mutation KV1.4 S123A we used the following oligonucleotides: AAG ATC CTT AGG GAG ATG GCC GAG GAG GAG (sense) and GTG GTA GAA AAT AGT TAA A (antisense). The PCR products were digested withSauI and ligated into KV1.4 pAKS2. The mutant KV1.4 T191A was generated by an overlay PCR using the oligonucleotides CTA CGC TTC GAA GCC CAA ATG AAA (sense) and TTT CAT TTG GGC TTC GAA GCG TAG (antisense). The two PCR fragments were digested with SauI and cloned into KV1.4 pAKS2. The mutant Kv 1.4 SS101/102AA was generated by using oligonucleotides CAC AGG CAG GCC GCT TTT CCT CAT TGC (sense) and G AGG AAA AGC GGC CTG CCT GTG GTG GAG (antisense). The product from the overlap PCR was cut withNcoI and cloned into Kv1.4 pAKS2. All mutants were verified by sequencing (Sanger et al., 1977) before use. For mRNA synthesis KV1.4 pAKS2, KV1.4 S123A pAKS2, and KV1.4 T191A pAKS2 were linearized with EcoRI.In vitro transcription was performed using the Sp6 Message Machine (Ambion, Austin, TX).
In vitro translation and in vitrophosphorylation. The radioactive in vitrotranslation of KV1.4 pAKS2 and KV1.4 S123A were performed using 1 μg cRNA in the FLEXI rabbit reticulocyte lysate system (Promega, Madison, WI) with 35S-methionine according to protocol. The reactions (final volume, 50 μl) were incubated for 1 hr at 30°C in the presence of 7.2μ canine pancreatic microsomal membranes. For in vitro phosphorylation, nonradioactivein vitro translation was set up using a full amino acid mixture and incubated under the same conditions as above. The microsomal membranes were sedimented at 4°C with an Eppendorf (Madison, WI) tabletop centrifuge at 15,300 rpm and afterward resuspended in CAMKII reaction buffer (20 mm Tris/HCl, pH 7.5, 10 mm MgCl2, 2 mmCaCl2, 0.5 mm dithiothreitol, 0.1 mm EDTA). For in vitro phosphorylation, 2.4 μm calmodulin, 100 μm ATP,32P-γ-ATP (specific activity of 100 μCi/μmol) and 250 U CaMKII (New England Biolabs, Beverly, MA) were added to the mixture. The reactions (final volume, 50 μl) were incubated at 30°C for 10 min in the presence of 1 mm Cyclosporin A and 100 μm okadaic acid. The reactions were stopped by addition of 50 μl of prewarmed (56°C) 2× Laemmli buffer. Ten microliters of the in vitro translation and 20 μl of the in vitro phosphorylation reaction were loaded per lane on 10% SDS-polyacrylamide gels. For control, in vitro translation and phosphorylation reactions without cRNA were performed to estimate the background of the in vitro phosphorylation reaction.
Electrophysiology and microinjection. Macroscopic currents were obtained from perforated-patch (Rae et al., 1991) and standard whole-cell patch-clamp recording (Hamill et al., 1981) using HEK 293-KV1.4 cells and cRNA-microinjected 293 cells. They were maintained with standard cell culture protocols and plated at a density of 5 × 104/ml on poly-l-lysine (50 μg/ml)-coated cellocate grids (Eppendorf) 1 d before the experiment. The bath solution for electrophysiological experiments contained (in mm): 135 NaCl, 5 KCl, 2 CaCl2, 2 MgCl2, 5 HEPES, 10 glucose, 20 sucrose, pH = 7.4 with NaOH. For perforated-patch recordings, the pipette solution contained (in mm): 70 K2SO4, 10 KCl, 10 NaCl, 7 MgCl2, 10 HEPES, pH = 7.2 with KOH, and 0.25 mg/ml amphotericin-B. For standard whole cell, the pipette solution contained (in mm): 95 K-aspartate, 20 KCl, 1 CaCl2, 11 EGTA, 10 HEPES, 2 glutathione, 2 Na2ATP, pH = 7.2 with KOH with a calculated (EQCAL, Biosoft, Cambridge, UK) free Ca2+ concentration of 20 nm. Total CaCl2 concentrations were increased according to EQCAL calculations to obtain free Ca2+concentrations between 100 and 1000 nm. Patch pipettes were pulled from borosilicate glass with resistances between 2 and 4mΩ. Only standard whole-cell experiments with series resistances <10 mΩ and perforated-patch experiments with <15 mΩ were included in this study. Series resistance compensation (0–80%) was used to obtain voltage errors <5 mV. Currents were recorded with an EPC9 (HEKA Elektronik) patch-clamp amplifier. The program package PULSE + PULSEFIT (HEKA Elektronik) was used for data acquisition and analysis. Leakage and capacitive currents were subtracted on-line using the P/4 subtraction method. P/4 pulses were applied 2 sec after the test pulses and the interpulse intervals were 30 sec. Records were digitized at 5 kHz and filtered with low-pass Bessel characteristic of 1 kHz cut-off frequency. For quantification of the time course of K currents, individual sweeps were fitted with a Hodgkin–Huxley related formalism (PULSEFIT): I =I 0 m(t)4 ht) with m(t) = 1-exp (−t/t m) andh(t) = h ∞ +a1a2 exp(−t/t h1) + (1−a1a2) exp(−t/t h2). The time constantst m t h1, andt h2 determine the activation and inactivation kinetics, respectively, h ∞ the steady-state inactivation. The two inactivation time constants are weighted by the variable a1/a2. Steady-state activation and inactivation curves were fitted with Boltzmann functions and time courses of recovery with monoexponential functions using a Marquardt-Levenberg algorithm in SIGMAPLOT (Jandel Scientific, San Rafael, CA). All experiments were performed at 30 ± 0.5°C. Data are given as mean ± SEM.
Autothiophosphorylation of purified CaMKII from rat brain (Calbiochem, La Jolla, CA) was induced by 5 min incubation (25°C) of 2 μg/ml CaMKII with 30 mg/ml calmodulin, 50 mm ATP gm-S, and 1 mg/ml BSA in a solution containing (in mm): 50 HEPES; 5 MgCl2; 0.3 CaCl2, pH 7.4. Aliqouts of CaMKII-solution were dialyzed on microdialysis membranes (Millipore, Bedford, MA) against a 150 mm KCl + 10 mm HEPES, pH 7.4, solution for 30 min at 4 C° and subsequently microinjected in 293 cells. Calcineurin autoinhibitory peptide [(CNIP) Bachem, Torrance, CA) was dissolved in water and microinjected. Stock solutions of KN-62 and KN-93, okadaic acid (Calbiochem), and cylosporin A (Calbiochem) were dissolved in dimethylsulfoxide.
For microinjection of mRNA, CaMKII or CNIP, KV1.4–293 or 293 cells were plated on poly-l-lysine-coated cellocate grids, and the Microinjector 5242 (Eppendorf) with an Automated Injection System (AIS) from Zeiss (Thornwood, NY) was used with solutions containing 50 ng/ml mRNA (Ikeda et al., 1992). KV1.4 wild-type and mutant cDNAs were cloned into pAKS expression vector for in vitro mRNA synthesis as described (Stühmer et al., 1989). Appropriate filling of individual cells was obtained with injection times of 200–300 msec and injection pressures between 80 and 140 hPa. In initial experiments, the filling of HEK 293-Kv1.4 cells was verified by co-injection of 0.1% fluorescein isothiocyanate-Dextran [(Sigma, St. Louis, MO) data not shown].
KV1.4 A-type currents in stably transfected 293 cells
Two hundred ninety-three cells were used as an in vitro model system to characterize the modulation of KV1.4 currents by CaMKII and protein phosphatases. The cells were stably transfected with a rat KV1.4 cDNA construct under control of the human metallothionin promoter (Kv1.4–293). KV1.4–293 cells gave rise to voltage-activated rapidly inactivating outward currents (Fig.1 A). They had properties similar to KV1.4 currents expressed previously in theXenopus oocyte expression system (Stühmer et al., 1989). In the perforated-patch whole-cell configuation, depolarization of KV1.4–293 cells to +40 mV produced KV1.4 current amplitudes in the 1–10 nA range. For comparison, amplitudes of slowly activating, noninactivating endogeneous outward currents were at +40 mV in the 50–300 pA range in both control and KV1.4–293 cells (data not shown). In our standard conditions, <5% of total outward current amplitude was attributable to endogenous 293 K channels. Activation of KV1.4 currents in KV1.4–293 cells was fast (time rise to peak <5 msec) and had a threshold at approximately −50 mV. KV1.4 currents inactivated rapidly with a residual steady-state current after a 200 msec depolarizing test-pulse representing ∼10% of the peak current (Fig. 1 A). Voltage-dependence of steady-state KV1.4 current activation was described with a Boltzmann function; mean half-maximal (V50) activation occurred at −22.6 ± 1.7 mV with a mean slope of 12.1 ± 1.2. mV (n = 8; Fig. 1 B). Mean half inactivation was at −44.5 ± 0.6 mV with a mean slope of 2.3 ± 0.2 mV (n = 8; Fig. 1 B).
A Hodgkin–Huxley related formalism (I =I 0 m(t)4 h(t); see Materials and Methods) with a single time constant (τm) for activation and two time constants (τh1,τh2) for inactivation was used to fit macroscopic KV1.4 currents. Within the studied voltage range of −40 to +60 mV, τm showed a marked voltage dependency, decreasing from a mean of 3.6 ± 0.3 msec (n = 8) at −40 mV to 0.7 ± 0.2 msec (n = 8) at +60 mV (Fig. 1 C). In agreement with previous single-channel data demonstrating the voltage independence of transitions from the open to the inactivated state inShaker KV-channels (Zagotta and Aldrich, 1990), time constants of KV1.4 inactivation showed no apparent voltage dependency between −20 to +60 mV (n = 6; Fig.1 D). Throughout this range, >70% (Iτh1/I = 73.5 ± 4.1%;n = 12) of a fast inactivation component (τh1= 16.6 ± 0.9 msec at +40 mV; n= 12) and <30% (Iτh2/I = 26.5 ± 4.1%; n = 12) of a slower inactivation component (τh2 = 43.3 ± 2.3 msec at +40 msec;n = 12) contributed to KV1.4 inactivation. Two distinct molecular mechanisms for K+ channel inactivation have been described: N-type inactivation, which depends on an amino terminal inactivation domain acting rapidly as a tethered intracellular K+ channel blocker (Hoshi et al., 1990), and C-type inactivation, which seems to involve slower structural changes at the extracellular mouth of the channel pore (Baukrowitz and Yellen, 1995, 1996). Deletion of the amino terminal of KV1.4 inactivation domain [KV1.4Δ110; (Rettig et al., 1994)] removes rapid N-type inactivation of KV1.4 currents and a pure C-type inactivation remains. Expression of KV1.4Δ110 in 293 cells exhibited slowly inactivating currents (data not shown). The inactivation time course was well described by a single time constant in the range of 200 msec (τh = 218 ± 16 msec; n = 5) that is significantly longer than the two time constants for KV1.4 wild-type inactivation. It seems that at least the dominant fast inactivation component of KV1.4 represents pure N-type inactivation.
It has been shown that recovery from inactivation of Shakerchannels involves at least two processes, a fast one in the millisecond time range and a slow one, which takes several seconds for completion (Zagotta and Aldrich, 1990; Demo and Yellen, 1991; Baukrowitz and Yellen, 1995). The fast component reflects recovery from N-type inactivation, the slow one that from C-type inactivated states (Baukrowitz and Yellen, 1995). Similarly, KV1.4 recovery from inactivation, determined in the perforated-patch configuration by a double pulse protocol of 200 msec depolarizing voltage steps to +40 mV separated by variable interpulse intervals at −80 mV (Fig.1 E), exhibited a fast and a slow component (Fig.1 F). Sixty-eight percent of the initial KV1.4 current recovered from inactivation via the fast recovery process [fraction (f) of fast recovery = 0.68], which could be described by a single exponential with a time constant τr1 of 598 msec (n = 5; Fig. 1 F). More hyperpolarized interval pulse potentials accelerated the initial component of KV1.4 recovery from inactivation saturating at potentials negative to −120 mV. The voltage dependence of the initial recovery process was described with a Boltzmann function with a meanV 50 of −53.5 ± 1.5 mV and slope of 17.8 ± 11.2 mV (n = 3). Complete recovery of KV1.4 from inactivation took up to 15 sec. The slow recovery process apparently had a time constant in the range of several seconds. Recovery of KV1.4Δ110 channels from C-type inactivation was similarly slow (τh = 2.4 ± 0.7 sec; n = 3). The slow recovery process showed no apparent voltage dependence. Because of this slow component of KV1.4 recovery, in all pulse protocols we used interval times of 30 sec between test pulses to avoid cumulative inactivation of KV1.4 currents.
Dephosphorylation accelerates KV1.4 inactivation
When calcineurin (protein phosphatase 2B) was inhibited in KV1.4–293 cells by microinjecting the calcineurin inhibitory peptide (CNIP, 10 μm) 1 hr before KV1.4 current recordings in the perforated-patch configuration, a delayed KV1.4 inactivation time course (Fig. 2 B) was observed in comparison to controls (Fig. 2 A). In contrast, time constants of activation were not affected by this treatment. Inactivation in the presence of CNIP could be well described with only one time constant (τh). Its value, 48.1 ± 2.5 msec at + 40 mV (n = 6), was similar to the slowly inactivating component τh2, which contributes <30% to the inactivation time course of KV1.4 currents in controls. Preincubation of KV1.4–293 cells with cyclosporin A, which blocks calcineurin via cyclophillins, also resulted in a slowed KV1.4 inactivation time course (46.1+ 1.9 msec;n = 3) similar to those observed with microinjected CNIP.
Calcineurin is able to initiate a calcineurin/inhibitor-1 protein phosphatase cascade where active calcineurin, by means of dephosphorylating an inhibitor, upregulates the activity of the calcium-independent protein phosphatase 1 [(PP1); for review, seeCohen, 1989]. PP1 activity is inhibited by nanomolar concentrations of okadaic acid (OA). To determine whether this cascade was also active in KV1.4–293 cells, they were preincubated with 100 nm OA for 30 min. In the presence of this protein phosphatase blocker, KV1.4 inactivation kinetics were also slow and similar to those observed with calcineurin inhibition (Fig.2 C). The inactivation time course of KV1.4 currents could again be well described by one time constant τh (51 ± 8 msec at + 40 mV; n = 6). Fitting the inactivation time constant of KV1.4 current responses to 1 sec depolarizing pulses gave similar time constants (57 ± 10 msec at + 40 mV, n = 5). In addition to PP1, 100 nm OA can also block PP2A, another calcium-independent protein phosphatase. PP2A has the highest affinity to this blocker and therefore can be selectively blocked with 1 nm OA (Cohen, 1989). Preincubation with 1 nm OA resulted in a less pronouced slowing of KV1.4 inactivation kinetics compared with the control (τh = 30.4 ± 2.8; n = 3) suggesting that in contrast with the calcineurin/inhibitor-1 protein phosphatase cascade, PP2A plays a minor role in regulation of KV1.4 inactivation kinetics.
When both the protein phosphatases and CaMKII were inhibited with 100 nm OA and 10 μm KN-62, respectively, KV1.4 currents were indistinguishable from those in control cells (Fig. 2 A,D). This showed that the effect of protein phosphatase inhibition on KV1.4 inactivation was occluded completely by the simultaneous application of a CaMKII inhibitor. Furthermore, incubation of KV1.4–293 cells with 100 nm okadaic acid and 10 μm KN-93, a more potent blocker of CaMKII compared with KN-62 (Sumi et al., 1991), yielded KV1.4 currents, which now inactivated approximately threefold faster than KV1.4 control currents (τh1 = 4.4 ± 0.3 msec at + 40 mV; n= 5) (Fig. 2 A,E). Similar to the results obtained in the simultaneous presence of OA and KN-93, the KV1.4 inactivation time course was dominated (OA+KN-93:Iτh1/I = 88.5 ± 1.5%;n = 5) by a fast inactivation time constant τh1 in the presence of KN-93 alone (4.2 ± 0.3 msec;Iτh1/I = 92.9 ± 1.0%;n = 6) (Fig. 2 F). These data suggested tentatively that KV1.4–293 cells contained a steady-state equilibrium between dephosphorylated and phosphorylated KV1.4 protein altered by either blocking the calcineurin/inhibitor-1 protein phosphatase cascade or CaMKII. Apparently, an increase in the concentration of phosphorylated KV1.4 protein induced by protein phosphatase blockers caused a slowing of inactivation kinetics. Conversely, a decrease in CaMKII-phosphorylated KV1.4 protein induced by CaMKII blockers caused an acceleration of KV1.4 inactivation kinetics. In contrast, the amplitude and kinetics of endogenous 293 currents were not effected by either inhibition of phosphatases (100 nm OA, n = 5) or CaMKII (10 μm KN-62; n = 5).
CaMKII phosphorylates KV1.4 channels
CaMKII might phosphorylate one or more of the three CaMKII phosphorylation consensus sites present in the KV1.4 amino terminus. Two of the motifs involving serine residues 101/102 and 123 are specific for KV1.4. They do not occur in other KV1 subfamily members (Chandy and Gutman, 1994). The third motif is related to threonine residue 191 located at equivalent positions within the amino terminal tetramerization (T) domain of all KV1 subfamily members (Shen and Pfaffinger, 1995). Serine 123, threonine 191, and serines 101/102 were replaced with alanine byin vitro mutagenesis of KV1.4 cDNA (KV1.4 S123A, KV1.4 T191A, KV1.4 SS101/102AA) to generate nonfunctional CaMKII consensus sites. The cDNAs were used as templates for in vitro mRNA synthesis. KV1.4 T191A mRNA did not express functional KV1.4 channels after microinjection into 293 cells and therefore was not analyzed further. KV1.4 SS101/102AA expressed only small currents with properties that were similar to wild-type KV1.4 (see below). KV1.4 S123A mRNA, on the other hand, expressed large currents after microinjection into 293 cells. The combined biochemical and electrophysiological characterization of KV1.4 S123A provided evidence that strongly suggested that serine 123 was the CaMKII modulatory site of KV1.4 (see below).
CaMKII-dependent phosphorylation of wild-type KV1.4 and KV1.4 S123A mutants were assayed in vitro. Wild-type KV1.4 and KV1.4 S123A mRNAs were translated into protein using a reticulocyte lysate supplemented with microsomes. The translated KV1.4 proteins were subsequently incubated with purified CaMKII and γ32P-ATP for biochemical phosphorylation studies. The 32P-labeled phosphorylated protein material was analyzed by denaturing SDS-PAGE followed by autoradiography (Fig. 3 A). The results showed that CaMKII had phosphorylated wild-type and KV1.4 S123A proteins to a different extent. Densitometric analysis of 32P-labeled KV1.4 protein normalized for protein concentration and averaged over seven experiments showed that ∼35% less 32P-phosphate had been incorporated into KV1.4 S123A than into wild-type KV1.4 protein (Fig. 3 A). Consistent with the presence of three CaMKII phosphorylation consensus sites in the KV1.4 amino terminus, the data suggested that CaMKIIin vitro phosphorylated more than one site on KV1.4 protein and that one of the functional sites had been eliminated in the S123A mutation.
Wild-type KV1.4 and KV1.4 S123A mRNA were microinjected into 293 cells and expressed currents were recorded in the perforated-patch configuration. The wild-type KV1.4 currents were indistinguishable from those elicited in stably transfected KV1.4–293 cells (compare Figs.1 A, 3B). The inactivation time course of KV1.4 S123A currents, on the other hand, was faster than wild type (Fig. 3 B,C). It was well described with two time constants (τh1 = 10.1 ± 0.3 msec; τh2= 36.4 ± 3.7 msec;Iτh1/I = 88 ± 4%;n = 6). Very similar time constants were obtained for KV1.4 S123A currents inactivation in the presence of 1 μm cyclosporin A (τh1 = 9.6 ± 1.4 msec; τh2 = 37.5 ± 1.5 msec; n = 5), 100 nm OA (τh1 = 10.5 ± 1.3 msec; τh2 = 32.0 ± 3.1 msec; n = 5), and the CaMKII inhibitor KN-62 (τh1 = 9.5 ± 0.4 msec; τh2 = 30.9 ± 1.6 msec, n = 3). These data demonstrated that inhibition of the calcineurin/inhibitor-1 protein phosphatase cascade or of CaMKII with KN-62 did not affect KV1.4 S123A inactivation kinetics in contrast to wild-type KV1.4. In the presence of the CaMKII inhibitor KN-93, wild-type KV1.4 (Figs. 2 E, 3 D) (τh1 = 4.2 ± 0.3 msec; τh2 = 26.0 ± 0.6 msec; n = 6) and KV1.4 S123A (Fig. 3 E) (τh1 = 3.8 ± 0.4 msec; τh2 = 26.9 ± 1.6 msec; n = 4) inactivation kinetics were similar and most rapid. In contrast, when the concentration of active CaMKII was increased by microinjecting the Ca2+-independent autothiophosphorylated CaMKII (for review, see Braun and Schulman, 1995) before electrophysiological experiments, KV1.4 and KV1.4 S123A inactivation kinetics were markedly different (Fig. 3 F,G). Inactivation of KV1.4 was delayed in comparison with controls (Fig.3 B,F) and was fitted by a single inactivation time constant (τh = 38.3 ± 3.3 msec; n = 6). This τh was similar to that obtained after blocking the calcineurin/inhibitor-1 protein phosphatase cascade (Fig.2 B,C). In contrast, the time course of KV1.4 S123A inactivation was not affected by microinjection of autothiophosphorylated CaMKII (Fig. 3 G). The inactivation time course was dominated by a fast inactivation time constant (τh1 = 7.9 ± 0.7 msec;Iτh1/I = 87.2 ± 3.3%;n = 5) as in control recordings. The results shown in Figure 3 demonstrated that KV1.4 is a substrate for CaMKII. When KV1.4 was phosphorylated at the serine 123 modulatory site, KV1.4-mediated currents inactivated 5–10 times slower than those mediated by dephosphorylated KV1.4. In addition, a comparison of KV1.4 Δ110 (Rettig et al., 1994) and KV1.4 Δ110 S123A-mediated currents showed that they possessed very similar slow inactivation kinetics (KV1.4 Δ110: τh = 218 ± 16 msec;n = 5; KV1.4 Δ110 S123A: τh= 213 ± 33 msec; n = 5). This observation suggested that the CaMKII phosphorylation at serine 123 apparently modulated N-type and not C-type inactivation of KV1.4.
KV1.4 recovery from inactivation modulated by CaMKII
KV1.4 currents that were recorded from KV1.4–293 cells microinjected with autothiophosphorylated CaMKII recovered from inactivation twofold faster and more completely (Fig. 4 A,B) than control (Fig.1 E,F). Eighty-seven percent (f = 0.87) of the KV1.4 current amplitude recovered from inactivation with a time constant τr1 = 309 msec (n = 5). This suggested that most KV1.4 channels in the CaMKII-phosphorylated form recovered directly from N-type inactivation and did not enter the long absorbing C-type inactivated state. In the presence of KN-93, however, initial KV1.4 recovery was twofold slower, with a time constant similar to control (τr1 = 601; n= 5) and recovered only to 49% with the initial fast component (Fig.4 C,D). Therefore, in the dephosphorylated form, ∼50% of the KV1.4 channels entered the long absorbing C-type inactivated state via N-type inactivation. The mean voltage dependency of the initial recovery component was not affected by CaMKII phosphorylation. In CaMKII-microinjected cells, it was described with a Boltzmann function with V 50 at −52.9 ± 1.9 mV and a slope of 18.9 ± 1.6 mV (n = 4). With KN-93, V50 was −55.2 ± 1.1 mV with a slope of 15.8 ± 0.8 mV (n = 3). Initial recovery of KV1.4 S123A currents from inactivation was even slower (τr1 = 755 msec; n = 3), and less completely (f = 0.41; n = 3) suggesting that ∼60% of the KV1.4 S123A channels recovered through a long absorbing state. These results indicated that the modulation of KV1.4 inactivation kinetics by CaMKII phosphorylation affected not only the rate of KV1.4 recovery from inactivation, but also the pathway of recovery. After CaMKII phosphorylation, more KV1.4 channels apparently returned directly from N-type inactivated state(s) and less entered long absorbing C-type inactivated state(s).
Macroscopic kinetics of voltage-dependent activation and the steady-state activation curve of KV1.4 were not affected by CaMKII phosphorylation (Fig. 5). Steady-state activations were described by Boltzmann functions and meanV 50 values and slopes obtained for KV1.4 currents from both CaMKII-microinjected cells (V 50 = −20.6 ± 2.0 mV; slope = 12.5 ± 1.4 mV; n = 6) and KN-93 treated cells (V50 = −23.7 ± 5.2 mV; slope = 6.4 ± 1.6 mV; n = 3) were similar to those obtained in control recordings. CaMKII phosphorylation, however, affected the apparent voltage dependence of KV1.4 steady-state inactivation. Microinjection of CaMKII shifted the apparent voltage-dependence of steady-state inactivation of KV1.4 currents by ∼7 mV to more positive membrane potentials (V 50 at −37.8 ± 1.2 mV; n = 6; Fig. 5 B) compared with both control recordings (Fig. 1 B) and those in the presence of KN-93 (V 50 was at −43.2 ± 1.8 mV; n = 4; Fig. 5 D). Similar to dephosphorylated KV1.4, theV 50 of KV1.4 S123A steady-state inactivation was also at more negative potentials (V 50 = −44.5 ± 0.6 mV; slope = 2.6 ± 0.5 mV; n = 4).
CaMKII phosphorylation modulates Kv1.4 cumulative inactivation
Standard voltage-clamp protocols reflect only poorly the physiological patterns of excitation occurring in vivo; i.e., repetitive discharge of short action potentials at varying frequencies. Under these conditions, A-type KV channels like KV1.4 are likely to undergo repetitive activation–deactivation and inactivation–recovery cycles. This is correlated with the occurrence of cummulative inactivation that involves both N-type and C-type inactivation (Baukrowitz and Yellen, 1995). To approximate physiological consequences that may be correlated with the CaMKII phosphorylation of KV1.4 channels, we simulated physiological discharge patterns by stepping the membrane potential of KV1.4–293 cells to +20 mV for 5 msec from interpulse potentials of −60 mV at frequencies between 1 and 100 Hz. Elicited KV1.4 current amplitudes were recorded in the perforated-patch configuration. In CaMKII-microinjected KV1.4–293 cells, 10 Hz stimulation induced only a small cumulative inactivation (Fig. 6 A). After 10 consecutive pulses, ∼80% of the initial KV1.4 current amplitude was still present. Throughout the studied frequency range (1–100 Hz), an acceleration of cumulative inactivation with increasing stimulation frequencies was apparent (Fig. 6 B), but even with 100 Hz, >50% of the initial KV1.4 amplitude was present after 10 pulses, demonstrating a relative resistence of CaMKII-phosphorylated KV1.4 channels to cumulative inactivation. In contrast, dephosphorylated KV1.4 channels obtained by either KN-93 incubation of wild-type KV1.4 (Fig. 6 C,D) or expression of KV1.4 S123 channels (Fig. 6 E,F) were sensitive to repetitive stimulation. In both cases, 10 Hz stimulation induced a significant cumulative inactivation after 10 pulses. A loss of ∼80% of the initial current amplitude was observed (Fig. 6 C,E). The increased sensitivity of dephosphorylated KV1.4 to cumulative inactivation was already apparent at 1 Hz stimulation frequencies and was accentuated at higher stimulation frequencies (Fig.6 D,F). After 10 pulses at 100 Hz, only ∼10% of the initially available KV1.4 channels remained active. Thus, the frequency–dependence of KV1.4 cumulative inactivation is determined markedly by CaMKII phosphorylation of KV1.4 at serine 123.
Inactivation kinetics of KV1.4 currents are Ca2+ dependent
The activation of CaMKII and calcineurin is regulated by intracellular Ca2+. In contrast with calcineurin, however, CaMKII can be converted to a Ca2+-independent form after autophosphorylation (Miller and Kennedy, 1986; Braun and Schulman, 1995). As we have shown, the phosphorylation status of the KV1.4 modulatory site depends on the relative activities of calcineurin and CaMKII and should thus depend on the intracellular Ca2+concentration. To examine this hypothesis directly, we recorded from KV1.4–293 cells in the whole-cell patch-clamp configuration to dialyze the cells with different free Ca2+ concentrations. With 200 nm free Ca2+ in the pipette solution, KV1.4 currents were similar to those observed in the perforated-patch configuration and showed no apparent change of inactivation kinetics throughout the course of the experiment (Fig. 7 A,E). Accordingly, inactivation time courses could be described by double exponential functions. A fast process with a time constant τh1 of 10.4 ± 1.3 msec at +40 mV (n= 6) dominated inactivation (Iτh1/I = 86.3 ± 3.1%) and the minor component (>15%) had a mean time constant, τh2 of 34.4 ± 2.2 msec at +40 mV (n= 6) (Fig. 7 F). Similar time constants were determined for whole-cell patch-clamp recordings with free Ca2+ concentrations between 0.1 and 1 μm in the pipette solution (Fig. 2 F). Also, steady-state activation and inactivation curves were similar to perforated-patch results (200 nm Ca2+: activation, V 50 = −23.1 ± 1.6; slope = 11.2 ± 0.6 mV; n = 8; inactivation:V 50 =−48.7 ± 2.2 mV, slope = 2.8 ± 0.3 mV; n = 8). In contrast, when the free Ca2+ concentration in the pipette solution was reduced to 20 nm, inactivation of KV1.4 currents progressively slowed after whole-cell dialysis (Fig. 7 B). The first KV1.4 current response, which was recorded immediatedly after breaking into the cell, still exhibited fast inactivation comparable with those recorded with higher Ca2+ concentrations. However, the KV1.4 current responses, which were subsequently recorded at 30 sec intervals, showed increasingly slower inactivation accompanied by an increase in sustained current at the end of the 200 msec pulse and an increase in total charge transfer (Fig. 7 E). The Ca2+-dependent modulation of KV1.4 inactivation reached a steady-state 6–10 min after breaking into the cell for standard whole-cell recording (Fig. 7 B). During dialysis of 20 nm free Ca2+, the slowing of the Kv1.4 inactivation time course could be described as a continous increase of the contribution of the slowly inactivating component (τh2 = 38.4 ± 1.6 msec, n= 5). This initially accounted for only ∼12% (τh2/I = 11.8 ± 3.4%;n = 5) of the inactivating Kv1.4 current and then rose to ∼65% (τh2/I = 65.0 ± 6.1%; n = 5). Accordingly, the contribution of the fast inactivating component (τh1 = 7.8 ± 1.1 msec; n = 5) had decreased during 20 nm Ca2+ dialysis from 88 to 35%. Preincubation with KN-62 blocked the transition from fast to slow KV1.4 inactivation with dialysis of 20 nm free Ca2+solutions (Fig. 7 C,E). When similar whole-cell patch-clamp experiments were performed with 293 cells expressing KV1.4 S123A currents, no Ca2+ sensitivity of the time course of inactivation was observed, and the fast component of inactivation was dominant (Iτh1/I>85%) with dialysis of both 20 nm (Fig. 7 D,E) and 200 nm free Ca2+ concentrations (20 nm: τh1 9.2 ± 0.3 msec; τh2 37.2 ± 2.5 msec;Iτh1/I = 90.4 +3.2%;n = 10). In contrast, 293 cells expressing KV1.4 SS101/102AA showed similar slowing after dialysis with 20 nm free Ca2+ concentrations compared with KV1.4 wild-type (n = 3) indicating that serines 101/2 are not involved in Ca2+ dependend modulation of KV1.4. These results suggested that the transition from fast to slow KV1.4 current inactivation observed with 20 nm Ca2+ dialysis depended on the presence of CaMKII-phosphorylated KV1.4 channels. Accordingly, the transition was blocked by preincubation with the CaMKII inhibitor KN-62 and apparently favored by the inhibition of the calcineurin/inhibitor-1 protein phosphatase cascade (Fig.2 B,C). The most likely explanation for these observations is that the balance between calcineurin and CaMKII activitities in KV1.4–293 cells was shifted at 20 nm Ca2+ in favor of the Ca2+independent autophosphorylated form of CaMKII, and consequently the concentration of CaMKII-phosphorylated KV1.4 was increased. Our data provide strong evidence that the Ca2+ sensitivity of KV1.4 inactivation kinetics is an expression of the dynamic equilibrium between CaMKII-phosphorylated and calcineurin/protein phosphatase 1 dephosphorylated KV1.4.
CaMKII phosphorylation of a KV1.4 amino-terminal site modulates inactivation kinetics
In this study, we have given biochemical and electrophysiological evidence that the Ca2+/CaMKII phosphorylated an important modulatory site of KV1.4 channels, located at serine 123 within the amino-terminal cytoplasmic domain. Furthermore, we have shown that this site is dephosphorylated by the Ca2+-dependent calcineurin/inhibitor-1 protein phosphatase cascade. CaMKII-phosphorylation of Kv1.4 has several functional consequences of KV1.4 activity. First, the time course of inactivation of CaMKII-phosphorylated KV1.4 channels (τh = 40–50 msec) (Fig. 3) is slowed by approximately one order of magnitude in comparison to that of dephosphorylated KV1.4 channels (τh = 4–8 msec) (Fig. 3). Second, CaMKII-phosphorylated KV1.4 showed an accelerated recovery from inactivation, a reduced propensity to enter long absorbing inactivated states, and a reduced tendency to undergo cumulative inactivation when stimulated with a train of action potential-like short depolarizations. Third, the voltage-dependent steady-state inactivation of CaMKII-phosphorylated KV1.4 channels was shifted to more positive potentials by ∼7 mV. This might increase a KV1.4 window current in the subthreshold range of between −50 and −30 mV.
Under control conditions in the perforated-patch configuration, KV1.4 currents had properties that were typical for dephosphorylated KV1.4; i.e., a dominant fast inactivating component, slow and incomplete initial recovery, and steady-state inactivation with a V50 at −45 mV. This indicated that KV1.4 was present in KV1.4–293 cells predominantly in the dephosphorylated form. The concentration of CaMKII-phosphorylated KV1.4 in 293 cells was increased by either microinjection of autothiophosphorylated CaMKII or by inhibition of calcineurin/protein phosphatase 1. Dialysis of KV1.4–293 cells with pipette solutions containing 20 nm free Ca2+ had a similar effect on KV1.4 inactivation kinetics as pharmacological inhibition of protein phosphatases. Both effects were occluded by the preincubation with the CaMKII inhibitors KN-62 or KN-93. These results suggest that in the 293 cellular enviroment KV1.4 channels exist in a steady-state equilibrium of CaMKII-phosphorylated and dephosphorylated forms. In aggreement with this notion are the properties of the currents mediated by KV1.4 S123A, which cannot be phosphorylated at the serine 123 modulatory site. These channels were locked in the fast inactivating mode and were not affected by microinjection of autothiophosphorylated CaMKII, protein phosphatase inhibitors, or dialysis with 20 nm free Ca2+ solutions.
CaMKII phosphorylation of KV1.4 modulates functional coupling of N- and C-type inactivation and frequency-dependent cumulative inactivation
Two principally different inactivation mechanisms have been identified for Shaker-related A-type channels. One is mediated by the amino-terminal inactivating “ball” domain acting as a intracellular tethered blocker of the open channel and has been termed N-type inactivation (Hoshi et al., 1990). The second mechanism, named C-type, is mediated by C-terminal domains and involves conformational changes at the extracellular mouth of the pore (Choi et al., 1991; Hoshi et al., 1991; Lopez-Barneo et al., 1993). Both types of inactivation can influence each other in a cooperative manner (Baukrowitz and Yellen, 1995, 1996). When the amino-terminal “ball” domain is deleted, e.g., in KV1.4 Δ110 (Rettig et al., 1994), KV1.4 channels inactivate only via C-type inactivation (τh = 213 msec). Comparison of the inactivation kinetics of CaMKII-phosphorylated KV1.4 channels with that of KV1.4 Δ110 shows that the former is still significantly faster than the latter. Also, the S123A mutation did not affect the time constant of C-type inactivation (KV1.4 Δ110 S123A: τh = 218 msec). This suggests that CaMKII-phosphorylation of the modulatory site at serine 123 may exert its effects on N-type inactivation of KV1.4. It is likely, however, that CaMKII phosphorylation influences C-type inactivation indirectly by reducing the functional coupling between N-type and C-type inactivated states. As pointed out by Baukrowitz and Yellen (1995), a short depolarizing pulse drives ShakerA-type K channels into N-type inactivation. From this state, they have two main pathways: they either recover directly via an open state (Ruppersberg et al., 1991) to resting states with rate λ or enter the long absorbing C-type inactivated state (Demo and Yellen, 1991) with rate μ. Therefore, the time constant for the fast phase of recovery is τr1 = [(λ + μ)−1 and the fraction (f) of channels recovering via the fast route is f = λ/(λ + μ) (Baukrowitz and Yellen, 1995)]. In the dephosphorylated state of KV1.4, these parameters are f = 0.5 and τr1 = 600 msec (Fig.4 D), i.e., both pathways have an equal rate of ∼0.8 sec−1. In comparison, the rate of recovery from C-type inactivation determined with the KV1.4Δ110 mutant is ∼0.4 sec−1. This comparison suggests that dephosphorylated KV1.4 channels are bound to accumulate in the C-type inactivated state via N-type inactivation. In contrast, the fraction of CaMKII-phosphorylated KV1.4 channels recovering directly from N-type inactivation is almost 90% (f = 0.87), with a time constant of 300 msec (Fig. 4 B). This predicts an accelerated rate λ of KV1.4 recovery from N-type inactivation (2.8 sec−1) and a reduced rate μ for entering the C-type inactivated state from N-type inactivation (0.5 sec−1). The latter is only slightly faster than the rate of leaving the C-type inactivated state (0.4 sec−1). This suggests that only a minor fraction of CaMKII phosphorylated KV1.4 channels can accumulate in the C-type inactivated state via N-type inactivation. Thus, CaMKII phosphorylation of KV1.4 is likely to interfere with the functional interplay of N-type and C-type inactivation. CaMKII phosphorylation of KV1.4 shifted the frequency-dependence of cumulative inactivation, induced by short repetitive depolarizations, toward higher stimulation frequencies (Fig.6). Only dephosphorylated KV1.4 channels showed an effective cumulative inactivation at low stimulation frequencies. This is in agreement with previous results that cumulative inactivation ofShaker A-type channels is most effective when N-type and C-type inactivation are both present and strongly coupled (Baukrowitz and Yellen, 1995).
Different localizations of modulatory phosphorylation sites inShaker-related A-type channels
It has been proposed from detailed kinetic analysis ofShaker channel inactivation combined with in vitro mutagenesis that N-type inactivation follows a “ball and chain” mechanism in which the ball is tethered to a flexible chain (Hoshi et al., 1990). Most likely, long-range electrostatic interactions are essential for the on-rate of inactivation (Murrell-Lagnado and Aldrich, 1993). They are apparently dominated by positive charges clustered within the inactivation domain and negative charges near or at the ball receptor region (Isacoff et al., 1991; Yool and Schwarz, 1995). Changes in net charge induced by phosphorylation in the ball region or other channel domains therefore are likely to affect inactivation kinetics. Covarrubias et al. (1994) showed for the cloned human A-type KV channel, hKV3.4 that PKC-dependent phosphorylation within the inactivation domain completely eliminates N-terminal inactivation. Conversely, Drain et al. (1994)demonstrated for Shaker KV channels that the phosphorylation of a C-terminal consensus site by PKA induces an acceleration of the macroscopic and microscopic transition from the open channel into the inactivated state. Therefore, the effect of phosphorylation on inactivation kinetics might be critically dependent on the localization of the phospho-acceptor in the channel protein. The CaMKII phosphorylation site in KV1.4 is obviously not part of the inactivation ball or ball receptor region, but rather is situated in the putative chain domain. It is possible that the density of negative charges near or at the CaMKII phosphorylation site might affect the flexibility of the “chain” connecting the inactivation ball with the core of the protein. Alternatively, phosphorylation at serine 123 might weaken the electrostatic ball–receptor interaction.
A role in synaptic frequency detection for CaMKII-dependent phosphorylation of KV1.4?
It is likely that KV1.4 channels are an in vivo substrate for CaMKII. In rat brain, KV1.4 channels may contribute, like Shaker channels in Drosophila, to action potential repolarization (Pardo et al., 1992). They are predominantly targeted to presynaptic compartments (Sheng et al., 1992,1993; Veh et al., 1995) where CaMKII (Braun and Schulman, 1995) and several protein phosphatases like calcineurin and protein phosphatase 1 are also abundant (Liu et al., 1994). The balance between CaMKII and the calcineurin/inhibitor-1 protein phosphatase cascade activities is known to regulate a number of pre- and postsynaptic key proteins (for review, see Schulman, 1995), including AMPA receptors (McGlade-McCulloh et al., 1993; Mulkey et al., 1994; Wyllie and Nicoll, 1994). It is an attractive possibility that CaMKII phosphorylation of KV1.4 channels contributes to the described CaMKII-induced shift of frequency-dependent properties of central synapses to higher discharge rates (Bear, 1995; Chapman et al., 1995; Mayford et al., 1995). Although it is not clear how CaMKII is involved in synaptic frequency detection, several studies favor the calmodulin trapping mechanism by autophosphorylated CaMKII as a good candidate (Meyer et al., 1992;Dosemeci and Albers, 1996). CaMKII-dependent modulation of KV1.4 inactivation kinetics might be an important additional factor. A synapse that makes use of KV1.4 activity, would require higher firing frequencies to drive CaMKII-phosphorylated KV1.4 channels into cumulative inactivation and, in turn, to increase the presynaptic action-potential duration facilitating Ca2+ entry (for review, see Byrne and Kandel, 1996). Only functional analysis of synapses, however, will clarify the relevance of CaMKII-dependent regulation of native fast-inactivating K channels for neural excitability.
This work was supported by a Grant of the Deutsche Forschungs-gemeinschaft (O. P.). We thank D. Clausen for excellent graphical services, S. Sewing for help with in vitromutagenesis, and D. Kuhl and C. Stansfeld for critical reading of the manuscript.
Correspondence should be addressed to Dr. Olaf Pongs, Zentrum für Molekulare Neurobiologie, Martinistrasse 52, D-20246 Hamburg, Federal Republic of Germany.
Dr. Lorra’s present address: Institut für Neurobiologie; INF 364, 69120 Heidelberg, Germany.