Phosphorylation and protonation of neighboring MiRP2 sites: function and pathophysiology of MiRP2-Kv3.4 potassium channels in periodic paralysis

MinK-related peptide 2 (MiRP2) and Kv3.4 subunits assemble in skeletal muscle to create subthreshold, voltage-gated potassium channels. MiRP2 acts on Kv3.4 to shift the voltage dependence of activation, speed recovery from inactivation, suppress cumulative inactivation and increase unitary conductance. We previously found an R83H missense mutation in MiRP2 that segregated with periodic paralysis in two families and diminished the effects of MiRP2 on Kv3.4. Here we show that MiRP2 has a single, functional PKC phosphorylation site at serine 82 and that normal MiRP2-Kv3.4 function requires phosphorylation of the site. The R83H variant does not prevent PKC phosphorylation of neighboring S82; rather, the change shifts the voltage dependence of activation and endows MiRP2-Kv3.4 channels with sensitivity to changes in intracellular pH across the physiological range. Thus, current passed by single R83H channels decreases as internal pH is lowered (pK a ~7.3, consistent with histidine protonation) whereas wild-type channels are largely insensitive. These findings identify a key regulatory domain in MiRP2 and suggest a mechanistic link between acidosis and episodes of periodic paralysis.—Abbott, G. W., Butler, M. H., Goldstein, S. A. N. Phosphorylation and protonation of neighboring MiRP2 sites: function and pathophysiology of MiRP2-Kv3.4 potassium channels in periodic paralysis. FASEB J. 20, 293–301 (2006)

Genes in the KCNE family encode single transmembrane (TM) domain peptides that do not function alone but coassemble with pore-forming potassium channel cx-subunits to form heteromeric complexes with the unique attributes of native potassium currents (1,2). Thus, MinK (KCNE1) associates with KCNQ1 to form a complex whose attributes closely resemble the cardiac current IKs (3,4); MinK-related peptide 1 (MiRP1, encoded by KCNE2) assembles with HERG to pass currents like cardiac IKr (5) and MiRP2 (KCNE3), the subject of the present study, associates with Kv3.4 to form a skeletal muscle potassium channel that activates at subthreshold membrane potentials (6). The necessity for KCNE peptides in native ion channels to achieve normal function is made apparent by their requirement to recapitulate the characteristics of native currents in heterologous expression studies and by genetic and biophysical evidence associating their alteration with human disease. Thus, missense mutations in KCNE1 and KCNE2 yield abnormal MiRPs that alter function and are associated with inherited and acquired long QT syndrome (LQTS), a cardiac arrhythmia that predisposes to sudden death (5,(7)(8)(9)(10)(11); KCNE3 is associated with periodic paralysis (PP) (6), a genetically and phenotypically mixed group of skeletal muscle disorders, characterized by episodic attacks of weakness or paralysis (12,13).
Disease-associated KCNE mutations identified to date reduce ability of channels to perform their normal role in repolarization due to diminished single channel conductance, a positive shift in the voltage dependence of activation, speeded deactivation, or a combination of these effects (6,8,14,15). Some mutations are revealed clinically in response to a secondary, environmental stimulus. Thus, KCNE mutants associated with drug-induced arrhythmia include those with normal sensitivity to drug blockade (5,16) and others with increased drug sensitivity (5). A common polymorphism associated with drug-induced arrhythmia manifests normal channel function that is altered only on drug exposure due to increased sensitivity to block (8). Analogously, a missense variant of the cardiac voltage-gated sodium channel that is common in African Americans, and linked with arrhythmia in aging adults (17) as well as sudden infant death syndrome, yields a channel that operates normally at baseline but shows abnormal gain-in-function with lowered intracellular pH (18).
Here, we describe the influence of altered intracellular environment on normal and dysregulated function of MiRP2-Kv3.4 channels. We find that phosphorylation of serine 82 of MiRP2 (MiRP2-S82 P ) is required for MiRP2 to produce the normal shift in voltage-dependent activation of Kv3.4 observed in vivo. The mutation of neighboring arginine 83 to histidine (MiRP2-R83H), associated with periodic paralysis, is found to mimic the effect of S82 dephosphorylation even though the change does not prevent phosphorylation of S82. Finally, the R83H variant confers sensitivity to lowered pH, offering a molecular rationale for the onset of periodic paralysis during activities that produce intracellular acidosis.

Electrophysiology
For electrophysiology, Chinese hamster ovary (CHO) cells were transfected using Superfect (Qiagen, Valencia, CA, USA) with vectors carrying the cDNA for GFP, wild-type, or variant MiRP2 and Kv3.4. Currents were recorded by patch-clamp in cell-attached or off-cell, inside-out mode, 1-2 days after transfection. Recordings were obtained using an Axopatch 200A Amplifier (Axon Instruments, Foster City, CA, USA), an IBM computer and CLAMPEX software (Axon Instruments). Data analysis was performed using CLAMPFIT, FETCHAN, PSTAT (Axon Instruments) and TAC software (Instrutech, Great Neck, NY, USA). All recordings were performed at room temperature. Curves were fit to a Boltzmann function: where V1/2 is the half-maximal voltage of activation and Vs the slope factor.

Protocols
Holding voltage was -80 mV unless otherwise stated. For quantification of the voltage dependence of activation, cells were held at -80 mV and pulsed in 10 mV, 3 s steps between -100 and +90 mV with a 5 s interpulse interval. For quantification of unitary current amplitude, cells were held at -40 mV for 5-20 min.

MiRP2-S82 is a phosphorylation site
When identified in patients with periodic paralysis, MiRP2-R83H was studied and found to significantly alter the gating of MiRP2-Kv3.4 channels (6,15). Specifically, wild-type MiRP2 shifts the midpoint voltage dependence of activation (V1/2act) of Kv3.4 channels to a subthreshold potential (~ -40 mV) while channels with R83H-MiRP2 show a suprathreshold value, V1/2act ~ -10 mV (15). Here, we explore the basis for the effect of the R83H mutation. Sequence analysis of human MiRP2 reveals that serine 82, neighboring R83, is within a predicted consensus protein kinase C (PKC) phosphorylation and that the site is conserved across the 5 members of the KCNE gene family (Fig. 1). We sought, therefore, to test the hypothesis that mutation at R83 altered function via an influence on the neighboring S82 phosphorylation site.
To determine if MiRP2 S82 is phosphorylated by PKC, COS cells were transfected with Kv3.4 and HA-tagged MiRP2 variants, for analysis by Western blotting using anti-phosphoserine antibodies. Following cell lysis and immunoprecipitation of wild-type MiRP2 (WT) using anti-HA, Western blot using anti-HA antibodies gave the expected bands (6) corresponding to the masses for the mature, immature and the carbohydrate-free glycosylation forms (Fig. 1C, lane 1). Western blotting using the anti-phosphoserine antibody revealed the same bands carried phosphoserine (Fig. 1D, lane 1). The same phosphoserine-positive bands were observed for MiRP2-R83H (Fig. 1D, lane 2) but not with MiRP2 variants, S82A and S82D, which remove the serine subject to phosphorylation (Fig. 1D, lanes 3 and 4). The results indicate that MiRP2 S82 can be phosphorylated by PKC, that this is the only MiRP2 site subject to modification, and that the R83H mutation does not prevent S82 phosphorylation.

MiRP2-S82 phosphorylation is required for subthreshold activation
The functional consequences of MiRP2 phosphorylation were assessed using patch clamp electrophysiology. Wild-type MiRP2-Kv3.4 channels readily reveal the most prominent influence of MiRP2 under base-line conditions, to produce subthreshold activation, when studied in cell-attached patches with symmetrical potassium across the membrane; wild-type MiRP2-Kv3.4 channels show significant inward current compared to channels with Kv3.4 alone ( Fig. 2A) (6). Compared to those with wild-type MiRP2, Kv3.4 channels with S82A-MiRP2, a variant that is not phosphorylated (Fig. 1D), exhibit much less inward current at negative potentials ( Fig. 2A ). This is particularly noticeable by comparison of traces at -40 mV (Fig. 2B, C) and suggests MiRP2 must bear phosphate to exert its full influence. Consistent with this conclusion, wild-type MiRP2-Kv3.4 channels expressed in cells treated with PMA to stimulate PKC activity show prominent inward currents at -40 mV while cells treated with BIS to inhibit PKC show reduced inward currents (Fig. 2D). Moreover, if phosphorylation of S82 mediates the effect, as expected, neither PMA nor BIS influence the operation of MiRP2-S82A-Kv3.4 channels (Fig. 2E ). Another argument for a role of phosphate on serine 82 in the effect of MiRP2 is recapitulation of the phosphorylated phenotype when a negative charge is permanently substituted at MiRP2 position 82; MiRP2-S82D-Kv3.4 channels show significant inward currents at -40 mV (Fig. 2B, C) in a manner that is insensitive to both PMA and BIS exposure (not shown).
Quantification of these effects (Table 1) shows that wild-type MiRP2-Kv3.4 channels in untreated cells activate with a V1/2act of ~-40 mV, as before (6). This value was relatively unaffected by application of PMA, but was shifted to ~-10 mV by BIS application. S82A channels showed a V1/2 act of ~-10 mV regardless of PMA or BIS treatment. S82D MiRP2Kv3.4 channels behaved similarly to wild-type MiRP2-Kv3.4 channels, with a V1/2 act of activation of ~-40 mV. For comparison, Kv3.4 alone channels were also assessed, giving a V1/2act of +9 mV, again similar to previous observations (6). These results are consistent with wild-type channels being constitutively phosphorylated by PKC at S82 and thus active at subthreshold potentials, and why they are largely unaffected by PKC stimulation, but show altered activity when PKC is inhibited. When S82 is mutated to A, PKC can no longer regulate MiRP2 via S82 and thus S82A channels are unresponsive to both activators and inhibitors of PKC. Conversely, the S82D change is found to mimic the effect of S82 phosphorylation yielding subthreshold activation.

R83H channels show abnormal activation, regardless of S82 phosphorylation state
R83H-MiRP2-Kv3.4 channels were shown to exhibit suprathreshold activation unlike wild-type MiRP2-Kv3.4 channels (6,15). Here, we find that phosphorylation of S82 in wild-type MiRP2 is required to produce subthreshold activation ( Fig. 2 and Table 1) and that R83H-MiRP2 is phosphorylated by PKC at S82, arguing that failure to achieve subthreshold activation is not due to suppression of this event (Fig. 1C,  D). Moreover, normal activation could not be restored in R83H channels by chemical or mutagenic addition of negative charge at S82. Thus, just as R83H-MiRP2 channels show suprathreshold activation (V1/2act ~-7 mV) (Fig. 3A, B), so, too, a double mutant carrying both the S82 P mimic S82D and R83H (S82D, R83H-MiRP2) showed V1/2act ~-12mV (Fig. 3C). Furthermore, application of PMA to channels with R83H-MiRP2 had no significant effect on the voltage dependence of activation (Fig. 3A, C); application of BIS did shift V1/2act channels to -19 mV, ~10 mV more negative than the value for wild-type channels with BIS application.

The R83H mutation confers sensitivity to inhibition by lowered intracellular pH
Lowered pH is a risk factor for some with periodic paralysis because attacks often occur with rest after exercise when serum and intracellular pH drop (19). Histidine can exist in both ionized and nonionized forms at neutral pH (pKa ~7) and can mediate pH sensitivity of proteins, including ion channels (20,21). The effects of altered intracellular pH (pHi) on wild-type and R83H-MiRP2-Kv3.4 single channels were therefore examined at -40 mV using inside-out patches from CHO cells; previously, we observed that channels with R83H-MiRP2 display lower unitary conductance than those with wild-type at physiological pHi (15).
While wild-type MiRP2-Kv3.4 channels showed little variation in unitary current amplitude over the range pHi 9.0 -7.0, R83H-MiRP2-Kv3.4 complexes showed a steep decline across the physiological range (Fig. 4A). This effect was quantified by all-points histograms for each channel type at various pHi levels and construction of proton-inhibition curves (Fig. 4B-D). The pH dependence of current suppression for wildtype MiRP2-Kv3.4 channels showed a half-maximal effect at ~ pHi 6.0, compared to ~ pHi 7.3 for R83H-MiRP2-Kv3.4 complexes (Fig. 4C, D). Dropping to pHi 7.0 from 8.5 at -40 mV reduced unitary current amplitude for channels with the R83H variant by 30% (to 1.20±0.21 pA from 1.70±0.14 pA) whereas wildtype channel amplitudes were reduced just 5% (to 1.90±0.18 pA from 2.0±0.1 pA). Kv3.4 channels with no MiRP2 showed no significant sensitivity over the pHi range of 9.0 to 6.5 (Fig. 4C, D).

MiRP2-Kv3.4 channel gating is tightly controlled by PKC phosphorylation
Wild-type MiRP2 shifts the potential for half-maximal activation of Kv3.4 channels ~ -50 mV (from V1/2act~+9 mV to -40 mV, Table 1; ref 6) to create subthreshold-activating channels; the disease-associated point change R83H suppresses this influence of MiRP2 yielding complexes with V1/2act ~-10 mV (Table 1;  ref 15). To assess the basis for this effect of R83H, we first studied neighboring S82, showing it to be an operational site for PKC phosphorylation, the only such site in the peptide, and to mediate the major effect of wild-type MiRP2 on Kv3.4 channels, that is, the ~ -50 mV shift in V1/2act. Mutation of MiRP2 S82 to A suppresses both phosphorylation of the peptide and its effects on Kv3.4 yielding V1/2act ~-7 mV. Further, an S82 to D mutation in MiRP2 allows it to serve like wild-type S82 P MiRP2 yielding V1/2act ~-42 mV.
A key role for PKC in Kv3.4 is already known; this pore-forming cx subunit has an amino-terminal inactivation domain of ~28 amino acids containing two PKC sites where phosphorylation acts to alter secondary structure and impair inactivation (22)(23)(24)(25). Thus, MiRP2-Kv3.4 channels can be regulated by action of PKC on both MiRP2 and Kv3.4 subunits, with phosphorylation leading to increased current due to activation at more negative potentials and reduced inactivation.

A mechanistic link between lowered pHi and attacks of periodic paralysis
To our surprise, the R83H change that suppressed the effects of S82 phosphorylation did not alter function by preventing phosphorylation of S82 or interfering with assembly of the MiRP2-Kv3.4 complexes ( Fig. 1; ref 6). Moreover, the effect of R83H is dominant since subthreshold function is not restored to R83H-MiRP2-Kv3.4 channels by stimulation with PKC or MiRP2 double substitution (R83H, S82D-MiRP2). However, while wild-type channels were relatively insensitive to changes in intracellular pH across the physiological range, R83H-MiRP2-Kv3.4 channels showed a significant decrease in current with lowered pHi. The most noticeable effect of the R83H mutation is the shift of ~-35 mV in activation voltage that decreases open probability at normal pH; lowering pHi further exacerbates this loss in function that may precipitate a clinical event. Although an S82 to D change mimics the effect of phosphorylated MiRP2 on Kv3.4 and suggests a negative charge is necessary and sufficient at the site, charge similarity does not appear adequate to restore native function to channels with R83H MiRP2, since lowered pH is likely to protonate the mutant histidine but further reduces similarity to wild-type. This suggests R83 may participate in important protein-protein contacts, perhaps with residues from Kv3.4.
Attacks of hypokalemic familial PP often occur during rest after exercise (26,27), as observed with one of our patients (6). While lowered serum potassium may be causative in these cases, another potential explanation is reduced pHi. During the first 5-10 s of exercise, metabolic alkalosis occurs apparently due to intracellular phosphocreatine hydrolysis, H + efflux and H + movement into sarcoplasmic reticulum (28 -33). Additional exercise leads to lower plasma and skeletal myocyte pH as lactic acid is produced, VO2 and local PCO2 increase, and H + transport processes are reversed (28,29,(31)(32)(33); acidosis facilitates oxygen dissociation from hemoglobin during exercise (31,34). In mammalian skeletal muscle pHi is reported to drop below 6.5 with exercise: 6.28 was recorded in canine gracilis muscle after 3 min of exercise (28); 6.5 was recorded in human gastrocnemius after 30 s of exercise (35); 6.4 was reported in human forearm after exercise to fatigue (36). Early in post-exercise recovery human muscle pHi is reported to become still lower (35,36). Thus, protonation of a substituted histidine in MiRP2 position 83 provides a possible link between exercise, acidosis, and the onset of paralysis. At baseline, channels with R83H MiRP2 pass less current than wild-type and as pH drops below 7.0 another ~1/3 of the current (assessed at -40 mV) will be lost.
Although sustained weakness can occur in some individuals with age, attacks of paralysis are not permanent and often manifest with an environmental demand. This type of superimposition of genetic factors predisposing to risk with a secondary environmental challenge is analogous to the apparent mechanisms for some acquired cardiac arrhythmias where events occur due to diminished repolarization capacity at baseline, exacerbated by normal (8) or increased sensitivity to drug blockade (5). Particularly relevant to this report is a missense variant in the cardiac voltage-gated sodium channel common in African Americans that operates normally at baseline but shows abnormal gain-in-function with lowered intracellular pH (18) and is linked to arrhythmia in aging adults (17) and sudden infant death syndrome (18). The sensitivity of R83H-MiRP2 channels to lowered pHi is also a possible explanation for apparently paradoxical results of treatment of some patients with the carbonic anhydrase inhibitor acetazolamide. While acetazolamide often decreases the frequency of PP attacks by causing metabolic acidosis and preventing intracellular potassium shifts (37), the agent can precipitate events in some (38). Thus, maintaining a normal pHi might prove more valuable in patients with R83H-MiRP2 than stabilizing intracellular potassium levels at the expense of pH. Acetazolamide slows the rate of recovery of normal intracellular pH during rest after exercise and reduces the absolute pH reached during exercise and recovery (36). This draws comparisons to treatment of cardiac arrhythmia, where some anti-arrhythmic drugs are proarrhythmic due to blockade of cardiac delayed rectifier potassium channels, especially IKr (8).

R83H-MiRP2 and linkage to PP
Periodic paralysis is a genetically and phenotypically heterogeneous group of episodic skeletal muscle disorders characterized by prolonged episodic attacks of weakness or paralysis (12,39). Given the overlap of symptoms, and as different forms of PP are associated with mutations in several skeletal muscle ion channel genes, a useful classification will be one based on genotyping (12,40) both as a strategy to target optimal therapy and avoid complications like those described above for acetazolamide.
PP has been associated with mutations in the L-type calcium channel gene (CACNLA3) (41,42), the voltage-gated sodium channel gene (SCN4A) (43)(44)(45)(46)(47), and MiRP2 (KCNE3) (6). The related muscle disorder myotonia congenita is linked to mutations in a chloride channel gene (CLCN1) (48,49). Andersen's syndrome, a disorder in which patients present with developmental defects, cardiac arrhythmia, and periodic paralysis, is caused by mutation of the inward rectifier potassium channel gene KCNJ7 (50). After the original report of association of KCNE3 mutation with PP (6), linkage of R83H-MiRP2 has been questioned based on observations of the genotype in 3 of 320 controls and a family with PP where it did not segregate with disease, although, that kindred had another mutation to explain the disease (51). We found the R83H variant in two small pedigrees in all individuals with PP and not in unaffected relatives, and did not observe the variant in 100 controls (6). This was suggestive for association with disease, but it was the functional effects of the amino acid change that made the case for linkage so strong (6). The possibility remains that R83H is predisposing for PP and must be present with some other genetic variant (such as a promoter variant that leads to a higher level of the R83H allele or changes the operation of some other protein); in that case, cosegregation in our kindreds could be because of linkage to the additional change or coincidence. The current report suggests that R83H-MiRP2 subunits might predispose to PP by creating channels that are sensitive to lowered intracellular pH, yielding a secondary challenge to normal muscle function from a common change in acid-base status in response to intense exercise. The data supporting the conclusion that R83H is a periodic paralysis gene are compelling: occurrence in patients from two affected families, presence of message and protein in the appropriate human tissue, interaction of MiRP2 with Kv3.4 in rat and mouse muscle, functional effects of assembly of the subunits, functional changes on expression of the mutation in muscle cells consistent with disease, and a mechanistic link between dysfunction of R83H-MiRP2-Kv3.4 channels and a known precipitating agent in some cases of PP, intracellular acidosis.

A KCNE domain critical to gating and conduction
The membrane-following region of MiRP2 contains a number of features highly conserved among KCNE peptides: a cluster of basic residues, a consensus protein kinase C (PKC) phosphorylation site, and a conserved aspartic acid (position 90 in MiRP2) (1). The basic residues are present in all five known KCNE peptides; the aspartic acid is conserved in MinK (D76), MiRP1 (D82), MiRP2 (D90) and MiRP4 (D92). In MinK and MiRP2 a D-N mutation at this position causes reduced unitary conductance and altered gating; in MinK this mutation causes Romano-Ward and Jervell and Lange-Nielsen syndromes (manifesting cardiac arrhythmia without and with sensorineural deafness, respectively) (10,14,15). Mutation of nearby S74 to L in MinK also reduces current density and is associated with cardiac arrhythmia (10,14).
Like R83 in MiRP2, MinK and MiRP1 contain lysine in the analogous site and mutant changes have similar effects. Thus, R83H-MiRP2, K69H-MinK, and K75H-MiRP1 have blunted effects compared with wild-type on cx-subunits in the resultant IKs and IKr channels (15). Further, R83H in MiRP2 affects not only complexes with Kv3.4 but the properties of MiRP2/KCNQ1 and MiRP2/HERG channels (15). While channel function is sensitive to changes in this conserved region, it should be noted that mutations do not