Polarized Axonal Surface Expression of Neuronal KCNQ Potassium Channels Is Regulated by Calmodulin Interaction with KCNQ2 Subunit

KCNQ potassium channels composed of KCNQ2 and KCNQ3 subunits give rise to the M-current, a slow-activating and non-inactivating voltage-dependent potassium current that limits repetitive firing of action potentials. KCNQ channels are enriched at the surface of axons and axonal initial segments, the sites for action potential generation and modulation. Their enrichment at the axonal surface is impaired by mutations in KCNQ2 carboxy-terminal tail that cause benign familial neonatal convulsion and myokymia, suggesting that their correct surface distribution and density at the axon is crucial for control of neuronal excitability. However, the molecular mechanisms responsible for regulating enrichment of KCNQ channels at the neuronal axon remain elusive. Here, we show that enrichment of KCNQ channels at the axonal surface of dissociated rat hippocampal cultured neurons is regulated by ubiquitous calcium sensor calmodulin. Using immunocytochemistry and the cluster of differentiation 4 (CD4) membrane protein as a trafficking reporter, we demonstrate that fusion of KCNQ2 carboxy-terminal tail is sufficient to target CD4 protein to the axonal surface whereas inhibition of calmodulin binding to KCNQ2 abolishes axonal surface expression of CD4 fusion proteins by retaining them in the endoplasmic reticulum. Disruption of calmodulin binding to KCNQ2 also impairs enrichment of heteromeric KCNQ2/KCNQ3 channels at the axonal surface by blocking their trafficking from the endoplasmic reticulum to the axon. Consistently, hippocampal neuronal excitability is dampened by transient expression of wild-type KCNQ2 but not mutant KCNQ2 deficient in calmodulin binding. Furthermore, coexpression of mutant calmodulin, which can interact with KCNQ2/KCNQ3 channels but not calcium, reduces but does not abolish their enrichment at the axonal surface, suggesting that apo calmodulin but not calcium-bound calmodulin is necessary for their preferential targeting to the axonal surface. These findings collectively reveal calmodulin as a critical player that modulates trafficking and enrichment of KCNQ channels at the neuronal axon.

Axonal rather than somatic KCNQ channels have been shown to suppress action potential firing in hippocampal CA1 neurons by regulating action potential threshold and resting membrane potential [14,17]. Computer modeling based on electrophysiological data has predicted that their axonal conductance must be 3-5 times greater than their somatic conductance in order to prevent spontaneous action potential firing in a ''near-realistic'' model of the CA1 pyramidal cell [14]. Indeed, KCNQ channels are enriched at the axonal surface with the highest concentration in the axonal initial segment (AIS) [18], the critical site for action potential initiation and modulation [19]. Localization of KCNQ channels at the AIS requires KCNQ2 and KCNQ3 interaction with ankyrin-G [6,18,20], an essential component of the AIS that maintains neuronal axon versus dendrite polarity [21]. Disrupting the normal channel localization at the AIS causes bursting and epileptiform activity [14]. Importantly, BFNC mutations in the cytoplasmic carboxy (C)-terminal tail of KCNQ2 decrease surface density of KCNQ channels at the AIS and distal axons [18]. Despite the significant implications of axonal KCNQ conductance in neuronal excitability, it is unclear how enrichment of KCNQ channels at the axonal surface is achieved.
The axon targeting signals have been shown to reside in the first 256 amino acid residues of the KCNQ2 C-terminal tail [18]. This region contains two helical domains (helices A and B) that bind to calcium (Ca 2+ ) sensor, calmodulin (CaM) [22,23] (Fig. 1A, 2A). Helix A contains the consensus CaM binding IQ motif whereas helix B mediates Ca 2+ -dependent CaM binding [22][23][24][25]. Since CaM interaction with KCNQ2 is required for functional expression of KCNQ channels in non-neuronal cells [22,24,26], we hypothesized that CaM regulates preferential targeting of KCNQ channels to the axonal surface. By utilizing a dominantnegative mutant CaM that is unable to bind Ca 2+ [27] or mutations that block or reduce CaM binding [24,26], we show that CaM interaction with the IQ motif of KCNQ2 is required for exit of KCNQ channels from the endoplasmic reticulum (ER), and their subsequent enrichment at the axonal plasma membranes in cultured hippocampal neurons.

Experimental animals
All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Illinois Urbana-Champaign in accordance with the guidelines of the U.S. National Institutes of Health (Protocols 10199, 12240). The timed-pregnant Sprague-Dawley rats were purchased from Charles River. To minimize stress and discomfort, the timed-pregnant rats were euthanized by inhalation of carbon dioxide followed by decapitation. The 18-19 day embryonic rats were quickly removed by caesarian section and decapitated. The hippocampi of embryos were dissected in ice-cold slice dissection solution containing (in mM): 10 HEPES, 82 Na 2 SO 4 , 30 K 2 SO 4 , 10 Glucose, 5 MgCl 2 (pH 7.4).

Neuronal Cell Culture and Transfection
Primary dissociated hippocampal cultures were prepared from hippocampi of 18-19 day embryonic rats as described [30] with the following modifications: Neurons were plated on 12 mm glass coverslips (Warner Instruments, 10 5 cells per coverslip), or 30 mm cell culture dishes (BD Biosciences, 7610 5 cells per dish) coated with poly L-lysine (0.1 mg/mL). Neurons were maintained in neurobasal medium supplemented with B27 extract, 200 mM Lglutamine, and 100 U/mL penicillin and streptomycin in a cell culture incubator (37uC, 5% CO 2 ) for 5-9 days in vitro. At 5-6 days in vitro, neurons were transfected with plasmids (total 0.8 mg) using Lipofectamine LTX (Invitrogen) according to manufacturer's protocol.

Pulse-Chase Assay after removal of brefeldin-A
Pulse-Chase assay after brefeldin-A (BFA) removal was performed as described [31] with the following modifications: 30 min post transfection with CD4-Q2C or HA-KCNQ3/ KCNQ2, half of the original medium per well was saved, and the transfected neurons were incubated with vehicle control (0.1% ethanol), or BFA (0.75 mg/ml) in the remaining medium and returned to the cell culture incubator. At 16 hr post-BFA treatment, the neurons were washed with fresh medium (37uC) and returned to the cell culture incubator with the saved original medium for 0, 1, 2, 4, 8, or 24 hr. The same experiments were repeated with the neurons co-transfected with GFP to visualize their entire neuronal morphology.

Image Acquisition and Quantification
Fluorescence and phase contrast images of transfected neurons were viewed using a Zeiss Axiovert 200 M inverted microscope with appropriate filter sets. The transfected neurons were examined for integrity and health using fluorescence imaging of MAP2 and/or GFP staining and differential interference contrast (DIC) imaging. If the transfected neurons had beaded or broken dendrites or axons, or damaged soma, then they were excluded from analysis. High-resolution gray scale images of healthy transfected neurons were acquired using 20X, 40X, or 63X objectives with a Zeiss AxioCam HRm Camera and Axiovert software and saved as 16-bit ZVI and TIFF files. Within one experiment, the images were acquired using the same exposure time to compare the fluorescence intensity of the neurons transfected with different constructs, and stored with no further modifications.
The fluorescence intensity profiles of the soma and the major axonal and dendritic processes from the 16-bit TIFF files were quantified using ImageJ Software (National Institutes of Health, USA, http://rsb.info.nih.gov/ij) as previously described [18]. The dendrites were identified as MAP2-positive processes from the transfected neuron (Figure S1, Figure S2) or the processes that were absent for the AIS markers, phospho IkBa Ser32 (14D4) [32] or ankyrin-G [21] in the GFP-transfected neuron ( Figure S3, Figure S4). The axon was identified as a MAP2-negative process from the transfected neuron (Figure S1, Figure S2) or a process that were labeled for the AIS markers in the GFP-transfected neuron ( Figure S3, Figure S4). Using ImageJ, 1 pixel-wide line segments were traced along all dendrites and the portions of axons as stated below and previously described [18,30]. Using DIC imaging and MAP2 staining or GFP fluorescence, the perimeter of the transfected soma was manually traced. Images displaying severe variability in focal plane were excluded from analysis. Regions where fasciculation or overlapping processes occurred were excluded from analysis.
We have observed that the axons originated directly from the soma in majority of the transfected neurons ( Fig. 1, Figure S2B), while some neurons had their axons originated from their proximal dendrites ( Figure S1, Figure S2B), consistent with the heterogeneity in the origin and location of the AIS [33][34][35]. Hence, the beginning of the axon was defined as the point where the axon originated from the soma or a proximal dendrite. The AIS was identified as an initial segment in the axon that was labeled with the AIS markers described above ( Figure S2B). Our manual tracing along the axons revealed that the soma-or CD4-Q2C were reduced to nearly 0 by L339R, I340E, and A343D mutations. (E) Background subtracted, mean intensity of surface CD4 fluorescence in the AIS, distal axons, soma, and major dendrites. AU, arbitrary unit. The sample number for each construct used in (D, E) was as follows: WT (n = 18), L339R (n = 20), I340E (n = 12), A343D (n = 25), and untransfected (n = 20). Ave 6 SEM (*p,0.05, **p,0.01, ***p,0.001). doi:10.1371/journal.pone.0103655.g002 dendrite-derived AIS started at 4.462.6 mm and ended at 29.860.7 mm from the beginning of the axon in cultured hippocampal neurons (n = 8, Figure S2B), consistent with the previous reports on the AIS length to be about 30 mm [33][34][35]. Thus, the background-subtracted mean fluorescence intensity of the soma, the axon within 0-30 mm of the beginning of the axon (AIS), the axon between 50-80 mm from the beginning of the axon (distal axon), and the major primary dendrites up to the disappearance of MAP2 staining (dendrite), were obtained for determination of the surface ''AIS/Axon'' and ''Axon/Dendrite'' ratios as previously described [18,30] ( Figure S2).
The color-merged images as well as inverted gray scale images were generated in Photoshop (Adobe Systems) as described [18]. Pseudo-color images were generated in ImageJ. For camera lucida drawings, the axon (identified by ankyrin-G or phospho IkBa Ser32 (14D4) staining of the AIS, or by the lack of MAP2 staining) from the inverted gray-scale image of the GFP-transfected neuron was traced using the neuronJ plugin for ImageJ as described [36] ( Figure S3, Figure S4). Similar to the previously published camera lucida drawings of transfected hippocampal neurons [37], the cellular processes of the GFP-transfected neurons were colored in gray to increase the visibility of the axon, which was traced in black. For camera lucida drawings of the neurons that were not transfected with GFP, the cellular processes of the CD4-or HApositive neurons were colored in gray ( Figure S1, Figure S2).

Immunoprecipitation
Immunoprecipitation was performed as described [22,26,39] with the following modification. HEK293T cells were plated on 60 mm cell culture dishes (BD Biosciences, 7610 5 cells per dish) maintained in Minimal Essential Medium containing 10% Fetal Calf Serum, 2 mM glutamine, 100 U/mL penicillin and 100 U/ mL streptomycin at 37uC and 5% CO 2 . At 24 hr post plating, the cells were transfected with plasmids (total 1.7 mg) using FuGENE6 transfection reagent (Promega) according to manufacturer's protocol. At 24 hr post transfection, the cells were washed with ice-cold PBS and solubilized in ice-cold immunoprecipitation (IP) buffer containing (in mM): 20 Tris-HCl, 100 NaCl, 2 EDTA, 5 EGTA, 1% Triton X-100 (pH 7.4) supplemented with Halt protease inhibitors (Thermo Fisher Scientific). The cells in IP buffer were incubated on ice for 15 min, followed by centrifugation at 17,0006g for 15 min at 4uC. The protein concentration of the resulting supernatant was determined using the BCA kit (Pierce). The lysate containing equal amount of proteins were incubated first with Protein A/G agarose beads (50 mL, Santa Cruz) for 1 hr 4uC. The pre-cleared supernatant was then incubated overnight at 4uC with Protein A/G-agarose beads (50 mL) and mouse anti-CD4 antibody (5 mg) or rabbit anti-KCNQ2 antibody (5 mg). After four washes with IP buffer, the immunoprecipitates were eluted with SDS sample buffer by incubating at 90uC for 5 min, and subjected to immunoblot analysis with anti-CD4 (1:500 dilution), anti-CaM (1:200 dilution), anti-KCNQ2 (1:200 dilution), or anti-b-actin (1:500 dilution).

Electrophysiology
Coverslips containing dissociated rat hippocampal neurons were transfected, and 24-48 hr after transfection, the coverslips were transferred to the whole-cell patch clamp recording chamber in external solution containing (in mM): 126 NaCl, 3 KCl, 2 CaCl 2 , 2 MgSO 4 , 1 NaH 2 PO 4 , 25 NaHCO 3 and 14 Dextrose, bubbled with 95% O 2 and 5% CO 2 (pH 7.4, 305-315 mOsm). The untransfected or GFP-positive pyramidal neurons were visually identified using an upright fluorescence microscope (Zeiss Axioscope) and the whole-cell patch clamp recordings were carried out immediately. All recordings were performed to obtain current clamp mode at room temperature (23-25uC) in the presence of the fast synaptic transmission blockers 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 20 mM), DL-2-amino-5-phosphonopentanoic (DL-AP5; 100 mM) and bicuculline (20 mM) in external solution. Recording pipettes were pulled from glass capillaries with an outer diameter of 1.5 mm on a micropipette puller (P-97; Sutter Instruments), and had a resistance of 3-5 MV when filled with internal solution containing (in mM): 130 KMeSO 4 , 10 KCl, 10 HEPES/K-HEPES, 2 MgSO 4 , 0.5 EGTA and 3 ATP (pH 7.3, 285-295 mOsm). Current-clamp recordings were performed as previously described [40] with the following modifications. Neurons were held at -60 mV. Action potential firing rates (Hz) were measured upon delivering constant current pulses of 500 ms in the range 0 to 200 pA, and were averaged from 3 to 5 individual sweeps per current injection. Neurons were eliminated from further analysis if the access resistance changed by more than 20% over the recording period. Whole-cell recordings were made using a Multiclamp 700B amplifier (Molecular Devices). Recordings were filtered at 2 kHz and digitized at 10 kHz. Data was acquired and analyzed with a Digidata 1440A interface (Molecular Devices) and the pClamp suite of software (version 10.2; Molecular Devices). Recording analyses were performed using Clampfit software (version 10.2; Molecular Devices).

Statistical Analyses
All fluorescence intensity and electrophysiology analyses were reported as mean 6 SEM. ANOVA and post-ANOVA Tukey's multiple comparison tests were performed to identify the statistically significant difference between groups of three or more, whereas the Student t test was performed for groups of two by Microsoft Excel with QI Macros 2013 plug-in. A priori value (p), 0.05 was considered statistically significant. The number of separate transfected cells for immunostaining and electrophysiology was reported as the sample size n.

Fusion of KCNQ2 C-terminal tail preferentially targets CD4 to the axonal surface
To test whether CaM regulates preferential targeting of KCNQ channels to the axonal surface, we transfected rat dissociated hippocampal cultured neurons with the KCNQ3 subunit tagged with an extracellular hemagglutinin (HA) epitope (HA-KCNQ3) and KCNQ2 at 5-6 days in vitro (DIV) (Fig. 1A) when CaM is abundantly expressed (Fig. 1B). Since antibodies that recognize extracellular domains of endogenous KCNQ2 and KCNQ3 are not available, surface immunostaining of HA-KCNQ3 proteins has been used to demonstrate enrichment of HA-KCNQ3/ KCNQ2 channels both at the AIS and more distally on axons [18]. The surface expression and function of HA-KCNQ3 has also been demonstrated in Xenopus oocytes [28]. The axon was identified as a MAP2-negative process from the transfected neuron (Fig. 1D, Figure S1). Surface immunostaining at 48 hr posttransfection revealed that HA-KCNQ3/KCNQ2 channels were preferentially expressed on the surface of axons originating directly from the soma compared to the somatodendritic surface (Fig. 1D). Consistent with the heterogeneity in the origin and location of the AIS [33][34][35], HA-KCNQ3/KCNQ2 channels were also enriched at the surface of axons originating from the proximal dendrites in some neurons ( Figure S1). They were also highly concentrated at the surface of the AIS (Fig. 1D-E, Figure S1), which is consistent with previous reports of the distribution and function of endogenous KCNQ2 and KCNQ3 [5,6,10,14,41,42].
The cluster of differentiation 4 (CD4), which is a non-neuronal single transmembrane glycoprotein, was used as a well-established trafficking reporter system [18,30,31] in which the C-terminal tail of KCNQ2 was fused to the C-termini of CD4 (CD4-Q2C) (Fig. 1C). Consistent with a previous report [18], we found that transfected CD4 proteins were uniformly distributed on the plasma membrane of hippocampal neurons whereas CD4-Q2C chimeric proteins were preferentially targeted to the axonal surface ( Fig. 1F-G, Figure S2C). To quantify the polarized surface expression, we calculated the ratio of the mean surface fluorescence intensity of the distal axon to the major dendrites (Axon/ Dendrite) [18,30] and found that the surface ''Axon/Dendrite'' ratio was 0.560.1 for CD4, and 1.660.3 for CD4-Q2C ( Figure  S2E). Axonal enrichment of CD4-Q2C ( Fig. 1F-G) closely mirrored the polarized axonal distribution of the intact HA-KCNQ3/KCNQ2 channels (Fig. 1D-E). Furthermore, the E810A and D812A mutations in KCNQ2 C-terminus, which disrupt ankyrin-G binding to KCNQ2 [6,18,20], reduced the surface ''AIS/Axon'' ratio of CD4-Q2C to 1, but had no effect on the surface ''Axon/Dendrite'' ratio ( Figure S2). These results together suggest that preferential targeting of CD4-Q2C proteins to the axonal surface precedes their ankyrin-G-dependent enrichment at the AIS surface.
Fusion of KCNQ2 C-terminal tail deficient in CaM binding fails to preferentially target CD4 to the axonal surface A previous report [18] as well as our results (Fig. 1D-E, Figure  S2) indicated that the membrane proximal region of the KCNQ2 C-terminal tail, upstream of the ankyrin-G binding domain, was sufficient to preferentially target CD4 to the axonal surface. This region of KCNQ2 contains helices A and B, which bind to CaM [22,23] (Fig. 2A). To test whether CaM interaction with KCNQ2 regulates the targeting of CD4-Q2C to the axonal surface, we introduced point mutations in helix A of KCNQ2 including two BFNC mutations, L339R and R353G [43] (Fig. 2A). Mutations in the CaM-binding consensus IQ motif (L339R, I340E, and A343D) abolished co-immunoprecipitation of CD4-Q2C with CaM from HEK293T cell lysate (Fig. 2B), consistent with their ability to disrupt KCNQ2 interaction with CaM [24,26].
To determine the extent to which these mutations affect targeting of CD4-Q2C to the axonal surface, neurons were cotransfected with GFP, which allows visualization of all neurites. The axon was identified by immunostaining for the AIS marker phospho IkBa Ser32 (14D4) (Fig. 2, Figure S3). Surface immu-nostaining revealed that mutations in the CaM-binding consensus IQ motif (L339R, I340E, and A343D) decreased surface expression of CD4-Q2C at the AIS and distal axons to background fluorescence intensities of untransfected neuronal dendrites (Fig. 2C, E) and reduced the surface ''Axon/Dendrite'' ratio to nearly 0 (Fig. 2D). The surface ''AIS/Axon'' ratios for these mutants were not calculated because the majority of these mutants had background-subtracted mean fluorescence values of 0 at the AIS and distal axon. Furthermore, these mutant proteins were intracellularly excluded from the AIS and distal axons ( Fig. 3A-C) and displayed reduced somatodendritic expression (Fig. 3B).
The R353G mutation located distal to the IQ motif ( Fig. 2A) has been shown to moderately reduce CaM binding to KCNQ2 in HEK293T cells [24,26]. Consistent with previous reports, the same mutation decreased but did not abolish co-immunoprecipitation of CD4-Q2C with CaM (Fig. 4A). In contrast to mutations in the IQ motif ( Fig. 2-3), the R353G mutation increased somatodendritic surface expression of CD4-Q2C without affecting axonal surface expression (Fig. 4B, D), thereby decreasing the surface ''Axon/Dendrite'' ratio to 1 (Fig. 4C). The surface CD4-Q2C R353G proteins also displayed a punctate distribution in the soma and dendrites compared to the CD4-Q2C WT (Fig. 4B). The surface ''AIS/Axon'' ratio was unaffected by the R353G mutation compared to wild-type CD4-Q2C (Fig. 4C). Permeabilized immunostaining revealed that total (surface and intracellular) expression of the R353G mutant CD4-Q2C proteins increased in the soma but modestly decreased in the axons (Fig. 5A-B). These results together indicate that fusion of KCNQ2 C-terminal tail deficient in CaM binding fails to preferentially target CD4 to the axonal surface.
Mutations that block CaM binding impair trafficking of CD4-Q2C from the ER to the axon KCNQ2 subunits deficient in CaM binding are retained in the ER of HEK293T cells [24,26], suggesting the possibility that preferential targeting of CD4-Q2C proteins to the axonal surface may involve their CaM-dependent trafficking from the ER to the axon in hippocampal neurons. To test this hypothesis, we treated the transfected neurons at 30 min post-transfection with brefeldin-A (BFA), which is a reversible inhibitor of anterograde transport from the ER to the Golgi complex [44] (Fig. 6A). Permeabilized immunostaining of transfected neurons treated with vehicle control for 16 hr revealed that total (surface and intracellular) expression of wild-type and R353G mutant CD4-Q2C proteins were found throughout the neuron, whereas expression of I340E mutant proteins were confined to the perinuclear region of the soma and proximal dendrites (Fig. 6B-D). In contrast, BFA treatment for 16 hr caused newly synthesized wild-type and all mutant proteins to accumulate at perinuclear regions in the soma and proximal dendrites (Fig. 6B-D). Wild-type and R353G mutant proteins but not I340E mutant proteins were also found in the distal dendrite of some transfected neurons at 16 hr post-BFA treatment (Fig. 6B-D). Since rough ER is prominent in the neuronal soma and proximal dendrites whereas smooth ER predominates in the distal dendrites and dendritic spines [45][46][47], these results suggest that newly synthesized CD4-Q2C proteins are retained mostly in the rough ER after BFA treatment.
BFA was then removed and the neurons were fixed at various time points to ''chase'' the trafficking of CD4-Q2C proteins from the ER by permeabilized immunostaining (Fig. 7A-B). The wildtype CD4-Q2C proteins gradually appeared in a diffused distribution as well as in punctate structures in the AIS and distal axons upon BFA removal (Fig. 7A, C). In contrast, I340E mutant proteins were mostly absent from the AIS and distal axon for the duration of the 8 hr BFA washout (Fig. 7A, C), consistent with our previous results for the I340E mutation, which blocked both surface and intracellular expression of CD4-Q2C throughout the axon (Fig. 2-3, 6C). In contrast, the R353G mutant proteins were found in the AIS and distal axon, albeit at reduced levels compared to wild-type CD4-Q2C at 8 hr post-BFA washout (Fig. 7A, C). BFA removal also caused wild-type and all mutant proteins to distribute diffusely in the dendrites with occasional accumulation in punctate structures at time points earlier than in the axons (Fig. 7A, C). Interestingly, the expression of I340E mutant CD4-Q2C was initially reduced in the dendrites compared to wild-type and R353G mutant proteins (Fig. 7A, C). However, the level of R353G mutant proteins at 8 hr post BFA washout was similar to the amount of I340E mutant protein in dendrites, which was approximately half the amount of wild-type CD4-Q2C (Fig. 7A, C). These results together suggest that disruption of CaM binding to KCNQ2 impairs the trafficking of CD4-Q2C proteins from the ER to the AIS and distal axons.
Disruption of CaM binding to KCNQ2 inhibits enrichment of HA-KCNQ3/KCNQ2 channels at the axonal surface Our results with CD4-Q2C proteins (Fig. 2-7) suggest that enrichment of intact KCNQ channels at the axonal surface may To test this hypothesis, we first examined the effect of A343D and R353G mutations on the interaction between CaM and intact KCNQ2 (Fig. 8A). Despite transfecting the same amount of each expression plasmid in HEK293T cells, we found that expression of the A343D mutant KCNQ2 subunits was consistently lower compared to wild type or the R353G mutant KCNQ2 subunits (Fig. 8A). Nonetheless, the immunoprecipitated amount of wild type and mutant KCNQ2 subunits using the same quantity of anti-KCNQ2 antibodies were nearly equal (Fig. 8A). This is most likely because the amount of antibodies used for immunoprecipitation was far less sufficient to immunoprecipitate all of transfected KCNQ2 proteins from the HEK lysate due to high level of expression of KCNQ2 proteins including A343D mutant proteins. Importantly, the A343D but not the R353G mutation abolished co-immunoprecipitation of KCNQ2 with CaM from HEK293T cell lysate (Fig. 8A).
Next we performed surface immunostaining in hippocampal neurons transfected with GFP, HA-KCNQ3, and wild-type KCNQ2 or mutant KCNQ2 (A343D and R353G). Robust surface expression of wild-type HA-KCNQ3/KCNQ2 channels was detected at the AIS and distal axons compared to dendrites (Fig. 8B, Figure S4) [18]. In contrast, the A343D mutation, which blocked CaM binding to KCNQ2 (Fig. 8A) [24,26], abolished surface and intracellular expression of HA-KCNQ3/KCNQ2 channels at the AIS and distal axons (Fig. 8B, D, 9A-B) and reduced the surface ''Axon/Dendrite'' ratio to below 1 (Fig. 8C). The A343D mutation also decreased surface expression in the soma and dendrites compared to the wild type (Fig. 8B, D). Despite a modest reduction of CaM binding to KCNQ2 by the R353G mutation (Fig. 8A) [24,26], the R353G mutant channels were localized to the AIS and axonal surface to a similar extent as the wild-type channels (Fig. 8B-D). Although the R353G mutation increased somatodendritic surface expression of CD4-Q2C (Fig. 4), the same mutation in full-length KCNQ2 had no effect on the surface and intracellular expression of HA-KCNQ3/KCNQ2 channels throughout the neuron compared to the wild type ( Fig. 8B, D, 9A, B). These results indicate that enrichment of HA-KCNQ3/KCNQ2 channels at the axonal surface was impaired by the A343D mutation in IQ motif that blocks CaM binding.

Disruption of CaM binding to KCNQ2 impairs trafficking of HA-KCNQ3/KCNQ2 channels from the ER to the axon
To examine whether the A343D mutation enhances the retention of HA-KCNQ3/KCNQ2 channels in the ER, we first examined their colocalization with the CD4 proteins harboring a KDEL ER retention signal (CD4-KDEL) [29,48,49]. The CD4-KDEL proteins were concentrated at the perinuclear region of the soma and displayed diffused distribution in the proximal and distal dendrites (Fig. 10A), consistent with the continuous network of the ER from the soma to the dendrites in hippocampal neurons [45,47]. Wild type and all mutant HA-KCNQ3/KCNQ2 channels displayed significant colocalization with CD4-KDEL proteins in the soma and dendrites (Fig. 10A), suggesting that majority of the channels reside in rough and smooth ER of hippocampal cultured neurons. The wild type and R353G mutant channels were also found at the AIS and distal axons, regions where the A343D mutant channels and the CD4-KDEL proteins were absent (Fig. 10A), indicating that the A343D mutation in the IQ motif of KCNQ2 enhances the localization of HA-KCNQ3/ KCNQ2 channels in the ER.
To test whether CaM binding to KCNQ2 promotes trafficking of HA-KCNQ3/KCNQ2 channels from the ER to the axon, we performed pulse-chase experiments after removal of BFA. At 16 hr treatment with vehicle control, the wild-type and R353G mutant HA-KCNQ3/KCNQ2 channels were expressed throughout the neuron including axons whereas the A343D mutant channels were found in the soma and dendrites but not axons (Fig. 10B), consistent with the total expression of these channels (Fig. 9, 10A). At 16 hr post BFA treatment, newly synthesized wild-type and mutant channels accumulated in the soma and proximal dendrites (Fig. 10B, Figure S5). Although they were found in distal dendrites, they were absent from the distal axons after BFA treatment (Fig. 10B, Figure S5).
Upon BFA removal, the wild-type HA-KCNQ3/KCNQ2 channels and the R353G mutant channels gradually appeared in a diffused distribution as well as in punctate structures in the AIS and distal axons for the duration of the 8 hr BFA washout ( Fig. 11A-C, Figure S5). In contrast, the A343D mutant channels displayed minimal expression at the AIS and distal axons for the duration of the 8 hr BFA washout ( Fig. 11A-C, Figure S5). Wildtype and mutant HA-KCNQ3/KCNQ2 channels were also found at the distal dendrites for the duration of the 8 hr BFA washout ( Fig. 11B-C, Figure S5), consistent with their steady-state localization in the ER (Fig. 10A). Although the expression of the R353G mutant channels appeared slightly elevated at the dendrites compared to the A343D mutant channels at 4 hr post BFA washout, the average dendritic expression of the wild-type channels was not statistically different from that of the R353G and A343D mutant channels by 8 hr post BFA washout (Fig. 11B-C). These results together indicate that disruption of CaM interaction with KCNQ2 by the A343D mutation impairs the trafficking of HA-KCNQ3/KCNQ2 channels from the ER to the axon.   Hippocampal neuronal excitability decreases upon expression of wild-type KCNQ2 but not mutant KCNQ2 deficient in CaM binding We next investigated the physiological relevance of CaMdependent axonal enrichment of KCNQ channels. Previous studies using whole-cell patch clamp recording of action potentials have demonstrated that axonal rather than somatic KCNQ channels suppress hippocampal neuronal excitability [14,17]. Furthermore, we showed that the wild-type KCNQ2 but not the A343D mutant KCNQ2 enriched HA-KCNQ3 at the axonal surface in hippocampal neurons cultured at 7-8 DIV (Fig. 8). Since expression of exogenous KCNQ2 has been previously shown to increase KCNQ current in hippocampal neurons cultured at 10 DIV when a relatively low level of native KCNQ2/3 channels are present [42], we hypothesized that expression of wild-type KCNQ2 subunits would reduce action potential firing by increasing axonal surface expression of both homomeric KCNQ2 channels and heteromeric channels formed with endogenous KCNQ3 subunits (Fig. 12A). In addition, we further hypothesized that expression of KCNQ2-A343D subunits deficient in CaM binding would have little effect on action potential firing, owing to their inability to exit the ER and express on the axonal surface (Fig. 12A). To test this hypothesis, we performed whole-cell patch clamp recording of action potentials as described [40] in cultured hippocampal neurons (6)(7)(8). Transient expression of wildtype KCNQ2 together with GFP decreased action potential firing rates compared to untransfected neurons for all current injections from 50 pA and neurons expressing GFP for all current injections from 100 pA (Fig. 12B-C). These results suggest that the transfected wild-type KCNQ2 subunits form functional homo- meric channels and/or heteromeric channels with endogenous KCNQ3 subunits, yielding outward K + current in transfected neurons. The neurons expressing wild-type KCNQ2 displayed a 26-30% decrease in the mean firing frequency induced by 100 pA current injection (19.762.8 Hz, p,0.05) compared to untransfected neurons (28.062.2 Hz) and GFP-transfected neurons (26.761.8 Hz) (Fig. 12D), which is consistent with wild-type KCNQ2 exerting its effect in the axon rather than in the soma [14,17].
In contrast to the expression of wild-type KCNQ2 which reduced action potential firing, expression of A343D mutant KCNQ2 and GFP had no effect on action potential firing frequency for all current injections as it was not statistically different from the firing rates displayed by untransfected neurons or neurons expressing GFP (29.762.2 Hz at 100 pA injection, p. 0.05, Fig. 12B-D). These data indicate that the A343D mutation completely inhibited the reduction in hippocampal neuronal excitability induced by expression of wild-type KCNQ2. Since the A343D mutation of KCNQ2 abolished axonal surface expression of heteromeric HA-KCNQ3/KCNQ2 channels (Fig. 8), these data indicate that the impaired targeting to the axonal surface by the A343D mutation of KCNQ2 has a functional consequence on how the exogenously expressed KCNQ2 subunits affect neuronal excitability.
We hypothesized that expression of KCNQ2 subunit harboring the BFNC R353G mutation would reduce the action potential firing to a similar extent as expression of wild-type subunits (Fig. 12A), since the HA-KCNQ3/KCNQ2-R353G mutant channels were enriched at the AIS and axonal surface to a similar extent as the wild-type channels (Fig. 8). To our surprise, neurons Figure 11. The A343D mutation impairs HA-KCNQ3/KCNQ2 trafficking from the ER to the axon. (A-B) Pulse-chase assay of wild-type (WT) or mutant (A343D and R353G) HA-KCNQ3/KCNQ2 channels after BFA washout. Permeabilized immunostaining was performed to visualize total (surface and intracellular) expression of HA-KCNQ3/KCNQ2 channels in the soma and AIS (A) as well as in the distal axons and dendrites (B) at indicated time points post-BFA removal (inverted images). The axon was identified by the AIS marker phospho IkBa Ser32 (14D4), whereas neuronal soma and dendrites were visualized by MAP2 immunostaining. Arrows indicate the AIS. Scale bars: 10 mm. (C) Background subtracted, mean intensity of total HA fluorescence in the AIS, distal axon, and major dendrites. Upon BFA removal, the A343D mutation but not the R353G mutation markedly reduced the appearance of HA-KCNQ3/KCNQ2 at the AIS and distal axon for the duration of the 8 hr BFA washout. The sample numbers were (n = 8-22) per time point for each construct. AU, arbitrary unit. Ave 6 SEM (*p,0.05, **p,0.01, ***p,0.001). doi:10.1371/journal.pone.0103655.g011 expressing R353G mutant KCNQ2 and GFP displayed a firing frequency that was indistinguishable from untransfected neurons and neurons expressing GFP (Fig. 12B-D). However, the difference in firing rates between neurons expressing R353G mutant KCNQ2 and wild-type KCNQ2 at all current injection amplitudes did not reach statistical significance (p = 0.051-0.093 at 90-200 pA current injection) except at 170 pA (p = 0.04) (Fig. 12B-D). These results indicate that the R353G mutation is sufficient to partially, but not fully, inhibit the reduction in the action potential firing rate caused by the expression of wild-type KCNQ2, even though the same mutation had no effect on axonal enrichment of surface HA-KCNQ3/KCNQ2 channels (Fig. 8). Membrane capacitance, input resistance, and resting membrane potential were unaffected by expression of GFP or KCNQ2 proteins compared to untransfected control (Table 1). Figure 12. Expression of wild-type KCNQ2 but not KCNQ2-A343D or KCNQ2-R353G decreases neuronal excitability. (A) Hypothesis by which exogeneously expressed KCNQ2 subunits affect neuronal excitability. At an early stage of hippocampal culture, the level of endogenous KCNQ2/KCNQ3 channels is low [42]. Transfection of wild-type KCNQ2 or R353G mutant KCNQ2 would reduce action potential firing by increasing axonal surface expression of KCNQ2-containing homomeric channels and heteromeric channels formed with endogenous KCNQ3 subunits. In contrast, expression of A343D mutant KCNQ2 subunits deficient in CaM binding would have little effect on action potential firing due to their inability to exit the ER and express on the axonal surface. Coexpression of Ca 2+ -insensitive CaM modestly reduces enrichment of HA-KCNQ3/KCNQ2 channels at the axonal surface Coexpression of Ca 2+ -insensitive CaM has been reported to moderately increase the ER retention of KCNQ2 by 20% in HEK293T cells [24,26], suggesting that enrichment of KCNQ channels at the axonal surface may be regulated by Ca 2+ -bound CaM. To test this, we performed surface immunostaining of HA-KCNQ3/KCNQ2 channels in cultured hippocampal neurons that were cotransfected with either wild-type CaM or dominantnegative mutant CaM1234 (Fig. 13). Although CaM1234 is unable to bind Ca 2+ [27], it has been shown to associate with KCNQ2 [22]. Given that CaM is abundantly expressed in cultured hippocampal neurons at 5 DIV, which is when transfections were performed (Fig. 1B), we hypothesized that coexpression of CaM1234 would displace endogenous CaM and make the CaM complex Ca 2+ -insensitive [27].
As previously demonstrated (Fig. 1D-E, 8) [18], HA-KCNQ3/ KCNQ2 channels were preferentially concentrated at the AIS and distal axons compared to the soma and dendrites in neurons cotransfected with control plasmid pcDNA3 (Fig. 13A, C, E). Coexpression of CaM1234 moderately decreased their surface expression at the AIS and distal axons but not at dendrites (Fig. 13A, D-E), leading to a 33% reduction in the surface ''Axon/Dendrite'' ratio (CaM1234 = 2.560.3) compared to control plasmid (pcDNA3 = 3.960.3, Fig. 13C). In contrast, coexpression of wild-type CaM had no effect on the surface expression of HA-KCNQ3/KCNQ2 channels at the AIS, axon, or dendrites (Fig. 13A, C-E). Interestingly, their surface but not intracellular expression in the soma was enhanced by coexpression with wildtype CaM or CaM1234 compared to pcDNA3 control (Fig. 13D-E). Notably, coexpression of wild-type CaM and CaM1234 had no apparent effect on neuronal polarity as evidenced by strong immunolabeling of MAP2 at the soma and dendrites as well as ankyrin-G at the AIS (Fig. 13B). Since the surface ''Axon/ Dendrite'' ratio of HA-KCNQ3/KCNQ2 channels was not reduced to 1 by coexpression of CaM1234 (Fig. 13C), our findings suggest that interaction with apoCaM but not Ca 2+ -bound CaM is necessary for preferential targeting of HA-KCNQ3/KCNQ2 channels to the axonal surface (Fig. 13F).

Discussion
The long C-terminal tail of KCNQ2 is a multi-modal region [50,51] that mediates assembly [52,53], localization at the AIS and distal axon [6,18,20], trafficking [24,26], and interaction with multiple signaling and adaptor proteins [22,23,[54][55][56][57]. Based on our new findings, we propose that CaM binding to the IQ motif of the KCNQ2 C-terminal tail is required for preferential targeting of CD4-Q2C chimeric proteins and intact KCNQ2/KCNQ3 channels to the axonal surface in hippocampal neurons and mediates their trafficking from the ER to the axon.
The role of CaM in the enrichment of KCNQ channels at the axonal surface The L339R, I340E, and A343D mutations at the IQ motif of KCNQ2 have been shown to disrupt KCNQ2 interaction with CaM [24,26]. We have shown that these same mutations abolish CaM binding to CD4-Q2C chimeric proteins as well as their axonal surface expression (Fig. 2). Although CaM can still interact with KCNQ3 subunits [23,58], the A343D mutation, which abolished CaM binding to KCNQ2 (Fig. 8A), prevented surface and intracellular expression of intact HA-KCNQ3/KCNQ2 channels at the AIS and distal axons (Fig. 8-9), consistent with a recent study reporting altered neuronal distribution of heteromeric channels by the I340A mutation [59]. In support of our trafficking results, transient expression of wild-type KCNQ2 significantly reduced action potential firing rates whereas expression of A343D mutant KCNQ2 did not (Fig. 12). These correlative findings suggest that enrichment of CD4-Q2C and HA-KCNQ3/KCNQ2 channels at the axonal surface requires CaM interaction with the IQ motif of KCNQ2.
In contrast, the BFNC R353G mutation located distal to the IQ motif modestly reduced but did not abolish CaM binding to KCNQ2 (Fig. 8A) and had minimal effect on the polarized axonal surface expression of HA-KCNQ3/KCNQ2 channels (Fig. 8).
Since the R353G mutation in KCNQ2 reduces CaM binding to heteromeric KCNQ2/KCNQ3 channels only by 20% [24,26], our results suggest that a greater degree of impairment in CaM binding to KCNQ2 is needed to prevent targeting of heteromeric channels to the axonal surface. Recently, the S511D mutation in KCNQ2 has been shown to inhibit CaM binding without affecting surface expression of homomeric channels and KCNQ2/KCNQ3 channels in non-neuronal cells [60]. However, the S511D mutation does not restore the surface expression of I340E and A343D mutant channels [60], which is consistent with our findings that KCNQ2 mutants deficient in CaM binding not only block axonal surface expression of CD4-Q2C proteins and HA-KCNQ3/KCNQ2 channels but also reduce their somatodendritic surface expression (Fig. 2, 8). Curiously, expression of R353G mutant KCNQ2 did not reduce action potential firing frequency to the same extent as expression of wild-type KCNQ2 (Fig. 12) although HA-KCNQ3/KCNQ2-R353G mutant channels were localized to the axonal surface to a similar extent as the wild-type channels (Fig. 8). This partial effect in neuronal excitability (Fig. 12) could be due to the reduced M-current density of R353G mutant KCNQ2 channels caused by their lower affinity to Phosphatidylinositol 4,5-bisphosphate (PIP 2 ) [61], an essential cofactor for M-current function [62][63][64].  We have also observed that the I340E mutation abolished the trafficking of CD4-Q2C proteins from the ER to the axons and dendrites (Fig. 7). Consistent with this result, the A343D mutation but not R353G mutation in KCNQ2 severely inhibited the trafficking of HA-KCNQ3/KCNQ2 channels from the ER to the AIS and distal axons (Fig. 11, Figure S5) where the CD4-KDEL proteins containing the ER retention signal were absent (Fig. 10A). Interestingly, wild-type and all mutant HA-KCNQ3/KCNQ2 channels were similarly distributed to the dendrites over the 8 hr time course following BFA removal (Fig. 11, Figure S5) and colocalized with the CD4-KDEL proteins (Fig. 10A). Such a distribution likely represents the diffusion and/or retention of the channels in the continuous network of the ER from the soma to the dendrites [45,47] as well as golgi outposts [65] in hippocampal neurons. Interestingly, the A343D mutation but not the R353G mutation reduced dendritic surface expression of KCNQ2 (Fig. 8), suggesting that the A343D mutation may also hinder channel exit from the ER to the plasma membrane in dendrites. These results collectively support the idea that disruption of CaM binding to the IQ motif of KCNQ2 retains intact heteromeric channels in the ER of cultured hippocampal neurons.
Importantly, coexpression of CaM1234 modestly reduced but did not abolish enrichment of HA-KCNQ3/KCNQ2 channels at the axonal surface (Fig. 13), suggesting that KCNQ2 binding to apoCaM but not Ca 2+ -bound CaM is required for targeting the channels to the axonal surface. Since CaM regulates the activity of many key ion channels and signaling proteins that modulate neuronal excitability [66][67][68][69][70][71], the effect of CaM1234 could be indirect. However, our results are consistent with the previous findings in HEK293T cells that CaM1234 modestly increases ER retention of wild-type KCNQ2 by 20% [24,26]. Given that KCNQ channel function can be potentiated in a Ca 2+ -dependent manner during activity-dependent, homeostatic intrinsic plasticity in hippocampal CA1 pyramidal neurons [72], the interaction between Ca 2+ -bound CaM and helices B of KCNQ2 and KCNQ3 [22][23][24][25]39,58,73] may mediate exocytosis or stabilization of KCNQ channels at the axonal surface upon Ca 2+ influx, in addition to Ca 2+ -dependent modulation of their current [39].
Considering a very low level of intracellular Ca 2+ in dissociated hippocampal cultured neurons at rest [74], we propose that apoCaM proteins bind simultaneously to helices A and B of KCNQ2 and KCNQ3 at the ER, with the IQ motif in helix A of KCNQ2 being the dominant domain for apoCaM interaction (Fig. 13F). Such an interaction would promote association of the channels with coat protein complex II (COPII)-coated vesicles at the ER exit sites for anterograde transport to the Golgi complex [75]. After exiting the ER, KCNQ channels can be preferentially targeted to the axonal surface by multiple trafficking pathways, including direct sorting to the axons [30,76], selective endocytosis and retention at the somatodendritic compartment [31,[77][78][79], or transcytosis from dendrites to axons [80][81][82]. The channels are then concentrated at the AIS by their interaction with ankyrin-G [6,18,20]. Further investigation may reveal how CaM regulates one of these trafficking pathways underlying polarized targeting to the axons.
Comparison between CD4-Q2C proteins and HA-KCNQ3/ KCNQ2 channels We have shown that fusion of the KCNQ2 C-terminal tail was sufficient to enrich CD4 at the axonal surface ( Fig. 1-2) regardless of KCNQ2 binding to ankyrin-G ( Figure S2). Our finding is consistent with previous reports that polarized axonal localization of heteromeric channels does not require ankyrin-G binding of KCNQ2 and KCNQ3 [18] and occurs prior to their ankyrin-G-dependent enrichment at the AIS [6,18,20]. These results together suggest that CD4-Q2C chimeric proteins serve as useful trafficking reporters to dissect the mechanisms underlying polarized axonal distribution of intact KCNQ channels [18]. Indeed, we found that surface and intracellular expression of both CD4-Q2C chimeric proteins and HA-KCNQ3/KCNQ2 channels at the axons were abolished by mutations at the IQ motif of KCNQ2, which disrupt CaM binding (Fig. 2-3, 8-9).
We did however observe several distinct differences in the trafficking of CD4-Q2C proteins and HA-KCNQ3/KCNQ2 channels from the ER to the axons (Fig. 7, 11, Figure S5). In comparison to wild-type CD4-Q2C proteins, the I340E mutant CD4-Q2C proteins trafficked from the soma to the dendrites to a lesser extent during the 8 hr time course following BFA removal (Fig. 7), whereas similar amounts of wild-type and A343D mutant heteromeric channels appeared at the dendrites (Fig. 11, Figure  S5). In contrast to the I340E mutant CD4-Q2C proteins, which were mostly absent from the AIS during the 8 hr time course (Fig. 7), a modest level of A343D mutant channels was detected at the AIS (Fig. 11, Figure S5). These trafficking differences could be due to channel retention in the smooth ER cisterns found at the proximal AIS [45,46,83], and/or due to the presence of KCNQ3, the dominant subunit that mediates KCNQ2/KCNQ3 channel localization to the AIS by binding to ankyrin-G [6,18,20].
In addition, the R353G mutation abolished polarized targeting of CD4-Q2C to the axonal surface by increasing dendritic surface expression (Fig. 4), whereas the same mutation had no effect on axonal enrichment of surface HA-KCNQ3/KCNQ2 channels (Fig. 8). Furthermore, the R353G mutation reduced trafficking of CD4-Q2C from the ER to the axons by half compared to the wild type (Fig. 7), whereas this mutation had no effect on the trafficking of heteromeric channels (Fig. 11, Figure S5). These results together indicate that the trafficking of CD4-Q2C protein and intact HA-KCNQ3/KCNQ2 channels to the axonal surface may not be equivalent. Despite the wide application of CD4 or CD8 chimeric proteins to understand polarized trafficking of endogenous protein complexes to axons or dendrites [18,30,31,77,84], the results from these chimeric proteins should be interpreted with caution and repeated using intact proteins.
Physiological significance of CaM-dependent enrichment of KCNQ channels at the axonal surface Axonal rather than somatic KCNQ channels have been shown to suppress action potential firing in hippocampal CA1 neurons [14,17]. We have shown that expression of wild-type KCNQ2 but not A343D mutant KCNQ2 allows robust axonal surface expression of HA-KCNQ3 (Fig. 8). Accordingly, our results from whole-cell patch clamp recordings have demonstrated that expression of wild-type KCNQ2 but not A343D mutant KCNQ2 significantly dampened action potential firing rate in cultured hippocampal neurons (Fig. 12). These results reflect the differences in the axonal enrichment of wild-type and A343D mutant channels, further supporting a critical role for CaM interaction with the IQ motif of KCNQ2 in regulating axonal surface expression of intact KCNQ channels. Consistent with our finding, dissociation of CaM from KCNQ2 or decreasing CaM levels has been shown to decrease hippocampal M-current and increase neuronal excitability [23,85]. Recently, phosphorylation of KCNQ2 by protein kinase C (PKC) has been shown to dissociate CaM from KCNQ2 channels, leading to a reduced affinity to PIP 2 and suppression of M-current [61]. Since A-kinase-anchoring protein (AKAP) 79 [54,86] regulates CaM binding to KCNQ2 [58] and serves as an adaptor protein for CaM and PKC [87,88], the mechanism responsible for CaM-mediated trafficking of KCNQ channels to the axonal surface may well converge with the ability of CaM to modulate PKC-dependent inhibition of Mcurrent via AKAP79/150 [61,89] and PIP 2 [61]. Considering the modest effect of CaM1234 on the axonal surface expression of HA-KCNQ3/KCNQ2 channels (Fig. 13), CaM bound to KCNQ2 may act as a Ca 2+ sensor for modulating not only the M-current, but also the channel density at the axonal membrane.
Recent studies have identified that de novo mutations in helices A and B of KCNQ2 are associated with neonatal epileptic encephalopathy [90][91][92] including drug-resistant Ohtahara syndrome [93]. Since inhibition of CaM binding to the IQ motif of KCNQ2 completely blocks the targeting of heteromeric HA-KCNQ3/KCNQ2 channels from the ER to the axonal surface ( Fig. 8-11), the KCNQ2 subunits with de novo [90][91][92] and BFNC [15] mutations in the IQ motif would likely exert a dominant negative effect by forming channels with KCNQ3 and retaining them in the ER. Complete lack of functional KCNQ channels at the axonal surface would therefore lead to burst and spontaneous firing of action potentials [10,14] and enhanced seizure susceptibility [8,94]. Thus, a deeper understanding of how CaM controls anterograde trafficking of KCNQ channels from the ER is expected to foster development of novel therapeutics that can enhance their axonal surface expression. Given that KCNQ channels are also implicated in hippocampal development [8], chronic inflammatory and neuropathic pain [95], anxiety [96], mania [97], and addiction [98], this novel therapeutic approach to increase KCNQ surface expression in combination with KCNQ agonist ezogabine/retigabine [16,99] may result in an efficacious therapy for a variety of neurologic disorders. and ''Axon/Dendrite'' ratios were determined as previously described [18,30] by obtaining background-subtracted mean surface CD4 fluorescence intensity of the axon between 0-30 mm (AIS) and between 50-80 mm (axon) from the beginning of the axon and the major primary dendrites. The ''Axon/Dendrite'' ratio shows that both WT (n = 13) and AIS mutant CD4-Q2C proteins (n = 13) were preferentially targeted to the axonal surface compared to non-polarized CD4 proteins (n = 8). The ''AIS/ Axon'' ratio reveals that enrichment of CD4-Q2C at the AIS surface is blocked by the E810A/D812A mutation. Ave 6 SEM (*p,0.05, **p,0.01, ***p,0.001). (TIF) Figure S3 Identification of axons and dendrites in neurons transfected with CD4-Q2C. Permeabilized immunostaining was performed in hippocampal neurons for the AIS using anti-phospho IkBa Ser32 (14D4) antibody to identify the axon after surface immunostaining for CD4-Q2C wild-type (WT) or mutant proteins (L339R, I340E, and A343D) was completed in