Calmodulin-dependent KCNE4 dimerization controls membrane targeting

The voltage-dependent potassium channel Kv1.3 participates in the immune response. Kv1.3 is essential in different cellular functions, such as proliferation, activation and apoptosis. Because aberrant expression of Kv1.3 is linked to autoimmune diseases, fine-tuning its function is crucial for leukocyte physiology. Regulatory KCNE subunits are expressed in the immune system, and KCNE4 specifically tightly regulates Kv1.3. KCNE4 modulates Kv1.3 currents slowing activation, accelerating inactivation and retaining the channel at the endoplasmic reticulum (ER), thereby altering its membrane localization. In addition, KCNE4 genomic variants are associated with immune pathologies. Therefore, an in-depth knowledge of KCNE4 function is extremely relevant for understanding immune system physiology. We demonstrate that KCNE4 dimerizes, which is unique among KCNE regulatory peptide family members. Furthermore, the juxtamembrane tetraleucine carboxyl-terminal domain of KCNE4 is a structural platform in which Kv1.3, Ca2+/calmodulin (CaM) and dimerizing KCNE4 compete for multiple interaction partners. CaM-dependent KCNE4 dimerization controls KCNE4 membrane targeting and modulates its interaction with Kv1.3. KCNE4, which is highly retained at the ER, contains an important ER retention motif near the tetraleucine motif. Upon escaping the ER in a CaM-dependent pattern, KCNE4 follows a COP-II-dependent forward trafficking mechanism. Therefore, CaM, an essential signaling molecule that controls the dimerization and membrane targeting of KCNE4, modulates the KCNE4-dependent regulation of Kv1.3, which in turn fine-tunes leukocyte physiology.

membrane targeting of Kv channels 13,14 . Thus, while KCNE1 interaction mostly improves Kv channel membrane targeting 27 , KCNE4 transfers to the Kv1.3 complex ER retention motifs crucial for the control of the surface expression of the channel 20, 21 . Because of the physiological relevance of these effects on the immune response, we characterized the intracellular retention mechanisms of KCNE4 by using KCNE1, the most documented KCNE subunit, as a reference. The membrane targeting of KCNE1 was low and similar to that observed with KCNE4 ( Fig. 2A), and both KCNEs markedly colocalized with the ER (Fig. 2B). Our data further support and extend the evidence suggesting that KCNEs share ER retention motifs in their structure. A sequence alignment of their primary structures clearly shows basic clusters canonically associated with ER retention located in the juxtamembrane region of the KCNE C-terminus (Fig. 2C). In this context, KCNE3 and KCNE4 motifs contained larger motifs, which could suggest a major capacity of transfer intracellular retention to the associated Kv complex. Interestingly, KCNE4 is unique within the KCNE family because it presents a tetraleucine motif (L69-72) known to participate in the CaM-dependent modulation of channels, such as Kv7.1 and Kv1. 3 15,20 . KCNE4 dimerization. Because dileucine motifs are critical for multiple protein-protein interactions, we wondered whether the tetraleucine signature mediates homo-oligomerizations of KCNE4. If it does, then this subunit is unique within the KCNE family, conferring distinctive structural features to this peptide that may fine-tune its Kv1.3-specific physiological functions. The molecular modeling of KCNE4, based on the KCNE1 structure (PBD ID: 2K21) 28 , suggested that the tetraleucine motif, containing four consecutive hydrophobic residues, forms an accessible loop for different protein-protein interactions, e.g. Kv1. 3 and CaM (Fig. 3A). This model is highly reliable because, although it is based on the KCNE1 structure 28 , it is quite similar to that obtained for KCNE3 by cryo-EM 29 . Protein analysis in the presence of the DMP cross-linking reagent revealed that KCNE4 and Kv1.3 form oligomeric structures, which indicates several protein-protein interactions (Fig. 3B). Thus, the tetrameric Kv1.3-YFP complex forms large molecules. Similarly, KCNE4 exhibited large forms, which indicate multimeric complexes detectable only with DMP. The formation of these structures was confirmed by coimmunoprecipitation showing that KCNE4-YFP and KCNE4-HA interact (Fig. 3C) and FRET studies of KCNE4-YFP and KCNE-4CFP ( Fig. 3D and E). This result was specific for KCNE4 because KCNE1 showed no positive FRET values (Fig. 3E).
To characterize the KCNE4 oligomers further, we performed a TIRF-derived single bleaching step assay 30 . Kv1.3-loopBAD-GFP, KCNE4-loopBAD-GFP and KCNE1-loopBAD-GFP were expressed in HEK 293 cells, and GFP fluorescent immobile spots were monitored (Fig. 4). Kv1.3 was used as the control (Fig. 4A-C 4B). We calculated the expected distribution of the tetrameric channels with different p values (probability of fluorescing GFP), and the best fit was obtained with a GFP folding efficiency of 67% (p = 0.67) (Fig. 4C), indicating a tetrameric architecture as previously described [30][31][32] . However, KCNE4 mostly exhibited two unique possible structures ( Fig. 4D-F). That is, monomeric and dimeric forms of KCNE4 were detected ( Fig. 4D and E). The best fit for the expected distribution and the experimental data revealed that KCNE4 is present in monomeric and dimeric forms (Fig. 4F). In contrast, using the same approach, we found that KCNE1 was present only in monomeric form (Fig. 4G, H), as previously demonstrated 33,34 .
The juxtamembrane tetraleucine motif facilitates CaM-dependent protein-protein interactions and membrane targeting of KCNE4. The tetraleucine motif of KCNE4 is an interacting platform that defines the CaM modulation of the KCNE4-dependent regulation of Kv1.3 and Kv7.1 15,20 . In this context, we wanted to know whether this unique motif among KCNE peptides facilitates KCNE4 dimerization.
To examine this possibility, we generated several KCNE4 mutants that disrupted the tetraleucine motif either alone (KCNE4(L69-72A)) or in combination with ERRM signaling (KCNE4(RM&L)). KCNE2, which neither contains tetraleucine motifs nor associates with Kv1.3, has been widely characterized 20 . Therefore, a KCNE2 mutant (KCNE2(L83-86)) with an embedded tetraleucine motif was also used (Fig. 5A). In contrast to WT KCNE4, disruption of the tetraleucine motif (KCNE4(L69-72A)) impaired KCNE4 dimerization, as observed by coimmunoprecipitation (Fig. 5B), non-denaturing polyacrylamide gel electrophoresis (Fig. 5C) and further supported by FRET assays (Fig. 5D). In addition, coimmunoprecipitation experiments with the KCNE2(L83-86) mutant showed that the addition of the leucine cluster to KCNE2 triggered the dimerization of KCNE2 with KCNE4 (Fig. 5E) www.nature.com/scientificreports/ Our data indicated that the tetraleucine motif of KCNE4 is a hub for protein-protein interactions governing the dimerization of this peptide. In addition, this cluster is involved in Kv1.3 and CaM associations 20 . Both Kv and CaM interactions with KCNE4 are important for the physiological function of Kv channels 15 . Therefore, we wanted to analyze whether KCNE4 dimerization has a competitive effect fine-tuning the association of KCNE4 -     www.nature.com/scientificreports/ between KCNE4-CFP/KCNE4-YFP, further supported that the dimerization of KCNE4 was also displaced by the presence of either Kv1.3 or CaM (Fig. 6D-G). Keeping all these results and the previous evidence in mind, we postulated a model (Fig. 6H). KCNE4 dimerizes via its juxtamembrane C-terminal tetraleucine motif. In addition, this region is involved in the Kv1.3 association and CaM interaction. The dimerization of KCNE4 would balance oligomeric interactions fine-tuning the KCNE4 physiological effects. Therefore, the association with Kv1.3 and CaM would be in competition with the KCNE4 dyad formation.
Recent evidence indicates that increasing amounts of KCNE4 steadily decrease Kv1.3 abundance at the cell surface 30 . Furthermore, CaM facilitates the membrane expression of Kv7 channels 22 . CaM also competes with Kv1.3 for KCNE4 tetraleucine motif association, facilitating Kv1.3 escape from KCNE4-dependent ER retention. We wondered whether CaM, impairing KCNE4 dimerization, affects the cellular distribution of KCNE4. The ER colocalization of KCNE4 in the absence (Fig. 7A-C) or presence of variable (low, Fig. 7D-G; high, Fig. 7H-K) CaM expression was analyzed. All images were captured using the same parameters of laser intensity and photomultiplication. Therefore, a pixel-by-pixel analysis of both CaM and KCNE4 intensities was performed, and the data were plotted against the ER localization of KCNE4. The data fitted with an exponential decay showed a strong correlation. Therefore, we found that increasing amounts of CaM triggered a notable decrease in KCNE4 ER localization concomitant with an increase in the cell surface staining of KCNE4 (Fig. 7H-L).
Our data indicated that CaM facilitated the targeting of KCNE4 to the plasma membrane. Thus, we wanted to identify the mechanism that promoted KCNE4 forward trafficking to the cell surface. Although KCNE peptides exhibit notable intracellular retention, KCNE1 reaches the cell surface via COPII-dependent machinery 27 . KCNE1 and KCNE4 behave similarly, but only the latter contains the tetraleucine signature that facilitates CaM-dependent membrane surface expression. Therefore, we analyzed whether the COPII machinery mediated KCNE4 membrane targeting (Fig. 8). KCNE4 exhibited intracellular retention and minor membrane staining (Fig. 8A-C). The coexpression of Sar1 (H79G), which disrupts ER-to-Golgi trafficking 27  www.nature.com/scientificreports/ of KCNE4 localized at the membrane ( Fig. 8A-C, Y). KCNE4(L69-72A) showed improved membrane targeting, similar to that obtained with KCNE4(ERRM), which has the ER retention motif disrupted, and KCNE4(RM&L), with both the ERRM and the tetraleucine signals altered (Fig. 8D-X). Interestingly, Sar1(H79G) clearly prevented membrane targeting under all conditions (Fig. 8Y). Therefore, our data suggest that, in addition to the tetraleucine motif, ERRM cooperates in the ER retention of KCNE4. In addition, we show that KCNE4, escaping the ER, is routed via conventional COPII-dependent anterograde trafficking to the cell plasma membrane.

Discussion
The voltage-gated potassium channel Kv1.3 is crucial in the physiology of the immune system. The channel controls the membrane potential, driving Ca ++ entry during the immune response. The abundance of Kv1.3 at the cell surface is in accordance with the magnitude of the cellular reaction 35 . KCNE4 is a regulatory subunit that controls the expression of Kv1.3 in the plasma membrane, fine-tuning this response 30 . Because KCNE4 is  The tetraleucine motif at the juxtamembrane domain of the KCNE4 C-terminus is a structural platform where essential proteins for Kv1.3-dependent physiology interface. Thus far, this KCNE4 feature is unique within the KCNE family. Dileucine motifs promote endocytosis, protein-protein interactions or basolateral trafficking [36][37][38][39] . In this context, leucine-enriched domains mediate the oligomerization of Toll-like receptors 40 . However, leucine residues are mainly located at every third amino acid position in these domains 41 . The KCNE4 tetraleucine signature forms a coil-like domain that can be transferred to other KCNE peptides. Moreover, our data demonstrated that this mechanism acts as a trafficking modulator. Mutation of this cluster, as well as the ERRM, facilitates plasma membrane targeting of KCNE4. Dimerization provides redundant signals to the KCNE4 dyad, retaining a pool of peptides at the ER. Thus, not only the interaction with Kv1.3, but also with CaM, is controlled 20 . Upon association with CaM or Kv1.3, KCNE4 units may travel either with CaM to the membrane or, by masking the forward traffic motifs of Kv1.3, retain the channel within in the cell, thereby tuning the channel expression at the cell surface. Both mechanisms reveal as important players controlling Kv1.3-asociated cellular responses 42,43 .
Evidence links dileucine clusters with protein trafficking. For instance, the ClC-2 chloride channel, which might compensate for the malfunction of CFTR in absorptive epithelia during cystic fibrosis, contains a dileucine motif. Mutations of this signature alter the basolateral targeting of the channel 44 . Similarly, the dileucine motif of human organic solute transporter beta plays a critical role in its association with the alpha subunit and cell surface polarization 37 . However, similar to that of KCNE4, other evidence raises an alternative possibility. The acid-sensing ion channel ASIC2a has a double dileucine motif (LLDLL) in the C-terminal juxtamembrane region. Altering the LLDLL sequence, either full or in part, increases the surface levels of the channel 45 . Moreover, mutation of the carboxyl terminal dileucine sequence of the LHβ subunit increases trafficking and exocytosis 46 . In this context, our data indicated that disrupting or masking the tetraleucine motif hinders dimerization, thereby facilitating membrane targeting of KCNE4. Therefore, in a KCNE4 monomer, although it contains an ERRM, the strength of the retention signal may be counterbalanced by associating with CaM. This mechanism is in agreement with the Ca 2+ /CaM-dependent regulation of KCNE4 of Kv channels, such as Kv7.1 or Kv1.3. CaM, upon calcium stimuli, can interact with KCNE4 at the tetraleucine motif 15,20 . Therefore, CaM fine-tunes the dimerization and promotes KCNE4 anterograde trafficking through a COPII-dependent mechanism. Thus, Ca 2+ signaling controls the final fate of KCNE4, which in turn modulates Kv1.3 function, which is profoundly related to the Ca 2+ trigger in leukocytes. The precise balance between several interactions in a unique cluster, masking the interaction sites to prevent competing associations, fine-tunes the final function of a multitask channel such as Kv1.3 16 . A similar cloaking mechanism is evident for E-cadherin. A dileucine motif in the juxtamembrane cytoplasmic domain is required for E-cadherin endocytosis. When mutated, this protein cannot be internalized. P-120, an E-cadherin interactor, associates and masks this motif, inhibiting endocytosis 47 . Furthermore, our data indicate that KCNE4 contains routing information to the plasma membrane via COPII-dependent mechanisms. KCNEs, which contain strong ERRM signals at their C-terminus, share a notable intracellular retention capacity. Surprisingly, without canonical sequences, the COPII-dependent mechanism is a major characteristic of KCNE peptides because, similar to KCNE4, KCNE1 uses this route 27 . In this sense, N-glycosylation facilitates KCNE anterograde trafficking through the secretory pathway, but the specific mechanisms have not been dissected 48 . The balance between forward trafficking and retention signaling is highly skewed toward the latter in KCNEs. Regardless of whether the assembly of KCNE and Kv channels takes place at the cell surface or through the secretory pathway, CaM controls cell surface channels by modulating KCNE4 trafficking 15,27,49 .
We demonstrated that Kv1.3, KCNE4 and CaM are closely related because they interact via the tetraleucine cluster in KCNE4. A balanced competition warrants no tripartite complexes 20 . Our data support the dimerization of KCNE4 as a regulating mechanism for such interactions and CaM displaced KCNE4 dyad formation. These data, similar to those of Kv7.1/KCNE1 and Kv7.1/KCNE3, further support that single monomeric forms of KCNE4 occupy the cleavage between subunits within the Kv tetrameric structure 20, 28,29
For confocal imaging and coimmunoprecipitation experiments, cells were seeded (70-80% confluence) in either 6-well dishes containing poly-D-lysine-coated coverslips or 100-mm dishes. Metafectene PRO (Biontex) was used for transfection according to the supplier's instructions. The amount of transfected DNA was 4 μg for a 100-mm dish and 500 ng for each well of a 6-well dish. Next, 4-6 h after transfection, the mixture was removed from the dishes and replaced with fresh culture medium. All experiments were performed 24 h after transfection 20, 21 .
For patch-clamp experiments, trypsinized cells from a confluent culture in a 100-mm dish were electroporated with 1 μg of DNA using a Bio-Rad Gene Pulser Xcell system with a 0.2-cm gap cuvette and a single 110-V 25-ms pulse 30 .
For total internal reflection fluorescence (TIRF) microscopy experiments, trypsinized cells from a confluent culture in a 100-mm dish were electroporated with 25-100 ng of the desired DNA plus 100 ng of BirA DNA (encoding a biotin ligase to biotinylate the loopBAD-tagged proteins) using a Bio-Rad Gene Pulser Xcell system, as previously described. Transfected cells were plated on 35-mm glass-bottom dishes (MatTek) previously coated with collagen and EZ-Link NHS-PEG12-Biotin (Pierce, Thermo Scientific). The next day, TIRF experiments were performed after incubation of the cells with neutravidin (50 nM) to immobilize channels onto glass as previously described 20,35 .
Protein extraction, coimmunoprecipitation and western blotting. HEK-293 cells were washed twice in cold PBS and lysed on ice with a lysis solution (1% Triton X-100, 10% glycerol, 50 mM HEPES, 150 mM NaCl at pH 7.2) supplemented with protease inhibitors (1 µg/ml aprotinin, 1 µg/ml leupeptin, 1 µg/ml pepstatin and 1 mM phenylmethylsulfonyl fluoride). Lysates were gently mixed for 10 min and spun (10 min at 12,000 × g). The supernatant was transferred to a new tube, and the protein contents were determined by using a Bio-Rad protein assay 20,21 .
For coimmunoprecipitation, 1 mg of protein added to lysis buffer at a final concentration of 500 µl and used for immunoprecipitation (150 mM NaCl, 50 mM HEPES, 1% Triton X-100 at pH 7.4) supplemented with protease inhibitors. Samples were precleared with 50 µl of protein A-Sepharose beads (GE Healthcare) for 1 h at 4 °C with gentle mixing. Next, each sample was incubated in a small chromatography column (Bio-Rad Micro Bio-Spin™ chromatography columns), which contained 2.5 µg of anti-GFP antibody previously cross-linked to protein A-Sepharose beads, for 2 h at room temperature (RT) with gentle mixing. The columns were centrifuged for 30 s at 1,000 × g. Supernatants were stored at -20 °C. The columns were washed four times with 500 µl of lysis buffer and centrifuged for 30 s at 1,000 × g. Finally, the columns were incubated with 100 µl of 0.2 M glycine (pH 2.5) and spun for 30 s at 1,000 × g for elution [19][20][21] .
Irreversible cross-linking of the antibody to the Sepharose beads was performed after incubation of the antibody with protein A-Sepharose beads for 1 h at RT. The beads were then incubated with 500 µl of 5.2 mg/ml dimethyl pimelimidate (Pierce) for 30 min at RT with gentle mixing. The columns were then washed four times with 500 µl of 1 × TBS, four times with 500 µl of 0.2 M glycine (pH 2.5) and three more times with 1 × TBS (0.1% Triton X-100, 10% glycerol, 150 mM NaCl, 50 mM HEPES at pH 7.4). Finally, the columns were incubated with protein lysates to perform immunoprecipitation as described above.
For the coimmunoprecipitation experiments in the presence of CaM, the samples were precleared for 1 h at 4 °C by mixing samples with 50 µl of protein A-Sepharose beads (GE Healthcare) previously washed with 1 × TBS and centrifuged for 30 s at 5,000 × g. Next, supernatants were incubated for 2 h at RT with 8 ng of anti-HA antibody (Proteintech) previously cross-linked to Protein A-Sepharose beads. The samples were centrifuged (30 s, 5.000 × g), and the pellet was washed 3 times with 1 × TBS buffer. Finally, proteins were eluted upon the www.nature.com/scientificreports/ cis-Golgi and trans-Golgi networks respectively. Next, the cells were incubated with Cy5-conjugated secondary antibodies for 2 h at RT and further mounted with Mowiol. The fluorescence resonance energy transfer (FRET) via the acceptor photobleaching technique was measured in discrete ROIs (regions of interest). Fluorescent proteins from fixed cells were excited with the 458 nm or 514 nm beam lines using low excitation intensities. Next, 475 to 495 nm bandpass and > 530 nm longpass emission filters were applied. The YFP protein was bleached using maximum laser power. We obtained acceptor intensity bleaching of approximately 80%. Posteriorly, images of the donors and acceptors were taken. The FRET efficiency was calculated using the equation [(F CFPafter -F CFPbefore )/F CFPbefore ]*100, where F CFPafter was the fluorescence of the donor after bleaching and F CFPbefore was the fluorescence before bleaching. The loss of fluorescence as a result of scanning was corrected by measuring the CFP intensity in the unbleached part of the cell 20 .
TIRF and bleaching steps of single fluorescent protein complexes. The single bleaching quantitative approach was first described by Isacoff and collaborators 31 and is based on TIRF microscopy to visualize single GFP-tagged proteins on the cell surface, as previously described 20, 35 . To immobilize proteins and analyze the number of bleaching steps, cells were transfected with loopBAD-tagged proteins in the presence of BirA, which encodes DNA for tag biotinylation. The cells seeded in a biotin-collagen-coated glass-bottom dish were incubated with neutravidin (50 nM) at 37 °C for 30 min before imaging. Neutravidin binds the cell surface biotinylated Kv1.3-, KCNE1-and KCNE4-loopBAD-GFP and, simultaneously, the biotinylated glass surface, fixing the cell surface proteins to monitor the fluorescently immobile GFP spots. The amount of DNA transfected was adjusted to achieve a low membrane density for imaging and counting. Multiple spots were imaged, but the density was low enough to minimize the probability of two channels lying within the same diffraction-limited spot. Transfected HEK-293 cells were imaged within 24 h after electroporation in HEK physiological saline buffer consisting of (in mM) 146 NaCl, 4.7 KCl, 2.5 CaCl 2 , 1 MgCl 2 , 10 glucose, and 10 HEPES at pH 7.4. Videos were processed and analyzed using Volocity software. The intensity of the 6 × 6 pixel ROIs was added and calculated for all durations of the videos. The intensity of the ROIs was plotted against time. The GFP bleaching steps were counted and statistically analyzed. GFP bleaching experiments were performed with a Nikon Eclipse Ti Perfect-Focus system equipped with a TIRF/wide-field fluorescence microscope and AOTF-controlled 405, 488, 561 nm diode lasers (100 mW each), and an Intensilight wide-field light source. A 100 × PlanApo TIRF, 1.49 NA, objective lens was used for image acquisition. Emissions were collected through a Sutter Lambda 10-3 filter wheel containing the appropriate bandpass filters. The microscope was equipped with an Andor iXonEMCCD DU-897 camera, 512 × 512. For TIRF image acquisition, an incident angle of 63.3° was used 30,35,52 . Statistics. The results are expressed as the means ± SE. Student's t-test, one-way ANOVA and Tukey's post hoc test and two-way ANOVA were used for statistical analysis (GraphPad PRISM v5.01). P < 0.05 was considered statistically significant.