Regulation of human cardiac potassium channels by full-length KCNE3 and KCNE4

Voltage-gated potassium (Kv) channels comprise pore-forming α subunits and a multiplicity of regulatory proteins, including the cardiac-expressed and cardiac arrhythmia-linked transmembrane KCNE subunits. After recently uncovering novel, N-terminally extended (L) KCNE3 and KCNE4 isoforms and detecting their transcripts in human atrium, reported here are their functional effects on human cardiac Kv channel α subunits expressed in Xenopus laevis oocytes. As previously reported for short isoforms KCNE3S and KCNE4S, KCNE3L inhibited hERG; KCNE4L inhibited Kv1.1; neither form regulated the HCN1 pacemaker channel. Unlike KCNE4S, KCNE4L was a potent inhibitor of Kv4.2 and Kv4.3; co-expression of cytosolic β subunit KChIP2, which regulates Kv4 channels in cardiac myocytes, partially relieved Kv4.3 but not Kv4.2 inhibition. Inhibition of Kv4.2 and Kv4.3 by KCNE3L was weaker, and its inhibition of Kv4.2 abolished by KChIP2. KCNE3L and KCNE4L also exhibited subunit-specific effects on Kv4 channel complex inactivation kinetics, voltage dependence and recovery. Further supporting the potential physiological significance of the robust functional effects of KCNE4L on Kv4 channels, KCNE4L protein was detected in human atrium, where it co-localized with Kv4.3. The findings establish functional effects of novel human cardiac-expressed KCNE isoforms and further contribute to our understanding of the potential mechanisms influencing cardiomyocyte repolarization.

the extent that the pumping function of the heart is disrupted, via a chaotic state termed ventricular fibrillation, which is often lethal 31,32 . Another potentially lethal cardiac arrhythmia syndrome, Brugada syndrome (BrS), is most commonly genetically linked to loss-of-function mutations in the SCN5A gene that encodes the primary cardiac sodium channel α subunits, Nav1.5 33 . However, more recently, sequence variants in the genes encoding subunits or putative subunits of cardiac channel complexes generating the transient outward Kv current (I to ) have been associated with BrS. These include the genes encoding KCNE3 and KCNE5, which can regulate the Kv4.3 α subunit that generates I to in human heart, and the KCND3 gene that encodes Kv4.3 itself [34][35][36][37][38] .
Given the necessity of KCNE subunits in some native channel complexes, the potential severity of disease states associated with some KCNE gene variants, and the possibility of remodeling of KCNE genes during disease states 39 , it is important to understand the molecular mechanisms underlying KCNE regulation of native Kv currents. Recently, I discovered exon 1 coding regions that generate longer (L) isoforms of human KCNE3 and KCNE4, detected both proteins in various epithelial tissues, and conducted an initial examination of their functional effects 40 . Here, I demonstrate the expression of KCNE4L protein in human heart, and describe the functional effects of KCNE3L and KCNE4L on the major human cardiac Kv channel α subunits.
For Chinese hamster ovary (CHO) cell biochemistry, CHO cells were transfected with cDNAs encoding human KCNE4S in pCINeo, or human KCNE4L in pCDNA3.1+ 40 . Two days post-transfection, cells were lysed in MiRP buffer, rotated end-over-end in Eppendorf tubes for 2 hours at 4 °C, then centrifuged (4000 × g; 10 min; 4 °C), and the resulting supernatants SDS-PAGE separated, western blotted and analyzed as described for heart samples above.
Immunofluorescence detection. Human normal atrial tissue slides (5 μ m thick, paraformaldehyde fixed and paraffin wax embedded) (2 per antibody combination) were purchased from Abcam (Cambridge, UK). Ventricular tissue samples from adult male mice (two per genotype) were similarly fixed, embedded and sectioned. Mice were housed and used according to the US National Institutes of Health Guide for the Care and Use of Laboratory Animals. Animal procedures were approved by the Animal Care and Use Committee at University of California, Irvine (IACUC Protocol #2011-2999). Human and mouse tissue sections were next deparaffinized and permeabilized, then immunofluorescence labeling was performed manually, with 3 × 5 minute washes in PBS between antibody incubation steps. Slides were incubated overnight at 4 °C with rabbit α -KCNE4L (in-house), goat α -KCNQ1 (Santa Cruz Biotechnology, Santa Cruz, CA), or mouse α -Kv4.3 (Neuromabs, Davis, CA) at 1/100 dilution in PBS with 1% BSA, followed by a 2 hour room-temperature incubation in Alexa-fluor-conjugated secondary antibodies (ThermoFisher Scientific, Waltham, MA) at 1/500 dilution in PBS. Following mounting using DAPI-containing slow-fade mounting medium (Life Technologies, Grand island, NY, USA), slides were viewed with an Olympus BX51 microscope and pictures were acquired using CellSens software (Olympus).
Xenopus laevis oocyte expression. cRNA transcripts encoding hKCNE3L and hKCNE4L were generated by in vitro transcription (T7 polymerase mMessage mMachine kit, Thermo Fisher Scientific) from synthetic genes sub-cloned into pCDNA3.1+ with Xenopus laevis β -globin 5′ and 3′ UTRs flanking the coding region to enhance translation and cRNA stability, after vector linearization. hKCNE3S and hKCNE4S cRNAs were similarly generated, after using site-directed mutagenesis (Agilent, Santa Clara, CA), to disrupt the exon 1 start site on the cDNAs of their long counterparts to facilitate direct comparisons of short versus long isoforms in otherwise identical constructs. Human Kv4.2, Kv4.3 (full-length) and KChIP2 were similarly transcribed from cDNA templates also incorporating Xenopus laevis β -globin 5′ and 3′ UTRs (a kind gift of Dr. Steve A. N. Goldstein), as were human HCN1, hERG and Kv1.1. cRNA was quantified by spectrophotometry. Defolliculated stage V and VI Xenopus laevis oocytes (Ecocyte Bioscience, Austin, TX) were injected with one, two or three of the subunit cRNAs ( per oocyte). Oocytes were incubated at 16 °C in SBB solution (Ecocyte) containing penicillin and streptomycin, with daily washing, for 2-3 days before two-electrode voltage-clamp (TEVC) recording.
TEVC. TEVC recording was performed at room temperature with an OC-725C amplifier (Warner Instruments, Hamden, CT) and pClamp8 software (Molecular Devices, Sunnyvale, CA). Oocytes were placed in a small-volume oocyte bath (Warner) and viewed with a dissection microscope. Bath solution was (in mM): 96 NaCl, 4 KCl, 1 MgCl 2 , 1 CaCl 2, 10 HEPES (pH 7.6); bath chemicals were from Sigma. TEVC pipettes were of 1-3 MΩ resistance when filled with 3 M KCl. Currents were recorded in response to a Voltage protocol consisting of pulses between − 120 mV or − 80 mV and 40 or 60 mV at 20 mV intervals, from a holding potential of − 80 mV, to yield current-Voltage relationships and for fitting of inactivation kinetics. For hERG and Kv1.1, plotting peak current during a tail pulse to − 30 mV, versus prepulse Voltage, permitted calculation of conductance-Voltage relationships. For quantification of Kv4.x steady-state inactivation, oocytes were held at − 100 mV and prepulsed to Voltages between − 120 and 0 mV followed by a tail pulse to + 40 mV. For quantification of inactivation recovery rates, Kv4.x-based channels were double-pulsed to + 40 mV with variable recovery times (10-5000 ms) at − 120 mV in between, and the magnitude of the second peak compared to that of the initial peak for each pair. TEVC data analysis was performed with Clampfit (Molecular Devices) and Origin 6.1 (OriginLab Corp., Northampton, MA) software. Values are stated as mean ± SEM. Normalized tail currents (Kv4.x steady-state inactivation; hERG and Kv1.1 G/V relationships) were plotted versus prepulse voltage and fitted with a single Boltzmann function according to: where g is the normalized tail conductance, A 1 is the initial value at − ∞ , A 2 is the final value at + ∞ , V 1/2 is the half-maximal voltage of activation and V s the slope factor. Kv4.x current inactivation curves were fitted with a standard (zero-shift) double exponential decay function with Chebyshev 4-point smoothing filter, whereas decay of Kv4.x-KChIP2 currents was amply described by a single exponential decay function. Inactivation recovery kinetics were fitted from mean normalized fractional recovery currents to a two-phase exponential association equation: and in cases where iterative fitting yielded identical τ values, a single exponential fit was reported. Because mean inactivation recovery curves were fitted, these data are reported as a value with no standard error, but rather a chi-squared test for goodness of fit. In all other cases, values are reported with standard error of the mean. Where informative, currents were compared with one another using one-way ANOVA to assess statistical significance (P < 0.05). If multiple comparisons were performed, a post-hoc Tukey's HSD test was performed following ANOVA.

Results and Discussion
Detection of KCNE4L protein in human atrium. The novel exon 1-encoded N-terminal additions to human KCNE3 and KCNE4 extend their predicted coding regions by 44 and 51 residues, respectively, yielding predicted full-length proteins of 147 (KCNE3L) and 221 (KCNE4L) (Fig. 1A,B). In support of their potential relevance to human cardiac function, transcripts coding for KCNE3L and KCNE4L were previously detected in human atrial tissue 40 .
The reported migration patterns of the two forms of mature KCNE4 protein on SDS-PAGE gels correspond to ~25 kDa for KCNE4S 41 and ~30 kDa for KCNE4L 40 . Here, an antibody raised to a peptide sequence present on only the long form of KCNE4 detected a ~30 kDa band in CHO cells expressing KCNE4L, but no bands in the 20-40 kDa size range in CHO cells expressing KCNE4S. Using aliquots of the same antibody stocks, a band at ~30 kDa (which corresponds to KCNE4L) was detected in homogenized human atrial tissue lysate (Fig. 1C). Thus, KCNE4L protein is expressed in human atrium. Similar studies were attempted for KCNE3, but did not yield unequivocal expression data for either KCNE3S or KCNE3L, and so are not included here.
Immunofluorescence staining of human atrial tissue with subunit-specific antibodies detected KCNE4L in a striated pattern in atrial myocytes, showing minimal overlap with KCNQ1 ( Fig. 1D) but co-localization with Kv4.3 (Fig. 1E). As a negative control, mouse ventricular myocytes (which do not express KCNE4L and exhibit little evidence of KCNQ1 activity) were similarly probed, showing negligible signal, as expected if the antibodies are specific (Fig. 1F).
KCNE3L inhibits hERG; KCNE4L inhibits Kv1.1; neither affects HCN1. I recently found that KCNE3L converts KCNQ1 to a constitutively active K + channel, as previously reported for KCNE3S 7 , although KCNE3L also diminished overall KCNQ1 current, in contrast to effects observed with KCNE3L 40 . KCNE3S inhibits KCNQ4 activity when expressed in Xenopus oocytes, whereas KCNE3L had no effect in similar experiments 7,40 . Previous studies also showed that in oocytes and in CHO cells, KCNE3S inhibits hERG 7,42 , which generates the major human cardiac repolarization current, I Kr 43 . Here, co-expression of hERG with KCNE3L in Xenopus oocytes recapitulated these findings, suggesting the extra N-terminal portion of KCNE3L does not substantively alter its interaction with hERG. As previously reported for KCNE4S 44 , KCNE4L did not alter hERG current ( Fig. 2A,B). Kv1.1 was recently reported to be expressed in human and mouse heart and to influence susceptibility to atrial fibrillation 45 . Here, KCNE4L inhibited hKv1.1 activity, as previously found for KCNE4S 46 , whereas KCNE3L had no effect (Fig. 2C,D). The current remaining after co-expression of hKv1.1 with KCNE4L had similar Voltage dependence to that of homomeric hKv1.1 (Fig. 2E). KCNE2 is known to modulate the Scientific RepoRts | 6:38412 | DOI: 10.1038/srep38412 hyperpolarization-activated, monovalent cation-nonselective "pacemaker" channel encoded by HCN1 47 ; here, neither KCNE3L nor KCNE4L exerted any functional effects when co-expressed with human HCN1, changing neither its instantaneous nor peak current magnitudes, nor its voltage dependence of activation ( Fig. 3A-D).

Potent inhibition of hKv4.3 by KCNE4L is diminished by KChIP2
. KCNE3L and KCNE4L were each inhibitors of hKv4.3 activity, with the latter producing the strongest inhibition (~60% versus > 90% at + 60 mV, P < 0.005; n = 13-14). This was in contrast to KCNE3S and KCNE4S, neither of which altered Kv4.3 peak current (n = 9-11) (Fig. 5A,B). As previously reported 51 , KChIP2 doubled the peak whole-cell current at + 40 mV of Kv4.3 (when 1 ng of each cRNA was co-injected per oocyte; Fig. 5D inset). Augmentation by KChIP2 was not apparent when a higher amount (5 ng per oocyte) of Kv4.3 cRNA was injected (P = 0.25, n = 12-14), suggesting that Kv4.3 membrane expression was already maximal after 5 ng cRNA injection, even without KChIP2 (Fig. 5C,D). Under these conditions, KChIP2 co-expression did not alter the extent of inhibition of hKv4.3 by KCNE3L (P = 0.89, n = 12) but was still able to partially alleviate inhibition by KCNE4L (from > 90% to ~80% at + 60 mV; P < 0.05, n = 11-12) (Fig. 5C,D).     KCNE3L and KCNE4L differentially influence hKv4.x inactivation gating. Fast inactivation is a signature of K + channels formed by Kv4.x α subunits and is crucial to their role in shaping early action potential morphology in cardiac myocytes in a range of species including Homo sapiens. Here, inactivation at + 40 mV of channels formed by Kv4.2 or Kv4.3 in the absence of KChIP2 was best fit with a double exponential function, while addition of KChIP2 produced inactivation that was well fit to a single exponential function. KCNE4L increased the τ of both the fast (by 50%) and slow (by 27%) components of hKv4.2 inactivation, and also slightly decreased the fractional amplitude of the fast component. KCNE4S exerted qualitatively similar effects but with quantitative differences that achieved statistical significance using ANOVA with post-hoc Tukey's HSD to correct for multiple subunit combinations (Fig. 6A). KCNE3L did not alter hKv4.2 inactivation rate, while KCNE3S slightly speeded the fast component of inactivation (Fig. 6A). In contrast, hKv4.2-KChIP2 channel inactivation was speeded by both KCNE3L (~10%) and KCNE4L (~20%) (Fig. 6B).
KCNE3L decreased the τ of both the fast (20% at + 40 mV) and slow (23% at + 40 mV) components of hKv4.3 inactivation; this was opposite to the effects of KCNE3S and KCNE4S, which each slowed both components of hKv4.3 inactivation (Fig. 6C). KCNE4L appeared to induce a bimodal spread of hKv4.3 inactivation rates that was not statistically significantly different from those of Kv4.3 alone but could hint at an additional regulatory component being imposed by KCNE4L (Fig. 6C). In direct contrast, KCNE3L did not alter Kv4.3-KChIP2 inactivation rate, yet KCNE4L decreased its τ of inactivation 30% (P < 1 × 10 −7 , n = 8-13), qualitatively similar to previous reports for KCNE4S 52 .
KCNE3L and KCNE4L each shifted the voltage dependence of Kv4.2 steady-state inactivation more positive, from a V 1/2 of − 76 ± 0.7 mV (Kv4.2 alone) to − 71.4 ± 1 mV (with KCNE3L) and − 67.9 ± 1 mV (KCNE4L). KChIP2 positively shifted Kv4.2 steady-state inactivation V 1/2 by > 20 mV, to − 55.1 ± 0.5 mV (similar to previous reports 53 ); this shift was slightly attenuated, by + 2 mV for KCNE3L and + 5 mV for KCNE4L. KChIP2 altered the slope from ~8 mV to ~5 mV but neither KCNE subunit impacted this property (Fig. 7A). KChIP2 also positive-shifted and steepened the slope of voltage dependence of Kv4.3 steady-state inactivation, while KCNE3L had no effects on either property for Kv4.3 or Kv4.3-KChIP2 channels (Fig. 7B). Increases in the fractional current remaining at − 50 mV to 0 mV associated with KCNE3L co-expression with Kv4.3 probably arose from increased relative contribution of endogenous current because of Kv4.3 inhibition by KCNE3L (Fig. 7B). For subunit combinations in which KCNE4L exerted strong inhibitory effects, steady-state inactivation could not be meaningfully quantified and these were omitted.  hKv4.2 recovery from inactivation fitted well with a double exponential fit regardless of the subunits regulating it. In the absence of KChIP2, recovery from inactivation of hKv4.2 was slow and unaffected by KCNE3L or KCNE4L (Fig. 7C,D). Differences in the amplitudes of fast and slow components of hKv4.2-KCNE4L recovery compared to those of hKv4.2 alone likely arise from difficulties in fitting the much smaller currents after KCNE4L inhibition and correspondingly increased relative contribution of endogenous current (Table 1). In contrast, KCNE3L and KCNE4L each slowed recovery from inactivation of hKv4.2-KChIP2 channels, which in the absence of KCNEs is much faster than that of hKv4.2 50 , both inducing increases in the τ of fast inactivation and a reduction in its relative amplitude. Consequently, at least double the fraction of channels remained inactivated after 100 ms of recovery at − 120 mV in hKv4.2-KChIP2 channels co-expressed with KCNE3L (P = 0.01) or KCNE4L (P = 1 × 10 −6 ), compared to those without (n = 6 per group) (Fig. 7C,D; Table 1). Interestingly, KCNE4L produced a minor overshoot in hKv4.2-KChIP2 channels (Fig. 7D). A complete comparison of effects of KCNE3L and KCNE4L on recovery from inactivation of hKv4.3 channels was not feasible because current inhibition for some subunit combinations precluded accurate fitting.
Conclusions, limitations and future studies. The current work expands our understanding of the potential modes of regulation of human cardiac potassium channels by demonstrating the effects of newly discovered longer forms of human KCNE3 and KCNE4 regulatory subunits on human channel properties. While some features of KCNE3L and KCNE4L function are shared with their shorter counterparts, other properties differ. Most markedly, KCNE4L is a potent inhibitor of Kv4.3 activity, even in the presence of the cytosolic regulatory subunit, KChIP2. This unexpected finding raises the possibility that KCNE4L could act as a repressor subunit for I to , the transient outward K + current thought to be primarily generated by Kv4.3-KChIP2 complexes in human cardiac myocytes. This might contribute to transmural gradients of I to across the myocardium, or perhaps allow dynamic regulation of I to density and other properties in response to specific signals or during different developmental states.
The recent discovery of extended forms of human KCNE3 and KCNE4 proteins opens up the potential for another level of complexity with regard to regulation of potassium channels by the KCNE subunits. Specifically, it may be possible for cells to respond to different signals by expressing either long or short forms of KCNE3 and KCNE4, with different functional outcomes. However, elucidation of whether splice variation occurs, versus solely the long forms being transcribed, and the mechanisms regulating this, will require future studies. Similarly, the question still remains of exactly which complexes can form in native tissues including the heart, and how the subunit composition might vary between cells or even within a single cell. Partly because KCNE proteins are small, transmembrane subunits, and partly because K + channels need not necessarily be expressed at high levels in excitable cells, cataloging of precise channel subunit composition at a regional level is highly challenging. The discovery of new, extended forms of KCNE3 and KCNE4 further complicates attempts at these assignments.
Sequence variations in the human KCNE3 gene have been discovered and tentatively functionally associated with propensity to cardiac arrhythmias including LQTS 54 , Brugada syndrome 34,36 and atrial fibrillation 55 . In addition, a single polymorphism in the human KCNE4 gene (E145D) is suggested to alter predisposition to atrial fibrillation in Chinese populations [56][57][58][59] . In future studies it will be of interest to determine whether the presence of the additional N-terminal residues in KCNE3L or KCNE4L, versus their shorter counterparts, alters the functional consequences of the disease-associated gene sequence variants.  Fig. 7.