MinK-dependent internalization of the I_Ks potassium channel

Abstract

Aims

KCNQ1–MinK potassium channel complexes (4α:2β stoichiometry) generate I_Ks, the slowly activating human cardiac ventricular repolarization current. The MinK ancillary subunit slows KCNQ1 activation, eliminates its inactivation, and increases its unitary conductance. However, KCNQ1 transcripts outnumber MinK transcripts five to one in human ventricles, suggesting KCNQ1 also forms other heteromeric or even homomeric channels there. Mechanisms governing which channel types prevail have not previously been reported, despite their significance: normal cardiac rhythm requires tight control of I_Ks density and kinetics, and inherited mutations in KCNQ1 and MinK can cause ventricular fibrillation and sudden death. Here, we describe a novel mechanism for this control.

Methods and results

Whole-cell patch-clamping, confocal immunofluorescence microscopy, antibody feeding, biotin feeding, fluorescent transferrin feeding, and protein biochemistry techniques were applied to COS-7 cells heterologously expressing KCNQ1 with wild-type or mutant MinK and dynamin 2 and to native I_Ks channels in guinea-pig myocytes. KCNQ1–MinK complexes, but not homomeric KCNQ1 channels, were found to undergo clathrin- and dynamin 2-dependent internalization (DDI). Three sites on the MinK intracellular C-terminus were, in concert, necessary and sufficient for DDI. Gating kinetics and sensitivity to XE991 indicated that DDI decreased cell-surface KCNQ1–MinK channels relative to homomeric KCNQ1, decreasing whole-cell current but increasing net activation rate; inhibiting DDI did the reverse.

Conclusion

The data redefine MinK as an endocytic chaperone for KCNQ1 and present a dynamic mechanism for controlling net surface Kv channel subunit composition—and thus current density and gating kinetics—that may also apply to other α–β type Kv channel complexes.

1. Introduction

Voltage-gated potassium (Kv) channels open in response to cellular depolarization to allow rapid, selective diffusion of K⁺ ions through an aqueous pore in the plasma membrane. This process repolarizes excitable cells, ending each action potential. Kv channel pore-forming α-subunits can form homotetrameric channels capable of voltage-sensing, gating, and selective K⁺ ion conduction, but many Kv channel complexes also contain one or more types of ancillary (β) subunit in vivo.

The MinK-related peptides (MiRPs, also called KCNE peptides) are single-transmembrane-domain ancillary (β) subunits encoded by KCNE genes, which co-assemble with Kv α-subunits to alter their function.¹ KCNQ1 (also referred to as Kv7.1) α-subunits form I_Ks channel complexes with MinK subunits, encoded by KCNE1, in human ventricular myocardium and inner ear.^2,3 MinK profoundly affects KCNQ1 function: it slows activation 5–10-fold, increases unitary conductance four-fold, eliminates inactivation, and provides an available reserve of closed states near the open state which can be recruited rapidly when required.^4–6 This latter property of I_Ks channels may provide a repolarization reserve to compensate for loss of the I_Kr repolarization current.⁷ Mutations in KCNQ1 and KCNE1 that cause loss of function of I_Ks channels are associated with long QT syndrome, which predisposes to life-threatening ventricular arrhythmias and sensorineural deafness due to reduced ventricular and inner ear I_Ks currents, respectively.^8–10 KCNQ1 is also regulated by MiRPs 1–4, encoded by KCNE2–5; these complexes are also postulated to occur in human heart and their dysfunction may contribute to ventricular and atrial arrhythmias.^1,11–13 Notably, KCNQ1 transcripts outnumber those of MinK five to one in human ventricles and atria,¹² whereas KCNQ1–MinK complexes adopt a 4α:2β stoichiometry,⁴ suggesting that homomeric KCNQ1 channels, or complexes involving KCNQ1 with one or more of the other MiRPs, contribute to ventricular and atrial repolarization. Given the critical importance of myocyte repolarization current density and functional attributes to normal cardiac rhythm, one would expect sophisticated control mechanisms in place to regulate the subunit composition of KCNQ1-based channels at the cell surface.

The subunit composition and lifetime at the cell surface of Kv channels dictate their contribution to the sum of repolarizing current and, therefore, cellular excitability. Mutant MinK was previously found to retain KCNQ1 in the endoplasmic reticulum (ER), demonstrating pathophysiological β-subunit perturbation of α-subunit forward trafficking.¹⁴ Recently, currents passed by KCNQ1 and KCNQ1–MinK complexes were found to be down-regulated by Nedd4.2-dependent ubiquitinylation.¹⁵ This down-regulation, and binding of Nedd4.2 to KCNQ1, were dependent on a PY motif in the C-terminal domain of KCNQ1 and proposed to involve channel internalization. This presented a mechanism for internalization of all channels containing KCNQ1, regardless of subunit composition.

Here, KCNQ1–MinK channels were found to also undergo clathrin- and dynamin-dependent internalization (DDI), mediated not by KCNQ1 but by MinK. This novel role for MinK, as an endocytic chaperone of its α-subunit partner, redefines this class of β-subunits and provides a novel mechanism for regulating net surface channel composition in addition to channel surface density.

2. Methods

Details of the following methods are described in Supplementary material online.

2.1. Molecular biology

Human MinK mutants were constructed using the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene). To facilitate immunofluorescence studies, MinK was tagged with a haemaglutinin (HA) epitope in the extracellular N-terminus, which we previously found not to alter MinK function¹⁶ and which was here also used for the electrophysiology studies, recapitulating the characteristic effects of wild-type minK on KCNQ1 (Figure 1). HA-tagged MinK with no other mutations is hereafter referred to as ‘MinK’ or ‘wild-type MinK’. COS-7 cells were transfected with cDNAs encoding wild-type or mutant MinK, and/or KCNQ1 alone or with wild-type or K44A dynamin 2 [and green fluorescent protein (GFP) for electrophysiology] using Superfect transfection reagent (Qiagen) 2 days before electrophysiology or confocal microscopy.

Figure 1.

I_Ks current density and MinK internalization are dynamin-dependent. (A) Example of current traces recorded in COS-7 cells expressing KCNQ1 alone or with MinK, ± wild-type (wt) or K44A dynamin as indicated. Voltage protocol and scale bars inset. Dashed line, zero current level. (B) Mean current density vs. voltage relationships for cells as in (A) expressing KCNQ1 alone (filled squares) or with wt dynamin (open squares) or K44A dynamin (open circles). Error bars indicate SEM; n = 11–17. There were no significant differences between groups. (C) Mean current density vs. voltage relationships for cells as in (A) expressing KCNQ1 and MinK alone (filled squares) or with wt (open squares) or K44A dynamin (open circles). Error bars indicate SEM; n = 10–11; *significant difference compared with other two groups, P < 0.0001. (D) Mean current density at +60 mV for cells expressing the longer variant of KCNQ1 (LQ1) alone or with wt or K44A dynamin (K44A). Error bars indicate SEM; n = 10–14. (E–G) Dyn, dynamin 2; EEA1, early endosomal antigen 1; MK, MinK; Q1, KCNQ1; Tf, transferrin; +, merge. Scale bars indicate 10 µm. White text inside images denotes fluorescent label; black text outside images denotes transfected cDNAs for each set. (E) Example of confocal images of COS-7 cells over-expressing MinK (green), alone (left), or with KCNQ1 (right) and incubated with fluorescent Tf (red) to label Tf receptor internalized in clathrin-coated pits. Merged image indicated by ‘+’; yellow indicates MinK-Tf co-localization. (F) Example of confocal image of COS-7 cells over-expressing MinK (green) and K44A dynamin 2, without (left) or with (right) KCNQ1. Cells were incubated with fluorescent Tf (red) to label Tf receptor. Merged image indicated by ‘+’. (G) Example of confocal images of a triple-labelled COS-7 cell over-expressing KCNQ1 (blue) and wild-type dynamin 2. The cell was fed with fluorescent Tf (green), and EEA1 was also labelled (red). Merged image indicated by ‘+’.

2.2. Electrophysiology

Whole-cell voltage clamp studies of COS-7 cells were performed as we described previously.¹⁷ Bath solution was (in mM): 135 NaCl, 5 KCl, 1.2 MgCl₂, 5 HEPES, 2.5 CaCl₂, and 10 d-glucose, pH 7.4. Pipettes were 3–5 MΩ resistance when filled with intracellular solution containing (in mM): 10 NaCl, 117 KCl, 2 MgCl₂, 11 HEPES, 11 EGTA, and 1 CaCl₂, pH 7.2. The KCNQ1 inhibitor XE991 (Tocris Biosciences) was applied at 2 µM via the bath where indicated. Whole-cell patch clamp recordings were performed at 22–25°C 2 days post-transfection using an IX50 inverted microscope equipped with epifluorescence optics for GFP detection (Olympus), a Multiclamp 700A Amplifier, a Digidata 1300 Analogue/Digital converter, and pClamp9 software (Axon Instruments). Leak and liquid junction potentials (<4 mV) were not compensated for when generating current–voltage relationships. For analysis of voltage dependence and activation kinetics, cells were held at −80 mV and subjected to 3 s test pulses from −120 to +60 mV in 20 mV increments, followed by a 2 s tail pulse to −30 mV. Current–voltage relationships were obtained by measuring peak current during depolarizing pulses. Statistical significance for differences observed between groups of cells in patch-clamp experiments was assessed by one-way ANOVA with P < 0.05 being indicative of significance.

2.3. Confocal microscopy

Confocal microscopy and immunocytochemistry were performed essentially as we described previously,¹⁷ using primary antibodies raised against KCNQ1, HA, LAMP2 (to label late endosomes and lysosomes), and early endosomal antigen 1 (EEA1) (to label early endosomes) and fluorescent secondary antibodies. cDNAs encoding fluorescent markers were transfected to facilitate detection of the ER and Golgi. Clathrin-mediated endocytosis (CME)-internalized transferrin (Tf) receptor was labelled by feeding live cells with Alexa-fluor 488-conjugated human Tf (Invitrogen) using the supplier-recommended protocol. After mounting onto slides, immunofluorescence single confocal section images from the centre of the cell were captured using a Zeiss LSM 510 laser scanning confocal microscope and analysed using Zeiss LSM proprietary software and MetaMorph software.

2.4. Antibody and biotin feeding, western blotting, and co-immunoprecipitation

Antibody feeding was performed essentially as described previously,¹⁸ using anti-HA antibody to bind to surface-exposed HA-tagged MinK in COS-7 cells transfected with HA-tagged MinK, KCNQ1, and wild-type or K44A dynamin 2. Hyperosmotic (0.45 M) sucrose was used to inhibit CME. Internalization over a 1 h time-course was quantified as reduction in signal at 492 nm using a spectrophotometer to detect the anti-HA antibody via goat anti-mouse horseradish peroxidase-conjugated secondary antibody and O-phenylene diamine substrate, in non-permeabilized cells. Permeabilized cells were used as a control to eliminate the possibility of other forms of signal loss (schematic in Supplementary material online, Methods). Western blotting and co-immunoprecipitation of KCNQ1, MinK, and β-actin (as a protein concentration control) were performed using transfected COS-7 cells essentially as we described previously¹⁷ (Supplementary material online, Methods). KCNQ1–MinK complexes were immunoprecipitated using anti-HA antibody to bind to HA-tagged MinK, then KCNQ1 was visualized by western blot using anti-KCNQ1 antibody. Biotin feeding (schematic in Supplementary material online, Methods) was performed using adult guinea-pig myocytes. Adult guinea-pigs were housed and utilized according to the NIH Guide for the Care and Use of Laboratory Animals and Weill Medical College of Cornell University animal care and use policies. Surface-exposed proteins were biotinylated with glutathione-cleavable EZlink NH-SS-Biotin (Pierce), then after 30 min at 37°C ± 0.45 M sucrose, remaining surface-exposed (non-internalized) proteins were de-biotinylated using glutathione. Internalized, previously surface-biotinylated MinK and KCNQ1 were avidin-purified, and then visualized using western blotting.

3. Results

3.1. Inhibition of dynamin-dependent internalization augments KCNQ1–MinK currents

CME involves internalization of proteins from the cell surface in clathrin-coated pits. A GTPase, dynamin, is required for formation of clathrin-coated pits.^19,20 Clathrin-independent DDI of membrane proteins is also thought to occur, via caveolae in glycolipid rafts,²¹ but CME requires dynamin.²² Here, KCNQ1 and KCNQ1–MinK (I_Ks) channels were expressed in COS-7 cells alone or with wild-type dynamin 2, or dominant-negative K44A dynamin 2—which inhibits CME.²² Homomeric KCNQ1 whole-cell current density was not significantly affected by co-transfection of wild-type or K44A dynamin (Figure 1A and B). In contrast, I_Ks current was up-regulated four-fold by K44A dynamin, whereas wild-type dynamin had no detectable effect (Figure 1A and C). This suggested that I_Ks channels were endocytosed via MinK by a dynamin-dependent process occurring constitutively in COS-7 cells.

A longer variant of KCNQ1 (‘LQ1’, see Supplementary material online, Methods) gave lower overall current density than the shorter (sKvLQT1) form used in Figure 1A–C but was similarly unaffected by wild-type or K44A dynamin (Figure 1D). This lack of effect was expected because mutagenic disruption of potential endocytosis motifs of the form YXXϕ (where ϕ is a bulky, hydrophobic residue) and a dileucine motif in the portion of the KCNQ1 N-terminus absent in sKvLQT1 was previously shown to either have no effect or reduce rather than increase current density. These sites were therefore not predicted to be mediators of KCNQ1 endocytosis, although one or more of them are thought to play a role in targeting in polarized epithelial cells.²³ The ‘sKvLQT1’ variant of KCNQ1 was therefore used in the remainder of this study. This variant contains the C-terminal ‘PY’ motif recently found to mediate Nedd4.2-dependent ubiquitylation.¹⁵

3.2. MinK undergoes dynamin- and clathrin-dependent internalization

Regardless of KCNQ1 co-expression, MinK underwent internalization by CME, as indicated by co-localization with fluorescent-labelled Tf fed to live cells, an established marker of the CME pathway²⁴ (Figure 1E); internalization of MinK and Tf was disrupted by co-expression with K44A dynamin, again characteristic of CME (although unlike MinK some Tf was also internalized, presumably by macropinocytosis, as previously reported²⁵) (Figure 1F). In contrast, KCNQ1 did not undergo significant DDI or clathrin-dependent internalization when expressed in the absence of MinK, as indicated by a lack of co-localization with internalized Tf or EEA1 (another marker of the CME pathway) even when co-transfected with wild-type dynamin 2 to stimulate DDI (Figure 1G).

The dependence of MinK internalization on dynamin and clathrin was also examined using antibody feeding, an established assay of endocytosis kinetics. Anti-HA antibodies were applied externally to live COS-7 cells expressing HA-tagged MinK, KCNQ1, and wild-type or K44A dynamin, then internalization kinetics determined by quantifying surface-exposed anti-HA antibodies during a 60 min, 37°C incubation (schematic in Supplementary material online, Methods). K44A dynamin reduced MinK internalization at all time points, although internalization still occurred with K44A dynamin (Figure 2A), suggesting that dynamin-independent internalization processes were also occurring, consistent with previous findings.¹⁵ Similar experiments using permeabilized cells showed no reduction in HA signal, indicating there was no significant cell detachment or other type of net antibody loss over the time-course (Figure 2B). Incubation with hyperosmotic (0.45 M) sucrose, which specifically inhibits CME by depleting clathrin-coated pits,²⁶ more than halved the dynamin-dependent fraction of internalization indicating that at least some of the DDI involved CME (Figure 2C).

Figure 2.

Antibody and biotin feeding: MinK undergoes clathrin-mediated endocytosis/dynamin-dependent internalization in COS-7 cells and guinea-pig myocytes. (A) Time-course of dynamin-dependent endocytosis of the I_Ks channel in COS-7 cells determined through anti-HA antibody feeding. Surface expression of the I_Ks channel was determined using COS-7 cells transfected with KCNQ1, HA-MinK, and either K44A or wild-type dynamin 2 under non-permeabilized conditions. Data are expressed as the ratio of the mean absorbance readings at set time intervals (T_t) and the mean initial absorbance reading (T₀), standardized to 1. Error bars for each time point (including T₀) represent the standard error between non-standardized values (n = 5 wells per time point; qualitatively similar results were obtained in two repeats of this experiment). (B) Experiment performed as in (A) but under permeabilized conditions as a control for cell surface expression, cell viability, and the assay accuracy. Data are expressed as a ratio of the mean absorbance readings at set time intervals (T_t) and the mean initial absorbance reading (‘T₀’) (37°C, n = 4 for each time point). Error bars indicate SEM. (C) Effect of inhibition of clathrin-mediated endocytosis on MinK internalization. Time-course experiments were performed as in (A) but in the presence or absence of hyperosmotic sucrose (0.45 M), to inhibit clathrin-mediated endocytosis. Data are expressed as the mean ratio of the normalized absorbance values (T_t/T₀), subtracted from 1 to obtain the fraction endocytosed, from COS-7 cells expressing K44A dynamin 2 compared with wild-type dynamin 2, at 60 min. n = 3 independent experiments per time point; error bars indicate SEM. (D) Example of western blot of guinea-pig cardiac myocyte lysates (avidin-purified fraction or total lysate as indicated) following steps as in Figure B (Supplementary material online, Methods), with increased band intensity indicative of protection of membrane proteins by internalization. Anti-KCNQ1 or MinK antibodies were used as indicated; numbers on left indicate marker migration distances. (E) Mean band intensities from blots as in (D) for MinK and KCNQ1 from guinea-pig myocytes ± sucrose treatment. Experiments were performed using a different heart for each repeat; n = 8 (MinK) and n = 5 (KCNQ1). Error bars indicate SEM. *P < 0.05; ***P < 0.001.

CME was also found to internalize native MinK and KCNQ1 in freshly isolated adult guinea-pig cardiac myocytes, using a biotin feeding assay in which internalized proteins that were previously exposed to surface biotinylation are protected from glutathione-mediated removal of that biotin and can thus be avidin-purified and detected by western blot. MinK and KCNQ1 were both internalized by CME (sucrose-inhibited) but also by a clathrin-independent pathway, with MinK being approximately two-fold more reliant upon the clathrin-dependent pathway for internalization (Figure 2D and E). The data correlate qualitatively with data in Figures 1 and 2 and suggest that some KCNQ1 may be expressed at the myocyte surface without MinK, either alone or perhaps with other KCNE subunits as suggested by others,¹² rendering it insensitive to MinK-mediated CME. Quantitative differences in MinK CME between COS-7 cells and myocytes may stem from differences in the range of internalization processes—or quantity and type of dynamin, clathrin, or adaptins—available to KCNQ1–MinK in the two systems. In addition, there may be inherent differences between internalization of native guinea-pig KCNQ1–MinK, which is probably part of a larger macromolecular complex in vivo,²⁷ and over-expressed human KCNQ1–MinK, which may saturate some endogenous COS-7 endocytic pathways, leading to alternative routes being adopted. Thus, the CME observed in COS-7 cells was recapitulated with native I_Ks complexes and was not an over-expression artefact.

3.3. Three MinK motifs are necessary and sufficient for dynamin-dependent internalization

KCNQ1 belongs to the S4 superfamily of six-transmembrane-domain α-subunits, whereas MinK belongs to the KCNE family of single-transmembrane-domain ancillary subunits (Figure 3A). The non-reliance of MinK DDI on KCNQ1 suggested that MinK harbours intrinsic motifs coordinating DDI. Motif searches identified three potential endocytic motifs on the MinK intracellular C-terminal domain (Figure 3A and B). The first, DPFNVY, contains FNVY which is reminiscent of the YXXϕ motif which in other proteins mediates binding to AP-2 for their internalization in clathrin-coated pits.^28,29 This motif also resembles the FLVI sequence which participates in CFTR internalization.³⁰ DPYXXY variants of this motif are also present in MiRP1 and MiRP2 (Figure 3B). The MinK DPFNVY sequence, and its equivalents on MiRP1 and MiRP2, all fall into the category of a motif highly conserved in G-protein coupled receptors—(D/N)PX_2-3Y—that has been implicated in functions ranging from internalization to activation.^31–33 The second, MinK motif, a consensus PKC phosphorylation site (S102) was previously shown to determine I_Ks current density by an unknown mechanism in vitro and in vivo in mammalian heart.^34,35 The third motif, detected using MINIMOTIF MINER,³⁶ is a consensus SH3-binding domain at the extreme C-terminus of MinK (PSP) (Figure 3B).

Figure 3.

Motifs coordinating dynamin-dependent internalization of MinK. (A) Membrane topology of KCNQ1 and MinK. MinK C-terminus domain is highlighted in yellow. (B) Predicted intracellular C-terminal domains of human MinK, MinK-related peptide 1 (MiRP1) and MiRP2 with potential endocytosis motifs boxed yellow. Numbering corresponds to MinK. (C) MinK C-terminal mutants, designated C1–C8, analysed in this study. Mutations are in yellow; dashes indicate wild-type residues; for putative motifs all residues are shown. (D) Example of current traces recorded in COS-7 cells expressing KCNQ1 with C8-MinK and wild-type or K44A dynamin; voltage protocol as in Figure 1. (E) Mean current density at +60 mV for cells expressing KCNQ1 with wild-type or mutant MinK as indicated, and wild-type (red) or K44A dynamin (blue crosshatch). Error bars indicate SEM; n = 10–17. Significant difference with K44A compared with wild-type dynamin for each MinK variant is indicated by asterisks: *P < 0.05; **P < 0.01; ***P < 0.001. (F–I) Example of confocal images of COS-7 cells expressing KCNQ1, wild-type or C8-MinK, and wild-type or K44A dynamin as indicated. C8, C8-MinK (green); ER, endoplasmic reticulum (red); EL, endosomes/lysosomes (red); G, Golgi (red); Q1, KCNQ1 (blue); MK, MinK (green); +, overlay. Triple co-localization, white. Arrows described in text.

Each of the putative sites was disrupted by mutagenesis individually or in combination to produce MinK constructs with variant C-terminal domains designated C1–C8 (Figure 3C). DPFNVY was mutated to DPAAVY (C1), guided by mutagenesis of a similar motif in CFTR;³⁰ S102 was mutated to alanine to prevent phosphorylation (C2) or to aspartic acid to mimic constitutive phosphorylation (C3); and the threonine in TKPSP was mutated to a stop codon to eliminate the putative SH3-binding motif (C4). Combinations of these mutants were designated C5–C8 (Figure 3C). Motifs important to DDI would be predicted to reduce effects of K44A dynamin on I_Ks current density, assessed by whole-cell patch clamp (Figure 3D and E). C1, C2, and C4 constructs each showed reduced difference in current density between wild-type vs. K44A dynamin groups, whereas with C3 a more than two-fold difference was preserved (Figure 3E). Double-mutants C5, C6, and C7 also showed reduced difference in current density between wild-type vs. K44A dynamin groups compared with the difference observed for wild-type I_Ks. Significantly, the triple mutation in MinK (C8) eliminated the dynamin-dependent difference in current density (Figure 3D and E). In support of a lack of non-DDI-related effects of the mutants, each of the mutants showed similar current density to wild-type MinK channels when co-expressed with K44A dynamin (Figure 3E). The data strongly suggest that each of the three sites is necessary, and the combination of the three sites sufficient, for DDI of KCNQ1–MinK complexes. Furthermore, the S102A (C2) and S102D (C3) results provide a potential mechanism for the prior observation that S102 phosphorylation regulates I_Ks current density.³⁴ These conclusions were reinforced using confocal microscopy. When co-expressed with wild-type dynamin and KCNQ1, MinK was localized almost exclusively in the Golgi and endosomes/lysosomes (Figure 3F). Some KCNQ1 co-localized with MinK in the Golgi (yellow arrow) and endosomes/lysosomes (white arrow), but much of it was detected alone in the ER or at the cell surface (grey arrow). The apparently inefficient surface trafficking of KCNQ1 is consistent with previous reports.³⁷ In contrast, when co-expressed with K44A dynamin and KCNQ1, MinK was more widely distributed throughout the cell (Figure 3G) and showed some co-localization with KCNQ1, at the plasma membrane (grey arrow) and inside the cell. The proposed DDI-deficient MinK mutant, C8, did not localize significantly with the Golgi or endosomes/lysosomes when co-expressed with KCNQ1 and wild-type dynamin (Figure 3H) or K44A dynamin (Figure 3I). Instead, C8 showed increased surface expression, and some co-localization with KCNQ1 there in both cases (Figure 3H and I—yellow arrow) and also internally (white arrow). These data also suggested that MinK coordinates KCNQ1–MinK DDI and that the three MinK C-terminal motifs are necessary and sufficient for DDI.

3.4. MinK-mediated dynamin-dependent internalization dictates surface net channel composition

The confocal data suggested that a significant fraction of KCNQ1–MinK complexes remain intact, but within the cell rather than at the cell surface, following DDI (Figure 3F). However, it was still possible that DDI, or the mutations we introduced into MinK, produced internalization-independent effects such as disruption of KCNQ1–MinK complex formation. This possibility was discounted by the findings that total KCNQ1 protein levels were not significantly affected by the co-expressed variant of MinK or dynamin (Supplementary material online, Figure S1A), neither were the total cell levels of KCNQ1–MinK complexes (Supplementary material online, Figure S1B and C).

Although total cell KCNQ1–MinK complex density was unaffected, the confocal and functional data suggested differences in the localization of these complexes—specifically that DDI decreased the surface expression of KCNQ1–MinK relative to homomeric KCNQ1 channels at the cell surface. This was testable because KCNQ1–MinK channels have 5–10-fold slower activation and are 14-fold less sensitive to inhibition by XE991,³⁸ than homomeric KCNQ1. Accordingly, KCNQ1 activation was significantly (more than three-fold) faster with wild-type dynamin vs. K44A dynamin when co-expressed with wild-type MinK (Figure 4A). With S102D-MinK (C3), this difference was preserved, whereas with other single MinK mutants (C1–C3), there was no longer a statistically significant difference with wild-type vs. K44A dynamin. With triple-mutant MinK (C8), the difference in kinetics was abolished completely (Figure 4A). In support of a lack of non-DDI-related effects of the mutants, none of the mutants showed significantly different activation kinetics to wild-type MinK channels when co-expressed with K44A dynamin (Figure 4A). Turning to pharmacology, homomeric KCNQ1 current (with no co-transfected MinK) was inhibited 84 ± 4% by 2 µM XE991. Current generated by co-expression of MinK and KCNQ1 with K44A dynamin was much less sensitive, being inhibited 4 ± 3% by 2 µM XE991, whereas with wild-type dynamin this current was inhibited 36 ± 3%, pushing the sensitivity closer to that of homomeric KCNQ1 (Figure 4B and C). Thus, data from both activation kinetics and XE991 pharmacology supported the hypothesis that DDI dictates the ratio of homomeric to heteromeric complexes at the cell surface, increasing the relative proportion of KCNQ1 compared with KCNQ1–MinK channels at the cell surface.

Figure 4.

Dynamin (Dyn)-dependent internalization dictates net surface I_Ks subunit composition. (A) Time-to-half-peak current at +60 mV for cells expressing KCNQ1 with MinK or mutant MinK (C1–C8 as indicated), and wild-type (wt) (solid) or K44A Dyn (crosshatch). Error bars indicate SEM; n = 10–17. Significant difference with K44A compared with wt Dyn for MinK variants: *P < 0.05; ***P < 0.001. (B) Example of current traces recorded in COS-7 cells expressing KCNQ1 with MinK and wt or K44A Dyn before (black trace) and after (gray trace) application of 2 µM XE991. Scale bars: vertical, 200 pA (wt Dyn) and 1 nA (K44A-Dyn); horizontal, 1 s. Voltage protocol as for +60 mV pulse in Figure 1. (C) Mean % inhibition by 2 µM XE991 at +60 mV of currents in COS-7 cells expressing KCNQ1, alone (Q1) or with MinK, and wt or K44A Dyn, as indicated. Error bars indicate SEM; n = 5–6. Significant difference between groups: *P < 0.0005; **P < 0.0001.

4. Discussion

A previous study of I_Ks trafficking focused on the movements of KCNQ1 in Xenopus oocytes, with the important finding that KCNQ1 (co-expressed with MinK) undergoes serum- and glucocorticoid-inducible kinase 1 (SGK1)-dependent forward trafficking through a Rab11 exocytotic pathway, connecting this process to the β-adrenergic pathway and thus coupling I_Ks current density to metabolic stimulation.³⁹ In that study, Rab5 was also shown to participate in internalization of KCNQ1; however, although MinK was co-expressed with KCNQ1, it was not tracked, neither was KCNQ1 expressed alone, therefore the channel subunit(s) and motifs mediating either trafficking process were not determined. More recently, KCNQ1 and KCNQ1–MinK currents were found to be reduced in cardiac myocytes by ubiquitylation of KCNQ1. This is thought to be due to subsequent internalization of KCNQ1 and presumably also MinK.¹⁵ The MinK-dependent internalization we now describe differs from these aforementioned processes because it can be used to alter surface channel composition, favouring surface expression of KCNQ1 complexes lacking MinK. Importantly, this change in channel composition, from MinK–KCNQ1 to homomeric KCNQ1, not only accelerates gating kinetics but also decreases current density, even if the net density of KCNQ1 at the cell surface remains constant due to KCNQ1 recycling. This is because MinK increases the unitary conductance of KCNQ1 four-fold.⁴ Thus, a shift in surface channel composition without altering net channel number at the surface can still significantly affect current density—explaining the relatively large increases in current density produced by K44A dynamin (Figure 1C).

How important this process is in shaping I_Ks current properties in human heart, and how significant this mode of internalization is compared with other, KCNQ1-mediated modes, remains to be seen. In human heart, measurement of cDNA copy number to quantify mRNA levels indicates that in both ventricular and atrial myocytes, KCNQ1 mRNA is present at more than five-fold higher levels than that of MinK,¹² which sets the stage for MinK-dependent endocytic sorting of channel composition. In some ligand-gated channels, such as somatostatin and opioid receptors, the relative amounts of the different constituent subunits that reach the plasma membrane affect CME rates of each of the subunits at the membrane, thus shaping the subunit composition of the surface-expressed population of these receptors.⁴⁰ AMPA receptors lacking GluR2 subunits are favoured over those with GluR2 subunits by CME of the latter in ischaemic neurons.⁴¹ Further, internalization of the Ca²⁺- and voltage-activated K⁺ channel hSlo is reportedly enhanced by KCNMB2 co-expression.⁴²

The ability to dynamically regulate MinK internalization potentially facilitates responsive control of Kv current density and gating kinetics. Two previous studies showed that PKC phosphorylation of MinK-S102 decreases I_Ks current in rat and mouse in vitro and in vivo, although the mechanism was not determined.^34,35 PKC also decreases human I_Ks current,⁴³ but not guinea-pig due to the absence of a serine at the equivalent position in the latter.^34,35 The data herein suggest the underlying mechanism is augmentation of I_Ks internalization by MinK-S102 phosphorylation. In a similar fashion, PKC phosphorylation of the human non-gastric H⁺–K⁺-ATPase causes it to be endocytosed by CME, whereas CME is not favoured when this protein is dephosphorylated.⁴⁴

In summary, cellular excitability is highly dependent upon the density and functional characteristics of the Kv channels expressed at the cell surface. Here, we show that β-subunit-dependent internalization of a Kv channel can dictate net surface channel subunit composition, gating kinetics, and pharmacology. It is important to mention that KCNQ1 α-subunits may undergo DDI without MinK, under specific conditions, in different cell types or with regulatory stimuli not present in our experiments. Future analyses will be aimed at assessing whether differential internalization of other cardiac-expressed MiRPs that regulate KCNQ1 plays a role in fine-tuning the cardiac I_Ks current by favouring surface expression of specific complexes, and whether this paradigm applies to α–β type Kv channels in general.

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Funding

This work was supported by the NIH (HL079275 to G.W.A.).