A minK–HERG complex regulates the cardiac potassium current I Kr

MinK is a widely expressed protein of relative molecular mass ~15K that forms potassium channels by aggregation with other membrane proteins 1–3 . MinK governs ion channel activation 4 , regulation by second messengers 5,6 , and the function and structure of the ion conduction pathway 7,8 . Association of minK with a channel protein known as KvLQT1 produces a voltage-gated outward K + current (I sK ) resembling the slow cardiac repolarization current (I Ks ) 9,10 . HERG , a human homologue of the ether-a-go-go gene of the fruitfly Drosophila melanogaster , encodes a protein that produces the rapidly activating cardiac delayed rectifier (I Kr ) 11,12 . These two potassium currents, I Ks and I Kr , provide the principal repolarizing currents in cardiac myocytes for the termination of action potentials 13,14 . Although heterologously expressed HERG channels are largely indistinguishable from native cardiac I Kr , a role for minK in this current is suggested by the diminished I Kr in an atrial tumour line subjected to minK antisense suppression 15 . Here we show that HERG and minK form a stable complex, and that this heteromultimerization regulates I Kr activity. MinK, through the formation of heteromeric channel complexes, is thus central to the control of the heart rate and rhythm. We + either alone (control) or in combination with HA–minK or HA–D77N-minK. Cx43– Myc was included in all transfections. Lysates were fractionated into plasma cytosolic and membrane fractions. HERG–Myc and Cx43–Myc were recovered exclusively in the membrane fraction. Co-expression of minK or D77N-minK did not affect the abundance of HERG (n = 3). f , CHO-HERG cells stably expressing wild-type HERG were fractionated as described and detected with antisera against HERG. Unlike actin, HERG localized exclusively in the membrane fraction. Co-expression of HA–minK or HA–D77N-minK did not affect the abundance of HERG (n = 3). g , Pulse-chase analysis. The rate of degradation of metabolically labelled HERG was not significantly different in cells transfected with HERG–Myc alone or in combination with minK.

Functional expression of HERG and HERG with minK. a, A CHO-HERG cell showing voltage-gated K + current typical for IKr. The holding potential was -70 mV, cells were depolarized to potentials between 0 and +60 mV and tail currents measured during repolarization to -40 mV. b, A CHO cell transiently transfected with minK showing no voltagegated currents. c, A CHO-HERG cell transiently transfected with minK showing HERG-minK, IKr-like K + currents. d, E4031 at 500 nM completely blocks all voltage-activated K + conductance in a cell expressing minK and HERG.

Figure 2
MinK alters functional expression of HERG in CHO cells. a, HERG current density is augmented by minK but not D77N-minK. The bar graph summarizes current densities for CHO-HERG cells (n = 55 cells), CHO-HERG cells transfected with wild-type minK (n = 39 cells), or CHO-HERG cells transfected with D77N-minK (n = 37 cells). Peak tail currents were measured as for Fig.1 and normalized for cell capacitance. b, Voltage-activation curves; normalized peak tail current density at -40 mV was plotted against test potential, demonstrating a left shift in voltage-dependent activation. Current density is greater for HERG-minK at all potentials that activate current (inset: data plotted without normalization). Data were fitted to the Boltzmann relation, y = 1/(1 + exp[(Vh -V)/k]), where V is the test potential, Vh the potential at half-maximal activation (+12 mV for HERG and +6 mV for HERG-minK) and k is the slope factor (11 mV for both). c, Current-voltage relation for HERG and HERG-minK cells. Normalized current during depolarization is plotted against the membrane potential. There is a hyperpolarizing shift in voltage dependence that is most pronounced at potentials greater than +20 mV. Current density is greater for HERG-minK than HERG alone at all voltages (inset: data plotted without normalization). d, Steady-state inactivation 30 was assessed by brief steps to various hyperpolarized potentials from +20 mV; the plot shows peak current released from inactivation at +20 mV against test potential. Inset shows voltage pulse protocol with X showing the voltage steps and Y indicating where current was measured. HERG and HERG-minK channels (τact: HERG, 282 ± 72 ms, n = 16; HERG-minK, 246 ± 38 ms, n = 17, P is NS; τfast deactivation: HERG, 477 ± 78 ms, n = 16; HERG-minK, 428 ± 64 ms, n = 17, P is NS).
As changes in gating kinetics did not explain the increased current density, single channel conductances were determined. Unitary conductances for HERG and HERG-minK channels between -30 and +40 mV were 10.3 ± 0.5 pS (n = 8) and 10.6 ± 0.4 pS (n = 14), respectively (Fig. 3). Thus, changes in i also failed to account for increased whole-cell current density. Neither did changes in membrane area lead to increased currents; the surface areas of cells expressing HERG or HERG-minK were not significantly different (34.1 ± 2.7 pF, n = 50 and 29.7 ± 2.7 pF, n = 46, respectively). Therefore, a change in the number of functional channels in the plasma membrane must underlie the effect of minK on HERG currents. These currents appear to result from HERG-minK channels rather than induction of non-HERG K + conductances because E4031, a specific blocker of IKr, abolished all voltage-gated K + currents in HERG-and HERG-minK-expressing cells (Fig. 1d) at concentrations used to distinguish between IKr and IKs (0.5 and 1.0 µ,M) 13 .
To assess whether minK influenced HERG by direct physical contact and formation of a stable heteromeric complex, the proteins were epitope-tagged for protein analysis. A c-Myc epitope at the carboxy terminus of HERG had no measurable effect on channel currents (data not shown). Similarly, a haemagglutinin (HA) epitope between minK amino acids 22 and 23 was also well tolerated, as shown before 5 , and did not alter its augmentation of HERG currents or formation of functional IKs channels with KvLQT1 (data not shown). Expression of HA-minK produced three specific bands corresponding to relative molecular masses (Mr) of ~15K-25K, which represent unmodified and glycosylated forms of minK ( Fig.  4a), as expected 4 . Expression of HERG-Myc yielded a single band of Mr ~ 130K, consistent with the predicted mobility of HERG (Fig. 4a).
When HERG-Myc and HA-minK were co-expressed, immune-precipitation of either protein led to recovery of both channel subunits (Fig. 4b,c; n = 3). Although D77N-minK did not augment HERG currents (Fig. 2a), HA-D77N-minK did associate with HERG to form a stable complex like the wild-type protein (Fig. 4c, lane 4; n = 2). The association between HERG and minK was specific; when Myc-tagged connexin 43 (Cx43-Myc), an unrelated membrane channel subunit, was co-expressed with both HA-minK and HERG-Myc, immunoprecipitation with anti-HA antibody recovered only HERG and no Cx43 (Fig. 4d). Although co-expression of HERG and minK increased the number of functional channels, it did not alter the amount of HERG protein on the cell surface. Western blot analysis was performed on membrane and cytosolic/microsomal fractions prepared from CHO cells transiently transfected with HERG-Myc and Cx43-Myc, either alone or in combination with HA-minK or HA-D77N-minK. In each case, all detectable HERG-Myc was recovered in the membrane fraction and the abundance of HERG-Myc (and Cx43-Myc) was unaffected by co-expression of HA-minK or HA-D77N-minK ( Fig. 4e; n = 3). The same results were obtained when either HA-minK or HA-D77N-minK was expressed in the stable CHO-HERG cell line; neither the membrane localization of HERG nor its level of expression was affected (Fig. 4f). Co-expression of minK with HERG also did not protect HERG from degradation. Pulse-chase analysis showed time constants of degradation of ~11.2 h and 11.6 h for HERG in the absence and presence of minK, respectively ( Fig. 4g; n = 2).
Our results show that minK physically associates with HERG and that the interaction leads to increased IKr current density. The fact that both non-glycosylated and glycosylated forms of minK bind to HERG suggests that association takes place early in the protein synthesis and trafficking pathway, probably in the endoplasmic reticulum. As minK increases the number of functional HERG channels in the plasma membrane but does not alter cell-surface expression of the HERG protein, minK appears to increase the active fraction in a membrane pool containing active and dormant channels. This type of shift in channel activity may underlie upregulation of minK function in oocytes after treatment with cyclic AMP1. Similarly, cAMP has been shown to regulate the proportion of functional nicotinic acetylcholine receptor channels in chick ciliary ganglion cells 17 . Such mode switching is reminiscent of the gating behaviour seen in RCK4 K + channels, where a fraction of channels in oocyte patches fail to open with depolarization when external K + is removed 18 .
The consequences of minK interaction with KvLQT1 and HERG are surprisingly similar. We have shown here that minK alters HERG channel gating kinetics and increases macroscopic IKr current density. By an as-yet unknown mechanism, minK also acts on KvLQT1 to alter its gating kinetics (to those characteristic of IKs) and to increase macroscopic current 9,10 . Although HERG produces IKr-like currents when expressed alone whereas KvLQT1 strictly requires minK to create recognizable IKs channels, this may be a quantitative rather than a qualitative difference. In this respect, it is notable that D77N-minK fails to augment IKr currents even though it physically associates with HERG protein (Figs 2a, 4c), just as it fails to produce IKs currents but can compete with wild-type minK when co-expressed in Xenopus oocytes 2 . This suggests that similar molecular features of minK are important for influencing the function of HERG and KvLQT1.
Inasmuch as minK is required for IKs and alters IKr, its expression and modulation by second messengers 5,6 will play a central role in determining the amount and character of the K + conductances in individual cardiac myocytes. Moreover, expression of minK, HERG, KvLQT1 and other K + channel proteins is region-specific in the heart and varies in pathological states, producing spatial heterogeneity in current densities and phenotypes 19,20 . Mutations in KvLQT1 and HERG produce hereditary long-QT syndrome (LQTS) by decreasing IKs and IKr (refs 21, 22), so minK should be expected to act as a modifier of genetically determined LQTS. As HERG has also been implicated in acquired arrhythmias 11,23 , minK may be involved in susceptibility to acquired LQTS and pro-arrhythmic drug phenomena. In the mouse, disruption of the minK gene results in shaker/waltzer behaviour and hair-cell degeneration 24 . This null phenotype suggests an association between mutations in minK and Jervell and Lange-Nielsen syndrome, an autosomal recessive form of LQTS that is associated with deafness 23 .
These results indicate a broad role for minK in determining the functional expression of delayed rectifier K + currents in the heart. The physical and functional interaction of minK with two diverse K + channel types (KvLQT1 and HERG) and its wide tissue distribution (heart, ear, kidney, pancreas, T cells and gastrointestinal tract) [25][26][27][28] suggest that the role of minK in normal physiology and disease pathogenesis may be even more extensive.

Methods
Plasmid construction and transfections. HERG was epitope-tagged by replacing the three C-terminal amino acids of the wild-type protein with ISMEQKLISEEDLN, derived from the c-myc gene recognized by monoclonal antibody 9E10. Construction of HA-tagged minK and D77N-minK have been described 1,2 . CHO cells stably expressing HERG were obtained by transfection with a pCI-based vector carrying HERG and selection with G418. Transient transfections were carried out by electroporation with the indicated minK constructs together with green fluorescent protein.
Protein analysis. Transfected cells were lysed in Tris lysis buffer (0.15 M NaCl, 1% NP-40, 1% CHAPS, 20 mM NaF, 10 mM Na3VO4, 1 mM PMSF and 50 mM Tris, pH 7.4). Lysates were cleared by centrifugation at 10,000g for 30 s. For direct immunoblot analysis, samples were separated by SDS-PAGE and epitope-tagged proteins detected with appropriate monoclonal antibodies conjugated to horseradish peroxidase (Boehringer Mannheim). For detection of untagged HERG, membranes were incubated with rabbit anti-HERG polyclonal antisera 29 , followed by horseradish peroxidase-conjugated protein A. Lysates were immunoprecipitated with either 5 µ,g anti-HA 12CA5 or 3 µ,g anti-Myc 9E10 monoclonal antibody for 2 h at 4 °C, followed by incubation with protein G-Sepharose for 1 h at 4 °C. For pulse-chase analysis, cells were labelled for 30 min with 35 S-methionine/cysteine, chased in medium with excess unlabelled methionine/cysteine for varying times and collected by detergent lysis. After immunoprecipitation and SDS-PAGE, gels were visualized by fluorography and quantified by densitometry (Molecular Analysis, Biorad). For subcellular fractionation, adherent cells were detached by incubation with EDTA and gently disrupted by Dounce homogenization. Cellular lysates were layered on 45% sucrose cushions and membrane and cytosolic/ microsomal fractions recovered after centrifugation at 7,000g for 20 min.