A superfamily of small potassium channel subunits: form and function of the MinK-related peptides (MiRPs)

1. INTRODUCTION 358 1.1 Summary 358 1.2 Overview 359 1.3 Four classes of pore-forming K+ channel subunits – necessary and (sometimes) sufficient 361 1.4 Soluble and peripheral membrane proteins that interact with P loop subunits to alter function 362 1.5 Integral membrane proteins that interact with P loop subunits to alter function 363 2. MinK DETERMINES THE FUNCTION OF MIXED CHANNEL COMPLEXES 363 2.1 The KCNE1 gene product (MinK) gives rise to K+-selective currents and controversy 363 2.2 MinK assembles with a P loop protein, KvLQT1, to form K+ channels with unique function 364 2.2.1 Single-channel conductance of KvLQT1 and MinK/KvLQT1 channels 366 2.2.2 Other differences between KvLQT1 and MinK/KvLQT1 channels 367 2.3 MinK assembles with HERG, another P loop subunit, to regulate channel activity 368 2.4 MinK does not form chloride-selective ion channels 368 3. EXPERIMENTAL AND NATURAL MinK MUTATIONS 369 3.1 Site-directed mutations 369 3.1.1 MinK mutation alters basic channel attributes and identifies key residues 369 3.1.2 MinK is a Type I transmembrane peptide 370 3.1.3 MinK is intimately associated with the IKs pore 370 3.1.4 The number of MinK subunits in IKs channel complexes 372 3.2 KCNE1 mutations associated with arrhythmia and deafness alter IKs channel function 373 3.3 Summary of MinK sites critical to IKs channel function 374 4. MinK-RELATED PEPTIDES: AN EMERGING SUPERFAMILY 374 4.1 KCNE2, 3 and 4 encode MinK-related peptides 1, 2 and 3 (MiRPs) 374 4.2 MiRP1 assembles with a P loop protein, HERG, to form K+ channels with unique function 375 4.2.1 MiRP1 alters activation, deactivation and single-channel conductance 376 4.2.2 MiRP1 alters regulation by K+ ion and confers biphasic kinetics to channel blockade 378 4.2.3 Stable association of MiRP1 and HERG subunits 380 4.3 KCNE2 mutations are associated with arrhythmia and decreased K+ flux 383 4.4 Summary of the evidence that cardiac IKr channels are MiRP1/HERG complexes 385 5. MinK-RELATED PEPTIDES: COMMONALTIES AND IMPLICATIONS 386 5.1 Genetics and structure 386 5.2 Cell biology and function 387 6. ANSWERS, SOME OUTSTANDING ISSUES, CONCLUSIONS 387 7. ACKNOWLEDGEMENTS 389 8. REFERENCES 389 MinK and MinK-related peptide 1 (MiRP1) are integral membrane peptides with a single transmembrane span. These peptides are active only when co-assembled with pore-forming K+ channel subunits and yet their role in normal ion channel behaviour is obligatory. In the resultant complex the peptides establish key functional attributes: gating kinetics, single-channel conductance, ion selectivity, regulation and pharmacology. Co-assembly is required to reconstitute channel behaviours like those observed in native cells. Thus, MinK/KvLQT1 and MiRP1/HERG complexes reproduce the cardiac currents called IKs and IKr, respectively. Inherited mutations in KCNE1 (encoding MinK) and KCNE2 (encoding MiRP1) are associated with lethal cardiac arrhythmias. How these mutations change ion channel behaviour has shed light on peptide structure and function. Recently, KCNE3 and KCNE4 were isolated. In this review, we consider what is known and what remains controversial about this emerging superfamily.


. Summary
MinK and MinK-related peptide  (MiRP) are integral membrane peptides with a single transmembrane span. These peptides are active only when co-assembled with pore-forming K + channel subunits and yet their role in normal ion channel behaviour is obligatory. In the resultant complex the peptides establish key functional attributes : gating kinetics, single-channel conductance, ion selectivity, regulation and pharmacology. Co-assembly is required to reconstitute channel behaviours like those observed in native cells. Thus, MinK\KvLQT and MiRP\HERG complexes reproduce the cardiac currents called I Ks and I Kr , respectively. Inherited mutations in KCNE (encoding MinK) and KCNE (encoding MiRP) are associated with lethal cardiac arrhythmias. How these mutations change ion channel behaviour has shed light on peptide structure and function. Recently, KCNE and KCNE were isolated. In this review, we consider what is known and what remains controversial about this emerging superfamily.

. Overview
Potassium channel proteins provide a path for rapid, highly selective diffusion of K + ions across cell membranes, controlling cellular electrical activity and thereby facilitating a range of fundamental physiological processes including nervous signaling, muscular contraction and fluid and electrolyte homeostasis (Hille, ). In electrically active tissues, like the heart, the primary roles of K + channels are to establish resting membrane potential, repolarize cells after excitatory depolarization, regulate action potential frequency and limit the effects of excitatory influences (Zipes & Jalife, ). Here, we describe the attributes of an emerging superfamily of small transmembrane peptides. Often referred to as ' accessory ' subunits, this appellation disguises their essential role in normal ion channel function. MinK peptide, encoded by the KCNE gene, was thought to be unique in form and function. This conclusion was based on the failure to identify MinK homologues (or proteins subserving similar function) in over a decade. Recently, three MinK-related genes were isolated and the first, MinK-related peptide  (MiRP) was evaluated in detail (Abbott et al. ). Below we compare the attributes of MinK and MiRP and find great overlap in the channel functions they influence. Both peptides associate with pore-forming K + channel subunits to determine how the resultant channel complex opens and closes, conducts ions, is regulated by second messengers, gains cell surface expression and is regulated by drugs and other small molecules ( Inheritance of one DN MinK encoding allele is associated with long QT syndrome (LQTS), a disorder that predisposes to torsades de pointes and sudden death ; with two mutant alleles, patients ' are subject to arrhythmia and congenital deafness (Schulze-Bahr et al.  ; Splawski et al.  ; Tyson et al.  ; Duggal et al. ). Similarly, individuals with QE MiRP form cardiac I Kr channels with abnormal activation, deactivation, regulation by external K + and sensitivity to the antibiotic clarithromycin, an agent well-tolerated by the general population ; patients with MiRP mutations have presented both with acquired and inherited arrhythmia (Abbott et al. ).
While generalizations can be extracted about the function and structure of ion channels containing MinK or MiRP from their study in wild-type and mutant forms, we approach this exercise with trepidation. The record is strewn with provocative MinK-related inquiries that arrive at mutually exclusive conclusions -is MinK an ion channel or a carrier-type transporter? ; does it function alone or in complexes? ; is MinK obligatory or accessory? ; is it pore-associated or peripheral? ; is it part of a K + channel and a Cl − channel? ; is it part of I Ks and I Kr  . Four classes of subunits have been identified that carry one or two P loops (Fig.  a). The first class of K + channel subunits to be recognized is well-represented by its founding member, Shaker, from Drosophila melanogaster (Kamb et al.  ; Papazian et al.  ; Tempel et al.  ; Pongs et al. ). Each subunit in this class has one P loop and six or more TMDs. Enjoying wide tissue distribution in both invertebrates and vertebrates, this class forms voltage-gated K + -selective channels and cyclic-nucleotide-gated ion channels with P\TMD subunits and Ca# + -activated K + channels with subunits predicted to have a P\TMD topology (Chandy & Gutman,  ; Kohler et al. ). It was studies of Shaker and its homologues that revealed the P loop dipping into and out of the membrane from the extracellular surface to create the external portion of the ion conduction pore ( (TMTTVGYG) was first identified and characterized (Heginbotham et al.  ; MacKinnon, ). In Shaker channels the S and S TMDs were shown to form the cytoplasmic vestibule of the pore by close association with residues of the P loop (Liu et al.  b, ). In voltage-gated Na + subunits and then in Shaker the S TMD was revealed as the primary sensor for changes in the transmembrane electric field (Stuhmer et al.  ; Papazian et al.  ,  ; Yang & Horn,  ; Larsson et al. ). Using Shaker subunits MacKinnon () demonstrated that subunits with one P loop form ion channels by tetrameric association.
A second class of pore-forming K + channel subunits is characterized by one P loop and two TMDs (P\TMD) (Fig.  a) The third and fourth classes of pore-forming K + channel subunits have two P loops and eight or four TMDs (Fig.  a) The two P domain subunits are expected to form channels by dimerization but even this is as yet unproven.

. Soluble and peripheral membrane proteins that interact with P loop subunits to alter function
The opening and closing of many ion channels is controlled by interaction with soluble or membrane-associated accessory proteins (Fig.  b)

. Integral membrane proteins that interact with P loop subunits to alter function
Three types of integral membrane proteins have been identified that co-assemble with P loop subunits to alter function ( Fig.  c). In each case, these subunits are required to form channels that function like those in native tissues. One type belongs to the ABC transporter superfamily ; these subunits have  TMDs and a wide variety of tissue-specific subtypes. Thus, cardiac and pancreatic I KATP channels form with four P\TMD subunits and four sulphonylurea receptors (SURs) ; this establishes the characteristic gating and drug sensitivity of Another ABC transport protein that regulates channels formed by P\TMD subunits is the cystic fibrosis transmembrane regulator (CFTR) (Ho, ). A second type of transmembrane protein is characterized by two predicted TMDs ; these subunits associate with the P\TMD subunits that form Ca# + -activated K + channels to alter sensitivity to Ca# + , channel pharmacology and fast inactivation The gene for rat MinK was cloned by Takumi and co-workers () based on its ability to produce K + currents in Xenopus laevis oocytes. Rat kidney mRNA injected into the cells induced a new outward current in response to membrane depolarization. By size-fractionation and iterative re-testing they isolated a single cDNA clone encoding a product smaller than any previously known ion channel subunit -the rat gene predicted an open reading frame of  amino acids, one TMD, and shared homology with no known proteins. The human gene ( However, the small size and unique attributes of MinK currents relentlessly engendered new concerns. First, there was worry that MinK might be a carrier rather than a channel-type transporter. This issue was raised because the unitary conductance was apparently quite small -during depolarizing pulses current developed smoothly without discernible single channel events. Second, multiple lines of investigation argued that MinK, whether transporter or channel, required another structural subunit or regulatory influence to function. Thus, MinK did not induce currents in some cell types despite expression of the protein at the cell surface (Lesage et al. ). Further, expression of increasing amounts of MinK on the oocyte plasma membrane did not lead to increasing current, rather, current reached a maximum as if some other factor was present in limiting amounts (Blumenthal & Kaczmarek,  ; Wang & Goldstein, ). Finally, currents in oocytes were blocked by a covalent chemical modifier in a fashion consistent with physical occlusion of the ion channel pore at a non-MinK site (Tai et al. ). Clarity seemed unattainable when it was then reported that MinK was capable of forming Cl − channels (Attali et al. ). These conundra have been resolved in the past three years.
. MinK assembles with a P loop protein, KvLQT , to form K + channels with unique function MinK does not work alone but with a pore-forming P\TMD subunit, KvLQT (Fig.  a)   Channels were formed by KvLQT subunits ; wild-type hMinK and KvLQT subunits ; DN hMinK and KvLQT subunits ; and, SL hMinK and KvLQT subunits. Patches were held at k mV and depolarized for  s to voltages from - mV in steps of  mV. (b) Activation of MinK\KvLQT channels at j mV, sampled at  kHz and filterd at  kHz, otherwise as in (a). (c) Variance-current relationship for data in (a) with KvLQT ( ) and hMinK\KvLQT () channels ; the conductances calculated in these patches were . and  pS, respectively. (d ) Unitary conductance as a function of experimental bandwidth for channels formed by KvLQT ( ), hMinK\KvLQT (), DN hMinK\KvLQT ($) and SL hMinK\KvLQT (X) from non-stationary variance analysis. The curves were obtained measuring the variance at  mV using cut-off frequencies from n to  kHz ; curves were scaled to the value of conductance determined at  kHz. Data were sampled at  kHz and digitally filtered at the indicated frequencies.
Each curve was generated from the average of three patches. Then, in , positional cloning was used to define a novel gene designated KvLQT that was associated in mutant form with the inherited cardiac arrhythmia LQTS (Wang et al.  b). The predicted product was homologous to other P\TMD K + channel genes with  residues, six TMDs and a classical poreforming P loop ; KvLQT mediated a rapidly-activating, non-inactivating K + current. It was the absence of a KvLQT-like current in native cardiac myocytes that led Sanguinetti and co-workers ( b) to co-express MinK and KvLQT subunits whereupon they observed currents with attributes like those of native I Ks channels. The function of MinK in oocytes (supposedly alone) was now rationalized by its co-assembly with a Xenopus laevis variant of KvLQT that is naturally expressed in oocytes (Sanguinetti et al.  b).
Identification of the 'missing ' subunit clarified many key issues and permitted detailed study of channels formed with MinK. Thus, cloning the gene for Unitary conductance estimates were dependent on analysis bandwidth due to rapid channel 'flicker ' between open and closed states (Fig.  b, d ). At  kHz in symmetrical  m KCl the singlechannel conductance of I Ks channels was "  pS (corresponding to " n pA at  mV) as judged by noise-variance analysis ; this was -fold greater than the conductance estimated for homomeric KvLQT channels (Fig.  a, c ; Table ). These results agree with those of Yang and Sigworth (). Moreover, Table 

. Unitary conductance and gating parameters for KvLQT and MinK\KvLQT channels
npn npn p npn npn npn Studies in excised patches from Xenopus laevis oocytes in symmetrical  m KCl, adapted from (Sesti and Goldstein,  b). Unitary conductance was determined by sampling at  kHz and filtering at  kHz at test voltages of - mV in steps of  mV using non-stationary noise analysis. Activation and deactivation kinetics were fit by a single exponential function : I ! jI " e (−t/ τ ) . Half maximal activation potentials (V " # ) and equivalent valences (z ! ) were determined by a fit to the Boltzmann function : \ojexp[ez(V " # kV )\kT]q where e, k and T have their usual meanings. Current rectification was estimated comparing currents at  and k mV. Conductance values for KvLQT and MinK\KvLQT correspond to unitary currents at  mV and  kHz of n and n pA, respectively. considering species and methodological differences, there was reasonable agreement between the conductance of native I Ks channels studied in guinea pig cardiac myocytes (Walsh et al. ) or stria vascularis (Shen & Marcus, ) and this estimate for human I Ks channels as they were "  % smaller and "  % larger, respectively. As expected for studies involving this exasperating subunit, another report had previously asserted the unitary conductance of MinK\KvLQT channels to be n pA, -fold smaller than that estimated for KvLQT channels (Romey et al. ). Perhaps, the rapid flicker of the channels explains why it remains so difficult to observe single channels or measure conductances accurately (Fig.  b).

.. Other differences between KvLQT and MinK\KvLQT channels
Like native I Ks channels, human MinK\KvLQT channels in membrane patches activated and deactivated more slowly than KvLQT channels and were less sensitive to voltage (Fig.  a)

. MinK assembles with HERG, another P loop subunit, to regulate channel activity
After I Ks channels were shown to form by assembly of MinK and KvLQT subunits, an unexpected regulatory role for MinK was detected in another channel complex. MinK-directed antisense nucleotides applied to AT- cells were found to suppress a cardiac current prominent in that cell line : I Kr (Yang et al. ). This suggested that MinK might also interact with HERG, the P\TMD subunit that forms the I Kr channel pore. Indeed, MinK was found to form stable assemblies with HERG ; however, unlike its role in I Ks channels, MinK served to modify HERG channel activity without significantly altering its electrophysiological attributes (McDonald et al. ). Thus, co-expression of MinK and HERG doubled the I Kr current density but had little effect on activation and deactivation kinetics or single-channel conductance. Neither did increased currents result from changes in membrane area or the amount of HERG protein on the cell surface. MinK appeared to alter the fraction of HERG channels in the plasma membrane that were active. This type of shift in channel activity has been postulated to underlie up-regulation of I Ks function in oocytes after cAMP treatment (Blumenthal & Kaczmarek,  a). Cyclic AMP has also been shown to regulate the proportion of functional nicotinic acetylcholine receptor channels in chick ciliary ganglion cells (Margiotta et al. ).

. MinK does not form chloride-selective ion channels
When Xenopus oocytes were injected with large amounts of cRNA for MinK (an amount equal to the native mRNA content in an oocyte, "  ng) it was possible to observe a hyperpolarization-activated Cl − current in addition to K + currents (Attali et al. ). While this suggested MinK might induce or contribute directly to forming a Cl − channel, two groups revealed this to be a non-specific effect. Tzounopoulos and co-workers () showed that heterologous expression of five different membrane proteins (but not a soluble protein) at high levels upregulated a Cl − channel endogenous to oocytes. Similarly, expression of three other non-MinK integral membrane proteins had the same effect while a fourth was not an activator (Shimbo et al. ).

 MinK mutation alters basic channel attributes and identifies key residues
Site-directed mutation has been used to identify those channel functions influenced by MinK and residues in MinK important for specific channel activities. Roles for MinK have been revealed in I Ks channel gating, ion selectivity, single-channel conductance, regulation by second messengers, and sensitivity to small molecule activators and inhibitors (such as, Class III antiarrhythmic agents).
A minimal MinK. Takumi and colleagues () expressed rat MinK mutants in oocytes to delineate the minimal peptide required for function. While residues - could be deleted without noticeable effect (removing the two glycosylation sites), the peptide did not function when residues - or - were deleted (Fig.   d ). Whereas truncation of the C-terminus to yield a  residue molecule was tolerated, the next three residues were required for function. Thus, a minimal MinK peptide of  residues was produced that included residues - and - ; this maintained the putative TMD (residues A-I).
Gating. To identify residues involved in gating, Takumi and colleagues () produced  mutants of rat MinK between residues - (Fig.  d ) ; all mutants were expressed at the plasma membrane and some produced moderate changes in channel activity (TV, IL, RQ, LI, EQ, EQ) while others produced severe effects (SA, KQ, HQ, DN) ; the authors attributed activity changes to altered channel activation. KCNE mutations associated with inherited arrhythmia have also been shown to modify channel gating, as discussed below This was manifested as a and -fold increase in the relative permeability of Cs + and NH % + , respectively, through channels formed with rat MinK mutated at residues  and  (in the midst of the putative TMD) (Goldstein & Miller, ) and increased permeation by Na + ions when MinK was altered at position  (Tai & Goldstein, ).
Single-channel conductance. Point mutations in KCNE associated with inherited cardiac arrhythmia change unitary conductance of I Ks channels, as discussed below (Sesti & Goldstein,  a).
Pharmacology. Mutations at a number of sites altered blockade by external TEA (rat MinK residues , , , , ) (Goldstein & Miller,  ; Wang et al.  a). As in other K + channels, inhibition by TEA was argued to be via a pore occlusion mechanism based on its voltage-dependence and the ability of ions on the inside of the membrane to diminish block external TEA in direct relationship to their position in the relative permeability series (Wang et al.  a) ; this was consistent with the notion that the ions traversed the channel to destabilize TEA on its external blocking site.
Mutating other residues in this region to cysteine (rat MinK residues , , ) produced channels susceptible to external pore blockade by the negativelycharged ethylsulphonate (ES) derivative of methanethiosulphonate (MTS), a sulphydryl-reactive molecule (Wang et al.  a). Further along the linear sequence cysteine substitution created pore blocking sites for external Cd# + (rat MinK residues , ) and internal Zn# + (residues , ). These findings supported the idea that MinK was in intimate association with the I Ks channel pore (more below). Investigators have also exploited natural species variations in MinK to identify residues that influence pharmacology. Differences were noted for external La$ + blockade such that  µ profoundly inhibited rat MinK while leaving the human isolate unaffected (Hice et al. ).
Regulation. While activators of PKC inhibited the current induced by expression of wild-type rat MinK in oocytes, inhibition was not seen in channels formed with SA MinK ; this suggested that inhibition of I Ks resulted from direct phosphorylation at this site (Busch et al.  b). Moreover, MinK-induced currents were increased by exposure to a Ca# + ionophore (A) or by intracellular injection of inositol ,,-trisphosphate (IP), two manipulations expected to increase the intracellular Ca# + concentration ; consistent with this idea, I Ks currents were decreased by microinjection of the Ca# + chelator BAPTA (Busch et al.  a).

.. MinK is a Type I transmembrane peptide
The topology proposed for rat MinK (external amino-terminus, one TMD, cytoplasmic carboxy-terminus, Fig.  d) is supported by the following observations. Takumi and colleagues () showed that both glycosylation sites (residues N and N) carry carbohydrate when the peptide emerges onto the plasma membrane indicating these sites are extracellular. Moreover, surface exposure of the amino-terminus was confirmed by binding of anti-epitope monoclonal antibodies to rat MinK variants with antigenic inserts between residues  and  . Two adjacent residues act as if separated by a portion of the pore that determines selectivity against transmembrane movement of Na + , Cd# + , and Zn# + ions (Fig. ). Thus, I Ks channels containing GC rat MinK are sensitive to block by external Cd# + but are not inhibited if the metal is applied from the cytosol ; whereas, channels formed with FC rat MinK are blocked only by exposure to internal Cd# + or Zn# + (Fig.  a, b) (Tai & Goldstein, ). Features of block by Cd# + at these sites argued strongly for a pore-blocking mechanism. Specifically, Cd# + inhibition was sensitive to transmembrane voltage, the presence of permeant cations on the opposite side of the pore (a trans-ion effect), or concurrent application of TEA, a reagent previously shown to block in the I Ks channel pore (Tai & Goldstein, ). Consistent with the idea that Gly- in the transmembrane stretch of rat MinK is in close proximity to the ion selectivity filter, channels with a cysteine at this site showed altered selectivity for Cs + and NH % + ions (Goldstein & Miller, ) and were sufficiently changed so that permeation of Na + ions could be measured (Tai & Goldstein, ). Two other MinK sites nearby in the linear sequence also appeared pore-associated based on their interaction with Cd# + or Zn# + ions. Thus, when rat MinK position  carried a cysteine, channels were susceptible to block by external but not internal Cd# + (Tai & Goldstein, ). Conversely, TC rat MinK channels were sensitive to internal but not external application of Cd# + or Zn# + (Fig.  b) (Tai & Goldstein, ) ; mutation of this site was previously associated with altered ionic discrimination of the channel (Goldstein & Miller, ).
The I Ks conduction pore has been suggested to widen in either direction away from the segments that mediate block by external and internal Cd# + (rat positions ,  and , , respectively) (Figs.  d,  b)

.. The number of MinK subunits in I Ks channel complexes
Three studies have assessed subunit stoichiometry using a mixing strategy in which wild-type and mutant MinK subunits with distinct properties were coexpressed. In this approach, a binomial distribution is assumed to apply to channel subunit association if the two subunits are processed equivalently and independently ; the analysis is simplified if the measured effect can be shown to depend on a single altered subunit in the complex, as for some toxin-channel studies (MacKinnon, ). Wang and Goldstein () mixed wild-type and DN rat MinK cRNAs in varying proportions. Antibody-based detection demonstrated that the mutation did not alter surface expression and that coexpression did not affect the subunits differently. While the DN mutant reached the membrane, channels containing the subunit passed no current and one mutant subunit appeared sufficient to ablate activity. Increasing the fraction of mutant subunits led to decreased current in a fashion most consistent with two MinK subunits per channel complex ; this stoichiometry was judged  times more likely than one DN subunit per complex and  times more likely than three. That channels were formed with the P loop subunit endogenous to oocytes rather than human KvLQT was thought unlikely to affect the estimate.
In a second study, Tzounopoulos and co-workers () made similar assumptions regarding subunit expression and mixing using wild-type and SA rat MinK. This mutation shifts the current-voltage relationship so that little current is observed at potentials more negative than  mV. Increasing the proportion of mutant in this study offered an estimate of at least  MinK subunits per channel complex (Tzounopoulos et al. ). However, neither surface expression levels nor the dominance of the mutant phenotype was determined.
After the heteromeric nature of I Ks channels was revealed, a third mixing study used wild-type and DN human MinK subunits and human KvLQT (Sesti & Goldstein,  b). There, the smaller unitary conductance of the channels formed with DN human MinK ( pS vs.  pS for wild type) was used to estimate stoichiometry. Assuming that one DN subunit per channel was sufficient to fully reduce the unitary conductance to  pS (as appeared to be the case), an upper limit of n MinK subunits per I Ks channel complex was calculated. It seems prudent to conclude that I Ks channels contain four KvLQT subunits (as yet unproven) and at least two but not more than four MinK subunits. I Ks channels formed with LQTS-associated MinK subunits were found to pass less current due to changes in voltage-dependent gating and unitary conductance (Splawski et al.  ; Sesti & Goldstein,  b). Thus, Sesti and Goldstein ( b) observed that channels formed with DN human MinK required "  mV greater depolarization to achieve half-maximal activation and deactivated -fold faster than wild-type I Ks channels ; those formed with SL human MinK required an additional "  mV depolarization and deactivated n-fold faster (Fig.  a, Table ). Both mutations lowered unitary channel currents ( Table ) but produced no significant change in relative permeability of the channels to monovalent cations (Sesti & Goldstein,  b). This suggested that mutations altered single-channel current at sites distinct from the ion selectivity apparatus. Patients carrying these mutant genes are therefore expected to have decreased K + flux through I Ks channels due to fewer channel openings, lowered single-channel conductance and speeded channel closings.

. Summary of MinK sites critical to I Ks channel function
Comparison of I Ks and KvLQT channels indicates that MinK establishes a number of fundamental channel attributes. Site-directed mutations and inherited mutations that cause human disease have identified a number of residues important for MinK function. These sites (enumerated by their rat MinK position, Fig.  d )  MinK was thought to be unique as no similar peptides (or molecules subserving a similar function) had been identified since its cloning in  (Takumi et al. ). However, three new genes encoding peptides related to MinK were recently isolated (Abbott et al. ). The new genes were found by searching for KCNE -related sequences in databases available through the National Center for Biotechnology Information. The search strategy targeted MinK sites shown to influence I Ks channel function and those physically exposed in the I Ks channel conduction pathway (as listed in Section .). Fragments of genes were identified on nine expressed sequence tags (ESTs) and full-length genes cloned. Sequences of the genes and their predicted protein products establish four KCNE subfamilies (Fig.  a, b). In Sections  and  of this review the attributes of KCNE , encoding MinK, and KCNE , encoding MiRP are compared.
The EST fragments for rat and human KCNE detected an abundant single message in cardiac and skeletal muscle by Northern Blot analysis and rt-PCR (Abbott et al. ). Using these EST sequences, multiple full-length MiRP clones were isolated from rat and human cardiac muscle cDNA libraries. Both rat and human KCNE cDNAs contain important consensus sequences near their predicted translation start sites. In-frame termination codons without intervening ATGs are found in the h upstream sequences of both (positions k and k for rat and human, respectively). Each has an A in the k position relative to the predicted initiator methionine. Open reading frames of  bp forecast that both proteins contain  amino acids (Fig.  b).
The proteins are predicted to have a single transmembrane segment (I-V), two N-linked glycosylation sites (N, N) and two consensus sites for protein kinase C-mediated phosphorylation (T, S) ; neither shows evidence for a cleaved leader sequence. Alignment of rat and human MiRP proteins shows they are  % identical and  % homologous. Rat isolates of MiRP and MinK show  % amino acid identity and  % homology if optimally aligned. Comparing the predicted peptides encoded by the new genes and MinK reveals  homologous residues (Fig.  b) that cluster in the transmembrane and membrane-following regions ;  of these sites were those used to identify KCNE gene fragments in the EST database based on their function (Abbott et al. ).
The human KCNE genomic clone is localized to q. (acc. no. AP) as is KCNE (acc. no. AP) (Abbott et al. ). The two genes are arrayed in opposite orientation, separated by  kb, and have open reading frames that share  % identity and are contained within a single exon. This suggests KCNE and KCNE are related by gene duplication and divergent evolution.

. MiRP assembles with a P loop protein, HERG, to form K + channels with unique function
To test whether rat MiRP (rMiRP) functioned as an ion channel subunit, rMiRP cRNA was injected into Xenopus laevis oocytes ; MinK induces K + currents under these conditions by its association with a pore-forming subunit Genetic and physiologic studies indicated that HERG (the human ether-a-gogo-related gene) encoded the pore-forming P loop subunit of cardiac I Kr channels and that its inheritance in mutant form was associated with long QT syndrome (Curran et al.  ; Sanguinetti et al. ). While channels formed of HERG subunits were similar in function to native I Kr channels they differed in their gating, single-channel conductance, regulation by external K + and sensitivity to antiarrhythmic medications (   Slope conductances were determined from single-channel current-voltage relationships for cell-attached patches in Xenopus laevis oocytes with  m KCl and n m Ca# + solution in the pipette (adapted from Abbott et al. ). Activation and deactivation kinetics were estimated in whole CHO cells in n m Ca# + ,  m KCl,  m NaCl  m HEPES, pH n bath solution and  m KCl,  m MgCl # ,  m EGTA,  m HEPES pH n in the pipette. Currents were measured as in Fig.  and fitted for activation parameters according to the Boltzmann function : is half maximal voltage and V the slope factor. The voltage-dependence for KvLQT channels is customarily reported as equivalent valence, z ! ( Table ) ; for comparison, z ! "  corresponds to V s "  mV. Deactivation was studied at k mV as in Fig.  and current relaxation fit with a double exponential function : (I ! jI f e (−t/ τ f ) j I s e (−t/ τ s ) ).

.. MiRP alters activation, deactivation and single-channel conductance
HERG channels open when depolarized to voltages that favour outward K + ) (a) Raw current traces using a protocol (inset) to assess steady-state activation ;  s voltage pulses from a resting potential of k to j mV in steps of  mV followed by a  s test pulse to k mV. Interpulse interval  s. Scale bars represent  µA and  s. (b) Tail currents from panel a measured at arrow are plotted (meanp... for groups of  oocytes normalized to j mV). Lines were fitted to the Boltzmann function : \ojexp[(V " # kV )\V s ]q where V " # is half maximal voltage and V s the slope factor. V " # was kp and kp mV and V s npn and np. for HERG and rMiRPjHERG channels, respectively. (b, inset) Activation rates for groups of three oocytes normalized to the rate at  mV using incremental prepulse durations from n to  s and voltages of  to j mV in  mV steps followed by test pulses to k mV. (c) Raw current traces using a protocol (inset) to assess deactivation ;  s voltage pulses from a resting potential of k mV to j mV, followed by  s test pulses to a range of voltages between k mV and j mV in steps of  mV. Scale bars represent  µA and  s. (d ) Deactivation rates at various voltages measured from (c), current relaxation was fit with a single exponential ( l Aekt\τ) for groups of eight oocytes. Under these ionic conditions, deactivation at k mV for HERG channels showed a τ l npn s whereas τ l npn s for rMiRP\HERG channels.
currents. They are, however, described as inwardly rectifying because net flux through the channels is inward over a depolarization-hyperpolarization cycle under symmetrical ionic conditions. Inward rectification in HERG channels results from rapid channel inactivation (Sanguinetti et al.  ; Trudeau et al.  ; Smith et al.  ; Wang et al.  a ; Zou et al. ). As seen in recordings performed in  m KCl, HERG channels activate from a closed to an open state upon depolarization but pass little outward current because they rapidly move to an inactive conformation (Fig.  a). Upon repolarization, channels rapidly recover from this inactive state to the open state and pass K + current until they deactivate to the closed state. Because deactivation is slow compared to recovery from inactivation, the time spent in the open state at negative potentials (and, so, the magnitude of current) can be significant.
MiRP alters the activation and deactivation of channels formed with HERG subunits (Figs. ,  ; Table ). When subunits were expressed in Xenopus laevis oocytes and the fraction of HERG and rMiRP\HERG channels leaving the closed state at equilibrium after depolarization compared, those containing rMiRP required a p mV greater depolarization to achieve half maximal activation (V " # ) with no change in slope factor (Fig.  a, b) ; this shift appeared to result from a slower rate of activation in channels formed with rMiRP ( Fig.  b, inset). Conversely, rMiRP increased channel deactivation rates markedly (Fig.  c). HERG channels did not deactivate appreciably until k mV and required a step that was  mV more negative to achieve the same rate of deactivation as channels formed with rMiRP (Fig.  d ). This increase in deactivation rate was also apparent at the single channel level : HERG channels remained open for many seconds in patches held at k mV while rMiRP\HERG channels closed rapidly (Fig.  a) Similar effects on gating were observed when wild type human MiRP (hMiRP) and HERG subunits were expressed in Chinese Hamster Ovary (CHO) cells using a bath solution with plasma-like ionic constituents ( m KCl and  m free Ca# + ) (Table ) (Abbott et al. ). Like channels formed with rMiRP, hMiRP\HERG complexes required depolarization to more positive potentials to achieve half-maximal activation and showed no change in slope factor compared to channels formed by HERG subunits alone. Neither hMiRP nor rMiRP altered steady-state inactivation. Like channels with rMiRP, hMiRP\HERG complexes deactivated faster than HERG channels, " -fold at k mV ( Table ). The effects of hMiRP on deactivation supported the idea that MiRP contributes to native I Kr channels as channels formed only by HERG subunits (or its murine homologue MERG) deactivated to -fold slower than I Kr channels recorded in human or murine ventricular myocytes ( Peak macroscopic currents generated in oocytes by co-injection of rMiRP and HERG cRNAs were  % lower than those generated using the same quantity of HERG cRNA alone. Single-channel analysis indicated that this was directly attributable to a  % reduction in unitary conductance (Fig.  b, c). Single HERG channels have a slope conductance of np pS (in symmetrical  m KCl solution at voltages positive to  mV) as compared to p pS for rMiRP\HERG channels and npn pS for hMiRP\HERG channels (Abbott et al. ). This also supported a role for MiRP in native channels since recordings with rabbit atrioventricular node cells under identical conditions exhibited a unitary conductance for I Kr channels of n pS (Shibasaki, ).

.. MiRP alters regulation by K + ion and confers biphasic kinetics to channel blockade
Another feature of native I Kr and HERG channels is their unique sensitivity to external K + ions. HERG channels are activated by external K + ions so that outward K + currents increase linearly with elevations in external K + concentration despite decreasing outward driving force (Sanguinetti & Jurkiewicz,  ; Sanguinetti et al. ). While MiRP\HERG channel complexes also display this K + -dependent up-regulation they are " -fold less sensitive to external K + than HERG channels (Abbott et al. ). Of note, native I Kr channels in murine atrial cells and guinea pig ventricular myocytes are also less sensitive than cloned HERG channels to regulation by bath K + levels ( conversely, native I Kr channels show significant inhibition with an initial test pulse and relax readily to equilibrium block with subsequent test pulses (Carmeliet,  , ). hMiRP\HERG channels expressed in CHO cells are blocked by E- with an apparent equilibrium constant of npn n, a value similar to that reported for native cardiac I Kr channels (Liu et al.  a) and " -fold lower than for HERG channels (Abbott et al. ) ; More striking, however, is that mixed complexes with hMiRP reproduce the distinctive biphasic kinetics that characterize blockade of native I Kr channels by E-. While HERG channels reached equilibrium slowly and only with repetitive test pulses (Fig.  a, c) channels formed with hMiRP were significantly inhibited on the first pulse and relaxed to equilibrium blockade readily (Fig.  b, c). For HERG channels, relaxation was best-approximated by a single exponential decay with a time constant (τ) of p pulse cycles (n l  cells) whereas block of channels with hMiRP was best-described as an initial fast block followed by a single exponential decay with τ l p pulse cycles (n l  cells). Thus, in contrast to HERG channels, mixed complexes reproduced the biphasic blocking kinetics observed with native I Kr channels.

.. Stable association of MiRP and HERG subunits
Subunit interaction between MiRP and HERG was evaluated by studying the two subunits modified with epitope tags and expressed in mammalian tissue culture cells (Abbott et al. ). Transient expression of MiRP with a nine-residue HA epitope (MiRP-HA) in COS cells, followed by Western blot analysis with anti-HA antibody, revealed three specific bands at migration distances appropriate for the mature protein and small amounts of its mono-and unglycosylated forms ; endoglycosidase F treatment supported this interpretation of the bands. Coexpression of MiRP-HA with HERG bearing a  residue cmyc epitope (HERG-cmyc) allowed recovery of rMiRP-HA by immunoprecipitation (IP) with an anti-cmyc monoclonal antibody. Recovery was shown to be specific because anti-cmyc IP gave no signal when HERG-cmyc was expressed alone, when rMiRP-HA was expressed alone, or when the channel protein connexin cmyc was expressed with rMiRP-HA (Abbott et al. ).  Table ) with voltage steps from k to  mV for  s followed by a pulse to k mV for  s using a n s inter-pulse interval. Cells were studied for four pulse cycles prior to drug application then held at k for  min in the presence of  µ E- (bar) followed by - cycles in the continued presence of the drug. (a) The first  traces are shown for a cell expressing HERG channels ; (b) the first  traces for a cell expressing hMiRP\HERG channels ; (c) relaxation to equilibrium blockade for cells in panels (a) and (b) ; HERG channels (, τ l  cycles) and hMiRP\HERG channels ( , τ l  cycles).  Table ). (a) Raw current traces elicited by a  s pulse from k to  mV in steps of  mV followed by a  s step to k mV with a  s inter-pulse interval for WT, QE or MT-hMiRP and HERG ; scale bars represent  pA for WT,  pA for MT and QE-hMiRP, and n s. (b) Tail currents elicited by depolarizing to  mV (not shown) and then repolarizing to voltages from k mV to k mV ; scale bars represent  pA for WT,  pA for MT and  pA for QE hMiRP, and n s. (c) Activation assessed by isochronal Po curves for WT ( ), QE ($) or MT-hMiRP (X) ; curves for groups of - cells and were fitted as in Fig.  b. (d ) Deactivation (fast component) for MinK and HERG-cmyc also co-assemble (McDonald et al. ). While MinK does not alter the biophysical attributes of HERG, its does regulate the number of functional channel complexes on the membrane. To compare the binding of MinK and MiRP to HERG-cmyc, an assay was performed using $&S-labelled MinK and MiRP subunits synthesized in vitro. Incubation of rMiRP and HERG-cmyc followed by anti-cmyc IP allowed strong recovery of rMiRP, as judged by autoradiography (Abbott et al. ). Similarly, incubation of rMinK and HERG-cmyc allowed strong recovery of rMinK. When rMiRP and rMinK were mixed in a  : ratio and incubated at -fold molar excess with HERG-cmyc, anti-cmyc IP led to strong recovery of rMiRP, like that seen in the absence of rMinK, while recovery of rMinK was poor. Thus, rMinK and rMiRP could each assemble with HERG-cmyc. However, in vitro, the presence of both peptides favored formation of stable rMiRP\HERG complexes in preference to those with rMinK.

. KCNE mutations are associated with arrhythmia and decreased K + flux
Four mutations in human KCNE have been associated with inherited or acquired cardiac arrhythmia. In a first study (Abbott et al. ), a panel of  patients with drug-induced arrhythmia and  patients with inherited or sporadic arrhythmias and no mutations in the known arrhythmia genes KvLQT , HERG, SCNA or KCNE were screened. A control population of  individuals was also evaluated. Analysis by SSCP and DNA sequencing revealed three abnormalities (QE, MT and IT hMiRP) and a polymorphism (TA hMiRP). In another study (Sesti et al.  ; Wei et al. ), a panel of  patients with drug-induced arrhythmia and no mutations in known arrhythmia genes were screened. This analysis identified the three abnormalities and one polymorphism described by Abbott et al. () and a mutation associated with quinidine-induced arrhythmia, AV hMiRP. The functional effects of one mutation associated with inherited disease and another associated with acquired arrhythmia are considered here.
Q E hMiRP . Of  patients with drug-induced arrhythmia in one study, one had a C to G transversion producing a Q to E substitution in the putative extracellular domain of hMiRP (Abbott et al. ) ; the mutation was not found in  control individuals. The patient is a -year-old African-American channels formed with WT ( ), QE ($) or MT-hMiRP (X) ; values for fast and slow rates and their weights were estimated by fitting raw current traces to a double exponential function ( Table ). (e) Blockade by clarithromycin ; raw current traces of channels formed with QE hMiRP and HERG subunits in the absence (control) and presence of n m clarithromycin (jclarithro) ; scale bars, n pA and n s. The plot shows the variation of peak tail current amplitude at equilibrium with varying doses of clarithromycin after activation at j mV ; half maximal blocking concentrations and Hill coefficients were npn m and npn and npn m and npn for WT () and QE hMiRP\HERG channels ($), respectively. female. Baseline electrocardiograms showed QT intervals that were borderline prolonged (QTc l  ms). Admitted to the hospital with pneumonia, she was treated with intravenous erythromycin ( mg every  h for  days) and then oral clarithromycin ( mg every  h). After two doses of clarithromycin, electrocardiography showed a QTc of  ms. She developed torsades de pointes and ventricular fibrillation requiring defibrillation. At the time, her serum potassium level was n mequiv.\l.
Wild-type hMiRP\HERG channels and those formed with QE hMiRP were compared by expression in CHO cells ( Fig.  ; Table ). Mutant channels were like those formed with wild-type subunits in their steady-state inactivation (not shown) and rate of deactivation (Fig.  b, d ). However, QE hMiRP channels required depolarization to more positive potentials to achieve half-maximal activation and had a diminished slope factor compared to wild type (Fig.  a, c). Moreover, QE hMiRP channels were -fold more sensitive to clarithromycin blockade than wild-type hMiRP\HERG channels with measured equilibrium inhibition constants (K i ) of npn and np. m, respectively (Fig.  e). Consistent with blockade of open channels, inhibition was observed at voltages positive to the threshold for activation and increased as prepulse potential became more positive. However, clarithromycin also caused a j mV shift in V " # (with no change in slope factor) for both wild type and QE-hMiRP channels (Abbott et al. ). At present, clarithromycin inhibition is best described as statedependent.
A suggested mechanism for acquired arrhythmia. The patient with QE hMiRP presented with a prolonged QTc prior to therapy and further QTc prolongation, torsades de pointes and ventricular fibrillation following clarithromycin administration (when her serum K + concentration was below normal). As noted for MinK mutants associated with LQTS above (Section .), decreased K + efflux slows myocardial repolarization and is reflected on the surface electrocardiogram as a prolonged QTc ; this predisposes to torsades de pointes and ventricular fibrillation (Sanguinetti et al. ). Compared to myocytes expressing wild-type channels, those with QE hMiRP are expected to pass less K + for three reasons. First, mutant channels activate less effectively in response to depolarization (Table ), perhaps the basis for an increased QTc at baseline. Secondly, mutant channels are more sensitive to blockade by clarithromycin (Fig.  e). Thirdly, concurrent hypokalaemia diminishes channel activity (Abbott et al. ). Moreover, female gender is an independent risk factor for drug-induced torsades de pointes in humans, possibly due to gender-specific differences in I Kr density, as seen in rabbit ventricular myocytes (Ebert et al. ). Our findings support the idea that acquired arrhythmia in otherwise asymptomatic individuals can result from mutant channel subunits that are well-tolerated under normal circumstances only to be revealed by provocative stimuli -in this case, the initiation of antibiotic therapy. These findings support the idea that a predisposition to arrhythmia can result from cumulative stressors that diminish cardiac repolarization reserve, that is, capacity of the myocardium to repolarize normally (Roden ). Stressors in this case include an inherited mutation that diminishes K + flux at baseline and concomitant therapy with clarithromycin, a known inhibitor of cardiac K + channels. A similar mechanism appears to be responsible for quinidine-induced arrhythmia in a patient with AV-hMiRP (Sesti et al.  ; Wei et al. ). M T hMiRP . One of  patients with inherited or sporadic arrhythmias had a T to C transition causing substitution of M for T in the predicted transmembrane segment ; this was not found in  control individuals (Abbott et al. ). This patient is a -year-old Caucasian female who was in good health and on no medications. This individual had ventricular fibrillation while jogging and her resuscitation required defibrillation. Electrocardiograms showed an atypical response to exercise with QTc intervals ranging from  to  ms and an automatic internal defibrillator was placed.
While mutant channels formed with MT-hMiRP were like wild type in their steady-state inactivation, they showed an increased voltage-dependence of activation due to diminished activation slope factor with no change in V " # ( Fig.   a, c ; Table ). In addition, channels formed with this mutant showed a speeded rate of closing : they deactivated " -fold faster than those with wild type hMiRP and " -fold faster than channels formed by HERG subunits alone (Fig.  b, d, Table ). As with QE hMiRP, increased voltage-dependence of activation results in fewer open channels for a given voltage step ; faster deactivation indicates that if channels formed with MT hMiRP subunits do open they will close more rapidly than wild type. In the heart, both these effects would reduce K + current, prolonging the cardiac action potential (and, so, the QTc interval measured on an electrocardiogram) thereby predisposing the patient to torsades de pointes and ventricular fibrillation. The KCNE superfamily now comprises four branches of putative single TMD peptides that range from  to  amino acids in length (Fig. ). MinK and MiRP subunits, the KCNE and KCNE gene products, share genetic, structural and functional features.

. Genetics and structure
Human KCNE and KCNE genes appear to be related by duplication and divergent evolution. They are both localized to q. where they are separated by just  kb. Both have a single open reading frame and these are  % identical at the nucleotide level. Human MinK and MiRP are predicted to be similar in length ( and  residues, respectively) and both appear to be Type I peptides : external amino-terminus, single TMD and internal carboxy-terminus. The two peptides share  % amino acid identity and  % homology. Both carry two asparagine-linked carbohydrates when expressed in mammalian tissue culture cells. Many similarities in their primary sequences coincide with positions important for I Ks channel function. Thus, both peptides have an FXF sequence in the TMD and a stretch of identical positively charged residues in the membrane-following region that are critical for function (Fig.  b). MinK and MiRP both assemble with pore-forming P\TMD subunits to alter their electrophysiological attributes (KvLQT and HERG, respectively). Assembly appears to occur co-translationally as stable complexes are formed prior to glycosylation (McDonald et al.  ; Abbott et al. ). That MinK and MiRP both bind to HERG subunits suggests similar molecular features are important for complex formation, however, the determinants of subunit interaction and specificity remain to be elucidated. At least two and perhaps four KCNE peptides contribute to a channel complex.

. Cell biology and function
MinK and MiRP perform similar tasks. Both assemble with P\TMD K + channel subunits to form complexes with unique functional characteristics. Found in numerous tissues, both peptides are present in the heart where they produce two essential currents that repolarize the myocardium to terminate the cardiac action potential : MinK\KvLQT channels underly the slowly activating current I Ks while MiRP\HERG complexes produce the rapidly activating current I Kr .
Their significance to normal cardiac rhythm is emphasized by the pathological consequences of inherited mutations in MinK and MiRP. MinK and MiRP determine key attributes of I Ks and I Kr channels, respectively. Compared to the channels formed of the respective P\TMD subunits alone, both peptides slow channel activation kinetics and alter its voltagedependence. While MinK also slows channel deactivation, MiRP speeds this gating transition. Both peptides alter single-channel current magnitude : MinK increases it -fold while MiRP decreases it -fold. Both alter channel pharmacology to determine the sensitivity and kinetics of blockade by Class III anti-arrhythmic agents.
It appears that KCNE peptides may also subserve a primarily regulatory role in some channel complexes. Thus, MinK assembles with HERG in mammalian tissue culture cells to alter the fraction of active channels in the membrane without significant effect on their biophysical function. Two studies support the idea that MinK regulates HERG in vivo.

. ,   , 
MinK and MinK-related peptide  (MiRP) are integral membrane peptides that co-assemble with pore-forming K + channel subunits to establish the gating kinetics, single-channel conductance, ion selectivity, regulation and pharmacology of the complex. Co-assembly is required to reconstitute channel behaviours like those observed in native cells. Thus, MinK\KvLQT and MiRP\HERG complexes reproduce the cardiac currents I Ks and I Kr , respectively. Inherited mutations of MinK and MiRP are associated with lethal cardiac arrhythmias. Studies of MinK and MiRP in wild type and mutant form have answered many questions.
MinK and MiRP are ion channel subunits rather than contributors to carriertype transporters. MinK and MiRP do not function alone but in assemblies with pore-forming subunits. MinK and MiRP are obligatory in some ion channel complexes -thus, mutant variants of the peptides that alter channel behaviour have been associated with cardiac rhythm disturbances. In other channels, the peptides subserve regulatory roles and may be accessory. MinK residues interact directly with ions traversing the I Ks conduction pathway placing these sites close to the core of the channel complex. MinK and MiRP are structural contributors to K + channels and do not form (or specifically regulate) Cl − channels. MinK is an obligatory constituent of cardiac I Ks channels. MiRP is a required component of cardiac I Kr channels. In addition, MinK appears to regulate cardiac I Kr channels. MinK is present two and perhaps four times in a channel complex. MinK increases and MiRP decreases unitary channel currents. While some issues appear to be clarified, others remain unclear.
Determination of the structure of the Streptomyces lividans K + channel KCSA at n A / has offered a first glimpse of a K + -selective pore (Doyle et al. ). Can this type of disciplined array of α-helices possibly accommodate two or four additional membrane-spanning stretches in close proximity to the ion-conduction pathway ? Our current hypothesis is that KCNE peptides intercalate between P loop subunits so that some peptide residues can interact with ions traversing the deep pore while others contribute to the outer and inner channel vestibules. Functional studies suggesting that the MinK transmembrane stretch adopts an αhelical conformation in the external vestibule (Wang et al.  a) and passes the narrow part of the pore in an extended structure (Tai & Goldstein, ) are consistent with studies using infrared and circular dichroism spectroscopy of a peptide corresponding to this region that showed predominant α-helical and minor β-strand structures (Mercer et al. ). This apparent correlation is tantilizing but is at best highly speculative : the conformation of peptides like MinK depends highly upon environment (including any interacting proteins) and the structure that KCNE peptides adopt in situ must be directly determined.
A list of many other questions that now require attack includes : what rules govern KCNE peptide and P loop subunit interaction ? Do obligatory and regulatory interactions between KCNE peptides and P loop subunits differ ? What degree of inter-subunit specificity exists ? What peptide : P loop subunit stoichiometry is predominant in vivo ? How many types of KCNE complexes are employed in native cells ? What factors control expression of KCNE genes ? What are the pore-forming partners for MiRP and MiRP ? It seems likely that answers to these questions can reveal much about how KCNE peptides contribute to the natural variety and authentic demeanor of ion channel complexes in vivo.

. 
This work was supported by grants to S. A. N. G. from the NIH and a Wellcome Trust Traveling Fellowship to G. W. A. S. A. N. G. is grateful to Chris Miller for suggesting that a study of MinK might make for an interesting postdoctoral project. We thank D. Goldstein for thoughtful feedback on the manuscript and F. Sesti for his essential and non-accessory role in many of the original studies considered herein.