Concerted All-or-none Subunit Interactions Mediate Slow Deactivation of Human ether-à-go-go-related Gene K+ Channels*

Background: The N terminus of hERG1 subunits modulates the rate of channel deactivation. Results: A single mutant subunit in a concatenated hERG1 tetramer accelerates channel deactivation. Conclusion: All four subunits cooperate fully to slow the rate of hERG1 channel deactivation. Significance: A single subunit harboring a mutation in the PAS domain or S6 segment can alter the gating properties of a tetrameric hERG1 channel. During the repolarization phase of a cardiac action potential, hERG1 K+ channels rapidly recover from an inactivated state then slowly deactivate to a closed state. The resulting resurgence of outward current terminates the plateau phase and is thus a key regulator of action potential duration of cardiomyocytes. The intracellular N-terminal domain of the hERG1 subunit is required for slow deactivation of the channel as its removal accelerates deactivation 10-fold. Here we investigate the stoichiometry of hERG1 channel deactivation by characterizing the kinetic properties of concatenated tetramers containing a variable number of wild-type and mutant subunits. Three mutations known to accelerate deactivation were investigated, including R56Q and R4A/R5A in the N terminus and F656I in the S6 transmembrane segment. In all cases, a single mutant subunit induced the same rapid deactivation of a concatenated channel as that observed for homotetrameric mutant channels. We conclude that slow deactivation gating of hERG1 channels involves a concerted, fully cooperative interaction between all four wild-type channel subunits.

During the repolarization phase of a cardiac action potential, hERG1 K ؉ channels rapidly recover from an inactivated state then slowly deactivate to a closed state. The resulting resurgence of outward current terminates the plateau phase and is thus a key regulator of action potential duration of cardiomyocytes. The intracellular N-terminal domain of the hERG1 subunit is required for slow deactivation of the channel as its removal accelerates deactivation 10-fold. Here we investigate the stoichiometry of hERG1 channel deactivation by characterizing the kinetic properties of concatenated tetramers containing a variable number of wild-type and mutant subunits. Three mutations known to accelerate deactivation were investigated, including R56Q and R4A/R5A in the N terminus and F656I in the S6 transmembrane segment. In all cases, a single mutant subunit induced the same rapid deactivation of a concatenated channel as that observed for homotetrameric mutant channels. We conclude that slow deactivation gating of hERG1 channels involves a concerted, fully cooperative interaction between all four wildtype channel subunits.
Voltage-gated K ϩ (K V ) channels are activated (opened) by depolarization of the cell membrane. Repolarization induces channel closure via a gating process called deactivation. The rate of deactivation varies considerably for different channel types but is unusually slow in human ether-à-go-go-related gene type 1 hERG1 2 K ϩ (K V 11.1) channels that conduct the rapid delayed rectifier K ϩ current, I Kr (1), in the heart (2,3). In car-diomyocytes, the extremely fast recovery from depolarizationinduced channel inactivation combined with a slow rate of deactivation allows I Kr to rebound during repolarization of the action potential from the plateau phase to the resting membrane potential (4). Reduction of outward K ϩ conductance caused by de novo or inherited loss of function mutations in hERG1 slows the rate of action potential repolarization, prolongs the QT interval on the body surface electrocardiogram and is associated with torsade de pointes arrhythmia and increased risk of sudden cardiac death (4 -6). The cardiac disorder caused by hERG1 mutations is called type 2 long QT syndrome (LQT2). The majority of the LQT2-associated mutations in hERG1 cause protein misfolding and reduced trafficking of the channel to the cell surface (7). Some LQT2-associated mutant channels traffic normally to the cell surface but alter biophysical properties of the channel. For example, some point mutations (e.g. R56Q) that are located in the cytosolic N-terminal region of the hERG1 subunit accelerate the rate of channel deactivation (8 -10).
Similar to other K V channels, hERG1 channels are formed by coassembly of four subunits. In mammals, including humans, the full-length subunit (hERG1a) can coassemble with hERG1b, an alternatively spliced subunit with a shorter N terminus. Western blot analysis indicates that both forms are expressed in human ventricle, and co-immunoprecipitation indicates that canine erg1a and erg1b subunits associate together in the T-tubules of cardiomyocytes (11). Erg1a homotetramers, with a fully intact N-terminal "eag" domain deactivate very slowly, whereas erg1b homotetramers deactivate ϳ10ϫ faster (12)(13)(14). It has been estimated that hERG1b represents about ϳ10 -25% of the total hERG1 mRNA in the human heart (14). Heterotetramers formed by coassembly of hERG1a and hERG1b subunits could either deactivate at a rate dominated by a subunit type or at an intermediate rate that was dependent on the relative number of each type to the fully assembled channel. When Xenopus oocytes were injected with variable amounts of hERG1a and hERG1b cRNA, the relative changes in the kinetics of deactivation of the resulting heterologously expressed channels were a linear function of the relative abundance of injected hERG1b cRNA (15). This finding suggests that full-length and N-terminal-truncated subunits cooperate during channel deactivation, but the nature of cooperative interactions (i.e. sequential versus fully concerted) has not been determined.
The N terminus of hERG1 is composed of 355 residues. Truncation of the entire N terminus (16 -18) or just the first 26 residues (9) accelerates the rate of deactivation to an extent similar to that observed for hERG1b homotetramers. Fast deactivation of homotetrameric hERG1 channels can also be achieved by mutations that neutralize the charge of just two basic residues, Arg-4 and Arg-5 located in the PAS-cap region (19). The initial 135 residues of the N terminus form the etherà-go-go (eag) domain, present in all members of the eag family of K ϩ channels. Residues 26 -135 form a PAS (Per-Arnt-Sim) domain (9), and residues 1-26 form the PAS-cap. The structure of the hERG1 PAS domain was solved years ago and was proposed to have an important regulatory function (9). More recent studies have revealed that the PAS domain interacts with the cytoplasmic C-terminal domain in hERG1 channels (10, 20 -22). Coexpression of eag domains together with N-terminal truncated channels can restore normal slow deactivation (22). With the exception of erg2, the other members of the eag K ϩ channel superfamily (K V 10-K V 12), including eag1 (23), eag2 (24), erg3 (25), eag-like (elk)1 (26), and elk2 (27) deactivate rapidly compared with mammalian erg1a. Here we ask how many wild-type (WT) N termini are required for the homotypic hERG1a channel to deactivate with its characteristic slow kinetics.

EXPERIMENTAL PROCEDURES
Construction of hERG1 Concatemers-WT and mutant forms of KCNH2 (HERG1 isoform 1a, NCBI reference sequence NM_000238) cDNAs were cloned into the pSP64 poly(A) oocyte expression vector (Promega Corp.). Concatenated tetramers were engineered to contain a variable combination of WT and/or mutant KCNH2 cDNAs with defined positioning as previously described (28). Two single mutations (R56Q, F656I) and one double mutation (R4A/R5A) known to accelerate deactivation were studied. The R4A/R5A double mutation is located in the PAS-cap region, and R56Q is located in the PAS domain. The mutation F656I is located in a region of the S6 segment that contributes to formation of the "S6 bundle crossing," the intracellular activation gate (29). In this study, channels formed from individual subunits expressed and assembled naturally in oocytes are called X monomer , where X is either a WT or mutant subunit. Homotypic concatenated tetramers formed by four WT or mutant (e.g. R56Q) subunits are designated WT 4 channels and R56Q 4 , respectively. Heterotypic concatenated tetramers indicate the relative positioning of the single WT and mutant subunits. For example, the R56Q 1 /WT 1 /R56Q 1 /WT 1 channel was engineered to contain WT subunits in the second and fourth positions together with subunits harboring the R56Q mutation in the first and third positions of the tetramer. Mutations were introduced into KCNH2 using the QuikChange site-directed mutagenesis kit (Agilent Technologies). Construction of dimers and fully concatenated tetramers was the same as previously described (28). Three types (1-3) of monomer constructs were used to make dimers. For all mono-mer constricts, a HindIII site was added just before the start codon. A KpnI site was added at the 3Ј end just before the stop codon in Type 2 monomers or 3Ј to the HindIII site in Type 3 monomers. Type 1 and 3 monomers were cut with HindIII then ligated together to form a Type I dimer (having a single KpnI site). Type 2 and 3 monomers were cut with HindIII then ligated together to form a Type II dimer (having two KpnI sites). Dimers were sequenced with forward and reverse primers based on vector sequences. Type I and II dimers were cut with KpnI then ligated together to create a tetrameric concatenated channel construct. Tetrameric constructs were verified by DNA sequence analyses using forward and reverse primers based on vector sequences, and the correct size of inserts was checked by digest with KpnI. Tetrameric KCNH2 plasmids were linearized with EcoR1 before in vitro transcription using the mMessage mMachine SP6 kit (Ambion).
Oocyte Isolation and cRNA Injection-Procedures used for harvesting oocytes from Xenopus laevis were approved by the University of Utah Institutional Animal Care and Use Committee. Ovarian lobes were surgically removed from adult female X. laevis anesthetized by immersion in 0.2% tricaine solution. Ovarian lobes were dissected into clumps of 10 -20 oocytes, which were then digested with 1 mg/ml each of type I and II collagenase (Worthington Biochemical Corp.) in Ca 2ϩ -free ND96 solution to remove the follicle cell layer. Ca 2ϩ -free ND96 contained 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES; pH was adjusted to 7.6 with NaOH. Stage IV and V oocytes were injected with 46 nl (0.5-2 ng) of cRNA of either WT or mutant hERG1 constructs. Oocytes were incubated in Barth's solution for 1-7 days at 17°C. Barth's contained 88 mM NaCl, 1 mM KCl, 0.41 mM CaCl 2 , 0.33 mM Ca(NO 3 ) 2 , 1 mM MgSO 4 , 2.4 mM NaHCO 3 , 10 mM HEPES and 1 mM pyruvate plus gentamycin (50 mg/liter) and ciprofloxacin (50 mg/liter); pH was adjusted to 7.4 with NaOH.
Electrophysiology-Whole cell currents were recorded from oocytes by using standard two-electrode voltage clamp techniques (30,31). Oocytes were bathed in ND96 extracellular solution that contained 96 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES; pH was adjusted to 7.6 with NaOH. Agarose cushion microelectrodes were prepared as described (32) using thin walled borosilicate glass (TW100F-4, World Precision Instruments, Inc.) and had a resistance of 0.5-1.5 megaohms when filled with 3 M KCl. A personal computer, GeneClamp 500 amplifier, Digidata 1322A, and pCLAMP 8.2 software (Molecular Devices) were used to acquire and digitize data.
The voltage dependence of current activation was determined by using a current-voltage (I-V) pulse protocol. From a holding potential of Ϫ90 mV, 5-s depolarizing pulses were applied to a test voltage (V t ) that ranged from Ϫ90 to ϩ60 mV. After each test pulse, the membrane potential was returned to Ϫ70 mV to elicit a tail current, I tail . The tail current amplitude (A tail ) for each V t was measured as the difference between the peak outward I tail and the current at Ϫ70 mV elicited during a short prepulse applied immediately before each test pulse. The voltage dependence of hERG1 channel activation was obtained by plotting A tail measured after each 5-s test pulse as a function of V t . Kinetics of current deactivation were evaluated using a two-step voltage pulse protocol. From a holding potential of Ϫ80 mV, a 1-s pulse to ϩ40 mV was used to activate channels and was followed by a repolarizing pulse to a variable return potential (V ret ) that ranged from ϩ40 to Ϫ140 mV and was applied in 10-mV increments.
Data Analysis-Voltage clamp data were off-line-analyzed with either pClamp 8.2 or 10.2 (Molecular Devices) and Prism 6 (Graph Pad) software packages. To estimate the voltage dependence of hERG1 channel activation, A tail was normalized to its maximum fitted value for each oocyte, plotted as a function of V t , and the resulting conductance-voltage (G-V) relationship was fitted with a Boltzmann function, where G/G max is relative conductance, V 0.5 is the half-point of the G-V relationship, z is the equivalent charge for activation, F is Faraday's constant, R is the gas constant, and T is the absolute temperature.
To derive the kinetics and amplitudes of the fast and slow components of deactivation, I tail elicited at each V ret was fitted with a double exponential function, where A f and A s are the amplitudes of the fast and slow components of I tail , deact-f , and deact-s are the fast and slow time constants of tail current decay, and C is steady state current at V ret . All data are presented as the mean Ϯ S.E., where n is the number of oocytes. Statistical differences were evaluated with twoway ANOVA and a Tukey multiple comparison test where appropriate. Differences were considered significant for p Յ 0.05.

Concatenated hERG1 Tetramers and Channels Formed Naturally from hERG1 Monomers Have Similar Properties of Activation and
Deactivation-In our previous study of hERG1 channel activators, we reported similar biophysical properties for WT monomer channels and concatenated WT hERG1 (WT 4 ) channels (28). These findings are confirmed for the present study in Fig. 1. Representative currents for WT 4 and WT monomer hERG1 channels are illustrated in Fig. 1A. Strong rectification for both channel types is indicated by the small size of the outward currents recorded at the V t of ϩ40 mV. The rate of tail current decay (deactivation) is very slow at Ϫ40 mV and is progressively increased at more negative levels of V ret . In Fig. 1, B and C, the values of deact-f and deact-s and their relative amplitudes for WT monomer and WT 4 channels are plotted as a function of V ret ranging from Ϫ80 to Ϫ40 mV. There was no significant difference between the channel types for the average values of either deact-f or deact-s . The voltage dependence of activation determined with 5-s depolarizing pulses for WT 4 and WT monomer hERG1 channels are compared in Fig. 1D. There was no significant difference between the V 0.5 values determined for the two channel types. Thus, the activation and deactivation properties of hERG1 channels are not altered by covalent linkage of four WT subunits.

Mutation of Residues in the N-terminal Domain of a Single Subunit Is Sufficient to Accelerate Deactivation of hERG1
Channels-Arg-4 and Arg-5 are located on one face of an amphipathic helix (33) in the PAS-cap of hERG1. Charge neutralization of these two residues by mutation to Ala accelerates channel deactivation by an order of magnitude (19) as shown in Fig. 2A where currents conducted by concatenated tetramers and R4A/R5A monomer channels are compared. To determine if one or more R4A/R5A subunits was required to increase the rate of deactivation, concatenated tetramers containing 1, 2, 3, or 4 mutant subunits were constructed, and their kinetics of deactivation ( Fig. 2A) and voltage dependence of activation (Fig. 2B, Table 1) compared. The presence of one or more R4A/ R5A mutant subunit shifted the voltage dependence of activation by ϩ6 to ϩ10 mV when compared with WT 4 channels. The average time constants for both fast and slow components of deactivation and their relative amplitudes are summarized in Fig. 2, C and D, respectively. Compared with WT 4 channels, the values of deact-f and deact-s were dramatically reduced at all voltages for concatenated channels containing one or more R4A/R5A mutant subunit. For example, at Ϫ60 mV deact-f and deact-s were 32 Ϯ 3 ms and 102 Ϯ 12 ms for (R4A/R5A) 1 /WT 3 channels (n ϭ 11) compared with 417 Ϯ 35 ms and 2367 Ϯ 230 s for WT 4 channels (n ϭ 10). There was no significant difference in deact-f and deact-s between channels containing one or more R4A/R5A subunits (two-way ANOVA). Thus, R4A/R5A acts in a dominant manner to disrupt the slow kinetics of hERG1 channel deactivation. The LQT2-associated point mutation R56Q, located in the PAS domain of hERG1, was previously reported to increase the rate of channel deactivation when expressed in Xenopus oocytes (8) and to increase the rate of activation and deactivation but not inactivation when channels were heterologously expressed in HEK-293 cells (34). To further explore how a mutation in the PAS domain affects the rate of deactivation, R56Q subunits were placed in either position 1 (R56Q 1 /WT 3 ) or position 3 (WT 2 /R56Q 1 /WT 1 ) of a concatenated tetramer, and the current conducted by these channels was compared with R56Q monomer and R56Q 4 channels. The mutant channels exhibited similar deactivation kinetics (Fig. 3A) and voltage dependence of activation (Fig. 3B, Table 1). The average values of deact-f and deact-s and their relative amplitudes at potentials ranging from Ϫ40 to Ϫ80 mV for WT 4 and channels containing R56Q mutant subunits are summarized in Fig. 3, C and D. There was no significant difference (ANOVA) between the time constants at any voltage for R56Q monomer and concatenated channels containing one or four R56Q subunits. Thus, similar to R4A/R5A subunits, the presence of a single R56Q subunit fully disrupted deactivation of hERG1 channels. More importantly, these findings indicate that all four WT subunits are required for hERG1 to deactivate with its characteristic slow kinetics.
A Single F566I Subunit Is Sufficient to Accelerate Deactivation of hERG1 Channels-In the closed state, the S6 segments of K V channels converge near their cytoplasmic ends (the S6 bundle crossing) to form a narrow aperture that prevents entry of intracellular ions into the central cavity (35,36). K V channel opening induced by membrane depolarization is mediated by a final, voltage-independent step that involves fully cooperative interactions between the S6 segments of all four subunits (37)(38)(39)(40). An outward splaying of the cytoplasmic ends of all four S6 segments widens this aperture (opens the pore) to permit K ϩ ion permeation. In hERG1a, Phe-656 is located in the S6 segment near the narrowest region of the pore formed by the bundle crossing. As previously reported (41), mutation of this residue to Ile (F656I) causes fast deactivation of homomeric channels formed from monomers (F656I monomer ; Fig. 4A).  Table 1). C and D, deact-f and deact-s and relative current amplitudes for both components plotted as a function of V ret for the indicated channels. Ϫ18.3 Ϯ 0.7 3.20 Ϯ 0.26 9 FIGURE 3. A single R56Q mutant subunit is sufficient to accelerate rate of hERG1 channel deactivation. A, voltage clamp pulse protocol and currents conducted by R56Q monomer and concatenated tetrameric channels as indicated. B, voltage dependence of activation for indicated channels. Normalized tail currents were plotted as a function of V t and fitted to a Boltzmann function (smooth curves) to estimate V 0.5 and z (summarized in Table 1). C and D, deact-f and deact-s and relative current amplitudes for both components plotted as a function of V ret for the indicated channels.  Table 1). C and D, deact-f and deact-s and relative current amplitudes for both components plotted as a function of V ret for the indicated channels.
To determine if the number or position of F656I subunits in a tetramer differentially alters the rate of channel deactivation, several concatenated tetramers were constructed. Tetramers that were formed from a single F656I subunit placed in either position 1 (F656I 1 /WT 3 ) or position 3 (WT 2 /F656I 1 /WT 1 ), from two mutant and two WT subunits positioned in a diagonal orientation (WT 1 /F656I 1 /WT 1 /F656I 1 ), and from four F656I subunits (F656I 4 ) deactivated with fast kinetics similar to F656I monomer channels (Fig. 4A). F656I monomer channels and all the mutant tetrameric channels, except F656I 1 /WT 3 , were activated at more positive potentials than WT 4 channels (Fig. 4B). There was no significant difference in deact-f and deact-s between channels containing one or more F656I subunit (twoway ANOVA). However, compared with WT 4 channels, the values of deact-f and deact-s were dramatically reduced at all voltages for concatenated channels containing one or more F656I mutant subunits. For example, at Ϫ60 mV deact-f and deact-s were 17 and 14 times faster for F656I 1 /WT 3 channels compared with WT 4 channels. The time constants for the fast and slow components of deactivation measured at a V ret of Ϫ70 mV for WT 4 and all the heterotypic concatenated tetramers analyzed in this study are presented in Fig. 5. At this potential the rate constants for closed to open state transitions are very slow compared with the reverse transition that mediates channel deactivation. For all three mutations examined, a single mutant subunit accelerated deactivation in a dominant manner, confirming the requirement for all four WT subunits to achieve slow deactivation gating.

DISCUSSION
Early studies of hERG1 channel gating established the importance of the N-terminal eag domain. Deletion (17) or truncation (12)(13)(14)42) of the N terminus or specific point mutations of residues within the eag domain (8 -10) were shown to greatly accelerate the rate of channel closing. Based on NMR structures, functional analysis of mutant channels, and disulfide linkage of introduced Cys residues, it was initially proposed that slow deactivation was dependent on an interaction between the eag domain and the S4-S5 linker (42)(43)(44). However, further investigation led to the current view that slow deactivation of hERG1 channels is dependent on an interaction between the N terminus and the C-terminal domain (composed of a C-linker and a cyclic nucleotide binding homology domain (CNBHD)) of adjacent subunits (45). A 2 Å resolution crystal structure of the mouse EAG1 eag domain and the CNBHD complex (46) has provided insight into the likely extensive structural interface of these two domains in the hERG1 channel. Specifically, an amphipathic ␣-helix of the PAS-cap domain lies between a hydrophobic patch of the PAS domain and the ␤4-␤5 strands and ␤8-␤9 loop of the CNBHD. Despite the presence of a CNBHD in the C-terminal domain of hERG1, these channels are not directly modulated by cyclic nucleotides. The C-linker/ CNBHD of a mosquito erg channel was recently solved (47) and provided a structural explanation for this apparent discrepancy. A short ␤-strand (an "intrinsic ligand") occupies the pocket where a cyclic nucleotide is bound in other (e.g. HCN) channels. Mutations in the intrinsic ligand speed the rate of hERG1a channel deactivation (47). Alignment with the homologous HCN2 channel indicates that the C-linker/CNBHD of eag or erg channels forms a concentrically arranged tetramer with the eag domains located at the periphery (46), and the PAS-cap of one subunit probably interacts with the CNBHD of an adjacent subunit (48).
Kinetic analysis of currents was used to distinguish between different models of subunit interactions that mediate inactivation (49). A similar approach can be used to analyze deactivation gating. If one assumes a two-state model for hERG1 channel deactivation (O to C), then deact is equal to the inverse of the sum of the forward and reverse rate constants: 1/(k f ϩ k b ). At sufficiently negative voltages, k b is negligible, and the deactivation rate constant k can be estimated by 1/ deact . If an independent conformational change in any one of the subunits is sufficient to accelerate deactivation, then k would be the sum of the individual rate constants contributed by each of the subunits within a tetramer. If deactivation is a cooperative process FIGURE 5. Slow deactivation of concatenated hERG1 channels requires four WT subunits. A, plot of the time constants for the fast (red bars) and slow (black bars) components of deactivation measured at a V ret of Ϫ70 mV versus the number of wild-type subunits contained within a concatenated tetramer. The position of WT subunits within each concatenated tetramer was as indicated in Fig. 2; the x axis label 3 indicates RA4:R5A mutant subunit was in position 1, whereas 3Ј indicates that the mutant subunit was in position 4. B, same as panel A for tetramers containing R56Q subunits, where 3 indicates mutant subunit in position 1, whereas 3Ј indicates that mutant subunit was in position 3. C, same as panel A for tetramers containing F656I subunits, where 3 indicates that the mutant subunit was in position 1, and 3Ј indicates the mutant subunit was in position 4. For each mutant channel type the deactivation time constants for tetramers containing 0, 1, 2, or 3 WT subunits were not significantly different from one another but were significantly different from WT 4 channels (2-way ANOVA with Tukey's multiple comparisons test).
involving sequential subunit transitions (39), then the deactivation rate of a heterotypic channel would be exponentially related to the sum of the energetic contributions of each of the WT and mutant subunits (49), and log k would be a linear function of the number of mutant subunits contained with a tetramer. Finally, if the rate constant of deactivation (i.e. 1/ deact ) is the same for channels that contained one or more of the mutant subunits, then WT subunits must interact cooperatively in an all-or-none, concerted fashion. Because hERG1 deactivation was biexponential rather than monoexponential, we were unable to use the above analysis that requires a two state model (and a single k). Nonetheless, our finding that there was no difference in either the fast or slow time constants for deactivation of any tetramer containing a mutant subunit greatly simplifies the analysis, as it satisfies the requirements for the extreme case of cooperativity where normal gating (i.e. slow deactivation) requires the concerted interaction of all four WT subunits.
The nature of subunit interactions during activation of K V channels has been extensively studied. Activation is initiated by outward movement of the voltage sensors in each of the four subunits. The final step in the activation pathway involves highly cooperative subunit interactions (37)(38)(39)(40)50). Thus, it is perhaps not surprising that the reverse process, channel deactivation, is also a highly cooperative process involving concerted subunit interactions. However, in most studies where mutant subunits were used to characterize channel activation, findings have been consistent with a sequential model of cooperative subunit interactions; i.e. equal energetic contribution from each subunit to the tetrameric channel function (37,38,40,50). In contrast, the concerted model of cooperativity assumes that a simultaneous conformational change in all four subunits occurs during channel opening (51). The concerted (all-or-none) model predicts that a single mutant subunit will alter activation to the same extent as channels containing 2, 3, or 4 mutant subunits per tetramer. This has been demonstrated by a point mutation (G466P) in the glycine hinge of the Shaker Kv1 channel, where a single mutant subunit is sufficient to induce channels to open at more negative potentials to the same extent achieved by mutation of Gly-466 in all four subunits (39). In hERG1, Phe-656 is also located in the S6 segment but below the Gly hinge and near the upper end of the S6 bundle crossing. We previously reported that mutation of this residue to Ile (F656I) accelerates deactivation (41). In the present study we found that a single F656I subunit was sufficient to accelerate the rate of deactivation to an extent similar to that attained with the F656I homotetramer. Slow deactivation of hERG1 requires the presence of WT subunits in all four positions of the tetrameric channel, similar to the effect of G466P on the voltage dependence of Shaker activation.
The role of subunit interactions during inactivation of K V channels has also been extensively studied. K V channels can inactivate by either an N-type or C-type mechanism. C-type inactivation of K V 1 channels is mediated by a subtle structural rearrangement of the selectivity filter (52)(53)(54). Similar to slow deactivation in WT hERG1a, C-type inactivation of K V 1 channels involves highly cooperative subunit interactions (49,55,56). N-type inactivation is mediated by blocking the channel pore by an N-terminal "ball" domain (57,58) or by the N terminus of an accessory ␤-subunit (59,60). A single N-terminal domain acts independently of others to plug the pore and prevent ion permeation. Thus, opposite to N-terminal-mediated slow deactivation, N-type inactivation is considered a fully independent process. As discussed above, hERG1 channels deactivate fast in the absence of an N terminus, partial truncation of the N terminus, or as a consequence of specific mutations in the PAS-cap or PAS domain. Other mammalian members of the eag K channel family, including eag1, eag2, erg3, elk1, and elk2 deactivate faster than mammalian erg1a. A recent comparison of the biophysical properties of erg channels from different species suggests that slow deactivation may be the ancestral phenotype for these channels (61). Erg1 of the sea anemone Nematostella vectensis deactivates slowly, similar to mammalian erg1a channels. In contrast, erg4 from this sea anemone, erg (unc103) from the nematode Caenorhabditis elegans and Drosophila erg channels lack the N-terminal eag domain, and these channels deactivate rapidly. Together with sequence analysis, these findings suggest that the ancestral eag channel deactivated slowly and that fast deactivation evolved later and independently in the Nematoda and Anthozoa by loss or disruption of the eag domain.
In this study we investigated the role of subunit interactions in the process of slow deactivation of hERG1a channels. Specifically, we sought to determine how many full-length hERG1 subunits are required for channels to deactivate with the slow kinetics characteristic of WT hERG1 homotetramers. To achieve this objective, we constructed concatenated hERG1 tetramers containing a variable number of WT and mutant subunits that contained a mutation known to accelerate the rate of deactivation of homotetramers. A potential limitation of the use of concatenated tetramers is that they can co-assembly as multiple units (62)(63)(64), but the problem is minimized when nearly identical subunits are covalently linked to one another (28,37,63). As discussed before (28), WT 4 and WT monomer channels have very similar biophysical properties, concatenated hERG1 tetramers expressed with a similar efficiency regardless of the number of mutant or WT subunits per tetramer, and with the exception of the voltage dependence of tetramers containing a single F656I subunit, the positioning of a single mutant subunit in a tetramer (first or third position) did not affect channel properties.
Potential interactions between subunits containing either full-length or altered N termini have been studied before by injecting Xenopus oocytes with variable quantities of cRNA encoding hERG1a and hERG1b subunit isoforms (15). In this study the authors found that the -fold changes in both deact-f and deact-s for deactivation at a V ret of Ϫ60 mV were a linear function of the ratio of hERG1a/hERG1b cRNA injected into the oocyte. Assuming that both proteins were expressed with similar efficiency and that subunits co-assembled in a random, non-biased fashion, then this result strongly suggests that hERG1a and hERG1b subunits interact in a cooperative manner to determine the rate of deactivation of heteromeric channels. However, although this approach mimics the physiological method of channel assembly, the resulting mix of heteromultimeric assemblies makes it impossible to quantify the specific effects of a variable number (1)(2)(3)(4) of hERG1b N termini on channel gating, a condition needed to fully characterize the nature of subunit cooperativity. We characterized mutations located in either the N terminus (R4A/R5A or R56Q) or S6 segment (F656I). In each case the presence of a single mutant subunit within a concatenated tetramer was sufficient to disrupt the normal gating process and increase the rate of channel deactivation.
In summary, our findings indicate that the slow rate of deactivation characteristic of a WT hERG1a homomeric channel involves an all-or-none or fully concerted, cooperative interaction between subunits. The PAS-cap of one hERG1a subunit interacts with the CNBHD in the C terminus of an adjacent subunit (48). Evidently, mutations in the N terminus that disrupt only one of the possible four intersubunit interactions is sufficient to revert the kinetics of deactivation to the fast mode characteristic of homomeric channels formed by coassembly of subunits that lack the entire N-terminal domain (16 -18).