Voltage-gated potassium (Kv) channels enable potassium efflux and membrane repolarization in excitable tissues. Many Kv channels undergo a progressive loss of ion conductance in the presence of a prolonged voltage stimulus, termed slow inactivation, but the atomic determinants that regulate the kinetics of this process remain obscure. Using a combination of synthetic amino acid analogs and concatenated channel subunits we establish two H-bonds near the extracellular surface of the channel that endow Kv channels with a mechanism to time the entry into slow inactivation: an intra-subunit H-bond between Asp447 and Trp434 and an inter-subunit H-bond connecting Tyr445 to Thr439. Breaking of either interaction triggers slow inactivation by means of a local disruption in the selectivity filter, while severing the Tyr445–Thr439 H-bond is likely to communicate this conformational change to the adjacent subunit(s).https://doi.org/10.7554/eLife.01289.001
Proteins are made from long chains of smaller molecules, called amino acids. These chains twist and bend into complex three-dimensional shapes, and sometimes two or more chains, or ‘subunits’, are packed into a protein. These shapes are often held together by hydrogen bonds between some of the amino acids. Moreover, since the shape of a protein defines its function, some proteins must be able to switch between different shapes to function properly.
Ion channels are proteins that form pores through cell membranes, allowing ions to flow in and out of the cell. Potassium ion channels, which are found in neurons and heart muscle cells, have four subunits that move to open or close the central pore in response to various signals.
The closing of the channels can be ‘fast’ or ‘slow’. When the channels are closed quickly (called fast inactivation), a small part of the protein ‘plugs’ the pore from the inside of the cell. However, the mechanism behind slow inactivation remained obscure. It was thought to involve hydrogen bonds between some of the bulky amino acids that are found at the edge the pore. However, testing this hypothesis—by replacing these amino acids with alternatives that cannot form hydrogen bonds—was tricky because none of the 20 naturally occurring amino acids were alike enough to be suitable replacements.
Now, Pless et al. have overcome this limitation by using synthetic amino acids that form hydrogen bonds that are stronger or weaker than those formed by the amino acids they are replacing. The results suggest that two types of hydrogen bond keep the pore open: one is a bond between two amino acids in the same subunit, and the other is an inter-subunit bond between amino acids in neighbouring subunits. Pless et al. suggest that opening the channel causes small movements that gradually weaken, and eventually break, these bonds in one of the four subunits. Specific amino acids within the pore are then free to twist and—via a cascade of similar movements in the other three subunits—block the pore and halt the flow of ions. As such, these networks of hydrogen bonds act as pre-set breaking points allowing channels to close, even in response to continued stimulation.
Since regulated potassium channel activity underpins healthy neurons and heart muscles; understanding what controls their inactivation rate may lead to new approaches to tune their activity and treatments for important diseases.https://doi.org/10.7554/eLife.01289.002
Enzymes and catalytic proteins have evolved to balance the thermodynamic challenges of stability and substrate throughput (Shoichet et al., 1995). Ion channels, for instance, must efficiently interconvert between open, closed and inactivated states to regulate ionic flux across biological membranes. Even small alterations in their function that change the rates of isomerization between states can underlie inherited or acquired diseases (Hille, 2001). Voltage-gated potassium (Kv) channels are tetrameric membrane proteins with a central ion-conducting pore domain, surrounded by four voltage-sensor domains (VSDs), which tightly regulate the conductive state of the pore domain (Figure 1A). The pore domain consists of two transmembrane helices (S5–S6) connected by a re-entrant pore helix, which forms the selectivity filter. Kv channels negatively regulate conductance after channel opening through a process termed inactivation. The rate and extent of inactivation exhibit considerable isoform-dependent differences, which are reflected in the physiological contributions of these channels to cellular excitability in neuronal and cardiac tissues (Bean, 2007; Smith et al., 1996; Spector et al., 1996; Sanguinetti and Tristani-Firouzi, 2006; Aldrich et al., 1979). Inactivation can be described by two kinetically and mechanistically distinct processes termed fast (or ‘N-type’) inactivation and slow (or ‘C-type’) inactivation (Hoshi et al., 1991; Kurata and Fedida, 2006; Hoshi and Armstrong, 2013). While the former results from a channel peptide docking within the open cytoplasmic entrance to the permeation pathway (Hoshi et al., 1990, 1991; Zhou et al., 2001), the latter is assumed to involve highly cooperative local conformational changes near the selectivity filter, a notion supported by electrophysiological, structural and computational approaches (Lopez-Barneo et al., 1993; Yellen et al., 1994; Liu et al., 1996, 1997; Starkus et al., 1997; Kiss et al., 1999; Cordero-Morales et al., 2006, 2007; Peng et al., 2007; Cuello et al., 2010a, b; Cordero-Morales et al., 2011; Ostmeyer et al., 2013; van der Cruijsen et al., 2013). However, and despite this available data, fundamental questions persist regarding the precise molecular determinants that mediate the rate of slow inactivation, the basis for cooperativity and the relationship between slow inactivation and the structural integrity of the selectivity filter (Hoshi and Armstrong, 2013). An ongoing challenge is to elucidate how these proteins precisely time the entry into the slow inactivated state in the presence of a sustained voltage stimulus.
The Kv channel pore domain contains intermeshed aromatic side chains, an arrangement termed the ‘aromatic cuff’ that is located at the extracellular end of the selectivity filter and pore helix (Figure 1). This highly conserved region has long been suggested to play a part in the stability of the pore and likely slow inactivation (Doyle et al., 1998; Larsson and Elinder, 2000; Kurata and Fedida, 2006), yet the dynamic rearrangements during inactivation have remained poorly resolved given that many side chains within this region are intolerant to replacement, likely due to the large chemical and steric changes produced by standard mutagenesis. Here, we bypass this experimental hurdle by employing subtle synthetic derivatives of naturally occurring side chains in combination with concatenated subunits to probe the inter- and intra-subunit atomic determinants that control the onset and cooperativity of slow inactivation in Kv channels.
Crystallographic data demonstrate the close physical proximity of highly conserved Asp and Trp side chains within the same subunit of potassium channels (Figure 2A) (Doyle et al., 1998; Long et al., 2005). Mutations at these positions (Asp447 and Trp434 by Shaker numbering, which is used throughout the manuscript) elicit drastic changes in channel function, including effects on inactivation (Perozo et al., 1993; Molina et al., 1997, 1998; Yang et al., 1997; Loots and Isacoff, 2000; Loboda et al., 2001; Cordero-Morales et al., 2011). However, functional studies have not demonstrated the chemical basis of the interaction between these two side chains. One previously proposed (Cordero-Morales et al., 2011; Hoshi and Armstrong, 2013), but untested possibility is that a H-bond is formed between the hydrogen on the indole nitrogen of Trp434 and the carboxylate moiety of Asp447, and disrupting this H-bond would promote conformational changes associated with slow inactivation. If correct, then a side chain such as Glu (with altered stereochemistry, but a conserved negatively charged carboxylate), might weaken the interaction but not abolish it completely. By contrast, the virtually isosteric but uncharged Asn side chain should not form a significant H-bond with Trp434 and should thus result in dramatically accelerated inactivation. Indeed, the Asp447Glu mutation leads to a rapid and complete decrease in ionic current (Molina et al., 1997, 1998), and this loss in conductance can be drastically slowed by addition of extracellular tetraethyl ammonium (TEA). This deceleration of current decay in the presence of extracellular TEA is a hallmark of slow inactivation and can serve to discriminate this process from other phenotypes, such as fast inactivation or generic protein dysfunction (Grissmer and Cahalan, 1989; Choi et al., 1991; Molina et al., 1997). The charge-neutralizing Asp447Asn mutation resulted in gating currents only, generally interpreted to reflect an inactivation phenotype too rapid to resolve (Figure 2B and Table 1) (Hurst et al., 1996; Yang et al., 1997; Loots and Isacoff, 1998). Conversely, manipulations of the putative H-bonding partner Trp 434 should cause complimentary effects. For example, if the hydrogen on the indole nitrogen of Trp434 does, indeed, play a critical role in slow inactivation, then removing it should drastically increase the rate of inactivation. However, there are no naturally occurring Trp derivatives that can faithfully test this hypothesis without perturbing the local structure. Other (smaller) aromatic side chains markedly alter channel function: Tyr in position 434 speeds up slow inactivation (and is sensitive to TEA, Figure 2—figure supplement 1), while Phe results in gating currents only (Figure 2C; Table 1) (Perozo et al., 1993; Yang et al., 1997; Cordero-Morales et al., 2011). Thus, while structural data suggests a possible H-bond between Trp434 and Asp447, the available functional data cannot definitely discriminate between the roles of side chain size and/or volume or hydrogen bonding ability being the major determinant of slow inactivation at position 434. We therefore employed synthetic derivatives of Trp to test the hypothesis that a H-bond between Trp434 and Asp447 is a rate-controlling interaction for slow inactivation. If this H-bond is a critical determinant of the rate of slow inactivation, predicted outcomes are that strengthening the interaction should decelerate inactivation, while weakening the putative H-bond would be expected to increase the rate of inactivation. To this end, we introduced F4-Trp (Figure 2D), a fluorinated Trp derivative that increases the acidity of the hydrogen on the indole nitrogen (Deutsch and Taylor, 1987, 1989), while leaving side chain size and hydrophobicity intact. This slowed the rate of slow inactivation by threefold (Figure 2E). However, this slowing is preceded by an initial faster inactivating component in the Trp434TAG + Trp trace (gray trace in Figure 2D), which may stem from a small degree of nonspecific incorporation of (endogenous) amino acids other than the one ligated to the tRNA (in this case Trp). This likely produces an underestimate of the true slowing of inactivation through fluorination as all naturally occurring amino acids other than Trp in position 434 lead to accelerated slow inactivation. As a chemical complement, we sought to incorporate 2-amino-3-indol-1-yl-propionic acid (Ind, Figure 1F) (Lacroix et al., 2012), a synthetic isosteric Trp derivative in which the indole hydrogen cannot act as a H-bond donor. As co-injection of Trp434TAG cRNA and Ind-coupled tRNA alone did not yield measurable ionic currents, Wild type cRNA was mixed with Trp434TAG cRNA and Ind-coupled tRNA. However, even in the presence of WT (Wild type) subunits, the resulting heteromeric channel population displayed a 70-fold increase in inactivation rate (Figure 2G), demonstrating that breaking the proposed Asp447–Trp434 H-bond in the absence of steric perturbation leads to substantially accelerated inactivation. Similar to our recordings with Trp434TAG + Trp alone, we observed an initial fast inactivating component with the Trp434TAG + Trp: WT mix (gray trace in Figure 2F), possibly indicating a small degree of nonspecific incorporation of endogenous amino acids in position 434. However, this also is unlikely to have a major impact on our results because, first, the resulting fast component is a minor component only visually discernible during the initial phase (less than 1 s) of the depolarization and, second, the overall measured effect is likely more affected (slowed) by the presence of the WT subunits that were necessary to obtain ionic currents with Ind-containing subunits. We thus conclude that slow inactivation is amenable to the simple atomic ‘push/pull’ of a single H-bond, and manipulation of the strength of this H-bond generates a spectrum of inactivation rates, that can be tuned to occur faster or slower than what is observed in WT channels.
Slow inactivation is highly cooperative (Ogielska et al., 1995; Panyi et al., 1995; Yang et al., 1997) and the single mutation, Trp434Phe, in the first of four concatenated subunits accelerates slow inactivation (Yang et al., 1997). We reasoned that the phenotype obtained by breaking the putative H-bond through manipulations at Trp434 should be mimicked by mutations at the complementary Asp447 site. Indeed, the Asp447Glu mutation in the first of four concatenated Shaker subunits accelerated inactivation rates, similar to those obtained from Trp434Phe concatemers (Figure 3A,B). Together, these data support the notion that the strength of the Asp447–Trp434 intra-subunit H-bond is directly correlated with slow inactivation rate, suggesting that breaking of this interaction is an intrinsic timing mechanism that tightly regulates Kv channel activity.
Structural evidence suggests that Trp435 (Figure 4A) forms an inter-subunit H-bond via its hydrogen on the indole nitrogen with the Tyr445 hydroxyl (Doyle et al., 1998; Larsson and Elinder, 2000; Kurata and Fedida, 2006), and therefore substitution of Tyr or Phe for Trp435 would be expected to disrupt this H-bond, and potentially accelerate inactivation (as observed for aromatic substitutions of the adjacent Trp434 residue). However, while the Trp435Ala mutation produced non-functional channels (as suggested by the absence of ionic or gating currents), Tyr and Phe substitutions at position 435 resulted in WT-like slow inactivation rates (Figure 4B,C), ruling out a role for Trp435 H-bonding in slow inactivation. However, the Tyr445Phe mutation results in a mix of gating current and ionic current, with markedly accelerated slow inactivation (Harris et al., 1998) (a phenotype antagonized by TEA) (Figure 4D, Figure 4—figure supplement 1). Furthermore, Tyr445Ala channels exhibited gating currents akin to Trp434Phe channels (Figure 4D; Table 1) (Heginbotham et al., 1994). Interestingly, crystallographic data (Doyle et al., 1998; Long et al., 2007) place the Tyr445 hydroxyl within 3 Å of the hydroxyl moiety of a conserved Thr or Ser side chain (Thr439 in Shaker, Figure 1B), raising the intriguing possibility of an uncharacterized inter-subunit interaction between Tyr445 and Thr439. Consistent with this possibility, the Thr439Val mutant exhibited exclusively gating currents (Figure 4E), while the Thr439Ser mutation resulted in a modest threefold faster inactivation than observed for WT channels, likely indicating a minor role for the Thr439 methyl group in slow inactivation (Figure 4E, Figure 4—figure supplement 3). Further, and unlike the Trp434Phe mutation (Kitaguchi et al., 2004), introducing Tyr445Ala, Tyr445Val or Thr439Val on the Thr449Val background did not slow inactivation to an extent that ionic currents could be observed (Figure 4—figure supplement 3). While these data point to an inter-subunit H-bond between Tyr445 and Thr439, they do not inform on the cooperativity between individual subunits or whether this interaction contributes to the same extent as the Asp447–Trp434 H-bond. However, we speculated that breaking this inter-subunit H-bond may have a more pronounced effect when introduced in only a single subunit, compared to the intra-subunit Trp434–Asp447 H-bond. Consistent with this possibility, either the Tyr445Ala or the Thr439Val mutation in the first of four concatenated Shaker subunits (Figure 5A) had similar phenotypes, with a clearly biphasic inactivation phenotype composed of fast (around 50 ms) and WT-like slow (around 3 s) components (Figure 5B). The fast component was affected by TEA, implicating a slow inactivation mechanism (Figure 5—figure supplement 1). The sizable gating currents at hyperpolarized potentials (Figure 5—figure supplement 2) suggest that either mutation (one per concatenated tetramer) reduces the ratio of ionic current to gating charge at a given voltage, an effect that would arise if a significant portion of channels rapidly adopt a non-conducting conformation. To further test this possibility, the pore blocker agitoxin II (Eriksson and Roux, 2002; Banerjee et al., 2013) was used to assay the gating currents as a metric for normalization of the number of channels present in the cell, and thus permitting an estimate of the relative reduction in ionic current in the mutant concatemers relative to WT concatemers. Indeed, we found the ratio of ionic current to gating charge to be significantly reduced in both mutant concatemers (Figure 5C), suggesting that a sizable proportion of channels rapidly enter an inactivated state upon depolarization. This behavior is further illustrated in Figure 5D, where currents from Tyr445Ala or Thr439Val concatemers were normalized to WT (by gating charge), thus emphasizing the very rapid and near-complete inactivation in Tyr445Ala and Thr439Val concatemers. These experiments establish a previously unidentified inter-subunit H-bond between Thr439 and Tyr445 that controls slow inactivation in Kv channels.
Thr441 and Thr442 are highly conserved amongst Kv channels and are favorably located at the junction of selectivity filter and pore helix (Figure 6A) for a possible role in pore stability and/or slow inactivation. We aimed to compare the relative contribution of Thr441 and Thr442 to slow inactivation with more extracellular structural elements of the selectivity filter. Interestingly, mutations here produce vastly different outcomes (depending on the amino acid substitution), including loss-of-function, alterations in open state stability, and the appearance of subconductance states with diminished selectivity (Yool and Schwarz, 1991; Heginbotham et al., 1994; Zheng and Sigworth, 1997). Consistent with these reports, we found that valine substitutions a 441 and 442 had severe consequences: while Thr442Val displayed a non-expressing phenotype (Figure 6C) (Zheng and Sigworth, 1997), the Thr441Val mutation resulted in voltage-dependent currents both in the inward and the outward direction, as well as reduced potassium selectivity (Figure 6B, Figure 6—figure supplement 1), suggesting a significant perturbation of the local structure. However, Thr441Ser channels displayed a WT-like GV whereas Thr442Ser channels displayed a modest left-shifted activation relationship, yet both had inactivation behaviors similar to WT channels (Figure 6B,C, Figure 6—figure supplement 2), supporting the notion that the hydroxyl moieties at positions 441 and 442 support normal pore function (in addition to the proposed role of the Thr442 methyl group in ion binding (Rossi et al., 2013)). However, despite the disruptive phenotypes observed in homotetrameric Thr441Val or Thr442Val channels, single subunit mutations had minimal effects in the background of a concatenated tetramer (Figure 6D,E). Thus, hydroxyl removal at either 441 or 442 produces channel phenotypes that are very mild compared to perturbations within the aromatic cuff, and importantly, their effects are not propagated to other subunits in the channel tetramer, suggesting they are unrelated to the cooperative mechanism of slow inactivation. We believe this is an important finding as it demonstrates that severe functional consequences of mutations at the selectivity filter are not necessarily linked to changes in slow inactivation.
Despite intense experimental scrutiny for almost 25 years, the molecular and atomic origin(s) of the ability of Kv channels to enter a non-conducting conformation in the presence of a sustained (voltage) stimulus has remained enigmatic. A major challenge with addressing the contribution of individual side chains in the selectivity filter and pore helix to slow inactivation is that many amino acids lack naturally occurring analogs that allow subtle manipulation without dramatic disruption of the overall structure of this critical protein region. Furthermore, since slow inactivation is tightly coupled to ion occupancy in the selectivity filter, it has been difficult to distinguish direct effects on the mechanistic underpinnings of slow inactivation from indirect effects arising from changes in the structural integrity of the selectivity filter. Here, we overcome this hurdle by employing subtle synthetic analogs of naturally occurring amino acids and by introducing isolated mutations in single subunits.
When using concatenated Shaker constructs to introduce single mutations in a fourfold symmetric channel, it is crucial to confirm that the concatenated subunits do not form functional channels that vary in their stoichiometry from that predicted from the cloning strategy. Albeit possible (McCormack et al., 1992; Hurst et al., 1995), we believe the constructs used here assemble correctly for three reasons. First, the Thr439Val concatemers and the Tyr445Ala concatemers showed almost complete inactivation over a period of only 200 ms (when corrected for the amount of gating charge). Second, we observed very low ratios of Imax to Qmax for Thr439Val concatemers and Tyr445Ala concatemers. Both scenarios are not compatible with the idea of a significant WT-only channel subpopulation; Lastly, Sigworth and co-workers have used the same concatemers and successfully demonstrated that channels containing only a single mutated subunit generally assemble in the correct stoichiometry (Yang et al., 1997). We conclude that the (vast majority of) concatemers assemble correctly, although we cannot ultimately rule out a small subpopulation of channels with WT-like slow inactivation.
Further, we employed fluorinated Trp derivatives, which have been used extensively to probe electrostatic (cation-pi) interactions (Dougherty, 1996) between Trp side chains and organic cations as fluorination allows a step-wise dispersion of the electronegative surface potential of aromatic side chains (Pless and Ahern, 2013). As such, our finding that F4-Trp in position 434 significantly slows channel inactivation could be interpreted as a result of a cation-pi interaction at Trp434 that is being diminished by fluorination. However, if this were true, Ind, a synthetic amino acid which lacks H-bonding ability, should have no effect on channel inactivation as it is isosteric and isoelectric to the native Trp side chain. By contrast, we observe a substantial increase in the rate of slow inactivation with Ind in position 434, a result not compatible with the notion of an energetically significant cation-pi interaction at Trp434. We thus conclude that it is the ability of the indole nitrogen to participate in a H-bond that regulates the strength of the intra-subunit interaction between Trp434 and Asp447.
Together, our experimental approaches provide strong evidence for two H-bonds that are critical for slow inactivation of Kv channels: one that confers stability within an individual subunit (the Trp434–Asp447 interaction), and a second that stabilizes the relative orientation of two adjacent subunits (the Tyr445–Thr439 interaction) (Figure 7). Although disruption of the Trp434–Asp447 interaction has profound effects on slow inactivation, breaking of the Tyr445–Thr439 interaction elicits more functionally significant phenotypes. We surmise that these comparatively more severe phenotypes seen when disrupting the Tyr445–Thr439 pair arise from its location at the inter-subunit interface, but cannot exclude the possibility that this difference arises from the fact that the Tyr445 backbone carbonyl is also directly involved in coordinating permeant ions. Furthermore, the Tyr445–Thr439 interaction is, to our knowledge, the first evidence for an inter-subunit interaction contributing to slow inactivation, possibly providing an explanation for the observed subunit cooperativity during slow inactivation. However, although previous studies have suggested evidence for both constriction (Baukrowitz and Yellen, 1996; Liu et al., 1996, 1997) and dilation (Hoshi and Armstrong, 2013) of the selectivity filter, the data here is not definitive in distinguishing these models of slow inactivation.
The notion that side chains critical to slow inactivation cluster around the ‘aromatic cuff’ (formed between the extracellular end of the selectivity filter and the pore helix) is further supported by the marked differences between side chains at the outer vs the inner end of the selectivity filter and pore helix: only those located around the ‘aromatic cuff’ result in notable effects on slow inactivation that are propagated to the entire channel (Figures 2–5), while those residing in the middle or lower section of the selectivity filter do not affect slow inactivation (see Figure 6 for positions 441 and 442; see (Heginbotham et al., 1994) for position 443).
Overall, the results point towards an intriguing molecular explanation for the mechanism of slow inactivation: upon depolarization and channel opening, the stability of the channel open state is proportional to the strength of two H-bonds that regulate entry into slow inactivation, thus endowing Kv channels with an intrinsic timing mechanism that tightly regulates their biological activity. During a sustained voltage stimulus, channels experience a sequential breaking of the Trp434–Asp447 and Tyr445–Thr439 H-bonds and given the relative arrangement of their hydroxyl moieties this would likely result in an anti-clockwise swivel movement of the Tyr445 backbone carbonyl away from the permeation pathway, ultimately disrupting the coordination and occupancy of potassium ions at the outer end of the selectivity filter. Such a scenario would lead to mutual repulsion between the Tyr445 backbone carbonyls of the remaining three subunits (Almers and Armstrong, 1980; Hoshi and Armstrong, 2013), further lowering filter-occupancy at its outer mouth. The resulting strain could trigger a cascade of disrupted H-bonds critical to inactivation near the extracellular end of the selectivity filter in all subunits, ultimately resulting in a fully inactivated channel.
Shaker IR (Inactivation Removed by deletion of amino acids 6–46) cDNA in pBSTA was used as the parent clone unless stated otherwise (note that Cys301 and Cys308 were present). For mutations, standard site-directed mutagenesis was employed in combination with automated sequencing to confirm successful incorporation of mutations. For experiments with concatenated Shaker tetramers (Yang et al., 1997), mutations were introduced in the first of the four subunits only: the first subunit was subcloned into pBSTA with SacI and XbaI, and standard site-directed mutagenesis was used to introduce mutations followed by sequence verification. Next, the mutated construct was subcloned (with SacI and XbaI) back into the parent concatemer.
For electrophysiology experiments, Stage V-VI Xenopus oocytes were prepared, and injected with cRNA transcribed with the T7 mMessage mMachine kit (Ambion, Austin, TX) as previously described (Pless et al., 2013). Oocytes were incubated at 18°C and all recordings were conducted within 12–72 hr after injection. The fluorinated Trp derivative F4-Trp (4,5,6,7-F4-Trp) was purchased from Asis Chem (Watertown, MA) and the Trp analog 2-Amino-3-indol-1-yl-propionic acid (Ind) was synthesized as described previously (Lacroix et al., 2012). The principle of the in vivo nonsense suppression methodology is outlined elsewhere (Pless and Ahern, 2013). In short, nitroveratryloxycarbonyl (NVOC) was used to protect the amine of the synthetic amino acid, while the carboxyl group was activated as the cyanomethyl ester for coupling to the dinucleotide pdCpA (Dharmacon, Lafayette, CO). The resulting product was stored in DMSO at −80°C before enzymatic ligation to a modified (G73) Tetrahymena thermophila tRNA, which was synthesized using an oligonucleotide by Integrated DNA Technologies (Coralville, IA) as a template. The NVOC protection group of the aminoacylated tRNA-UAA was removed directly prior to co-injection with the cRNA by UV irradiation for 8 min at 400 W. In a typical experiment, 10–80 ng of tRNA-UAA and 25–50 ng of cRNA were co-injected in a 50 nl vol. In control experiments, the tRNA coupled to pdCpA (without an appended synthetic amino acid) was co-injected with the Trp434TAG cRNA. The control did not yield currents larger than for uninjected oocytes, ruling out significant levels of non-specific amino acid incorporation, or re-charging of the tRNA with endogenous amino acids. Note that incorporation of Ind at position 434 did not result in measurable ionic currents. This was expected given the results with the conventional Trp434Tyr and Trp434Phe mutants, as side chains in position 434 with no propensity to contribute to a H-bond (such as Ind) were shown to result in gating currents only. However, the nonsense suppression method is generally not efficient enough to establish current levels of necessary magnitude to resolve gating currents, even for sites with exceptionally high incorporation efficiency (Pless et al., 2011, 2013). We thus co-injected WT Shaker cRNA with the Trp434TAG cRNA for incorporation of Ind. Despite the high degree of cooperativity between subunits (Figure 3 and Figure 5), this likely results in slower and less complete slow inactivation than would be expected for Ind-containing subunits alone and the 70-fold increase in inactivation rate (Figure 2) is likely to be an underestimate of the real acceleration of inactivation induced by the Ind side chain in position 434.
Two electrode voltage-clamp recordings were conducted with an OC-725C voltage clamp (Warner, Hamden, CT) in standard Ringers solution (in mM): 116 NaCl, 2 KCl, 1 MgCl2, 0.5 CaCl2, 5 HEPES (pH 7.4). TEA (St. Louis, MO) and agitoxin-II (Alomone Labs, Jerusalem, Israel) were dissolved in Ringers and stored at −20°C until use. Glass microelectrodes with resistances of 0.1–1 MΩ were backfilled with 3 M KCl. Currents were acquired using leak subtraction and from a holding potential of −80 mV, unless stated otherwise. To obtain conductance-voltage (GV) relationships, isochronal tail current amplitudes were plotted vs the depolarizing pulse potential. All data = mean ± SEM; Student’s t test was used to determine statistically significant differences. To obtain gating charge-voltage (QV) relationships, the total area of the off gating charge was plotting against the depolarizing pulse potential.
Mechanism of frequency-dependent broadening of molluscan neurone soma spikesJ Physiol 291:531–544.
Tetraethylammonium blockade distinguishes two inactivation mechanisms in voltage-activated K+ channelsProc Natl Acad Sci USA 88:5092–5095.https://doi.org/10.1073/pnas.88.12.5092
Molecular determinants of gating at the potassium-channel selectivity filterNat Struct Mol Biol 13:311–318.https://doi.org/10.1038/nsmb1069
Molecular driving forces determining potassium channel slow inactivationNat Struct Mol Biol 14:1062–1069.https://doi.org/10.1038/nsmb1309
Intracellular pH as measured by 19F NMRAnn N Y Acad Sci 508:33–47.https://doi.org/10.1111/j.1749-6632.1987.tb32892.x
Ion channels of excitable membranes (3rd edition)Sunderland, Mass: Sinauer.
Potassium channel assembly from concatenated subunits: effects of proline substitutions in S4 segmentsReceptors Channels 3:263–272.
A structural interpretation of voltage-gated potassium channel inactivationProg Biophys Mol Biol 92:185–208.https://doi.org/10.1016/j.pbiomolbio.2005.10.001
Effects of external cations and mutations in the pore region on C-type inactivation of Shaker potassium channelsReceptors Channels 1:61–71.
Pore mutations in Shaker K+ channels distinguish between the sites of tetraethylammonium blockade and C-type inactivationJ Physiol 499:361–367.
Role of methyl-induced polarization in ion bindingProc Natl Acad Sci USA 110:12978–12983.https://doi.org/10.1073/pnas.1302757110
A relationship between protein stability and protein functionProc Natl Acad Sci USA 92:452–456.https://doi.org/10.1073/pnas.92.2.452
Fast inactivation causes rectification of the IKr channelJ Gen Physiol 107:611–619.https://doi.org/10.1085/jgp.107.5.611
Ion conduction through C-type inactivated Shaker channelsJ Gen Physiol 110:539–550.https://doi.org/10.1085/jgp.110.5.539
Importance of lipid-pore loop interface for potassium channel structure and functionProc Natl Acad Sci USA 110:13008–13013.https://doi.org/10.1073/pnas.1305563110
Richard AldrichReviewing Editor; The University of Texas at Austin, United States
eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.
Thank you for sending your work entitled “Hydrogen bonds as molecular timers for slow inactivation in voltage-gated potassium channels” for consideration at eLife. Your article has been favorably evaluated by a Senior editor, a Reviewing editor, and 2 reviewers.
The Reviewing editor and the two reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.
Using the atomic structure of Kv1.2/2.1 as the roadmap, Pless et al. ask a series of specific questions regarding the mechanism of C-type inactivation focusing on the hydrogen bonds near the selectivity filter. For example, is the (probable) hydrogen bond between W434 and D447 important in C-type inactivation of Shaker potassium channels? Despite the efforts by many, the exact mechanism of C-type inactivation remains mysterious and any specific and firm mechanistic information will be appreciated by those in the field and this study has potential to make an important contribution. The use of unusual amino acids using the in vivo nonsense suppression method brings a fresh approach to study C-type inactivation. The overall experimental rationale/design is straightforward, the experiments are elegant and the conclusions are for the most part solid.
There are two major central issues that must be addressed:
1) The overly simplistic treatment of inactivation kinetics must be corrected, taking into account the existence of non-exponential decay.
Perhaps the most problematic issue with the study described in this manuscript may lie in the data analysis/interpretation and the data description (e.g., the manuscript text). Many of the sample ionic currents involving the in vivo nonsense suppression method exhibit double-exponential inactivation kinetics (e.g., Figure 2D blue, Figure 2F gray, and Figure 2H W434F blue). Except for Figure 1F gray and blue, it is not clear what is responsible for the double exponential character. Is it possible that the nonsense suppression method was not perfect? Given the double exponential character, it is also unclear how the currents were analyzed to estimate the time constant values plotted in the figures. It is also uncertain how the “rate of entry into slow inactivation” often mentioned in the text is estimated. Additionally, the manuscript does not report any attempt to separate the entry and recovery rates – this would be important because many of the sample currents shown (e.g., Figure 2D, 2F, 2H; Figure 3C) suggest that the inactivated state at depolarized test voltages was not an absorbing state and the recovery rate constant value was appreciable (as evidenced by “steady-state” currents in the sample data shown). If the authors really wish to address the stability of the open state (as opposed to the stability of the inactivated state), they may wish to separate them out.
2) The work has little power (or relevance) in distinguishing between the pore collapse and pore dilation hypotheses of inactivation. Speculation along these lines should be eliminated. Likewise the distinction between pore stability and inactivation is weak and should be toned down.
The data presented on the perturbation analysis done on positions Thr441 and Thr442 although extremely interesting are not conclusive evidence to rule out “pore stability and selectivity filter from C-type inactivation”. These far-reaching conclusions have not been systematically addressed by the authors. The fact that the authors introduce the mutation on the first subunit of the tandem tetramer and they saw no effect implies that one subunit is not sufficient to affect slow inactivation and it might need more than one, but concluding that we should “conceptually dissociate pore stability and selectivity filter from C-type inactivation” is mere speculation that is not demonstrated in this paper.https://doi.org/10.7554/eLife.01289.022
1) …Many of the sample ionic currents involving the in vivo nonsense suppression method exhibit double-exponential inactivation kinetics (e.g., Figure 2D blue, Figure 2F gray, and Figure 2H W434F blue). Except for Figure 1F gray and blue, it is not clear what is responsible for the double exponential character. Is it possible that the nonsense suppression method was not perfect?
The basis for the multi-exponential nature of slow inactivation is not known but this complication has not hampered the numerous previous studies that have nonetheless approximated the decay as a single exponential process. In this particular case, we believe the reviewers are referring to the small rapid components at the very beginning of the depolarizing pulses. The reviewers are most likely correct in attributing this to imperfect nonsense suppression. Indeed, we often suspect that a small degree of non-specific incorporation of amino acids other than the one ligated to the synthetic tRNA that we inject. Since Trp434 mutations generate extremely rapid inactivation (cooperatively), even a small amount of non-specific incorporation could be sufficient to translate into a rapid component in macroscopic currents. A small rapid component was observed irrespective of the ligated amino acid – the ‘WT rescue’ experiment that introduces Trp by nonsense suppression shows a similar rapid component, and so we think that this property is best attributed to some imperfections of the nonsense suppression. We were unsure of the importance of going into explicit detail about this in the original submission, so we thank both reviewers for highlighting this point.
We hope to emphasize that this issue does not significantly impact on the interpretation of our results: it is known from other studies as well as ours that 1) amino acids other than Trp induce a rapidly inactivating phenotype, and 2) that the process of slow inactivation is highly cooperative. Therefore, any ‘imperfections’ of the in vivo nonsense suppression method would lead to phenotypes that inactivate faster than WT. As we still observe a significant slowing in the rate of inactivation for F4-Trp, the values obtained by us are actually quite likely to serve as an underestimate of the real extent of slowing induced by fluorination of F4-Trp. Similarly, we cannot rule out that some of the fast inactivating component of the Ind-containing channels is ‘contaminated’ through non-specific incorporation of amino acids other than Ind. However, the fast component is very small compared to that observed with Ind. Moreover, Ind-containing subunits had to be mixed with WT subunits to obtain ionic currents (which will significantly slow inactivation). Therefore, values obtained through this approach also reflect acceleration induced through incorporation of Ind. We have now explicitly mentioned these limitations of our approach in the Results section, and we hope this will add significant candor and clarity to the text:
“However, this slowing is preceded by an initial faster inactivating component in the Trp434TAG + Trp trace (gray trace in Figure 2D), which may stem from a small degree of nonspecific incorporation of (endogenous) amino acids other than the one ligated to the tRNA (in this case Trp)…”
Given the double exponential character, it is also unclear how the currents were analyzed to estimate the time constant values plotted in the figures.
We have now clarified this by including explicit information on how the data were fit in all figure legends.
Regarding the specific case of the Trp434Phe concatemers, we did observe varying rates of inactivation in both Trp434Phe and Asp447Glu concatemers, with some being well-fit by a single exponential, while others may have been better fit with multiple time constants. To account for this observation (and to facilitate comparison with other mutants) we had limited the analysis to the first 2 seconds of the depolarization (for the Trp434Phe and Asp447Glu concatemers), and this is now explicitly mentioned in the figure legend of Figure 3. Further, and to exclude any bias introduced by this approach, we also used in parallel a ‘time-to-half- maximal-current’ metric as a more generic quantification of inactivation and obtained virtually identical results (shown in Figure 3–figure supplement 3). Finally, we would like to again remark that the molecular basis for the occurrence of multiple time constants during inactivation does not only show up in our data set, but is a generally accepted phenomena that remains ill- defined, significantly complicating the interpretation of potential multi exponential decays.
Additionally, the manuscript does not report any attempt to separate the entry and recovery rates – this would be important because many of the sample currents shown (e.g., Figure 2D, 2F, 2H; Figure 3C) suggest that the inactivated state at depolarized test voltages was not an absorbing state and the recovery rate constant value was appreciable (as evidenced by “steady-state” currents in the sample data shown). If the authors really wish to address the stability of the open state (as opposed to the stability of the inactivated state), they may wish to separate them out.
We appreciate this comment and realize the merits of separating the entry and recovery rates from inactivation. However, the recovery from inactivation in potassium channels has been recently shown to be a highly complex molecular orchestration that not only involves protein conformational changes, but is also primarily controlled by the water occupancy behind the selectivity filter (Ostmeyer et al., Nature, 2013). We thus believe that it would be beyond the scope of the current study to investigate effects on recovery from inactivation in a detailed manner. Furthermore, while our current data set does not allow to definitively exclude potential effects on recovery from slow inactivation, we believe major effects on recovery are unlikely for the following reason: in the simplified context of a two state model (open-inactivated), a faster recovery rate from inactivation would also be predicted to result in a faster initial decay that would result in an increased steady-state current (compared to WT). These criteria were not met by any of the mutants tested. Only Trp434TAG + F4-Trp displayed an increased steady-state current, but the concomitant initial acceleration of inactivation was due to the ‘imperfect’ incorporation of F4-Trp (as verified by the ‘imperfect’ incorporation of Trp; see also our above comments), making a major effect on recovery unlikely.
2) The work has little power (or relevance) in distinguishing between the pore collapse and pore dilation hypotheses of inactivation. Speculation along these lines should be eliminated. Likewise the distinction between pore stability and inactivation is weak and should be toned down.
The data presented on the perturbation analysis done on positions Thr441 and Thr442 although extremely interesting are not conclusive evidence to rule out “pore stability and selectivity filter from C-type inactivation”. These far-reaching conclusions have not been systematically addressed by the authors. The fact that the authors introduce the mutation on the first subunit of the tandem tetramer and they saw no effect implies that one subunit is not sufficient to affect slow inactivation and it might need more than one, but concluding that we should “conceptually dissociate pore stability and selectivity filter from C-type inactivation” is mere speculation that is not demonstrated in this paper.
We agree with the reviewers that our data has limited power to discriminate between these disparate concepts and in hindsight we feel we may have overstepped this interpretation in our data. Consequently, we have now removed all such conclusions regarding a potential discrimination between pore dilation vs collapse and pore stability vs slow inactivation. Specifically, we have made the following changes to the manuscript:
The final sentence of the Abstract has now been changed to: “...triggers slow inactivation by means of a local disruption in the selectivity filter...”
The second-to-last sentence of the Results section (“These data therefore serve to conceptually dissociate the notion of pore stability and selectivity filter from the mechanism of slow inactivation”) has been removed and the final sentence has been changed to:
“We believe this is an important finding as it demonstrates that severe functional consequences of mutations at the selectivity filter are not necessarily linked to changes in slow inactivation.”
Finally, we have now changed the final sentence of the second-to-last paragraph of the Discussion section to: “However, although previous studies have suggested evidence for both constriction and dilation of the selectivity filter, the data here is not definitive in distinguishing these models of slow inactivation.”https://doi.org/10.7554/eLife.01289.023
- Christopher A Ahern
- Harley T Kurata
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We would like to thank Dr Fred Sigworth for kindly providing the concatenated Shaker WT clone.
- Richard Aldrich, The University of Texas at Austin, United States
- Received: July 25, 2013
- Accepted: October 29, 2013
- Version of Record published: December 10, 2013 (version 1)
© 2013, Pless et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.