The carboxyl tail forms a discrete functional domain that blocks closure of the yeast K⁺ channel

Abstract

Non-targeted mutagenesis studies of the yeast K⁺ channel, TOK1, have led to identification of functional domains common to other cation channels as well as those so far not found in other channels. Among the latter is the ability of the carboxyl tail to prevent channel closure. Here, we show that the tail can fulfill this function in trans. Coexpression of the carboxyl tail with the tail-deleted channel core restores normal channel behavior_. A Ser/Thr-rich region at its amino end and an acidic stretch at its carboxyl end delineate the minimal region required for tail function. This region of 160 aa apparently forms a discrete functional domain. Interaction of this domain with the channel core is strong, being recalcitrant to removal from excised membrane patches by both high salt and reducing agents. Although the use of a cytoplasmic domain to regulate channel is common among animal channels, by using it as a “foot-in-the-door” to maintain open state appears unique to TOK1, the first fungal K⁺ channel studied in depth.

The K⁺-specific ion channel in the plasma membrane of the budding yeast Saccharomyces cerevisiae (TOK1) is one of the first microbial K⁺ channels to be examined in detail. Genetic analysis, coupled with biophysical analysis, of TOK1 has assisted in the identification of what we call the PP region, the cytoplasmic end of the P-region-following membrane domain, which has been found to be intimately involved with gating of cation channels (1, 2, 3). Our analyses have also pointed to the existence of functional domains that, on first glance, appeared unique to TOK1, including a filter-specific gating (Fig. 1A; ref. 22) and a carboxyl tail stabilization of the inner-pore gate (Fig. 1B; ref. 4). Further characterization of these “unique” properties are important not just for the sake of understanding TOK1 function, but for possible general relevance. Both filter-specific gating (5–8) and cytoplasmic domain influence of channel gating (9–18) are emerging as common themes amongst K⁺ channels. Here, we report on a more specific analysis of the ability of the carboxyl tail to influence TOK1 gating.

Figure 1.

Simplified model of TOK1 gating. In this model of TOK1 gating, the R state results from a collapse of the filter in response to an inward Δμ whereas the C state results from occlusion of the inner pore akin to deactivation gating of other cation channels. The K/Vm dependency of C is modeled as resulting from K⁺ forcing the R-state filter gate open, which blocks opening of the inner gate. The carboxyl tail stabilizes the non-C states by dynamically blocking inner-gate closure à la a “foot-in-the-door” method. See ref. 2 for a detailed description of this model.

TOK1 is predicted to contain eight transmembrane domains with canonical P-region pore loops following both the fifth and seventh spans (19, 20). The biophysics and genetics of this channel have been explored, but its physiological function remains largely unknown except as a target for killer toxin (6, 7). Although commonly referred to as “voltage-regulated” (21), TOK1's gating is regulated by the K⁺ electrochemical potential (Δμ) rather than merely voltage alone (19, 20).

The front line of this Δμ-dependent regulation is a near-instantaneous gating process that prevents inward current flow, the “R” state (1). Because it is the only place where the transmembrane Δμ can be efficiently monitored, the R state is most readily accounted for as an intrinsic gating property of the filter region (22). We proposed a model in which inward Δμ results in K⁺ occupancy state(s) of the filter that favor a non-conducting collapsed state whereas outward Δμ conversely favors a conducting non-collapsed state (Fig. 1). Recently, K⁺-evacuation-dependent collapse of the related KcsA filter has been directly observed in crystal structures (8).

TOK1 also dwells in a set of more stable closed “C” states that are dependent on negative voltage and high external K⁺ (8). Mutations exclusively in the “PP” region (the end of M6 and M8 in the case of TOK1) specifically disrupt the C states (1) suggesting that they result from a constriction of the inner mouth of the pore, akin to deactivation-type gating in other cation channels (2, 3, 23, 24). We have proposed that the K/Vm dependence of C could simply result from the open non-collapsed filter preventing inner pore gating transitions (4). The C-locking open filter would be maintained by its K⁺ occupancy in the face of high external K⁺ and negative Vm when the inner gate is closed (Fig. 1). In addition to R and C, TOK1 also dwells in a voltage-independent “IB” closed state that could reflect stochastic inner gate closure in the absence of subsequent filter gate opening (4).

In a screen for intragenic suppression of a PP mutant, it was found that deletion of the carboxyl tail dramatically decreased dwell in the non-C states (4). This resulted both from an increase in deactivation rate to IB, causing shortened open burst durations, and an increase in the rate of deactivation to C, reflected in a dramatic positive shift in the C-state/voltage relationship. In contrast, activations from either IB or C were not substantially altered by tail deletion. This unidirectional effect favors a model in which the tail acts like a “foot in the door” dynamically preventing closure to both IB and C (Fig. 1), presumably by interacting with the inner-mouth gating region in such a way as to block inner gate closure.

Our model was partly deduced from the behavior of tailless mutant channels (1). A deletion can be effective by removing only one of several necessary parts. Thus, the previous experiments did not show that the C-terminal tail performs all of the functions necessary to maintain the open state. To test whether the C tail forms a discrete and fully competent domain and to delineate its boundary, we coexpressed the tail with the tailless channel to test for possible functional restoration. Here, we present results that further define the action of the carboxyl tail in TOK1 gating, including the ability of the tail to act in trans, a delineation of the minimal region required for its function, and evidence of a strong but undefined interaction between the tail and the channel core.

Materials and Methods

DNA Manipulation and Oocyte Expression.

All RNA injected into oocytes was produced in vitro from plasmid templates as described previously (1). Shortened TOK1 templates, both carboxyl and/or amino deletions, were produced by using standard PCR techniques to generate ORFs that were inserted between the 5′ and 3′ untranslated sequences of the Xenopus β-globulin gene in the oocyte expression vector, pGH19 (9). All PCR-generated constructs were single-strand sequenced to verify the absence of mutations.

In vitro RNA synthesis and oocyte isolation, injection, and maintenance were as described previously (9). For wild-type TOK1 analysis, ≈10 ng of RNA was injected per oocyte. For tail deletants, ≈50 ng RNA was injected, because observation of tailless-type currents required injection of greater amounts of RNA (4). For coinjection experiments, ≈30 ng of RNA encoding the cytoplasmic tail was coinjected with ≈10 ng of RNA encoding the end-545 channel core.

Electrophysiological Recordings.

Macropatch and single channel patch recordings were performed as described elsewhere (1). In all cases, pipette (external) solutions contained 140 mM KCl (1 M in the case of Fig. 5A), 1 mM MgCl₂, 1 mM CaCl₂, 5 mM Hepes (pH 7.5), and the bath (internal) solutions all contained 140 mM KCl, 4 mM MgCl₂, 1 mM CaCl₂, 5 mM Hepes, and 5 mM EGTA (pH 7.5). All recordings were carried out at 21–22°C.

Figure 5.

Standardized G/V plots of critical tail deletions. (A) G_standard/V plot for end-670, end-672, and end-674, shown to illustrate the carboxyl extent of the functional tail domain. Plots were generated as in Fig. 3B. (B) Similar plots of stop-545 tailless channel core coexpressed with tails of differently length as diagramed.

Results

Carboxyl Tail Blocks Closure in Trans.

Tail deletion dramatically increases dwell in the C states (4). When depolarized from strongly negative holding potentials in 140 mM external K⁺, the majority of wild-type channels activate with relatively slow kinetics indicative of C dwell at such holds (Fig. 2A, C→O). Depolarization from less negative holds results in larger fractions of the channels activating in less than a millisecond (appearing as instantaneous activation) reflecting significant dwell in R at mildly negative holds (Fig. 2A, R→O). Tail deletion causes channels to stay in C at much more positive potentials. Even at +80 mV, tailless channels are almost exclusively partitioned into the C states (Fig. 2B).

Figure 2.

Tail restores wild-type C-state/V behavior in trans. Activations of wild-type (A) end-545 tailless (B) and tailless channels coexpressing the cytoplasmic amino acids 456–691 tail (C) were assessed from various 10-mV incremented holding potentials. Activations from C state are distinguished by their slow time course, whereas activations from R appear instantaneous on these time scales. All traces are 1 s long, with 15-s inter-episode delays. The precise voltage protocol is shown in Insets (note the tailless channels have a more positive protocol than the others because channels stay in C state until very positive potentials). All currents were recorded from excised macropatches exposed to symmetrical 140 mM KCl solutions. Calibration bars represent 1 nA × 200 ms. Horizontal carets represent 0-current level.

The carboxyl tail can restore normal C-state behavior when expressed in trans. Coinjection of tail mRNA with that of the tailless core produces channels that activate with wild-type C-state/V relationships (Fig 2C). Like the wild-type channels, the majority of channels resulting from the coexpression of the tail and core are partitioned out of the R state at positive holding potentials.

Coexpression of the tail has profound effect on the steady-state current/voltage (I/V) profiles (Fig. 3A). The conductance (G/V) profile of wild-type channels increases between 0 mV (reversal potential) and 100 mV and then slightly decreases at more positive potentials. The reason for this negative slope conductance was not investigated, but it should be noted that the G/V relationship reflects at least two gating processes, R-to-O and C-to-O, the former of which is not well understood, at least at the quantitative level. Because tailless channels remain in C until highly positive potentials, tail deletion causes a large rightward shift in the steady state conductance. Coinjection of the tail RNA restores the wild-type G/V relationship. The lower average current levels from coinjection channels compared with wild-type channels (Figs. 2C and 3) likely results from the fact that an excess of tail-encoding RNA was coinjected in these experiments to saturate the channel cores expressed from the smaller amount of RNA.

Figure 3.

Tail restores wild-type rectification in trans. (A) Near-steady-state currents were elicited from oocytes expressing TOK1 channels described in Fig. 2 by applying 30-s voltage ramps between −90 and +200 mV. (B) These currents were converted to conductances standardized to the values at 200 mV based on a 0-mV reversal potential given the presence of symmetrical K⁺ as the sole permeant ion.

Delineation of the Minimal Functioning Tail.

In an attempt to ascertain the ends of the domain that alters C-state gating, the abilities of shortened tails to maintain wild-type G/V profiles were assessed. To this end, recursively more focused tail deletions were engineered and assayed.

In the first round, stop codons were placed at three approximately equidistant intervals between the initially isolated mutant end-545 (i.e., final amino acid is 545) and the wild-type end-691 channel: end-578, end-624, and end-663. All three of these deletions produced channels with tailless-type conductances (Fig. 4A, upper half), indicating that a region within the final 28 aa of the tail is necessary for inhibiting closure.

Figure 4.

Delineation of minimal contiguous functioning tail. (A) Summary of recursive deletion experiments designed to pinpoint the minimal extent of the carboxyl end of the tail required to maintain wild-type rectification as in Fig. 3. Numbers represent the terminal amino acid. Individual rounds of experiments are grouped. (B) Summary of similar coinjection experiments designed to delineate amino end of the restoration-potent tail. (C) Amino acid sequence of carboxyl tail with the minimal contiguous rescuing domain underlined, Thr/Ser residues bold, and acidic residues bold and italicized. The amino terminus of the functional domain is marked by a Thr/Ser-rich region, and the carboxyl terminus is marked by a stretch of acidic residues. +, Restoration of the wild-type G/V profile as in Fig. 3; −, failure to restore; +/−, intermediate restoration.

Successive deletion sets were then recursively engineered to pinpoint the carboxyl end of the tail's functional domain (Fig. 4A, lower half). It was found that deletion of the terminal 22 aa (end-669) produces a channel with typical tailless conductance profile, whereas deleting 4 aa fewer (AA 673) produces a current with typical wild-type behavior (Fig. 5A). An intermediate deletion following AA 671 produced a channel with an intermediate conductance profile. This region between 669 and 673 is notable in having four acidic amino acids, EEDE (Fig. 4C).

To ascertain the amino end of the tail domain, the abilities of shorter coexpressed tails to restore wild-type current voltage profiles to the tailless channel core were assessed. As with deletion of the carboxyl terminus, recursive deletions of the amino terminus were examined. Whereas a tail starting at AA 494 works, one starting at 524 does not. All of the tails starting between 502 and 514 restore wild-type behavior. A tail starting three amino acids downstream, at 517, did not work (Figs. 4B and 5B). Thus, the region between amino acids 514 and 517 is critical for the tails function in trans. AA514 is at the start of a region notably rich in Ser and Thr residues. However, exposure of inside/out macro patches of oocytes expressing wild-type channels to either bacterial or shrimp alkaline phosphatase had no effect (data not shown).

A 160-aa tail extending from the established amino to carboxyl termini of the functional domain, amino acids 514 to 673, was coexpressed. As expected, it also restores a wild-type current/voltage profile to the tailless channel. Thus, the minimal contiguous domain of the carboxyl tail responsible for inhibiting channel closure is 160 aa long, extending from a Ser/Thr-rich region at its amino end, to an acidic stretch at its carboxyl end.

Tail Interacts Strongly with the Channel Core.

The carboxyl tail is predominantly hydrophilic, dictating that it interacts with the channel peripherally rather than as an additional transmembrane domain. Wild-type channel activity from excised membrane patches from oocytes coexpressing the tailless core and the tail does not revert to a tailless state after extensive perfusion of the bath recording solution, suggesting that this interaction is stable. Because the functional tail domain is highly charged, it might be predicted that the interaction of the tail with the channel core is primarily electrostatic. Attempts were made to interrupt such electrostatic interaction by rinsing excised coexpressing patches with 1M KCl. Even after perfusion for 10 min, channels maintained wild-type G/V profiles (Fig. 6A), indicating that the tail's interaction with the core was not disrupted. Visual inspection of the macropatch indicated that the membrane face was clean and being subject to perfusion, although the presence of tail-entrapping structures cannot be ruled out.

Figure 6.

Tail effect cannot be rinsed away from excised patches. (A).Current/voltage relationship measured from excised patches coexpressing end-545 core with amino acids 456–691 tail before (dark trace) and after 10-min perfusion (gray trace) of 1M KCl. (B) Similar experiments with 10 mM DTT instead of extra KCl. In both cases, excised patches were exposed toward the direction of perfusion and the patch contained no accompanying cellular material observable under a microscope.

The coexpressed tail contains two cysteine residues at positions 580 and 599. To test the possibility that the tail was being covalently linked to the core by disulfide bond, coexpressing patches were rinsed with 10 mM DTT. After 10 min perfusion, channels still maintained wild-type G/V profiles, indicating that disulfide bond formation was probably not anchoring the tail to the core (Fig. 6B).

Attempts were made to rinse the tail from the core by using high pH, which releases tightly bound peripheral membrane proteins from Drosophila membranes (25), as well as 1 M guanididium, a chaotropic agent that disrupts hydrogen bonding. Both treatments resulted in rapid loss of seal integrity so that their ability to extricate the tail could not be determined. In summary, the hydrophilic carboxyl tail appears to bind quite stably to the excised membrane patch, but the nature of this binding could not be determined.

Discussion

Previous deletion experiments indicate that the carboxyl tail of TOK1 functions to maintain the channel in the open state by inhibiting two closures: closure to voltage-independent IB state and closure to the voltage- and external-K⁺-dependent C states (4). Here, we have shown that the tail can function even when it is not covalently attached to the core channel. We defined the minimal contiguous region of the tail necessary for tail function, and shown that the coexpressed tail is intimately associated with the membrane.

In our model of TOK1 gating (ref. 4; Fig. 1), the tail dynamically interacts with the mouth of the channel to prevent closures to both IB as well as C, acting as a “foot-in-the-door” per se. In its simplest interpretation, it might be predicted that the tail binds to the channel in its open and IB conformations but dissociates when the channel is more fully closed in C. The inability of extensive perfusion to rinse away the unlinked tail, however, suggests that it remains associated with the channel even in the C states. In some respects, similarities may exist between the mechanism of TOK1 tail function and that of Shaker-type β subunits containing intrinsic N-type inactivation domains. Both the Shaker β subunit and the TOK1 tail constitutively associate with the channel core through state-independent noncovalent interactions (26). Their gating effects, presumably mediated by further interaction with the cores (4, 26), require opening of the inner gate. These state-dependent interactions must be distinct. Whereas the hydrophobic terminus of the N-inactivation gate appears to intimately bind to the channel pore as an extended peptide (27) such that it blocks ion flux, however, the TOK1 tail does not require a hydrophobic terminus and its proposed interaction does not occur deep enough in the pore to block conduction but only enough to block inner-gate closure.

The mechanism by which unlinked TOK1 tail associates with the core remains elusive. That the tail contains numerous charged residues and that the functional domain is demarcated by four acidic residues hints at an electrostatic interaction, but the fact that high salt perfusion could not remove the tail argues against this interpretation. That Thr and Ser residues demarcate the amino terminus of the tails functional domain may indicate that hydrogen bonding is important, but unfortunately we were unable to directly test the ability of chaotropic agent to extricate the tail. It could well be the case that the tail is not bound to the core or other protein, but to the membrane itself. Although it is clearly too hydrophilic to be a transmembrane protein itself, posttranslational modification of the protein such as myristylation could lead to its anchoring to the membrane. Coincidentally, expression of the tail results in the production of both the predicted and higher molecular weight peptides in yeast (data not shown). Finally, although good perfusion did appear to be occurring over clean membrane patches, the presence of invisible macroscopic structure that prevents its extraction cannot be ruled out.

It, of course, is not necessarily the case that the demarcating terminal residues are required for association, but for some other aspect of tail function instead. The requisite stretch of acidic residues at the end of the tail could be involved in Ca²⁺ binding. Such a “Ca²⁺-bowl” binding motif is required for the gating function of BK K⁺ channel (15). Coincidentally, the C-terminal domain of the BK channel can also function in trans. It was reported that Ca²⁺ has a complex effect on the gating of TOK1 (28), but those investigations varied cytoplasmic-side Ca²⁺ between 10 μM to 10 mM, a presumably nonphysiological spectrum. Others have reported no effect of Ca²⁺ on TOK1 gating in yeast (21).

Phosphorylation of the carboxyl tail of plant and animal K⁺ channels regulates their gatings (9, 13, 14, 18). Despite the absence of recognizable kinase-substrate consensus sites, the demarcating Ser/Thr stretch of amino acids could likewise indicate an importance of phosphorylation in tail function. Treatment of excised patches with alkaline phosphatase failed to alter TOK1 currents. This result could indicate either that phosphorylation in fact has no role in TOK1 gating, that the dephosphorylated form is the functional form, or that the phosphatase failed to dephosphorylate the channel. Consistent with the notion that cytoplasmic-side phosphorylation is important, TOK1 channel rundown could result from loss of tail function and can be prevented when ATP is added to the bath of membrane patches excised from yeast (29).

The ability of discrete cytoplasmic domains to affect the gatings of a broad spectrum of ion channels is emerging as an important general theme in ion-channel regulation (4, 9–18). In most cases, little is known about the mechanism and sites of interaction of these domains with the gating mechanism of the channel itself. The carboxyl tail of TOK1 has a profound effect on gating. That tail function can influence yeast proliferation coupled with the power of microbial genetics makes TOK1 an ideal system for an unbiased survey of such cytoplasmic-domain regulation. Although we have begun to uncover the mechanisms involved, much is still to be learned.