Independence and Cooperativity in Rearrangements of a Potassium Channel Voltage Sensor Revealed by Single Subunit Fluorescence

The model of Hodgkin and Huxley 1952 for the activation of voltage-dependent potassium channels postulates that voltage controls the conductance of membranes by changing the equilibrium between two states (resting and activated) of four identical and independent charged membrane particles. This hypothesis was supported by the finding that potassium channels are composed of four subunits (MacKinnon 1991), each containing an S4 segment that consists of a sequence of basic residues conserved within the primary structure of voltage-gated ion channels, and therefore hypothesized to confer voltage sensitivity (Noda et al. 1984; Greenblatt et al. 1985; Guy and Seetharamulu 1986). The idea that S4 is the Hodgkin and Huxley voltage sensing gating particle has been supported by recent evidence based on S4 accessibility to internal and external solutions (Yang and Horn 1995; Larsson et al. 1996; Yang et al. 1996; Yusaf et al. 1996; Starace et al. 1997; Baker et al. 1998), on total gating charge measurements after neutralization of its basic residues (Aggarwal and MacKinnon 1996; Seoh et al. 1996), and on real-time fluorescence measurement of its motion (Mannuzzu et al. 1996; Cha and Bezanilla 1997, Cha and Bezanilla 1998; Baker et al. 1998). These studies showed that S4 traverses the membrane, moves across the membrane electric field in the direction expected for the voltage sensor, and can account for the majority of the gating current that is generated during channel activation (acidic residues in S2 and S3 may also contribute to the gating charge; Seoh et al. 1996; Cha and Bezanilla 1997).

While the description of Hodgkin and Huxley captures some of the essential features of voltage-dependent gating (four identical gating particles, a sequence of transitions that yields sigmoidicity to channel opening), it has required two major modifications to account for the details of potassium channel gating. First, voltage sensor movement occurs not in one but in several steps (Stühmer et al. 1991; Schoppa et al. 1992; Bezanilla et al. 1994; McCormack et al. 1994; Schoppa and Sigworth 1998a,Schoppa and Sigworth 1998b,Schoppa and Sigworth 1998c). At least two of these steps are generated by sequential movements of S4 (Baker et al. 1998). Second, several lines of evidence indicate that late transitions that follow the major gating charge carrying steps are cooperative (Hurst et al. 1992; Tytgat and Hess 1992; Bezanilla et al. 1994; Lin et al. 1994; Sigg et al. 1994; Zagotta et al. 1994a,Zagotta et al. 1994b; Zheng and Sigworth 1997, Zheng and Sigworth 1998; Schoppa and Sigworth 1998a,Schoppa and Sigworth 1998b,Schoppa and Sigworth 1998c; Smith-Maxwell et al. 1998a,Smith-Maxwell et al. 1998b; Ledwell and Aldrich 1999). The most successful models for gating of the Shaker potassium channel describe it as a series of independent charge carrying transitions followed by at least one concerted or highly cooperative transition.

Cooperativity in channel activation represents an issue of fundamental importance for understanding the mechanics of channel function. Subunit interaction in activation is expected to have important physiological consequences by influencing voltage sensitivity in the same way that cooperative interaction among binding sites affects the sensitivity of an allosteric protein to ligand concentration. Such cooperativity could account for the discrepancies previously described between voltage sensitivity and number of charges in S4 (Stühmer et al. 1989; Papazian et al. 1991; Sigworth 1994; Tang and Papazian 1997; Smith-Maxwell et al. 1998a,Smith-Maxwell et al. 1998b).

Recent studies from the Aldrich laboratory have demonstrated a role for S4 in cooperativity in channel activation (Smith-Maxwell et al. 1998a,Smith-Maxwell et al. 1998b; Ledwell and Aldrich 1999). Smith-Maxwell et al. 1998a,Smith-Maxwell et al. 1998b found that a triple mutation in the Shaker S4, inserting the sequence of the Shaw S4, makes the voltage dependence of activation shallow and shifts it to positive voltages, while making the ionic current activate slowly, with a single exponential time course that lacks the sigmoidicity of Shaker. These effects were accounted for by slowing a final cooperative transition that follows the two independent charge-carrying steps per subunit according to the model of Zagotta et al. 1994b. As a consequence, this step, which itself appears to carry a small amount of gating charge (Smith-Maxwell et al. 1998b; Ledwell and Aldrich 1999), becomes rate limiting for channel opening. In other words, mutations in S4 can alter cooperative interactions between subunits in late steps of channel activation.

In this study, we used a novel approach to examine subunit interaction during activation. We did this by taking advantage of the ability of voltage-clamp fluorometry (Mannuzzu et al. 1996) to monitor particular gating rearrangements in site-specifically labeled protein segments of individual subunits in heterotetrameric Shaker channels. Our test consisted in determining whether or not S4 movement of a labeled subunit is influenced by coassembly with other subunits having different S4 gating properties. Absence of influence was taken to indicate independence and presence of influence to indicate cooperativity. This approach allowed us to evaluate whether cooperative interactions occur during the major S4 charge carrying movements that have been previously thought to be independent. We find that, of S4's two major charge-carrying steps, the first takes place independently in each subunit, whereas the second is influenced by cooperative interaction between S4 segments. Our results suggest that at least some of this cooperativity is intrinsic to the second major S4 movement. We interpret our results in light of the earlier kinetic models.

Experiments were performed on nonconducting, ball-deleted ShH4 (W434F/Δ6-46) Shaker channels (Kamb et al. 1987; Hoshi et al. 1990; Perozo et al. 1993) after removal of two native cysteines (C245V and C462A) (Mannuzzu et al. 1996) to ensure that membrane-impermeant fluorescent thiol reagents would attach exclusively to a known position of cysteine addition (S352C) near the outer end of S4. Site-directed mutagenesis, dimer and tetramer fusion gene construction, cRNA synthesis, and cRNA injection into Xenopus oocytes were as described previously (Isacoff et al. 1990) except that T7 Ambion kits MEGAscript or mMESSAGE mMACHINE (Ambion Inc.) were used for the transcriptions.

For the coinjection experiments, a S352C:L382V or S352C/L382V:wild-type ratio of 1.5:8.5 was used to maximize the fraction of channels containing only one fluorophore binding subunit. For free association between subunits, these ratios are predicted to yield 52, 37, 10, 1, and ∼0% channels with zero, one, two, three, or four labeled subunits, respectively, so that ∼62% of the fluorescence will come from channels with one labeled subunit.

We ensured that linked heterotetrameric constructs with one labeled subunit produced channels with the subunit stoichiometry defined by their tandem linkage in two ways: (a) by comparing linked tetramers, such as A-B-B-B, to coinjections of cRNAs A + B at a ratio that favored B by 5.7-fold (this comparison produced identical results for studies with the L382V mutation); (b) by attempting to reduce the opportunity for intermolecular assembly of tetramers and favoring intramolecular assembly. This was done by injecting progressively more diluted cRNA encoding the linked tetramers to express channel protein at low enough levels so that the highest chance of interaction in endoplasmic reticulum would be between subunits of the same nascent protein. For the linked tetramer C*/S-W-W-W (S352C-TMRM/R365S linked to three wild-type subunits; Fig. 1) at high levels of expression there was a positive f1, with no voltage shift, and a negative f2, with a very minor leftward voltage shift, consistent with channels containing multiple C*/S subunits due to intermolecular assembly—precisely the situation we need to avoid. At fourfold lower expression levels of C*/S-W-W-W, both f1 and f2 were positive, f1 was not shifted, and f2 was shifted far to the left. Further lowering of expression did not change the behavior. The results are consistent with intertetramer assembly at high expression levels, but not over the range of lower expression levels on which we based our analysis.

Voltage clamp fluorometry, oocyte preparation and incubation, and tetraglycine maleimide (TGM) blocking were performed as described earlier (Mannuzzu et al. 1996). In brief, defolliculated, cRNA-injected oocytes were incubated for 4–6 d at 12°C, either in MBSH [88 mM NaCl, 1 mM KCl, 0.41 mM CaCl₂, 0.33 mM Ca(NO₃)₂, 0.82 mM MgSO₄, 2.4 mM NaHCO₃, 10 mM HEPES, pH 7.5] or ND96 (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 5 mM HEPES, pH 7.6). At this temperature, channel protein is made, but little is expressed at the plasma membrane. Cysteines of native oocyte membrane proteins were then blocked by a 30-min incubation at room temperature in 0.1–0.5 mM TGM. The oocytes were subsequently washed, and then incubated for ∼14 h at room temperature to incorporate the channels into the plasma membrane. Fluorescent labeling was done for 30 min on ice with 5–10 μM tetramethylrhodamine-5-maleimide (TMRM) (Molecular Probes).

TGM was synthesized by incubating 25 mM tetraglycine (TG; Aldrich Chemical Co.), dissolved in 0.1 M NaCl, 0.1 M Naphosphate, pH 7.25, with 10 mM succinimidyl trans-4-(N-maleimidylmethyl)cicloexane-1-carboxylate (SMCC; Molecular Probes) for 1 h at 37°C. TGM was purified from the unreacted components by HPLC, using a semipreparative C18 reversed phase column, and an acetonitrile gradient (+0.1% trifluoroacetic acid). TGM concentration was measured according to Sedlak and Lindsay 1968.

The two-electrode voltage clamping was performed with a CA-1 amplifier (Dagan Corp.). External solution was NaMES (110 mM NaMES, 2 mM Ca(MES)₂, 10 mM HEPES, pH 7.5). Capacitance compensation was performed from a holding potential of +60 mV (or +100 mV for channels carrying the L382V mutation). A photomultiplier tube (HC120-05; Hamamatsu Phototonics) was used for fluorescence measurements.

Since fluorophore attachment could alter gating, and interaction between fluorophores bound to neighboring subunits could alter the fluorescence report of gating movements, two means were used to ensure that differences in fluorescence–voltage relations (f-v's) between channels did not arise from differences in the number of labeled subunits per channel. First, f-v's were compared, and found to be similar, in channels composed of subunits with similar voltage dependence, but with a different number of labeled subunits (e.g., see Fig. 7 B). Second, photodestruction of TMRM molecules bound to the 352C homotetramer (labeled to saturation) produced only a small (<5 mV) leftward shift in the f-v (Fig. 2). This indicates that TMRM–TMRM interaction between subunits had no significant effect on our comparisons, and that, if anything, it led to a slight underestimate of the degree of the cooperative influence. Nevertheless, to avoid this complication, comparisons were also made between channels made of distinct subunits, but in which only one subunit was labeled (Fig. 7 B).

All comparisons were made within the same batch of oocytes in the same experiment under the same labeling conditions. Values are reported as mean ± SEM.

Charge–voltage (q-v) and fluorescence–voltage (f-v) relations were obtained using Clampfit 6 (Axon Instruments). q-v's were constructed from the integrated “off” gating currents, evoked by repolarizations to −80 mV, after long enough depolarizations (25–500 ms) for the “on” gating current to decay to completion. f-v's were constructed from the amplitudes of the on fluorescence changes. Although detailed models of Shaker activation may require functions more complicated than Boltzmann to describe Shaker's q-v, it has been shown that Boltzmann functions fit very well the q-v curves of wild-type and mutant Shaker channels (Stefani et al. 1994; Bezanilla et al. 1994; Baker et al. 1998). Therefore, we estimated the voltage parameters of both q-v and f-v curves by fitting the data points to Boltzmann functions using Origin 3.0 (Microcal).

We designed a method to exclusively monitor a specific set of Shaker protein rearrangements in single subunits during channel gating transitions. The method was based on three elements.

First, a cysteine substitution at S352 near the outer end of S4 allowed for the site-specific attachment of a sulfhydryl-reactive fluorescent probe. As shown earlier (Baker et al. 1998), and further documented below, fluorescence of this labeled site tightly correlates with S4's voltage-sensing transmembrane rearrangements. Second, coinjection was used to express channels made of a combination of subunits with and without the cysteine mutation (Fig. 1). Coinjections were weighted to favor the unlabeled subunit, yielding a large fraction of channels with only one labeled subunit that provided the bulk of the fluorescent signal (see materials and methods). These constraints on stoichiometry were further assured by tandem linkage of subunits (Isacoff et al. 1990; Fig. 1). Third, mutations that alter specific activation transitions were employed to make heteromeric channels in which the labeled and unlabeled subunits differ in known gating steps (Fig. 1). The expectation was, if certain activation transitions are cooperative, then gating movements in the labeled subunit would be altered by the presence of unlabeled subunits with distinct energetics of specific S4 gating steps.

To monitor the activation movements of S4, position S352C in S3–S4, near the NH₂ terminus of S4, was labeled with TMRM (see materials and methods). Homotetramers made of S352C labeled with TMRM (C*) showed large fluorescence changes in response to voltage steps (Fig. 3 A), with a steady state fluorescence–voltage relation that was statistically indistinguishable from the voltage dependence of gating charge movement (q-v) (Fig. 3 B). This indicates that C* fluorescence can serve as a reporter of S4's charge-carrying conformational changes in a labeled subunit.

The first voltage-shifting mutation chosen was L382V (Lopez et al. 1991) (referred to as V in this paper), which decreases the steepness of the voltage dependence of a late component of the gating charge, and shifts its v_1/2 to positive values, without altering the total gating charge (Schoppa et al. 1992). This effect has been accounted for by a reduction in the forward bias of a concerted step that follows the main charge carrying steps in each subunit (Schoppa et al. 1992; Sigworth 1994; Schoppa and Sigworth 1998b,Schoppa and Sigworth 1998c). As previously shown for comparisons of wild-type and L382V homotetramers (Schoppa et al. 1992), the voltage dependencies of gating of C* and V homotetramers differed considerably in the positive range (Fig. 4 A). In addition, the foot of the f-v (and q-v) of C* homotetramers was shifted to the right compared with the q-v of V homotetramers (Fig. 4 A). Such a difference at the foot of the q-v is not seen between V and wild-type homotetramers (Schoppa et al. 1992), and it was due to a right shift of the C* q-v compared with both the V and wild-type q-v curves. The different behavior of the V and C* homotetramers in the whole voltage range of gating charge movement made it possible to look for interactions that affect both early and late transitions.

To examine how interactions between wild-type and V subunits influence S4 activation rearrangements, subunits that were wild type except for the cysteine labeling site substitution were coexpressed with unlabelable V subunits, or labelable V subunits were coexpressed with unlabelable wild-type subunits (W; Fig. 1 A).

Complementary RNAs (cRNAs) encoding the S352C subunit and the V subunit were coinjected into oocytes in a ratio chosen to optimize the number of channels containing three V subunits and one fluorophore-binding S352C subunit (Fig. 1 A, and see materials and methods), and then labeled with TMRM. The f-v of the S352C-TMRM (C*) subunit in these heterotetramers (C* + V in Fig. 4 A) was shallower and shifted to the right with respect to the C* homotetramers, approaching the shallow depolarized component of the V homotetramers (Fig. 4 A). The same result was obtained when subunit stoichiometry was constrained by using the linked tetramer C*-V-V-V (Fig. 4 B). The effect on the single C* subunit of the three V subunits was strong enough to make the f-v indistinguishable from that of the homotetramer of the double mutant 352C*/382V (C*/V) (Fig. 4A and Fig. C).

In contrast to the pronounced influence of the V subunits on the C* subunit's rearrangements at positive voltages, the foot of the f-v of the C* subunit was not shifted toward the q-v of the V homotetramer (Fig. 4 A). This suggests that interactions between subunits affect late transitions, but not early transitions.

The voltage-shifting influence of the three V subunits on a single C* subunit could result from a strong cooperative influence of the V subunits, or, alternatively, the L382V mutation may eliminate a cooperative interaction that normally makes the q-v comparatively steep. To distinguish between these possibilities, we determined whether a V subunit remains sensitive to the cooperative influence exerted by wild-type subunits. This was done in channels composed of a single labeled C*/V subunit and three W subunits. In both coinjection of C*/V cRNA with an excess of W cRNA, and expression of the linked tetramer C*/V-W-W-W, the f-v's generated by the single C*/V subunit combined with three W subunits were virtually the same as the f-v of the C*/V homotetramers (Fig. 4 C). The gating currents of the C*/V-W-W-W channels, three quarters of which should be generated by W subunits, were intermediate in kinetics between those of the C*/V homotetramer and the W homotetramer (Fig. 5). These results demonstrate that the L382V mutation renders the subunit insensitive to cooperative influence on its S4 movement by wild-type subunits (Fig. 6), a remarkable effect that leaves the L382V subunit operating independently.

The above results suggest that early steps in activation are independent and that late steps are cooperative. However, the early and late charge carrying transitions could not be resolved in the f-v's of C*-V-V-V and C*/V-W-W-W (Fig. 4), preventing us from precisely identifying the cooperative step(s). This loss of resolution is due to the combined influence of the cysteine mutation at S352 and to its conjugation with TMRM, as found earlier for the double mutant A359C-TMRM/L382V (Mannuzzu et al. 1996).

To determine more clearly which activation movement of S4 is cooperative, we employed another voltage-shifting mutation, R365S. Like other neutralizations of this second arginine in S4 (Perozo et al. 1994; Aggarwal and MacKinnon 1996; Seoh et al. 1996), R365S shifts apart the voltage dependencies of the two major charge-carrying steps (q1 and q2) of channel activation (Bezanilla et al. 1994). The q1 and q2 steps correspond to two sequential outward movements of S4 between three distinct transmembrane topologies corresponding to the deactivated, intermediate, and activated states of the channel (Baker et al. 1998). Neutralization of R365 provided the possibility of directly relating cooperative interactions to transitions between these known conformations of S4. In addition, unlike L382V, R365S appears to maintain its cooperativity, since, as for wild-type channels, despite the similarity in the magnitude of charge carried by q1 and q2, the steepness of q2 is greater than that of q1 (Perozo et al. 1994; Fig. 7 A). This could be explained by positive cooperativity in q2, or, alternatively, by negative cooperativity in q1.

As we have previously shown (Baker et al. 1998), TMRM conjugated at S352C in the double mutant S352C/R365S (C*/S) experienced a two-step fluorescence change (f1 and f2) that well correlates in their voltage dependence with the two steps of charge movement (q1 and q2). This correspondence is clearly demonstrated in Fig. 7 A, which shows that the voltage parameters from the double Boltzmann fit to the q-v equally fits the f-v (solid lines in Figure 7A). This correlation was also confirmed by examining the kinetics of the C*/S fluorescence change. Fig. 8 (left) shows fluorescence traces from the linked tetramer C*/S-S-S-S, which had a kinetic behavior (data not shown) and steady state behavior (Fig. 7 B) indistinguishable from the C*/S homotetramer. From very negative voltages up until −30 mV, only a single component that increases fluorescence with depolarization was present. Further depolarization changed the fluorescence during the step (on fluorescence) to one in which a fast fluorescence increase was followed by a slower fluorescence decrease to a steady state level. As the step potential was made more positive, both the magnitude and rate of activation of the downward fluorescence component increased, resulting in a progressively smaller and shorter lasting upward-going transient (Fig. 8 B, arrow). In the same voltage range (from about −20 to +50 mV), two components were also evident in the fluorescence upon repolarization (off fluorescence), a rapid fluorescence increase followed by a slower fluorescence decrease. This sequence of events in the on and off fluorescence change is consistent with the behavior, in wild-type channels, of q1 and q2 (Bezanilla et al. 1994), which we have shown to be better resolved, and preserved in character, in channels with an R365 neutralization (Baker et al. 1998).

The good correlation between f1 and q1, and between f2 and q2 (Fig. 7 A), indicates that TMRM at 352 reports on both steps of transmembrane S4 movement, making it possible to determine if the cooperative interactions occur during one or both of these charge carrying transitions.

In the following experiments, the subunit carrying both the S352C and R365S mutations was labeled with TMRM (C*/S), and the S4 movements of this voltage-shifted subunit were examined in three contexts (Fig. 1 B): (a) in C*/S homotetramers, (b) in linked C*/S-S-S-S heterotetramers, made of one C*/S subunit and three unlabeled voltage-shifted R365S subunits (S), and (c) in linked C*/S-W-W-W heterotetramers, made of one C*/S subunit and three unlabeled wild-type subunits. The rationale behind these experiments was that comparison of the fluorescence of C*/S in these channels should make it possible to resolve the influence of wild-type subunits on the q1 and q2 steps of S4 movement.

The f-v of the C*/S homotetramers was virtually indistinguishable from that of the linked tetramer C*/S-S-S-S (Fig. 7 B and Table). In contrast, activation gating of the single C*/S reporter subunit was clearly influenced by its coassembly with three W subunits (Fig. 7 B). This influence was seen not only in the voltage dependence of fluorescence, but also in the direction of the f2 fluorescence change, which became positive (Fig. 7 B and 8, right). A free parameter fit of the C*/S-W-W-W f-v to the sum of two Boltzmann functions gave an estimate for the v_1/2 of f2 that was significantly different (P = 1.2 × 10⁻⁶) from that of C*/S-S-S-S. The three W subunits shifted the v_1/2 of the C*/S f2 in the hyperpolarized direction, so that it approached the v_1/2 of the f2 of W homotetramers (Fig. 9 A, solid line; Table). In other words, the W subunits exerted a cooperative influence in f2. Unlike the evidence for cooperativity in f2, the f1 estimated from the free fit of the C*/S-W-W-W f-v to the sum of two Boltzmann functions was not significantly different from that of C*/S-S-S-S (Table), indicating that the q1 transition in the reporter subunit is not substantially influenced by the other subunits in the channel. This is consistent with the absence of a shift toward more depolarized voltages in the foot of the f-v of the C*/S-W-W-W tetramer.

The above conclusion that W subunits affected q2 but not q1 motion of the C*/S subunit S4 is supported by the fact that the f-v of C*/S-W-W-W was closely approximated by the sum of two fixed Boltzmann functions, one having the voltage parameters of the f1 of C*/S-S-S-S, and the other having the voltage parameters of the q2 of W homotetramers (Fig. 9 B, solid line). In contrast, a double Boltzmann composed of the opposite combination of parameter values (the f1 from the q1 of W homotetramers and f2 of C*/S-S-S-S), failed to fit the f-v of C*/S-W-W-W (Fig. 9 B, dotted line). Finally, a sum of two Boltzmann functions made of both f1 and f2 of C*/S-S-S-S, but with reversed f2 polarity, also did not fit the data (Fig. 9 B, dashed line).

In summary, the good agreement between the f1 of C*/S-W-W-W and that of C*/S-S-S-S points to an absence of cooperativity in q1. The large shift to the left of the C*/S subunit's f2 in C*/S-W-W-W channels, towards the q2 values of W homotetramers, is consistent with a strong cooperativity in the q2 movement of S4.

In this paper, we used a new approach to study interactions between subunits in heteromeric membrane proteins. This method relies on the use of cysteine mutagenesis to introduce an attachment site for a thiol-reactive fluorophore into a specific functional domain of a subunit. These modified subunits are coexpressed with subunits that cannot be labeled because they do not contain accessible cysteines. In this manner, fluorescence measurements only monitor the structural rearrangements of the labeled domain in particular subunits without interference from signals in unlabeled subunits. This approach makes it possible to define how rearrangements in a specific domain of one subunit, such as those evoked by voltage change or ligand binding, are coupled to functional rearrangements in other subunits of a multimeric protein. The advantage of studying cooperativity in this manner is that it provides a clear indication of subunit interaction, circumventing the need to dissect apart gating or ionic currents, which are governed by the complex ensemble properties of all of the gating domains in each of the channel's subunits.

For this method to work, two criteria need to be met: (a) the fluorophore should report only on the conformational changes of the subunit to which it is bound, and (b) the fluorescence report of conformational rearrangements should not be distorted by interaction between fluorophores bound to multiple subunits of the same channel. In this context, we have shown that Shaker channels carrying the substitution S352C labeled with TMRM (C*) fulfill these two criteria. With respect to the first criterion, we showed that the fluorophore bound to this site is not sensitive to gating movements in neighboring subunits, since the f-v's of two distinct subunits (C*/V and C*/S) are different although they have the same three neighboring subunits (three Ws in the tetramers C*/V-W-W-W and C*/S-W-W-W; Fig. 4 and Fig. 7; Table). Further, the f-v of C*/V was the same whether associated with the same subunits in the homotetramer C*/V or with different subunits in the heterotetramer C*/V-W-W-W (Fig. 4 C). The second criterion was met in that, despite the difference in the number of labeled subunits, the f-v's of C*/V and C*/S homotetramers, with four labeled subunits, were found to be equivalent to those of the linked heterotetramers C*/V-W-W-W and C*/S-S-S-S, with only one labeled subunit (Fig. 4 C and 7 B). This data is consistent with evidence that the S3–S4 loop of one subunit is far enough away from the S3–S4 loops of other subunits to allow for their simultaneous binding of multiple molecules of the peptide hanatoxin (Swartz and MacKinnon 1997a,Swartz and MacKinnon 1997b). Additionally, since basic residues in S4 have been shown to interact with acidic residues in the S2 and S3 of the same subunit (Tiwari-Woodruff et al. 1997), the microenvironment around an S3–S4 fluorophore is also likely to be defined by the same subunit.

We used this approach to determine whether activation during gating involves cooperative interactions between subunits of the Shaker potassium channel. Our results, using two mutations that exert distinct effects on channel activation, indicate that a specific step of voltage sensing is cooperative, while another is independent. This cooperativity influences both the energetics of specific S4 activation states and the conformation of S4 or of the protein that immediately surrounds it.

Our results with both the mutation L382V and R365S indicate that cooperative interactions occur in late activation transitions, but not in early ones. The results were particularly informative with the mutation R365S, which neutralizes the second basic residue in S4. This arginine is one of the key positive charges in S4. During activation it traverses practically the entire membrane-spanning protein and membrane electric field (Aggarwal and MacKinnon 1996; Larsson et al. 1996; Seoh et al. 1996; Starace et al. 1997; Baker et al. 1998). In the resting and intermediately activated S4 membrane positions, it is buried deeply enough (Baker et al. 1998) to probably require electrostatic interactions with acidic residues in S2 and S3 for structural stability (Papazian et al. 1995; Tiwari-Woodruff et al. 1997). Based on the kinetics of q1 and q2 charge movements and on the relationship between magnitude of the charge and slope, we have concluded that the mutation R365C closely resembles wild type, except that the energetics of the three transmembrane topologies of S4 (associated with the deactivated, intermediate, and activated states) are altered, resulting in large voltage shifts (Baker et al. 1998).

We took advantage of the easily resolved q1 and q2 gating charge movements in S352C-TMRM/R365S subunits to correlate cooperative interactions with specific S4 activation movements using fluorescence. We found that the q1 gating rearrangement is insensitive to the identity of the other subunits in the tetramer. In contrast, the q2 rearrangement is significantly influenced by the other subunits. Thus, in the context of the q1–q2 model of gating charge movement, the results are consistent with cooperative interactions occurring only in q2.

As shown in Fig. 3, upon depolarization, the S352C-TMRM homotetramers produce a fluorescence increase throughout the voltage range. The same is true for homotetramers of S352C-TMRM/L382V, despite the shift by the L382V mutation of the late charge carrying step to more depolarized voltages (Fig. 4 C). Although it is difficult to resolve distinct charge carrying components in these two homotetramers, it appears that all the major steps of S4 movement produce increases in fluorescence with activation for this labeling position. In contrast, although the f1 component of S352C-TMRM/R365S homotetramers also increases fluorescence with depolarization, f2 decreases fluorescence with depolarization (Fig. 7 A). Since fluorescence depends on the local fluorophore environment, this reversal of the direction of f2 indicates that the mutation R365S not only alters the relative energetics of S4 gating states, but that it also alters the conformation of the S3–S4 loop at position 352, and/or of its protein surround.

Interestingly, we found that the identity of the amino acid at position 365 is not the sole determinant of the ratio of brightness of the intermediate state versus the fully activated state for that subunit's S352C-TMRM. Coassembly of one C*/S subunit with three W subunits restored the direction of the f2 fluorescence change to that of C*, with the fully activated state being brighter than the intermediate state (Fig. 7 B and 8). This indicates that subunit interaction during gating affects the protein conformation either of the S3–S4 or of the protein microenvironment around S3–S4. In summary, interaction between subunits influences: (a) the energetics of an S4's intermediate and fully extruded states, and (b) the conformation of the segment just external to S4 (or of its protein surround).

The cooperativity observed in q2 could be “intrinsic” to S4, in the sense that the transmembrane state of one S4 could directly or indirectly influence that of S4 in other subunits. Alternatively, the cooperativity could result from an “extrinsic” influence of other coupled transitions. The basis of this cooperativity is examined in the next two sections.

Measurements of ionic and gating currents by the Aldrich and Sigworth labs have led to models of Shaker channel activation that consist of two or three major charge carrying transitions that occur independently in each subunit. These are followed by a concerted step that: (a) either directly opens the channel, or immediately precedes channel opening, (b) carries between 1.4 and 1.8 e₀ per channel, and (c) is strongly forward biased (Hoshi et al. 1994; Zagotta et al. 1994a,Zagotta et al. 1994b; Schoppa and Sigworth 1998a,Schoppa and Sigworth 1998b,Schoppa and Sigworth 1998c; Smith-Maxwell et al. 1998a,Smith-Maxwell et al. 1998b; Ledwell and Aldrich 1999). Mutations that are modeled to reverse this forward bias separate the main voltage sensing movements from channel opening (Schoppa and Sigworth 1998b,Schoppa and Sigworth 1998c; Ledwell and Aldrich 1999), whereas a mutation modeled to increase the forward bias can actually cause the channel to open before all the voltage sensors have moved, leading to subconductances (Zheng and Sigworth 1997, Zheng and Sigworth 1998; and see Chapman et al. 1997).

Our experiments are consistent with the q2 phase of S4 charge movement being cooperative. According to the models described above, q2 movements are intrinsically independent, but could be made functionally cooperative by being tightly coupled to an ensuing forward biased concerted transition. In this context, the apparent cooperativity of S4 would depend on the ratio of the forward and backward rate constants of the concerted transition for each contributing subunit. A prediction of such a model is that heteromeric channels composed of mixtures of W subunits (having a strongly forward-biased concerted transition) and V subunits (in which the same transition is slightly backward biased) would be expected to have an intermediate behavior weighted by the number of each subunit type. If the q2 movement of S4 borrowed cooperativity from the concerted step, then the S4 of a single V subunit should still experience cooperative influence from the three W subunits. Therefore, channels made of one V and three W subunits should still have a strongly forward-biased concerted step. Our results do not support this prediction. Instead, they show that the S4 of a V subunit is not influenced at all by other subunits with which it coassembles. This implies that the L382V mutation affects more than just the rates of the concerted step. One possibility is that the mutation directly eliminates an intrinsic cooperativity of q2.

Our results suggest that the intermediate to fully extruded motion of S4 is intrinsically cooperative. But how does the transmembrane state of one S4 affect that of S4 in other subunits? We have two possible indications about this mechanism. The first is the dependence of the q2 cooperativity on L382. L382 is located at the internal end of S4, in a segment that is not buried in the protein even when S4 is fully extruded at positive voltages (Larsson et al. 1996). Although the NH₂-termini of S4s appear to be too far apart from each other (Swartz and MacKinnon 1997a,Swartz and MacKinnon 1997b) to directly interact, nothing is known about the distance between the intracellular ends of S4. Thus, L382 could be part of a structure that either directly couples S4 segments to one another when these are in the intermediate and/or fully extruded state (see discussion on the putative leucine zipper motif; McCormack et al. 1991; Durell et al. 1998) or indirectly couples S4s via interaction with Shaker's internal activation gate (see Liu et al. 1997). The second point to consider is that interactions between subunits also change the environment of the S3–S4 loop near the NH₂ terminus of S4 (Fig. 7 B and 8). This suggests that S4 movements may also be coupled via structural rearrangements of the part of the channel protein surrounding S4, including the pore-forming S5 and S6.

We have used single subunit fluorescence to obtain evidence for cooperativity in a charge-carrying voltage sensor movement of S4 during the activation of Shaker potassium channel. Of the two sequential outward steps that S4 takes in response to membrane depolarization, the first appears to occur independently in the four subunits, and the second step appears to be cooperative. This cooperative interaction between subunits not only affects the relative energetics of specific S4 gating states, but also changes the conformation of the S3–S4 loop with respect to its protein environment. With this method in hand, it should now be possible to identify the structures that couple gating movements of the S4s of different subunits, and S4 movements to Shaker channel gates.

We thank L. Llewelyn for making the dimer and tetramer linked constructs, O. Baker, P. Larsson, H. Lecar, J. Ngai, M. Moronne, E. Loots, C. Gandhi, M. McGrath, and all the other members of the lab for helpful discussions.

L.M. Mannuzzu was supported by a post-doctoral fellowship from the American Heart Association (AHA), California Affiliate (95-41). This work was supported by grants from the National Institutes of Health (NS35549-03) and AHA, California Affiliate (96-248A).