Mutation in pore-helix modulates interplay between filter gate and Ba²⁺ block in a Kcv channel pore

Opening and closing of K⁺ channels is in most cases determined by two types of gates: the so-called inner gate is formed by dynamic hydrophobic barriers that control entry into and exit from the cavity on the cytosolic side (Yellen, 2002; Kim and Nimigean, 2016; Kopec et al., 2019). The filter gate faces the extracellular side of the channel protein and is part of the selectivity filter structure where it controls the current through the filter (Cuello et al., 2010; Kim and Nimigean, 2016; Kopec et al., 2019). After discovering filter gating in KcsA (Cordero-Morales et al. 2006, 2007), structural and functional studies have implied that this type of gating is a general phenomenon in K⁺ channels including physiologically important Kir (Sackin et al., 2009) and Kv channels (Köpfer et al., 2012) as well as K2P channels (Natale et al., 2021).

Gating in the filter is generated and modulated by critical interactions between the selectivity filter and surrounding parts of the protein (Cordero-Morales et al., 2007; Cuello et al., 2010; Kim and Nimigean, 2016; Kopec et al. 2019, 2023). This mechanical interplay between the filter and supporting scaffold causes small positional changes in the delicate filter geometry, which then favor or disfavor current through the filter. A prominent example of this kind of filter gating is C-type inactivation. It occurs for instance in several Kv channels after voltage-dependent opening of the inner gate (Peters et al., 2013). Also, in K2P channels, the selectivity filter interacts with the surrounding protein (Niemeyer et al., 2016; Piechotta et al., 2011; Lolicato et al., 2020) and translates in this manner the effect of external regulatory inputs into a modulation of filter gating kinetics in these channels.

In a recent study, we used mutagenesis and single-channel recordings of the small channel Kcv_NTS to obtain more general insight into structure/function correlates of filter gating (Rauh et al., 2022). This channel is particularly suitable for addressing this question because it exhibits distinct voltage-dependent closures at negative voltages, which can be assigned to filter gating (Rauh et al., 2018). Also, since the Kcv_NTS channel has no inner gate (Rauh et al., 2017), these fast open–closed transitions at negative voltages must be entirely generated by operation of the filter gate.

It was already shown that the position S42 in Kcv_NTS is structurally equivalent to the amino acid E71 in KcsA and V55 in MthK (Fig. 1 A and Fig. S1 A), a position which is critical for filter gating in the latter channels (Rauh et al., 2022; Kopec et al., 2023; Chakrapani et al., 2011). As in the case of KcsA and MthK, mutations of this critical position in the pore-helix of Kcv_NTS also affect filter gating. It was shown that mutations S42T and S42V slow down the closing of the filter gate in the latter channel (Rauh et al., 2022). In addition to its effect on fast gating, these two mutations but also the S42A mutation elicited an additional gating phenomenon: all three mutants introduce long-lasting and voltage-dependent closures at positive voltages. The data suggested at first glance that mutations at the critical position 42 modulate in addition to the aforementioned fast gating, a second gate. In the absence of an inner gate in this channel, the data suggest that voltage-dependent closure at positive voltages is the result of an additional filter gating mode. Here, we show that this additional voltage-dependent closing at positive voltages is not generated by an additional gating mode of the filter gate but by an increase in the sensitivity of the selectivity filter to a voltage-dependent block by another ion. Our data show that the S42T mutation generates a 200-fold increase in the sensitivity of the channel to cytosolic Ba²⁺. Because the binding site for the Ba²⁺ block is in these channels, part of the selectivity filter we predict that manipulations of the coupling between pore-helix and selectivity filter are not only affecting the process of gating but also the fine structure of the selectivity filter, which determines the propensity of an entering ion to permeate or block the channel pore.

Serine in position 42 in the Kcv_NTS channel was mutated into all other 19 proteinogenic amino acids by site-directed mutagenesis using a protocol based on the method described in (Braman et al, 1996). The coding regions of all constructs were sequenced. In vitro protein expression and purification of the Kcv_NTS channel and its mutants were performed as described previously (Rauh et al., 2022) in an in vitro reaction with the Expressway Mini Cell-Free Expression System (Invitrogen). The nascent channel proteins were thereby directly embedded into lipid nanodiscs. The latter MSP1D1-His DMPC nanodiscs (Cube Biotech GmbH) contain multiple His-tags, which served in the next step for purification by metal chelate affinity chromatography (0.2 ml HisPur Ni-NTA spin columns, Thermo Fisher Scientific). Different from the manufacturer’s instructions, the washing and elution buffers were salt-free. The column was washed three times with 400 μl of 20 mM imidazole, and the nanodisc–channel complexes were eluted in three fractions (200 μl each) with 250 mM imidazole. This improves the reconstitution efficiency into the bilayer. For details of the procedure see Rauh et al. (2021).

Functional reconstitution of channel proteins in planar lipid bilayers was performed as in previous studies (Rauh et al. 2017, 2021). Cis and trans chambers were filled with standard recording buffer (100 mM KCl [AppliChem GmbH], 10 mM HEPES [AppliChem GmbH] in double-distilled water, and adjusted to pH 7.4 with KOH). Bilayers were formed from 1,2-diphytanoyl-sn-glycero-3-phosphocholine (DPhPC, Avanti Polar Lipids) by the pseudo painting/air bubble technique. For single-channel recordings, an elution fraction was diluted in 250 mM imidazole solution 1:1,000–1:100,000 and 1–3 μl was added directly below the bilayer in the trans compartment with a bent Hamilton syringe. Insertion and orientation of a single-channel protein were monitored during short-voltage steps. The pronounced asymmetry of Kcv_NTS channel conductance allowed the determination of channel orientation in the bilayer. The trans compartment was grounded, and membrane voltages were applied to the cis compartment. With a strong bias of the Kcv_NTS protein for inserting exclusively with the cytoplasmic side first into the bilayer, positive currents correspond to an outward current in the in vivo situation.

Steady-state single-channel currents were recorded at room temperature (∼25°C) over a voltage window from +160 to −160 mV in steps of 20 mV for 1–5 min. Both chambers were connected via Ag/AgCl electrodes to an amplifier (L/M-EPC-7; List-Medical). Currents were filtered with a 1-kHz four-pole Bessel filter before digitization with a sampling frequency of 5 kHz (LIH 1600; HEKA Elektronik).

For block experiments with divalent cations, an appropriate amount of a stock solution containing 10 mM of the divalent ion with Cl⁻ as anion was added to the cis (= intracellular) or trans (= extracellular) chamber and carefully mixed by repeated pipetting. Contaminations of Ba²⁺ in the KCl buffer were eliminated according to Neyton (1996) by adding 200 µM of the chelating macrocyclic polyether (+)-18-Crown-6-tetracarboxylic acid (18C6TA; Aldrich Chem. Co.) to the buffer solution.

The use of g_t⁸ instead of the simple cumulative sum g_t improves the attenuation of high-frequency noise.

A jump is detected as soon as the test value g_t⁸ exceeds the threshold value λ⁸. The incidence of the jump is estimated as the time at which g_t last adopted the value 0. Subsequently, g_t is set back to 0 and the algorithm starts again with the opposite sign, as it now searches for a jump from µ₁ to µ₀.

For dwell time analysis, only measurements at membrane voltages greater than or equal to +60 mV were used. The lowest SNR value was achieved at +60 mV and was always at least SNR = 20. Consequently, the effective temporal resolution limit of our analysis was not determined by the H.O.H.D. but by the 1-kHz four-pole Bessel filter used in our measurement setup. The latter has a rise time T_r of about 400 µs. We therefore assumed that virtually all events with a duration ≥1 ms were detected correctly and that the majority of missed events and events registered incorrectly in terms of their duration fall within the interval t = [0, 1 ms]. For this reason, we only considered events with a duration of >1 ms for the dwell-time analysis.

To be able to assume that the mean lifetimes of the closed-time populations correspond to the true mean lifetimes, only those measurements were used whose mean open time was >20 ms. In this way, the number of open events with t < 1 ms should be <5% and therefore negligible. A missed events correction of the mean closed times was therefore not carried out.

Each closed-time population was then treated as an individual closed-state Cj that can be reached exclusively from the open state O as shown in Scheme 1.

The assumption that there are no transitions between the closed states is a simplification resulting from the lack of information about possible hidden C–C transitions. If additional information on the state transitions becomes available in the future, which requires a modification of this model topology, algorithms are available to transform equivalent models into each other (Kienker, 1989).

Experimental data are generally presented as arithmetic/geometric mean ± arithmetic/geometric standard deviation (SD) of N independent experiments.

The analysis of single-channel traces was performed using the program Kielpatch (https://www.zbm.uni-kiel.de/aghansen/software.html). Matlab 2014b (MathWorks, Inc.) was used for dwell-time analyses. The Matlab files are available from the corresponding author upon reasonable request by email.

Metal ion contaminations in measuring buffer were analyzed using inductively coupled plasma mass spectrometry (ICP-MS, Analytik Jena Plasma Quant MS Elite). Sample preparation (double-distilled water or 100 mM KCl, 10 mM HEPES/KOH, pH 7) included filtration with a 0.45-µm regenerated cellulose (RC) filter (Perfect Flow syringe filter, Wicom) and acidification with ultrapure HNO3 (ROTIPURAN ≥69 %, Carl Roth) to pH < 2. Calibration standards in a concentration range of 0.2–100 µg/l for barium were prepared from a multielement standard (Analytik Jena). A 10-µg/l internal standard solution of Sc, Y, In, and Bi (diluted 100 mg/l Analytik Jena Internal standard solution) was added online via a peristaltic pump to all samples and standards to compensate for drift of the ICP-MS system.

Fig. S1 shows crucial amino acids for filter gating in Kcv_NTS and other K⁺ channels. Fig. S2 shows the effect of different symmetrical K⁺ concentrations on the voltage-dependent inactivation of Kcv_NTS S42T. Fig. S3 shows that the addition of 18C6TA to the cytosolic solution abolishes the voltage-dependent closures in Kcv_NTS S42A and Kcv_NTS S42V.

Single-channel activity of wild type Kcv_NTS and its S42T mutant (Kcv_NTS S42T) was recorded like in previous studies using a combination of in vitro translation into nanodiscs and functional reconstitution into DPhPC bilayers. In symmetrical 100 mM KCl, Kcv_NTS has a high open probability over the entire window of test voltages between −160 and +160 mV (Fig. 1, B and D). Typically, the channel shows fully resolvable open and closed events at positive voltages and very fast open/closed transitions at negative voltages (Fig. 1 B). The latter is the result of sub-millisecond gating, which can be causally linked to filter gating in this channel (Rauh et al., 2018). Functional reconstitution of Kcv_NTS S42T reveals channel activity with a twofold increase in unitary conductance (Fig. 1 C) as well as distinct impacts on gating (Fig. 1 B). In a previous study, we have already shown that the S42T mutation stabilizes the filter gate, resulting in a slow-down and full resolution of open/closed transitions at negative voltages (Rauh et al., 2022).

At positive voltages, Kcv_NTS S42T generates distinct long closures, which increase in frequency with depolarizing voltages (Fig. 1 B). At first glance, this channel closing at positive voltages has similarities to C-type inactivation in that the voltage dependency of channel inactivation is shifted toward more positive voltages with increasing KCl concentrations (Fig. S2). A plot of the P_O values as a function of voltage can be fitted for low and medium KCl concentrations with a single Boltzmann function (Fig. S2 B). From the fit parameters, it is evident that increasing KCl concentrations shifts the voltage for half-maximal inactivation (V_1/2) positive and decreases the value of the apparent gating valence δz (Fig. S2 C).

In a systematic study of voltage-dependent closure in the S42T mutant, we observed a large variability between different experiments. Fig. 2 shows examples of current traces and P_O values from two independent sets of experiments performed several months apart. The P_O/V relationships from both sets of experiments exhibit the same voltage-dependent inactivation but the position of the V_1/2 value is shifted by about 35 mV between the two experiments.

To understand the mechanism of inactivation in Kcv_NTS S42T at positive voltages and the reason for the variability in the data, we performed a dwell-time analysis. The data show that the mean open lifetime decreases exponentially with positive voltages while the mean closed lifetime has a weak bell-shaped voltage dependency with a maximum of about +140 mV (Fig. 3 B). Close inspection of the closed dwell-time histograms for voltages between +80 and +160 mV reveals three populations C1–C3 (Fig. 3 A). C1 has a mean lifetime τ_C1 of about 1 ms at the edge of resolution and disappears at voltages greater than or equal to +160 mV (Fig. 3 A). The frequency of events attributed to populations C2 and C3 increases with depolarization (Fig. 3 A). Calculation of the occupation probabilities reveals that the decrease in P_O is mainly caused by an increase in the occupation probability of C3 (P_C3) (Fig. 3 D). Notably, the latter is not the result of an increase in the mean lifetime τ_C3, but solely caused by an increase in closing frequency. In fact, the mean lifetimes of closed events C2 and C3 decrease almost exponentially with depolarizing voltages (Fig. 3 C). This behavior is rather unusual for a voltage-dependent gate and more reminiscent of a charged pore blocker, which can permeate through the pore as the electrical driving force increases. Therefore, we speculated that the S42T mutation may have not affected a gate in Kcv_NTS but increased the sensitivity of the channel to a blocker that is present in the solutions used in an undefined concentration.

To test this assumption, we performed experiments in which we clamped the WT and the mutant channel in symmetrical 100 mM KCl buffer to extreme positive voltages. This causes in Kcv_NTS WT only a small drop in P_O to about 0.8 at +300 mV (Fig. 3, E and F). The mutant channel on the other hand exhibits the expected voltage-dependent decrease in P_O toward zero as expected from the Boltzmann equation before rising again at voltages more positive than +200 mV (Fig. 3 F). The current traces in Fig. 3 E reveal that P_O rises again at extreme positive voltages because the closed times shorten to a greater extent than the closing frequency increases. This is a strong indication of a so-called punch-through behavior (French and Shoukimas, 1985; Nimigean and Miller, 2002) in which the blocking ion is relieving beyond a critical positive voltage the block by exiting through the filter to the extracellular solution rather than back toward the cytosol; this favors an increasing net conduction of K⁺ at extreme positive voltages.

In search for unknown blockers in the solution, we screened the effect of divalent cations in the cytosolic solution on channel activity. These experiments were motivated by three reasons: first, the decreasing lifetimes of channel closures in the mutant is reminiscent of a Ba²⁺ block in K⁺ channels (Neyton and Miller, 1988). Second, previous data have shown that divalent cations in the cytosolic solution cause a voltage-dependent decrease in open probability in the MthK channel (Thomson et al., 2014) with a nanomolar affinity for Ba²⁺ (Guo et al., 2014). Third, high-quality KCl salts contain, according to the manufacturer’s certificates of analysis, traces of alkaline earth metal ions; the Ba²⁺, Ca²⁺, and Mg²⁺ concentrations in a 100 mM KCl solution for example can be in the range of hundreds of nanomolar.

If the voltage-dependent closure of the mutant is caused by a divalent ion present as a contaminant in the solution, the mutant and the WT channel should differ significantly in their sensitivity to this ion. The data in Fig. 4 show that at a reference voltage of +120 mV only 100 μM Ba²⁺ and Sr²⁺ cause a severe block of both Kcv_NTS WT and Kcv_NTS S42T, different from the MthK channel (Thomson et al., 2014), and Ca²⁺ has at the same concentration neither an appreciable effect on the open probability of the WT nor on the mutant channel (Fig. 4, A and B). Scrutiny of the voltage-dependent inhibition by cytosolic Ba²⁺ (Ba²⁺_cyt) and cytosolic Sr²⁺ (Sr²⁺_cyt) at a lower concentration further reveals that the S42T mutant channel exhibits a higher sensitivity to both blocking ions compared with the WT channel (Fig. 4, C and D). This difference between the mutant and WT channels is particularly pronounced for Ba²⁺, while 1 μM Ba²⁺_cyt inhibits the WT channel at +160 mV by about 30%, the mutant shows inhibition of >50% already at +50 mV and of almost 100% at +160 mV (Fig. 4 D).

Sr²⁺_cyt induces in the WT and mutant channel short closing events (<100 ms) whose frequency increases with increasing Sr²⁺ concentration (Fig. 5 A). The K_I value of the Sr²⁺ block decreases for WT and S42T mutant almost exponentially from 1 mM at +60 mV by three orders of magnitude to 1 μM at +160 mV (Fig. 5 D); the K_I value of the mutant drops only at membrane voltages greater than +100 mV below the reference value of the WT channel. This minor difference in Sr²⁺ affinity cannot explain the decrease in P_O of the S42T mutant at positive voltages.

An additional argument against Sr²⁺ as blocking contaminant comes from a comparative analysis of the dwell times: the addition of Sr²⁺ to the cytosolic solution elicits only one additional population of block events (C2_Sr, Fig. 5 E), while the decrease in P_O of the S42T mutant can be attributed to the voltage-dependent appearance of two closed-time populations (C2 and C3) (Fig. 3). At first glance, the mean lifetime of C2_Sr falls within the range of the C2 population (Fig. 5 E); at moderate positive membrane potentials, C2 and C2_Sr are indistinguishable (Fig. 5 E). However, at high voltages C2_Sr is no longer identical to C2, and the closed time histograms at +160 mV and 1 μM Sr²⁺ can only be described in a satisfying way by adding a fourth exponential function (Fig. 5 E). If we assume that each closed-time population Cj represents a separate closed state that can be reached exclusively from the open state (for details see Materials and methods and Scheme 1), we can calculate the apparent rate constants (k_OCj) for the transitions from the open to the closed states. As expected for a concentration-dependent block, the apparent rate constant k_OC2Sr of the C2_Sr population increases linearly with [Sr²⁺]_cyt (Fig. 5 F). In contrast, the rate constant k_OC3 of population C3 is unaffected by an increase in [Sr²⁺]_cyt. However, the [Sr²⁺]_cyt-dependent increase in the number of C2_Sr events causes a decrease in the amplitudes of populations C2 and C3, indicating that Sr²⁺ and the unknown contaminant cannot block the pore simultaneously. The fact that C2_Sr is indeed not identical with C2 becomes once again apparent from the mean lifetimes; while τ_C2 of the closed time population C2 decreases from 41 ms at +80 mV to 6 ms at +160 mV (Fig. 3 D and dashed blue line in Fig. 5 G), the mean lifetime τ_Sr of Sr²⁺ block events displays the inverse voltage-dependence and increases from 5 ms at +80 mV to 50 ms at +160 mV (Fig. 5 G).

Analysis of the Ba²⁺ block shows that the S42T mutation significantly increases the affinity for Ba²⁺_cyt (Fig. 6, A–D). The current traces depicted in Fig. 6 A illustrate this impressively: while 1 μM Ba²⁺ has almost no effect on the open probability of the WT channel at +120 mV, the S42T mutant is almost completely blocked at the same concentration. Furthermore, the dose-response curves reveal that the S42T mutant is for voltages greater than +60 mV sensitive to [Ba²⁺]_cyt far below 1 µM (Fig. 6 C). This is also reflected in the calculated K_I values (Fig. 6 D). For the WT channel, the K_I value decreases from 22 μM at +20 mV to 1.9 μM at +160 mV. In the case of the S42T mutant, the K_I value decreases in the same voltage range by almost four orders of magnitude from 21 µM to 25 nM. These results strengthen the suspicion that the drop in P_O of Kcv_NTS S42T at positive voltages is caused by a Ba²⁺ contamination in the test solutions used. This hypothesis is further supported by the results of the dwell-time analyses (Fig. 6, E–G). Ba²⁺_cyt generates two closed-time populations (C2_Ba and C3_Ba), which coincide with C2 and C3 at all tested voltages (Fig. 6 E). The assumption that both populations are caused by Ba²⁺ can be demonstrated by calculating the apparent rate constants k_OC2Ba and k_OC3Ba: both increase linearly with [Ba²⁺]_cyt (Fig. 6 F) as expected for a bimolecular reaction. Moreover, the mean lifetimes of the populations C2_Ba and C3_Ba are identical to the mean lifetimes of C2 and C3 at all voltages tested (Fig. 6 G). These results strongly suggest that the voltage-dependent decrease in P_O of the channel mutant Kcv_NTS S42T is caused by traces of Ba²⁺ in the KCl solutions used.

The data advocate a scenario in which the S42T mutation increases the affinity of the channel for a block by Ba²⁺. The aforementioned variability in the voltage-dependent closure of the mutant (Fig. 2) may in that case only report different amounts of Ba²⁺ contaminations in the KCl salt between different lots. This predicts that a complete removal of Ba²⁺ from the buffer should eliminate the voltage-dependent closures in Kcv_NTS S42T. For a test of this prediction, we measured Kcv_NTS S42T activity before and after adding 200 µM of the macrocyclic polyether 18C6TA to the cytosolic solution. 18C6TA has a very high affinity to Ba²⁺ (K_D < 3.5 × 10⁻¹⁰ M, see Neyton, 1996) and was already used before for removing traces of Ba²⁺ contaminations from the buffer (Neyton, 1996; Diaz et al., 1996; Vergara et al., 1999). The single-channel traces (Fig. 7 A) and the corresponding P_O/V relations (Fig. 7 B) show that 18C6TA almost completely abolishes long-channel closures at positive voltages. The mutant channel acquires the same high P_O value as the WT channel. Analysis of the closed dwell times reveals that the increase in the open probability is due to a reduction in the occupation probabilities of populations C2 and C3 (Fig. 7, C and D). C3 is even no longer detectable at voltages less than +120 mV (Fig. 7 D). The finding that C2 does not decrease to the same extent as C3 is most likely due to the fact that the S42T mutant, like the WT channel (Rauh et al., 2017), has a Ba²⁺-independent closed-time population with a similar mean lifetime to C2_Ba. This will be discussed in detail below. The results of these experiments underscore that the voltage-dependent closures of the Kcv_NTS S42T mutant are not generated by a conventional gating mechanism; they are the result of a voltage-dependent block by Ba²⁺.

Previously, we found that not only the mutation S42T but also the mutations S42A and S42V result in the occurrence of voltage-dependent closures at positive voltages (Rauh et al., 2022). These closures can be suppressed by adding the crown ether 18C6TA to the cytosolic solution (Fig. S3), which clearly demonstrates that the voltage-dependent decrease in the open probability of the mutants S42A and S42V is also in these cases caused by traces of Ba²⁺ in the measurement solution. Interestingly, all these mutations also caused an increase in unitary conductance. To investigate the causal relationship between the two functional parameters, we substituted S42 with 13 amino acids of different sizes, hydrophobicity, and charge.

Functional reconstitution of these Kcv_NTS mutants reveals that position 42 at the C-terminal end of the pore-helix has a key role in modulating conductance and filter gating of this channel pore: all substitutions result in functional channels and in an amino acid-specific change in unitary conductance, fast gating at negative membrane voltages, and slow gating at negative and/or positive voltages (Fig. 9). Fig. 9 A shows representative single-channel traces of four mutants in which S42 was replaced by the polar amino acid threonine (S42T), the aliphatic amino acid leucine (S42L), and the negatively and positively charged amino acids glutamate (S42E) and lysine (S42K), respectively. At positive voltages, the substitution of S42 with glutamate (E) has a similar effect on unitary conductance and P_O as the mutation S42T (Fig. 9, B–D). But in contrast to S42T, S42E leads to a reduction of the apparent single-channel amplitude at negative voltages, presumably due to an acceleration of fast gating kinetics (Fig. 9, A and B). Interestingly, inserting the positively charged amino acid lysine has no effect on the open-channel conductivity at positive voltages (Fig. 9, A and B) but converts Kcv_NTS into a strong outward rectifier (Fig. 9 C). The mutant has an open probability of virtually 0 at negative voltages, which increases with a V_1/2 of +62.2 ± 5.5 mV and an apparent gating charge of 1.37 ± 0.16 e₀ to a maximal Po value of about 0.7 at high positive voltages (Fig. 9 C). This outward rectification is most likely not related to the positive charge of the lysine residue since the substitution of amino acid 42 with the aliphatic amino acid leucine (L) produces a similar, albeit somewhat weaker, voltage dependence with V_1/2 of +53.1 ± 8.4 mV and an apparent gating charge of 0.96 ± 0.33 e₀ (Fig. 9, A and C).

The experimentally determined open-channel amplitudes and open probabilities at −160 and +160 mV of all tested channel variants are summarized in Fig. 9, D and E, respectively. It is astonishing that even (apparently) small changes in the physicochemical properties of amino acid 42 can evoke large changes in the functional properties of the channel pore (e.g., S versus T or V versus I), whereas amino acids with very different properties, such as K and L, can give rise to very similar phenotypes. Thus, it can be concluded that Kcv_NTS is highly sensitive to mutations at position 42. Although a simple correlation between the physicochemical properties of the amino acid at position 42 and the electrophysiological properties of the mutant cannot be identified, we find a correlation between K⁺ conductance at positive membrane voltages and the sensitivity against cytosolic Ba²⁺. When the open-channel amplitude at +160 mV is plotted against the corresponding open probability (as a measure for the sensitivity to Ba²⁺), there is a strong negative correlation with a Pearson correlation coefficient of −0.79 (P = 0.001). This means that a mutation that increases the outward conductivity of the filter for K⁺ simultaneously decreases the permeability of the filter for Ba²⁺. In contrast, the effect of mutations on the K⁺ conductance at positive voltages only weakly correlates with the apparent open-channel amplitude at negative voltages (Fig. 9 G); the latter value is an indirect measure of the stability of the filter gate at negative voltages (Rauh et al., 2022). This difference in correlation suggests that the effects of a mutation at position 42 on conductivity and gating of the K⁺ channel pore at positive and negative membrane voltages are unrelated.

The miniature size Kcv channels represent, in their basic structural and functional properties, the pore module of all K⁺ channels. Among >100 known Kcv channel variants (Murry et al., 2020), Kcv_NTS is most useful for understanding structure/function correlates in the selectivity filter of K⁺ channels because this channel lacks a cytosolic gate and its gating is entirely dominated by structural rearrangements directly in or in the vicinity of the selectivity filter (Rauh et al., 2017). This unique combination of small size and functional reduction to a single defined gate makes Kcv_NTS a perfect model system for studying filter gating.

We had previously reported that mutations in the critical position 42 in Kcv_NTS, which is equivalent to a functional hot spot for filter gating in KcsA (E71), MthK (V55), and Kv channels (V370 in Kv1.2 [Chao et al., 2010]), alter key properties of this Kcv channel. In particular, the S42T mutant shows an increase in unitary conductance as well as a stabilization of the SF gate at hyperpolarizing voltages (Rauh et al., 2022). The same mutation also generates a voltage-dependent and K⁺-sensitive closing at depolarizing voltages. This phenomenon, which was further studied here, resembles at first glance SF-mediated C-type inactivation in other K⁺ channels such as KcsA or MthK. However, the key message from the present analysis is that this voltage-dependent phenomenon in Kcv_NTS S42T is not generated by an intrinsic gate. Different from the MthK channel where Ca²⁺ and Sr²⁺ augment C-type inactivation (Thomson et al., 2014), channel closures in Kcv_NTS at positive voltages are the consequence of a hypersensitivity of the mutant to Ba²⁺_cyt. We find that also Sr²⁺, which triggers in the MthK channel C-type inactivation (Thomson et al., 2014), causes in Kcv_NTS S42T voltage-dependent closure at positive voltages. But this mechanism can be excluded as an explanation for the voltage dependency of Kcv_NTS S42T in KCl buffer without additional divalent cations. First, the required concentrations of Sr²⁺ for this effect are much higher than the impurity of 80 ± 1 nM in a 100 mM KCl buffer as measured by ICP-MS. Also, the effect of Sr²⁺ is mechanistically different from the voltage-dependent gating of Kcv_NTS S42T in a KCl buffer. The data on Kcv_NTS S42T are hence best explained by a scenario in which nanomolar Ba²⁺ contaminations in the KCl solutions cause block events that follow a double-exponential probability density function. These block events can be significantly reduced by chelating Ba²⁺ contaminations from the solution with the Ba²⁺ selective crown ether 18C6TA. The interpretation of two distinct populations of closed times in the S42T mutant as a result of a block by Ba²⁺ traces in the KCl solutions is further supported by experiments in which the population of the very two closed dwell-time populations can be augmented by adding additional Ba²⁺ to the intracellular solution.

Such a voltage-dependent block of K⁺ channels by Ba²⁺ contaminations in the experimental solution has already been described for BK channels (Neyton, 1996; Diaz et al., 1996). Also, long-lived closed events occurring at hyperpolarizing voltages in Kir1.1 could be causally related to a Ba²⁺ block arising from contaminations in the extracellular solution (Choe et al., 2001). Considering that any experimental KCl solution contains in the absence of an appropriate chelator unavoidable traces of other ions, it must be assumed that such distinct channel closures caused by blocking events might have also been interpreted in other studies as the consequence of intrinsic gating.

The filter sequence of Kcv channels as well as their geometry is very similar to other generic K⁺ channels like KcsA or MthK (Fig. 1 A and S1 A). This implies that ion permeation as well as the mechanism of channel block by ions in the pore should be like that in other K⁺ channels. This assumption is in good agreement with experimental data showing that mutations in a position that forms the binding site for the inner filter ion lowers the affinity for the Ba²⁺ block not only in a Kir channel but also in a Kcv channel (Chatelain et al., 2009). Also, from a mechanistic point of view, the Ba²⁺ block of the Kcv_NTS channel shares similarities with many other K⁺ channels. Like in the case of the well-studied BK, KcsA, or MthK channels, the Ba²⁺ block of Kcv_NTS and its S42T mutant is very asymmetrical (Guo et al., 2014; Vergara and Latorre, 1983; Piasta et al., 2011). While an extracellular concentration of 100 μM Ba²⁺ reduces the open probability in Kcv_NTS and its S42T mutant by ∼50% (Rauh et al., 2022), the same concentration nearly abolishes channel activity when provided from the cytosolic side (Fig. 4).

Our data suggest a scenario in which the S42T mutation at the C-terminal end of the pore-helix augments the Ba²⁺ sensitivity of the SF in an asymmetrical manner. While the affinity to external Ba²⁺ at negative voltages remains unaffected, the sensitivity to Ba²⁺_cyt at positive voltages increases by several orders of magnitudes giving rise to a Ba²⁺ block at nanomolar concentrations. This impact on the structural properties of the SF must be highly selective because it does not affect the sensitivity of the channel to Mg²⁺ and Ca²⁺. While the mutation also slightly increases the sensitivity to Sr²⁺, the data underpin that the mode of blocking between the two similar divalent cations is fundamentally different.

A detailed dose–response analysis reveals that Ba²⁺ blocking events in the S42T mutant are caused by the binding of a single Ba²⁺ ion inside the pore. This is consistent with findings for other K⁺ channels (e.g., Piasta et al., 2011; Guo et al., 2014; Choe et al., 2001). Since the dwell-time analyses revealed that Ba²⁺ generates two well-distinguishable populations of closed dwell times, it must be assumed that the SF harbors two Ba²⁺ binding sites, b₁ and b₂, that cannot be occupied simultaneously (Fig. 10 B). The double-exponential distribution of Ba²⁺ block times is then the result of hidden transitions of the blocking ions between these two binding sites before dissociating from the pore toward the intracellular or extracellular space. A two-state sequential Ba²⁺ blocking model has already been proposed for KcsA based on single-channel recordings (Piasta et al., 2011).

In line with this model, we assume that the closed dwell-time populations C2 and C3 of Kcv_NTS S42T in 100 mM KCl solution are caused by nanomolar Ba²⁺ impurities in the cytosolic solution and that these block events can be explained by a two-state sequential blocking model. This predicts that the reduction of the free cytosolic Ba²⁺ concentration by the crown ether 18C6TA abolishes these components in the closed dwell-time histograms. This effect was indeed observed for C3, whereas the C2 component was only slightly reduced. This result can be explained by the existence of an intrinsic closed state C2* (Fig. 10 A), which happens to have a similar mean lifetime as the Ba²⁺_cyt-induced block-time population C2_Ba. Dwell-time analyses of the single-channel traces recorded in the presence of 200 µM cytosolic 18C6TA support this assumption: in contrast to the mean lifetime of C2_Ba, the mean lifetime of C2* is almost voltage-independent (Fig. 10 C). The observation that the mean lifetimes of the Ba²⁺-induced closed-time population C2_Ba coincide for all tested voltages with the mean lifetimes of C2 (Fig. 6 G) further suggests that the C2 events occurring in standard KCl solution are essentially caused by Ba²⁺ block events masking the activity of the intrinsic gate.

To further test the accuracy of the two-state blocking model, we analyzed the Ba²⁺ dependency of C2_Ba and C3_Ba. Since the transitions between the blocked states B₁ and B₂ (rate constants c and d in Fig. 10 B) and the dissociation into the intracellular and extracellular space (rate constants b and z in Fig. 10 B) are according to the model [Ba²⁺]-independent, the renormalized amplitudes a_f of the fast component C2_Ba and a_s of the slow component C3_Ba should be [Ba²⁺]-independent. The renormalized amplitudes a_f and a_s can be calculated from Eqs. 3a and 3b given in Piasta et al. (2011) from the rate constants b, c, d, and z of the blocking model (Fig. 10 B) and the mean lifetimes τ_f (τ_C2Ba) and τ_s (τ_C3Ba) of the fast and slow components, respectively. Renormalization (i.e., mathematical removal of the fast component C1) was done according to Eqs. 14a and 14b.

To test this model prediction we calculated a_f and a_s from dwell-time data as shown in Fig. 6 for different voltages and cytosolic Ba²⁺ concentrations. The results shown in Fig. 10 D indeed confirm the model prediction in that the renormalized amplitudes are independent of the Ba²⁺ concentration. A Ba²⁺ dependency would have been expected if C2_Ba and C3_Ba were caused by two independent processes.

The equivalent model to that in Fig. 10 B for the Ba²⁺ block of KcsA outward current (Piasta et al., 2011) suggests that the S42T mutation may have increased in Kcv_NTS the Ba²⁺ affinity of the SF by creating a deeper energy well for Ba²⁺ at one of these two binding sites. This model predicts that the same mutation should also augment the sensitivity for the Ba²⁺ block from the extracellular solution. However, since previous experiments had already shown that the S42T mutant does not increase the sensitivity of the channel to external Ba²⁺ (Rauh et al., 2022), this model can be refused. On the contrary, the S42T mutation decreases the binding rate for extracellular Ba²⁺ at moderate negative voltages (Rauh et al., 2022). This supports an alternative model in which the increased asymmetrical Ba²⁺ block is explained by an elevated energy barrier for the release from the outermost binding site into the external solution (rate constant z in Fig. 10 B). In the context of the debate on whether K⁺ channels—including Kcv_NTS—have one or two binding sites, the experimental data can also be explained by a single Ba²⁺ binding site in which the energy barrier for blocker release fluctuates between two states with a low and high barrier. The S42T mutation may in this case augment the propensity for the high energy barrier state. The question of a suitable blocking model and the structural alterations underlying the hypersensitivity of Kcv_NTS S42T to Ba²⁺_cyt at positive voltages will be addressed in a separate study.

All K⁺ channels have the same overall architecture in which the selectivity filter is anchored to the pore helix. A critical position in this arrangement is E71 in KcsA and the respective amino acids are in the equivalent position in other channels. This can be a valine in Shaker and in MthK, but also a serine in Kcv channels. Our data and those from others impressively show that any kind of mutation, even the very conservative exchange of Ser for Thr, has profound impacts on functional properties of the channel in which the mutation was made. While the original data on channel gating with the E71A mutation in KcsA were interpreted in the context of an essential hydrogen bond, which anchors the filter to the pore helix, it was already obvious from other mutations in this work, that the structure/function correlates must be much more complex and include other types of interactions. The high degree of sensitivity of this site to mutations and the functional diversity with respect to unitary conductance, gating, and blocker sensitivity, which can be generated by single amino acid substitutions in different K⁺ channels (e.g., for Kcv, Fig. 9 in this study or KscA, Fig. 2 in Chakrapani et al. [2011]), suggest that this site creates together with its interaction partners a central modulator for a fine-tuning of selectivity filter properties. The amino acid diversity of this site in different types of K⁺ channels (Prole and Marrion, 2012) presumably determines the immense functional diversity of different native K⁺ channels. Because of this sensitive anchoring of the non-flexible filter domain to the pore helix, any small conformational change in this area can result in dramatic effects on the filter and filter-related properties such as conductance, gating, and voltage dependency.

The present data also offer conceptual insights into the mechanism of filter gating in that they advocate a transient channel block as a form of gating. Our results suggest a scenario in which minute alterations in the filter structure, which are here introduced by a mutation in the critical connection between the pore helix and filter, can result in extended dwell times of Ba²⁺ in filter binding sites. This voltage-dependent filter block appears then on the functional level as a form of channel gating. Based on data with the ROMK2 channel it is reasonable to speculate that this type of filter block is not restricted to Ba²⁺; also, the permeant K⁺ ion itself may remain at certain filter configurations for longer times in its binding sites giving rise to distinct gating events (Choe et al, 1998). In fact, the short closures of the Kcv_NTS S42T channel at a positive voltage in the presence of crown ether (Fig. 7 A) might reflect such K⁺ block-induced gating. The fact that the equivalent position S42 in Kcv_NTS has also in other K⁺ channels critical importance for filter gating lets us speculate, that this kind of gating may not only be introduced by mutations. Also, external cues, which modulate, like in K2P channels, filter gating, may in this way generate subtle changes in the filter with the effect of introducing conformations in which K⁺ or other ions that can penetrate the SF bind and transiently block the ion flow.

The data are directly available from the corresponding author upon reasonable request by email.

Crina M. Nimigean served as editor.

We thank Mrs. Stefanie Schmidt and Prof. Christoph Schüth (Inst. Hydrology, Applied Geosciences, TU Darmstadt) for the ICP-MS measurements.

The work was funded in part by the Deutsche Forschungsgemeinschaft (to G. Thiel; TH558/34-1).

Open Access funding provided by Universitäts- und Landesbibliothek Darmstadt.

Author contributions: Conceptualization: O. Rauh. Data curation: O. Rauh. Formal analysis: N. Tewes, B. Kubitzki, F. Bytyqi, N. Metko, S. Mach, and O. Rauh. Funding acquisition: G. Thiel. Investigation: N. Tewes, B. Kubitzki, F. Bytyqi, N. Metko, S. Mach, and O. Rauh. Methodology: O. Rauh. Project administration: G. Thiel and O. Rauh. Resources: NG. Thiel. Software: O. Rauh. Supervision: G. Thiel and O. Rauh. Validation: O. Rauh. Visualization: O. Rauh. Writing—original draft: G. Thiel, O. Rauh. Writing—review & editing: N. Tewes, B. Kubitzki, G. Thiel, and O. Rauh.