Oxidation modulates LINGO2-induced inactivation of large conductance, Ca2+-activated potassium channels

Ca2+ and voltage-activated K+ (BK) channels are ubiquitous ion channels that can be modulated by accessory proteins, including β, γ, and LINGO1 BK subunits. In this study, we utilized a combination of site-directed mutagenesis, patch clamp electrophysiology, and molecular modeling to investigate if the biophysical properties of BK currents were affected by coexpression of LINGO2 and to examine how they are regulated by oxidation. We demonstrate that LINGO2 is a regulator of BK channels, since its coexpression with BK channels yields rapid inactivating currents, the activation of which is shifted ∼−30 mV compared to that of BKα currents. Furthermore, we show the oxidation of BK:LINGO2 currents (by exposure to epifluorescence illumination or chloramine-T) abolished inactivation. The effect of illumination depended on the presence of GFP, suggesting that it released free radicals which oxidized cysteine or methionine residues. In addition, the oxidation effects were resistant to treatment with the cysteine-specific reducing agent DTT, suggesting that methionine rather than cysteine residues may be involved. Our data with synthetic LINGO2 tail peptides further demonstrate that the rate of inactivation was slowed when residues M603 or M605 were oxidized, and practically abolished when both were oxidized. Taken together, these data demonstrate that both methionine residues in the LINGO2 tail mediate the effect of oxidation on BK:LINGO2 channels. Our molecular modeling suggests that methionine oxidation reduces the lipophilicity of the tail, thus preventing it from occluding the pore of the BK channel.

Ca 2+ and voltage-activated K + (BK) channels are ubiquitous ion channels that can be modulated by accessory proteins, including β, γ, and LINGO1 BK subunits. In this study, we utilized a combination of site-directed mutagenesis, patch clamp electrophysiology, and molecular modeling to investigate if the biophysical properties of BK currents were affected by coexpression of LINGO2 and to examine how they are regulated by oxidation. We demonstrate that LINGO2 is a regulator of BK channels, since its coexpression with BK channels yields rapid inactivating currents, the activation of which is shifted −30 mV compared to that of BKα currents. Furthermore, we show the oxidation of BK:LINGO2 currents (by exposure to epifluorescence illumination or chloramine-T) abolished inactivation. The effect of illumination depended on the presence of GFP, suggesting that it released free radicals which oxidized cysteine or methionine residues. In addition, the oxidation effects were resistant to treatment with the cysteine-specific reducing agent DTT, suggesting that methionine rather than cysteine residues may be involved. Our data with synthetic LINGO2 tail peptides further demonstrate that the rate of inactivation was slowed when residues M603 or M605 were oxidized, and practically abolished when both were oxidized. Taken together, these data demonstrate that both methionine residues in the LINGO2 tail mediate the effect of oxidation on BK:LINGO2 channels. Our molecular modeling suggests that methionine oxidation reduces the lipophilicity of the tail, thus preventing it from occluding the pore of the BK channel.
Large conductance (G), Ca 2+ , and voltage-activated potassium channels (BK, MaxiK, or KCa1.1) are allosterically activated by an increase in intracellular calcium and depolarization (1,2). These channels play crucial roles in different physiological functions such as neuronal excitability, neurotransmitter release, and smooth muscle contraction (3,4). The pore of the BK channel is formed by the tetrameric assembly of α subunits, and its structure has recently been solved using cryo-EM (5,6). The differing physiological roles of BK channels can be mediated through their association with the regulatory β 1-4 , γ 1-4 , and LINGO1 subunits. Structurally, the β-subunits have two transmembrane helices which are connected extracellularly via a large loop and have short intracellular N & C termini tails (7,8). In contrast, the γ-subunits have a single transmembrane helix with a large extracellular N terminal domain and an intracellular C terminal tail (9)(10)(11). Functionally, the β subunits alter the Ca 2+ sensitivity of the BK channels to different extents (12,13), whereas coexpression of either β 2 or most β 3 splice variants also induces the inactivation of BKα subunits (14)(15)(16). The γ-subunits shift the voltage-dependent activation of BK channels to negative potentials by varying amounts in the absence of Ca 2+ . Thus, the γ 1 subunit shifted activation V 1/2 by -140 mV, whereas γ 2 , γ 3 , and γ 4 shifted the V 1/2 by −100 mV, −50 mV, and −20 mV respectively (10).
Recently, LINGO1 has been shown to be a new auxiliary subunit of BK channels, which reduced BK channel plasmalemmal expression, shifted the activation V 1/2 -60 mV and induced rapid inactivation of the BK channels (11). LINGO1, like the γ 1-4 subunits, is a leucine-rich repeat-containing protein with a single transmembrane helix and a short intracellular tail. However, in contrast to the γ subunits, LINGO1 contains 12 rather than six LRR domains and possesses an IgI1 domain that is absent in the γ subunits (17,18). Furthermore, LINGO1 has three homologues named LINGO2, LINGO3, and LINGO4 which share 61%, 56%, and 44% sequence similarity with LINGO-1, respectively (19,20). However, no study has investigated if any of these homologues also modulate BK channels. In this study, we examined the effects of coexpressing BK and LINGO2 in human embryonic kidney (HEK) cells. The data presented demonstrate that LINGO2 induces the inactivation of BK channels, and this effect can be abolished by oxidation of the LINGO2 subunit.

Results
Currents were recorded from excised inside out patches of membrane from HEK cells transfected with BKα complementary DNA (cDNA). As shown in Figure 1, A-C depolarizations from −100 mV to +200 mV in 20 mV increments from a Panels A-C show typical currents recorded from HEK cell patches expressing BKα subunits. Currents were elicited from −100 mV to +200 mV in 20 mV increments from a −60 mV holding potential. Patches were pre-pulsed to −100 mV before evoking currents, and tail currents were elicited by repolarizing back to −80 mV. The red dashed lines in this and other current traces represent zero current. Panel D shows a summary, GV curve constructed and fitted with the Boltzmann equation (solid line, n = 7). With increased Ca 2+ concentrations, summary activation V 1/2 was shifted towards more negative potentials (100 nM, 1 μM, and 10 μM Ca 2+ -white, grey, and black circles, respectively). Panels E-G show that coexpression of LINGO2 with BKα generated rapidly inactivated currents in 100 nM and 1 μM Ca 2+ concentrations, but currents were practically abolished in 10 μM Ca 2+ . The same voltage clamp protocol was used to record currents as described for holding potential of −100 mV evoked noisy, sustained currents. These activated at more negative potentials as the [Ca 2+ ] at the cytosolic surface of each patch was increased. The effect of Ca 2+ on the half maximal voltage of activation (V 1/2 ) is summarized in the activation curves plotted in Figure 1D (n = 7). However, when BKα and LINGO2 cDNA were cotransfected and the cytosolic surface of patches was bathed with 100 nM Ca 2+ (Fig. 1E), the currents activated rapidly at potentials positive to +60 mV and inactivated completely over the course of several milliseconds. In eight experiments summarized in Figure 1H (open symbols), the mean V 1/2 in 100 nM Ca 2+ was +130 ± 6 mV, compared to (+155 ± 8 mV) recorded in patches expressing only BKα subunits (n = 7, p < 0.0001). When [Ca 2+ ] was increased to 1 μM (Fig. 1F), the currents activated at more negative potentials (V 1/2 =53 ± 4 mV, Fig. 1H, grey circles, n = 8), and inactivated more rapidly (grey symbols, Fig. 1L), compared to the currents recorded in 100 nM Ca 2+ (open symbols, Fig. 1L). In the presence of 10 μM Ca 2+ (Fig. 1G), only small inward currents were recorded at negative potentials, and outward currents were practically abolished, as demonstrated previously with BK:LINGO1 currents (11). The recovery from inactivation was rapid when patches were stepped to −120 mV for increasing durations, where in six experiments the τ recovery was 0.83 ± 0.28 ms (Fig. S1). We attempted to assess plasmalemmal expression of BK:LINGO2 channels by measuring BK current amplitude in a series of patches obtained from 5 MΩ pipettes and comparing these with currents recorded under identical recording conditions (100 nM Ca 2+ ) from cells transfected with either BKα alone or in combination with LINGO1 cDNA. As Fig. S2A suggests, patches from cells transfected with BKα cDNA and depolarized to +160 mV in 100 nM Ca 2+ produced large, noisy, sustained currents. In contrast, although both BK:LINGO1 (Fig. S2B)-and BK:LINGO2 (Fig. S2C)-containing patches showed rapid inactivation, as previously shown, the amplitude of the BK:LINGO1 currents was much smaller (11) than that recorded from patches obtained from cells transfected with BKα alone or BK and LINGO2 cDNA. This was reflected in the summary data shown in Fig. S2D where the mean peak current in BKα patches (3313 ± 3260 pA, n = 43) was significantly larger than that recorded from BK:LINGO1 patches (166 ± 215 pA, n = 29), but not significantly different from that recorded from BK:LINGO2 patches (2081 ± 2605 pA, n = 43, ANOVA).
When Flag-tagged BKα proteins were coimmunoprecipitated from HEK cell lysates, the HA-tagged LINGO2 protein of the expected size was immunoblotted in three separate experiments (Fig. S3), suggesting that both proteins closely associate when coexpressed in HEK cells.
To characterize the steady-state voltage dependence of the inactivation in BK:LINGO2 currents, we applied a double pulse protocol (shown inset in Fig. 1I), where the patches were subjected to a series of conditioning pre-pulses (from −120 mV to +120 mV in 20 mV increments for 200 ms) prior to being depolarized to a test potential of +140 mV for 25 ms. The peak amplitude of these currents was unaltered when conditioning potentials negative to 0 mV were applied in 100 nM Ca 2+ (Fig. 1I). However, with more positive conditioning potentials, the amplitude of the current evoked by a test step to +140 mV began to decline. For example, when a conditioning potential of +40 mV was applied, peak current amplitude was reduced by 50% and was abolished when conditioning potentials were more positive than +80 mV. The rate of inactivation increased and its shifted to more negative potentials when [Ca 2+ ] was elevated to 1 μM, as shown in Figure 1J. In the experiments summarized in Figure 1K the V 1/2 of inactivation was 35 ± 6 mV (n = 9) and −39 ± 6 mV (n = 8), in the presence of 100 nM Ca 2+ (open symbols) and 1 μM Ca 2+ (black symbols), respectively (p < 0.0001).
Interestingly, we noted that the inactivation observed in the continued presence of 100 nM Ca 2+ appeared to wane over the course of several minutes in some patches. We investigated this further and found that it was caused by the epifluorescent illumination being left on during some recordings. Figure 2A shows typical currents recorded from an excised patch before and during two periods of illumination, applied for 2 min, as denoted by the blue bars in Figure 2B. The experimental protocol consisted of a 3 min control period (minutes 1-3) before the first period of illumination was applied (minutes 4 and 5). The second period of illumination was applied during minutes 8 and 9 Figure 2B shows summary data for peak (black symbols) and sustained (blue symbols) current amplitude, measured during the first and last 5 ms of the depolarization to +160 mV, respectively. In the absence of illumination, there was no evidence of inactivation waning. However, when the epifluorescence was switched on for the two periods denoted by the blue bars in Figure 2B, it was clear that illumination caused a rapid and irreversible removal of inactivation by the end of the second period of illumination. For example, before illumination, the sustained current was 7 ± 5% of the peak current (measured in the first 5 milliseconds of the depolarization) and this increased to 97 ± 10% after a total of 4 min of illumination, (p < 0.05, n = 6). The peak current was also increased significantly (by 23 ± 10%, p < 0.05). To test if the effects on LINGO2 were dependent on the presence of GFP, we repeated the illumination experiments in patches from cells in which the GFP plasmid was omitted at transfection (Fig. 2, C and D). Under these conditions, epifluorescent illumination for 4 min failed to significantly increase the sustained current (3 ± 2% versus 5 ± 4%) in 5 cells. Although the sustained current increased to 10 ± 7% after a total of 8 min illumination, this effect failed to reach statistical significance (p = 0.06). These data suggested that removal of GFP practically abolished the rundown caused by epifluorescent illumination.
We reasoned that illumination could lead to the production of free radicals which may oxidize residues in either the BK channels or the LINGO2 subunits. Consequently, we next tested if the effects of illumination were mimicked by the oxidizing agent, Chloramine T (Ch-T, 200 μM). As Figure 3, A and B suggest, Ch-T caused rapid abolition of inactivation, such that the sustained current amplitude increased from 11 ± 8% of the peak current to 59 ± 39% (p < 0.05) and 130 ± 19% (p < 0.05) after 60 s and 120 s of its application, respectively. The amplitude of the peak current was also significantly increased by 35 ± 31% compared to control after 120 s in the EDITORS' PICK: Redox modulation of BK:LINGO2 channels presence of Ch-T (p < 0.05, Wilcoxin signed-rank test). The effects of Ch-T on peak and sustained current were maintained for the duration of the recording, despite the removal of the oxidizing agent. When we examined the effect of Ch-T on BK currents in the absence of LINGO2, we found that its application for up to 210 s failed to significantly increase current amplitude (21 ± 27%) as shown in Figure 3D (n = 4-5, p = 0.43).
The effects of Ch-T were not prevented by pretreating patches with the reducing agent DTT (100 μM, Fig. 4, A and B), nor were they reversed by coapplication of DTT and Ch-T ( Fig. 4, C and D) suggesting that the effects of Ch-T were not mediated via oxidation of cysteine residues. Consequently, we tested the possibility that other oxidizable residues, such as methionine, were involved.
LINGO2 contains two methionine residues in its distal Cterminus (marked with asterisks in Fig. 6A), and we examined if their oxidation abolished inactivation. We have previously shown that the last eight residues of LINGO1 were sufficient to mimic the inactivation induced by the full-length LINGO1 protein (11). Therefore, we first synthesized a series of acylated amidated LINGO2 tail peptides comprised of the last eight residues of the LINGO2 protein. As Figure 5, A and B suggest, the Ac-RRFNMKMI-NH2 peptide caused inactivation, the rate of which was concentration-dependent, and the summary data in Figure 5C shows the mean reduction in current amplitude measured in the first (white circles) and last 5 ms (black circles) of the depolarizing pulse. The concentration of the peptide that produced half maximal block in current amplitude recorded in the last 5 ms of a pulse (IC 50Last5ms ) to +160 mV was 260 ± 78 nM (95% Confidence Intervals (CI) 223-303 nM, n = 6). Similar effects were observed with the RRFNMKM(O)I peptide, in which the methionine equivalent to M605 in the full-length peptide was oxidized. However, the Figure 2. Epifluorescent illumination abolishes the inactivation of BK:LINGO2 currents and is dependent on GFP. A, BK:LINGO2 currents were elicited by stepping to +160 mV from −100 mV, and tail currents were generated by repolarizing back to −80 mV. The black and blue symbols denote where the peak and sustained current were measured, respectively, to produce the plot shown in Panels B and D. Panel B shows the summary of six experiments where normalized peak (black symbols) and sustained (blue symbols) current amplitudes were measured during the first and last 5 ms of the depolarization and compared using the Wilcoxon signed-rank test. Patches were illuminated with epifluorescence twice, consecutively for 2 min as denoted by the blue bars. At the end of the 4 min of epifluorescence illumination, inactivation was completely abolished in BK:LINGO2 currents (p < 0.05 compared to control, Wilcoxon signed-rank test). Panel C shows currents evoked (using the same protocol as Panel A) from a patch taken from cells in which the GFP plasmid was excluded. Note that illumination failed to remove inactivation when GFP was absent, even when applied continuously for up to 8 min (p = 0.06, Wilcoxon signed-rank test). Panel D shows summary data for this set of experiments in which the normalized peak (black circles) and sustained (blue diamonds) current amplitudes were plotted (n = 4-5), where it is clear that the effects of illumination on inactivation were practically abolished. BK, voltage-activated potassium.
rate of inactivation was slower compared to that observed with the WT peptide (compare currents in 1 μM of each peptide in Fig. 5, A and C), and the IC 50Last5ms was increased (2.3 ± 0.7 μM, 95% CI, 2.0-2.7 μM, p < 0.001, n = 5, extra sum of squares F-test), suggesting that its affinity for BK channels was reduced. Application of the RRFNM(O)KMI peptide, in which the methionine equivalent to M603 was oxidized, had much less effect, as illustrated in Figure 5, E and F. Thus, the inactivation in the presence of 1 μM of this peptide was slower than that of the WT peptide and the IC 50Last5ms was increased 27 fold (to 6.9 ± 0.5 μM, 95% CI 6.2 μM -7.8 μM) compared to the WT peptide (n = 6, p < 0.001, extra sum of squares Ftest). Finally, we applied a peptide in which both methionine residues were oxidized (RRFNM(O)KM(O)I) and found that inactivation was practically abolished at all concentrations tested (Fig. 5, G and H). The IC 50Last5ms was increased 50 fold to 13.9 ± 1.1 μM (95% CI 11 μM -17 μM)) compared to the WT tail peptide (p < 0.001, n = 7, extra sum of squares F-test).
We next tested if substitution of each methionine in the Cterminus (see Fig. 6A) with leucine could protect the full-length LINGO2 from oxidation induced by Ch-T. We chose leucine because its side chain is similar in size and hydrophobicity but is more resistant to oxidation than the sulphur-containing side chain of methionine. Figure 6B shows a typical current from an experiment in which the M605L mutation was introduced into the full-length LINGO2 protein and cotransfected with BK cDNA. The currents evoked by a step to +160 mV retained the typical inactivation characteristics of WT LINGO2 ( Fig. 2A), but the inactivation was affected by Ch-T. Thus, after 60 s application of Ch-T to the M605L mutant, the sustained current was increased from 6 ± 4% to 15 ± 14% of the peak current amplitude (Fig. 6C, n = 10, p < 0.01, Wilcoxin signed-rank test). After an additional 90 s exposure to the oxidizing agent, sustained current amplitude increased further to 47 ± 25% (p < 0.01, Wilcoxin signed-rank test), but this was 50% less effective than it was on WT BK:LINGO2 currents (p < 0.001, Mann-Whitney test). As Figure 6D suggests, the M603L mutant was even more resistant to the effects of Ch-T treatment, as evidenced by the relatively small increase in the amplitude of the sustained currents, following its application. Summary data in Figure 6E show that the mean sustained current under control conditions still significantly increased from 10 ± 9% to 17 ± 9% and 20 ± 12% in the presence of Ch-T for 60 s and 150 s, respectively (p < 0.05, n = 9, Friedman's test), but this effect was significantly smaller than that observed in the LINGO2 M605L mutant (p < 0.05, Mann-Whitney test). As shown in Fig. S5, the inactivation of this mutant was also much more resistant to the effect of epifluorescent illumination (n = 6), compared to the WT BK:LINGO2 currents shown in Figure 2A. We next examined the effects of Ch-T on BK:LINGO2 currents recorded at different voltages and in varying [Ca 2+ ] i . Figure 7A shows a family of currents evoked by stepping from −100 mV to +200 mV in 20 mV increments for 50 ms from a holding potential of −100 mV and recorded with 100 nM Ca 2+ bathing the cytosolic surface of the patch. Figure 7B shows the currents recorded from the same patch after Ch-T had its maximal effects and demonstrates that inactivation was practically abolished, as evidenced by the amplitude of the currents at the end of the depolarizing pulses. The summary conductance versus voltage (GV) curves in Figure 7C were constructed from seven similar experiments conducted before (black symbols) and after applying Ch-T (blue symbols) for 2 min, which shifted the half maximal activation voltage from 128 ± 3 mV to 86 ± 11 mV. Figure 7D shows currents recorded from a different patch, which was bathed in 10 μM Ca 2+ at its cytosolic surface and stepped from −100 mV to +200 mV in 20 mV steps. Under these conditions, inward currents were recorded at negative potentials and small, inactivating, outward currents were recorded at more positive potentials. However, after treatment with 200 μM Ch-T (Fig. 7E), the current amplitude was increased at all voltages recorded, and inactivation was practically abolished. Figure 7F shows summary GV data from six similar experiments recorded in 10 μM Ca 2+ in the absence (black symbols) and presence (blue symbols) of Ch-T and illustrated that oxidation led to large increases in G across the voltage range.
To examine if the steady-state voltage-dependent inactivation was altered, we used a double pulse protocol (inset Figure S-4) before and after removal of inactivation with Ch-T. As Fig. S4, B and C demonstrate, no time-or voltagedependent inactivation was observed after Ch-T treatment.
In an attempt to explain how oxidation might lead to the removal of inactivation and in the absence of published BK:LINGO2 structures, we performed docking studies of LINGO2 to the BK channel, as shown in Figure 8. Initially, the 8-residue peptide and the transmembrane helix of LINGO2 were docked to the BK channel separately. From the flexible docking of the peptide, the final peptide docking pose in the BK channel was chosen to be able to link to the LINGO2 helix. The LINGO2 helix model was taken from the AlphaFold database (https://alphafold.ebi.ac.uk/), and its position in the channel was identified by rigid-body docking. LINGO2 was docked similarly to the auxiliary β4 protein in the crystal complex with the BK channel (21). Next, both parts of LINGO2 were linked by a flexible loop, which was predicted by a loop prediction approach with consideration of its interactions with the BK channel and is shown in Figure 8A. As it suggests, the tail of LINGO2 gained access to the BK channel pore through a side portal between the transmembrane domain and the intracellular RCK domains of the BK channel. The LINGO2 tail docked centrally in the BK channel pore as shown in Figure 8B. The optimized BK:LINGO2 channel complex showed that the positively charged residues of the LINGO2 C-terminal peptide formed ionic interactions with the E324 residues of the BK subunits, whereas hydrophobic residues of the peptide sat in the hydrophobic pockets of the pore. Interestingly, the docking suggested that while M605 was buried deep in a hydrophobic pocket of the pore, M603 had a more solvent-exposed position (Fig. 8C). These configurations were also preserved when the M603L and M605L mutants were modeled (Fig. S6). A side-on view of the optimized BK:LINGO2 complex is presented in Figure 8D and illustrates that the eight terminal residues of the LINGO2 tail, shown in green sticks, penetrated deep into the pore and the I606 residue sat just below the selectivity filter. This could help provide a structural explanation of how the C-terminus of LINGO2 occluded the BK channel pore, but structural studies will be required to confirm this. Interestingly, as the blue sticks shown in Figure 8E suggest, when the M603 oxidized form of the peptide was docked, it failed to penetrate as deeply into the channel pore.

Discussion
The results of the current study demonstrate that when BK channels were coexpressed with LINGO2, the resultant currents inactivated rapidly and activated at more negative potentials than BKα subunits, similar to the effects of LINGO1 on BK channels (11). This is not surprising as LINGO1 and LINGO2 look qualitatively similar, sharing 61% identity and 77% homology (22). However, there are a number of distinguishing features between BK:LINGO1 and BK:LINGO2 currents. Firstly, coexpression of LINGO2 with BK channels did not significantly reduce the amplitude of the inactivating currents (Fig. S2), suggesting that this subunit does not alter plasmalemmal expression of BK channels, in contrast to that observed previously with LINGO1 (11). Secondly, the activation V 1/2 for BK:LINGO2 currents was 130 mV in 100 nM Ca 2+ , compared to 100 mV for BK:LINGO1 under identical recording conditions (11). Thirdly, BK:LINGO2 currents inactivated more slowly than BK:LINGO1 currents (inactivation time constant at +100 mV of 10 ms for BK:LINGO2 and 5 ms for BK:LINGO1). Similarly at +200 mV, LINGO2 inactivated with a τ of 4 ms compared to 2 ms observed in BK:LINGO1 channels (11). Fourthly, when steady-state inactivation was investigated with a double pulse protocol, we found that BK:LINGO2 channels half maximally inactivated at voltages 30 mV more positive than those previously recorded in BK:LINGO1 channels. Thus, in the present study we found that the inactivation V 1/2 for BK:LINGO2 was +35 mV in 100 nM Ca 2+ and this EDITORS' PICK: Redox modulation of BK:LINGO2 channels shifted to -40 mV when the cytosolic [Ca 2+ ] was increased to 1 μM. Given that the inactivation observed with LINGO1 and LINGO2 appears to be due to open channel block, we speculate that the differences in the above biophysical properties are due to the differences in activation V 1/2 observed between BK:LINGO1 and BK:LINGO2. Li et al. (23) demonstrated that a phenylalanine residue (F273) in the transmembrane domain of the leucine-rich repeat-containing protein, γ3, was responsible for the shift in BK activation V 1/2 . Interestingly, LINGO1 has a phenylalanine (the red box in Fig. 6A) at the equivalent position to F273 of the γ3 subunit, whereas LINGO2 does not. This may at least partly explain why BK:LINGO2 channels do not activate as negatively as BK:LINGO1 channels. Another interesting observation is that LINGO2 has three positively charged residues in the juxta-membrane region, whereas LINGO1 only has two. Li et al. (23), demonstrated that positively charged juxta-membrane residues contributed to the shift in activation V 1/2 observed with γ3 subunits; therefore, it will be interesting to examine if this accounts for the differences in V 1/2 between BK:LINGO1 and BK:LINGO2 currents.
Another important observation on BK:LINGO2 currents was that the inactivation waned rapidly in patches exposed to epifluorescent illumination. The ability of illumination to remove inactivation was practically abolished when GFP was omitted from the transfection. A number of studies have demonstrated that GFP can alter the redox state of cells (24,25), and its illumination has been shown to generate reactive oxygen species (26,27), which can oxidize proteins, including ion channels, and modulate their function (28)(29)(30)(31)(32)(33)(34)(35)(36). For example, Pooler (28) demonstrated that photo-oxidant stress slowed inactivation of Na + currents, and Ciorba et al. (31) showed that Ch-T reduced the inactivation of Shaker C/B K + channels but failed to alter the inactivation of the Shaker B splice variant. Ciorba et al. (31) also demonstrated that oxidation mediated its effects via methionine rather than cysteine residues in the N-terminus of the Shaker C/ B channel tail. They also showed that the Shaker B splice variant contained only one methionine residue (M1) in its amino terminus ball domain (M1), compared to the Shaker C/B splice variant, which, like LINGO2, contained two methionines (M1 & M3) in its tail. Our results suggest that the removal of inactivation by Ch-T was not due to an effect on cysteine residues, given the insensitivity of the responses to DTT treatment. Instead, a number of lines of evidence support the idea that Ch-T mediated its effect via oxidation of M605 and M603 in the Cterminus of LINGO2. Firstly, synthetic peptides containing Met(O) at either positions, showed a decreased rate of inactivation and a reduced affinity for the BK channels. illumination (Fig. S5), suggesting that this residue may play a more prominent role in this response. The modeling data shown in Figures 8 and S6 are consistent with this idea and potentially provide a structural basis for these observations. Thus, in the absence of oxidation, the C-terminus of the LINGO2 protein docks deep in the BK channel pore and M605 is buried in a hydrophobic region, which would reduce its ability to be modified by oxidizing stimuli. In contrast, the M603 residue is exposed to the aqueous environment of the pore and thus could be more easily oxidized. Such an arrangement is consistent with the different rates of removal of inactivation by Ch-T shown in Figure 6, C and E. The resultant oxidation of either of these methionine residues would reduce their lipophilicity and disrupt any lipophilic interactions in the pore, which presumably reduces the affinity of the inactivating particle and consequently decreases inactivation. Figure 8E shows an example of the consequences of having M603 oxidized, where it is clear that the last eight residues (blue sticks) of the tail peptide fail to dock as deeply in the pore, compared to the WT LINGO2 tail shown in Figure 8D. This more surfaced pose of the oxidized LINGO2 tail may provide a structural hypothesis for the lack of channel inactivation observed during oxidation, although this will require experimental confirmation in future structural studies. The sensitivity of BK:LINGO2 channels to oxidation may provide an additional mechanism for dynamically modulating BK channel activity. The functional consequences of such modulation are illustrated in Figure 7, D-F, where Ch-T in 10 μM Ca 2+ significantly increased the open probability of BK:LINGO2 channels, even at physiological potentials. Interestingly, in contrast to early reports that LINGO proteins were exclusively expressed in the CNS, recent immunohistochemical studies have demonstrated that LINGO proteins are also expressed on human intestinal epithelial cells (LINGO2, (37); LINGO3, (38)), endometrial smooth muscle (LINGO2, (39)), and in the human respiratory system (LINGO1, (39); LINGO3, (38)). Given that LINGO1-3 proteins possess the same pattern of two methionines in the last four residues of their C-termini (Fig. 6A), it is possible that they all can be modulated to various extents by oxidation. We speculate that this may be particularly important in the tissues exposed to a highly oxidizing environment such as the epithelial cells of the respiratory system. We surmise that the oxidation of BK:LINGO1-3 channels could dynamically regulate BK channel function to increase their open probability, but this will require experimental confirmation.
In conclusion, we present evidence that LINGO2 interacts with BK channels to shift their voltage of activation negatively and induce inactivation, which is subject to regulation by oxidation of two methionine residues in the C-terminus tail of LINGO2.

Experimental procedures Solutions
All excised patch experiments were performed at 37 C in 140 mM symmetrical K + solutions which contained 140 mM KCl, 10 mM glucose, 10 mM Hepes, and either 1 mM ethylene glycol-bis(β-amino ether)-N,N,N 0 ,N 0 -tetraacetic acid (EGTA (for free [Ca 2+ ] 100 nM to 300 nM) or 1 mM hydroxyethylethylenediaminetriacetic acid (HEDTA for free [Ca 2+ ] 1 μM to 10 μM) and the pH of all solutions was adjusted to 7.2 with KOH. These solutions were made with double-distilled, deionized, filtered water from a MilliQ water purification system. The pipette solution contained 100 nM free Ca 2+ . We used "Chelator" (https://www.ru.nl/animal/research/chelator/) to calculate the total amount of Ca 2+ required, as per Schoenmakers et al., (40)  , 10 glucose, 2.9 sucrose, and 10 Hepes, and its pH was adjusted to 7.4 with NaOH. In addition, the patch under study was continuously superfused by means of a close delivery system consisting of a pipette (tip diameter 200 μm) placed approximately 300 μm away from the cell. This could be switched, with a dead-space time of around 5 s, to a solution containing a drug.

Electrophysiology
Electrodes were pulled from Corning borosilicate glass (1.5 mm O.D. × 0.86 mm I.D.) using a Sutter P-97 pipette puller and fire polished using a Narashige MF 83 microforge. Pipettes had resistances of 2 to 5 MΩ when filled with recording solutions and series resistance was compensated by up to 80%. Standard single-channel patch clamp recording methods were used. Voltage clamp commands were delivered via an Axopatch 200B patch clamp amplifier (Axon Instruments) connected to a Digidata 1322A AD/DA converter (Axon Instruments) interfaced to computers running pClamp software (Axon Instruments). Data was acquired at 100 KHz and filtered at 2 KHz or 5 KHz. Patches were held at either −60 mV or −100 mV and depolarized in 20 mV increments to 200 mV. Residual capacitance and leakage currents were subtracted using either a P/4 protocol or offline by manual leak subtraction.

Illumination
We used either a cool LED PE100 470 nM LED (CoolLED) light source (at 75% intensity) attached via a fibre optic guide to the rear port of a Nikon Diaphot microscope, through a standard FITC filter set, to a Nikon Plan 40/0.55 objective, or via a Nikon Mercury Lamp (C-SHG1) through the rear port of a Nikon Eclipse TE2000S microscope and passed through a standard FITC filter set to a Nikon Pan Fluo 40×/0.6 objective. Inactivation of BK:LINGO2 currents in excised patches was abolished within 4 min of illumination.

Ion channel cloning and mutagenesis
The human LINGO2 transcript used [NM_001354575.2] was untagged in a pcDNA 3.1 vector (VectorBuilder). The α subunit of the rabbit BK channel was isolated from rabbit urethral smooth muscle and cloned using the pcDNA TOPO DNA cloning kit (Life Technologies). The identifiesd EDITORS' PICK: Redox modulation of BK:LINGO2 channels transcript corresponded to the ZERO variant of mouse BKα and to variant 2 (NM_002247.3) of human BKα. We used 100 ng ml −1 cDNA for BKα only transfections, whereas a 100:500 ng ml −1 cDNA ratio was used for BKα:LINGO2 and BKα:LINGO1 cotransfections for inside-out patch recordings. GFP (150 ng ml −1 ) was included in all transfections except those used for Figure 2, C and D. cDNA was transiently transfected into HEK cells using lipofectamine. Site-directed mutagenesis used the Phusion kit (Thermo Fisher Scientific) method (41) and all mutations were confirmed by sequencing.

Insertion of HA sequence into LINGO2
A PCR strategy was used to incorporate the coding sequence of the HA epitope YPYDVPDYA into full-length LINGO2 between residues 27 and 28 with the HA sequence shown in bold -STIGYPYDVPDYACPAR. Oligonucleotides were designed coding for the HA sequence and a small complementary section of LINGO2 sequence to insert the HA tag downstream of the signaling sequence. Reaction mixes containing template, Phusion Pfu (Thermo Fisher Scientific), GC buffer, dNTPs, DMSO, and either forward or reverse primers in separate tubes were cycled in a PCR machine for 10 cycles of 98 C − 10 s, 65 C − 30 s, and 72 C − 3 min 30 s. Forward and reverse reaction mixes were then combined for 20 cycles of PCR using the same reaction conditions. PCR products were treated with DpnI (New England Biolabs) and transformed into XL10 competent cells. Resultant colonies were then DNA extracted and the plasmids sequenced in both directions to check for HA insertion and sequence integrity.

Generation of the tagged BK channel
The ZERO variant of the BK channel with an N-terminal Flag-epitope and C-terminal-HA epitope (Flag-ZERO-HA) was constructed in pcDNA3 as previously described by Chen et al 2005 (42). The N-terminal sequence, thus, begins immediately upstream of MDALI start site with the sequence MDYKDDDDKMDALI.. (Flag-sequence in bold italic). The HA tag at the C-terminus replaced the normal stop codon of the ..REVEDEC tail thus the C-terminus sequence is .REVE-DECYPYDVPDYA* (-HA sequence in bold italic).
To engineer the Flag-ZERO-myc construct, a short sequence of the C-terminal tail spanning the internal SfiI site and a NotI site in the vector backbone was synthesized de novo (Twist Bioscience) and ligated into the Flag-ZERO-HA construct to replace the HA tag with a C-terminal-myc tag. The C-terminal amino acids of Flag-ZERO-myc was, thus, … REVEDECEQKLISEEDL*.
Electrophysiology curve fitting and statistics G was derived from peak currents according to Ohm's law where, E K = 0 mV in symmetrical [K + ]. Summary data was expressed as the mean ± SEM. G-V relationships were fitted with the Boltzmann equation where, V 1/2 is the voltage of half-maximum activation, S is the slope of the curve, V m is the test potential, G is the conductance, and G max is the maximal conductance. For BK constructs, data from each patch was normalized to the peak conductance measured in either 1 μM or 10 μM Ca 2+ to obtain G max , and curves were constrained to the G max value obtained in this. In full-length BKα:LINGO2 channels, currents were normalized to the peak G max recorded in 1 μM Ca 2+ or 100 nM Ca 2+ . Inactivation curves were fitted with a similar Boltzmann function: where, I was the current recorded at the test step, I max was the maximal current recorded, V c was the conditioning potential (see RESULTS), and K was the slope factor. Concentration-effect data were fitted with a Hill-Langmuir equation: where, I was the current recorded in the drug, I control was the current in the absence of drug, IC 50 was the half-maximal effective concentration, and [Drug] was the concentration variable.

Statistical tests
All data is presented as mean ± SD and was analyzed with Prism software (GraphPad). Individual data points are also shown in bar charts. Statistical differences were assessed using either paired or unpaired nonparametric tests (Wilcoxin signed-rank test, Mann-Whitney test), or the extra sum of squares F-test, as appropriate. Comparisons of three or more nonparametric data sets used Friedman's test followed by Dunn's test for multiple comparisons, whereas ANOVA was used for three or more parametric data sets. p < 0.05 was taken as significant.

Molecular modeling
The structure of the Ca 2+ bound human BK channel in complex with the β4 subunit with PDB code: 6V22 (21) was used for protein-peptide docking. The eight C-terminus residues of LINGO2 in the WT, M603 oxidized form, M603L, and M605L were first docked to the BK channel opening using the Glide module of Schrodinger software (Release 2020-1; (43)). The protein and peptide structures were prepared with the protein preparation module of Schrodinger software using a standard protocol. Residues involving L312, F315, S317, V319, P320, E321, I323, and E324 were selected as the center of the docking box. Docking poses were evaluated with the Glide docking score. The frequency of docking pose and the possibility of linking the peptide with the rest of LINGO2 were considered in the final docking pose selection. The N-terminal transmembrane helical part of LINGO2 was taken from the AlphaFold database (44) and docked to the BK channel, using the HEX protein-protein docking program (45) with a default protocol. The peptide linker between the transmembrane helix and the C-terminal 8-residue peptide was built taking into account the architecture of the BK channel using the Prime module of Schrodinger (46). Next, the transmembrane part of LINGO2 was manually connected to the 8-residue peptide, and the entire structure was optimized. Minimization of docking complexes was carried out with the MacroModel module of Schrodinger software. A default protocol of minimization in the implicit solvent with a distance-dependent dielectric constant was used to obtain the final complexes. The hydrogen bonds of the helices were constrained to avoid the structure deformation. The OPLS2005 force field was used in all calculations. The 3D images were created in Maestro 2020 to 1.

Coimmunoprecipitation assay
HEK293T cells were seeded out at 4 × 10 6 in a 10 cm dish in complete media (Dulbecco's modified Eagle's medium; 10% fetal calf serum and 1% PenStrep) and incubated overnight at 37 C in 5% CO 2 . The next day, Flag-Zero-myc and HA-LINGO2 cDNA (each at 441 ng ml −1 to maximize protein expression) were cotransfected into the cells using the TransIT-293 transfection reagent (3:1; carrier: DNA) (Mirus Bio). The transfected cells were incubated for 72 h at 37 C in 5% CO 2 .
For harvesting lysates, cell-conditioned media were aspirated, and cells were scraped into 500 μl ice-cold immunoprecipitation (IP) lysis buffer (20 mM Tris, 150 mM NaCl, 0.2% NP-40, 10% glycerol, pH 7.0, plus protease inhibitor cocktail and phosphostop tablets). The cell suspension was incubated on ice for 15 min with intermittent vortexing, before centrifugation (4 C, max speed for 10 min) and collection of the supernatant. The lysates were quantified by bicinchoninic assay (BCA) and kept on ice until IP.
For the immunoprecipitation, the Dynabeads Protein G immunoprecipitation kit (Invitrogen, 10007D) was used, following manufacturer's instructions. Briefly, 1 mg beads were transferred to fresh Eppendorf's, and storage buffer was aspirated using magnet. Antibody binding buffer was added (200 μl) to each tube, followed by 8 μg antibodies (anti-Flag-M2; Sigma F3165 or anti-mouse IgG1; BioLegend 401,402) and incubated on a rotator for 1 h at room temperature (RT). Conjugated Ab-Bead complexes were then washed 3× in wash buffer before storing in 50 μl binding buffer. Lysates were added to the bead mixtures at 1 mg per tube. Inputs were retained in Eppendorf's at 20 μg. The IP tubes were filled up to 750 μl final volume using binding/wash buffer and rotated for 1 h at RT. Post incubation, the beads were washed 3× using the wash buffer and resuspended in elution buffer with lithium dodecyl sulfate (4×) (final volume 25 μl). Input and IP samples were then denatured at 94 C for 10 min before separation by SDS-PAGE.
For immunoblotting, lysates were separated using 4 to 12% NuPage BT gel for 1 h at 200 V, transferred onto nitrocellulose using iBlot2 (25V for 8 min), transfer confirmed by Ponceau S staining, membrane was blocked in 5% milk for 1 h at RT and the membrane incubated in anti-HA antibody (CST; 3724) overnight on a roller at 4 C. For imaging, the membranes were washed 3× in TBST before incubating in anti-rabbit-800 secondary Ab (LiCor; 926-32211) for 1 h at RT in a dark box. The membrane was washed 3× in TBST before imaging using LiCor Odyssey imager.

LINGO2 tail peptide synthesis
All peptides were prepared manually using standard Fmoc strategy on a rink amide MBHA resin (47). The resin (100 mg) was swollen in dimethylformamide (DMF) (2 ml) before use, in a polypropylene-fritted chromatography column. Deprotection of the resin was effected by shaking in a solution of piperidine in DMF (1:4 v/v) for two periods of 25 min. The resin was then washed with DMF (x4), DCM (x4), and DMF (x4). For amino acid coupling, Fmoc-protected amino acid derivatives (3 equiv.) were activated with di-isopropylethylamine (3 equiv.) and COMU (3 equiv.) (48) in DMF, loaded onto the resin, and shaken for 45 min. The resin was then washed with DMF (x4), DCM (x4), and DMF (x4). Deprotection of the N-terminus of the growing peptide chain was accomplished by two 25-min periods of shaking in a solution of piperidine in DMF (1:4 v/ v). Successful coupling and deprotection steps were confirmed using the ninhydrin test (49,50). N-terminus capping was carried out using a solution of acetic anhydride:pyridine:DMF (2:2:6 v/v) for 2 h. Global deprotection and cleavage of the peptide from the resin was performed using a cocktail containing trifluoracetic acid/water/phenol/ethanedithiol/thioanisole (82.5/5/5/2.5/5 v/v) for 2 h under a blanket of nitrogen. Peptides were precipitated in ice-cold ether and centrifuged at 3000 rpm for 10 min. The resulting solid was isolated by filtration, washed several times with ether, dried under vacuum, and stored at −20 C. Peptides were analyzed by reverse-