Chinese hamster ovary (CHO) cells were obtained from the American Type Culture Collection (Manassas, VA) and transfected with wild-type K_V1.3 and enhanced green fluorescent protein (EGFP) encoding plasmids (OriGene Technologies, MD, United States) using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA) as described in Supplementary Material.

Methyl-β-cyclodextrin (MβCD), hydroxypropyl-β-cyclodextrin (HPβCD), hydroxypropyl-γ-cyclodextrin (HPγCD), and “inverted” cyclodextrins (iCDs) including hexakis (3,6-anhydro)-α-cyclodextrin (iαCD), heptakis (3,6-anhydro)-β-cyclodextrin (iβCD) and octakis (3,6-anhydro)-γ-cyclodextrin (iγCD) were obtained from CycloLab Cyclodextrin R&D Laboratory (Budapest, Hungary). For further characterization of iCDs, see Supplementary Methods. In general, cells were treated with CDs dissolved in standard extracellular solution at concentrations of 1 or 5 mM at room temperature. Incubation times and other details of CD applications for the given experiments are described in the following sections.

Cholesterol contents of control samples and those treated with different CDs for 1 h were determined using a fluorometric cholesterol quantitation kit (Sigma-Aldrich) according to instructions of the manufacturer with a Synergy HT Microplate Reader (BioTek Instruments, Winooski, VT, United States). For kinetic examination of membrane cholesterol extraction, cells were pre-loaded using MβCD complexed with NBD-conjugated cholesterol (25-[N-[(7-nitro-2-1,3-benzoxadiazol-4-yl)methyl]amino]-27-norcholesterol, Avanti Polar Lipids, Alabaster, AL and CycloLab Cyclodextrin R&D Laboratory) for 1 h at a concentration of 195 µM sterol, followed by incubation with 1 or 5 mM of CDs in the sample holder of a NovoCyte 3000RYB flow cytometer (ACEA Biosciences, San Diego, CA) at room temperature. NBD fluorescence intensity was determined using excitation at 488 nm and an 530/30 nm emission filter. Measurements started immediately after addition of CDs and continued for 10 min. Time-correlated fluorescence intensity values were quantified with FCS Express (De Novo Software, Pasadena, CA) and a custom-written Matlab program as described previously (Batta et al., 2021).

Patch-clamp measurements were carried out in whole-cell or outside-out configuration using a KF-based pipette solution with a final K⁺ concentration of 160–165 mM and a standard extracellular solution containing 150 mM NaCl (see also Supplementary Material). CDs were dissolved in standard extracellular solution that were directly applied on the cells with a gravitation propelled perfusion system. The proper working of the perfusion system was validated by applying a high potassium (150 mM KCl) containing extracellular solution before the measurement of every single cell to exclude false negative cases originating from the inappropriate flow of CD-containing solutions.

For characterizing direct inhibitory effects of CDs, 15 ms depolarizing pulses to +50 mV were applied every 15 s from a holding potential of –100 mV. The current inhibitory effects of CDs at given concentrations were determined as remaining current fractions (RCF) using the following equation RCF = I I 0 (1) where I ₀ is the leak-corrected peak current in standard extracellular solution before CD applications and I is the leak-corrected peak current of the same patch at equilibrium block, at a given CD concentration. To demonstrate the washing-in kinetics of CDs, leak-corrected peak currents at every time point were normalized to the maximal peak. To characterize the extent of washing-out of CDs, recovered current fractions (RF) were determined as a leak-corrected peak current in standard extracellular solution after washing-out the given CDs (I_CD) over the initial leak-corrected peaks (I₀) before CD applications according to RF = I CD I 0 (2)

To measure changes in membrane fluidity, hydration and lipid order in response to CDs, fluorescence anisotropy of 4′-(trimethylammonio)-diphenylhexatriene (TMA-DPH) and generalized polarization (GP) of two indicators, 6-dodecanoyl-N,N-dimethyl-2-naphthylamine (Laurdan) and 4-[2-(6-Dibutylamino-5-fluoro-naphthalen-2-yl)-vinyl]-1-(3-triethylammonio-propyl)-pyridinium dibromide (PY3174), were quantified as described previously and in Supplementary Material (Kwiatek et al., 2013; Kovács et al., 2017; Zakany et al., 2021).

In silico docking was performed with AutoDock Vina (Trott and Olson, 2009). The conformation of MβCD was extracted from the crystal structure of the extracellular domain of human Gastric inhibitory polypeptide receptor (PDB 2QKH) that was cocrystallized with MβCD (Parthier et al., 2007). The target of docking was the cryo-electron microscopic structure of K_V1.3 (PDB 7EJ1) (Liu et al., 2021) and the search space was defined to include the extracellular orifice of the pore domain (Supplementary Figure S1). AutoDock Vina was run with an exhaustiveness parameter of 40, and ten binding modes were recorded. Amino acids taking part in ligand-target interactions were identified and displayed using LigPlot+ (Laskowski and Swindells, 2011) and PyMol.

Measured data are represented as mean ± SEM obtained from n independent samples indicated in figure legends. Differences were considered significant (*) when p < 0.05 calculated based on ANOVA followed by Tukey’s HSD test.

In order to test membrane cholesterol depletion induced by CDs with different cavity charge profiles (Figure 1A), we examined cellular cholesterol levels using a commercially available fluorometric cholesterol quantitation kit after treatment of CHO cells with 1 and 5 mM of CDs for 1 h. In CDs with conventional cavity polarity we obtained data consistent with literature (Zidovetzki and Levitan, 2007; Szente et al., 2018). MβCD and HPβCD, two derivatives most widely used and most effective in cholesterol extraction, induced significant and comparable dose-dependent decreases in cholesterol levels reaching ∼50% extraction at a concentration of 5 mM (Figure 1B). HPγCD with much lower affinity for forming complexes with cholesterol also resulted in reduced cholesterol contents, however, these reductions were much smaller thus reaching the level of significance only at 5 mM. Concordant with our expectations, iCDs with a hydrophilic cavity did not cause any significant alterations in membrane cholesterol levels.

To characterize the time dependence of cholesterol extraction induced by MβCD, HPβCD, and HPγCD, we preloaded cells with NBD-cholesterol, a fluorophore-conjugated sterol derivative, and examined changes in fluorescence intensities of cells using time-resolved flow cytometry. Treatments with 5 mM of CDs started in the sample holder of the flow cytometer immediately before the initiation of measurements. With this technique we were able to follow reductions in levels of exogenous cholesterol molecules incorporated into cellular membranes with a time resolution of 20 s in the first 10 min of treatments. We observed time-dependent decreases in fluorescence in response to all three examined CDs (Figure 1C). Again, consistent with literature, cholesterol-extracting abilities of MβCD and HPβCD were superior to that of HPγCD (Zidovetzki and Levitan, 2007; Szente et al., 2018). Since only ∼12% of the cholesterol extracted by a 1-h CD treatment was depleted from cells in the first 3 min, these experiments suggested that cholesterol depletion in response to conventional CDs is dominant mainly after longer incubation periods. Considering a recent report suggesting similar efficiency of MβCD to extract unlabeled and NBD-cholesterol in living cells (Ostašov et al., 2013), this piece of information was utilized for the design of patch-clamp measurements to exclusively separate in time the direct, ligand-like and indirect effects of CDs showing significant cholesterol-depleting abilities.

The potential direct, ligand-like effect of CDs on K_V1.3 was tested by patch-clamp. Currents were elicited by 15-ms depolarizing steps to +50 mV from a holding potential of –100 mV (Figures 2A,B, top panels) to fully activate K_V1.3 current (Figures 2A,B, bottom panel, black solid lines). Only those experiments were evaluated, where the proper working of the perfusion apparatus was verified by changes in current parameters in response to high potassium (150 mM) solution (Figure 2A, grey solid line), which could be subsequently reverted in one step by re-applying standard extracellular solution (not shown for clarity). The exclusive investigation of direct, ligand-like interactions between CDs and K_V1.3 was ensured in two ways. 1) If patch-clamp measurements are performed in 3 minutes after applying the CD-containing extracellular solutions, no significant cholesterol extraction in response to the cholesterol-depleting CDs takes place (Figure 1B). 2) Application of iαCD, iβCD and iγCD is a suitable tool for the exclusive examination of direct ligand-like effects of CDs since these derivatives are not able to deplete membrane cholesterol even after 1 h (Figure 1A). As indicated by representative current traces (Figures 2A,B, blue traces) and decreases in RCF values (Figure 2E), a current block can be detected in the presence of both 1 and 5 mM iαCD. Similarly, MβCD and iγCD also resulted in dose-dependent current blocking effects, as shown by decreases in RCF values, while neither the cholesterol-depleting HPβCD and HPγCD, nor the non-cholesterol-depleting iβCD induced any remarkable changes in RCF (Figure 2E). According to the wash-in kinetics for MβCD (Figure 2C, black lines), iαCD and iγCD (Figure 2D blue and cyan lines), current-blocking effects were completed within 90 s after the initiation of CD exposure in the absence of significant cholesterol depletion according to Figure 1B. In all three cases, current blocking effects were only partially reversible (Figures 2A,B dotted lines), as indicated by RF values (Figure 2F).

Although experiments described in the previous section suggest direct interaction between CDs and K_V1.3 ion channels, we examined whether CDs can possibly exert their effects through changes in biophysical properties of the cell membrane. Therefore, we examined membrane fluidity, hydration and lipid order using environment-sensitive fluorophores. First, to test membrane fluidity, we treated cells with 1 or 5 mM CDs for 1 h and determined TMA-DPH fluorescence anisotropy negatively correlating with the fluidity of the cell membrane (Batta et al., 2021; Zakany et al., 2021). As can be expected from the intimate relationship between membrane cholesterol content and fluidity, TMA-DPH fluorescence anisotropy of control cells was significantly decreased by cholesterol-extracting CDs (Figure 3A). On the other hand, iCDs not capable of depleting membrane cholesterol induced no significant changes in membrane fluidity. Similar effects were observed when examining membrane hydration through quantification of Laurdan GP inversely correlating with membrane hydration (Batta et al., 2021; Zakany et al., 2021). Laurdan GP of control cells was significantly reduced by CDs causing cholesterol depletion, while it was not affected by iCDs (Figure 3B). In order to remove the potential contribution of internalized Laurdan to the measured signal and to ascertain that the calculated GP values reflect the hydration of the plasma membrane where the potassium channels are expressed, membrane lipid order was also investigated with PY3174. Analysis of PY3174 GP, a parameter that positively correlates with the degree of lipid order (Kwiatek et al., 2013; Zakany et al., 2021), was restricted to the plasma membrane using segmentation of confocal microscopic images. PY3174 GP of control cells was significantly lowered by conventional CDs, but not the inverted derivatives (Figure 3C).

Furthermore, to characterize the time-dependence of changes induced by MβCD, HPβCD and HPγCD in membrane fluidity, we repeated our measurements with TMA-DPH by pre-labeling cells with the fluorophore, which was followed by incubating them with 5 mM CDs in the cuvette used in the spectrofluorometer. Treatments started immediately before the initiation of measurements that included repeated quantification of fluorescence anisotropy every minute in the first 10 min of incubation and, as an end-point of the experiments, after 60 min. Consistent with results of our kinetic cholesterol quantitation experiments, time-dependent reductions were observed in fluorescence anisotropy values for all three examined CDs (Figure 3D). Membrane fluidizing effects of MβCD and HPβCD were similar and much higher in magnitude than those of HPγCD. For MβCD and HPβCD, significant decreases occurred after 8 min, while in the case of HPγCD reductions were not significant in the first 10 min. Based on the kinetic traces it could be concluded that only relatively small, 15–20% reductions occurred in the first 5 min. These data suggest that alterations in membrane biophysical parameters can be expected in response to cholesterol-extracting CDs mainly after longer incubation periods not in the time scale of patch-clamp experiments.

To quantitatively characterize the correlation between K_V1.3 current-inhibiting propensities of the examined CDs and their abilities to deplete membrane cholesterol and modify membrane biophysical parameters, we performed linear regression analysis on results described in previous sections, which were obtained after 1-h incubation in the presence of 5 mM CDs. R ² and p values revealed no significant correlation between RCF values determined in patch-clamp measurements and cholesterol levels, TMA-DPH fluorescence anisotropy, Laurdan GP or PY3174 GP values (Figure 4A), further supporting the existence of a cholesterol-independent current-blocking effect of certain CDs.

Furthermore, to examine potential direct interactions between CDs and K_V1.3, we performed in silico molecular docking analysis between MβCD and Kv1.3 (Figure 4B). Since electrophysiological evidence argues against binding of MβCD intracellularly (current-inhibiting effects of all CDs appeared in 15 s and completed within 90 s) or to the voltage-sensor (no shift in activation threshold, data not shown), the extracellular orifice of the pore domain was chosen as a search space for docking covering the most frequent binding sites of toxins and small molecule inhibitors (Kavanaugh et al., 1991; Holmgren et al., 1997; Gilquin et al., 2005; Karbat et al., 2019) (Supplementary Figure S1). Ten binding modes were recorded, which were characterized by estimated K_ds consistent with the dose-dependence of current inhibition described above (binding affinities were in the range of −4.3 to −3.9 kcal/mol corresponding to estimated K_d values between 618 and 1,229 µM). These modes identified a common pattern of direct binding involving a network of hydrogen bonds and hydrophobic interactions between MβCD and pore residues. The association was typically mediated by hydrogen bonds formed by His399, Gly396 and Thr373 or Asp397, and hydrophobic interactions through Asp397, Gly375 or Tyr395 of one subunit. The structure was usually further stabilized by a hydrogen bond with His399 and hydrophobic interaction with Gly396 on the adjacent subunit, and additional hydrophobic interactions with Gly396 and Asp397 of the opposing subunit (Figure 4B).