Coupling of functionality to trafficking of KCNQ2/3 potassium channels at the axon initial segment

Significance

KCNQ2/3 voltage-gated potassium channels are critical determinants of neuronal excitability. The impact of KCNQ2/3 on neuronal activity is regulated not only by the functionality of channels but also by their localization on the axon initial segment (AIS), the site of action potential initiation. However, there has been no comprehensive study on the linkage between these two regulatory mechanisms governing KCNQ2/3 functionality and trafficking. Here, we uncover a molecular mechanism inherent to the KCNQ2/3 channel, in which the conformational state of the KCNQ3 subunit dictates its binding kinetics to ankyrinG, the AIS master organizer, and thereby regulates the AIS localization of KCNQ2/3 complexes. This unified framework for the KCNQ2/3 gating and trafficking provides insights into AIS quality control and KCNQ2/3-related epilepsy.

Abstract

KCNQ2/3, a major voltage-gated potassium channel at the axon initial segment (AIS), plays a crucial role in controlling neuronal excitability. While the functionality of KCNQ2/3 is regulated by conformational changes from voltage sensing, the AIS localization of KCNQ2/3 is regulated by ankyrinG (ankG). However, the potential coupling between the mechanisms governing channel functionality and trafficking remains unresolved. Here, we combine genetic engineering of channel functionality with advanced imaging techniques of channel trafficking to uncover a coupling of KCNQ2/3 functionality to trafficking. Single-molecule imaging reveals that reduced KCNQ3 functionality alters the entire trafficking pathway, including exo/endocytosis and lateral diffusion, reducing AIS localization of KCNQ2/3. Furthermore, we develop a live-cell assay to quantify the interactions between full-length KCNQ3 and ankG, demonstrating that the active conformation of KCNQ3 is essential for the stable ankG binding. Our findings establish a mechanistic basis for the integration of KCNQ2/3 gating and trafficking in regulating neuronal excitability.

KCNQ2/3 is a primary type of neuronal voltage-gated potassium channel playing a crucial role in regulating membrane potential and firing rates of action potentials in neurons (1–3). Dysfunction of KCNQ2/3 precipitates various neurological diseases, including benign familial neonatal convulsions (BFNC) and early infantile epileptic encephalopathy (EIEE) (4–6). The physiological function of KCNQ2/3 depends on both channel gating and its spatial distribution at specific subcellular sites such as the axon initial segment (AIS) (7–9). Although molecular mechanisms underlying KCNQ2/3 gating are well understood at the structural level (10), how these channels are trafficked to the AIS remains elusive.

The AIS localization of KCNQ2/3 channels is mediated by multiple trafficking regulatory sites (11, 12). The critical requirement for AIS localization of KCNQ2/3 is the interaction between the scaffolding protein ankyrinG (ankG) and KCNQ3 via the ankG binding motif (ABM) near its C-termini (7, 8). The KCNQ3 subunit exhibits a higher ankG binding affinity than the KCNQ2 subunit (13) and is regarded as the dominant trafficking regulator in the heteromeric KCNQ2/3 complex (9). In addition, several mechanisms governing KCNQ2/3 functionality simultaneously contribute to the regulation of their AIS localization. CaM binding to the A-B helices and heteromeric channel assembly via the C-D helices are not only critical for the gating control of KCNQ2/3 but also necessary for their efficient surface expression (9, 14–21). Furthermore, epilepsy-related mutations in the pore domain (PD) hinder AIS targeting of KCNQ2/3 (22). Taken together, these reports suggest that the functional state of KCNQ2/3 may be an essential determinant of their proper trafficking. Accordingly, all the regulatory processes in the sequential activation of KCNQ2/3, including voltage sensing and electromechanical (E-M) coupling (23–25), may concurrently govern their AIS localization. However, the causal role of KCNQ2/3 functionality in their trafficking has yet to be examined. There is still a lack of fundamental information on the kinetics of three trafficking pathways (i.e., exocytosis, lateral diffusion, and endocytosis) that shape the spatial distribution of KCNQ2/3 in neurons. While ankG is recognized as the master regulator of AIS proteins including KCNQ2/3 (26), it is unclear whether sites other than ABM of KCNQ2/3 also affect their interaction with ankG. A significant limitation is the absence of a method to quantify the interaction between full-length KCNQ and ankyrins in living cells. Therefore, a key mechanistic challenge is to determine whether and how the functional state of KCNQ2/3 influences their AIS targeting in neurons, and to uncover the underlying molecular mechanisms.

Here, we aim to elucidate the coupling of functionality to trafficking of KCNQ2/3. First, by combining the genetic manipulation of voltage sensing and E-M coupling of KCNQ3 and imaging analysis of KCNQ2/3 trafficking, we provide compelling evidence for a positive causal effect of KCNQ3 functionality on AIS localization of KCNQ2/3. Moreover, single-molecule imaging reveals that reduced KCNQ3 functionality suppresses exocytic insertion, enhances lateral diffusion, and accelerates endocytic internalization of KCNQ2/3 on the AIS surface, thereby reducing their AIS localization. Finally, we develop a method for quantification of interaction between full-length KCNQ3 and ankG in living cells, demonstrating that the active conformation of KCNQ3 is required for more stable ankG binding. These findings identify the functional state of KCNQ3 as an essential factor determining the AIS localization of KCNQ2/3 via ankG binding, and provide insight into the molecular basis of AIS quality control and epilepsy associated with KCNQ2/3 dysfunction.

Results

Halo-TAC-KCNQ3 Retains Channel Functionality and Enables Live-Cell Imaging of Heteromeric KCNQ2/3 Complexes.

To separately visualize the surface and intracellular pools of KCNQ2/3 in neurons, we added a HaloTag-fused single-pass transmembrane domain (TAC) to the N-terminus of mouse KCNQ2 and KCNQ3 (Halo-TAC-KCNQ2 and Halo-TAC-KCNQ3), adapting a previously reported strategy (27). This design ensures that the HaloTag is displayed extracellularly. To confirm that the Halo-TAC fusion does not alter the channel functionality, we expressed these constructs in HEK293T cells and performed whole-cell patch-clamp recordings. For KCNQ3, we utilized a KCNQ3-A316T mutant (KCNQ3T) to promote efficient trafficking and functional expression of homotetrameric channels (28). The electrophysiological analysis revealed no significant differences in current density, V_1/2, and activation kinetics between Halo-TAC-fused and -unfused channels (Fig. 1 A–C and SI Appendix, Figs. S1 A–C and S2). We next evaluated whether the Halo-TAC fusion enables live-cell imaging of heteromeric KCNQ2/3 complexes in cultured hippocampal neurons. We coexpressed Halo-TAC-KCNQ2 with KCNQ3 or Halo-TAC-KCNQ3 with KCNQ2 in neurons at 10 d in vitro (DIV), when the periodic cytoskeletal structure and the diffusion barrier at the AIS are established (29, 30). We then sequentially labeled HaloTag-fused channels with a cell-impermeable fluorescent ligand (AF660) to mark surface pools, followed by a cell-permeable fluorescent ligand (TMR) to label intracellular channels. This dual-labeling strategy enabled simultaneous discrimination of surface versus cytoplasmic HaloTag-fused channels. The AIS was marked by coexpressed EGFP-Nav1.2 II-III linker (31–33). Confocal imaging showed that both Halo-TAC-KCNQ2 and Halo-TAC-KCNQ3 were diffusively distributed throughout the cytoplasm, but selectively enriched on the distal AIS surface (Fig. 1 D and F and SI Appendix, Fig. S1D), recapitulating the distribution of endogenous KCNQ2/3 (8). We quantified AIS localization as the AF660 intensity ratio between the AIS and the dendrite of the same neuron (Fig. 1 G–I and SI Appendix, Fig. S1 F–H). The AIS localization of Halo-TAC-KCNQ3 was observed only when coexpressed with KCNQ2 (Fig. 1 D and F–I and SI Appendix, Fig. S1 D and F–H), indicating that Halo-TAC-KCNQ3 labeled with AF660 predominantly reflects KCNQ2/3 heteromer. Furthermore, the reduction of ankG binding affinity of KCNQ3 by mutations in ABM, namely E838A/T839A/D840A (ETD/AAA) (8, 9), not only suppressed the AIS localization of Halo-TAC-KCNQ3 itself (Fig. 1 D–I) but also exerted a dominant-negative effect on that of coexpressed Halo-TAC-KCNQ2 (SI Appendix, Fig. S1 D and F–H). In contrast, the ETD/AAA mutation of Halo-TAC-KCNQ2 did not impair its AIS targeting when coexpressed with KCNQ3 (SI Appendix, Fig. S1 D–H). These results reconfirm the dominant regulatory role of KCNQ3 in KCNQ2/3 trafficking (9). In summary, our data demonstrate that Halo-TAC-KCNQ3 retains functional channel gating and trafficking properties, making it a suitable probe for tracing heteromeric KCNQ2/3 complexes to study the relationship between their functionality and AIS localization.

Fig. 1.

Halo-TAC-KCNQ3 retains channel functionality and enables live-cell imaging of heteromeric KCNQ2/3 complexes. (A) Representative current traces of KCNQ3T and Halo-TAC-KCNQ3T obtained by whole-cell patch-clamp recordings in HEK293T cells. The dotted line indicates the zero-current level. (B) Current densities at 80 mV for KCNQ3T (Q3T) and Halo-TAC-KCNQ3T (Halo-TAC-Q3T). P = 0.090. (C) G–V curves of KCNQ3T and Halo-TAC-KCNQ3T. V_1/2 for KCNQ3T: −35.11 ± 1.67 mV and for Halo-TAC-KCNQ3T: −35.38 ± 1.87 mV. P = 0.927. (B and C) n = 4 cells (Q3T) and 5 cells (Halo-TAC-Q3T). P-values were obtained by Welch’s t test. (D) Representative confocal microscopic images of primary cultured hippocampal neurons expressing EGFP-Nav1.2 II-III linker and Halo-TAC-KCNQ3 with KCNQ2 or an empty vector. (E) Representative confocal microscopic images of primary cultured hippocampal neurons expressing EGFP-Nav1.2 II-III linker, Halo-TAC-KCNQ3-ETD/AAA, and KCNQ2. (D and E) HaloTag was labeled with AF660 and TMR. The arrow indicates the position of AIS. (Scale bar, 20 µm.) (F) EGFP and AF660 intensity profiles along the axon in neurons expressing EGFP-Nav1.2 II-III linker with following constructs: KCNQ2 and Halo-TAC-KCNQ3 (Q2 + Q3); an empty vector and Halo-TAC-KCNQ3 (Mock + Q3); KCNQ2 and Halo-TAC-KCNQ3-ETD/AAA (Q2 + Q3-ETD/AAA). The fluorescence intensities were normalized to the maximum values of the profiles for EGFP and AF660 in neurons expressing KCNQ2 and Halo-TAC-KCNQ3 (Q2 + Q3). (G–I) AF660 intensities at the AIS and dendrite, ratios of AF660 intensity between AIS and dendrite, and TMR intensities at the AIS. (F–I) n = 6 neurons (Q2 + Q3), 5 neurons (Mock + Q3), and 5 neurons (Q2 + Q3-ETD/AAA). *P < 0.05 and ***P < 0.001 (Dunnett’s test, comparisons were made with the “Q2 + Q3” column for the same group). Data are mean ± SE.

KCNQ3 Functionality Determines AIS Targeting of KCNQ2/3.

To test how functionality of KCNQ3 is critical for AIS targeting of KCNQ2/3, we then genetically manipulated the KCNQ3 functionality. For the manipulation of VSD activation, we mutated the triple amino acid residues that constitute the VSD, namely F168R/D203N/Q234E (FDQ/RNE), to inhibit the transition of VSD conformation from the resting state to the activated state (Fig. 2A) (25). For the manipulation of E-M coupling, we introduced several mutations, namely R243A, W248R, H258N, K260A, Y267C, S343F, and K359A/R365A/K367A (KRK/AAA), to inhibit the interaction between VSD and PD (Fig. 2A) (23–25). Of these, W248, Y267, and S343 residues directly mediate interactions between transmembrane segments (25), while other residues control E-M coupling by binding to phosphatidylinositol 4,5-bisphosphate (PIP₂) (23, 24). Whole-cell patch-clamp recordings in HEK293T cells demonstrated that these mutations decreased the current density and shifted V_1/2 of Halo-TAC-KCNQ3T to more positive potentials (Fig. 2 A and B and SI Appendix, Fig. S3 A–C). Surface expression levels of KCNQ3T and its mutants were comparable in HEK293T cells (SI Appendix, Fig. S4), indicating that the observed functional deficits are not due to altered channel trafficking. Moreover, when coexpressed with KCNQ2, each Halo-TAC-KCNQ3 mutant still increased current density and/or shifted V_1/2 to more positive potentials relative to KCNQ2 homomers (SI Appendix, Fig. S3 D–H), confirming that Halo-TAC-KCNQ3 mutants retain the ability to form heteromeric KCNQ2/3 complexes. In neurons, these Halo-TAC-KCNQ3 mutants failed to efficiently localize to the AIS, although they were expressed in the cytoplasm (Fig. 2 C–F). The AIS localization of Halo-TAC-KCNQ3 and its mutants positively correlated with the current densities of the corresponding Halo-TAC-KCNQ3T constructs measured in HEK293T cells (Fig. 2G). We also examined whether functional impairment of KCNQ3 affects KCNQ2 trafficking. Coexpression of low-activity KCNQ3-H258N significantly impaired Halo-TAC-KCNQ2 targeting to the AIS in a dominant-negative manner (SI Appendix, Fig. S1 D and F–H). Taken together, these data indicate that incorporation of dysfunctional KCNQ3 subunits into heteromeric KCNQ2/3 complexes disrupts their proper AIS localization. Thus, the functionality of KCNQ3, governed by voltage sensing and E-M coupling, is essential for AIS targeting of heteromeric KCNQ2/3 complexes in neurons.

Fig. 2.

KCNQ3 functionality determines AIS targeting of KCNQ2/3. (A) Schematic representation of the KCNQ3 activation process. (B) Representative current traces of Halo-TAC-KCNQ3T and its mutants obtained by whole-cell patch-clamp recordings in HEK293T cells. The dotted line indicates the zero-current level. (C) Representative confocal microscopic images of primary cultured hippocampal neurons expressing EGFP-Nav1.2 II-III linker and KCNQ2 with Halo-TAC-KCNQ3 or its mutants. HaloTag was labeled with AF660 and TMR. The arrow indicates the position of AIS. (Scale bar, 20 µm.) (D–F) AF660 intensities at the AIS and dendrite, ratios of AF660 intensity between AIS and dendrite, and TMR intensities at the AIS of neurons expressing EGFP-Nav1.2 II-III linker and KCNQ2 with Halo-TAC-KCNQ3 or its mutants (n = 5 to 11). *P < 0.05, **P < 0.01, and ***P < 0.001 (Dunnett’s test, comparisons were made with the “WT” column for the same group). (G) Positive correlation between the current densities of the Halo-TAC-KCNQ3T constructs measured in HEK293T cells (SI Appendix, Fig. S3A) and the AIS localization of the corresponding Halo-TAC-KCNQ3 constructs (E). Spearman’s rank correlation coefficient (R) = 0.933. Data are mean ± SE.

Coupling of KCNQ2/3 Functionality to Trafficking Is Not Attributable to Plastic AIS Remodeling Dependent on Neuronal Excitability.

Why do low-activity mutations in KCNQ3 reduce the AIS localization of KCNQ2/3? Previous studies have shown that neuronal excitability critically drives plastic remodeling of AIS structure (33, 34). In particular, inhibition of KCNQ2/3 current rapidly induces AIS plasticity within an hour (35). These suggest that expression of low-activity KCNQ3 mutants may alter neuronal excitability, trigger AIS remodeling, and thereby change the spatial distribution of KCNQ2/3. To test this possibility, we first examined whether expression of functionally impaired KCNQ3 affects intrinsic neuronal excitability. Whole-cell current-clamp recordings showed no significant differences in the firing rate, the resting membrane potential (RMP), and the waveform of action potentials (amplitude and half-width) between neurons expressing wild-type KCNQ3 and those expressing KCNQ3-H258N (Fig. 3 A–G). Therefore, KCNQ3-H258N expression does not significantly affect endogenous KCNQ2/3 and neuronal excitability under our experimental conditions. Next, we examined whether expression of low-activity KCNQ3 mutants affects the AIS structure. The length and start position of AIS structure were quantified using EGFP-Nav1.2 II-III linker. The AIS structure was not rearranged by expression of any of these low-activity mutants (Fig. 3 H–J). In summary, these data indicate that the loss of AIS localization for KCNQ2/3 complexes with low-activity KCNQ3 mutants is not attributable to AIS remodeling driven by altered neuronal excitability. Therefore, the coupling of KCNQ2/3 functionality to trafficking (Fig. 2) is conceptually distinct, both in phenomenon and mechanism, from the well-known form of AIS plasticity in which changes in neuronal activity remodel the entire AIS structure through specific signaling cascades and globally redistribute multiple AIS components, including KCNQ3, Nav channels, FGF14, neurofascin, βIV spectrin, and ankG (33, 35, 36).

Fig. 3.

Coupling of KCNQ2/3 functionality to trafficking is not attributable to plastic AIS remodeling dependent on neuronal excitability. (A) Whole-cell patch-clamp recordings from primary cultured hippocampal neurons expressing KCNQ2 with Halo-TAC-KCNQ3 (WT or H258N). Representative action potentials evoked by 200 pA current injection for 800 ms. The dotted line indicates the −65 mV level. (B and C) Firing rates evoked by current injections with increments of 50 pA (50 pA, P = 0.363; 100 pA, P = 0.466; 150 pA, P = 0.707; 200 pA, P = 0.599; 250 pA, P = 0.677; 300 pA, P = 0.706) and resting membrane potentials (RMP) (P = 0.151) in neurons expressing KCNQ2 with Halo-TAC-KCNQ3 (WT, n = 7; H258N, n = 6). P-values were obtained by Welch’s t test. (D–G) Waveforms, phase plots, maximum amplitudes (P = 0.249), and half-widths (P = 0.114) of spontaneous action potentials at resting membrane potential in neurons expressing KCNQ2 with Halo-TAC-KCNQ3 (WT, n = 5; H258N, n = 5). P-values were obtained by Welch’s t test. (H) Representative confocal microscopic images of primary cultured hippocampal neurons expressing EGFP-Nav1.2 II-III linker and KCNQ2 with Halo-TAC-KCNQ3 or its mutants. The arrow indicates the position of AIS. (Scale bar, 20 µm.) (I and J) The lengths and start positions of EGFP-Nav1.2 II-III linker localized region (n = 5 to 12). P > 0.05 for WT versus all mutants (Dunnett’s test, comparisons were made with the “WT” column). Data are mean ± SE.

AnkG Binding and Functionality of KCNQ3 Govern Insertion, Immobilization, and Lifetime of KCNQ2/3 on the AIS Surface.

To explore the mechanisms underlying the coupling of KCNQ2/3 functionality to trafficking, we performed single-molecule imaging of TMR-labeled Halo-TAC-KCNQ3 and its mutants with total internal reflection fluorescence microscopy (TIRFM) (37–39). This approach enabled us to quantify the trafficking dynamics of KCNQ2/3 throughout neurons, which had not been fully characterized. We then analyzed KCNQ3-ETD/AAA to define the role of ankG binding in each trafficking process of KCNQ2/3. Subsequently, we examined how functional impairment of KCNQ3-H258N perturbs KCNQ2/3 trafficking dynamics. Single KCNQ3 molecules were visualized as fluorescent spots when they appeared on the basal membrane of neurons (Fig. 4A and SI Appendix, Fig. S5 and Movies S1–S3). From the trajectories of spots at the AIS, soma, and dendrites (Fig. 4B and SI Appendix, Figs. S7A and S8A), we quantified the apparent appearance rate constant (k_on^app) of KCNQ2/3 in each compartment by measuring the density of KCNQ3 appearing on the membrane per unit time. The k_on^app reflects the exocytic kinetics of KCNQ2/3. The k_on^app of wild-type KCNQ3 was significantly higher at the AIS than at the somatodendritic region (Fig. 4C and SI Appendix, Figs. S7B and S8B). Both KCNQ3-ETD/AAA and KCNQ3-H258N showed a significant reduction in k_on^app at the AIS compared with wild type, whereas no significant differences were observed at the somatodendritic region (Fig. 4C and SI Appendix, Figs. S7B and S8B). The k_on^app potentially depends on the density of KCNQ3 in the cytoplasmic pool; however, we have confirmed that ETD/AAA and H258N mutations do not significantly alter the protein expression levels of KCNQ3 in the cytoplasm (Figs. 1 and 2). Therefore, the observed differences in k_on^app solely reflect the differences in the efficiency of exocytosis. These findings indicate that both proper ankG binding and functionality of KCNQ3 are required for the exocytic insertion of KCNQ2/3 at the AIS.

Fig. 4.

AnkG binding and functionality of KCNQ3 govern insertion, immobilization, and lifetime of KCNQ2/3 on the AIS surface. (A, Left) A representative TIRFM image of a primary cultured hippocampal neuron expressing EGFP-Nav1.2 II-III linker, KCNQ2, and Halo-TAC-KCNQ3 labeled with TMR. (Scale bar, 5 µm.) (Right) Magnified view of the area within the yellow square in the Left panel, showing the trajectories of Halo-TAC-KCNQ3 molecules for 40 s as yellow lines. (Scale bar, 2 µm.) (B) Representative diffusion trajectories for 40 s of Halo-TAC-KCNQ3 (WT, ETD/AAA, and H258N) labeled with TMR at the AIS. Red, yellow, and blue lines indicate trajectories of immobile, slow-diffusion, and fast-diffusion states of the molecules within an interval of 33 ms, respectively. (Scale bar, 2 µm.) (C) k_on^app of Halo-TAC-KCNQ3 and its mutants on the cell surface at the AIS (WT, n = 10; ETD/AAA, n = 9; H258N, n = 9). *P < 0.05 and **P < 0.01 (Steel–Dwass test, only comparisons with the “WT” column are depicted). (D) Probability density distributions of the displacements of Halo-TAC-KCNQ3 and its mutants during 33 ms at the AIS. (E and F) Diffusion coefficients and occupancies of immobile, slow-diffusion, and fast-diffusion states of Halo-TAC-KCNQ3 and its mutants at the AIS. (D–F) n = 10 (WT); n = 9 (ETD/AAA); n = 7 (H258N). (G and H) Disappearance curves and k_off for immobile, slow-diffusing, and fast-diffusing Halo-TAC-KCNQ3 (WT, n = 10; ETD/AAA, n = 9; H258N, n = 9) on the cell surface at the AIS. (E, F, and H) *P < 0.05 and **P < 0.01 (Steel–Dwass test, only comparisons with the “WT” group for the same state or molecular group are depicted). (C–H) Data are mean ± SE. (I) Fractions (p) for immobile, slow-diffusing, and fast-diffusing Halo-TAC-KCNQ3 (WT, N = 1,474 molecules from 10 neurons; ETD/AAA, N = 698 molecules from 9 neurons; H258N, N = 713 molecules from 9 neurons) on the cell surface at the AIS. **P < 0.01 and ***P < 0.001 (post hoc residual analysis following chi-square test, only comparisons with the “WT” group for the same molecular group are depicted). (J) Schematic models summarizing the differential trafficking dynamics of Halo-TAC-KCNQ3 (WT and H258N) at the AIS. Parameters shown in red indicate significant differences between WT and H258N.

To analyze the lateral diffusion of KCNQ2/3, we fitted the displacement distributions of KCNQ3 (Fig. 4D and SI Appendix, Figs. S7C and S8C) with a sum of two-dimensional diffusion equations and determined that a three-state model is optimal according to the Akaike information criterion (AIC) (SI Appendix, Fig. S6; see SI Appendix, Materials and Methods). Using a hidden Markov model (HMM), we classified KCNQ3 trajectories into immobile, slow-diffusion, and fast-diffusion states (Fig. 4B and SI Appendix, Figs. S7A and S8A). Three diffusion coefficients of wild-type KCNQ3 at the AIS were estimated to be 0.01, 0.03, and 0.11 µm² s⁻¹ for the immobile, slow-diffusion, and fast-diffusion states, with no significant differences compared with the somatodendritic region (Fig. 4E and SI Appendix, Figs. S7D and S8D). In contrast, the immobile state occupancy was significantly higher at the AIS (0.34) than at the soma (0.23) or dendrites (0.22), while the fast-diffusion state occupancy was significantly lower at the AIS (0.16) than at the soma (0.31) and dendrites (0.30) (Fig. 4F and SI Appendix, Figs. S7E and S8E). Mean squared displacement (MSD) analysis further revealed that the KCNQ3 diffusion is more confined at the AIS than at the somatodendritic region. The confinement radii of the immobile and slow-diffusion states of KCNQ3 increased from 33 and 70 nm at the AIS to 172 and 298 nm at the soma and to 142 and 257 nm at the dendrite (SI Appendix, Fig. S9). Both ETD/AAA and H258N mutations decreased immobile state occupancy and increased fast-diffusion state occupancy of KCNQ3 at the AIS (Fig. 4F and SI Appendix, Figs. S7E and S8E), with no significant differences in diffusion coefficients or confinement radii (Fig. 4E and SI Appendix, Figs. S7D, S8D, and S9). These findings not only reconfirm that ankG binding to KCNQ3 is required for KCNQ2/3 immobilization at the AIS (27) but also indicate that proper KCNQ3 functionality is essential.

AnkG binding and functionality of KCNQ3 also affected the lifetime of KCNQ2/3. Because each diffusion state of KCNQ3 exhibited high self-transition probabilities (83 to 98%), individual molecules were classified as immobile, slow-diffusing, or fast-diffusing based on the predominant diffusion state in their trajectories (SI Appendix, Fig. S10). Thus, the observed differences in diffusion state occupancies (Fig. 4F) are attributable to differences in the disappearance rate constants (k_off) and/or the fractions (p) of each category. We quantified the k_off of KCNQ2/3 as the reciprocal of the mean lifetime of KCNQ3 in each category. The k_off reflects the endocytic kinetics of KCNQ2/3. In this analysis, the photobleaching rate constant of TMR (0.150 ± 0.005 s⁻¹, n = 3) sets a conservative lower bound on the observed k_off. Across all subcellular regions, the k_off correlated positively with the diffusion coefficients, while no significant differences in k_off were observed between the AIS and the somatodendritic region (Fig. 4 G and H and SI Appendix, Figs. S7 F and G and S8 F and G). In contrast, the fraction of immobile molecules with the longest lifetime was significantly higher at the AIS (6.31%) than at the soma (3.79%) and the dendrite (3.39%), whereas the fraction of fast-diffusing molecules with the shortest lifetime was significantly lower at the AIS (26.87%) than at the soma (38.94%) and the dendrite (35.36%) (Fig. 4I and SI Appendix, Figs. S7H and S8H). Both ETD/AAA and H258N mutations decreased the fraction of immobile molecules and increased the fraction of fast-diffusing molecules at the AIS without affecting k_off (Fig. 4 G–I and SI Appendix, Figs. S7 F–H and S8 F–H). These results indicate that proper ankG binding and functionality of KCNQ3 are both required to suppress the endocytic internalization of KCNQ2/3 and thereby prolong their lifetime at the AIS.

The above results suggest that physiological modulators of KCNQ2/3 functionality, such as neuronal activity, may also affect its trafficking at the AIS. To test this, we acutely stimulated or silenced neurons with glutamate plus glycine (Glu/Gly) (27) or tetrodotoxin (TTX) (40), respectively, and examined the effects on KCNQ2/3 diffusion at the AIS. Glu/Gly treatment significantly decreased and increased the diffusion coefficients of the immobile and fast-diffusion states, respectively (SI Appendix, Fig. S11A). Nevertheless, the occupancy-weighted average diffusion coefficient remained unchanged (WT, 0.033 ± 0.001 µm² s⁻¹; +Glu/Gly, 0.034 ± 0.003 µm² s⁻¹; P = 0.927, Welch’s t test). By contrast, TTX treatment significantly increased the diffusion coefficient of the fast-diffusion state and significantly decreased and increased the occupancies of the immobile and fast-diffusion states, respectively (SI Appendix, Fig. S11A). This suggests that neuronal silencing may shift KCNQ2/3 toward a resting state, thereby increasing its diffusivity, similar to the low-activity mutant (Fig. 4). However, neuronal activity could also affect KCNQ2/3 diffusivity indirectly by altering AIS cytoskeletal dynamics through specific signaling cascades (33, 35, 36, 40). To test whether the observed pharmacological effects depend on the functional state of KCNQ3, we examined the effects of Glu/Gly and TTX on the diffusivity of a constitutively active KCNQ3 mutant with VSD conformation locked in the activated state, namely KCNQ3-F168R/D203N/R240E (KCNQ3-FDR/RNE) (25). KCNQ3-FDR/RNE localized to the AIS similarly to wild type, but neither Glu/Gly nor TTX significantly altered its diffusivity at the AIS (SI Appendix, Fig. S11 B–D). These findings suggest that pharmacological stimulation or silencing of neuronal activity influences KCNQ2/3 diffusion dynamics at least in part dependent on the functional state of KCNQ3.

Functional State of KCNQ3 Determines Its Binding Affinity for AnkG.

Our single-molecule imaging of Halo-TAC-KCNQ3 illustrates the three-dimensional trafficking dynamics of KCNQ2/3 throughout neurons and shows that both ankG binding and functionality of KCNQ3 are required for the insertion, anchoring, and retention of KCNQ2/3 on the AIS surface (Fig. 4J and SI Appendix, Figs. S7I and S8I). The similar trafficking profile of KCNQ3-ETD/AAA and KCNQ3-H258N raises the hypothesis that loss of KCNQ3 functionality disrupts its interaction with ankG. To test this, we first performed single-molecule imaging of Halo-TAC-KCNQ3 upon coexpression of ankG in HEK293T cells. AnkG coexpression increased the immobile state occupancy of wild-type KCNQ3 from 0.16 to 0.29 without altering its diffusion coefficients (SI Appendix, Fig. S12 A and B). This anchoring effect was absent for the KCNQ2–ankG and KCNQ3–ankyrinB (ankB) pairs and abolished in KCNQ3-ETD/AAA, confirming that our assay detects the preferential and direct interaction between KCNQ3 and ankG (SI Appendix, Fig. S12 C–E). The diffusion dynamics of low-activity KCNQ3 mutants (KCNQ3-FDQ/RNE, KCNQ3-Y267C, and KCNQ3-H258N) also remained unchanged upon ankG coexpression (SI Appendix, Fig. S12 F–H), indicating reduced ankG binding affinity in these mutants.

We further performed dual-color TIRF imaging of Halo-TAC-KCNQ3 labeled with TMR and SNAP-tagged ankG (ankG-SNAP) labeled with 647-SiR expressed in living HEK293T cells (Fig. 5A and Movie S4; see SI Appendix, Materials and Methods). In this measurement, we compared the ankG colocalization kinetics between wild-type KCNQ3 and low-activity KCNQ3-H258N. To separately identify the ankG colocalization kinetics of KCNQ3 in the resting and activated state, we also examined KCNQ3-FDQ/RNE and KCNQ3-FDR/RNE (25). The colocalization probabilities for KCNQ3-H258N and KCNQ3-FDQ/RNE were lower than those for wild-type KCNQ3 and KCNQ3-FDR/RNE, respectively (Fig. 5B). We quantified the association rate constant (k_a) from the frequency of colocalization events, showing that the k_a for KCNQ3-FDQ/RNE was lower than for KCNQ3-FDR/RNE, whereas no significant difference was observed between wild-type KCNQ3 and KCNQ3-H258N (Fig. 5C). The dissociation kinetics was quantified by fitting the dissociation curve generated as the survival function of the colocalized lifetimes with a double exponential function (Fig. 5D). The fraction-weighted average dissociation rate constants (k_d) for KCNQ3-H258N and KCNQ3-FDQ/RNE were higher than those for wild-type KCNQ3 and KCNQ3-FDR/RNE, respectively (Fig. 5E). This decreased k_a and/or increased k_d resulted in a higher dissociation constant (Kd = k_d/k_a), indicating reduced ankG binding affinity of these low-activity mutants (Fig. 5F). Thus, our data demonstrate a molecular mechanism intrinsically embedded in the KCNQ2/3 channel itself: The conformational state of KCNQ3 determines its binding kinetics to ankG (Fig. 6), thereby regulating the AIS localization of KCNQ2/3 (Figs. 2 and 4).

Fig. 5.

Functional state of KCNQ3 determines its binding affinity for ankG. (A) Representative TIRFM images of HEK293T cells expressing Halo-TAC-KCNQ3 labeled with TMR and ankG-SNAP labeled with 647-SiR. (Scale bar, 5 µm.) (B) Colocalization probabilities of Halo-TAC-KCNQ3 and its mutants with ankG-SNAP. P = 0.038 for WT versus H258N; P = 0.034 for FDR/RNE versus FDQ/RNE. (C) k_a of Halo-TAC-KCNQ3 and its mutants colocalized with ankG-SNAP. P = 0.924 for WT versus H258N; P = 0.026 for FDR/RNE versus FDQ/RNE. (D) Dissociation curves of Halo-TAC-KCNQ3 and its mutants from ankG-SNAP which were fitted with a double exponential function. (E) k_d obtained from Halo-TAC-KCNQ3 and its mutants colocalized with ankG-SNAP. P = 0.016 for WT versus H258N; P = 0.007 for FDR/RNE versus FDQ/RNE. (F) Kd calculated as k_d/k_a for Halo-TAC-KCNQ3 and its mutants in (C and E). P = 0.043 for WT versus H258N; P = 0.003 for FDR/RNE versus FDQ/RNE. n = 8 cells (WT), 7 cells (H258N), 6 cells (FDR/RNE), and 7 cells (FDQ/RNE). P-values were obtained by Welch’s t test. Data are mean ± SE.

Fig. 6.

Dynamic equilibrium of the binding reaction between KCNQ3 and ankG. The three kinetic parameters (k_a, k_d, and Kd) for KCNQ3-FDQ/RNE (resting state) and KCNQ3-FDR/RNE (activated state) are shown. The equilibrium between KCNQ3 and ankG depends on the functional state of KCNQ3. KCNQ3 has higher ankG binding affinity in the activated state than in the resting state.

Discussion

Here, we demonstrate the coupling of KCNQ2/3 functionality to trafficking in neurons. We first found that impaired VSD activation and E-M coupling of KCNQ3 cause insufficient localization of KCNQ2/3 at the AIS (Fig. 2). Furthermore, we successfully achieved the visualization and quantification of the spatiotemporal dynamics of KCNQ2/3 at the single-molecule level. This analysis revealed that both impaired ankG binding and reduced functionality of KCNQ3 diminish AIS localization of KCNQ2/3 by affecting all three trafficking processes: suppressing exocytic insertion, enhancing lateral diffusion, and accelerating endocytic internalization on the AIS surface (Fig. 4). Finally, we provide a dual-color TIRF imaging method to assess the binding kinetics between full-length KCNQ3 and ankG in living cells, establishing that the active conformation of KCNQ3 is essential for its stable ankG binding (Figs. 5 and 6).

Kinetic and Molecular Basis for AnkG-Mediated AIS Targeting of KCNQ2/3.

To assess how KCNQ2/3 functionality affects their AIS targeting mechanism, we first quantitatively characterized the fundamental trafficking dynamics of KCNQ2/3 across all subcellular compartments by single-molecule imaging. This confirmed the ankG-mediated immobilization of KCNQ2/3 at the AIS as previously reported (27) (Fig. 4). Of note, the confinement radius of the slow-diffusion state (70 nm) (SI Appendix, Fig. S9) was comparable to half the periodic spacing of actin rings at the AIS (95 nm) (41), suggesting that lateral diffusion of KCNQ2/3 is additionally governed by diffusion barriers formed by the actin meshwork as described by the “picket-and-fence” model (30, 42). Furthermore, we uncovered functional evidence that ankG binding drives the exocytic insertion of KCNQ2/3 while simultaneously inhibiting their endocytic internalization on the AIS surface (Fig. 4). Thus, KCNQ2/3 employs both mechanisms, the ankG-mediated exocytosis reported for Nav1.6 (43) and the ankG-mediated inhibition of endocytosis reported for neurofascin186 (44). The higher ankG binding affinity of KCNQ3 relative to KCNQ2 (SI Appendix, Fig. S12 A–C) (13) accounts for its dominant role in KCNQ2/3 trafficking (Fig. 1 D–I and SI Appendix, Fig. S1 D–H) (9). Moreover, we found that KCNQ3 binds more weakly to ankB than to ankG (SI Appendix, Fig. S12D), despite over 80% sequence conservation across their 24 ankyrin repeats (45). This differential binding affinity explains why KCNQ2/3 is properly confined to the AIS even though ankB is distributed outside the AIS (46). Taken together, these data establish the kinetic and molecular basis underlying preferential KCNQ2/3 targeting to the AIS. Our findings suggest a developmental switch in the AIS targeting pathways of KCNQ2/3 from lateral diffusion at 3 DIV, before the AIS formation (47), to exo/endocytosis at 10 DIV, after the AIS establishment (Fig. 4 and SI Appendix, Figs. S7 and S8). Building on a previous strategy (27), the use of HaloTag enabled simultaneous identification of surface and intracellular KCNQ2/3 populations and allowed us to fully exploit single-molecule imaging analysis. Although these exogenously expressed constructs recapitulated the endogenous AIS localization of KCNQ2/3 (Fig. 1) (8), future endogenous tagging approaches will be important to fully rule out the possibility that overexpression itself may artificially influence their single-molecule dynamics.

Mechanisms Underlying Coupling of KCNQ2/3 Functionality to Trafficking.

Single-molecule imaging demonstrated that KCNQ3 dysfunction impairs insertion, immobilization, and retention of KCNQ2/3 on the AIS surface, thereby reducing their AIS localization (Figs. 2 and 4). Moreover, dual-color TIRF imaging has enabled a live-cell assay to quantify the dissociation constant between full-length AIS proteins and ankyrins, overcoming limitations of the conventional in vitro assay using purified protein fragments (45). This approach revealed the ankG binding kinetics of KCNQ3 dependent on its functional state, establishing the molecular basis coupling KCNQ3 functionality to KCNQ2/3 trafficking (Figs. 5 and 6). This mechanism consistently explains why the reduced surface density of low-activity KCNQ3 was exclusively observed at the ankG-rich AIS, but not at the somatodendritic region (Figs. 2 and 4) or in non-neuronal cells such as HEK293T cells (SI Appendix, Fig. S4), CHO cells (24), and Xenopus oocytes (25). Given that ankG binding modulates KCNQ2/3 functionality (8), it is reasonable to assume that conformational changes in KCNQ2/3 could affect its binding to ankG, either directly or indirectly by altering interactions with regulators of its ankG binding affinity, such as CK2 (13). Nevertheless, the full-length structure of KCNQ3 remains unresolved, and available structures of KCNQ1, KCNQ2, and KCNQ4 lack the C-terminal ABM region (10, 48–51). Therefore, further progress in structural analysis of the KCNQ family is needed to elucidate the KCNQ3 conformational changes enabling ankG binding. The mechanism by which KCNQ3 functionality dictates its ankG binding may serve as a unifying framework for understanding KCNQ2/3 trafficking, which is intricately regulated by diverse sites such as the PD (22), the AB helix for CaM binding (14–19), and the CD helix for heteromer formation (9, 14, 20, 21). Thus, in future experiments, it would be of interest to examine the ankG binding affinities of additional KCNQ2 and KCNQ3 mutants outside the ABM that exhibit trafficking or functionality defects (9, 14–22). Moreover, it is tempting to speculate that physiological modulators of KCNQ2/3 functionality could also affect its trafficking in neurons. Acute stimulation or silencing of neurons differentially affected KCNQ2/3 diffusion at the AIS, in part dependent on the functional state of KCNQ3 (SI Appendix, Fig. S11). This highlights the importance of investigating how different patterns of neuronal activity and other modulators (e.g., PIP₂) (52–54) affect KCNQ2/3 trafficking at high spatiotemporal resolution. Given channel degeneracy (55) and the competitive or synergistic effects of ankG binding proteins on other AIS components (21, 56), it will be necessary to assess the impact of changes in KCNQ2/3 functionality and trafficking on other AIS proteins and to evaluate how KCNQ2/3 functionality–trafficking coupling contributes to AIS plasticity and, ultimately, neuronal activity in various contexts.

Physiological and Pathophysiological Significance of KCNQ2/3 Functionality–Trafficking Coupling.

Dysfunction of KCNQ2/3 has been linked to neurological diseases, including BFNC and EIEE (4–6). Notably, several residues (R243, H258, Y267, and R365) that we identified to regulate the functionality and trafficking of mouse KCNQ3 (Fig. 2) correspond to mutated residues in epilepsy-associated variants of human KCNQ3, specifically R242Q/ W, H257Y, Y266C (57), and R364H/C (ClinVar accession ID: VCV000934080.14; VCV001376796.5; VCV000937869.5; VCV000205964.2; VCV000934997.12; VCV000656938.6). Therefore, our findings highlight a key perspective: Functional impairment of KCNQ2/3 tends to entail trafficking abnormalities, and both could simultaneously contribute to the pathogenesis of epileptic diseases. The coupling of channel functionality to trafficking may hold physiological and pathophysiological significance when defective KCNQ3 is partially expressed as a result of heterozygous mutations, transcriptional/translational errors, or protein misfolding. Low-activity KCNQ3 exhibits reduced ankG binding affinity and is poorly delivered to the AIS surface (Figs. 2 and 4–6), thus avoiding competition with normal channels for ankG binding. This may preserve normal KCNQ2/3 occupancy at the AIS and help maintain neuronal excitability (Fig. 3). In this context, the functionality–trafficking coupling of KCNQ2/3 can serve as a protein surveillance mechanism on the neuronal surface, complementing the ER protein quality control mechanism (58). Conversely, in patients homozygous for KCNQ3 mutations, the combined defects in KCNQ2/3 functionality and trafficking may lead to severe epileptic phenotypes or lethality, which could in part contribute to the rarer incidence of epilepsy-associated KCNQ3 variants compared with KCNQ2 (6). The combined defects further pose a significant challenge to conventional antiepileptic strategies that target only the gating of KCNQ2/3 on the neuronal surface and may underlie drug-resistant epilepsy. However, the functionality–trafficking coupling also suggests that compounds restoring the functionality of intracellular KCNQ3 mutants may simultaneously correct their trafficking defects, offering broad therapeutic avenues. Of note, retigabine (RTG), a first-in-class antiepileptic drug (59), primarily targets KCNQ3 in CA1 pyramidal neurons (60), and rescues the functional impairment of the human KCNQ3-Y266C mutant (25). In the future, it would be interesting to examine the potential of neuronal KCNQ openers, such as RTG, to rescue not only functionality but also ankG binding and AIS localization of epilepsy-associated KCNQ3 variants.

Materials and Methods

DNA Constructs.

Mouse KCNQ2 (852 amino acids; UniProt accession number: B7ZBW1) and mouse KCNQ3 (873 amino acids; UniProt accession number: Q8K3F6) were cloned into the pCS5+ vector for expression in mammalian cells. A HaloTag (Promega) and a single-pass transmembrane domain (TAC) from human IL2RA (UniProt accession number: P01589) were fused to the N-terminus of KCNQ2 and KCNQ3 using the In-Fusion HD Cloning Kit (Takara). Site-directed mutations were introduced into KCNQ2 and KCNQ3 by standard PCR techniques using PrimeSTAR Max DNA Polymerase (Takara). For whole-cell patch-clamp recording of KCNQ3 channels in HEK293T cells, the A316T mutation was introduced to facilitate efficient trafficking and functional expression of the homomeric KCNQ3 channel complex (28). The pl-Synapsin-YFP-NavII-III construct was obtained from Addgene (Plasmid #91246), and the fluorescent tag was replaced with EGFP. EGFP-fused 270 kDa ankG and EGFP-fused 220 kDa ankB were gifts from Vann Bennett (Duke University, Durham, US). EGFP of ankG was replaced with SNAP-tag (New England BioLabs). The DNA sequences of all constructs were confirmed through Sanger sequencing.

Culture of HEK293T Cells.

HEK293T cells were cultured in DMEM (10% FBS, 0.1% penicillin/streptomycin) at 37 °C, 5% CO₂. The cells were passaged every 2 to 3 d. For whole-cell patch-clamp recordings, a total of 300 ng of plasmid was transfected into the cells in a 35-mm dish using Lipofectamine 2000 (Invitrogen). For the measurement of the homomeric KCNQ2 or KCNQ3T, 200 ng of KCNQ2 or Halo-TAC-KCNQ3T (wild type or the mutants) was cotransfected with 100 ng of EGFP. For the measurement of the heteromeric KCNQ2/3 complex, 100 ng each of KCNQ2, Halo-TAC-KCNQ3 (wild type or the mutants), and EGFP were cotransfected. In both cases, whole-cell patch-clamp recordings were performed 24 h posttransfection. For fluorescence imaging, 200 ng of Halo-TAC-KCNQ3 (wild type or the mutants), Halo-TAC-KCNQ3T (wild type or the mutants), or Halo-TAC-KCNQ2 was transfected, and imaging was performed 24 h later. For coexpression of ankyrins, 800 ng of ankG or ankB was cotransfected, and imaging was performed 48 h later. The imaging data in SI Appendix, Fig. S5 A–C were obtained 24 h after cotransfection with 100 ng of EGFP. The cells transfected with the specified plasmid DNA were plated onto glass coverslips (Matsunami) coated with collagen (TMTCC-050, TOYOBO) a few hours before measurement. The subsequent fluorescent labeling and imaging were conducted after the cells adhered to the coverslips.

Primary Culture of Hippocampal Neurons.

The hippocampi were dissected from the brain of fetal C57BL/6 mice (E17). The use of mice was approved by the Animal Care and Use Committee of The University of Osaka. The sex of the mice was not considered for the primary culture of hippocampal neurons, and the tissues derived from both male and female mice were mixed. For digestion, the tissues were washed with PBS on ice, and then treated with a 3 mL PBS solution additionally containing 25 mg glucose, 16.5 mg papain (164-00172, FUJIFILM), 1 mg L-cysteine (10309-41, Nacalai Tesque), 1 mg BSA (A9418-50G, SIGMA), and 1,500 U DNase I (DN25, SIGMA) for 20 min at 37 °C. Subsequently, the digested tissues were washed with 5 mL of DMEM (10% FBS) to halt the enzymatic reaction of papain. The supernatant was subsequently removed, and the tissues were mechanically triturated by pipetting in 1 mL of PBS containing 3 mg MgSO₄·7H₂O and 500 U DNase I Type II. The suspension of dissociated neurons was collected into a new tube containing 5 mL of DMEM (10% FBS) and centrifuged at 1,000×g for 5 min at 4 °C. After removing the supernatant, the pellet was resuspended in Neurobasal medium (Gibco) supplemented with GlutaMAX (Sigma) and B-27 (Gibco). The cell density was adjusted to 2 × 10⁵ cells mL⁻¹, and the 1 mL suspension of neurons was plated onto glass coverslips (Matsunami) coated with poly-L-lysine (28360-14, Nacalai Tesque) and collagen (354236, CORNING) in a 12-well plate. After plating, half of the medium was replaced with fresh medium twice a week. At 3 DIV, 2 µM Ara-C (030-11951, FUJIFILM) was added to the medium to prevent the proliferation of non-neuronal cells. At 7 DIV, a total of 1.2 µg of plasmid DNA was transfected into the neurons using Lipofectamine 2000. Each plasmid DNA (e.g., Halo-TAC-KCNQ3, KCNQ2, EGFP-Nav1.2 II-III linker) was transfected at the same amount (400 ng).

Electrophysiology.

Whole-cell configuration patch-clamp recordings were conducted in HEK293T cells and cultured neurons using an amplifier (EPC 800, HEKA) and a digitizer (Digidata 1550A, Molecular Devices). The signals were sampled at 20 kHz and filtered at 5 kHz. The junction potential was not adjusted. The series resistance compensation was not conducted. Microelectrodes were made from borosilicate glass capillaries (B150-86-10, Sutter Instrument) using a puller (P-97, Sutter Instrument).

For recordings from HEK293T cells, the extracellular solution was composed of 160 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, and 10 mM HEPES (pH 7.4 adjusted with NaOH). The microelectrodes with a resistance of 2 to 5 MΩ were filled with the intracellular solution composed of 160 mM KCl, 5 mM MgCl₂, 2 mM Na₂ATP, and 10 mM HEPES (pH 7.4 adjusted with KOH). The current was examined by holding the membrane potential at −80 mV and applying a 300 ms depolarizing pulse from −80 mV to 60 mV in 10 mV increments every 3 s. For recordings from primary cultured neurons (10 to 13 DIV), the extracellular solution was composed of 160 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 10 mM glucose, 1.2 mM MgCl₂, and 1.8 mM CaCl₂ (pH 7.3 adjusted with NaOH). The microelectrodes with a resistance of 2 to 6 MΩ were filled with the intracellular solution composed of 135 mM KCl, 2 mM EGTA, 1.1 mM CaCl₂, 10 mM HEPES, and 5 mM glucose (pH 7.25 adjusted with KOH). The data were processed and analyzed using Clampfit 10.7 (Molecular Devices).

Subcellular Localization Imaging Under Confocal Microscopy.

For the measurement of HEK293T cells, HaloTag was labeled with 1 µM AF660 (Promega) in medium for 15 min at 37 °C, 5% CO₂. HaloTag was subsequently labeled with 1 µM TMR (Promega) in medium for 15 min at 37 °C, 5% CO₂. The cells on a coverslip in a 35-mm dish were then washed three times with 1 mL of HBSS (H8264, Sigma). The coverslip was mounted in a metal chamber (Attofluor Cell Chamber, Invitrogen). For the measurement of primary cultured neurons (10 DIV), HaloTag was labeled with 1 µM AF660 (Promega) in medium for 15 min at 37 °C, 5% CO₂. HaloTag was subsequently labeled with 1 µM TMR in medium for 15 min at 37 °C, 5% CO₂. The neurons on a coverslip in a 12-well plate were then washed three times with 1 mL of artificial cerebrospinal fluid solution (aCSF; 125 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 10 mM glucose, and 10 mM HEPES, pH 7.4 adjusted with NaOH). The coverslip was mounted in a magnetic chamber (CM-B18-1, Live Cell Instrument).

The spatial distribution of fluorescent molecules in HEK293T cells and neurons was imaged using a confocal laser scanning microscope (FV3000, Olympus) equipped with an oil immersion objective lens (UPLSAPO60X, 60×, Oil, NA 1.42, Olympus). The pixel size was 414 nm.

Single-Molecule Imaging Under TIRFM.

For the measurement of HEK293T cells, HaloTag and SNAP-tag were labeled with 2.5 pM TMR and 5 nM 647-SiR (New England BioLabs), respectively, for 15 min at 37 °C, 5% CO₂. The cells on a coverslip in a 35-mm dish were then washed three times with 1 mL of HBSS and incubated for a further 15 min at 37 °C, 5% CO₂. For the measurement of primary cultured neurons (10 to 11 DIV), HaloTag was labeled with 2.5 pM TMR for 15 min at 37 °C, 5% CO₂. The neurons on a coverslip in a 12-well plate were then washed three times with 1 mL of an aCSF and incubated for a further 15 min at 37 °C, 5% CO₂. The coverslip was mounted in a metal chamber (for HEK293T cells) or a magnetic chamber (for neurons) and then additionally washed two times with 0.5 mL of HBSS (for HEK293T cells) or aCSF (for neurons) before imaging under TIRFM.

Single-molecule imaging was performed using an inverted microscope (IX71, Olympus) equipped with an oil immersion objective lens (UPLAPO100XOHR, 100×, Oil, N.A. 1.50, Olympus) and a 1.6× intermediate magnification lens, resulting in a total magnification of 160× (the pixel size was 100 nm). The molecules labeled with TMR and 647-SiR near the basal surface were excited with 532 nm (OBIS 532 nm, 50 mW, COHERENT) and 637 nm (OBIS 637 nm, 70 mW, COHERENT) lasers through the objective lens by a dichroic mirror (ZT532/640rpc, CHROMA). The images were captured by an electron-multiplying charge-coupled device (EM-CCD) camera (ImagEM, Hamamatsu) after passing through a single band pass filter (FF01-593/40-25, Semrock). Fluorescence images (512 × 512 pixels) were acquired using image software (MetaMorph, Molecular Devices) with an exposure time of 33 ms and an EM gain of 1,200. All measurements were performed at room temperature.

Details of data analysis and statistics are provided in SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Single-molecule imaging of Halo-TAC-KCNQ3 at the AIS. Time-lapse TIRFM images depict single molecules of TMR-labeled Halo-TAC-KCNQ3 (Left, WT; Center, ETD/AAA; Right, H258N) on the surface at the AIS of the primary cultured hippocampal neurons. The images correspond to Fig. 4. The images are displayed at intervals of 33 ms. Time, s. Scale bar, 1 µm.

Movie S2.

Single-molecule imaging of Halo-TAC-KCNQ3 at the soma. Time-lapse TIRFM images depict single molecules of TMR-labeled Halo-TAC-KCNQ3 (Left, WT; Center, ETD/AAA; Right, H258N) on the surface at the soma of the primary cultured hippocampal neurons. The images correspond to Fig. S7. The images are displayed at intervals of 33 ms. Time, s. Scale bar, 1 µm.

Movie S3.

Single-molecule imaging of Halo-TAC-KCNQ3 at the dendrite. Time-lapse TIRFM images depict single molecules of TMR-labeled Halo-TAC-KCNQ3 (Left, WT; Center, ETD/AAA; Right, H258N) on the surface at the dendrite of the primary cultured hippocampal neurons. The images correspond to Fig. S8. The images are displayed at intervals of 33 ms. Time, s. Scale bar, 1 µm.

Movie S4.

Single-molecule imaging of Halo-TAC-KCNQ3 and ankG-SNAP in the HEK293T cell. Time-lapse TIRFM images depict single molecules of Halo-TAC-KCNQ3 labeled with TMR (green) and ankG-SNAP labeled with 647-SiR (magenta) on the surface of the HEK293T cell. The images correspond to Fig. 5. The images are displayed at intervals of 33 ms. Time, s. Scale bar, 5 µm.

Data, Materials, and Software Availability

Custom Python code for analyzing the diffusion dynamics from single-molecule trajectories using an HMM has been deposited in GitHub (https://github.com/DaisukeYoshioka99/HMM) (61). Study data are included in the article and/or supporting information.