Nanometer-Scale Imaging of Compartment-Specific Localization and Dynamics of Voltage-Gated Sodium Channels

Membrane excitability and cell-to-cell communication in the brain are tightly regulated by diverse ion channels and receptor proteins localized to distinct membrane compartments. Currently, a major technical barrier in cellular neuroscience is lack of reliable methods to label these membrane proteins and image their sub-cellular localization and dynamics. To overcome this challenge, we devised optical imaging strategies that enable systematic characterization of subcellular composition, relative abundances and trafficking dynamics of membrane proteins at nanometer scales in cultured neurons as well as in the brain. Using these methods, we revealed exquisite developmental regulation of subcellular distributions of voltage-gated sodium channel (VGSC) Nav1.2 and Nav1.6, settling a decade long debate regarding the molecular identity of sodium channels in dendrites. In addition, we discovered a previously uncharacterized trafficking pathway that targets Nav1.2 to unmyelinated fragments in the distal axon. Myelination counteracts this pathway, facilitating the installment of Nav1.6 as the dominant VGSC in the axon. Together, these imaging approaches unveiled compartment-specific trafficking mechanisms underpinning differential membrane distributions of VGSCs and open avenues to decipher how membrane protein localization and dynamics contribute to neural computation in the brain.

channels in dendrites. In addition, we discovered a previously uncharacterized trafficking pathway 23 that targets Nav1.2 to unmyelinated fragments in the distal axon. Myelination counteracts this 24 pathway, facilitating the installment of Nav1.6 as the dominant VGSC in the axon. Together, these 25 imaging approaches unveiled compartment-specific trafficking mechanisms underpinning 26 differential membrane distributions of VGSCs and open avenues to decipher how membrane 27 protein localization and dynamics contribute to neural computation in the brain. 28

INTRODUCTION 29
Information processing in the brain is regulated at the molecular level by diverse membrane 30 proteins such as ion channels and receptor proteins (Catterall, Goldin, & Waxman, 2005;Hodgkin 31 & Huxley, 1952). Two broad types of ion channels are: voltage-gated ion channels and ligand-32 gated ion channels. In mammals, it was estimated that ~140 genes encode voltage-gated K + , Na + 33 and Ca 2+ channels (Yu, Yarov-Yarovoy, Gutman, & Catterall, 2005). In addition to ion channels, Currently, one of the major challenges to probe membrane proteins in the brain is lack of 44 reliable methods to label endogenous ion channel or receptor proteins. Specifically, traditional 45 immuno-labeling is associated with several limitations: 1) nonspecific cross-reaction, especially 46 for antibodies against closely related channels and receptors; 2) insufficient sensitivity when the 47 copy number of the target protein is low; 3) subcellular localization information is obscured by 48 high packing density of neurites in the brain (Mikuni, Nishiyama, Sun, Kamasawa, & Yasuda, 49 2016). As a result, mapping sub-cellular localization of membrane proteins poses tremendous 50 challenges for neuroscientists (Baker, 2020). 51 To address these limitations, here we combined CRISPR/Cas9 in vivo genome editing with 52 high affinity peptide tags (V5 (GKPIPNPLLGLDST) or HA (YPYDVPDYA)) and self-labeling 53 tags (e.g. HaloTag) to label membrane proteins. Sparse cell labeling and high sensitivity of 54 monoclonal antibodies enable us to reconstruct subcellular localizations of membrane proteins 55 with high spatial resolution. Using brain-enriched voltage-gated sodium channel Nav1.2 and 56 Nav1.6 as the model, we found that Nav1.2 is highly enriched in the AIS, dendrites and 57 unmyelinated distal axon branches during early development. As animals develop into adults, 58 Nav1.6 levels increase while Nav1.2 levels decrease in dendrites, accompanied by myelination 59 dependent exclusion of Nav1.2 from the axon and an eventual installment of Nav1.6 as the 60 dominant VGSC at the AIS and nodes of Ranvier. Super resolution and live-cell single molecule 61 imaging in cultured neurons enables real time investigation of VGSC trafficking dynamics at 62 nanometer scales. We found that while localization of Nav1.2 and Nav1.6 to the AIS is dependent 63 on Ankyrin G binding domain (ABD) as previously described (Garrido et al., 2003;Lemaillet, 64 Walker, & Lambert, 2003), the targeting of Nav1.2 to unmyelinated fragments in the distal axon 65 requires separated signals within the intracellular loop 1 (ICL1) between transmembrane domain 66 I and II. Specifically, Nav1.2 ICL1 suppresses AIS retention and permits the membrane loading of 67 Nav1.2 at the distal axon. Together, these results unveiled compartment-specific localization and 68 trafficking mechanisms for Nav1.2 and Nav1.6, which could be modulated independently to fine 69 tune membrane composition and physiological functions of VGSCs in the brain. 70 the tag insertion did not affect electrophysiological properties of Nav1.2 and Nav1.6 (Figure 1-94 figure supplement 5). Consistent with previous immune-staining results (Xu,Zhong,& Zhuang,95 2013), super-resolution STED imaging revealed that tagged Nav1.2 and Nav1.6 form ~200 nm 96 periodic striations that showed anti-phased exclusion from actin rings at the AIS, further validating 97 the labeling strategy (Figure 1B, C). 98 To quantify relative abundance of Nav1.2 and Nav1.6 in distinct neuronal compartments, 99 we tagged both channels with the V5 tag, followed by labeling and imaging under the same 100 condition. By cross referencing with AIS (Ankyrin G) and dendrite (MAP2) markers, we found 101 that both channels showed highest enrichment in the AIS (Figure 1-figure supplement 3), 102 consistent with previous reports (Hu et al., 2009;Lorincz & Nusser, 2010). Interestingly however, 103 we found that the relative abundance of Nav1.2 are much higher than Nav1.6 in the distal axon and 104 dendrites ( Figure 1D, Figure 1-figure supplement 4). To confirm that what observed are not 105 influenced by cell-type specific expression, we employed sequential HITI editing and achieved 106 dual labeling of Nav1.2 (V5) and Nav1.6 (HA) in the same cell population (Figure 1-figure  107 supplement 1B). Nav1.2 and Nav1.6 staining patterns in the dual labeling condition were 108 consistent with what were observed in separate populations with highest levels of Nav1.2 and 109 Nav1.6 in the AIS and Nav1.2 as the dominant VGSC in the distal axon and dendrites ( Figure 1A, 110 Nav1.2 is indeed inserted into cell membrane in the distal axon and dendrites (Figure 1-figure  112 supplement 2). Thus, here we were able to unambiguously confirm the localization of Nav1.2 in 113 dendrites of cultured hippocampal pyramidal neurons. 114 The differential localization patterns of Nav1.2 and Nav1.6 prompted us to probe underlying 115 trafficking mechanisms. Super-resolution Airyscan imaging and computer-aid segmentation revealed no co-labelled fraction between Nav1.2 and Nav1.6 positive trafficking vesicles ( Figure  117 1E, Video 2), suggesting that once synthesized, Nav1.2 and Nav1.6 are sorted into distinct vesicle 118 populations potentially coupled with separated trafficking and membrane loading pathways. 119 120 Developmental regulation of Nav1.2 and Nav1.6 subcellular localizations in vivo 121 To map Nav1.2 and Nav1.6 localizations in vivo, we used in utero electroporation to deliver the 122 HITI construct into heterozygous H11-SpCas9 mouse embryos expressing Cas9 in all cell types 123 Because both Nav1.2 and Nav1.6 were tagged with the V5 peptide, we were able to estimate 131 their relative abundances with high spatial resolution. At P15, both Nav1.2 and Nav1.6 were 132 Nav1.6 is concentrated at the distal part of the AIS with a ~15 µm gap between their concentration 136 peaks ( Figure 2F), consistent with a previous report (Hu et al., 2009). Interestingly however, 137 Nav1.2 levels decreased significantly at the proximal AIS with the Nav1.6 concentration peak 138 shifting inwards at P30 and P90 (Figure 2-figure supplement 1).
We were also able to confirm the localization of Nav1.2 in dendrites of cortical and 140 hippocampal pyramidal neurons in vivo (Figure 2A, Figure 2-figure supplement 2A). In addition, 141 we found that Nav1.2 is the dominant VGSC in dendrites during early development and its 142 concentration gradually decreases, accompanied by an increase of Nav1.6 levels at this region as 143 mice mature ( Figure 2G). 144 Previous electrophysiology experiments revealed that Nav1.6 has much lower activation 145 threshold and larger persistent currents than Nav1. One consistent observation across all developmental stages is that the axonal coverage by Nav1.2 156 is largely uninterrupted in neurons with high Nav1.2 expression levels, suggesting that these cells 157 are unmyelinated (Figure 2A, B, Video 3). Indeed, when using myelin basic protein (MBP) to co-158 stain samples, we found that Nav1.2 was preferentially expressed (~60%) in unmyelinated neurons 159 with a smaller fraction (~35%) of detectable expression in partially myelinated neurons and the 160 lowest fraction (~5%) in fully myelinated neurons ( Figure 2B, C, E). By contrast, Nav1.6 has 161 similar fractions of detectable expression across all 3 populations ( Figure 2E). In addition, we found that the axonal coverage by Nav1.6 is restricted to Ankyrin G positive regions such as the 163 AIS and nodes of Ranvier ( Figure 2D, Video 4 and Video 5), whereas Nav1.2 broadly covers 164 unmyelinated axonal fragments ( Figure 2B, Video 3). Our results support that localizations and 165 expression levels of these two channels alter with the myelination status of a neuron. Specifically, 166 myelination excludes Nav1.2 and decreases its expression levels in the axon, with an eventual 167 installment of Nav1.6 as the dominant VGSC at the AIS and nodes of Ranvier in fully myelinated 168 neurons. Previous reports showed that, upon neuronal injury, the large persistent currents of Nav1.6 169 at demyelinated sites trigger reverse action of Na + -Ca 2+ exchanger, leading to Ca 2+ influx that 170 further damages the axon (Craner et al., 2004;Rush et al., 2005). This result explains the 171 physiological need of coating unmyelinated axonal fragments with a VGSC that conducts smaller 172 persistent currents such as Nav1.2. 173 174 Compartment-specific Targeting Mechanisms for Nav1.2 and Nav1.6 175 The differential localization patterns and separated vesicle populations associated with Nav1.2 and 176 Nav1.6 suggest that they are trafficked by different pathways (Figure 1E and 2A). To study the 177 underlying mechanism, we established a 2-color imaging assay in which Nav1.2 and Nav1.6 178 (tagged with V5 and HA respectively) were co-expressed in cultured hippocampal neurons where 179 their localization patterns can be directly compared in distinct subcellular compartments. We found 180 that exogenously expressed Nav1.2(V5) and Nav1.6(HA) displayed similar localization patterns as  To further dissect the function of ICL1, we fused it to GFP-P2A-mCherry. Strikingly, we 198 found that Nav1.2 ICL1 itself was able to broadly target GFP to cell membrane across different 199 compartments (soma, axon and dendrites) ( Figure 3D To build a physical model to explain how differential subcellular localizations of Nav1.2 and 209 Nav1.6 are dynamically established at the molecular level, we sought to utilize live-cell single-210 molecule imaging approaches that we established previously (Chen et  terminus of these two VGSCs, followed by staining with bright, membrane permeable Janelia 215 Fluor dyes (Grimm et al., 2015). We found that localization patterns of HaloTag-labeled Nav1.2 216 and Nav1.6 in the AIS were similar to these tagged with V5 and HA tag ( To establish a simple and effective method to quantify dynamic states (stable binding, local 224 exploration, diffusion and active transport) associated with trafficking and membrane loading, we 225 took advantage of the "Radius of Confinement (RC)" parameter which we used successfully to 226 study binding and diffusion states of diverse transcription factors (Lerner et al., 2020). Specifically, 227 the RC is defined as the distance from the center of mass to the furthest point of the trajectory 228 ( Figure 4B). Intuitively, fast diffusion and active transport events along the neurites should 229 correlate with larger RCs compared with bound and local exploration states (Figure 4B, C). Indeed, 230 we found that ABD deletion in Nav1.2 or Nav1.6 led to a dramatic reduction of shorter RC fractions and an increase in longer RC fractions, reflecting less binding but more active transport events in 232 the AIS, consistent with known functions of ABD ( Figure 4C). Conversely, inhibiting active 233 transport by ATP analog (AMPPNP) significantly reduced active transport (longer RC) fractions 234 but increased short RC fractions in distal axon (Figure 4C), confirming the ability of the RC 235 analysis to separate distinct dynamic states. 236 To dissect the molecular basis underlying each dynamic state, we next coupled the RC 237 analysis with genetic perturbations. We found that Nav1.2 displayed significantly less binding and 238 more active transport events in AIS than Nav1.6 ( Figure 4C). Similarly, replacing ICL1 in Nav1.6 239 with Nav1.2 ICL1 deceased binding and induced more active transport in the AIS and soma. The 240 opposite is true as Nav1.2 with Nav1.6 ICL1 has more binding but less active transport events than 241 Nav1.2 ( Figure 4D). The remarkable consistency in these results support that Nav1.2 ICL1 242 promotes active transport and suppresses retention in the AIS, counterbalancing the anchoring 243 effect of ABD. Complementary with these results, we found that Nav1.2 with the membrane 244 anchoring domain ICL-36aa (AA725-760) replaced with the same region from Nav1.6 showed 245 much less binding at the distal axon, suggesting that this domain is critical for membrane insertion 246 of Nav1.2 ( Figure 4E), consistent with its ability to anchor GFP to cell membrane. Taken together, 247 these results suggest that localization of Nav1.2 to the distal axon requires two distinct functions 248 of ICL1: one for reducing anchoring at the AIS; the other for promoting membrane insertion at the 249 distal axon (Figure 4G). 250 Next, we used fluorescence recovery after photobleaching (FRAP) to examine lateral 251 diffusion of membrane bound Nav1.2 and Nav1.6 across different compartments. For this assay, 252 we utilized Pitstop 2 to block endocytosis mediated exchanges on the membrane so that fluorescent 253 recovery is largely dependent on lateral diffusion. We found that both Nav1.2 and Nav1.6 showed slow exchanging rates at the AIS with only ~15% recovery 1 hour after photobleaching, whereas, 255 in the distal axon and dendrites, VGSC is more dynamic, with ~60 percent recovery in the distal 256 axon and ~50 percent recovery in the dendrites ~0.5 hour after photobleaching ( Figure 4F, Video  257   6). These results are consistent with that lateral diffusion of VGSCs is also regulated in a 258 compartment-specific fashion ranging from minimal mobility in the AIS to faster diffusion in 259 dendrites and the distal axon ( Figure 4G). It is likely that the reduced VGSC lateral diffusion in 260 the AIS could be related to the unique, ring-like cytoskeleton structures at this region showing 261 anti-phase, exclusive distributions to VGSC striations ( Figure 1B, C). 262

DISCUSSION 263
Here, we demonstrated nanometer-scale imaging strategies to characterize sub-cellular 264 localization, relative abundances and tracking dynamics of membrane proteins in the brain. We  Lemaillet et al., 2003). Interestingly, we found that separated signals located in ICL1 are responsible for targeting and membrane loading of Nav1.2 to/at the distal axon. 286 Strikingly, Nav1.6 with Nav1.2 ICL1 gained access to the distal axon. Nav1.2 ICL1 alone targets 287 GFP molecules to cell membrane. Single molecule imaging revealed that Nav1.2 ICL1 promotes 288 active transport, suppresses retention at the AIS and promotes membrane loading at the distal axon.

Primary Culture of Hippocampal Neurons 320
We prepared dissociated hippocampal neurons from P0 to 1 Sprague-Dawley rat or C57Bl/6 mouse 321 pups. Briefly, the hippocampi were dissected out and digested with papain (Worthington 322 Biochemical). After digestion, the tissues were gently triturated and filtered with the cell strainer. 323 The cell density was counted and ~2.5 × 10 5 cells were transfected with indicated constructs by 324 using P3 Primary Cell 4D-Nucleofector X kit (Lonza). After transfection, neurons were plated 325 serum in a 37°C incubator with 5% CO2 and were grown in 60-mm culture dishes. Plasmids 375 encoding wild type (WT) or V5-labeled Nav1.2, or wild type Nav1.6 or V5-labeled Nav1.6 (4 μg) 376 were co-transfected with Scn1b (2 µg), Scn2b (2 µg) and eGFP (0.3 µg) using Lipofectamine 2000 377 (ThermoFisher Scientific). Whole-cell voltage-gated sodium (Na + ) currents were measured 48 378 hours after transfection at room temperature under voltage patch-clamp configuration with an 379 Axopatch 200B amplifier (Molecular Devices) and sampled at 10 kHz and filtered at 2 kHz. Na + 380 currents were elicited with a 50 ms depolarization step from -100 mV with 5 mV increment at a 381 holding potential of -100 mV. Steady-state inactivation were tested by a two-pulse protocol with 382 the first pulse of 500 ms from -100 mV to -10 mV at 5 mV increment followed by a second pulse 383 fixed at -10 mV. Gating activation and steady-state inactivation curves were obtained using a 384 Comparisons between two groups were performed with Student's t test. Comparisons among 432 multiple groups were performed with one-way ANOVA and post hoc Bonferroni test. Differences 433 were considered to reach statistical significance when p < 0.05. 434 435