Nanoscale architecture and dynamics of CaV1.3 channel clusters in cardiac myocytes revealed by single channel nanoscopy

16 The clustering of L-type calcium channels in cardiac myocytes presents an important mechanism for 17 functional regulation of calcium signaling. Here we applied targeted super-resolution imaging tech-18 niques for the study of atrial-specific CaV1.3 channel clusters in human iPSC-derived atrial cardiomyo-19 cytes (hiPSC-aCM). We thereby clarified cluster localization, dimensions, architecture, and dynamics, 20 which were largely unexplored previously. Live-cell STimulated Emission Depletion (STED) imaging 21 identified that cell surface-localized clusters contained 9 channel molecules within 120 nm diameter 22 on average. DNA Points Accumulation for Imaging in Nanoscale Topography (DNA-PAINT) optimized 23 for molecular mapping revealed an irregular arrangement of channels with significant spacing. Single 24 Particle Tracking (SPT) further evidenced that clustered channels do not associate into rigidly packed 25 structures (oligomers or lattices), but rather co-diffuse in confined and stationary membrane 26 nanodomains. Immunofluorescence showed consistent cell-surface colocalization with Ryanodine Re-27 ceptor type 2 and Junctophilin-2 forming stable calcium release units, similar to dyadic junctions con-28 taining CaV1.2 in ventricular cardiomyocytes. Lastly, novel genetic constructs for live-cell imaging 29 showed that the cytosolic C-terminal tail of CaV1.3 by itself is sufficient for cluster formation. In con-30 clusion, a novel strategy for LTCC clustering studies in atrial cells was established, suitable for a wide 31 range of super-resolution imaging techniques. Based on live-cell STED, DNA-PAINT and SPT data, we 32 propose that CaV1.3 channel clusters consist of mobile individual channels inside defined membrane 33 nanodomains. 34 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint

35 L-type calcium channels (LTCC) are essential for maintenance and regulation of heart contractility. In 36 cardiac myocytes, LTCC opening is triggered by action potential depolarization and leads to rapid, 37 transmembrane calcium influx and subsequent myofilament activation. As one of two cardiac LTCC, 38 CaV1.3 drives both pacemaking and contractility (Zaveri et al., 2023), supported by its selective ex-39 pression in atrial cells and activation at more negative potentials compared to CaV1.2 (Z. Zhang et al., 40 2005). LTCC form subdiffraction-sized clusters, which facilitate calcium release, cooperative gating and 41 protein interactions (Dixon et al., 2022). Previous studies on CaV1.3 clustering in neurons showed that 42 alternative splicing affects CaV1.3 function and cluster formation possibly through C-terminal protein 43 interactions with Calmodulin (CaM; Moreno et al., 2016), PDZ-binding proteins (Jenkins et al., 2010; 44 Stanika et al., 2016; Yang et al., 2023) and Junctophilin isoforms (Sahu et al., 2019). These findings 45 point towards analogous modulatory mechanisms governing CaV1.3 channel function in cardiomyo-46 cytes. 47 Recent studies revealed new regulatory mechanisms of the cardiac channel homolog CaV1.2 (Del Vil-48 lar et al., 2021; Liu et al., 2020) and it has become clear that the ‘classic’ model of functional upregu-49 lation by direct channel phosphorylation is incorrect: β-adrenergic upregulation of CaV1.2 currents is 50 mediated by the small GTPase Rad, even when all potential phosphorylation sites on the α- and β-51 channel subunits have been removed (Papa et al., 2022). An alternative and possibly converging 52 model of LTCC regulation involves the modulation of cooperativity via channel clustering (Baudel et 53 al., 2022; Del Villar et al., 2021), but the mechanisms and molecular dynamics of channel clustering 54 are not well understood. Earlier studies by the Santana group showed dimer-like bridging of channel 55 C-termini mediated by Calmodulin (Dixon et al., 2015; Moreno et al., 2016) and proposed a stochastic 56 self-assembly model of cluster formation (Sato et al., 2019), however there is no direct experimental 57 evidence for oligomerization-like cluster assemblies. Recently, increased clustering upon phosphoryla-58 tion of the C-terminal CaV1.2 residue S1928 by PKA in vascular cells was reported (Baudel et al., 59 2022). Clustering of the CaV1.3 isoform in cardiomyocytes was so far not characterized, as the channel 60 is not expressed in ventricular cells and presents the challenge of combining adequate cell isolation, 61 protein labeling and super resolution microscopy. 62 In this study, we show that human induced pluripotent stem cell-derived atrial cardiomyocytes 63 (hiPSC-aCM) expressing tagged CaV1.3 channels present a valuable experimental approach for unrav-64 eling LTCC clustering mechanisms. We introduce a HaloTag-CaV1.3 fusion protein for live-cell STimu-65 lated Emission Depletion (STED) imaging and Single Particle Tracking (SPT). Further, we use a corre-66 sponding GFP fusion protein to perform DNA Points Accumulation for Imaging in Nanoscale Topogra-67 phy (DNA-PAINT) at molecular-scale resolution (Schnitzbauer et al., 2017), which was not previously 68 reached in LTCC imaging studies. Combining the results of these super-resolution imaging techniques, 69 we address the molecular spatial arrangement of clustered channels within membrane nanodomains, 70 which was previously unclarified for LTCC in contrast to other ion channels. These findings serve as an 71 important foundation for future studies aiming to correlate cluster structure and its modulation with 72 functional readouts. 73 74 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint

75 Halo-tagged CaV1.3 channels form clusters in the plasma membrane of atrial cardiomyocytes 76 Human induced pluripotent stem cell-derived atrial cardiomyocytes (hiPSC-aCM) were used as a 77 model for the cellular physiology of atrial heart muscle (Cyganek et al., 2018). To investigate the clus-78 tering of CaV1.3 calcium channels, an expression construct encoding the pore-forming subunit α1D of 79 human CaV1.3 fused to an N-terminal HaloTag (Halo-CaV1.3) was expressed in hiPSC-aCM using transi-80 ent transfection (Figure 1A, upper left). While the N-terminal tagging position was previously re-81 ported as functionally inert (Grabner et al., 1998; Liu et al., 2020; H. Zhang et al., 2005), we confirmed 82 a physiological voltage-current response for our construct by patch clamp measurement (Figure S1). 83 For live-cell imaging of hiPSC-aCM expressing Halo-CaV1.3, the cells were labeled with a cell-permea-84 ble HaloTag ligand (HTL) conjugated to a fluorogenic JF646 fluorophore. Subsequent confocal imaging 85 revealed spot-like signals (Fig. 1A, magenta) representing CaV1.3 channel clusters. The signals local-86 ized predominantly to the plasma membrane, which was confirmed by co-staining with fluorescently 87 labeled Cholesterol (green). Spot-like CaV1.3 signals were only observed in transfected cells and not in 88 apparently untransfected neighboring cells, highlighting the specificity of our labeling approach. The 89 focal plane was then shifted to the coverslip-adherent plasma membrane of cardiomyocytes, result-90 ing in the highest density of spot-like Halo-CaV1.3 signals. To accurately resolve individual clusters be-91 low the diffraction limit, super-resolution STED imaging was applied (Fig. 1B). Compared to confocal 92 imaging, more distinct and smaller signal shapes were detected by STED, presumably representing 93 individual clusters. Interestingly, cluster signals often appeared in grouped arrangements, which cor-94 responded to unresolved, single spots in the confocal image. 95 To quantify the abundance and size of cluster signals, image analysis was applied to a larger dataset 96 of equivalently recorded STED images (Fig. 1C). Due to the high variability of signal spot intensities, 97 thresholding-based methods were insufficient for segmentation and a custom approach based on ro-98 bust peak finding and expansion was implemented (see Methods section). This segmentation method 99 led to reliable cluster detection even for low-intensity or directly adjacent signals, resulting in a spa-100 tial density of 2.0 ± 0.5 (mean ± s.d.) clusters/µm². Exemplary segmentation outlines are presented 101 on raw image data in Fig. 1C. Across the dataset, these outlines encompassed cluster areas of 0.013 ± 102 0.008 µm² corresponding to equivalent diameters of 122 ± 35 nm when assuming circular shapes. The 103 histogram of measured areas shows a right-skewed frequency distribution, demonstrating that cluster 104 diameters were typically around 100 nm and rarely exceeded 200 nm. 105 Using the same image data, molecular counting of dye molecules was applied to segmented clusters 106 by referencing their brightness against calibration samples with defined dye numbers (Fig. S2). Halo-107 Tag labeling is well-suited for this approach, since precisely one dye molecule is covalently bound to 108 each labeled channel. Hence the approximate number of CaV1.3 channels within each cluster signal 109 can be extracted with low statistical variance, despite not considering that a fraction of channels is 110 unlabeled. The mean background signal surrounding each cluster signal was subtracted from the con-111 tained signal intensity. As a result, we found 9 ± 12 channels per cluster in a right-skewed distribution 112 (median = 5, Fig. 1D). By relating the channel count of each cluster to its area, we were able to calcu-113 late intra-cluster channel densities, which amounted to 612 channels/µm². Notably this result greatly 114 differs from a theoretical limit of ~ 10,000 channels/µm² expected for oligomer-like dense channel 115 packing, which was not nearly reached in our measurement (99th percentile: 1625 channels/µm²). 116 DNA-PAINT resolves channel arrangements and confirms loosely packed cluster structure 117 The molecular architecture of LTCC clusters has not been resolved thus far. Ground State Depletion 118 (GSD) and STochastic Optical Reconstruction Microscopy (STORM) were previously used for super-res-119 olution cluster imaging but did not reach true molecular-scale resolution. Moreover, antibody-based 120 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint labeling has been a hindering factor, due to the large physical displacement (so-called linkage error) 121 between label and epitope. We aimed to surpass these limitations by combining direct CaV1.3 chan-122 nel tagging with DNA-PAINT, a technique that reaches molecular-scale resolution through the use of 123 exchangeable fluorophores – the main limiting factor in single-molecule fluorescence microscopy 124 (Schnitzbauer et al., 2017). 125 Analogous to our Halo-CaV1.3 construct, we expressed GFP-CaV1.3 in hiPSC-aCM and labeled for DNA-126 PAINT in fixed cells using the commonly used GFP nanobody (Fig. 2A; Ries et al., 2012; Sograte-Idrissi 127 et al., 2019). Imaging was performed using a custom-built Total Internal Reflection Fluorescence 128 (TIRF) setup with single-molecule sensitivity to image CaV1.3 channels selectively in the coverslip-ad-129 herent plasma membrane. Single molecule binding events of Atto 643 (or Atto 550)-labeled imager to 130 its complementary docking strand were highly specific and sparsely distributed (Fig. 2B). The emitter 131 positions were localized over time series of 30,000 to 50,000 frames to build a super-resolution image 132 reconstruction. The resulting DNA-PAINT images showed clustered signal distributions, which were in 133 full agreement with GFP fluorescent signals recorded at diffraction-limited resolution, confirming the 134 specificity of DNA-PAINT binding events. A magnified image region shown in Fig. 2C demonstrates 135 that clusters of GFP-CaV1.3 were super-resolved by DNA-PAINT, leading to groups of puncta corre-136 sponding to each diffraction-limited GFP-fluorescence spot. Individual cluster magnifications (Fig. 2D) 137 revealed a disordered arrangement of clearly separable puncta. Since puncta appearance was mostly 138 uniform and non-overlapping, countable puncta were assumed to reflect single channel positions. 139 We benchmarked our reconstruction quality by Nearest Neighbor Analysis (NeNa; Endesfelder et al., 140 2014) as the basis for further optimization. For initial reconstructions, a localization precision σ = 12.5 141 ± 1 nm was measured (Fig. 2E), which was improved to 9.8 ± 1.2 nm after applying drift and vibration 142 correction based on a recently published algorithm (Cnossen et al., 2021). Moreover, localization 143 merging (as described by Martens et al., 2021) and filtering (see Methods section) led to a drastic im-144 provement of localization precision to 4.1 ± 0.3 nm, which was deemed sufficiently small to resolve 145 individual CaV1.3 channels with a channel diameter of 10 nm and hence similar expected minimal 146 spacing (Yao et al., 2022). Along with these optimization steps, we observed a successive improve-147 ment of resolution in the reconstructions without a noticeable loss of spot detection sensitivity (Fig. 148 S3A). 149 To characterize clustering quantitatively, molecular mapping was performed by identification of signal 150 maxima in DNA-PAINT reconstructions, presumably indicating single channel positions (Fig. 2F). Nota-151 bly, adjacent maxima with a distance of 12 nm were reliably resolved (white arrows in Fig. 2F). Chan-152 nel positions were then subjected to DBSCAN clustering (Ester et al., 1996; Siddig et al., 2020). The 153 optimal parameter value ε = 100 nm was chosen and used to detect the highest number of clusters 154 (2.5 ± 0.5 µm-2), whereas increased ε values led to merging of adjacent clusters and increased vari-155 ance (Fig. S3B,C). We note that the detected cluster density is close to the value of 2.0 µm-2 obtained 156 by STED-based cluster analysis in living cells. Using the obtained molecular maps, we computed me-157 dian nearest-neighbor distances (NND) of 47 nm and 32 nm when considering all or only clustered 158 channels, respectively. The frequency distribution of NND for all channels (Fig. 2G) showed a local 159 plateau at 100 nm, thus reaffirming the chosen ε value. Notably, only 19% of clustered channels were 160 in close mutual proximity defined by NND values below 20 nm. Lastly, we quantified DBSCAN-based 161 cluster detections, which were defined by areas of 0.013 ± 0.020 µm² (Fig. 2H) containing 7 ± 7 chan-162 nel spots (Fig. 2I), with both distributions showing an exponential falloff. Taken together, these results 163 cross-validate our STED-based cluster analysis (see Table 1) and support a model of a widely spaced, 164 disordered distribution of clustered CaV1.3 channels. 165 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint CaV1.3 channels are laterally mobile despite static cluster positions 166 To explore the dynamics of individual CaV1.3 channels within clusters, a dual labeling approach was 167 introduced for living hiPSC-aCM expressing Halo-CaV1.3: First, sparse labeling of single channels was 168 attained by application of HTL-JF646 at minimal concentration (see Methods section). Directly after, a 169 concurrent ensemble labeling of clusters was achieved by applying HTL-JF549 at saturating concentra-170 tion. The resulting signal distributions were evaluated by live-cell, single-molecule TIRF imaging of the 171 basal plasma membrane (Fig. 3A, left), leading to whole-cluster labeling in the JF549 channel, and al-172 ternatively well-separated single-molecule signals in the JF646 channel. For either channel, untrans-173 fected control cells showed almost no signals in comparison (Fig. 3A, center). 174 To evaluate single channel and cluster mobilities by single particle tracking, movies of each red 175 (JF646) and green (JF549) channel fluorescence were recorded consecutively for each cell. The result-176 ing particle tracks for an exemplary cell are shown as a temporal overlay (Fig. 3A, right). Notably, 177 tracks of both labeling modes were restricted to small domains, but JF646-labeled single channel 178 tracks occupied larger areas compared to JF549-labeled cluster tracks. In the latter case, signals origi-179 nated from multiple labeled channels within subdiffraction-sized domains; therefore, an averaged, 180 central cluster position was detected and tracked. As indicated by nearly point-like track overlays, all 181 recorded cluster positions were strongly confined or immobile. 182 The mobility of individual channels and clusters was then quantified by track-based diffusion analysis 183 (Fig. 3B). For each track, we evaluated the mean jump distance (MJD) over the 30 ms frame interval 184 and applied mean squared displacement (MSD) analysis to retrieve the diffusion coefficient. MJD of 185 84 ± 43 nm were measured for single channels, showing a broad distribution of values ranging from 186 20 to 210 nm (98% of data). In contrast, cluster positions showed MJD of only 50 ± 19 nm, which dis-187 tributed rather symmetrically ranging from 10 to 100 nm (98% of data). This indicates that individual 188 channels diffused more rapidly and showed more heterogeneous movement compared to cluster po-189 sitions. Notably, mathematical modeling of immobile positions considering localization uncertainty 190 (see Methods section) resulted in MJD of 44 ± 12 nm, which indicates by comparison that cluster po-191 sitions showed little to no mobility. Next, we examined channel and cluster diffusion by MSD analysis. 192 The fit of individual MSD curves along the first five lag times generated short-term diffusion coeffi-193 cients (D), which are a more robust measure of diffusivity that accounts for localization error (Micha-194 let, 2010). For both single channels and cluster positions, the frequency distributions of D were ap-195 proximately lognormal (Fig. 3B, right). Median D values were more than threefold higher for single 196 channels (D = 0.0054 µm² /s) as compared to cluster positions (D = 0.0015 µm² /s), confirming a signifi-197 cant mobility of single channels. For reference, a threshold value of 0.001 µm² /s (Di Biase et al., 2011; 198 Folci et al., 2018; Hansen et al., 2018) is commonly used to define immobile spots, which classified 199 46% of clusters but only 20% of single channels as immobile based on short-term diffusion. Inde-200 pendently, confocal time lapse data confirmed immobility of cluster positions at lower temporal but 201 higher spatial resolution for time scales of up to 10 minutes (Fig. S4). 202 Nanodomain traversal of CaV1.3 channels corresponds to dynamic channel clustering 203 Importantly, we validated our assumption of single-molecule labeling: We almost exclusively ob-204 served single-step bleaching in intensity traces of long tracks (Fig. S5A) and found the distribution of 205 mean track intensity to be monomodal (Fig. S5B, top-left) reflecting that multi-labeled channels were 206 only rarely measured and thus did not interfere with the interpretation of diffusion coefficients. We 207 also assured that both imaging modes obtained similar spot brightness and track lengths to ensure an 208 unbiased comparison (Fig. S5B top versus bottom). 209 When looking into the shape of long single channel tracks (Fig. 3C), we found two predominant mo-210 tion types: First, we mostly identified mobile channels, which appeared highly confined to one or 211 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint multiple membrane domains and showed variable diffusion speed. Second, with lesser abundance, 212 we identified immobile channels, showing much smaller and consistent displacements around a de-213 fined position. For the first motion type, multi-domain diffusion was observed for particularly long 214 tracks, with a clearly higher channel mobility across inter-domain spaces (examples shown in Fig. 3D). 215 The observed switching of motion types across consistent nanodomains seems to reflect the occa-216 sional transit of channels from one cluster to another. Notably, the high diffusivity state generally 217 lasted less than one second before returning to the confined state for longer time periods. 218 To quantify the confinement of single channels in terms of domain size, we noted that power-law fit-219 ting of MSD is unsuitable for rather short track lengths and low diffusivity compared to the localiza-220 tion error (Kepten et al., 2015). Instead, we determined the convex hull area and radius of gyration 221 for each track (Fig. 3E), which are direct geometrical measurements and thus do not rely on curve fit-222 ting (Golan & Sherman, 2017; Luković et al., 2013). We limited our analysis to a time window length 223 of 300 ms (10 frames), which ensured consistency across variable track lengths (Fig. S4B). We thereby 224 measured convex hull areas of 0.015 ± 0.026 µm² and radii of gyration of 63 ± 36 nm for single chan-225 nel tracks, which is consistent with previously determined cluster dimensions obtained by STED and 226 DNA-PAINT (Table 1). In contrast, JF549-labeled cluster positions showed vastly smaller convex hull 227 areas of 0.004 ± 0.003 µm² and radii of gyration of 39 ± 16 nm, which primarily reflect the localization 228 error around immobile positions (see Methods section). 229 CaV1.3 clusters robustly assemble as Ca2+ release units with RyR2 and Junctophilin-2 230 Next, we examined the organization of calcium release units (CRU) in hiPSC-aCM. CRUs are special-231 ized membrane sites characteristic for primary cardiomyocytes, where sarcoplasmic reticulum-con-232 tained Ryanodine Receptor type 2 (RyR2) and sarcolemmal LTCC are juxtaposed as functional units, 233 which are scaffolded by Junctophilin-2 (JPH2; Brandenburg et al., 2019; Weninger et al., 2022; Z. F. 234 Yang et al., 2022) and mediate Ca2+-induced Ca2+ release from the sarcoplasmic reticulum. While 235 CRUs in adult atrial cardiomyocytes of highly developed species are found both at the cell surface and 236 in intracellular tubular membrane networks (Brandenburg et al., 2018), immature cardiomyocytes in-237 cluding hiPSC-aCM do not feature these membrane networks and thus inherently rely on cell-surface 238 localized domains for calcium release. This suits well to quantitative imaging of the coverslip attached 239 membrane in a consistent, reproducible focal plane (compare Fig. 1). 240 Accordingly, three-channel confocal immunofluorescence of hiPSC-aCM labeled for Halo-CaV1.3, JPH2 241 and RyR2 showed spot-like signals for each target protein at the cell surface (Fig. 4A). All three pro-242 teins colocalized to a large degree, which resulted in white spot coloring in the overlay image. Colo-243 calization was confirmed in the basal imaging plane (Fig. 4B), which rendered the lateral distribution 244 of each protein across the cell surface and revealed a high number of three-channel-colocalized 245 spots. The pattern of spatial correlation was exemplified by intensity line profiles (Fig. 4C), showing 246 that indeed most signal peaks constituted all three proteins. To quantify the recruitment of CaV1.3 to 247 calcium release units (CRUs), three-channel image segmentation (Fig. 4D) and colocalization analysis 248 (Fig. 4E) were applied to a dataset of similarly recorded cell-surface images: We first analyzed the 249 fraction of colocalized CaV1.3 signal and thus found that 65% of CaV1.3 signal mass was RyR2-colocal-250 ized, 61% was JPH2-colocalized, and 51% was double-colocalized, indicating an efficient channel re-251 cruitment to CRUs. The results were reproduced by inversion of secondary antibodies and imaging 252 channels for RyR2 and JPH2 in the same experiment (Fig. 4E). Lastly, we analyzed CaV1.3-colocaliza-253 tion from the perspective of CRU composition, defining CRU signal as the product of JPH2 and RyR2 254 signal. Thereby we found that 86% ± 3% of CRU signal mass colocalized with CaV1.3, indicating that 255 atrial CRUs very consistently harbor CaV1.3 channels. 256 We also determined the colocalization of Halo- and GFP-tagged CaV1.3 with several other subcellular 257 compartments and proteins (Fig. S6). Live-cell co-staining with fluorescently labeled Cholesterol as a 258 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint nanodomain marker showed an exclusion-like pattern rather than colocalization in the basal plasma 259 membrane, highlighting that CaV1.3 clusters are independent of cholesterol-containing lipid domains 260 (Fig. S6A). Immunodetection of endogenous Caveolin-3 (Cav3), a marker of caveolar nanodomains, 261 showed only mild colocalization with CaV1.3 (Fig. S6B). Labeling of the endoplasmic reticulum showed 262 no colocalization (Fig. S6C), indicating that GFP-CaV1.3 was efficiently expressed and transported to 263 the plasma membrane. No colocalization between Halo-CaV1.3 and either sarcomeric alpha-actinin or 264 Junctophilin-1 was observed (Fig. S6D,E). To test whether the association of CaV1.3 and JPH2 requires 265 other cardiac proteins, both proteins were expressed in HEK293 cells and found to strongly colocalize 266 in the basal plasma membrane, pointing towards a tissue-independent, intrinsic association of both 267 proteins. 268 The C-terminal cytosolic tail of CaV1.3 is sufficient for cluster formation 269 Alternative splicing of human CaV1.3 primarily truncates the C-terminal protein sequence, leading to 270 a shortening of the cytosolic C-terminal tail (CTT) from 694 to 180 amino acids in the short (42A) iso-271 form. Since the CTT contains several important protein interaction sites (e.g. for CaM, JPH2, AKAP) 272 and its splicing was shown to modulate channel clustering (Moreno et al., 2016; Stanika et al., 2016), 273 we hypothesized that the CTT may be involved in, necessary, or even sufficient to confer clustering of 274 CaV1.3 channels. We conceived an accessible experimental approach to address this by generating 275 novel fusion proteins concatenating extracellular N-terminal HaloTag (Svendsen et al., 2008) with in-276 tracellular CTT sequence of either CaV1.3 isoform. We then expressed these constructs in hiPSC-aCM 277 and applied surface-selective labeling using HTL-Alexa488 for live-cell confocal imaging to observe the 278 extent of cluster formation. 279 For both constructs and a control construct lacking CTT sequence, signals were predominantly found 280 at the cell surface, as confirmed by co-staining of the plasma membrane (Fig. 5). Interestingly, basal 281 plane imaging showed a segregation of signals into cluster-like shapes for both CTT constructs (Fig. 282 5A, magenta), but an increased abundance of clustered signals for the long versus short CTT isoform 283 (Fig. 5B). As expected, the control construct lacking CaV1.3 sequence showed a highly homogenous 284 membrane signal (Fig. 5C), indicating that there is no intrinsic ability of the extracellular HaloTag to 285 form clusters. The clustering of these constructs was quantified by image analysis, which confirmed a 286 significantly higher cluster abundance for the long versus short CTT construct (mean 17.4% versus 287 7.9%, p < 0.0001, Fig. 5D). Notably, the surface expression level (Fig. 5E) and relative cluster bright-288 ness indicative of cluster size (Fig. 5F) were not significantly different, reflecting that CTT truncation 289 mainly affected cluster abundance but not other parameters. The control construct of transmem-290 brane-anchored HaloTag without CTT notably showed a much higher expression level, but a near-ab-291 sence of clustering. 292 Since the observed clustering of CTT constructs may depend on pre-existing endogenous CaV channel 293 clusters in hiPSC-aCM, we similarly investigated CTT cluster formation in transfected HEK293 CT6232 294 cells (expressing only the accessory channel subunits β3 and α2δ1, but not α1D). In these cells lacking 295 cardiac proteins, we again observed robust cluster formation of both CTT constructs as compared to 296 the control construct (Fig. S7). Similar to hiPSC-aCM, a higher cluster abundance for the long versus 297 short CTT isoform was observed. These results confirmed that CaV1.3 CTT is sufficient to form CaV1.3 298 clusters even in the absence of cardiac-specific accessory proteins and pre-existing clusters. 299 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint

300 Herein nanoscale imaging of human CaV1.3 channel clusters was pioneered in the hiPSC-aCM expres-301 sion system with labeling strategies novel to L-type calcium channels. Consequently, we provide the 302 first model of Cav1.3 cluster assembly in cardiomyocytes, which can be readily compared to existing 303 data in neuronal model systems. The use of hiPSC-aCM provided a physiological cellular framework 304 for the assembly of functional calcium channels, resembling spontaneously contracting cardiomyo-305 cytes in an early developmental stage (Emanuelli et al., 2022; Pourrier & Fedida, 2020). The atrial sub-306 type-directed differentiation implicates an endogenous expression of LTCC in these cells (Chapotte-307 Baldacci et al., 2023; Cyganek et al., 2018). In contrast to primary cardiomyocytes, hiPSC-CM are ame-308 nable to gene transfection (Yuan et al., 2022), which was harnessed in our study to transiently trans-309 fect with tagged Cav1.3 variants. Since we selectively labeled these tagged CaV1.3 variants and not 310 endogenous CaV1.3 in our imaging experiments, simultaneous occurrence of both types of channels 311 within clusters cannot be completely excluded. Offsetting this, we can assume that relative protein 312 cell surface abundance was clearly weighted towards tagged CaV1.3 due to a titration effect: in con-313 trast to endogenous channels, tagged Cav1.3 was overexpressed, while LTCC surface trafficking and 314 residency depends on molecular assembly with β-subunits, limited by the available endogenous pool 315 (Conrad et al., 2021; Stanika et al., 2016). In addition, ER-resident tagged Cav1.3 channels and aggre-316 gates were hardly observed in transfected cells, owing to a lower protein biosynthesis rate compared 317 to heterologous expression systems and the sensitive unfolded protein response for ion channels in 318 hiPSC-CM (Liu et al., 2018). Consequently, potential overexpression artifacts on cluster analysis are 319 considered non-significant. 320 Cluster analysis was performed on the canonical, full-length human CaV1.342 sequence, the most 321 abundant isoform in human cardiomyocytes (Singh et al., 2008). The pore-forming subunit α1D was 322 tagged at the N-terminus preserving channel voltage-gating, while the C-terminus is crucial for regu-323 latory functions that may be perturbed by fusion-tagging. In addition, we considered that alternative 324 tagging of the accessory β-subunit (Conrad et al., 2021; Del Villar et al., 2021; Liu et al., 2020) was un-325 suitable for our study since CaVβ can bind other Ca2+ channel isoforms and performs intracellular 326 functions (Vergnol et al., 2022). We developed independent and synergistic cluster analysis workflows 327 on hiPSC-aCM expressing tagged CaV1.3 proteins. Live-cell quantitative STED and DNA-PAINT imaging 328 determined cluster size and geometry data, while live-cell SPT produced the first mobility data on 329 Cav1.3 channels in cardiomyocytes. Both STED and DNA-PAINT detected cell surface channel clusters 330 of on average ~ 120 nm diameter containing 7–9 channel molecules. Notably live-cell STED excluded 331 potential fixation artifacts previously reported (Huebinger et al., 2018; Sograte-Idrissi et al., 2020) and 332 introduced channel counting based on brightness referencing (Schmied et al., 2012). Brightness cali-333 bration assumed a similar labeling efficiency and linear signal to dye number relation for calibration 334 beads and within cluster nanodomains in situ. The method offers a higher counting range and live-cell 335 compatibility in contrast to widely used photobleaching analysis (Hummert et al., 2021). Importantly, 336 spatial fluorophore densities observed in this study vastly argue against quenching effects known for 337 directly adjacent fluorophores (Schröder et al., 2019). 338 While STED-based size metrics were limited by a spatial resolution of ~ 70 nm, our DNA-PAINT imag-339 ing approach using GFP-targeted CaV1.3 (Fig. 2) attained molecular-scale resolution down to 4 nm lo-340 calization precision, which was not achieved by any previous study of LTCC clustering, e.g. compared 341 to 16 nm in GSD imaging (Moreno et al., 2016). We note that additional, variable displacement errors 342 caused by the undetermined mobility of the cytosolic CaV1.3 N-terminus and the physical size of the 343 nanobody (~ 4nm) used for detection possibly reduced the effective resolution. However, these ef-344 fects were deemed insignificant compared to previous approaches using indirect immunodetection 345 (Sograte-Idrissi et al., 2019). The developed procedures for DNA-PAINT and subsequent data analysis 346 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint thus are sufficient to resolve adjacent channels of 10 nm diameter (Yao et al., 2022) and resulted in 347 similar cluster metrics as live-cell STED imaging, confirming the validity of our approach (Table 1). 348 DNA-PAINT revealed a rather large, non-uniform spacing of clustered channels (median NND = 32 nm) 349 and a low incidence of directly adjacent channels (19% of channels with NND < 20 nm). These obser-350 vations argue against both oligomerization-like isotropic packing of channels and constitutive dimeri-351 zation. Interestingly, we did not observe any grid-like arrangements, which are characteristic for skele-352 tal muscle CaV1.1 and RyR1 clusters (Schredelseker et al., 2005). In this regard, we found that CaV1.3 353 channel arrangement rather corresponds to the stochastic nature of cardiac RyR2 channel clusters 354 (Asghari et al., 2014; Jayasinghe et al., 2018). Interestingly, a previous study in primary hippocampal 355 neurons measured similar CaV1.3 channel counts per cluster by bleach step counting and the same 356 exponential distribution of cluster areas (Moreno et al., 2016), although we measured approximately 357 threefold larger cluster areas on average in cardiomyocytes. Tissue-dependent characteristics were 358 previously not evidenced between neurons and cardiomyocytes, but rather for cochlear inner hair 359 cells, which contain multifold larger CaV1.3 clusters (Neef et al., 2018). Differences in cluster size 360 could rather arise from different segmentation strategies, especially since comparison of the pre-361 sented images indicates similar dimensions of DNA-PAINT and GSD-imaged clusters. 362 Additional data for our Cav1.3 clustering model was gathered by SPT analysis of channel mobility (Fig. 363 3), that included both sparse single-channel labeling and ensemble labeling for tracking whole clus-364 ters. HaloTag-based covalent labeling enabled the use of bright and photostable organic fluoro-365 phores, achieving robust motion tracking by state-of-the-art algorithms (Kuhn et al., 2021). The re-366 sulting trajectories were suitable for track-based diffusion analysis and readily comparable between 367 both imaging modes, owing to matched spot intensities and tracking durations (Fig. S5). A direct com-368 parison revealed that individual channels exhibited much higher diffusivity than independently 369 tracked cluster positions (median D 0.0059 versus 0.0015 µm² /s), thus excluding the possibility of rig-370 idly packed cluster structures. Notably, the absence of multi-step bleaching events in our SPT data in-371 dicates that the majority of tracks correspond to single fluorophores, and thus single channels. Given 372 that we identified confined mobility as the major motion type across all single channel SPT tracks, we 373 interpreted the occupied nanodomains as being equivalent to clusters. These cluster domains were 374 described by an average gyration radius across all single channel tracks of 64 nm, which fits to cluster 375 diameters of around 120 nm reported by STED imaging and DNA-PAINT. Thus, our multi-faceted 376 methodological approach is devoid of major technical compromises and shows consistent independ-377 ent readouts and output parameters, building up a new model of Cav1.3 cluster configurations. Our 378 SPT approach measured channel mobilities that were also in line with previously reported values of 379 neuronal CaV1.2 channels (D = 0.005 µm² /s; Di Biase et al., 2011). Similar to the cited study, we evi-380 denced the traversal of channels across multiple confined domains, which implies that clusters might 381 dynamically recruit and disband individual channels within relatively short dwell times on the order 382 of seconds. The possibility for partial disassembly of clusters stands in contrast to previous mathe-383 matical modeling of LTCC clustering (Sato et al., 2019), however a stochastic assembly process may 384 hold true. 385 Interestingly, CaV1.3 cluster positions in hiPSC-aCM were nearly immobile over long time scales. This 386 is likely due to scaffolding at defined membrane locations, given the highly organized nature of cardi-387 omyocytes including multi-protein CRUs with membrane tethering proteins like Junctophilin-2. In this 388 line we confirmed that CaV1.3, RyR2 and JPH2 consistently associate at the cell surface (Fig. 4), form-389 ing stable calcium release units, equivalent to peripheral dyadic junctions containing CaV1.2 in ven-390 tricular cardiomyocytes (Franzini-Armstrong et al., 1999). A direct interaction site between LTCC and 391 JPH2 was recently postulated (Gross et al., 2021; Z. F. Yang et al., 2022), which is supported by our 392 data showing strong colocalization of CaV1.3 and JPH2 not only in hiPSC-aCM but also upon co-ex-393 pression in HEK293 cells lacking cardiotypical proteins. In parallel these interactions could account for 394 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint the observed fraction of immobile Cav1.3 channels in SPT. However, our Cholesterol and Cav3 stain-395 ings did not reproduce CRU association with lipid rafts (Poulet et al., 2021). Interestingly, previous 396 studies reported the association of potassium channels KV2.1, KCa1.1 and KCa3.1 with neuronal CRUs, 397 which further promoted LTCC clustering and function through direct interactions (Sahu et al., 2019; 398 Vierra et al., 2019; Vivas et al., 2017). This demonstrates that CRUs are dynamic, heterogenous struc-399 tures enabling multifold protein interactions, in line with the emerging clustering model for LTCC. As 400 indicated additionally by significant channel spacing and mobility, clustered CaV1.3 channels are pre-401 sumably intermixed with relevant interactors in the same nanodomain. This organization enables effi-402 cient channel regulation through transient, rather than constitutive interactions. Whether correla-403 tions exist between spatial arrangements and dynamics for distinct CRU constituents remains specu-404 lative but appears to be likely. Analogously these principles have been more extensively researched 405 for CaV2 channels in neuronal systems, however underlying a different functional context (Heine et 406 al., 2020; Perni & Beam, 2021). Notably, the confinement of presynaptic CaV2 channels to cluster do-407 mains was shown to be dependent on alternative C-terminal splicing (Heck et al., 2019; Held et al., 408 2020) and similar organizational mechanisms were found for KV channel clustering (Lewin et al., 2020; 409 Sadegh et al., 2017). 410 For CaV1.3 channels, there are two broadly expressed splice variants: Full-length, canonical CaV1.342 411 and C-terminally truncated CaV1.342A, which have distinct electrophysiological properties and a tissue-412 specific relative abundance, implying a cell-context specific fine-tuning of channel activation 413 (Kuzmenkina et al., 2019). Interestingly, endogenous cytosolic peptides of CaV1.3 distal C-termini 414 (DCT) competitively bind to CaV1.3 channels and downregulate their function (Y . Yang et al., 2022). To 415 assess the relevance of CaV1.3 C-termini for cluster formation, we took advantage of our flexible 416 hiPSC-aCM expression system and introduced novel, artificial constructs of membrane-anchored 417 CaV1.3 C-terminal tail (CTT) for live-cell imaging. Cluster analysis surprisingly revealed that CaV1.3 CTT 418 intrinsically formed cell-surface clusters. Moreover, cluster abundance was clearly reduced for the 419 shorter CTT sequence (180 aa of CaV1.342A) as opposed to the canonical CTT sequence (694 aa of 420 CaV1.342) (Fig. 5). This clearly demonstrates that while the proximal, structured C-terminus is suffi-421 cient for cluster formation, there are additional protein interactions on the extended CTT of CaV1.342 422 that appear to enhance cluster formation or maintenance in the plasma membrane. Given that we 423 observed a similar experimental outcome for ectopic expression in HEK293 cells (Fig. S7), it seems 424 that the required protein interactions underlying this effect are not specific to cardiac myocytes. 425 This leads us to a comparison of putative protein interactions on the proximal versus extended (full-426 length) CTT to relate our overall findings to possible clustering mechanisms. The extended CTT of 427 CaV1.342 contains binding sites for JPH2, AKAP family and PDZ-binding proteins, which are relevant 428 scaffolds for LTCC and may spatially define cluster nanodomains or mediate channel tethering (Choi 429 et al., 2022; Sahu et al., 2019; Stanika et al., 2016; Yang et al., 2023). While these interactions may 430 enhance clustering, none of them appear to be strictly required for clustering, given that we showed 431 cluster formation for the short CTT sequence of CaV1.342A. This proximal, structured C-terminus only 432 contains the EF-Hand and IQ domains, which are important binding sites for modulation by CaM and 433 CaBP (Scharinger et al., 2015). CaM binding was previously reported to mediate C-terminal channel 434 coupling through dimer-like bridging (Moreno et al., 2016) and confirmed to be transient in nature 435 (Kuzmenkina et al., 2019). Whether this mechanism is sufficient for multifold channel associations 436 beyond dimerization was so far unknown but is supported by our data showing clustering of short 437 CaV1.3 CTT constructs. However, we observed a strongly (~ 50%) reduced clustering of the short ver-438 sus long CTT construct. This indicates that a mechanism involving CaM may be sufficient to form clus-439 ters, but not stabilize them, leading to disassembly or internalization. Interestingly, in tsA-201 cells 440 and primary neurons (Moreno et al., 2016; Stanika et al., 2016), the equivalent channel isoform 441 CaV1.342A showed only a mild (~ 20%) reduction of cluster areas compared to full-length CaV1.342. The 442 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint small effect size compared to our study of corresponding CTT constructs indicates that additional, 443 CTT-independent mechanisms may increase the stability of CaV1.3 channel clusters. Further investiga-444 tion of clustering mechanisms and their functional impact using super-resolution and live-cell imaging 445 studies presents a promising future direction for the research of LTCC regulation and pathophysiology. 446 In conclusion, we established a novel approach for researching LTCC clustering based on N-terminal 447 channel tagging in hiPSC-aCM tailored for innovative multiscale super-resolution microscopy. Based 448 on complementary results including DNA-PAINT, live-cell STED and SPT data, we propose that CaV1.3 449 channel clusters consist of relatively mobile individual channels with large interspacing in defined 450 membrane domains, which facilitate transient channel interactions with regulatory and scaffolding 451 proteins to effectively regulate calcium signaling. 452 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint

453 Plasmids 454 CaV1.3 human cDNA (accession number NM_001128840.2, UniProt Q01668-1) was de-novo synthe-455 sized including an N-terminal ‘GGS’ linker. This cDNA was assembled into a vector encoding N-termi-456 nal fusion to the HaloTag (pHTN_HaloTag_CMV-neo, Promega G7721) using restriction cloning, yield-457 ing the Halo-CaV1.3 plasmid. To alternatively generate an N-terminal mEGFP fusion, HaloTag was ex-458 changed by restriction cloning to yield the GFP-CaV1.3 plasmid. 459 A sequence encoding N-terminal HaloTag fused to the transmembrane sequence of non-clustering 460 Integrin β1 (NM_002211.4) was generated as described by Svendsen et al. (2008) and assembled into 461 an EFS-promotor driven vector (pRP-EFS, Vectorbuilder) to generate the control construct ‘Hal-462 oTM_Ctrl’. For CaV1.3 CTT constructs, the insert sequence was fused with cDNA encoding CTT se-463 quence of either human CaV1.342 (NM_001128840.2) or CaV1.342A (XM_047448874.1), starting at 464 amino acid position D1468, generating the long and short CTT plasmids ‘HaloTM_Cav13CT-L’ and ‘Hal-465 oTM_Cav13CT-S’. 466 Cell culture and transfection 467 Human induced pluripotent stem cells (hiPSC) that were derived from healthy human donor cells 468 (isWT1.14, kindly provided by the UMG Stem Cell Unit) were cultured in StemFlex medium (Gibco 469 A3349401) and differentiated to atrial cardiomyocytes according to an established protocol 470 (Kleinsorge & Cyganek, 2020). Cardiomyocytes were purified in glucose-free selection medium after 471 differentiation and then cultured in RPMI 1640 medium (Gibco 72400021) with B-27 supplement 472 (Thermo Fisher 17504044) lacking antibiotics. All differentiations showed spontaneous contractility. 473 For imaging experiments, cells were seeded on Matrigel-coated glass-bottom imaging dishes (ibidi 474 81158) at subconfluent density. Two days after seeding, growth medium containing 2 µM CHIR99021 475 (Merck 361559) and 10% fetal bovine serum (Gibco 16140071) were added to enhance transfection 476 efficiency (Yuan et al., 2022). Cells were transfected the following day using 0.6–1 µg of plasmid and 477 Viafect reagent (Promega E4981, 6 µl/µg plasmid). The medium was exchanged the following day to 478 regular culture medium. Cells were imaged at 4–7 days after transfection. 479 HEK293 cells with constitutive expression of CaV subunits β3 and α2δ1 and inducible expression of α1D 480 (Charles River Laboratories CT6232) were cultured in DMEM/F12 medium containing selection antibi-481 otics and 0.6 µM isradipine (Sigma I6658). For imaging experiments, cells were seeded on fibronectin-482 coated glass-bottom imaging dishes (ibidi 81158) in a growth medium lacking selection antibiotics. 483 Transfection was performed using Lipofectamine 3000 reagent (Invitrogen L3000008) according to 484 the manufacturer’s instructions with 0.6–1 µg plasmid DNA per imaging dish. A washing step was per-485 formed 3 h after transfection using a fresh culture medium. Microscopy experiments were carried out 486 1–2 days after transfection. For electrophysiology, the same protocol was applied to 6-well plates 487 with 2 µg plasmid being added per well. The induction of α1D subunit expression with Tetracycline was 488 generally not performed, unless indicated. 489 Cell labeling for microscopy 490 For Halo-CaV1.3 imaging by confocal and STED microscopy, a labeling solution containing 100 nM 491 JF646-HTL (Promega GA1121) in phenol red-free culture medium was freshly prepared. Live-cell la-492 beling was performed by incubation of hiPSC-aCM in labeling solution for 30 min at 37°C, optionally 493 followed by co-labeling with Cholesterol-Star488/ -StarOrange (Abberior 0206, 40 nM) for 10 min or 494 ER-Tracker Red (Invitrogen E34250, 1 µM) for 30 min in cell culture medium at 37°C. After labeling, a 495 wash-out step was performed by incubation with fresh culture medium for 2 h at 37°C. Afterwards, 496 cells were washed thrice and imaged in live-cell imaging solution (Thermo Fisher A14291DJ). 497 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Dual channel labeling for SPT experiments was performed in phenol red-free culture medium: First, a 498 solution of 250 pM JF646-HTL was applied for 10 min at 37°C. Then, a solution containing 50 nM 499 JF549-HTL (Promega GA1110) was applied for 15 min at 37°C. This was followed by two washing steps 500 and incubation in culture medium for 2 h at 37°C to achieve effective wash-out of unbound ligands. 501 The cells were then washed four times with live-cell imaging solution for 5 min each and subse-502 quently imaged. 503 For confocal imaging of CTT constructs, hiPSC-aCM or HEK293 cells were incubated with a labeling 504 solution containing both cell-impermeant HTL-Alexa488 (Promega G1001, 1 µM) and Cellmask Deep 505 Red Plasma Membrane stain (Invitrogen C10046, 5 µg/mL) in phenol-red free culture medium for 10 506 min at 37°C. Then, cells were washed twice with medium and twice with live-cell imaging solution be-507 fore imaging. Cellmask stain was optionally exchanged for Cholesterol-PEG-KK114 (Brandenburg et 508 al., 2018), yielding similar membrane staining. Notably, alternative HaloTag labeling using Alexa660-509 HTL (Promega G8471) at 3.5 µM concentration did not generate a sufficient labeling outcome. 510 For immunofluorescence of cells expressing Halo-CaV1.3, live-cell incubation with JF646-HTL was per-511 formed as described above. Afterward, cells were washed twice with phosphate buffered saline (PBS) 512 and fixed for 10 min by 4% paraformaldehyde (PFA) diluted in Dulbecco’s PBS containing Ca2+ and 513 Mg2+ (DPBS; Gibco 14040083), then blocked and permeabilized for 1 h in blocking buffer (10% bovine 514 calf serum and 0.1% Triton X-100 in DPBS). Cells were then incubated with primary antibody diluted 515 in blocking buffer overnight at 4°C. This was followed by a secondary antibody incubation in a block-516 ing buffer for 90 min at room temperature. Cells were imaged in DPBS, SlowFade Diamond (Invitrogen 517 S36967) or ProLong Gold (Invitrogen P36930). 518 DNA-PAINT labeling of GFP-CaV1.3 transfected cells was carried out according to the immunofluores-519 cence protocol, but cells were fixed in 4% PFA for 20 min at RT, both before and after the labeling pro-520 cedure to achieve post-fixation. The blocking buffer was supplemented with 0.1 mg/mL sheared 521 salmon sperm DNA (Thermo Fisher 15632011) and 0.05% w/v dextran sulfate (Merck D4911) and Im-522 age-iT FX reagent (Invitrogen I36933) was applied for 10 min after blocking to reduce nonspecific im-523 ager binding (Koester et al., 2022; Youn et al., 2023). GFP nanobody conjugated to R4-docker DNA 524 (sequence 5’ → 3’: ACACACACACACACACACA, Metabion) was applied in dilution buffer (3% BCS, 0.1% 525 Triton X-100, 0.05 mg/mL sheared salmon sperm DNA in DPBS) for 1 h at RT. 526 The following antibodies were used in this study: 527 Species Target Cat number Dilution Rabbit Mouse RyR2 RyR2 Sigma HPA 020028 Thermo Fisher mA3-916 1:500 1:500 Mouse Rabbit JPH2 JPH2 SantaCruz sc377086 Invitrogen 40-5300 1:500 1:100 Nanobody GFP NanoTag N0304 1:500 Mouse Cav3 BD Biosciences 610421 1:500 Mouse ACTN2 Sigma A7811 1:500 Rabbit JPH1 Thermo Fisher 40-5100 1:200 528 Microscopy setup and image acquisition 529 Confocal and STED imaging were performed using an Abberior Expert Line inverted microscope 530 equipped with an oil immersion objective lens (Olympus UPlanSApo 100x NA 1.4), pulsed excitation 531 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint lasers at wavelengths 640/591/485 nm, pulsed STED laser at wavelength 775 nm, Abberior QUAD 532 scanner and avalanche photodiode detectors (Excelitas Technologies SPCM-AQRH). Acquisition set-533 tings for quantitative JF646 STED imaging were as follows: 30% excitation laser power at 640 nm, 12% 534 STED laser power, pixel size 25 nm, pixel dwell time 64 µs, time gating window 0.5–6 ns. Confocal im-535 ages were generated at variable excitation powers depending on the experiment and a pixel size of 80 536 nm. Image channels were recorded separately in line steps to avoid fluorescence crosstalk. All imag-537 ing was performed at RT. 538 DNA PAINT and SPT measurements were performed on a custom-built TIRF optical setup, as de-539 scribed elsewhere (Sograte-Idrissi et al., 2019). The optical configuration is shown in Figure S8. Briefly, 540 488 nm (Omicron PhoxX+ 488-100), 561 nm (Changchun MGL-FN-561-100), and 638 nm (Omicron 541 PhoxX+ 638-150) lasers were used for sample excitation. A neutral density filter (Thorlabs NE10A-A) 542 in tandem with the variable neutral density filter ND (Thorlabs NDC-50C-4-A) were used to adjust the 543 laser excitation power. The laser beam was coupled into a single-mode optical fiber (SMF, Thorlabs 544 P1-460B-FC-2) with a typical coupling efficiency of 40%. After exiting the optical fiber, the collimated 545 laser beam was expanded by a factor of 3.6X using telescope lenses (TL1 and TL2). The typical excita-546 tion intensity at the sample was ~ 1 kW/cm2 for high-photon flux DNA-PAINT imaging. 547 The laser beam was focused onto the back focal plane of the TIRF objective (Olympus UAPON 100X 548 oil, 1.49 NA) using achromatic lens L1 (Thorlabs AC508-180-AB). Mechanical shifting of the beam with 549 respect to the optical axis was done through a translation stage (TS, Thorlabs LNR25/M) to allow for a 550 change between different illumination schemes: EPI, HILO, and TIRF. The smooth lateral positioning of 551 a sample was achieved by using a high-performance two-axis linear stage (Newport M-406). In addi-552 tion, an independent one-dimensional translation stage (Thorlabs LNR25/M) together with a differen-553 tial micrometer screw (Thorlabs DRV3) was used to shift the objective along the optical axis for focus-554 ing on different sample planes. The spectral separation of the collected fluorescence light from the 555 reflected excitation light was achieved using a multi-band dichroic mirror (DM, Semrock Di03 556 R405/488/532/635), which directed the emitted fluorescence light towards the tube lens L2 (Thorlabs 557 AC254-200-A-ML). The field of view was physically limited in the emission path by an adjustable slit 558 aperture (OWIS SP60) positioned in the image plane. Lenses L3 (Thorlabs AC254-100-A) and L4 559 (Thorlabs AC508-150-A-ML) were used to re-image the emitted fluorescence light form the slit onto 560 an emCCD camera (Andor iXon Ultra 897). A band-pass filter (BPF, BrightLine HC 692/40) was used to 561 further block the scattered excitation light. The total magnification of the optical system on the 562 emCCD camera was 166.6X, resulting in an effective pixel size in the sample space of 103.5 nm. 563 DNA-PAINT acquisition was performed in a TIRF mode, with the exposure time of 30 ms and EM gain 564 of 500. First, cells with GFP expression level were selected and then DNA PAINT movies of 30–50k 565 frames were acquired. The imager concentration was in the range of 0.5–1 nM. Single particle track-566 ing was performed in a TIRF mode with the exposure time of 30 ms and EM gain of 500 (JF646) or 100 567 (JF549). Typically, the acquisition of a single movie took 2–5 minutes. All experiments were done at 568 22°C temperature, which was crucial for the mechanical stability of the optical setup. 569 Confocal and STED image analysis 570 Image analysis was performed in ImageJ Fiji (Schindelin et al., 2012) version 1.54f. All analyses were 571 applied selectively to annotated cell areas. For STED image segmentation, the FFT bandpass filter 572 (2.5–20 px) was applied to remove high-frequency noise and unstructured background signal. Candi-573 date signal spots were identified in the filtered image by maxima detection and peak expansion to 574 half-maximal intensity (FWHM) using the ImageJ plugin Interactive H-Watershed. Resulting candidate 575 regions of interest (ROI) were discarded if containing less than 5 pixels or a mean intensity less than 576 50% above the background signal. All remaining ROIs representing specific signals were used for area 577 and brightness measurements. Cluster diameters were calculated from segmented areas using 𝑑 =578 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint 2√𝐴/𝜋 assuming circular shape. Signal brightness was measured in raw image data and corrected for 579 local background by subtraction of the mean brightness of a ring-like ROI obtained by differential en-580 largement of each cluster ROI by 4 versus 2 pixels. 581 For molecular counting in STED images (Fig. S2), DNA Origami reference structures containing 23 ± 3 582 and 7 ± 1 JF646 dye binding sites were purchased from GATTAquant and immobilized on the surface 583 of ibidi glass-bottom imaging dishes coated with BSA-biotin. A calibration measurement was per-584 formed by applying the same quantitative nanoscopy workflow as for Halo-CaV1.3 cluster samples. 585 The resulting distribution of single particle brightness was used to determine a single dye brightness 586 value for molecular counting of channel molecules in cluster ROIs. 587 For confocal-based colocalization analysis, three-channel images were binarized using FFT bandpass 588 filtering (4–40 px) followed by automated local thresholding (‘Otsu’, radius 20 px) in each channel. 589 Colocalization was determined in raw image data as the fraction of above-threshold CaV1.3 signal 590 mass in binarized JPH2-, or RyR2-positive area, or JPH2-RyR2 double-positive area (as defined by 591 Manders M1 colocalization coefficient). To analyze CRU composition, CRU signals were defined as the 592 pixel-wise product of JPH2 and RyR2 signals. Then, the fraction of above-threshold CRU signal mass 593 colocalized with binarized CaV1.3 signals was calculated (corresponding to Manders M2). 594 For confocal-based cluster analysis of CTT constructs, images were smoothed by a Gaussian filter (σ = 595 1 px) and then binarized using automated thresholding (‘Moments’). Resulting spots were quantified 596 by ‘Analyze Particles’ with particle size 6–120 px². Larger particles were not analyzed as they were 597 atypical for clusters and may originate from endosomes. 598 The ImageJ macro code used in this section is provided as Supplementary Software. 599 DNA-PAINT image reconstruction and analysis 600 Raw DNA-PAINT image sequences were processed in ImageJ using the ThunderSTORM plugin (Ovesny 601 et al., 2014). The following parameters were used for emitter localization: 602 Lowered Gaussian filter with sigma = 1.6 px, peak intensity threshold = 1.5*std(Wave.F1), 603 weighted least squares fit, PSF = integrated Gaussian, initial sigma = 1.6 px, fitting radius = 3 px 604 For drift and vibration correction, the recently published DME algorithm (Cnossen et al., 2021, version 605 1.2.1) was implemented using MATLAB R2022A. We increased the robustness of long-term drift track-606 ing by applying DME iteratively for decreasing time bins, which performed better compared to redun-607 dant cross correlation (RCC), which was suggested in the original implementation. Our optimization 608 efforts resulted in the following parameters: 609 iterative: usercc = false, coarse_frames_per_bin = 5000/ 500/ 10, gradientstep = 5e-5/ 5e-5/ 610 5e-6, crlb = 4σ/ 2σ /σ 611 final step: framesperbin = 1, gradientstep = 1e-6, crlb = σ 612 The obtained improvements in localization precision were evaluated by NeNa, which was applied to 613 the central image region (code adapted from Martens et al., 2021). To further increase localization 614 precision, detections were merged in ThunderSTORM with parameters max. dist. = 30 nm, max. 615 frames = 4, max. off frames = 0 and then filtered by the criteria detections > 2 & uncertainty < 10 nm. 616 Subsequently, a density filter with radius = 12 nm, n_min = 4 was applied to discard spurious signals. 617 The reconstruction was then generated by Gaussian rendering with sigma = 5 nm, magnification = 26, 618 resulting in 4 nm image pixel size. 619 For molecular mapping and cluster analysis, a custom-written analysis was applied using MATLAB 620 R2022A. The analysis was limited to manually annotated cell footprints. Briefly, reconstructions were 621 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint smoothed by the H-maxima transform (imhmax) followed by local maxima detection (imregion-622 almax), thus detecting putative channel positions by combining localizations and binding events origi-623 nating from the same binding site. These positions were then subjected to DBSCAN clustering with 624 minPts = 3 and ε = 100 nm (see Fig. S3C). The points comprising each cluster were then outlined using 625 the boundary function. The resulting polygons were expanded by d = 10 nm using the polybuffer func-626 tion to account for physical channel dimensions and localization error. Nearest-neighbor distances 627 across all points or only clustered points were determined using the knnsearch function. 628 The MATLAB code used in this section is provided as Supplementary Software. 629 Single particle tracking and motion analysis 630 Localization and tracking of single-molecule image series was performed using the TrackIt package 631 (Kuhn et al., 2021) running in MATLAB R2022A. The following parameters were used: 632 JF646: u-track random motion, threshold factor = 1.25, tracking radius = 4 px, min track length 633 = 10, gap frames = 3 634 JF549: u-track random motion, threshold factor = 2, tracking radius = 2 px, min track length = 635 10, gap frames = 1 636 Mean jump distance (MJD) and diffusion coefficient (D) were determined for each track using the 637 same software. D values were derived using a linear fit of the first 5 MSD values based on the equa-638 tion 𝑀𝑆𝐷(𝜏) = 4 𝐷 𝜏 + 𝑏 where MSD is the mean squared displacement over all localization pairs for 639 any given lag time τ (based on the frame interval), and b is an offset term reflecting the experimental 640 localization error (Michalet, 2010). For confocal time series in Fig. S4 containing relatively long tracks, 641 the logarithmic fit function 𝑀𝑆𝐷(𝜏) = 4 𝐷 𝜏𝛼 + 𝑏. was applied, where α is an exponent indicative of 642 the motion type (Lerner et al., 2020). For visualization, tracks were rendered with temporal time cod-643 ing using the color map turbo. 644 Confinement metrics were calculated for each track containing at least 10 localizations by averaging 645 over a sliding time windows of 300 ms duration. Convex hull areas were calculated using the convhull 646 function. The radius of gyration (RGyr) was calculated as the root mean square of pointwise distances 647 to the center of mass. Notably, an apparent motion of immobile particles is always detected in experi-648 mental SPT data due to localization errors. Mathematical modeling of this effect was used to estimate 649 the influence on our measurements. We determined an approximate localization precision (σ) of 25 650 nm in our SPT images using ThunderSTORM. Subsequently, the following formulae (Golan & Sherman, 651 2017) were used to determine MJD and RGyr values for immobile particles: 652 MJD = √𝜋 𝜎 = 44.3 nm 653 RGyr = √2 𝜎 = 35.4 nm 654 To generate value distributions resembling our experiment as a point of comparison, we performed a 655 Monte-Carlo simulation (n = 10,000 runs) of immobile particles according to the experimental track 656 length distribution (Fig. S5B) and estimated localization precision σ = 25 ± 5 nm, resulting in: 657 MJD = 44 ± 11 nm 658 RGyr = 34 ± 8 nm 659 The MATLAB code used in this section is provided as Supplementary Software. 660 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Patch clamp measurements 661 HEK293 CT6232 cells were seeded on fibronectin -coated 6-well plates and transfected with 2 µg of 662 plasmid per well encoding Halo- or GFP-tagged Ca V1.3 as described earlier. For control conditions, 663 CaV1.3WT (α1D) was induced by adding 1 µg/mL Tetracycline (Sigma T7508) to the culture medium in 664 parallel to transfections. Cells were harvested one day after transfection as follows: After washing with 665 pre-warmed PBS, cells were dissociated by incubation with 0.25% Trypsin -EDTA solution (Thermo 666 Fisher 25200056) for 30 s at 37°C. The cells were then suspended in growth medium and centrifuged 667 at 100 g, 4°C for 5 min. The cell pellet was then resuspended in divalent-free HBSS (Gibco) at 4 °C. 668 Automated patch-clamp experiments were conducted with the SyncroPatch 384 (Nanion Technologies 669 GmbH) device with thin borosilicate glass, single aperture 384-well chips (NPC384T 1 x S-type). Appli-670 cation of negative pressure (150–250 mbar) allowed for whole-cell access. CaV1.3 currents were elicited 671 with a voltage-step protocol at 0.5 Hz with a holding potential of -80 mV followed by a 100 ms test-672 pulse at -40 mV to +70 mV with 5 mV increments. Experiments were carried out at 22–24 °C. Internal 673 solution contained (in mmol/L): EGTA 10, HEPES 10, CsCl 10, NaCl 10, CsF 110, pH 7.2 (with CsOH). Bath 674 solution contained (in mmol/L): HEPES 10, NaCl 140, glucose 5, KCl 4, CaCl 2 2, MgCl2 1, pH 7.4 (with 675 KOH). Currents were recorded with a Tecella amplifier controlled by PatchControl 384 software. Re-676 cordings were analy zed offline with DataControl 384 software (both: Nanion Technologies GmbH) 677 (Seibertz et al., 2022). 678 General data analysis 679 Data was visualized and statistically analyzed using GraphPad Prism version 10 and MATLAB R2022A. 680 Data and materials availability statement 681 The generated biological materials will be provided by the authors upon request. The generated code 682 for image data processing and analysis is provided as Supplementary Software. 683

684 We are grateful for the excellent technical assistance by Timo Schulte, Brigitte Korff and Elli Rehbein-685 Bode. 686 Stephan E. Lehnart is a principal investigator of DZHK (German Centre for Cardiovascular Research). 687 Funding 688 Funded by the Deutsche Forschungsgemeinschaft (DFG) under Germany’s Excellence Strategy−EXC 689 2067−390729940 to SEL, JE and NV. 690 This work was supported by the Deutsche Forschungsgemeinschaft to NV (VO 1568/3-1, VO 1568/4-691 1, IRTG1816 project 12, SFB1002 project A13) and by the DZHK to NV (German Center for Cardiovas-692 cular Research, 81X2300189 and 81X4300102, “DNAfix”). 693 JE and SB are grateful to the European Research Council (ERC) via project “smMIET” (Grant agree-694 ment No. 884488) under the European Union’s Horizon 2020 research and innovation program. 695 Author contributions 696 Conceptualization: NS, TK, JE, SEL Methodology: NS, RT, TK, FS Software: NS Investigation: NS, RT, SB, FS Formal analysis: NS, RT Visualization: NS, SB Writing – original draft: NS Writing – review & editing: NS, TK, SEL Supervision: TK, NV, JE, SEL Funding acquisition: SEL, JE, NV .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint

697 Asghari, P ., Scriven, D. R., Sanatani, S., Gandhi, S. K., Campbell, A. I., & Moore, E. D. (2014). Nonuniform and variable 698 arrangements of ryanodine receptors within mammalian ventricular couplons. Circ Res, 115(2), 252-262. 699 https://doi.org/10.1161/CIRCRESAHA.115.303897 700 Baudel, M. M. A., Flores-Tamez, V. A., Hong, J., Reddy, G. R., Maillard, P ., Burns, A. E., Man, K. N. M., Sasse, K. C., Ward, S. M., 701 Catterall, W. A., Bers, D. M., Hell, J. W., Nieves-Cintron, M., & Navedo, M. F. (2022). Spatiotemporal Control of 702 Vascular CaV1.2 by alpha1C S1928 Phosphorylation. Circ Res, 131(12), 1018-1033. 703 https://doi.org/10.1161/CIRCRESAHA.122.321479 704 Brandenburg, S., Pawlowitz, J., Eikenbusch, B., Peper, J., Kohl, T., Mitronova, G. Y ., Sossalla, S., Hasenfuss, G., Wehrens, X. H., 705 Kohl, P ., Rog-Zielinska, E. A., & Lehnart, S. E. (2019). Junctophilin-2 expression rescues atrial dysfunction through 706 polyadic junctional membrane complex biogenesis. JCI Insight, 4(12). https://doi.org/10.1172/jci.insight.127116 707 Brandenburg, S., Pawlowitz, J., Fakuade, F. E., Kownatzki-Danger, D., Kohl, T., Mitronova, G. Y ., Scardigli, M., Neef, J., Schmidt, 708 C., Wiedmann, F., Pavone, F. S., Sacconi, L., Kutschka, I., Sossalla, S., Moser, T., Voigt, N., & Lehnart, S. E. (2018). Axial 709 Tubule Junctions Activate Atrial Ca2+ Release Across Species. Frontiers in Physiology, 9, 1227. 710 https://doi.org/10.3389/fphys.2018.01227 711 Chapotte-Baldacci, C. A., Pierre, M., Djemai, M., Pouliot, V., & Chahine, M. (2023). Biophysical properties of NaV1.5 channels 712 from atrial-like and ventricular-like cardiomyocytes derived from human induced pluripotent stem cells. Sci Rep, 713 13(1), 20685. https://doi.org/10.1038/s41598-023-47310-6 714 Choi, S., Vivas, O., Baudot, M., & Moreno, C. M. (2022). Aging Alters the Formation and Functionality of Signaling 715 Microdomains Between L-type Calcium Channels and beta2-Adrenergic Receptors in Cardiac Pacemaker Cells. 716 Frontiers in Physiology, 13, 805909. https://doi.org/10.3389/fphys.2022.805909 717 Cnossen, J., Cui, T. J., Joo, C., & Smith, C. (2021). Drift correction in localization microscopy using entropy minimization. 718 Optics Express, 29(18), 27961-27974. https://doi.org/10.1364/OE.426620 719 Conrad, R., Kortzak, D., Guzman, G. A., Miranda-Laferte, E., & Hidalgo, P . (2021). CaVbeta controls the endocytic turnover of 720 CaV1.2 L-type calcium channel. Traffic, 22(6), 180-193. https://doi.org/10.1111/tra.12788 721 Cyganek, L., Tiburcy, M., Sekeres, K., Gerstenberg, K., Bohnenberger, H., Lenz, C., Henze, S., Stauske, M., Salinas, G., 722 Zimmermann, W. H., Hasenfuss, G., & Guan, K. (2018). Deep phenotyping of human induced pluripotent stem cell-723 derived atrial and ventricular cardiomyocytes. JCI Insight, 3(12). https://doi.org/10.1172/jci.insight.99941 724 Del Villar, S. G., Voelker, T. L., Westhoff, M., Reddy, G. R., Spooner, H. C., Navedo, M. F., Dickson, E. J., & Dixon, R. E. (2021). 725 beta-Adrenergic control of sarcolemmal CaV1.2 abundance by small GTPase Rab proteins. Proceedings of the 726 National Academy of Sciences of the United States of America, 118(7). https://doi.org/10.1073/pnas.2017937118 727 Di Biase, V., Tuluc, P ., Campiglio, M., Obermair, G. J., Heine, M., & Flucher, B. E. (2011). Surface traffic of dendritic CaV1.2 728 calcium channels in hippocampal neurons. Journal of Neuroscience, 31(38), 13682-13694. 729 https://doi.org/10.1523/JNEUROSCI.2300-11.2011 730 Dixon, R. E., Moreno, C. M., Yuan, C., Opitz-Araya, X., Binder, M. D., Navedo, M. F., & Santana, L. F. (2015). Graded 731 Ca2+/calmodulin-dependent coupling of voltage-gated CaV1.2 channels. Elife, 4. https://doi.org/10.7554/elife.05608 732 Dixon, R. E., Navedo, M. F., Binder, M. D., & Santana, L. F. (2022). Mechanisms and physiological implications of cooperative 733 gating of clustered ion channels. Physiological Reviews, 102(3), 1159-1210. 734 https://doi.org/10.1152/physrev.00022.2021 735 Emanuelli, G., Zoccarato, A., Reumiller, C. M., Papadopoulos, A., Chong, M., Rebs, S., Betteridge, K., Beretta, M., Streckfuss-736 Bomeke, K., & Shah, A. M. (2022). A roadmap for the characterization of energy metabolism in human 737 cardiomyocytes derived from induced pluripotent stem cells. J Mol Cell Cardiol, 164, 136-147. 738 https://doi.org/10.1016/j.yjmcc.2021.12.001 739 Endesfelder, U., Malkusch, S., Fricke, F., & Heilemann, M. (2014). A simple method to estimate the average localization 740 precision of a single-molecule localization microscopy experiment. Histochemistry and Cell Biology, 141(6), 629-638. 741 https://doi.org/10.1007/s00418-014-1192-3 742 Ester, M., Kriegel, H.-P ., Sander, J., & Xu, X. (1996). A density-based algorithm for discovering clusters in large spatial 743 databases with noise. kdd, 744 Folci, A., Steinberger, A., Lee, B., Stanika, R., Scheruebel, S., Campiglio, M., Ramprecht, C., Pelzmann, B., Hell, J. W., 745 Obermair, G. J., Heine, M., & Di Biase, V. (2018). Molecular mimicking of C-terminal phosphorylation tunes the 746 surface dynamics of CaV1.2 calcium channels in hippocampal neurons. Journal of Biological Chemistry, 293(3), 1040-747 1053. https://doi.org/10.1074/jbc.M117.799585 748 Franzini-Armstrong, C., Protasi, F., & Ramesh, V. (1999). Shape, Size, and Distribution of Ca2+ Release Units and Couplons in 749 Skeletal and Cardiac Muscles. Biophysical Journal, 77(3), 1528-1539. https://doi.org/10.1016/s0006-3495(99)77000-750 1 751 Golan, Y ., & Sherman, E. (2017). Resolving mixed mechanisms of protein subdiffusion at the T cell plasma membrane. Nat 752 Commun, 8, 15851. https://doi.org/10.1038/ncomms15851 753 Grabner, M., Dirksen, R. T., & Beam, K. G. (1998). Tagging with green fluorescent protein reveals a distinct subcellular 754 distribution of L-type and non-L-type Ca2+ channels expressed in dysgenic myotubes. Proceedings of the National 755 Academy of Sciences of the United States of America, 95(4), 1903-1908. https://doi.org/10.1073/pnas.95.4.1903 756 Gross, P ., Johnson, J., Romero, C. M., Eaton, D. M., Poulet, C., Sanchez-Alonso, J., Lucarelli, C., Ross, J., Gibb, A. A., 757 Garbincius, J. F., Lambert, J., Varol, E., Yang, Y ., Wallner, M., Feldsott, E. A., Kubo, H., Berretta, R. M., Yu, D., Rizzo, V., . 758 . . Houser, S. R. (2021). Interaction of the Joining Region in Junctophilin-2 With the L-Type Ca2+ Channel Is Pivotal for 759 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Cardiac Dyad Assembly and Intracellular Ca2+ Dynamics. Circulation Research, 128(1), 92-114. 760 https://doi.org/10.1161/circresaha.119.315715 761 Hansen, A. S., Woringer, M., Grimm, J. B., Lavis, L. D., Tjian, R., & Darzacq, X. (2018). Robust model-based analysis of single-762 particle tracking experiments with Spot-On. Elife, 7. https://doi.org/10.7554/eLife.33125 763 Heck, J., Parutto, P ., Ciuraszkiewicz, A., Bikbaev, A., Freund, R., Mitlohner, J., Andres-Alonso, M., Fejtova, A., Holcman, D., & 764 Heine, M. (2019). Transient Confinement of CaV2.1 Ca2+-Channel Splice Variants Shapes Synaptic Short-Term 765 Plasticity. Neuron, 103(1), 66-79 e12. https://doi.org/10.1016/j.neuron.2019.04.030 766 Heine, M., Heck, J., Ciuraszkiewicz, A., & Bikbaev, A. (2020). Dynamic compartmentalization of calcium channel signalling in 767 neurons. Neuropharmacology, 169, 107556. https://doi.org/10.1016/j.neuropharm.2019.02.038 768 Held, R. G., Liu, C., Ma, K., Ramsey, A. M., Tarr, T. B., De Nola, G., Wang, S. S. H., Wang, J., van den Maagdenberg, A., 769 Schneider, T., Sun, J., Blanpied, T. A., & Kaeser, P . S. (2020). Synapse and Active Zone Assembly in the Absence of 770 Presynaptic Ca2+ Channels and Ca2+ Entry. Neuron, 107(4), 667-683 e669. 771 https://doi.org/10.1016/j.neuron.2020.05.032 772 Huebinger, J., Spindler, J., Holl, K. J., & Koos, B. (2018). Quantification of protein mobility and associated reshuffling of 773 cytoplasm during chemical fixation. Sci Rep, 8(1), 17756. https://doi.org/10.1038/s41598-018-36112-w 774 Hummert, J., Tashev, S. A., & Herten, D. P . (2021). An update on molecular counting in fluorescence microscopy. Int J 775 Biochem Cell Biol, 135, 105978. https://doi.org/10.1016/j.biocel.2021.105978 776 Jayasinghe, I., Clowsley, A. H., Lin, R., Lutz, T., Harrison, C., Green, E., Baddeley, D., Di Michele, L., & Soeller, C. (2018). True 777 Molecular Scale Visualization of Variable Clustering Properties of Ryanodine Receptors. Cell Rep, 22(2), 557-567. 778 https://doi.org/10.1016/j.celrep.2017.12.045 779 Jenkins, M. A., Christel, C. J., Jiao, Y ., Abiria, S., Kim, K. Y ., Usachev, Y . M., Obermair, G. J., Colbran, R. J., & Lee, A. (2010). Ca2+-780 Dependent Facilitation of CaV1.3 Ca2+ Channels by Densin and Ca2+/Calmodulin-Dependent Protein Kinase II. 30(15), 781 5125-5135. https://doi.org/10.1523/jneurosci.4367-09.2010 782 Kepten, E., Weron, A., Sikora, G., Burnecki, K., & Garini, Y . (2015). Guidelines for the fitting of anomalous diffusion mean 783 square displacement graphs from single particle tracking experiments. PloS One, 10(2), e0117722. 784 https://doi.org/10.1371/journal.pone.0117722 785 Kleinsorge, M., & Cyganek, L. (2020). Subtype-Directed Differentiation of Human iPSCs into Atrial and Ventricular 786 Cardiomyocytes. STAR Protoc, 1(1), 100026. https://doi.org/10.1016/j.xpro.2020.100026 787 Koester, A. M., Szczepaniak, M., & Nan, X. (2022). Fast and Multiplexed Super Resolution Imaging of Fixed and 788 Immunostained Cells with DNA-PAINT-ERS. Curr Protoc, 2(11), e618. https://doi.org/10.1002/cpz1.618 789 Kuhn, T., Hettich, J., Davtyan, R., & Gebhardt, J. C. M. (2021). Single molecule tracking and analysis framework including 790 theory-predicted parameter settings. Scientific Reports, 11(1). https://doi.org/10.1038/s41598-021-88802-7 791 Kuzmenkina, E., Novikova, E., Jangsangthong, W., Matthes, J., & Herzig, S. (2019). Single-Channel Resolution of the 792 Interaction between C-Terminal CaV1.3 Isoforms and Calmodulin. Biophysical Journal, 116(5), 836-846. 793 https://doi.org/10.1016/j.bpj.2019.01.025 794 Lerner, J., Gomez-Garcia, P . A., McCarthy, R. L., Liu, Z., Lakadamyali, M., & Zaret, K. S. (2020). Two-Parameter Mobility 795 Assessments Discriminate Diverse Regulatory Factor Behaviors in Chromatin. Molecular Cell, 79(4), 677-688 e676. 796 https://doi.org/10.1016/j.molcel.2020.05.036 797 Lewin, L., Nsasra, E., Golbary, E., Hadad, U., Orr, I., & Yifrach, O. (2020). Molecular and cellular correlates in KV channel 798 clustering: entropy-based regulation of cluster ion channel density. Sci Rep, 10(1), 11304. 799 https://doi.org/10.1038/s41598-020-68003-4 800 Liu, G., Papa, A., Katchman, A. N., Zakharov, S. I., Roybal, D., Hennessey, J. A., Kushner, J., Yang, L., Chen, B. X., Kushnir, A., 801 Dangas, K., Gygi, S. P ., Pitt, G. S., Colecraft, H. M., Ben-Johny, M., Kalocsay, M., & Marx, S. O. (2020). Mechanism of 802 adrenergic CaV1.2 stimulation revealed by proximity proteomics. Nature, 577(7792), 695-700. 803 https://doi.org/10.1038/s41586-020-1947-z 804 Liu, M., Shi, G., Zhou, A., Rupert, C. E., Coulombe, K. L. K., & Dudley, S. C., Jr. (2018). Activation of the unfolded protein 805 response downregulates cardiac ion channels in human induced pluripotent stem cell-derived cardiomyocytes. J Mol 806 Cell Cardiol, 117, 62-71. https://doi.org/10.1016/j.yjmcc.2018.02.011 807 Luković, M., Geisel, T., & Eule, S. (2013). Area and perimeter covered by anomalous diffusion processes. New Journal of 808 Physics, 15(6), 063034. https://doi.org/10.1088/1367-2630/15/6/063034 809 Martens, K. J. A., Turkowyd, B., & Endesfelder, U. (2021). Raw Data to Results: A Hands-On Introduction and Overview of 810 Computational Analysis for Single-Molecule Localization Microscopy. Front Bioinform, 1, 817254. 811 https://doi.org/10.3389/fbinf.2021.817254 812 Michalet, X. (2010). Mean square displacement analysis of single-particle trajectories with localization error: Brownian 813 motion in an isotropic medium. Physical Review. E: Statistical, Nonlinear, and Soft Matter Physics, 82(4 Pt 1), 041914. 814 https://doi.org/10.1103/PhysRevE.82.041914 815 Moreno, C. M., Dixon, R. E., Tajada, S., Yuan, C., Opitz-Araya, X., Binder, M. D., & Santana, L. F. (2016). Ca2+ entry into 816 neurons is facilitated by cooperative gating of clustered CaV1.3 channels. Elife, 5. https://doi.org/10.7554/elife.15744 817 Neef, J., Urban, N. T., Ohn, T.-L., Frank, T., Jean, P ., Hell, S. W., Willig, K. I., & Moser, T. (2018). Quantitative optical 818 nanophysiology of Ca2+ signaling at inner hair cell active zones. Nature Communications, 9(1). 819 https://doi.org/10.1038/s41467-017-02612-y 820 Ovesny, M., Krizek, P ., Borkovec, J., Svindrych, Z., & Hagen, G. M. (2014). ThunderSTORM: a comprehensive ImageJ plug-in 821 for PALM and STORM data analysis and super-resolution imaging. Bioinformatics, 30(16), 2389-2390. 822 https://doi.org/10.1093/bioinformatics/btu202 823 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Papa, A., Zakharov, S. I., Katchman, A. N., Kushner, J. S., Chen, B.-X., Yang, L., Liu, G., Jimenez, A. S., Eisert, R. J., Bradshaw, G. 824 A., Dun, W., Ali, S. R., Rodriques, A., Zhou, K., Topkara, V., Yang, M., Morrow, J. P ., Tsai, E. J., Karlin, A., . . . Marx, S. O. 825 (2022). Rad regulation of CaV1.2 channels controls cardiac fight-or-flight response. Nature Cardiovascular Research, 826 1(11), 1022-1038. https://doi.org/10.1038/s44161-022-00157-y 827 Perni, S., & Beam, K. (2021). Neuronal junctophilins recruit specific CaV and RyR isoforms to ER-PM junctions and 828 functionally alter CaV2.1 and CaV2.2. Elife, 10. https://doi.org/10.7554/eLife.64249 829 Poulet, C., Sanchez-Alonso, J., Swiatlowska, P ., Mouy, F., Lucarelli, C., Alvarez-Laviada, A., Gross, P ., Terracciano, C., Houser, S., 830 & Gorelik, J. (2021). Junctophilin-2 tethers T-tubules and recruits functional L-type calcium channels to lipid rafts in 831 adult cardiomyocytes. Cardiovascular Research, 117(1), 149-161. https://doi.org/10.1093/cvr/cvaa033 832 Pourrier, M., & Fedida, D. (2020). The Emergence of Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes (hiPSC-833 CMs) as a Platform to Model Arrhythmogenic Diseases. Int J Mol Sci, 21(2). https://doi.org/10.3390/ijms21020657 834 Ries, J., Kaplan, C., Platonova, E., Eghlidi, H., & Ewers, H. (2012). A simple, versatile method for GFP-based super-resolution 835 microscopy via nanobodies. Nature Methods, 9(6), 582-584. https://doi.org/10.1038/nmeth.1991 836 Sadegh, S., Higgins, J. L., Mannion, P . C., Tamkun, M. M., & Krapf, D. (2017). Plasma Membrane is Compartmentalized by a 837 Self-Similar Cortical Actin Meshwork. Phys Rev X, 7(1). https://doi.org/10.1103/PhysRevX.7.011031 838 Sahu, G., Wazen, R. M., Colarusso, P ., Chen, S. R. W., Zamponi, G. W., & Turner, R. W. (2019). Junctophilin Proteins Tether a 839 CaV1-RyR2-KCa3.1 Tripartite Complex to Regulate Neuronal Excitability. Cell Rep, 28(9), 2427-2442 e2426. 840 https://doi.org/10.1016/j.celrep.2019.07.075 841 Sato, D., Hernandez-Hernandez, G., Matsumoto, C., Tajada, S., Moreno, C. M., Dixon, R. E., O'Dwyer, S., Navedo, M. F., 842 Trimmer, J. S., Clancy, C. E., Binder, M. D., & Santana, L. F. (2019). A stochastic model of ion channel cluster formation 843 in the plasma membrane. Journal of General Physiology, 151(9), 1116-1134. https://doi.org/10.1085/jgp.201912327 844 Scharinger, A., Eckrich, S., Vandael, D. H., Schonig, K., Koschak, A., Hecker, D., Kaur, G., Lee, A., Sah, A., Bartsch, D., Benedetti, 845 B., Lieb, A., Schick, B., Singewald, N., Sinnegger-Brauns, M. J., Carbone, E., Engel, J., & Striessnig, J. (2015). Cell-type-846 specific tuning of CaV1.3 Ca2+-channels by a C-terminal automodulatory domain. Frontiers in Cellular Neuroscience, 847 9, 309. https://doi.org/10.3389/fncel.2015.00309 848 Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, 849 B., Tinevez, J. Y ., White, D. J., Hartenstein, V., Eliceiri, K., Tomancak, P ., & Cardona, A. (2012). Fiji: an open-source 850 platform for biological-image analysis. Nature Methods, 9(7), 676-682. https://doi.org/10.1038/nmeth.2019 851 Schmied, J. J., Gietl, A., Holzmeister, P ., Forthmann, C., Steinhauer, C., Dammeyer, T., & Tinnefeld, P . (2012). Fluorescence and 852 super-resolution standards based on DNA origami. Nature Methods, 9(12), 1133-1134. 853 https://doi.org/10.1038/nmeth.2254 854 Schnitzbauer, J., Strauss, M. T., Schlichthaerle, T., Schueder, F., & Jungmann, R. (2017). Super-resolution microscopy with 855 DNA-PAINT. Nat Protoc, 12(6), 1198-1228. https://doi.org/10.1038/nprot.2017.024 856 Schredelseker, J., Di Biase, V., Obermair, G. J., Felder, E. T., Flucher, B. E., Franzini-Armstrong, C., & Grabner, M. (2005). The 857 beta1a subunit is essential for the assembly of dihydropyridine-receptor arrays in skeletal muscle. Proceedings of the 858 National Academy of Sciences of the United States of America, 102(47), 17219-17224. 859 https://doi.org/10.1073/pnas.0508710102 860 Schröder, T., Scheible, M. B., Steiner, F., Vogelsang, J., & Tinnefeld, P . (2019). Interchromophoric Interactions Determine the 861 Maximum Brightness Density in DNA Origami Structures. Nano Letters, 19(2), 1275-1281. 862 https://doi.org/10.1021/acs.nanolett.8b04845 863 Seibertz, F., Rapedius, M., Fakuade, F. E., Tomsits, P ., Liutkute, A., Cyganek, L., Becker, N., Majumder, R., Clauss, S., Fertig, N., 864 & Voigt, N. (2022). A modern automated patch-clamp approach for high throughput electrophysiology recordings in 865 native cardiomyocytes. Commun Biol, 5(1), 969. https://doi.org/10.1038/s42003-022-03871-2 866 Siddig, S., Aufmkolk, S., Doose, S., Jobin, M.-L., Werner, C., Sauer, M., & Calebiro, D. (2020). Super-resolution imaging reveals 867 the nanoscale organization of metabotropic glutamate receptors at presynaptic active zones. Science Advances, 868 6(16), eaay7193. https://doi.org/10.1126/sciadv.aay7193 869 Singh, A., Gebhart, M., Fritsch, R., Sinnegger-Brauns, M. J., Poggiani, C., Hoda, J. C., Engel, J., Romanin, C., Striessnig, J., & 870 Koschak, A. (2008). Modulation of voltage- and Ca2+-dependent gating of CaV1.3 L-type calcium channels by 871 alternative splicing of a C-terminal regulatory domain. Journal of Biological Chemistry, 283(30), 20733-20744. 872 https://doi.org/10.1074/jbc.M802254200 873 Sograte-Idrissi, S., Oleksiievets, N., Isbaner, S., Eggert-Martinez, M., Enderlein, J., Tsukanov, R., & Opazo, F. (2019). Nanobody 874 Detection of Standard Fluorescent Proteins Enables Multi-Target DNA-PAINT with High Resolution and Minimal 875 Displacement Errors. Cells, 8(1), 48. https://doi.org/10.3390/cells8010048 876 Sograte-Idrissi, S., Schlichthaerle, T., Duque-Afonso, C. J., Alevra, M., Strauss, S., Moser, T., Jungmann, R., Rizzoli, S. O., & 877 Opazo, F. (2020). Circumvention of common labelling artefacts using secondary nanobodies. Nanoscale, 12(18), 878 10226-10239. https://doi.org/10.1039/d0nr00227e 879 Stanika, R., Campiglio, M., Pinggera, A., Lee, A., Striessnig, J., Flucher, B. E., & Obermair, G. J. (2016). Splice variants of the 880 CaV1.3 L-type calcium channel regulate dendritic spine morphology. Sci Rep, 6, 34528. 881 https://doi.org/10.1038/srep34528 882 Svendsen, S., Zimprich, C., McDougall, M. G., Klaubert, D. H., & Los, G. V. (2008). Spatial separation and bidirectional 883 trafficking of proteins using a multi-functional reporter. BMC Cell Biology, 9, 17. https://doi.org/10.1186/1471-2121-884 9-17 885 Vergnol, A., Traore, M., Pietri-Rouxel, F., & Falcone, S. (2022). New Insights in CaVbeta Subunits: Role in the Regulation of 886 Gene Expression and Cellular Homeostasis. Front Cell Dev Biol, 10, 880441. 887 https://doi.org/10.3389/fcell.2022.880441 888 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Vierra, N. C., Kirmiz, M., van der List, D., Santana, L. F., & Trimmer, J. S. (2019). KV2.1 mediates spatial and functional coupling 889 of L-type calcium channels and ryanodine receptors in mammalian neurons. Elife, 8. 890 https://doi.org/10.7554/eLife.49953 891 Vivas, O., Moreno, C. M., Santana, L. F., & Hille, B. (2017). Proximal clustering between BK and CaV1.3 channels promotes 892 functional coupling and BK channel activation at low voltage. Elife, 6. https://doi.org/10.7554/elife.28029 893 Weninger, G., Pochechueva, T., El Chami, D., Luo, X., Kohl, T., Brandenburg, S., Urlaub, H., Guan, K., Lenz, C., & Lehnart, S. E. 894 (2022). Calpain cleavage of Junctophilin-2 generates a spectrum of calcium-dependent cleavage products and DNA-895 rich NT1-fragment domains in cardiomyocytes. Sci Rep, 12(1), 10387. https://doi.org/10.1038/s41598-022-14320-9 896 Yang, Q., Perfitt, T. L., Quay, J., Hu, L., Lawson-Qureshi, D., & Colbran, R. J. (2023). Clustering of CaV1.3 L-type calcium 897 channels by Shank3. Journal of Neurochemistry. https://doi.org/10.1111/jnc.15880 898 Yang, Y ., Yu, Z., Geng, J., Liu, M., Liu, N., Li, P ., Hong, W., Yue, S., Jiang, H., Ge, H., Qian, F., Xiong, W., Wang, P ., Song, S., Li, X., 899 Fan, Y ., & Liu, X. (2022). Cytosolic peptides encoding CaV1 C-termini downregulate the calcium channel activity-900 neuritogenesis coupling. Communications Biology, 5(1). https://doi.org/10.1038/s42003-022-03438-1 901 Yang, Z. F., Panwar, P ., McFarlane, C. R., Tuinte, W. E., Campiglio, M., & Van Petegem, F. (2022). Structures of the 902 junctophilin/voltage-gated calcium channel interface reveal hot spot for cardiomyopathy mutations. Proceedings of 903 the National Academy of Sciences of the United States of America, 119(10), e2120416119. 904 https://doi.org/10.1073/pnas.2120416119 905 Yao, X., Gao, S., & Yan, N. (2022). Structural basis for pore blockade of human voltage-gated calcium channel CaV1.3 by 906 motion sickness drug cinnarizine. Cell Research, 32(10), 946-948. https://doi.org/10.1038/s41422-022-00663-5 907 Youn, Y ., Lau, G. W., Lee, Y ., Maity, B. K., Gouaux, E., Chung, H. J., & Selvin, P . R. (2023). Quantitative DNA-PAINT imaging of 908 AMPA receptors in live neurons. Cell Rep Methods, 3(2), 100408. https://doi.org/10.1016/j.crmeth.2023.100408 909 Yuan, Q., Maas, R. G. C., Brouwer, E. C. J., Pei, J., Blok, C. S., Popovic, M. A., Paauw, N. J., Bovenschen, N., Hjortnaes, J., 910 Harakalova, M., Doevendans, P . A., Sluijter, J. P . G., van der Velden, J., & Buikema, J. W. (2022). Sarcomere 911 Disassembly and Transfection Efficiency in Proliferating Human iPSC-Derived Cardiomyocytes. J Cardiovasc Dev Dis, 912 9(2). https://doi.org/10.3390/jcdd9020043 913 Zaveri, S., Srivastava, U., Qu, Y . S., Chahine, M., & Boutjdir, M. (2023). Pathophysiology of CaV1.3 L-type calcium channels in 914 the heart. Frontiers in Physiology, 14, 1144069. https://doi.org/10.3389/fphys.2023.1144069 915 Zhang, H., Maximov, A., Fu, Y ., Xu, F., Tang, T. S., Tkatch, T., Surmeier, D. J., & Bezprozvanny, I. (2005). Association of CaV1.3 L-916 type calcium channels with Shank. Journal of Neuroscience, 25(5), 1037-1049. 917 https://doi.org/10.1523/JNEUROSCI.4554-04.2005 918 Zhang, Z., He, Y ., Tuteja, D., Xu, D., Timofeyev, V., Zhang, Q., Glatter, K. A., Xu, Y ., Shin, H. S., Low, R., & Chiamvimonvat, N. 919 (2005). Functional roles of CaV1.3 (alpha1D) calcium channels in atria: insights gained from gene-targeted null 920 mutant mice. Circulation, 112(13), 1936-1944. https://doi.org/10.1161/CIRCULATIONAHA.105.540070 921 922 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure 1: Live-cell STED imaging resolves clustering of Halo-CaV1.3 at the cell surface of hiPSC-aCM. A) Halo-CaV1.3 fusion protein was transiently expressed in hiPSC-aCM and labeled with HTL-JF646. Live-cell confocal imaging of medial cell sections revealed predominantly spot-like signal patterns of Halo-CaV1.3 (magenta) distributed along the plasma membrane (green, Cholesterol-Star488), as highlighted by magnification of the indicated area. B) Quantitative live-cell imaging was performed centering the focal plane on the basal plasma membrane (‚Fire‘ LUT). STED imag- ing (right column) revealed sub-diffraction size and spacing of Halo-CaV1.3 clusters, which could not be resolved by confocal mi- croscopy (left). The indicated image region is magnified in the lower images, showing representative signal distributions. C) STED images were analyzed by automated image segmentation to detect individual signal clusters at a density of 2.0 µm-2. Cluster sizes averaged to areas of 0.013 ± 0.008 µm2, which were equivalent to diameters of 122 ± 35 nm assuming circular shape (n = 10875 clusters, N = 30 cells). D) The signal mass of each cluster was referenced to calibration samples for molecular counting of labeled Halo-CaV1.3 inside these clusters (see Fig. S2). On average clusters contained 9 ± 12 CaV1.3 channels contributing to an intra-cluster channel density of 612 µm-2. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure 2: Super-resolution DNA-PAINT imaging of GFP-CaV1.3 resolves channel arrangement within clusters. A) Illustration of DNA-PAINT labeling of GFP-tagged CaV1.3 expressed in hiPSC-aCM. After fixation, GFP was detected by docker- DNA-coupled anti-GFP nanobodies (NbGFP) and reversibly binding imager-DNA labeled by Atto643 or Atto550. B) Single raw data TIRF imaging frames of imager dye signals were used to localize DNA-PAINT binding events, which were accu- mulated over 30,000–50,000 frame sequences to generate a reconstruction (‚Fire‘ LUT, blurred to enhance cluster visibility at low magnification). Simultaneous recordings of GFP fluorescence confirmed basal membrane imaging planes in transfected cells ex- pressing GFP-CaV1.3. C) Magnification of the indicated image region in (B), showing clusters of DNA-PAINT localization spots (left) and GFP fluores- cence (center). Spatial correlation is evident in overlay images (right), confirming specificity of anti-GFP labeling with nanobodies. D) Further magnification of exemplary clusters highlighted in (C) confirms separate, countable localization spots as a prerequisite for molecular mapping and counting of CaV1.3 channel numbers. E) Drift correction (DME) and spot merging (detection time binning across 4 frames) improved the NeNA-measured localization precision from 12.5 to 4.1 nm (N = 10 cells). Significance was shown by repeated measures ANOVA with indicated pairwise com- parisons (**** = p < 0.0001). F) Molecular mapping of single channel positions was achieved by local maxima detection (indicated as green circles) from DNA- PAINT reconstructions with ≤ 5 nm localization precision (left). White arrows indicate the resolution of adjacent spots at 12 nm distance. Next, DBSCAN clustering was applied to define individual channel clusters, containing 15 channels in this example (right). DBSCAN parameters were set to ε = 100 nm and minPts = 3 (see Figure S3). Molecular mapping was applied to a dataset of n = 18129 clusters in N = 17 cells. G) Molecular maps were used to compute nearest-neighbor distances (NND) across all channel positions and restricted to DBSCAN-defined clusters. The blue-dotted line indicates a plateau of NND values at 100 nm in line with the optimal ε parameter. H) Molecular maps and DBSCAN cluster outlines were used to determine the distribution of cluster area and channel counts (I) in reconstructed DNA-PAINT data. Additional median and interquartile range (IQR) values are shown in Table 1. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure 3: Single channel tracking quantifies the mobility and confinement of clustered CaV1.3 channels. A) Single-molecule TIRF images of a living hiPSC-aCM expressing Halo-CaV1.3 channels (first column) and a corresponding mock- transfected control cell (second column). Cells were concurrently labeled for individual channels (250 pM HTL-JF646, first row) and multichannel clusters (50 nM HTL-JF549, second row). For each recorded cell, an image series of multichannel clusters (JF549) was recorded, followed by a second image series of single channels (HTL-JF646). The image data of both labeling modes was independently processed by single-particle-tracking (SPT). The resulting tracks are shown as an overlay for the exemplary cell (third column). B) The diffusion of individual channels and clusters is compared by histograms depicting the mean jump distance of tracks and the diffusion coefficient originating from a fit of time-dependent mean-squared-displacement (MSD) curves. Insets show the same data on a logarithmic scale with gaussian fit curves in blue and median values indicated above. D value histograms include under- and overflow bins for values outside of the axis range. The dataset includes N = 15 cells with n = 6519 single channel tracks and n = 5873 cluster tracks. C) Exemplary tracks of JF646-labeled single channels with duration > 10 s demonstrate the two predominant motion types: Mo- bile, confined (left) and immobile (right). The corresponding MSD curves are shown on the right side, with a red line indicating the linear fit used to retrieve the diffusion coefficient. The grid interval is 100 nm. D) Exemplary single-channel tracks show occasional domain and motion type switching, characterized by intermittently high mo- bility traversal between multiple domains of lower mobility and high confinement. E) Confinement analysis was performed on sufficiently long tracks containing at least 10 localizations. Two metrics were used to characterize local confinement: The convex hull area surrounding each complete track, and the radius of gyration. Both metrics were computed over a 300 ms (10 frame) sliding time window to avoid the inclusion of multi-domain track segments. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure 4: Colocalization of CaV1.3 with JPH2 and RyR2 in surface-localized calcium release units. A) Live-cell labeling of Halo-CaV1.3 expressed in hiPSC-aCM was combined with subsequent RyR2 and JPH2 immunofluorescence for confocal imaging. Medial confocal sections display CaV1.3 (red), RyR2 (green) and JPH2 (blue) only in the cell periphery, where they show extensive colocalization (white coloring), representing calcium release units (CRU) localized to the plasma membrane. ‘2AB-’ indicates the secondary antibody conjugate and corresponding imaging channel (580 = Abberior Star580, 488 = Abberior StarGreen). B) Imaging of the adherent, basal plasma membrane in the same cell reveals a homogeneous 2D distribution of spot-like signals, representing adjacent, plasma membrane resident CRUs. C) Representative line profiles across cluster signals demonstrate extensive spatial correlation between fluorescent signals corre- sponding to all three analyzed CRU proteins. The Pearson correlation coefficient (r) indicates one-dimensional correlation of each RyR2 and JPH2 to Halo-CaV1.3 signal. D) Three channel images from planar membranes (magnification from B) were segmented for signal-spots and binarized. Conse- quently, white signal color indicates three-channel colocalization, magenta indicates CaV1.3-JPH2, yellow indicates CaV1.3-RyR2 and cyan indicates RyR2-JPH2 colocalization, respectively. E) Colocalization analysis quantified the fraction of CaV1.3 signal mass overlapping with binarized areas of either RyR2, JPH2 or both (left graph). Specific colocalization was confirmed with an inversion of fluorophores on secondary antibodies. The right graph shows the fraction of CRU signal mass (defined by the product of RyR2 and JPH2 signal) that is colocalized with CaV1.3- binarized area (N = 16 cells). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure 5: CaV1.3 C-terminal construct expression in hiPSC-aCM leads to cluster formation. A) A fusion protein of CaV1.342 C-terminal tail (long CTT) attached to transmembrane-HaloTag was expressed in hiPSC-aCM using transient transfection. After live-cell labeling with cell-impermeable HTL-Alexa488, distinct spot-like signals resembling CaV1.3- like clusters were revealed at the cell surface by confocal imaging in the medial (left) and basal (right) imaging plane (displayed in magenta). The localization was confirmed by co-staining with the plasma membrane marker Cellmask-DeepRed (displayed in green). B) Expression of the equivalent fusion protein containing the short CTT of CaV1.3 splice variant 42A also lead to cluster-like signal shapes at the cell surface. However, cluster-like spots appeared less abundant compared to the long isoform shown in (A). C) A control construct containing only transmembrane-HaloTag showed a homogenous protein distribution in the plasma mem- brane without the formation of distinct, cluster-like spots. D) Evaluation of confocal HTL-Alexa488 signal distributions shown in (A-C) by cluster analysis. Cluster signals were detected by automatic thresholding after gaussian filtering of images and then quantified by size and brightness. Box plots show Median, IQR and Tukey-based whiskers. The fraction of clustered signal within basal plasma membranes was significantly larger for long vs short CTT (**** = p < 0.0001, Welch‘s unpaired t-test, N = 28, 29, 25 cells). The control condition showed an absence of signifi- cant clustering and was therefore not statistically compared. E) Membrane expression levels measured by the average fluorescence intensity showed no significant difference between long and short CTT (p = 0.47). F) Similarly, the relative signal intensity of clusters compared to the whole plasma membrane showed no significant difference between long and short CTT (p = 0.7). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Table 1: Comparison of cluster analysis results generated by STED imaging and DNA-PAINT .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S1: Automated patch clamp measurements of wild-type and tagged CaV1.3 channels expressed in HEK293 cells confirm similar electrophysiological characteristics. HEK293 CT6232 cells expressing accessory β3 and α2δ1 subunits were induced to express CaV1.3WT or transfected with plasmid encoding Halo- or GFP-tagged CaV1.3 channels. Whole-cell calcium currents were measured using the Nanion SyncroPatch 384 device. The resulting IV curves of individual cells are shown as gray lines and averaged for each condition (colored lines). The last graph shows an overlay of the averaged IV curves for each condition after normalizing current amplitude minima to -1 a.u. to account for the variation in channel expression levels. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S2: Brightness referencing method for molecular counting of JF646 fluorophores. DNA Origami linked to 7 or 23 JF646 dye molecules were immobilized on coverslips and recorded by STED imaging under equal conditions as for cellular Halo-CaV1.3 cluster imaging (A). By image analysis of spot-like signals, a distribution of integrated pho- ton counts across all detected spots was determined for each sample (B). The histograms were fitted by normal distributions to retrieve mean brightness values, which was used for a linear fit of spot brightness to dye molecule counts in (C). The determined conversion factor (32.5 photon counts per fluorophore) was used for image analysis of Halo-CaV1.3 samples to retrieve labeled channel counts within clusters. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S3: Optimization of DNA-PAINT image reconstruction and DBSCAN clustering. A) Representative reconstruction of a GFP-CaV1.3 channel cluster imaged by DNA-PAINT in TIRF mode. A successive improvement of the localization-based image reconstruction over standard reconstruction (1) was achieved by applying a customized version of drift correction by entropy minimization (DME, 2), followed by either histogram-based localization filtering (3), or followed by temporal merging of subsequent localizations and local density filtering (4). B) The point clustering algorithm DBSCAN was applied to DNA-PAINT molecular map data. Three exemplary values for the param- eter ε give rise to distinct clustering results. C) Graph showing the change in DBSCAN cluster density and points per cluster as a function of ε parameter values at minPts = 3. For ε = 100 nm, the highest cluster density (2.5 µm-2) is observed, while higher ε values lead to merging of pre-existing clusters and increased heterogeneity. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S4: Confocal timelapse imaging demonstrates immobility of Halo-CaV1.3 clusters across time scales. A) Confocal timelapse imaging of hiPSC-aCM expressing Halo-CaV1.3 (row 1+2) or GFP-CaV1.3 (row 3) shows cluster positions in the basal plasma membrane. Images series were recorded in intervals of 400 ms, 1.5 s and 10 s. For each timelapse, the first, second, middle and last frame are shown. B) Representative trajectory of a single cluster position shown on a 30 nm pixel grid, generated by SPT of a timelapse at 1.5 s intervals and 30 nm pixel size. Quantitative analysis of SPT data reveals low jump distances of ~ 35 nm reflecting the localization uncertainty and MSD fit-derived diffusion coefficients of less than 10-4 µm² /s, thus confirming immobility of the tracked cluster positions. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S5: Supporting data for single particle tracking analysis. A) Intensity time traces for the exemplary tracked spots shown in Figure 3C. No bleaching steps were observed in the majority of long tracks. B) Mean spot intensity and track length distributions indicative of tracking performance were calculated. Both metrics show simi- lar distributions for both imaging modes, which excludes a potential bias in the comparative diffusion analysis. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S6: Halo-CaV1.3 colocalization with nanodomain and compartment markers. A) Live-cell confocal images of hiPSC-aCM expressing Halo-CaV1.3, labeled by HTL-JF646 and Cholesterol-StarOrange. CaV1.3 clus- ters and Cholesterol both localized to the plasma membrane (top), but dual-channel STED imaging in the basal membrane focal plane (bottom) showed rather exclusion-like arrangement with Cholesterol-containing nanodomains. B) Immunofluorescence of hiPSC-aCM expressing EGFP-CaV1.3 showed rather low colocalization with Caveolin-3 (Cav3). C) Live-cell imaging of hiPSC-aCM showed a mutually exclusive distribution of EGFP-CaV1.3 and endoplasmic reticulum, labeled by ER-Tracker Red. D) Immunofluorescence of hiPSC-aCM expressing Halo-CaV1.3 showed no colocalization with cardiac α-actinin or Junctophilin-1 (JPH1) (E). F) Immunofluorescence of HEK293 CT6232 cells transfected with Halo-CaV1.3 and JPH2-CFP showed extensive colocalization of clustered spots in the basal membrane focal plane. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S7: CaV1.3 C-terminal construct expression in HEK293 leads to cluster formation independent of the cardiac proteome. A) CaV1.3 C-terminal cytosolic tail (CTT, long isoform) fused to cell-surface HaloTag was expressed in HEK293 CT6232 cells and labeled with cell-impermeable HTL-Alexa488. The cells were co-stained with the plasma membrane marker Cholesterol-PEG- KK114 and imaged by live-cell confocal microscopy. B) Expression of the equivalent fusion protein containing the short C-terminal tail splice variant. C) Expression of a control construct containing only cell-surface HaloTag without CTT sequence, serving as a negative control. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint Figure S8: Illustration of the custom-built optical setup (described in the Methods section) .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 28, 2024. ; https://doi.org/10.1101/2024.02.26.582084doi: bioRxiv preprint