Engineered bacterial voltage-gated sodium channel platform for cardiac gene therapy

All animal studies were performed in accordance with the animal protocol A064-21-03 approved by the Duke University Institutional Animal Care and Use Committee.

Plasmid construction

All lentiviral transfer plasmids were constructed from the pRRL-CMV vector (a gift from Inder Verma, Salk Institute). Human codon optimization of bacterial Na_vSheP D60A^7,12 (bSheP) gene was performed via Genscript OptimumGene algorithm⁸³ (hSheP) and ATUM Gene-GPS™ algorithm⁸⁴ (h2SheP). Wild-type and codon-optimized sequences are listed in Supplementary Information. Human codon-optimized cDNAs synthesized by respective companies were subcloned into the pRRL-CMV vector where they were linked with GFP via the T2A peptide (pRRL-CMV-hSheP-2A-GFP and pRRL-CMV-h2SheP-2A-GFP). The lentiviral plasmid containing wild-type channel sequence co-expressed with GFP (pRRL-CMV-bSheP-2A-GFP) served as the control. For optimization of transcription efficiency in cardiomyocytes, two additional lentiviral transfer plasmids were constructed by replacing the CMV promoter in pRRL-CMV-h2SheP-GFP with MHCK7¹⁷ and cTnT¹⁶ promoters (pRRL-MHCK7-h2SheP-2A-GFP and pRRL-cTnT-h2SheP-2A-GFP). Single-stranded and self-complementary AAV transfer plasmids were constructed from the pAAV-CAG-eYFP (a gift from Viviana Gradinaru, Addgene plasmid #104055) and pscAAV-CAG-GFP (a gift from Mark Kay, Addgene plasmid #83279), respectively. For mouse tail-vein injection studies, h2SheP-2A-GFP and MHCK7-h2SheP-HA fragments were amplified from lentiviral plasmids and subcloned into AAV vectors to generate pAAV-MHCK7-h2SheP-HA-2A-GFP, pAAV-MHCK7-GFP, pAAV-CAG-h2SheP-2A-GFP, and pscAAV-MHCK7-h2SheP-HA. Plasmid pscAAV-CAG-GFP⁸⁵ (a gift from Mark Kay, Addgene plasmid #83279) was used to generate control scAAV9 for optimization of in vivo delivery method.

Flow cytometry

HEK293 (ATCC, CRL-1573) monolayers were rinsed with phosphate-buffered saline (PBS) then dissociated using 0.05% Trypsin-EDTA (Thermo Fisher Scientific) at 37 °C for 3 min. Trypsin was quenched with DMEM high glucose (Thermo Fisher Scientific) containing 10% FBS (Hyclone) and 20 µg/ml DNase I (Millipore 260913). The cell suspension was centrifuged at 500 × g for 5 min, then resuspended in 4% paraformaldehyde (PFA) diluted in PBS. Cells were incubated in 4% paraformaldehyde (PFA) for 10 min at room temperature (RT), centrifuged again, then resuspended in PBS containing fluorescence-activated cell sorting (FACS) buffer (PBS with 0.5% BSA (Sigma), 0.1% Triton-X 100 (Thermo Fisher Scientific), and 0.02% Azide (VWR)). FACS was performed using either BD DiVA or B-C Astrios cell sorter at the Flow Cytometry Shared Resource Core Facility at Duke University. The analysis was performed using FlowJo v10.7.1.

Lentivirus production and titration

High-titer lentiviruses were prepared using second-generation lentiviral packaging system as described previously¹³. Specifically, 293T cells (ATCC, CRL-3216) were co-transfected with lentiviral transfer plasmid, packaging plasmid psPAX2, and envelope plasmid pMD2.G (6:3:1 mass ratios) using JetPRIME transfection reagent (Polyplus). Seventy-two hours after transfection, the supernatant containing lentiviral particles was collected, centrifuged (500 × g, 10 min), and filtered through 0.45 μm cellulose acetate filter (Corning) to remove cell debris before being combined with Lenti-X Concentrator (Clontech) at 3:1 volume ratio and incubated overnight at 4 °C. Concentrated lentiviral particles were harvested following 45 min centrifugation (1500 × g, 4 °C) and resuspended in DPBS. Plasmids psPAX2 and pMD2.G were obtained from Didier Trono (Addgene plasmids #12260 and #12259). To determine the functional titer of lentiviruses expressing fluorescence reporter, 293T cells were transduced with serial dilutions of concentrated lentiviral stock and the percentage of transduced cells was determined via flow cytometry 72 h post-transduction. Functional titer in transduction units per mL (TU/mL) was estimated from dilutions that yielded 5–30% transduction efficiency, by dividing the total number of transduced cells by the volume of virus added in mL. Transduction in HEK293, NRVM, and hiPSC-CM monolayers was performed with the multiplicity of infection (MOI) of 1, 7, and 2, respectively, and functional studies were conducted 3–5 days after transduction.

Neonatal rat ventricular myocyte culture

Ventricles of both male and female 2-day-old Sprague-Dawley rats (Charles River Laboratories, Wilmington MA) were excised, minced, and incubated with 0.1% trypsin (Thermo Fisher Scientific) overnight and dissociated in four sequential steps using 0.1% collagenase³⁶. Dissociated cells were centrifuged for 5 min at 200 × g and further enriched by a 45 min preplating step. Isolated cardiomyocytes were seeded onto Aclar coverslips (21 mm diameter, Electron Microscopy Sciences) coated with 30 µg/ml fibronectin (Sigma) at 8 × 10⁴ cells/cm² in DMEM/F12 medium (Gibco, 11320-033) supplemented with 10% fetal bovine serum (FBS), 0.2% penicillin, and 0.2% B12. The following day (day 1), cells were treated with 10 µg/ml mitomycin-C (Sigma) for 2 h before media change to fresh seeding media. At day 2, media was changed to serum-free maintenance media (DMEM/F12 + 0.2% penicillin + 0.2% B12 + 2.5 µg/ml L-ascorbic acid + 5 nM Triiodo-L-Thyronine + 1X Insulin-Transferrin-Selenium supplement) and lentivirus was added into the cultures. Complete maintenance media change was performed every 2 days and cultures were studied on days 4–6. For optimization of highly arrhythmogenic cultures, four cell seeding numbers were tested—300K, 400K, 600K, and 800K, which correspond to seeding densities of 8 × 10⁴, 1.1 × 10⁴, 1.6 × 10⁵, and 2.2 × 10⁵ cells/cm², respectively. For patch-clamp and sharp electrode recording studies, maintenance media consisted of DMEM/F12, 0.2% penicillin, 0.2% B12, and 5% FBS. For experiments in hypertrophic NRVM monolayers, 100 µM phenylephrine (Sigma–Aldrich) was added for 24 h at Day 3 and Day 8 of culture and monolayers were optically mapped and immunostained at culture Day 9³⁷.

hiPSC-CM differentiation and culture

Human-induced pluripotent stem cells (hiPSCs) were reprogrammed from commercially available BJ fibroblasts (ATCC cell line, CRL-2522) at the Duke University iPSC Core Facility and named DU11 (Duke University clone #11)⁸⁶. The DU11 hiPSC line was authenticated by pluripotency marker expression using IF and FACS, karotyping to confirm genomic integrity, and teratoma formation⁸⁶. DU11 hiPSCs were differentiated into cardiomyocytes (hiPSC-CMs) using small-molecule modulation of the Wnt pathway⁸⁷ and purified via metabolic selection⁸⁸ on day 10 post induction⁸⁶. Specifically, DU11 hiPSCs were plated at 2 × 10⁵/cm² with 5 µM Y-27632 (ROCK inhibitor, Tocris) and induced to differentiate 2 days after seeding. To induce cardiac differentiation (on day 0, d0), cells were treated with 10–14 μM CHIR99021 (SelleckChem) in RPMI-1640 with B27(−) insulin (ThermoFisher Scientific). Exactly 24 h later, CHIR was removed and replaced with basal RPMI/B27(−) medium. On d3, half of the old medium was collected and mixed with fresh RPMI/B27(−) medium containing 5 μM (final concentration) IWP-4 (Tocris). On d5, IWP-4 was replaced with a basal RPMI/B27(−) medium. From d7 onward, cells were fed with RPMI/B27(+)-insulin every 2–3 days, with spontaneous beating generally starting on d7–d10 of differentiation. Differentiating CM cultures were purified via metabolic selection between d10 and d12 by rinsing cultures with PBS, followed by incubation in “no glucose” medium for 48 h (glucose-free RPMI (ThermoFisher Scientific 11879020) supplemented with 4 mM lactate (Sigma L4263), 0.5 mg/mL recombinant human albumin (Sigma A6612), and 213 μg/mL L-ascorbic acid 2-phosphate (Sigma A8960))⁸⁸. Metabolically purified hiPSC-CMs were dissociated into single cells and plated onto 21 mm diameter Aclar coverslips coated with Corning Matrigel hESC-Qualified Matrix (Corning, 354277) at 2 × 10⁵ cells/cm² (for optical mapping) or 10⁴ cells/cm² (for patch clamp) in 3D RB+ medium, which contains RPMI-1640 (Sigma, R8758), 2% B27 supplement (Gibco, 17504044), 2 mg/mL aminocaproic acid (Sigma, A2504), 50 µg/mL ascorbic acid 2-phosphate (Sigma, A8960), 1% penicillin-streptomycin (Thermo Fisher,15140), 1% non-essential amino acids (Thermo Fisher, 11140), 1% sodium pyruvate (Thermo Fisher, 11360), 0.45 µM 1-thioglycerol (Sigma, M6145), and 5µM Y-27632 (Tocris, 1254). One day post-seeding, the medium was replaced with 3D RB+ medium without Y-27632 (maintenance medium) and h2SheP or control lentivirus was added. The medium was exchanged every other day and cells underwent patch-clamp or optical mapping 72–96 h after lentiviral transduction.

Quantitative RT-PCR

Total RNA was extracted using RNeasy Plus Mini Kit according to the manufacturer’s instructions (Qiagen) and the concentration was measured using NanoDrop One (Thermo Scientific). Reverse transcription was run on equal amounts of RNA using iScript cDNA Synthesis Kit (Bio-Rad). Standard quantitative PCR was performed using an iTaq Universal SYBR Green Supermix kit (Bio-Rad). The relative expression of indicated genes was quantified by the ΔCT method⁸⁹. The primers used are listed in Supplementary Table 1.

Whole-cell patch-clamp recordings

Dissociated single cells were plated onto Aclar coverslip and left to attach for 5 h in 37 °C incubators. Coverslip was then transferred to a glass-bottom patch-clamp chamber perfused with bath solution. Patch pipettes were fabricated with tip resistances of 1–2 MΩ when filled with pipette solution. Whole-cell patch-clamp recordings were acquired at room temperature (25 °C) or 37 °C using the Multiclamp 700B amplifier (Axon Instruments), filtered with a 10 kHz Bessel filter, digitized at 40 kHz, and analyzed using WinWCP software (John Dempster, University of Strathclyde). To measure activation properties of voltage-gated sodium channels, membrane voltage was stepped from a holding potential of -80 mV to varying 500 ms test potentials (−50 to 60 mV, increments of 10 mV). Inactivation of voltage-gated sodium channels was derived from peak currents measured at 0 mV after varying 3-s prepulse potentials (−160 to −30 mV, increments of 10 mV). Steady-state I_K1–V curve was generated from the current responses to varying 1 s test potentials (−90 to 50 mV, increments of 10 mV) from a holding potential of −40 mV. Action potentials were triggered by injecting a 1 ms current pulse at 1.1× threshold amplitude. For I_K1 and AP recordings, Tyrode’s solution was used as bath solution, containing (in mM): 135 NaCl, 5.4 KCl, 1.8 CaCl₂, 1 MgCl₂, 0.33 NaH₂PO₄, 5 HEPES, and 5 glucose; and pipette solution containing (in mM): 140 KCl, 10 NaCl, 1 CaCl₂, 2 MgCl₂, 10 EGTA, 10 HEPES, and 5 MgATP. For sodium current recordings, bath solution consists of (in mM): 135 NaCl, 1.8 CaCl₂, 1.2 MgCl₂, 2 NiCl₂, 10 HEPES, and 10 glucose; pipette solution consists of (in mM): 115 CsCl, 10 NaCl, 0.5 MgCl₂, 10 TEA-Cl, 10 EGTA, 10 HEPES, and 5 MgATP. Tetrodotoxin (TTX) in micromolar concentrations was also included in bath solution in studies where blockade of Na_v1.5 was desired.

Sharp intracellular recordings

Coverslip plated with confluent NRVM monolayer was transferred into a patch-clamp chamber perfused with Tyrode’s solution at 37 °C. The cell monolayer was paced at 1 Hz by a bipolar point electrode and propagated APs at cells remotely situated from the stimulus site were recorded with a high-access resistance electrode (50–100 MΩ) filled with 3M KCl. Data were acquired and processed in similar manners to whole-cell current-clamp recordings. AP parameters, including resting membrane potential (RMP), AP amplitude (APA), AP duration at 80% repolarization (APD₈₀), and maximum AP upstroke velocity, were extracted using a custom-made MATLAB script.

Optical mapping of action potential propagation and reentry induction in cardiomyocyte cultures

Confluent cell monolayers were optically mapped with a 20 mm diameter hexagonal array of 504 optical fibers (Redshirt Imaging), as previously described^18,36,90. Specifically, monolayers were stained with 10 μM Di-4-ANEPPS (Biotium, 61010) for 5 min at room temperature before being transferred to a temperature-controlled (37 °C) recording chamber filled with Tyrode’s solution. Illumination via a solid-state excitation light source (Lumencor, SOLA SM) was passed through a 520 ± 30 nm bandpass filter to excite the dye, and emitted red fluorescence signals (λ > 590 nm) were collected by the optical fiber array, converted to voltage signals by photodiodes, and recorded at a 2.4 kHz sampling rate with a 750 μm spatial resolution. Action potential propagation was initiated by 10 ms, 1.2 × threshold, 1 Hz stimuli from a bipolar point electrode connected to a Grass Stimulator (Grass Technologies). Light shutter control, data acquisition, and electrical stimulation were synchronized using LabView 8.5. Maximum capture rate (MCR) was determined as the highest pacing rate at which tissue responded in 1:1 fashion. Generation of isochrone maps and calculation of CV and APD₈₀ were performed for 1 Hz pacing using custom MATLAB software, as previously described^91,92. For reentry induction⁹³, NRVM monolayers were stimulated with 15 pulses, at the maximum 1:1 capture rate (MCR). If reentry was not induced and 1:1 capture during pacing was maintained, the pacing rate was increased by 0.5 Hz in the next induction attempt. If 1:1 capture during pacing was lost, the rate was decreased by 0.25 Hz and the monolayer stimulated again. The resulting success or loss of 1:1 capture was then followed by an increase or decrease of pacing rate by 0.125 Hz, respectively, as the last attempt at induction. In the case of successful reentry induction, the recording was performed 1, 2, and 5 min later to assess if reentry was sustained long term. Incidence of reentry induction was calculated as the fraction of total monolayers in which sustained reentry (>1 min) was successfully induced.

Computational modeling

BacNa_v model was adapted from Nguyen et al.¹² with updated voltage-clamp and current-clamp experimental data. Modifications were made to the time constant (({tau }_{m}) and ({tau }_{h})) and steady-state functions for activation and inactivation (({{{{{{rm{m}}}}}}}_{{{infty }}}) and ({{{{{{rm{h}}}}}}}_{{{infty }}})) as follows:

The new form of time constant functions was chosen for its ability to produce a wide variety of curve shapes including Gaussian distribution but with asymmetry defined by the shape-fitting parameters. Modeling of BacNa_v effects in different adult cardiomyocyte models was achieved by inserting the BacNa_v equations directly into the Rudy lab models of human²⁵, dog^30,31, and guinea pig^29,30 ventricular myocyte. One-dimensional (1D) cable simulations of AP propagation were performed as described previously¹² using 100 µm cell length, 10 µm cell radius, 1 cm total cable length (100 total cells), and 0.4 kΩ.cm intracellular resistivities. All of human, dog, and guinea pig models were paced at their respective sinus rhythm rates (1 Hz for human, 2 Hz for dog, and 3.33 Hz for guinea pig) until reaching equilibrium (defined when all state variables had variability of <0.001%/beat) and parameters of the last induced AP (in the single-cell model; AP upstroke, APA, and APD₈₀) or conducted AP (in the 1D cable; CV) were determined and used for comparisons among different conditions. Two-dimensional (2D) human cardiac tissue simulations were implemented as a continuous monodomain model. Nonconducting obstacles were randomly generated using a custom MATLAB GUI. After the locations of obstacle nodes were determined, the conductivity was set to 0 for the connections from obstacle nodes to all other nodes and vice versa. Domains were discretized into 100 patches by 100 patches with dx=dy=0.01 cm for a total domain dimension of 1 cm by 1 cm. Human cardiomyocyte formulation and all other conduction parameters (e.g., intracellular resistivity) were maintained from the 1D cable model.

For Brugada syndrome simulations, transmural conduction in a cable of 165 guinea pig ventricular myocytes was modeled as previously described³⁵. Specifically, the cable was divided into the endocardial (cells 1–60), midmyocardial (cells 61–105), and epicardial (cells 106–165) region and stimulated at the endocardial end. The three regions were differentiated by the density of the transient outward potassium current (I_to) and the ratio of current density between the slow and rapid rectifying potassium currents (I_Ks:I_Kr). Endocardial cells had zero I_to and an 11:1 ratio of I_Ks:I_Kr; midmyocardial cells had a max I_to of 0.2125 pA/pF and a 4:1 ratio of I_Ks:I_Kr; and epicardial cells had an I_to of 0.25 pA/pF and a 35:1 ratio of I_Ks:I_Kr. Brugada severity was simulated at two levels by increasing both I_to maximum conductance and the speed of fast inactivation for the endogenous Na_v1.5 current as previously described³⁴. Specifically, mild Brugada case was modeled with 1.5X faster I_Na inactivation and 3X max I_to conductance and severe Brugada case was modeled with 3.5X faster I_Na inactivation and 7X maximum I_to conductance. Simulated BacNa_v current was incorporated into the cable at the 0.2X and 0.5X conductance levels as described. The virtual ECG electrode was placed 2 cm away from the epicardium along the fiber axis³⁵. The pseudo-ECG signal was calculated using the following integral expression taken from Plonsey and Barr⁹⁴:

Where (x’,’y’,z’) is the location in Euclidean space of the simulated point electrode, a is the radius of the fiber (10 µm), σ_i and σ_e are the intracellular and extracellular conductivity, respectively, and r is the Euclidean distance from the source point (x,y,z) to the simulated point electrode. For each simulated Brugada case, ECG deviation <ECG − ECG_Healthy> was calculated by taking the sum of absolute voltage differences overall time points between the Brugada ECG waveform and the healthy ECG waveform.

AAV production and titration

All recombinant AAV viruses were generated using the standard triple transfection method as described previously⁹⁵. Specifically, 293T cells (ATCC, CRL-3216) were co-transfected with the adenoviral helper plasmid pALD-X80 (Aldevron), the packaging plasmid AAV2/9 (gift from James M. Wilson, Addgene plasmid #112865), and the transfer ITR plasmid (1:1:1 molar ratios) using polyethylenimine (PEI) 40K Max transfection reagent (Polysciences). Transfected cells were supplied with fresh media 48–72 h after transfection and both cells and supernatant containing virus particles were collected 120 h after transfection. Collected cells were centrifuged (500 × g, 10 min) and the cell pellet was resuspended in cell lysis buffer (0.15 M NaCl + 0.05 M Tris-HCl, pH 8.5) and lysed through four sequential freeze-thaw cycles (15 min in dry ice/ethanol bath followed by 5 min in 37 °C water bath). AAV-containing cell lysate was collected following centrifugation at 3900 × g and 4 °C for 30 min to remove cell debris. Collected media supernatant was filtered through 0.45 mm cellulose acetate filter (Corning) before being combined with 40% polyethylene glycol (PEG) solution at 4:1 volume ratio for overnight incubation at 4 °C. Concentrated AAV particles were harvested following 15 min centrifugation (2818 × g, 4 °C), resuspended in cell lysis buffer, and combined with viral particles collected from the cell pellet. Benzonase (Millipore Sigma) was added to the virus-containing solution at a final concentration of 50 U/ml with subsequent incubation at 37 °C for 30 min. Viral particles were purified via iodixanol density gradient⁹⁶ ultracentrifugation at 166,880 × g and 17 °C for 15–17 h (WX Ultra 80, Thermo Fisher Scientific). Fractions containing AAV9 were collected and subjected to subsequent phosphate-buffered saline (PBS) buffer exchange using Zeba Spin (40-kDa-molecular-weight cutoff [MWCO]) desalting columns (Thermo Fisher Scientific). Viral titers of purified viruses were determined by quantitative PCR with primers that specifically amplify the AAV2 ITR regions (forward primer, 5’-AACATGCTACGCAGAGAGGGAGTGG-3’; reverse primer, 5’-CATGAGACAAGGAACCCCTAGTGATGGAG-3’) (Integrated DNA Technologies).

Mouse tail-vein injection

All mice were housed in 12 h light/dark cycles, at ambient temperatures of 68–79 degrees Fahrenheit, at a humidity range between 30 and 70%, and with access to food and water ad libitum. Male 6–10-week-old CD-1 mice (Charles River Laboratories) were injected via tail vein with 200 µl of AAV9 solution (2 × 10¹² vg/mouse for AAV9-CAG-h2SheP-2A-GFP, and 1 × 10¹² vg/mouse for AAV9-MHCK7-h2SheP-HA-2A-GFP, scAAV9-MHCK7-h2SheP-HA and scAAV9-MHCK7-GFP). Mice were euthanized by isoflurane inhalation 4–6 weeks post-injection and the hearts were harvested for cardiomyocyte isolation or histology.

Isolation of adult mouse ventricular myocytes

Adult mouse ventricular cardiomyocytes were isolated and cultured according to a previously published Langendorff procedure⁹⁷. Briefly, the heart was excised and enzymatically digested by perfusion of 40 ml prewarmed enzyme solution (Collagenase II 475 U/ml (Worthington), Blebbistatin 15 μmol/L (Stemcell Technologies)) at a rate of 2 ml/min. The collagenase activity was inhibited with fetal bovine serum (FBS) to a final concentration of 10% and the cell suspensions were passed through a 200 μm filter (BD Biosciences). Calcium concentration was gradually restored using 3 intermediate calcium reintroduction buffers (prepared as previously described³) and the cells were allowed to settle by gravity each round for 15 min. The final cell pellet was resuspended in a plating medium and plated onto 21 mm diameter Aclar (Ted Pella) coverslips coated with laminin (5 μg/mL, Thermo Fisher Scientific) and incubated for 4 h at 37 °C before patch-clamp studies.

Mouse electrocardiograms

All measurements were conducted and analyzed in a blinded fashion. Mice were anesthetized using a volatile anesthetic system with an induction chamber (R5835, RWD Life Science, Dover, Delaware, United States). Electrocardiographic (ECG) measurements were performed using four subdermal leads: I, aVR, aVL, and aVF. ECG parameters, such as RR, PR, QRS, QT, and corrected QT for heart rate (Bazett’s QT correction) were measured and/or calculated at baseline and following adrenergic stimulation and ryanodine receptor sensitization with 200 μg/g caffeine and 1 μg/g isoproterenol IP. Rhythm detection was captured by an iWorx-RA-834 Eight Channel 16-bit Data Acquisition System (iWorx, Dover New Hampshire, United States). Data were viewed using a custom-built ECG Analysis Module software program for LabScribe v4.

SAN dissection for immunostaining

Hearts from heparinized mice (200U i.p.) were perfusion-fixed with 4% PFA and immersed in 30% (w/v) sucrose overnight. Ventricles were removed and atria were pinned on a PDMS mold and visualized using a stereomicroscope (DFC7000T; Leica). The SAN region was identified using the superior and inferior vena cava, the right atrial appendage, the crista terminalis, and the interatrial septum as landmarks. The SAN preparation including right and left atria were embedded and frozen in OCT compound (VWR) using a dry ice/isoproterenol bath, cut into 10 µm sections using a cryostat (Leica), and immunostained as described below.

Immunostaining and imaging

Cell monolayers were fixed in 4% paraformaldehyde (PFA) for 10 min at room temperature. Hearts were fixed and sectioned as described for the SAN tissue. Fixed monolayers or heart sections were permeabilized and blocked in blocking solution (5% chicken serum + 0.1% Triton-X, 30 min). The following primary antibodies (1 h incubation) were used: anti-sarcomeric α-actinin (Sigma, a7811, 1:200), anti-vimentin (Abcam, ab92547, 1:500), anti-Cx43 (LSBio, LS-B9770, 1:300), anti-HCN4 (Alomone, APC-052, 1:200), anti-cardiac troponin T (Abcam, ab45932, 1:200), and anti-HA tag (Cell Signaling Technology, C29F4, 1:200). Secondary antibodies (1 h incubation) included: Chicken anti-Mouse Alexa Fluor 488 (Thermo Fisher Scientific, A-21200/A-21441, 1:200), Chicken anti-Mouse Alexa Fluor 594 (Thermo Fisher Scientific, A-21201/A-21442, 1:200), Chicken anti-Mouse Alexa Fluor 647 (Thermo Fisher Scientific, A-21463, 1:200), Donkey anti-Rabbit Alexa Fluor Plus 594 (Thermo Fisher Scientific, A-32754, 1:200), Donkey anti-Rabbit Alexa Fluor Plus 647 (Thermo Fisher Scientific, A-32795, 1:200), Alexa Fluor 488-conjugated phalloidin (Thermo Fisher Scientific, A12379, 1:300), Alexa Fluor 647-conjugated phalloidin (Thermo Fisher Scientific, A22287, 1:300), DAPI (Sigma, D9542, 1:300). All immunostaining steps were performed at room temperature. Fluorescence images were acquired using inverted fluorescence (Nikon TE2000) or confocal (Leica SP5, Andor Dragonfly) microscope, and processed with ImageJ software.

Statistics and reproducibility

All statistical analyses and data plotting were performed using Prism (GraphPad Software Inc.). D’Agostino–Pearson test was used to confirm data normality. Data are presented as mean ± s.e.m. and represent a minimum of three independent experiments with at least three biological and technical replicates unless otherwise stated. For comparisons of two experimental groups, statistical significance was evaluated with a standard unpaired Student t-test (two-tailed; P < 0.05) or Chi-square test (two-tailed; P < 0.05). For multiple-comparison analysis, statistical significance was determined by one-way or two-way ANOVA, followed by Tukey’s post-hoc test to calculate P-values. Statistical significance was defined as P < 0.05 (95% confidence). For all results, the exact P-value, number of biological replicates, and statistical test used are reported in figure legends. All shown images are representative of three independent experiments with at least three biological and technical replicates.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.