GJA1-20k Arranges Actin to Guide Cx43 Delivery to Cardiac Intercalated Discs

Abstract

Rationale

Delivery of connexin 43 (Cx43) to the intercalated disc is a continuous and rapid process critical for intercellular coupling. By a pathway of targeted delivery involving microtubule highways, vesicles of Cx43 hemichannels are efficiently trafficked to adherens junctions at intercalated discs. It has also been identified that actin provides rest stops for Cx43 forward trafficking, and that Cx43 has a 20kDa internally translated small C-terminus isoform (GJA1-20k) which is required for full-length Cx43 trafficking, but by an unknown mechanism.

Objective

We explored the mechanism by which the GJA1-20k isoform is required for full-length Cx43 forward trafficking to intercalated discs.

Methods and Results

Using an in-vivo AAV9-mediated gene transfer system, we confirmed in whole animal that GJA1-20k markedly increases endogenous myocardial Cx43 gap junction plaque size at the intercalated discs. In micropatterned cell pairing systems, we found that exogenous GJA1-20k expression stabilizes filamentous actin (F-actin) without affecting actin protein expression, and that GJA1-20k complexes with both actin and tubulin. We also found that F-actin regulates microtubule organization as inhibition of actin polymerization with a low dose of latrunculin A (LatA) disrupts the targeting of microtubules to cell-cell junctions. GJA1-20k protects actin filament from LatA disruption, preserving microtubule trajectory to the cell-cell border. For therapeutic implications, we found that prior in vivo AAV9-mediated gene delivery of GJA1-20k to the heart protects Cx43 localization to the intercalated discs against acute ischemic injury.

Conclusions

The internally translated GJA1-20k isoform stabilizes actin filaments which guides growth trajectories of the Cx43 microtubule trafficking machinery, increasing delivery of Cx43 hemichannels to cardiac intercalated discs. Exogenous GJA1-20k helps to maintain cell-cell coupling in instances of anticipated myocardial ischemia.

INTRODUCTION

The mechanism by which membrane proteins are localized to their respective subdomain remains largely unknown. Seminal trafficking studies in cell biology have uncovered the major organelles of the cell¹, zip codes which allow proteins to recognize their destination once they arrive², and the lipid transport machinery which performs the transport³. However, we have yet to understand how pre-arrival delivery specificity of proteins is achieved.

The ventricular cardiomyocyte is a highly organized and polarized cell with distinct membrane subdomains, and thus an excellent model for studying delivery specificity. Furthermore, connexin 43 (Cx43) gap junction proteins tend to cluster and have very high turnover rates, both are features which can be exploited in live cell, interventional, and in vivo studies to uncover mechanisms of membrane protein trafficking. Physiologically, Cx43 gap junctions comprise of hexameric hemichannels from neighboring cells that form gap junctions at the intercalated discs which are essential for electrical coupling in cardiac ventricles^{4, 5}, synchronizing beat-to-beat heart contractions^{6, 7}. Considering the specific localization of Cx43 and rapid protein turnover which is within several hours⁸, the intracellular movement of Cx43 is highly regulated, disruptions of which are causal for lethal arrhythmias⁹.

Growing evidence supports a cytoskeletal-based trafficking mechanism for Cx43 delivery directly to the intercalated disc^{10, 11}. It is understood that Cx43 hemichannels oligomerize in the trans-Golgi network¹², and are subsequently packaged into vesicles transported by motor proteins along microtubules^13–15. Our previous work identified a targeted delivery paradigm^16–18, by which the specificity of Cx43 trafficking is in part achieved by the interaction between microtubule plus-end binding proteins and the membrane scaffolding proteins of the intercalated disc^{17, 18}. We have also discovered that actin is a necessary component of the Cx43 forward trafficking machinery, serving as rest stops at which Cx43 vesicles pause and slow down before they are delivered to the cell surface along microtubules¹⁹. However, the interrelationship between the actin and microtubule machineries for Cx43 trafficking is not understood. Adding to the complexity of regulation of Cx43 trafficking is the autoregulation of Cx43 localization by smaller accessory subunits²⁰ generated through internal translation of the GJA1 mRNA²¹. In hearts and cell lines, up to six N-terminal truncated isoforms containing the Cx43 C-terminal cytoplasmic tail have been identified^21–23, of which the 20-kilodalton isoform (GJA1-20k) is the most abundantly expressed. We have found that GJA1-20k aids in Cx43 gap junction localization²¹, but how trafficking regulation is accomplished by this particular isoform, and whether GJA1-20k contributes specificity to cytoskeletal-based Cx43 transport remain unknown.

In this study, we present the finding that GJA1-20k contributes to the specificity of Cx43 trafficking by stabilizing actin polymerization to organize the growing paths of microtubules towards the intercalated discs, onto which Cx43 containing vesicles are loaded for delivery. Using an in-vivo AAV9-mediated gene transfer system in adult mice, we find that exogenous GJA1-20k markedly increases endogenous Cx43 gap junction plaque size at the intercalated discs. In Hela cells, exogenous GJA1-20k not only complexes with both actin and tubulin, but also stabilizes filamentous actin (F-actin) in micropatterned cells. These results are further supported in micropatterned neonatal mouse ventricular cardiomyocytes, where GJA1-20k stabilizes F-actin fibers and preserves microtubule trajectory to the cell-cell border when cells are subjected to actin depolymerization by low dose LatA. Moreover, by maintaining the actin and microtubule cytoskeletal system, GJA1-20k protects Cx43 localization to the intercalated discs following acute ischemic injury in isolated mouse hearts. These findings reveal that delivery specificity of Cx43 gap junction is dependent on the internally translated GJA1-20k isoform that stabilizes F-actin, which in turn is required to organize microtubule directed trajectories to the cell-cell border for Cx43 transport. The results further indicate that exogenous GJA1-20k can serve to preserve cardiac gap junction coupling in the setting of anticipated ischemic insults.

METHODS

An expanded Methods section is available in the Online Data Supplement at http://circres.ahajournals.org.

Mice

C57BL/6 mice used for heart preparations were maintained under sterile barrier conditions. All procedures were reviewed and approved by Cedars-Sinai Medical Center Institutional Animal Care and Use Committee.

Molecular biology

Human GJA1 cDNAs encoding full-length Cx43 and smaller isoforms were cloned as previously described²¹ with internal methionine start sites in GJA1-43k, and GJA1-20k mutagenized to leucine (QuickChange Lightning Mutagenesis Kit, Agilent) to ensure single isoform expression. C-terminal V5-, HA- or GFP-tagged proteins were subsequently made.

AAV9 gene delivery

AAV9 vectors expressing GST-GFP, GJA1-20k-GFP, or GJA1-43k-GFP driven by the CMV promoter were produced and purified at Welgen Inc. (Worcester, MA). 8–12 week old Male C57BL/6 mice received 100 microliters of 3 × 10¹⁰ vector genomes (vg) through retro-orbital injection^{24, 25}. 4 weeks post injection, the hearts were excised and processed for immunofluorescence imaging, biochemical analysis, or RNA expression.

Langendorff-perfused mouse heart preparation

4 weeks post AAV9 injection, a subset of mouse hearts was excised, ex vivo perfused on a Langendorff-perfusion apparatus, and subjected to acute ischemic injury study as previously described¹⁹. Briefly, hearts were perfused using a Langendorff apparatus (ADInstruments) at a constant rate of 2.6 ml/min with Krebs-Henseleit (K-H) buffer, which was constantly gassed with 95% O2/5% CO2, and maintained at 37°C. Equilibration was carried out for 20 minutes before hearts were subjected to 30 minutes of global ischemia. During no-flow ischemia, the heart was immersed in warm K-H buffer to maintain temperature. Control hearts were perfused continuously throughout the protocol. Immediately after the procedure, hearts were embedded in OCT media (Sakura Finotek) and snap frozen for cryosectioning.

Tissue immunofluorescence

For tissue immunofluorescence, cryosections (10 μm) were fixed in 4% PFA for 20 minutes at room temperature, blocked and permeabilized at room temperature for 1 hour with 10% normal goat serum (NGS) and 0.5% TritonX-100 in PBS and then incubated with primary antibodies diluted in 5% NGS and 0.1% TritonX-100 in PBS (antibody diluent) at 4°C overnight. Slides were then incubated with Alexa fluor secondary antibodies (Thermo Fisher Scientific) in antibody diluent for 1 hour at room temperature and mounted using ProLong gold-containing DAPI (Thermo Fisher Scientific) for image acquisition.

Quantification of Cx43 at the intercalated disc

Quantification of Cx43 expression at N-Cadherin-containing intercalated disc regions was carried out as previously described^{18, 19} using average intensity projections of 13.5 μm confocal z-stacks. N-Cadherin images were used to generate binary masks of intercalated disc regions. Masks were image-multiplied by the corresponding Cx43 image, thus excluding all Cx43 signal except those at the intercalated disc. Cx43 plaque fluorescence intensity was subsequently measured and normalized to GST control or to N-Cadherin signal at the intercalated disc regions.

Triton solubility assay

Snap-frozen mouse ventricular tissue samples were weighed and homogenized in 1% Triton X-100 buffer at 100 mg tissue/ml as previously described¹⁸. Protein lysate was added to an equal volume of 1% Triton X-100 buffer containing 8 M urea and 2 M thiourea (Sigma-Aldrich) to generate the total protein fraction. The remaining lysate was centrifuged and an equal volume of 1% Triton X-100 buffer containing 8 M urea and 2 M thiourea was added to the supernatant for the soluble fraction (non-junctional Cx43). The pellete resulting from the previous centrifugation was mixed with 1% Triton X-100 buffer containing 4 M urea and 1 M thiourea at a final concentration of 30 mg/ml to generate the non-soluble fraction (junctional Cx43). Sample buffer (ThermoFisher Scientific) containing 100mM DTT was added to all fractions and the samples were reduced for 45 minutes at room temperature before Western blotting.

Western blotting

Western blotting was performed as previously described²¹. Briefly, samples were subjected to SDS-PAGE electrophoresis using NuPAGE Bis-Tris gels and MES (Thermo Fisher Scientific) buffer according to the manufacturer’s instructions. Gels were transferred to FluoroTrans PVDF membranes (Pall) and subsequently fixed by soaking in 100% methanol and air drying before rewetting with methanol and blocking for 1 hour at room temperature in 5% nonfat milk in TNT (50 mM Tris, pH8.0, 150 mM NaCl, 0.1% Tween-20) buffer. Membranes were probed overnight with primary antibodies in 5% milk in TNT buffer then incubated for 1 hour at room temperature with secondary antibodies in 5% milk in TNT. Membranes were then immersed in 100% methanol, air dried and imaged using the ChemiDoc MP fluorescent western detection system (BioRad).

Quantitative Real Time PCR

Total RNA was isolated from mouse hearts 4 weeks post AAV9 retro orbital injection, using Trizol and PureLink RNA mini kit (Thermo Fisher Scientific) according to manufacturer’s instructions. The purified RNA was further treated for any DNA contamination using Turbo DNA-Free Kit (Thermo Fisher Scientific). cDNA was synthesized using SuperScript IV VILO Master Mix (Themro Fisher Scientific). Taqman gene expression assays, using Taqman Universal Master Mix and Taqman probes (Thermo Fisher Scientific), were prepared for quantitative Real Time PCR reactions and ran using Biorad CFX Connect Real-Time System.

Micropatterning of cell pairs

We used soft-microlithography ²⁶ to pattern HeLa cells and neonatal mouse myocytes, as previously pubished.^27–32.

Hela cell transfection

Hela cells were transfected with pDEST-GST-GFP or pDEST-GJA1-20k-GFP using Lipofectamine 2000 according to manufacture’s instructions (Thermo Fisher Scientific). Transfection efficiency was assessed by examining GFP fluorescence at 16–24 hours following transfection. The cells were detached using EDTA and seeded at a density of 10⁵ cells/ml on the micropatterned coverslips and cultured overnight in the 37°C incubator. The following day, the cells were washed with 1x PBS (3x5 min washes), and fixed in 4% PFA (Electron Microscopy Services) for 20 minutes at room temperature before immunofluorescence studies

Electron microscopy

For transmission electron microscopy (TEM), HeLa cells were transfected with either GST-GFP or GJA1-20k-GFP plasmids as described above. Cells were fixed in 2% glutaraldehyde in PBS, scraped and centrifuged and the resulting pellet was post-fixed with 1% osmium tetroxide and incubated in 3% uranyl acetate. Samples were then dehydrated in ethanol, treated with propylene oxide, embedded in Spurr resin (Electron Microscopy Services), and sectioned using an ultramicrotome (UCT, Leica). Sections were mounted on EM grids and stained with uranyl acetate and lead citrate followed by imaging. All electron microscopy work was done by the core facility at the Electron Imaging Center of The California NanoSystems Institute, UCLA.

Immunofluorescence staining

Cells were fixed in either 4% paraformaldehyde (PFA, Electron Microscopy Services) or in ice cold methanol. The cells were permeabilized and blocked at room temperature and incubated with primary antibody solution at 4°C overnight. The cells were then incubated at room temperature with antibody diluent containing Alexa fluor secondary antibodies (Themro Fisher Scientific) with phalloidin (Thermo Fisher Scientific) or with wheat germ agglutinin (WGA) (Thermo Fisher Scientific). ProLong gold-containing DAPI (Thermo Fisher Scientific) was used to mount slides for image acquisition.

Fiber quantification

Fiber number and length were quantified using imageJ (single z-plane with background subtracted).

Co-immunoprecipitation

Co-IP was carried out using transiently transfected HeLa cells expressing HA-tagged GJA1-20k. Cells were lysed in lysis buffer (150 mM KCl, 20 mM HEPES (pH7.4), 2 mM MgCl₂, 2 mM K₂HPO₄, 1 mM DTT, 25 μM phalloidin, and Halt protease phosphatase inhibitors) containing 0.5% NP40. The cell lysate was precleared using Dynabeads protein G (Thermo Fisher Scientific) and immunoprecipitation was undertaken at 4°C overnight, using 5μg of either mouse anti-HA (4C12, Abcam), or mouse anti-GST (B-14, Santa-Cruz Biotechnology) as negative isotype control. Dynabeads protein G were added to each reaction, and tubes were rotated for an additional hour at 4°C. Protein complexes were washed with lysis buffer containing 0.1% NP40, and pelleted using a Dynamag-2 magnet. Proteins were then eluted with 2X NuPAGE sample buffer supplemented with 100 mM DTT, incubated at 37 degrees for 20 minutes and subjected to SDS-PAGE electrophoresis and Western blotting as described above.

Cardiomyocyte isolation, transduction and latrunculin A treatment

Primary neonatal mouse ventricular cardiomyocytes were isolated and maintained in culture medium as previously described³³. Briefly, ventricles were dissected from postnatal hearts and digested with trypsin (Invitrogen) and type II collagenase (Worthington) at 37°C with constant gentle stirring. Cells were pre-plated on TC-treated culture dishes (Corning) to remove fibroblasts, before they are seeded on the micropatterned coverslips in culture media containing arabinofuranosyl cytidine and 5-bromo-20-deoxyuridine to inhibit rapidly proliferating non-myocytes. Cardiomyocytes were then transduced overnight with adenovirus carrying GJA1-43k-V5, GJA1-20k-V5 or GFP-V5 (produced at the CURE Vector Core Facility at UCLA) and then treated with either 0.1% DMSO control or 250 nM latrunculin A for 1 hour, and fixed in 100% ice-cold methanol (5 minutes at −20°C) or in 4% PFA (Electron Microscopy Services) for 20 minutes at room temperature and stored in PBS at 4°C for immunofluorescence studies.

Quantification of α-tubulin and EB1 comets at the cell-cell border

The cell-cell border was labeled using either N-Cadherin or WGA and the cell counter plugin for imageJ was used to quantify the number of EB1 comets or α-tubulin molecules reaching the cell-cell border. The number of EB1 comets and α-tubulin molecules at the border were normalized to the cell border length.

Statistical analysis

All quantitative data were expressed as mean +/− s.e.m. and analyzed using Prism 6 software (GraphPad). For comparison between two groups, unpaired two-tailed student’s t-test was performed. For comparison among three and more treatment groups, one-way ANOVA followed by either Bonferroni’s or Tukey’s post-test was performed.

RESULTS

GJA1-20k increases Cx43 gap junction plaque size at the intercalated disc in vivo

Based on our previous in vitro studies indicating that the GJA1-20k improves Cx43 gap junction plaque size²¹, we tested whether GJA1-20k regulates Cx43 gap junction plaque size in vivo. We introduced GST-GFP control, GJA1-43k-GFP (lacking all internal methionines to generate only the full-length protein), or GJA1-20k-GFP (also lacking downstream internal methionines to generate only the 20k isoform) into adult mouse heart by recombinant AAV9-mediated gene transfer³⁴. Each mouse received 3 × 10¹⁰ vector genomes (vg) on experimental day 0 (Figure 1A). 4 weeks after AAV9 delivery, hearts were harvested and assessed for Cx43 localization at the intercalated discs. Using an antibody specific to the Cx43 N-terminus, thus only detecting full length protein, we examined Cx43 localization in each AAV9 group by confocal imaging. N-terminal Cx43 immunofluorescence signal (green) is detected at intercalated discs which are demarcated by N-Cadherin (red, Figure 1B). Introduction of full-length GJA1-43k (Figure 1B, middle panel) increased the overall intracellular accumulation of Cx43 signal, whereas GJA1-20k (Figure 1B, lower panel) robustly increased Cx43 immunofluorescence mainly at the intercalated discs. Quantification of Cx43 intensities at intercalated disc regions revealed that AAV9-mediated expression of GJA1-20k-GFP alone significantly increases Cx43 gap junction localization at the intercalated discs, as compared to GJA1-43k-GFP and GST-GFP groups (Figure 1C, all data normalized to the GST group). The increase in GJA1-20k holds when normalized to mean N-Cadherin intensity (Figure 1D), which is not surprising given no significant difference in N-Cadherin between the test groups (Online Figure I).

Figure 1. GJA1-20k increases Cx43 gap junction plaques at the intercalated discs in vivo.

(A) Schematic of the AAV9 gene delivery protocol. Mice received via retro-orbital injection 3×10¹⁰ vg of AAV9 vectors expressing GST-GFP, GJA1-43k-GFP, or GJA1-20k-GFP. Localization of Cx43 at the intercalated discs was assessed at 4 weeks post injection. (B) Immunofluorescence detection of N-Cadherin (red, marking intercalated disc structures) and Cx43 (anti-N-terminus, green) in heart sections from the different AAV9 treatment groups. Average intensity projections of confocal z-stacks (13.5 μm thickness), scale bar = 50 μm. (C) Quantification of Cx43 fluorescence intensity in intercalated disc regions. Data are presented as mean ±SEM relative to GST group (n = 3 hearts per group, 6–15 images analyzed per heart), *p < 0.05, by one-way ANOVA followed by Tukey’s post-hoc-test. (D) Quantification of Cx43 fluorescence intensity normalized to N-Cadherin signal. Data are presented as mean ±SEM (n = 3 hearts per group, 6–15 images analyzed per heart), *p < 0.05, **< 0.01, by one-way ANOVA followed by Tukey’s post-hoc-test.

The Cx43 distribution was further examined using a Triton solubility assay to biochemically separate the heart tissue lysates into soluble (non-junctional Cx43) versus insoluble (junctional Cx43) fractions^{18, 35}. As seen in representative Western blots in Figure 2A (quantified in B and C), GJA1-43k transduction led to full-length Cx43 protein accumulation primarily in the non-junctional fraction (soluble Cx43) whereas GJA1-20k preferentially increased Cx43 amount in the junctional fraction (insoluble Cx43). This result supports the increase of Cx43 localization at plaques (Figure 1). To examine whether GJA1-20k is regulating Cx43 trafficking and not influencing overall expression of Cx43, we assessed the transcriptional level of Cx43 in the hearts of mice from each group. Cx43 mRNA levels were not significantly changed in the GJA1-20k-GFP animals when compared to the GST-GFP AAV9 control mice (Figure 2D).

Figure 2. GJA1-20k increases junctional Cx43 protein levels without affecting Cx43 transcription.

(A) Adult mice received via retro-orbital injection 3×10¹⁰ vg of AAV9 vectors expressing either GST-GFP, GJA1-43k-GFP or GJA1-20k-GFP. 4 weeks following AAV9 gene expression, mouse hearts were subjected to TritonX-100-based tissue fractionation of soluble (non-junctional) and insoluble (junctional) proteins and probed for Cx43, α-tubulin, and N-Cadherin using Western blot analysis. (B–C) Quantification of the amount of Cx43 in the soluble (B) and insoluble (C) fractions, normalized to Cx43 input and expressed as fold change relative to GST control. Data are presented as the mean ± SEM (n = 3 hearts per group) *p < 0.05, ** < 0.01, by one-way ANOVA followed by Tukey’s post-hoc-test. (D) 4 weeks following AAV9 gene expression of GST-GFP or GJA1-20k-GFP (retro-orbital injection 3×10¹⁰ vg), Cx43 mRNA level was assessed in the mouse heart using Taqman gene expression assay and Real-Time PCR. Data are presented as mean ±SEM relative to GST group (n = 3 hearts per group), no significance was shown by student’s t-test

Taken together, the imaging and biochemical data both indicate that exogenous GJA1-20k alone is sufficient to aid in targeted delivery of endogenous Cx43, increasing gap junction plaques at the intercalated discs in vivo. Interestingly, the increase in Cx43 plaques is significantly greater with exogenous GJA1-20k than with exogenous full length GJA1-43k, highlighting the value of enhancing the trafficking apparatus rather than the protein substrate itself for its efficient localization into appropriate membrane subdomain.

GJA1-20k stabilizes F-actin fibers

We have previously found that actin in general facilitates Cx43 forward trafficking¹⁹. Actin is known to be responsive to the C-terminal tail of Cx43³⁶ which is homologous with GJA1-20k¹⁰. We therefore used GJA1-20k overexpression to screen for the effects of GJA1-20k on the actin cytoskeleton. HeLa cells were used because they have low endogenous levels of Cx43 and are amendable to imaging experiments. Using transmission electron microscopy (TEM) imaging in unpaired cells, we found that GJA1-20k overexpression results in a dramatic increase in both actin fiber length and number when compared to GST transfected control cells (Figure 3A, top panel arrows). The individual fibers are traced in red for better visualization (Figure 3A, bottom panel). On average, GJA1-20k increases average actin fiber length by 50% and doubles fiber number (Figure 3A, right panels).

Figure 3. GJA1-20k isoform promotes actin polymerization and stabilizes F-actin.

(A) Representative TEM images of HeLa cells transfected with GST-GFP control or GJA1-20k-GFP, showing fiber structures indicated by red arrows (top panels) and red line traces (lower panels). Scale bar = 500 nm. Fiber length (GST n=193 and GJA1-20k n=368 fibers measured from 13 cells for each treatment) and fiber number (n=13 images per group) are quantified in the graphs. Data are mean ± SEM. * p < 0.05, ****<0.0001 by student’s t-test. (B) Confocal images of micropatterned Hela cells transfected with GST-GFP control or GJA1-20k-GFP and treated with phalloidin for F-actin labeling. Scale bar = 10μm. Actin fiber length (GST n=420 and GJA1-20k n=253 actin fibers measured from 10–12 cells) and actin signal intensity (GST n=23 and GJA1-20k= 25 cells) are quantified in the graphs. Data are mean ± SEM. * p < 0.05, ****<0.0001 by student’s t-test. (C) Western blotting of GJA1-20k, actin, and GAPDH in HeLa cells transfected with GFP tagged GST or GJA1-20k. (D) Co-immunoprecipitation assay with exogenous GJA1-20k-HA (IP with anti-HA or anti-GST antibodies) in HeLa cells and immunoblotting (IB) with anti-Cx43-CT, anti-actin or anti-α-tubulin antibodies.

Given that full length Cx43 concentrates at cell-cell borders, the effect of GJA1-20k on actin fibers directed toward cell-cell borders was then studied in a micropatterned cell pairing system with controlled cell morphology and actin fiber orientation³³. Confocal images of micropatterned HeLa cells reveal that exogenous GJA1-20k expression can markedly promote the stability of F-actin fibers when compared to GST transfected cells (Figure 3B) and untransfected cells (Online Figure II). GJA1-20k significantly increases actin fiber length and actin signal intensity as seen in the bar graphs (Figure 3B) without altering total actin protein levels (Figure 3C), indicating that GJA1-20k stabilizes polymerized actin fibers rather than protein synthesis. These data suggest that GJA1-20k and actin fibers can be involved in making trafficking routes towards membrane subdomains. Using biochemical immunoprecipitation with HA-tagged GJA1-20k expressing HeLa cells, we find that GJA1-20k complexes with both actin and tubulin (Figure 3D), further indicating that GJA1-20k is a critical component of the actin-microtubule trafficking machinery.

GJA1-20k stabilizes -actin microfilaments in cardiomyocytes

Our previous findings indicate that Cx43 complexes with non-sarcomeric β-actin¹⁹ which is a central component of intracellular microfilament fibers responsible for cellular roles other than β-actin thin filament mediated contraction³⁷. Similar to our finding that GJA1-20k stabilizes F-actin fibers in HeLa cells (Figure 3), we find that GJA1-20k stabilizes β-actin microfilament fibers in micropatterned neonatal mouse ventricular cardiomyocytes (Figure 4A). The cardiomyocytes were transduced with GFP-V5 control, GJA1-43k-V5 or GJA1-20k-V5 adenoviruses (4 MOI) and treated with a low dose of LatA (250 nM) to impair actin polymerization. LatA caused about a 60% decrease in the number of β-actin filaments as compared to that of the GFP-V5 and GJA1-43k-V5 adenovirus transduced cardiomyocytes (Figure 4A, middle panels, and Figure 4B). Transducing cardiomyocytes with GJA1-20k-V5 resulted in partial rescue (protection) of β-actin filaments from impairement induced by LatA treatment (Figure 4A, bottom panel and Figure 4B), indicating stabilization of polymerized actin fibers by GJA1-20k, which is consistent with results from HeLa cells (Figure 3).

Figure 4. GJA1-20k promotes β-actin polymerization, preserving and protecting the F-actin structure in cardiomyocytes.

(A) Confocal images of micropatterned neonatal mouse ventricular cardiomyocytes transduced with GFP-V5, GJA1-43k-V5 or GJA1-20k-V5 adenovirus and treated with DMSO or 250nM of Latrunculin A (LatA) and labeled for β-actin by immunofluorescence. Box inserts show zoomed in areas of the cardiomyocytes. Scale bar =5μm. β-actin fiber number is quantified in (B) and data are presented as mean ±SEM (number of cells quantified for each group shown on the bars), *p < 0.05, **< 0.01, ***< 0.001, ****< 0.0001 by one-way ANOVA (Kruskal-Wallis test) followed by Dunn’s multiple comparisons test.

F-actin stabilization is required to orient microtubule growth towards the cell-cell junctions

To reach adherens junctions at cell-cell borders, Cx43 hemichannels undergo targeted delivery, whereby Cx43-containing vesicles are trafficked along microtubules which anchor at N-Cadherin containing membrane for offloading^{17, 18}. Using high resolution confocal imaging and micropatterned neonatal mouse ventricular cardiomyocyte pairing, we explored the role of GJA1-20k and actin cytoskeleton in regulating directionality and attachment of the microtubule trafficking machinery (Figure 5). Inhibition of actin polymerization with a low dose of LatA greatly decreased the number of microtubules reaching the fascia adherens at the cell-cell junctions (Figure 5A–D). Microtubules were marked with either the plus-end tracking protein (+TIP) EB1 (Figure 5A–B) or α-tubulin (Figures 5C–D). EB1 marks rapidly growing tips of microtubules and is essential for Cx43 forward trafficking^{17, 18}. Quantification of EB1 and α-tubulin (green) labeled microtubules reaching the N-Cadherin defined cell-cell border (red) is shown in Figures 5B and 5D respectively. Strikingly, in these micropatterned cardiomyocyte pairs (Figure 5E), only exogenous GJA1-20k but not GFP or GJA1-43k is protective of microtubule trajectories from LatA effects, normalizing microtubule targeting (α-tubulin labeling, arrows in the skeleton image) to the cell-cell borders (defined by WGA labeling, dotted line). Quantification of microtubule number reaching the cell-cell border in the different treatment groups is shown in Figure 5F. These results indicate that GJA1-20k organizes the actin cytoskeleton which in turn is required for the specific targeting of microtubules to the cellular junctions, placing the GJA1-20k arranged actin cytoskeleton upstream of the microtubule-based Cx43 trafficking machinery.

Figure 5. Actin polymerization as regulated by GJA1-20k is required to orient microtubule growth trajectories towards the cell-cell border.

(A) Confocal images of micropatterned neonatal mouse ventricular cardiomyocyte pairs treated with DMSO or 250 nM of LatA and labeled for EB1 (green, marking rapidly growing tips of microtubules) and N-Cadherin (red, marking cell-cell border) by immunofluorescence. (B) Number of EB1 molecules touching the cell-cell border is quantified and normalized to border length as shown in the graph. Data are presented as mean ±SEM, n= 25 cell pairs analyzed for each group from 2 dishes (10–15 images per dish),****p< 0.0001 by students’ t-test. (C) Confocal images of micropatterned neonatal mouse ventricular cardiomyocytes treated with DMSO or 250 nM of LatA and labeled for α-tubulin (green, marking microtubule structures) and N-Cadherin (red) by immunofluorescence. (D) Number of α-tubulin molecules touching the cell-cell border is quantified and normalized to border length as shown in the graph. Data are presented as mean ±SEM, n= 19 cell pairs for each group from 2 dishes (9–10 images per dish),****p< 0.0001 by students’ t-test. (E) Confocal images of micropatterned neonatal mouse cardiomyocyte pairs transduced with GFP-V5, GJA1-43k-V5 or GJA1-20k-V5 adenovirus and treated with DMSO or 250 nM of LatA and labeled for α-tubulin (green) by immunofluorescence. The cell-cell border is labeled with WGA and shown as a traced dotted line. The microtubules reaching the border are traced with arrows (right panels). (F) Quantification of microtubules number that are touching the cell-cell border, normalized to border length. Data are presented as mean ±SEM (n= 10 cells per group), *p < 0.05, ***< 0.001 by one-way ANOVA followed by Tukey’s post-hoc-test.

GJA1-20k maintains Cx43 gap junction localization at the intercalated discs after acute ischemia

Our previous findings indicate that the acute ischemia in Langendorff-perfused mouse hearts or actin disruption with LatA treatment, decreased Cx43 localization at the intercalated discs by disrupting Cx43/β-actin interaction. The quantified level of disruption was similar between ischemia and actin disruption¹⁹. Given that GJA1-20k maintains the Cx43 trafficking machinery in the setting of actin disruption (Figure 5), we tested whether GJA1-20k is protective against Cx43 plaque disruption following acute ischemia. For these experiments, we again utilized the recombinant AAV9-mediated gene transfer system to deliver GST-GFP control, GJA1-43k-GFP, or GJA1-20k-GFP into the mouse heart in vivo. Each mouse received 3 × 10¹⁰ vector genomes (vg) on experimental day 0. Four weeks later, hearts were rapidly excised and perfused ex vivo on a Langendorff system. Hearts were then subjected to either 30 min of no flow ischemia or were continually perfused as a control (Figure 6A). Immediately after ischemia, hearts were mounted in OCT medium and snap frozen for later cryosection preparation. Using confocal imaging of immunofluorescence labeled fixed slices, we assessed Cx43 (N-terminus antibody, green signal) localization at the intercalated discs marked with N-Cadherin (red) (Figure 6B). Results are that the Cx43 signal detected at intercalated discs is reduced in hearts transduced with GST-GFP and GJA1-43k-GFP AAV9 vectors after ischemia (Figure 6B, middle 2 panels) as compared to the no ischemia control hearts (Figure 6A, left panel). Strikingly, in hearts transduced with GJA1-20k AAV9, robust Cx43 immunofluorescence is still detected at the intercalated discs after ischemia (Figure 6A, right panel). Quantification of Cx43 intensities at intercalated disc regions revealed that only AAV9-mediated expression of GJA1-20k-GFP maintained Cx43 localization to the intercalated discs following ischemia, to a level not significantly different from that is observed in the control non-ischemic hearts (Figure 6C–D; Cx43 plaque fluorescence normalized to N-Cadherin signal at the intercalated discs). Together these data indicate that in an acute ischemic injury setting, full length GJA1-43k protein by itself is not sufficient to maintain Cx43 localization at the cell-cell junctions while the GJA1-20k isoform is protective of Cx43 forward delivery to the intercalated discs thus preserving the gap junction plaques.

Figure 6. GJA1-20k maintains Cx43 gap junction localization to the intercalated discs after acute ischemia.

(A) Adult mice received via retro-orbital injection 3×10¹⁰ vg of AAV9 vectors expressing GST-GFP, GJA1-43k-GFP, or GJA1-20k-GFP. 4 weeks post injection, the hearts were excised and maintained using Langendorff perfusion apparatus for 20 minutes followed by 30 min of no flow ischemia or continuous perfusion for control hearts. (B) Immunofluorescence detection of N-Cadherin (red, marking intercalated disc structures) and Cx43 (anti-N-terminus, green) in heart sections from the different AAV9 treatment groups after ischemia or control perfusion. Average intensity projections of confocal z-stacks (13.5 μm thickness), scale bar = 50 μm. (C) Quantification of Cx43 fluorescence intensity at intercalated disc regions. Data are presented as mean ±SEM relative to GST ischemia group (number of hearts analyzed per group is shown on bars, 5–10 images analyzed per heart), *p < 0.05, **< 0.01, ***< 0.001, ****<0.0001, by one-way ANOVA followed by Tukey’s post-hoc-test. (D) Quantification of Cx43/N-Cadherin signal at intercalated disc regions. Data are presented as mean ±SEM (number of hearts analyzed per group in shown on bars, 5–10 images analyzed per heart), ***p< 0.001, ****<0.0001, by one-way ANOVA followed by Tukey’s post-hoc-test.

DISCUSSION

Targeted delivery of ion channels to the appropriate subdomain requires three components: a cytoskeleton delivery apparatus, a membrane anchor, and the channel itself^{11, 38}. Previously, we identified that key components of targeted delivery of Cx43 hemichannels to the ventricular intercalated disc include microtubules which are captured by the membrane anchor adherens junctions complex including N-Cadherin, β-catenin and p150 (Glued)^{17, 18}. While these trafficking components serve as the highways to the destination of intercalated disc, it has not been clarified how the channel itself is involved in specificity of delivery. In this study, we find that GJA1-20k, an alternatively translated isoform of Cx43²¹, organizes the actin cytoskeleton (Figures 3A, 3B, 4A, 4B and Online Figure II) which in turn is required to orient microtubule trajectories to the cell-cell border (Figure 5) for Cx43 delivery. These findings indicate that the specificity of Cx43 in delivery is derived from one of its smaller alternatively translated isoforms.

LatA binds with 1:1 stoichiometry to the monomeric G-actin near its nucleotide binding cleft and shifts the dynamic equilibrium to F-actin depolymerization³⁹. It is worth noting that our treatment of cardiomyocytes with a low dose of LatA (250 nM) only partially inhibits actin polymerization, which might explain why actin fibers (Figure 4) and microtubule growth towards the cell-cell border (Figure 5) are not completely abolished in the LatA treated GST and GJA1-43K transduced cardiomyocytes. GJA1-20k interacts with actin (Figure 3D) and stabilizes polymerized actin fibers (Figures 3A, 3B, 4A, 4B and Online Figure II), shifting the equilibrium towards formation of F-actin opposing the effect of LatA. We previously identified the actin cytoskeleton as a necessary component of forward trafficking¹⁹. Because Cx43 is relatively stationary when associated with actin, our paradigm from the earlier work was that actin serves as rest stops along the microtubule highway at which Cx43 vesicles slow down and dwell. The rest stops collect channels in an ever-ready reservoir able to be rapidly mobilized to ensure continuous supply en route to the intercalated disc. Our findings in this study indicate that GJA1-20k serves as an upstream stabilizer of these actin rest stops which in turn act as active organizers of the microtubule highways by orienting microtubule trajectories toward the cell-cell border for cargo delivery.

Precedent exists, in plants and neurites, for actin laying the scaffolds by which microtubules are patterned^{40, 41}. Our finding that GJA1-20k interacts with both actin and tubulin (Figure 3D) and that it stabilizes the actin cytoskeleton (Figures 3 and 4) thus regulating the microtubule network (Figure 5), provides a powerful tool to manipulate Cx43 trafficking. Using an in vivo AAV9 delivery system, we find that GJA1-20k offers high potency of facilitating the Cx43 trafficking pathway, as exogenous GJA1-20k alone increases endogenous Cx43 gap junction localization at the intercalated discs even greater than over-expression of full-length GJA1-43k itself (Figure 1).

We have previously found that in mouse heart, either acute ischemia or LatA treatment resulted in disruption of the actin cytoskeleton and disruption of Cx43/β-actin complexing, limiting localization of Cx43 to the intercalated discs¹⁹. In this study, we extend these findings with evidence that, in vivo AAV9 mediated introduction of GJA1-20k maintains Cx43 gap junction coupling following acute ischemia (Figure 6). These data indicate that in the stressed myocardium, where microtubule and actin organization is perturbed ^{18, 19}, GJA1-20k can protect and stabilize the actin cytoskeleton and as a result maintain targeted delivery of Cx43. These findings indicate that GJA1-20k has therapeutic potential in situations of anticipated cardiac ischemia such as cardiac surgery or planned percutaneous interventions.

It should be noted that while the focus of this study has been on forward trafficking, ischemia will also induce changes in Cx43 phosphorylation of the C-terminus⁴², including dephosphorylation of the casein kinase serines⁴³ and phosphorylation of 14-3-3 binding motif including serine 373⁴⁴, both of which will accelerate internalization. It is possible that the protection of GJA1-20k against ischemia may occur not only in the acute conditions studied but is also involved in protecting the Cx43 C-terminus against pathologic post-translational modification. Future studies will examine the effects of GJA1-20k in the setting of subacute and chronic ischemia.

Cx43 has been implicated in the forward trafficking of other cardiac channels and junctional proteins to the cell-cell border, including Nav1.5 and N-Cadherin^45–48. It remains unclear where, between the Golgi and intercalated disc membrane, Cx43 facilitates their trafficking. It may be that rather than Cx43, GJA1-20k is a common facilitator of forward trafficking by stabilizing actin to organize microtubule growth trajectories for cargo delivery, or by mediating actin-to-microtubule package handoff to aid in the delivery of multiple ion channels and membrane proteins. Although Cx43 is not known to contain a direct actin-binding domain, it interacts with F-actin indirectly via binding to ZO-1 at the last four amino acid residues of the C-terminus⁴⁹. In an elegant study by Lubkemeier and colleagues, the last five amino acids were removed in a (Cx43D378stop) knock-in mouse⁴⁶. Cx43D378stop lead to reduction in sodium and potassium current densities as well as decreased Nav1.5 protein from intercalated discs. The impaired sodium and potassium trafficking in the Cx43D378stop mouse could be due to a truncated GJA1-20k isoform as discussed above. An additional F-actin interaction site downstream of the microtubule-binding domain in Cx43 has also been reported, which directly interacts with the actin binding protein drebrin (DRB)⁵⁰. This proximal F-actin interacting site is retained in GJA1-20k and may potentially aid in the regulation of the actin cytoskeleton by GJA1-20k but this still requires further study as does mechanism by which GJA1-20k regulates actin-to-microtubule transfer of Cx43 vesicles.

Together our results indicate that the internally translated GJA1-20k isoform contributes specificity to Cx43 delivery by stabilizing polymerized F-actin to organize the growth trajectories of microtubules, guiding full-length Cx43 protein traffic towards the cell-cell junctions. Given its ability to maintain Cx43 localization at intercalated discs in the setting of acute ischemia, GJA1-20k is a potential therapy to modulate the cytoskeletal trafficking machinery in the setting of anticipated ischemic cardiac injury.

NOVELTY AND SIGNIFICANCE.

What Is Known?

• Heart disease and heart failure cause a decrease of gap junction communication channels between heart cells, increasing the chance of sudden cardiac death.

• The intracellular cytoskeleton is the delivery highway for gap junction channels to arrive to their correct location between muscle cells, but specificity of delivery is still being determined.

• A small truncated isoform of Connexin43 (Cx43), called GJA1-20k, is created by ribosomal alternative translation and may help with delivery of full length channel to cell-cell borders.

What New Information Does This Article Contribute?

• GJA1-20k contributes specificity to full length protein delivery by promoting organization of the actin cytoskeleton, which directs the microtubule network to deliver cargo to the right location.

• Alternatively translated GJA1-20k is a critical link to Targeted Delivery of gap junction proteins, increasing the amount of native full length protein at cell-cell borders.

• GJA1-20k could be a novel therapeutic to limit the reduction of gap junction coupling and limit arrhythmogenesis in situations of anticipated ischemia of the heart.

Cx43 forms gap junctions between ventricular cardiomyocytes, which facilitate coordinated contraction of each heartbeat. Altered Cx43 trafficking in failing heart contributes to deadly arrhythmias. The Cx43 trafficking pathway has been identified to include microtubule highways and actin rest stops, yet it is unclear how these cytoskeleton components interact. In addition, the alternatively translated GJA1-20k can aid Cx43 forward trafficking, but how trafficking regulation is accomplished by this isoform is unknown. Here we report that GJA1-20k guides Cx43 full length channel trafficking by laying actin fibers en route to direct microtubule growth trajectories to cell-cell borders. GJA1-20k contributes specificity to Cx43 full-length channel trafficking by stabilizing filamentous actin which in turn is necessary to organize microtubule growth trajectories to direct Cx43 hemichannel delivery. This is the first report of an alternatively translated protein guiding full length proteins to their appropriate destination. Moreover, we demonstrate that cardiac expression of GJA1-20k can maintain cell-cell coupling in the setting of acute ischemia. This study not only offers a small isoform-based mechanism for Cx43 in channel trafficking, but also provides a new therapeutic which can be used to rescue gap junction coupling in failing heart.

Nonstandard Abbreviations and Acronyms

Cx43

Connexin 43

F-actin

Filamentous actin

KDa

Kilodalton

Lat A

Latrunculin A

AAV9

Adeno-associated virus serotype 9

mRNA

Messenger RNA

EB1

End binding protein 1