Plakophilin-2 (PKP2) is an abundant protein of the desmosome, an intercellular adhesion structure (Grossmann et al., 2004). Mutations in PKP2 associate with most cases of gene-positive arrhythmogenic right ventricular cardiomyopathy (ARVC), a pleiotropic disease that can manifest as mainly electrical, structural or both depending on stage progression (Ohno, 2016). As such, while ARVC is best recognized as a cardiomyopathy of right ventricular predominance, catecholaminergic sudden cardiac arrest is common during the subclinical (or “concealed”) phase of the disease (Cadrin-Tourigny, 2019; Müssigbrodt et al., 2019). The molecular/cellular mechanisms responsible for these arrhythmias remain unclear and current treatment options are still lacking. Recent studies have demonstrated that in addition to cell-cell adhesion, PKP2 and its intercalated disc (ID) partners translate information initiated at the site of cell-cell contact into intracellular signals that maintain structural and electrical homeostasis (Cerrone et al., 2017; Austin et al., 2019).

To study the role of cardiomyocyte PKP2 expression in cardiac function, we developed a cardiomyocyte-specific, tamoxifen (TAM)-activated, PKP2 knockout murine line [PKP2cKO; (Cerrone et al., 2017)]. Adult PKP2cKO mice present, in a compressed timeline, aspects of the history of human ARVC. Following TAM injection these hearts progress from a normal state, to an arrhythmogenic cardiomyopathy of right ventricular (RV) predominance and eventually biventricular dilated cardiomyopathy and end-stage failure (Cerrone et al., 2017). In Kim et al., PKP2cKO mice were studied 14 days post-TAM, a time point at which hearts present a phenotype of concealed ARVC (i.e., an arrhythmogenic phenotype but no overt structural disease) (Kim et al., 2019). At this stage, ventricular myocytes present increased Connexin43 (Cx43)-mediated membrane permeability. The latter is considered responsible for an excess entry of calcium (Ca²⁺) into the cells, eventually leading to the arrhythmogenic/cardiomyopathic phenotype.

Gap junction plaque formation is known to depend on proper intercellular adhesion (Beardslee et al., 2000). A recent study from our group showed that PKP2 deficiency can increase cell membrane permeability to ATP, an event prevented by silencing Cx43 expression (Cerrone et al., 2018). In addition, it is known that Cx43 hemichannels (Cx43-Hs) reside in the perimeter of the gap junction plaques (the perinexus) (Rhett et al., 2011) and are capable of flickering with very low probability (Contreras et al., 2003; Verma et al., 2009). We postulate that loss of PKP2 disrupts gap junction plaque integrity, thereby increasing the population of “orphan” Cx43-Hs in the perinexus. The latter would have important translational implications to the molecular mechanisms of inheritable ARVC. Despite functional evidence, visualization of these “orphan” (i.e., non-paired in a gap junction configuration) Cx43-Hs remains lacking.

With the development of confocal microscopy, live cell imaging and super resolution microscopy, the spatiotemporal characteristics of protein landscapes can now be unraveled in high detail. However, fluorescent microscopy has a restricted imaging range within 1 µm of the coverslip, a maximal resolution of ∼10–30 nm and is constrained to a small number of photo switchable fluorophores. Therefore, it lacks the ability to detect subcellular protein structures at sub-diffraction limit resolution. In that regard, is super resolution microscopy not the most optimal tool for detailed reconstruction of for example individual gap junction plaques, of which the intercellular space is between 2–4 nm. Immuno-electron microscopy (IEM) is an extremely powerful tool to localize, with nanometric resolution, a protein within its native structural landscape. Yet, challenges for IEM rely in preservation of antigenicity of molecular targets, accessibility for antibodies, while maintaining the cellular/tissue ultrastructure. Colloidal gold particles as post-embedding immunoprobes, combined with immunoglobulins or immunoglobuline-binding proteins (protein A), have been widely used for subcellular localization of specific molecules using EM (Faulk and Taylor, 1971; Roth et al., 1978). Previous studies show that smaller gold particles provide a higher labeling density, but gold particles with a diameter below 4 nm become hard to distinguish on ultrathin biological specimen under transmission EM (TEM). Therefore, gold particles with a diameter ranging from 5 to 15 nm are most convenient for on section immunolabeling (Slot and Geuze, 1981; van Bergen en Henegouwen et al., 1986; Yokota, 1988). Given the fact that colloidal gold particles have limited penetration capabilities into cells or tissues (even when using smaller size gold particles, such as 1–3 nm in diameter) and lack stability during the labeling process, they are not considered an optimal tool for pre-embedding and volumetric imaging (Hainfeld and Furuya, 1992). Attempts to prepare probes combining fluorescent labeling and colloidal gold for correlative microscopy resulted in limited success over the years, in part caused by restricted tissue penetration capabilities and hampered fluorescence intensity of these probes (De Brabander et al., 1985, De Brabander et al., 1986; Powell et al., 1997).

Nanogold particles provide an effective alternative. Nanogold is a neutral molecule with a diameter of 1.4 nm that is stable in a wide range of pH’s and ionic charges (Hainfeld and Furuya, 1992). Previous studies have shown that these 1.4 nm gold–Fab’ conjugates can penetrate tissue sections up to 40 µm of thickness, providing a perfect approach for pre-embedding immunolabeling (Sun et al., 1995). Fixation is important to maintain cell structure, but strong fixation and vigorous dehydration (as is routine for EM) can alter protein structure, thus impairing antigen-antibody binding. Since immunolabeling is performed on weakly fixed cells or tissues prior to dehydration, nanogold pre-embedding has a maximum antibody-antigen binding efficiency, comparable to light microscopy labeling. By combining fluorescent probes with covalent bound 1.4 nm clusters of gold atoms, FluoroNanogold allows the visualization of any given antigen using correlative microscopy and/or multimodal microscopy (Cheutin et al., 2007; Giepmans, 2008). Therefore, pre-embedding using nanogold or FluoroNanogold is considered one of the optimal tools for volumetric electron microscopy (volume EM, vEM).

In this study our objective was to visualize and examine, in three-dimensions (3D), the gap junction plaque morphology present in the adult murine heart. We used 200 µm thick sections of adult ventricular free walls of control and PKP2cKO cardiac muscle, pre-embedded with nanogold or FluoroNanogold labeling of Cx43. We measured the characteristics of the intercellular space and of the perinexus using a 3D-vEM approach. Our data show that loss of PKP2 expression led to widening of the intercellular space in the gap junction plaque, and increased abundance of unpaired (orphan) Cx43 channels in the gap junction. We conclude that pre-embedding and serial block-face scanning electron microscopy allowed visualization of cardiac Cx43 plaques in their native environment, providing for the first time a visual complement of functional data indicating the presence of orphan Cx43 hemichannels resulting from loss of desmosomal integrity in the heart.

Continuous variables are presented as box and whisker plots, reaching from the minimum to maximum value. Categorical variables are presented as percentage or total length. All data sets were tested for normal distribution by the Shapiro-Wilk and Kolmogorov-Smirnov tests, data were considered normally distributed when passing one of these two tests. Normally distributed data were compared using the Student’s t-test. Non-normally distributed data were compared using the Mann–Whitney Test. Statistical significance was assumed when p < 0.05. Statistical tests were performed using GraphPad Prism.

Over the past years, several investigators have studied loss of desmosomal integrity in hearts of ARVC patients by desmosomal protein mutations (Basso, 2006; Gomes et al., 2012; Noorman et al., 2013). Functional data has been largely supported by immunohistochemistry labeling of desmosomal proteins, though the complete structure of the PKP2-deficient ID in three dimensions remains unresolved. As a first step, here we describe the complete spatial ultrastructure of the gap junction plaques in control, and in PKP2-deficient myocytes prior to the development of a structural disease.

Studies in cell culture systems and in tissue from ARVC-affected patients have previously indicated that loss of expression or mutations in PKP2 can alter the structure and function of gap junctions (Oxford et al., 2007; Saffitz, 2009; Asimaki and Saffitz, 2014). Previous studies showed that PKP2 deficiency in the murine heart increases Cx43-mediated membrane permeability, and that the latter may facilitate ATP efflux (Cerrone et al., 2018) and excessive entry of Ca²⁺ into the cells (Kim et al., 2019). It is also known that Cx43 hemichannels reside in the perimeter of the gap junctions (Rhett et al., 2011) and are capable to flicker with very low probability when undocked (Contreras et al., 2003; Verma et al., 2009). Consistent with those observations, ablation of Cx43 prevented the accumulation of Ca²⁺ _i, and GAP19, a peptide known to block Cx43-Hs without affecting junctional conductance, normalized the intracellular Ca²⁺ homeostasis (Kim et al., 2019). These data support the hypothesis that Cx43-Hs are likely important players in the arrhythmogenic/cardiomyopathic phenotype of these mice. Here, we provide structural data in support of the notion that Cx43 hemichannels are present with a higher abundance in the membrane of PKP2-deficient hearts. We thus postulate that these single strings of Cx43 immunoreactive molecules are the structural counterpart to the excess Cx43 hemichannels detected by functional assays. Of note, previous studies have shown an absence of changes in Cx43 abundance (Cerrone et al., 2017; Kim et al., 2019). Given the timing of our experiments vis a vis the development of a PKP2-dependent cardiac phenotype in these mice, we conclude that the nanostructural changes in the ID described here are among the earliest events following loss of PKP2 expression.

We show that the multimodal imaging application offered by nanogold or FluoroNanogold labeling can be applied for vEM in thick cardiac sections. Highlighting it as novel tool for future use in 3D imaging of protein localization in different tissue types. FluoroNanogold is a probe containing two different markers (Takizawa and Robinson, 2000). As such, it opens the possibility of imaging the same sample at both the optical and the EM level (Robinson and Vandré, 1997; Takizawa and Robinson, 2000). The method can be used for multimodal studies, either on different cell types or on the same cell, visualized by correlative microscopy. The very small size of the gold atom clusters allows efficient absorption within fixed tissues prior to embedding (Robinson and Vandré, 1997). Quantum dots are largely used to identify and localize proteins simultaneously with different imaging approaches; however, they can only penetrate up to 4 μm deep into the tissue (Giepmans et al., 2005). Our experiments show that nanogold could easily penetrate in up to 20 μm on both sides of the cardiac vibratome sections. The use of nanogold and FluoroNanogold labeling in adult heart therefore represents a step forward in the development of novel methodologies for visualization of specific protein aggregates in native tissue, at nanometric resolution and in three dimensions.

Making use of FIB-SEM, we previously identified a nanometric separation of the ID membrane in RV samples of PKP2cKO mice (Kim et al., 2019). In hearts of heterozygous PKP2 knock out mice, we implemented high-pressure freezing tissue preservation for tomographic electron microscopy (tomographic EM) purposes (Leo-Macías et al., 2015). Besides clear detection of ID membrane separation, ultrastructural characteristics of the intercellular space could be visualized as well. Tomographic EM, however, did not allow detailed identification of individual gap junction plaques. Labeling Cx43, prior to imaging, is therefore critical to detect orphan hemichannels and asymmetric gap junctions in the ID membrane. In this study, IEM allowed us to examine both parameters (separation of the ID membrane and presence of Cx43-Hs) in parallel, adding an extra visual component to the previously obtained microscopy and functional data reported in Kim et al.

We emphasize that our experimental model (PKP2cKO) neither is intended to “recapitulate” ARVC nor is intended to be a surrogate for studies of human hearts affected with disease. This is in fact true for all animal or cellular models of ARVC that have been published, and an obvious consequence of the limitations inherent to the study of a human disease. The significance of our studies is not in having reproduced conditions that lead to human ARVC, but in having studied a molecule that, when mutated, causes the disease. Whether the structural changes described here occur in human hearts with PKP2 mutations remains undefined. Extrapolation from our data to the specific case of ARVC needs to be done with caution. That said, our model has shown consistency with human informatics data (Montnach et al., 2018) and has been useful in proposing therapeutic approaches to patients with ARVC (e.g., the pilot clinical trial on use of flecainide in ARVC patients; NCT03685149).

In conclusion, the results described in this study support the notion that, at this stage in the progress of the phenotype (i.e., before histologically-detectable disease), we detect a nanometric separation of the gap junction plaque, and the presence of unopposed Cx43-immunoreactive single strings (orphan Cx43-Hs). We propose that these orphan hemichannels are the structural correlate to the functional Cx43-Hs previously detected by (Kim et al., 2019). Our data illustrate the value and the potential of the novel methodology described here for visualization. It also provides further support to the notion that the gain-of-function of Cx43-Hs constitutes an important arrhythmogenic substrate in the setting of PKP2 deficiency. We speculate that block of Cx43-Hs may represent a novel and effective therapeutic strategy to prevent life-threatening arrhythmias in patients with PKP2 deficiency, particularly in the concealed stage of the disease.