Generation and Characterization of an Inducible Cx43 Overexpression System in Mouse Embryonic Stem Cells

Connexins (Cx) are a large family of membrane proteins that can form intercellular connections, so-called gap junctions between adjacent cells. Cx43 is widely expressed in mammals and has a variety of different functions, such as the propagation of electrical conduction in the cardiac ventricle. Despite Cx43 knockout models, many questions regarding the biology of Cx43 in health and disease remain unanswered. Herein we report the establishment of a Cre-inducible Cx43 overexpression system in murine embryonic stem (ES) cells. This enables the investigation of the impact of Cx43 overexpression in somatic cells. We utilized a double reporter system to label Cx43-overexpressing cells via mCherry fluorescence and exogenous Cx43 via fusion with P2A peptide to visualize its distribution pattern. We proved the functionality of our systems in ES cells, HeLa cells, and 3T3-fibroblasts and demonstrated the formation of functional gap junctions based on dye diffusion and FRAP experiments. In addition, Cx43-overexpressing ES cells could be differentiated into viable cardiomyocytes, as shown by the formation of cross striation and spontaneous beating. Analysis revealed faster and more rhythmic beating of Cx43-overexpressing cell clusters. Thus, our Cx43 overexpression systems enable the investigation of Cx43 biology and function in cardiomyocytes and other somatic cells.


Introduction
Connexins (Cx) are a large family of transmembrane proteins that are primarily localized in the plasma membrane, but also in intracellular organelles, such as mitochondria [1,2]. In mice, 20 different isotypes of Cx have been identified [3]; they are widely expressed in all organs and tissues starting from early embryonic development [4]. All Cx share the same structural features consisting of an amino-and carboxy-terminus located in the cytosol, one intracellular loop and four transmembrane domains. Six Cx form a connexon, and two adjacent connexons in the plasma membrane of neighboring cells can form functional intercellular channels, the gap junctions [5]. Numerous gap junctions form clusters, called plaques [6]. Cx biology is complex, as different Cx isotypes can assemble and form heteromeric connexons and/or even heterotypic gap junctions [7]. More recent work proposes the existence of hemichannels, which are thought to open under pathophysiological conditions [8].
Gap junctions are known to have pleiotropic functions, such as the exchange of nutrients, signaling components, and ions below a MW of 1.5 kDa. Therefore, these proteins mediate electrical and metabolic cell-to-cell coupling. This explains their vital role and the early embryonic lethality upon knocking out specific Cx genes, e.g., Cx45 [9]. CAG promoter and a floxed stop cassette into the FseI sites of the Ai6 vector (Addgene, Watertown, MA, USA, #22798) [30], thereby exchanging the zsGreen reporter. The Ai6 vector is a Rosa26 (Rs26) targeting vector that allows homologous integration of expression and NeoR/KanR cassettes into the mouse Rs26 locus by co-transfection with specific zinc finger nucleases. The Ai6 vector contains the post-transcriptional regulatory element of woodchuck hepatitis virus (WPRE) to increase transgene expression [31]. All DNAs were prepared using the EndoFree Plasmid Maxi Kit from Qiagen (Venlo, The Netherlands). To verify correct cloning of the CAG-floxSTOP-Cx43-P2A-mCherry-Ai6 construct, restriction analyses and sequencing (Eurofins Genomics, Ebersberg, Germany) were performed. For molecular cloning procedures and sequence analyses, SnapGene Viewer 5.3.2 (from Insightful Science, available at snapgene.com, San Diego, CA, USA) and Clone Manager Professional 9 (Sci Ed Software LLC, Denver, CO, USA) were used.
All cells were cultivated at 37 • C and 5% CO 2 .
HeLa cells (60% confluency) were transfected with the PacI digested linearized CAG-floxSTOP-Cx43-P2A-mCherry-Ai6 construct using the TransIT-HeLaMONSTER Transfection Kit (Mirus Bio LLC, Madison, WI, USA) according to the manufacturer's instructions. Two days after transfection, transgenic HeLa cells were selected by adding 600 µg/mL G418 to the medium for twelve days.
Resistant ES cells and HeLa colonies were isolated and expanded in 24-well plates; in case of the ES cells, these were coated with irradiated neomycin-resistant fibroblasts. Colonies of resistant transgenic 3T3 cells were pooled, and a single cell dilution of 3T3fibroblasts was performed, as depicted in Figure 3a.
Neomycin-resistant colonies were analyzed for the presence of the expression cassette by PCR (see 2.9) and mCherry fluorescence after AAV2.1-Cre transduction.
For the generation of the Rs26-specific DNA probe (pRs26-5 , see Soriano et al. [36]), the plasmid pDonor MCS Rosa26 vector (Addgene, #37200) was digested with EcoRI, and the resulting 450 bp fragment was isolated. pRs26-5 probe sequence: For Southern blot analysis, 10 µg genomic DNA were digested with AflIII or double digested with EcoRV and HindIII and the addition of 2.5 mM of spermidine at 37 • C overnight. Enzymes were inactivated by incubation for 20 min at 60 • C and 80 • C, respectively, and digested DNAs were separated overnight (30 V) in 0.8% agarose gels containing 0.01% ethidium bromide. A Lambda HindIII digested DNA marker (Promega, Madison, WI, USA) was used to determine DNA fragment sizes. After 15 min in 0.25 N HCl, 5 min in distilled water, and two 15 min periods in 0.5 M NaOH, the DNA was transferred to a positively charged nylon membrane treated with distilled water and 0.5 M NaOH (BrightStar-Plus, Ambion, Thermo Fisher Scientific) via capillary blotting method (transfer buffer: 1M NaCl and 400 mM NaOH) for 3 h at room temperature. Following this, the membrane was neutralized for 15 min in 2× SSC buffer (0.3 M NaCl and 0.03 M sodium citrate tribasic dihydrate) and dried for 15 min at 80 • C (Heratherm incubator, Thermo Fisher Scientific) to immobilize the DNA. The membrane was then prehybridized for 3 h in a glass hybridization tube (Analytik Jena GmbH, Jena, Germany) placed in a hybridization oven (Analytik Jena GmbH) at 60 • C in an ULTRAhyb ultrasensitive hybridization buffer (Invitrogen, Thermo Fisher Scientific). DNA probes were labelled with α 32 P-dCTP (PerkinElmer 6000 Ci/mmol 20 mCi/mL Lead, 500 µCi, Waltham, MA, USA) using the Prime-It II random primer labeling kit (Agilent Technologies, Santa Clara, CA, USA). Labelled probes were purified with Bio-Spin P30 Tris Chromatography Columns (Bio-Rad) added to the prehybridized membrane, and hybridization was performed under stringent conditions (68 • C) overnight. Filters were washed four times (each 30 min, 55 • C) in 2× SSC containing 0.1% SDS and were exposed to hyperfilm MP8 films (GE Healthcare GmbH, Solingen, Germany) using intensifier screens (Amersham, GE Healthcare). The hyperfilms were exposed for 6-8 days before they were developed and fixed (Kodak GBX Fixer and Replenisher, Sigma-Aldrich, Merck; Kodak GBX Developer and Replenisher, Sigma-Aldrich, Merck; both 1:5 diluted with distilled water).

Selection of Suitable G4 ES and 3T3 Cell Clones Using PCR Analysis
Cells of the different G4 ES and 3T3 clones were lysed, and DNA was extracted using the Gentra Puregene Cell Kit (Qiagen). DNA was amplified applying 5× HOT To test the functionality of the Cre-loxP system RNA of Cre − as well as Cre + G4 ES cell clones 29 and 31 were isolated using the RNeasy Mini Kit (Qiagen), and cDNA was synthesized with the SuperScript VILO cDNA Synthesis Kit (Invitrogen, Thermo Fisher Scientific). CAG-driven expression of the transgene using a forward primer that binds to the first exon of the CAG promoter in combination with a reverse primer binding to Cx43 sequence demonstrated functionality of the Cre-loxP system: fw: CTGACTGACCGCGTTACTC, rev: AGCTGACTCAACCGCTGTCC.
Agarose gel electrophoreses were run to separate PCR products, which were visualized by ethidium bromide (Carl Roth) using an UV light imager Intas GelSTICK Touch (Intas Science Imaging Instruments GmbH, Göttingen, Germany) in combination with Intas GDS Touch 2 software 1.1.5.4 (Intas Science Imaging Instruments GmbH).

Copy Number Determination by qPCR
To analyze actual transgene copy numbers of G4 ES cells, the genomic DNA of three Cre − clones was isolated (EchoLUTION CellCulture DNA kit, BioEchoLife Sciences GmbH, Cologne, Germany) according to manufacturer's instructions. Then, qPCR was performed as triple determination with 200 ng of genomic DNA, TaqMan Gene Expression Master-Mix (Applied Biosystems, Thermo Fisher Scientific), a Tfrc-VIC housekeeper TaqMan probe (TaqMan Copy Number Reference Assay, mouse, Tfrc, Applied Biosystems, Thermo Fisher Scientific, #4458366), and a custom-made mCherry binding TaqMan probe (Thermo Fisher Scientific). A CFX 96 Real-Time System (Bio-Rad) and 96 well PCR plates (Bio-Rad) with seals (Bio-Rad) (PCR program: (I) 95 • C for 10 min, 35 cycles of (II-IV): (II) 95 • C for 15 s, (III) 60 • C for 30 s, (IV) 72 • C for 30 s) were used. CT values for the target (mCherry) and reference (Tfrc, two copy numbers) gene were computed with Bio-Rad CFX Manager 3.1 software (Bio-Rad). The ∆CTTR values (CT target − CT reference ) and copy numbers (2 −∆CTTR × 2) of the mCherry-containing transgene were calculated.

Fluorescence Recovery after Photobleaching (FRAP)
3T3-fibroblasts and G4 ES cells were incubated for 20 min at 37 • C with calcein AM (0.38 µM, Invitrogen, Thermo Fisher Scientific, #C3100MP) diluted in 10% DMEM medium Cells 2022, 11, 694 9 of 25 (3T3, see above) or 15% KnockOut-DMEM (G4 ES cells, see above). Fluorescence intensity of a cluster of calcein + 3T3-fibroblasts or G4 ES cells was measured using a confocal microscope (Nikon Eclipse Ti, 40× objective; CFI Apo Lambda S LWD 40XWI, NA 1.25, Nikon). Next, a single fibroblast or G4 ES cell in the cluster was bleached with a 561 nm laser pulse (5 s, 2.5 mW), fluorescence intensity was measured, and pictures were taken over 10 min at 15 s intervals. Fluorescence intensities were converted into values between 1 (fluorescence intensity before bleaching) and 0 (fluorescence intensity immediately after bleaching).

Video Microscopy of Cell Clusters of Dissociated Beating EBs
Cre + and Cre − beating EBs were harvested at d12/d13 of differentiation. After partial dissociation with trypsin for 20 min at 37 • C, cell clusters were seeded on 0.1% gelatin-(Merck) coated 24-well plates filled with 20% IMDM medium. One day later, frequency measurements were performed via video microscopy of beating clusters on a Nikon Eclipse Ti2 inverted microscope system with a 4× objective (Fluor, Nikon) using a brightfield setting. EB clusters plated in the 24 wells were heated by an Ibidi temperature controller (Ibidi GmbH, Gräfelfing, Germany), and CO 2 was maintained at 5% through an Ibidi gas mixer (Ibidi GmbH). Spontaneous contractions were recorded with a Grasshopper 3 Camera (GS3-U3-23S6C-C, Teledyne Flir LCC) analyzed online and offline using a customdesigned software (LabView, National Instruments, Austin, TX, USA), as described earlier in Makowka, Bruegmann et al. [39]. Time points of individual beats were recorded by a PowerLab system through the sound recorder of the PC (PowerLab 4/35 and Labchart 8 software, AD Instruments, Sydney, Australia) and used to calculate the frequency in offline analysis. For statistical analysis, the mean of beats per min (bpm) and its standard deviation (SD) of 5 min recording (sampling every second) was calculated. The mCherry expression of Cre + cell clusters was verified before the experiment with a Cy3 filter (F 46-004 ET, Nikon) using a prime BSI camera (Teledyne Photometrics, Tuscon, AZ, USA) controlled by the NIS-Elements AR 5.21.03 software (Nikon).

Statistical Analyses
Relative Cx43 expression data of Western blots were either compared using an unpaired two-tailed t-test (Cre + vs. Cre − ES cells) or an ordinary one-way ANOVA test followed, in case of significance, by a Tukey's multiple comparisons test (comparison of WT-, Cre − -, and Cre + 3T3-fibroblasts). Ordinary one-way ANOVA and Tukey's multiple comparisons tests (comparison of WT-, Cre − -, and Cre + 3T3-fibroblasts) or an unpaired two-tailed t-test (Cre + vs. Cre − ES cells) were determined to evaluate statistical differences of FRAP experiments. Beating frequencies and standard deviations of beating frequencies were compared using an unpaired two-tailed t-test. Cx isoform expression of RT qPCR analysis was compared using an unpaired two-tailed t-test (Cre + vs. Cre − EBs). All error bars are provided as SEMs. Statistical significance was considered a p-value of ≤0.05. All statistical analyses were calculated using GraphPad Prism 9.3.0 (GraphPad Software Inc., San Diego, CA, USA).

Generation and Characterization of Inducible Cx43 Overexpression in Mouse G4 ES Cell Lines
We have opted to generate an inducible Cx43 overexpression system in pluripotent mouse embryonic stem (ES) cells using a fluorescent reporter (mCherry), marking the overexpressing cells and a small detectable peptide (P2A), which is fused to the C-terminus of exogenous Cx43 (Figure 1b). For this purpose, we have generated mouse G4 ES cells, in which the CAG promoter drives the inducible overexpression of Cx43. The CAG promoter was chosen because of its strong expression in muscle cells [34,40,41] at all stages of mouse embryonic development and after birth. The inducibility of the expression cassette is provided by a loxP flanked stop cassette with three SV40 polyadenylation signals that can be removed by Cre protein activity ( Figure 1a). As depicted in Figure 1a, a DNA fragment composed of the murine Cx43 cDNA followed by the GSG-P2A DNA-sequence and the mCherry cDNA in frame was cloned downstream of the stop cassette of the Ai6 vector (CAG-floxSTOP-Cx43-P2A-mCherry-Ai6) (Figure 1a,b). This vector contains Rosa26 (Rs26) genomic sequences to drive homologous recombination of the expression cassette into the endogenous Rs26 locus of G4 ES cells, allowing targeted single integration (Supplementary Figure S1a) [42].   To obtain a single integration into the genomic Rs26 locus, G4 ES cells were cotransfected with this construct and specific zinc finger nucleases (Figure 1c and Supplementary Figure S1a). We have characterized the transfected cells and assessed successful single integration of the transgene into the Rs26 locus of individual cells with various approaches: First, ten (52 clones in total) independent transgenic G4 ES cell clones were isolated after ten days of selection with neomycin and further cultivated (Figure 1c). The mCherry DNA could be detected in nine out of ten analyzed ES cell clones (Supplementary Figure S1b). PCR analysis of the Rs26 locus integration site using primer pairs that bind to the endogenous Rs26 sequence at the 5 -or 3 -end and to sequences within the construct (Supplementary Figure S1e Figure 1d). Karyotyping of these six different cell clones revealed best results for clones 13, 29, and 31, as the large majority of counted cells (80% of clone 13, 75% of clone 29 and 79% of clone 31) contained 40 chromosomes (Figure 1e).
In order to investigate whether one or both Rs26 alleles were affected by integration of the insert, a Rs26-specific primer pair was used. PCR analysis resulted in a 297 bp band, indicating one unaffected Rs26 allele (Figure 1f). We also performed Southern blot analysis, proving the integration of the entire expression cassette (Figure 1g Figure S1f) and confirming one WT allele for each tested clone and one integration of the transgene. Then, we used qPCR to confirm single integration of the transgene in individual clones (Figure 1h). These experiments revealed that G4 ES cell clones 29 and 31 carry one copy of the transgene. We also found that clone 13 most likely carries a second copy (Figure 1h) that appeared integrated outside of the Rs26 locus, as one remaining WT Rs26 allele could be detected by Southern blot analysis (using the pRs26-5 probe) (Figure 1g; Supplementary Figure S1f), as well as by qualitative PCR (Figure 1f; Supplementary Figure S1e). We therefore used for the further experiments clones 29 and 31.
Next, we assessed the functionality of the Cre-loxP system by transducing the undifferentiated murine G4 ES cells of clones 29 and 31 with AAV2.1-Cre (Figure 2a). Cell suspensions of both transgenic clones were mixed with AAV2.1-Cre and plated on irradiated fibroblasts. After three days several, mCherry-expressing colonies were visible, demonstrating that the floxed stop cassette was excised by the Cre recombinase activity. Since not all cells were successfully transduced, we obtained a mixed cell population containing mCherry positive (Cre + ) and mCherry negative (Cre − ) colonies (Figure 2b, middle picture). To obtain homogenous Cre + transgenic cell lines, we isolated single Cre + subclones and expanded these in culture (Figure 2b, middle and right pictures). All these cells were found to be mCherry + , whereas the control cells (Cre − ), not AAV2.1-Cre transduced, were mCherry − (Figure 2b). The functionality of the Cre-loxP system was further corroborated by RT-PCR. Using a primer pair matching exon 1 of the chicken beta actin (CAG promoter) and the Cx43 cDNA, we could show that the stop cassette was removed and the chimeric intron was spliced out (band of 417 bp; Supplementary Figure S1d,e). Thus, both transgenic G4 ES clones C29 and C31 were functioning; for the following experiments, we used the (Cre − ) and (Cre + ) cells of clone 31.  We explored the Cx43, mCherry, and P2A expression patterns in ES cells using immunostainings ( Figure 2c) and Western blot analysis (Figure 2d and Supplementary Figure S7a). These experiments evidenced that P2A-tagged Cx43 and mCherry could be detected in the Cre + but not in the Cre − control ES cell lines (Figure 2c,d). Based on staining for P2A, we found that the cellular distribution pattern of the exogenous Cx43 protein was almost identical to the endogenous one: for both components, cytoplasmic and membrane location were observed (Figure 2c). Importantly, quantification of Cx43 expression in Western blot analysis proved an approximately 2.5-fold overexpression following Cre-induction upon normalization with GAPDH (Figure 2d; n = 6). In contrast, immunostainings and Western blot analysis in Cre − control cells showed lack of P2A and mCherry expression, underscoring appropriate function of the loxP stop cassette without major leakiness (Figure 2c,d). Taken together, we could generate an inducible Cx43 overexpressing G4 ES cell line carrying a single copy of the CAG-floxSTOP-Cx43-P2A-mCherry construct, which is integrated into the Rosa26 locus.

Formation of Functional Cx43 Gap Junction Channels after Expression of the Inducible Construct in HeLa Cells, 3T3-Fibroblasts, and Undifferentiated G4 ES Cells
Even though we could demonstrate inducible overexpression of Cx43 protein in pluripotent cells, this was no proof for functional gap junction formation. This is a critical issue, as tagged gap junction channels were reported to show altered cell biological properties [43,44]. Therefore, we established a transgenic HeLa cell line using the construct described above to assess gap junction functionality. We chose HeLa cells because they do not express endogenous gap junction channels. WT HeLa cells were transfected with the CAG-floxSTOP-Cx43-P2A-mCherry-Ai6 construct and subsequently selected with neomycin for twelve days. Single HeLa cell clones were picked, and transgenic HeLa cell lines (Cre − ) established by propagating the cells. Next, cells of a transgenic HeLa cell line (Cre − ) were transduced with the AAV2.1-Cre virus, and mCherry + HeLa colonies could be observed after three days. To establish Cx43 expressing Cre + HeLa cell lines, we performed single cell dilution (Figure 3a). The lines were further characterized by immunostainings and Western blot analysis for Cx43, P2A, and mCherry (Figure 3b,c). We found that Cre + HeLa cells co-expressed P2A-tagged Cx43 protein, which was localized in the cytoplasm and the cell membrane. We also observed that the Cre − transgenic control cells neither displayed mCherry fluorescence nor Cx43 or P2A expression (Figure 3b). These findings were underscored by the Western blot analysis (Figure 3c). Cells 2022, 10, x FOR PEER REVIEW 15 of 27  Given that we could generate inducible Cx43 expression in HeLa cells, we next explored functional Cx43 gap junction formation using fluorescent dye diffusion. To discriminate between gap junctions and cytosolic bridges, we used two different dyes: Alexa 350 with a MW of 349 Da and Alexa 647 dextran with a MW of 10 kDa, both at a concentration of 0.5 µg/µL. Single cells were dialyzed with the respective dye using a patch clamp pipette under microscopic control, and we found that Alexa 350 had diffused within 0.5 min to several Cre + HeLa cells, that were in close contact. Within approximately 5 min, the whole cell colony (10-20 cells) was found to be loaded with fluorescent dye (Figure 3d Besides HeLa cells, we also tested transfection and transduction of 3T3 mouse fibroblasts, which are known to endogenously express a low density of Cx43 gap junction channels. These experiments were performed to investigate inducible Cx43 gap junction channel formation in the presence of endogenous Cx43 gap junctions. 3T3 cells were stably transfected with the CAG-floxSTOP-Cx43-P2A-mCherry-Ai6 vector and selected with neomycin for twelve days. Through single cell dilution, a new transgenic 3T3 line was established (Cre − ), as proven by PCR analysis (Supplementary Figure S3a). Cells were transduced with AAV2.1-Cre, and three days after transduction, mCherry + 3T3 cells could be detected (Supplementary Figure S3b). Immunostainings of the Cre + and Cre − 3T3-fibroblasts proved, as would be expected, large amounts of Cx43 and P2A (Figure 4a, upper panel) in the mCherry + cells but no P2A and only low amounts of endogenous Cx43 in the controls (Figure 4a, lower panel). The mCherry and Cx43 overexpression in Cre + 3T3-fibroblasts compared to control cells was confirmed by Western blot analysis (n = 4) (Figure 4b). We also studied functional Cx43 gap junction formation using fluorescence recovery after photobleaching (FRAP), which enables, in contrast to dye diffusion, quantification of the diffusion velocity of the calcein AM dye. Cells were first loaded with calcein AM, followed by the bleaching of a single fibroblast in a cell cluster using a 561 nm laser for 5 s with an intensity of 2.5 mW. Thereafter, recovery of the fluorescence intensity in the bleached fibroblast was determined every 15 s over a period of 10 min (Figure 4c   We also tested the functional exogenous Cx43 gap junction formation in undifferentiated Cre + G4 ES cells using FRAP and compared the results with Cre − control ES cells. Our experiments revealed, at 4 min after bleaching, a significantly faster dye recovery in the Cre + G4 ES cells than in the controls (two-tailed unpaired t-test after 4 min: p(Cre + vs. Cre − ) = 0.0005; n = 5), whereas at later time points, the controls also showed a clear fluorescence recovery (two-tailed unpaired t-test after 10 min: p(Cre + vs. Cre − ) = 0.4905; n = 5) ( Figure 5, Supplementary Videos S5 and S7) due to prominent endogenous Cx43 expression (Figure 2c, lower pictures). Owing to its large size, mCherry fluorescence could not be recovered, as already mentioned above (Supplementary Figure S4a, Supplementary Video S6).

In Vitro Differentiation of Cx43-overexpressing ES Cells into Spontaneously Beating Cardiomyocytes
A key advantage of using pluripotent cells as a Cx43 overexpression system is that they can be differentiated in vitro, and the consequences of overexpression can be tested regarding development, differentiation, and function in different somatic cell types. As a proof of concept, we have focused on the in vitro differentiation into cardiomyocytes using the well-established hanging drop protocol (Figure 6a; adapted from Boheler, Czyz et al. [33]). Embryoid bodies (EBs) were generated from either Cre + or Cre − CAG-floxSTOP-Cx43-P2A-mCherry G4 mouse ES cells (Figure 6a,b). Western blot analysis and immunofluorescence stainings proved that P2A, mCherry, and Cx43 are overexpressed only in Cre + EBs (Figure 6c,d, middle panel, Supplementary Figure S7a). Furthermore, immunostaining against the muscle-specific marker cardiac α-actinin indicated a similar degree of cardiomyocyte differentiation, suggesting that cardiomyocyte development and differentiation was not greatly altered by Cx43 overexpression (Figure 6d, left panel). This was supported by the finding that both mCherry + (Cre + ) and mCherry − (Cre − ) EBs started to beat spontaneously at days 10-12 of differentiation (Supplementary Videos S8 and S9), underscoring the viability of the cardiomyocytes despite Cx43 overexpression. This was further investigated by performing immunostainings with the apoptosis marker cleaved caspase 3. Apoptotic areas were preferentially detected, as expected, in central areas of Cre + , as well as Cre − EBs, but there was no obvious difference in apoptosis rates in transgenic and control EBs (Figure 6d, right panel; n = 5). We next wondered whether Cx43 overexpression resulted in functional differences and therefore measured spontaneous beating rates in EB-derived cell clusters using microscopy-based video recordings. Our measurements revealed that the Cre + cell clusters displayed significantly higher beating rates (beating frequency = 75.9 bpm; n = 16) than did Cre − controls (beating frequency = 60.3 bpm; n = 19; unpaired t-test: p < 0.0001) (Figure 7a; Supplementary Videos S10 and S11). In addition, we also found that the Cre + cell clusters were beating more regularly than the Cre − controls, as their beating rates at 1 Hz over a 5 min period yielded a standard deviation (SD) of 1.9 vs. 5.3, respectively (unpaired t-test: p < 0.0001; n(Cre + ) = 16, n(Cre − ) = 19) (Figure 7a). We therefore wondered whether this could be due to changes in the expression of cardiac Cx isoforms. RT qPCR analysis of Cre + and Cre − undifferentiated ES (data not shown) and EB-derived cells revealed expression of Cx43 and Cx45, but not of Cx30.2 and Cx40 (Supplementary Figure S5a). Besides the expected upregulation of total (endogenous + exogenous) Cx43 in Cre + EBs (unpaired t-test: p = 0.0028; n = 4), there were no prominent changes in the other Cx isoforms (Supplementary Figure S5a). Thus, these experiments proved intact in vitro differentiation characteristics of Cre + cardiomyocytes despite Cx43 overexpression. dotted line and a yellow arrow); and dye recovery at 17 s, 37 s, and 57 s after bleaching; calcein AM dye (green). Scale bars 20 µm. ** p value ≤ 0.01, *** p value ≤ 0.001, **** p value ≤ 0.0001, ns (not significant) p value > 0.05.
We also tested the functional exogenous Cx43 gap junction formation in undifferentiated Cre + G4 ES cells using FRAP and compared the results with Cre − control ES cells. Our experiments revealed, at 4 min after bleaching, a significantly faster dye recovery in the Cre + G4 ES cells than in the controls (two-tailed unpaired t-test after 4 min: p(Cre + vs. Cre − ) = 0.0005; n = 5), whereas at later time points, the controls also showed a clear fluorescence recovery (two-tailed unpaired t-test after 10 min: p(Cre + vs. Cre − ) = 0.4905; n = 5) ( Figure 5, supplementary Videos S5 and S7) due to prominent endogenous Cx43 expression (Figure 2c, lower pictures). Owing to its large size, mCherry fluorescence could not be recovered, as already mentioned above (supplementary Figure  S4a, supplementary Videos S6).

Establishment of a Virus-Based Cx43 Overexpression System
Viruses are powerful tools for the targeting of cells and tissues. We wanted to have another tool for Cx43 overexpression in vitro and in vivo and have therefore generated an AAV2.6-Cx43-P2A-mCherry virus. This harbored the identical expression cassette, which was used above (Supplementary Figure S6a) under direct control by the CAG-promoter. We have chosen the AAV2.6 type, as this serotype is known to transduce not only muscle cells, but also other cell types, such as those of the central nervous system [45]. We tested the feasibility and utility of this virus in neonatal mouse cardiomyocytes (NNCMs), which are frequently used as an in vitro model because adult cardiomyocytes dedifferentiate rapidly in culture and therefore cannot be used for this type of experiment [46]. Enriched NNCMs (Figure 7b) were treated with the virus for two days and three days post-transduction; mCherry + NNCMs were observed (Figure 7c, upper picture). The transduction efficiency of the AAV2.6-Cx43-P2A-mCherry virus amounted to 47.7% in NNCMs (n = 3, 531 mCherry + NNCMs out of a total of 1113 NNCMs) (Figure 7c, graph). Cell clusters were observed to beat spontaneously before and after viral transduction, underscoring the viability of NNCMs (Supplementary Videos S12 and S13). Immunofluorescence stainings confirmed P2A-tagged Cx43 overexpression in mCherry + NNCMs (Figure 7d, upper panel). Cellular distribution of Cx43 was mainly restricted to the membrane in both control (Figure 7d, second and fourth panel) and transduced cardiomyocytes (Figure 7d, upper and third panel); to a small degree, cytosolic localization was also detected. Positive cardiac α-actinin staining of mCherry + cells corroborated the successful viral targeting of NNCMs (Figure 7d,  third panel). These experiments demonstrated the successful development of a virus-based transduction system that allows targeted overexpression of Cx43, as well as mCherry and P2A labelling.

Discussion
Herein we report the generation of a new genetic model that allows inducible overexpression of Cx43 in mouse ES cells. Cx43 is widely distributed in mammals, and we have chosen the ES cell system because in vitro differentiation of pluripotent cells allows us to assess the effects of Cx43 overexpression in different somatic cell types, such as cardiomyocytes and astrocytes. We have also established an AAV virus, which enables Cx43 overexpression of cells, tissues, and organs in vitro and in vivo.
We have opted for an inducible genetic system because CMV-driven ubiquitous Cx43 overexpression in mice was reported to cause morbidity and postnatal mortality due to neural tube and conotruncal heart defects [29], a phenotype which was reminiscent of Cx43 KO models [21,47]. We tested our genetic construct in ES cells, as well as HeLa cells and 3T3-fibroblasts, and observed no leakiness in the absence of Cre, whereas AAV-Cremediated excision of the floxed stop cassettes resulted in Cx43 overexpression in cells with and without endogenous Cx43 expression. A double reporter system was implemented to detect both Cx43 overexpressing cells based on mCherry fluorescence and the location and distribution pattern of exogenous Cx43 molecules based on their fusion to the small self-cleaving peptide P2A [48], which is detectable by antibodies (Figure 2c,d). P2A is a powerful tool, as it allows co-translational expression of P2A-linked genes in equal amounts in contrast to IRES sites [49]. Earlier work reported that the tagging of Cx43 is challenging, as fusion of proteins to its N-terminus resulted in the expression of nonfunctional hemichannels and gap junctions [43], fusion to its C-terminus in altered gating properties [50], and altered regulation of connexons and gap junctions [44,51]. We therefore chose to fuse the small P2A peptide to the C-terminal site of Cx43, and our data clearly show that this did neither alter the expression pattern nor the biology of the exogenous protein. This was underscored by dye diffusion in HeLa cells and FRAP experiments in 3T3-fibroblasts and G4 ES cells, proving the formation of functional Cx43 gap junction channels. As perturbation of the C terminus of Cx43 could lead to alterations of interactions with proteins involved in plaque dynamics [52], this aspect needs be explored in the future by performing specific biochemical assays [53,54]. Similarly, phosphorylation is known to be a key modulator of Cx43 gap junction function, and, therefore, the phosphorylation pattern of exogeneous Cx43 should be investigated using phospho-specific antibodies [55]. When analyzing the exogeneous Cx43 distribution pattern with confocal microscopy, we also found that it overlapped with endogenous Cx43 (Figure 2c, upper panel; Figure 4a, upper panel). Super-resolution microscopy and biochemical approaches are required to determine whether exogenous P2A + -and endogenous P2A − -Cx43 molecules assemble to form connexons and functional gap junctions.
Because of the reported adverse effects of Cx43 overexpression in vivo, we have explored this aspect in our system and differentiated mouse ES cells into spontaneously beating EBs in vitro upon Cre-induction. We could not observe prominent cell biological differences, as neither cardiac α-actinin immunostainings (Figure 6d, left panel), nor the degree of apoptosis (Figure 6d, right panel), nor the time point of initiation of spontaneous beating differed strongly between Cx43 overexpression and control EBs. Interestingly, when quantifying beating rates in Cx43 overexpression and control cell clusters, we noticed that mutant clusters were beating at faster rates and in a clearly more rhythmic fashion. At this point, we can only speculate on the mechanism and propose that the ubiquitous Cx43 overexpression provides better overall conduction between cardiomyocytes and also with non-cardiomyocytes within EBs, as we did not detect changes in the expression of other cardiac Cx isoforms. Recent experiments in human-induced pluripotent cellderived cardiac microtissues found that Cx43 gap junctions are critical for the maturation of cardiomyocytes in these constructs and that this depends on the diffusion of cAMP through Cx43 gap junction channels [56]. It is possible that also, in our case, the observed higher beating rates are related to more mature cardiomyocytes, which is difficult to assess based on purely morphological criteria. As expected for cultured cardiomyocytes, endogenous and exogenous Cx43 was found to be distributed in ES cell-derived cardiomyocytes and in NNCMs diffusely on the cell surface, whereas in the mature heart, its location is primarily restricted to intercalated discs.
Taken together, we have generated and characterized an inducible Cx43 overexpression system in murine ES cells and an AAV-Cx43 overexpression system enabling the investigation of both the cell biological and functional consequences of overexpressing Cx43 in vitro and in vivo. We envision for both systems multiple applications. Our group could show that grafting of Cx43-expressing muscle cells into the infarct area or overexpression of Cx43 in resident myofibroblasts of the scar strongly reduced the incidence of post-infarction ventricular tachycardias in mice in vivo [37,57]. Despite the great translational potential of these strategies for specific patient cohorts, potential cell biological (e.g., cell differentiation, maturation, viability) and functional (e.g., expression of ion channels and transporters) drawbacks of Cx43 (over)expression in cardiomyocytes and non-cardiomyocytes need to be investigated. The in vitro ES cell differentiation system is ideally suited for this purpose. Similarly, it has been suggested that some forms of epilepsy could benefit from the increased opening of Cx43 gap junction channels in astrocytes [58]. Therefore, novel Cx43 gap junction activators and openers could be studied in ES cell-derived astrocytes in regard to their effects on Cx43 gap junction channels. In addition, potential adverse effects of such Cx43 modulators could be investigated in terms of safety pharmacology in transgenic ES cell-derived cardiomyocytes. The transgenic G4 ES cells containing the CAG-floxSTOP-Cx43-P2A-mCherry construct can be used to generate transgenic mice. Such a mouse model would allow for the testing of the therapeutic potential of Cx43 overexpression in any cell type, depending on crossings with Cre-deleter mouse lines. For instance, for us and other groups working on cardiac arrhythmias, it would be of great interest to induce homogenous overexpression of Cx43 in border zone cardiomyocytes or myofibroblasts of the scar area and to study the effects on ventricular arrhythmias after myocardial infarction.