Overexpression of Wild-Type TMEM43 Improves Cardiac Function in Arrhythmogenic Right Ventricular Cardiomyopathy Type 5

The data that support the findings of this study are available from the corresponding author upon reasonable request.

WT male and female C57BL/6J mice were obtained from the animal facility of Centro Nacional de Investigaciones Cardiovasculares. TMEM43wt and TMEM43mut mice, in the C57BL/6J background, express human WT and human S358L-TMEM43, respectively, specifically in cardiomyocytes under the control of the α-myosin heavy chain promoter, as described previously.¹¹ Double transgenic mice were obtained by crossing TMEM43wt and TMEM43mut mice. WT littermates were used as controls. Male and female mice were used throughout the study. Genotyping was performed by Sanger sequencing after polymerase chain reaction (PCR) amplification of human TMEM43 with the following primers: human TMEM43 forward 5’-TCCCAAGTATCCAGAGGTGGGAGACT-3’ and human TMEM43 reverse 5’-AGGCCGGCAATGAGGAGGG-3’ (450 bp amplicon size). Thermocycler conditions were set as follows: initial denaturation at 95 °C for 5 minutes, followed by 37 cycles of denaturation at 98 °C for 30 seconds, annealing at 63 °C for 30 seconds, and extension at 72 °C for 40 seconds.

Mice were housed in an air-conditioned room with a 12-hour light/dark cycle and free access to water and chow. Mice were euthanized in a carbon dioxide chamber. After sacrifice, tibial length was measured, and hearts were weighed. Power calculations were conducted using G*Power software to determine the appropriate sample size required to detect a 10% improvement in ejection fraction, with an SD of ±6.5%, according to previous data.¹¹ A significance level of 0.05 and 80% power was used for the calculations. Based on these parameters, the required sample size per group was found to be at least 8 animals. Given that ARVC5 often presents differently in women and men, a minimum of 8 animals per sex and group were included in the studies. All procedures were approved by the Ethics Committees of the Centro Nacional de Investigaciones Cardiovasculares and the Regional Government of Madrid (PROEX191.5/22).

Transthoracic echocardiography was performed in mice aged 5, 10, 16, and 24 weeks. Mice were placed on a heating pad and kept under light anesthesia with isoflurane administered via a nose cone. The anesthesia was adjusted to obtain a target heart rate of 500±50 bpm. Cardiac function, chamber dilation, and wall thickness were analyzed from 2-dimensional and M-mode acquisitions. Image acquisitions were performed by a blinded operator using a high-frequency ultrasound system with a 30-MHz probe (Vevo 2100, Visualsonics). Left ventricular ejection fraction (LVEF) and LV end-diastolic volume measurements were obtained from the long axis view, and LV posterior wall thickness in diastole values were analyzed in the short axis view. Right ventricular systolic function was assessed indirectly from the tricuspid annular plane systolic excursion, estimated from maximum lateral tricuspid annulus movement obtained from a 2-dimensional 4-chamber apical view. Images were analyzed offline by an expert blinded to genotype and treatment using the Vevo 2100 Workstation software.

Mice were anesthetized with isoflurane via nose cone for echocardiography. Surface ECGs were obtained by using bipolar limb leads (leads I, II, and III) and unipolar limb leads (leads aVR, aVL, and aVF) for 90 seconds. Recordings were acquired and analyzed by a blinded operator using Acqknowledge 4.1.1 for MP36R (BIOPAC Systems Inc). Mean values of P-wave duration and amplitude, R-wave and S-wave amplitudes, and QRS duration and amplitude were calculated from 3 nonconsecutive beats from lead II. Mice that exhibited premature ventricular contractions in at least one 90-second ECG recording during follow-up were quantified. For experiments with flecainide, 90-second baseline ECG recordings were obtained before flecainide administration via intraperitoneal injection (40 mg/kg). ECG recordings continued for ≈5 minutes postadministration.

TMEM43 co-immunoprecipitation experiments were performed in the P19 cell line. P19 cells were transfected with both HA-tagged and GFP (green fluorescent protein)-tagged TMEM43 WT and S358L constructs using Lipofectamine 2000 (11668019; Thermo Fisher Scientific). Cells were lysed with tris-buffered saline buffer supplemented with 1% IGEPAL, 5 mmol/L EDTA, 5 mmol/L MgCl₂, and protease and phosphatase inhibitors (5892791001 and 04906845001, respectively, from Roche) for 30 minutes at room temperature. After centrifugation, the supernatants were precleared with Protein G Dynabeads (10003D; Thermo Fisher Scientific). Proteins were immunoprecipitated with Protein G Dynabeads coupled with a rabbit anti-GFP antibody (632592; Takara Bio), rotating at 4 °C overnight. Beads were washed 3× with 0.05% IGEPAL lysis buffer and 5× with lysis buffer without added detergent. Proteins were eluted from beads by boiling for 5 minutes at 95 °C in Laemmli sample buffer. Input, supernatant, and output samples from the co-immunoprecipitation were resolved by SDS-PAGE.

For TMEM43 codon-optimization experiments, HL1 cells were transfected with plasmids carrying a chicken cardiac troponin T promoter driving GFP, human codon-optimized TMEM43, or human WT TMEM43 expression using jetPRIME transfection reagent (POL101000046; Polyplus).

LV myocardium samples or transfected HL1 cells were homogenized by manual grinding in radioimmunoprecipitation assay buffer in the presence of protease and phosphatase inhibitors (5892791001 and 04906845001, respectively, from Roche). Lysates were separated on 10% SDS-PAGE gels (acrylamide:bisacrylamide 29:1), transferred to polyvinylidene fluoride membranes, and blocked with 3% nonfat dry milk in 1X tris-buffered saline for 1 hour. Membranes were incubated with primary antibodies overnight, followed by appropriate horseradish peroxidase-labeled secondary antibodies (anti-mouse P044701 and anti-rabbit P044801-2; Dako). Horseradish peroxidase activity was detected using Radiance ECL chemiluminescent horseradish peroxidase substrate (AC2204; Azure Biosystems). Primary antibodies were as follows: TMEM43 (ab184164; Abcam), GFP (11814460001; Roche), HA (11583816001; Roche), and vinculin (V4505; Sigma). Blot images were obtained and quantified with the Invitrogen iBright CL750 Imaging System and analysis software (Thermo Fisher Scientific).

LV samples were snap-frozen in liquid nitrogen right after sacrifice. Total RNA was isolated using TRIzol reagent (15596026; Thermo Fisher Scientific). First-strand cDNA was synthesized using 100 ng of total RNA and a High Capacity cDNA Reverse Transcription Kit (4368814; Thermo Fisher Scientific). Quantitative reverse-transcribed PCR was performed in a QuantStudio 5 real-time PCR thermocycler (Thermo Fisher Scientific) using SYBR Green (4367659; Thermo Fisher Scientific) for double-stranded DNA detection. Primer pairs used for quantitative reverse-transcribed PCR are LOX (lysyl oxidase) forward 5’-GCTGCGGAAGAAAACTGC-3’ and LOX reverse 5’-CCTTGGTTCTTCACTCTTTGC-3’ (104 bp amplicon), POSTN (periostin) forward 5’-AACGTCTGTGCCCTCCAG-3’ and POSTN reverse 5’-AGCCTTTCATCCCTTCCATT-3’ (151 bp amplicon), COL1A1 (collagen type I alpha 1 chain) forward 5’-GTGCCACTCTGACTGGAAGA-3’ and COL1A1 reverse 5’-CTGACCTGTCTCCATGTTGC-3’ (100 bp amplicon), COL3A1 (collagen type III alpha 1 chain) forward 5’-CACCCTTCTTCATCCCACTC-3’ and COL3A1 reverse 5’-ATGTCATCGCAAAGGACAGA-3’ (159 bp amplicon), ACTA1 (alpha actin 1) forward 5’-GACCACAGCTGAACGTGAGA-3’, ACTA1 reverse 5’-TGTTGTAGGTGGTCTCATGGAT-3’ (239 bp amplicon), and Luciferase forward 5’-TATCATGGCCTCGTGAAATCC-3’ and Luciferase reverse 5’-TCCTGGGTCCGATTCAATAAAC-3’ (131 bp amplicon). Results were analyzed with the LinReg PCR software,¹² which calculates the PCR efficiency of each sample independently. The average efficiency of each mRNA was then used to calculate the relative expression of each gene, which was normalized to that of β-actin.

Hearts were fixed in 4% paraformaldehyde in PBS for 48 hours, washed in PBS, dehydrated, included in paraffin, and sliced in 5 µm sections. The collagen content was analyzed using Masson trichrome protocol.¹³ For immunofluorescence, heart sections were subjected to sodium citrate antigen retrieval, permeabilized with 0.1% Tween-20/PBS for 20 minutes, and blocked with blocking solution (5% goat serum, 2% MgCl₂, 3% BSA in 0.3% Tween-20/PBS). Sections were incubated with anti-TMEM43 (ab184164; Abcam) diluted 1:100, followed by secondary antibody Alexa 568-labeled goat anti-rabbit (A-11011; Thermo Fisher Scientific) diluted 1:200, and wheat germ agglutinin (WGA) Alexa Fluor 488 conjugate (W11261, Thermo Fisher Scientific) diluted 1:2000. Nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole; D1306; Thermo Fisher Scientific). Samples were mounted in Vectashield mounting medium (H-1000; Vector Laboratories). Antibody-negative tissue sections were used as a control for the primary antibody, while a secondary-only control was included to account for nonspecific binding of the secondary antibody.

Immunofluorescence images were acquired with a Leica SP5 confocal microscope, an HCX PL APO CS 10× 0.4 dry objective, and Leica LAS-AF 2.7.3 software. Images for immunohistochemistry were digitized with NanoZoomer-2.0RS (Hamamatsu). All images were analyzed offline with Fiji, and brightness and contrast were linearly adjusted.

Brillouin spectra were acquired with a home-modified Olympus BX51 reflected light microscope with a diode-pumped solid-state (DPSS) laser at 532 nm (SpectraPhysics) and a light power below 3.5 mW on the sample coupled to a 3+3 TandemFabry-Perot spectrometer (J. Sandercock, Table Stable Ltd.).^14,15 An Olympus MPlan 10X objective with a numerical aperture of 0.25 was used to focus the incident light beam and to collect the scattered light simultaneously. Measurements were performed on dried samples to avoid having to account for water content variability. The tip of the heart apex was carefully sliced to allow the entry of agarose into the ventricular lumen, and samples were embedded in agarose (0.2 g/10 mL distilled water) and sliced into 500 µm sections using a vibratome (Campdem Instruments). A cover slip was placed on top of the sample without sealing. Transparent nail polish was applied on the sides to fix the cover slip. Samples were dried out at room temperature for 5 hours and stored at 4 °C overnight. Brillouin frequency shift, its error, and the half width half maximum were obtained from the Brillouin spectra by a nonlinear least squares fitting procedure using a Pseudo-Voigt function and a cubic polynomial background in MATLAB environment. Laboratory temperature monitoring, displacement of the Märzhäuser Wetzler automated table, and the GHOST multichannel analyzer were controlled with our own developed software using LABVIEW 4 platform.

Codon optimization of human TMEM43 was performed using VectorBuilder design studio. Self-complementary AAV vectors codon-optimized wild-type TMEM43 AAV vector (AAV-WT-OPT), AAV vector expressing S358L-TMEM43 (AAV-MUT), and AAV vector expressing luciferase (AAV-LUC) expressing optimized human WT-TMEM43, S358L-TMEM43, and luciferase, respectively, were constructed by VectorBuilder. Centro Nacional de Investigaciones Cardiovasculares’s Viral Vector Unit produced the recombinant AAVs with the AAVMYO myotropic capsid¹⁶ in-house using a modified Rep/Cap vector. Briefly, the cap gene from the AAV9 variant in the pAAV2/9n helper plasmid (gift from James M. Wilson, addgene number 112865) was replaced by digestion and ligation with a 3014-bp SwaI+SbfI fragment encoding the AAVMYO viral particle 1 (VP1) gene (obtained by synthesis from the public GenBank sequence deposited under accession: MN365014).

For viral administration, neonatal mice at postnatal day 1 (P1) were lightly anesthetized with ice, and 30 uL of the viral preparation, including 5e10 VP, were injected through the temporal vein in a single injection (1.7e13 VP/kg). Neonates recovered on a heating mat before placing them back together with the dams.

For experiments involving symptomatic adult mice, a subgroup of AAV-MUT mice was randomized into 2 subgroups based on LVEF and QRS amplitude at 10 weeks of age. Mice without signs of disease at this time were excluded from the study (n=7). These subgroups were then treated with 1.7e13 VP/kg of either AAV-LUC or AAV-WT-OPT via the femoral vein.

All statistical analyses were performed using GraphPad Prism (version 10.0.0; GraphPad Software). Data sets were first assessed for normality using the Shapiro-Wilk test. For data meeting the assumption of normality, statistical significance was evaluated using regular 1-way or 2-way ANOVA, followed by Bonferroni posttest for multiple comparisons. For non-normally distributed data, the Kruskal-Wallis test followed by Dunn multiple comparisons test, or multiple Mann-Whitney U tests with Bonferroni-Dunn posttest for multiple comparisons, were used as specified in the figure legends. Survival curves were analyzed using the log-rank (Mantel-Cox) test.

Statistical significance was defined as P<0.05, and exact P values were provided where applicable. Data are presented as mean±SD.

We have previously shown using in silico models that WT-TMEM43 and S358L-TMEM43 can form heterodimers. To confirm the models experimentally, we conducted IP experiments using tagged proteins to increase specificity. We found that GFP-tagged S358L-TMEM43 co-immunoprecipitated with HA-tagged S358L-TMEM43 as well as with HA-tagged WT-TMEM43 (Figure 1A). This result validated the in silico models and led us to hypothesize that the overexpression of WT-TMEM43 might quench the deleterious dominant effects of the mutant protein by forming a heterodimer.

To explore whether WT-TMEM43 overexpression could reduce the toxicity of S358L-TMEM43, we crossed our ARVC5 mouse model overexpressing human S358L-TMEM43 (TMEM43mut) with mice overexpressing human WT-TMEM43 (TMEM43wt mice) to generate a double transgenic mouse line that overexpresses both forms of the human protein in a cardiac-specific manner. The functional effects were evaluated through echocardiography and ECG over 24 weeks, and double transgenic mice were compared with TMEM43mut, TMEM43wt, and WT littermates.

The life span of double transgenic mice increased significantly by >10 weeks compared with mice overexpressing the mutant TMEM43, with a median life span of 34.6 versus 23.9 weeks, respectively (Figure 1B). Echocardiography analysis revealed improved LV contraction (LVEF) and reduced dilation in double transgenic mice (Figure 1C and 1D, respectively), while LV wall thickness remained unaltered in all of the groups (Figure 1E). Right ventricular contractility was also improved in double transgenic mice, without any significant deterioration observed until 24 weeks of age (Figure 1F). In addition, double transgenic mice exhibited reduced heart weight to tibial length ratio compared with TMEM43mut mice (Figure 1G).

Overexpression of WT-TMEM43 also improved ECG abnormalities caused by the mutant protein (Figure 2A through 2D). In contrast to the progressive P-wave prolongation observed in TMEM43mut mice, double transgenic mice exhibited normal P-wave duration (Figure 2B). Similarly, double transgenic mice showed reduced QRS duration and increased QRS amplitude compared with TMEM43mut mice at both 16 and 24 weeks of age (Figure 2C and 2D). Furthermore, the proportion of mice showing premature ventricular contractions, a common pathological manifestation in ARVC5 patients, was reduced to levels found in WT mice among double transgenic mice (≈12.5%), compared with TMEM43mut mice (≈35%) (Figure S1A and S1B). In line with these results, the mislocalization of connexin 43 observed in TMEM43mut mice was also partially prevented in double-transgenic mice (Figure S2).

As we previously demonstrated, ARVC5 mice develop extensive myocardial fibrosis as a result of cardiomyocyte death, evidenced by an increase in circulating cardiac troponin I levels.¹¹ As shown in Figure 3A, double transgenic mice showed a delayed rise in circulating troponin I compared with mutant mice, in agreement with the delayed onset of the disease. Moreover, double transgenic mice presented reduced expression of fibrosis markers at week 16, including LOX, which is responsible for collagen and elastin crosslinking, POSTN, and collagens 1 and 3 (COL1A1 and COL3A1), together with a decrease of the cardiac stress marker ACTA1 (Figure 3B through 3F). Masson trichrome staining confirmed a reduced fibrotic content in the heart of double transgenic mice compared with TMEM43mut mice at both 16 and 24 weeks (Figure 4A through 4E). Consistent with these results, TMEM43mut mice showed an increased Brillouin frequency shift, indicating increased stiffness of the myocardial tissue compared with double transgenic, WT, and TMEM43wt mice (Figure 4F). Together, these results demonstrate that WT-TMEM43 overexpression reduces cardiomyocyte death, fibrotic replacement, and myocardial stiffness, leading to improved cardiac function.

After the positive impact of WT-TMEM43 overexpression in our ARVC5 mice, we developed a gene therapy tool based on AAVs carrying WT-TMEM43 to further explore the therapeutic potential of this approach. To enhance vector efficiency, we first investigated the influence of cDNA codon optimization on the expression of human WT-TMEM43. This process includes substituting rare or nonpreferred codons in the original cDNA sequence with synonymous codons that are more frequently used by the host, likely improving mRNA stability and protein translation efficiency without altering the encoded protein’s amino acid sequence.¹⁷ HL1 cells were transfected with control (GFP), human TMEM43 (TMEM43WT), and codon-optimized human TMEM43 (TMEM43WT-OPT) constructs, and protein expression was assessed through western blot. Our results revealed a positive impact of codon optimization on TMEM43 protein expression (Figure S3A and S3B), prompting the utilization of this optimized sequence for the development of a self-complementary myotropic AAV. This vector expresses codon-optimized TMEM43 specifically in the cardiac tissue under the control of the cardiac troponin T promoter, hereafter referred to as AAV-WT-OPT (Figure S3C).

To determine the efficacy of our gene therapy approach and ensure a similar ratio of WT to mutant human TMEM43, we used systemic delivery of myotropic AAVs in neonates through the temporal vein and followed the animals for 24 weeks (Figure 5A). Each neonate received a total of 5e10 VP distributed as follows: 2.5e10 VP of AAV-WT-OPT+2.5e10 VP of AAV-LUC, 2.5e10 VP of AAV carrying S358L-TMEM43 (AAV-MUT)+2.5e10 VP of AAV-LUC, or 2.5e10 VP of AAV-WT-OPT+2.5e10 VP of AAV-MUT.

Western blot and PCR analyses revealed sustained TMEM43 overexpression in mice infected with AAV-WT-OPT or AAV-MUT (Figure 5B; Figure S4). Similarly, mice infected or co-infected with AAV-LUC expressed high levels of luciferase mRNA, as determined by quantitative reverse-transcribed PCR (Figure 5C). Immunofluorescence analysis of human TMEM43 showed overexpression of the protein in 10% to 30% of cardiomyocytes (Figure 5D through 5K), indicating wide biodistribution of the viral infection and cell type-specific protein expression.

Consistent with our TMEM43mut transgenic mice, mice transduced with AAV-MUT+AAV-LUC showed reduced life span (Figure 6A) and a progressive decline in LVEF as assessed by echocardiography (Figure 6B). This cardiac function impairment and life span decline were successfully prevented in mice co-infected with AAV-WT-OPT+AAV-MUT, providing additional support for the therapeutic potential of WT-TMEM43 overexpression. Right ventricular function showed similar trends, although the differences were milder and did not reach statistical significance (Figure 6C). AAV-MUT mice also developed ECG abnormalities, including increased P-wave duration, progressive widening of the QRS, and a gradual decline of the QRS amplitude, mirroring the pathological changes observed in our TMEM43mut transgenic mice. Remarkably, coadministration of AAV-WT-OPT proved sufficient to prevent all these ECG abnormalities (Figure 6D through 6G). In addition, the proportion of mice showing premature ventricular contractions during follow-up was reduced in AAV-WT-OPT+AAV-MUT coinfected mice compared with AAV-MUT mice (Figure S5).

AAV-MUT mice also showed a significant increase in circulating cardiac troponin I levels, albeit to lower levels than TMEM43mut transgenic mice (Figure 7A). This cardiomyocyte death was accompanied by mild but significant increases in the expression of genes related to fibrosis, including LOX, POSTN, and collagens 1 and 3, and the cardiac stress marker ACTA1 (Figure 7B through 7F). Importantly, coinfection with AAV-WT-OPT+AAV-MUT strongly reduced cardiomyocyte death and partially prevented the induction of fibrosis and stress genes (Figure 7A through 7F). Collectively, these findings support that AAV-WT-OPT gene therapy can prevent cardiomyocyte necrosis and the upregulation of fibrosis-related genes, improving cardiac contraction and reducing ECG abnormalities, in agreement with the results obtained in the double transgenic mice.

To date, we have demonstrated the potential of WT-TMEM43 overexpression as a preventive therapeutic strategy. To investigate whether late administration of the therapy could also ameliorate the ARVC5 phenotype, we administered AAV-WT-OPT to adult AAV-MUT mice after overt symptoms of the disease were already evident (10 weeks of age), including low LVEF, reduced QRS amplitude, and increased QRS duration. These mice were randomized into 2 groups with matched median LVEF and QRS amplitude. Mice from each group received the same dose of either AAV-LUC or AAV-WT-OPT used in neonates (1.7e13 VP/kg) via the femoral vein (Figure 8A). PCR analyses from 24-week-old mouse hearts exhibited sustained AAV-MUT and AAV-WT-OPT expression in both groups (Figure S6B).

Mice injected with AAV-WT-OPT showed a significant increase in LVEF compared with AAV-LUC-treated mice, lasting for up to 24 weeks (Figure 8B). Regarding QRS amplitude and duration, no significant improvement was observed (Figure 8C and 8D). During the 24-week follow-up period, none of the AAV-WT-OPT treated mice died, whereas 3 of 20 mice (15%) in the AAV-LUC group were found dead (Figure 8E), in line with our previous results at this age (Figure 6A). To uncover any subtle changes in electrical conduction, we administered a supratherapeutic dose of flecainide (40 mg/kg), a class I antiarrhythmic drug that can induce arrhythmias and produce AV block at high doses.¹⁸ As shown in Figure 8F and 8G, AAV-WT-OPT treatment reduced the incidence of AV blocks in response to flecainide compared with treatment with AAV-LUC. These findings further support the potential of WT-TMEM43 overexpression therapy in improving cardiac function and electrical manifestations in ARVC5, even when administered in adult symptomatic mice.

Although knock-in models are generally preferred for therapy development, heterozygous TMEM43-S358L knock-in models often show inconsistent results, with an exceedingly delayed onset or even an absence of cardiac phenotypes. Consequently, we used both transgenic and AAV-based ARVC5 models, which reliably reproduce the key functional and electrophysiological abnormalities of the disease. Notably, the AAV-based model exhibited a milder phenotype, which we attribute to lower mutant protein expression compared with the transgenic model.

We propose that overexpression of WT TMEM43 confers a therapeutic benefit by increasing the ratio of WT to mutant subunits within TMEM43 oligomers; however, it remains unclear whether this benefit stems from the oligomerization process itself or from competitive interactions between WT and mutant subunits.

Furthermore, while our mouse gene therapy models have provided promising insights, significant challenges remain in translating these findings to larger mammals and ultimately to human patients. Differences in tissue structure, immune responses, and cardiac electrophysiology may impact the efficiency of AAV-based gene delivery, making future studies in larger animal models essential for clinical application.