Connexin Gene Transfer Preserves Conduction Velocity and Prevents Atrial Fibrillation

Study protocol

Thirty Yorkshire swine (20–30 kg) were randomized into 2 groups (sinus rhythm (SR) and AF groups), and within each group into 3 subgroups: sham-operated control, gene therapy with adenovirus expressing connexin (Cx) 40 and Cx43 (n=5 per subgroup). During an initial procedure, animals underwent invasive electrophysiology study (EPS), gene painting with a solution containing 2g/L poloxamer-F127, 0.05g/L trypsin, and 1 × 10⁹pfu/ml of the indicated virus; and for AF animals, implantation of an atrial pacemaker. Burst atrial pacing was activated immediately after the procedure. On a daily basis, animals had clinical status assessment and 2-minute telemetry recording. On post-gene transfer day 7, animals underwent EPS and cardiac extraction for optical mapping, histology and molecular studies. All of these procedures were performed using standard methods, as previously reported.^8,12–14 Cardiac rhythm was interpreted by two clinical cardiac electrophysiologists (TI, JKD). For subjective analyses (clinical assessment, rhythm analysis, histology and immunohistochemical studies), the investigators were blinded to specimen identity. For increased consistency, all histology and immunohistochemical analyses were performed in a single session.

Statistical Analysis

All data are presented as mean±SD. For rhythm analysis, probability of SR was assessed by Poisson regression using generalized estimating equations in order to address the animal-level clustering. In optical mapping, histology, immunohistochemistry and Western blot analyses, statistical differences were determined by related-measures ANOVA (post-hoc Tukey correction), which assumed that the outcome was correlated over time for each animal. All statistical tests were conducted at the 0.05 significance level. Statistical analyses were performed using SAS9.2 GLM and GENMOD procedures.

Clinical observations and rhythm analysis

We assessed heart rhythm with daily 2-minute telemetry recordings. All animals in the SR group started in SR, remained in SR, and never had any observed rhythm other than SR on the daily recordings. In the AF group, we continued burst pacing with the 2-second on/off duty cycle through the telemetry recordings, and calculated %SR as the percent age of off-pacing segments showing SR per animal per day (Figure 1). In the AF group, we observed no rhythm other than AF or SR in any animal. In AF-controls, SR progressively declined over time. All control animals were in continuous AF 5.8±0.6 days after onset of burst pacing. The mean time to continuous AF for AF-controls in this study did not differ significantly from the time course we previously reported for control animals with this model.^8,14,15

We analyzed probability of SR between AF animals using a generalized estimating equations model. We assumed that the probability of SR followed a Poisson distribution and working correlation matrix was first-order autoregressive. The results showed a significantly increased probability of SR for AF-AdCx40 (relative risk; 2.16, 95% confidence interval; 1.07 to 4.36, p<0.05) and AF-AdCx43 animals (relative risk; 3.69, 95% confidence interval; 1.97 to 6.93, p<0.01) compared to AF-controls. There was no statistically significant difference in SR% between AF-AdCx43 and AF-AdCx40.

We evaluated animal clinical status on a daily basis, and assessed echocardiograms at baseline and sacrifice (online supplement tables 1–2). Other than increased weight in the SR-controls and SR-Cx43 animals relative to the other subgroups, there were no baseline differences between subgroups. The baseline weight difference in these SR subgroups had no apparent impact on the study as it did not correlate with measures of cardiac mechanical or electrical function. At sacrifice, echocardiography showed significantly increased left atrial size and decreased left atrial and left ventricular function in AF-controls relative to SR-controls. These structural alterations did not manifest in clinically apparent heart failure. Within the SR group, echo measurements did not differ between subgroups. In the AF group, left atrial size and function were improved in the Cx-treated animals relative to controls, consistent with the observed reduction in AF for those animals.

Effects on Atrial Conduction

We determined the effects of Cx gene transfer on atrial conduction using three methods: P-wave duration on 12-lead ECG, in vivo pacing at the SN and measurement of activation time at the tip of the LAA, and ex vivo optical mapping of an anterior left atrial tissue wedge. In the SR group, the P-wave duration did not differ between subgroups (Figure 2A). The P-wave duration was significantly longer in AF-controls compared to any other group. The P-wave duration in the AF-AdCx40 and AF-AdCx43 animals was significantly shorter than AF-controls, and not different than SR animals.

We observed similar behavior for both the in vivo atrial conduction time assessment (Figure 2B) and the ex vivo optical mapping (Figure 2C–D). In the SR group, intra-atrial conduction did not differ between subgroups. Intra-atrial conduction was significantly slower in AF-controls compared to all other groups, and conduction did not differ between AF-connexin subgroups and SR-controls. Interestingly, the atrial conduction data for AF control animals in this report are similar to those we previously reported in Amit et al., suggesting reproducibility across different animal groups over time with this model.⁸

Histological Evaluation

To assess for possible confounding of the conduction results from differences in structural remodeling between subgroups, we analyzed atrial histology (Figure 3). We examined previously reported histological parameters associated with AF: nuclear and cellular hypertrophy, interstitial inflammation, cellular myolysis and fibrosis. Nuclear enlargement, inflammation and myolysis were greater in AF-controls compared to SR-controls. Cellular hypertrophy and fibrosis did not differ between control groups. Within the AF and SR groups, histological measures did not differ significantly between subgroups. These data suggest that our observed differences in CV and AF prevention were a direct result of Cx gene transfer and not due to differences in remodeling.

Connexin Expression

We assessed Cx expression using Western blot analysis. We used an anti-Cx43 antibody that recognized total Cx43, and we distinguished the differentially phosphorylated variants by their molecular weights. Similar to other reports, we observed only a single Cx40 band, so we could not assess Cx40 phosphorylation.¹⁶

In the SR group, the expression of total Cx40 was greater in SR-AdCx40 animals (Figure 4A–B) and the expression of both total and phosphorylated Cx43 were greater in SR-AdCx43 animals (Figure 4C–D) compared to SR controls.

Comparing AF to SR controls, expression of total Cx40 was not significantly different (Figure 4A–B), but expression levels of total and phosphorylated Cx43 were markedly diminished in the AF-controls (Figure 4C–D). We saw significantly increased Cx43 expression in the AF-AdCx43 animals, and significantly increased Cx40 in the AF-AdCx40 animals relative to AF-controls. The levels of both total and phosphorylated Cx43 expression in the AF-AdCx43 animals were similar to the levels found in SR-controls.

Connexin Localization

We evaluated the percentage of expressed connexin located at the intercalated disk using immunohistochemistry and confocal imaging (Figure 5A). In the SR group, the relative percentage of connexin at the intercalated disk did not change with Cx43 overexpression. There was a slight but significant decrease in the relative percentage of intercalated disk Cx40 for the SR-Cx40 group (Figure 5B–C). In both the AF-control and AF-Cx40 subgroups, we saw an increase in the relative percentage of lateralized Cx43 compared to SR animals. AdCx43 gene therapy restored intercalated disk localization of Cx43 to the same level observed in SR group animals. There were no differences in Cx40 lateralization between AF subgroups.

Cross-regulation of Cx40 and Cx43

We used Western blot analysis to examine for changes in Cx40 with AdCx43 exposure or changes in Cx43 with AdCx40 exposure. We confirmed that the antibodies were sensitive and specific to the indicated Cx proteins (online supplement figure). In the SR group, animals receiving AdCx40 had no significant difference in total or phosphorylated Cx43 expression compared to SR-controls (Figure 6A–B). However, the SR-AdCx43 animals had significantly increased Cx40 expression compared to SR-controls (Figure 6C–D). In the AF group, we saw no difference in Cx43 for the AF-AdCx40 animals or in Cx40 for the AF-AdCx43 animals relative to AF-controls. Thus, AdCx40 had no effect on Cx43 levels in either SR or AF. AdCx43 increased Cx40 expression during SR, but this effect was lost with atrial burst pacing.

Toxicity evaluation

The possibility of proarrhythmia is a principle concern when manipulating cardiac electrophysiology. We evaluated for atrial or ventricular proarrhythmia or signs of other toxicity during the in-life phase by assessing daily clinical measures and telemetry for atrial or ventricular arrhythmias, and at termination study by assessing 12-lead ECG parameters, echocardiography, tachycardia inducibility, MAPD₉₀, and by looking for early afterdepolarizations (online supplement table 2). We saw no evidence for toxicity in any animal.

AF/CHF model

To put the current findings into context, it is important to briefly review our previous observations with this model, and to compare the porcine AF/CHF model to the available data from humans and from other experimental models. The dominant features of our model are progression to sustained AF with continued burst pacing over several days and aggressive development of heart failure from the uncontrolled ventricular rate. In prior study, we found four-chamber dilation and LVEF of 28% 3 weeks after initiation of burst pacing, and LVEF 22% at 5 weeks.^14,15 In the current study, we found evidence of left atrial dilation, and slightly reduced LVEF (53%) without change in LV diameter after 1 week of burst pacing. Evaluation of atrial histology in previous study showed severe myolysis, inflammation, cellular and nuclear hypertrophy and only modest fibrosis at the 3- and 5-week timepoints.^14,15 In our current data, we saw the beginnings of myolytic, inflammatory, and hypertrophic responses after 1 week, but we sawno difference in fibrosis between AF and SR controls. EP study from a previous cohort revealed conduction velocity slowing (P-wave duration baseline 54 ms, 1 week 90 ms, 3 weeks 100 ms; SN to LAA conduction time baseline 72 ms increasing to 90 ms at 1 week) in a similar range as that reported in our current study.⁸

Histological and conduction data for human AF are limited. The majority of available data come from measurements on tissue samples taken during cardiac surgery, and comparisons are generally made between SR and AF patients with valve disease or coronary atherosclerosis. Data on patients with any significant degree of LV dysfunction are extremely limited. General trends across the human studies include evidence of atrial fibrosis, inflammation, myolysis, and reduced intra-atrial conduction velocity.^17,18

Animal models potentially supplement the limited human information. It is important to keep in mind when reviewing the animal data, however, that each model reflects a different facet of the human disease, but none of these models (including ours) provides a complete picture of the diverse set of conditions that make up humanAF.

The goat atrial burst pacing model is remarkable for sustained AF in the absence of structural ventricular disease, an important distinction between the goat and pig models. The goat model likely emulates lone AF in humans. In a series of investigations, Allessie and colleagues reported extensive atrial myolysis and inflammation but no atrial fibrosis or CV slowing even after months of AF.^19–21

Several canine models have been reported, each with distinct features. A sterile pericarditis model, potentially analogous to post-operative AF, has CV slowing and inflammation.²² An atrial tachypacing model simulates sustained atrial tachycardia, with increased AF inducibility but without sustained AF. In one study, CV decreased approximately 25% over several weeks, but half of that change occurred within the first 24 hours of atrial tachypacing.²³ Another study reported a 50% increase in P-wave duration, consistent with significant CV slowing, after 2–6 weeks of burst pacing.²⁴ A ventricular tachypacing model with increased AF inducibility but without sustained AF provides the only data for atrial remodeling in the setting of impaired LV function. Key features of that model included atrial fibrosis and either heterogeneous or reduced atrial CV.^16,24,25

Overall, then, the relevant features of our model include rapid progression to persistent AF over several days, steady progression of severe atrial myocyte pathology and ventricular failure over a few weeks, and reduction in atrial conduction velocity prominently in the first week with continued decline over at least a 3-week period. Similarities between our model and the available human data include decreased atrial conduction velocity and prominent myocyte pathology. Important differences include limited fibrosis and aggressive onset of heart failure in our model. Relative to the other animal models, direct comparisons are difficult because no other model has the combined features of persistent AF and heart failure. The goat model has persistent AF, but without heart failure. The canine ventricular tachypacing model has prominent ventricular failure, but without sustained AF. As such, our model likely has some features driven by interaction between persistent AF and heart failure that are not captured by other animal models.

Connexins and AF

Loss-of-function mutations in both Cx40 and Cx43 have been associated with lone AF, suggesting a role for cellular coupling in vulnerability of structurally normal hearts to AF.^26,27 Connexin data are inconsistent across studies of diseased human hearts. Kostin et al. found decreased and lateralized Cx43 expression and heterogeneous Cx40 expression (increased in the right atrial free wall, decreased in the right atrial appendage) without overall change in Cx40 levels.¹⁷ Rucker-Martin et al. found reduced expression of phosphorylated Cx43 and total CX40, with lateralization of Cx43.²⁸ Nao et al. and Gaborit et al. saw no significant change in Cx43 levels and decreased Cx40, and Wetzel et al. found increased levels of both connexins.^29–31 The heterogeneous expression observed by Kostin suggests that biopsy site may account for the differing Cx40 expression levels observed in the other reports. No explanation is available for the variability in Cx43 data. Overall, in spite of the variability in these reports, consistent findings included lateralization and dephosphorylation of Cx43 (where assessed) suggesting that the level of functional Cx43 is diminished in human AF.

In the goat atrial burst pacing model, Van der Velden et al. saw no change in overall expression levels of Cx40 or Cx43, but they noted Cx40 expression heterogeneity similar to Kostin, dephosphorylation of Cx43, and lateralization of both connexins.^20,21 In the canine sterile pericarditis model, Cx43 expression was diminished throughout the myocardium, and essentially undetectable in the epicardial layer.²² In the canine atrial tachypacing model, Cx43 mRNA expression was reduced approximately 67% in one study, but another study found a 2-fold increase in Cx43 protein level with a comment that Cx43 was more prominent at the lateral cell borders.^32,33 In the canine ventricular tachypacing model, both dephosphorylation and lateralization of Cx43 were uniformly observed.^16,24,25 Cx43 levels were variably reported to be either unchanged or increased, and Cx40 levels were unchanged compared to SR controls.^16,33 Bikou et al. reported a decrease in Cx43 expression in a model of porcine AF similar to ours.³⁴ Thus, the animal data are similar to the human studies in that Cx expression levels varied by study but Cx43 lateralization and dephosphorylation were consistent findings across multiple reports in different animal models.

Reports of peptides that modulate connexin activity have shown inconsistent effects in canine models of acutely induced AF. Shiroshita-Takeshita et al. found that rotigaptide had no effect on normal canine atria, slight but statistically significant improvement in CV for atrial or ventricular tachypacing animals, and considerable improvement in CV when administered to dog atria prior to ischemic insult.³⁵ The duration of self-limited, acutely induced AF was shortened by rotigaptide only in the ischemic model. In contrast, Guerra et al. found that rotigaptide significantly improved conduction velocity in normal dogs and dogs with chronic mitral regurgitation but not ventricular tachypacing dogs.³⁶ The mitral regurgitation dogs also had significant reduction in the duration of acutely induced AF episodes. Discrepancies between these results (e.g. improved CV in normals for one study and not the other) are difficult to explain based on the methods used, since they are similar between the studies. Relevance of the rotigaptide data to our results are difficult to assess given these discrepancies in canine results, and the unclear comparability of acute AF episode duration to the persistent AF of the porcine model. The drug mechanism for connexin-modulating peptides is incompletely understood, but it seems to require receptor-binding, protein kinase C activation and modulation of connexin phosphorylation state (see review by Dhein et al. for a more comprehensive discussion of drug mechanism).³⁷ Any of these steps could be impaired in the canine models, thus limiting drug efficacy. Connexin expression levels are another important parameter to consider for the canine studies of rotigaptide. If the protein is not there, then the drug is unlikely to affect its function. Basic pharmacokinetics also need to be considered when assessing drug effect. Is the drug getting to the target at the time and in the concentration necessary to achieve its effect? None of these issues is investigated in the rotigaptide studies.

In our AF-heart failure model, decreased Cx43 expression occurred in the setting of CV slowing and onset of persistent AF. Gene transfer-induced increase in either Cx40 or Cx43 prevented the CV decrease and stopped the fibrillation, suggesting a mechanistic link between connexin expression, CV and AF in this model, a finding supported by recent data from Bikou et al.³⁴ In our study, the AF-controls had left atrial dilation and reduced ejection fraction relative to AF-Cx40 and AF-Cx43 animals, likely a result of differences in AF burden. These gross structural changes did not manifest in differences at a microstructural or cellular level. Atrial fibrosis and cellular myopathic changes were similar between groups as was left ventricular structure and function, so those potential confounders could not explain the observed result. Since Cx43 was decreased but Cx40 was not, our data suggest that Cx40 and Cx43 may be interchangeable in their ability to preserve cellular connectivity and conduction.

Absence of a phenotype in SR animals

An interesting observation in this report is the absence of any discernable increase in CV for the SR animals paired with the CV preserving effects in the AF animals. These data verify a role for connexins in determining CV but suggest that other obstacles to conduction, unrelated to cellular connectivity, are the principle determinants of CV under normal circumstances.

Our data are the first report of atrial CV after Cx 40 and 43 overexpression. The prior literature focused primarily on decreasing or eliminating either Cx40 or Cx43, and the resulting effects on atrial CV have been inconsistent. Thomas et al. found no difference for in vivo measures of atrial CV when comparing wild type to heterozygous Cx43-deleted mice.³⁸ In contrast, Beauchamp et al. observed a 44% reduction in CV with in vitro measurement of atrial myocytes from mice homozygous for Cx43 deletion.³⁹ Beauchamp et al. also observed an increase in CV in vitro with atrial myocytes isolated from Cx40 knock-out mice. Leaf et al. saw regional heterogeneity but overall no change in atrial CV with ex vivo optical mapping of either homozygous or heterozygous Cx40 knock-out mouse hearts,⁴⁰ and Verheule et al. found a 30% decrease in atrial CV in vivo with heterogeneous Cx40 deleted mice.⁴¹ In a study inducing limited Cx43 overexpression, Bikou et al. found increased CV in right but not left atria.³⁴

Methodological differences likely account for some variability in this literature. The in vivo methods used by Thomas and Verheule are the least precise or controllable but arguably the most relevant. The ex vivo mapping of Leaf has the advantage of better resolution and therefore potentially more reliable CV measurement, but these data need to be considered within the context of the metabolic changes resulting from loss of blood perfusion, the use of voltage-sensitive dyes and excitation-contraction uncouplers, and the potential tissue damage from harvest and manipulation. Beauchamp’s in vitro mapping has the greatest likely precision and control, but the extreme manipulations necessary to create this model limit application to the in vivo environment. To overcome these limitations, we evaluated atrial conduction with a combination of in vivo and ex vivo methods. We saw similar trends in the overall effect across methods (no change in SR, decrease in AF, preservation to SR levels in AF with Cx gene transfer), but the magnitude of conduction effects differed with method and conditions.

Another potential confounder is compensatory change in expression of untargeted connexins. In particular, Beauchamp’s observations could potentially be explained by the level of Cx43 expression alone, since Cx43 increased in the Cx40 knock-out mice where CV increased, and decreased in the Cx43 knock-out along with decreased CV. Thomas and Leaf saw no compensatory changes and no change in overall CV. We saw an increase in Cx40 expression for the SR-Cx43 group, but no compensatory change in any other group. The potential interaction between Cx40 and Cx43 was explored by Cottrell et al. who evaluated various combinations of Cx40 and/or Cx43 expressing cells in vitro.^42,43 They found a shift to a lower conductance state when mixed Cx40/43 channels docked with pure Cx43 channels but not when the mixed channels docked with pure Cx40 channels. These results suggest a more complex relationship than can be estimated by looking only at Cx40:Cx43. Further study is warranted to increase our understanding of this issue, but that is beyond the scope of the current report.

In our data, the absence of a phenotype in the SR group does not appear to be connected to Cx localization, phosphorylation or function. We saw no change in the relative amount of Cx43 at the intercalated disk for the SR-Cx43 subgroup, and only a subtle alteration in Cx40 localization for the SR-Cx40 subgroup. We did not see any change in the ratio of phosphorylated to total Cx43 with gene transfer. Successful preservation of conduction in the relevant AF groups further suggested appropriate processing and function of the transgenic connexins. The SR result suggests that under normal operational circumstances, gap junctional resistance is not the principle limitation to conduction, and that increased expression beyond the normal level cannot overcome other determinants of conduction velocity.

Conclusion

In this study, we have shown that atrial overexpression of Cx40 and Cx43 had no discernable effect on conduction in SR animals, but both transgenes preserved CV and prevented sustained AF. We also found considerable evidence for safety of this approach. For these proof-of-principle data, we focused on the short-term connexin gene transfer to assess conduction velocity and anti-fibrillatory effects. Future study with permanently expressing vectors like AAV or lentivirus would allow evaluation over the longer-term to assess interaction with varying levels of atrial pathology. Our results show that atrial-specific gene therapy with gap junction protein provides a novel paradigm for AF treatment based on improvement of intercellular communication.