Extracellular Perinexal Separation Is a Principal Determinant of Cardiac Conduction

Abstract

Background:

Cardiac conduction is understood to occur through gap junctions. Recent evidence supports ephaptic coupling as another mechanism of electrical communication in heart. Conduction via gap junctions predicts a direct relationship between conduction velocity (CV) and bulk extracellular resistance. By contrast, ephaptic theory is premised on the existence of a biphasic relationship between CV and the volume of specialized extracellular clefts within intercalated discs such as the perinexus. Our objective was to determine the relationship between ventricular CV and structural changes to micro- and nano-scale extracellular spaces.

Methods:

Conduction and connexin43 (Cx43) protein expression were quantified from optically mapped guinea pig whole-heart preparations perfused with the osmotic agents albumin, mannitol, dextran 70kDa, or dextran 2MDa. Peak sodium current was quantified in isolated guinea pig ventricular myocytes. Extracellular resistance was quantified by impedance spectroscopy. Intercellular communication was assessed in a heterologous expression system with fluorescence recovery after photobleaching. Perinexal width was quantified from transmission electron micrographs.

Results:

CV primarily in the transverse direction of propagation was significantly reduced by mannitol, and increased by albumin and both dextrans. The combination of albumin and dextran 70kDa decreased CV relative to albumin alone. Extracellular resistance was reduced by mannitol, unchanged by albumin, and increased by both dextrans. Cx43 expression and conductance, and peak sodium currents were not significantly altered by the osmotic agents. In response to osmotic agents, perinexal width, in order of narrowest to widest, was: albumin with dextran 70kDa, albumin or dextran 2MDa, dextran 70kDa or no osmotic agent, and mannitol. When compared in the same order, CV was biphasically related to perinexal width.

Conclusions:

Cardiac conduction does not correlate with extracellular resistance, but is biphasically related to perinexal separation, providing evidence that the relationship between CV and extracellular volume is determined by ephaptic mechanisms under conditions of normal gap junctional coupling.

Graphical Abstract

INTRODUCTION

Understanding cardiac conduction is critical to the diagnosis, treatment, and prevention of arrhythmogenic diseases of the heart. Cardiac conduction is traditionally understood to follow a cable-like model of electrical conduction.^1–4 In this paradigm, electrical current is envisaged as flowing into myocytes, through gap junction (GJ) channels at the intercalated discs (ID), back across cell membranes through outward potassium channels, and returning via the extracellular space. Importantly, experimental evidence has provided support for this long-established theoretical conception by demonstrating decreasing extracellular resistance can increase conduction velocity (CV).^2,5,6

Since electrical propagation occurs in the hearts of lower vertebrate animals in which GJs are scarce,^7–9 and persists in mammalian hearts with the principal GJ protein connexin43 (Cx43) knocked out,¹⁰ computational models of conduction were revised to propose an additional method of electrical intercellular communication.^11–16 In these models of ephaptic coupling (EpC: non-GJ, non-synaptic coupling), strongly depolarizing inward currents facing a 10–30 nm cleft separating adjacent cardiomoycytes significantly decrease the extracellular potential to depolarize nearby membranes. In the case of dense sodium channel localization to the ID between cardiomyocytes, and more specifically in the GJ adjacent perinexus,¹⁷ sodium channels self-activate in a positive feedback process that rapidly decreases the extracellular potential and activates more sodium channels¹⁸ while simultaneously increasing intracellular potentials. Importantly, the rapid decrease of extracellular potentials, increase in intracellular potentials, and depletion of extracellular sodium in the cleft^13,19 decrease the driving force during intercellular propagation in a process known as self-attenuation.¹² All computational models with EpC reveal that starting from a very narrow cleft, widening clefts between cardiomyocytes will first increase CV by reducing self-attenuation and then decrease CV by reducing self-activation. Importantly, the relationship between CV and cleft width is biphasic when GJ conductance is closer to experimentally measured values of gap junctional coupling (GJC).^20–27

In order for conduction to occur, current must return to depolarizing myocytes through the bulk extracellular space, and cardiac CV should be directly proportional to extracellular volume if conduction were only dependent on cable-like propagation through GJs without EpC. However, we previously demonstrated that cardiac CV can be inversely correlated to extracellular volume.²⁸ Subsequent studies revealed that the osmotic agent mannitol, which increases bulk extracellular volume, also increases perinexal volume and reduces CV.²⁹ Since we found that ventricular conduction slowed with perinexal expansion,^30,31 we concluded that EpC is a significant determinant of CV. However, the relative contributions of cable-like propagation through bulk interstitial volume and ephaptic-like conduction through perinexal domains remain unresolved. The purpose of this study was to determine the relationship between ventricular CV and structural changes to bulk extracellular volume and perinexal nanodomains. Results demonstrate osmotic agents alter CV primarily via an ephaptic mechanism.

METHODS

Data Availability

Data supporting the findings of this study are available from the corresponding author upon reasonable request. An expanded Materials and Methods section can be found in the Data Supplement.

RESULTS

Whole-Heart Electrophysiology

We quantified ventricular conduction velocity (CV) by optical mapping during perfusion with our historical lab standard solution³¹ (baseline) and after 15 minutes perfusion with the addition of one of the osmotic agents. In Figure 1A, representative optical maps demonstrate how CV is affected by time and osmotic agents. After 15 minutes, CV may modestly increase in time control optical maps, is decreased with mannitol (as evidenced by crowding of isochrones), is increased by albumin (as evidenced by enhanced isochrone spacing), is modestly increased by dextran 70kDa, and increased by dextran 2MDa. Summary data in Figure 1B demonstrate the effect of the osmotic agents relative to baseline measures in the same heart, as well as a comparison of the paired change in conduction to time controls. In short, CV in the transverse direction (CV_T) was not significantly different after 15 minutes of time control perfusion (Diff: 1.1 cm/sec, 95%CI: ±1.3). Mannitol significantly decreased CV_T relative to baseline measurements (Diff: −2.7 cm/s, 95%CI: ±1.8), and the change in CV_T was significantly lower (Diff: −3.7 cm/s, 95%CI: ±3.0) than the change in CV_T observed in the 15-minute time control measurement. Albumin, on the other hand, significantly increased CV_T (Diff: 7.2 cm/s, 95%CI: ± 2.4) relative to baseline, and the change was significantly greater (Diff: 6.1 cm/s, 95%CI: ± 2.8) than the change found for the time control. Interestingly, both dextran 70kDa and dextran 2MDa increased CV_T in paired comparison (Diff: 1.0 cm/sec, 95%CI: ±0.9 and Diff: 2.7 cm/sec, 95%CI: ±1.6), but the change in CV_T was not significantly different (Diff: −0.1 cm/s, 95%CI: ±3.0 and Diff: 1.7 cm/sec, 95%CI: ±3.0) from the time control. CV in the longitudinal direction (CV_L) significantly increased after 15 minutes of time control perfusion (Diff: 4.1 cm/sec, 95%CI: ±3.9). Mannitol did not significantly change CV_L (Diff: −2.1 cm/sec, 95%CI: ±3.5). Albumin (Diff: 8.7 cm/sec, 95%CI: ±4.1) and dextran 70kDa (Diff: 4.5 cm/sec, 95%CI: ±4.4) significantly increased CV_L. Dextran 2MDa did not significantly change CV_L (Diff: 1.9 cm/sec, 95%CI: ±2.9). However, none of the magnitudes of change in CV_L were significantly different from time control. In general, CV changes induced by the osmotic agents in this study measurably changed transverse conduction. In summary, mannitol decreased, while albumin and both dextrans increased epicardial conduction measured by optical mapping relative to their paired pre-treatment baseline.

Figure 1. Osmotic agents modulate transverse conduction velocity (CV).

A. Representative 3 ms activation time isochrones demonstrate the effect of various osmotic agents on the spread of activation. A detailed explanation can be found in the methods supplement B. CV plotted as individual points (left axis) and the change in CV (right axis) from paired baseline experiments demonstrates mannitol (N=10) significantly decreases, albumin (N=14) significantly increases, dextran 70kDa (N=10) increases, and dextran 2MDa (N=10) increases CV preferentially in the transverse direction *p<0.05 relative to paired baseline measurement, paired Student’s t-test. #p<0.05 relative to time control, ΔCV across groups compared with ordinary one-way ANOVA applying Dunnett’s correction.

Neither APD₉₀ nor maximum APD dispersion were significantly different from baseline in any treatment group (Figure S4).

Cx43 and Gap Junctions

In order to determine possible effects of osmotic agents on gap junctional coupling (GJC), levels of the principal ventricular gap junction (GJ) protein connexin43 (Cx43) were quantified by western blotting. Total Cx43 expression was not expected to change in response to 15 minute exposure to the osmotic agents, and so Cx43 phosphorylated at serine 368 (pCx43-S368) was also quantified since it has been associated with changes in GJ conductivity.^50–52 Across all conditions, neither total nor pCx43-S368 (Figure 2A,B) significantly changed in response to osmotic agent perfusion. Relative to time control tissue, total Cx43 was not significantly different in tissue perfused with mannitol (Diff: 0.1 AU, 95%CI: ±0.4), albumin (Diff: 0.3 AU, 95%CI: ±0.4), dextran 70kDa (Diff: −0.2 AU, 95%CI: ±0.4), nor dextran 2MDa (Diff: −0.02 AU, 95%CI: ±0.5). With respect to time control, pCx43-S368 was also not significantly different in tissue perfused with mannitol (Diff: 0.1 AU, 95%CI: ±0.6), albumin (Diff: −0.2 AU, 95%CI: ±0.6), dextran 70kDa (Diff: −0.1, 95%CI: ±0.6), or dextran 2MDa (Diff: −0.1 AU, 95%CI: ±0.5).

Figure 2. Osmotic agents do not significantly alter Cx43 protein expression or function.

A. Representative western blots suggest that neither total connexin43 (Cx43) nor Cx43 phosphorylated at serine-368 (pCx43-S368) significantly change in response to acute exposure to osmotic agents. Lane 7 contains a positive control using hearts exposed to 60 min of room temperature ischemia, a condition shown to upregulate pCx43-S368.⁹¹ B. Summary data do not reveal significant differences in either total or pCx43-S368 for all conditions (N=3 hearts per condition. p=n.s. via ordinary one-way ANOVA applying Dunnett’s correction. Center is the mean and whiskers are standard deviation). C. Representative fluorescent images of calcein-AM fluorescence in Control cultures (N=13 plates, n=220 cells) and cultures exposed to Mannitol (N=7 plates, n=116 cells), Albumin (N=4 plates, n=77 cells), Dextran 70kDa (N=4 plates, n=102 cells), Dextran 2MDa (N=4 plates, n=95 cells), and Carbenoxolone (CBX, N=6 plates, n=110 cells). D. Fluorescence recovery after photobleaching (FRAP) plot of bleached cells during 300s post-bleaching. Solid lines denotes the average and shaded areas the standard error of the mean. E. Only CBX significantly decreased FRAP relative to control. *p<0.05, nested ANOVA applying Dunnett’s correction with culture plate nested under treatment type.

Since protein expression does not always correlate with function, functional GJC was also assessed by Fluorescence Recovery After Photobleaching (FRAP) in heterologous cells stably expressing Cx43 using an approach that we have reported previously.^43,44,53 Representative fluorescent images taken before photobleaching (prebleach, Figure 2C), immediately after bleaching (t=0s), and at 150 and 300s post-bleaching, demonstrate fluorescence recovers to approximately the same extent as control cells. Specifically, FRAP, as can be seen in summary data plotted in Figures 2D and E, was not significantly different for cells incubated with mannitol (Diff: 1.0%, 95%CI: ±16.2), albumin (Diff: −4.2%, 95%CI: ±20.9), dextran 70kDa (Diff: 3.8%, 95%CI: ±20.7), or dextran 2MDa (Diff: 3.6%, 95%CI: ±20.8). As a positive control, the GJ uncoupler carbenoxolone (CBX, 10 μM) decreased FRAP as expected (Diff: −23.8%, 95%CI: ±5.9). Despite a visual drop in FRAP measured in cells treated with albumin, summary data in Figure 2E reveal that only CBX significantly decreased GJC estimated by FRAP. Even if 1-hour perfusion with albumin modestly decreased GJC, that finding would be inconsistent with the observed increased CV in hearts perfused with albumin. In summary, we did not measure a change in Cx43 level or function in response to the osmotic agents employed in this study at the concentrations or incubation times noted.

Peak Sodium Current

Next, we sought to determine whether the osmotic agents affected the sodium current, which is another important determinant of cardiac conduction. Representative current traces from a single cell, the average of average current-voltage measurements from multiple cells in each heart, and summary peak current for all cells and hearts are plotted in Figure 3A–D. In summary, the osmotic agents did not significantly change peak sodium currents relative to control replicates obtained from the same hearts. Specifically, the peak sodium current relative to control was not significantly different in cells treated with mannitol (Diff: 4.1 pA/pF, 95%CI: ±11.0), albumin (Diff: −1.1 pA/pF, 95%CI: ±13.9), dextran 70kDa (Diff: −2.6 pA/pF, 95%CI: ±14.1), or dextran 2MDa (Diff: −0.7 pA/pF, 95%CI: ±14.0).

Figure 3. Osmotic agents do not significantly alter peak sodium current.

Representative currents from a single cell, current-voltage curves averaged from a single heart, and peak current at −20 mV step potential for all hearts and cells reveal none of the test agents significantly decrease sodium currents relative to Control A. Mannitol (N=5 hearts, n=34 cells, red) B. Albumin (N=3 hearts, n=25 cells, green) C. Dextran 70kDa (N=3 heart, n=21 cells, purple) D. Dextran 2MDa (N=3 hearts, n=24 cells, blue) All treatment groups were compared to same day measurements from control group cells using a nested ANOVA applying Šidák’s correction, where the heart is the nested variable.

Bulk Interstitium

To test whether the volume of bulk interstitial space (VIS) correlated with changes in CV, the percentage of tissue unoccupied by cardiac cells was quantified with a positive-pixel analysis of hematoxylin and eosin (H&E)-stained tissue. Histological quantification did not reveal significant differences between tissues subjected to 15 minutes of perfusion with any osmotic agent (Figure 4). Relative to control, VIS was not significantly changed by perfusion with mannitol (Diff: 1.6%, 95%CI: ±11.4), albumin (Diff: 3.0%, 95%CI: ±11.9), dextran 70kDa (Diff: −6.1%, 95%CI: ±11.1), or dextran 2MDa (Diff: −1.0%, 95%CI: ±12.4). This was unexpected since our previous results from ventricular tissue analyzed from glutaraldehyde fixation and snap frozen samples demonstrated albumin decreases bulk extracellular volume and mannitol increases it.²⁸

Figure 4. Osmotic agents do not significantly change interstitial space analyzed from hemotoxalin and eosin (H&E) stained ventricular tissue.

A. Representative images of H&E-stained tissue B. Volume of bulk interstitial space (VIS) as quantified by a positive-pixel algorithm for Control (N=8 hearts, n=84 images), Mannitol (N=8 hearts, n=73 images), Albumin (N=7 hearts, n=71 images), Dextran 70kDa (N=9 hearts, n=91 images), Dextran 2MDa (N=6, n=62 images). p=n.s. nested one-way ANOVA applying Šidák’s correction, with heart nested under treatment type.

We therefore sought to quantify tissue impedance at baseline and after 20 minutes of treatment with the osmotic agents using a 4-electrode technique as previously described.^46,47,54 Extracellular tissue resistance (R_e) was measured in each heart without interventions (time control) and after serial perfusion with the osmotic agents. This permitted a pair-wise comparison to reduce errors associated with different depths of electrode penetration and orientation of electrodes relative to myocardial ultrastructure and transmural rotational anisotropy. The specific values of R_e are less important than whether the osmotic agent changed R_e and if R_e increased or decreased. Summary data in Figure 5A reveal that R_e did not significantly change over 20 minutes relative to baseline for time control measurements (Diff: 13.6 Ω, 95%CI: ±89.7), mannitol significantly reduced R_e (Diff: −4.9 Ω, 95%CI: ±5.0), albumin did not change R_e (Diff: 14.3 Ω, 95%CI: ±85.7), dextran 70kDa significantly increased R_e (Diff: 20.6 Ω, 95%CI: ±10.9), and dextran 2MDa significantly increased R_e (Diff: 70.3 Ω, 95%CI: ±68.4). Importantly, CV changes observed above do not correlate with R_e.

Figure 5. Osmotic agents modulate extracellular resistance.

A. Neither extracellular resistance (R_e), nor the constant-phase element (CPE) parameter η varied during Time Control (N=4). B. Mannitol decreased R_e but did not change η (N=7). C. Albumin did not affect either R_e or η (N=4). D. However, Dextran 70kDa significantly increased R_e but not η (N=4). E. Similarly, Dextran 2MDa increased R_e but not η (N=4). *p<0.05, paired 2-tailed Student’s t-test, relative to baseline.

Intercellular parameters, including cell membrane capacitance, cytosolic resistance and GJ resistance were modeled as a constant phase element (CPE). The distributivity of the time constant of the CPE element, the parameter η, did not significantly change relative to paired baseline measurements: time control (Diff: 0.05 AU, 95%CI: ±0.9), mannitol (Diff: 0.11AU, 95%CI: ±0.10), albumin (Diff: 0.00 AU, 95%CI: ±0.23), dextran 70kDa (Diff: −0.03AU, 95%CI: ±0.16), and dextran 2MDa (Diff: −0.03AU, 95%CI: ±0.20), as can be seen in Figure 5. These data suggest the R_e changes described above are unlikely to be the result of changes to cellular or tissue factors beyond R_e.

Molecular Exclusion from the Intercalated Disc

In order to estimate whether the different sized osmotic agents can permeate the intercalated disc (ID), the ID distribution of different sized fluorescently tagged dyes (0.5, 3, and 10kDa) was measured by laser scanning confocal microscopy. Representative images in Figure 6A reveal the presence of the 0.5kDa dye in the ID, and the exclusion of the 3 and 10kDa dyes. Figure 6B demonstrates how ID relative to bulk extracellular dye distribution was quantified by taking fluorescent intensity profiles of the dye through the Cx43 immunolabelled ID. By quantifying the intensity of dye beyond the Cx43 signal (lateral membrane) relative to the dye at the peak of the Cx43 signal, summary data on relative ID localization reveal that 0.5kDa dyes can permeate the ID with a normalized fluorescence intensity of 3.6 (95%CI: ±0.88), which is significantly greater than 1, indicating the dye accumulates in the ID. Relative normalized fluorescence intensity for 3kDa dye was 0.30 (95%CI: ±0.20) and 10kDa dye was 0.12 (95%CI: ±0.04), which are both significantly below 1, indicating these compounds were largely excluded from the ID. The data suggest that even if 66.5kDa albumin freely diffused into the bulk interstitium, it would not penetrate the ID, nor would dextran 70kDa or 2MDa. In short, small molecular weight molecules <3kDa can penetrate the ID and cause osmotically mediated dehiscence, while larger molecules may be size restricted from the ID and prevent or reduce fluid accumulation in these nanodomains.

Figure 6. Molecules smaller than 0.5kDa can permeate the intercalated disc.

A. Representative images of ventricular tissue slices perfused with 0.5, 3, and 10kDa fluorescent dyes. Panels demonstrate extracellular diffusion. Cx43 signal is used to identify the intercalated disc (ID), and nuclei are shown alone and with overlay of all signals. B. A region of interest is identified around the Cx43 ID signal and intensity of signal is averaged to identify the peak of the Cx43 signal. The same spatial averaging with 0.5, 3, and 10kDa signals demonstrates a dip in 3kDa and 10kDa signals with a peak in the 0.5kDa dye signal demonstrating the 0.5kDa dye permeates the ID, but the other two do not. C. The intensity of the dye fluorescent signal at the peak of the Cx43 signal is normalized to the intensity of signal beyond the ID to reveal that fluorescent signal intensities for 3kDa (n=3 images) and 10kDa (n=4 images) are significantly lower than for 0.5kDa (n=7 images) signal intensity. *p<0.05 relative to 0.5kDa, one way ANOVA with Dunnett’s correction.

Ventricular Perinexus

To test the hypothesis that different sized osmotic agents differentially modulate ID volume, we quantified the width of the cardiac perinexus (W_P) with transmission electron microscopy (TEM). Representative perinexi from the left ventricle of hearts are shown in Figure 7A at 150,000X magnification. The osmotic agents differentially modulated cell membrane separation immediately adjacent to the GJ plaque within 15 minutes of osmotic agent perfusion. Specifically, and relative to time control, mannitol significantly increased W_P by 4.3 nm (95%CI: ±3.1), albumin significantly narrowed W_P by −5.1 nm (95%CI: ±2.9), dextran 70kDa did not significantly change W_P (Diff: 0.5 nm, 95%CI: ±2.1), and dextran 2MDa significantly narrowed W_P by −3.9 nm (95%CI: ±1.8), as can be seen in Figure 7B. Importantly, perinexal quantification revealed an inverse correlation between W_P and conduction. Specifically, mannitol widened W_P and decreased CV consistent with previous reports,²⁸ and albumin and dextran 2MDa narrowed W_P and increased CV.

Figure 7. Cardiac perinexal width inversely correlates with CV.

A. Representative TEM images of ventricular perinexi (shaded yellow) suggest osmotic agents differentially modulate perinexal width (W_P). B. Summary data reveal that relative to Time Control (N=9 hearts, n=140 images), Mannitol significantly increases (N=10 hearts, n=170 images), Albumin significantly decreases (N=12 hearts, n=135 images), Dextran 70k Da does not change (N=5 hearts, n=74 images), and Dextran 2MDa significantly decreases (N=5 hearts, n=75 images) W_P. Dashed line represents mean W_P for time control. *p<0.05, as evaluated with a generalized linear mixed effects model specifying a gamma distribution and identity link between samples collected from the same heart.

Self-Attenuation

To this point, presented data demonstrate that CV is inversely correlated to W_P. Yet, the relationship between W_P and CV is biphasic. For very narrow perinexi, CV can be directly proportional to W_P by altering sodium channel self-attenuation.^12,35,55 In a final set of experiments, we hypothesized that combining dextran 70kDa with albumin would further decrease W_P and slow CV relative to albumin alone, consistent with previous studies.⁶ Representative perinexal images and summary data in Figure 8A demonstrate that albumin and dextran 70kDa perfused concurrently significantly decreased W_P relative to albumin alone by 2.3 nm (95%CI: ±2.0).

Figure 8. Dextran 70kDa in the presence of albumin can narrow perinexal width and slow conduction relative to albumin alone.

A. Representative TEM images of ventricular perinexi (shaded yellow) and B. summary data reveal that simultaneous perfusion of Albumin and Dextran 70kDa (N=4 hearts, n=60 images) significantly narrows perinexal width (W_P) relative to Albumin alone (N=10 hearts, n=150 images), #p<0.05 relative to albumin, as evaluated with a generalized linear mixed effects model specifying a gamma distribution and identity link between samples collected from the same heart. C. Representative activation time isochrones. D. Summary data reveal that cardiac conduction is significantly slower in hearts treated with both Albumin and Dextran 70kDa (N=15 hearts) relative to albumin alone (N=14 hearts). #p<0.05, unpaired two-tailed t-test. Note CV and Wp data for albumin alone is replicated from figures 1 and 7 for comparison.

On average, albumin perfusion increased CV_T relative to baseline by Diff: 38% (95%CI: ±13.3). Whereas the combination of albumin and dextran 70kDa did not significantly increase CV_T (Diff: 4.8%, 95%CI: ±11.8). Therefore, albumin and dextran 70kDa slowed CV_T relative to albumin alone, by Diff: 33.3% (95%CI: ±14.9). On average, albumin perfusion increased CV_L relative to baseline by Diff: 16.2% (95%CI: ±7.1). Whereas the combination of albumin and dextran 70kDa did not significantly increase CV_T (Diff: 4.5%, 95%CI: ±6.2). Therefore, albumin and dextran 70kDa slowed CV_L relative to albumin alone, by Diff: 11.7% (95%CI: ±9.8). These data demonstrate that dextran 70kDa can slow conduction in the presence (Figure 8B), but not the absence of albumin.

DISCUSSION

Herein, we present two primary observations: 1) The cardiac perinexus, a specialized nanodomain of extracellular space within the intercalated disc (ID), can be osmotically expanded and collapsed on a timescale of less than 15 minutes, and 2) The regulation of cardiac epicardial conduction velocity (CV) is biphasically related with perinexal width (W_P). Among the conditions tested, we did not observe significant changes in total connexin43 (Cx43) expression or phosphorylation at serine 368 (pCx43-S368), nor did we find altered gap junction coupling (GJC) estimated by fluorescence recovery after photobleaching (FRAP). Additionally, the lack of a monotonic correlation between CV and altered tissue resistance argues for a minimal role in cable-like conduction mediated by gap junctions (GJ) alone. Furthermore, the osmotic agents did not significantly alter peak sodium current, although the change could have been below the resolution of detection. Our interpretation of these data is that extracellular volume changes in sodium channel-rich nanodomains such as the ID perinexus, rather than bulk extracellular resistance (R_e), play a substantial role in modulating CV by an ephaptic mechanism in the absence of GJ remodeling.

In the present study, CV measurements for both mannitol and albumin were similar to those described earlier.²⁸ Mannitol expanded W_P as we reported earlier.^19,29 Albumin and dextran 2MDa both decreased W_P and increased CV. When conduction changes were noted, they were primarily observed in the transverse direction of propagation. This is consistent with electrical wavefronts encountering more ID, with associated GJs and ephapses, along the short versus the long axis of cardiomyocytes for a given unit length. More specifically, as cardiomyocytes are highly polarized and IDs are located primarily on the edges of the long-axis, transverse propagation requires the electrical wavefront to “zig-zag” through the lateral edges. Since conduction across the ID is slower than through the myocyte,⁵⁶ the increased number of junctions in the transverse, relative to the longitudinal direction will multiply the effect of modulating junctional communication on cardiac conduction. Interestingly, 40 g/L dextran 70kDa alone modestly but significantly increased CV in the whole-heart preparation, which seems to be in contrast to a previous study that found conduction slows with dextran 70kDa.⁶ Unlike our study, albumin was present in the baseline perfusion solution of the previous independent study. This being said, we replicated previous findings that dextran 70kDa in the presence of albumin will prevent the conduction increase associated with albumin alone. We propose the mechanism of CV slowing with dextran 70kDa and albumin is more dependent on perinexal narrowing resulting in ephaptic self-attenuation than a decrease in bulk extracellular impedance. Self-attenuation is predicted to occur when sodium channels decrease extracellular potential or sufficiently deplete extracellular sodium such that the inward driving force of this ion is reduced in ephaptic nanodomains. The results of the experiments with concurrent perfusion of albumin and dextran 70kDa are consistent with previous demonstrations of self-attenuation,^19,35 and reconciles conduction differences observed during osmotic stress by our group^28,29,31,37 and others.⁶

Gap Junctions

Cx43-formed GJs and R_e are widely accepted as major determinants of cardiac conduction.^57–59 Since the half-life of cardiac Cx43 is on the order of 1–2 hours,⁶⁰ and we observed CV changes at 15 minutes, it is not surprising that we do not measure a significant change in overall connexin protein expression. The further absence of associated changes in pCx43-S368 level is consistent with the lack of observable change in functional GJC estimated from FRAP assays. We previously demonstrated that Cx43 expression levels were not significantly changed by the same concentrations of albumin and mannitol perfused in this study.²⁸ Though we cannot definitively rule out changes to GJ assembly or other Cx43 post-translational modifications, our data, combined with the short time scale of conduction changes, suggest that the primary mechanism for conduction changes may not be significantly attributed to GJC.

Most computational models of ephaptic coupling (EpC) and GJC do not demonstrate a biphasic relationship between CV and cleft width at “nominal” GJC,^12–15 leading to the hypothesis that EpC is not particularly relevant during conditions of “normal” GJC. These models often reveal CV slowing secondary to widened cleft widths when GJC is reduced by more than 80%.^12–15 As a result, a recurring question is whether EpC has a significant impact on cardiac conduction when GJC is nominal. In addressing this question, it is important to note that early computational models of cardiac conduction mediated by EpC and GJC often assumed normal GJ conductivity that is significantly greater^12–15,56 than has been measured experimentally between ventricular cardiomyocytes. Specifically, direct measurement of GJ conductances by dual-cell patch clamp yields estimates ranging from 3 to 2530 nS, with the majority of studies reporting an average GJC of less than 550 nS, independent of species.^20–27 Based on this consensus average, actual levels of GJC between myocytes are likely an order of magnitude smaller than the 2000nS values typically used as a nominal estimate of coupling in many computational models. Furthermore, even in models that do not incorporate EpC, GJC is often tuned to higher values than those experimentally measured for the purpose of producing in silico CV values consistent with measurements made in heart.^61,62 It is possible that unaccounted for EpC could be a reason for non-physiologic GJC tuning that is necessary to replicate experimentally measured CV in hearts. Importantly, when one retrospectively re-evaluates these key early models in the context of consensus levels of GJC experimentally reported in the literature, EpC emerges as a major determinant of CV under normal, as well as pathological conditions of GJC.

Perinexal Width

In addition to the aforementioned biphasic predictions of the CV-cleft width relationship in computational models of cardiac conduction, computational models that include both EpC and GJC also predict that CV sensitivity to GJC is reduced when intercellular separation is below approximately 20 nm and significantly enhanced at much wider cleft widths. These predictions have since been validated in mice and guinea pigs.^29,35,63 Additionally, expansion of the perinexus has been associated with chronic atrial fibrillation in humans.⁴⁹

Based on the observations that CV is inversely correlated with W_P during perfusion with the individual osmotic agents, and that combining albumin and dextran 70kDa narrowed W_P more than albumin alone and prevented the albumin mediated CV increase, the data can be interpreted to support an ephaptic mechanism of osmotically regulated cardiac conduction. Specifically, very narrow perinexi are associated with relatively slow conduction (albumin and dextran 70kDa). Expanding the perinexus first increases CV (albumin or dextran 2MDa), then starts to slow CV (control or dextran 70kDa), and eventually dramatically slows CV as perinexi are dramatically disrupted (mannitol). These observations are consistent with an expected biphasic CV-W_P relationship predicted by models which include both EpC and GJC.^11–14,64

While we frequently report W_P near 20 nm, we also have demonstrated that the absolute value is dependent on the blinded experimentalist who traces the boundary of the perinexus in transmission electron microscopy images.⁴⁹ Additionally, estimating a 3-dimensional volume from 2-dimensional images limits the precision of estimating perinexal volume, and therefore correlation between CV and W_P. Even so, the relationship between CV and W_P is biphasic and therefore non-monotonic. Regardless, modelling and experimental data should be compared cautiously when attempting to establish numerical relationships between something as simple as channel conductance, whole-cell electrophysiology, and tissue level responses. For example, models often assume a cylindrical geometry of myocytes with a surface area that significantly underestimates the tortuosity of the ID. Incorporating ion tracking¹³ or simple structural detail into the model such as interdigitation or the known disposition of larger GJs to encircle the ID¹⁵ modulates the response and range of CV observed with altered cleft width. The estimated CV values in EpC models are also dependent on sodium channel density in the ID,¹⁵ and the location of sodium channel clusters and their relative apposition within the ID.¹⁸ While it is tempting to infer a nominal starting point for any parameter on a predicted curve and speculate if CV should increase or decrease with a specific change in W_P, this approach is likely premature. New studies utilizing models with enhanced structural detail, biophysics of spatial interactions, and ion channel clustering will likely yield important new insights into the effect of EpC on electrophysiology.^55,65

Extracellular Impedance

The electrochemical impedance experiments revealed significant R_e changes at a similar time scale when conduction changes, as reported previously.²⁸ Further, the data were collected on a similar temporal scale to the measurements of previous studies that carefully quantified tissue resistance at each point when conduction was measured.⁶ The physics of cable-like conduction through GJs alone is not compatible with our results, as R_e did not directly correlate with conduction. Notably, whereas both the 70kDa and 2MDa dextrans increased R_e and should have slowed conduction, they did not: Specifically, 70kDa and 2MDa dextran significantly increased CV. Furthermore, mannitol decreased R_e while simultaneously slowing CV as reported previously,²⁸ but according to models only including GJC alone, mannitol should have increased CV.

Importantly, dextran 70kDa increased R_e by 21% in guinea pig whole-heart, similar to what was reported by others in isolated rabbit papillary muscle (a change of roughly 45%).⁶ To address the possibility that one or more osmotic agents altered GJC, cell size, membrane capacitance or any additional intercellular electrical parameters, we modeled the cellular pathway using a constant phase element (CPE). Tissue is composed of various cell types with different shapes and membrane properties, resulting in a distributed cell membrane time constant – the response speed of the cell membrane to electrical stimulation. This distributivity is what we quantified in the parameter, η. Since this parameter did not significantly change with time control, or with the addition of any of our osmotic agents, we conclude that the cell membrane time constant distribution on the tissue level was not substantially affected by our interventions. Therefore, the changes to R_e described above are likely a result of changes to the extracellular compartment, i.e. interstitial space. The lack of change in η is also consistent with non-significant changes induced by the osmotic agents in Cx43 expression or FRAP coupling assays. Importantly, macroscopic impedance measurement and its relationship to the cable theory is at best complicated and requires additional study.

Sodium Channels

Despite the non-significant change in peak sodium current (I_Na) with osmotic agents, it is possible that conduction changes reported in this and previous studies^28,30 could be attributed to a change in peak sodium current below our resolution of detection. It should be noted here that the statistical approach may play a role in the ability to report a significant change in peak conductance. Specifically, if each current obtained from a cell is considered an independent measurement, regardless of which heart it was obtained from, a simple non-paired Student’s t-test would suggest that mannitol modestly decreases peak I_Na on the order of 20%. If mannitol does significantly decrease peak I_Na, this and previous studies may over-associate conduction slowing with perinexal widening. It is possible that a relatively small sodium channel inhibitory effect of mannitol exacerbates conduction slowing under these conditions as well. Yet, perinexal expansion associated with a peptide developed to disrupt sodium channel beta subunit trans-adhesion did not affect peak sodium current density and also slowed conduction in the same isolated guinea pig whole-heart preparation.³¹

The computationally predicted relationship of CV to W_P is also non-symmetric in the self-attenuation (positive correlation) and self-activation (negative correlation) phases of conduction. It is therefore difficult to know where the “normal” heart may reside on the CV-W_P curve because this relationship is also modulated by factors such as sodium and potassium channel localization to the ID, channel unitary conductances, channel availability, cell size, and extracellular sodium and potassium.^{12–15,19,66–69} Importantly, we provide experimental evidence that the degree of perinexal narrowing or expansion is associated with different magnitudes of CV increase or decrease, respectively. Thus, at present, it is difficult to state to what degree the different osmotic agents should change CV for a given change in W_P. We can say with confidence that the change in CV with narrower and wider W_P is not uniform. Regardless, evidence continues to support the conclusion that cardiac conduction is well-described by models including both EpC and GJC.

Membrane and Intracellular Pathways

The 4-electrode distributivity measurement, a metric of cell membrane capacitance, cytosolic resistance and GJ resistance, does not significantly differ between controls and cells or tissue incubated with the osmotic agents. This inconclusive result does not support nor exclude the possibility that osmotically induced changes in cellular parameters are significant determinants of CV. The lack of distributivity change is inconsistent with previous results when cell volume was measured by cell morphometry, for example. Specifically, we and others reported that mannitol can decrease isolated ventricular myocyte volume by approximately 7% to 25% for concentrations of mannitol between 50 mM and 143 mM.^28,70,71 The discrepancy and lack of significant differences in metrics of intracellular impedance in this study could therefore be a result of: a change that is below the detection resolution of the technique, intra- and interventricular heterogeneity,⁷² inter-animal variability, and/or variability secondary to cellular isolation.⁷³ Importantly, a previous computational model of conduction found that the relationship between cell size and CV is complexly dependent on both forms of electrical coupling.^16,69 Specifically, with relatively high GJC (2500 nS), decreasing cell size can slow conduction, and EpC can modestly steepen that relationship. In contrast, with relatively nominal but closer to normal GJC (100 nS), decreasing cell size can increase CV, but once again EpC modestly increases the steepness of this relationship.⁶⁹ Since GJC has been measured over a large range of values, and the average value in literature is closer to 550 nS,^20–27 we do not know which result to expect. Therefore, the effect of parameters such as cell volume on the observed changes in CV still requires significantly more independent investigation for a number of reasons including: 1) Morphometric changes in myocyte geometry are likely different in an isolated, unloaded cell relative to cells in intact myocardium, 2) The osmotic agents will partition differently between vascular, extracellular, and ID compartments based on osmolyte permeability and size, and 3) The molarity of the osmotic agents used in this study are not identical.

Limitations

No intervention can be applied without experimental consequence. Osmotic agents employed in this study have well-established effects in addition to changing oncotic pressure, and some are used clinically for these and other reasons.

Albumin, for example, has been used to regulate blood pressure due to its binding affinity to the vascular endothelium, leaving vessels at a rate of approximately 5% per hour. Albumin furthermore regulates vascular function and can bind to a number of cations, fatty acids and hormones, in addition to acting as an antioxidant.⁷⁴

Mannitol, a sugar alcohol, has been used as a hyperosmotic agent to treat high intracranial pressure.^75,76 With appropriate dosing and time courses of treatment, mannitol will decrease interstitial edema by drawing fluid out of the cerebral parenchyma; though at higher doses and over long time courses, mannitol will also cross the blood-brain barrier. Once across the blood-brain barrier, mannitol will then have the opposite intended effect, more similar to what we observe in cardiac tissue: drawing fluid into tissue and exacerbating edema.

Dextrans of various sizes have been used, like albumin, as a blood volume expander, with larger dextrans having a more pronounced effect on blood viscosity. Dextran 70kDa has antithrombotic properties and reduces the inflammatory response and troponin-I release after cardiac surgery.^77,78 It remains unknown, however, whether the other non-osmotic properties affected our results over the much more acute time frame of less than 15 minutes studied here.

Despite our efforts to select osmotic agents with minimal binding or metabolic effects, we cannot definitively exclude any such effects and advise the reader to cautiously interpret our results about any implications for clinical treatment.

As mentioned previously, experiments were conducted in excised, isolated, male retired breeder, guinea pig ventricles that were perfused with di-4-ANEPPS and preserved for hematoxylin and eosin staining and electron microscopy with formalin and glutaraldehyde, respectively. As such, all values of structural dimensions presented – e.g. W_P and interstitial volume, should be interpreted without implication of explicit in vivo values. Additionally, while our CV results are similar to those published previously,²⁸ we observed no significant differences in interstitial volumes between interventions, though all of our values were similar to the controls reported previously. It is worth noting that the reported interstitial volume differences may be attributed to a shorter protocol (15 minutes in this study vs 60 minutes in previous studies), different fixatives (formalin vs glutaraldehyde), different analysis programs (positive-pixel vs color deconvolution), and the statistical approach. Interestingly, previous work suggests that 30 minutes of perfusion is required for changes in R_e to reach steady-state.⁶

Furthermore, albumin is difficult to work with in optical mapping experiments because signal quality degrades quickly, as demonstrated previously³⁰ and in Figure S2A. The reasons for this are not entirely understood since the absorption peak of albumin is approximately 280 nm^79,80 and peak excitation for di-4-ANNEPS is near 480 nm.

Optical mapping also provides fluorescent signals from a volume of tissue that includes intramural cells, and volume averaging may “blur” the upstrokes of optical action potentials and capture transmural propagation patterns.^81–85 The values of conduction between the left and right ventricle may also be affected by transmural rotational anisotropy given the difference in myocardial thickness.⁸⁶ Additionally, our CV measures represent an average across the anterior epicardial surface, and the averages contain an indeterminate ratio of points from the left and right ventricles, as well as points that are more apical or basal relative to the pacing electrode. Regional heterogeneities in tissue response and which regions of tissue are quantified in a given image may have unanticipated effects on the magnitude of the observed effects, particularly as our measurement field of view and pacing site is biased towards the left ventricle. Therefore, caution should be taken when considering the absolute CV values presented herein. However, our group is careful about conducting positive and time control measurements with optical mapping under a variety of conditions previously reported to affect conduction: sodium channel inhibitors,^29,87 GJ inhibition and Cx43 genetic loss of function^35,68 mannitol or albumin,^28,30 and changes in extracellular potassium.⁶⁸

With regards to whether edema confounds epicardial measures of CV due to volume averaging, we currently discount this hypothesis for a few reasons. First, dehydrating tissues should bring more cells into the field of view epicardially and transmurally. Epicardial recruitment of cells into an imaging volume should decrease the speed of conduction by increasing the number of IDs in each pixel, even if the dimmer mid-myocardium would increase conduction. The epicardial recruitment effect is inconsistent with our finding that solutions associated with increased or decreased R_e, a metric of bulk hydration, do not correlate with the directional change in conduction. Furthermore, we provide confirmatory evidence that dextran 70kDa in the presence of albumin is associated with slower conduction when compared to hearts perfused only with albumin, and this is consistent with studies using microelectrodes, the same osmotic agents, and papillary muscles that lack rotational anisotropy.⁶

Finally, within our tissue impedance experiments, electrode array orientation was visually aligned with local fiber orientation. However, as fiber orientation changes in 3-dimensional myocardial tissue, we cannot compare tissue impedance changes between experiments, only between interventions in the same heart. As a result, only the finding that R_e changed has practical meaning, but the absolute values of R_e and magnitude of change do not have direct physiologic relevance. Additionally, the measurement of R_e does not distinguish between bulk extracellular volume expansion or changes to conductivity in the ID. Both volume changes could change R_e in the same direction.^88,89 However, our observations that CV does not change proportionally in response to R_e argues against bulk extracellular volume expansion as the principal determinant of CV in these experiments.

Conclusions

Clinically relevant osmotic agents may impact cardiac conduction by a previously unappreciated mechanism of modulating nanodomain separation between cardiomyocytes in the perinexi of ID. Specifically, this study demonstrates that expanding ID nanodomains densely expressing ion handling proteins can both increase and decrease cardiac conduction consistent with computational predictions of cardiac conduction dependent on both EpC and GJC. While we cannot definitively exclude the possibility of off-target effects from the experimental interventions studied, our data suggest that EpC modulation of one type of ID nanodomain, the perinexus, is a major determinant of cardiac conduction. The perinexus therefore continues to emerge as a therapeutic target for managing cardiac arrhythmias, and further research is needed to better understand the conditions under which perinexal expansion represents a beneficial adaptation or pathologic remodeling.

Novelty and Significance.

What is Known?

• Changes in extracellular volume affect cardiac conduction and arrhythmia susceptibility.

• Cardiac conduction occurs via intercellular electrical communication that is thought to occur exclusively through gap junction coupling.

• Under such a model, cardiac conduction velocity (CV) should be directly proportional to extracellular resistance.

What New Information Does This Article Contribute?

• Exposure to clinically relevant osmotic agents leads to edema within the cardiac intercalated disc nanodomain known as the perinexus.

• During treatment with these osmotic agents, cardiac conduction does not correlate with extracellular resistance, but is biphasically related to perinexal separation.

• These findings are consistent with computational predictions of cardiac conduction dependent on both ephaptic and gap junctional coupling, but not gap junctional coupling alone.

Clinical efforts to improve cardiac conduction by targeting gap junctions have met with limited success. An alternative form of electrical communication, called ephaptic coupling has been proposed, but controversies about similar experimental conditions yielding different results prevented the reconciliation of electrotonic coupling through gap junctions with ephaptic coupling in the heart. Specifically, electrotonic coupling predicts that conduction is positively correlated with extracellular resistance, and ephaptic coupling predicts a biphasic correlation between conduction and cellular separation in the intercalated disc. This study reconciles previous controversies, reveals that cardiac conduction is biphasically related to perinexal separation, and ephaptic coupling is an important determinant of cardiac conduction. The results support ephaptic coupling as a novel target for improving cardiac conduction and possibly preventing arrhythmias.

Non-standard Abbreviations and Acronyms

APD/APD90

Action Potential Duration/APD as measured at 90% repolarization

CBX

Carbenoxolone

CV_T/L

Conduction Velocity, Transverse/Longitudinal direction

EpC

Ephaptic Coupling

η

Distributivity of the time constant of the CPE

FRAP

Fluorescent Recovery After Photobleaching

GJ

Gap Junction

GJC

Gap Junction Coupling

H&E

Hematoxylin and Eosin

I_Na

Peak Sodium Channel Current

ID

Intercalated Disc

pCx43-S368

Cx43, Phosphorylated at serine 368

R_e

Extracellular tissue resistance

TEM

Transmission Electron Microscopy

VIS

Volume of Interstitial Space

W_P

Width of the Perinexus