The objective of the study was to investigate the incidence, characteristics, and ionic mechanism of AP alternans developing in an experimental model in which we eliminated the CaT, a Ca²⁺ signal that has been proposed to be the source of AP alternans. All procedures were approved by the Animal Care and Use Committee of the University of Utah and complied with the American Physiological Society’s Guiding Principles in the Care and Use of Animals. The experiments were carried out using isolated rabbit (male and female) ventricular myocytes subject to continued pacing at progressively faster rates to promote the development of AP alternans. Myocytes were superfused with a physiological solution which was untreated (group I, Control) or treated with a I_CaL agonist (group II) to determine the effect of stimulating I_CaL on the formation of alternans in cells with conserved intracellular Ca²⁺ cycling. Myocytes were also treated with a Ca²⁺ buffer loaded into the intracellular compartment to determine the effect of buffering the CaT on the formation of AP alternans (groups III and IV, each at a distinct concentration of the buffer), and to determine how buffering CaT interacted with a I_CaL agonist. In the latter condition, the unexpected effect of combining the two test drugs prompted a pacing protocol modification such that rather than subjecting the cells to progressive increases in the pacing rate, we had to reduce the pacing rate applied to the cell progressively in order achieve a pacing response of one captured beat for every stimulus. Pacing induced AP responses were recorded continuously using the patch clamp technique and intracellular Ca²⁺ changes measured in tandem for brief intervals (∼2s) during pacing by means of the Ca²⁺ sensitive fluorescent probe fluo-4. Provided that paced myocytes did not always develop alternans, the total number of cells per group was adjusted in order to obtain 8–14 alternans positive cases in each experimental condition. The investigators were not blinded to the experimental interventions. However, experimental protocols and data analysis were designed to include all myocytes (i.e., alternans positive and alternans free), and an offline computationally driven data analysis implemented so as to remove bias in the quantification of AP alternans by selecting an objectively predetermined section of the AP recordings for analysis, which was equivalent for all experimental conditions. For each myocyte, computational analysis of each AP in response to the complete pacing protocol was implemented via custom software, and used to identify AP alternans sequences.

Adult New-Zealand White rabbit (1.5–1.7 Kg) ventricular myocytes were isolated by combining enzymatic tissue digestion and mechanical trituration (supplement for details) (Saegusa et al., 2013). Transmembrane voltage (V_m) was continuously recorded using a glass pipette (2–4 MΩ tip) and an Axoclamp 2B amplifier (Molecular Devices, Ca) in bridge mode (supplement for details) (Zaniboni et al., 2000). To elicit an AP, we injected a rectangular depolarizing current pulse (duration 3–4 ms) into the myocytes through the patch pipette (supplement for details). Using a previously described approach we loaded the fluorescent probe fluo-4 into the isolated myocytes to track the changes in intracellular Ca²⁺ (Saegusa et al., 2013; Saegusa et al., 2011). Fluo-4 fluorescence signals were recorded using an EMCCD camera (iXon 860, Andor Technology, Belfast, United Kingdom) configured to record images at a resolution of 64 × 64 pixels and 860 frames/s (supplement for details) (Warren et al., 2010). Simultaneously recorded APs and CaTs were aligned for analysis (supplement for details) (Supplementary Figure S1). Where indicated, myocytes were loaded through the glass pipette with 10.0 mM or 20.0 mM BAPTA [1,2-Bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid; Tocris Bioscience, United Kingdom)] (Saegusa et al., 2013), a Ca²⁺ buffer used to effectively reduce the CaT beyond detection. Experiments were carried out at 36.5 ± 1.0°C using Rod-shaped myocytes with well-defined striations.

Once a myocyte was attached to the patch pipette, stimulus pulses were delivered at a pacing cycle length (PCL) of 2000 (or 1000) ms for 2–7 min to establish steady state conditions. Therein, the cells were subject to progressively shorter PCL with the objective of inducing AP alternans (Supplementary Figure S2A). The target PCLs were (in ms): 2000, 1000, 500, 300, 180. Given that each cell responded uniquely to changes in rate, a number of cells were subject to intermediate PCL during implementation of the pacing protocol. At each PCL (including intermediate rates) the cells were constantly paced for ∼30–∼90 s to allow for adaptation before shortening the PCL (Supplementary Figure S2A). Ultimately, the PCL was reduced to attain a final value of 180 ms unless the development of alternans occurred before.

The pacing protocol (Supplementary Figure S2A) was applied to myocytes from four experimental groups: I) myocytes patched with normal pipette solution and superfused with normal Tyrode solution, Control group (n = 22); II) cells patched with normal pipette solution and superfused with Tyrode’s solution containing 25 nM Bay K 8644, Bay K 8644 group (n = 9); III) cells patched with pipette solution containing 10 mM BAPTA and superfused with normal Tyrode solution, 10 mM BAPTA group (n = 10); IV) cells patched with pipette solution containing 20 mM BAPTA and superfused with normal Tyrode solution, 20 mM BAPTA group (n = 9). In the Bay K 8644 group, the pacing protocol was initiated once delivery of Bay K 8644 at baseline elicited a visible effect on the AP (i.e., AP prolongation).

To test the effect of Bay K 8644 on myocytes lacking CaT, myocytes loaded with 10 mM BAPTA bathed in control solution were exposed to 25 nM Bay K 8644 during constant pacing at 500–1000 ms PCL (n = 8). The PCL was adjusted to ensure 1:1 capture at the fastest possible rate as required by the drug effects.

Data reported were generated from the last 40 activations recorded at each PCL (Supplementary Figure S2A). For each AP in this ‘analysis window’ we determined the APD at 90% repolarization (APD₉₀) and the duration of the preceding diastolic interval (DI) (supplement for details). Representative AP’s and APD₉₀ sequences during long-short-long-short type responses are shown in Supplementary Figure S2B (upper and lower, respectively). For each 40 activation-long ‘analysis window’ we determined the maximum number of consecutive APD₉₀ pairs locked in a long-short-long-short type pattern, which we will also refer to as 2:2 type response/pattern. If seven or more consecutive APD₉₀ pairs exhibiting 2:2 type responses were identified we designated the myocytes as ‘alternans positive’, following the probabilistic reasoning introduced by Dilly and Lab (1988) The longest applied PCL satisfying the alternans positive criteria was defined as the ‘alternans threshold’ (ALT-TH). Myocytes devoid of APD₉₀ sequences satisfying the ‘alternans positive’ criteria were classified as ‘alternans free’. Activation in alternans free sequences is also referred to as 1:1 type response.

Data are presented as mean ± standard deviation. For each PCL applied to a given myocyte, we computed the mean DI (Mean-DI) and mean APD₉₀ (Mean-APD₉₀) as the average DI and APD₉₀ values of activations conforming the 40-activation-long analysis window (Supplementary Figure S2A). We then plotted Mean-APD₉₀ versus Mean-DI to generate an APD restitution curve for each myocyte. Myocytes where classified into either an ‘alternans positive” or ‘alternans free’ category, and the segregated data used for group-wise and/or category-wise statistical analysis by means of the two-sided Student’s t-test. We used the Chi-squared test to compare the group-wise fraction of alternans positive myocytes and the group-wise distribution of alternans types. For each 40 activation-long activation sequences meeting the ‘alternans positive’ criteria we determined the alternans magnitude (ALT-MAG) as follows. Initially we identified each APD₉₀ pair locked in a long-short-long-short type pattern. For each pair we then calculated the difference in APD₉₀ values of the first AP in the pair minus the second AP. Finally, by averaging the absolute values of these differences over all pairs exhibiting the 2:2 type response, we determined ALT-MAG. To determine the slope of the APD₉₀ restitution curve at the ALT-TH we determined the slope of the linear function connecting two specific data points of the APD₉₀ restitution curve: data-point#1 consisted of the Mean-DI/Mean-APD₉₀ pair corresponding to the ALT-TH, and data-point#2 consisted of the Mean-DI/Mean-APD₉₀ pair with next largest Mean-DI value. Statistical comparisons yielding p values less than 0.05 were considered significant.

Action potential (AP) alternans could be induced by rapid pacing in some but not all myocytes. Figure 1 depicts representative transmembrane potential changes recorded from control ‘alternans free’ (Figure 1A) and ‘alternans positive’ (Figure 1B) myocytes. Note that the AP sequence of the ‘alternans positive’ myocyte exhibits a long-short-long-short APD₉₀ pattern (Figure 1B), whereas the ‘alternans free’ does not (Figure 1A). Group-wise analysis showed that there were no significant differences in the fraction of ‘alternans positive’ cells between control cells (14/22) and cells perfused with the I_Ca,L agonist Bay K 8644 at 25 nM concentration (8/9; p = 0.16), or cells loaded with either 10 mM BAPTA (7/10; p = 0.73 vs. control) or 20 mM BAPTA (8/9; p = 0.16 vs. control) to buffer CaT.

It has been hypothesized that CaT alternans drive AP alternans (Qu and Weiss, 2023). Therefore, to determine the necessity of CaT in the formation of AP alternans in our model of ‘free-running’ AP alternans developing at physiological temperature, we recorded AP with current clamp and CaT with optical methods simultaneously in alternans free and alternans positive sequences for all experimental conditions. Figure 2A depicts simultaneous AP and CaT signals acquired during PCLs above (upper pair) or below (lower pair) the alternans threshold (ALT-TH) from an alternans positive Control myocyte. Specifically, note how shortening the PCL from 500 ms to 200 ms shifted the response of both the AP and the CaT signals from a non-alternans (1:1) to an alternans (2:2) type response. Due to the finite and relatively short optical CaT recordings relative to continuous patch clamp recordings, we could not determine if either the AP or CaT alternans preceded one another, or if alternans developed simultaneously in both signals. All dual AP and CaT recordings acquired during the development of AP alternans (12 dual recordings from eight different cells) revealed an in-phase 2:2 capture in the AP and CaT signals, such that long APD was accompanied by large amplitude CaT, and conversely short APD with a small amplitude CaT (Figure 2A, lower two traces). Simultaneous AP and CaT recordings from a representative alternans free Control myocyte depict the absence of AP alternans in the 500–180 ms range of PCLs (Figure 2B). Note that the simultaneously recorded CaT signals were also alternans free (Figure 2B, blue traces).

In myocytes perfused with the I_Ca,L agonist Bay K 8644, both AP and CaT developed in-phase alternans (Figure 2C). Importantly, intracellular loading of either 10 or 20 mM of the Ca²⁺ chelator BAPTA reduced the CaT below detection, yet AP alternans were readily inducible (Figures 2D, E, respectively). Finally, the addition of Bay K8644 to agonize I_Ca,L did not restore the CaT buffered by 10 mM BAPTA, yet AP alternans were still inducible (Figure 2F). These data demonstrate that AP and CaT can alternate in phase, and AP alternans can occur in the absence of a measurable CaT.

Representative AP recordings and the corresponding APD₉₀ over time for two PCLs are presented in Figure 3 for untreated Control cells (Figure 3A, leftmost and rightmost panels, respectively), cells treated with Bay K 8466 (Figure 3B, leftmost and rightmost panels, respectively), and cells treated with BAPTA (Figures 3C, D, leftmost and rightmost panels, respectively). For each experimental condition, transmembrane potential recordings were acquired during pacing at a PCLs above the ALT-TH (upper trace, leftmost panels) or equal to or below the ALT-TH (lower trace, leftmost panels). Importantly, the data demonstrate the ability to measure an uninterrupted 2:2 type pattern of APD₉₀ alternans lasting at least 40 activations for all conditions (Figures 3A–D lower trace, rightmost panels).

To determine the relationship between the transmembrane potential properties and the development of AP alternans in the various experimental groups, we tabulated and compared the average values of the Mean-DI and of the Mean-APD₉₀ at a PCL of 500 ms, a rate close to the intrinsic heart rate of rabbits (normocardia) and above the ALT-TH. Summary data in Table 1 shows that Mean-APD₉₀ at 500 ms PCL (normocardia) was 252 ± 53 ms in controls and similar in cells treated with Bay K 8466 (296 ± 65 ms, p = 0.059) (Table 1). Note, an effect of Bay K 8466 was observed as evidenced by Mean APD₉₀ prolongation from 270 ± 107 ms to 323 ± 116 ms (p < 0.0001) during bradycardic pacing (PCL = 1000 ms). Both 10 mM BAPTA and 20 mM BAPTA increased normocardic Mean-APD₉₀ to 367 ± 56 ms and 337 ± 43 ms respectively (p < 0.0005, Table 1).

We next quantified the ALT-TH as previously described (Pruvot et al., 2004). In Controls, the ALT-TH was 229 ± 36 ms (Table 1). Consistent with the lack of effect of Bay K 8466 on normocardic Mean-APD₉₀, the I_CaL agonist did not significantly increase the ALT-TH relative to Controls (262 ± 72 ms, p = 0.16) (Table 1). In contrast, AP alternans were induced at longer PCLs in the absence of the CaT. Specifically, the average ALT-TH in cells perfused with 10 mM BAPTA and 20 mM BAPTA were both significantly greater than the ALT-TH for Controls (288 ± 91 ms and 284 ± 50 ms; p = 0.045 and p = 0.0065 respectively) (Table 1).

At the ALT-TH, Mean-APD₉₀ was yet again not significantly different between Control (182 ± 34 ms) and cells perfused with Bay K 8466 (206 ± 60 ms, p = 0.24) (Table 1). In contrast, the average Mean-APD₉₀ at the ALT-TH was significantly increased by 10 mM BAPTA (240 ± 79 ms, p = 0.029) and 20 mM BAPTA (232 ± 44 ms, p = 0.007), consistent with the significantly greater ALT-TH observed in both CaT buffered groups (Table 1).

To further assess the nature of the AP instabilities developing at the ALT-TH, we quantified the number of consecutive APD₉₀ alternans pairs developing in the 40 activation-long window corresponding to the ALT-TH. Of note, we did not detect a significant difference between the number of consecutive 2:2 AP’s identified in Controls and cells treated with 25 nM Bay K 8466 (p = 0.99), 10 mM BAPTA (p = 0.42), or 20 mM BAPTA (p = 0.42) (Table 1). Additionally, the alternans magnitude (ALT-MAG) measured at the ALT-TH did not exhibit group-wise differences (p = 0.45 for Control vs. Bay K 8644; p = 0.77 for Control vs. 10 mM BAPTA; and p = 0.54 for Control vs. 20 mM BAPTA) (Table 1). Taken together, our data indicate that agonizing I_CaL with 25 nM Bay K 8466 does not promote significant AP alternans. However, removing CaT with BAPTA is an alternans promoting intervention.

Given that the development of AP alternans driven by AP instabilities has been tied to the slope of the APD restitution curve (Nolasco and Dahlen, 1968; Qu and Weiss, 2023), we next plotted Mean-APD₉₀ versus Mean-DI for each experiment of the Control (Supplementary Figures S3, S4), Bay K 8466 (Supplementary Figure S5), 10 mM BAPTA (Supplementary Figure S6), and 20 mM BAPTA (Supplementary Figure S7) groups. First, it should be noted that the Mean-DI at the ALT-TH was similar between interventions. Specifically, Mean-DI was 47 ± 17 ms in Control, 56 ± 14 ms with 25 nM Bay K 8466 (p = 0.21), 49 ± 13 ms with 10 mM BAPTA (p = 0.86), and 53 ± 11 ms with 20 mM BAPTA (p = 0.40) (Table 1), suggesting that a critical common ionic time-dependent event occurring during the DI may underlie AP alternans formation.

Using a first derivative approach to calculate the restitution slope at the ALT-TH for each cell, we determined that at the onset of alternans the average slope was >1 for all groups (Figure 4). While average values remained similar in the Control, Bay K 8644, and 20 mM BAPTA cells, the slope in cells buffered with 10 mM BAPTA was significantly larger than Controls (Figure 4).

With current clamping alone, we were able to measure AP alternans for significantly longer periods of time. Inspection of the AP dynamic response to a pacing rhythm change at the scale of tens-to-hundreds of sequential activations revealed that the pattern of developing alternans was diverse (see example in Supplementary Figure S8), in line with previous observations (Dilly and Lab, 1988). For the purpose of visualizing the dynamic response of alternans over extended periods of time we plotted APD₉₀ against time during uninterrupted pacing at a fixed PCL (Figure 5). Analysis of such prolonged sequences revealed that alternans could be classified into three major categories: overdamped, critically damped, and underdamped (Figure 5).

Specifically, overdamped alternans were present at a rhythm change, and disappeared (aka resolved) after a variable amount of time. A representative trace of overdamped alternans is shown in Figure 5A. On the other hand, critically damped alternans were either stable (i.e., the alternans magnitude was constant over time) or exhibited lower frequency harmonics (i.e., amplitude modulated aka quasiperiodic) as shown in Figure 5B. Similar quasiperiodic type alternans have been previously described in experimental models (Gilmour et al., 1997) and in simulations (Otani and Gilmour, 1997; Shiferaw et al., 2005). Finally, underdamped alternans were those in which the amplitude of oscillations became progressively larger (Figure 5C), ultimately prompting the conversion of a 2:2 type response (alternans) to a 2:1 type response (a consecutive captured and non-captured beat).

To determine if the formation and pattern of alternans over extended time intervals was modified by the experimental conditions, we classified each AP alternans sequence recorded for every myocyte into one of the above-mentioned categories. Analysis of summary data (Table 2) reveals that AP alternans were more frequently critically damped in alternans positive cells of both the Control (77%) and Bay K 8644 (43.75%) groups. Of note however, the proportion of alternans types in the Bay K 8644 group was significantly altered (p = 0.046), with the I_CaL agonist lowering the proportion of critically damped alternans while also increasing the proportion of underdamped alternans from 9% to 31.25% (Table 2).

On the other hand, while 10 mM BAPTA myocytes developed a nominally equal proportion of critically damped alternans and overdamped alternans (Table 2), the distribution of alternans types was not significantly different from Controls (p = 0.069). Interestingly, overdamped alternans that resolved to non-alternating rhythms during rapid pacing in myocytes perfused with 10 mM BAPTA occurred frequently (5/7 alternans positive myocytes). This finding is consistent with other reports that AP alternans are absent at steady-state pacing in the presence of BAPTA (Goldhaber et al., 2005). Importantly, the AP alternans in 7/8 alternans positive cells treated with 20 mM BAPTA were either critically damped (43.75%) or underdamped (43.75%) (Table 2), and this distribution is significantly different from that of Controls (p = 0.012).

As an additional exploration of the feedback loop between AP, CaT, and I_Ca,L, cells were first loaded with the lowest BAPTA concentration (10 mM) and subsequently perfused with 25 nM Bay K 8466. We hypothesized that the agonizing I_Ca,L in the absence of CaT would further increase cellular susceptibility to AP alternans as well as the complexity of those alternans. The results were unexpected, yet still consistent with the hypothesis. Specifically, perfusion of Bay K 8644 upon myocytes with buffered CaT prompted prominent changes in the dynamic response to pacing (Figure 6). During bradycardic constant pacing (1000 ms PCL), AP alternans were not observed with 10 mM BAPTA (Figure 6A), consistent with the lack of alternans observed during normocardic pacing. However, the addition of Bay K 8644 to the CaT free myocyte (Figure 6A, dotted line and darkest grey bar) elicited a fast and prominent increase of APD₉₀ (from ∼500 ms at baseline to ∼850 ms) followed by the formation of AP alternans (Figure 6A, stimulus #284–302). Of note, the magnitude of the alternans progressively increased and became unstable. The unrestricted increases in the magnitude of alternans during constant pacing ultimately led to activation block (i.e. 2:1 type response) on stimulus #303. APD₉₀ values were not reported during 2:1 pacing. A subsequent attempt to reduce the development of underdamped alternans by increasing the PCL to 1200 ms was successful, but upon readjusting the PCL down towards 1100 ms, an additional episode of underdamped alternans leading to activation block was triggered (Figure 6A dotted box). For better visualization of this transition, the episode of underdamped alternans enclosed in dotted box is replotted with an expanded x-axis (orange trace, inset). The final events leading to the activation block in this episode are labeled ‘a’ to ‘j’ in the zoomed-in APD₉₀ sequence, and can be traced to the corresponding dynamic transmembrane potential changes depicted in Figure 6B. Note here that the progressive increase in the magnitude of AP alternans leads to the final abortive response to pacing which effectively initiates the 2:1 response sequence (Figure 6B, last unlabeled AP complex and pacing artifact during repolarization). The above-described phenomenon occurred two additional times upon resumption of pacing (not shown). Eventually, by increasing the PCL to 1600 ms, activity devoid of 2:2 type responses was achieved. In all, similar responses to combined 10 mM BAPTA and Bay K 8644 were observed in six additional myocytes. In these myocytes, the proportion of underdamped alternans increased prominently from 9% in control to 79% (p < 0.0001) (Table 2). Importantly, the ALT-TH in the CaT buffered myocytes subject to Bay K 8644 was markedly increased to 990 ± 338 ms (p < 0.0002 compared to control) (Figure 6C). Taken together, agonizing I_Ca,L is not proarrhythmic on its own. Importantly, agonizing I_Ca,L in the absence of CaT is proarrhythmic at a cellular level.

From a theoretical point of view, continuous shifts back and forth from an electromechanically concordant to an electromechanically discordant AP and CaT type relationship have been implicated in the development of quasiperiodic alternans (Shiferaw et al., 2005).

See Figure 7 for a quantification of the relationship between APD₉₀ and CaT during alternans formation in various examples. Note from the plots in Figure 7 depicting percentage changes of AP and CaT parameters measured on a beat-to-beat basis, that some of the examples (#1, #2, and #3) exhibited a clear demarcation between the two groups of data points forming part of the scatter plots of AP vs. CaT. From a theoretical standpoint (Shiferaw et al., 2005), this distribution is consistent with positive AP-to-Ca²⁺ coupling, negative Ca²⁺-to-AP coupling, and an “instability dominated by voltage”. On the other hand, the scatter plots from a different set of examples (#4, #9, and #10), exhibit a smearing of the distribution of data points in the scatterplot. From a theoretical standpoint (Shiferaw et al., 2005), this type of distribution is partly consistent with shifting dominance in the mechanism (either voltage or Ca²⁺) driving the alternans. Such a shifting dominance scenario has been suggested to underlie the formation of quasiperiodic alternans (Shiferaw et al., 2005). However, a shifting dominance scenario would also require a shift in the AP and CaT relationship from electromechanically concordant to electromechanically discordant, which was not observed. Interestingly, inspection of AP alternans at the scale of tens-to-hundreds of sequential activations, showed that the BAPTA loaded myocytes without a detectable CaT exhibited quasiperiodic oscillations (Supplementary Figure S9; Table 2), indicating that an exclusively sarcolemmal ionic mechanism likely plays an important role in the development of quasiperiodicity.

The purpose of this study was to investigate the ionic basis of voltage driven AP alternans. Based on earlier simulation studies that highlighted I_CaL gating as a determinant of AP alternans (Fox et al., 2002), we tested the hypothesis that the CaT dampens action potential alternans. Herein we demonstrate: 1) Ca²⁺ alternans are not required for the development of AP alternans; 2) an AP alternans promoting effect of the Ca²⁺ chelator BAPTA; 3) increasing BAPTA concentrations favor the destabilization of AP alternans; 4) CaT removal in combination with the I_CaL agonist and inotropic agent Bay K 8644 can promote the formation of severely underdamped alternans leading to recursive activation block, an arrhythmia triggering event. In short, AP alternans can occur without CaT alternans, the CaT negatively feeds back to attenuate AP alternans via a mechanism involving the I_CaL. This paradigm is in contrast with a number of studies which suggest that instabilities in the CaT are the cause of AP alternans (Qu and Weiss, 2023; Kulkarni et al., 2019). Importantly, our results are in line with previous computer simulations that identified the I_CaL and its sensitivity to changes in intracellular Ca²⁺ as a promoter of steep APD restitution and AP alternans (Fox et al., 2002; Shiferaw et al., 2005). Our study is clinically relevant because it identifies a principle ionic mechanism underlying the formation of voltage driven AP alternans, the cellular driver of T-wave alternans, a known harbinger of ventricular arrhythmia. The protective role of Ca²⁺ may bring about a new paradigm for therapy.

Our results may appear to be in contrast with previous studies demonstrating that CaT removal with BAPTA protected against the development of alternans (Walker et al., 2003; Goldhaber et al., 2005). While the results of both studies are nominally opposite to ours (protective vs. detrimental) for the same concentration of BAPTA (10 mM), we additionally report that 10 mM BAPTA shifted alternans from ‘critically’ to ‘overdamped’ type alternans. Given that AP alternans may resolve over time at this degree of Ca²⁺ chelation, it is possible that the time of measurement affected the results in each of the studies. Partially supporting this hypothesis is a previous observation of transient alternans after a single premature beat and no alternans over much longer pacing durations (Goldhaber et al., 2005). An alternative and complementary hypothesis could be that we used larger step sizes when reducing PCL and did not pace below 180 ms, while others reduced PCL in 5–20 ms decrements achieving, at least in some occasion, final PCLs close to 50 ms (Goldhaber et al., 2005). The different pacing protocols could also affect the temporal formation of alternans as a result of cardiac memory, different rates of intracellular ion accumulation and loss, or other time-dependent electrophysiological properties.

Importantly, when we elevated BAPTA to 20 mM (twice the concentration of previous studies), we made a novel observation that this level of Ca²⁺ buffering promoted underdamped alternans. Under this condition, the applied PCL could not be shortened beyond the ALT-TH. In short, buffering Ca²⁺ appears to shift the complexity of alternans from critically damped, to overdamped, and at increased buffering potency to underdamped.

Aside from the origin of discrepancies amongst studies, our experiments using BAPTA to buffer the CaT demonstrate that Ca²⁺ alternans is not required for the development of rapid-pacing induced AP alternans at physiological temperatures. More importantly, the alternans threshold increased (i.e., slower rate required) when we buffered the CaT, indicating that Ca²⁺ buffering increases the propensity to form alternans. This principle finding must be viewed in tandem with earlier studies demonstrating the close to reverse scenario, where Ca²⁺ alternans can form during an alternans-free AP-clamp (Chudin et al., 1999). Others, using a similar approach, showed that CaT could develop both electromechanically concordant or discordant alternans, in a manner which was independent of AP dynamics (Kanaporis and Blatter, 2015). Taken together, the data indicate that both signals are capable of generating alternans inherently when voltage and Ca²⁺ are uncoupled.

Another method to determine whether AP or CaT are the primary driver of alternans would be to compare the ALT-TH of each signal. Our data place the AP alternans threshold of the BAPTA loaded myocyte prominently higher (∼280 ms) than that of the unclamped AP which is ∼220 ms, in accordance with previous studies (Mahajan et al., 2008). To the best of our knowledge, the equivalent measure for the CaT, the inherent Ca²⁺ alternans threshold, has not been experimentally determined in rabbit ventricular myocytes at physiological temperatures. In canine myocytes the alternans-free AP-clamped threshold for Ca²⁺ alternans was ∼258 ms (Wilson et al., 2009), which is still lower than the ∼280 ms AP alternans threshold in the BAPTA loaded myocytes of our study. Side-by-side comparison of the available information on the inherent thresholds for the AP and the CaT signals suggests that AP alternans occur at higher ALT-TH. Given the above, and recognizing the limitations of interstudy data comparisons, our data is consistent with a model in which rapid-pacing induced AP alternans is governed by a voltage driven instability which is dampened by intracellular Ca²⁺ cycling. In line with this notion, our dual AP and CaT recordings always exhibited a relationship between the signals such that long/short APD was associated with a large/small CaT, which is expected for a graded type Ca²⁺ release triggered by voltage. This type of interaction between the signals has been previously denominated a ‘positive voltage to calcium type coupling’ (Qu and Weiss, 2023).

Given that BAPTA prolonged APD both at baseline and at the ALT-TH, it is likely that the intracellular Ca²⁺ mediated -inactivation of I_CaL is the signal responsible for dampening the AP alternans. In line with this, simulations by Fox et al. showed that the magnitude and incidence of alternans could be altered by varying the time constant for Ca²⁺ induced inactivation of I_CaL (Fox et al., 2002). In addition, our observation that the higher BAPTA concentration promotes underdamped alternans indicates that the extent of Ca²⁺ buffering is an important determinant of alternans stability. Specifically, we propose that this observation may be tied to the local regulation of I_CaL by calmodulin and changes in Ca²⁺ concentration which take place very near or at the channel pore (Christel and Lee, 2012). Supporting the notion that changing BAPTA concentration from the 10 mM–20 mM may influence the formation of alternans via a change in the local regulation of I_CaL, previous studies reported that buffering provided by 10 mM BAPTA reduces, but does not abolish CaT induced inactivation of I_CaL (Sham, 1997; Dick et al., 2008).

In considering a voltage driven mechanism of alternans, the specific ion channels generating the instability that cause AP alternans remain to be further delineated. Of note, our data show that changes in the ALT-TH closely tracked the changes in APD. Given that changes in the APD are tied to reciprocal changes in the DI, the finding supports the presence of an ionic time-dependent event occurring during the DI as the cause of AP alternans. Indeed, the finding that values of DI measured at the ALT-TH was similar across experimental groups exhibiting distinct APD values at ALT-TH support this. Additionally, the fact that cells with intact CaT and cells loaded with BAPTA had similar DI values at ALT-TH suggests that the voltage instability causing alternans may be less dependent on the presence or absence of Ca²⁺. Of note also, the marked alternans promoting effect that the I_CaL agonist Bay K 8644 had on the BAPTA loaded myocytes indicates that the voltage and time-dependent properties of I_CaL may play a key role in the formation of alternans.

Figure 8 depicts schematically the proposed AP alternans dampening mechanism involving a negative feedback loop connecting the membrane and intracellular Ca²⁺ signals, for the four experimental conditions that we tested. Under this alternans-driven-by AP scheme, Ca²⁺ does not directly cause alternans but plays a critical role in the reduction and stabilization of the membrane oscillations. Following this mechanistic reasoning, myocytes with conserved CaT exposed to Bay K 8644 would be expected to exhibit a modest change in the formation of alternans, as indeed we and others observed (Saitoh et al., 1988). By the same reasoning, the absence of a dampening cue in the CaT buffered myocytes subject to Bay K 8644 underlies the dramatic formation of underdamped alternans that effectively increased the ALT-TH by fourfold in this group of cells. Given the scenario proposed in Figure 8, we further speculate that the determinants of intracellular Ca²⁺ release restitution (Cely-Ortiz et al., 2020) are key in modulating AP alternans through dynamic dampening of the AP oscillations. Experiments designed to measure the inherent capability of CaT to develop alternans at physiological temperatures as well as tandem measurements during transition to alternans will be required to further clarify the mechanism driving AP alternans. Towards this, continuous tandem recordings of Vm and CaT during the transition to alternans carried out at room temperature showed that the onset of APD alternans coincided in time with the onset of CaT alternans (Kanaporis and Blatter, 2015). In contrast, simultaneous optical recordings of AP and CaT during rapid pacing induced alternans in intact rabbit hearts showed a temporal hierarchy in the development of alternans with CaT amplitude alternans occurring before APD alternans (Visweswaran et al., 2013). Others, however, reported the simultaneous initiation of both APD and CaT alternans in dog hearts (Wilson et al., 2009). Given the temporal and spatial limitations of optical mapping, particularly the inability to resolve where in 2-dimensional tissue a phenomenon starts, isolated myocytes may be useful towards determining if any of the signals is the alternans precursor.

Towards determining how removing the CaT affects the formation of alternans, others used either caffeine (Kihara and Morgan, 1991; Saitoh et al., 1989; Hirayama et al., 1993), ryanodine (Kihara and Morgan, 1991; Saitoh et al., 1989; Kanaporis and Blatter, 2015), thapsigargin (Pearman et al., 2018), or ryanodine in combination with thapsigargin (Goldhaber et al., 2005). All listed approaches, which were reported as alternans protective, hinge on preventing sarcoplasmic reticulum Ca²⁺ accumulation, which should markedly increase cellular free Ca²⁺ in contrast to the buffering effect of BAPTA. The contrasting conclusion of whether inhibiting Ca²⁺ cycling is pro and anti-alternans may therefore be dependent on the method to attenuate CaT. It is plausible that differing free Ca²⁺ levels between approaches may result in distinct effects on the Ca²⁺ -dependent inactivation of I_CaL, with approaches increasing the cytosolic Ca²⁺ levels promoting I_CaL inactivation (protective) versus those approaches reducing cytosolic Ca²⁺ levels preventing I_CaL inactivation (detrimental). Additional studies designed to test how a range of cytosolic Ca²⁺ concentrations alter AP alternans could be insightful.