Recent experimental studies have shown that fibroblasts can electrotonically couple to myocytes via gap junctions. In this study, we investigated how this coupling affects action potential and intracellular calcium (Cai) cycling dynamics in simulated fibroblast-myocyte pairs and in two-dimensional tissue with random fibroblast insertions. We show that a fibroblast coupled with a myocyte generates a gap junction current flowing to the myocyte with two main components: an early pulse of transient outward current, similar to the fast transient outward current, and a later background current during the repolarizing phase. Depending on the relative prominence of the two components, fibroblast-myoycte coupling can 1) prolong or shorten action potential duration (APD), 2) promote or suppress APD alternans due to steep APD restitution (voltage driven) and also result in a novel mechanism of APD alternans at slow heart rates, 3) promote Cai-driven alternans and electromechanically discordant alternans, and 4) promote spatially discordant alternans by two mechanisms: by altering conduction velocity restitution and by heterogeneous fibroblast distribution causing electromechanically concordant and discordant alternans in different regions of the tissue. Thus, through their coupling with myocytes, fibroblasts alter repolarization and Cai cycling alternans at both the cellular and tissue scales, which may play important roles in arrhythmogenesis in diseased cardiac tissue with fibrosis.
fibroblasts comprise the majority of nonmyocyte cells in the heart and play diverse roles in cardiac function (2, 3, 5, 25). Structural remodeling associated with aging, ischemia, inflammation, hypertension, hypertrophy, and heart failure invariably involves cardiac fibrosis, caused by fibroblast proliferation and collagen production, and is associated with an increased risk of arrhythmias (1, 3, 4, 31, 45, 49, 52). Traditionally, fibroblasts and fibrosis have been thought to promote reentrant arrhythmias in cardiac tissue by creating obstacles that slow and block electrical pulse propagation. However, recent cell culture experimental studies (5–7, 9) have documented that fibroblasts express connexins and are able to form functional gap junctions not only with each other but also with myocytes, although the extent to which this occurs in native cardiac tissue is controversial. This raises the possibility that fibroblasts play an active role in cardiac electrophysiology beyond acting as passive electrical insulators.
The effects of fibroblasts on cardiac electrophysiology have been investigated in experimental studies of cultured monolayers (13, 32, 33, 47, 57) and computational models (19, 20, 29, 40, 41, 45, 57). Gaudesius et al. (13) showed that insertion of fibroblasts into cultured myocyte strands did not block electrical conduction but delayed conduction until a critical length of inserted fibroblasts was reached, demonstrating that myocytes and fibroblasts are electrotonically coupled and that an electrical pulse can conduct through a certain distance in fibroblasts. Both experimental and computational studies (19, 32, 57) have demonstrated that fibroblast-myocyte coupling caused nonmonotonic changes (first an increase and then a decrease) in conduction velocity (CV) as fibroblast content or coupling strength increased. It has also been shown in computational studies (19, 29, 41) that fibroblast-myocyte coupling modulated action potential morphology and action potential duration (APD), depending on the resting potential of the fibroblasts. Fibroblast-myocyte coupling could also depolarize myocyte membrane potential sufficiently to induce automaticity (33).
In this study, we used mathematical modeling and computer simulation to explore the mechanisms by which fibroblast-myocyte coupling affects repolarization and intracellular calcium (Cai) alternans dynamics. Our study shows that fibroblast coupling promotes both repolarization and Cai cycling alternans as well as novel alternans dynamics at the cellular and tissue scales. Since repolarization alternans and Cai alternans have been mechanistically linked to cardiac arrhythmogenesis (8, 12, 34, 36, 39), the latter findings may be relevant to the increased arrhythmia risk in heart disease (12, 35, 54), in which structural remodeling promotes increased fibrosis.
The membrane voltage (V) of a myocyte (Vm) or a fibroblast (Vf) in coupled myocyte-fibroblast pairs (Fig. 1, A and B) or in tissue (Fig. 1C) is governed by the following: (1) where C is the membrane capacitance of a myocyte (Cm) or a fibroblast (Cf), t is time, Iion is the corresponding membrane current [of a myocyte (Im) or a fibroblast (If)], n is the number of coupled neighbors (either myocytes or fibroblasts), and G is the gap junction conductance between a cell (either a myocyte or a fibroblast) and its kth neighbor (either a myocyte or a fibroblast). The size of the myocyte was set to 125 × 25 × 25 μm, with Cm = 125 pF. For Im, we used two different action potential models to distinguish between the effects of fibroblast coupling arising purely from membrane ionic currents and those arising from the bidirectional interaction between membrane ionic currents and Cai cycling. For the former model, we used phase I of the Luo and Rudy (LR1) model (28) with modifications in parameters (see Supplemental Material for details),1 whose action potential properties (including APD alternans) are mediated exclusively by voltage since Cai merely passively follows the calcium current. For the latter model, we used the rabbit ventricular action potential model developed by our group (30), with detailed Cai cycling that is bidirectionally coupled to voltage, as in the realistic physiological setting. The size of the fibroblast was set to 25 × 25 × 25 μm. For If, we used a “passive” model (26), as follows: (2) where Gf is the membrane conductance of a fibroblast and Ef is the resting potential of a fibroblast. We also carried out simulations using an “active” model developed by MacCannell et al. (29) for comparison; these results are described in the Supplemental Material. We used the passive model as our primary model for the following reasons: 1) Gf and Ef could be altered independently to explore their effects on APD and alternans dynamics; 2) the experimentally measured range of Ef varies from −50 to 0 mV (5, 18, 22, 23, 38), which can be directly adjusted in the passive, but not the active, model; and 3) as we show in this study, the two models give rise to the same results qualitatively. In addition, many of the electrophysiological properties of the fibroblast are still unknown; therefore, the use of a phenomenological (passive) model is reasonable for qualitative experiments. In this study, we fixed Cf = 25 pF, which is in the experimentally estimated range of 6.3–75 pF (29, 48). We varied Gf [range: 0.1–4 nS based on experimental estimates (26)] and Ef to explore the effects of fibroblasts on the action potential dynamics of the myocyte.
In the two-dimensional (2-D) ventricular tissue model, fibroblasts were randomly inserted between myocytes, as shown in Fig. 1C. To generate this tissue model, we first generated a mesh with a spatial resolution of 25 × 25 μm. A fibroblast accounts for one such box, whereas a myocyte accounts for five boxes in a row. Once a fibroblast-myocyte ratio (α) is assigned, then the probability of a box being a fibroblast can be determined as α/(α + 5), and a random number is generated to determine if the present box is a fibroblast or not. If the box is not a fibroblast, then an additional four consecutive boxes in the row will be used to form a myocyte (composed of five boxes). This procedure is repeated until all boxes are assigned as either fibroblasts or myocytes. A tissue model with diffuse fibroblast distribution, as shown in Fig. 1C, is then generated. The gap junction conductance between myocytes was set to 600 nS when two myocytes were coupled end to end and 1,000 nS when they were fully overlapped side to side. When two myocytes were partially overlapped side to side, gap junction conductance was proportional to the overlap. We used a gap junction conductance in the same range as measured by Yao et al. (55), which averaged 588 ± 104 nS (range: 226–981 nS) for side-to-side coupling and 558 ± 92 nS (range: 262–937 nS) for end-to-end coupling. These values of gap junction conductance gave rise to a longitudinal CV of 0.56 m/s and a transverse CV of 0.13 m/s when no fibroblasts were present. The gap junction conductance between a myocyte and fibroblast or between two fibroblasts is denoted as Gj in this article. Rook et al. (38) reported that Gj ranged from 0.3 to 8 nS in cultured cells, whereas 0 < Gj < 100 nS was used in other modeling studies (19, 29, 41). In this study, we used 0 ≤ Gj ≤ 20 nS. When Gj = 0, the fibroblasts are not coupled with myocytes and are simply passive obstacles.
Measurements of APD and CV.
Since fibroblasts elevate the resting membrane potential of the coupled myocyte and change the action potential profile, APD was measured using a voltage threshold that was 10% above the resting potential (Vmr), i.e., VAPD = 0.9Vmr. CV was defined as the tissue length divided by the time for the action potential to conduct from the left end to the right end of the tissue.
Fibroblast-myocyte coupling gives rise to an early transient outward current and a late background current for the myocyte.
When either the LR1 model or the rabbit ventricular model is coupled to a fibroblast, the resting potential of the fibroblast becomes markedly hyperpolarized and that of the myocyte slightly elevated. The extent of these alterations depends on the membrane conductance of the myocyte, Gf, and Gj (see Supplemental Material), as also shown in a previous study by Jacquemet and Henriquez (19). For example, when we coupled the LR1 model to a fibroblast with a low Gf, the resting potential of the fibroblast was only slightly higher than that of the myocyte (Fig. 2A). After a stimulus was applied to the myocyte, its voltage increased from around −80 mV to above 0 mV in <1 ms. However, since the inexcitable fibroblast is equivalent to a leaky capacitor, it took several milliseconds for the myocyte to charge the fibroblast and elevate its voltage from −80 mV to close to 0 mV. Due to the time delay, the initial voltage difference between the myocyte and fibroblast was very large, giving rise to a large gap junction current (Igap), which flowed from the myocyte to the fibroblast (Fig. 2B). After that, the voltage of the fibroblasts was close to that of the myocyte, and Igap became small. Whether this late component of Igap flows from the myocyte to the fibroblast or vice versa depends on the properties of the fibroblast. Since the early phase of Igap is a short pulse and outward for the myocyte, it resembles the fast transient outward current (Ito) in cardiac myocytes (15, 56), which is most prominent in the epicardial layer (27, 56). Figure 2, C and D, shows If for Gf = 0.1 and 4 nS, which are inward currents during the diastolic phase but outward during the systolic phase. Thus, rather than a pure passive load or driver, the fibroblast acts like a leaky capacitor, charging during the resting potential and upstroke, and then leaking current throughout the action potential plateau of the myocyte.
Using a voltage-clamp model of the myocyte, we can analytically derive the decay time of the pulse [time constant (τ)], which is τ = Cf/(Gf + Cj) (see Supplemental Fig. S1 and the Supplemental Material). The peak of Igap is insensitive to Gf and increases with Gj almost linearly unless Ef is high (Fig. 2E). In the late phase, Igap can be approximated as follows (see the derivation of Eq. S27 in the Supplemental Material): (3) where Vc is the voltage at which Vm crosses Vf, defining the point at which Igap changes from an outward current (tending to shorten APD) to an inward current (tending to lengthen APD). Equation 3 agrees well with the simulation results of the LR1 and rabbit ventricular models coupled to the passive fibroblast model (Fig. 2F). Similar linear relations occurred when the LR1 and rabbit ventricular models were coupled to the active fibroblast model (Supplemental Fig. S1). Note that in the repolarizing phase, Igap is proportional to both Gj and Gf.
These two characteristics of Igap give rise to different and nontrivial action potential dynamics. In the following sections, we show how the fibroblast-myocyte coupling affects APD, APD alternans, and Cai alternans at both the cellular and tissue scales.
Fibroblast-myocyte coupling alters APD.
Previous modeling studies (19, 29, 41) have shown that fibroblast-myocyte coupling can either prolong or shorten APD, depending on the specific ionic models of myocytes and fibroblasts, their membrane and coupling parameters, and also on the threshold used to define APD. However, the general mechanisms by which APD is affected by fibroblast-myocyte coupling are not clear.
Here, we used the APD at 90% repolarization (see methods for the definition) and scanned the Ef from −80 to 0 mV and Gf from 0 to 4 nS to show how APD is altered in a coupled cell pair of a passive fibroblast and the LR1 myocyte model. APD was shortened only when Ef was low and Gf was high (Fig. 3A). Note that APD was still prolonged even for an (unphysiological) value of Ef = −80 mV, close to the resting potential of the myocyte.
Based on the properties of Igap, we can understand the general mechanisms by which fibroblast-myocyte coupling affects APD as follows. When Gf is small, Igap in the repolarizing phase is small and becomes inward at a Vc, regardless of Ef, tending to prolong APD. The early phase of Igap is similar to the fast Ito, which lowers the action potential notch and prolongs APD in the LR1 model, as also shown in other models (15). Therefore, APD always prolongs when Gf is small irrespective of Ef. When Gf is large, however, the late phase of Igap is large and exerts a major effect on APD. Since Vc ∼ Ef for large Gf, then if Ef is high, this current is mainly inward and lengthens APD; however, if Ef is low, this current is mainly outward and thus shortens APD. This explains why APD only shortens when Gf is large and Ef is low (see Fig. 3A, top left corner).
When the rabbit ventricular model was coupled to the passive fibroblast model, we observed exactly the same behaviors as in the LR1 model (Supplemental Fig. S2). Similar results were also obtained when the LR1 model or rabbit ventricular model was coupled to the active fibroblast model formulated by MacCannell et al. (29): APD shortened when the resting potential of the fibroblast was low but prolonged when the resting potential was high, and Igap was similar to that of using the passive fibroblast model (Supplemental Fig. S3).
Fibroblast-myocyte coupling modulates voltage-driven APD alternans.
Since fibroblast-myocyte coupling alters APD, it will also change the critical pacing cycle length (PCL) at which alternans occurs. For small Gf, APD prolongs and APD alternans occurs at a longer PCL, whereas for large Gf, APD shortens and APD alternans occurs at a shorter PCL (Fig. 3B). We scanned Ef from −80 to 0 mV and Gf from 0 to 4 nS for the difference in critical PCL (ΔPCLc) from that of the control myocyte (Fig. 3C), which was similar to the pattern of ΔAPD in Fig. 3A. Note that the change in PCLc was much larger than the change in APD (ΔPCLc ranged from −107 to +60 ms in Fig. 3C, whereas ΔAPD ranged from −45 to +22 ms in Fig. 3A). Therefore, the change in the alternans onset is due not just to the prolongation or shortening of APD. To understand this difference, we plot APD restitution curves for the control, small, and large Gf in Fig. 3D and the corresponding slopes in Fig. 3E. When APD was prolonged, the APD restitution curve became steeper, and when APD was shortened, the APD restitution curve became less steep. APD alternans is due to the steep slope of the APD restitution curve in this model, which explains why the change in the alternans onset PCL is much larger than the change in APD. The same effects were observed when the active fibroblast model was coupled to the LR1 model (Supplemental Fig. S4).
Fibroblast-myocyte coupling results in novel APD alternans dynamics.
In addition to altering the onset of APD alternans during pacing at fast heart rates, fibroblast-myocyte coupling causes a new dynamical instability that leads to alternans at slow heart rates. For the small Gf case, when the number of fibroblasts attached to a myocyte was increased (equivalent to scaling up Cf, Gf, and Gj for a single virtual fibroblast), the notch of the action potential became progressively lower, and APD first prolonged and then shortened as the number of attached fibroblasts increased (Fig. 4, A and B). The curve relating APD to the number of attached fibroblast was not smooth but shortened suddenly at two critical points (Fig. 4B). These features are similar to fast Ito, which first increases APD and then shortens APD as its conductance increases (10, 15, 17), ultimately causing “all-or-none” repolarization. The mechanism (15) is that Ito first lowers the action potential notch, increasing the driving force for inward current through L-type calcium channels and delaying the activation of time-dependent potassium currents, thus lengthening the APD. As Ito becomes larger, however, the early repolarization phase interferes with the full activation of the L-type calcium current (ICa,L), leading to action potential shortening and, with further increases, to all-or-none repolarization. Since for small Gf the dominant component of Igap is the early Ito-like phase, it is not surprising that coupling more fibroblasts to the myocyte has the same effect as increasing the conductance of Ito.
The effects of increasing the number of attached fibroblasts on the onset of APD alternans are shown in Fig. 4C. PCLc increased dramatically and then suddenly dropped as the number of attached fibroblasts increased. In addition to alternans at short PCLs (∼200 ms), the bifurcation diagram shown in Fig. 4D (see also the representative voltage trace at PCL = 500 ms in Fig. 4E) demonstrates that a new APD alternans region occurs at long PCLs. Unlike APD alternans at short PCLs, which in the LR1 model is due to the slope of the APD restitution exceeding 1 (Fig. 3), the slope of APD restitution is <1 during APD alternans at long PCL. Since Igap is similar to a fast Ito, the mechanism of alternans is likely to be the same as demonstrated in a recent modeling study by Hopenfeld (17), who showed that the interplay between ICa,L and Ito can give rise to a novel voltage-driven mechanism of APD alternans and complex APD dynamics. This same effect of Ito on alternans dynamics at slow heart rates has been shown in experiments by Lukas and Antzelevitch in canine hearts during ischemia (27).
Fibroblast-myocyte coupling promotes Cai alternans.
In the above simulations, we used the LR1 model to characterize the direct influence of Igap on membrane ionic currents during the action potential in the absence of complicating effects of Cai, since this model does not contain detailed Cai cycling. In real myocytes, however, Cai cycling is a major factor driving APD alternans (14, 51). To investigate how the addition of Cai cycling affects alternans, we used the rabbit ventricular model in place of the LR1 model. When fibroblasts were coupled to the rabbit ventricular model, the effects of Igap on action potential morphology were generally similar to those in the LR1 model (Supplemental Fig. S2). Differing from the LR1 model, however, fibroblast coupling promoted alternans whether the APD was prolonged (small Gf) or shortened (large Gf) (Fig. 5). This is not completely surprising, since in the rabbit ventricular model, the onset of alternans is primarily driven by a Cai cycling instability rather than APD restitution slope (30). Under control conditions in the rabbit ventricular model, alternans occurred at PCLc = 250 ms (Fig. 5A) and was electromechanically concordant, i.e., a long APD was associated with a large Cai transient, and vice versa. When coupled to two fibroblasts of Gf = 1.2 nS, alternans occurred at a longer PCLc of 300 ms (Fig. 5B) and remained electromechanically concordant. When coupled with two fibroblasts of Gf = 0.1 nS, however, alternans occurred at PCLc = 305 ms and was electromechanically discordant, i.e., a long APD is associated with a small Cai transient, and vice versa (Fig. 5C). The alternans remained discordant until PCL = 272 ms and then became concordant.
Since Ito augments the Cai transient amplitude by increasing the driving force through ICa,L during the early action potential (16, 30, 42, 43), we examined whether the Ito-like early phase of Igap accounted for the promotion of alternans. To increase the amplitude of the early Ito-like phase of Igap, we increased Gj and then examined how this affected PCLc (Fig. 5D). As Gj was increased, PCLc for the onset of alternans first increased from 250 to ∼400 ms and then decreased for Gf = 0.1 nS. If Gf was larger (1.2 nS), however, PCLc increased and saturated, without a decrease. When the conductance of the fast Ito (Gto,f) in the rabbit ventricular model (Gto,f = 0.11 mS/cm2 in the original model) was increased in the absence of fibroblast coupling, the PCLc for alternans increased (Fig. 5E). When we fixed Gto,f = 0.15 mS/cm2 and plotted APD and peak Cai versus PCL, the bifurcation diagrams (Fig. 5F) were very similar to the ones with fibroblast-myocyte coupling (Fig. 5C). Therefore, by comparing the two sets of simulation studies, we conclude that the Ito-like effect of the fibroblast-myocyte coupling may cause the promotion of Cai alternans. However, in the case of fibroblast-myocyte coupling, increasing Gj also increases the late phase of Igap, which may account for the difference between the two sets of simulations. Although we have not pinpointed the exact mechanism that Ito or the Ito-like effect due to fibroblast-myocyte coupling promotes Cai alternans, Jordan and Christini (21) have shown that action potential morphology modulates the onset of Cai alternans, which may explain this effect.
Fibroblast-myocyte promotes spatially discordant alternans in tissue.
In cardiac tissue, APD alternans can be spatially concordant or discordant (34, 36), with the latter being more arrhythmogenic (34). Since CV restitution (the dependence of CV on heart rate) has been shown to be important in the development of spatially discordant alternans (36), we examined the effects of fibroblast-myocyte coupling on CV restitution. In homogenous 2-D tissue using the LR1 model, as shown in Fig. 1B, CV was ∼0.56 m/s without fibroblasts. When fibroblasts were randomly inserted at a 1:1 ratio with myocytes, the CV longitudinal to fiber direction decreased to ∼0.33 m/s for either coupled or uncoupled fibroblasts, whereas CV in the transverse direction decreased from 0.13 to 0.1 m/s. CV started to decrease at longer PCLs when fibroblasts were coupled with the myocytes than when fibroblasts were uncoupled or absent (Fig. 6A), indicating that fibroblast-myocyte coupling broadens the PCL range over which CV changes. Under control conditions without fibroblasts (Fig. 6B, solid squares) or with randomly inserted uncoupled fibroblasts (Fig. 6B, open squares), the PCLc at the onset of alternans was similar. When the fibroblasts (with a small Gf) were coupled, however, alternans occurred at a longer PCLc, as in the single cell case. In the 2-D tissue, without fibroblasts, a single nodal line formed beyond the middle of the tissue when PCL = 280 ms (Fig. 6C). With uncoupled fibroblasts inserted at 1:1 fibroblast-to-myocyte ratio, a single nodal line formed closer to the pacing site (Fig. 6D), consistent with the principle that slowing CV effectively increases the tissue size. With fibroblasts coupled to myocytes, however, two nodal lines formed in the tissue and the amplitude of alternans was increased (Fig. 6E). Thus, fibroblast-myocyte coupling promotes spatially discordant APD alternans by slowing CV and broadening CV restitution.
To examine how Cai cycling and electromechanical concordance/discordance affects spatially discordant alternans, we simulated 2-D tissue using the rabbit ventricular model, with randomly inserted passive fibroblasts with small Gf (0.1 nS) at a fibroblast-to-myocyte ratio of 2:1. At a PCL of 200 ms, for which alternans was electromechanically concordant (Fig. 5C), spatially discordant APD and Cai alternans formed and electromechanical concordance was maintained (Fig. 7A). The formation of spatially discordant alternans in this case was due to the engagement of CV restitution. In the same tissue at a longer PCL of 290 ms, in which CV restitution was not yet engaged, APD alternans remained spatially concordant but was electromechanically discordant (Fig. 7B). An interesting case occurred at an intermediate PCL of 260 ms, in which the fibroblast-to-myocyte ratio was increased linearly from 1:9 to 2:1 from left to right. This produced electromechanically concordant alternans at both the left and right sides but electromechanically discordant alternans at the middle, resulting in spatially discordant APD alternans yet spatially concordant Cai alternans (Fig. 7C). This is a novel mechanism for the formation of spatially discordant alternans, consistent with the observations of Wilson et al. (53, 54), who showed that regional differences in electromechanical concordance and discordance in heart failure led to spatially discordant alternans.
In this study, we analyzed the mechanisms by which fibroblast-myocyte coupling modulates action potential morphology, repolarization, and Cai cycling and showed how these effects to give rise to novel alternans dynamics at both the cellular and tissue scales. Contrary to the conventionally held viewpoint that fibroblast coupling is largely equivalent to attaching a simple passive load or diver to a myocyte, we show that the effects are more complex, due to the combination of two effects: the early Ito-like component of Igap related to charging of the fibroblast's membrane capacitance, followed by the late sustained component due to the current leak through the fibroblast's membrane conductance. Since the amplitudes of these two components vary markedly depending on the specific fibroblast membrane properties and coupling parameters, a richly complex interaction results. The mathematical analysis and findings, obtained using several different models, reveal generic mechanistic insights into how fibroblast-myocyte coupling amplifies the proarrhythmic effects of fibrosis, which is a common factor in many forms of heart disease.
Effects on repolarization and Cai dynamics.
As we show in this study, a fibroblast coupled to a myocyte is equivalent to a leaky capacitor generating an Igap with two major components. The initial component is the capacitive charging current, which is outward and largest when the myocyte voltage changes rapidly (i.e., during the action potential upstroke), equivalent to a transient outward current, such as Ito. The second component is equivalent to a nonselective background current whose amplitude is proportional to the leakiness of the capacitor (Gf). Based on our analysis, the early Ito-like current decays exponentially with τ = Cf/(Gf + Gj). Rook et al. (38) reported that Gj ranges from 0.3 to 8 nS in cultured cells, whereas Cf ranges from 6.3 to 75 pF (29, 48) and Gf ranges from 0.1 to 4 nS (26). Using these estimates, τ can potentially range from 0.5 to 190 ms, so that a value on the order of 10–20 ms, similar to the decay time of the fast Ito (15), is reasonable. For small Gf in which the early phase of Igap is large and the late phase is small, the effects of fibroblast coupling on action potential dynamics are very similar to those of Ito, as we have shown by direct comparison. In contrast, when Gf is large, the second, later phase of Igap (equivalent to the capacitor becoming more leaky and unable to hold charge) becomes important. If Ef is low (e.g., −50 mV), the current is mainly outward and shortens APD. If Ef is high (e.g., −10 mV), the current during the later phase of repolarization becomes inward sooner and prolongs APD.
The effects of fibroblast-myocyte coupling on APD restitution parallel to the effects on APD, with APD restitution becoming less steep when APD is shortened and steeper when APD is prolonged. These effects, in turn, cause the onset of restitution-driven APD alternans to occur at a shorter PCL when APD is shorter and at a longer PCL when APD is prolonged, although these effects are more pronounced than can be explained solely by the degree of change in APD. Due to the fast transient component of Igap, a novel type of APD alternans that occurs at slow heart rates is observed, similar to the alternans caused by Ito shown in experimental (27) and modeling (17) studies.
In addition to the effects on voltage-driven APD alternans, we also found that fibroblast-myocyte coupling promoted Cai alternans in the rabbit ventricular model, irrespective of whether it prolonged or shortened APD. The mechanisms of Cai-driven alternans have been analyzed in many previous studies (11, 37, 44, 46). The exact dynamical mechanism by which fibroblast-myocyte coupling promoting Cai alternans is not completely clear, but the early Ito-like phase of Igap appears to play a key role, since increasing Ito in the rabbit ventricular model had the same dynamical effects (Fig. 5). By potentiating the driving force for calcium entry through the ICa,L during the action potential notch, Ito has been shown to significantly augment the Cai transient amplitude (16, 42, 43), which is replicated in the rabbit ventricular action potential model (30). Presumably, Cai alternans potentiation by Ito is related to this effect. Similarly, the ability of Igap to induce electromechanically discordant alternans was also reproduced by increasing Ito. Electromechanically discordant alternans has been experimentally observed in heart failure (53), in which Ito is typically downregulated, so that fibroblast-myocyte coupling may provide an alternative explanation.
Fibroblasts promote the dispersion of refractoriness in cardiac tissue.
Because fibroblast proliferation and fibrosis in the diseased heart are inherently heterogeneous, the local effects of fibroblast-myocyte coupling on APD and APD restitution will promote increased dispersion of refractoriness. In addition, dispersion of refractoriness can be dynamically generated even in homogeneous tissue by spatially discordant APD and Cai alternans, which we found is also promoted by myocyte-fibroblast coupling through at least two mechanisms. First, fibroblasts broaden the heart rate range over which CV changes, a well-known factor promoting spatially discordant alternans (36). Second, as shown in Fig. 7C, spatial heterogeneity in the fibroblast density can cause alternans to shift from electromechanically concordant to electromechanically discordant alternans over space. A novel mechanism of spatially discordant alternans arising from regional electromechanical concordance/discordance may provide an explanation for the experimental observations of Wilson et al. (53) in an animal model of heart failure, without needing to postulate that primary myocyte remodeling caused the observed regional differences in electromechanical concordance/discordance. Rather, the regional fibroblast density may have been the critical factor.
Several limitations of this study should be mentioned. Although different fibroblast models have been developed (20, 29, 41), there is limited direct information on fibroblast electrophysiology, such that many elements in the existing models have been formulated phenomenologically. On the other hand, by using multiple models, we demonstrated that the effects of fibroblast-myocyte coupling on APD and alternans dynamics identified here are generic rather than model specific, leading us to believe that the findings are relevant to real cardiac tissue. Regarding the myocyte action potential models, we considered exploring the pure voltage-mediated effects of fibroblast-myocyte coupling using the rabbit ventricular action potential model with its Cai buffered for comparison with the normal (Cai unbuffered) condition. However, Cai buffering with EGTA or BAPTA in rabbit ventricular myocytes has been reported to flatten APD restitution and suppress APD alternans (14), making this option untenable. Therefore, we used the LR1 model as an alternative, despite its simplicity. However, the generally similar effects of fibroblast coupling in both models strengthens the argument that the findings are not likely to be highly model dependent. Another limitation involves the tissue models. As shown by Kohl and Camelliti (24), there are at least three different types of fibroblast-myocyte coupling in cardiac tissue (zero-, single- and double-sided connections), and the number and size of attached fibroblasts are likely to be highly variable, which may form heterogeneous patterns of fibroblast distribution. In addition, whereas fibroblasts are not excitable, they do respond to various neurohumoral factors and manifest their own Cai cycling properties. We did not considered these factors here because information about their electrophysiological responses is limited. Finally, the most controversial aspect relates to whether fibroblasts form functional gap junctions with myocytes in intact cardiac tissue, as they do in fibroblast-myocyte cocultures. In normal hearts, fibroblasts comprise the majority of nonmyocyte cells, accounting for up to two-thirds of all cells in the heart (2, 5). However, direct experimental evidence on fibroblast-myocyte coupling is scarce in the intact working myocardium, having only been demonstrated in the sinoatrial node (7). However, it has been speculated that pathophysiological conditions may promote coupling (50), which is indirectly supported by a recent study (47) showing that cardiac fibroblasts isolated from ischemic hearts expressed significantly more connexin40 and connexin43 than those isolated from normal hearts.
In summary, despite some limitations, our present study provides novel mechanistic insights into how the early and late phases of Igap generated by fibroblast-myocyte coupling affect cardiac action potential dynamics in a complex manner and thereby contributes to our understanding of the mechanisms by which fibroblasts may be proarrhythmic in the diseased heart.
This work is supported by National Heart, Lung, and Blood Institute Grant P01-HL-078931 and Laubisch and Kawata Endowments.
↵1 Supplemental material for this article is available at the American Journal of Physiology-Heart and Circulatory Physiology website.
- Copyright © 2009 the American Physiological Society