Optogenetic current in myofibroblasts acutely alters electrophysiology and conduction of co-cultured cardiomyocytes

Interactions between cardiac myofibroblasts and myocytes may slow conduction and generate spontaneous beating in fibrosis, increasing the chance of life-threatening arrhythmia. While co-culture studies have shown that myofibroblasts can affect cardiomyocyte electrophysiology in vitro, the extent of myofibroblast-myocyte electrical conductance in a syncytium is unknown. In this neonatal rat study, cardiac myofibroblasts were transduced with Channelrhodopsin-2, which allowed acute and selective increase of myofibroblast current, and plated on top of cardiomyocytes. Optical mapping revealed significantly decreased conduction velocity (− 27 ± 6%, p < 10–3), upstroke rate (− 13 ± 4%, p = 0.002), and action potential duration (− 14 ± 7%, p = 0.004) in co-cultures when 0.017 mW/mm2 light was applied, as well as focal spontaneous beating in 6/7 samples and a decreased cycle length (− 36 ± 18%, p = 0.002) at 0.057 mW/mm2 light. In silico modeling of the experiments reproduced the experimental findings and suggested the light levels used in experiments produced excess current similar in magnitude to endogenous myofibroblast current. Fitting the model to experimental data predicted a tissue-level electrical conductance across the 3-D interface between myofibroblasts and cardiomyocytes of ~ 5 nS/cardiomyocyte, and showed how increased myofibroblast-myocyte conductance, increased myofibroblast/myocyte capacitance ratio, and increased myofibroblast current, which occur in fibrosis, can work in tandem to produce pro-arrhythmic increases in conduction and spontaneous beating.

www.nature.com/scientificreports/ Furthermore, the extent of electrical coupling between MFBs and CMs has only been measured between cell pairs 13,14 , which differs from the situation in vivo in that the cells are sparse and interact along a narrow interface. In this study, MFBs were plated on top of CMs, so that their interactions with CMs occurred over a large area in 3-D, and were transduced with Channelrhodopsin-2 (ChR2), a relatively non-selective cation channel that opens in response to light 15 , to acutely depolarize them. This MFB-specific perturbation enabled the study of acute effects of MFB depolarizing current on the macroscopic electrophysiological properties of syncytia containing co-cultured MFBs and CMs, and was used in tandem with computational modeling to estimate MFB-CM electrical conductance. The model was then used to better understand the mechanism of these effects and to show how changes in MFB-CM conductance, ratio of MFB to CM capacitance, and level of endogenous MFB currents may increase arrhythmia in fibrosis. These results have been reported in large part in a preprint of this work 16 .

Results
Co-culture of cardiomyocytes with ChR2-transduced myofibroblasts. To assess whether inward current in MFBs can alter CM electrophysiology at a tissue level, neonatal rat cardiac fibroblasts were transduced with ChR2 and differentiated into MFBs by treatment with TGF-β1, then plated on top of neonatal rat ventricular CM monolayers. Confocal imaging of MFBs with CMs demonstrated continued expression of α-SMA by MFBs two days after plating on CMs and concomitant cessation of TGF-β1 treatment ( Fig. 1A-C). Wide-field (~ 2 mm) imaging of co-cultures of ChR2-transduced MFBs (ChR2-MFBs) showed that they formed a homogeneous, dense network over a wide area of CMs (Fig. 1D) and continued to express ChR2 during co-culture (Fig. 1E). Confocal imaging showed confluent CMs ( Myofibroblast current can cause electrophysiological changes in cardiomyocytes. In ChR2-MFB co-cultures with CMs, electrically paced at 500 ms cycle length (CL, Fig. 2Ai), application of continuous blue light to open ChR2 channels could induce diastolic depolarization and spontaneous beating at a rate faster than the 500 ms paced CL (Fig. 2Aii), whereas this did not occur in MFB co-cultures with CMs. Cessation of light (and therefore ChR2 current) caused spontaneous beating to stop (Fig. 2Aiii). This spontaneous beating in response to light occurred in ChR2-MFB co-cultures, but in none of the MFB co-cultures (Fig. 2B), thus precluding thermal or other non-specific effects of the applied light. Data across multiple samples showed that 3*I 0 light (I 0 = 0.0057 mW/mm 2 , approximately the lowest light level at which effects could be seen) caused spontaneous beating faster than the 500 ms paced CL in 4/10 (four of ten) samples, and that beating CL decreased as ChR2 current increased (Fig. 2B). Under 10*I 0 light, 6/7 co-cultures beat spontaneously, and the change in CL www.nature.com/scientificreports/ was − 36 ± 18% from a paced baseline of 500 ± 1 ms (n = 7; p = 0.002 vs. MFB co-cultures, Fig. 2B). Activation maps showed spontaneous beating was focal and in 5/6 cases originated in a location different than the pacing site (See Supplementary Fig. 2 for example). Addition of ChR2-MFBs trended towards reducing CV relative to CM-only cultures during 500 ms CL pacing, although not at the level of significance (17.7 ± 5.3, n = 8 vs. 20.9 ± 4.3, n = 14; p = 0.17). For ChR2-MFB co-cultures ( Fig. 3Ai), application of light slowed CV (Fig. 3Aii), and this slowing was reversed when the light was turned off (Fig. 3Aiii). CV decreased further as light intensity was increased, until the onset of spontaneous beating prevented further comparison of CVs (since CV can change from beating rate changes alone) (Fig. 3B). There was significant slowing in ChR2-MFB co-cultures at light levels as low as I 0 (ΔCV = -12 ± 11% from a baseline of 17.7 ± 5.3 cm/s, n = 8; p = 0.01 vs. MFB co-cultures). Slowing was even greater at 3*I 0 (ΔCV = -27 ± 5% from a baseline of 14.8 ± 3.6 cm/s, n = 5; p < 10 -3 vs. MFB co-cultures, Fig. 3B).
Acute application of a ChR2-saturating level of light (1.2 mW/mm 2 , equal to 210*I 0 ) to ChR2-MFBs did not affect MFB contractility and force generation, as measured for single cells seeded on flexible micropost arrays ( Supplementary Fig. 3), excluding the possibility that inward current in MFBs caused them to contract and potentially influence CMs by activating mechanosensitive channels.
Insights from a mathematical model of co-cultures of cardiomyocytes and ChR2-transduced myofibroblasts. To better understand these results and use them to estimate MFB-CM electrical conductance, G MFB-CM , in a syncytium, which is very difficult to measure experimentally, a numerical model of a cable of 30 neonatal rat ventricular CMs connected to MFBs transduced with ChR2 current was created (Fig. 5, Supplementary Tables 2-4). Modifications were made to the neonatal rat Korhonen model 17 to allow for stimulation at 500 ms CL in the cable, as well as to match experimental data (see Supplementary Tables 2-7 for details). The changes lowered resting potential from − 67 to − 77 mV, which is similar to other published data 13,18,19 . Minimum  showed that this alternate current path had little effect on the results ( Supplementary Fig. 5). Using our model, a broad parameter space was explored, where G MFB-CM and light intensity were varied. The model showed that at light levels < 3*I 0 or G MFB-CM < 1.7 nS/CM, no spontaneous beating was produced, since either there was too little ChR2 current produced, or this current was unable to depolarize CMs, respectively ( Fig. 6A and Supplementary Fig. 6A). However, at G MFB-CM = 5.3 nS/CM (the value for which the model best fit our experimental data), spontaneous beating could exceed the pacing rate (500 ms CL) in response to a light level > 3*I 0 ( Fig. 6A and Supplementary Fig. 6A). This behavior agreed with our experiments, in which 4/10 of our samples beat spontaneously at 3*I 0 , while 6/7 samples beat faster than the paced rate when stimulated with 10*I 0 . This higher level of light reduced CL by 30% in the model, similar to our experiments in which CL was reduced by 36 ± 18% (Supplementary Figs. 7 and 8A). Furthermore, the model predicted that CMs become inexcitable at high light levels and high G MFB-CM (Fig. 6A, dark gray area), but we were unable to record transmembrane voltage at such high light levels due to optical crosstalk.
The model also showed conduction slowing with addition of MFBs, and that this slowing increased with increasing G MFB-CM or light intensity (Fig. 6B). With G MFB-CM = 5.3 nS/CM, addition of MFBs decreased CV by 3.5 cm/s (Fig. 6B, compare box marked by black dot to top row, where G MFB-CM = 0, which is equivalent to CMs-only in the model), similar to the 3.2 cm/s found experimentally (Supplementary Table 8). Application of I 0 and 3*I 0 light decreased CV by 17% and 29%, respectively (Fig. 6B, compare CV values in boxes marked by blue dots to the box marked by the black dot), similar to experiments (− 12 ± 11% and − 27 ± 5%, respectively, Supplementary Fig. 8B), while higher levels of light (blue dot at 10*I 0 ) caused loss of capture due to spontaneous beating, as discussed for Fig. 6A.
Addition of MFBs to CMs decreased APD 80 by 8 ms (Fig. 6C, black dot vs. top row, which is equivalent to the absence of MFBs), somewhat less than the decrease found experimentally (26 ms, Supplementary Table 8). Application of I 0 and 3*I 0 light decreased APD 80 by 5% and 14%, respectively (Fig. 6C, compare blue dots with black dot), matching the experimental results (− 6 ± 6% and − 14 ± 7%, Supplementary Fig. 8C). Further investigation showed that the decrease in APD 80 in response to light observed experimentally (Fig. 4) could be partly attributed to a more positive MDP and decreased action potential amplitude ( Supplementary Fig. 9A) rather than simply an increased repolarization rate, as might be interpreted from normalized optical recordings (Supplementary Fig. 9B).
Modeling showed that increased G MFB-CM caused CMs to depolarize and MFB and CM potentials to become more similar, while increased light intensity made the MDP of both more positive (Fig. 6D). At the estimated G MFB-CM and 500 ms CL pacing, MFB potential changed from − 23 mV for unconnected MFBs to − 67 mV (black   Application of light produced additional inward ChR2 current that depolarized MFBs to − 60 and − 53 mV and further depolarized CMs to -67 or -61 mV for I 0 or 3*I 0 , respectively (blue dots at I 0 and 3*I 0 , Fig. 6D). Higher light level (10*I 0 ) produced more depolarized MDPs (− 42 and − 55 mV for MFBs and CMs, respectively; rightmost blue dots, Fig. 6D-E) as well as spontaneous beating (dashed lines, Fig. 6D-E). The inward sodium current in CMs, I Na , decreased from 199 pA/pF for CM-only cables to 154 pA/pF (black dot on purple curve, Fig. 6E) for ChR2-MFB/CM cables, and to 81 and 9 pA/pF with the further addition of I 0 or 3*I 0 light, respectively (blue dots on purple curve, Fig. 6E). The reduced inward current resulted in lower CV (black to blue dots, Fig. 6F). Sodium current was highly inactivated at higher light levels, resulting in calcium-mediated conduction during spontaneous beating, since calcium current, I CaL was much less affected by changes in MDP (black to blue dots on gold curve, Fig. 6E inset). At 3*I 0 , peak I CaL was 68% of peak I Na , whereas at 10*I 0 , it was 490% peak I Na (Fig. 6E inset).
To estimate effects of changes in MFB/CM ratio, we examined the effects of changing the MFB/CM capacitance ratio, r C , in tandem with G MFB-CM , since both are expected to directly vary with MFB/CM ratio. We also modeled the effects of different levels of endogenous MFB currents, by scaling them by a factor of k I,endo (Supplementary Tables 2-3), where k I,endo = 1 represents their normal, baseline values. In doing so, we found that increasing r C reduced excitability, resulting in intermittent capture at r C = 4 (Fig. 7A). In our model, there was no spontaneous beating faster than the 500 ms paced rate at k I,endo = 1, regardless of r C ; however, it did occur at k I,endo > 2 ( Fig. 7A and Supplementary Fig. 6B). Furthermore, while a minimum r C was necessary to produce spontaneous beating, further increase of r C generally caused progressive slowing of spontaneous beating (Fig. 7A). Increasing either r C or k I,endo could slow conduction significantly (to < 8 cm/s from 21 cm/s, Fig. 7B). Also, increasing r C could either increase or decrease APD 80 , while increased k I,endo always decreased it (Fig. 7C). Comparing the effects of ChR2 current and increased endogenous MFB currents (k I,endo > 1), the model showed that the light intensities used in experiments (I 0 and 3*I 0 ) resulted in MFB-CM current during diastole equal to that produced by an 84% and 117% increase in endogenous MFB current, respectively (Fig. 7D). www.nature.com/scientificreports/ Modeling also showed that as with G MFB-CM , increased r C caused MFB and CM potentials to become more similar, while increased MFB currents made the overall MDP of both more positive (Fig. 7E). Plots of peak I Na Figure 7. Modeling effects of changes in MFB/CM ratio and MFB currents. MFB-CM conductance was scaled with MFB capacitance ratio (r C ), to model changes in MFB/CM ratio. Changes in endogenous MFB current scaling (k I,endo , per MFB capacitance) were also modeled. Color maps show beating cycle length (A), conduction velocity (B), and APD 80 (C), at different MFB-CM conductances and light intensities. Parameters are varied in third-log 10 increments around the baseline values. Gray region in A denotes that spontaneous beating, if present, was slower than the 500 ms paced CL. CV and APD 80 were not calculated for the purple region in B and C since spontaneous beating prevented capture at 500 ms CL pacing. Light gray region denotes that beating occurred at less than the 500 ms paced cycle length due to intermittent capture, sometimes with spontaneous beating. Black www.nature.com/scientificreports/ and I CaL versus CM MDP almost completely overlap at different G MFB-CM (Fig. 6E) and r C (Fig. 7F), showing that these currents are essentially determined only by MDP. However, plots of CV versus peak I Na show that these curves do not overlap and that CV decreases with increased G MFB-CM (Fig. 6F) or r C (Fig. 7G) independently of peak I Na . Thus, CV is decreased both by reduced I Na caused by elevated resting potential as well as by the loading effects of MFBs, including capacitance (illustrated clearly in the first column of Fig. 7B, where MFB currents = 0).

Discussion
Since the initial findings of Miragoli et al. 18 showing that addition of MFBs to CM cultures causes MDP elevation, conduction slowing, and spontaneous beating, a number of studies have attributed MFB-induced conduction slowing to electrical coupling between CMs and MFBs 6,7 . Interventional experiments have knocked down Cx43 in MFBs, and showed that doing so increased CV in co-cultures, compared with CMs co-cultured with unaltered MFBs, to provide direct evidence of an electrical mechanism for CV reduction by MFBs 11,22 . However, decreasing Cx43 expression also inhibits fibroblast differentiation to MFBs 23,24 and presumably the number of MFBs present, so that CV would still increase even if MFB-induced suppression of CV is by non-electrical means (e.g., paracrine or mechanical signaling). Studies that dynamically and specifically alter the electrophysiology of MFBs and monitor the subsequent changes in CM electrophysiology can circumvent such confounding effects. Previous studies have shown that CM electrophysiology can be modulated by acutely altering exogenous potassium current in co-cultured 3T3 fibroblasts 25,26 . Another used 3T3 fibroblasts transduced with ChR2 and applied mW/mm 2 light flashes to pace CMs 27 . One other study used sphingosine-1-phosphate to increase MFB inward current, and found that it suppressed CM excitability in co-cultures with MFBs, but not CMs alone 28 .
In this study, light was used to produce steady inward current specifically in ChR2-transduced cardiac MFBs. MFBs were sufficiently connected electrically to CMs for their coupling to produce diastolic depolarization ( Fig. 2A and 4A), spontaneous focal beating ( Fig. 2 and Supplementary Fig. 2), conduction slowing (Fig. 3), decreased APD 80 (Fig. 4A-B), and decreased upstroke rate (Fig. 4C), all of which can contribute to arrhythmia in the context of fibrosis. While ChR2 is not naturally present in MFBs, it is similar to the TRP channels that are upregulated during MFB differentiation 29 in that it is a relatively non-selective cation channel 15 , and our modeling shows that at the light levels used experimentally, the current it produces has a magnitude similar to that of endogenous MFB channels (Fig. 7D). The rapid time scale of these changes (within seconds, the time interval between measurements) eliminates the possibility that changes in cardiac ion channel expression in response to the presence of MFBs underlie these effects. Additionally, the absence of changes in force generation with application of light to ChR2-MFBs ( Supplementary Fig. 3) rules out the possibility that CV slowing occurred secondary to acute changes in MFB tugging forces 30 . The fact that light had no effect on control MFB co-cultures (Figs. 2-4) also supports the notion that CV effects were due to light-induced ChR2 current and not to off-target effects such as heating or photochemical reactions. While confocal imaging suggests that the observed electrical connection between MFBs and CMs is due to Cx43 (Fig. 1F-H), other possibilities, including tunneling nanotubes 31 , remain. Furthermore, in addition to electrical mechanisms, paracrine and mechanical signaling may also contribute to MFB effects, especially changes in APD 80 , which we found decreased significantly in the presence of ChR2-MFBs in our experiments (from 191 ± 22, n = 14 to 165 ± 28 ms, n = 8; p = 0.03), but not in our model (from 181 to 173 ms, Fig. 6C).
The effects of addition of MFBs to CM syncytia have been previously explored experimentally 6,7 and computationally 32 . Although experimental data have been derived mostly from cultured neonatal rat ventricular CMs, computational studies have generally relied on CM models of other species. In this work, ChR2 current 33 was added to a MFB model 34 parameterized directly from data from cultured neonatal rat MFBs 13 , which was coupled to a modified neonatal rat ventricular CM model 17 (Supplementary Tables 2-7) to enable G MFB-CM to be estimated. This approach extends the OptoGap method previously adopted by the authors 35 . Examination of different values of G MFB-CM and light intensities showed that, provided G MFB-CM is large enough, all of the phenomena seen in our experiments occur, including spontaneous beating (Fig. 6A), conduction slowing (Fig. 6B), and APD 80 reduction (Fig. 6C). We then determined the syncytial G MFB-CM that best matched our experimental data. While previous studies have used FRAP to determine the relative coupling between CMs and other cells, including MFBs 36 , these cannot be directly translated to a value of electrical conductance 35 . Dual-cell patch clamp has been used to quantify the electrical connection between CM and MFB cell pairs 13,14 , but estimated values have varied over a very wide range (0.31 14 to 165 nS 13 ). Also, cells grown in isolation have structural and electrophysiological properties that differ from cells grown in a syncytium 37 . Furthermore, dual-cell patch clamp measures the conductance between adjacent cells spread on a substrate, where because of the small height of the cells relative to their area, only a small fraction of the cell surface is available for connection, which is not the case for cells in 3-D tissue. Indeed, this limited interface also occurs in co-cultures of cells intermixed in 2-D. In this study MFBs were layered on top of CM monolayers, so that the MFB-CM interface was effectively in 3-D and occurred over a large area, allowing us to estimate a syncytial G MFB-CM of 5.3 nS/CM in a geometry more similar to that in vivo, and resulting in excellent agreement between our model and experiments (Supplementary Figs. 7 and 8, Supplementary Table 8).
We also proceeded to model the effects of important changes that can occur in fibrosis -increased G MFB-CM , increased r C due to increased number of (myo)fibroblasts, and increased k I,endo . The results showed how these contribute to arrhythmogenic spontaneous beating (Fig. 6A and 7A) and conduction slowing (Fig. 6B and 7B). The model also substantiates the previously proposed mechanism 38 , as well as our experimental findings, that endogenous (Fig. 7E) or exogenous (Fig. 6D) currents in MFBs can depolarize CMs, thereby inactivating their sodium channels (Fig. 6E and 7F), reducing upstroke rate (Fig. 4C), and slowing conduction (Fig. 6F and Fig. 7G). It also shows that increased MFB inward current causes spontaneous beating (Fig. 6A and 7A), which has the highest rate at r C = 0. www.nature.com/scientificreports/ greatest risk of ectopic beating by this mechanism. During spontaneous beating, the model shows conduction is mediated significantly by calcium current (Fig. 6E, inset) due to block of sodium current, in agreement with a previous study that found little effect of the sodium channel blocker TTX on CV in CM strands co-cultured with large numbers of MFBs 18 . Our findings show that the CM MDP and the relative amount of MFB to CM currents are important determinants in effects of MFBs on CV and spontaneous beating. Our neonatal rat CM model has a more positive MDP and lower resting membrane resistance than in adult rat or human cells, which makes them more susceptible to MFB electrical effects. However, elevated MDP and reduced I K1 current (reducing resting CM current) have been reported in heart failure 39 , which is often a comorbidity of fibrosis 4,5 , suggesting that our model may well apply in this case.
Our findings complement two recently published studies which used 3T3 fibroblasts transduced with a depolarizing or hyperpolarizing opsin to investigate the relative contributions of current sinking vs. diastolic depolarization effects of the fibroblasts on neonatal rat CM electrophysiology 40 , and show that light-activated fibroblasts can increase or decrease beating rate in human pluripotent stem cell-derived CMs, especially after fibroblast differentiation to MFBs by TGF-β1 41 . Our work used isolated neonatal rat cardiac myofibroblasts, which have different electrophysiology than 3T3 fibroblasts 42 , together with the depolarizing opsin ChR2 to produce spontaneous beating, conduction slowing, decreased upstroke velocity, and reduced APD 80 , and estimate G MFB-CM in a syncytium across a 3-D interface between MFBs and CMs. Previous studies 43,44 have also probed electrical connections between MFBs and CMs in vivo by creating genetically engineered mice that expressed an optogenetic voltage sensor specifically in non-CMs, and found time-varying signals specifically near sites of injury, suggesting that non-CMs can electrically connect to CMs in those areas. However, this does not necessarily demonstrate the G MFB-CM is large enough to significantly affect CM electrophysiology, since non-CMs are better voltage followers than drivers when coupled to CMs because of their higher sarcolemmal resistance and lower sarcolemmal currents 35 . Indeed, our modeling suggests that MFBs have significant action potential-like deflections in membrane voltage even under conditions where they have little effect on CM electrophysiology. For example, when light intensity = 0 and G MFB-CM = 0.53 nS/CM (one-tenth that suggested by this study), MFB amplitude is 55% of CM amplitude ( Supplementary Fig. 6A), while CV decreases by only 7% (Fig. 6B) and APD 80 by only 3% (Fig. 6C). Therefore, using a similar design with an optogenetic actuator, as done in this study, instead of an optogenetic sensor may be better suited to determine whether G MFB-CM is large enough to cause arrhythmogenic conduction slowing and spontaneous beating in vivo.

Conclusion
This study used optogenetic actuation of inward current in myofibroblasts to show that they can acutely cause spontaneous beating, conduction slowing, decreased upstroke rate, and decreased action potential duration in co-cultured cardiomyocytes, clearly demonstrating functional electrical coupling between these cells. Computational modeling of the experiments allowed the estimation of myofibroblast-myocyte coupling in a syncytium in which myofibroblasts and cardiomyocytes interact in 3-D. It also demonstrated how increased strength of myofibroblast electrical coupling to cardiomyocytes, increased myofibroblast/myocyte capacitance ratio, and increased myofibroblast current levels, all of which may occur in fibrosis, can work in tandem to generate a proarrhythmic reduction in conduction velocity and increase in spontaneous beating.

Methods
An expanded description of the methods is available in the supplementary material.
Cell culture. This study is in compliance with the ARRIVE guidelines for in vivo study on animals. All animal procedures were approved by the Johns Hopkins Animal Care and Use Committee and were performed in compliance with guidelines of federal and state laws and regulations. Neonatal rat ventricular CMs were produced as described previously 45 , with minor modifications, and plated onto coverslips coated with 25 μg/mL fibronectin at 1 million cells per well for a 12-well plate or 500,000 per well for a 24-well plate (approximately 250,000/cm 2 ). During isolation, CMs were separated from fibroblasts using two 1-h preplating steps. Fibroblasts from the first preplate were passaged twice. Some were transduced during the second passage with ChR2-YFP adenovirus at a multiplicity of infection of 2,000, as determined by experiments ( Supplementary Fig. 10), with media changed 4-6 h later to remove virus as described previously 46 . Transduced and non-transduced fibroblasts were treated with 5 ng/mL TGF-β1 (R + D Systems) for 2-3 days to differentiate them into MFBs.
Co-culture, imaging, and optical mapping. ChR2-MFBs or non-transduced MFBs were added to 4 to 5-day-old CM monolayers at 400,000/well in 12-well plates or 200,000/well in 24-well plates (approximately 100,000/cm 2 ), giving a MFB:CM cell ratio of 0.4. High levels of TGF-β1 treatment were used to irreversibly differentiate fibroblasts into MFBs 47 to ensure maintenance of an MFB phenotype without application of exogenous TGF-β1 during co-culture, which could have directly affected CMs. On days 5-8, co-cultures were imaged under phase contrast and fluorescence microscopy (Eclipse TE2000U, Nikon) to examine their morphology and continued expression of ChR2, and then placed in a custom optical mapping system 48 and stained for 5 min with 35 μM of the voltage-sensitive dye di-4-ANBDQBS (obtained from Dr. Leslie Loew, University of Connecticut), which is excited by red light (λ = 655 nm) 49 and therefore spectrally different than excitation of ChR2 15,50,51 . Tyrode's solution at 35 °C was then continuously flowed over the cells. Pacing threshold was determined to within 1 V, and cells were point paced at 1.1 × threshold for 5 min at 500 ms CL to reach steady state. A baseline optical recording was taken, and then continuous blue light (λ = 448 nm) was applied across the entire monolayer to activate ChR2 channels for approximately 2 s before and throughout the duration of a 2 s recording, after www.nature.com/scientificreports/ which the light was switched off, and a post-activation recording was collected within seconds. This was done for different light intensities, starting at approximately the lowest intensity for which changes could be detected (I 0 = 0.0057 mW/mm 2 ), and increasing to 3*I 0 and 10*I 0 , at which point 6/7 samples beat spontaneously faster than the paced rate. Co-cultures were fixed and stained for α-actinin (Sigma), α-SMA (DAKO), connexin43 (Cx43) (Sigma), YFP (Invitrogen GFP Ab), and/or DAPI before confocal imaging (LSM 710NLO-Meta, Zeiss).
Mathematical model. Experimental data were modeled in MATLAB (The MathWorks) using a modified version of the Korhonen model of neonatal rat ventricular CMs 17 (see Fig. 5, expanded methods, Supplementary  Fig. 11 (Fig. 5A).
In addition to endogenous currents, the Williams model for ChR2 current 33 was added to MFBs. MFBs were considered to have similar capacitance per cell as CMs, as done previously 52 , and so r C = 0.36, similar to the cell number ratio in experiments. Each CM was 50 μm long, and had a 50 µm-long interface with the following cell, based on previous measurements 52 . A 30-cell-long 1-D cable of ChR2-MFB/CM cell pairs was modeled ( Fig. 5B and Supplementary Tables 2-4). Based on previous data which found conductance between two CMs to be 7.7 nS/μm 13 , neighboring CMs were connected by lumped gap junctions with a conductance of 50 μm*7.74 nS/ μm = 387 nS.
Data processing and statistics. Optical mapping data were processed by custom MATLAB software. Cocultures with initial CV below 10 cm/s or that could not be paced at 500 ms CL (either due to inexcitabality or spontaneous beating faster than the paced CL) were excluded from analysis. Only samples that were not beating faster than the paced 500 ms CL were included in CV and APD analysis. Confocal images were processed by FIJI 53 and Zen Black (Zeiss) software. Background values from empty wells were subtracted from plate reader measurements. All data are presented as mean ± SD. Paired or unpaired t-tests with unequal variances were used to determine statistical differences, where appropriate. Differences were considered statistically significant at p < 0.05. Additional statistics on experimental data are shown in Supplementary Table 1.

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.