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

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 (Fig. 1F) with ChR2-MFBs resting on top of them (Fig. 1G and H; see Supplementary Fig. 1 for full z-stack), as well as Cx43 puncta in the MFB cell layer (Fig. 1G), suggesting expression of Cx43 by MFBs.

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*I₀ light (I₀ = 0.0057 mW/mm², 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*I₀ light, 6/7 co-cultures beat spontaneously, and the change in CL 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 I₀ (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*I₀ (ΔCV = –27 ± 5% from a baseline of 14.8 ± 3.6 cm/s, n = 5; p < 10^–3 vs. MFB co-cultures, Fig. 3B).

Addition of ChR2-MFBs to CM cultures significantly reduced action potential duration at 80% repolarization (APD₈₀, from 191 ± 22, n = 14 to 165 ± 28 ms, n = 8; p = 0.03). Light-induced inward current further decreased APD₈₀, which reversed upon removal of light (Fig. 4A). Across multiple samples, application of 3*I₀ light resulted in a significant decrease in APD₈₀ in ChR2-MFB co-cultures (− 14 ± 7%, n = 5; p = 0.004 vs. MFB co-cultures, Fig. 4B) and significantly decreased upstroke rate (− 13 ± 4%, n = 5, from a baseline of 3.8 ± 0.4%/ms; p = 0.002 vs. MFB co-cultures, Fig. 4C). P-‍values for the experimentally measured values calculated using other methods (paired and equal variance, instead of unequal variance t-test) are provided in Supplementary Table 1.

Acute application of a ChR2-saturating level of light (1.2 mW/mm², equal to 210*I₀) 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 model¹⁷ 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 capture CL decreased from 420 to 200 ms (− 80 pA/pF, 0.5 ms stimulation), which is similar to what we have reported previously (~ 250 ms²⁰). CV increased from 4.5 to 21.0 cm/s, the same as the experimentally measured CV (20.9 ± 4.3 cm/s). APD₈₀ was also reduced from 208 to 181 ms (Supplementary Fig. 4 and Supplementary Table 8), more closely matching our measured APD₈₀ of 191 ± 22 ms. While the addition of ChR2-MFBs to the CM cable caused 2:1 block at 500 ms CL in the original Korhonen model, ChR2-MFB/CM cables using the modified model could be stimulated at 500 ms CL, and had a CV and APD₈₀ of 17.5 cm/s and 173 ms, respectively, similar to our measured values of 17.7 ± 5.3 cm/s and 165 ± 28 ms (Supplementary Table 8). Neighboring MFBs were not connected to each other to prevent them from having “double-sided” effects on CMs²¹ by creating an alternate current path. However, additional modeling (with MFB-MFB conductance = CM-CM conductance) 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*I₀ 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*I₀ (Fig. 6A and Supplementary Fig. 6A). This behavior agreed with our experiments, in which 4/10 of our samples beat spontaneously at 3*I₀, while 6/7 samples beat faster than the paced rate when stimulated with 10*I₀. 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 I₀ and 3*I₀ 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*I₀) caused loss of capture due to spontaneous beating, as discussed for Fig. 6A.

Addition of MFBs to CMs decreased APD₈₀ 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 I₀ and 3*I₀ light decreased APD₈₀ 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 APD₈₀ 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 dot on green curve, Fig. 6D), while depolarizing CMs from − 73 to − 71 mV (black dot on red curve, Fig. 6D). 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 I₀ or 3*I₀, respectively (blue dots at I₀ and 3*I₀, Fig. 6D). Higher light level (10*I₀) 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 I₀ or 3*I₀ 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*I₀, peak I_CaL was 68% of peak I_Na, whereas at 10*I₀, 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 APD₈₀, 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 (I₀ and 3*I₀) resulted in MFB-CM current during diastole equal to that produced by an 84% and 117% increase in endogenous MFB current, respectively (Fig. 7D).

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 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).

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 previously⁴⁵, 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/cm²). 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 previously⁴⁶. 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/cm²), giving a MFB:CM cell ratio of 0.4. High levels of TGF-β1 treatment were used to irreversibly differentiate fibroblasts into MFBs⁴⁷ 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 system⁴⁸ 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 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 (I₀ = 0.0057 mW/mm²), and increasing to 3*I₀ and 10*I₀, 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 CMs¹⁷ (see Fig. 5, expanded methods, Supplementary Fig. 11 and Supplementary Tables 2–7 for model details). With the modifications, the model had a much better fit to the experimental data (Supplementary Figs. 4, 7 and 8 and Supplementary Table 8). CMs were connected via a lumped gap junction to a lumped MFB unit, with currents defined by the Sachse fibroblast model³⁴ with current conductances fitted to the I/V curve measured by Salvarani et al. for TGF-β1-treated MFBs¹³ (Fig. 5A). In addition to endogenous currents, the Williams model for ChR2 current³³ was added to MFBs. MFBs were considered to have similar capacitance per cell as CMs, as done previously⁵², 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 measurements⁵². 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/μm¹³, 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. Co-cultures 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 FIJI⁵³ 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.