Simultaneous Widefield Voltage and Dye-Free Optical Mapping Quantifies Electromechanical Waves in Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes

Abstract

Coupled electromechanical waves define a heart's function in health and diseases. Optical mapping of electrical waves using fluorescent labels offers mechanistic insights into cardiac conduction abnormalities. Dye-free/label-free mapping of mechanical waves presents an attractive non-invasive alternative. In this study, we developed a simultaneous widefield voltage and interferometric dye-free optical imaging methodology that was used as follows: (1) to validate dye-free optical mapping for quantification of cardiac wave properties in human iPSC-cardiomyocytes (CMs); (2) to demonstrate low-cost optical mapping of electromechanical waves in hiPSC-CMs using recent near-infrared (NIR) voltage sensors and orders of magnitude cheaper miniature industrial CMOS cameras; (3) to uncover previously underexplored frequency- and space-varying parameters of cardiac electromechanical waves in hiPSC-CMs. We find similarity in the frequency-dependent responses of electrical (NIR fluorescence-imaged) and mechanical (dye-free-imaged) waves, with the latter being more sensitive to faster rates and showing steeper restitution and earlier appearance of wavefront tortuosity. During regular pacing, the dye-free-imaged conduction velocity and electrical wave velocity are correlated; both modalities are sensitive to pharmacological uncoupling and dependent on gap-junctional protein (connexins) determinants of wave propagation. We uncover the strong frequency dependence of the electromechanical delay (EMD) locally and globally in hiPSC-CMs on a rigid substrate. The presented framework and results offer new means to track the functional responses of hiPSC-CMs inexpensively and non-invasively for counteracting heart disease and aiding cardiotoxicity testing and drug development.

Figure 1

Schematic diagram of the simultaneous voltage and dye-free macroscopic imaging system using oblique transillumination with two low-cost Basler CMOS cameras synchronized by a trigger. A laser or a 660 nm LED was used for oblique illumination at 30°. At 0.315× magnification, the FOV was about 16 × 12 mm² (confined to a circular region by an iris or cropping). CMOS: CMOS camera; DM: dichroic mirror; TL: tube lens; EF: emission filter; MF: magnification factor; OL: objective lens. Example fluorescence and dye-free images are shown.

Figure 2

Dye-free only optical mapping and simultaneous calcium–dye-free widefield imaging of reentrant waves. (A) Optical mapping of paced mechanical waves in unlabeled hiPSC-CMs using oblique illumination and dye-free imaging: activation maps and global and local traces shown. (B) Simultaneous optical mapping of calcium waves and mechanical waves using the system in Figure 1 with appropriate light sources and filters; samples labeled with Rhod-4 calcium-sensitive dye. The anatomical reentrant wave was induced by removing cells from the center of the dish to create an obstacle and applying rapid pacing until a sustained self-propagating reentrant wave was observed.

Figure 3

Comparison of activation maps from simultaneous widefield-imaged voltage and dye-free signals in hiPSC-CMs under different stimulation frequencies. (A) Normalized global (averaged over the FOV) action potentials imaged by the NIR voltage-sensitive dye BeRST1 (red) and dye-free (DF) signals—original and TD-boosted (black) under 0.5, 0.75, and 1 Hz pacing. Shown at the bottom are the respective detected activation times (marked) by a TD filter. (B) Macroscopic NIR fluorescence (top) and dye-free (bottom) images of the hiPSC-CM samples in the same FOV (left). Small triangles indicate the centers of the matched regions of interest (ROIs, 70 μm × 70 μm each) for the local trace plots. Shown on the right are voltage (top) and dye-free (bottom) activation maps along with the extracted local traces at the indicated triangle points under 0.5, 0.75, and 1 Hz pacing from a point in the upper right corner. Isochrones are 0.05 s apart. All measurements were done at room temperature.

Figure 4

Quantification of basic parameters of electromechanical waves in hiPSC-CMs by simultaneous widefield voltage and dye-free optical mapping. (A) Restitution (frequency response) of action potential and dye-free signal duration (APD and DFD) for all samples; voltage in red (slope: −0.26; R²: 0.45) and dye-free in black (slope: −1.48; R²: 0.7). The inset shows that for voltage, the measurement was at 80% level, i.e., APD80, and for the original dye-free signals, the transient duration measurement was between the start of contraction and the end of the relaxation. (B) Frequency response of the relaxation–contraction (R–C) ratio (see the inset) for all samples based on dye-free signals (slope: −1.24; R²: 0.33). (C) Electromechanical signal duration ratio, i.e., APD-DFD ratio, as a function of pacing frequency (slope: 0.44; R²: 0.52). (D) Conduction velocity (CV) restitution (voltage and dye-free slopes: −5.24 and −4.78 and R²: 0.25 and 0.28, respectively). (E) Wavefront morphology: voltage and dye-free wavefront segmentation from an activation map (obtained at 0.5 Hz pacing) using frames that are 0.15 s apart. Close-up insets showing the local dye-free velocity vectors; green arrows indicate the wavefront. (F) Wavefront tortuosity (WT) index (for details, see Methods) quantified for voltage and dye-free waves for all samples as a function of the pacing frequency (dye-free slope: 22.77; R²: 0.25). All measurements were at room temperature. Data are presented as scatter plots (mean values per sample) with overlaid mean ± S.E. (n = 6 samples). Regression was statistically significant (two-way ANOVA test, p < 0.05). The WT index for voltage did not have a significant dependence on frequency.

Figure 5

Correlation of electrical and mechanical conduction properties, detection of pharmacologically inhibited conduction by voltage and dye-free imaging, and correlation of electromechanical conduction to gap junctional (Cx43) protein levels. (A–C) Correlation of electromechanical properties for individual samples (n = 6): (A) Voltage–dye-free correlation of CV across pacing frequencies. Linear regression was applied for data fitting (slope: 0.83; R²: 0.94). (B) Voltage–dye-free correlation of transient duration DFD vs APD across pacing frequencies; linear regression line (slope: 3.09; R²: 0.46). (C) Voltage–dye-free correlation of the WT index across pacing frequencies. (D, E) Pharmacological inhibition of conduction by cell uncoupler heptanol and the response seen by voltage and dye-free optical mapping (n = 3). (D) Voltage and dye-free activation maps predrug and after 30 min of 0.5 mM heptanol application; pacing frequency was 0.75 Hz; left: global voltage and dye-free signals. (E) Parameter comparison from the dual imaging predrug and after heptanol application. Both modalities sense cell uncoupling in a similar way; electromechanical delay is increased by cell uncoupling, most pronounced as a prolongation of the global aEMD. (F–H) Correlation of electromechanical conduction to Cx43 protein levels (n = 5) (two of the samples are from the same set as in Figure 4, and three of them are additional samples). Panel (F) presents a pipeline for correlating functional properties to Cx43 protein by western blot. After functional optical measurements, protein was extracted, and western blot was done by the automated capillary electrophoresis Wes system, keeping track of the sample identity. Panel (G) shows the Wes-generated digital lanes (obtained from electropherograms) for Cx43 and a housekeeping protein GAPDH, along with the quantified ratios for the samples from several runs/technical replicates. Negative and positive controls for Cx43 were lysates from WT HeLa cells and from human female heart ventricular tissue. Shown underneath are voltage and dye-free activation maps for samples 3 and 5. Panel (H) shows individual sample correlation of voltage (red) and dye-free (black) CV to the Cx43/GAPDH ratios. Linear regression lines for voltage had a slope of 24.52 and R² = 0.97, and those for the dye-free signal had a slope of 18.82 and R² = 0.97. Biorender was used for parts of the figure.

Figure 6

Apparent electromechanical delay (aEMD) distribution over space and pacing frequency. (A) Space segmentation guided by voltage wave propagation. Shown from top to bottom: voltage activation maps, segmentation images, and local voltage action potentials (red) and dye-free original traces (black) extracted from each segment under 0.5, 0.75, and 1 Hz pacing. (B) Local aEMD distribution over space (over the 11 segments from the pacing to the distal site) under each pacing frequency. The inset shows that aEMD was defined by the time lag between the onset of voltage action potential and the onset of dye-free contraction. (C) Local aEMD as a function of pacing or spontaneous frequency (slope for pacing data: 0.19; R²: 0.76); shown are averaged values for the 11 segments across all samples. (D) Global aEMD as a function of pacing or under spontaneous frequency (slope for pacing data: 0.26; R²: 0.41). Plotted data are presented with overlaid mean ± S.E.; n = 6 for pacing data and n = 5 for spontaneous data (one sample did not show spontaneous activity). The averaged spontaneous frequency is 0.17 ± 0.03 Hz. The pacing site was consistently at the coverslip edge for all samples. Regression was statistically significant for the local aEMD (linearity test, p < 0.0001) and global aEMD (linearity test, p < 0.01) dependence on frequency.