Cardiovascular diseases remain a major cause of death globally. Basic research using optical mapping to measure action potentials, calcium concentrations and conduction velocity at high spatiotemporal resolution has had a major impact on our understanding of such diseases (1–5). And because this tool permits the investigation of cardiac electrophysiology at a macroscopic spatial scale, much has been learned about cardiac excitation wave dynamics during fibrillation. In addition, the use of drugs to perturb normal functioning of ion-channels and gap junctions provides mechanistic insights into pathophysiological behavior (6, 7). However, studies involving the use of existing and new drugs at various concentrations and combinations requires the use of scalable cardiac substrates and scalable measurement technologies. This requirement is further underscored when considering the immense time and money invested in early drug discovery (8, 9).

Currently, optical mapping/imaging is the only multi-parameter measurement technology that is scalable. It has been applied to whole-hearts (ex vivo and in vivo), tissue-slices and cardiac monolayers/tissue-constructs (1–5). Cardiac monolayers/tissue-constructs (also known as engineered heart tissues) are scalable macroscopic substrates that permit the measurement of several key electrophysiological parameters (10–14). Although many recent advances have been made in optical mapping technologies for these substrates (15–21), there is a scarcity of published work on fully-automated optical mapping systems that not only perform optical mapping, but also perform fluid-handling, electrical/optical stimulation and mechanical positioning, all inside an incubator. To truly increase measurement throughput, all tasks involved in an optical mapping experiment must be automated without human intervention, similar to what has been done in the production and maintenance of human pluripotent stem cell-derived cultures (22, 23) and the analysis of optical mapping data (24).

As a formative project, we aimed at developing a barebones, scalable and fully-automated monolayer optical mapping robot that requires no human intervention, specifically for macroscopic cardiac monolayer substrates large enough to study excitation wave dynamics. High-throughput optical mapping systems for “point” measurements of cardiac cell cultures plated in 96-well plates, for example, have already been developed (16, 25). In effect, each well yields one action potential/calcium transient measurement point (26). Here, we describe and validate the robot using the well-established and inexpensive neonatal-rat-ventricular-myocyte (NRVM) monolayer substrate (27) plated on standard 35 mm dishes by loading cells with dye and treatment and then performing macroscopic optical mapping, all in a parallelized fashion inside a 37°C incubator. To demonstrate future applicability to human pluripotent stem cell-derived cardiomyocyte (hPSC-CM) monolayers/tissue-constructs (28), we performed parallelized macroscopic optical mapping of voltage dynamics in hPSC-CM monolayers also plated on standard 35 mm dishes using a relatively fast genetically encoded voltage indicator (GEVI) and a commonly-used voltage sensitive dye (VSD) (25, 29–31). The total cost of the system components is <$15,000 USD for the oblique excitation configuration and <$20,000 USD for the perpendicular excitation configuration, which requires the use of more components. The total cost of the system is lower than the cost of one optical mapping camera system typically used in the field. By overcoming the limitations of cost and human intervention, we believe the fully-automated high-throughput electrophysiological measurements of macroscopic cardiac constructs can be operated seamlessly.

Here we present data from a sample 16-monolayer experimental run. The robot tasks were pipelined because of the dye and treatment incubation times and the time needed to perform fluid-handing and optical mapping tasks (Supplementary Figure S2). For drugs, we used (1) nifedipine, a calcium-channel inhibitor that reduces the calcium transient amplitude (37) and (2) flecainide, a sodium-channel inhibitor that reduces conduction velocity (38). The vehicle for both drugs was DMSO. In this sample run, 5 out of 16 NRVM monolayers entered spontaneous fibrillatory activity, which is a well-known disadvantage of using NRVM monolayers plated on 35 mm dishes. However, this disadvantage is tolerated because of the simplicity and low-cost of this preparation.

The 11 non-fibrillating monolayers were divided into 3 treatment categories: vehicle (0.1% DMSO, 4 monolayers), nifedipine (3 µM, 3 monolayers) and flecainide (0.6 µM, 4 monolayers). Figures 3A,B shows sample results comparing the effects of nifedipine and vehicle treatments on the calcium transient amplitude during 2 Hz pacing. The calcium transient amplitude was measured right before (control) and then 15 minutes after treatment using ΔF/F as a measure. The sample traces in Figure 3A show a decrease in amplitude after nifedipine treatment. The value of ΔF/F went from 0.1143 ± 0.0178 (control) to 0.0591 ± 0.0073 after nifedipine treatment (n = 12; 3 dishes and 4 points per dish) whereas ΔF/F went from 0.1441 ± 0.0430 (control) to 0.1363 ± 0.0343 after vehicle treatment (n = 16; 4 dishes and 4 points per dish). Figure 3B shows the decrease in calcium transient amplitude after nifedipine treatment compared to vehicle treatment.

Figures 3C–E shows sample results comparing the effects of flecainide and vehicle treatments on the conduction velocity during 3 Hz pacing. The conduction velocity was measured right before (control) and then 15 minutes after treatment along the diameter of the dish. We used the time fluorescence rises to 50% of its peak change as the “activation time” for a propagating wavefront. Figure 3C shows normalized fluorescence intensity maps at progressive time points during 3 Hz electrical pacing and Figure 3D shows the corresponding activation maps. It can be seen that flecainide slows the propagation speed of the activation front. The conduction velocity went from 22.9 cm/s ± 1.6 cm/s (control) to 17.9 cm/s ± 2.2 cm/s after flecainide treatment (n = 4; 4 dishes) whereas conduction velocity went from 22.9 cm/s ± 2.4 cm/s (control) to 24.0 cm/s ± 2.1 cm/s after vehicle treatment (n = 4; 4 dishes). Figure 3E shows the decrease in conduction velocity after flecainide treatment compared to vehicle treatment. Supplementary Movie S2 shows parallelized optical mapping with 4 cameras recording simultaneously under one column of NRVM monolayers during a control measurement.

As a proof-of-principle experiment for future development, we performed parallelized macroscopic optical mapping of hPSC-CM monolayers plated on 35 mm dishes using the GEVI ASAP1 (n = 4) and the synthetic VSD FluoVolt (n = 8). Figure 4A shows the lower-cost imaging head used (oblique excitation configuration). This imaging head is made up of 4 modules, which are in turn made up of a high-speed CMOS camera, camera lens, emission filter (EM), excitation filter (EX), high-power LED and a collimating lens for the LED light. Figure 4B shows sample action potentials from an ASAP1-expressing hPSC-CM monolayer paced at 3 Hz (Supplementary Movie S3). For a higher ΔF/F and faster kinetics, we loaded wild type hPSC-CM monolayers with a commonly-used synthetic VSD FluoVolt. Figure 4C shows sample action potentials during 1 Hz paced activity (Supplementary Movie S4).

To accelerate the elucidation of disease mechanisms and development of effective non-proarrhythmic drugs, high-throughput and low-cost electrophysiological measurement systems must be developed. This necessitates a scalable cardiac substrate and a scalable measurement technology. Here, we described and validated a robot using the inexpensive NRVM monolayer substrate and the clinically-relevant hPSC-CM monolayer substrate, plated on standard 35 mm dishes, by automating all tasks involved in a basic optical mapping experiment. With the total cost of the system components being <$15,000 USD for the oblique excitation configuration, we believe that fully-automated high-throughput electrophysiological measurements of macroscopic cardiac constructs will become more widespread and will enable the investigation of many variations and combinations of drug perturbations in a significantly reduced time frame.

The cost difference between the two imaging head options was significant and the oblique excitation configuration is recommended in cases where there are no geometric constraints. Perpendicular excitation is a more widely-used configuration in fluorescence imaging and combines excitation and emission light into a single optical path. Such a configuration is advantageous in cases where there are geometric constraints on oblique excitation but disadvantageous in certain multi-parametric imaging scenarios where the excitation wavelength of a fluorophore is longer than the emission of another (39–41). The oblique excitation configuration avoids spectral congestion and more easily takes advantage of multi-band emission optical filter technology but has more geometric constraints because the excitation light must illuminate the substrate at an angle.

To more thoroughly study cardiac electrophysiology mechanisms, longitudinal measurements are needed. Such measurements will require the integration of an off-the-shelf or custom 5% CO₂ incubator, which has become a more affordable option with the development and availability of optical gas sensors (42, 43). And with the addition of microscopy, we believe that the next generation of optical mapping robots will yield new observations important to both the drug-development and basic science research communities.

Our barebones system has limitations and can certainly be improved upon based on experimental needs. Other parameters, such as force/contractility can be measured acutely, or longitudinally over several days and even weeks. As noted earlier, longitudinal measurements will require the integration of an incubator that maintains a 5% CO₂ environment and a method for maintaining high humidity over cell cultures while not damaging surrounding electrical, mechanical and optical instrumentation. With the emergence of relatively affordable commercial sources of human induced pluripotent stem-cell-derived cardiomyocytes (iPSC-CMs) (17) and the large-scale production of mature hPSC-CMs (31), NRVM monolayers are largely being replaced as a standard experimental cell culture model (10, 14). Any future development of high-throughput systems should focus on the use of human stem-cell-derived cardiomyocytes. The use of brighter GEVIs, like ASAP3 (44), and genetically encoded calcium indicators, like GCaMP (45, 46), can also simplify longitudinal measurements by eliminating the need for repeated dye loading.

To reduce the incidence of spontaneous fibrillatory activity (Figure 1C), we also investigated the use of small-width rectangular monolayers by gluing custom-made polydimethylsiloxane (PDMS) channels to off-the-shelf tissue-culture-treated polystyrene slides and found the incidence of fibrillation drastically reduced (Supplementary Figure S3). Not only does this easily-implemented geometry reduce the number of cells needed, but it simplifies conduction velocity measurements and more readily permits the imaging of multiple monolayers from a single camera sensor (Supplementary Movie S5). Lastly, if one has the capacity to generate hundreds or thousands of hPSC-CM monolayers/tissue-constructs at a macroscopic scale, a more advanced robot should be able to handle trays of cell culture dishes to enable higher-throughput optical mapping experiments.