High-speed mechano-active multielectrode array for investigating rapid stretch effects on cardiac tissue

Systematic investigations of the effects of mechano-electric coupling (MEC) on cellular cardiac electrophysiology lack experimental systems suitable to subject tissues to in-vivo like strain patterns while simultaneously reporting changes in electrical activation. Here, we describe a self-contained motor-less device (mechano-active multielectrode-array, MaMEA) that permits the assessment of impulse conduction along bioengineered strands of cardiac tissue in response to dynamic strain cycles. The device is based on polydimethylsiloxane (PDMS) cell culture substrates patterned with dielectric actuators (DEAs) and compliant gold ion-implanted extracellular electrodes. The DEAs induce uniaxial stretch and compression in defined regions of the PDMS substrate at selectable amplitudes and with rates up to 18 s−1. Conduction along cardiomyocyte strands was found to depend linearly on static strain according to cable theory while, unexpectedly, being completely independent on strain rates. Parallel operation of multiple MaMEAs provides for systematic high-throughput investigations of MEC during spatially patterned mechanical perturbations mimicking in-vivo conditions.


Supplementary Note 1: Dielectric Elastomer Actuators: Principles of Operation
Dielectric elastomer actuators (DEAs) consist of stretchable electrodes that are used to generate an electrostatic pressure that acts on a dielectric membrane consisting of polydimethylsiloxane (PDMS). As the PDMS is deformable and at the same time ν = 0.5), the attraction of the DEA electrodes causes a lateral deformation which results in a strain field. 1 To enable the DEA electrodes to follow expansion during operation, they are implemented by using carbon powder bound to the PDMS 2 . Supplementary Fig. 1a illustrates the actuation principles and Supplementary Fig. 1b illustrates the implementation of a DEA as used in this study. The device shown exhibited a large strain response (19% at 4 kV). After seeding of the cells, the strain response is reduced to ~ 10-12% at 4 kV which is likely due to stiffening of the membrane following UV activation that is required for cell attachment (cf. method section).
Prior to mounting to the holder, the PDMS membrane is pre-strained for 2 reasons: 1) It has been shown that pre-strain improves the performance by delaying mechanical buckling 2  and 1.2-fold (x-direction) sufficiently stiffens the membrane along one axis 5 as to produce linear strain along the low pre-stretch axis only.
The stamping method 6 allows the transfer of patterned electrodes onto the suspended and pre-strained membrane (see Supplementary Note 2 for details on the fabrication steps and materials used). The geometry of the DEA electrodes is chosen such that the center of the membrane remains transparent for imaging purposes. This central region experiences uniform uniaxial tensile strain (Zε+) during DEA actuation. With a total well diameter of 20 mm, the expansion of the membrane between the center DAE electrodes over a distance of ~ 2 mm is compensated for by compressive strain in the ~8 times larger outer regions (Zε-) which results in a correspondingly lower compressive strain in the Zεzones. Typical data describing strain field homogeneity and residual strain perpendicular to the main axis are presented in the supplementary material of Ref. 5 Of note, it has been shown previously that strain perpendicular to the direction of impulse conduction does not alter ATDs. 7

Supplementary Note 2: Device Fabrication and Preparation
The fabrication steps of the mechano-active multielectrode array (MaMEA) are summarized in Supplementary Fig. 2 with further details regarding DEA fabrication given elsewhere. 5 Fig. 2a,b). After curing the carbon electrodes at 80°C for 60 minutes in an oven, they are passivated by stamping ~2 μm of LSR 4305 on top of the electrodes. This insulates the high voltage electrodes both in respect to the culture medium (top) and the air (bottom) which is critical for improving the lifetime and reliability of the device. Moreover, the bottom passivation reduces electrical noise that may arise from micro-burst of currents secondary to the ionization of air.
Extracellular recording and stimulation electrodes (RecEl and StimEl) are fabricated using gold ion implantation techniques and steel shadow masks of the desired shapes (Supplementary Fig. 2c). 9 At the site of prospective contacts with the biological preparation, recording electrodes are 200 μm wide. Electrodes are passivated using stamping techniques with the exception of the tip of StimEls and the center region of RecEls.
Supplementary Figure 2. Fabrication steps after the PDMS membrane is stretched and mounted to the carrier ring. a) Carbon electrodes stamped on front side (ground electrode) and, b), on backside (high voltage electrode). Both sides of the membrane are subsequently passivated by stamping, with the exception of the contacts (pink shading). c) Ion gold implanted stimulation and recording electrodes. Electrodes are passivated (blue shading) except for the end (StimEl) and center (RecEl) which will establish contact to the excitable tissue. d)-h) Transfer of the PDMS membrane to the PCB frames with electrical contact being made with carbon containing silicone. h)-i) The device is sealed using LSR and the PMMA wall forming the culture well is added.
The transfer of the fully structured membrane to the PCB frames and the assembly of the cell culture well is illustrated in Supplementary Fig. 2d Fig. 2d,e). Subsequently, the same procedure is repeated for the high-voltage bottom PCB ( Supplementary Fig. 2f,g), thereby sandwiching the membrane between the two PCBs.
LSR 4305 is used to seal the interface between the PDMS membrane and the PCBs. This prevents the occurrence of leaks and improves device durability and reliability by ensuring a gradual transition between the free-standing elastic membrane and the solid PCBs. Exposed HV-vias are also sealed by RTV (room temperature vulcanized silicone rubber, Silpuran 4200; Wacker Chemie AG) to prevent electrical breakdown (cf. Supplementary Fig. 2h). The device is completed by fixing the wall of the cell culture well (polymethyl methacrylate, PMMA) to the device with LSR 4305.
Prior to the cell seeding, the PDMS membrane is immobilized by adding a sheet of Mylar to the backside.
Immobilization has proven to be critical for optimal cell adherence possibly by removing vibrations of the thin membrane. A second Mylar mask carrying the desired growth pattern is added to the top of the membrane. At this point, the MaMEA is complete and ready for cell seeding as described in the methods section. It has been observed that cell growth is improved if the wells are placed in the incubator for ~ 48 hours prior to cell seeding (without growth medium). The exact reason for this is not known but may involve saturation of the PDMS membrane with water in the humid atmosphere.

Supplementary Note 3: Multi-MaMEA System with Signal Conditioning
Because the MaMEA architecture lends itself for conducting multiple experiments in parallel, we designed the multi-MaMEA dish shown in Supplementary Fig. 3. In contrast to the single well MaMEAs, the multiwell configuration included signal conditioning circuits for each well for buffering and low gain amplification (3.5x) of the extracellular signals. On-board signal conditioning close to the RecEls reduces the requirements on the DAQ and reduces noise pickup. Signal conditioning close to the recording sites also minimizes the capacitive load which can affect APEC amplitudes. Accordingly, the signals shown are qualitatively at least similar to the ones recorded in single well devices as used for the experiments shown in the main manuscript. Conceptually more important is the ability to run experiments in parallel, which improves throughput and allows for more comprehensive studies and/or redundancies. Unlike cell stretchers based on pneumatics or conventional motors for actuation, the MaMEA system is intrinsically parallelizable in the sense that a single high voltage source and appropriate transistors can be used to generate a whole range of drive signals which can then be applied individually to each well because each well has a dedicated HV input for strain actuation.

Supplementary Note 4: Characterization of Mechanical Strain in the Voltage and Time Domain
To determine the extent and dynamics of strain imposed during experiments, the devices need to be calibrated under experimental conditions, i.e., the MaMEA must be held at physiological temperature and be filled with growth medium. Strain profiles in response to specific actuation voltage profiles as measured by a high-speed camera under these conditions are depicted in Supplementary Fig. 3.
Step increases in voltage result in an approximately exponential mechanical response with dynamics being limited by the slew Preparations were stimulated by a bi-polar current pulse provided by a current source that was galvanically insulated from the bath/sensing circuit to reduce electrical crosstalk with the recording electrodes. during which the preparation is stimulated twice with Vε being ramped to 3.6 kV during 5 ms between the two stimulated APs. The red trace depicts the raw data recorded by the DAQ at 10 kHz sampling rate. The peak-to-peak noise of the recording is <10 µV which corresponds to a spectral noise density of 0.1 V/√Hz.
The standard deviation of the noise is 2 µV, as can be seen in the close-up plot of the first 300 ms of the recording in Supplementary Fig. 7b. In order to detect and characterize APECs during propagated activity, the following analysis steps were performed (Matlab): 1. Artifacts due to electrical crosstalk with stimulation and DEA electrodes were blanked (data set to 0).
Typical blanking periods to remove the Istim artifact lasted 5 ms, blanking for removing the Vε artifact were chosen to be slightly longer than the ramp duration. 3. From the derivative, APECs were identified using peak detect functions (Matlab function findpeaks())and thresholds tuned to each RecEl for a given experiment. Filters were applied to remove noise spikes.
4. Once a peak was was measured by summing up the squares of its amplitude over the downstroke period of the APEC as defined by the negative derivative. This provided a second threshold for peak selection and was effective in excluding noise spikes which may be large but are shorter than APECs.

5.
ECs were selected based on the plausibility argument that APECs must occur within a limited time period after electrical stimulation. This time varies with  and with the distance between the StimEl and a given RecEl and lies within 8-100 ms of the stimulation trigger.  time, tDS, (Supplementary Fig. 8d), and for the maximum downstroke velocity, | ECP | max , (Supplementary Fig. 8e) The results show that Vpp, tDS and maximal downstroke velocity of the APECs are, with the exception of moderate drift present at RecElVI, stable to within ±1% over the entire measurement period. Inter-electrode variability is largest for Vpp, which is the expected result because this parameter is sensitive to local conditions defining the coupling between the cell culture and the recording electrodes (2) The offset of ATDs present at the end of each series of strain modulations appears to largely relax during the ~ 1 min pause before the subsequent strain protocol is applied. To which extent this phenomenon is due to relaxation processes within the PDMS membrane and whether short term and reversible alterations in cellular electrophysiology contribute to the build-up and relaxation of ATDs during the strain protocols remains to be shown. Supplementary Fig. 9b shows an excerpt of the data for tramp = 5 ms which illustrate a return to initial conditions within 0.25% over the six subsequent measurement protocols. These data suggest that, other than short-term memory effects, the CMC strand is not adversely affected by repeated mechanical perturbations.

Ventricular Cardiomyocytes and hi-PSC Derived Cardiomyocytes
To demonstrate the suitability of the MaMEA platform to record repolarization related extracellular signals in strand preparations of neonatal rat ventricular cardiomyocytes, we increased the coupling time constant of the AC-coupling circuit that connects the gold ion-implanted electrodes to the amplifiers. With the setting used for the experiments shown in this study (time constant, c, of 8 ms), the fast initial signal caused by the action potential upstroke was followed by a stable baseline, i.e., the repolarization related signal was entirely suppressed by the high-pass filtering effect of the coupling circuitry (Supplementary Figure 10a).
When increasing c, to 1600 ms, the cut-on frequency of the circuitry was reduced accordingly and the repolarization signal was revealed (Supplementary Figure 10b). This signal followed without delay on the initial fast, upstroke-related signal which is in accordance to data presented by others and is to be expected given that the rat cardiomyocyte action potential has a rather triangular shape. 12 Functional proof that this signal was due to action potential repolarization was achieved by demonstrating presence of electrical restitution of action potential duration (Supplementary Figure 10c).
To furthermore demonstrate that the MaMEA system can be used with cell types different from rat cardiomyocytes, we generated strands consisting of hi-PSC derived cardiomyocytes. As shown in Supplementary Figure 11 and similar to rat cardiomyocytes, strands constructed from these stem-cell derived cardiomyocytes produced fast signals related to the action potential upstroke. When increasing c, from 8 ms to 1600 ms, the initial fast signal was followed by a second biphasic signal representing action potential repolarization which was again functionally validated by showing presence of action potential duration restitution. In contrast to the rat cardiomyocyte repolarization signal following immediately after the first fast signal, the repolarization related signal of stem-cell derived cardiomyocytes occurred with a distinct delay that reflects the extended plateau phase of these cells. 13,14 Supplementary Note 9: Deriving  from Activation Time Differences The data presented in

Supplementary Note 10: Cable Equation Applied to a Linearly Strained Cell Strand
Here we consider how the Cable Theory predicts the scaling of impulse propagation velocity with regards to strain. Supplementary Fig. 13    Given these scaling laws, becomes independent of strain which is supported by the finding that maximal downstroke times and velocities were not or only barely affected by strain. Hence, the velocity scales with the length constant . By matching the theory to the data, one can determine the relative contribution of the cytosolic and the gap junctional resistance by considering the only remaining free parameter, x.
Comparing theory with experimental results: The derived cable theory and the empirically found results can be compared after a series expansion (around = 0) of supp. eqn. (S8) and the result is summarized in Supplementary Table 1.
where x (the ratio of cytosol and gap junction) is a free parameter and is chosen to fit the experimental results. Hence, including the first two terms and assuming a strain of 10% one can find x by supplementary equating (S11) with (S10):  (S11) This analysis suggests that, for the preparation described in the main manuscript, the increase of ATDs during strain can be explained solely by the deformation induced increase in axial resistance if the cytosolic and gap junctional resistance contribute 29% and 71%, respectively, to overall axial resistance. These numbers are between generally assumed relative contributions (50/50%) 16 and recently published values using cultured cardiomyocyte strands (25/75%) 7 . While this does not a priori rule out additional contributions from stretch activated channels, changes in cell capacitance and effects related to supernormal conduction, these contributions seem to be minor.

Strands -Additional Experiments
To validate the results presented in the main manuscript, identical experiments were performed with 5 additional preparations cultured in 4 different devices with the results being shown in Supplementary Fig.   13b-f (Source data are provided 17 ). Supplementary Fig. 13a reproducing the findings presented in the main manuscript for comparison purposes. For each preparation, ATD vs. strain rate, ATD vs. strain, and  vs.

Membranes
To test whether exposure to high voltage fields affects the electrical function of cultured cardiomyocytes in the absence of mechanical strain, the underside of the PDMS membrane of MaMEAs was mechanically stabilized with Mylar before subjecting the cultures to standard high voltage protocols. This type of immobilization has previously been described. 5 Optical control measurements confirmed that there was no strain development in presence of Mylar. As shown in Supplementary Fig. 14, applying high voltage pulses Supplementary Figure 13. Replication of the experiments shown in the manuscript with 5 independent additional preparations subjected to identical strain amplitude -strain rate protocols. Data are presented according to Figure 5 of the manuscript which is reproduced for comparison purposes (panels a1-a3). The left panels depict changes in activation time difference (ATD) as a function of different strain rates and strain amplitudes corresponding to specific DEA drive voltages (color coded). The middle panels show the dependence of ATDs on strain modulation for different strain rates (color coded) whereas the right panels show the same data after transformation of the ATD data ('preparation coordinates') into conduction velocity data ('observer coordinates', i.e., distance) with values being fitted by cable equations that take into account myoplasmic (ri) and gap junctional (rgj) resistivity. Source data are provided.
to mechanically immobilized DEAs had no measurable effect on impulse conduction as demonstrated by the complete overlap of these data (purple trace: this trace contains 5 instead of the standard 10 strain modulations, which was deemed sufficient as no change in ATDs could be observed) to those obtained in absence of DEA activation (black trace). Upon release of the mechanical immobilization, preparations showed normal modulation of ATDs with regards to modulated strain (red trace). The summary data shown in Supplementary Fig. 14b confirm that ATDs measured in the absence of DEA actuation and ATDs measured in presence of electrically driven but mechanically immobilized membranes were similar. This finding is relevant if DEAs are to be used as motors for mechanical modulation because it eliminates the concern that the observations are affected by the high electrical fields produced by DEAs.

Supplementary Note 13: Effects of High Electrical Field Strengths on Cardiomyocytes
The field strength in a DEA can reach 10 2 V/µm which may cause concerns regarding its effect on cardiomyocytes close by. Although most of the field is shielded by the ground electrodes, one would expect the fringe fields in close proximity to the edge of the DEA to be still significant (~1 V/µm). The literature provides little information on the effect of such fields on excitable cells as most experiments refer to fields within the culture medium during current flow. In our case, the field is only capacitively coupled to the cells with exchange of charge carriers occurring exclusively during the up and down ramps of V . Furthermore, the presence of the conductive cell culture medium is expected to shield the time varying fields from reaching the cardiomyocytes.
Supplementary Fig. 15 compares typical environmental field strength to those caused by the DEA. Of note, cell membranes of cardiomyocytes at rest withstand very high fields in the order of 10 V/µm. As is evident in the figure, this does not exclude that smaller external fields can affect cell behavior as well.
To determine whether capacitively coupled electrical fields adversely affect cell viability, cardiomyocytes were grown on PDMS (100 µm thick) with gold electrodes having the same geometry as the DEAs being patterned beneath the PDMS, which permitted a direct capacitive coupling of the cells to the electrodes, hence leading to their exposure to full fields (not only fringe fields as present in the DEAs).
From 9 cultures, 3 served as controls while 6 were exposed to modulated electrical fields (1 Hz, 4 kV) for 6 hours. Cell viability was assessed with a Biotium Kit (PI-#30066) that can differentiate between apoptotic and necrotic cells. As illustrated by the phase contrast and fluorescence images in in Supplementary Fig.   16, exposure to modulated electrical fields affected neither the gross morphology of the cardiomyocyte monolayers nor did it cause obvious changes in the density of apoptotic and necrotic cells. This is in line with the findings presented in the main manuscript that demonstrate that function in respect to θ and APEC characteristics was not affected by prolonged exposure to (admittedly smaller) electrical fields.
A further question in the context of fast changing electrical fields concerns their ability to trigger electrical activity. Such triggering was occasionally observed in the test cultures shown in Supplementary Fig. 16. By