Micro-electrode channel guide (µECG) technology: an online method for continuous electrical recording in a human beating heart-on-chip

To provide a means for the precise positioning of 3D micro-electrodes within microfluidic platforms, a dead-end microchannel (named 'micro-electrode coaxial guide' or 'micro-guide') geometry was designed as a guiding track for the insertion of 3D wire-shaped probes (figure 1(a)). This constrains lateral and vertical electrode movements while allowing it to move towards the target area. The target area is separated from the dead-end channel by a thin wall of perforable and self-sealing material (figure 1(a)). To accurately position the micro-electrode tip in the target area, (1) the micro-electrode is first inserted in the micro-guide until the tip reaches the perforable wall, (2) the wall is pierced, (3) the micro-electrode tip reaches the desired location. The whole process allows repeatable spatial control of the tip position without need for micromanipulation. At first, to validate µECG concept, we designed a microfluidic platform consisting of a single coaxial guide to assess its compatibility with the recording of 3D microtissues electrical activity (figure 1(b)). The platform (figure 1(b)(i)) features a central channel for injection of a cell-laden hydrogel, separated from the lateral medium channels by means of two rows of pillars designed to confine the cell construct [25]. A 150 µm wide micro-guide (figure 1(b)(i) in gray) was included for positioning a single micro-electrode (figure 1(b)(iii)), separated from the microtissue by means of a 200 µm wall. The electrode is considered to be correctly inserted when just its tip reaches the target area. Recording the electrical activity of cardiac microtissues.

Zoom In Zoom Out Reset image size

The microfluidic platform was first exploited to monitor 3D cardiac microtissues generated from neonatal rat cardiomyocytes (CMs) embedded in fibrin hydrogel. After 5 days of culture, cardiac microtissues showed a spontaneous beating, and the electrical signals generated were recorded by keeping the devices in a CO₂ incubator and connecting the electrodes to an extra-cellular amplifier (figures S1(a) and (b) (available online at stacks.iop.org/BF/13/035026/mmedia)). The typical cardiac FP morphology was recorded using either tungsten [26] sharp electrodes [27] or stainless steel needles (SSNs). Although the signal to noise ratio was comparable (figure S2), SSN resulted easier to handle, and were thus chosen for all subsequent experiments. To exploit the versatility of the µECG technology, we designed multiple micro-guides able to precisely positioning SSN electrodes and monitoring cellular electrical activity in specific areas of the microtissue. FP signals were successfully recorded from multiple regions, detailed in figure 1(c): position 'a' in contact with the extremity of the microtissue, position 'b' at the center of the microtissue and position 'c' in the lateral medium channel. As expected, signals in positions 'a' and 'b' showed a higher amplitude with peak to peak values among 500 µV and 800 µV, respectively. On the contrary, signals in position 'c', showed a reduced amplitude (less than 100 µV peak to peak amplitude), due to the tip of the electrode not being in direct contact with the microtissue (figure 1(c)(ii)). In this configuration, the separating wall was positioned along the micro-guide, while its end was placed in direct contact with the target area by means of a channel striction allowing for a consistent positioning of the electrode tip, as confirmed by optical visualization under the microscope (figure 1(c)(i)).

The µECG technology was coupled with a previously developed technology, named uBeat [28], which allowed to apply controlled mechanical stretching [18] to 3D cardiac microtissues. A new platform (named 'uHeart') was designed to provide (a) online measuring of electrical activity of beating cardiac microtissues trained by mechanical stimulation and (b) increased operational throughput comprising three biologically independent cell culture chamber units, each integrated with micro-guides. Of note, all three units were coupled with a single uBeat actuation mechanism, allowing to simultaneously apply mechanical training to three cardiac microtissues. As detailed in figures 2(a) and (b), uHeart is composed by two PDMS layers and one coverslide: (a) an upper cell culture layer containing the cell culture chambers (in red), each integrated with two lateral micro-guides to measure electrical activity at the two ends of the cardiac microtissues (in blue, 'a' and 'a'' positions), and (b) a bottom actuation layer (in green) containing pressure-actuated compartments. Each cell culture chamber unit, integrating the µECG technology, comprises two rows of overhanging posts designed both to host the 3D cardiac microtissue and to control the intensity of the mechanical stretching, as previously described [18]. Briefly, the application of a positive pressure of 0.35 atm in the actuation chamber, which was pre-filled with sterile Phosphate Buffer Saline (PBS), causes the movement of the membrane (i.e. 50 µm of displacement corresponding to 1/3 of the total microtissue height) towards the free-end of posts in the culture chamber. Being the cell-laden hydrogel confined between the two rows of the over-hanging posts, membrane deflection causes its deformation through the space between the posts, eventually resulting in a uniaxial stretching of the microtissues, as previously described [18]. Once the pressure is released, the elastic recoil of PDMS allows the membrane and the microtissue to gain the rest position. To mimic the in vivo like heartbeat, the cyclic mechanical stimulation is provided with 1 Hz frequency, with 50% of duty cycle. Figure 2(c) shows a representative image of uHeart platform.

Zoom In Zoom Out Reset image size

The uHeart platform was used to culture rat cardiac cells in 3D under dynamic culture conditions (10% of uniaxial strain, 1 Hz), which successfully organized in functional cardiac microtissues (figure S3) confirming previously obtained results [18]. To demonstrate the advantages of the self-sealing properties of the µECG, we exploited the technology to monitor the electrical activity of rat cardiac microtissue during the entire culture period, characterizing the physiological adaptation of the cells while forming the syncytium. In particular, simultaneous measurements were performed in two different locations of a cardiac microtissue (left part, position 'a'' in red; right part, position 'a' in gray, in figure 3(a)). µECG successfully recorded the occasional spontaneous beating at both sides of the microtissue already at day 1 in culture, as highlighted by two simultaneous recorded FP measured from the right and left areas of each microtissue (figure 3(a)). The captured signals evidenced a gradually reinforced cell electrical activity during time, reaching a stable and synchronous beating after 4 days in culture, as evidenced by the regular amplitude (e.g. 300–400 µV at day 4) and frequency (i.e. ≈1 Hz) of the spikes depicted in the representative signals. These results were confirmed by Poincaré plots (figure 3(b)), pointing out how RR intervals variability decreased during time (i.e. points started concentrating closer on the identity line) and how RR intervals synchronized in both part of microtissues (i.e. overlapping of the gray and red lines). Moreover, from the obtained signals, the gradual achievement of regular beating can be also quantified (figure S4(a)) and the beating synchronicity across the entire length of the microtissues can be evaluated by computing the consistency of the delay between R peaks measured from left and right parts of the microtissues during the culture period (figure S4(b)). Owing to the continuously monitoring capability of the µECG technology, the detection of arrhythmic events was also achieved. As evidenced in figure S4(c), before the establishment of a regular electrical activity (i.e. day 4 of culture), trains of smaller peaks in the FP signals between the regular beatings were present in the tracks, probably due to groups of cells not yet completely coordinated within the entire construct.

Zoom In Zoom Out Reset image size

To exploit the potentiality of uHeart in assessing different aspects and parameters of 3D cardiac microtissues beating, we generated human models by injecting a fibrin hydrogel loaded with a combination of h-iPSC-CMs (iCell Cardiomyocytes from Fujifilm) and human fibroblasts (Lonza) in a 3:1 ratio [15]. After 7 days of mechanical training (figure S5(a)), CMs exhibited a 3D elongated morphology, showing typical sarcomeric striations (cardiac troponin I staining, figure 4(a)(i)) and appeared interconnected with randomly distributed fibroblasts (vimentin-stained in red, figure 4(a)(ii)), used as supportive population [15] (phalloidin and cardiac troponin I staining, figure 4(a)(iii)). Gene expression analyses on ion-channel related genes (i.e. KCNH2; KCNQ1; CACNA1C; SCN5A) were performed after 7 days of mechanical training (figure 4(b)). Compared to the baseline (i.e. mixed population of h-iPSC-CMs and human fibroblast before fibrin embedding and seeding within uHeart), the trained microtissues showed a statistically significant upregulation of both hERG (encoded by KCNH2 gene) and IKs (encoded by KCNQ1) voltage gated ion channels, which are responsible for the outward potassium current during the cardiac repolarization phase. Conversely, sodium (i.e. encoded by SCN5A) and calcium (i.e. encoded by CACNA1C) ion channels expression were not altered by the culture within uHeart.

Zoom In Zoom Out Reset image size

Cardiac microtissues started beating already after 3 days in culture and after 7 days of mechanical stretching cells exhibited a spontaneous and synchronized beating, as shown in the video S1. µECG technology was thus exploited to verify and quantify the synchronism of the whole microtissues at day 7. As highlighted in figure 4(c), real time simultaneous recording of the cell electrical activity in the two opposite positions 'a'' (in yellow) and 'a' (in blue), revealed two FP traces overlapping in time. Of note, owing this syncytium, the typical depolarization and repolarization phases of the microtissues could be recognized and measured directly from the curves without the need of particular signal processing. As highlighted in the representative Poincaré plot of figure 4(d), human cardiac microtissues showed regular RR intervals and acquired synchronized beating at both their extremities (i.e. plot of lines from both left and right parts overlapped). Electrophysiological parameters were also calculated from FP recorded among 13 cardiac human microtissues after 7 days in culture, accounting for spontaneous beating period (i.e. 1.7 ± 0.45 s), coefficient of variation (i.e. 9.5 ± 4.4%), FPD (i.e. 0.6 ± 0.2 s) and spike amplitude (590 ± 440 µV) (figure 4(e)).

The uHeart platform was exploited for real-time, on-chip, drug screening tests on both rat and human microtissues, focusing on compounds causing well-known electrical alterations [8, 29]. The drug administration protocol is shown in figure S5(b). Three calibration compounds (i.e. aspirin, sotalol and verapamil) and one vehicle (i.e. DMSO) were tested on human cardiac microtissues, trained with mechanical stimulation for 7 days and showing a spontaneous synchronous beating. Results in figure 5(a) show representative FP curves recorded before and after the administration of increasing compound concentrations. As evidenced in figure 5(b), neither DMSO nor aspirin caused significant FPDc alterations (i.e. variations in percentage of FPD corrected with Fridericia's method, FPDc, above 10%) [29, 30] in human microtissues at any concentration tested (i.e. up to 0.6% and up to 300 µmol, respectively). Sotalol, a Kr-channel blocker [31], caused a significant 20% prolongation of the human microtissues' FPDc at 15 µmol, which doubled at 30 µmol and reached a value of about 80% at 60 µmol. Conversely, verapamil, a well-characterized class 4 antiarrhythmic compound [32], decreased the human microtissues' FPDc of about 25% at 50 nmol and of 40% at the maximum tested concentration of 1 µmol. Data on RR intervals and FPD with no correction are provided in figure S6. The obtained results were confirmed also investigating the variations in percentage of FPD corrected with Bazzet's method (figure S7) [33].

Zoom In Zoom Out Reset image size

To further exploit the potentiality of uHeart and to test its specificity in detecting inter-species variability, the effect of the two positive compounds was also assessed on rat microtissues mechanically trained for 5 days. In the rat model (figure S8), sotalol prolonged the repolarization wave peak in a dose dependent manner, with concentrations above 7.5 µmol causing an increase of the FPDc. Verapamil also increased the repolarization interval, showing an opposite trend in murine as compared to human microtissues (i.e. FPDc prolongation close to 50% at doses ranging between 10 and 1000 nM). In details, as evidenced in figure S9, the treatment of human cardiac cells with a high concentration of verapamil (i.e. 1000 nM) caused a statistically significant decrease of the FPDc (%) with respect to the increased values of rat FPDc (%) obtained at 100–1000 nM.

The layouts of the microfluidic devices were drawn in CAD software (AutoCAD, Autodesk Ink) and the corresponding optical masks for photolithography were printed at high resolution (64 000 DPI). Master molds were fabricated in a cleanroom environment (Polifab, Politecnico di Milano) by means of the conventional photolithography technique. As already described [20], the pattern of each layer was transferred on SU8‐2050 photoresist (MicroChem, USA), previously spin-coated on 4 inch silicon wafers. The culture chamber master mold integrating the µECG technology was composed by two layers of photoresist: (a) a 50 μm thin layer representing the bottom region of the cell culture chamber, forming the gap space underneath the posts, with the micro-guides and (b) a 100 μm thick layer representing the upper part of the cell culture chamber including the posts and identical micro-guides, aligned on top of the thin layer. Specifically, to obtain 50 μm and 100 μm photoresist layers, the spinning velocity of the spin coater (Karl Süss RC8, Süss Microtec, Germany) were 3840 rpm and 1900 rpm, respectively. The hanging posts formed a central channel for cell-laden hydrogel injection 300 µm wide and 1000 µm long [18]. The micro-guides, designed to be either in one (figure 1(b)) or six different positions (figure 1(c)), had a squared cross-section (side of 150 µm) and were designed with a widening of about 1 mm at the outer border to facilitate the manual insertion of the micro-electrode. Each micro-guide was separated from the target area by a 200 µm perforable and self-sealing wall. The pressure actuated compartment was obtained by spinning a single photoresist layer, 50 μm thick, with a velocity of 3480 rpm [20].

The final devices were made of PDMS (Sylgard 184; Dow Corning): the elastomer was mixed with the curing agent at a ratio of 10:1, respectively, degassed and casted on the master molds. After curing (65 °C for 3 h), PDMS stamps were peeled off the mold and assembled by permanently bonding surfaces after treatment with air plasma (i.e. high level plasma for 50 s at 0.420 torr) through the plasma cleaner (Harrick Plasma, PDC-002). Wells for media reservoirs and access ports for hydrogel injection were created within the cell culture chamber through biopsy punchers of 5 mm and 1 mm, respectively. Specifically, to produce the single chamber devices (figure 1), the culture chamber layer was plasma bonded to a 1 mm PDMS membrane. Instead, to assemble the uHeart platform (figure 2), the cell culture layer was firstly carefully aligned and bonded to the corresponding pressure-actuated compartment and then bonded to a 150 μm thick coverslide. The access port to the pressure-actuated compartment was made by punching a hole of 1.5 mm, to fit a Tygon tube (i.d. 0.50 mm, o.d. 1.5 mm) connecting the compressed air source. Once uHeart platform was assembled, each pressure-actuated compartment resulted separated from the culture chamber by means of a 800 μm thick PDMS membrane, able to transfer the mechanical stimulation to the cardiac microtissue when the pressure in the actuation chamber increases.

In this study, neonatal rat CMs were isolated from 2 d old Sprague Dawley rats (Charles River, Wilmington, MA, USA) following a previously described protocol [20]. The isolated population was enriched for CMs by performing a 1 h pre-plating. The harvested non-adhered cells, mostly CMs, were cultured in Dulbecco's modified essential medium high glucose (DMEM, 4.5 g l⁻¹ glucose, Gibco) supplemented with 10% v/v fetal bovine serum (FBS, Hyclone), 100 U ml⁻¹ penicillin (Life Technologies), 100 g ml⁻¹ streptomycin (Life Technologies), 10 mmol hepes (Gibco), 2 μg ml⁻¹ insulin (Sigma), 50 μg ml⁻¹ ascorbate (Sigma) and 5 μmol cytosine-B-D-arabino-furanoside hydrochloride (AraC, Sigma). Cells were used the day after to generate the cardiac microtissues.

Neonatal rat CMs were embedded in fibrin gel at a density of 75 × 10⁶ cells ml⁻¹ and injected in the microfluidic platforms. In detail, a human fibrinogen (Sigma Aldrich) solution was mixed with the cardiac cells suspended in DMEM containing human thrombin (Tisseel, Baxster), in order to achieve a final fibrin gel of 20 mg ml⁻¹ fibrinogen with 5 U ml⁻¹ of thrombin. The cell-laden fibrin solution was pipetted into the central channel of each cell culture chamber (∼2 μl each unit) and incubated for 5 min (5% CO₂; 37 °C) to allow the gel cross-linking, before adding the culture medium in the lateral channel. The microtissues were cultured for 5 d in cardiac complete medium, composed by a 1:1 (v/v) mix of DMEM (Gibco, 4.5 g l⁻¹ glucose) and EGM-2MV (Lonza) supplemented with 10% v/v FBS (Hyclone), 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin, 10 mmol hepes and 2 mg ml⁻¹ of aminocaproic acid (ACA, Sigma). The amount of ACA in the medium was gradually diminished as follow: day 1 (2 mg ml⁻¹), day 2 (1.6 mg ml⁻¹), day 3 (1.4 mg ml⁻¹), day 4 (1.2 mg ml⁻¹), day 5 (1 mg ml⁻¹). Cardiac tissues were mechanically trained (10% uniaxial strain, frequency 1 Hz) for 5 d, starting 4 h from the seeding.

Human cardiac models were generated by combining h-iPSC-CMs (iCell Cardiomyocytes from FUJIFILM Cellular Dynamics Inc.) and adult human dermal fibroblasts (Lonza) in a ratio of 75% h-iPSC-CMs and 25% human fibroblasts, as previously reported [15]. H-iPSC-CMs were thawed according to standard protocol and immediately used. Human fibroblasts were expanded in DMEM (Gibco, 4.5 g l⁻¹ glucose) supplemented with 10% v/v FBS, 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin and 10 mmol hepes and used at confluency. Human fibroblasts were used till passage 11. Cells were embedded at a density of 125 × 10⁶ cells ml⁻¹ in fibrin hydrogel prepared as previously described to reach a final concentration of 10 mg ml⁻¹ of human fibrinogen and 1.25 U ml⁻¹ of human thrombin. The hydrogel solution (i.e. roughly 2 µl per each cell culture chamber) was subsequently dispensed in uHeart and let crosslink at 37 °C for 8 min before manually adding culture media. Plating medium (FUJIFILM Cellular Dynamics Inc.) supplemented with 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin and 2 mg ml⁻¹ of ACA was used for the first 24 h of culture. Subsequently, RPMI (Sigma Aldrich) medium supplemented with 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin, ACA and B-27® was used and changed every day for the following 6 days. The amount of ACA in the medium was gradually diminished as previously described for rat microtissues. Mechanical stimulation [18] (10% uniaxial strain, frequency 1 Hz) was started after 4 h from seeding.

Experiments were performed by maintaining uHeart platforms in the incubator at 37 °C with 5% CO₂. FP of the developed rat and human microtissues were recorded as soon as the cells started to spontaneously beat. The measuring electrodes (stainless-steel microelectrode of 120 µm in diameter and 3 cm long or parylene coated tungsten sharp electrodes with a tip diameter of 10 µm) were firstly manually inserted in the micro-electrode coaxial channels till the separating wall was punched and the tip reached the desired position close to the microtissue. An AgCl ground electrode was positioned into one of the four cell culture medium reservoirs. An extracellular amplifier (Ext-02b, Npi Electronic GmbH, Germany) was used to amplify (10 k Gain) and filter (3 Hz high pass, 10 kHz low pass) simultaneously up to two electrophysiological signals. FPs were acquired and recorded with a rate of 2000 sample s⁻¹ by means of Waveform software of an electronic board (Analog Discovery 2, Digilent, Washington, USA) connected to a personal computer.

Immunofluorescence analyses were performed directly within the microdevices on rat and human cardiac microtissues after 5 or 7 days of culture, respectively. Samples were fixed in 4% paraformaldehyde for 30 min and subsequently permeabilized for 1 h at room temperature with 0.1% v/v Triton X-100 (Thermo Fisher Scientific) in a solution of 2% w/v bovine serum albumin (BSA, Sigma Aldrich) dissolved in PBS, to block nonspecific binding. Microtissues were incubated overnight at 4 °C with primary antibodies in a solution of 0.5% w/v BSA: cardiac cells were stained with troponin I (mouse, dilution 1:100; Santa Cruz), fibroblasts and non-myocytes cells with vimentin (rabbit, dilution 1:500; Abcam) and cell–cell electrical coupling was assessed by connexin-43 (rabbit, dilution 1:500; Abcam). Goat anti-mouse (Thermo Fisher Scientific, rodamine conjugated) and goat anti-rabbit (Thermo Fisher Scientific, FITC conjugated) secondary antibodies were diluted 1:200 in 0.5% w/v BSA and incubated in the dark for 6 h at 4 °C. 4',6-diamidino-2-phenylindole (DAPI, Invitrogen) nuclear staining was used to identify cell nuclei. Images were acquired by means of a confocal microscope (Fluoview FV10i, Olympus).

Total RNA extraction using TRIzol (Sigma), complementary DNA synthesis and quantitative real-time reverse transcriptase-polymerase chain reaction (RT-qPCR; 7300 AB Applied Biosystems) were performed according to standard protocols. In detail, after two steps of PBS washing, the uHeart platform was disassembled by manually separating the cell culture layer and the actuation, so to expose the microtissue. An aliquot of 8 µl of TRIzol were then dropped on it and cells were mechanically scratched by means of a pipette tip. The resulting cell-laden fibrin gel dissolved in TRIzol was finally harvested and transferred in a 1.5 ml Eppendorf tube containing 300 µl of TRIzol, to proceed with the RNA extraction, as previously described [18, 60]. Expression levels of the following genes of interest were specifically quantitated (Applied Biosystems): CACNA1C (Hs00167681_m1), SCN5A (Hs00165693_m1), KCNH2 (Hs04234270_g1) and KCNQ1 (Hs00923522_m1). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as reference housekeeping gene (Hs02758991_g1). Three and six biologically independent samples were considered for t0 (i.e. mix of iCell-CM and fibroblast before their embedding in fibrin) and t7 (i.e. 7 days mechanically trained cardiac microtissues within uHeart), respectively.

All drugs were purchased from Sigma Aldrich and stock solutions were prepared in dimethyl sulfoxide (DMSO) at 1000× the maximum target concentration. Serum-free medium was used to serially dilute the drug stock solution to reach the desired concentrations for the test. A schematic of the experimental protocol is represented in figure S4(b). Before starting the assay, the cell culture medium was exchanged with fresh serum-free medium: cardiac complete medium for rat microtissues and RPMI supplemented with B27 for human cells. After 30 min, a 3 min baseline signal was recorded. In each microtissue, the drugs were administered at increasing concentrations and were incubated for 10 min before performing a 3 min recording of the FP. After each drug concentration, a 5 min washing with fresh media was performed to prevent drug accumulation. Compounds belonging to different risk category, according to CiPA [61], to develop TdP were tested. In details, for human microtissues the selected compounds were: sotalol (high risk; 1–3.25–7.5–15–30–60 µmol), verapamil (low risk; 1–10–30–50–100–1000 nmol), aspirin (negative control; 1–3–10–30–100–300 µmol) and DMSO (vehicle; 0.1%–0.2%–0.3%–0.4%–0.5%–0.6%). To assess the inter-species differences, rat microtissues were tested with: sotalol (high risk; 7.5–15–30–60 µmol) and verapamil (low risk; 1–10–100–1000 nmol). For each compound, different concentrations were chosen taking into account studies related to MEA system and C_max in clinics.

Data were analyzed by means of a custom-made MATLAB code. In order to obtain the characteristic FP curves for each drug concentration, a single recorded signal was firstly smoothened to decrease the noise and then averaged. For each drug, baseline curve and representative traces of averaged FP recorded at incremental drug concentrations were plotted overlaid. From the recorded signals, the following parameters were also quantified: FPD (ms), as the time from the start of the depolarizing spike to the peak of the repolarization wave, the spontaneous beating (RR interval, ms) and the spike amplitude (µV). The FPD was corrected with the Fridericia's method (FPDc = FPD (RR)^−1/3) [62]. The percent change in FPDc following each concentration of drug with respect to the baseline was computed and plotted. FP recordings were collected from at least three individual cardiac microtissues (n ⩾ 3) for each drug tested.

The results of the electrical activity monitoring (i.e. the quantified delays between peaks from the two extremities of the microtissues) of rat cardiac microtissues are presented as mean with SD.

The results of RT-qPCR, are presented as mean with SD. Statistical analysis was made by Mann–Whitney test (non-normal distributions; *P < 0.05; **P < 0.01; ***P < 0.001).

The results of FPDc (%) after drug administration are presented as mean with SE. Statistical analysis was made through Kruskal–Wallis with Dunn's multiple comparison test (non-normal distributions; *P < 0.05; **P < 0.01; ***P < 0.001).

Statistical analyses were performed using GraphPad Prism 8.