Heart Muscle Microphysiological System for Cardiac Liability Prediction of Repurposed COVID-19 Therapeutics

Despite global efforts, it took 7 months between the proclamation of global SARS-CoV-2 pandemic and the first FDA-approved treatment for COVID-19. During this timeframe, clinicians focused their efforts on repurposing drugs, such as hydroxychloroquine (HCQ) or azithromycin (AZM) to treat hospitalized COVID-19 patients. While clinical trials are time-consuming, the exponential increase in hospitalizations compelled the FDA to grant an emergency use authorization for HCQ and AZM as treatment for COVID-19, although there was limited evidence of their combined efficacy and safety. The authorization was revoked 4 months later, giving rise to controversial political and scientific debates illustrating important challenges such as premature authorization of potentially ineffective or unsafe therapeutics, while diverting resources from screening of effective drugs. Here we report on a preclinical drug screening platform, a cardiac microphysiological system (MPS), to rapidly identify clinically relevant cardiac liabilities associated with HCQ and AZM. The cardiac MPS is a microfabricated fluidic system in which cardiomyocytes derived from human induced pluripotent stem cells self-arrange into a uniaxially beating tissue. The drug response was measured using outputs that correlate with clinical measurements such as action potential duration (proxy for clinical QT interval) and drug-biomarker pairing. The cardiac MPS predicted clinical arrhythmias associated with QT prolongation and rhythm instabilities in tissues treated with HCQ. We found no change in QT interval upon acute exposure to AZM, while still observing a significant increase in arrhythmic events. These results suggest that this MPS can not only predict arrhythmias, but it can also identify arrhythmias even when QT prolongation is absent. When exposed to HCQ and AZM polytherapy, this MPS faithfully reflected clinical findings, in that the combination of drugs synergistically increased QT interval when compared to single drug exposure, while not worsening the overall frequency of arrhythmic events. The high content cardiac MPS can rapidly evaluate the cardiac safety of potential therapeutics, ultimately accelerating patients’ access to safe and effective treatments.


Supplementary Figure 2. Drug absorption in the device components and free drug in media.
Absorption studies for dose escalation exposure to HCQ alone (A), or AZM alone (B). A dose escalation study without cells was setup as shown in Supplementary Fig. 1. Effluent collection was started 3 minutes after switching to the respective dose to allow washout of the previous dose and stopped as the pump switched to the next higher dose. Graphs show normalized values to HCQ and AZM controls of the respective dose without perfusion through the microfluidic system. No significant drug loss was observed at any dose. (C) Graph showing the percentage of drug (HCQ or AZM) free fraction in either experimental media (maturation media) or buffer (blank, isotonic sodium phosphate). Differences between the media and buffer were not significant.

In-house Python script
Post-experiment processing was performed with an in-house Python library. This library has integrated automated background subtraction and normalization, and eliminates bleaching baseline drift by fitting a polynomial to the baseline via half-quadratic minimization (Mazet et al., 2005). The resulting traces were used for quantitative analysis of the action potential by calculating metrics such as 80% and 30% action potential duration (APD80 and APD30), triangulation ((APD80-APD30)/ APD80) and beat rate. Similar metrics were calculated for calcium transient (CaD80 and CaD30). Poincaré plots were generated by the same library, by plotting CaD80 of each (n th ) beat, against CaD80 of the preceding beat (n-1) th , normalized to the CaD80 mean. Identical CaD80 values in sequence appear as a single point, stable CaD increase or decrease (anti-arrhythmic) will cluster around the center of the graph and large deviations between successive CaD80 (pro-arrhythmic), points will deviate from the center giving rise to disorganized polygons (Hondeghem et al., 2001).

Drug absorption into PDMS
In order to quantify the actual drug concentration available to the microtissues, drug absorption into the device (PDMS, Polydimethylsiloxane, or tubing) was measured. Drug absorption experiments were performed and analyzed via liquid chromatography-mass spectrometry (LC-MS/MS) for HCQ (Supplementary Fig. 2A) or AZM (Supplementary Fig. 2B). The drug doses were prepared the same way as for the cell experiments. The experimental setup of acute studies was replicated in the absence of cells and the effluent collection was started 3 min after switching to the respective dose. Additional samples from freshly prepared drug doses were collected for controls. All samples were immediately frozen at -80°C.

Drug free fraction in media
The binding of test compounds to plasma proteins is an important factor affecting drug efficacy, metabolism and pharmacokinetic properties. In many cases, drug efficacy is determined by the concentration of free drug (unbound), rather than the total concentration in plasma. If the drug is highly bound to plasma proteins, the amount of drug available to reach the target is reduced. We therefore measured the portion of free drug in the media with RED (rapid equilibrium analysis) assays (Thermo Scientific, 90006) to determine the media-bound versus free fraction of drug following manufacturer's instructions (Supplementary Fig. 2C). Briefly, experimental media (maturation media, MM) spiked with test compound was added to the left chamber of a commercial plate based RED (rapid equilibrium dialysis) device. Blank, isotonic sodium phosphate buffer was added to the outer chamber of the RED device and the plate was incubated at 37°C. Aliquots of the buffer and MM were taken at predetermined time points and the concentration of free and bound test compound was determined by LC-MS/MS analysis.

LC-MS/MS analysis
Samples (10 µL) were chromatographed on a ZORBAX SSHD Eclipse Plus C18 column (3.0 x 50 mm, 1.8 μm, catalog no. 959757-302; Agilent, Santa Clara, CA) with a guard column (2.1 x 5 mm, 1.8 µm, catalog no. 821725-901; Agilent) via an Agilent 1290 Infinity II LC system. Column temperature and the LC flow rate were set at 40°C and 0.4 ml/min. Initial chromatographic condition was maintained at 95% mobile phase A (water with 0.1% formic acid, v/v) and 5% mobile phase B (acetonitrile with 0.1% formic acid, v/v) for one min, then increased to 80% B by 3 min, then to 95% B by 4 min, and then returned to initial condition at 5 min until 8 min for sufficient equilibrium. All MS/MS analyses were performed in positive ion mode with an electrospray ion (ESI) source using an Agilent 6470 triple-quadruple mass spectrometer. The capillary voltage was set at 3500 V. The nebulizer gas pressure and gas temperature were set at 35 psi and 350°C, respectively. The MS/MS transitions, collision energy (CE), and fragmentor energy (FE) were set for the detection of AZM (m/z 749.5→591.4, CE=29 V, FE=160 V) and HCQ (m/z 336.2→247.1, CE=21 V, FE=106 V).