Development of a drug screening system using three-dimensional cardiac tissues containing multiple cell types

We hypothesized that an appropriate ratio of cardiomyocytes, fibroblasts, endothelial cells, and extracellular matrix (ECM) factors would be required for the development of three-dimensional cardiac tissues (3D-CTs) as drug screening systems. To verify this hypothesis, ECM-coated human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs), ECM-coated cardiac fibroblasts (CFs), and uncoated cardiac endothelial cells (CEs) were mixed in the following ratios: 10:0:0 (10CT), 7:2:1 (7CT), 5:4:1 (5CT), and 2:7:1 (2CT). The expression of cardiac-, fibroblasts-, and endothelial-specific markers was assessed by FACS, qPCR, and immunostaining while that of ECM-, cell adhesion-, and ion channel-related genes was examined by qPCR. Finally, the contractile properties of the tissues were evaluated in the absence or presence of E-4031 and isoproterenol. The expression of ECM- and adhesion-related genes significantly increased, while that of ion channel-related genes significantly decreased with the CF proportion. Notably, 7CT showed the greatest contractility of all 3D-CTs. When exposed to E-4031 (hERG K channel blocker), 7CT and 5CT showed significantly decreased contractility and increased QT prolongation. Moreover, 10CT and 7CT exhibited a stronger response to isoproterenol than did the other 3D-CTs. Finally, 7CT showed the highest drug sensitivity among all 3D-CTs. In conclusion, 3D-CTs with an appropriate amount of fibroblasts/endothelial cells (7CT in this study) are suitable drug screening systems, e.g. for the detection of drug-induced arrhythmia.

Many drugs are withdrawn from the market due to adverse side effects 1 . Cardiotoxicity, including cytotoxicity and proarrhythmic effect, is a major cause of discontinuation of drug development [2][3][4][5] .
The current screening methods are based on animal models, but species-specific differences in drug effectiveness and safety may be observed 6,7 . Moreover, also due to growing political and social pressure, various attempts to develop systems alternative to animal models have recently been made [8][9][10][11] . A particularly promising approach is represented by innovative in vitro drug screening systems based on human cells, including human induced pluripotent stem cells (hiPSCs).
We have previously developed a new drug screening system to test drug-induced cardiotoxicity in vitro, based on hiPSC-derived three-dimensional cardiac tissues (3D-CTs), which showed good response to some cardiovascular agonists 12 . However, the contractile properties of these systems are still poor as compared to those of myocardial tissues in vivo. In addition, we previously observed that although E-4031 administration decreases in vitro contractility, it does not elicit arrhythmia or Torsades de Pointes (TdP), which are normally observed in vivo under the same conditions. In light of the above results, we concluded that 3D-CTs did not completely reproduce the characteristics of in vivo myocardial tissues and that, therefore, appropriate adjustments may be necessary before using 3D-CTs as accurate tools for drug response prediction.

Discussion
In this study, we created three-dimensional cardiac tissues containing various proportions of hiPSC-CMs, www.nature.com/scientificreports/ fibroblasts, and endothelial cells, and investigated their pathophysiology and suitability as a drug screening tool. The expression of cardiac-specific proteins and ion channel-related genes in 3D-CTs increased with the proportion of hiPSC-CMs, whereas the expression of ECM components and cell-adhesion factors depended on the proportion of fibroblasts and the presence of endothelial cells. 7CT exhibited the best contractility in the absence of any cardiovascular agonists. When exposed to low concentrations of proarrhythmic drugs, 7CT and 5CT exhibited a decrease in relaxation velocity and an increase in CRPI and CRDcf compared to 10CT. In particular, 7CT and 5CT showed higher reactivity to these drugs than did the other 3D-CTs. In addition, when exposed to a positive inotropic drug, 10CT and 7CT exhibited the strongest and dose-dependent changes in contraction-relaxation velocity and CRPI. www.nature.com/scientificreports/ Overall, 7CT and 5CT showed better contractility than did 10CT and 2CT, indicating that 3D-CTs with extremely low or high proportions of fibroblasts may not be suitable for drug screening.
Fibroblasts and endothelial cells directly and indirectly regulate cardiac electrophysiology. The interaction between cardiomyocytes and these cells allows for electrical propagation in detached cardiomyocytes, and promotes electrical conduction. The binding of cardiomyocytes to depolarized fibroblasts increases the resting membrane potential, which may result in increased conduction velocity, provided that fibroblasts are present at the appropriate density [17][18][19][20][21][22][23] . In addition, the adhesion of these cells to cardiomyocytes affects the electrical conduction. By producing ECM components and matrix metalloproteinases, fibroblasts and endothelial cells promote cell-to-cell adhesion and fill the gaps between cells [24][25][26][27] . Insufficiently strong cell-to-cell adhesion results in increased cardiomyocyte resting membrane potential and conduction velocity. An excessive proportion of fibroblasts or endothelial cells may result in abnormally strong cell-to-cell adhesion, possibly slowing down, and eventually impairing, electrical conduction, as suggested by previous reports 18,22,24,28 . Therefore, it can be concluded that an optimal proportion of fibroblasts and endothelial cells is crucial to preserve electrical conduction.
Our results showed that 2CT, which is rich in fibroblasts, highly expressed ECM components and exhibited the weakest contractility. 2CT also showed a lower inter-region correlation coefficient than did the other 3D-CTs, indicating the synchronization of cardiomyocyte contraction. 2CT may not be suitable as a model to study contraction, due to the low number of cardiomyocytes and the excessive number of fibroblasts, causing insufficient contractility and impaired conduction, respectively.
7CT, in which the proportions of hiPSC-CMs, CFs and CEs were 70%, 20% and 10%, respectively, showed the best contractile performance. In this 3D-CT, the proper strength of cell-to-cell adhesions and the appropriate proportions of ECM factors, fibroblasts, and endothelial cells ensured an optimal balance between contraction and electrical conduction.
In myocardial tissue in vivo, cardiomyocytes and fibroblasts account for 30% and 70% of the total cell number, respectively, with 70% of the cell volume accounted for by cardiomyocytes and 30% by fibroblasts 15 . In our optimal 3D-CT model, 70% of the cells were cardiomyocytes and 20% were fibroblasts, while the 80% of the cell volume was constituted by cardiomyocytes and 20% by fibroblasts, presenting a model histologically comparable to the native heart 15,18 .
Next, the suitability of the different 3D-CTs as drug screening models was explored. To evaluate proarrhythmic and negative chronotropic effects, a HERG potassium channel blocker, E-4031, acting on ion channels at the cell surface, was employed. Our results showed that 7CT and 5CT, with CF proportions of 20% and 40%, respectively, reproduced proarrhythmic and negative chronotropic effects in an E-4031 dose-dependent manner, whereas 10CT, containing only hiPSC-CMs, showed little response to E-4031. Previous studies reported that TdP do not occur in cardiac tissues that contain only cardiomyocytes, in line with our findings [28][29][30] . In the present study, the exposure of 2CT, containing 70% CFs, to a high concentration of E-4031 resulted in progressive weakening and in the eventual stop of contraction. The block of action potential in 2CT might be due to the lack of depolarization in the presence of high E-4031 concentrations. This behavior was different from that observed in vivo, suggesting that 2CT was not suitable for the evaluation of drugs causing QT prolongation.
To evaluate positive inotropic effects, we used isoproterenol, a non-selective beta adrenoceptor agonist, which increases cardiac contractility via catecholamine receptors. Since catecholamine receptors are present in cardiomyocytes, we reasoned that 3D-CTs with a high content of hiPSC-CMs would show a better contractile performance than those with lower hiPSC-CM proportions. As expected, 10CT and 7CT, containing an hiPSC-CM proportion of 100% and 70%, respectively, as well as a high content of cardiomyocytes, exhibited a concentration-dependent increase in contractility after treatment with isoproterenol. Although 7CT exhibited a sensitive response to typical positive and negative inotropic drugs in this study, it is necessary to test this system using more drugs that act through various mechanisms. In addition, we plan to further investigate drug response by analyzing calcium transient, extracellular potential, and contraction force, in addition to tissue contractile motion, to provide a more complete analysis of cellular physiology. In this study, 5CT showed a drug response that showed proarrhythmic and negative chronotropic effects as sensitively as 7CT, suggesting that it may be used as a cardiac fibrosis model in the future. Standardization of this model for drug screening requires homogeneous cell quantity and quality. In the construction of the described 3D-CTs, the cell proportions can be precisely controlled and, therefore, tissue constructs with approximately the same number of cells can be obtained. Regarding the cellular quality, further optimization is needed to ensure phenotypic homogeneity and controlled maturation of cardiomyocytes after cardiac differentiation.
In conclusion, 3D-CTs with various cellular compositions were tested and 7CT (70% hiPSC-CMs, 20% CFs, 10% CEs) showed optimal contractility and good response to inotropic drugs. In vitro cellular models capable of recapitulating cardiac physiology are strongly desired for drug screening. Thus, 3D-CTs with appropriate cellular combinations are reliable systems for in vitro drug screening.
Cardiac differentiation of hiPSCs. Cardiac differentiation was induced as previously described with modifications 12 . In brief, hiPSCs were dissociated using Accumax (Innovative Cell Technologies, San Diego, Histological stain. The 3D-CTs were fixed with formalin and embedded in paraffin. The paraffin 3D-CTs blocks were then sectioned (0.5 µm thickness) using a microtome (HM430; ThermoFisher Scientific) and the sections were then stained with hematoxylin-eosin and assessed using a microscope (DM4000B; Leica, Germany). Quantitative real-time polymerase chain reaction. Total RNA was extracted from 3D-CTs using a PureLink RNA Mini Kit (ThermoFisher Scientific). Quantitative real-time polymerase chain reaction (qPCR) were performed as previously described 12 . All data were normalized using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a control, and assessed using the delta CT method. The expression of β-MHC and α-MHC were normalized using cTnT. All qPCR data are shown as relative values compared to 10CT. The primer sequences are listed in Table 1.

Cell motion analyses.
Cell motion analyses were conducted using a Cell Motion Imaging System (SI8000; SONY, Tokyo, Japan). Motion images of the 3D-CTs were recorded at a rate of 150 frames per s, a resolution of 1024 × 1024 pixels, and a depth of 8 bits. Cell motion was recorded for 10 s after cumulative exposure for 10 min to each E-4031 concentration (1, 3, 10, 30, 100, 300, 1000 nM; Calbiochem Merck Millipore, Darmstadt, Germany) or isoproterenol (Calbiochem Merck Millipore). The beating rate, contraction velocity, relaxation velocity, contraction-relaxation peak interval (CRPI), contraction-relaxation duration (CRD), and inter-region correlation coefficient were analyzed using an SI8000C analyzer (SONY). The value of CRD was normalized to the beat rate, using Fredericia correction (CRDcf) 34 . The beating rate, contraction velocity, relaxation velocity, CRPI, and CRDcf are shown as relative values compared to vehicle-treated controls. The effects of treatments on these parameters were examined by comparison with vehicle-treated controls. When an abnormal waveform due to the addition of E-4031 was observed, the subsequent values were excluded from the analysis (Fig. 5B).
Since the inter-region correlation coefficient indicates cooperative cardiomyocyte contraction 35 , only the negative value, but not the abnormal waveform, was excluded.
Statistical analyses. The data were expressed as the mean ± standard deviation. Statistical significance was determined by Student's t-test (2-tailed) adjusted with the Bonferroni correction in the experiments for gene expression, in comparison with 10CT. Dunnett test was used to assess significance in the assessments of contractile properties, performed using R software (freely available from https ://www.r-proje ct.org/). *P < 0.05 was considered statistically significant and **P < 0.01.

Data availability
All data generated or analyzed during this study are included in this published article.