**3D Stem Cell Culture**

Editor

**Joni H. Ylostalo**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editor* Joni H. Ylostalo University of Mary Hardin-Baylor USA

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Cells* (ISSN 2073-4409) (available at: https://www.mdpi.com/journal/cells/special issues/3D culture).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-03943-803-7 (Hbk) ISBN 978-3-03943-804-4 (PDF)**

c 2020 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**



## **About the Editor**

**Joni H. Ylostalo** is an Associate Professor at the University of Mary Hardin-Baylor (UMHB). Dr. Ylostalo obtained his BSc in Biochemistry and MSc in Biochemistry, with a focus in Biotechnology and Molecular Biology, from the University of Oulu. His MSc research involved biochemical studies of type IX collagen. Dr. Ylostalo obtained his PhD in Biomedical Sciences from Tulane University, where he worked with mesenchymal stem cells (MSCs). His research was focused on gene expression changes of MSCs during differentiation and culture expansion. Dr. Ylostalo completed his postdoctoral work at Texas A&M, studying the aggregation and 3D culture of MSCs to enhance their therapeutic potential. He continued his MSC research at Texas A&M as a faculty member focusing on the therapeutic applications of 3D culture-activated MSCs. Since transitioning to UMHB, he has continued his MSC research among other research and service opportunities. Dr. Ylostalo has published over 50 articles that span the fields of biochemistry, cell biology, bioinformatics, regenerative medicine, and biology education. He has served as an editor and reviewer for numerous journals while giving plenary talks at various conferences.

## *Editorial* **3D Stem Cell Culture**

#### **Joni H. Ylostalo**

Department of Biology, University of Mary Hardin-Baylor, 900 College Street, Box 8432, Belton, TX 76513, USA; jylostalo@umhb.edu; Tel.: +1-254-295-5534

Received: 16 September 2020; Accepted: 24 September 2020; Published: 27 September 2020

**Abstract:** Much interest has been directed towards stem cells, both in basic and translational research, to understand basic stem cell biology and to develop new therapies for many disorders. In general, stem cells can be cultured with relative ease, however, most common culture methods for stem cells employ 2D techniques using plastic. These cultures do not well represent the stem cell niches in the body, which are delicate microenvironments composed of not only stem cells, but also supporting stromal cells, extracellular matrix, and growth factors. Therefore, researchers and clinicians have been seeking optimal stem cell preparations for basic research and clinical applications, and these might be attainable through 3D culture of stem cells. The 3D cultures recapitulate the in vivo cell-to-cell and cell-to-matrix interactions more effectively, and the cells in 3D cultures exhibit many unique and desirable characteristics. The culture of stem cells in 3D may employ various matrices or scaffolds, in addition to the cells, to support the complex structures. The goal of this Special Issue is to bring together recent research on 3D cultures of various stem cells to increase the basic understanding of stem cells and culture techniques, and also highlight stem cell preparations for possible novel therapeutic applications.

**Keywords:** stem cell; 3D; culture condition; expansion; niche; regenerative medicine; scaffold; organoid

Stem cells are cells that demonstrate the abilities to self-renew and differentiate. Many types of stem cells can be isolated from embryonic or adult tissues, varying in their potency from pluripotent to unipotent depending on the stem cell type. Many cells can be found throughout the human body, often localized in niches that provide a nurturing microenvironment for the stem cells while directing their proliferation and differentiation. Stem cells are typically cultured in 2D tissue culture plastic for ease and maximal expansion, as large numbers are often needed for translational research and therapies. The 2D cultures do not, however, represent the natural environment of stem cells in the body well. Many beneficial properties of stem cells might be lowered or even lost in 2D cultures. Therefore, the use of 3D culture techniques has become more common in basic and translational research. The 3D cultures often provide more complete cell-to-cell and cell-to-matrix interactions, mimicking the natural environment in which the stem cells reside better than the traditional 2D cultures. Furthermore, many desired cellular characteristics are maintained or even promoted in 3D cultures, further supporting their use in basic and translational research. In this Special Issue, recent advances in the 3D cultures of stem cells are highlighted, bringing together a collection of articles presenting various stem cell types and their characteristics in 3D environments.

This Special Issue includes articles highlighting the effective use of induced pluripotent stem cells (iPSCs) to develop cardiac microtissue [1], neurospheres [2], and cortical progenitors [3]. In addition, this Special Issue contains work on scaffolds with mesenchymal stem cells (MSCs) [4] and embryonic stem cells (ESCs) [5]. Furthermore, research on generating intestinal organoids from stem cells [6,7] and nephrogenesis studies utilizing ESCs [8] are included in this Special Issue, along with a comprehensive review on the 3D culture of hematopoietic stem cells (HSCs) [9].

A very relevant problem in cardiology is cardiac fibrosis. In the work by Blyszczuk et al., a model of fibrotic cardiac microtissue was generated using iPSC-derived cardiomyocytes and cardiac fibroblasts by treatment with transforming growth factorβ1 (TGF-β1) or by use of cardiac fibroblasts from heart failure patients [1]. The authors demonstrated that activated cardiac fibroblasts could, via direct stimulation of β-adrenoreceptor signaling, promote cardiac contraction rate. Furthermore, the generated model could be used as a high-throughput model for drug testing or general studies of cardiac fibrosis [1].

A 3D in vitro model utilizing iPSCs was generated by Kobolak et al. to study neurotoxicity [2]. In this work, iPSC-derived, 3D, free-floating neurospheres, exhibiting the various cells of the nervous system, were developed and tested in neurotoxicity studies. Furthermore, these neurospheres could open some further opportunities for the detection of developmental neurotoxicity, and hence support the existing animal models [2].

The work by Cutarelli et al. employed iPSC-derived cortical progenitors and scaffolds to mimic the radially oriented cortical radial glia fibers that play an important role in the development of the cerebral cortex [3]. The authors used silicon vertical micropillar arrays to promote the expansion and preservation of stemness of the cortical progenitors. Furthermore, the described model of iPSC-derived cortical progenitors and silicon micropillars could be used in cortical tissue engineering [3].

Another in vitro 3D model was generated by Ciardulli et al. for the study of MSCs and their tenogenic differentiation [4]. This work employed a hyoluronate/poly-lactic-co-glycolic acid (PLGA)/fibrin 3D scaffold with MSCs that were studied under static and cyclic-strain conditions. The research demonstrated that MSCs grown in this scaffold increased their tenogenic marker and pro-repair cytokine expression, supporting the notion of this model as a potential predictive system to be used in future studies employing the sustained release of biochemicals [4].

To circumvent some of the challenges with the 2D culture expansion of ESCs, McKee et al. developed a 3D scaffold to mimic in vivo stem cell niches [5]. The authors used polyethylene glycol (PEG) polymers with thiol and acrylate end-groups to guide the self-assembly of the scaffolds. ESCs grown in these scaffolds were able to maintain their viability, self-renewal, and differentiation potential. ESCs in the scaffolds also exhibited a high expression of pluripotency markers and some mechanosensitive genes, supporting the notion of these scaffolds as potentially helpful for many ESC studies [5].

To aid in the study of intestinal disorders, Kramer et al. developed intestinal organoids from stem cells obtained from various parts of the intestine [6]. These intestinal organoids showed long-term expansion of the cells while inhibiting cell differentiation when cultured in expansion media, while differentiation into goblet and enteroendocrine cells was promoted with culture in differentiation media. These organoids could possibly be employed in various in vitro models of functional intestinal disorders [6].

Another study utilizing intestinal stem cell organoids explored the cellular origin of carcinogenesis [7]. In this study, Yen et al. developed a tumorigenesis model based on carcinogenesis and genetically engineered mice. The authors were able to demonstrate the interplay between extrinsic carcinogen and intrinsic genetic modification and their contribution towards transformation while elucidating the involvement of molecular factors, such as protein phosphatase 2A (PP2A), in this process [7].

A study by Tan et al. utilized ESCs to generate ureteric bud progenitor cells [8]. The authors then used these progenitor cells to induce nephrogenesis in co-culture with primary metanephric mesenchyme. These kidney organoids exhibited nephron structures with collecting ducts connected to nephron tubules. This study demonstrated a relatively simple and reproducible way of generating ureteric bud progenitors [8].

This Special Issue includes articles highlighting recent discoveries in 3D stem cell cultures using MSCs [4], ESCs [5,8], iPSCs [1–3], and intestinal stem cells [6,7]. This Special Issue also includes a review article by Ribeiro-Filho et al. that discusses the 2D and 3D culture of hematopoietic stem cells (HSCs) as a relevant model to study both normal and abnormal hematopoiesis [9]. Furthermore, this Special

Issue includes articles that utilize various scaffolds together with stem cells from various sources to study 3D cultures [3–5].

**Funding:** This research received no external funding.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Activated Cardiac Fibroblasts Control Contraction of Human Fibrotic Cardiac Microtissues by a** β**-Adrenoreceptor-Dependent Mechanism**

**Przemysław Błyszczuk 1,2,\*, Christian Zuppinger 3, Ana Costa 4, Daria Nurzynska 5, Franca Di Meglio 5, Mara Stellato 1, Irina Agarkova 6, Godfrey L. Smith 4, Oliver Distler <sup>1</sup> and Gabriela Kania 1,\***


Received: 13 April 2020; Accepted: 19 May 2020; Published: 20 May 2020

**Abstract:** Cardiac fibrosis represents a serious clinical problem. Development of novel treatment strategies is currently restricted by the lack of the relevant experimental models in a human genetic context. In this study, we fabricated self-aggregating, scaffold-free, 3D cardiac microtissues using human inducible pluripotent stem cell (iPSC)-derived cardiomyocytes and human cardiac fibroblasts. Fibrotic condition was obtained by treatment of cardiac microtissues with profibrotic cytokine transforming growth factor β1 (TGF-β1), preactivation of foetal cardiac fibroblasts with TGF-β1, or by the use of cardiac fibroblasts obtained from heart failure patients. In our model, TGF-β1 effectively induced profibrotic changes in cardiac fibroblasts and in cardiac microtissues. Fibrotic phenotype of cardiac microtissues was inhibited by treatment with TGF-β-receptor type 1 inhibitor SD208 in a dose-dependent manner. We observed that fibrotic cardiac microtissues substantially increased the spontaneous beating rate by shortening the relaxation phase and showed a lower contraction amplitude. Instead, no changes in action potential profile were detected. Furthermore, we demonstrated that contraction of human cardiac microtissues could be modulated by direct electrical stimulation or treatment with the β-adrenergic receptor agonist isoproterenol. However, in the absence of exogenous agonists, the β-adrenoreceptor blocker nadolol decreased beating rate of fibrotic cardiac microtissues by prolonging relaxation time. Thus, our data suggest that in fibrosis, activated cardiac fibroblasts could promote cardiac contraction rate by a direct stimulation of β-adrenoreceptor signalling. In conclusion, a model of fibrotic cardiac microtissues can be used as a high-throughput model for drug testing and to study cellular and molecular mechanisms of cardiac fibrosis.

**Keywords:** cardiac microtissues; iPSC-derived cardiomyocytes; cardiac fibroblasts; cardiac fibrosis; cardiac rhythm; TGF-β signalling; drug screening; in vitro model

#### **1. Introduction**

Cardiac fibrosis refers to an excessive accumulation of stromal cells and extracellular matrix (ECM) proteins in the myocardium and represents a common pathophysiological scenario in a broad range of cardiovascular conditions, including myocardial infarction, hypertension, myocarditis, and hypertrophic or dilated cardiomyopathy [1]. Cardiac fibroblasts and myofibroblasts represent the

most extensively characterised stromal cell types involved in fibrotic processes in the heart [2]. In the traditional view, resident cardiac fibroblasts become activated and overproduce ECM proteins such as collagen type I and III and fibronectin. Fibrogenesis is regulated by multiple profibrotic inputs, which activate a complex signalling network. In cardiac fibrosis, transforming growth factor β (TGF-β) is recognised as the key profibrotic cytokine activating quiescent cardiac fibroblasts. Constitutive overexpression of TGF-β was shown to induce interstitial cardiac fibrosis and cardiac hypertrophy in transgenic mice [3], whereas targeting TGF-β or its downstream mediators successfully reduced or prevented cardiac fibrosis in various animal models [4–6]. Progressive fibrosis causes tissue stiffening and may affect cardiac rhythm. Clinical data associate fibrotic heart condition with increased resting heart rate [7] and with elevated risk of developing life-threating arrhythmias [8,9]. The underlaying mechanisms remain obscure, however.

In the body, heart rhythm established by pacemaker cells located in the sinoatrial node of the heart is mainly under control of the autonomic nervous system. Sympathetic stimulation increases the rate, whereas parasympathetic stimulation decreases it. Neurotransmitter noradrenaline released by the sympathetic neurons increases heart rate by activating β-adrenergic receptors on cardiomyocytes. Under stress condition, β-adrenoreceptors are further activated by adrenaline—a catecholamine produced by the adrenal glands and released into the circulation. Binding of agonists to β-adrenergic receptors causes the formation of cyclic adenosine monophosphate and activation of protein kinase A that phosphorylates a number of proteins including myosin light chains and thus increases the ability of cardiomyocytes to relax following excitation contraction coupling [10]. In the heart, the increased relaxation is associated with increased rate and strength of contraction.

Because extracardiac stimuli are important regulators of the heart rhythm, it is difficult to study intracardiac mechanisms of cardiac contractility using in vivo systems only. Data from ex vivo and in vitro cardiac models may shed more light on specific aspects of intracardiac rhythm regulation. Disease models using human cells offer a natural advantage over animal models by providing a proper genetic and cellular context. In cardiovascular research, difficulties in obtaining and culturing contractile primary human cardiomyocytes limited the development of in vitro human models for many years. The human induced pluripotent stem cell (iPSC) technology opened new perspectives and became a commonly used, unlimited source of spontaneously beating cardiomyocytes for regenerative medicine, drug discovery and toxicity screening [11].

A lack of the physiologically relevant experimental setting is a major drawback of many in vitro models. In general, traditional two-dimensional (2D) cell culture systems poorly mirror the biomechanical and biochemical microenvironment of the tissue. Three-dimensional (3D) cell culture technologies, instead, much better reflect the organ complexity and represent an attractive alternative to the commonly used 2D models. 3D cell culture systems, called microtissues or organoids, include cellular models on various bioscaffolds, but also scaffold-free spheroids comprising of one or more cell types. In recent years, various types of cardiac microtissues have been successfully used predominantly in regenerative medicine, but also as platforms for drug screening or as disease models [12]. Self-aggregating, scaffold-free cardiac microtissues consisting of human iPSC-derived cardiomyocytes and human primary cardiac fibroblasts cultured in nonadherent plates represent a high-throughput in vitro model of human cardiac tissue. These spontaneously beating cardiac microtissues show typical structures including well-developed myofibrils and recapitulate cardiac functionality including responsiveness to electrical stimulation [13]. In this work, we aimed to address how fibrotic changes affect contractility of human cardiac microtissues. We found that profibrotic condition and, in particular, activation of cardiac fibroblasts increase beating rate of cardiac microtissues, and this effect is at least partially mediated by endogenous activation of β-adrenergic receptors. Furthermore, we demonstrated that human cardiac microtissues represent an easy-to-use, reproducible high-throughput method for small molecule screening.

#### **2. Material and Methods**

#### *2.1. Cellular Sources*

Frozen human iPSC-derived cardiomyocytes (>99% troponin-positive) were purchased from Cellular Dynamics International, and were used directly upon thawing. Foetal human cardiac fibroblasts were purchased from Sigma (Cell Applications) and cells from passages 10–15 (>99% vimentin-positive, >99% collagen I-positive and <5% positive for filamentous form of α-SMA) were used. Adult cardiac fibroblasts were isolated from cardiac tissue samples obtained from the left atria of hearts from the patients with end-stage heart failure due to ischemic heart disease undergoing heart transplantation and from the waste fragments of donor normal hearts that were trimmed off while adjusting atrium size and shape during transplantation. Specimens were collected without patient identifiers according to the protocol revised by the Ethics Committee of the University of Naples Federico II (approval number 79/18) and in conformity with the principles outlined in the Declaration of Helsinki. Patients characteristic is available in Supplementary Table S1.

#### *2.2. Isolation of Human Adult Cardiac Fibroblasts*

Adult cardiac fibroblasts were isolated as described previously [14]. Briefly, cardiac tissue samples were minced and then enzymatically disaggregated by incubation in 0.25% trypsin and 0.1% (*w*/*v*) collagenase II (both Sigma-Aldrich, Basel, Switzerland) for 30 min at 37 ◦C. The digestion was stopped by adding Hanks' Balanced Salt solution supplemented with 10% foetal bovine serum (FBS, Gibco, Paislay, UK). The tissue was further disaggregated by pipetting; noncardiomyocyte cells were separated from debris and cardiomyocytes by sequential centrifugation and passage through 20 μm cell sieve. Fibroblasts were isolated from cell suspension by immunomagnetic cell sorting through positive selection with anti-fibroblast MicroBeads (MiltenyiBiotec, Bergisch Gladbach, Germany). Following selection, fibroblasts were cultured in Dulbecco's Modified Eagle's Medium (DMEM, Sigma, Basel, Switzerland) supplemented with 10% FBS, penicillin 10,000 U, and streptomycin 10 mg/mL (all from Sigma, Basel, Switzerland) at 37 ◦C in 5% CO2. The aCFs from patients with heart failure and unaffected donors were used individually. Cells from passages 4–6 were used.

#### *2.3. Fabrication and Culture of Microtissues*

Three types of microtissues containing: (1) iPSC-derived cardiomyocytes and fibroblasts mixed in ratio 4:1, (2) fibroblasts only, and (3) iPSC-derived cardiomyocytes only were generated by self-assembling in 96-well GravityTRAP plates (InSphero, Schlieren, Switzerland), and cultured as one microtissue per well. Accordingly, for generation of one microtissue, 5000 cells suspended in 70 μL maintenance medium were placed in the GravityTRAP plates for three days in the slanted position. Maintenance medium contained 2% FBS (Gibco, Paislay, UK), 50 μM phenylephrine hydrochloride (Sigma), 0.3 μM L-ascorbic acid (Sigma), 50 μM 2-mercaptoethanol (Gibco, Paislay, UK), 50 U/mL penicillin/streptomycin (Gibco, Paislay, UK) in high glucose DMEM (Sigma). Upon self-assembly, microtissues were further cultured in the GravityTRAP plates in the maintenance medium for 10 days. Microtissues were cultured under standard culture conditions (at 37 ◦C, 5% CO2 in a humidified incubator). Medium was changed at days 0, 2, 4, and 7. Fibrotic differentiation of cardiac microtissues was induced at day 0 with 10 ng/mL recombinant human TGF-β1 (Peprotech, London, UK) and TGF-βR1 was blocked with 10<sup>−</sup>8–10−<sup>5</sup> g/mL SD208 (Tocris, Zug, Switzerland). To address β-adrenoceptor-dependent mechanisms, microtissues were treated with 10 nM isoproterenol (Tocris) and/or 1 μM nadolol (Sigma) 15–90 min prior to video recording. Controls received solvents only.

#### *2.4. Live Imaging and Video Analysis*

Randomly chosen live microtissues were analysed using microscopes equipped with a humidified chamber at 37 ◦C and 5% CO2. Pictures of microtissues were captured using the Live Cell Imaging System (Olympus, Shinjuku, Japan) and size of microtissues was measured with the Xcellence Pro software (Olympus) as area in pixels. Increase in size was calculated as the size of microtissue at day 10/size of microtissue at day 0. Videos of contracting microtissues were recorded using the AxioObserver Z1 (Zeiss, Hombrechtikon, Switzerland) and the ZEN software. Videos were processed using Fiji software and custom-made macro. Contraction analysis was performed using Fiji software and MUSCLEMOTION macro [15].

#### *2.5. FluoVolt Measurements*

Voltage experiments were conducted on day 11. The microtissues were washed in serum-free medium consisting of DMEM (Gibco, Paislay, UK) supplemented with 10 mM D-Galactose (Sigma) and 1 mM sodium pyruvate (Sigma). The microtissues were loaded with voltage-sensitive dye: 0.1% FluoVolt and 1% PowerLoad (ThermoFisher, Waltham, WA, USA) in the above listed serum-free medium for 25 min at 37 ◦C 5% CO2. The voltage-sensitive dye was removed by washing in serum-free medium, and the multiwell plate was placed in an environmentally controlled stage incubator (37 ◦C, 5% CO2, >75% humidity) in the CellOPTIQ® platform (Clyde Biosciences, BioCity Scotland) 30 min before experimentation. The fluorescent signal was recorded with an excitation at 470 nm, and emitted light was collected from the entire microtissue using a 10× Fluor objective and the intensity recorded by a photomultiplier tube at 510–560 nm at 10 kHz. A 15 s recording was taken of each microtissue. Offline analysis was performed using CellOPTIQ®.

#### *2.6. Electrical Pacing*

Prior to experiment, microtissues were transferred into a glass bottom cell culture dish. Electrical field pacing of microtissues was performed using a modified glass bottom culture dish with inserted platinum wires attached to a MyoPacer (IonOptix, Westwood, MA, USA) set to a frequency of 3 Hz. Microtissues were recorded using the Eclipse TE2000-U microscope (Nikon, Tokyo, Japan) equipped with heating chamber, temperature controller (Ibidi, Graefelfing, Germany) and HeroBlack6 GoPro camera (Back-Bone Gear, Ontario, Canada).

#### *2.7. Quantitative RT-PCR*

For one sample, 12 microtissues were pooled. Total RNA was isolated using the Quick RNA Micro Kit (Zymo Research, Irvine, CA, USA). cDNAs were amplified using the cDNA Synthesis Kit (Roche). Gene expression was detected using oligonucleotides complementary to transcripts of the analysed genes and the GoTaq qPCR Master Mix (Promega, Dubendorf, Switzerland) using the 7500 Fast Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). Oligonucleotide sequences are available in the Supplementary Table S2. Transcript levels of human RPLP0 were used as endogenous reference, and relative gene expression was analysed using the 2−ΔΔCt method.

#### *2.8. Immunocytochemistry*

In total, 25–30 microtissues were pooled, fixed overnight with 4% paraformaldehyde and embedded in 2% agarose and in paraffin. Microtissue sections of 3–5 μm were cut (Sophistolab, Muttenz, Switzerland) and deparaffinised following standard methods. Sections were blocked with 10% goat serum (Vector Laboratories, Burlingame, CA, USA) and stained with primary/secondary antibodies: rabbit anti-human α-SMA (Sigma, dilution 1:750)/goat anti-mouse AP (Dako, Glostrup, Denmark, dilution 1:50), rabbit anti-human vimentin (Abcam, Cambridge, UK, dilution 1:250)/goat anti-rabbit (Vector Laboratories, Burlingame, CA, USA, dilution 1:200), rabbit anti-human periostin (Abcam, dilution 1:250)/Polymer anti-rabbit (Histofine, Nichirei Biosciences Inc. Tokyo, Japan), rabbit anti-human fibronectin (Abcam, dilution 1:100)/Polymer anti-rabbit (Histofine), rabbit anti-human Ki67 (Abcam, dilution 1:100)/goat anti-rabbit (Vector Laboratories, dilution 1:200), rabbit anti-human connexin 43 (Abcam, dilution 1:200)/goat anti-rabbit (Vector Laboratories, dilution 1:500), and rabbit anti-human troponin T (Origene Technologies, Rockville, Maryland, USA, dilution 1:2000)/Polymer anti-rabbit (Histofine). Detection was performed with the Bond Polymer HRP Refine Detection kit (Leica, Heerbrugg, Switzerland) according

to the manufacturer's guidelines for troponin T, and with DAB or Vector Red (Vector Laboratories) for all other antibodies. Nuclei were counterstained with haematoxylin (J.T.Backer, Gliwice, Poland). Direct Red Sirius Red (Sigma-Aldrich, Basel, Switzerland ) staining was used to detect collagen deposition. Stained sections were analysed using the Olympus DP80 microscope and the cellSens Standard imaging software (Olympus, Shinjuku, Japan). Intensity of immunopositive signals and nuclei counterstaining for each section were quantified using the ImageJ software. Staining intensity is presented in arbitrary units (AU) representing immunopositive signal intensity corrected by the counterstained nuclei value.

#### *2.9. Procollagen Type I and IL-6 ELISA*

Supernatants from individual microtissues were collected at day 10 and stored at −80 ◦C. Undiluted supernatants were analysed with the human Procollagen I alpha 1 DuoSet ELISA (R&D Systems, Abingdon, UK) and the human IL-6 DuoSet ELISA (R&D Systems) following manufacturer's protocol. Optical density was measured with the Synergy microplate reader (Biotek, Winooski, Vermont, USA) and the Gen5 software.

#### *2.10. Caspase 3*/*7 Activity*

Supernatants containing one microtissue each were collected at day 10 and immediately measured for apoptosis with the Caspase-Glo 3/7 Assay (Promega, Dubendorf, Switzerland) following the manufacturer's protocol. Luminescence was measured with the Synergy microplate reader (Biotek, Winooski, Vermont, USA) and the Gen5 software.

#### *2.11. Western Blotting*

Cellular proteins of foetal human cardiac fibroblasts were extracted with RIPA buffer (Sigma-Aldrich) supplemented with protease inhibitor cocktail (Complete ULTRA Tablets, Roche) and phosphatase inhibitors (PhosphoStop, Roche) from cultivated cells. Protein concentration was quantified by colorimetric bicinchoninic acid assay according to the manufacturer's protocol (Thermo Fisher Scientific). SDS-PAGE electrophoresis and wet-transfer method were used to separate and transfer proteins on nitrocellulose membranes, followed by 45 min incubation in blocking solution (tris buffered saline, Tween-20 (TBST, Thermo Fisher Scientific) containing 5% skim milk powder (Becton Dickinson AG, Allschwil, Swizterland). Membranes were incubated overnight with the following primary antibodies: anti-αSMA (1:1000, A2547, clone 1A4, Sigma-Aldrich), anti-Fibronectin (1:2000, Abcam: ab2413) or GAPDH (1:10,000, clone 14C10, #2118, Cell Signalling). Horseradish peroxidase (HRP)-conjugated secondary antibodies (1:5000) were used for detection with ECL substrate (SuperSignal West Pico Plus, Thermo Fisher Scientific) and development on the Fusion Fx (Vilber, Collegien, France). Densitometric analyses were performed with ImageJ 1.47t. Fold changes were computed after normalization to GAPDH.

#### *2.12. Cell Contraction Assay*

To assess the contractile properties of foetal human cardiac fibroblasts, the Contraction Assay Kit (Cell Biolabs, San Diego, CA, USA) was used following the manufacturer's protocol. Cells were cultivated with or without 10 ng/mL TGF-β for 72 h and reseeded in collagen gels for a further 72 h. Each condition was analysed in triplicates or quadruplicates. Images were taken at time 0, 24, 48 and 72 h after reseeding in a collagen gel. Areas of the gels were measured by ImageJ. Percentage of contraction of all conditions was measured compared to the average of unstimulated cells at day 0.

#### *2.13. Statistics*

Normally distributed data were analysed by two-tailed unpaired or paired Student's *t*-test and unpaired two-tailed ANOVA followed by uncorrected Fisher's LSD post-hoc test. All analyses were computed using GraphPad Prism 8 software. Differences were considered as statistically significant for *p* < 0.05.

#### **3. Results**

#### *3.1. Cardiac Fibroblasts Improve Integrity and Contractility of Human Cardiac Microtissues*

Using human iPSC-derived cardiomyocytes (iCMs) and human foetal cardiac fibroblasts (fCFs), we developed a high throughput in vitro model of human cardiac tissue [16]. Cardiac microtissues consisting of 5000 cells each were generated by self-assembly (Figure S1). iCMs assembling without fibroblasts, however, formed loose cell aggregates (Figure 1A, Video S1). Addition of fCFs to iCMs in a ratio of 1:4 (fCFs:iCMs) allowed for formation of compact, spontaneously contracting cardiac microtissues (Figure 1A,B, Video S2) In the next step, we performed transcriptional profiling of microtissues by qPCR for genes characteristic for cardiomyocytes and fibroblasts. Seven genes characteristic for cardiomyocytes were detected in microtissues containing iCMs (iCM and iCM:fCF microtissues), but not in those generated with fCFs only. In contrast, nine out of twelve genes associated with fibroblasts and fibrosis were consistently detected in all types of microtissues (Figure 1C). However, expression levels of most of fibroblastic genes were substantially higher in fCFs than in iCMs microtissues. As expected, the expression profile of iCMs:fCFs microtissues showed high levels of all cardiac and fibroblastic genes (Figure 1C).

**Figure 1.** Characteristics of cardiac microtissues. Panel (**A**) illustrates typical morphologies (bar = 50 μm) and panel (**B**) shows typical contraction patterns of cardiac microtissues generated using iPSC-derived cardiomyocytes (iCMs) only (top) or iCMs mixed with foetal cardiac fibroblasts (fCFs) in a ratio of 4:1 (iCMs:fCFs, bottom). Representative contractions are available in the Supplementary Materials (Videos S1 and S2). A heat map in panel (**C**) indicates expression of cardiomyocyte and fibroblast genes in microtissues containing iCMs only (left), fCFs only (center) or iCMs:fCFs (right). Each segment indicates the average (*n* = 3–5) expression of one experiment. Lower −Δ Ct values indicate higher relative expression. n.d.—not detected.

#### *3.2. TGF-*β*1 Induces Fibrotic Changes in Human Cardiac Microtissues*

TGF-β1 represents a potent inducer of profibrotic changes in cardiac fibroblasts. In the first step, we exposed microtissues to TGF-β1 for 10 days. In the presence of TGF-β1, both fCFs and iCMs:fCFs microtissues significantly increased their size (Figure 2A, Figure S2).

In the next step, we measured procollagen type I secretion, which is an important hallmark of the ongoing fibrotic processes. As expected, fCFs microtissues produced markedly more collagen I upon TGF-β1 stimulation. Under this stimulatory condition, significant increase in procollagen type I production was also observed for iCMs:fCFs microtissues (Figure 2B). Importantly, iCMs microtissues did not produce detectable levels of procollagen I (Figure 2B) pointing to fibroblasts as the main source of collagen I in cardiac microtissues. Increased collagen production in iCMs:fCFs microtissues treated with TGF-β1 was confirmed by picrosirius red staining (Figure 2C). To address whether treatment with TGF-β1 induced apoptosis in this model, we measured caspase 3/7 activity in iCMs:fCFs microtissues. We observed similar caspase 3/7 activity in iCMs:fCFs microtissues treated with and without TGF-β1 (Figure 2D). Additionally, iCMs:fCFs microtissues treated with TGF-β1 produced significantly less IL-6 (Figure 2E).

Cardiac fibrogenesis is a transcriptionally regulated process. Upon TGF-β1 stimulation, genes characteristic for cardiomyocytes were only slightly dysregulated in iCMs microtissues. In iCMs:fCFs microtissues, instead, cardiomyocyte-specific genes were downregulated following TGF-β1 treatment (Figure 2F). Analysis of fibrotic genes showed that most of them were strongly upregulated in fCFs and iCMs:fCFs microtissues in response to TGF-β1 (Figure 2F).

TGF-β1-induced changes in gene expression was followed by the analysis of the respective proteins in cardiac microtissues. To this aim, selected markers were analysed by immunohistochemistry in iCMs:fCFs microtissues cultured with or without TGF-β1. We found significantly less expression of the cardiac cell marker troponin T and gap junction protein connexin 43 in iCMs:fCFs microtissues exposed to TGF-β1. Instead, proteins produced by quiescent (vimentin, fibronectin) or activated (α-SMA, periostin) fibroblasts were more abundant in TGF-β1-treated cardiac microtissues. The number of proliferative (Ki67-positive) cells was also higher in iCMs:fCFs microtissues cultured with TGF-β1 (Figure 3, Figure S3). In conclusion, our results indicate that TGF-β1 effectively induced fibrogenesis in cardiac microtissues.

#### *3.3. Pharmacological Targeting of TGF-*β*R1 Signalling Prevents from Fibrotic Changes in Human Cardiac Microtissues*

In the next step, we asked whether the fibrotic cardiac microtissue model is relevant to test anti-fibrotic compounds. To this aim, we used the TGF-βR1 inhibitor SD208 and tested it in the iCMs:fCFs microtissues cultured in the presence or absence of TGF-β. We observed that SD208 inhibitor effectively blocked TGF-β1-mediated microtissue growth (Figure 4A).

Furthermore, we found that SD208 suppressed procollagen type I production in iCMs:fCFs microtissues cultured not only with, but also without TGF-β1 (Figure 4B). Given the effect of the inhibitor in the absence of exogenous TGF-β1, we analysed gene expression in iCMs:fCFs microtissues cultured with and without SD208. Expression of 12 out of 18 analysed genes were significantly changed in response to SD208 treatment (Figure 4C). Similar results were obtained for treatment of iCMs:fCFs microtissues with SD208 in the presence of TGF-β1 (Figure 4D). Next, we analysed dose-dependent responses of microtissues to SD208. In these experiments, iCMs:fCFs microtissues were cultured in the presence of TGF-β1 and serial dilutions of SD208. Our results showed dose-dependent responses of iCMs:fCFs microtissues to SD208 in terms of microtissue size (Figure 4E), collagen type I secretion (Figure 4F) and expression of profibrotic genes (Figure 4G). Taken together, these data suggest that cardiac microtissues represent a useful model for pharmacological studies.

**Figure 2.** TGF-β1 activates foetal cardiac fibroblasts in microtissues. Panel (**A**) demonstrates changes in size of microtissues generated with fCFs only (fCFs, left) and iCMs mixed with fCFs in ratio 4:1 (iCMs:fCFs, right) cultured in the presence (red) or absence (black) of TGF-β1 (10 ng/mL) for 10 days. Panel (**B**) shows relative levels of procollagen I (measured by ELISA), at day 10 in supernatants of all three microtissue types: fCFs (left), iCMs:fCFs (middle) and iCMs (right). Graphs show cumulative data of 2–5 independent experiments. Each dot represents data of one microtissue. Panel (**C**) illustrates representative picrosirius red staining in iCMs:fCFs microtissues at day 10 (bar = 10 μm). Panel (**D**) shows caspase 3/7 activity measured at day 10 in iCMs:fCFs microtissues. Graphs show cumulative data of 3 independent experiments. Panel (**E**) shows IL-6 levels measured by ELISA, at day 10 in supernatants of iCMs:fCFs microtissues. Graphs show cumulative data of 3 independent experiments. Each triangle represents data of one microtissue. Panel (**F**) summarizes fold changes in gene expression in indicated microtissues in the presence of TGF-β1 (in relation to expression in the absence of TGF-β1). \* *p* < 0.05. Graphs show cumulative data of 3–4 independent experiments, *n* = 11–20. For all graphs, *p* values were calculated with the Student's *t*-test. n.d.—not detected.

**Figure 3.** TGF-β1 induces fibrotic phenotype in cardiac microtissues. Immunohistochemistry of iCMs:fCFs microtissues cultured in the presence or absence of TGF-β1 (10 ng/mL) at day 10. Panel (**A**) illustrates representative staining for the indicated proteins at the indicated condition (bar = 50 μm). Higher magnification pictures are available in the Supplementary Materials (Figure S3). Panel (**B**) shows quantification of the respective staining for microtissues cultured in the presence (red) or absence (black) of TGF-β1. Graphs show cumulative data of 3 independent experiments. Each triangle represents data for one microtissue. *p* values were calculated with the Student's *t*-test.

**Figure 4.** Pharmacological targeting of microtissues with TGF-βR1 inhibitor SD208. (**A**–**D**) iCMs:fCFs microtissues were cultured in the presence or absence of TGF-β1 (10 ng/mL) and SD208 (10 μg/mL). Panel (**A**) shows changes in size of microtissues and panel (**B**) normalized levels of secreted procollagen I (measured by ELISA in supernatants) at the indicated conditions at day 10. Each dot represents data for one microtissue. *p* values were calculated with ANOVA followed by uncorrected Fisher's LSD tests for selected groups, \* *p* < 0.05. Panels (**C**,**D**) show changes in gene expression in response to treatment with SD208 of microtissues cultured in the absence (**C**) or presence (**D**) of TGF-β1. *n* = 6, *p* values were calculated with the Student's *t*-test, \* *p* < 0.05. (**E**–**G**) iCMs:fCFs microtissues were cultured in the presence of TGF-β1 (10 ng/mL) and serial dilutions of SD208. Panel (**E**) shows changes in microtissue size, panel (**F**) normalized levels of secreted procollagen I and panel (**G**) relative expression of selected genes at the indicated conditions at day 10. Each dot represents data for one microtissue. Each triangle represents data for one microtissue. *p* values were calculated with ANOVA followed by uncorrected Fisher's LSD tests versus +TGF-β group, \* *p* < 0.05.

#### *3.4. Activated Cardiac Fibroblasts Increase the Contraction Rate of Cardiac Microtissues*

Contractility represents the main functional feature of cardiomyocytes. Using high-speed movies and motion-tracking analysis, we examined contractile properties of iCMs:fCFs microtissues at day 10. Fibrotic changes triggered by TGF-β1 substantially deteriorated the contraction pattern of iCMs:fCFs microtissues. We consistently observed significantly increased contraction rate and reduced contraction amplitudes of fibrotic microtissues in all experiments. A more detailed analysis of contraction pointed mainly to the shortening of relaxation phase in iCMs:fCFs microtissues upon treatment with TGF-β1 (Figure 5A, Figures S4 and S5, Video S3).

**Figure 5.** Contractile properties of cardiac microtissues containing foetal or adult cardiac fibroblasts. Panel (A) shows quantification of contraction parameters of iCMs:fCFs microtissues cultured in the presence (red) or absence (black) of TGF-β1 (10 ng/mL) at day 10. Quantification of contraction parameters of microtissues containing fCFs pretreated with TGF-β1 for 3 days prior microtissue formation (blue) or untreated fCFs (black) recorded at day 10 are shown in panel (B). Each dot represents average data of one experiment. Data of individual experiments are available in the Supplementary Materials (Figures S5 and S8). *p* values were calculated with the paired Student's *t*-test. Panel (C) shows quantification of contraction parameters of cardiac microtissues containing aCFs from unaffected hearts (white) or heart failure (HF) patients (grey). Each dot represents average data of one experiment (*n* = 14–18). *p* values were calculated with the Student's *t*-test. Representative contraction records are available in the Supplementary Materials (Figures S5, S10B and S11 and Videos S2–S6).

We hypothesized that TGF-β1 affected the contractility of iCMs:fCFs microtissues by activation of fCFs, rather than by a direct effect on iCMs. To validate this hypothesis, we generated cardiac microtissues using fCFs pretreated with TGF-β1 for three days (prior microtissue formation) and analysed their contractility in the absence of exogenous TGF-β1 at day 10. Pretreatment with TGF-β1 effectively activated fCFs as indicated by increased production of collagen, α-SMA and fibronectin in 2D cultures (Figure S6). We observed that iCMs:fCFs microtissues containing activated fCFs displayed increased size and secreted significantly more procollagen I (Figure S7), showed increased contraction rate, reduced contraction amplitudes and shortened relaxation phase (Figure 5B, Figures S4 and S8, Video S4), and successfully reproduced the contraction pattern observed in fibrotic (+TGF-β1) cardiac microtissues.

To further address the relevance of CFs to cardiac microtissue contractility, we compared cardiac microtissues containing adult CFs (aCFs) obtained from unaffected hearts of mostly young individuals (healthy aCFs) to aCFs from patients with heart failure (HF aCFs). Cardiac microtissues containing healthy aCFs and HF aCFs showed similar sizes at day 10 (Figures S9 and S10A). We found that iCMs:aCFs microtissues with HF aCFs showed increased contraction rate and shortened relaxation phase in comparison to cardiac microtissues containing healthy aCFs (Figure 5C, Figures S4 and S10B, Videos S5 and S6). Of note, microtissues containing healthy iCMs:aCFs more often responded to TGF-β1 stimulation by increasing contraction rate than those with HF aCFs (four of six vs. one of six, Figure S11). All these data indicate that contractile properties of cardiac microtissues depend on the activation status of CFs.

#### *3.5. Una*ff*ected Repolarization Phase in Fibrotic Cardiac Microtissues*

In cardiomyocytes, mechanical contraction is triggered by electrical excitation. Next, we addressed the effect of external electrical stimulation of cardiac microtissues. Stimulation at 3 Hz induced a beating rate of 178–179 bpm in responsive microtissues, reduced contraction amplitude and substantially shortened contraction duration in both control and TGF-β1-treated iCMs:fCFs microtissues (Figure 6A, Videos S7 and S8), pointing to their full electrical responsiveness. Next, we analysed changes in membrane potential of control and fibrotic iCMs:fCFs microtissues using the relevant fluorescent probe (Figure 6B).

**Figure 6.** Cardiac microtissue electrophysiology. iCMs:fCFs microtissues were cultured in the presence (red) or absence (black) of TGF-β1 (10 ng/mL) for 10 days. Panel (**A**) shows quantifications of the contraction parameters recorded during spontaneous contractile activity (triangles, ctr) and then upon electrical stimulation with 3 Hz (circles, 3 Hz). Each dot represents data for one microtissue at the indicated condition, lines match data obtained from the same microtissue. *p* values were calculated with the paired Student's *t*-test. Panel (**B**) illustrates representative action potentials recorded with FluoVolt probe (left) and the respective contractions (right). Quantifications of action potential parameters are shown in panel (**C**). Graphs show cumulative data of 3 independent experiments. Each dot represents data for one microtissue. *p* values were calculated with the Student's *t*-test. APD—action potential duration, TRise—depolarisation phase.

We found significantly lower action potential amplitudes in TGF-β1-treated iCMs:fCFs microtissues in comparison to controls (Figure 6C), suggesting reduced number of electrically active cells in fibrotic microtissues. Instead, there was no difference in the action potential duration, although treatment with TGF-β1 slightly prolonged the depolarization phase (Figure 6C). It seems that electrical activity of iCMs in cardiac microtissues was not significantly affected by profibrotic changes.

#### *3.6. Endogenous* β*-Adrenergic Receptor Signalling Controls Increased Contraction Rate in Fibrotic Cardiac Microtissues*

Action potential triggers cardiomyocyte contraction, but β-adrenoceptor-coupled mechanisms can modulate it. First, we used iCMs:fCFs and iCMs:aCFs microtissues and treated them with β-adrenoceptor agonist isoproterenol. As expected, stimulated microtissues showed increased contraction rate (Figure 7A, Figure S12). Isoproterenol significantly shortened both the contraction and relaxation phases of spontaneously contractingmicrotissues. In contrast, contraction amplitude remained unaffected. Addition of theβ-adrenergic receptor blocker nadolol effectively prevented isoproterenol-induced changes in microtissue contraction (Figure 7A). These data confirmed functional β-adrenoceptors in our cardiac microtissues.

**Figure 7.** β-adrenergic receptor signalling in cardiac microtissues. Panel (**A**) shows quantification of contraction parameters of iCMs:fCFs microtissues stimulated with β-adrenoceptor agonist isoproterenol (iso, 10 nM) in the presence or absence of β-adrenoceptor blocker nadolol (1 μM). Unstimulated microtissues were used as controls. Graphs in panels (**B**,**C**) show contraction parameters of iCMs:fCFs (**B**) and iCMs:aCFs (**C**) microtissues treated with vehicle or 1 μM nadolol in the absence of isoproterenol. Each triangle represents data for one microtissue. *p* values were calculated with ANOVA for (**A**) and with the Student's *t*-test for (**B**,**C**).

Next, we addressed the relevance of endogenous β-adrenoceptor signalling in cardiac microtissues. To this aim, we analysed the contractility of iCMs:fCFs and iCMs:aCFs microtissues in the presence and absence of nadolol. Treatment with nadolol only slightly reduced contraction of iCMs:fCFs microtissues

(Figure 7B). In the next experiment, we used iCMs:aCFs microtissues showing high spontaneous contraction rate. In this model, nadolol significantly reduced contraction rate, whereas contraction amplitude remained unchanged (Figure 7C). We found that nadolol specifically affected relaxation, but not contraction phase. These results suggest activation of endogenous β-adrenoceptor signalling in iCMs:aCFs microtissues.

#### **4. Discussion**

In 3D microtissue models, fibroblasts have been recognized to improve microtissue architecture and biomechanics by providing proper ECM. We confirm that CFs significantly increased the integrity and synchronized contractility of human cardiac microtissues. As recently demonstrated, the use of a 4:1 ratio (iCMs:CFs) results in cardiac microtissues with stable phenotype and beating properties for up to one month [16]. Our data further demonstrated that contractility of these cardiac microtissues was substantially affected under profibrotic condition (i.e., in the presence of TGF-β1). TGF-β1 is a profibrotic cytokine that enhances fibroblast proliferation and production of ECM proteins. In our model, TGF-β1 induced fibrotic phenotype in cardiac microtissues, as indicated by elevated levels of fibrotic genes, secretion of procollagen I and fCFs proliferation. Previously, exogenous TGF-β1 has been shown to induce fibrosis in rat [17,18] or human [19] cardiac microtissue models. Furthermore, fibrotic phenotypes in cardiac microtissues or in biowire were also produced by collagen/fibroblast enrichment [20,21] or by constitutive activation of profibrotic pathways [22].

Our data demonstrated that pretreatment of fCFs with TGF-β1 prior to microtissue formation resulted in a similar effect on microtissue contractility as continuous stimulation with TGF-β1. This result suggested that activation of CFs, rather than a direct effect of TGF-β1 on iCMs, was a main mechanism responsible for deteriorated contraction of fibrotic cardiac microtissues. This idea was further strengthened by the observation that aCFs induced a contraction pattern (high contraction rate with low amplitude) of cardiac microtissues similar to those containing activated fCFs. CFs become activated following heart failure and with aging [23]. We observed that cardiac microtissues containing aCFs obtained from aged heart failure patients showed significantly higher beating rate than microtissues containing aCFs from young, healthy hearts. All these data suggested that activation status of CFs was a key factor determining contraction pattern of cardiac microtissues.

Fibrotic condition in the heart is associated with arrhythmias. Published data showed that activated rodent CFs could cause asynchronous contraction [18] and induce proarrhythmic changes in cardiomyocytes by affecting action potential profile [22,24]. In our model, fibrotic microtissues showed arrhythmic contractions only occasionally and their action potential profile remained unaffected. Instead, the fibrotic condition was associated with reduced action potential signal amplitude, suggesting a lower number of electrically active iCMs in these microtissues. Impaired conduction and contraction of these microtissues might be a consequence of iCMs decoupling and coupling of iCMs with CFs, reduced levels of gap junction protein connexin 43 and/or overproduction of cytokines and ECM by activated CFs. Secreted factors by activated CFs, biomechanical signalling through excessive ECM deposition and remodelling of mechanical junctions are known factors disturbing proper propagation of electrical impulses [25]. Signal conduction abnormalities, such as re-entry, can cause cardiac arrhythmias and may have serious clinical implications for patients with fibrotic hearts [9]. Clinical data suggest that ventricular fibrosis represents a strong predictor of life-threatening ventricular arrhythmia and sudden cardiac death in ischemic and nonischemic heart disorders [8].

In our model, profibrotic condition was consistently associated with increased beating rate of cardiac microtissues. Similar data were obtained in human cardiac microtissues containing mesenchymal stem cells instead of CFs following treatment with TGF-β1 [19]. Physiologically, increased beating rate of the heart is achieved by activation of β-adrenoreceptors. In normal adult cardiomyocytes, β1-adrenoreceptors are dominant, whereas iCMs contractility can be regulated by both β1- and β2-adrenoreceptors [26]. Previous data showed that the beating rate of human iCMs could be increased by treatment with the β-adrenergic agonist isoproterenol [26,27]. In line with these data, we could demonstrate the responsiveness of human cardiac microtissues containing iCMs to isoproterenol. In case of cardiac microtissues showing high beating rate (due to presence of activated aCFs), blockade of β-adrenergic receptors in the absence of exogenous agonists significantly decreased frequency. This finding indicated that activated CFs triggered endogenous β-adrenoreceptor signalling in cardiac microtissues. Mechanistically, stimulation of β-adrenoreceptors increases beating rate by enhancing cardiomyocyte relaxation [10]. Indeed, we observed that the reduced beating rate observed in the presence of β-adrenoreceptor blockers was associated with longer relaxation time. Accordingly, increased contraction rate of fibrotic cardiac microtissues correlated with shortened relaxation time. These results may suggest that activated CFs produce β-adrenergic agonists. Alternatively, fibrotic condition may sensitize β-adrenoreceptors. Cellular response to β-adrenergic stimulations is further modulated by multiple factors, such as cell surface receptor density, activity of the downstream signalling pathway, or by specific costimulations [28]. For example, fibrotic hearts of TGF-β-overexpressing mice showed increased density of β-adrenoreceptors on cardiomyocytes and altered β-adrenergic signalling [3]. Furthermore, collagen receptor β1-integrin [29], connective tissue growth factor [30] or tissue stiffness [31] have also been shown to regulate β-adrenergic signalling.

A successful preclinical platform has to combine physiological-like relevancewith the high-throughput processing. Our fibrotic cardiac microtissue model is established in a 96-well format and certain parameters, such as size, contractile properties, changes in membrane potential, procollagen-I production or apoptosis can be measured for individual microtissues. In particular, recently developed software measuring the contractile activity of the whole microtissue allows for quick and easy analysis of the key properties of beating cardiac microtissues [15]. In our experiments, transcriptomic data were obtained from 10–12 pooled microtissues, but we used a simple RT-PCR to analyse gene expression. Application of the next-generation sequencing technology would allow obtaining transcriptomic data from one microtissue. Thus, this model could be used as a high-throughput platform for pharmacological screening with a timeframe for up to at least 10 days. The proof-of-concept experiments with the TGF-βR1 inhibitor SD208 confirmed that fibrotic cardiac microtissues represent a robust model for drug screening.

Although our human cardiac microtissue-based fibrotic platform combines technical ease, high throughput and reproducibility with physiological-like relevance, it also needs to be discussed in the context of its limitations. iCMs are characterized by their foetal developmental status [32]. Currently, there is no satisfactory alternative to iCMs. The usefulness of adult primary human cardiomyocytes is limited, as these cells contract upon chemical or electrical stimulation only [33]. Despite recent advances in the maturation of iCMs through improved biomechanics and cocultures with other cell types, so far, their maturation status does not reflect adult cardiomyocytes [13,34–36]. Another limitation of our model is the heterogeneity of iCMs. Conventional differentiation of iPSCs into iCMs results in populations of pacemaker, ventricular-like, atrial-like cells as well as nonbeating progenitors. The use of defined types of differentiated cardiomyocytes could further improve the relevance of the cardiac microtissue model. Moreover, addition of other cell types, such as endothelial cells or macrophages, to iCMs and fCFs might result in formation of microtissues that better mimic the structure and function of human cardiac tissue.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/9/5/1270/s1, Figure S1: Schematic presentation of experimental setup with time lines, Figure S2: Kinetics of microtissue growths, Figure S3: TGF-β1 induces fibrotic phenotypes in cardiac microtissues, Figure S4: Contraction patterns of cardiac microtissues containing foetal or adult cardiac fibroblasts, Figure S5: Effect of exogenous TGF-β1 on contractility of cardiac microtissues - individual experiments, Figure S6: Activation of foetal cardiac fibroblasts with TGF-β1, Figure S7: Effect of activated foetal cardiac fibroblasts on size and procollagen I secretion, Figure S8: Effect of activated foetal cardiac fibroblasts on contractility of cardiac microtissues - individual experiments, Figure S9: Size of cardiac microtissues containing adult cardiac fibroblasts, Figure S10: Size and contractile properties of cardiac microtissues containing adult cardiac fibroblasts, Figure S11: Effect of exogenous TGF-β1 on contractility of healthy iCMs:aCFs and HF iCMs:aCFs microtissues, Figure S12: b-adrenergic receptor stimulation of iCMs:aCFs microtissues; Table S1: Clinical characteristics of patients from which aCFs were obtained, Table S2: Primers used for quantitative RT-PCR; Video S1: Control iCMs only, Video S2: iCMs\_fCFs control d10, Video S3: iCMSfCFs TGF-β1, Video S4: iCMsfCFs pre-TGF-β-3d\_d10 D10, Video S5: iCMs\_healthy aCFs control d10, Video S6: iCMS\_HF aCMs control d10, Video S7: iCMs\_fCFs control, Video S8: iCMs\_fCFs 3Hz.

**Author Contributions:** Conceptualization: P.B., G.K.; Methodology: P.B., C.Z., A.C., D.N., F.D.M., I.A., M.S., G.K.; Data curation: P.B., C.Z., A.C., M.S., G.K.; Formal analysis: P.B., G.L.S., C.Z., G.K.; Funding acquisition: G.K.; Supervision: O.D., G.K.; Writing—original draft: P.B., G.K.; Writing—review & editing: C.Z., G.L.S., O.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was founded by the Swiss Heart Foundation and Vontobel Foundation.

**Acknowledgments:** The authors acknowledge financial support of the Swiss Heart Foundation and Vontobel Foundation as well as technical assistance of the Center for Microscopy and Image Analysis of the University of Zurich and Sophistolab AG.

**Disclosures:** Irina Agarkova is employed by InSphero company, which produces microtissues. Godfrey Smith is a founder, shareholder, executive and honorary Chief Scientific Officer of Clyde Biosciences Ltd. company, which provides an assay service to the pharmaceutical industry using commercial iPSC-derived cardiomyocytes.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Human Induced Pluripotent Stem Cell-Derived 3D-Neurospheres Are Suitable for Neurotoxicity Screening**

**Julianna Kobolak 1, Annamaria Teglasi 1, Tamas Bellak 1, Zofia Janstova 1, Kinga Molnar 2, Melinda Zana 1, Istvan Bock 1, Lajos Laszlo <sup>2</sup> and Andras Dinnyes 1,3,4,\***


Received: 26 March 2020; Accepted: 29 April 2020; Published: 1 May 2020

**Abstract:** We present a hiPSC-based 3D in vitro system suitable to test neurotoxicity (NT). Human iPSCs-derived 3D neurospheres grown in 96-well plate format were characterized timewise for 6-weeks. Changes in complexity and homogeneity were followed by immunocytochemistry and transmission electron microscopy. Transcriptional activity of major developmental, structural, and cell-type-specific markers was investigated at weekly intervals to present the differentiation of neurons, astrocytes, and oligodendrocytes. Neurospheres were exposed to different well-known toxicants with or without neurotoxic effect (e.g., paraquat, acrylamide, or ibuprofen) and examined at various stages of the differentiation with an ATP-based cell viability assay optimized for 3D-tissues. Concentration responses were investigated after acute (72 h) exposure. Moreover, the compound-specific effect of rotenone was investigated by a panel of ER-stress assay, TUNEL assay, immunocytochemistry, electron microscopy, and in 3D-spheroid based neurite outgrowth assay. The acute exposure to different classes of toxicants revealed distinct susceptibility profiles in a differentiation stage-dependent manner, indicating that hiPSC-based 3D in vitro neurosphere models could be used effectively to evaluate NT, and can be developed further to detect developmental neurotoxicity (DNT) and thus replace or complement the use of animal models in various basic research and pharmaceutical applications.

**Keywords:** induced pluripotent stem cells; neurospheres; 3D culture; neurite outgrowth; neurotoxicity

#### **1. Introduction**

Environmental stressors, such as chemicals or the medical drugs could have a toxic effect on humans which may occur at any stage of their life, during fetal development, childhood, or adult life. The toxic effects of environmental agents coupled with inherited susceptibility of individuals make the toxicology prediction difficult. Conventional animal-based toxicity and safety tests have high costs, use large numbers of animals (mainly rats) and in many cases, they do not provide clearly translatable results for humans [1–3]. Consequently, in line with legislation, there is an increasing need to develop alternative testing methods which could handle thousands of drugs or chemicals with affordable time and cost and with human-relevant neurotoxicology (NT) outcome [4–7]. Development of new approach methods (NAM), would be important both for NT and developmental neurotoxicology (DNT) tests, providing data on the effect of chemicals and the potential adverse outcomes (AOs) [6,8–13].

An increasing number of studies use primary cell cultures and recently, pluripotent stem cells (PSCs) to create in vitro systems for NT and DNT screenings. It is widely accepted concept that 3D cell cultures can mimic better the original tissue environment including tissue-specific architecture, mechanical and biochemical features, cell-to-cell communication and signaling, and differentiation capability, while 2D cell culture systems are less complex and more artificial in this sense [14–17], although the existing assays provide meaningful readouts for specific neurodevelopmental processes [6,8,11]. Only one in vitro assay will not be able to cover the complexity of the in vivo development, therefore a battery of assays and fit-for-purpose applications should be used to cover the relevant processes [10,18]. Human induced pluripotent stem cell-derived (hiPSC) neurospheres, grown on 3D scaffolds or self-forming, can establish cell–cell interactions and model certain neurodevelopmental processes, therefore, might be used as an in vitro screening platform not only for NT but in DNT studies or drug development [19–21].

Neural stem cells (NSCs) are considered as a multipotent and self-renewing pool of cells in the mammalian central nervous system (CNS), occurring in vivo in the developing embryonic neural tissue as early as the neural tube formation happens. These cells have the capacity to differentiate into all neuronal cell types, as well as into glial cells, therefore they are able to emulate some fetal neurodevelopmental processes [22,23]. Moreover, NSCs can be differentiated in vitro from pluripotent stem cells (PSCs) thus providing an attractive and almost unlimited in vitro tool for toxicology studies including drug development [24,25].

Despite the developments mentioned above, still, a limited number of human iPSC derived 3D neuronal culture-based studies were published focusing on the development of NT or DNT models [21,26–31]. A limiting factor for the further development of such test methods is the lack of high-throughput screening (HTS) read-outs on 3D cell cultures [16], despite the developments using high-content image analysis (HCA) [27].

Here, we present a reliable model system where hiPSC-derived NSCs are differentiated towards subtypes of neurons, astrocytes, and oligodendrocytes, forming free-floating 3D neurospheres in 96-well plate format. Over 6 weeks the various differentiation stages are characterized, and three selected stages are exposed with different compounds to investigate their cytotoxic effect. The generated neuronal spheroids used for NT measurements at different time-points resembling various differentiation stages, therefore, provide an excellent platform for further DNT test system developments.

#### **2. Methods**

#### *2.1. Chemicals and Plasticware*

The chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA) and all cell culture reagents and plasticware from Thermo Fisher Scientific Inc. (Waltham, MA, USA), unless otherwise specified.

#### *2.2. Human iPSC Culture*

Human iPSC line (Ctrl-2), derived from a healthy mid-age Caucasian female donor peripheral blood mononuclear cells (PBMCs), established and characterized earlier [32,33], was used in this study. Cells were cultured on BD Matrigel™ matrix (BD Biosciences, Franklin Lakes, NJ, USA) with mTeSR™1 medium (Stem Cell Technologies, Vancouver, Canada), using Gentle Cell Dissociation Reagent for passages, according to the manufacturer's instruction. Representative hiPSC colony morphology, pluripotency staining for POU5F1, NANOG, SSEA4, and TRA1-60 and karyotype analysis are presented in Figure S1. For mycoplasma screening, the Venor®GeM-Advance (Minerva Biolabs) Mycoplasma Detection Kit was used according to the manufacturer's protocol in every fifth passage during maintenance and before freezing. Cells were cultured at 37 ◦C in a humidified atmosphere containing 5% CO2. In the current study cultures from passage 16 and 17 were used for differentiation.

#### *2.3. Neuronal Di*ff*erentiation and Maintenance*

The Ctrl-2 hiPSC line was differentiated to neural stem cells (NSCs) by dual SMAD inhibition procedure [34], following the detailed protocol of Shi et al. [35]. Briefly, when hiPSC cultures reached 90% confluence, the culture was passaged onto poly-l-ornithine and laminin (POL/L; 0.002%/1 μg/cm2) coated plates, using conventional iPSC passage (as detailed above). On the next day media was changed to neural induction medium (NIM; DMEM/F12: Neurobasal medium, supplemented with 1× N2, 2× B27, 2 mM glutamine, 1× non-essential amino acid (NEAA), 100 μM β-mercaptoethanol, 5 μg/mL insulin) supplemented with 10 μM SB431542, 500 ng/mL Noggin (R&D Systems, Inc., Minneapolis, MN, USA) and 5 ng/mL basic fibroblast growth factor (bFGF) to induce the neuroectodermal lineage. Neuronal induction media was applied until day 10 of the differentiation. The forming neural rosette-like structures were manually picked under a stereomicroscope (Olympus SZX2; Olympus Ltd. Tokyo, Japan) and re-plated onto POL/L (POL/L; 0.002%/1 μg/cm2) plates. NSCs were expanded in neural maintenance medium (NMM; DMEM-F12: Neurobasal medium, supplemented with 1× N2, and 2× B27, 2 mM glutamine, 1× NEAA), supplemented with 10 ng/mL bFGF and 10 ng/mL Epidermal growth factor (EGF) and maintained in a monolayer on dishes coated with POL/L (0.002%/1 μg/cm2). When reached 100% confluence, cells were passaged using Accutase (Sigma) and seeded as single cells (50,000 cells/cm2) for further expansion on POL/L (0.002%/1 μg/cm2) coated dishes. After 4 passages NSCs were frozen in 1 mL freezing medium (90% Fetal Bovine Serum, FBS; heat-inactivated, Thermo Fisher Sci.; Cat N.: 10500-064, LOT: 08F1180K; 10% DMSO; 2 million cells/vial), using Accutase passage. Before freezing the NSCs, a mycoplasma test was performed (see above), repeated every fifth passage during maintenance.

#### *2.4. 3D Neurosphere Culture*

After thawing, NSCs (generated as detailed above) were cultured on POL/L (0.002%/1 μg/cm2) coated plates (50,000 cells/cm2). When reached 100% confluence NSCs were propagated with Accutase treatment and single-cell suspension was plated onto a low-adherent 96-well plate with 10,000 cells/well in NMM medium. Neurospheres were formed within 48 h after plating and NMM half of the media was changed in every 3 days until analyzing the samples. Samples were collected at Day 0 (D0), Day 2 (D2), and weekly intervals from the end of the 1st week (D7) until the end of the 6th week (D14, D21, D28, D35, and D42).

#### *2.5. Cryosectioning and Immunocytochemistry (ICC) Staining*

3D neurospheres (untreated, vehicle or compound-treated) were fixed with 4% paraformaldehyde (PFA) in 0.1 mol/L phosphate buffer for 1 h at RT and washed 3 times with PBS. The fixed samples were cryoprotected in 30% sucrose in PBS containing 0.01% sodium azide at 4 ◦C until embedding in Shandon Cryomatrix gel (Thermo Fischer Scientific). The 16 μm parallel sections were made using cryostat (Leica CM 1850 Cryostat, Leica GmbH), mounted to Superfrost™ Ultra Plus Adhesion Slides (Thermo Fisher Scientific) and stored at −20 ◦C until use. After 10 min air-drying, the sections were blocked for 1 h at RT with blocking solution (3% BSA in PBS), supplemented with 0.2% TritonX-100. The sections were then incubated with primary antibodies (Table S1) overnight at 4 ◦C. Next day, sections were washed in PBS 3 times, and isotype-specific secondary antibodies (Table S1) were diluted in blocking buffer and applied for 1 h at RT. The sections were washed 3 times with PBS and covered using Vectashield® mounting medium containing DAPI (1.5 μg/mL; Vector Laboratories), that labelled the nuclei of the cells (at least 1 h at RT). Negative controls for the secondary antibodies were performed by omitting the primary antibodies. Immunoreactive sections were analyzed using a BX-41 epifluorescent microscope (objectives: 20× 0.50 NA; 40× 0.75 NA; Olympus) equipped with a DP-74 digital camera and its CellSens software (V1.18; Olympus). For confocal imaging, Olympus Fv10i-W compact confocal microscope system (objective: 60× 1.35 NA; Olympus) with Fv10i software

(V2.1; Olympus) was applied. All images were further processed using the GNU Image Manipulation Program (GIMP 2.10.0) and NIH ImageJ analysis software (imagej.nih.gov/ij).

Quantification of the immunocytochemistry data was performed using ImageJ software according to Tieng et al. [36]. Briefly, images were taken by confocal microscopy at 120× magnification. The numbers of Ki-67-, NESTIN-, GFAP-, AQP4-, TUBB3-, NF200kD-, MBP-, VAMP2-, and MAP2-immunoreactive pixels were measured in 5 neurosphere (highest diameter middle sections, 5 randomly selected fields/slide) at every time points. Data was normalized with DAPI positive nuclei number. Data were expressed as a percentage of marker/DAPI ratio ± SEM (*p* < 0.05).

#### *2.6. Apoptosis Assay*

Embedding and cryosectioning of 3D samples were performed as above. To detect apoptotic activity, the DeadEnd™ Colorimetric TUNEL System (Promega) was used on the middle cryosections (highest diameter) of the spheroids, following the instructions of the manufacturer. In brief, apoptosis was detected by immersing the slides in PBS for 5 min (at RT), adding 20 μg/mL Proteinase K solution and incubating for 10–30 min (at RT). After 5–10 min treatment in Equilibration buffer, recombinant terminal deoxynucleotidyl transferase (rTdT) was added to the reaction mixture. Next, the sections were incubated for 60 min at 37 ◦C inside of a humidified chamber to allow the end-labelling reaction to occur. The reaction was terminated by immersing the slides in saline-sodium citrate for 15 min (RT). Endogenous peroxidases were blocked by immersing the slides in 0.3% hydrogen peroxide in PBS for 3–5 min (RT). Streptavidin-HRP was added to slides, incubated for 30 min (RT), stained with diaminobenzidine (DAB) solution for 5 min until a light brown background appeared. For hematoxylin–eosin (HE) staining Mayer's Hematoxylin solution was used for 3 min. Sections were rinsed with tap water and placed into distilled water for 30 s, then into 96% alcohol for 30 s. One percent Eosin solution in distilled water was used for 3 min. Stained sections were dehydrated through alcohols, clear in xylene and mount in DPX. Microphotographs were made with DP-74 digital camera (Olympus) using a light microscope (BX-41, objectives: 20× 0.50 NA; 40× 0.75 NA; Olympus) and CellSens software (V1.18; Olympus). For counting the apoptotic and total (Hematoxylin-stained) number of cells, NIH ImageJ analysis software was used. Five Ctrl and five ROT-treated spheroids were randomly selected, and middle sections were analyzed from each differentiation stage (D21, D28, and D42) samples in three experiments (*n* = 3).

#### *2.7. Transmission Electron Microscopy (TEM)*

Neurospheres (untreated, vehicle or compound-treated) were fixed at different differentiation stages in a fixative solution containing 3.2% PFA, 0.2% glutaraldehyde, 1% sucrose, 40 mM CaCl2 in 0.1 M cacodylate buffer (pH 7.4) for 12 h at 4 ◦C. Samples for ultrastructural analysis were embedded in 1.5% agar (dissolved in dH2O), post-fixed in 1% ferrocyanide-reduced osmium tetroxide [37], then dehydrated using graded series of ethanol, finally embedded in Spurr low viscosity epoxy resin medium. Ultrathin sections were collected from the middle region of the spheroids (highest diameter) on copper slot grids coated with formwar (Agar Sci., Essex, UK) and counterstained with uranyl acetate and Reynolds's lead citrate. Sections were examined with a JEOL JEM 1011 transmission electron microscope (JEOL Ltd., Tokyo, Japan) equipped with a Morada 11-megapixel camera using iTEM software (Olympus).

#### *2.8. RT-qPCR Analysis*

For each sample, 12 spheroids were pooled, and 3 biological replicates were performed (*n* = 3). Total RNA was isolated using the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). For the reverse transcription, 600 ng of the isolated RNA was used applying the Maxima First Strand cDNA Synthesis Kit for RT-qPCR with dsDNase (Thermo Fisher Scientific) according to the manufacturer's instructions.

Gene-specific primers were designed using the Primer3 software [38], specified with mFOLD software [39] and Primer-BLAST software [40]. Primers were optimized using two-fold serial dilution standard curves (Table S2). As a reference, gene GAPDH was used (Table S2). Each real-time PCR reaction contained 5 ng RNA-equivalent cDNA template, 400 nM of each primer and 50% SYBR Green JumpStart Taq ReadyMix (Sigma Aldrich) in a total volume of 15 μL. PCR reactions were set up using QIAgility liquid handling robot and performed on a Rotor-Gene Q cycler (Qiagen). The cycling parameters were as follows: 94 ◦C for 3 min initial denaturation followed by 40 cycles of 95 ◦C for 5 s, 60 ◦C for 15 s, and 72 ◦C for 30 s. Melting curve analysis and agarose gel electrophoresis confirmed the specificity of the primers and the absence of gDNA contamination. Data of three replicates were analyzed for each gene, using the ddCT method [41].

#### *2.9. XBP1-Assay of Endoplasmic Reticulum Stress*

Upon accumulation of unfolded proteins in the endoplasmic reticulum (ER), a 26-nucleotide fragment from the X-box binding protein 1 mRNA (XBP1(U)) is removed with a special splicing mechanism [42]. This shorter mRNA (XBP1(S)) is a frequently used marker of ER-stress. To study the expression of XBP1(S), previously described primers [43] and our own primers were used [44] (Table S2). As a positive control, cells were treated with 5 μM and 10 μM Tunicamycin to induce ER-stress. RT-PCR reactions were performed and analyzed as above, using Phusion Hot Start II High-Fidelity DNA Polymerase and 20 ng of the cDNA samples.

#### *2.10. Toxicity Treatments and ATP Viability Assay*

Eleven well-known compounds were tested in 7 different concentrations to generate concentration-response curves. The tested compounds with the used concentrations are detailed in Table S3. The highest dosage was determined based on the solubility of the compounds, and care was taken not to reach above 0.1% DMSO levels after diluting the compound in the culture media. In each assay plate, 4 technical replicates were applied from each sample and 3 biological replicate assays were run (*n* = 3). The 3D cell cultures were exposed to the toxicants for 72 h (exposure schemes are detailed in Figure 5A or Figure 6A or Figure 7A). Vehicle control was used as a regular medium supplemented with 0.1% DMSO.

ATP viability assay was performed with CellTiter-Glo® 3D Cell Viability Assay, according to the manufacturer's protocol (Promega). Neurospheres were lysed with 100 μL CellTiter-Glo® 3D Reagent for 60 min at RT. Luminescence signal was recorded with a Thermo VarioScan Flash (Thermo Fisher Scientific) plate reader.

#### *2.11. Diametric and Total Protein Determination of the Spheroids*

3D spheroids grown in a 96-well plate were captured using the 4x objective (0.1 NA) of Olympus IX71 microscope and DP21 camera (Olympus). The images were analyzed by measuring the diameters of the spheroids using the Olympus CellSens Dimension software (V1.11). Each value represents the average of 3 experiments (*n* = 3) in each 96 spheroids were measured. The 3D spheroids were lysed individually with RIPA Lysis and Extraction Buffer supplemented with Halt™ Protease and Phosphatase Inhibitor Cocktail and Pierce™ Universal Nuclease for Cell Lysis, sonicated and the total protein concentration determined using a Pierce BCA Protein Assay Kit according to the manufacturer's instructions. Due to the small spheroid size of the D2, D7, and D14 samples, three spheroids were pooled, then the individual values were calculated accordingly. In total, three experiments were performed (*n* = 3), in which 24 spheroids were analyzed for each timepoint.

#### *2.12. Neurite Outgrowth Assay*

For neurite outgrowth measurement the method of Harris et al. was modified [45]. 3D spheroids were treated with different concentrations (vehicle, 0.1 μM, 0.5 μM, and 0.75 μM) of Rotenone (ROT) at D21 differentiation stage using the 72 h exposure scheme, similarly to the previous concentration-response experiments. After the treatment, the spheroids were plated on Matrigel-coated plates in NMM medium without compound, and 24 h later fixed with 4% PFA and immunostained against TUBB3, as described in detail under the immunostaining paragraph. First, the number of neurites was counted for each sample. Then, the "edge" of the spheroids was determined by drawing an "edge line" and every neurite length was measured from this "edge line" to the tip of each neurite. Total neurite length was calculated for each spheroid using the ImageJ software's Neurite Tracer plugin. In each experiment eight spheroids were treated in each experimental group (5 groups) and three independent experiments were performed (*n* = 3).

#### *2.13. Statistical Analysis*

All results were analyzed using Prism 5 (GraphPad Software, La Jolla, CA, USA) and handled in Microsoft Office 2010 (Microsoft, Redmond, WA, USA) software. For normalization of the concentration response, the average values of positive control were used as 0% and the average of vehicle control served as 100%. Four parameter curve fitting methods were used to determine EC10 and EC50 values, where the data on the graphs represent the average of three biological replicates. The "n" value corresponds to the number of biological replicates for each tested concentration. Analysis of data is presented as the mean ± SEM. Significance of data was determined with paired T test for RT-qPCR (\* *p* < 0.01) and with One-way ANOVA for concentration response (\* *p* < 0.05; \*\* *p* < 0.01).

#### **3. Results**

#### *3.1. Three-Dimensional Spheroid Di*ff*erentiation of iPSCs-Derived NSCs Revealed Complex Neuronal Cultures*

Our first aim was to characterize the timewise differentiation of 3D neuronal spheroids free-floating in suspension culture, originating from hiPSCs. The neuronal differentiation capacity of the starting neural stem cell (NSCs) population, derived by the dual-SMAD inhibition protocol, was comprehensively characterized by us previously [32,33]. These cells expressed the major NPC markers (NESTIN, SOX1 (SRY-Box 1), PAX6 (Paired Box 6) investigated on protein level with ICC (see details in Figure S2). When the NPCs were terminally differentiated in a 2D culture system over 5 weeks, cortical neurons and glial cells were differentiated and formed a neuronal network, as we published recently [32]. In the present study differentiation in a 3D culture system was investigated in the course of 6 weeks (42 days).

In suspension culture upon withdrawing the two mitogens EGF and bFGF, NSCs formed compact 3D spheroids, the so-called neurospheres within 48 h (Figure 1A). Growth properties of the structures were investigated by measuring the diameter of the individual spheres and their total protein content. Results showed a continuous growth of the spheres during the 6 weeks culture period, in terms of diametric growth and protein content (Figure 1). While the diameter increased linearly, the protein content increased more dynamically after week 4, resembling the cellular and structural changes during differentiation (Figure 1B,C).

**Figure 1.** Growth properties of 3D neurospheres during the 6 weeks of differentiation (**A**) Representative microscopic view of 3D spheroids, investigated at weekly intervals (4×, scale bar: 200 μm). (**B**) The average diameter of 3D spheroids in μm measured by CellSens Dimension software (Olympus) (*n* = 3, in each experiment 96 spheroids were compared weekly trough 6-weeks). (**C**) Average total protein content of spheroids determined by Pierce BCA Protein Assay. Note that D2, D7 and D14 samples were measured by pooling three spheroids and individual values were calculated, while in all other timepoints spheroids were measured individually (*n* = 3, in each stage 24 spheroids were measured). ±SEM values are presented on graphs. (\* *p* < 0.05).

This observation was in accordance with the expression of the cell proliferation marker Ki-67 (KI67), representing the proportion of dividing cells, monitored by ICC on protein level (Figure 2A,B) or by RT-qPCR on transcript level (Figure 3). The expression of Ki-67 and the number of Ki-67-positive cells were the highest in D2 and D7 samples, which showed a continuous decline over the differentiation period (Figure 2A,B or Figure 3). The forming 3D cell aggregates mainly expressed the early neuronal markers such as *PAX6*, *NESTIN*, and *SOX1*, which expressions decreased with maturation on a time-wise manner (RT-qPCR, Figure 3). Tubulin beta-3 chain (TUBB3) expression appeared very early in the samples, already at D2 as neuronal processes started to grow inside the spheres (ICC, Figure 2A,B). These changes were obvious when investigated at transcript level as well (RT-qPCR, *TUBB3*; Figure 3). It is important to note that no necrotic regions were detected in the center of the spheroids or in other regions, not even the highest diameter of spheroids reached 800 μm in average in D42 samples (Figure 1B or Figure 2 first line). This fact was corroborated in semithin sections where cell density proved to be high and consistent during the whole examined period. Layer organized fibers had become distinguishable on the surface from the 21st day (Figure S3a–g).

**Figure 2.** Immunocytochemical analysis of 3D spheroids. (**A**) Spheroids were fixed and cryosectioned then immunostained at weekly intervals from D2 until D42 stage. The first line represents the overview of the cryosectioned spheroids, while the rest of the panel shows higher magnifications. Relevant markers of proliferation (KI67), neural stem cells (NESTIN), neuronal differentiation (TUBB3 and MAP2), an intermediate filament of dendrites and axons (NF200), synaptic vesicles of neurons (VAMP2), astrocyte (AQP4), and oligodendrocyte (MBP) specific proteins were stained. Protein name IDs are indicated with colors, representing the color of the fluorophore used (e.g., green as Alexa 488; red as Alexa 594) Nuclei were counterstained with DAPI (in blue). Scale bar: 100 μm (first line only) and 25 μm.

(**B**) Quantitative analysis of the immunostainings on confocal images. The numbers of Ki-67, NESTIN, AQP4, TUBB3, NF200kD, MBP, VAMP2, and MAP2 immunoreactive pixels were measured in 5 neurospheres (middle sections, 5 randomly selected fields/slide) at every time points. Data was normalized with DAPI positive nuclei number. Data were expressed as percentage of marker/DAPI ratio ± SEM (\* *p* < 0.05).

**Figure 3.** Real-time PCR measurements of relevant markers during the neuronal differentiation of the spheroids. Twelve 3D spheroids were pooled, lysed with RLT-buffer and used in RT-PCR analysis at each timepoints. Where the genes are already expressed in D2 samples, it is referred as 1. Day 2, 7, 14 and 21 values for *GRIN1, CHAT, TH, GAD1,* and *SLC6A4* are not indicated as these genes were not expressed at those time points. Graphs represent normalized relative expression values, analyzed for each gene using the ddCT method [41]. Mean values and ± SEM of three biological replicates (*n* = 3) are presented on graphs, significance are determined by paired T test (∗ *p* < 0.01). Note that y axis's scales alter according to the different relative expression values.

A weekly investigation of the gene expression revealed a continuous progress in neuronal differentiation and increase in the level of maturation markers, such as the dendritic marker Microtubule-associated protein 2 (*MAP2*) or Microtubule-associated protein tau (*MAPT*), and RNA Binding Fox-1 Homolog 3 (*RBFOX3*, also known as NeuN) (Figure 3), which were in accordance with the protein data of immunofluorescent staining of the 3D spheroids, MAP2 and Neurofilament 200 kDa (NF200); (ICC, Figure 2A,B).

Terminal differentiation resulted in the formation of neuronal networks and synapses, from the stage of D28, post-synaptic marker PSD95 (now DLG4, Discs Large MAGUK Scaffold Protein 4) (Figures 3 and 4), and the major synaptic vesicle protein p38, Synaptophysin (SYNP) and Vesicle-associated membrane protein 2 (VAMP2) (Figures 2 and 4) expression appeared in maturing neurons. Synapse formation was assessed by transmission electron microscopy (TEM) revealing the presence of synaptic connections, presynaptic vesicles and postsynaptic density (Psd). Matured synapses were detected in D42 samples, where docked presynaptic vesicles were determined by identifiable contact point between the vesicle and the presynaptic membrane (Figure S3h,i).

**Figure 4.** Immunocytochemical detection of neuronal subtypes in 3D neurospheres. Presence of synapses was determined with post-synaptic marker PSD95 and synaptic protein Synaptophysin (SYNP) double staining. Glutamatergic (VGLUT1/2), GABAergic (GAD65/67), cholinergic (VAChT) and dopaminergic (TH) neurons were detected in the developing 3D neurospheres from D28. All samples were stained with TUBB3 (in red) to label the neurites. Protein name IDs are indicated with colors, representing the color of the used fluorophore (e.g., green as Alexa 488; red as Alexa 594). Nuclei were counterstained with DAPI (in blue). Scale bar: 25 μm.

Neuronal subtypes were investigated by the expression of glutamatergic (VGLUT1/2), GABAergic (GAD65/67), cholinergic (VAChT) and dopaminergic (TH) neurons with ICC and qRT-PCR (*GRIN1*: Glutamate Ionotropic Receptor NMDA Type Subunit 1; *GAD1*: Glutamate Decarboxylase 1; *CHAT*: Choline *O*-Acetyltransferase; *TH*: Tyrosine Hydroxylase; *SLC6A4*: Solute Carrier Family 6 Member 4) (Figures 3 and 4). The expression of these markers was detectable mainly from D28 stage and gradually increased until the last investigated time point of the differentiation (D42), in line with the maturation process. It has to be remarked that the spontaneous differentiation of dopaminergic neurons was rare in the cultures, we identified only a few TH positive cells in the cultures (Figure 4).

During the neuronal differentiation, astrocytes started to emerge after week 2, based on the Aquaporin-4 (AQP4) or the Glial fibrillary acidic protein (GFAP) expression both at mRNA and protein level (Figure 2, Figure 3 and Figure S4 ). Oligodendrocyte differentiation followed a similar course, from D21 it was clearly detectable both at transcript (*Claudin 11*, *CLDN11*, formerly known as Oligodendrocyte marker 4; Figure 3) and protein level (Myelin basic protein, MBP; Figure 2A,B) which gradually increased from D35 with the maturation of the cells, however, still remained in a low expression level (Figure 2A,B or Figure 3). Our ICC results clearly show that the expression of mature neuronal and glial markers and neuron-neuron and neuron-glial interactions increased during the differentiation period, thus emulating the development of the human fetal neural tissue.

Overall, results at the transcript level were in accordance with those of the proteins detected by ICC, confirming the effective differentiation of hiPSC-derived NSCs towards the neural lineage. The 3D neurospheres represented a neuronal tissue-like differentiation, containing neurons, astrocytes and oligodendrocytes, what was presented by transcript-, protein-, and ultrastructural level as well. We can conclude that the 3D spheroid system provides a complex neuronal cell culture which can serve as a model of the early neuronal differentiation.

#### *3.2. Early 3D Neurospheres as a Neurotoxicity Model*

Our 3D spheroid-based model was tested in cytotoxicity assay, monitoring the viability of the cells within the neurospheres. Eleven compounds with well-known effects (drugs, pesticides, and chemicals) were selected and applied on the neurospheres in 7 different concentrations (compounds and concentrations are detailed in Table S3) at D21 by acute (72 h) exposure. Concentration-response curves were generated and evaluated. The exposure scheme is presented in Figure 5A.

According to the results, Paraquat (PQ; EC50: 1.89 log μM), Rotenone (ROT; EC50: −0.61 log μM), Mercury(II) chloride (HgCl2; EC50: 1.87 log μM) and Doxorubicin (DOX; EC50: 0.67 log μM) caused maximum cell death, Hexachlorophene (HE; EC50: 1.26 log μM) and Colchicine (COL; EC50: −0.49 log μM) had a strong effect on the viability, but did not kill all the cells in the investigated concentration range. Acrylamide (ACR; EC50: 3.5 log μM) and Rifampicin (RIF; EC50: > 2 log μM) also reduced the viability, while Valproic acid (VPA; EC50: > 2.6 log μM) and Paracetamol (PAR; EC50: > 2 log μM) had minimal effect on the viability of 3D neurospheres at D21 stage. Ibuprofen (IBU; EC50: > 2 log μM) as a non-neurotoxic agent was applied as negative control which indeed, did not decrease the viability of D21 3D neurospheres (Figure 5B, Table S4). Overall, the different compounds induced different levels of cytotoxicity in a concentration-dependent manner on the 3D neurosphere cultures.

**Figure 5.** Cell viability measurement on D21 3D neurospheres after 72 h exposure. (**A**) Exposure scheme at D21 stage. (**B**) Concentration response curves of compounds, tested in 7 different concentrations (see concentrations listed in Table S3), representing the cell viability (%) of treated D21 3D neurospheres (*n* = 3). Concentration values are presented in log μM ± SEM. EC10 and EC50 values are presented on graphs where applicable.

*3.3. Di*ff*erent Age of the 3D Neurospheres Represent Distinct Di*ff*erentiation Stages in the Cytotoxicity Model*

Next, we investigated the concentration-response of different differentiation stage-derived 3D neurospheres. D28 and D42 samples were analyzed in exposure schemes similar to that of the D21 samples (Figure 6A or Figure 7B) for all the previously tested compounds. As detailed above, in D28 samples differentiated astrocytes expressing GFAP and AQP4 are present (Figure 2 and Figure S4), axonal outgrowth is prevalent, and neuronal subtype-specific proteins start to appear (Figures 2–4). The D42 samples represent a more mature cell culture where synapsis of neurons was formed with established subtypes, mature astrocytes and a few oligodendrocytes are already present (Figure 2, Figure 4, Figures S3 and S4).

**Figure 6.** Cell viability measurement on D28 3D neurospheres after 72 h exposure with toxicants. (**A**) exposure scheme at D28 stage. (**B**) Concentration-response curves of compounds, tested in 7 different concentrations (see concentrations listed in Table S3), representing the cell viability (%) of treated D28 3D neurospheres (*n* = 3). Concentration values are presented in log μM ± SEM. EC10 and EC50 values are presented on graphs where applicable.

**Figure 7.** Cell viability measurement on D42 3D neurospheres after 72 h exposure with toxicants. (**A**) exposure scheme at D42 stage. (**B**) Concentration-response curves of compounds, tested in 7 different concentrations (see concentrations listed in Table S3), representing the cell viability (%) of treated D42 3D neurospheres (*n* = 3). Concentration values are presented in log μM ± SEM. EC10 and EC50 values are presented on graphs where applicable.

The experiments showed that both D28 and D42 stages are suitable for the viability assays without modifications in the culture system, using a 96-well plate format-based analysis. The EC50 and EC10 values for D28 and D42 samples were determined for each compound, as described above for D21 samples (Figure 6B or Figure 7B, Table S4). Some compounds (Figures 5–7, Table S4) showed less toxic effect on more matured samples (e.g., ACR: EC50D21 = 4.1 μM vs. EC50D42 = 3.6 μM) while it was the opposite action for other compounds (e.g., HgCl2: EC50D21 = 1.7 μM vs. EC50D42 = 1.9 μM). Based on the data obtained, a concentration-related ranking of the toxicant could be generated for each differentiation stage (Figure 8, Table S4).

**Figure 8.** Comparison of the toxic effect of compounds. (**A**) Ranking of compounds at different differentiation stages based on EC50 values. Darker background color represents higher toxicity, while white color represents non-toxic compounds. (**B**) Radar chart showing the comparison of compounds based on the order of EC10 or EC50 values, at various differentiation stages (yellow line with rectangular: D21; orange line with circle: D28; green line with square: D42) (see also EC values in Table S4). Concentration values are presented in log μM.

In conclusion, we can state that a compound-specific concentration-response was detected in all stages. At the same time, a differentiation stage-relevant difference was observed, suggesting that the in vitro system can mimic the differential responses of the developing fetal human neuronal tissue, caused by distinct toxicants.

#### *3.4. Compound-Specific Cellular Events Can be Detected in the 3D Neurospheres*

ATP-based viability assay can detect cell death but is unsuitable for detecting other specific cellular events. Due to the lack of validated tests for 3D tissues, we decided to investigate the effect of a toxic compound at a subcellular level. We have chosen rotenone (ROT), a well-studied neurotoxic compound, known to interfere with the electron transport chain in mitochondria, to investigate the ultrastructure of the cells and especially mitochondria, followed by TEM. In parallel, a TUNEL assay was also applied to detect cell death. Based on the previously determined concentration-response curves and EC50 values, we treated the 3D neurospheres with 0.5 μM ROT for 72 h and analyzed the samples in the 3 differentiation stages. The data showed that ROT significantly increased the cell death in the spheroids (Figure 9A or Figure S6), compared to the vehicle-treated controls (Figure 9B), in accordance with the results of the ATP measurement (see Figure 5B or Figure 6B). Moreover, TEM revealed a change in the ultrastructure of the mitochondrial inner membrane, the cristae: Both shape and complexity were changed in the ROT treated samples, compared to the controls. Two types of mitochondria were detected, often in the same cell. Organelles with darker matrix were usually narrow and elongated, whereas lighter ones were more rounded. In control cells, cristae were straight, narrow, and long in both types. Matrix granules were observable more frequent in darker than in lighter mitochondria (Figure 9Ca–c). Effect of rotenone treatment was different on the two types of mitochondria. In darker organelles, cristae swelled and formed swirls. Several wide crista junctions became identifiable. In lighter mitochondria, cristae disintegrated and often became unrecognizable (Figure 9Cd–i). Matrix granules disappeared or showed decreased density (Figure 9Cc,i).

**Figure 9.** Effect of Rotenone (ROT) exposure on 3D neurospheres at three differentiation stages. (**A**) 3D neurospheres were treated with 0.5 μM ROT concentration for 72 h at three different maturation timepoints (D21, D28, and D42), fixed, sectioned and analyzed to detect the cellular effect of ROT by TUNEL assay (DeadEnd™ Colorimetric TUNEL System, Promega), compared to the vehicle (0.1% DMSO) treated control (scale bar: 100 μm). (**B**) ROT treatment revealed in average a 15% increase in the apoptotic cell number compared to the control in each stage (\* *p* < 0.05). Average values are presented on graphs (*n* = 3). (**C**) Ultrastructure of mitochondria in control (panel a, b, c) and ROT treated (d–i) neurons in the 3D spheroids. See the alteration of the internal membrane (cristae) morphology (white arrowheads) in ROT treated cells (d, e, h, i). Black star: unidentifiable cristae morphology in lighter mitochondria (h, i); black arrowhead: membrane swirl in darker organelles (e, f); white arrows: matrix with and without matrix granules (control cells: a, c; treated cells: g, i) Note the density difference between these granules in control (c) and ROT treated mitochondria (panel i) (D21: a, d, f, h; D28: i; D42: b, c, g) (scale bar: 250 nm).

Finally, we used an ER-stress assay to detect if the observed change is compound-specific or an "overall" cell death event is detected with the ATP or TUNEL assay. The XBP1 assay was performed to identify ER-stress in the ROT treated samples, where ER stress was not expected. As a positive control, tunicamycin was used. The XBP1 assays demonstrated that ROT (using EC50 concentration) has no ER-stress inducing effect in the D21, D28 or D42 neurospheres (Figure S6B), while cell death

was observed in the cryosectioned ROT-treated samples (Figure 9A), providing a strong evidence that compound-specific effects can be determined in 3D spheroids upon treatment. Immunocytochemical investigation reflected some disorganization of the treated group of spheroids compared to the controls, however other marked differences were not observed (Figure S6A).

#### *3.5. Neurite Outgrowth Assay is Suitable to Determine the E*ff*ect of NT Compounds in the 3D Model*

Functional readouts which connect the effect of a tested compound to a given cell type are essential to determine their tissue-specific effect. 3D spheroids represent a complex cell culture with a network of neurons, astrocytes, and oligodendrocytes, and this complexity makes more challenging to analyze minor differences in neurite length affected by chemical exposure due to the limitations of the current detection systems. Therefore, we investigated if neurite growth could be investigated using 3D neurospheres. We found that Matrigel-coated surface sufficiently supports the rapid and strong attachment of free-floating D21 3D neurospheres in order to generate a robust and reproducible procedure, where the effect of toxic compounds on neurite outgrowth can be investigated. The effect of different ROT concentrations were demonstrated comparing the total neurite outgrowth and the average number of neurites/spheroid ratio. ROT administration resulted significantly shorter and decreased number of neurites compared to the untreated and vehicle-treated control groups in all concentration levels in D21 spheroids (\*\* *p* < 0.01) (Figure 10B).

**Figure 10.** Neurite outgrowth measurement on D21 3D spheroids, exposed for 72 h with ROT. (**A**) Representative photograph of control (untreated) and ROT (0.5 μM) treated D21 spheroid immunolabeled with TUBB3 (in green). White lines represent the border of the spheroids where the neurite outgrowth was determined from, using ImageJ software (scale bar: 100 μm). (**B**) Total neurite length/spheroid (presented in μm ± SEM) and the average number of neurites/spheroids were determined 24 h after plating the treated spheroids. Different symbols denote treatment groups. (*n* = 3, in each experiment 8 spheroids were treated in each group) (\*\* *p* < 0.01).

#### **4. Discussion**

The development of in vitro platforms for neurotoxicology screenings is driven by the urgent needs of the chemical, food, cosmetic, and pharma industries. Most NT studies are carried out in rodents or rodent derived primary cells, resulting in relatively high cost and lower translational value of the results due to the species differences [3,46]. Major international initiatives have started to convert the traditional animal-based neurotoxicity tests to in vitro assays using both mammalian brain cells and human cells to detect and predict chemical hazards [47,48]. However, there is only a limited number of human neuronal cell lines (e.g., carcinoma cell lines such as SH-SY5Y; BE2-M17 or immortalized cell lines like LUHMES) are available and hard to obtain primary human CNS tissue suitable for NT studies. Overall, the highly complex structure of the human brain makes in vitro modeling very difficult. Human iPSCs could fill this niche and offer the advantage that other cells and tissue types (e.g., kidney, liver, cardiac, neuronal, intestinal) sharing the same individual genetic background can be created using specific differentiation protocols in a replicable manner. This may provide a very effective in vitro tool for toxicologist for capturing the individual variability in the human population [49]. For example, in 2D neurotoxicity screening, the usage of human iPSC derived neuronal cultures, especially the commercially available QC controlled neuron and astrocyte cultures, where the differentiation of iPSCs is not required for the "users", is dynamically increasing [48].

In recent years, numerous in vitro models have been created in order to study the human CNS in a more physiological way, but the field of NT and DNT using 3D neuronal tissues did not develop as fast as disease or developmental modeling. It is due to the difficulty to find a compromise between biological complexity and technical reproducibility which are necessary for drug or toxicity screening [29].

An important approach was the application of human cell lines in NT assay development. For example, the LUHMES immortalized human fetal tissue-derived mesencephalic cell line can be efficiently differentiated towards dopamine-like neurons upon tetracycline administration [50,51] and provide a suitable system for neurotoxicology screenings [52,53] or Parkinson's disease-related drug testing [54] both in 2D or 3D cultures [45,55–57]. Despite the advantages in straightforward handling, they represent only one specific cell type of the CNS. New developments using stem cell-derived astrocytes or microglia in co-culture with LUHMES cells could reveal new potential both in NT and DNT tests [58], but also highlight that full-PSC derived systems could provide great advantages over the conventional cell lines.

To develop an efficient, highly reproducible neurotoxicology test system, an improved 3D culture of microtissues differentiated from human PSCs would be beneficial. For example, in neuronal disease modeling, a major step that opened new perspectives, was the development of cortical layer-organized 3D brain microtissues, providing a complex system forming under in vitro conditions from human PSCs [59,60]. However, for toxicological studies, the very low throughput potential of such complex systems is a significantly limiting factor at the moment.

Huang and his colleagues demonstrated that human brain organoids could be applied as an in vitro model for CNS drug screening to evaluate structural, cellular, and molecular changes. They used neurotoxic tranylcypromine in hiPSC-derived brain organoids leading to decreased proliferation activity and apoptosis induction [61]. HiPSC-derived cerebral organoids were treated with different concentrations of vincristine for 48 h, and the expansion of the treated organoids was measured, showing concentration-dependent neurotoxicity. Vincristine inhibited fibronectin, tubulin, and MMP10 expression in the cerebral organoids, which was specific for its well-known effect on microtubule dynamics [62]. Recently, iPSC-derived cortical neurons and astrocytes were co-cultured in 3D to detect calcium oscillations upon a chemical compound treatment, analyzing multiple parameters, highlighting the potential of such readouts in neurotoxicity assessment [31]. Another study has presented a novel 3D heterotypic glioblastoma-brain sphere (gBS) model applied for screening new anti-glioblastoma agents [63]. This new application highlighted the flexibility of iPSC-derived platforms to be used in disease modeling, drug testing, and toxicology.

In this study, we have analyzed the complexity of our hiPSC-derived 3D neurosphere system demonstrating intensive glial-neuron interaction with the astrocytes and oligodendrocytes present in the neurospheres. Complex characterization was performed with gene expression, protein level analysis in line with immunocytochemical investigations, including morphological evidence of neurocrine communication at the end of the examined period. Morphologically, synapses are composed of precisely opposed pre- and postsynaptic membranes decorated with electron-dense thickening, synaptic cleft filled with a fine meshwork of electron-dense material, and presynaptic vesicles [64]. One of the two populations of presynaptic vesicles belong to the active zone matrix, is the docked vesicles [65]. The membrane of them is in contact with the presynaptic membrane, therefore they form the readily releasable pool (RRP) of synaptic vesicles. [66].

Our system showed similarities in morphology and maturation properties with a previous study where 3D organoids were generated and characterized over 8 weeks [29]. In our system, there was no need to use BDNF, GDNF, or a special electrophysiology media; a basal medium was sufficient to promote the differentiation of complex spheroids within 6 weeks of culture. Although the spheroids showed a continuous growth and development/maturation, the homogeneity of the plates was not compromised, both the diameter and total protein content of the spheroids within the 96-well plate showed very low variation at a given time point, represented by the low SEM values (Figure 1). Low variability of samples and high homogeneity is crucial when the aim is the development of a reliable HTS for DNT studies.

Here, we performed a medium-throughput 96-well plate assay on 3D neurospheres to detect the cytotoxic effect of the selected compounds, based on ATP-release measurements. We tested drugs, pesticides and well-known chemicals with or without neurotoxic effect in different stages of the neuronal differentiation. Our results highlighted a compound-specific and differentiation stage-related effect of the tested chemicals, providing the possibility to determine EC50 and EC10 values for the compounds (Table S3). For example, colchicine has a strong toxic effect on neurospheres with very similar EC50 values in all differentiation stages. Its effect can be detrimental, it blocks basic cellular (protein assembly, endocytosis, exocytosis, cellular motility, etc.) and neuronal functions at the same time (assembly of tubulin filaments in the microtubules of neurites), which we supposed to happen in all differentiation stages in our neurospheres. A comparable neurotoxic effect was described on iPSC derived 2D cultures of NSCs, neurons, and astrocytes [67], and LUHMES cells [57]. By comparing EC10 or EC50 values, another example is doxorubicin, where D42 samples were the most sensitive for the treatment when intensive protein synthesis could happen both in neurons and glial cells (e.g., neurotransmitter synthesis, axonal growth, oligodendrocyte maturation), therefore the potential blocking of the transcription machinery might have a major effect on cell viability (see Table 1 for a summary).


**Table 1.** Known effects of the used compounds.


#### **Table 1.** *Cont.*


**Table 1.** *Cont.*

Note: probable lethal dose data is provided upon PubChem database (https://pubchem.ncbi.nlm.nih.gov/), in vitro multicellular cytotoxicity "comptox" data were collected from EPA's Chemical dashboard (https://comptox.epa.gov/ dashboard).

Continuing the line with hexachlorophene that blocks the electron transport chain, by acting on GLUD1, it affects the turnover of an excitatory neurotransmitter, glutamate. This effect could explain the observed differentiation stage-related sensitivity differences of the neurospheres, still under an acute exposure scheme. Likewise, less cytotoxic effect was reported on iPSC derived 2D cultures of NSCs, early neurons (14 days old) and astrocytes when 10 or 100 μM HE was exposed for 24 h [67], while more pronounced cytotoxicity and neurite outgrowth inhibition of LUHMES cells were reported [57].

Valproate is known as a DNT positive compound. However, in our cell viability assay, it has no detectable impact on the investigated concentration range, which might be in correlation with the maturity of the treated cells, they are not in early neurodevelopment (neural tube stage), but maturing neurons and astrocytes. When LUHMES cells were exposed the EC50 value of VPA was determined as >100 μM, and significant inhibition of neurite outgrowth was detected. While the EC50 value was in a similar range with our 3D neurosphere-driven data, we did not investigate its effect on neurite outgrowth. Here, we also have to clarify that cytotoxicity read-out alone is not suitable to predict DNT, although it is suitable to predict the neurotoxic effect of compounds or detect necrosis in differentiating cultures, more relevant end-points (e.g., proliferation, migration, apoptosis, network formation, synaptogenesis, and growth of neurites) must be investigated for evaluating the DNT effect of a compound [18].

In the case of pesticide rotenone, which inhibits mitochondrial Complex I of the electron transport chain, strong cell death was detected. The effect of ROT exposure was investigated extensively. In iPSC-derived neuron-astroglia 2D cell cultures the activation of the Nrf2/ARE pathway was documented upon ROT-induced oxidative stress, which led to the activation of astrocytes and cell death of neurons [67,78,79]. Using LUHMES cells this robust cytotoxic effect was not obvious while a prominent neurite growth inhibition was reported [57]. Using "BrainSpheres" system [30], where ICC-based morphology analysis, ROS measurements and viability were compared in different developmental times and exposure schemes, dopaminergic-neuron selective toxicity and general cytotoxic concentration were determined [30]. Importantly, increased cell death and mitochondrial dysfunction were detected in our ROT exposed 3D neurospheres, similar to others' results [30]. In contrast, ibuprofen has no effect on neuronal cell viability in the investigated concentration range, as it was expected upon known in vitro and in vivo data.

Neurite outgrowth assays are well established in conventional 2D neuronal cultures, derived from cell lines, primary tissues, or iPSCs [57,80,81], however, the adaptation to 3D cultures is less tested. Determining the radial migration and neuronal density distribution within the migration area of NPCs grown as spheroids is rather described [27]. Here we adapted a simple system which provides a tissue-specific functional readout for neurospheres upon exposure, which can be automated using high content imaging (HCI) systems [45]. Based on our results, it is sensitive enough to distinguish different concentration of toxic compounds, does not require complicated read-out systems which was proved by testing the effect of ROT, previously confirmed to inhibit neurite growth of LUHMES cells [57]. Although these results agree with the reported neurite outgrowth inhibition effect, we cannot exclude the influence of cytotoxicity on the reduction of the overall neurite length we observed. Combining with LDH or lactate sampling, metabolite measurements from the media or terminal transcriptomic or proteomic assays, a complex dataset can be collected, and the effect of a given compound can be investigated systematically on the molecular level, as well to identify adverse effects. Similar to the RT-qPCR and ICC investigation that we used to characterize our 3D neurospheres these methods can be used when the effect of a given compound is investigated. Although we performed immunocytochemistry on treated spheroids using rotenone only, others successfully used such investigations to gain DNT readouts, for example detecting cell migration, neurite outgrowth, Ca2<sup>+</sup> reabsorption, synaptogenesis, and PPAR pathway disruption [29,30,57,82,83].

Evaluating our concentration–response results, we faced with the lack of available in vitro human data, which makes the comparison challenging. This demonstrates that due to the lack of EC values and very different readouts, the appropriate comparison between the various test systems is often not possible. As summarized in Table 1, it is not possible to provide human in vitro NT values for every compound, the databases are incomplete. Although there are several initiatives and efforts on the field to collect and harmonize the available test methods and evaluate the data produced so far, there is still a lot to do in the field of in vitro neurotoxicity test method development [1,16].

#### **5. Conclusions**

We have established and evaluated a hiPSC-based 3D in vitro cellular model for the study of the neurotoxic effect of different compounds. Concentration–response was determined in different differentiation stages, using a well-known read-out the cell viability. For a certain compound, apoptotic activity, ER-stress assay and TEM was performed, providing possibilities to generate novel human CNS-relevant data for other compounds using hiPSC derived cells. Tissue-specific readouts such as neurite outgrowth were also investigated, however further, read-outs and DNT-relevant training compound set should be applied to evaluate the suitability of our 3D spheroids in DNT screening Nevertheless, due to the pluripotent nature of the hiPSC, this model offers an excellent tool for drug testing, gene therapy studies and toxicology studies parallelly on the same genotype using other cell or tissue types at the same time. Moreover, new advancement in gene manipulation such as CRISPR/Cas9 mediated gene targeting, makes it possible to target specific pathways and generate reporter cell lines for toxicological or other applications. Combining these new approaches with 3D cell culture-based assays could revolutionize the field of toxicology including DNT studies in the near future.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/9/5/1122/s1, Figure S1: Characterization of CTRL-2 hiPSC line, Figure S2: Characterization of NPCs, Figure S3: Summary of the morphological differentiation of neurospheres, Figure S4: Representative immunostaining of cryosectioned neurospheres, Figure S5: The initial delta Ct distributions of relevant neuronal markers in D2 stage samples by Real-time PCR measurements, Figure S6: Effect of ROT treatment on 3D spheroids, Table S1: Antibodies used for immunocytochemistry and flow cytometry, Table S2: Primers used in the study, Table S3: Tested compounds and their concentration, Table S4: Effective concentration (EC) values of compounds tested on D21, D28 and D42 3D neurospheres.

**Author Contributions:** Data curation, M.Z. and I.B.; Funding acquisition, A.D.; Investigation, J.K., A.T., T.B., Z.J., K.M., I.B. and L.L.; Supervision, L.L. and A.D.; Writing—original draft, J.K.; Writing—review and editing, M.Z., L.L. and A.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 681002 (EU-ToxRisk).

**Acknowledgments:** We are grateful for the help of Balázs Mihalik in study design and assay development. We thank Mónika Truszka for the technical help in the preparation of samples for TEM.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Ethics approval and consent to participate:** Written informed consents have been obtained from all subjects providing samples for iPSC derivation. Ethical approvals were obtained from the Hungarian Medical Research Council (in Hungarian: *Egészségügyi Tudományos Tanács; ETT*) to establish and maintain hiPSC lines (ETT-TUKEB 834/PI/09, 8-333/2009-1018EKU).

**Consent for publication:** Not applicable.

**Availability of Data and Material:** Additional files are made available online along with the manuscript.

#### **Abbreviations**


#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
