Multi-Lineage Human iPSC-Derived Platforms for Disease Modeling and Drug Discovery

Abstract

Human induced pluripotent stem cells (hiPSCs) provide a powerful platform for disease modeling and have unlocked new possibilities for understanding the mechanisms governing human biology, physiology, and genetics. However, hiPSC-derivatives have traditionally been utilized in two-dimensional monocultures, in contrast to the multi-systemic interactions that influence cells in the body. We will discuss recent advances in generating more complex hiPSC-based systems including three-dimensional organoids, tissue-engineered constructs, microfluidic organ-chip platforms, and humanized animal systems. While hiPSC differentiation still requires optimization, these next-generation multi-lineage technologies can augment the biomedical researcher’s toolkit and enable more realistic models of human tissue function.

eTOC:

Sharma et al review recent advances in multi-cell lineage, hiPSC-derived platforms for disease modeling and drug screening. These complex model systems, which include organoids, assembloids, organ-chips, and tissue-engineered constructs, are more realistic representations of human tissues containing multiple cell types that interact during development, aging, and disease.

Introduction

The human body is comprised of at least 200 cell types (Regev et al., 2017), and a variety of model systems have been used to study multi-cell type and systemic interactions (Figure 1). However, animal models in particular exhibit shortcomings when extrapolating findings to the human condition due to species-specific differences in physiology, metabolism, and genetics. These differences can lead to inaccurate results in preclinical drug safety or efficacy studies and ultimately lead to drug failure during clinical trials. Indeed, this may give credence to the “valley of death” scenario driving massive attrition during drug development (Butler, 2008). Preclinical human cell-based assays for drug toxicity evaluation traditionally utilize monocultured cells in isolation, such as immortalized cancer cell lines overexpressing a gene or protein of interest. While important for understanding fundamental cellular mechanisms, the relevance of these cells to human disease is questionable. Thus, there is interest in developing a new generation of multi-lineage platforms that can safely, accurately, and simultaneously evaluate interactions between multiple cell types in the context of studying human disease and for drug discovery.

Figure 1: Catalog of cell types and publications associated with hiPSC-derived multi-lineage model systems.

Human iPSCs can model development and disease for a variety of somatic tissues. Differentiation protocols for the cell types comprising these somatic tissues have been developed and refined over the past two decades, initially through two-dimensional culture. In the past 5 years, the hiPSC field has experienced rapid development in organoid, tissue engineering, and microfluidic chip models. Key studies utilizing hiPSC-derived cells in a multi-lineage context are listed according to tissue type and culture platform.

Over the past 15 years, cutting-edge technologies have been developed and refined to model the complex interactions occurring between multiple cell types in the human body. A variety of model systems harboring multiple cell lineages have been employed, ranging from isolated cells in two-dimensional cell co-culture to complex animal models with humanized tissues (Figure 1) (Watson et al., 2017). While several adult human cell types such as neurons and cardiomyocytes are difficult to isolate and culture ex vivo, human induced pluripotent stem cell (hiPSC)-derived cells can now overcome this hurdle (Takahashi et al., 2007; Yu et al., 2007). Treatment with critical pluripotency genes or proteins can transform almost any adult tissue into pluripotent stem cells. These hiPSCs exhibit continued proliferation while maintaining the ability to differentiate into an array of somatic tissues using optimized differentiation protocols. Critically, hiPSCs differentiated into somatic cell types of interest can recapitulate disease phenotypes associated with genetic mutations (Passier et al., 2016).

While analysis of cell subtypes in a cell culture dish provides an excellent resource to study lineage-specific disease mechanisms, in reality cells in the human body do not function in mono-lineage isolation. Thus, understanding disease may require multi-cellular, integrated platforms, which exhibit distinct advantages over mono-lineage analyses (Table 1). Genetic and environmentally-influenced diseases are commonly multi-lineage or multi-systemic in impact, with even a single affected cell type causing a domino-effect that influences downstream pathophysiology in diverse tissues. Similarly, drug toxicity can manifest in multiple organ systems, especially in the case of chemotherapeutic compounds known for their potentially catastrophic off-target toxicities (Burridge et al., 2016). Multi-cell lineage analysis using hiPSCs can also be integrated with personalized medicine, to enable both drug efficacy and toxicity predictions within specific patient populations with unique genetic profiles. A new generation of model systems is being developed to meet rising demand for multi-lineage analysis. These may vary based on their scalability and physiological relevance, but they can be used in combination to elicit a more accurate mechanistic understanding of human biology (Figure 2).

Table 1:

Advantages and challenges of multi-cell lineage model systems using hiPSC-derivative cell types

Multi-lineage platform	Advantages	Challenges
2D co-culture	Highly scalable and amenable to rapid interrogation of cellular, transcriptional, genetic, and signaling mechanisms	Unable to match the 3D physiological complexity associated with other in-vitro systems
In-silico	Large computational power enables highly scalable predictive modeling of biological processes and drug responses.	Non-biological platform and dependent on input from accurate biological datasets to establish a useful predictive model
Microfluidic organ and body chips	More accurate mimic of in vivo physiological forces than traditional 2D cell culture. Integrates the concept of systemic vasculature through microfluidic channels. Simultaneous interrogation of multiple cell types in a fully-integrated system.	Modest scalability, expensive, and necessary to find appropriate biocompatible materials for cell culture
Tissue engineering and 3D printing	Three-dimensionality enables more realistic mimicking of in vivo tissues with potential functional uses for human transplantation	Less scalable than traditional monoculture and cost of bioprinting and bioengineering platforms can be a limiting factor. Biocompatible bioinks must be further refined.
Organoids and assembloids	More realistically mimic in vivo developmental processes in an in vitro setting while retaining moderate scalability	Maximum size is limited without adequate vascularization. Potential for tissue necrosis within organoid interior.
Humanized animals and chimeras	A physiologically relevant in vivo platform for studying development and potentially serving as a system for mass producing human tissues for transplantation purposes	Expensive and limited throughput. Significant bioethical concerns with growing human reproductive, neural, and other tissues in animals.

Figure 2: Evolution of multi-lineage hiPSC-derived platforms and comparison of model throughput versus physiological relevance.

Since the discovery of hiPSCs in 2007, hiPSC-derived multi-lineage platforms have grown in complexity from simple, two-dimensional cultures to engineered platforms and complex organoids. These hiPSC-derived multi-lineage platforms strive to maintain a balance between high throughput and high physiological relevance, with an ideal model exhibiting both characteristics. Future work may incorporate hiPSCs in humanized animal models and organ chimeras to enable human tissue production in large animals.

Two-dimensional co-culture of hiPSC-derivatives

The traditional two-dimensional co-culture of hiPSCs and cell types derived from them provides the simplest form of a multi-lineage, hiPSC-based, in vitro model. In fact, the first hiPSCs were grown in two-dimensional co-culture with irradiated mouse embryonic fibroblasts, which provided pluripotency-maintaining signaling to hiPSC colonies. In some instances, feeder cells are still used to maintain the multipotency of hiPSC-derived tissue progenitors (Toyohara et al., 2015; Trott et al., 2017). During differentiation, hiPSCs have also been co-cultured with supporting somatic cell types from immortalized cell lines, since morphogens secreted from support cells can enhance differentiation towards a lineage of interest (Choi et al., 2009; Dias et al., 2011; Freyer et al., 2017; Schweizer et al., 2017; Yang et al., 2014a; Zhang et al., 2018a). These support cells can also be derived from primary mouse or human tissues (Qu et al., 2013; Songstad et al., 2017; Taguchi et al., 2014; Tajiri et al., 2018; Teotia et al., 2017; Xia et al., 2013). However, efforts to simplify and standardize hiPSC-derived cell type production have led the stem cell field to shift away from co-culture based methods in favor of chemically or genetically defined, feeder-free approaches. Today, the growth and differentiation of hiPSCs is mainly induced with small molecules, growth factors, or ectopic transcription factor gene expression (Burridge et al., 2014).

HiPSC-derived cell types can be grown together in two dimensions with other cells to mimic a tissue of interest. This has been shown to more accurately recapitulate phenotypes associated with cellular maturity, toxicity, and function (Berger et al., 2015). Astrocytes and neurons have been grown together in two-dimensional co-cultures in an effort to recapitulate the support cell network found in the brain (Brennand et al., 2011; Israel et al., 2012; Wen et al., 2014). Such glial-neuronal co-culture has resulted in improved neural network development as well as enhanced cellular maturity and action potential propagation (Amiri et al., 2013; Ishii et al., 2017). Co-culture with neuronal cell types is used to assess the ability of hiPSC-derivatives to establish synaptic connections (Chen et al., 2018a; Demestre et al., 2015; Gunewardene et al., 2016). For other ectodermal derivatives, hiPSC-derived skin precursors upregulated expression of mature keratinocyte markers when co-cultured with human dermal cells (Veraitch et al., 2013). HiPSC-derived cardiomyocytes (hiPSC-CMs) exhibited improved functionality when co-cultured with mesenchymal stem cells, which secrete soluble growth factors that may enhance cellular maturation through activation of receptor-mediated signaling cascades (Yoshida et al., 2018). Fibroblasts co-cultured with hiPSC-CMs influence cardiomyocyte hypertrophy, function, and gene expression (Josowitz et al., 2016; Zhang et al., 2019). Maturation of hiPSC-CMs in particular has represented an ongoing challenge worthy of being addressed. Since functionally mature hiPSC-CMs would elicit the force output and electrophysiological properties found in adult heart muscle cells, hiPSC-CMs have promising applications for cardiac cell therapy and disease modeling. Although scalability has improved for a majority of hiPSC-derived cell types, immaturity of cells such as hiPSC-CMs can limit their clinical potential (Tu et al., 2018). For the heart, recent work demonstrates that co-cultured hiPSC-CMs and epicardial cells could enhance maturity and potentially lead to more long-term functional integration into heart tissue (Bargehr et al., 2019).

Two-dimensional, multi-lineage co-culture could also represent a more accurate model than monoculture for assessing drug toxicity. Neurotoxicity, hepatotoxicity and cardiovascular toxicity represent major causes of drug withdrawal from the pharmaceutical market, and platforms to accurately assess drug toxicity and efficacy in one or more cell types are in high demand. For example, astrocytes induced a neuroprotective phenotype in neurons subjected to neurotoxic chemicals (Amiri et al., 2013). Co-culture of hiPSC-derived hepatocytes with fibroblasts has shown to more accurately model liver metabolism of toxic drug byproducts (Ware et al., 2015). Indeed, a hepatocyte co-culture may be required for in vitro drug efficacy and toxicity screening assays to mimic liver-mediated metabolism of a prodrug into the active form experienced by the body (Jang et al., 2019; Takayama et al., 2019). Additionally, chemotherapeutic compounds such as tyrosine kinase inhibitors elicit simultaneous cardiac and endothelial toxicity and, therefore, co-culture could be valuable for assessing drug toxicity (Sharma et al., 2017). Endothelial cells represent a major cell type of interest for multi-lineage models, given their integral role in lining the blood vessels that supply nutrients and oxygen to all tissues of the body. However, modeling a barrier-interface system in a two dimensional multi-lineage system is difficult, and may be more amenable to complex three-dimensional and microfluidic model systems. Additionally, immune system-mediated inflammation is associated with numerous cardiovascular, neurological, and infectious disorders (Chen et al., 2018b). Through hiPSC technology, it is now possible to introduce simplified aspects of the immune system into a two-dimensional co-culture system. For example, hiPSC-derived endothelial cells are able to interact with and bind leukocytes both in static conditions and in mimicked blood flow (Adams et al., 2013; Patsch et al., 2015). Overall, two-dimensional co-culture represents an easily accessible platform for studying multi-cell type interactions, though there is clearly a need for more complex, three-dimensional model systems to attain greater physiological relevance.

hiPSC-Derived Organoids

Improvements in the directed differentiation of hiPSC-derived organoids have taken the stem cell research community by storm. Organoids are three-dimensional, miniaturized, self-organizing cellular aggregates containing multiple cell types, and hence represent more structurally accurate analogues of human organs and tissues. Thus, they may provide more faithful models of development and disease, when compared to standard two-dimensional cell culture (Takebe and Wells, 2019). The naturally-occurring counterparts to hiPSC-derived organoids are teratomas, masses containing cells derived from all three germ layers. Teratomas are similar to hiPSC-derived organoids, and indeed can serve as an in vivo assay to validate hiPSC differentiation potential, but arise via spontaneous, undirected differentiation. Under the right growth, survival, and high-density culture environments, three-dimensional hiPSC-organoid differentiation enables the production of an exponentially greater number of cells than in two-dimensional monoculture. Due to their small size, organoids still retain the scalability of traditional two-dimensional monoculture, enabling their use in high-throughput screening (Czerniecki et al., 2018).

Ideally, hiPSC-derived tissue-specific organoids can self-assemble, recapitulate the fundamental developmental mechanisms that ultimately give rise to an organ of interest, and harbor the appropriate cell types found in that tissue. A variety of hiPSC-derived organoids representing different bodily tissues have been recently developed, with the notable exception of cardiac organoids. Spherical cardiac microtissues, although not completely able to self-organize and accurately recapitulate the structural and spatial dynamics of the developing heart, have been developed via bioengineering or forced amalgamation of hiPSC-derived cardiomyocytes and used for drug screening (Hoang et al., 2018; Mills et al., 2019; Mills et al., 2017; Voges et al., 2017). Although these cardiac microtissues do not contain functional vasculature, miniaturized blood vessel organoids containing hiPSC-derived endothelial cells and pericytes have been able to model mechanisms of diabetic vasculopathy (Wimmer et al., 2019). Staying within the realm of blood, T-cells have been generated from pluripotent stem cell-derived organoids in an effort to create off-the-shelf products for chimeric antigen receptor T-cell (CAR-T) therapy (Montel-Hagen et al., 2019; Motazedian et al., 2020). Lung organoids from pluripotent stem cells show rudimentary bronchiole-like structures and express alveolar-cell markers (Dye et al., 2015; Gotoh et al., 2014; Miller et al., 2019; Strikoudis et al., 2019; Wilkinson et al., 2017). Pancreatic organoids are able to secrete insulin in response to glucose and are rapidly advancing towards the clinic as a potential treatment for diabetes (Hohwieler et al., 2017; Pagliuca et al., 2014; Raikwar et al., 2015). Similarly, vascularized liver organoids generate albumin and appropriately metabolize drugs of interest (Guan et al., 2017; Takebe et al., 2013; Takebe et al., 2017). Retinal and corneal organoids are also functionally appropriate and harbor key cell types such as light-responsive photoreceptor cells (Foster et al., 2017; Hallam et al., 2018; Susaimanickam et al., 2017; Zhong et al., 2014). HiPSC-derived kidney organoids have rudimentary nephrons, and 3D culture combined with active fluid flow has led to their enhanced vascularization and maturation (Forbes et al., 2018; Low et al., 2019; Morizane et al., 2015; Takasato et al., 2015; van den Berg et al., 2018). Multiple components of the digestive tract, including the esophagus (Trisno et al., 2018; Zhang et al., 2018b) colon (Crespo et al., 2017; Munera et al., 2017), and stomach (Broda et al., 2019; McCracken et al., 2017; McCracken et al., 2014) have been replicated using hiPSC-derived organoids. Intestinal organoids in particular are noted for their advanced substructure development, and display villus-like and crypt-like structures as well as enterocytes, goblet, enteroendocrine, and Paneth cells (Mithal et al., 2020; Spence et al., 2011; Takahashi et al., 2018; Watson et al., 2014). Directed differentiation of hiPSCs using Wnt and BMP signaling modulation has produced organoids composed of fallopian tube epithelium (Yucer et al., 2017). Even inner ear organoids, containing functional hair cells, have been derived from hiPSCs (Jeong et al., 2018; Koehler et al., 2017; Mattei et al., 2019). These studies demonstrate the explosive growth in the last decade in protocols for differentiating and culturing hiPSC-derived organoids from many lineages.

Although there have been tremendous advancements in hiPSC-derived organoids from a variety of somatic tissue types, the brain organoid field has perhaps experienced the most rapid progress in the past decade. Cerebral organoids contain not only neurons but also the glial support cell types normally found in the brain (Lancaster et al., 2013). For example, oligodendrocytes responsible for myelination have been generated within three-dimensional neural cultures (Madhavan et al., 2018; Marton et al., 2019). Recent studies have been able to recapitulate in a dish, advanced developmental processes such as neurulation using micropatterned organoids stimulated with morphogens under precise temporal control (Haremaki et al., 2019). Organoids can also represent an ideal cellular model to study the effect of developmental teratogens. For instance, Zika virus, known to cause microcephaly, can readily infect hiPSC-derived brain organoids and induce premature differentiation in neural progenitors (Gabriel et al., 2017; Qian et al., 2016; Wells et al., 2016). Similarly, using hiPSC differentiation as an analog for development, organoids can address the divergence in human brain development from other primate species. This is a unique application of organoids towards evolutionary biology, and has revealed that chromatin accessibility differs between human and non-human primate brain organoids in multiple cell types over the course of differentiation (Kanton et al., 2019; Pollen et al., 2019). Thus, brain organoids provide a new opportunity to examine neurodevelopmental processes.

In spite of rapid adoption by the stem cell community, major limitations of organoid cultures include hiPSC differentiation and culture inconsistency and inadequate vascularization. Nutrients are not evenly distributed within an organoid by diffusion alone, thus leading an organoid to often develop a necrotic inner core and limiting the maximum possible size. Cellular, genetic, and bioengineering-based approaches have attempted to address this issue by introducing endothelial cells or printing vascular structures within an organoid (Cakir et al., 2019; Pham et al., 2018). Vascularization can also be induced by transplanting an organoid into an in vivo model system to examine integration of organoids with the host vasculature (Mansour et al., 2018), and these hiPSC-chimera approaches are discussed further later in this review. Vascularization or prolonged culture may improve organoid development and maturation, leading to larger and more functional organoids. Over long-term culture, hiPSC-derived cortical organoids can dynamically change their constituent cell populations and mature electrophysiologically to develop enhanced neural networks (Trujillo et al., 2019). Preliminary comparative studies have shown that these primitive organoid neural networks exhibited characteristics that vaguely resemble those found in electroencephalograms from pre-term human infants. However, using a longer time course to increase organoid and hiPSC-derivative maturation may not always work. If in vitro growth conditions are not comparable to the in vivo situation, cell maturation may create unintended cell types. Further, the maturation of hiPSC-derived multi-lineage organoids, cerebral organoids in particular, raises interesting bioethical questions (Chen et al., 2019). Some in the general public have referred to these immature cerebral organoids as “brains-in-a-dish”, which we believe is a gross oversimplification of the current state of the technology. But should the scientific community more closely consider advanced properties such as higher neural function and consciousness in these brain organoids? And if so, should they be more strictly regulated (Farahany et al., 2018)? Although we must closely evaluate such bioethical concerns, there is no doubting the scientific power of multi-lineage hiPSC-derived organoids in revolutionizing our understanding of human disease and developmental principles.

hiPSC-Derived Forced Aggregates and Assembloids

Taking developmental organoids a step further in complexity, recent studies have focused on artificially assembling multiple organoids or combining organoids together with cells from diverse tissue lineages to create forced-aggregate “assembloids” (Pasca, 2019). These advanced co-cultures can better model the interactions between multiple sub-regions and cell types found in complex organs. For example, hiPSC-derived dorsal and ventral forebrain organoids can be combined in vitro to recapitulate the migration of interneurons in the developing human brain and model neurodevelopmental disorders such as Timothy’s syndrome (Birey et al., 2017; Xiang et al., 2017). Similarly, fusing hPSC-derived thalamic and cortical organoids provides a platform for studying inter-regional circuit organization in the human brain (Xiang et al., 2019). Introducing rudimentary anteroposterior and dorsoventral axes in vitro using so-called “organizers” have attempted to recapitulate the complex three-dimensional neural architecture present in vivo (Cederquist et al., 2019). These organizers could be morphogens or cell types of interest that direct the formation of brain axes or the spinal cord. Creating a signaling gradient of the molecular organizer Sonic Hedgehog in forebrain organoids yielded self-assembly around the aforementioned axes, establishing forebrain subdivisions mimicking in vivo brain organization. Improved polarized assembloids will require a more complete understanding of the developmental signaling mechanisms required to properly specify the spatiotemporal axes in the developing nervous system. Similarly, multi-cell lineage brain assembloids can be made by introducing cell types specified outside the nervous system, such as microglia, endothelial cells comprising the blood vessels in the brain, or other support cells such as pericytes. Such multi-lineage hiPSC-derived assembloids have been used to study neural-glial interplay in Alzheimer’s disease (Lin et al., 2018) and could potentially be used to study the role of the immune system in autoimmune diseases affecting the nervous system, such as multiple sclerosis.

Assembloid-based studies extend beyond the brain as well. For instance hiPSC-derived intestinal organoids can be combined with neural crest cells to recapitulate intestinal enteric nervous system development and study motility disorders of the gastrointestinal tract (Workman et al., 2017). Similarly, hPSC-derived axial stem cells can simultaneously specify spinal cord neurons and skeletal muscle cells that self-organize to form hybrid neuromuscular organoids (Faustino Martins et al., 2020). Additionally, combining anterior and posterior gut hiPSC-derived organoids enabled the simultaneous, dynamic morphogenesis of hepatic, biliary, and pancreatic structures (Koike et al., 2019). Patient-derived glioblastoma tumor cells have been combined with cerebral organoids to study the mechanisms driving tumor metastasis (Ogawa et al., 2018). Extending this work further, cancer assembloids consisting of tumor cells and hiPSC-derived organoids of various tissue types could be used as a predictive model for metastasis to elucidate tissue types most susceptible to tumor cell migration and invasion, or conversely, chemotherapy. Assembloids could advance targeted, tissue-specific drug treatment in scenarios where drugs differentially affect subsections of an organ or tissue. For example, some drugs more potently affect the atrial versus ventricular chamber cardiomyocytes in the heart, and could cause serious cardiac arrhythmias or atrial fibrillation (Kaakeh et al., 2012). In fact, the heart could be one organ where complex assembloids faithfully recapitulate the multiple chambers that confer its physiological function. As mentioned, existing cardiac microtissues do not accurately recapitulate the three-dimensional, chamber-specific organization of cell types found in the heart, even though atrial and ventricular-like cardiomyocytes can be specified from hiPSCs (Cyganek et al., 2018). Perhaps atrium and ventricle-specific organoids, containing atrial or ventricular hiPSC-derived cardiomyocytes and nodal conduction cells, could be combined to generate a primitive, four-chambered heart consisting of two atrial organoids and two ventricular organoids. While a heart assembloid would not functionally be on par with the in vivo heart, given that hiPSC-derived cardiomyocytes only elicit a fraction of the force output of true cardiomyocytes, it could be a model to study how the heart chambers interact during development. Any organ with multiple chambers, sub-structures, or compartments could be modeled using hiPSC-derived assembloids, so long as the cells in those sub-structures can be produced in culture. The potential of forced-aggregate assembloids as a developmental model is only just beginning to be realized.

Using hiPSCs for Tissue Engineering Applications

At the intersection of bioengineering and stem cell biology is the concept of using hiPSCs for tissue engineering applications. In contrast to tissue-specific organoids that arise by mimicking in vivo developmental processes, tissue-engineered hiPSC-derived platforms are used to arrange cells in specific locations using biocompatible materials that can be natural, synthetic, or a combination of both. While this enables the tissue-engineer to align cells in specific locations, the field must improve how to accurately mimic interactions between different parts of an organ such as the brain. Alternatively, the creation of assembloids relies on nature to establish connections between different parts of an organ. Assembloids are created by the manual forced assembly and fusion of individual organoids or cell types, possibly by using morphogen gradients as substructure “organizers”. In contrast, when combined with hiPSCs or their derivatives, biomaterials can create customizable scaffolds that maintain cell integrity and viability (Wang et al., 2011). Such platforms can harbor multiple cell types and can be used for disease modeling, drug screening, or cell therapy. Bioactive scaffolds can direct stem cell differentiation into a lineage of interest (Willerth and Sakiyama-Elbert, 2008). Scaffolds comprised of cross-linked hydrophilic polymers and matrix proteins are useful for maintaining stem cell pluripotency and can also be injected for in vivo applications (Abazari et al., 2018; Enderami et al., 2018). Hydrogel scaffolds can contain growth factors, nanoparticles, or small molecules that can actively influence hiPSC-derivatives residing within the scaffold (Kuo et al., 2017; Zhang et al., 2016). Other scaffolds can be comprised of protein-based biomaterials such as fibrin or collagen commonly found in natural extracellular matrix (Batchelder et al., 2015; Fischer et al., 2017; Kitano et al., 2017). Fibrin and fibronectin are found naturally during the blood clotting process, making them highly biocompatible materials for stem cell adhesion within scaffolds (Wan et al., 2017). Similarly, collagen-based biomaterial scaffolds are common in stem cell tissue engineering applications, given that collagen is prevalent in the human body and facilitates cell-cell adhesion (Benedetti et al., 2018). Organ decellularization can generate a natural collagenous scaffold upon which hiPSC-derivatives can be introduced, and this process has been particularly useful for generating biomechanically appropriate, but delicate scaffolds for lung tissue engineering (Ghaedi et al., 2013; Ghaedi et al., 2018; Gilpin et al., 2014) Collagen can also be modified to produce gelatin, another material found in biocompatible hydrogels as “GelMA” or gelatin methacrylate (Kang et al., 2014). Improvements in biocompatible materials will facilitate improved cell survival and function in scaffolded and non-scaffolded tissue-engineered constructs.

Three-dimensional bioprinting is an extension of tissue engineering, whereby mechanical devices combine cellular building blocks with biocompatible “bioinks” to deposit scaffolded or scaffold-free constructs (Romanazzo et al., 2019). Using computer-aided design, a mechanical printer can deposit bioink containing hiPSC-derived cells at designated spatial coordinates, velocity, and volume. Bioinks are integral to the printing process as they must maintain cellular viability, and in the context of multi-lineage tissues, must be compatible across multiple cell types. The previously described hydrogels containing biocompatible materials such as collagen can be used as bioinks for printing undifferentiated hiPSCs or hiPSC derivatives.

Over the past 5 years, a number of hiPSC-derived, multi-lineage platforms have been bioprinted. Scaffold-free three-dimensional printing has generated cardiac platforms by printing cardiac spheroids comprised of endothelial cells, fibroblasts, and hiPSC-derived cardiomyocytes. As the native human myocardium is comprised of these cell types, it is hypothesized that a co-cultured, three-dimensional printed system would represent a more accurate model of the human heart. Printed cardiac spheroids were able to function as a rudimentary cardiac pump (Arai et al., 2018). Similarly, multi-lineage cardiac spheroids have been printed into patches and implanted into the native rat myocardium, which elicited some signs of vascularization (Ong et al., 2017). In addition to scaffold-free systems, scaffolded platforms have also gained traction in the cardiac bioprinting field. These include “engineered heart muscle” harboring hiPSC-derived cardiomyocytes and fibroblasts bound by collagen (Tiburcy et al., 2017), cardiac patches co-cultured with hiPSC-derived cardiomyocytes, smooth muscle cells, and fibroblasts, encapsulated in fibrinogen hydrogels (Shadrin et al., 2017), or heteropolar cardiac tissues composed of both atrial and ventricular hiPSC-CMs seeded in hydrogels (Zhao et al., 2019). Scaffolded platforms can also incorporate micropatterning to force cells such as hiPSC-derived cardiomyocytes to adopt more physiologically relevant shapes in hopes of enhancing cell maturity (Ribeiro et al., 2015; Serpooshan et al., 2017). Additionally, endothelial cell incorporation provides not only a potential support structure and enhanced cardiomyocyte survival through cellular crosstalk, but also is the basis for the development of an endogenous network of vasculature in tissue engineered cardiac constructs. As is the case with any large, thick tissue platform, nutrients must be able to diffuse to the tissue interior to prevent necrosis. Alternatively, vasculature itself could be engineered, and cutting-edge work has demonstrated that it is possible to reproducibility create hiPSC-derived vascular channels that could transport blood, air, or other critical nutrients (Grigoryan et al., 2019; Gui et al., 2016; Samuel et al., 2013). Sacrificial bioinks can be degraded to leave behind elaborate vascular channels (Skylar-Scott et al., 2019). The cardiovascular field has seemingly taken the forefront in the realm of multi-lineage tissue engineering and bioprinting, given the interest in remuscularization and revascularization of the heart after injury.

However, other tissue types have also benefitted from recent advances in tissue engineering and 3D bioprinting for multi-lineage construct generation. For example, 3D-cultured hiPSC-derived hepatocytes, when combined with support cells such as endothelial cells, demonstrate improved morphologies, hepatic gene expression, and liver-specific metabolism over two-dimensional cultures (Faulkner-Jones et al., 2015; Ma et al., 2016). Improvements in differentiation and scaffolding of hiPSC-derived osteoclasts and osteoblasts have unlocked new possibilities for generating replacement bone grafts using tissue engineering approaches (de Peppo et al., 2013; Jeon et al., 2016; Wang et al., 2014b). In the neural space, hiPSC-derived spinal neural progenitors and oligodendrocyte progenitors have been co-cultured to develop three-dimensional spinal cord platforms for modeling nervous tissue repair (Joung et al., 2018). HiPSC-derived neural progenitors can be scaffolded in biomaterials of varying stiffness, confirming that matrix stiffness can influence neural differentiation (DeQuach et al., 2011). Clinical trials for macular degeneration and other ocular disorders are being conducted using hiPSC-derived retinal cells as a cell therapy. Tissue engineering approaches enable long-term scaffolding and axonal growth in hiPSC-derived retinal cells, which may be amenable to transplantation (Calejo et al., 2020; Li et al., 2017; Worthington et al., 2017; Yang et al., 2017). Finally, complex skin analogs have been engineered using hiPSC-derived keratinocytes, endothelial cells, and fibroblasts (Abaci et al., 2016; Gledhill et al., 2015; Itoh et al., 2013; Kim and Ju, 2019; Kim et al., 2018; Petrova et al., 2014). The possibilities become endless for creating multi-lineage, hiPSC-derived platforms as bioprinters become more accessible, bioink perfected to be compatible with multiple cell types, and vascularization improved to enhance nutrient diffusion. It remains to be seen whether complex organs can be perfectly replicated using tissue engineering and bioprinting, but for the purpose of disease modeling, these technologies have substantially improved during the past decade. The decision between allowing natural development to do the work of organizing complex tissues or instead artificially engineering the correct orientation of cells remains an exciting and evolving area of scientific discussion. We feel that it is important to continue to advance both areas and evaluate models based on their predictive value to understand and treat human disease.

Organ and Body-on-a-Chip

Advances in biocompatible materials engineering combined with microfabrication technologies have enabled the development of microfluidic organ-chips (Bhatia and Ingber, 2014). Organ-chips are designed for the growth of multiple organ-specific cell types in a fully-integrated system, often with the different cell types being cultured in separate compartments that are interconnected via perforated membranes or juxtaposed microchannels. This approach allows for the co-culture of the multiple cell lineages that underlie an organ’s function and enables fluid flow, and in some cases mimicked stretch, to further simulate the in vivo environment. Since the body is not a static system, but rather perfused by flowing blood vessels, a lymphatic system, mucus flow across the lungs, and fluid flow through the gastric system, organ-chips advance in vitro models to the next level of physiology.

Importantly, these organ-chip systems must be devised from highly biocompatible materials that can sustain long-term culture of multiple cell types. Some studies have used silicon-based chips, comparable to those used in computer processors. Silicon chips have remained useful for high-throughput, microfluidic sorting of blood cells for CAR-T cell therapy (de Wijs et al., 2017). Another commonly used material is polydimethylsiloxane (PDMS), which is a silicon-based organic polymer. PDMS is preferred in organ-chip manufacturing because of its high biocompatibility, optical transparency for facilitating imaging applications, and ease of assembly. Drug permeability and absorbance associated with PDMS may be a concern, although this remains an area of debate since the process is highly drug- and tissue-dependent. Other materials are still frequently used for specialized applications, including polyurethane, glass, and various biopolymers, although the flexibility of PDMS still makes it one of the most attractive candidates in this field.

A major advantage of microfluidic co-culture systems is that they enable evaluation of hydrodynamic forces conferred by flow of the in vivo circulatory systems in a simplified, in vitro format (Table 1). This allows mechanistic interrogation of vessel-organ interfaces, as in the case of the blood-brain barrier (Appelt-Menzel et al., 2018; Vatine et al., 2019; Wang et al., 2017). Cells grown within microfluidic channels can be subjected to controllable shear stresses by altering the flow rate of media into and out of the chip, mimicking the circulatory system. This is particularly valuable for accurately studying the biology and maturation of endothelial and smooth muscle cells, which are constantly subjected to such biomechanical stresses by virtue of residing within blood vessels (Zanotelli et al., 2016). For example, hiPSC-derived smooth muscle cells from patients with progeria exhibited an inflammatory response to strain when cultured on a microfluidic chip (Ribas et al., 2017). Mimicked fluid flow and endothelial cell co-culture has also shown to improve maturation and function of endothelial and neuronal cell types grown within organ-chips (DeStefano et al., 2017; Park et al., 2019; Sances et al., 2018; Vatine et al., 2019). Fluid flow may enhance cellular alignment to mimic physiologically-relevant conditions or induce maturation. Mature cardiomyocytes harbor linearly aligned sarcomeres, a hallmark of adult cardiomyocytes that is typically absent in hiPSC-derived cardiomyocytes grown in two-dimensional culture (Mathur et al., 2015). Recent heart-on-a-chip studies have shown improved alignment of hiPSC-derived cardiovascular cells under flow and have also modeled cardiomyocyte functional abnormalities experienced during cardiomyopathies (Ellis et al., 2017; Wang et al., 2014a). In addition to creating sheer stress similar to the in vivo environment, blood or serum can be flowed through the blood channel of microfluidic channels to mimic a realistic physiological situation (Vatine et al., 2019). This allows the administration of drugs to the blood channel while observing the effects on the tissue channel, further recapitulating the in vivo environment and adding power to the model. Such drug validation have also been demonstrated independently in hiPSC-derived retina-on-a-chip and kidney-on-a-chip systems, which were able to reproduce the retinotoxic and nephrotoxic side-effects of select compounds, respectively (Achberger et al., 2019; Musah et al., 2017). Another enhancement would be the incorporation of pulsatile flow to mimic both the blood flow after a heartbeat and the constant periodic change in sheer force (Atchison et al., 2020; Atchison et al., 2017). Organ-on-a-chip systems can be integrated into standard hiPSC-derived cell analysis pipelines, conferring the possibility of evaluating patient-specific disease mechanisms or drug responses in multiple cell types simultaneously. Primary cells or immortalized cell lines can also be grown in conjunction with hiPSC-derivatives within organ-chips, depending on the research question or application. For example, microfluidic organ-chips harboring hiPSC-derived and primary brain microvascular endothelial cells, human astrocytes, and pericytes have recapitulated the barrier function of the blood-brain barrier (Brown et al., 2016; Park et al., 2019). Similarly, organ-chips containing hiPSC-derived gut organoids were more faithfully able to model human intestinal epithelial signaling responses than organ-chips containing Caco-2 immortalized cells (Workman et al., 2018). As recently demonstrated by intestine-, pancreas-, stomach-, and liver-chips, microfluidic organ-chip and hiPSC-derived organoid technologies can be easily coupled to enable a new generation of more physiologically-relevant cell culture (Lee et al., 2018a; Schepers et al., 2016; Sidar et al., 2019; Tao et al., 2019; Wang et al., 2018). Scalability still remains a concern for organ-chips, as these systems tend to sacrifice throughput for the sake of physiological complexity. However, if concerns related to chip miniaturization and fabrication cost are alleviated, it may be possible to achieve simplified and disposable organ-chips in high-throughput screening pipelines. Microfluidics-enabled, automated systems for media exchange, drug delivery, and collection of the media from the organ-chips will remove manual manipulation as a requirement and, therefore, reduce variability due to human error.

Perhaps the most exciting culmination of this technology is the body-on-a-chip concept, whereby chips derived from different bodily tissues are connected by a mimicked vascular system. Multiple organ-chips can be interconnected by microfluidics or supernatant media transfer to achieve an analogous body-chip platform (Esch et al., 2011; Ramme et al., 2019; Wikswo et al., 2013). Interconnected chips can enable transfer of metabolites, signaling intermediates, and growth factors from compartments harboring distinct cell types, establishing a more accurate model of how different bodily tissues influence each other in health and disease. Recently, up to 10 independent microphysiological systems mimicking different tissue types have connected in this manner to develop a “physiome-on-a-chip” (Edington et al., 2018; Novak et al., 2020). Body-chip systems also are particularly relevant to study drug toxicity, whereby a drug can be metabolized by hepatocytes in the liver-mimicking compartment of the body-chip (Esch et al., 2011). The metabolized drug can then be introduced to other organ-mimicking compartments on the same chip, establishing a more realistic model of tissue-specific drug toxicity and efficacy. Indeed, fully-integrated body-chips may prove useful for accurate preclinical identification of types of drug toxicity, including cardiotoxicity, hepatotoxicity, and neurotoxicity (Ronaldson-Bouchard and Vunjak-Novakovic, 2018).

The future remains bright for organ-chip technology. Ultimately, long-term success will depend on their integration into preclinical drug discovery and toxicity analysis pipelines for the pharmaceutical industry, as well as using these systems to obtain drug approval from the United States Food and Drug Administration (FDA). Validation of the methodology will require that predictions made in the organ-chip model have direct clinical relevance in human subjects. If successful, hiPSC-based organ-chips have the potential to significantly reduce the need for animals in medical research and increase the efficacy and safety of new medicines. Eventually this may even lead to patient-specific treatments where individual subjects could have their own hiPSC-based organ-chip system for testing drugs prior to use. While this is currently far from reality, the technology is moving at a very fast pace. Both the United States FDA and National Institutes of Health have major initiatives in place to accelerate the integration of organ-chips into drug testing pipelines (Tagle, 2019).

hiPSC Animal Models and Chimeras

In vivo multi-cell type model systems include animal models hybridized with hiPSC-derived tissues. Animal models have the advantage of being living biological systems in which biology and pathophysiology can be interrogated. Mammalian models such as rodents and non-human primates have long served in parallel with human in vitro systems to validate human disease-relevant mechanistic studies. Some of these models have been developed in direct response to regulatory requirements for cellular therapies. Preclinical animal studies are required by the FDA to examine the safety and efficacy of hiPSC-derived cell therapy products prior to use in patients. Indeed, hiPSC-derivatives have routinely been introduced into non-human models, including rodents, pigs, and primates, to investigate mechanisms and treatments for diseases of the heart (Chong et al., 2014; Kadota et al., 2017; Ye et al., 2014), nervous system (Chen et al., 2016; Xu et al., 2019), eye (Carr et al., 2009; Kamao et al., 2014; Kanemura et al., 2014; Zhao et al., 2015) and ear (Takeda et al., 2018). In particular, mice with inhibited immune systems are commonly used as recipients of human donor tissues to test the engraftment of hiPSC-derived cell types including muscle, neural tissues, and pancreatic cells (Funakoshi et al., 2016; Jeon et al., 2012; Kroon et al., 2008; Nori et al., 2011). Animal models can determine whether hiPSC-derivatives have appropriate in vivo functionality, such as the ability of beta cells to secrete insulin under glucose stimulation (Haller et al., 2019; Ma et al., 2018), hematopoietic stem cells to differentiate into blood lineages (Espinoza et al., 2018; Tan et al., 2018) and hepatocytes to metabolize drugs and generate albumin (Liu et al., 2011; Nagamoto et al., 2016; Pettinato et al., 2019). Human iPSC-derived neural tissues have been shown to integrate into rodent central nervous system tissues (Sareen et al., 2014) and in some cases have positive effects in models of neurodegenerative and eye diseases (Nori et al., 2011; Wang et al., 2013). Vascularized hiPSC-derived human brain organoids can merge with mouse host vasculature (Cakir et al., 2019). In a similar fashion, hiPSC-derived tissues have been introduced into non-human primates to study neurodegenerative and musculodegenerative disorders (Sundberg et al., 2013). Pig models have more commonly been used in cardiac cell transplantation studies, due to physiological similarities between human and porcine hearts (Templin et al., 2012). However, long-term functional integration of hiPSC-derived cardiomyocytes and other tissues have been challenging due to the established immaturity of hiPSC-derived cell types. In the case of the heart, human stem cell-derived cardiomyocytes have been able to remuscularize the myocardium after an ischemic event, but in some circumstances, have caused arrhythmias (Liu et al., 2018). Recent work has shown, however, that simultaneously providing epicardial support tissues may enhance remuscularization and integration of hiPSC-derived cardiomyocytes into the heart (Bargehr et al., 2019). Given the propensity of hiPSCs to differentiate into nearly all somatic cell types, other unique therapeutic applications have also been validated in animal models, such as using hiPSC-derivatives for hair growth (Liu et al., 2019; Yang et al., 2014b) and bone regeneration (Phillips et al., 2014; Sheyn et al., 2016). Humanized animal models have been staples of biomedical research for decades and their utility will likely continue to complement in vitro systems.

Large animal models remain critical to evaluating the safety and efficacy of hiPSC-derived cell types used for cell therapy, but use of animal models in this context also requires unique bioethical considerations, in particular for neural and reproductive applications. Human-animal neural chimeras remain an intriguing model system for studying neurological disease and development, but have been heavily regulated and banned in some instances (Sharma et al., 2015). For decades, scientists have considered whether transplanted human brain tissue would confer enhanced cognitive function to other species (Han et al., 2013), although existing data to evaluate this possibility is limited (Mariani et al., 2019; Windrem et al., 2017). Still, animal models engrafted with primary or hiPSC-derived human brain tissue have readily served as in vivo models for studying neurodevelopment or neurodegeneration. Human brain extracts have been injected into the mouse brain for studying tauopathies (Clavaguera et al., 2013) and hiPSC-derived oligodendrocytes have been able to restore myelination in mice with injured spinal cords or congenital hypomyelination (Wang et al., 2013). Similarly, hiPSC-derived glial cells have been injected into the cortex of mice to study the mechanisms of schizophrenia (Windrem et al., 2017). Developmental hiPSC-animal chimeras remain much rarer, although some countries such as Japan have recently eased restrictions on this type of work. An ultimate goal for human-animal developmental chimeras could be the creation of large animals, such as pigs, harboring human organs that could be used for transplantation (Figure 3). The global shortage for human organs is a well-recognized crisis, which could be alleviated by the generation of chimeric pigs containing select human organs such as heart and lungs. To accomplish this, hiPSCs could be introduced at an early developmental timepoint in an animal embryo knocked out for a critical developmental regulator of an organ of interest. This way, the hiPSCs would be able to complement the animal embryo, restoring the developmental potential only in a single organ of interest, and ideally without contributing to neural or reproductive tissues. In reality, hiPSC-animal blastocyst complementation remains extremely technically challenging, although some preliminary studies have been successful. For example, hiPSCs can contribute to early stage pig embryos, suggesting that the human-porcine chimera for human organ generation may be in the realm of possibility (Wu et al., 2017). In addition, mouse blastocysts complemented with hiPSCs are able to generate chimeric human-mouse early stage embryos (Mascetti and Pedersen, 2016). Similarly, mice harboring a rat pancreas have been generated by introducing rat PSCs into a mouse blastocyst knocked out for a key pancreatic developmental regulator, PDX1 (Kobayashi et al., 2010). Importantly, no human chimeric animals have been carried to term due to ethical and governmental restrictions, although this may change as the technology becomes refined. These studies demonstrate the power of hiPSCs to contribute to chimeric development, both for modeling human disease and potentially for translational purposes in organ generation. At the same time, they highlight the challenges of performing ethically acceptable experiments and carefully weighing the risks and benefits of animal chimera studies to advance human therapies.

Figure 3: Assays for interrogating cells within hiPSC-derived multi-lineage model systems and potential downstream applications.

A variety of technical readouts can be considered when evaluating the properties of multiple cell types within an hiPSC-derived multi-lineage model system. Analyses are widely applicable across multi-lineage model systems, in particular those analyses that examine functional, visual, electrophysiological, and –omics level properties of cells. Ideal assays are able to interrogate multiple cell types simultaneously, such as in the case of single-cell transcriptomic, epigenomic, proteomic, and metabolomic analyses. Downstream applications for multi-lineage hiPSC-derived systems include patient-specific disease modeling, personalized drug response screening, and human tissue production.

Readouts to enhance the fidelity of hiPSC models

Assays for evaluating cellular function in multi-lineage, hiPSC-based model systems exhibit significant overlap with those used for evaluating cells in traditional monoculture (Figure 3). Cell viability assays, which could have bioluminescence, fluorescence, or metabolic readouts, can be used to examine cell survival across multiple cell types (Figure 3). However, some multi-lineage platforms may be more amenable to visual evaluation of cellular viability than others. In rodent models containing human cells, modified bioluminescence or fluorescence imaging can be used examine cell engraftment and viability (Jung et al., 2018; Shafa et al., 2018), although zebrafish, given their transparent nature, are more naturally suited for in vivo imaging (Orlova et al., 2014). Organoids are known to vary in cell viability, due to potential necrosis in the center of a spherical organoid depending on nutrient diffusability. This necrotic core may confound results if viability assays are conducted on intact organoids. Indeed, tissue necrosis due to incomplete nutrient diffusability into organoid interiors is one reason why organoids are unable to achieve a diameter above a few millimeters, although bioengineering of vasculature and introduction of artificial microfluidic systems can enhance tissue viability in the organoid interior (Cakir et al., 2019). Similarly, real-time imaging of organoid interiors may be difficult due to tissue thickness, although fluorescence or bioluminescent reporters improve imaging possibilities. Depending on the biomaterial used, cells grown on organ-chips may or may not be amenable to imaging at high resolution. Thus, assays that quantify cell proliferation, for example using imaging-based approaches, may be influenced by limited imaging resolution. To facilitate imaging, custom fluorescent reporter hiPSC lines can be generated using CRISPR/Cas9 genome editing, and can be tailored to cell type-specific expression of fluorescent markers (Sharma et al., 2018). In this manner, transgenic, stably-integrated reporters can be developed to provide cell fluorescence at different wavelengths depending on the differentiated cell types (Kim et al., 2019). This can be useful for live imaging of cell processes in hiPSC-derivatives, as opposed to immunofluorescence as a terminal assay. Functional readouts such as cardiomyocyte contractility or neuronal excitation can also be used in conjunction with fluorescent reporter hiPSC lines (Figure 3) (Toepfer et al., 2019).

Innovative approaches to other functional readouts have also been developed with an emphasis on real-time monitoring, which offers significant advantages due to minimal invasiveness. Real-time monitoring enables long-term analyses of multi-lineage models over the course of extended differentiation or drug treatment. Organoids harboring integrated multi-electrode arrays enable a real-time readout of electrophysiological function during organoid differentiation (Figure 3) (Kalmykov et al., 2019; Zhang et al., 2017). Similarly, organ-chips with integrated electrophysiological assessment assays enable real-time evaluation of action potentials, particularly important for functionally active cell types such as cardiomyocytes and neurons (Maoz et al., 2017). Recently, metabolomics of secreted extracellular molecules has become a staple assay in multi-lineage systems whereby multiple cell types are connected by an integrated nutrient-supplying system (Figure 3) (McAleer et al., 2019). In organ-chips, which typically harbor only a small amount of cellular material, evaluation of secreted growth factors, bioactivated small molecule drugs, and other metabolites in media can be extremely useful as an alternative approach. For example, organ-chips can model the barrier permeability of metabolites analogous to the blood-brain barrier, and have been able to identify previously unknown metabolic coupling between the blood-brain barrier and neurons (Maoz et al., 2018). Thus, multi-lineage platforms vary in their amenability to downstream cellular readouts, but innovative assays and approaches are constantly being developed and optimized for use with each system.

Computational Analyses for Interrogating Multi-Lineage Models

Over the past two decades, the increased accessibility of next-generation sequencing technologies, and their reduced cost, has been paralleled by dramatic advancement in computational power. Improvements in sequencing and computational power have provided the foundation for the field of bioinformatics, which intersects the two technologies. But perhaps the key advance related to simultaneously evaluating different cell types within hiPSC-derived multi-lineage systems has been the emergence of single-cell transcriptomic analysis. This revolutionary technology has spurred the creation of global consortia to discover and characterize all of the cell types found within individual tissues, organs, and eventually, the entire human body (Regev et al., 2017; Thomsen et al., 2016). “Cell Atlas” projects using single-cell analysis are also useful in the study of developmental processes, such as the differentiation of pluripotent stem cells into multipotent progenitors and terminally-differentiated cells. For example, single-cell analysis is an excellent tool for evaluating the specification of the various lineages found in human blood (Villani et al., 2017). Single-cell transcriptomics can be coupled with any multi-lineage, hiPSC-derived platform, and is particularly useful in analyses of heterogeneous cell populations in organoids mimicking developmental processes (Kanton et al., 2019; Subramanian et al., 2019). Single-cell transcriptomic and epigenomic analyses allow for finer dissection of both the distribution of multiple cell types in a bulk population of hiPSC-derived cells and the defining characteristics of cells in that population. Even within a population of undifferentiated hiPSCs, single-cell transcriptional profiling has revealed heterogeneity in gene expression, something that would not have been detectable by bulk RNA-sequencing (Narsinh et al., 2011). In a multi-lineage co-culture system of hiPSC-derived neurons and astrocytes, single-cell RNA-seq has distinguished cell type-specific responses after treatment with anti-convulsant drugs (Ishii et al., 2017). Single-cell RNA-seq has also identified unique subpopulations in differentiating hiPSC-derived endothelial cells, lending further evidence to the heterogeneity of the hiPSC differentiation process (Paik et al., 2018). Using single-cell expression data from hiPSC differentiation in combination with bioinformatics analysis tools, cell lineages can be artificially recreated, providing a “roadmap” for cellular maturity or a key with which unknown cells can be identified (Gong et al., 2018).

Although transcriptomics is currently the most readily-used single-cell –omics level analysis, recent technical advances have enabled reliable single-cell epigenomic (Bravo Gonzalez-Blas et al., 2019) and proteomic analysis (Bravo Gonzalez-Blas et al., 2019). Data from –omics level, single-cell analyses of multiple cell lineages could be used to establish purely computational blueprints for development, differentiation, or drug response. One could envision a scenario whereby a perturbation of interest, such as a drug or genetic mutation, could be fed into an –omics level in silico dataset compiled from analyses of hiPSC-derivatives. Incorporating machine learning, an algorithm could be able to predict the biological downstream effect, such as disease, developmental abnormality, or drug response, of a perturbation of interest completely in silico. As tantalizing as such a prospect may be, a substantial number of standardized multi-omics datasets from multiple cell lineages, in combination with cellular functional data, will be needed to accurately inform such a predictive algorithm. Nevertheless, recent studies have shown that in silico analyses built upon hiPSC-cardiomyocyte electrophysiological datasets can evaluate and predict drug effects (Paci et al., 2018). Such in silico analyses will likely become more common as disease modeling and drug screening datasets from hiPSC-derived cell types continue to grow in number. Systems-level approaches with integrated analyses of multiple functional readouts will provide the best use of multi-lineage models.

Conclusions and Future Directions

An abundance of recent literature on multi-lineage hiPSC-based models such as organoids, assembloids, microfluidic chips and even animal chimeras demonstrates the excitement around these fields. It is also clear that the biotechnology industry is now utilizing hiPSC-based systems for effective drug production, based on a wealth of unpublished data at leading pharmaceutical companies worldwide. There is hope that these new technologies will more closely resemble tissues of the human body, thus becoming more accurate in both modeling human disease and testing drugs for efficacy and toxicity. During reprogramming, hiPSCs wipe the epigenetic slate clean and allow the researcher to repeatedly replay tissue development. If patient-specific disease phenotypes are found in hiPSC-derived cells, they are likely the result of genetic perturbations within the patient’s DNA (Laperle et al., 2020). Conversely, if the hiPSC model based on patient-specific cells has no observed phenotype, it suggests that there was likely no genetic contribution to the disease, but rather, environmental factors were involved. For most diseases, there will be a complex interaction between the genetics and the environment, potentially with disease initiation from the genetics and exacerbation from the environment. To address this question in models lacking a phenotype, the next step would be to introduce environmental factors to “stress” the system and examine if phenotypes emerge only within the patient cells.

The turning point for multi-lineage systems will be validation in human patients. Did the pathology in the model reflect the pathology in the human disease? Were the -omics changes similar to those seen in patients? Did the hiPSC model accurately predict a drug effect in the same patient? This last point will be invaluable for the future of precision medicine, potentially with a patient’s hiPSCs providing a personalized avatar for personalized disease modeling and drug screening. These validation studies must be performed in advanced models to establish whether they are simply interesting in vitro observations, or an important next step in understanding and treating human disease. This will require a precision health-oriented approach and the further development of “clinical trials in a dish” to predict outcomes in the same patient group. As a related first step, multiple high-profile studies have shown that patient-derived tumor organoids using primary cells can serve as a predictive platform to develop personalized therapies for pancreatic (Huang et al., 2015), colorectal (Vlachogiannis et al., 2018; Yao et al., 2019), prostate (Gao et al., 2014), and bladder cancers (Lee et al., 2018b). Likewise, hiPSC-derived cell types are perfectly poised to help develop personalized drug treatments, given their inherent patient-specificity. Combinations of multi-lineage modeling in vitro, in vivo, and even in silico will likely yield the most accurate, reproducible, and translatable results. Incorporation of other cutting-edge technologies such as genome editing, multi-omics and single-cell gene expression analysis, will enable greater flexibility in both establishing customized models and eliciting more accurate downstream analysis of target cell types. Genome editing in particular unlocks an array of possibilities for generating customized, multi-lineage hiPSC-derived cells for use in personalized medicine or even for synthetic biology approaches, whereby these cells can be engineered to produce useful materials or behave in unique ways.

Against this backdrop of enthusiasm for multi-lineage hiPSC-derived systems, significant questions clearly remain (Table 1). There are skeptics who argue that using these hiPSC-based systems will never replace traditional animal models. We believe that each disease has its own challenges, and all hiPSC-derived models require validation as described. Perhaps first and foremost among the prominent issues (Table 1) is cellular immaturity of hiPSC-derived cell types. Human iPSC-derived tissues are typically fetal-like in nature based on transcriptional and functional analyses, and further maturation often requires complex additions to the culture media or over-expression of aging genes (Ho et al., 2016; Miller et al., 2013). Recent studies have also utilized biophysical stimulation via electromechanical inputs to further mature and age functionally-active hiPSC-derived cell types such as cardiomyocytes (Ronaldson-Bouchard et al., 2019; Ruan et al., 2016). Gradually increasing electromechanical stimulation in hiPSC-derived cardiomyocytes dramatically improves their structural and functional properties. Another approach to maturation could be to transplant hiPSC-derived cell types in vivo so that these tissues could benefit from the systemic milieu and physiologically relevant setting. For example, human intestinal organoids transplanted into mice showed more adult-like phenotypes including increased expression of adult digestive enzymes (Finkbeiner et al., 2015; Watson et al., 2014; Workman et al., 2017). Other unique methods to induce hiPSC-derivative maturation have recently been illustrated, such as training human PSC-derived pancreatic islets according to a circadian rhythm (Alvarez-Dominguez et al., 2020). These studies suggest that it is possible to create more mature hiPSC-derivatives, although the field still has not seen a study demonstrating that an hiPSC-derived tissue is functionally and transcriptionally identical to its adult counterpart. If true adult maturity of hiPSC-derivatives is unattainable, and if a disease is only evident in aging patients, one could argue current hiPSC models will never have any value. However, if the late-onset disease has a genetic component, it could very well affect cells during development and affect other support cells during the lifetime of the patient, both leading to overt clinical phenotypes later in life, as we have recently shown for Huntington’s disease (Garcia et al., 2019; Mathkar et al., 2019). We believe that hiPSC models combined with the multi-lineage technologies described here may reveal very early changes and molecular targets in cells prior to their ultimate dysfunction and death. Indeed, we have discovered a very early molecular phenotype in hiPSC-derived dopaminergic neurons from patients with early onset Parkinson’s disease that can be reversed with phorbal ester compounds (Laperle et al., 2020). These types of studies lead to the future hope of prophylactic drug treatments given early in life to at-risk patients that can mechanistically correct the biological dysfunction and thus resolve the disease before onset. An analogy would be the use of statins that are given years before heart disease in order to lower cholesterol and avoid atherosclerosis. Thus, patient hiPSC-derived models, even with their current immaturity, may provide important clinical information.

For certain cells, such as primary neural and cardiac tissues, it remains exceptionally difficult to obtain and scale in a manner that would be useful for large-scale disease modeling and drug discovery projects. Scalability is not a concern for hiPSC-derived cell types, as hiPSCs can be mass produced in two-dimensional and bioreactor-based platforms and differentiated efficiently. However, the efficiency of directed hiPSC differentiation is still variable between donor lines and even individual experiments and typically leads to varying populations of multiple cell types that arise spontaneously, even in the most refined differentiation protocols (Sances et al., 2016). This remains a serious challenge, which may be helped by better sourcing of small molecules and proteins added to serum free cultures. One step to enhance hiPSC-derived systems is to eliminate extraneous lineages from mixed cultures either manually, metabolically, chemically, or genetically in favor of a single lineage. While simplifying the model and reducing variability, the drawback of this approach is the loss of multiple cell type interactions that may be required for normal tissue function, as emphasized in this review. Multi-lineage in vitro or in silico models, if validated in humans, may serve to reduce a reliance on animals for research purposes, an ethically and societally advantageous situation. But incorporation of these model systems into industrial drug discovery pipelines will certainly depend on the financial feasibility of such platforms. Will organoids, organ-chips, or other systems provide enough scientific utility and value to drug development to justify potentially significant reagent and biomanufacturing costs? These questions remain to be fully addressed, but major advances are on the horizon.

Ultimately, no single model can perfectly replicate the complexity of the human body. However, when used thoughtfully and with well-structured validation studies, multi-lineage model systems offer new tools that can be used in the battle against human disease. In summary, we believe that the multi-lineage, hiPSC-based model systems presented here represent the future of precision medicine.