Multi-cellular engineered living systems to assess reproductive toxicology.

The development of multi-cellular engineered living systems involves mimicking the in vivo structure and function of cells, differentiating cell types for MPS incorporation or organoid development, and having various cellular samples to study from a population are all still difficult to maintain [ 97 ]. These systems can recreate the same mechanical, chemical, and physiological cues of mammalian reproductive biology that are observed in vivo. Multi-cellular engineered living systems that focus on incorporating the models with human cell lines for further optimization offer the advantage of providing measurement of hormone signaling, cell proliferation, and matrix regulation in varying degrees. Additionally, to further advance this technology, there is a need for the manufacturing of inexpensive sensor systems to monitoring cell-to-cell and cell-to-protein interactions to better monitor the functional properties of the tissue. This is especially important as the cellular complexity of the models make high throughput assays difficult to conduct and large numbers of cellular samples are needed to thoroughly screen environmental pollutant and contaminant toxicity [ 98 ]. Engineered approaches need to be simplified and standardized to fully recapitulate reproductive organ and tissue culturing needs that mimic in vivo conditions, while also providing faster fabrication for higher throughput to meet the needs toxicological testing models and potentially explore therapeutic treatment pathways. The MPS and organoids are important methods to evaluate specific cell types affected by toxicants and identify the mechanism of toxicity. To date, most studies have been descriptive, but validation is needed in order for these models to be useful. The MPS systems need to reproduce key structure and function of the tissue. Fox toxicants, that also means establishing levels of the various enzymes and transporters involved in metabolism and relating concentrations at which toxicity occurs with in vivo and in vitro values. Specific comparisons may be challenging and a focus on pathways may be useful to understand cell type toxicity, as was done recently with cardiac toxicity [ 99 ]. Machine-learning algorithms have had increasing prevalence in pharmacological drug screening for predictive toxicology testing. For example, they have been formulated to detect the neurotoxicity of unknown compounds was developed from RNA-seq data of human pluripotent stem cell derived neural constructs exposed to known toxic compounds [ 100 ]. As micro-physiological systems and technologies combine to form synergistic multi-unit models for drug development and toxicology models, different types of data need to be monitored and analyzed. Artificial intelligence (AI)-computerized systems that can be paired with chip-embedded biosensors can help in this endeavor by simultaneously adjusting parameters to create a self-sustaining micro engineered system [ 101 ]. Biosensors can monitor flow, biomechanics, receptor signaling, pH, gas content, macromolecule distribution and electrophysiology to assess engineered tissue function and architecture [ 102 ]. Data derived from these models can then be further analyzed by computational algorithms for predictive toxicity. A physiologically-based pharmacokinetic model (PBMK) can be developed to represent an organ-on-chip (or body-on-chip) as a quantitative approach to measuring drug sensitivity and metabolism [ 103 ]. For such multiorgan or multi-tissue systems, a common media is needed to support all tissue types. This can be done by creating an endothelial barrier and using a common media similar to plasma which each tissue type in in equilibrium with media specific to that tissue [ 19 ]. Tissue sizes and tissue-specific flow rates must be scaled to reflect the organization and physiological ratios. The analysis needs to incorporate protein binding of drug or toxicant to reflect available free concentration. Clearance of the drug or toxin is obtained by incorporating kidney and liver MPS units and key organs are need for simulation of distribution. Finally, application of physiologically based pharmacokinetic models can be used to predict drug and toxin levels in specific tissues. Tatosian and Shuler incorporated PBMK analysis to a micro-engineered chip containing liver, bone marrow, cancerous tissue, and multi-drug resistant cancer tissue in four different, separated compartments as a way of comparing drug-dosage responses to doxorubicin and acute cytotoxicity [ 104 ]. Similar types of experimental methods and mathematical approaches can be explored to evaluate toxicant response in female and male reproductive tissues to compare healthy and disease conditions. The next stage of micro-physiological advancements involves the incorporation of novel technologies to address current limitations and improve in vitro organ system modeling. For instance, organoids can be placed into organ-on-chip platforms (called organoid-on-chips) as organoids alone still pose some complications since they do not demonstrate complete tissue properties and cellular composition can vary dramatically due to less defined 3D matrix components and self-differentiation processes. Current organoid-on-chip models have been instrumental for understanding numerous diseases and has made organoid culturing more feasible in biomedical studies. These systems are more effective at reproducing organoid differentiation and vascularization, and enable the possibility of examining the interactions of organoids for different systems [ 105 ]. Incorporating hydrogel and microfluidic methods into organoid-on-chip platforms can also help in standardizing platform design and attaining reproducible results. Rajan et al. recently developed used a microfluidic platform containing hyaluronic acid derived from ECM to study tissue interactions between six different organoid cultures of the liver, heart, endometrium [ 106 ], testes, brain, and lung as a potential model for testing drug toxicity in vitro [ 106 ]. By exposing liver organoid first to drug compound, metabolites released from the liver organoid were then able to affect other organoids. Although this model can potentially be clinically translatable as it is an inexpensive tool to create and use, this model is highly dependent on including a representative liver for analyzing downstream drug-induced affects that could be toxic. Organ-on-electronic-chips (organ-on-e-chip) that involve using electric 3D biosensors is another approach that can enhance development of MPS’s by measuring cellular electrophysiological communication. Kalmykov et al. [ 107 ] developed a chip to study electrochemical signals of cardiac organoids using passive microelectrode sensors and active metallic sensors. Not only can these sensors be incorporated to study other types of electrically active cells, but also activity of immune cells. This can be potentially utilized to explore reproductive MPS systems that incorporate immune cell factors. Ethical considerations must be considered when discussing the development and engineering of MPS’s. Incorporating patient stem cells into these systems for modeling disease and toxicology can have huge benefits, but also serious implications to consider. It is extremely important to establish and maintain culturing/differentiation standards to ensure the stability of engineered tissues [ 108 ]. If using or incorporating embryonic stem cells (ESCs), it is important to gage guidance and oversight over its use due its controversy and lack of consensus for its use [ 109 ]. Overall, although there are still challenges posed into creating MPS’s for reproductive biological purposes, and development of these systems is still very much novel and ongoing, there is much potential for their inclusion in future toxicological studies because of insights they can provide for understanding toxicity-induced pathways and developing therapies to target these effects. As environmental chemicals and contaminants continue to prevail, reproductive disorders and pathologies related to infertility and premature birth are on the rise. This is a wide-spread health issue that impacts millions of women across the globe and can continue to impact human health and longevity for decades.

Several drugs as well as natural toxins and human-made toxicants adversely affect the structure, function, and fertility of male and female reproductive systems, as well as offspring development. Specific changes include altered reproductive hormone production and receptor signaling by toxicants that affect gamete maturation and release [ 1 ]. In the male reproductive system, toxicants can decrease sperm count and motility by changing development, function, and death of germ/sperm and somatic/Sertoli cells or heightening inflammation and genotoxicity in the testes [ 2 ]. Exposure to drugs or toxicants during fetal development can lead to birth defects and neurodevelopmental disorders in newborns and young children. While drugs undergo clinical trials to assess safety and efficacy, the impact on the fetus is often not known at the time of drug approval. Further, there are about 150,000 synthetic chemicals in use and only a small fraction of these compounds has been tested for toxicity [ 3 ]. Tox21 was developed to address this need, as well limitations of epidemiological and animal studies. Tox 21 is a concerted effort to develop human models that are rapid, cost effective and faster [ 3 ]. Further, these models must be able to provide mechanistic information that can assist in understanding the critical pathways by which classes of compounds disrupt organ structure and function. Given ethical issues of assessing toxicity directly in humans, toxicity is assessed currently through a combination of epidemiological studies, animal experiments, human tissue samples, in vitro experiments and computational. Epidemiological studies provide information on the statistical relationship between toxicant or drug and specific functional or pathological changes. These studies do not provide biological or biochemical mechanism behind the pathology caused by exposure to the toxicant. Often the individual exposure is not known, and the studies are not controlled, so epidemiological studies are only a first step to identify the potential toxicity of a compound. Animal models provide information on the response of the entire organism and the processes administration (i.e. routes of entry), distribution throughout the animal, metabolism, and excretion (collectively known as ADME) can be assessed. Further, well controlled experiments can be performed. However, significant differences exist between animals and humans in physiology and specific biochemical and cellular responses in many organs, including the reproductive organs and placenta [ 4 ]. For example, only primates, such as gibbons, orangutans, gorillas, chimpanzees, and only a one rodent species menstruates [ 1 ]. A number of drugs successfully used to treat disease in animal models fail in clinical trials for safety reasons [ 5 ], likely to due to differences in drug metabolism between the animal model and humans and inaccuracies in extrapolating between species. These studies are expensive and lengthy, limiting throughput in testing large numbers of chemical over a range of doses. Alternative animal systems or assays that have been recently incorporated into reproductive teratogenic and toxicity studies are the Frog Embryo Teratogenesis Assay (FETA), zebrafish larvae and embryos, mouse embryonic stem cells (ESCs), and rodent whole embryo culture (WEC) [ 6 ]. Advantages of such models include the time scale to establish the necessary responses, each of assessing outputs, and ability to perform higher through screens. Such models are suitable if they provide accurate information that can be related to existing animal models and the results are relevant to human reproductive biology, especially for embryo-fetal malformations caused by toxicants. Ex vivo models using human clinical specimens are regularly available for the placenta but are difficult to developed for reproductive organs, due to difficulties in obtaining tissue samples. New alternative approaches, like the ex vivo placental perfusion model can provide information on the transplacental permeability of toxicants and some mechanistic information about the transport process. To date, this model has been used in a limited number of studies [ 7 ], potentially due to the complexity of establishing the circuit and maintaining placental viability. While such a model is not suited for high or moderate throughput, it can be used to confirm animal or higher throughput studies with cultured human cells. Ex vivo placental explant models preserves all cells for up to 28 days and can be used to evaluate toxic effects of drugs for long periods of time [ 8 ]. However, it is still not fully understood if ex-vivo placentas and explants can fully model the mechanical and dynamic microenvironment needed for placental function [ 3 ]. Traditional two-dimensional (2D) cell culture has been used to identify biochemical and biophysical pathways. In many cases, these 2D systems do not recapitulate the same bioactivities and cues found in vivo. While techniques like cell-adhesive islands, coating with hydrogels, microwells, and micropillars have been developed to mitigate these issues, these methods often induce unnatural apical-basal polarity that changes cellular architecture [ 9 ]. Trans well co-culture systems have provided opportunities to explore cell-to-cell and extracellular matrix interactions that occur in barrier tissues important in the placenta and male and female reproductive tracts, as well as pregnancy and embryo implantation. However, Transwell systems do not allow for biomechanical control of cellular or tissue properties. Other limitations of 2D cell culture systems are that they do not replicate tissue architecture, provide limited insight on cell-to-tissue and cell-to-matrix interactions, and cells often completely or partially dedifferentiate in culture. Multi-Cellular Engineered Living Systems [ 10 ], also referred to as organ constructs [ 11 ], are three-dimensional in vitro models of the functional units of organs. These systems consist of multiple cell types and represent the structure and function of the tissue being modeled. Multi-cellular engineered living systems can be produced by spontaneous organization of the cells in a tissue to form organoids, or by a more-top-down engineered approach, known as microphysiological systems (MPS), in which the tissue organization is established by use of natural or synthetic extracellular matrix and specific placement of cells. In some cases, a combination of both methods has been used. Organoids arise when embryonic or induced pluripotent stem cells are cultured in a defined three-dimensional (3D) geometry that upon addition of specific molecules (morphogens) spontaneously initiate differentiation to a specific tissue lineage [ 12 ]. Encapsulating pluripotent cells inside an extracellular matrix may recapitulate the in vivo microenvironment needed for proper differentiation. Key features of any organoid are that they form by self-assembly, express the cell types found in the tissue, and mimic the in vivo tissue architecture [ 11 , 13 ]. Although organoids demonstrate tissue heterogeneity and cellular complexity, their limitations include a lack of vasculature and uncontrollable morphogenesis during self-assembly [ 11 , 12 ]. Organization of the developing structure to reflect better the in vivo organ can be addressed by better control of the morphogen concentration, geometry and biomechanical environment [ 10 ]. MPS, often referred to as organ-on-chips or tissue engineered cellular/organ constructs, are made with different cell sources such as primary cells, immortalized cell lines, or cells derived from induced pluripotent stem cells (iPSCs) or embryonic stem cells. The constructs allow for the growth of heterogeneous cells in 3D extracellular matrices that recapitulate realistic stiffness, mechanical, electrical, and chemical signals, and perfusion of tissues [ 14 ]. A common format appropriate for barrier tissues in the lung, microcirculation, GI tract, and some reproductive tissues is a microfluidic chip with two or three channels in which organ-specific cells and endothelial cells form monolayers on opposite sides of a porous elastic membrane or extracellular matrix hydrogel ( Figure 1 ). Recreating the physical microenvironment of the tissue is affected by media flow and mechanical forces applied to the tissue [ 14 ]. MPS are more complex to establish than 2D cell culture, limiting throughput. Since multi-cellular engineered living systems can more accurately model the tissue organization and function than do 2D models, they have been utilized to evaluate drug toxicity [ 15 ] and the impact of xenobiotics and environmental pollutants [ 16 ]. Multiorgan MPS to model human ADME are assembled by interconnection of specific organ in a manner consistent with normal physiology [ 17 – 19 ]. Pumpless systems reduce fluid volumes introduced by pumps and tubing for interconnections and enable higher throughout [ 20 ]. Pumpless systems may pose challenges in maintaining specific flow rates and shear stresses or regulating the temporal flow waveform of the perfusion media [ 21 ]. To test efficacy, diseases can be modeled by differentiating iPSCs derived from individuals with genetic diseases or by recreating disease conditions [ 22 , 23 ]. Multi-cellular engineered living systems have the potential to provide important information on reproductive toxicology. To serve as new approach methodologies (NAMs), the specific context of use of the MCELS needs to be clearly defined, the approach validated using standardized procedures, and the strengths and limitations identified. The OECD guidance document 34 is an important starting point [ 24 ]. Experience with NAMS indicates that development, validation and adoption of NAMs requires collaboration among developers, users, and regulatory agencies [ 25 ]. In general, MCELS are best suited for screening a limited number of compounds and providing mechanistic information, since they more accurately replicate the in vivo environment than simpler systems. For the most part, MCELS for reproductive toxicology are in an early stage of development and few have been validated. We review existing multi-cellular engineered living systems for the male and female reproductive systems and during embryonic and fetal development, provide specific examples in which they are used to evaluate toxicity, and discuss needed advances so that current and potential MPS systems can enhance our understanding of the effects of toxins and drugs on the reproductive system and become validated NAMs.

Placental Development Due to the number of synthetic molecules in the environment and the high frequency at which pregnant women take medications for pre-existing or fetal conditions [ 79 , 80 ], the developing fetus may be exposed to doses of drugs or toxicants that can adversely affect development. For drugs to enter the fetal circulation, they must mass through the placenta, which consists of villous structures through which the fetal microcirculation passes ( Figures 6A , B ). The syncytiotrophoblasts form a multinucleated, terminally differentiated syncytium in contact with maternal blood. The apical surface contains microvilli which facilitate gas and nutrient exchange. These cells also secrete hormones critical for maintaining pregnancy. Beneath the syncytial layer of trophoblasts are proliferative cytotrophoblasts that can replenish the syncytial layer and are attached to the basement membrane [ 81 ]. Between the capillaries that feed into the umbilical vessels and the cytotrophoblasts is a basement membrane to which endothelial cells attach [ 82 ]. The Pgp efflux transporter which can remove drugs and toxins from cells, is present in the apical microvilli of syncytial trophoblasts. By limiting fetal exposure to xenobiotics and aiding in maternal-to-fetal transfer of nutrients and waste, Pgp serves to tightly regulate and protect fetus from damage and maintain circulation of vital molecules [ 83 ]. A number of other transporters regulate organic molecule transport across the placental barrier ( Figure 6C ) [ 84 ]. Multi-Cellular Engineered Living System Models of the Placenta Organ-on-chip and microfluidic technology has been used to model fertilization, implantation, and placentation ( Table 1 ). A recently established oviduct-on-a-chip platform using bovine oviduct epithelial cells supports in vitro fertilization [ 86 ]. It also allowed for live imaging of sperm and oocyte and oocyte-sperm binding to form embryos. While this highlights the first bioprinting approach to recreate the bovine oviduct, further advances in utilizing human oviduct epithelial cells and adding a mixture of cells with and without cilia can help to further advance the clinical translatability of this platform. Park et al. micro-engineered an MPS chip platform that recreates the invasion of trophoblasts into the maternal uterine arteries for studying implantation [ 87 ]. They attempted to recreate the physiological interface between mother and fetus by using a microfluidic platform embedded with ECM hydrogels to create microvasculature between uterus, trophoblasts, and embryo. Although this study represents a significant advance towards creating an in vitro implantation-on-chip, the authors recognized that the structural and molecular biology may not be fully represented without incorporating smooth muscle cells and decidual macrophages that make up the maternal microvasculature. Additionally, this model does not include hormonal signaling molecules; which play a significant role in implantation. Nonetheless, the development of this chip serves as an important starting point to better understand developmental processes involved in early pregnancy. Several placental barrier models have been developed using a chip format shown in Figure 1A . Using BeWo cells, a placental epithelial cell line from an individual with choriocarcinoma, and human placental villous endothelial cells, Blundell et al. [ 88 ] fabricated a microfluidic device to replicate placental barrier structure and physiology. After treatment with forskolin which elevates cyclic AMP, BeWo cells fused to form epithelium similar to syncytiotrophoblasts. Flow induced microvilli formation. GLUT1 was localized to the apical surface and the glucose transfer rate from maternal to fetal tissues was similar to in vivo values and higher than what was obtained with Transwell systems. Using the b30 clone of BeWo, which more accurately resembles villous trophoblasts, these investigators obtained transepithelial electrical resistance (TEER) values of ~5020 Ω cm 2 which indicates the higher placental barrier integrity measured for an in vitro placental model ( Figure 6D – H ) [ 85 ]. They demonstrated glyburide drug transport across the placenta-on-chip that demonstrated enhanced trophoblast efflux transport activation that serves to limit maternal drug exposure to the fetus in vivo. While this provides context that would be beneficial for toxicity studies, this platform is limited to only studying placental barrier physiology and function during late stages of pregnancy. Recently, MPS platforms have been used for fetal and placental toxicological studies. An organ-on-chip using amnion epithelial and mesenchymal cell lines was used to recreate the fetal amnionic membrane that serves a physiological barrier between mother and fetus [ 89 ]. They were able to study cellular communication and migration between healthy, oxidatively stressed cells (by cigarette smoke extract exposure), and antioxidant-treated cells. This system was able to replicate uterine oxidative stress that inhibit cell migration and increases inflammation; and show that antioxidant, N-acetyl-L-cysteine can prevent the occurrence of these effects. Hence, this chip system was used in a later study to investigate the cellular and molecular mechanisms behind cadmium exposure [ 90 ]. Including human maternal decidual cells and fetal chorion cells to further recapitulate the maternal-fetal lining into the chip, Kim et al. [ 91 ] analyzed how cadmium impacted apoptosis and necrosis. Decidual cells were more affected by inflammatory markers and cell death compared to the other cells. The results suggest that cadmium-induced fetal preterminal labor is a result of maternal damage to the chemical. Using a multichannel microfluidic chip, this group examined the effect of statins on cytotrophoblasts (BeWo cells), syncytiotrophoblasts (BeWo cells treated with forskolin) and human umbilical vein endothelial cells [ 92 ]. They observed that cells could transport and metabolize the statins and that the statins had anti-inflammatory effects in the chip. Although the cells were not in their normal anatomical arrangement, this system allowed assessment of the contribution of each cell type. These studies represent an important example about how MPS systems can help both understand and predict toxicant effects. Placental organoid technology can also be of aid in elucidating biochemical pathways in placental function and dysfunction. One such organoid was created using human cytotrophoblasts (CTB) during first-trimester pregnancy and expressed native placental tissue transcription factors that promote CTB fusion and differentiation [ 93 ]. As = trophoblast developmental differentiation and self-renewal in the placenta is still being widely investigated, this organoid culture can provide a model to study factors that induce or prohibit this process. A different trophoblast organoid capable of differentiating into syncytiotrophoblasts or extra-villous trophoblasts was also developed as another placenta model [ 94 ]. These cultures are representative of in vivo placental tissues because they contain elements are “villous-like” and they secreted peptides and hormones like primary trophoblasts. As sufficient amounts of primary cytotrophoblast cells for placental study are limited in availability and they can be difficult to maintain over long periods of time, organoids offer the opportunity to better study placental development and physiology. However, seeding number and timing need to be further optimized to aid in control of organoid self-differentiation [ 95 ]. Recently, a commercially available multichannel chip, Organoplate ® was used to establish a placental barrier model ( Figure 7A , B ) [ 96 ]. Endothelial cells and trophoblasts were cultured around channels that were separated by a collagen I/IV ECM. A Human iPSC cell line, ChiPS4, was differentiated to a mixed population of cytotrophoblasts, syncytiotrophoblasts, and extravillous trophoblasts. When grown on the chip, syncytium formation was demonstrated by E-cadherin staining and reduced permeability to 155 kDa dextran, although the barrier was not complete. β-hCG production was also increased. Gene expression studies increased expression of several ABC transporters and solute carriers for glucose, amino acids, fatty acids, metals, and ions. While further validation of the chip and iPSC-derived cells is needed, this system does appear suited to test toxicity.