Stem cell derivatives, organoids, and microphysiological systems to model normal and abnormal tissue physiology.

Stem cells and their differentiated progeny can be used to model tissue physiology, but more complex in vitro models are needed to recapitulate higher- level anatomical and physiological or pathological aspects of human biology. Organoid and organ-on-a-chip technologies (also referred to as microphysiological systems) are rapidly advancing platforms for such complex in vitro models. These models, from 2D to printed technologies, represent different aspects of human organs and tissues, and promise to reproduce human physiology that resembles the human situation well enough for predictive testing of interventions. Guidelines for best practice are needed to realize opportunities and address limitations of the models. These technologies aim towards the same overall goal of being physiologically reliable, albeit simplified, tissue representations.

Crucial to ensuring that these human model systems are widely adopted by academia and industry is confirming their reproducibility between developers and end-users, and individual laboratories and operators. The goal of this section is to improve the utility of model systems in fundamental research by:

• Improving the rigor and interpretability of model systems.

• Improving their reproducibility by reducing variability in their derivation, composition and use.

• Assessing the quality and validity of model systems, including their ability to recapitulate human (patho)physiology.

We provide comprehensive indications of which aspects of the models need particular attention to achieve these goals and ensure that they can be used optimally to understand and advance human health.

Understanding Your Starting Material

Recommendation 4 .1 .1: Consider the cell line or tissue of origin and, if known, identify the cell type of the starting material for the model.

There are several important points to consider when sourcing material for the generation of stem cell-based models as in vitro platforms to study normal physiology or disease pathology.

The starting material can influence variability and reproducibility. Firstly, the starting cell type(s) (e.g., fibroblast, epithelial cell) should be considered and documented. Secondly, whether the cells are obtained from fetal or adult tissues or banked material should be noted. Thirdly, consider the site that the sample was derived from, describing the anatomical location as well as possible. Fourthly, the isolation procedure should be carefully considered as this can lead to enrichment of specific cell and different types. Given the inherent nature of clinically derived material, it is essential that the origin is consistently described. The tissue or cell of origin of the starting material should be characterized as early as possible. Lastly, researchers should carefully consider the culture conditions since differences, even if small, can change the phenotype of the cell of origin or lead to selective growth of different cell populations.

Recommendation 4 .1 .2: Consider the sex, age, ethnic and genetic background, health status, risk factors and any additional clinical signs or symptoms of the donor, where available and as permitted by local regulations.

The background of the donor cells for derivation of model systems can influence the outcome of these models and may affect the generalizability of findings. Thus, details of the donor background should be documented even in the case of donors with no known relevant diseases. Laboratories should aspire to collect as much metadata as possible to assess how broadly applicable the findings from these model systems are. Include sex, age, health status and histopathological analysis of the starting material where available. Whenever possible, the donor’s information should also be expanded to include self-reported lifestyle risk factors (e.g., smoking, diet, and exercise), known infections, prior treatments and/or disorders, and family history/genetic predisposition. For disease modeling, disease status and health of the donor should be provided, including genetic mutation and associated clinical data. Consider best practices in data management and donor privacy in your jurisdiction, noting that this may mean that not all information collected will be publicly available at the end of the study.

Generating the Model System

Recommendation 4 .2 .1: Quality control metrics of the method and the intended model should be established, fully documented, and validated across different stem cells and donors.

Researchers should use and establish, where not already available, metrics to assess the quality of components used to generate the model, and basic quality control metrics of the model system itself. Components and reagents for the development and maintenance of the model system should be tested, either by the manufacturer or the experimenter, for key metrics relevant to the model system.

A model system based on an engineered device should ideally be produced from ready-to-use devices and components. If this is not possible, the devices should be manufactured using processes that are widely available at academic institutions or via companies. Detailed protocols should include methods of device manufacture, companion reagents and their source. Descriptions of the fabrication of the devices should indicate potential problems and provide troubleshooting advice. Each step should be specified so that the fabrication processes can be reproduced. The success rate in producing the final model system should be stated so as not to create false expectations.

Recommendation 4 .2 .2: Ensure reproducibility within and between laboratories by describing the operational microenvironment and identifying conditions that affect variability for the given model system.

Methods of measurement need to be described and repeated to document any technical variability. The conditions that can affect variability, such as cell seeding density, culture reagents, fluid flow rate in microfluidic devices, oxygen tension, exogenous extracellular matrix components, frequency of media changes, and media batches, should be elaborated, and details of quality control metrics should be recorded. Sufficient replicates should be performed to assess technical and biological variability. For example, independent experiments are needed to assess intra-batch, batch-to-batch and line-to-line variability.

Validating Stem Cell-Derived Models

Recommendation 4 .3 .1: Demonstrate that the cellular model is functionally and phenotypically representative of the native cell/tissue by multiple, appropriate criteria.

The successful application of stem cells for modelling requires verification of cell specialization to target cells that recapitulate native cellular phenotypes. This is achieved by systematically evaluating differentiation to lineage- specific cell and tissue morphology, function, and expression of cellular markers (such as cell surface antigens and RNA transcripts). Importantly, in addition to being dependent on the genetic makeup of the donor organism, cellular phenotype is influenced by the surroundings to which the cells are subjected, including various epigenetic processes. As such, recurrent assessments during the development (i.e., at different developmental timepoints) and maintenance of cellular models are necessary.

Different levels of complexity in the model system may be needed to represent different aspects of the physiological system. There are four basic tissue- and component cell-types derived from stem cells for modelling, which are defined by their morphology, function, and cellular markers: epithelium, connective tissue, muscle and nervous tissue/cells. Epithelial cells and tissues form organ boundaries and are involved in protection, secretion, absorption, excretion, filtration, diffusion, and sensory reception. Connective tissues and cells (including cartilage, adipose, bone, and blood/lymph vessels) support and provide structure to other tissue types and help transfer nutrients and other substances between tissues and organs, repair damaged tissue, and defend the body against infection and disease. Muscle tissue (skeletal, cardiac, or smooth) is composed of cells that contract to produce movement of body parts, while nervous tissue, including neurons and glial cells, transmits and integrates information through the central and peripheral nervous systems. Models of these tissues and constituent cells should exhibit established native cellular morphological and functional traits. Morphological assessment should, therefore, confirm the shape, structure, form, and/or size of target cells, with alterations in the morphology of cells, potentially indicative of changed cellular function, such as during stem cell differentiation, tumor formation, and cell-pathogen interactions. Cell functionality should similarly recapitulate in vivo cellular processes underpinning the fundamental activities (intra- and intercellular) and the role of a target cell or tissue, such as metabolism, proliferation, respiration, diffusion, osmosis, active transport, ion flux, motility, and electrophysiology.

It is important that phenotypes identified in stem cell disease modelling are relevant to the human disease. Researchers should make necessary efforts to corroborate cell and molecular features that emerge from disease stem cell studies in a patient with the disease, through comparison with post-mortem tissue, relevant patient cells or tissues, or published data.

Assessment of cell specific markers can be performed by common immunophenotyping methods, including flow cytometry for cell antigen analysis of cell suspensions, or immunocytochemistry to evaluate single cell layers of fixed or unfixed cells, or immunohistochemistry, which is performed on fixed-, whole- or sectioned tissue specimens. Common methods for transcript analysis include qPCR, single-cell RNA sequencing (scRNAseq) and gene expression microarrays.

Recommendation 4 .3 .2: Where the development of new benchmarking tools is required to assess a stem cell model, the readout should be extensively validated against reference tissue panels and reproduced in multiple stem cell derivatives.

The highly coordinated networks of gene products that underpin cell states and functions present opportunities to use high-dimensional data such as RNA-sequencing to phenotype cultured cells. However, single markers should not be used to indicate cell lineage, stage, or specific cell identity. Molecular phenotypes should consist of panels of gene products whose correlated behavior is reproducibly associated with functional behaviors. New marker panels should be generally applicable across multiple experiments, or between laboratories; these should be derived using appropriate statistical approaches to benchmark their sensitivity and specificity.

Recommendation 4 .3 .3: Phenotypes that are associated with perturbation assays should ensure phenotypic measurements can distinguish general stress responses from targeted changes . Measurements should control for density-dependent phenotypes if cell cycle, cell growth or cell death are altered.

Generalized stress responses should be expected any time the environment of a cell is altered, and the resulting phenotypes may overwhelm the impact of the planned perturbation. In any screen, controls should be included to monitor for phenotypes that are a consequence of perturbation, or measurement. For example, anti-viral responses should be expected in samples exposed to nucleic acids, including guide-RNAs for genome editing, and may activate cellular shutdown or cell death pathways. If evaluating the perturbation of a specific gene, then controls should include guide RNAs of similar purine/pyrimidine composition. This requires careful evaluation of expected phenotypes (e.g., changes in drug sensitivity, response to a differentiation agent), proliferative responses, morphology changes, or cell death. If any of these form part of the target phenotyping panel, then understanding and reporting on specific types of proliferative signal, or form of cell death (e.g., apoptosis vs pyroptosis) is essential to identify target activities over background phenotypes.

Recommendation 4 .3 .4: Where the model is assessing the impact of a known genotype on the phenotype, it is essential to confirm the stem cell-derived disease model carries the expected genotype.

Genetic validation of patient-derived stem cells is important to confirm the known mutation(s) and may also be useful in authentication of the sample and its origin (see Recommendation 1.3.1). Genetic instability, as well as genetic mosaicism of donor tissue, may contribute to stem cell pools of mixed genotypes.

Proper Controls

Recommendation 4 .4 .1: Consider variability when determining the necessary number of disease and control stem cell derivatives to be included in a study.

Power analysis should be used to determine sample size. This will be impacted by the effect size and penetrance of the phenotype. If effect sizes of the biological readouts are unknown, then aim for the largest sample size available. A strong rationale should be provided for the chosen sample size. If replication is not possible, then variability should be reduced by using isogenic controls. For disease modelling, it is important to select appropriate non-diseased samples to establish base lines for controls (see Recommendations 4.4.2 and 4.4.3).

Recommendation 4 .4 .2: In comparing disease models with healthy controls, the meaning of “healthy” should be clearly defined.

Many models require comparisons of diseased states with non-diseased states. Therefore, it is important to define the parameters of the disease that are being assessed to choose an appropriate control, so that conclusions of any study can be properly placed in context. Given the penetrance of most diseases is age-related, and that many control samples may carry additional risk factors towards the given disorder or be chosen from well-characterized stem cells derived from an unrelated disease cohort, the term healthy is subjective. Therefore, the choice of controls requires consideration of age-matched, ethnicity, sex, familial associations, genotype, and clinical history.

Recommendation 4 .4 .3: The genetic background should be considered when selecting cells to introduce or correct disease-associated mutations.

Stem cells may manifest disease-associated traits because of polymorphisms in the donor or because of culture acquired genetic changes. Beyond considerations of controls outlined in Recommendation 4.4.1, these genetic aspects can confound the phenotypic readout. (See also Recommendation 3.1.1).

Recommendation 4 .4 .4: When comparing isogenic cells derived from genome manipulation, multiple independent clones should be assessed. Where bulk cultures but not independent clones are used, this should be documented.

All current methods for gene editing/gene correction risk introducing unintended genetic changes. These include CRISPR-based approaches, prime-, or base editing, multiple conventional methods of homologous recombination, etc. Researchers should take necessary measures to identify genetic changes and to select lines for phenotypic characterization (of differentiated derivatives) in which these common or rare mutations have not been introduced. Further, it should be noted that healthy individuals and patients can be mosaic in cell composition and the relative proportions of mutant and healthy cells can vary depending on the tissue sample available for stem cell generation. Ideally, the products of independent genetic modification experiments would be compared. In the case of time consuming- or costly experiments, or long differentiation protocols, it may be sufficient to carry out only key experiments relevant to the goals of the study. ‘Independent lines’ are considered to mean derivatives of a single cell selected during reprogramming or gene modification (correction or mutation) arising in an independent well.