Microphysiological Systems Evaluation: Experience of TEX-VAL Tissue Chip Testing Consortium

Abstract

Much has been written and said about the promise and excitement of microphysiological systems, miniature devices that aim to recreate aspects of human physiology on a chip. The rapid explosion of the offerings and persistent publicity placed high expectations on both product manufacturers and regulatory agencies to adopt the data. Inevitably, discussions of where this technology fits in chemical testing paradigms are ongoing. Some end-users became early adopters, whereas others have taken a more cautious approach because of the high cost and uncertainties of their utility. Here, we detail the experience of a public-private collaboration established for testing of diverse microphysiological systems. Collectively, we present a number of considerations on practical aspects of using microphysiological systems in the context of their applications in decision-making. Specifically, future end-users need to be prepared for extensive on-site optimization and have access to a wide range of imaging and other equipment. We reason that cells, related reagents, and the technical skills of the research staff, not the devices themselves, are the most critical determinants of success. Extrapolation from concentration-response effects in microphysiological systems to human blood or oral exposures, difficulties with replicating the whole organ, and long-term functionality remain as critical challenges. Overall, we conclude that it is unlikely that a rodent- or human-equivalent model is achievable through a finite number of microphysiological systems in the near future; therefore, building consensus and promoting the gradual incorporation of these models into tiered approaches for safety assessment and decision-making is the sensible path to wide adoption.

A [VERY] BRIEF HISTORY OF MICROPHYSIOLOGICAL SYSTEMS

The shift of biomedical engineering science from the focus on improving the properties of biocompatible materials and development of bioartificial organs, to the development of microphysiological systems, largely occurred in the 1990s. At that time, the efforts in miniaturization of bioartificial organs had reached the microscale owing to progress in both materials science and understanding of the cellular and molecular biology (Prokop, 2001). Recent advances in the miniaturization of more complex tissue-scale biology, combined with the advances in microfluidics (Whitesides, 2006), have led to a transition of biomedical engineering of microphysiological systems into a robust and burgeoning interdisciplinary field (Mitchell, 2001). Accordingly, microphysiological systems development has rapidly evolved from micropatterning and scaffold-based designs to elaborate microchannel and reservoir-containing devices that can be fabricated from various synthetic polymers, and may be perfused using either mechanical or gravitational force (Marx et al., 2020). As the result, biomedical science and funding trends followed the microphysiological systems-focused excitement of the 2010s that promised development of various “organ-on-a-chip” model and even as much as a “human-on-a-chip” (Khetani and Bhatia, 2008; Marx et al., 2012).

In the past decade, government funding agencies, philanthropists, and private equity fueled the rapid and massive inflow of funding into the field (Ishida, 2021; Low et al., 2021; Low and Tagle, 2017; Vinken, 2020). As a result, not only are the number of publications and discoveries on the rise, but so is the number of companies that are offering a range of microphysiological systems (Wu et al., 2020). One key driver for the “value proposition” in the commercialization of these technologies is the belief that microphysiological systems may address the shortcomings in current drug development, such as (1) improving clinical efficacy, (2) eliminating the need for animal-to-human extrapolation, and (3) reducing the costs in drug development and safety evaluation by increasing efficiencies and human relevance of the model systems (Franzen et al., 2019; Roth and MPS-WS Berlin 2019, 2021). Most recently, an evolution is emerging from using these for internal discovery and de-risking (Allwardt et al., 2020; Ewart et al., 2017, 2018; Marx et al., 2016) to submitting the data from microphysiological systems in support of registrations for pharmaceuticals (Marx et al., 2020; Vulto and Joore, 2021) and pesticides (U.S. EPA, 2018).

REGULATORY DECISION CONTEXTS: WHERE COULD MICROPHYSIOLOGICAL SYSTEMS FIT?

One common argument for supporting development of microphysiological systems is that it is a sensible path toward generating human-relevant data and reducing, or quite possibly eliminating the need for the use of animals in research and regulatory testing (Smirnova et al., 2018). To achieve this goal, the developers need to appreciate the considerable inconsistencies of the regulatory requirements across different sectors and types of chemicals, and the intricacies of drug development and approval workflows. For example, the regulatory regimes are quite diverse depending on the use of chemical. Drugs and pesticides typically require extensive and prescribed data packages, cosmetic ingredients cannot be tested in animals in many parts of the world, and commodity chemicals have varying data requirements for their registration and use. To increase mutual understanding among researchers and regulators, joint meetings are being held between the developers, end users, and regulators (Anklam et al., 2022; Dourson et al., 2022; Fabre et al., 2014; Piergiovanni et al., 2021a). A number of recent scholarly publications and road maps provided some clarity to the views of the regulatory agencies on the utility of microphysiological systems in their decision contexts (Avila et al., 2020; ICCVAM, 2018; Thomas et al., 2019; U.S. EPA, 2018).

Because much effort in development and commercialization of microphysiological systems has been directed at their potential use in the pharmaceutical industry, the position of the United States Food and Drug Administration (U.S. FDA) has been considered the most relevant in the United States. The U.S. FDA has endorsed efforts to develop and use qualified alternatives to animal and human studies in its 2017 Predictive Toxicology Roadmap (U.S. FDA, 2017). The Agency is actively engaged in discussions with the developers and vendors, is testing the technology in their laboratories (Rubiano et al., 2021; Wang et al., 2021), and also published a perspective on the decision contexts where microphysiological systems and other new models may be most useful (Avila et al., 2020). Sponsors who wish to present data from microphysiological systems in their submissions, or discuss with the U.S. FDA staff in presubmission consultations, are encouraged to do so. Increased familiarity of the U.S. FDA reviewers with the technology and greater confidence of the companies in taking these data to the U.S. FDA have resulted in several informative examples of when and how microphysiological systems can serve as “fit-for-purpose” information in regulatory submissions (Marx et al., 2020; Roth and MPS-WS Berlin 2019, 2021; Vulto and Joore, 2021). It is yet to be determined, however, if these data have made meaningful contributions in regulatory decisions because there are very few examples of these data being presented to the regulators.

Regulators in chemical toxicology and environmental health have also expressed cautious interest in the utility of microphysiological systems (Andersen et al., 2014). Although the landmark National Academies report (National Research Council, 2007) that is credited with initiating the pivot of regulatory toxicology toward embracing alternatives to animal and human testing did not explicitly mention microphysiological systems, the follow-up report (National Academies of Sciences Engineering and Medicine, 2017) did note that these are an important tool to be considered in risk-based evaluations. Since 2007, large-scale in vitro toxicity testing campaigns have been conducted by the Federal agencies (Thomas et al., 2019), albeit all of the data collected thus far on thousands of chemicals and drugs, in hundreds of cellular and molecular assays, are yet to include microphysiological systems (Williams et al., 2017). The concept of the New Approach Methods has recently gained popularity in the context of using data from alternative methods in regulatory decisions (Bal-Price et al., 2018). To this effect, both U.S. EPA (Thomas et al., 2019) and the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM, 2018) declared commitments to support the development of the New Approach Methods and increase confidence in their use among U.S. regulatory agencies. It is envisioned that microphysiological systems may be used as a second- or third-tier testing option where organ- or tissue-specific context may be needed to elucidate the potential mechanisms of adverse effects that were “flagged” in high-throughput screening campaigns of specific molecular targets.

The U.S. EPA has taken a position (U.S. EPA, 2021) that (1) their major environmental regulations do not specifically require the use of animals, but rather call for “the best available science” to be used, which may include the New Approach Methods; and (2) the authority for U.S. EPA’s research programs arising from these statutes is sufficiently broad and does not constrain the Agency from developing or advancing the use of new data streams. The U.S. EPA is required to reduce or replace the use of vertebrate animals in the testing of chemical substances or mixtures and promote development and use of alternative test methods or strategies that do not require new vertebrate animal testing (U.S. EPA, 2018). Accordingly, the U.S. EPA released the work plan to implement these mandates which includes, amongst other objectives, a focus on research and analysis to establish scientific confidence in the New Approach Methods in the context of use for regulatory decisions (U.S. EPA, 2021). The European Chemicals Agency has advocated a similar approach (ECHA, 2020) and encourages submissions of such data in chemical registrations or when animal testing waivers are requested; albeit acceptance of the New Approach Methods data in REACH dossiers is currently rare.

Beyond the general interest of multiple agencies, the only informative example of a microphysiological systems-related model that was considered in a regulatory context for U.S. EPA (eg, pesticide regulation) is a case study of refining inhalation risk assessment for spray application of the pesticide chlorothalonil, a contact irritant. This case study was recently presented to the EPA Federal Insecticide, Fungicide, and Rodenticide Act Scientific Advisory Panel (FIFRA SAP, 2019). Specifically, a point of departure (POD) was derived for chlorothalonil using “a three-dimensional in vitro test system of human respiratory tissues” (U.S. EPA, 2021) because severe local irritation effects in traditional short term animal toxicity studies, “the usual approach for assessing inhalation health risks was considered unworkable” (FIFRA SAP, 2019). The data from the MucilAir model (Crespin et al., 2011) were used to derive a POD, which was then coupled with a computational fluid dynamics model to directly predict the corresponding human equivalent POD based on calculation of the deposition of chlorothalonil in the human respiratory tract. In their opinion, the FIFRA advisors agreed that conceptually, using an in vitro test system such as MucilAir to assess the toxicological profile of irritants or cytotoxicants is appropriate and could eventually be widely implemented, but only when a number of shortcomings in the overall approach are addressed (FIFRA SAP, 2019). In particular, the need for repeat exposure studies, direct testing of the aerosols, and testing of more relevant regions of the small airway (the model was seeded with nasal epithelial cells) were identified as critical gaps. To date, no further disposition of these data in a decision on chlorothalonil has occurred, although research to further develop and evaluate these models with gases and aerosols is continuing (Mistry et al., 2020).

PERSPECTIVES OF THE END-USERS OF MICROPHYSIOLOGICAL SYSTEMS

End-users of this technology in the pharmaceutical, and to a lesser degree, chemical industries have embraced the microphysiological systems as a valuable tool for de-risking of compounds and improving target selection (Allwardt et al., 2020; Marx et al., 2020). The recognition of the potential for microphysiological systems has been broad, it also has been highly un-even in terms of how different stakeholders define the potential context(s) of use. Recent surveys (Leite et al., 2018; Prior et al., 2019; Schneider et al., 2021) show that although a number of stakeholders were early adopters through both academic collaborations and partnerships with vendors, others have taken a more measured approach of focused investments in a limited range of microphysiological systems, or just waiting for more clarity to emerge. A common hesitation from the end-users has been the nascent state of the technology and the brisk pace of development that often outpaced the merits of investments in a particular technology. It has also become very clear that the developers seldom take the potential “purpose” for their models into consideration as early as the conceptual design phase. This has created a classic “fitting a square peg into a round hole” scenario where most microphysiological systems that were developed would have no immediate application because they were not designed to be “fit-for-purpose.” To address this major challenge, a pharmaceutical industry consortium concerned with promoting the adoption of microphysiological systems has published a series of manuscripts that aimed to detail the overall needs of the industry and provided specifics on what phenotypes need to be modeled and assayed, and defining the appropriate benchmarks for Microphysiology systems to be deemed useful (Fabre et al., 2020). Combined with the regulator’s perspectives (Avila et al., 2020; ICCVAM, 2018; Thomas et al., 2019; U.S. EPA, 2018), such information is invaluable to bridging the gap between very promising technology and its application, and to address pressing practical needs of the end-users.

In addition to the “mutual education” between the end-users and developers that is focused on better defining the fit-for-purpose contexts, 2 other common threads have emerged from the experience of the user community. Initially, there is a need to characterize the robustness, the ability to withstand rigorous testing, of the technology when it is being used outside of the primary developer’s laboratory. This led to the discussion about increasing scientific confidence in specific models, with a focus on establishing the most sensible approach to standardization of the technology. To address both of these, several concerted efforts have been taken in the United States, Europe, and Japan. In the United States, the National Center for Advancing Translational Sciences (NCATS) in 2016 funded 2 “Tissue Chip Testing Centers,” one at the Massachusetts Institute of Technology and one at Texas A&M University, as well as the Microphysiology Systems Database (MPS-Db) Center at the University of Pittsburgh (Schurdak et al., 2020). The goal of these grantees was to transfer microphysiological system platforms and cells from the developers to their laboratories, to independently replicate published findings, assess model’s robustness, develop best practices and standard operation protocols for each model that could be successfully transferred, and provide input for further improvements, all information that is critically lacking and impedes wider adoption of these models (Livingston et al., 2016; Low and Tagle, 2017). Similar, albeit less formalized efforts to test reproducibility of microphysiological systems are underway in Europe (Mastrangeli et al., 2019) and Japan (Figarol et al., 2020). These efforts identified a number of common barriers to successful technology transfer of microphysiological systems in a reproducible manner: generally low throughput, high cost, the need for highly specialized equipment, and the availability of cells and other necessary reagents and materials to establish and maintain it in a functional state (Marx et al., 2020).

In addition to addressing the challenges with technology transfer, there is a need for building confidence through standardization and qualification of microphysiological systems (Ewart and Roth, 2021). Although the 2 are not the same, for a given intended purpose, studies can be designed and conducted to qualify standardized models as acceptable surrogates. A number of initiatives and discussions are underway to consider if existing procedures for harmonizing regulatory acceptable test methods would be applicable to microphysiological systems. Because these are to be used not only in discovery and biomedical research, but also for application as potential prognostic tools, a variety of options exist. These may range from the international standards for clinical laboratories (Antonelli et al., 2017), to the processes used by the Organization for Economic Cooperation and Development (Rasmussen et al., 2019), or the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (Seldrup, 2011). Various “good practice” documents have also been recently developed for the New Approach Methods as a whole, these could also be adopted to microphysiological systems technology (Kolle and Landsiedel, 2021; Krebs et al., 2019). However, microphysiological systems are typically more complex than other in vitro methods, and the various polymers they are made from, presence of microfluidics and other technical considerations, make choices for standardization and subsequent qualification a formidable challenge (Piergiovanni et al., 2021b). It is not clear whether standardization of such models is sensible at this time because of the rapid pace of development of the technology; narrow-focus qualification for specific context(s) of use may be a more practical path. However, as these models are sufficiently qualified for narrowly defined “contexts of use,” benchmarks for standardization will be important for regulatory applications.

TEX-VAL TISSUE CHIP TESTING CONSORTIUM: LEARNINGS FROM THE TECHNOLOGY TRANSFER OF MICROPHYSIOLOGICAL SYSTEMS

The creation of the Tissue Chip Testing Centers program by NCATS in 2016 (Hargrove-Grimes et al., 2021; Livingston et al., 2016) allowed for discussions of the challenges with technology transfer in the microphysiological systems field to be conducted with systematically collected relevant data. Two laboratory-based centers were funded from 2016 to 2020. The center at Texas A&M University evolved into a university-based collaboration known as the TEX-VAL Tissue Chip Testing Consortium (TEX-VAL Consortium) that began operations in January 2020. The center at the Massachusetts Institute of Technology is now operating as Javelin Biotech, Inc. (Woburn, Massachusetts). The Database Center remains supported by NCATS and is based at the University of Pittsburgh (Schurdak et al., 2020).

TEX-VAL Consortium brought together pharmaceutical and consumer products companies, the trade association of chemical manufacturers, and government agencies that develop, promote and use the New Approach Methods. Member organizations collectively decide on the annual work plan for selected models, whereas the research staff at Texas A&M University conduct experiments and share their findings with the member organizations. All data generated by TEX-VAL Consortium, as well as detailed study protocols, is being deposited into the database established by NCATS specifically for the data from microphysiological systems (Schurdak et al., 2020). This ensures compliance with the “Policy on enhancing public access to archived publications resulting from NIH-funded research” (Steinbrook, 2005) and with the FAIR (Findable, Accessible, Interoperable, Reusable) principles (Wilkinson et al., 2016).

Because the scientists who participate in the Consortium’s research activities and work-in-progress discussions have accrued a unique set of knowledge and expertise in evaluation of the utility of microphysiological systems for a number of regulatory applications, we reason that our perspective is unique and informative to the broad audience of scientists and regulators alike. It is our opinion that the literature on microphysiological systems development and potential use in regulatory decisions has been dominated by the opinions that can be best described by the words “very promising” and “rapidly developing.” Although we do not disagree with these sentiments, we believe that the practical challenges and realities of microphysiological systems adoption remain largely under-reported. There is also a risk that that the most optimistic opinions may be coauthored by those who are affiliated with either the agencies that fund this research, the companies that are vendors of these devices, or the organizations that promote alternatives to animal testing. It is often not clear whether the authors have first-hand experience with technology transfer of these very diverse technologies, and thus can offer a balanced view of the state of the art on the utility of microphysiological systems.

 

Prioritization of Microphysiological Systems for Testing and Use

Some concerns existed when the Consortium was being organized that it may be difficult to ensure that all member organizations are confident in the “value added” by the collaboration. Member organizations had a varying level of experience with microphysiological systems and also are quite diverse in their “context of use” considerations. However, the Consortium was able to easily establish annual work plans that would incorporate the wishes of all members because the focus on major organ systems and contexts of data use presented common opportunities to “try before you buy.” The organs tested by the Consortium are listed in Table 1. In general, although many of the organ/tissue models were based on commercially available devices, none of the testing was a straightforward reproducibility study of the existing/published model; in every case, additions or changes were made based on the input from the Consortium members to enhance each model’s “fit for purpose.”

Table 1.

Microphysiological Systems Tested by TEX-VAL Consortium in 2020–2021

Organ	Cell Types (Platform)	Studiesa	Chipsb
2020
Kidney	Glomerulus (Mimetas 3-lane)	6	360
	Proximal tubule (Mimetas 3-lane)	7	320
Liver	Multicellular (Nortis Bioc)	11	90
Gut	Caco-2 (Mimetas 3-lane)	8	206
	Enteroids (Mimetas 3-lane)	4	405
Lung	Small airway (custom devicesd)	6	115
	Total	42	1500
2021
Kidney	Glomerulus (Mimetas 3-lane)	4	160
	Proximal tubule (Mimetas 3-lane, CNBio TC-12, Transwells)	7	524
Liver	Single-/multicellular (Mimetas 2-lane, CN Bio LC-12)	23	1182
Gut	Caco-2 (Transwells, CN Bio TC-12)	7	243
	Enteroids (Transwells, CN Bio TC-12)	11	456
BBB	Multicellular (Transwells)	3	32
	Total	55	2600

^aNumber of the individual studies for each cell type/platform combination. See Supplementary Table for the links to the study descriptions, protocols, and data in the Microphysiology Systems Database.

^bNumber of the individual microphysiological systems devices (ie, “chips”) in each study.

^cThe study is detailed in Sakolish et al. (2021a,b).

^dThe study is detailed in Sakolish et al. (2022).

Overall, there are several learnings from the experience of establishing a broad consensus among diverse stakeholders interested in microphysiological systems. One, there is great deal of good will and interest in collegial discussions regardless of company or sector affiliation; the microphysiological systems hold promise for all potential stakeholders and diverse participants are more than willing to view this technology as a commodity that needs to be tested before some of the proprietary applications can be considered in-house. Two, there are few options for what is actually “ready” to be onboarded/tested. The reality is that the quantity of publications has yet to translate into quality in terms of readiness for technology transfer. Even though the rapid development of the technology is widely appreciated, it was somewhat surprising to all members that the options that are ready to be put through independent evaluation were relatively few. Three, even if there are microphysiological systems that are ready for onboarding/testing, most of them need to be further adapted, some far more than others, in terms of how they are assembled or evaluated. Four, there is a general sense of doubt regarding the advertised level of microphysiological systems readiness and utility from both academic publications and marketing materials from vendors. The “try before you buy” needs are very acute and the Consortium provided a valuable opportunity to determine whether a certain platform is ready to be onboarded and put into research and development pipelines.

Cell Sourcing Is the Top Vulnerability of High-Fidelity Microphysiological Systems

It has been long recognized by biomedical engineers that it is far more difficult to replicate human biology than to make reproducible tissue chips. This challenge is not unique to the field of biomedical engineering, the difficulty in reproducing findings from in vitro experiments that used a particular lot of primary cells or a certain donor is common knowledge (Hirsch and Schildknecht, 2019). It is virtually impossible to fully replicate a study if it used a specific lot of primary cells that is no longer available. Even for some of the immortalized cells that still retain basic functions of their origin, offerings may differ among vendors. From a practical point of view of an end-user, the need to qualify every new lot of primary cells, a preferred cell type, in a complex model is a major barrier to adoption. Unfortunately, the promise of induced pluripotent stem cell (iPSC) technology has not been realized yet in full to support the reproducibility of microphysiological systems due to challenges in cell differentiation and maturation.

These challenges were addressed by the Consortium in several ways. For tissues where qualified iPSC-derived cells are available, like the liver, we compared different lots of primary human hepatocytes to iPSC-derived hepatocytes from established vendors. For tissues where various immortalized cell types and stably overexpressing clones are available, like renal proximal tubule epithelial cells, we compared those with different lots of the primary cells. For other tissues where cell offerings are limited, the Consortium procured cells from different vendors and conducted marker-assisted cell sorting to evaluate purity of the preparations, as well as the viability of cells after cryo-storage. In cases when established cell lines, like Caco-2, are a common model, we used those cells for comparison. Altogether, these considerations were not the only reason why numerous controls need to be tested before study test compounds, but also the dependence of the overall “success” of each experiment typically hinging upon which cells go into the device, in addition to the device itself. Regardless of the cell type or tissue, in general these models utilize a higher number of cells than corresponding experiments in multi-well plates (eg, 96 and greater). Because cells are often the most expensive components in the overall microphysiological system, albeit the cost of the tissue chips varies widely, this can further limit throughput to create sufficient context for interpretation, and reduce the opportunity for intra- and interexperiment replication.

Therefore, the primary learning here is that a reliable cell source may outweigh other factors when considering what microphysiological systems to use for a particular application. The high variability of the results from the same experiments conducted by different laboratories is likely the result of challenges with cell sourcing. Given the limited availability of some cell types (eg, primary cells) and the large number of cells required for some models, there are limits to conduct of reproducibility studies. The options for resolving this challenge exist, such as development of the methods for at-scale production of human iPSC-derived cells from different tissues (Ryu et al., 2021); however, we are likely several years away from robust and reproducible supply of iPSC-derived cells for many tissue types. This is also one reason why large-scale standardization efforts of microphysiological systems may be premature at this time, the biological variables should be standardized first.

Microphysiological Systems Configuration and Flow

Some in the field of microphysiological systems development argue over definitions and terminology. The prevailing view is that unless there is flow, the model is not microphysiological. These discussions have merits and there are many valid reasons for why flow improves tissue function. The benefits of “flow” are largely attributed to enhanced nutrient availability and waste exchange that eliminate “static” culture conditions that likely differ across microphysiological systems (eg, scaffold-free, free-floating organoids increasingly identified in the spectrum of microphysiological systems). With chip-based platforms, the benefits of flow come with technological complexity as considerations need to be made for the hydrodynamics of each model and whether physiologically relevant parameters can be achieved. There is a range of options for what constitutes flow in various microphysiological systems. The Texas A&M University Center has tested tissue chips where flow was driven by pumps (large or small) or gravity (Liu et al., 2020; Sakolish et al., 2018, 2020a,b, 2021b). Similarly, the Consortium decided to test microphysiological systems that facilitated media flow in various ways, ranging from the use of laminar to peristaltic pumps, resulting in pulsatile or continuous gravity-driven media movement, or completely static conditions. In general, it is no surprise that throughput is highest when no flow is needed and the lowest when syringe pumps need to be placed close to or inside the incubators. But consideration of the cost-benefit of increased complexity is often neglected in the discussions of what flow is physiological.

The learning from the Consortium is that flow-based models generally tend to display somewhat more physiological performance; however, the amount or type (ie, continuous or pulsatile) of sheer stress in those models is far from being physiologically relevant. Thus, the discussions we had as a group about the added value of flow were always accompanied by the considerations of throughput and cost of adding this feature to the model. How these considerations are resolved is highly model dependent, but our collective experience is that the jury is still out on the relative benefits of the added technical complications for most of the models that may be used routinely, rather than in one-off special cases.

Studies of the Barrier Function in Microphysiological Systems: Gel Layer Is a Formidable Barrier

To assemble multilayered tissues in vitro, the common approach is to create a physical barrier using various gel matrices or scaffolds that provide physical separation between cell types, and are meant to simulate basement membranes in different barrier tissues. Because of the physical and topological configurations of various existing microphysiological systems that contain gel-based compartments, such a barrier is unphysiologically thick. It also may significantly impede small molecular transport, a requirement for estimating human pharmacokinetics and modeling organ/tissue function. Little data are available in the published literature or from the vendors to address the utility of such models for studies of pharmacokinetics or the distribution of chemicals over time in their respective microphysiological systems. Although modeling internal exposure dynamics is not the only reason to use these models, it is a very important consideration for a number of tissues where the transfer of chemicals between compartments or cell types is the cause of the adverse effect. Different cell types, or cell preparations of the same cell type, may also affect the scaffold and/or gel barriers in different ways. Invasion into the matrix would have a profound effect on apparent pharmacokinetic properties of a test compound, yet it is difficult to quantify between replicates or experiments. In this regard, the learning from the Consortium is that some gel-layer containing microphysiological systems are poorly suited for studies of pharmacokinetics; ADME modeling with these models may be limited to metabolism and to a lesser degree include evaluation of re-uptake/secretion at this point in their development. Its widely recognized that nominal concentration is not the best metric for quantifying dose response in vitro (Armitage et al., 2021), yet the challenges for predicting the distribution of test articles in microphysiological systems are considerable.

Imaging Remains the Primary Means of Phenotyping Microphysiological Systems

It has been our experience that fluorescence-enabled immunocytochemistry is the most robust phenotyping method for microphysiological systems to establish the biological relevance and to monitor changes in performance under exposed conditions. Although biochemical assays can be performed on the media collected over the time course of the experiment, for example, for the liver models albumin and metabolic markers can be used to assess the basic functional state of each device before treatments are applied, other tissues however generally lack similarly informative biomarkers. In this regard, immunostaining is a valuable, albeit terminal assay, but is not suitable for continuous monitoring of the functionality (eg, decreased effective throughput). A comparison can be made to a terminal sacrifice of the animals in an in vivo study, the data collected will be representative of a particular time point and condition; however, histopathology would be available from dozens of tissues in an animal study, but only from one “tissue” in microphysiological system. Another important learning here is that those who are interested in using microphysiological systems shall consider robustness and versatility of their imaging capabilities. Some of the microphysiological systems require high-resolution microscopes, others can be imaged in plate-enabled instruments. Most valuable are confocal imagers because most tissue chips are multilayer constructs.

Cost of the Experiments

Although the cost is not the primary deciding factor when human health and safety are at stake, microphysiological systems have not yet been shown to be the only option, or even the preferred option, for investigating a particular condition/phenotype or for establishing dose-responses in target cells/organ systems. There are numerous published examples of microphysiological systems’ advantages as compared with other existing models and methods (Anklam et al., 2022; Dourson et al., 2022). It is not, however, widely appreciated how formidable the differences in set-up and routine use costs for various options may be. It may be perceived that collaborations with academic labs are the low-cost option. Yet this is generally not the case because most tissue chips that may be manufactured by the academic laboratories would be the least ready for technology transfer. The quality of such manufacturing varies widely, costs to cover personnel and materials are high, the Universities typically include overhead charges, and the complexities of the technology transfer are many, especially when the receiving party is a commercial entity. The time needed to realize any advantages from such collaborations is typically the primary deterrent to end-users.

Commercial platforms may spare the end-users from some of these challenges. These are typically transactional relationships and vendors bear responsibility for manufacturing and other deficiencies in their products. The costs, although generally high, are predictable, at least in the short term. The risks are also many, as most device providers are small businesses that are at risk of discontinuing their technology at any time. The Consortium considered a number of commercial vendors and chose models manufactured by Mimetas (Leiden, Netherlands) and CN Bio (Cambridge, United Kingdom) for the experiments in the most recent studies, in part, because of the reasonable set-up and maintenance costs. Initial set-up was most economical for the Mimetas devices, albeit some auxiliary equipment may be quite costly. This model also comes with options for multiplexing and choices of configuration (2- or 3-lanes) where the cost of materials per experiment is generally lower and throughput is greater. The CN Bio set-up cost was greater and the throughput is lower; however, this model offered additional potential advantages such as the ability to add flow to trans-well cultures toward modeling barrier function that are already familiar to many researchers (eg, drug metabolism and pharmacokinetics).

A hidden, but equally important factor is the cost of skilled labor that is needed to conduct experiments with microphysiological systems. Special training is needed not only for appreciation of the greater technical complexity of these models as compared with traditional cell-based models, but also for understanding when various parts of the system, not just the device with cells, are malfunctioning. Our experience is that those who were trained as biomedical engineers have an easier time adapting to different platforms and are more attentive to the performance of the complete system. Those who were trained in molecular biology and other fields would also become highly proficient, but the learning curve is typically longer. Collectively, the learning from the Consortium is that our collaborative model offers a unique opportunity to assess the realistic performance and needs of each platform and to be able to make correct decisions when investments are made into technology and staff. The Consortium also is in a position to train member organization’s research staff to increase efficiencies and assist in troubleshooting.

CONCLUSIONS

Microphysiological systems are considered within virtually any and all discussions of alternatives to animal testing or the New Approach Methods. Both the developers and the prospective end-users are very enthusiastic about the prospects of replacing costly tests that can require large numbers of laboratory animals and that may be of questionable biological relevance to humans with miniature gadgets that, one day, may serve as acceptable proxies for human physiology. The opinions of experienced users of these models are more nuanced, the goodwill interest and guarded optimism are confounded by the realities of working with a rapidly evolving technology that requires major investments of time, money, and personnel training. Not to mention the still evolving considerations of developing scientific confidence in what this technology may be good for—replacing existing tests and models in a context of reducing/replacing/refining the use of animals, filling in the gaps where current models are known to fail, or creating new opportunities for modeling rare disease modalities? It is likely that the answer is “all of the above.” Still, to every end-user, those who were early adopters and charged ahead with a firm belief in the potential, and those who are still waiting on the sidelines because the cost of entry is forbidding, the outlook is still cloudy of how the data from microphysiological systems might be used. They face similar challenges—the need to define what is practically achievable and what is aspirational. It is difficult to size up the whole, or even major parts, of the still expanding universe of device options.

It is with these considerations the TEX-VAL Consortium was conceived as a collaboration of the end-users who are genuinely interested in the technology’s potential, but wished to benefit from the shared cost structure, common expertise, and collective considerations of the strengths and challenges with each technology, and their use in “next generation” human health assessments (Andersen et al., 2019; Baltazar et al., 2020; Cote et al., 2016). Collectively, we offer a number of perspectives from the practical point of view of a very realistic “median” user of microphysiological systems technology. We reason that these experiences and considerations will be informative and encouraging to those who are fully immersed and those who are still on the sidelines.

First, the Consortium is not taking a position on the merits of each individual device that was tested, whether it was from an academic lab or a commercial vendor. The utility of each model is to be decided by each member organization based on the data that was procured through the Consortium’s research program. No model is perfect, but whether they are useful depends on the priorities and needs of each end-user. Most importantly, experience gained through the Consortium shows prospective end-users need to appreciate that it is highly unlikely that they will be able to acquire a plug-and-play system; some form of on-site optimization and quality control of the cells and reagents will be needed. Similarly, the phenotyping of the microphysiological systems will fall on the end-user as well and we encourage the prospective users to consider the whole “eco-system,” not just what a vendor or collaborator can, or promises to, deliver in terms of the device itself. The potential to use 1 platform for multiple organs/tissues is attractive, but our experience demonstrates that one-size-fits-most is still an aspirational goal. It is also likely that the end-users will need to hedge their bets across several technologies because almost all vendors are small companies that may face challenges with performance of their equipment and/or availability of the reagents.

Second, we learned that it is not the device itself that is the most critical determinant of achieving success. It is the cells, related reagents, intended application, and the technical skills of the research staff who need interdisciplinary training and have broad skills from engineering, to cell biology, imaging, and chemistry. These factors may be overlooked when strategic management decisions are made regarding choices of microphysiological systems, but they will become the most important considerations for their routine use.

Third, studies of pharmacokinetics of drugs and chemicals in microphysiological systems are considered as a very promising area of their application (Sung, 2021). Although many studies of microphysiological systems address kinetics through a narrow focus on nonspecific binding of the chemicals to the devices themselves, it is only one part in addressing the needs for extrapolation from concentration-response effects in microphysiological systems to human blood or oral exposures. A number of publications demonstrated the potential of tissue chips to provide informative data for physiologically based pharmacokinetic and pharmacodynamic modeling of human drug responses, as well as for in vitro-to-in vivo extrapolation (Prantil-Baun et al., 2018). Such data are critical for translation of measurements in microphysiological systems, or any other in vitro model, to predict human exposure conditions (National Academies of Sciences Engineering and Medicine, 2017). Thus, differences among microphysiological systems in their ability to provide data for dosimetry are a critical consideration. For example, in multicell microphysiological systems that separate cells by a gel barrier, options for collecting pharmacokinetics data may be limited.

Fourth, the regulatory schemes are often focused on the data needs for certain “toxicities” and adverse effects on a finite number of tissues. Although the options are abundant for a number of such tissues, the majority of other important ones remain under-developed (ie, reproductive, developmental, immune, etc.). Therefore, a battery of organotypic models or a “whole system” (eg, human or animal studies) may be needed to determine efficacy and safety. A related challenge is the need for subchronic and chronic exposures (equivalent to 90 day or 2-year regulatory studies in animals), something that microphysiological systems are unlikely to address, at least in the short- to medium term.

Finally, most end-users are facing the prospect of deciding what constitutes the “right sizing” of each model they are planning to introduce into their existing pipelines. This is especially true for microphysiological systems, because both miniaturization and microfluidics increase complexity, especially when the goal is to couple organ-specific chips together to further improve physiological relevance. It is unlikely that a rodent- or human equivalent can be put together through a finite number of devices, despite the elegant work of many labs that are trying to develop interconnected systems. The softer the microphysiological systems or nothing rhetoric, the more realistic it may be for more end-users to become welcoming of this technology as part of an integrated approach comprised of a suite of fit-for purpose investigative tools (eg, computational methods, mechanistic assays, 2D or 3D submerged cell culture methods, microphysiological systems, and targeted in vivo studies in lab animal models). Polarizing statements will not help with building consensus, expanding use in everyday routines in research and development of safer medicines and chemicals, and promoting the gradual adoption of microphysiological systems-derived data into tiered approaches for safety assessment and regulatory decision-making.

SUPPLEMENTARY DATA

Supplementary data are available at Toxicological Sciences online.

FUNDING

National Center for Advancing Translational Sciences (U24 TR001950, U24 TR002633); and by equitable monetary contributions from TEX-VAL Consortium member organizations.

DECLARATION OF CONFLICTING INTERESTS

The authors who are affiliated with Texas A&M University, the National Toxicology Program, and U.S. EPA declare no relevant conflicts of interest. Other authors are employed by various commercial entities or nonprofit organizations who may use various microphysiological systems or have commercial relationships with vendors of the models mentioned in this manuscript. The views expressed in this manuscript do not necessarily represent those of the author’s affiliated organizations. The use of specific commercial products in this work does not constitute endorsement by the author’s affiliated organizations or the funding agencies.