1. Introduction

Microphysiological systems (MPS, also known as ‘organ-on-a-chip’, ‘body-on-a-chip’, or ‘human-on-a-chip’) have emerged over the last fifteen years as attractive systems to probe response to pharmaceutical or chemicals. While applicable to animals, such systems are particularly powerful in predicting human response prior to clinical testing of a drug or as augmentation of clinical studies to test underlying mechanisms. The rate of development of such systems has increased exponentially, particularly over the last three years. Organ-on-a-chip systems have evolved in sophistication and ability to model details of organ physiology, allowing better understanding of underlying mechanisms of response to drugs and chemicals. Body-on-a-chip (BOC) systems are multi-organ systems, often designed to emulate human physiological response to drugs and have the potential to capture both efficacy of a drug and potential toxicity in other organs. While the focus has been on human response to pharmaceuticals, such systems can evaluate response to general chemical exposure, which is important in evaluating the safety of chemicals, food ingredients, and cosmetics. Here, we review recent progresses in the development of model systems over the last three years, with particular focus on BOC systems.

The physiological relevance of the model systems is a key factor that determines the success of drug development.¹ Although animal models have been the gold standards for preclinical drug testing, animal models often do not predict human response effectively,² increasing the demand for more advanced model platforms. In addition to the high cost and the ethical issues associated with using animal models, accurate extrapolation of data from animal experiments to humans is also an important limitation. In vitro cell-based models with a single cell type in static culture require lower cost and are more adaptable to high-throughput format but cannot recapitulate many of the complex biological phenomena as well as BOC systems.

Microphysiological systems (MPS) offer several advantages, such as recapitulation of tissue architecture, diffusion kinetics of drugs and signaling factors, and physiological flow conditions.³ Since the initial development in early 2000s, various organ-on-a-chip (OOC) devices have been developed, some of them showing notable successes in recapitulating the physiological functions of in vivo tissues, which was not possible using traditional cell culture models.⁴

One of the fundamental issues with current cell-based single tissue/organ in vitro models is that they cannot reproduce complex interactions between different organs or tissues in the body. This can be a critical issue, since such interactions in the body play essential roles in maintaining homeostasis, as well as in many cases of pathologies. Traditional cell culture models or current OOC systems targeting a single organ or tissue cannot achieve this level of complexity. Such a limitation of traditional in vitro models becomes more critical when dealing with complex pathologies, such as metabolic diseases, obesity, and immunological disorders. MPS technology, which relies heavily on microfabrication and microfluidics, is ideal for mimicking such interactions in a reductionist way, by connecting and integrating multiple ‘modules’ of OOC systems.⁵ These multi-organ systems are often termed as BOC, human-on-a-chip, multi-organ microphysiological systems (MOM), or multi-organ-on-a-chip (MOC), and have emerged as potential tools to evaluate both a drug’s efficacy and side effects that may limit the drug’s usefulness. BOC systems are constructed to mimic the physiological ratios of organ sizes and flow,⁶ while MOC systems may not necessarily attempt to emulate physiological conditions.

Proof-of-concept studies of multi-organ systems were first reported almost 15 years ago,^7–9 and were used to understand the mechanism of toxicity of naphthalene in rodents and to explain differences in response of rats and mice to naphthalene. Since then, recent advances in MPS systems, organoids and stem cell technology, and in vitro vascularization techniques have contributed to the development of improved BOC systems. In this review, we highlight recent progresses in MPS systems aimed at reproducing multi-organ physiology of the human body. We discuss recent advances in novel microfluidic platforms that connect multiple organ modules, and on-chip sensing techniques for real-time analysis of BOC systems, and BOC systems aimed at modeling diseases and drug testing. As described in this review, we have seen remarkable developments in BOC systems. Overcoming several existing challenges promises more advanced systems that can be potentially applicable to clinical situations or incorporated into the drug development process. We will review first advances in the microfluidic platforms for BOCs, then techniques, particularly on-chip systems to monitor responses, then selected single organ systems with multiple cell types, then devices with multiple interacting organ modules.

2. Recent advances and trends in microfluidic platforms for BOCs

Microfluidic technologies enable integration of multiple organ models in a single system to simulate the dynamic organ-organ interactions in a living body. Various BOC systems generally fall into four major categories based on their configurations of fluidic interconnections: i) static microscale platforms; ii) single-pass microfluidic platforms; iii) pump-driven and iv) pumpless recirculating microfluidic platforms, as have been thoroughly reviewed previously.³ During the last two years, there has been substantial growth in the development of BOC systems. Several previously reviewed multiorgan microfluidic platforms have moved beyond the demonstration stage into the development stage. They have been adapted to develop a broader range of BOC models. Some examples include a two-compartment microtunnel platform used to create a functional neuromuscular junction (NMJ) model for phenotypic screening for motoneuron diseases (Figure 1A).¹⁰ A two-chamber, transwell-based recirculating BOC platform driven by pneumatic actuation from the Marx group¹¹ was applied to a pancreatic islet-liver model to simulate insulin and glucose regulation (Figure 1B)¹² and to a tumor-skin model to evaluate the safety and efficacy of antibody therapies.¹³ Multiple organ units or chips have also been connected in series using tubing and/or an integration chip to form BOC systems.^14,15 Such configurations have recently been employed to construct a neurovascular unit to evaluate neurotherapeutics (Figure 1C),¹⁶ a lung-heart-liver BOC to simulate drug response (Figure 1D),¹⁷ as well as a multiorgan model that included the ovary, fallopian tube, uterus, cervix and liver to emulate female reproductive tract and the endocrine loops (Figure 1E).¹⁸ The gravity-driven pumpless recirculating platforms developed by the Shuler group¹⁹ have been employed for multiple BOC systems of two to four organs, including gut-liver models,^20–22 a cardiac-liver chip (Figure 1F),²³ a tumor liver metastasis model,²⁴ and a cardiac-liver-neuron-skeletal muscle chip,²⁵ as well as a whole-body BOC model with 13 explicitly considered organs (Figure 1G).²⁶ New platforms are also rapidly emerging. Here we review the most recent (within the last two years) advances and new trends in microfluidic platform development for BOCs, mainly focusing on the areas of fluidic control, platform materials, platform throughput, and compatibility of circulating cells.

3. Recent advances in on-chip sensing/analysis for BOCs

The potential of organ-on-chip models to predict drug outcomes and reveal underlying disease mechanisms can be vastly improved by the type, quality and non-invasive nature of readouts obtained from these systems. End point readouts on cell viability, cell health, biomarkers, and environmental conditions provide reliable data for highly toxic compounds or for permanent cellular changes.¹⁵ Toxicity to cells, in terms of disruption of normal function, manifests in ways outside of cell death. Temporal mechanical, electrical, and biochemical readouts make the detection of such sub-lethal interactions possible. Dynamic changes to cellular health or functions at multiple timepoints over the lifecycle of the system can be captured by mechanical, electrical and biochemical readouts.²³ Additionally, integrated sensors to monitor the environmental conditions in the system ensure reliability of the readouts obtained from cellular responses to stimuli. Recent advances have been made in sensing environmental conditions, cellular products, and functional outputs on multi-organ chips, with most on-chip sensors developed in single organ systems. Several multi-organ chip systems with on-chip sensing have been reported, and many of the technologies developed in single organs can be adapted for multi-organ chips.

Temperature, pH, and oxygen are three primary environmental conditions that affect cellular function, and changes in these environmental conditions may be indicative of a change in biological function. Temperature impacts metabolic rates, protein folding, and spontaneous levels of action potential frequency in electrically-active cells. Temperature sensors have been produced in which the change in resistance of temperature-sensitive metals such as Platinum is measured, and commercial sensors have also been incorporated as on-chip temperature sensors.¹⁵ Extracellular pH affects a variety of biological processes, extracellular matrix synthesis, cardiac contraction, and immune function. Perturbations in pH can be indicative of metabolic distress, loss of kidney/liver organoid function, or tumor activity. To this end, pH in multi-organ systems can be detected optically via introduction and monitoring of a pH-sensitive dye that is otherwise physiologically inert.⁴⁴ Medium pH can also be monitored by voltage changes in an electrode that has been electro-deposited with a pH-sensitive material.¹⁵ Oxygen levels play a critical role in micro-biological activity. A decrease in oxygen availability can cause deficiencies in cellular metabolism while high oxygen concentrations may lead to increased generation of reactive oxygen species. Optical methods can be used to take real-time measurements of oxygen levels in integrated multi-organ systems. These can be based on the oxygen-sensitive fluorescence of a ruthenium dye, which is chosen for its high photostability and moderate brightness.^7,44 Recently, an inkjet printing technique was used to produce electrochemical dissolved oxygen sensors and integrated into a microfluidic system with hepatocytes and reported oxygen gradients due to metabolism.⁴⁵

Muscle tissues are responsible for generating mechanical forces to varying degrees (cardiac, skeletal muscle, smooth muscle). Cardiac dysfunctions in response to stimuli (disease, environmental, chemical compounds) can include arrhythmia, tachycardia, bradycardia, QT elongation, conduction velocity deficits, and contractile force deficits (Figure 6B). ^{1,15,23,25,46–49} Consequently, cardiac beat dynamics serve as one of the primary metrics for assessing cardiac health over cellular viability.²³ Optical assessment combined with pixel analysis have been used to study cardiac beat dynamics for drug testing in a system containing liver, heart, and lung tissues¹; a heart and liver system¹⁵; and a cardiac only platform⁴⁷. This method can be used to quickly and easily identify changes in beat frequency and regularity. However, this method does not elucidate electrophysiological dynamics (such as QT duration, conduction velocity) or contractile strength and requires imaging during the assay period. Several methods of measuring contraction strength of contractile tissues have been reported for on-chip measurements. Cardiac force changes to drug response has been measured using cells grown on microscale silicon cantilevers or thin film cantilevers. The silicon cantilevers reported used an incident laser directed at the tip of the cantilever, and the deflection of the reflected beam during cellular contraction was directly correlated to force using a known transfer function in multi organ systems with heart, liver, neuron and skeletal muscle (Figure 6A).^23,25 Using thin film cantilevers, the force has been determined less directly using a video capture and image analysis system to correlate bending of the thin film to force.⁵⁰ Skeletal muscle contractions have also been captured under electrical stimulation by optical pixel analysis,²⁵ and by using an optical force transducer and a computer controlled linear actuator.⁵¹ Integrated on-chip impedance electrodes were shown to measure the cardiac contractile force in a fully integrated static culture chamber ⁵². Additionally optical pixel methods have been employed to assess the change in contraction force and frequency dynamics,¹⁰ displacement length, and contraction frequency in neuromuscular junction models.⁵³

Electrophysiological dynamics of cardiac and neuronal tissues can be assessed via their culture onto micro-electrode arrays (MEA’s) to elucidate drug effects in a heart-liver ²³, vascular endothelium-heart⁴⁹, and heart only⁵² systems. This technology enables the high-temporal-resolution measurement of cardiac action potentials with a direct electrical readout. Under the correct conditions, these measurements can be used to determine field potential duration, a QT interval analogue.^49,54 When coupling with defined chemical patterns, the conduction velocity as the contraction propagates throughout the cardiac tissue can be assessed. Additionally, via electrical stimulation at one electrode, the frequency dynamics of the cardiac syncytium can be measured^23,54 Patterned microelectrode arrays have been recently developed to measure cardiac response in a microfluidic system that additionally included liver and to observe changes in these measurements in response to drug compounds.²³ Moutaux et al reported a MEA-integrated microfluidic device with two independent chambers to demonstrate action potential propagation between the presynaptic and postsynaptic chambers via axons growth between the chambers through microtunnels.⁵⁵

Transendothelial electrical resistance (TEER) is an electrical technique implemented to quantify integrity of barrier tissues non-invasively.⁵⁶ This has primarily been used in single organ systems such as the blood brain barrier (BBB),^34,49,51 GI tract,^27,32 skin models.^57,58 This method is versatile, with a variety of electrode materials, electrode configurations, and measurement conditions (such as input frequency) being reported. Cells release and uptake biomolecules as part of routine maintenance, metabolic activities as well as in response to external stimuli. The release of biomarkers in response to stimuli can be immediate or delayed. Monitoring the presence of these biomolecules over time provides a description of cell function including efficient working or disruption of cellular pathways and sheds light on mechanism of drug actions.⁵⁹ Zhang et al. reported an integrated biochemical sensing platform in a multi organ system composing heart and liver. Electrochemical biosensors that operate by an enzymatic catalysis reaction to generate an electric current was employed for on-chip detection of albumin, glutathione A transferase for monitoring liver toxicity and creatine kinase to assess cardiac viability.¹⁵ Misun et al. developed an integrated multi-analyte biosensor system for the monitoring of glucose and lactate to assess cellular metabolism of cancer microtissues, in which the working electrodes were functionalized with enzymes and detection was achieved through amperometry.⁵⁹ Bavli et al. reported on-chip microprobes to measure glucose and lactate for monitoring the metabolic pathways in a liver on chip model.⁶⁰ Curto et al used an on-chip organic electrochemical transistor to report oxygen measurements in a microfluidic system with canine kidney cells ⁶¹.

The most exciting advancement in on-chip sensing in BOC systems has been the development of systems incorporating multiple on-chip sensors for non-invasive measurements of several tissue systems simultaneously. Oleaga et al. combined MEAs for electrical measurements and silicon cantilevers for mechanical measurements in a system with cardiac and liver components, and used the system for measuring off-target cardiac toxicity of compounds which undergo hepatic metabolism, demonstrating organ-organ interactions.²³ Specifically, in this multi-organ system, the liver reduced the cardiotoxicity of terfenadine, as measured by electrical and mechanical function, as the liver metabolized the cardiotoxic parent to non-toxic metabolites. Additionally, the liver increased the cardiotoxicity of cyclophosphamide as the liver metabolized the non-toxic parent to toxic metabolites such as acrolein. The varying dose effects of doxorubicin a chemotherapeutic to different organ systems were shown in a 4-organ BOC system comprising liver, heart, skeletal muscle and neurons using readouts for cardiac, skeletal muscle contraction force and neuron electrophysiological data.²⁵ Zhang et al showed cardiotoxicity in a heart-liver system to acetaminophen with real time on-chip monitoring of biomarkers and cardiac beat frequency.¹⁵ Skardal et al. demonstrated an assortment of bioengineered tissue organoids and tissue constructs that are integrated in a closed circulatory perfusion system, facilitating inter-organ responses between the liver, heart, and lung, using integrated TEER electrodes and optical monitoring (Figure 6D).¹⁷

Recent advances in techniques for on-chip analysis of electrical, mechanical, and chemical activity of tissues have contributed to the improvement in the performance of recent BOC models, as well as easier and more in-depth validation of the BOC models. As noted earlier, further improvements in the resolution of the detection, as well as non-invasive, real-time, and multiplexed detection in more devices will contribute to a more physiologically-relevant BOC models.

4. Recent advances in multi-organ models

The human body is composed of multiple organs, connected via circulatory networks. The organs in the human body function in an inter-dependent manner, which is orchestrated by complex, yet largely unknown biological mechanisms. In addition to the organs comprising the body, the vascular networks play an important role of supplying nutrients and removing wastes from organ tissues. Vascular networks enable communications between different organs and tissues, via soluble signals such as cytokines and hormones. Mimicking the biological mechanisms of the human body requires development of a model system that emulates such complex multi-organ interactions. Current in vitro model systems are generally based on static single organ or tissue systems and are incapable of capturing such mechanisms. There exist some microplate well-based in vitro model systems for modeling interactions between different organs or different cells, by co-culturing cells or transferring media in between different cell cultures.^62,63 However, these systems do not capture the real-time dynamics of multi-organ interaction and consequently may miss essential underlying mechanisms.

The importance of real-time dynamics of multi-organ interaction can be manifested with several examples. Xenobiotics entering the body may undergo a series of transformations involving multiple organs. For example, the fate of an orally administered drug is largely determined by interactions between the gut, liver, kidney, and even the microbiota in the gut.⁶⁴ How the drugs interact with these organs profoundly affect the pharmacokinetics (PK) of the drugs (that is, absorption, distribution, metabolism and elimination (ADME) characteristics), as well as the toxicity of the drugs, which are often observed in organs other than the target sites. Maintaining homeostasis in the body requires organs interacting with one another via several different mechanisms, soluble signaling being one of them. For example, the glucagon and insulin are endocrine hormones that control the balance in glucose metabolism.⁶⁵ Cytokines regulate the immune response to foreign invaders.⁶⁶ Many of metabolic diseases, such as diabetes mellitus, progress as the interaction between multiple organs goes awry.⁶⁷ Organs and tissues often interact with one another through more direct mechanisms, such as cell migration or cell to cell communication via direct contact. For example, immune cells can migrate into a tissue with inflammation as part of the body’s immune response to foreign invaders. Cancer cell metastasis occurs via malignant cells transmigrating the vascular endothelium and invading tissues at distant site. Many of the biological phenomena described above are difficult to recapitulate with current in vitro models. For example, recirculation of immune cells between organs is difficult due to shear damage. In this section, we start with an in vitro model of a single organ primarily focusing on the models of the liver and the skin as organs representing a solid organ with metabolic functions and a barrier tissue, respectively. Then we expand our discussion to a more complicated model of multiple organ interactions.

5. Remaining challenges and future directions

Despite the considerable advances on BOC systems made during the past decade, several aspects remain to be improved. Overall, BOC systems should be able to provide reproducible results that have significant physiological relevance to human data. Validation of these in vitro models that can substitute or complement current animal models will be critical to development of this technology.

First, the development of a blood surrogate, a common medium for multiple cell types, is essential for advancing to more complete BOC systems. Currently, different cell types may require different nutrients and growth factors, which makes it challenging to co-culture these cells in the same systems for an extended time. This problem becomes more serious as there is often a desire to use stem cells and primary cells in BOC systems, as they often require specialized media. A modular approach, where each organ module can be integrated or detached as needed, can be a partial solution to this issue, because cells usually require specialized medium during the differentiation phase, and it is often possible to maintain a short-term co-culture of multiple cell types with basal medium.¹²⁶ In some cases, mixing different media at specific ratios has been used with some success.¹⁵ However, these strategies are only partial solutions since it will be more difficult with increasing number of organs, and obviously, the development of a universal, as well as chemically well-defined media that can support growth and differentiation of all cell types would be a more complete solution. Hickman et al. demonstrated using a serum-free medium for supporting co-culture of muscle cells derived from human stem cells and sensory neurons,¹²⁷ and co-culture of motoneurons and oligodendrocyte,¹²⁸ as well as in their 4-organ systems,²⁵ which show great promise for a co-culture systems with a larger number of organs.²⁵

Secondly, development of biomaterials that meet specifications of different cell types is also needed to further improve current BOC models. Although current researchers are equipped with diverse families of natural and synthetic biomaterials,¹²⁹ chemical and physical interaction of cells with surrounding 3D environment is an important factor in the cellular microenvironment, and fine-tuning of this microenvironment is needed for different cell types. Another importance aspect of using appropriate materials in OOC technology is the issue of unwanted absorption and adsorption of chemicals to the inner surface of MPS devices, which can distort the experimental observation. Since different materials have widely varying degrees of interaction with different molecules, choice of materials in developing MPS devices should be made with caution. In general, stainless-steel, glass, Teflon, and polystyrene are considered to be inert to most of chemicals,¹³⁰ whereas PDMS (polydimethylsiloxane) is known for its hydrophobic surface and tendency to both adsorb and absorb many molecules ¹³¹. Coating of PDMS surface with various materials such as extracellular matrix proteins or paraffin has shown to be helpful.^22,132,133 Using thermoplastic materials such as polystyrene, polymethyl methacrylate (PMMA), and polycarbonate offers advantages in both aspects of mass production and relatively low adsorption and absorption, but require more complicated fabrication techniques.

Thirdly, construction of physiologically realistic fluidic microenvironment is important. The role of fluidic shear stress on cell morphology, alignment, and cytoskeleton organization of vascular endothelial cells is well known,¹³⁴ and microfluidic devices have been extensively used in investigating the underlying mechanisms.¹³⁵ The fluid dynamics of circulating medium may also carry important implications for circulating cells, such as circulating tumor cells (CTC) and immune cells, as a high level of shear stress can damage the cells in the device.¹³⁶

Finally, correctly scaling the BOC systems is essential to achieve maximum physiological relevance. In allometric scaling principles, it is known that the sizes of different organs scale differently between animals of varying sizes.¹³⁷ Basal metabolic rates of different organisms are also known to follow a three quarter power relationship with the masses of organisms.¹³⁸ This issue of scaling becomes serious since BOC systems are generally highly miniaturized system of the human or animal body. Maintaining correct ratios of organ sizes, as well as transport rates between the organs is essential for replicating physiological responses, for example, pharmacokinetic response to drug administration ^22,108. Several strategies of scaling and designing BOC systems have been suggested, but a more wide consensus needs to be formed.^137,139,140 Establishing physiologically realistic ratios of liquid to cell volumes is also an important issue. In most cases, biochemical reactions are concentration dependent, which means that the ratio of liquid to cells can alter considerably the rates of reactions. In a traditional microwell plate-setting, cells are usually immersed in excess amount of liquid, resulting in the dilution of molecules. Comparing the liquid-to-cell ratios of various BOC systems revealed that the values of liquid per cells were considerably smaller than traditional macroscale systems, but still larger than human physiological values.³ However, attempts to meet the physiological level of liquid-to-cell ratio poses additional engineering problem, such as need to achieve sufficient supply of circulating medium, and analytical techniques to detect molecules with small quantities of sample.

6. Concluding remarks

As described, MPS and particularly BOC systems have been developing rapidly. The primary focus has been on using such systems in preclinical drug development to select which drug candidates are most likely to enter human clinical trials and become approved drugs. The focus has been on BOC systems using human cells to potentially predict efficacy and off-target toxicity during preclinical drug development studies. Almost all major pharmaceutical companies are exploring this technology, because they recognize its potential to improve the success rate of drug candidates in clinical trials. Even a modest improvement in success rate would provide more useful drugs to society and at a potentially reduced cost.

The acceptance of such new technology by the pharmaceutical industry will be involved with acceptance by regulatory bodies. Fortunately, FDA and its European counterpart, EMA, have been involved in six workshops with academic and industrial leaders on this technology.¹⁴¹ These workshops have begun a dialog among technology developers, regulators, and potential users on approaches to best bring this technology forward. In cases where the safety of a drug is already established, it is possible to use only in vitro data to extend its use to new applications.

While pharmaceutical firms are exploring this new technology, it is critical for MPS and BOC systems to reach the point where they are adopted as a standard component of the drug development process. The fact that it has been demonstrated that BOC systems could have predicted the failure of drugs in clinical trials or even those drugs that had to be withdrawn from the market is encouraging. As companies develop more confidence from tests of their own proprietary compounds, they will become more willing to adopt this technology in preclinical testing.

Of course, further improvement in MPS and BOC technologies is required for wide spread adoption of this technology. A primary constraint will be to reduce the cost of those systems, increase throughput and maintain relevant predictions. Many companies are working on methods to reduce cost significantly and increase throughput. Self-contained multiorgan systems with embedded sensors may be ideal. Additionally, advanced manufacturing techniques should reduce per unit costs significantly. A critical issue, and often a bottleneck, is development of biological organ modules that are robust, work in a common medium, and replicate key aspects of organ physiology. For example, iPS cell technology is improving rapidly. If one could derive all the relevant organs from a single genotype, it would make testing more reproducible. It would also open doors to use in precision medicine to find the best treatment option for individuals. Also, many of the drugs that fail in the market do so because a subpopulation responds adversely to the drug. One can imagine the use in drug development of iPS cells derived from a spectrum of subpopulations. These subpopulations can differ from one and another in many ways; for an example, they may differ in certain cytochrome P450 enzymes, which would profoundly affect drug metabolism. You could picture using perhaps 10 or 12 models from humans from diverse geographic locations or highly inbred populations. From such studies you might decide that a drug would be ideal for one subpopulation, but not useful or potentially toxic for another.

While this review has focused on preclinical drug development, those systems have the ability to test for the potential toxicity of chemicals, particularly in cosmetics and foods. Indeed, some firms in the cosmetic industry have invested significantly in MPS and BOC technology to predict human response to chemical additives. Ultimately, we believe that these systems are likely to reduce significantly the reliance on animal testing and to improve predictions concerning potential toxicity.

Overall, we believe that this technology is beginning to mature and to approach commercial adoption. There still remains much to do to increase reliability and decrease cost, but the large number of academic groups and companies involve suggest that creative solutions to any problems will be found.

Acknowledgments

Biographies

Footnotes

References