Hype or hope – microphysiological systems?

Listen to this article on the DDW Podcast:

Dr Tomasz Kostrzewski, Director of Biology at CN Bio Innovations, examines MPS in drug discovery as part of a preclinical toolbox. 

Drug development remains slow and costly for the pharmaceutical industry, driven by the financial impact of late-stage failures of drug candidates¹. Although our understanding of the molecular basis for many human diseases has greatly improved, the number of FDA drug approvals, per million US dollars spent, has decreased consistently since 1950². The current drug discovery process involves lengthy and costly lead discovery and optimisation campaigns, often on poorly validated targets in cellular models with weak translational relevance to human disease. The primary reason for clinical trial termination remains lack of human efficacy, whilst preclinical and Phase I clinical failures are more prominently due to issues with safety³. Drug development therefore needs new approaches and tools to deliver on the promise of science for patients. Faster, more predictive and disease-relevant in vitro and in silico models are required, which will identify risks early in discovery programs and enable a focus on molecules which directly target key pathways associated with human disease.  

Capturing the complexity of human disease states in robust translational in vitro assays is no small task. Recently, through the convergence of sophisticated engineering, computing, and biological research approaches, new approach methodologies (NAMs) for nonclinical testing have started to be developed and validated⁴. NAMs is an umbrella term describing a range of novel approaches including in silico mathematical prediction models, alternative non-mammalian in vivo models (e.g. zebrafish) and advanced in vitro tissue models, which are all starting to be adopted into the drug development process^4,5. NAMs aim to better mimic human physiology and provide more accurate predictions of human responses. Such capabilities help reduce drug attrition by aiding target identification and lead optimisation, enabling more promising candidates to be optimised before they reach clinical trials. 

Microphysiological systems (MPS) are a branch of NAMs used to provide valuable insights across a range of research areas, including metabolic and infectious disease, ADME and toxicology. MPS are integrated systems that recapitulate organ function in vitro and are synonymous with terms such as organ-on-a-chip, and body-on-a-chip⁵. The financial impact of adopting MPS technology into drug discovery workflows is profound: cost reductions are predicted to be 10-26% over a five-year period, equivalent to up to $700 million⁶. MPS therefore clearly offer great potential to strengthen the drug development pipeline, but misconceptions around how they fit into the broader preclinical landscape remain. However, to realise their true value, we need to define the contexts in which MPS are most beneficial and explore how these tools can be used alongside other preclinical models to improve the accuracy and efficiency of drug discovery. 

Features of MPS technologies 

MPS bring together previously disparate biological and engineering advances such as the use of microfluidics, microfabrication engineering, induced-pluripotent stem cells (iPSCs), 3D-bioprinting and multicellular cell cultures, to permit the study of human tissues and organs in vitro. These miniature engineered devices support human cells grown in 3D structures with tissue-level functions and human-relevant responses to biological cues, stimuli, and/or compounds⁷. 

MPS typically mimic vascular-like perfusion – fluid flow supplies nutrients to the tissue structures, creates microenvironmental biomolecular gradients and provides relevant mechanical cues (e.g. shear stress)⁸. These features constitute a major advantage of MPS, differentiating them from conventional (static) cell cultures. With MPS, temporal and spatial gradients are precisely controlled, and there is a high level of control over cell type and placement. Therefore, the resulting tissues in manufactured platforms can be produced with low variability and high physiological relevance. Furthermore, MPS provide the opportunity for long-term in vitro studies, compared to 2D cultures, which don’t offer the same level of longevity. Overall, MPS design is diverse⁷, and a range of sophisticated approaches has been employed to recapitulate the structure and function of a range of human organs, including the liver^9-11, kidney^12,13, intestine¹⁴, and lung¹⁵. 

MPS can also include multi-organ models, sharing a common medium, allowing inter-organ crosstalk by cellular or endocrine signalling^16-18. Importantly, multi-organ MPS can extend the benefits of single-organ technologies to form in vitro models that better represent how organs interact and communicate as part of a complex system. This approach is highly complex as it requires the optimisation of many parts; for example, determining which system microfluidics should be used to recreate a physiologically relevant cardiac output. While some early, bold initiatives were aimed at linking many organs together¹⁸, MPS solution providers of today focus on producing more usable models with specific groups of tissue types. This might mean incorporating liver, intestinal and kidney tissues to enable studies of drug absorption, metabolism and excretion¹⁹.  

Adoption of MPS in drug discovery  

MPS offer value at multiple stages of the drug discovery process, including in studies of human-specific diseases, target identification, lead optimisation and preclinical safety assessment. Most MPS are not high-throughput platforms, so are unlikely to be used for large-scale screening activities. Instead, they can be used to obtain deep mechanistic insights using a high number of endpoints, for tens or hundreds of molecules. The likely success of later stage preclinical development can be vastly increased by using MPS to better predict human-specific ADME properties, toxicities and efficacy.  

MPS for target identification 

Having research models that closely replicate human disease states is critical to the successful identification of novel drug targets and pathways. Such capabilities are needed to study high-burden diseases, such as non-alcoholic steatosis (NASH), a disease associated with fat deposits in the liver, inflammation, fibrosis, and insulin resistance. If left untreated, NASH can develop into cirrhosis and hepatocellular carcinoma. Many companies have embarked on drug discovery programmes to tackle the problem, yet their high-profile compounds continually fail to meet primary endpoints in late-stage clinical trials. To this day, NASH remains a disease of unmet medical need.  

In a recent study, bone morphogenetic protein 8B (BMP8B) was identified as a major contributor to NASH progression²⁰. BMP8B was shown to promote inflammation and pro-fibrotic pathways in hepatic stellate cells in an MPS NASH model (showing NASH key hallmarks), and the absence of BMP8B was found to promote a NASH phenotype in several other preclinical models^{20, 21}. Human liver biopsies showed increased BMP8B expression correlated with NASH disease stage progression, together confirming a key role for BMP8B in NASH²⁰. The transcriptomic profile of the MPS NASH model was also shown to corroborate closely with data from human tissue, further indicating the relevance of the MPS model to the clinical disease and justifying the pursuit of BMP8B inhibitors as NASH drug targets²⁰.  

MPS in lead optimisation 

Predicting drug efficacy in preclinical models remains a major challenge in drug discovery, as many current animal models do not replicate the pathophysiology of human disease. Recreating disease pathology affecting the pulmonary vasculature, for example, is highly challenging, but MPS have already shown their usefulness in this context. The architecture of the lung in idiopathic pulmonary fibrosis, with stiff fibrotic tissue, has been modelled by a lung-on-chip model that recreates a functional alveolar–capillary interface²². The model enables the study of anti-fibrotic drugs in greater detail than is possible in vivo, such as the inhibitory effect of nintedanib on neo-vascularisation²².  

Human MPS models of disease have also been developed in cases where preclinical animal models do not exist, e.g. Hepatitis B virus (HBV) infection. Using a liver MPS incorporating primary human hepatocytes and immune cells, all the stages of the HBV life cycle were recapitulated, creating opportunities to study potential biomarkers and specific pathways for immune evasion¹¹. The model can be employed to study in detail the molecular effects of anti-viral therapeutics targeting the immune response (e.g. interferon α) or directly targeting the virus (e.g. viral entry inhibitors)¹¹. 

By using a variety of translational endpoints in MPS studies, the functionality of a particular tissue model and the effects of a particular compound can be accurately determined. For this reason, MPS solution providers are collaborating with research groups around the world to address the urgent need for Covid-19 therapies. To assess SARS-CoV-2 antiviral therapeutics and prophylactics, human-lung responses to viral infection were recently modelled in lung MPS cultures²³. SARS-CoV-2 pseudoparticles containing the SARS-CoV-2 spike protein, which faithfully reflect key aspects of the native virus’ entry into host cells, were applied to the lung MPS model, and the number of pseudoparticles in the infected cells was measured²³.  

Optimising leads  

In later stages of preclinical development, advanced in vitro technologies offer a way to identify and optimise the most promising candidates. Through studies of human intestinal absorption, liver clearance, and kidney excretion, MPS can be used to predict human bioavailability and pharmacokinetics/pharmacodynamic (PK/PD) relationships, allowing better predictions for first-in-man dosing. 

Intestinal MPS that use primary human intestinal cells or organoids have been developed; these reflect in vivo functionality, with well-maintained tight junctions, GI barrier functionality and CYP and P-gp expression at higher levels²⁴. These intestinal MPS can be used to study molecule absorption (equivalent of an oral drug dose) and their physiological relevance can be further enhanced if they are combined with human gut bacteria to mimic the microbiome^14,25. Liver MPS can play an important role during drug development by providing valuable insights into population variability in hepatic drug metabolism²⁶. Predictions of liver drug clearance rates from MPS studies have been shown to translate with observed human clearance rates²⁶. The predictive capability of liver MPS is superior to that of standard 2D cultures, as their metabolic capacity can be maintained over longer time periods²⁷. This is advantageous for studies of low clearance compounds and extends the potential scope of in vitro studies²⁷. Studies of kidney MPS currently remain more limited (most focusing on nephrotoxicity), although some involve kidney MPS which can express various metabolic enzymes and transporters required for ADME studies²⁸. 

Predicting human-specific toxicity  

Hepatoxicity is a major cause of lead compound attrition in drug development²⁹. Hepatic MPS have been developed to provide insights into the mechanistic causes of drug-induced liver injury (DILI). An immunocompetent liver-on-a-chip model incorporated hepatocytes and nonparenchymal cells to assess the acute form of DILI resulting from diclofenac secondary metabolite formation³⁰. As these cells typically lack functional longevity in standard static cell cultures, using the MPS enabled the formation of tissue-like structures, which are critical for maintaining functionality³⁰. In the model, diclofenac produced a metabolism and toxicity profile comparable to that observed in humans, but not in animal models³⁰. An alternative liver MPS approach demonstrated how cross-species toxicity profiles could be predicted by culturing liver cells from multiple species in the same set-up and comparing effects to different hepatotoxicants³¹. Importantly, toxicity in these MPS can also be evaluated using clinically relevant biomarkers such as human albumin, alanine transaminase and aspartate transaminase, the same endpoints as measured in clinical trials to monitor liver function. 

Along with the liver, the kidneys are a major site of toxicity and numerous kidney MPS have now been developed to enable predictions of nephrotoxicity. These models culture human proximal tubular cells in several formats and configurations and, as with the liver MPS, use a range of standard and clinical endpoints to assess toxicity³². One recent study showed how a renal proximal tubule-on-a-chip could be used to predict toxicity profiles of a wide range of molecules in a highly robust manner and was amendable to long-term chronic exposure studies³². 

How can MPS deliver on their promise? 

MPS clearly have the potential to supplement datasets and bridge current shortfalls in the drug discovery workflow. Single- and multi-organ MPS are ideal platforms for addressing questions that cannot be answered using other approaches and serve as a complimentary cross-validation tool. To maximise the success and value of MPS going forward, key guiding principles should be followed. MPS models should be built using primary cells, or from a source of stem cells that can be consistently shown to fully differentiate. Ideally, these cells would be incorporated into a co-culture to better mimic physiological conditions. The inclusion of immune cells can further enhance translation, but this is dependent on the specific context of use for the MPS model. 

Cellular exposure to therapeutics and other compounds should be guided by careful considerations of free drug availability, protein binding, and tissue metabolism, to help ensure exposure closely matches in vivo conditions. Finally, the endpoints used to assess tissue function should be the same, or at least comparable, to those used in clinical studies. Appropriate selection of clinical biomarkers, e.g. blood-based biomarkers, is essential to streamlining the translation of in vitro data into clinical outcomes.  

The development of MPS technologies has accelerated in recent years and has seen significant success, but there remain hurdles to overcome. The technology’s limitations must be considered in the same way as those of other approaches (e.g. in silico, high throughout in vitro, or other NAMs). All preclinical models should be judged by the same set of standards to enable end users to be clear about which approach is right for them. MPS approaches must be developed with a clear context of use in mind, so the solution they provide can be clearly articulated. This will be ever more important with the rise of more complex multi-organ MPS, as these will never be able to capture all the complexity of an in vivo model. The pharmaceutical industry will also require increased throughput for the technology and for the approaches to be further automated so they can be used in an increasingly routine fashion. 

Making the change from entrenched gold-standard 2D cell culture and animal models to NAMs such as MPS will require a coordinated effort from suppliers, industry, and regulators. For regulatory bodies to accept MPS data within IND submissions, there is first a need to demonstrate the robustness, reliability, and performance of MPS models. Here, strong collaborations are essential, and significant progress has been made in this area¹⁰. Regulatory authorities are now acknowledging the potential of MPS and are investing in initiatives to help underpin their use through research and collaboration. The recent first co-publication between FDA scientists and MPS developers was a significant milestone for the technology’s development. The publication showed that a liver MPS can be used reproducibly for the assessment of drug toxicity, metabolism, and intracellular accumulation¹⁰. Of note, the liver MPS systematically detected the toxicity of trovafloxacin, a broad-spectrum antibiotic, which was withdrawn from the market due to its hepatotoxic potential¹⁰.  

MPS and the future of drug discovery? 

The use of MPS has increased dramatically over the last decade and they are becoming more common in drug discovery workflows. However, some misconceptions remain, and many are yet to understand the true value of these technologies. This confusion is understandable, given the large amount of media hype that has surrounded their development. Many headlines imply that MPS technology will now replace animal testing and serve as a simple fix to all the challenges associated with bringing a drug to market. Those in the MPS field are frequently asked: “Will MPS replace the use of animals? How are MPS better than in vivo models?” More pertinent questions would be: “What are the disadvantages of animal models?” and “How can MPS be used in combination to fill knowledge gaps?”. MPS are important tools to be used by preclinical researchers alongside other NAMs such as in silico modelling and organoid technology. Together, the use of these NAMs will help plug the gaps left by traditional approaches. While mice models capture the complexity of a complete organism, MPS models demonstrate how a disease mechanism, or drug effects, will differ in a human setting. Using both models as part of a research programme gives a much stronger and broader insight that supports development of more successful therapeutics.  

With drug development being notoriously inefficient and costly, there is a need for new approaches, paradigms, and tools⁷. Finding the optimal balance of research approaches can be achieved by asking the right questions of the right tool, whether that be an advanced in vitro model, animal model, in silico approach or otherwise.  

In other words, the whole is greater than the sum of its parts. Together with other NAMs, MPS are paving the way for more insightful decisions, and a more efficient and cost-effective drug discovery pipeline. The adoption of MPS will likely be encouraged by regulatory agencies to reduce the reliance on animal testing, so care must also be taken to consider the true merits of each novel MPS assay in its context of use. While NAMs will not suddenly replace or eliminate traditional tools such as animal models, we should upgrade our overall methodology by integrating NAMs such as MPS in support of a new era of productive drug discovery. 

Volume 22, Issue 4 – Fall 2021

About the author 

As the Director of Biology at CN Bio, Dr Kostrzewski is responsible for biological tissue model development, applications for these models and collaborative research projects with academic and commercial partners. With more than 15 years of experience in molecular and cellular biology research, Dr Kostrzewski has a PhD in Molecular Immunology from Imperial College London and experience working within large pharma developing biopharmaceutical therapeutics. 

References 

1. Paul, S. M. et al. How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nat. Rev. Drug Discov. 2010; 9, 203. doi: 10.1038/nrd3078.
2. Scannell, J. W., Blanckley, A., Boldon, H. & Warrington, B. Diagnosing the decline in pharmaceutical R&D efficiency. Nat. Rev. Drug Discov. 2012; 11, 191-200. doi: 10.1038/nrd3681.
3. Hay, M.; Thomas, D. W.; Craighead, J. L.; et al. Clinical Development Success Rates for Investigational Drugs. Nat. Biotechnol. 2014, 32, 40–51. doi: 10.1038/nbt.2786.
4. Avila, A. M. et al. An FDA/CDER perspective on nonclinical testing strategies: Classical toxicology approaches and new approach methodologies (NAMs). Regul. Toxicol. Pharmacol. 2020; 114, 104662. doi: 10.1016/j.yrtph.2020.104662.
5. Truskey G. Human microphysiologicalsystems and organoids as in vitro models for    
6. toxicological studies. Frontiers in Public Health. 2008; 6:185. doi:10.3389/fpubh.2018.00185 
7. Franzen N, van Harten WH, RetèlVP, Loskill P, van den Eijnden-van RaaijJ, IJzerman M. Impact of organ-on-a-chip technology on pharmaceutical R&D costs. Drug Discovery Today. 2019; 24(9):1720-1724. doi:10.1016/j.drudis.2019.06.003 
8. Low, L. A., Mummery, C., Berridge, B. R., Austin, C. P. & Tagle, D. A. Organs-on-chips: into the next decade. Nat. Rev. Drug Discov. 2021; 20 (5): 345-361. doi:10.1038/s41573-020-0079-3.
9. Esch EW, BahinskiA, Huh D. Organs-on-chips at the frontiers of drug discovery. Nature Reviews Drug Discovery. 2015;14(4):248-260. doi:10.1038/nrd4539.
10. Ribeiro AJS, Yang X, Patel V, MadabushiR, Strauss DG. Liver microphysiological systems for predicting and evaluating drug effects. Clinical Pharmacology & Therapeutics. 2019;106(1):139-147. doi:10.1002/cpt.1458. 
11. Rubiano A, IndapurkarA, YokosawaR, et al. Characterizing the reproducibility in using a liver microphysiological system for assaying drug toxicity, metabolism, and accumulation. Clinical and Translational Science. 2021;14(3):1049-1061. doi:10.1111/cts.12969. 
12. Ortega-Prieto AM, Skelton JK, Wai SN, et al. 3D microfluidic liver cultures as a physiological preclinical tool for hepatitis B virus infection. Nature Communications. 2018;9(1). doi:10.1038/s41467-018-02969-8.
13. VriendJ, Vormann M, Lanz H, Joore J, Trietsch S. Nephroscreen: A robust and versatile renal tubule-on-a-chip platform for nephrotoxicity assessment. Current Opinion in Toxicology. 2021; 25:42-48. doi:10.1016/j.cotox.2021.03.001. 
14. Weber EJ, ChapronA, ChapronBD, et al. Development of a microphysiological model of human kidney proximal tubule function. Kidney International. 2016; 90(3):627-637. doi:10.1016/j.kint.2016.06.011. 
15. Kim HJ, Li H, Collins JJ, Ingber DE. Contributions of microbiome and mechanical deformation to intestinal bacterial overgrowth and inflammation in a human gut-on-a-chip. Proceedings of the National Academy of Sciences. 2015;113(1): E7-E15. doi:10.1073/pnas.1522193112.
16. Huh D, Leslie DC, Matthews BD, et al. A Human disease model of drug toxicity–induced pulmonary edema in a lung-on-a-chip microdevice. Science Translational Medicine. 2012; 4(159):159ra147-159ra147. doi:10.1126/scitranslmed.3004249.
17. Chen WLK, Edington C, Suter E, et al. Integrated gut/liver microphysiologicalsystems elucidates inflammatory inter-tissue crosstalk. Biotechnology and Bioengineering. 2017; 114(11):2648-2659. doi:10.1002/bit.26370.
18. TrapecarM, Communal C, Velazquez J, et al. Gut-liver physiomimetics reveal paradoxical modulation of IBD-related inflammation by short-chain fatty acids. Cell Systems. 2020;10(3). doi:10.1016/j.cels.2020.02.008. 
19. Edington CD, Chen WLK, GeisheckerE, et al. Interconnected microphysiologicalsystems for quantitative biology and pharmacology studies. Scientific Reports. 2018; 8(1). doi:10.1038/s41598-018-22749-0. 
20. Tsamandouras, N. et al. Integrated Gut and Liver MicrophysiologicalSystems for Quantitative In Vitro Pharmacokinetic Studies. AAPS J. 2017;19, 1499–1512. 
21. Vacca M, Leslie J, Virtue S, et al. Bone morphogenetic protein 8B promotes the progression of non-alcoholic steatohepatitis. Nature Metabolism. 2020; 2(6):514-531. doi:10.1038/s42255-020-0214-9.
22. Kostrzewski T, MaraverP, Ouro‐GnaoL, et al. A microphysiological system for studying nonalcoholic steatohepatitis. Hepatology Communications. 2019; 4(1):77-91. doi:10.1002/hep4.1450. 
23. Zeinali S, Bichsel C A ,Hobi N, Funke M et al. Human microvasculature-on-a chip: anti-neovasculogenic effect of nintedanibin vitro. Angiogenesis 21. 2018; 861–871. doi: 10.1007/s10456-018-9631-8. 
24. Si L, Bai H, Rodas M, et al. A human-airway-on-a-chip for the rapid identification of candidate antiviral therapeutics and prophylactics. Nature Biomedical Engineering. 2021:1-15. doi:10.1038/s41551-021-00718-9.
25. Chen H J, Miller P, and Shuler M L.  A pumpless body-on-a-chip model using primary culture of human intestinal cells and a 3D culture of liver cells. Lab Chip, 2018, 18, 2036–2046.
26. Zhang J, Huang Y, Yoon Y J et al. Primary Human Colonic Mucosal Barrier Crosstalk with Super Oxygen-Sensitive Faecalibacteriumprausnitziiin Continuous Culture. Med 2, 2021; 74-98. doi: 10.1016/j.medj.2020.07.001. 
27. TsamandourasN, Kostrzewski T, Stokes C L, et al. Quantitative assessment of population variability in hepatic drug metabolism using a perfused three-dimensional human liver microphysiological system. The Journal of Pharmacology and Experimental Therapeutics. 2017;360(1):95-105. doi:10.1124/jpet.116.237495. 
28. VivaresA, Salle-Lefort S, Arabeyre-Fabre C, et al. Morphological behaviour and metabolic capacity of cryopreserved human primary hepatocytes cultivated in a perfused multiwell device. Xenobiotica. 2014;45(1):29-44. doi:10.3109/00498254.2014.944612. 
29. Weber E J, ChapronA, ChapronB D, et al. Development of a microphysiological model of human kidney proximal tubule function. Kidney Int., 2016, 90, 627–637. doi: 10.1016/j.kint.2016.06.011. 
30. Rowe C, Shaeri M, Large E, et al. Perfused human hepatocyte microtissues identify reactive metabolite-forming and mitochondria-perturbing hepatotoxins. Toxicology in Vitro. 2018;46:29-38. doi:10.1016/j.tiv.2017.09.012
31. Sarkar U, Ravindra K C, Large E, et al. Integrated Assessment of Diclofenac Biotransformation, Pharmacokinetics, and Omics-Based Toxicity in a Three-Dimensional Human Liver-Immunocompetent Coculture System. Drug MetabDispos. 2017; 45, 855–866. doi: 10.1124/dmd.116.074005.
32. Jang K J, Otieno M A, RonxhiJ, et al. Reproducing human and cross-species drug toxicities using a Liver-Chip. Sci. Transl. Med. 2019; 11 (517). doi: 10.1126/scitranslmed.aax5516.
33. VormannM K, Vriend J, Lans H L, et al. Implementation of a Human Renal Proximal Tubule on a Chip for Nephrotoxicity and Drug Interaction Studies. J. Pharm. Sci. 2021; 110, 1601–1614. doi: 10.1016/j.xphs.2021.01.028.