We've updated our Privacy Policy to make it clearer how we use your personal data. We use cookies to provide you with a better experience. You can read our Cookie Policy here.


How Far Can Organs-on-Chips Go On Their Own?

Listen with
Register for free to listen to this article
Thank you. Listen to this article using the player above.

Want to listen to this article for FREE?

Complete the form below to unlock access to ALL audio articles.

Read time: 7 minutes

Most people have heard about Moore’s law, the golden rule of the electronics industry, stating that computers’ processing power doubles every two years. But have you heard of Eroom’s law? This trend shows that R&D costs per drug approved roughly doubled every ~nine years between 1950 and 2010.1 This, added to the fact that today’s drug candidates are more likely to fail in clinical trials than those in the 1970s2, goes to show how difficult bringing a new drug to the market can be. Different reasons have been put forth explaining the high failure rate, one of them being that preclinical models might not reflect properly the human drug response.

This is why a lot of hope has been placed on organs-on-chips (OOCs) as more complex and predictive in vitro models that can better recapitulate the human drug response. It is however a daunting task, and might be a lot to ask from them. Now that OOCs have been around for close to a decade, it is worth asking whether they can live up to expectations, and in particular, whether they can do it alone.

Worth the wait

In 2011, Huh and Ingber baptized organs and chips.3 After publishing their ground-breaking paper describing a lung on a chip, they define organs on chips as “integrat[ing] microfluidics technologies with living cells cultured within 3D devices created with microfabrication techniques from the microchip industry to study human physiology in an organ-specific context, and to develop specialized in vitro disease models”.4 But really, they were years in the making5-7 and were kick-started by the onset of microfluidic cell culture in the late 1990s.8

OOC are the result of more than 100 years of cell culture research. In the early 1900s scientists started manipulating cells outside of the organism for several reasons:
  • To see more easily through a microscope how cells behave and lead to diseases
  • To create in vitro surrogates that you could test drugs on instead of animals/humans
  • To engineer tissues as a means to replace failing organs

OOCs might have never been meant to replace whole organs (since they were designed to remain small) but they have so far delivered on their first promise: they are providing unique insight into the mechanisms underlying pathologies.

The community is off to the races: there is a chip version of more and more individual organs. Cancer has been one of the diseases heavily studied in microfluidic models,9 and studying it in OOCs has revealed key organ-specific insights: for example, Hassell et al. showed in the lung-on-a-chip that cyclic stretching of the lung epithelium suppresses cancer cell proliferation, raising the possibility that loss of lung motion due to filling of the alveolar spaces by growing cancer cells in patients may further help the disease progression.10

Finally, OOCs are at the cusp of becoming streamlined in the drug discovery process, especially now that studies have proved that the tight micro-environmental control in OOCs impacts drug testing: Hassell et al. also show in the lung-on-a-chip that tumor cells become resistant to a drug because of cyclical stretching. Is it then just a matter of time before it is streamlined in the drug discovery process, or could there be other challenges?

Less is not always more

Originally, a key driver for OOC was the desire to design small minimalist assays that recreate one or few key function of organs in vitro. As Cynthia Hadjal, PhD student from MIT, points out: “ the sequential addition of different cell types and physiological flows/pressures allows the user to decouple the effect of each parameter on the system overall. In vitro organ-specific models allow for high spatio-temporal resolution imaging which tends to be difficult to achieve using in vivo models”. This also minimized costs, cell numbers and increased throughput.

However, in its quest for simplicity, OOCs suffer from several limitations explained by Cynthia Hajal: “OOCs cannot fully recapitulate the human physiology since the organ model often lacks several major cell types and is not connected to other organs via a circulatory system as it would be in vivo”. Indeed, we now know that normal tissues are highly heterogeneous and that heterotypic cellular interactions are key to many homeostatic and pathological processes that need to be recapitulated in vitro.

In addition, modeling only one organ at a time fails to capture interactions between organs that are important for drug metabolism (e.g. between the kidney and the liver11). Cynthia Hajal further adds that “the entire immune system found in vivo cannot be fully recapitulated across one OOC even though its importance in modelling disease and organ function has been more emphasized recently”.

Now that multi-organ systems are emerging, scaling OOCs is also needed more than ever to validate their human pharmacokinetic and pharmacodynamic (PK/PD) properties, especially regarding the relative sizes of individual OOCs. Typically miniaturization and alloscaling were chosen, but these are limited for several reasons.12 Instead, using functionality as a scaling metric has shown good agreement with in vivo results13: future studies should also design OOCs as a function of their PK/PD properties from the beginning to ensure valid drug testing.

In addition, OOCs in their current “closed” format might not always be the best logistical option to access cells, media and for modularity. Instead, a recent study has used an open-fluidic system,14 based on the use of well plates connected by microfluidic channels, each of which corresponds to a different organ, and can be switched around easily. This interesting paper poses the interesting following question: at what cost should we miniaturize OOC assays?

A current challenge might therefore be to find the fine line where OOCs' simplicity is maximized without reducing its predictive value. According to Cynthia Hajal, the personalization of OOCs might be key to maximizing their potential: “if we use patient-derived cells inside OOCs, in vivo validation will not provide additional insight into how the patient will respond since the personalized OOCs would expose the full picture and give scientists and physicians the ability to perform full panels of drug testing”. Thus, scientists are aware of the next steps to take to improve the experimental validity of OOCs. But are we aware of the work that needs to be done to improve its analytical power?

Organs on computer chips

Indeed, Professor Martinelli, from the University of Rome Tor Vergata, points out to another interesting limitation of OOCs: lack of high content analysis, to which there is according to him, a clear solution: “artificial intelligence (AI): it will have a fundamental role in the field of OOC in the near future.” This is particularly important with the advent of time-lapse microscopy that is generating huge datasets which can be hard to exploit by biologists.15

These data are crucial to study cell motility, a feature that is key to many processes such as immune cross-talk, or pathologies such as cancer. It can be quite hard to recognize patterns from cell motility, especially in the ever more complex micro-environment recreated in OOCs. As OOCs become more sophisticated, this could apply to data other than images, coming from sensors for example. He explains that “machine learning has the capability to extract and obtain relevant and unexploited information of the high throughput OOCs experiments. In this way, the scaling issue of the OOCs necessary to do reliable and reproducible experiments is partially compensated.” Interestingly, machine learning could also solve the issue of reproducibility and standardization in bringing unbiased automatic analysis, and Deep Learning could even help increase the quality of the time-lapse videos.

Even more interestingly, the use of AI opens “the possibility to control in an automatic and dynamic way several parameters of different experiments and modify their conditions which is something that can help to reduce the gap between the in vivo and ex vivo experiment.” In this way, AI could accelerate OOCs' impact on drug discovery by “serving as feedback for further OOC experiments”,15 helping scientists decide which variables to maintain or not in their model, and minimize experimental trial and error.

In a recent article, Comes et al. recreated entirely virtually cell displacements to adapt the experimental protocol of acquisition and ensure valid conclusion can be withdrawn from OOCs experiments.16 At any rate, what seems to be clear to Prof. Martinelli is that in silico experiments using AI will not dispense the use of in vitro OOCs, but rather merge with it: “I am confident that in the future machine learning will make it possible to simulate in-silico part of the experiments and to reduce the experimental prediction time in a high throughput environment.“

A likely shift in focus

In 2014, Bhatia and Ingber concluded in their review that OOCs were still in their infancy, stating that “researchers in the field must recognize that there are major hurdles to overcome before this technology will have widespread acceptance and impact”.17 Five years later, one could say the field has matured and come to grips with its shortcomings, enough to start seeing beyond its own reach.

On one hand this means OOCs are probably going to evolve faster, by focusing less on miniaturization and throughput and more on multi-scale predictive value. On the other hand, this could push OOCs to seek help from other digital technologies such as AI to become less of a labor-intensive and time-consuming experimental tool. Instead of trying to meet every expectation on its own, OOCs could use this opportunity to become a stronger link in a series of in vivo, in vitro and in silico tools to accelerate drug discovery in a more virtual laboratory.

  1. Scannell J, Blanckley A, Boldon H, Warrington B. Diagnosing the decline in pharmaceutical R&D efficiency. Nat Rev Drug Discov. 2012; 11: p. 191–200
  2. Hay M, Thomas J, Craighead J, Economides C, Rosenthal J. Clinical development success rates for investigational drugs. Nat Biotechnol. 2014; 32: p. 40–51
  3. Huh, D. et al. Reconstituting organ-level lung functions on a chip. Science 328, 1662–1668 (2010).
  4. Huh et al. From 3D cell culture to organs-on-chips. Trends Cell Biol. 2011 Dec;21(12):745-54
  5. Carraro A, et al. In vitro analysis of a hepatic device with intrinsic microvascular-based channels. Biomed Microdevices. 2008;10:795–805
  6. Sin A et al. The design and fabrication of three-chamber microscale cell culture analog devices with integrated dissolved oxygen sensors.Biotechnol Prog. 2004 Jan-Feb;20(1):338-45.
  7. Baudoin R et al. Trends in the development of microfluidic cell biochips for in vitro hepatotoxicity.Toxicol In Vitro. 2007 Jun;21(4):535-44
  8. Folch, A. et al. Molding of Deep Polydimethylsiloxane Microstructures for Microfluidics and Biological Applications. J Biomech Eng. 1999 Feb;121(1):28-34.
  9. Sontheimer-Phelps A et al. Modelling cancer in microfluidic human organs-on-chips. Nat Rev Cancer. 2019 Feb;19(2):65-81
  10. Hassell, B. A. et al. Human Organ Chip Models Recapitulate Orthotopic Lung Cancer Growth, Therapeutic Responses, and Tumor Dormancy In Vitro. Cell Rep. 21, 508–516 (2017).
  11. J. M. Prot et al., First pass intestinal and liver metabolism of paracetamol in a microfluidic platform coupled with a mathematical modeling as a means of evaluating ADME processes in humans. Biotechnol. Bioeng., vol. 111, no. 10, pp. 2027–40, Oct. 2014.
  12. Wikswo, J. P. et al. Scaling and systems biology for integrating multiple organs-on-a-chip. Lab on a Chip 13, 3496–3511
  13. Maas C et al. Multi-functional scaling methodology for translational pharmacokinetic and pharmacodynamic applications using integrated microphysiological systems (MPS).Integr Biol (Camb). 2017 Apr 18;9(4):290-302.
  14. Edington CD et al. Interconnected Microphysiological Systems for Quantitative Biology and Pharmacology Studies.Sci Rep. 2018 Mar 14;8(1):4530.
  15. Mencattini A et al. From Petri Dishes to Organ on Chip Platform: The Increasing Importance of Machine Learning and Image Analysis. Front Pharmacol. 2019 Feb 26;10:100.
  16. Comes et al. The influence of spatial and temporal resolutions on the analysis of cell-cell interaction: a systematic study for time-lapse microscopy applications. Sci Rep. 2019 May 1;9(1):6789.
  17. Bhatia, S. N., & Ingber, D. E. (2014). Microfluidic organs-on-chips. Nature Biotechnology, 32(8), 760–772