The Uphill Development for Organs on a Chip in Drug Discovery

Over the last five years, we have watched the advanced cell culture model space mature dramatically as it has gone from primarily an academic pursuit to an area of research that almost every pharmaceutical and biotech company in the world is engaged in. At first, these models appeared as spheroids and then organoids and now we are pushing into the space of organs on a chip and bodies on a chip. The rationale for this rapid change and adoption is that advanced cell culture models have shown a tremendous ability to fail therapeutics faster during the drug discovery and development process far before they enter into expensive animal or human studies. They are able to accomplish this feat and also address many basic research questions simply because they can better recapitulate the in vivo microenvironment as the more relevant structure within these in vitro models tends to lead to more relevant function.  

While some advanced models have shown the ability to better predict in vivo outcomes compared to traditional more simplistic models, this has unfortunately led to the axiom that more complicated in vitro models are always better. In chatting with various experts and founders in the space, it appears that this is fait accompli but the reality is far more complicated for a number of reasons: 

Efficacy of Organs on a Chip

When I talk to companies and researchers in the advanced cell culture space, my first question about their model is always, "what is it better than?" For many, they believe the question is subjective and they say that their model is better than the current solution because of Feature A or Feature B which for example are longer culture time, more relevant cell populations, improved sensitivity, etc. I then typically repeat the question, but "what is better about it?" and they repeat their answer. I then have to explain that even though their model has many features that make it more like the in vivo reality, none of these necessarily mean that their model is better than the current solution and they could actually mean that it is much worse. This is a really important distinction which I think scientists are not trained well to understand as they tend to assume features mean value. The NIH and NSF even developed their I-Corps program to help with addressing this problem and misconception. As researchers, we are trained to make small line-extension improvements on research thrusts to acquire publications and we assume these are of course intrinsically valuable.  

A model needs to better predict in vivo outcomes, be more inexpensive than the current approach, address a question that cannot be answered today or some combination of these. Therefore, a researcher needs to demonstrate for example that their model's improved culture time (i.e., a feature) results in better evaluating slowly metabolized compounds and their metabolites like is the case with the Visikol HUREL™ micro liver models. Otherwise, the model simply does not have efficacy in helping with improving the drug discovery and development process compared to the current standard. In the case of microfluidic systems, it is simply not true that adding a microfluidic component improves all systems for all applications and in fact quite the contrary is likely true. There are researchers that have shown of course the fluidics alters phenotype but this specific benefit for each application needs to be evaluated. It's critical to prove features add value and not to assume this is true which many in this space have done. 

The Increasing Complexity of Organs on a Chip Makes Validation Challenging 

Anyone who has developed an in vitro assay knows how challenging it is to develop a robust assay with good translation to in vivo data. Further, anyone who has done this knows how challenging it is to have reproducibly in in vitro assays which perform within the same tight acceptance criteria over time which is required for any successful program. One of the key challenges inherent with microfluidic systems is that you are taking an in vitro model and adding a mechanical and physical component as well as a complex geometry in which the cells need to thrive. This additional complexity means multiple new degrees of freedom for the assay and statistically a more challenging time in achieving reproducibility and a tight assay window. While this may be technically feasible for some models, it will add substantial cost to the development, manufacturing and validation of these systems. This problem is further intensified when researchers start discussing multi-organ systems in which many different types of in vitro microfluidic systems need to be brought together at precisely the right time. While not technically impossible, complexity is not always a good thing in developing an assay and adding these degrees of freedom greatly increases cost and the ease-of-validation.  

The Cost of Organs on a Chip

For any cell culture system as mentioned in my prior article on "The Financial Paradox of Developing an Advanced Cell Culture Model," cost is the main driver of whether the system will be successful or not. On one end of the spectrum of cost for drug discovery assays to evaluate a therapeutic are clinical subjects and on the other end are enzymatic assays and even in silico work. Ideally, we want to achieve clinical-like relevance for in silico cost but of course this is not necessarily possible. Therefore, as we progress through the drug discovery pipeline, we are continually balancing cost and relevancy where the idealized goal of microfluidic cell culture models is to be close to clinical models in relevancy but to have a reasonably low price. The problem with this is that a microfluidic model will always cost more than a non-microfluidic model per N and likely more than many simpler animal models per N. To have a true place in this market, these microfluidic models will have to comprehensively demonstrate that they are more predictive of in vivo results than current animal models which is a very high technical bar to clear given the decades of animal data. Otherwise, these models fall between advanced 3D cell culture models and animal models in a small niche market. In this market position, the market for the developed application likely would not justify the cost of developing the microfluidic model in the first place and thus the microfluidic platform would not be financially viable as a business proposition. The pie in the sky aspiration of course for microfluidic models is that they will supplant animal models, but it is more likely that they will only be developed for a small subset of applications and be a tool in the advanced cell culture space albeit a small and seldom used one.  

Naivety

I find it disingenuous when scientists seriously talk about their body-on-a-chip or organ-on-a-chip models unless they caveat the description with the fact that these are merely marketing terms and not of course depictions of reality. The reason for this is that how can we possibly recapitulate a system which we don’t fully understand in its entirety nor do we understand the complex interplay between the physical, chemical and mechanical properties that make an organ an organ? A ball of cells within a tube carrying fluid is little more an organ than a ball of cells in a plate (i.e., a spheroid) being swashed around in solution. I find the more proper term for these systems to be microfluidic cell culture models or microphysiological systems which distinctly and accurately describes their nature. I see that we do a big disservice to this space if we inaccurately describe what we are doing as these models are certainly today not close to organs or bodies on a chip.

So Where Do Microfluidics Cell Culture Models Fit In?

These microfluidic cell culture models will by their nature be more expensive than their more simplistic counterparts and thus they are relegated to research questions that these more simplistic models cannot effectively evaluate. This means that their place within the drug discovery and development process will be small and will only be for a select few target applications. The antithesis to this line of thinking is that a research group develops an approach to quickly, easily and inexpensively combine multiple organ models together and the resulting creation is more relevant than simple animal models. From my perspective, thinking we can recapitulate the body on a chip in such a way in the short term is naïve and while we will make progress in this direction as a research community in the coming years, routine and widespread use of microfluidic models in drug discovery is a long way away. Further, building a viable business around this concept is extremely challenging due to the high technical bar of replacing animal models and the paradox described in my prior article. However, I hope I am wrong because a body on a chip sure would disrupt the entire drug discovery and development field!