Are we Rushing too much?

Editorial:

There is a great push in the research arena to reduce, refine and replace animals with an array of alternative techniques, such as computer modeling, artificial tissue models, 3D bioprinting, and alike. Indeed, a worthy cause, but we must ask, does it jeopardize human health, and can we wholly rely on such methods to assess risk associated with therapeutics or environmental exposures?

As early as 1,000 years ago Ibn Sina¹, a Persian polymath who is regarded as one of the most significant physicians, astronomers, writers and thinkers of the Islamic Golden Age, and the father of early modern medicine, advocated for human studies over animals. Alexander Pope², one of the foremost English poets of the eighteenth century echoed with the well-known maxim “The proper study of mankind is man”.²

Consider that on many occasions, even whole animal experimentations have been poor predictors of human responses to environmental exposures or drugs. Take isotretinoin, more commonly known as acutane. In rabbits and monkeys, as well as in humans, but not in mice or rats, it causes birth defects³. Take corticosteroids! They are not teratogenic in humans, but are in experimental animals⁴. And how about thalidomide? It is a teratogen in humans but not in many experimental animal species⁵. Other examples abound!

There are many reasons why animal studies are poor predictors of human outcomes, and why at times they fail to translate to human responses. Few reasons include the following:

1. The studies are poorly designed (length of experiments, dose, methods of randomization, distinctions in laboratory techniques) and methodologically inadequate;

2. The studies are not replicated and are rarely subjected to meta analyses;

3. The tested species or strains differ in metabolic pathways and drug metabolism from those in humans; and

4. Disease manifestation in the animals are distinct from those encountered in humans; to name a few.

Many of the above caveats are inherent to the newly developed methods as well.

Take for example computational models. The enormous growth of statistical software packages and chemical descriptors clearly permits the development of new models. However, safeguards that fully account for the pitfalls of the underlying features and models may be lacking. As recently noted by Luechtefeld and Hartung (2017)⁶, this may lead to a misrepresentation of the accuracy of these models, and by default, the conclusions on safety or toxicity.

Let’s consider another example, namely, printed neurons. They have already been utilized to study brain cortical patterning and traumatic brain injury, to name a few. One could argue they would fail to predict the neurotoxicity of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), a model drug that has been commonly used to elicit Parkinson’s disease in experimental animals. Why? Because, the toxicity of MPTP is related to the generation of its active metabolite 1-methyl-4-phenylpyridinium (MPP+), a process catalyzed by monoamine oxidase-B⁷. This takes place largely in astrocytes, a glial ell type. Hence, a 3D printed neuronal layer would fail to predict the neurotoxicity of MPTP, given the absence of astrocytes. Even in the presence of 3D-printed astrocytes the neurons will still likely not succumb to MPTP toxicity, since MPTP and its metabolite MPP are selective for dopaminergic cells (not cortical neurons). You get the point! Since the brain contains billions of neurons across multiple scales, not to mention a plethora of other specialized cells (astrocytes, microglia, oligodendrocyte, endothelia, pericytes) in order to have any confidence in the testing results, especially when testing unknown compounds, one would have to generate thousands (perhaps even more) permutations of 3D-printed cellular co-culture models to conclude with confidence a given compound does not represent a human health risk.

In conclusion, one should applaud the exciting and new developments, yet be mindful of their limitations that severely diminish their utility in the assessment on causal relationships between human exposure to a substance or drug tested and the subsequent clinical condition or outcome. Due to the artificiality of the inherent environment, the response of tissues or cells, as well as the pitfalls inherent computer modeling in assessing exposure outcomes may often differ from the response in an intact organism. While potentially informative on mechanisms, transport, and other parameters, discussion on human risk vis-à-vis these novel approaches seems at this point non sequitur.