The conundrum with animal models of the human microbiome
Posted 15th March 2018 by Jane Williams
‘The most that can be expected from any model is that it can supply a useful approximation to reality: All models are wrong; some models are useful.’
This aphorism, simply articulated here by prominent British statistician George Box, gets to the crux of the dilemma we face when considering animal models for the study of human microbiome dynamics and mechanisms.
On the one hand we have a pretty good handle on how to manipulate particular pathways and phenotypes in rodent, other non-primate, and non-human primate animal models to recreate human disease-associated genotypes and/or phenotypes.
On the other hand, we have to contend with the fact that by most estimates, the human component of a human is only ~10 to 50 percent of the whole, and that this percent plays host to the 100 trillion or so microbes that make up the microbiome and complete the picture of what makes a human human.
Against this backdrop, we face the challenge of generating animal models that recreate the complex interactions between the microbiome and its host’s pathologies in ways that are translationally relevant.
“Can anything other than a human recapitulate the complexity…?” “Do we need disease models…?” “Move directly into humans and forget animal models!”
These are comments we have heard from some in academia, biotech, pharma, and the investment community during the lead-up to Microbiome Futures. But many also believe animal models of human disease still play an important role in microbiome research.
“Animal models remain relevant for performing reductionist studies that enable dissection of biological mechanisms and allow for sufficiently powered studies under controlled conditions,” says Alexander Maue, Director of Microbiome Products and Services at Taconic Biosciences. Taconic is a global provider of genetically engineered rodent models and services based in Albany, NY.
“The key to making further progress through the use of animal models is not to reject them but to understand more precisely what their limitations are, and how to make these findings translatable to humans,” says José Clemente, assistant professor at Mount Sinai Icahn School of Medicine in New York.
Undoubtedly, animal models of disease generated through human microbiome transplantation – currently the closest we can get to recreating the complexity of the human microbiome in an animal – have led to valuable insights in our understanding of how the microbiome interacts with its host.
However, while such models have shown the key role the microbiome can play in conditions such as obesity and inflammatory bowel disease, identifying actual mechanisms and microbial biomarkers that translate from the model animal to humans has proven elusive.
Genetic, physiological, and anatomical differences between humans and animals affect engraftment, naturally favoring those microbiome phylotypes best adapted to survive in the new host. More importantly, the host’s immune system, which sits at the host-microbiome interface and channels signals from the microbiome to direct the host’s response to environmental cues and to, in turn, regulate the microbiome composition and dynamics by the host, varies greatly among species.
Human microbiomes have been transplanted into both conventionally raised animals and germ-free animals. With conventionally raised animals, the purpose is to displace and hopefully replace the host’s microbiome with the transplanted human microbiome in a ‘natural and gradual’ way. With germ-free animals, the goal is to create gnotobiotic animals displaying the human microbiome only.
In both instances, however, the transplanted human microbiome needs to adapt to a set of host-specific physiological and topological niches and an immune system it has not co-evolved with, and it needs to engraft in an environment where the host’s immune system has not developed from birth with a human microbiome, or any microbiome in the case of germ-free animals.
Two ways to start addressing some of the above issues are “to use animals with humanized immune systems, and to perform experiments on the offspring of transplanted parental animals that would have co-developed their immune systems with the transplanted microbiome – the diversity of the transplanted microbiome has been shown to be conserved in the F1 generation,” says Taconic’s Maue.
Beyond co-evolution and co-development, physiological and topological niche disparities could further be addressed by moving up from rodents, on the evolutionary scale, and enlist non-primate animal models such as pigs or even non-human primates with immune systems, physiologies and anatomical characteristics closer to humans. Practically, however, such efforts are hampered by the complexity and expense of performing experiments with large animals.
But in spite of all these approximations, it will remain challenging to determine whether a dysbiotic state recreated in any animal model is truly representative of the human pathology. In most cases, both the transplanted dysbiotic microbiome and the transplanted control or healthy microbiome experience ecological shifts upon engraftment. The resulting animals may still present compositional dissimilarities in their microbiomes between the healthy and the dysbiotic state, but these may not necessarily be representative of the original pathological state and simply reflect new ecological patterns of the engrafted microbiomes.
And last but not least, “there is an intrinsic bias when speaking of ‘microbiome’ that naturally pivots discussion towards gut-centric models,” says Taconic’s Maue. “It is maybe not a challenge per se, but it has resulted in a paucity of models, if any, that capture different microbiome sites (e.g. lung, gut, skin), let alone integrate them into one singular translational animal model.”
A lack of good translational animal models that mirror the complexity of the human microbiome makes animal studies less informative and is driving exploration and R&D into early human prototyping, and thus, says Christian Grøndahl, “better models will likely not be the breakthrough, but rather early clinical testing and identification of pivotal culprit bacteria, or pivotal beneficial bacteria and their molecular signals, directly in humans.” Grøndahl is CEO of Copenhagen-based microbiome play SNIPR Biome and Director at Oxford-based BioMe Oxford, a startup developing technologies for non-invasive microbiome sampling in humans.
Three studies recently published in Science illustrate this point: Routy et al., Matson et al., and Gopalakrishnan et al. used a combination of microbiome profiling in patients and experiments with gnotobiotic mice to pinpoint specific gut bacteria that influence patient response to anti–PD-1–based cancer therapy.
What breakthroughs will have to happen in terms of model design to make animal models of disease ever more translationally relevant, and conversely, what technical advances will have to happen for microbiome studies to safely take place directly in humans, are questions that remain unanswered and are pivotal for microbiome translation. We are looking forward to discussing them further at Microbiome Futures later in May this year.
Gaspar Taroncher-Oldenburg is Consultant-in-Residence for Global Engage. He was previously Founding and Managing Editor of Nature’s SciBX: Science-Business eXchange (now BioCentury Innovations) and scientific editor of Nature Biotechnology.
Leave a Reply