Defining the environmental determinants of dysbiosis at scale with zebrafish

Abstract

The gut microbiome, critical to maintaining vertebrate homeostasis, is susceptible to a various exposures. In some cases, these exposures induce dysbiosis, wherein the microbiome changes into a state conducive to disease progression. To better prevent, manage, and treat health disorders, we need to define which exposures induce dysbiosis. Contemporary methods face challenges due to the immense diversity of the exposome and the restricted throughput of conventional experimental tools used for dysbiosis evaluation. We propose integrating high-throughput model systems as an augment to traditional techniques for rapid identification of dysbiosis-inducing agents. Although high-throughput screening tools revolutionized areas such as pharmacology and toxicology, their incorporation in gut microbiome research remains limited. One particularly powerful high-throughput model system is the zebrafish, which affords access to scalable in vivo experimentation involving a complex gut microbiome. Numerous studies have employed this model to identify potential dysbiosis triggers. However, its potential could be further harnessed via innovative study designs, such as evaluation of synergistic effects from combined exposures, expansions to the methodological toolkit to discern causal effects of microbiota, and efforts to assess and improve the translational relevance of the model. Ultimately, this burgeoning experimental resource can accelerate the discovery of agents that underlie dysbiotic disorders.

The gut microbiome is a critical determinant of how we experience many exposures. This community of trillions of microbes that reside in the gastrointestinal tract collectively execute a variety of molecular functions that contribute to our health and homeostasis. For example, gut microbes produce enzymes, compounds, and metabolites that aid in digestion, impact host metabolism, modulate systemic immune state, innervate the central nervous system, and mediate infection¹. But this community is also dynamic, responding to the environmental stimuli it experiences, such as nutrients, drugs, toxicants, and biologics that enter the gastrointestinal tract. In some cases, these exposures are capable of altering the composition or functioning of the gut microbiome such that its contribution to homeostasis is perturbed^2–4. In such situations, the microbiome is said to have entered dysbiosis, which can increase the risk of various diseases and disorders, if not cause them outright. For example, variation in dietary fat content affects the composition of the gut microbiome into a form that metabolically increases bioavailability of dietary lipids to drive cardiometabolic disorders⁵. Additionally, exposure to lead disrupts the microbiome in a way that links to neurotoxicity (and incidentally microbiome manipulation through probiotic administration can alleviate these injuries)⁶. Furthermore, some pesticides that enter the gut can stimulate expression of metabolites by microbiota that increase the risk of autoimmune disorders such as inflammatory bowel diseases^7,8. Determining which exposures affect the microbiome to drive pathologies and how they do so are emergent objectives for exposure science, as they hold potential to transform understanding of exposure induced disease or toxicity mechanisms. Moreover, they may help clarify the interindividual variation in exposure effects given the personalization of the human gut microbiome^9,10.

Our resolution of the exposures that induce dysbiosis, however, remains limited by the complexity of the problem. For example, consider the vast diversity and variation of the exposome. Humans are on average exposed to myriad chemicals each day, and these exposures are highly personalized owing to variation in lifestyle and geographic factors¹¹. Moreover, the parameters of a given exposure, such as its duration, concentration, and developmental timing, may dictate how it effects the microbiome. Furthermore, exposures do not occur in a vacuum, but rather manifest in the presence of diverse mixtures or combinations of exposures, which could elicit additive or synergistic effects, possibly because one exposure affects the transcriptional state of a microbial cell that affects how it then responds to a secondary exposure. Additionally, because the microbiome is also highly personalized, the effects of a particular exposure on the gut microbiome may be quite varied¹². These observations collectively mean that zeroing in on dysbiosis inducing exposures potentially requires consideration of many specific exposures or their combination across a wide array of exposure parameters, as well as the use of large sample sizes to resolve statistical effects. In short, there exists a tremendous search space of potential dysbiotic leads for us to explore.

The impact of this complexity on our ability to advance understanding of dysbiosis is magnified by the contemporary tools typically used to study putative dysbiotic agents. Most investigations of exposure impacts on the microbiome have utilized systems with high translational impact, such as human cohorts of mouse models. While these are certainly critical resources for the community to utilize, they are also expensive to experimentally implement and tend to poorly scale. As a result, studies that seek to define the basis of dysbiosis have typically leaned on prior knowledge to direct their investigation. These efforts wisely considered that there are particular types of exposures, such as clinical administration of antibiotics¹³, large changes in dietary macronutrient composition¹⁴, or well studied cellular toxicants (e.g., heavy metals¹⁵) that are likely to disrupt the gut microbiome given what we know about how these exposures broadly impact microbial growth and metabolism. These investigations will often evaluate a single exposure parameter to determine its impact on microbiome composition, and in some cases will use stool transplantation into germ-free mice to discern causal impacts of the disrupted microbiome on mammalian physiology. This area of research has rapidly grown, and we now posses useful knowledge about how a growing list of exposures affects the microbiome in ways that drive or link to health endpoints, including exposure to plastics like bisphenol A¹⁶, air pollutants¹⁷, nonsteroidal anti-inflammatory drugs¹⁸, proton pump inhibitors¹⁹, cruciferious vegetables²⁰, and ketogenic diets²¹. However, this overarching experimental strategy remains biased in the types of exposures it tends to consider, leaving exploration of the vast landscape of exposure parameters off the table. Moreover, these studies are rarely capable of considering interactions between exposure parameters, typically only looking at a single exposure at a time. While insightful, the total context of exposure can determine how the microbiome responds to a given parameter, as has been observed in studies that modulate heavy metal exposure on diets that differ in their micronutrient content²². Consequently, the insight obtained through single exposure challenges of the microbiome may ultimately be in their translational utility.

Other areas of research, namely pharmacology and toxicology, have battled similar challenges. In these fields, efforts to uncover drug or toxicity leads have benefitted from the use of high-throughput screening approaches. By leveraging relatively inexpensive model systems that offer timely insight into biological effects, researchers in these fields have been able to scale out exploration to rapidly assess the effects of specific chemicals on endpoints of interest. Due to the low cost of screening, this approach enables evaluation of exposures that might otherwise go untested due to limitations in standing knowledge. Where traditional methods of screening chemicals can be slow, time-consuming, and expensive, high-throughput systems enable researchers to test thousands of compounds quickly, efficiently, and cost-effectively. For example, in drug discovery, high-throughput models allow researchers to screen large libraries of compounds for potential therapeutic effects, enabling them to identify promising candidates for further study²³. As a result, this strategy serves to compliment knowledge-based approaches, as these screens can reveal new insight about a poorly understood compounds, or their combination, that propels deeper exploration in more translationally relevant models.

High-throughput screening holds potential to transform the discovery of dysbiotic agents. By adapting high-throughput model systems used in pharmacology and toxicology, or by innovating new models, we can rapidly evaluate how different exposure conditions impact model microbial communities and, depending on the design of the system, whether these impacts relate to health endpoints or biomarkers of interest. In so doing, we may be able to massively expand the diversity of dysbiotic leads that we evaluate and improve the prioritization of leads that we test in translational investigations. There currently exist a variety of high-throughput resources that can serve this objective in microbiome research. These include the use of in vitro, ex vivo, and in vivo model systems, as well as data mining approaches. For example, several studies have used in vitro models to evaluate how specific exposures affect the growth and metabolism of diverse human gut microbiota isolates as well as combinations of these isolates, often under growth conditions that simulate the human gut^24–26. Some of these tools incorporate host tissue culture, often using microfluidic technologies (e.g., organ-on-a-chip), to improve inference about how the interaction between an exposure and a cultured isolate impacts host cellular physiology^27,28. However, a limitation of these approaches is their reliance on in vitro assays, which do not necessarily represent the complexity of host physiology and its contribution to exposure outcomes, Additionally, while the cultured isolates or consortia used in these studies represent the gut microbiome, they rarely model its complexity, which could impact exposure outcomes due to ecological or metabolic interactions across taxa. Ex vivo models typically incubate human stool samples in growth media to clarify how the native gut community responds to specific exposures. These approaches range from relatively naïve ‘poop soup’ studies, which have for example clarified how stool communities respond to changes in dietary composition²⁹, to the use of technologies that apply media and growth conditions that simulate the human gut, including minibioreactors³⁰. While these approaches model complex microbiome found in humans, they tend to do so in vitro and without the context of host physiology, which may affect the microbiome’s composition and functional response to exposure. Low-cost in vivo model systems afford opportunities to evaluate how exposures directly impact the microbiome in the context of host health. In particular, several studies have used the large sample sizes and germ-free derivation approaches inherent to various invertebrate models to disentangle exposure, the microbiome, and host physiological interactions. For example, an investigation using flies (Drosophila) found that the microbiome plays a role in defining atrazine toxicity³¹, while a study using worms (C. elegans) defined the microbiome’s role in cancer chemotherapy impacts on the host³². While powerful, these invertebrate models tend to carry relatively simplified microbial communities as compared to their vertebrate counterparts, and often exhibit significant differences in physiology as compared to humans, which limits their translational impact. An alternative strategy relies on data analytical methods, which can advance lead generation by integrating diverse exposome measures (e.g., untargeted metabolomic data) with microbiome and health outcomes across a human cohort to resolve specific exposures that associate with health endpoints³³. While such approaches are useful for resolving translationally relevant dysbiotic agents, they can suffer from confounding factors due to their reliance on inferential approaches. Collectively, these resources are useful tools that should be increasingly relied upon to resolve the mechanisms of dysbiosis.

One additional high-throughput model system that has rapidly grown as a resource for resolving dysbiotic agents is zebrafish. Unlike the aforementioned systems, zebrafish affords access to a highly scalable in vivo model of the gut microbiome that incorporates representation of vertebrate physiology as well as a complex microbial community. As a result, this model holds potential to serve as a key intermediary step in the translational pipeline between the models mentioned above and less scalable mammalian systems. Owing in part to their small size, high fecundity, and genetic tractability, zebrafish have catalyzed discovery of myriad toxicants and drugs (e.g.,³⁴ ). Moreover, zebrafish embryos are transparent and develop rapidly, allowing for non-invasive monitoring of developmental processes in real time. In addition, zebrafish share many physiological and genetic similarities with mammals, including humans, making them a relevant model for studying chemical effects on vertebrates. There also exist a rich array of automated technologies for phenotyping zebrafish, including a battery of behavioral assays³⁵, which mitigates major bottlenecks often experienced in other in vivo experimental systems. These technologies become particularly powerful when combined with robotic platforms that deposit embryos across 96-well plates, wherein individual embryos can be subjected to precisely controlled exposures conditions and develop into larvae in isolation. Fish can also be grown out in tanks with members of their cohort to evaluate how exposure at various points along their lifespan affects their maturation, survival, development, and behavior, even across generations³⁶. Recent advances in CRISPR/CAS technology in zebrafish also facilitate discovery of the genetic determinants of exposure outcomes³⁷.

In recent years, microbiome researchers have advanced the resources available in zebrafish to result in a powerful vertebrate model system for understanding how exposure impacts host-microbiome interactions³⁸. Zebrafish can be inexpensively reared germ-free, at least as larvae, and colonized with specific microbiota to define the role gut microbes play in modulating exposure and host health³⁹. Additionally, in adult fish, stool samples can be collected using passive sampling techniques to enable longitudinal studies of the gut microbiome⁴⁰. Moreover, low-cost experimental strategies for characterizing the larval zebrafish microbiome in high-throughput screening studies have been developed⁴¹. Recent analyses also demonstrate that zebrafish carry a phylogenetic core gut microbiome that is robust to experimental and facility level variation⁴². These and related resources have been used determine that this microbiome drives fish development, physiology, immunity, and behavior^43,44. Recent studies have even mapped out how the gut microbiome effects the intestinal cellular composition of the developing zebrafish gut⁴⁵.

Zebrafish have increasingly been used to resolve potential agents of dysbiosis, especially in the form of pollutants, drugs, and nutrients. For example, Gaulke et al. found that oral triclosan exposure significantly altered gut microbiome composition in adult zebrafish⁴⁶. Pindling et al. showed that chronic exposure to subclinical concentrations of antibiotics by larval zebrafish disrupts microbiome assembly and is linked to increased larval mortality⁴⁷. Catron et al showed that developmental exposure to bisphenol analogs can induce developmental toxicity and disrupt microbiome assembly in a concentration dependent manner⁴⁸. Recent years have built upon this prior body or research to result in an explosion of research in the use of zebrafish to understand how different environmental and drug chemical exposures impact the gut microbiome, including the evaluation of various forms of exposures such as water soluble crude-oil components⁴⁹ , methylparaben⁵⁰, amitriptyline⁵¹, phthalates⁵⁰, MPTP⁵², polychlorinated biphenyls⁵³, and enrofloxacin⁵⁴. This experimental model system is rapidly advancing our understanding of the exposure-microbiome-health relationship.

But despite the rapid growth of zebrafish as a resource for resolving dysbiotic agents, there exist several areas where innovations in study design or experimental methodology are need to advance discovery. For example, little work in zebrafish has considered how their microbiomes respond to combined or mixed exposures, despite the fact that such mixtures are increasingly explored in traditional zebrafish toxicity assays^55,56. These types of study designs could help clarify how the background chemical context of a given exposure affects the microbiome and whether these effects inherently impact exposure outcomes. Additionally, one of the benefits of the zebrafish model system is its relatively rapid growth rate, a feature that microbiome dysbiosis investigations have not yet fully capitalized upon. Growing out and longitudinally monitoring the microbiomes of zebrafish that were embryonically or developmentally exposed to specific agents can clarify the long-term impact that particular exposures have on microbiome succession and whether these impacts drive health outcomes later in life. Relatedly, the short generation times inherent to the zebrafish model afford an opportunity to consider the intergenerational effects of exposure on the gut microbiome. While zebrafish are subject to such considerations in terms of chemical toxicity assays^36,57, little work has considered the microbiome’s role in such relationships. Furthermore, experimental tools can be expanded or adapted into the zebrafish microbiome research toolkit to improve understanding of the mechanisms and outcomes of dysbiosis. For example, while germ-free fish can be inexpensively reared³⁹, researchers currently struggle to maintain zebrafish germ-free past ~14 days post fertilization. This limitation challenges assessment of the functional impact of the microbiome in later stages of the fish’s lifespan. Tools to robustly deplete the microbiome antibiotically or maintain germ-free fish later into life are needed advance exploration in this regard. Transplantation of adult fish stool into germ-free larval fish may be an additional experimental resource whose innovation can advance these efforts. Additionally, incorporating CRISPR/CAS technologies into zebrafish studies can prove to be a useful resource for understanding how specific host genes contribute to an underlying dysbiosis. These technologies are relatively easily utilized in zebrafish studies⁵⁸, and accordingly advanced toxicological investigations⁵⁹, but have only seen limited application in microbiome investigations⁶⁰. Finally, efforts to expand the zebrafish microbiome culture collection and colonize fish with specific microbiota or consortia can help clarify how specific microbiota respond to specific exposures and possibly lead to clarification of the microbial genetic mechanisms that underlie these their sensitivity to exposure.

Future innovation of the zebrafish system for microbiome research needs to also focus on improving the translational relevance of the model. While zebrafish exhibit useful genetic and chemical homology to humans, zebrafish vary from humans in several key areas that could impact the translational relevance of discoveries about the zebrafish microbiome. For example, zebrafish contain unique genetic paralogs of human genes (e.g., the AHR genes⁶¹ ), the pH and electrolyte composition of the zebrafish gut lumen differs from humans, and the differential tissue morphology along the length of the gastrointestinal tract in zebrafish varies from that of humans⁶². Based on culture collections, the zebrafish gut microbiota appear to be more aerotolerant than mammalian gut microbes, possibly reflecting a difference in the aerobic capacity of the gastrointestinal tract⁶³, and these two host species consume different nutritional regimes. Furthermore, the cold-blooded nature of fish means their cardiometabolic states and internal temperatures do necessarily reflect those of their warm-blooded mammalian counterparts. These lineage-specific differences in host biology could result in host-microbiome interactions that are also unique to each lineage. Moreover, in accordance with these observations – and the fact that the ecology of zebrafish and their microbiota is aquatic – the types of microbial taxa that colonize the zebrafish gut tend to be distinct from those that colonize human guts (though many taxonomic groups are consistent). Efforts to define the phylogenetic, genetic, and functional homology between zebrafish and human microbiomes will help clarify the translational relevance of this model system. Moreover, recent work to colonize zebrafish with mammalian microbes indicate that microbiome humanization of zebrafish may be possible⁶⁴, and work should build upon these findings to determine if specific zebrafish strains or growth conditions improve the ability to represent specific human microbiota within the zebrafish gut. But ultimately, as with all models, we need to define how well leads generated in the zebrafish model system translate to mammals and humans through follow-up investigations in these other host systems, something that has to date been sorely lacking in the literature.

In summary, high-throughput model systems offer a powerful means of efficiently and ethically screening for exposures that interact with microbiota to potentially drive health outcomes. These efforts compliment traditional experimentation in translationally relevant models: when such screens reveal new dysbiotic leads, they can be prioritized in studies that use pre-clinical models or involve human cohorts. In so doing, we can more rapidly pinpoint the underlying exposure agents and parameters that drive dysbiosis, which can help advance new microbiome-derived tools for monitoring, preventing, and possibly even treating exposure-induced disorders. While we think zebrafish plays a key role in this experimental pipeline, there are myriad resources currently available to researchers looking to conduct high-throughput investigations of dysbiosis. Going forward, the community should work to innovate study designs that fully capitalize on the throughput afforded by these models (e.g., text mixtures), expand the toolkit available to discern the health consequences of exposure induced changes to the microbiome, and collaborate with other investigators the define the translational relevance of leads identified through these investigations.

Figure 1 –

An overview of the strategies for resolving dysbiotic agents. A) A schematic illustrating the role that dysbiotic agents, which may be genes or various environmental factors (the focus of this narrative), play in disrupting microbiome functioning to drive health impairment. Note that the effect of these agents on the gut microbiome may depend on their developmental timing. B) The major strategies for resolving dysbiotic agents, as well as their advantages and disadvantages. These strategies are not mutually exclusive and various methodologies lay at different points along the continuum of the strategies listed here.

Figure 2 –

An experimental workflow that illustrates how zebrafish can be utilized to screen for dysbiotic agents. Exposures can be precisely administrated a various stages throughout zebrafish development and the effect of these exposures can be rapidly assessed through automated phenotyping technologies and microbiome sequencing approaches. Exposed and control fish can also be subject to spawning for studies intended to evaluate intergenerational effects. Not shown here is the additional technique of germ free derivation of larval fish, which can be utilized to determine the microbiome’s role in defining the effects of specific exposure outcomes.

Highlights.

Knowing which environmental factors perturb the microbiome to induce dysbiosis can help advance our understanding of the origins of diseases and disorders.

The diversity in human exposure types, combinations, concentrations, and durations poses a complex combinatorics problem in the search for dysbiotic agents.

Traditional, low-throughput methods to identify dysbiotic agents often incur strong prior bias. High-throughput screening tools offer a relatively unbiased approach to exposome exploration, which can unlock novel dysbiotic leads.

The zebrafish model system, with its experimental resources and access to vertebrate genetics, physiology, and a complex gut microbiome, offers a pivotal resource for large-scale discovery of dysbiotic agents.

Augmenting the utility and translational relevance of the zebrafish model system in dysbiotic agent and mechanism discovery could revolutionize our ability to discern the causes of microbiomic disorders.