The Microbiome

Abstract

The microbiome is increasingly recognized as a critical component in human development, health, and disease. Its relevance to toxicology and pharmacology involves challenges to current concepts related to absorption, metabolism, gene:environment, and pathways of response. Framing testable hypotheses for experimental and epidemiological studies will require attention to study designs, biosampling, data analysis, and attention to confounders.

Recent research as well as designated funding, including the establishment of the Human Microbiome Project within the National Institutes of Health, has generated considerable attention to the importance of the microbiome for many aspects of human health and development (Turnbaugh et al. 2007). Among its newly explored roles in human biology, the microbiome may be an important modulator of host response to xenobiotics, including both toxicants and drugs. When the microbiome is located on the external side of the interface between the environment and portals of entry, such as the skin, gut, and lung, there are 2 critical roles for site-specific microbiomes: as a gatekeeper and as a watchman (Dietert and Silbergeld 2015). As gatekeeper, the microbiome is at the physiological front line of exposure prior to absorption by inhalation, ingestion, or dermal contact. In this space, the microbiome can function as “first pass” metabolism, prior to absorption, and thereby influence the nature and amount of agents that will be absorbed and distributed to the host’s internal physiological and cellular systems (Klaassen and Cui 2015; Klunemann, Schmid, and Patil 2014).

The microbiome also functions as a watchman at these portals of entry, generating signals that are transduced to distal organs and organ systems through cross talk between the circulating immune system and the endothelial cells that form the physiological barriers of skin, gut, and lung (Hooper, Littman, and Macpherson 2012). As a watchman, the microbiome also responds to signals from the host that are transduced through the immune system (Dishaw et al. 2014; Hooper, Littman, and Macpherson 2012; Mortha et al. 2014). Thus, the role of the microbiome is a two-way street, sending and receiving signals to and from the host. Additionally, the microbiome may be directly affected by xenobiotic agents not limited to those with antibiotic properties (Lu, Abo, et al. 2014).

The connections between host and microbiome are complex. The host genome can influence the microbiome in terms of phenotypes and function (Donovan et al. 2014; Lu, Mahbub, et al. 2014; Khachatryan et al. 2008). These interactions support the hypothesis that including the microbiome in toxicity studies is likely to increase our fundamental understanding of the molecular and pathological pathways of toxicity and our ability to predict organ and organism-level responses to nanomaterials and other agents of concern.

Needed: New Definitions of “ADME”

The concept of ADME—absorption, distribution, metabolism, and excretion—is fundamental to pharmacology and toxicology. ADME is based upon the assumption that these critical events occur in a physiological sequence following absorption and that absorption depends upon chemical and biophysical characteristics of an agent encountered at the boundary between external and internal milieu. This paradigm has enabled extensive research and publications modeling structural aspects that influence absorption as well as distribution within the body, which contributes to the evaluation of new molecules by the pharmaceutical and chemical industry (Wang and Hou 2015). None of this work includes the microbiome.

The Microbiome—Absorption and Metabolism

By occupying the physiological front lines of exposure prior to absorption by inhalation, ingestion, or dermal contact, the microbiome of the respiratory tract, the gut, or the skin can act as first pass metabolism and thereby influence the nature and amount of the actual absorbed dose of an agent that is delivered to internal physiological and cellular systems. The gut microbiome possesses many, if not of all, of the genes present in the human genome that encode enzymes that influence toxicity through methylation of metals, activation of procarcinogens, and other pathways relevant to the potential health risks of environmental and occupational exposures as well as drug metabolism (e.g., Kruger et al. 2013). This was dramatically demonstrated by Van de Wiele et al (2010) in studies exposing an explanted human gut microbiome to inorganic arsenic and assessing the efficiency of arsenic methylation by the microbiome. This study effectively challenged the prevailing assumptions by toxicologists and epidemiologists that arsenic metabolism occurs within human cells and organs only after absorption and that host genetics drive interindividual differences in arsenic metabolism. This group has since developed a complex staged simulation of the entire gut microbiome and the host landscape as represented by a mucosal layer and gut epithelial cells, which could be used in in vitro toxicological studies in a manner similar to the addition of liver homogenates to generate metabolites in the Ames test of genotoxicity (Marzorati et al. 2014).

The Microbiome and Gene:Environment Interactions

It is now well understood, since exploding the “black box” of epidemiology first proposed in 1983, that there are a number of variables that influence individual and population responses to xenobiotics (National Research Council 1987). These are defined as inherent factors mostly related to genes, that is, genes in the human genome. This concept needs to be reexploded by adding the microbiome to the exposure/effect continuum (Dietert and Silbergeld 2015). Once we let the microbiome into the picture, we must consider the significance of the microbial metagenome as compared to the human genome (Olsen, Choffner, and Mack 2012). Not only is this metagenome vastly larger than the human genome, it changes as the microbiome changes over development and in response to diet, drugs, infections, and chemical agents (Lu et al. 2013). These changes in the microbial metagenome may thus introduce high degree of dynamics to gene:environment interactions.

The Microbiome and Pathways of Response

The microbiome is itself a target for the action of xenobiotics, including drugs and other agents, changing the composition of the microbiome as well as its contribution to metabolomics. There are several studies in the literature reporting on changes in the microbiome after exposures to antibiotics, nanomaterials, and environmental toxicants including metals and pesticides. However, the hypothesis that the microbiome could be a watchman in addition to a gatekeeper—that is, that it could be a part of the domain of adverse outcome or toxicity pathways, and thereby modulate host responses to toxicant exposure—has received less attention to date. Few published studies have tested signal transduction from the microbiome to target organs or from target organs to the microbiome and the associations between these signals and system toxicity or response. The rationale for this hypothesis is nevertheless strongly supported by studies in the clinical literature, in which alterations in the gut microbiome have been reported to affect health status related to a range of diseases including diabetes, obesity, neuropsychiatric conditions, liver, and kidney disease (Shreiner, Kao, and Young 2015). The status of the gut influences the metabolome (Wang, Ehrlich, and Pedersen 2013). This correlation was demonstrated in a study of mice exposed to arsenic (Lu, Abo, et al. 2014). Another study of mice exposed to the pesticide chlorpyrifos reported alterations in the relative prevalence of major bacterial families in the gut and also in the pattern of metabolomics measured in urine (Zhao et al. 2016). Importantly, this latter study noted associations between alterations in metabolomics signatures and the health status of the host.

It Won’t Be Easy: Putting the Microbiome into Toxicology and Pharmacology

Introducing the microbiome into toxicology and pharmacology is new to research in toxicology and epidemiology and will require considerable work on developing focused hypotheses, study designs that can test these hypotheses, and molecular and informatics methods to generate and represent information from these studies. It is likely that there are “critical windows” in the development of the microbiome, as there are for other complex systems of the human body such as the immune and nervous systems. The gut microbiome is transferred from mother to infant by vertical transfer during vaginal delivery and then undergoes a prolonged period of postnatal development, during which, like other systems, the microbiome may be particularly sensitive to xenobiotic influence (Dietert 2014; Planer et al. 2016).

Our current research has been informed by the questions below.

What Is the Microbiome?

There are at least 3 levels of definition that can be applied to “microbiome.” The most parsimonious is that used in most biomedical research: the microbial organisms (usually restricted to bacteria) located in or on a defined anatomic space or surface. The second definition includes the energetics and metabolism—the inputs and outputs of a specific microbial community. The third definition adopts the perspective of microbial ecology to include the host and microbiome, through the ecological concept of the microbiome as a spatiotemporally defined microbial community existing within and communicating with a physiologically defined “landscape” (an ecological niche or a physiological niche such as the gut; Olsen, Choffner, and Mack 2012; Hooper, Littman, and Macpherson 2012; Shade et al. 2013; Dietert and Silbergeld 2015; Martin et al. 2007). This perspective is the most appropriate to the goals of research in toxicology and pharmacology since we seek to connect events in the microbiome with host biology. Adopting the ecological perspective also enables us to draw upon the range of robust statistical methods that have been developed in microbial ecology to support inferences and hypothesis testing based on metagenomics data (Schloss and Handelsman 2008). For example, this expands the analytic repertoire beyond principal components analyses that are widely used in studies of the microbiome and toxicant exposures (such as Lu et al. 2013; Zhao et al. 2016).

What and Which Is the Microbiome That We Are Studying?

We utilize terms such as “gut” or “skin” or “nasopharyngeal” to denote a microbiome of interest. However, these anatomic notations are misleading because they cover enormous landscapes within which a range of microbiomes are located with considerable differences among them. The gut microbiome, which is most often the object of study, consists of many microbiomes from the esophagus to the rectum. Different dermal regions—axilla, genital, arm, and so on—are highly varied in terms of microbial communities (Turnbaugh et al. 2007). We need to define the microbiomes we are sampling with more precision. In addition, some of our methods introduce undefinable heterogeneity in the sampled microbiome. For example, the “gut microbiome” is usually sampled in terms of excreted feces or rectal swabs. This is substantially different from directly sampling the microbiome of the stomach or the intestines, and excreta are likely to represent undefined and potentially variable biosamples. The ideal study would incorporate the complex spatial organization of the gut and the gut microbiome (Earle et al. 2015), as is done in the in vitro system described by Marzorati et al. (2014). A study of the effects of nanosilver on the gut microbiome examined microbiomes from specific regions of the gut, but this required sacrificing animals to dissect these regions (Williams et al. 2015), but this design limited their analyses to only one time point for comparing treated and control animals. In longitudinal studies, biosampling methods are limited to those that are not lethal to the animals.

What Is the Appropriate Experimental Model?

Selection of an animal model and study design depends upon the hypotheses being tested (Fritz et al. 2013). Biomedical researchers have adopted a model in which the host animal functions much as a petri dish to support a human microbiome in a host stripped of its own microbiome and in murine hosts housed in carefully controlled sterile conditions. While the humanized model (a gnotobiotic animal—usually a rodent—seeded with an individual or representative human microbiome) has demonstrated value for many biomedical studies, it may not be optimal for research on interactions between host and microbiome in responding to xenobiotics (Rawls et al. 2006). As toxicologists and pharmacologists, our commitment is to developing an experimental model that explicitly tests the importance of the microbiome as part of systemic responses of the host to an exposure or treatment. For that reason, a different model from the biomedical system is likely to be of value, that is, a nonhumanized model for in vivo studies of drug or toxicant exposures. There are other reasons to be cautious about the gnotobiotic model based on an extensive study that compared the gut microbiomes and metabolic profiles of germfree mice colonized by human microflora to data from the same strain of conventional mice with their normal mouse microbiome (Martin et al. 2007). Considerable differences were observed in microbiological and metabolic profiles as well as associations between the microbiome and tissue-specific metabolic profiles from liver. In general, the gnotobiotic mouse:human microbiome model demonstrated a “pre-pathologic state” in the terms of the authors, whereas the conventional mouse:mouse microbiome model resulted in “more refined” interactions between the microbiome and host physiology. These findings strengthen the rationale for utilizing what we term a homologous model for toxicology. To study effects of exposures on the human microbiome, the explanted system reviewed above may be preferable (Mazorati et al. 2015).

What Is the Information We Need to Acquire?

We need to utilize sophisticated methods that generate in-depth knowledge beyond phyla, particularly if we are interested in identifying specific genes and gene expression within the microbiome (such as arsenic methylation pathways). We also require methods for obtaining simultaneous information on the microbiome and on the host, as well as on cross talk between the microbiome and the host. If our goal is to examine the role of the microbiome in toxicology or pharmacology, we also need biosamples that characterize in vivo responses of target organs. Our goal is to utilize fecal pellets for microbiome analyses and for isolating host epithelial cells for analyses of host response at the interface of the microbiome and the host (McDermott and Huffnagle 2014). We can assess microbiome response by bacterial transcriptomics and assays of blood and urine for metabolomics as biomarkers of the functional status of the microbiome beyond taxonomics. We can assess host response in terms of epithelial cell gene expression as well as immunologic biomarkers measured in blood. And to assess overall response (toxicity or beneficial effect), we can use the most sensitive and specific methods in toxicology, including molecular and physiologic biomarkers of internal dose and response, functional measures, and pathology including immunohistopathology.

Because of the need for longitudinal studies (see below), it is important to utilize methods for biosampling blood and urine with minimal stress to animals. Dried blood and urine spots can be collected from droplets with minimal stress; these biosamples can be used to derive information on metabolomics (blood and urine) and cytokines (blood) as markers of microbiome and host response, respectively (Le Chatelier et al. 2013; Wang, Ehrlich, and Pedersen 2013; McDermott and Huffnagle 2014).

How Do We Deal with Unknowns Such as Variability Among and Within Subjects?

Due to our currently limited understanding of factors that contribute to inter- and intrasubject variability in terms of the microbiome (which is in large part due to the small sample size of most studies published to date), a longitudinal design may be preferable to cohort or group designs to characterize responses of the microbiome and the host over time. This requires methods that do not require necropsy or submitting animals to excessive stress. Study designs other than cross-sectional comparisons will also be needed in epidemiology.

What Are the Confounders and Interactions That Are Important to the Microbiome?

This question concerns animal studies as well as epidemiology. Clearly, just as we have learned to examine the constituents of diets in relation to studying endocrine disruption, characterizing, and standardizing diets will be important for studies of the gut microbiome. Infectious diseases—possibly even subclinical infections—will also be important covariates, as will the use of antimicrobial medications. This has been demonstrated in a study of the effects of helicobacter infections in mice exposed to arsenic (Lu et al. 2013). We do not know enough about age and sex, although studies are suggesting that these are important covariates in toxicological studies involving the microbiome (Williams et al. 2015).

Discussion and Conclusion: Do We Really Need to Do This?

Given the challenges listed above, it is not unreasonable to question the scientific value of including the microbiome in toxicological and epidemiological studies. Nevertheless, for more accurately defining the transition from external exposure to internal dose, it is hard to dispute the information that microbiomes at portals of entry play a role in the metabolism and absorption of xenobiotics, including drugs, chemicals, and natural compounds. It is equally hard to dispute the potential role of the microbial metagenome in testing gene:environment interactions. Finally, we share the excitement of researchers in clinical medicine in exploring the potential for involving the microbiome in prevention and treatment of diseases and dysfunctions associated with drug and chemical exposures.