Chapter 459: The Human Microbiome

Nearly two and a half millennia after Hippocrates made this statement, we are just coming to truly appreciate its profundity. Since the beginning of humankind, scholars have been investigating the underpinnings of disease with an almost singular focus on the human side of the equation. Microbes were not recognized as an important cause of disease until the inception of the “germ theory” in the late nineteenth century. During the first century of medical microbiology, research largely centered on the role of microbes as pathogens. Only recently has there been a resurgence of interest in understanding how commensal organisms—the bacteria, viruses, fungi, and Archaea that make up the microbiota—impact human physiology. The idea that these microorganisms are vital to the well-being of humans has challenged our traditional notions of “self.” Indeed, a human being can most accurately be described as a holobiont: a complex assemblage of human cells and microorganisms interacting in an elaborate pas de deux that drives normal physiologic processes.

Aimed at a better understanding of this relationship, myriad studies during the past decade have begun to catalogue the microbiota at various body sites and in a multitude of disease conditions. Diseases in virtually every organ system have been associated with changes in the microbiota. Indeed, the microbiota has been linked to intestinal disorders, disturbances in metabolic function, autoimmune diseases, and psychiatric conditions and has been shown to influence susceptibility to infection and the efficacy of pharmaceutical therapies. Knowledge of the specific mechanism(s) underlying most of these microbe–disease associations is lacking; it remains unclear whether the disease-associated alterations in the microbiota represent mere biomarkers of disease, a causal relationship, or a combination of the two. Although cause-and-effect relationships are still being elucidated for many diseases, it is clear that humans coexist in an intricate relationship with commensal organisms. This chapter explores in detail the nature of these host–commensal interactions, focusing on how this information might be translated into clinically meaningful interventions.

Massive undertakings, such as the Human Microbiome Project (HMP) sponsored by the National Institutes of Health and MetaHIT sponsored by the European Commission, have catalogued all the bacteria present at multiple body sites in people with and without disease. Coupled with the confluence of advances in sequencing technologies (Chap. 474), gnotobiotic animal availability, and microbial culture, significant progress has been made toward an understanding of the interplay between the microbiota and human health. However, recent findings were foreshadowed by work done centuries ago.

The human microbiota was first explored in 1683 when Antony van Leeuwenhoek described in a letter to the Royal Society of London the “very little living animalcules, very prettily a-moving” that he had observed in the plaque between his teeth. Leeuwenhoek went on to perform the first comparative “microbiota” studies by assessing how fecal and oral bacteria differ, how oral microbes change in the setting of disease (e.g., alcoholism and tobacco use), and how microbial composition changes across the age spectrum (e.g., in young children versus old men). He attempted—unsuccessfully—to eliminate these bacteria. Although Leeuwenhoek was not taken seriously when he first reported his findings, his studies laid the groundwork for what is now the field of microbiome research, and investigators are still trying to answer many of the same overarching questions that he raised more than three centuries ago.

Although Leeuwenhoek first reported the existence of bacteria and their association with humans at the end of the seventeenth century, the significance of commensal bacteria was not realized until late in the nineteenth century. In 1885, Pasteur suggested that animals could not survive if they were “artificially and completely deprived of the common microbes.” Although Pasteur’s preconceived ideas were proven incorrect in 1912 by the advent of germ-free animals (animals raised without exposure to any microorganisms), the underlying concept that commensal organisms are critical to health has held up. Élie Metchnikoff made another conceptual advance in this field by suggesting at the beginning of the twentieth century that clinical outcomes could be altered by the administration of specific beneficial organisms (probiotics). In particular, Metchnikoff believed that aging was caused by toxic bacteria in the gut and that lactic acid–producing bacteria (e.g., Lactobacillus species) present in sour milk and yogurt could mitigate against this process. The data behind this specific claim are still lacking, but recent discoveries offer continued hope that the microbiome can be effectively harnessed to protect against and treat a variety of diseases. Thus, although the field of microbiome research is sometimes considered to have emerged over the last one or two decades, the basic tenets—that the microbiota varies according to body site and clinical characteristics, that microbes are critical for human health, and that specific modulation of the microbiota may lead to improved clinical outcomes—are far from new.

Given that microbiome-based studies have identified and compared microbes at different levels of taxonomic resolution (Fig. 459-1), some understanding of taxonomy is essential for better comprehension of the implications of these studies. Of the ~100 bacterial phyla that exist in nature, only five (Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Proteobacteria) are dominant members of the human microbiome. Each of these phyla can be further categorized into multiple classes, orders, families, genera, and species. Early studies on the microbiota focused on changes in the relative abundance at the phylum level between different groups (e.g., obese versus normal-weight patients); however, these comparisons are at such a broad taxonomic level that they often provide little or no biological insight. As illustrated in Fig. 459-1, drawing comparisons between organisms in two different bacterial phyla is analogous to comparing humans to sea stars: the evolutionary distance between the two is tremendous. The limitations of current bioinformatic tools require lumping together of taxonomically related strains and thus cloud the richness of microbial ecology. Examining microbial profiles at the phylum, family, or even genus level—as is often done at present—ignores the great heterogeneity within different strains of the same bacterial species. The analytical pipelines are just now beginning to enable strain-level comparisons, and these improvements will likely facilitate our ongoing investigation of host–commensal interactions.

OVERVIEW OF THE HUMAN MICROBIOTA

The overwhelming majority of microbiota studies have focused on stool, given that this sample type represents the most ecologically rich anatomic site, is easy to obtain, and can readily be followed longitudinally in the same individual. A landmark study by the HMP sought to define the “normal” microbiota throughout the entire body in healthy Western adults. To this end, the microbial populations at 15–18 body sites were characterized in 242 people. One striking finding was that all samples from a given body region (e.g., skin) were more similar to each other than they were to samples from a different body region (e.g., stool), even in the same individual (Fig. 459-2A). In essence, the effect of the anatomic site on microbial composition is far greater than the effect of heterogeneity between individuals. That said, there was a remarkable amount of inter-individual variation at any given body site (Fig. 459-2B). In stool, for example, the abundance of the phylum Bacteroidetes ranged from ~10% in some individuals to >90% in others. Remarkably, even with person-to-person variability and differences among body sites, the functional capacity of the microbiota—assessed using metagenomic data to identify gene pathways—was quite similar across different people and different body sites (Fig. 459-2C). This discrepancy between the substantial differences in microbial composition and the little or no resulting change in the functional properties of the microbiota reflects an important ecologic property of the microbiota: the microbial communities at different body sites and in different people assemble in such a way that all the core metabolic functions are maintained. This finding also hints at the likely possibility of significant functional redundancy within the microbiota, with different species executing the same biological functions at different anatomic sites.

While the HMP provided the first large-scale catalogue of the microbiome in multiple people and at many different body sites, the amount of data generated by what, at the time, was by far the largest microbiome study has been dwarfed by subsequent studies. These more recent studies have confirmed the HMP’s major tenets: the composition of the microbiota differs by body site, there is tremendous inter-individual variation, and the microbial gene content is relatively conserved irrespective of the body site or individual. No microbial species are ubiquitous in all individuals and at all body sites, but some species are highly prevalent at a given body site: in the HMP study, Staphylococcus epidermidis was present in 93% of nares samples and Escherichia coli in 61% of stool samples. These findings highlight the remarkable personalization of the human microbiome. While the human genome is typically >99.5% identical in different people, the microbiotas of two individuals may not overlap at all. Although the “precision medicine” approach currently focuses on teasing out how differences in the human genome relate to different clinical end points, the human microbiome clearly represents a critical component for consideration.

THE MICROBIOTA BY THE NUMBERS

It has long been known that the human-associated microbiota is numerically dense. Leeuwenhoek estimated that there were more “animals living in the scrum on the teeth in man’s mouth than there are men in a kingdom.” Specific enumeration of the components of the microbiota has been challenging, in part because of its variability across time, space (body region), and clinical conditions. Moreover, the majority of human-associated microbes are not readily cultivable—a situation that raises questions about the best methodology for such quantitation. Initial back-of-the-envelope calculations performed in the 1970s suggested that there were roughly tenfold more bacteria in the body than there were human cells. This rather astounding estimate suggested that humans are really only ~10% “human” and that by far the greatest part of the holobiont is represented by microbes. This stark numerical discrepancy has prompted some to question “who parasitizes whom.” However, a more recent estimate has suggested that there are “only” ~1.3 times more bacteria in the body than there are human cells and thus that humans are ~56% “bacterial.” Of note, this more recent study does not include the numbers of viruses (known to generally be approximately tenfold more abundant than other microbes), fungi, or Archaea. Given these additional microorganisms, the notion that microbes constitute >90% of the cells present in a human body is likely correct. These ratios are even starker when one considers the genetic potential of human cells versus that of commensal organisms. In contrast to the ~20,000 genes in the human genome, the estimated total number of genes in the microbiota (which together constitute the microbiome)—i.e., >2,000,000—indicates that the human genome contributes <1% to the total genetic potential of the overall holobiont. Most microbiome studies to date have focused almost exclusively on the bacterial component; much remains to be learned about the functional interplay of bacteria, viruses, fungi, and Archaea and how these other classes of microorganisms impact human health.

In terms of overall diversity, >10,000 different bacterial species are present in the human microbiota; the intestines alone contain >1000 species. At any given time, the body of any given individual harbors 500–1000 bacterial species, with 100–200 bacterial species in the gut alone. If one considers different strains of the same bacterial species, which may be functionally different from one another, the diversity of the microbiota is probably at least a magnitude greater. Although marked diversity exists at the strain and species level, only limited bacterial phyla are generally found in the human microbiota at any given body site (Fig. 459-3).

INFLUENCES ON THE MICROBIOTA

An individual’s specific microbial configuration is dynamic and is quickly altered in response to subtle changes in the microenvironments in which the bacteria reside. On a day-to-day basis, these changes usually reflect alterations in the relative abundance of the various microbes. However, some exposures have a greater effect on the microbiota and can shift the microbial population to a new equilibrium via the loss of specific species and/or the acquisition of others; this new microbial equilibrium can be associated with either health or a disease state (Fig. 459-4). Identification of the factors that influence the microbiota’s composition is critical to an understanding of what leads to and controls intra- and inter-individual variation. Moreover, an understanding of the influences on the microbiota will facilitate the design and proper interpretation of microbiota studies. While it is clear that the microbiota can be altered through these various mechanisms, it is not yet clear whether these changes are biologically significant.

Genetics

Studies of monozygotic and dizygotic twins have revealed that host genetics have a small but statistically significant effect on the microbiota’s composition. Notably, some taxa, such as Christensenella species, are more heritable than others. Thus, some inter-individual variation may be due to underlying differences in host genetics; however, it is likely that other factors explain more of the observed variability. That said, the host’s genetic contribution to the microbiota, albeit small, may be meaningful. Studies in mice have demonstrated that genetic variation in the major histocompatibility complex, a specific set of immune-related genes, leads to changes in the microbiota that alter susceptibility to an autoimmune disease. These studies offer a proof of concept for the notion that the genetic predisposition observed for certain diseases may actually be mediated by indirect alterations in the microbiota.

Age

Burgeoning evidence now indicates that microbial exposure begins in utero: bacterial DNA has been identified in otherwise healthy placentas, in amniotic fluid obtained at early stages of gestation, and in meconium of term newborns. Although some controversy persists about whether these results reflect contamination and/or the presence of nonviable bacteria, they raise the possibility that human exposure to the microbial world begins before birth. The delivery mode (vaginal versus cesarean section) and the method of feeding (breast milk versus formula, timing of solid food introduction) are major determinants of an infant’s early microbiota. After birth, the infant’s microbiota goes through a stereotyped succession process; with increases in bacterial diversity and functional capacity, the child’s microbiota resembles that of an adult by the age of 2–3 years. Cross-sectional studies that have examined the microbiota across the entire age spectrum have revealed a general stability of the fecal microbiota after 2–3 years of age; however, the microbiota of the elderly (persons >80 years of age) demonstrates notable differences from those of their younger counterparts, with increases in Bacteroides and Eubacterium species and decreases in the bacterial family Lachnospiraceae.

Diet

Diet is a strong determinant of human health. The impact of diet is mediated, in part, by its effects on the composition of the gut microbiota. This makes intuitive sense, as the human diet provides nutrients needed not only by our own cells but also by the microbes living in the alimentary tract. In young children, this dietary influence is marked by major shifts (e.g., a decrease in Bifidobacterium species) in the intestinal microbiota that occur at weaning and with the introduction of solid food. In adults, long-term dietary patterns are associated with relatively stable microbial compositions. However, drastic changes in short-term macronutrient availability cause rapid (within 1 day) and reproducible fluctuations in the fecal microbiota that reflect the biological processes needed to degrade and metabolize the nutrients in the new diet. For example, vegetarian diets are associated with a microbiota that has an increased ability to metabolize plant polysaccharides (e.g., Roseburia species, Eubacterium rectale, Ruminococcus bromii), while animal-based diets result in an increased abundance of bile-tolerant organisms (e.g., Alistipes, Bilophila, and Bacteroides species). At the completion of dietary interventions and the resumption of the individual’s normal dietary pattern, the microbial communities revert back to their previous states, probably because the individual resumes his or her typical diet. Taken together, dietary studies confirm that the microbiota is highly adaptable and varies in relation to changes in the diet. Of note, virtually all of these studies have focused on how the diet influences the fecal microbiota. It will be interesting to determine whether dietary changes similarly influence the microbiota at nonintestinal sites.

Drugs

Virtually all drugs have the capacity to change the microbiota by altering the chemical landscape in which the microorganisms live (e.g., statins, bile acid sequestrants), modulating the host’s ability to recognize and react to microbes (e.g., immunosuppressants), and/or directly interfering with the microbiota’s constituents (e.g., antibiotics). These potential effects have made critical interpretation of microbiota studies much more difficult. A prominent study that claimed to identify a fecal microbiota signature associated with type 2 diabetes was later found actually to have identified a signature for patients taking metformin instead; the effects of this drug on the microbiota were far greater than the effects of the disease itself. These results highlight the importance of controlling for clinical variables in microbiota studies.

Antibiotics are the most obvious and best-studied class of drugs that modulate the microbiota. Multiple groups have demonstrated that antibiotics exert a considerable effect on the gut microbiota by depleting antibiotic-sensitive strains. What is more surprising is that many strains resistant to the antibiotic tested are also eliminated. This observation highlights the intricate microbe–microbe interactions that are fundamental to maintenance of the overall microbial community. For example, treatment with ciprofloxacin, which has little to no activity against clinically relevant anaerobes, leads to a loss of roughly one-third of the bacterial taxa in the gut. This broad effect is likely mediated by the depletion of certain “keystone” species that are required for the persistence of other, unrelated species. While many of the observed antibiotic effects (e.g., loss of specific taxa) are shared across many different individuals, some effects vary greatly among people. For example, studies found that microbiota recovery following antibiotic treatment differed significantly in terms of timing and degree. The microbiota of most healthy people who received ciprofloxacin for 5 days had completely recovered within 4 weeks, whereas microbiologic changes lasted up to 6 months in other individuals. Moreover, the degree of variation was compounded by repeated antibiotic administration, with fewer individuals reverting to their baseline microbiota after a second course of ciprofloxacin given 6 months after the first. These findings are consistent with those of microbial ecology experiments, which also showed that this type of repeated disturbance leads to less predictable results.

Lifestyle

Many seemingly innocuous lifestyle decisions can impact the human microbiota. For example, a person’s skin and fecal microbiotas are more similar to those of their household members, regardless of genetic relatedness, than to those of residents of different households. The degree of similarity in skin microbiotas is even greater if a dog also lives in the home; in contrast, the presence of a young child does not accentuate this microbial relatedness. The presumption is that the dog serves as a more effective “vector” for transmitting microbes during its frequent direct contact with adults in the household. The type of setting in which a person lives also impacts the microbiota. Living in a rural or farm setting leads to a different fecal microbiota than living in an urban environment. Similarly, the individual’s country of residence affects the microbiota. An analysis of daily fecal samples from an individual who temporarily (i.e., for a couple of months) moved from the United States to Thailand demonstrated a large shift in the fecal microbiota that coincided with arrival in Thailand and a reversion in most respects to the “American” microbial configuration upon return to the United States. These geography-driven changes probably reflect a combination of environmental and dietary differences between locations.

Circadian Rhythms

Many human biological processes follow a circadian clock; aspects of physiology are tuned by external cues, including the degree and timing of ambient light, temperature, and availability of nutrients. This endogenous biological clock enables animals to efficiently adapt to changing environmental conditions. Similarly, the microbiota maintains a circadian rhythm that is linked to the host’s circadian clock. If circadian oscillations are disrupted in the host, they are similarly disrupted in the microbiota, and vice versa. These bacterial vacillations occur at the level of spatial localization within the intestine, relative species abundance, and bacterial metabolite secretion. Work in the 1960s showed that mice exhibited daily periodicity of susceptibility to infection with either S. pneumoniae or E. coli lipopolysaccharide. Although the fundamental basis for this difference was not known at the time, it is likely to be related, in part, to the microbial circadian clock. Derangements of these microbial oscillations have also been linked to the development of metabolic diseases and may underlie some of the health hazards associated with shift work and jet lag.

THE HYGIENE HYPOTHESIS

Over the past few decades, abundant epidemiologic data have revealed an inverse correlation between exposure to microbes and the incidence of autoimmune and/or atopic diseases (Fig. 459-5). This type of epidemiologic correlation led to the proposal of the “hygiene hypothesis” in 1989. Initially, this hypothesis focused on the development of atopic diseases in young children, with the idea that these epidemiologic observations could “be explained if allergic diseases were prevented by infection in early childhood, transmitted by unhygienic contact with older siblings, or acquired prenatally from a mother infected by contact with her older children.”¹ In fact, this notion that differences in living conditions and environmental exposures contribute to susceptibility to hay fever (summer catarrh) dates back to at least the early nineteenth century. The hygiene hypothesis has continued to evolve over the past three decades and now posits that inadequacies in microbial exposure—in combination with genetic susceptibilities—lead to a collapse of the normally highly coordinated, homeostatic immune response. At its core, the hygiene hypothesis holds that specific early-life microbial exposures are required to prevent subsequent disease and that the “Westernization” of society has led to a decrease in such exposures. This concept is now being applied beyond atopic diseases to other inflammatory and autoimmune diseases and is thought to reflect processes that occur in later life as well.

RELATIONSHIP BETWEEN THE MICROBIOTA AND SPECIFIC DISEASE STATES

The ideas inherent in the hygiene hypothesis—in sum, that microbial exposure can affect long-term health outcomes—laid the theoretical foundation for translational microbiome studies. While most of the studies described earlier sought to describe how the microbiota responds to specific and often transient influences (e.g., a course of antibiotics, dietary interventions, travel), a multitude of studies have characterized the microbiota in patients with various diseases in the hope that a better understanding of the nature of disease-specific microbial communities will provide insight into disease pathogenesis and potentially uncover novel treatment modalities. Remarkably, virtually all of these studies have demonstrated differences between the microbiotas of healthy controls and patients, irrespective of the specific disease process examined. Although it is difficult to generalize across all studies, a couple of general themes have emerged. First, disease states are typically associated with microbiotas that are less diverse than those of healthy individuals. This loss of diversity can be measured either as a decrease in the number of species (alpha diversity; often measured as the number of operational taxonomic units, which are the bioinformatic equivalent of species) or as a reduction in the microbial relatedness of the species present (beta diversity). Often, both alpha and beta diversity decrease in the setting of disease. Second, states of inflammation—regardless of site or underlying disease process—are often associated with a relative increase in the abundance of the bacterial family Enterobacteriaceae and a decrease in the abundance of Lachnospiraceae.

Dissecting Correlation and Causality

Given that most of these investigations have been designed as case–control studies, it is difficult to determine whether microbiologic findings are the cause or the effect of the disease. Even studies that examine treatment-naïve patients at the time of initial diagnosis are still confounded by this “chicken or egg” issue. Moreover, prospective, longitudinal clinical studies—still rare in the microbiome field—may simply yield correlations between the microbiome and subclinical disease rather than necessarily proving causality. Experiments in animals—specifically, studies using gnotobiotic mice (germ-free mice that have been colonized with specified microbial communities)—have been critical in this regard as they allow investigation of specific differences in microbial components while controlling for the host’s genetics, diet, and housing conditions. Moreover, human microbes can be transplanted into gnotobiotic mice to permit in-depth mechanistic studies of how these microbial communities affect disease pathogenesis. This marriage of human samples and animal experiments has facilitated the identification of causal roles played by some microbes in disease pathogenesis; these findings provide a critical proof of concept for the interplay of the microbiota with human health. However, the vast majority of microbiome studies are still at the level of correlation. The next several sections describe the clinical and animal data for many different disease processes. Given the voluminous and rapidly changing nature of this field, it is impossible to cover all of the disease associations known to date; rather, the following discussion represents a combination of the leading exemplars of microbiome data and nascent areas of significant clinical interest. In all cases, the hope is that further study of the role of the microbiota will provide novel diagnostics, new therapeutic modalities, and/or additional insight into disease pathogenesis.

Gastrointestinal Diseases

Given that the intestines harbor the largest number and greatest diversity of organisms in the body, much work has focused on how the microbiota impacts gastrointestinal diseases. Even though the luminal surface area of the gastrointestinal tract is 30–40 square meters (~90% of which is contained within the small intestine) and features marked anatomic and functional differences that result in many discrete macro- and micro-ecosystems, stool is often used as a surrogate for the intestinal microbiota given the relative ease of collecting samples. A few studies that have compared the microbial profile in stool with the mucosa-adherent organisms present in biopsy samples have demonstrated that stool is, in fact, a reasonable proxy for biopsy samples; however, the relative microbial “noise” present in stool can sometimes overwhelm the “signal,” making biopsy samples more informative for some scientific questions. The key issue is to ensure that the biopsy samples evaluated represent relatively similar intestinal regions, as there are significant differences between the organisms present in the crypt and the tip of the villus and between microbes found in the ascending versus the descending colon.

OBESITY

Obesity is a worsening epidemic throughout the world, and multiple studies have linked the composition of the intestinal microbiota to the development of obesity in animal models and in humans. Indeed, many of the initial translational microbiome studies performed in mice at the beginning of the twenty-first century focused on obesity. These early studies suggested that the ratio of the relative abundance of Bacteroidetes to Firmicutes was lower in obese mice than in control animals. Moreover, a causal relationship between the microbiota and obesity was established by the finding that gnotobiotic mice colonized with the microbiota from obese individuals had more rapid and more extensive weight gain than gnotobiotic mice colonized with the microbiota from lean individuals. Biologically, it is posited on the basis of metagenomic surveys that the obesity-associated microbiome has an increased capacity to harvest energy from the diet. Notably, the relationship between the Bacteroidetes/Firmicutes ratio and obesity did not hold in initial human studies; however, the finding that this ratio increased in obese patients who lost weight while on a fat- or carbohydrate-restricted diet suggested some generalizability between mice and humans. Beyond this ratio of major bacterial phyla, obesity was linked to a microbiome with a lower alpha diversity. Over the past ~15 years, numerous human studies examining the relationship between the microbiome and obesity have been completed, all with mixed results. A recent meta-analysis of 10 studies including nearly 3000 individuals revealed an apparent lack of relationship between the Bacteroidetes/Firmicutes ratio and obesity, though there is ~2% lower diversity associated with obesity that is statistically significant but of unclear biological significance. This finding highlights a problem common to microbiome studies: i.e., there is no sense as to what magnitude of change is biologically meaningful. Ultimately, although murine studies have indicated a causal link between the microbiota and obesity, the human data are less convincing, and their significance may be limited by the studies’ having primarily examined only high-level taxonomic information rather than also assessing transcriptional or metabolic differences.

The rise in obesity has elicited a plethora of ideas about the type of diet that might be most successful in leading to sustained weight loss. It has become clear that the same dietary ingredient can have highly diverse effects on blood glucose measurements in different people and that this effect is mediated largely by the microbiome. These observations suggest that the “optimal” diet needs to be individualized in the context of the person’s microbiome, which itself may continue to change over the course of the diet. An intriguing parallel question is whether the microbiota may also influence dietary preferences; such an influence would suggest important feedback loops between the microbiome and diet.

MALNUTRITION

Representing the other end of the metabolic spectrum from obesity, malnutrition is also linked to an altered microbiome. Analysis of Malawian twin pairs (≤3 years of age) who were discordant for kwashiorkor—a severe form of malnutrition—revealed that kwashiorkor is associated with a microbiologically “immature” fecal microbiota that resembles that of a chronologically younger child. Transplantation of the fecal microbiota from these discordant twins into gnotobiotic mice that were fed a diet similar in composition to a typical Malawian diet established that the kwashiorkor-associated microbiome is causally related to poor weight gain. Subsequent studies demonstrated these same general trends in malnourished Bangladeshi children. Investigators were able to identify five bacterial species (Faecalibacterium prausnitzii, Ruminococcus gnavus, Clostridium nexile, Clostridium symbiosum, and Dorea formicigenerans) that—when administered together as a “cocktail” to mice colonized with a kwashiorkor-associated microbiome—was able to prevent growth impairments. These results demonstrate that rationally designed modulation of the murine microbiota can lead to improved health outcomes. The clinical significance of these findings for humans remains to be clarified.

INFLAMMATORY BOWEL DISEASE

Ulcerative colitis and Crohn’s disease, the two predominant forms of inflammatory bowel disease (IBD), are chronic gastrointestinal inflammatory conditions that differ in their locations and patterns of inflammation (Chap. 319). The following observations have led to the suggestion that IBD is the result of an immune response to a dysbiotic microbiota in a genetically susceptible individual: genes account for only ~20% of susceptibility to IBD (and many of the relevant genes are related to host–microbe interactions), antibiotic treatment reduces the clinical severity of disease, and relapses of Crohn’s disease are prevented by diversion of the fecal stream. While the microbiota clearly is not the only driver of disease, it is considered to be an important element. Accordingly, numerous animal and clinical studies have been designed to tease out the nature of the relationship between the microbiota and IBD.

Most of these studies have focused on comparing the microbiome’s composition in IBD patients with that in healthy controls, concentrating on microbial diversity and specific bacterial taxa that are associated with health or disease. Unfortunately, few, if any, results have been universally obtained, probably because of differences in study design, inclusion criteria, and methodology (e.g., the use of stool, rectal swabs, or biopsy samples; the choice of sequencing primers; the analysis pipeline). Even with these differences among studies, patients with IBD have been shown typically to have reduced alpha and beta diversity in their fecal microbiotas. Moreover, Clostridium clusters IV and XIVa, which are polyphyletic and encompass several different bacterial families, are generally reduced in patients with IBD. F. prausnitzii is a notable example from Clostridium cluster IV that is often underrepresented in the stool of patients who have Crohn’s disease, with more mixed results in biopsy samples. The bacterial family Lachnospiraceae, which is largely contained in Clostridium cluster XIVa, and other butyrate-producing organisms are also reduced in the stool of patients with IBD. Some of these species produce butyrate by using acetate generated by other members of the microbiome, and some of these acetate-producing species are similarly reduced (e.g., Ruminococcus albus). These complex interactions and dependencies among bacterial species pose unique challenges to definitive ascertainment of the cause–effect relationships between microbes and disease. Even before researchers were able to assess the entire microbiome at once, they often noted that patients with Crohn’s disease had a higher representation of adherent invasive E. coli in the ileal mucosa, an observation consistent with the increased abundance of Enterobacteriaceae seen in more recent microbiome studies. Beyond bacteria, burgeoning evidence supports a role for Caudovirales bacteriophages in IBD pathogenesis, though these findings may merely reflect the underlying dysbiosis related to the loss of bacterial diversity in IBD. Moreover, some data suggest that IBD is also associated with fungal dysbiosis; several studies have demonstrated an increased ratio of Basidiomycota to Ascomycota. It is still unclear whether any of these microbial associations reflect the cause of IBD or merely serve as biomarkers of disease.

Studies of antibiotic-treated mice and gnotobiotic mice colonized with IBD-associated microbiotas have been useful in confirming that the microbiota affects colitis severity. Several bacterial species have been identified as either promoting colitis in mice (e.g., Klebsiella pneumoniae, Prevotella copri) or protecting against it (e.g., Bacteroides fragilis, Clostridium species); however, these organisms do not always correlate with the taxa identified as differentially abundant across multiple clinical studies. In contrast, IgA-coated commensal organisms isolated from patients with IBD promote more severe colitis in mice than either IgA-uncoated bacteria from patients with IBD or IgA-coated bacteria from healthy controls. These data suggest that functional categorization of the microbiota based on immune recognition (e.g., IgA coating) may be a useful approach for identifying pathogenic organisms.

Cardiovascular Disease

Inflammation helps drive the pathogenesis of atherosclerosis, and it has long been postulated that microbes are involved in the atherosclerotic process. Early work demonstrated that patients with cardiovascular disease have higher titers of antibody to Chlamydia pneumoniae than control patients, that C. pneumoniae is present within atherosclerotic lesions, and that C. pneumoniae can both initiate and exacerbate atherosclerotic lesions in animal models. This type of analysis has been extended to other bacteria, such as Porphyromonas gingivalis, with the idea that multiple different bacteria may play some role in the pathogenesis of atherosclerosis.

More recent studies have demonstrated clinical correlations between serum levels of trimethylamine N-oxide (TMAO) and atherosclerotic heart disease. Given that red meat, eggs, and dairy products are important sources of carnitine and choline (both precursors of TMAO), it is not surprising that levels of TMAO are higher in omnivores than in vegans. Animal studies have confirmed that transfer of the gut microbiota from atherosclerosis-susceptible strains of mice to atherosclerosis-resistant animals leads to increased serum levels of TMAO and a dietary choline-dependent increase in atherosclerotic plaques; this observation confirms the role of the gut microbiota in the generation of TMAO and atherosclerosis. Moreover, treatment of atherosclerosis-susceptible strains of mice with a structural analogue of choline that inhibits the first enzymatic step in TMAO formation leads to decreased circulating TMAO levels and, more importantly, restrains macrophage foam-cell formation and atherosclerotic lesion development. In a study of more than 4000 patients, plasma TMAO levels were also predictive of incident thrombosis risk (myocardial infarction, stroke). Gnotobiotic animals were used to demonstrate that this risk was dependent on the microbiota; although eight bacterial taxa were identified as being associated with both plasma TMAO levels and thrombotic risk, organisms with choline-utilization genes that represent the first step of TMAO production were not more abundant in animals at greater risk for thrombosis. This discrepancy highlights the complexity of the microbiota and suggests that other aspects of the overall dynamics of the microbial community may be in play.

Oncology

Recent studies exploring the link between the microbiota and cancer have demonstrated that specific members of the microbiota can affect treatment efficacy in both a positive and a negative manner. For example, therapy with antibody to programmed cell death ligand 1 (anti-PD-L1) has proven highly effective for a number of different cancers (Chap. 69); however, a significant proportion of patients do not respond even when their tumors have high PD-L1 expression levels, a prerequisite for this type of checkpoint blockade inhibition. Using a murine melanoma model, investigators showed that variations in the microbiota resulted in differences in melanoma growth, an impact that was accentuated by anti-PD-L1 therapy. Ultimately, Bifidobacterium species were bioinformatically associated with improved anti-tumor responses, and administration of a “cocktail” of Bifidobacterium species (B. bifidum, B. longum, B. lactis, and B. breve) to melanoma-susceptible mice resulted in improved tumor-specific immunity and responses to anti-PD-L1 therapy. In a separate set of studies involving both patient data and animal experiments, the efficacy of therapy with antibody to cytotoxic T lymphocyte–associated antigen 4 (anti-CTLA-4) was associated with T-cell responses specific for either Bacteroides theta-iotaomicron or B. fragilis. In particular, administration of B. fragilis to germ-free or antibiotic-treated mice restored the normally absent anti-cancer response to anti-CTLA-4 therapy. While both of these examples demonstrate potentiation of anti-cancer therapies by the microbiota, other therapies can be antagonized. Some cancers, such as pancreatic ductal adenocarcinoma, contain intratumor bacteria, particularly Gammaproteobacteria, that can metabolize the chemotherapeutic agent gemcitabine and thereby contribute to the drug resistance of these tumors. Overall, these examples highlight the microbiota’s critical impact—both direct and indirect—on the efficacy of drugs. Many other notable examples have been described (e.g., involving cyclophosphamide, digoxin, levodopa, and sulfasalazine), and many more likely remain to be discovered.

The application of microbiome science to hematopoietic stem cell transplantation (HSCT) is an area of expanding interest, particularly given the significant morbidity and mortality related to graft-versus-host disease (GVHD). In light of studies in the 1970s showing that germ-free mice developed less frequent and less severe gut GVHD than wild-type mice, clinicians began to use antibiotics to decontaminate the gut of patients undergoing HSCT. This decontamination approach yielded mixed results, probably because of differences in the antibiotic regimens used. The natural history of patients undergoing allogeneic HSCT includes a substantial loss of diversity in the fecal microbiota, with lower levels of bacterial diversity associated with increased mortality. Moreover, a retrospective analysis of ~850 patients undergoing allogeneic HSCT revealed that receipt of imipenem-cilastatin or piperacillin-tazobactam for neutropenic fever was associated with increased GVHD-related mortality at 5 years; this observation suggested that specific bacteria may help protect against GVHD-related mortality. More detailed analyses revealed an association between the abundance of Blautia species and protection against GVHD and mortality, though this correlation is still being examined with regard to its causal relationship. Despite significant interest in examining these microbial relationships with HSCT, little has yet been studied in the context of solid organ transplantation, which likely represents the next frontier of transplantation-related microbiome investigation.

Autoimmune Diseases

The dramatic rise in the incidence of many autoimmune diseases over the past few decades has been far more rapid than can be explained simply by genetic factors (Fig. 459-5). It is increasingly thought that environmental triggers, including the microbiome, are partially responsible for the development of these autoimmune diseases.

TYPE 1 DIABETES

Type 1 diabetes (T1D) is an autoimmune disorder characterized by T cell–mediated destruction of insulin-producing pancreatic islets (Chap. 396). There is a clear genetic predisposition for the disease: ~70% of patients with T1D have human leukocyte antigen (HLA) risk alleles. However, only 3–7% of children with these risk alleles actually develop disease, an observation that suggests a role for other environmental factors. Studying a prospective, densely sampled, longitudinal cohort of at-risk, HLA-matched children from Finland and Estonia, investigators detailed changes in the microbiota prior to development of disease. Although only 4 of the 33 children studied developed T1D within the time frame of the study, a marked decrease of ~25% in alpha diversity occurred after seroconversion but before disease diagnosis. The low number of cases in this study unfortunately precluded identification of any specific disease-associated taxa. A follow-up study compared the microbiomes of a larger cohort of these high-risk northern European children with those of low-risk Russian children who lived in geographic proximity. Bacteroides species were more abundant in the high-risk group than in the low-risk group, particularly at early ages. This difference was postulated to be associated with an altered structure of the bacterial lipopolysaccharide to which children were exposed at a young age. It was further suggested that Bacteroides-derived lipopolysaccharide was not able to provide the immunogenic stimulus necessary to prevent T1D. These two studies offer attractive—though logistically complicated—options for future clinical investigations aimed at exploring the role of the microbiome. The first approach—longitudinally following individuals who are at high risk for a given disease—may provide insight into host–microbe relationships by mapping temporal changes in the microbiome with disease onset. An important caveat with this type of study, though, is that the associations identified may reflect preclinical disease rather than specifically indicating causality for any observed changes. The second approach illustrates how careful selection of study participants may offer an opportunity to uncover more meaningful associations that can subsequently be experimentally verified.

RHEUMATOID ARTHRITIS

Similar to many other autoimmune diseases, rheumatoid arthritis (RA) is a multifactorial disease that comes to clinical attention after an environmental factor triggers symptoms in an individual with pre-existing autoantibodies. Multiple lines of evidence support the notion that RA pathogenesis is reliant on the microbiota, including the findings that germ-free mice do not develop symptoms in several RA models and that antibiotic treatment of mice mitigates against RA development. Several taxa (e.g., Bacteroides species, Lactobacillus bifidus, and segmented filamentous bacteria) have been implicated in promoting RA in murine models, and analysis of the fecal microbiota of patients with newly diagnosed RA have indicated that P. copri is a biomarker of disease. That this association with P. copri does not exist for chronic, treated RA or for psoriatic arthritis suggests some specificity for new-onset RA. A major limitation of this approach is that the identified association is shown to be a biomarker of disease (and, in this case, potentially of response to treatment) but no added insight is gained into a possible causal relationship between P. copri and RA. In fact, many of the patients with new-onset RA had no Prevotella detected, and several of the healthy controls had significant levels of Prevotella. The lack of a strict concordance between the presence (or absence) of a specific taxon and a given disease state argues against a possible causal role.

MULTIPLE SCLEROSIS

Epidemiologic studies of twin pairs and at-risk individuals moving between high- and low-risk geographic areas indicate that genetics plays a minor component in multiple sclerosis (MS) susceptibility relative to environmental factors. For example, in monozygotic twin pairs in which one sibling has MS, the other sibling also develops MS in only ~30% of cases. Although MS is a disease of the central nervous system (CNS), there is growing evidence of a link between MS and the microbiota, specifically that of the gut. Germ-free animals and antibiotic-treated animals display reduced disease incidence and severity in an MS model. Similarly, some clinical studies suggest improved disease outcomes in patients with MS who have been treated with minocycline, while patients treated with long-term penicillin appear to have an increased disease risk. Although several studies have compared the fecal microbiotas of healthy controls to those of patients with MS, these studies have all been relatively small and have yielded few results (if any) that are common throughout. Although work relating the microbiome to MS is ongoing, it has opened the door to exploring this link with other neurologic diseases. Already, there are animal data demonstrating links between the microbiota and both Parkinson’s disease and autism, and there are clinical data assessing fecal microbiomes in relation to a variety of neurologic conditions. It is not quite clear how the gut microbiota is communicating with the CNS—i.e., whether communication takes place via bacterial metabolites that travel in the bloodstream and cross the blood–brain barrier, via migration of whole organisms into the CNS, or via feedback through the vagus nerve. Although our understanding of this brain–gut axis is still in its infancy, research in this area has elicited tremendous excitement as a tractable approach to potential treatments for these challenging diseases.

Atopic Diseases

The incidence and prevalence of allergic diseases continue to steadily increase, as do more severe clinical presentations. Life-threatening food allergies are now such a public health issue that nut-free classrooms are the norm in many cities. The development of allergic diseases often follows a stereotyped progression that begins with atopic dermatitis (AD) and continues, in order, with food allergy, asthma, and allergic rhinitis. The microbiome has been linked to all of these conditions and has the potential to modulate effects anywhere along this spectrum.

ATOPIC DERMATITIS

The skin is the largest organ in the body, and its different anatomic sites (e.g., antecubital fossa, volar forearm, alar crease) represent distinct ecologic niches and harbor unique microbial communities. Moreover, given that the skin serves as a critical interface between the body and the external environment (e.g., microbes), it must be able to respond to unwanted microbes with an adequate immune response. AD is an inflammatory skin disorder involving immune dysfunction and a dysbiotic skin microbiota that is typically marked by greater abundances of Staphylococcus aureus and a lesser degree of bacterial diversity. Effective treatment of AD does not require complete elimination of S. aureus but is associated with restoration of the normal level of diversity. It is likely that this increase in bacterial diversity re-establishes normal immune homeostasis in the skin; specific members of the skin microbiota have been shown to induce protective skin-restricted immune responses. Coagulase-negative staphylococci (CoNS; primarily S. epidermidis and S. hominis) obtained from lesional and nonlesional skin of patients with AD were functionally screened and compared to CoNS from healthy controls; AD-lesional CoNS were much less often able to produce antimicrobial peptides (lantibiotics) directed against S. aureus. To demonstrate that these lantibiotic-producing CoNS were biologically relevant, they were incorporated into a lotion and applied to the arms of patients with AD. Surprisingly, a single application of the probiotic-laced lotion led to a decrease in the abundance of S. aureus recovered; no such decrease was observed when lantibiotic-negative strains were used. The authors of this study did not specifically comment on the clinical improvement of the AD lesions. Nevertheless, this is one of a limited number of studies that is beginning to extend microbiome-related findings into clinical trials.

ASTHMA

Asthma is characterized by the clinical triad of airflow obstruction, bronchial hyperresponsiveness, and inflammation in the lower respiratory tract. Although the long-standing dogma was that the lungs are sterile, there is now convincing evidence for a constant ebb and flow of bacteria within the lower airways. In healthy states, the mucociliary escalator continually eliminates these bacteria soon after they land in the airways; in disease states (e.g., cystic fibrosis, chronic obstructive pulmonary disease), these bacteria establish long-term colonization of the airways and influence disease pathogenesis. In asthma specifically, both fecal and airway microbes have been linked to clinical outcomes.

Early studies of the microbiome’s influence on asthma used culture-based methods to assess the hypopharyngeal microbiota of asymptomatic 1-month-old infants. Intriguingly, in one study, early-life colonization with Streptococcus pneumoniae, Haemophilus influenzae, Moraxella catarrhalis, or a combination of these organisms—but not S. aureus—was significantly associated with persistent wheeze and asthma at 5 years of age. Eosinophilia and total IgE levels at 4 years of age were also increased in children who were neonatally colonized with these organisms. Although this study examined a fairly focused set of bacteria, it laid the experimental groundwork indicating that early-life microbial exposures influence subsequent development of asthma. A later longitudinal investigation of the fecal microbiota in a general-population birth cohort of more than 300 children demonstrated that lower abundances of the genera Lachnospira, Veillonella, Faecalibacterium, and Rothia at 3 months of age were associated with an increased risk for development of asthma. The fact that these bacterial changes were no longer apparent when the children were 1 year of age is consistent with the notion that microbial exposures early in life are important to disease pathogenesis later in life. Transplantation of stool samples from 3-month-old children at risk for asthma into gnotobiotic mice resulted in significant airway inflammation in a murine model of asthma; pre- and postnatal exposure of mice to a four-species cocktail (F. prausnitzii, Veillonella parvula, Rothia mucilaginosa, and Lachnospira multipara) inhibited airway inflammation, with a marked reduction in neutrophil numbers in bronchoalveolar lavage fluid. These data suggest that early-life modulation of the microbiome may be an effective strategy to help prevent asthma, though the specific logistics (e.g., strains, dose, timing of exposure, patient selection) remain to be clarified.

Infectious Diseases

The increased susceptibility of antibiotic-treated mice to infection with a wide range of enteric pathogens was initially observed in the 1950s and led soon thereafter to the concept of colonization resistance, which holds that the normal intestinal microbiota plays a critical role in preventing colonization—and therefore disease production—by invading pathogens. Seminal work in the 1970s demonstrated that this protection is largely reliant on clostridial organisms, and the subsequent half-century has been spent trying to identify the specific microbes involved. Although much of the work relating the microbiota to infection has focused on enteric pathogens, the intestinal microbiota has also been clearly linked to bacterial pneumonia in mouse models, and changes in the microbial composition of the gut have been causally related to changes in the severity of disease. Although this gut–lung axis clearly exists in animals, its relevance in humans is still unclear. Several groups are beginning to study the human lung microbiome in the context of pneumonia and tuberculosis. Moreover, the relationships between the microbiota and both systemic infections (e.g., HIV infection, sepsis) and the response to vaccination are starting to be explored.

ENTERIC INFECTIONS

Clostridium difficile infection (CDI) represents a growing worldwide epidemic and is the leading cause of antibiotic-associated diarrhea (Chap. 129). Roughly 15–30% of patients who are successfully treated for CDI end up with recurrent disease. The strong association between antibiotic exposure and CDI initially raised the idea that the microbiota is inextricably linked to acquisition of disease, presumably because of the loss of colonization resistance. Consistent with the epidemiologic data, characterization of the fecal microbiota of patients with CDI revealed that it is a markedly less diverse, dysbiotic community. Fecal microbiota transplantation (FMT)—the “transplantation” of stool from a healthy individual into patients with disease—was successfully used in the 1950s to treat four patients with severe CDI and has recently been demonstrated in numerous studies to be an effective therapy for recurrent CDI, with clinical cure in 85–90% of patients (as detailed below). Thus, FMT for recurrent CDI has become the “poster child” for the idea that microbiome-based therapies may transform the management of many diseases previously considered to be refractory to medical therapy. Although FMT is agnostic as to the underlying mechanism of protection, work is ongoing to identify specific microbes and host pathways that can protect against CDI. Studying mice with differential susceptibilities to CDI due to antibiotic-induced changes in their microbiota, investigators identified a cocktail of four bacteria (Clostridium scindens, Barnesiella intestihominis, Pseudoflavonifractor capillosus, and Blautia hansenii) that conferred protection against CDI in a mouse model. Intriguingly, treatment of mice with just C. scindens offered significant, though not complete, protection in a bile acid–dependent manner. Clinical data from patients who underwent HSCT also associated C. scindens with protection from CDI, an observation that suggests the possibility of translating these findings from mice to humans. This study provides another example of the identification of relevant bacterial factors through examination of microbial differences in populations that differ in disease risk.

Microbiome-related changes associated with Vibrio cholerae infection include a striking loss of diversity (largely due to V. cholerae’s becoming the dominant member of the microbiota) and an altered composition that rapidly follows the onset of disease. These changes, which occur in a reproducible and stereotypical manner, are reversible with treatment of the disease. This recovery phase involves a microbial succession that is similar to the assembly and maturation of the microbiota of healthy infants. In addition to V. cholerae, streptococcal and fusobacterial species bloom during the early phases of diarrhea, and the relative abundances of Bacteroides, Prevotella, Ruminococcus/Blautia, and Faecalibacterium species increase during the resolution phase and mark the return to a healthy adult microbiota. Analysis of these microbial changes occurring in patients with cholera and in healthy children led to the selection of 14 bacteria that were transplanted into gnotobiotic mice, which were then challenged with V. cholerae. Bioinformatic analysis of specific taxa changing during cholera determined that Ruminococcus obeum restrained V. cholerae growth. Subsequently, this relationship was experimentally confirmed, and the R. obeum quorum-sensing molecule AI-2 (autoinducer 2) was found to be responsible for restricting V. cholerae colonization via an unclear mechanism. These studies highlight the potential for use of microbiome-based therapies to prevent and/or treat infectious diseases. Moreover, they suggest that temporal analysis of longitudinal microbiome data may be an effective strategy for identifying microbes with causal relationships to disease.

HIV INFECTION

The augmentation of HIV pathogenesis by some viral, bacterial, and parasitic co-infections suggests that a patient’s underlying microbial environment can influence the severity of HIV disease. Moreover, it has been hypothesized that the intestinal immune system plays a significant role in regulating HIV-induced immune activation; this seems particularly likely since the intestines are an early site for viral replication and exhibit immune defects before peripheral CD4+ T cell counts decrease. Several studies have examined the intestinal microbiotas of HIV-infected individuals. Initial studies performed in nonhuman primates infected with simian immunodeficiency virus found no alteration in the bacterial components of the fecal microbiota; however, there were profound changes in the enteric virome. In contrast, many recent studies exploring this issue in patients have identified substantial differences in the HIV-associated fecal microbiota that correlate with systemic markers of inflammation. Curiously, these microbial changes do not necessarily normalize with antiretroviral therapy; this finding suggests that the microbiota may have some “memory” of the previously high HIV loads and/or that HIV infection helps reset the “normal” microbiota. This memory-like capacity of the microbiota has been demonstrated in animal models in the context of other infections and in response to dieting.

Given that the majority of new HIV transmission events follow heterosexual intercourse, there has been significant interest in examining the relationship between the vaginal microbiota and HIV acquisition. A longitudinal study of South African adolescent girls who underwent high-frequency testing for incident HIV infection facilitated the identification of bacteria that were associated with reduced risk of HIV acquisition (Lactobacillus species other than L. iners) or with enhanced risk (Prevotella melaninogenica, Prevotella bivia, Veillonella montpellierensis, Mycoplasma, and Sneathia sanguinegens). In mice inoculated intravaginally with Lactobacillus crispatus or P. bivia, the latter organism induced a greater number of activated CD4+ T cells in the female genital tract, a result suggesting that the increased risk of HIV acquisition associated with P. bivia may be secondary to the increased presence of target cells. In a separate study, the composition of the vaginal microbiota was shown to modulate the antiviral efficacy of a tenofovir gel microbicide. Although tenofovir reduced HIV acquisition by 61% in women who had a Lactobacillus-dominant vaginal microbiota, it reduced HIV acquisition by only 18% in women whose vaginal microbiota comprised primarily Gardnerella vaginalis and other anaerobes. This difference in efficacy was due to the ability of G. vaginalis to metabolize tenofovir faster than the target cells can take up the drug and convert it into its active form, tenofovir diphosphate. These findings illustrate how microbial ecology can be an important consideration in choosing effective treatment regimens.

RESPONSE TO VACCINATION

Second only to the provision of clean water, vaccination has been the most effective public health intervention in the prevention of serious infectious diseases. Its effects are mediated by antigen-specific antibodies and, in some cases, effector T-cell responses. Although vaccines are clearly effective on a population scale, the magnitude of the immune response to vaccines can vary among individuals by tenfold to a hundredfold. Although many factors (e.g., genetics, maternal antibody levels, prior antigen exposures) can affect vaccine immunogenicity, the microbiota is now recognized as another important factor. Analysis of the fecal microbiotas of ~50 Bangladeshi children identified specific taxa that exhibited positive associations (e.g., Actinomyces, Rothia, and Bifidobacterium species) and negative associations (e.g., Acinetobacter, Prevotella, and staphylococcal species) with responses to vaccines against polio, tuberculosis (bacille Calmette-Guérin), tetanus, and hepatitis B. A study of infants from Ghana revealed an inverse relationship between the fecal abundance of Bacteroidetes and a response to the rotavirus vaccine. Moreover, the nasal microbiota has been implicated as a factor that contributes to the IgA response to live, attenuated influenza vaccines. These correlations based on clinical data have been partially confirmed in animal studies. The best example is the demonstration that the responses to non-adjuvanted viral subunit vaccines (inactivated influenza and polio vaccines) are reliant on the microbiota, whereas the responses to live or adjuvanted vaccines (live attenuated yellow fever, Tdap/alum, an HIV envelope protein/alum vaccine) are not. An interesting note is that the antibody response to inactivated influenza vaccine is dependent on recognition of the microbiota by Toll-like receptor 5, presumably via flagellin-expressing microbes. These data suggest that the microbiota may serve as an adjuvant for certain vaccine types. Confirmation of these findings in clinical settings may suggest ways to improve vaccine efficacy in the future.

As highlighted in the examples above, numerous associations have been made between the microbiome and various disease states. These correlations have often been established at broad taxonomic levels, with little or no insight into causality. Given that most clinical studies of these relationships have a fairly small sample size (often <100) and are simultaneously comparing numerous variables (i.e., each of the bacterial species in the microbiota is effectively a different feature being compared), many of these studies may not be adequately powered and therefore may yield false-positive results. Testing of these correlations in animal models of disease has been critical in demonstrating a causal relationship between microbes and specific phenotypes. Because microbiome-wide association studies typically result in a long list of bacterial taxa that are correlated with a disease, it has been challenging to know which organism to test further in mechanistic studies. Moreover, even if a specific bacterial species is identified in these analyses, there is potentially enough strain-to-strain variation that the “functional” isolate may need to be recovered from the individuals studied; a publicly available representative of the species may not confer the same phenotype.

Despite all these difficulties, a handful of specific microbes have now been linked to disease effects; some examples have been mentioned above. The next layer of challenges relates to identification of the specific mechanisms that underlie these causal relationships. Successes along these lines have been more limited, but approaches are being developed to tackle the issue of defining specific bacterial factors and metabolites that are responsible for the phenotypic changes. Complicating factors are that many organisms, particularly those in the phylum Firmicutes, are not readily genetically tractable and that many of the phenotypes are not easy to assess with high-throughput screening.

BACTERIAL FACTORS

B. fragilis polysaccharide A (PSA) is perhaps the best-studied commensal-derived molecule that has been demonstrated to influence disease outcomes in mouse models. PSA—one of at least eight capsular polysaccharides expressed by B. fragilis—has a unique zwitterionic structure that incorporates both a positive and a negative charge within each repeating unit. Studies in which mice have been treated either with isogenic strains of B. fragilis that differ in PSA expression or with purified PSA have shown that PSA confers protection—prophylactically and therapeutically—against experimental colitis and MS. PSA is recognized by Toll-like receptor 2 on antigen-presenting cells, particularly plasmacytoid dendritic cells, and—in the setting of inflammation—induces interleukin 10 (IL-10)–producing regulatory T cells (Tregs) that help restrain inflammation.

B. fragilis is also the source of the only other microbiota-based bacterial factor identified thus far: an immunomodulatory glycosphingolipid that affects the numbers of invariant natural killer T (iNKT) cells. It is not clear whether these glycosphingolipids activate or inhibit iNKT cells; results have been discordant, probably because different glycosphingolipid species have been tested. Analysis of a specific purified glycosphingolipid (Bf717) demonstrated that it inhibits endogenous iNKT cell agonists in vitro and in vivo. Treatment of neonatal mice with Bf717 leads to a decreased number of colonic iNKT cells in adulthood and to improved outcomes in a model of colitis.

BACTERIAL METABOLITES

The use of mass spectrometry to detect and profile tens of thousands of different metabolites present in different bodily fluids has offered the promise of deeper insight into microbially mediated processes that underlie disease susceptibility. However, the fact that the overwhelming majority of these metabolites are not annotated, coupled with the sheer volume of data generated, has so far limited the general utility of these untargeted approaches. Instead, interest in more targeted approaches has increased, with a current emphasis on examining the role of short-chain fatty acids (SCFAs) and bile acids.

Short-Chain Fatty Acids

Several groups have demonstrated that SCFAs, the intestinal levels of which are largely determined by bacterial metabolism, are important for the induction of Tregs, though there is not agreement on which specific SCFA (propionate, acetate, or butyrate) is most relevant. Wild-type mice colonized with bacteria known to induce colonic Tregs have elevated cecal levels of SCFAs. Colonization with any of three Bacteroides species (B. caccae, B. massiliensis, and B. thetaiotaomicron) increases levels of acetate and propionate, whereas colonization with Parabacteroides distasonis or a mix of 17 human-derived Clostridium species elevates levels of all three SCFAs. In all of these cases, though, the SCFAs inhibit histone deacetylase, with a consequent increase in Foxp3 expression. Notably, microbe-induced SCFA production has not been shown to be critical for Treg induction by any of these organisms. In contrast, there appears to be no correlation between SCFA levels and Treg numbers in mice monocolonized with various Treg-inducing bacterial species. Taken together, these data suggest important heterogeneity in the mechanisms underlying Treg development and do not rule out the possibility of other, redundant mechanisms for Treg induction.

Bile Acids

Bile acids are produced in the liver but then are metabolized by intestinal bacteria to form deconjugated and secondary bile acids. These microbially produced bile acid profiles act through complex signaling pathways to balance the metabolism of lipids and carbohydrates and to affect immune responses. Therefore, bile acids are now being investigated as microbial metabolites that are critical to maintaining human health. As mentioned above, C. scindens helps protect mice against CDI through a bile acid–dependent process. Alterations in bile acid profiles due to underlying microbial dysbiosis have also been associated with hepatic and colonic inflammation, hepatic cellular carcinoma, colorectal cancer, and impaired gut motility. Almost all of these relationships have been documented at the level of correlation and, at best, reflect a partial change in phenotype in the setting of bile acid sequestrants (e.g., cholestyramine). Work is ongoing to determine causal relationships between bacterial metabolism of bile acids and changes in host physiology.

Other Bacterial Metabolites

Although most work has thus far focused on SCFAs and bile acids, a few notable examples of other bacterial metabolites have been implicated in maintaining health. ­Taurine enhances NLRP6 inflammasome–induced colonic IL-18 secretion, while histamine, spermine, and putrescine suppress IL-18 secretion. Desaminotyrosine produced by Clostridium orbiscindens confers protection from influenza by inducing type I interferon activity. In both of these cases, the microbiota was initially shown to influence the phenotype, with either untargeted metabolomics or a more targeted screen indicating a potential role for the indicated metabolites. Given the thousands of different bacterial metabolites throughout the body, many more metabolites will undoubtedly be linked to health and disease.

The numerous microbiome–disease associations identified thus far have generated a great deal of hope that understanding the relevant microbe–host interactions will open the door to unlimited therapeutic applications. Microbiome-based therapies offer several potential benefits. Patients often view such treatment as more “natural” than conventional drug therapy and are therefore more likely to comply with it. Biologically, microbiome-based therapies are more likely to address one of the root causes of disease (microbial dysbiosis) rather than simply affecting the downstream sequelae. Finally, a given microbiome-based therapy may serve as a “polypill” that is effective against several different diseases stemming from similar microbial changes. Despite tremendous interest in therapeutically exploiting the microbiome, there have thus far been few clinical successes along these lines.

The most successful therapeutic application of microbiome science has been the use of FMT, particularly for CDI. As mentioned earlier, FMT involves “transplanting” stool from a healthy individual to a diseased patient, with the idea that the “healthy” microbiota will correct whatever derangement may exist in the ill patient and therefore will alleviate symptoms. Fundamentally, this notion is agnostic as to the specific microbial dysbiosis and holds that any healthy microbiota will be curative. The idea of FMT dates back to at least the fourth century, when traditional Chinese doctors used a “yellow soup” (fresh human fecal suspension) to successfully treat food poisoning and severe diarrhea. The continued use of FMT through the centuries for the treatment of diarrheal illnesses in both humans and animals, along with the growing appreciation in recent years of the importance of the microbiota, laid the groundwork for using FMT to treat CDI. Since the first major prospective trial assessing FMT for recurrent CDI in 2013, most of the numerous studies of FMT for CDI have demonstrated remarkable efficacy, with an average clinical cure rate of ~85%. The donor stool can be fresh or frozen (use of the latter allows biobanking of samples from a limited number of pre-screened donors) and can be administered via nasogastric tube, nasoduodenal tube, colonoscopy, enema, or oral capsules; the cure rate is slightly higher with lower-gastrointestinal administration than with upper-gastrointestinal treatment. The optimal screening, preparation, and concentration of infused donor stool have not yet been determined. The most common adverse effects of FMT include altered gastrointestinal motility (with constipation or diarrhea), abdominal cramps, and bloating, all of which are generally transient and resolve within 48 h. At least 80 immunosuppressed patients have undergone FMT with no serious adverse events noted during 3 months of follow-up.

The successful use and the favorable short-term safety profile of FMT for CDI have led to its expanded application for other indications. At the end of 2017, 190 trials (listed at ClinicalTrials.gov) were investigating the efficacy of FMT for a range of indications, including CDI, IBD (ulcerative colitis and Crohn’s disease), obesity, eradication of multidrug-resistant organisms, anxiety and depression, cirrhosis, and type 2 diabetes. The few published studies regarding indications other than CDI have generally included small sample sizes and have offered mixed results. In contrast to the successes in CDI, the results have been more varied for patients with IBD, which is perhaps the second-best-studied indication. It is not clear whether these discrepancies are due to heterogeneity in recipients (e.g., in terms of underlying disease mechanisms or endogenous microbiotas), the donor material, and/or the logistical details of FMT administration (e.g., route, frequency, dose). However, these results demonstrate that—under the right circumstances—modulation of the microbiota can be an effective therapy for IBD.

Although FMT offers an important proof of concept that microbiome-based therapies can be effective, treatment is difficult to standardize across large populations because of variability among stool donors and among the endogenous microbiotas of recipients. In addition, FMT is fraught with safety concerns, and its mechanism(s) of action are unclear. FMT likely represents the first generation of microbiome-based therapies; subsequent generations will include the use of more refined bacterial cocktails, single strains of bacteria, or bacterial metabolites as the therapeutic intervention. The field of probiotics has a complicated history: many different strains have been tested against a multitude of diseases. Several meta-analyses have combined results across bacterial strains and/or disease indications and have generally concluded that the data are not yet convincing enough to support the use of the tested regimens. It should be noted that the tested organisms have been chosen mainly on the basis of their presumed safety profile rather than in light of a plausible biological link to disease. The hope is that more focused, mechanistic microbiome studies will identify specific commensal organisms—and their underlying mechanisms of action—that are involved in disease pathogenesis and that will serve as the basis for the next wave of rationally chosen probiotics, a few of which are currently in clinical trials. The main hurdle in this endeavor has been identifying specific microbes that are causally related to protection from disease.

The medical view of microbes has changed radically, moving from the early-twentieth-century notion that we are engaged in a constant struggle with microbes—an “us-versus-them” mentality that focused on the necessity of eradicating bacteria—to the more recent understanding that we live in a carefully negotiated state of détente with our commensal organisms. Instead of holding a simple view of microbes as enemies to be eliminated with antibiotics, scientists are increasingly recognizing the critical role these organisms play in maintaining human health; loss of these host–microbe interactions in the increasingly sterile environment typical of Western civilization may have predisposed to the increased incidence of autoimmune and inflammatory diseases. The field of microbiome research has made great strides over the past decade in cataloguing the normal microbiota and is now on the cusp of being able to identify clinically actionable microbe–host relationships.

The recent explosion of “–omics” technologies (e.g., metagenomics, metatranscriptomics, metabolomics) has enabled the generation of vast amounts of data, but it is not yet clear how best to integrate datasets in order to gain useful insights into host–microbe relationships. The use of FMT has demonstrated that modulation of an individual’s microbiota can effectively treat certain diseases; however, models with which to predict specifically how a microbiota will change after modulation—and what potentially untoward effects these changes might have—are still lacking. Implicit in this limitation is our ignorance about what microbial configuration is optimal and how a given microbiota should be rationally altered to obtain an ideal outcome.

Despite initial hyperbolic hype and a few false starts, microbiome research now stands at the precipice of an ability to treat the fundamental basis of many diseases. As the field continues to mature, it will need to move beyond correlations and address causation. The identification of causal microbes and their mechanisms of action will create a “microbial toolbox” from which relevant bioactive strains can be chosen on a per-patient basis to correct specific underlying microbial dysbioses. In the near future, our knowledge base regarding the microbiome and its relationship to health and disease will be robust enough that this information can be applied in making important treatment decisions.

Additional Online References