Methods in Microbiome Research: Past, Present and Future

Abstract

The human microbiome is impressively immense and participates in many aspects of our health and wellness, particularly involving development and maintenance of a healthy immune system. Not only do our microbes teach the immune system to fight infection, they also teach immune tolerance and help maintain homeostasis. From this knowledge, we have learned that the loss of tolerance to microbiota in both innate and adaptive processes play an important role in immune-mediated and autoimmune disease. In this chapter, we will discuss methods used to study the microbiome, both old and new, fundamental concepts that have taken hold within the field, and how these principles relate to rheumatology including thoughts on how microbiome research may be focused in the next decade.

1). Introduction: Historical Perspective

Microbiome refers to the community of microbes that exist on and in a species as well their collective genes. The terms ‘microbiota’ and ‘microbiome’ are often used interchangeably but some argue that the microbiota are the organisms themselves and microbiome includes the organisms and their collective genomes. These microbes live on our skin, along our gastrointestinal tract, at other mucosal surfaces, and sometimes even inside our cells. The majority are bacteria, but microbiota also includes fungi, archaea and viruses (1). Human intestinal tracts have been estimated to contain up to 100 trillion (10^14) microbes which is about ten times the number of cells in the human body and it has been suggested that these microbes encode 100 times more genes than the human genome (1–3). For most of the earth’s 4.5-billion-year history, microbes were the only living organisms in existence. They have been around long before humans existed and as a result, we have evolved among them – and even from them. Therefore, though the term ‘microbiome’ is a relatively recent one, it is not surprising that the concept of the microbiome is centuries old.

The first microscopes were developed as early as 1675 when scientists discovered the existence of the microscopic creatures that we now call microbes. As these tools improved over the next 200 years, scientists like Pasteur and Koch taught us about the importance of these creatures as ‘germs’ or pathogens discovering the causes of leprosy, gonorrhea, typhoid, tuberculosis, cholera, diphtheria, tetanus and plague. Nevertheless, even Pasteur thought that bacteria were important to life (4, 5). Around the turn of the 20^th century the idea of symbiosis was coined with newspapers at the time reporting on ‘good germs’ that could nourish soil as well as make alcohol and yogurt (4). However, most of the early 20^th century was spent trying to kill microbes with the discovery of antibiotics.

In the 1960s Carl Woese began to analyze 16S ribosomal RNA (16S rRNA) genes, which encode a component of the 30S small subunit of ribosomes in prokaryotes. These genes have both highly conserved as well as variable regions and therefore can be used to accurately determine microbial phylogeny. Woese profiled 16S rRNA genes from many different bacteria, changing the field of microbiology as we know it, and creating the field of molecular phylogenetics (4, 6, 7). This paved the way for the idea of metagenomics, or the genomics of communities, by direct extraction and cloning of DNA to capture the yet-uncultured microorganisms present among us and provide insight into their function within the community (8).

Characterizing the diversity and complexity of the microbes to which we play host is daunting and, at present, cataloging this diversity remains an incomplete task. However, recent technological innovations that we will discuss here have enabled us to begin to understand the role of microbiome in human health and disease. While gut microbiota are required for host immune education and colonic nutrition, microbial ‘dysbiosis’ or maladaptation, has been linked with complex genetic immune-mediated disorders such as inflammatory bowel disease (IBD) and spondyloarthritis (SpA). In this chapter, we will discuss the advancements in gut microbial profiling and its impact on understanding disease development.

2). Bacterial Bar Codes – 16S rRNA Sequencing

In 1977, Woese and Fox suggested that the 16S rRNA gene could be used as a marker in taxonomic identification (6, 7). Profiling microbial communities in a culture-independent manner enabled the identification of gut microbiota that are highly refractory to culture-based assays. This method utilizes selective binding of universal primer pairs to highly conserved sequences within the hypervariable regions of the 16S rRNA gene (such as V3–V5), enabling amplification and sequencing of a region that contains sufficient phylogenetic information for taxonomic classification. This information is used to determine operational taxonomic units (OTUs) having 97% sequence identity with a bacterial taxa.

Sequence-based clustering is often guided by bacterial reference genomes; however, common methods include de novo clustering to identify previously unknown species (9). Importantly, natural genetic variations are not easily distinguished from sequencing errors, which affect about 0.1% of all sequenced nucleotides, despite the use of the mentioned highly accurate sequencing platforms (9). Therefore, it is important to consider and account for incorrect base calls in analyses. Statistical denoising methods such as Deblur or Dada2 are available and implemented in QIIME2, an open-source software for 16S rRNA sequence analysis (10, 11).

The 16S rRNA gene sequencing approach is the most utilized method in the field of microbiome research to date. Costs are moderate, enabling large-scale microbiota analyses on relatively modest budgets (9). In addition to low costs, there are advantages to 16S rRNA gene sequencing. First, 16S genes are universally distributed with a high abundance exceeding other bacterial genes, thus facilitating sequencing and measuring relationships across taxa (12). Further, there is a low risk of horizontal 16S rRNA gene transfer. Disadvantages of using the 16S rRNA approach include the inability to detect variation among some strains and sub-strains; copy numbers per genome can vary in a taxon-specific fashion. There is also a risk of PCR amplification bias resulting in overinflated or inaccurate diversity estimates. Because 16S rRNA gene variation is sometimes not sufficient to differentiate closely related species, it provides limited information compared to whole bacterial genome sequencing or metagenomic sequencing (12, 13).

When considering microbiome sequencing studies, it is important to note the spatial heterogeneity of community profiling along the intestinal tract (14). Using fecal samples to provide a community profile delivers a valuable picture of the diversity of the gut microbiota, however focusing on abundance measurements within the feces neglects the importance of mucus and tissue-associated organisms (14). Bacteria can access and adhere to the epithelium only if they have the ability to cross the mucus layer and bacterial adhesion influences the composition of the gut microbiota, particularly in the small intestine (14). Futhermore, sampling only the feces does not account for spatial distributions along the GI tract due to the variety of distinct microbial habitats (14). The cecum and colon contain the most diverse and dense microbial communities within the entire body (14). Whereas, the small intestine, which has more oxygen, lower pH and more antimicrobials than the colon, is host to fast-growing facultative anaerobes that can tolerate these conditions (14). Future work considering the microhabitats within the gut will have an essential impact on our understanding of microbial function as it relates to health and disease.

3). Bacterial Genome Sequencing – Metagenomics

Metagenomic sequencing, which refers to random sequencing of sheared DNA fragments that are then realigned, generally affords a more comprehensive taxonomic and functional analysis of the entire genomes of viral, bacterial, and eukaryotic microbiota (9). This method captures the entire genome of organisms instead of focusing on fragments of 16S rRNA genes (9). Therefore, shotgun metagenomic approaches provide information about phages, viruses, archaea, fungi and other eukaryotes, as well as plasmids and other extra-chromosomal elements in addition to host, chloroplast and mitochondrial DNA. As a result, this method requires much more sequencing reads to obtain the depth necessary to characterize rare microbiota. The data generated often reaches several terabases per study, which adds to the costs and increases bioinformatic demands on sequence assembly, mapping and analysis (9, 15). However, due to the possibility of characterizing a microbiome with taxonomic resolution down to individual strains, this method is growing in popularity.

Since shotgun metagenomic sequencing provides complete genetic information about the microbes, it can reveal a functional profile of the microbe as well as specific taxonomic information. The microbes in our gut are regularly undergoing transcription and translation and interacting with each other as well as the human intestine. Therefore, sequencing of their genetic material can focus on phylogenetic markers (e.g., 16S rRNA) or expression of specific traits like enzyme activity (8). Using functional metagenomic techniques, the distribution and redundancy of functions in a community can be inferred and insights can be made into how microorganisms form symbioses, how they compete and communicate with other microbes, and how they use and produce energy (8).

Metagenomic data are analyzed by either comparing reads individually to reference databases, or after de novo assembly (8, 15, 16). One major disadvantage is that using a pre-existing database does not allow for characterization of previously unknown microbiota. And, de novo assembly requires substantial sequencing depth; therefore, this is restricted to only highly abundant species or strains within the microbial community (15, 16). This is particularly challenging if the sample being analyzed does not have a large microbial load. Therefore, the major disadvantages of metagenomics include cost and the risk of missing previously uncharacterized species or strains of bacteria (9).

Interestingly, there are statistical methods which use the more affordable 16S rRNA sequencing data to predict metagenomics and provide functional information from taxonomic and phylogenetic data. These statistical tools (phylogenetic investigation of communities by reconstruction of unobserved states, or PICRUSt) offer predictive inferences as opposed to measured data but have been found to produce accurate data (17). PICRUSt uses an ancestral-state reconstruction algorithm in order to predict which gene families are present and then uses this predicted information to estimate the metagenome (17). This method has been recently used by Gill and colleagues in an HLA-B27 induced experimental spondyloarthritis model in rats linking specific microbes to dysregulated cytokine pathways (18).

4). Metabolomics

Microbiota produce and consume metabolites that serve as a crucial link between microbial function and host physiology, in both health and disease states. Metabolomics, as applied to microbiome research, aims to comprehensively identify and quantify the metabolic consequences of microbe-host interactions (19). Nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry are the main technologies utilized in this field. Metabolic profiling can be targeted or non-targeted for discovery purposes. Targeted analysis of a disease state determines perturbation of metabolites associated with disease and will focus on a specific class of compounds (e.g., amino acids, fatty acids, lipids, carbohydrates, or bile acids). Non-targeted discovery approaches provide a greater overview of metabolic differences that may be a consequence of quantitative or functional differences between microbes in a community and can lead to discovery of unexpected biomarkers or therapeutic targets (19). Best practice recommendations for microbiome analysis, including laboratory and bioinformatic procedures are available, for example, from the U.S. Microbiome Quality Control project (20).

5). Host Response to the gut microbiome

a. Barrier Function & Innate Immune Reactivity

Our understanding of how the immune system relates and responds to both commensal and pathogenic bacteria in the gut in healthy individuals is incomplete but has grown significantly over the last two decades. The conventional view has been that the gut barrier prevents dissemination of nearly all bacteria from the intestine and this view has been supported by rare detection of live commensal bacteria in extra-intestinal organs in an immunocompetent host (21). However, we now know that the gut and the microbiota within it also play a critical role in the development of a healthy immune system (22). In fact, the host response to microbes is directed in part by the gut barrier. The gut epithelial barrier serves as the interface between the microbes within the gut lumen and the host. The degree of epithelial permeability is regulated by the integrity of the epithelial cell layer and the basement membrane, the surface mucus layer, the autonomic nervous system, and by the secretion of host factors, such as defensins. Toll-like Receptors (TLRs), in addition to their significant role in defense against microbial infection, also play a role in prevention of epithelial injury and permeability (23). TLRs encompass a family of pattern-recognition receptors that detect molecular products of microbes, or microbe-associated molecular patterns (MAMPs) including lipopolysaccharide (LPS) and lipoteichoic acid (LTA) as well as muramic acid, capsular polysaccharides, flagellin and unmethylated bacterial DNA (24). By recognition of these MAMPs, TLRs sense the presence of microbes in order to initiate an inflammatory response (23). Following TLR stimulation, a cascade of signals is initiated which then leads to the release of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) which then activates transcription of many host genes including chemokines, cytokines and acute phase reactants (24). However, the molecular products recognized by TLRs, like LPS and other MAMPS, are not unique to pathogens but rather are shared by entire classes of bacteria, including commensals. TLR activation by commensal MAMPS will in some cases signal inhibition of inflammatory reactions which is crucial to maintain intestinal epithelial homeostasis and symbiosis between the host and its commensals (23–25). Some bacteria increase gut permeability, like Escherichia coli, whereas others can decrease permeability, like lactobacilli (26) potentially due to the different MAMPs they express, like LPS. Furthermore, there exist LPS subtypes which exhibit various levels of endotoxicity or capacities to elicit an innate immune response (27). For example, LPS from E. coli is a more potent immune activator than LPS from Bacteroides species due to subtle structural differences (28).

There is some correlative evidence for TLR activation by MAMPs playing a role in the pathogenesis of autoimmune disease (28). A study in 2016 characterized the development of the infant gut microbiome in populations from Finland, Estonia and Russia – who have different rates of autoimmune disease – and identified significant differences between populations of intestinal microbes with different LPS subtypes (28). This group proposes that early microbial communities producing different LPS subtypes might contribute to immune modulation and alter the course of autoimmunity by their differing abilities to initiate tolerance (28). The importance of TLRs in both gut homeostasis and immune defense underscores both the delicate balance the immune system must achieve in maintaining gut health and the potential role of the microbiome in development of autoimmune disease.

Some systemic autoimmune diseases, like inflammatory bowel disease (IBD), are characterized by increased gut permeability which then allows for mucosal penetration of MAMPS and antigens and may trigger an aberrant immune response as a result. However, altered intestinal permeability has been documented in healthy relatives of patients with Crohn’s Disease (CD) (29), which may suggest that increased permeability is only one factor that predisposes an individual to disease. Overproduction of IgG directed against commensals occurs in patients with CD which could either reflect an increase in gut permeability or a state of hyperresponsiveness to normally harmless bacteria (30). Whether these antibodies are a result of nonspecific bacterial translocation through a damaged and epithelial layer or reflect a primary pathogenic process is currently not known (26) and is an important question that is currently being explored in other autoimmune diseases as well. These immunologic observations, among many others, highlight the potential relevance of the microbiome in immune-related disease.

b. Host response to bacterial metabolites

An important concept in relating the microbiome to host health is the role of bacteria in digestion, creation of nutrients, and the resulting byproducts of their metabolism that influence our health. Human co-evolution with symbiotic gut microbes has provided us with beneficial microbial metabolites, required for the degradation of indigestible components including plant-derived pectin, cellulose, and resistant starches (31, 32). The anaerobic microbial communities of the mammalian intestine digest dietary fiber which is, in turn, used by the gut microbiota of the colon as their main source of energy (33, 34). Fibrolytic bacteria degrade polysaccharides into smaller carbohydrates, which are then fermented into short-chain fatty acids (SCFAs) such as acetate, propionate, and butyrate. These SCFAs are carried in the bloodstream to a variety of different organs, where they are used as substrates for oxidation, lipid synthesis, and energy metabolism particularly by hepatocytes, which use propionate for gluconeogenesis (35, 36). Additionally, butyrate can inhibit cell growth or promote cell differentiation, and it can induce apoptosis in tumor cells (34). And butyrate is a major source of energy for colonocytes and transported preferentially across the epithelium. However, all three SCFAs have demonstrated important immunomodulatory properties (33).

The main producers of butyrate belong to the classes Clostridia, Eubacteria and Roseburia (35). Butyrate has anti-inflammatory effects due to inhibition of NF-kappaB transcription factor activation which consequently reduces the formation of pro-inflammatory cytokines (34). Acetate, propionate, and butyrate inhibit TNFalpha-mediated activation of the NF-kappaB pathway and inhibit LPS-induced TNFalpha release from neutrophils (37). Furthermore, SCFAs can regulate neutrophil functions and migration, inhibit expression of vascular cell adhesions molecule-1 and increase expression of tight junction proteins in the colonic epithelia (35). Regulatory T cells (Tregs) expressing transcription factor Foxp3 have an important role in modulating inflammatory responses in the intestine and it has also been found that both butyrate and propionate facilitate extrathymic differentiation of Treg cells in mice (38). SCFAs help to regulate the size and function of the Treg pool within the colon and protect against colitis in mice (39). Accordingly, the notion that dysbiosis may be related to a disrupted ratio of fermenter versus non-fermenters (i.e., bacteria that do and don’t ferment polysaccharides to create SCFAs) as part of the pathogenesis of IBD has been proposed. Due to these anti-inflammatory properties, SCFAs have been explored as options for treatment in IBD and colon cancer (40). Recent rodent models of inflammatory disease, like experimental autoimmune encephalomyelitis and allergic airway disease, have revealed evidence that SCFA administration can modulate or inhibit disease by increasing Tregs (41, 42). Ultimately, the immunomodulatory properties of these microbial and dietary metabolites might be exploited for the treatment of inflammatory disease in humans.

c. Adaptive Immune Reactivity

i. T cell responses

The adaptive immune response is crucial in the maintenance of intestinal immune homeostasis. Germ-free mice fail to develop a normal immune system, but rather develop lymphoid architecture with abnormalities of multiple cell lineages including B cells, T cells (Th17, Treg, CD8+ T), and Th1/Th2 balance (43). The gastrointestinal tract microbiota have been shown to affect differentiation of T cells into various T helper cell types, like Th1, Th2, Th17 or Tregs (24). The Th17 population secretes cytokines, like IL-22 and IL-17A, which have an impact on immune homeostasis and gut epithelial inflammation (24). The purified capsular polysaccharide of the commensal Bacteroides fragilis suppresses production of IL-17 and stimulates T cell production of IL-10, thereby protecting colonic epithelium from inflammation due to bacterial antigens (24). The colon also promotes expansion of Tregs, mediators of immune tolerance, (in part due to SCFA production as previously discussed) which limits an inappropriately high inflammatory response (24).

CD4+ T cells in particular, discriminate between harmless microbes (e.g., commensals) versus toxic ones (e.g., pathogens) (44, 45). CD4+ T cells establish a well-honed crosstalk with the gut microenvironment to maintain tolerance yet simultaneously and efficiently eradicate local and systemic infections (44, 45). In order to generate tolerance to commensal antigens, CD4+ Tregs dampen inflammation through release of anti-inflammatory cytokines, like IL-10 and TGFβ (44, 45). To this end, they orchestrate the immune response via release of either pro- or anti-inflammatory cytokines and expression of co-stimulatory molecules as well as driving or repressing a response of macrophages, CD8+ T cells or B cells (44, 45). Importantly, the breakdown of these mechanisms may play a role in chronic inflammatory disease.

ii. Sequencing IgA-Targeted Microbes (IgA-Seq)

The gut microbiota also induce local secretory immunoglobulin A (sIgA) which contributes to gut homeostasis and protection from enteric pathogens (46). IgA is the most abundant immunoglobulin isotype produced in mammals, largely secreted across mucous membranes and by extension is a critical mediator of intestinal immunity (46). Secretory IgA production is induced by antigen capture within Peyer’s patches M cells, subsequent activation of T cells and then B cell class switch recombination within mesenteric lymph nodes and gut-associated lymphoid tissue (GALT). Various cytokines, including IL-10, stimulate sIgA production and are crucial in maintaining mucosal tolerance, providing a link between sIgA, immunity and intestinal homeostasis (24). A T cell-independent IgA response to commensal bacteria is also locally induced in the intestine under homeostatic conditions (47, 48).

IgA is also an effective marker of potentially inflammatory microbiota, not just commensals. The intestinal immune system will recognize enteric pathogens and thereby produce high-affinity, T-cell-dependent, pathogen-specific IgA, which is transcytosed into the intestinal lumen. In the lumen, IgA can then bind and ‘‘coat’’ offending pathogens and provide protection against infection through neutralization and exclusion (49). Due to the overwhelming diversity of the mammalian microbiome and difficulty in associating specific microbes with disease, another experimental methodology was identified which has been termed “IgA-Seq”. The IgA-Seq method takes advantage of these differences in IgA binding between different microbes. Using flow cytometry, one can identify fluorescence intensity shifts of IgA-bound microbes compared to those not bound to IgA and thereby identify a specific taxon or group of taxa that is potentially related to dysbiosis. The microbes in the high-binding gate and low-binding gate are separated using flow cytometry sorting or magnetic-activated cell sorting (MACS) and the two different fractions are analyzed for taxonomic differences using either 16S rRNA sequencing or shotgun metagenomics. This approach has identified inflammatory taxa in various disease states in both mice and humans (49).

However, despite IgA being the predominant antibody isotype produced at mucosal surfaces it only interacts with a fraction of the intestinal microbiota (50–52). Furthermore, normal flora within the intestinal microbiota, or commensals, also can stimulate IgA production and can become coated with IgA (50, 51, 53). As compared to pathogen-induced IgA, commensal-induced IgA is thought to be of low affinity and specificity (50, 51). Nevertheless, the commensal bacteria bound by IgA are poorly characterized and the type of humoral immunity they elicit remains elusive. One study, using bacterial flow cytometry coupled with 16S rRNA gene sequencing (IgA-Seq) in murine models of immunodeficiency, identified IgA-bound bacteria to elucidate mechanisms of commensal IgA targeting. They found that residence in the small intestine, rather than bacterial identity, dictated induction of specific IgA (52). They reported that commensals elicited strong humoral T-independent immune responses that originated from the B1b lineage in particular. Atypical commensals including segmented filamentous bacteria and Mucispirillum evaded T-independent responses but elicited T-dependent IgA. These data demonstrate targeting of distinct commensal bacteria by multiple layers of humoral immunity, and suggest IgA “coating” may be more suggestive of bacterial location along the gastrointestinal (GI) tract rather than indicate a role in pathogenesis of disease (52). Nevertheless, the use of IgA-Seq has helped isolate potential colitogenic bacteria in IBD. Noah Palm and colleagues aimed to characterize taxa-specific coating of gut microbes with IgA with the hypothesis that IgA coating can identify potentially inflammatory commensals that drive intestinal disease (49). Using this method, Palm and other groups have shown particular bacterial taxa as uniquely potent drivers of intestinal disease. However, it is important to note that associations between species, or even phyla, and dysbiosis in the setting of disease states are often unable to be replicated; this fact may suggest that the role of the microbiome in driving disease is complex and is likely related to multiple microbial factors and an ecological model of dysbiosis.

iii. Sequencing IgG-Targeted Microbes (IgG-Seq)

Immunoglobulin G (IgG) is not native to the gut and it is not clear whether human gut microbiota are capable of inducing a systemic IgG immune response under healthy homeostatic conditions. However, circulating IgG against commensals has been reported in healthy humans and at higher concentrations in those with Crohn’s disease and diabetes (21, 54–56). This higher IgG response to symbiotic bacteria in patients with disease may be due to increased gut permeability leading to increased bacterial translocation. However, there is also evidence from mouse studies that under healthy homeostatic conditions (in the absence of a “leaky gut”) some symbiotic gram-negative bacteria (Salmonella spp. and E. Coli) are able to disseminate systemically and induce an IgG response. This IgG was found to directly coat the bacteria to promote killing by phagocytosis, and thus confer protection against systemic infection. This suggests a possible compensatory protective role for gut bacterial translocation (21). Recent work suggests that mice generate broadly reactive T-independent IgG to commensals with the help of TLRs (57). Using a similar approach to IgA-Seq to characterize the antibody response to the microbiota, this group used murine feces as the microbial source and stained the bacteria with paired sera (from the same mouse). This allows profiling of the complete antibody response rather than restricting analysis to antibodies present within the lumen that are bound to commensals. Using isotype-specific secondary antibodies, you can distinguish between IgA, IgM and IgG binding. These mouse models have shown that fecal samples with paired serum demonstrate IgG binding of the microbiota without IgG ever entering the gut, suggesting that the microbes or a portion of the microbes have translocated the gut barrier (57, 58). However, whether this approach will be useful in humans is not yet known.

6). Important Concepts in Microbiome Research

a. Richness, diversity & redundancy

The human microbiome is both vast and diverse. The gut is dominated by four major bacterial phyla: Bacteroidetes, Firmicutes, Proteobacteria and Actinobacteria (33). However, an individual’s gastrointestinal tract likely contains on the order of 10¹⁴ microorganisms with over 2,000 species and 12 different phyla represented (3, 24, 59, 60). Any given fecal sample, for example, contains at least 1,000 species of bacteria; these species vary among individuals. In one of the largest studies of 16S rRNA sequence diversity in a gut community, investigators revealed more than 395 bacterial phylotypes in samples from 3 subjects (61). As described by Jeff Gordon and his group, the gut microbiota can be visualized as a grove of eight palm trees (or divisions) with deeply divergent lineages represented by the fans of closely related bacteria at the top of each trunk (31). This diversity is often measured two ways: alpha and beta diversity. Alpha diversity, when described by an OTU count, provides information about how many different species are present in one ecosystem, or species richness. The Shannon alpha diversity index also takes into account relative abundance, and therefore it measures diversity and evenness, or whether the microbes within the ecosystem are balanced or instead if there are species that dominate. Beta diversity provides information on how different one environment’s microbial community is as compared to another environment and generally focuses on taxonomic abundance between samples. The question of whether increased or decreased alpha diversity is associated with inflammatory disease has not yet been resolved. The data are mixed; differences may be due to divergent pathogenesis between diseases and/or alpha diversity may be too simplistic of a measure to characterize disease or wellness.

In order to characterize this profound richness and diversity, an integrated catalog of microbiota from various cohorts in Europe and the United States has been created (the Integrated Gene Catalog) which describes almost 10 million sequenced genes (1). While we do not yet have a global comprehensive microbial database, regional microbiome projects that are linked as a single global network, such as the International Human Microbiome Consortium, the European Commission’s Metagenomics of the Human Intestinal Tract project, the US National Institutes of Health’s Human Microbiome Project (HMP), and the Canadian Microbiome Initiative, are a step in this direction.

The HMP Consortium as well as other work has addressed the question of whether, despite the significant diversity among individual microbiomes, healthy individuals all share a core microbiome. Insights from these studies suggest that we do share many of the same microbes, but the relative quantities of species vary greatly among individuals (3, 59, 62). Qin et al demonstrated that 57 bacterial species were present in >90% and 75 species in > 50% of the individuals in their study (3, 63). But, despite these interindividual similarities, the idea of a universal healthy gut microbiome or a core microbiome is likely too simplistic in light of the interindividual differences observed in many other studies (64). Therefore, it is essential to sample a broad population of healthy individuals over time to uncover features of the microbiome unique to geography, age and cultural traditions.

Additionally, despite the significant diversity between individual microbiomes, some data suggest that the functional profile, or the relative abundance of different metabolic pathways, is quite even (62). This provides evidence for a functional core microbiome (ecological model) and suggests that microbial function is redundant (62, 63). However, despite these data, our knowledge of the genetic and functional diversity of the gut microbiota remains incomplete.

b. Inter-individual differences

i. The role of genetics

It is clear that both host genetics and the gut microbiome can influence metabolic and immunologic phenotypes in animals and humans (65). However, the relative influence of host genetics on the host microbiome is not known (65). The gut microbiome may function as an environmental factor that interacts with host genetics to shape a phenotype, or the microbiome itself may be genetically determined and subsequently influence the host (or both). Gut microbiomes differ significantly across adults(3, 59) but family members have more similar microbiota than individuals who are not related (62, 66). These familial similarities are often attributed to similar environments (e.g., cohabitation & diet). However, when these factors are controlled for, family members have more similar microbiota profiles than unrelated individuals; therefore, genetics does appear to play a role (62). Animal and twin studies provide an interesting way to examine this question. One study providing evidence for a role of genetics reported that the concordance rate for carriage of the methanogen Methanobrevibacter smithii is higher for monozygotic (MZ) twins than dizygotic (DZ) twins (67). A highly powered twin study comparing MZ and DZ twins found certain taxa that were likely “heritable” compared to other taxa that were more likely to be shaped by environment (65). They found that the most highly heritable taxon was Christensenellaceae, a member of the Firmicute phylum. Members of the Bacteroidetes phylum were not found to be heritable and likely more influenced by environment (65). Other studies have compared known genetic loci differences between humans (e.g., different NOD2 risk alleles) and shown differences in relative frequencies of bacterial taxa between these groups (68–70).

In a similar study design to Goodrich and colleagues (65), albeit less powered, other groups have reported, in contrast, that characterization of the fecal microbiota of adult female monozygotic and dizygotic twins using 16S rRNA sequencing has shown that an individual’s GI microbial community will vary in specific bacterial lineages present but the degree of variation between adult twin pairs is comparable between monozygotic and dizygotic twins. Therefore, they suggest that the overall heritability of the microbiome is low, emphasizing the importance of environment(62, 66).

ii. The importance of environment

1. Diet

There are many studies that underscore the importance of environment on the gut microbiota and several population-based studies have shown diet is a key determinant of microbiota variation among individuals (71–73). Studies of hunter-gatherer populations with seasonal variation in diet have shown changes in their gut microbial communities that parallel these cyclical dietary changes (71). For example, in seasons where more fiber is consumed, bacteria that specialize in complex carbohydrate digestion (Bacteroidetes) become more abundant (74). Mouse studies have shown that changing their diet to a high fat chow (a “Western” diet) will dramaticall alter the gut microbiome within hours (75).

As the prevalence of metabolic diseases, like diabetes and obesity, continue to rise and the complexity of these diseases are slowly unveiled, the role of the gut microbiota has been offered as an explanation for part of their pathophysiologic intricacy. Because the gut microbiota can affect energy balance, and as a result have an impact on body mass and metabolism, by extension, the GI microbiota can influence the efficiency of calorie harvest from the diet and how this energy is used and stored (75). In early seminal microbiome studies, Jeff Gordon and his group found that when germ-free mice are colonized with the microbial community of conventionally raised mice, they have a dramatic increase in body fat within two weeks despite decreased caloric intake (32). They also note that it is repeatedly found that obesity is associated with a significant decrease in the level of diversity of the microbiome as compared to non-obese individuals (75). Though their work highlights the influence the microbiota may have on development of obesity, it also demonstrates how diet dramatically affects the composition of one’s microbiome.

Both diet and geographical environment are two major determinants of the microbial structure of the gut and comparisons of rural indigenous populations versus industrialized populations shows significant differences in microbial community makeup (66, 76, 77). In a recent study from Vangay et al, the effects of immigration to the United States from Thailand on the gut microbiome was explored(78). They collected stool from individuals both before and after immigration and found that moving from a non-Western country to the United States is associated with loss of gut microbiome diversity with replacement of bacterial strains for local ones(78).

2. Cohabitation

In addition to diet, a major environmental factor that influences the microbiome is cohabitation. Song et al. compared microbiota of humans and dogs in 60 families and found by comparing human pairs, dog pairs and human-dog pairs, that co-habitation resulted in more similar microbiomes, particularly of the skin (79). They also found that the skin microbiome was similar between people of different ages (whereas this is not the case in the gut). Similarly, genetically unrelated couples living together are microbially more similar to one another than family members living apart (66, 80). One study found in analysis of spouse and sibling pairs that spouses have more similar microbiota to one another than to their siblings even after accounting for dietary factors (80). The phenomenon of the “cage effect” in murine studies is a pronounced example of this, where mice housed in the same cage will share microbiota due to coprophagia (81).

iii. What’s more important, genetics or environment?

The evidence clearly shows there are many factors that shape our microbial composition, but it is not clear which factor or factors dominate. Many studies support a greater role for the environment than genetics. In a 2018 study looking at over 1,000 healthy individuals of distinct ancestries in a relatively common environment, it was concluded that over 20% of the variability between individual microbiomes is associated with diet, drugs and body size, and that genetic ancestry played a minor role in gut microbial composition (82).

1. HLA-B27 models demonstrate dual role of genetics and environment

The initial studies linking microbiota to a rheumatic disease phenotype came from HLA-B27, human β₂-microglobulin (hβ₂m) transgenic (HLA-B27 Tg) rats in the early 1990s (83, 84). These animals develop joint, gut, and skin inflammation, and males develop epididymoorchitis, thus resembling a spondyloarthritis phenotype. However, in germ-free facilities neither gut nor joint inflammation is seen (84). Yet, re-colonization with normal gut flora, especially containing Bacteroides spp., was permissive for the gastrointestinal phenotype, whereas E. coli was not (85). These studies established the importance of environmental triggers, particularly gut microbiota, acting in the context of genetic predisposition to cause spondyloarthritis-like disease.

Subsequent studies in related strains of HLA-B27 Tg rats demonstrated that this allele along with hβ₂m can alter the composition of gut microbiota, raising the possibility that this mechanism might be involved in disease predisposition (86). These studies also revealed a greater abundance of Bacteroides vulgatus in HLA-B27 Tg rats compared to wild-type (WT) animals (85, 86). Since this work was performed using low copy number HLA-B27 Tg rats that do not develop disease, this suggested that changes in gut microbiota did not require the presence of overt gut inflammation (86). They also suggest that changes in Bacteroides vulgatus (and other microbiota) are not sufficient to elicit the inflammatory phenotype. A recent study has shown that effects of HLA-B27 on gut microbiota in experimental SpA in rats are highly dependent on host genetics and/or environment. They showed that while HLA-B27-induced microbial dysbiosis varied between strains housed in different facilities, the dysregulated immune pathways and cytokines were background independent (87). This suggests there may be inter-individual differences in patterns of dysbiosis in humans that are nevertheless disease-related and important.

c. Other variables

i. Early life microbial acquisition and delivery conditions

Infants begin to acquire their microbiota at birth, with the mode of delivery impacting which microbes will colonize their gut and skin (88). The gastrointestinal microbiota of infants delivered by C-section is significantly less similar to their mothers than vaginally delivered infants (64, 88). In contrast to vaginally delivered infants, the fecal microbiome of infants delivered via C-section, who lack vaginal exposure, was enriched in bacteria commonly found on the skin like Staphylococcus aureus and other Staphylococcus species as well as species of Streptococcus, among others, suggesting that these newborns are first colonized by microbes in the surrounding environment during delivery. Vaginally born infants are enriched in genera including Bacteroides, Bifidobacterium and Escherichia which are seen in the vaginal and perineal milieu of their mothers. These differences gradually decrease over the first year of life (64) but may have long-lasting consequences on susceptibility to certain pathogens, like methicillin-resistant staphylococcus aureus (MRSA), or on the risk of obesity (88). Recent findings also suggest that microbial colonization of the gut may even start before birth (24, 89, 90). The placenta demonstrates a unique microbiome profile with Firmicutes, Proteobacteria, Tenericutes, Bacteroidetes and Fusobacteria (90). Additionally, meconium of full-term infants has been determined to play host to multiple genera often found in the amniotic fluid, vagina, and oral cavity (89).

i. Longitudinal variation & development

Human colonizing microbiota are established early in life but can shift with changes in age, diet, geographical location, intake of food supplements and drugs, and likely other causes as well (64). Bacterial communities are structured primarily by their habitat, meaning the oral, gut, skin and vaginal regions have a distinct community (88). Longitudinal studies of the human gut microbiota have suggested that the adult human bacterial community is stable over time in the absence of perturbations (e.g., antibiotics). In a longitudinal study of five years, it was shown that the microbial community of an adult individual was largely stable and persistent (91, 92). However, the gut microbiota does undergo significant changes as humans move from infancy to old age.

In the TEDDY study in 2018, stool samples from 903 children were collected longitudinally over a few years revealing rapid changes in three phases: the “developmental” phase from ages 3–14 months, the “transitional” phase from ages 15–30 months, and the “stable” phase from ages 31–46 months (93). Breastmilk was the most significant factor influencing microbial structure, with breastfeeding associated with higher levels of Bifidobacterium species (93). The cessation of breastmilk was associated with more rapid maturation of the gut microbiome as marked by presence of Firmicutes (93). They also found the birth mode significantly influenced the microbiome of the developmental phase.

ii. Sex

Given hormonal differences between sexes as well as differences in body fat percentages, sex has been considered as another possible factor influencing the microbiome. One group used a mouse model of type 1 diabetes (NOD mouse) to examine the relationship of microbial exposures and sex hormones to development of autoimmune disease (94). They found that transfer of adult male gut microbiota to immature females not only altered the microbiota of the host, but resulted in elevated testosterone levels, reduced pancreatic islet inflammation, reduced autoantibody production and protection from development of type 1 diabetes (94). There is not much known in humans about the sex-microbiome relationship but recent data has suggested that gut microbial diversity may be inversely associated with the android fat ratio (central adiposity) (95). This group suggested that certain taxa may be associated with fat distribution in men and women, helping to link obesity to the gut microbiome (95).

d. Dysbiosis

Dysbiosis is a condition where the normal microbiome population structure is disturbed, generally due to extrinsic factors like medications, diet, and disease (96), or even genetic variants, and is in contrast to eubiosis. The composition and functional characteristics of a “healthy” microbiome, or eubiosis, have yet to be defined, but in theory could be described in many ways (63). For example, a “healthy” microbiome may be defined by stability, the ability to resist community structure change under stress, or how quickly the community returns to baseline after a stressor. It could be defined by an idealized, health-associated, composition. And additionally, it may be defined by a desirable functional profile (63). Dysbiosis is then the loss of these characteristics due to a stressor, like acute or chronic illness, or perhaps loss or gain-of-function of critical genes.

In individuals with dysbiosis, the host may become more susceptible to infection from both exogenous and endogenous sources, or immunotolerance may be affected leading to development of autoimmunity, allergy, and/or chronic inflammation (24). Dysbiosis can therefore be a cause or an effect of disease. In dysbiosis, immune responsiveness could be up- or down-regulated to promote the return to eubiosis. This could be achieved through specific effects of sIgA, less specific effects of innate immune effectors (such as defensins) or local environment changes (i.e., diarrhea) (24).

i. Ecological & single organism models

The observation that the gut microbiota play a role in conditions like IBD, specifically Crohn’s disease and ulcerative colitis, was proposed as early as the 19^th and early 20^th century with descriptions of multiple potential infectious agents as triggers (97, 98). Early evidence for this hypothesis includes the use of oral antibiotics to treat IBD, epidemiologic studies that demonstrate IBD clustering similar to the way some infectious diseases cluster, mouse models of colitis requiring the presence of bacteria, and diversion of the fecal stream from inflamed bowel associated with improvement in IBD (26, 99). Additionally, in contrast to healthy subjects, colonic biopsy samples from subjects with Crohn’s disease reveal a higher concentration and thicker mucosal layer of bacteria that associate with intestinal surface, suggestive of a dysbiosis (26). Despite significant evidence supporting the role of the microbiome in IBD, identifying bacteria that “cause” dysbiosis has been a challenge. While many microbes have been postulated to play a role, there is no definitive evidence or consensus that microorganisms are the cause (26).

Altered gut microbial composition, or dysbiosis, is also implicated in the pathogenesis of several chronic immune-mediated rheumatic diseases, from ankylosing spondylitis (AS) and other forms of spondyloarthritis (SpA) including psoriatic arthritis (PsA), to rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) (100–102). Similar to IBD, these studies have found dysbiosis characterized by perturbations in many species of bacteria rather than single or small groups of organisms. These studies are reviewed in greater depth in other sections of this issue, and will not be detailed here. However, it is interesting to note that dysbiosis in HLA-B27 transgenic rats that serve as an experimental system for studying SpA, conforms to an ecological model (87). In other words, the effect of HLA-B27 is very much dependent on the strain of animal used and the housing environment. The effect of HLA-B27 expression in different rodent strains is to cause substantial shifts in the composition of gut microbiota associated with gut inflammation. The changes in gut microbiota caused by HLA-B27 on three different genetic backgrounds (DA, Lewis and Fischer) are almost distinct – there is very little overlap – despite the immunological and inflammatory changes being almost entirely overlapping and dominated by activation of inflammatory pathways involving IL-23/IL-17, IFNγ, IL-1, and TNF). This study would suggest that different HLA-B27 positive individuals in the human population with different communities of gut microbiota are likely to exhibit different patterns of dysbiosis. By focusing on microbiota that are associated with HLA-B27 and disease across a large population may ignore important individual differences. These results implicate a model of dysbiosis involving multiple microbes contributing to the aberrant immune response, rather than a single or small number of microbes driving pathogenesis (18, 87).

While changes in communities of microbes are of importance, there is also evidence that a particular genus or species may be important. For example, Dialister prevalence in ileal biopsies correlates with disease activity in spondyloarthritis (103) in one study, while R. gnavus is over-represented in fecal samples from patients with spondyloarthritis, particularly in a sub-group with coexisting IBD (104). These and many other studies are laying the groundwork for evidence of the influence of the microbiota on disease by providing associations of microbial communities or species with disease states. However, association does not provide causation.

ii. Association studies evolve to mechanistic studies

Over the past few decades, clinical association studies are moving closer to the root of pathogenesis as the experimental techniques change. Methods previously discussed, like sequencing of IgG targeted microbes and metagenomics, used together, get us closer to pathogenesis than simply categorizing community structures and making associations between groups. However, even these methods still provide correlations. We are now in need of future investigations that will provide a deeper and more mechanistic understanding of how microbial dysbiosis influences disease pathogenesis.

Summary:

In the past few decades we have made leaps in our knowledge about the human interaction with and dependence upon the microbes that live within us. Though some of the current data appear contradictory, with various different microbes, from species to entire phyla, associated with pathology from study to study, as we develop our understanding of the functional microbiome, we are learning that these differences may not be contradictory at all. In fact, the ecological model of dysbiosis implicating multiple microbes at once may explain these differences. Multiple different microbes may not only contribute to an aberrant immune response but may also provide similar functions to the host and when these functions are disrupted in some way may precipitate disease. The ecological model of dysbiosis requires further study as it is currently simply a hypothesis, but we are now entering an era of microbiome research where we have more and more tools at our disposal. As we move toward examining which microbes activate the immune system using IgA- and IgG-coated microbe sequencing and toward a more functional exploration of the disrupted microbes associated with disease, we will be more able to fully understand how the microbiome impacts disease.

Practice Points:

• 16S rRNA gene sequencing is the most cost-effective sequencing approach and currently the most commonly used methodology to explore the microbiome; certain statistical analyses of 16S rRNA data (PICRUSt) can be used to obtain predictive inferences about metagenomics in addition to phylogenetic data

• Metagenomic sequencing is a more comprehensive and informative method of microbiome analysis with increased capture of species and strains as well as providing a functional profile of the microbes’ potential activities

• The gut epithelial barrier is highly regulated by various components of the immune system in order to maintain gut homeostasis and symbiosis between the host and its commensals while also recognizing pathogens

• The gut is a major site for development of immune tolerance, but it is not yet clear how much of a role the microbiome plays in breaking tolerance during the development and propagation of autoimmune disease

• Our growing understanding of how the immune system interacts with microbes in the gut has allowed for newer methodologies for microbiome analysis, like sequencing IgA- or IgG-targeted microbes, which may provide more functional information than 16S rRNA gene and even metagenome sequencing

• The human microbiome of the GI tract is rich, diverse and also redundant, and is affected by innate factors like genetics and sex but even more so by our environment, including diet, geographical location, antibiotic use, birth method and other factors

• The redundancy of the microbiome in addition to the different types of dysbiosis associated with disease in different studies suggests that dysbiosis is likely not related to perturbation of a small number of microbes but instead implicates an ecological model where multiple microbes with important functions contribute to an aberrant immune response

Research Agenda:

• The promising newer methodologies of microbiome analysis which implicate function over taxonomy must be further explored in human studies with an emphasis on pathogenesis and establishing cause and effect relationships rather than association studies

• We must work to develop multidimensional, interdisciplinary approaches to scrutinize the complexities of the relationship of the host to its microbiota

• Larger, better powered and controlled studies with improved control group matching are needed. Increased sampling along the GI tract due to the spatial heterogeneity of the gut microbiome is important

• We must work to continue to grow our current understanding of bacterial metabolites, like SCFAs, with the ultimate goal of using these metabolites to understand their role in development of or protection from disease and then further manipulate them to use as treatments for disease