Exploring Bacterial Attributes That Underpin Symbiont Life in the Monogastric Gut

ABSTRACT

The large bowel of monogastric animals, such as that of humans, is home to a microbial community (microbiota) composed of a diversity of mostly bacterial species. Interrelationships between the microbiota as an entity and the host are complex and lifelong and are characteristic of a symbiosis. The relationships may be disrupted in association with disease, resulting in dysbiosis. Modifications to the microbiota to correct dysbiosis require knowledge of the fundamental mechanisms by which symbionts inhabit the gut. This review aims to summarize aspects of niche fitness of bacterial species that inhabit the monogastric gut, especially of humans, and to indicate the research path by which progress can be made in exploring bacterial attributes that underpin symbiont life in the gut.

THE RELATIONSHIP BETWEEN THE GUT MICROBIOTA AND THE MAMMALIAN HOST

The digestive tract of mammals provides habitats for a variety of microbial species (1, 2), which, through interspecies relationships, form communities (microbiotas) that interact with host features to produce intricate physiological connections (3–7). The resulting composite of host and associated microbes has been termed the “holobiont” (8, 9). In the case of humans, a gradual assembly of the microbiota begins early in life and culminates in a community with reasonable compositional stability in the large bowel (colon) within a few years of birth (10–12). The relationship between the microbes and their human host is therefore long term, a characteristic of a symbiosis (13, 14). The climax community is mainly composed of bacterial taxa and has a characteristic, yet somewhat individualistic, composition in the adult human (15). The gut microbiota of adult humans has high diversity (contains many different bacterial species), which is considered to promote good functional resilience (16, 17). The microbiota of humans has been studied by means of effective, anoxic, culture-based methods since the 1970s and in depth, since the 1990s, by using bioinformatics analysis of microbial DNA sequences obtained by high-throughput sequencing methodologies (18–21). By these means, a relatively comprehensive view of the composition and metabolic capacity of the microbiota of the human colon in infancy and adulthood has been obtained through the investigation of fecal samples collected from thousands of healthy humans originating mostly in Western countries. Comparisons of the fecal microbiotas of healthy and ill people have also been made, which have sometimes indicated association of altered microbiota compositions with particular diseases (such as inflammatory bowel diseases) and conditions (such as metabolic syndrome) (22–25). Indeed, much of gut microbiota research has been driven by the search for evidence of “dysbiosis” (altered relationship between microbiota and human host) in relation to disease (26–29). While gut microbiota research has a strong speculative basis with respect to human health, there is no doubt that aspects of host physiology are influenced by the gut microbiota. These “microbiota-associated characteristics” (MACs) were identified in studies in which comparisons of the anatomy, physiology, and biochemistry of gnotobiotic (defined microbial status) animals were made (30–32). These animals were raised in the absence of microbes (germfree), in the presence of complex microbiotas (“conventional” or “conventionalized”), or ex-germfree animals with defined bacterial associates. Germfree, monogastric animals can be maintained in the absence of microbial associates, and therefore in the absence of MACs, indicating that the symbiotic association is not obligate with respect to the host. In contrast, in adult ruminants, whose diet contains a vast preponderance of plant glycans, the biochemical activities of the rumen microbiota located at the anterior end of the digestive tract provide practically all the nutritional necessities for herbivorous life (33) and, indeed, constitute an obligate symbiosis (33–35).

THE SYMBIOTIC CONTINUUM

The symbiotic relationship between humans and the gut microbiota requires further discussion. The term “symbiosis” encompasses a spectrum of ecological relationships between partner organisms, from obligate copartnerships and mutualism to commensalism and parasitism (13, 36). In all cases, one partner cannot exist within a given context without the other. A parasitic partnership involves harm to one partner, whereas in mutualism both partners benefit. Commensalism is a useful term in situations where there are obvious benefits to one partner whereas the other partner is (as far as can be currently discerned) neither harmed nor benefited. For much of the large number of bacterial species comprising the gut microbiota of an individual human, their precise symbiotic relationship with the host is unknown because beneficial or harmful interactions with the host have not been directly demonstrated by experiment. The situation is complicated by the trophic interrelationships that exist between members of the microbiota in which syntrophic interactions between microbiota members, such as hydrogen transfer and short-chain fatty acid (SCFA) assimilation, are important factors in the maintenance of the gut community (37–42). Clearly, the provision of a habitat in the gut by the human partner is beneficial to an individual microbial species even though the quid pro quo, if any, is not known. Overall, it is fair to conclude that the monogastric host-microbiota relationship contains multilevel, symbiotic relationships.

Since they make a home in the gut, we need to discover how particular bacterial species cope with conditions in the gut, fit into the interactive landscape, and so thrive in their specific habitat. We cannot hope to correct dysbiosis without acquiring this ecological information. This review will therefore explore the fundamental question as to how particular bacteria make a life for themselves in the large bowel of monogastric mammals, with special reference to humans. Of necessity, information drawn from the experimental study of nonhuman subjects and in vitro experiments will be utilized in this text because the study of colonization attributes of individual bacterial types usually requires defined, experimental systems for studies. These cannot readily be accomplished in humans. Overall, the outcomes of this review can provide a basis for future studies in this field, help to develop ecological concepts of life in the gut, and help extend gut microbiota research beyond the descriptive to experimental science.

PROSPECTING METAGENOMIC DATA TO LEARN ABOUT LIFE IN THE GUT

Sequencing of bulk DNA extracted from the fecal microbiota, and genome sequencing of cultured bacterial species, together provide a general overview of how bacteria live in the gut (43, 44). However, to live in the gut, the symbionts must first reach the digestive tract by means of human-to-human transmission. Mother-to-infant transmission is very important and is facilitated by the intimacy of the birth process (12, 45). Transfer from other humans is likely to occur as well, and this may pose problems for most gut bacteria because they are highly sensitive to exposure to aerobic conditions. Fascinatingly, using analysis of genomic (cultured taxa) and metagenomic data (the collection of genomes represented in a community) by means of a “spore gene signature,” Browne and colleagues (46) revealed the phylogenetically widespread existence of genes associated with sporulation within the human microbiota. Fifty to 60 percent of bacterial genera comprising the microbiota are likely to produce resilient spores. Thus, the transmission of a large proportion of the symbiont bacteria in the gut is probably facilitated by resistant, metabolically inert forms. In vitro testing of cultured representatives of the spore-forming bacteria showed that sporulation enhanced survival under aerobic conditions and that germination of spores was initiated by exposure to bile acids, particularly taurocholate (46). Hence, at least some of the sporing microbiota taxa can specifically trigger vegetative growth once environmental conditions characteristic of the human gut are encountered.

With respect to bacterial metabolism, pioneering metagenomic studies provided a broad overview, in terms of Clusters of Orthologous Genes (COG) and Kyoto Encyclopedia of Genes and Genomes (KEGG) assignments, of the biochemical pathways characteristic of the microbiota as an entity (15). The most abundant pathways identified by this work were housekeeping functions required for cellular life. More compelling was the information relating to the presence of genes encoding hydrolytic enzymes (glycoside hydrolase and glycosyltransferase families; CAZymes classification) that catalyze the degradation of plant glycans that are the principal sources of carbon and energy for the microbiota (15). Overall, the metabolic pathways as proportions of metagenomes were found to be relatively constant between human subjects, in contrast to the widely varying taxonomic compositions of individual microbiotas. Since the number of biochemical pathways associated with the fermentation of specific substrates that are common in the gut is limited in the bacterial world, there must be considerable metabolic redundancy among gut species; the same function can be performed by different bacterial species (47, 48). This explains the constancy of metabolic capacity despite taxonomic variation between individual microbiotas (“the players may change but the game remains”) (49). Metabolic redundancy, therefore, assists in maintaining a high level of functional resiliency in microbiotas (50, 51).

Reconstruction of microbial genomes from metagenomic sequences extracted from the environment is possible as initially demonstrated by Tyson and colleagues (52). They assembled nearly complete genomes from Leptospirillum group II and Ferroplasma type II from shotgun-sequenced DNA extracted from acidic mine waste that contained a simple microbial community. The metabolic pathways by which the microbes lived in their harsh environment could be inferred from the reconstructed genomes. With respect to the mammalian digestive tract, several thousand metagenome-assembled genomes have been obtained from cattle rumen samples (53). These “rumen-uncultured genomes” can be as much as 80% complete and provide a rich source of information concerning the capacity of rumen microbes to produce proteins involved in the degradation and utilization of plant-derived components of the diet. Recent metagenome-assembled genome investigations using sequence data from human feces confirm that some functional attributes of the microbiota are characteristic of specific human populations (for example, seaweed degradation in Japanese) as well as providing pipelines for the discovery of new taxa, many of which have small genomes with predicted auxotrophies (for example, requirement for exogenous fatty acids) (44, 54–57). Metagenomic strategies also offer the possibility to discover tightly regulated biosynthetic gene clusters (BGCs) that encode the production of secondary metabolites that may have value in human medicine (for example, antimicrobial substances) (58). In metagenomics studies with a narrow focus, the potential importance of bacteriocins, bacteriophages, bile salt hydrolase, biosynthetic gene clusters, exopolysaccharide, flagella, and type IV pili as factors that influence the ability of bacteria to colonize the gut has been investigated. Summaries of this work are given in Table 1.

TABLE 1.

Specific bacterial attributes investigated by metagenomic analysis

Attribute	Summary	References
Antimicrobials and other substances	Extensive searches of metagenomic sequences derived from human gut microbiota (and microbiotas of other body sites) reveal the common presence of biosynthetic gene clusters (BGCs; tightly linked sets of mostly nonhomologous genes participating in a common, discrete metabolic pathway). The genes are located together on the genome, and their expression is often coregulated. The gene clusters encode pathways for the synthesis of small molecule, secondary metabolites, some of which might be useful antimicrobial agents for treatment of infections or pathogen carriage. Hundreds of BGCs are present in gut metagenomes and are associated with a diversity of bacterial taxa. They include genes for saccharide biosynthesis (such as capsular polysaccharides of Bacteroides vulgatus and Bacteroides ovatus). A critical feature of BGC research is finding gene regulatory systems to enable expression of BGCs in commercial scale systems. Similar metagenomic screens can reveal the presence of antimicrobial resistance determinants that might provide evidence of horizontal transfer of these genes in ecosystems. Other bacterial properties of interest that are advanced by metagenomic screens include enzymes involved in tryptophan conversion to tyramine (a potential neurotransmitter) and enzymes to assist in degradation of plant substances that could be used in preprocessing biofuels. The importance of BGC-associated properties in maintaining stable bacterial communities is currently unknown but their common occurrence in bacterial genomes indicates a potential role in ecological fitness. Overall, these studies show the potential importance of the gut microbiota in biotechnological bioprospecting for useful “natural products.”	3, 58, 198–205
Bacteriocins	Bacteriocins are (at least in vitro) antibacterial molecules that are diverse in structure and are encoded by complex, variable biosynthetic gene clusters (BGCs). They are proteins or peptides that are classified as class I bacteriocins (ribosomally synthesized and modified posttranslationally) or class II which are ribosomally synthesized but unmodified. Analysis of gut metagenomic data indicate that the most common kind of bacteriocins that are detected in gut bacteria are lantibiotics (polycyclic peptide antibiotics that contain the characteristic thioether amino acids lanthionine or methyllanthionine) and sactipeptides (sulfur-to-alpha carbon thioether cross-linked peptides). They are greater than 10 kDa, and the biosynthetic genes occur mainly in Enterococcus and Lactobacillus genomes. Modulation of bacterial populations by bacteriocins in vivo has yet to be demonstrated.	206, 207
Bacteriophages	Bacteriophage genomes are frequently detected in metagenomic studies of the gut microbiota. Bacteriophages can be divided into viral classes and 280 of these are detected in fecal metagenomes of humans worldwide. Most of these genomes are derived from classes of nonenveloped DNA bacteriophages (dsDNA Caudovirales or ssDNA Microviridae), and most are lysogenic (dormant prophages). The most abundant bacteriophages infect Bacteroides and Clostridiales species. Extensive bioinformatic analysis shows that most bacteriophages in the gut have high host specificity (infect only a single bacterial species), but examples of bacteriophages with wider host ranges have been reported. Unlike other environments (such as the ocean), free bacteriophage particles are low in no. in the gut relative to the no. of bacterial cells, possibly because of rapid destruction/sequestration of free bacteriophage particles in the gut environment. Bacteriophages contribute to horizontal gene transfer in bacteria. Recent studies with gnotobiotic mice and defined microbiota/bacteriophage mixtures indicate that lytic bacteriophages can reduce the population sizes of target bacterial species but do not totally deplete the microbiota of them, and a proportion of the population is subsequently resistant to bacteriophage infection. Nontarget bacterial populations can also be affected because of nutritional interactions within the microbiota.	208 – 212
Bile salt hydrolase (choloylglycine hydrolase)	Bile salt hydrolases (BSH) associated with members of the gut microbiota catalyze the cleavage of amino acid residues (taurine, glycine) from conjugated bile acids synthesized in the human liver and released into the gut in bile. Unconjugated bile acids (cholic acid and chenodeoxycholic acids) in the colon are further metabolized by gut bacteria through dihydroxylation, dehydrogenation, and sulfation. Metagenomic screens of the gut microbiota indicate that BSH are encoded mainly by members of the Firmicutes, Bacteroidota, and Actinobacteria. BSH can be divided into four clusters by comparison of amino acid sequences, but the catalytic site is highly conserved in all of these proteins. Cluster 1 and cluster 2 BSH are mostly represented in Proteobacteria and Bacteroidota, whereas the remaining clusters are mainly represented in the Firmicutes. The expression of BSH genes by the fecal microbiota is related to dietary composition of the human host (higher in association with increased fat/protein). Conjugated bile salts tend to inhibit bacterial growth whereas unconjugated bile acids less so, and unconjugated forms are further metabolized by gut bacteria.	(213–216).
Exopolysaccharide	Exopolysaccharides produced by members of the gut microbiota have a role in the regulation of the host immune system, notably the development of immune tolerance due to the production of regulatory T cells. The capsular polysaccharides of Bacteroides fragilis have been especially well studied. The genome of B. fragilis contains genes encoding 8 different capsular types that are phase variable (the adaptive process by which bacteria undergo reversible phenotypic changes). It seems that the production of any one of the 8 capsular polysaccharides promotes ecological fitness in a gnotobiotic mouse model, but production of the full complement of polysaccharides is associated with greater ability to establish in the gut. An acapsular mutant strain is quickly eliminated from the gut of the mice.	217, 218
Flagellin proteins	Flagellin proteins are the structural units of the bacterial flagellum filament, associated with cell motility. Cell movement directed by chemotaxis and achieved by motility assists bacteria in nutrient foraging and should therefore be a fitness attribute. Metagenomes from the gut microbiota consistently indicate “cell motility” as a prevalent COG category, although at relatively low abundance. However, there is an unexpected depletion of flagellin biosynthetic genes and chemotaxis genes in gut metagenomes relative to other habitats. This may be related to the production by the host of anti-flagellin antibodies that immobilize bacterial cells, perhaps to minimize association of the bacteria with the mucosal lining.	219 – 222
Pili	Pili are protein, multisubunit structures associated with bacterial cell surfaces. They are of particular interest as mediators of adhesive interactions that may be interbacterial but also between bacteria and mammalian cells and so could be important colonization attributes. Two categories of pili have received particular attention with respect to gut bacteria. 1. Sortase-dependent pili have been described for lactobacilli, including L. ruminis, and bifidobacteria (B. breve and B. bifidum). 2. The genomic signatures of type IV pilus systems are detectable in about 30% of gut bacteria genomes. These pilus systems are associated with adherence, biofilm formation, substrate binding, twitching motility, and DNA transfer in various bacteria but exptl verification of these functions is so far limited, mainly to examples from pathogenic bacteria. However, adherence of Ruminococcus albus cells to cellulose seems to be mediated by type IV pili.	223 – 232

MEASURING BIOCHEMICAL PATHWAYS IN ACTION: METATRANSCRIPTOMICS, METAPROTEOMICS, AND METABOLOMICS

The detection of genetic loci encoding particular functions in bulk DNA does not necessarily mean that the genes are “active” (transcribed) in cells at the time of sampling (59). Gene transcription, in general, is tightly regulated according to need under prevailing environmental conditions. While a “gene catalog” may be maintained within a microbiota over long periods of time, many of the genes will be “switched on” by bacteria only under specific conditions. Thus, measuring the transcription of genes in temporal studies may be much more informative of the biochemical processes that are critical to life in a habitat such as the gut. Likened by Moran (60) to “eavesdropping on the microbial ecology,” “metatranscriptomics” is the analysis of the active gene set in the community by sequencing of mRNA molecules (gene transcripts) extracted from the microbiota. In essence, it is the extension of whole-transcriptome shotgun sequencing (RNAseq) from analysis of a bacterial culture to the whole microbiota and a successor to DNA microarrays as screening assays in transcriptomic studies (61). Metatranscriptomics identifies the subset of the metagenome responsible for carrying out specific functions under prevailing conditions, so it could be very useful in measuring, for example, the functional response of microbiota members when the composition of the diet is changed (62). These functional responses may not involve changes to the taxonomic composition of the microbiota. There is, however, little correlation between the abundance of a protein in the cell and the abundance of the transcript that encodes its synthesis. Several factors are responsible for poor mRNA-protein correlation, including posttranslational processing and regulation, random fluctuations in low-copy mRNA, uneven partitioning of macromolecules during cell division, and translation efficiencies. The main factor is the difference in half-life between proteins (about 20 hours) and mRNA (a few minutes). Therefore, the possibility of using analysis of bacterial proteins extracted from the feces (“metaproteomics”) is an attractive proposition in that the proteins will mostly have been produced in habitats in the colon. Analysis of these proteins might provide indications about the functioning of the microbiota in situ (63, 64). In contrast, some mRNAs detected in the feces may be due to stress responses of bacteria when feces are exposed to aerobic conditions after collection and before low-temperature storage.

“Metabolomics” was originally an investigative approach that utilized nontargeted, holistic, quantitative analysis of changes in the complete set of metabolites (the metabolome) in a cell in response to environmental or cellular alterations (65). The most common analytical platforms used in obtaining quantitative metabolomic data are liquid chromatography combined with mass spectrometry and nuclear magnetic spectroscopy. The scope of metabolomics has since been expanded to include not only metabolites within cells, but also extracellular metabolites. Metabolomic investigations of the microbiota reveal substances of bacterial origin to which the bowel community, bowel mucosa, and even systemic tissues may be exposed. Importantly, metabolomics offers the possibility of revealing alterations to the emergent properties of the microbiota. These alterations may be relevant to understanding the etiology of some human diseases or conditions (66, 67).

These three “-omics” approaches have, to date, been technically difficult to use and, in the case of metaproteomics and metabolomics, lack comprehensive databases to provide major conceptual advances about the microbial ecology of the gut microbiota. However, metatranscriptomics has revealed much detailed information about “niche differentiation” in the gut microbiota. Niche differentiation relates to the coexistence of potentially competing species in a community. The competitive exclusion principle states that if two species with identical niches compete, then one will inevitably drive the other to extinction (68). This problem is negated if the breadth of “realized niches” (actual activity in the community) of the species is reduced in the habitat relative to “fundamental niches” (potential capacity). For example, Limosilactobacillus (previously Lactobacillus) (69) reuteri 100-23 and Lactobacillus johnsonii 100-33 cohabit the murine forestomach. According to the niche exclusion principle, this should not be possible because both strains utilize the two main fermentable carbohydrates present in the stomach digesta: glucose and maltose. Gene transcription analysis, in vitro physiological assays, and in vivo experiments show that the two strains can coexist in the forestomach habitat because 100-23 grows more rapidly using maltose, whereas 100-33 preferentially utilizes glucose (70). Mutation of the maltose phosphorylase gene (malA) of strain 100-23 prevented its growth in maltose-containing culture medium, and resulted in the numerical dominance of 100-33 in the forestomach. The fundamental niche of Lm. reuteri 100-23 in the mouse forestomach can therefore be defined in terms of “glucose and maltose trophism.” However, its realized niche when L. johnsonii 100-33 is present is “maltose trophism.” Hence, nutritional adaptations provide niche differentiation that assists cohabitation by the two strains through resource partitioning in the mouse forestomach.

With regard to the human microbiota, Plichta and colleagues (71), using microarray technology, identified in situ transcriptional interactions between pairs of bacterial species in the microbiota of 233 fecal samples. Among 102 interacting pairs of species, transcriptional changes occurred that led to a reduced transcription of orthologous functions between the cohabiting species, suggesting a decreased overlap of transcribed functions in interacting pairs of species and thus niche differentiation at the transcriptional level. The use of the metatranscriptomic approach could also lead to an increased understanding of food webs in the gut ecosystem in which resources are shared between bacterial species occupying different trophic levels, from those based on complex glycans to oligosaccharides to simple carbohydrates that become available through syntrophic interactions (72). Thus, in the rat gut, Bacteroides uniformis utilizes inulin for growth, whereas Blautia glucerasea, Clostridium indolis, and Bifidobacterium animalis use fructo-oligosaccharide and monosaccharides derived from inulin hydrolysis (73).

In addition to the “-omics” investigatory methods mentioned above, the glycan substrates that feed gut microbes need to be analyzed (“glycomics,” glycobiology) in greater depth. As Amicucci and colleagues (74) have pointed out, the structure of plant glycans in foods is poorly advanced mainly due to the lack of rapid, high-throughput technologies. Plant glycans are diverse in chemical content (monosaccharide content), have different monosaccharide spatial arrangements (sequences), and different glycosidic linkages and stereochemistry. These features are variable between plant species even when considering general categories of polysaccharide (for example, “xylans”) (75). In-depth analytical techniques could enhance studies of the temporal degradation of glycans by bacterial consortia in the gut, reveal substrate preferences, and, most importantly, allow regulation of bacterial enzyme activities to be studied in synthetic and natural communities (76, 77).

PROSPECTING THE METAGENOME USING BACTERIAL ARTIFICIAL CHROMOSOMES

The construction and sequencing of large-insert clone libraries from bulk DNA, together with DNA shotgun sequencing, were used in achieving the goal of the human genome project (78). Large-insert cloning techniques utilize DNA molecules such as “bacterial artificial chromosomes” (BACs). They are not really chromosomes but are modified plasmids containing the origin of replication derived from the Escherichia coli F factor plasmid. The replication of the BACs is strictly controlled, keeping the copy number to one or two per cell. They are stably retained in the bacterial host cell and can replicate DNA inserts of up to 300 kb. They have a low level of chimerism because of two genes inserted in the plasmid: parA and parB (79). Thus, relatively large DNA inserts (generally between 80 and 200 kb) can be cloned from bulk DNA to produce extensive libraries that can be examined for biochemical activities associated with environmental bacteria. The huge investigative potential of BACs was revealed in the ground-breaking work of Beja and colleagues (80) who uncovered a novel light-driven proton pump (proteorhodopsin, a bacteriorhodopsin homolog) that was detected on a 150-kb BAC fragment originating from an uncultured marine bacterial group (SAR86). In addition to the identification of the proteorhodopsin gene by DNA sequencing of the insert, the predicted functional activity of the clone was confirmed in the E. coli surrogate host. Thus, the use of BACs can be used to mine metagenomic information about novel bacterial functions, but perhaps more importantly, to screen microbiotas for specific biochemical activities (“functional metagenomics”) (81). These activities might be relevant to the development of “natural product” remedies to diseases but also to understanding the ecology of microbial communities. The prototype research using BACs in gut microbiota research identified novel β-glucanase genes in the large bowel metagenome of mice (82). A community genomic library consisting of 5,760 BAC clones was prepared in E. coli DH10B. DNA inserts detected in 61 randomly chosen clones averaged 55 kb (range: 8 to 150 kb) in size. A functional screen was conducted, and three clones with β-glucanase activity were further investigated and characterized. Based on DNA homology, one of these clones was likely to have originated in a Bacteroides species. Screening of the same BAC clone library using a biofilm-forming assay identified two clones (40–50 kb inserts) that contained DNA probably originating in Bacteroides spp. The bacterial clones had increased adherence ability relative to the control (83). Mutation of the putative adherence-associated regions of the cloned DNA resulted in a reduction in the ability of the surrogate host to colonize the murine intestine. These two studies of a murine microbiota clone library demonstrate the potential technical capacity of the BAC approach for studying not only genomic features of the gut microbiota but also the functions required for life in the gut.

THE NEED TO EXPERIMENT WITH CULTURED MEMBERS OF THE GUT MICROBIOTA

A common plaint of microbial ecologists is the lack of cultured representatives of prokaryotes that are common in the environment (84, 85). This is especially true of marine and terrestrial ecosystems in which the laboratory cultivation of several key taxa has now been prioritized because they have a widespread distribution in the natural environment. The human gut microbiota is generally said to be composed of largely uncultivated bacteria because metagenomic studies indicate that there are taxa that do not yet have a cultivated representative (86). Additionally, the global diversity of fecal microbiotas has so far been underrepresented because African, Asian, and South American samples were not analyzed in any number until relatively recently (57, 87, 88). Fortunately, due mostly to the efforts of Moore and Holdeman, Finegold, and Mitsuoka and their colleagues in the 1970s, common, numerically predominant members of the human fecal microbiota have been cultivated and are extant in culture collections (18–20). A more recent effort to assemble a diverse collection of bacterial isolates from the feces of 11 humans has resulted in the Broad Institute-OpenBiome Microbiome Library (BIO-ML) that contains 7,758 isolates belonging to six abundant bacterial phyla. Metagenomic data of the donors have also been determined. The collection is open-access and is clearly a step in the right direction to obtain strains that could be used in laboratory studies (89). This approach could be emulated by others, including using donors inhabiting non-Western countries. Characterized bacterial isolates provide the basis for experimental studies of the interactions that occur between members of the gut microbiota. Nevertheless, low abundance species and species with high nutritional interdependency are still uncultivated, so this topic needs much attention in gut microbiota research (44, 90–92).

Culture of bacteria is important because it enables predictions from genomic studies about cell biology and physiology to be substantiated. About 90% of the prokaryote genome encodes proteins (93). Although metagenomic analysis of the gut microbiota provides an indication of the biochemical capacity of the microbiota, it is probably not completely reliable. This is because assigning function to the encoded proteins (annotation) mostly results from automated pipeline comparisons of sequences in databanks (94, 95). As a result, errors in the sequence may not be detected, the start of protein-coding regions may not be indicated correctly and gene naming may be spurious or may not be appropriate in the biological context (96, 97). Additionally, “unknown proteins” deduced from sequence annotation but that have no known function, and “orphan enzymes” (described in the literature and often cataloged in the EC database but for which a corresponding gene has not been validated by experimentation) comprise significant proportions of genome information (98). Enzymes revealed by annotation of genomic sequences (for example, glycosyl hydrolase families) may belong to categories that have only a single member with an elucidated structure and may be associated with a range of substrate specificities (99).

Pure cultures enable microbial traits to be studied in replicates, thus improving reproducibility and giving statistical confidence (100). Mutation of specific genes can be used to test the fitness of bacterial strains for life under conditions prevailing in gut habitats (101). Metabolic interactions between cultivated species can be tested in cocultures of bacterial strains that provide information about growth substrate utilization, competition, trophic interdependence, and adaptations to changing conditions in the habitat (102–104). Intriguingly, coculture experiments with taxa from the human gut microbiota (a member of the Ruminococcaceae and Bacteroides fragilis) led to the discovery that the latter species is a major producer of the inhibitory neurotransmitter γ-aminobutyric acid (GABA) (105). This observation is pertinent to the concept of the “gut-brain axis,” postulated to link emotional and cognitive centers of the brain with peripheral intestinal functions and which may be influenced by the gut microbiota (106). Therefore, coculture experiments can reveal not only novel bacterial growth factors but also the synthesis of biologically active molecules of significance to human health. Coculture studies are essential in defining the consortia (microbial assemblages that are enriched from a taxonomically diverse inoculum) that mediate the degradation and fermentation of specific plant glycans and other substrates in the human gut (107). Continuous cocultures in chemostats, and in multistage fermenters simulating the gut, are important tools because the bacteria can be maintained as stable populations (108, 109). Chemostats provide steady-state conditions (growth occurs at a constant rate and in a constant environment) in which all the bacterial cells have a similar metabolic state under conditions of carbon and energy limitation (110). Planning of culture-based studies needs to recognize strain-to-strain variation in relation to the utilization of particular substrates by members of a bacterial species. This is a well-known phenomenon in culture-based, taxonomic investigations where variable fermentation of carbohydrates is a feature of identification schema (111). Thus, researchers need to be aware of the “pan-genome” (the sum of all genes present in all strains, or the global gene repertoire) of a species which is composed of the “core genome” (the pool of genes shared by all strains) plus the “dispensable genome” (pool of genes present in only some strains) (112–114). With regard to the gut microbiota, Lactobacillus ruminis provides a well-described example of strain variation regarding the use of tetrasaccharides derived from barley β-glucan as growth substrates (115).

The development of multistrain cocultures in a liquid medium to investigate the enrichment of consortia with specific degradative/fermentative capacities taxes the traditional method of selective culture of bacterial strains on plates of solid medium to obtain differential CFU values. A possible solution is to label bacterial strains with constitutively expressed reporter molecules. β-d-glucuronidase-constitutively expressed promoter systems have been developed for bacteria (116–118), including the gut inhabitant Bifidobacterium longum, but the use of the diverse collection of fluorescent proteins, now available, as constitutively expressed tags might be more useful (119–121). Bacterial strains in a coculture could be differentially tagged using molecules with different excitation and emission wavelengths. This kind of technology utilizes fluorescence microplate readers and allows high-throughput experimentation (122). Novel reporters that fluoresce under anaerobic conditions will be essential in these coculture studies.

IDENTIFYING SPECIFIC BACTERIAL FEATURES OF THE BACTERIAL CELL THAT CONFER ECOLOGICAL FITNESS: EXAMPLES FROM THE HIGHLY HOST-ADAPTED SPECIES LM. REUTERI 100-23

An extensive amount of knowledge is available concerning the proposed type strain of Lm. reuteri subsp. rodentium (100-23; DSM 17509^T; plasmid-free derivative 100-23C) that conforms to the definition of an autochthonous resident of the rodent gut (123, 124). Strain 100-23 cells adhere to the stratified squamous epithelium of the murine forestomach forming a biofilm. The bacteria are shed from the forestomach habitat and can be detected at predictable population levels in the remainder of the digestive tract throughout the life of the murine host. Critical to studies of this strain was the derivation of a mouse colony that lacked lactobacilli as a component of the gut microbiota but had an otherwise conventional gut microbiota (125). The animals could be inoculated with cultures of lactobacilli, including mutant derivatives of 100-23C, and bacterial colonization fitness could be evaluated by comparison with the wild type. In other work, the influences of colonization with the strain on murine host characteristics were revealed, particularly effects on the immune system (126–128). Effective strategies in identifying fitness traits of Lm. reuteri 100-23 were as follows: the use of In Vivo Expression Technology (IVET) to identify genes that were specifically transcribed in the mouse gut, the use of specific insertional mutagenesis, the sequencing of the genome, and in vivo and in vitro transcriptional analysis using microarrays (129–133). Experimental outcomes from work with strain 100-23/100-23C using these technologies are summarized in Table 2. The culmination of this work was the comparison of the strain 100-23 genome with that of other rodent isolates of Lm. reuteri that showed traits that separated them from isolates from other vertebrate hosts, including humans (132). Inactivation of seven of eight rodent-specific genes (associated with adherence, secretion of cell surface proteins, and environmental sensing) in strain 100-23 had impaired colonization fitness in the murine gut. A striking feature of the studies was the importance of acid tolerance mechanisms, including urea hydrolysis, in strain 100-23 (133–135). Acid tolerance is presumably an important fitness attribute because of the potential build-up of lactic acid in the forestomach biofilm, as well as the generally acidic nature of gastric environments.

TABLE 2.

Examples of genetic loci of Lm. reuteri 100-23 associated with ecological performance in the murine gut

Genetic locus	Summary	Reference
xylA (xylose isomerase), msrB (methione sulfoxide reductase), and met (methionine synthase II)	The three genes were shown to be transcribed under conditions prevailing in the mouse gut using In Vivo Expression Technology (IVET). Subsequent insertional inactivation of the msrB gene showed that mutation impaired the ability of the strain to colonize the mouse gut.	129, 130
lsp (a high molecular mass cell surface protein)	Lsp of 100-23 encodes a large surface protein that shows homology to proteins associated with adherence of other bacteria to epithelial cells. Insertional inactivation of lsp resulted in less adherence in vitro and ex vivo to the forestomach epithelium and impaired colonization ability in the mouse gut.	130
dltA (d-alanine-d-alanyl carrier protein ligase)	The dlt operon encodes proteins required for incorporation of d-alanine esters into cell wall-associated teichoic acids. Insertional inactivation of the 100-23C dltA gene resulted in complete depletion of d-alanine substitution of lipoteichoic acids. The mutant did not form a biofilm in the murine forestomach, had smaller populations in the gut compared to wild type, and was almost eliminated from the habitat in competition with the parental strain. The mutant had impaired growth under acidic conditions.	134
luxS (AI2 synthase)	Autoinducer2 (AI2) mediates quorum sensing that affects transcription of some genes of ecological importance in some bacterial species. The production of AI2 is associated with transcription of the luxS gene. Insertional mutagenesis of the luxS gene of strain 100-23C prevented production of AI2. Biofilms formed in vitro and in vivo by the mutant were thicker than those produced by the wild type. However, biofilm morphology was not restored to wild-type values by addition of AI2 to culture medium. LuxS inactivation seemed to have transcriptional and metabolic consequences rather than a quorum sensing role because, in exponentially growing cells, the mutant had 65% ATP content of wild-type cells, and transcription of genes associated with cysteine biosynthesis/oxidative stress response, urease activity, and sortase-dependent proteins was altered in the mutant according to microarray analysis. Metabolomic analysis showed that luxS mutation affected cellular levels of fermentation products, fatty acids, and amino acids.	131, 233
ftf (fructosyl transferase)	Strain 100-23 produces an extracellular polysaccharide (EPS; a levan) from sucrose. Insertional inactivation of the ftf gene led to a loss of EPS production and impaired colonization of the mouse gut. Biofilm formation was unaffected.	128
ureC (urease subunit)	Comparative transcriptome analysis of 100-23C cells harvested from cultures and mouse forestomach showed upregulation of genes associated with acid tolerance, including urease production, in the mouse forestomach. Insertional inactivation of ureC reduced acid tolerance and impaired colonization ability in the mouse gut.	133

FITNESS TRAITS BASED ON HUMAN SECRETIONS: BIFIDOBACTERIUM LONGUM SUBSP. INFANTIS AND AKKERMANSIA MUCINIPHILA

Most bacterial species inhabiting the human colon depend, directly or indirectly, on growth in the gut on the hydrolysis of plant glycans (136). Niche specializations are achieved because of the different structures and complexity of polysaccharides, which are matched by the hydrolytic capacities of different bacterial species (137, 138). However, adaptations to life in the human gut are not limited to the utilization of plant glycans. Oligosaccharides in human milk and mucins in mucus secreted by the gut mucosa provide prime examples of host-derived substances that are utilized by highly metabolically adapted bacteria. Human milk oligosaccharides (HMOs) are synthesized in the mammary glands and form an abundant component of human milk (139). There are about 150 different molecular types of HMOs that can be separated from human milk into acidic and neutral fractions (140). The acidic fraction contains mostly sialylated HMOs, whereas the neutral fraction contains fucosylated and nonfucosylated forms (141). Most HMOs are not digested in the small bowel of infants but pass to the colon where at least some become substrates for bacterial growth (142). Several Bifidobacterium species can utilize some of the HMOs for growth, but the best adapted for this purpose is B. longum subsp. infantis, seemingly because it has a genomic cluster of genes that form an HMO utilization locus (143). This locus is conserved across strains and is associated with efficient use of HMOs and their component parts for growth (144). Thus, B. longum subsp. infantis strains utilize HMOs efficiently because they can sequester these carbohydrates through their complement of ATP-dependent transporters, cation symporters, and phosphotransferase systems as well as other translocation mechanisms that facilitate the progression of oligosaccharides to the bifidobacterial fructose-6-phosphate phosphoketolase central pathway. In general, bifidobacteria are favored in the gut ecosystem of human milk-fed babies. They comprise, on average, about 60% of the total abundance of bacteria in the feces of these infants whereas a more diverse microbiota is present in the feces of formula-fed infants (no HMOs) (145). Bacteroides thetaiotaomicron, Bacteroides vulgatus, and Bacteroides fragilis are also able to degrade HMOs and are more abundant in the feces of infants fed human milk compared to those of formula-fed infants (145, 146). There are similarities between some HMO structures and those of oligosaccharides present in mucus (see below), and they are metabolized by the same unique, inducible pathways in Bacteroides species (146).

Akkermansia muciniphila can utilize HMOs for growth (147), but most research about this species has concerned its ability to live on the components of intestinal mucus, which covers the surface of the mucosa and forms a protective and lubricating blanket. The mucus blanket has two layers: a dense layer close to the epithelium and a more diffuse layer on the luminal side (148). The mucus layer is in a dynamic state, constantly liberated from association with the mucosal surface and replaced by secretion from goblet cells. Glycoproteins known as “mucins” make up 2% to 10% of the mucus, with the remainder being mostly water. Mucins are high molecular weight, linear, heavily glycosylated molecules consisting of a protein backbone containing regions of tandem repeats of threonine/serine, to which oligosaccharides are attached through O-linkages. The overall structure of a mucin is like a bottle brush: a protein backbone with oligosaccharide bristles. Galactose, N-acetylglucosamine, and N-acetyl-galactosamine form branching chains terminated by fucose, sialic acid (N-acetyl-neuraminic acid), or galactose residues at the peripheral (exposed) end. The peripheral groupings are variable and confer blood group specificity (for example, Lewis antigens) (149, 150). They are often terminated by sulfate, especially in regions where bacteria are numerous (oral cavity and colon). These “sulfomucins” rate limit hydrolysis of mucins: chemical desulfation enhances microbial degradation (151). Genomic investigations indicate that about 80% of 397 bacterial genomes represented in the gut microbiota have genes that encode enzymes that could be important in mucin degradation, either through hydrolysis of oligosaccharides using extracellular glycosyl hydrolases or through intracellular catabolic pathways that utilize the monosaccharide constituents of the oligosaccharides (152). In general, it is probable that interspecies cooperation is required for the complete hydrolysis of mucins (153). However, A. muciniphila obtains all its carbon and nitrogen requirements from mucin and is a common and abundant member of human fecal microbiotas (154–158). Intriguingly, a comparison of DNA sequences obtained from palaeofaeces, and feces collected from members of nonindustrialized societies, shows less representation of A. muciniphila and lower abundance of genes associated with mucin degradation compared to modern (industrialized) microbiotas (159). Wibowo and colleagues (159) speculate that these differences are due to higher dietary intake of plant glycans in the ancient and nonindustrial societies with concomitant lower dependence by the microbiota on mucins as alternative growth substrates.

SPECIALIZED MECHANISMS TO UTILIZE PLANT GLYCANS: BACTEROIDES SPP.

In general, Bacteroides species inhabiting the human gut have an expansive capacity to degrade complex plant glycans (160). The bacteria have relatively large genomes that encode numerous glycosyl hydrolases and polysaccharide lyases that confer the molecular ability to recognize and metabolize complex, polymeric carbohydrates (for example, xylans, pectins, starch) derived from plants (161, 162). These substances are major contributors to “dietary fiber.” Utilization of these substrates by Bacteroides spp. is mostly associated with discrete genomic features known as polysaccharide utilization loci (PULs) that encode proteins for the highly specific capture, importation, and degradation of specific glycans in coordinated pathways (163–166). Generally, even highly decorated glycans are minimally degraded at the bacterial cell surface with most of the specialized degradative machinery located within the cell; possibly a mechanism to avoid “leaking” growth substrates to potential competitors. Each kind of PUL targets a different glycan structure (165, 166). The archetypal PUL is the starch utilization system (Sus) that contains eight colocalized and coregulated genes (163, 164). The genome of Bacteroides cellulosyliticus WH2 encodes the most CAZymes associated with glycan digestion of any Bacteroidetes genome (101). The genes encoding these enzymes are distributed among 113 PULs. Specific glycans in the diet influence which of these loci are critical fitness attributes under prevailing conditions. For example, when B. cellulosyliticus was provided with citrus pectin in the diet of gnotobiotic mice, only PUL83-encoded proteins had increased abundance (101).

During in vitro experiments, B. cellulosyliticus, Bacteroides fragilis, Bacteroides ovatus, Bacteroides vulgatus, Bacteroides xylanisolvens, Bacteroides caccae, Bacteroides uniformis, Bacteroides finegoldii, and Bacteroides intestinalis all grew in a hemicellulose-rich (mostly xyloglucan and heteroxylan) medium (138). Xyloglucan PULs have been identified in B. ovatus, B. cellulosilyticus, and B. uniformis, but the genomes of B. ovatus and B. cellulosilyticus additionally contain xylan-degrading PULs (167). Chemical analysis of spent media showed almost complete consumption of the hemicellulose-rich substrate by these latter two species. In contrast, the other species studied that do not contain xyloglucan PULs consumed only heteroxylan, although minor changes in xyloglucan-related linkages indicated that some of these Bacteroides species may be able to cleave specific xyloglucan side chain residues (138). Additionally, when presented with a variety of utilizable glycans, Bacteroides prioritize the degradation of these complex substrates. B. ovatus, for example, preferentially uses unsubstituted pectin (homogalacturonan) before β-glucan and then substituted pectin (rhamnogalacturonan) and arabinoxylan (77). Thus, Bacteroides species provide prime examples of gut symbionts that use a spectrum of specialized molecular processes, including the ability to form biofilms on particulate material in the colonic digesta, to adapt to potential day-to-day changes in the glycan composition of the host diet (137, 168, 169).

METABOLIC SYNTROPHY BETWEEN MEMBERS OF THE MICROBIOTA

The overall prevalence of syntrophic interactions (one species obtaining nutrition from the metabolic products of another species) between members of the gut microbiota is unknown. However, it is likely to range from the provision of simple carbohydrates derived from hydrolysis of plant glycans to peptides, amino acids, and vitamins (170). Cross-feeding of carboxylic acids (fermentation products such as succinate, propionate, and lactate), between members of the gut microbiota has been investigated in detail and the results highlight the important role that mutualism plays in the maintenance of the microbial community (37, 40, 47). Fermentations also produce the gases carbon dioxide, hydrogen, hydrogen sulfide, and methane (104, 171–173). The production of hydrogen is an effective way of disposing of reducing equivalents so that fermentations yield more ATP. The concentration of hydrogen in the gut is, however, very low because it is used in the metabolism of methanogens, sulfate-reducing bacteria, and acetogenic bacteria (104, 174, 175). Methanogens, notably Methanobrevibacter smithii, are detectable in the microbiotas of about 50% of adult humans (174). These archaea produce methane by using carbon dioxide (or formate) and hydrogen produced by fermentative bacteria. In doing so, the concentration of hydrogen is reduced in the gut digesta, minimizing inhibition of the reoxidation of pyridine nucleotides, which permits the fermentative bacteria to oxidize NADH by a more energetically favorable process than would be the case if hydrogen accumulated (174). As a result of this “interspecies hydrogen transfer,” the activity of fermentative bacteria is enhanced through more efficient and complete oxidation of growth substrates.

“Micronutrients” are essential substances required by organisms as cofactors needed to facilitate a range of metabolic functions (176). The B group of vitamins are micronutrients that have been extensively studied in both eukaryotes and prokaryotes (177). Mammals obtain B vitamins (biotin, cobalamin, folate, niacin, panthothenate, pyridoxine, riboflavin, thiamine) from dietary sources but potentially also from members of the gut microbiota since vitamin prototrophy is relatively widespread (about 70% of bacterial strains) in the community (178, 179). From a microbial ecology perspective, the presence of both prototrophic and auxotrophic bacterial species in the microbiota raises the prospect of B vitamin cross-feeding as a factor in the maintenance of stable communities (180). Metagenomic analysis indicates that complete biosynthetic pathways for B vitamins are encoded in some gut inhabitants whereas mechanisms enabling the uptake of vitamins by other species can be recognized (181). These genetic predictions are supported by the results of laboratory experiments using bacterial cultures and biochemically defined media (181). Intriguingly, incomplete but complementary biosynthetic pathways are present in some species, suggesting that multiple species could collaborate in vitamin synthesis (176, 182–184). The results of a study of cocultures of three common members of the human gut microbiota, showed that the genomes of Subdoligranulum variabile and Hungatella hathewayi encoded biosynthetic genes for cobalamin, whereas the genome of Bacteroides ovatus did not. However, the B. ovatus genome encoded a cobalamin uptake mechanism. Transcription of cobalamin genes of S. variabile was less when cocultured with H. hathewayi, suggesting that H. hathewayi was the principal source of cobalamin in the cocultures (185). Overall, current research on this topic points to the utility of the combination of genetic analysis, in silico reconstruction of biochemical pathways, culture-based experimentation, transcriptomics, and metabolic profiles of bacteria as a useful approach in characterizing the interactive structure of the gut microbiota.

THE RESEARCH PATH AHEAD

Since the genome of a gut bacterium is 3 to 6 Mb in size, there are potentially thousands of gene products that impinge on ecological fitness (44, 182, 186). To date, only a relatively few of these have been identified experimentally. It can be proposed that the main hurdle to understanding the role of the gut microbiota in health and disease is that the metaphorical “jigsaw puzzle” representing the taxonomic and functional complexity of the microbial community has not yet been completely assembled. This is due, in part, to the lack of a detailed ecological context on which to base the observational data derived from metagenomic and other studies of the gut microbiota. It would be helpful to construct an interactive schema of the human bowel ecosystem. To produce such a schema, gut microbiota research needs to prioritize questions such as “what are the microbes doing and how are they doing it?” (functional analysis) and “what are the regulatory mechanisms relating to specific functions?” (analysis of metabolic integration) rather than focusing on “who’s who in the zoo?” (taxonomic analysis). Detailed understanding of the operation of the ecosystem is only likely to be generated using data obtained from physiological experiments in the laboratory. It should be possible to use synthetic ecological communities in conducting these experiments (6, 187). Synthetic communities provide a way to limit experimental systems to a manageable number of microbial constituents. This provides control of microbial interactions and makes analysis of the model system easier. Although synthetic communities do not reproduce the taxonomic complexity of the gut microbiota, principles of community function learned from them should be transferable to complex communities. Careful choices of species regarding biochemical capacity need to be made in devising the synthetic community. For example, Liu and colleagues (103) chose bacterial species (Bacteroides ovatus, Bifidobacterium longum subspecies longum, Megasphaera elsdenii, Ruminococcus gnavus, and Veillonella parvula) that could be grown in five-member cocultures that produced SCFAs by the common pathways represented in the gut microbiota. Preferential order of use of plant glycans in a mixture containing β-glucan, pectin, xyloglucan, arabinoxylan, arabinan, and galactan was observed in the coculture, and variation in specific glycan availability on community function was tested. Propionate, as a proportion of total SCFAs, was augmented when glycan mixtures contained galactan, resulting in greater succinate production by B. ovatus and conversion of succinate to propionate by V. parvula. The development of laboratory microcosms like this removes sources of variation, provides data for modeling (mathematical and metabolic reconstructions), and allows exact control of the dynamic nutrient landscape in order to develop concepts of ecological niches in gut communities.

There is a need to consider gut microbiotas as cooperative societies based on “leakage” of nutrients resulting from cell structure and cell physiology. The basic premise is that a community contains “helpers” (that perform public services) and “beneficiaries” (use the services without obvious cost to themselves) (188–192). The arrival of a complex glycan in the colon, for example, may cause “nutrient ripples” due to the hydrolysis of the carbohydrate by some members of the microbiota and the release of oligosaccharides that become growth substrates for other bacteria. In this way, the cooperative society can be predicted to be based on the activities of functional consortia that form and change according to the prevailing nutrient landscape. The composition of consortia may differ between individual humans, but functional redundancy allows conservation of function. In some instances, “keystone” species may be universal members of particular consortia (193). Synthetic communities, therefore, coupled with the use of mutant strains, metatranscriptomics and metabolomics, and integrated systems biology, offer the prospect of defining the rules by which these phenomena occur (194–197). Exploration of bacterial attributes that underpin symbiont life in the human gut will enhance our ability to define these rules and will support efforts to understand how dysbiosis of the microbiota occurs and how it can be rectified.