Gnotobiotic experimentation helps define symbiogenesis in vertebrate evolution

ABSTRACT

The common presence of microbial communities (microbiotas) that form life-long associations with particular body sites of vertebrates are considered to represent symbiotic partnerships. These multi-organismal associations are different than ‘holobiont’ partnerships defined by Margulis which have restricted diversity and feature high fidelity, intergenerational transmission of symbionts. Comparison of the nutritional and physiological properties of germfree, defined microbiota, and conventional mice help to reveal the extent of the symbiosis between the vertebrate host and microbial communities of high diversity. The mouse-microbiota association is not obligate, but gut microbiota function enables additional energy harvest from the diet. This represents a major evolutionary innovation and therefore could be considered to be an example of symbiogenesis. However, this advantage is balanced by the expenditure of energy required to maintain structures and physiology that prevent invasion of murine tissues by the microbiota. Extrapolation of observations from gnotobiotes to humans helps to develop an appreciation of the human-gut microbiota relationship with emphasis on factors that may have assisted human survival.

KEYWORDS:

• Gut microbiota
• germfree
• gnotobiotic
• holobiont
• symbiogenesis

Gnotobiotic animals

Experimental animals that are ‘pure’ vertebrate (‘germfree’, ‘axenic’ animals) offer valuable tools for the study of the impacts of microbiotas on host properties (Gordon and Pesti 1971). Comparisons can be made of the characteristics of germfree animals (absence of microbial associates) with those of gnotobiotic (defined biota) counterparts. These latter may be ex-germfree animals inoculated with a single kind of microbe (monoassociated), or a defined collection of microbes, or with a preparation of digesta obtained from a conventional animal (the recipient is then said to be conventionalised) (Gordon and Pesti 1971). Any variation in characteristics between germfree animals and the other experimental groups must be due to the absence/presence of the microbial associates, all other factors held constant.

Work at the Lobund Institute, USA, during the 1940s and 1950s resulted in the derivation of colonies of germfree rats and mice by sterile Caesarean delivery of offspring and hand-feeding with sterile milk (Luckey 1963). Germfree breeding colonies of rodents were available by the 1950s and this allowed the use of surrogate, germfree mothers in raising newly derived rodent strains (Wostman 1996; Schoeb and Eaton 2017). Modern derivations may use in vitro fertilisation and then embryo transfer to pseudo-pregnant, germfree females under sterile conditions (Schoeb and Eaton 2017). Germfree mice are nowadays available from commercial and some university sources. Gnotobiotic animal experimentation is conducted in sterile isolators with a filtered air supply and the animals are provided with sterile food, water and bedding. This provides a controlled experimental setting to assess the causal involvement of specific microbes or complex microbiotas across a spectrum of nutritional, physiological (metabolic) and immunological functions.

Microbiotas, holobionts and hologenomes

Microbial communities, often of complex and more or less predictable taxonomic composition (microbiotas), associate with the epithelial surfaces and contents of the interior of hollow organs of vertebrate species to such an extent that they are described as at least ‘indigenous’ (belonging to a place) and often ‘autochthonous’ (found where formed) (Savage 1977). It has been estimated that the healthy human body in adulthood is colonised by at least an equal number of bacterial cells as there are human cells in the body (3.8 × 10¹³ bacterial cells; 3.0 × 10¹³ human cells) (Sender et al. 2016). Intimate associations between healthy vertebrates and their microbiotas are life long and therefore are commonly classed as symbiotic, especially since the host clearly benefits the microbes by supplying them with habitats (Tannock 2022).

Obligate, symbiotic associations between particular insect hosts and bacterial species are well studied and include the pea aphid-Buchnera aphidicola relationship (the bacteria provide amino acids) (Moran et al. 2003; Moran and Degnan 2006), and Wolbachia that cause four distinct reproductive phenotypes in a range of arthropod orders (feminisation, parthenogenesis induction, male killing, cytoplasmic incompatibility) (Werren et al. 2008). In some insect partnerships, the bacterial partner is located within insect cells (bacteriocytes) and transmission from mother to offspring is ensured by vertical transmission (Baumann 2005). Symbioses such as these between insects and microbes support the concept of highly integrated partnerships between disparate species. These composite partnerships have been described as forming a ‘holobiont’ (holos whole, biont living thing) and considered to be a unit of natural selection in evolution. In originating the term, Margulis described a ‘holobiont’ as a ‘symbiont compound of recognizable bionts’ (Magulis 1991; O’Malley 2017). Eukaryotes are the prime examples of holobionts because, in the endosymbiotic theory, their cells contain at least two genomes that result from the permanent integration of species (heterogenomic): host genome and mitochondrial genome in the case of all eukaryotes, and host genome, mitochondrial genome and chloroplast genome in the case of plants (Magulis 1991; Margulis 1993; Collins et al. 2019). Implicit in the Margulis viewpoint is the ‘appearance of new tissues, organs, physiologies, or other new features’ (symbiogenesis) as a result of protracted symbiotic association (Margulis 2004; Guerrero et al. 2013). Also critical to the holobiont concept is evidence that the symbiogenic partnership is stable and transmitted with high fidelity, inter-generationally (Margulis 1993). Thus, long-term symbiosis together with DNA mutations could result in the appearance of new forms of life (Guerrero et al. 2013; O’Malley 2017).

The ‘hologenome’ concept has been devised mostly because of the lack of firm evidence that microbiota species are usually inherited other than randomly (Tannock 2021). Humans, for example, could be considered to have a hologenome encoding 20,000–25,000 human genes plus a bacterial metagenome of about 2 million to 20 million different genes providing most of the combined genetic capacity (Qin et al. 2010; Grice and Segre 2012; Ezkurdia et al. 2014). It has been proposed that these composites of mammalian and microbial genes, rather than the holobiont per se, could be considered a unit of evolutionary selection (Rosenberg and Zilber-Rosenberg 2016; Carrier and Reitzel 2017). However, metabolic outcomes mediated by bacteria can be achieved by different biochemical pathways. These pathways, despite having a common goal, are encoded by different sets of genes. Thus propionate, a major fermentation product in the mammalian gut, may be formed by three different biochemical pathways (succinate pathway, acrylate pathway, propanediol pathway) encoded by the genomes of different bacterial species (Reichardt et al. 2014). This functional diversity is an extension of functional redundancy (more than one species can carry out the same function) (Louis et al. 2004). Therefore, the selection of functions characteristic of eukaryote-prokaryote partnerships, rather than taxa (holobiont) or genes (hologenome) per se, could be a more useful focus of study.

Energy harvest

Eating less by using microbial enzymes

A diversity of vertebrate species has evolved to consume different types of environmental resources as their food, and this is reflected in gut anatomy, from the multi-chambered rumen of cows to the straight tube gut of carnivores (Furness et al. 2019). Mice are monogastric species that are omnivorous but in the wild consume large amounts of seeds containing starch. Lab chow fed to laboratory mice is also rich in starch because it is cereal-based (Wenderlein et al. 2022). Like humans, conventional mice harbour a complex microbiota in the large bowel (caecum and colon). Until recently, accurate comparisons were somewhat difficult to achieve between murine and human microbiotas. Analysis of murine microbiotas has been based on PCR-amplified, variable region sequences of 16S rRNA genes, whereas human studies favour DNA shotgun sequence data (includes 16S rRNA genes and other bacterial genes useful in taxonomy) (Nguyen et al. 2015). Taxonomic variation in microbiotas between mouse colonies housed in different facilities and the use of different chows to feed the animals are among confounding factors (Krych et al. 2013). ‘Cage effect’ (an increased intra-cage similarity within treatment groups) is a recognised cause of irreproducibility of microbiota experimentation using mice (Ericsson and Franklin 2021). Animals within each cage are subject to the effects of close contact and coprophagy. Gut microbiotas of cage mates begin to more closely resemble each other over time than the microbiota of mice in the same treatment group but in other cages. Although the most abundant bacterial phyla in both humans and mice are the Firmicutes and Bacteroidota, the dominant genera are different. Lactobacillus, Alistipes and Turicibacter are more abundant in mice, and Mucispirillum is a common inhabitant of intestinal mucus of mice (Hugenholtz and de Vos 2018). In human microbiotas, Prevotella, Faecalibacterium and Ruminococcus are more abundant than in mice (Hugenholtz and de Vos 2018). There is 62% overlap between human and murine microbiotas at the genus level, but only 2–10% overlap at species level (Beresford-Jones et al. 2022; Kieser et al. 2022). The abundance of lactobacilli in the large bowel microbiota of mice is doubtless due to colonisation of the non-secretory region of the murine stomach by these bacteria and cells shed from this population inoculate the digesta (Tannock 2004). Thus taxonomically, mouse and human microbiotas are distinct.

As proposed in the previous section, determining the biochemical functions associated with the gut microbiota is more important than the bacterial taxonomy because it reveals the mechanisms by which the microbial community is maintained and microbe-host interactions are facilitated. Exposure of germfree mice to material containing microbial communities from humans, zebrafish, termites, soil, or estuarine microbial mats shows that colonisation success (ability to form populations in germfree mouse gut) is associated with ability of bacteria to utilise carbohydrates from food (plant glycans) or host (mucins) sources, and chemical transformation of conjugated bile salts (Seedorf et al. 2014). Using the CAZyme (carbohydrate-active enzymes) – microbiota database, 304 enzyme families were annotated in the mouse catalogue and 301 of these were shared with the human catalogue. Common functional modules included carbohydrate-binding, carbohydrate esterases, cellulase, glycoside hydrolases, glycosyl transferases and polysaccharide lyases (Xiao et al. 2015). There is, therefore, conservation of function involved in carbohydrate utilisation in the murine and human microbiotas, even though bacterial taxonomic compositions are distinct (Xiao et al. 2015). Thus, it can be implied that the mouse gut microbiota carries out the same digestive processes as does the much better studied human gut microbiota. It is clear, moreover, that degradation of complex carbohydrates by the gut microbiota markedly influences digestion of food in mice; germfree mice consume 20-30% more laboratory chow than conventional or conventionalised animals (Gordon 1968; Gordon and Pesti 1971; Bäckhed et al. 2004). So, in the presence of microbial associates, conventional mice consume less food but, even so, harvest sufficient energy from the diet to satisfy their bodily requirements.

Eating less through human ingenuity, natural selection and microbial enzymes

Humans are monogastric and omnivorous but are heavily dependent on cereals for nutrition (rice, wheat, corn) (Copeland et al. 2009). The evolution of the digestive tract has been influenced by the ability to process and cook cereals prior to ingestion (cucinivores). The subsequent lessening in the amount of raw food that had to be ingested to obtain sufficient energy allowed a reduction in gut size which, in turn, aided the attainment of erect posture (Furness et al. 2019). Increased amylase gene number in the human genome that occurred at least 20–30 thousand years ago possibly coincided with access to starches in grains and tubers through grinding and cooking (Perry et al. 2015). Thus, humans rely mainly on mechanical, enzymatic and acid breakdown in the stomach and proximal small bowel after ingestion to harvest energy from pre-processed food. Nevertheless, 80% of dietary fibre (starches and hemicelluloses) ingested by humans is degraded and fermented by the colon microbiota (Cummings and Stephen 1980) and this harvests up to an additional 10% of energy from the diet in the form of bacterial fermentation products. The fermentation products are mainly short chain fatty acids (SCFAs; mostly acetate, propionate, butyrate) as well as gases (hydrogen, carbon dioxide, methane), minor amounts of branched-chain fatty acids (isobutyrate, isovalerate) derived from fermentation of amino acids in the distal colon, a minor amount of the SCFA valerate, and pungent volatiles such as indoles (derived from tryptophan), hydrogen sulphide and thiols (derived from sulphur-containing amino acids such as methionine) that give faeces a distinctive odour (Cummings and Macfarlane 1991; Yadav et al. 2018). The release of ‘phytochemicals' (polyphenols, anthocyanins, phenolics, flavins) through the degradation of plant-derived substances by the microbiota also occurs (O’Keefe 2019). Ninety-five per cent of acetate, propionate and butyrate produced by the microbiota are absorbed from the colon and have caloric value (Høverstad 1986; Cummings and Macfarlane 1991). The metabolites absorbed in the colon are carried by the portal vein to the liver where they are often chemically transformed. Acetate passes to the peripheral circulation and can be detected throughout the body and is an important source of acetyl-CoA that supports fatty acid synthesis (Ballard et al. 1969) and may also activate parasympathetic nerve responses (Perry et al. 2016). Propionate is transformed to glucose in the liver (gluconeogenesis) and is therefore also an energy source (Morrison and Preston 2016). Butyrate seems mostly to be used as energy source by the cells forming the gut mucosa (butyrate provides up to 70% of colonocyte energy source) (Zeng et al. 2014). This harvest of additional calories would not otherwise occur in humans and was likely to have been important for survival in times past when food sources were sometimes scarce, and malnourishment and starvation frequent. However, except in the case of very high dietary fibre intakes characteristic of palaeolithic humans, the contribution of the microbial energy harvest to total caloric intake is relatively small and, under conditions of consuming a constant amount of food, more dietary fibre compared to digestible carbohydrates results in a net decrease in total energy gain; there is less net energy yield per gram from fibre, and fibre affects satiety (Flint 2020).

Physiological effects

The impossible dream: eat more food but stay lean

Germfree C57BL/6J(B6) mice consume more food, have a lower metabolic rate and less body fat than conventional counterparts (Bäckhed et al. 2004). In other words, they eat more food, burn less energy, yet do not store as much fat. The mechanism by which body fat deposition is affected by the microbiota of conventional mice (eat less food, higher metabolic activity, more body fat) is complex, involving increased harvest of carbohydrates from plant glycans by the microbiota, absorption of monosaccharides from the gut lumen, and subsequent induction of hepatic lipogenesis. Additionally, the gut microbiota suppresses the production of fasting-induced adipocyte factor (Fiaf), also known as angiopoietin-like protein 4 (ANGPTL4), in the ileal epithelium. Fiaf inhibits the activity of lipoprotein lipase (LPL) by tissues, so reduction in circulating levels of Fiaf result in increased uptake of fatty acids by adipocytes and triglyceride accumulation (Bäckhed et al. 2004). Enhanced fat reserves as a result of the activities of the gut microbiota could be important in survival if there are fluctuations in food availability, but the same mechanism operating in humans would be considered detrimental in modern times due to the worldwide problem of obesity (World Health Organisation 2021). Interestingly, Fiaf gene transcription increases with age of germfree mice coincident to weaning to a solid, laboratory chow diet (Bäckhed et al. 2004, supplementary data). Chubby pups but lean adults appear to be the preferred ‘murine’ morphotypes in the absence of the microbiota. Unfortunately, the body fat results reported by Backhed et al. (2004) were not reproduced in experiments conducted by other researchers using C3H germfree mice (Fleissner et al. 2010). Reports of even earlier germfree studies are also variable with respect to body fat data, indicating that genetically-programmed, host physiological regulatory actions may negate any impact of the gut microbiota (Wostmann 1981).

Microbial metabolites produced in the gut are potential signalling molecules in mammalian physiology because they are agonists of G protein-coupled receptors (GPCRs). Some of these receptors are located in the intestinal epithelium and several bind small-molecule therapeutic agents (Cohen et al. 2017). They are, therefore, of physiological significance. GPCR clone-libraries can be prepared and used in high throughput screens of bacterial metabolites produced in cultures of bacteria characteristic of the human faecal microbiota (Cohen et al. 2017; Chen et al. 2019). Although the selection of bacterial strains is so far very limited, bacterial metabolites that have the potential to modulate host physiology have been identified (for example, SCFAs, N-acyl amides [fatty acid compound mimicry], histamine, phenylalanine, tyramine, cadaverine, nicotinic acid) (Colosimo et al. 2019). Studies with gnotobiotic mice confirm that in vitro (culture) observations have physiological relevance, at least to the extent that the metabolites are enriched in gnotobiotic gut relative to germfree (Chen et al. 2019; Kumar et al. 2020). Experimental results to date are interesting but not conclusive as to the role of bacterial metabolites in physiological signalling and its role in vertebrate evolution.

A natural vitamin supplement

In general, germfree rodents can be maintained without nutritional problems as long as the vitamin content of food is at a sufficient level following heat sterilisation (Wostmann 1981). Food sterilised by gamma irradiation can also be used. Vitamin-deficient diets result in poor growth rates and signs of deficiency disease in germfree rodents as occurs in conventional animals prevented from coprophagy (the innate behaviour of ingesting faeces directly from the anus) by tail cupping (Ebino 1993). Thus, when a vitamin K-deficient diet (for example) is fed to tail-cupped animals, plasma prothrombin times lengthen and death may occur (Barnes and Fiala 1959). The vitamin K-deficient diet causes disease in germfree animals even when coprophagy is not prevented (Gustafsson 1959). Collectively, these observations indicate that gut bacteria provide supplementary vitamin K (in the form of menaquinones) which is sufficient to prevent deficiency disease (Karl et al. 2017).

Vitamins and the human gut microbiota

Humans obtain B vitamins (biotin, cobalamin, folate, niacin, pantothenate, pyridoxine, riboflavin, thiamine) from dietary sources and the vitamins are absorbed from the small bowel using specific carrier-mediated processes (Said and Mohammed 2006). In principle, B vitamins could also be obtained from members of the gut microbiota inhabiting the colon since B vitamin prototrophy is relatively widespread (about 70% of bacterial strains) and absorptive mechanisms (except for cobalamin) occur in the colonic mucosa (Tramontano et al. 2018). However, humans are coprophobic rather than coprophagic, so bacterial synthesis of vitamins in the colon is probably of minor nutritional consequence except to the members of the microbiota that are unable themselves to synthesise B vitamins (Magnúsdóttir et al. 2015; Das et al. 2019; Sharma et al. 2019; Soto-Martin et al. 2020).

Tissue maintenance

While interest in the impact of the microbiota on the gut environment mostly focuses on the large bowel, Savage and colleagues (Savage 1972) compared enzyme activities associated with epithelial cells (enterocytes) on duodenal villi of germfree and conventionalised mice. The total activity of some enzymes (for example, alkaline phosphatase) tended to be higher in the small intestinal mucosa of germfree animals compared to conventionalised mice (Whitt and Savage 1981). Differential removal of enterocytes from villi showed that the enzyme activities were higher only in enterocytes from the extrusion zone at the tips of villi compared to those nearer the base (Whitt and Savage 1988). Germfree mice had about one and a half times as many enterocytes per villus compared to conventionalised animals and this was associated with longer villi (Savage and Whitt 1982). Enterocytes are formed in the crypts of Lieberkuhn and migrate along the villus until they are extruded at the villous tip. The rate of this turnover of enterocytes was twice as fast in conventionalised mice compared to germfree animals (2 versus 4 days) (Savage et al. 1981). These factors were considered to explain much of the phenomenon relating to enzyme activities; slower enterocyte migration rate in germfree mice resulted in a larger number of mature enterocytes per villus, therefore more enzyme activity. The bacterial mechanisms involved remain elusive because only conventionalisation of the mice affected enzyme levels; individual strains of bacteria had little or no effect (Yolton and Savage 1976; Whitt and Savage 1987).

Comparison of gut wall features between germfree and conventional animals shows a thinner lamina propria in the former due to less reticuloendothelial cells (phagocytic cells such as macrophages) (Gordon and Pesti 1971). Indeed, the conventional gut, relative to germfree, is in a state of mild inflammation (Gordon and Pesti 1971). More energy must be expended to sustain these effects in the conventional animal. Indeed, as mentioned previously, the metabolic rate of germfree mice is lower than that of conventional animals. The ‘growth-promoting’ effect of the administration of subtherapeutic concentrations of antimicrobials in the food of farm animals (poultry, pigs, calves) may, in part, be due to the suppression of bacterial populations that stimulate the inflammatory response or influence cell replacement rates (Visek 1978). More rapid turnover of intestinal epithelium and immune cells could require that the host divert energy from ‘growth’ (muscle gain) to maintenance of gut tissues (Visek 1978). The growth-promoting effect of antibiotics is certainly due to an effect on the microbiota because germfree animals do not show a growth response when fed antibiotics (Coates et al. 1955). Therefore, studies with germfree animals point to a potentially detrimental impact of the gut microbiota on vertebrate energy balance.

Exclusion of pathogens

The presence of the gut microbiota confers resistance (barrier effect) to low-level exposure to enteric pathogens. Changes to the gut environment may deregulate the ecosystem to an extent that normal microbiota functions are affected, allowing a pathogen to establish in the gut, or could encourage the proliferation of some types of bacteria relative to others, producing an ‘overgrowth’ or ‘bloom’ of particular bacteria that can flourish in the altered ecosystem. For example, pre-treatment of conventional mice with antibiotics enables the establishment of Salmonella Typhimurium or Clostridiodes difficile in the gut (Bohnhoff et al. 1954; Lawley and Young 2013). Germfree mice are readily colonised by Salmonella Typhimurium or C. difficile with resulting characteristic pathology (Tannock et al. 1975; Pawlowski et al. 2010). Although the alterations to the gut ecosystem that permit the establishment of these pathogens are not well described, diminished concentrations of short chain fatty acids that normally provide a non-specific, basic level of protection against the establishment of pathogenic loads are likely to be involved (Meynell 1963; Rolfe 1984).

Gut-brain axis

The vagus nerve provides connection between the central nervous system (CNS) and the neck, chest, and abdomen, and contributes to innervation of the viscera. It contains between 80 and 90% of the afferent nerves, mostly conveying sensory information about the state of the organs to the CNS. Nowadays, the gut microbiota is sometimes invoked as an active participant in the communication between the gut and CNS and germfree mouse experimentation has played a vital part in the development of this idea. For example, specific-pathogen-free (SPF), germfree, and mono-associated mice (‘Bifidobacterium infantis’) were subjected to ‘restraint stress’ induced by constraint in narrow tubing. The concentrations of adrenocorticotropic hormone (ACTH) and corticosteroid in blood were measured: germfree mice produced more ACTH and corticosteroid than did SPF mice or ex-germfree mice harbouring bifidobacteria (Sudo et al. 2004). This indicated that the hypothalamic–pituitary–adrenal (HPA) response to stress was greater in the absence of a microbiota. Another study (Bercik et al. 2011) was divided into two parts. In the first part of the study, two strains of SPF mice (BALB/c, NIH Swiss) were administered a mixture of nonabsorbable antimicrobial drugs in their drinking water for 7 days. Behaviour was compared between treated and nontreated animals using step-down tests and light/dark preference tests, which evaluate the effects of anxiogenic and anxiolytic drugs. Antimicrobial drug treatment altered the composition of the microbiota and animals showed increased exploratory behaviour and less apprehensive behaviour than controls, while hippocampal expression of brain-derived neurotropic factor (BDNF) was increased. Specifically, the treated mice stepped down faster from an elevated platform, spent more time in an illuminated compartment, and displayed an increased number of zone entries between dark and light compartments. Intraperitoneal injection of antimicrobials to SPF mice and oral administration to germfree mice did not alter animal behaviour. These results supported the view that gut microbiota affected CNS activities. In the second part of the study, germfree BALB/c mice were colonised with the gut microbiota of NIH Swiss mice, which resulted in increased exploratory behaviour and increased hippocampal levels of BDNF relative to germfree. In contrast, colonisation of germfree NIH Swiss mice with BALB/c microbiota reduced exploratory behaviour. This result was consistent with previous observations that BALB/c mice are generally more timid and anxious compared to other mouse strains, such as NIH Swiss. Therefore, transferring intestinal microbiota from one mouse strain to another could alter the behavioural phenotype of the recipient. The behavioural differences were unrelated to inflammation, gastrointestinal neurotransmitters, or the integrity of the vagus nerve.

The majority of studies using gnotobiotic mice have tested anxiety. Unfortunately, there is poor reproducibility of results between these studies (Luczynski et al. 2016). It is noteworthy that germfree life in an isolator is itself anxiogenic (Crumeyrolle-Arias et al. 2014). Moreover, the animals are removed from isolators for testing and, even if an equilibration period is used, the removal to a different environment that may contain odours of unfamiliar mice could affect murine behaviour. Humans and mice share about 70% of genomic protein-coding gene sequences but the activity of genes involved in stress response is different between mice and humans so it is difficult to transpose knowledge between the two species (Yue et al. 2014). To date, microbiota effects on human anxiety and other conditions that involve neural circuitry are highly speculative and are correlative, so not proven to be causative (Hooks et al. 2019; Schächtle and Rosshart 2021; Cryan and Mazmanian 2022; Liu et al. 2022).

Immunology

The fundamental development of the immune system occurs in utero where tolerance of self-antigens by deletion of autoreactive T cell clones in the thymus (central tolerance) and the suppressive influence of CD4+ CD25+ FoxP3+regulatory T cells (Tregs) in the periphery prevents development of autoimmune diseases. Foetal CD4+ T cells have a strong predisposition to differentiate into tolerogenic Tregs that actively promote self-tolerance, as well as tolerance to non-inherited antigens on chimeric maternal cells that reside in foetal tissues. As birth nears, a transition between the tolerogenic foetal immune system and a more defensive adult-type immune system that is able to combat pathogens occurs (Li et al. 2020a).

About 80% of the immune-system cells of the body are associated with the digestive tract (Vighi et al. 2008). This observation reflects the position of the alimentary canal as a portal for the entry of food-associated antigens – and a potential portal for pathogens – into the body. After birth, mechanisms are required to ‘tolerise' the body to immunostimulatory molecules in the diet, and to prevent the invasion of the mucosa and the systemic spread of pathogens. Additionally, the trillions of bacteria that make up the microbiota provide a plethora of antigens associated with bacterial cells and their extracellular products. Germfree animals maintain a smaller store of immune cells so at least some of the microbiota-derived antigens are detected by the immune system of conventional animals (Gordon and Pesti 1971). The presence of the microbiota drives further development of the human immune system for at least 2 years following birth (Sjögren et al. 2009). This is the period during which the gut microbiota develops in complexity because new and chemically diverse substrates from solid foods become available to bacteria in the gut (Tannock 2021). The immune system seems to be regulated so that the microbiota is tolerated and not eliminated from the bowel by immune mechanisms, yet bacteria not characteristic of the microbiota elicit danger signals and immunological responses. Microbiota and pathogens are recognised by the same pathogen-associated molecular patterns (PAMPs) and pattern-recognition receptors (PRRs) found on epithelial and immune cells (Chu and Mazmanian 2013). It has not been conclusively explained how this differential phenomenon occurs. It could be that pathogens, by binding to and/or invading epithelial cells, present stimuli to PRRs (some of which are intracellular) in sufficient quantity to trigger an immunological response, while gut microbiota, which lack specific virulence factors, do not. Prevented from contact with the epithelial surface by the mucus blanket and by antibacterial molecules such as defensins, much of the antigenic mass of the microbiota might be largely invisible to the sentinels of the immune system (e.g. mucosal dendritic cells), which may require an ‘above-threshold’ stimulus to react. The microbiota is not completely invisible, because as mentioned previously, relative to germfree, conventional animals have a low-level, chronic inflammation of the bowel mucosa, which is considered ‘normal’ (Gordon and Pesti 1971). Part of the ‘invisible microbiota’ phenomenon might be the transport into the gut lumen of low-specificity secretory immunoglobulin A (S-IgA) by cells in the mucosa. About half of the bacterial cells in human faeces are coated with S-IgA. Such cells are unable to enter the mucosa and are therefore restricted to a luminal existence (van der Waaij et al. 1996).

A mucosal (local) immune response to the microbiota, without the need to suppress a systemic immune response (which might be necessary if penetration of bacteria occurred), is assisted by the distinct immune ‘geography’ of the bowel (as compared to the rest of the body) (Macpherson et al. 2005). Critical barriers are present, first in the form of the gut epithelium, and second in the mesenteric lymph nodes. The evidence for this is that when mice are given high doses of microbiota bacteria, small numbers of live bacteria can be detected in dendritic cells of the Peyer's patches and (eventually) in the mesenteric lymph nodes. As long as the mesenteric lymph nodes are intact, dendritic cells loaded with microbiota bacteria do not penetrate any further and do not reach systemic secondary lymphoid structures. If the mesenteric lymph nodes are absent, a single intestinal dose with microbiota bacteria results in live bacteria in the spleen. Repeated challenge causes dramatic enlargement of the spleen, indicating a proliferation of immune cells. This compartmentalisation of the mucosal and systemic immune systems, such that unnecessary immune events are prevented, has an energy cost to the host: synthesis and transport of S-IgA, production of mucus and defensins, and maintenance of low-grade mucosal inflammation divert energy from other biosynthetic processes.

Limiting energy expenditure by minimising the extent of immune-cell activation by gut bacteria would obviously be useful. This might involve the derivation of T-lymphocytes with a regulatory function (Treg) in the immune system. Bacteroides fragilis produces an extracellular polysaccharide: a neutral molecule with both positively and negatively charged regions (zwitterion) on the cell surface, encoded by one of eight genetic loci (Liu et al. 2008). Transcriptional switching enables the bacteria to alter the kind of polysaccharide produced. Polysaccharide A (PSA) production in the bowel of gnotobiotic mice mediates the conversion of CD4+ T cells into Foxp3+ Treg cells, which produce the anti-inflammatory cytokine interleukin 10 (IL-10) (Round and Mazmanian 2010). In a disease model of gut inflammation, PSA administration increased the abundance of Treg cells and lessened inflammation compared to controls. It seems likely that many gut bacteria will have this property, since it may be an adaptation of the cell-surface architecture that helps avoid elimination of the microbiota from the gut by immune mechanisms (Sims et al. 2011).

Conclusions

Since mice can be derived germfree and maintained in this state over generations, the microbiota is clearly not essential for their life. Several other vertebrate species have been derived germfree, including chickens, pigs, dogs, and fish. In the 1970s, human infants with severe combined immunodeficiency disease, highly susceptible to infection by usually innocuous microbes, were maintained in isolators following sterile Caesarean delivery prior to checking immune status and, if necessary, finding compatible bone marrow donors for transplantations (Malinak et al. 1973; Simons et al. 1973; Teller 1973; Bealmear et al. 1985; Kirk 2012). Based on a single report containing bacteriological information from that time, microbial contamination (skin) of an infant was transient, and a gut microbiota was not established (Barnes et al. 1969). A germfree human existence is, therefore, possible even though only tested for a short duration for most of these children.

There is clearly a vast difference between the ‘holobiont’, as envisaged by Margulis, and the much more complex and incompletely defined ‘gut microbiota relationship’. Therefore ‘holobiont’, sometimes referred to in vertebrate research, is a heavily nuanced term that has been used inappropriately in some microbiological contexts (Moran and Sloan 2015; Douglas and Werren 2016; O’Malley 2017; Stencel and Wloch-Salamon 2018; Simon et al. 2019). Clearly, ‘holobiont’ should be utilised cautiously in relation to discussions about the gut microbiota as a driver of vertebrate evolution.

Consideration of the structure of digestive tracts of vertebrates clearly points to adaptations in the gut of animals whose diet contains abundant grasses, cereals, and tubers. The bulky nature of the diet selected the evolution of expanded gut regions where the food could be stored (for example, rumen, caecum, colon). These anatomical features are apparent in the foetus, so adaptations of gut structure are evolutionary features programmed in the germline. Much of the stored ‘food’ cannot be digested by animal genome-encoded processes, so the colonisation of the gut with microbial species that degrade and ferment plant glycans became an associated feature of food retention. SCFAs resulting from the fermentation of plant glycans increased caloric extraction from the diet. This seems to be a pivotal, evolutionary event because, as seen in comparing CAZyme categories of vertebrate (mouse, rumen, human) gut microbiotas, bacterial functional adaptation is similar (Li et al. 2020b; Badhan et al. 2021). These adaptations have resulted in the formation of consortia of bacterial species that power the overall metabolism of the gut microbiota (Tannock 2017).

The specialised consortia may be coincidences, constantly repeated generationally but, alternatively, could have arisen long ago in evolutionary time and continue to be specialised colonisers of the gut of modern vertebrates. Human faecal microbiotas, even of monozygous twins, although showing some similarity, are not identical (Zoetendal et al. 2001; Turnbaugh et al. 2009; Tims et al. 2013; Koo et al. 2019). Therefore, the gut microbiota of the parents is not reproduced identically in their children. It is a recognizably human microbiota, but as far as we know, it is not clonal to the extent that the exact same microbiota is passed from generation to generation. Allogenic factors are likely to be important in shaping the microbiota, most obviously the components of the diet that provide nutritional advantages for specific bacteria.

Vertebrate physiology must also have a role because the taxonomic composition of the gut microbiota would not otherwise be recognisably ‘human’ or ‘mouse’. Gut transit time is likely to be a major influence because bacteria whose doubling time is not compatible with the rate of passage of digesta through the large bowel will be ‘washed-out’ of the system (Tigchelaar et al. 2016; Vandeputte et al. 2016; Müller et al. 2020; Asnicar et al. 2021). Bile salt profiles are also different between vertebrate species and some forms are inhibitory to microbial growth (Floch et al. 1971; Hofmann et al. 2010). Overall, it seems that enrichment of specialised bacterial consortia by allogenic and autogenic factors ensures that a functionally appropriate microbiota, even though adventitiously obtained from several sources, is achieved.

In conclusion, it is fair to say that microbial energy harvest from the host diet has been a major evolutionary feature in the evolution of many vertebrate species. All other features of host-microbiota associations revealed by gnotobiotic studies (such as immune system programming) are likely to be consequences of this evolutionary step.

Acknowledgements

Gerald Tannock is a Professor Emeritus of the University of Otago and is hosted by the Department of Microbiology and Immunology.

Disclosure statement

No potential conflict of interest was reported by the author(s).