What is the microbiome?

Man lives in a microbial world and it is increasingly recognised that the human organism is host to a very large microbial community, which colonises every surface of the body: skin; mouth; gut⁽Reference Palmer, Bik and DiGiulio⁵⁾; vagina. Even within each of these surfaces there is diversity; the skin of the axilla is host to a very different microbial community from that of the skin of the forearm or perineum and the stomach has a different microbial flora from that of the mouth or rectum. Man is therefore in a sense the sum of many parts; in a much-quoted estimate an individual's own cells are outnumbered by the microbial residents of that individual by possibly an order of magnitude and their genetic material is dwarfed by that of the microbiota. The microbiome is an analogy with the genome, which is the sum of man's germ line-derived genetic material. Understanding the microbiome is just beginning now that new techniques are available for understanding it without the need for bacterial culture of each species of the microbiota⁽Reference Gill, Pop and de Boy^6–Reference Rawls, Mahowald and Ley⁸⁾. The reliance until very recently on culture to identify bacteria has always been a problem, as many of the bacteria present cannot be cultured. Culture has a further problem; it yields information on the biochemical and growth characteristics of the species present, but that is not necessarily the required information. Certainly, from a nutritional point of view what is of more interest is the sum of the metabolic activities of the flora (the ‘metabolome’) or the composition of the total gene pool, which represents man's genes and that of the microbiota (the ‘metagenome’). Man has recently been released from a dependence on bacterial culture by the development of advanced DNA sequencing and hybridisation methods that are outside the scope of the present article, but which allow quantification of the relative abundance of different bacteria at a genetic level in a way that is much more representative of the total bacterial pool than has ever been possible before.

Bacterial taxonomy is complex and rapidly evolving, but generally it is possible to divide the bacterial kingdom into about seventy major divisions (each of which corresponds to a phylum), twelve of which are represented in the human gut⁽Reference Gill, Pop and de Boy⁶⁾. In the intestine ten of these phyla have been described (Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, Fusobacteria, Verrucomicrobia, Cyanobacteria, TM7, Spirochaetes, and VadinBE97). In the stomach eight phyla are present, although in much lower abundance (Proteobacteria, Firmicutes, Bacteroidetes, Actinobacteria, Fusobacteria, TM7, Deferribacteres, and Deinococcus-Thermus). The flora of the oral cavity is different again. Probably 15 000 species are represented in the human gut, so description of the microbiome at a species level is an impossibly complex task and currently changes in the abundance of major phyla are all that can reasonably be encompassed. In addition to bacteria there are viruses (including phages that parasitise bacteria), yeasts and other fungi, protozoa and helminths. All these microbes, even helminths when in low abundance, can be considered to be commensals. There is also one dominant representative of the primitive prokaryotes, the Archaea, in Methanobrevibacter smithii.

In the gut this vast microbial community is not inactive. It carries out several important functions including fermentation and catalysis of complex polysaccharides (particularly those containing xylan, pectin and arabinose residues), deconjugation of bile salts, salvaging urea and synthesising essential amino acids⁽Reference Jackson, Gibson and Bundy⁹⁾ and vitamin K; there are probably many others. Together, the collection of metabolic activities in a non-host compartment makes the flora of the gut a separate metabolic organ or ‘metabolome’.

One such effect is the trophic effect that the flora has on intestinal tissue in the host. The SCFA liberated by polysaccharide breakdown in the right colon of human subjects have effects on colonic integrity as they represent an important metabolic fuel for the colonic epithelium⁽Reference Le Leu, Brown and Hu¹⁰⁾. There is also increasing evidence from animal models that the microbiota is required for proper development of the gastrointestinal tract and associated immune cells (see later).

The establishment of the microbiota varies greatly depending on circumstances, but once established it appears to be remarkably stable. In utero the gut is sterile but after birth it is rapidly (in <3 d) colonised by microbial populations. Infants born by caesarean section have a different flora to those born vaginally, with delays in colonisation following caesarean section of 10 and 30 d for Lactobacillus-like and Bifidobacterium-like species respectively⁽Reference Gronlund, Lehtonen and Eerola¹¹⁾. The implications for human health are as yet unknown. These findings have been confirmed and extended by an analysis from a Dutch birth cohort, in which mode of delivery, breast- or formula feeding and the use of antibiotics were all found to be major determinants of the composition of the flora⁽Reference Penders, Thijs and Vink¹²⁾. Antibiotics can have long-lasting effects on the microbiota; deep sequencing of the microbiome from three adults before and after ciprofloxacin treatment has shown that most taxa recover by 4 weeks, but some taxa take much longer to recover⁽Reference Dethlefsen, Huse and Sogin¹³⁾. Similar results have been obtained in mice⁽Reference Croswell, Amir and Teggatz¹⁴⁾, in which it has also been shown that an intestinal pathogen (in this case Citrobacter rodentium) perturbs the microbiota but the pretreatment status quo returns within 28 d⁽Reference Lupp, Roberton and Wickham¹⁵⁾. Importantly, the microbiome tends to recover a composition characteristic of that individual even after treatment with a broad spectrum antibiotic, despite the apparent opportunity that this treatment represents for a change in the flora and that is sometimes exploited by Clostridium difficile ⁽Reference De La Cochetiere, Durand and Lalande¹⁶⁾.

Most interesting of all, there are several examples of the ability of the microbiota to transmit a phenotype from one experimental host to another. Mice that are deficient in the transcription factor T-bet develop colitis. T-bet defines T-cells of the T-helper 1 subset, which are pro-inflammatory T-helper cells that are important in the clearance of viral and other intracellular infections and typically secrete the cytokine interferon-γ. By transferring the colonic flora from mice with T-bet deficiency and colitis to wild-type animals the colitis can be transferred too, suggesting that the flora has become ‘colitogenic’⁽Reference Garrett, Lord and Punit¹⁷⁾. In a second example, mice deficient in the leptin gene develop obesity and transferring their flora to wild-type mice also transfers a change in energy balance so that the recipient mice gain increased fat⁽Reference Turnbaugh, Ley and Mahowald¹⁸⁾. This outcome can be referred to as a ‘transplantable metabolome’ and it may have enormous implications for understanding human obesity, as one of the changes that can be seen in the microbiome of the ob/ob (leptin deficient) mouse is an increase in Firmicutes:Bacteroidetes, which has also been described in obese human subjects⁽Reference Ley, Backhed and Turnbaugh¹⁹⁾. It is possible that an obesity-generating diet in some way transforms the metabolic activity of the microbiome to make obesity more inevitable, and it can be speculated that this effect might help explain why cycles of dieting often fail to control obesity and why some individuals do appear to be condemned to sustain higher body mass than others, even when consuming less energy.

The microbiome influences host defence

As mentioned earlier, the intestinal flora appears to be required for development of the gut and for full maturation of the mucosal immune system. Experiments in germ-free animals (which are raised in a sterile environment and have no endogenous flora) are instructive⁽Reference Round and Mazmanian²⁰⁾. Such animals have reduced mucosal thickness and underdeveloped mucosal immune systems with hypocellular lymphoid tissue (Peyer's patches, isolated lymphoid follicles and intraepithelial lymphocyte and lamina propria plasma cell compartments are all less cellular). Mesenteric lymph nodes are hypoplastic and there is a specific reduction in CD4⁺CD25⁺Foxp3⁺ regulatory T-cells. Production of IgA (the dominant secreted Ig class in the gut) is reduced, probably because secretion of IgA seems to require the phagocytosis and carriage of bacteria of the flora from the epithelium to the mesenteric lymph node⁽Reference MacPherson and Uhr²¹⁾. Expression of certain molecules fundamental to innate immunity (Toll-like receptor 9, angiogenin 4 and RegIIIγ) and adaptive immunity (MHC class II, IL-25) is reduced⁽Reference Round and Mazmanian²⁰⁾.

Studies using experimental animals with specific gene deletions confirm the dependence of the development of the gut on the microbial flora. Animals with deletions of the gene that encodes MyD88 (a central signalling molecule of innate immunity), and that cannot therefore recognise the presence of the microbial flora, develop spontaneous colonic inflammation⁽Reference Rakoff-Nahoum, Paglino and Eslami-Varzaneh²²⁾. When this deletion is specific to Paneth cells, which are the cellular source of α-defensins, the endogenous secretion of antimicrobial peptides is impaired, indicating the importance of the gut flora for maintaining constitutive expression of antimicrobial peptides and host defence against pathogens⁽Reference Vaishnava, Behrendt and Ismail²³⁾. One molecule that can confer protection from inflammation has recently been shown to be the polysaccharide A from Bacteroides fragilis ⁽Reference Mazmanian, Round and Kasper²⁴⁾. In another example expression of the antimicrobial peptide RegIIIγ is reduced in germ-free mice but can be restored by conventional flora⁽Reference Cash, Whitham and Behrendt²⁵⁾. Colonisation with the single species Bacteroides thetaiotaomicron does not fully restore expression in wild-type animals, but in IgA-deficient animals this single species fully restores RegIIIγ expression, suggesting that in the absence of IgA the close association between bacteria and epithelium drives constitutive expression of the antimicrobial molecule⁽Reference Cash, Whitham and Behrendt²⁵⁾.

The microbiome influences nutrition

The transplantable metabolome observations referred to earlier⁽Reference Turnbaugh, Ley and Mahowald¹⁸⁾ strongly suggest that the microbiome can influence nutrition. Further evidence from an animal model may have identified one mechanism by which the flora influences fatty acid metabolism. SCFA are ligands for a G protein-coupled receptor, Gpr41, present on enteroendocrine cells in the colon that release the gut hormone peptide YY, which in turn drives SCFA absorption⁽Reference Samuel, Shaito and Motoike²⁶⁾. Gpr41-knock-out mice are leaner than wild-type animals, suggesting that energy salvage from SCFA is less efficient without this receptor, but this difference in adiposity is only observed in gnotobiotic (i.e. known composition, in this case Bacteroides thetaiotaomicron and Methanobrevibacter smithii) animals and not in germ-free animals⁽Reference Samuel, Shaito and Motoike²⁶⁾. This key paper demonstrates that the flora interacts with host physiology to modulate host nutrition, and the mechanism is beginning to be elucidated. Undoubtedly, over the coming years further data will elucidate the details of this fascinating interaction.

Nutrition influences host defence

It has become axiomatic in teaching physicians, nurses and dietitians that malnutrition leads to increased susceptibility to infectious disease. The term ‘malnutrition’ in this context is generally considered to mean reduced body mass. However, the evidence base for this apparently simple statement is slender. In classic experiments on starvation in human volunteers⁽Reference Keys, Brozek and Henschel²⁷⁾ no evidence was found that incidence of infectious disease is higher in the volunteers than in the carers who looked after them, despite severe depletion of body cell mass. Furthermore, there is no evidence that patients with anorexia nervosa have an increased incidence of infections⁽Reference Golla, Larson and Anderson²⁸⁾, although case reports exist of exotic, particularly fungal, infections that appear to complicate this disorder. In one of these studies IL-2 synthesis was reduced by 49%⁽Reference Bessler, Karp and Notti²⁹⁾. In another study T-lymphocyte populations were found to be normal, with lymphocyte proliferation in response to phytohaemagglutinin and concanavalin A showing a marginal increase⁽Reference Golla, Larson and Anderson²⁸⁾. Conversely, another study has observed high circulating levels of IL-1β and TNFα, together with reduced T-cell activation as expressed by CD2 and CD69⁽Reference Allende, Corell and Manzanares³⁰⁾. While this report is one of very few to identify an immunological change, it may not be important if an increase in susceptibility to infection is not observed. These examples are of ‘pure’ or primary undernutrition, a result of simple energy deprivation rather than undernutrition secondary to disease. Thus, altogether there is very little evidence that pure malnutrition influences host defence mechanisms.

In a more complex situation, in a trial in which Kenyan schoolchildren were randomised to several different food supplementation regimens (meat-based, milk-based, vegetable oil-based or none) antibody titres to Helicobacter pylori, rotavirus, tetanus toxoid and malaria merozoite surface proteins were found to show very little change⁽Reference Watson, McMurray and Martin³¹⁾. In severe malnutrition in children there is good evidence of atrophy of the thymus and lymphatic tissues⁽Reference Purtilo and Connor^32–Reference Parent, Chevalier and Zalles³⁴⁾, and an interesting hypothesis has been put forward that thymic impairment programmed in early life carries forward into susceptibility to infection in adult life⁽Reference Moore, Collinson and N'Gom³⁵⁾. More recently, dendritic cell function has been shown to be markedly impaired in severely-malnourished children⁽Reference Hughes, Amadi and Mwiya³⁶⁾ but these studies nicely illustrate the difficulty of interpreting data in severe clinical malnutrition. The most recent study has shown that lipopolysaccharide is correlated most closely with the immune impairment⁽Reference Hughes, Amadi and Mwiya³⁶⁾, suggesting that it is the coexistent infections that do the real immunological damage. In ‘pure’ primary malnutrition, because of pure energy deprivation as in famine and conflict situations, survival is better at extreme low BMI than would be expected in patients with malnutrition complicated, or induced by, infectious or malignant disease⁽Reference Collins³⁷⁾.

However, as there is abundant evidence that malnourished children in developing countries have a high coexistent burden of infectious disease⁽Reference Hughes and Kelly³⁸⁾, how is this situation explained? There appear to be two possible explanations. One, as suggested earlier, is that much of the immunodeficiency in severe malnutrition is a consequence of, as well as a cause of, infection. The second is that the immunodeficiency seen in severely-malnourished patients is related to micronutrient deficiencies. There is much better evidence for the hypothesis that micronutrient deficiencies impact on host defence than for the hypothesis that reduced body cell mass (energy deprivation) impacts on host defence. This second hypothesis would fit with the lack of evidence for substantially increased infection susceptibility in pure malnutrition, as micronutrient depletion in anorexia nervosa, for example, would be in a sense in proportion to the reduced body cell mass until an inflammatory response supervenes, when losses of micronutrients as a result of inflammation⁽Reference Thurnham, McCabe and Northrop-Clewes³⁹⁾ would precipitate immunological consequences. The evidence that micronutrients can alter host defence in the gut will be reviewed, with an emphasis on human studies and a focus on four dominant elements of host defence in the intestine: physico-chemical barriers; innate immunity; humoral immunity; cell-mediated immunity. The micronutrients for which there is the most compelling evidence are vitamin A and Zn, and also briefly discussed are Se, other antioxidants and Fe. In much of what follows selected evidence has been used that does not directly relate to intestinal defence mechanisms because direct evidence for an effect in the gut is scant.

Host defence influences the microbiome

There is very little evidence in human subjects for this contention. While it is certainly an important question, it has only recently become possible to address it; two recent important experiments have demonstrated that host defence can modulate the microbiome.

It has been shown that a transgenic mouse that expresses a human intestinal α-defensin 5 (in addition to its own complement of antimicrobial peptides) is protected against salmonellosis⁽Reference Salzman, Ghosh and Huttner⁶⁸⁾. Furthermore, the composition of the flora is altered⁽Reference Salzman, Hung and Haribhai⁶⁹⁾ and it is tempting to speculate that this outcome may be an important role of intestinal antimicrobial peptide expression, as it has been established that such expression responds to the microbiota⁽Reference Vaishnava, Behrendt and Ismail²³⁾.

Very interesting recent data from an in vitro model of oral plaque biofilm formation indicates that antimicrobial molecules can influence the relative abundance of many species in the buccal flora (D Devine, personal communication).

Can this interaction be manipulated? The case for and against probiotics

It would be useful if the susceptibility of individuals or populations against intestinal infection could be reduced by up regulating host defence or if the metabolome could be manipulated. However, very few instances have been demonstrated.

Probiotics are orally-administered cultures of viable bacteria, which are claimed to re-populate the microbiota. Prebiotics are intended to modify the microbiota through provision of a balance of nutrients that favours the expansion of one bacterial group. Probiotics have been demonstrated to be efficacious in the prevention of childhood diarrhoea, which by implication means successful manipulation of some element of this interaction⁽Reference Allen, Okoko and Martinez⁷⁰⁾. However, of all the trials included in the review only two were conducted in tropical populations who most need the interventions. Nevertheless, this efficacy of probiotics demonstrates the principle that manipulation is possible. A full discussion of the large literature relating to probiotics is outside the scope of the present article but some recent data are worth noting. In a cross-over trial a probiotic preparation has been shown to modify the composition of the flora, but the same effect is obtained with a heat-killed control, suggesting that its true effect might have been prebiotic rather than probiotic in nature⁽Reference Garcia-Albiach, Jose and de Felipe⁷¹⁾. The likely effect of probiotic preparations may actually be prebiotic, rather than the commonly supposed effect, for two reasons. First, prebiotics can undoubtedly manipulate the microbiota to good effect, such as when oral rehydration therapy containing resistant starch is used to shorten childhood diarrhoea⁽Reference Ramakrishna, Subramanian and Mohan⁷²⁾. Second, it is not immediately apparent that a dose of 10¹⁰ colony-forming units of a probiotic (many of which will be killed by gastric acid) could greatly influence a microbiota that comprises ≥10¹⁵ bacteria. More work is needed on that point.

Can this interaction be influenced? The evidence relating to nutritional modification

Nutritional manipulation of intestinal defence clearly works; Zn is without question an important element of treatment of diarrhoeal disease and vitamin A has an important role to play in prevention. However, their mechanism of action is not known; for vitamin A, even the metabolic fate of mega-dose supplements is unknown. There is a strong rationale for giving multiple micronutrients together when attempting to achieve physiological nutritional restitution, as free-living organisms do not consume single nutrients to excess (or very rarely). This situation is different from using single nutrients as pharmacological modulators of specific physiological processes. In relation to Zn supplementation during diarrhoea, it is not clear whether the dose given represents physiological restitution at a time when requirements are increased or whether the dose is supraphysiological and in fact a pharmacological manipulation of, for example, defensin expression. Future studies should try and dissect out which of these processes is in operation in each circumstance.

Of the different elements of intestinal defence that might mediate these nutrient effects, effects on T-cells and antimicrobial peptides seem to be the strongest candidates. A placebo-controlled trial in healthy volunteers in The Netherlands has found that a combined supplement of vitamins A, C and E with Zn increases delayed-type hypersensitivity responses but not lymphocyte subsets, oxidative burst or responses to tetanus toxoid vaccine⁽Reference Wolvers, van Herpen-Broekmans and Logman⁷³⁾. As part of a recent randomised controlled trial of a daily multiple micronutrient supplement in Zambian adults⁽Reference Kelly, Katubulushi and Todd⁷⁴⁾ an analysis was also carried out of antimicrobial peptide expression and an up-regulation of the α-defensin HD5 was observed in malnourished (BMI <18·5 kg/m²) individuals (P Kelly and T Shawa, unpublished results). Effects of micronutrients on T-cell and antibody responses at rest, during intestinal infection or following vaccination remain largely unexplored.

Concluding remarks

Obesity is emerging as a public health crisis that will shape the health challenges of the 21st century, yet one billion individuals are undernourished and often these conditions coexist in the same populations⁽Reference Prentice, Gershwin and Schaible⁷⁵⁾. While this position is driven by inequality in an unjust world, the key to understanding these disorders may lie in the human metabolome⁽Reference Prentice, Gershwin and Schaible⁷⁵⁾, the microscopic and physiological mechanism through which human behaviours drive some of the burden of disease. Understanding the human metabolome may hold the key to colo-rectal cancer, to inflammatory disorders and to effective treatments for diarrhoeal disease (such as oral rehydration therapy improved with resistant starch⁽Reference Ramakrishna, Subramanian and Mohan⁷²⁾). The benefits of incisive research into the microbiome and its effects on physiology could be very important indeed.